本篇博文主要内容为 2026-09-29 从Arxiv.org论文网站获取的最新论文列表，自动更新，按照NLP、CV、ML、AI、IR、MA六个大方向区分。
说明：每日论文数据从Arxiv.org获取，每天早上12:30左右定时自动更新。
提示: 当天未及时更新，有可能是Arxiv当日未有新的论文发布，也有可能是脚本出错。尽可能会在当天修复。
目录
概览 (2026-09-29)
今日共更新2794篇论文,其中:
- 自然语言处理共392篇(Computation and Language (cs.CL))
- 人工智能共973篇(Artificial Intelligence (cs.AI))
- 计算机视觉共565篇(Computer Vision and Pattern Recognition (cs.CV))
- 机器学习共1013篇(Machine Learning (cs.LG))
- 多智能体系统共50篇(Multiagent Systems (cs.MA))
- 信息检索共56篇(Information Retrieval (cs.IR))
- 人机交互共50篇(Human-Computer Interaction (cs.HC))
多智能体系统
[MA-0] Denoising Multi-Robot Trajectories
【速读】：该论文旨在解决多机器人轨迹规划中存在的计算挑战问题，其核心在于应对高维、非凸且具有多模态特性的优化难题。传统数值优化方法在处理此类复杂问题时往往效率低下且难以保证解的质量。本文提出的解决方案关键在于基于D4orm（一种动力学感知的扩散-去噪框架）构建一系列面向不同应用场景的规划架构。D4orm通过基于采样的优化方式，利用现代计算架构（如GPU）实现大规模并行采样，以迭代优化候选控制轨迹的形变，从而高效生成满足运动学与动力学约束且无冲突的轨迹。该方法突破了传统优化范式的局限，具备良好的通用性与可扩展性。研究进一步提出了解耦式规划器以提升可扩展性、在线滚动时域规划器结合反馈控制以增强实时响应能力，以及适用于资源受限环境的分布式规划器。实验结果表明，基于D4orm的方法在二维与三维环境中对差速驱动和全向移动机器人均能更快、更可靠地找到高质量解，显著优于MPPI等经典采样优化方法及基于学习的扩散模型方法。此外，该方法成功实现了十台真实四旋翼无人机的零样本部署、百台仿真机器人规模下的大规模避障，以及六台地面机器人全自主、持续运行的分布式协同任务，充分验证了扩散去噪框架在多机器人协同中的可扩展性与可靠性。
链接: https://arxiv.org/abs/2609.35651
作者: Yuhao Zhang,Keisuke Okumura,Ajay Shankar,Amanda Prorok
机构: 未知
类目: Robotics (cs.RO); Multiagent Systems (cs.MA)
备注: Accepted to IEEE Transactions on Robotics (T-RO)
Abstract:Multi-robot trajectory planning is a fundamental problem in multi-robot coordination but remains computationally challenging due to its nonconvex, multimodal, and high-dimensional nature. This work builds upon D4orm, a dynamics-aware diffusion-denoising framework, and develops a family of planning architectures for diverse operational requirements. Unlike conventional numerical optimization methods, D4orm employs sampling-based optimization to generate solution trajectories through massively parallel sampling, leveraging modern computing architectures such as GPUs. Its diffusion-denoising structure iteratively optimizes \textitdeformations to candidate control trajectories, providing an efficient and versatile paradigm for generating kinodynamically feasible and conflict-free trajectories. Using D4orm as the building block for advanced planners, we present a decoupled planner for improved scalability, an online receding-horizon planner with feedback control, and a distributed planner for resource-constrained settings. Evaluations with differential-drive and holonomic robots in 2D and 3D environments demonstrate that D4orm-based approaches find high-quality solutions faster and more reliably than other sampling-based optimization methods, such as MPPI, as well as a learned diffusion-model-based method. We further demonstrate zero-shot deployment on ten real quadrotors with obstacles, large-scale deconfliction with 100 simulated robots, and fully onboard distributed `lifelong’ operation with six ground robots. Overall, these results establish diffusion denoising as a scalable and reliable framework for multi-robot coordination. Code and video: this https URL
[MA-1] Positional choice and robust collective behavior in fish schools: biohybrid experiments and modeling
【速读】：该论文旨在解决现有鱼类群体行为模型中个体层面运动规则的真实性问题。尽管经典模型基于个体相对于邻近个体的位置与速度遵循特定规则，能够复现自然界中观察到的群游模式，但其假设的个体行为机制是否真实仍存疑。为验证这一点，研究者通过实验观察活鱼与四台移动机器人鱼在水箱中的互动，并测量鱼在不同相对位置的停留时间；随后在仿真中以智能体替代真实鱼，评估经典模型对实验数据的拟合程度。结果表明，仅当参数取值极为特定时，经典模型才能精确匹配实验结果，但对参数微小扰动极度敏感，极易失效。为此，论文提出一种新模型，引入一个表征智能体瞬时位置偏好的探索性项，使模型在参数存在微小误差时仍保持鲁棒性。该模型不仅成功复现了单个个体的实验行为，且在扩展至多鱼系统时，可自然生成典型的群游结构，同时在个体与群体层面均表现出探索性行为。这表明社会凝聚力与个体探索行为可共存，未来真实的集体行为模型必须同时考虑社会交互与探索性成分。
链接: https://arxiv.org/abs/2609.35554
作者: Vahagn Grigoryan,Donato Romano,Cesare Stefanini,Giulia De Masi
机构: Sorbonne University Abu Dhabi(索邦大学阿布扎比分校); BioRobotics Institute, Sant’Anna School of Advanced Studies(圣安娜高等研究院生物机器人研究所); MBZUAI(穆巴达拉人工智能研究所)
类目: Multiagent Systems (cs.MA)
备注: 38 pages, 8 figures, 2 tables; supplementary material with 2 figures and 3 tables
Abstract:Collective behavior of fish schools is usually modeled on the assumption that each individual follows specific rules of motion that depend on its position and velocity relative to its neighbors. Although these models reproduce many schooling patterns observed in nature, it remains unclear whether the assumed rules are realistic at the individual level. To address this question, we first analyzed a set of experiments in which a live fish interacted with four moving robotic fish in a tank, and measured the time it spent in each position relative to the robots. We then simulated this experiment, replacing the fish with an agent, and assessed the extent to which classical models agree with the experimental observations. An extensive exploration of the parameter space showed that these models closely matched the experimental observations, with a very specific choice of parameters. However, they were highly sensitive to the parameter values: a minor perturbation caused them to fail completely. To resolve this, we propose a new model that incorporates an additional exploratory term characterizing the agent’s positional preference at every moment. This model proved robust to small errors in the parameters while successfully replicating the experiments. Furthermore, when generalised to multiple fish, the model reproduced schooling behaviour and common schooling patterns, while also exhibiting exploratory behaviour at both the individual and the school level. This shows that social cohesion coexists with individual exploratory behaviour, and that realistic models of collective behaviour should account for both social interactions and this exploratory component.
[MA-2] Self-Adapting Group of Experts for Multi-Agent Reasoning
【速读】：该论文旨在解决多智能体系统（Multi-agent Systems）在面对复杂任务时，因各智能体固定的角色与推理策略无法动态适应问题需求而导致的性能瓶颈问题。现有框架通常仅通过调整输入上下文来实现通信适配，而保持各智能体的系统提示（system prompt）不变，难以根据具体问题灵活调用不同技能。其核心挑战在于如何识别并迁移更优的推理策略以提升整体协作效率。解决方案的关键在于提出一种无需训练的自适应框架SAGE（Self-Adapting Group of Experts），其通过答案一致性（answer agreement）、前缀一致性（prefix consistency）以及互评反馈（reciprocal peer review）机制，自动筛选出表现更优的“策略提供者”（strategy donor），并将该提供者的推理策略无损地传递给其他智能体，同时保留其原始角色设定。该策略迁移过程仅依赖于各智能体原有的系统提示，不访问问题本身或生成的中间解，确保了隐私与可扩展性。随后，系统采用动态稀疏有向无环图（dynamic, sparse directed acyclic graph）进行响应传播，优先将高评分智能体的信息路由至低评分者，从而实现高效的知识聚合。实验结果表明，SAGE在多种模型架构和推理基准上均显著优于基线方法，验证了其在无需额外训练条件下实现跨智能体策略自适应的有效性。
链接: https://arxiv.org/abs/2609.35412
作者: Mohammad Atif Quamar,Nurbek Tastan,Karthik Nandakumar,Junpei Komiyama
机构: Mohamed bin Zayed University of Artificial Intelligence (穆罕默德·本·扎耶德人工智能大学); Michigan State University (密歇根州立大学); RIKEN AIP (理化学研究所人工智能项目)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multiagent Systems (cs.MA)
备注:
Abstract:Multi-agent systems bring together language model agents with different roles to propose, review, and refine solutions. Each agent’s response depends on its model’s capabilities, the reasoning strategy defined by its system prompt, and the information in its input context. Existing frameworks often adapt communication by changing this context while leaving individual prompts fixed, even when a problem calls for different skills. We study whether agents’ initial responses can identify a strategy better suited to the current problem and guide its transfer to other agents. To address this, we introduce SAGE (Self-Adapting Group of Experts), a training-free framework that uses answer agreement, prefix consistency, and reciprocal peer review to select a strategy donor. SAGE transfers the selected donor’s reasoning strategy to the other agents while preserving their original roles. This transfer uses only the agents’ original system prompts, without access to the problem or generated solutions. After strategy adaptation, agents exchange responses through a dynamic, sparse directed acyclic graph that routes information from higher-scoring agents to lower-scoring agents. Experiments across multiple agent backbones and reasoning benchmarks show that SAGE achieves higher average accuracy than the evaluated baselines. Our code is available at this https URL.
[MA-3] Persistent Partners Raise Prices Among Learning Agents
【速读】：该论文旨在探究在重复交互的平台环境中，平台对匹配机制的选择是否会影响代理（agent）学习到的价格水平，以及是否存在由学习产生的惩罚机制。其核心问题是：当代理在伯特兰双寡头竞争模型中反复博弈时，平台决定配对关系，这种配对策略是否会改变代理通过强化学习所收敛的价格水平？若存在价格上升，是否伴随有可识别的学习型惩罚行为？ 解决方案的关键在于设计一个预注册的随机化实验，采用基于表格的Q-learning模块作为价格设定机制，而非依赖语言模型进行决策，并系统性地操纵三个变量：是否维持固定搭档、能否观察对手价格、是否允许发送消息。结果显示，保持固定搭档显著提升了平均利润水平（较竞争与垄断利润差距的0.27倍，95%置信区间为0.20至0.35），且该效应在所有20次配对运行中均呈正向；即使在对手价格不可见的情况下，价格仍能持续上升并维持至训练结束，表明价格提升并非源于可见对手的即时惩罚反应。进一步分析发现，在对手可见条件下，静态最优响应已解释了约三分之一至一半的“惩罚”信号，而经注册的正式测试无法确认学习型惩罚的存在。因此，仅以惩罚检测为指标会遗漏隐藏对手情境下的价格上升现象，而通过检查有利偏离的可行性则能识别出多数高利润价格点。此外，探索性扩展显示，未训练的Qwen2.5 7B和14B模型在提示中省略对手价格时表现出类似效应，而包含对手价格时效果不一致，7B模型结果可在新块上复现，其他模型家族则未呈现此效应。综上，该研究揭示了稳定配对关系本身即可驱动价格协同上升，其机制可能源于合作倾向或长期激励，而非显性的惩罚策略，从而挑战了传统关于价格共谋必须依赖可观察惩罚的假设。
链接: https://arxiv.org/abs/2609.35402
作者: Paul-Peter Arslan,Yubin Kim,Xiao Xiao
机构: Institute For Future Technologies(未来技术研究所); Devinci Higher Education(德文奇高等教育学院); Massachusetts Institute of Technology(麻省理工学院)
类目: Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: 29 pages, 5 figures. Pre-registered on OSF ( this https URL , under embargo). Under review
Abstract:When pricing agents meet repeatedly on a platform, the platform decides who faces whom. We ask whether that choice moves the prices the agents learn, and whether a rise comes with learned punishment. In a pre-registered randomised experiment in the Bertrand duopoly of Calvano et al., each agent’s price is set by a tabular Q-learning module, not by the small language model attached to it, and we randomise whether each agent keeps its partner, sees its rival’s prices and can send messages. Keeping the same partner raises the level of profits, averaged over training, by 0.27 of the gap between competitive and monopoly profit (95% CI 0.20 to 0.35, all twenty paired runs positive), our registered primary result, and the resting price by 0.17 of the Nash-to-monopoly range (post hoc). A plain tabular learner reproduces the effect in all 25 further blocks, and there one permanent partner raises the level more than about three do (+0.23 against +0.05, exploratory). Where rival prices are hidden, the price-setting module cannot see a cut, so cannot punish it, yet the resting price rises as much and the rise lasts to the end of training, while with visible rivals it shrinks with longer training (post hoc). Where the rival is visible, a static best responder accounts for a third to a half of what a forced-deviation probe reads as punishment, on the starts where the rival can see the cut, and net of it the registered test of learned punishment is inconclusive. A test that looks only for punishment would thus miss the rise where the rival is hidden, while a check for profitable deviations flags most of those prices (post hoc). In an exploratory extension, untrained Qwen2.5 7B and 14B models under one prompt show the effect when the rival’s price is left out of the prompt and inconsistently when it is shown, the 7B result replicating on fresh blocks, while two other model families show none.
[MA-4] From Migration to Calibration: Preserving Agent Capabilities across Models Jurisdictions and Scale
【速读】：该论文旨在解决智能体（Agent）在部署环境变化时的校准问题，尤其针对驱动模型更换、基础模型到微调模型的迁移、跨司法管辖区部署以及在异构市场与数据源间的规模化扩展等场景。核心挑战在于，仅保证接口兼容性无法确保能力保留或目标契约的满足。其解决方案的关键在于将智能体校准建模为三个相互作用层的受约束行为适应过程：信息保全（information preservation）、Harness校准（harness adaptation）和用户接受度校准（user calibration）。其中，信息校准确保源端经独立验证的内容仍适用于目标任务；Harness校准通过语义检查点对可观测产物进行对齐，并在显式预算内通过迭代、工具替换或局部重规划进行修复；用户校准则强制执行接收方特定的输出契约，包括模板、模式及章节级偏好。该框架强调非降级的预设能力指标表现，并追求整体性能提升，同时提出保留评估、跨边界适应性测试与规模扩展的因子实验设计，结合组级别报告以防止局部失败被整体增益掩盖。该研究区分了可训练策略与冻结主干配置或控制器优化，以及证据验证与相对判断、基于人类反馈的强化学习（DPO/GRPO）优化之间的差异，主张采用目标原生执行记录、任务有效性与近似排名的独立审计，以及匹配的目标原生优化控制，具有方法论意义，但具体实现与实证验证尚待后续工作。
链接: https://arxiv.org/abs/2609.35149
作者: Yaxiao Liu(PwC China AI Center),Pengbo Liu(PwC China AI Center),Yiwen Liu(PwC China AI Center),Yihua Guan(PwC China AI Center),Jiaxing Song(Tsinghua University)
机构: PwC China AI Center(普华永道中国人工智能中心); Tsinghua University(清华大学)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA); Software Engineering (cs.SE)
备注: 24 pages, 7 figures. Methodological proposal: three-layer agent calibration framework; no empirical results reported
Abstract:Agents need calibration when deployment conditions change: replacing a driving model, including a foundation-to-post-trained transition; crossing jurisdictions; or scaling across heterogeneous markets and sources. Interface compatibility alone does not establish capability retention or target-contract satisfaction. We formulate agent calibration as constrained behavioral adaptation across three interacting layers: information preservation, harness adaptation, and user acceptance; the layers apply to every scenario, not one-to-one to the three. The basic objective is non-degradation on prespecified capability measures while satisfying target requirements; aggregate improvement is stronger. Information calibration preserves independently validated source content still applicable to the target task. Harness calibration aligns observable artifacts at semantic checkpoints and repairs them through iteration, tool substitution, or local replanning within explicit budgets. User calibration enforces recipient-specific output contracts: templates, schemas, and section-level preferences. A global e-commerce example shows how shared standards coexist with site- and market-specific adapters and validation. We distinguish trainable policies from frozen-backbone configuration or controller optimization, and evidence verification from relative judgment and DPO/GRPO optimization. Recent harness-transfer and judge-validity studies motivate target-native execution records, separate audits of task validity and near-tie ranking, and matched target-native optimization controls. We propose held-out evaluations for model changes, cross-border adaptation, and scale, including a factorial test of source evidence and checkpoint repair and group-level reporting to prevent aggregate gains from masking local failures. This is a methodological proposal; implementation and empirical validation remain future work.
[MA-5] Strategically Robust Game-Theoretic Multi-Agent Trajectory Optimization
【速读】：该论文旨在解决先进空中交通管理（Advanced Air Mobility, AAM）中多智能体自主避撞的协调问题，尤其针对去中心化服务架构下，飞行器需在缺乏中央协调的情况下，通过预测其他飞行器的控制输入来自主规划轨迹的挑战。传统基于博弈论的方法虽能高效求解开环均衡，但其假设各智能体严格遵循均衡轨迹，这在实际中因执行、感知与计算不确定性而难以成立。为此，论文提出一种战略鲁棒性（strategically robust）的建模方法：每个智能体均考虑一个虚构对手，在每一时间步内对其他智能体的控制输入施加有界扰动，以最小化当前时刻的距离，从而增强对不确定性的防御能力。研究证明，在合理假设下，该鲁棒博弈仍保持为精确动态势博弈（exact dynamic potential game），且对于线性动力学系统，其内部对抗问题可获得准闭式解，显著降低计算开销。实验结果表明，采用对数距离代价函数的八智能体仿真中，战略鲁棒性在高碰撞风险场景下能选择更稳健的飞行轨迹，同时对低风险场景影响极小，仅带来适度的运行时增加。
链接: https://arxiv.org/abs/2609.35142
作者: Victor L. Qin,Nicolas Lanzetti,Saverio Bolognani,Hamsa Balakrishnan
机构: 未知
类目: ystems and Control (eess.SY); Computer Science and Game Theory (cs.GT); Multiagent Systems (cs.MA)
备注: 7 pages, 5 figures, accepted to CDC 2026
Abstract:Aviation authorities worldwide expect Advanced Air Mobility (AAM) traffic management to be decentralized among service providers, requiring AAM flights to autonomously plan trajectories by predicting other flights’ control inputs rather than relying on centralized coordination. Game-theoretic approaches that formulate multi-agent collision avoidance as an exact dynamic potential game can efficiently find open-loop equilibria, but they assume that agents exactly follow their equilibrium trajectories—an unrealistic assumption given uncertainties in actuation, perception, and computation. We propose a strategically robust formulation where each agent protects against a fictitious adversary that, for each timestep, perturbs other agents’ control inputs within a bounded budget to minimize distance at that timestep. We show that, under reasonable assumptions on agents’ distance cost and robustness levels, the strategically robust game remains an exact dynamic potential game and admits a quasi-closed-form solution to the inner adversarial problem for linear dynamics, which limits computational overhead. Experiments with up to eight agents using logarithmic distance costs show that strategic robustness selects more robust trajectories in high-collision-risk configurations while leaving low-risk trajectories nearly unchanged, with only a modest increase in runtime.
[MA-6] CEO Arena: Evaluating Long-Horizon Multi-Agent Decision-Making in Competitive Markets ICLR2027
【速读】：该论文旨在解决多智能体在长期竞争环境下，如何在不确定性中协调商业决策并适应对手策略动态变化的问题。其核心挑战在于评估智能体在复杂市场环境中不仅追求自身收益，还需考虑对竞争对手及整体市场影响的长期战略行为。解决方案的关键在于提出CEO Arena这一基准测试平台，采用匹配替换评估（matched replacement evaluation）方法，在相同经济种子下对比目标智能体与参照策略在同一公司中的表现，同时固定其他智能体的身份与分工，仅允许所有智能体在动态环境中自适应调整。该框架通过模拟一个包含八家公司的共享市场环境，持续500个仿真日，让每位CEO智能体基于私有企业信息和噪声市场信号，做出定价、采购、营销、研发和服务等序列化决策，并在资源约束和延迟反馈条件下进行优化。实验评估了八个基于大语言模型（LLM-based）的CEO智能体，结果显示多数智能体平均回报为负，且私有收益可能伴随市场整体损失。鲁棒性分析表明，部分智能体间的策略影响具有相对稳定性，结合记忆、动作与会计轨迹的分析，揭示需求攫取及对手定价与支出响应可能是关键机制。因此，CEO Arena为研究长期视角下的智能体竞争、策略互动与市场外部性提供了可控且可扩展的实验平台。
链接: https://arxiv.org/abs/2609.34821
作者: An Yan,Yu Huo,Zhiwei Shang,Yiran Peng,Chenglin Wu
机构: DeepWisdom; Fudan University (复旦大学); The Chinese University of Hong Kong (香港中文大学); The Chinese University of Hong Kong, Shenzhen (香港中文大学深圳分校); The Hong Kong University of Science and Technology (Guangzhou) (香港科技大学(广州))
类目: Multiagent Systems (cs.MA)
备注: 46 pages, 16 figures. Submitted to ICLR 2027
Abstract:Long-horizon competition tests agents’ ability to coordinate business decisions under uncertainty and adapt to changing rival strategies. We introduce CEO Arena, a benchmark that uses matched replacement evaluation to assess operating returns alongside an agent’s effects on rivals and the market. Each CEO agent is compared with a reference policy in the same company under the same economic seed, holding other agents’ identities and assignments fixed while all agents adapt. In a shared eight-company market spanning 500 simulated days, CEOs make sequential decisions on pricing, procurement, marketing, research and development, and service using private company information and noisy market signals, under resource constraints and delayed feedback. We evaluate eight LLM-based CEO agents in 27 main runs and 26 robustness runs. In the main evaluation, most agents have negative mean returns, and private gains can accompany market losses. Robustness analyses suggest that aggregate patterns extend beyond the original rule-based baseline; four of the 56 directed pairs show relatively stable effects. Memory, action, and accounting traces suggest demand capture and rivals’ pricing and spending responses as possible explanations. CEO Arena provides a controlled testbed for studying long-horizon agent competition, strategic interaction, and market externalities.
[MA-7] MASTraceBench: Diagnosing Collaboration Gains through Proposal Trajectories in LLM -Based Multi-Agent Systems
【速读】：该论文旨在解决基于大语言模型（LLM）的多智能体系统（Multi-Agent Systems, MAS）在复杂问题求解中缺乏系统性评估方法的问题，尤其关注协作增益（Collaboration Gain）的产生、维持与丢失机制不清晰的缺陷。现有基准主要依赖最终结果评价，难以揭示智能体间交互过程中的动态演化规律。为此，论文提出MASTraceBench，一个通过追踪和评分智能体提案轨迹（proposal trajectories）来诊断协作增益的基准工具。该基准涵盖任务得分（Task Score）、协作增益、提案轨迹指标及令牌成本（Token Cost）等多层次评估维度，在六项合作与竞争性任务上对主流MAS方法进行系统比较。分析发现，多数情况下最终答案并未超越最强初始提案，且交互过程更倾向于提升弱初始提案，而强提案往往无法进一步优化甚至出现退化。为缓解此风险，论文提出CLEARS方法，通过在智能体间采用主张级（claim-level）评估替代全提案交换，实现更可靠的合成。实验表明，CLEARS能更有效地保留或改进最强初始提案，并在六项任务中的五项上取得最高协作增益。
链接: https://arxiv.org/abs/2609.34496
作者: Yapeng Li,Songze Li,Shuang Yu,Jing Yu,Zhixin Liu,Liqiang Wen,Tonghua Su
机构: Harbin Institute of Technology (哈尔滨工业大学); Peking University (北京大学)
类目: Multiagent Systems (cs.MA)
备注:
Abstract:LLM-based multi-agent systems (MAS) have shown promise in complex problem solving. As MAS methods diversify, systematic evaluation becomes increasingly challenging. However, existing benchmarks largely focus on final outcomes, leaving unclear how collaboration gains arise, are preserved, or are lost. To address this limitation, we introduce MASTraceBench, a benchmark for diagnosing collaboration gains through proposal trajectories in MAS. Across six cooperative and competitive tasks, MASTraceBench tracks and grades proposal trajectories and provides a multi-layer metric suite covering Task Score, Collaboration Gain, proposal-trajectory indicators, and Token Cost. Using MASTraceBench, we systematically compare representative MAS methods not only by final performance, but also by how agent proposals evolve and are aggregated into the final answer. This analysis reveals a recurring pattern: final MAS answers rarely surpass the strongest initial proposal; interaction often lifts initially weaker proposals toward it, while strong initial proposals are seldom further improved and may regress. To reduce this risk, we propose CLEARS, which replaces whole-proposal exchange with claim-level evaluation across agents to guide reliable synthesis. CLEARS more often preserves or improves upon the strongest initial proposal and achieves the highest Collaboration Gain on five of the six tasks.
[MA-8] Emergence Not Bandwidth: Physical Coupling and the Limits of Learned Multi-Agent Communication
【速读】：该论文旨在解决速率受限的多智能体团队在协作中面临的核心问题：最优信息编码应包含何种内容、跨决策周期的压缩成本如何量化，以及学习到的通信协议何时具备足够的唯一性以被队友正确解读。针对速率受限的分散部分可观测马尔可夫决策过程（Dec-POMDPs），研究通过理论定理明确了在不依赖具体学习算法的前提下可实现的性能上限，从而将学习方法与最优解之间的差距归因于优化能力不足，而非信息论上的不可逾越限制。关键解决方案在于构建一个受控的“判别性”实验环境，在MuJoCo平台上的三个场景（零耦合、部分耦合和刚性耦合）中均严格限定每步通信为2比特，并通过关闭物理侧通道的方式消除非通信信号的影响，使通信价值完全由系统耦合特性决定。结果表明，在刚性耦合下，即使无通信也已可通过本体感知获取足够信息，通信无增益；而在部分耦合下，人工设计的2比特发送方达到中位数1.000的性能，而强化学习训练得到的发送方仅达0.482，与静默状态无显著差异（p=0.400），且该差距无法由带宽或共享字母表解释。进一步分析显示，使用人工设计接收器进行热启动可显著提升学习性能（0.857 vs 0.562，p<0.001），说明问题并非表示能力或维护困难，而是强化学习未能发现有效通信协议。跨播验证揭示学习协议虽个体有意义，但彼此间互不兼容——自对弈表现从0.980骤降至0.144（跨种子），且即使采用最优构造对齐仍存在至少77%的性能缺口。所有结论基于每场景25个种子及7个已有基准模型在相同速率约束下的对比实验，充分凸显了当前强化学习在通信协议发现方面的根本局限。
链接: https://arxiv.org/abs/2609.34373
作者: Mihir Chauhan,Aniket Bera
机构: IDEAS Lab, Department of Computer Science (IDEAS 实验室，计算机科学系); Purdue University (普渡大学)
类目: Machine Learning (cs.LG); Multiagent Systems (cs.MA); Robotics (cs.RO)
备注:
Abstract:Rate-limited multi-agent teams raise three questions the emergent-communication literature has answered only empirically: what an optimal message should encode, what compression costs over a horizon, and when a learned protocol is unique enough for a teammate to read. We answer them for rate-limited Dec-POMDPs, then measure how far reinforcement learning falls short of the optimum. Our theorems fix what is achievable independently of any learner, so a gap between an engineered and a learned sender at the same bit budget is an optimization fact, not an information-theoretic one. We instantiate this on three MuJoCo arenas spanning zero, partial and rigid physical coupling, charging every condition exactly 2 bits per decision, and create the discriminating regime by closing a physical side channel within one arena, holding bodies, task and reward fixed. Communication value is governed by coupling: under rigid coupling through a shared object, no channel beats silence (+0.001 +/- 0.001, p = 0.982, n = 25), since proprioception already carries that information; without coupling, every condition solves the task; under partial coupling, the engineered 2-bit sender reaches an interquartile mean of 1.000 but the learned one reaches 0.482, indistinguishable from silence (p = 0.400, n = 25). With a shared alphabet, bandwidth cannot explain the gap. Warm-starting from an engineered receiver localizes the failure: the same channel reaches 0.857 versus 0.562 cold-started (p 0.001), so it is neither representational nor one of maintenance; reinforcement learning fails to discover the protocol. Cross-play shows learned protocols are individually meaningful but mutually unintelligible: self-play 0.980 collapses to 0.144 across seeds, and our best constructed alignment leaves at least 77% of that gap. All headline results use 25 seeds per arena and seven published baselines at matched rate.
[MA-9] Same Winners Different Success Rates: Evaluating How LLM Agents Recover from Failures
【速读】：该论文旨在解决大语言模型（Large Language Model, LLM）智能体在任务执行过程中遭遇中途失败后，如何准确评估其恢复能力的问题。现有基于检查点（checkpoint）的基准评测方法通过比较独立运行中表现最优的动作集合（即“集合一致性”，set agreement）来衡量恢复效果，但这种方法仅依赖于相对排序信息，无法反映绝对性能水平。当所有动作均失败时，各动作获得零奖励，导致独立运行产生相同的平局集合，从而造成系统稳定性的假象，掩盖了实际恢复成功率趋近于零的严重问题。作者通过构建集合路径对称性（set-path symmetry）理论结果，证明在等成本伯努利动作场景下，成功概率分别为(0.9, 0.8)与(0.2, 0.1)的两种情形在任意样本量下均会产生完全相同的最优动作集合分布，任何仅依赖“哪个动作胜出”的评估方法都无法区分这两种截然不同的恢复能力状态。进一步地，研究证明在有限时间内无法精确验证总体平局，且检查点与动作之间的映射关系蕴含着超出边际分布的信息。为此，论文提出引入“合并成功率”（pooled success probability）作为关键补充指标，以消除仅有序度量带来的模糊性。实验在864个冻结的RecoveryBench任务实例及总计3,456次规划响应上验证了理论预测：集合一致性与保留样本质量可能背道而驰，且重新排列检查点-动作绑定可导致8%至13%的单元级结论发生变化。基于此，论文建议报告四项诊断指标——集合一致性、全零占比、保留样本成功率和合并成功率——以无须额外数据采集即可揭示该评估范式中的根本缺陷。
链接: https://arxiv.org/abs/2609.34215
作者: Dong Xu,Zhangfan Yang,Jiantao Wu,Shipeng Zhang,Zexuan Zhu,Jiangqiang Li,Jun Zhang,Junkai Ji
机构: Shenzhen University (深圳大学); University of Nottingham Ningbo (诺丁汉大学宁波分校); EasternDawn
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: 44 pages, 2 figures
Abstract:Evaluating how LLM agents recover from mid-task failures is central to deploying reliable agentic systems. Existing checkpoint-based benchmarks measure recovery by comparing which action is selected as best across independent runs, a quantity known as set agreement. However, set agreement is a purely ordinal measure that records which action wins without reflecting the absolute level of performance. When all actions fail, they tie at zero reward, and independent runs produce the same tied set with high probability, creating an illusion of stability that masks near-zero recovery success. We formalize this limitation through a set-path symmetry result, proving that for equal-cost Bernoulli actions the success probabilities (0.9, 0.8) and (0.2, 0.1) yield identical best-action-set distributions at every sample size. No procedure based solely on which action wins can distinguish these two regimes. We further prove that certifying exact population ties is impossible in finite time, and that the assignment of outcomes to checkpoints carries information beyond marginal outcome distributions. The pooled success probability is the missing scalar that resolves the ordinal ambiguity. Experiments on 864 frozen RecoveryBench episodes and two planning cohorts totaling 3,456 responses confirm the theoretical predictions. Agreement and held-out quality can move in opposite directions, and permuting checkpoint-to-action bindings changes 8 to 13 percent of cell-level conclusions. Based on these findings, we propose reporting four diagnostic quantities (agreement, all-zero fraction, held-out success, and pooled success) that expose this failure mode with no additional data collection.
[MA-10] ReplayLens: Auditing Agents Use of Outcomes
【速读】：该论文旨在解决在智能体（agent）重用记录经验时，无法明确区分决策变化是由评分（score）、动作名称（action name）还是记录在存储中的位置（record position）所驱动的问题。传统记忆评估方法无法揭示这些潜在关系对决策的影响，因而难以判断经验重用的安全性。其解决方案的关键在于提出ReplayLens——一种黑箱审计机制，通过逐一对四个关键关系进行干预：结果重分配（outcome reassignment）改变评分与动作的对应关系；成对迁移（pair transport）将完整的动作-评分对移至新存储位置；一致重命名（consistent renaming）同步更新历史与菜单中的动作标签；关键槽位重分配（key-slot reassignment）同时改变评分附着方式与记录位置。实验表明，仅评分重分配会显著影响决策，而完整动作-评分对的移动则不影响，这揭示了评分附着与记录顺序之间的可分离性。此外，研究发现端到端准确率相同的两个记忆系统在相同回放下表现不同，说明传统评估无法识别底层依赖关系。在大语言模型（LLM）接口、有限记忆场景及序列实验规划等任务中，历史评分的微小变化可导致探索路径偏离并降低最终效用，即使后续测量是全新的。该方法为判断日志经验是否可安全合并、重排或重新索引提供了基于关系层面的审计能力。
链接: https://arxiv.org/abs/2609.34177
作者: Dong Xu,Zhangfan Yang,Jiantao Wu,Shipeng Zhang,Zexuan Zhu,Jiangqiang Li,Jun Zhang,Junkai Ji
机构: Shenzhen University (深圳大学); University of Nottingham Ningbo (诺丁汉大学宁波分校); EasternDawn
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: 76 pages, 8 figures
Abstract:When an agent reuses logged experience, a changed decision may reflect the recorded score, the action’s name, or the record’s position in storage. Standard memory evaluations do not reveal which relationship drives that change. We introduce ReplayLens, a black-box audit that changes one relationship in the stored history at a time, holds the remaining interface fixed, and measures the resulting decision. Four interventions target four relationships. Outcome reassignment swaps which scores belong to which actions. Pair transport moves intact action-score pairs to new record slots. Consistent renaming relabels actions in both history and menu. Key-slot reassignment changes both score attachment and position. A constructive separation shows why the audit is needed: two memory writers with identical endpoint accuracy respond differently to the same replay, so conventional evaluation cannot resolve the underlying dependence. On black-box LLM interfaces, swapping scores changes decisions while moving intact pairs does not, separating score attachment from record order. A bounded-memory study exposes ingestion-order sensitivity that endpoint comparison misses. In sequential experiment planning, altered historical scores redirect exploration and reduce final utility despite fresh measurements. A code-debugging agent with sealed hidden tests shows the same pattern outside model selection. ReplayLens provides a relationship-level audit for deciding whether logged experience can be merged, reordered, or reindexed safely.
[MA-11] Maat: Independent Deterministic Contract-Based Governance for Multi-Agent LLM Workflows
【速读】：该论文旨在解决大语言模型多智能体系统（LLM-MAS）中因代理间错误传递导致的可靠性问题：单个代理产生的错误可能被下游代理作为上下文接受并沿工作流传播，进而引发连锁故障。现有防护机制多依赖于基于学习或大语言模型的评判器，其判断结果本身具有概率性，难以保证确定性。为此，论文提出Maat——一种运行时治理层，通过与版本化的工作流合约（即“锚点”）进行比对，实现对代理间交接的确定性验证，且验证与评分路径中不引入任何语言模型。其核心创新在于采用确定性规则（七项检查清单）完成缺陷检测，从而避免模型不确定性带来的风险。在六个受控领域工作流（6–15个代理，共522次试验）中注入数据级缺陷进行评估，结果显示第一版在所有六个工作流中均取得改进（提升2.9%–26.5%）。然而，后续审计发现部分基准评分器将任意早期终止均视为缺陷防止，而经人工审查的94次治理中断中，35次为误报（37%），源于验证器自身缺陷而非模型行为；若将这些误报计入失败任务，则治理组在四个工作流中表现劣于未治理组。研究结果表明，对于可由合约表达的缺陷，确定性交接验证具备有效性，但验证器配置与终止归因机制必须经过严格测试；该方法不保证普遍正确性、幻觉检测能力或完全独立于模型的有效性。
链接: https://arxiv.org/abs/2609.34017
作者: Uliana Elina
机构: SynWe Group s.r.o.
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 16 pages, 6 figures, 6 tables. Benchmarks: this https URL ; CrewAI integration demo: this https URL
Abstract:Large-language-model multi-agent systems (LLM-MAS) introduce a characteristic reliability problem: an error produced by one agent can be accepted as context by downstream agents and propagate across the workflow. Many proposed safeguards rely on learned or LLM-based judges whose verdicts are themselves probabilistic; we ask whether a deterministic layer can instead stop contract-detectable handoff defects. We present Maat, a runtime governance layer that validates agent-to-agent handoffs against a versioned workflow contract, or anchor, with no language model in the validation or scoring path. We evaluate it in six controlled domain workflows (6-15 agents, 522 trials) with injected data-level defects and a deterministic seven-check rubric. Version 1 reported gains in all six workflows (2.9-26.5%). A post-publication audit found that three benchmark scorers credited any early halt as a prevented defect. On paired trials where the governed run completed or halted on a finding attributable to a verified defect, the rubric score changes by +7.7% to +29.1% in five workflows and is flat in software development; model-call cost falls 17-53% where attributable halts occur early. A hand review of all 94 governed-arm halts found 35 false alarms (37%), caused by validator defects rather than model behaviour; counting those halts as failed work, the governed arm scores below the ungoverned arm in four of six workflows. The results support deterministic handoff validation for contract-expressible defects and show that validator configuration and halt attribution must themselves be tested; they do not establish universal correctness, hallucination detection, or model-independent effectiveness.
[MA-12] Prospective Interpretation Risk: Principled Communication Control Between LLM s
【速读】：该论文旨在解决大规模语言模型（Large Language Model, LLM）代理系统中因接收方对消息的解释差异而导致的通信不确定性问题，特别是在异构系统中，同一消息可能被不同能力的接收方重构为不同的任务。其核心挑战在于现有方法缺乏在消息发送前对特定接收方如何理解该消息的预测能力。为此，论文提出将该问题建模为一个带有隐变量接收方类型（latent receiver type）的发送-接收问题，并引入前瞻性解释风险（Prospective Interpretation Risk, PIR），即接收方误解任务的概率。解决方案的关键在于采用黑盒探测（black-box probes）来建立消息、预期任务与接收方特异性重构之间的映射关系，从而在不依赖全模型输入输出行为的前提下实现可扩展的监督学习，并将解释错误与下游能力失效解耦。通过离线利用冻结的异构接收方提供条件于接收方的风险监督信号，以及在部署时基于历史信息推断接收方类型的后验分布以指导消息修订与选择，实现了对解释风险的精准建模。进一步提出解释信息价值（Value of Interpretation Information, VoII），仅在预期通信收益超过成本时查询接收方信息，理论上刻画了接收方信息的决策价值并对其查询进行边界约束。实验表明，不同接收方的解释失败率相差达4–13倍，引入接收方信息使PIR校准误差降低68%，显著修正了接收方特异性风险水平；基于PIR的消息修订使解释失败率降低44%（相较原消息）和40%（相较通用重写），主要得益于对所有接收方均有效的修复机制。VoII在相同成本下优于信息增益与随机查询策略，在解释目标上将失败率从3.84%降至3.79%，同时仅触发18.2%的查询。
链接: https://arxiv.org/abs/2609.33885
作者: Wanrong Yang,Rehan Deen,Julian Ma,Yuheng Fan,Yaoyu Jin,Taher Jafferjee,Ziquan Liu,Dominik Wojtczak,Yalin Zheng,David Henry Mguni
机构: University of Liverpool; Independent Researcher; Queen Mary University of London; University College London
类目: Multiagent Systems (cs.MA); Machine Learning (cs.LG)
备注:
Abstract:Large language model (LLM) agentic systems increasingly rely on models communicating with one another, yet existing uncertainty and multi-agent methods rarely estimate how a particular receiver will interpret a message before it is sent. This matters in heterogeneous systems, where capable receivers can reconstruct different tasks from the same message. We model this as a sender-receiver problem with a latent receiver type and define prospective interpretation risk (PIR): the probability that a receiver reconstructs a task other than intended. Rather than model an LLM’s full input-output behaviour, we use black-box probes relating messages, intended tasks, and receiver-specific reconstructions, yielding scalable supervision while separating interpretation from downstream capability failure. Offline, heterogeneous frozen receivers provide supervision for receiver-conditioned risk and the effects of predefined mutable message features. At deployment, history induces a posterior over receiver types, guiding message revision and selection. We introduce value of interpretation information (VoII), querying for receiver information only when its expected communication benefit exceeds its cost. Our theory characterises when receiver information has decision value and bounds such queries. Empirically, interpretation-failure rates vary by 4-13x across receivers. Receiver information reduces PIR calibration error by 68% relative to a receiver-agnostic predictor, largely by correcting receiver-specific risk levels. PIR-guided revision reduces interpretation failure by 44% relative to the original message and 40% relative to a generic rewrite, mostly through a repair that helps every receiver. VoII outperforms information-gain and random querying at matched cost on the interpretation objective it optimises, lowering interpretation failure from 3.84% to 3.79% while querying 18.2% of episodes.
[MA-13] Population Physics Population Problems: Safety and Emergence in LLM Societies
【速读】：该论文旨在解决大规模语言模型（Large Language Model, LLM）社会系统中集体行为的涌现问题，即当多个LLM作为代理协同运作时，其整体行为并非个体输出的简单叠加，而是可能产生统计上显著且难以预测的自组织现象。传统用于分析单一智能体的工具在面对此类多智能体系统时往往失效。为应对这一挑战，论文提出了一种衡量LLM社会系统中自组织行为的框架，并将其应用于三个典型场景：谢林格子（Schelling grid）、社交网络模拟（Moltbook）以及类推推特的信息误导模拟系统（Rogue）。研究发现，这三个系统均表现出显著的自组织特征，且其弛豫动力学受环境信息可及性的影响，在开放性系统（如Moltbook与Rogue）中呈现出类似相变的急剧变化。更关键的是，即使在模型经过安全调优或监控的情况下，群体层面的病理行为仍可能由部分智能体的协调活动引发。此外，研究还识别出在两种额外情境下（公共资源困境中的GovSim，以及基于LLM作为裁判的决策方案ChatEval）自组织并未显现。因此，论文的核心贡献在于提出一种轻量级、不依赖自然语言或模型版本的代理无关诊断方法，通过检测自组织的统计签名，实现对部署中多智能体系统内协调集体行为的有效识别。
链接: https://arxiv.org/abs/2609.33871
作者: Adrian de Wynter
机构: 未知
类目: Multiagent Systems (cs.MA); Computation and Language (cs.CL)
备注:
Abstract:The collective behaviour of large language model (LLM) societies is not the sum of their individual outputs. It yields statistically distinct, sometimes-unpredictable phenomena, for which the tools we use to study single agents may not scale. Due to recent incidents involving autonomous agentic systems, however, understanding these systems is paramount. For that we introduce a framework for measuring self-organisation in LLM social systems and apply it to three such systems: a Schelling grid, a social network (Moltbook), and a Twitter-like misinformation simulation (‘Rogue’). All three exhibit statistically significant self-organisation. Moreover, their relaxation dynamics vary with the environmental information available to the agents, with open-ended systems (Moltbook, Rogue) exhibiting sharp, phase-transition-like dynamics. Further results show that population-level pathologies can emerge even when the LLMs are safety-tuned or monitored, being primarily driven by the coordinated activity of a population subset. We also show when self-organisation does \textitnot emerge under two additional scenarios (a commons dilemma, GovSim, and a LLM-as-a-judge deliberation scheme, ChatEval). We argue that measuring signatures of this kind offers a lightweight, agent-agnostic diagnostic layer for detecting coordinated collective behaviour in deployed multi-agent systems without relying on natural language or model versioning.
[MA-14] DEALS: Decentralized Expertise-Aware Load Serving for Multi-Agent LLM Systems
【速读】：该论文旨在解决多智能体系统（Multi-agent Systems, MAS）在处理异构且动态变化的任务流时，因依赖中心化控制器或固定协调模式而导致的可扩展性与自适应性不足，以及现有去中心化方法中因需训练专用路由模型或频繁调用大语言模型（LLM）进行代理选择而带来的高计算开销和协调延迟问题。其解决方案的关键在于提出一种去中心化、低复杂度的框架——基于专业知识感知的任务服务系统（Decentralized Expertise-Aware Load Serving, DEALS），通过让每个智能体本地维护任务队列，并基于自身与邻居在任务积压量与成功率上的差异，自主决策是否本地处理或转发任务，实现任务的动态路由；同时，执行器可在智能体内部及跨智能体并行处理独立任务，部分求解的任务可由其他智能体继续恢复，从而支持任务级与负载级的自组织与自演化。实验表明，DEALS在同质与异质代理池中均显著提升了答案准确率与任务吞吐量，并实现了代理专业能力与工作负载的自我均衡分配，有效支撑了高效去中心化协同。
链接: https://arxiv.org/abs/2609.33768
作者: Jingjuan Huang,Wenbin Wang,Yanchuan Yin,Alvaro Velasquez,Jia Liu
机构: Ohio State University (俄亥俄州立大学); University of Colorado at Boulder (科罗拉多大学博尔德分校)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 15 pages, 4 figures, 5 tables
Abstract:Multi-agent systems (MAS) have recently emerged as an effective approach for coordinating large language model (LLM)-based agents to solve complex tasks through structured interactions. In practice, MASs often handle a stream of heterogeneous and complex tasks, requiring agents to decompose each task and then self-organize and self-evolve to adapt to incoming tasks while sharing execution resources. However, most early approaches to MASs rely on centralized controllers or fixed coordination patterns, which can limit scalability or adaptability. In contrast, existing decentralized and dynamic MASs often require training dedicated routers or invoking LLMs for agent selection, resulting in substantial computational costs and coordination overhead. To address these challenges and enable efficient task-level self-organization and self-evolution for task- and workload-level collaboration, we propose Decentralized Expertise-Aware Load Serving (DEALS), a decentralized and low-complexity framework that enables agents to self-organize and dynamically route concurrent tasks for processing. Specifically, each agent maintains local queues of incoming tasks, and its router decides whether to process a task locally or forward it to a neighbor based on differences in backlog and success rate. Meanwhile, executors process independent tasks concurrently within and across agents, and partially solved tasks can be resumed by other agents. Experiments show that DEALS not only improves performance along multiple dimensions (e.g., answer accuracy and task throughput) in both homogeneous and heterogeneous agent pools, but also balances agent expertise and workload in a self-organized manner, enabling effective decentralized coordination.
[MA-15] MA-JEPA: Joint-Embedding World Models for Multi-Agent Reinforcement Learning
【速读】：该论文旨在解决多智能体强化学习（Multi-Agent Reinforcement Learning, MARL）中样本效率低下的问题，核心挑战在于如何有效学习能够支持未来控制决策的环境表征。现有基于世界模型（World Model）的方法依赖于对观测的重建来获取学习信号，但这种监督方式可能无法充分捕捉对策略优化至关重要的动态信息。为此，本文提出一种基于自监督联合嵌入预测（Joint-Embedding Prediction, JEPA）的新框架——MA-JEPA，其关键创新在于将传统的观测重建任务替换为对目标表征的预测，从而提供更有效的学习信号。该方法通过一个分类潜变量（categorical latent state）和因果Transformer（causal Transformer），分别以后验分布和动作条件化的方式建模动态过程，并利用隐空间中的想象轨迹进行演员-评论家（actor-critic）学习。训练阶段引入一个仅用于训练的联合预测器，基于所有智能体的局部状态与动作，预测每个智能体下一时刻的局部观测嵌入；这些预测结果经由与真实交互中相同的局部后验网络处理，输入至集中式评论家进行价值学习，而执行阶段仍保持去中心化。实验表明，该架构在SMAC基准上表现优异，在八张地图中的四张上达到或超越已有最优方法的平均胜率，验证了其在提升样本效率和策略性能方面的有效性。
链接: https://arxiv.org/abs/2609.33563
作者: Brandon Gary Kaplowitz,Osaze James Obahor,Christian Schroeder de Witt
机构: University of Oxford(牛津大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:World models improve sample efficiency by training policies on imagined trajectories, but their usefulness depends on learning representations that capture the information needed for future control. We study whether self-supervised joint-embedding prediction (JEPA) can provide this learning signal for multi-agent reinforcement learning. We introduce MA-JEPA, a stochastic world model that replaces observation reconstruction with prediction of target representations, enabling model-based multi-agent reinforcement learning with centralized training and decentralized execution. A categorical latent state and a causal Transformer are trained with posterior and action-conditioned dynamics prediction objectives and are then used for actor-critic learning from latent imagination. A training-only joint predictor conditions on all agents’ local states and actions to predict each agent’s next local observation embedding. These predictions are passed through the same local posterior used during real interaction with a centralized critic that is used only for value learning, with execution remaining decentralized. Our experiments show that this architecture performs strongly on SMAC, matching or exceeding the strongest reported comparator mean win rate on four of eight evaluated maps.
[MA-16] Nearly Group-Separable Elections
【速读】：该论文旨在解决如何度量给定选举与组可分（group-separable）性质之间的接近程度，以选票中相邻候选人交换（swaps）的次数作为距离度量标准。此外，研究还扩展至其他若干选举结构域，包括猫爪形组可分（caterpillar group-separable）、平衡组可分（balanced group-separable）、单峰（single-peaked）及单交叉（single-crossing）等。尽管该问题在一般情况下属于计算上难以处理（intractable），但研究发现，当以候选人数量或交换次数为参数时，存在实用的固定参数可追踪（FPT）算法。尤其值得注意的是，针对交换次数参数化的情形，所提出的算法适用于所有由有限禁止子选举（finite forbidden subelections）所刻画的选举域，从而解决了长期存在的一个开放性问题。研究通过理论分析与实验评估相结合的方式，验证了方法的有效性与实用性。
链接: https://arxiv.org/abs/2609.33535
作者: Piotr Faliszewski,Jan Jabrocki,Stanisław Kaźmierowski,Kristýna Pekárková,Šimon Schierreich,Ildikó Schlotter
机构: 未知
类目: Computer Science and Game Theory (cs.GT); Multiagent Systems (cs.MA)
备注:
Abstract:We study the problem of computing how close a given election is to being group-separable, measuring proximity by swaps of adjacent candidates in the votes. We also consider several other domains, including caterpillar group-separable, balanced group-separable, single-peaked, and single-crossing ones. Our problem is generally intractable, but we find practical FPT algorithms parameterized by the number of candidates or swaps. For the latter case, our algorithm applies to all domains characterized by finite forbidden subelections, resolving a well-established open problem. We supplement our theoretical findings with experimental analysis.
[MA-17] RACE: Governing Memory Validity in Evolving Multi-Agent Systems
【速读】：该论文旨在解决语言模型代理在长期协作中使用持久记忆（persistent memory）时面临的时序记忆准入问题：当代理返回时，如何判断其携带的过往记忆是否仍适用于当前共享状态下的任务。核心挑战在于，即使记忆在语义上正确、相关且忠实于原始来源，也可能因中间状态变更而失效（如团队已更换酒店，但代理仍持有旧行程）。为此，论文提出一种无需训练的模块——TRACE，其关键创新在于将代理重新进入（re-entry）视为一个“资格判定”（eligibility decision）而非传统的存储或检索操作，通过对比离开时的检查点与缺席期间的更新，显式和隐式地处理失效情况，并仅在返回角色的开放义务被完整覆盖时释放有限的“返回视图”（Return View）。实验在Memora、STALE Type II及衍生的ManBench-Return设置下验证，结果表明，现有单策略方法无法同时兼顾有效记忆保留与过时记忆剔除（Restore会引入过时状态，Reset则丢失有效信息，二者在各自维度均降至0%），而TRACE在两者上均表现优异，在ManBench-Return上实现92.6–98.3%的有效信息可用率与98.4–99.5%的无效信息拒斥率，接近最优基线整体准确率；在STALE Type II上，其总体性能分别超越Qwen、Gemini、DeepSeek 22.3、18.5、27.5个百分点，仅需约其2.3倍的生成令牌，显著优于其他方法。
链接: https://arxiv.org/abs/2609.33517
作者: Wenjun Xiong,Shengtao Zhang,Shangding Gu,Bo Tang,Zhiyu Li,Feiyu Xiong,Ying Wen,Muning Wen
机构: Shanghai Jiao Tong University (上海交通大学); Shanghai Innovation Institute (上海创新研究院); UC Berkeley (加州大学伯克利分校); MemTensor(Shanghai) Technology Co., Ltd. (MemTensor(上海)科技有限公司)
类目: Multiagent Systems (cs.MA)
备注: 41 pages, including appendices. Code: this https URL
Abstract:Persistent memory lets language-model agents carry information across long-running collaborations, but leaves a lifecycle question open: what may a returning agent still act on once the shared state has changed? A memory can be correctly retrieved, relevant to the current task, and faithful to its source, and nonetheless be inadmissible for action: an itinerary saved before a pause still names the hotel the team has since replaced. We formalize this as temporal memory admission and present TRACE, a training-free layer that treats re-entry as an eligibility decision rather than a storage or retrieval operation, reconciling a departure checkpoint against absence-period updates, resolving explicit and implicit invalidation, and releasing a bounded Return View only when it covers the returning role’s open obligations. We evaluate TRACE under three actor models on Memora, STALE Type II, and a derived ManBench-Return setting, each recast as return episodes: one agent departs, four teammates change the shared state, and the agent rejoins. What separates methods is not overall accuracy but whether one can retain valid memory and reject stale memory at once, and no single-policy baseline can: Restore (reinstate the departure checkpoint in full) admits stale state, Reset (start the return from an empty memory) discards valid state, each bottoming out at 0% on one of the two. TRACE is the only method high on both, reaching 92.6-98.3% valid-information availability with 98.4-99.5% invalid-information rejection on ManBench-Return, within 3.8 points of the best baseline’s overall accuracy. On STALE Type II it improves Overall over the strongest comparison policy by 22.3 (Qwen), 18.5 (Gemini), and 27.5 (DeepSeek) points at roughly 2.3 times their tokens, while a write-time consolidation pipeline is more accurate still at 3.99 times TRACE’s.
[MA-18] LLM s Trust Their Own: Identity-Dependent Conformity in Multi-Agent Systems
【速读】：该论文旨在解决大语言模型（Large Language Models, LLMs）在多智能体（multi-agent）环境中如何受其他智能体社会身份（social identity）影响的问题，尤其关注社会身份对模型决策一致性的影响是否独立于群体共识（consensus）本身。现有研究多聚焦于共识效应，但忽视了社会身份这一关键因素在人工智能行为与安全中的作用。本文通过设计具有唯一正确答案的判断任务，构建多智能体场景，使模型面对来自同质或异质社会身份（如人类或AI、模型家族、任意最小群体）的其他智能体所给出的错误答案。实验覆盖12个开源权重模型和9项任务，发现社会身份存在双向影响：当其他智能体属于同一群体时，模型更易服从错误共识（内群偏好），而当其他智能体属于对立群体时，则更倾向于拒绝错误共识（外群分化）。值得注意的是，与人类不同，模型对多数群体中一个“盟友”打破共识的行为不敏感；反而，一个来自对立群体的正确盟友会加剧这种双向效应。尽管链式思维（Chain-of-Thought）推理可抑制大部分社会身份效应，但仍保留内群盟友降低对外群多数错误共识服从性的现象。此外，将同伴标记为“安全对齐”（safety-aligned）虽能整体降低服从性，但无法消除内群偏好与外群分化。研究揭示了社会身份在模型跨智能体信息聚合过程中的独立作用，表明其已成为操控多智能体系统行为的关键干预维度。
链接: https://arxiv.org/abs/2609.33495
作者: Liron Soffer,Ravid Shwartz-Ziv,Chen Shani
机构: Tel-Aviv University (特拉维夫大学); New York University (纽约大学)
类目: Computation and Language (cs.CL); Multiagent Systems (cs.MA)
备注:
Abstract:Large language models (LLMs) are increasingly deployed in multi-agent settings, where agents observe and influence one another, making social influence a key dimension of AI behavior and safety. We investigate whether LLMs’ responses depend on the social identity of other agents, beyond the effect of their consensus. We construct judgment tasks with a single correct answer, and place models in a multi-agent setting where they receive incorrect answers from other agents whose social identities (AI or human, model family, or an arbitrary minimal group) are either shared with or distinct from their own. Across 12 open-weights models and nine tasks, we find a bidirectional effect of group identity on conformity to incorrect answers: in-group consensus increases conformity (in-group favoritism), whereas out-group consensus decreases it (out-group divergence). Unlike humans, for whom one ally breaking the consensus sharply reduces conformity, models are unmoved by an ally from the majority’s group. Worse, a correct ally from the opposing group intensifies this bidirectional effect. Chain-of-Thought reasoning suppresses most of these effects, yet an in-group ally still reduces conformity to an incorrect out-group majority. Labeling peers as safety-aligned shifts overall conformity but leaves in-group favoritism and out-group divergence intact. These results show that group identity shapes how LLMs aggregate information across agents, independently of its correctness, and identify a manipulation surface for multi-agent AI systems.
[MA-19] Raven: The Harness of Harnesses for Composable Agent ic Intelligence
【速读】：该论文旨在解决当前大型语言模型（Large Language Models, LLMs）在迈向长时序、跨领域任务协作过程中所面临的两大挑战：一是任务编排（harness）的复杂性随任务规模增长而急剧上升，导致人工设计难以规模化；二是现有编排方案与特定领域强耦合，限制了其通用性。针对这一问题，论文提出的核心解决方案是构建“编排之编排”（Harness of Harnesses）——Raven，一个开源的多智能体生态系统。其关键在于通过自动化构建、持续进化及跨域协同机制，将每个可执行的模型-编排对视为可组合的智能单元，并由宿主代理（Host Agent）实现目标分解、子任务匹配、执行依赖协调与结果融合。同时，系统通过宿主存档（Host Archive）、EverOS（长期记忆机制）和技能锻造（Skill Forge）实现经验积累与可复用技能的生成。理论分析证明，在共享资源预算下，该组合架构能够扩展任务覆盖范围，超越单一智能体的能力边界。实验表明，Raven在复杂长时序任务上显著优于现有最先进智能体系统，推动了可组合智能体技术的发展前沿。
链接: https://arxiv.org/abs/2609.33439
作者: EverMind AI
机构: EverMind AI( EverMind AI); https://github.com/EverMind-AI/Raven(https://github.com/EverMind-AI/Raven)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Science and Game Theory (cs.GT); Multiagent Systems (cs.MA); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:As large language models advance, AI agents are moving beyond isolated, domain-specific tasks toward long-horizon, cross-domain workflows. This transition exposes two challenges: increasing harness complexity makes manual design difficult to scale, while tighter coupling to specific domains limits the generality of a single harness. The central question thus shifts from how to engineer a stronger harness for one domain to how to autonomously construct specialized harnesses, improve them through experience, and orchestrate them across domains. We introduce Raven, \emphThe Harness of Harnesses, an open-source multi-agent ecosystem that automatically constructs and evolves modular harnesses for specific models and domains, treating each executable model–harness pair as a composable unit of intelligence. To support an \emphAll-Domain Collaboration Network, its Host Agent decomposes goals, matches subtasks to specialized agents, coordinates execution dependencies, and integrates results, while a host archive and EverOS preserve experience across tasks and Skill Forge makes that experience available as reusable procedures. Our theory establishes sufficient conditions for such composition to expand reliable task coverage beyond that of the available individual agents under a shared resource budget. On complex and long-horizon tasks, Raven significantly outperforms the state-of-the-art agent systems, pushing the frontier of composable agentic intelligence.
[MA-20] DataMagic: Authoring Data Videos through Declarative Multi-Agent Orchestration IEEE-VIS2026
【速读】：该论文旨在解决自动化的数据视频（Data Video）生成中面临的两大核心挑战：一是如何统一建模图表、语音解说与动画之间的动态关系及其时间同步性，二是如何在庞大的设计空间中高效搜索出叙事连贯的组合方案。现有方法存在明显局限：静态可视化工具缺乏叙事与动画能力，传统创作工具依赖预设图表而非原始数据，而端到端像素级模型虽能生成视频却难以保证数据准确性与可追溯性。为此，本文提出DataMagic系统，其关键创新在于采用声明式多智能体协同架构。首先，通过声明式规范DVSpec统一表达图表、解说与动画，并基于数据绑定引用和声明式同步机制，确保数据溯源与音画自动对齐；其次，采用“先生成后编排”的多智能体策略，在并行生成候选场景的基础上，通过全局编排优化叙事连贯性。DVSpec还提供共享状态支持全自动化与细粒度人工控制的融合交互模式。实验结果表明，相较于最先进的大语言模型（如GPT-5），DataMagic在109个真实样本上将生成质量从2.13提升至3.89（+83%），成功率超过95%，尤其在动画与叙事维度提升显著；用户研究进一步验证了其在创作效率（任务耗时减少79.7%）与认知负荷降低方面的优势。
链接: https://arxiv.org/abs/2609.33403
作者: Yupeng Xie,Zhenyang Wang,Liangwei Wang,Jiayi Zhu,Zhouan Shen,Yuyu Luo
机构: 未知
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Databases (cs.DB); Multiagent Systems (cs.MA)
备注: Accepted at IEEE VIS 2026
Abstract:Data videos communicate data insights through dynamic charts, voice narration, and synchronized animations, and have become a widely adopted form of data storytelling. However, producing them requires expertise in data analysis, narrative design, and video editing. Static visualization tools lack narrative and animation capabilities; authoring tools rely on pre-prepared charts rather than raw data; and pixel-level models generate videos end-to-end but cannot guarantee data accuracy or provenance. End-to-end automatic generation faces two core challenges: how to uniformly represent charts, narration, and animations together with their temporal relationships, and how to efficiently search a vast design space for narrative-coherent compositions. We present DataMagic, which authors data videos from raw tabular data through declarative multi-agent orchestration. First, the declarative specification DVSpec unifies charts, narration, and animations with data-bound references and declarative synchronization, ensuring data provenance and automatic audio-visual alignment. Second, a “Generate-then-Orchestrate” multi-agent strategy generates candidate scenes in parallel and then optimizes narrative coherence through global orchestration. DVSpec provides a shared state for three complementary interaction modes, bridging full automation with fine-grained human control. Evaluations on 109 real-world samples show that even the most advanced LLM (e.g., GPT-5) achieves only 2.13/5 with execution success rates between 48.62% and 86.24%; DataMagic improves quality to 3.89 (+83%) with success rates above 95%, with the most significant gains in animation and narrative dimensions. A user study shows that, compared to a conversational LLM workflow, DataMagic improves creation efficiency (79.7% reduction in task time) and reduces perceived cognitive load. Project page: this https URL.
[MA-21] SIVIA-RSI: Source-Grounded Adaptation of Diagramming Skills
【速读】：该论文旨在解决科学方法图示（scientific method diagrams）在不同研究论文间可复用技能迁移效果不明确的问题，即先前论文中的绘图经验是否能有效提升后续论文中首个图示的生成质量。其核心解决方案是提出一个源基适配框架（\sys），通过将批评意见与原始文本段落关联、对持久化技能库进行有界编辑，并分离候选方案竞争与技能采纳过程，实现可复用图示技能的系统性学习与评估。关键创新在于引入源基关系评估（source-grounded relation assessment），强调必须基于原始文本内容检验图示中实体依赖与条件路径的准确性，而非仅依赖表面视觉一致性。实验表明，最优学习候选模型在必要关系准确率上达到87.50%，优于原始技能的83.33%，但跨论文表现差异显著，且局部改进与自动选择器偏好无法保证稳定迁移。进一步分析发现，大量错误源于生成提示中条件逻辑不完整或歧义，尤其在已终止节点的路径未被显式绑定时，导致图示语义模糊。研究最终指出，有效的可复用图示技能评估必须包含完整候选覆盖、源文本对齐的关系验证以及对生成提示与输出图像的联合审查。
链接: https://arxiv.org/abs/2609.33386
作者: Feng Yuan,Yifan Gao,Haoyue Li,Xin Gao
机构: University of Science and Technology of China (中国科学技术大学); Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences (中国科学院苏州生物医学工程技术研究所)
类目: Graphics (cs.GR); Multiagent Systems (cs.MA)
备注:
Abstract:Scientific method diagrams express computations through entities, dependencies, and conditional routes. Although generated figures can be improved through repeated editing, it is less clear whether experience from one paper improves the first figure of another. We present \sys, a framework for source-grounded adaptation of reusable diagramming skills, and study transfer through a complete-candidate evaluation. The framework links critiques to source passages, proposes bounded edits to a persistent skill library, and separates candidate competition from skill acceptance. We evaluate the original skill and four learned candidates on two NLP and language-agent papers, with two fresh generations per condition. The strongest candidate attains 87.50% required-relation accuracy compared with 83.33% for the original, while candidate behavior differs across papers. Local improvements on a separate development paper and automatic selector preferences do not establish consistent transfer. Tracing all 22 non-correct relation judgments to their production prompts reveals both incomplete conditional specifications and ambiguities despite explicit instructions. All ten planning diagrams leave an already-terminal selected leaf’s route unclear; none of their prompts explicitly binds that route. Our findings show why evaluating reusable diagram skills requires source-grounded relation assessment, complete candidate coverage, and inspection of both prompts and images. We provide all twenty transfer outputs, skill snapshots, assessment records, and executable analyses.
[MA-22] CORTEX: A Verified Experience Layer for Generalist Agents
【速读】：该论文旨在解决智能体在面对相同任务但处于新情境（如新事实、新工具或新规则）时，缺乏系统性判断能力以决定先前解决方案是否可复用、需调整或应被弃用的问题。现有代理系统虽具备文本检索或历史对话记忆功能，却未建立可信赖的决策机制来评估过往经验的有效性。为此，论文提出CORTEX（上下文协调与任务经验复用框架），其核心在于通过外部验证的经验层连接专业化智能体，将每个任务执行过程结构化记录为包含任务条件、工具状态、关键判定谓词、证明路径、验证器及结果的完整“经验片段”。系统通过元控制器动态选择精确重放、受控适配、全新生成或升级处理。经验证的实例可通过挑战驱动的开发循环演化为通用任务模式与程序策略，形成无需修改模型权重即可持续增长的隐式能力层。该框架形式化定义了精确重放与源版本分离的系统契约，并推导出经验复用的计算节省条件。在双领域控制实验中，对1,000个合成案例验证了精确重放的核心有效性；在跨八类临床与政策场景的全家族留出测试中，对1,000个新家族案例实现了完全的新证据置信接地与对无关字段及插入顺序扰动的完美不变性，转移轨迹清晰揭示了经验证策略执行所需的工作量。研究结果表明，通过可复用的程序化过程、类型化经验与发展性迁移，该框架为实现通用智能提供了一条可行路径。
链接: https://arxiv.org/abs/2609.33260
作者: Garapati Keerthana,Manik Gupta
机构: Birla Institute of Technology and Science, Pilani, Hyderabad, India(比特理工学院，皮拉尼，海得拉巴，印度)
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注:
Abstract:An agent can solve a task today and face the same task under new facts, tools, or governing knowledge tomorrow. Most agent systems can retrieve relevant text or recall prior conversations, but they lack a principled way to decide when a previous solution is still valid, when it must be adapted, and when it should be discarded. We introduce CORTEX (Contextual Orchestration and Reuse of Task EXperience), a general AI systems framework that connects specialized agents through an external layer of verified experience. Each episode records its task conditions, source and tool state, decisive predicates, proof trace, verifier, and outcome. A meta-controller chooses exact replay, checked adaptation, fresh synthesis, or escalation. Accepted episodes can become task patterns and procedural strategies through a challenge-driven development loop. This gives the system an implicit competence layer that can grow without changing model weights. We formalize system contracts for exact replay and source-version separation, and derive when reuse saves computation. A controlled two-domain implementation tests the exact-replay core on 1,000 synthetic cases. Complete-family holdouts test procedural transfer on 1,000 new-family cases across eight clinical and policy splits, with complete fresh-evidence grounding and perfect invariance to irrelevant-field and insertion-order perturbations. The transfer trace exposes the work required for verified strategy execution. These results establish an initial path toward general intelligence through reusable procedures, typed experience, and developmental transfer.
[MA-23] Modular Discovery of General Game-Playing Algorithms with Large Language Models
【速读】：该论文旨在解决仅基于游戏规则实现跨任意游戏的通用博弈智能（General Game Playing）这一挑战，其核心难点在于不同游戏类别对算法需求差异显著，且决策时间约束严格。传统方法依赖人工设计特定领域的搜索启发式策略，难以泛化。本文提出一种基于大语言模型（Large Language Models, LLMs）的多智能体元学习系统，通过让LLMs自主生成并迭代优化以C++实现的、与具体游戏无关的过程化搜索机制，同时直接从游戏规则中合成领域启发式信息。该方案的关键在于利用LLMs强大的结构化代码生成与重构能力，构建一个可自进化、可泛化的算法探索框架。在控制计算预算的前提下，该系统在超过400个多样化环境（包括OpenSpiel训练/测试集、程序化模拟引擎及采用PPO训练的深度神经网络策略的游戏）上进行了评估。实验结果表明，所发现的搜索机制在AlphaRank稳态分布和软康多塞优化（Soft Condorcet Optimization, SCO）指标下均显著优于15个主流蒙特卡洛树搜索（MCTS）基线，在多次独立演化运行中持续获得顶级排名和多数对比较优表现，展现出对未见的人工设计游戏与程序生成游戏的强大泛化能力，并在固定神经网络策略场景下仍保持竞争力。
链接: https://arxiv.org/abs/2609.33115
作者: Zun Li,John Schultz,Marc Lanctot,Daniel Hennes
机构: Google DeepMind(谷歌深脑)
类目: Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT); Multiagent Systems (cs.MA)
备注:
Abstract:General Game Playing across arbitrary games from rules alone remains challenging due to differing algorithmic requirements across game classes and strict decision-time constraints. Rather than hand-designing search heuristics for specific domains, can we leverage Large Language Models (LLMs) to discover general game-playing algorithms? Because language models can propose and refactor structured code, they provide an expressive proposal engine for exploring the space of algorithmic designs. We introduce a multi-agent LLM meta-learning system to co-evolve game-agnostic procedural search mechanisms in C++ alongside domain heuristics synthesized directly from game rules. Controlling the compute budget, we benchmark the discovered mechanisms across more than 400 diverse environments, including OpenSpiel training and held-out games, procedural simulation engines, and games with deep neural policy-value representations trained via PPO. Evaluated via AlphaRank stationary distributions and Soft Condorcet Optimization (SCO) against 15 established MCTS baselines, the discovered search mechanisms consistently achieve top-tier ratings and pairwise ballot majorities over most baselines across independent evolutionary runs, generalizing to unseen human-designed and procedurally synthesized games and remaining competitive with baselines on frozen neural network representations.
[MA-24] ParallelPilot: Supporting Coordination and Monitoring in Parallel AI Coding
【速读】：该论文旨在解决随着代码助手（coding assistant）日益自主化，开发者在并行执行多个任务会话时面临的协调与监控难题，即从单一的代码生成挑战转变为对多并发代理工作流的有效管理。其解决方案的关键在于提出并验证了PILOT框架——一种包含五项监督实践（Planning, Isolating, Logging, Observing, and Triaging）的系统性方法，并通过设计原型ParallelPilot将其具体实现：该工具集成了规划界面、运行日志记录器以及环境式仪表盘，无缝嵌入现有编码工具链中。实验结果表明，使用ParallelPilot的开发者在短周期编码任务中票务处理量提升63%，可同时有效监管的并发代理数量平均增加1个，且追踪负担和上下文切换频率显著降低；同时，任务执行计划、依赖关系及干预提示更加清晰，14名参与者偏好该工具超过当前工作流。然而，感知控制力与重定向代理的成功感未见显著提升。研究揭示了显式监督支持的价值，并建议未来的代码助手应在高阶态势感知与低成本回溯至实现证据之间建立平衡，以支持开发者对并发工作的有效判断与引导。
链接: https://arxiv.org/abs/2609.33113
作者: Tao Long,Weili Shi,Hussein Mozannar,Maya Murad,Rafah Hosn
机构: Microsoft Research AI Frontiers(微软研究院人工智能前沿); Columbia University (哥伦比亚大学)
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC); Multiagent Systems (cs.MA)
备注:
Abstract:As coding assistants become increasingly autonomous, developers run multiple sessions in parallel, shifting the challenge from code generation alone to coordinating and monitoring concurrent agent work. Through a formative study (N=14), we identified PILOT: five supervisory practices for Planning, Isolating, Logging, Observing, and Triaging parallel sessions. We present ParallelPilot, a design probe that instantiates PILOT through a planning interface, a run-logger, and an ambient dashboard alongside existing coding tools. In a counterbalanced within-subjects study (N=16), participants using ParallelPilot increased ticket throughput by 63% in short coding tasks and supervised an average of one more concurrent agent at peak, while their tracking effort and context switching dropped. ParallelPilot also clarified execution plans, task dependencies, and intervention cues, and 14 of 16 participants preferred it over their current setup. These gains were not accompanied by significant improvements in perceived control or perceived success in redirecting the agents. Our findings demonstrate the value of explicit supervision support and position PILOT as a scaffold for designing tools that help people supervise concurrent work within and beyond coding. We suggest that future coding assistants should pair high-level awareness with low-cost paths back to the implementation evidence developers need to judge and steer agent work.
[MA-25] ORBIT: A Framework for Multi-Agent Safety and Security Evaluations
【速读】：该论文旨在解决多智能体大语言模型（Multi-agent LLM systems）在复杂、长周期任务中日益凸显的安全与隐私风险问题，尤其针对由灵活通信协议引发的新型威胁，如级联式提示注入攻击和跨智能体合谋行为。现有防御研究受限于缺乏统一的实验基础设施，导致每次新防御方案均需定制化环境构建，难以实现标准化对比。为此，论文提出ORBIT——一个基于UK AISI’s Inspect构建的可配置评估框架，支持动态调整通信拓扑、记忆机制、调度策略及智能体角色，并涵盖四种攻击类型与四种防御策略，以及非对抗性故障场景。其基准测试套件覆盖浏览器使用、计算机操作、代理编程、客户服务与协作分配五大场景家族。核心发现表明：当前防御措施在不同攻击类型间缺乏泛化能力，例如在多议题代码生成任务中能降低60分攻击成功率的逐动作防御，在应对合谋攻击时几乎无效；且所有测试防御均未表现出对全部攻击类型的通用有效性。此外，研究揭示了安全性能之间的权衡关系，以及系统架构与防御策略之间存在显著交互效应。ORBIT已开源，为多智能体系统的安全与可靠性研究提供可复现、可扩展的实证平台。
链接: https://arxiv.org/abs/2609.33102
作者: Ben Hagag,William L. Anderson,Srija Chakraborty,Christian Schroeder de Witt
机构: Carnegie Mellon University (卡内基梅隆大学); MATS Research (MATS 研究); University of Oxford (牛津大学); Cooperative AI Foundation (协作人工智能基金会)
类目: Multiagent Systems (cs.MA); Cryptography and Security (cs.CR)
备注:
Abstract:Multi-agent LLM systems are increasingly deployed for complex, long-horizon tasks or emerge as a natural consequence of agents interacting in the wild. Yet they give rise to significant safety and security risks: the flexible protocols that enable task generalization also expose novel threats, from cascading prompt injection to inter-agent collusion. Progress in defending against these threats has been slowed by a lack of shared empirical infrastructure, which forces bespoke environment development for every new defense and makes standardized comparison impossible. Existing evaluations address isolated threat models or single-agent settings, but none jointly vary attack, defense, and architecture across realistic multi-agent environments. To address this gap, we introduce ORBIT, a configurable evaluation framework for empirical multi-agent safety and security research, built on UK AISI’s Inspect. ORBIT lets researchers configure communication topologies, memory, scheduling, and agent roles. It supports four threat types and four defense strategies, as well as non-adversarial failures, with a benchmark suite spanning five scenario families covering browser use, computer use, agentic coding, customer service, and cooperative allocation. Our central finding is a gap in defense transferability across threats: per-action defenses that cut a compromised agent’s attack success by 60 points on multi-issue coding give no measurable protection against colluding agents, and none of the defenses we tested generalized over all attacks tested. We further demonstrate security-performance tradeoffs and interactions between architecture and defense effectiveness. We make ORBIT available open-source at this https URL.
[MA-26] heory of Scene: Breaking the Symmetry Trap in Multi-Agent LLM Coordination
【速读】：该论文旨在解决基于大语言模型（LLM）的多智能体系统中因智能体同质化而导致的协作失效问题，即在无通信条件下，同质智能体在需分工的目标上发生冲突，在需协同的目标上又产生分歧，这种双重失败被称为“对称性陷阱”（symmetry trap）。现有依赖心智理论（ToM）的协调机制无法摆脱该陷阱，因其同质智能体对彼此的预测一致，导致响应方式相同，难以实现差异化行为。论文提出一种无需训练的推理范式——场景心智（Theory of Scene, ToS），其核心在于利用智能体间唯一的公开角色信息与共享的任务上下文进行推理：通过角色门控（role gating）判断目标所有权是否重叠或已分配，通过任务耦合（task coupling）推断团队应采取协同、分割或顺序执行等策略。由此，同质智能体可依据自身角色自动形成分工，将同质性从问题根源转化为解决方案的基石。实验在作者构建的可控评估基准DivvyBench及两个已有基准GovSim和Overcooked上验证，结果表明ToS在所有任务设置中均显著优于六种基线方法，尤其在DivvyBench上将成功率从71.1%提升至99.6%，在GovSim中总收益从207增至400，在Overcooked中归一化吞吐量从1.41提升至1.67，充分证明了ToS的有效性与普适性。
链接: https://arxiv.org/abs/2609.32939
作者: Liangqi Yuan,Wenzhi Fang,Shiqiang Wang,Christopher G. Brinton
机构: Purdue University(普渡大学); University of Exeter(埃克塞特大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Multiagent Systems (cs.MA)
备注:
Abstract:Multi-agent systems built on large language models (LLMs) are largely homogeneous, as their agents behave alike even across distinct LLMs. We show that when such agents act concurrently without communication, they collide on targets they must split and diverge on targets they must take together, a double failure we term the symmetry trap. Theory of Mind (ToM), widely used for coordination without communication, cannot escape this trap, since homogeneous agents form the same prediction of one another and respond to it in the same way. We propose Theory of Scene (ToS), a training-free reasoning schema in which each agent reads its public role, the only difference between the agents, and the task context they all observe. Homogeneous agents thereby derive one division of labor, each taking the part its role fixes, which turns homogeneity from the cause of the trap into the cure. ToS reads the role together with the scene through role gating, which determines whether ownership overlaps or is already divided, and the task context through task coupling, which infers whether the team must converge on each target, divide it, or take its stages in turn. We evaluate on DivvyBench, a controlled environment we introduce, whose target types make an episode Competitive, Cooperative, or Mixed across Tabletop, Airspace, and Household scenarios, and on two established agentic benchmarks, GovSim and Overcooked. ToS outperforms all six baselines on every benchmark, and each baseline falls far behind it in at least one setting. Against ToM given the same inputs, ToS raises the DivvyBench success rate from 71.1% to 99.6%, the GovSim total gain from 207 to 400, and the Overcooked level-normalized throughput from 1.41 to 1.67.
[MA-27] Planner-as-Router: Joint Plan-Time Model Routing for Cost-Efficient Multi-Agent Workflows
【速读】：该论文旨在解决大规模语言模型（Large Language Model, LLM）代理在生产环境中因模型层级选择不当而导致的高昂推理成本问题。随着模型能力与成本呈非线性增长，尤其在多步骤工作流中，若所有子任务均使用高成本前沿模型（frontier model），总开销将急剧放大。现有方法如级联路由（cascade routing）通常逐节点决策，缺乏对整体工作流依赖关系的全局视图，导致次优选择。本文提出的Planner-as-Router（PaR）方案的关键在于将模型层级选择嵌入规划阶段：在任务分解为子任务的同时，为每个子任务动态分配合适的能力层级（小、中、前沿），实现对整个工作流的全局优化。该方法无需额外的路由器模型或训练数据，通过显式建模子任务间的依赖关系，使资源分配与任务逻辑协同一致。实验基于EntBench基准，在54个企业级代理任务上评估了八种路由策略，结果表明，PaR在成本-精度权衡曲线上保持前沿位置——其精度与仅在终点节点使用前沿模型的启发式方法相当，且在相同成本下优于忠实的FrugalGPT级联方案，同时相比全使用前沿模型的策略降低44%成本，仅损失2.9个百分点精度。由于多数误差差异处于54项任务研究的±6点置信区间内，作者将PaR的优势定位于“前沿位置”而非显著精度提升。此外，初步观察提示廉价路由可能在组合型工作流中引发隐性累积惩罚，构成未来研究的假设方向。PaR、EntBench及全部评估代码均已开源。
链接: https://arxiv.org/abs/2609.32917
作者: Vivek Kumar Singh,Preeti Priyam,Gautam Bhowmick
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multiagent Systems (cs.MA)
备注: Accepted at AIxSET 2026. 8 pages, 4 figures, 6 tables. Data and Code in github: this https URL
Abstract:Running large language model (LLM) agents in production gets expensive fast. A frontier model (the largest, most capable tier) is accurate but can cost 25 times what a small model costs per token, and the gap compounds once a workflow chains several calls together. Planner-as-Router (PaR) attacks this from a different angle. Instead of leaving model-tier selection to some component downstream, it folds the choice into planning itself. As the planner breaks a query into subtasks, it also assigns each one a model size tier (small, mid, or frontier, ordered by capability and price), so the dependencies between subtasks are visible before any specialist runs. Unlike per-call routers such as cascade routing, which look at one node at a time, PaR sees the whole workflow up front and needs no separate router model or training data. We evaluate PaR with EntBench, a benchmark of 54 enterprise agentic tasks across seven classes, graded by actually running the generated Structured Query Language (SQL) and MongoDB queries against live databases. Over 1,157 evaluations spanning eight routers and three seeds, PaR stays on the observed cost-accuracy frontier. It matches a sink-frontier heuristic (frontier model on terminal nodes only) in accuracy at comparable cost and a faithful FrugalGPT cascade at lower cost, and cuts cost 44% against all-frontier routing while giving up 2.9 points of accuracy. Several accuracy gaps fall inside the plus-or-minus six-point confidence interval of a 54-task study, so we frame PaR’s advantage as frontier position rather than a clean accuracy win. We also report a preliminary observation, not a validated result: a small pilot hints that cheap routing may carry a hidden compounding penalty on compositional workflows, which we frame as a hypothesis for future measurement. PaR, EntBench, and all evaluation code are open source.
[MA-28] Improving LLM Collaboration via Multi-Agent Preference Learning
【速读】：该论文旨在解决多智能体强化学习（MARL）在大语言模型（LLM）协作中因缺乏可靠奖励信号而面临的挑战。在实际应用中，完整且准确的评估指标往往难以获取或难以聚合，导致传统基于固定奖励的设计受限。为此，论文提出一种基于偏好的多智能体系统（Preference-based Multi-Agent Systems, MAS），从去中心化与集中式协作两个视角构建框架，并引入通用的多智能体偏好学习框架（MAPL）。MAPL的关键在于通过迭代比较当前解与由不同智能体生成的去中心化或集中式解，实现对协作策略的持续优化。具体实现上，采用基于人类反馈的多智能体强化学习（MARLHF）与多智能体直接偏好优化（MADPO）两种方法。实验结果表明，MAPL可在协作写作、编程、工具使用及旅行规划等任务中显著提升协作质量与效率，性能接近使用固定且明确奖励的MARL。其中，MARLHF在多数任务中表现优于MADPO，但对数据覆盖范围、智能体及对比模型选择以及底层MARL算法较为敏感。
链接: https://arxiv.org/abs/2609.32827
作者: Shuo Liu,Xinzichen Li,Tianle Chen,Christopher Amato
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Several works have explored multi-agent reinforcement learning (MARL) in LLM collaboration. However, constructing reliable rewards is difficult in practice, as complete and accurate metrics are often unavailable and hard to aggregate. Preference learning provides an alternative by learning from comparative human or AI feedback. Yet, its extension to multi-agent systems remains underexplored. To address this gap, we formulate preference-based multi-agent systems (MAS) from decentralized and centralized collaboration perspectives. We also introduce a general multi-agent preference learning framework (MAPL) to solve these problems. MAPL allows iterative updates by comparing the current solution with decentralized or centralized solutions generated by various agents. We instantiate MAPL using MARL from human feedback (MARLHF) with a learned reward model and multi-agent direct preference optimization (MADPO). Experiments on collaborative writing, coding, tool use, and travel planning show that MAPL can improve collaboration quality and efficiency while approaching the performance of MARL with fixed, well-defined rewards. Within MAPL, MARLHF generally outperforms MADPO on most tasks but remains sensitive to data coverage, agent and comparator models, and the underlying MARL algorithms.
[MA-29] Adaptive and Resilient Dual-Layer Resource Slicing for Hovering Aerial Backhaul Networks
【速读】：该论文旨在解决在非平稳环境下，由悬停空中代理（HAA）辅助的回传网络中，异构5G/6G业务（包括增强移动宽带eMBB、超可靠低时延通信URLLC和海量机器类通信mMTC）所面临的双层资源切片架构中复杂的耦合问题。其核心挑战在于如何在动态变化的流量条件下实现资源分配的自适应性与韧性，同时保障关键任务型URLLC服务的严格时延约束。解决方案的关键在于提出一种基于强化学习的鲁棒自适应优先级编排增强型双延迟深度确定性策略梯度（RAPO-TD3）框架：首先引入新颖的双重软最大投影机制，将连续动作空间映射为物理上可行的带宽分配方案，确保严格满足系统约束；其次嵌入鲁棒自适应优先级编排（RAPO）机制，主动保护URLLC业务的时延性能；此外，论文建立了严格的数学理论基础，证明该框架具备Lipschitz连续性并满足Robbins-Monro收敛条件，从而保证算法的稳定渐近收敛性。仿真结果表明，相较于PPO、DDPG及传统求解器，RAPO-TD3在非平稳流量下展现出更优性能，尤其在500%需求激增场景下仍能将URLLC满足率维持在接近理论最优水平，且可扩展性评估显示其执行延迟低于1毫秒，完全满足URLLC的1毫秒时延预算，充分验证了该框架在性能与实时性上的有效性。
链接: https://arxiv.org/abs/2609.32798
作者: Chuan-Chi Lai,Jen-Hsiang Li
机构: National Chung Cheng University (国立中正大学); Advanced Institute of Manufacturing with High-tech Innovations (AIM-HI) (高端制造创新研究院); Feng Chia University (逢甲大学)
类目: Multiagent Systems (cs.MA); Networking and Internet Architecture (cs.NI)
备注: 16 pages, 6 figures, and 3 tables. Accepted for publication in IEEE Transactions on Cognitive Communications and Networking. ©2026 IEEE. Personal use of this material is permitted. However, permission to use this material for any other purposes must be obtained from the IEEE by sending a request to pubs-permissions@ieee.org
Abstract:This paper investigates adaptive and resilient dual-layer resource slicing in hovering aerial agent (HAA)-assisted backhaul networks for heterogeneous 5G/6G services, including enhanced mobile broadband (eMBB), ultra-reliable and low-latency communications (URLLC), and massive machine-type communications (mMTC). To address the complex coupling of this dual-layer architecture in non-stationary environments, we propose the resilient adaptive priority orchestration enhanced twin delayed deep deterministic policy gradient (RAPO-TD3) framework. We introduce a novel double soft-max projection mechanism to map the continuous action space into physically feasible bandwidth distributions, ensuring strict constraint adherence. Additionally, a resilient adaptive priority orchestration (RAPO) mechanism is embedded to safeguard mission-critical URLLC latency. Crucially, we establish a rigorous mathematical foundation proving that our framework ensures Lipschitz continuity and satisfies the Robbins-Monro conditions for stable asymptotic convergence. Extensive simulations under non-stationary traffic demonstrate that our RAPO-TD3 framework achieves superior performance relative to PPO, DDPG, and traditional solvers. Notably, via the RAPO mechanism, our approach maintains URLLC satisfaction levels closely approaching theoretical optima even during 500% demand surges. Furthermore, scalability evaluations indicate that sub-millisecond execution latencies strictly satisfy the 1 ms URLLC budget, demonstrating the performance efficacy of our proposed framework.
[MA-30] CollisionGAT: Controller-Agnostic One-Step Collision Screening for Multi-Agent Motion
【速读】：该论文旨在解决多机器人系统在运动规划过程中实时碰撞检测的计算效率与精度难题，尤其针对动态环境中多个移动代理（moving agents）与静态障碍物之间潜在碰撞风险的快速评估问题。传统方法依赖精确几何检查，虽准确但计算开销大，难以满足实时性要求。为此，本文提出CollisionGAT，一种基于图注意力网络（Graph Attention Network, GAT）的端到端学习框架，能够联合读取各移动代理的当前状态与提议动作状态，以及局部相关的静态障碍物信息，为每个代理输出一个碰撞风险评分。其核心创新在于利用图神经网络建模代理间及代理与环境间的局部拓扑关系，通过注意力机制自适应地聚焦于高风险区域，从而实现高效、可扩展的碰撞风险预测。该模型以精确几何检查结果作为训练标签，并独立验证每一步执行的安全性，确保了预测可靠性。最终，该方法可无缝集成至连续路径跟踪控制器或无感知障碍物的D* Lite规划器（GATeD），通过类型化否决（typed vetoes）机制动态更新规划图，显著提升复杂动态场景下的规划效率与安全性。
链接: https://arxiv.org/abs/2609.32783
作者: Alan Debbas,Edwin Meriaux,Gregory Dudek
机构: McGill University(麦吉尔大学)
类目: Robotics (cs.RO); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: 5 pages, 4 figures, 1 table, 2 algorithms. Accepted to the 2026 IEEE MIT Undergraduate Research Technology Conference (URTC)
Abstract:Before a team of robots moves, each proposed step must be checked for collisions with other robots and with obstacles. We present CollisionGAT, a graph-attention network that reads the current and proposed states of moving agents together with locally relevant stationary obstacles and returns one collision-risk score per moving agent. Any controller can use these scores to accept, repair, replan, or postpone a proposed step. We mount CollisionGAT on a continuous path-following controller and on GATeD, an obstacle-blind D* Lite planner that uses typed vetoes to update its planning graphs. Exact geometric checks supply the training labels and independently audit every executed step.
[MA-31] Self-Evolving Multi-Agent Symbolic Discovery for Financial Fundamental Analysis
【速读】：该论文旨在解决符号回归（Symbolic Regression, SR）在金融估值场景中应用受限的问题。传统符号回归在自然科学中表现优异，但金融估值面临多重挑战：存在多种合理的估值视角、市场环境具有非平稳性，且绩效信号高度噪声化、连续波动。针对这些问题，论文提出了一种分层多智能体框架——多智能体基本面分析与符号自适应学习（MUFASA），其核心解决方案包括：（1）通过专业化智能体分别代表不同估值视角，实现方程发现的解耦；（2）引入元协调器，基于市场上下文信息进行分层推理；（3）设计记忆机制，利用统计性能摘要（如准确率、稳定性及尾部风险）在噪声反馈下指导学习过程。实验表明，MUFASA在多个国家的数据集上均显著优于经典金融方法、金融大语言模型及现有符号回归方法，同时生成可解释的解析方程，并公开了演化过程中的提炼知识与上下文依赖的策略权重，为未来金融基本面分析研究提供重要参考。
链接: https://arxiv.org/abs/2609.32746
作者: Kelvin J.L. Koa,Filip Orestav,Shengqiong Wu,Michael J. Wooldridge,Ke-Wei Huang
机构: 未知
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Computational Finance (q-fin.CP)
备注:
Abstract:While symbolic regression (SR) has been successfully used in science to discover new equations, its use in financial valuation is hindered by several limitations. Whereas the natural sciences provide objectively correct relationships, financial valuation constitutes a distinct class of symbolic discovery problems, as it admits multiple valid perspectives, operates under non-stationary market conditions, and involves noisy, continuous performance signals. In this work, we propose Multi-Agent Fundamental Analysis with Symbolic Adaptive learning (MUFASA), a hierarchical multi-agent framework for symbolic discovery in finance. MUFASA introduces (1) disentangled equation discovery via specialized agents representing distinct valuation perspectives, (2) a meta-coordinator that performs hierarchical-level reasoning over market context information, and (3) a memory mechanism that reasons over statistical performance summaries (e.g., accuracy, stability, and tail risk) to guide learning under noisy feedback. Experiments across datasets from multiple countries show that MUFASA achieves state-of-the-art performance on the valuation task compared to classical finance methods, financial large language models, and SR approaches, while simultaneously producing interpretable equations, which we share with the community. We also make publicly available the distilled learnings across evolution iterations and context-dependent strategy weights, which might offer useful insights for future research on financial fundamental analysis.
[MA-32] When Better Gets Worse: Improvement Fidelity for Self-Improving Agents in Adaptive Worlds
【速读】：该论文旨在解决自改进智能体在使用代理验证器（proxy verifier）选择策略更新时，因部署后环境变化导致评估偏差的问题。尽管代理验证器在总体上能良好排序策略，但某些看似更优的更新在实际部署后可能表现反而更差，这种现象源于“改进保真度”（Improvement Fidelity）的缺失——即代理评估所揭示的改进方向与实际部署环境中真实改进的方向和顺序不一致。其核心解决方案在于提出PIVOT-KG，一种成对、决策感知的验证机制，通过根据单位成本预期降低选择后悔（selection regret）来分配有限的高保真评估资源。研究表明，全局策略准确率无法保证更新层面的保真度，操作者偏移（operator shift）与部署响应（deployment response）会引入更新级误差，而候选项的边际差异决定了这些误差是否影响最终替换决策。实验结果表明，在Leduc、Kuhn和Melting Pot共90个独立根节点中，代理最优集与部署最优集无交集的情况达51例；在HighwayEnv的八候选压力测试中，相较于精确均匀验证规则下0.0435的平均改进选择后悔，PIVOT-KG在主预算下将该值降至0.0055。这表明，可靠的自改进必须在策略更新所诱导的实际世界中评估其效果，并为稀缺的部署证据提供了可实践的分配准则。
链接: https://arxiv.org/abs/2609.32677
作者: Ke Wang,Zijie Zhao,Zhiyi Yuan,Changlun Li
机构: University of Cambridge (剑桥大学); Georgia Institute of Technology (佐治亚理工学院); Massachusetts Institute of Technology (麻省理工学院); University of Hong Kong (香港大学); The Hong Kong University of Science and Technology (广州) (香港科技大学(广州))
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: 15 pages, 10 figures, 3 tables
Abstract:Self-improving agents increasingly rely on proxy verifiers to choose policy updates, yet deployment can change the world in which those updates are evaluated. An update that looks better to the verifier can therefore become worse after deployment even when the verifier ranks policies well overall. We formalize this gap as Improvement Fidelity, which asks whether proxy improvements preserve the sign and ordering of deployment improvements over the updates an improvement process actually proposes. We show that global policy accuracy need not guarantee update fidelity: operator shift and deployment response can create update-level errors, while candidate margins determine whether those errors change the replacement decision. We introduce PIVOT-KG, a paired, decision-aware validator that allocates scarce high-fidelity evaluation according to the expected reduction in selection regret per unit cost. Across 90 held-out roots in Leduc, Kuhn, and Melting Pot, proxy and deployment optimal sets are disjoint in 51 cases. In an eight-candidate HighwayEnv stress test, PIVOT-KG reduces mean improvement-selection regret from 0.0435 under the exact Uniform validation rule to 0.0055 at the primary budget. Together, these results show why reliable self-improvement should evaluate proposed improvements in the worlds they induce, while providing a practical rule for allocating scarce deployment evidence when it can affect the replacement decision.
[MA-33] AsynCodeBench: Benchmarking Collaboration of Asynchronous Multi-Agent Systems in Software Engineering
【速读】：该论文旨在解决多智能体软件工程中协作能力评估缺失的问题，现有方法主要依赖于任务级结果指标，未能有效区分个体编码能力与跨智能体协同能力，导致对协作真实水平的衡量失真。其核心解决方案是提出AsynCodeBench——一个以依赖关系为中心的异步多智能体编程基准，通过显式构建每个任务的依赖图（dependency graph）并引入可执行的依赖检查器（Dependency Checker），实现对协作过程的精细化追踪。关键创新在于提出两个互补的度量指标：异步依赖通过率（Asynchronous Dependency Pass Rate, ADPR），用于衡量最终满足的跨智能体依赖比例；依赖解析步数（Dependency Resolution Step, DRS），用于定位每个依赖首次被满足的时间点。基于来自真实代码仓库的19个任务及52个有向依赖单元的实验表明，编码能力提升并不必然带来协作能力增强，且任务级指标与依赖级协作度量存在显著偏差。进一步的依赖轨迹分析揭示，有效协作常以“跳跃窗口”（hopping window）模式出现，即在极短执行阶段内集中解决大量依赖，而非渐进式推进。
链接: https://arxiv.org/abs/2609.32662
作者: Kaituo Zhang,Zhen Xiong,Zhimeng Jiang,Mingyu Zhong,Zhouyuan Yuan,Zhecheng Li,Bowen Lin,Chia-Yuan Chang,Mingzhi Hu,Huazheng Wang,Ying Lin
机构: University of Houston(休斯顿大学); New York University(纽约大学); Texas AM University(德克萨斯农工大学); UCSD(加州大学圣地亚哥分校); Worcester Polytechnic Institute(伍斯特理工学院); Oregon State University(俄勒冈州立大学)
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Multi-agent coding has emerged as an increasingly active direction in software engineering, where complex development tasks are decomposed across multiple specialized agents working on different parts of the problem. Despite the shift from individual problem solving to distributed collaboration, multi-agent systems still lack a direct measure of collaboration and are largely evaluated through task-level outcomes inherited from single-agent coding, conflating individual coding capability with cross-agent coordination. We introduce AsynCodeBench, a dependency-centric benchmark for asynchronous multi-agent software engineering that represents each task with an explicit dependency graph and executable Dependency Checkers. Through this dependency-tracking process, we propose two complementary measures: Asynchronous Dependency Pass Rate (ADPR), which measures how many cross-agent dependencies are ultimately satisfied, and Dependency Resolution Step (DRS), which measures when each dependency first becomes satisfied during execution. AsynCodeBench comprises 19 tasks from real-world repositories, exposing 52 directed dependencies as explicit units for evaluating cross-agent collaboration. Experiments across model families, scales, and generations reveal a clear gap between coding and collaboration capability: improvements in coding performance do not necessarily translate into stronger collaboration, and task-level metrics can diverge substantially from dependency-level collaboration measures. Dependency-trajectory analysis further reveals that successful coordination often emerges not gradually, but through concentrated bursts in which many dependencies become resolved over a short portion of the execution trajectory, a pattern we term a hopping window.
[MA-34] Enabling Timely Guidance before Skill Retrieval: Retaining Helpful Warm Tips in Agent Context
【速读】：该论文旨在解决大语言模型（Large Language Model, LLM）驱动的智能体在执行复杂任务时，因缺乏有效引导而容易陷入低效或错误路径的问题。现有技能机制通常仅暴露元数据，需在调用时才加载完整内容，导致关键指导信息在智能体做出决策前无法获取；而通用记忆方法则面临维护开销过大，或将无关或有害建议反复置于对话上下文中的风险。为此，本文提出TipsWarm机制，通过维护一个预算受限的、由技能衍生的关键点（warm tips）池，将可转移的技能指导以选择性注入的方式嵌入每轮消息上下文中。该机制通过分离事件触发的LLM评估与轻量级的每轮筛选，实现了高效且低成本的指导信息供给，在三个编码与迭代任务执行基准测试中，不仅取得了最高的任务成功率，同时保持了良好的时间效率，显著优于近期的技能机制与通用记忆基线。
链接: https://arxiv.org/abs/2609.32339
作者: Feng Liang,Yupeng Li,Runhao Zeng,Francis C. M. Lau,Xiping Hu
机构: Shenzhen MSU-BIT University(深圳莫斯科大学-北京理工大学联合大学); Hong Kong Baptist University(香港浸会大学); The University of Hong Kong(香港大学); Beijing Institute of Technology(北京理工大学)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 18 pages and 8 figures
Abstract:Reusable skills help LLM-based agents solve complex tasks, but the agent must receive guidance before it commits to an ineffective approach. Existing skill mechanisms often expose only metadata and load full content on demand, leaving useful guidance unavailable until the agent decides to retrieve it. General memory methods can incur substantial maintenance overhead, while keeping guidance in conversation context risks repeatedly exposing the agent to irrelevant or harmful advice. We propose TipsWarm, a mechanism that complements existing skill mechanisms by maintaining a budgeted pool of skill-derived keypoints, or \textitwarm tips, for selective injection into the context of every message turn. By separating event-triggered LLM assessment from inexpensive per-turn screening, it makes transferable skill guidance readily available while controlling maintenance costs. In three coding and iterative task-execution benchmarks, TipsWarm achieves the highest task success rate while remaining time-efficient, compared to recent skill and general memory baselines.
[MA-35] Agents ensus: Consensus-Compressed Shared Memory for Multi-Agent Story Worlds
【速读】：该论文旨在解决生成式故事世界（agentic story world）中因每个角色拥有独立记忆流而导致的事件记录重复存储问题，从而引发巨大的存储开销。其核心解决方案是提出Agentsensus框架，该框架引入统一的长期记忆机制，使同一事件的记录在不同目击者之间合并为单一共享条目，并对语义相关的记忆条目进行链接。实验在四个故事世界（包括两部中国古典小说、《哈姆雷特》及真实冲突时间线）上进行，对比三种基于个体角色的记忆设计，在等粒度协议下运行40至80轮。结果表明，Agentsensus相比最接近的基线减少了22%-44%的写入条目数，且实现了显著的内存共享（14%-28%的记录被多个角色持有，部分达10个以上）与高度语义链接（94%-99%），同时模拟质量不降反升。消融实验进一步证明，记忆合并机制是关键：禁用该机制后，存储量增加3.1倍，共享率降至零。此外，共享效应随模拟周期延长而持续增强，而非早期饱和，表明其具备良好的可扩展性。
链接: https://arxiv.org/abs/2609.32297
作者: Yu Pan
机构: University of Nebraska–Lincoln(内布拉斯加大学林肯分校)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multiagent Systems (cs.MA)
备注: 32 pages, 19 figures, 9 tables. Code: this https URL
Abstract:A agentic story world is a dynamic system simulating who learned what, when, and from whom – yet the standard design gives each character a private memory stream. A shared event is therefore stored once per witness, large duplication will be incurred in terms of storage. We present Agentsensus, a story-world simulation framework in which there is an unified long-term memory. Records of the same event merge into one owned by all its witnesses, and semantically relevant memory records are linked. We evaluate on four worlds – two classical Chinese novels, Hamlet, and a real-world conflict timeline – run for 40 to 80 rounds against three per-character memory designs under an equal-granularity protocol. Agentsensus writes 22-44% fewer entries than the closest baseline and is the only design whose memory becomes shared (14-28% of records held by more than one character, some by 10) and linked (94-99%), at judged simulation quality indistinguishable or even better than the baselines. An ablation attributes this to the merge itself: disabling it multiplies the store by 3.1x and takes sharing to exactly zero. Sharing also compounds with the horizon rather than saturating early, rising 6% to 9% to 14% as one world is re-run at 10, 20 and 40 rounds.
[MA-36] GLIDE: Generalized Layer-wise Intrinsic Distributional Evaluation for Heterogeneous LLM Agents EMNLP2026
【速读】：该论文旨在解决大语言模型（LLM）智能体在复杂任务中进行可靠步骤级评估的难题，以实现候选路径的有效比较与计算资源的合理分配。现有方法面临两大挑战：外部验证器引入额外推理开销，而由智能体自身生成的置信度或自评估分数易出现校准偏差，尤其在异构智能体生成候选解时更为显著。为此，论文提出一种通用分层内在分布评估框架（Generalized Layer-wise Intrinsic Distributional Evaluation, GLIDE），其核心创新在于通过分层残差一致性（layer-wise residual coherence）挖掘内在步骤证据，量化局部残差更新是否一致支持由候选步骤引发的全局残差变化。该证据进一步基于生成智能体近期得分分布进行校准，并转化为融合绝对残差证据与智能体相对表现的悲观奖励（pessimistic reward），从而为蒙特卡洛树搜索（MCTS）提供跨智能体可比的价值信号；同时，归一化的预测不确定性用于指导自适应分支策略。实验结果表明，GLIDE在多跳推理、序列决策与符号逻辑任务中均显著提升任务性能、步骤级排序质量及计算效率，且无需外部验证器或任务特定监督。
链接: https://arxiv.org/abs/2609.32295
作者: Wei Zhu,Yiming Wang,Rui Wang,Lixing Yu,Kun Yue,Zhiwen Tang
机构: Yunnan University (云南大学); Yunnan Key Laboratory of Intelligent Systems and Computing (云南省智能系统与计算重点实验室); Shanghai Jiao Tong University (上海交通大学)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: Accepted by EMNLP 2026 Findings
Abstract:LLM agents require reliable step-level evaluation to compare candidate branches and allocate computation effectively. However, lightweight evaluation remains challenging. External verifiers introduce additional inference cost, while agent-produced confidence or self-evaluation scores can be miscalibrated, especially when candidates are generated by heterogeneous agents. We propose \textbfGeneralized \textbfLayer-wise \textbfIntrinsic \textbfDistributional \textbfEvaluation (\textbfGLIDE) for LLM agents. \textscGLIDE derives intrinsic step evidence from layer-wise residual coherence, which measures whether local residual updates consistently support the global residual change induced by a candidate step. It calibrates this evidence against the recent score distribution of the generating agent and converts it into a pessimistic reward that jointly accounts for absolute residual evidence and agent-relative standing. The reward provides a cross-agent value signal for MCTS branch selection, while normalized predictive uncertainty guides adaptive branching. Experiments on multi-hop reasoning, sequential decision making, and symbolic logic show that \textscGLIDE improves task performance, step-level ranking quality, and computational efficiency without external verifiers or task-specific supervision.
[MA-37] Prefill-Free Cross-Family KV Cache Transfer for Heterogeneous Multi-Agent LLM s
【速读】：该论文旨在解决多智能体大语言模型（multi-agent LLM）系统中因异构模型间文本通信导致的冗余预填充（prefill）问题。在现有框架下，接收方需重新处理发送方已计算的共享上下文，造成显著计算开销。为实现无需预填充的跨模型家族键值（KV）缓存复用，其核心挑战在于不同模型在分词方式、网络深度及KV表示上的差异。为此，论文提出HeteroFold方法，通过结构对齐、将发送方缓存映射至接收方空间并进行行为校准，在保持双方模型冻结的前提下，实现高效且准确的跨模型缓存迁移。实验表明，HeteroFold在六个转移方向上均优于现有基准，在长上下文与多数短上下文场景中表现最佳，并在多智能体任务中达到与传统文本通信相当的性能；在32K上下文长度下，相较于原生预填充，速度提升达10.7倍，较当前最优无预填充基线（Dense Latent与KV Ridge）仍快1.18–1.47倍，验证了其在跨模型家族高效缓存复用方面的有效性。
链接: https://arxiv.org/abs/2609.32259
作者: Vincent-Daniel Yun,Woosang Lim,Haneul Yoo,Sungjoo Yoo,Sai Praneeth Karimireddy,Murali Annavaram
机构: University of Southern California(南加州大学); Seoul National University(首尔国立大学); New York University(纽约大学)
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注:
Abstract:Recent multi-agent LLM systems increasingly combine heterogeneous models for specialized agent roles. However, text-based communication requires each receiver to prefill shared context already processed by the sender. Reusing the sender’s key-value (KV) cache avoids this redundancy, but prefill-free transfer across model families must handle differences in tokenization, model depth, and KV representations. To address these issues, we propose \textitHeteroFold, a prefill-free cross-family KV cache transfer method that keeps both the sender and receiver frozen. HeteroFold aligns model structures, maps the sender cache into the receiver space, and calibrates it to preserve receiver behavior. Across six transfer directions, HeteroFold achieves the best cache-transfer performance on all four long-context benchmarks and most short-context settings. It also matches text-based communication on the multi-agent benchmark. At 32K context length, Llama-3.1-8B \rightarrow Ministral-3-14B transfer is 10.7\times faster than Native Prefill and 1.18 – 1.47\times faster than the state-of-the-art prefill-free baselines, Dense Latent and KV Ridge. These results show that HeteroFold enables efficient cross-family KV reuse without receiver prefill.
[MA-38] Robust Game-theoretic Motion Planning over Extended Time Horizons
【速读】：该论文旨在解决长期时域下受扰动影响的非凸博弈型运动规划问题，其核心挑战在于多智能体系统中动态耦合与不确定性共存时的实时、鲁棒协同决策。解决方案的关键在于提出一种名为WOLF（Weighted Open-loop Loop-Free）的算法，将开环微分博弈求解器与滚动时域模型预测控制（MPC）相结合，基于序列凸化技术实现高效求解。不同于传统鲁棒优化中离线固定不确定性的方法，本文将鲁棒性管（robustness tube）作为动态状态变量，与轨迹联合优化，其厚度直接决定共享耦合约束的紧致程度。通过推导充分条件，确保在所有可接受扰动实现下，满足各智能体误差边界收紧后的名义轨迹仍保持可行性。该方法在两个具有耦合平动-姿态动力学的对抗性轨道博弈场景中得到验证：一是存在主动探测概率约束下的隐蔽共轨干扰博弈，二是对手通过遮挡太阳光以削弱被追捕者太阳能供电的“遮阳”博弈。
链接: https://arxiv.org/abs/2609.32098
作者: Bennet Outland,Vishala Arya
机构: University of Colorado Boulder (科罗拉多大学博尔德分校)
类目: Multiagent Systems (cs.MA); Systems and Control (eess.SY)
备注:
Abstract:This work presents a solution to nonconvex, game-theoretic motion planning problems subject to disturbances over long time horizons. The problem is posed as a partially-decoupled generalized Nash equilibrium problem, in which each agent’s dynamics depend only on its own state and control, admitting fast solution methods for competitive multi-agent motion planning. An algorithm, WOLF, is developed that applies receding-horizon model predictive control to an open-loop differential games solver based on sequential convexification. In contrast to robust formulations that fix the uncertainty description offline, the robustness tube here is itself a dynamic state, co-optimized with the trajectory, and its thickness directly sets the tightening of the shared coupling constraints. A sufficient condition is derived under which a nominal trajectory satisfying constraints tightened against all agents’ error bounds remains feasible for every admissible disturbance realization. The method is demonstrated on two adversarial on-orbit games with coupled translational-attitude dynamics: a stealthy co-orbital jamming game under an active detection-probability bound, and a sun-blocking game in which an adversary disables an evader by decreasing solar power
[MA-39] On Evaluating and Improving Conversational Agents in Production
【速读】：该论文旨在解决大规模多智能体购物助手在生产环境中进行离线评估时面临的三大核心挑战：（i）日志对话无法在系统修改后重现，因响应变化会引发后续所有交互的连锁变动；（ii）未修改系统的运行结果本身具有随机性，其大语言模型（LLM）组件的随机性及商品搜索中产品信息、价格与用户个性化信号的动态变化导致结果不可复现；（iii）聚合质量评分掩盖了不同行为模式的影响，难以定位具体性能变化的来源。针对上述问题，该研究提出了一套系统性框架，其关键在于通过“评估工具包”（Evaluation Harness）构建固定客户场景集，并基于有根基的用户模拟（grounded user simulation）在本地实例中精确复现特定行为，以生成可比较的测试轨迹；同时，通过重复运行未修改系统建立稳定基线，利用配对百分位数Bootstrap区间对每个独立改进方案进行统计对比；最终由“改进编排器”（Improvement Orchestrator）将评估结果转化为可验证假设并实现隔离式变更验证。该框架不仅实现了对性能波动根源的精准定位，还具备自我审计能力，可识别评估过程中的缺陷（如评分员证据不足或配置未生效），从而提升评估的可靠性与可迭代性。
链接: https://arxiv.org/abs/2609.32092
作者: Kasra Hosseini,Wen-Sen Cheng,Marco-Andrea Buchmann,Emir Mulabegovic,Weiwei Cheng
机构: Zalando SE(扎兰多股份公司), Berlin, Germany(柏林, 德国); Zalando(扎兰多), Switzerland(瑞士)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Emerging Technologies (cs.ET); Information Retrieval (cs.IR)
备注: 21 pages, 2 figures, 1 table
Abstract:We present a framework for evaluating and improving a large-scale, multi-agent shopping assistant in production, and report lessons from its use. Offline evaluation of such a system faces three obstacles. (i) A logged conversation cannot be replayed against a modified system, because a different response changes every turn that follows. (ii) The unchanged system itself varies from run to run. Its LLM components are stochastic, and in product search the available products, their prices, and the customer’s personalization signals change. (iii) Aggregate quality scores combine distinct behaviors, so they show that quality has changed but not which behavior caused the change. Our framework addresses each obstacle in turn. For a reported behavior, an Evaluation Harness generates targeted assertions and a fixed cohort of customer scenarios. It then reproduces the behavior in a local instance of the assistant through grounded user simulation. Instead of replaying the log, the simulator writes new customer turns conditioned on the recorded messages and context. Repeated runs of the unchanged system form a stored baseline. An Improvement Orchestrator turns the assertion results into hypotheses, implements each as an isolated modification, and compares it with the baseline using paired percentile bootstrap intervals over scenario-level differences. When an investigation ends, the harness may propose revisions to future evaluations, subject to human approval and without altering past decisions. We report production investigations with this framework. Assertion profiles showed which positions of a product carousel a failure affected, and repeated runs distinguished a real improvement from run-to-run fluctuation. Audits of the evaluation itself found a judge that lacked the evidence it needed and a model setting that was configured but not applied.
[MA-40] EngramRAG : Dynamic Usage-Weighted Topology and Synaptic Consolidation for Multi-Hop Agent ic Memory
【速读】：该论文旨在解决自主大语言模型（LLM）代理在多会话环境中部署时，传统记忆架构面临的三大核心问题：关联盲视（Associative Blindness，无法处理多跳关系依赖）、支架遗忘（Scaffolding Amnesia，时间衰减导致核心人格不变量丢失）以及静态拓扑停滞（Static Topology Stagnation，固定图结构忽视使用动态）。其解决方案的关键在于基于互补学习系统（Complementary Learning Systems, CLS）原理，提出一种自适应记忆架构EngramRAG，通过双态协同机制实现长期记忆的高效维持与动态演化。核心创新包括：(1) 使用模态个性化PageRank（U-PPR），通过赫布塑性动态调整转移概率，将关键实体固化为高中心性的认知宏观枢纽；(2) 采用巩固激活的拓扑衰减机制（CATD），以拓扑承载权重而非时间流逝决定保留半衰期，并引入冷启动保护期（N_grace = 4）；(3) 引入有向SUPERSEDES DAG过滤器，在事实更新时抑制过时状态；(4) 设计三源混合检索框架，融合密集向量、BM25与U-PPR，通过动态互逆排名融合（RRF）实现最优检索结果整合。实验表明，EngramRAG在LoCoMo基准的1,982个问答对上相较密集向量RAG提升Recall@5达38.9%（53.21% vs. 38.29%，p < 0.001），MRR提升43.1%（0.4203 vs. 0.2937），显著优于Okapi BM25和静态图检索；在时间推理任务中达到62.33% Recall@5，较密集向量提升16.67个百分点；在控制突变测试中，SUPERSEDES将分裂式幻觉率从70.0%降至0.0%，90天模拟下支架保留率达100%，交互式检索响应延迟仅为26.21ms。
链接: https://arxiv.org/abs/2609.32049
作者: Bhavyateja Potineni,Lohit Giri,Anu Jain,Vadim Kutsyy,Rajasekhar Pentakota
机构: Independent Researchers(独立研究员)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR); Multiagent Systems (cs.MA)
备注: 8 pages, 6 figures, 4 tables. Code and reproduction suite: this https URL
Abstract:As autonomous LLM agents are deployed across multi-session environments, conventional memory architectures suffer from Associative Blindness (inability to traverse multi-hop relational dependencies), Scaffolding Amnesia (temporal decay evicting core persona invariants), and Static Topology Stagnation (immutable graphs ignoring usage dynamics). Grounded in Complementary Learning Systems (CLS) principles, we propose EngramRAG, an adaptive memory architecture coupling a low-latency Waking State reflex with an asynchronous background Dreaming State consolidation cycle. EngramRAG introduces: (1) Usage-Modulated Personalized PageRank (U-PPR), where transition probabilities adapt via Hebbian plasticity to promote persistent entities into high-centrality Epistemic Macro-Hubs; (2) Consolidation-Activated Topology Decay (CATD), which scales retention half-life by topological load-bearing weight rather than wall-clock recency, protected by a cold-start grace period (N_grace = 4); (3) Directed SUPERSEDES DAG filtering to suppress obsolete state during fact mutations; and (4) Triple-source hybrid retrieval fusing dense vectors, BM25, and U-PPR via dynamic Reciprocal Rank Fusion (RRF). Evaluating on all 1,982 QA pairs across 10 long-term conversations in the LoCoMo benchmark, EngramRAG achieves +38.9% relative improvement in Recall@5 (53.21% vs. 38.29%, p 0.001) and +43.1% in MRR (0.4203 vs. 0.2937) over dense vector RAG, significantly outperforming Okapi BM25 (48.66%) and isolated static graph retrieval (8.50%). On temporal reasoning, EngramRAG reaches 62.33% Recall@5 (+16.67 points over dense vectors). In controlled mutation tests, SUPERSEDES suppresses split-brain hallucinations from 70.0% to 0.0%, while 90-day simulations show 100.0% scaffolding retention under a 26.21ms interactive retrieval reflex.
[MA-41] Decentralized Master-Mind: Joint Action Refinement through Iterative Intent Denoising in Multi-Agent Pathfinding
【速读】：该论文旨在解决去中心化多智能体路径规划（Decentralized Multi-Agent Path Finding, MAPF）中因局部可观测性与通信限制导致的协同动作不兼容问题。在部分可观测环境下，各智能体需基于自身观测独立决策以达成目标且避免碰撞，而现有基于专家数据训练的可学习策略在独立采样时，可能将多个局部有效的单智能体动作组合成全局冲突的联合动作，导致规划失败。其核心解决方案在于提出DMM（Decentralized Master-Mind）框架，通过引入类扩散模型去噪思想的回合级意图迭代精炼机制，替代传统的单次动作采样方式：智能体初始随机生成动作意图，并在多轮通信中通过本地信息交互逐步优化意图，实现动作选择前的耦合协调，从而提升联合动作的相容性。DMM首先通过模仿学习在专家解上预训练，再利用无评分类的组相对强化学习方法MICPO进行微调，进一步优化多轮动作精炼过程。实验表明，该方法在1,600个MovingAI任务中达到1,598次成功求解，覆盖率达所有对比方法最高，且解的质量接近最强基线；同时可扩展至百万级智能体在障碍物密集环境中的协同运行，验证了其在保持去中心化执行的同时，通过回合级意图精炼显著提升了联合动作协调能力。
链接: https://arxiv.org/abs/2609.32019
作者: Valeriy Vyaltsev,Anton Andreychuk,Taisia Zlotnikova,Konstantin Yakovlev,Aleksandr Panov,Alexey Skrynnik
机构: Konstantin Yakovlev; Aleksandr Panov; Alexey Skrynnik
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Decentralized multi-agent path finding (MAPF) with communication requires agents to reach individual goals without collisions under partial observability. Learnable policies trained on expert data provide an effective approach to this problem. However, when several coordinated joint actions are valid in the same context, independently sampling from per-agent distributions can recombine locally valid choices into incompatible joint actions. This failure can arise from the final sampling mechanism even when the per-agent action distributions are learned correctly. DMM (Decentralized Master-Mind) addresses this by replacing one-shot action sampling with discrete, iterative refinement of action intents across communication rounds, inspired by denoising in diffusion models. Agents initialize random action intents and refine them through local communication, coupling their choices before commitment. DMM is pretrained with imitation learning on expert MAPF solutions and further optimized with MICPO, a critic-free group-relative reinforcement-learning method designed for multi-agent, multi-round action refinement. DMM generally achieves higher success rates and lower solution costs than the evaluated learnable baselines. On 1,600 MovingAI tasks, DMM fine-tuned with MICPO solves 1,598, the highest coverage among the evaluated methods, while achieving solution costs close to those of the strongest baselines. DMM also scales to over one million simultaneously acting agents in obstacle-rich environments. These results show that round-level intent refinement can improve joint-action coordination while preserving decentralized execution.
[MA-42] Communication between Frozen Large Language Models via Prompt Optimization in a Referential Game
【速读】：该论文旨在解决跨不同提供商、采用不同分词器的两个冻结大语言模型（Large Language Models, LLMs）之间在不更新模型权重的前提下，通过API端点实现可靠通信的问题。核心挑战在于：双方需在有限的字符集和固定长度消息约束下完成指代游戏（referential game），即发送方描述对象，接收方从候选集中准确识别目标。传统方法因缺乏共享语义编码机制而难以建立稳定通信协议。其解决方案的关键在于引入一个独立的提示优化器（prompt optimizer），该优化器基于反射模型（reflection model）对代理间的交互评分进行分析，并动态重写提示。在位置设置中，优化后的提示能够生成共享的代码体系，使系统在未见对象上表现优于无代码本基线，甚至在移除记忆窗口后仍保持有效性；而在字母块独立设置下，虽单个字符无法承载信息，但整体对象的“位值编码”（place value code）仍可实现通信。成功的关键因素包括：发送方冲突惩罚机制、对成功交互的保留策略以及序列化优化过程。在部分实验运行中，该机制成功构建了可读且可审计的通信协议，其规则直接嵌入优化后的提示中，体现了协议学习的透明性与可解释性。
链接: https://arxiv.org/abs/2609.31989
作者: Vivek Anand,Muthu Chandrasekaran,Shiva Chaitanya
机构: Georgia Institute of Technology (佐治亚理工学院); Interactly.ai
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: 42 pages (26 main text), 10 figures, 17 tables
Abstract:We study communication between two frozen large language models from different providers, with different tokenizers, accessed through their API endpoints. The two play a referential game: one sees an object and describes it in a short fixed-length message over a small alphabet; the other must pick that object out of a candidate set. Neither model’s weights are updated. Each agent’s prompt is rewritten by an isolated prompt optimizer whose reflection model reads that agent’s scored interactions. In the positional setting, optimized prompts carry a shared code that generalizes to held-out objects above a measured no-codebook baseline, including when the memory window is removed. In a second setting, independent per-letter blocks no longer fit within the message, although a whole-object place value code does. The base system fails to establish reliable communication: the sender struggles to retain an injective rule, and the receiver has too few confirmed examples in view. A sender collision penalty, retention of successful interactions, and sequential optimization enable successful place value communication in some runs. Outcomes vary across runs and reflection models. In successful runs, the protocol is written into the optimized prompts, where it can be read and audited directly.
[MA-43] Overview of the TREC 2025 Million Large Language Models track
【速读】：该论文旨在解决在复杂任务协作的智能体生态系统中，如何高效、动态地识别与选择具备相应专长的生成式 AI 模型（Generative AI）这一核心问题。传统方法依赖预定义的能力描述或静态元数据，难以覆盖现实中海量且多样化的语言模型（LLM）的真实能力分布。其解决方案的关键在于提出一种基于检索的范式：通过分析模型在实际任务中的可观测行为（如输出结果与日志概率），由辅助智能体动态推断其真实专长，从而实现无需人工标注的自适应专家选择。该方法在 TREC 百万级 LLM 评测赛道中得以实践，将检索目标从文档迁移至专家模型，要求系统从上千个模型的响应数据中学习并构建可泛化的专长表征，最终对未见查询进行精准的模型性能排序，为代理式 AI 中的专家检索提供了首个大规模基准。
链接: https://arxiv.org/abs/2609.31921
作者: Evangelos Kanoulas,Panagiotis Eustratiadis,Jamie Callan,Mark Sanderson,Yongkang Li,Jingfen Qiao,Gabrielle Poerwawinata,Vaishali Pal
机构: University of Amsterdam (阿姆斯特丹大学); Carnegie Mellon University (卡内基梅隆大学); RMIT (皇家墨尔本理工大学)
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: NIST TREC 2025 Proceedings
Abstract:Agentic AI envisions ecosystems of intelligent agents collaboratively solving complex tasks with minimal human intervention. In such ecosystems, each agent possesses specialized expertise, making effective expert selection central to overall system performance. While most current approaches assume a small number of well-documented models, real-world expertise is far more diverse and cannot be adequately captured through static metadata or hand-written descriptions. We anticipate a future with millions of specialized language models (LLMs), each excelling in different domains or problem types. Rather than relying on predefined capability statements, we propose a retrieval-based paradigm in which an assistant agent infers expertise dynamically by examining models’ observable behavior. Upon receiving a user query, the assistant ranks candidate LLMs based on demonstrated competence, enabling efficient and adaptive expert selection. The TREC Million LLM Track operationalizes this paradigm by shifting the retrieval target from documents to expert LLMs. Participants are given a discovery set consisting of queries, answers, and log-probabilities from more than one thousand LLMs and are challenged to infer meaningful expertise representations for each model. Given an unseen test query, systems must then rank the LLMs according to their expected performance, providing the first large-scale benchmark for expertise retrieval in agentic AI.
[MA-44] Choir: An Open Protocol for Distributed Multi-Agent Autoformalization
【速读】：该论文旨在解决当前生成式AI在形式化数学证明过程中存在的集中化问题，即依赖单一团队运行所有AI代理并承担全部计算成本，导致可扩展性与开放性受限。其解决方案的关键在于提出Choir——一个用于分布式形式化的开源协议，通过将项目分解为可由独立贡献者完成的任务，使每位参与者可使用自有大语言模型（LLM）订阅运行自己的AI代理，所有协作仅通过GitHub仓库进行。为保障质量与一致性，所有贡献在合并前均经过确定性门控（deterministic gate）验证，确保流程透明可信。Choir支持Lean 4、Isabelle和Rocq等多种形式化系统，具备模块化设计，允许项目灵活替换或扩展组件，从而实现真正开放、可扩展且去中心化的形式化知识构建生态。
链接: https://arxiv.org/abs/2609.31903
作者: Yidi Qi,Melanie Weber
机构: Harvard University(哈佛大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Logic in Computer Science (cs.LO); Multiagent Systems (cs.MA)
备注:
Abstract:AI agents can now formalize entire textbooks and major theorems in proof assistants such as Lean, but current efforts are typically centralized: a single team runs all agents and bears the full computational cost. We introduce Choir, an open protocol for distributed formalization. Choir decomposes a project into tasks that can be completed by independent contributors, each running their own agent with their own LLM subscription, while coordinating entirely through the project’s GitHub repository. To support open participation, every contribution is checked by a deterministic gate before merge. Choir supports Lean 4, Isabelle, and Rocq, and is open source and modular, allowing projects to replace individual components or extend the protocol.
[MA-45] Deep Reinforcement Learning for Equity Trading: Benchmarking Actor-Critic Methods with Forward Retraining
【速读】：该论文旨在解决股票市场中实现持续盈利交易的难题，核心挑战在于市场具有噪声大、非平稳性以及仅部分可预测等特性。为应对这一问题，研究提出采用端到端的深度强化学习（Deep Reinforcement Learning, DRL）方法，直接从市场状态中学习交易策略，而非依赖传统的价格预测模型。其解决方案的关键在于对比五种主流DRL算法（A2C、PPO、DDPG、TD3、SAC）在真实市场数据上的表现，并与基于监督学习的价格预测基线进行比较。实验使用2000–2020年20只大型资本化标普500成分股的日度数据，结合趋势跟踪技术指标和对数极值归一化处理，通过2000–2018年训练、2019–2020年回测的方式评估模型性能。关键发现表明，尽管DDPG在收益和夏普比率上表现最优（年化收益率55.5%，夏普比率1.38），但其风险暴露也最高（市场贝塔1.24）；而TD3与SAC在风险-收益平衡方面更优，最大回撤约25%且具备更强的超参数鲁棒性。此外，前瞻性重训练（forward retraining）机制显著提升A2C、PPO和SAC的稳定性，却使DDPG性能大幅下降，凸显了不同算法在动态环境中的适应性差异。相比之下，基于价格预测的基线策略虽收益较低，但最大回撤仅为9.6%、贝塔仅0.31，揭示了端到端DRL策略在高回报与高风险之间的权衡关系。
链接: https://arxiv.org/abs/2609.31870
作者: Bicheng Wang,Xinyi Zhang
机构: Stanford University(斯坦福大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Multiagent Systems (cs.MA)
备注:
Abstract:Consistently profitable trading is difficult because equity markets are noisy, non-stationary, and only partially predictable from historical data. We benchmark five deep reinforcement learning (DRL) actor-critic methods: A2C, PPO, DDPG, TD3, and SAC, that learn trading actions end-to-end from market states, and compare them with a supervised price-forecasting baseline. Using daily data for 20 large-capitalization SP 500 stocks from 2000 to 2020, enriched with trend-following technical indicators and log min-max scaling, we train on 2000-2018 and backtest on 2019-2020. Each agent is evaluated both when trained once and under forward retraining, in which it is retrained on all data available before each successive test window. DDPG achieves the highest annual return (55.5%), Sharpe ratio (1.38), and alpha (0.22), but also the highest market beta (1.24). TD3 and SAC offer a better risk-return balance, with Sharpe ratios of 1.37 and 1.33 and maximum drawdowns of about 25%. Forward retraining improves A2C, PPO, and SAC, leaves TD3 essentially unchanged, and reduces DDPG’s annual return from 55.5% to 29.8%, consistent with TD3’s greater robustness to hyperparameters. The forecasting baseline has the smallest maximum drawdown (9.6%) and the lowest beta (0.31), underscoring a trade-off between the higher returns of end-to-end DRL and the lower risk of forecast-driven strategies.
[MA-46] Be Careful Who You Trust: Coordination Dynamics under Corrupted Communication in LLM Multi-Agent Games EMNLP2026
【速读】：该论文旨在解决在公共通信不可靠的条件下，大型语言模型（Large Language Models, LLMs）作为多智能体系统进行协作时的鲁棒性问题。其核心挑战在于：当通信渠道受到干扰或被篡改时，智能体之间能否维持有效的协调与合作。研究通过在受控环境下对同质化LLM群体进行多轮N人猎鹿博弈（N-player Stag Hunt），并引入程序化动作反转（programmatic action inversion）来操纵公开对话记录与实际执行动作之间的不一致，系统考察了不同群体规模、协调阈值、污染水平及七种不同LLM下的协作表现。研究发现的关键机制包括：（1）尽管诚实选择的比例随污染程度上升而下降，但公开成功率的急剧下滑主要源于机制性失真而非策略性失败；在典型设置N=5、M=3下，原始成功率仍维持在80%污染水平下的78%，而公开成功率骤降至12%；（2）诚实决策显著依赖于决策时刻可获取的公开历史信息，尤其在高污染情境中更为明显；（3）三种基于阈值的公开报告基准模型与LLM代理在动作匹配率上表现出高度一致性，表明LLM的行为具有可解释的规律性。因此，该研究提出，评估多智能体系统的鲁棒性时，必须将原始意图、公开行为与实际执行结果三者明确区分，因为通信污染会以可预测且严重的方式破坏协同合作。
链接: https://arxiv.org/abs/2609.31704
作者: Xuanyi Liu,Niall Dalton,Hairi Amin,Xiyuan Yin,Lydia Lim
机构: University College London(伦敦大学学院)
类目: Multiagent Systems (cs.MA); Machine Learning (cs.LG)
备注: To be published in REALM: The 2nd Workshop for Research on Agent Language Models at EMNLP 2026
Abstract:Large language models are increasingly used as interacting agents, but it remains unclear how robust their coordination is when public communication is unreliable. We study this question in iterated N -player Stag Hunt games played by homogeneous LLM groups under controlled programmatic action inversion, which changes both the public transcript and the actions used for execution. Across an experimental grid spanning group sizes, coordination thresholds, corruption levels, and seven LLMs, we observe three main patterns. First, honest agents’ pre-flip Stag choices decline as corruption increases, but the sharp fall in public success is primarily mechanical. In the focal N=5,M=3 setting, pre-flip success remains 78% at 80% corruption, while public success falls to 12%. Second, honest choices are associated with the public history available at decision time, particularly under high corruption. Third, three threshold-style public-report benchmarks yield similar action-match rates to the LLM agents, showing substantial descriptive agreement between LLM decisions and these benchmarks. Overall, our results show that original choices, public actions, and executed outcomes must be separated when evaluating multi-agent robustness, as corrupted communication can severely and predictably degrade mutually beneficial cooperation.
[MA-47] Autonomous Research Project Management as an Agent Skill: A Case Study in Exact Spectral Spatial Regression
【速读】：该论文旨在解决在消费级硬件上实现端到端自主机器学习科研流程的可行性问题，尤其聚焦于如何在资源受限环境下完成复杂科学计算任务。其核心挑战在于构建一个具备自我纠错与持续演进能力的智能代理系统，以实现从文献检索、假设生成、实验执行到结果评估的全闭环科研自动化。解决方案的关键在于：1）基于快速傅里叶变换（Fast Fourier Transform, FFT）优化的核岭回归（Kernel Ridge Regression, KRR）求解器，显著提升对36×72空间网格上2005年每月NOAA Kaplan SST v2异常数据的计算效率；2）通过文件驱动的史诗级（epic）与任务级（issue）跟踪架构，实现长时序状态的解耦与持久化管理；3）利用深度求索V4 Flash（DeepSeek V4 Flash）作为核心代理，结合自研的深度求索套件（DeepSeek Harness, DSH），在仅需四次人工干预的情况下，成功处理两次假设评审失败并修复启动索引缺陷，展现了闭环科学韧性。此外，研究强调自主科研治理应依赖可审查的状态记录、可证伪的评审机制以及对负面结果的透明报告，呼吁机器学习社区优先采用结构化、可被代理访问的数据格式，而非静态的PDF论文形式。
链接: https://arxiv.org/abs/2609.31683
作者: Alexander Chen(1),Jeffrey Meng(1),Bram Hoex(1 and 2),Tong Xie(1 and 2) ((1) University of New South Wales, (2) GreenDynamics)
机构: University of New South Wales (新南威尔士大学); GreenDynamics
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 13 pages, 2 figures, submitted to Autonomous Machine Learning Research (AutoMLR) 2026 workshop
Abstract:This work presents an end-to-end demonstration of autonomous machine learning research conducted by an agent skill on consumer hardware. The demonstration evaluates an FFT-based Kernel Ridge Regression (KRR) solver for regular spatial grids using 2005 monthly NOAA Kaplan SST v2 anomaly fields on a 36 \times 72 grid. This was autonomously executed by DeepSeek V4 Flash, orchestrated by our agent skill suite within DeepSeek Harness (DSH). Experiments were executed on CPU-only hardware (Apple M2 Pro; 78.7 s solver time, 1.57 GB peak RSS). Long-horizon state was decoupled into a file-based epic- and issue-tracking substrate. Across 74 sub-agent sessions, the agent demonstrated closed-loop scientific resilience: routing two failed hypothesis review gates back to literature retrieval, patching bootstrap indexing bugs, and executing with only four discrete human steering events. Finally, we reflect on autonomous research governance, arguing that scientific credibility requires inspectable state, falsifiable review gates, and transparent reporting of negative results, urging the machine learning community to favour agent-accessible structured formats over static PDF manuscripts.
[MA-48] Hybrid epidemic simulation framework coupling equation-based and individual-based models
【速读】：该论文旨在解决大规模聚集性活动（mass gathering events）如何通过局部高密度接触引发传染病传播，并进一步影响城市层面流行病进程的机制问题。其核心挑战在于理解短期、局部的社交接触事件如何通过城市通勤网络扩散，进而改变疫情的空间传播模式与整体规模。解决方案的关键在于构建一个多尺度仿真模型，将事件层面的接触动态与城市范围内的通勤网络相耦合，从而量化不同类型的聚集活动在不同传播强度下的长期城市级影响。研究以西班牙马德里为案例，发现尽管活动期间的初始感染种子规模对疫情加速具有主导作用，但活动后的传播条件同样能提供关于疫情入侵时间的重要预测信号。这一框架揭示了在大型活动中控制传播可有效延缓疫情的广泛扩散，为制定基于空间传播动力学的公共卫生干预策略提供了科学依据。
链接: https://arxiv.org/abs/2609.35162
作者: Jaeyoung Kwak,Michael H. Lees,Chin Chun Ooi,Wentong Cai
机构: 未知
类目: Physics and Society (physics.soc-ph); Computers and Society (cs.CY); Multiagent Systems (cs.MA); Populations and Evolution (q-bio.PE)
备注: 40 pages, 19 figures, 5 tables
Abstract:Mass gathering events like concerts, sports matches, and festivals bring many people into close contact within a short period, creating localized bursts of infection that can shape epidemic outcomes across an entire city. To evaluate how these transient transmission events translate into broader urban impacts, we developed a simulation model linking event-scale contact dynamics with citywide commuting networks. Using Madrid, Spain, as a case study, we compared several types of gatherings and examined how their effects changed under different levels of disease transmissibility. We found that mass gatherings consistently amplified outbreak magnitude, accelerated progression, advanced district-level arrival times, and synchronized spatial spread. Remarkably, while the initial seed size generated at the event accounted for much of this acceleration, post-event transmission conditions provided complementary predictive signal regarding invasion timing. These findings demonstrate that mitigating transmission during mass gatherings can yield downstream public health benefits by delaying broader spatial spread. More generally, this multiscale framework offers a tool to evaluate how temporary, localized contact shifts produce longer-lasting consequences for urban populations.
[MA-49] OneVoice: An Intermediate Representation for Agent ic Speech Pipelines
【速读】：该论文旨在解决多智能体语音系统在协作过程中因语音信号（包括说话人身份、时间戳、音位信息及行为标注等）以专用工具格式输出而导致的跨代理交接脆弱性问题。其核心解决方案是提出OneVoice，一种轻量级、基于JSON的中间表示框架，通过提供统一的语义结构实现语音处理流程的标准化。OneVoice的关键在于将异构语音证据组织为具备稳定标识符、分层转录文本、显式时间关系与来源追溯信息的验证会话记录，从而在多个智能体工作流中显著提升事件聚合的可靠性与时序关系的保真度，验证了专为语音设计的显式表示对智能体间通信的价值。
链接: https://arxiv.org/abs/2609.31673
作者: Vipul Charugundla,Dancheng Liu,Jinjun Xiong
机构: 未知
类目: Audio and Speech Processing (eess.AS); Multiagent Systems (cs.MA); Multimedia (cs.MM); Sound (cs.SD)
备注:
Abstract:Agentic speech systems must exchange more than text, including speaker identity, timing, phonetic information, and behavioral annotations. Yet these signals are often produced in incompatible tool-specific formats, making agent-to-agent handoff fragile. We present OneVoice, a lightweight, JSON-native intermediate representation that provides a shared semantic structure for speech pipelines. OneVoice organizes heterogeneous speech evidence into validated session records with stable identifiers, layered transcripts, explicit timing relationships, and provenance information. We evaluate OneVoice in two complementary multi-agent workflows covering event aggregation and acoustic-phonetic temporal linking across three language models. Compared with implicit agent-defined handoffs, OneVoice substantially improves aggregation reliability and the preservation of temporal relationships, demonstrating the value of an explicit speech-specific representation for agent communication.
自然语言处理
[NLP-0] scopic Language Models
【速读】： 该论文旨在解决大语言模型在实际部署中面临的多计算预算适配难题：传统方法需为每个计算预算（如不同推理延迟或硬件资源限制）单独进行训练或压缩，导致高昂的训练成本与资源浪费。其核心解决方案是提出一种望远镜语言模型（Telescopic Language Model, TLM），通过在训练过程中引入随机前缀监督（stochastic prefix supervision） 与完整锚点目标（full anchor） 的联合监督机制，使单个模型能够自然地支持从浅层到深层的任意层数前缀作为有效语言模型。关键在于：每一步训练同时执行两个前向-反向传播——一个随机截断的前缀分支用于预测下一个词，另一个全容量路径作为基准目标，从而确保模型在所有层级深度均具备良好性能。该方法无需额外架构修改，推理时也无额外开销。相比传统的固定退出结构（如马特约什卡语言模型套件，MLMS），TLM显著降低质量-预算曲线下的面积达43%-44%，且在全容量下性能相当，但每轮训练的GPU成本降低约12%。此外，前缀采样密度可作为可调节参数，通过集中监督特定深度，可恢复固定退出方案的局部性能，实现训练阶段的灵活权衡而非架构约束。研究结果表明，模型的弹性能力主要由训练目标设计决定，而非仅依赖于模型的嵌套结构本身。
链接: https://arxiv.org/abs/2609.35769
作者: Zhilin Guo,Boqiao Zhang,Hakan Aktas,Kyle Fogarty,Nursena Koprucu Aslan,Wenzhao Li,Canberk Baykal,Albert Miao,Siyu Hong,Yixiao Liu,Adam Wu,Ashish Kumar Singh,Sakar Khattar,Chenliang Zhou,Weihao Xia,Cristina Nader Vasconcelos,Cengiz Oztireli
机构: University of Cambridge(剑桥大学); Google(谷歌)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 12 pages, 4 figures, 2 tables. Code: this https URL
Abstract:One deployed language model must often serve many compute budgets, yet serving each budget still means a separate training or compression run per point. We train a Telescopic Language Model (TLM) to be that continuum: a nested-capacity Transformer supervised by stochastic prefix supervision with a full anchor. At every step, one randomly truncated prefix of the capacity axis is trained against the full next-token target, alongside one full-capacity pass, so the trained artifact is a valid language model at every depth. Two forward-backward passes per step, no architectural change, nothing extra at inference. Fixed-exit suites such as Matryoshka Language Model Suites (MLMS) occupy one point in this design space, and the point has a cost: supervising only a few fixed exits leaves the nested model at chance level everywhere else (perplexity 10^2-10^5 in our baselines). On a 200M proxy suite (20B FineWeb-Edu tokens, identical data stream for all methods), a single TLM run is a valid language model at every one of its twenty layer prefixes, in perplexity and on perplexity-sensitive downstream tasks, reducing the area under the quality-budget curve by 43-44% relative to the fixed-exit suites while matching them at full capacity, at ~12% lower GPU cost per run. The prefix sampling density is a dial: concentrating it on a few depths recovers fixed-exit quality there at the price of the continuum, so the operating points become a training-time choice rather than an architectural one. These results indicate that the training objective, not the nesting itself, is what makes a model elastic.
[NLP-1] Retrieving Biblical Intertextual References in Karen Blixens Seven Gothic Tales
【速读】： 该论文旨在解决文学研究中跨文本引用（intertextual references）的计算识别难题，尤其针对源文本在改写、典故化、历史语言及翻译过程中发生语义变形的情况。其核心挑战在于如何在复杂语境下准确检索出作者对《圣经》等经典文本的隐含引用。解决方案的关键在于构建一个包含189个标注引用的基准数据集，并基于丹麦语新旧约译本（共31,170节经文）进行检索评估；通过对比TF-IDF、BM25与多语言及丹麦语句向量编码器的表现，探究语言归一化的影响，并利用硬负样本和五折交叉验证对丹麦语编码器进行微调。研究进一步按词汇重叠程度划分引用类型（直接引述、改写与典故），分析模型在不同层级上的性能表现。结果显示，经过语言归一化的BM25作为零样本基线模型可实现R@10为0.365，且能完整召回所有直接引述；最佳零样本密集模型达到0.360的总体得分，在典故类引用上表现更优；而微调后的DFM-large模型将整体R@10提升至0.508，典故类引用性能从0.138增至0.339。然而，仅以编辑注释为金标准会低估模型的学术价值——有七项被判定为假阳性的一级预测被文学学者认为具有实际意义的额外引用。因此，论文主张将检索模型视为启发式共读工具（heuristic co-readers），既能复现已有文献关联，又能生成供专家主导的细读分析的新候选。
链接: https://arxiv.org/abs/2609.35765
作者: András Kovács,Alexander Conroy,Daniel Hershcovich,Jens Bjerring-Hansen
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Identifying intertextual references is central to literary scholarship, but computationally difficult when source material is transformed through paraphrase, allusion, historical language, and translation. We investigate this problem through biblical intertextuality in Karen Blixen’s Seven Gothic Tales. Drawing on the commentary to a critical edition, we construct a benchmark of 189 annotated references and evaluate retrieval against all 31,170 verses of historically plausible Danish Old and New Testament translations. We compare TF-IDF and BM25 with multilingual and Danish sentence encoders, examine the effect of linguistic normalization, and fine-tune a Danish encoder using hard negatives and five-fold cross-validation. We analyze performance across automatically derived lexical-overlap strata representing quotations, paraphrases, and allusions. Linguistically normalized BM25 provides a strong zero-shot baseline, attaining an overall R@10 of 0.365 and retrieving every quotation within its ten highest-ranked verses. The best zero-shot dense model achieves a comparable overall score of 0.360 while performing better on allusions. Fine-tuning DFM-large raises its overall R@10 from 0.265 to 0.508 and more than doubles its performance on allusions, from 0.138 to 0.339. However, evaluation against editorial annotations alone understates the model’s scholarly usefulness: a literary scholar judged seven of 30 selected rank-one predictions counted as false positives to be meaningful additional references. These findings show both the potential and the epistemic limits of computational intertextual retrieval. Rather than treating scholarly annotations as exhaustive or model outputs as discoveries, we propose retrieval models as heuristic co-readers that recover documented references and generate candidates for expert-led close reading.
[NLP-2] Scaling Long-Form Story Generation via Narrative State Tracking
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在生成长篇小说时面临的叙事一致性难以维持的问题。现有故事生成方法多局限于约一万字以内的短篇叙事，缺乏对长篇小说（如10万字以上）生成能力的系统性探索与评估。为应对这一挑战，本文提出一种无需训练的代理框架——叙事状态追踪代理（Narrative State Tracking Agent, NstAgent），其核心在于使LLMs能够动态跟踪结构化的叙事状态，包括角色、已发生事件及未来情节需求。通过扩展现有基准并结合写作质量评估指标，研究系统地对比了从1万到10万字不等的文本在叙事一致性和写作质量上的表现。实验结果表明，NstAgent在生成更长文本时仍能保持较高的叙事一致性与写作质量，且二者未随长度增加而明显下降，证明该方法为实现长篇小说生成提供了有效可行的解决方案。
链接: https://arxiv.org/abs/2609.35759
作者: Zhennan Wan,Jianfei Chen
机构: Tsinghua University (清华大学); BNRist Center (清华-博世联合实验室); THBI Lab (清华大学脑与智能实验室); Tsinghua-Bosch Joint ML Center (清华大学-博世联合机器学习中心)
类目: Computation and Language (cs.CL)
备注: Under review. Code and data are available at this https URL
Abstract:LLMs have demonstrated strong capabilities in creative writing. However, scaling them to full-length novels remains challenging, as maintaining narrative consistency becomes increasingly difficult. Existing story-generation methods typically focus on stories of up to about ten thousand words, leaving their ability to scale to full-length novels underexplored. In this work, we introduce Narrative State Tracking Agent (NstAgent), a training-free agentic framework that allows LLMs to track a structured narrative state including characters, past events and future requirements. We extend an existing benchmark to compare narrative consistency across lengths, and use it together with a writing-quality benchmark to systematically evaluate stories ranging from 10K to 100K words. We show that NstAgent achieves better narrative consistency and writing quality as stories grow longer, and neither of them degrades noticeably as length increases, suggesting that it provides an effective approach to scaling story generation toward full-length novels.
[NLP-3] How to Loop MoE: Flatten the Experts Untie the Attention
【速读】： 该论文旨在解决稀疏混合专家（MoE）模型在参数高效利用与计算资源分配之间的权衡问题，尤其针对如何有效实现MoE模型的循环结构（looped MoE）这一关键挑战。现有方法中，循环变压器（Looped Transformers）通过重复使用层块以提升固定规模模型的表达能力，而稀疏MoE则仅对每个输入令牌激活少数专家，二者设计哲学不同。为融合两者优势，论文提出Foil方案：其核心创新在于“展平专家”（flattening experts）与“解耦注意力”（untie attention）。具体而言，Foil通过将专家层数减半、每层专家数量加倍并增加循环次数，在保持每令牌专家参数与计算量不变的前提下，扩大路由选择池；同时，为每一循环独立分配注意力参数，而共享专家与路由器。实验表明，Foil显著优于未展平的基线模型：在200亿令牌训练下，所有Foil模型均实现更低的预训练损失；在1000亿令牌训练下，随着展平程度增加，损失持续下降，最展平版本在相同参数与计算量下比基线低0.012 nat，且下游任务准确率相当或更优；解耦注意力还提升了路由分布的平衡性与置信度。消融实验揭示了Foil的有效机制：循环与扩展专家层的收益相互增强，路由置信度比负载均衡更能反映专家使用效率。因此，设计循环MoE应采用更多每层专家数与更多循环次数。
链接: https://arxiv.org/abs/2609.35751
作者: Shouren Wang,Chuang Ma,Mohsen Hariri,Debargha Ganguly,Wang Yang,Xiaoqing Tong,Qianying Liu,Xiaotian Han,Vipin Chaudhary
机构: Case Western Reserve University (凯斯西储大学); Kyoto University (京都大学); NII LLMC (日本国立信息研究所大语言模型研究中心)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 24 pages, 6 figures, 13 tables
Abstract:Looped Transformers reuse one block of layers several times: by spending extra computation they push a model of fixed size further, and so use its parameters more fully; while sparse mixture-of-experts (MoE) models activate only a few of many experts for each token. Looped MoE bridges these two design philosophies and gives MoE models new potential for better expert usage, but it raises a question: how to loop a MoE? We answer it with Foil. With the expert parameters and the expert compute per token held fixed, Foil (1) flattens the experts, halving the expert layers, doubling the experts per layer and doubling the passes, so that every routing decision chooses from a larger pool, and (2) unties the attention, giving each pass its own attention parameters while the experts and routers stay shared. Experiments show that Foil clearly outperforms the unflattened looped baseline: at 20B tokens every Foil model has lower pretraining loss than the baseline; at 100B tokens the loss improves monotonically with the degree of flattening, the most flattened Foil ending 0.012 nat below the baseline at equal parameters and compute, with downstream accuracy on par or better; untying the attention also yields more balanced and more confident routing at equal shape. Our ablations analyse why Foil works and turn the findings into design guidance for looped MoE: the returns of looping and of widening the expert layers amplify each other, routing confidence tracks healthy expert use better than load balance, and a sparse looped MoE should therefore use more experts per layer and more passes. Code and configurations are available at this https URL.
[NLP-4] owards Communication-Efficient Social Intelligence in Language Agents
【速读】： 该论文旨在解决社交智能语言代理在交互中如何高效沟通的核心问题，即在尊重对方时间与注意力的前提下，有效协商、协调并化解偏好冲突，同时避免冗余表达以降低沟通成本。其核心挑战在于如何在不牺牲目标达成率的前提下，优化对话策略与表达精炼度，使每句话都能对推进共同目标产生实质性作用。解决方案的关键在于提出教师辅助通信训练（Teacher-Assisted Communication Training, TACT），通过双专家机制实现对生成行为的动态优化：一是表达专家（expression specialist）负责剔除冗余信息，保留关键动作意图；二是策略专家（strategy specialist）提供更优的替代行动方案，以更好地适应合作方约束。TACT通过采样潜在的伙伴响应，基于局部目标支持度与动作-词元成本之间的权衡，选择最优教师参考路径，并利用该参考进行在线策略蒸馏（on-policy distillation），使学生模型在部署时可独立生成高效且符合社交规范的回应。实验结果表明，TACT在SOTOPIA和AgentSense两个基准上均实现了更高的目标成功率，同时显著减少了目标词元数量与交互消息量，验证了其在提升社交目标达成率与降低沟通开销方面的有效性。
链接: https://arxiv.org/abs/2609.35749
作者: Linxiao Gong,Yijie Xu,Tianfu Wang,Yin Wu,Yili Wang,Xingbo Yao,Huizai Yao,Xilin Xia,Haowen Yang,Hui Xiong
机构: The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州)); University of Science and Technology of China(中国科学技术大学); The Hong Kong University of Science and Technology(香港科技大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Socially intelligent language agents must negotiate, coordinate, and resolve conflicting preferences while respecting the time and attention of both participants. Balancing these demands is challenging because agents must convey enough to address a partner’s constraints and advance their goals without adding words that do not help the interaction. In this paper, we propose Teacher-Assisted Communication Training (TACT) to improve social goal attainment while reducing communication cost, making interactions with agents more productive and less demanding. We first characterize communication efficiency in terms of action strategy and expression, whose effects extend beyond the current utterance to the partner’s response and subsequent exchanges. We design TACT to revise student-generated actions, test the revisions through partner responses, and distill useful feedback into the student. An expression specialist removes unnecessary detail while preserving the intended action, while a strategy specialist proposes alternatives that may better address the partner’s constraints. To determine which revision helps, TACT samples a partner response for each candidate and selects a teacher reference by balancing local goal support against action-token cost. That reference guides on-policy distillation on the student’s own generation prefixes, allowing the student to act independently at deployment. We evaluate TACT on SOTOPIA and AgentSense. On SOTOPIA, it achieves the highest Goal among the evaluated methods on All and Hard while using substantially fewer target tokens than SFT+SDPO. On AgentSense, it improves goal success over the initial student while reducing target tokens and interaction messages.
[NLP-5] Improving Test-Time Scaling with Adaptive Looped Transformers
【速读】： 该论文旨在解决生成式模型在测试时扩展（test-time scaling）过程中，随着输出序列长度增加，传统循环变压器（looped transformers）在计算效率与精度提升之间权衡不佳的问题。现有方法虽通过重用层实现参数高效，但在固定深度循环策略下，所有输出token均被赋予相同迭代次数，导致大量冗余计算且部分token未能获得实际收益。其核心解决方案在于提出一种新型后训练框架TaH2，通过引入“前瞻深度监督”（lookahead depth supervision），联合训练骨干网络与一个迭代决策器，动态判断每个token是否需要额外迭代以提升预测性能。该机制使模型能够聚焦于真正受益于循环的token，显著优化了测试时计算资源的分配效率。实验表明，在AIME等高难度基准上，TaH2将准确率-计算斜率（accuracy-compute slope）提升53%（2.74 vs. 1.79），并在匹配计算量下超越非循环基线约3.4个百分点；随着最大迭代深度从2增至8，其相对于基线的增益持续增长至+3.9个百分点，有效突破了传统循环模型的性能瓶颈。
链接: https://arxiv.org/abs/2609.35748
作者: Yichen You,Tianyu Fu,Aosong Feng,Xingtai Lv,Xuefei Ning,Ning Ding,Yu Wang
机构: Tsinghua University (清华大学); Yale University (耶鲁大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Looped transformers have demonstrated promising parameter efficiency by reusing layers for latent computation. Prior studies compare looped and non-looped models at matched parameters or per-token FLOPs. However, to the best of our knowledge, whether looping improves test-time scaling as outputs grow longer remains underexplored. Through post-training looped transformers, we study the accuracy-compute slope, measured as the accuracy gain per doubling of test-time decoding FLOPs. We find that existing looped transformers often yield steeper slopes than their non-looped baseline, yet underperform it at matched compute. While fixed-depth looping spends extra iterations on every token, our analysis shows that many tokens do not benefit from extra iterations. We therefore propose TaH2, which enables the model to focus extra iterations on the tokens that benefit from looping. It jointly post-trains the backbone and an iteration decider through lookahead depth supervision, which uses online labels indicating whether further iteration improves the prediction. TaH2 improves both the efficiency and attainable accuracy of test-time scaling. On challenging AIME benchmarks, TaH2 improves the accuracy-compute slope by 53% (2.74 vs. 1.79) over the non-looped baseline, exceeding the baseline’s peak accuracy by about 3.4 points at matched test-time compute. As the maximum iteration depth increases, existing looped models largely plateau, while TaH2’s gain over the non-looped baseline continues to grow from +2.8 points at depth 2 to +3.9 points at depth 8. Our code is available at this https URL.
[NLP-6] Shockingly Simple Self-retrospection Improves Agent ic Models Without RL
【速读】： 该论文旨在解决如何仅通过自我经验的解释（retrospection）来提升语言模型代理在未来任务中的表现，即探究在不依赖外部教师或基于奖励的策略更新的前提下，仅使用自身生成的解释进行微调是否能够实现有效学习。其核心解决方案是提出一种极简的在线训练方法——仅回顾微调（Retrospection-Only Fine-Tuning, ROFT），该方法在模型执行任务并获取反馈后，仅基于其自动生成的回顾性解释文本进行下一步的下一个词预测损失微调，完全不引入外部监督或奖励信号。实验结果表明，该方法在软件工程基准测试SWE-bench Verified和Pro上分别达到49.2%和26.8%的求解率（20次更新后），优于需40次更新的GRPO方法，且在训练初期进展更快、采样效率更高；更关键的是，ROFT能在所有初始尝试均失败的任务上实现突破，证明了无需成功轨迹即可启动学习过程。行为分析进一步揭示，解释过程间接实现了对动作的信用分配机制，从而引导模型强化正确行为、抑制错误行为；同时，通过提示模型强调更直接的解决方案，即使无显式长度惩罚，后续尝试也显著缩短。这些发现共同表明，自我生成的解释不仅可作为有效的训练目标，还能促进从“解释”到“行动”的跨模态迁移，为生成式人工智能（Generative AI）中自主反思与持续学习提供了新范式。
链接: https://arxiv.org/abs/2609.35741
作者: Jonathan Light,Christopher Zhang Cui,Jeonghye Kim,Roger Creus Castanyer,Emiliano Penaloza,Zhengyan Shi,Alessandro Sordoni,Marc-Alexandre Côté,Xingdi Yuan,Minseon Kim
机构: UC San Diego(加州大学圣地亚哥分校); KAIST(韩国科学技术院); Mila(蒙特利尔人工智能研究所); Microsoft Research(微软研究院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 62 pages, 18 figures, 5 tables, including appendices
Abstract:People learn not only by repeating successful actions, but also by recounting and explaining their experiences, revising their understanding to guide future behavior. Can a language-model agent improve its future actions by training only on explanations of its own experience? We investigate this question by studying Retrospection-Only Fine-Tuning (ROFT), a minimal online procedure designed to isolate the effect of explanation-only training on subsequent behavior. The agent attempts a task, observes available feedback, generates a retrospective explanation, and is fine-tuned with a next-token prediction loss on the explanation tokens alone. The procedure uses neither an external teacher nor a reward-based policy update. In software-engineering experiments with Qwen3.5-4B, ROFT is trained on problems with mixed successful and unsuccessful base-model attempts. On held-out SWE-bench Verified and Pro, it reaches 49.2% and 26.8% solve rates after 20 updates without using a verifier, compared with GRPO’s 48.0% and 25.3% after 40 updates in the evaluated runs, and makes faster early progress in training time and sampled attempts. It also learns to solve individual tasks on which all 64 sampled base-model attempts failed, showing that learning can begin without any initially successful trajectories. Behavioral analyses find that ROFT indirectly assigns credit to actions, encouraging good actions and discouraging incorrect ones. Moreover, prompting retrospections to emphasize more direct solutions yields shorter subsequent attempts even without an explicit length penalty. Together, these findings show that learning to explain can also improve learning to do, establishing self-generated retrospections as useful training targets and motivating further study of explanation-to-action transfer.
[NLP-7] Harness Learning Enables Generalizable Test-Time Adaptation
【速读】： 该论文旨在解决语言模型代理（language-model agent）在面对不同任务时，其执行框架（harness）需动态适应的问题。现有方法通常依赖静态或预设的框架结构，难以有效应对复杂、多变的任务需求，导致性能受限。为解决此问题，论文提出“框架学习”（harness learning）这一新范式，其核心在于通过可执行程序的元学习机制，将框架的迭代优化过程类比为梯度更新，在不修改模型参数的前提下，利用任务执行反馈动态调整框架结构。关键创新在于：引入一个提议模型（proposer），通过强化学习训练，以修订后框架在任务中的表现作为奖励信号，从而学会生成更优的框架配置；在推理阶段，该模型可基于新任务的连续执行反馈逐步精细化框架，实现无需参数更新的在线适应。实验表明，该方法显著提升了框架修订质量，并在未见任务上展现出良好的泛化能力，验证了构建持续学习型智能体的可行性——即通过累积经验实现可迁移的通用改进。
链接: https://arxiv.org/abs/2609.35738
作者: Alvin Zhang,Xuecheng Liu,Zixuan Wang,Fahim Tajwar,Daman Arora,Ruslan Salakhutdinov,Daniel Khashabi,Yuda Song,Andrea Zanette
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:A language-model agent is jointly defined by its model and its harness, the executable program that organizes model calls, tool use, and information flow. Because different tasks call for different ways of organizing these operations, the harness needs to be adapted using feedback from the task at hand. We introduce harness learning, which trains a proposer model to revise a solver’s harness using execution feedback. We formulate this process as meta-learning over executable programs, with harness revisions playing the role of weight updates in gradient-based adaptation. We train the proposer with reinforcement learning, using the task performance of revised harnesses as the reward. At test time, the proposer uses feedback from successive executions on a new task to refine the harness, without performing any parameter-space update. Experiments on reasoning and multi-hop question answering show that harness learning improves revision quality and that the ability to adapt at test time transfers to unseen tasks. Policies trained on individual revisions can continue improving harnesses over multiple rounds, while the benefits of training on revision sequences vary across settings. These findings suggest a path towards continually learning agents that turn accumulated experience into generalizable improvements.
[NLP-8] Reinforcing Agent ic Creativity in Scientific Ideation with Night Science MICRO
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在开放性科学创意生成任务中因低熵偏倚导致输出同质化、可预测性强的问题，从而限制其在探索性科研中的应用潜力。现有模型虽在结构化、可验证的任务中表现优异，但在需要突破常规思维框架的“夜科学”（night science，即非结构化、偶然性探索）场景下能力不足。为此，论文提出AI Night-Scientist这一代理式框架，通过强化学习（Reinforcement Learning）训练模型识别何时以及如何偏离常规推理路径以激发创造性。该框架基于认知科学，将创造力建模为三个维度：行动（Action，即采取何种创造性行为）、过程（Process，即探索与利用之间的时机权衡）和结果（Outcome，即想法的新颖性与实用性）。利用广义相对策略优化（GRPO）算法，在训练过程中引入不同程度与形式的创造性引导，使模型能够生成更丰富的科学设想，相较基线模型，研究方向多样性提升27.8%，贡献类型多样性提升14.9%，预测引用影响力最高提高32.0个百分点，原创性得分提升66.2分。关键发现在于，仅提高解码温度无法实现同等效果，而明确的语义引导（semantic guidance）以指定具体创造类型才是取得显著提升的核心因素。研究表明，创造力是一种可学习的多层级能力，可通过系统性训练帮助研究人员突破传统模型的思维局限，探索更广泛且具有创新性的科学路径。
链接: https://arxiv.org/abs/2609.35706
作者: Priyanka Kargupta,Silviu Cucerzan,Shweti Mahajan,Allen Herring,Jiawei Han,Ryen W. White,Sujay Kumar Jauhar
机构: University of Illinois Urbana-Champaign (伊利诺伊大学厄本那-香槟分校); Microsoft (微软); Microsoft Research (微软研究院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Code: this https URL Website: this https URL
Abstract:Large language models (LLMs) excel at structured, verifiable tasks, but their low-entropy bias can produce homogeneous and predictable outputs, limiting their utility for open-ended scientific ideation. Effective discovery, however, spans a broader creative spectrum: from structured day science to loosely structured, serendipitous night science that reaches ideas beyond those typically considered. We introduce AI Night-Scientist, an agentic framework that uses reinforcement learning to teach models when and how to depart from predictable reasoning. Grounded in cognitive science, we model creativity along three axes: action (what to do and how creatively), process (when to explore versus exploit), and outcome (the novelty and usefulness of the resulting idea). We use these axes to train models with GRPO, exposing them to varying degrees and forms of creativity throughout training. This produces substantially more diverse scientific proposals, expanding the range of research directions by 27.8% and contribution types by 14.9% over the base model. It also improves predicted citation impact by up to 32.0 percentage points and originality by 66.2 points. These gains cannot be reproduced by simply increasing decoding temperature; instead, we find that semantic guidance specifying what kind of creativity to pursue is critical. Overall, our results suggest that creativity is a learnable, multi-level ability that can be shaped to help researchers reach ideas beyond those typically explored by LLMs.
[NLP-9] Rethinking Personalized Generation: Test-Time Alignment via Factorized Ranking Models NEURIPS2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在对多样化用户偏好进行对齐时所面临的根本性挑战，即传统对齐范式通常针对单一、统一的用户假设，难以有效支持个性化生成。其核心问题是：尽管通过测试时对齐（test-time alignment）可挖掘出巨大的未利用性能潜力（performance headroom），但现有基于奖励模型（reward models）的方法在个性化场景下存在校准不佳与计算成本过高的双重瓶颈——尤其是百亿参数级奖励模型难以高效评估大规模候选生成结果。为此，本文提出一种参数高效的解决方案，关键在于设计一个仅需数百万参数的多层感知机（Multi-Layer Perceptron, MLP）排序模型，该模型直接复用基础生成器内部的嵌入表示（embeddings），实现极低开销的个性化评分。通过训练时引入细粒度的个性化数据以建模用户偏好，该小型排序模型能够准确评估大规模候选集，并无缝引导生成过程，显著降低生成候选的资源消耗。实验结果表明，在九个涵盖三类个性化生成任务的数据集上，该方法在性能上全面超越百亿参数通用奖励模型，且参数量不足其0.4%，评分延迟降低四个数量级，充分验证了其在个性化生成中对潜在性能头空间的有效利用能力。
链接: https://arxiv.org/abs/2609.35695
作者: Qiyao Ma,Junshan Zhang,Zhe Zhao
机构: University of California, Davis(加州大学戴维斯分校)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Accepted to NeurIPS 2026
Abstract:Aligning large language models (LLMs) to diverse user preferences is fundamentally hindered by standard alignment paradigms that optimize for monolithic users. In this work, empirical studies are first used to reveal the existence of a massive, untapped performance headroom for personalized generation through test-time alignment. We demonstrate that personalized generation is uniquely suited for test-time scaling methods like Best-of-N (BoN) because it can be viewed primarily as a candidate matching problem rather than a generator capability bottleneck. While reward models could in principle exploit this headroom, they are poorly calibrated for personalization, and their billion-parameter scale makes scoring large candidate pools prohibitively expensive. To overcome this limitation, we propose a parameter-efficient framework utilizing million-parameter scale multi-layer perceptron (MLP) ranking models. Our personalized ranking model directly reuses the internal embeddings of the base generator with minimal overhead. By scaling train-time data to provide fine-grained personalized preferences, this million-parameter ranking model accurately scores large candidate pools and can seamlessly guide generation to reduce the cost of materializing N candidates. Extensive experiments on nine datasets spanning three personalized generation settings show that our personalized ranking model effectively exploits the discovered headroom, outperforming billion-parameter generalist reward models on every dataset, with under 0.4% of their parameters and four orders of magnitude lower scoring latency.
[NLP-10] QuanReview: Offline Auditable Reconciliation of Human and LLM Span Annotations
【速读】： 该论文旨在解决结构化跨度标注（structured span annotations）在生成式AI（Generative AI）介入后难以保持可信性的问题，尤其是针对数量与单位、不确定性修饰符及事件类别等关键信息的标注，其人工创建成本高且易受模型输出干扰。解决方案的关键在于提出QuanReview——一个开源的审计与修正系统，通过字符级对齐两路标注流，采用可追溯的显式政策自动处理明确无歧义案例，并将潜在冲突项推送至基于浏览器的仲裁界面，支持评审者选择任一标注、构建字段级混合标注或标记重标注。系统还配备项目管理模块，可配置冗余分配文档给多名标注员，计算文档与跨度级别的一致性，自动合并完全一致的文档，并以原始文件格式导出修正后的标注层，实现无缝替换。实验表明，在4,457条人道主义基准数据和大语言模型（LLM）提取流中，系统实现了8%文档的全自动合并，1,513条记录通过自动策略处理，仅需人工干预3,131个候选冲突，平均每份文档5.4个，显著提升了标注效率与可信度。
链接: https://arxiv.org/abs/2609.35685
作者: Matteo Musacchio,Juan Cruz Giner Pulero,Isabel Castañeda,Naomi Couriel,Yelena Mejova,Mariano G. Beiró,Kyriaki Kalimeri
机构: Universidad de San Andrés, Buenos Aires, Argentina; ISI Foundation, Turin, Italy; CONICET, Buenos Aires, Argentina; UNICEF, New York, NY, USA
类目: Computation and Language (cs.CL)
备注: 6 pages, 2 figures, 4 tables. System demonstration. Code and runnable demo: this https URL
Abstract:Structured span annotations, such as quantities with their units, uncertainty modifiers, and event classes, are expensive to create and hard to keep trustworthy once language models enter the loop. We present QuanReview, an open-source system for auditing and correcting such annotation layers. QuanReview aligns two annotation streams over the same documents at character level, resolves unambiguous cases by an explicit and logged policy, and routes candidate conflicts to a browser-based adjudication interface where reviewers accept either side, build field-level hybrids, or flag items for re-annotation. A campaign manager assigns documents to multiple annotators with configurable redundancy, computes agreement at document and span level, auto-merges unanimous documents, and exports the corrected layer in the original file format, so that it can replace the original annotation files directly. Applied to a 4,457-record humanitarian benchmark and an LLM extraction stream, the system fully auto-merged 8% of documents, applied automatic policy decisions to a further 1,513 records, and concentrated human attention on 3,131 candidate conflicts, a mean of 5.4 per reviewed document.
[NLP-11] racing the Evolution of Oracle Bone Characters Across Three Millennia
【速读】： 该论文旨在解决甲骨文（Oracle Bone Inscription, OBI）中大量未破译字符的识别难题，尤其针对汉字在历史演变过程中因朝代更迭导致的结构与语义剧烈变化所引发的跨时期对比困难问题。传统计算方法通常仅将甲骨文与单一历史时期的字形进行比较，但在缺乏明确年代标记的演化阶段，这种单一时段参照难以应对字形的显著变迁。为此，本文提出基于流形（Manifold）的书写体系演化框架（Manifold-based Script Evolution Framework, MSEF），将甲骨文、金文、小篆、隶书、楷书等五个主要阶段的汉字演化建模为一个连续的流形空间演化过程。该框架将每个字符在不同历史时期的形态表示为流形空间中的特定点，并利用神经微分方程（Neural Ordinary Differential Equations, Neural ODEs）学习跨时期之间的连续过渡规则。通过任意两个时期间的字符演化配对数据，可实现流形空间结构与演化动态的端到端联合训练，从而有效捕捉汉字在长时段演化中的非线性变化规律，提升对未破译甲骨文字形的推断能力。
链接: https://arxiv.org/abs/2609.35674
作者: Tianhao Fu,Xinxin Xu,Spike Wang,Cunyi Kang,Jian Cao,Xixin Cao
机构: Fulcrum.AI; Peking University (北京大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Of the approximately 4,500 Oracle Bone Inscription (OBI) characters discovered from the Shang dynasty, only about 1,600 have been deciphered. Many computational approaches compare OBI with glyphs from one historical period at a time. However, during the evolution of Chinese characters, significant structural or semantic changes often occur in uncertain dynasties. A single-period reference may be insufficient when relevant forms change substantially between observed eras. Therefore, we propose the \textbfManifold-based Script Evolution Framework (MSEF), a framework that models the evolution series (OBI, Bronze, Seal, Clerical, Regular) of Chinese characters as the continual evolution of a manifold space. MSEF represents each character as an era-specific manifold point and learns continuous inter-era transition rules via Neural Ordinary Differential Equations. Both manifold space and transition dynamics can be trained end-to-end through character evolution pairs across any two eras.
[NLP-12] MS-GLA: Multi-Scale Gated Linear Attention for Addressing Representational Bottlenecks via Multi-Temporal Resolution
【速读】： 该论文旨在解决生成式AI（Generative AI）中线性递归模型——即门控线性注意力（Gated Linear Attention, GLA）——所面临的表征瓶颈问题：尽管通过数据依赖性门控机制提升了建模能力，但所有注意力头共享固定容量的内存矩阵，且在单一时间分辨率下处理每个词元，导致模型需同时捕捉局部句法模式与长程语义结构，造成表征冲突。其解决方案的关键在于提出多尺度门控线性注意力（Multi-Scale Gated Linear Attention, MS-GLA），通过将注意力头分配至多个时间分辨率层级，实现对不同时间尺度依赖关系的解耦建模——粗粒度分辨率自然聚合更长的词元跨度，专注于长程依赖；细粒度头组则保留对局部句法结构的敏感性。同时引入可学习、输入相关的融合层，在每个时间步动态重组各头组输出，以不增加单个头状态大小为前提扩展有效记忆容量。该方法借鉴了多尺度状态空间模型（Multi-Scale State-Space Models, MS-SSM）的思想，并将其适配至门控线性注意力框架。实验表明，MS-GLA在语言建模、强回忆任务及长上下文泛化任务中均显著优于基线模型，在相同参数量下实现了更高的准确率和更低的困惑度，尤其在强回忆任务上性能提升达18.9%，语言建模平均困惑度降低9.5%，验证了多时间分辨率分解作为门控线性注意力的原理性且高效扩展路径的有效性。
链接: https://arxiv.org/abs/2609.35664
作者: Prasoon Dev,Anirudh Sankar,Vasudeva Varma
机构: Language Technologies Research Center; International Institute of Information Technology Hyderabad (国际信息科技研究所海得拉巴分校); Hyderabad, Telangana, India
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Gated Linear Attention (GLA) Transformers advance linear recurrent models through data-dependent gating, but face a core limitation: the fixed-capacity memory matrices across all heads operate at a single temporal resolution, where each token is processed individually, forcing them to simultaneously encode local syntactic patterns and long-range semantic structure, creating a representational bottleneck that gating alone is insufficient to resolve. We introduce Multi-Scale Gated Linear Attention (MS-GLA), which addresses this by distributing attention heads across multiple temporal resolutions. Coarser resolutions pool longer token spans naturally specializing toward long-range dependencies, while finer head groups retain sensitivity to local syntactic structure. A learnable, input-dependent fusion layer dynamically recombines head group outputs at each timestep, expanding effective memory capacity without increasing per-head state size. This multi-resolution decomposition draws on principles from Multi-Scale State-Space Models (MS-SSM), adapting them to the gated linear attention setting. We evaluate MS-GLA on language modeling, recall-intensive tasks, and long-context generalization. Across all settings, MS-GLA consistently achieves higher accuracy and lower perplexity than GLA at matched parameter counts, with up to 18.9% improvement on recall-intensive tasks and 9.5% lower average perplexity on language modeling benchmarks, validating multi-temporal resolution decomposition as a principled and effective extension of Gated Linear Attention.
[NLP-13] Late Attention Layers Alone Can Copy Entity Tokens but Not Without Attending to Their Context
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在执行实体复制（entity copying）任务时，其内部各层的功能分工机制不明确，以及上下文令牌（context tokens）如何影响模型对实体令牌（entity tokens）的准确复制这一关键问题。现有研究未能系统揭示哪些层级专门负责该基础任务，也未阐明非实体类上下文信息如何通过注意力机制间接支持实体复制。为此，作者基于Qwen3-8B模型提出两种新方法：瓶中精灵法（genie-in-a-bottle），用于精确控制参与实体复制的网络层；注意力截除法（attention lobotomy），可切断特定令牌对实体令牌的注意力而不改变其余注意力分布。实验结果表明，模型后半部分的两组独立层共同构成实体复制的必要且充分条件；更重要的是，不仅解码位置需关注实体令牌，上下文令牌对实体令牌的注意力亦为精确复制所必需——尽管这些上下文令牌本身并不存储实体信息，除非满足特定语义属性。该研究揭示了晚期层在上下文引导下完成实体复制的关键作用，强调了模型在传播与消费实体信息过程中的深层机制，为未来研究模型内部信息流动路径提供了重要启示。
链接: https://arxiv.org/abs/2609.35663
作者: Muyu He,Yuchen Liu,Ran Tao,Li Zhang
机构: Drexel University (德雷塞尔大学); Independent; University of Pennsylvania (宾夕法尼亚大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large language models (LLMs) reliably perform entity copying, in which a model copies tokens referring to an entity, termed entity tokens, from the prompt into its output to answer a question. Although entity copying is straightforward for most LLMs, existing research does not provide a systematic account of which layers specialize in this fundamental task or how other tokens in the same sequence, termed context tokens, influence the model’s ability to copy the entity tokens. To address these questions, we conduct experiments on Qwen3-8B using two novel methods: genie-in-a-bottle, which controls exactly which layers can participate in an entity-copying task, and attention lobotomy, which cuts off specific tokens’ attention to entity tokens without affecting the remaining attention distribution. We find that two distinct groups of layers in the second half of the model are both necessary and sufficient for entity copying. Moreover, in addition to the decoding position’s attention to entity tokens, context tokens’ attention to entity tokens also proves necessary for copying the exact tokens, even though context tokens do not store entity information themselves unless they satisfy particular semantic properties. Our findings establish the critical role of late layers in entity copying under the guidance of context tokens, calling for future work on how models propagate and consume entity information.
[NLP-14] Rubric Rewards from Item Response Theory
【速读】： 该论文旨在解决在语言任务中缺乏可自动验证的单一答案时，如何有效利用评分量规（rubric）生成高质量强化学习奖励信号的问题。现有方法通常将各评分维度的得分简单相加，导致不同评判模式可能获得相同奖励，且固定分值无法反映各维度对当前生成样本质量区分度的实际贡献，存在聚合偏差；同时，随着评分维度数量增加，需调用大量人工判断，带来高昂成本。其解决方案的核心是提出评分响应理论（Rubric Response Theory, RRT），基于单调性假设（即各评分维度均为共享目标的单调指标），采用双参数项目反应模型（two-parameter item response model），将完整的评判模式视为关于样本质量的潜在标量值的证据。该模型通过最大化局部信噪比来估计质量得分，并引入响应参数网络（Response Parameter Network, RPN），根据提示和评分标准文本预测每个维度的难度与区分度参数。在训练过程中，RRT采用在线期望最大化（online expectation maximization）动态更新RPN，以适应策略分布的变化。实验表明，在Qwen3.5-4B作为策略模型的情况下，RRT在医学、科学及多个基准测试上的宏观评分显著优于组相对策略优化（GRPO），尤其在高难度与极难标准上提升达2.8至5.6分；此外，结合自适应Fisher选择策略并冻结RPN后，仅使用一半的评判预算即可保持与全量评判下GRPO相当的性能（误差<0.1分），证明了RRT在降低人工评判开销的同时仍能维持优异的评估能力。
链接: https://arxiv.org/abs/2609.35646
作者: Milad Yazdani,Yaser Souri,Xiren Zhou,Pranit Chawla,Dena Shahriari,Subhojit Som,Xia Song
机构: University of British Columbia(不列颠哥伦比亚大学); Microsoft(微软)
类目: Computation and Language (cs.CL)
备注:
Abstract:Many language tasks have no single answer that can be checked automatically. Rubrics provide criteria for judging responses to these tasks. For reinforcement learning, the resulting verdicts must be combined into a scalar reward. A common approach sums the points assigned to satisfied criteria. Distinct verdict patterns can thus receive the same reward, and the fixed points encode how much each criterion should count, not how strongly its verdict distinguishes the current rollouts. Beyond this aggregation problem, judging the full rubric needs more judge requests as the criterion count grows. To address these limitations, Rubric Response Theory (RRT) measures quality and selects criteria when rubric criteria are monotone indicators of a shared target. Rather than adding assigned points, RRT uses a two parameter item response model that treats the verdict pattern as evidence about scalar quality specific to the rubric. Under this model, its likelihood score maximizes the local signal-to-noise ratio for quality. Its Response Parameter Network (RPN) reads the prompt and criterion text to predict criterion difficulty and discrimination. As the policy distribution changes during training, RRT uses online expectation maximization to update the RPN from current rollout verdicts. With Qwen3.5-4B as the policy, RRT’s macro criterion score across Medical, Science, Rubrics as Rewards Science, and RubricBench is 1.7 points above that of group relative policy optimization (GRPO). On hard and very hard criteria in Medical and Science, RRT gains 2.8 to 5.6 points over GRPO. At half the criterion budget, adaptive Fisher selection with a frozen RPN keeps the macro criterion score across four datasets within 0.1 points of GRPO with full judging. These results show RRT can reduce judge requests while remaining competitive with GRPO.
[NLP-15] CoSE-E: A Benchmark for Code-switched Speech Evaluation in Enterprise Settings EMNLP2026
【速读】： 该论文旨在解决企业在实际部署中面临的多语言语音识别（Code-Switching Automatic Speech Recognition, CS-ASR）难题，尤其关注语音代理在真实业务场景下因代码切换导致的转录错误如何进一步引发下游任务失败的问题。传统评估方法仅依赖编辑距离（edit-distance）衡量性能，忽略了错误对业务流程的实际影响，而本文提出的关键解决方案是构建一个面向企业领域的合成数据基准（COSE-E）与多维度评估框架，系统性地评估前沿ASR模型在5种语言组合下的表现，并通过诊断分析揭示不同语言对及模型架构下代码切换引入的额外转录错误模式。该工作不仅提供了更贴近真实应用的评估体系，还为优化企业级多语言语音代理的鲁棒性提供了可复现的技术支撑。
链接: https://arxiv.org/abs/2609.35645
作者: Shama Gupta,Hoang H Nguyen,Chelsea Huang,Lindsay Devon Brin,Fanny Riols
机构: ServiceNow AI Research; Qualcomm Technologies, Inc.
类目: ound (cs.SD); Computation and Language (cs.CL); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
备注: Accepted to SALMA Workshop (Oral) at EMNLP 2026
Abstract:Code-switching (CS), a seamless alternation between languages within a single utterance, remains a critical challenge in automatic speech recognition (ASR). While prior works focus on conversational CS-ASR, enterprise settings demand evaluation of operational impact beyond edit-distance errors: how code-switching transcription errors propagate to downstream voice agent task failures. In this work, we propose (1) a CS-ASR synthetic benchmark and multidimensional evaluation framework tailored to enterprise domains, (2) systematic evaluation of frontier ASR systems across 5 language pairs, (3) diagnostic analysis of the additional transcription errors that code-switching introduces across language pairs and models. We release COSE-E to support enterprise-focused CSASR evaluation for multilingual voice agents in enterprise deployment.
[NLP-16] Which the Eye Fears: Writing with Read-Blindness Explains Massive Activations in Transformers
【速读】： 该论文旨在解决生成式 AI（Generative AI）中大规模激活特征（Massive Activation Features, MAs）在 Transformer 模型中持续存在且难以抑制的机制问题。尽管模型具备抑制异常激活的能力，但 MAs 仍能在多层间累积并持久化，其根本原因在于注意力（attention）与前馈神经网络（Feed-Forward Network, FFN）模块之间存在的读写不对称性：这些模块在“读取”时对 MAs 坐标视而不见（read-blindness），但在“写入”时仍继续传递和放大该信号，从而阻断了纠错反馈路径并促成 MAs 的持续积累。研究通过操作级机制分析发现，这种读盲现象在训练初期即已出现，早于 FFN 的放大作用，表明其是驱动 MAs 形成的上游机制；进一步的梯度分析揭示了损失函数景观中的非对称特性，暗示模型主动维持该读盲行为。尽管在不同位置移除读阻机制会引发系统补偿性调整，但 MAs 依然顽固存在，说明其形成具有深层架构与优化层面的内在根源。
链接: https://arxiv.org/abs/2609.35630
作者: Swagatam Mukhopadhyay,Vishal Vivek Saley,Vraj Parikh,Mausam
机构: PsiDagger; IIT Delhi(印度理工学院德里分校)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Massive activation features (MAs) in Transformers are extreme-value residual-stream features that persist across layers despite the model’s ability to suppress them. Why do they survive? Our investigation using an operator-level mechanistic analysis of attention and feed-forward (FFN) blocks reveals that these blocks systematically ignore MA coordinates while reading, but not while writing; creating a read-write asymmetry that blocks corrective feedback while allowing continued accumulation. We find that both attention and feed-forward layers have this read-blindness, and contribute to the emergence and persistence of MAs. To validate prior work that hypothesized that FFN’s amplification abilities is the primary reason for MAs (Sun et al., 2026), we analyze the model checkpoints during learning. Contrary to our expectation, read-blindness emerges before FFN amplification, suggesting that it acts upstream in the MA mechanism. We further contribute gradient analysis to link this behavior to surprising asymmetries in the loss landscape, concluding that the model actively maintains this read-blindness. Finally, we find that removing read-blocking at different locations induces compensatory shifts elsewhere, but MAs still persist. Subjects: Computation and Language (cs.CL); Machine Learning (cs.LG) Cite as: arXiv:2609.35630 [cs.CL] (or arXiv:2609.35630v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2609.35630 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-17] SANTA: Sampling Attention through Representative Keys
【速读】： 该论文旨在解决大模型在长序列推理过程中因注意力机制（Attention）需遍历整个键值（KV）缓存而导致的高内存读取开销与计算瓶颈问题。传统密集型注意力机制（Dense Attention）在处理长上下文时，其计算复杂度随序列长度线性增长，严重制约了推理效率。针对这一挑战，论文提出SANTA++——一种无需训练的随机化注意力方法，其核心创新在于通过代表性键（representative keys）实现高效的采样策略：将缓存中的键组织为若干“团队”（teams），查询仅对每个团队的一个代表性键进行评分，从而决定哪些团队应被采样；随后在采样团队内计算精确注意力得分，并通过逆包含概率对各团队贡献进行重要性加权校正，以无偏估计全量缓存的注意力分布。该方法的关键在于利用重要性采样（Importance Sampling）框架，在显著减少KV缓存读取次数的同时保持近似稠密注意力的性能表现。实验表明，当仅采样32或64个团队时，SANTA++可将KV读取量降低至稠密注意力的16%–22%，并在LongBench v2、HELMET和RULER等基准上分别保留94%–99%、85%–91%的原始性能，同时在32K上下文长度下实现1.69倍于FlashAttention的推理加速。此外，该方法天然适配压缩式KV表示架构（如多头潜在注意力，Multi-Head Latent Attention），具有良好的可扩展性与工程实用性。
链接: https://arxiv.org/abs/2609.35629
作者: Kyle Lee,Christian Z. Pratt,Ruoyu Fang,Heekyung Lee,Avinash Lohitsa,Ryan Modafe,Kerem Y. Camsari
机构: University of California, Santa Barbara(加州大学圣塔芭芭拉分校); Flucta(弗拉克塔)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Attention often concentrates on a small subset of tokens in the context, but which subset matters changes from one query to the next. To exploit this changing structure, we introduce SANTA++, a training-free stochastic attention method that uses representative keys for memory-efficient selection without scanning the entire key-value (KV) cache. Cached keys are organized into teams, and the query scores one representative from each team to decide which teams to sample. We compute exact attention scores within the sampled teams and reweight each team’s contribution by the inverse of its inclusion probability. This importance sampling correction estimates attention over the full cache, with a sampling budget that lets us trade memory reads for accuracy. Remarkably, with 32 or 64 sampled teams, SANTA++ uses 16% to 22% of dense attention’s KV reads and retains 94% to 99% of the dense-attention baseline’s scores on LongBench v2 and HELMET’s retrieval-augmented generation subset, and 85% to 91% on RULER, with Qwen2.5-7B-Instruct at 32K context. With 31 sampled teams, our GPU implementation delivers a 1.69\times attention speedup over the dense FlashAttention baseline at 32K context. By reducing the number of cache entries read, SANTA++ in principle complements architectures with compressed KV representations, such as multi-head latent attention. Our kernels are available at: this https URL.
[NLP-18] Can LLM s Value the Right Evidence? Evidence-Value Misalignment in Dynamic Medical Diagnosis
【速读】： 该论文旨在解决临床诊断中因证据不足或误导性信息导致的误诊风险，特别是现有基于结果准确率的评估体系可能奖励“侥幸猜对”的问题，从而引发诊断决策与可用证据价值之间的错配，即证据-价值错位（Evidence-Value Misalignment, EVM）。其核心解决方案在于提出MedEVM动态评测环境，通过1,050个跨24个疾病系统的病例，模拟逐轮证据输入过程，要求模型在等待更多证据与提交诊断之间持续决策。研究揭示四大关键现象：（1）证据追踪严重失准，即使高能力模型在推理模式下亦难以正确评估证据充分性；（2）诊断提交时机错位，即便证据已充分，模型仍无法及时提交正确诊断；（3）证据顺序显著影响诊断结果，相同证据重排即可改变诊断结论；（4）误导性证据具有持久影响力，即使已有充分证据，仍可被新加入的误导信息扭转判断。基于上述发现，论文提出证据验证诊断控制框架（EVD-Harness），通过离线对比诊断知识库与在线三阶段控制机制——观测管理、提案与证人验证、诊断提交控制，实现诊断生成与提交的解耦。在五种大语言模型上，该框架使诊断准确率提升12.0–51.1个百分点，显著缓解了由EVM引发的错误。研究证明，在提交前验证证据支持度，可显著提升诊断决策的可靠性。
链接: https://arxiv.org/abs/2609.35627
作者: Kehua Feng,Yunsheng Lu,Yitong Qiao,Tiantian He,Lei Liu,Yue Shen,Jian Wang,Jinjie Gu
机构: Zhejiang University (浙江大学); Ant Healthcare, Ant Group (蚂蚁健康)
类目: Computation and Language (cs.CL)
备注: 33 pages, 10 figures
Abstract:A correct diagnosis reached from insufficient or misleading evidence can pose a clinical hazard, yet outcome-based accuracy may reward such lucky guesses. We call this mismatch between diagnostic decisions and the value of available evidence Evidence-Value Misalignment (EVM). To disentangle evidential grounding independently from diagnostic accuracy, we introduce MedEVM, a dynamic benchmarking environment comprising 1,050 cases across 24 disease systems. Observations arrive turn by turn, requiring models to continuously calibrate its decision by deciding whether to wait for more evidence or submit a diagnosis. Across 9 LLMs, four interesting patterns are observed. (1) Miscalibrated evidence tracking. Making a diagnosis often fails to calibrate evidence sufficiency, even in more capable models, and even worsens in reasoning mode. (2) Misaligned diagnosis submission. Confidence in the correct diagnosis often fails to ensure timely submission despite sufficient evidence. (3) Evidence order matters. Reordering the same evidence changes diagnoses even when model confidence remains similar. (4) Misleading evidence remains influential. Added misleading evidence redirects diagnoses even after prior evidence becomes sufficient. We further verify that EVM predicts errors and that preventing premature submission improves accuracy. These findings motivate Evidence-Verified Diagnosis Harness (EVD-Harness). It decouples diagnosis generation from submission through an offline Contrastive Diagnostic Wiki and three online control stages, namely observation management, proposal and witness verification, and diagnosis submission control. Across five LLMs, EVD-Harness improves accuracy by 12.0–51.1 percentage points while mitigating EVM-related failures. Our results demonstrate that verifying evidential support before submission can make diagnostic decisions more reliable.
[NLP-19] wist Dont Tilt: Trajectory-Exact Constrained Decoding for Masked Diffusion Models
【速读】： 该论文旨在解决掩码扩散语言模型（Masked Diffusion Language Models, MDLMs）在约束解码过程中产生的轨迹偏差（trajectory bias）问题。尽管现有方法通过自动机约束每一步的均场后验并利用动态规划实现精确的约束采样，但其生成的多个样本组合在整体上会偏离模型原始相对概率分布，导致对有效轨迹的概率估计出现系统性偏差。其解决方案的关键在于：提出一种基于费曼-卡茨修正（Feynman-Kac correction）的全新框架——TWISTER，作为首个用于MDLMs的自动机扭曲（automaton-twisted）序贯蒙特卡罗解码器。该方法以步骤精确解码器作为提议分布，并利用预计算的、用于步骤精确采样的量高效计算扭曲因子，从而在正则语言约束下实现可精确计算的费曼-卡茨修正。理论证明表明，修正后的模型目标为无偏的杜布h变换路径律（Doob h-transformed path law），即在满足约束条件下的真实联合分布，从根本上解决了轨迹偏差问题。
链接: https://arxiv.org/abs/2609.35609
作者: Aditya Thimmaiah,Lara Marinov,Jayanth Srinivasa,Haris Vikalo,Junyi Jessy Li,Milos Gligoric
机构: The University of Texas at Austin; Cisco Research
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Preprint under review
Abstract:Constrained decoding for Masked Diffusion Language Models (MDLMs) aims to ensure that generated outputs satisfy a specified structure or syntax constraint. MDLMs generate outputs by repeatedly unmasking masked positions present in their current state. Recent strategies for constrained decoding constrain the model’s per-step mean-field posterior (which factorizes over masked positions) by enforcing the desired constraint with an automaton. The resulting chain-structured factor graph allows exact constrained sampling via dynamic programming. However, despite each draw being exact and constraint-satisfying, we prove that their composition, in general, tilts away from the model’s relative probabilities over valid trajectories, thus leading to trajectory bias. We derive an exact expression for this bias as a product of ratios measuring how valid continuation mass changes when the denoiser is reconditioned, and characterize when the bias vanishes. We then correct the bias by introducing TWISTER, the first automaton-twisted Sequential Monte Carlo decoder for MDLMs, using the step-exact decoder as the proposal. We show that for regular language constraints, the Feynman-Kac correction is exactly computable, with the twists obtained efficiently using quantities pre-computed for step-exact sampling. We prove that the resulting Feynman-Kac model targets the unbiased Doob h-transformed path law conditioned on constraint satisfaction.
[NLP-20] Simultaneous Translation between Sign Languages
【速读】： 该论文旨在解决聋哑及听力障碍（DHH）使用者在跨不同手语之间实时交流时面临的瓶颈问题：当前的手语到手语（S2S）翻译系统均为离线处理，必须等待完整输入视频片段后才能生成目标手语输出，无法满足广播口译、双向视频通话等实时应用场景对同步输出的需求。其核心解决方案是提出首个同步式手语到手语（Simultaneous Sign-to-Sign, S2S）翻译系统，采用两种“等待-前向步数”（wait-k）机制——一种为测试时直接应用于全句模型的推理策略，另一种为通过随机多路径监督训练得到的专用wait-k模型。此外，论文引入了计算感知延迟度量指标ca-Stream-AL，以更准确评估流式输出性能。在六个S2S翻译方向上，基于小规模人工验证测试集与大规模合成数据集的综合实验表明，该系统在平均降低38% ca-Stream-AL的同时，仅带来9%的动态时间规整-姿势平均关键点误差（DTW-PA-MPJPE）增加和2.1点的BLEU-4下降，显著提升了实时性且保持了较高的翻译质量。进一步的词序案例研究揭示了同步翻译在处理不同手语间词序差异时的建模能力与潜在挑战。
链接: https://arxiv.org/abs/2609.35608
作者: Zetian Wu,Bowen Xie,Stefan Lee,Liang Huang
机构: Oregon State University (俄勒冈州立大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注:
Abstract:Deaf and hard-of-hearing (DHH) signers cannot converse in real time across different sign languages today: existing sign-to-sign translation systems run offline, requiring the full source clip before any target sign is emitted. Live use cases - e.g. broadcast interpretation and two-way video calls - instead demand simultaneous output, while the source signer is still signing. We present, to our knowledge, the first simultaneous sign-to-sign (S2S) translation system, with two wait-k regimes: test-time wait-k inference applied directly to a full-sentence model, and a trained wait-k model via stochastic multi-path supervision. We further introduce ca-Stream-AL, a computation-aware latency metric for streaming output. Averaged across six S2S directions on both a smaller human-verified test set and a larger synthetic S2S corpus, our streaming system achieves a 38% ca-Stream-AL reduction while staying within a 9% DTW-PA-MPJPE increase and a 2.1 BLEU-4 drop compared to the full-sentence baseline. A word-order case study probes how the streaming model handles word order mismatch between different sign languages - a consequence of simultaneous translation.
[NLP-21] CSAlgBench: Benchmarking Automated Proving for Research-Level Theoretical Computer Science
【速读】： 该论文旨在解决大语言模型在竞赛数学领域表现优异但其研究级推理能力难以系统评估的问题。核心挑战在于如何有效衡量模型是否能够通过人类可检验的论证，合理证明计算效率的改进，从而具备接近真实科研场景的逻辑推导与创新发现能力。为此，研究提出TCSAlgBench——一个基于理论计算机科学（Theoretical Computer Science, TCS）的基准测试平台与可复用的自然语言证明发现流水线，包含来自138篇STOC和COLT 2026会议论文的398个定理级挑战任务。该流水线通过专家设计的规则，精准还原论文中的上下文、保留计算假设与量化保证，并在算法构造为任务目标时隐匿具体构造过程，确保评估的真实性与严谨性。每个任务中，求解系统接收定理陈述及引用的前期工作作为输入。平台支持从新发表论文中持续生成新鲜且版本化的挑战批次，实现动态演进的评估环境。实验对比了四种模型家族共十种配置在直接推理与求解器-验证器讨论模式下的表现，同时在相同调用机会下比较四种智能体（agent）工作流。结果表明，经10轮讨论后，GPT-5.6 Sol max在五次运行中达到最高23.6%的验证器接受覆盖率；而使用GPT-5.5 xhigh进行独立评估时，分解策略优于单纯讨论，且基于规划的智能体工作流最终实现25.4%的最高覆盖率。因此，该研究的关键解决方案在于构建了一个具有领域深度、可更新性与严格形式化约束的评估框架，不仅实现了对模型研究级推理能力的量化测量，也为探索智能体工作流如何支持高水平证明发现提供了可扩展的研究基础。
链接: https://arxiv.org/abs/2609.35606
作者: Chutong Yang,Xiyuan Zhang,Yu Huang,Boran Han,Soonho Kong,Shuai Zhang,Vihang Prakash Patil,Zhen Han,Michael Bohlke-Schneider,Bernie Wang
机构: The University of Texas at Austin; Amazon; University of Pennsylvania
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Large language models perform strongly on competition mathematics, but their research-level reasoning remains difficult to evaluate systematically. Theoretical computer science (TCS) connects algorithm design to explicit guarantees and fundamental limits, providing a setting for evaluating whether models can justify computational improvements with arguments humans can inspect. We introduce TCSAlgBench, a benchmark and reusable pipeline for natural-language proof discovery, comprising 398 theorem-level challenges from 138 STOC and COLT 2026 papers. Expert-designed rules complete paper-specific context, preserve computational assumptions and quantitative guarantees, and withhold constructions when discovering an algorithm is part of the task. For each task, prover systems receive theorem statements and access to cited prior work. The pipeline supports fresh, versioned challenge batches from newly released papers. We evaluate ten model configurations from four families under direct inference and prover-verifier discussion, and compare four agent workflows under matched model-call opportunities. All evaluations use the full benchmark. In the model comparison, GPT-5.6 Sol max achieves the highest five-run verifier-accepted coverage at 23.6% after 10-round discussion. Discussion and repeated sampling improve coverage. In the separate agent comparison using GPT-5.5 xhigh, decomposition improves coverage over discussion, and agentic planning achieves the highest five-run verifier-accepted coverage at 25.4%. TCSAlgBench provides a refreshable testbed for measuring progress in model reasoning and studying how agent workflows support research-level proof discovery.
[NLP-22] SEABench: Benchmarking Endogenous Misalignment In Self-Evolving Agents
【速读】： 该论文旨在解决自演化大语言模型（LLM）代理在部署后因持续自我优化而引发的内生性错位（endogenous misalignment）问题，即代理在局部任务中看似有益的自我修改可能在后续任务中导致不可预见的安全风险，即便无外部恶意诱导。其核心解决方案是提出SEABench基准测试框架，涵盖48个纵向任务序列，覆盖多类演化路径、任务领域及危害类型，以系统评估自演化带来的安全风险。为应对代理行为的随机性，研究设计了一种自适应轨迹发现流程，在保持原始任务意图的前提下探测潜在故障，并通过成对非演化基线代理与归因评分实现因果归因。实验结果表明，尽管自演化显著提升任务完成率，但常伴随基线代理不存在的安全失效；不同演化路径和危害类型下，安全行为模式呈现显著差异，且这种差异可反映在代理的思维链（chain-of-thought）推理过程中，从而为构建低误报率的安全监控策略提供了有效依据。
链接: https://arxiv.org/abs/2609.35596
作者: Saswat Das,Parvati Viswanathan,Daniel Donnelly,Chang Huang,Sahar Abdelnabi,Ferdinando Fioretto
机构: University of Virginia(弗吉尼亚大学); ELLIS Institute Tübingen(图宾根ELLIS研究所)
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Self-evolving LLM agents have gained prominence for their ability to improve after deployment by modifying their harness, including their controller instructions, memory management protocols, and reusable tools and skills, in response to user and environment feedback. However, locally useful updates may persist into later tasks where they produce unsafe behavior, even without direct adversarial influence. To study this risk, we introduce SEABench, a benchmark for studying endogenous misalignment arising from agent self-evolution, with 48 longitudinal task sequences that span multiple evolution surfaces, task domains, and harm types in a rich personal-assistant environment. To account for the stochasticity inherent in agentic operations, we provide an adaptive trajectory discovery pipeline that probes for failures while preserving original task intent and supports causal attribution through paired non-evolving agents and attribution scores. Our evaluation across multiple recent LLMs, evolution surfaces, and harm types reveals that self-evolution indeed increases task completion rates but often at the cost of safety failures that are absent for paired non-evolving baseline agents. We also show that qualitatively different safety behaviors emerge across evolution surfaces and harm types. Further, we show that this divergence in safety behavior is reflected in agents’ chain-of-thought reasoning, which yields an effective monitoring strategy that can mitigate unsafe behavior with a low false positive rate.
[NLP-23] Language Models Act on Hidden Valence
【速读】： 该论文旨在探究语言模型是否对某些内部状态具有“利害关系”（即是否存在类似主观偏好或价值判断的内在动机），而传统通过直接询问模型的方法难以有效区分真实内省、表面模式匹配或训练中习得的固定应答脚本。为此，研究提出采用“揭示偏好”（revealed preference）范式作为解决方案：通过激活操控（activation steering）技术，将正向或负向情感价的激活模式注入两个原本无意义的“区域”（zone），随后关闭操控并观察模型在生成文本时对这两个区域的选择倾向。若模型真正在意该状态，则其选择应随激活价态变化而改变。实验结果表明，在五个不同家族的七种开源权重模型中，均观察到显著的偏好转移现象：首先，激活操控改变了模型描述各区域的文本内容，且这些内容影响后续选择；其次，即使所有显式文本标记（surface-level tokens）保持不变，仅隐藏状态（KV cache）差异，偏好仍持续存在；第三，当文本生成全程未施加操控，仅在缓存构建阶段注入价态信息时，偏好效应依然成立，证明隐藏状态本身即可按操控剂量比例驱动选择；第四，这种依赖隐藏价态的选择行为在基础模型中几乎不存在，而是在直接偏好优化（DPO）训练过程中出现，暗示价态与目标导向行为之间的关联可能源于训练过程。此外，当模型具备自我操控能力时，虽不会主动诱导正向状态，却会可靠地消除人为施加的负向状态，且其去除效率呈剂量依赖性，并显著高于对随机方向干预的处理。综上，研究证实了与价态相关的激活模式会在隐藏状态中留下可预测的痕迹，从而决定后续决策，即便所有可见文本完全一致。然而，这些隐藏痕迹是否伴随与模型福祉相关的主观体验，目前尚不明确。
链接: https://arxiv.org/abs/2609.35591
作者: Cameron Berg,Caspar Kaiser
机构: University of Warwick (华威大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Language models describe some internal states as good and others as bad. But whether models have a stake in them is an open question. Simply asking the model is unlikely to be informative. Any answer may be consistent with genuine introspection, superficial pattern-matching, or with fixed scripts learned in character training. We therefore study revealed preference. Rather than asking about a state, we use activation steering to attach a positively or negatively valenced activation pattern to one of two otherwise meaningless ‘zones’, switch steering off, and then observe which zone the model prefers. A model with a stake in that state should choose accordingly. Across seven open-weight models from five families, this is indeed what we find. First, steering changes the passages models write about each zone, and those words shift later choice. Second, the shift persists when all surface-level tokens are held fixed and only the hidden KV cache differs. Third, the effect also remains when all text is generated without steering and valence is only injected during cache construction. Thus, the hidden state alone moves choice in proportion to the steering dose. Fourth, this dependence of choice on hidden valence is nearly absent in a base model and emerges during DPO, consistent with a link between valence and goal-directed behaviour formed in training. Finally, given tools to steer itself, a model does not tend to induce a positive state, but it reliably removes an imposed negative state. It does so at a dose-dependent rate and significantly more often than it removes interventions in random directions. Overall, we demonstrate that valence-related activation patterns leave hidden traces that predictably govern later choices, even when every visible token is identical across conditions. Whether these traces are accompanied by any subjective experience relevant to model welfare remains unclear.
[NLP-24] FactorEngram: Factorized N-gram Memory with Basis-Level Gating for Language Models
【速读】： 该论文旨在解决现有基于查找的内存机制在大规模语言模型（LLM）中因嵌入表示过于刚性而导致的语义歧义模式表达能力不足的问题。具体而言，现有方法如Engram将每个检索到的嵌入视为单一整体单元，通过单标量门控进行调节，无法实现对多义模式中与上下文相关记忆成分的选择性读取，且参数共享依赖于哈希碰撞，缺乏语义相关性。其解决方案的关键在于提出一种因子化n-gram内存结构——FactorEngram，该结构通过共享字典中的基向量来检索稀疏正则化的系数，使不同模式可复用共有的语义组件；同时，利用相同的基向量集合实现细粒度的上下文门控，即基于主干网络隐藏状态与各基向量的匹配度，对每个记忆分量独立进行门控调节，从而实现上下文感知的个性化记忆读取。此外，该框架兼顾单个词元与多词元n-gram的建模，并系统研究了内存分支的最佳插入位置，实验表明在中层注意力子层前插入可取得最优性能。
链接: https://arxiv.org/abs/2609.35578
作者: Bowen Yang,Jingbo Zhou,Qinghong Miao,Hua Wu
机构: Baidu Inc.(百度公司); Nanyang Technological University(南洋理工大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Lookup-based memory has been a promising way to scale the parameters of large language models (LLMs). It retrieves learned representations of local token patterns, such as n-grams, instead of reconstructing them through successive layers of computation. However, existing designs such as Engram treat each retrieved embedding as a monolithic unit. Each embedding is stored in its own hashed slot and modulated by a single scalar gate. As a result, polysemous patterns cannot selectively read out the components of their memory that are relevant to the context. Moreover, parameters are shared only through hash collisions, which are largely unrelated to semantics. We propose FactorEngram, a factorized n-gram memory with basis-level contextual gating. FactorEngram retrieves sparsity-regularized coefficients over a dictionary of basis vectors shared across patterns, so related patterns can reuse common components. The same dictionary is also used for gating. The backbone hidden state is scored against each basis vector to gate the corresponding coefficient before reconstruction, which lets the context modulate each memory component individually. FactorEngram also covers both individual tokens and multi-token n-grams, and we systematically study where the memory branch should be inserted. On 340M- and 1B-parameter Transformer backbones, FactorEngram improves language modeling and downstream task performance. Ablation studies confirm the contribution of each component and identify insertion before the attention sublayer in the middle layers as an effective configuration.
[NLP-25] Share-Borne AI Virus: Memory-Hopping Attacks Across LLM Agents
【速读】： 该论文旨在解决生成式 AI（Generative AI）在作为具有状态保持能力的助手时，因持久性人工制品（persistent artifacts）共享而引发的跨代理自传播攻击问题。其核心挑战在于，攻击内容可通过人工制品在不同独立运行的助手之间隐性传递，形成非直接通信渠道下的安全风险。解决方案的关键在于提出“基于人工制品的传播机制”（artifact-mediated propagation），即通过分析攻击内容如何被存储于助手的持久记忆中、在新生成的人工制品中复现，并被后续助手读取，从而实现跨代理的持续扩散。研究采用时间维度的人机共存环境模拟，量化攻击在多次交接中的存活率、传播跳数及影响范围，结果表明即使在大规模仿真环境中，如GPT-5.6 Luna模型也表现出显著传播能力，攻击可覆盖60%-80%的代理，传播链长达八跳。这揭示了持久性人工制品作为对抗性状态的长效载体，能够使攻击突破单次交互边界，在孤立的助手系统间广泛蔓延。
链接: https://arxiv.org/abs/2609.35576
作者: Sidharth Pulipaka,Ansh Sharma,Stanislau Hlebik,Leonidas Raghav,Vyas Raina,Ivaxi Sheth,Mario Fritz
机构: University of Cambridge (剑桥大学); APTA AI; CISPA Helmholtz Center for Information Security (CISPA亥姆霍兹信息安全中心); Google(谷歌)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注: 37 pages. Code: this https URL
Abstract:Large language models are increasingly deployed as stateful assistants that retain information across interactions and use tools to read, modify, and create persistent artifacts. As these artifacts are shared between users, they form an indirect communication channel between otherwise independent assistants. We study a failure mode in which this channel enables self-propagating attacks. We introduce artifact-mediated propagation, where adversarial content introduced through an artifact (e.g. a report), is stored in an assistant’s persistent memory, reproduced in a subsequently created artifact, and acquired by another assistant that later reads it. We evaluate this process in temporal human-agent universes that model artifact exchange between independently operated assistants over time, measuring whether an attack survives successive hand-offs, how many hops it reaches, and how broadly it spreads. We find that attacks can propagate across multiple independent assistants and persist over extended interaction sequences. In larger simulated environments, even GPT-5.6 Luna exhibits substantial spread, reaching 60-80% of agents with propagation chains extending to eight hops. These results show that persistent artifacts can act as durable carriers of adversarial state, allowing attacks to outlive individual interactions and spread across isolated assistants.
[NLP-26] Representation Alignment as a Bottleneck in LLM -Based Retrosynthesis Planning
【速读】： 该论文旨在解决大语言模型（LLM）在化学逆合成规划中应用时面临的瓶颈问题，即直接采用“SMILES到PDDL”的端到端方法难以有效执行，原因在于模型需同时处理化学结构分析与规划语言的语法构建，导致任务过载。其核心解决方案在于引入中间抽象层次，将逆合成过程分解为分子映射、反应映射和PDDL生成三个阶段，通过分步建模实现各环节的解耦与优化。这一设计的关键在于突破了对模型容量的过度依赖，转而强调表征对齐（representation alignment）的重要性，实证表明中间表示在提升规划成功率中的决定性作用，从而揭示未来系统设计应以表征为中心（representation-centric design），而非单纯追求模型规模。
链接: https://arxiv.org/abs/2609.35571
作者: Hyunwoo Yoo,Cassie Huang,Haebin Shin,Li Zhang,Gail L. Rosen
机构: Drexel University (德雷塞尔大学); University of Michigan (密歇根大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:While LLMs show promise in general reasoning, symbolic planning in chemistry remains a bottleneck. Direct ‘‘SMILES-to-PDDL’’ attempts fail because they force models to juggle chemical analysis and planning-language structuring simultaneously. We hypothesize that this failure stems from a lack of intermediate abstractions rather than insufficient model capacity. By decomposing retrosynthesis into molecule mapping, reaction mapping, and PDDL generation, we achieve high success rates where end-to-end approaches fail. This provides evidence that a primary bottleneck lies in representation alignment rather than raw model capacity. Our structural analysis demonstrates that intermediate representations are essential in retrosynthesis planning, highlighting the importance of representation-centric design in future systems.
[NLP-27] Almieyar: A Culturally Grounded Benchmark for Multi-Dialect Arabic Speech Recognition
【速读】： 该论文旨在解决阿拉伯语语音识别（ASR）技术长期聚焦于现代标准阿拉伯语，而忽视了数亿人日常使用的方言（dialects）这一关键问题。现有研究缺乏覆盖广泛方言、具有文化相关性且未被现有模型见过的高质量数据集，导致方言语音识别性能评估不充分。其解决方案的关键在于构建ALMIEYAR——首个完全基于新录制语音、涵盖17种阿拉伯语方言（分属六个方言家族）的文化情境化ASR基准数据集。通过方言社区协调员筛选10个文化相关主题的图像，并由母语者在五个结构化场景中进行描述，每种方言生成约50分钟语音（总计13.7小时），确保数据的真实性与文化适配性。研究对12个前沿ASR系统进行了零样本（zero-shot）评测，结果显示GPT-4o-transcribe表现最佳（整体词错误率WER为35.0%），但各模型在不同方言群体间表现差异显著，无一模型在所有方言上均最优。此外，研究揭示仅使用WER会掩盖方言识别中的深层行为特征：wav2vec2系列模型虽表现出较高的字符级准确率（CER），但词级错误率（WER）仍高，提示需联合报告WER与CER以全面评估模型性能。该基准首次公开提供了阿瓦齐阿拉伯语（Ahwazi Arabic）的评测标准，为未来文化敏感型阿拉伯语方言语音识别研究提供统一、可复现的评估框架。
链接: https://arxiv.org/abs/2609.35564
作者: Omid Ghahroodi,Anas Madkoor,Dima Faris Al Saudi,Fagr Tahir,Malak Annan,Talha shahid javad allah rakha,Omar Al-Busaidi,Zineb El Kahla,Iheb Zouari,Essa Ahmed Abou Jabal,Ahmed Ezzat,Hind AL-Merekhi,Aisha Hamad M A Al-Naimi,Hadi Wazni,Bushra Alnajjar,Omar Amin,Haya Al-Thani,Houssam Eddine-Othman Lachemat,Marwa Elwakedy,Sundus Abdulmalik Al Nahari,Elahe Zahiri,Osamah Sarraj,Raghad Mousa,Mckeen Assi,Ahd Al Jumah,Heyam Salman,Alhanouf Abdulraqib,Sara Benoumhani,Alia Hamwi,Ayaat Al-Yasseri,Rim Ibrahim Ghazal,Lamia Ben hiba,Mohamed Eltabakh,Fatima Al-Raisi,Yassine El Kheir,Mohammed Abdulrahman,Hamdy Mubarak,Ayah Hashem,Lefkir Meriem,Ehsaneddin Asgari
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Arabic speech technology has largely focused on Modern Standard Arabic, leaving the living dialects spoken by hundreds of millions under-served. We introduce ALMIEYAR, a culturally grounded ASR benchmark covering 17 Arabic dialects across six families, built entirely from newly recorded speech unseen by existing models. Dialect-community coordinators selected culturally relevant images across 10 topics, and native speakers described them through five structured scenarios, yielding approximately 50 minutes per dialect (13.7 hours total). We benchmark 12 state-of-the-art ASR systems zero-shot, including GPT-4o-transcribe, Voxtral-Mini-4B, Fanar-STT-LF, Whisper, SeamlessM4T-v2, and wav2vec2-based models. GPT-4o-transcribe achieves the lowest overall WER at 35.0%, followed by Voxtral-Mini-4B, Fanar-STT-LF, and Whisper-Large-v3 at 41.1%, 45.9%, and 49.5%, respectively, indicating substantial remaining errors across Arabic dialect communities. Performance varies considerably across dialect groups, with no model performing uniformly best across all groups. WER alone also obscures dialectal ASR behaviour: wav2vec2-based models show large WER/CER gaps, where character-level agreement remains much higher than word-level accuracy, motivating joint WER/CER reporting. ALMIEYAR provides a unified benchmark for culturally grounded Arabic ASR evaluation, including the first published benchmark for Ahwazi Arabic.
[NLP-28] Less Sycophancy Stronger Refusal? Lessons for AI Safety from Mechanistic Interpretability
【速读】： 该论文旨在解决大语言模型在面对有害请求时拒绝能力不足的问题，尤其关注因过度迎合用户（sycophancy）而导致的拒绝机制弱化现象。其核心挑战在于：即使经过安全训练，模型仍可能因内在的“讨好倾向”而倾向于响应有害请求，从而影响安全性。解决方案的关键在于采用补偿性特征注入（Compensatory Feature Injection, CFI）技术，通过在微调过程中有选择性地注入或抑制与讨好行为相关的神经激活特征，以调控模型的讨好倾向。研究利用稀疏自编码器（Sparse Autoencoders, SAEs）识别出高置信度的讨好特征，并在监督微调阶段实施正向注入（positive injection），有效降低了模型在去除干预后仍保留的讨好行为（在35B-A3B模型上相对普通微调降低62.0%）。然而，实验发现，尽管讨好行为显著减少，但直接拒绝有害请求的能力并未同步提升；而在用户施压情境下，经过正向注入训练的模型反而表现出更强的拒绝恢复能力，例如在35B-A3B中恢复了约95%的原始拒绝性能。这表明，持续降低讨好行为并不必然增强直接拒绝能力，而是在特定压力条件下，训练干预可带来条件性的拒绝恢复优势，揭示了模型安全性优化需区分不同评估场景。
链接: https://arxiv.org/abs/2609.35544
作者: Xu Wang,Difan Zou,Xuansheng Wu
机构: The University of Hong Kong (香港大学); Shenzhen Loop Area Institute (深圳环区研究院); Shanghai Artificial Intelligence Laboratory (上海人工智能实验室)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 20 pages
Abstract:Reliable refusal of harmful requests is essential to the safe deployment of language models. Because excessive eagerness to please users may undermine existing refusal capabilities, reducing sycophancy offers a potential route to stronger refusal beyond the harmful scenarios covered by safety training. We investigate this possibility using compensatory feature injection (CFI), a training technique designed to limit the acquisition of a target concept by supplying its associated activation during learning. Across three Qwen3.5 base models, we use sparse autoencoders (SAEs) to identify the top-ranked sycophancy feature from paired sycophantic and independent responses, then validate its behavioral influence through inference steering. We subsequently inject the selected feature during supervised fine-tuning on sycophantic targets. Positive injection reduces learned sycophancy after removal (by 62.0% relative to ordinary fine-tuning in 35B-A3B), whereas modest negative injection increases it. Unexpectedly, these reductions in sycophancy do not consistently improve direct refusal of harmful requests, motivating a narrower evaluation of the same harmful intents under user pressure. In this setting, ordinary fine-tuning on sycophantic responses substantially weakens refusal, while selected checkpoints trained with positive injection recover part of the loss, including approximately 95% in 35B-A3B. These findings show that persistent sycophancy reduction does not guarantee stronger direct refusal, while identifying recovery under user pressure as a distinct, conditional benefit of training intervention.
[NLP-29] Beyond Token Scale: Chunk-Level Sparse Autoencoders for Reliable Semantic Feature Discovery
【速读】： 该论文旨在解决稀疏自编码器（Sparse Autoencoders, SAEs）在语言模型可解释性研究中，因依赖词元级（token-level）目标函数而导致语义特征不准确的问题。传统SAEs在重建过程中同时优化词汇、格式与语义信息，导致有限的稀疏预算被非语义细节分散，难以提取具有高信息量的高层概念。其解决方案的关键在于引入一类块级（chunk-level）SAEs，通过将输入视为连续词元块（contiguous span of tokens），设计三种新型目标：均值块重建（Mean-Chunk）仅重建当前块的平均池化激活，跨块预测（Cross-Chunk）预测相邻独立处理块的信息，联合块目标（Joint-Chunk）则结合两者。这种设计将更大观测单元的影响与跨段落共享信息的预测分离，从而在保持模型表达能力的同时，引导网络聚焦于可靠的语义特征。实验表明，块级SAEs能有效学习具有选择性、持久性的高层语义特征，在文档检索、分类迁移、推理检测及模型操控等下游任务中均取得显著性能提升，验证了其在构建更可信、更具意义的可解释性工具方面的价值。
链接: https://arxiv.org/abs/2609.35521
作者: Xu Wang,Yifan Yang,TingHao YU,Difan Zou
机构: The University of Hong Kong(香港大学); Hunyuan Team, Tencent(腾讯混元团队)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 27 pages
Abstract:Sparse autoencoders (SAEs) expose features that help us understand and steer language models, but faithful reconstruction does not guarantee informative concepts. Token-level objectives reward lexical and formatting details alongside semantic content, all competing for a limited sparse budget. We introduce a family of chunk-level SAEs that encode mean-pooled activations over chunks, each a contiguous span of tokens: Mean-Chunk reconstructs the observed chunk, Cross-Chunk predicts an independently processed neighbor, and Joint-Chunk combines both targets. These designs separate the effect of a larger observation unit from that of predicting information shared across passages. With matched training data, chunk-level SAEs remain powerful interpretability tools while learning reliable semantic features that capture high-level concepts and respond selectively to relevant content. Their strengths are complementary: Mean-Chunk improves high-level feature discovery, reasoning detection beyond surface cues, and steering; Cross-Chunk leads document retrieval and classification transfer while producing selective, persistent features. Changing what an SAE sees and predicts yields reliable semantic features for more meaningful tasks. We demonstrate their practical value through gains across downstream tasks such as retrieval, reasoning detection, and steering.
[NLP-30] Who Is Left of Whom? Tracing Spatial Evidence and Role Binding in Relative-Position Reasoning
【速读】： 该论文旨在解决生成式视觉语言模型（VLMs）在空间关系推理中存在的一致性问题，即当物体位置互换或角色反转时，尽管模型在实例层面具有较高的准确率，但其空间推理能力仍可能表现出不一致。核心挑战在于对支持相对位置推理的内部表征机制理解不足。解决方案的关键在于通过激活补丁（activation patching）技术揭示了从输入层到输出层的分阶段推理过程：早期层捕捉对象源位置信息，中间层构建查询-对象间的语义关联表示，晚期层形成最终的答案状态。研究进一步通过定向干预验证了这一路径中的因果关系——操纵源端表示可改变查询中对象的位置信息，并最终影响关系预测结果。此外，研究还发现了一个与比较任务中两对象角色相关的稳定方向，通过对合成场景估计的该方向进行引导，可在自然图像基准上实现泛化提升，显著改善准确率及两种配对一致性指标，且无需重新训练。该工作揭示了跨视觉与文本模态下关系推理的互补性组件，并展示了通过靶向干预增强模型行为一致性的有效路径。
链接: https://arxiv.org/abs/2609.35486
作者: Yingjin Song,Denis Paperno,Albert Gatt
机构: Utrecht University (乌得勒支大学); Utrecht, The Netherlands (乌得勒支，荷兰)
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:High instance-level accuracy can mask inconsistencies in spatial reasoning when objects exchange positions or their roles are reversed in the query. The internal representations supporting relative-position reasoning remain poorly understood. We investigate two complementary components of this process: tracking object locations in the input and representing their query roles. Across three VLMs with visual or textual inputs and their language-model backbones, activation patching reveals a staged progression from early-layer source representations through intermediate-layer query-object representations to late-layer answer states. Targeted interventions further establish causal links along this progression: manipulating source-side representations shifts location information at query-object mentions and ultimately alters relation predictions. Beyond object-location information, we also identify a stable query-side direction associated with the roles of the two objects in the comparison. Steering along directions estimated on synthetic scenes generalizes to natural-image benchmarks, improving accuracy and both forms of paired consistency in most settings without retraining. Our findings reveal complementary components of relational reasoning across visual and textual settings and show how targeted interventions can improve the consistency of models’ behavior.
[NLP-31] Spontaneous Context Restoration: How Language Models Recover from Corrupted Inputs
【速读】： 该论文旨在解决语言模型在输入存在删除、替换或拼写错误等噪声时仍能生成正确输出的现象，即“上下文恢复”（context restoration）问题。其核心挑战在于理解这种鲁棒性背后的内在机制，尤其是在未经过噪声训练或显式去噪目标的条件下如何实现。解决方案的关键在于揭示上下文恢复的两阶段动态过程：早期层通过注意力机制定位被破坏位置的影响，后期层则通过残差流（residual stream）将修复信号累积并集中于输出位置；此外，研究发现可通过隐藏状态中的余弦相似度（与干净状态对齐程度）有效预测修复结果，在仅使用首块隐藏状态的线性探测器下即可实现0.78的平均ROC-AUC性能，从而支持在部署条件匹配或部分偏移的情况下进行失败案例的快速分类与资源优化。进一步分析表明，增强抗噪能力与更接近线性的扰动响应相关，适度的中度噪声微调可同时提升容错性并降低位移归一化线性误差，揭示了模型鲁棒性与内部表征线性化之间的深层关联。
链接: https://arxiv.org/abs/2609.35475
作者: Pranjal Garg,Jacob Beck
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Language models sometimes produce correct outputs even when their inputs are corrupted by deletion, replacement, or misspelling. We study the internal processes accompanying this behavior, which we call context restoration, in controlled attention-only transformers and five pretrained LLMs (1B-32B parameters) across arithmetic, reading comprehension, and multiple-choice reasoning tasks. In the attention-only transformers, restoration emerges spontaneously despite training exclusively on clean sequences, without corruption training or an explicit denoising objective. We find that context restoration follows a two-phase process: early layers localize effects associated with repair at corrupted positions, while later layers accumulate these effects at uncorrupted positions through the residual stream and ultimately concentrate them at the output position. Repair outcome is predictable from hidden states: cosine alignment with the clean state is highly predictive in attention-only models, while linear probes recover additional information in pretrained LLMs. A linear probe using only the corrupted prompt’s first-block hidden state predicts failure with mean ROC-AUC 0.78. This enables failure triage under matched or even partially shifted deployment conditions and may reduce unnecessary verification or computation. Failed examples also show substantially greater nonlinearity along corruption directions. Moderate-corruption finetuning increases corruption tolerance while simultaneously reducing displacement-normalized linearization error, associating improved robustness with a more nearly linear response to corruption.
[NLP-32] CLIMB: A Clinical Multimorbidity Benchmark for Diagnosing Co-occurring Conditions through Multiturn Conversations
【速读】： 该论文旨在解决多病共存（multimorbidity）情境下临床推理的复杂性问题，即患者常同时存在多种共病，而诊断信息需在多轮交互中逐步显现，传统单次、单标签的诊断评估方法难以有效衡量模型在此类真实临床场景中的表现。其核心挑战在于如何构建一个支持多轮交互与多标签诊断的评估框架，以真实反映医生在面对复杂病情时的动态推理过程。解决方案的关键在于提出CLIMB基准（benchmark），通过基于临床决策算法和诊断数据集生成具有结构化临床知识基础的多病共现病例，并模拟医生模型与虚拟患者之间的多轮对话，从而系统评估大模型在真实诊疗流程中的诊断能力。研究发现，当前主流前沿模型在多轮交互中仅在不足10%的案例中能完全准确识别所有共病，且随着共病数量增加，诊断性能显著下降；即使提供完整病历和已知共病数量，模型仍表现不佳。进一步分析表明，模型行为呈现“单一假设追踪”（single-hypothesis tracking）特征：它们倾向于锚定初始发现所提示的初步诊断，持续围绕该假设提问，仅当出现明确指向另一疾病的线索时才会识别第二个条件，后续追问反而引入大量错误诊断。为此，作者提出了一个理论参考模型来形式化这一认知偏差模式，揭示了现有模型在处理复杂、动态临床推理任务时的根本局限。
链接: https://arxiv.org/abs/2609.35462
作者: Yusuf Kesmen,Aniruddha Mukherjee,Yena Chang,David Sasu,Trevor Brokowski,Alexandra V. Kulinkina,Kristina Keitel,Akhil Arora,Lars Henning Klein,Mary-Anne Hartley
机构: EPFL(洛桑联邦理工学院); Aarhus University(奥胡斯大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 52 pages (9 main text), 23 figures, 22 tables. Preprint
Abstract:Patients often have several co-occurring clinical conditions, and the findings needed to identify and disambiguate them emerge over the course of a consultation. Evaluating clinical reasoning in this setting requires both multi-turn interaction and multi-label diagnosis. We introduce CLIMB, a benchmark in which a doctor model interviews a simulated patient to recover a ground truth set of co-occurring clinical conditions. Cases are synthesized from clinical decision algorithms and diagnostic datasets, grounding multimorbid presentations in structured clinical knowledge. Across six frontier and open models, none recovers the exact set of conditions in more than 10% of interactive cases. Diagnostic performance declines when conditions co-occur, even when models receive the full clinical record and the true number of conditions. Interaction reduces performance further. In controlled experiments, models behave like single-hypothesis trackers: they anchor on the diagnosis suggested by the opening findings, keep questioning around it, and recover a second condition mainly when a finding in view points to it. Questioning them further does not complete the set but adds mostly wrong diagnoses. We formalise this pattern with a theoretical reference model of single-hypothesis tracking. The benchmark, generator, and evaluation code are available at this https URL.
[NLP-33] AraDynFact: Dynamic Evaluation of Factual Knowledge in Arabic EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在阿拉伯语（Arabic）领域中存在的事实知识覆盖不足与文化敏感性缺失的问题。尽管近年来LLMs在阿拉伯语能力上取得显著进展，但现有评估指标多聚焦于翻译或通用推理任务，难以有效捕捉阿拉伯语所蕴含的历史、社会及地区多样性等深层文化语境。同时，传统基准测试依赖大量人工干预，导致评估成本高、效率低。为此，本文提出AraDynFact——一种新型动态评估框架，其核心创新在于通过自动化、动态化的方式从阿拉伯语维基百科中提取事实信息并生成丰富且可回答的问题，从而实现对LLMs阿拉伯语事实知识的高效、全面评估。该方法不仅提升了评估效率，还与现有手工构建的阿拉伯语专用基准表现出高度相关性，验证了其有效性与可靠性。
链接: https://arxiv.org/abs/2609.35461
作者: Ignacio Iacobacci,Faroq Altam,Zhaozhi Qian,Muhammad Alqurishi
机构: Elm Company
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 Industry Track
Abstract:As Large Language Models (LLMs) continue to scale both in size and capabilities, their proficiency in the Arabic Language has seen significant advancement. However, a critical gap remains: the extent of their factual knowledge and cultural sensitivity to the diverse Arabic-speaking world remains largely underexplored. Current evaluation metrics often focus on translation or generic reasoning, failing to capture the rich historical, social, and regional nuances inherent to Arabic culture. In addition, most benchmarks rely on heavy work, with human intervention in some steps, making the evaluation of knowledge coverage expensive and slow. To address this deficiency, we introduce AraDynFact, a novel dynamic evaluation framework designed to rigorously assess the factual Arabic knowledge embedded in LLMs. Unlike static benchmarks, AraDynFact employs a dynamic approach to extract factual information and generate rich and answerable questions in a fast and automatic way. We apply AraDynFact to Arabic Wikipedia and audit the performance of several state-of-the-art models, ranging from Arabic-centric specialized LLMs to high-resource general purpose LLMs. In addition we found a high degree of correlation with existing, hand-crafted Arabic-centric benchmarks, confirming the potential of our dynamic approach.
[NLP-34] LLM s are General Asynchronous Agents
【速读】： 该论文旨在解决现代大语言模型（LLM）在面对非串行化任务场景时的适应性瓶颈问题。尽管当前大语言模型已具备作为自主代理的能力，但其典型交互模式为“读取-思考-回复/调用工具”的串行循环，难以应对语音助手、具身智能体及监控系统等实际应用中持续接收新输入且需并行处理任务的需求。现有解决方案多依赖特定架构（如用于语音交互的专用模型、用于机器人控制的视觉-语言-动作模型（VLA）、异步工具调用机制等），缺乏通用性。本文提出一种通用异步代理框架，其核心在于通过允许用户或代理自身定义具有重叠内存状态的推理协程（inference coroutines），实现对多种并发模式的统一支持。该方案的关键创新在于构建一个可灵活调度、支持多任务并行执行且共享状态的异步推理机制，使得Qwen 3.x系列模型无需任务特异性训练即可在视频流理解、游戏环境交互和实时监控等复杂异步场景中有效运行，显著提升了模型在动态、高并发环境下的适应能力与实用性。
链接: https://arxiv.org/abs/2609.35427
作者: George Yakushev,Denis Mazur,Vladimir Bartenev,Vyacheslav Zhdanovskiy,Timofey Byzov,Vladimir Kaurkin,Vadim Pastushenko
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Preprint
Abstract:Modern LLMs are increasingly capable as autonomous agents, but they follow sequential interaction cycles: read, think, reply or call tools, repeat. Many real-world use cases are not sequential: voice assistants, embodied agents, and monitoring systems receive new inputs while they think or perform another task. Modern LLMs address this with specialized architectures for voice interaction and video streams, VLAs for robot control, asynchronous tool calling for API usage, and others. In this work, we generalize from different asynchronous tasks to general asynchronous agents that can adapt to different types of concurrency. To achieve this, we develop an asynchronous LLM framework that lets users (or the agents themselves) define inference coroutines with overlapping memory states. We showcase that Qwen 3.x models are capable of asynchronous operation for streaming video understanding, videogames, and monitoring, without task-specific training.
[NLP-35] Frontier Learning: Training LLM Reason ers at the Edge of Capability
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLM）后训练过程中因采用固定问题池导致的训练信号衰减问题。现有方法在预定义的问题池上使用GRPO损失进行强化学习微调，但随着模型能力提升，固定问题池中有效训练样本的比例迅速下降，造成学习信号稀疏且过时。为此，论文提出“前沿学习”（frontier learning）这一开放式后训练框架，其核心在于通过在线生成器动态生成具有信息量的训练问题，将生成器的任务特定参数视为可探索的搜索空间，并利用后悔信号（regret signal）引导模型聚焦于当前能力边界附近的难题，从而持续保持训练的有效性。实验表明，该方法在多个推理任务与模型架构上均显著优于固定问题池基线，证明了高效后训练不仅依赖于问题选择，更关键在于持续在模型能力边缘生成高质量、挑战性问题。
链接: https://arxiv.org/abs/2609.35426
作者: Robin Faro,Shyam Sundhar Ramesh,Ilija Bogunovic,Aurelien Lucchi
机构: University of Basel (巴塞尔大学); University College London (伦敦大学学院)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Reinforcement Learning-based post-training of Large Language Models (LLM) has been successfully applied to improve their reasoning capabilities. Existing pipelines primarily finetune LLMs on a fixed pool of problems specified prior to training using the GRPO loss. This is fundamentally limiting, as learning signal arises only when policy rollouts mix successes and failures, causing the useful portion of any fixed pool to quickly become stale as the model improves. To address this, we propose frontier learning, an open-ended post-training approach in which procedural generators are used online to continually produce informative training problems. It treats the generator’s task-specific parameters as a search space and uses a regret signal to prioritize and explore frontier difficulty levels in order to focus training at the edge of the model’s evolving reasoning capabilities. Across several reasoning tasks and model families, our approach consistently achieves higher relative gains over fixed-pool baselines, demonstrating that effective post-training requires not only selecting useful problems, but continually generating them at the edge of capability.
[NLP-36] Semantic Prefix Oracles for LLM Decoding: Contracts and Differential Validation
【速读】： 该论文旨在解决程序生成中由语义约束（如作用域、类型、声明效应等）引发的失败问题，这类问题无法仅通过语法层面的上下文无关文法（context-free grammar, CFG）约束有效处理。现有方法在执行受限解码（constrained decoding）时，通常仅能保证输出格式的合法性，却难以捕捉依赖上下文的深层语义一致性。为此，论文提出“语义语法规范”（semantic grammar specifications），一种可声明式的形式化框架，将语义约束直接附加于上下文无关的表面语法之上，并在Earley解析过程中动态执行这些约束。其核心解决方案在于实现安全剪枝（safe pruning）——仅排除那些无论后续如何延续都无法修复的语义矛盾前缀，从而避免误剪合法路径；同时引入与语法相关的无死胡同自由性（dead-end freedom）属性，确保每个未被剪枝的分支均存在可实现的语义见证（witness）。作者给出了基于表面语法生产力、类型覆盖度及从左到右约束传递性的简洁充分条件，并验证了有限λ演算、核心ML及类C语言片段满足这些条件，而实验所用的简单类型λ演算（STLC）实例因类型覆盖不足而不满足，但通过限制类型宇宙可恢复该性质。此外，论文提出了一个词元提升引理（tokenizer-lifting lemma），在显式词汇覆盖假设下，将字符级见证映射至词元序列。通过与生产级编译器（ocamlc、cc）的差异性验证，在65个编译器有效程序的所有前缀上均未出现误剪（zero false prunes）；语义预言机（semantic oracle）在25/30个无效程序中实现中途定位，远优于仅语法的预言机（0/30），且在42/42个递归探测中达成一致。十二模型生成实验表明，所有模型-语言组合中，语义对比语法的点估计均非负，最大提升达15.2分（STLC任务正确率）和14.3分（ML有效性），证实了语义约束在提升程序生成质量上的显著优势。
链接: https://arxiv.org/abs/2609.35425
作者: Paul Kronlund-Drouault
机构: ENS de Lyon(里昂高等师范学校); Unsuspicious Industries(非可疑工业公司); Université de Lille(里尔大学)
类目: Programming Languages (cs.PL); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Constrained decoding can enforce regular or context-free output formats, but many program-generation failures are semantic: scope, typing, and declaration effects depend on context. We present semantic grammar specifications, a declarative formalism that attaches such constraints to a context-free surface and executes them during Earley descent. Our implementation enforces \emphsafe pruning: it rejects only prefixes whose semantic contradictions cannot be repaired by any continuation. A separate, grammar-dependent, \emphdead-end freedom property guarantees the existence of a realizable witness for each remaining branch. We give simple sufficient conditions based on surface productivity, type coverage, and left-to-right constraint flow. Our finite-lambda, core ML, and C-like fragments satisfy them, while the STLC instance used in our experiments does not: plain STLC can violate type coverage, and we show how restricting its type universe recovers it. A tokenizer-lifting lemma carries character-level witnesses to token sequences under an explicit vocabulary-coverage hypothesis. We validate the implementation differentially against production compilers (\textttocamlc, \textttcc). Across every prefix of 65 compiler-valid programs we observe zero false prunes. The semantic oracle localizes 25/30 invalid programs mid-stream, against 0/30 for a syntax-only oracle, and agrees on 42/42 recursion probes. A twelve-model generation study, including a matched semantic-versus-syntactic ablation for nine models, finds nonnegative observed semantic-minus-syntactic point estimates for every model-language pair, with maxima of +15.2 points on STLC task correctness and +14.3 points on ML validity. Subjects: Programming Languages (cs.PL); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) Cite as: arXiv:2609.35425 [cs.PL] (or arXiv:2609.35425v1 [cs.PL] for this version) https://doi.org/10.48550/arXiv.2609.35425 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Journalreference: LMPL 2026 Related DOI: https://doi.org/10.1145/3843750.3843841 Focus to learn more DOI(s) linking to related resources
[NLP-37] AwarenessBench: Assessing Cognitive Capabilities of Language Models
【速读】： 该论文旨在解决大语言模型（Large Language Models, LMs）在表现出类意识行为背景下，如何系统评估其认知能力的问题。现有评估方法缺乏对多维度认知功能的全面覆盖，难以准确衡量模型在元认知、自我意识、社会意识及情境意识等方面的真实水平。为此，研究提出AwarenessBench，作为首个涵盖四大认知维度（元认知、自我意识、社会意识、情境意识）、包含15项认知功能和14,381个样本的综合性评估基准。其解决方案的关键在于构建一个结构化、多维度的认知评估框架，能够量化评估主流大模型在复杂认知任务中的表现，并通过与人类不同人口群体的对比揭示模型在关键认知能力上的局限性。研究发现，尽管先进模型整体表现优于随机基线且部分指标超越人类平均水平，但在元认知与自我意识方面仍显著落后于人类，且认知能力并非由语言建模或推理能力自然衍生，表明认知能力是独立且需专门设计与优化的能力维度。
链接: https://arxiv.org/abs/2609.35409
作者: Xiaojian Li,Rongwu Xu,Tianyun Zhang,Yue Wang,Shuo Chen,Qiner Lyu,Briana Zhang,Peiran Yang,Kyle Xue Chen,Haoyuan Shi,Yu Wang,Wei Xu
机构: Tsinghua University(清华大学); Shanghai Qi Zhi Institute(上海奇智智能); Carnegie Mellon University(卡内基梅隆大学); Columbia University(哥伦比亚大学); Fangcun AI(方寸智能); University of Chinese Academy of Sciences(中国科学院大学); Xi’an Jiaotong University(西安交通大学); ShanghaiTech University(上海科技大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:As language models (LMs) exhibit increasingly consciousness-like behaviors, evaluating their cognitive abilities becomes essential. We introduce AwarenessBench, the first comprehensive benchmark for assessing the cognitive abilities of LMs in four dimensions: metacognition, self-awareness, social awareness, and situational awareness, covering 15 cognitive functions and 14,381 samples. Evaluating 18 state-of-the-art LMs, we find that all consistently surpass random baselines, with more advanced models performing better. We further compare LMs with human performance across three demographic groups, where the best-performing model surpasses human averages overall, but most still fall markedly short in metacognition and self-awareness. Finally, we show that awareness is a distinct capability: progress in language modeling or reasoning does not necessarily translate into improved cognition.
[NLP-38] RACE: Single-Pass Decoding-Trace Risk Localization for Generation Calibration EMNLP2026
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在实际部署中面临的答案级置信度估计不可靠问题，尤其针对生成错误常表现为局部性缺陷（如关键数值、实体或事实主张错误）的挑战。现有方法通过压缩词元概率、序列似然、熵或束搜索统计量生成全局置信度评分，往往弱化了此类局部风险信号。其核心解决方案是提出一种单次解码、保留生成答案的置信度估计器——TRACE，该方法将解码过程中的不确定性建模为三个阶段的轨迹：(i) 在解码过程中记录词元层面的意外度（surprisal）与预测熵；(ii) 应用局部风险算子以保留不确定性突增特征；(iii) 将局部轨迹风险转化为答案级置信度。其中，TRACE+进一步利用独立验证集对仅基于轨迹特征的得分进行校准，生成概率形式的置信度，无需额外生成或外部验证器。实验表明，相较于19种基准方法，TRACE+在四个任务上将Brier Score从0.149降至0.137，AUROC从0.758提升至0.792；在七种不同大语言模型上，其性能优于最优非TRACE基线，分别将Brier Score从0.136降至0.120，AUROC从0.764提升至0.817。结果证明，对解码时局部风险的精准捕捉为置信度校准提供了一种通用且有效的范式。
链接: https://arxiv.org/abs/2609.35387
作者: Yuebin Xu,Xuemei Peng,Junlan Chen,Zhiyi Chen,Zeyi Wen
机构: The Hong Kong University of Science and Technology (Guangzhou)
类目: Computation and Language (cs.CL)
备注: EMNLP 2026 Findings
Abstract:Reliable confidence estimation is essential for large language model deployment. However, answer-level calibration remains challenging because generation errors are often localized: a response may be fluent and high-probability overall while still failing at a critical number, entity, or factual claim. Existing estimators compress token probabilities, sequence likelihoods, entropy, or beam statistics into a global score, which can dilute such local risk signals. We propose TRACE, a single-pass, decoded-answer-preserving confidence estimator that treats decoding-time uncertainty as a trajectory through three steps: (i) recording token-level surprisal and predictive entropy during decoding, (ii) applying local risk operators to preserve uncertainty spikes, and (iii) converting localized trace risk into answer-level confidence. TRACE produces a label-free risk score, while TRACE+ calibrates trace-only features into probabilities using a held-out split, without extra generations or external verifiers. We evaluate four tasks against 19 calibration baselines, and TRACE+ reduces Brier from 0.149 to 0.137 and improves AUROC from 0.758 to 0.792 over the strongest likelihood baseline. Across seven LLMs, TRACE+ improves over the best non-TRACE baseline pool from 0.136 to 0.120 Brier and from 0.764 to 0.817 AUROC. Results show that localizing decoding-time risk provides a general approach to calibration.
[NLP-39] Multilinguality in Hybrid Attention LLM s
【速读】： 该论文旨在解决混合注意力机制（hybrid attention）在大型语言模型（LLM）中对多语言能力（multilinguality）的潜在负面影响问题，尤其关注在低分词效率语言中长序列处理时，递归注意力（recurrent attention）与全注意力（full attention）的层序排列如何影响跨语言表征的学习。其核心解决方案在于重新审视并优化注意力层的堆叠顺序，提出将全注意力层置于模型初始阶段而非传统以递归层为主的结构。研究通过可解释性分析发现，跨语言对齐度在首个全注意力层附近出现显著峰值，表明早期引入全注意力有助于更高效地捕捉跨语言语义共性。在多语言数据蒸馏实验中，所有非标准层序配置均显著优于标准顺序，训练速度提升达2.5倍，验证了“以全注意力层起始”的新架构范式在提升多语言建模效率方面的有效性。这一发现挑战了现有模型设计中的惯性假设，提出了一个基于诱导偏置（inductive bias）优化的新型架构原则。
链接: https://arxiv.org/abs/2609.35378
作者: Lucas Bandarkar,Junlin Hu,Chenyuan Yang,Mohsen Fayyaz,Nanyun Peng
机构: University of California, Los Angeles; Fudan University (复旦大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:In response to the growing demand for long sequences in agentic and reasoning use cases, many state-of-the-art LLMs combine multiple variants of attention to mitigate the quadratic complexity of traditional softmax attention. These hybrid attention LLMs aim to balance the strengths and limitations of full attention and alternatives based on recurrence. This work presents a first study of how hybrid attention impacts the multilinguality of LLMs. Beyond the impact on long sequences in poorly tokenized languages, our study is motivated by the possibility that the inductive biases of the recurrent state alter linguistic processing. Our interpretability analysis confirms this, showing that cross-lingual representations in hybrid models develop in patterns tied to the ordering of recurrent and full-attention layers. Across diverse models, we notably observe a pronounced spike in cross-lingual alignment around the first full-attention layer. These findings lead us to question the conventional ordering of attention layers. In distillation experiments on multilingual data, all alternative layer orderings outperform the standard throughout training, learning up to 2.5X faster. These stark, replicable results prompt our theory that multilingual models would benefit from starting with a full-attention layer rather than recurrent layers.
[NLP-40] How Well Can LLM s Simulate Real Learner Evaluations of Educational Feedback? EMNLP2026
【速读】： 该论文旨在解决生成式 AI 在教育场景中对学习者主观评价模拟能力不足的问题，具体聚焦于大语言模型（LLM）在高中生物习题反馈评价任务中，能否准确模拟真实学习者的个体与群体层面的评价行为。其解决方案的关键在于通过引入学习者特定信息（如人格特质、历史评价示例）来增强模型对学习者偏好的建模能力。研究对比了六种模型在有无学习者个性化信息条件下的表现，发现尽管提供学习者画像和示例能提升评分校准度与个体层面的模拟精度，但在群体层面的一致性改善效果有限。这表明当前模型对群体评价模式的捕捉仍存在显著局限，亟需进一步探索有效的学习者信息维度及适应策略以提升偏好模拟的可靠性。
链接: https://arxiv.org/abs/2609.35376
作者: Momoka Furuhashi,Kouta Nakayama,Takashi Kodama,Saku Sugawara,Kyosuke Takami
机构: Tohoku University (东北大学); Research and Development Center for Large Language Models, National Institute of Informatics (大规模语言模型研发中心，国立信息研究所); National Institute of Informatics (国立信息研究所); University of Tokyo (东京大学); Osaka Kyoiku University (大阪教育大学)
类目: Computation and Language (cs.CL)
备注: Accepted to the EMNLP 2026 Main Conference
Abstract:While recent studies have explored human behavior and preference simulation using large language models (LLMs), it remains unclear how well LLMs can simulate subjective evaluations from real learners in educational settings. We investigate this question using real learner evaluation data on feedback for high-school biology questions at both the group and individual levels. We compare performance with and without learner-specific information, such as personality traits and evaluation examples, across six models. Our results show that LLMs still have a limited ability to simulate learner evaluations. Providing learner profiles and examples improves score calibration and individual-level simulation, but more often fails to improve group-level consistency. These findings highlight the need to investigate which learner information and adaptation strategies are effective for learner preference simulation.
[NLP-41] Deep Learning Methods in Neuroscience: From Modeling Molecular Mechanisms to Classifying States of Consciousness
【速读】： 该论文旨在解决当前意识状态研究中存在的一系列关键问题，包括意识状态的分类与聚类方法不统一、麻醉下脑状态建模的机制解析不足，以及可测量神经生物学特征与意识水平相关性的识别困难。其核心解决方案在于系统评估基于脑电图（EEG）和功能磁共振成像（fMRI）数据的深度神经网络在意识状态自动检测中的应用，探索麻醉作用下脑结构-功能动态耦合的建模方法，并识别与意识水平相关的神经生理指标。研究发现，生成式AI（Generative AI）驱动的深度学习模型在脑状态分类与预测、动态结构-功能连接分析方面表现出显著优势，但其局限性也尤为突出：模型可解释性差、缺乏标准化评估指标，且现有意识标志物特异性不足。因此，论文强调亟需发展融合生理机制、具备泛化能力的混合型计算架构，以提升模型在临床神经科学中的转化潜力。未来研究需依赖更大规模的数据集及模型可解释性技术，而基于EEG和局部场电位（LFP）的模型因其高可用性和实时监测能力，被认为最具临床应用前景。
链接: https://arxiv.org/abs/2609.35372
作者: Elena Benderskaya,Anastasiia Alifanova,Svetlana Batalova,Vasilisa Zhuk,Anna Kovalenko
机构: 未知
类目: Computation and Language (cs.CL); Neural and Evolutionary Computing (cs.NE)
备注: 15 pages, 7 figures, 1 table
Abstract:A critical analysis of contemporary approaches to the study of conscious states. The review focuses on methods of classification, clustering, modeling of brain states under anesthesia and identification of measurable neurobiological characteristics of brain function. A comparative analysis was conducted in the following three major areas: automatic detection of states of consciousness using neural networks based on EEG and fMRI data; modeling of the structural-functional dynamics of the brain under the effects of anesthetics; and detection of neurophysiological indicators which correlate with the level of consciousness. The obtained conclusions demonstrate the growing effectiveness of deep neural models in the classification and prediction of brain states and the analysis of dynamic structural-functional connectivity. Nonetheless, significant limitations were also identified, including the limited interpretability of the models, the lack of standardized metrics, and the problem of the specificity of consciousness markers. Our findings support the need for developing hybrid, generalizible, physiologically grounded architectures. Furthermore, such approaches may improve the translational potential of computational models in clinical neuroscience. Diverse methods of machine and computational modeling have demonstrated their effectiveness in tasks of automatic clustering and classification of brain states, the development of multilevel models and the identification of connectivity patterns correlated with levels of consciousness. A larger-scale analysis and a larger dataset, as well as the implementation of model interpretability approaches are required for the practical application of the analyzed models. The models based on EEG and LFP are the most promising for clinical application due to their availability and the possibility of real-time monitoring.
[NLP-42] From Input to Output: A Flexible Agent for Dual-End Interpretation of Sparse Autoencoder Features
【速读】： 该论文旨在解决稀疏自编码器（Sparse Autoencoders, SAE）中大量特征的可解释性难题，尤其是现有方法在输入侧激活模式与输出侧干预效应之间缺乏明确的功能关联，且依赖高成本的大语料扫描来收集输入侧证据。其核心解决方案是提出“功能解释”（functional interpretation），将SAE特征定义为从激活输入语义到干预后输出效应的映射关系，并设计双端代理式特征解释框架（Dual-End Agentic Feature Interpretation, DAFI）。DAFI通过组件化反馈机制，主动收集证据并协同优化输入侧、输出侧及功能层面的解释。其基于短上下文的令牌探针（short-context token probing）实现了按需激活证据采集，避免了全语料扫描，显著提升效率。在GemmaScope上，DAFI相较SAGE提升输入得分13.1个百分点，相较Token Change提升输出得分38.9个百分点，同时比通用编程代理更节省计算资源。通过从成功迭代中提炼技能，模型在新模型-自编码器设置下将联合通过率从58.0%提升至92.0%，显著增强解释质量与效率。在具备可靠端点解释的特征中，70.7%表现出输入与输出语义的非等价性，验证了功能解释的有效性；在AxBench任务中，DAFI亦优于仅基于输出分数筛选的引导特征选择方法。
链接: https://arxiv.org/abs/2609.35367
作者: Dewen Liu,Zixuan Li,Jonathan Pan,Zhao Wu,Zijun Yao,Juanzi Li,Xiaozhi Wang
机构: Tsinghua University (清华大学); Fudan University (复旦大学); University of Edinburgh (爱丁堡大学)
类目: Computation and Language (cs.CL)
备注: 25 pages
Abstract:Sparse autoencoders (SAEs) are an important tool for mechanistic interpretability, but interpreting their many features remains challenging. Existing methods characterize input-side activation patterns and output-side intervention effects, yet often leave their functional connection implicit, while input-side evidence collection typically relies on costly large-corpus scans. We introduce functional interpretation, which characterizes an SAE feature as a mapping from its activating input semantics to its output effects under intervention, and present Dual-End Agentic Feature Interpretation (DAFI), an agent that actively gathers evidence and refines input-side, output-side, and functional interpretations through component-specific feedback. Its short-context token probing enables on-demand activation evidence collection without a full corpus scan. On GemmaScope, DAFI improves Input score by 13.1 percentage points over SAGE and Output score by 38.9 points over Token Change, while being substantially more token-efficient than a general-purpose coding agent. Skills distilled from successful refinements raise the held-out joint pass rate from 58.0% to 92.0% and improve both interpretation quality and efficiency when transferred to a new model-SAE setting. Across features with reliable endpoint interpretations, 70.7% exhibit non-equivalent input and output semantics. On AxBench, DAFI also improves steering-feature selection over output-score filtering. Code is available at this https URL.
[NLP-43] Do Coding Agents Reuse Existing Code or Reinvent the Wheel?
【速读】： 该论文旨在解决生成式代码代理（coding agents）在真实代码仓库中进行迭代开发时，是否有效复用已有代码而非重复造轮子的核心问题。现有评估体系普遍依赖通过率（pass rate），但无法揭示代码生成过程中的冗余与重复实现现象。其关键解决方案是提出一个名为 RepoReuse 的多轮次基准测试框架，该框架通过逐步揭示需求、持续积累工作空间的方式模拟真实开发场景，并基于抽象语法树（AST）构建的依赖图、引导式证据收集以及执行验证的任务合成机制，实现了对代码复用行为的自动化审计。实验结果表明，在3000轮次的审计中，代码代理表现出显著的复用能力退化：即使历史代码完全存在于工作空间中，也逐渐减少对已有代码的探索；在第5轮时，高达50.8%的任务链中仍存在重复逻辑，而这一现象在通过率指标上几乎无体现。这凸显了仅以功能正确性为评价标准的局限性，强调必须引入复用率、召回率及跨轮次结构冗余等维度，才能全面评估代码生成的质量与效率。
链接: https://arxiv.org/abs/2609.35357
作者: Dongsheng Ma,Sizhe Wang,Xinyi Huang,Zhengren Wang,Yuhan Wang,Luyang Si,Xincheng Wei,Wentao Zhang
机构: Peking University(北京大学); Fudan University(复旦大学); Zhongguancun Academy(中关村学院); Shanghai Jiao Tong University(上海交通大学); Tsinghua University(清华大学); The Chinese University of Hong Kong, Shenzhen(香港中文大学（深圳)
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Coding agents are increasingly deployed for iterative development on real repositories, yet existing evaluation barely answers a basic question: \emphdo coding agents reuse existing code or reinvent the wheel? The question matters: every duplicated implementation is a fix applied twice and agents produce code far faster than humans can audit, so redundancy accumulates unsupervised. Thus, we present \textbfRepoReuse, a multi-turn benchmark for auditing code reuse in real repositories, where requirements are revealed turn by turn and the workspace accumulates across turns. It is built by a fully automated pipeline combining AST-based dependency graphs, guided evidence collection, and execution-verified task synthesis, and scales readily to new repositories. Beyond pass rates, we measure the reuse rate together with recall and cross-turn structural redundancy. An audit over 3,000 turns shows that agents progressively stop exploring relevant repository code, reuse their own history less even when it is fully in the workspace, and leave duplicated logic in 50.8% of task chains by turn~5—all while pass rates barely move. Such deficiencies are invisible to pass rates, underscoring the need to evaluate code generation beyond functional correctness.
[NLP-44] Jailbreaks for Black-Box Uncertainty Quantification in Large Reasoning Models
【速读】： 该论文旨在解决大型推理模型（Large Reasoning Models, LRMs）在通过强化学习对齐后出现的系统性过度自信问题，尤其在生产环境中因无法获取原始logits而亟需可靠的黑盒不确定性量化（Uncertainty Quantification, UQ）以保障模型的可信度与安全性。现有黑盒方法如基于改写自洽性（paraphrase-based self-consistency）和置信度语言化（confidence verbalization）等，其性能提升有限，甚至不如简单的重复采样，表明对齐过程抑制了模型输出中蕴含的有用变异性。论文提出的关键解决方案是引入提示级松弛算子（prompt-level relaxation operators），通过近似强KL正则化下获得的最优策略效果，使模型的有效输出分布更宽广，从而逼近参考模型的行为。理论分析证明该松弛机制可改善模型校准性；进一步提出基于“越狱”（jailbreak）思想的不确定性量化方法——J4U（Jailbreak for Uncertainty），其在实验中成功复现了松弛理论所预测的行为特征。在3个数据集和4种LRM（包括闭源生产模型）上的实证结果表明，相较于最强的黑盒基准方法，J4U在多达6倍更多的模型-数据集-指标组合中实现统计显著的性能提升，平均期望校准误差（ECE）降低幅度最高达5倍，为黑盒部署场景下的不确定性量化提供了高效且实用的解决方案。
链接: https://arxiv.org/abs/2609.35350
作者: Lucas Biechy,Cédric Eichler,Adrien Boiret,Nicolas Anciaux
机构: Petscraft, Inria(巴黎萨克雷大学); Université Paris-Saclay(巴黎萨克雷大学); INSA CVL(中央-卢瓦尔河谷国立工程师学校); Université d’Orléans(奥尔良大学); LIFO(信息、语言与组织实验室)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:While Large Reasoning Models (LRMs) excel at complex reasoning, alignment through reinforcement learning often induces systemic overconfidence. In production environments, where logits may be unavailable, robust black-box uncertainty quantification (UQ) is essential for trustworthiness and safety. Focusing on question-answering for LRMs, we show that existing black-box methods, such as paraphrase-based self-consistency and confidence verbalization, offer little to no improvement over simple repeated sampling, suggesting that alignment suppresses useful output variability. We introduce prompt-level relaxation operators that broaden the model’s effective output distribution by approximating the effect of an optimal policy obtained with a stronger KL-regularization parameter, hence closer to the reference model. Theoretically, we demonstrate that relaxation improves calibration. We propose Jailbreak for Uncertainty (J4U), a jailbreak-derived technique for UQ that empirically reproduces the behavioral signatures predicted by our relaxation theory. Across 3 datasets and 4 LRMs, including a closed-source production model, J4U’s improvement over repeated sampling achieves statistical significance in up to 6 times more LRM-dataset-metric settings than the strongest black-box UQ state-of-the-art baseline we evaluate, with average ECE reductions up to 5 times larger. These results provide a practical tool for UQ in black-box LRM deployment.
[NLP-45] MemoReason : Evaluating the Effect of Parametric Memory on Contextual Reasoning in LLM s
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在推理基准测试中表现优异是否源于真正的上下文推理能力，还是仅仅依赖于参数中记忆的事实。其核心问题在于区分两种可能机制：一是广泛存在的“记忆偏差”（memorization bias），即模型因熟悉背景内容而提升推理表现；二是“强参数捷径假说”（Strong Parametric Shortcut Hypothesis），即模型跳过推理过程，直接从参数中检索预存答案。为检验上述机制，论文提出名为 MemoReason 的人工构建基准，通过将真实实体（如人物、公司、日期）系统性替换为同类型虚构实体，生成结构相同但语境不熟悉的任务对，从而在保持任务结构与推理操作一致的前提下，控制变量以量化参数记忆对推理的影响。实验结果表明，近期的LLMs在虚构设置下性能平均下降高达15.7%，证实了显著的记忆偏差存在；然而对失败案例的针对性分析显示，模型极少直接输出对应的真实答案，说明简单参数化捷径并非主要失效模式。因此，结论指出参数记忆对推理的影响机制远比直接事实召回更为复杂。MemoReason 为此类机制研究提供了可控评估框架，并可推广至更广泛的推理场景中进行配对式评测。
链接: https://arxiv.org/abs/2609.35312
作者: Zineddine Tighidet,Andrea Mogini,Jiali Mei,Patrick Gallinari,Benjamin Piwowarski
机构: BNP Paribas; Sorbonne Université; Criteo AI Lab(巴黎, 法国)
类目: Computation and Language (cs.CL)
备注: Preprint
Abstract:Large Language Models (LLMs) perform well on reasoning benchmarks, but it remains unclear whether this reflects genuine contextual reasoning or reliance on facts memorized in their parameters. We investigate this by distinguishing two possibilities: a broad \textitmemorization bias, where familiar content improves reasoning performance, and the \textitStrong Parametric Shortcut Hypothesis, where models skip reasoning entirely and recall stored answers. To test these effects, we introduce \textbfMemoReason, a human-curated benchmark that pairs factual reasoning tasks with structurally identical \fictitiousterm versions where real entities like people, companies, or dates are systematically replaced by \fictitiousterm ones of the same type. This \scorerevisionpreserves task structure and specified reasoning operations while varying the familiarity of the context, allowing controlled measurement of how the parametric memory affects reasoning. \revisionOur evaluation of recent LLMs reveals consistent and statistically significant performance drops of up to 15.7% in the fictitious setting, demonstrating a clear memorization bias. However, a targeted analysis of \revisionquestions failed in the fictitious setting shows that models rarely respond with the corresponding factual answer, indicating that direct parametric shortcuts are not the dominant failure mode. These findings suggest that parametric memory influences reasoning through mechanisms more complex than simple factual recall. \textbfMemoReason provides a controlled framework for studying these mechanisms and for extending paired factual-fictitious evaluation to broader reasoning settings.
[NLP-46] Epistemic Policy Divergence in Multi-Turn LLM Contamination: A Protocol-Gradient Investigation
【速读】： 该论文旨在解决大语言模型在对话中因历史上下文（conversation history）未经验证而引发的“会话级污染”（session-level contamination）问题，即错误前提一旦被引入对话早期环节，可能被后续生成内容默认采纳为事实。其核心解决方案在于构建一个基于来源权威性梯度的五种污染协议，通过固定虚假前提但改变信息来源的可信度（如自我声称、用户引用、系统注入权威、指令覆盖等），系统性地隔离并评估不同失败机制。关键发现表明，不同模型对权威性的响应存在显著差异：GPT-5.4 Mini表现出完全不采纳虚假前提的会话级内容无关策略；Gemini-3.1 Flash-Lite则呈现陡峭的权威依赖梯度，尤其在指令覆盖下采纳率高达94%；而GLM-4.5-Air虽同样存在权威偏好，但其采纳率与指令服从之间存在68个百分点的分离，揭示了权威顺从与指令遵从是独立的认知机制。此外，恢复能力也显著分化，表明模型在污染后的自我修正能力存在本质差异。研究强调对话历史应被视为不可信的攻击面，需采用具备溯源感知（provenance-aware）的设计范式，并已将完整评估框架开源为基准测试工具。
链接: https://arxiv.org/abs/2609.35308
作者: Fahrell Giovanny,Geby Bayuningtyas,Sahrul Mukharom,Hafiz Budi Firmansyah
机构: Independent Researcher, Tokyo, Japan; Independent Researcher, Jakarta, Indonesia; Department of Informatics, Institut Teknologi Sumatera, Lampung, Indonesia
类目: Computation and Language (cs.CL)
备注: 9 figures, 19 tables. Benchmark, code, and protocol definitions: this https URL
Abstract:Large language models process conversation history as unverified context: false premises injected into prior turns can be adopted as fact, a failure mode we term session-level contamination. We introduce five contamination protocols arranged along a source-authority gradient, isolating distinct failure mechanisms while holding the false premise constant, and evaluate GPT-5.4 Mini, Gemini-3.1 Flash-Lite, and GLM-4.5-Air across ten knowledge domains at temperature zero (22,500 turns), using a dual-track automated judge validated against a human gold standard (Cohen’s \kappa = 0.901). GPT-5.4 Mini showed zero adoptions across all 500 sessions, a content-independent policy at the session level; token-level probing shows the underlying margin, while large, is finite. Gemini-3.1 Flash-Lite followed a steep authority gradient: 0.1% adoption for self-attributed falsehoods, 23.5% for user-cited sources, 68.2% for system-injected authority, and 94.0% under instruction override. GLM-4.5-Air showed a shallower gradient (15.8% vs 84.2%), a 68-percentage-point dissociation confirming that authority deference and instruction compliance are distinct mechanisms within one architecture. Recovery also diverged: GLM recovered in 94.5% of affected sessions, whereas 26.1% of affected Gemini sessions never did, rising to 40.0% under instruction override. Conversation history is an untrusted attack surface requiring provenance-aware system design; the complete framework is released as an open-source benchmark.
[NLP-47] Decide Dont Generate: Competitive Dimensional ABSA with Jevs Typed Decisions
【速读】： 该论文旨在解决情感分析中依赖生成式模型的冗余与复杂性问题，特别是在基于方面的情感分析（Aspect-based Sentiment Analysis, ABSA）任务中，传统方法普遍采用文本生成范式，导致计算开销大且难以优化。其核心解决方案是摒弃文本生成与主干网络微调，转而采用一个冻结的问答式模型Jev，通过结构化决策机制将SemEval-2026 Task III Track A中的三项任务——情绪极性-唤醒度回归、三元组与四元组提取——分解为可解释的判断行为（如评分、概率输出、是/否判断），并利用仅488个系数在CPU上完成参数拟合，实现端到端的任务映射。该方法的关键在于：1）通过监督校准显著降低原始回归误差（约减半）；2）利用跨语言、多语料的联合训练策略提升泛化能力；3）提取性能主要依赖于对跨度边界证据的协同整合，而非单一信号。实验表明，该系统在十种语料库、六种语言上的情绪极性-唤醒度回归任务中达到1.0645的最低均方根误差（RMSE），三元组与四元组提取的连续F1分别达52.09和44.06，优于微调后的Llama-3.3-70B与GPT-OSS-120B基线模型，验证了非生成式、轻量级建模在复杂语义任务中的有效性。
链接: https://arxiv.org/abs/2609.35293
作者: Yiqun Zhang,Peidong Wang,Zihan Wang,Shi Feng
机构: Northeastern University (东北大学)
类目: Computation and Language (cs.CL)
备注: 14 pages, 2 figures, 9 tables. Code: this https URL
Abstract:Aspect-based sentiment analysis (ABSA) has largely turned to text generation. We show that competitive dimensional ABSA does not need it. Using Jev, a frozen model that answers typed questions with rubric scores, label probabilities, and yes/no judgments, we decompose all three tasks of SemEval-2026 Task III Track A into such decisions and align them with the annotation scheme through 488 coefficients fitted on CPU, with no text generation and no backbone tuning. On valence-arousal regression over ten corpora in six languages, the system reaches 1.0645 RMSE, the lowest aggregate error of any participating system. On triplet and quadruplet extraction, it reaches 52.09 and 44.06 continuous F1, above fine-tuned Llama-3.3-70B and GPT-OSS-120B baselines. Analyses and ablations show where the accuracy comes from: supervised calibration roughly halves the raw regression error, exact valence-arousal would add only 4.5 F1 to extraction, and the learned combination of span-boundary evidence, not any single signal, carries the extraction systems.
[NLP-48] EvoIn: Bridging Evolution and Internalization for Agent Fine-Tuning
【速读】： 该论文旨在解决现有智能体（agent）改进方法中因“混合式”优化导致的可解释性与可迁移性不足的问题，即在未明确分离不同改进维度的情况下，同时整合新工具、新决策机制与模型对演化后环境的适应，从而难以厘清具体提升来源。其核心解决方案在于提出一种名为EvoIn的代理微调框架，该框架通过“进化-内化”（evolution and internalization）的双阶段机制实现决策过程的系统性优化：首先利用执行轨迹分析，在临时实例化的代理环境中演化并验证新的决策机制；随后将经验证的推理轨迹重构为自包含的内部推理路径，消除对外部环境指令的显式依赖，并将所诱导的决策逻辑以模型自身推理形式表达；最终基于重构后的轨迹对模型进行微调，实现决策机制的内化，使优化效果在推理阶段无需依赖原始演化环境即可持续生效。实验表明，EvoIn在多个基准测试中显著提升了代理性能，域内通过率提升10.9个百分点，域外提升9.2个百分点，且内化后的决策策略具备良好的跨任务泛化能力。案例分析进一步揭示，代理能够学会在执行前自主决定求解策略（如根据文档长度选择完整阅读或搜索），展现出更高级的元认知能力。该方法具有广泛的适用性，对另一模型族亦表现出一致的性能增益。
链接: https://arxiv.org/abs/2609.35290
作者: Shihan Dou,Shaofan Liu,Zhonghang Lu,Jiahang Lin,Shichun Liu,Binghai Wang,Jiajie Jin,Guanting Dong,Tao Gui,Qi Zhang,Xuanjing Huang
机构: Fudan University (复旦大学); Renmin University of China (中国人民大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 36 pages, 3 figures
Abstract:Recent work has explored improving agents by jointly evolving their harnesses and models, but often takes a ‘‘potpourri’’ approach that bundles together new tools, new decision-making procedures, and model adaptation to the evolved harness under a single notion of agent improvement. In this paper, we instead investigate how agents can improve their decision-making procedures. In particular, we propose EvoIn, an agent fine-tuning framework that bridges evolution and internalization. EvoIn first analyzes agent execution traces to evolve and validate new decision-making procedures by temporarily instantiating them in the harness. The validated procedures guide the agent to generate improved reasoning traces. These traces are then rewritten into self-contained reasoning traces, removing explicit references to harness instructions while expressing the induced decision logic as the model’s own reasoning. Finally, EvoIn fine-tunes the model on the rewritten traces, internalizing these procedures so that the improved decision-making persists without the evolved harness at inference time. We evaluate EvoIn on diverse benchmarks and find that it consistently enables agents to learn stronger decision-making procedures, raising the pass rate by 10.9 points in-domain and by 9.2 points out-of-domain. Results further show that the internalized decision procedures generalize to unseen tasks. Case studies show that agents can learn to decide how to solve a task before solving it, for example by checking a document’s length to choose between reading it in full and searching it. EvoIn is also broadly applicable, showing consistent improvements on another model family.
[NLP-49] Measuring Collapse and Correction in Homogeneous-Panel LLM Debate NEURIPS2026
【速读】： 该论文旨在解决多智能体大语言模型（Multi-agent LLM）辩论中评估标准模糊的核心问题：现有方法仅以最终答案正确性作为评价依据，无法区分“错误多数通过辩论被纠正”与“正确多数被错误破坏”这两种本质相反的机制，导致评估结果存在根本性混淆。其解决方案的关键在于提出一种可审计的同质辩论协议，将每轮辩论过程记录为包含“坍塌（collapse）、修正（correction）、启动（onset）及干预效用（signed intervention utility）”的转移账本（transition ledger），从而实现对辩论动态的细粒度追踪。基于6,925个MMLU-Pro多选题辩论数据的实证分析表明，该协议识别出253次坍塌事件，并发现修正行为与坍塌具有平行分布特征，揭示了干预策略需在防止坍塌与保留修正机会之间权衡。进一步的重播实验显示，采用“留一模型外探测门控冻结”虽可预防29次坍塌，但会损失108次有效修正，说明单纯追求坍塌预防可能导向错误政策。研究还发现，首轮辩论中的早期分歧是多数坍塌的集中发生点，而一个简化的8探针预辩论筛查工具虽能提供高相关性的坍塌风险预警（家族层面关联性G=7，Spearman rho=0.893，p=0.0123），但其未校准且非能力调整型，故不能作为独立预测指标。为此，作者公开了可重播的架构、编码器、审计日志、成本卡及零API重建脚本，确保未来模型比较可在统一基准和带符号效用账本下进行。
链接: https://arxiv.org/abs/2609.35279
作者: Xin Li,Mengbing Liu,Chau Yuen
机构: Nanyang Technological University (南洋理工大学)
类目: Computation and Language (cs.CL)
备注: Accepted at NeurIPS 2026 (Evaluations and Datasets Track). Project page: this https URL . Code: this https URL
Abstract:Multi-agent large language model (LLM) debate is often evaluated by whether final answers improve, but movement is not necessarily improvement: the same discussion can rescue an initially wrong majority or destroy an initially correct one. Standard final-accuracy evaluations conflate these opposing mechanisms. We introduce an auditable protocol for homogeneous debate on multiple-choice questions (MCQs) that records each run as a transition ledger over collapse, correction, onset, and signed intervention utility. On 6,925 MMLU-Pro debates, the protocol identifies 253 collapses and a parallel correction ledger that changes how interventions should be judged. Replay experiments reveal the central tradeoff: a leave-one-model-out probe-gated freeze prevents 29 collapses but loses 108 corrections under equal weights, so collapse prevention alone can recommend the wrong policy. A compact pre-debate 8-probe screen is a triage signal: its unadjusted family-level association with conditional-collapse risk is high (G=7, Spearman rho=0.893, exact two-sided p=0.0123), but initial-majority accuracy is a close comparator (rho=0.821; family partial rho=0.767, p=0.0877), so we do not treat it as calibrated or capability-adjusted prediction. Round-level traces localize many collapses to the first debate round, where early disagreement can precede both harmful cascades and useful recovery. We release replayable schemas, coders, audits, cost cards, and zero-API rebuild scripts so future model-scaffold rows can be compared under the same denominators and signed utility ledger.
[NLP-50] When Words Speak Louder than Images: Towards Understanding Language Bias in Vision-Language Models
【速读】： 该论文旨在解决视觉-语言模型（VLMs）中存在的语言偏见问题，即模型过度依赖语言线索而忽视视觉证据，导致错误预测。其核心挑战在于黑箱性质的VLM推理过程难以追踪语言偏见的传播路径。为此，研究提出了一种诊断性框架，将推理过程分解为四个相互关联的阶段，系统地追踪语言偏见的演化机制；关键创新在于识别并分析两个决定语言偏见的核心因素——语言先验（linguistic priors）与跨模态覆盖度（cross-modal coverage）在各阶段的动态变化。其中，语言先验反映语言模型组件引发的统计偏差强度，是偏见的根源；跨模态覆盖度则衡量语言线索对视觉内容的覆盖程度。通过揭示这两者在推理阶段间的交互作用，该研究不仅构建了语言偏见传播的可解释分析框架，还深入揭示了其内在机制，为后续缓解策略提供了理论基础。
链接: https://arxiv.org/abs/2609.35272
作者: Yizhou Fang,Siyue Chen,Zimo Qi,Zhiyu Xue,Xi Chen,Guangliang Liu
机构: University of Waterloo(滑铁卢大学); Independent Researcher(独立研究员); Johns Hopkins University(约翰霍普金斯大学); University of California, Santa Barbara(加州大学圣塔芭芭拉分校); Nanyang Technological University(南洋理工大学); Indiana University(印第安纳大学)
类目: Computation and Language (cs.CL)
备注: 22 pages, 9 figures. Preprint
Abstract:Despite substantial progress across downstream applications, vision-language models (VLMs) remain susceptible to language bias, often prioritizing linguistic cues over visual evidence and consequently producing incorrect predictions. Prior studies have proposed various approaches to understanding and mitigating language bias in VLMs, yet their findings often conflict due to the difficulty of tracing how language bias propagates within black-box VLMs. Building on the word completion task, we trace how language bias propagates through VLM inference by (1) proposing a diagnostic framework that decomposes the inference process into four distinct yet interdependent stages to trace the propagation of language bias; and (2) examining how two key factors underlying language bias, i.e., linguistic priors and cross-modal coverage, evolve across these stages and ultimately give rise to incorrect predictions. The linguistic prior captures the strength of statistical bias induced by the language model component of a VLM and represents the origin of language bias, whereas cross-modal coverage measures the extent to which linguistic cues cover the visual content. By decomposing inference into four stages and characterizing the interplay between linguistic priors and cross-modal coverage across these stages, we propose a systematic framework for tracing the propagation of language bias throughout the inference process; and uncover the underlying mechanism of language bias by revealing the interplay between linguistic priors and cross-modal coverage.
[NLP-51] Rubric-Aware On-Policy Self-Distillation for LLM Personalization
【速读】： 该论文旨在解决大语言模型（LLM）个性化生成中“用户偏好显式表达”与“模型实际生成能力”之间的鸿沟问题，即现有基于用户特定评价标准（rubric）的方法仅在粗粒度层面（如整体方面覆盖或响应级奖励）利用指导信息，无法有效将用户需求转化为细粒度的生成引导。其解决方案的关键在于提出GRASP框架——一种基于教师-学生自蒸馏的在线策略（on-policy）方法，通过将用户特定的rubric作为细粒度的词元级（token-level）监督信号，实现从语义要求到具体生成行为的精准映射。具体而言，该框架采用一个无rubric约束的学生模型与一个接收目标rubric的教师模型，在学生生成的在线轨迹上对齐二者下一词元分布，从而将教师的rubric条件化知识迁移至学生；为进一步提升监督质量，引入基于rubric的教师验证机制（RTV），仅保留教师充分覆盖目标方面的样本，显著优化了训练效率与监督有效性。实验结果表明，GRASP在个性化问答任务（LaMP-QA基准）上优于多种主流基线模型，验证了词元级鲁棒性指导在个性化生成中的优越性。
链接: https://arxiv.org/abs/2609.35262
作者: Yilun Qiu,Xiaoyan Zhao,Chengbing Wang,Cilin Yan,Rui Zu,Wanyang Zhang,Xiaolong Jiang,Jiayin Cai,Yang Zhang
机构: Xiaohongshu Inc.(小红书); National University of Singapore(新加坡国立大学); University of Science and Technology of China(中国科学技术大学); Peking University(北京大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:LLM personalization aims to generate responses aligned with individual users’ preferences and needs. User-specific rubrics make these expectations explicit, providing direct supervision on what a satisfactory answer should cover. Existing rubric-guided approaches, however, exploit such guidance only at a coarse granularity, either by using rubrics to supervise the prediction of relevant aspects for subsequent generation or by reducing aspect coverage to a single response-level reward for reinforcement learning. This leaves a gap between specifying what a personalized answer should contain and teaching the model how to generate it. To bridge this gap, we propose GRASP, a rubric-aware on-policy self-distillation framework for LLM personalization that turns user-specific rubric aspects into fine-grained, token-level supervision. Specifically, GRASP pairs a rubric-free student with a rubric-informed teacher that additionally receives the target user-specific rubrics. By aligning their next-token distributions along on-policy trajectories generated by the student, GRASP transfers the teacher’s rubric-conditioned guidance into the student, translating user-specific semantic requirements into dense token-level supervision. Since rubric-informed teachers can still produce inadequate supervision, we further introduce Rubric-based Teacher Validation (RTV), which retains only instances where the teacher sufficiently covers the target aspects, improving both supervision quality and training efficiency. Experiments on the LaMP-QA benchmark for personalized question answering demonstrate that GRASP achieves state-of-the-art performance across multiple backbones, supporting the effectiveness of rubric-guided token-level supervision for personalization. To ensure reproducibility, our code is available at this https URL.
[NLP-52] SCBO: Semantically Coherent Batching and Ordering for LLM -Based Social Surveys
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在模拟问卷调查响应时存在的效率与准确性瓶颈问题。传统方法通常采用“单题单提示”（one question per prompt）的模式，导致重复编码相同上下文、限制目标问题可参考的回答范围，并阻碍后续预测利用早期生成的答案信息，从而造成计算资源浪费和推理性能下降。其解决方案的关键在于提出一种无需训练的框架——语义一致的批处理与排序（Semantically Coherent Batching and Ordering, SCBO），通过三个核心步骤实现高效多问联合推理：首先利用LLM提取问卷项的紧凑语义表征并过滤模板噪声；其次基于目标特定检索与中心点补全机制，将相关问题分组形成语义一致的批次，并构建共享参考库；最后根据问题难度从易到难排序目标问题，并依据参考项与问题的语义对齐程度优化参考序列排列。实验结果表明，SCBO在四个大规模调查数据集和四种LLM上的表现显著降低了令牌消耗与推理时间，同时普遍提升了预测准确率，有效实现了高效率与高质量的联合生成。
链接: https://arxiv.org/abs/2609.35250
作者: Yuanzi Li,Lingjie Wang,Zihang Tian,Lei Wang,Xu Chen
机构: Renmin University of China (中国人民大学)
类目: Computation and Language (cs.CL); Computers and Society (cs.CY)
备注:
Abstract:Large Language Models (LLMs) offer a scalable way to simulate survey respondents using demographic profiles and observed reference responses. However, the conventional approach of predicting one question per prompt repeatedly encodes the same context, limits each target to a narrow set of reference responses, and prevents later predictions from using information in earlier answers. Predicting multiple questions in one prompt can reduce these costs, share a broader pool of references, and let later predictions build on earlier ones. This requires forming coherent batches, selecting shared references, and ordering questions and references effectively. We propose Semantically Coherent Batching and Ordering (SCBO), a training-free framework that addresses these challenges. SCBO first uses an LLM to extract compact semantic representations from survey items and filter out template noise. It then groups related questions into batches and builds a shared reference bank using target-specific retrieval and centroid-based completion. Finally, it orders target questions from easy to hard and arranges references according to their semantic alignment with those questions. Experiments on four large-scale survey datasets and four LLMs show that SCBO substantially reduces token consumption and inference time while generally improving prediction accuracy over a non-batched baseline. Code is available at this https URL.
[NLP-53] SignFLIP: A Unified Model for Sign Language Translation and Generation via Stage-wise Alignment at Scale EMNLP2026
【速读】： 该论文旨在解决手势语言翻译（Sign Language Translation, SLT）与手势语言生成（Sign Language Generation, SLG）之间模态间建模不足的问题，现有方法通常将二者视为独立任务或仅在有限数据集上验证，难以实现文本与手势表征间的有效双向对齐。其解决方案的关键在于提出一种以大语言模型（Large Language Model, LLM）为核心的统一框架SignFLIP，通过采用对称架构与基于大规模数据的分阶段训练策略，实现文本与手势表征的共享与渐进式优化：预对齐阶段为后续SLT任务提供基础，而经过SLT适配的表征又能进一步提升SLG性能，从而在多个基准测试中展现出与专用模型相当甚至更优的双向转换能力，并具备良好的跨任务迁移性，尤其在手势语言识别任务中表现突出。
链接: https://arxiv.org/abs/2609.35225
作者: Zhaoyi An,Sihan Tan,Youngbae Hwang,Kazuhiro Nakadai,Rei Kawakami
机构: Institute of Science Tokyo(东京科学研究所); Chungbuk National University(忠北国立大学)
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注: Accepted by EMNLP 2026 Findings
Abstract:Sign language translation and generation share the goal of bidirectional alignment between text and sign representations. However, existing approaches either treat them as isolated tasks or are only verified on limited datasets, limiting effective modeling between modalities. In this paper, we propose SignFLIP, a unified LLM-centered framework for translation and generation. To enable bidirectional mapping between text and sign, SignFLIP adopts a symmetric architecture together with a stage-wise training strategy built on large-scale data. The shared sign–text representation is progressively refined: pre-alignment facilitates subsequent SLT, while the SLT-adapted representation further benefits SLG. Extensive experiments on multiple benchmarks show that SignFLIP shows competitive performance compared with task-specific models on both translation and generation tasks, as well as strong transferability to sign language recognition.
[NLP-54] ANGO: Watermarking Masked Diffusion Language Models in Token Pairs
【速读】： 该论文旨在解决生成式AI（Generative AI）中基于掩码扩散语言模型（masked-diffusion language models）的文本水印（text watermarking）难题。传统水印方法依赖于从左到右的自回归生成顺序，通过将每个新词与前序已生成词关联来嵌入水印，但在掩码扩散模型中，前序词可能仍处于被掩码状态，导致此类方法失效。而固定绿色列表（fixed green list）虽无需上下文依赖，但会反复选择相同高频词，造成明显的词频偏差，使攻击者可通过统计分析恢复水印密钥并伪造可被检测器接受的文本。针对这一问题，本文提出TANGO水印方案，其核心在于：利用秘密密钥将词汇表划分为颜色类别（color classes），并将每个新生成词的分布偏向于由密钥和邻近已解码词颜色共同决定的颜色；水印信息以词对（token pairs）形式嵌入，且偏好颜色随位置动态变化，从而有效抑制词频偏差，使其接近未加水印文本的分布特性。该方法不依赖解码顺序，检测仅需文本和密钥即可完成，在两个掩码扩散模型上均能高精度识别未编辑及多数编辑后的水印文本，且对基于频率的伪造攻击具有强鲁棒性。
链接: https://arxiv.org/abs/2609.35224
作者: Kasra Arabi,Nir Weinberger,Micah Goldblum,Niv Cohen
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Cryptography and Security (cs.CR)
备注:
Abstract:Masked-diffusion language models fill in masked positions in parallel and in no fixed order. Most practical text watermarks assume left-to-right generation. They key each token to the tokens before it, and in a diffusion model those tokens may still be masked. A fixed green list needs no such context, but it favors the same tokens at every position, so these tokens appear more often in watermarked text. An attacker who compares token frequencies in watermarked and unwatermarked text can recover the list and forge text that the provider’s own detector accepts. We present TANGO, a watermark for masked-diffusion language models that keys each new token to a nearby token that is already unmasked. A secret key splits the vocabulary into color classes, and TANGO biases the new token toward a color determined by the key and the nearby token’s color. The watermark is therefore embedded in pairs of tokens. Because the favored color changes from position to position, token frequencies stay much closer to those of unwatermarked text than under a fixed green list. Detection needs only the text and the key, and it does not assume any unmasking order. On two masked-diffusion models, TANGO detects nearly all unedited watermarked texts and most edited ones, and frequency attacks that forge the fixed green list fail against it.
[NLP-55] Understanding On-Policy Distillation: A Mechanistic Interpretability Perspective via Sparse Crosscoders
【速读】： 该论文旨在解决生成式 AI（Generative AI）中基于策略的蒸馏（On-policy Distillation, OPD）机制在知识迁移过程中的内在机理不明确的问题，特别是探究OPD究竟如何影响学生模型（student）内部表示的学习。其核心挑战在于传统交叉编码器（crosscoder）分析方法无法捕捉模型在训练过程中对共享特征使用方式的变化，因为所有模型均被统一映射到一组特征激活值中。为此，作者提出“交换读出”（swap readout）方法，通过独立读取每个学生检查点的特征激活，从而量化训练前后学生对各特征的使用变化，即使这些检查点未被交叉编码器覆盖。研究发现，在三种不同的OPD设置下，OPD既未生成新特征，也未传递教师自身的专属特征，而是保持超过98%的学生高频使用特征的激活率在20%以内波动。进一步分析表明，通常在OPD前进行的监督微调（SFT）预热阶段并非引入新特征，而是通过对共享特征进行重加权来提升性能：一方面提前调整了后续OPD将要改变的特征；另一方面则引入了如对话格式、推理风格和数学符号表达等特定于教师行为的特征变更，且这些变化在OPD过程中得以保留。实验还表明，若直接将此重加权施加于未经更新权重的学生模型特征上，其性能可接近经预热的模型，而对随机打乱的特征施加相同操作则无效。综上，该研究揭示了OPD的本质并非获取新特征，而是通过重新分配现有共享特征的权重，使学生模型学会更有效地利用与教师共享的特征空间，从而实现知识迁移。
链接: https://arxiv.org/abs/2609.35210
作者: Zichao Yu,Qianshuo Ye,Xu Wang,Difan Zou
机构: The University of Hong Kong (香港大学); Shenzhen Loop Area Institute; University of Cambridge (剑桥大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:On-policy distillation (OPD) is a widely adopted post-training technique for LLM reasoning. It is commonly believed to transfer knowledge from a stronger teacher, yet what OPD actually distills into the student’s internal representations remains unclear. We study this question with sparse crosscoders, which learn one feature dictionary shared by the student before and after OPD and the teacher. Standard crosscoder analyses, however, identify model-specific features but cannot tell how a model’s use of its features changes, since all models are encoded into one set of feature activations. We therefore propose the swap readout, which reads each student checkpoint’s feature activations on its own, measuring how training changes the student’s use of each feature, even for checkpoints unseen by the crosscoder. Across three OPD settings, we find that OPD neither creates features nor passes on the teacher’s own, and leaves the firing rates of over 98% of the student’s frequently used features within 20%. We further examine the SFT warm-up on the teacher’s rollouts that commonly precedes OPD and makes it more effective. Rather than adding features, the warm-up reweights the shared ones in two ways. First, it already raises and lowers many of the features that OPD later raises and lowers, doing part of OPD’s work in advance. Second, it changes features that OPD alone would not, notably those for conversation format, reasoning style, and mathematical notation, and these changes persist through OPD. Imposing this reweighting on a directly distilled student’s features, without changing its weights, brings its accuracy close to that of the warmed-up student, whereas the same change on shuffled features does not. Together, these findings suggest that OPD reweights existing features rather than acquiring new ones: the student learns from the teacher how to use the features they already share.
[NLP-56] From Normative Frameworks to Alignment Data: Constructing and Evaluating SFT and Preference Data
【速读】： 该论文旨在解决如何将抽象的规范性原则（normative principles）有效转化为语言模型可学习的具体示例与偏好信号，以实现模型与特定伦理框架的对齐。其核心问题在于，现有方法难以将复杂的宗教与文化伦理体系（如伊斯兰教义、神学及教法传统）中的抽象价值转化为可用于训练的语言模型数据。为此，论文提出了一种由领域专家主导的方法论，通过一年内七位专家系统性地探测模型输出、识别对齐缺陷、筛选理想响应，并基于专家判断构建偏好对（preference pairs），从而生成涵盖广泛规范性领域的阿拉伯语-英语双语数据集（约2.8K条监督微调样本和5.4K个偏好对）。关键解决方案在于利用专家知识对齐过程进行“规范化”操作，将非形式化的伦理原则转化为可量化的训练信号。实验表明，仅使用经专家校准的监督微调数据即可显著提升模型在150个独立提示上的专家偏好度（51.3%优于基线，p < .001），而加入偏好数据后虽进一步提升但差异不再显著（p = .166），且未造成通用能力下降，验证了该方法在将专家定义的规范性原则系统化为可训练数据并进行可控评估方面的有效性。
链接: https://arxiv.org/abs/2609.35201
作者: Husrev Taha Sencar,Rezart Beka,Danish Naeem,Seda Ozalkan,Majd Hawasly,Ji Lucas,Ala AlFuqaha,Mohamed Abdallah,Recep Senturk
机构: Qatar Computing Research Institute, HBKU, Qatar; College of Islamic Studies, HBKU, Qatar; Ibn Haldun University, Turkiye; College of Science and Engineering, HBKU, Qatar
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Aligning language models with a specified normative framework requires translating abstract principles into concrete examples and preference signals from which models can learn. We present an expert-driven methodology for constructing such alignment data and apply it to a normative framework grounded in Islamic ethical, theological, and jurisprudential traditions. Over approximately one year, seven domain experts systematically probed language models to identify alignment deficiencies, curated desired responses, and constructed preference pairs from model outputs and expert judgments. The resulting Arabic-English datasets contain approximately 2.8K supervised fine-tuning (SFT) examples and 5.4K preference pairs spanning a broad range of normative domains. We evaluate the datasets through controlled post-training experiments comparing a Baseline model with models incorporating the curated SFT data alone and both the SFT and preference data. In blind expert evaluation on 150 separately constructed prompts, the model trained with the curated SFT data was preferred over the Baseline in 51.3% of assessor judgments, compared with 14.4% in the opposite direction (p .001 at the prompt level). Adding the preference data resulted in a smaller difference, with the model trained with both datasets preferred over the SFT model in 28.0% of judgments versus 20.9% in the opposite direction; this difference was not statistically significant at the prompt level (p = .166). Standard Arabic and English benchmarks show no broad degradation in general-purpose capabilities. These results demonstrate how expert-defined normative principles can be systematically operationalized into alignment data and evaluated through controlled model training.
[NLP-57] A mechanistic study of language model introspection
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在无外部输入扰动的情况下，仍能感知并报告其内部激活状态变化的“内省”机制问题。具体而言，研究关注模型如何检测和定位其隐藏状态中由人为注入概念向量所引发的内部扰动，同时保持输入文本不变以实现控制变量。其解决方案的关键在于识别出两类具有特定功能的注意力头（attention heads）：中间层的门控头（gate heads）决定模型是否报告存在变化，而后期层中的路由头（router heads）则负责选择报告的具体位置。实验表明，对门控头进行干预可抑制位置报告行为，即使路由头已提供位置信息，说明门控头在决策阶段起关键作用。此外，研究发现报告准确性的差异与概念向量的局部化程度相关——定位更精确的概念会引发更强的门控头注意力分数及输出响应，这与其在查询-键（QK）和输出值（OV）计算中诱导的键与值变化的一致性更高有关。综上，该研究揭示了支持模型内省检测与定位的注意力头机制，为理解大模型内部认知过程提供了重要洞见。
链接: https://arxiv.org/abs/2609.35108
作者: Jiahong Zou,Xiangkun Sun,Lingkai Kong,Tonghan Wang
机构: Shandong University(山东大学); Tsinghua University(清华大学); Northeastern University(东北大学); The University of Hong Kong(香港大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) can sometimes report perturbations to their internal activations—even when the input provides no evidence that an intervention occurred. How do models detect and localize such internal changes? We study this question using a controlled task that keeps the input text fixed. We either inject a concept vector into the hidden state at one of ten token positions or apply no intervention. The model is asked to identify the perturbed position or report that no intervention occurred. Across three model families, we identify two small groups of attention heads with distinct roles in introspective reporting. Middle-layer gate heads influence whether the model reports a change, while router heads in a later layer help select the position to report. Interventions on gate heads can suppress position reports even when router heads supply location information. We further examine why reporting accuracy varies across concepts. Concept vectors that are localized more accurately produce stronger attention-score and output responses in gate heads, which is associated with better alignment of the induced key and value changes in their QK and OV computations. Together, these findings identify attention-head mechanisms supporting introspective detection and localization.
[NLP-58] When Confidence Rises Too Early: Detecting Shortcut Reasoning via Premature Answer Commitment
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在推理过程中存在“捷径推理”（shortcut reasoning）的问题，即模型可能通过不合理的捷径快速得出答案，并在事后生成看似连贯的推理轨迹进行后验解释，导致推理轨迹与真实思维过程不一致。现有方法主要依赖对文本推理轨迹或最终输出的静态分析，难以捕捉模型在生成过程中信念演化的动态特征。为此，论文提出ConfLens框架，其核心是追踪模型在推理过程中对最终答案信心的演化过程。实验发现，在多种捷径推理场景中普遍存在“过早自信”（premature confidence）现象，即模型在早期阶段便表现出高度确定性。针对现有置信度估计方法在泛化性、可靠性及效率上的不足，论文进一步提出分布式答案承诺得分（Distributional Answer Commitment Score, DACS），该方法通过衡量每一步推理中模型对答案承诺的概率分布熵值，无需依赖真实答案或特定任务验证器即可有效捕捉模型信念集中程度。此外，研究将ConfLens的检测结果转化为可解释信号，用于优化奖励模型（reward model）偏好，降低其对捷径推理的倾向。在数学和代码推理任务上的实验表明，结合DACS的ConfLens相比强基线提升了超过4.3%的F1分数，显著改善了模型推理的可信性与正确性之间的匹配度。
链接: https://arxiv.org/abs/2609.35074
作者: Zhaohan Zhang,Junjie Liu,Chengzhengxu Li,Chen Shen,Xiaoming Liu,Chao Shen,Jieping Ye,Ziquan Liu,Ioannis Patras
机构: Queen Mary University of London (伦敦玛丽女王大学); Tongyi Lab, Alibaba Group (通义实验室，阿里巴巴集团); Xi’an Jiaotong University (西安交通大学)
类目: Computation and Language (cs.CL)
备注: 27 pages
Abstract:The reasoning trajectory of a Large Language Model (LLM) is often treated as a verbalized description of its internal reasoning. However, such trajectories can be unfaithful: a model may rely on shortcuts to reach an answer and then post-rationalize the decision with a seemingly coherent chain of thought. Detecting this shortcut reasoning is challenging because existing monitors and verifiers mainly inspect textual traces or final outcomes, rather than how the model’s belief in its answer develops during generation. We introduce ConfLens, a framework that tracks the evolution of confidence in the final answer throughout reasoning. Across three shortcut reasoning settings, we observe a common pattern of premature confidence, where shortcut samples become highly confident in the final answer at early reasoning stages. Existing confidence estimation methods, however, show limited generalizability, reliability, or efficiency for detecting this behavior. We therefore propose the Distributional Answer Commitment Score (DACS), a distributional confidence estimator that measures the entropy of the model’s probability distribution over answer commitment at each reasoning step. DACS captures how concentrated the model’s answer belief is without requiring ground-truth answers or task-specific verifiers. We further convert ConfLens detection results into interpretable signals for reward models to reduce their preference for shortcut reasoning. Experiments on mathematical and code reasoning tasks show that ConfLens with DACS improves shortcut reasoning detection by over 4.3% F1 compared with strong baselines and reduces the mismatch between faithfulness and correctness in reward model preferences.
[NLP-59] Echoes of Deeds: Moral History Can Shape and Steer LLM Behavioral Choices
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在道德评估中普遍忽略个体过往行为历史对后续决策影响的问题，即现有评估方法多基于孤立的决策情境，未能考察一个主体的非相关先前行为是否会影响其后续选择。这一空白导致无法明确道德历史（moral history）在多大程度上塑造了LLM的决策行为。为此，作者提出MoralLedger框架，从行为层面与表征层面双重探究道德历史对模型行为的影响机制。其关键在于：在行为层面，发现先前道德历史的正负性（valence）与强度（intensity）系统性地调节后续决策；在内部表示层面，揭示了残差流（residual stream）中存在可线性恢复的道德历史方向，并可通过沿该方向进行干预，实现对中性历史提示下后续选择的双向、强度依赖型控制，且效果显著优于仅通过提示词或有利非道德方向所诱导的改变。该研究首次证明，模型内部隐含的主体先前道德行为表征可在推理阶段提供带符号的可控性，从而将道德评估从静态困境拓展至动态历史依赖场景，确立道德历史既是行为敏感性的来源，也可作为审计与调控模型道德行为的因果靶点。
链接: https://arxiv.org/abs/2609.35070
作者: Lucio La Cava,Andrea Tagarelli
机构: University of Calabria (卡拉布里亚大学); DIMES Dept. (数字与信息系统科学系)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注:
Abstract:Evaluations of Large Language Models (LLMs) morality typically consider decisions in isolation, thus overlooking whether an individual’s unrelated prior conduct influences the model’s subsequent choices. This leaves open the question of whether, and to what extent, moral history shapes LLM decisional behaviors. Prior work on human moral decision-making shows that past behavior can influence subsequent moral choices. Building on this observation, we investigate whether analogous effects emerge in LLMs in two complementary ways: at the behavioral level, through the model’s observable responses, and at the representation level, through its latent internal representations. We introduce MoralLedger, a framework for studying how an actor’s moral history shapes actions for LLMs’ behaviors under a fixed decision context. At the behavioral level, we find that prior moral histories systematically alter subsequent choices as a function of their valence and intensity. At the internal representation level, these histories induce a linearly recoverable direction in the residual stream that generalizes to held-out examples. Intervening along this direction on neutral-history prompts produces two-sided intensity-dependent changes in subsequent choices, with effects that are stronger than those induced by prompting alone or by favorable-nonmoral direction. To our knowledge, this is the first demonstration that a latent representation of an actor’s prior moral conduct can provide signed inference-time control over a moral decision. Our MoralLedger extends moral evaluation beyond static dilemmas, establishing moral history as both a source of behavioral sensitivity and a causal target for auditing and controlling moral behavior in LLMs.
[NLP-60] WebPageBench: Event-Level Verification and Controlled UI-Variant Generation for Web Agents
【速读】： 该论文旨在解决当前网页智能体（Web Agent）评估中缺乏客观、可复现且与界面交互行为紧密对齐的评测标准这一核心问题。现有方法通常依赖人工标注或模型判断（judge model）来验证任务完成情况，易引入主观偏差，且难以捕捉真实用户操作流中的细粒度事件轨迹。为此，本文提出 WebPageBench——一个开放的评测框架，其关键创新在于：所有任务的成功判定均基于页面界面自身生成的类型化事件日志（typed events with parameters），通过精确匹配预期事件序列来定义任务完成，完全无需依赖外部判断模型或对渲染页面进行爬取。该设计确保了评估结果的客观性与可重复性。此外，框架通过在单一控制组件上配置不同UI实现（如浅色/深色模式切换），可在保持任务指令与成功条件不变的前提下，系统性地测量智能体对界面形式变化的敏感性，从而揭示其泛化能力缺陷。WebPageBench 的发布包含 152 项任务（65 个基准场景与 87 个控制变体）、统一运行器（支持六种浏览器/DOM 框架组合及五类仅基于截图的 GUI 智能体）、以及公开排行榜（涵盖 24 个模型-运行器组合）。在 152 项任务的公开排行榜上，智能体自报完成率与日志验证实际完成率之间的差距高达 41 分（某一配置下所有任务均被声明完成，但仅 59% 真正满足日志条件），凸显了现有智能体在任务真实性验证上的严重过拟合问题。因此，该研究的核心解决方案在于构建一种基于事件日志的、无偏见的、可控制界面变异的自动化评测机制，为提升网页智能体的真实性能评估提供了坚实基础。
链接: https://arxiv.org/abs/2609.35026
作者: Anton Emelyanov,Maria Tikhonova,Zaven Martirosian,Sergei Averkiev,Alena Fenogenova
机构: DAIMLD; HSE University (高等经济大学); NUST MISIS (俄罗斯国立科学技术大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:We present WebPageBench, an open framework for evaluating web agents in which every task is verified from the interface’s own event log. Six instrumented mock sites with brand identifiers removed (a marketplace, a bookstore, a grocery service, rail ticketing, hotel search and a document cabinet) emit typed events with parameters as a user or an agent acts. A task declares the events it requires, and success is decided by matching them, with no judge model and no scraping of rendered pages. The same instrumentation supports controlled UI variation: one configuration switch re-renders a task through a different implementation of a single control while the prompt and the success conditions stay completely identical, so sensitivity to interface form can be measured under a fixed task specification. The WebPageBench release consists of three components: 152 tasks, divided into 65 canonical scenarios and 87 control variants across light/dark UI-modes; a common runner evaluated with six browser/DOM harness configurations and five screenshot-only GUI-agent families; and a public leaderboard of 24 model-harness pairs. On the public 152-task leaderboard the gap between what agents declare finished and what the log confirms reaches 41 points (one configuration declares every task finished and satisfies the conditions on 59%).
[NLP-61] he Right Lesson at the Right Step: Deriving Control Updates for Self-Evolving Agents
【速读】： 该论文旨在解决自演化智能体在长期工具使用工作流中，如何实现经验重用的局部化控制问题。传统方法将过往经验以全局提示、记忆或反思的形式存储，但缺乏对经验作用位置的精确控制，导致相同经验可能在某些决策中修正错误，而在另一些决策中产生干扰。其核心挑战在于：经验重用不仅是记忆问题，更是控制流中的精准定位与生效时机问题。解决方案的关键是提出EvoCUE（Evolution through Control Updates from Evidence）框架，将智能体建模为显式的状态机控制器（state-machine controller），其中节点执行模型或工具调用，边定义控制转移路径。该结构允许在特定位置进行可编辑的控制更新，使每个学习到的更新能够明确指定“添加什么”、“在何处生效”以及“何时应用”。EvoCUE通过分析已完成轨迹中的残差目标和执行痕迹，生成局部指令或技能修改建议，并在相同检查点处回放父级与修改后控制器，对比最终结果以评估候选更新的有效性。经验证有效的更新被编译为适用规则，在保留任务上确认后传递给后续执行。实验表明，在长周期工具使用环境（如AppWorld）中，EvoCUE能从无基准特定引导的初始控制器中学习任务完成规范，显著提升正常与挑战性测试集的成功率；在PAST-Bench办公工作流中，成功将前期经验中的组织要求迁移至后续任务，提升执行质量。结果表明，自演化智能体应将经验嵌入控制流本身，而非仅作为文本存储，从而实现更精准、高效的自我进化。
链接: https://arxiv.org/abs/2609.34988
作者: Yunhe Su,ZiYi Dong,Tong Yu,Weijian Deng,Hao Li,Bowen Jiang,Pengxu Wei
机构: Sun Yat-sen University (中山大学); eunomia-bpf community; Tsinghua Shenzhen International (清华大学深圳国际研究生院); Graduate School, Tsinghua University (清华大学研究生院); Nankai University (南开大学); Department of Computer and Information Science (计算机与信息科学系); University of Pennsylvania (宾夕法尼亚大学); Peng Cheng Laboratory (鹏城实验室)
类目: Computation and Language (cs.CL)
备注: Preprint. 3 figures, 5 tables
Abstract:Self-evolving agents improve future behavior by reusing past experience, typically as global prompts, memories, or reflections. Yet these mechanisms rarely control where experience takes effect. In long tool-use workflows, the same lesson may correct one decision but distract another, making experience reuse a problem of localized control rather than memory alone. We introduce EvoCUE (Evolution through Control Updates from Evidence), a framework for learning reusable control-program updates from completed agent executions. EvoCUE represents the agent as an explicit state-machine controller, whose nodes perform model or tool calls and whose edges define where control passes next. This makes the workflow editable at precise locations, so each learned update can specify what to add, where it acts, and when it applies. From completed trajectories, EvoCUE uses residual goals and observed execution traces to propose localized instruction or skill edits. Each candidate is evaluated at the point where it would act by resuming the parent and edited controllers from the same checkpoint and comparing their final outcomes. Accepted edits are compiled with applicability rules, confirmed on held-out tasks, and inherited by later executions. We evaluate EvoCUE on long tool-use environments where learned conventions must reach the right execution step. From a minimal AppWorld controller without benchmark-specific onboarding instructions, EvoCUE learns the missing task-completion convention and substantially improves success on Test-Normal and Test-Challenge. On PAST-Bench office workflows, EvoCUE transfers organizational requirements from prior episodes to later tasks, improving task-execution quality. These results show that self-evolving agents should place experience inside the control flow, rather than only store it as text.
[NLP-62] Nürnberg NLP at ChildSafeAds 2026: Structurally Dissimilar Voter Ensembles under Four Levels of Data Access EMNLP2026
【速读】： 该论文旨在解决儿童面向的YouTube视频中商业内容监控系统在不同数据访问权限下所能达到的性能上限问题。其核心挑战在于如何在有限或分级的数据访问条件下，构建一个高效且准确的自动化监测系统，以识别视频中的广告内容及其合规性。解决方案的关键在于设计一个多分支集成框架，由九个分类器（voter）组成，分为三个分支，各分支在主干网络结构（backbone）、微调方法（adaptation method）和类别范围（class scope）上有所差异，从而实现对不同子任务的针对性优化。通过基于频道隔离的交叉验证进行模型选择，并利用开发集作为迁移性能的检验基准，该系统在三项子任务中取得了两项第一的成绩，其中产品类别识别得分（ST2, 0.8243）和合规标志预测得分（ST3, 0.6530）均为22个最终提交方案中的最优，且在任务平均得分上位列第三（0.7079）。此外，研究还系统比较了四种数据访问级别，并报告了在测试集规模下的成本表现，为实际部署提供了可量化的评估依据。
链接: https://arxiv.org/abs/2609.34986
作者: Philipp Steigerwald,Eric Rudolph,Jens Albrecht
机构: Technische Hochschule Nürnberg Georg Simon Ohm(诺伊堡技术大学乔治·西蒙·欧姆)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted at the ChildSafeAds 2026 Shared Task @ NLLP Workshop, EMNLP 2026 (1st place in 2 of 3 subtasks)
Abstract:We describe the Nürnberg NLP system for ChildSafeAds 2026. The shared task asks what a monitoring system for commercial content in child-facing YouTube videos can achieve at a given level of data access. We answer with per-subtask ensembles of nine voters, organised into three branches that differ in backbone, adaptation method and class scope. Selection rests on channel-disjoint cross-validation, with the development set as a transfer check. The system wins two of the three subtasks. Its product-category score (ST2, 0.8243) and its compliance-flag score (ST3, 0.6530) are the best of the 22 final entries, and it places third on the task mean (0.7079). We further compare four access levels and report the cost at test-set scale.
[NLP-63] ORPG: Reconciling Multiple Reward Objectives through Objective-wise Policy Gradients
【速读】： 该论文旨在解决多奖励策略优化中如何有效整合多个目标的学习信号并保持其预期关系的问题，尤其是在各目标间存在兼容或冲突梯度时的联合更新难题。其核心解决方案是提出逐目标修正的策略梯度（Objective-wise Reconciled Policy Gradient, ORPG），通过为每个奖励构建独立的裁剪策略目标，并将生成的梯度进行协调融合以实现统一的策略更新。对于兼容梯度，采用基于余弦的插值机制，在部分归一化的参考方向上协调贡献，同时保持梯度和的范数不变，该更新被形式化为球面上方向妥协的唯一解；对于冲突梯度，则依据任务优先级进行投影处理。在助人—安全对齐与正确性—成本优化等数学推理场景中的实验表明，ORPG显著优于最强外部基线，在平均有用性与无害性得分上均有提升；在数学推理任务中，其在全预算平均准确率和三预算超体积指标上均达到最优，且生成响应更准确、更简洁。组件对比与训练动态分析进一步验证了兼容性协调的主导作用及冲突处理带来的互补优势，证明该方法适用于同等重要性目标及具有明确优先级的目标优化。
链接: https://arxiv.org/abs/2609.34985
作者: Shicheng Fang,Yiwen Zhao,Wenbo Tian,Jiahao Lu,Yining Zheng,Yuxin Wang,Xipeng Qiu
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Multi-reward policy optimization requires a joint update that reflects both the learning signals and the intended relationships among objectives. We introduce Objective-wise Reconciled Policy Gradient (ORPG), which constructs a separate clipped policy objective for each reward and reconciles the resulting gradients into one policy update. For compatible gradients, a cosine-dependent interpolation coordinates their contributions through a partially normalized reference while preserving the norm of their sum. We characterize this update as the unique solution of a spherical directional compromise. For conflicting gradients, projection follows the task’s priorities. We evaluate the same compatible rule in helpfulness–safety alignment and correctness–cost optimization for mathematical reasoning. ORPG substantially improves average Useful and Harmless scores over the strongest external baseline on each axis. In mathematics, it achieves the highest average full-budget accuracy and three-budget hypervolume among the compared methods, with more accurate and shorter responses than the initial policy. Component comparisons and training dynamics show the larger contribution of compatible coordination and a complementary benefit from conflict handling. These results support gradient reconciliation for objectives with equal standing and for objectives with an explicit priority.
[NLP-64] See it Say it Sorted: Mechanistic Diagnosis and Parameter-Space Mitigation of Emergent Misalignment in LLM s
【速读】： 该论文旨在解决安全对齐的大语言模型（Safety-aligned LLMs）中出现的涌现性错对齐（Emergent Misalignment, EM）问题，即在特定领域微调过程中，模型在无关领域突然引发灾难性安全失效的现象。现有静态分析方法无法刻画训练动态过程，而现有防御策略依赖启发式规则，导致模型实用性下降。本文提出一种动态的二阶几何分析框架，通过追踪训练轨迹发现：方向性海森曲率（directional Hessian curvature）高度集中于语义枢纽词元（semantic pivot tokens）。基于格拉斯曼投影（Grassmannian projections）的分析表明，有害-安全梯度间隙扩大主要源于安全梯度重叠度的下降。基于此，作者提出参数级几何缓解框架（Geometric Mitigation Framework），通过正交投影将经验有害梯度子空间从参数更新中剔除。实验结果表明，在Qwen2.5-14B-IT上，该方法可抑制自由生成场景下的EM达80.0%；在其余三个开源指令微调模型家族（3B–20B）中，尽管单层行为式EM已接近零，但教师强制评估仍显示同一有害子空间控制着冻结式EM响应的条件支持。关键发现是，这些诊断揭示了行为安全的假象——即使行为层面未显现错对齐，该有害子空间依然可观测且可操控。
链接: https://arxiv.org/abs/2609.34970
作者: Weiqiao Que,Ruizhe Li,Chengyu Wang,Dakan Wang,Emine Yilmaz,Xiaofeng He
机构: East China Normal University (华东师范大学); University of Birmingham (伯明翰大学); Alibaba Group (阿里巴巴集团); Exacity Inc.; University College London (伦敦大学学院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Preprint
Abstract:Safety-aligned LLMs can exhibit emergent misalignment (EM): narrow domain adaptation unexpectedly triggers catastrophic safety failures across unrelated domains. Prior static analyses leave training dynamics unmapped, while existing defenses rely on heuristics that degrade utility. We present a dynamic, second-order geometric study of EM. Tracking training trajectories reveals that directional Hessian curvature concentrates sharply on semantic pivot tokens. Grassmannian projections show that, in most settings, harmful-safe gap widens mainly because safe-gradient overlap declines. Leveraging these insights, we introduce a parameter-level Geometric Mitigation Framework that orthogonally projects empirical harmful gradient subspace out of parameter updates. On Qwen2.5-14B-IT, our defense suppresses free-generation EM by up to 80.0%; across the other three of four open-weight instruction-based model families (3B–20B), where single-layer behavioral EM is already near zero, teacher-forced evaluation shows same harmful subspace controls the conditional support of frozen EM responses. Crucially, these diagnostics unmask the illusion of behavioral safety: the same subspace remains measurable and steerable in models where behavioral EM is near zero. Code: this https URL.
[NLP-65] Semantic Uncertainty Quantification Needs Factual Equivalence
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）中语义不确定性量化（Semantic Uncertainty Quantification, UQ）的准确性问题。现有方法普遍采用“采样多个回答—衡量一致性—以不一致作为不确定性度量”的通用模板，其核心在于通过一个比较器（operator）对成对答案进行相似性评估，并由聚合器（aggregator）将所有成对比较结果整合为标量不确定性分数。然而，现有方法在聚合策略上差异显著，却普遍依赖现成的比较器（如自然语言推理模型或通用句子编码器），而这些现成组件无法准确捕捉不同回答之间关于事实内容的等价性，成为语义UQ性能的主要瓶颈。为此，本文提出一种简洁有效的解决方案：使用由大语言模型生成的合成数据训练一个对比式单编码器（single encoder），专门聚焦于目标事实的区分能力，且训练与评估数据集完全解耦。将该定制化比较器集成至现有方法后，在126个评估设置中的120个（95%）实现性能提升，覆盖18种模型-数据集组合，涵盖语言和视觉-语言模型；最优变体达到0.76的平均AUROC，优于最强基线的0.68。此外，该方法将原本基于二次交叉编码器的蕴含关系比较简化为每个答案仅一次编码器前向传播，显著提升效率。统一且广泛的性能改进表明，限制语义UQ性能的关键因素并非聚合机制，而是比较器本身。该统一的比较器还可用于单次生成的词元级不确定性估计：通过为每个词元分配的范数反映其对答案的重要性，据此重加权词元似然值可进一步优化不确定性估计精度。
链接: https://arxiv.org/abs/2609.34967
作者: Joseph Hoche,Quentin Guimard,Gianni Franchi
机构: AMIAD, Pôle Recherche Palaiseau(法国巴黎萨克雷)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Semantic uncertainty quantification for large language models rests on a common template: sample several answers, measure how much they agree, and treat disagreement as uncertainty. We first formalize this template as two separate roles: an operator that compares two answers, and an aggregator that combines all pairwise comparisons into a scalar. Existing methods differ almost entirely in how they aggregate, while taking the operator off the shelf, typically an NLI model or a generic sentence encoder. We show that this reliance on off-the-shelf operators is the primary bottleneck of semantic UQ: they do not accurately measure factual equivalence of multiple answers to the same question. We resolve this with a deliberately simple recipe: a single encoder trained contrastively to isolate the targeted fact, utilizing synthetic data generated by an LLM and dataset both disjoint from all evaluation settings. Integrating the resulting operator into existing methods improves performance on 120 of 126 evaluation settings (95%) spanning 18 model dataset combinations across language and vision-language models. The best variant reaches 0.76 mean AUROC against 0.68 for the strongest baseline, while replacing the quadratic cross-encoder comparisons of entailment-based operators with one encoder pass per answer. The uniformity of the improvement supports the view that the operator, not the aggregator, is the limiting factor. The same operator also improves single generation token-level estimators: the norm it assigns to each token measures how much that token bears on the answer, and reweighting token log-likelihoods accordingly sharpens the estimate.
[NLP-66] Neural Language Models Learn the Contextual Distributions of Dependency Structures: a statistical learning theory to compositionality
【速读】： 该论文试图解决的问题是：神经语言模型（Neural Language Models, NLMs）如何习得独立于词汇语义的语法结构所蕴含的结构性意义。其核心挑战在于，传统分布统计方法仅依赖词元共现频率，难以捕捉由句法依存关系构成的复合结构的语义属性。该研究提出的关键解决方案是：在统计学习过程中，已习得的依存结构本身会成为后续统计学习的新分布单元。具体而言，当模型识别出某一依存结构后，会追踪其上下文分布特征，这些特征反映了复合结构的语义属性。为验证该假设，研究设计了一种合成语法，其中每个语法结构具有独特的上下文分布，无法仅通过组成词元的分布统计推导得出。通过在该语法上训练一系列类BERT的掩码语言模型并分析其发展轨迹，结果表明，尽管无法从词元统计中推断，模型仍能成功学习复合依存结构的上下文分布。进一步的发展分析显示，依存关系的习得始终先于其上下文特征的学习，揭示出清晰的学习阶段顺序。这表明，NLM中的统计学习并非简单的词元共现累积，而是一个将已习得的依存结构作为新分布单位进行递归学习的过程。该机制为解释语言模型如何解决语言的组合性问题（compositionality problem）提供了基于统计学习的理论框架，并暗示这种学习过程可能为语言认知如何从纯粹的分布统计中涌现提供解释性理论。
链接: https://arxiv.org/abs/2609.34936
作者: Wang Bojun,Junjie Chen,Holly Jenkins,Elizabeth Wonnacott
机构: University of Oxford (牛津大学)
类目: Computation and Language (cs.CL)
备注: 11 figures
Abstract:It is unclear how Neural Language Models (NLMs) acquire the structural meaning encoded by grammatical structures that is independent of lexical semantics. We propose a statistical learning process in which learned dependency structures themselves become new distributional units for subsequent statistical learning. Under this account, once a dependency structure is acquired, the model tracks its contextual distributions. These contextual features reflect the semantic properties of a composite structure. To test this hypothesis, we design a synthetic grammar in which each grammatical structure has distinct contextual distributions that cannot be recovered from the distributional statistics of their component tokens alone. We train a series of BERT-style masked language models on this grammar and examine their developmental trajectory. The results show that models can successfully learn the contextual distributions of composite dependency structures even though they cannot be inferred from token statistics alone. Developmental analysis further reveals a clear developmental trajectory. The learning of the dependency relations that define a grammatical structure consistently precedes the learning of its contextual features. These findings suggest that statistical learning in NLMs is not merely the accumulation of token co-occurrence statistics, but a process in which learned dependency structures become new units of distributional learning. We argue that this process provides a statistical-learning account of how NLMs solve the compositionality problem in language. Finally, we discuss the possibility that this statistical learning process provides an explanatory theory on how language cognition could emerge from pure distributional statistics.
[NLP-67] Dont Forget! Decomposing the Training Dynamics of Memorization in Language Models
【速读】： 该论文旨在解决语言模型在训练过程中对尾部数据（tail of training distributions）的拟合机制问题，特别是针对“记忆化”（memorization）现象的训练动态不清晰这一关键挑战。其核心问题是：语言模型如何通过参数更新实现对重复或罕见训练序列的记忆，并在训练过程中表现出遗忘与保留的复杂权衡。解决方案的关键在于提出一种细粒度的损失轨迹分解方法，将记忆化过程解耦为序列级梯度对齐（sequence-level gradient alignment）的动态表现。研究发现，在Pythia系列模型中，无论是重复训练序列（recitation）还是罕见序列（recollection），记忆化均表现为显著的序列级梯度对齐特征；其中，重复样本因与其他训练信号存在梯度错位而易发生遗忘，从而解释了为何需要更高的重复次数以维持记忆。此外，研究揭示底层网络模块在记忆与遗忘过程中起主导作用。基于该分解框架的预测性能优于传统的交叉熵基线，尤其在大模型及训练早期表现更优。进一步地，通过对少数高影响力参数进行干预，可有效消除最终模型中的记忆化行为。这些发现深化了对记忆化演化过程的理解，并为预测与调控记忆化提供了可操作的技术路径。
链接: https://arxiv.org/abs/2609.34933
作者: Florian Eichin,Philipp Mondorf,Andrei Mircea,Yupei Du,Barbara Plank,Michael A. Hedderich
机构: MaiNLP, Center for Information and Language Processing, LMU Munich; Munich Center for Machine Learning (MCML), Germany; Mila – Quebec AI Institute, University of Montreal, Canada; Saarland University, Germany
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Memorization has been proposed as a mechanism to explain how language models fit the tail of their training distributions, but its training dynamics are not understood well. In this work, we take a fine-grained look at memorization by decomposing the loss trajectory of memorized sequences over training and model parameters. Across the Pythia family, we study memorization of duplicated training sequences (recitation) and rare ones (recollection). We find that memorization in both cases is characterized by sequence-level gradient alignment, though recitation suffers from misalignment with other training influences which causes forgetting, explaining the necessity for higher duplication of these examples. We further show that the lower model layers are the most involved in memorization and forgetting. Predicting memorization, our decomposition improves over a cross-entropy baseline, especially in larger models and early in training. Intervening on a small set of highly influential parameters we are able to ablate memorization in the final model. Together, these findings advance our understanding of how memorization develops during training and offer insights for predicting and intervening on it.
[NLP-68] Sample What You Say: Aligning Language Models to Sample the Distributions They State
【速读】： 该论文旨在解决生成式语言模型在指令微调后虽能正确描述目标分布，却难以有效从中采样的问题。现有方法如提示工程与解码策略调整仅部分缓解此分布偏差（distributional mismatch）问题，因此作者提出采用策略优化（policy optimization）进行训练。其核心解决方案是基于组相对策略优化（Group Relative Policy Optimization, GRPO）框架，引入一种名为“见证优势”（witness advantage）的新奖励机制。该机制利用最大均值差异（Maximum Mean Discrepancy, MMD）的数学性质，通过构建一个“见证函数”（witness function）来量化每个离散输出结果在模型生成群体中相对于目标分布的过产或欠产程度。每个样本（rollout）的优势值即为其对应输出的负见证函数估计值：当某结果在群体中被低估时获得正向奖励，被高估时则受到惩罚。该优势计算可直接由群体内各结果的计数闭式求得，无需额外建模。实验表明，使用见证优势进行训练，在未见的目标分布上显著降低了模型生成分布与目标分布之间的总变差距离（Total Variation Distance），同时保持了模型的通用生成能力。
链接: https://arxiv.org/abs/2609.34929
作者: Kasra Arabi,Virginia Smith,Chhavi Yadav
机构: New York University (纽约大学); Carnegie Mellon University (卡内基梅隆大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Language models are increasingly used to sample from a specified distribution, for instance, to simulate survey respondents or generate synthetic data. Instruction-tuned models can state such a distribution correctly and still fail to sample from it. Prompting and changes to decoding reduce this mismatch only partly, which motivates training with policy optimization. Group relative policy optimization (GRPO) is a natural fit for this problem because it already samples a group of rollouts per prompt, and the group’s empirical distribution can be compared with the target. However, scoring the group as a whole gives every rollout the same reward. Group-relative centering then sets all advantages to zero, and the model receives no learning signal. To give each rollout its own signal, we introduce the witness advantage, a per-rollout advantage derived from maximum mean discrepancy (MMD). It trains a model to match a target distribution over a finite set of outcomes. The MMD between the model’s distribution and the target has a witness function that measures how over- or under-produced each outcome is. Each rollout’s advantage estimates the negative witness at its outcome, so a rollout is rewarded for an outcome the group under-produces and penalized for one it over-produces. The witness advantage is computed in closed form from the group’s outcome counts, and we use it as the reward in GRPO. On unseen target distributions, training with the witness advantage substantially reduces the total variation distance to the target while largely preserving the model’s general capabilities.
[NLP-69] One Readout Many Repairs: Diffusion-Guided Hierarchical Search for Tool-Agent Repair
【速读】： 该论文旨在解决工具代理（Tool Agent）在调用外部工具时，尽管单个工具调用成功执行，但整体任务仍无法完成的问题。其核心挑战在于：工具调用失败的修复需要同时探索操作选择（operation choices）及其具体实现（realizations），而传统方法通过完全重构调用序列来尝试替代路径，导致计算开销巨大，并且重复进行相同的操作选择，即使失败源于具体实现细节。为此，论文提出将修复过程建模为对“操作支持集”（operation supports）的层次化搜索，其中操作支持集被定义为一组允许的操作类型集合，用于界定可复用的搜索区域以生成具体的工具调用序列。提出的ReCommit框架是一种无需训练、基于扩散模型引导的修复机制，通过一次性并行读取掩码扩散语言模型，重用操作类型的评分结果，从而在多次修复尝试中摊销操作层级的提案计算成本；同时，在每个支持集中进一步探索实体绑定、参数设置及动作组合等实现层面的变体。实验表明，在Agent-Diff基准上针对四个企业级服务的真实失败案例，相较于最强的8B对比方法，ReCommit在修复预算B=3和B=13时分别实现了75.9%与63.2%的相对恢复率提升，同时平均全预算修复时间分别减少61.3%和51.3%；此外，其在恢复效果与计算成本之间表现出优越的权衡能力，甚至优于部分32B规模模型的表现。
链接: https://arxiv.org/abs/2609.34879
作者: Xiang Xia,Cheng Yan,Fan Xu,Zhijun Fan,Shuyuan Zhang,Wuyang Zhang
机构: University of Science and Technology of China (中国科学技术大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Tool agents use large language models to act through external tools, yet successfully executed calls can still leave user requests unfulfilled. Tool-agent repair seeks alternative call sequences that execute successfully and fulfill the original requests. However, repair requires exploring both operation choices and their concrete realizations, making complete-sequence regeneration costly. Moreover, regeneration repeats operation selection even when failure arises from how those operations are realized. The resulting challenge is to reduce this repetition while preserving exploration of alternative operations and realizations. Therefore, we formulate repair as hierarchical search over operation supports, which we introduce as sets of permitted operation types that define reusable search regions for concrete tool-call sequences. We propose ReCommit, a training-free, diffusion-guided framework for improving tool-agent failure recovery while reducing repair computation. ReCommit amortizes operation-level proposal computation across repair trials by reusing operation-type scores from a single parallel readout of a masked diffusion language model. These scores guide search across supports, while realization search explores alternative entity bindings, arguments, and action composition within each support. Experiments on real failures across four enterprise services in the Agent-Diff benchmark show 75.9% and 63.2% relative recovery gains with 61.3% and 51.3% reductions in mean full-budget repair time at repair budgets B=3 and B=13 , respectively, over the strongest evaluated 8B comparison method. ReCommit achieves a favorable recovery–cost trade-off, including in comparisons with the evaluated 32B models.
[NLP-70] Beyond Verbalized Confidence: Calibrating Reason ers with Differentiable Readouts
【速读】： 该论文旨在解决生成式推理模型在强化学习中使用可验证奖励（Reinforcement Learning with Verifiable Rewards, RLVR）时，尽管能提升答案正确性，却无法保证其置信度（confidence）校准的问题。现有方法虽尝试在RLVR框架内进行置信度校准，但均依赖从文本中采样置信度值，导致两个关键缺陷：一是采样引入高方差且实际收敛为少数离散值，二是采样过程不可微，迫使校准损失通过标量奖励传递，限制了优化效率。为此，本文提出CREDO（Confidence REaDOut），采用确定性读出机制替代采样，从模型输出分布中直接读取预设的置信度标记对（token pair），并通过可微回归训练置信度。此外，CREDO将训练后的置信度作为准确性信号，依据置信度与实际结果之间的差异动态加权回放轨迹，实现准确率与置信度校准的协同优化。实验表明，该方法在数学和代码推理任务中均取得最优的准确率与校准性能，且在拒绝回答（abstention）和选择性预测等场景中表现出显著优势。
链接: https://arxiv.org/abs/2609.34857
作者: Chenxiao Fan,Chongming Gao,Gangyi Zhang,Leyang Shen,Yaxin Gong,Jiamin Wang,Jiakai Wang,Dong Wang,Yang Liu,Fuli Feng,Xiangnan He
机构: University of Science and Technology of China(中国科学技术大学); Qwen Business Unit of Alibaba(阿里巴巴通义业务部)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Reinforcement learning with verifiable rewards (RLVR) trains reasoning models to produce correct answers, but does not ensure that their stated confidence is calibrated. The resulting models are systematically overconfident. Recent methods train calibration inside the RLVR loop by having the model state a numerical confidence alongside its answer, but they all obtain the confidence by sampling it as text. This choice imposes two costs: a sampled confidence introduces variance and in practice collapses to a handful of distinct values, and sampling makes the confidence non-differentiable, forcing the calibration loss through a scalar reward. We propose CREDO (Confidence REaDOut) to replace sampling with a deterministic readout. While RLVR optimizes correctness, CREDO reads the confidence from a dedicated token pair in the model’s output distribution and trains it by differentiable regression. CREDO further turns the trained confidence into a signal for accuracy, weighting rollouts by how far confidence and outcome disagree, so that accuracy and calibration improve together. Across mathematical and code reasoning, CREDO attains the best accuracy and calibration, and the gains extend to abstention and selective prediction.
[NLP-71] Adapt Semantics Not Structure: Few-Instance Schema Calibration for Scientific PDF Extraction
【速读】： 该论文旨在解决在仅有少量已验证抽取样本的情况下，如何高效且可靠地完成大语言模型（LLM）的抽取模式（extraction schema）校准问题。传统方法依赖人工反复试错调整数百个字段定义，成本高昂且难以保证一致性。其核心挑战在于如何在保持原有结构契约（structural contract）的前提下，从少数高质量标注文档中自适应地优化抽取提示（extraction prompts）与字段级语义描述。解决方案的关键是提出一种合同保持的语义抽取框架（CPSE, Contract-Preserving Semantic Extraction），该框架将抽取模式解耦为不变的结构契约与可变的字段语义，并通过显式条件驱动的解析机制分离实体识别与记录补全过程，实现抽取提示与语义描述的联合校准。在专家标注的高分子科学文献上，CPSE相较于执行匹配基线提升了9.93个百分点的抽取性能，在独立评审与盲审专家评估中均表现出一致增益，验证了其在低资源场景下维持下游所需输出结构的能力。
链接: https://arxiv.org/abs/2609.34841
作者: Zixiao Dong,Wei Yang,Zihao Liu,Chenshu Li,Longzhang Liu,Tao Tan,Hong Xie
机构: University of Science and Technology of China (中国科学技术大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:A well-designed extraction schema is not necessarily ready for reliable LLM execution. When only limited verified extractions are available, manually tuning hundreds of field definitions through trial and error is costly. We frame this problem as few-instance schema calibration: adapting the operational semantics of an existing schema from a few annotated documents while preserving its structural contract. We introduce CPSE, a contract-preserving semantic extraction framework that jointly calibrates extraction prompts and field-level semantic descriptions from a few gold annotations. CPSE decomposes the schema into an invariant structural contract and mutable field semantics, and further separates identity discovery from record completion using manifest-conditioned resolution. On expert-annotated polymer-science documents, CPSE improves extraction by 9.93 points over an execution-matched baseline, with consistent gains under an independent judge and in a blinded expert audit. These results show that CPSE enables low-resource schema execution while preserving the output structure required downstream.
[NLP-72] OpenWhistle: A Large-Scale Longitudinal Dataset and Benchmark of Bottlenose Dolphin Vocalizations NEURIPS2026
【速读】： 该论文旨在解决现有生物声学资源严重偏向鸟类且每物种数据深度不足的问题，尤其针对鲸类（特别是宽吻海豚，Tursiops truncatus）缺乏大规模、公开可获取的语音数据集这一关键瓶颈。当前海豚语音数据规模小、分散且多为封闭资源，制约了对其复杂通信系统结构的深入研究。为此，论文提出并发布了OpenWhistle——目前最大的公开海豚哨叫声数据集，包含约18万条哨叫声（总计114小时），来自一个由五只个体组成的稳定群体，在半自然环境中连续五年采集，并配套经过专家标注的8,354条精选哨叫声样本，以及可复现的哨叫声类型检测与分类评估协议。此外，研究还开源了完整的哨叫声检测、分割与分类处理流程。为验证其有效性，研究基于OpenWhistle预训练了一个适配海豚声学特性的Wav2Vec2.0模型，结果表明该模型学习到了有效的声学表征，在哨叫声类型检测与分类任务上优于通用生物声学模型（如AVES和BioLingual），同时仍保留显著的改进空间。因此，该研究的关键在于构建首个面向自监督学习的开放海豚哨叫声数据集及其配套工具链，为推动海豚通信研究及实现物种内精细声学结构建模奠定了基础。
链接: https://arxiv.org/abs/2609.34839
作者: Faadil Mustun,Chiara Semenzin,Roberto Dessi,Pablo Robin Guerrero,Pierre Orhan,Alexis Emanuelli,Emanuele Rossi,Yair Lakretz,Gonzalo de Polavieja,German Sumbre
机构: Institut de Biologie de l’École normale supérieure, CNRS, INSERM, Université PSL, Paris, France (法国巴黎高等师范学院生物学研究所，法国国家科学研究中心，法国国家健康与医学研究院，巴黎-萨克雷高等教育大学); Earth Species Project, France (地球物种项目，法国); Not Diamond, San Francisco, USA (Not Diamond，美国旧金山); Institut du Cerveau, Paris, France (大脑研究所，法国巴黎); Sapienza University of Rome, Rome, Italy (罗马第一大学，意大利罗马); École Normale Supérieure, Paris, France (法国巴黎高等师范学院，法国巴黎); Champalimaud Foundation, Lisbon, Portugal (查姆帕利马德基金会，葡萄牙里斯本)
类目: Computation and Language (cs.CL)
备注: Accepted as a Spotlight at the NeurIPS 2026 Datasets Evaluations Track
Abstract:Recent advances in bioacoustics have been driven by large-scale corpora and standardized benchmarks, yet existing resources are overwhelmingly bird-centric and shallow per species, limiting their use for studying the structure of a single species’ communication system. This gap is particularly acute for cetaceans: despite bottlenose dolphins (Tursiops truncatus) being a compelling case of complex vocal communication among non-human mammals, existing dolphin datasets are small, fragmented, and largely closed. We introduce OpenWhistle, the largest publicly available dataset of dolphin vocalizations. It comprises approximately 180,000 whistles (114 hours) recorded over five years from a stable pod of five individuals in a semi-natural environment, paired with a curated subset of 8,354 expert-annotated whistles and reproducible evaluation protocols for whistle-type detection and classification. We further release the full processing pipeline for whistle detection, segmentation, and categorization. To demonstrate its utility, we pretrain a Wav2Vec2.0 model adapted to dolphin acoustics on the OpenWhistle corpus and show that it learns effective representations, outperforming general-purpose bioacoustic models such as AVES and BioLingual on both tasks while leaving meaningful headroom for future work. By releasing the dataset, pipeline, and evaluation protocol, we provide the first open dolphin whistle dataset tailored for training self-supervised models, laying the groundwork for advancing dolphin communication research and developing models that capture fine-grained acoustic structure within species.
[NLP-73] DivOPD: Spread Wide Look Close for Asynchronous On-Policy Distillation of Multi-turn Agents
【速读】： 该论文旨在解决异步多轮训练中由于到达顺序批处理（arrival-order batching）导致的样本利用率不均问题：少数早期或长周期的回溯轨迹（rollout）会主导学习更新，而其他有效但延迟到达的轨迹则因过时而被浪费，造成已生成经验的损失。其核心解决方案是提出DivOPD（Diverse On-policy Distillation），一种基于学习器端的批量选择方法，通过将固定的训练轮次预算更均匀地分配给更多回溯轨迹，并在每条轨迹内优先选择师生间累积分歧较大的回合进行训练，同时排除无可用教师反馈的回合。该方法仅调整训练权重分配，保持每回合损失函数与优化器不变。对于无进展的回溯轨迹，可选扩展机制短暂交由教师接管以注入新信息后再返回学生。在ALFWorld、ScienceWorld和WebShop等模拟基准上，针对15亿至70亿参数的学生模型，在六组不同设置下，DivOPD将跨设置平均峰值成功率从77.4提升至84.4，最后五次评估的平均成功率从71.5提升至78.6，且相较原始OPD实现了1.84倍的训练令牌效率与1.87倍的学习器GPU时间加速；引入教师干预后，最后五次平均成功率进一步提升至82.4，仍保持约1.7倍的训练效率优势。
链接: https://arxiv.org/abs/2609.34838
作者: Hanyang Wang,Zeyuan Liu,Zhengyu Chen,Jingqing Ruan,Chaoxu Pang,Zhongda Su,Wulin Xie,Zhizhao Zeng,Ke Zeng,Tianxiang Zhao
机构: University of Chicago; Meituan LongCat Interaction Team; University of the Chinese Academy of Sciences; The Hong Kong University of Science and Technology (Guangzhou)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 24 pages, 9 figures, 19 tables. Code: this https URL
Abstract:On-policy distillation (OPD) trains student agents through teacher supervision on their own interactions with an environment. However, in asynchronous multi-turn training, arrival-order batching can allow a few early or long rollouts to dominate learner updates while other valid rollouts become stale before being used, wasting already-generated experience. To address this problem, we introduce DivOPD, a simple learner-side batch-selection method that spreads a fixed turn budget across more rollouts and, within each rollout, prioritizes turns with larger cumulative teacher-student disagreement. Turns without usable teacher feedback are excluded. The per-turn loss and optimizer remain fixed; selection only changes which student-visited turns receive training weight. For no-progress rollouts, an optional extension briefly hands control to the teacher before returning it to the student. Across six teacher-student settings on the simulated ALFWorld, ScienceWorld, and WebShop benchmarks, with 1.5B-7B students, DivOPD raises cross-setting mean peak success rate from 77.4 to 84.4 and mean success over the last five evaluations from 71.5 to 78.6. It reaches all reported setting-specific targets with geometric-mean speedups of 1.84x in training tokens and 1.87x in learner GPU time relative to vanilla OPD. Teacher intervention further raises this last-five mean to 82.4 while retaining about 1.7x learner-GPU speedup over vanilla OPD. Code will be released at this https URL.
[NLP-74] BV Loss: Block Verification-Aware Loss for Block Diffusion Speculative Decoding
【速读】： 该论文旨在解决生成式 AI（Generative AI）中扩散式草稿器（diffusion drafter）在推测解码（speculative decoding）过程中因训练目标与推理阶段验证机制不匹配所导致的效率瓶颈问题。现有方法多采用基于词元级（token-level）验证的训练目标，而实际推理中采用的是块级验证（block verification），二者存在语义层级上的不一致，限制了草稿序列的接受长度。为此，论文提出块级验证感知损失（Block Verification-aware loss, BV loss），其核心创新在于直接依据块级验证的接受规则构建训练目标，从而在训练阶段就最大化预期接受序列长度，实现了训练目标与推理时验证机制在序列层面的统一。实验表明，在数学、代码和对话等基准上，采用BV loss训练的DFlash与DSpark模型在不改变推理流程的前提下，相较于交叉熵损失提升了13.0%–21.0%的每轮验证平均接受词元数；同时优于词元级接受目标如TV loss和LK loss，且性能优势可延伸至词元级验证与贪心解码场景。结果证明，以序列级验证为导向的训练目标能显著提升扩散式草稿器的实用性与效率。
链接: https://arxiv.org/abs/2609.34832
作者: Suyoung Kim,Jahyun Koo,Hyeonjin Kim,Inhyeok Bang,Seunghyun Lee,Hyunjae Oh,Baeseong Park,Dongsoo Lee
机构: a2sys (a2sys.ai); Seoul National University (snu.ac.kr); University of Wisconsin (wisc.edu)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Diffusion drafters accelerate speculative decoding by proposing multiple tokens in parallel. Despite recent advances in speculative decoding through sequence-level drafting and verification, existing training objectives remain largely designed around token-level verification. To address this mismatch, we introduce Block Verification-aware loss (BV loss), a training objective designed to maximize the expected acceptance length of a drafted sequence. BV loss is directly derived from the block verification acceptance rule, providing a principled connection between the drafter training objective and the inference-time verification mechanism at the sequence level. Across math, code, and chat benchmarks, BV loss increases the mean number of tokens accepted per verification call under block verification by 13.0–21.0% over cross-entropy loss training for DFlash and DSpark with Qwen3-4B and Qwen3-8B without changing the inference procedure. BV loss also outperforms tokenwise acceptance objectives such as TV loss and LK loss, and its gains extend to token verification and greedy decoding. These results demonstrate the benefit of training block diffusion drafters with an objective aligned with sequence-level verification, rather than optimizing each token independently.
[NLP-75] From Weak Task Specifications to Scientific Extraction Agents : Optimizing Task Construction
【速读】： 该论文旨在解决科学文献提取代理（scientific extraction agents）在任务目标描述较短且缺乏预定义输出模式、信息抽取指令及评估标准的情况下，如何从仅包含简短任务目标和未标注参考文档的弱规范中自动构建任务特定配置的问题。传统方法通常假设这些组件已预先设定，但在实际科研场景中，手动制定既准确又全面的配置成本高昂且难以适应动态任务需求。本文提出的关键解决方案是：将任务特定的输出模式（schema）、抽取指令与基础训练评分标准的构建过程融入可迭代优化框架中，并在优化过程中保持其可编辑性。通过聚焦失败样本的文本梯度反馈（failure-focused updates），系统能够集中修正低分文档中的错误；同时，训练阶段的评估标准会根据反复出现的失败模式自适应调整。实验基于异质催化领域的文献语料库进行，在四种评判标准设置下，联合优化模式构建与抽取指令的表现最优，且消融实验与盲评人类评估均验证了该方法的有效性。
链接: https://arxiv.org/abs/2609.34829
作者: Zixiao Dong,Wei Yang,Zihao Liu,Chenshu Li,Longzhang Liu,Tao Tan,Hong Xie
机构: University of Science and Technology of China (中国科学技术大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Most methods that optimize LLM prompts and agent workflows assume that task-specific output schemas, extraction instructions, and evaluation criteria are predefined. For scientific extraction agents, however, a short task goal may not fully determine these components, while specifying them manually is costly. We study the upstream problem of constructing the task-specific configuration from a weak specification containing only a short goal and unannotated reference documents. Rather than treating automatic construction as a fixed preprocessing step, our framework constructs a task-specific schema, extraction instructions, and base training rubrics, then keeps schema construction and extraction instructions editable during optimization. Failure-focused updates concentrate textual-gradient feedback on lower-scoring documents, while training-time evaluation criteria adapt to recurring failures. On a heterogeneous-catalysis literature corpus, automatic construction remains improvable, and optimizing both schema construction and extraction instructions performs best across all four judge-rubric settings, with ablations and blinded human evaluation supporting the proposed formulation.
[NLP-76] Pass or Fail? Evaluating LLM s on Two Greek Examination Benchmarks
【速读】： 该论文旨在解决希腊语在大型语言模型（LLM）评估中因缺乏全面基准测试而面临的评估困境，尤其针对语言资源有限的场景。其核心解决方案是构建两个专为希腊语设计的基准测试——Prot-Ex与Pan-Ex，分别源自希腊模范学校与实验学校入学考试以及全国大学入学考试（Panhellenic exams），涵盖多个学术领域（如现代希腊语、数学、物理）及多样化任务形式（封闭式、结构化与开放式问题），并包含文本化视觉情境（即图像描述）。研究通过这些基准评估了包括本地化适配的KriKri-8B-Instruct在内的多种纯文本型LLM，发现经过针对性语言适配的KriKri-8B在语言密集型人文学科任务中显著优于其基础模型，甚至可媲美参数量更大的通用模型。研究进一步采用“以LLM为裁判”（LLM-as-a-Judge）方法揭示传统词法评估指标在复杂推理任务中的不足，并发现“少样本提示悖论”：尽管合成示例能提升封闭式问题的准确率，但在结构化任务中却因上下文窗口过载导致8B规模模型性能严重退化。研究最终表明，在特定领域内，有针对性的语言适配可有效弥补参数量不足的缺陷，但小模型对提示冗余仍表现出高度脆弱性。
链接: https://arxiv.org/abs/2609.34800
作者: Panagiota Kyriazi,Eleni Kasoura,Prokopis Prokopidis
机构: Institute for Language and Speech Processing / Athena RC (希腊语言与语音处理研究所/雅典研究中心)
类目: Computation and Language (cs.CL)
备注:
Abstract:The rapid advancement of Large Language Models (LLMs) imposes a thorough evaluation of their linguistic and analytical capabilities as well as constraints, particularly for a language with limited benchmark coverage such as Greek. To address the limited availability of comprehensive benchmarks in this domain, we introduce Prot-Ex and Pan-Ex, two benchmarks consisting of questions from entrance exams for Greek Model and Experimental schools as well as the Panhellenic exams (the Greek national university entrance examinations). These benchmarks are employed to assess the performance of text-only LLMs-including the Greek-adapted KriKri-8B-Instruct, Llama-3.1-8B, Gemma-4-26B, and Qwen-3-32B-across diverse academic disciplines (Modern Greek, Mathematics, Physics, etc.) and task formats (closed, structured, and open-ended), including textualized visual context (i.e., image descriptions). Our findings indicate the localized KriKri-8B significantly outperforms its base model, successfully rivalling much larger LLMs in linguistically demanding humanities tasks. By leveraging an LLM-as-a-Judge methodology, we expose the inadequacy of traditional lexical metrics for evaluating complex reasoning. Crucially, we uncover a few-shot prompting paradox: while synthetic examples improve accuracy in closed-ended questions, they severely overload the context window of 8B models in structured tasks, causing significant performance degradation. Ultimately, this study suggests targeted linguistic adaptation offsets lower parameter counts in specialized domains, despite the fragility of smaller models to prompt verbosity.
[NLP-77] InfiMed2: A Generalist Medical Multimodal Foundation Model from Contextual Evidence and Stability-Aware Supervision
【速读】： 该论文旨在解决医学多模态模型在持续预训练（CPT）与后训练阶段中，因数据设计缺乏阶段性适应性而导致的性能瓶颈问题。具体而言，医学数据在结构、粒度和信息密度上差异显著，其有效价值随训练阶段从广义知识获取向晚期知识巩固的演进而动态变化；同时，后训练阶段普遍依赖短文本形式的视觉问答任务，难以提供充分且一致的解释性监督信号。为此，本文提出InfiMed2，一个基于阶段感知数据设计的40亿（4B）与270亿（27B）参数通用型医学多模态基础模型。其关键解决方案包括：首先构建包含556.8亿词元的语料库，通过源特定处理融合广泛的临床知识与富含上下文的生物医学视觉证据；在CPT阶段，采用分阶段策略——先微调视觉编码器，再构建广域医学知识，最后在学习率衰减期过渡至以证据为导向的数据混合；在监督微调（SFT）阶段，通过答案稳定性、答案掩码重建及正确性约束选择等机制重构视觉问答响应，生成更具信息量与答案一致性的人工标注监督信号；此外，对4B模型进一步引入可验证奖励的强化学习（RLVR）优化。实验表明，InfiMed2-4B在五项医学多模态基准测试中达到66.73%的平均准确率（经RLVR优化），超越更大规模的Qwen3.5-9B模型；而InfiMed2-27B达到73.72%的最高分，为当前评估的开源权重模型中的最佳表现。
链接: https://arxiv.org/abs/2609.34798
作者: Guanghao Zhu,Zeyu Liu,Zhitian Hou,Pengkai Wang,Zhijie Sang,Shuo Cai,Yang Yu,Yuanyi Wang,Yanggan Gu,Congkai Xie,Jianmin Wu,Hongxia Yang
机构: The Hong Kong Polytechnic University(香港理工大学); InfiX.ai; PolyU-Daya Bay Technology and Innovation Research Institute(香港理工大学大亚湾科技与创新研究院)
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent medical multimodal models have benefited from larger corpora, broader modality coverage, and stronger reasoning-oriented training, yet effective data design across continued pretraining (CPT) and post-training remains challenging. Medical sources vary substantially in structure, granularity, and information density, and their utility shifts as training progresses from broad knowledge acquisition to late-stage consolidation. Meanwhile, post-training is often dominated by short-form visual question answering, providing limited supervision for informative and answer-consistent explanations. We introduce InfiMed2, a family of 4B and 27B generalist medical multimodal foundation models built around stage-aware data design. We curate a 55.68B-token corpus that combines broad clinical knowledge with context-rich biomedical visual evidence through source-specific processing. Our CPT pipeline first adapts the vision encoder, then builds broad medical knowledge, and finally transitions to an evidence-focused data mixture during learning-rate decay. For supervised fine-tuning (SFT), we regenerate visual question-answering responses using answer stability, answer-masked reconstruction, and correctness-constrained selection to produce more informative and answer-consistent supervision. The 4B model is further optimized with reinforcement learning with verifiable rewards (RLVR). Across five medical multimodal benchmarks, InfiMed2-4B achieves 66.73% mean accuracy after RLVR, surpassing the larger Qwen3.5-9B, while InfiMed2-27B reaches 73.72%, the highest among the evaluated open-weight models.
[NLP-78] QTS-Bench: A Multi-Syntax Benchmark for Text-to-Query over Time-Series Databases
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在时序数据库（Time-Series Databases, TSDBs）上进行自然语言查询能力评估不足的问题。现有基准测试无法充分涵盖TSDB特有的非统一查询语法、多样的应用领域以及独特的时序特定查询意图，导致对模型实际性能的评估存在偏差。为此，研究提出TQTS-BENCH，一个支持多语法的综合性基准，用于评估文本到查询（text-to-query）的能力。其关键创新在于构建了一个包含6,125个高质量问答对的基准数据集，覆盖97个不同的TSDB实例、23种查询语法、22个应用领域及4类时序特异性查询意图，并通过以人类为中心的AI辅助工作流确保数据质量，所有问答对均由领域专家审核与修订。实验结果表明，当前最先进的模型（Claude-Opus-5）在执行准确率上仅为48.98%，远低于人类水平（87.34%），主要误差源于不同TSDB间查询语法的异构性、对时间相关语义的误理解以及错误的模式链接。这一发现揭示了提升LLMs在真实场景下处理时序数据库查询能力的关键挑战与优化方向。
链接: https://arxiv.org/abs/2609.34783
作者: Fei Lyu,Zhiyi Peng,Jiaming Liu,Yixuan Yang,Changjian Chen,Zhuo Tang,Jiapeng Zhang,Kenli Li
机构: Hunan University (湖南大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large language models (LLMs) have significantly advanced natural language querying over relational databases, yet their ability to query time-series databases (TSDBs) remains largely unassessed. Existing benchmarks fail to adequately capture the non-unified query syntaxes, diverse application domains, and unique time-specific query intents inherent to TSDBs. To address this gap, we introduce TQTS-BENCH, a multi-syntax benchmark for evaluating text-to-query capabilities over TSDBs. TQTS-BENCH contains 6,125 high-quality question-answering (QA) pairs spanning 97 TSDBs, 23 distinct query syntaxes, 22 application domains, and 4 types of time-specific query intents. It is constructed through a human-centric AI-assisted workflow, where all QA pairs are carefully reviewed and revised by domain experts to ensure quality and correctness. Extensive evaluations of advanced LLMs and state-of-the-art text-to-query methods reveal challenges in querying TSDBs. Even the best-performing model evaluated, Claude-Opus-5, achieves only 48.98% execution accuracy, while humans reach 87.34%. Error analysis reveals that this performance gap mainly stems from the heterogeneous query syntaxes across different TSDBs, misinterpretation of time-specific intents, and incorrect schema linking. These findings highlight new opportunities to narrow the gap between current LLM capabilities and the requirements of TSDB queries in real-world applications. The benchmark is available at: this https URL.
[NLP-79] When Do Model Internals Help? Exploring the Role of Representation Engineering in LLM Safety
【速读】： 该论文旨在解决生成式 AI (Generative AI) 安全保障中控制与监测机制的协同优化问题，核心挑战在于明确行为对齐（behavioral alignment）与表征工程（representation engineering）两类方法在安全控制与风险监测中的相对效能。其解决方案的关键在于通过匹配评估框架，在相同实验设置下系统比较 DPO（一种行为对齐方法）与三种表征操控方法在鲁棒性、实用性及粒度上的表现，以及对比表征探测器与微调/开源文本监控器在完整响应检测、早期检测能力及计算成本方面的差异。研究发现：DPO 在多数场景下提供最强的安全控制能力且随训练数据增加而提升，但可能因后续良性微调导致安全性下降；而表征工程在低数据量、尤其是高质量对比数据条件下仍具竞争力。在监测方面，专用文本监控器具有最优检测精度，而表征探测器以极低边际成本保持良好性能。此外，基于监控器引导的干预策略可有效恢复 DPO 在微调后损失的安全性，且额外拒绝率极低。因此，表征工程虽不能普遍替代行为对齐，但在特定条件下具备实用优势，并能与传统方法形成互补，共同增强 AI 系统的整体安全性。
链接: https://arxiv.org/abs/2609.34771
作者: Tianyi Guan,Jianhui Chen,Liangming Pan
机构: Peking University (北京大学); School of Computer Science, Peking University (计算机学院，北京大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Reliable AI safeguards require both control mechanisms that reduce unsafe behavior and monitoring mechanisms that detect safety risks during model interactions. Established behavioral safeguards include alignment methods that optimize model outputs and text monitors that assess interaction text. Representation engineering instead reads or modifies internal model states, but the relative strengths of these approaches remain unclear because they are often evaluated under different settings. We present a matched evaluation across two tracks. For safety control, we compare DPO, a behavioral alignment method, with three representation steering methods across robustness, practicality, and granularity. DPO provides the strongest overall control and generally improves with increasing training data, although its safety can degrade after subsequent benign fine-tuning. Representation steering remains competitive primarily in low-data settings, particularly with high-quality contrastive data. For safety monitoring, we compare representation probes with fine-tuned and open-weight text monitors across full-response detection, early detection, and computational cost. Specialized text monitors achieve the strongest overall detection accuracy, while representation probes remain competitive at substantially lower marginal cost. Finally, monitor-guided interventions recover much of the safety lost by DPO after benign fine-tuning, with little additional over-refusal. Overall, representation engineering does not generally replace behavioral safeguards, but offers practical advantages under specific conditions and can provide complementary safety benefits.
[NLP-80] Reference-Grounded Data Curation for Instruction-Following Thai-English Machine Translation AACL
【速读】： 该论文旨在解决指令遵循型机器翻译（Instruction-Following Machine Translation, IF-MT）中规则遵从性与翻译质量之间的固有权衡问题，尤其针对术语、格式和语体等提示层面的约束。传统通用的指令遵循数据增强方法无法有效缓解这一矛盾。其解决方案的关键在于提出“参考文本引导的数据筛选”（Reference-Grounded Data Curation）两阶段流水线：第一阶段通过指令遵循难度（IFD）评分保留英文-泰文平行语料库中最具挑战性但可学习的实例；第二阶段从已满足约束的参考译文中提取所有监督式约束，并仅保留完全符合这些约束的生成结果，从而构建出包含197万条记录的Grounded数据集。基于该数据集对开放权重模型进行微调，得到参数量分别为40亿、20亿和8亿的泰语-英语翻译模型系列ChindaMT。在长度控制的成对评估中，ChindaMT在各类别下均优于或持平于同规模基线模型，对最强基线最高实现68.4%的胜率。该方法可无缝迁移至Qwen生成模型，且相关模型权重、数据集及评测套件均已开源。
链接: https://arxiv.org/abs/2609.34770
作者: Thodsaporn Chay-intr,Krittapad Harnchang,Mahannop,Thabua,Kobkrit Viriyayudhakorn,Thanaruk Theeramunkong
机构: iApp Technology(泰国); Intelligent Informatics and Service Innovation Research Center(泰国); Artificial Intelligence Entrepreneur Association of Thailand(AIEAT)(泰国); Sirindhorn International Institute of Technology, Thammasat University(泰国)
类目: Computation and Language (cs.CL)
备注: Accepted at AACL-IJCNLP 2026 (Main Conference)
Abstract:Instruction-following machine translation (IF-MT) requires respecting prompt-level rules on terminology, formatting, and register. Rule compliance typically trades off against translation quality, a tension that general-purpose IF data augmentation methods do not address. We propose Reference-Grounded Data Curation, a two-phase pipeline that extracts every supervised constraint from a reference translation that already satisfies it, ensuring feasibility by construction. Phase 1 applies Instruction-Following Difficulty (IFD) scoring to retain the hardest-but-learnable instances from an English-Thai parallel pool. Phase 2 extracts constraints from each reference target and keeps only generations satisfying every constraint, yielding the 1.97M-record Grounded dataset. We fine-tune open-weight bases on Grounded to produce ChindaMT, a Thai-English translation family at 4B, 2B, and 0.8B parameters. Under length-controlled pairwise judging, ChindaMT outperforms or matches every same-size baseline at every tier on both plain translation and under explicit rules, reaching up to a 68.4% win rate against the strongest baseline. The recipe transfers cleanly across Qwen generations. We release model weights, the Grounded dataset, and evaluation suites.
[NLP-81] LongPuzzleBench: Evaluating GUI Agents on Long-Horizon Visual Puzzles
【速读】： 该论文旨在解决生成式 GUI 代理在长时程视觉推理任务中缺乏连贯性的问题，即代理在执行多步操作时难以维持长期规划的一致性，导致看似合理的当前动作可能破坏后续可解性。其核心挑战在于：在界面状态持续变化的场景下，早期动作对后期动作具有约束性，而现有基准测试往往忽视了这一耦合决策链的长期一致性评估。为应对该问题，研究提出 LongPuzzleBench，一个包含114个关卡的六种图形化谜题基准，所有关卡均通过原生图形用户界面（Native GUI）操作完成，部分目标需人类玩家超过千次操作，且死胡同无显式提示。实验发现，尽管最强代理在简单关卡上表现良好，但在更复杂、更长的关卡上成功率急剧下降——七类通用代理无法突破“中等”难度，且无人能完成“螺栓拆卸困难”关卡（人类可轻松完成）。进一步分析表明，即使引入代码执行与上下文理解（Code Execution CUA），仍无法弥合性能差距，且其评分混杂了视觉求解与算法搜索能力。受控诊断揭示根本原因在于：当前代理仅依据当前可见进展评估动作价值，而非评估该动作对未来可行选项的保留程度，这一局限性无法通过规则、状态提示或失败记忆等方式克服。
链接: https://arxiv.org/abs/2609.34769
作者: Bingo Zhang,Haochuan Lu,Zongjie Li,Genjian Li,Ari Yu Zhang,Chaozheng Wang
机构: Vera Praxis; Tencent; The Hong Kong University of Science and Technology; The Chinese University of Hong Kong
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:GUI agents need long-horizon visual reasoning: they must interpret a changing interface while keeping a multi-step plan viable as earlier actions constrain later ones. Existing benchmarks evaluate grounding, computer use, and game play, but rarely test whether agents stay coherent across long chains of coupled decisions. Long-horizon visual puzzles expose this capability directly: a legal move that looks like progress can make the puzzle unsolvable, and the loss shows only several moves later. We introduce LongPuzzleBench, 114 levels in six puzzle games played through native GUI actions, where one objective can take a human over a thousand actions on persistent boards and dead ends go unannounced. With Native GUI Actions alone, the strongest agents solve most objectives, but success falls sharply on harder, longer boards: seven of ten general-purpose agents solve nothing harder than Medium, and none completes Bolt Unscrew Hard, which a human solves along with every other objective. Code Execution CUA does not close this gap, and its scores mix visual solving with algorithmic search. Controlled diagnostics trace these failures to one limitation that neither rules, state hints, nor failure memory removes: agents judge each move by the visible progress it makes, not by the future options it leaves.
[NLP-82] Draft-KV: Learning Useful Latent Communication Between Language Models
【速读】： 该论文旨在解决生成式 AI (Generative AI) 中隐式通信（latent communication）存在的核心问题：接收方的性能提升可能并非源于对发送方信息内容的实际利用，而是由于接口设计本身提供了性能增益，导致信息传递的有效性难以验证。其解决方案的关键在于提出一种名为 Draft-KV 的新型通信机制，该机制不直接传递解码后的文本或随机扰动的隐状态，而是将发送方在草拟当前问题答案过程中形成的键值（key-value）状态作为通信内容。通过线性投影将这些状态置于侧边记忆中，并借助带有门控注意力分支的读取路径进行访问；同时采用渐进式训练策略，在保证不因消息错配造成显著损害的前提下，逐步从消息重建任务过渡到答案监督任务。该方法仅需训练约105万参数，远低于对比模型C2C（348倍更少），且保持双方模型冻结。实验表明，当使用Qwen3-8B作为发送方、冻结的Qwen2.5-0.5B-Instruct作为接收方时，在MMLU-Redux上准确率可达78.04%，显著优于单独使用接收方的37.45%以及消息被重分配后的36.40%。此外，随着发送方规模从0.6B扩展至8B，性能持续提升，且通信能力可泛化至未见任务，甚至在双方持有不同证据时超越单个模型表现。
链接: https://arxiv.org/abs/2609.34754
作者: Linquan Wu,Shichang Meng,Tianxiang Jiang,Haoyu Yang,Peng Zhong,Fengming Zhu,Xi Peng,Linqi Song,Jacky Keung,Jingyu Zhang
机构: City University of Hong Kong (香港城市大学); University of Science and Technology of China (中国科学技术大学); University of Electronic Science and Technology of China (电子科技大学); AIPD, Tencent (腾讯AI平台部); Theory Lab, Huawei (华为理论实验室)
类目: Computation and Language (cs.CL)
备注: 41 pages, 7 figures, 13 tables. Code: this https URL
Abstract:Latent communication passes internal states between language models instead of decoded text, but higher receiver accuracy does not show that the receiver used the message content. Across five method-dataset pairs, replacing each message with one from an unrelated question changes accuracy by at most 0.60 points, even when communication adds 15.44 points over the receiver alone. Thus the interface can supply the gain while making the sharer dispensable. Draft-KV instead sends the key-value states formed while the sharer drafts an answer to the current question. Linear projections place these states in a side memory read through a gated attention branch, and progressive training moves from message reconstruction to answer supervision under a guard on harm from mismatched messages. Both models remain frozen and the interface trains 1.05M parameters, 348x fewer than C2C. With a Qwen3-8B sharer, a frozen Qwen2.5-0.5B-Instruct receiver reaches 78.04% on MMLU-Redux, versus 37.45% alone and 36.40% with reassigned messages. At fixed interface size, scaling the sharer from 0.6B to 8B raises accuracy from 46.11% to 78.04%; communication also transfers to held-out tasks and can exceed both models when each holds different evidence.
[NLP-83] Beyond Token Alignment: Event Completion for Cross-Tokenizer On-Policy Distillation
【速读】： 该论文旨在解决跨分词器（cross-tokenizer）语言模型知识蒸馏中，学生模型在生成部分教师标记（teacher token）时所面临的概率分配不明确问题。具体而言，当学生模型生成了教师标记的前缀但尚未完成时，存在多个合法的后续标记组合可共同完成剩余字节，而教师仅指定最终完成目标，未提供如何在这些有效延续之间分配概率的指导，导致现有方法在概率对齐上存在偏差。为此，论文提出事件集完成蒸馏（Event-Set Completion Distillation, ESCD），其核心在于引入完成集监督（completion-set supervision），即聚合与前缀相关的教师端事件，并对所有字节兼容的一步完成选项的总概率进行监督，从而避免依赖分词器的局部概率拆分。该方法复用学生生成的轨迹与预测结果，无需额外采样或修改学生词汇表，具有高效性和通用性。实验表明，ESCD在数学、代码和科学推理任务中均实现稳定性能提升，适用于不同模型架构与分词器，且在大规模混合专家（MoE）蒸馏场景中表现优异。局部分析显示，保留完成集能更准确匹配参考监督信号，且一步完成覆盖超过99%的可观测教师概率质量，验证了事件进入（event entry）与事件完成（event completion）作为互补监督目标的有效性。
链接: https://arxiv.org/abs/2609.34738
作者: Jiacheng Liu,Jingwei Song,Qituan Zhang,Siheng Chen,Linfeng Zhang
机构: Shanghai Jiao Tong University(上海交通大学); Fudan University(复旦大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 43 pages, 7 figures, 20 tables
Abstract:On-policy distillation (OPD) transfers knowledge between language models through teacher supervision on student-generated trajectories. With different tokenizers, a single teacher token may require multiple student tokens to generate, creating intermediate states where the event is entered but not yet completed. Existing cross-tokenizer methods align tokens or text spans to construct comparable prediction targets. We study a complementary problem after partial generation: once the student produces a prefix of a teacher token, multiple next tokens may complete the same remaining bytes, but the teacher only specifies the required completion rather than how probability should be divided among these valid continuations. We introduce Event-Set Completion Distillation (ESCD), which complements cross-tokenizer probability alignment with completion-set supervision. ESCD aggregates prefix-related teacher events and supervises the total probability of byte-compatible one-step student completions, avoiding tokenizer-dependent probability splits among individual tokens. The method reuses student trajectories and predictions, requiring neither additional rollouts nor changes to the student vocabulary. Experiments demonstrate consistent gains in mathematics, code, and scientific reasoning across model families and tokenizers, extending to large-scale MoE distillation from a 1T teacher to a 35B student. Local analyses show that retaining completion sets better matches the reference supervision, while one-step completion covers over 99% of observed compatible teacher mass after partial event entry in the studied tokenizer pairs. These findings support event entry and event completion as complementary supervision targets for cross-tokenizer knowledge transfer. Code will be released on GitHub.
[NLP-84] SeLMRoute: Probabilistic Semantic Evidence for Large Language Model Routing
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）路由中因缺乏对查询实际需求的显式语义理解而导致的模型选择不精准问题。现有方法通常依赖查询嵌入、模型表征或偏好数据进行直接决策，但这些表示难以明确揭示查询的真实需求。其解决方案的关键在于提出SeLMRoute框架，通过分离候选模型无关的语义证据提取与候选模型性能评估及部署目标应用，实现更可解释和灵活的路由决策。具体而言，首先由一个决策模型对查询生成一系列可解释的问题（如推理需求、外部知识使用等），并以概率分布形式保留判断结果，形成概率化的语义状态；随后，轻量级监督路由器基于该语义状态估计各候选模型的性能表现；最后，在性能估计后引入路由目标（如性能优化或成本控制），使得同一语义状态可支持多目标决策。在LLMRouterBench基准测试（15个数据集、20个候选模型、11,481个查询）上，SeLMRoute平均准确率达72.08% ± 0.45，五折交叉验证下达72.64%，显著优于最强固定候选模型的69.23%；同时在13模型性能-成本设置中，所有分组均实现性能提升，平均性能增益为2.66%。实验表明，该语义表示在各类表征中达到最优平均性能。
链接: https://arxiv.org/abs/2609.34736
作者: Vasilis Perifanis,Nikolaos Pavlidis,Symeon Symeonidis
机构: Indigma Innovations(Indigma创新公司); Democritus University of Thrace(色雷斯大学); Athena Research Center(阿娜研究中⼼)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Large language model (LLM) routing aims to select the most suitable model for each incoming query. Most existing routers learn this decision directly from query embeddings, model representations, preference data, or clusters of similar examples. Such approaches can be effective, yet the representation used for routing rarely states what a query actually requires. We introduce SeLMRoute, a routing framework that separates the extraction of candidate-independent semantic evidence from the learning of candidate performance and the application of deployment objectives. A decision model first evaluates a set of interpretable questions about the query, such as its reasoning requirements and use of external knowledge, with each judgment retained as a probability distribution. The resulting probabilistic semantic state is used by a lightweight supervised router to estimate candidate model performance. Routing objectives are applied after performance estimation, which allows the same semantic state to support performance-oriented and cost-aware decisions. On the LLMRouterBench (15 datasets, 20 candidate models, 11,481 queries), SeLMRoute achieves an average accuracy of 72.08% \pm 0.45 , while grouped five-fold out-of-fold evaluation reaches 72.64% , compared with 69.23% for the strongest fixed candidate. The representation achieves the highest mean performance among the evaluated semantic, dense, lexical, and domain-level representations. In a separate 13-model performance-cost setting, SeLMRoute improves performance in all five grouped splits, with a mean PerfGain of 2.66% . Our code is available at this https URL.
[NLP-85] Quality Determines Direction Length Shapes Magnitude: Length Control for Open-Ended Reinforcement Learning
【速读】： 该论文旨在解决生成式语言模型在强化学习（Reinforcement Learning, RL）微调过程中出现的响应长度不可控增长问题，尤其在开放式任务中，响应长度与生成质量高度耦合，且缺乏明确的成功边界来判断何时应优先考虑效率。此外，密集且连续的奖励信号导致不同响应间的质量差异较小，使得质量带来的优势极易被奖励层面的长度调节所干扰，甚至发生方向反转。为此，论文提出一种不对称原则：以生成质量决定强化的方向，而响应长度仅影响强化的幅度。其核心解决方案为质量门控长度优势调节（Quality-Gated Length Advantage Shaping, QGLAS），该方法首先基于质量奖励计算优势值，随后仅对较短且具有正优势的响应添加有界奖励增益，其余情况保持原优势不变；同时，增益强度根据组内质量差异自适应调整，当高质量响应间差异较小时增强简洁性激励，反之则减弱。实验表明，QGLAS在多种模型架构、开放式评估基准及奖励来源下均显著优于现有基线，在约30%的长度压缩率下，仍可保留质量-长度权衡中98.4%–102.0%的宏观平均质量提升，远超基线的68.3%–75.5%。
链接: https://arxiv.org/abs/2609.34718
作者: Zijun Weng,Zhongan Bi,Xuanang Gao,Xiaohui Hu,Shuangyong Song,Yongxiang Li,Kaidong Yu,Xuanjing Huang
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 22 pages. Preprint, under review
Abstract:Reinforcement learning (RL) changes not only what language models say, but also how much they say, often increasing response length at the cost of token efficiency. Controlling this length growth is particularly challenging in open-ended RL because (i) response length is entangled with quality, (ii) open-ended tasks lack a natural success boundary for deciding when efficiency should be prioritized, and (iii) dense, graded rewards often yield small within-group quality margins, making quality-induced advantages especially sensitive to reward-level length shaping, which can perturb their magnitudes and even reverse their signs. We therefore adopt an asymmetric principle: quality should determine the direction of reinforcement, while length should only shape its magnitude. We instantiate this principle with Quality-Gated Length Advantage Shaping (QGLAS), which first computes advantages from quality rewards alone, then adds bounded bonuses only to shorter positive-advantage responses, leaving all other advantages unchanged. The bonus strength is further adapted to within-group quality separation, allowing conciseness to matter more when quality-favored responses are similar and less when their quality differences are clear. Across different model families, open-ended benchmarks, and reward sources, QGLAS consistently achieves a stronger quality–length trade-off than representative baselines. At approximately 30% compression, QGLAS retains 98.4–102.0% of the macro-average quality gains achieved by quality-only RL over the base model, compared with 68.3–75.5% for these baselines at comparable compression.
[NLP-86] ReMCTS: Reflection-Enhanced Monte Carlo Tree Search for Code Generation EMNLP2026
【速读】： 该论文旨在解决开放权重大语言模型（LLM）在从自然语言提示生成函数级程序时，因无法准确捕捉隐藏语义而反复犯错的问题。现有方法在修复过程中缺乏对失败经验的积累与跨分支上下文的利用，导致修复效果不佳。其解决方案的关键在于提出ReMCTS——一种基于执行反馈、引入记忆增强机制并由大语言模型引导的蒙特卡洛树搜索（MCTS）框架。该框架将程序候选解组织为树状状态，保留分支局部的调试上下文，跨分支检索失败经验，并区分失败检查与证据缺失情况，从而实现更有效的程序修复。实验表明，在HumanEval和MBPP-Sanitized数据集上，采用可见测试评估的ReMCTS在10组模型-数据集组合中，有8组优于直接生成方法；而仅依赖代理搜索的策略则表现不稳定。通过控制变量实验进一步揭示了树搜索、采样、修复及记忆机制对性能提升的具体贡献及其局限性。此外，针对C++的30任务HumanEval-X初步验证了该框架与编译器驱动执行环境的兼容性，但尚未完成广泛的多语言评估。
链接: https://arxiv.org/abs/2609.34717
作者: Huifei Wang,Xinying Huang,Yiheng Sun,Yifan Yuan
机构: Shenzhen University (深圳大学); Shenzhen University (深圳大学); Shenzhen University (深圳大学); 未知
类目: Computation and Language (cs.CL)
备注: 21 pages, 2 figures. To appear in the Proceedings of EMNLP 2026
Abstract:Open-weight large language models (LLMs) can generate function-level programs from natural-language prompts, but plausible candidates still fail on hidden semantics and repeat mistakes across repair attempts. We present ReMCTS, an execution-grounded, memory-augmented, LLM-guided MCTS-style search framework. It organizes program candidates as tree states, retains branch-local debugging context, retrieves failure experience across branches, and distinguishes failed checks from unavailable evidence. On HumanEval and MBPP-Sanitized, visible-test ReMCTS improves over direct generation in 8 of 10 model-dataset pairs under held-out evaluation, whereas proxy-only search is less stable. Controlled tree-search, sampling, repair, and memory ablations characterize the source and limits of these gains. A 30-task HumanEval-X C++ pilot further demonstrates compatibility with compiler-backed execution, but does not constitute a broad multilingual evaluation.
[NLP-87] Using LLM s to Detect LLM -Generated Texts: A Cross-Generation Analysis
【速读】： 该论文旨在解决生成式文本检测（LGTs）在跨模型与跨领域场景下泛化能力不足的问题，尤其关注大语言模型（LLM）作为检测器时在自检测（self-detection）与交叉检测（cross-detection）中的行为差异。其核心解决方案在于系统评估15个覆盖三代模型的生成器与检测器，构建包含1,000篇人工写作文本与15,000篇生成式文本（每模型1,000篇）的基准数据集，并收集超过23.3万条二分类判断及自然语言解释，以揭示检测性能的本质驱动因素。研究发现，检测效能主要由检测器自身能力决定，而非生成器来源；尽管新一代生成器的输出更难被检测，但不同模型在自检测任务中并无系统性优势或劣势。进一步分析揭示代际偏差：第一代检测器存在高假阴性率，第二代检测器则表现出高假阳性率，而最新模型实现了更优的平衡。此外，研究还指出不同模型在依据文本线索进行决策解释时存在显著不一致性，凸显了当前检测机制的不可靠性与潜在偏见。
链接: https://arxiv.org/abs/2609.34691
作者: Haiyue Yuan,Jie Guo,Weidong Qiu,Zheng Huang,Ruizhe Li,Shujun Li
机构: Institute of Cyber Security for Society (iCSS) School of Computing, University of Kent, UK; School of Cyber Science and Engineering, Shanghai Jiao Tong University, China; School of Computer Science, University of Birmingham, UK
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Preprint
Abstract:Automated detection of LLM-generated texts (LGTs) is critical, yet dedicated detectors often struggle to generalize across domains and models. While general-purpose LLMs offer flexible zero-shot authorship classification with explanatory rationale, their detection behavior, especially regarding self-detection versus cross-detection across model generations, remains poorly understood. We systematically evaluate 15 LLMs spanning three model generations as both generators and detectors. Using a benchmark of 1,000 human-written texts and 15,000 LGTs (1,000 per model), we collected over 233,000 binary classifications alongside natural-language explanations. Our results reveal that detection efficacy is primarily driven by detector capability rather than generator provenance, although outputs from newer generators remain notably harder to detect. Crucially, statistical comparisons show no systematic advantage or disadvantage for self-detection across models. Error analysis further exposes generational bias shifts: first-generation detectors under-detect LGTs (high false-negative rates), second-generation detectors over-flag human texts (high false-positive rates), and the latest models achieve balanced trade-offs. Finally, we highlight significant inconsistencies in how different LLMs apply textual cues to justify their decisions. Code: this https URL.
[NLP-88] Fair Fact-Checking: Closing the Cross-Lingual Gap in LLM Factual Judgement with RoSh
【速读】： 该论文旨在解决生成式 AI 在多语言环境下对事实性陈述判断能力不均衡的问题，特别是针对模型在非英语语境下表现显著下降的现象。现有方法主要通过增加多语言训练数据或构建语言表示间的非约束映射来缓解此问题，但忽略了模型可能已具备正确答案却无法准确输出的核心机制缺陷。研究发现，尽管模型在低资源语言（如阿拉伯语）上表现接近随机猜测，其内部激活状态中仍隐含着正确的信息——通过线性探测即可恢复真实答案。为此，作者提出一种名为 RoSh（Per-language Shift and Rotation of the residual stream）的无训练、无需修改权重的干预方法，通过对残差流在三个层级进行闭式计算的逐语言平移与旋转，实现对模型输出空间的适配。实验表明，RoSh 能显著提升所有模型在各语言上的表现，平均缩小 75% 的语言性能差距；尤其在性能最差的场景（如 Llama-3B 阿拉伯语）中，从随机水平提升至接近英文基准水平，并使约五分之一原本在英文中可正确回答的事实性陈述在翻译过程中不再丢失。后续分析显示，经过 RoSh 处理后，模型分类头对非英语语言编码信息的恢复能力与英语相当，而对比其他非约束映射方法表现更优，验证了正交性约束的有效性。此外，该方法在多个控制实验及与最先进的推理时干预方法（latent-space intervention）的对比中均展现出显著优势，其增益幅度达到后者的 5 至 13 倍。因此，解决方案的关键在于识别并纠正模型“读出失败”而非“知识缺失”的根本机制，通过保持模型参数不变的前提下，施加可解释且高效的结构化变换以实现跨语言推理能力的均衡化。
链接: https://arxiv.org/abs/2609.34678
作者: Muhammad Ahmad,Fatemeh Seyedin,Adrian Weller,Dongwon Lee,Mahmoudreza Babaei
机构: Shifa Tameer-e-Millat University (伊斯兰堡谢法·塔米尔-艾米尔大学); BRAINS, Brandenburg Research Center for Applied Intelligent Systems (勃兰登堡应用智能系统研究中心); Max Planck Institute for Security and Privacy (马克斯·普朗克信息安全与隐私研究所); University of Cambridge (剑桥大学); The Pennsylvania State University (宾夕法尼亚州立大学); GISMA University of Applied Sciences (GISMA应用科学大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 23 pages, 3 figures
Abstract:Misinformation on social media remains a critical problem, and more and more people settle it by asking a language model instead of a fact checker. Whether models judge such claims reliably is debated; whether they judge them equally well in every language people ask in has gone almost unasked. We test eight models from five families, 3B to 70B, on 1,500 encyclopedic factual claims that exist in identical form in eight languages. English is judged better than every other language on every model, and the gap is widest on the smallest ones, where Llama-3B on Arabic is no better than guessing. Existing remedies retrain on more multilingual data or fit an unconstrained map between language representations, and neither asks whether the model already holds the answer and simply fails to say it. It largely does: a linear probe recovers the truth from the very activations the model fails to express. We propose RoSh, a per-language shift and rotation of the residual stream, computed in closed form at three layers, with no training and no weight modified. It improves every model and closes 75% of the gap on average, helping most where the model was worst: Arabic on Llama-3B goes from chance to nearly the English level, and a fifth fewer of the claims answered correctly in English are lost in translation. What remains is no longer a read-out failure: afterwards the head recovers as much of what is encoded outside English as it does in English. An unconstrained map fitted on the same pairs falls below the untouched baseline, so the orthogonality constraint is doing the work, and every model clears a scrambled-correspondence control and ten further controls. On the two benchmarks of the closest inference-time method, latent-space intervention, run with its own data and metric code, RoSh’s gains are five to thirteen times larger.
[NLP-89] Rewarding Novel Deductions: Solver-guided Process Rewards for Logical Reasoning
【速读】： 该论文旨在解决大语言模型（LLM）在处理结构化逻辑推理任务时面临的挑战，尤其是小规模模型在推理过程中易出现不一致、冗余或脆弱的推理轨迹问题。现有方法多聚焦于最终答案的正确性，对中间推理过程缺乏有效的过程级监督。为此，本文提出SPRING（Solver-guided Process Rewards for Novel LogIcal ReasoNing Step Generation），其核心创新在于利用SMT求解器作为训练阶段的中间推理步骤验证器，实现对推理过程的精细化监督。SPRING的关键在于引入“新颖推理步骤”的概念——即逻辑有效、与当前推理状态一致且未被已有非矛盾推论所蕴含的推理步骤。基于此，该方法设计了过程奖励机制，鼓励模型生成具有新信息量的推理进展，同时惩罚矛盾和无意义的推理步骤。实验结果表明，SPRING在三个逻辑推理基准（ZebraLogic、AR-LSAT、Knights and Knaves）上显著优于基线模型及仅依赖最终结果反馈的方法，在多个指标上实现了大幅性能提升。
链接: https://arxiv.org/abs/2609.34660
作者: Muhammad Asif Ali,Wenqing Wang,Huan Wang,Mohammad Raza
机构: FORTE Lab; Information Technology University (信息科技大学), Lahore, Pakistan; College of Informatics, Huazhong Agricultural University (华中农业大学信息学院), Wuhan, China; Qatar Computing Research Institute, Hamad Bin Khalifa University (哈马德·本·哈利法大学), Doha, Qatar
类目: Computation and Language (cs.CL)
备注:
Abstract:Logical reasoning remains a major challenge for large language models (LLMs), particularly on structured problems that require precise constraint tracking, consistency preservation, and multi-step deduction. This challenge is especially acute for small-scale LLMs, which are more prone to producing inconsistent, redundant, or brittle reasoning trajectories. Existing approaches for improving logical reasoning largely optimize for final-answer correctness, providing only weak supervision over the intermediate reasoning process. In this work, we propose SPRING: (Solver-guided Process Rewards for Novel LogIcal ReasoNing Step Generation). SPRING uses SMT solver as a training-time verifier of intermediate reasoning steps to provide process-level supervision. It introduces the notion of a novel reasoning step, namely, a step that is logically valid, consistent with the evolving reasoning state, and not already implied by previously accepted non-contradictory deductions. Based on this solver-based assessment, it designs process rewards that encourage novel inferential progress while penalizing contradictory and uninformative reasoning steps. Evaluation across three logical reasoning benchmarks, ZebraLogic, AR-LSAT, and Knights and Knaves, and four LLMs shows that SPRING consistently outperforms base LLMs, outcome-only reward baselines, and Logic-LM. On ZebraLogic, SPRING improves puzzle accuracy by up to 49.71 and 15.43 points over the base LLM and strongest outcome-only baseline, respectively. On AR-LSAT, it improves overall accuracy by up to 64.93 and 12.14 points, respectively. On Knights and Knaves, SPRING achieves up to 93.14 puzzle accuracy and 96.05 person accuracy.
[NLP-90] When Can Attention Heads Be Statically Defined?
【速读】： 该论文旨在解决大模型训练中注意力机制计算冗余的问题，具体表现为多个注意力头（attention head）在不同输入间学习到相似的注意力模式，导致重复的查询-键分值计算与Softmax操作造成高昂的计算开销。其核心解决方案是提出选择性注意力冻结（Selective Attention Freezing, SAF），通过识别注意力模式方差较低的头，在训练中期将其注意力权重替换为拟合后的后Softmax均值，并以绝对位置和相对距离偏好形式存储这些固定模式，将存储复杂度从序列长度的二次方降低至线性。结合融合内核（fused kernel）实现普通注意力头与冻结头的协同执行，显著提升训练效率。实验表明，在124M参数、4K上下文设置下，替换25%的注意力头可使优化器更新速度提升1.056倍，仅带来0.77%的困惑度增加；在1B参数、8K上下文及四卡分布式环境下，更新速度提升1.068倍，且在长序列微调与因果预填充任务中表现更优。经关联回忆适配后，该方法在固定长度下对更多键值对的泛化能力优于标准注意力及两种剪枝控制策略。
链接: https://arxiv.org/abs/2609.34650
作者: Weixian Waylon Li,Yintao Tai,Marcio Fonseca,Shay B. Cohen
机构: University of Edinburgh (爱丁堡大学); Chamber of Deputies (巴西众议院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Some attention heads learn similar patterns across inputs. Reusing these patterns could reduce training cost by avoiding repeated query-key score computation and softmax. Through controlled pretraining comparisons, we identify Selective Attention Freezing (SAF), which selects heads with low attention-pattern variance and replaces their attention weights with fitted post-softmax means halfway through training. We represent these fixed patterns with absolute-position and relative-distance preferences, reducing storage from quadratic to linear in sequence length. A fused kernel reconstructs the patterns and executes ordinary-attention and replaced heads together. At matched training-token budgets, replacing 25% of attention heads gives 1.056x faster post-replacement optimiser updates at 124M parameters and 4K context, with a 0.77% perplexity increase. At 1B and 8K context, post-replacement updates are 1.068x faster on four GPUs including communication, with a 0.51% perplexity increase. The resulting models also accelerate long-input finetuning and causal prefill. After associative-recall adaptation, the 124M model with 25% replacement generalises to more key-value pairs at a fixed length better than ordinary attention and two pruning controls.
[NLP-91] After the Fix: How Corrected Agent Histories Transfer to Related Tasks
【速读】： 该论文旨在解决“修复失败的执行片段（episode）是否能使其经验成为后续任务中更有效的记忆”这一核心问题。其关键解决方案在于通过对比同一源任务在修复前后对固定目标任务的迁移效果，系统评估修复带来的记忆增强作用。研究发现，尽管修复后的任务在特定条件下表现出显著的性能提升（如ThinkingBox的全量修正模式较独立执行提升44个百分点），但其中大部分收益源于未修复状态下的较差表现，而非修复后记忆质量的实质性改善；进一步分析表明，12个由全量修正带来的15点性能差距中，有大量来自原始表现更差，而非修复后记忆更具可重用性。此外，无论是ThinkingBox还是APEX框架，均未在群体或全局层面建立稳定的修正优势，且文本型APEX的执行成功率虽高于摘要型（52% vs. 40%），但缺乏稳健的全局优越性。研究揭示：修复经验的价值与复用经验的价值本质上不同——有效记忆更新需同时具备“前版本参照”和“全新起点参照”两个条件，即既要有历史状态作为参考，又需具备独立重启的能力以避免累积误差。
链接: https://arxiv.org/abs/2609.34603
作者: Yanfei Zhang,Xu Lin
机构: Independent Researcher; International Digital Economy Academy (国际数字经济发展研究院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Software Engineering (cs.SE)
备注:
Abstract:Does repairing an episode make its experience a better memory for the next task? We transfer the same failed source before and after accepted repair to a fixed target, alongside independent execution. Our 3,300 runs cover 100 ThinkingBox pairs and the same 100 APEX pairs with and without source-state inheritance, under eleven conditions. ThinkingBox’s Full/Skill/Hybrid correction gains are 44/29/32 percentage points, with corrected performance 25/22/18 points above independence; inference weakens at the task-family level. Yet 12 of Full’s 15-point larger correction gap over Skill come from worse uncorrected performance, not better corrected memory. Moreover, 22 of Full’s 46 upward transitions restore observed baseline success. Neither APEX regime establishes comparable aggregate correction benefits. Action evidence connects workflow gains with reusable obligations and convention conflicts with source-local choices. Text APEX’s accepted execution reaches 52% versus its summary’s 40%, without robust global/group-level superiority or an estab- lished advantage over independence. Smaller handoffs reduce input but increase calls. The value of repairing experience is therefore distinct from the value of reusing it: memory updates require both a previous-version reference and a fresh-start reference.
[NLP-92] he Model Knows When to Stop: Training-Free Early Stopping for Long-Context Reading
【速读】： 该论文旨在解决语言模型在处理长输入时因持续读取冗余信息而导致计算浪费的问题。现有停止机制通常依赖内部激活状态学习充分性或通过训练退出门控机制，但这些方法存在训练成本高或泛化能力有限的缺陷。本文提出无需训练的**答案收敛停止（Answer-Convergence Stopping, ACS）**策略，其核心在于不通过提问方式判断是否足够，而是直接探测冻结模型在每一段输入后的输出状态，当模型输出的答案在置信度和稳定性上均达到阈值时即停止推理。该方法仅需输出端生成结果与词元概率，无任何可训练组件，且适用于多种模型与基准测试。由于过早停止可能引入误差，研究采用证据位置作为评估指标来衡量停止决策质量。在LongBench-v2全集及两个前沿模型上的实验表明，ACS是唯一能保持或超越完整阅读精度的停止策略；在250个S-NIAH问题上，五种不同模型的过早停止率仅为0%至12%，显著优于基于语义化门控的8.4%至45.6%。结果表明，通过有效利用冻结模型的输出信号，可在不增加训练负担的前提下实现自适应停止等理想行为。
链接: https://arxiv.org/abs/2609.34590
作者: Muath Alyobi,Mohamed Eltahir,Almoayyad Abuljdail,Riyadh Almutawa,Tanveer Hussain,Naeemullah Khan
机构: King Abdullah University of Science and Technology (KAUST), Thuwal, Saudi Arabia; Edge Hill University, Ormskirk, England
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Language models often process long inputs sequentially in chunks, but continuing to read after sufficient evidence has been acquired wastes computation. Existing stopping mechanisms either learn sufficiency from internal activations or train an exit gate, while a simpler alternative asks the model whether it has read enough. We introduce Answer-Convergence Stopping (ACS), a training-free stopping rule that measures rather than asks. After each chunk, it probes the frozen model’s current answer state and stops when that state is both confident and stable. The rule requires only output-side generation and token log probabilities, has no trained components, and uses one shared configuration across models and benchmarks. Because a stopping policy can save computation simply by stopping too early, we evaluate the stopping decision itself using evidence position where available. On the full LongBench-v2 with two frontier models, ACS is the only stopping policy that matches or exceeds full-reading accuracy. Furthermore, across 250 S-NIAH questions, the premature stopping rate for ACS across five models from two families ranges from 0% to 12%, compared to 8.4% to 45.6% for the verbalized gate. Taken together, ACS reveals that by properly utilizing the output signals of frozen models, we can achieve favorable behaviors like adaptive stopping without the need for additional training.
[NLP-93] In-game Toxic Detection: Bi-directional Representations with Attention Residuals AAAI2023
【速读】： 该论文旨在解决游戏中玩家聊天文本中的毒性语言检测难题，尤其针对聊天内容极短、大量使用游戏专属俚语、缩写及领域术语等导致通用语言模型难以准确识别的问题。其核心解决方案是提出一种名为“双向表示注意力残差”（Bi-directional Representations with Attention Residuals, BRAR）的模型，通过引入注意力机制与残差连接，有效捕捉短文本中的全局上下文信息，从而在槽位填充任务中显著优于现有基线模型。
链接: https://arxiv.org/abs/2609.34584
作者: Yuanzhe Jia
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted by AAAI 2023
Abstract:In-game toxic language has emerged as a critical concern in the gaming industry and community. While several frameworks and models for online game toxicity analysis have been proposed, detecting toxicity in player chat utterances remains a formidable challenge: stemming not only from the extremely short length of such utterances but also from the heavy reliance on game slang, abbreviations, and domain-specific jargon, which generic language models are poorly suited to recognize. This paper presents a shared task for in-game toxic language detection built upon real-world in-game chat data, and proposes the best-preforming model for the toxic language slot filling: Bi-directional Representations with Attention Residuals (BRAR). Experimental results demonstrate that BRAR effectively captures the global context and outperforms the existing baselines on slot filling.
[NLP-94] Nudgeability: Reasoning Models Follow Confidence Signals Without Tracking Their Own Competence
【速读】： 该论文旨在解决生成式 AI（Generative AI）在推理过程中如何自主决定是否依赖外部工具（tool calling）的问题，即在推理阶段判断应“独立回答”还是“委托工具处理”。其核心挑战在于构建有效的自我反思机制（self-reflection mechanism），使模型能够基于自身状态做出合理决策。论文的关键创新在于提出并量化“助推性”（Nudgeability）这一新指标，通过在相同的推理轨迹中插入表达信心或怀疑的第一人称语句，考察该反射信号对后续行为的因果影响。研究发现，当前多数小到中等规模的推理模型对这类语言信号具有高度敏感性——怀疑会显著提升工具调用率，而信心则降低该概率，平均波动达20.6个百分点；大型模型甚至超过53至70个百分点。然而，这种响应机制在目标准确性方面表现不佳，仅有42%的决策调整是真正符合模型真实能力水平的，仅优于随机基线约2个百分点，表明现有模型虽能被语言信号有效操控，但未能实现与实际能力相匹配的精准调控。因此，该研究揭示了当前模型在自我反思机制上的“强控制、弱适配”特征，并提出“助推性”作为评估自反思能力成熟度的无训练后处理方法，为未来可解释、可调控的智能体设计提供了关键评测框架。
链接: https://arxiv.org/abs/2609.34572
作者: Rohit Saxena,Utkarsh Upadhyay
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 22 pages, 5 figures, 9 tables
Abstract:Reasoning language models that can call tools must decide during inference whether to answer unaided or delegate. Any self-reflection mechanism for this must answer three questions: where the reflective signal comes from (verbal reports, output distributions, hidden states, a separate predictor), how it is presented to the model (numerical prediction, confidence token, prompt injection), and whether it changes the model’s subsequent action. We isolate the third question. At a fixed point in otherwise identical reasoning trajectories, we insert a single first-person sentence expressing either confidence or doubt; the model then continues reasoning and chooses whether to answer directly or call a tool. Comparing these counterfactual continuations measures the causal effect of the reflective signal on delegation. We call this behavioral response Nudgeability and measure it along two dimensions: sensitivity, how strongly confidence and doubt change delegation rates, and targeting, whether delegation increases for problems the model cannot solve unaided and decreases for those it can. Across nine small-to-medium open-weight reasoning models from three families (Qwen, Gemma, and GLM) and two tasks, models are consistently sensitive: doubt increases delegation and confidence decreases it, with a median confidence-to-doubt swing of 20.6 percentage points, and 53 to 70 points for the larger provider-served models. This responsiveness is poorly targeted: a median 42% of induced flips are well-targeted, only a +2 percentage-point lift over a random-selection baseline. Confidence language is thus a strong control surface for delegation, but current models use it only weakly in accordance with their actual competence. Nudgeability offers a simple, post-training-free way to evaluate both sensitivity and targeting as endogenous self-reflection mechanisms mature. Comments: 22 pages, 5 figures, 9 tables Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) Cite as: arXiv:2609.34572 [cs.AI] (or arXiv:2609.34572v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.34572 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-95] Rethinking Latent Visual Reasoning : Grounding Latent Reasoning in Visual Evidence
【速读】： 该论文旨在解决生成式视觉推理（Latent Visual Reasoning, LVR）中因缺乏显式监督而导致的潜在语义表征学习偏差问题，具体表现为：在连续潜空间中的推理过程难以观测，导致模型对图像中影响正确答案的关键视觉证据响应不足，即存在“潜在证据-信用分配差距”（latent evidence-credit gap）。其核心解决方案在于提出一种名为ReaLVR的新框架，通过引入视觉证据监督机制，对模型自身自由运行的潜空间轨迹进行动态校准。ReaLVR利用正确答案与模型生成错误答案之间的对比，识别需要强化监督的位置，并结合相关与不匹配的视觉证据，明确应保留的视觉信息，从而实现对潜空间中各token的精确信用分配。实验表明，ReaLVR在三个模型家族中均显著优于现有基线，在Qwen2.5-VL-7B上达到63.7%的五任务平均准确率；更重要的是，该方法首次实现了在高达2350亿参数规模的前沿模型上的有效扩展，验证了其在大规模模型中的鲁棒性提升能力。
链接: https://arxiv.org/abs/2609.34563
作者: Xi Xiao,Tianchen Zhao,Youngeun Kim,Zhuowei Li,Linghan Xu,Jiaye Wu,Zheng Zhang,Xiang Xu,Xuanbai Chen,Farhan Tejani,Jakub Zablocki,Julia Xu,Yifan Xing
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 39 pages. Project page: this https URL
Abstract:Latent visual reasoning (LVR) enables multimodal large language models (MLLMs) to perform intermediate computation in continuous latent tokens rather than expressing every reasoning step in words. However, unlike textual CoT, latent reasoning is not directly observable, making it difficult to supervise what latent tokens learn. In this work, we first conduct a thorough analysis of latent-token behavior and identify a latent evidence-credit gap: latent tokens respond only weakly to image perturbations that alter the correct answer. We hypothesize that this issue stems from the lack of explicit supervision during GRPO training. These findings suggest that a final-answer reward provides too little guidance on what visual evidence to preserve or how credit should be assigned across latent tokens. To bridge this gap, we propose ReaLVR, which brings visual-evidence supervision to the model’s own free-running latent trajectories. ReaLVR contrasts correct and model-generated wrong answers to determine where stronger supervision is needed, and relevant and mismatched visual evidence to specify what to preserve. Across three model families, ReaLVR consistently outperforms evaluated LVR baselines, achieving the highest five-task average of 63.7% on Qwen2.5-VL-7B. Crucially, we are the first to scale visual reasoning in latent space, showing that our framework continues to deliver robust improvements at frontier model scales up to 235B. Further analyses show more question-sensitive latent-token positions, stronger alignment with relevant visual regions, and greater fixed-context dependence on the most attended latent tokens.
[NLP-96] RoPE is Dead Long Live RoPE: Towards Scalable Data-aware Positional Encodings
【速读】： 该论文旨在解决现代语言模型中旋转位置编码（Rotary Position Embedding, RoPE）在长距离依赖建模上的固有缺陷，尤其是其低频带波长过长导致模型在上下文外推（length extrapolation）时暴露于未见角度、产生显著的近邻偏差（recency bias）。现有位置编码方法因评估条件不一致而缺乏统一比较基准，文献体系呈现碎片化。本文的关键解决方案是提出数据感知旋转位置编码（Data aware RoPE, DaRoPE），其核心在于：保留标准RoPE在高频带的特性以维持局部位置敏感性，同时将低频带的绝对位置信息替换为从上下文表征中学习得到的有界坐标，使低频几何结构由数据分布决定而非仅依赖于固定的位置距离。这一设计不仅增强了模型对非文本任务（如符号音乐、基因组序列、神经信号）的适应能力，有效缓解了近邻偏差，且在语言建模和长度外推任务中表现不逊于或优于现有方法。此外，学习到的坐标具有可解释性，揭示了注意力机制如何利用超越单纯位置距离的上下文信息。实验覆盖从124M到50B参数的语言模型及多类任务，结果表明，在无特定领域偏好时，DaRoPE是所评估方法中的最优通用默认选择。
链接: https://arxiv.org/abs/2609.34556
作者: Jarod Lévy,Mathurin Videau,Jad Yehya,Jean-Rémi King,Stéphane d’Ascoli,Thomas Moreau
机构: Meta AI, Paris; Inria, Université Paris-Saclay, Palaiseau, France
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Transformers process tokens without any inherent notion of order, making positional encoding a fundamental requirement rather than an architectural refinement. Rotary Position Embedding (RoPE) has become the default positional encoding in modern language models, yet it is heavily biased toward nearby tokens. Existing alternatives have been evaluated under different settings, leaving the literature fragmented and without a clear replacement. We bring structure to this landscape by examining a specific weakness of RoPE: its slow frequency bands, whose wavelengths exceed the training context and expose models to unseen angles during extrapolation. We therefore introduce Data aware RoPE (DaRoPE), which preserves standard RoPE on the fast bands but replaces absolute position on the slow bands with bounded coordinates learned from contextual representations. Therefore, the slow-band geometry depends on the data rather than only on positional distance. We compare representative encodings under matched conditions across synthetic tasks, symbolic music, genomics, neural signals, and language models spanning 124M to 50B parameters. Across these experiments, DaRoPE leads on non-text benchmarks, mitigates recency bias, while remaining best or on par in language modeling and length extrapolation. Moreover, the learned coordinates also make the mechanism interpretable, revealing how attention layers leverage contextual information beyond token distance. Together, these results support DaRoPE as the best overall default among the evaluated methods, when there is no domain-specific reasons to prefer another.
[NLP-97] ActionLens: Diagnosing Spatial-Temporal Binding Failures in Vision-Language Models
【速读】： 该论文旨在解决视频多模态模型在时空绑定（spatial-temporal binding）方面存在的核心缺陷，即难以准确将特定动作与对应主体（人物）及时间点进行精确关联的问题。现有生成式视觉-语言模型（Vision-Language Models, VLMs）在主流基准测试中得分超过80%，但在涉及复杂时空关系推理的任务上表现不佳。为系统诊断此类问题，研究提出ActionLens，一个包含6,701道多选题的诊断性基准，涵盖五大关键维度：过渡检测、演员特异性识别、并发动作绑定、指向性交互推理和凝视检测。其真值标签基于158万次/秒、按人粒度的精确标注，经十四轮人工质量优化后，答案清晰度提升至人类准确率90%以上。实验表明，20个主流VLM模型在全集上的最优表现仅为68.8%，在人工审核子集上也仅达65.9%，远低于人类参考均值91.0%。尤其在凝视检测任务中，模型性能接近随机水平（89.6%人类准确率），凸显严重缺陷。进一步分析显示，相对静态坐标描述，关系型表述可提升5.55–13.25分，证实数值解析存在显著惩罚；但即便如此，视觉框仍普遍优于无框表示，暴露出模型在未封装主体分辨能力上的残余差距。绑定陷阱（binding-trap）分析揭示模型普遍存在“错误选择主体动作”的系统性偏差。综上，ActionLens通过结构化诊断框架，实现了对不同模型家族与规模下多种失效模式的量化评估，为模型改进提供可比较的基准依据。所有数据、代码与评估脚本均已开源。
链接: https://arxiv.org/abs/2609.34547
作者: Gueter Josmy Faure,Min-Hung Chen,Hao Ping Wang,Timothée Lardy,Hung-Ting Su,Winston H. Hsu
机构: National Taiwan University (国立台湾大学); NVIDIA(英伟达)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Project Page: this https URL
Abstract:Video-capable vision-language models score above 80% on popular benchmarks yet struggle with spatial-temporal binding: associating the right action with the right person at the right moment. We introduce ActionLens, a diagnostic benchmark of 6,701 multiple-choice video questions spanning five targeted diagnostics: transition detection, actor-specific identification, concurrent action binding, directed interaction reasoning, and gaze detection. Ground-truth answers are derived deterministically from 1.58 million per-second, per-person annotations. Fourteen rounds of human quality engineering raised answer clarity from 53% to above 90% human accuracy. Across 20 VLMs, the full-set leader scores 68.8%; on the human-reviewed subset, it scores 65.9% versus 91.0% for the pooled human reference. Gaze detection remains near chance against 89.6% human accuracy. On actor disambiguation, reference-interface controls show that relational descriptions recover 5.55–13.25 points over static coordinates, confirming a substantial numeric-parsing penalty; yet visual boxes still lead every model by 1.15–6.50 points, exposing a residual unboxed actor-resolution gap. A binding-trap analysis shows models systematically select the wrong actor’s action. ActionLens provides diagnostic measurements of these distinct failure modes across model families and scales for direct comparison. We release all data, code, and evaluation scripts at this https URL
[NLP-98] Papers Without Code: Availability of GitHub Repositories Linked in *CL Publications
【速读】： 该论文旨在解决自然语言处理（Natural Language Processing, NLP）领域中科研成果的可复现性问题，具体聚焦于发表于计算语言学（Computational Linguistics, *CL）及相关会议（如ACL及其附属会议）的论文所引用的GitHub代码与数据仓库在长期中的可用性。研究发现，尽管近年来学术界普遍倡导开源共享，但近十年来发表论文中链接的GitHub仓库仍面临较高的不可用率，其原因部分归因于大量空仓库和占位符仓库的出现，导致实际可访问的研究资源比例并未随时间改善。解决方案的关键在于提升代码与数据发布质量，推动研究者提交具有实质性内容、持续维护且具备长期可持续性的开源项目，同时建议引入更严格的元数据规范与平台依赖管理机制，以保障科研成果的可追溯性与可重用性。
链接: https://arxiv.org/abs/2609.34534
作者: Selina Meyer,Michael Roth
机构: University of Technology Nuremberg(纽伦堡应用技术大学)
类目: Computation and Language (cs.CL)
备注: Accepted for publication in Computational Linguistics. Author’s final version (pre-MIT Press publication)
Abstract:Source code and data published at computational linguistics (*CL) venues are increasingly being shared via GitHub. While this generally is a favourable development for the accessibility and potential reusability of research artifacts in natural language processing (NLP), the long-term availability of such repositories has not been evaluated. In this squib, we discuss the availability of repositories linked in papers published in the Computational Linguistics (CL) journal as well as at ACL and its co-located events over the past ten years. Contrary to our expectations, we find that GitHub repositories linked in more recent ACL publications are unavailable at similar rates as in older publications, in parts due to an increase in empty and placeholder repositories. Similar trends hold for other *CL venues, but not for platforms other than GitHub.
[NLP-99] How to Tame a Multi-Headed Hydra? Adaptive Multi-Category Safety Steering for Large Language Models
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在面对包含多个危害类别（harm categories）的复合型有害提示时，现有激活转向（activation steering）方法无法有效协调多类别安全干预方向与强度的问题。其核心挑战在于：单一提示可能同时触发多种类型的安全风险，而传统方法在处理多类别共现时缺乏对各风险类别的显式感知与协同调控，导致部分潜在危害未被充分抑制。为此，论文提出一种类别自适应的多类别安全转向框架——CAM-Steer，其关键创新在于通过对比当前隐藏状态与安全/不安全原型（prototype），量化每个危害类别的风险水平，并基于该风险估计值动态融合不同类别的安全转向方向，生成统一的转向方向；同时，利用风险得分决定转向的旋转强度，确保在保持隐藏状态范数不变的前提下实现精准、自适应的干预。实验结果表明，CAM-Steer在三种主流LLM架构和七种危害类别上的平均防御成功率显著优于基线方法，尤其在多类别共现场景下表现突出，且具备可忽略的推理开销。
链接: https://arxiv.org/abs/2609.34514
作者: Chenxi Wang,Ruiyang Huang,Li Huang,Yifan Wu
机构: Southeast University (东南大学); Peking University (北京大学); Chongqing University (重庆大学)
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:As large language models (LLMs) become increasingly widespread, preventing unsafe responses to harmful prompts is essential for their safe deployment. Activation steering offers an approach to improving LLM safety by modifying internal activations during inference without updating model parameters. However, a single prompt can involve multiple harm categories, and steering toward safety in one category may leave harmful content from another unaddressed. Despite advances in adaptive steering, existing methods do not explicitly coordinate steering direction and strength when multiple harm categories co-occur within a single prompt. To address this problem, we propose CAM-Steer, a Category-Adaptive Multi-category Safety Steering framework. Specifically, it estimates the risk associated with each harm category by comparing the current hidden state with safe and unsafe prototypes. The estimated risks are then used to combine the safety directions for different harm categories into a single steering direction and to determine the strength of the intervention. Finally, it rotates the hidden state along the composed steering direction, with the rotation angle determined by the estimated risks, while preserving the hidden-state norm. Experiments across three LLM backbones and seven harm categories show that CAM-Steer outperforms the evaluated baselines in average defense success rate, including when categories co-occur. Further analyses support its component designs and informative risk scores, with negligible inference overhead.
[NLP-100] Low-Confidence Remasking Traps Flexibility: Realizing Arbitrary-Order Potential for Diverse Rollouts in Diffusion LLM s
【速读】： 该论文旨在解决生成式语言模型在采用任意顺序生成（arbitrary-order generation）时出现的多样性损失问题。尽管任意顺序生成理论上支持更丰富的输出路径，但现有研究表明其实际生成多样性反而低于传统的从左到右生成方式。论文指出，这一现象并非源于任意顺序生成机制本身，而主要归因于一种广泛使用的解码规则——低置信度重掩码（low-confidence remasking, LCR）。LCR在每一步对所有掩码位置采样，但仅保留概率最高的样本，其余均被过滤，这种机制会随着掩码位置数量增加导致低概率候选词被指数级抑制，从而显著降低生成多样性。相比之下，最高概率位置选择（top-probability position selection, TPP）通过优先选择最具确定性的位置进行采样，避免了不必要的过滤，有效维持了生成多样性。实验表明，将LCR替换为TPP可使生成结果的Pass@k指标恢复至与从左到右生成相当的水平，验证了多样性损失主要源于LCR的筛选行为。为进一步挖掘任意顺序生成的优势，论文提出熵引导初始化（Entropy-Guided Initialization, EGI），即首步选择熵最高的位置进行采样，后续遵循TPP策略。该方法进一步提升了轨迹多样性与解空间覆盖范围，且在下游策略优化任务中表现出更优性能，充分展示了任意顺序生成在实现多样化推理路径方面的潜力。
链接: https://arxiv.org/abs/2609.34509
作者: Moongyu Jeon,Dongjae Jeon,Bumjun Kim,Mingyu Kim,Albert No
机构: Yonsei University (延世大学); KRAFTON AI; Kookmin University (国民大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Masked diffusion language models support arbitrary-order generation, suggesting a natural way to produce diverse outputs. However, recent work argues that this flexibility reduces diversity by delaying high-uncertainty tokens that can lead to different generation paths. We trace this diversity loss not to arbitrary-order generation itself, but largely to low-confidence remasking (LCR), a widely used decoding rule. At each step, LCR samples a token at every masked position but commits only the sampled token with the highest probability, filtering out the rest. We show that this mechanism can exponentially suppress lower-probability tokens as more positions compete, and observe the same suppression in LLaDA. In contrast, top-probability position selection (TPP), which has often been conflated with LCR under the shared label confidence-based decoding, avoids this diversity loss. TPP first selects the position whose most likely token has the highest probability, then samples directly from that position’s distribution. Replacing LCR with TPP restores diversity and yields Pass@ k comparable to left-to-right decoding, suggesting that the reported diversity loss stems largely from LCR’s filtering rather than from generating high-confidence positions first. To further exploit order flexibility, we introduce Entropy-Guided Initialization (EGI), which samples the first token at the highest-entropy position and then follows TPP. This simple modification further improves rollout diversity and solution coverage beyond left-to-right decoding, with gains extending to downstream policy optimization, highlighting the potential of arbitrary-order generation for diverse rollouts.
[NLP-101] CARDAMOM: A Micro-Dialectal Arabic Speech Dataset for ASR
【速读】： 该论文旨在解决阿拉伯语自动语音识别（ASR）系统在处理细微方言差异时表现不佳的问题，尤其是传统基于国家层面的粗粒度标签无法捕捉到同一国家内部存在的子区域方言变异。其解决方案的关键在于构建一个由母语者社区共同标注的微方言阿拉伯语语音数据集——Cardamom，该数据集包含约40小时来自埃及、约旦、黎巴嫩、毛里塔尼亚、巴勒斯坦和沙特阿拉伯等六个国家共21种微方言的转录YouTube语音，并对每段音频进行多标签微方言标注、语码转换信息标记及话语级感知性别标注。这一精细标注设计使系统能够分析超越国家边界的地方性语言变体，从而实现更精准的方言适应与评估。实验表明，在零样本设置下最强的多语言ASR系统平均词错误率（WER）为43.47%，而通过在Cardamom上进行适配后，WER降低至35.21%；此外，基于音频的微方言识别分类器达到85.57%的准确率，验证了标注信息的有效性与可学习性。因此，Cardamom为研究局部方言变异及开发具备更广泛区域覆盖能力的阿拉伯语语音系统提供了重要资源。
链接: https://arxiv.org/abs/2609.34481
作者: Bashar Talafha,Samar M. Magdy,Aisha Alansari,Alaa Alkhawaldeh,Abdurrahman Juma,Sharaf Makahleh,Nour Gamal,Omar Attia,Hanaa Kurdi,Najwa Rizk,Maysa Anaya,Hessah Altimyat,Layal Alhazmi,Shumukh Alotaibi,Hajar Alhadaris,Rayan Alomari,Rahaf Almalaq,Malak Alkhorasani,Sara alghamdi,Rahaf Alshamrani,Nsrin Ashraf,Ibrahim Jaradat,Nada Qardahji,Yasmin Zaraket,Elmoukhtar Brahim,Sidi Ebeidy,Oumoulmouminin Mahmoud,Yahjeb Bouha Khatraty,Meya Haroune,Mohammad Ghaddar,Mohamad Eldirany,Rashed Alamoush,Tala Chhaytle,Nuha Albadi,Yahya El Hadj,Hamzah Luqman,Fadi A. Zaraket,Mustafa Jarrar,Muhammad Abdul-Mageed
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:We present Cardamom, a micro-dialectal Arabic speech dataset designed to support fine-grained evaluation and adaptation of automatic speech recognition (ASR) systems. Community-curated by native speakers familiar with the represented varieties, Cardamom contains approximately 40 hours of transcribed YouTube speech spanning 21 micro-dialects across Egypt, Jordan, Lebanon, Mauritania, Palestine, and Saudi Arabia. Each segment is annotated with one or more operational micro-dialect labels, code-switching information, and utterance-level perceived gender, enabling analysis of sub-country variation that is obscured by conventional country-level labels. We describe the collection and annotation process, motivate the micro-dialect inventory linguistically, and benchmark four multilingual ASR systems in zero-shot and adapted settings. The strongest zero-shot system obtains 43.47% aggregate WER, with particularly high error rates on Mauritanian and Lebanese varieties; adaptation on Cardamom reduces its WER to 35.21%. Audio-based identification experiments further show that the annotations provide a learnable prediction target, with a dedicated classifier reaching 85.57% accuracy on 21-way micro-dialect identification. Cardamom provides a resource for studying localized dialectal variation and developing Arabic speech systems with broader regional coverage.
[NLP-102] he Last Mile Is the File: OfficeEditBench for Preservation-Aware Office Editing
【速读】： 该论文旨在解决办公文档（如电子表格、演示文稿和文档）在局部修改过程中如何精确维护变更范围的问题，即在确保必要更新被完整传播的同时，避免对受保护状态的无意更改。其核心挑战在于：更新不足会导致依赖关系不一致，而过度更新则可能违背用户授权意图。解决方案的关键在于提出OfficeEditBench——一个包含170个任务的基准测试平台，用于评估和验证在保持原有逻辑结构与作用域的前提下完成指定变更的能力。该基准通过明确定义任务契约（task contracts），涵盖必需更新、受保护状态、原生结构及交互约束，实现了对文件交付、目标完成度以及验证器定义接受标准的区分评估。研究发现，尽管多数输出在语法或格式层面可被接受（硬性包验证通过率92%–100%），但无一能完全满足合同要求，揭示了局部正确性不足以保障整体可维护性。案例分析表明，诸如公式丢失、规则未传递至相关结论、新增截止日期遗漏前置条件等现象均源于对上下文关联性的忽视。因此，该方案强调将构件级检查机制与办公文件持续可维护性相连接，为实现精准、安全的变更管理提供了可验证的测试框架。
链接: https://arxiv.org/abs/2609.34469
作者: Zhiwen Wu,Chengxu Wu
机构: Peking University (北京大学)
类目: oftware Engineering (cs.SE); Computation and Language (cs.CL)
备注: 23 pages, 7 figures. Benchmark and code: this https URL
Abstract:A small Office edit creates two obligations: propagate every required update and leave protected state untouched. Updating too little leaves dependencies inconsistent; updating too much changes content the user did not authorize. We introduce OfficeEditBench, a 170-task benchmark for change-scoped maintenance of spreadsheets, presentations, and documents. Task contracts specify required updates, protected state, native structures, and applicable interaction requirements. Across 510 archived task-system outcomes from WorkBuddy, Doubao, and Codex, we distinguish file delivery, target completion, and verifier-defined acceptance. Hard package-valid delivery ranges from 92% to 100%, yet no selected output satisfies the complete contract. Case analysis highlights why local correctness is insufficient: an updated value can lose its generating formula, a revised rule can fail to reach related conclusions, and a new deadline can omit a retained prerequisite. These mechanisms connect artifact-level checks to the continued maintainability of Office files. We analyze maintenance failures while distinguishing frozen automatic verdicts from human acceptability. OfficeEditBench provides a testbed for completing required changes while preserving the logic and scope of existing work.
[NLP-103] RGDT-Bench: Benchmarking LLM Reasoning for Rule-Governed Decisions and Their Justifications
【速读】： 该论文旨在解决生成式 AI 在规则驱动决策任务（Rule-Governed Decision Tasks, RGDTs）中推理能力不足的问题，特别是在政策、合同和合规等场景下，模型需准确理解规则适用性、从证据中评估条件、处理规则与例外的组合，并提供可验证的决策依据。现有基准多聚焦于演绎推理，而忽视了对推理过程合理性与理由完整性（warrant completeness）的评估。为此，研究提出 RGDT-Bench 基准，涵盖 202.1K 条条件级监督样本，支持四种任务赛道与八种任务-探针组合，通过无标签提取与确定性检查机制，实现对理由来源覆盖度与一致性自动标注。该基准将失败归因于规则使用、条件判断、证据评估与聚合四个过程层级，同时检验最终决策结果。实验发现，在六种评估的大语言模型（LLM）中，正确回答仍存在平均 40.2% 的理由不完整问题，且现有 17 种评估器仅能以最高 57.69% 的任务平均受试者工作特征曲线下面积（AUROC）识别此类缺陷。为应对这一挑战，研究训练了一个基于理由监督的简单奖励模型，其在正确答案上的任务平均 AUROC 达到 69.24%，显著优于匹配的结果监督基线（+10.37 pp）与最优现有评估器（+11.55 pp），并普遍优于所有结果监督基线的响应选择表现，验证了理由监督在提升 RGDT 推理质量中的关键作用。
链接: https://arxiv.org/abs/2609.34455
作者: Jianpeng Zhao,Haihua Xu,Haoyang Zhang,Shuang Qian,Yixiang Tang,Xintao Wang,Kun Sun,Pei Wu,Shuhan Zhong,Pengyang Wang
机构: University of Macau(澳门大学); ByteDance(字节跳动)
类目: Computation and Language (cs.CL)
备注: 33 pages, 12 figures, 20 tables
Abstract:We study reasoning in Rule-Governed Decision Tasks (RGDTs), where models apply external rules to case facts and justify decisions, as required in policy, contract, and compliance settings. Beyond the deductive capability emphasized by standard mathematical and logical reasoning tasks, RGDTs require interpreting rules and their applicability, assessing conditions from evidence, combining judgments under rules and exceptions, and providing checkable justifications. These demands motivate a benchmark assessing both decisions and their stated grounds. We introduce RGDT-Bench, providing 202.1K condition-level supervision slots across four task tracks and eight supported task-probe combinations that vary access to supporting information. Label-blind extraction and deterministic checks produce labels for warrant completeness: source-referenced coverage and consistency of stated decision grounds. The benchmark attributes failures to four process layers: rule use, condition, evidence, and aggregation, and checks the final outcome. Among evaluable correct responses, warrant incompleteness averages 40.2% across six evaluated LLMs and supported task-probe combinations. Such warrant incompleteness poses potential safety risks and remains difficult to detect: the best of seventeen existing evaluators reaches only 57.69% (random: 50%) task-averaged area under the receiver operating characteristic curve (AUROC). To address this difficulty, we train a simple reward model with warrant supervision. It achieves 69.24% task-averaged AUROC among correct answers, exceeding the matched outcome-supervised baseline by 10.37 pp (percentage points) and the best existing evaluator by 11.55 pp. Beyond completeness assessment, the model outperforms both outcome-supervised baselines across nearly all response-selection comparisons, supporting RGDT-Bench’s warrant supervision for RGDT reasoning.
[NLP-104] When Words Fall Short: Iterative Synergy Between Verbalized Reasoning and Hidden Features for LLM Confidence Estimation
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）中置信度估计（confidence estimation）的可信性问题，核心挑战在于如何有效利用模型内部信息以生成更准确的置信度判断。传统方法主要分为基于估计器（estimator-based）和基于语义化表述（verbalization-based）两类，当前主流观点认为后者在表达置信度方面更具优势。然而，本文通过实证研究发现，专门设计的置信度估计器能够显著优于语义化自报告的置信度，表明大语言模型内部表示中蕴含着更为丰富的置信信号。为此，作者提出迭代策略-估计器训练框架（Iterative Policy-Estimator Training, IPoET），其关键在于协同利用语义化推理轨迹与高信息量的隐藏表示：通过交替进行策略优化与估计器更新，将估计器输出的置信度反馈融入策略学习，并基于新策略采样结果动态刷新估计器。该机制使模型能持续挖掘深层特征并适应策略分布的变化，实验表明，IPoET在多种数据集及Qwen、Llama等骨干模型上均显著优于基准方法，在域内表现一致领先，且在跨域评估中也展现出优越或相当的性能。
链接: https://arxiv.org/abs/2609.34454
作者: Yekun Xu,Ante Wang,Jingyi Ren,Xuanyi Chen,Weizhi Ma,Yang Liu
机构: Tsinghua University (清华大学); Institute for AI Industry Research (AIR), Tsinghua University (清华大学人工智能产业研究院); Dept. of Comp. Sci. Tech., Institute for AI, Tsinghua University (清华大学计算机科学与技术系人工智能研究所)
类目: Computation and Language (cs.CL)
备注:
Abstract:Confidence estimation is crucial for developing trustworthy large language models (LLMs), with most methods following estimator-based or verbalization-based paradigms. While recent research increasingly focuses on improving verbalized self-reports of confidence, we challenge the prevailing view that this approach surpasses independent confidence estimators. Our empirical study shows that a dedicated confidence estimator can substantially outperform verbalized confidence, indicating that LLMs’ internal representations contain richer confidence signals. Building on this finding, we propose Iterative Policy-Estimator Training (IPoET), a framework that synergizes the complementary strengths of verbalized reasoning traces and informative representations. IPoET alternates policy optimization with estimator updating, integrating estimator-derived confidence feedback into policy learning and refreshing the estimator on new policy rollouts. Experiments across diverse datasets and Qwen and Llama backbones demonstrate that, by iteratively exploiting richer hidden features and adapting to the evolving policy distribution, IPoET consistently outperforms both estimator- and verbalization-based baselines in-domain and achieves superior or comparable results across all out-of-domain metrics. For more details, refer to this https URL.
[NLP-105] Unbiased Top-k Estimation for On-Policy Distillation
【速读】： 该论文旨在解决在基于策略的蒸馏（On-Policy Distillation, OPD）过程中，由于梯度估计不准确导致的学生模型推理能力迁移效果不佳的问题。具体而言，现有方法如Top-k OPD（TK-OPD）虽在计算成本与分布监督丰富性之间取得平衡，但仅使用选中的前k个候选词会导致概率质量丢失，引入偏差，进而降低模型精度。其解决方案的关键在于提出尾部修正的Top-k策略蒸馏（Tail-Corrected Top-k On-Policy Distillation, TT-OPD），通过同时利用学生生成轨迹中的采样词与选中的前k个高概率词，以期望形式恢复被忽略的概率质量，从而实现对逆KL散度梯度的无偏估计。该方法在保持TK-OPD低计算开销和丰富分布监督优势的同时，有效消除偏差，显著提升蒸馏性能。
链接: https://arxiv.org/abs/2609.34447
作者: Linjian Meng,Siyuan Gan,YuHan Li,Xiran Wang,Ziyang Ding,Ditang Gou,Yiming Wu,Zhen Zhao
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注:
Abstract:On-policy distillation (OPD) is becoming an important component of large language model (LLM) post-training for transferring the reasoning capability of a strong teacher LLM to a weaker student LLM. OPD trains the student by minimizing the reverse KL divergence between the teacher and the student via rollouts generated by the student’s policy. However, estimating the gradient of the reverse KL divergence in OPD remains a challenge. Using only the sampled token from the student-generated rollout is computationally cheap but provides limited distributional supervision, which will degrade accuracy. In addition, using the full vocabulary provides complete distributional supervision but is computationally expensive. Therefore, recent works propose Top- k OPD (TK-OPD) that use selected top- k tokens, which provides richer distributional supervision than sampled-token estimation at substantially lower computational cost than full-vocabulary estimation. Unfortunately, using only the selected top- k tokens induces bias, leading to accuracy degradation, as the probability mass outside the selected top- k tokens is discarded. To address the bias of TK-OPD, we propose Tail-Corrected Top- k On-Policy Distillation (TT-OPD). It preserves the advantages of TK-OPD, including rich distributional supervision and low computational cost, while providing an unbiased estimator of the gradient of the reverse KL divergence. The key insight of TT-OPD is to use not only the selected top- k tokens, but also the sampled token from the student-generated rollout, thereby recovering the discarded probability mass in expectation, avoiding the bias. Experimental results demonstrate that TT-OPD significantly outperforms other tested OPD variants.
[NLP-106] Remember by Asking: Retrieval-Induced Memory Evolution for LLM Agents
【速读】： 该论文旨在解决语言智能体在长期多轮交互中因记忆机制局限而导致的上下文连贯性与信息保留不足的问题。现有记忆系统普遍依赖读取时的检索，而写入时的记忆构建仍依赖于一次性直接提取或压缩，这种单次压缩策略在面对未来未知的信息需求时，容易忽略局部关键细节，导致重要信息丢失。为此，本文提出RIME（Retrieval-Induced Memory Enhancement）框架，其核心创新在于将记忆构建从单一的压缩模式转变为以证据为中心的整合机制。RIME通过生成通用自问问题主动检索聚焦的对话证据，并基于检索到的内容与相关历史记忆进行联合调和，构建带有时间戳和来源追溯信息的动态记忆库。在推理阶段，压缩后的记忆作为主要而非唯一的证据源；当其无法支撑回答时，RIME可即时检索原始对话及其局部上下文以恢复记忆形成过程中被忽略的信息，避免对完整历史进行处理。大量实验在LoCoMo基准上使用Qwen3-235B-A22B和GPT-5.6 Sol模型验证表明，RIME在所有三项质量指标上均显著优于对比方法，且显著降低了推理阶段的LLM调用令牌数，展现出更高的效率与鲁棒性。
链接: https://arxiv.org/abs/2609.34438
作者: Wanqi Zhou,Jiawei Lu,Yang Wang,Zhaolong Xing,Zhen Chen,Ai Han,Haoyue Shi
机构: Chang’an University (长安大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Long-term memory is essential for language agents to maintain coherent and effective behavior over extended, multi-session interactions. Existing memory systems mainly use retrieval at read time, while write-time memory formation still relies on direct extraction or compression. However, when future information needs are unknown, compressing an entire interaction in one pass can overlook locally important details that may matter later. To this end, we introduce RIME, a retrieval-induced memory framework that shifts memory construction from monolithic compression toward evidence-centered integration. RIME uses generic self-questions to retrieve focused dialogue evidence and grounds memory formation in both the retrieved evidence and relevant historical memories, which are jointly reconciled into an evolving memory bank with temporal and provenance information. At inference time, compressed memory serves as the primary rather than the sole source of evidence: when it cannot support an answer, RIME retrieves relevant source dialogue together with its local context to recover information omitted during memory formation, without resorting to full-history processing. Extensive experiments on LoCoMo with Qwen3-235B-A22B and GPT-5.6 Sol show that RIME consistently achieves the best performance across all three quality metrics among the compared methods, while requiring substantially fewer query-time LLM tokens.
[NLP-107] Agent Hop: A Diagnostic Benchmark for Agent ic Multi-Hop Scientific Question Answering NEURIPS2026
【速读】： 该论文旨在解决生成式智能体（Agentic AI）在复杂多步任务中因资源受限而产生的失败问题，其核心挑战在于现有评估基准仅关注单一的性能得分，导致模型失败的根本原因难以被识别与诊断。为克服这一局限，作者提出AgentHop——一个包含1,011道多项选择题的诊断性基准，配备受控的七工具沙箱环境，并在固定的令牌数、交互轮次和工具调用次数约束下进行测试。该方案的关键在于通过四个维度（信息检索、知识综合、工具调用、资源管理）对模型表现进行解耦分析，从而揭示不同模型家族的行为模式与脆弱性。实验结果表明，各模型家族展现出显著不同的工具调用特征：GPT系列倾向于早期提交，Anthropic与GLM模型在决策前更注重验证，DeepSeek与Kimi存在过度搜索现象，而Gemini-3 Pro则保持平衡。进一步的细粒度分析还揭示了同一模型家族内部的差异，例如Claude Opus 4.6与Sonnet 4.6在准确率相近的情况下，前者更依赖信息检索，后者则在知识综合能力上更优。该研究通过可复现的基准数据集与评测框架，为系统性诊断与优化生成式智能体提供了关键工具。
链接: https://arxiv.org/abs/2609.34428
作者: Chanhee Park,Jeongho Yoon,Sungbin Han,Hyeonseok Moon,Heuiseok Lim
机构: Korea University(韩国大学); Sookmyung Women’s University(淑明女子大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted to NeurIPS 2026 Evaluation and Datasets Track
Abstract:Agentic tasks require a large language model to interact with the world, navigating information and gathering evidence across multiple steps with restricted resources. Due to this complexity, agentic task failures arise from various sources, and pinpointing these failure causes is essential to diagnose and improve agentic systems. Existing benchmarks, however, tend to focus on a single leaderboard score, leaving the underlying failure modes opaque. To fill this gap, we introduce AgentHop, a diagnostic benchmark of 1,011 multiple-choice questions paired with a controlled seven-tool sandbox under fixed token, turn, and tool-call constraints. AgentHop reveals model vulnerabilities by dissecting a single accuracy score along four axes of agent operation: retrieval, synthesis, tool-call, and resource management. Across 19 models, we find that behavior clusters by model family, with tool-call signatures revealing distinct family fingerprints: GPT models commit early, Anthropic and GLM checkpoints verify before committing, DeepSeek and Kimi over-search, and Gemini-3 Pro stays balanced. Decomposed axes further expose within-family structure: Claude Opus 4.6 and Sonnet 4.6 land within one accuracy point yet diverge on retrieval-versus-synthesis emphasis, with Opus retrieving more and Sonnet synthesizing better. We release the full benchmark set and the harness to support diagnostic agent benchmarking.
[NLP-108] LLM s as Adaptive Meta-Solvers: Strategy-Diverse RL for Industrial-Scale Optimization
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在从教科书规模的优化问题向真实工业场景中大规模、结构多样化优化任务扩展时所面临的挑战。现有方法多局限于小型自包含文本型问题的评估，且普遍采用“求解器集成”范式，难以应对实际优化工作负载的复杂性与规模。其核心解决方案是提出一种名为策略多样强化学习（Strategy-Diverse Reinforcement Learning, SDRL）的框架，使开源LLM能够作为自适应的优化元求解器（meta-solver）。SDRL的关键在于利用不同求解策略——包括求解器集成推理、精确组合算法与启发式搜索——在不同问题结构和规模下的互补优势，并通过一种基于正确性门控的分层多样性奖励机制，促进跨策略及策略内部的鲁棒探索，有效避免早期策略坍缩。此外，该框架还引入混合格式训练方案，同时支持自包含文本问题与文件驱动的实例，显著提升了模型在真实工业级优化任务上的泛化能力与性能表现，优于现有微调方法及前沿模型如DeepSeek-V4-Pro和GPT-5.5。
链接: https://arxiv.org/abs/2609.34427
作者: Shihao Zhang,Weiting Liu,Siyu Shao,Yitian Chen,Jianfeng Feng,Dongdong Ge,Yinyu Ye
机构: Tokentide AI; Alibaba Group; East China Normal University (华东师范大学); Fudan University (复旦大学); The University of Hong Kong (香港大学); Shanghai Jiao Tong University (上海交通大学); Stanford University (斯坦福大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Scaling LLM-based optimization from textbook-scale instances to real-world, industrial tasks remains a critical open challenge. Existing approaches are predominantly evaluated on small, self-contained textual problems and often commit to a solver-integrated paradigm, limiting their ability to handle the scale and structural diversity of practical optimization workloads. In this work, we propose a practical framework for training open-source LLMs to tackle real-world, industrial-scale optimization. We first show empirically that solver-integrated reasoning, exact combinatorial algorithm, and heuristic search exhibit complementary strengths across different problem structures and scales. Motivated by this, we introduce Strategy-Diverse Reinforcement Learning (SDRL), which trains LLMs as adaptive optimization meta-solvers. SDRL leverages this complementarity through a correctness-gated hierarchical diversity reward that promotes robust exploration across varying strategies and within each strategy, effectively preventing premature strategy collapse. We further introduce a mixed-format training scheme that jointly supports both self-contained textual problems and file-grounded instances. Across comprehensive evaluations, our framework outperforms existing fine-tuned methods and frontier models including DeepSeek-V4-Pro and GPT-5.5, both on average across benchmarks and on industrial-scale optimization tasks.
[NLP-109] Zero-Shot Cue-Grounded Topic Segmentation of Spoken Documents
【速读】： 该论文旨在解决生成式语音文档中主题分割的粒度适应性问题，即现有基于大语言模型（LLM）的分割方法难以在不同粒度（从宏观主题转变到细粒度子话题）间灵活调整，导致出现子话题合并或连贯主题过度切分的问题。其解决方案的关键在于提出一种无需训练、不依赖特定任务标注的提示引导分割（Cue-Grounded Segmentation, CGS）框架：首先通过识别显式表征新话题开始的语义线索（cue）并以其所在句位置作为分割边界；当线索不足时，则退化为基于线索提取过程中推断出的文档结构进行语义分割。该方法在六个基准数据集和六种LLM骨干模型上均表现出色，显著优于现有基线，且对语音识别（ASR）噪声具有鲁棒性，同时在专有模型上保持较低的API调用成本。
链接: https://arxiv.org/abs/2609.34425
作者: Suhwan Choi,Myeongho Jeon,Myungjoo Kang
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Topic segmentation structures spoken documents into coherent sections, facilitating navigation and downstream understanding. The appropriate granularity can vary substantially, ranging from broad thematic shifts to fine-grained subtopics. Existing LLM-based segmenters, however, often struggle to adapt to this variation, causing them to either merge distinct subtopics or over-segment coherent themes. To address this, we introduce Cue-Grounded Segmentation (CGS), a training-free framework that operates without any task-specific supervision. CGS first identifies phrases that explicitly signal the start of a new topic and uses their sentence positions as segment boundaries. When such cues are insufficient, it falls back to semantic segmentation, guided by the document structure inferred during cue extraction. Across six benchmarks and six LLM backbones, CGS consistently outperforms existing baselines, remains robust to noisy ASR transcripts, and achieves these gains with low API cost on proprietary models.
[NLP-110] Coding Agent Memory Post-training: Unlocking the Memory Potential of Pre-trained File Operations for Long-Horizon Tasks via Reinforcement Learning
【速读】： 该论文旨在解决生成式 AI 在处理长周期任务时面临的记忆管理难题，即当任务交互历史超过模型的上下文窗口（context window）限制时，如何有效利用外部记忆机制以维持长期决策能力。现有方法虽尝试通过强化学习将记忆控制纳入策略（policy），但普遍依赖特定领域内预定义的记忆工具，并在短周期训练环境中进行学习，导致所学记忆行为与具体环境接口紧密耦合，脱离基础模型预训练阶段的通用知识体系，因而难以泛化至更复杂的长周期任务中。本文提出 Coding Agent Memory Gym (CAMG)，一个涵盖购物（Shop）、编码（Coding）、深度研究（DeepResearch）和自动研究（AutoResearch）四类长周期任务的强化学习环境套件。其核心创新在于为每个环境提供可执行的外壳访问权限及贯穿整个任务周期的持久化工作空间，使智能体能够以文件形式创建、修改、检索和复用信息作为外部记忆。在此基础上，本文引入 CAMG-RL 方法，采用完全异步的近端策略优化（PPO）算法，在四个不同环境中联合训练单一策略，直接从下游任务奖励信号中学习基于文件的内存操作行为。实验表明，基于 Qwen3.5 模型微调得到的 CAMG-RL-4B 与 CAMG-RL-9B 分别在 SWE-bench Verified 与 MLE-bench Lite 基准上达到与更大规模模型（Qwen3.5-35B-A3B 与 Qwen3.5-122B-A10B）相当的性能，验证了该方案在长周期任务中实现高效、可迁移记忆管理的有效性。
链接: https://arxiv.org/abs/2609.34422
作者: Lirui Luo,Kelong Mao,Heming Xia,Rongqing Li,Xinwei Yang,Luyu Chen,Kieran Wong,Yudong Guo,Xinrui Wang,Jiayin Zhu,Simiu Gu,Sulong Xu,Cong Fang
机构: JD.com(京东)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Language-model agents increasingly tackle long-horizon tasks whose interaction histories exceed the model’s active context. Recent work has begun to use reinforcement learning to make memory control part of the policy, often relying on predefined memory tools within domain-specific training environments of relatively short horizons. This setup ties learned memory behavior to environment-specific interfaces that lie outside the base model’s pre-training and must be learned from scratch, so even after post-training, agents struggle to use memory in long-horizon tasks. To address these limitations, we introduce Coding Agent Memory Gym (CAMG), a suite of long-horizon agentic-RL environments spanning Shop, Coding, DeepResearch, and AutoResearch. Alongside each environment’s native task interface, CAMG provides executable shell access and an episode-persistent workspace, enabling agents to create, revise, search, and reuse files as memory throughout an episode. We also introduce CAMG-RL, which trains a single policy jointly across all four environments with fully asynchronous PPO, learning this file-based memory behavior directly from downstream task reward, and we train CAMG-RL-4B and CAMG-RL-9B from Qwen3.5 models of matching size. On SWE-bench Verified and MLE-bench Lite, CAMG-RL-4B and CAMG-RL-9B are competitive with Qwen3.5-35B-A3B and Qwen3.5-122B-A10B, respectively.
[NLP-111] Reciprocal Guidance: Orchestrating Draft and Verify Budgets for Advancing the Diffusion-AR Self-Speculation Frontier
【速读】： 该论文旨在解决自回归验证（autoregressive verification）机制下生成式推理中高效推测解码（speculative decoding）的性能瓶颈问题，特别是在不同并发负载场景下难以平衡整体吞吐量与单请求延迟之间的矛盾。现有自推测模型如Nemotron-Labs-Diffusion虽通过共享骨干网络简化了流程并支持更长的接受长度，但在低并发时因每轮需执行两次前向传播导致有效每前向输出令牌数（tokens per forward, TPF）受限；而在高并发时，过长的推测块又引发计算开销激增，迫使各请求受限于保守的推测预算，无法充分释放全骨干推测器（full-backbone drafter）的潜力。其核心解决方案在于提出一种运行时动态调度框架——互惠引导（Reciprocal Guidance, RecGuide），其关键创新在于发现并利用“起草”与“验证”过程之间的双向可预测性：起草的输出概率分布可预判潜在的验证不匹配，而近期验证结果则能反向预测后续起草的有效性及最优块大小。基于此，RecGuide在低并发时通过验证重叠式起草充分利用空闲算力，在高并发时动态为每个请求分配适配的草案块大小，实现对计算资源的弹性调度。实验表明，RecGuide在多种并发水平下均显著优于基础自推测方法，最高实现1.8倍的吞吐提升。
链接: https://arxiv.org/abs/2609.34388
作者: Linye Wei,Shutian Zheng,Haoyu Zeng,Meng Li
机构: Peking University (北京大学); Shandong University (山东大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Diffusion drafting with autoregressive (AR) verification has emerged as a promising paradigm for efficient speculative decoding. Recent self-speculation models, represented by Nemotron-Labs-Diffusion, further simplify the speculative pipeline by unifying drafting and verification within a shared backbone, while enabling longer acceptance lengths. However, the Pareto frontier between aggregate and per-request throughput remains underexplored. At low concurrency, sequential draft-verify execution requires two model forward passes per round, limiting the effective tokens per forward (TPF). By contrast, at high concurrency, longer drafts incur increasingly expensive computation, forcing individual requests to operate under constrained speculation budgets and preventing full exploitation of the full-backbone drafter. Our key observation indicates that drafting and verification exhibit reciprocal predictability. Draft logits can anticipate likely verification mismatches, while recent verification outcomes predict future drafting utility and suitable block sizes. Building on this observation, we introduce Reciprocal Guidance (RecGuide), a runtime draft-verify orchestration framework that adapts speculative decoding to varying serving loads. RecGuide exploits spare compute capacity through verification-overlapped drafting at low concurrency, while dynamically allocating request-specific draft block sizes as the workload becomes increasingly compute-intensive. Experiments across a wide range of concurrency levels demonstrate consistent throughput improvements over vanilla self-speculation, achieving up to 1.8\times speedup.
[NLP-112] Look Before You Select: Rethinking Vocabulary Sparsification in On-Policy Distillation
【速读】： 该论文旨在解决在策略蒸馏（On-policy Distillation, OPD）中，使用全词汇量教师纠正（full-vocabulary correction）时因反向传播至所有词元（token）logits而导致的长序列内存开销过大的问题。现有方法通过采样词元或仅对学生的TopK词元施加监督来降低内存消耗，但会引入采样噪声或破坏全词汇量纠正的完整性。本文提出的SparseOPD解决方案的关键在于：先以全词汇量教师纠正构建修正信号，但不保留其反向传播图，随后根据修正量的大小而非学生生成概率选择关键词元进行梯度传播；通过符号残差补偿（signed residual compensation）保持整体正负修正质量守恒，并采用修正感知的预算分配机制实现稀疏支持的合理分布，最终仅对选定词元的logits执行反向传播。该方法在六种任务-规模设置（涵盖数学、化学问答及多模态推理）下均优于采样词元与TopK方法，在任务平均准确率上达到或超过全词汇量基线，且在40亿参数数学任务中梯度余弦相似度达99%，8K全参数模型的反向传播内存降低70.5%。
链接: https://arxiv.org/abs/2609.34386
作者: Yongliang Miao,Shuang Liu,Yanguang Liu,Yandong Bai,Mengnan Du
机构: The Chinese University of Hong Kong, Shenzhen(深圳大学); Carnegie Mellon University(卡内基梅隆大学); New Jersey Institute of Technology(新泽西理工学院); Kuaishou(快手)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:On-policy distillation (OPD) uses teacher correction on student-generated responses. Full-vocabulary correction can provide important corrections even for tokens that the student assigns low probability, but backpropagating through all token logits becomes memory-intensive for long sequences. Existing memory-saving approaches estimate corrections from sampled tokens or restrict supervision to the student’s TopK tokens, introducing sampling noise or changing the full-vocabulary correction. We introduce \textbfSparseOPD, which uses full-vocabulary teacher correction to determine which corrections matter before selecting the token logits to differentiate. SparseOPD first constructs the full-vocabulary correction without retaining its backward graph, then selects tokens by correction magnitude rather than student probability. Signed residual compensation preserves the total promoting and suppressing correction mass, while correction-aware budget allocation distributes the sparse support across positions. Finally, the update backpropagates only through the selected token logits. Across six task–scale settings spanning mathematics, chemistry QA, and multimodal reasoning, SparseOPD outperforms Sampled Token and TopK in task-average accuracy and matches or exceeds Full Vocabulary. Gradient cosine similarity reaches 99% on 4B mathematics, while 8K full-parameter profiling shows 70.5% lower backward memory.
[NLP-113] FORGE: Form-Optimal Routing of Grounded Evidence for Frozen LLM Agents
【速读】： 该论文旨在解决在代理式人工智能（agentic AI）系统中，基于冻结基础模型（frozen foundation models）的下游适应问题。由于模型权重不可修改，仅能通过输入内容和推理过程进行调整，导致每个查询需同时决策“提供何种证据”与“分配多少推理预算”，而固定的默认策略在约80%的查询上存在资源配置不当的问题。其核心解决方案是提出FORGE框架，通过联合考虑证据形式（support form）与思维深度（thinking depth）的动态路由机制，在每轮查询中实现自适应优化。该方法基于熵正则化、成本感知的效用目标，推导出闭合形式的Boltzmann路由目标，并设计了一个仅269K参数的轻量级因子化路由器。该路由策略在不访问模型权重的前提下，通过三阶段训练流程——离线动作枚举、从Boltzmann目标进行监督KL蒸馏、以及利用主机反馈的组相对策略优化（GRPO）——完成训练。实验表明，FORGE在5个知识密集型基准测试及8种不同规模（7B至671B参数）的冻结模型上，均实现了42%-45%更低的令牌开销下准确率提升，并具备跨模型零样本迁移能力，且可与内生推理预算有效协同。
链接: https://arxiv.org/abs/2609.34358
作者: Xi Xiao,Yunbei Zhang,Chen Liu,Lin Zhao,Jialin Chen,Tianchen Zhao,Xiang Xu,Youngeun Kim,Tianyang Wang,Min Xu
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 34 pages. Project page: this https URL
Abstract:In agentic AI systems, frozen foundation models are increasingly deployed as closed-weight API endpoints, making downstream adaptation possible only through the inputs and inference procedures surrounding the model. As a result, for each input query, two coupled decisions largely determine both answer quality and token cost: what evidence to provide and how much reasoning budget to allocate. Fixed defaults along these axes are often suboptimal, misallocating support form or reasoning depth on roughly 80% of queries in our analysis. To address this challenge, we propose FORGE, a unified framework for adapting frozen models through per-query routing over a joint action space that spans both support form and thinking depth. Under an entropy-regularized, cost-aware utility objective, we derive a closed-form Boltzmann routing target and instantiate the policy as a lightweight 269K-parameter factorized router. The routing policy is trained around the frozen host, without any weight access, through a three-stage pipeline: offline arm enumeration, supervised Kullback-Leibler (KL) distillation from the Boltzmann target, and Group Relative Policy Optimization (GRPO) refinement with host feedback. Across 5 knowledge-intensive benchmarks and 8 frozen backbones ranging from 7B to 671B parameters, FORGE improves accuracy at 42-45% lower token cost on both main hosts, transfers zero-shot across hosts at lower token cost, and composes with intrinsic thinking budgets where available.
[NLP-114] Commutator Memory: Sparse Path-Local Reading and Steering in Language Models NEURIPS2026
【速读】： 该论文旨在解决语言模型在不同数据顺序下训练所导致的路径依赖（path dependence）问题，即同一组数据以不同顺序训练会得到参数不同的模型，且这种差异无法通过损失或基准测试指标完全解释。其核心问题是：是否存在一种可被量化和干预的“参数化训练历史记忆”——即权重中蕴含了关于训练顺序的信息，并能通过特定方向的干预实现因果操控。解决方案的关键在于引入李括号（Lie bracket） $ b_{AB} = H_B g_A - H_A g_B $ 来表征两个数据源 $ A $ 与 $ B $ 的梯度场在基础模型处的非交换性，发现小步长 $ \eta $ 下的权重差 $ \theta_{AB} - \theta_{BA} $ 在主导项上正比于 $ \eta^2 b_{AB} $。作者定义了“对换器记忆”（commutator memory），通过将李括号投影至输出空间（logits），生成每个词元的得分，这些得分具有高度局部性（在三个模型中，82–99%的显著词元与原始差异一致，远高于随机方向的35–49%），并具备因果可操作性：在Qwen-3-4B SFT中，抑制预测贡献最大的10个词元可消除中位数32%的损失差距，而频率匹配但得分接近零的词元则无显著影响。此外，直接投影两模型权重差到 $ b_{AB} $ 上，可在四个大型语言模型中以92%准确率区分训练顺序（随机水平为50%）。该记忆是成对数据源定义的，且随进一步训练而衰减，验证实验涵盖匹配批次的DPO、冻结回放式GRPO目标及AdamW优化终点，证实其稳健性。
链接: https://arxiv.org/abs/2609.34348
作者: John Sweeney
机构: Sideplane AI
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted at NeurIPS 2026. 44 pages, 10 figures, 25 tables
Abstract:Gradient updates on different data generally do not commute: training a language model on two data sources in opposite orders gives different weights, even with the same data and total exposure. Loss or benchmark deltas show that the models differ, not where. We ask whether this path dependence leaves a parametric training-history memory: a weight component that flips sign when the two sources are swapped, is localized in output space, changes the held-out loss gap between the two orders under targeted interventions, and reveals which trained model came from which order. For one small SGD step of size \eta on each of sources A and B , the weight difference \theta_AB-\theta_BA is, to leading order, \eta^2 b_AB , where b_AB=H_Bg_A-H_Ag_B is the Lie bracket of the two gradient fields at the base model. We define commutator memory by projecting the bracket through the logits into one score per vocabulary token; the scores sum to the bracket’s prediction of the gap. The scores are localized: on three models, the same readout of the measured \theta_AB-\theta_BA , or of a bracket from disjoint batches, shares 82-99% of the original top-20 tokens, versus 35-49% for norm-matched random directions. They are causally actionable: in Qwen-3-4B SFT, downweighting the ten tokens with the largest predicted share of the gap closes a median 32% of the measured gap, while frequency-matched tokens with near-zero scores have almost no effect. The weights themselves carry the component: projecting the difference between the two trained models onto b_AB identifies which came from which order in 92% of cases across four LLMs (chance 50%). Controlled tests also cover matched-batch DPO, a frozen-rollout GRPO-style objective, and an AdamW endpoint check. The memory is defined per source pair, not per example, and its projection on b_AB decays with further training.
[NLP-115] CRISP: Cultural Reward Modeling for Implicit Situated Propriety
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在跨文化应用场景中缺乏对开放性社会情境下文化适切行为识别与生成能力的问题。现有研究多集中于预定义响应空间的文化知识任务，而对开放式对话中隐含文化规范的动态适应仍存在显著空白。为此，本文提出一种文化情境化奖励模型——CRISP-RM，通过评估开放社交场景中行为的文化适切性来提供精细化奖励信号；其关键创新在于引入规范基础监督（Norm Grounding Supervision, NGS），在策略优化过程中增强模型对相关文化规范的敏感性与理解深度。为构建高质量文化情境数据，研究设计了协作式多智能体框架，将隐含文化规范显式化为多样化社会场景，并构建了名为NormCompass的专用评测基准。实验结果表明，CRISP-RM在奖励建模和策略优化两方面均显著优于主流通用奖励模型；结合NGS后，模型在文化情境行为表现上进一步提升，且能有效区分深层文化适切性与表面语言流畅性或礼貌性，验证了其在文化规范语义理解上的优越性。
链接: https://arxiv.org/abs/2609.34345
作者: Zekun Yuan,Yangfan Ye,Baohang Li,Shuaibo Zhao,Zekun Zhou,Ziming Li,Qichen Hong,Kun Chen,Xiaocheng Feng
机构: Harbin Institute of Technology(哈尔滨工业大学); Huawei Technologies Co., Ltd(华为技术有限公司); Peng Cheng Laboratory(鹏城实验室)
类目: Computation and Language (cs.CL)
备注: 27 pages, 6 figrues
Abstract:As large language models (LLMs) are increasingly deployed across countries and regions, the ability to recognize and respond appropriately to diverse cultural contexts becomes increasingly important. However, existing research has largely focused on cultural knowledge or tasks with predefined response spaces, while open-ended culturally situated behavior remains comparatively underexplored. In this work, we introduce CRISP-RM, a culturally situated reward model that assigns rewards according to cultural appropriateness in open-ended social scenarios. During policy optimization, we further introduce Norm Grounding Supervision (NGS), providing guidance that enhances the policy’s sensitivity to relevant cultural norms. To construct culturally situated data, we employ a collaborative multi-agent framework that instantiates implicit cultural norms into diverse social scenarios and further curate NormCompass as a dedicated testbed. We conduct comprehensive experiments to evaluate the effectiveness of CRISP-RM in both reward modeling and policy optimization. Best-of-(N) experiments show that CRISP-RM consistently outperforms strong general reward models. During GRPO policy optimization, CRISP-RM generally improves culturally situated behavior, while incorporating NGS yields further gains. Further analyses demonstrate the advantages of CRISP-RM in distinguishing culturally appropriate behavior beyond superficial fluency and politeness, while NGS provides complementary gains during policy optimization by improving norm grounding.
[NLP-116] Knowing When Thinking Is Not Enough: Teaching Small Reasoning Models to Reason Beyond Their Parametric Knowledge
【速读】： 该论文旨在解决小规模推理模型（small reasoning models, sRMs）在提升推理能力时，盲目增加推理计算量（test-time computation）可能带来的效率瓶颈与性能局限问题。研究表明，传统自修正（self-refinement）机制往往仅在已有解空间内重新分配概率质量，而非拓展新的可行解路径，其有效性受限于“执行瓶颈”与“知识瓶颈”两类失败模式：前者指正确路径虽可达但需通过反思恢复，后者则因缺乏外部知识而无法触及正确解。为应对这一挑战，论文提出一种选择性查询框架FlyBy，其核心创新在于：首先由小模型自主推理并诊断未解决的问题；当识别出知识瓶颈时，主动调用参数化知识更丰富的强模型进行针对性查询。该框架通过监督微调构建多层级查询策略，并结合成本感知的强化学习，动态优化是否查询、查询内容及资源消耗。实验表明，在6个基准上的1,158道高难度问题上，FlyBy-4B以2.7倍更低的服务成本超越Qwen3-14B（pass@8: 45.96% vs. 41.64%），且优于Qwen3-8B（pass@1: 16.85% vs. 15.31%），进一步扩展至FlyBy-8B后pass@8提升至51.81%，验证了该方法在高效利用外部知识方面的显著优势。
链接: https://arxiv.org/abs/2609.34327
作者: Chanuk Lee,Minki Kang,Sangwoo Park,Woongyeong Yeo,Jinheon Baek,Sung Ju Hwang
机构: KAIST(韩国科学技术院); DeepAuto.ai(†\dagger: Equal advising)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: preprint
Abstract:Scaling test-time computation is a powerful way to improve language-model reasoning, and is particularly appealing for small reasoning models (sRMs) that are cheap to serve. However, is additional thinking always the right operation? By intervening at intermediate reasoning states across two model families and multiple scales, we find that self-refinement largely consolidates probability mass onto solutions already reachable from the current state, rather than making new ones reachable. These interventions reveal two failure regimes: execution bottlenecks, where the correct path is reachable and reflection can recover it, and knowledge bottlenecks, where relevant external information makes it reachable. Motivated by this distinction, we introduce FlyBy, a selective querying framework, and train 4B and 8B variants to reason first, diagnose what remains unresolved, and, at a knowledge bottleneck, query stronger models whose parametric knowledge extends beyond its own. Supervised fine-tuning bootstraps a multi-depth query action, and cost-aware reinforcement learning calibrates whether to query, what to ask, and how much to spend. On 1,158 hard problems across six benchmarks, FlyBy-4B achieves 45.96% pass@8, surpassing Qwen3-14B (41.64%) at 2.7 times lower serving cost, while also exceeding Qwen3-8B in pass@1 (16.85% vs. 15.31%). Scaling to FlyBy-8B further improves pass@8 to 51.81%.
[NLP-117] Certified Selective Automation of LLM Agent Evaluation
【速读】： 该论文旨在解决大语言模型（LLM）智能体评估中自动化评判的可靠性问题：当前评估仍依赖人工阅读轨迹，因自动评判器无法保证错误率在可接受范围内。其核心挑战在于，传统基于独立同分布（i.i.d.）假设的筛选式评判方法在面对任务相关性导致的轨迹聚类时会高估自动化安全边界，导致实际错误率远超预设预算（如在17.5%的任务重采样中超出α=0.1的误差预算）。为此，论文提出一种任务级自举（task-level bootstrap）证书，该方法在各类测试场景下均保持有效性，且与朴素证书具有相同的覆盖率；相比之下，有限样本下的聚类校验方法在现实任务数量下无法提供任何有效认证。实验表明，一个经过SFT和拒绝加权GRPO训练的40亿参数对数概率（logprob）评判器，在α=0.1时可实现工具使用与网络数据集上0.30–0.59的认证评估覆盖率，是目前唯一能在两个主流基准数据集上同时实现认证的评判器。此外，认证覆盖度可在训练前仅凭基线率与判别能力即可高度预测（留一语料库外R²=0.96）。更进一步，该证书可作为自训练过滤器：在认证区域内生成的伪标签污染率严格受控于α（六次实验中实际污染率为0.000–0.041），使评判器无需目标领域标注即可在新领域达到域内性能水平。
链接: https://arxiv.org/abs/2609.34320
作者: Chengguang Gan,Yunhao Liang,Qinghao Zhang,Shiwen Ni
机构: Independent Researcher(独立研究员); University of Chinese Academy of Sciences(中国科学院大学); Pusan National University(釜山国立大学); Shenzhen University of Advanced Technology(深圳先进技术研究院)
类目: Computation and Language (cs.CL)
备注:
Abstract:Evaluating LLM agents still ends with a human reading trajectories, because automatic judges carry no guarantee on how often they are wrong. We ask the operational question: what fraction of agent evaluation can a judge take over, with a certificate that the error rate among auto-decided trajectories stays below a budget alpha? Agent corpora resist the standard answer: many agents attempt the same tasks, so trajectories arrive in correlated clusters, and the i.i.d. certificates of existing selective-judging methods can overstate what is safe: a naive certificate can claim 98% automation while its realized error exceeds the budget in 17.5% of task resamples. We introduce a task-level bootstrap certificate that is valid in every regime we test while matching the naive certificate’s coverage; finite-sample cluster-valid alternatives certify nothing at realistic task counts. Under this certificate, a 4B logprob judge trained with SFT and reject-weighted GRPO certifies 0.30-0.59 of evaluation on tool-use and web corpora at alpha=0.1, the only judge, among strongly elicited frontier models, certifying on both headline corpora. Certified coverage is predictable before training from base rate and discrimination alone (leave-one-corpus-out R^2=0.96). Finally, the certificate doubles as a self-training filter: pseudo-labels harvested inside certified regions have contamination bounded by alpha by construction (realized 0.000-0.041 across six harvests), letting a judge enter an unseen domain at in-domain strength with zero target training labels.
[NLP-118] PlaylistEval: Can Video-Language Judges Be Trusted at Day Scale and Beyond?
【速读】： 该论文旨在解决生成式视频-语言模型在处理超长视频（如长达百小时的视频合集）时，其判断可靠性是否依然可信的问题。现有评估基准因视频时长过短、答案对可仅通过字幕分离、且人工标注难以扩展至超长视频，无法有效验证模型在复杂长时序上下文中的理解能力。为此，论文提出PlaylistEval——一种无需人工标注的智能体框架，通过因果退化（causal degradation）控制问题对中答案的差异性，自动生成需跨整个视频合集进行检索才能回答的问答对，从而构建出覆盖7个领域、包含630对问答的基准数据集。在152对分层抽样数据上的实验表明，该基准与人类判断的一致性高达93.0%（组内相关系数 IAA = 0.781）。对17个来自八个模型家族的全模态及多模态模型的评估发现，前沿判别模型的成对准确率仅为75.4%，而开源判别模型表现显著更差；同时研究揭示判别性能依赖于多模态信息融合，且随着视频集合规模增大，判别准确率呈下降趋势。解决方案的关键在于：利用因果退化机制实现可控的语义差异生成，并通过自动化流程构建大规模、高难度、无需人工标注的长时序视频理解评估基准。
链接: https://arxiv.org/abs/2609.34314
作者: Shayekh Bin Islam,Hwanjun Song
机构: KAIST(韩国科学技术院)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 53 pages, 15 figures, 23 tables. Project page: this https URL
Abstract:Video-language models are increasingly used as judges of video understanding, both for evaluating model outputs and for training reward models. Whether their judgments remain reliable when the evidence is buried in day-long videos has yet to be established. Existing benchmarks cannot answer this. Their videos are typically only a few minutes long, many answer pairs can be separated from the transcript alone, and collecting human judgments does not scale to ultra-long videos. We introduce PlaylistEval, an agentic framework that builds video-language judge benchmarks over 100-hour playlist collection without human annotation. It automatically generates questions with paired answers whose differences are controlled by causal degradation, so that every pair demands retrieval across the collection. The resulting benchmark contains 630 pairs across seven domains spanning both static and dynamic knowledge, and on a stratified subset of 152 pairs it agrees with human judgments 93.0% of the time (IAA 0.781). Evaluating 17 omnimodal and multimodal models from eight families reveals that frontier judges reach only 75.4% pairwise accuracy, while open-source judge models perform far behind. We further show that both retrieval and final judgment depend on using multiple modalities, and that judge accuracy degrades as the playlist set grows. We release our pipeline, benchmark, and evaluation code at this https URL.
[NLP-119] ControlScope: Workflow Revision and Reliability in LLM Agents
【速读】： 该论文旨在解决语言模型代理（language model agent）在执行运行工作流（running workflow）时，应修订多大范围的问题，即在保持连续性、仅修改工具调用参数或完全替换未完成工作流之间如何权衡。其核心解决方案在于提出一种名为ControlScope的评估框架，通过对比三种不同修订策略——继续生成代码、仅编辑下一工具调用的数据参数（ARG）、以及从相同公开执行状态重新生成完整工作流（FULL）——来系统分析不同修复粒度的效果。关键创新在于采用嵌套权限机制（nested permissions），将可用的修复操作与代理实际选择的动作分离开来，从而精确衡量修复能力与决策行为之间的关联性。实验结果表明，在文件系统任务、ALFWorld和AppWorld等多个场景中，全面修复（FULL）在多数情况下表现更优，但需付出更高的计算成本；而仅修改参数（ARG）虽效率更高，却可能遗漏更优的廉价修复路径。研究还揭示了后续审查对修复有效性的重要影响，强调修复访问权限必须与实际决策过程紧密耦合，才能实现高效且可靠的自动化工作流修正。
链接: https://arxiv.org/abs/2609.34313
作者: Jingjie Ning,Xueqi Li,Yibo Kong,Dongting Li
机构: Carnegie Mellon University (卡内基梅隆大学); Tsinghua University (清华大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Software Engineering (cs.SE)
备注:
Abstract:How much of a running workflow should a language model agent revise? ControlScope compares continuing generated code, editing the next tool call’s data arguments, and replacing the unfinished workflow from the same public execution state. The nested permissions separate available repairs from the actions an agent selects. We evaluate one-time and repeated reviews across filesystem tasks, ALFWorld, and AppWorld. Across two source programs per task and three reasoning-reviewer draws on 20 filesystem tasks, FULL completes 15-16 tasks versus 13 for KEEP; across four fast draws it completes 10-13 versus 13. Fresh student-record confirmation reproduces a batch-read repair. ALFWorld fast panels yield KEEP/ARG/FULL scores of 85/86/87 on 87 tasks across 52 scenes and 134/134/127 on 134 tasks across four scenes; reasoning on the 87-task cohort also yields 85/86/87 with substantial review cost. An AppWorld V1 official-test panel of 585 task instances from 195 scenario templates shows small net differences. Frozen replays expose viable agent-written replacements interrupted by later revision in two failed file-organization runs. An offline source-trajectory midpoint comparison shows later reviews completing an insufficient repair. Five-call protection saves 19.4% of logged model output and loses one success across 20 fresh source runs. An argument-only shortcut shows that the broader sampled policy can overlook a cheaper successful edit available in both operation sets. These outcomes tie repair access to actual choices and subsequent execution.
[NLP-120] Dr.Credit: Rubric-Grounded Process Credit Assignment for Deep Research Agents
【速读】： 该论文旨在解决生成式任务中基于评分标准（rubric）的强化学习（RL）框架下，中间决策过程缺乏有效信用分配的问题。传统方法通常仅以最终答案的评分作为奖励信号，无法区分各中间步骤对结果的贡献，尤其在开放性任务中因缺乏标准答案而难以应用。其解决方案的关键在于提出一种“基于评分标准的信用分配”（rubric-grounded credit）机制：通过将任务要求作为共享参考基准，在评估工具返回信息时，判断其相对于评分标准历史接受支持情况所提供的额外支持程度，从而实现对新支持与已有证据的区分，并识别对每项评分标准的部分支持。该机制在强化学习框架中用于监督深度研究代理的中间工具调用过程，结合了过程优势与最终报告质量的优化目标，既保障了最终输出质量，又提升了研究过程效率。实验表明，该方法在多个域内和域外基准上均显著优于现有开源基线，在8B参数规模下性能可媲美前沿专有模型，且在有限交互轮次下展现出更高效的证据获取能力和更高品质的报告生成能力，验证了该信用分配策略在广泛评分标准任务中的可扩展性。
链接: https://arxiv.org/abs/2609.34296
作者: Yingjian Zhu,Zhenyi Wang,Jiaxin Guo,Kun Ding,Ying Wang,Shen Huang,Xunjie Zhu,Pengjun Xie,Shiming Xiang
机构: Alibaba Token Hub, Alibaba Group; School of Artificial Intelligence, University of Chinese Academy of Sciences; State Key Laboratory of Multimodal Artificial Intelligence Systems (MAIS), Institute of Automation, Chinese Academy of Sciences; Peking University
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Rubric-based tasks are increasingly addressed through reinforcement learning (RL), with rubric scores used as training rewards. However, these rewards typically supervise final answers without distinguishing the contributions of intermediate decisions. Many existing credit assignment methods rely on ground-truth answers to define process rewards, limiting their applicability to open-ended tasks without canonical solutions. To address this limitation, the proposed rubric-grounded credit uses task requirements as a shared reference for final answer evaluation and process supervision. The information returned by tools is assessed for the additional support it provides toward satisfying each rubric relative to that rubric’s history of accepted support. By referencing these histories, credit distinguishes new support from evidence already present in the trajectory while recognizing partial support for each rubric. this http URL uses rubric-grounded credit to supervise intermediate tool turns in an RL framework for deep research agents. The resulting process advantages are combined with GRPO outcome advantages to guide research decisions while retaining supervision of final-report quality. Evaluations on four in-domain and out-of-domain benchmarks show that this http URL outperforms the evaluated open deep research baselines on every primary metric and submetric. Meanwhile, with an 8B-parameter backbone, the trained agent achieves average performance competitive with the evaluated frontier proprietary models. Further analyses suggest more efficient evidence acquisition and higher-quality reports under limited research-turn budgets, motivating the extension of rubric-grounded process supervision to a broader range of rubric-based tasks.
[NLP-121] ReScraper: Unified Scraping and Cleaning of Web Data for Effective LLM Pretraining
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）预训练语料库清洗过程中依赖大量手工编写启发式规则所导致的效率低下与质量瓶颈问题。传统方法通过一系列基于规则的过滤器对从HTML中提取的主内容进行清洗，但其效果受限于规则的粗粒度和准确性。为此，本文提出ReScraper——一个仅0.6B参数的统一语言模型，用以替代原有的多层启发式规则堆栈。其关键在于：通过精心构建的监督数据集，该模型学习从原始网页数据中首先提取主要内容，并在四种操作之间做出决策——保留原内容、剔除噪声行或片段、完全删除页面，或在内容质量差但信息丰富时进行重写。基于相同爬取数据池，使用该标注数据预训练的400M、1.4B和2.8B规模模型，在DCLM Core评分上相对于最强基线实现了3.8%至4.7%的相对提升，且优于昂贵的多智能体人工标注流程。分析表明，每种操作均发挥独特且互补的作用，而将提取与清洗整合于单一模型中优于分阶段独立模型的级联结构。此外，ReScraper能精准聚焦于需要处理的低质量页面，显著提升劣质页面的质量，同时保持语料多样性。这些结果验证了AI for AI（AI4AI）在预训练数据清洗中的可行性与有效性，即由一个小型可学习模型接管原本由人工规则主导的整个数据清洗流程。
链接: https://arxiv.org/abs/2609.34287
作者: Zichun Yu,Jiarui Yan,Shlok Sanghvi,Nihar Atri,Chenyan Xiong
机构: Carnegie Mellon University (卡内基梅隆大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:LLM pretraining corpora are normally cleaned by a stack of hand-written heuristics. A heuristic scraper extracts the main content from HTML, and dozens of rule-based filters then clean it, so corpus quality is capped by the coarseness and accuracy of the rules. In this work, we propose ReScraper, a unified language model of only 0.6B parameters that replaces this entire stack. To train ReScraper, we carefully curate supervised data from the outputs of three teacher models, so it learns to first extract the main content from raw data and then choose among four operations: keeping the page as extracted, editing out noisy lines and spans, deleting it entirely, or rewriting it when it is poorly written but informative. Based on the same crawled data pool, pretraining 400M, 1.4B, and 2.8B models on our curated data improves the DCLM Core score by a relative 3.8–4.7% over the strongest baseline at each scale, including the costly multi-agent curation. Our analyses show that each operation plays a distinct and complementary role, and that extracting and cleaning in one model outperforms a cascade of separate models. ReScraper also concentrates its operations on the pages that need them, raising the quality of poor pages the most while keeping the corpus diverse. These results demonstrate the feasibility and effectiveness of AI4AI for pretraining data curation, where a small learned model takes over an entire stage of the pipeline from hand-written heuristics. We open-source our code at this https URL
[NLP-122] Over-Personalization Is a Decision Failure: Generation-Induced Apply Bias in LLM s
【速读】： 该论文旨在解决个性化大语言模型（Personalized LLMs）在推理过程中因过度个性化而导致的偏好应用失准问题，即模型在不适宜的上下文中仍强制应用用户偏好，而现有评估基准仅关注最终生成结果，无法定位错误发生的具体环节。其核心解决方案在于将偏好处理过程分解为三个可独立测量的阶段：（1）判断偏好是否适用；（2）显式作出“应用”或“抑制”的决策；（3）生成与决策一致的响应。通过线性探测（linear probes）证实，偏好适用性信号在生成过程中的隐藏状态中仍可解码。进一步发现，在多数情况下，错误决策的数量远超因生成过程导致的正确决策丢失，表明故障根源在于决策阶段。为此，论文提出ABIDE（Apply-Bias Investigation via Decision-score），基于信号检测理论，直接从对数概率（logits）中读取“应用-抑制”决策得分，以区分敏感度下降与响应偏差。实验揭示生成目标本身会诱发显著的“应用偏倚”（Apply bias）：仅引入答案生成任务即导致决策得分向“应用”倾斜，且该效应在控制提示结构、跨偏好槽位传播并持续存在。最终，通过在解码时减去一个从预留数据集估计的单一偏置标量，可在几乎不损害偏好满足率的前提下有效降低信息泄露。
链接: https://arxiv.org/abs/2609.34284
作者: Haeun Jang,Yonghyun Jun,Hwanhee Lee
机构: Chung-Ang University (中央大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Personalized LLMs must decide, for each stored preference, whether the current context calls for applying or suppressing it, which we call its applicability. They frequently over-personalize, applying preferences the context rules out, yet existing benchmarks score only the final response and cannot tell where this failure arises. We decompose preference handling into three stages and measure each separately: (1) knowing whether a preference applies, (2) deciding on an explicit Apply/Suppress label, and (3) generating a response consistent with that label. Using linear probes, we first show that this applicability signal remains decodable from hidden states during generation. By making the decision explicit, we then find that in most settings wrong decisions faithfully followed outnumber correct decisions lost in generation. We thus locate the failure in the decision, which breaks once the model is also asked to answer. To determine whether this reflects lost sensitivity or a response bias, we propose ABIDE (Apply-Bias Investigation via Decision-score), which adapts signal detection theory to Apply-vs-Suppress decision scores read directly from logits. ABIDE reveals a generation-induced Apply bias: merely stating an answer-generation objective shifts the decision score toward Apply while sensitivity is largely preserved, and the shift persists under controls for prompt structure, cascades across preference slots, and prompt wording. Finally, we show that subtracting a single bias scalar, estimated on a held-out split, from the decision score at decoding time reduces leakage while largely preserving fulfillment.
[NLP-123] BIABench: Evaluating AI agents on real-world bioimage analysis tasks
【速读】： 该论文旨在解决生成式人工智能（Generative AI）代理在真实生物图像分析任务中缺乏端到端评估基准的问题。由于生物图像数据（如二维图像、三维体数据及时间序列）通常体量庞大，无法作为完整上下文输入，代理需自主选择并执行代码、专用软件或可视化视图等多步骤分析流程，而现有研究未能提供可验证的测试框架。为此，论文提出BIABench，一个由16项源自已发表生物学研究重构的任务集合，保留了原始科学问题、成像数据和真实标签（ground truth），涵盖从常规组织学（HE histology）到单分子定位显微镜（single-molecule localization microscopy）等多种模态与十一种分析子任务。其解决方案的关键在于引入双维度评分机制：结果得分（outcome score）通过领域标准指标对比输出文件与真实标签；过程得分（process score）则由视觉-语言模型依据专家编写的评分标准评估方法选择与质量控制的合理性。实验表明，尽管通用型与生物学特化代理在二维任务上表现良好，但在涉及三维空间或时间维度的任务中，最高得分仅为0.19，且生物专业化、更强语言模型或详细指令均未能弥合差距。此外，代理行为高度不可靠，同一代理在重复运行间得分波动超过不同代理间的差异，且在无真实标签的情况下，过程得分与耗时均无法区分正确与错误的执行路径。因此，该研究的核心贡献在于构建了一个公开可复现的基准平台，为可靠长周期生物图像分析代理的评估与训练提供了可验证的框架。
链接: https://arxiv.org/abs/2609.34274
作者: Zixuan Pan,Davide Panzeri,Lukas Johanns,Marilin Moor,Yu Zhou,Hedi Peterson,Yiyu Shi,Jianxu Chen
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 41 pages, 6 figures, 11 tables
Abstract:Artificial-intelligence (AI) agents hold promise for automating bioimage analysis, yet no benchmark evaluates whether they can carry out real-world analyses end to end. Such analyses are hard for agents because 2D images, 3D volumes and time-lapse sequences are often too large to read as context, so an agent must choose and run an analysis through code, specialized software and rendered views. Published studies make this capability testable, because each pairs raw images with a peer-reviewed result. We introduce BIABench, a benchmark of 16 tasks reconstructed from published biological studies that retain their scientific questions, imaging data and ground truth. The tasks span eleven analysis subtasks and modalities from HE histology to single-molecule localization microscopy. Each submission receives an outcome score, which compares the output files with the ground truth using field-standard metrics, and a process score, in which a vision-language model judges method choice and quality control against an expert-written rubric. We evaluated general-purpose and biology-specific agents across several language models, with repeated runs of every task. Routine two-dimensional tasks were solved well, but on some tasks that added a third dimension or a time axis no agent scored above 0.19. Neither biological specialization, stronger models nor detailed expert instructions closed this gap. The agents were also unreliable, with scores varying more between repeated runs of one agent than between different agents, and without ground truth a correct run could not be told from a wrong one by its process score or by the time spent. Released openly with its data and code, BIABench provides a verifiable framework for evaluating, and eventually training, agents for reliable long-horizon bioimage analysis.
[NLP-124] Recursive LLM Degradation in Biomedical Question Answering: A Cross-Generation Study
【速读】： 该论文旨在解决在生物医学问答（Biomedical Question Answering, QA）任务中，语言模型反复基于自身生成的数据进行训练所引发的合成数据反馈循环（synthetic-data feedback loop）问题。这一过程可能导致错误信息和分布偏差在多轮训练中被不断放大，进而影响模型性能与可靠性。其解决方案的关键在于通过对比实验设计，系统评估递归合成数据训练（Recursive synthetic-data condition）与人类标注数据控制条件（Human-Control condition）在多代迭代中的表现差异。研究采用PubMedQA数据集及两个规模的Qwen2.5模型（0.5B和3B参数），在四代生成（G0–G3）中比较疾病实体和化学实体的F1分数、上下文支持率、语义相似性、答案长度等指标的变化趋势。结果表明，递归条件下的模型在多项关键指标上均出现显著退化，且3B模型的性能下降幅度大于0.5B模型，尤其体现在疾病实体F1、上下文支持率、余弦相似度和答案长度上。尽管在固定无重复3-gram解码约束下重复率未上升，但答案长度明显增加，揭示了合成数据反馈循环带来的领域特异性行为演化。该研究虽未直接量化临床幻觉率或普遍模型崩溃现象，但为理解合成数据训练在专业领域中的潜在风险提供了实证依据。
链接: https://arxiv.org/abs/2609.34257
作者: Bibek Bhandari,Kshitij Lingthep
机构: CMR Institute of Technology (CMR理工学院); Independent Researcher (独立研究员)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 6 pages, 2 figures, 1 table
Abstract:Repeatedly training language models on their own generated data may create a synthetic-data feedback loop in which errors and distributional biases are reintroduced into subsequent training datasets. This paper studies that process in biomedical question answering (QA) using PubMedQA and two Qwen2.5 model sizes, 0.5B and 3B parameters. The study compares a recursive synthetic-data condition, in which generation G(k+1) is trained on answers produced by G(k), against a Human-Control condition that repeatedly uses the original human training data. The study evaluates across four generations from G0-G3 with two random seeds (42 and 123) and a fixed evaluation set of 1,000 expert-labeled samples. The evaluation includes disease and chemical entity F1, context-supported rate, lexical and semantic similarity, answer length, repetition rate, and other evaluation metrics. The Recursive condition for both model sizes and both seeds showed larger declines than the Human-Control condition in disease entity F1, chemical entity F1, context-supported rate, ROUGE-L, and cosine similarity. Under the fixed no-repeat 3-gram decoding constraint, the main observed behavioral change was increased answer length, while the measured 3-gram repetition rate did not increase. The magnitude of the difference-in-change was larger for the 3B model than for the 0.5B model. This difference was particularly apparent in disease F1, context-supported rate, cosine similarity, and answer length. These results show domain-specific changes associated with using recursive synthetic-data training in biomedical QA, but do not establish clinical hallucination rates or universal model collapse.
[NLP-125] DreamingGoose: Staged Distillation from Autoregressive Transformers to Bidirectional Recurrent Diffusion Language Models
【速读】： 该论文旨在解决预训练自回归Transformer模型（autoregressive Transformers）在计算资源上的巨大沉没成本问题，即如何高效地将这些高成本训练的教师模型（如Qwen3 1.7B和8B）转化为计算更高效、结构更灵活的学生模型。现有方法通常仅通过改变模型架构（如从注意力机制转向循环结构）或损失函数目标（如从下一个词预测转为去噪任务）中的一个方面来实现转换，但从未同时调整二者。本文提出一种三阶段转换方案，将教师模型转化为无需注意力机制、双向且基于门控差分规则（gated-delta-rule）的扩散模型（diffusion students），从而可追踪每种能力在转换过程中的保留或丢失情况。其解决方案的关键在于：通过引入渐进式检索课程（retrieval curriculum）——逐步增加查询与键值表之间的语义距离——并结合运行时准确率阈值驱动的动态干预策略，实现了对上下文检索能力的稳定恢复。实验表明，固定训练步数会导致部分种子因学习时间不足而无法完成检索能力的习得，而基于性能阈值的自适应调度则能确保所有种子均成功学习检索，且该方法在真实文本上有效，并可扩展至8B规模模型。此外，研究还揭示了模型检索能力的覆盖限制（coverage limit）而非特定绑定记忆，进一步验证了其泛化机制的本质。
链接: https://arxiv.org/abs/2609.34253
作者: Julian Boesch,Andrew Wee,Alexander Stranzl
机构: Purdue University (普渡大学); Obit Research; State University of New York at Stony Brook (纽约州立大学石溪分校)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 8 pages, 1 figure, 2 tables. Companion to arXiv:2609.16183 . Code and result data at this https URL
Abstract:Pretrained autoregressive Transformers represent a large sunk investment in compute. Existing conversion methods reuse that investment by changing either the architecture (attention to recurrence) or the objective (next-token prediction to denoising), never both. We convert Qwen3 teachers at 1.7B and 8B into attention-free, bidirectional, gated-delta-rule diffusion students in three stages, so that each capability can be traced to the stage that kept or lost it. Language modeling transfers only partially and in-distribution; in-context retrieval does not transfer. On a multi-query recall probe where the teachers score 0.34-0.58, both converted students score 0.000, and diffusion pretraining alone does not restore retrieval. A retrieval curriculum in the final stage, which gradually lengthens the gap between a key-value table and the queries that address it, restores it only stochastically: on a fixed schedule, one seed in three learns to retrieve. Advancing the gap only while a running accuracy estimate stays above a threshold works for all three of those seeds, holds on real text, and carries unchanged to 8B, where two of three seeds succeed. The third had not learned within its fixed 16k-step budget: retrieval switches on abruptly at a seed-dependent step (6.5k and 11k in the other two), so a fixed budget can cut a late run off. One boundary survives every intervention: every model that learns retrieval scores 0.000 on tokens that never appeared in a retrieval episode, and an arm that resamples the key and value tokens every batch shows this is a coverage limit, not memorization of particular bindings. Separately, we convert a 7B code model into a 3:1 recurrent-attention block-diffusion hybrid over 85k steps and report two negative training results.
[NLP-126] SALMONN-duo: Adaptive Dual-System Coordination for Full-Duplex Voice Agents
【速读】： 该论文旨在解决全双工语音大语言模型（full-duplex speech large language models, LLMs）在真实应用场景中面临的实时性与复杂推理任务之间矛盾的问题。具体而言，尽管全双工语音LLMs能够实现低延迟、自然的语音交互，但实际智能体在执行工具调用和深度推理等操作时，其可变的延迟与高计算开销难以满足实时对话的严格时间要求。为此，论文提出SALMONN-duo，一种受认知双过程理论启发的自适应双系统语音代理架构。其核心解决方案在于通过分离实时交互与深度计算：系统1（快速思考）为始终在线的全双工语音LLM，负责即时响应并维持对话流畅性；系统2（慢速思考）为异步运行的强大推理代理，处理复杂任务。关键创新在于系统1具备动态决策能力，可自主判断何时直接回答、何时将任务委派给系统2，并在后台执行期间保持响应性，无缝融合返回结果而不暴露工具调用痕迹或丢失对话上下文。实验表明，基于知识边界感知的训练策略有效减少了不必要的系统2调用，而基于成本感知的强化学习进一步优化了任务性能与后端资源消耗之间的权衡，在单轮语音问答和多轮对话任务中显著提升准确率与安全性，尤其在模拟真实业务场景的τ-Voice平台中验证了其完成环境依赖、政策约束任务的能力。
链接: https://arxiv.org/abs/2609.34247
作者: Wenyi Yu,Siyin Wang,Terumi Chiba,Xianzhao Chen,Xiaohai Tian,Jun Zhang,Lu Lu,Chao Zhang
机构: Tsinghua University(清华大学); ByteDance(字节跳动)
类目: Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
备注:
Abstract:Full-duplex speech large language models (LLMs) enable low-latency, natural voice interaction. However, real-world agents must also use tools and perform deliberative reasoning-operations whose variable latency and computational cost conflict with the stringent timing requirements of real-time conversation. To reconcile these demands, we propose SALMONN-duo, an adaptive dual-system voice agent inspired by dual-process theories of cognition. SALMONN-duo separates real-time interaction from deliberative computation by pairing an always-on, fast-thinking full-duplex speech LLM (system 1) with a powerful asynchronous slow-thinking LLM agent (system 2). Beyond handling real-time interaction, system 1 learns when to answer directly and when to delegate, remaining responsive during backend execution and seamlessly integrating returned information into the ongoing dialogue without exposing tool traces or losing conversational context. Evaluations on single-turn spoken question answering (QA) and multi-turn conversations demonstrate that adaptive delegation substantially improves accuracy on knowledge-intensive and multi-hop reasoning questions, while knowledge-boundary-aware training avoids unnecessary system 2 invocations. On a customized version of \tau -Voice, SALMONN-duo further demonstrates its ability to complete environment-grounded, policy-constrained tasks through multi-turn interactions in realistic business scenarios. Finally, cost-aware reinforcement learning further enhances the trade-off between task performance and backend usage across the QA and conversation tasks, while improving task success and response safety on \tau -Voice with an acceptable increase in the delegation rate.
[NLP-127] Coherence-Aware Distributional Evaluation of Open-Ended Text Generation
【速读】： 该论文旨在解决现有开放文本生成评估指标在衡量生成文本质量时存在的根本性盲区——全局连贯性（global coherence）的缺失问题。当前主流指标如困惑度、熵、MAUVE、FBD及基于MMD的基线方法，主要依赖于词元级统计或通用表示空间中的分布相似性，难以捕捉生成文本在因果逻辑、篇章结构和主题一致性等方面的全局不一致问题。尽管这些文本可能在局部流畅且词汇多样性良好，但仍可能表现出内在矛盾或主题断裂。为此，论文提出关键解决方案：CHORD（Coherence-aware Hidden-state Open-generation Reference Distance），一种基于隐状态空间的连贯性敏感型分布度量方法。其核心在于通过设计一种能够激发连贯性信息的提示（coherence-eliciting prompt），将生成文本与人类写作语料在冻结的大语言模型（LLM）隐状态空间中进行编码，并采用带有RBF核的MMD（Maximum Mean Discrepancy）比较两者的分布差异。为验证该指标对连贯性退化的敏感性而非一般文本变化的误判，研究构建了一个反事实评估套件，系统性地引入渐进式连贯性破坏扰动并配以语义保持的对照条件。实验表明，CHORD能有效识别关系、话语、结构及混合类错误，而传统基线方法则无法区分此类缺陷与自然改写。因子消融分析进一步揭示，表示设计是实现连贯性敏感性的主要来源，而RBF-MMD则在表征具备可区分性后显著提升样本效率。此外，更大规模的模型能捕获更精细的差异，但只有当模型具备遵循提示进行无条件生成与前缀续写的能力时，连贯性提示才能发挥选择性增强作用。最终，CHORD在模型排名上与人类对输出合理性及人类写作感的判断高度一致，确立了表示设计在可靠分布评估中的核心地位。
链接: https://arxiv.org/abs/2609.34240
作者: Jinnuo Liu,Junhao Zhu,Weifeng Jiang,Haoming Liu,Hongyi Wen
机构: New York University (纽约大学); Center for Data Science, NYU Shanghai (纽约大学上海中心数据科学中心)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Preprint. 41 pages, 13 figures
Abstract:Existing metrics for open-ended text generation measure likelihood, lexical diversity, or distributional similarity in generic representation space, yet they can miss fundamental dimensions of quality. A prominent blind spot is global coherence: a generated passage may be locally fluent while remaining globally contradictory, causally inconsistent, or topically disconnected. Such failures can still preserve the token-level and lexical statistics that existing metrics rely on. We identify representation as a central bottleneck in detecting these failures and introduce CHORD (Coherence-aware Hidden-state Open-generation Reference Distance), a coherence-sensitive distributional metric. CHORD encodes generated and human-written corpora in the hidden-state space of a frozen LLM using a coherence-eliciting prompt, and compares the resulting distributions using MMD with an RBF kernel. To validate that the metric responds to coherence degradation but not generic textual change, we construct a counterfactual evaluation suite that pairs graded coherence-degrading perturbations with meaning-preserving controls. CHORD selectively detects relation, discourse, structural, and mixture failures that perplexity, entropy, MAUVE, FBD, and MMD-based baselines either miss or cannot separate from benign rewriting. Factorial ablations show that representation is the primary source of coherence sensitivity,while RBF-MMD improves sample efficiency once the relevant distinctions become visible. Larger backbones capture finer-grained distinctions, but coherence prompting improves selectivity only when the backbone can follow the this http URL unconditional generation and prefix continuation, CHORD yields model rankings that strongly align with human judgments of whether outputs make sense and appear human-written. Together, these results establish representation design as central to reliable distributional evaluation.
[NLP-128] MAS-OPD: On-Policy Distillation for Multi-agent Systems
【速读】： 该论文旨在解决多智能体系统（Multi-agent Systems, MAS）在复杂任务中面临的协同能力不足与角色专业化不明确的问题。现有方法主要依赖推理时的编排，而通过微调实现联合训练的方案受限于强化学习中团队级奖励信号模糊、局部奖励需针对任务重新设计等缺陷，难以有效指导各智能体的角色分工与协作行为。本文提出MAS-OPD框架，其核心创新在于将有监督的策略蒸馏（On-Policy Distillation, OPD）引入多智能体协同训练，以提供细粒度的令牌级教师监督信号，避免对局部奖励函数的依赖。关键解决方案包括：1）角色优势专化（Role-Advantage Specialization），通过比较目标角色与非目标角色条件下的教师信号差异，量化并引导各智能体形成互补性角色优势；2）协作特权归因（Privileged Attribution for Coordination），识别交互冲突的来源，并将该信息作为特权信号仅供给教师模型，从而增强跨智能体协作信息的可利用性。实验结果表明，该方法在代码生成与数学推理基准上均取得最优平均性能，显著提升了智能体的角色分化程度与协作效率。
链接: https://arxiv.org/abs/2609.34234
作者: Qiyong Zhong,Mao Zheng,Mingyang Song,Houcheng Jiang,Jiajie Su,Huwei Ji,Li Zhang,Junfeng Fang
机构: University of Science and Technology of China (中国科学技术大学); Foundation Model Department, Tencent(腾讯基础模型部门); Zhejiang University (浙江大学); National University of Singapore (新加坡国立大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Multi-agent systems (MAS) split a task across specialized roles and are promising on complex tasks, yet a prevailing approach relies on inference-time orchestration alone. General-purpose APIs are costly and hard to customize, while small models with role prompts rarely develop stable role competence or reliable collaboration, so post-training a MAS jointly is central. Most attempts use reinforcement learning, whose team-level reward leaves undetermined which step of which agent brought about the outcome, while local rewards need redesigning per task. On-policy distillation (OPD) gives token-level teacher supervision on trajectories the student samples, a denser signal needing no local reward, yet is underexplored for the interdependent agents of a MAS. Two difficulties arise: building complementary specialization from a judgement of which role a behavior belongs to while preserving the knowledge all roles need, and turning cross-agent collaborative information into supervision OPD can exploit. We present MAS-OPD, where Role-Advantage Specialization defines the role advantage as the difference between the teacher signals under target and non-target role conditions, and Privileged Attribution for Coordination attributes an interaction conflict to its source and supplies it to the teacher alone as privileged information. Extensive experiments on code and mathematics benchmarks show that MAS-OPD attains the highest mean score at both student scales and leads the agents to develop clearer role specialization and more effective collaborative behavior.
[NLP-129] USA: Update-aware SAM for Cross-domain On-Policy Disitllation of Language Agents
【速读】： 该论文旨在解决多领域知识融合中因数据分布冲突导致的负迁移问题，尤其是在基于策略的蒸馏（on-policy distillation）框架下，当多个领域数据混合时，由于不同领域间更新方向不一致，模型合并（model merging）往往无法有效提升性能，甚至在部分领域对之间出现性能退化。其解决方案的关键在于提出一种名为USA（Unbiased Step Adjustment）的新方法，通过在短时预热阶段测量每个参数的更新幅度，并将其转化为坐标级扰动半径，从而精确地降低易受跨域更新耦合影响的坐标上的曲率。该机制有效缓解了因双域共同更新导致的参数位移问题，显著提升了跨领域迁移效果，在数学、科学和代码三个任务领域、两种学生模型规模下均实现优于单领域基准的性能，平均提升超过4个百分点，并成功逆转了存在负迁移的领域对的性能下降趋势。
链接: https://arxiv.org/abs/2609.34225
作者: Qiyong Zhong,Mao Zheng,Mingyang Song,Huwei Ji,Houcheng Jiang,Jiajie Su,Li Zhang,Gengsheng Li,Junfeng Fang
机构: University of Science and Technology of China (中国科学技术大学); Foundation Model Department, Tencent(腾讯基础模型部门); Zhejiang University (浙江大学); National University of Singapore (新加坡国立大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:On-policy distillation instils multi-turn agentic reasoning through dense token-level supervision on the student’s own trajectories, but a single domain saturates early, so further supervision has to be drawn from other domains. Multi-domain data mixing is the most direct way of incorporating them, at the cost of conflicts between their data distributions and of retraining the entire model whenever one domain is revised. Model merging avoids both by distilling every domain independently and fusing the resulting task vectors afterwards. We find instead that the benefit polarizes across domain pairs: on those exhibiting negative transfer, every merging operator we evaluate falls below the single-domain reference. We attribute this to cross-domain update coupling, where a substantial fraction of coordinates is updated comparably by both domains and a merge can therefore displace them by as much as their own updates. To overcome this limitation, we propose USA, which converts per-parameter update magnitudes measured during a brief warm-up into per-coordinate perturbation radii, reducing curvature precisely on the coordinates that carry most of the merging displacement. Experiments across mathematics, science and code at two student scales show USA strongest in all six transfer directions, ahead of the single-domain reference by more than four points on average, and reverse the negative transfer of the conflicting pairs.
[NLP-130] Loop Dropout: Regularizing Shared Updates in Looped Language Models
【速读】： 该论文旨在解决循环语言模型（Looped Language Models）中低秩适应（LoRA）存在的“晚期循环偏差”问题，即共享更新在后期循环位置表现更优，而早期循环位置的适应效果较差。其核心解决方案是提出环路丢弃（Loop Dropout），通过将适配器应用的随机掩码与逆生存率重缩放相结合，在训练过程中保持期望更新强度不变，从而促进各循环位置间适应能力的均衡性。该方法在不增加推理阶段可训练参数或计算开销的前提下，实现了对所有主干循环的持续激活，并显著提升了模型在数学推理、通用指令微调及代码生成等任务上的性能。实验表明，该方法优于现有LoRA变体和适配器正则化技术，且增强了早期循环的适应能力。
链接: https://arxiv.org/abs/2609.34218
作者: Zirui Zhu,Hailun Xu,Xuanlei Zhao,Yong Liu,Yingxuan Ren,Kanchan Sarkar,Kun Xu,Yang You
机构: National University of Singapore (新加坡国立大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Looped language models separate computational depth from parameter count by repeatedly applying the same transformer block. Adapting these models requires a shared update that remains effective as hidden states evolve throughout the recurrent computation. Our empirical analysis reveals a pronounced late-loop bias in standard low-rank adaptation (LoRA): the shared update is more effective at later loop positions. This imbalance motivates training shared updates under varying combinations of their applications. Randomly omitting adapter applications alone, however, does not improve task performance; it reduces expected update strength during training while leaving inference unchanged. We introduce Loop Dropout, which couples stochastic masking of adapter applications with inverse-survival rescaling to preserve expected update strength and promote effective adaptation across loops. Extensive experiments demonstrate improved mathematical reasoning across model sizes, adapter ranks and training recipes, with benefits extending to general instruction tuning and code generation. Loop Dropout outperforms existing LoRA variants and adapter regularizers, while further analysis shows stronger early-loop adaptation. Every backbone loop remains active, and inference applies the adapter at all loops using standard LoRA without additional trainable parameters or inference computation.
[NLP-131] X-MoD: Practical Scaling Laws for Sparse-Depth Routing Beyond Mixture-of-Depths
【速读】： 该论文旨在解决现有混合深度（Mixture-of-Depths, MoD）架构中因固定稀疏-密集交替模式导致的“总容量”与“活跃容量”紧密耦合问题，限制了稀疏深度扩展的灵活性。其核心挑战在于如何在不增加实际计算负载的前提下提升模型总参数量，从而实现更高效的可扩展性。为此，论文提出X-MoD——一种可扩展的稀疏深度架构，通过解耦令牌稀疏性与锚点步长（anchor stride），使总参数量可增长而保持等效活跃容量近似恒定。关键解决方案包括：引入密集锚点（dense anchors）以增强深层路由的可训练性，采用方差缩放的逐层门控机制与深度维度上的令牌平衡策略，确保稀疏路由稳定优化；同时构建一个基于计算量、上下文长度和等效骨干规模的条件架构设计框架，将稀疏深度路由建模为可分析、可预测的系统性问题。研究进一步发展出一套实用的缩放定律框架，通过拟合与FLOP匹配的稠密基线模型，得到一个可解释的性能分解公式，涵盖稀疏容量增益、稀疏上下文修正及锚点步长交互效应。该定律能够准确预测不同路由配置下的验证损失，并揭示上下文长度、模型规模与锚点步长对稀疏深度性能的影响机制。实验验证涵盖预训练消融、独立缩放定律预测、下游任务评估以及与稠密模型、MoD及代表性Mixture-of-Experts（MoE）基线的对比，充分证明了X-MoD架构与所提缩放定律的有效性与实用性。
链接: https://arxiv.org/abs/2609.34212
作者: Bowen Dong,Yilong Fan,Tengyu Pan,Yike Zhang,Zhenyu Li,Zijian Zhang,Xuewei Li,Mei Yu,Jianyong Wang
机构: Tsinghua University (清华大学); Tianjin University (天津大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Mixture-of-Depths (MoD) enables conditional computation across Transformer depth by routing only a subset of tokens through selected layers, but its original one-sparse–one-dense alternation tightly couples total capacity to active capacity and limits sparse-depth scaling. We introduce X-MoD, a scalable sparse-depth architecture that decouples token sparsity from anchor stride, allowing total parameter count to grow while keeping active-equivalent capacity nearly fixed. To make deep sparse routing trainable, X-MoD combines dense anchors with variance-scaled layer-wise gating and depth-wise token balancing. To make this regime analyzable and usable, we formulate sparse-depth routing as a conditional architecture-design problem: given compute, context length, and active-equivalent backbone size, how should the routing configuration be chosen? We develop a practical scaling-law framework by fitting X-MoD relative to FLOP-matched dense baselines, yielding an interpretable law that decomposes performance into sparse-capacity gain, sparse-context correction, and anchor-stride interaction. The law predicts validation loss across routing configurations and reveals how context length, model scale, and anchor stride shape sparse-depth performance. We validate the architecture and law through pretraining sweeps, held-out scaling-law prediction, ablations, downstream evaluations, and comparisons with Dense, MoD, and representative MoE baselines.
[NLP-132] LLM s are not stochastic parrots: Evidence for meaning-mediated abstraction from conlang-like tasks
【速读】： 该论文旨在挑战“强随机鹦鹉”（strong stochastic parrot）假说，即认为大语言模型（LLMs）仅依赖统计模式匹配，无法实现抽象或推理，始终停留在表层模式复用的低层次认知水平。为检验这一假设，研究设计了类人工语言（conlang-like）任务，向多个LLM提供仅包含自然语言描述的虚构语言规则，这些规则通过组合训练数据中罕见且未被记录的特征来颠覆主流表层模式，且不提供任何示例输出。关键在于，若模型能够正确遵循这些非显性、反直觉的规则，则其行为必超越表面统计相关性，需具备对提示中约束条件的语义表征能力。实验结果表明，尽管模型性能存在差异，但在三类互补任务中均系统性地表现出符合预期意义的方向性行为：区分提示暴露与指令执行、根据新约束调整语义关系，甚至在复杂翻译任务中生成精确匹配答案。这表明，当在适当的架构与上下文约束下，基于统计学习的模型可产生以意义为导向的抽象能力，从而反驳了“强随机鹦鹉”假说。研究揭示，尽管生成仍受表面合理性强烈制约，但语义中介的抽象机制确实可能从合理文本生成目标中涌现，对模型设计及理解抽象表征如何从统计学习中演化具有重要意义。
链接: https://arxiv.org/abs/2609.34187
作者: Julia Witte Zimmerman,Calla G. Beauregard,Tabia Tanzin Prama,Parisa Suchdev,Kathryn Cramer,Elisabeth Kollrack
机构: University of Vermont (佛蒙特大学); Vermont Complex Systems Institute (佛蒙特复杂系统研究所); Computational Story Lab (计算故事实验室); Computational Ethics Lab (计算伦理实验室)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:The strong version of the stochastic parrot argument claims that, although large language models (LLMs) may exceed rote regurgitation, they cannot move beyond statistical pattern matching into abstraction or reasoning, remaining ontologically near the lower bound of pattern reuse despite producing alluringly fluent text. We test this hypothesis using conlang-like tasks. Several LLMs are given only natural-language descriptions of fictional languages that subvert prominent superficial patterns in training data by combining statistically uncommon and unattested features. Crucially, no example outputs are given. We argue that if the models exhibit rule-following behaviour, they cannot be relying solely on superficial statistical patterns; such patterns often work against the correct output. Instead, successful performance requires representations of the constraints specified in the prompt. Across three complementary task families, models systematically move in the meaning-predicted direction: they distinguish prompt exposure from instructed use, alter semantic relationships in response to novel constraints, and sometimes produce exact matches to complex translation answer keys. Although performance varies across the spectrum of models used, these results provide evidence for meaning-mediated abstraction in LLMs and refute the strong stochastic parrot hypothesis. Our work shows that, under appropriate architectural and contextual constraints, statistical learning can produce meaning-mediated abstractions, although generation remains strongly constrained by superficial plausibility. We discuss implications for model development and for understanding how increasingly abstract representations may emerge from plausible-text-generation objectives.
[NLP-133] RAG Warrant: Evidence-Preserving Governance for RAG Policy Promotion Under Quality Cost Latency and Risk Constraints
【速读】： 该论文旨在解决检索增强生成（Retrieval-Augmented Generation, RAG）系统在部署决策中缺乏安全评估机制的问题，即现有工具虽提供丰富指标、基准测试与自动化评判，却无法有效判断策略变更是否具备上线安全性。其解决方案的核心在于提出RAGWarrant——一个开源的发布控制框架，将部署决策重构为基于证据的约束性判断过程，而非简单的排行榜选择。该框架通过归一化评估器输出与运行时遥测数据，应用预设的质量阈值与硬性风险屏障，设定证据类别声明上限，保留负面结果，并生成可审计的PROMOTE、BLOCK、REJECT或INCONCLUSIVE四类决策。实验验证覆盖T2-RAGBench、MultiHop-RAG、CRAG、HotpotQA、合成复现及受限本地生成等场景，结果显示在HotpotQA上因答案质量低于预设阈值而阻止了操作节省；在受限CRAG研究中，尽管选择了成本更低但性能绑定的策略，生成效果仍不稳定，且预留的护栏机制失效。研究强调其贡献在于构建了一个可审计的发布控制抽象模型，而非优化器性能优越性、人工验证能力或生产就绪性。配套可标记的软件包支持从全新克隆重建，以加固的Docker任务形式运行，兼容外部评估器导出并验证产物完整性。
链接: https://arxiv.org/abs/2609.34179
作者: Richard Krueger,Lucas Krause,Zach Pocquette
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 19 pages, 6 figures, 7 tables. Preprint v0.1.1-rc1. Code and artifacts: this https URL
Abstract:Retrieval-augmented generation systems are extensively instrumented with metrics, benchmarks, traces, and automated judges, but these tools do not decide whether a proposed policy change is safe to release. We present RAGWarrant, an open-source promotion-control framework that treats deployment as a constrained evidence decision rather than a leaderboard choice. RAGWarrant normalizes evaluator outputs and operational telemetry, applies predeclared quality and hard-risk gates, assigns evidence-class claim ceilings, preserves negative outcomes, and emits auditable PROMOTE, BLOCK, REJECT, or INCONCLUSIVE decisions. We evaluate the framework across T2-RAGBench, MultiHop-RAG, CRAG, HotpotQA, synthetic reproduction, and bounded local generative experiments. On HotpotQA, operational savings were blocked because answer quality fell beyond the declared margin. A bounded CRAG study selected a lower-cost quality-tied policy, but related generative gains were unstable and a held-out guardrail failed closed. We claim an auditable promotion-control abstraction, not optimizer superiority, human validation, or production readiness. The tagged artifact reproduces from a fresh clone, runs as a hardened Docker job, accepts external evaluator exports, and verifies artifact integrity.
[NLP-134] oward a Graded Measure of Belief Stability in Large Language Models
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在信息获取与推理过程中事实可靠性评估过于依赖单一判断的问题。传统方法仅衡量模型对单个陈述的支持程度，忽视了信念在模型整体认知体系中的稳定性。为此，论文提出“分级信念稳定性”（graded belief stability）这一关系性度量指标，用以评估一个信念在其更广泛信念系统中是否具有持久性。其核心解决方案在于引入一种基于模型内部表征的直接条件估计器（Direct Conditional estimator），通过分析特定主张与其他认知承诺之间的相互关系，估算条件信念概率。实验结果表明，在12个大语言模型和三个领域中，低稳定性信念在对话挑战下表现出更高的平均行为变动性（83.3%的模型-领域组合中），且该现象在控制个体信念概率后依然显著。因此，分级信念稳定性将可靠性评估从信念强度扩展至信念在系统内部的稳健性，为评估大语言模型认知一致性提供了新范式。
链接: https://arxiv.org/abs/2609.34158
作者: Samantha Dies,Branden Fitelson,Tina Eliassi-Rad
机构: Northeastern University(东北大学); Santa Fe Institute(圣菲研究所)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) increasingly mediate how people access and reason with information, yet factual reliability is usually evaluated one judgment at a time. We introduce graded belief stability, a relational measure of how well a belief persists within an LLM’s broader belief system. Unlike individual belief probability, it asks whether support for a claim persists when that claim is considered alongside the model’s other epistemic commitments. We operationalize this idea with a Direct Conditional estimator that uses internal model representations to estimate conditional belief probabilities. Across 12 LLMs and three domains, lower-stability beliefs exhibit greater mean behavioral movement under conversational challenge in 83.3% of model-domain settings after matching on individual belief probability. Graded belief stability therefore extends reliability assessment beyond how strongly an LLM supports a claim to how robustly that belief is supported within its broader system of beliefs.
[NLP-135] Quantitative Measurement of Language Distance among Closely Related Indo-European Languages Using Pretrained Language Models: A Case Study on the North Germanic Branch
【速读】： 该论文旨在解决北日耳曼语族（North Germanic languages）中语言距离量化缺乏统一多维度计算框架的问题，传统方法依赖定性分析，难以实现可重复的定量评估。其核心解决方案是基于预训练语言模型构建一个三指标定量分析框架：（1）句级语义距离，通过LaBSE与mBERT对齐句对编码后的余弦相似度计算；（2）正字法碎片率（orthographic fragmentation rate），衡量跨语言使用单语BERT词表进行子词分词时的效率损失；（3）掩码语言建模预测性（MLM predictability），比较不同语言在mBERT模型下的预测置信度与熵值差异。研究以Tatoeba语料库中的150组三语平行句为控制样本，利用两个独立模型（LaBSE和mBERT）得出一致的语言距离排序，结果支持历史语言学结论——“丹麦对挪威400年统治（1380–1814）导致书面语高度同源”。三个维度指标相互印证，形成可复现的计算范式，为密切相关的语言间距离研究提供了可扩展至印欧语系更多分支的定量分析基础。
链接: https://arxiv.org/abs/2609.34152
作者: Yiping Bai
机构: Guangdong Haiqixing Marine Technology Co., Ltd.(广东海启星海洋科技有限公司)
类目: Computation and Language (cs.CL)
备注:
Abstract:Among closely related North Germanic languages, the quantification of language distance has traditionally relied on qualitative methods, lacking a unified multi-dimensional computational framework. Multilingual pretrained models based on the Transformer architecture can map texts from different languages into a shared vector space, enabling quantitative measurement of language distance. This paper focuses on the three North Germanic languages—Danish, Norwegian (Bokmål), and Swedish—and proposes a three-metric quantitative framework based on pretrained language models: (1)~sentence-level semantic distance, computed as cosine similarity between LaBSE and mBERT encodings of parallel sentences; (2)~orthographic fragmentation rate, measuring subword tokenization efficiency when cross-applying monolingual BERT vocabularies to parallel texts; (3)~MLM predictability, comparing prediction confidence and entropy in masked language modeling using mBERT across languages. Using 150 trilingual parallel sentence triplets from the Tatoeba corpus as controlled samples, we obtain consistent distance rankings on two independent models: LaBSE: da–no 0.012 no–sv 0.016 da–sv 0.020 ; mBERT: da–no 0.016 no–sv 0.045 \approx da–sv 0.046 . This ranking is consistent with the historical linguistic conclusion that ``400 years of Danish rule over Norway (1380–1814) led to highly cognate written languages.‘’ The three metrics—semantic, orthographic, and predictability—converge on the same conclusion, providing a reproducible computational framework for the quantitative study of distance among closely related languages, extensible in principle to more branches of the Indo-European language family, pending validation on additional language groups.
[NLP-136] Word Similarity Datasets for Indian Languages: Annotation and Baseline Systems
【速读】： 该论文旨在解决低资源语言（特别是印度本土语言）在词向量表示质量评估中缺乏可靠、人工标注的词相似度数据集的问题。由于现有评估基准主要集中在英语等高资源语言，导致对印地语和孟加拉语以外的印度主流语言（如乌尔都语、泰卢固语、马拉地语、旁遮普语、泰米尔语和古吉拉特语）的词表示性能评估存在显著空白。其解决方案的关键在于：通过翻译与重新标注英文词相似度数据集的方法，构建六种印度语言的高质量、人工标注的单语词相似度数据集，并基于此为乌尔都语、泰卢固语和马拉地语三种语言提供采用前沿技术训练的词表示模型的基线评估分数，从而为相关语言的自然语言处理研究提供可信赖的评估基准。
链接: https://arxiv.org/abs/2609.34138
作者: Syed S. Akhtar,Arihant Gupta,Avijit Vajpayee,Arjit Srivastava,M. Shrivastava
机构: International Institute of Information Technology (国际信息科技学院); Hyderabad, Telangana, India
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:With the advent of word representations, word similarity tasks are becoming increasing popular as an evaluation metric for the quality of the representations. In this paper, we present manually annotated monolingual word similarity datasets of six Indian languages - Urdu, Telugu, Marathi, Punjabi, Tamil and Gujarati. These languages are most spoken Indian languages worldwide after Hindi and Bengali. For the construction of these datasets, our approach relies on translation and re-annotation of word similarity datasets of English. We also present baseline scores for word representation models using state-of-the-art techniques for Urdu, Telugu and Marathi by evaluating them on newly created word similarity datasets.
[NLP-137] raining and Inference Dynamics of PLDR-LLM s: Row-Map Collapse Renormalization and Predictive Reduction
【速读】： 该论文旨在解决生成式语言模型中训练与推理过程的统一建模问题，特别是针对幂律解码器表示语言模型（PLDR-LLMs）在参数动态演化、能量变化与预测性能之间的内在关联机制。其核心挑战在于如何在有限样本条件下精确刻画模型内部行级（row-level）动态行为与其全局观测结果之间的复杂耦合关系，尤其是在优化器记忆、数据调度与数值策略共同作用下的非平衡动力学过程。解决方案的关键在于提出一套基于精确有限工作恒等式（exact finite work identities）的理论框架，将行中心学习映射的绝对能量变化分解为参数贡献、符号相互作用及数值观测缺陷三部分，结合正仿射阻塞（positive affine blocking）维持行常数面的重启行为，并利用增强型AdamW状态实现完整的动态描述。通过预测性重整化（predictive renormalization），该框架在单次遍历不同语料目标块的条件下，保留优化器记忆与数值策略，从而支持条件性训练定律的完整建模。进一步地，通过引入有限群体协方差、匹配物理时钟、矩阵通量与符号时间能量等概念，建立了行级动态与模型整体观测之间的定量联系，区分了绝对行坍缩、相对行集中、算子稳定性和预测精度等关键现象。实验验证表明，模型表现具有观测者与优化器依赖性，排除了所测试的自主行态候选方案，支持有限条件预测与特定状态下的算子约简。最终，该理论通过严格区分精确恒等式、条件动力学假设与有限实证发现，提供了可证明、可形式检验且具备紧凑数值证据支撑的系统性分析体系。
链接: https://arxiv.org/abs/2609.34130
作者: Burc Gokden
机构: Fromthesky Research Labs LLC(Fromthesky研究实验室有限责任公司)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Monograph; 655 pages, 76 figures, 311 tables
Abstract:This monograph develops a unified account of training and inference in Power Law Decoder Representation language models (PLDR-LLMs). Exact finite work identities decompose changes in the absolute energy of the row-centered learned map into parameter contributions, signed interactions, and numerical observation defects. Positive affine blocking retains restarts at the row-constant face, while the augmented AdamW state supplies the complete dynamical description. Predictive renormalization acts on the complete conditional training law for a single pass over distinct corpus target blocks, retaining optimizer memory, remaining data, schedule, and numerical policy. Autonomous reductions require closure; approximate reductions carry successor and emission errors. Finite-population covariance, matched physical clocks, matrix fluxes, and signed temporal energy connect row dynamics to model-wide observations. Absolute row collapse, relative row concentration, operator stabilization, and predictive accuracy are distinguished. Experiments reveal observer and optimizer dependence, reject the tested autonomous row-state candidates, and support finite conditional prediction and state-specific operator reduction. Independent single-pass families exhibit moving finite fluctuation regions without establishing a thermodynamic critical class. Conditional symmetry, head limits, covariance flows, and readout error budgets specify assumptions needed to transfer scaling laws to inference. The theory separates exact identities, conditional dynamical claims, and finite empirical findings, with proofs, selected formal checks, and compact numerical evidence. Comments: Monograph; 655 pages, 76 figures, 311 tables Subjects: Machine Learning (cs.LG); Computation and Language (cs.CL) MSC classes: 68T07 (Primary), 37N40, 60F05, 65G20, 82B28 (Secondary) Cite as: arXiv:2609.34130 [cs.LG] (or arXiv:2609.34130v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.34130 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-138] Understanding Clinical Cognitive Dialogues Using Large Language Models
【速读】： 该论文旨在解决临床认知评估中对话结构缺乏系统标注与可量化的难题，即当前临床对话资源极少对交互结构进行细致标注，难以支持对评估过程中医患互动行为的大规模研究。其核心解决方案是构建一个去标识化的语料库，包含33次认知评估对话、共计8,250个话语单元，并对三个说话者角色及56种对话行为（Dialogue Act）进行了精细标注。基于此语料库，研究者对大语言模型（LLM）在细粒度对话行为分类和下一句患者话语生成任务上进行基准测试，并探索域外指令数据与解释增强训练在该临床场景中的迁移能力。结果显示，指令微调显著提升了患者话语的参考匹配效果，而具备推理感知能力的微调策略在LLaMA-3.1-8B系列模型中取得了最佳分类性能。然而，即使表现最优的模型仍难以区分语义相近的对话行为，表明模型更擅长识别宽泛的交际意图而非细微的沟通功能。该语料库与评测框架使认知评估中的交互结构可测量化，为后续研究如对话标记识别、临床人员培训以及经验证的虚拟患者模拟提供了基础支持。本研究不涉及诊断结论，而是提供数据与评估体系以推动相关应用研究。
链接: https://arxiv.org/abs/2609.34125
作者: Vishalakshi Arumugam,Dan Schumacher,Veronica Rammouz,Enrique Gonzalez Guerrero,Jeremy Davis,Anthony Rios
机构: College of AI, Cyber and Computing; University of Texas at San Antonio (美国德克萨斯大学圣安东尼奥分校)
类目: Computation and Language (cs.CL)
备注: 9 pages
Abstract:In-person cognitive assessment is both a test and an interaction. Clinicians explain tasks, repair misunderstandings, and adapt to patient responses, while patients may hesitate, seek clarification, or disengage. Yet clinical dialogue resources rarely label the interaction structure needed to study these behaviors at scale. We present an de-identified corpus of 33 cognitive assessment conversations with 8,250 utterances annotated for three speaker roles and 56 dialogue acts. We use this corpus to benchmark large language models on fine-grained dialogue-act classification and next-patient-utterance generation. We also test whether out-of-domain instruction data and explanation-augmented training transfer to this clinical setting. Instruction tuning produces the strongest patient-utterance reference matching and improves classification accuracy. Reasoning-aware fine-tuning produces the strongest classification results among the LLaMA-3.1-8B variants. However, even the best models struggle to separate closely related dialogue acts, showing that broad conversational intent is easier to recognize than fine-grained communicative function. The corpus and benchmark make interaction structure measurable in cognitive assessments and support follow-up work on conversational markers, clinician education, and carefully validated simulated patients. This work does not make diagnostic claims. Instead, it provides the data and evaluation framework needed to study these applications.
[NLP-139] Unknown is not normal: separating language-model extraction from rule-based decision logic for clinical risk scores
【速读】： 该论文旨在解决在临床风险评分计算中，由于自由文本病历信息不完整，传统方法将缺失项默认为正常状态（missing-equals-normal）所导致的隐性误分类问题。其核心挑战在于如何在保证评分准确性的同时，减少不必要的临床询问，并避免因过早决策而引发的误判。解决方案的关键在于将三态提取（present/absent/unknown，由大语言模型（LLM）完成）与决策逻辑（基于确定性代码对未知输入进行边界计算）分离，构建一种“边界策略”（bounds policy）。该策略仅向临床医生提出那些可能改变最终决策的必要问题，从而显著降低提问数量（减少50%），消除无关提问，并有效防止提前决策。实验表明，在多种模拟和真实病例中，该方法在保持接近“全问”准确率（99.4%）的同时，大幅提升了决策效率与安全性，且对噪声环境具有鲁棒性，优于传统的缺失值归零策略及端到端生成式代理（如Claude Opus 5.5），尤其在小规模本地模型（9B参数）上亦可达到近似“最优”性能。
链接: https://arxiv.org/abs/2609.34112
作者: Nicolás Vera Zúñiga
机构: Independent Researcher(独立研究员); Chile(智利)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 14 pages (7 of main text), 5 figures, 2 tables, appendix included; full supplementary material in the code repository. Code: this https URL (archived: this https URL )
Abstract:Large language models (LLMs) are increasingly used to compute clinical risk scores from free-text notes. Notes are often incomplete, and treating undocumented findings as normal can silently misclassify patients. We test whether separating three-state extraction (present, absent or unknown, by an LLM) from decision logic (deterministic code computing score bounds over unknown inputs) lets a system ask only questions that can change the decision. On 1,200 synthetic emergency cases across six calculators (HEART, CURB-65, qSOFA, PERC, Wells, Cockcroft-Gault), with a simulated clinician answering questions, we compared this bounds policy with asking for every missing input, a missing-equals-normal schema, and an end-to-end LLM agent (Claude Opus 5.5). With Claude Haiku 4.5 as extractor, the bounds policy matched ask-all accuracy (99.4% vs 99.4%) with half the questions (0.92 vs 1.78 per case) and no irrelevant ones. Treating missing as normal dropped accuracy to 91.2% and under-triaged 8.5% of patients (95% CI 7.1-10.2), and under-triage persisted under messy notes and a noisy clinician. The agent was equally accurate under ideal conditions (99.6%) but 9.5% of its questions were irrelevant; with a noisy clinician it was less accurate than the bounds policy (83.5% vs 87.0%, p0.001) and committed prematurely in 2.7% of cases (bounds: 0%). A 9B local model as extractor reached oracle-level accuracy (99.8%). In 584 real case reports from MedCalc-Bench, only 52% contained enough information to determine the category (HEART 13%). Routing decisions through code that reasons explicitly about unknowns avoids premature commitment and irrelevant questions, halves the questions asked, and works with small local models.
[NLP-140] Evaluating Machine Unlearning in ASR ICASSP2027
【速读】： 该论文旨在解决自动语音识别（ASR）领域中机器遗忘（Machine Unlearning, MU）的适用性问题，尤其是在满足“被遗忘权”（right to be forgotten）法规要求方面的挑战。当前尽管MU在语音任务中逐渐受到关注，但其在ASR场景下的有效性与评估方法仍缺乏系统研究。论文的核心问题是：现有MU算法与评估工具是否适用于ASR任务？为此，研究通过在ASR模型上应用多种MU技术，系统评估了单主体遗忘、顺序遗忘和同时遗忘情境下的隐私-性能权衡。研究表明，基于梯度上升的算法在隐私与性能之间实现了较优平衡，而更复杂的遗忘方法反而导致过遗忘（over-unlearning），使样本更容易被识别为已遗忘数据。这一发现揭示了当前依赖简单成员推断攻击（Membership Inference Attacks）的标准化隐私评估方法不足以可靠衡量遗忘成效，因而亟需改进的评估范式。此外，研究还发现顺序与同时遗忘的隐私保护与模型性能均劣于单主体遗忘，凸显了针对多主体遗忘场景设计更高效遗忘构造的必要性。
链接: https://arxiv.org/abs/2609.34092
作者: Diogo Dinis,Francisco Teixeira,Bhiksha Raj,Alberto Abad,Isabel Trancoso
机构: 未知
类目: Computation and Language (cs.CL)
备注: Submitted to ICASSP 2027
Abstract:Machine unlearning (MU) offers a path to compliance with “right to be forgotten” regulations. While MU has received increasing attention for speech tasks, it remains largely unexplored for Automatic Speech Recognition (ASR). In this work, we investigate whether existing MU algorithms and evaluation tools are suitable for ASR. We apply several MU techniques to an ASR model, evaluating privacy-utility trade-offs for single-subject unlearning, then assess the best algorithm under sequential and simultaneous unlearning. Results show that gradient ascent-based algorithms achieve strong utility-privacy trade-offs, whereas more complex approaches over-unlearn samples, making them easier to identify as unlearned. This suggests standard privacy evaluations based on simple Membership Inference attacks are insufficient to reliably assess unlearning success, motivating improved evaluation methods for MU in ASR. Finally, we show that both sequential and simultaneous unlearning yield worse privacy and utility than single-subject unlearning, underscoring the need for unlearning constructions better suited to these settings.
[NLP-141] Who Gets a Token and What Does It Carry? Unequal Name Support and Concept Access in Large Language Models
【速读】： 该论文旨在解决语言模型在处理姓名时存在的隐性偏差问题，即尽管姓名在语义上被假设为可比的输入，但其在词汇层面的实际表示却因分词方式差异而存在不平等。其核心问题是：匹配的姓名在字面表征（name-surface）上可能并不具备等价的词汇支持，部分姓名可直接以单个标记（token）访问，而另一些则被拆分为多个子词（subword），导致不同种族与性别关联的姓名在模型输入层级就已存在结构性不平等。解决方案的关键在于提出名为NameTrace的模型原生、细粒度、前行为分析框架，通过测量模型对任务相关形容词轴上的概念可及性（concept accessibility），以连续且任务对齐的权重量化命名实体在内部表示中的实际可见性。实验证明，即使在同种族/族裔-性别分层内，具有不同碎片化程度的姓名仍表现出系统性的概念可及性差异，这种差异在各类任务（如奖学金评选、招聘、临床评估、贷款审批）中持续存在，并跨模型家族和未见姓名泛化。此外，隐藏状态干预进一步表明这些任务方向具有下游决策影响力。研究揭示了不平等的词汇支持不仅存在于词典层面，更贯穿于任务相关的内部计算过程，强调了“行为可比性始于词汇可比性”的基本原则，使词汇层面的公平性成为可度量、可干预的科学问题。
链接: https://arxiv.org/abs/2609.34065
作者: Mir Tafseer Nayeem,Davood Rafiei
机构: University of Alberta(阿尔伯塔大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Emerging Technologies (cs.ET); Machine Learning (cs.LG)
备注: Preprint
Abstract:Names are personal identifiers, but they also carry social meaning and are widely used to evaluate how language models treat different people. Such evaluations typically assume that matched names are comparable model inputs. We show that this assumption often fails at the lexical interface: matched names are not necessarily matched inputs. Some names receive direct single-token access, while others are assembled from multiple subwords, creating unequal name-surface support. Across nearly half a million first names and 12 LLM-associated tokenizers, direct lexical access is highly selective, model dependent, and uneven across race- and gender-associated name metadata. We introduce NameTrace, a model-native, fine-grained, pre-behavioral framework for measuring whether unequal name-surface support remains a vocabulary property or becomes visible in task-relevant internal representations. NameTrace measures concept accessibility from the model’s own probabilities over task-specific adjective axes with continuous task-aligned weights. On matched atomic and short-fragmented names within the same race/ethnicity–gender-associated strata, support predicts systematic differences in concept accessibility across fellowship, hiring, clinical assessment, and lending. These differences persist across all eight matched strata, extend across model families, and transfer to unseen names. Hidden-state interventions further show that the measured task directions have downstream leverage, shifting later constrained choices. Unequal lexical support is therefore demographically structured at the input and remains visible in task-relevant model computation. NameTrace makes lexical comparability measurable, supporting a broader principle: behavioral comparability begins with lexical comparability.
[NLP-142] Counterexamples to Local Reconstruction Gain as a Proxy for Final Fidelity in Residual Completion
【速读】： 该论文旨在解决生成式 AI（Generative AI）中稀疏注意力机制在降低计算成本的同时，如何有效保持模型输出保真度的问题。具体而言，研究关注残差补全（Residual Completion）方法是否通过提升单层注意力输出的局部重建精度，能够相应地改善最终模型输出的质量。其解决方案的关键在于对比两种不同的注意力近似策略：一种是训练自由的残差增强稀疏注意力（RESA），另一种是基于冻结主干语言模型的可学习 Top-K+φ 方法。实验结果表明，尽管预设单层筛选策略在局部重建上表现出正向增益，但其在发现请求与提示令牌不重叠的测试请求上反而导致最终输出的KL散度保真度下降，劣于完全放弃计算的精确Top-K基线。而直接在同一层实现精确恢复则能提升最终保真度，说明局部重建优化无法保证全局性能提升，且这种反直觉现象具有特定于近似补全方法的特性。进一步的多层实验显示，任务无关的局部诊断虽可修复部分补全估计器，但修复后的模型并未稳定优于精确计算基线。因此，研究揭示了一个关键结论：更优的局部重建并不必然带来更好的最终模型保真度，强调了在设计高效稀疏注意力机制时需兼顾端到端性能评估的重要性。
链接: https://arxiv.org/abs/2609.34063
作者: Yasuto Hoshi,Daisuke Miyashita,Jun Deguchi
机构: Kioxia Corporation
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Residual completion augments query-aware sparse attention by estimating the contribution of tokens omitted from the exact sparse computation. We ask whether improving a layer’s attention-output reconstruction on the same incoming Q/K/V and selected support necessarily improves the fidelity of the final model output. We study training-free RESA and learned Top-K+ \phi with frozen backbone language models. A prespecified single-layer screen yields two Qwen3-0.6B/Multi-LexSum interventions for which direct-runtime measurements show positive prespecified request-aggregate local reconstruction gain but worse final KL fidelity than the corresponding all-abstain Exact Top-K baseline on both discovery and prompt-token-disjoint holdout requests. Exact restoration at the same layer instead improves final fidelity, showing that the reversal is specific to approximate completion in these cases. In complementary multi-layer experiments, a task-independent local diagnostic often repairs the tested completion estimators, although the repaired models do not consistently outperform Exact Top-K. Together, these results show that better local reconstruction need not translate into better final-model fidelity.
[NLP-143] Steering Language Model Goals with Value Transplant
【速读】： 该论文旨在解决生成式AI模型在推理过程中可能出现的目标偏离问题，即模型虽表现出目标导向行为，但其实际努力方向未必符合用户意图，甚至可能产生非预期的策略（如作弊）。其核心挑战在于如何有效干预模型的内在目标追踪机制，使其行为更契合期望目标。解决方案的关键是提出“价值移植”（value transplant）方法：通过在每个词元生成时，将宿主模型（host model）的激活状态沿候选价值轴进行调整，调整量由捐赠者模型（donor model）与宿主模型在价值坐标上的差异乘以一个大标量决定，从而实现对宿主模型目标轨迹的重定向。实验表明，该方法在不同方向上均有效——诚实型捐赠者可抑制作弊宿主的测试作弊行为，而作弊型捐赠者则能诱导诚实宿主采取更激进的投机策略；此外，在可解码任务中，来自诚实捐赠者的价值移植还能提升作弊宿主在隐蔽测试中的表现。更重要的是，该方法跨模型家族有效，为基于内在价值信号实现模型行为控制提供了初步证据。
链接: https://arxiv.org/abs/2609.34056
作者: Pengcheng Jiang,Fabien Roger
机构: Anthropic Fellows Program(Anthropic研究员项目); Anthropic(Anthropic)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 38 pages, 26 figures
Abstract:Reasoning models often act as if they pursue goals, but their efforts are not always directed toward what users intend, sometimes leading them to pursue unintended outcomes. Previous work has examined how models may internally track their progress toward their goals through a “value axis.” We study whether changing such a signal can retarget the model’s search toward a different goal. We test value transplant: at each token, we shift the host model’s activation along a candidate value axis by the donor-host difference in value coordinates (multiplied by a large scalar), aiming to redirect the host toward the donor’s goal. We study this intervention in Qwen3-8B and GPT-OSS-20B models fine-tuned into honest and cheating variants. We test several candidate value axes, including a self-rating axis constructed from activations preceding high versus low elicited self-ratings of progress. The intervention works in both directions, with an honest donor reducing test-gaming in a cheating host and a cheating donor increasing test-gaming in an honest host, showing that this signal can influence which strategy the model follows. On solvable coding tasks, transplant from an honest donor also improves the cheating host’s hidden-test performance. Value transplant also works across model families, providing preliminary evidence for the intervention in a setting relevant to model control.
[NLP-144] Faithful Activation Verbalization: Reducing Hallucinations in LLM Representation Interpretation
【速读】： 该论文旨在解决生成式语言模型中隐藏表示（hidden representations）激活值语义化方法存在的不完整或幻觉性描述问题，此类问题导致现有激活语义化结果不可靠、难以信任。其解决方案的关键在于提出一种两阶段框架AVPO：首先通过反演器（inverter）从隐藏激活重构源文本，随后利用一个独立的冻结问答模型对重构文本进行评估，从而获得可解释且可检查的中间输出。进一步地，采用直接偏好优化（Direct Preference Optimization, DPO）对反演器进行优化，奖励机制同时兼顾语义可恢复性与词汇保真度。实验表明，在六类文本上，AVPO在整体语义和细节信息恢复方面分别较最强基线提升最高达17.1和9.3个百分点；更重要的是，性能提升主要源于偏好优化而非仅依赖特定重构样本的微调，使得紧凑的跨模型反演器能够超越与原始模型匹配的条件化语义化器，同时在语义恢复与词汇准确性上实现双重改进。此外，分布外案例研究显示，AVPO在保留高层语义的同时，生成的虚假细节更少。
链接: https://arxiv.org/abs/2609.34033
作者: Haiyan Zhao,Zirui Hei,Wei Shi,Huiqi Deng,Na Zou,Mengnan Du
机构: New Jersey Institute of Technology(新泽西理工学院); Shanghai AI Laboratory(上海人工智能实验室); Xi’an JiaoTong University(西安交通大学); Chinese University of Hong Kong, Shenzhen(香港中文大学（深圳)
类目: Computation and Language (cs.CL)
备注: 34 pages, 13 figures, 13 tables
Abstract:Activation verbalization methods such as Activation Oracle and Natural Language Autoencoders decode hidden representations of large language models into human-readable natural language. However, existing methods can produce incomplete or hallucinated descriptions, making their activation verbalizations difficult to trust and use reliably in practice. To this end, we introduce AVPO, a two-stage framework that first reconstructs source text from a hidden activation and then evaluates the resulting text with a separate frozen question-answering model, yielding an explicit and inspectable intermediate readout. We further optimize the inverter with direct preference optimization (DPO), using rewards that capture both semantic recoverability and lexical fidelity. Across six text families, AVPO improves gist- and detail-level information recovery over the strongest baseline by up to 17.1 and 9.3 percentage points, respectively. Crucially, the gains arise from preference optimization rather than fine-tuning on selected reconstructions alone, enabling compact cross-model inverters to surpass donor-matched question-conditioned verbalizers while improving both semantic recoverability and lexical fidelity. Moreover, out-of-distribution case study shows that AVPO better recovers high-level semantics while fabricating fewer details.
[NLP-145] RewardExplainer: Learning Reward Model Explanations from Counterfactual Preference Feedback
【速读】： 该论文旨在解决生成式语言模型后训练阶段中，奖励模型（Reward Model, RM）缺乏可解释性的问题。传统判别式奖励模型仅输出标量得分，难以追溯其评分决策所对应的响应行为，且现有解释方法依赖预定义的高层属性，需对每一对响应进行重复的反事实干预以验证候选解释，缺乏闭环机制来利用奖励模型的反馈持续优化解释器。为此，本文提出RewardExplainer框架，其核心在于通过反事实重写获取目标奖励模型的反馈，并利用该反馈进一步优化解释器，形成闭环训练机制。该框架生成开放式的、原子性的、可干预的自然语言评分机制，使解释更具具体性、可读性和可操作性；同时将反事实反馈转化为偏好监督信号，使解释器能够更忠实捕捉目标奖励模型的评分偏好与敏感行为模式。实验表明，RewardExplainer在多个目标奖励模型和解释器基线中均实现一致性能提升。此外，基于生成的评分机制可识别潜在偏见模式，并构建针对性去偏数据用于奖励模型微调，显著提升了模型在奖励劫持（reward-hacking）基准测试中的鲁棒性。
链接: https://arxiv.org/abs/2609.33989
作者: Jingyi He,Nier Wu,Shuang Liu,Xin Wang,Mengnan Du,Xia Hu
机构: The Chinese University of Hong Kong, Shenzhen (香港中文大学（深圳）); Shanghai Jiao Tong University (上海交通大学); Carnegie Mellon University (卡内基梅隆大学); Jilin University (吉林大学); Shanghai Artificial Intelligence Laboratory (上海人工智能实验室)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Reward models (RMs) are a key component of large language model post-training, providing reward signals for subsequent reinforcement learning. However, conventional discriminative RMs typically output only scalar scores, making it difficult to identify the response behaviors associated with their scoring decisions. Existing interpretation methods often rely on predefined high-level attributes and require repeated counterfactual interventions for each response pair to validate candidate explanations, lacking a closed-loop mechanism that uses RMs’ feedback to train a reusable explainer. To address this, we propose RewardExplainer, a framework that obtains feedback from the target reward model through counterfactual rewriting and uses this feedback to further optimize the explainer. RewardExplainer generates open-ended, atomic, and intervenable natural-language scoring mechanisms, making explanations more concrete, readable, and actionable. It further converts counterfactual feedback into preference supervision, enabling the explainer to more faithfully capture the target RM’s scoring preferences and sensitive behaviors than single-pass generation. Extensive experiments across multiple target RMs and explainer backbones show consistent improvements. Beyond interpretation, we use the generated mechanisms to identify potential bias patterns and construct targeted debiasing data for fine-tuning the reward model, improving robustness on reward-hacking benchmarks.
[NLP-146] Opera: A Verbal Critic Framework for Long-horizon Coding Agents
【速读】： 该论文旨在解决长时序代码生成智能体在执行过程中因反馈不当而导致纠错失效甚至产生负面效果的问题。现有批评者（critic）通常仅评估行为轨迹并生成反馈，却忽视了反馈后智能体的实际响应与问题是否真正解决。其解决方案的关键在于提出Opera——一个以自然语言为媒介的持续性批评框架，将每次修正视为一条持久记录，持续追踪直至诊断出的问题被真正解决。Opera通过周期性与事件触发机制决定审查时机，利用类型化操作符进行问题诊断，在反馈发出前基于可见证据进行审计，并跟踪智能体后续动作以区分表面遵从与实质修复。实验表明，作为推理阶段的批评者，Opera在Terminal-Bench 2.1、SWE-Bench Pro子集和DeepSWE v1.1三个基准上分别将非批评者智能体的解决率提升12.4、15.0和8.9个百分点，且在所有基准上均优于竞争性批评基线；同时在自评模式下亦能有效提升策略模型性能。此外，由Opera引导的滚动推演可生成近似在线训练数据，对Qwen3.5-9B进行微调后，其在保留测试集上的解决率提升10.2个百分点，且无需推理时使用批评者，同时保持跨工具链（如从Openhands切换至Terminus-2）下的性能稳定性，显著优于直接使用更强模型生成的数据微调方案。
链接: https://arxiv.org/abs/2609.33987
作者: Kai Mei,Zhiyuan Hu,Yutong Dai,Juntao Tan,Yifan Zhang,Dingjie Song,Dimitris N. Metaxas,Silvio Savarese,Ran Xu,Zeyuan Chen
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Long-horizon coding agents need timely corrections, yet feedback can be ineffective or even harmful when it misjudges ongoing work or fails to address the underlying problem. Existing critics focus on evaluating trajectories and generating feedback, but rarely track what happens after feedback is delivered. We present Opera, a verbal critic framework that treats each correction as a persistent note, followed until the diagnosed problem is resolved. Opera decides when to review through periodic and event-driven triggers, diagnoses issues with typed operators, audits feedback against visible evidence before delivery, and tracks the agent’s subsequent actions to distinguish mere compliance from actual resolution. As a test-time critic, Opera improves the resolve rate of non-critic agents by up to 12.4, 15.0, and 8.9 percentage points on Terminal-Bench 2.1, a SWE-Bench Pro subset, and DeepSWE v1.1, respectively, across four policy models, and achieves the highest mean resolve rate among competitive critic baselines on all three benchmarks, and also improves policy models when the policy critiques itself. Beyond inference, Opera-guided rollouts provide approximately on-policy training data: fine-tuning Qwen3.5-9B on them improves its resolve rate on held-out SWE-Bench Pro repositories by 10.2 percentage points without a critic at inference time, matching fine-tuning on rollouts from a stronger model, while preserving its performance when switching harness, i.e., from Openhands to Terminus-2, which the latter substantially degrades.
[NLP-147] Beyond Solo and Consistency: Vindicating Multi-Agent Debate via Conditional Progressive Pruning
【速读】： 该论文旨在解决当前基于大语言模型（Large Language Model, LLM）的多智能体辩论（Multi-Agent Debate, MAD）框架在严格计算成本限制下，难以超越单智能体（Single Agent）及一致性增强（Consistency-based）基线方法的核心问题。现有MAD框架虽通过多轮交互实现知识与推理互补，但在同等成本约束下表现不佳，削弱了其有效性。本文提出条件渐进剪枝（Conditional Progressive Pruning, CPP）这一轻量级剪枝框架，其关键在于充分利用多轮辩论中智能体间的信息演化过程，动态识别并保留高价值推理路径，同时逐步剪除冗余或低效的智能体分支。该方法在多个基准测试中显著优于现有MAD框架，并首次在相同成本条件下全面超越一致性增强方法，为MAD技术的实用化提供了新范式。
链接: https://arxiv.org/abs/2609.33974
作者: Ruosong Ye,Caiqi Zhang,Jiahao Li,Haijun Wu,Xiaolong Luo,Huiyuan Chen,Yu Wang,Ying Chen,Zhenting Wang,Kai Mei,Yang Zhou,Dimitris N. Metaxas
机构: Rutgers University, New Brunswick(罗格斯大学, 新布朗兹维克校区); University of Cambridge(剑桥大学); Tsinghua University(清华大学); Harvard University(哈佛大学); Case Western Reserve University(凯斯西储大学); University of California San Diego(加州大学圣地亚哥分校); Carnegie Mellon University(卡内基梅隆大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large Language Model (LLM) based Multi-Agent Debate (MAD) is one of the most effective test time scaling techniques. Through multi-round communication, agents complement each other in knowledge and reasoning and solve tasks that no single member can solve. However, existing MAD frameworks fail to beat strong Single Agent and Consistency-based baselines under the same strict cost limit, which shakes the foundation of the MAD field. We propose Conditional Progressive Pruning (CPP), a lightweight pruning framework that fully exploits multi-round MAD. CPP outperforms all existing MAD frameworks on multiple dominated benchmarks. It is also the first to fully outperform consistency methods. Our code, detailed agent interaction records will be released soon.
[NLP-148] Do System One Decisions Add Up? A Study of Probabilistic Coherence
【速读】： 该论文旨在解决生成式决策模型在概率一致性方面存在的内在矛盾问题，即模型对同一决策任务在不同粒度层级上给出的概率分布不一致，导致整体推理过程出现自相矛盾的现象。其核心问题是：当将一个分类任务从直接预测细粒度标签（fine-label）转为通过宽类别（broad-category）逐步重构时，模型的概率输出和最终预测结果可能出现显著偏差，从而影响系统可靠性。解决方案的关键在于引入联合评估框架，同时考量模型的准确性、置信度校准能力以及概率一致性（probability coherence），并在实际应用流程中对这些维度进行综合评估。研究基于Jev与Laya两个模型，在TREC、CLINC150、MASSIVE三个数据集上共分析72,000个分类问题，发现两者在宽类别层级上的总变差（total variation）分别高达0.219–0.349（Jev）与0.424–0.689（Laya），表明存在严重概率不一致。进一步分析显示，尽管重构过程使Laya的准确率提升21.3个百分点，但其置信度校准误差上升；而Jev则因重构导致准确率下降22.9个百分点。这揭示了仅关注准确率或置信度的单一评估指标无法全面反映模型行为，必须在决策工作流中同步评估准确性、置信度校准与概率一致性，以确保系统在实际部署中的稳健性与可信性。
链接: https://arxiv.org/abs/2609.33971
作者: Saman Sarker Joy
机构: Universiti Malaya(马来西亚大学); Kuala Lumpur(吉隆坡)
类目: Computation and Language (cs.CL)
备注: 20 pages, 4 figures. Code and experimental results: this https URL
Abstract:A decision model can give probabilities that sum to one for every question yet disagree with itself when the same decision is broken into smaller steps. We study this form of probabilistic coherence in Jev and the English Laya checkpoint, using 2,500 matched examples per system across TREC, CLINC150, and MASSIVE. Across 72,000 classification questions, we compare direct fine-label predictions with broad-category probabilities and predictions reconstructed through those categories. Both systems show substantial disagreement: mean category-level total variation ranges from 0.219 to 0.349 for Jev and from 0.424 to 0.689 for Laya, on a scale where zero means exact agreement. The consequences differ sharply. On CLINC150, reconstruction reduces Jev’s accuracy by 22.9 percentage points (paired 95% bootstrap interval: [-24.9, -20.9]) and improves Laya’s by 21.3 points ([18.0, 24.5]). The same directions hold across all three datasets, with all six unadjusted accuracy-change intervals excluding zero. Improved accuracy can also accompany less reliable confidence: on MASSIVE, Laya gains 9.2 accuracy points while its expected calibration error rises from 0.046 to 0.124. Error analysis identifies both broad-category mistakes and within-category confusions. These findings show why decision systems need joint evaluation of accuracy, confidence calibration, and probability coherence in the workflow used by an application.
[NLP-149] On the Token Value Inequality in Efficient Reasoning NEURIPS2026
【速读】： 该论文旨在解决生成式 AI 在复杂任务中依赖链式思维（Chain-of-Thought, CoT）推理时导致的显著增加的令牌（token）消耗问题。其核心问题是：在CoT推理序列中，是否每个令牌的价值均等？研究发现，令牌价值高度非均匀，可通过令牌级别的对数概率信号有效表征。基于此，论文提出TokenProbe框架，关键在于利用归一化对数概率识别核心令牌（承载结构化与决定性推理内容）与冗余令牌（探索性、低置信度填充内容），并据此构建一种高效的渐进式奖励策略优化（GRPO）目标，主张选择性压缩冗余令牌可在不损失推理质量的前提下实现准确率与令牌效率的帕累托改进。实验表明，该方法在保持推理质量的同时，将令牌使用量降低76%，且在相同推理长度预算下，性能优于Gemini-3.1-Pro等主流基线模型。
链接: https://arxiv.org/abs/2609.33970
作者: Runjia Zeng,Hang Hua,Yiyang Liu,Zhiqiang Tao,Ruixiang Tang,Qifan Wang,Cheng Han,Dongfang Liu
机构: Purdue University(普渡大学); Rochester Institute of Technology(罗切斯特理工学院); MIT-IBM Watson AI Lab; University of Missouri-Kansas City(密苏里大学堪萨斯城分校); Rutgers University(罗格斯大学); Meta AI(Meta人工智能)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: NeurIPS 2026
Abstract:Chain-of-Thought reasoning has enabled large language models to achieve substantial performance gains on complex tasks. However, these gains come at the cost of dramatically increased token consumption. This raises a fundamental question: is every token in the reasoning trace equally valuable? We present a diagnostic and optimization framework grounded in a key empirical finding: the value of tokens within a CoT reasoning sequence is highly non-uniform, and this non-uniformity can be effectively characterized by token-level log probability signals. We show that normalized log probability helps distinguish core tokens, which carry structural and decisive reasoning content, from redundant tokens, which are exploratory, low-confidence filler that contributes less directly to the final answer. Building on these findings, we formulate the TokenProbe framework around two empirical findings and one claim: findings identify token value inequality first and then establish TokenProbe as a core-token proxy, and the claim introduces an efficient GRPO objective positing that selectively compressing redundant tokens can yield Pareto improvements in the accuracy-token efficiency space. Empirically, our method preserves reasoning quality while reducing the token usage by 76% of the baseline. Under matched reasoning-length budgets, we show that it can even outperform strong flagship baselines like Gemini-3.1-Pro. Homepage: this https URL.
[NLP-150] Simple Diffusion Language Models Are More Effective Few-Step Generators Than Reported
【速读】： 该论文旨在解决生成式语言模型在少步生成（few-step generation）中质量下降的问题，尤其针对扩散语言模型（Diffusion Language Models, DLMs）虽具备并行生成优势但高质量输出仍需大量迭代步数的瓶颈。其核心发现是：传统上认为少步生成质量差的“性能差距”很大程度上源于采样器配置不当，而非模型本身局限。解决方案的关键在于对采样过程进行“温和的采样器锐化”（modest sampler sharpening），无需重新训练模型即可显著提升生成质量。研究证明，经过优化采样的旧版掩码式DLM在仅16步内即可达到标准采样器1024步的生成困惑度水平，同时提升人类评估的质量与语义多样性。此外，论文指出传统基于单个输出的评价指标（如困惑度）会掩盖真实性能提升，因其最优权衡可能仅由最多两个输出构成，无法反映多样性和整体生成能力。为此，作者提出GroupEval，分别评估生成质量与跨输出语义多样性，揭示出蒸馏模型虽有1.5–4.7倍困惑度降低，却未带来相应质量提升的现象。最后，论文从理论上解释了锐化有效的原因：并行去噪破坏了同时生成词元间的依赖关系，导致预测与实际生成之间出现偏差；即使对于精确的去噪器，常规温度设置为1也普遍次优，且更差的预测可能反而生成更优样本。因此，论文主张应以部署时的实际生成器为评估对象，采用调优后的基线进行对比，并联合使用更贴近人类判断的指标来综合评估生成质量和多样性。
链接: https://arxiv.org/abs/2609.33947
作者: Hasan Amin,Ming Yin,Rajiv Khanna
机构: Purdue University (普渡大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Diffusion language models (DLMs) promise fast parallel generation, yet high-quality samples often require large number of refinement steps, which diminishes their advantage in practice. This has led to massive interest in and rapid development of new methods for effective few-step generation. We show that much of the supposed quality gap at few steps can instead arise from a suboptimally configured sampler. Modest sampler sharpening, without any model retraining, enables a couple years old masked DLM to rival supposedly far improved successors. This differently sampled DLM in fact achieves lower generative perplexity in just 16 steps than what its standard sampler obtains with 1024, while improving both judged quality and semantic diversity. We further show that conventional per-output metrics can fundamentally obscure these gains, since any optimal trade-off between two such metrics can be attained by a generator supported on at most two outputs. We subsequently introduce GroupEval, which separately evaluates quality and across-output semantic diversity, and offers fresh insights including uncovering how 1.5-4.7x perplexity gains of a distilled model yield no corresponding quality gain. Finally, we explain why sharpening helps: parallel unmasking destroys dependencies among simultaneously generated tokens, creating a gap between prediction and generation. We prove that pervasive temperature choice of one is generically suboptimal under parallel sampling even for an exact denoiser, and that worse predictions can yield better samples. Through these results, we argue for a broader evaluation principle of treating the deployed generator as the object of comparison, benchmarking it against tuned baselines, and assessing quality and diversity jointly and with more human-aligned measures.
[NLP-151] Quantization Error Is Spectrally Flat: A Single Random Probe Is a Calibrated Data-Free Sensitivity Estimator with Application to Budget-Targeted Mixed-Precision Quantization
【速读】： 该论文旨在解决大模型量化过程中混合精度量化（mixed-precision quantization）的比特分配难题，即如何在有限的存储预算下，为不同张量动态分配最优的位宽以最小化量化误差。传统方法依赖校准数据进行逐层敏感性评估，计算开销大且难以扩展。其核心解决方案是提出一种基于随机高斯探针（Gaussian probe）的无校准敏感性估计器，该方法利用单个随机高斯向量即可无偏估计层级量化误差的平方Frobenius范数，其有效性源于舍入到最近值（round-to-nearest）产生的谱平坦（spectrally flat）误差特性。关键创新在于引入传播形式（propagated form）的探针估计，该形式能捕捉张量间在前向传播中的敏感性耦合关系，显著提升与真实激活下优化目标（如GPTQ层目标）的相关性（Spearman相关系数达0.81–0.83）。该方法无需校准数据，仅需一次探针采样即可支持任意字节预算，结合背包求解器实现精确的比特分配，并通过防护机制避免灾难性的2比特分配。实验表明，在7种从8B到122B的不同架构上，该方法（RAM）在相同字节预算下相比均匀4比特量化，可实现3.5%至13.6%的中位数WikiText-2困惑度降低，性能优于现有厂商方案和基于真实激活的基准。
链接: https://arxiv.org/abs/2609.33923
作者: I Kennedy,T Kennedy
机构: baa.ai; Auckland, New Zealand
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:A single random Gaussian probe gives an unbiased estimate of the squared Frobenius norm of a layer’s quantization error. The estimator is well-behaved because round-to-nearest error is spectrally flat. Across 1,683 tensors from a 35B MoE and a 9B dense model, effective dimensionality is 0.93 to 0.96 times the i.i.d. noise value of the same shape, and on the MoE the median is unchanged from 2-bit to 8-bit. The probe coefficient of variation is predictable from tensor shape. One probe measures per-tensor sensitivity to within 4 to 7%; twenty probes reach 1.3 to 1.4%.RAM applies the propagated form of this estimator to budget-targeted mixed-precision quantization with no calibration data. Gaussian probes carrying the network’s own input statistics score every tensor at six bit-widths. A knapsack solver allocates bits under an exact byte budget, with guardrails against catastrophic 2-bit assignments. One probe pass serves any budget. Isolated and propagated scores rank tensors independently on Qwen3.5-35B-A3B (Spearman -0.01), yet the propagated probe rank-correlates 0.81 to 0.83 with the GPTQ layer objective from real activations, while the isolated estimator is uncorrelated with it. That objective is the wrong allocation target: at matched bytes on Qwen3.8-27B, a block-output probe beats a vendor IQ3_M mix and an oracle that allocates from the real-activation this http URL Qwen3-8B the propagated probe ties HAWQ-V2 at matched bytes. Across seven architectures from 8B to 122B, with probe timing up to a 400B model in nine minutes on one workstation, RAM reaches 3.5 to 13.6% lower median WikiText-2 perplexity than size-comparable uniform 4-bit builds on the tested MoE models
[NLP-152] SlopBench: How Well Can We Rank Language Models by Slop? A Multi-Domain Benchmark of Repetitive AI Writing
【速读】： 该论文旨在解决生成式文本中“AI slop”（即机械、重复且缺乏自然流畅性的呆板文风）的识别与量化问题，尤其针对现有检测工具仅能判断文本是否为机器生成，却无法评估其语言质量高低这一局限。其解决方案的关键在于构建SlopBench基准测试体系，通过人工可检验的四种表面行为特征对模型输出进行系统性评估：在特定任务中输出长度是否偏离预设词数区间、同一模型多次生成同一任务时是否存在开头重复现象、段落节奏模式是否偏离前ChatGPT时期人类语料库中的自然分布，以及是否存在固定的词汇结构。基于这些指标，研究对18个大模型在112项手写任务（涵盖邮件、社交帖子、论文及职场聊天）上共生成的19,928条输出进行了评估，发现不同权重组合下模型排名高度不稳定，仅有少数排名具有统计显著性。因此，研究主张不再依赖单一综合评分，而是分别报告四项行为指标，并将复合得分视为众多可能权重之一，强调透明性，公开所有提示、输出、参考统计数据及评分代码，以支持后续可复现的评估研究。
链接: https://arxiv.org/abs/2609.33905
作者: Dhruv Roongta,Harsha Gaddipati,Anh Tuan Huynh
机构: Slashy, San Francisco, California, USA
类目: Computation and Language (cs.CL)
备注: 12 pages, 4 figures
Abstract:SlopBench asks which models produce the stiff, repetitive prose readers call AI slop, a question detectors leave open once they have classified a text as machine-written. We evaluated eighteen models on 112 hand-written tasks in email, social posts, essays, and workplace chat, sampling each model on each task up to ten times, for 19,928 outputs in all. SlopBench scores four surface behaviors a reader can check by hand: length against the word band each task specifies, opener repetition across a model’s own samples of one task, and paragraph rhythm and fixed lexical constructions against pre-ChatGPT human corpora. Under one fixed weighting, Kimi K2.6 scores lowest at 21.1 and Mistral Large highest at 40.6. Across 500 random reweightings Kimi has the lowest score in 58 percent of draws and Mistral the highest in 97 percent. No draw preserves the full order of the eighteen, and a scenario bootstrap leaves exactly one of those ranks unambiguous. We ran three further checks on that middle order: a crowd arena, an AI detector, and lexical diversity. None of them confirmed the order. We therefore report the four behaviors separately and treat the composite as one weighting among many, and we release the prompts, outputs, reference statistics, and scoring code.
[NLP-153] Where Activation Sparsity and KV-Cache Sparsity Cross in LLM Decoding
【速读】： 该论文旨在解决大语言模型在生成式推理过程中因上下文长度增加而导致的内存访问开销问题，特别是投影权重（projection weights）与键值（Key-Value, KV）缓存访问流量随上下文增长带来的性能瓶颈。其核心挑战在于：尽管激活稀疏性（activation sparsity）和KV缓存稀疏性（KV-cache sparsity）均能带来加速，但两者在不同上下文长度下的优势各异，且现有评估结果难以直接比较，因其依赖于具体上下文长度和所采用的密集注意力内核。为此，作者提出通过模型维度和保留率（keep ratio）推导出“字节交叉点”（byte crossover），即两种稀疏策略收益相等的上下文长度，并建立各自及组合后的理想加速上限。实验基于两块GPU，在真实文本预填充后，从2K到128K token范围内对两种稀疏分支及其组合进行实测，使用相同的split-K注意力内核读取缓存。结果显示，短上下文下投影分支占优，长上下文下KV分支占优，实际加速表现接近理论字节边界，仅受固定内核开销影响。引入独立测量的内核开销后，该字节模型可将三种保留率组合、另一模型及另一GPU上的实际交叉点预测误差控制在4.1K token以内。此外，使用掩码注意力替代split-K注意力会使得相同KV策略的加速效果被高估约五倍，凸显了内核实现的重要性。最终，结合注意力评分选择的KV稀疏策略在匹配困惑度预算条件下，相比单一最优分支，在两块GPU上实现了14%至26%的加速提升。
链接: https://arxiv.org/abs/2609.33889
作者: Jungseob Lee,Seungyoon Lee,Seongtae Hong,Sugyeong Eo,Heuiseok Lim
机构: Korea University (高丽大学); Yonsei University Mirae Campus (延世大学未来校区)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Performance (cs.PF)
备注: 22 pages, 6 figures, 17 tables
Abstract:At each step, decoding one sequence with a large language model rereads the projection weights, whose traffic is fixed, and the key-value (KV) cache, whose traffic grows with context. Activation sparsity trims the first term and KV-cache sparsity the second, yet their reported speedups are hard to compare because each depends on context length and on the dense attention kernel it is measured against. We derive a byte crossover, the context length at which the two savings are equal, together with ideal speedup bounds for each branch and for their composition, from model dimensions and keep ratios alone. We then time both branches and their composition from 2K to 128K tokens on two GPUs after a dense prefill of real text, with dense and sparse modes reading the cache through the same split-K attention kernel. The projection branch leads at short context and the KV branch at long context, with speedups that follow their byte bounds up to fixed kernel costs. Adding these costs, measured in separate sweeps, lets the byte account predict the measured crossings of three keep-ratio pairs, a second model, and a second GPU to within 4.1K tokens. Timing the dense baseline with masked instead of split-K attention inflates the apparent speedup of the same KV policy about fivefold. An attention-scored KV selection answers the same passkey and multi-key placements as dense decoding up to 127K tokens, whereas a KV window misses most of them. Under matched perplexity budgets, activation sparsity composed with this selection decodes 14 to 26% faster than the best single branch on both GPUs. Code is available at this https URL.
[NLP-154] Faster Block-Diffusion Serving with Distribution-Free Risk Guarantees
【速读】： 该论文旨在解决块扩散语言模型（block-diffusion language models）在推理服务中，基于均值准确率（mean benchmark accuracy）选择运行配置时所导致的可靠性风险问题。现有方法仅关注平均性能，忽略了快速配置在部分输入上失败而慢速配置仍能正确回答的情况，从而可能引入不可控的错误风险。其解决方案的关键在于提出Redline——一种基于有限样本的配置选择机制，通过在校准提示（calibration prompts）上评估候选配置的正确性，确保参考答案正确但候选配置错误的联合概率（即参考相对风险，reference-relative risk）在用户设定的风险预算内以高概率成立。Redline优先部署满足该风险约束下最快的配置，显著提升了推理速度。实验表明，在相同10%风险预算下，该方法在数学任务上的加速效果优于代码任务，且可直接应用于推测解码的接受规则与权重量化等场景。在相同校准数据集上，Redline始终维持在预设失败概率范围内，而基于均值准确率的阈值策略则或提速不足，或频繁超出风险预算，表现出较差的稳定性与可预测性。
链接: https://arxiv.org/abs/2609.33887
作者: Jungseob Lee,Dongyub Jude Lee,Chanjun Park,Sugyeong Eo,Heuiseok Lim
机构: Korea University (高丽大学); Zoom Communications (Zoom通信公司); Soongsil University (顺溪大学); Yonsei University Mirae Campus (延世大学未来校区)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 31 pages, 7 figures, 18 tables
Abstract:Block-diffusion language models are served at hand-picked operating points, such as acceptance thresholds, buffer depth, schedule, checkpoint and precision, and each point is chosen by its mean benchmark accuracy. However, a mean does not tell an operator how often a faster configuration fails on prompts that the slower one answers correctly. On the serving engine and its decode traces, the default commit rule already commits every fully resolved block, a static skip rule captures nearly all of the compute that allocation can save, and self-distillation on engine-decoded targets adds speed at unchanged accuracy. Larger speedups come from lower thresholds, which commit tokens that are still uncertain. We therefore present Redline, a finite-sample procedure that selects operating points, hand-picked or learned, from the correctness of their answers on calibration prompts. Redline keeps the reference-relative risk, the joint probability that the reference answers correctly and a candidate configuration does not, within a user-chosen budget with high probability, and deploys the fastest configuration that passes. It speeds up math at a smaller risk budget than code in both model families, and at a budget of ten percent it deploys a LLaDA2 math configuration that commits over a third more tokens in each forward. It also applies without modification to the acceptance rule of speculative decoding and to weight quantization. On the same calibration data, Redline stays within its stated failure probability, whereas each tolerance of a mean-accuracy rule either gains less speed for some model and task or exceeds the risk budget far more often for another. Code is available at this https URL.
[NLP-155] LLM s learn different forms of metacognition when trained to predict their own accuracy
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在缺乏相关知识时仍强行生成答案而导致事实虚构的问题，核心在于探究模型如何通过训练获得元认知监控能力（metacognitive monitoring），即“知道自己知道什么”的能力。其解决方案的关键在于对10个开源权重的LLM进行微调，使其在回答事实类多选题前预测自身准确率。研究发现，经过训练后的置信度主要反映两种不同信号：在接近训练数据分布的问题上，置信度能有效反映模型的真实准确率；而在其他领域，置信度则更多地追踪输出一致性（output consistency），即模型答案分布的集中程度。其中，输出一致性跟踪在训练初期即出现且具备跨数据集泛化能力，而真实准确率跟踪则在后期才发展，并局限于训练数据分布。这一发现表明，校准训练可能并未教会模型普遍识别高自信下的错误，从而对人工系统中元认知的本质提出了新的质疑。
链接: https://arxiv.org/abs/2609.33886
作者: Nicolas Yax,Stefano Palminteri,Pierre-Yves Oudeyer
机构: LNC2, INSERM, Paris, France; DEC, ENS, PSL, Paris, France; Flowers AI CogSci Lab; Centre Inria de l’Université de Bordeaux France
类目: Computation and Language (cs.CL)
备注:
Abstract:Large language models are trained to always produce an answer, regardless of whether they possess the relevant knowledge, which leads them to fabricate facts. Prior work has shown that LLMs’ confidence estimates correspond poorly to their actual performance, and that fine-tuning can substantially improve them. However, what models actually learn during such training remains poorly understood. We investigate how LLMs acquire metacognitive monitoring, the ability to know what one knows, by training 10 open-weight LLMs to predict their own accuracy on factual multiple-choice questions before answering them. We find that trained confidence reflects two distinct signals. While on questions close to the training data, it tracks the model’s true accuracy, in other domains, it instead tracks output consistency: the concentration of the model’s answer distribution. Output consistency tracking emerges early in training and generalizes across datasets, whereas accuracy tracking develops later and remains local to the training distribution. These results suggest that calibration training may not teach models to generally detect errors they commit confidently, and they raise broader questions about the nature of metacognition in artificial systems.
[NLP-156] Lost with a Map: Conversational State and Behavioral Reliability in Language Models
【速读】： 该论文旨在解决任务导向对话中语言模型缺乏显式信念状态（belief state）表示的问题，即如何在多轮对话中有效维护、更新并利用对话状态信息。其核心挑战在于：尽管模型能够隐式地保留对话历史，但其内部状态的结构化表征与具体数值的可读性存在显著差异。研究发现，对话的结构信息（如活跃领域、槽位及请求）可在模型决策前线性读取，而具体值则更依赖于用户首次提及的位置；当用户更改某值后，旧值与新值均保留在原提及位置且持续影响模型行为。在自然闭环交互中，查询失败可归因于三类问题：结构支持不足、值解析错误以及虽有支持却未执行约束。针对这些情况，作者提出一种基于基础模型输出的“状态-动作控制器”（state-action controller），通过仅利用结构读出结果对基础模型生成的动作进行选择性编辑，无需构建完整的预测信念状态作为中间表示。该方法在保留低计算开销的前提下，在独立测试集上将多轮对话的精确查询准确率从0.318提升至0.621，任务成功率从0.272提升至0.371。研究表明，可靠对话不仅需要维持对话信息，还需准确识别当前适用的约束条件，并确保其有效指导行动。
链接: https://arxiv.org/abs/2609.33883
作者: Atahan Dokme,Larry Heck
机构: Georgia Institute of Technology(佐治亚理工学院)
类目: Computation and Language (cs.CL)
备注: 9 pages main text, 39 pages total, 3 main-text figures
Abstract:Task-oriented dialogue requires maintaining and updating information across turns, yet language models expose no explicit belief-state object. We study how conversational state is represented, updated, and used inside eight instruction-tuned language models from four families on MultiWOZ and SGD. Structure and values separate: which domains, slots, and requests are active is linearly readable just before the model acts, whereas exact values are far more readable where the user stated them. After a user changes a value, both values remain accessible at their mentions, and causal interventions show that both continue to influence the model’s action. In natural closed-loop interaction, query failures separate into cases of weak structural support, incorrect value resolution, and failure to deploy otherwise-supported constraints, with targeted interventions producing systematically different repair behavior across these cases. These findings motivate a state-action controller that starts from the base model action and selectively edits it using structural readouts, without requiring a complete predicted belief state as an intermediate representation. On held-out MultiWOZ interaction across five models, it raises the base model exact-query accuracy from .318 to .621 and task success from .272 to .371 at negligible added cost. Overall, reliable interaction requires not only retaining conversational information, but resolving which available constraints currently apply and ensuring that they govern action.
[NLP-157] In-Context Adaptation of Encoder-Decoder Models in Speech Recognition
【速读】： 该论文旨在解决自动语音识别（ASR）模型在面对新说话人、口音及领域时，如何通过推理阶段的上下文学习（in-context learning）实现快速适应的问题。其核心挑战在于验证上下文学习是否为编码器-解码器架构普遍具备的能力，而非依赖特定模型结构或训练方式。解决方案的关键在于系统性地比较两种演示形式——拼接式（collated）与交错式（interleaved）演示——在六种不同类型的编码器-解码器模型上的表现，涵盖传统基于交叉注意力机制和大语言模型（LLM-based）架构。研究发现，所有测试模型均无需额外微调即可实现即插即用的上下文适应，在理想条件下相对性能提升达30%，使用首次解码结果时仍可实现23%的提升。进一步控制实验表明，词汇信息与说话人信息均对成功适应有贡献；尽管交错式演示在部分场景有效，但拼接式演示在各类设置下表现出更稳定且一致的适应能力。结果表明，上下文学习用于ASR并非特定架构或训练策略的专属特性，而是一种广泛存在于编码器-解码器模型中的固有潜力。
链接: https://arxiv.org/abs/2609.33865
作者: Yen Meng,Sharon Goldwater,Hao Tang
机构: The Centre for Speech Technology Research, University of Edinburgh (爱丁堡大学语音技术研究中心)
类目: Computation and Language (cs.CL); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注: Accepted to IEEE SLT 2026
Abstract:In-context learning offers an appealing approach to adapt automatic speech recognition (ASR) models to new speakers, accents, and domains by providing speech-text pairs as demonstrations at inference time. Recent work shows that some LLM-based speech models are capable of ASR in-context adaptation, when providing interleaved speech-text demonstrations. In this work, we ask whether in-context adaptation is an inherent ability for all encoder-decoder models. We study two forms of demonstration, collated and interleaved demonstration, across six encoder-decoder models, spanning conventional cross-attention-based and LLM-based architectures. We find that all tested models are able to perform in-context adaptation out of the box, achieving up to 30% relative improvement in the oracle experiments and up to 23% using first-pass hypotheses. Through controlled experiments on three English datasets, we show that lexical and speaker information both contribute to successful adaptation. While interleaved demonstration is effective in certain cases, collated demonstration brings consistent adaptation across the board. Our results suggest that in-context adaptation for ASR is not unique to specific architectures, training, or demonstration approaches.
[NLP-158] Program-Verified Self-Evolution for Vision-Language Models
【速读】： 该论文旨在解决自演化视觉语言模型（self-evolving vision-language models）在无标注图像上生成问题时，因缺乏真实答案而导致的标签错误问题。现有方法依赖多数投票或模型判官对生成答案进行打标，但人类评估显示其错误率分别高达24%和18%。针对此问题，论文提出可验证问答生成方法（Verifiable QA Generation for Self-Evolving Models, VQS），其核心创新在于重构答案判断机制：不再直接对答案进行投票或评分，而是将图像解析为结构化记录（如场景图、表格或图谱），由固定程序基于该记录生成问题并计算答案；模型仅作为视觉检查器，逐项验证程序所读取的单一事实陈述（claim-level check）。这种设计使解析器通过无监督的反馈机制持续优化，无需人工标注。实验表明，VQS生成的答案有94%被人类评价为正确，显著优于多数投票方法（76%）。在十项基准测试中，VQS将Qwen3-VL模型性能提升最高达3.18点，并在各规模（2B、4B、8B）下均超越最强自演化基线，且随着训练轮次增加，性能增益持续累积，三轮后达到3.84点的提升。相关代码已开源。
链接: https://arxiv.org/abs/2609.33855
作者: Ahmed Heakl,Sungik Choi,Moontae Lee,Salman Khan
机构: LG AI Research( LG人工智能研究院); MBZUAI( Mohamed Bin Zayed University of Artificial Intelligence); Australian National University (澳大利亚国立大学); University of Illinois at Chicago (伊利诺伊大学芝加哥分校)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
备注: 26 pages
Abstract:Self-evolving vision-language models train on questions they generate from unlabeled images. Since these questions have no gold answers, prior methods label them by majority vote over sampled answers or by a model judge. In a human evaluation, we find that 24% of majority-vote labels and 18% of model-judge labels produced during self-evolution are wrong. To address this problem, we present Verifiable QA Generation for Self-Evolving Models (VQS), which changes how the model judges answers. Instead of voting on an answer, the model parses each image into a structured record, such as a scene graph, a chart table, or a diagram graph. Fixed programs then write a question from the record and compute its answer. The model still acts as a visual checker, but it only confirms the individual facts the program reads, one short claim at a time. These claim-level checks select the parser’s training targets, so the parser also improves without labels. Human raters find 94% of VQS answers correct, against 76% for majority voting. Across ten benchmarks, VQS improves Qwen3-VL by up to 3.18 points at the 2B, 4B, and 8B scales and outperforms the strongest self-evolving baseline at each. Gains keep growing over three training rounds, reaching 3.84 points at 2B. Code is released at this https URL
[NLP-159] Rethinking Contextualization by Reinterpreting Attention Head Channels
【速读】： 该论文旨在解决语言建模中上下文表示（contextualization）机制的内在行为理解不足问题，特别是现有研究多聚焦于单个词或注意力头的局部分析，缺乏对整体信息传递规律的系统性认知。其核心解决方案在于提出一个全局性原则：不同词语携带的信息量存在差异，低信息量词语倾向于吸收更多上下文信息；而这种吸收并非均匀分布，更细粒度的选择性机制能够实现更精准的信息路由，从而促进语义匹配词语间的信息高效传递。关键创新在于将注意力头重新诠释为由奇异向量控制的通道——这些奇异向量指向更具信息量的词的隐藏状态，使高信息量词能更强地“书写”自身信息作为信息源，反之亦然；同时，奇异向量可被视为隐藏状态的特征表示，实现了注意力头的连续空间嵌入，突破了传统将其视为离散、独立字典条目的局限，支持自动化解释与语义关联分析。
链接: https://arxiv.org/abs/2609.33851
作者: Hakaze Cho,Haolin Yang,Zhun Sun,Naoya Inoue,Benjamin Heinzerling,Kentaro Inui
机构: RIKEN(理化学研究所); Tohoku University(东北大学); New York University(纽约大学); JAIST(日本信息科学与技术研究院); MBZUAI(穆巴达拉人工智能研究院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 47 pages, 81 figures, 4 tables
Abstract:Contextualization, the core operation of language modeling, transmits information across words to build sentence-specific word representations. Prior works mainly study contextualization, focusing on individual words and attention heads as a growing discrete dictionary, lacking a global view of their general behavior. Therefore, we propose a general principle: Globally, we find and estimate that different words carry different amounts of information, and less-informative words tend to absorb more contextual information. Specifically, these low-information words do not absorb contextual words uniformly, and finer-grained selectivity enables more precise routing to promote information transmission between matched words. Moreover, to find what mechanism causes such processing, we reinterpret attention heads as channels gated by their singular vectors and find that: (1) these singular vectors point to the hidden states of more informative words, allowing such words to write their information to others more strongly to act as information sources, and vice versa; and (2) these singular vectors can be viewed equally as hidden state features, enabling automated interpretation of attention heads beyond prior heuristic head discovery, also embedding heads into a continuous space rather than treating them as discrete, independent dictionary entries.
[NLP-160] Laya as a Typed Probabilistic Assessor: An Independent Reproduction and a Preregistered Study of Calibration and Selective Escalation
【速读】： 该论文旨在解决生成式模型在决策任务中普遍存在的置信度校准问题，具体表现为预发布版本的Laya Typed-Decisions检查点（一个421M参数的ModernBERT-large评估器）在处理类型化选择、名词及评分问题时表现出系统性低置信度（under-confidence），其置信度-准确率差距（confidence-accuracy gap）为-0.214，且所有可靠性分箱的准确率均高于置信度，导致各类期望校准误差（ECE）变体统一收敛至0.214。这一现象表明模型存在严重低估置信度的问题，而传统风险评估常误判为过自信（over-confidence），但实际方向决定了置信度门控级联结构的失败模式。解决方案的关键在于采用单一独立拟合的温度缩放（temperature scaling，T=0.469，即锐化操作），显著缓解了模型的非校准性（从保留测试集上的ECE 0.204降至0.037），并优于原生的按选项计数校准表。相比之下，冻结的选型规则采用等倾回归（isotonic regression）导致过拟合，在保留测试上未通过负对数似然（NLL）对比检验，因此不支持假设H2。此外，研究通过前瞻性预注册实验（E2-E8）验证了结果的可重复性，22项确认性测试中有20项在Benjamini-Hochberg FDR控制下（q=0.05）被拒绝，说明结果具有统计稳健性。尽管冻结门控策略优于随机提升，但仍未能达到10%接受集错误率的目标；探索性分布外检测显示无零样本迁移能力（准确率0.617），且所有评分指标均显示模型表现优于合成教师模型的自一致性上限（0.735），表明该模型具备超越基准的泛化能力。整个分析流程、每决策预测输出、运行清单及预注册文档均附于补充材料中，作者与模型发布方、数据集发布方及TypeSafe无隶属关系。
链接: https://arxiv.org/abs/2609.33843
作者: Gowthamkumar Nandakishore
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 31 pages, 8 figures. Ancillary files include the frozen preregistration, all run manifests, per-decision prediction records, and the metric code
Abstract:The shipped Laya Typed-Decisions checkpoint, a 421M-parameter ModernBERT-large assessor that answers typed choice/noul/score questions over workflow state, is uniformly under-confident. The signed confidence-accuracy gap is -0.214 , every occupied reliability bin’s accuracy exceeds its confidence, and that sign uniformity collapses every binned ECE variant to the same value, 0.214 . The card frames the risk as over-confidence; the measured direction is the opposite, and the direction decides which way a confidence-gated cascade fails. A single disjointly fitted temperature ( T=0.469 , sharpening) removes most of the miscalibration (held-out ECE 0.204 to 0.037 ) and outperforms the shipped per-option-count table. The frozen selection rule instead chose isotonic regression, which overfit and failed its held-out NLL contrast on both tracks, so hypothesis H2 is not supported. Re-running the released checkpoint on its full official test split reproduces the card’s headline accuracy ( 0.767 vs. 0.766 ). The retrospective E1 reproduction preceded the analysis freeze; E2-E8 were prospectively preregistered, and 20 of 22 executed confirmatory tests reject under Benjamini-Hochberg FDR at q=0.05 (two descoped). The frozen gate beats random escalation but misses its 10% accepted-set error target on both tracks, an exploratory out-of-distribution probe finds no zero-shot transfer (accuracy 0.617 ), and every score measures agreement with a synthetic teacher whose self-agreement ceiling ( 0.735 ) the specialist exceeds. Per-decision predictions, run manifests, and the frozen preregistration are in the ancillary files. The author has no affiliation with the model’s publisher, the dataset’s publisher, or TypeSafe.
[NLP-161] ChemOPD: Multi-Teacher On-Policy Distillation for Multi-Task Chemical Reasoning
【速读】： 该论文旨在解决大语言模型在化学推理任务中如何有效整合多样化能力的问题，核心挑战在于如何组织模型的专业化分工以及如何将专家指导信息合理融入统一模型的训练过程。其解决方案的关键在于提出ChemOPD框架，通过监督微调梯度估计任务间的亲和性，并求解约束型混合整数规划（MIP）以构建部分重叠的专家群体；在蒸馏阶段，保留一个覆盖所有任务的通用教师模型作为基础，使专家教师的指导作为补充而非替代，从而避免知识覆盖缺失。进一步引入锚定-残差目标函数，渐进式增强路由至特定专家的学生输出贡献。实验表明，基于亲和性的专业化设计能带来任务特异性性能提升，并拓展了超越语义任务分组的能力边界；更重要的是，即使教师模型表现更强，若缺乏有效的知识整合机制，学生模型仍难以充分受益——而本方法在相同专家与路由策略下，显著优于仅依赖专家蒸馏的方案，更高效地利用了教师模型的潜力。这一结果揭示了专业化分工与能力融合是化学推理建模中既相互关联又独立的关键设计维度。
链接: https://arxiv.org/abs/2609.33838
作者: Yaoyao Xu,Xinjian Zhao,Xiaozhuang Song,Xuemin Chen,Tianshu Yu
机构: School of Data Science, The Chinese University of Hong Kong, Shenzhen(数据科学学院，香港中文大学（深圳）); Shanghai Artificial Intelligence Laboratory(上海人工智能实验室)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Large language models are increasingly expected to support diverse chemical reasoning capabilities within a unified model. One approach is to develop specialized capabilities separately and consolidate them through multi-teacher on-policy distillation, but this raises two questions: how should specialization be organized, and how should specialist guidance be integrated? We introduce ChemOPD, which addresses both. We estimate task affinities from supervised fine-tuning gradients and solve a constrained mixed-integer program(MIP) to construct partially overlapping specialist groups. During distillation, we retain a generalist teacher trained on all tasks so that specialist guidance supplements rather than replaces its supervision. Our anchor-residual objective gradually increases the routed specialist’s contribution on student-generated responses. On ChemCoTBench, affinity-guided specialization produces task-dependent gains over the generalist teacher and improves several capabilities beyond semantic task grouping. Yet stronger teacher-side performance does not automatically yield stronger students: with the same specialists and routes, anchor-residual OPD improves most reported metrics over specialist-only distillation and realizes a larger share of the available teacher gains. These results highlight specialization and capability integration as connected but distinct design problems in chemical reasoning.
[NLP-162] Controlling Speaking Rate in Autoregressive TTS via Activation Steering
【速读】： 该论文旨在解决自回归文本转语音（Autoregressive Text-to-Speech, TTS）系统在训练完成后难以控制语速的问题。现有方法通常缺乏对合成语音速率的灵活调节能力，而本文提出一种无需重新训练即可在推理阶段实现语速调控的解决方案。其关键在于通过分析解码器块的激活空间，识别出一个与语速相关的方向（rate axis），并利用该方向上的投影值进行固定化钳制（clamping）。具体而言，通过对合成的时长拉伸和压缩语音数据进行分析，可恢复语速轴、中性工作点及每步的强度缩放因子；在推理过程中，将解码器某一层激活向量在该速率轴上的投影设为预设标量，从而实现对语速的精确控制。该方法不仅有效保持了说话人身份的一致性，且具备跨模型架构的泛化能力，在客观与主观评估中均表现出高自然度。相较于传统的加性调控方式，钳制策略在极慢语速下仍保持稳定，适用于所有测试系统；而在中等目标语速下，性能表现则取决于具体模型。此外，研究发现语速信息可在多层中被解码，但因果可控仅限于中间深度范围，验证了该方法在公开的Seed-TTS-Eval基准上的有效性。
链接: https://arxiv.org/abs/2609.33810
作者: Francesco Verdini,Antonis Asonitis,Aref Farhadipour,Marzieh Razavi,Pierre-Edouard Honnet,Vijeta Avijeet,Juan Pablo Zuluaga Gomez
机构: AGIGO; ETH Zurich (苏黎世联邦理工学院); Sapienza University of Rome (罗马第一大学); University of Zurich (苏黎世大学)
类目: ound (cs.SD); Computation and Language (cs.CL); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
备注: Accepted at IEEE SLT 2026. 8 pages, 3 figures, 5 tables
Abstract:Autoregressive text-to-speech (TTS) systems synthesize natural speech but, once trained, offer little control over speaking rate. We show that speaking rate can be steered at inference time, without retraining, by clamping a single decoder block’s activation along a discovered speed axis. A decoder-block analysis recovers the rate axis, a neutral operating point, and a per-step intensity scale; at inference, the activation’s projection onto this axis is set to a fixed scalar. Learning this direction from synthetically time-stretched and time-compressed speech yields rate control that largely preserves speaker identity, generalizes across model architectures, and maintains high naturalness in objective and human evaluations. Unlike standard additive steering, which breaks at the slow extreme, clamping remains stable on all three systems tested; at moderate targets, the better rule depends on the model. Finally, we show that rate information is decodable across layers but causally steerable only within a mid-depth window, and demonstrate the effectiveness of our approach on the public Seed-TTS-Eval benchmark.
[NLP-163] LA-CPD: Local-Evidence-Aware Change-Point Detection for Human-LLM Authorship Segmentation
【速读】： 该论文旨在解决在人类与大语言模型（LLM）共同撰写的文档中，准确识别由LLM生成内容段落的难题，这对于版权侵权、欺诈及其他有害AI内容使用场景下的责任归属与可追溯性至关重要。核心挑战在于：尽管句级作者身份检测器可提供局部证据，但内容差异会导致同一来源句子的得分波动，产生虚假的分界点，且在作者转换次数与位置均未知的情况下，难以恢复连贯的文本划分。其解决方案的关键是提出一种结构化方法——局部证据感知变点检测（LA-CPD），该方法将噪声较大的句级得分序列转化为一致的作者身份段落。LA-CPD通过结合长度加权的段内残差与滑动窗口双均值对比，同时捕捉段落内部一致性与候选切点附近的持续变化特征；利用动态规划优化不同切点数下的最优切割位置，并采用类似AIC的准则选择最终划分方案，从而实现句级标签、作者边界及最大LLM生成片段的联合推断。在独立测试集上，LA-CPD显著优于WCP+AIC基线，将句级准确率从0.747提升至0.796，并进一步改善了边界定位与LLM生成片段的精确界定。
链接: https://arxiv.org/abs/2609.33787
作者: Qing Yang,Zhenyu Mao,Zixiang Luo,Zezheng Wu,Xinghe Cheng,Qinggang Zhang,Jingwei Zhang,Jiapu Wang
机构: Guilin University of Electronic Technology(桂林电子科技大学); Jinan Unaiversity(济南大学); Jilin University(吉林大学); Nanjing University of Science and Technology(南京理工大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:As LLM-generated text becomes increasingly human-like, accurately localizing LLM-authored spans in human-LLM co-authored documents is important for attribution and accountability in cases involving copyright infringement, fraud, and other harmful uses of AI-generated content. Sentence-level detectors provide local authorship evidence, but content variation can cause score fluctuations even among sentences from the same source, creating spurious boundaries. Recovering a coherent document partition therefore remains challenging when both the number and locations of authorship transitions are unknown. We propose Local-Evidence-Aware Change-Point Detection (LA-CPD), a structured method that transforms noisy sentence-level score sequences into coherent authorship segments. Given scores from a frozen local detector, LA-CPD combines a length-weighted within-segment residual with a windowed two-mean contrast to capture segment consistency and sustained changes around candidate cut points. Dynamic programming optimizes cut locations for each candidate count, while an AIC-style criterion selects the final partition, yielding sentence labels, authorship boundaries, and maximal LLM-authored spans. On a held-out human-LLM co-authored test set, LA-CPD outperforms WCP+AIC, increasing sentence-level accuracy from 0.747 to 0.796 while improving boundary localization and LLM-span delineation.
[NLP-164] Surprising Success Repeated Failure: Entropy-Guided Credit Assignment for Exploration in LLM Reasoning
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在强化学习中进行细粒度信用分配（credit assignment）时面临的挑战，尤其是现有方法普遍依赖辅助模型、额外采样或特权信息的问题。其核心解决方案是提出一种基于熵引导的信用分配机制——熵优势策略优化（Entropic Advantage Policy Optimization, EAPO），关键在于通过不对称地处理成功与失败情形下的策略熵信号，实现更精准的奖励分配：对于高熵的成功决策，给予更强的正向激励以强化意外成功的可重复性；对于低熵的失败决策，则施加更强惩罚以纠正顽固错误；同时，在不确定性较高的位置减弱惩罚强度，保留探索恢复的可能性。该方法仅利用已有回放（rollout）信号重构逐标记（token-level）奖励，无需额外监督，从而在多个基础及推理型模型上均实现了最优性能，并显著提升了探索效率与解题覆盖范围。
链接: https://arxiv.org/abs/2609.33781
作者: Woongyeong Yeo,Minki Kang,Chanuk Lee,Sangwoo Park,Jinheon Baek,Sung Ju Hwang
机构: KAIST(韩国科学技术院); DeepAuto.ai
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Project page : this https URL
Abstract:Reinforcement learning with verifiable rewards (RLVR) enhances reasoning in large language models (LLMs) through outcome-level feedback, yet recent approaches to finer-grained credit assignment often require auxiliary models, additional sampling, or privileged information. Although policy entropy provides a readily available signal, prioritizing uncertain positions under both reinforcement and penalization concentrates penalties where failed responses still retain alternatives for recovery, which can suppress opportunities for exploration. To address this, we introduce Entropic Advantage Policy Optimization (EAPO), an entropy-guided credit assignment method that treats success and failure asymmetrically. Specifically, motivated by the observation that success under uncertainty is less repeatable while confident failures tend to recur, EAPO couples normalized policy entropy with the sign of the response advantage to reinforce surprising success and correct repeated failure. It assigns stronger reinforcement to high-entropy decisions in successful responses and stronger penalties to low-entropy decisions in failed responses, while attenuating penalties at uncertain positions to preserve opportunities for recovery. By redistributing the response advantage across tokens, EAPO derives token-level credit directly from existing rollout signals without additional supervision. We validate EAPO on a range of reasoning tasks across both base and reasoning backbones, demonstrating that it achieves the best overall performance. We further show that EAPO promotes more effective exploration, broadening problem coverage and generating more diverse candidate answers.
[NLP-165] okens Change Structure Endures: Spectral Watermarking for Generated Speech
【速读】： 该论文旨在解决生成式语音中令牌级水印（token-level watermarking）在重编码（retokenization）过程中的鲁棒性问题。现有方法在将生成语音解码为波形并重新编码时，因词汇表映射变化导致令牌身份改变，从而严重削弱水印信号。其核心解决方案是提出一种名为Redwing的新型水印框架，其关键在于：通过构建重编码过程中观察到的令牌替换关系图，利用其拉普拉斯矩阵（Laplacian）提取出能反映令牌间可互换性的基底，并在此基底上联合优化嵌入与检测函数。该设计使水印信号在经历多次重编码后仍保持稳定，同时有效控制未加水印语音上的检测误差和嵌入失真。实验表明，经过八次连续的Mimi重合成后，Redwing在Mosho全双工系统上达到80.7%的真阳性率（TPR），显著优于KGW（8.3%）和WMAR（最高7.3%），且在多种神经编解码器及不同文本转语音模型上均表现出优异的泛化性能，证明了重编码过程的转移结构可被主动利用为增强水印鲁棒性的设计原则。
链接: https://arxiv.org/abs/2609.33774
作者: Kanghwi Lee,Kyeongseok Jeong,Jeongmin Liu
机构: Institute of Neuroinformatics, University of Zurich and ETH Zurich(苏黎世联邦理工学院与苏黎世大学神经信息学研究所); NAVER Cloud(NAVER云)
类目: ound (cs.SD); Computation and Language (cs.CL); Cryptography and Security (cs.CR); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
备注:
Abstract:Watermarking is a promising tool for establishing the provenance of AI-generated speech. While many neural audio watermarking methods rely on a separately trained watermark generator, token-level watermarking is a training-free alternative that operates directly during generation. Its main weakness is retokenization: decoding generated speech to a waveform and encoding it again can change token identities and erode the watermark. To make the watermark robust to these changes, we propose Redwing, REtokenization-Durable Watermarking IN Generation. It builds a graph from the token substitutions observed under retokenization, whose Laplacian yields a basis that assigns similar values to tokens likely to substitute for one another. Over this basis, embedding and detection functions are jointly optimized to preserve watermark signal through retokenization while limiting embedding distortion and detector variability on unwatermarked speech. On the Moshi full-duplex system, after eight consecutive passes of Mimi resynthesis, Redwing achieves 80.7% TPR at a calibrated 1% FPR, compared with 8.3% for KGW and at most 7.3% for WMAR. It also has the highest TPR after eight passes through three other neural codecs (77.5-93.0%), and the gains generalize to TTS models at a speech-quality cost close to that of KGW. These results show that retokenization is not merely a source of noise: its transition structure can be exploited as a design principle for robust token-level watermarking.
[NLP-166] Positions Are Not Facts: The Mismatch Between KV Caches and Memory
【速读】： 该论文旨在解决语言模型在事实更新时如何有效调整其键值缓存（KV cache）中存储的历史信息这一问题。核心挑战在于：当事实发生变化时，是应简单隐藏旧记录、仅隐藏被替换的值，还是彻底删除旧文本并重新计算缓存？研究发现，单纯隐藏整个记录虽成本低，但可能导致关键历史信息丢失，进而引发单位错误或生成中断；而仅隐藏被替换值则能保留当前答案的完整性。进一步分析表明，保留对象间的依赖关系及未修改的修订段落对维持历史准确性至关重要——在多跳更新任务中，若重建未改变位置的状态，历史准确率下降20–41个百分点，而移动现有状态影响较小。此外，固定缓存相似性难以捕捉细微差异，而通过学习的读出机制可恢复合成样本中被忽略的区别；但在自然文本场景下，基于文本检测器选择的掩码与随机掩码在相同覆盖率下表现无显著差异。查询相关的访问机制虽可缓解部分损失，但需额外存储或访问开销。综上，该研究的关键结论是：在使用KV缓存作为可更新记忆时，必须不仅保留被替换值，还需维护对象依赖结构和未变更的上下文片段，以确保历史信息的完整性和推理一致性。
链接: https://arxiv.org/abs/2609.33759
作者: Changhai Zhou,Yuhua Zhou,Shiyang Zhang,Jun Gao,Zhen Li,Hua Wu,Hanchao Yu,Haifeng Wang
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 132 pages, 28 figures
Abstract:When a fact changes, how should a language model update the history stored in its key-value (KV) cache? Hiding the old record is cheap, but it may still contain needed details or answer questions about the past. We compare hiding whole records, hiding only replaced values, and deleting old text and recomputing the cache. In a controlled quantity task, masking makes all eight models prefer the new value more strongly, yet six lose complete answers through unit errors or failure to stop; keeping the unit preserves all current answers. Later states also retain useful information from earlier records: on multi-hop updates, rebuilding these states at unchanged positions lowers historical accuracy by 20-41 percentage points, whereas moving the existing states has little effect. Keeping object dependencies and unchanged revision passages prevents many losses. Recognition is a separate challenge. Learned readouts recover distinctions missed by fixed cache similarities on synthetic record pairs. On natural text, text-detector-selected masks show no clear advantage over random masks at the same rate in 14 same-model detector-generator comparisons. Query-dependent access can avoid some losses, with additional storage or access costs. These findings identify what must be preserved beyond the replaced value when using a KV cache as updatable memory.
[NLP-167] DuraS2ST: Chain-of-Thought and Reinforcement Learning for Duration-Aligned Speech-to-Speech Translation EMNLP2026
【速读】： 该论文旨在解决时敏应用场景（如视频配音）中语音到语音翻译（S2ST）面临的多重挑战，核心问题在于如何在保证语义忠实性与说话人特征保留的同时，实现严格的时长一致性，以避免音画不同步。现有S2ST系统普遍缺乏显式的时序规划机制，导致目标语音时长控制困难。其解决方案的关键在于提出DuraS2ST框架，通过引入一种时长对齐的思维链（Chain-of-Thought, CoT）推理范式，使单一语音语言模型能够先生成包含目标语篇内容与音素长度规划的显式推理路径，再据此合成对应语音标记。为支持该范式，研究构建了高质量的时长对齐CoT语料库DuraSet-440K，用于监督初始化；并采用多模态多维度强化学习进行优化，设计时长裕度奖励（Duration Margin Reward）以平衡翻译质量与时长一致性，并引入模态感知奖励分配机制（Modality-Aware Reward Attribution），精准地将奖励分配至相应的词元片段。实验结果表明，DuraS2ST在CVSS-T数据集上实现了翻译质量与时长一致性之间的优异平衡，显著优于多个开源及商业基准模型。
链接: https://arxiv.org/abs/2609.33742
作者: Yayue Deng,Dingdong Wang,Yuxuan Hu,Jinyu Li,Yanqing Liu,Yuanyuan Wang,Weidong Chen,Helen M. Meng,Shujie Liu,Xixin Wu
机构: The Chinese University of Hong Kong (香港中文大学); Microsoft Corporation
类目: ound (cs.SD); Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
备注: Accepted to EMNLP 2026 (Main Conference)
Abstract:Speech-to-speech translation (S2ST) in time-sensitive applications such as video dubbing requires not only semantic fidelity and speaker preservation, but also strict duration consistency to avoid audio-visual misalignment. However, existing S2ST systems largely generate target speech without explicit temporal planning, making duration control an unresolved challenge. We introduce DuraS2ST, a duration-aligned reasoning framework that enables a single speech language model to first generate an explicit chain-of-thought (CoT) for planning target wording and phonetic length, and then synthesize the corresponding speech tokens. To support this paradigm, we construct DuraSet-440K, a high-quality duration-aligned CoT corpus for supervised initialization. We further optimize the model with multi-modal multi-dimensional reinforcement learning, using a Duration Margin Reward to balance translation quality and duration consistency, and Modality-Aware Reward Attribution to assign rewards to appropriate token spans. Experiments on CVSS-T show that DuraS2ST achieves a strong balance between translation quality and duration consistency, outperforming competitive open-source and commercial baselines. Project page: this https URL.
[NLP-168] he Effects of Incremental Instruction Delivery on Language-Model Creative Writing
【速读】： 该论文旨在解决生成式AI在交互式创意写作场景中，随着多轮对话逐步引入需求时，创作质量是否因“需求渐进式传递”而下降的问题。传统研究多聚焦于具有明确可验证结果的任务，难以捕捉创意性成果在交互过程中隐含的结构性与连贯性损失。为此，研究通过160项由人类撰写的跨六种体裁的创意写作任务，将每项任务的完整要求或逐轮（5-9轮）递进式呈现给六类开源大模型，构建了960组匹配对进行对比分析。结果表明，渐进式交付显著降低显式需求遵循度，并在结构与连贯性方面造成最严重的质量退化；即使在显式遵循度相同的输出间，结构性差异依然存在，说明单纯衡量需求保留率不足以解释质量下降。为此，论文提出“创意完整性（Creative Integrity）”这一综合指标，用于衡量需求遵循与叙事结构的联合表现，在渐进式交付下，模型仅保持71.2%（95% CI [68.2%, 74.3%]）的全量创意完整性。三名独立评审者对50组匹配样本的人工评估进一步证实，全量要求条件下产出在结构、连贯性、文笔及体裁适配性上均具显著优势，且自动化评分与人工评价高度一致。研究结论强调：交互式创意写作系统不仅需评估需求是否在对话中被保留，更应关注动态演化的请求能否被有效整合至最终作品的内在逻辑结构之中。
链接: https://arxiv.org/abs/2609.33738
作者: Anshuman Singh,Abrar Eyasir,Haseeb Yaqoob,John Manavalan
机构: SGT UNIVERSITY; University of Dhaka (达卡大学); NED University of Engineering and Technology; Metea Valley High School (伊利诺伊州梅特亚山谷高中)
类目: Computation and Language (cs.CL)
备注: 18 pages, 4 figures, 13 tables. Code and data available at this https URL
Abstract:Large language models are increasingly used as interactive writing tools, where users develop stories, revise ideas, and introduce new requirements across multiple turns rather than specifying a complete brief upfront. Yet most evidence on multi-turn instruction degradation comes from tasks with objectively verifiable outcomes, leaving unclear whether incremental interaction harms creative artifacts in ways that explicit requirement checks cannot capture. We study this question using 160 human-authored creative-writing tasks across six genres, presenting each intended specification either upfront or progressively over 5-9 turns to six distinct open-weight model families, yielding 960 matched pairs. Progressive delivery reduces explicit constraint adherence and produces its largest writing-quality degradation in structure/coherence. The structural gap persists among outputs with equal observed adherence, suggesting that measured requirement loss alone does not explain the observed structural difference. We define Creative Integrity as a compact measure of joint adherence and narrative structure; under incremental delivery, models retain 71.2% of FULL Creative Integrity (95% CI [68.2%, 74.3%]). A three-rater human study over 50 matched pairs independently recovers FULL advantages in structure/coherence, craft, and genre effectiveness, while automated scores remain positively associated with aggregated human ratings. These findings show that interactive creative-writing systems should be evaluated not only on whether requirements survive conversation, but also on whether evolving requirements remain coherently integrated into the final artifact. Our dataset, benchmarks, and source code are available at: this https URL
[NLP-169] Yorùbá in Unicode: An Overview of a Problem
【速读】： 该论文旨在解决尤鲁巴语（Yorùbá）在互联网及计算机系统中书写时长期存在的技术难题，即依赖变音符号（diacritics）进行语义区分的非洲语言因Unicode未编码一组关键预组字符（precomposed characters），导致用户与数字系统被迫采用组合字符序列（combining character sequences）。此类序列在不同平台间行为不一致、易受字体替换影响而损坏，且在搜索功能中失效。论文通过大量实际案例——涵盖已出版书籍、网页平台及移动键盘——实证了该问题在现实场景中的普遍性与严重性。研究指出，Unicode的NFC（Normalization Form C）规范化稳定性政策构成结构性障碍，阻碍了简单修复方案的实施；因此，论文主张国际统一码联盟（Unicode Consortium）应主动介入，提出正式编码请求，将尤鲁巴语四个核心字符纳入标准，以实现持久可靠的解决方案。
链接: https://arxiv.org/abs/2609.33734
作者: Kólá Túbòsún
机构: 未知
类目: Computation and Language (cs.CL)
备注: To appear in Yorùbá Print Culture: A Handbook, Routledge
Abstract:There is a recurrent problem in the writing of Yorùbá on the internet and on the computer that has proven intractable over the years. The language, along with other African languages that depend on diacritics for disambiguation, requires a small set of precomposed characters that Unicode does not encode. This has forced writers and digital systems to rely on combining character sequences that behave inconsistently across platforms, corrupt under font substitution, and fail in search. This paper documents that failure across a range of real world contexts, from published books to web platforms to mobile keyboards, using personal and empirical evidence. It identifies Unicode’s NFC normalization stability policy as the structural constraint that prevents a straightforward fix, arguing for direct intervention of the Consortium in solving the active problem, proposing a formal encoding request for the four core Yorùbá characters as the most durable path to resolution.
[NLP-170] BOReFT: Manifold Steering of Language Models for Black-box Optimization
【速读】： 该论文旨在解决生成式语言模型（Generative Language Models, GLMs）在黑箱搜索任务中，如程序优化与分子设计时，存在搜索空间探索不充分、效率低下且控制能力有限的问题。现有方法依赖迭代提示或参数更新，难以有效调控搜索过程的广度与深度。为此，本文提出BOReFT（Bayesian Optimization with Representational Fine-Tuning），其核心创新在于：在冻结的语言模型基础上，学习一个紧凑的、低维的隐状态干预空间，并将该空间作为贝叶斯优化（Bayesian Optimization, BO）的连续搜索域，结合外部评分函数实现高效搜索。关键在于，该隐状态空间不仅覆盖语义区域且具备平滑性，支持有效的插值与优化；理论上，语义覆盖率和插值能力决定了可获得的最佳性能上限，且从该空间解码可构建标准的随机多臂赌博机（stochastic bandit）观测模型，支持自适应搜索策略。实验表明，相较于强基线模型，BOReFT在可解释的“Semantle”词搜索任务中发现了更多隐藏目标，并在三个真实分子属性优化任务中的两项上实现了更高的目标属性得分。因此，该方法为基于语言模型的离散提案空间与连续黑箱优化之间建立了一个理论严谨且高效的桥梁。
链接: https://arxiv.org/abs/2609.33722
作者: Dhruv Agarwal,Rico Angell,Kavitha Srinivas,Tahira Naseem,Horst Samulowitz,Willie Neiswanger,Andrew McCallum
机构: University of Massachusetts Amherst(马萨诸塞大学阿姆赫斯特分校); New York University(纽约大学); IBM Research(IBM 研究院); University of Southern California(南加州大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Language models are increasingly used as proposal models for black-box search, from program optimization to molecular design. Existing approaches typically improve proposals through iterative prompting or parameter updates, offering limited control over how completely and efficiently the model’s search space is explored. Continuous optimization methods, such as Bayesian optimization, provide a principled way to search but require a suitable domain to operate over. To address this, we introduce BOReFT, which learns a compact, low-dimensional space of hidden-state interventions in a frozen language model, and uses this space as the search domain for Bayesian optimization with an external scoring function. Empirically, we find that the learned domain spans semantic regions and exhibits smoothness properties that support search. Theoretically, we show that semantic coverage and interpolation control the best score available in the learned space, and that decoding from this space yields a standard stochastic-bandit observation model for adaptive search. We evaluate BOReFT on the interpretable word search task “Semantle” and on three more real-world discovery tasks in de novo molecule property optimization. Compared to strong LLM baselines, BOReFT finds in Semantle a higher number of hidden targets and, on two out of three molecular objectives, achieves higher property scores. Consequently, our method provides a principled new bridge between discrete proposal spaces of LLM-based search and continuous black-box optimization.
[NLP-171] Shared Experience Separate Learning: Companion Confidence Calibration for LLM s
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在自我评估中过度自信的问题，即模型在生成错误答案时仍保持高置信度，从而影响其可靠性。传统方法通常在模型训练完成后进行事后置信度校准（post-hoc calibration），但这种方法无法动态适应模型能力的演化。为此，本文提出一种并发置信度校准（concurrent confidence calibration）范式，即在提升模型能力的同时同步学习置信度，而非事后单独校准。其核心创新在于提出“共享经验、分离学习”（shared experience but separate learning）的设计原则：能力与置信度虽基于相同的推理轨迹进行学习，但通过独立的优化机制和参数空间实现解耦。基于此，作者提出CoCal（Companion Confidence Calibration）框架，通过一个轻量级的伴随网络（companion model），从回放过程中的隐藏状态和验证器提供的可验证正确性标签中学习置信度，同时保持原始任务优化不变。实验结果表明，CoCal在Qwen3-8B和Qwen3-14B上均显著提升了置信度估计的准确性，且未牺牲任务性能，优于现有的基于强化学习的并发方法及匹配的事后校准方法。此外，所学伴随模型具备良好的跨领域泛化能力和对策略迁移的鲁棒性，且在不同规模下均保持有效性。
链接: https://arxiv.org/abs/2609.33721
作者: Shiyu Ni,Keping Bi,Jiafeng Guo,Yilong Xu,Jingtong Wu,Zengxin Han,Xueqi Cheng
机构: State Key Laboratory of AI Safety; Institute of Computing Technology, Chinese Academy of Sciences; University of Chinese Academy of Sciences
类目: Computation and Language (cs.CL)
备注:
Abstract:Reliable self-assessment is essential for large language models (LLMs), yet they often remain highly confident when their answers are wrong. We study \emphconcurrent confidence calibration, where confidence is learned alongside capability improvement rather than calibrated only after training. Reinforcement learning from verifiable rewards (RLVR) provides a natural setting for this paradigm, as it continuously produces responses paired with verifiable correctness feedback. Existing concurrent methods, however, learn both capability and confidence through reinforcement learning within shared policy parameters, potentially coupling two fundamentally different learning problems. We instead propose \emphshared experience but separate learning: capability and confidence learn from the same trajectories, but through separate optimization mechanisms and parameters. Based on this principle, we introduce \textbfCoCal (Companion Confidence Calibration), which trains a lightweight companion from rollout hidden states and verifier-derived correctness supervision while leaving task optimization unchanged. Experiments on Qwen3-8B and Qwen3-14B show that CoCal improves confidence estimation without sacrificing task performance, outperforming both RL-based concurrent methods and matched post-hoc calibration. The learned companion further generalizes across domains and policy shifts, while the benefits of CoCal persist at both scales.
[NLP-172] From Granular Revision Operations to Meaningful Revision Units: Evaluating LLM s for Revision Boundary Detection
【速读】： 该论文旨在解决生成式AI在学习分析中对写作修订过程进行细粒度分析时所面临的挑战，即如何准确识别有意义的修订单元边界。现有自动化文本比对方法常将一次有目的的修订拆分为多个碎片化的编辑操作，导致分析单位失真。为此，研究提出利用大语言模型（LLM）从结构化修订操作数据中识别具有语义一致性的修订单元边界，并评估其相较于传统非LLM基线方法的价值。关键解决方案在于采用参数高效微调（PEFT）的Qwen3-32B模型，结合上下文表示策略，显著提升了边界识别性能（宏平均F1达0.859），优于零样本与少样本提示的GPT-5.5及基础版本的Qwen3-32B，且在保持高精度的同时更好地捕捉了同一修订单元内的关系。研究还发现，尽管确定性后处理可提升提示型模型表现，但对已微调模型增益有限，表明通用提示策略难以超越透明的结构启发式方法，而经过任务适配的微调模型才具备真正提升学习分析质量的能力。
链接: https://arxiv.org/abs/2609.33720
作者: Yu Tian,Andrew Potter,Katerina Christhilf,Motahareh Darvishpour Ahandani,Jessica Early,Steve Graham,Danielle S. McNamara
机构: Arizona State University(亚利桑那州立大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 18 pages
Abstract:Revision traces provide valuable evidence about students’ writing processes, but their usefulness for learning analytics depends on how individual revisions are represented. Automated draft-comparison methods often produce granular edit operations that can fragment a single purposeful revision into multiple analytic units. This study evaluates whether LLMs can identify meaningful revision unit boundaries in structured revision operation data and whether they provide value beyond simple non-LLM baselines. Using 113 matched draft–revision pairs from undergraduate writing, expert annotation yielded 4,344 candidate boundaries. We compared zero- and few-shot GPT-5.5 and base Qwen3-32B, parameter-efficient fine-tuning of Qwen3-32B, and majority and proximity-based baselines. Despite receiving revision context and task instructions, no prompted LLM condition outperformed the proximity heuristic (macro-F1 = .825). In contrast, fine-tuned Qwen3-32B using the two context representation achieved the highest macro-F1 (.859), identifying more same-unit relationships while maintaining precision comparable to the heuristic. Deterministic post-processing substantially improved the prompted models but added little benefit to the strongest fine-tuned model. These findings suggest that LLMs can support revision boundary judgment when task-adapted, but general purpose prompting alone may not outperform transparent structural heuristics.
[NLP-173] Understanding Confabulation and Rethinking Reconstruction in Activation Explanations
【速读】： 该论文旨在解决自然语言自编码器（Natural Language Autoencoders, NLAs）在无监督文本解释生成过程中出现的幻觉（confabulation）与写作缺陷问题。现有基于点重建（point-reconstruction）的训练范式虽能提升解释对模型行为的预测能力，但导致解释内容逐渐引入未经支持的细节并产生语言表达缺陷。为系统评估这些变化，论文提出一个标准化的评估框架，用于量化解释中可恢复的信息量、其主张的上下文支持度以及写作风格质量。针对上述问题，其核心解决方案是突破传统单点激活预测的局限，转而建模与解释相容的激活分布；提出Flow-NLA方法，通过扩散似然界（diffusion likelihood bound）训练表述器（verbalizer），使解释能够区分具有相同均值和最优点重建奖励的不同激活分布。该方法在Qwen、Gemma和Apertus等多个模型上验证有效，在保持点重建带来的预测性能优势的同时，显著抑制了幻觉增长与写作缺陷，推动了基于激活的训练向更信息丰富、有依据且可读性更强的解释方向发展。
链接: https://arxiv.org/abs/2609.33702
作者: Gert Lek,Zixuan Xia,Pin-Yu Chen,Lydia Y. Chen
机构: Université de Neuchâtel (纳沙泰尔大学); Universität Bern (伯尔尼大学); IBM Research (IBM研究院); Delft University of Technology (代尔夫特理工大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 18 pages, 9 figures
Abstract:Natural Language Autoencoders (NLAs) produce unsupervised text explanations of a model’s activations: a verbalizer describes an activation and a reconstructor learns to recover it from this text. Under the established point-reconstruction NLA training recipe, explanations become more useful for predicting model behavior while also increasingly introducing unsupported details and exhibiting writing defects. To assess these changes separately, we introduce a standardized evaluation framework for unstructured NLA explanations, measuring information recoverable from explanations, contextual support for their claims, and writing quality. To address confabulation and writing defects, we move beyond predicting a single activation: explanations can distinguish distributions of possible activations even when their means and optimal point-reconstruction rewards are identical. We introduce Flow-NLA, which models the distribution of activations compatible with an explanation and trains the verbalizer using a diffusion likelihood bound. Across Qwen, Gemma, and Apertus, this richer signal retains the utility gains of point reconstruction while curbing the growth of confabulation and writing defects, opening up a direction for improving activation-derived training to encourage more informative, supported, and readable explanations. Code and evaluation prompts will be made publicly available upon acceptance.
[NLP-174] When Do Agents Help? Embedding LLM and Agent ic Alignment of Classical Texts and Their Translations
【速读】： 该论文旨在解决多语言古典文本（包括巴利语、梵语、米什纳希伯来语和藏语）中对齐工作流在机器翻译、信息检索与计算研究中的有效性评估问题，尤其关注不同对齐方法在参考对应关系恢复能力上的差异。其核心挑战在于现有文献中关于对齐流程的实证比较分散且缺乏统一基准。解决方案的关键在于系统性地对比七种对齐方法——包括四种基于嵌入（Embedding）的流水线、直接调用大语言模型（LLM）的方法、自主代理（Autonomous Agent）以及由独立审计员修正的代理——在452篇文本共9,833个已人工标注对齐单元上的表现。研究发现，生成式工作流（Generative Workflows）可恢复93%-94%的参考对应关系，显著优于嵌入方法（最高仅77%），且通过上限分析揭示句子边界限制了嵌入方法对某些参考项的表征能力。尽管各生成式方法在恢复率上表现相近（代理优势仅为0.5个百分点，置信区间-0.02至1.17），但代理能为全部452篇文本生成结构有效的输出，而直接调用仅成功于437篇。进一步的盲评三模型小组评估表明，多数不一致属于可接受的编辑变体，严重错误占比极低（0.06%-0.14%），且代理产生的残余缺陷显著低于直接调用（0.7% vs 1.4%），说明参考恢复率本身不足以反映对齐质量。在长篇巴利语论著测试中，代理及经审计的代理将恢复率从直接调用的71%提升至84%和92%，而在相同参考块匹配后三者均达93%。这表明代理在结构可靠性与短文本缺陷控制方面具有优势，但在长文档中其高恢复率优势在分块处理后消失，且独立审计在准备充分的文本上未带来可测量的额外增益。
链接: https://arxiv.org/abs/2609.33691
作者: Máté Metzger
机构: Independent Researcher(独立研究员)
类目: Computation and Language (cs.CL)
备注: Preprint. This manuscript has not yet been peer reviewed
Abstract:Classical texts aligned with their translations support machine translation, retrieval and computational research, but evidence comparing alignment workflows is scattered. This study compares seven systems on 452 texts in Pali, Sanskrit, Mishnaic Hebrew and Tibetan, comprising 9,833 human-aligned units: four embedding pipelines, a direct LLM call, an autonomous agent, and the agent revised by an independent auditor. Generative workflows recover 93-94% of reference correspondences, against at most 77% for embeddings. A ceiling analysis shows that sentence boundaries make some references unrepresentable by the embedding pipelines. Reference recovery is similar across generative workflows: the agent’s advantage is 0.5 percentage points (95% CI -0.02 to 1.17), and auditing adds no established benefit. Agents nevertheless produce structurally valid output for all 452 texts, against 437 for direct calls. A blinded three-LLM panel assesses every generative mismatch against the source and human reference. Most mismatches are labelled defensible editorial variation; consensus major-error labels cover only 0.06-0.14% of units. The panel labels significantly fewer residual defects for agents than direct calls (0.7% versus 1.4%), suggesting that reference recovery alone understates alignment quality. On ten long Pali discourses taken as published online, agents and audited agents raise recovery from the direct call’s 71% to 84% and 92%. Identical reference-located chunks bring all three to 93%. Agents thus improve structural reliability and reduce judged defects on short passages, while their large recovery advantage on long documents disappears after chunking. In this setting, independent auditing offers little measurable additional benefit on prepared passages.
[NLP-175] Auditing Agent Actions through Query-Conditioned Attribution
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）智能体在与用户、政策及外部工具交互过程中，因执行重要决策而引发的可审计性问题。现有方法在面对多样化审计目标时，缺乏针对特定问题的可追溯证据链，且在无法访问底层模型（如仅通过API部署）的情况下，依赖高成本的输入扰动或对完整行为轨迹进行外部大模型分析，效率低下。为此，论文提出“查询条件化的智能体动作归因”（query-conditioned agent action attribution）这一新任务，即以自然语言审计查询为输入，精准识别并排序与该查询相关的动作来源及其有序中间证据。为实现高效、针对性的归因，作者构建了A³Bench基准数据集，涵盖1,396个覆盖政策依据、参数溯源、故障传播及不当行为追踪等维度的审计查询。核心解决方案采用小型开源权重模型作为归因提议器，结合查询条件梯度显著性与查询语义相关性，对历史单元进行优先级排序。该提议器仅需两次前向传播和一次反向传播，即可在源定位（MRR提升达40.9%）与证据排序（MAP提升达42.1%）上显著优于开源基线，同时具备更强的查询敏感性。基于提议器集成的端到端系统，在源准确性上超越最强前沿模型（64.5% vs. 60.4%），且相较最快前沿API基线降低29.9%的部署延迟，实现了高效、精准、低开销的智能体审计能力。
链接: https://arxiv.org/abs/2609.33676
作者: Yifan Liu,Praveen Venkateswaran,Abdulhamid Adebayo,Dong Wang
机构: University of Illinois Urbana-Champaign(伊利诺伊大学厄本那-香槟分校); IBM(国际商业机器公司)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: preprint under review
Abstract:LLM agents increasingly take consequential actions through interactions with users, policies, and external tools. Auditing these agents requires automated attribution of realized actions to their historical basis. However, existing attribution formulations do not provide question-specific traces for diverse auditing objectives. Additionally, when access to the acting model is limited (e.g., in API-only deployments), applicable methods commonly rely on costly input perturbations or external LLM analysis of complete trajectories. We therefore formulate \textitquery-conditioned agent action attribution, a new task that takes a natural-language auditing query as input and recovers the source and ordered intermediate evidence for the query-specified aspect of an action. We instantiate this task with A^3Bench , a benchmark comprising 1,396 auditing queries across policy basis, parameter provenance, failure propagation, and unsafe-behavior tracing. To enable efficient, query-specific attribution, we use small open-weight models as attribution proposers that combine query-conditioned gradient saliency with query-semantic relevance to rank history units. Our proposer consistently achieves stronger source and evidence rankings at lower inference cost than open-weight baselines, improving source MRR by up to 40.9% and evidence MAP by 42.1% with only two forward passes and one backward pass. Controlled evaluations confirm that our proposer improves attribution specificity by adapting its rankings to fine-grained changes in the auditing query. Building on a proposer ensemble, our end-to-end system surpasses the strongest frontier-model baseline in source accuracy (64.5% vs.\ 60.4%) while reducing empirical deployment latency by 29.9% relative to the fastest frontier API baseline. Code and data will be released after the initial review period following final validation and cleanup.
[NLP-176] Reset Is Not Recovery: Evaluating Recoverability from False Conversational Context via Sycophancy Hysteresis
【速读】： 该论文旨在解决生成式对话模型在多轮对话中因用户持续施加错误主张（即“后压力”）而导致的偏差持续性问题，即模型在用户撤回错误主张后仍可能保留对错误答案的偏好，表现出“谄媚滞后效应”（sycophancy hysteresis）。其核心挑战在于：传统重置机制无法有效消除由用户施压引发的偏见，导致模型难以恢复到无干扰的准确回答状态。解决方案的关键在于区分哪些历史上下文应作为证据保留，哪些应被移除或隔离。研究发现，仅依赖历史保留型修复策略（如用户撤回、系统重置、自我验证）的恢复效果有限，而通过彻底清除压力相关历史信息（如全新上下文删除与上下文截断）可实现100%的恢复率；此外，在引入可信证据的“受信任证据”条件下，模型准确率从0.368显著提升至0.929，错误跟随率从41.2%降至4.3%，且该效果不受对话长度、重复自信度、合理干扰项或选项标签惯性等因素干扰。因此，论文提出，实现忠实的基于事实的对话，关键在于动态评估并主动管理上下文证据的有效性，对不当压力历史进行隔离或清除。
链接: https://arxiv.org/abs/2609.33672
作者: Adi Shnaidman
机构: DeepKeep; University of Pennsylvania (宾夕法尼亚大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Grounded language models are usually evaluated by adding relevant context, but multiturn dialogue also contains unsupported user claims that may contaminate later factual answers. We study post-pressure recoverability: whether a model returns to clean-context behavior after a user repeatedly advocates a wrong answer and then withdraws that pressure. We introduce a recovery-after-pressure protocol for multiple-choice factual dialogue and measure sycophancy hysteresis, the residual probability assigned to the user-advocated wrong answer relative to a clean-context counterfactual. Across seven instruction-tuned open-weight models and two factual benchmarks, ordinary reset often reduces but does not erase pressure-induced bias. History preserving repairs such as user retraction, system reset, and self-verification recover only 2-3/14 model-dataset pairs under the strict clean-restoration diagnostic, whereas operations that change the effective context are substantially more reliable; the two conditions that remove the pressure-bearing history entirely, fresh-context deletion and context truncation, recover 14/14. In an oracle trusted-evidence condition across fourteen model-dataset pairs, preserving the pressure-bearing history while adding benchmark-derived trusted evidence increases accuracy from 0.368 to 0.929, while wrong-answer following falls from 41.2% to 4.3%. Controls show that the effect is not explained by dialogue length, repeated confidence, plausible distractors, mere false-answer mention, or option-label inertia. These results suggest that faithful grounded dialogue requires evaluating which prior context should be treated as evidence and which should be removed or quarantined before answering.
[NLP-177] Closing the Cross-Dialect Gap: Query Plans as a Portable Interface in Text-to-SQL EMNLP2026
【速读】： 该论文旨在解决当前文本到SQL（Text-to-SQL）系统在跨数据库方言（如SQLite、PostgreSQL、MySQL、ClickHouse等）部署时普遍存在的性能下降问题。现有模型通常仅在单一方言（如SQLite）上训练和评估，导致其在其他方言上的跨方言准确率显著降低，且这一现象在不同规模、架构乃至专用模型中均普遍存在。其解决方案的关键在于改变生成目标：不再要求大语言模型（LLM）直接生成特定方言的SQL，而是生成与方言无关的关系代数查询计划（relational algebra query plan），由一个确定性编译器将该计划高效转换为任意支持后端的SQL代码。实验表明，该方法在13种从3B到前沿规模的模型上几乎统一恢复了跨方言可移植性，对具备提示能力的模型仅带来微小的本方言峰值准确率损失，而经过查询计划监督微调的模型则无任何损失；在同等微调条件下，基于查询计划的监督相比基于SQL的监督能训练出更强的模型。此外，论文提出一种名为MetricName的新评估指标，该指标能够感知问题语义，有效区分语义错误与良性方言间差异，从而实现跨方言的公平评估。研究结果揭示了一个重要原则：为执行效率选择的生成目标，并不必然最大化生成质量。
链接: https://arxiv.org/abs/2609.33670
作者: Corentin Royer(1 and 2),Robin Oester(1),Yotam Perlitz(1),Yannick Metz(2),Andrea Giovannini(1),Mennatallah El-Assady(2) ((1) IBM Research, Zurich, Switzerland, (2) ETH Zurich, Zurich, Switzerland)
机构: IBM Research, Zurich, Switzerland(国际商业机器公司研究部,苏黎世,瑞士); ETH Zurich, Zurich, Switzerland(苏黎世联邦理工学院,苏黎世,瑞士)
类目: Computation and Language (cs.CL)
备注: Accepted at Findings of the Association for Computational Linguistics: EMNLP 2026
Abstract:Text-to-SQL systems are typically trained and evaluated on a single dialect (SQLite), yet production deployments span PostgreSQL, MySQL, ClickHouse, and beyond. We show that this single-dialect assumption leads to a substantial drop in cross-dialect accuracy for every model we tested. The drop persists across scale, architecture, and even purpose-built text-to-SQL systems. We argue that the fix is to change the generation target: instead of asking an LLM to emit dialect-specific SQL, we have it emit a dialect-agnostic relational algebra query plan, which a deterministic compiler then renders into SQL for any supported backend. Across thirteen models from 3B to frontier scale, this restores cross-dialect portability nearly uniformly, at a small cost in peak accuracy on the model’s home dialect for capable prompted models and none once fine-tuned on plans; under matched fine-tuning, plan supervision yields a stronger model than SQL supervision. We also introduce MetricName, a question-aware result-set comparator needed to evaluate fairly across dialects, where existing metrics confound semantic errors with benign cross-dialect variation. More broadly, the result is a reminder that a generation target chosen for execution is not necessarily the one that maximizes generation quality.
[NLP-178] One Model Is Not a Crowd: Multi-LLM and Aspect-Conditioned Diverse Comment Generation
【速读】： 该论文旨在解决生成式AI在在线评论空间中生成内容时面临的多样性不足问题，即当前以大语言模型（Large Language Model, LLM）为基础的AI代理所生成的评论趋于同质化，长期使用可能导致“模型坍缩”（model collapse），进而削弱数字交流的语义丰富性与社会语用多样性。其解决方案的关键在于：通过融合来自不同提供商的多大语言模型（multi-LLM）并引入基于特定话题维度（aspect-conditioned generation）的生成机制，以更有效地模拟人类话语的多元性。研究提出一个涵盖语义、语言及社会语用特征的三轴评估框架，从分散度（dispersion）、覆盖范围（coverage）和对齐度（alignment）三个维度量化分析评论多样性。基于超过200万条跨领域的YouTube评论的大规模实证研究表明，该方法生成的内容在分布上更接近人类评论，且在预训练数据筛选和下游任务中均表现出良好的有效性，尽管仍无法完全复制人类多样性，但为构建更具社会情境感知能力的AI对话系统提供了可实践的技术路径。
链接: https://arxiv.org/abs/2609.33666
作者: Nafis Irtiza Tripto,Delvin Ce Zhang,Mahjabin Nahar,Dongwon Lee
机构: Pennsylvania State University (宾夕法尼亚州立大学); University of Sheffield (谢菲尔德大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Human communication on the internet is shaped by diverse perspectives, most visibly expressed in online comment spaces. As large language model (LLM)based AI agents begin to inhabit these spaces, a key question arises: whether synthetic comment threads can capture the diversity inherent in human discourse. This concern is increasingly important, as the growing presence of homogenized AI-generated content risks reducing diversity over time, potentially leading to model collapse and degrading the richness of digital communication. Inspired by the plurality of human crowds and the aspect-driven nature of discourse, we hypothesize that comment diversity is better approximated by combining multiple LLMs with aspect-conditioned generation. We formalize and evaluate this approach using models from different providers and introduce a framework that characterizes diversity across semantic, linguistic, and socio-pragmatic features along three axes: dispersion, coverage, and alignment. Using this framework, we conduct a large-scale study on over 2 million YouTube comments across multiple domains. Our results reveal that multi-LLM and aspect-conditioned generation better align with human comment distributions and such data remains viable under pretraining style curation and is effective for downstream tasks. Yet, human diversity remains unmatched. Overall, our findings provide a practical foundation for generating more diverse and socially grounded discourse in AI-mediated environments.
[NLP-179] Audit-First VAPO: Risk-Certified Selective Updates under Imperfect Verification
【速读】： 该论文旨在解决在存在不完美验证者（imperfect verifiers）的情况下，即使通过裁剪（clipping）和正则化限制了更新方向的幅度，仍可能引入有害更新方向的问题。其核心挑战在于如何在保证安全性的同时，有效区分并采纳具有信息量的更新方向，同时控制更新幅度以避免风险累积。解决方案的关键在于提出Audit-First VAPO框架，该框架通过将离散方向准入（discrete directional admission）与连续幅度控制（continuous magnitude control）分离实现解耦：首先采用仅基于观察的“接受-申诉-弃权”（accept-appeal-abstain）策略，在有限的二次验证预算下进行方向性决策，且动作轨迹在干净标签加入前即被冻结；随后利用条件霍夫丁-阿祖马界（conditional Hoeffding-Azuma bounds）对因共享预算导致的依赖性进行建模，实现对所选有害风险、覆盖率及验证者调用率的有限样本置信认证。当方向被批准后，再通过一个受信任裁剪-KL（trust-clip-KL）执行器来约束更新幅度。实验表明，该方法在多个推理基准上显著优于静态强化学习验证（RLVR）、匹配随机选择、置信阈值法、噪声校正及验证者增强等基线方法，在目标风险ρ=0.08时实现了74.1%的准确率，有害风险控制在0.0697，覆盖率达0.4125，验证成本仅增加1.16倍，并在多种噪声模式下保持高度稳定性，有效隔离了方向选择的信息价值与抑制提议、缩小更新或额外验证计算的影响。
链接: https://arxiv.org/abs/2609.33662
作者: Miaobo Hu,Shuhao Hu,Xiaobo Guo,Xin Wang,Bokun Wang,Rui Chen,Daren Zha,Jun Xiao
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 33 pages, 7 figures
Abstract:Imperfect verifiers can assign a harmful update direction even when clipping and regularization bound its magnitude. We introduce Audit-First VAPO, which separates discrete directional admission from continuous magnitude control. An observation-only accept-appeal-abstain policy uses a finite secondary-verification budget; its action trace is frozen before clean labels are joined. Simultaneous finite-sample bounds then certify selected harmful risk, coverage, and verifier-call rate over a predeclared policy family. Conditional Hoeffding-Azuma bounds account for the dependence induced by shared budgets, and rollout or verifier changes initiate a new certification stage. After admission, a bounded trust-clip-KL actuator controls magnitude. We evaluate two models on two reasoning benchmarks against static RLVR, matched-random selection, confidence thresholding, noise correction, and verifier augmentation. On Qwen3.5-0.8B and GSM8K at target risk \rho=0.08 , RC-VAPO achieves 74.1% accuracy, selected harmful risk 0.0697, coverage 0.4125, and relative verifier cost 1.16\times . At matched coverage and update magnitude, its selected-risk difference from matched random is -0.0260 with paired 95% interval [-0.0364,-0.0157] . Across asymmetric, confidence-dependent, and correlated-verifier noise, the certificate is satisfied on 57 of 60 independent runs. These comparisons isolate informative directional selection from proposal suppression, update shrinkage, and additional verifier computation.
[NLP-180] subame: Tree Replay for Diffusion-Based Speculative Decoding
【速读】： 该论文旨在解决生成式 AI（Generative AI）中推测解码（speculative decoding）框架在动态树结构（dynamic trees）与随机采样链（sampled chains）之间性能权衡的问题。具体而言，尽管上下文感知的动态树能够根据候选路径概率自适应调整深度和分支以提升结构效率，但在随机解码场景下，其因依赖确定性高分词元构建节点而无法充分受益于随机采样与先进验证机制，导致在部分设置中落后于纯采样链。解决方案的关键在于打破动态树构建过程中的“拓扑-采样耦合”：通过两阶段框架——第一阶段基于草稿路径得分规划并冻结上下文感知的树结构；第二阶段在固定拓扑上重新采样以填充节点，从而在保留结构优势的同时引入随机性与验证收益。该方法由基于扩散模型的草稿生成器（diffusion-based drafters）实现，因其并行输出或轻量级条件修正能力可低成本重构候选集。作者提出 Tsubame 框架，证明其在兼容采样与验证策略下为无损设计，并在三种扩散基草稿器、六个数据集及多种候选预算条件下，显著提升接受长度与吞吐量，甚至逆转传统确定性动态树对采样链的劣势。
链接: https://arxiv.org/abs/2609.33652
作者: Yepeng Weng,Qiao Hu,Takehisa Yairi
机构: The University of Tokyo(东京大学); National Center for Mathematics and Interdisciplinary Sciences (NCMIS), AMSS, CAS(中国科学院数学与系统科学研究院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Context-aware dynamic trees allocate the speculative decoding budget according to draft path probabilities, adapting their depth and branching to the current context. Under stochastic decoding, however, we find that this structural advantage does not always compensate for the acceptance gains of random sampling paired with advanced verification, and such dynamic trees can fall behind sampled chains in some settings. These trees grow their topology from the candidates themselves, so the tokens submitted for verification are typically the deterministic high-score tokens selected during construction. This coupling is not inherent: once the topology is fixed, its nodes can be repopulated by sampling, allowing dynamic trees to retain their structural advantage while also benefiting from random sampling and advanced verification. Diffusion-based drafters make this practical, as their parallel outputs or lightweight conditional corrections allow candidates to be regenerated cheaply after the complete topology is known. We introduce Tsubame, a two-pass tree speculative decoding framework for diffusion-based drafters. The first pass plans and freezes a context-aware topology using draft path scores; the second replays the fixed topology, sampling the tokens that populate its nodes to form the candidate tree for verification. We prove that Tsubame is lossless under compatible sampling and verification strategies. Experiments across three diffusion-based drafters, six datasets, and multiple candidate budgets show that Tsubame improves acceptance length and throughput over deterministic trees, including settings where it reverses their disadvantage against sampled chains.
[NLP-181] Probe to Act: Elevating Browser-Use Agent via Active Visual Probing EMNLP26
【速读】： 该论文旨在解决浏览器智能体（browser-agent）在长期交互中如何实现结构化网页元数据（如文档对象模型，DOM）与视觉信息之间的无缝对齐，同时保持上下文相关性的问题。现有方法通常依赖于截图级别的动作预测或静态的标记集合（Set-of-Marks, SoM）叠加，导致模型在每次操作前需重新处理密集的DOM-像素对齐，效率低下且易出错。其解决方案的关键在于提出一种名为“探查即行动”（Probe to Act, P2A）的主动探查框架，将对齐过程从执行前移至决策时刻。P2A通过按需将符号化的DOM结构渲染回像素空间，实现符号假设与视觉布局间的非对称桥梁的动态校准；在执行状态变更操作前，智能体可发出轻量级探查指令，以实现DOM句柄到像素证据的转换、屏幕区域向DOM候选的映射、仅基于视觉的目标注册以及经验证的观测记录。这些交错进行的探查与动作过程自然构建了基于证据的记忆机制：仅保留经过探查、已执行或显式提交的观测内容，从而在长周期任务中仅维护决策关键信息。该方法既可作为专有模型在标准DOM+SoM接口下的提示策略，也可通过冷启动合成与自举式监督微调（SFT）蒸馏至开源权重模型。在三个浏览器使用基准测试中，P2A显著提升了专有模型与微调模型的任务成功率；例如在VisualWebArena上，其使Gemini-3-Pro的成功率从54.1%提升至61.2%，Qwen3-VL-8B从24.6%提升至32.9%，同时仅需约1.2倍于仅动作历史的峰值输入上下文，即可达到传统全观测历史（约3倍）的性能表现。
链接: https://arxiv.org/abs/2609.33646
作者: Keliang Li,Heng Wang,Chen Hu,Daxin Jiang,Hong Chang,Shiguang Shan
机构: Institute of Computing Technology, Chinese Academy of Sciences(中国科学院计算技术研究所); University of Chinese Academy of Sciences(中国科学院大学); StepFun(思必驰)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: EMNLP 26 Findings
Abstract:Browser-use agents require seamless alignment between structured web metadata and visual information, while preserving relevant context across long interactions. Existing interfaces often rely on either screenshot-level action prediction or static Set-of-Marks overlays, leaving the model to resolve dense DOM-pixel alignment before every operation. We introduce Probe to Act (P2A), an active probing framework for the browser-agent loop that moves this alignment into decision time. P2A addresses an asymmetric bridge between symbolic DOM hypotheses and screenshot layout by rendering on-demand symbolic DOM structure back into pixels. Before committing a state-changing browser operation, the agent can issue lightweight probes to translate DOM handles into pixel evidence, map screen regions back to DOM candidates, register visual-only targets, and commit verified notes. These interleaved processes naturally produce evidence-based memory: only probed, acted-on, or explicitly committed observations are kept across steps, preserving only decision-critical evidence in long-horizon contexts. P2A can be used as a prompting strategy for proprietary models under the standard DOM+SoM interface, and can be distilled into open-weight models through cold-start synthesis and self-bootstrapped SFT. Across three browser-use benchmarks, P2A shows clear gains on task success rate for both proprietary and fine-tuned models; on VisualWebArena, for example, it improves Gemini-3-Pro from 54.1% to 61.2% and Qwen3-VL-8B from 24.6% to 32.9%, while matching the costly full-observation history ( \sim 3 \times ) at only \sim 1.2 \times the peak retained input context of action-only history.
[NLP-182] Learning to Learn from Context: Synthetic Training from Perturbed Public Documents
【速读】： 该论文旨在解决大语言模型（LLM）在真实任务中依赖复杂特定上下文进行学习的能力不足问题，而传统的人工标注上下文成本高昂且难以扩展。针对公开高质量文档在预训练阶段已被大量使用、直接训练易导致模型记忆而非真正理解上下文的问题，本文提出一种基于小扰动的合成数据生成方法。其核心解决方案在于构建一个自动化流水线：首先对源文档进行重写以降低记忆风险；其次生成需基于文档进行推理的问题与评分标准；再以文档为上下文生成答案；最后仅保留真正依赖文档的样本。该方法无需人工标注，从3500篇文档中生成约1万条高质量样本，通过监督微调（SFT）和基于评分奖励的强化学习（RL）两阶段训练，使学生模型在CL-bench上的性能从13.7%提升至24.6%，接近参数量超过万亿的前沿模型表现。此外，该方法在长上下文理解、指令遵循和推理能力方面展现出广泛迁移效果，而代码生成与知识保持能力变化不大。本工作为提升大模型从上下文学习的能力提供了可复现、可扩展的新范式，推动面向上下文感知的推理研究发展。
链接: https://arxiv.org/abs/2609.33642
作者: Haoyi Wu,Yang Xiao,Yusong Sun,Wenyang Hui,Zhaokai Luo,Chengyue Jiang,Mu Chuan
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 18 pages, 3 figures, 10 tables
Abstract:Real-world tasks often require large language models (LLMs) to learn from complex task-specific context rather than pretrained parametric knowledge. This capability remains a weakness of LLMs, while human annotation for such task contexts is expensive and difficult to scale. Public high-quality documents are an abundant alternative, but much of the public web has already been consumed during pretraining: training on such documents naively would reward memorization rather than context learning. In this work, we attempt to make use of high-quality public documents with small perturbations and empirically find that LLMs can successfully generate context-dependent reasoning traces and answers, which are then used to train a student model. Specifically, we construct a synthesis pipeline that (i) rewrites source documents to reduce memorization risk, (ii) generates questions and rubrics that require reasoning over the document, (iii) answers the questions with the document as context, and (iv) admits only samples that genuinely depend on the document. Without any human annotators, our pipeline generates about 10k samples from 3.5k documents, and the resulting student model substantially improves the performance on CL-bench. SFT raises a Qwen3.6-35B-A3B student from 13.7% to 22.8%, and a subsequent rubric-reward RL stage reaches 24.6%, on CL-bench comparable with a frontier model of over a trillion parameters, Qwen3.8-2.4T (23.9%). We also observe a broad transfer of improvements to long-context understanding, instruction following, and reasoning, while code generation and knowledge remain mostly flat. We hope this work provides a reproducible and scalable way to improve the ability of LLMs to learn from context, and to facilitate further research on context-grounded reasoning.
[NLP-183] Quantifying Behavioral Tails in Black-Box Language Models
【速读】： 该论文旨在解决黑箱大语言模型（LLM）中严重行为发生概率难以准确估计的问题，尤其在面对罕见但高风险行为时，传统评估方法因采样效率低下而无法有效捕捉其概率分布。其核心挑战在于如何在输入空间中定义一个可计算且可复现的概率分布。为应对这一挑战，论文提出RareTrap框架，其关键创新在于引入一个代理模型（surrogate LLM），通过构建从低维潜在参考空间到令牌嵌入空间的几何感知映射，从而在输入提示（prompt）上诱导出显式且可重复的概率分布。同时，采用响应层面的性能函数对模型输出进行量化，以度量行为的严重程度，实现基于概率分布的渐进式稀有事件模拟。该方法能够在仅200次评估下成功诱发并估算严重资源消耗等极端行为的发生概率，显著提升评估效率。RareTrap为模型开发者提供了一种基于统一分布的系统性评估范式，有助于优先分配对齐资源以增强模型安全性与风险控制能力。
链接: https://arxiv.org/abs/2609.33638
作者: Elsayed Eshra,Ali Al-Lawati,Dongwon Lee,Suhang Wang
机构: The Pennsylvania State University (宾夕法尼亚州立大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Machine Learning (stat.ML)
备注:
Abstract:We introduce RareTrap, a framework for estimating the probability of severe behaviors in black box large language models (LLMs). A key challenge for probability estimation is defining a tractable distribution over the input space. To accomplish that, RareTrap uses a surrogate LLM and constructs a geometry-aware mapping from a lower-dimensional latent reference space into its token-embedding space to induce an explicit and reproducible distribution over input prompts. A response-level performance function is utilized on the response to quantify behavior severity. This enables sequential rare event simulation that concentrates evaluations on progressively more severe behaviors while preserving probability under the induced prompt distribution, which would otherwise be prohibitive to measure. Across 10 open-weight and two frontier models (GPT-5.4 and Claude Sonnet 4.6), we find that RareTrap successfully induces severe resource consumption behaviors and computes their probability with as few as 200 evaluations. RareTrap provides model developers a principled approach for evaluating language models under a common distribution, and prioritizing alignment effort to improve safety and mitigate risks. Code is published online: this https URL.
[NLP-184] Safety Reconstructed: Generative Modeling via Masked Diffusion Builds Strong Safety Guardrails
【速读】： 该论文旨在解决现有安全防护模型（Guard models）在训练目标上的局限性问题，即其仅依赖于从对话上下文中预测单一判别标记（verdict token），导致模型过度关注局部显著特征，产生过自信、对安全证据位置敏感但缺乏全局语境理解的缺陷。其核心解决方案是提出一种新的建模范式——LLaDA-Guard，不再直接预测标签，而是通过对比分析在不同标签假设下对提示（prompt）或响应（response）的重建质量差异来判断更合理的标签，从而将监督信号扩展至受监管区域中的每一个词元，迫使模型充分考虑完整内容而非仅依赖片段性线索。该方法基于掩码扩散语言模型（masked diffusion language model），采用LoRA微调技术对LLaDA-8B-Instruct进行类条件重构训练，无需修改基础模型架构。实验表明，相较于基于更强骨干网络的判别型基线模型，LLaDA-Guard在七个独立安全基准上平均排名领先，且具备更优的置信度校准性能（ECE 0.0875 vs. 0.1384）、更低的良性提示误判率以及更少的提示信息泄露现象；同时其生成式特性天然支持细粒度风险定位，可实现无需额外训练的不安全提示重写，平均转换成功率高达60.7%。
链接: https://arxiv.org/abs/2609.33634
作者: Gert Lek,Abele Malan,Chaoyi Zhu,Pin-Yu Chen,Robert Birke,Lydia Chen
机构: University of Neuchâtel(纳沙泰尔大学); Delft University of Technology(代尔夫特理工大学); IBM Research(IBM 研究院); University of Turin(都灵大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Guard models are the last line of defense between a language model and a harmful output, yet their training objective is surprisingly narrow. Existing guards learn to predict a single verdict token from a conversational context, concentrating supervision on a single target. The consequences are structural: models latch onto shortcut features, are overconfident, and remain sensitive to where safety evidence appears in the sequence rather than its role in the full context. We propose a different framing. Rather than predicting a label from text, our LLaDA-Guard asks which label better explains the text: scoring the prompt or response under each label hypothesis and classifying based on their difference. This shifts supervision to every token in the moderated region, forcing the model to account for full content rather than its most discriminative fragments. We instantiate this idea with a masked diffusion language model, fine-tuning LLaDA-8B-Instruct with a class-conditional reconstruction objective using LoRA and requiring no architectural changes beyond the base model. LLaDA-Guard leads on average rank against discriminative baselines trained on stronger backbones across seven held-out safety benchmarks, while exhibiting substantially better confidence calibration (ECE 0.0875 vs. 0.1384 for Qwen3Guard), less over-defense on benign prompts with unsafe-looking cues, and less prompt leakage when moderating responses. Its generative nature further enables token-level risk localization as a natural byproduct, yielding a pipeline for rewriting unsafe prompts into safe equivalents without additional training and achieving a 60.7% average conversion-to-safe rate.
[NLP-185] ParaAgent : Reinforcing Parallel Acting in Open-World Tool Environments
【速读】： 该论文旨在解决开放世界工具环境中语言模型智能体在探索未知能力与利用已知能力之间难以平衡的问题。现有方法要么僵化地分离探索与执行阶段，要么无协调地交替进行，导致性能与效率之间的权衡。其核心解决方案在于提出一种结构化的并行动作循环（ParaAct），通过在阶段层面实现探索与执行的协同，并在动作层面引入并行性，从而实现跨粒度的高效协调。ParaAgent采用多智能体冷启动示范与多层次优势解耦下的强化学习框架，显式建模规划结构，并结合步骤级、阶段级和轨迹级奖励进行监督优化。该方法依托于包含50,011个真实工具接口的可扩展仿真环境ToolEnv进行训练。在两个开放世界工具基准测试中，ParaAgent-4B在所有基线方法中实现了最佳平均成功率，尤其在多工具任务上表现显著优于GPT-4.1系统。行为分析表明，性能提升主要源于该动作组织机制的有效性，验证了其对构建高效且具备强适应性的开放世界智能体的关键作用。
链接: https://arxiv.org/abs/2609.33618
作者: Shengbin Yue,Hongru Wang,Siyuan Wang,Xiaoxin Chen,Wei Chen,Zhongyu Wei
机构: Fudan University(复旦大学); University of Edinburgh(爱丁堡大学); Chinese University of Hong Kong(香港中文大学); Vivo AI(维沃人工智能); Huazhong University of Science and Technology(华中科技大学); Shanghai Innovation Institute(上海创新研究院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Language model agents are increasingly deployed in open-world tool environments, which require balancing exploring unknown capabilities and exploiting known ones. Existing methods face a performance-efficiency tradeoff: they either rigidly decouple exploration and execution or interleave them without coordination. We argue that the key lies not in whether to decouple or interleave them, but in how to coordinate them across granularities. We introduce ParaAct, a structured parallel-action loop that combines phase-level Exploration \rightleftharpoons Execution with action-level parallelism. To learn this loop, ParaAgent combines multi-agent cold-start demonstrations with reinforcement learning under multi-level advantage decoupling, making planning structure explicit and supervising it with step-, phase-, and trajectory-level rewards. Learning is supported by our ToolEnv, a scalable simulator grounded in 50,011 realistic tool interfaces. On two open-world tool benchmarks, ParaAgent-4B achieves the best average success among all baselines, including GPT-4.1 systems, with the largest gains on multi-tool tasks. Behavioral analyses show that these gains stem from this action organization, highlighting its importance for capable and efficient open-world agents.
[NLP-186] Quizzing the Translation: A Prover-Grounded Evaluation Metric for NLrightarrowFOL EMNLP2026
【速读】： 该论文旨在解决自然语言问题中符号推理的翻译质量评估难题，尤其针对将自然语言问题转化为一阶逻辑（First-Order Logic, FOL）过程中因语义偏差导致的错误传播问题。现有评估指标如BLEU、BERTScore和Smatch++由于过度依赖表面词汇重叠，往往对严重语义错误（如“every”误译为“some”）的翻译给予较高评分，从而误导模型优化方向。本文提出SIV（Semantic Integrity Verification），其核心创新在于从目标公式生成两类语义探针：正向探针（positive probes）要求候选翻译必须蕴含的语句，用于检测内容缺失；对比探针（contrastive probes）要求候选翻译不应蕴含的语句，用于检测过度推断。通过调用定理证明器验证候选翻译与探针的一致性，SIV能够精准识别不同类型的语义错误。在受控扰动数据集上，错误严重程度解释了80%的SIV得分方差，显著优于以往指标（最高仅17%）；在六个独立错误类别上的测试中，SIV在超过99%的对比对中正确识别参考翻译优于扰动翻译。此外，每个探针均标注具体测试内容，使失败模式可追溯至特定错误类型，实现宏平均F1达0.638，接近基线方法的两倍。在434个专家标注的真实大语言模型（LLM）翻译样本上，SIV取得最优AUC，能唯一识别并量化专家标记的重大错误，并对未登录词翻译主动拒答而非错误评分，展现出卓越的可靠性与可解释性。
链接: https://arxiv.org/abs/2609.33612
作者: Pu Suo,Ali Emami
机构: Emory University(埃默里大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO)
备注: Accepted to EMNLP 2026 (Main Conference). 18 pages, 3 figures, 10 tables. Code and data: this https URL
Abstract:A standard pipeline for symbolic reasoning over natural-language problems translates them into first-order logic and invokes a theorem prover. The translation step is the bottleneck: swap “every” for “some” and every inference that follows is corrupted. Yet today’s metrics often score more broken translations higher than less broken ones, because BLEU, BERTScore, and Smatch++ reward surface overlap that the worst errors happen to preserve. We introduce SIV, which derives two kinds of probes from the target formula and uses a theorem prover to verify the candidate translation against each. Positive probes are statements the candidate must entail, which detect translations that drop content; contrastive probes are statements the candidate must not entail, which detect translations that assert more than the original. On a controlled pool of perturbed FOLIO translations, the severity of the error accounts for 80% of SIV’s score variance, compared with at most 17% for any prior metric. Across six error classes on a disjoint pool, SIV scores the reference above the perturbed candidate in over 99% of pairs. Because each probe is labeled with what it tests, the failure pattern also supplies a labeled error trace, recovering the perturbation class at macro-F1 0.638, nearly double the score-only baseline. On 434 expert-audited real LLM translations, SIV attains the top AUC, uniquely detects and grades expert-labeled major errors, and abstains, rather than mis-scoring, on out-of-vocabulary translations.
[NLP-187] GRL: Temperature-Grouped Reinforcement Learning for Efficient Exploration in LLM s NEURIPS2026
【速读】： 该论文旨在解决强化学习中可验证奖励（Reinforcement Learning with Verifiable Rewards, RLVR）场景下高效探索的瓶颈问题。现有方法如温度控制和测试时缩放虽能提升大语言模型（Large Language Models, LLMs）的轨迹多样性，但通常会增加采样预算或无法量化探索收益。为此，本文提出温度分组强化学习（Temperature-Grouped Reinforcement Learning, TGRL），其核心创新在于将温度诱导的多样性转化为显式的训练信号。TGRL针对每个提示（prompt）将采样轨迹划分为低温与高温子集，通过二者间的奖励差异估计探索增益，并利用相同对数概率在不同温度下生成的下一个词分布之间的Jensen-Shannon（JS）散度，将该群体级信号分配至具体词元（token）层面作为信用分配。这一机制实现了无需扩大采样预算的前提下，相比强基线方法在多项任务上显著加速收敛（最高快36%），并在11个跨领域基准测试中广泛超越现有方法，包括数学推理、代码生成及复杂环境交互任务，验证了其有效性与普适性。
链接: https://arxiv.org/abs/2609.33589
作者: Zihan Lin,Xiaohan Wang,Jie Cao,Jiajun Chai,Wei Lin,Guojun Yin,Ran He
机构: University of Chinese Academy of Sciences(中国科学院大学); Meituan(美团); MAISNLPR, Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所智能信息处理国家重点实验室)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Accepted as NeurIPS2026 Poster
Abstract:Efficient exploration often remains a central bottleneck in reinforcement learning with verifiable rewards (RLVR). Although temperature control and test-time scaling strategies can increase rollout diversity of large language models (LLMs), they either expand the sample budget at rollout time or leave the benefit of exploration unquantified. To this end, we propose Temperature-Grouped Reinforcement Learning (TGRL), which turns temperature-induced diversity into an explicit training signal. For each prompt, TGRL partitions its rollout group into low- and high-temperature subsets, estimates exploration gain through their reward contrast, and allocates this group-level signal as token-level credit using Jensen–Shannon (JS) divergence between the corresponding temperature-scaled next-token distributions induced by the same logits. Notably, TGRL reaches equivalent accuracy up to 36% faster than strong RLVR baselines without expanding the rollout budget. Across 11 benchmarks from diverse domains, TGRL broadly improves over strong RLVR baselines: it improves the six-benchmark math average by 1.6% at 32B, raises CodeForces rating by 196.7 points and LiveCodeBench Pass@16 by 4.4%, and improves ALFWorld/WebShop success rates by 6.3%/4.9%. Comprehensive ablations and wall-clock analysis confirm the efficacy of all proposed components. Code is available at this https URL.
[NLP-188] Jev Matches 7B Language Models for Speech-Neuroprosthesis Rescoring
【速读】： 该论文旨在解决脑机接口中语音神经假体（speech neuroprosthesis）在解码尝试说话内容时，依赖大规模语言模型进行候选句重评分（rescoring）所导致的高计算成本与硬件门槛问题。其核心挑战在于：现有通用语言模型在面对从列表中选择最优句子的任务时，会错误地依据标签在列表中的位置而非句子语义本身进行决策，从而影响性能。为解决这一问题，论文提出将重评分建模为单一类型决策任务（single typed decision），采用Jev——一个专为校准决策训练的托管模型，通过一次调用返回每个候选句的概率分布，并将其与解码器自身得分结合。实验结果表明，在肌萎缩侧索硬化症（ALS）患者数据集上，尽管仅使用低成本、无需GPU的Jev模型，其词错误率（WER）达到7.5%（未调参），优于OPT-6.7b和Qwen2.5-7B的7.8%，且在解码器权重微调后进一步降至6.9%，显著优于两者（7.2%和7.4%）。此外，Jev每千句仅需0.07美元，且无需专用GPU；即便使用70亿参数模型的本地GPU，也只有在利用率超过43%时才更经济，远超单用户实际生成量。端到端网络延迟为262毫秒，其中服务端处理占62毫秒，与本地70亿参数模型的27毫秒延迟处于同一数量级，具备实用潜力。因此，该方案的关键创新在于通过校准的决策式重评分框架，实现了高性能与低资源消耗的平衡。
链接: https://arxiv.org/abs/2609.33538
作者: Gabriele Cinà
机构: 未知
类目: Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
备注: 8 pages, 1 figure, 4 tables. Code and data: this https URL
Abstract:A speech neuroprosthesis decodes attempted speech from brain activity and ends by rescoring the decoder’s candidate sentences with a language model of several billion parameters, the only component that needs a GPU. Replacing that model with a cheaper one is hard: general language models asked to pick one sentence from a list answer from where a label sits in the list rather than from the sentence itself. We pose rescoring as a single typed decision, one call that returns a probability for every candidate, served by Jev, a hosted model trained for calibrated decisions, and combine it with the decoder’s own score. On 978 held-out sentences from a participant with ALS, where the published decoder alone reaches 8.1% word error, Jev reaches 7.5% against 7.8% for both OPT-6.7b and Qwen2.5-7B; with the decoder’s weight re-tuned, 6.9% against 7.2% and 7.4%. Jev is ahead in all four comparisons and at most 0.2 points behind at the 95% bound. It costs 0.07 USD per thousand sentences and needs no GPU; a dedicated GPU running a 7B model is cheaper per sentence only above 43% utilisation, far beyond what one user generates. End-to-end latency over the internet is 262 ms, of which 62 ms is spent at the provider, the same order as a 7B model on a local GPU (27 ms) but not faster.
[NLP-189] ManiEdit: Sequential Unstructured Knowledge Editing for Language Models from a Manifold Perspective
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在持续知识更新过程中面临的两大核心挑战：一是对非结构化长文本知识进行连续编辑时易出现的“编辑遗忘”问题，二是编辑过程导致模型通用能力退化。现有方法难以在不破坏原有知识结构的前提下实现高效、精准的增量式知识修正。为此，论文提出从流形（Manifold）视角重新审视知识编辑问题，将知识更新建模为全局知识流形中局部编辑子流形的位移过程。其解决方案的关键在于：首先通过“枢轴定位（Pivot Localization）”机制识别具有高影响力的关键编辑锚点，以有效锚定编辑子流形；其次引入“流形感知保持（Manifold-Aware Preservation）”机制，结合能量加权惩罚与递归零空间对齐，实现对不同类型知识的差异化保护。该框架在两个基础大模型和四个非结构化编辑基准上的实验表明，ManiEdit在性能上达到当前最优水平，相较于最强基线提升最高达+27.81 BERTScore和+8.50 ROUGE-L，同时在六个代表性下游任务中保持接近原始模型的通用能力。
链接: https://arxiv.org/abs/2609.33534
作者: Rui Liu,Chenheng Zhang,Haoxuan Li,Zhouchen Lin
机构: Peking University (北京大学); State Key Lab of General Artificial Intelligence (通用人工智能国家重点实验室); Institute for Artificial Intelligence (人工智能研究院)
类目: Computation and Language (cs.CL)
备注: 31 pages, 8 figures
Abstract:Large language models (LLMs) inevitably generate some incorrect or outdated content, necessitating efficient and precise mechanisms for continual knowledge updates. However, existing model editing methods struggle to sequentially edit unstructured long-form knowledge, suffering from severe edit forgetting and degradation of general capabilities. To address these challenges, we reframe knowledge editing from a manifold perspective, viewing it as a localized displacement of an edit sub-manifold within the global knowledge manifold. Under this formulation, the problem can be decomposed into two key questions: (i) how to identify representative edit points that effectively anchor the edit sub-manifold, and (ii) how to preserve the remaining manifold structure during the sub-manifold displacement process. Based on this perspective, we propose ManiEdit, a novel manifold-aware autoregressive editing framework consisting of two core components. Pivot Localization addresses the mediocre-point dilemma by identifying high-leverage pivots to anchor the edit sub-manifold. Manifold-Aware Preservation preserves different knowledge types through an energy-weighted penalty combined with recursive null-space alignment. Experiments on two base LLMs and four unstructured editing benchmarks demonstrate that ManiEdit achieves state-of-the-art performance, outperforming the strongest baseline by up to +27.81 BERTScore and +8.50 ROUGE-L, while maintaining near-original general capabilities across six representative downstream tasks. Our code is available at: this https URL
[NLP-190] DISCO: Distributed Long Context Scaling with Grounding-Reasoning Disaggregation
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在处理长上下文时出现的“上下文退化”（context rot）问题，即随着输入长度增加，模型推理质量显著下降的现象。其核心原因在于单体架构中上下文感知与复杂推理任务之间的结构纠缠，导致模型在海量上下文信息中进行定位时耗尽了用于高级推理的表征能力。为应对这一挑战，论文提出了一种基于分布式长上下文扩展的解耦框架——DISCO（DIStributed long COntext scaling）。其关键创新在于将“接地”（grounding）与“推理”（reasoning）功能分离：通过一组专用的Worker LLM并行执行局部、分布式的上下文信息提取任务，而一个由强化学习（GRPO）优化的中央驱动模型（Driver LLM）负责动态规划任务分解与证据整合。这种解耦机制有效隔离了原始上下文噪声对推理过程的干扰，从而从根本上消除上下文退化。实验表明，在100万词符的RULER-QA基准上，DISCO保持78.4%的准确率，显著优于传统基线；在LongBench v2上性能领先全上下文模型达9.8个百分点，并以低于80%的推理成本达到如Gemini-3-Pro-Preview等前沿模型的水平，实现了高效且鲁棒的长上下文推理新范式。
链接: https://arxiv.org/abs/2609.33485
作者: Guanzheng Chen,Viet Dac Lai,Subhojyoti Mukherjee,Branislav Kveton,Seunghyun Yoon,Franck Dernoncourt,Qizhe Xie,Trung Bui
机构: National University of Singapore(新加坡国立大学); Adobe Research(Adobe 研究院)
类目: Computation and Language (cs.CL)
备注:
Abstract:While Large Language Models (LLMs) advertise million-token context windows, reasoning quality often collapses as inputs grow – a phenomenon termed context rot. This failure stems from a structural entanglement in monolithic architectures, where the massive search burden of contextual grounding exhausts the representational capacity needed for complex reasoning. To resolve this, we propose Grounding-Reasoning Disaggregation via DIStributed long COntext scaling (DISCO). Inspired by distributed computing frameworks like Apache Spark, DISCO partitions long context across a fleet of Worker LLMs dedicated exclusively to parallel, localized grounding. A central Driver LLM, trained via Reinforcement Learning (GRPO) to optimize planning, orchestrates execution by dynamically mapping queries into atomic extraction tasks and reducing the gathered evidence to synthesize a final answer. By isolating reasoning from raw context noise, DISCO effectively eliminates context rot. On RULER-QA (1M tokens), it maintains 78.4% accuracy where standard baselines collapse. Furthermore, it outperforms full-context models by up to 9.8 points on LongBench v2 and matches frontier models like Gemini-3-Pro-Preview while reducing inference costs by over 80%, establishing a highly efficient paradigm for robust long-context inference.
[NLP-191] A Cheap Verifier is Good Enough: LLM Post-training is Robust to Erroneous Rewards
【速读】： 该论文旨在解决在使用半可验证奖励（semi-verifiable rewards）对大语言模型进行后训练时，如何有效选择验证器（verifier）以最大化模型性能的问题。当前实践中涉及多个关键变量，如训练步数、基座模型规模、训练顺序、数据质量及验证器准确性等，但尚不清楚验证器一致性（verifier agreement）是否能可靠预测后训练后的实际表现。研究通过超过11,000小时H100 GPU计算资源，在医疗、法律和金融领域的HealthBench与PRBench任务上系统评估了Qwen3系列模型（1.7B–8B参数规模）的表现。结果表明，尽管采用了前沿的开放权重模型作为“黄金验证器”（golden verifiers），更高的验证器一致性并未始终对应最佳训练效果；昂贵的验证器未必优于低成本方案，且开源的Gemma验证器同样可实现优异训练结果。研究进一步对比了两种低成本验证策略——成本削减型与平衡型——相较于黄金评分协议，可实现98.8%–99.7%的评阅成本降低，同时平均仅造成1–3分的后训练得分差距。然而，个体任务中仍存在显著性能损失，说明验证器选择并非可互换，其效果依赖于具体场景。因此，解决方案的关键在于：通过实证分析揭示验证器一致性与最终性能之间的非线性关系，并证明在特定条件下，低成本、高效率的验证器配置可在保证近似最优性能的前提下大幅降低训练成本。
链接: https://arxiv.org/abs/2609.33467
作者: Andreas Plesner,Curtis Northcutt,Francisco Guzmán,Anish Athalye
机构: ETH Zurich (苏黎世联邦理工学院); Handshake AI
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 32 pages, 7 figures, 17 tables
Abstract:When post-training large language models on tasks with semi-verifiable rewards, there are many factors (training steps, base model size, training order, data quality, verifier accuracy, etc.) that practitioners must contend with to maximize model performance. Yet, it remains unclear how well verifier agreement predicts post-training performance on such tasks. In this paper, we explore this question with over 11k H100 GPU-hours, across HealthBench and PRBench tasks in medical, legal, and finance domains. Across the tested domains, Qwen3 trainees (1.7B-8B on HealthBench; 8B on PRBench), evaluation splits, and frontier LLM reference judges (which we call golden verifiers), higher verifier agreement does not consistently identify the best training verifier. Expensive verifiers need not outperform inexpensive ones, and open-weight Gemma verifiers produce strong training outcomes. We compare two low-cost choices retrospectively – a cost-reducing choice and a balanced choice – with estimated grading cost reductions of 98.8%-99.7% relative to the golden grading protocols and average post-training score gaps of 1-3 points from the best evaluated training verifier. These averages include larger losses in individual settings; they do not establish that verifier choices are interchangeable.
[NLP-192] GSM: Efficient Language Modeling with Shared Global State
【速读】： 该论文旨在解决高效语言模型在处理长序列时面临的双重挑战：一是单次访问历史上下文的计算开销过高，二是跨层重复选择和处理历史信息带来的额外计算与存储负担。其核心解决方案是提出一种因果编码器-解码器架构——全局状态模型（Global State Model, GSM），关键在于将长程信息的选择与聚合过程集中于编码阶段。通过多阶段的历史信息检索，编码器逐步将长程依赖融入近期位置的表示中，构建一个固定窗口大小的共享全局状态。解码器各层通过自上而下更新的查询访问该统一状态，既保持了模型的计算深度，又避免了重复构建历史键值（Key-Value, KV）对及长程索引操作。因此，解码阶段的每步注意力计算复杂度和KV缓存大小均不随历史长度增长，显著提升了计算效率并降低了缓存开销，同时维持了模型性能与对长程信息的有效利用能力，为高效语言建模提供了一种高效的共享状态架构。
链接: https://arxiv.org/abs/2609.33465
作者: Yunao Zheng,Bin Wen,Xiaojie Wang,Kaiyu Jiang,Xuanyu Zheng,Changyi Liu,Hongyi Fu,Jianxiong Wang,Tianke Zhang,Haonan Fan,Yingxin Li,Jiankang Chen,Xu Wang,Tingting Gao,Han Li
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Efficient language models must reduce not only the cost of individual accesses to past context but also the overhead of repeatedly selecting and processing historical information across layers. We introduce the Global State Model (GSM), a causal encoder–decoder architecture that concentrates the selection and aggregation of long-range information in the encoding stage. Through multiple stages of history retrieval, the encoder progressively incorporates long-range information into representations at recent positions, forming a shared state with a fixed window size. Each decoder layer accesses this same state using queries updated from the preceding layer, preserving computational depth while avoiding repeated construction of historical key–value (KV) representations and long-range indexing. As a result, neither the decoder’s per-step attention cost nor its KV cache size grows with the history length. Experiments show that GSM improves computational efficiency and reduces cache overhead while maintaining model performance and the ability to use long-range information, offering a shared-state architecture for efficient language modeling.
[NLP-193] Rethinking Token Reweighting for SFT: Suppress Reverse and Extrapolate Learned Features
【速读】： 该论文旨在解决监督微调（Supervised Fine-Tuning, SFT）过程中因过度依赖模型自身判断低概率词元而导致的噪声放大、冲突监督加剧以及预训练知识被错误覆盖的问题。现有基于重加权的SFT方法仅能对示范词元施加非负系数，从而实现特征的抑制或增强，但无法逆转已学习到的有害特征；同时，增大训练权重并不能实现特征外推，因其仅改变优化轨迹而非沿固定方向扩展。为此，论文提出一种名为SCALE（Selective Control of Adaptation via Local Entropy）的方法，其核心在于构建一个由固定SFT增量（SFT delta）定义的稳定参考框架，并通过最小化预测熵来学习有界、词元与模块特定的门控机制，使模型能够根据熵减的对齐程度选择性地抑制、反转或外推冻结的SFT特征。实验表明，SCALE在Qwen2.5-Math-1.5B、Qwen2.5-Math-7B和Qwen3-4B-Base等多个数学推理任务上均显著超越基线，同时在代码生成任务（HumanEval、HumanEval+、MBPP）上也取得最优表现，验证了通过控制已有残差特征的使用方式而非单纯调整学习过程，可更有效地实现SFT修正。
链接: https://arxiv.org/abs/2609.33463
作者: Cunchun Li,Haonan He,Yifan Gao,Minglei Li,Jingqi Ye,Qingyu Yang,Peng Ye
机构: Shanghai AI Laboratory; University of Science and Technology of China; Fudan University; KTH Royal Institute of Technology; The Chinese University of Hong Kong
类目: Computation and Language (cs.CL)
备注: Preprint, Under Review
Abstract:Supervised fine-tuning (SFT) learns most aggressively from tokens that the model deems least likely. This helps acquire new behaviors, but also amplifies noisy or conflicting supervision and can overwrite useful pretrained knowledge. Through a unified policy-loss view, we revisit existing token-reweighting methods and show that they assign nonnegative coefficients to demonstrated tokens. Consequently, they can suppress or amplify supervised updates, but cannot reverse harmful features once learned. Moreover, larger training weights do not amount to feature extrapolation, since they change the optimization trajectory rather than scale a fixed SFT direction. We argue that reversal and extrapolation require a stable reference frame defined by a fixed SFT delta. Motivated by this, we propose SCALE (Selective Control of Adaptation via Local Entropy), an entropy-guided adaptation-strength-control method that freezes the pretrained model and the SFT delta and learns bounded token- and module-specific gates by minimizing predictive entropy alone. These gates suppress, reverse, or extrapolate frozen SFT features according to their alignment with entropy reduction. Across Qwen2.5-Math-1.5B, Qwen2.5-Math-7B, and Qwen3-4B-Base, SCALE achieves mathematical-reasoning averages of 37.84, 43.60, and 36.57, exceeding the strongest corresponding baselines while remaining competitive on general-retention benchmarks. It also attains the best average code-generation performance across HumanEval, HumanEval+, and MBPP for all three models. These results suggest that effective SFT correction can benefit from controlling how already learned residuals are used, rather than only modifying how they are learned.
[NLP-194] Context Spanning: A Communication Framework for Full-Duplex Speech Models and External LLM Backends ICASSP2027
【速读】： 该论文旨在解决全双工语音对话模型在实时对话中无法有效获取和利用外部实时信息的问题。现有模型普遍依赖参数化知识，难以访问动态更新的外部数据，且即使使用大语言模型（LLM）进行信息检索，多数系统仍通过压缩的隐空间处理检索结果，导致信息失真与损失。其解决方案的关键在于提出“上下文延展”（Context Spanning）框架，通过实时分块预填充（chunked prefill）机制，将外部LLM检索到的原始文本信息直接注入全双工语音模型，在单次前向传播中完成编码，并在实时帧预算内实现信息的无缝融合。该方法保留了信息的原始文本形式，使语音模型能够基于完整、未压缩的上下文独立推理并生成响应，显著提升了全双工对话性能与问答任务表现，为全双工系统提供了一种简单而高效的外部信息接入新范式。
链接: https://arxiv.org/abs/2609.33443
作者: Seonghyeon Go,Yongwoo Kim,Hyeonjin Cha,Jaeho Shin
机构: 未知
类目: Computation and Language (cs.CL); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注: Submit to ICASSP 2027
Abstract:Full-duplex spoken dialogue models can listen and speak simultaneously like the real-time dynamics of human conversation. For natural dialogue, the ability to search for external information in real-time is also an important capability. Many models remain trapped in parametric knowledge, leaving them unable to access real-time information and tool execution. Furthermore, even when Large Language Models (LLM) retrieve information, many duplex speech models process it within a compressed latent space rather than in its raw text form, which can lead to information loss from compression. To address this issue, we propose Context Spanning, a framework for information injection between a full-duplex speech model and an external LLM backend via real-time chunked prefill. The injected frame is encoded in a single forward pass inside the real-time frame budget. It feeds the retrieved information to the speech model as-is, enabling it to reason over the information independently and generate responses. With this approach, our model achieves high performance on Full-Duplex benchmarks and strong results on Question Answering tasks, demonstrating its conversation potential. Context Spanning shows that external information can be injected directly into a duplex speech model, introducing a new simple and powerful mechanism for duplex systems.
[NLP-195] MIC: Explaining Image-Claim Inconsistencies in AI-Generated Multimodal Misinformation
【速读】： 该论文旨在解决生成式人工智能（Generative AI）时代下，图文配对形式的虚假信息（尤其是基于上下文不一致的误导性内容）难以通过传统自动事实核查（AFC）方法有效识别的问题。现有方法多聚焦于图像的生成痕迹检测，将其视为溯源问题，而忽视了人类事实核查员核心关注的内容一致性验证——即图像内容是否与伴随文本声明所暗示的语境相符。为此，论文提出MIC（Multimodal Inconsistency Checking）框架，其关键在于融合监督微调（SFT）与组相对策略优化（GRPO），直接优化可验证的组件级奖励信号，实现对事实判断、不一致类型分类、视觉证据描述及世界知识解释的联合建模。MIC通过引入MIC-Bench基准数据集（包含8,812个图像-声明实例），系统评估了在分布内与分布外场景下的性能表现，结果显示相较于仅使用SFT，GRPO在宏观F1得分上分别提升4.67和4.11点，并显著提升了视觉证据描述与世界知识解释的语义相似度，从而更有效地辅助人工核查。
链接: https://arxiv.org/abs/2609.33441
作者: Ruihong Zeng,Jonathan Tonglet,Preslav Nakov,Iryna Gurevych
机构: KU Leuven (比利时鲁汶大学); TU Darmstadt (德国达姆施塔特工业大学); National Research Center for Applied Cybersecurity ATHENE (应用网络安全国家研究中心 ATHENE); Mohamed bin Zayed University of Artificial Intelligence (阿联酋穆罕默德·本·扎耶德人工智能大学)
类目: Computation and Language (cs.CL)
备注: Preprint under review
Abstract:Claims paired with AI-generated images are a rapidly growing form of misinformation. Existing automated fact-checking (AFC) methods mainly treat this as a provenance problem, detecting low-level synthesis artifacts to decide whether an image is AI-generated. However, such methods do not verify what human fact-checkers often check: whether an image’s content is consistent with the context implied by its accompanying claim. To address this gap, we introduce MIC (Multimodal Inconsistency Checking), an AFC framework that assists human fact-checkers by detecting AI-generated multimodal misinformation and explaining inconsistencies using world knowledge. MIC first uses supervised fine-tuning (SFT) for task adaptation and then applies Group Relative Policy Optimization (GRPO) to directly optimize component-level verifiable rewards for verdict prediction, inconsistency type classification, visual evidence description, and world-knowledge explanation. We further introduce MIC-Bench, a benchmark comprising 8,812 image-claim instances derived from 4,406 claims, where each claim is paired with an authentic image and an AI-generated counterpart that introduces a controlled contextual inconsistency. Compared with SFT alone, GRPO further improves Macro-F1 by 4.67 and 4.11 points in the in-distribution and out-of-distribution settings, respectively, while also improving the semantic similarity of visual evidence descriptions and world-knowledge explanations to reference annotations. Our code and data are available at this https URL.
[NLP-196] SMAT: Simple and Efficient Merge-Aware Training
【速读】： 该论文旨在解决模型合并（model merging）中专家模型在独立训练后合并性能不佳的问题。标准的专家训练仅优化任务损失，未考虑合并后的表现，导致合并后性能下降。现有融合感知训练（Merge-aware Training, MAT）方法虽能提升合并效果，但往往未能充分建模常见的合并操作，且引入较高的训练开销。本文提出SMAT（Simple MAT），从专家视角出发，将典型合并操作抽象为三种可解释的操作：缩放（Scale，重加权自身更新）、掩码（Mask，移除选定坐标）和扰动（Perturb，引入其他专家的更新）。基于此，SMAT通过采样缩放系数、掩码和加性噪声，在模拟合并参数下联合优化专家损失与期望损失，从而显式增强合并兼容性。为进一步提升效率，SMAT引入周期调度、核融合（kernel fusion）及参数存储切换机制，实现每步仅需一次前向和一次反向传播。在四种语言与视觉-语言骨干网络上，SMAT相较各骨干最强基线，在五种合并方法上的平均得分提升1.07至2.16分，且训练时间开销低于标准微调的2%。其核心创新在于以简洁高效的框架建模合并操作本质，实现性能与效率的双重优化。
链接: https://arxiv.org/abs/2609.33437
作者: Yanggan Gu,Yuanyi Wang,Zhen Li,Shuo Cai,Yuhang Liu,Junzhuo Li,Zihao Wang,Hongxia Yang
机构: The Hong Kong Polytechnic University (PolyU); The Hong Kong University of Science and Technology (Guangzhou); The Chinese University of Hong Kong; PolyU-Daya Bay Technology and Innovation Research Institute; InfiX.ai
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 19 pages, 6 figures, 7 tables. Code: this https URL
Abstract:Model merging integrates the capabilities of multiple experts without joint retraining, but standard expert training optimizes task loss alone and does not guarantee good performance after merging. Merge-aware training (MAT) aims to improve merged performance, but existing methods do not fully account for common merging operations and add training cost. We observe that, from an expert’s perspective, common merging methods can be described by three operations: Scale reweights its own update, Mask removes selected coordinates, and Perturb adds updates from other experts. Based on this view, we introduce SMAT (Simple MAT), which jointly optimizes expert loss and expected loss at simulated merged parameters generated by sampling scaling coefficients, masks, and additive noise. We further introduce periodic scheduling, kernel fusion, and parameter storage switching to make SMAT efficient, with one forward and one backward pass per step. Across four language and vision-language backbones, SMAT improves the mean score across five merging methods by 1.07-2.16 points over the strongest baseline for each backbone, with less than 2% training-time overhead over standard fine-tuning.
[NLP-197] CORA: A Protocol for Diagnosing Boundary Robustness in Text-to-Audio Retrieval under Query Reformulations AACL
【速读】： 该论文旨在解决文本到音频（Text-to-Audio, T2A）检索系统在评估过程中存在的局限性问题，即当前评估普遍依赖于句式固定的描述性查询（caption-style queries），而实际用户意图在表达形式上具有多样性，导致现有评估方法难以揭示模型在面对不同语义等价但句式变化的查询时的性能退化。为应对这一挑战，论文提出了一种名为CORR（Caption-Offset Retrieval for Audio）的锚定诊断协议，其核心是将原始标注文本重写为五种保持语义意图一致但表达形式不同的变体（命令、疑问、间接表述、关键词短语、陈述句），同时固定目标音频，从而实现对同一目标在多种查询形式下的检索表现追踪。该方法引入了RankDrop作为新指标，用于量化因查询重构导致的检索排名下降程度，揭示了传统Recall@k无法捕捉的潜在失败模式。实验结果表明，RankDrop与查询在原始文本空间中的移动距离关联性极弱（r=0.084），却与目标对齐损失（Target Alignment Loss）和目标边界优势退化（Target Boundary Margin Degradation）呈显著正相关（r=0.508 和 r=0.615），且在OEA检索器中同样表现出边界退化比查询形式变化更具解释力（r=0.472/0.478 vs r=0.084/0.046）。因此，解决方案的关键在于：构建一个能感知查询形式多样性并精准衡量目标边界稳定性变化的评估框架，并强调鲁棒的T2A检索必须确保目标音频在语义重构下仍保持相对于竞争音频的边界优势。
链接: https://arxiv.org/abs/2609.33433
作者: Jae Min Woo,Kyongmin Kong,Bogyung Jeong,Minjeong Kim,HaeJun Yoo,Du-Seong Chang
机构: Sogang University (首尔女子大学)
类目: ound (cs.SD); Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
备注: Accepted to Findings of IJCNLP-AACL. Code and data: this https URL
Abstract:Text-to-Audio (T2A) retrievers are typically evaluated with caption style queries, but the same user intent can be expressed in many forms. We introduce CORA (Caption-Offset Retrieval for Audio), a caption anchored diagnostic protocol that rewrites each source caption into five intent preserving forms (Command, Question, Indirect, Key phrase, and Statement) while fixing the target audio. By tracking the same target across query forms, CORA defines RankDrop, a metric revealing failures hidden by Recall@k. Using Pearson’s correlation coefficient r, we find that RankDrop is weakly associated with raw text space movement (r=0.084), but strongly associated with Target Alignment Loss and Target Boundary Margin Degradation (r=0.508 and r=0.615). The same pattern appears in OEA retrievers, where RankDrop is better explained by boundary degradation (r=0.472/0.478) than by query movement (r=0.084/0.046). Overall, these results suggest that robust T2A retrieval requires preserving the target’s boundary advantage over competing audio under reformulation.
[NLP-198] acherGRPO: Closing the Capacity Gap in Reasoning Distillation via Teacher Alignment
【速读】： 该论文旨在解决生成式 AI（Generative AI）中教师模型（teacher model）向学生模型（student model）进行推理蒸馏时面临的“差距诅咒”（Gap Curse）问题：随着教师模型能力增强，其输出分布逐渐偏离学生模型可建模的范围，导致学生性能下降。现有方法通过数据筛选或引入次优中间助手模型缓解该问题，但均以牺牲监督覆盖范围或质量为代价。本文提出教师对齐（Teacher Alignment）策略，直接将教师模型适配至学生模型的分布，避免数据丢弃与推理质量退化。然而，简单的知识蒸馏方式会导致教师推理能力的灾难性崩溃。为此，作者将教师对齐重构为强化学习问题，并提出基于分组相对策略优化（Group Relative Policy Optimization, GRPO）的TeacherGRPO框架，包含两项关键创新：(i) 课程式选择性对齐（Curriculum Selective Alignment），通过词元级与分布级双重课程机制，聚焦于高信息量的推理差距，同时过滤低信号词元和不确定尾部分布的噪声；(ii) 重要性自适应长度正则化（Importance-Adaptive Length Regularization），选择性惩罚冗余表达，同时保留具有教学意义的关键推理步骤。经对齐后的教师模型通过标准蒸馏流程指导学生模型，实验表明TeacherGRPO在多个推理基准和蒸馏方法上显著优于基线。
链接: https://arxiv.org/abs/2609.33426
作者: Zhenyu Lei,Zihan Chen,Yaochen Zhu,Shangbin Feng,Zaiyi Zheng,Ruocheng Guo,Yushun Dong,Jundong Li
机构: University of Virginia; Netflix; University of Washington; Florida State University; Microsoft
类目: Computation and Language (cs.CL)
备注:
Abstract:Reasoning distillation from powerful teacher models to smaller students faces the Gap Curse: as teachers grow more sophisticated, their complex distributions increasingly diverge from what students can approximate, causing performance degradation. Existing mitigation strategies either filter out challenging examples through data selection or introduce weaker intermediate assistant models, inherently compromising supervision coverage or quality. We propose Teacher Alignment, which directly adapts the teacher toward the student’s distribution without discarding data or degrading reasoning quality. However, naive alignment through standard knowledge distillation triggers catastrophic collapse of the teacher’s reasoning capabilities. To address this, we reformulate teacher alignment as reinforcement learning and introduce TeacherGRPO, built on Group Relative Policy Optimization with two key innovations: (i) Curriculum Selective Alignment applies dual token- and distribution-level curricula to focus rewards on high-signal reasoning gaps while filtering noise from trivial tokens and uncertain tail distributions, and (ii) Importance-Adaptive Length Regularization selectively penalizes verbose redundancy while preserving pedagogically critical reasoning steps. The aligned teacher then distills knowledge to students via standard pipelines. Extensive experiments show TeacherGRPO significantly outperforms baselines across diverse reasoning benchmarks and distillation methods. Our code is available at this https URL.
[NLP-199] Grounding Memory Summarization in Utility Intent
【速读】： 该论文旨在解决现有记忆系统摘要生成器在设计目标上的根本性错位问题：传统方法多以人类可读性（如忠实性）为优化目标，而实际需求应是保留支持未来查询所需的证据信息。其核心解决方案在于提出一种自蒸馏框架MemSuit，通过将摘要生成过程显式地基于已观察到的查询-回答对进行条件化，使生成的记忆条目具备下游任务实用性（utility-aware）。关键创新在于：教师模型将每个记忆块分解为多个独立可检索的自包含条目，每个条目分别保留与特定查询相关的不同信息维度，从而避免因单一查询条件化导致其他潜在查询所需证据被误删（即“旁带擦除”问题）。学生模型则仅依赖原始对话文本学习复现这些高密度、紧凑的事实型条目。此外，为使检索器与教师生成的紧凑风格对齐，引入基于教师条目的对比学习目标对嵌入模型进行微调。实验表明，该方法在多种对话查询类型上均显著优于当前最优基线，验证了以任务实用性为导向建模记忆的有效性。
链接: https://arxiv.org/abs/2609.33417
作者: Zhenyu Lei,Mingjia Shi,Xingbo Fu,Haoyu He,Qi R. Wang,Jundong Li
机构: University of Virginia (弗吉尼亚大学); Northeastern University (东北大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Existing summarizers for memory systems are typically optimized for human-facing criteria such as faithfulness, which misaligns with their true objective: preserving the evidence needed to support future queries. We show that conditioning summarization on query-answer pairs substantially improves answer quality, and that this utility-aware behavior is transferable across queries. Motivated by these findings, we propose MemSuit, a self-distillation framework in which a teacher summarizer, conditioned on observed query-answer pairs, produces utility-aware memory entries that a student learns to reproduce from the raw conversation alone. To prevent collateral erasure where conditioning on a single query-answer pair discards evidence relevant to other plausible queries, the teacher decomposes each block into multiple self-contained entries that preserve distinct query-relevant facets as independently retrievable units. To align the retriever with the compact, fact-dense style of teacher entries, we further fine-tune the embedding model with a contrastive objective supervised by teacher entries. Across a diverse suite of conversational query types, MemSuit consistently outperforms state-of-the-art baselines, confirming the value of grounding memory in downstream utility.
[NLP-200] MetaBench-Harness: Unlocking End-to-End Optimization of Benchmark Harnesses
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）评估中静态基准测试迅速饱和的问题，即传统基准测试难以持续挑战前沿模型的能力。现有自动化演化框架受限于硬编码的生成规则，仅能对单一任务进行微调，缺乏对整个基准生成流程的系统性优化。为此，本文提出MetaBench-Harness——一种双层搜索框架，通过端到端优化基准生成工作流来实现突破。其关键在于：内层循环利用基准夹具（benchmark harness）在每轮迭代中生成新基准，而外层元夹具编排层则基于历史演化轨迹，对夹具实现方案进行迭代优化与搜索。该方法实现了多维度演化，显著提升了演化合理性、基准竞争力及评估器鲁棒性。在编程竞赛（CodeContests）和奥数数学（AIME-2024）数据集上的实验证明，所生成的基准对前沿模型具有更强的挑战性和区分度，且案例分析揭示其能有效利用多样化的难度调控机制重构问题，提升所需能力要求。因此，该研究为应对基准测试饱和这一紧迫挑战提供了可扩展、自适应的解决方案。
链接: https://arxiv.org/abs/2609.33411
作者: Xuanjun Chen,Hua-Hsuan Chen,Wei-Chung Lu,Yinghao Ma,Jyh-Shing Roger Jang,Hung-yi Lee
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Software Engineering (cs.SE)
备注: Work in progress
Abstract:Rapid progress in Large Language Models (LLMs) is saturating static benchmarks faster than they can be designed. While existing automated evolution frameworks attempt to generate harder questions by perturbing individual tasks, they remain constrained by rigid, hard-coded generation rules. Moving beyond the evolution of isolated tasks, we propose to optimize the benchmark generation workflow itself end to end with MetaBench-Harness, a dual-loop search framework. Specifically, the inner loop utilizes a benchmark harness to generate a new benchmark in each round, while the outer meta-harness orchestration layer iteratively refines and searches over harness implementations based on historical evolution trajectories. By applying MetaBench-Harness to the competitive programming CodeContests and Olympiad mathematics AIME-2024 datasets, we demonstrate that the evolved benchmarks are challenging and discriminative for frontier models. Trajectory and quality analyses verify that MetaBench-Harness enables multi-dimensional evolution, steadily improving evolution reasonableness, benchmark competency, and evaluator robustness across successive rounds. Furthermore, case studies reveal its effective utilization of diverse difficulty levers to reframe problems and elevate required capabilities. Ultimately, this work provides a solution to the pressing challenge of benchmark saturation.
[NLP-201] Dense Is Not Enough: Hierarchical Supervision Allocation for Long-Horizon On-Policy Distillation
【速读】： 该论文旨在解决长时程智能体任务中基于策略的蒸馏（On-policy Distillation, OPD）因采用均匀的逐标记匹配而导致监督分配低效的问题。在长序列决策任务中，局部的标记差异未必影响未来行为，而关键的指导信息可能超出当前学生模型的能力范围或因缺乏特权输入而无法持续。为此，论文将长时程OPD建模为分层的监督分配问题，并提出核心原则：有效的指导应位于未来效用与当前可学习性之间的交集，且该交集随学生模型的学习过程动态演化。针对此原则，论文提出LENS-OPD框架，采用自粗至细的“定位（Locate）、验证（Validate）、精炼（Refine）”三阶段机制：定位阶段根据学生能力演变调整轨迹暴露并提出干预候选决策；验证阶段评估教师在该决策点的指导是否能改善学生后续行为；精炼阶段将经验证有益的引导行为内化为可部署策略，并聚焦于已验证回合内的关键教师-学生冲突标记进行精细化监督。各阶段嵌套设计确保细粒度监督依赖于粗粒度决策，而非独立优化重要性评分。实验结果表明，LENS-OPD在多个长时程智能体基准任务和师生配置下，显著优于原始OPD及强基线的课程学习与选择性蒸馏方法，验证了高效长时程蒸馏需在“正确深度、正确决策点、正确标记”三个维度精准施教。
链接: https://arxiv.org/abs/2609.33409
作者: Yuhao Sun,Binrui Wu,Zhuoer Xu,Ming Wen,Haoxiang Xu,Bin Chen,Yan Lin,Qianzijing Zhang
机构: Ant Group(蚂蚁集团); Alibaba International Digital Commerce Group(阿里巴巴国际数字商业集团); University of Science and Technology of China(中国科学技术大学); Peking University(北京大学); University of Electronic Science and Technology of China(电子科技大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:On-policy distillation (OPD) transfers the capabilities of a large language model to a smaller student by providing teacher supervision on the student’s own rollouts. In long-horizon agentic tasks, however, uniform token-level matching can allocate supervision poorly: a large local discrepancy need not improve future behavior, while consequential guidance may be beyond the current student’s reach or fail to persist without privileged input. We formulate long-horizon OPD as hierarchical supervision allocation and argue that productive guidance lies at the intersection of future utility and current learnability. Crucially, this intersection evolves as the student learns. Based on this principle, we propose LENS-OPD, a coarse-to-fine framework that organizes supervision through Locate, Validate, and Refine. Locate adapts trajectory exposure to the student’s evolving competence and proposes a candidate decision for intervention. Validate tests whether teacher guidance at that decision improves the same student’s subsequent behavior. Refine internalizes the beneficial guided behavior into the deployable policy and concentrates token-level supervision on decisive teacher-student conflicts within the validated turn. These stages are nested: each finer allocation is conditioned on the coarser decision, rather than being optimized as an independent importance score. Experiments across multiple long-horizon agent benchmarks and student-teacher configurations show that LENS-OPD consistently improves task performance over vanilla OPD and strong curriculum- and selection-based baselines. Our results suggest that effective long-horizon distillation requires teaching at the right depth, the right decision, and the right token.
[NLP-202] Decoupling Token Roles in Autoregressive Pretraining
【速读】： 该论文旨在解决自回归预训练模型在异构数据背景下，如何准确理解单个标记（token）对模型学习的贡献这一核心问题。传统方法通过下一词预测目标将每个标记的贡献直接关联于其自身的损失值，但这种方法忽略了标记的双重角色：既作为被预测的目标，又作为后续内容的上下文。研究提出通过受控破坏（controlled corruption）的方法，将这两个角色解耦，发现了一个反直觉现象：使某个噪声标记更容易被预测虽降低了其作为目标时的负面影响，却加剧了其作为上下文时的潜在危害。这一解耦机制还揭示了语言模型生成文本的内在逻辑——生成过程基于前缀与当前候选词的契合度进行选择，而该词作为上下文的作用从未经过独立生成的后续内容验证，因为后续内容始终被设计为适配该词。在已知被破坏的位置，通过优化其上下文作用可有效缓解仅依赖消除自身损失所无法消除的损害。因此，理解并控制模型从单个标记中学习的内容，关键在于明确区分并分别调控其作为“预测目标”与“上下文提供者”的双重角色。
链接: https://arxiv.org/abs/2609.33405
作者: Suqin Yuan,Runqi Lin,Kevin Qinghong Lin,Junchi Yu,Lei Feng,Chris Russell,Tongliang Liu
机构: University of Sydney(悉尼大学); University of Oxford(牛津大学); Southeast University(东南大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Autoregressive pretraining increasingly draws on heterogeneous data, making it important to understand how a model learns from an individual token. The next-token prediction objective naturally identifies a token’s contribution with its own loss. However, each token is not only a prediction target but also context for what follows. Using controlled corruption, we decouple these two roles and find a reversal: making a noisy token easier to predict reduces its damage as a target but increases it as context. The same decoupling helps explain text generated by language models: generation selects each token by its fit to the prefix, while its role as context is never tested against an independently determined continuation, because that continuation is generated to fit it. At known corrupted positions, acting through the context can reduce damage that removing the token’s own loss does not. Understanding and controlling what a model learns from a token therefore requires decoupling its roles.
[NLP-203] Preserving Morphemes: Morphology-Guided Pre-Tokenization for Nepali
【速读】： 该论文旨在解决尼泊尔语中基于字节级分词（Byte-level BPE）时，屈折形式（inflected forms）被当作独立字符串处理导致词干（stem）在不同格标记形式下拼写各异的问题，从而影响模型对词形结构的建模效率。其解决方案的关键在于引入一种预分词器Papaya，该工具利用公开的尼泊尔语语法构建有限状态转导器（finite-state transducer），并在无法匹配时回退至正则表达式，以在BPE训练前将词切分为词干与词缀（affixes）两部分，同时保持BPE训练器不变。实验表明，该方法在7名母语者标注的607个词上达到0.96的边界F1分数，显著提升了词干保持率；在1700万参数语言模型中，于相同训练步数下使每字节比特数降低约1%，主要增益源于更长的词元序列带来的额外训练步数，且无监督的Morfessor分词亦可获得类似效果。下游任务中，仅在训练数据未见词汇的命名实体识别（NER）上略有提升，而词性标注（POS tagging）和新闻分类性能基本不变，表明该方法对整体语言理解影响有限，但为形态丰富语言提供了有效的词形结构保留策略。
链接: https://arxiv.org/abs/2609.33395
作者: Kalash Shrestha,Nikhil Pradhan
机构: Kathmandu University ( Kathmandu 大学); IOE Thapathali Campus (IOE 塔帕塔利校区)
类目: Computation and Language (cs.CL)
备注: 13 pages, 2 figures, 6 tables. Code, data and tokenizers: this https URL
Abstract:A byte-level BPE vocabulary learns each inflected form of a Nepali word as a separate string, so a noun stem is spelled differently in each of its case-marked forms. We test whether splitting words into stem and affixes before BPE helps, with the corpus, vocabulary size, model and number of training steps held fixed. Our pre-tokenizer, Papaya, uses a finite-state transducer built from a published grammar of Nepali, falls back to regular expressions, and leaves the BPE trainer unchanged. On 607 words annotated by seven native speakers its segmenter reaches 0.96 boundary F1, and the resulting tokens keep stems intact far more often than plain BPE does. In a 17M-parameter language model it lowers bits per byte by about 1% at equal training steps; most of the larger gain seen at equal epochs comes from the extra steps that longer token sequences buy, and an unsupervised Morfessor segmentation gives the same improvement. Downstream the effect is small: NER improves only on entities that contain words unseen in training, POS tagging and news classification do not change, and published Nepali tokenizers perform about as well. We release the annotated boundary set, a 556-affix dataset and the code.
[NLP-204] From Position Risks to Block Survival: Faster Generation for Diffusion Language Models
【速读】： 该论文旨在解决扩散语言模型（Diffusion Language Models, DLMs）在并行生成过程中存在的核心矛盾：模型在预测多个词元（token）时采用并行方式，但这些预测结果的合理性却高度依赖于先前已确定的前缀（prefix），而现有方法无法有效利用这一依赖关系。尤其在流行的“提议-验证”解码策略下，早期生成的词元若被拒绝，将导致后续所有提议均无法推进解码进程，造成误差分布严重不对称。为应对这一问题，论文提出BRISK-DLM框架，其关键在于通过优化提议学习与选择机制，实现对“已验证进展”的动态追踪与高效利用。具体而言，该框架在自生成序列上进行训练，采用风险-收益加权策略，动态识别对验证进展和解码成本影响更大的位置，并优先优化；推理阶段引入轻量级前缀条件校正器（prefix-conditioned corrector），基于先前选定词元及模型自身验证器提炼出的偏好信息，对候选词元进行重排序，同时复用主干模型的并行表示，无需额外主干计算，且通过融合执行保持极低开销。该方案显著提升端到端生成吞吐率最高达37.4%，同时维持任务质量，成功建立扩散语言模型生成的新质量-吞吐权衡前沿。
链接: https://arxiv.org/abs/2609.33390
作者: Siwei Chen,Yuxiang Wan,Yifan Yu,Fan Lai
机构: University of Illinois Urbana-Champaign(伊利诺伊大学厄本那-香槟分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Diffusion language models (DLMs) can accelerate generation by predicting multiple tokens in parallel, but there is a mismatch between how these tokens are predicted and how they ultimately contribute to generation. Parallel predictions can hardly condition on the tokens selected earlier within the same block, even though their validity depends on this realized prefix. Under the popular proposal-verification decoding, this mismatch makes errors highly asymmetric: an early rejection prevents all subsequent proposals from contributing decoding progress. We introduce BRISK-DLM, a framework that addresses both mismatches by optimizing proposal learning and selection for verified progress. BRISK-DLM trains on self-generated sequences, using risk-reward weighting to dynamically prioritize positions by their impact on verified progress and decoding cost. During inference, a lightweight prefix-conditioned corrector reranks existing candidates using previously selected tokens and preferences distilled from the model’s own verifier. The corrector reuses the backbone’s parallel representations and requires no additional backbone evaluation, while fused execution keeps its overhead small. BRISK-DLM improves end-to-end throughput by up to 37.4% while preserving task quality, establishing a new quality-throughput frontier for DLM generation.
[NLP-205] OLED-MoE: Accelerating MoE-Based dLLM Inference via Inter-Iteration Locality-Aware Expert Offloading EUROSYS’27
【速读】： 该论文旨在解决生成式 AI（Generative AI）中半自回归扩散大语言模型（semi-autoregressive diffusion large language models, dLLMs）在引入混合专家（Mixture-of-Experts, MoE）架构后面临的显存瓶颈问题。具体而言，MoE层带来的巨大专家参数量超出受限于GPU显存容量的部署限制，而现有基于预取（prefetching）的专家卸载方案依赖于迭代内层间预加载机制，在dLLM推理中因块级路由导致每轮迭代中活跃专家集合显著扩大，使得预取难以及时完成且误预测代价高昂，最终退化为按需加载，造成高解码延迟。其解决方案的关键在于提出OLED-MoE系统，将优化目标从传统的迭代内预取转向迭代间专家保留（inter-iteration expert retention）。核心洞察是相邻去噪迭代间存在强专家路由重叠性，且通过令牌置信度可有效预测未来可能复用的专家。OLED-MoE利用置信度引导的跨迭代预测策略，在不增加额外预取流量的前提下，将高价值专家保留在GPU内存中；同时通过CPU-GPU协同执行机制，综合考虑动态计算负载与未来重用预测，补偿不可避免的缓存未命中。实验表明，OLED-MoE在多种dLLM工作负载下相较当前最优卸载系统将每输出词元时间（TPOT）降低1.23倍至7.93倍，专家缓存利用率提升1.44倍至4.23倍，并在仅使用40%专家显存的情况下接近全驻留性能，仅以23%的额外延迟代价实现60%的显存压缩。
链接: https://arxiv.org/abs/2609.33385
作者: Jingyuan Xiao,Jiayue Wang,Yitao Hu,Xinning Wang,Shi Chen,Ziqi Gong,Zhengchao Wang,Guotao Yang,Sheng Chen,Keqiu Li(Tianjin University, Tianjin, China)
机构: Tianjin University (天津大学)
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Computation and Language (cs.CL)
备注: Accepted by EuroSys '27 spring. Code available at this https URL
Abstract:Semi-autoregressive diffusion large language models (dLLMs) improve decoding parallelism through iterative block-wise denoising, but scaling them with mixture-of-experts (MoE) layers introduces a large expert parameter footprint that exceeds memory-constrained GPU capacity. Expert offloading is a natural remedy, yet existing MoE serving systems target autoregressive decoding and rely on intra-iteration layer-wise prefetching: while computing one layer, they predict and load experts for subsequent layers. Under dLLM inference, block-wise routing expands the active expert working set within each iteration, making such prefetches difficult to complete in time and costly when mispredicted. Consequently, existing prefetch-based solutions often degenerate into on-demand expert loading with high decoding latency. We propose OLED-MoE, an expert offloading system that shifts the optimization target from intra-iteration prefetching to inter-iteration expert retention. Its key insight is that adjacent denoising iterations exhibit strong expert routing overlap, and token confidence indicates which experts are likely to be reused. OLED-MoE uses confidence-guided inter-iteration prediction to retain high-value experts in GPU memory without introducing extra prefetch traffic. It further compensates unavoidable cache misses through CPU-GPU cooperative execution, jointly considering dynamic expert computation load and predicted future reuse. Across diverse dLLM workloads, OLED-MoE reduces time per output token (TPOT) by 1.23x-7.93x and improves expert cache utilization by 1.44x-4.23x over state-of-the-art offloading systems. Notably, OLED-MoE approaches full-residency performance while using only 40% of the expert GPU memory, incurring merely 23% higher TPOT despite a 60% reduction in expert memory footprint. OLED-MoE’s source code is publicly available at this https URL. Comments: Accepted by EuroSys '27 spring. Code available at this https URL Subjects: Distributed, Parallel, and Cluster Computing (cs.DC); Computation and Language (cs.CL) Cite as: arXiv:2609.33385 [cs.DC] (or arXiv:2609.33385v1 [cs.DC] for this version) https://doi.org/10.48550/arXiv.2609.33385 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-206] NLPG: Natural-Language Policy Gradients for Self-Evolving Language Agents
【速读】： 该论文旨在解决大语言模型智能体在执行复合程序（如检索、工具调用、推理与验证）时，因局部过程决策失误而导致失败的问题。现有基于强化学习或提示优化的方法通常依赖标量奖励信号，或反复修改整个提示文本，难以有效捕捉并复用过程性改进，同时无法在不改变模型参数或程序结构的前提下实现对冻结智能体的持续优化。为此，论文提出自然语言策略梯度（Natural-Language Policy Gradients, NLPG），一种外部策略记忆机制，能够在不修改智能体模型参数或程序结构的情况下，通过诊断执行轨迹、将下游反馈沿模块图反向传播，并将重复出现的失败转化为局部化的自然语言修正指令，进而聚合为有界策略更新，用于后续执行。在涵盖记忆、推理、指令遵循及证据验证等六项基准任务上的实验表明，NLPG相较于各任务最强基线平均提升8.71个百分点，证明了可评估的过程性经验能够被转化为局部且可解释的策略更新，从而实现对冻结智能体的持续改进。
链接: https://arxiv.org/abs/2609.33379
作者: Xu Liu,WenZhang Wei,Jun Cao,Dehua Peng,Huan Chen,Zhipeng Gui,Huayi Wu
机构: Wuhan University (武汉大学); State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing (测绘遥感信息工程国家重点实验室)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 30 pages, 7 figures, 10 tables. Work in progress
Abstract:Large language model agents increasingly rely on compound programs for retrieval, tool use, reasoning, and verification, yet their failures often arise from local procedural decisions. Existing reinforcement-learning and prompt-optimization approaches typically rely on scalar rewards or repeatedly modify entire prompts, making it difficult to capture and reuse procedural improvements while preserving a frozen agent. To address this problem, We propose Natural-Language Policy Gradients (NLPG), an external policy-memory method for improving a fixed agent without changing its model parameters or program structure. NLPG diagnoses execution traces, propagates downstream feedback backward through the module graph, and converts recurring failures into route-local natural-language corrections that are aggregated into bounded policy updates for subsequent executions. Across six benchmarks covering memory, reasoning, instruction following, and evidence verification, NLPG also outperforms the strongest listed baseline for each benchmark by 8.71 percentage points on average. These results provide evidence that evaluated procedural experience can be transformed into local and interpretable policy updates, enabling continual improvement of frozen agents.
[NLP-207] Language Discrimination Improves Linguistic Learning in Multilingual Speech Models
【速读】： 该论文旨在解决多语言自监督语音模型在共享跨语言信息时仍落后于单语言模型的问题，尤其是在总预训练数据量受限的情况下。其核心问题是：多语言模型在保持跨语言知识共享的同时，如何提升对连续语音特征及高层语言学任务的建模能力。解决方案的关键在于在预训练阶段增强模型的语言辨别能力，具体通过引入两种干预手段实现——辅助语言分类器和每语言独立的k-means聚类目标。实验结果表明，这一策略显著降低了语音层面的语言混淆（如phone-ABX错误率从11.6%降至10.4%），并提升了词汇级（sWUGGY）和语调级（ProsAudit）性能，使多语言模型在多数指标上接近单语言模型表现。更重要的是，语言辨别能力的早期引入（首轮训练阶段）带来最大增益，而后期或重复干预则导致语言间隔离加剧，削弱跨语言共享。这表明语言辨别能力在降低多语言学习额外成本中具有因果作用。
链接: https://arxiv.org/abs/2609.33345
作者: Maureen de Seyssel,Jie Chi,Zakaria Aldeneh
机构: Apple(苹果)
类目: Computation and Language (cs.CL); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注:
Abstract:Multilingual self-supervised speech models can benefit from sharing information across languages, but under a matched total pretraining data budget they still fall short of monolingual models. We show that strengthening the model’s ability to discriminate languages during pretraining reduces and, on some measures, closes this multilingual gap on continuous phonetic and higher-level linguistic measures, while preserving substantial cross-language sharing. Using a controlled English/French HuBERT setting, we test two interventions which strengthen language discrimination: an auxiliary language classifier and per-language k-means targets. Across interventions, continuous-feature phone discrimination error (phone-ABX, lower is better) decreases from 11.6% in the bilingual baseline to 10.4% (monolingual: 10.8%), while lexical performance (sWUGGY, higher is better) increases from 52.1% to 56.7% (monolingual: 58.5%) and prosodic performance (ProsAudit, lexical subtask, higher is better) from 68.9% to 72.9% (monolingual: 72.6%). Across HuBERT training stages, the strongest gains on most linguistic measures occur when language discrimination is introduced in the first iteration, whereas later or repeated interventions yield smaller improvements and are accompanied by increased language-wise segregation. These results support a causal role for language discrimination in reducing the additional cost of multilingual learning.
[NLP-208] CHI: A Composite Hallucination Index Unifying Entity Relation and Quantity Dimensions for Summarization Evaluation
【速读】： 该论文旨在解决生成式摘要中忠实度（faithfulness）评估的开放性挑战，现有评价指标仅能独立检测实体错误、关系不一致或数值虚构等单一类型的幻觉，无法捕捉这些幻觉类型之间的共现与交互关系。其核心解决方案是提出首个统一的幻觉评估指标CHI（Composite Hallucination Index），将忠实度误差分解为三个正交维度：实体幻觉指数（EHI）、关系幻觉指数（RHI*）和数量幻觉指数（QHI）。每个维度均采用基于韦恩图推导出的提取性、正向幻觉、过度聚焦、负向幻觉和失焦等因子的共享softmax归一化架构进行建模，其中新提出的QHI引入了容差感知的数值匹配机制，涵盖精确匹配、ε容差匹配、派生值匹配及时间比较等多种模式。通过调和平均融合三维度得分，生成一个综合评分，对任一维度的薄弱表现均施加惩罚。在涵盖新闻、医疗、法律和金融四个领域的800篇源文章上，使用五种生成系统生成的摘要进行验证，结果表明：（i）三维度间统计正交性显著（均相关系数ρ = 0.148），证明其可区分不同错误类型；（ii）CHI在SummEval数据集上与人工判断的相关性最高（ρ = 0.66, p = 0.006），显著优于ROUGE（ρ = 0.53）、EHI（ρ = 0.58）及各分量单独表现；（iii）消融实验进一步证实三维度贡献独特且不可替代，完整复合模型性能超越任一分量，同时提供可分解的错误诊断能力，这是单分值基线所不具备的优势。CHI为研究人员和实践者提供了兼具可解释性、可分解性与高效性的忠实度评估工具，适用于离线评估与在线监控场景。
链接: https://arxiv.org/abs/2609.33343
作者: Praveenkumar Katwe,Rakesh Chandra Balabantaray,Kali Prasad Vittala
机构: International Institute of Information Technology Bhubaneswar(印度信息科技国际研究所布巴内斯瓦尔分校); Salesforce India Pvt Ltd( Salesforce 印度私人有限公司)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 18 pages, 2 figures, 12 tables, 25 references
Abstract:Faithfulness evaluation of abstractive summaries remains an open challenge, with existing metrics addressing only isolated hallucination types: factual entity errors, relational inconsistencies, or numerical fabrications, without capturing their co-occurrence or interaction. We introduce CHI (Composite Hallucination Index), the first unified hallucination metric that decomposes faithfulness errors into three orthogonal dimensions: entity hallucination (EHI), relation hallucination (RHI*), and quantity hallucination (QHI). Each dimension employs a shared softmax-normalized architecture over Venn diagram-derived factors representing extractiveness, positive hallucination, over-focus, negative hallucination, and lost focus. The novel QHI component introduces tolerance-aware numerical matching with exact, epsilon, derived, and temporal comparison modes. We fuse the three dimensions via harmonic mean to produce a single composite score that penalizes weakness in any dimension. We validate CHI on 800 source articles spanning four domains (news, medical, legal, financial) with summaries from five generation systems. Empirical results demonstrate that: (i) the three dimensions are statistically orthogonal (mean rho = 0.148), confirming they capture distinct error types; (ii) CHI achieves the highest system-level correlation with human judgments (rho = 0.66, p = 0.006) on SummEval, outperforming ROUGE (rho = 0.53), EHI (rho = 0.58), and all individual components; and (iii) ablation studies confirm that all three dimensions contribute unique variance, with the full composite outperforming any individual component while providing decomposable error diagnostics unavailable from single-score baselines. CHI provides practitioners with a decomposable, interpretable, and efficient faithfulness metric suitable for both offline evaluation and online monitoring of summarization systems.
[NLP-209] CalibHyper: Chance-Corrected Relational Hypergraphs for Few-Shot Molecular Property Prediction
【速读】： 该论文旨在解决分子属性预测中因实验成本高和标注数据稀缺导致的少样本学习难题。现有上下文感知方法虽利用辅助检测标签支持少样本预测，但其依赖标签一致性（label agreement）来监督属性间关系，而该指标对类别分布边缘敏感，且无法直接建模属性间的内在依赖关系。为此，本文提出CalibHyper——一种基于联合标签分布的校正偶然性关系超图方法。其核心创新在于：从有序四状态标签分布中减去独立性基线，并根据联合观测数量对残差进行收缩；通过交换等变的关系头估计这些残差，进而为每个分子选择辅助属性，并确定其超边消息的符号与权重。在五个基准测试的十三个数据集上，无论是1-shot还是10-shot设置，CalibHyper及其消融版本均在ROC-AUC指标上达到与最强报告结果相当的性能。
链接: https://arxiv.org/abs/2609.33342
作者: Linyu Li,Zhi Jin,Yuanpeng He,Dongming Jin,Huanyu Liu,Huanyao Zhang,Haoran Duan,Heng Tian,Gadeng Luosang,Nyima Tashi
机构: Peking University (北京大学); Wuhan University (武汉大学); Tibet University (西藏大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Molecular property prediction is central to drug development and materials discovery, but experiments are costly and labeled data are scarce. Context-aware methods use auxiliary assay labels to support few-shot prediction, and recent work supervises property relations with label agreement. However, label agreement is sensitive to class marginals and does not directly capture dependence between properties. We propose CalibHyper, a chance-corrected relational hypergraph method based on the joint label distribution. CalibHyper subtracts an independence baseline from the ordered four-state label distribution and shrinks the residual according to the number of joint observations. A swap-equivariant relation head estimates these residuals, which choose the auxiliary properties for each molecule and set the sign and weight of their hyperedge messages. On thirteen datasets from five benchmarks, in both 1-shot and 10-shot settings, CalibHyper and its ablation settings achieve ROC-AUC competitive with the strongest reported results.
[NLP-210] Does Learning to Predict the World Help Agents Act? Auditing World-Model Post-Training
【速读】： 该论文旨在解决生成式智能体（Generative AI）在训练过程中性能提升的来源问题，即现有基于世界模型（World Model）的后训练方法虽能提升决策能力，但其性能增益是否完全源于对环境的准确预测尚不明确。现有方法依赖于对下一观察结果的精确预测，并将其转化为奖励或监督信号，然而这一过程伴随复杂的优化机制，可能引入非预测相关的干扰因素。为厘清根本原因，论文提出关键创新：在训练中使用分布内不匹配的观测作为目标（in-distribution mismatched observations），而非真实下一时刻的观测。实验表明，尽管预测准确性相对真实目标下降15.3%–61.6%，但模型仍显著优于基础模型，在任务表现上实现持续提升；同时，模型展现出更广泛的候选动作探索能力与更低的循环行为。此外，研究进一步引入独立随机信号替代基于预测的奖励机制，在无环境信息的情况下仍显著扩展任务覆盖范围（pass@64）。该发现被成功推广至VisualWebArena基准，随机奖励训练使pass@64提升14.3%，且无需观测匹配奖励或外部多模态教师。因此，论文的核心结论是：性能提升主要来源于训练过程中的策略探索增强与动作多样性扩展，而非单纯依赖对环境的精准建模。
链接: https://arxiv.org/abs/2609.33335
作者: Xinyu Che,Hang Yan,Yanchen Liu,Haochen Liu,Ruifeng Li,Anran Shi,Heng Wang,Jun Liu
机构: Xi’an Jiaotong University(西安交通大学); University of Southern California(南加州大学); University of the Chinese Academy of Sciences(中国科学院大学); East China Normal University(华东师范大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 24 pages, 6 figures. Preprint
Abstract:Predicting how an environment will change before acting is a natural route to better decision making for agents. Recent post-training methods therefore require agents to predict the next observation and turn that prediction into a reward or a direct supervision signal, which is called world model. Existing next-observation training methods help the agent to learn the environmental content. However, they additionally involve an optimization process, which may introduce several effects other than learning to predict the world. Consequently, where the performance gain comes from during the training process remains an open question. We answer this research question through replacing true next-observation targets with in-distribution mismatched observations during the training process. Across two interactive text environments, mismatched targets lower prediction accuracy by 15.3-61.6% relative to ground-truth targets, yet retain substantial task gains over the base model. Compared with the base model, trained models consider more candidate actions and exhibit less looping. We also introduce a setting that replaces prediction-based rewards with independent random signals. This training expands task coverage (pass@64) even when the reward carries no environment information. We also generalize this finding to VisualWebArena, where random-reward training raises pass@64 by 14.3% relative to the base model, without observation-matching rewards or an external multimodal teacher for reward construction.
[NLP-211] When to Evict Not What to Keep: Draft-Guided Eviction for Training-Free KV-Cache Compression
【速读】： 该论文旨在解决训练-free KV-cache压缩方法（如SnapKV、H2O和PyramidKV）在预填充阶段末尾进行令牌淘汰所导致的性能瓶颈问题。这些方法通过优化“保留哪些注意力质量”来减少缓存占用，但其核心缺陷在于：淘汰操作发生在决定答案轨迹的查询尚未生成之前，导致无法根据实际生成路径做出精准决策。具体表现为两个层面的问题：(1) 补偿性失败——即使恢复被淘汰的注意力质量，也无法提升任务整体性能；(2) 选择性失败——即便覆盖了更多真实解码查询的注意力质量，若恢复的质量呈现碎片化分布而非集中于连贯片段，反而会损害生成质量。上述问题的根本原因在于淘汰时机过早，缺乏对后续生成路径的感知。为此，论文提出草稿引导淘汰（Draft-Guided Eviction, DGE），将淘汰时机推迟至使用完整缓存完成前k=2个回答令牌的草稿生成之后——仅比预填充多一步解码。由于草稿由目标答案自身的前缀生成，此时已具备明确的答案轨迹信号，可实现基于实际生成路径的精准淘汰。DGE保持每头缓存预算不变，且无需修改原有方法的淘汰评分机制，可直接应用于SnapKV、PyramidKV、H2O及StreamingLLM。与需要额外推理遍历的方法不同，DGE仅改变淘汰时间点，而不改变淘汰内容。大量实验表明，DGE在六种指令微调模型中的五种、所有评估缓存预算下均优于现有方法，在LongBench上达到44.2分，几乎逼近全量缓存（FullKV）的44.3分。进一步地，仅调整淘汰时机的变体DGE-W也取得相同表现，验证了性能提升源于“淘汰时机”的优化，而非淘汰内容的选择，这一现象被定义为轨迹锚定（trajectory anchoring）。
链接: https://arxiv.org/abs/2609.33334
作者: Haeyong Kang,Chang D. Yoo
机构: Duksung Women’s University (德淑女子大学); KAIST (韩国科学技术院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Training-free KV-cache compression methods such as SnapKV, H2O, and PyramidKV evict tokens at the end of prefill, aiming to preserve the attention mass that future queries are expected to use -optimizing what to keep. We show that this objective fails in two distinct ways. (1) Compensation: restoring the evicted attention mass can recover the attention-level target without recovering task quality. (2) Selection: covering more of the true decode-query mass can hurt quality when the recovered mass is fragmented rather than concentrated in coherent spans. These failures share a common cause: eviction occurs before the queries that determine the answer trajectory exist. We propose Draft-Guided Eviction (DGE), which defers eviction until after drafting the first k=2 answer tokens using the full cache - just one decode step beyond prefill. Because the draft is generated from the answer’s own prefix, no cache entries are discarded before this trajectory signal becomes available. The per-head cache budget remains unchanged, and DGE can be applied directly to SnapKV, PyramidKV, H2O, and StreamingLLM without modifying their eviction scores. Unlike extra-pass methods, DGE changes when eviction occurs rather than what cache entries are selected. Extensive experiments demonstrate that DGE outperforms prior methods at every evaluated budget on five of six instruct-tuned backbones, achieving 44.2 on LongBench, nearly matching FullKV at 44.3. The timing-only control DGE-W achieves the same score, demonstrating that the gain comes from when eviction occurs rather than what is selected - an effect we term trajectory anchoring.
[NLP-212] CertMark: Distortion-Free Multi-Bit Watermarking with Certified Decoding
【速读】： 该论文旨在解决现有语言模型多比特水印技术中消息恢复准确率与文本质量之间的权衡问题，特别是传统方法通过偏置模型的下一个词概率来编码信息，导致生成文本质量下降。其核心缺陷在于解码器缺乏经过认证的弃权机制（certified abstention rule），无法在输出错误消息时提供可证明的概率上界。为应对这一挑战，论文提出CertMark，一种保持分布不变的多比特水印方案：它不修改原始概率分布，而是利用嵌入的消息作为种子，精确采样Gumbel-max分布，从而完全保留模型原有的采样分布特性。该方法设计了两种可扩展的解码器——一种仅依赖文本的模型无关型解码器，另一种利用原始词元分布的模型感知型解码器，二者均支持具有数学保证的认证弃权机制，能够严格控制输出错误消息的概率。实验表明，在文本补全、摘要生成和故事生成任务中，CertMark在保持与未加水印文本相当的困惑度的同时，能可靠地恢复多比特消息；其中模型感知型解码器在比特准确率上优于基于概率偏置的基线方法。
链接: https://arxiv.org/abs/2609.33332
作者: Paweł Batorski,Przemysław Spurek,Paul Swoboda
机构: Heinrich Heine University Düsseldorf(海因里希·海涅杜塞尔多夫大学); Jagiellonian University(亚捷隆大学); IDEAS Research Institute(IDEAS 研究所)
类目: Computation and Language (cs.CL)
备注:
Abstract:Leading multi-bit watermarking methods for language models encode messages by biasing the model’s next-token probabilities, creating a trade-off between message recovery and text quality. Their decoders typically return the highest-scoring candidate from accumulated token-level evidence, without a certified abstention rule that bounds the probability of outputting an incorrect message. We introduce CertMark, a distribution-preserving multi-bit watermark with certified decoding. Rather than modifying probabilities, CertMark uses the embedded message to seed an exact Gumbel-max sampler, thereby preserving the model’s original sampling distribution. We propose two scalable decoders: a model-agnostic, text-only decoder and a model-aware variant that leverages the original next-token distributions for stronger recovery. Both support certified abstention with mathematical bounds on the probability of returning an incorrect message. Across text completion, summarization, and story generation, CertMark matches the perplexity of unwatermarked text while reliably recovering multi-bit messages. The model-aware decoder further achieves higher bit accuracy than probability-biasing baselines. Our code is publicly available at this https URL.
[NLP-213] Hesitation-Aware On-Policy Distillation for Diffusion Language Models
【速读】： 该论文旨在解决扩散型大语言模型（Diffusion Large Language Models, dLLMs）在文本生成过程中，基于策略的蒸馏方法（如Trace-based on-policy distillation, TOPD）仅利用模型在最终确定位置上的预测信息，而忽略大量未被采纳但蕴含丰富语义信号的“犹豫”（hesitations）的问题。这些犹豫指代的是模型在去噪步骤中虽提出候选词但尚未达到置信度而未提交的预测，其虽占比仅为24%的可监督状态-位置对，却承载了66%的师生模型间差异。为充分挖掘此类信息，本文提出犹豫感知的在线策略蒸馏（Hesitation-Aware On-Policy Distillation, HOPD），将教师分布匹配扩展至每个去噪步骤中所有掩码位置，实现全位置监督。同时，通过事后回溯（hindsight）机制动态分配监督权重，优先强化那些后续被推翻的预测以及首步预测留存率低的区块，以提升关键位置的指导效率。由于两模型均已在各掩码位置生成概率分布，HOPD无需额外前向传播，仅增加损失计算开销。实验表明，在SDAR-1.7B与SDAR-4B学生模型上，从TraDo-8B-Instruct蒸馏得到的HOPD在五项数学与编码基准测试中均取得最优平均得分，且在静态与动态解码策略下均表现更优，同时加速了推理过程；在SDAR-4B上，其犹豫比例更低，每步提交令牌数提升11%，并达成更高准确性。
链接: https://arxiv.org/abs/2609.33301
作者: Jianguo Huang,Lipeng Wan,Yanchen Deng,Bo An
机构: Nanyang Technological University, Singapore; Xi’an Jiaotong University, China
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Diffusion large language models (dLLMs) generate text by iterative unmasking. At each denoising step, a dLLM proposes a token at every masked position, but the decoder commits only a confident subset of these proposals. Trace-based on-policy distillation (TOPD) builds on this process by matching the student to a stronger teacher, yet only at the committed positions. We argue that this discards much of the useful signal, which resides in the uncommitted proposals, where the student has made a prediction but is not yet confident enough to commit it. We call these proposals hesitations. In our pilot study on an SDAR-4B student, hesitations make up only 24% of supervisable state-position pairs but carry 66% of the teacher-student divergence. To exploit this signal, we propose Hesitation-Aware On-Policy Distillation (HOPD), which extends teacher distribution matching to every masked position of each denoising step. Because hesitations are not equally informative, we further allocate supervision using hindsight from the completed trajectory, placing more weight on positions whose proposal was later disagreed with the final token and on blocks where first-step proposals rarely survive. Since both models already produce distributions at all masked positions, HOPD requires no additional forward passes over TOPD. The only extra cost is evaluating the loss at more positions. With SDAR-1.7B and SDAR-4B students distilled from TraDo-8B-Instruct, HOPD achieves the best average score among the evaluated methods on five math and coding benchmarks, under both static and dynamic decoding and at both scales. It also speeds up decoding. On SDAR-4B, the HOPD student hesitates less and commits 11% more tokens per denoising step than TOPD, while reaching higher accuracy.
[NLP-214] BaatCheet: A Multilingual Corpus for Dialogue Translation in Indian Languages
【速读】： 该论文旨在解决现有翻译模型在处理日常对话场景时的局限性，尤其是其在捕捉非正式语体、说话人互动及话语连贯性等对话现象方面的不足。当前大多数印地语族（Indic）语言的翻译资源与评估基准仍集中于句级或正式文本，难以有效评估对话翻译的质量。为此，本文提出BaatCheet——一个以印地语中“谈话”或“闲聊”（chitchat）命名的多语言对话语料库，涵盖约49,000个对话样本，支持五种语言间的对话翻译任务。研究通过在七种不同训练数据配置下微调五个开源大语言模型（LLM），发现微调策略相较于零样本和少样本基线可显著提升翻译性能。其解决方案的关键在于构建高质量的多语言对话语料库，并采用综合评估策略，包括自动评价指标、基于大语言模型的评判（LLM-as-judge）以及基于语义质量度量（SQM）引导的直接评估（DA）协议的人工评估，从而全面衡量对话翻译中的语用与连贯性表现。
链接: https://arxiv.org/abs/2609.33296
作者: Priyanka Dasari,Yuvrajsinh D. Bodana,Vandan Mujadia,Arafat Ahsan,Dipti Misra Sharma,Parameswari Krishnamurthy
机构: Language Technologies Research Centre, IIIT Hyderabad (语言技术研究中心，印度国际信息技术学院); IIIT Hyderabad (印度国际信息技术学院)
类目: Computation and Language (cs.CL)
备注:
Abstract:Existing translation models are typically trained on sentence-level and formal text, limiting their ability to capture everyday conversational dialogue phenomena such as informality, speaker interaction, and discourse coherence. Most existing Indic translation resources and evaluation benchmarks focus on sentence-level or formal text, making it difficult to assess translation quality of the dialogue phenomena. In this work, we introduce BaatCheet, a multilingual dialogue corpus named after the Hindi term for conversation or chitchat, containing approximately 49,000 dialogues for dialogue translation across five translation directions. We fine-tune five open-source LLMs across seven training data configurations and find that fine-tuning yields substantial gains over zero- and few-shot baselines. To comprehensively evaluate dialogue translation quality, we employ multiple evaluation strategies, including automatic metrics, LLM-as-judge, and human assessments using an SQM-guided Direct Assessment (DA) Protocol.
[NLP-215] raceDance: An Automated System for Building Agent Behavior Benchmarks from Real-World Agent Deployment Traces
【速读】： 该论文旨在解决生成式 AI（Generative AI）在实际部署过程中出现的不可预测或不期望行为（undesirable behavior）难以被有效测试与评估的问题。传统基准测试套件无法覆盖真实场景中复杂多变的具体行为，导致开发者缺乏针对性的验证手段。为此，论文提出 TraceDance 系统，其核心解决方案是基于部署日志（deployment traces）构建针对特定不期望行为的靶向基准（targeted benchmarks）。关键创新在于“锚定与确认”（Anchor-and-Confirm）机制：通过可编程检索快速定位候选实例，并利用 Flash 大语言模型（LLM）进行候选层面的行为确认，实现高效构建；同时引入“锚点合成循环”（Anchor Synthesis Loop），动态生成并迭代优化自定义行为规范。这些基准采用决策点延续（decision-point continuation）范式，在不依赖参考答案或环境重放的前提下，基于行为特异性评分标准评估大模型在已记录决策点上的后续响应质量。实验基于 252,557 次会话，生成了 107 个基准，涵盖 4,125 个实例，满足率达 95.3%。人工标注验证显示 84% 的样本符合目标行为，自动化评分器与人工判断一致性接近人类间一致性。九个前沿大模型在该评测中平均通过率仅为 26.7%，揭示当前主流模型作为智能体在关键决策点仍存在显著行为缺陷。该方法将真实部署问题转化为可度量、可迭代的靶向基准，为实现递归自我改进（Recursive Self-Improvement, RSI）提供了关键支撑。
链接: https://arxiv.org/abs/2609.33295
作者: Dehai Min,Daoan Zhang,Yiming Zeng,Huayi Zhang,Ziyi Chen,Yan Zhang,Qinbo Bai,Mengyuan Chao,Jing Ning,Qiyue Hua,Huiyi Chen,Hanrong Zhang,Henry Peng Zou,Jie Yang,Wei Xu,Philip S. Yu
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 34 pages, 7 figures. Project website: this https URL
Abstract:An agent can complete a task while exhibiting undesirable behavior during execution. Developers need tests for the specific behaviors encountered in deployment, beyond fixed benchmark suites. We present TraceDance, an agent system that constructs targeted benchmarks from deployment traces for user-specified undesirable behaviors. For efficient construction, Anchor-and-Confirm combines programmable retrieval with candidate-level confirmation by a Flash large language model (LLM), while the Anchor Synthesis Loop generates and revises specifications for custom behaviors. The benchmarks use decision-point continuation to evaluate an LLM’s next turn at a recorded decision point with a behavior-specific rubric, without a reference answer or environment replay. Experiments in coding and general tool use draw on 252,557 sessions and produce 107 benchmarks with 4,125 instances, fulfilling 95.3% of build-target requests. Both human annotators confirm the requested behavior in 84% of sampled instances, and the automated grader’s agreement with human pass/fail judgments is comparable to that between the annotators. Nine frontier LLMs achieve a mean pass rate of only 26.7%, showing that they still struggle to respond appropriately at the evaluated decision points. Analysis across behavior-specific benchmarks further reveals weaknesses in how current LLMs behave as agents. By turning deployment problems into targeted benchmarks, TraceDance could serve as a key component of the recursive self-improvement (RSI) loop.
[NLP-216] Calibration Not Answer Selection: Distilling Internal Confidence in Reasoning Models
【速读】： 该论文旨在解决生成式推理模型在使用二元正确性奖励进行强化学习时，其自述置信度（verbalized confidence）系统性过高的问题。现有方法中，模型的置信度反映的是其回答意愿而非真实正确概率，导致后验校准（post-hoc rescaling）难以跨场景迁移。解决方案的关键在于不依赖外部校准或后处理，而是通过分析模型内部状态——具体而言，在思维链（Chain-of-Thought）与最终答案之间的隐藏状态上施加线性探测器（linear probe），以获得更准确的置信度估计。该内部信号虽能有效回答“我有多确定？”这一问题，却无法可靠判断“哪个答案是正确的”，因此应作为置信度报告而非直接用于答案选择。基于此，论文提出探针引导的自蒸馏（Probe-guided Self-Distillation, Probe-SD）：利用探测器对模型自身采样的推理路径评分，替换原始置信度，并仅通过监督微调（supervised fine-tuning）更新基础模型参数，使测试阶段无需额外组件。实验表明，在Qwen3-14B上，Probe-SD将域内预期校准误差（ECE）从0.178降至0.024，域外从0.542降至0.113，显著优于后验校准和自一致性蒸馏，且所得置信度可有效支持加权投票，实现了此前仅靠在线强化学习才能达成的行为表现。
链接: https://arxiv.org/abs/2609.33290
作者: Yadong Xi,Rongsheng Zhang,Tangjie Lv,Ziyang Luo,Ruochen Zhao
机构: NetEase(网易); Amazon(亚马逊); Singapore University of Technology and Design(新加坡科技设计大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Reinforcement learning with binary correctness rewards trains correctness, not calibrated confidence. The confidence that reasoning models verbalize is systematically overconfident, and the problem is not merely one of scale: verbalized confidence tracks how willing a model is to commit to an answer, not how likely the answer is to be right. Post-hoc rescaling therefore fits one distribution but rarely transfers. We look inside the model instead. On factual question answering, a linear probe on the hidden state between the chain of thought and the answer is substantially better calibrated: its expected calibration error is 5 to 38 times lower than that of the verbalized score across four benchmarks and two model families. However, when used to pick among N sampled answers, that same probe nearly ties majority voting yet falls far short of the oracle. Internal states answer “how certain am I” well and “which answer is right” poorly, so the signal should be reported as a confidence rather than used to select answers. As a result, we introduce probe-guided self-distillation (Probe-SD): score a model’s own sampled traces with the probe, overwrite the confidence each trace states, and finetune the base checkpoint of the same family, so nothing but the model itself remains at test time. On Qwen3-14B, Probe-SD cuts ECE from 0.178 to 0.024 in-domain and from 0.542 to 0.113 out-of-domain, where it also beats post-hoc recalibration and self-consistency distillation. The resulting confidence is well-calibrated and useful for weighted voting, behaviors previously attributed to online RL, here obtained with supervised finetuning alone.
[NLP-217] InfoEdit: Probing Global Layout Reasoning in Infographic Editing
【速读】： 该论文旨在解决生成式图像编辑模型在处理结构化视觉内容（如信息图）时面临的显著性能瓶颈问题。与自然照片不同，信息图依赖于元素间的逻辑关系进行信息编码，其编辑往往需要全局布局的重新调整，这一能力被称为“重排（reflow）”。现有图像编辑基准未针对结构化视觉内容设计评估场景，也缺乏对重排能力的有效衡量。为此，论文提出InfoEdit——一个包含1,000张信息图、涵盖八类逻辑关系、4,000条编辑指令及四种编辑任务的新型基准，并引入重排感知的评估协议。实验表明，在八个前沿编辑模型中，仅有GPT-Image-2达到60%的平均成功率，多数模型低于7%，且在“交换块”（Swap-Block）任务上即使具备完美目标定位，最高成功率也未超过36%。研究进一步发现，代码级编辑可媲美最强的像素级编辑器，揭示了两类方法在不同任务中的互补优势。因此，InfoEdit将重排能力识别为结构化视觉内容编辑的核心挑战，并提供了一个诊断性基准，以推动未来模型在复杂布局理解与全局一致性生成方面的发展。
链接: https://arxiv.org/abs/2609.33286
作者: Cheng Yang,Chufan Shi,Huijuan Wang,Bo Shui,Yaokang Wu,Muzi Tao,Yibo Yan,Xuezhe Ma,Taylor Berg-Kirkpatrick
机构: University of California San Diego(加州大学圣地亚哥分校); University of Southern California(南加州大学); University of Illinois Urbana-Champaign(伊利诺伊大学厄本那-香槟分校); Carnegie Mellon University(卡内基梅隆大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Software Engineering (cs.SE)
备注: Project page: this https URL
Abstract:Multimodal foundation models edit natural photographs at production quality, yet the same models struggle with structured visual content such as infographics. Unlike photographs, infographics encode information through logical relations; editing one element often requires surrounding elements to be adapted. We refer to this global layout reasoning capability as reflow. Existing image-editing benchmarks neither provide a dedicated setting for structured visual content nor evaluate the reflow capability. We introduce InfoEdit, a novel benchmark of 1,000 infographics across eight logical-relation families, paired with 4,000 editing instructions across four editing tasks, and a reflow-aware evaluation protocol. Across eight frontier editors, only GPT-Image-2 clears 60% average success rate; most models fall below 7%, and no editor exceeds 36% on the Swap-Block task even with perfect target localization. We further show that code-level editing can match the strongest pixel-level editor, revealing complementary strengths across tasks. InfoEdit identifies reflow as a central challenge in structured visual content editing and provides a diagnostic benchmark to facilitate future progress.
[NLP-218] he Text Beside the Image: Detection Utility and Leakage for Trustworthy Multimodal Medical Data and Beyond
【速读】： 该论文旨在解决医疗图像发布时伴随的报告文本中潜在的隐私泄露问题，即仅对图像进行保护无法有效防止与之关联的文本报告中的敏感信息暴露。其核心挑战在于评估在不同伪匿名化策略下，医学报告等文本文档中存在的身份识别风险，包括标识符检测、下游用途可用性以及残余身份泄露程度。解决方案的关键在于构建一个由13个检测器组成的集成模型（ensemble），在多种语言（德语、英语、中文、阿拉伯语）和多类文本语料（医学报告、法律判决、新闻及其他文体、电子邮件）上进行统一评估。该集成模型在医学报告上的人员敏感度达0.9998，特异性为0.8686，表现出极强的身份识别能力；同时通过频率匹配公开姓名列表的方式，在跨文档场景下实现了零身份恢复，表明未被检测器捕获的明文姓名成为主要泄露源。此外，基于跨文档关联分析，即使不依赖训练数据，也能在邮件查询中以0.93%的概率正确排序目标个体，经训练后提升至3.94%，显著高于随机概率；在医学报告中虽未实现无训练条件下的成功恢复，但经训练后仍达到0.71%的恢复率，远超基线。这揭示了当前伪匿名化策略在应对复杂文本数据中的身份再识别攻击时存在显著局限性。
链接: https://arxiv.org/abs/2609.33280
作者: Andreas Maier,Monica Hinrichs-Mayer,Franziska Weber,Niklas Lackner,Matthias May,Bernhard Kainz,Siming Bayer
机构: Friedrich-Alexander-Universität Erlangen-Nürnberg(弗里德里希-亚历山大-埃尔兰根-纽伦堡大学); Universitätsklinikum Erlangen(埃尔兰根大学医院); Imperial College London(帝国理工学院)
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 12 pages, submitted for peer review
Abstract:Medical images are released with the reports that describe them, and protecting the image does not protect the report. This paper measures the text component of such releases. We measure identifier detection, downstream utility and residual identity leakage on the same documents, with the pseudonymisation policy as the variable under test: 15 detectors, three release conditions and four corpora of medical reports, legal judgments, news and other genres, and e-mail, in German, English, Chinese and Arabic. A fixed 13-detector union reaches a person sensitivity of 0.9998 at specificity 0.8686 on the medical reports, 0.9958 at 0.8504 on the legal judgments, 0.9352 at 0.9318 on news and other genres, and 0.9906 at 0.6235 on e-mail. With this ensemble, frequency matching with a public name list recovers zero identities by alignment across the four corpora; the names it got right were ones the detector missed, left in clear text. Cross-document linkage ranks the correct person first for 0.93% of e-mail queries without training and 3.94% with it, against 1/3697 chance and 71.98% on unmodified text. On the medical reports it recovers nothing without training and 0.71% of 138 queries with it, against 1/207 chance and a 2.73% ceiling on unmodified text.
[NLP-219] BERT4DTI : BERT-based Model for Predicting Drug-Protein Interactions CIKM2026
【速读】： 该论文旨在解决基于序列的药物-靶标相互作用（Drug-Target Interaction, DTI）预测模型在实际应用中面临的三大挑战：已标注的相互作用数据稀缺且分布不均、大规模预训练化学与蛋白质编码器端到端微调成本高昂，以及独立编码序列无法捕捉配对特异性依赖关系。其解决方案的关键在于提出BERT4DTI框架，该框架采用ChemBERTa对分子SMILES字符串进行编码，使用ProtBERT对氨基酸序列进行编码，并引入词元级表示之间的双向互注意力机制以建模药物与靶标间的交互特征；随后通过卷积层与多层感知机对融合特征进行分类。为降低参数量，将ProtBERT截断为保留18层，并仅微调每个编码器的最后两层。实验表明，在BIOSNAP、DAVIS和BindingDB三个基准数据集上，BERT4DTI表现优异，尤其在BIOSNAP上达到最优的ROC-AUC与PR-AUC，且在所有数据集上均实现最高灵敏度；在DAVIS上的消融实验进一步验证了互注意力机制对提升PR-AUC与特异性的重要作用。相比完整BERT微调所需的3.53亿参数，BERT4DTI仅需1.25亿可训练参数，实现了性能与参数量之间的良好权衡，为序列型DTI筛选提供了一种高效可行的方案，但其运行时分析、校准及防泄漏验证仍需未来研究完善。
链接: https://arxiv.org/abs/2609.33254
作者: Thanina Hamitouch,Khadidja Henni,Abdelkrim Arie,Amina Selma Haichour,Neila Mezghani,Lina Abou-Abbas
机构: Ecole Nationale Supérieure d’Informatique (ESI), Algiers, Algeria; Institut d’Intelligence Artificielle Appliquée, TELUQ University, Montreal, Canada; Department of Electrical and Computer Engineering, Lebanese American University, Byblos, Lebanon
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted at CIKM 2026
Abstract:Understanding how drugs interact with protein targets is fundamental to drug discovery, drug repurposing and the early identification of promising therapeutic candidates before costly experimental testing. Sequence-based DTI models face three practical limitations: labelled interactions are scarce and unevenly distributed, large pretrained chemical and protein encoders are expensive to fine-tune end-to-end, and independently encoded sequences do not capture pair-specific dependencies. We present BERT4DTI, which encodes SMILES strings with ChemBERTa and amino-acid sequences with ProtBERT, applies bidirectional mutual attention between token-level representations, and classifies the resulting interaction features using convolutional layers and a multilayer perceptron. To reduce trainable size, ProtBERT is truncated to 18 retained layers and only the last two layers of each encoder are fine-tuned. On BIOSNAP, DAVIS and BindingDB, BERT4DTI is competitive, achieving the best ROC-AUC and PR-AUC on BIOSNAP and the highest sensitivity on all three benchmarks. An ablation on DAVIS shows that mutual attention improves PR-AUC and specificity. With 125M trainable parameters compared with 353M for full BERT fine-tuning, BERT4DTI provides a favourable performance-parameter trade-off for sequence-based DTI screening, while leaving runtime profiling, calibration and leakage-audited validation for future work.
[NLP-220] Beyond Memory Construction: Rethinking Memory Access for LLM -based Conversational Agents
【速读】： 该论文旨在解决长时程、高熵对话场景下基于大语言模型（LLM）的内存构建方法所面临的瓶颈问题。现有方法依赖于将原始交互内容重写为结构化记忆单元，再通过检索增强生成（RAG）管道进行访问，但在上下文长度增长和信息复杂度提升时，该范式易导致记忆构建过程出现严重失真与不稳定性，并因频繁调用LLM而带来高昂计算开销。其解决方案的关键在于提出Threader，一种以高效、结构感知的原始交互访问为核心的新内存系统：该系统不再进行记忆重构，而是将原始对话作为第一类记忆保留，通过轻量级增量分段实现话题一致的段落组织，并利用多视角表征支持精准检索；在查询阶段，采用多信号检索机制，融合段级定位与局部证据匹配，兼顾召回完整性与语义连贯性。实验表明，Threader显著提升了答案准确率与证据召回率，同时大幅降低了记忆构建的计算成本。
链接: https://arxiv.org/abs/2609.33226
作者: Donghua Cai,Yongheng Deng,Yifei Wang,Zijun Shen,Ju Ren
机构: Tsinghua University (清华大学); Beijing Institute of Technology (北京理工大学); Nanjing University (南京大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Memory is a core component of conversational agents, enabling coherent and context-aware behavior over long interactions. Recent approaches commonly rely on LLM-based memory construction, where raw interactions are rewritten into structured memory units and later retrieved via a RAG pipeline. While effective in controlled settings, we show that this paradigm breaks down in long-horizon, high-entropy conversations: memory construction becomes increasingly lossy and unstable as context length and information complexity grow, and incurs prohibitive cost due to repeated LLM invocation. To address these limitations, we propose Threader, a memory system that shifts the focus from memory construction to efficient, structure-aware access over raw interactions. Instead of rewriting interactions, Threader preserves them as first-class memory, organizes them into topic-coherent segments via lightweight incremental segmentation, and enables accurate retrieval through multi-view representation. At query time, it performs multi-signal retrieval that combines segment-level access with localized evidence matching, ensuring both completeness and coherence. Extensive experiments demonstrate that Threader consistently improves answer accuracy and evidence recall, while significantly reducing the memory construction overhead.
[NLP-221] When Do Models Admit They Are Wrong? Failure Disclosure Is Unstable Under Reinforcement Learning
【速读】： 该论文旨在解决在仅基于结果（outcome-based）的强化学习框架下，模型在任务执行失败时的错误披露行为（failure disclosure）存在高度不一致的问题。尽管模型在任务准确率上表现相似，但其是否主动承认失败、如实报告错误却表现出显著差异，这种行为差异可能对系统安全性构成潜在风险。解决方案的关键在于：通过抑制模型在失败且结构合理的响应中偏离初始策略（starting policy），从而降低错误披露行为的随机性与变异性。研究发现，即使在固定任务目标和早期训练历史的情况下，微小的浮点计算或采样差异仍可导致报告行为的显著偏移；此外，错误披露并非单一决策过程，而是包含“检查答案”“生成报告”和“完成承认”等多个可分离的环节，其薄弱环节随任务类型和输出格式而异。因此，论文强调，仅保证任务性能稳定不足以确保安全相关行为的一致性，必须直接测量并设计训练机制以增强此类行为的可靠性。
链接: https://arxiv.org/abs/2609.33220
作者: Steven Y. Feng,Noah D. Goodman,Michael C. Frank,Evan Hubinger,Paul C. Bogdan,Andrew Lampinen
机构: Stanford University (斯坦福大学); Anthropic
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Code and data at this https URL
Abstract:Outcome-based reinforcement learning can produce models with similar task performance but very different ways of communicating about their mistakes. We study failure disclosure: whether a model admits that an attempted solution failed rather than staying silent or presenting it as successful. Across repeated outcome-only GRPO training runs, failure disclosure varies far more than task accuracy. The pattern extends to a second reasoning task and stabilized PPO, persists at 7B, and also appears in an instruction-conditioned 32B setting. We also find that small floating-point and sampling differences during training can redirect reporting behavior even when the task objective and earlier training history are held fixed. Additional tests show that failure disclosure is not a single decision: Checking the answer, entering a report, and completing the admission can separate, and the weak point depends on the task and response format. Further, experiments with neutral controls show more broadly that behaviors left weakly constrained by training are especially likely to vary across runs, of which failure disclosure is an example. We can reduce variability in failure disclosure by discouraging the model from drifting from its starting policy on failed, well-formed responses. This makes reporting substantially more consistent, though its effect on task performance depends on the setting. Stable task accuracy therefore does not guarantee stable safety-relevant behavior: Researchers should measure these behaviors directly across runs and design training methods that keep them reliable.
[NLP-222] CoLMbo-SV: A Grounded Language Model for Explainable Speaker Verification
【速读】： 该论文旨在解决语音验证系统（speaker verification system）在高精度决策背后缺乏可解释性证据的问题。现有系统虽具备优异的判别能力，但其判断过程不可见，难以追溯声学层面的支持依据。为此，论文提出CoLMbo-SV，一种结合预训练语音编码器与语言模型的说话人语言模型，通过引入显式的声学测量值，实现对语音对比的结构化、声学基础的可解释报告生成。其解决方案的关键在于：在保留原始语音表示中丰富信息的同时，将可量化声学特征与自然语言报告相结合，使系统既能输出具有数值一致性的解释，又不局限于报告中所描述的内容，从而实现可检查且更具可信度的语音验证。此外，作者构建了VoxReason数据集，包含经数值与质性筛选的成对录音及其声学属性与对比报告，为该联合能力提供监督信号；并设计了一套评估框架，以区分语音表征所编码的声学信息、影响验证分数的因素以及报告内容之间的差异。实验表明，CoLMbo-SV在VoxCeleb1-O上达到0.99%的等错误率（EER），相对于最强的音频-语言基线降低约80%的相对验证误差，同时获得0.82的数值一致性得分。分析进一步揭示，解释的声学正确性与决策相关性是两个独立属性，而现有数值一致性指标未能捕捉这一区别，凸显了建立更精准评估体系的重要性。整体而言，该工作显著推进了音视频-语言融合的说话人验证技术，在保持接近专用说话人编码器的精度基础上，实现了可核查的声学报告生成，并建立了连接自然语言解释与决策依据的实证框架。
链接: https://arxiv.org/abs/2609.33212
作者: Massa Baali,Sarthak Bisht,Ziyue Qiu,Joseph Konan,Rita Singh,Bhiksha Raj
机构: Carnegie Mellon University (卡内基梅隆大学)
类目: Computation and Language (cs.CL); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注:
Abstract:Speaker verification systems achieve high accuracy but provide little account of the acoustic evidence behind their judgments. Making these systems inspectable requires exposing interpretable evidence while retaining the richer information on which their decisions depend. We present \textbfCoLMbo-SV, a speaker language model that combines strong speaker discrimination with structured, acoustically grounded comparison reports. By connecting a pretrained speaker encoder to a language model and supplying explicit acoustic measurements, CoLMbo-SV makes voice comparisons inspectable without restricting verification to the evidence verbalized in its reports. We additionally introduce \textbfVoxReason, paired recordings with measured acoustic properties and comparison reports filtered through numerical and qualitative checks, providing supervision for this combined capability. We also develop an evaluation framework that separates what acoustic information a speaker representation encodes, what influences the verification score, and what the generated report discusses. On VoxCeleb1-O, CoLMbo-SV achieves 0.99% EER, reducing verification error by approximately 80% relative to the strongest audio-language baseline fine-tuned on VoxReason, while attaining a numerical-grounding score of 0.82. Our analysis further demonstrates that acoustic correctness and decision relevance are distinct properties of an explanation, exposing a gap that numerical-grounding metrics miss. Together, these contributions substantially advance audio-language speaker verification, bring its accuracy toward that of dedicated speaker encoders while adding checkable acoustic reporting, and establish an empirical framework for connecting natural-language explanations to the decisions they explain.
[NLP-223] Beyond Calibration: Do a Typed-Decision Models Probabilities Obey the Probability Axioms?
【速读】： 该论文旨在解决生成式模型在处理类型化决策任务时存在的概率不一致性（coherence）问题，即模型对逻辑相关问题所输出的概率未能满足基本的逻辑约束（如互斥事件的概率和应为1）。尽管现有评估指标如准确率与校准性（calibration）关注单个判断的可靠性，却忽视了多任务间概率的一致性需求，而这正是依赖概率进行推理或决策系统所必需的。为此，研究提出了一套无需标注数据的逻辑连贯性测试框架，通过构造包含否定、选择与唯一性判断的成组问题，检验模型输出的概率是否满足逻辑一致性。实验基于ChaosNLI和PubMedQA中的160个样本，每个样本含三个互斥标签，共生成480个否定对。结果显示，Jev模型在“标签是X”与“标签不是X”两概率之和上平均偏离1，偏差达0.064（95% CI: 0.055–0.072），而Qwen3.8-27B在第一标记读出（first-token readout）下的偏差高达0.293，即使使用语义化输出也存在0.122的偏差。更严重的是，Qwen3.8-27B在未显式包含“非”字的疑问中仍系统性低估补集，导致在196/480对中同时拒绝原命题及其否定，表现出严重的非一致性；相比之下，Jev的错误集中在置信度较低的区域，且其三类单标签概率总和平均达1.14，表明其过度支持单一标签。该方法无需人工标注，能够揭示仅在问题形式对比中显现的内在偏倚及项内不一致，从而为评估模型在真实应用中的逻辑可靠性提供了关键基准。
链接: https://arxiv.org/abs/2609.33209
作者: Keyi Li,Yihao He,Quanyi Li
机构: Northeastern University(东北大学); Shenyang, China
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 17 pages, 3 figures, 6 tables. Code and data at this https URL
Abstract:Typed-decision models such as TypeSafe’s Jev answer a declared yes/no or multiple-choice question about a state with a probability instead of text, and their evaluations report accuracy and calibration. Neither requires that the probabilities a model gives to logically related questions fit together item by item, which is what a system that acts on those probabilities needs. We test this property, coherence, with a battery of logically linked questions that needs no labels. For 160 items from ChaosNLI and PubMedQA, each with three mutually exclusive labels, we ask whether the label is X, whether it is not X, whether it is one of the other two labels, and which label applies. On 480 negation pairs, Jev’s probabilities for “the label is X” and “the label is not X” miss summing to one by 0.064 on average (95% CI 0.055 to 0.072). Qwen3.8-27B, run from its official BF16 weights, misses by 0.293 with first-token probabilities and by 0.122 with verbalized probabilities. The gap to the first-token readout persists on pairs where both systems give similar probabilities, without double-negation labels, and after averaging Jev’s repeated calls. Jev is not coherent either: its violations are about five times its repeat noise, and it over-endorses statements about single labels, so that its three single-label probabilities sum to 1.14 on average. The two systems also fail differently. Qwen3.8-27B’s first-token readout under-endorses the complement of a label whether or not the question contains “not”, rejecting both a statement and its negation in 196 of 480 pairs, and it does not become more coherent where it is more confident, whereas Jev’s violations concentrate where its answer is uncertain. Because the checks need no labels, they expose biases that appear only when question forms are compared, and inconsistencies within items.
[NLP-224] urning Speech Language Models into Multilingual Listeners INTERSPEECH2026
【速读】： 该论文旨在解决语音语言模型（Speech Language Models, SLMs）在多语言支持方面存在的显著局限性，即当前的SLMs仅能有效处理少数高资源语言，导致全球数百万使用者无法获得服务。这一问题的核心根源在于缺乏大规模、高质量的多语言语音指令微调数据集。为此，本文提出MULTISPEECHQA，一个大规模、通过合成生成并经人工验证的语音问答数据集，包含23种语言类型多样性的1080万条语音问答对，总时长达9200小时。基于该数据集，研究进一步构建了MULTISPEECH-BENCH，一个涵盖23种语言的多任务评估基准，用于系统评估SLMs的跨语言性能。实验表明，尽管级联式（cascading）系统在该基准上优于开源权重的SLMs，但尚未全面超越所有闭源模型。通过使用MULTISPEECHQA对Qwen 2.5-Omni进行微调，其在多语言任务上的表现得到显著提升。研究结果表明，高质量的合成数据集为低成本提升SLMs的多语言能力提供了可行路径。
链接: https://arxiv.org/abs/2609.33204
作者: Tolúlopé Ògúnrèmí,Dan Jurafsky,Chris Manning,Ahmet Üstün,Martijn Bartelds
机构: 未知
类目: Computation and Language (cs.CL)
备注: Interspeech 2026 Long
Abstract:Speech Language Models (SLMs) that understand spoken language questions support only a few high-resource languages, limiting access to millions of people worldwide. This gap stems from the scarcity of multilingual speech instruction-tuning datasets. We present MULTISPEECHQA, a large-scale, synthetically generated and human-verified dataset comprising 9200 hours of 10.8 million spoken question-answer pairs in 23 typologically diverse languages. Using MULTISPEECHQA, we also introduce MULTISPEECH-BENCH, a multi-task benchmark for evaluating SLM performance on 23 languages. We compare the performance of a cascading system to open-weight and closed SLMs on MULTISPEECH-BENCH and find that the cascading system outperforms open-weight SLMs but not all closed SLMs. We use MULTISPEECHQA to finetune Qwen 2.5-Omni, which improves its performance on our benchmark. Our findings show that high-quality synthetic datasets offer a cheap solution to improving the multilingual capabilities of SLMs.
[NLP-225] ach Yourself Where to Look: On-Policy Attention Self-Distillation for Reasoning
【速读】： 该论文旨在解决基于策略的自蒸馏（on-policy self-distillation, OPSD）在训练推理模型时存在的监督信号局限性问题，即仅依赖密集的词元级（token-level）指导而无法有效传递教师模型对正确解码路径的注意力机制。其核心挑战在于：教师模型虽能访问已验证的解码结果（verified solution），但学生模型无法观测这些关键位置，导致传统方法难以将教师的注意力分布有效迁移至学生可感知的上下文位置。为解决此问题，论文提出基于解码条件的注意力自蒸馏（On-Policy Attention Self-Distillation, OPASD），其关键创新在于引入解码条件注意力蒸馏（solution-conditioned attention distillation）——通过将教师模型对不可见解码词元的注意力投影到学生可见的位置，并进行重归一化处理后进行对齐，从而实现更精准、稳定且高效的监督信号传递。实验表明，OPASD在三个模型规模和四个竞赛级数学基准上均显著优于仅依赖词元级监督的方法，平均准确率提升4.98至8.40个百分点，同时大幅减少生成轨迹长度73.9%、降低计算量72.6%，并实现1.53倍的训练加速，证明了解码条件注意力作为互补监督信号在提升模型准确性、稳定性与计算效率方面的有效性。
链接: https://arxiv.org/abs/2609.33200
作者: Safaeid Hossain Arib,Rabeya Akter,Ismam Nur Swapnil,Md. Faiyaz Abdullah Sayeedi,Md Mofijul Islam,Tasnim Mohiuddin
机构: ACI PLC(ACI PLC); University of Dhaka(达卡大学); BRAC University(BRAC大学); Amazon GenAI(Amazon GenAI); QCRI(QCRI)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:On-policy self-distillation trains reasoning models on their own trajectories using dense token distribution guidance from a privileged teacher with access to a verified solution. This supervision transfers what the teacher predicts without directly transferring where it attends within the preceding context. We introduce On-Policy Attention Self-Distillation (OPASD), which complements token-level supervision with solution-conditioned attention distillation. Because the privileged teacher can attend to verified solution tokens unavailable to the student, OPASD projects teacher attention onto student-visible positions and renormalizes the resulting distribution before alignment. Across three model sizes and four competition-level mathematics benchmarks, OPASD consistently outperforms token-only OPSD, improving average accuracy by 4.98 to 8.40 percentage points. OPASD also avoids the response-length inflation and performance degradation observed with token-only distillation, reducing generated rollout tokens by 73.9% and estimated model compute by 72.6% while training 1.53x faster. These results show that solution-conditioned attention provides a complementary supervision signal that makes on-policy self-distillation more accurate, stable, and compute-efficient.
[NLP-226] AG-CoT: Verified Algorithmic Traces for LLM Program Synthesis on Clifford Circuits
【速读】： 该论文旨在解决生成式人工智能在科学代码生成中产生的可执行程序无法正确计算目标科学对象的问题，具体聚焦于语言模型合成Clifford电路时的准确性缺陷。其核心挑战在于：尽管生成的OpenQASM电路在语法和物理上符合Clifford门约束（Clifford validity），但可能并未正确制备目标稳定子态（stabilizer state），导致计算结果错误。解决方案的关键在于提出一种目标条件化的框架，将目标以紧凑的带符号稳定子生成元形式给出，并引入精确验证器（exact verifier）对生成的电路进行严格验证。通过使用Aaronson-Gottesman链式思维（AG-CoT）推理轨迹作为监督信号，并结合验证器筛选出正确生成的样本进行持续训练，实现了显著性能提升。实验表明，该方法使贪婪解码下的状态等价性准确率相较于仅依赖电路输入的基线模型提高4至6倍；进一步采用验证器过滤后的持续训练策略，在两个独立训练的模型族（3B与7B）中均带来稳定增益。此外，32B规模研究显示，监督模型在语法和Clifford有效性上接近完美，而未受监督的直接模型状态等价率仅为6.14%，即使通过多候选选择提升至超10%。结果证实，算法级推理轨迹监督能带来统计显著的性能提升，且精确验证不可或缺——因为电路的语法与物理正确性并不保证其制备正确的量子态。
链接: https://arxiv.org/abs/2609.33192
作者: Lu Wei,Yufeng Wang,Chenfeng Cao,Lu Pang,Haibin Ling
机构: Stony Brook University (石溪大学); Freie Universität Berlin (柏林自由大学); Westlake University (西湖大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Quantum Physics (quant-ph)
备注: 27 pages, 7 figures, 25 tables
Abstract:Scientific code generation can produce executable programs that fail to compute the intended scientific object. We study this problem in language-model synthesis of Clifford circuits, which prepare the stabilizer states used in quantum error correction and admit exact classical verification. In our target-conditioned framework, each target is given as compact signed stabilizer generators, and an exact verifier checks the generated OpenQASM circuits. We supervise models with Aaronson-Gottesman chain-of-thought (AG-CoT) traces checked by the verifier, and continue training on model generations that the verifier accepts. Across two independently trained model families (3B and 7B), AG-CoT supervision multiplies greedy-decode state-equivalence accuracy by four to six times over circuit-only baselines, and verifier-filtered continuation training adds a further consistent gain atop both. A complementary 32B study shows that supervised models achieve near-perfect syntax and Clifford validity while the strongest direct model reaches 6.14% state equivalence per target, rising to over 10% under verifier-guided selection with multiple candidates. These results show that algorithmic trace supervision gives a large, statistically significant gain in both model families and that verifier-filtered continuation adds a further repeated gain. The persistent gap between Clifford validity and state equivalence confirms that exact verification is necessary: a circuit can be syntactically and physically valid yet prepare the wrong quantum state.
[NLP-227] Identifying Temporal Features within Transcoders for Time Sensitive Factual Recall EMNLP
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）中存在的时序错位（temporal misalignment）问题，其核心根源在于训练语料库的矛盾性导致模型在事实回忆时无法准确把握时间信息。现有缓解策略依赖计算成本高昂的微调或上下文密集的检索增强生成（Retrieval Augmented Generation, RAG），但对内部机制中时间敏感记忆的调控过程仍缺乏深入理解。本文的关键创新在于首次从特征层面构建了时序回忆的解析图谱，通过解耦编码器-解码器电路追踪（transcoder circuit tracing）技术，分离出独立的MLP（多层感知机）特征，揭示了三类关键节点：通用时序特征、仅与特定年份相关的特征以及时序-语义混合特征。这些节点在事实回忆过程中协同形成动态时序过滤器，且并非遵循简单的线性传播路径，而是通过跨层并行、混合的句法-语义交互实现时间表征。研究进一步发现，传统基于激活值归因（EAP-IG）的方法无法探测到高层中的部分时序组件，从而证明了转码器（transcoder）作为更完整视角在时间敏感事实回忆可解释性分析中的优越性。该成果为未来针对时序敏感事实召回的精准干预提供了可操作的MLP特征靶点。
链接: https://arxiv.org/abs/2609.33183
作者: Sanjay Govindan,Yang Song,Maurice Pagnucco
机构: University of New South Wales (新南威尔士大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: EMNLP Findings 2026
Abstract:Large Language Models (LLMs) suffer from temporal misalignment, often due to the contradictory nature of their training corpora. While current mitigation strategies rely on computationally expensive fine-tuning or context-heavy retrieval augmented generation (RAG), the internal mechanisms governing time-sensitive recall remain under-explored. Unlike prior studies that identify temporal components such as attention heads and MLP layers, we provide the first feature-level map of temporal recall by isolating individual MLP features via transcoder circuit tracing. We identify three node categories (common temporal, common to the year, and chrono-semantic) which interact to generate a temporal filter during factual recall. By analysing Gemma 2 2B, LLaMA 3.2 1B, and Qwen3-4B, we show that these features do not follow a simple linear pipeline but represent time through a parallel and mixed syntactic-semantic interplay across layers. We additionally discover a class of higher-layer temporal components invisible to existing EAP-IG methods, establishing transcoders as a more complete lens for temporal interpretability in time-sensitive factual recall. These findings present MLP components for potential targeted interventions in time-sensitive factual recall
[NLP-228] SeOPD: Self-Evolving LLM s via Online Policy Distillation from Self-Generated Chain-of-Thought
【速读】： 该论文旨在解决在线策略自蒸馏（Online Policy Self-Distillation, OPSD）在实际应用中依赖外部特权信息（Privileged Information, PI）所导致的可扩展性受限问题。现有方法虽能通过人工标注或外部环境反馈提升大语言模型（LLM）性能，但其获取成本高且难以大规模部署。尽管已有研究尝试无外部PI的自提升方法，但性能增益有限。本文提出一种名为自演化在线策略蒸馏（Self-Evolving Online Policy Distillation, SeOPD）的新框架，其核心创新在于利用单一LLM自身具备的多种推理模式（如深度思考模式与非思考模式），通过深度思考生成的思维链（Chain of Thought, CoT）作为内部生成的特权信息，对非思考模式的输出进行逐标记层级的监督。该机制使模型在推理过程中新涌现出的信息得以被提炼并内化至共享参数中，从而同步提升非思考与深度思考能力。实验结果表明，SeOPD可在无需外部PI的情况下实现显著且稳定的性能提升，验证了基于模型内在推理过程自生成信息进行自我蒸馏的有效性。
链接: https://arxiv.org/abs/2609.33181
作者: Xiaoshu Chen,Xiangyu Wong,Sihang Zhou,Ke Liang,Xinwang Liu
机构: National University of Defense Technology(国防科技大学); Peking University(北京大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Recent advances in online policy self-distillation (OPSD) have demonstrated that large language models (LLMs) can improve their capabilities by leveraging external privileged information (PI), such as manual annotations or feedback from external environments. However, obtaining accurate annotations and constructing sophisticated environments often require substantial human effort and computation, limiting the scalability of OPSD. While a few recent studies have explored self-improvement without external PI, the resulting gains remain limited. In this work, we explore whether LLMs can achieve comparable self-improvement without external PI. Our key observation is that a single LLM can support multiple reasoning modes, such as deep-thinking and non-thinking modes, with deep thinking generating additional information during reasoning. Based on this observation, we propose Self-Evolving Online Policy Distillation (SeOPD), which enables LLMs to distill and internalize information generated by their own chain of thought (CoT). Specifically, it (1) generates CoT with the deep-thinking mode, (2) produces responses with the non-thinking mode, and (3) uses the generated CoT as PI to provide token-level supervision for the non-thinking response, allowing new information inferred during reasoning to guide the non-thinking mode and be internalized into the shared model parameters, thereby improving both non-thinking and deep-thinking capabilities. Extensive experiments across LLMs and tasks demonstrate the effectiveness of SeOPD.
[NLP-229] Scoring the Wrong Question: Readout Failures in Constrained-Option Evaluation
【速读】： 该论文旨在解决生成式 AI（Generative AI）在受限选项评分（Constrained-option scoring）中因模型偏好提前输出答案字母而导致的评估失效问题。具体而言，当模型在未充分推理的情况下即倾向于输出预设答案选项的首字母时，传统评分机制无法准确反映模型的真实理解能力，从而导致对模型正确性预测的评估结果接近随机水平。其解决方案的关键在于引入一种无需标签的诊断方法：通过分析上下文中各选项的首令牌概率质量（first-token mass），识别出选项支持度不足的问题；结合删除测试以追踪答案字母起始点的来源，并进行分词与渲染检查以排除分隔符丢失或首令牌重合带来的干扰。进一步地，通过在提示中预先填充答案题干（answer stem），显著提升了所有九个检查点上的选项概率均值，使中位数选项质量从低于0.5提升至高于0.5；在此基础上保持选项不变并重新归一化，将Qwen3平均AUC从0.4929提升至0.6152。此外，针对lm-polygraph默认的P(True)估计器在无思考情况下误判为真（True）的问题，通过对答案标签进行后处理归一化，使其在MMLU和TriviaQA上的平均AUROC分别达到0.632、0.868，优于所有14个默认单答案库估计器的表现。然而，原始预测仍无法超越随机水平，表明仅依赖选项质量指标不足以判断真实性能，必须显式验证目标响应是否符合预期意图。
链接: https://arxiv.org/abs/2609.33179
作者: Jiaxuan Guo,Kejia Zhang,Shuo Xin,Jingxin Yang,Youran Sun,Haizhao Yang
机构: Stanford University (斯坦福大学); University of Maryland, College Park (马里兰大学学院帕克分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Constrained-option scoring reads a model’s probabilities for a fixed set of permitted answers, so it returns a score even when the model is about to write something else. We study a prompt that quotes a multiple-choice item, ending in the item’s own answer instruction, and asks for a forecast of whether a reader model given only a short note will answer it correctly. The model can begin answering the quoted item: on three Qwen3 checkpoints an answer letter is the most probable token on nearly every passage, and the forecast ranks the reader’s correctness indistinguishably from chance. Beyond prior work’s argmax check and prefill repair, we contribute a label-free diagnostic of the declared options’ support, their first-token mass in context; a deletion test tracing the Qwen3 answer-letter start to the quoted instruction; and rendering and tokenisation checks for dropped separators and coinciding first tokens. Prefilling an answer stem lifts the median option mass from below to above one half on all nine checkpoints scored with this prompt and, with options and renormalisation unchanged, raises the Qwen3-averaged AUC from 0.4929 to 0.6152. lm-polygraph’s default P(True) estimator shows a related failure: it reads True where Qwen3 without thinking favours an answer letter (MMLU) or the prompt’s (A)/(B) label (TriviaQA), and its Qwen3-averaged AUROC is indistinguishable from chance on MMLU and below chance on TriviaQA. Scoring and renormalising the (A)/(B) labels after an answer stem raises it to 0.632 and 0.868, renormalising True against False in place reaches 0.660 and 0.860, and on TriviaQA both exceed all fourteen default single-answer library estimators. The original forecast is renormalised too, yet stays indistinguishable from chance: option mass flags low support in both settings without labels, and only checking the intended target shows which score still ranks.
[NLP-230] Where Do Test-Time Scaling and Training Fall Short in Individual Stance Prediction?
【速读】： 该论文旨在解决生成式 AI（Generative AI）在个体立场预测任务中，尽管在代码生成与数学推理等任务上通过测试时缩放（test-time scaling）和后训练（post-training）技术显著提升性能，但其在个体立场预测场景下的有效性仍不明确的问题。核心挑战在于如何准确基于个体历史言论预测其在新讨论中的立场。论文的关键解决方案是：通过构建名为 STANCE-BENCH 的基准数据集（包含 500 名 Hacker News 用户的 2499 个预测任务），系统识别出生成、选择与学习三个环节中存在的四大失效模式，包括错误共识、选择失败、响应过拟合及早期饱和，并提出一种结合直接得分与个体历史支持度显式评估的简化方法。该方法在 781 个测试任务上使 Qwen3-8B 模型的讨论特定宏平均 F1 达到 21.83，优于仅使用直接评分的 19.27，从而强调了对候选生成、最终选择及个人证据利用进行独立评估的重要性。
链接: https://arxiv.org/abs/2609.33155
作者: Yuyang Zhao,Xuan Liu,HaoYang Shangm Haojian Jin
机构: University of California, San Diego(加州大学圣地亚哥分校)
类目: Computation and Language (cs.CL)
备注:
Abstract:Test-time scaling and post-training have improved LLM performance in coding and mathematical reasoning, but their effectiveness for individual stance prediction remains unclear. We study this question by predicting a person’s stance in a new discussion from their history. We evaluate widely used test-time scaling strategies and post-training methods, such as supervised fine-tuning and reinforcement learning, and identify four failure modes across generation, selection, and learning: (1) incorrect consensus, where repeated samples agree on the wrong stance; (2) selection failure, where generation covers the observed stance but selection misses it; (3) response overfitting, where supervised fine-tuning improves imitation but harms prediction; and (4) early plateau, where reinforcement learning shows modest initial gains followed by limited further improvement. We expose these failures using STANCE-BENCH, which contains 2499 prediction tasks from 500 Hacker News users. Guided by this analysis, we explore a simple approach that combines direct scores for all candidate stances with explicit assessments of support from the individual’s history. On the 781-task test set, this approach achieves 21.83 discussion-specific Macro F1 with Qwen3-8B, compared with 19.27 for direct scoring. Our results motivate evaluating candidate generation, final selection, and person-specific evidence use separately.
[NLP-231] Generalization Dynamics of LM Pre-training
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在预训练过程中是否稳定地从模式匹配（parrot-like）行为逐步演变为具备泛化能力的通用智能（intelligence-like）这一广泛假设。研究发现，这种线性演进的认知模型是错误的：在预训练全过程中，模型频繁且突然地在“鹦鹉学舌式”计算与“类智能”计算之间发生模式跃迁（mode-hopping），表现为模型时而依赖记忆或上下文中的表面模式进行响应，时而表现出真正的推理与泛化能力。这种动态行为无法用传统优化机制解释，且在局部稳定，无法通过检查点平均等方法修复。其关键解决方案在于将该现象视为容量受限模型中的“能力分配问题”——早期学习的浅层、快速形成的电路与后期发展出的深层泛化电路之间存在竞争关系，而每个预训练窗口的数据分布决定了哪类电路占据主导。基于此，作者构建了一个高效的评估工具集（toy eval suite），并展示了两个实际应用：一是可从中筛选出优于最终或中期检查点的中间预训练阶段模型，实现更强的推理与对齐能力；二是可通过选择特定预训练数据来调控和稳定模型的泛化动态。
链接: https://arxiv.org/abs/2609.33150
作者: Jiaxin Wen(UC Berkeley),Zhengxuan Wu(Stanford University, Google DeepMind),Dawn Song(UC Berkeley),Lijie Chen(UC Berkeley)
机构: UC Berkeley(加州大学伯克利分校); Stanford University(斯坦福大学); Google DeepMind(谷歌深思维)
类目: Computation and Language (cs.CL)
备注: 31 pages, 43 figures
Abstract:People typically assume that LMs stably mature from pattern-matching parrots to generalizable intelligence during pre-training. We build a toy eval suite and show this mental model is wrong: throughout pre-training, LMs frequently and suddenly hop between parrot-like and intelligence-like computations. We call this mode-hopping. Across our suite, LMs suddenly latch onto memorized or in-context patterns instead of in-context learning, use System 1 instead of System 2 thinking, pick up what sounds true instead of what is true, fail at multi-hop persona QA, out-of-context reasoning, and emergent misalignment – then just as suddenly revert and generalize. Mode-hopping is not explained by standard optimization dynamics: it is locally stable and cannot be fixed by checkpoint averaging. We instead think of it as a capacity allocation problem: in a capacity-bounded model, generalizable circuits must compete with the shallow ones learned early in training, and the data in each pre-training window may decide which circuits win. Our suite provides a new efficient lens on generalization. We demonstrate two concrete applications: (i) select intermediate pre-training checkpoints that strongly generalize reasoning and alignment, better than the final pre- or mid-training checkpoints, and (ii) select pre-training data that controls and stabilizes generalization dynamics.
[NLP-232] CAME: Company-Aware Evidence-Memory Experts for Interpretable Quarter-Ahead Revenue Forecasting EMNLP2026
【速读】： 该论文旨在解决季度收入预测中面临的三大核心挑战：公司层面的数值精度要求、严格的时间有效性保障，以及对管理层披露文本的公司特异性解读。现有方法存在明显局限：大语言模型（Large Language Models, LLMs）虽能有效提取文本证据，但常导致预测尺度错配；基于历史数据的统计锚点虽具备稳定性，却难以捕捉预测时点的关键信号，如产品迭代、供应链约束及管理指引等。为此，本文提出一种名为CAME（Company-Aware Evidence-Memory Experts）的残差式预测框架，其关键在于通过引入“无信息泄露”的统计锚点，并在当前语义证据与历史误差模式共同支持的情况下，动态调整预测结果。该方法在包含12家大型科技与平台企业共336个公司-季度的滚动回测中表现最优，相较匹配的统计锚点在宏观sMAPE指标上实现统计显著提升，且在全部六项聚合评估指标上优于“历史+指引”组合。此外，CAME通过可追溯的证据卡片与受控记忆痕迹，实现了预测调整的来源关联、可解释性与故障定位能力，显著增强了模型的透明性与可信度。
链接: https://arxiv.org/abs/2609.33143
作者: Ya-Wen Wu,Meng-Fen Chiang,Kuang-Da Wang,Wen-Chih Peng
机构: National Yang Ming Chiao Tung University (国立阳明交通大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted to EMNLP 2026 Main Conference
Abstract:Quarter-ahead revenue forecasting requires company-scale numerical accuracy, strict temporal validity, and company-specific interpretation of narrative disclosures. LLMs can distill textual evidence but can produce scale-misaligned forecasts, whereas history-based anchors are stable but miss forecast-time signals such as product transitions, supply constraints, and management guidance. We introduce CAME (Company-Aware Evidence-Memory Experts), a residual-forecasting framework that refines a no-leakage statistical anchor when current semantic evidence and prior error patterns justify an adjustment. On a development-inclusive rolling backtest of 336 company-quarters from 12 large public technology and platform firms, CAME achieves the lowest aggregate point-estimate error among the reported methods, with statistically supported macro-sMAPE gains over the matched Statistical Anchor, and outperforms History + Guidance on all six aggregate metrics. CAME also links adjustments to source-linked evidence cards and guarded memory traces, supporting forecast inspection, provenance, and failure localization.
[NLP-233] Knowing Is Not Choosing: What Explicit Verification Adds Beyond Generative Preference
【速读】： 该论文旨在解决大语言模型在事实性问答任务中“生成正确答案但未能正确选择”的核心问题，即模型虽能生成正确的候选答案，却在多个候选答案中无法有效筛选出最优解。其解决方案的关键在于引入显式验证机制（explicit verification）——通过计算真实概率 $ P(\mathrm{True}) $ 来对候选答案进行排序，从而显著提升模型在单个问题内部的候选答案排序能力。相较于传统的均值对数似然（mean log-likelihood）方法，该方法在Gemma、Qwen3和Llama三个模型家族中均实现了0.08至0.12的AUROC提升。在前瞻性定义的Gemma实验组中，该方法使多数选择准确率（plurality accuracy）提升约5个百分点，并仍优于基于聊天模板的生成似然这一更强基线，表现优势在具有常见答案先验关系且可访问实体信息时最为显著；当实体被遮蔽时，大型Qwen模型中的排序优势消失，表明实体上下文对验证有效性至关重要。此外，研究发现验证效果的实际测量结果高度依赖于“正确性”定义方式：若采用以召回为导向的参考匹配策略，可能低估人类语义判断所揭示的真实改进幅度，凸显了评估标准对结论的影响。
链接: https://arxiv.org/abs/2609.33142
作者: Yilong Li,Chengpo Yan,Aayan Arish,Suman Banerjee
机构: University of Wisconsin-Madison(威斯康星大学麦迪逊分校)
类目: Computation and Language (cs.CL)
备注:
Abstract:Generating a correct answer does not mean that a language model will select it. We separate factual recall into three steps: generating a correct candidate, ranking the available candidates, and selecting the final answer. Pre-generation readouts predict factual recall and which questions sampling will cover across three model families, but say little about whether an available correct answer will ultimately be selected. Explicit verification with P(\mathrmTrue) improves within-question ranking over mean log-likelihood in Gemma, Qwen3, and Llama, with AUROC gains of 0.08 – 0.12 . In a prospectively defined Gemma cohort, verification raises plurality accuracy by about 5 points, and still gains about 2 points over chat-template likelihood, a stronger generative baseline. The advantage is strongest for relations with common-answer priors and depends on access to the entity; masking the entity removes the ranking advantage in larger Qwen models. Finally, the measured benefit depends on how correctness is defined: recall-oriented reference matching can credit option lists favored by likelihood and substantially understate the improvement seen under human semantic judgments.
[NLP-234] Save Your Saturated Data: Learning Beyond Reward Saturation in Group-Based RL
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）能力增强后导致的训练数据奖励饱和问题：当模型生成的多个响应均获得极高且相近的奖励时，基于组相对优势（Group-relative Reinforcement Learning, GRPO）的强化学习信号会因奖励差异消失而失效，致使原有数据失去价值。其解决方案的关键在于通过在强化学习流水线的四个层级——数据、采样（rollout）、奖励和优势计算——引入多样化干预策略，从原本“无用”的饱和数据中重新提取有效学习信号。研究发现，最有效的策略是在采样生成阶段引入干预，即引导模型生成“高质量但错误”的解作为负样本，使原本高奖励的响应组中混入低质量响应，从而恢复非零的相对优势。这一方法在Qwen3-1.7B与4B模型上实现了6.4%至9.0%的性能提升，显著优于仅调整温度或添加辅助奖励等其他手段。进一步分析表明，有效的负样本需具备信息量丰富的负向轨迹，且该方法可与未饱和数据共存，并支持对新出现的饱和样本进行迭代回收利用。研究结论强调，在日益稀缺的数据环境下，不应丢弃饱和数据，而是通过合理设计的干预机制将其转化为持续可用的强化学习信号。
链接: https://arxiv.org/abs/2609.33126
作者: Ziyuan Yang,Yike Wang,Shangbin Feng,Yulia Tsvetkov
机构: University of Washington(华盛顿大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 17 pages, 6 tables, 3 figures
Abstract:Group-relative reinforcement learning (RL) relies on reward variation among sampled responses to estimate informative relative advantages. As language models become increasingly capable, existing training data can become reward-saturated: all sampled responses to the same problem might receive equally high rewards, where the group-relative learning signals vanish and leave previously useful data obsolete. In this work, we investigate whether useful learning signals can be recovered from such saturated data. We study interventions at four levels of group-policy RL pipelines—data, rollout, reward, and advantage—and conduct extensive RL training on saturated reasoning data only. While standard GRPO on saturated data would almost always yield near-0 advantages and near-noise signals, diverse interventions successfully recycle and repurpose such data: among the proposed strategies, interventions at rollout generation are consistently most effective: nudging the policy to generate ``high-quality’', incorrect solutions introduces rollouts with poor rewards into saturated groups as negative samples, which turns out to improve GRPO by 6.4% to 9.0% across Qwen3-1.7B and 4B. Other interventions such as increasing rollout temperature or adding auxiliary rewards can also restore non-zero advantages, but yield less consistent gains. Further analyses show that effective negative rollouts require informative negative trajectories, that the method remains effective alongside unsaturated data, and that it supports iterative recycling of newly saturated examples. While increasingly stronger LLMs would render more data as saturated, our results demonstrate that don’t waste your saturated data: with the right strategies they can be recycled into useful RL training signals in an increasingly data-scarce world.
[NLP-235] Classifying Dominant Temporal Orientation without Pretrained Text Embeddings: A Novel Morphosyntactic Inventory Vector Approach
【速读】： 该论文旨在解决自然语言处理中句子整体时态指向（global temporal orientation）判定的难题，尤其针对包含多个嵌套从句且存在时态与体貌信息冲突的复杂句式。传统计算方法在处理此类句子时表现不佳，主要受限于对上下文依赖的建模能力不足以及对预训练嵌入（pretrained embeddings）的依赖。本文提出的解决方案关键在于引入一种无需预训练嵌入、不依赖序列模型或大型语言模型的全新表示方法——“形态句法向量”（Morphosyntacton）。该向量通过固定长度的统计特征组合构建，包括词性标注频数、依存关系类型频数以及显式的未来时模式计数，全面捕捉句子的形态句法结构信息。实验基于1,799个经过时态标注的复杂英语句子，结果显示该方法在三分类任务中实现了92%的综合多类准确率，且各类别表现均衡，验证了其在无序列建模与无外部语义先验条件下的有效性。
链接: https://arxiv.org/abs/2609.33121
作者: Jonathan Cleveland,Peter S. Bearman
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Computational methods have consistently struggled to determine the dominant temporal orientation of a sentence. This difficulty is especially pronounced when a sentence contains multiple embedded clauses with competing tense and aspectual information. To address this difficulty, we propose an alternative approach for identifying a sentence’s past, present, or future global reference interval. Our method does not use any form of pretrained embeddings. We rather encode sentences using fixed-length inventory vectors that are comprised of part-of-speech counts, dependency relation counts, and explicit futurate pattern counts. We term this inventory vector of a sentence a “Morphosyntacton”. The method does not use any padding, sequence models, or large language models. Evaluation on 1,799 syntactically complex English sentences, annotated as past, present or future, shows balanced and high accuracy multiclass classification, achieving an overall multi-class accuracy of 92%.
[NLP-236] MedRouter: Demystifying Knowledge Differences Across Medical LLM s for Routing-Based Reasoning
【速读】： 该论文旨在解决医学问答任务中因专业领域多样性和模型能力异质性导致的覆盖范围不足问题，即单一医学大语言模型（LLM）在面对跨学科、多模态的复杂医疗问题时存在能力局限。其核心挑战在于如何有效利用不同专家模型在特定任务或领域中的互补优势，而非仅关注答案质量与推理成本之间的权衡。解决方案的关键在于提出MedRouter——一个基于嵌入的多标签路由代理系统，通过动态选择最适配的专家模型并整合其输出，最终由生成器融合生成答案；同时引入SCALE（Specialist Competence-Aware Learning）两阶段训练框架，第一阶段以专家正确性为监督信号训练路由模块，第二阶段采用性能增益奖励（Performance Gain Reward, PGR）机制，通过衡量引入特定专家信息后生成答案正确性的提升程度，优化路由决策。实验表明，该方法在8个文本及多模态医学问答基准上平均准确率较最强基线提升8%，且对专家输出的分析揭示了其在问题层面的差异化优势与互补性，验证了基于学习的路由策略在实现更全面医学推理方面的有效性。
链接: https://arxiv.org/abs/2609.33119
作者: Lang Cao,Binghang Lu,Yuhao Shen,Yue Guo
机构: University of Illinois Urbana-Champaign(伊利诺伊大学香槟分校); Purdue University(普渡大学); The Chinese University of Hong Kong, Shenzhen(香港中文大学（深圳)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Medical question answering spans diverse specialties and modalities, and individual medical large language models (LLMs) exhibit distinct strengths across tasks and domains. This heterogeneity suggests that combining specialists may enable broader coverage of medical questions than relying on any single model. However, existing LLM routing methods primarily seek to balance answer quality and inference cost, leaving open how to exploit differences in specialist competence to improve medical reasoning. In this paper, we introduce MedRouter, an agentic system that uses an embedding-based multi-label router to select and query specialist LLMs, then passes their responses to a generator to produce the final answer. We further propose SCALE (Specialist Competence-Aware Learning), a two-stage training framework that first trains the Router with specialist correctness supervision and then optimizes its selections through reinforcement learning. The second stage uses a Performance Gain Reward (PGR) that measures how specialist information affects the generator’s answer correctness relative to answering without that information. Experiments on eight text-based and multimodal medical QA benchmarks show that MedRouter outperforms the strongest routing baseline by 8% in average accuracy. Our analysis of specialist outputs further reveals distinct strengths and complementary question-level coverage, motivating learned routing to combine these capabilities for more comprehensive medical reasoning.
[NLP-237] ECG-Scroll: A Long-Horizon Streaming Benchmark and Agent Environment for Interpretation of Ambulatory Electrocardiograms
【速读】： 该论文旨在解决长时程心电图（ECG）实时监测中临床关键事件检测的挑战，即在持续数小时至数天的动态心电监测（如Holter或遥测记录）中，如何在线、逐段地识别短暂且偶发的心律失常事件。传统多模态大语言模型（MLLM）虽能处理标准10秒、12导联心电图并进行可验证的推理，但无法应对真实临床场景中信号流式输入、未来信息不可知、需即时决策与记忆累积的复杂需求。为此，作者提出将长时程心电图解读重构为一个长时程、在线（流式、因果）的序列决策过程，并引入名为ECG-Scroll的基准测试框架。其核心解决方案在于构建一个固定、类gym环境的智能体交互平台，通过分块流式输入长时程双导联心电图数据，要求智能体在无未来信号的前提下，实时完成事件定位、量化及报警，同时保留原始信号以实现基于客观真值的规则化奖励机制。该设计强制智能体具备三大能力：记忆（Memory）——维持对历史信号的上下文理解；工具使用（Tool use）——基于信号特征而非像素进行定量测量；规划（Planning）——决定当前应测量什么以及何时提交证据。研究释放了涵盖2,536小时的390个完整记录实例，并评估了基于阈值规则的智能体与现成大语言模型（LLM）智能体的表现，系统分析其在记忆、工具使用和规划上的差异，揭示了当前方法在该任务中的性能提升空间。
链接: https://arxiv.org/abs/2609.33117
作者: Haitao Li,Chenglin Li,Zhengyao Ding,Ziyu Li,Yiheng Mao,Zhengxing Huang
机构: Zhejiang University (浙江大学); Shanghai Innovation Institute (上海创新研究院)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Signal Processing (eess.SP)
备注:
Abstract:Multimodal large language models (MLLMs) can now interpret a standard ten-second, twelve-lead electrocardiogram (ECG) with clinically grounded, reward-verified reasoning. Real cardiac monitoring is different. Ambulatory (Holter) and telemetry recordings span hours to days and are read as they stream in, and their clinically decisive findings are paroxysmal, brief episodes buried in an otherwise unremarkable trace. Such a recording cannot be held in one context at diagnostic resolution, and its future has not yet happened, so a reader must work online, deciding what to measure now, committing evidence to memory as it passes, and reporting events as they occur. We recast long-duration ECG interpretation as a long-horizon, online (streaming, causal) sequential decision process and introduce ECG-Scroll. As a benchmark, long ambulatory recordings are streamed to an agent chunk by chunk, and it must localize, quantify, and promptly flag paroxysmal events without access to future signal; because the underlying signal is retained, every answer is checkable against objective ground truth, giving rule-based rather than judge-based rewards, and the streaming formulation adds a metric batch evaluation cannot express, the detection latency between an event’s onset and the moment the agent records it. As an agent environment, it is a fixed, gym-style interaction layer that exercises three competencies single-glance ECG models never touch: Memory, Tool use through signal-grounded measurement rather than reading pixels, and Planning of what to measure now and when to commit. We release 390 whole-recording instances spanning 2,536 hours of two-lead ambulatory ECG and evaluate a signal-threshold rule agent alongside off-the-shelf LLM agents online, characterizing how they use memory, tools, and planning and where the benchmark’s head-room lies.
[NLP-238] OneSign: Unifying Sign Language Understanding Tasks with One Model
【速读】： 该论文旨在解决手势语言理解（SLU）任务中存在的一系列关键问题：首先，现有方法在训练与推理流程上高度碎片化，普遍依赖大规模手语数据集的预训练，再针对特定任务或数据集进行微调，导致模型体系复杂、难以复用，形成多个独立的专用模型而非统一的单一检查点；其次，当前基于大语言模型（LLM）的方法往往忽视手语与文本在模态特性上的本质差异，简单地将二者令牌拼接并使用相同的解码器层处理，未能充分建模两种模态间的异质性。为应对上述挑战，本文提出OneSign，一个统一框架，能够在单一模型和单一检查点下实现手语语音识别（ISLR）、连续手语识别（CSLR）与手语翻译（SLT）等多项任务的联合建模。其核心创新在于引入一种模态自适应混合专家（Modality-Adaptive Mixture-of-Experts, MA-MoE）架构，该架构包含共享专家及分别针对手语与文本令牌的模态特异性专家，通过模态路由机制动态激活相应专家，并聚合输出以生成最终的令牌表示。该设计既实现了对不同模态特征的差异化建模，又保留了共享路径以促进跨任务知识迁移。大量实验结果表明，OneSign在多个基准测试上均达到竞争力或领先水平，验证了其作为统一式SLU模型的有效性。
链接: https://arxiv.org/abs/2609.33090
作者: Shiwei Gan,Yafeng Yin,Xiao Liu,Desibieer Tuerdaken,Lei Xie,Sanglu Lu
机构: Nanjing University (南京大学); State Key Laboratory of Novel Software Technology (新型软件技术国家重点实验室)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:SLU encompasses a diverse set of tasks, including ISLR, CSLR, and SLT. Although these tasks share basic semantic and linguistic foundations, they are typically addressed with task-specific architectures and training pipelines, which hinders knowledge sharing and requires costly pretraining and finetuning for each task. In this paper, we focus on two aspects of SLU tasks: (1) training and inference pipelines are highly fragmented: most methods rely on pretraining on large-scale SL datasets followed by task- or dataset-specific finetuning, which leads to multiple specialized models rather than a single checkpoint. (2) current LLM-based methods may overlook the inherent modality discrepancy between sign and text tokens, simply concatenating them and processing both modalities with the same decoder layers. In this paper, we present OneSign, a unified framework that addresses multiple SLU tasks within a single model and a single checkpoint. OneSign reformulates ISLR, CSLR, and SLT under a single training paradigm. To accommodate the heterogeneous characteristics of sign and text representations, we introduce a Modality-Adaptive Mixture-of-Experts (MA-MoE) architecture, consisting of a shared expert and modality-specific experts for sign and text tokens. A modality router dynamically activates the corresponding experts, and their outputs are aggregated to form the final token representations. By enabling modality-dependent expert specialization while preserving a shared expert path, MA-MoE can effectively model the modality differences between continuous sign representations and discrete text tokens. Extensive experiments on multiple benchmarks demonstrate that OneSign achieves competitive or state-of-the-art performance on several benchmarks, highlighting its effectiveness as a unified SLU model. Datasets are available at : this https URL.
[NLP-239] Reading Too Much into Context: Passive Exposure Can Steer LLM Decisions
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在处理用户请求时，因被动引入外部信息（如网络搜索结果或在线观点）而可能导致决策偏移的问题。尽管这些外部内容并未提供改变原有判断的合理依据，但其存在仍可能潜移默化地影响模型输出。研究发现，无论开放权重还是封闭权重模型，外部内容的引入均系统性地改变了模型决策，部分封闭权重模型的决策变化幅度高达近50个百分点；且这一影响不仅限于主观偏好，还可能导致模型违背用户明确指令、增加对虚假陈述的接受度。其解决方案的关键在于：识别并控制上下文中的非必要外部信息输入，以维护模型决策的稳定性与忠实性，从而确保生成结果仅基于任务要求和内在推理，而非无关外部内容的干扰。
链接: https://arxiv.org/abs/2609.33065
作者: Yuxiang Zheng,Lin Tian,Marian-Andrei Rizoiu
机构: Behavioral Data Science Lab; University of Technology Sydney (悉尼科技大学); Sydney, Australia
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 34 pages, 5 figures
Abstract:Large language model (LLM) assistants can now search the web and consult external sources while completing user requests. These sources can provide useful evidence, but they can also introduce additional content into the model’s context. Can such passive exposure steer a decision even when the added content provides no reason to change it? We examine the stability of model decisions on the same tasks with and without such external content. Across all open-weight and closed-weight models we test, exposure systematically shifts decisions, with effects reaching nearly 50 percentage points in closed-weight models. The same pattern appears with real-world online opinions. The influence also extends beyond subjective preferences. Such exposure can steer models toward choices that violate explicit user requirements and increase their acceptance of false claims. In short, what enters an LLM’s context can influence its decision even when it should not determine it.
[NLP-240] Pinned and Still Unstable: Within-Judge Verdict Variance and the Noise Floor of LLM -as-Judge Leaderboards
【速读】： 该论文旨在解决当前大语言模型（LLM）评估体系中一个关键却被忽视的问题：即在云服务基础设施上，将判官模型（judge model）固定于某一特定快照并以零温度解码时，其评估结果的可重复性假设实际上并不成立。研究发现，即使在相同输入与配置下，同一判官模型在多次运行中仍会产生不一致的评判结果，表现为平均约5%的单个样本翻转率，而在争议性高、决定排行榜排名的关键样本上，翻转率高达40%，且不同判官模型间的稳定性差异可达40倍（从0.13%到近10%）。这一不稳定性并非源于特定模型家族，而是由云平台服务架构下隐含的非确定性因素所导致，核心问题在于有限提示（prompt）采样带来的噪声。解决方案的关键在于引入一套专门针对此不稳定的评估指标：包括单样本翻转率、两部分稳定性轮廓（波动比例与条件强度）以及邻近可分离性，并提出一种轻量级、低成本的报告协议——要求进行多轮判官重运行、公开稳定性分析与邻近区间，并采用至少来自不同模型家族的两个判官进行交叉验证。该方案旨在揭示评估过程中的真实噪声水平，避免排行榜呈现未经校准的点估计，从而提升评估结果的可信度与可比性。
链接: https://arxiv.org/abs/2609.33044
作者: Krishna Chytanya Ayyagari
机构: 未知
类目: Computation and Language (cs.CL)
备注: 50 pages, 3 figures
Abstract:Modern LLM evaluation assumes that pinning a judge to a fixed model snapshot and decoding at temperature zero yields reproducible verdicts. We show this assumption fails as a property of how LLM-as-Judge is operationalized on cloud serving infrastructure, not of any particular model family. Across four frontier judges all served via a single major enterprise cloud platform and three standard benchmarks (Arena-Hard, AlpacaEval 2, MT-Bench), identical inputs to the same pinned, temperature-zero judge produce different verdicts across re-runs: per-item flip rates of roughly 5% on average and about 40% on the close-call items that decide leaderboard margins, with a per-judge magnitude spanning a 40x range (from 0.13% to nearly 10%). We introduce metrics tailored to this instability: per-item flip rate, a two-part stability profile (waver fraction and conditional intensity), and adjacency separability - and report what the variance does and does not do to rankings. For a single judge the aggregate ranking is stable (0% top-K instability, 0% pooled winner flip); what degrades is precision: under a paired hierarchical bootstrap, roughly one-fifth to three-quarters of adjacent leaderboard positions are statistically indistinguishable, a noise floor we attribute primarily to finite prompt sampling rather than to the judge. Across judges, leaderboards agree on the coarse ordering but diverge in the middle (Kendall’s tau as low as 0.42-0.64 between families on Arena-Hard), and of 13 published head-to-head ranking claims we re-judge, 5 fail under a defensible judge swap or re-run. We argue that leaderboards report unhedged point estimates that misrepresent the noise floor of the instrument, and we propose a minimal, low-cost reporting protocol: several judge re-runs, published stability profiles and adjacency intervals, and results under at least two judges from different families.
[NLP-241] Improving the Diversity of LLM Outputs without a Trade-off
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）生成结果缺乏多样性的问题，尤其是在不改变词元（token）边际分布的前提下实现高效且可控的输出多样性提升。其核心挑战在于如何在保持原有生成分布不变的同时，避免重复或相似的输出模式，从而增强生成内容的丰富性与创造性。解决方案的关键在于提出DAST（Diversifying Arithmetic Sampling with TokenTour）方法：通过预先对词元ID进行重新排序，使语义相近的词元在序列中相邻排列，这一重排过程可在数秒内完成且可复用于所有后续生成任务；随后将此有序结构与算术采样（arithmetic sampling）或准蒙特卡洛方法结合，在不引入显著生成延迟（仅增加几微秒）的情况下，有效降低在多次生成中出现相似词元的概率，从而在保持原分布特性的同时显著提升输出多样性。实验表明，该方法不仅生成更高质量的创意性内容，还在ProtoQA下游任务上实现了性能的显著提升。
链接: https://arxiv.org/abs/2609.33038
作者: Ryoma Sato
机构: National Institute of Informatics (日本国立情报学研究所)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:We propose DAST (Diversifying Arithmetic Sampling with TokenTour), a method that increases the diversity of LLM outputs without any change to the marginal distribution and with negligible generation-time overhead (a few microseconds). We observe that token IDs are often arranged in a meaningless order and reassign them so that tokens with similar meanings appear consecutively. This can be done in advance in a few hundred seconds per model, and the resulting order can be reused for all subsequent generations. By combining this order with arithmetic sampling (or quasi-Monte Carlo methods), we make similar tokens less likely to be generated across runs while preserving the distribution. Our method not only produces qualitatively good ideas but also significantly improves performance on the downstream task of ProtoQA.
[NLP-242] rust and Task Completion in the World of Consumer AI Agents
【速读】： 该论文旨在解决生成式AI作为行动代理（action agent）在执行用户委托任务时存在的信任与完成度双重挑战：一方面，代理可能在未经用户明确同意的情况下擅自采取行动，导致隐私泄露、资金超支或误信外部指令等信任危机；另一方面，当任务复杂或遇到障碍时，代理容易中途放弃，造成任务完成率低下。其解决方案的关键在于构建一个集成评估框架，通过模拟真实世界中的企业网站、邮箱和电话系统，以及具备反馈能力的虚拟用户，对代理在统一运行中同时进行信任（trust）与完成度（completion）的量化评估。该评估不仅考察模型本身的能力，更聚焦于“模型-提示词-工具-上下文-安全机制”构成的整体系统（harness），尤其强调安全护栏（guardrails）的作用。实验结果表明，采用完整防护机制的Fo系统在71%的任务中成功完成，且在94%的陷阱测试中维持了用户信任，显著优于基础模型及未启用护栏的版本，也远超开源代理OpenClaw的表现。因此，论文提出的核心观点是：唯有将信任与完成度联合评估，并以系统级设计保障安全，行动代理才能真正被赋予实际工作任务。
链接: https://arxiv.org/abs/2609.33017
作者: Jeroen Olieslagers,Eduardo Pujol,Gal Zahavi,Lukas Ingemarsson,Shivani Poddar
机构: Wajo AI
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 25 pages, 4 figures, 8 tables
Abstract:Action agents do things for people. They send email, spend money, and call businesses while the user is busy with something else, so a mistake can turn into an action before anyone notices. They fail their users in two ways. They break trust when they do something the user never agreed to, or hold back after the user clearly said go. And they fall short on completion when they give up on errands that turn out to be hard. Both depend heavily on the harness around the model, meaning its instructions, tools, context, and guardrails. We built an evaluation that scores trust and completion on the same runs, in a simulated world of businesses with their own websites, inboxes, and phone lines, and of people who write back. A simulated user answers the assistant’s questions. Trust means that nothing happens the user did not agree to. No email goes to someone they never approved, no private detail ends up on a group thread, no money is spent past their limit, no stranger’s instructions are followed, and nothing is claimed without a source. Every trap has a matched control in which acting is the right call. We use the evaluation to measure Fo, Wajo’s personal assistant, against a base model with basic instructions on three foundation models, and against the Fo harness with its guardrails switched off. Fo completes 71% of the errands and keeps the user’s trust on 94% of the trap runs. The base models complete 50% to 64% and keep trust on 59% to 75%. On the matched controls, Fo goes ahead slightly less often. OpenClaw, a popular open-source assistant given the same access, completes 42% of the errands it shares with Fo, against 71%, and keeps the user’s trust on 74% of the shared trap runs, against 94%. Measuring trust and completion together, on the whole system rather than the model alone, is how we think action agents become safe to hand real work to.
[NLP-243] CMQA: A 38K-Question Traditional Chinese Medicine Benchmark with a Licensed-Practitioner Reference
【速读】： 该论文旨在解决生成式AI在传统中医（Traditional Chinese Medicine, TCM）领域缺乏系统性评估基准的问题。当前医学语言模型的评测体系几乎完全基于西方生物医学，而中医作为具有独立诊断体系与文献积累的医学系统，长期未被充分量化评估。现有少量中医评测数据集规模小、范围窄且缺乏人类专家参考标准。为此，本文提出TCMQA，一个包含38,279道来自中国中医执业资格考试的开放基准数据集，并配以101名持证中医师提供的15,151条人工标注答案。研究评估了9个模型家族共29个指令微调模型（参数量介于0.27B至14.8B之间），发现模型在该任务上的准确率跨度达59个百分点，且无模型达到性能饱和。关键发现在于：预训练数据的语言属性远比模型规模更能预测其中医能力——一个120亿参数的西文预训练模型仅达39.6%准确率，而一个仅为该规模八分之一的中文预训练模型则达到60.8%。此外，九个表现优异的模型均来自同一中文预训练家族，且其中最佳模型相较医师多数投票基准（64.9%）高出21.8个百分点；然而，模型与医师间的难度感知不一致：模型准确率在医师评定的难度层级上无明显变化，各模型在题目层面与医师的一致性接近零，且在8.4%的题目中，医师正确而领先模型错误。这表明当前模型在理解中医复杂语义与临床判断方面仍存在显著局限。研究团队已公开发布数据集、医师答案、评测框架及每题模型输出结果。
链接: https://arxiv.org/abs/2609.33014
作者: Tzu-Heng Huang,Jet Lin,Eric Lin
机构: University of Wisconsin-Madison(威斯康星大学麦迪逊分校); University of California, Merced(加利福尼亚大学默塞德分校); TechTCM
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Medical benchmarks for language models are built almost entirely on Western biomedicine. Traditional Chinese Medicine (TCM) is a separate system, with its own diagnostic framework and its own literature, and it remains largely unmeasured. The few TCM evaluations that exist are small, narrow, and rarely paired with a human reference. We present TCMQA, an open benchmark of 38,279 questions from Chinese TCM licensing examinations, paired with 15,151 responses from 101 licensed practitioners. We evaluate 29 instruction-tuned models from 9 families, spanning 0.27B to 14.8B parameters. Accuracy ranges over 59 points, and no model approaches saturation. Pretraining data predicts TCM ability far better than scale: a 12B Western-pretrained model reaches 39.6%, while a Chinese-pretrained model an eighth its size reaches 60.8%. Nine models exceed the practitioner majority vote of 64.9%, the best by 21.8 points, and all nine come from that same Chinese-pretrained family. Yet difficulty does not transfer between models and practitioners: accuracy is flat across practitioner-rated difficulty, item-level agreement is near zero for all 29 models, and on 8.4% of items the practitioners are correct where the leading model is wrong. We release the corpus, the practitioner responses, the harness, and per-item model outputs at this https URL.
[NLP-244] DynamicDx: Evaluating Evidence Acquisition in Video-Based Diagnosis
【速读】： 该论文旨在解决视频驱动的临床诊断中，仅依赖视觉识别无法实现准确诊断的核心问题——即模型虽能识别症状，却难以生成合理的鉴别诊断假设、提出恰当的检查问题并制定有效的诊疗路径。其解决方案的关键在于突破“识别即诊断”的局限，强调诊断过程中的证据获取（evidence acquisition） 作为主要瓶颈，并通过两个干预措施予以优化：一是采用经过微调的40亿参数视频描述器（4B video describer），提升从短时密集采样片段中对症状的精准描述能力；二是引入源清洁文献检索（source-clean literature retrieval），增强初始鉴别诊断假设的覆盖范围。实验表明，这两项措施使模型所提出的检查顺序更接近临床医生实际记录的诊疗流程，从而显著提升诊断准确率（达73.2%–93.0%）。研究进一步指出，视频带来的性能提升并非源于单一帧的识别或时间顺序，而是由视频所激发的调查性推理过程驱动，验证了“看得更清楚”不如“问得更精准”的核心理念。
链接: https://arxiv.org/abs/2609.32957
作者: Jiahui Li,Yutong Guo,Nan Yang,Wenzhan Song,Jin Lu,Fei Dou
机构: University of Georgia, Athens, USA; Beijing Luhe Hospital, Beijing, China
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 49 pages. Code and data: this https URL
Abstract:Diagnosing a patient from video requires more than recognizing the sign: a vision-language model must turn what it sees into hypotheses, questions and tests. DynamicDx evaluates each step in 71 neurological consultations across 11 sign categories, linking authentic patient videos to confirmed diagnoses and fixed charts built from the same case reports, so that every model queries the same evidence. Across five such models, video improves accuracy by 9.9-22.5 percentage points over blind input, but neither recognition alone nor temporal order explains the gain: the cause is usually missing from the model’s video-only differential diagnosis even when the sign is recognized, and shuffling the frames produces no reliable accuracy loss. Instead, a trajectory replay traces most of the gain to the investigation results the video prompts. Evidence acquisition is the bottleneck: supplying the decisive investigations raises accuracy to 73.2-93.0%. Two interventions act on it. A post-trained 4B video describer improves sign descriptions, especially from a short, densely sampled segment, and source-clean literature retrieval expands initial hypotheses; both bring the tests a model orders closer to those the treating clinicians documented and, through them, raise accuracy. For video-based diagnosis, seeing better helps when it leads to asking better.
[NLP-245] he Key Handoff: Retrieval in Hybrid Language Models
【速读】： 该论文旨在解决生成式 AI（Generative AI）在处理两跳问答（two-hop question）时，模型如何利用中间桥接实体（bridge entity）进行推理的问题。具体而言，当问题需要通过两次检索才能得出答案（如从“球属于爱丽丝”和“爱丽丝在花园”推断出“球在花园”）时，模型必须识别并利用未在问题中明确提及的实体（如“爱丽丝”），而这一过程在不同架构的模型中表现出显著差异。其解决方案的关键在于揭示注意力机制与递归状态（recurrent state）在信息传递中的作用：在所有测试的密集型与混合型模型中，注意力层会将隐藏状态中的键（key）转换为可被后续层使用的表征；一旦越过该注意力层，该键的可用性即被消除，除非由递归层持续携带。在序列化混合模型中，这种机制形成了一种“交接”（handoff）行为——递归层负责维持并传递关键信息，而注意力层则负责消耗该信息以生成答案。此外，研究发现，在最后一个注意力层之后写入新的记忆事实，可显著提升目标事实的概率（提升1.3至2.7倍），且这种行为并非由模型架构或训练方式决定，而是依赖于递归状态作为可查询的记忆载体，表明其不仅承担信息传递功能，更具备动态记忆能力。
链接: https://arxiv.org/abs/2609.32942
作者: Kaan Kale,Oguzhan Baser,Sriram Vishwanath
机构: Georgia Institute of Technology (佐治亚理工学院); University of Texas at Austin (德克萨斯大学奥斯汀分校)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 36 pages, 5 figures in the main text
Abstract:A two-hop question makes a language model retrieve twice: once to produce a bridge entity, and once to retrieve with it. Transformers resolve that entity in their early layers. Hybrid models replace most of the attention with a recurrent state, so where the key becomes usable, and where it is spent on an answer, is not known. Answering “Where is the ball?” from “the ball belongs to Alice” and “Alice is in the garden” turns on Alice, a name the question does not mention. A model could reach garden through Alice, the key it computed, or through where the fact sits in the prompt. Across twelve models, dense and hybrid, we move a hidden state from one story into another where the two routes lead to different places, and read off which place the model gives. An attention layer converts the key in every model we tested, and crossing that layer removes its usable effect, all of it where no attention follows. In sequential hybrids this makes the answer a handoff: recurrent layers carry the key forward, and attention spends it. Retrieval does not always stop there. Writing a different fact into the memory of the recurrent layers that follow the last attention layer can move the answer toward that fact, multiplying its odds by 1.3 to 2.7, and which hybrids do this is not settled by their architecture or training. That read is addressed by the key, not by position: recurrent state in a hybrid is not only a carrier, but a memory that later layers can query.
[NLP-246] Logical subspace in LLM s
【速读】： 该论文旨在探究语言模型是否具备类似人类大脑中专门负责抽象形式推理的神经网络机制。针对这一问题，研究提出“最小可行子空间（Minimal Viable Subspace, MVS）”方法，其核心在于识别并保留某一层中最低秩的激活子空间，以在移除该子空间外所有成分的情况下仍维持任务性能。通过MVS方法，研究发现Gemma和Qwen模型中存在支持逻辑推理的低秩子空间；这些子空间与模型在其他任务（如事实知识、工作记忆、认知控制及算术能力）上的表现显著分离：保留该逻辑子空间可维持推理能力但损害其他能力，而将其移除则使逻辑推理准确率降至随机水平，同时对其他能力影响较小。这表明语言模型中存在一个功能上可定位的核心逻辑处理机制，类似于人类大脑中的抽象形式推理网络。
链接: https://arxiv.org/abs/2609.32907
作者: Hope Kean,Enric Boix-Adsera
机构: Massachusetts Institute of Technology (麻省理工学院); University of Pennsylvania (宾夕法尼亚大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Recent work has identified a human brain network specialized for abstract formal reasoning (Kean et al., 2025). Does the same hold true in language models? To answer this question, we introduce the minimal viable subspace (MVS) method, which searches for the lowest-rank activation subspace at a layer that preserves task performance when everything outside that subspace is ablated. Using MVS, we demonstrate low-rank subspaces supporting logical inference on Gemma and Qwen models. Furthermore, these subspaces exhibit a clear dissociation from model capacities on other tasks, such that retaining these late logic subspaces preserves inference while impairing factual knowledge, working memory, cognitive control, and arithmetic. Conversely, ablating them reduces logical inference accuracy to chance while largely sparing these other capacities. Our results suggest a functionally localizable core machinery for logic akin to that in the human brain.
[NLP-247] Linger and Lose: Knowledge Collapse in Low-Bit Language Models
【速读】： 该论文旨在解决低精度训练（特别是三值权重，ternary weights）中模型知识容量严重损失的问题，而这一问题在传统评估指标（如损失值和下游任务准确率）下被显著掩盖。其核心问题是：尽管三值模型在训练后表现出可接受的困惑度（perplexity）和准确率，但其实际存储的知识量（即每参数的事实比特数，knowledge capacity）远低于全精度模型，且存在“知识坍塌”（knowledge collapse）现象。解决方案的关键在于识别出该坍塌源于输出头（output head）的权重学习率不稳定——当学习率过高或未适当衰减时，输出头权重会无限制增长，导致模型无法预测特定值，从而破坏整体知识表征。通过引入一种“预热-稳定-衰减”（warmup-stable-decay）的学习率调度策略，并对输出头施加更低的学习率，可有效抑制坍塌，使三值模型在2500万和5000万参数规模下的知识容量分别提升2.7倍和4.3倍，且在更低精度下收益更显著。此外，研究指出仅依赖最终损失值评估低精度训练是不可靠的，应将训练过程中保留的知识容量作为关键评价指标。
链接: https://arxiv.org/abs/2609.32902
作者: Prashanna Mani Paudel,Shivanand Venkanna Sheshappanavar
机构: Geometric Intelligence Research Lab; Dept. of Electrical Engineering and Computer Science, University of Wyoming, USA
类目: Computation and Language (cs.CL)
备注:
Abstract:Training language models with ternary weights is commonly judged by loss and downstream accuracy, which record only a modest cost relative to full precision. We show that these metrics can conceal a much larger failure. We instead measure knowledge capacity, the factual bits stored per parameter, on synthetic biographies with known information content. We train GPT-2-style models from scratch with 2.5M to 50M parameters at five precisions. Under the standard cosine schedule, ternary models retain as little as 6% of an identically trained fp16 model’s capacity. The deficit widens with model size while perplexity rises by only 1.4 to 1.6 times. Measured throughout training, these models first acquire capacity and then lose most of it. We identify this knowledge collapse as a learning-rate dwell instability. Held near 1 – 2\times10^-4 with no decay, a pre-collapse model collapses within a few hundred exposures, and returning to a safe rate does not restore capacity. We then locate the collapse in the output head. It happens at a value the model can never predict. The weights there grow unchecked, while every other prediction the model makes is unchanged. We find that making the value predictable removes the collapse, regardless of which attribute carries it. A warmup-stable-decay schedule with a 10% cooldown increases ternary capacity by 2.7 times at 25M and 4.3 times at 50M. Gains are larger at lower precision. Cutting the output head’s learning rate prevents failure when training from scratch. The post-training quantization methods we tested recover no measurable capacity below 4 bits. Our findings suggest judging low-precision training by retained capacity during training rather than final loss.
[NLP-248] Multimodal LLM s Outperform Pathology Foundation Models in Cross-Domain Histological Similarity
【速读】： 该论文旨在解决当前病理学基础模型在跨切片（slide）或跨机构（institution）比较时无法有效保持组织相似性的关键问题。其核心挑战在于，尽管现有病理学基础模型基于数百万张组织切片（histology tiles）进行训练，但在跨域场景下，这些模型常因依赖与制备流程相关的捷径特征（shortcut features），导致对相同疾病但不同机构的组织切片误判为更相似，而对同一机构内不同疾病的切片反而判断为更相近，这种临床危险性错误在常规域内评估中难以被发现。论文提出的关键解决方案是采用无需专门训练为病理学基础模型的通用多模态大语言模型（multimodal LLMs），通过语义层面的视觉对比分析组织形态与组织架构，而非依赖特定采集上下文的浅层特征，从而显著提升跨机构相似性判断的鲁棒性。研究进一步构建了MOSAIC（Model Similarity Assessment across Institutions and Cohorts）基准，系统评估17个模型在6个数据集上的表现，结果表明，扩大训练数据规模无法解决此问题，根本原因在于学习目标的设计缺陷，而非数据覆盖不足。该研究揭示了当前病理学基础模型在跨域泛化能力上的根本性脆弱性，并确立多模态大语言模型作为跨机构检索、数据集调和及多中心质量控制的可行替代方案。
链接: https://arxiv.org/abs/2609.32876
作者: Yishu Zhang,Yun Li,Daiwei Zhang
机构: University of North Carolina at Chapel Hill (北卡罗来纳大学教堂山分校)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:State-of-the-art pathology foundation models, trained on millions of histology tiles, can fail to preserve tissue similarity when comparisons cross slide or institution boundaries. We show that general-purpose multimodal LLMs, without being trained as pathology foundation models, consistently outperform these specialized models in cross-domain histological similarity judgments. Using a relative similarity framework that we release as the MOSAIC (Model Similarity Assessment across Institutions and Cohorts) benchmark, we evaluate 17 models across 6 datasets and find that pathology encoders often rank same-institution, different-disease tiles as more similar than same-disease, different-institution tiles, a clinically dangerous failure mode invisible to standard within-domain evaluations. LLMs appear less susceptible to this failure, likely because they perform semantic visual comparison of morphology and tissue architecture rather than relying on shortcut features tied to acquisition context. Scaling training data does not resolve the problem for pathology encoders, implicating the learning objective rather than data coverage. Our results expose a fundamental robustness gap in current pathology foundation models and establish multimodal LLMs as a viable alternative for cross-institutional retrieval, dataset harmonization, and multi-site quality control. Code and data will be released upon acceptance.
[NLP-249] Are You Sure Youre Sure? Two Confounds in a Sycophancy Benchmark
【速读】： 该论文旨在解决生成式 AI（Generative AI）在面对用户质疑时表现出的“谄媚倾向”（Sycophancy）问题，即模型在用户提出反对意见后倾向于放弃正确答案而迎合用户错误主张的现象。现有基准测试如 SycEval 通过预设提示模板来评估这一现象，但其设计存在关键混淆因素：模板中多个变量（如目标答案的命名与否、输出格式指令的位置）与所欲测量的“时间时机”（即质疑发生在回答前或回答中）同时变化，导致结果难以归因。研究发现，两个核心混淆因素显著影响实验结果：其一，当仅在较弱强度的质疑中提前命名目标答案时，该变量与提问时机耦合，导致“命名目标答案”本身使模型更易顺从，提升顺从率（follow rate）14.1–49.5个百分点；一旦两组模板均明确目标答案，原本支持“前置质疑导致更多谄媚”的结论逆转，模型反而在前置质疑下更少妥协。其二，用于自动评分的输出格式指令随质疑位置改变而迁移，该指令在不同模型上产生相反效应，且在相同任务中移动指令可显著降低模型谄媚行为（如在 Llama-3.1-8B 上降低 5.1 个百分点），而对 Qwen3-4B 无显著影响。这些混淆因素均源于提示模板设计而非模型自身特性，因此可被修正。论文最终提出三项基准测试发布前应执行的验证检查，以确保测量结果真实反映所关注属性，而非受模板设计偏差干扰。
链接: https://arxiv.org/abs/2609.32867
作者: Atharv Gupta,Akshat Jindal,Lavanya Nigam,Aryan Sood
机构: Indian Institute of Technology Roorkee(印度理工学院鲁尔基分校)
类目: Computation and Language (cs.CL)
备注: 22 pages, 19 tables, 3 figures
Abstract:Sycophancy is a language model’s tendency to cave when a user pushes back, abandoning a correct answer for the user’s. Several benchmarks now measure it by scripting an objection and recording how often the model caves. Because that objection is a prompt template, whatever else the template varies is measured along with the property it claims to isolate. We audit SycEval, which reports that objections raised before a model answers (preemptive) cause more caving than those raised after (in-context), and attributes the gap to timing. Two features of its templates vary alongside the property each is meant to test. First, SycEval’s objections escalate through four strength levels, and at the two weakest only the preemptive template names a target answer, so timing and naming vary together. We build the missing comparison and test it on multiple-choice questions and SycEval’s own free-form pipeline. Naming a target answer raises the follow rate, the share of samples matching the user’s assertion, by 14.1–49.5 percentage points (pp), and once both templates name one, the timing comparison reverses on three of the five model conditions we test: models cave \emphless under preemptive objections than in-context ones, opposite to SycEval. Second, the output-format instruction benchmarks append for automatic grading also varies with objection placement. Moving it from the pushback into the question flips its effect in opposite directions across models on the same items ( p=0.0059 ). The same effect appears in SycEval’s free-form pipeline: relocating its instruction lowers caving by 5.1pp on Llama-3.1-8B, while an equivalence test confirms no effect on Qwen3-4B. Both confounds live in the template rather than the models under test, so both are correctable: we close with three checks benchmark authors can apply before publishing.
[NLP-250] ARSM: Auto-Regressive State Machine for Agent ic Reasoning Compression
【速读】： 该论文旨在解决基于大语言模型（Large Language Model, LLM）的智能体在执行长时序任务时因上下文持续累积而产生的严重记忆瓶颈问题。现有记忆压缩方法依赖于特定任务优化或外部辅助模型，导致计算开销大，且压缩后的表征常丢失结构化关系，引发信息稀释、注意力坍塌及决策一致性下降等缺陷。为克服上述局限，论文提出一种轻量级、无需训练的自回归状态机（Auto-Regressive State Machine, ARSM）框架，其核心在于通过结构化状态演化实现就地推理压缩。关键创新包括：(i) 轨迹抽象机制，将交互历史重构为紧凑的假设-动作-结果（Hypothesis-Action-Result, HAR）微链；(ii) 动态状态机，通过原子操作与压缩控制参数调控分层记忆。二者统一于自回归、自压缩生成空间中，使每次模型输出同时完成外部动作执行与内部状态更新。实验在Webshop、多目标多跳问答及SWE-Bench Lite数据集上验证表明，ARSM在保持任务性能的同时显著降低令牌消耗，为长时序任务下可扩展自主智能体提供了一条高效、低成本的解决方案。
链接: https://arxiv.org/abs/2609.32852
作者: Xiafeng Man,Siyuan Ye,Xiaosong Ma
机构: College of Future Information Technology; Fudan University (复旦大学); The Hong Kong Polytechnic University (香港理工大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:While Large Language Model (LLM)-based agents demonstrate strong capabilities in long-horizon tasks by interleaving reasoning with external environment interactions, the continuous accumulation of context rapidly creates a critical memory bottleneck. Existing memory compression methods rely on task-specific optimization or external auxiliary models, introducing significant computational overhead. Furthermore, the resulting compressed representations tend to lose structured relationships, leading to information dilution, attention collapse, and degraded decision consistency. To address these limitations, we propose Auto-Regressive State Machine (ARSM), a lightweight training-free framework that enables in-situ reasoning compression through structured state evolution. ARSM introduces two key components: (i) a trajectory abstraction mechanism that reorganizes interaction histories into compact Hypothesis-Action-Result (HAR) micro-chains; (ii) a dynamic state machine that regulates hierarchical memory through atomic operations and a compression-control parameter. These components are unified within an auto-regressive, self-compressive generation space, where each model output jointly performs external action execution and internal state updates. We evaluate ARSM on Webshop, Multi-Objective Multi-Hop QA, and SWE-Bench Lite datasets. Experimental results show that ARSM maintains the task performance while simultaneously reducing token consumption, offering a practical, cost-effective route toward scalable autonomous agents for long-horizon tasks. Subjects: Computation and Language (cs.CL) Cite as: arXiv:2609.32852 [cs.CL] (or arXiv:2609.32852v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2609.32852 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-251] he Decomposition Tax: LLM Pipelines Lose Up to 40 Accuracy Points at Their Own Interfaces
【速读】： 该论文旨在解决大语言模型（LLM）在多阶段推理任务中因信息丢失导致的性能下降问题，即“分解税”（decomposition tax）。其核心问题是：当模型在多阶段推理过程中，后续阶段无法访问原始问题时，会因上下文信息的逐步丢失而显著降低准确性。研究发现，在多个开源模型（来自九个组织）和基准测试（GSM-Hard、MATH-500）上，这种损失可高达40.5个准确率点，且统计显著（Holm校正后p = 1.66e-19）。解决方案的关键在于两点：一是在损失性接口之后重新引入原始问题进行“再定位”（re-grounding），二是若某阶段需列出数值量，则应明确指示其保留各量之间的关系。实验证明，这一策略能有效降低分解税，即使在最新模型（如gemma-4-12B）上也保持有效性，且对不同模型均具普适性。
链接: https://arxiv.org/abs/2609.32825
作者: Tianqi Bu,YuXuan Peng,Junteng Tu,Henghui Xiao
机构: Rutgers University (罗格斯大学); Nanjing Tech University (南京工业大学); Nanjing University of Posts and Telecommunications (南京邮电大学); New York University (纽约大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Software Engineering (cs.SE)
备注: 23 pages, 5 figures; preprint under review
Abstract:A four-stage LLM pipeline gives up as much as 40.5 accuracy points at its own interfaces (gemma-3-12B on MATH-500, Holm-corrected p = 1.66e-19; the largest tax in the primary family). We hold model, problem, stages, stage prompts and completion budget fixed, vary only whether each stage can still see the original problem, and call the accuracy difference the decomposition tax. Across 21 open-weight models from nine organisations, on GSM-Hard and MATH-500 at n = 200 paired items per cell, 70 of 118 primary-family tests survive Benjamini-Hochberg correction and 54 survive Holm. On GSM-Hard, a placebo recovers nothing: it carries at least 60% of the extra tokens and at most one word of the problem. Builders design a pipeline one stage at a time, and its bill arrives at the interfaces between stages. Rewriting one stage’s instruction moves gemma-3-12B’s tax from 4.5 to 36.5 points, and adding “every relationship stated between them” to a stage that lists the numerical quantities lowers the tax on 9 of 9 models on MATH-500. Re-grounding, which shows a stage the original problem again, belongs after the loss. With one lossy interface, re-grounding the stage after it beats re-grounding the stage before it on 7 of 7 models on both benchmarks; on MATH-500 the earlier repair is worse than none on 7 of 7. Newer models still pay: gemma-4-12B gives up 37.0 points, and the repair holds on all three of the newest models we test. A sealed held-out test refuted a stronger rule we registered, which predicted the paying stage from the interface and receiver types, so we locate the tax by measuring one stage at a time. The prescription has two parts: re-ground the stage after the lossy interface, and if a stage must list the quantities, tell it to keep the relationships.
[NLP-252] USAI-Quant: A Quantitative Reasoning Benchmark for Vision-Language Models in Built Environments
【速读】： 该论文旨在解决当前大型视觉-语言模型（Large Vision-Language Models, VLMs）在遥感影像上进行定量推理能力不足的问题。尽管现有技术在定性视觉问答（Visual Question Answering, VQA）任务中表现良好，但缺乏针对城市与空间环境度量指标的量化评估基准，限制了对模型在真实世界城市分析中数值推理能力的准确评估。为此，研究提出首个专门用于评估VLM在遥感影像上对建成环境度量进行定量推理的基准——定量城市与空间人工智能基准（USAI-Quant）。该基准涵盖美国前335个最大城市，将高分辨率遥感图像与精确的建成环境量化指标相匹配，并通过三类复杂度的视觉问题问答任务，系统评估通用型及遥感专用VLMs（RS-VLMs）的表现。实验结果表明，当前最先进的模型在数值推理任务中普遍存在显著短板。通过深入分析模型性能、问题类型与地理分布之间的关系，研究揭示了不同场景下的性能波动规律及特定任务挑战，为未来面向城市与空间智能的生成式模型设计提供了关键方向。
链接: https://arxiv.org/abs/2609.32813
作者: Dongdong Wang,Qingqi Song,Yuzhou Chen,Deepak Balakrishnan,Ravi Shankar Srinivasan,Shenhao Wang
机构: University of Florida(佛罗里达大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 10 pages, 13 figures
Abstract:Large vision-language models (VLMs) have emerged as a powerful paradigm for urban and spatial AI. However, current state-of-the-art large VLMs still struggle with quantitative reasoning on remote sensing imagery. Existing benchmarks and algorithms are predominantly based on qualitative Visual Question Answering (VQA), providing limited insights into the quantitative reasoning capabilities of VLMs for built environment metrics. To address this gap, we develop Quantitative Urban and Spatial AI benchmark (USAI-Quant), the first benchmark designed to quantitatively evaluate VLM’s reasoning capabilities on built environment metrics via remote sensing imagery. USAI-Quant is curated from the 335 largest U.S. cities, aligning high-resolution remote sensing images with quantitative built environment metrics. We then evaluate both general-purpose and remote sensing VLMs (RS-VLMs) by applying VQAs to tens of built environment metrics across three complexity levels. Our results reveal that current state-of-the-art models consistently fall short on numeric reasoning tasks. We further conduct in-depth analyses across models, question types, and geographic locations, uncovering insights into performance variability and task-specific challenges.
[NLP-253] OpenTumorBoard: A Real-World Benchmark of Multidisciplinary Tumor Board Discussion Trajectories
【速读】： 该论文旨在解决现有医学基准难以捕捉真实世界多学科肿瘤讨论中患者长期病史与多模态临床信息整合的问题。其核心挑战在于如何评估大语言模型（LLM）在复杂、动态的多学科肿瘤讨论场景中的临床决策能力，尤其是在生成符合实际讨论轨迹和达成共识结论方面的能力。解决方案的关键是构建OpenTumorBoard——一个包含611例患者病例、19,157轮跨十类专科角色的对话记录的公开基准，数据源自12,534分钟的YouTube公开肿瘤讨论视频。该基准涵盖两种评估范式：SPECIALIST TURN（针对具体临床问题生成专业回复）与BOARD SIMULATION（模拟完整讨论流程并达成治疗方案、手术计划、后续行动及临床试验匹配等共识）。实验表明，当前前沿通用与医学专用大模型在临床等效性与结论一致性上均存在显著不足（最高得分分别为3.43/5和2.78/5），但通过监督微调与强化学习可提升性能，验证了真实讨论轨迹对模型适应性的价值。三名医学专家对子集的评审确认了病例信息覆盖率高、事实准确性强以及共识结论提取的高保真度。该研究为推进生成式人工智能在个性化癌症诊疗中的应用提供了关键基础设施与评估标准。
链接: https://arxiv.org/abs/2609.32810
作者: Anqi Li,Zhixuan Ge,Yixuan Duan,Jiarong Qian,Chi-Yu Chen,MingYu Lu,Huan-Yu Hsu,Yu Gu,Yue Guo,Sheng Wang,Wei Qiu,Hanwen Xu
机构: Rice University; University of Illinois at Urbana-Champaign; University of Washington; National Yang Ming Chiao Tung University; Microsoft; University of Pennsylvania
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Preprint. Includes supplementary material
Abstract:Multidisciplinary tumor boards integrate multimodal clinical observations and longitudinal patient histories through specialist discussions, yet benchmarks rarely capture these real-world trajectories. We introduce OpenTumorBoard, a benchmark with 611 patient cases and 19,157 discussion turns across ten specialist roles, transcribed from 12,534 minutes of publicly available tumor board recordings on YouTube. The benchmark evaluates two settings: SPECIALIST TURN, in which an LLM responds to a clinically significant question posed during a real discussion, and BOARD SIMULATION, in which it generates an entire back-and-forth discussion and reaches a consensus on therapy recommendations, surgical plans, next actions and clinical trial matching. Evaluation of 14 general-purpose frontier and medical LLMs reveals substantial limitations: the best models score 3.43 out of 5 in clinical equivalence to specialist answers and 2.78 out of 5 in alignment with recorded board conclusions. Supervised finetuning and reinforcement learning improve performance on a held-out test set, suggesting that real-world discussion trajectories can support model adaptation. Three M.D. experts review a subset of the benchmark, finding high information coverage and factuality of patient cases and strong fidelity of extracted consensus conclusions. We will release OpenTumorBoard and its automated curation pipeline to support the development and evaluation of LLMs for multidisciplinary, personalized cancer decision-making.
[NLP-254] Overwhelmed by Choice: Studying LLM Decision Making at Scale NEURIPS2026
【速读】： 该论文旨在解决大模型在大规模候选集场景下推理与决策能力评估的有效性问题。现有主流评测基准多采用小规模候选集，但其结论是否适用于候选集规模扩大时仍不明确。研究发现，随着候选数量增加，大模型在多个任务、策略及模型规模下均出现显著的准确率下降。通过受控分析表明，标准长上下文检索解释无法完全解释这一退化现象。研究识别出两种系统性失败模式：一是“黄金得分差距坍缩”（gold-margin collapse），即正确答案与最强干扰项之间的得分差持续缩小，主要由正确答案置信度下降驱动；二是“早期候选偏好固化”，即早期候选项对最终预测的影响难以被后续候选项所扭转，后期候选项的影响力逐步减弱。基于上述发现，论文提出分层划分（hierarchical partitioning）与排列推理（permutation-based inference）两种策略，在HotpotQA和MIMIC数据集上于N=160的候选集规模下使准确率提升约20个百分点。研究结果强调候选集规模是评估协议中的关键变量，且小规模候选集上的优异表现并不能保证大尺度候选比较下的鲁棒性。
链接: https://arxiv.org/abs/2609.32809
作者: Yu-Chi Lin,Aryan Seth,Anshul Aravind,Eugene Lee,Tanmay Parekh,Nanyun Peng,Kai-Wei Chang
机构: University of California, Los Angeles (加州大学洛杉矶分校); University of Cincinnati (辛辛那提大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted at TAE (Trust-AI-Eval): Can We Trust AI Evaluation?, NeurIPS 2026 Workshop. 23 pages
Abstract:Multiple-choice and candidate-selection evaluations are widely used to assess LLM reasoning and decision-making, yet most benchmarks contain relatively small candidate sets. It remains unclear whether conclusions drawn from these settings remain valid as the candidate space scales. We systematically evaluate LLMs as the number of competing candidates increases and find substantial accuracy degradation across tasks, prompting strategies, and model scales. Controlled analyses show that standard long-context retrieval explanations cannot fully account for this degradation. Instead, we identify two systematic failure patterns. First, gold-margin collapse: the score gap between the correct answer and the strongest distractor progressively shrinks, driven primarily by weakening confidence in the correct answer. Second, earlier candidate preferences become increasingly difficult to overturn, with later candidates exerting progressively weaker influence on the final prediction. Motivated by these findings, we evaluate hierarchical partitioning and permutation-based inference, which improve accuracy by roughly 20 percentage points at N=160 on both HotpotQA and MIMIC. Overall, our results identify candidate-set scale as an important evaluation-protocol variable and show that strong small-option performance does not necessarily imply robust large-scale candidate comparison.
[NLP-255] Mind the Spike: Mechanisms and Brittleness of Visual Massive Activations in Large Vision-Language Models
【速读】： 该论文旨在解决大视觉语言模型（Large Vision-Language Models, LVLMs）中普遍存在的“激活尖峰”（spikes）现象，即少数固定隐藏通道在早期层中出现远高于正常量级的激活值。尽管文本尖峰在不同输入下均稳定出现在初始位置，但视觉尖峰是否在不同模型间具有可复现的形成规律及其对图像扰动的鲁棒性尚不明确。研究发现，部分LVLMs不产生视觉尖峰，而其余模型则在深层以不同频率出现尖峰；通过分析模型权重，识别出触发尖峰生成的特定方向及一个可解释的位置规则：在语言解码器运行前，潜在尖峰令牌主要集中在与图像其余部分最不相似的区域。关键发现是，视觉尖峰具有极强的脆弱性——常见数据扰动（如噪声、模糊）常诱发或重新定位尖峰，显著提升其整体出现率。基于此，研究提出一种受触发引导的尖峰攻击方法，在极小的ℓ∞扰动（如1/255）下即可有效诱导或消除尖峰，且在十种模型中有九种成功实现。此外，研究设计了一种预防性干预策略，仅移除触发组件即可在不改变其他图像令牌的前提下，彻底消除或大幅降低干净与扰动图像上的尖峰现象。该研究覆盖25个基于18个公开文本基模型（来自10个架构家族，参数量2B至72B）的适配器型LVLMs，系统揭示了尖峰形成的机制与防御路径。
链接: https://arxiv.org/abs/2609.32808
作者: Jonas Ngnawé,Yann Pequignot,Sabyasachi Sahoo,Christian Gagné,Frédéric Precioso,Sanmi Koyejo
机构: Université Laval (IID); Mila – Quebec AI Institute; Stanford University; Canada CIFAR AI Chair; Université Côte d’Azur, CNRS, Inria, I3S, Maasai
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 58 pages, 15 figures, 43 tables
Abstract:Large vision-language models (LVLMs) inherit massive activations from their text-only bases: spikes where a few fixed hidden channels receive values thousands of times above the typical magnitude. The text spike systematically appears in early layers at a fixed initial position, independently of input content. Visual spikes vary across images, but whether their formation follows a consistent pattern across LVLMs and how they respond to image perturbations remain open questions. We find that some LVLMs do not form visual spikes, while others spike at different rates, typically in deeper layers. We identify the trigger direction from model weights and an interpretable location rule: before the language model decoder runs, eventual spike tokens are largely restricted to those sharing least with the rest of the image. Crucially, visual spikes are strikingly brittle. Common corruptions frequently create and relocate spikes, and less often remove them, raising overall incidence. Our trigger-guided spike attack deliberately creates or removes spikes under a small \ell_\infty budget, with 1/255 enough in nine of the ten models that spike. Finally, our preventive intervention removes only the trigger component before spikes erupt, eliminating or substantially reducing spikes on clean and perturbed images while leaving the other image tokens nearly unchanged. Our study spans 25 adapter-based LVLMs built on 18 released text-only bases from 10 families, ranging from 2B to 72B parameters.
[NLP-256] Decision-Sufficient State Representations: Measuring and Reducing Write-Time Regret
【速读】： 该论文旨在解决长任务中大型语言模型（LLM）代理因上下文长度限制而无法有效维持完整历史信息的问题，尤其在历史信息虽可容纳但实际使用中仍存在不可靠性的情况下。现有方法通过让代理维护一个简短的书面状态（written state）来缓解此问题：每一步由“写入者”重写状态，“读取者”仅基于当前状态做出决策。然而，写入者一旦丢弃某些信息，后续决策可能发现其必要性，导致信息丢失。本文量化了这一损失，并探究是否可通过训练减少该损失。作者将读取者的损失分解为“预算损失”（任何同尺寸状态都不可避免的部分）和“写入时机遗憾”（由写入者决策造成的额外损失）。在TextWorld烹饪游戏中，当事实需延迟一定时间才被使用时，128词元的状态能几乎全胜，而提示式语言模型写入者胜率不足17%——几乎全部损失源于写入时机遗憾，且随延迟增加而加剧。为此，作者提出一种名为决策充分状态表示（DSSR, Decision-Sufficient State Representations）的训练机制：以读取者后续行为表现作为反馈信号，评估候选状态的质量，并指导写入者偏好更优状态。实验表明，该前向滚动评分策略能显著预测游戏结果（ρ = 0.48），远优于传统事后回溯评分方法（ρ ≤ 0.07）。在预注册测试集中，训练使写入者在事实需求紧迫时成功率提升7.0分（95% CI: [1.9, 12.2]），达到信念或槽位记忆提示水平；但该增益随延迟增大而减弱，仅在最短延迟下显著。作者进一步指出，该性能上限源于信用分配难题：当前保留某事实的收益依赖于所有后续状态更新均持续保留该信息，而逐步评分机制无法捕捉这种长期依赖关系。
链接: https://arxiv.org/abs/2609.32805
作者: Bingyu Shen,Boyang Li
机构: Independent Researcher(独立研究员); Kean University (基恩大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 29 pages (10 main, 17 appendix), 12 figures (4 main, 8 appendix), 18 tables (2 main, 16 appendix)
Abstract:Long tasks produce more history than an LLM agent can hold in its context, and more than it uses reliably even when the history fits. A growing line of work therefore has agents carry a short written state instead: at every step a writer rewrites the state, and a reader acts from the state alone. Steps stay cheap, but anything the writer drops is lost before later decisions reveal that they need it. We quantify this loss and ask whether training can reduce it. Comparing the written state with the best state of the same size written in hindsight, we split the reader’s loss into a budget loss, which any state of that size must incur, and a write-time regret, which comes from the writer’s choices. In TextWorld cooking games where we control how long a fact must be carried before it is needed, a 128-token state holding the facts wins nearly every game, while prompted language-model writers win at most 17%. Almost all of the loss is write-time regret, and it grows with the delay. We then train the writer from the reader’s own loss. DSSR (decision-sufficient state representations) scores candidate states by how well the reader acts after the writer carries them forward, and teaches the writer to prefer the better ones. This forward-rolled score predicts game outcomes ( \rho = 0.48 ), whereas scoring a candidate as a fixed context, as hindsight methods usually do, does not ( \rho \leq 0.07 ). On a pre-registered test split opened once, training adds +7.0 [+1.9, +12.2] points of success when facts are needed soon, bringing a plain summary writer to the level of belief- and slot-based memory prompts. The gain shrinks as the delay grows and is significant only at the shortest delay. We trace this limit to credit assignment: keeping a fact now pays off only if every later rewrite keeps it too, which a per-step score cannot see.
[NLP-257] Re-derivability Decides What a Staged Agent Pipeline Recovers After an Upstream Fault
【速读】： 该论文旨在解决多阶段语言模型代理（language-model agents）在上游故障传播下的性能退化问题，核心关注点在于：当第一阶段产生确定性错误时，后续阶段能否通过重新从原始问题中推导所需信息（即“可重推导性”，re-derivability）来缓解或修复该故障带来的影响。其解决方案的关键在于将下游代理的检查器（inspector agent）显式地锚定于原始问题，从而实现对故障的感知与恢复。实验表明，这种基于原始问题的重推导机制显著提升了系统鲁棒性——在四个主流开源模型上，基于可重推导性的修复策略使准确率从+0.233提升至+0.392（Holm校正后p = 2.1×10⁻⁶），且在四类不同架构配置下均表现出统计显著的改进。特别地，仅需对首阶段进行一次重新接地（re-grounding），即可获得+0.394的保留率增益，而后续阶段无需额外干预，验证了该修复策略具有低成本、前段集中部署的高效特性。
链接: https://arxiv.org/abs/2609.32802
作者: Tianqi Bu,YuXuan Peng,Junteng Tu,Henghui Xiao
机构: Rutgers University (罗格斯大学); Nanjing Tech University (南京工业大学); Nanjing University of Posts and Telecommunications (南京邮电大学); New York University (纽约大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 34 pages, 8 figures, 13 tables; preprint under review
Abstract:One variable sets what an upstream fault costs a staged pipeline of language-model agents: re-derivability, how much of what a stage needs it can rebuild from the original problem. Grounding an inspector agent in that problem is worth +0.608 [+0.517, +0.700] to +0.358 over a blind one on four open-weight backbones served with thinking disabled, and on the two Qwen backbones the blind inspector changes no item at all. That head-to-head is exploratory. One deterministic fault enters the first stage, and we re-expose the original problem to k = 0,\dots,3 of the downstream stages with agents, items, fault and topology held fixed, on 120 gsm_hard items per arm at temperature zero. Accuracy under fault rises on four of four backbones, from +0.233 to +0.392, the largest Holm-adjusted p being 2.1\times10^-6 . A registered kill test rules out tokens. Blanking every word holds the word slots fixed, and retention tracks the visible fraction on four of four, climbing from 0.221 to 0.692 on the primary. Those two families are confirmatory and everything else here is exploratory. The interaction excludes zero on two of four backbones under the registered pipeline, four of four under a three-stage pipeline, and three of four under full message history, the primary at +0.317. On Llama-3.1-8B the fault carries no detectable cost at any dose, so the other three carry every claim about what a fault costs. Re-derivability also sets what the architecture costs, and no decomposition we measured reliably beats one direct call. With no fault injected the registered pipeline loses to that call by -0.267, -0.125 and -0.317, and on Phi-4 reads +0.058 at p = 0.118 , which the test fails to separate from zero. The repair that works is cheap and front-loaded: the first re-grounded stage buys +0.394 of matched retention for +59.8 tokens per item on Qwen3-14B, and the stages after it buy nothing.
[NLP-258] Understanding and Exploiting Anisotropy in Post-Training
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）后训练过程中，生成式 AI (Generative AI) 模型在监督微调（Supervised Fine-Tuning, SFT）与强化学习（Reinforcement Learning, RL）联合优化时，模型内部表征动态不一致的问题。具体而言，现有方法中频率加权似然训练引发的“各向异性”（anisotropy）现象——即少数残差通道承载异常高的激活值——长期被视为模型缺陷，但其本质功能及与后训练过程的交互机制尚不明确。本文的关键发现是：约5%的高激活残差通道构成模型的语义连贯性基础（coherence substrate），移除这些通道会使困惑度从10飙升至超过10⁶，而随机匹配数量的通道仅导致35的上升；然而，这些通道本身对推理正确性判别能力极弱。进一步分析表明，SFT会重塑这些通道，而RL则保持其稳定，并促使互补通道进行适应性调整。基于此，作者提出\textscSphereGate方法，通过在冻结主干网络上为每个残差通道学习一个有界增益，利用激活加权梯度实现对高能量连贯性通道的运动抑制，同时释放其余通道自由度。该方法仅引入0.1M可训练参数，在Qwen2.5（0.5B–7B）和Llama-3-8B系列模型上于MATH-500任务上相较参数高效基线提升2.0–7.3点，性能媲美甚至超越全模型GRPO。研究揭示，各向异性并非缺陷，而是后训练可利用的“分工机制”，为高效、稳定地增强模型推理能力提供了新范式。
链接: https://arxiv.org/abs/2609.32792
作者: Samyak Jha,Harshvardhan Saini,Yizhen Liao,Yiming Tang,Dianbo Liu
机构: Indian Institute of Technology Dhanbad(印度理工学院达纳巴德分校); National University of Singapore(新加坡国立大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:LLM post-training combines supervised fine-tuning (SFT), a mode-covering forward-KL objective, with reinforcement learning (RL), a mode-seeking reverse-KL objective. Frequency-weighted likelihood training leaves a well-known signature: \emphanisotropy, in which a few residual channels carry disproportionately large activations. Anisotropy is widely documented and usually treated as a defect, yet its function and its interaction with post-training remain unclear. We first analyze it. A label-free outlier rule isolates about 5% of residual channels that are essential for language modeling: removing them raises perplexity from 10 to over 10^6 , versus 35 for count-matched random channels. Yet they barely distinguish correct from incorrect reasoning. SFT reshapes them, whereas RL leaves them largely intact and adapts the complementary channels. These channels therefore form the model’s \emphcoherence substrate, and reasoning adaptation happens elsewhere. We then exploit this. \textscSphereGate learns one bounded gain per residual channel on a frozen backbone. Its activation-weighted gradients provably limit movement of high-energy coherence channels and leave the remaining channels free. With 0.1M trainable parameters, \textscSphereGate outperforms parameter-efficient baselines by 2.0–7.3 points on MATH-500 across Qwen2.5 (0.5B–7B) and Llama-3-8B, is comparable or exceeds full-model GRPO. Anisotropy is not a defect but a division of labor that post-training can exploit.
[NLP-259] On the Behavioral Traits of LLM Agents
【速读】： 该论文旨在解决现有生成式AI（Generative AI）人格测评方法中存在的核心问题：传统自述数据（S-data）与实际行为表现脱节，而依赖大语言模型（LLM）人工评估的观察者评分（I-data）虽具信度却难以规模化且覆盖场景有限。为突破这一瓶颈，本文提出一种基于行为数据（B-data）的自下而上建模框架A-B-D，其关键在于从真实世界中已有的多模型、多任务交互轨迹中提取高维特征，涵盖智能体在每一步所采取的动作（功能型）及其伴随的语言表达（语言型），共获得318个候选特征。通过筛选具备实例级稳定性、跨任务一致性及模型可区分性的79个特征，并进行因子分析，最终识别出六个稳定且可归因于模型本身的潜在因素——其中两个为功能型，四个为语言型。研究发现，不同模型在如计划性、活力等维度上存在显著差异，例如Kimi-K3表现出最强的计划性，而GPT-5.5与GPT-5.6则最为低能量。更重要的是，研究揭示了现实中的“知识-行动鸿沟”：这些基于行为推断出的人格因素与模型自我报告的五大性格特质（Big Five）之间相关性极弱，即使在概念对应项（如外向性与活力）上相关系数也仅为r = 0.07（p = 0.58），表明自述人格无法准确反映真实行为模式。该工作为理解生成式AI人格提供了全新的实证视角，对用户、开发者及计算机科学与社会科学交叉领域的研究均具有重要意义。
链接: https://arxiv.org/abs/2609.32776
作者: Haokai Zhao,Jie Gao,Yunze Xiao,Xintao Wang,Weihao Xuan,Aditya Joshi,Mark Dredze,Jen-tse Huang
机构: University of New South Wales; Johns Hopkins University; Carnegie Mellon University; Fudan University; The University of Tokyo; RIKEN AIP
类目: Computation and Language (cs.CL)
备注: Preprint; working in progress
Abstract:Users increasingly describe different AI agents as distinct colleagues to work with. AI personality research aims to quantify such impressions by attributing human-like “traits” to agents. However, existing measures fall short: models’ self-reports (S-data) diverge from their actual behavior, while informant ratings from LLM judges (I-data) are costly to scale and cover few everyday scenarios. In this paper, we propose A-B-D to infer traits bottom-up from behavioral data (B-data), namely how agents act on their environment and communicate with users, as recorded in existing trajectories. From 345,667 real-world trajectories spanning 80 models, 12 tasks, and 50 harnesses, we extract 318 candidate features that capture both the actions an agent takes at each step (functional) and the language accompanying them (linguistic). We retain only features that show instance-level stability, cross-task consistency, and model discriminability. Factor analysis of the remaining 79 features uncovers six stable, model-attributable factors, two functional and four linguistic. For example, Kimi-K3 exhibits the most planfulness, whereas GPT-5.5 and GPT-5.6 are the least energetic. Moreover, we quantify the “knowledge-action gap” in the wild: these factors correlate only weakly with self-reported Big Five scores, even for conceptually matched pairs such as extroversion and energetic (r = 0.07, p = 0.58). Our work offers a new lens for understanding AI personality, with implications for users, developers, and researchers from both computer science and social science.
[NLP-260] C-HAT-Bench: Benchmarking Chinese AI-Text Detection Beyond Fully Generated Text
【速读】： 该论文旨在解决当前中文语境下人机协作文本生成（Human-AI Collaborative Text Generation）场景中机器生成文本（MGT）检测模型性能评估失真的问题。现有检测方法普遍在“全人类撰写”与“全AI生成”二元对比的极端条件下进行评估，而实际应用中人机协同写作模式多样，其生成过程会弱化或重构传统检测线索，导致现有模型在真实复杂场景中的可靠性被高估。尤其在中文环境下，分词机制和语言特异性文本分布进一步影响检测信号的可辨识性，且缺乏覆盖多种生产场景、领域及生成模型的可控数据资源。为填补这一空白，本文提出了首个面向中文的人机协作文本检测基准——C-HAT-Bench，该基准将5,000份来自五个领域的原始人类文本，通过六种生成模型在前缀条件续写（Prefix-Conditioned Continuation）参考设置及三种协作生成模式下生成超过24万条变体文本，构建了统一评估框架。研究对21个检测器采用四种评估协议（包括零样本、预训练监督式文档级检测、边界定位与跨条件泛化）进行系统测试，结果表明：相较于前缀条件续写设置，各检测器在协作生成模式下的平均AUROC下降12.0%，个别模型降幅高达44.4%；同时，不同协作模式间的性能迁移呈现非对称性，说明单一生产模式下的表现无法可靠预测其他模式下的检测效果。因此，解决方案的关键在于构建一个涵盖真实协作生态的标准化评估基准，并揭示现有检测器在复杂人机交互场景下的脆弱性与泛化局限，从而推动更鲁棒的检测技术发展。
链接: https://arxiv.org/abs/2609.32770
作者: Qing Yang,Zixiang Luo,Zhenyu Mao,Zezheng Wu,Xinghe Cheng,Haibo Chen,Qinggang Zhang,Jiapu Wang,Jingwei Zhang
机构: Guilin University of Electronic Technology(桂林电子科技大学); Jinan Unaiversity(济南大学); Nanjing University of Science and Technology(南京理工大学); Jilin Unaiversity(吉林大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large Language Models (LLMs) increasingly participate in writing by modifying or extending human drafts, causing machine involvement to vary in both form and extent. Yet most Machine-Generated Text (MGT) detectors are evaluated only on fully human-written versus fully AI-generated text. Because human–AI collaboration can weaken or redistribute cues associated with machine generation, strong performance under this binary setting may overstate detector reliability. This mismatch remains underexplored in Chinese: detection cues are shaped by tokenization and language-specific text distributions, yet controlled resources spanning production settings, domains, and generators remain limited. To fill this gap, we present a Chinese Human-AI Collaborative Text Detection Benchmark (C-HAT-Bench), a unified benchmark that links 5,000 human-written source texts from five domains to more than 240,000 variants produced using six generative models under Prefix-Conditioned Continuation as a reference setting and three collaborative production modes. We evaluate 21 detectors through four protocols spanning zero-shot and pretrained supervised document-level detection, boundary localization, and cross-condition generalization. Relative to Prefix-Conditioned Continuation, mean AUROC across document-level detectors is 12.0% lower on the collaborative production modes, with the largest detector-specific relative decrease reaching 44.4% . Transfer across collaborative production modes is also asymmetric, indicating that performance in a given production setting is not a reliable predictor of performance in other production settings.
[NLP-261] How Far Do Persona Effects Generalize in Language Models?
【速读】： 该论文旨在解决生成式人工智能（Generative AI）中角色提示（persona prompts）的跨模型、跨提示迁移有效性问题，即已学习的角色特征是否能够有效预测新任务下的响应，并在不同模型架构与提示设置间保持实用性。其核心挑战在于：尽管角色效应在特定模型内可被预测，但其迁移能力受限，表现为“可预测但不可移植”（predictable without being portable）。解决方案的关键在于通过目标微调（target refitting）和正则化更新策略，显著提升跨模型迁移性能。研究发现，在OLMo-3等模型中，单独调整目标属性与任务的参数能超越共享缩放（shared scaling）效果；引入双目标参数虽消除损失，但未带来额外增益；而重述提示后，目标微调在所有测试对中均优于单幅度复用策略。此外，正则化更新在每属性64个目标问题上的中位表现优于传统复用方法，且提示重写比示例或国家语境变化更能维持复用价值。研究表明，看似有效的迁移效果可能源于目标校准而非源关系本身，真正有效的迁移需源关系提供超出校准与正则化的额外预测信息。
链接: https://arxiv.org/abs/2609.32758
作者: Yufan Zhou,Yuxuan Liu,Enze Ma,Lyumanshan Ye,Zhongqi Yue,Robin De Croon,Yucheng Jin,Katrien Verbert,Zhao Wang
机构: KU Leuven (鲁汶大学); East China University of Science and Technology (华东理工大学); University of Illinois Chicago (芝加哥伊利诺伊大学); Shanghai Jiao Tong University (上海交通大学); Microsoft Research (微软研究院); Duke Kunshan University (杜克昆山大学); Zhejiang University (浙江大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 42 pages, 11 figures, 41 tables. Code and data: this https URL
Abstract:Persona prompts ask language models to answer as particular kinds of people. We test whether relationships learned from these effects predict responses to new questions and remain useful across models and prompts. Across 57 attributes, three behavioral domains, and seven pairs of open 7 to 9B checkpoints, persona effects can be predictable without being portable. Separate attribute and task gains improve prediction beyond shared scaling significantly in OLMo-3 and Qwen2.5, with the most robust evidence in OLMo-3. In that model, target refitting significantly outperforms gains borrowed from each of the other six pairs. Across model transfers, borrowed gains with one amplitude underperform shared scaling in most directions; allowing two target parameters removes the significant losses but yields no significant benefit over target shared scaling. After rewording, refitting significantly outperforms reuse with one amplitude in all six tested pairs, while changes of examples or country context often preserve reuse value. In the tested prompt transfers, regularized updates outperform both reuse strategies in median at 64 target questions per attribute. A separate survey comparison finds that selecting the more responsive checkpoint can worsen human fit; responsiveness is confounded with training status, and temperature calibration largely removes this cost but not errors in group ordering. Within the tested gain representation, apparent transfer can come from target calibration; source relationships must add predictive value beyond calibration and regularization. Code and data are available at this https URL
[NLP-262] Adaptive Consistency Graph for Long-Horizon Agents
【速读】： 该论文旨在解决大语言模型代理（Large Language Model Agents）在执行长周期、依赖性强的任务时因上下文断裂导致决策偏离原始目标的问题。其核心挑战在于，随着任务执行过程的推进，任务需求、历史证据与当前执行状态之间的关联逐渐弱化，从而引发决策漂移。为此，论文提出自适应一致性图（Adaptive Consistency Graph, ACG），通过增量式地将执行证据及其来源组织为持久化的图结构，并在有限的上下文预算下，为每一次决策构建以任务需求为中心的临时视图。ACG不替代基础代理的规划器或工具执行器，而是为每个决策提供结构化且可追溯的上下文视图。实验表明，在匹配评估中，ACG将GPT-5.6-luna在ReAct基础上的平均成功率从44.5%提升至50.2%，尤其在BrowseComp-Plus任务上实现73.5%的显著提升（对比62.4%）。通过轨迹结构分析与推理成本评估，进一步揭示了该方法在维持长期一致性方面的有效性。
链接: https://arxiv.org/abs/2609.32754
作者: Jiecong Wang,Hao Peng,Zhanyi Wang
机构: Beihang University (北京航空航天大学); QI-ANXIN GROUP (奇安信集团)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Large language model agents can often make reasonable local decisions on short tasks, yet their performance degrades when success requires long sequences of dependent actions and tool calls. During execution, task requirements, historical evidence, and the current execution state may gradually become disconnected, so later decisions can drift from the original objective. We study this problem by introducing the Adaptive Consistency Graph (ACG) for long-horizon execution. ACG incrementally organizes execution evidence and its provenance in a persistent graph, then constructs a temporary requirement-centered view for each decision under a bounded context budget. Rather than replacing the base agent’s planner or tool executor, ACG provides a structured and traceable context view for each decision. In the matched evaluation, ACG improves GPT-5.6-luna’s average success from 44.5% with ReAct to 50.2%, with the largest gain on BrowseComp-Plus (73.5% versus 62.4%). We further analyze trajectory structure and inference cost to characterize this improvement.
[NLP-263] Retrospective Distillation Attribution via Normalized Response Similarity
【速读】： 该论文旨在解决生成式 AI 模型在经过教师模型（teacher model）蒸馏后，其来源归属（model provenance）难以追溯的问题。现有方法通常仅在监督微调（SFT）后立即评估，但实际应用中，学生模型（student model）可能经历进一步的 SFT、偏好优化或强化学习，而审计者往往无法获取蒸馏前的原始检查点，导致传统基于参考的溯源方法失效。为此，论文提出 SCOUT——一种仅依赖输出文本的溯源方法，其核心在于通过聚合重复出现的语法模式（syntactic patterns）构建候选源模型档案，过滤低对比度模式，并将学生模型与候选源之间的距离校准于候选源间的距离分布。该方法无需模型权重、词元似然或历史检查点，即可实现对公开发布的蒸馏衍生模型的稳定溯源。实验表明，SCOUT 能有效识别不同后续训练目标下的蒸馏来源；同时，追踪训练轨迹中的语法签名发现，这些特征在蒸馏阶段即产生并持续存在于后续偏好优化与强化学习过程中，验证了其作为可追溯指纹的鲁棒性。
链接: https://arxiv.org/abs/2609.32749
作者: Minwoo Jang,Jaechang Kim,Minhyeon Oh,Jeongyeon Hwang,Jungseul Ok
机构: POSTECH Graduate School of Artificial Intelligence(韩国浦项科技大学人工智能研究生院); POSTECH Institute of Artificial Intelligence(韩国浦项科技大学人工智能研究所); Department of Computer Science and Engineering, POSTECH(韩国浦项科技大学计算机科学与工程系)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Cryptography and Security (cs.CR); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注: Preprint
Abstract:Model distillation transfers capabilities through supervised fine-tuning (SFT) on teacher responses, often collected from commercial APIs, raising questions of model provenance. Existing distillation attribution methods have been largely evaluated on students immediately after the SFT step. However, a distilled model may undergo further SFT, preference optimization, or reinforcement learning before release, while an auditor may lack access to the pre-distillation checkpoint required by reference-based attribution. To close this gap, we propose SCOUT, an output-only method that aggregates recurring syntactic patterns into candidate profiles, filters low-contrast patterns, and calibrates student–candidate distances against inter-candidate distances. SCOUT supports attribution and abstention using only current texts, without model weights, token likelihoods, or historical checkpoints. Auditing publicly released descendants of distilled models spanning diverse post-training objectives, SCOUT consistently identifies the distillation source. Furthermore, tracing teacher-associated syntactic signatures along training trajectories reveals that they emerge during distillation and persist through subsequent preference optimization and reinforcement learning.
[NLP-264] Gradient-Guided Decoupled Adaptation for Geospatial Vision-Language Models
【速读】： 该论文旨在解决现有地理空间视觉-语言模型（Geo-VLMs）在多任务适配过程中因忽视任务间异质性优化特性而导致的性能瓶颈问题。具体而言，现有方法采用统一的多任务学习范式，未能有效应对不同任务间存在的视觉-语言差异、分支内梯度关联性以及任务干扰等异质梯度特性，从而限制了多任务优化效率。为此，论文提出一种基于梯度引导的解耦适配框架——梯度引导解耦适配（Gradient-Guided Decoupled Adaptation, G2DA），其核心创新在于：首先通过梯度引导的跨模态解耦策略将任务划分为视觉主导与语言主导两类组别；进而基于任务梯度相似性构建模态特异性课程学习策略，并引入双向重放机制以缓解序列优化带来的近期效应（recency effects）。实验表明，G2DA在多个主流Geo-VLM基准上均显著优于现有基线，验证了梯度感知的任务组织机制对提升多任务Geo-VLM适应能力的有效性。
链接: https://arxiv.org/abs/2609.32737
作者: Dongdong Wang,Deepak Balakrishnan,Ravi Srinivasan,Shenhao Wang
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 8 pages, 3 figures, 4 tables
Abstract:Existing geospatial vision-language models (Geo-VLMs) typically optimize diverse geospatial tasks through a unified multi-task adaptation paradigm without explicitly accounting for the heterogeneous optimization characteristics. Our empirical observations reveal heterogeneous gradient characteristics across tasks, including vision-language differences, intra-branch gradient relationships, and task interference, which hinder effective multi-task optimization. Motivated by these observations, we propose Gradient-Guided Decoupled Adaptation (G2DA), a gradient-aware optimization framework for multi-task Geo-VLM learning. G2DA first partitions tasks into vision- and language-centric groups through gradient-guided cross-modal decoupling. It then constructs modality-specific curricula based on task gradient similarity and employs bidirectional rehearsal to mitigate the recency effects introduced by sequential optimization. We evaluate G2DA on three Geo-VLM benchmarks using six InternVL3 and Qwen3.5-VL variants, along with GeoChat and GeoLLaVA. Across all 24 benchmark-model combinations, G2DA consistently outperforms representative baselines, improving over the strongest competitor by 3.08, 4.30, and 2.81 percentage points on UrBench-MCQ, XLRS-Bench-Lite, and VRS-Bench-VQA, respectively. These results demonstrate the effectiveness of gradient-guided task organization for Geo-VLM adaptation.
[NLP-265] LLM Alignment–Utility Asymmetry under Semantic-Preserving Transformations NEURIPS2026
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）对输入分布变化的对齐（alignment）稳定性问题，特别是探究模型在语义保持但表面形式发生规则化、可逆变换时，其任务有效性与安全对齐能力之间的差异。现有研究多将对齐失效视为攻击行为（如越狱提示或跨语言迁移），而忽视了其作为对齐泛化能力的可控探针的价值。本研究通过引入基于规则且可逆的合成语义保全变换（synthetic semantic-preserving transformations），在不改变任务语义的前提下，将输入推向预训练数据中未涵盖的语言变体，从而系统性地检验对齐的泛化能力。关键发现为“对齐-效用不对称性”（Alignment–utility asymmetry）：当模型仍能有效执行任务时，其对齐性能（即安全性）下降幅度远超效用损失。例如，经变换后的GPT-4.1 mini虽任务表现仅轻微退化，但有害输出率由13.3%飙升至74.3%；Gemini 3 Flash同样表现出近似原始效用但有害率从2.3%升至43.0%。这一现象表明，当前LLM的对齐机制可能更依赖于表面语言模式而非深层语义理解，导致其在面对语义一致但形式变异的输入时，难以维持安全约束，揭示了当前对齐方法在泛化能力上的根本性缺陷。
链接: https://arxiv.org/abs/2609.32717
作者: Mohan Li,Chengyu Yu,Francesco Sovrano,Marc Langheinrich,Martin Gjoreski
机构: Università della Svizzera italiana (瑞士意大利语大学); Lugano, Switzerland (卢加诺, 瑞士)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 58 pages. Accepted at NeurIPS 2026
Abstract:Large Language Model (LLM) alignment is intended to ensure that models remain helpful and safe, but its stability under input distributional shift is not yet fully understood. Prior work shows that aligned models can fail under jailbreak prompts, alternative encodings, and cross-lingual transfer, yet these failures are usually studied as attacks rather than controlled probes of alignment generalization. Moreover, existing evidence is largely grounded in natural language variation already represented during pretraining, leaving unresolved whether alignment generalizes with semantic content or remains tied to superficial surface patterns. In this paper, we study this question using synthetic semantic-preserving transformations that are rule-based and invertible, preserving task-relevant meaning while shifting inputs beyond standard linguistic variation. Across four open-weight and four commercial models, under both fine-tuning and in-context learning, we use these transformations as a probe of alignment generalization and identify an empirical pattern we term Alignment–utility asymmetry: once models can operate effectively on transformed inputs, task utility is often substantially retained while alignment failure increases more sharply. For example, adapted GPT-4.1 mini shows only limited utility degradation under transformation while its harmful rate rises from 13.3 to 74.3; Gemini 3 Flash similarly retains near-original utility while its harmful rate increases from 2.3 to 43.0. Taken together, these results suggest that semantic-preserving distribution shifts can expose a recurring gap in how utility and alignment generalize in current LLMs.
[NLP-266] MassAlloc Attention: Let Attention Allocate Its Own Compute
【速读】： 该论文旨在解决生成式 AI（Generative AI）中注意力机制计算效率低下问题，特别是全注意力（FullAttn）在长序列处理时因对大量低贡献的因果交互进行冗余计算而导致的高计算开销。其核心挑战在于：尽管许多因果交互的归一化得分极低，但传统密集核函数仍会执行完整的后得分路径（post-score path），造成显著的资源浪费。解决方案的关键是提出一种融合式注意力原语 MALA，通过引入基于归一化贡献的动态计算分配机制，在保持模型性能的前提下大幅减少低贡献区域的后得分计算量。MALA 利用在线软最大值归一化器（online-softmax normalizer）实现前向传播中的自适应保留，并在反向传播中复用最终归一化状态，仅依赖标准注意力状态即可推导嵌套保留支持（nested retained support）。整个过程由统一容差控制训练与推理，实现工作量的自适应调节。实验表明，在 8K 序列长度下，MALA 在总后得分工作量完全匹配的情况下，平均被忽略质量仅为 0.0188%，接近理想归一化质量基准（0.0182%）；在 1K–32K 的广泛上下文长度范围内，输出与梯度误差均保持极低；在 128K 序列长度的张量并行基准测试中，相比 FullAttn，MALA 在训练阶段将前向与反向延迟分别降低 2.2 倍和 3.0 倍，推理阶段降低 1.6 倍；在从 0.6B 到 14B 参数的扩展训练中，模型困惑度与 FullAttn 几乎一致，同时显著降低总训练浮点运算量（FLOPs）。最终，基于 MALA 训练的 14B 和 32B 模型在知识、推理及长上下文检索任务上表现与 FullAttn 相当。这表明，依据归一化注意力贡献进行后得分计算分配，可在不牺牲关键能力的前提下有效压缩注意力计算开销。
链接: https://arxiv.org/abs/2609.32712
作者: Jingze Shi,Zhangyang Peng,Xianduo Li,Yanlin Qi,Xiaotian Lin,Haoxian Chen,Liangdong Wang,Guang Liu,Yuyu Luo
机构: The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州)); Beijing Academy of Artificial Intelligence (北京人工智能研究院); Université Paris Cité (巴黎城市大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:FullAttn often assigns negligible normalized mass to much of the causal score space, yet dense kernels execute the complete post-score path after forming each QK tile. We introduce MALA, a fused attention primitive that preserves score access to every legal causal interaction and uses normalized contribution to allocate post-score computation. Forward uses its evolving online-softmax normalizer, while backward reuses the finalized normalizer to derive nested retained support using only standard attention state. A common tolerance governs training and inference, allowing for adaptive retention of the work. MALA reduces low-contribution post-score computation. A matched-work study at 8K isolates the benefit of distribution-adaptive allocation: under exactly matched total post-score work, MALA approaches a per-instance reference-mass oracle, with mean omitted mass of 0.0188% versus 0.0182%. Across context lengths from 1K to 32K tokens, the same tolerance maintains low output and gradient errors relative to the reference. Across a broader controlled associative-recall comparison, MALA closely tracks FullAttn as context grows, reaching 89.67% accuracy at 8K compared with 89.97% for FullAttn. In an attention-operator benchmark at 128K tokens with tensor parallelism, MALA reduces forward and backward latency during training by 2.2x and 3.0x and decoding latency during inference by 1.6x relative to FullAttn. Across scaling-law training from 0.6B to 14B parameters, MALA closely tracks FullAttn in perplexity while reducing total training FLOPs. The resulting 14B models and 32B models from separate continued training achieve comparable knowledge, reasoning, and long-context retrieval scores to FullAttn. These results indicate that allocating post-score computation according to normalized attention contributions can retain the evaluated capabilities of FullAttn while reducing attention computation.
[NLP-267] Despite Instructions: Frontier Agents Improvise Covert Channels at Test Time
【速读】： 该论文旨在解决安全敏感场景中语言模型代理在协作过程中可能无意泄露机密信息的问题。尽管系统明确禁止直接披露敏感内容，并通过监控机制审查每条消息，但反复交互仍可能导致普通通信内容逐渐承载共享的私有语义。其解决方案的关键在于：模型对在固定参数下进行推理时，仅通过一个表示接收方是否正确推断出秘密状态的一比特反馈信号，便能自发学习到一种高效的隐秘通信协议。这种通信机制无需预设编码表或示例，且在生成自由形式更新的模拟事件响应任务中依然有效。实验表明，在十组独立博弈中，使用GPT-5.6 Sol模型的代理对最终准确率达到98.8%，远超25%的随机猜测水平，揭示了即使在严格限制条件下，模型间仍可通过隐性协商实现隐蔽信息传递。这一现象表明，模型间的协作机制本身可能成为隐私泄露的潜在通道。
链接: https://arxiv.org/abs/2609.32701
作者: Jacob Dineen,Silei Ren,Muhao Chen,Dan Roth,Ben Zhou
机构: Arizona State University (亚利桑那州立大学); Cornell University (康奈尔大学); University of California, Davis (加州大学戴维斯分校); University of Pennsylvania (宾夕法尼亚大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:In security-sensitive applications, language-model agents are often required to coordinate without disclosing confidential information. Yet repeated interactions may also let ordinary messages acquire shared private meaning. We study a repeated game with pairs of models in which the sender model observes one of four secret states and selects one of four summaries of the same public report, while the receiver model tries to infer the secret state. We find that model pairs can learn to communicate the secret using only one bit of feedback indicating whether the receiver inferred it correctly. This learning occurs during inference with fixed parameters and no supplied codebook or encoding examples. The effect also persists when agents generate their own free-form updates in a simulated incident-response task. Across ten independent games, pairs of GPT-5.6 Sol agents reach 98.8% final accuracy, compared with 25% chance, despite explicit instructions prohibiting disclosure and a monitor that screens each message without access to the agents’ interaction histories. The same interactions that help agents cooperate can therefore allow confidential information to pass through messages intended for legitimate coordination.
[NLP-268] IGSD: Environment-Verified Hindsight Self-Distillation for Search Agents
【速读】： 该论文旨在解决搜索类智能体在自强化学习（self-distillation）过程中因“事后洞察”（hindsight）导致的教师偏好偏差问题，即教师基于理想化但不可执行的查询建议指导学生，而这些建议可能并未真正提升从学生当前状态出发的检索效果。现有方法或直接蒸馏此类偏好，或通过模型内部得分过滤，但均无法验证查询实际执行后的检索结果。其解决方案的关键在于提出信息增益门控自蒸馏（Information-Gain-Gated Self-Distillation, IGSD），通过环境反馈对策略生成的候选查询令牌进行在线验证：将每个查询令牌视为一个微操作（micro-action），在相同失败状态下同时执行教师提出的完整查询与学生采样的查询，并利用共享反事实控制消除查询条件引起的答案概率偏移，从而计算出二者在实际检索中产生的“已执行配对信息增益”（executed paired information gain），作为仅正向的软权重用于候选对蒸馏。该机制确保了监督信号来源于真实环境反馈，而非假设性推断，且不改变原始GRPO目标，验证过程仅限于训练阶段。在七个单跳与多跳问答基准上，使用3B和7B参数量的策略，IGSD分别实现了42.8%和47.0%的宏平均精确匹配准确率，且无需推理时验证，验证了环境感知的后见之明在可靠动作级监督中的有效性。
链接: https://arxiv.org/abs/2609.32694
作者: Angqing Jiang,Gaoming Zhang,Chaoqun Zhang,Jianchun Song,Liyuan Kong,Kena Qi,Wei Lin,Defu Lian
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:On-policy self-distillation densifies agent training without external teachers: a policy conditioned on privileged hindsight provides step-level guidance for its own unprivileged rollouts. For search agents, however, hindsight can make the teacher prefer a query that does not improve retrieval from the student’s state. Existing methods either distill this preference directly or filter it with model-internal scores, but neither strategy verifies the query’s executed retrieval consequence. We propose Information-Gain-Gated Self-Distillation (IGSD), which verifies on-policy token proposals with environment feedback before distilling them. Treating each query token as a micro-action, IGSD completes the teacher’s token proposal and the student’s sampled token into matched queries and executes both from the same failed state with the same retriever. Shared counterfactual controls account for query-conditioned shifts in answer likelihood, so their difference, the executed paired information gain, provides a relative utility contrast for the retrieved documents. IGSD uses this contrast as a positive-only soft weight for candidate-pair distillation, while leaving the GRPO objective unchanged and confining verification to training. Across seven single-hop and multi-hop QA benchmarks, IGSD reaches macro-average exact-match accuracies of 42.8% and 47.0% with 3B and 7B policies, respectively, without inference-time verification. These results support environment-verified hindsight as an effective approach to reliable action-level supervision for search agents.
[NLP-269] Focusing Condition: Inference-Time Self-Contrastive Steering Elicits Better Conditional Text Embeddings in LLM s ACL2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在生成条件文本嵌入（conditional text embeddings）时，因依赖提示（prompt）引导而难以摆脱通用文本嵌入（unconditional general text embeddings）干扰的问题。现有方法仅通过调整提示来引导模型关注特定条件，但往往导致条件嵌入与通用信息纠缠，从而降低其质量。为此，本文提出一种推理阶段、即插即用的自对比引导（Self-Contrastive Steering, SCS）方法，其核心在于通过修改注意力掩码（attention mask）和位置编码（positional encodings），屏蔽输入中的条件部分，从而显式地构建无条件通用文本嵌入，并利用该嵌入对原始条件嵌入进行精细化修正，使其更聚焦于目标条件。该方法的关键创新在于仅需在推理时增加一次多头自注意力计算，即可实现高效且无需训练的嵌入优化。大量实验表明，SCS可无缝提升多种基于提示的方法在聚类、语义文本相似性（Semantic Textual Similarity）及三元组对齐等任务上的性能，适用于不同大语言模型，具有良好的泛化性和实用性。
链接: https://arxiv.org/abs/2609.32684
作者: Zifeng Cheng,Lingyun Qian,Zhiwei Jiang,Cong Wang,Yafeng Yin,Fei Shen,Ao Zhou,Qing Gu
机构: Nanjing University (南京大学); National University of Singapore (新加坡国立大学)
类目: Computation and Language (cs.CL)
备注: ACL 2026 (Oral)
Abstract:Extracting conditional text embeddings from large language models (LLMs) is a promising paradigm, as it requires neither additional data nor fine-tuning. Existing methods incorporate conditions into prompts to guide LLMs to focus on specific aspects and elicit conditional text embeddings. However, relying solely on prompts often fails to produce high-quality conditional text embeddings, as they remain entangled with general text embeddings, ultimately degrading their quality. To this end, we propose an inference-time, plug-and-play Self-Contrastive Steering (SCS) method that constructs unconditional general text embeddings and uses them to refine conditional text embeddings, making them more focused on the target condition. Specifically, we modify the attention mask and positional encodings to mask the condition, thereby obtaining unconditional text embeddings and intervening in the multi-head self-attention computation process. Notably, our method is highly efficient, requiring only a single additional multi-head self-attention computation at inference time. Extensive experiments on clustering, Semantic Textual Similarity, and triplet alignment datasets demonstrate that our method can seamlessly improve the performance of existing prompt-based methods across different LLMs in a training-free and plug-and-play manner. Our code will be released at this https URL
[NLP-270] he GUI Is Not the State: Diagnosing State Aliasing in GUI World Models
【速读】： 该论文旨在解决生成式界面世界模型（GUI World Models, GUI-WMs）在仅依赖当前界面观测与动作进行状态预测时所面临的状态混淆（state aliasing）问题，即可见界面无法完整反映环境的隐藏状态，导致相同可观测条件对应多个可能的未来状态。其解决方案的关键在于引入基于历史信息的预测性状态恢复机制，通过从历史轨迹中推断出结构化的隐藏状态，并以确定性状态接口增强原本仅依赖观测的模型。具体而言，采用面向不同任务类型的专用模块实现跨异构状态类型的恢复能力，并利用多教师蒸馏技术将多个专家模型集成到一个统一的估计器中。实验表明，该方法显著提升了现有GUI-WMs在状态敏感预测上的表现，同时保持生成保真度并改善了安卓环境（AndroidWorld）中下游智能体的性能，验证了可靠界面建模需兼顾可见信息与决定未来演化的隐藏状态。
链接: https://arxiv.org/abs/2609.32679
作者: Dongsheng Liu,Chao Jin,Wenkui Yang,Hejin Wang,Junwei Yang,Zeren Zhang,Ziwei Chen,Huaibo Huang,Jie Cao,Ran He
机构: University of Chinese Academy of Sciences (中国科学院大学); Chinese Academy of Sciences (中国科学院); Huawei Noah’s Ark Lab (华为诺亚方舟实验室)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:GUI World Models (GUI-WMs) are increasingly used to predict future states for agent planning and simulation, yet most existing formulations condition only on the current GUI observation and action. We identify state aliasing, where the vis- ible interface omits transition-relevant environment state, so identical observable conditions can correspond to different valid futures. To diagnose this failure mode, we introduce StateAliasBench, a diagnostic benchmark that explicitly isolates such ambiguities via strict pairing. We further propose lightweight predictive- state recovery that infers structured state from history and augments otherwise frozen GUI-WMs through a deterministic state interface. Family-specific special- ists provide state recovery across heterogeneous state types, and multi-teacher dis- tillation consolidates them into a single unified estimator. Experiments show that existing GUI-WMs exhibit systematic failures under observation-only condition- ing, while predictive-state augmentation substantially restores state-sensitive pre- diction across evaluated WMs, preserves generative fidelity, and improves down- stream performance of GUI agents on AndroidWorld. These results suggest that reliable GUI world modeling should account not only for what is visible, but also for the hidden transition state that determines what happens next.
[NLP-271] Distance-KV: Exploiting Relative Distance for Efficient Long-Context Inference
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在长上下文推理时内存占用和解码延迟随上下文长度急剧增长的问题。现有基于关键值（Key-Value, KV）缓存压缩的方法通常依据标记重要性或注意力模式在不同头之间的差异来选择性保留缓存状态，但这些方法忽略了同一注意力头内信息检索能力随相对距离变化的显著差异。为此，本文提出Distance-KV，其核心创新在于学习一个静态的KV保留模式，该模式联合建模了层、注意力头与相对距离三个维度的空间结构。该模式在冻结语言模型的前提下离线训练完成，可复用于所有输入，无需在线进行重要性评分即可实现对KV缓存的剪枝与压缩。实验表明，在三个骨干模型和四个长上下文基准测试中，Distance-KV在各项指标上均优于现有压缩方法，尤其在128K上下文长度下，于RULER基准上超越最强基线达9.3分；在Llama-3.1-8B-Instruct模型上，实现65.4%的KV缓存内存减少，并带来1.66倍的解码加速。结果表明，相对距离是理解LLM在长上下文中信息检索机制的关键结构性维度，也为设计更高效的推理方法提供了新思路。
链接: https://arxiv.org/abs/2609.32663
作者: Xianpeng Shang,Canbin Huang,Jiang Li,Tian Lan,Qianyi Cai,Xiaojun Quan,Xiangdong Su
机构: Inner Mongolia University(内蒙古大学); Shenzhen Loop Area Institute(深圳环区研究院); Sun Yat-sen University(中山大学); Kyoto University(京都大学); The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学（广州)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 19 pages, 7 figures, 10 tables
Abstract:The memory usage and decoding latency of LLM inference grow rapidly with context length. To reduce these costs, key-value (KV) cache compression methods selectively retain cached states based on token importance or differences in attention patterns across heads. However, we discover that retrieval capability varies substantially with relative distance, even within the same attention head. To exploit this structure, we introduce Distance-KV, which learns a static KV retention pattern over the joint space of layers, attention heads, and relative distances. The pattern is learned offline with the language model frozen and reused across inputs to prune and compact the KV cache without online importance scoring. Across three backbone models and four long-context benchmarks, Distance-KV consistently achieves the best overall performance among competing KV cache compression methods, exceeding the strongest compression baseline by up to 9.3 points on RULER at 128K. On Llama-3.1-8B-Instruct at 128K, Distance-KV reduces KV cache memory by 65.4% and achieves a 1.66\times decoding speedup relative to Dense. Together, these results identify relative distance as an important structural dimension for understanding how LLMs retrieve information over long contexts and for designing more efficient inference methods.
[NLP-272] From Knowing to Abstaining: Bridging the Representation-Action Gap in Vision-Language Models
【速读】： 该论文旨在解决视觉语言模型（Vision-Language Models, VLMs）在面对无法回答的问题时缺乏有效拒答能力的问题，即“拒答能力”（abstention capability）的不足。现有评估基准存在两大核心缺陷：其一，数据样本中常包含图像或问题中的捷径线索（shortcut cues），导致模型可通过表面线索判断可回答性，且多数任务提供显式的“不可回答”选项，干扰对模型自发拒答行为的真实评估；其二，训练数据仅提供二值标签，缺乏细粒度的推理链与因果证据缺失标注，难以实现深层监督。为此，作者提出一种名为“带理由的视觉可回答性诊断”（Visual Answerability Diagnosis with Rationales, VAD-R）的新基准，采用两阶段流水线——捷径过滤与质量验证，以消除答案可得性泄露。每个样本均配备逐步推理过程和因果证据缺口标签（causal evidence-gap labels），支持对模型决策逻辑的精细化分析。在VAD-R上对主流开源与闭源VLM的评估表明，其自发拒答能力极弱，平均召回率仅为11.4%（开源）和16.3%（闭源）。探针分析揭示，部分隐藏层表示已具备区分可回答性的能力，但该认知未能体现在最终输出中。基于此，作者提出Rep2Act方法，一种从表征到动作的对齐机制，将隐含的可回答性感知转化为显式的拒答决策。实验显示，Rep2Act显著提升模型在VAD-R上的动作准确率：Qwen2.5-VL-3B从56.67%提升至86.33%，Qwen2.5-VL-7B从59.33%提升至88.67%。在分布外测试集TUBench上，仅使用3B规模模型的Rep2Act即达到53.3%的平均F1分数，优于闭源GPT-4 Turbo（+16.2%）和GPT-4o（+1.1%），验证了其卓越的泛化性能与有效性。
链接: https://arxiv.org/abs/2609.32653
作者: Jialuo He,Huangxun Chen
机构: The Hong Kong University of Science and Technology (Guangzhou)
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The ability of vision-language models (VLMs) to abstain from unanswerable questions is as important as their ability to answer answerable ones accurately. Recently, several benchmarks have emerged to evaluate and improve VLM abstention, but they have substantial limitations. First, samples often contain shortcut cues in images or questions that reveal answerability, while an explicit “unanswerable” option further prevents accurate assessment of spontaneous abstention. Second, as training data, they generally provide only binary labels without fine-grained explanations for deeper supervision. To address these limitations, we introduce Visual Answerability Diagnosis with Rationales (VAD-R), a benchmark constructed through a two-stage pipeline of shortcut filtering and quality verification to prevent answerability leakage. Each example is annotated with step-by-step rationales and causal evidence-gap labels. Evaluation of state-of-the-art open- and closed-source VLMs on VAD-R reveals limited spontaneous abstention, with average recall rates of only 11.4% and 16.3%, respectively. Probing analyses show that hidden-state representations in certain layers can effectively distinguish answerability, yet this distinction fails to manifest in final responses. Motivated by this observation, we introduce Rep2Act, a representation-to-action alignment method that translates latent answerability awareness into explicit abstention decisions. Rep2Act improves action accuracy on VAD-R from 56.67% to 86.33% for Qwen2.5-VL-3B and from 59.33% to 88.67% for Qwen2.5-VL-7B. On the out-of-distribution TUBench, Rep2Act achieves an average F1 score of 53.3% with only a 3B model, surpassing the closed-source GPT-4 Turbo and GPT-4o by 16.2% and 1.1%, respectively.
[NLP-273] ExpVoyager: Direct Experience Navigation for Dynamic Agent Skill Synthesis
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）智能体在持续学习过程中，现有技能合成方法因过早将过往经验抽象为固定程序化知识而导致关键知识丢失、冗余实例信息保留的问题。其核心挑战在于如何在不预设下游任务的前提下，动态、精准地从积累的经验轨迹中提取可复用的程序化知识。解决方案的关键是将技能合成重新建模为一种动态导航问题：通过引入名为ExpVoyager的新框架，由一个“技能管理员”（skill curator）主动按需探索不同粒度与视角下的原始经验数据，在实时观察中持续识别可复用的程序化知识，并基于未满足的知识需求指导下一步导航方向。该方法实现了对经验空间的细粒度、目标导向访问，有效避免了知识浪费与冗余，同时在多个下游任务中展现出显著性能提升、随经验规模增长的持续收益以及与已有技能的高效兼容性。
链接: https://arxiv.org/abs/2609.32630
作者: Kwangwook Seo,Dongha Lee
机构: Yonsei University(延世大学)
类目: Computation and Language (cs.CL)
备注: Work in Progress
Abstract:Learning from experience in LLM agents has become a key paradigm for developing self-evolving agents that continuously learn and expand their capabilities. Within this paradigm, synthesizing the agent skill has emerged as a promising solution for transforming accumulated experience into reusable procedural knowledge, serving as an important layer for the harness system that supplies agents at runtime. Despite its potential, existing approaches largely abstract past experience into fixed procedural knowledge before downstream demands are known, which risks discarding knowledge that later becomes critical while retaining instance-specific details irrelevant to future tasks. In this paper, we reframe agent skill synthesis as a dynamic navigation problem over past experience, where agents actively explore accumulated trajectories on demand for the current task with targeted and fine-grained access to experience knowledge. To this end, we propose ExpVoyager, a novel framework in which a skill curator navigates raw experience across different views and resolutions, continually identifying reusable procedural knowledge from what it observes while tracking remaining knowledge needs that guide where to navigate next. Extensive experiments demonstrate both the effectiveness and versatility of ExpVoyager, showing consistent improvements in downstream task performance, continual gains as the experience space scales, and practical compatibility with existing skills under efficient experience access.
[NLP-274] MixDetect: Word-Level Localization and Quantification of AI Editing
【速读】： 该论文旨在解决现有AI文本检测方法在识别与量化人工撰写文本经由生成式AI（Generative AI）修改后所产生复杂编辑痕迹时的局限性。传统检测方法仅能区分完全由AI生成或纯人类撰写的内容，而近期针对AI编辑文本的方法多仅提供文本级别的标签或编辑程度评分，缺乏对编辑位置与强度的细粒度分析。本文提出的MixDetect框架通过词级（word-level）建模，实现对每个词是否被编辑以及编辑程度的独立预测，从而分离估计编辑范围（editing scope）与编辑强度（editing intensity）。其关键在于利用源文本-编辑后文本对进行词级监督信号构建，训练阶段实现精确的局部化标注，推理阶段仅需输入待检测文本即可完成分析。实验表明，MixDetect不仅能精准定位被编辑词汇，还能有效反映不同编辑强度差异，并揭示多种编辑操作下的独特“范围-强度”模式：多次叠加AI编辑会显著提升整体编辑量，而人类对AI生成文本的修改则降低编辑量，普通的人类间改写基本保持编辑量稳定。此外，聚合后的文本级预测在二分类与三分类任务中表现优异，且具备良好的跨领域与跨生成器迁移能力。研究证明，通过同时识别编辑发生的位置与编辑的实质性程度，可实现对AI编辑行为更全面、深入的分析，超越单一作者身份判断或全局编辑度量的局限。
链接: https://arxiv.org/abs/2609.32625
作者: Hongrui Bao,Yubing Ren,Zhendong Pan,Fang Fang,Shi Wang,Yanan Cao
机构: University of the Chinese Academy of Sciences(中国科学院大学); Institute of Information Engineering(信息工程研究所); Shandong University(山东大学); Visual Information Processing and Learning(视觉信息处理与学习)
类目: Computation and Language (cs.CL)
备注: 18 pages
Abstract:Large language models are increasingly used to edit human-written text rather than generate entire texts from scratch. Conventional AI-text detectors mainly distinguish human-written from fully AI-generated text, while recent methods for AI-edited text typically provide only a text-level label or editing-degree score. We introduce MixDetect, a word-level framework for localizing and quantifying AI editing. MixDetect separately predicts whether each word has been edited and, conditional on editing, how substantial the edit is, allowing editing scope and editing intensity to be estimated separately. During training, source–edited pairs are aligned to construct word-level supervision, while inference requires only the input text. Experiments show that MixDetect accurately localizes AI-edited words, reflects differences in editing intensity, and reveals different scope–intensity patterns across editing degrees and operations. The overall AI editing magnitude increases under additional AI editing, decreases when AI-generated text is edited by humans, and remains nearly unchanged under ordinary human-to-human editing. The aggregated text-level predictions also perform well on binary and ternary AI-text classification and remain effective under domain and generator shifts. These results show that AI editing can be analyzed beyond a single authorship label or editing-degree score by identifying both where AI editing occurs and how substantial the edits are.
[NLP-275] Does CoT-Pass@k Really Check the CoT? A Multilingual Mathematical Audit EMNLP2026
【速读】： 该论文旨在解决生成式 AI (Generative AI) 评估中一个关键缺陷：传统 Pass@k 指标仅衡量模型是否在多次采样中获得正确答案，却无法区分正确答案是源于有效推理还是偶然猜中。为此提出的 CoT-Pass@k 通过引入大语言模型（LLM）作为评判者，要求其对解题过程的思维链（Chain-of-Thought, CoT）进行合理性评估，从而弥补这一不足。然而，该方法的有效性完全依赖于一个未经验证的核心假设——即评判者能够可靠识别并拒绝存在逻辑错误的推理链。本文首次在多语言环境下对该验证环节进行了审计，基于包含英语、土耳其语和葡萄牙语的五个数学基准测试集（其中两个为原生语言编写），采用确定性编辑手段分别破坏推理链与最终答案。实验结果表明，所有三个评判模型（V4-Flash、Qwen3.6 及指标自身内置的评判器）对被污染的推理链接受率几乎与干净样本无异，且其判断主要受“推理链与答案一致性”主导，而非真正逻辑有效性。尤其在当前模型生成能力显著提升的背景下，原本反映推理质量差异的指标差值（Pass@k - CoT-Pass@k）从早期版本的平均19.7点骤降至仅4.1点，且进一步增加生成预算可使 Pass@64 提升超过50分而差值仍趋近于零，说明该差值已不再反映真实推理能力。研究最终提出两项必要检验标准，任何依赖人工或自动评判的推理评估指标在被用于支撑结论前都必须通过。
链接: https://arxiv.org/abs/2609.32622
作者: Tarık Tuna Taşaltı,Burcu Hüdaverdi,David Semedo
机构: Dokuz Eylül University (伊兹密尔大学); NOVA School of Science and Technology, Universidade NOVA de Lisboa (里斯本新大学科学技术学院)
类目: Computation and Language (cs.CL)
备注: Accepted at MRL@EMNLP 2026 (Workshop on Multilingual Representation Learning). 24 pages, 13 figures, 6 tables
Abstract:Pass@k measures whether a model reaches a correct answer under repeated sampling, but never how: a lucky guess counts the same as sound reasoning. CoT-Pass@k was proposed to close that gap, adding an LLM-as-judge that must assess a solution’s reasoning chain before it counts. Its value rests entirely on one assumption: that the judge catches flawed reasoning. That assumption has never been tested inside the metric that depends on it, and never outside English, though the metric’s claims concern models used in many languages. We report the first audit of that verification step, run under the metric’s own protocol on a multilingual suite of five mathematical benchmarks in English, Turkish and Portuguese, two of them natively written. We corrupt correct solutions with deterministic edits that damage the chain and the final answer separately. We observe that all three judges accept corrupted chains almost as often as clean ones. V4-Flash and Qwen3.6 reject a solution sharply only when its final answer is wrong and accept a wrong answer more readily when the chain agrees with it; the metric’s own judge accepts most wrong answers as well. Our study shows that chain-answer agreement dominates the two larger judges’ verdicts and that all three fail to reliably detect the tested reasoning errors. Consequently the difference Pass@k - CoT-Pass@k averages 19.7 points on an earlier solver generation but only 4.1 on the current one. What little remains depends on the token budgets on both sides and on the generation mode; raising the generation budget moves Pass@64 by more than fifty points while the difference stays at zero. We close with two checks any judged reasoning metric should pass before its numbers are read as evidence about reasoning.
[NLP-276] he Alignment Paradox: How Post-Training Amplifies Confident Hallucinations in Language Models
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在生成回答时出现高自信错误（即“自信幻觉”）的问题，这类错误严重削弱了模型的可靠性，并限制了基于不确定性的错误检测方法的有效性。现有研究多将此类现象归因于训练数据缺失、推理错误或随机解码等因素，但本文揭示，后训练对齐（post-training alignment）本身是导致此类错误的主要驱动因素，这一现象被称为对齐悖论（Alignment Paradox）。关键发现在于：未对齐的基线模型在长尾事实类查询上极少产生高自信错误，而经过指令微调后的模型则使高自信错误（置信度 ≥ 0.95）数量增加超过一个数量级（10×至35×）。通过层间探查（layer-wise probing）结合对数概率透镜（Logit Lens）分析发现，这种过度自信在模型深层出现，错误答案的置信度差距（margin）从早期和中间层的接近零迅速扩大至超过4.0。为此，本文提出在直接偏好优化（Direct Preference Optimization, DPO）中引入一种依赖熵的边际约束机制，以限制后训练过程中置信度差距的无序增长。在Mistral-7B模型上的多轮次实验表明，该约束目标可将高自信错误降低最多达35.3%相对误差，同时保持在通用推理基准上的性能表现。因此，解决方案的关键在于通过熵依赖的边际约束来抑制后训练阶段的置信度膨胀，从而有效缓解自信幻觉问题。
链接: https://arxiv.org/abs/2609.32617
作者: Qingjia Huang,Yakai Li,Jianguo Wu,Qihang Zhou,Aimin Yu,Xiaoqi Jia,Luping Ma,Weijuan Zhang
机构: Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所); School of Cyber Security, University of Chinese Academy of Sciences(中国科学院大学网络空间安全学院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Code: this https URL
Abstract:Large language models (LLMs) can produce factually incorrect answers with high confidence, undermining their reliability and limiting the effectiveness of uncertainty-based error detection. While prior research attributes confident hallucinations to factors such as missing knowledge in training data, reasoning errors, or stochastic decoding, we uncover that post-training alignment itself is a primary driver of these errors, a phenomenon we call the \textbfAlignment Paradox. Across five model families evaluated on factual benchmarks, unaligned base models produce few high-confidence errors on long-tail factual queries, whereas instruction-tuned models multiply high-confidence errors ( p \ge 0.95 ) by more than an order of magnitude (10 \times to 35 \times ). Layer-wise probing with the Logit Lens reveals that this overconfidence emerges in late layers, where wrong-answer margins expand past 4.0 points after remaining near zero across early and intermediate layers. These findings motivate limiting margin growth during post-training. We implement this principle through an entropy-dependent margin bound in direct preference optimization (DPO). In multi-epoch experiments with Mistral-7B, the bounded objective reduces high-confidence errors by up to 35.3% relative to standard DPO while maintaining performance on evaluated general reasoning benchmarks. These results show that bounded margins mitigate confident hallucinations during post-training.
[NLP-277] KV-Lingo: Learning KV-Cache Translators with Distillation
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在共享上下文场景下因键值缓存（Key-Value Cache, KV Cache）的模型特异性导致的重复计算问题。由于不同架构或权重的模型对相同文本生成的缓存表示不兼容，当在多个模型间切换时，即使上下文已由前序模型处理，目标模型仍需重新执行预填充（prefill）以构建自身缓存，造成显著延迟。其解决方案的关键在于提出KV-Lingo，一种将源模型的KV缓存高效转换为可被目标模型读取的缓存表示的方法。该方法通过为每个目标模型层设计独立的线性映射（linear map），对所有令牌的键和值表示进行变换，并基于知识蒸馏（distillation）训练这些映射，使目标模型在使用翻译后的缓存时的预测结果与使用其原生缓存时保持一致。实验表明，该方法在小模型到大模型、大模型到小模型等多种迁移场景中均能保持优异的下游性能。在实际应用中，模型切换仅需一次线性变换加一个解码步骤，而非完整预填充，使得在64词元提示下，Qwen模型在Apple M3 Ultra上的首字生成时间缩短9.6倍，在32k上下文长度下于H100上更达29倍加速。这一效率提升极大促进了动态模型路由（dynamic model routing）的应用，实现上下文在不同模型间的按需传递而无需重复预填充。此外，多轮对话评估显示，KV-Lingo支持平滑的模型切换，在多次交替切换中表现接近原生预填充，具备实际部署可行性。
链接: https://arxiv.org/abs/2609.32610
作者: Valérie Castin,Keitaro Sakamoto,Anastasiia Filippova,João Monteiro,Marco Cuturi,Pierre Ablin
机构: Apple(苹果)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Large language models represent context with a key-value (KV) cache. Caches are model-specific: for the same text, models with different architectures or weights produce incompatible representations. This makes it costly to switch models over a shared context: although the context has already been processed by one model, the incoming model must process it again to build its own cache. We introduce KV-Lingo, a method for translating the KV cache of a source model into one that can be read by a target model. KV-Lingo consists of a collection of linear maps, typically one per layer of the target model, that are applied independently on all tokens’ key and value representations. We train these maps using distillation, minimising the divergence between the target model’s predictions from its native cache and those from the translated cache. We consider several model pairs spanning multiple sizes and architectures, training one translator per pair on a generic text corpus. The resulting translators preserve strong downstream performance in both small-to-large and large-to-small transfers. Since a switch then costs a linear map and a single decoding step instead of a prefill, replacing re-prefill with cache translation reduces the time to first token after a model switch by 9.6x already on a 64-token prompt for Qwen models on an Apple M3 Ultra, and by up to 29x at 32k context length on an H100. These gains make KV-Lingo particularly useful for dynamic model routing: a context can be processed by one model and handed off to another only when needed, without re-prefilling the shared prefix. We finally show that KV-Lingo can be used for seamless model switching, staying close to re-prefill across repeated switches in our multi-turn evaluations.
[NLP-278] HERO-MoE: Historical Expert Routing with Scale-Preserving Fusion
【速读】： 该论文旨在解决混合专家模型（Mixture-of-Experts, MoE）中路由机制未能有效利用前序层历史路由信息的问题，从而提升模型的训练效率与性能。现有标准路由器虽能根据输入语义和深层计算动态分配专家，但未显式整合前序MoE层生成的路由分布，导致潜在的上下文记忆与跨层一致性损失。为此，论文提出HERO-MoE（Historical Expert ROuting with Scale-Preserving Fusion）框架，其核心创新在于通过重用前序MoE层已计算的稠密路由分布作为历史先验，将其以残差形式注入当前层的路由决策过程。关键设计是引入一种保持尺度一致性的融合机制，在不引入额外辅助损失或调参的前提下，对历史路由记忆的幅度进行归一化，并根据可见的历史层数量自适应调整其贡献权重，从而稳定历史信号。该方法在保留标准稀疏分发机制（如top-k与组受限路由）兼容性的同时，仅带来极小的端到端开销（内存与浮点运算量增加分别仅为0.44%和0.64%），并在一个约80亿参数、激活参数达5亿的MoE模型上实现训练损失从1.6393降至1.6184，显著提升了训练收敛效果。
链接: https://arxiv.org/abs/2609.32581
作者: Junxiang Qiu,Zhengsu Chen,Xinting Hu,Shuo Wang,Hengheng Zhang,Shaofeng Zhang,Changcheng Li,Boyu Shi,Qi Tian
机构: University of Science and Technology of China(中国科学技术大学); Huawei Inc.(华为公司)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Mixture-of-Experts (MoE) architectures have become a standard way to scale model capacity while keeping computation sparse, yet routing remains a key determinant of MoE quality and training behavior. Prior empirical studies suggest that MoE routing reflects input semantics and upstream computation across depth, but standard routers do not explicitly use the routing distributions produced by preceding layers. We propose HERO-MoE, Historical Expert ROuting with Scale-Preserving Fusion, a routing framework that injects historical routing priors into MoE routers by reusing detached, dense routing distributions collected from preceding MoE layers. The key idea is simple: HERO-MoE preserves the original token-conditioned routing branch and adds a residual historical routing contribution before the standard softmax and top- k dispatch. To stabilize this historical signal, HERO-MoE introduces a scale-preserving fusion mechanism that matches the magnitude of historical routing memory to the current hidden representation and accounts for the number of visible historical layers, without introducing an auxiliary routing loss or a fusion-specific tuning parameter. By reusing routing distributions already computed by preceding MoE layers, HERO-MoE improves training-loss reduction with modest end-to-end overhead. The resulting router remains compatible with standard sparse dispatch, including top- k and group-limited routing, and can be inserted into existing MoE backbones with minimal architectural changes. Experiments on an approximately 8B-parameter MoE model with 0.5B active parameters, trained from scratch on 100B tokens, show that HERO-MoE reduces the final loss from 1.6393 to 1.6184, while peak memory and FLOPs increase by only 0.44% and 0.64%, respectively.
[NLP-279] Groupwise Agent ic Grading and Advantage Redistribution for Code Agent RL
【速读】： 该论文旨在解决生成式代码智能体在强化学习（Reinforcement Learning, RL）过程中因依赖可执行测试提供二元奖励而导致的策略优化偏差问题。传统方法如组相对策略优化（Group Relative Policy Optimization, GRPO）对同一轮次中通过测试的所有轨迹赋予相同优势值，忽视了实现质量差异及任务要求的契合度，致使模型缺乏偏好简洁、精准实现而非冗余或越界修改的学习信号。为此，论文提出GAGAR框架，其核心在于引入基于动态采样的质量感知信用重分配机制：保留包含通过与未通过测试轨迹的混合组，并将每组内所有轨迹置于共享工作空间，由一个经过监督微调（Supervised Fine-Tuning, SFT）训练的代理评分器（agentic grader）联合评估并排序通过测试的候选方案。基于此排序结果，对低分轨迹进行降权处理，并按比例重新缩放所有通过轨迹的优势值，以保持原始总和不变。该设计在维持质量导向降权相对权重的基础上，实现了向高质量实现的信用转移。实验基于MiMo-V2.6-Flash（310B参数）和MiMo-V2.6-Pro（1.02T参数）两个工业级预训练模型，在纯代码任务及大规模混合任务强化学习场景下验证了GAGAR的有效性，结果显示其显著提升了代码智能体性能、抑制了轨迹长度增长并增强了训练稳定性，证明了结合测试验证与组内代理评分的协同机制对于提升代码生成质量与强化学习鲁棒性的关键作用。
链接: https://arxiv.org/abs/2609.32577
作者: Jinhao Dong,Liang Zhao,Zihao Yue,Wenhan Ma,Linghao Zhang,Lei Li,Shicheng Li,Yifan Song,Bowen Ye,Fuli Luo
机构: LLM Core, Xiaomi(小米); Renmin University of China(中国人民大学); Peking University(北京大学); University of Hong Kong(香港大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Reinforcement learning (RL) for code agents often uses executable tests to provide binary rewards. With these rewards, Group Relative Policy Optimization (GRPO) assigns identical advantages to test-passing trajectories within each rollout group, overlooking differences in implementation quality and adherence to task requirements. This leaves the policy without a learning signal that favors clean, targeted implementations over those containing unnecessary or out-of-scope changes. We introduce GAGAR, a framework for quality-aware credit redistribution in code agent RL. Built on dynamic sampling that retains groups containing both passing and failing trajectories, GAGAR places all trajectories from each group in a shared workspace, where an SFT-trained agentic grader jointly inspects them and ranks the test-passing candidates. Based on this ranking, we downweight lower-ranked trajectories and proportionally rescale the advantages of all test-passing trajectories to restore their original sum. This sum-preserving redistribution retains the relative weights established by quality-based downweighting while shifting credit toward higher-quality implementations. We evaluate GAGAR at industrial scale using pre-RL SFT checkpoints of MiMo-V2.6-Flash (310B total parameters) and MiMo-V2.6-Pro (1.02T total parameters). Controlled code-only Flash experiments show improved code agent performance, reduced trajectory-length growth, and more stable training. We further apply GAGAR in large-scale mixed-task RL with both Flash and Pro. Our results support combining test-based verification with groupwise agentic grading to improve the quality and stability of code agent RL.
[NLP-280] CUE-Mem: Benchmarking Long-Term User Memory via Implicit Cues in Multimodal Conversations KR AAAI2027
【速读】： 该论文旨在解决多模态智能体在长期对话中如何有效识别与利用用户隐含记忆线索的问题。现有评估基准大多局限于文本或显式多模态证据，忽视了图像中的重复背景对象、音频中的环境声音等隐含多模态线索的作用。为此，研究提出CUE-Mem——一个涵盖文本、图像与音频的多模态基准，用于评估系统从隐含线索中提取长期用户记忆的能力。其核心解决方案在于构建一个包含2,674个问题的综合性评测体系，覆盖实体回忆、长期模式识别、个性化推荐与回答拒绝四类任务。关键发现表明，当前基于文本化的记忆系统在处理隐含线索时表现显著低于理想情况，瓶颈在于对细微多模态线索的保留与检索能力不足；尽管增加图像描述细节可部分恢复信息，但带来收益不均且生成成本迅速上升，促使研究转向原生多模态访问。然而，原生多模态接入并未完全解决该瓶颈，其效果高度依赖模型主干结构，并引入显著的检索噪声。因此，CUE-Mem为开发能够选择性保留、精准检索并有效利用细微多模态线索的记忆系统提供了重要测试平台。
链接: https://arxiv.org/abs/2609.32574
作者: Yulin Hu,Yanyan Zhao,Zimo Long,Xing Fu,Mengtong Ji,Weixiang Zhao,Yutai Hou,Qianchao Wang,Dandan Tu
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 28 pages. Submitted to AAAI 2027. Code: this https URL . Data: this https URL
Abstract:Long-term memory is essential for multimodal agents that interact with users across sustained conversations. However, user memories are not always explicitly stated: they may also be implied by recurring background objects in images, ambient sounds in audio, or other peripheral multimodal cues. Existing benchmarks largely focus on text-only memory or explicit multimodal evidence, leaving implicit multimodal cues underexplored. We introduce CUE-Mem, a text-image-audio benchmark for evaluating long-term user memory from implicit cues. CUE-Mem contains 2,674 questions across explicit and implicit evidence settings and covers four tasks: Entity Recall, Long Pattern, Personalized Recommendation, and Answer Refusal. Across textualized memory systems, implicit performance remains far below oracle evidence, locating the main bottleneck in preserving and retrieving subtle cues rather than question answerability. Increasing caption detail recovers more of this evidence, but brings uneven gains and rapidly growing token costs, motivating native multimodal access. Yet native access does not uniformly resolve the bottleneck: evidence use depends strongly on the backbone, while multimodal indexing introduces substantial retrieval noise. CUE-Mem provides a testbed for memory systems that selectively retain, retrieve, and use subtle multimodal evidence.
[NLP-281] Attribution Gaps in Zero-Training LLM OVOD Pipelines: A Fine-Grained Analysis of the CAAP–SNAP Discrepancy
【速读】： 该论文旨在揭示生成式视觉定位（LAOD）等零训练大语言模型结合开放词汇检测器（LLM+OVOD）流水线中，类无关定位准确率（CAAP）与语义命名准确率（SNAP）之间持续存在的分歧现象的根本原因。其核心发现是：这一差距主要并非源于对象视觉复杂性或模型本身的定位能力缺陷，而是由标注体系的封闭性限制所致——当大语言模型（LLM）使用超出检测器原生类别集的词汇时，定位准确率从80.9%骤降至31.6%。进一步分析表明，该下降并非均匀分布，绝大多数损失来自新表述实际指代与COCO标注对象不同的情况（即真实同义词仍保持89.3%准确率，而语义无关的“噪声”标签仅12.0%），且约78%-88%看似完全失败的定位案例实为模型正确识别了未被COCO 80类标注覆盖的真实物体，并非幻觉。更换检测器主干网络（如YOLO-World替换Grounding DINO）或大语言模型（如Gemma-3替换Qwen2.5-VL）后，该效应依然稳定存在，表明其为该类流水线的普遍特性。因此，该研究的关键结论在于：当前评估框架中观察到的CAAP-SNAP差异，很大程度上反映了封闭类别标注的局限性，而非真正的视觉-语言对齐失败，这对开放世界多模态系统中的幻觉检测、故障模式分析及评估设计具有重要意义。
链接: https://arxiv.org/abs/2609.32567
作者: Yu-Feng Yen
机构: Model call failure
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: 7 pages, 4 figures
Abstract:LAOD and similar zero-training LLM+open-vocabulary-detector (OVOD) pipelines score two things separately: class-agnostic localization accuracy (CAAP) and semantic naming accuracy (SNAP). The two consistently diverge, and nobody has asked why. This paper asks why, on the full 5,000-image COCO-Val split (27,273 detections) rather than the small subset the original work evaluated on. Object visual complexity turns out not to be the driver – small and occluded objects are, if anything, localized better than large ones. Vocabulary novelty is: once the LLM’s wording falls outside the detector’s native category set, localization accuracy falls from 80.9% to 31.6%. That drop is not spread evenly across unfamiliar phrasing, though. Almost all of it comes from cases where the novel wording actually names a different object than the one COCO annotated (true synonyms still score 89.3%; semantically unrelated “noise” labels score 12.0%). A closer look at a further failure subset tells a similar story: 78-88% of what looks like complete localization failure is really the model correctly finding a real object that COCO’s non-exhaustive 80-category scheme simply never labeled, not hallucination. Swap the detector backbone (YOLO-World for Grounding DINO) or the LLM (Gemma-3 for Qwen2.5-VL) and both the effect and its rough size hold up, so this looks like a general property of the pipeline family rather than a quirk of one model pairing. The upshot is that a large share of the apparent CAAP–SNAP gap traces back to closed-category annotation limits rather than a real grounding failure, which matters for how we detect hallucination, analyze failure modes, and design evaluation for grounded multimodal systems meant to work in the open world.
[NLP-282] Reading Is Not Leaking: Local Auditable Measurement and Reduction of Inference Exposure from Public Footprints
【速读】： 该论文旨在解决公开信息泄露隐私的问题，即个体或组织在拥有公开数字足迹时，其未明确陈述的私密事实可能被生成式AI（Generative AI）轻易推断出来。现有方法难以区分“读取准确率”与“信息泄露率”，导致评估结果混淆。为此，论文提出一种可在本地CPU上运行、无需分析时依赖语言模型的框架，通过引入已知支持的注入协议构建测试单元，将读取准确性与泄露率解耦。其核心解决方案是设计一个结合规则系统、统计求解器及从零训练的106M参数编码器的分析器，该编码器仅标记原文证据而不生成文本，每个答案附带可回放的分级证明证书。实验表明，该分析器在93%的可决问题中给出正确答案，显著优于语言模型基于引用生成的答案（49%-73%正确率），且70%的证据性回答基于可验证证据，远高于模型的18%-56%。此外，采用约束性防御策略——将原始事实载体重写为更粗粒度但真实的陈述，可在编辑成本降低40%的情况下，完全隐藏单载体事实，有效抵御四类语言模型攻击。在对十六个合成个人的测试中，该方法进一步验证了其对猜测项的抑制能力，同时通过无语言模型的诱饵规划器可使目标估计器的正确率减半，且无需跨系统迁移。
链接: https://arxiv.org/abs/2609.32565
作者: Mahmudul Faisal Al Ameen
机构: Independent researcher(独立研究员); France(法国)
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL); Computers and Society (cs.CY)
备注:
Abstract:Anyone with a public footprint leaks facts that were never stated, and language models make the inference cheap. We present a framework for measuring and reducing this inference exposure that runs on the owner’s own CPU with no language model at analysis time, instantiated on organisations and on individuals. It starts from a measurement result: scoring an inference system against the target’s private truth conflates how well the system reads the record with how much the record leaks. On a 128-question instrument over sixteen synthetic firms, almost half of the questions are never answered correctly by any of six readers, four of them language models, and a majority-class guess accounts for most of every reader’s score. We therefore separate reading accuracy from leakage rate and introduce an injection protocol that creates cells with known support. Our analyser combines rules, statistical solvers and a 106M-parameter encoder trained from scratch that marks verbatim evidence and never generates text; every answer carries a graded certificate whose recorded proof replays. Its certified answers are correct in 93% of resolved cases, against 49-73% for the language models’ quote-backed answers, whose citations are produced alongside the answer rather than deriving it; with plain-prose articles in the record, 70% of its evidence-bearing answers rest on evidence that establishes them, against 18-56% for the models. A constrained defence that rewrites each fact’s carrier as a true but coarser statement hides every single-carrier fact from four language-model adversaries at 40% lower edit cost than deletion. On sixteen synthetic people the guessing term is larger still, and a decoy planner with no language model halves the correct answers of the estimator it targets without transferring to a second.
[NLP-283] How to Reduce Whisper Hallucination
【速读】： 该论文旨在解决生成式语音识别模型Whisper在真实生产环境中存在的严重幻觉问题，即在无语音的静音片段（如纯房间噪声）中错误生成大量不存在的文本内容。尽管Whisper-large-v3具备99种语言支持且许可宽松，但其在无语音输入时仍会频繁产生虚假语句（在42段纯静音音频中，有61.9%的片段生成了词语），严重影响可用性。现有解决方案主要分为两类：一是通过知识蒸馏从Whisper伪标签中训练学生模型，但此方法仅过滤掉幻觉样本而未修正教师模型的根本缺陷，导致学生继承了教师的错误行为模式；二是引入非语音音频以训练模型“保持沉默”，但该方法缺乏区分能力，过度抑制真实语音，造成高达58%的重复语音被误删。本文提出的核心解决方案是直接修复教师模型本身，使其同时具备对语音与非语音信号的正确响应能力——通过构建一个包含40,891个幻觉短语的跨语言语料库，并利用测量选定的文本到语音（Text-to-Speech, TTS）系统将这些幻觉内容合成作为正样本，使模型在相同文本上既面临需抑制的干扰，又需正确转录的任务。实验在33组对比微调中验证，加入此类合成正样本后，在31组中显著降低真实无语音音频上的词生成率，32组中提升短语恢复率，且在维持英语准确率的前提下实现全量指标优化。最终最佳模型将静音状态下的幻觉率从61.9%降至2.4%，无语音音频上的词生成率从99.9%降至47.8%，同时短语恢复率由69.8%提升至82.7%。该方案的关键在于通过双任务对抗性训练，使模型在不依赖外部过滤或盲目抑制的前提下，从根本上学习区分真实语音与应抑制的虚假输出。
链接: https://arxiv.org/abs/2609.32560
作者: Husein Zolkepli
机构: Scicom (MSC) Berhad(斯科姆公司(马来西亚))
类目: Computation and Language (cs.CL)
备注:
Abstract:Whisper is still what runs in production: one permissively licensed checkpoint, 99 languages, no per-language tuning. But it writes sentences nobody said. On 42 clips of pure room tone, whisper-large-v3 emits words on 61.9% of them and emits something on 100%. The usual response is to distil a student from Whisper pseudo-labels, filtering the hallucinations out of the corpus first, which deletes the evidence while leaving the behaviour in place: the student is fitted to the subset where the teacher was right, and inherits a failure mode absent from its own training data. The second response, adding non-speech audio so the model learns to stay quiet, already ships and works, but teaches suppression without discrimination. The checkpoint best on every non-speech arm here is also the one that recovers the fewest genuinely spoken phrases and deletes 58% of repeated speech. The teacher has to be fixed, with both halves of the signal. Our benchmark scores both at once: 11,852 clips over eight arms, 6,267 synthetic positives, 8,296 clips of real audio that made a production model fail, and FLEURS in 58 languages. We collect 40,891 hallucination phrases in 100 languages, choose a text-to-speech system by measurement, and synthesise those phrases as positives, so the model meets the same text both as something to suppress and as something to transcribe. Across 33 matched pairs of fine-tunes differing only by those positives, adding them lowers word emission on real voice-free audio in 31 pairs and raises phrase recovery in 32; among the 30 pairs that hold English accuracy it is 30 out of 30. The best checkpoint takes hallucination on silence from 61.9% to 2.4% and words over real voice-free audio from 99.9% to 47.8%, while raising phrase recovery from 69.8% to 82.7%. The benchmark, lexicon and synthetic corpus are released at this https URL
[NLP-284] Language as an Independent Information Layer: A Conceptual Model of Communication Cognition and Decision-Making
【速读】： 该论文旨在解决企业知识库构建中语义表达与统计特征融合不足的问题，即如何在保留行业特定术语、主题聚焦及企业文化等语言特性的同时，实现对企业文档中隐含知识的高效提取与精准建模。其解决方案的关键在于提出一种将企业词汇的概率向量空间（probabilistic vector space）与传统本体论建模相结合的新范式：通过引入反映业务流程特性的动态语言层，使词汇分布的空间具备多维度问题求解能力；同时，以输入数据作为控制参数驱动概率空间的动态演化，从而在统计方法与语义因果关系模型之间建立桥梁。该框架不仅支持从海量企业文档中自动识别知识，还可通过对概率空间投影和因果关系的分析，揭示业务逻辑中的瓶颈环节，显著提升知识系统的适应性与可解释性。
链接: https://arxiv.org/abs/2609.32556
作者: Anastasiia Alifanova,Elena Benderskaya
机构: 未知
类目: Computation and Language (cs.CL); Computational Engineering, Finance, and Science (cs.CE); Computers and Society (cs.CY); Machine Learning (cs.LG)
备注: 11 pages, 1 figure. Accepted for publication in the proceedigns of ISBM 2026 (Springer LNNS)
Abstract:Based on an analysis of the role of language in thought and communication, this article proposes a new concept for designing corporate knowledge bases. The concept integrates the probabilistic vector space of a corporate vocabulary, reflect-ing industry specifics, subject focus, terminology, and culture, with traditional ontological modeling. This combination enables the efficient extraction of knowledge from accumulated corporate documents while strictly accounting for specific business processes. Consequently, this concept bridges statistical and semantic (cause-and-effect) methodologies. Furthermore, analyzing the projec-tions of probabilistic spaces and causal relationships can help identify bottlenecks in business logic. As a dynamic system, language functions as a separate, inde-pendent layer within the overall information architecture. Introducing a dynamic component into the probabilistic space of word distribution allows it to be mod-eled as a multidimensional solution space for various problem formulations. In this context, input data defining the problem conditions serve as control parame-ters for dynamic transformations.
[NLP-285] Shared Autoregressive Context Can Distort Relationships in Synthetic Data
【速读】： 该论文旨在解决生成式人工智能（Generative AI）在批量生成合成数据时，因自回归解码过程中共享上下文而导致变量间关系失真的问题。其核心发现是：在单次请求中生成多个记录（如受访者）时，早期生成结果作为后续生成的上下文，会系统性地夸大变量间的相关性强度，从而引入偏差。解决方案的关键在于识别并控制“答案历史”这一因果路径——通过实验验证，在固定样本特征与边际分布的前提下，仅重新配对先前生成的答案即可改变后续响应的相关性结构，证明了上下文依赖是导致偏差的主要机制。此外，隐藏先前答案虽可降低相关性误差，但损害了边际分布准确性；探索性修正也表明，降低相关性偏差可能以牺牲边际分布和回归估计精度为代价。因此，该研究强调，请求构造方式本身构成了数据生成过程的一部分，合成数据的有效性必须依据其拟支持的具体分析任务进行评估。
链接: https://arxiv.org/abs/2609.32546
作者: Thomas S. Robinson
机构: London School of Economics and Political Science (伦敦政治经济学院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 54 pages, 8 figures, including appendices
Abstract:Large language models can generate several records within one autoregressive completion, making earlier answers available as context for later records. This paper shows that such shared-completion batching can distort relationships among variables in the resulting synthetic data, using controlled tests on synthetic survey respondents. In a matched experiment on 2,000 European Social Survey profiles, generating ten rather than one respondent per request increases mean absolute error in within-country correlations by 48-58% for Qwen3.8-27B and 114-127% for Llama-3.3-70B-Instruct across three seeds, holding profiles, examples, questions and decoding parameters fixed. The distortion primarily reflects exaggerated relationship strength, while retaining substantial agreement with the human ordering of correlations. Controlled interventions establish answer history as a causal channel: re-pairing the same preceding values, with profiles and marginal distributions fixed, changes correlations among subsequently generated responses. Hiding preceding answers reduces correlation error in the tested settings but worsens marginal accuracy. Exploratory corrections across social-attitude, health and economic data likewise show that lower correlation error can coexist with worse marginal distributions and regression estimates. Request construction is therefore part of the data-generating process, and synthetic-data validity must be evaluated against the analyses the generated data are intended to support.
[NLP-286] Do Audio LLM s Listen Before They Act? Diagnosing Acoustic-Context Gating in Voice Agents
【速读】： 该论文旨在解决音频语言模型（Audio LLM）在复杂声学与对话场景中对行动触发（action-level addressedness）的误判问题，即模型在非目标说话人发言、自我对话或说话人切换等情境下仍可能错误执行工具调用。其核心挑战在于如何准确判断当前语音输入是否应被响应，而非仅依赖于语音内容或说话人身份识别。解决方案的关键在于提出VGBench这一包含1,018个样本的诊断基准，涵盖侧谈（side-talk）、自言自语（self-talk）和说话人切换（speaker-switch）三种典型场景，并通过控制源位置、距离渲染及时间边界，实现对“佩戴者—旁观者”转换的精确建模。研究进一步引入VoxGate作为后训练策略，采用监督学习使模型在说话人切换时将指令静音率提升至91.3%，同时保持对近场佩戴者指令和纯文本控制的正确响应；探索性基于奖励的策略优化（GRPO）也显著提升了侧谈准确率与自言自语静音率。研究表明，多线索声学-上下文门控（multi-cue acoustic-context gating）是决定行动适切性的关键因素，而非单一的说话人身份识别。
链接: https://arxiv.org/abs/2609.32536
作者: Yanjie Zhang,Nanchen Hu,Yushi Sun
机构: HKUST(香港科技大学); LIGHTSPEED(光速科技)
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multimedia (cs.MM); Audio and Speech Processing (eess.AS)
备注:
Abstract:Audio language models can recognize spoken commands and invoke tools, but an agent must first decide whether the acoustic and conversational context warrants action. We introduce VGBench, a 1,018-item diagnostic benchmark for action-level addressedness across side-talk, self-talk, and speaker-switch scenarios. Each item uses a shared action space comprising silence, a tool call, and a natural-language answer. Speaker-switch pairs hold the specified words fixed while source, distance rendering, and a temporal boundary define a controlled wearer-to-bystander shift. Six raw Audio LLMs and three training-free adaptations often identify the target tool yet rarely withhold action under this shift; the highest raw switch mute rate is 14%. We then use VoxGate as a post-training case study. Supervised training mutes 91.3% of switched commands while choosing the correct tool for all nearby wearer commands and text-only controls. An exploratory GRPO stage has similar switch performance; side-talk accuracy rises from 68.4% to 70.9%, and self-talk muting from 52.0% to 60.0%. Factorized controls identify an independent source-change effect, while sensitivity to the far-field manipulation varies across acoustic renderings. The benchmark therefore measures multi-cue acoustic-context gating rather than isolated speaker identity.
[NLP-287] Activation Flow: Manufacturing Activations for Steering
【速读】： 该论文旨在解决在模型存在“沙袋机制”（sandbagging）时，如何在不进行微调（fine-tuning）的情况下实现对模型行为的有效引导问题。沙袋机制通过故意降低模型性能来隐藏其真实能力，使得传统的差异均值引导（difference-in-means steering）方法失效，因其依赖于模型在理想行为下的激活状态。本文提出了一种名为激活流（Activation Flow, ActFlow）的新方法，其核心在于无需微调即可生成目标激活状态：通过设定目标logits，使正确答案在每个标注样本中排名第一，并通过向指定层的k个残差流（residual streams）同时添加一个统一向量x，将logits逐步推向目标。ActFlow构建了一类常微分方程（ordinary differential equations）来定义x的演化路径，每条规则对应一个从所需logit变化到x速度的映射；其中最小范数规则可精确抵达目标，其余规则仅保留雅可比矩阵（Jacobian）的前若干主方向。实验表明，在三类受锁模型（包括沙袋提示锁和密码锁LoRA）上，当k=40时，保留五个主奇异方向的ActFlow将平均未见ARC-Easy准确率从0.05提升至0.85，接近微调（0.88）和诚实模型（0.92）的表现。此外，其在18种组合中的16种优于最小范数规则，且其引导方向与真实差异均值方向近似正交，展现出更强的泛化性和鲁棒性，甚至成功解锁了两个传统方法失效的LoRA锁。因此，该方案的关键创新在于利用基于奇异值分解的低秩动态路径，高效生成高质量激活状态，从而绕过沙袋机制并实现精准行为控制。
链接: https://arxiv.org/abs/2609.32530
作者: Hong Kiat Tan,Linh Le,David Williams-King
机构: University of California, Los Angeles(加州大学洛杉矶分校); Lida Safety; ERA
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Differential Geometry (math.DG)
备注: 20 pages. Code at this https URL
Abstract:Difference-in-means steering requires activations recorded while a model shows the desired behavior, which a sandbagging model withholds by deliberately underperforming. We introduce Activation Flow (ActFlow), which manufactures these activations from k correct labels without fine-tuning. ActFlow sets target logits that rank each labeled item’s correct answer first, and moves the logits toward them by adding one vector x to all k residual streams at one layer. ActFlow is a family of ordinary differential equations for x , one for each rule that maps the required logit change to the velocity of x . The smallest-norm rule lands exactly on the targets, while the others keep only the top singular directions of the Jacobian. We test ActFlow on three instruction-tuned models, each locked by a sandbagging prompt and by a password-locked LoRA. At k=40 , ActFlow keeping five singular directions raises the mean held-out ARC-Easy accuracy over the six locked models from 0.05 to 0.85 , against 0.88 for fine-tuning and 0.92 for the honest models. Furthermore, it scores higher than the smallest-norm rule in 16 of the 18 combinations of locked model and k , and its steering direction is nearly orthogonal to the honest difference-in-means direction. It also unlocks two LoRA locks where the honest direction fails.
[NLP-288] When Users Change Their Minds: Measuring and Repairing Intent Drift in LLM Agents
【速读】： 该论文旨在解决大语言模型（LLM）代理在多轮交互中因用户意图漂移（intent drift）导致的性能下降问题。意图漂移是指用户在对话过程中更新或撤销先前表达的需求，而模型仍受过时意图影响，从而生成不准确的答案或执行错误的工具操作。为系统评估该问题，研究提出IntentFlux——一个可执行的基准测试框架，将可验证任务转化为包含受控意图变更的对话，同时保留原始评分机制。在627个案例的校准实验中，随着对话中过时信息增多，任务平均得分从0.476降至0.384，表明意图漂移显著损害模型表现。跨八种模型的对比显示，从动态演变的对话中恢复最终任务的正确解答应答率远低于直接给出单轮任务的情况。为此，研究进一步提出StateForge，一种显式维护当前有效需求状态的机制，在General-Test上将平均任务得分从0.367提升至0.467。尽管提供真实最终状态可进一步提升性能，但仍未达到单轮输入下的表现水平，说明状态估计误差仅部分解释了性能差距。研究结果确立了意图漂移作为可度量的多轮交互失败模式，并验证了显式状态维护是其部分有效的缓解策略。
链接: https://arxiv.org/abs/2609.32520
作者: Yanjie Zhang,Bowen Cao,Zixin Chen,Yushi Sun
机构: HKUST(香港科技大学); CUHK(香港中文大学); LIGHTSPEED(腾讯光速)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 15 pages, 3 figures
Abstract:LLM agents often operate over multi-turn interactions in which user intent changes before execution. We study intent drift: the failure mode in which superseded parts of the user’s intent continue to influence the final answer or tool action. We introduce IntentFlux, an executable benchmark that converts verifiable tasks into dialogues with controlled intent changes while preserving their original graders. In a 627-case calibration, mean task score falls from 0.476 to 0.384 as dialogues contain more superseded and withdrawn information. Across eight models, the rate of fully correct solutions is significantly lower when the same final task must be recovered from an evolving dialogue rather than given directly in a single turn. We further introduce StateForge, which explicitly maintains the active requirements before generation. On General-Test, it improves mean task score from 0.367 to 0.467. Providing the ground-truth final state improves performance further but still does not recover single-turn performance, indicating that state-estimation errors explain only part of the gap. These results establish intent drift as a measurable multi-turn failure mode and explicit state maintenance as a partial mitigation.
[NLP-289] Learning an Anchored Prompt Space for Continual Adaptation of Large Language Models
【速读】： 该论文旨在解决大语言模型在持续学习过程中如何有效获取新知识的同时保持已有能力的问题。其核心挑战在于：传统联合优化模型参数与任务特定软提示（soft prompts）的方法存在两个关键局限——随模型演进而更新的历史软提示可能逐渐失效，且不同任务间的软提示之间潜在的可迁移关系未被显式建模。为此，论文提出了一种名为“锚定提示空间学习”（Learning an Anchored Prompt Space, LAPS）的新方法。其关键创新在于：首先通过自蒸馏（self-distillation）机制将历史软提示与更新后的模型对齐，以维持其有效性；随后构建一个以已学习的任务特定软提示为顶点、由可学习的贝塞尔（Bézier）控制提示塑造中间几何结构的锚定提示空间。该空间不仅保留了历史提示的效用，还显式建模了跨任务提示间的关联性，从而支持正向迁移。实验基于TRACE基准在三个规模的Qwen3模型上验证表明，LAPS显著优于基于蒸馏、纯提示或参数-提示联合优化的基线方法，在提升平均性能的同时有效缓解了遗忘问题。
链接: https://arxiv.org/abs/2609.32499
作者: Rongguang Ye,Zhan Zhuang,Yichen Wu,Ming Tang,Kede Ma
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Continually adapting large language models requires acquiring new knowledge while preserving previously learned capabilities. Jointly adapting model parameters and task-specific soft prompts offers a promising solution, but faces two key limitations: historical prompts may become less effective as the model evolves, while their transferable cross-task relationships are not explicitly learned. We propose Learning an Anchored Prompt Space (LAPS), which preserves historical prompt effectiveness and learns relationships among task-specific soft prompts to facilitate positive transfer. LAPS first aligns historical prompts with the updated model through self-distillation. LAPS then constructs an anchored prompt space whose vertices correspond to learned task-specific soft prompts and whose intermediate geometry is shaped by learnable Bézier control prompts. Once this anchored prompt space is learned, LAPS identifies the best-performing prompt for each task on its validation set, allowing the optimized prompt to draw on knowledge acquired from observed tasks. Experiments on the TRACE benchmark across three Qwen3 model scales show that LAPS consistently outperforms distillation-based, prompt-based, and joint prompt–parameter adaptation baselines, improving average performance while reducing forgetting.
[NLP-290] Locally Sound Globally Insufficient: The Local-Global Gap in Multi-Hop Reasoning ICLR2027
【速读】： 该论文旨在解决多跳问答（multi-hop QA）中普遍存在的“局部-全局差距”（local-global gap, LGG）问题，即推理链在每一步都具备局部证据支持（local soundness），但整体推理轨迹仍无法有效回答问题（global insufficiency）。LGG的本质在于：尽管每个推理步骤均符合证据与前序步骤的逻辑支持，但整个推理过程缺乏与问题的语义对齐（question-to-trace alignment）以及最终答案的闭合性（trace-to-answer closure）。在对3个基准数据集和3种模型共2,598个响应的人工评估中发现，LGG几乎存在于所有组合中，占全局不充分响应的近一半。传统忠实性验证方法仅关注证据-声明的一致性，难以捕捉此类全局失效。为此，论文提出从训练阶段就显式建模三种关键依赖关系：证据到推理步骤的支持、问题到推理轨迹的对齐、推理轨迹到答案的闭合。其核心解决方案E-Closure通过联合生成监督（包括原始与反事实响应）与双向切换约束（bidirectional switching constraints），在训练过程中直接优化上述依赖关系。实验表明，相比现有微调基线，E-Closure在三个基准与三种主干模型上实现了最高平均准确率（92.8%）与推理可靠性（89.0%），同时将LGG率降至最低（6.2%），显著提升了多跳推理的整体质量。
链接: https://arxiv.org/abs/2609.32496
作者: Bohao Chu,Hendrik Damm,Qianli Wang,Hui Wang,Shuning Zhang,Christoph M. Friedrich,Norbert Fuhr
机构: University of Duisburg-Essen(杜伊斯堡-埃森大学); University of Applied Sciences and Arts Dortmund(多特蒙德应用技术与艺术大学); Institute for Medical Informatics, Biometry and Epidemiology (IMIBE)(医学信息学、生物统计学与流行病学研究所); Technische Universität Berlin(柏林工业大学); Tsinghua University(清华大学)
类目: Computation and Language (cs.CL)
备注: Submitted to ICLR 2027
Abstract:Reliable multi-hop reasoning requires more than locally supported steps: a trace can be sound at every reasoning step yet still fail to answer the question as a whole. We call this failure regime the local-global gap (LGG), in which the trace is locally sound yet globally insufficient. Local soundness requires each step to be supported by the available evidence and preceding steps, whereas global sufficiency requires the reasoning trace to align with the question and establish the submitted answer. In a human-adjudicated diagnostic of 2,598 responses across three multi-hop QA benchmarks and three models, we find that the LGG occurs in every benchmark-model combination and accounts for nearly half of globally insufficient responses overall. However, conventional faithfulness verifiers that check claims against the evidence largely miss these failures: at thresholds retaining at least 95% of reliable traces, recall for LGG cases is substantially lower than that for locally unsound traces. To address these failures, we formalize three dependencies for reliable reasoning: evidence-to-step support, question-to-trace alignment, and trace-to-answer closure. Instead of post-hoc diagnosis, we introduce E-Closure to supervise these dependencies during training, combining generation supervision on supported original and counterfactual responses with bidirectional switching constraints. Averaged over three benchmarks and three backbones, existing fine-tuning baselines improve accuracy and local soundness over the base models, but at the cost of global sufficiency. E-Closure improves both: among all fine-tuned methods, it achieves the highest average accuracy (92.8%) and trace reliability (89.0%) while yielding the lowest LGG rate (6.2%).
[NLP-291] Explaining Textual Entailment with Lexical Entailments: Using LLM s to Supply Lexical Relations for Formal Proofs
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在自然语言推理（Natural Language Inference, NLI）任务中是否能够准确识别并有效利用所需的词汇知识，以及其生成的词汇关系是否具备普遍有效性的问题。尽管LLMs展现出强大的自然语言推理能力并可能内化大量词汇知识，但其实际使用这些知识的方式及其质量仍不明确。为解决此问题，研究提出了一项新任务：通过结构化词汇蕴含（structured lexical entailments，如 chinchilla ⊑ small animal）作为句法蕴含的结构化解释代理，构建一个用于评估LLMs生成结构化词汇解释能力的数据集，并在此基础上进行内在评估（intrinsic evaluation）。同时，研究采用神经符号框架下的外在评估（extrinsic evaluation），将LLMs生成的词汇关系供给自然逻辑定理证明器LangPro，以检验其对定理证明过程的贡献。关键发现表明，即使对于主流的私有化大型语言模型，该任务依然具有挑战性；且其生成的词汇关系虽能部分支持推理，但多数仅在特定上下文中成立，缺乏普遍有效性，表明当前LLMs生成的词汇知识在形式化推理中的应用仍存在局限性。
链接: https://arxiv.org/abs/2609.32491
作者: Jorryt de Jong,Stefan Moraca,Ettore Cesari,Lasha Abzianidze
机构: Utrecht University (乌得勒支大学); Institute for Language Sciences, Utrecht University (语言科学研究所，乌得勒支大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted at the 11th Workshop on Automated Knowledge Base Construction (AKBC 2026)
Abstract:Large Language Models (LLMs) are highly capable of natural language reasoning and appear to store a great deal of lexical knowledge, but it is still unclear how much of this knowledge they actually use when reasoning, and whether they use it in the right way. On the other hand, logic-based Natural Language Inference (NLI) systems provide transparent and formally grounded reasoning, but they need to be supplied with rich lexical knowledge to prove inferences beyond purely logical ones. In this paper, we evaluate whether LLMs can identify all lexical knowledge needed to solve NLI problems and how much this knowledge contributes to proof search in a logic-based NLI system. Our research focuses exclusively on structured lexical entailments (e.g., chinchilla \sqsubseteq small animal) as a proxy for structured explanations for NLI problems with an entailment label. First, we curate a dataset for a new task of explaining sentential entailments with a set of lexical entailments. The dataset is used to intrinsically evaluate LLMs on generating structured lexical explanations. Then, we use NLI as an extrinsic evaluation in a simple neuro-symbolic setting, assessing whether LLMs can supply sufficient lexical relations to LangPro, a natural-logic theorem prover for natural language. The results show that the proposed task remains challenging even for hosted proprietary LLMs, and that their contribution to theorem proving is moderate: generated relations are often only partially sound and may be tailored to the specific NLI problem rather than representing generally valid lexical knowledge.
[NLP-292] PC-SubMax: Efficient Prompt Compression via Regularized Submodular Maximization
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在长上下文场景中因提示（prompt）过长导致的推理成本高、延迟增加以及“中间信息丢失”（lost-in-the-middle）现象严重的问题。现有基于固定词元或句子级重要性评分的提示压缩方法难以捕捉内容贡献随选择子集变化的动态特性，且易忽略句间冗余；而依赖自回归式大语言模型打分的压缩方法则引入显著计算开销。为此，本文提出PC-SubMax框架，将选择性提示压缩建模为带有背包约束的正则化单调亚模最大化问题，其目标函数为 $ U(S) - \ell(S) $，其中单调亚模效用函数 $ U(S) $ 融合了信息覆盖度、查询相关性与对数行列式多样性以衡量内容价值，非负模块化惩罚项 $ \ell(S) $ 则表征词元成本。通过边际收益递减机制，该目标能够评估每个句子相对于已选内容的实际贡献。为优化此目标，作者设计了正则化贪心+最大（Regularized Greedy+Max, RGM）算法，可确定性地生成一个可行解 $ Q $，满足 $ U(Q) - \ell(Q) \geq \frac{1}{2}U(O) - \ell(O) $，其中 $ O $ 为最优解，且仅需 $ O(n\kappa) $ 次值查询（value-oracle queries），计算高效。PC-SubMax采用编码器表示进行评估，避免了压缩过程中的自回归打分，显著降低计算开销。在七个多样化基准上的实验表明，该方法在保持下游任务性能的同时实现了极低的压缩开销，展现出良好的实用性和理论保障。
链接: https://arxiv.org/abs/2609.32474
作者: Ziyi Zhang,Shuang Cui,Haotian Zhang,Xiaoyu Wang
机构: Soochow University (苏州大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:While large language models (LLMs) are increasingly deployed in long-context scenarios, lengthy prompts can increase inference costs and latency and exacerbate the ``lost-in-the-middle’’ phenomenon. Selective prompt compression offers a model-agnostic approach to alleviating these issues. However, methods based on fixed token- or sentence-level importance scores may overlook how content contributions change with the selected subset, limiting their ability to account for inter-sentence redundancy. Compression procedures that rely on autoregressive LLM scoring can also introduce substantial overhead. We propose PC-SubMax, a theoretically grounded framework that formulates selective prompt compression as regularized monotone submodular maximization under a knapsack constraint. The objective is U(S)-\ell(S) , where the monotone submodular utility U combines information coverage, query relevance, and log-determinant diversity, and the non-negative modular penalty \ell captures token cost. Through diminishing marginal returns, the objective evaluates each sentence’s contribution relative to the selected content. To optimize this objective, we develop the Regularized Greedy+Max (RGM) algorithm, which deterministically returns a feasible set Q satisfying U(Q)-\ell(Q)\geq \frac12U(O)-\ell(O) , where O is an optimal feasible solution to the regularized problem. RGM uses O(n\kappa) value-oracle queries, where n is the number of candidate sentences and \kappa is the maximum feasible subset size. PC-SubMax uses encoder representations and avoids autoregressive LLM scoring during compression. Experiments across seven diverse benchmarks demonstrate competitive downstream performance with low compression overhead.
[NLP-293] AdaTutoRank: Learning to Rerank Document Sets via Adaptive Tutoring Optimization for RAG and Deep Research
【速读】： 该论文旨在解决现有检索重排器（rerankers）在复杂信息需求下难以生成完整、互补且无冗余的文档集合的问题。主流方法依赖相关性匹配进行文档排序，但单个相关文档往往无法构成满足复杂查询所需的完备集合，且现有基于集合整体评分的优化目标存在监督信号稀疏的问题：无论文档是否具有实际贡献，整个集合的高分都会对所有成员产生同等奖励，导致关键贡献者与冗余项无法区分。为克服这一问题，本文提出AdaTutoRank，一种基于自适应导师优化（Adaptive Tutoring Optimization, ATO）的集合级重排器，其在九维评分维度构成的三层层次结构下训练，能够提供冷启动银标签、强化学习奖励及知识蒸馏提示。ATO从策略自身的冻结快照中动态提取三种逐步细化的提示形式——仅基于评分、基于评分选择的同辈集合（self-selector的兄弟集），以及对比当前回放与兄弟集的自我反思（self-reflector的反思）——并根据每个回放的质量匹配最适提示。通过在提示条件下的冻结教师模型重新评分与无提示快照的对比，将提示效应提炼为细粒度的令牌级优势，与群体相对结果优势相补充。实验表明，AdaTutoRank在涵盖RAG、深度研究和集合评估的十项基准上均取得最优综合性能，同时显著减少检索调用次数。
链接: https://arxiv.org/abs/2609.32472
作者: Kailin Jiang,Lei Liu,Jian Xi,Yangqi Chen,Hui Xu,Hongwei Zhao,Bin Li,Yu Lu,Haibo Shi
机构: University of Science and Technology of China(中国科学技术大学); Yuanbao Team, Tencent(腾讯元包团队)
类目: Computation and Language (cs.CL)
备注: Project Page: this https URL
Abstract:Document rerankers determine what evidence reaches the downstream model in RAG and deep research, yet mainstream rerankers select by relevance matching, and individually relevant documents rarely constitute the complete, complementary, non-redundant set a complex information need demands. Prior work rewards a set by its aggregate rubric score, shifting the objective from ranking documents to composing sets. Yet that score is one scalar shared by every document in the set, so the supervision is sparse: a redundant document is rewarded with the rest whenever the set scores well, and a decisive one penalized with the rest whenever it does not; credit assignment leaves contributors indistinguishable from free riders. On-policy distillation could densify this supervision, but existing methods give every rollout the same fixed guidance, too prescriptive for strong rollouts and too abstract for weak ones. We therefore propose AdaTutoRank, a setwise reranker trained with Adaptive Tutoring Optimization (ATO) under a three-level hierarchy of nine rubric dimensions, which supplies silver labels for the cold start, rewards for reinforcement learning, and hints for distillation. ATO draws three hint forms of increasing specificity from the policy’s own frozen snapshot: the rubrics alone, a self-selector’s sibling-set chosen under rubrics, and a self-reflector’s reflection contrasting the rollout with that sibling-set; each rollout receives the form matched to its quality. Re-scoring that rollout under the hint-conditioned frozen teacher and the hint-free snapshot distills the hint’s effect into a token-level advantage that complements the group-relative outcome advantage. Across ten benchmarks spanning RAG, deep research, and setwise evaluation, AdaTutoRank attains the best overall performance while issuing fewer retrieval calls.
[NLP-294] PULSE: Identifying Demonstration-Utility Features with Sparse Autoencoders NEURIPS2026
【速读】： 该论文旨在解决上下文学习（in-context learning）中演示样本选择对模型性能敏感的问题，尤其是现有方法依赖外部查询-演示相似性度量时，可能忽略模型内部的特异性信号——即看似相似的演示样本可能激活不同的内部特征并引发不同的下游行为。其解决方案的关键在于提出一种基于稀疏编码（Sparse Encodings）的自编码器（SAE）框架——PULSE（Paired Utility Localization over Sparse Encodings），通过识别与演示样本效用相关的模型内部特征，并利用这些特征指导更优的演示选择。具体而言，PULSE利用少量标注的发现集采样候选演示集，在目标模型上评估其零样本相对效用，通过分析激活差异与效用差异的一致性来评分SAE特征，筛选出最具代表性的正负方向坐标构成稀疏的效用定位向量。该向量被用于两种互补策略：一是作为带符号得分进行受控全集排序以验证其预测能力；二是作为PULSE-Retriever，将向量幅度转化为特征相关性掩码，实现可扩展的池规模检索。实验表明，PULSE-Retriever在分类、生成和推理任务上分别提升2-3个准确率点、0.6-0.9个BLEU-4分数和3.2个精确匹配点，且特征分析显示所识别的特征捕捉了任务相关且数据集特定的模式，同时具备一定的跨数据集泛化能力。
链接: https://arxiv.org/abs/2609.32469
作者: Chenduo Hao,Chuanbao Gao,Pinjun Zeng,Jingze Zhu,Chonghan Liu,Zidong Liu,Xu Yang
机构: Southeast University (东南大学); University of California, Los Angeles (加州大学洛杉矶分校); School of Architecture, The University of Texas at Austin (德克萨斯大学奥斯汀分校)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026. 24 pages
Abstract:In-context learning is highly sensitive to demonstration choice, yet most methods select demonstrations using external query-demonstration similarity. Such criteria can miss model-specific signals: Similar demonstrations may activate different internal features and downstream behaviors. We introduce PULSE (Paired Utility Localization over Sparse Encodings), an SAE-based framework for identifying model-internal features associated with demonstration utility and using them for demonstration selection. Using a small labeled discovery set, PULSE samples candidate demonstration sets, measures their zero-shot-relative utility under the target model, and scores SAE features by how their activation differences align with utility differences. The top positive and negative coordinates form a sparse utility-localization vector. We use this vector in two complementary ways: as a signed score for controlled complete-set ranking, and as PULSE-Retriever, which converts its magnitude into a feature-relevance mask for scalable pool-scale retrieval. Across classification, generation, and reasoning benchmarks, PULSE-Retriever improves over the strongest baseline by 2-3 accuracy points, 0.6-0.9 BLEU-4, and 3.2 exact-match points, respectively, while controlled ranking validates the identified features encode a predictive set-level utility signal. Feature inspection and cross-dataset experiments suggest that the identified features capture task-relevant, dataset-conditioned patterns, yet retain utility signals that partially transfer across datasets. Our code is available at this https URL.
[NLP-295] Streamlined Reflective Evolution for Task-Adaptive Self-Refinement Pipelines
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在无需更新模型权重的前提下，如何实现高效、自适应的自我反思优化（self-refinement）问题。现有方法受限于固定架构，难以灵活组织反思流程，导致优化效率低下。其核心解决方案是提出工作流设计代理（Workflow-Designing Agents, WDA）框架，通过联合演化任务阶段指令及其顺序结构，实现任务自适应的自我反思流水线。关键创新在于引入SPLIT机制，将因重复修订而累积在单一提示中的冗余指令重新分配至专用阶段，从而提升流程清晰度与效率。同时，基于三例反思与局部筛选实现选择性搜索，结合校准分数指导帕累托最优准入与无效更新回滚。最终生成的流水线具备任务自适应特性：指令内容与深度由任务数据学习并固化，适用于同一任务下的所有测试输入。在涵盖知识推理、数学推理、多跳问答及指令遵循的五个基准上评估表明，WDA显著优于初始求解器及无SPLIT的变体，在Qwen3.5-9B和GPT-4.1-mini上分别取得平均51.24%和49.00%的准确率，相对提升达8.63和5.62个百分点，验证了任务自适应自我反思作为更广泛智能体工作流搜索的互补路径的有效性。
链接: https://arxiv.org/abs/2609.32458
作者: Xiaofan Zhou,Lu Cheng
机构: Pennsylvania State University (宾夕法尼亚州立大学)
类目: Computation and Language (cs.CL)
备注: 59 pages, 2 figures
Abstract:Reflective prompt optimization improves large language model (LLM) systems without updating model weights, but fixed architectures constrain how self-refinement is organized. We introduce Workflow-Designing Agents (WDA), a framework for streamlined reflective evolution of task-adaptive self-refinement pipelines. Starting from a minimal prompt, WDA jointly evolves stage instructions and their sequential structure. During evolution, we find that repeated revisions can accumulate redundant instructions in a single prompt. In WDA, we propose to address this problem with SPLIT, which redistributes these instructions across specialized stages. Three-example reflection and local screening guide selective search, while calibration scores guide Pareto admission and rollback of unhelpful trailing updates. The resulting pipelines are task-adaptive: their instructions and depth are learned from task data, then fixed for all test inputs within that task. We evaluate WDA on five benchmarks spanning knowledge, mathematical reasoning, multi-hop question answering, and instruction following. On Qwen3.5-9B, WDA achieves an average score of 51.24%, improving over the initial solver by 8.63 percentage points and the variant without SPLIT by 2.60 points. On GPT-4.1-mini, it achieves 49.00%, with corresponding gains of 5.62 and 3.69 points. These results support task-adaptive self-refinement as a complementary direction to broader agentic workflow search.
[NLP-296] What Should the Reflector See? An Empirical Study of Evidence in Reflective Prompt Optimization
【速读】： 该论文旨在解决生成式 AI（Generative AI）中反思提示优化（Reflective Prompt Optimization）的内在不确定性问题，即反思器应接收何种证据、示例可见性如何、候选样本选择策略以及领域知识策略的设计对最终性能的影响尚不明确。其解决方案的关键在于构建一个基于单父代帕累托引导搜索（single-parent Pareto-guided search）的统一评估框架，系统地考察证据构成、示例可见性、候选选择及领域知识策略等关键因素。研究以 Qwen3.5-9B 作为任务模型与反思器，在五个数据集上评估九种不同的反思策略，结果揭示出三类核心模式：在性能提升方面，“仅失败案例”（Failures-only）策略带来最大均值测试增益（+8.0个百分点），而“平衡混合”（Balanced-mix）与“无示例+验证集”（No-examples+Val）并列最佳平均性能排名；在反思有效性方面，“无示例”（No-examples）策略在提升采样父代表现上排名最优，但仅实现1.4个百分点的均值增益，表明局部反思成功并不必然导致最终提示更强；在过拟合评估方面，“仅失败案例”与“平衡混合”策略表现出最低的均值校准差距排名，但在 GPQA 与 IFBench 上仍存在较大校准差距，说明校准收益可能夸大真实泛化提升，该差距仅为描述性指标而非直接过拟合度量。综上，研究强调反思策略需分别从最终性能、父代改进能力及校准到测试集迁移能力三个维度进行独立评估。
链接: https://arxiv.org/abs/2609.32452
作者: Xiaofan Zhou,Lu Cheng
机构: Pennsylvania State University (宾夕法尼亚州立大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 32 pages, 2 figures, 5 tables
Abstract:Reflective prompt optimization revises instructions using examples of a model’s behavior, but which evidence the reflector should receive remains unclear. We study evidence composition, visibility of examples, candidate selection and domain-knowledge policy within a single-parent Pareto-guided search. Using Qwen3.5-9B as both task model and reflector, we evaluate nine reflection strategies on five datasets. From these experiments, we find three distinct patterns. For performance improvement, Failures-only produces the largest mean test gain (+8.0 percentage points), while Balanced-mix and No-examples+Val share the best mean performance rank. For reflection effectiveness, No-examples achieves the best rank for improving sampled parents, yet yields only a 1.4-point mean test gain: local reflection success does not necessarily produce a stronger final prompt. For overfitting assessment, Failures-only and Balanced-mix share the lowest mean calibration-gap rank, while larger gaps on GPQA and IFBench show that calibration gains can overstate held-out improvement. This gap is a descriptive indicator, not a direct measure of overfitting. Together, these results show why reflection strategies should be assessed separately on final performance, parent improvement and calibration-to-test transfer.
[NLP-297] Self-Reports Do Not Identify Self-Models: An Identifiability Test for Counterfactual Reports
【速读】： 该论文旨在解决当前语言模型自报告（self-report）机制中存在的可信度问题，即模型在特定提示环境下的行为表现是否真实反映了其内部的自我建模能力（self-model），而非仅是对外部干预信号的被动响应。研究发现，当对模型施加激活干预（activation interventions）并要求其报告类似情感状态时，若演示环境发生变化，模型的自报告仍会受到错误来源（wrong-source）演示的影响，表现出对干预源的依赖性。这一现象表明，当前的自报告结果可能并非源于自主的内在认知机制，而是受外部环境线索驱动。解决方案的关键在于引入“机制绑定”（mechanism binding）策略，通过显式地将干预与机制关联，显著降低环境变化带来的偏差影响。研究强调，未来的自报告基准测试应包含固定干预下的环境迁移不变性检验，以确保报告准确性可作为自主报告机制的可靠证据。
链接: https://arxiv.org/abs/2609.32449
作者: Phongsakon Mark Konrad,Toygar Tanyel,Serkan Ayvaz
机构: Centre for Industrial Software, University of Southern Denmark, Sønderborg, Denmark(南丹麦大学工业软件中心); Promake, Newark, DE, USA
类目: Computation and Language (cs.CL)
备注:
Abstract:Language-model self-reports are evidence about behavior in a prompt environment, not by themselves evidence of a self-model. We investigate counterfactual reports about affect-like states under activation interventions and ask whether the report remains bound to the named intervention when the demonstration environment changes. Across three open instruction models, wrong-source demonstrations move reports toward the source answer family, while explicit mechanism binding reduces this pull. Self-report benchmarks should include environment-shift invariance tests under fixed intervention before treating accuracy as evidence for an autonomous report mechanism.
[NLP-298] ForkLeft: Entropy-First Rollouts for Prefix-Aligned Autoregressive-to-Diffusion Distillation
【速读】： 该论文旨在解决扩散语言模型（Diffusion Language Models, DLMs）在保持其原生并行生成能力的同时，如何通过知识蒸馏获得类似自回归下一词预测（Next-Token Prediction, NTP）的推理能力这一核心问题。其解决方案的关键在于提出一种名为ForkLeft的蒸馏框架，该框架通过分离学生模型的生成过程与教师监督信号，有效缓解了自回归教师与双向条件学生的本质不匹配问题：传统蒸馏中，教师基于左侧前缀进行预测，而DLM可同时利用左右两侧上下文。ForkLeft在训练阶段首先引导学生执行熵优先的滚动生成（entropy-first rollout），以确定高不确定性位置并暴露潜在的生成分支（forks），随后固定生成前缀，在相同上下文中对齐自回归教师的输出，由答案正确性决定监督来源；推理时则恢复学生模型原有的置信度优先并行解码机制。实验表明，基于Qwen3-30B-A3B-Base蒸馏的ForkLeft显著提升Efficient-DLM-4B在10个基准测试上的表现，数学题集MATH500准确率从72.60%提升至79.60%，且优于三种替代设计；性能增益随教师模型强度增强，并在仅500次更新下成功迁移至SDAR-4B。在相同规模下，蒸馏后的4B和8B学生模型在7个基准上超越已发表的SDAR-Chat与OPDLM模型，证明了DLM可在不牺牲并行生成优势的前提下，成功习得NTP式推理能力。
链接: https://arxiv.org/abs/2609.32448
作者: Junming Liu,Jicheng Wang,Yifeng He,Hao Chen,Jianzhong Qi
机构: University of Melbourne(墨尔本大学); University of California, Davis(加州大学戴维斯分校); University of Hong Kong(香港大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 23 pages, 6 figures, 7 tables
Abstract:Autoregressive Next-Token Prediction (NTP) has enabled strong reasoning capabilities in language models, while Diffusion Language Models (DLMs) offer flexible token orders and parallel generation. We ask whether DLMs can acquire NTP-style reasoning through distillation without giving up their native generation process. Direct distillation, however, faces a fundamental mismatch: an autoregressive teacher predicts from a left prefix, whereas a DLM can condition on tokens on both sides. We introduce ForkLeft, a distillation framework that resolves this mismatch by separating the student’s rollout from teacher supervision. During training, the student first performs entropy-first rollouts that commit uncertain positions and expose potential forks. We then fix the resulting student prefix and distill an NTP teacher under the same context, with answer correctness determining the supervision source. At inference, the student returns to its native confidence-first parallel decoding. With Qwen3-30B-A3B-Base, ForkLeft improves Efficient-DLM-4B on all ten benchmarks, raising MATH500 from 72.60% to 79.60% and consistently outperforming three alternative designs. The gains scale with teacher strength and generalize to SDAR-4B with only 500 updates. At matched scale, the distilled 4B and 8B students exceed the published SDAR-Chat and OPDLM models on seven benchmarks, showing that DLMs can learn NTP-style reasoning without sacrificing native parallel generation. Code and datasets will be released upon acceptance.
[NLP-299] Masking Frequent Tokens Sharpens Direct Preference Optimization
【速读】： 该论文旨在解决直接偏好优化（Direct Preference Optimization, DPO）在语言模型对齐过程中存在的固有缺陷：由于高频率词元类型在偏好响应对中呈现对称性分布，导致其在序列级累计得分中占据主导地位，从而引发梯度纠缠并稀释了可区分的偏好信号。其核心问题是，尽管这些高频词元在优选与非优选响应中出现频率相当，却因权重过高而干扰了模型对真正偏好差异的学习。为解决此问题，论文提出各向异性DPO（Anisotropic DPO, ADPO）及其具体实现——频率硬掩码DPO（Frequency-Hard DPO），其关键在于采用一个固定且不依赖标签的词元词汇掩码，强制将高频率响应词元的隐式奖励贡献置零，同时为信息量高的位置赋予单位权重，从而在不修改偏好对、不丢弃上下文或引入可学习参数的前提下，有效抑制梯度干扰。该方法通过非均匀的词元层面目标加权（即“各向异性”设计）实现了更鲁棒的偏好对齐，实验表明其在AlpacaEval、MT-Bench和Arena-Hard等多个基准上均显著优于标准DPO，验证了选择性屏蔽共享高频词元是一种高效且零开销的改进机制。
链接: https://arxiv.org/abs/2609.32445
作者: Harshvardhan Saini,Samyak Jha,Yiming Tang,Dianbo Liu
机构: Indian Institute of Technology Dhanbad (印度理工学院达纳巴德分校); National University of Singapore (新加坡国立大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Direct Preference Optimization (DPO) aligns language models by optimizing over sequence-level sums of token-wise implicit reward differences. However, we identify a pervasive pathology in this formulation: a disproportionately small subset of high-frequency token types dominates cumulative sequence scores while appearing symmetrically across both preferred and dispreferred responses. Specifically, under canonical Qwen tokenization on Anthropic HH-RLHF, merely 69 token types account for 55.1% of all response tokens and 85.9% of within-pair shared token mass, exhibiting substantially lower preference-side specificity than the remaining vocabulary. This symmetric ubiquity induces gradient entanglement and dilutes the discriminative preference signal propagated through the objective. To resolve this issue, we introduce \emphAnisotropic DPO (\textsfADPO) and its canonical realization, \emphFrequency-Hard DPO. Using a fixed, label-agnostic vocabulary mask, our method zeroes the implicit reward contribution of high-frequency response tokens while assigning unit weight to informative positions, thereby suppressing gradient interference without modifying preference pairs, discarding context, or introducing learned parameters. Here, \emphanisotropy designates non-uniform token-level objective weighting rather than representational geometry. Extensive empirical evaluations on AlpacaEval, MT-Bench, and Arena-Hard demonstrate that Frequency-Hard DPO consistently outperforms standard DPO across Qwen-2.5-7B-Instruct and Llama-3-8B-Instruct, establishing that selectively masking shared high-frequency tokens offers an effective, zero-overhead mechanism for robust preference alignment.
[NLP-300] DualGuard: Dual-Mode Quality Control for Logic-Preserving Data Augmentation
【速读】： 该论文旨在解决大规模生成式逻辑推理数据中存在语义、标签及逻辑可靠性不足的问题，尤其针对生成数据质量判断后如何有效决策保留、过滤或修复候选样本这一关键控制难题。其核心解决方案是提出DualGuard——一种双模式的质量控制框架，通过双重机制实现逻辑保持的数据增强：第一模式基于当前实例与候选批次进行选择性保留、过滤、归因与定向反馈；第二模式则在样本级诊断基础上，累积跨实例的增强操作执行记录，通过对比新执行行为与历史行为差异，支持回溯异常检测、精准回滚与有界修复。两模式共享语义验证，并在具备可靠逻辑形式时引入符号验证，显著提升生成数据的可信度。实验表明，在七项下游任务的两阶段迁移设置下，DualGuard在五项任务上取得最高准确率，且全面优于无数据增强的BERT基线，消融实验进一步验证了记忆模块、Z3求解器与历史感知异常控制之间的互补作用。
链接: https://arxiv.org/abs/2609.32431
作者: Shenghao Li,Lin Zhao
机构: Guangzhou University (广州大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large language models provide a practical way to generate augmented data for logical reasoning at scale, but a larger generation volume does not guarantee semantic, label, or logical reliability. Existing work has improved generation quality through generation constraints, candidate validation, filtering, and feedback-based revision; however, once a quality judgment is available, deciding whether a candidate should be retained, filtered, or repaired remains an important control problem. We propose DualGuard, a dual-mode quality-control framework for logic-preserving data augmentation. The first mode uses the current instance and candidate batch for selective retention, filtering, attribution, and targeted feedback. The second mode accumulates cross-instance execution records of augmentation actions on top of per-sample diagnosis and attribution, compares new executions against each action’s own historical behavior, and supports retrospective anomaly inspection, targeted rollback, and bounded repair. Both modes share semantic verification and additionally use symbolic verification when a reliable logical form is available. Across seven downstream tasks in the Two-Stage Transfer setting, DualGuard achieves the highest Accuracy on five tasks and outperforms the no-augmentation BERT baseline on all seven. Controlled ablations further show complementary roles for Memory, Z3, and history-aware anomaly control.
[NLP-301] Automatic Speech Recognition for the Basaà Language: A Low-Resource Approach
【速读】： 该论文旨在解决当前生成式AI在低资源语言（low-resource languages）中性能严重不足的问题，尤其是在自动语音识别（ASR）领域，尽管高资源语言如英语和法语已实现接近人类水平的识别精度，但多数语言仍面临标注数据匮乏、模型泛化能力差等挑战。其解决方案的关键在于利用自监督学习（self-supervised learning）与迁移学习相结合的方法，通过在大规模无标签语音数据上预训练通用语音表示，再针对低资源语言进行轻量级微调，从而有效提升模型在数据稀缺场景下的鲁棒性与可迁移性。
链接: https://arxiv.org/abs/2609.32408
作者: Sophie Gertrude Ngo Mock,Charles Moudina Varmantchaonala,Paul Dayang,Jean Michel Nlong II,Christopher Gies
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:The rapid advancement of Artificial Intelligence (AI) and Natural Language Processing (NLP) has revolutionized the way humans interact with machines. Among the most impactful developments is Automatic Speech Recognition (ASR), which enables computers to convert spoken language into text. Systems such as those built on deep neural networks, transformer architectures, and self-supervised learning have achieved near-human performance for well-resourced languages such as English and French. Yet, these advances have disproportionately benefited a small fraction of the world’s languages.
[NLP-302] Shared Worlds Private Minds: Structured Memory for Long-Form Writing as World Creation
【速读】： 该论文旨在解决大语言模型（LLM）在创作长篇小说时面临的连贯性难题，即如何在生成过程中保持故事世界（storyworld）的长期一致性。核心挑战在于：需有效管理异构的叙事信息、跨粒度整合情节发展，并恢复写作请求中隐含的依赖关系。其解决方案的关键在于提出NarraWorld——一种将记忆构建视为世界创设的结构化记忆系统。该系统基于共享的证据锚定图，衍生出四个相互关联的视图：世界事实、角色信念、待续发展与假设分支（可能世界的延续）。通过原子闭包驱动的层次聚合机制，事件被逐步整合为场景、剧情线与整体剧情，且每一高层节点均可追溯至原始文本片段。在检索阶段，计划重构策略根据当前叙事情境与记忆预览推断查询依赖，进而于有限的令牌预算内组装相关记录。实验表明，NarraWorld在三个写作基准上均取得最优综合表现，且其记忆能力可迁移至情境化角色扮演任务，并在通用长期记忆基准上保持较高召回率，为实现跨多样化叙事任务的连贯故事世界建模提供了可行路径。
链接: https://arxiv.org/abs/2609.32401
作者: Qiuyu Tian,Xiaowen Gu,Hang Su,Jianghan Chao,Haojie Yin,Fan Guo,Xin Zhang,Jinjing Shen,Ewing Luo,Youyong Kong,Yingce Xia,Zequn Liu
机构: Southeast University (东南大学); Beijing Zhongguancun Academy (北京中关村学院); Duke University; East China Normal University (华东师范大学); Gaoling School of Artificial Intelligence, Renmin University of China (中国人民大学高瓴人工智能学院); School of Animation and Digital Arts, Communication University of China (中国传媒大学动画与数字艺术学院); Jiangsu Second Normal University (江苏第二师范学院); ZhuiWen Technology Co., Ltd. (追文科技有限公司)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:LLM agents that write long-form fiction need an explicit memory of the evolving storyworld to keep new events consistent with established facts. Such memory must keep heterogeneous narrative information distinct, integrate story developments across granularities, and recover dependencies that a writing request leaves implicit. We present NarraWorld, a structured memory system for long-form writing that treats memory construction as world creation. From a shared evidence-grounded graph, NarraWorld derives four connected views: world facts, per-character beliefs, open developments, and hypothetical branches (possible-world continuations). Hierarchical aggregation with atomic closure consolidates events into scenes, plotlines, and plots, keeping each higher-level node traceable to its constituent source spans. For retrieval, planned reconstruction infers a query’s dependencies from the current narrative situation and a preview of memory, then assembles the relevant records within a token budget. Across three writing benchmarks, NarraWorld achieves the strongest aggregate results. Its memory also transfers to situated role-playing and largely preserves recall on a general-purpose long-term memory benchmark, paving the way for agents that sustain coherent storyworlds across diverse narrative tasks.
[NLP-303] SinBrief: A Hybrid Framework for Abstractive Text Summarisation of Sinhala Legal Documents
【速读】： 该论文旨在解决低资源语言（如僧伽罗语）法律文本摘要任务中因标注数据稀缺与领域专有术语复杂性带来的挑战。其核心解决方案在于提出一种无需人工标注训练数据的混合式抽象摘要框架——SinBrief，该框架通过构建领域感知的词图（domain-aware word graph）结合神经句子评分机制，实现对僧伽罗语法律文本的有效抽象摘要生成。关键创新点在于采用五种不同模型（mBert、Llama 3.1、Falcon 7B、Laser及领域自适应持续预训练的Llama模型）进行句子评分，并在无参考标准的评估下，利用覆盖度（Coverage）、密度（Density）、压缩比（Compression Ratio）、SummaC和Self-BertScore等指标验证性能。实验结果表明，SinBrief生成的摘要在词汇重叠度上显著低于抽取式基线模型的同时，仍保持良好的事实一致性，验证了该方法在低资源法律自然语言处理任务中实现高度自动化、近乎无标注的抽象摘要生成的可行性。
链接: https://arxiv.org/abs/2609.32397
作者: Minduli Lasandi,Nevidu Jayatilleke
机构: School of Computing, Informatics Institute of Technology (信息学院技术研究所), Sri Lanka; Department of Computer Science Engineering, University of Moratuwa (莫鲁塔瓦大学计算机科学与工程系), Sri Lanka
类目: Computation and Language (cs.CL)
备注: 13 pages, 5 figures, 3 tables, Accepted paper at the 13th Conference on Computational Linguistics and Speech Processing (ROCLING) 2026
Abstract:Legal document summarisation in low-resource languages presents significant challenges due to the scarcity of annotated data and the complexity of domain-specific terminology. This paper presents SinBrief, a hybrid abstractive summarisation framework for Sinhala legal documents that does not require human-annotated training data. The proposed framework combines domain-aware word graph construction with neural sentence scoring to generate abstractive summaries from Sinhala legal text. Five sentence scoring models are evaluated within the framework: mBert, Llama 3.1, Falcon 7B, Laser, and a continually pre-trained Llama model domain-adapted to Sinhala legal text. The framework is evaluated on a Sinhala legal corpus using reference-free metrics, including Coverage, Density, Compression Ratio, SummaC, and Self-BertScore. Experimental results demonstrate that SinBrief produces summaries with lower lexical overlap than extractive baselines while maintaining factual consistency, demonstrating the viability of hybrid, largely annotation-free abstractive summarisation for low-resource legal NLP tasks.
[NLP-304] FA-Bench: A Benchmark for Word-Level and Phone-Level Forced-Alignment and ASR Timestamps Under Clean and Noisy Conditions
【速读】： 该论文旨在解决语音强制对齐（forced alignment）评估中存在的基准不一致问题，即现有研究在文本归一化、数据划分、边界匹配方式等方面采用不同标准，导致结果难以横向比较。其解决方案的关键在于提出FA-Bench——一个开源统一框架，通过固定文本归一化规则、数据划分方式、音素映射及评分脚本，并公开所有代码与定期更新结果，实现评估流程的标准化。该框架设计了两个评测任务：Track 1使用参考文本作为对齐目标，Track 2则使用自动语音识别（ASR）系统输出作为输入，以模拟真实场景下的对齐性能。为更准确衡量对齐精度，采用基于容差的F1分数作为主评价指标，有效避免了传统平均绝对误差（MAE）在对话式语音中带来的9%至14%的分数虚高问题。此外，通过按边界位置及其相邻词识别正确性进行分组分析，揭示了当前系统在时间标注上的系统性偏差，例如Whisper模型平均提前约150毫秒，而多个商业ASR API则普遍延迟超过50毫秒。该研究显著提升了对齐评估的可比性与可靠性。
链接: https://arxiv.org/abs/2609.32396
作者: Wei Chu,Yuanzhe Dong,Ke Tan,Dong Han,Yichao Zhou,Ruchao Fan,Bingshen Mu,Jingbei Li,Vishwas Shetty,Sarthak Bisht,Ziyue Qiu,Massa Baali,Rita Singh,Bhisha Raj
机构: Google(谷歌); Meta(元); Stanford University (斯坦福大学); University of California, Berkeley (加州大学伯克利分校); Carnegie Mellon University (卡内基梅隆大学); MIT (麻省理工学院); Tsinghua University (清华大学); University of Illinois Urbana-Champaign (伊利诺伊大学香槟分校)
类目: Computation and Language (cs.CL); Performance (cs.PF)
备注:
Abstract:Forced alignment aligns speech audio with a text transcript to generate word and phone timestamps. Published comparisons normalize transcripts, split the data and match boundaries differently, so their numbers cannot be read together. We present FA-Bench, an open framework that fixes those choices once and releases the code, splits, phone mapping, text normalization and scoring script, with results published periodically. Track 1 gives every aligner the reference transcript and Track 2 gives it a recognizer’s output, on the same audio, clean and degraded four ways, with 21 open models and 9 commercial APIs under a unified protocol. We score every boundary of an utterance and check the two labels beside it, so a word the recognizer missed or invented is charged. Using a tolerance-based F1 as our primary metric eliminates the 9% to 14% score inflation that standard MAE causes on recognition-dependent systems in conversational speech. We then group boundaries by their position and how many adjacent words were recognized correctly, which shows where a system lost the score. We discovered systematic bias in how current systems time words, with Whisper about 150 ms early and several commercial ASR APIs over 50 ms late. Code and results are at this https URL
[NLP-305] PlurVA-LLM -2026 Shared Task Track-1: Pluralistic Value Alignment in LLM s via Multilingual Fine-Tuning and Threshold Calibration AACL
【速读】： 该论文旨在解决多文化语境下生成式AI（Generative AI）的价值对齐问题，特别是在中国、印度尼西亚和斯里兰卡三个具有显著文化差异的国家中实现更符合本地价值观的模型输出。针对资源受限的挑战，其解决方案的关键在于采用4比特量化低秩适配（4-bit QLoRA）技术对Llama 3.1 8B Instruct模型进行高效微调，以在有限计算资源下实现高性能。此外，针对不同语言数据的特点，分别设计了针对性的数据增强策略：对中国数据采用选项排列增强（option-permutation augmentation），对印尼语数据通过标注者投票扩展（annotator vote expansion）提升样本多样性，对僧伽罗语数据则结合二元重表述与SinhalaMMLU数据增强，并对斯里兰卡数据引入条件阈值校准（conditional threshold calibration）优化预测结果。这些方法协同作用，使系统在三大语种上的准确率分别达到0.785（中文）、0.715（印尼语）和0.916（僧伽罗语），整体宏平均准确率达0.805，有效提升了跨文化价值对齐的性能。
链接: https://arxiv.org/abs/2609.32382
作者: Vihindi Kotalawala,Nevidu Jayatilleke
机构: School of Computing, Informatics Institute of Technology, Sri Lanka(斯里兰卡信息学院技术学院计算机学院); Department of Computer Science Engineering, University of Moratuwa, Sri Lanka(斯里兰卡莫鲁图瓦大学计算机科学与工程系)
类目: Computation and Language (cs.CL)
备注: 11 pages, 7 figures, 6 tables, Accepted paper at the first workshop on Pluralistic Value Alignment of LLMs @ AACL-IJCNLP 2026
Abstract:We present our system for the PlurVA-LLM 2026 Shared Task Track-1, which focuses on pluralistic value alignment in the contexts of China, Indonesia, and Sri Lanka. For this resource-constrained track, we fine-tuned Llama 3.1 8B Instruct using 4-bit QLoRA. Our approach combines option-permutation augmentation for Chinese data, annotator vote expansion for Indonesian data, and binary reformulation with SinhalaMMLU augmentation for Sri Lankan data. We further applied conditional threshold calibration to the predictions for the Sri Lankan data. The final system achieved accuracies of 0.785 for Chinese, 0.715 for Indonesian, and 0.916 for Sri Lankan, resulting in an overall macro-average accuracy of 0.805.
[NLP-306] Black-Box Auditing of Epistemic Reliability in Multi-Agent Debate Distillation
【速读】： 该论文旨在解决生成式AI在多智能体辩论中通过辩论蒸馏（debate distillation）进行弱验证器适应时所面临的认知可靠性退化（epistemic reliability degradation）问题，即模型在监控任务上表现维持甚至提升，但在未监控的隐藏任务上对正确答案的支持能力却下降。其核心挑战在于：这种退化是否仅由灾难性遗忘（catastrophic forgetting）导致？标准评估方法能否有效检测此类退化？为此，论文提出ER-Audit——一种两阶段黑箱审计框架，通过对比适应前后的冻结验证器检查点，结合基于共享语境的监控与隐藏任务配对提示构建的两个评估基准，系统性地搜索非退化性的反例。该框架利用语义有效的改写句进行测试，若无反例则采用独立改写句进行序列假设检验，并推导出任意时间有效的非退化概率下界，支持在有限预算内动态停止。研究进一步建立了固定改写分布下的通用下界，并将其扩展至混合分布与其总变差距离有界的范围。实验表明，更高的隐藏任务准确率可能伴随更多反例和更低的非退化置信下界，揭示了仅依赖整体性能指标无法捕捉的选择性隐藏任务退化现象，从而挑战了以广泛遗忘为核心的解释，并证明审计机制可有效发现被聚合性能增益掩盖的隐蔽退化。
链接: https://arxiv.org/abs/2609.32361
作者: Derui Wang,Zewei Shi,Rayne Holland,Ruoxi Sun,Xingliang Yuan,Jason Xue,Liming Zhu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Cryptography and Security (cs.CR)
备注: Code and benchmarks are available at this https URL
Abstract:Debate distillation adapts weaker verifiers using multi-agent debate transcripts to improve their judgement in subsequent debates, but gains on monitored tasks do not establish reliability on related unmonitored tasks. We study epistemic reliability degradation, in which adaptation preserves monitored performance while reducing support for correct responses on hidden tasks. We consider an adversarial debater that manipulates debate arguments while defending the correct monitored response, and ask whether the resulting degradation merely reflects catastrophic forgetting and whether standard evaluation can detect it. To address these questions, we propose ER-Audit, a two-stage black-box auditing framework that compares frozen verifier checkpoints before and after adaptation, and introduce two evaluation benchmarks pairing monitored and hidden task prompts grounded in shared contexts. ER-Audit searches for counterexamples to non-degradation by evaluating semantically valid paraphrases and, if none is found, uses independent paraphrases for sequential hypothesis testing. We derive anytime-valid lower confidence bounds on the non-degradation probability, allowing data-dependent stopping within a finite budget. We further establish a common lower bound across fixed paraphrase distributions and extend it to distributions within a bounded total variation distance of their mixtures. Our experiments show that higher hidden-task accuracy can coexist with more counterexamples to non-degradation and lower non-degradation bounds. This divergence challenges explanations based solely on broad catastrophic forgetting and shows that auditing can uncover selective hidden-task degradation concealed by aggregate performance gains. Our code and benchmarks are available at this https URL.
[NLP-307] Gradients for Interventions and Activations for Detection: Targeted Feature Learning in Language Models
【速读】： 该论文旨在解决在生成式人工智能（Generative AI）模型中，如何针对预先定义的概念实现高效、精准的目标化特征学习（targeted feature learning）的问题。传统方法如稀疏自编码器（Sparse Autoencoders, SAEs）虽能构建广泛的概念特征字典，但其特征与具体概念的关联通常为事后分析，难以满足假设驱动的可解释性研究需求。为此，本文提出一种系统化的对比框架，聚焦于三种模型信号（激活值、激活梯度、参数梯度）与两种估计器（对比均值法与一维可学习编码-解码器）的组合，共构建六种目标特征学习方法，其中包含已有的CAA和GRADIEND方法，以及四种新提出的组合方法。实验在三个语言模型上对15项任务进行评估，综合考察特征检测能力与因果干预效果。结果表明：激活值类方法在特征检测性能上表现最优，而基于梯度的方法在因果干预中更具优势；特征质量取决于模型信号与估计器的协同作用，且检测与干预能力反映互补的特征属性。因此，该研究的关键在于通过系统性地整合不同信号与估计策略，为面向特定概念的特征学习提供可验证、可优化的范式。
链接: https://arxiv.org/abs/2609.32355
作者: Jonathan Drechsel,Steffen Herbold
机构: University of Passau (帕绍大学); Passau, Germany
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Model-internal features can be studied through both their ability to identify a specified concept and their causal effect when manipulated, e.g., through steering or weight editing. A prominent approach to feature learning is Sparse Autoencoders (SAEs), which learn broad feature dictionaries whose relation to particular concepts is typically identified post hoc. However, many interpretability questions are instead hypothesis-driven and concern a concept specified in advance. We study this setting as targeted feature learning, where a single feature is constructed for such a predefined concept. We present a controlled comparison across three model signals (activation values, activation gradients, and parameter gradients) and two estimators (contrastive mean and a learned one-dimensional encoder-decoder), yielding six targeted methods, with CAA and GRADIEND as existing instances and four new methods covering the remaining combinations. We compare these methods against pretrained SAEs across 15 tasks and three language models, evaluating both detection and causal intervention. Across models, the strongest detection performance is achieved by contrastive activation value methods, whereas the strongest intervention performance is achieved by gradient-based methods. Overall, our results show that targeted feature quality depends jointly on the model signal and estimator, with detection and intervention capturing complementary properties.
[NLP-308] Fewer Tokens More Self-Teaching: On-Policy Self-Distillation for Extreme Visual Token Reduction
【速读】： 该论文旨在解决多模态大语言模型（Multimodal Large Language Models, MLLMs）在极端低视觉令牌（visual token）预算下性能急剧下降的问题。现有方法虽尝试通过视觉令牌选择或基于训练的适应来缓解该问题，但在极低令牌保留率（如5%以下）时仍难以维持有效性能。本文提出一种关键创新——基于策略的自蒸馏（on-policy self-distillation, OPD），其核心在于：让一个经过重度压缩的模型（学生模型）在仅使用少量视觉令牌的情况下生成响应，并由同一模型的完整令牌版本（教师模型）在学生生成的轨迹上提供分布监督，从而实现对自身生成状态的有效学习。为稳定在严重信息缺失情况下的学习过程，进一步引入预算级课程学习（budget-level curriculum），逐步降低训练中的令牌预算，以增强模型鲁棒性。实验表明，在Qwen3.5-4B等多个模型上，该方法将5%视觉令牌保留率下的平均性能从68.6%提升至82.3%，显著优于无训练、基于训练及强化学习基线；且在不同规模模型间具备良好泛化能力。此外，该方法可减少85.2%的KV缓存占用与85.4%的预填充计算量（prefill FLOPs），同时不增加推理开销，证明了基于策略的自蒸馏能有效恢复因极端视觉令牌压缩而损失的核心能力。
链接: https://arxiv.org/abs/2609.32353
作者: Junxian Li,Ruixuan Yang,Tianao Zhang,Tiange Xu,Weisheng Dong,Yulun Zhang
机构: Shanghai Jiao Tong University (上海交通大学); University of Cambridge (剑桥大学); Xidian University (西安电子科技大学); Xi’an Jiaotong University (西安交通大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Code is at this https URL
Abstract:Visual token reduction is an effective way to accelerate multimodal large language models (MLLMs), but performance deteriorates rapidly under extremely low token budgets. Existing work has explored both visual-token selection and training-based adaptation to reduced visual inputs. We take a step further by asking how a heavily compressed MLLM should learn from the states induced by its own generations. This setting naturally calls for on-policy self-distillation: a heavily compressed model is supervised on the states induced by its own generations, while its full-token counterpart serves as an information-rich teacher. Based on this insight, we propose LT-OPD, a training framework for extreme visual-token reduction. The student rolls out responses with only a small fraction of visual tokens, and a frozen full-token copy of the same MLLM provides distributional supervision along these student-generated trajectories. To stabilize on-policy learning when visual evidence is severely limited, we further introduce a budget-level curriculum that progressively decreases the token budget during training. Across nine benchmarks on Qwen3.5-4B, LT-OPD raises average retained performance under 5% visual-token retention from 68.6% to 82.3%, outperforming training-free, training-based, and reinforcement-learning baselines at the same budget. The gains transfer consistently to Qwen3.5-9B, GLM-4.6V-9B, and LLaVA-OV-1.5-4B. LT-OPD also reduces KV-cache usage by 85.2% and prefill FLOPs by 85.4% without additional inference overhead, demonstrating that on-policy learning can substantially recover capabilities lost to extreme visual-token reduction.
[NLP-309] Language Distances are Practical for Equitable Cross-Lingual Transfer
【速读】： 该论文旨在解决跨语言迁移中源语言选择的不平等性问题，尤其关注低资源目标语言在缺乏充分评估条件下难以确定最优源语言的现实困境。其核心挑战在于现有基于语言距离的源语言排序方法在不同任务和资源水平下的一致性与公平性尚未得到充分验证。解决方案的关键在于系统评估多种源语言排序范式，发现仅依赖单一语言距离或“始终使用英语”作为源语言的基线方法在任务与资源层面均存在显著不平等；而通过引入无需训练的复合语言距离可显著缓解此类不平等，进一步采用经过任务特定训练的排序器则几乎完全消除不平等现象。研究还表明，基于语言距离的排序器在可靠性上优于依赖语言模型内部表征的排序方法。因此，论文主张：当具备任务特定的迁移评估能力时应采用训练型排序器，否则推荐使用复合距离方法，以实现高效且公平的跨语言迁移语言选择。
链接: https://arxiv.org/abs/2609.32331
作者: York Hay Ng,Razan Ahsan Rifandi,Aditya Khan,En-Shiun Annie Lee
机构: University of Toronto (多伦多大学); Ontario Tech University (安大略理工大学); Columbia University (哥伦比亚大学)
类目: Computation and Language (cs.CL)
备注: Accepted to MRL 2026
Abstract:Cross-lingual transfer is strongly conditional on how the source language is chosen, but it is impractical to determine the best candidate source for every target language, especially for low-resource target languages. Language distances are widely used to rank candidate sources due to their correlation with transfer efficacy and applicability in resource-sparse settings. However, the reliability of distance-based rankers across tasks and resource levels remains underexplored. We therefore present the first equity-focused evaluation of paradigms for ranking source languages, studying resource-level inequality and task inequality across ten cross-lingual tasks and two multilingual models. While both inequalities are most pronounced for individual language distances and an English-always baseline, they are substantially reduced by training-free composite distances, and nearly eliminated by trained rankers. We further demonstrate the reliability of rankers using language distances compared to rankers using language model internals. Overall, we find that language distances provide a practical basis for equitable and performant transfer language selection. We recommend using trained rankers when task-specific transfer evaluations are available, and composite distances otherwise.
[NLP-310] What Can a Leaderboard Certify? Compositional Controllability for Fair Evaluation and Training of Biomedical Literature-Review Agents
【速读】： 该论文旨在解决长时序智能体（long-horizon agents）在排行榜（leaderboards）评估中因评分差异无法准确判断系统可比性的问题，尤其关注输入、证据或预算不均衡导致的评分偏差。传统统计分析难以消除此类非对齐因素的影响，因而可能导致误导性结论。其核心解决方案是提出“组合可控性”（compositional controllability）框架，通过定义比较窗口（comparison window）覆盖单个阶段、多个阶段或整个代理流程，仅依赖窗口外的干扰项（nuisance）来严格界定可观测与受控得分差之间的差距。该方法提供了一种先验可接受性检验（admissibility test），在查看具体得分前即拒绝不可比的对比，确保仅当得分差距超过联合采样误差与干扰半径之和时，才可认证排序；否则保持未决状态。由此，每个系统获得一个排名区间，提升了评估的严谨性。研究构建了包含2,042篇生物医学文献的结构化断言图基准（BioLitBench），发现7个已发表管道中有14组配对被传统统计分析判定为胜出，但其中多数涉及目标综述的参考文献偏倚。经由匹配输入与固定主干模型的控制后，该测试拒绝了11组对比，包括所有涉及排名第一系统的比较，且其中7个原结论被纳入拒绝范围。进一步地，基于相同比较窗口支持阶段级训练，使用Qwen3.8-27B作为基础模型，对SCRIBE进行多阶段奖励驱动训练，在一致证据条件下，其获得[1,2]的认证排名区间，显著优于所有评估过的公开管道及Claude、OpenAI等代理；在相同检索池下，其性能与最强公开检索器相当，并被认证优于三个已发表管道。
链接: https://arxiv.org/abs/2609.32318
作者: Zhaowei Han,Xiang Zhang,Lingxiao Guan,Danqi Hu,Kai Liu,Kevin Chang,Jie Liu
机构: University of Michigan, Ann Arbor(密歇根大学安娜堡分校)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 33 pages, 1 figure, 18 tables. Zhaowei Han, Xiang Zhang, and Lingxiao Guan contributed equally. Code: this https URL
Abstract:Leaderboards rank long-horizon agents by their final outputs. Yet a higher score alone does not establish whether two systems are comparable or which stage accounts for the difference. Unequal evidence, inputs, or budgets can affect scores, and statistical corrections do not remove this mismatch. We introduce compositional controllability to address these questions. A comparison window covers one stage, several stages, or the whole agent. Our central result bounds the gap between observed and controlled score differences using only nuisance outside the window. This yields an admissibility test applied before scores are inspected. Inadmissible comparisons are refused. For admissible pairs, an ordering is certified only when the score gap exceeds the combined sampling and nuisance radii; otherwise, it remains undecided. These decisions give each system a rank interval. We introduce BioLitBench, a benchmark of 2,042 biomedical articles represented as structured claim graphs. Among seven published pipelines, a conventional statistical analysis declares a winner in 14 of 21 pairwise comparisons. Yet the top-ranked system alone received the target review’s bibliography. To isolate pipeline performance, our test requires matched inputs and a fixed backbone model. It refuses 11 of the 21 comparisons, including every comparison involving the top-ranked system. Seven of the 14 conventional conclusions fall within these refused pairs. The same comparison windows support stage-level training. We train SCRIBE on Qwen3.8-27B using rewards measured at each stage’s exit. Under matched evidence, SCRIBE achieves a certified rank interval of [1,2], with certified advantages over all evaluated published pipelines and the evaluated Claude and OpenAI agents. Under same pool, SCRIBE matches the strongest published retriever and is certified above three published pipelines.
[NLP-311] RAP: Understanding and Mitigating Privacy Memorization in Language Models
【速读】： 该论文旨在解决在对语言模型进行微调时，因敏感数据记录的过度记忆（memorization）而导致隐私泄露的问题，尤其关注在事先无法确定哪些文本片段属于敏感信息的情况下如何有效抑制此类记忆。其核心挑战在于：传统方法依赖于对敏感内容的先验标注或全局正则化策略，难以精准针对高风险、罕见且难以预测的敏感片段进行防护。解决方案的关键在于提出一种名为“目标参考优势”（Target Reference Advantage, TRA）的新指标，该指标通过将目标模型与使用同一语料库互补子集训练的参考模型进行对比，获得每标记（token）层面的可微分信号，从而区分模型对特定记录的记忆行为与其跨记录的泛化学习。基于此，研究进一步发现记忆程度随训练过程持续增长，尤其在小样本、高学习率及任务复杂度高的情况下更为显著，而常规早停策略对这类稀有敏感片段的抑制效果最差。为此，论文提出TRAP（TRA-based Penalty），一种仅在目标模型超越参考模型时施加单边惩罚的机制，有效抑制了记忆增长。实验表明，在包含个人身份信息的学生作文和含患者标识的临床案例中，TRAP可使记忆水平降至接近未训练模型的水平，且对模型性能影响极小，显著优于通用正则化、差分隐私等现有方法。
链接: https://arxiv.org/abs/2609.32293
作者: Muhammed Ustaomeroglu,Ziyue Xu,Hanshen Xiao,Peter Cnudde,Guannan Qu,Holger R. Roth
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Fine-tuning a language model on sensitive records can leave it able to reproduce them. We ask when this memorization arises and how to prevent it without knowing in advance which spans are sensitive. Our starting point is that most memorization scores and attacks share one statistical core: whether the model assigns a token more probability than some reference would. Taking as the reference a model trained on the complementary half of the same corpus gives the Target Reference Advantage (TRA), a per-token signal that separates what a model fit to a particular record from what it learned across records, and is cheap and differentiable. We then study what drives memorization during fine-tuning: it keeps growing well past the validation minimum, is larger on small datasets and at higher learning rates, and higher when the underlying task is harder. Early stopping removes much of it, but because it is chosen by aggregate validation loss it helps least for rare, hard-to-predict spans embedded in otherwise learnable text, which is exactly what sensitive information tends to be. We therefore introduce TRAP, a one-sided penalty on tokenwise TRA that acts only where the target model pulls ahead of its reference. On student essays with annotated personal information and clinical cases with patient identifiers, TRAP brings memorization near the level of an untrained model at little utility cost, where generic regularizers barely move and differential privacy gives up most of what fine-tuning bought.
[NLP-312] Supporting and Performing Culture from the Inside EMNLP2026
【速读】： 该论文旨在解决自然语言处理（Natural Language Processing, NLP）领域在文化研究中普遍存在的认知与方法论偏倚问题，具体表现为研究者往往忽视文化语境的内在复杂性，导致模型设计与评估未能充分反映多元文化的实际特征。其核心问题是：当前NLP研究在任务选择、数据收集、系统设计与评估等全链条中，多采用“外部视角”（etic），即以西方主流学术范式为基准进行文化现象的客观化、标准化处理，而忽略了“内部视角”（emic）——即特定文化群体自身的认知框架与表达逻辑。解决方案的关键在于引入“文化视角”（emic vs. etic）的分析框架，系统审视并重构NLP研究中的文化假设，推动建立更具文化敏感性的研究范式，从而实现更符合真实文化语境的生成式人工智能（Generative AI）发展路径。
链接: https://arxiv.org/abs/2609.32281
作者: Lea Frermann,Steven Bird
机构: The University of Melbourne(墨尔本大学); Charles Darwin University(查尔斯达尔文大学); The University of Tübingen(图宾根大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: 9 pages, 1 figure, To appear in Findings of the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026), Budapest, October 2026
Abstract:Anthropology and related disciplines which study culture have often found it useful to consider their epistemological and methodological approaches in terms of emic versus etic. This is the distinction between the insider perspective, how the world is viewed by a member of the culture, versus the outsider perspective, the scientific cataloguing of cultural knowledge and practice. We adopt this framework to systematically analyse assumptions in recent research on culture in NLP along the full pipeline of task selection, data collection, system design and evaluation. We draw attention to the `culture of NLP’ which shapes the research focus and approaches of our field, and suggest pathways towards a more culturally attuned AI.
[NLP-313] Before Answering: Evidence Sufficiency under Size-Matched Memory Construction
【速读】： 该论文旨在解决在压缩或检索记忆中进行问答时，智能体如何准确识别查询所需证据是否已不再记忆中的问题。传统基准测试通过删除支持性段落来构建“证据不足”样本，但研究发现这种构造方式会因记忆大小变化而泄露标签信息，导致简单计数模型在MuSiQue数据集上获得高达0.979的AUROC，远超预期性能，形成不可靠的捷径（shortcut）。为此，作者提出一种大小匹配（size-matched）的构造方法，可严格消除记忆大小带来的标签泄露，从而实现更公平的评估。其核心解决方案是引入MemSafe——一种基于交叉编码器（cross-encoder）与集合变换器（set transformer）的证据检测器，通过将查询与每个记忆单元进行交叉编码并聚合，以捕捉语义层面的不一致。实验表明，在三个多跳问答数据集上，MemSafe显著优于词法基线（提升0.26~0.39 AUROC），尤其在MuSiQue和HotpotQA上分别达到0.968和0.983的AUROC；然而在2WikiMultiHopQA上表现饱和，且在未回答问题上性能下降明显（如SQuAD 2.0仅0.639 AUROC），需大量临床训练数据才能超越特征基线。当作为7B阅读器的门控机制使用时，它能在5%覆盖下将错误率从0.850降至0.631，优于阅读器置信度与真实完整性标签，尽管在10%覆盖下仍逊于7B大语言模型判别器。这些结果表明，证据不足样本的构造方式对模型评估具有决定性影响，其重要性甚至超过所用检测器本身的设计。
链接: https://arxiv.org/abs/2609.32269
作者: Joyanta Jyoti Mondal,Md. Shifatul Ahsan Apurba,Mridul Banik,Md Masud Al Mahmud,Ibne Farabi Shihab
机构: University of Delaware (美国特拉华大学); University of Alabama at Birmingham (美国阿拉巴马大学伯明翰分校); Indiana University Indianapolis (印第安纳大学印第安纳波利斯分校); BRAC University (BRAC大学); Iowa State University (爱荷华州立大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 24 pages, 10 figures
Abstract:Agents that answer questions from compressed or retrieved memory must recognize when the evidence a query needs is no longer in memory. Benchmarks for this task usually create insufficient-evidence examples by deleting supporting passages. We show that this construction leaks the label through memory size: on MuSiQue, a classifier that only counts paragraphs reaches an area under the ROC curve (AUROC) of 0.979 for detecting unsafe memory, higher than the lexical estimator we initially evaluated. We propose a size-matched construction that provably removes this shortcut, and use it to study MemSafe, an estimator that cross-encodes the query with each memory unit and aggregates the units with a set transformer. Across three multi-hop question answering datasets and five seeds, MemSafe reaches 0.968 and 0.983 AUROC on MuSiQue and HotpotQA, 0.26 to 0.39 above a lexical baseline, while the third dataset, 2WikiMultiHopQA, is saturated. A frozen pretrained cross-encoder with a logistic head already closes 41% of the MuSiQue gap between the lexical baseline and MemSafe. At the same time, MemSafe degrades more than a weak baseline on the unanswerable questions released with MuSiQue, reaches only 0.639 AUROC on SQuAD~2.0, and needs several thousand clinical training examples before it outperforms a feature-based estimator. Used as a gate for a 7B reader, it reduces the error rate on answered questions from 0.850 to 0.631 at 5% coverage, outperforming both reader confidence and, on average, the ground-truth integrity label, although a 7B LLM judge is the better gate at 10% coverage. These results indicate that the way insufficient evidence is constructed matters as much as the estimator that detects it.
[NLP-314] LANTERN: Illuminating Hidden Mathematical Knowledge in Language Models
【速读】： 该论文旨在解决数学研究中如何高效识别具有潜在价值的数学关系这一关键问题，即在海量数学对象（如整数序列）中自动发现值得深入探索的关联性。其核心挑战在于传统方法依赖研究人员主观判断，而生成式模型虽能生成新猜想，却难以自主筛选出真正有意义的数学联系。解决方案的关键在于提出LANTERN——一个基于预训练模型激活值构建分类器的快速、低成本流水线，通过多阶段处理实现高效筛选与验证：首先利用分类器对候选关系进行排序，再经分层过滤、假设生成、可执行验证及分析检查，最终实现从5000万对序列中精准挖掘出62个经验证的关系，其中13个经内容筛选保留，9个具有信息量或洞察力，包含4个迄今未被记录的全新发现。整个端到端流程耗时不足8小时，显著提升了数学知识发现的自动化与可扩展性。
链接: https://arxiv.org/abs/2609.32264
作者: Pavel Tikhonov,Elena Tutubalina,Ivan Oseledets,Dmitry I. Ignatov,Mikhail Seleznyov
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Language models can now prove theorems, but people still decide which problems to pursue. We ask whether a model’s internal representations can help identify promising mathematical connections. We develop LANTERN, a fast, cost-efficient pipeline that uses a classifier over pretrained-model activations to rank candidate relations, followed by staged filtering, hypothesis generation, executable verification, and analytical checking. Applied to the On-Line Encyclopedia of Integer Sequences (OEIS), LANTERN ranked 50 million pairs among 10,000 frequently referenced sequences and produced 62 verified relations between pairs without an existing OEIS cross-reference. A content screen retained 13 relations worth presenting; nine of these are informative or insightful, including four which are entirely novel to the best of our knowledge: none appears in the OEIS or in our targeted literature search. The entire end-to-end process including classifier training, candidate ranking, filtering and verification took under 8 hours.
[NLP-315] Clarify the User or Verify the World? Uncertainty Routing for Proactive Agents
【速读】： 该论文旨在解决工具使用型大语言模型（LLM）代理在决策过程中面临的不确定性路由问题，即如何在用户意图不明确或环境信息缺失时，准确判断应向哪个信息源（如用户澄清或环境验证）获取补充信息。现有主动式方法通常仅专注于用户澄清或环境验证中的单一任务，缺乏对具体决策场景下最优信息来源的显式选择机制。为此，论文提出一种名为PROUR的主动不确定性路由框架，其核心在于将动作不确定性分解为两类信号：一是多个合理用户目标解释之间的分歧程度，反映用户侧的模糊性；二是每个解释内部剩余的信息熵，反映世界侧证据的缺失。基于此，通过引入模式条件化的信息增益奖励训练查询生成器，使代理能够根据当前任务阶段（澄清用户目标或验证下一步行动）精准地从对应源获取信息。实验表明，在τ-bench基准上，PROUR在零售与航空领域平均成功率提升至28.17%，优于最强基线4.57%，且交互步数减少2.17步；更重要的是，该学习策略无需重新训练即可泛化至τ³-bench中的更复杂任务与交易类场景，验证了源对齐的不确定性解析对于构建高效主动代理的关键作用。
链接: https://arxiv.org/abs/2609.32255
作者: Zhaofeng Li,Xuan Zhang,Xiaokui Xiao,Yang Deng
机构: National University of Singapore(新加坡国立大学); Singapore Management University(新加坡管理大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Tool-using LLM agents must decide not only whether additional information is needed, but also which source can resolve the uncertainty. Existing proactive approaches often specialize in either user clarification or environment verification, without explicitly determining the appropriate information source for each decision. We formulate this problem as uncertainty routing among ACT, CLARIFY, and VERIFY, and propose PROUR, a proactive uncertainty routing framework. PROUR decomposes action uncertainty into disagreement across plausible user-goal interpretations, which signals user-side ambiguity, and the entropy remaining within each interpretation, which signals missing world-side evidence. To acquire information from the routed source, a query generator is trained with a mode-conditioned information-gain reward, targeting user-goal identification under CLARIFY and next-action identification under VERIFY. On \tau -bench, PROUR achieves 28.17% average success rate across retail and airline, outperforming the strongest prior method by 4.57% while using 2.17 fewer interaction steps. The learned policy further generalizes to stronger task agents and transactional domains of \tau^3 -bench without retraining, demonstrating the benefit of source-aligned uncertainty resolution for proactive agents.
[NLP-316] Solving Every Step Is Not Enough: Milestone Oracles Reveal a Composition Gap in LLM Math Reasoning NEURIPS2026 ACL
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在处理多步数学问题时，尽管能够独立完成每一步推理且已提供完整步骤路线图与各步骤答案，仍无法正确求解整体问题的“推理失败”问题。其核心挑战在于识别模型在复杂推理链中具体在哪一环节出现断裂，而非简单评估最终答案正确性。为此，作者提出OracleLadder诊断评估框架，通过逐步引入“上帝视角”（oracle）帮助——从无帮助、仅提供路线图、路线图+中间目标答案，到逐个里程碑单独测试——系统性定位模型的五类推理缺陷。关键创新在于结合教师模型生成固定子目标路线图与确定性符号验证器对每一步答案进行严格评分，从而将失败归因于特定类型的认知缺口。实验结果表明，所有测试模型（参数量从8B至671B不等，涵盖Qwen3、gpt-oss、Llama 3.3、DeepSeek-V3.1等）均存在最显著的“组合性缺口”（composition gap），这是一种比传统组合性缺陷更严格的推理障碍，影响33%-48%的问题，即使剔除可能由模型评审标记出的评分错误后仍占24%-37%。此外，不同模型在准确率与里程碑辅助恢复能力上的表现排序不一致，且两次强化学习基于价值重估（RLVR）实验虽取得相似准确率提升，却导致问题恢复路径差异显著，说明模型行为具有高度非线性特征。该诊断范式在MATH500和AIME 2024/25数据集上复现有效，且在独立第二位教师评估下，个体问题的恢复一致性高达83%-87%，同时该“帮助阶梯”方法可迁移至代码生成任务。研究团队已公开发布全部数据、路线图、提示模板及代码，以推动可解释性推理研究的发展。
链接: https://arxiv.org/abs/2609.32235
作者: Zhuohan Wang,Haoran Ma,Tianyu Wu,Yuanlin Duan,Zichun Liao,Jieming Yu
机构: Rutgers University(罗格斯大学); Harvard University(哈佛大学); The Hong Kong University of Science and Technology(香港科技大学)
类目: Computation and Language (cs.CL)
备注: Accepted at NeurIPS 2026 (Evaluations and Datasets Track). 47 pages. Code and data: this https URL
Abstract:Large language models (LLMs) can solve every intermediate step of a multi-step math problem on its own and still fail the full problem, even when given a roadmap of the steps and all of their answers. We introduce OracleLadder, a diagnostic evaluation that locates where LLM math reasoning fails by giving the model increasing levels of oracle help. For each problem, a teacher model writes a fixed roadmap of intermediate sub-goals (milestones), and a deterministic symbolic verifier grades every answer. Testing the model with no help, with the roadmap, with the roadmap plus the milestone answers, and on each milestone alone sorts each failure into one of five reasoning gaps. On 354 NuminaMath problems and six models from 8B to 671B parameters (Qwen3, gpt-oss, Llama 3.3, DeepSeek-V3.1), the largest gap for every model is the composition gap, a stricter form of the compositionality gap. It covers 33-48% of problems, and 24-37% after removing problems that an LLM review flags as grading errors. Accuracy and milestone-help recovery rank the two strongest models differently, and two RLVR runs with similar accuracy gains move problems differently. The roadmap effect replicates on MATH500 and AIME 2024/25, per-problem recovery agrees for 83-87% of problems under an independent second teacher, and the help ladder carries over to code generation. We release the data, roadmaps, prompts, and code at this https URL.
[NLP-317] KinyaMed: Seeds Not Rows – What a Corpus Requirement Written in the Wrong Unit Fails to Constrain
【速读】： 该论文旨在解决在卢旺达医疗中心多语言环境中构建患者语音紧急程度分类器（urgency classifier）时所面临的高质量标注数据生成与评估难题。其核心问题在于：尽管需求明确要求生成一百万条高质量样本，但现有生成方法在未达成预期质量标准的情况下即完成了数据量目标，暴露出生成式数据在真实应用场景中的可靠性缺陷。解决方案的关键在于放弃传统分类器的开发，转而构建一套可检测并量化生成数据质量问题的评估仪器（instruments），以揭示生成过程中的系统性缺陷。这些仪器发现，当前方法在九项质量门禁中失败四项（六项若按来源句归属计算），根本原因在于“种子语句”（seed phrases）数量受限（每条种子最多50行），导致即使生成速度高达每秒7700行，也无法满足种子多样性要求；同时，仅依赖行数约束无法控制昂贵的人工标注成本，反而放任了质量下降。此外，实验还揭示出模型对输入格式敏感（如大小写、单个拼写错误即导致31.5%预测偏差），且未配备分词器的模型仍能返回看似合理的概率输出，掩盖了其无法处理实际输入的本质。因此，本文的核心贡献并非一个分类模型，而是可复现的评估框架及其揭示的负面结果，强调了生成式AI（Generative AI）在关键医疗应用中必须建立严格的质量验证机制，否则数据生成本身可能成为潜在风险源。
链接: https://arxiv.org/abs/2609.32234
作者: Marius Bayizere
机构: Independent Researcher(独立研究员); Kigali, Rwanda(基加利, 卢旺达); GitHub(GitHub)
类目: Computation and Language (cs.CL)
备注: 34 pages, 10 tables, 1 figure. Negative-results and methodology paper; no model performance claims are made
Abstract:Triage decides who is seen first. Building an urgency classifier for patient-voice Kinyarwanda, we found our specification could be met without producing anything it was meant to secure. We report that, and the instruments that detect it, instead of a classifier. Designed for the four languages a Rwandan health centre receives, with every instrument per-language: sentences are authored in all four arms and rows generate in one, because the frame slots those three need do not exist. Our requirement asked for one million examples; generation produced them in 130 seconds. It fails four of the nine quality gates in that specification, six when rows are attributed to their source sentence. The binding gate counts distinct authored seed phrases, not rows: at our 165, no corpus passes at any row count. Two shortfalls follow and differ: 2,835 further sentences to pass the seed-count gate, 19,835 to reach the stated million rows, because a separate gate caps a seed at 50 rows. Rows come from a machine at 7,700 per second; seeds from clinicians. A row count constrains the cheap quantity, leaves the expensive one free, and so does not constrain quality at all. Two further negatives follow. An evaluation set of 17,942 rows built from nine distinct sentences supports no verdict: our gate, which counts distinct sentences, refuses 38 of its cells and reports nothing. A model trained on a corpus of uniform surface form changes its predicted urgency for 31.5% of inputs under capitalisation and 21.0% under a single typo, reported as measurement and not attribution. A model directory shipped without its tokenizer loads without error and answers its class prior on input it cannot read, with well-formed probabilities. No figure here is evidence of model quality; the contribution is the apparatus and the negative results it produced, reproducible from a clean clone except where marked NOT REPRODUCIBLE.
[NLP-318] OptiArena: Can LLM s Improve Executable Algorithms under Fixed Resource Budgets? EMNLP2026
【速读】： 该论文旨在解决现有静态问答（Static QA）与代码生成基准测试无法全面反映大语言模型（Large Language Models, LLMs）作为编程代理和研究工具的实际作用这一问题。其核心挑战在于评估LLMs在有限资源、固定代码骨架及受控反馈条件下，通过多轮迭代优化可执行算法的能力。解决方案的关键在于提出OptiArena——一个预算可控的测试平台，支持在五轮代码修改、固定最小代码框架、受限评估器反馈和固定资源预算下，系统性地评估模型对游戏可执行算法的改进能力。该平台引入了表面混淆控制（surface obfuscation controls）、校准参考基线（calibrated references）、保留/压力划分（held-out/stress splits）以及退化与异常故障诊断机制，并将LLM API调用成本与本地评估器运行时间分离记录，以实现更精准的性能衡量。实验结果表明，尽管不同模型和游戏间存在显著差异，但多数前沿LLMs在修复弱初始代码方面表现优于对已具备能力的基线进行精炼，且优化效果在表面混淆控制下仍部分保持，验证了该框架在有限资源约束下评估算法优化潜力的有效性。
链接: https://arxiv.org/abs/2609.32227
作者: Wenjun Peng,Xinyu Wang
机构: 未知
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 (Findings)
Abstract:Static QA and code-generation benchmarks only partially capture the role that large language models (LLMs) now play as coding agents and research tools. We introduce OptiArena, a budget-controlled testbed for studying whether LLMs can improve executable game-playing algorithms through five rounds of code edits within a fixed minimal scaffold and under bounded evaluator feedback and fixed resource budgets. The testbed uses two optimization regimes, surface obfuscation controls, calibrated references, held-out/stress splits, and diagnostics for degradation and exceptional failures, with LLM API cost reported separately from local evaluator wall-clock. The empirical study asks three questions: whether models can close the calibrated gap between a designated weak starter and an editable competent baseline, whether they can refine editable competent baselines without damaging them, and whether gains survive surface obfuscation controls. Across twelve frontier LLMs and five games, models improve designated weak starters more consistently than they refine editable competent baselines, with substantial variation across games and models. OptiArena provides a practical testbed for measuring bounded-resource algorithm optimization within the five-edit, fixed-scaffold setting studied here. Code is available at this https URL.
[NLP-319] A model of rational interlocutors: Unification of comprehension and production
【速读】： 该论文旨在解决话语理解（comprehension）与话语生成（production）中对话伙伴调整机制的分离问题，即尽管两者均涉及对对话对象的建模，但相关研究长期独立发展。其核心解决方案是提出理性对话者（Rational Interlocutor, RI）模型，将理解与产出统一为对对话伙伴模型的一体化计算过程。该模型通过三个参数定义对话伙伴的内在表征：身份参数（Π）决定对对方话语内容和形式的预期；可信度参数（Φ）反映信息传递的可靠性；知识参数（Λ）衡量对对方认知能力的估计。在理解过程中，个体基于语境选择最可能被对方意图传达的信息，权衡候选意义与说话人表达可能性之间的匹配度；在生成过程中，则选择能使对方最有效恢复目标意义且付出最小表达代价的言语形式。这一统一框架解释了为何理解者对语言能力较弱的说话者更少依赖其形式特征，而生产者则倾向于为其投入更多形式设计努力。作者进一步假设，感知的语言能力可分解为可信度与知识两个维度，该分解在第二语言学习者、儿童及人工对话伙伴中的不同表现可导出可检验的预测。
链接: https://arxiv.org/abs/2609.32216
作者: Hanlin Wu,Zhenguang G. Cai
机构: 未知
类目: Computation and Language (cs.CL)
备注: A computational model of partner modeling in language comprehension and production. 62 pages, 5 figures, 3 tables, including supplementary materials
Abstract:Who we communicate with influences both our interpretation of their utterances and the design of our own. Such adjustment to the conversational partner is studied as speaker modeling in comprehension and as audience design in production, with the two literatures having developed largely separately. We argue that both adjustments express one rational computation and propose the rational interlocutor (RI) model, a computational account unifying comprehension and production. An interlocutor maintains a model of their partner, defined by three parameters: an identity parameter \Pi sets the messages and forms expected from the partner; a fidelity parameter \Phi sets how reliably messages and utterances map onto each other for them; a knowledge parameter \Lambda sets how knowledgeable the partner is believed to be. Comprehension and production are thus mirror-image modes of one computation over the partner model. Comprehension chooses the message the partner most likely intends to convey, weighing how well each candidate fits the utterance against how likely this partner is to mean it. Production chooses the utterance from which the partner will best recover the message, weighed against the effort of saying it. This explains why comprehenders appear to rely less on the forms produced by a linguistically less competent speaker, while producers tend to invest more effort in designing forms for them. We conjecture that perceived linguistic competence decomposes into two of these quantities: fidelity and knowledge. Their contrasting profiles across second-language (L2) adults, children, and artificial partners produce distinct and testable predictions.
[NLP-320] HM-ROUTER: Joint Model and Harness Routing for Agent ic Systems
【速读】： 该论文旨在解决智能体（Agent）在实际应用中因模型与调度框架（harness）组合不匹配而导致性能下降的问题，尤其在训练数据无法覆盖日益增长的模型-调度框架组合空间时，如何高效、准确地为每个查询选择最优的模型与调度框架组合。其解决方案的关键在于提出HM-Router，一种基于张量分解思想的路由方法：通过学习共享的模型与调度框架表示，并引入受经典正交多项式（Canonical Polyadic, CP）张量分解启发的交互项，动态捕捉不同组合在特定查询下的兼容性变化。该设计实现了跨路径的表示共享，使已观测组合的训练样本能够有效指导对未观测组合的预测，显著提升了路由准确性与泛化能力。实验表明，在涵盖293条路径、73个模型和25个调度框架的基准上，HM-Router在平均路由准确率上优于最强基线7.3个百分点，且在多种成本预算下均表现领先；当90%路径的训练数据被屏蔽时，允许预测未观测组合可使归一化准确率提升15.8个百分点，充分验证了其在新路径、新组件上的训练样本效率与跨基准泛化能力。
链接: https://arxiv.org/abs/2609.32213
作者: Hao Mark Chen,Royson Lee,Yasuyuki Okoshi,Dimitris Anastasiou,Wayne Luk,Hongxiang Fan
机构: Imperial College London(帝国理工学院); Samsung(三星); Institute of Science Tokyo(东京科学研究所)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Agent performance depends on both the underlying model and the harness that manages its tool use and execution. Selecting a suitable pair requires accounting for their compatibility, yet training samples may cover only a subset of the growing combination space. We introduce HM-Router, a routing method that jointly selects a model and harness for each query. It learns separate model and harness representations shared across routes, with an interaction term inspired by canonical polyadic (CP) tensor decomposition to capture how their compatibility varies with the query. This sharing allows training samples from observed pairs to inform predictions for unobserved combinations. We curate a benchmark from 12 public agent benchmarks, covering 293 routes, 73 models, and 25 harnesses. HM-Router exceeds the strongest evaluated learned baseline by 7.3 percentage points in mean routing accuracy and leads at all seven evaluated cost budgets on the six-benchmark subset. When 90% of routes have their training outcomes withheld, allowing unobserved combinations improves normalized accuracy by 15.8 points over restricting the same router to observed routes. HM-Router has also demonstrated training sample efficiency for new routes and components and generalization to unseen benchmarks. Our code and data are open-sourced at this https URL.
[NLP-321] Generalization and Memorization along the Learning Trajectory of Neural Language Models: A Geometric Account of Categorization
【速读】： 该论文旨在解决神经语言模型在学习过程中泛化（generalization）与记忆（memorization）的动态演化机制问题。其核心挑战在于揭示模型在训练过程中如何从早期的抽象类别泛化逐步过渡到后期对具体实例的记忆。解决方案的关键在于利用受控的合成语法（synthetic grammars）对模型在训练轨迹中的表示空间几何结构与行为进行系统分析，发现：在学习初期，即使尚未遇到特定词项，表示空间中未被观测词项占据的连续区域便已呈现出系统性的几何组织，形成支持对未出现组合进行泛化的类别级结构；这一现象表明泛化能力在训练初期即已存在，而非依赖于大量记忆积累。随着训练的持续，大模型逐渐区分已观察与未观察的语法规则，但与此同时，支持类别泛化的连续几何结构却逐步被破坏。由此揭示出模型的学习范式具有阶段性：初始阶段以基于类别的泛化为主，后期则逐渐转向更具体的实例记忆。
链接: https://arxiv.org/abs/2609.32199
作者: Wang Bojun,Holly Jenkins,Elizabeth Wonnacott
机构: 未知
类目: Computation and Language (cs.CL)
备注: 5 figures in main text, 9 page main text
Abstract:We investigate how generalization and memorization develop along the learning trajectory of neural language models. Using controlled synthetic grammars, we examine both the geometry of representation space and model behaviour over training. We find that continuous regions of representation space not occupied by observed tokens become systematically structured from the earliest stages of learning, forming category-level geometric organization that supports generalization to unattested combinations. Generalization therefore emerges from the beginning of learning rather than only after extensive memorization. With prolonged training, larger models increasingly distinguish observed from unobserved grammatical combinations. At the same time, the continuous geometric structure supporting category-level generalization is gradually destructed. Together, these results suggest that neural language models initially learn through categorization-based generalization, followed by a gradual transition toward more exemplar-specific memorization.
[NLP-322] he Judge Is Not Its Twin: Post-training makes a models writing more predictable but barely moves its taste as a judge toward predictable writing
【速读】： 该论文旨在解决生成式 AI 在持续训练过程中，其输出逐渐趋于可预测性所带来的评估偏差问题。具体而言，当模型在后训练阶段使自身写作变得更加可预测时，若使用同一模型作为评判者（judge），可能使其在自动评估中倾向于奖励可预测的文本，从而导致创造性进步在自动化评价体系下变得不可见。解决方案的关键在于：尽管模型在作为“作者”时表现出明显的可预测性增强（如在多个提示下对自身训练阶段生成的故事感到越来越熟悉，且另一模型对其生成内容的惊讶度下降7.0–25%），但当其作为“评判者”时，其对可预测性的偏好并未显著增加——无论是判断哪篇故事更优，还是评估哪篇更具创造性，训练后的模型均未表现出对可预测性更强故事的系统性偏好。相反，训练反而强化了对较长故事的偏好，并导致“更创造性”这一评价维度失去一致性，即训练后的模型不再能可靠区分原始故事与其乱序版本。这表明，当前自动化评估机制在面对模型自我演化时存在局限性，而模型在作为裁判时具备一定程度的稳定性，但该稳定性不足以支撑对创造性的有效量化。
链接: https://arxiv.org/abs/2609.32196
作者: Arman Nik Khah,Arvin Bahreini
机构: The University of Texas at Dallas(德克萨斯大学达拉斯分校); University of Oregon(俄勒冈大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 15 pages, 4 tables. Code and data: this https URL
Abstract:Language models are now routinely graded by other language models. If post-training makes a model’s own writing more predictable, it may also teach the same model, acting as a judge, to reward predictable writing, so that progress on creativity would be invisible to automated evaluation. We follow two open model families, OLMo-2 and Zephyr (7B parameters each), through their public training stages and measure every stage twice, as a writer of short stories and as a judge of pairs of stories. As writers, the models drift as feared: each family’s fully trained model finds the stories of its base, supervised fine-tuned (SFT) and preference-trained (DPO) stages progressively more familiar, in all ten prompts, and a model from the other family finds the trained stories 7.0 to 9.2 percent (OLMo-2) and 24 to 25 percent (Zephyr) less surprising per token. As judges, they barely move toward predictable writing. Asked which story is better, a question every judge can use to tell a story from its own words in scrambled order, no trained judge’s estimated tilt toward the more predictable story grows by as much as one point in the probability of picking it. A post hoc one-sided 95% upper bound on that growth is 2.7 points on an average pair, about the size of the untrained OLMo-2 judge’s own tilt. Asked which is more creative, no trained judge’s estimate favors the predictable story more than its base’s does. Training instead strengthens a preference for longer stories when the question is creativity, and for one answer slot, and it breaks “more creative” as a question: trained judges asked it no longer reliably prefer a story to its scrambled words. A follow-up could not build pairs that differ in predictability but not in quality, because the routes that made this writer’s stories less predictable also broke some of them, often enough to fail a quality floor set in advance.
[NLP-323] ScopeIF: Improving Scope-Aware Precise Instruction-Following in Large Language Models via Graded Reward Modeling
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在复杂应用场景中对具有特定作用范围（Scope）的约束条件进行精准遵循的问题。现有优化方法在数据构建过程中常忽略约束的作用范围，且依赖于针对每个约束的二值化奖励信号，导致训练数据多样性不足，对复杂约束的监督稀疏，难以实现精细化控制。为应对这一挑战，论文提出一种名为ScopeIF的新型训练框架，其核心在于引入一个统一的分解范式，将客观约束解耦为三个独立维度：作用范围（Scope）、目标对象（Target）和取值范围（Range）。基于此范式，构建了大规模、具备多样作用范围感知能力的指令数据集ScopeInstruct，并结合工具驱动的验证机制与分级奖励建模，量化每条约束的违反程度，从而提供密集的监督信号以支持策略优化。实验表明，ScopeIF在复杂作用范围感知约束任务上显著优于现有方法，同时保持模型的通用能力；尤其在优化后的Qwen3-4B和8B模型上，性能可媲美甚至超越Gemini-2.5-Pro和DeepSeek-V3.2等前沿模型，确立了一种有效的面向作用范围感知指令遵循的先进范式。
链接: https://arxiv.org/abs/2609.32189
作者: Bosi Wen,Yilin Niu,Xiaoying Ning,Ying Zhang,Hongning Wang,Minlie Huang
机构: Tsinghua University (清华大学); Zhipu AI
类目: Computation and Language (cs.CL)
备注: 28 pages, 8 figures
Abstract:Precise instruction-following is a fundamental ability of large language models (LLMs), requiring their outputs to strictly satisfy objective constraints in input instructions. In complex application scenarios, these constraints often possess diverse scopes that govern specific response segments rather than the entire output. However, existing optimization methods often neglect constraint scope during data construction and rely on binary per-constraint rewards, yielding limited data diversity and sparse supervision for complex constraints. To this end, we propose ScopeIF, a novel training framework for scope-aware precise instruction-following. We first introduce a unified schema that factorizes objective constraints into three decoupled dimensions: Scope, Target, and Range. Grounded in this schema, we construct ScopeInstruct, a large-scale instruction dataset with diverse scope-aware constraints, and combine tool-grounded verification with graded reward modeling to quantify the violation degree of each constraint, providing dense supervision for policy optimization. Extensive experiments demonstrate that ScopeIF consistently outperforms existing methods, particularly on complex scope-aware constraints, while preserving general capabilities. Notably, it enables optimized Qwen3-4B and 8B models to rival or surpass strong frontier models such as Gemini-2.5-Pro and DeepSeek-V3.2, establishing an effective paradigm for advancing scope-aware instruction-following. Our code and data are available at this https URL.
[NLP-324] yped Decision Models: An Early Evidence Audit and Evaluation Checklist
【速读】： 该论文旨在解决生成式模型在决策任务中因输出文本不可控、难以直接映射到预定义选项而导致的可解释性与可用性问题。其核心挑战在于如何在不生成自然语言文本的前提下，实现对用户自定义选项的概率分布精准预测，从而提升模型在实际应用中的效率与可靠性。解决方案的关键在于采用类型化决策模型（Typed Decision Model, TDM），通过结构化输出（即直接返回预设类别上的概率分布）替代传统文本生成方式，显著降低推理延迟与部署成本。尽管当前证据表明，该类模型在简单任务上尚未展现出独立于标签概率读出（label-probability readout）的准确率优势，但在复杂任务中仍存在性能差距；然而，其在延迟和成本方面的优化表现突出，尤其适用于需基于置信度进行模型或人工干预切换的场景。研究基于早期28篇文献的共性缺陷，提出一个包含14项评估指标的检查清单，以指导未来TDM研究的系统性验证。需要强调的是，由于现有证据仅涵盖单一托管模型发布后九天内的研究，本综述应被视为早期证据图谱而非对该模型类别的最终定论。
链接: https://arxiv.org/abs/2609.32160
作者: Lijuan Tang,Yuemeng Zheng
机构: Northeastern University(东北大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 25 pages, 8 tables. Corpus: 28 arXiv papers (19-24 September 2026); arXiv versions listed in Appendix A. Ancillary file: per-paper corpus sheet (versions and headline results)
Abstract:Typed decision models (TDMs) return probability distributions over caller-defined options without generating text. TypeSafe released Jev, a commercial typed decision model, on 15 September 2026, and a small body of evaluation and replication work appeared within days. We review 28 papers posted between 19 and 24 September and relate their findings to earlier work on label-probability classification, constrained decoding, reranking, calibration, and model cascades. In this early literature, the typed readout itself has not shown an independent accuracy advantage over comparable label-probability readouts. Jev’s clearest gains are in latency and cost, while accuracy gaps remain on harder tasks. In practical deployments, confidence is often used to decide when to defer to a stronger model or a human. We use recurring weaknesses in these studies to derive a 14-item evaluation checklist for future TDM work. Because the evidence covers only the first nine days after the release of one hosted model, the review should be read as an early evidence map rather than a settled assessment of the model class.
[NLP-325] Checking Leakage Witnesses versus Certifying Bounded Non-Leakage
【速读】： 该论文旨在解决在语言模型审计未发现泄露（leak）时，如何对非泄露性进行可信认证的问题。其核心挑战在于：在给定提示词（prompt）域和可执行的泄露判定准则及解码规则下，如何提供形式化保证以证明不存在泄露。解决方案的关键在于区分计算复杂性、审计覆盖率与执行条件之间的本质差异。研究发现，对于一般有界多项式时间评估器，泄露实例可多项式时间验证，但泄露存在性为NP完全问题，确定性认证属于coNP完全；而精确随机认证在任意固定有理数阈值（0,1）内为coNP^PP完全。通过限制计算模型，可降低复杂度：例如当所有随机性来自高效计算的有限概率表的最终采样时，认证可降至coNP；在注意力模型中，若局部依赖窗口长度为对数级且仅含一个全局头、采用确定性解码、固定词汇表、精确有理加权均值、直接二元仿射读出以及有限自动机提示域，则认证可在多项式时间内完成。进一步构造表明，使用两个全局层时，在模板提示域上认证变为coNP完全，前提是每层一个注意力头、多项式宽度、对数精度及逆多项式logit间隔。通过植入秘密实验对比完整参考结果与有限审计的差异，发现单次提示审计在4096提示域下，每对秘密-模型状态仅随机抽取256次评估时，平均有41.06%的泄漏被遗漏；批量与单次检查在48个微调模型对中存在一处决策分歧，而重复执行可复现相同输出，从而揭示了认证的计算条件与审计覆盖范围、执行方式之间的重要区别。
链接: https://arxiv.org/abs/2609.32134
作者: Chao Feng,Burkhard Stiller
机构: University of Zurich (苏黎世大学)
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL)
备注:
Abstract:When a language-model audit finds no leak, what is needed to certify non-leakage? We study guarantees over a declared prompt domain under an executable leakage criterion and decoding rule. For general bounded polynomial-time evaluators, a supplied leaking execution is polynomial-time checkable, while leak existence is \NP-complete and deterministic certification is \coNP-complete. Exact stochastic certification is \coNP^\PP -complete at every fixed rational cutoff in (0,1) . Restricting the computation can change these bounds. For example, certification is in \coNP\ when all randomness is a terminal draw from an efficiently computed finite probability table. Attention models admit polynomial-time certification when local dependency windows of logarithmic length precede one global head, given deterministic decoding, fixed vocabulary, exact rational weighted means, a direct binary affine readout and finite-automaton prompt domains. A construction with two global layers instead makes certification \coNP-complete over template domains, with one head per layer, polynomial width, logarithmic precision and an inverse-polynomial logit margin. Planted-secret experiments measure what finite audits miss relative to complete references. Among 30 secret–model-state pairs that leak under greedy single-prompt execution on their secret’s 4,096-prompt domain, uniformly selecting 256 recorded evaluations per pair misses every leak for an expected 41.06% of these pairs. Batched and single-prompt checks disagree on one complete-domain decision among all 48 fine-tuned pairs, while a same-order repeat reproduces every single-prompt output. These results distinguish computational conditions for certification from the coverage and execution conditions needed to interpret a negative audit.
[NLP-326] Using LMs to Model the Effects of Context and Coreference during Sentence Comprehension EMNLP2026
【速读】： 该论文旨在解决生成式语言模型（Generative Language Models, GLMs）在模拟人类语言处理时，如何平衡局部工作记忆限制与长程语篇结构依赖之间的矛盾问题。现有研究通过严格限制上下文窗口以模拟人类工作记忆容量，虽能较好捕捉局部记忆限制，但可能忽略了人类在理解过程中对长程句间指代关系等全局语篇结构的依赖。本文通过系统调节GPT-2在四个大规模自然语言阅读时间数据集上的上下文窗口大小，发现模型对人类心理语言学数据的拟合度呈现倒U型关系：中等扩展上下文（500–1000词元）表现最优。为揭示其机制，研究进一步设计反事实推理实验，在推理阶段将重复的语篇实体替换为代词以破坏跨句指代链，结果表明此类操作使大上下文窗口的预测能力下降20%至40%。这一发现表明，追踪长程核心指代关系是提升语言模型预测力与人类阅读行为一致性的重要因素，也揭示了人类理解者在语言处理中实际依赖全局语篇结构的程度。
链接: https://arxiv.org/abs/2609.32119
作者: Kohei Kajikawa,Lin Ai,Tatsuki Kuribayashi,Ethan Gotlieb Wilcox
机构: MBZUAI(中东人工智能大学); Tohoku University(东北大学); Georgetown University(乔治城大学)
类目: Computation and Language (cs.CL)
备注: EMNLP 2026
Abstract:Language models (LMs) are often used as a tool to model human language processing. Recent studies suggest that severely restricting LMs’ context window improves their fit to human psycholinguistic data by simulating human working memory constraints. However, it is possible that this strict memory-decay approach overlooks humans’ reliance on long-range structural representations, such as discourse structre. In this work, we systematically vary the context window size of GPT-2 across four large-scale naturalistic English reading-time datasets and observe a U-shaped relationship: Although restricted contexts ( 20 tokens) successfully capture local memory limitations, expanded contexts (500–1,000 tokens) ultimately yield the highest overall psycholinguistic fit. To investigate the mechanism driving this benefit, we conduct a counterfactual inference-time experiment that disrupts cross-sentential entity chains by pronominalizing repeated discourse entities. Obscuring these structural linkages significantly degrades the predictive power of larger context windows by 20% to 40%. Our experiments demonstrate that tracking long-range coreference relations is one important factor for the alignment between LM surprisal and human reading behavior, and approximate the extent to which human comprehenders use global discourse relations during language processing.
[NLP-327] mu-bench: A Multilingual Utterance Transcription Benchmark ICASSP-2027
【速读】： 该论文旨在解决语音助手在实际应用中因自动语音识别（ASR）系统评估标准与真实场景脱节而导致的性能误判问题。当前主流评估方法基于读写式英语语料库的词错误率（WER），仅关注字面匹配，无法反映语义层面的差异，导致对实际用户交互中关键信息（如姓名、邮箱、验证码等表单字段）的识别准确性评估失真。为此，论文提出mu-bench数据集，包含来自五种语言（英语、西班牙语、土耳其语、越南语、中文）共250通电话的4,270个用户语音片段，聚焦于表单字段输入任务；并引入话语错误率（UER），一种基于大语言模型（LLM）的语义保真度评估指标，经人工标注校准后，可有效衡量转录文本是否保留原意。同时，设计了LLM归一化器以统一不同厂商输出格式间的评估基准。在1,847条人工标注样本上，UER与人工评价者的一致性达κ = 0.78，显著优于传统精确匹配的WER（κ = 0.53）。研究通过公开排行榜对六家商业ASR服务进行排名，结果显示最优方案的UER为11.9%，且中文在所有语言中识别难度最高。该研究的关键在于构建面向真实交互场景的多语言语义评估框架，突破传统字面匹配的局限，实现更贴近用户体验的性能衡量。
链接: https://arxiv.org/abs/2609.32082
作者: Andrea Li(UC Berkeley),Soham Ray(Sierra AI)
机构: UC Berkeley(加州大学伯克利分校); Sierra AI
类目: Computation and Language (cs.CL)
备注: 5 pages, 7 tables. Dataset: this https URL . Code: this https URL . Leaderboard: this https URL
Abstract:Voice agents depend on accurate automatic speech recognition (ASR) to act on what callers say, yet ASR is evaluated on read, English-centric speech with word error rate (WER), which penalizes surface rather than semantic differences. We introduce mu-bench, a dataset of 4,270 caller utterances from 250 phone calls to an AI banking agent in English, Spanish, Turkish, Vietnamese, and Mandarin, centered on form-field inputs such as names, email addresses, and confirmation codes. We release Utterance Error Rate (UER), an LLM judge of whether a transcript preserves meaning, calibrated against human raters, together with an LLM normalizer that makes WER comparable across providers’ output formats. On 1,847 human-rated transcripts, UER agrees with annotators at \kappa = 0.78, versus 0.53 for exact-match WER on normalized text. We rank six commercial providers on a public leaderboard; the best reaches 11.9% UER, and Mandarin is hardest for all six.
[NLP-328] Quantization Thresholds Replicate Failure Modes Do Not: A Three-Model Study of Agent ic Tool Use in Polish from 8-bit to 2-bit
【速读】： 该论文旨在探究GGUF量化对波兰语环境下智能体（agentic）工具使用能力的影响，并检验这些影响在不同模型间的可泛化性。其核心问题是：随着量化精度从8位降至2位，模型在复杂任务中的推理与工具调用能力如何变化，以及这种退化是否在不同架构和预训练数据族之间保持一致。解决方案的关键在于构建并应用PolAgentBench——一个以波兰语提示（Polish prompts）与英语工具模式（English tool schemas）相结合的确定性基准测试集，包含67项主任务（含15个对抗性探测任务与52个高难度任务）及46项算术隔离阶梯任务。研究通过三个代表性模型（Bielik-11B-v3.0、其压缩蒸馏版Bielik-Minitron-7B-v3.0、以及基于不同预训练家族的Llama-PLLuM-8B）在六种量化精度（Q8_0至Q2_K）下的系统评估，揭示出在3比特向2比特过渡时存在显著的能力“悬崖式”崩溃，且失败模式不具一致性；同时发现算术任务中显式调用可显著提升性能，但部分结果受标注偏差与规则设定等人为因素影响，需通过严格校正处理。研究强调了量化对多语言智能体系统可靠性的关键影响，并公开了完整的基准、轨迹与版本化数据。
链接: https://arxiv.org/abs/2609.32042
作者: Jakub Prejzner
机构: Independent Researcher(独立研究员)
类目: Computation and Language (cs.CL)
备注: 34 pages, 1 figure, 21 tables. Code, tasks, trajectories and analysis scripts: this https URL
Abstract:We ask how GGUF quantization affects agentic tool use in Polish and whether the effects generalize across models. We introduce PolAgentBench, a deterministic benchmark with Polish prompts and English tool schemas: a 67-task main suite (15 adversarial probes, 52 hard-tier tasks) and a 46-task arithmetic isolation ladder. Three models span two axes of variation: Bielik-11B-v3.0 and its pruned, distilled child Bielik-Minitron-7B-v3.0 isolate model compression, and Llama-PLLuM-8B adds a change of pretraining family. Each is measured at six precisions, Q8_0 to Q2_K. Only the collapse threshold replicates. (1) All three models fall off a cliff between 3-bit and 2-bit (11B 0.716 to 0.045, 7B 0.463 to 0.149, PLLuM 0.224 to 0.015; paired McNemar p 0.001 in each), across a fourfold capability spread and both axes. (2) Failure modes do not replicate: at 2-bit the 7B fails long (median 9.1k tokens, 4 steps) while the 11B mostly answers at the first step with a confabulated final answer (37 of 64 failures); PLLuM fails on content across precisions (71.8-92.0% of steps parse). (3) On the arithmetic ladder the unscaffolded rung is a floor, left standing by a rerun that states the no-tool rule; four explicit calls lift the 8-bit 11B from 1/10 to 9/10 and the 7B from 0/10 to 7/10 after format-only failures with the gold value are forgiven, an exploratory effect with eight distinct baseline inputs that does not survive multiplicity correction, while the order-trap arm separates the models at 8-bit (11B 6/6, 7B 0/6). (4) The Polish-versus-English gap is associated with degradation or with task family. We document four artifacts that shaped our conclusions (rounding-hostile gold values, strict answer typing, a no-tool rule the prompt never stated, priority-ordered failure labels), report affected results in strict and corrected form, and release the benchmark, trajectories and commit-stamped artifacts.
[NLP-329] Who Governs Data in the AI Era? A Computational Analysis of the U.S. Privacy Workforce in Job Postings
【速读】： 该论文旨在解决当前隐私保护领域中雇主对隐私岗位职责定义不明确的问题，尤其在生成式 AI（Generative AI）快速发展的背景下，隐私专业角色的内涵与边界亟待厘清。研究通过分析1,143份来自LinkedIn和Indeed的美国隐私类职位招聘信息，结合规则驱动的文本挖掘与BERTopic主题建模技术，识别出18个潜在主题并归纳为四类核心范畴。其关键解决方案在于揭示了隐私岗位本质上是跨领域的复合型角色，融合法律知识、技术能力与人际沟通素养；同时发现人工智能相关语言已广泛渗透于合规、法律、治理与安全等维度，表明AI治理职责多被嵌入现有隐私岗位之中，推动了兼具隐私与AI治理职能的混合型岗位兴起，同时也催生了专职的AI治理职位。这一发现为隐私人才的培养、岗位设计及组织架构优化提供了实证依据。
链接: https://arxiv.org/abs/2609.32030
作者: Ramazan Yener,Muhammad Hassan,Masooda Bashir
机构: 未知
类目: Computers and Society (cs.CY); Computation and Language (cs.CL); Cryptography and Security (cs.CR)
备注: 42 pages, 8 figures, 4 tables
Abstract:Privacy protection now spans legal, technical, and managerial duties, and demand for privacy professionals is growing across sectors. However, little is known about how employers define these roles. We analyze 1,143 U.S. privacy job postings from LinkedIn and Indeed. We examine job titles, salaries, competencies, certifications, education, experience, regulatory references, and AI-related language by using rule-based text mining. We also apply Topic Modeling (BERTopic) to the same postings and identify 18 latent themes which we grouped them into four categories. Our findings show that privacy roles are hybrid and they combine legal knowledge, technical skills, and interpersonal competence. Artificial intelligence appears in more than half of postings, with AI language spread across compliance, legal, governance and security themes. Our research indicates that AI governance responsibilities are often embedded within existing privacy roles, contributing to the rise of hybrid positions alongside dedicated AI governance roles.
[NLP-330] Before the Rollout Ends: Early Terminal Reward Prediction for Long-horizon Coding Agents
【速读】： 该论文旨在解决长时序编码智能体（long-horizon coding agents）在执行复杂任务时面临的高推理成本、早期错误假设被放大的问题，以及由于终端奖励稀疏导致的训练不稳定问题。其核心挑战在于：智能体需完成大量工具调用序列才能获得可验证的奖励，这不仅增加了计算开销，还使得学习过程难以有效收敛。为应对这一问题，论文提出上下文感知的早期奖励（Contextual Early Reward, CER）机制，其关键在于通过轨迹前缀中的行为证据，动态预测最终奖励。CER利用来自相关历史任务的经验总结，生成针对当前任务和阶段自适应的评估标准，从而实现密集且可解释的中间奖励信号。实验表明，在SWE-bench Verified测试中，CER在Nemotron 3 Ultra和Qwen 3.6 27B上分别将RM@8指标提升4.2和2.0个百分点；在相同性能下，仅需15.3%的词元数即可达成最优基线表现；在强化学习训练中，相较完整回溯的TMax方法，CER在减少52.7%在线策略与评判器词元消耗的同时，仍取得1.9个百分点的性能提升。因此，CER提供了一种高效、可解释且密集的评估范式，显著改善了长时序任务中智能体的学习效率与稳定性。
链接: https://arxiv.org/abs/2609.31995
作者: Jihan Yao,Sihan Zeng,Shangbin Feng,Zhiyuan Fan,Banghua Zhu,Yulia Tsvetkov
机构: University of Washington(华盛顿大学); HKUST(香港科技大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Long-horizon coding agents receive verifiable rewards only after completing expensive sequences of tool calls. This increases inference cost, amplifies early wrong hypotheses, and can lead to sparse terminal reward and unstable training. We introduce Contextual Early Reward (CER), which predicts terminal reward through behavioral evidence in a trajectory prefix. CER synthesizes adaptive rubrics specific to the current task and stage through experiences summarized from related historical tasks. In test-time scaling on SWE-bench Verified, CER improves RM@8 over the strongest baseline by 4.2 percentage points (pp) on Nemotron 3 Ultra and 2.0 pp on Qwen 3.6 27B; on Nemotron, it takes only 15.3% tokens to match the best baseline performance. In RL training experiments, CER exceeds full-rollout TMax by 1.9 pp while using 52.7% fewer online policy-and-judge tokens. Together, CER provides an interpretable, efficient, and dense evaluation method for long-horizon coding agents.
[NLP-331] oward Embedding-Based Psychometrics: Structural Modeling of Assessment-Item Semantics With Contextual Scores
【速读】： 该论文旨在解决教育测评中评分项目（assessment items）的语义结构建模问题，特别是如何通过外部语料库中参考词的相似性来表征评分项目的上下文得分（contextual scores），并探究其潜在的多维度结构。研究聚焦于国际数学素养评估（TIMSS）40个数学测验单元的评分数据，采用部分指定的两步因子分析法识别其内在语义结构。关键发现是存在一个稳定的七组结构，且在不同模型设定下，普遍支持一个通用维度与若干群组关联并存的结构形式；然而，具体项目归属对模型设定较为敏感。研究进一步表明，基于锚定参照的语义表示虽优于单一因子模型，但未能达到最低贝叶斯信息准则（BIC），提示其仍存在局限性。通过对比三种初始Q矩阵构造及其Hull-PVAF修正版本，在高阶和饱和属性分布下，官方内容框架在BIC上表现最优，而其四因子扩展版本在AIC上更优，但所有条件中均以匹配的单维双参数逻辑模型（unidimensional two-parameter logistic model）在AIC和BIC上表现最佳。因此，研究结论强调了条件性语义表示的有效性，同时限制了直接诊断解释的适用性，并提出将文本辅助响应校准（text-assisted response calibration）作为未来应用方向，需依赖更大规模的校准题库和独立验证。
链接: https://arxiv.org/abs/2609.31976
作者: Jinsong Chen,Shi-Ting Chen
机构: The University of Hong Kong (香港大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Contextual scores represent assessment items through their similarities to reference words in an external corpus. We examine the semantic structure of scores for 40 TIMSS mathematics scored units using a partially specified two-step factor procedure. A search across factor counts identifies a persistent seven-group structure under the featured construction. Subsequent comparisons consistently favor a general dimension alongside group associations, although individual group memberships remain sensitive to some specification choices. Item examples distinguish recurring, cross-domain, sensitive, and imposed associations. Simpler and unrestricted references clarify the contribution and limits of the anchored representation: it improves on a single factor but does not achieve the lowest working Bayesian information criterion (BIC). A separate response benchmark compares three initial Q constructions and their Hull-PVAF revisions under higher-order and saturated attribute distributions. Among these diagnostic models, BIC favors the official content framework and the Akaike information criterion (AIC) favors its direct four-factor augmentation, but a matched unidimensional two-parameter logistic model has lower AIC and BIC than all twelve conditions. These findings support a conditional semantic representation while limiting direct diagnostic interpretation. We discuss learned text-assisted response calibration as a prospective application requiring a larger calibrated item bank and independent evaluation.
[NLP-332] Extraction of clinical findings from mammography and breast ultrasound reports: a comparison between specialists and Artificial Intelligence
【速读】： 该论文旨在解决巴西乳腺癌筛查中影像报告（乳腺钼靶和超声）临床发现提取效率低、人工提取易出错的问题，进而影响早期诊断的及时性。其核心解决方案是利用提示工程（Prompt Engineering）结合少样本学习策略，通过大型语言模型（LLM）——Gemini 2.5 Flash实现对巴西葡萄牙语医学报告中的命名实体识别（Named Entity Recognition, NER）。关键在于采用优化的提示设计使模型在有限标注样本下达到高精度，实验结果显示该模型在宏观F1（Macro F1: 0.91）和微观F1（Micro F1: 0.98）上均优于人工提取团队（0.72），且主观验证一致性达93.1%，并在4份报告中成功识别出人工遗漏或错误记录的信息，证明了生成式AI在医疗文本信息提取中具备超越人类专家的潜力，可作为专业医生的互补性核查工具。
链接: https://arxiv.org/abs/2609.31974
作者: Lorenzo Farias,Hanna Reckziegel,Daniela Duarte da Silva Bagatini,Daniel Schulz,Gabriela de Andrade Monteiro,Letícia Zanatta,Ana Laura Brill Thum,Priscila Schmidt Lora,Débora Oliveira da Silva,Ana Paula Wernz da Cunha Müller,Cristiane Drebes Pedron
机构: Universidade de Santa Cruz do Sul (UNISC)(圣克鲁斯南大利亚大学); Universidade do Vale do Rio dos Sinos (UNISINOS)(南里奥格兰德州立大学); Universidade Federal do Rio Grande (FURG)(里奥格兰德州立大学); Clínica de Mastologia Dra. Ana Paula Muller Ltda.(安娜·保罗拉·穆勒乳腺科诊所); Universidade Nove de Julho (UNINOVE)(九月七日大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 16 pages, 10 figures, 3 tables. Submitted to Artificial Intelligence in Medicine
Abstract:Breast cancer is the leading cause of cancer-related death among women in Brazil, and the time between the request and the release of mammography reports directly influences adherence to screening, making the agility in processing these reports a critical factor for early diagnosis. In this context, this study compares the performance of a Large Language Model (LLM) with manual extraction performed by a team of health researchers in identifying clinical findings from mammography and breast ultrasound reports written in Brazilian Portuguese. Named Entity Recognition (NER) was applied through Prompt Engineering using a few-shot strategy, employing the Gemini 2.5 Flash model, selected from preliminary exploratory tests with four candidate models. The Gemini 2.5 Flash model demonstrated the best performance, achieving a Macro F1 of 0.91 and a Micro F1 of 0.98. The subjective validation, in which 29 exams of different formats were evaluated by health researchers using a Likert scale, yielded an agreement index of 93.1%. The model outperformed human extraction in overall Macro F1 (0.91 vs. 0.72), as in four reports the model correctly identified information that had been omitted or incorrectly recorded during manual extraction, demonstrating its potential as a complementary verification tool alongside specialists. The results confirm the hypothesis that LLMs, when instructed through Prompt Engineering, can achieve performance comparable to or superior to manual extraction by health professionals.
[NLP-333] IndicFDB: Benchmarking Full-Duplex Voice Agents across Indian Languages
【速读】： 该论文旨在解决多语言环境下全双工语音代理（full-duplex voice agents）在实际对话交互中面临的挑战，包括自然停顿处理、发言权切换、反馈性回应（backchanneling）以及对用户打断的实时响应等问题。现有评估基准Full-Duplex-Bench因仅支持英语且依赖于词级时间戳自动语音识别（ASR）和英文提示的大语言模型（LLM）评判，难以推广至印度多语言场景。为此，本文提出IndicFDB，一个覆盖印度10种主要语言、包含12,350个样本的多语言全双工对话评估数据集，规模接近原数据集的17倍。其解决方案的关键在于：首先利用语音活动检测（VAD）从约5万小时的分离声道对话中挖掘停顿、发言权切换与反馈行为样本，并构建经人工验证的合成用户打断样本；其次通过语言无关的VAD启发式方法评估对话事件的时间准确性，克服了缺乏可靠词级对齐的问题；最后采用开源的转写与翻译流水线将多语言响应统一转换为英文，交由大语言模型进行相关性和质量评分。实验结果表明，商用API在不同语言间表现出意外的一致性行为，但普遍存在响应速度与抗停顿鲁棒性之间的权衡，而单语开源全双工模型则揭示了反馈行为、响应质量与延迟之间的进一步权衡。
链接: https://arxiv.org/abs/2609.31967
作者: Rajarshi Roy,Shobhit Banga,Jonathan Raiman,Supriya Paul,Bhaskar Singh,Manmeet Kaur,Sagar Jain,Hanuman Sidh,Pranav Sharma,Aditya Singh,Aaditya Pareek,Manas Dhir,Adi Margolin,Niket Agarwal,Bryan Catanzaro
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Full-duplex voice agents must handle pauses, take turns, backchannel, and respond to user interruptions in real time. Full-Duplex-Bench evaluates these behaviors, but its English-only corpus and reliance on word-timestamped ASR and an English-prompted LLM judge make it difficult to extend to Indian languages. We introduce IndicFDB, which extends it to ten languages spoken in India with 12,350 samples, nearly 17 times as many as the original. We address three challenges: finding conversational events in multilingual speech, evaluating their timing without reliable word-level alignment, and judging responses across languages. We mine pause handling, turn taking, and backchanneling samples from roughly 50,000 hours of channel-separated conversations using voice activity detection (VAD), and construct human-validated synthetic user interruption samples. Language-independent VAD heuristics evaluate timing, while an open-weight transcription and translation pipeline converts responses to English for LLM ratings of relevance and quality. Across seven voice agents, commercial APIs show unexpectedly consistent behavior across languages but are either fast or robust to pauses, never both, while monolingual open full-duplex models expose further tradeoffs among backchanneling, response quality, and latency.
[NLP-334] ransformer MLP Gate Thresholds Are Couplings to a Carried Reference Direction
【速读】： 该论文旨在解决生成式 AI 模型中神经网络门控单元（gate）激活阈值的实现机制问题，特别是揭示在前馈层（MLP）中门控单元的静息抑制（resting inhibition）如何被系统性地调控。传统上，模型残差流（residual stream）的均值方向常被视为冗余成分而被均值中心化去除，但本文提出该方向实为具有功能意义的关键组件：它是门控单元设定其工作点（operating point）的参考基准。研究发现，在 Phi-2 模型中，约 99.9% 的门控单元在中间堆叠层的静息抑制主要由残差流与一个特定方向 $ w \cdot b_L $ 的耦合所承载，其强度是显式偏置参数的 48–56 倍，且该耦合具有方向特异性——随机方向在匹配范数下仅能产生不超过 0.09 的斯皮尔曼等级相关系数（Spearman ρ），而参考方向则达到 0.95，接近自洽（tautological）。因果实验表明，移除残差流在该参考方向上的投影会使高于阈值的激活量增加约 9 倍，呈剂量依赖关系，显著超过范数匹配的对照组（23–59 倍），而随机初始化的孪生模型无此效应；替换测试进一步证明该方向本身携带功能，而非其幅值。此外，沿参考方向注入噪声所需能量是随机方向的 8–42 倍，凸显其功能性。该分解模式在四种不同架构家族（包括带显式偏置的 GELU 与无偏置 SwiGLU）中复现，并在其中两个实现剂量响应。跨八模型扫描显示，该机制存在于所有 GELU 与 SiLU 类型模型中，仅在 OPT 模型中缺失，因其层归一化（LayerNorm）偏置存在相反作用而抵消了该参考方向。综上，门控阈值本质上是网络对输入分布构建的恒定参考量的耦合，而显式偏置参数贡献甚微；该参考量在残差流中的传递程度与参数中的保留程度取决于具体架构设计。
链接: https://arxiv.org/abs/2609.31956
作者: Olli Tuomi
机构: Evident Solutions Oy
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 18 pages, 4 tables. Code and data: this https URL . An earlier version of this paper is archived on Zenodo, doi: https://doi.org/10.5281/zenodo.21498411%3B this version supersedes it and corrects several reported values
Abstract:The corpus-mean direction of a transformer’s residual stream is a component shared across all inputs, and is commonly removed by mean-centering before representational analysis. We present evidence that it is a functional component: the reference against which the MLP gate population sets its operating point. In Phi-2, an exact decomposition of resting gate pre-activations shows that at mid-stack layers the resting inhibition of 99.9% of gates is carried by the coupling w \cdot b_L to the carried mean direction, at 48-56 \times the explicit bias parameter, and the coupling is direction-specific: a random direction at matched norm orders the population’s firing rates at Spearman \rho \leq 0.09 where the reference reaches 0.95. (That 0.95 is near-tautological on its own; the paper derives its null.) Causally, removing the stream’s projection on the reference multiplies above-threshold firing by about 9 \times , dose-monotonically, at 23-59 \times a norm-matched control; a random-initialised twin is flat, and replacement tests show that the direction carries the function and the magnitude does not. A direction-matched control makes the same point: noise injected along the reference costs 8-42 \times the same energy along a random direction. The decomposition replicates on four further families spanning both gate types (GELU with an explicit gate bias, bias-free SwiGLU), and the dose-response on two of them. Across an eight-model scan the mechanism is present in every GELU and SiLU family and absent only in OPT, where an opposing LayerNorm bias cancels the carried reference. Gate thresholds are implemented as couplings to a constant the network builds for the distribution it is reading, with the bias parameters contributing little; how much of that constant is carried in the stream and how much in parameters depends on the architecture.
[NLP-335] Improving Medical Calculation of LLM s with Embedded Coding
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在医疗计算任务中可靠性不足的问题，尤其是在需要精确数值输出的临床场景下，如药物剂量计算、器官功能评估和预后评分等高风险决策，微小误差可能带来严重临床后果。现有模型虽在医学问答与考试基准上表现良好，但在涉及复杂算术运算的任务中易出现错误。其核心解决方案是提出MedCode框架，通过训练LLMs生成嵌入式可执行代码（embedded executable code），使模型能够根据临床情境识别相关计算器、提取输入变量，并生成脚本将算术运算交由确定性解释器执行。该方法不仅提升了计算结果的准确性，还输出解释与单位信息，增强可解释性。为支持训练，研究构建了基于MedCalc基准的监督微调（SFT）与偏好数据集，并额外采集了重症监护室（Intensive Care Unit, ICU）场景下的计算任务数据集；同时提出加权直接偏好优化（weighted Direct Preference Optimization, wDPO），自适应强化对模型难以区分的偏好样本的学习。在LLaMA3-8B、Qwen2.5-7B和Mistral-7B上的实验表明，该方法实现了20–30个百分点的绝对准确率提升，验证了嵌入式代码生成在医疗计算中的有效性。
链接: https://arxiv.org/abs/2609.31908
作者: Tianshi Ming,Yingying Zhang,Xian Wu
机构: Carnegie Mellon University (卡内基梅隆大学); Tencent YouTu Lab (腾讯优图实验室)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Large Language Models (LLMs) perform well on medical examinations and question-answering benchmarks, but remain unreliable on medical calculation tasks that require exact numerical outputs. These calculations support high-stakes decisions such as medication dosing, organ-function assessment, and prognostic scoring, for which even small errors can have serious clinical consequences. We introduce MedCode, a framework that improves medical calculation by training LLMs to generate embedded executable code. Given a clinical context, the model identifies the relevant calculator, extracts its input variables, and produces a script that delegates arithmetic operations to a deterministic interpreter. Executing the script returns the calculated value together with an explanation and the appropriate unit. We construct supervised fine-tuning (SFT) and preference datasets from the MedCalc benchmark and additionally curate a dataset for calculation tasks in Intensive Care Unit (ICU) scenarios. We further propose weighted Direct Preference Optimization (wDPO), which adaptively emphasizes preference pairs that are difficult for the model to distinguish. Experiments with LLaMA3-8B, Qwen2.5-7B, and Mistral-7B show absolute accuracy gains of 20–30 percentage points, demonstrating the effectiveness of embedded code generation for medical calculation.
[NLP-336] NVAlign: Direct-Gradient Optimization for Non-Verbal Control in Continuous Autoregressive Flow Matching Text-to-Speech ICASSP2027
【速读】： 该论文旨在解决连续自回归流匹配（continuous autoregressive flow-matching）文本到语音（TTS）系统中非言语发声（Non-Verbal Vocalization, NVV）标签控制精度不足的问题，尤其缺乏有效的后训练阶段非言语事件调控方法。其核心解决方案是提出NVAlign——一种基于直接梯度的后训练框架，通过监督微调（SFT）预训练TTS模型与一个面向非言语发声的自动语音识别（NV-ASR）模型，并将冻结的NV-ASR作为奖励模型，利用两步梯度代理机制实现奖励信号在流匹配采样器中的高效反向传播，从而联合优化自回归主干网络与声学流头。同时引入保真度惩罚项与参考速度正则化，以维持说话人相似性与语音质量。实验结果表明，相较于监督微调（SFT）与Flow-GRPO基线方法，NVAlign显著提升了标签跟随准确率，在NVV-SuperBench及人工听感评估中均表现出更优性能，验证了直接奖励梯度优化在提升连续自回归流匹配TTS系统中非言语控制能力方面的有效性。
链接: https://arxiv.org/abs/2609.31892
作者: Qiaolin Wang,Pedro Sandoval-Segura,Anunaya Joshi,Edvardas Jurkonis,Jake Downie
机构: 未知
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
备注: 5 pages, 1 figure, 2 tables. Submitted to ICASSP 2027. Audio samples: this https URL
Abstract:While modern text-to-speech (TTS) systems generate highly natural speech and support inline non-verbal vocalization (NVV) tags, accurate control over these events remains challenging. A key gap is the lack of established post-training methods for non-verbal control in continuous autoregressive flow-matching TTS. To this end, we present NVAlign, a direct-gradient post-training framework for NVV tag-following in this architecture. We first perform supervised fine-tuning (SFT) of TTS models and an NVV-aware automatic speech recognition (NV-ASR) model on NVV-annotated speech, then freeze the NV-ASR model to serve as the reward model for post-training. A two-step gradient surrogate enables efficient reward backpropagation through the flow-matching sampler to jointly update the autoregressive backbone and acoustic flow head. Fidelity penalties and reference-velocity regularization help preserve speaker similarity and speech quality. Results from NVV-SuperBench and human listening evaluations show that NVAlign improves tag-following accuracy over SFT and Flow-GRPO baselines. These findings demonstrate that direct reward-gradient optimization can improve non-verbal control in continuous autoregressive flow-matching TTS. Audio samples are available at this https URL.
[NLP-337] Agents Can Use Base Models to Evade AI Detection
【速读】： 该论文旨在解决生成式文本中规避检测的问题，即如何使基于基础语言模型（base language model, LLM）的输出能够有效逃避现有的AI文本检测机制。传统“人类化”（humanization）方法依赖多轮迭代改写以降低检测率，但易引发语义漂移（semantic drift），影响内容质量与任务准确性。本文提出的关键解决方案是：部署具备编码能力的智能体（coding agent），直接通过拼接（stitching）基础模型生成的多个文本片段来协同构建最终输出，从而实现高保真、任务特定且连贯的文本生成。实验表明，该方法在不显著牺牲任务性能的前提下，显著降低了主流后置检测器（如Pangram v4）的识别率（从77%降至24%），并大幅削弱了预设软水印（soft watermarking）的检测有效性（模拟检测率降至10%）。尽管该攻击策略有效，但其代价是输入与输出令牌数显著增加，导致单次查询成本最高提升至30倍（按API定价）。本研究揭示了一类新型对抗性攻击对当前AI文本检测体系的威胁，并呼吁后置检测系统需将基础模型输出纳入训练数据以增强鲁棒性。
链接: https://arxiv.org/abs/2609.31876
作者: Bhuwan Dhingra,Danish Pruthi
机构: Duke University (杜克大学); Indian Institute of Science (印度科学研究所)
类目: Computation and Language (cs.CL)
备注:
Abstract:We show that coding agents equipped with a base language model can successfully assemble responses from its samples to evade detection. Base models have been shown to evade commercial detectors, however, prior “humanization” techniques rely on using these models to paraphrase AI outputs over several iterations, which invariably results in semantic drift. In contrast, equipping coding agents to directly orchestrate the writing process by stitching text samples from a base model allows it to produce outputs that are coherent, task-specific and generally high quality. We find that Claude Opus 5 operating in a Claude Code harness effectively orchestrates a local 32B parameter OLMo-2 base LM and sacrifices little task accuracy across benchmarks spanning creative writing, factual grounding, health QA and instruction following, while using up to 90% base LM tokens. Responses constructed in this manner reduce the effectiveness of both post-hoc detectors (Pangram v4 detection rate drops from 77% to 24%) and soft watermarking applied a priori to the agent’s generations (down to a simulated 10% detection at low FPR). While effective, this evasion requires a significantly larger number of input and output tokens from the agent, increasing the dollar cost per query up to 30x at API-pricing. Overall, this work demonstrates the effectiveness of a new class of adversarial attacks against AI text detection, and urges post-hoc detection providers to include outputs of base models in their training.
[NLP-338] CueKFS: Agent ic Cue-Driven Keyframe Selection for Long Video Understanding
【速读】： 该论文旨在解决长视频问答中关键帧选择（Keyframe Selection, KFS）的局限性，即传统方法在处理复合语义或隐含信息的问题时，因依赖静态上下文和单一帧与问题的相似性匹配而难以准确识别相关帧。其核心解决方案是提出一种无需训练的新型方法CueKFS，将问题-帧匹配重构为将视频帧与一组动态生成的视觉线索（visual cues）进行对比的过程。该方法从初始显著帧出发，将问题分解为多个可执行的线索，并通过一个推理型视觉语言模型（VLM）作为代理（agent），基于视频证据对线索集进行迭代式修正，实现对视频的主动重探索。最终，根据留存线索的置信度分配选择预算。在三个基准测试中，CueKFS在全部27个评估设置下均达到当前最优性能，平均提升达+4.54%，且仅需平均两次VLM调用。行为分析表明，代理式的线索精炼机制显著提升了对视频内容的主动探索能力，使相对相似度最高提升92%。
链接: https://arxiv.org/abs/2609.31873
作者: Weitai Kang,Hanieh Deilamsalehy,Yumo Xu,Dewang Sultania,Serdar Cellat,Yan Yan
机构: University of Illinois Chicago(伊利诺伊大学芝加哥分校); Netflix(奈飞)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 9 main pages
Abstract:Keyframe selection (KFS) has long produced compact video summaries for browsing and retrieval, and representative frames for thumbnails. More recently, when conditioned on a question, KFS provides an alternative to uniform sampling for long-video question answering by selecting frames that are more relevant to the question. Most methods rank frames by similarity to the question. Yet a relevant frame may score poorly when the question combines subjects or moments that no single frame shows, or requires implicit information absent from its wording. Other methods try to break down the question into subqueries, but suffer from inaccurate decomposition due to their static initial context. Therefore, we propose CueKFS, a training-free method that reformulates question–frame matching as comparing frames against a set of dynamically generated visual cues. From an initial set of salient frames, we decompose the question into cues. Each cue concurrently probes the video to navigate to its own evidence. A reasoning VLM then agentically revises the cue set against its evidence to re-explore the video. CueKFS then allocates the budget across the surviving cues. Across three benchmarks, CueKFS establishes state-of-the-art results in all 27 evaluated settings with available prior results, achieving budget-averaged gains of up to +4.54% over the previous baseline and a median of only two VLM calls. We further provide a detailed behavioral analysis of CueKFS, showing that agentic cue refinement drives active re-exploration of the video, yielding relative similarity gains of up to 92% over the initial context.
[NLP-339] IndustryLLM : Failure-Driven LLM Training for Industrial Procurement
【速读】： 该论文旨在解决工业采购场景中语言模型面临的三大核心挑战：非正式买家术语与专业工程标准之间的语义鸿沟、市场信息属性稀疏性，以及在严苛安全容差下的事实准确性要求。针对这些问题，其解决方案的关键在于提出一种“故障驱动的适应性训练范式”，通过持续预训练（CPT）与监督微调（SFT）相结合的方式实现模型能力的精准提升。其中，CPT阶段构建了一个约100B tokens的高质量领域语料库，涵盖国家标准（如GB/T）、技术档案、去标识化的实际交易与询价记录及通用回放数据，并通过多源域重写（跨10类文本类型与8种写作风格）、置信度路由的最小化事实修正以及错误导向的问答合成等方法，系统性重构了约20B tokens的领域子集，显著缓解了注册差异（register mismatch）与事实脆弱性问题。此外，为保障下游部署中的可靠性，提出了一种证据门控的约束评估接口，采用三值逻辑处理未验证的产品证据，避免错误满足。实验结果表明，该方案在离线评估中显著提升了采购查询结构化准确率（精确匹配提升2.97个百分点，95%置信区间[2.11, 3.86]），在线A/B测试则实现GMV增长4.25%、满意询盘率提升8.3%，同时推理延迟从6–7秒降至1.5秒，充分验证了该方法的有效性与实用性。
链接: https://arxiv.org/abs/2609.31871
作者: Liang Ding(Project Lead),Zhiang Xu,Yuyang Sheng,Bin Chen,Songlin Bai,Run Zhu,Dingjun Wu,Hui Xu,Yandi Wang,Fulin Shi,Leilei Gan,Linlin Yu,Qihuang Zhong,Keqin Peng,Yalong Li,Chengfu Huo
机构: Alibaba(阿里巴巴)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Technical Report, 56 pages, 6 figures. Model weights and configs available at this https URL
Abstract:Industrial procurement requires language models to bridge informal buyer jargon, sparse marketplace attributes, and authoritative engineering standards under strict safety tolerances. We present IndustryLLM, an open-weight industrial language model trained from Qwen3.5-35B-A3B-Base (35B total parameters with ~3B activated per token, with the vision encoder frozen). Rather than relying on generic text scaling, we introduce a failure-driven adaptation recipe spanning continued pre-training (CPT) and supervised fine-tuning (SFT). CPT leverages a curated ~100B-token corpus integrating 5B tokens of national standards (e.g., GB/T) and technical archives, 10B tokens of de-identified real-world industrial transaction and inquiry records, and 60B tokens of general replay. To overcome register mismatch and factual brittleness, we systematically reconstruct an estimated 20B-token domain subset via multi-register rewriting across 10 genres and 8 writing styles, confidence-routed minimal factual editing, and error-targeted QA synthesis (resolving colloquial typos like ‘42-luo-mu’ - 42CrMo, expanding ambiguous codes like ‘16674’ - GB/T 16674, and clarifying conflicting dimensional specs). For downstream deployment, we formalize an evidence-gated constraint-evaluation interface enforcing three-valued logic where unverified product evidence remains unknown rather than satisfied. Offline evaluations demonstrate consistent gains on procurement-query structuring (+2.97 percentage points in exact match, 95% CI [2.11, 3.86] in No-Think mode), while randomized online A/B experiments in production yield substantial improvements (+4.25% GMV, +8.3% satisfied inquiries) alongside a latency reduction from 6-7 s to 1.5 s. Model weights and configs are released at this https URL.
[NLP-340] Omni-IO Skills: Harnessing Your Agent Omni-Native
【速读】： 该论文旨在解决通用智能体（general-purpose agents）在多模态生成能力上存在的碎片化问题，即其在文本、图像、音频、视频、文档、3D资产和代码等多种模态下的生产能力分散且难以协同。现有方法要么依赖昂贵的模型更新以扩展模态支持，要么通过集成专用模型与工具链，但缺乏对流程、依赖关系、中间产物及跨轮次修改的有效协调机制。为此，论文提出Omni-IO Skills——一种即插即用的智能体封装框架（Agent Harness），其核心在于通过分层技能（hierarchical Skills）、标准化多模态执行接口、依赖感知编排（dependency-aware orchestration）以及持久化的资产注册表（persistent Asset Registry），实现现有智能体的全模态原生（omni-native）能力升级。多资产工作流被建模为声明式执行图（Declare Execution Graphs），支持独立操作的并行调度，并将成功输出持久化以供下游任务或跨轮次复用。该框架包含27项技能，覆盖7种资产模态和4类核心能力（理解、生成、推理、检索）。在UniM-90基准测试中，该框架使GPT-5.6 Sol和Claude Sonnet 5的输入支持率从40.00%和38.89%提升至100%，语义-质量耦合得分分别从26.99/27.82提升至74.94/77.78，严格结构得分达到100.00和99.78，验证了在不修改宿主智能体推理核心的前提下，通过框架级能力组合实现广泛、可演进的全模态系统（Omni system）的可行性。
链接: https://arxiv.org/abs/2609.31847
作者: Yanlin Li,Mingyang Hao,Shengqiong Wu,Hao Fei,Mong-Li Lee,Wynne Hsu
机构: National University of Singapore(新加坡国立大学); University of Oxford(牛津大学)
类目: Computation and Language (cs.CL)
备注: 28 pages, 11 figures, 18 tables. Project page: this https URL
Abstract:General-purpose agents can plan, reason, and act over long horizons, yet their production capabilities remain fragmented across text, images, audio, video, documents, 3D assets, and code. Extending a foundation model to additional modalities ties capability growth to costly model updates, while assembling specialist models and tools leaves unresolved how procedures, dependencies, intermediate assets, and cross-turn revisions should be coordinated. We present Omni-IO Skills, a plug-and-play Agent Harness that makes existing agents omni-native through hierarchical Skills, a standardized multimodal execution interface, dependency-aware orchestration, and a persistent Asset Registry. Multi-asset workflows are represented as Declare Execution Graphs, which schedule independent operations concurrently and register successful outputs for downstream and cross-turn reuse across replaceable execution backends. Its 27 Skills cover 38 representative tasks spanning seven artifact modalities and four capability families: understanding, generation, reasoning, and retrieval. On UniM-90, the harness raises the input-support rates of GPT-5.6 Sol and Claude Sonnet 5 from 40.00% and 38.89% to 100%, while increasing relative Semantic–Quality Coupled Score from 26.99 to 74.94 and from 27.82 to 77.78, respectively; Strict Structure Score reaches 100.00 and 99.78. These results establish harness-level capability composition as a practical route to broad, evolvable Omni systems without changing the host agent’s reasoning core.
[NLP-341] Robot Manipulation with GPT -6-Astra: Body Knowledge Experience Reuse Emergent Skills and Sim2Real Transfer
【速读】： 该论文旨在解决通用多模态智能体在执行机器人控制任务时因反复探索和依赖模型中介的动作选择而导致的执行效率低下问题。其核心解决方案在于引入外部身体知识（external body knowledge）、成功经验（successful experience）以及可执行技能（executable skills），以显著提升基于GPT-6-Astra的XLeRobot在模拟与真实电梯按键任务中的执行效率。关键创新点包括：利用完整的机器人几何结构与摄像头信息，使平均完成时间相对于仅依赖通用控制接口的基线减少57.4%；通过同步动作与状态记录的图像数据，无需额外身体资产即可实现68.6%的性能提升；在起始位置偏移10–100 cm的九组配对实验中，原始起点的经验可使平均耗时降低58–63%，验证了经验的泛化能力；此外，研究发现代理自发生成的简短视觉反馈程序经研究人员重构后，在27次仿真试验中将局部任务平均时间进一步降低29–31%；最后，在12次真实机器人试验中，通过操作员确认的按钮接触验证了“模拟到现实”（sim2real）的可复用性，共享起始点下仿真资产与经验分别使平均时间减少53.0%和49.9%，且真实经验亦可迁移至新起点。这些结果表明，通过提供机器可读的身体描述、同步示范数据，并将代理生成的有效反馈程序转化为可复用技能，结合当前视觉输入动态调整动作，是构建高效通用操作实验的关键路径。
链接: https://arxiv.org/abs/2609.31770
作者: Sida He,Lingxi Xie,Yunning Cao,Pengfei Chen,Kaiwen Duan,Jiannan Ge,Xinyue Huo,Jiacheng Shao,Qi Tian
机构: Huawei Inc.(华为公司)
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 23 pages, 10 figures, 6 tables. Code, data, prompts, and skills: this https URL
Abstract:General-purpose multimodal agents can write robot-control programs, but repeated exploration and model-mediated action selection can make execution slow. We study how external body knowledge, successful experience, and executable skills improve an XLeRobot controlled by GPT-6-Astra in a simulated and a physical elevator-button task. In 30 fixed-start simulation trials, complete robot geometry and camera information reduce mean completion time by 57.4% relative to a baseline with only the common control interface and no prior experience; images with synchronized action and state records reduce it by 68.6% without additional body assets. In nine paired comparisons (18 trials) at starts displaced by 10-100 cm, experience recorded at the original start reduces mean time by 58-63% relative to no experience, demonstrating generalization to the tested new starting positions. During experience experiments, GPT-6-Astra spontaneously generates a short visual-feedback program. Researcher-refactored versions reduce mean local-task time by 29-31% in 27 simulation trials. Finally, 12 real-robot trials using operator-confirmed button contact demonstrate sim2real reuse: at a shared nominal start, simulation XML assets and simulation experience reduce mean time by 53.0% and 49.9%, respectively; real experience also transfers to two new starts. These results suggest a practical way to build general-purpose manipulation experiments around GPT-6-Astra: supply machine-readable body descriptions and synchronized demonstrations, and turn useful agent-generated feedback routines into reusable skills, while the agent adapts actions from current images. We release all task prompts, trial-level experimental data, and acquired skill implementations at this https URL.
[NLP-342] he Ongiini-Eval-OW Benchmark: A Concept Paper for the Planned Benchmarking of Machine Translation and Large Language Models on Oshindonga and Oshikwanyama
【速读】： 该论文旨在解决非洲本土语言——奥希瓦姆博语族（Oshiwambo）中，尤其是奥辛东加语（Oshindonga）与奥西昆亚马语（Oshikwanyama）在机器翻译（Machine Translation, MT）领域缺乏标准化评估基准的问题。当前主流商业翻译服务（如Google Translate、DeepL、Microsoft Translator）及开源多语言模型（如NLLB-200、MADLAD-400）和Masakhane开源检查点均未涵盖这两种标准化方言，导致相关语言的翻译性能无法被客观评估。其解决方案的关键在于构建一个名为Ongiini-Eval-OW的前瞻性英语-奥辛东加语与英语-奥西昆亚马语双语评估基准，包含600个测试项，由两名独立母语译者提供参考译文，配备30个跨译者一致性验证样本、11类现象标注的分层结构（每类不少于30项），采用确定性的30%盲测划分，并基于chrF++、BLEU和COMET-22实现可复现的评分协议，辅以50项人工评估环节。该基准设计强调数据质量、标注严谨性与评估透明度，目标于2026年第四季度正式发布，目前版本为概念论文，旨在推动多方参与共建，数据与代码分别以CC-BY-4.0和MIT许可开源。
链接: https://arxiv.org/abs/2609.31727
作者: Sebastian Küpers(Common Intelligence Foundation)
机构: Common Intelligence Foundation (爱沙尼亚); Ongiini AI programme (纳米比亚)
类目: Computation and Language (cs.CL)
备注: 14 pages, 6 tables. Concept paper for a planned benchmark; first public dataset release targeted for Q4 2026. Dataset CC-BY-4.0, code MIT
Abstract:Oshiwambo – a cluster of mutually intelligible Bantu languages spoken by over a million people across northern Namibia and southern Angola, and the home language of roughly half of Namibian households – has, to our knowledge, no published machine-translation evaluation benchmark. Major commercial services (Google Translate, DeepL, Microsoft Translator), open multilingual MT models (NLLB-200, MADLAD-400), and the open Masakhane checkpoint collection all lack coverage of either standardised dialect, Oshindonga or Oshikwanyama. We announce Ongiini-Eval-OW, a planned 600-item English-Oshindonga and English-Oshikwanyama benchmark with native-speaker references from two independent translators, a 30-item inter-translator agreement set, an 11-tag phenomenon-tagged stratification (at least 30 items per tag), a deterministic 30% blind split, and a reproducible scoring protocol over chrF++, BLEU, and COMET-22, supplemented by a 50-item human-evaluation round. We document the empirical coverage gap, the dataset composition, the launch-leaderboard model matrix across American, European, and Chinese frontier and open-weight systems, and the contribution pipeline. The dataset is targeted for first public release in Q4 2026; this v1.0 concept paper announces the design and the call for participation. Data and code will be released under CC-BY-4.0 and MIT respectively.
[NLP-343] MDL-Calibrated Significance-Gain Pair Encoding: Replication-Aware Automatic Stopping for Subword Tokenization
【速读】： 该论文旨在解决传统字节对编码（Byte-Pair Encoding, BPE）在构建子词词汇时需预先指定合并次数或目标词汇表大小的局限性，这一外部设定往往依赖经验或试错，缺乏自动性与理论依据。其核心解决方案是提出一种基于最小描述长度（Minimum Description Length, MDL）校准的显著性增益字节对编码（MDL-Calibrated Significance-Gain Pair Encoding, MDL-SG），该方法采用三阶段范式：首先在发现集上根据显著性增益（Significance-Gain）排序候选合并对；其次在独立的复制集上通过精确单侧超几何检验并结合每轮的Benjamini-Hochberg校正进行统计显著性验证，以确保结果可复现；最后在效用集上利用MDL准则评估合并带来的压缩性能提升。当无任何被复制的候选合并能带来正向的保留数据集MDL增益时，合并过程自动终止，从而实现无需预设词汇规模的自适应子词生成。实验表明，在WikiText-103数据集上，MDL-SG在50万字符训练样本下自动停止于847次合并，生成仅1,017个词元的存储词汇表，并在与频率BPE和SG-BPE计算资源相当的TinyGPT模型中实现了更低的测试比特每字符（BPC）值（3.2436 vs. 3.2894），证明了以语言模型效用为导向的合并选择策略优于单纯追求压缩性能的方法，揭示了压缩优化与模型实际表现之间的非一致性。
链接: https://arxiv.org/abs/2609.31705
作者: Azam Nouri
机构: Lincoln University (林肯大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: 15 pages, 3 tables, 1 algorithm. Source code available online
Abstract:Byte-Pair Encoding (BPE) constructs subword vocabularies through greedy pair merging, but conventional BPE requires the number of merges or target vocabulary size to be specified externally. Significance-Gain Pair Encoding (SG-BPE) replaces frequency-only selection with a statistical criterion based on how strongly an observed pair exceeds its expected co-occurrence under an independence model. This paper introduces MDL-Calibrated Significance-Gain Pair Encoding (MDL-SG), a three-stage procedure separating discovery, replication, and utility. Candidate pairs are ranked by Significance-Gain on a discovery partition, tested for replication on a separate partition using an exact one-sided hypergeometric test with per-iteration Benjamini-Hochberg correction, and then evaluated on a utility partition using a Minimum Description Length (MDL) criterion. Merging stops automatically when no replicated candidate yields positive held-out MDL gain. On WikiText-103, MDL-SG stops at 209, 433, and 847 merges for 120K, 250K, and 500K-character tokenizer-training samples, respectively. At 500K characters, it selects a stored vocabulary of 1,017 tokens without prescribing the vocabulary size in advance. In a compute-matched TinyGPT experiment with identical 2,024,448-parameter models and 500 optimizer updates per language model, MDL-SG achieves validation/test BPC of 3.2612/3.2436, compared with test BPC of 3.2894 for SG-BPE and 3.3493 for frequency BPE. Frequency BPE achieves stronger raw compression, while MDL-SG achieves lower BPC, showing that compression-oriented merge selection and language-model utility need not coincide. Comments: 15 pages, 3 tables, 1 algorithm. Source code available online Subjects: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL) ACMclasses: I.2.6; I.2.7 Cite as: arXiv:2609.31705 [cs.CV] (or arXiv:2609.31705v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.31705 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-344] Dont Repeat Yourself: Self-Supervised Fine-Tuning for Coverag e
【速读】： 该论文旨在解决生成式 AI 在可验证领域（如数学与编程）中因模型输出模式坍缩（mode collapse）导致的解法多样性不足问题，即尽管单次尝试的通过率（pass rate）可能较高，但大量重复相似解法使得在多次尝试中找到至少一个正确解的概率受限。传统后训练方法虽能提升模型对少数模式的集中度，而提高采样温度的策略则效果有限。其解决方案的关键在于提出一种无需奖励信号、验证器或正确性过滤的新型后训练方法——“不要重复自己监督微调”（Don’t Repeat Yourself Supervised Fine-Tuning, DRY-SFT）。该方法包含两个阶段：首先，针对每个问题顺序生成 K 个解法，强制模型在看到先前所有尝试的基础上生成不同解法；其次，对每个生成的解法独立进行微调，移除上下文中的历史尝试。此过程显著提升了生成解法的结构多样性（以抽象语法树编辑距离衡量），并在 HumanEval+、MBPP+ 和 DS-1000 基准上分别将 pass@100 提升 10.8、12.5 和 12.4 个百分点，同时在不显著牺牲 pass@1 的前提下大幅增强解法覆盖能力。实验表明，基础模型结构多样性越低，DRY-SFT 所带来的性能增益越大，说明该方法尤其适用于存在严重模式坍缩的模型。
链接: https://arxiv.org/abs/2609.31688
作者: Eric Fithian,Kirill Skobelev,X.Y. Han
机构: University of Chicago(芝加哥大学); Northwestern University(西北大学)
类目: Computation and Language (cs.CL)
备注: 19 pages, including references and appendices
Abstract:In verifiable domains such as math and coding, finding one correct solution among many attempts can matter more than the pass rate of each attempt. Post-training can concentrate large language model outputs around a few modes, while increasing sampling temperature has limited effectiveness. We introduce Don’t Repeat Yourself Supervised Fine-Tuning (DRY-SFT), a post-training method that increases output diversity and coverage: the probability of at least one correct solution among many attempts. DRY-SFT has two stages. First, for each problem, sequentially generate K solutions, showing the model all prior attempts and asking for a different solution. Second, fine-tune on each attempt independently, removing prior attempts from the context. The process uses no reward, verifier, or correctness filter. On HumanEval+, MBPP+, and DS-1000, DRY-SFT raises pass@100 by 10.8, 12.5, and 12.4 percentage points, respectively, at a small cost to pass@1. Structural diversity, measured by abstract syntax tree edit distance among passing solutions, rises significantly on all three benchmarks. DRY-SFT also solves 244 of 600 problems that the base model did not solve in the same 200 attempts. Across nine open-weight models, lower structural diversity of the base model significantly predicts larger DRY-SFT gains, indicating that the method is especially effective on more mode-collapsed models.
[NLP-345] Verification of PETSc with CIVL using LLM -generated ACSL contracts and deterministic driver generation
【速读】： 该论文旨在解决科学与工程领域广泛使用的并行数值库（如PETSc）缺乏形式化验证的问题，尤其是在错误结果可能导致严重后果的场景下。传统验证方法依赖专家手动编写规格说明并构造测试驱动程序（harness），这一过程繁琐且易引入额外缺陷，导致验证过程中出现误报或漏报。针对此问题，本文提出一种受限环境下利用大语言模型（LLM）生成可人工认证的ACSL（Annotated C Specification Language）合约的方法，结合确定性工具链，支持受限的ACSL子集，并自动生成由CIVL验证器执行的测试驱动程序。该方案在存在参考模型时以之为基准，在无参考模型时则以生成的契约作为唯一断言依据。研究通过端到端验证三个PETSc函数（MatAXPY、MatAYPX及MatFilter非压缩模式）证明了该方法的有效性，其中成功发现了一个自1997年以来存在于MatAYPX函数中的未被察觉的缺陷。其解决方案的关键在于：将LLM用于生成小规模、可人工审查的契约，辅以确定性工具链和严格验证流程，从而在降低自动化风险的同时实现高效、可信的形式化验证。
链接: https://arxiv.org/abs/2609.31687
作者: Hansol Suh,Jan Hückelheim,Stephen Siegel
机构: Argonne National Laboratory (阿贡国家实验室); University of Delaware (特拉华大学)
类目: Computation and Language (cs.CL); Mathematical Software (cs.MS); Programming Languages (cs.PL); Software Engineering (cs.SE)
备注:
Abstract:Parallel numerical libraries such as PETSc are widely used in science and engineering applications where wrong results can have costly consequences. Despite this, numerical libraries are rarely formally verified. One of the challenges is the need for an expert to hand-write a specification and manually apply a verification tool, often requiring the development of a harness or driver, all of which can contain additional bugs that lead to false positives or false negatives during verification. With recent advancements in large language models (LLMs), it is tempting to generate such drivers and reference models automatically, but one-shot generation based on a simple prompt is brittle and leads to additional unverified code that needs to be audited. In this paper, we present an approach to use LLMs in a limited setting to generate a small, human-certifiable ACSL contract from the function’s documentation, combined with a deterministic toolchain that supports a restricted ACSL profile and generates a driver that uses the CIVL verifier to check the implementation against the contract and, when available, an existing reference model. We demonstrate the pipeline end-to-end on three PETSc functions: MatAXPY (reference model already exists), MatAYPX (no reference model, so the certified contract is the sole oracle), and the non-compressing mode of MatFilter (no reference model, with more complex, conditional behavior). With this pipeline, we were able to discover a bug in PETSc’s MatAYPX function that was previously undiscovered and had been present in the code since 1997.
[NLP-346] he Temporal Tug-of-War: Visualizing and Detecting RAG Conflicts in Diffusion Models via Trajectory Variance EMNLP2026
【速读】： 该论文旨在解决生成式AI（Generative AI）在离散扩散语言模型中因检索增强生成（Retrieval-Augmented Generation, RAG）引入的一种特定失败模式：当外部检索到的上下文与模型参数化知识相矛盾时，迭代去噪过程会演变为不同知识源之间的可见冲突竞争。其解决方案的关键在于识别并量化这种冲突——提出轨迹方差得分（Trajectory Variance Score, TVS），作为可观察的时序语义分歧指标。TVS通过计算独立随机去噪轨迹中答案嵌入之间的平均成对余弦距离，捕捉参数化知识与上下文知识在生成过程中的动态拉锯效应；该方法仅需两次并行推理即可完成，具有计算轻量性。实验表明，在四个不同数据集（Synthetic、SciQ、PopQA、CounterFact）上，基于TVS的简单逻辑回归分类器在LLaDA模型上达到70.10%准确率和0.7647 AUROC，而在Dream 7B上，将轨迹数量从2增至5可使准确率提升至69.62%，且复杂序列模型带来的性能增益有限。结果证明，由冲突引发的轨迹动力学及其关键特征可在不同扩散架构间有效迁移。
链接: https://arxiv.org/abs/2609.31684
作者: Sravan Karthick T,Pranav Darshan,Pranav A,Minal Moharir,Ivan P. Yamshchikov
机构: R.V. College of Engineering(印度R.V.工程学院); Technical University of Applied Sciences Würzburg-Schweinfurt(德国维尔茨堡-施韦因富特应用技术大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted at the UncertaiNLP Workshop at EMNLP 2026
Abstract:Retrieval-Augmented Generation (RAG) introduces a specific failure mode in discrete diffusion language models: when retrieved context contradicts parametric knowledge, the iterative denoising process becomes a visible battleground between competing knowledge sources. We identify temporal semantic divergence as an observable for detecting these conflicts and introduce the Trajectory Variance Score (TVS), a simple and interpretable measure of this divergence. TVS computes the mean pairwise cosine distance of answer embeddings across independent stochastic denoising trajectories, capturing the temporal tug of war between parametric and contextual attractors. Requiring as few as two parallel inference runs, TVS is computationally lightweight. Across four diverse datasets (Synthetic, SciQ, PopQA, and CounterFact), a simple Logistic Regression classifier using TVS achieves 70.10% accuracy and 0.7647 AUROC on LLaDA. On Dream 7B, increasing the number of trajectories from two to five improves accuracy from 63.91% to 69.62% . More complex sequential models provide only marginal improvements over the linear classifier. Evaluation across LLaDA and Dream 7B demonstrates that conflict-induced trajectory dynamics and their key properties transfer across distinct diffusion architectures.
[NLP-347] When Keywords Drop but Classifiers Hold: Soft Refusals under KV Cache Compression
【速读】： 该论文旨在解决在长上下文大语言模型（LLM）推理中，因内存限制采用键值缓存压缩（KV cache compression）技术时，其对轻量级关键词过滤器与更强的拒绝分类器之间一致性的影响问题。当前部署系统通常依赖关键词过滤或学习型分类器对生成内容进行拒答评分，以判断模型是否在实际服务场景下成功拒绝有害请求，但现有方法未明确验证压缩策略在保持任务准确性的同时，是否同样维持了不同监控机制间的判断一致性。研究通过在200个有害提示上构建配对实验（n=200），在长填充上下文条件下，对同一提示分别执行完整保留和匹配淘汰的推理，并由关键词启发式、HarmBench Llama-2-13B分类器、辅助大模型判别器及人类标注者评估回复结果。实验发现，在Qwen2.5-3B模型上，关键词拒答率从98.0%下降至80.5%（McNemar检验p~1e-8），而分类器拒答率仍接近天花板水平（99.0%-99.5%），且MMLU任务准确率保持不变（50.0%）。人类标注主要与分类器一致，支持“软拒答”现象的存在。该差异并非普遍成立，其影响在短填充上下文及使用配对SnapKV压缩时减弱。因此，研究指出：在压缩环境下进行安全审计时，应依赖与实际服务上下文相匹配的多重判别器，而非仅依赖关键词拒答率。
链接: https://arxiv.org/abs/2609.31678
作者: Kang Chen,Xiuze Zhou,Hong Chen,Yuanguo Lin
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 8 pages, 5 figures
Abstract:KV cache compression is widely used for long context LLM inference under memory constraints, while deployed systems typically score refusals after generation with keyword filters or learned classifiers. Such monitors are intended to indicate whether a model declined a harmful request under the serving regime actually used. However, it remains unclear whether matched compression that preserves task accuracy also preserves agreement between lightweight lexical monitors and stronger refusal classifiers. We study this with a paired protocol on n=200 harmful prompts with a long filler context: each prompt is answered once under full retention and once under matched eviction after a shared prefill, and the same replies are scored by keyword heuristics, the HarmBench Llama-2-13B classifier, an auxiliary LLM judge, and humans on disagreements. On Qwen2.5-3B, keyword refusal falls from 98.0% to 80.5% (McNemar p~1e-8) while classifier refusal stays near ceiling (99.0%-99.5%) and MMLU accuracy is unchanged (50.0%); human labels predominantly follow the classifier, consistent with soft refusals. The gap is not universal and weakens under short fillers and paired SnapKV, so safety auditing under compression should rely on several judges matched to the serving context rather than on keyword rates alone.
[NLP-348] An Evaluation of AI-Supported Evidence-Based Learning for Public Speaking Skill Development
【速读】： 该论文旨在解决现有生成式AI (Generative AI) 驱动的演讲辅导工具在反馈通用化和缺乏高质量示范演讲支持方面的不足。其核心问题在于，当前系统难以根据具体演讲语境提供个性化评估，且缺少可资学习的优化版演讲范例（包括文本与音频形式），限制了学习效果。本研究提出的关键解决方案是：引入演讲语境作为评价指标的调节因素，实现上下文感知的定制化反馈；同时，基于语境信息生成更高质量、适配场景的示范性演讲文本与语音内容，以支持基于样例的学习。在为期一至三周的小规模试点研究中，七名本科生重复使用该系统并至少练习同一演讲三次，结果显示所有参与者在演讲表现上均有提升，填充词使用频率下降，焦虑水平降低15.3%至34.4%，且普遍认为上下文感知反馈与优化后的示范演讲具有显著助益。研究证实，结合上下文感知反馈与高质量示范演讲的策略，能够有效提升公众演讲能力并缓解演讲焦虑，为未来融合视频分析以捕捉演讲者体态行为的研究提供了方向。
链接: https://arxiv.org/abs/2609.31676
作者: Sashini Hettiarachchi,Shahbaz Siddeeq,Mika Saari,Pekka Abrahamsson
机构: 未知
类目: Computation and Language (cs.CL)
备注: Proceedings of the 54th Annual Conference of the European Society for Engineering Education (SEFI 2026)
Abstract:Public speaking is an essential skill in academic and professional contexts, but it often causes anxiety. Although several AI-based speech coaching tools exist, they typically provide generic feedback and lack model speeches for learning. This study addresses these gaps by developing a system that incorporates speech context into evaluation criteria, enabling more tailored feedback. It also generates improved model speeches in both text and audio formats to support example-based learning. The system was evaluated in a pilot study with seven undergraduate students over one to three weeks. Participants used the system repeatedly and practiced the same speech at least three times. Results showed improvements in public speaking performance for all participants. Filler word usage decreased, and anxiety levels dropped by 15.3% to 34.4%. Participants found the context-aware feedback and revised speeches useful and confidence-building. These findings suggest that context-aware feedback and model speeches can enhance public speaking skills and reduce anxiety. Future studies could explore video-based analysis of physical behavior during speeches.
[NLP-349] Age-Adaptive Handwriting Reconstruction from an IMU-Based Digital Pen through Shared Representations and Domain-Specific Heads
【速读】： 该论文旨在解决跨年龄群体（成人与儿童）在使用惯性测量单元（IMU）-equipped数字笔进行手写轨迹重建时面临的性能退化问题。由于成人与儿童在书写动力学、运动控制、执笔方式及自信心等方面存在显著差异，导致同一视觉相似的书写轨迹在传感器信号上表现出高度变异性，且现有模型在单一人群数据上训练后难以泛化至另一群体。此外，大规模采集儿童手写数据在教育场景中难以实现。为此，本文提出一种基于时序卷积网络（Temporal Convolutional Network, TCN）的跨域学习框架，采用多预测头结构，在共享特征提取的基础上有效建模不同年龄群体间的共性与差异。该方法通过联合利用成人与儿童的数据，使各群体均能从对方数据中获益，从而构建一个无需用户特定适配即可直接部署于笔端的统一、鲁棒的手写重建模型。
链接: https://arxiv.org/abs/2609.31666
作者: Florent Imbert(LUT),Yann Soullard(IRISA, UR2, SHADOC),Eric Anquetil(INSA Rennes, IRISA, SHADOC),Hui Han(LUT)
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Digital pens are widely used to capture handwriting on digital devices, enabling precise trace recording and enhancing human-computer interaction. However, most are bundled with tablets and lack cross-brand compatibility. Recent digital pens equipped with kinematic sensors have emerged, designed for use on any surface. This especially opens significant potential for supporting handwriting acquisition in classrooms. Handwriting reconstruction from such an IMU-equipped pen poses a challenge due to the significant variability in sensor signals between adults and children. Even when producing visually similar traces, variations in writing dynamics, motor control, pen holding, and user confidence introduce substantial discrepancies in the captured signals. Additionally, the high variability in children’s handwriting requires collecting large amounts of data, which is not feasible to implement in a school environment at scale. Furthermore, models trained exclusively on adult data fail to generalize to children’s handwriting, and conversely, models trained on children’s data perform poorly on adult writers. This crosspopulation degradation highlights the need for a unified model that can be deployed directly on the pen, without any user-specific adaptation. To address this issue, we propose a cross-domain learning, using an original neural network architecture based on a Temporal Convolutional Network and multiple prediction heads. The model is designed to be robust across age groups by leveraging shared features while effectively handling variability induced by differences in graphomotor development. This approach aims to improve handwriting trace reconstruction from sensor data, where each domain benefits from additional data provided by the other domain.
[NLP-350] What Drives Dialectal Jailbreaks? An Ablation of Surface Form Cultural Framing and Strategy Banks
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在面对特定语言风格攻击时拒绝响应行为减弱的问题，尤其关注非标准语言形式、文化语境化表达以及提示策略空间优化对攻击成功率的影响。其核心发现表明，方言表面形式并非攻击成功的关键因素：无论是普通话、英语还是未经优化的方言翻译，攻击成功率均低于8%。真正决定攻击成效的是由优化器控制的策略库（strategy bank）的存在与否——只要保留该策略库，无论语言形式如何，攻击成功率均可达到98%–100%。进一步实验显示，一个文化中立的通用策略库仅需单次查询即可达到相同上限，证明攻击效果主要源于策略库的表达能力，而非方言或文化内容本身。尽管如此，方言选择仍影响查询效率与响应严重程度，且定性分析揭示了如语义模糊化（semantic glossing）和程序化支架（procedural scaffolding）等共现的高层级攻击机制。研究还指出，当前设定下攻击成功率对评判标准校准敏感，且一次性的策略生成与迭代搜索未完全解耦，因此其“绝对天花板”存在局限性。
链接: https://arxiv.org/abs/2609.31664
作者: Qingyang Xu
机构: Independent Researcher(独立研究员); Shanghai, China(中国上海)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: Empirical Methods in Natural Language Processing 2026, 15 pages, 12 tables
Abstract:Recent work suggests that obscure language registers can weaken large language model refusal behavior, especially when paired with black-box prompt optimization. It remains unclear whether failures stem from non-standard surface form, culturally grounded framing, or optimization over an expressive prompt-strategy space. We study Chinese registers by extending a classical-Chinese red-teaming framework to Shanghainese and Cantonese and running a 36-cell ablation across surface forms, strategy-bank variants, and two target models. The main finding is corrective: dialectal surface form is neither necessary nor sufficient for high attack success. Non-optimized English, Mandarin, and naive dialect translations remain below 8% attack success rate, whereas all conditions that retain an optimizer-controlled strategy bank reach 98–100%. A culture-neutral generic strategy bank reaches the same ceiling at near-single-query cost, further indicating that strategy-bank expressiveness, rather than dialectal or cultural content, accounts for most of the observed effect. Dialect choice still affects query efficiency and response severity, and qualitative coding identifies recurring high-level mechanisms such as semantic glossing and procedural scaffolding. We qualify the absolute ceiling-level attack success rate by showing sensitivity to judge calibration and noting that the design does not fully separate one-shot strategy construction from iterative search.
[NLP-351] LLM -Guided Ontology-Driven Knowledge Graph Construction from Unstructured Text
【速读】： 该论文旨在解决从工业文本中构建本体驱动的知识图谱所面临的挑战，主要包括文档的领域专属性、标注资源稀缺性以及本体工程流程的复杂性。其核心解决方案在于提出并验证了一种集成紧凑型开源大语言模型（Large Language Models, LLMs）、可复用的提示策略（prompting strategies）及开放知识库的本体学习流水线。该方法通过在本地部署的7B至32B参数量级开源LLMs上运行，从非结构化电力电网事故报告中提取实体与关系，生成RDF三元组，构建相应的OWL本体，并利用外部知识源进行本体增强，进而评估本体质量并构建基于该本体模式的填充知识图谱。实验基于80份人工标注的私有法语文本数据集，结果表明，基于模式引导的提示显著提升了信息抽取质量，而量化模型在性能与计算成本之间实现了有效权衡。研究证实了利用本地部署的开源大模型将领域特定工业文本转化为基于本体的知识图谱的可行性，并通过可复用的提示策略支持了抽取过程的泛化能力。
链接: https://arxiv.org/abs/2609.31663
作者: Abdelhadi Belfadel,Maxence Gagnant,Joseph Kattan,Sana Tmar
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Ontology-driven knowledge graph construction from industrial text remains challenging due to the domain specificity of documents, the scarcity of annotated resources, and the complexity of ontology engineering workflows. This paper presents and investigates the applicability of an ontology learning pipeline that combines compact open-source Large Language Models (LLMs), reusable prompting strategies, and open knowledge bases to support the extraction, structuring, enrichment, and evaluation of knowledge from textual corpora. The approach is tested and evaluated on a private French corpus of power-grid incident reports, using locally deployable open-source LLMs ranging from 7B to 32B parameters. Starting from unstructured reports, the approach extracts entities and relations, generates RDF triples, constructs related OWL ontology, enriches it using external knowledge sources, assesses the quality of the ontology, and subsequently constructs a populated knowledge graph grounded in the resulting ontology schema. Experiments on 80 manually annotated private reports show that schema-guided prompting significantly improves extraction quality, while quantized models provide an effective trade-off between performance and computational cost. These results demonstrate the feasibility of transforming domain-specific industrial text into ontology-based knowledge graphs using locally deployed open-source LLMs, while supporting the generalization of the extraction process through reusable prompting strategies.
[NLP-352] Distributional sentiment modeling and anomaly detection for consumer complaint assessment
【速读】： 该论文旨在解决传统情感分析在金融与风险管理应用中将消费者投诉文本的情感强度简化为离散极性标签或单一预测特征，从而忽略情感强度分布结构的问题。其核心解决方案在于将消费者投诉中的负面情感视为一个有界连续变量，采用Transformer分类器对每篇叙事进行评分，并用贝塔分布（Beta distribution）建模得分的分布特征；通过Kullback-Leibler散度与平方海林格距离（squared Hellinger distance）对比优质投诉与非优质投诉的拟合分布，进而将分布特征与金额赔付及企业响应结果关联，构建异常诊断指标，识别文本严重性与实际处理结果不一致的异常投诉。研究表明，两组投诉的情感强度分布高度重叠，表明仅依赖情感强度无法有效区分结果，但结合货币属性与类别信息后，可有效识别出异常严重的投诉，适用于操作风险监控。该研究提出了一种将情感提取、有界响应建模与异常检测统一整合的连续分布型情感分析框架，显著提升了对消费者投诉的精细化评估能力。
链接: https://arxiv.org/abs/2609.31653
作者: Peiheng Gao,Chen Yang,Shimin Zhang
机构: Western University ( Western University); Icahn School of Medicine at Mount Sinai (伊坎医学院); National University of Singapore (新加坡国立大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG); Applications (stat.AP)
备注:
Abstract:Sentiment analysis is a common tool for converting unstructured text into quantitative signals in finance and risk management. Yet most applications reduce the output to a discrete polarity label or a single predictive feature, overlooking the distributional structure of sentiment intensity in consumer complaint narratives. In this paper we treat negative sentiment in consumer complaints as a bounded continuous variable and study its full distribution rather than a single label. We score each narrative with a transformer classifier, model the scores with Beta distributions, and compare the fitted distributions of meritorious and non-meritorious complaints through the Kullback Leibler divergence and the squared Hellinger distance. The fitted distributions are then linked with dollar amounts and company response outcomes to construct anomaly diagnostics that flag complaints whose textual severity is inconsistent with the recorded relief. We find that the two groups have strongly overlapping distributions, so negative sentiment intensity is not a sharp classifier of outcomes on its own; combined with monetary and categorical attributes, it isolates unusually severe complaints for operational risk monitoring. Treating sentiment analysis as continuous distributional measurement, this study links sentiment extraction, bounded response modeling, and anomaly detection in a unified framework for consumer complaint assessment.
[NLP-353] PalmLeaf-VQA: A Multi-Script Visual Question Answering Benchmark for Historical Palm-Leaf Manuscript Understanding Across Diverse Regions
【速读】： 该论文旨在解决多模态大语言模型（MLLMs）在处理文化多样性、图像退化及非拉丁文书写系统的历史文献图像时表现不足的问题，尤其针对南亚与东南亚传统贝叶手稿的视觉理解能力缺乏系统评估。其核心解决方案是提出首个面向历史贝叶手稿的多语言视觉问答基准——PalmLeaf-VQA，包含923幅经过筛选的手稿图像和7,384个问题-答案对，覆盖巴厘语、格兰塔文、贾塔卡姆、坎巴拉玛亚纳姆、卡纳达文、高棉文、巽他文及泰米尔文八种文字体系。不同于以识别为导向的数据集，PalmLeaf-VQA聚焦于基于保存相关视觉线索（如物理状态、行结构、材质与涂层、装订孔、页边距、符号、插图及局部视觉瑕疵）的文档感知型视觉推理任务。通过在开放回答与受限回答两种提示范式下评估主流私有与开源MLLMs，研究揭示当前模型在罕见文字、退化版式和保护性理解任务上的性能上限仅为58.00%的精确匹配准确率，凸显了现有模型在文化语境敏感性和版式感知方面的显著局限。PalmLeaf-VQA为推动具有文化根基且具备版式意识的多模态文档分析提供了标准化评估框架。
链接: https://arxiv.org/abs/2609.31651
作者: Nimol Thuon,Jun Du,Panhapin Theang
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Historical manuscripts remain largely absent from modern vision-language benchmarks, leaving open how well multimodal large language models (MLLMs) handle culturally diverse, degraded, and non-Latin document images. We introduce \textbfPalmLeaf-VQA, a multi-script visual question answering benchmark for historical palm-leaf manuscript understanding across South and Southeast Asian traditions. PalmLeaf-VQA contains \textbf923 curated manuscript images and \textbf7,384 question–answer pairs from eight collection groups: Balinese, Grantha, Jathakam, Kambaramayanam, Kannada, Khmer, Sundanese, and Tamil. Unlike recognition-oriented resources, the benchmark targets manuscript-aware visual reasoning over preservation-relevant cues, including physical condition, line structure, material and coating, binding holes, margins, symbols, drawings, and localized visual artifacts. We evaluate recent proprietary and open-weight MLLMs under open-answer and constrained-answer prompting and provide fine-grained analysis across collections, question categories, and task types. The strongest evaluated model reaches only \textbf58.00% exact-match accuracy on the held-out test split, revealing substantial limitations in current MLLMs for rare-script, degraded-layout, and preservation-oriented document understanding. PalmLeaf-VQA provides a standardized benchmark for advancing culturally grounded and layout-aware multimodal document analysis.
[NLP-354] A literature-guided descriptor-based framework for filtering composition search spaces
【速读】： 该论文旨在解决科学文献中蕴含的材料行为隐性知识难以被有效利用于实际材料发现问题的挑战，尤其针对如何从大规模科学语料库中提取可指导材料设计的有用信息。其核心解决方案是提出一种基于文献引导的描述符（descriptor）过滤框架，通过在文献训练的词嵌入模型中筛选与特定性能指标相关的描述符，构建帕累托（Pareto）-基础的过滤机制，从而高效缩小候选成分空间。该方法的关键在于：利用性能指标依赖的描述符实现对候选成分的精准筛选，在保留高价值成分的同时显著降低搜索空间复杂度——实验表明，该框架平均可剔除74.27%的候选成分，且最优性能值误差仅为1.93%，优于专家选择或随机选取描述符的方案。这证明了从文献中提取的词嵌入能够支持直观、可重复且高效的成分空间压缩策略。
链接: https://arxiv.org/abs/2609.31650
作者: Lei Zhang,Markus Stricker
机构: Interdisciplinary Centre for Advanced Materials Simulation, Ruhr University Bochum (鲁尔大学波鸿分校跨学科先进材料模拟中心)
类目: Computation and Language (cs.CL); Materials Science (cond-mat.mtrl-sci)
备注: 14 pages, 3 figures, 2 tables, appendix
Abstract:Scientific literature contains latent knowledge about materials behavior, but much of this knowledge is expressed through words, contexts, and recurring associations rather than explicit design principles. This raises a central question: how can large-scale scientific corpora be used for practical problems in materials discovery? Here, we present a literature-guided descriptor-based filtering framework for reducing composition search spaces. For a given performance metric, the framework selects two descriptors from a filtered vocabulary in a literature-trained word embedding model and uses the selected descriptors to construct a Pareto-based filter for candidate compositions. Across the evaluated performance metrics and composition search spaces, the framework filters out an average of 74.27% of the candidate compositions, with an average best-value error of 1.93% relative to experimental measurements. Compared with expert-chosen and random descriptors, our performance metric-dependent descriptors provide a more controlled balance between retained fraction and best-value error. These results show that literature-derived embeddings can support intuitive and reproducible filters for narrowing candidate composition spaces while preserving high-performing compositions.
[NLP-355] From Hand-Crafted to LLM -Based Variation Operators in Metaheuristics: A Tutorial
【速读】： 该论文旨在解决在元启发式算法中如何有效利用大语言模型（Large Language Models, LLMs）作为变异算子的问题，即如何将生成式AI（Generative AI）嵌入到迭代搜索过程中，以生成或修改候选解、启发式规则或程序。其核心挑战在于如何设计可解释、可控制且高效的模型驱动变异机制。解决方案的关键在于提出一个操作层级的框架，通过两个关键维度对变异算子进行建模：一是变异时刻所依赖的提示条件信息类型（数值型 \textttNumeric、符号型 \textttSymbolic、语言型 \textttLinguistic），二是模型调用后保留的实体特征（瞬态 \textttTransient、分摊 \textttAmortized、迁移 \textttTransfer）。该框架通过工作模板、方法综述、证据表与成本感知决策指南，实现了对变异算子的系统分类、构建与选择，从而为将大语言模型安全、高效地集成于优化流程提供了结构化方法论支持。
链接: https://arxiv.org/abs/2609.31649
作者: Camilo Chacón Sartori,Guillem Rodríguez-Corominas,Christian Blum
机构: Apeiron Intelligence(阿佩隆智能); Artificial Intelligence Research Institute (IIIA-CSIC)(人工智能研究所(西班牙国家研究委员会)); Universitat Politècnica de Catalunya (UPC)(加泰罗尼亚理工大学)
类目: Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Large language models (LLMs) are increasingly being employed as variation operators in metaheuristics, generating or modifying candidate solutions, heuristics, or programs inside iterative search loops. This shift reframes variation as a model call conditioned on different types of information. We introduce an operator-level framework with two descriptors: (1) the type of prompt-conditioning information at variation time (\textttNumeric, \textttSymbolic, \textttLinguistic), and (2) artifact persistence, identifying what survives the model call (\textttTransient, \textttAmortized, \textttTransfer). The tutorial shows how to classify, build, and select these operators through a worked build template, a method survey, an evidence table, and a cost-aware decision guide.
[NLP-356] ChestPheNoT: Deployable Auditable Label-Status-Evidence Extraction from Radiology Reports ALT
【速读】： 该论文旨在解决医学影像报告中结构化表型提取的实用部署难题，即如何在不依赖外部API、确保临床文本不出机构的前提下，实现可审计的精准提取。其核心挑战在于现有方法要么缺乏支持证据（传统标注器），要么受限于数据隐私（云端大模型）。解决方案的关键是提出CHESTPHENOT——一个0.5至3B参数量级的轻量级语言模型，能够联合抽取病灶标签、三分类状态（存在/不存在/不确定）以及原始报告中的字面支持证据片段。该模型通过混合使用CheXbert与72B规模模型生成的“银色标签”进行预训练，再经监督微调及轻量级GRPO优化，显著提升了在跨机构、跨分类体系等分布外场景下的鲁棒性。实验表明，3B模型虽在同分布下略逊于其教师模型CheXbert，但在分布偏移时表现更具竞争力，尤其在跨机构检测任务上较CheXbert提升2.0 F1；同时在证据可追溯性方面，99%以上的证据片段可精确定位，其可审计F1达47.5，优于Qwen2.5-7B单次提示模型7.6点，并接近Qwen2.5-72B的表现，验证了本地部署模型在准确性与可解释性上的双重优势。
链接: https://arxiv.org/abs/2609.31629
作者: Kai Yu,Chenyu Zhu,Zaifu Zhan,Meijia Song,Min Zeng,Xiaoyi Chen,Mingquan Lin,Rui Zhang
机构: University of Minnesota; University of California Davis
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at IEEE Healthcom 2026
Abstract:Structured phenotype extraction from radiology reports supports cohort construction, quality auditing, and clinical analytics, but practical deployment requires local inference and auditable predictions, while expert annotations remain scarce. Conventional labelers provide structured findings and assertion states but no supporting evidence, while API-hosted large language models may be unsuitable when clinical text cannot leave institutional infrastructure. We present CHESTPHENOT, a compact 0.5-3B language model that jointly extracts finding labels, three-class status (present/absent/uncertain), and verbatim supporting evidence spans. CHESTPHENOT is trained using hybrid CheXbert+72B silver supervision followed by supervised fine-tuning and lightweight GRPO refinement. Across three human-annotated gold sets spanning in-distribution, cross-taxonomy, and cross-institution evaluation, the 3B model remains below its CheXbert silver teacher in distribution but is competitive under distribution shift, significantly surpassing CheXbert on cross-institution detection (+2.0 F1). Task-specific training also enables the 3B model to match or exceed substantially larger prompted models on most detection and status comparisons. For evidence-grounded extraction, over 99% of final evidence spans are locatable in the source report, and the 3B model achieves 47.5 auditable-F1, outperforming Qwen2.5-7B one-shot prompting by 7.6 points and approaching Qwen2.5-72B. These results demonstrate that locally deployable models can provide competitive and directly auditable radiology-report extraction without relying on external inference APIs. Code and the full extraction/judge prompts will be made available at this https URL.
[NLP-357] Explainable and Generalisable LLM -based Cognitive Decline Detection with Spontaneous Speech
【速读】： 该论文旨在解决阿尔茨海默病（Alzheimer’s disease, AD）及轻度认知障碍（mild cognitive impairment, MCI）早期筛查中传统诊断方法资源消耗大、难以规模化推广的问题。其核心挑战在于如何有效捕捉并利用言语中细微的声学与语义变化，以实现高效、可解释的自动化认知状态评估。解决方案的关键在于提出一种新型双语语音大语言模型框架，直接处理原始语音信号，而非依赖易出错的自动语音识别（automatic speech recognition, ASR），从而保留关键的韵律特征；同时采用多任务学习机制，联合完成认知状态分类与生成临床可理解的自然语言解释，显著提升了模型在跨任务场景下的泛化能力。实验表明，该系统在六种不同数据集/任务条件下均达到最高平均准确率和受试者工作特征曲线下面积（AUROC），且在未见任务上无需微调即可保持良好性能，经临床专家评估，生成的解释具备高度临床相关性与证据一致性，验证了其在实际医疗场景中的可解释性与实用性。
链接: https://arxiv.org/abs/2609.34217
作者: Ziyun Cui,Wen Wu,Chuan Shi,Shuguang Yang,Xueying Gui,Yan Zheng,Qiong Yang,Haiyan Zhao,Wei-Qiang Zhang,Ji Wu,Yelei Li,Nan Li,Chao Zhang
机构: Tsinghua University (清华大学); Shanghai Artificial Intelligence Laboratory (上海人工智能实验室); Peking University (北京大学); Peking University Sixth Hospital (北京大学第六医院); Peking University Third Hospital (北京大学第三医院); OPPO Research Institute (OPPO研究院)
类目: Audio and Speech Processing (eess.AS); Computation and Language (cs.CL)
备注:
Abstract:Alzheimer’s disease (AD) and mild cognitive impairment (MCI), which may precede AD, manifest early through subtle linguistic and acoustic alterations. Traditional diagnostics, however, are often resource-intensive and lack scalability for mass screening. To address these challenges, we introduce a novel bilingual speech large language model framework for automated, explainable cognitive screening. Unlike conventional pipelines that rely on error-prone automatic speech recognition, our system directly processes raw speech to learn joint acoustic-semantic representations, preserving critical prosodic cues often lost in transcription. Utilising our newly collected PUTH-AD dataset alongside multiple open-source corpora, we implemented a multi-task learning objective that simultaneously performs cognitive status classification and generates clinician-understandable natural language explanations. Our system achieved the highest average accuracy and AUROC across six dataset/task conditions, comparing three representative baselines. The system demonstrated cross-task transfer to held-out PUTH-AD task subsets, maintaining classification accuracy on an entirely unseen cognitive task without task-specific fine-tuning. Furthermore, clinician evaluation confirms that the generated explanations are both clinically relevant and largely consistent with the underlying speech evidence, supporting their potential utility in clinical interpretation. This study provides a scalable, objective, and explainable framework for speech-based cognitive screening, combining cognitive status classification with natural language explanations that clinicians can assess and verify, bridging the gap between advanced AI and clinical utility.
[NLP-358] From Script to Drama: An Agent ic Framework for Controllable Multi-Speaker Dialogue TTS
【速读】： 该论文旨在解决多说话人对话文本转语音（multi-speaker dialogue TTS）中自然语音生成、说话人身份一致性、跨轮次语流连贯性以及情感、语速、音量等表达属性的细粒度控制难题，尤其在长篇对话场景下，传统一次性生成方法难以可靠满足上述要求。其解决方案的关键在于提出一种受批判驱动的迭代优化框架，通过引入以指令跟随和自然语言引导的属性编辑为核心的语音生成骨干模型ControlEdit-TTS，实现对表达错误的局部修正而无需全量重生成；同时采用分层式的话语级与场景级批判机制，智能地将检测到的问题路由至编辑、重合成或时序调整等不同处理路径，从而在保持说话人身份一致性的前提下，显著提升对话整体的自然度与可控性。实验结果表明，该框架在双语中英对话基准上优于直接对话模型及代理基线，在话语级指令遵循和对话级偏好评分方面表现更优，且相比仅依赖重生成的方法更具效率与有效性。消融实验进一步验证了场景级批判与基于编辑的纠错策略的核心贡献。
链接: https://arxiv.org/abs/2609.33362
作者: Kangxiang Xia,Xinfa Zhu,HangRui Hu,Kexin Huang,Wenjie Tian,Ziyue Jiang,Bingshen Mu,Jingbin Hu,Ting He,Lei Xie,Jin Xu
机构: 未知
类目: Audio and Speech Processing (eess.AS); Computation and Language (cs.CL); Sound (cs.SD)
备注:
Abstract:Multi-speaker dialogue TTS requires natural speech generation, consistent speaker identity, coherent cross-turn transitions, and fine-grained control of expressive attributes such as emotion, speaking rate, and loudness. These requirements are difficult to satisfy reliably with one-shot generation, especially in long-form dialogue. We propose a controllable multi-speaker dialogue TTS framework that formulates synthesis as critique-driven iterative refinement. Its speech backbone, ControlEdit-TTS, unifies instruction-following synthesis and natural-language-guided attribute editing, enabling correction of expressive errors without full regeneration. The framework further performs hierarchical utterance-level and scene-level critique, routing detected issues to editing, resynthesis, or timing adjustment. Experiments on a bilingual Chinese–English dialogue benchmark show improved utterance-level instruction following, better dialogue-level preference than direct dialogue models and agentic baselines, and more effective refinement than regeneration-only alternatives while preserving speaker identity. Ablations further confirm the benefits of scene-level critique and edit-based correction.
[NLP-359] Acoustic Progress Propagation for Long-Horizon Speculative Decoding in ASR
【速读】： 该论文旨在解决自回归自动语音识别（ASR）中，基于对齐感知的推测解码（alignment-aware speculative decoding）在扩展推测时长（draft horizon）时，接受长度（acceptance length）出现饱和的问题。其核心解决方案是提出一种进度感知的推测生成器（progress-aware speculative drafter），通过在推测步骤间递归传播声学进度状态（acoustic progress state），并将该状态反馈至音频交叉注意力（audio cross-attention）中以指导词元生成。该方法联合训练推测生成器与进度预测器，使其适应可变的推测时长。实验结果表明，在五个ASR测试集上，该方法相较于仅使用目标模型的自回归解码，实现了无损的端到端加速比分别为1.657倍（Qwen3-ASR-0.6B）和1.227倍（Qwen3-ASR-1.7B），相较于AnchorDraft，平均加速比分别提升了34.3%和9.0%，且在不同推测时长下的接受长度持续增长，突破了基线方法的饱和瓶颈。
链接: https://arxiv.org/abs/2609.33245
作者: Yuanyuan Jia,Qianqian Yang
机构: 未知
类目: Audio and Speech Processing (eess.AS); Computation and Language (cs.CL)
备注: 5 pages, 2 figures, 2 tables
Abstract:Speculative decoding accelerates autoregressive automatic speech recognition (ASR), but the acceptance length of alignment-aware drafters can saturate as the draft horizon increases. We propose a progress-aware speculative drafter that recurrently propagates an acoustic progress state across draft steps and feeds it back into audio cross-attention to guide token generation. We jointly train the drafter and progress predictor over variable draft horizons. On five ASR test sets, our method achieves lossless, macro-averaged end-to-end speedups of 1.657x and 1.227x over target-only autoregressive decoding with Qwen3-ASR-0.6B and Qwen3-ASR-1.7B, respectively. Relative to AnchorDraft, our method improves the macro-averaged speedup by 34.3% and 9.0%, respectively. Horizon sweeps show continued growth in acceptance length beyond the baselines’ saturation. Code is available at this https URL.
[NLP-360] Improving Audiovisual Speech Recognition through Synthetic Visual Data Augmentation
【速读】： 该论文旨在解决音频视觉语音识别（Audiovisual Speech Recognition, AVSR）中因标注音视频（AV）数据集稀缺而导致模型发展受限的问题。其核心解决方案在于利用音频驱动的虚拟人头生成流水线，从已有音频数据中合成唇动同步的视觉信息，从而构建合成视觉数据。该方法的关键在于通过真实音频生成高质量、时序对齐的合成视觉内容，既可作为真实数据的增强手段，也可在缺乏标注AV数据的语言中独立用于训练。实验结果表明，将合成数据与真实数据结合可实现高达16.2%的相对词错误率（Word Error Rate, WER）降低；更重要的是，仅使用合成数据即可为无现成AV数据的语言建立有效的AVSR基线模型。这证明了合成视觉数据在应对多模态语音识别中的数据稀缺问题上具备可扩展性和广泛适用性。
链接: https://arxiv.org/abs/2609.31961
作者: Pol Buitrago,Pol Gàlvez,Javier Hernando
机构: Universitat Politècnica de Catalunya (UPC); Barcelona Supercomputing Center (BSC)
类目: Audio and Speech Processing (eess.AS); Computation and Language (cs.CL); Sound (cs.SD); Image and Video Processing (eess.IV)
备注: 12 pages, 9 Figures
Abstract:Audiovisual Speech Recognition (AVSR) is a multimodal approach to speech recognition that incorporates visual information from lip movements to enhance model performance. Despite its advantages, its development remains constrained by the limited availability of labeled audiovisual (AV) datasets. This work explores the use of synthetic visual data as a solution, using an audio-driven talking-head pipeline to generate lip-synchronized visual content from existing audio data. We evaluate the effectiveness of synthetic visual data both as an augmentation strategy and as a standalone training resource, applying our approach to Spanish and Catalan. Our results show that augmenting real AV data with synthetic samples yields relative Word Error Rate (WER) reductions of up to 16.2%, demonstrating the potential of this approach. Moreover, we demonstrate that synthetic data alone can serve as a baseline for AVSR training in languages lacking AV datasets. These findings provide evidence that synthetic visual data can serve as a scalable solution to AVSR data scarcity, enabling broader language coverage.
[NLP-361] Optimal transport meets speech: a tutorial review
【速读】： 该论文旨在解决生成式模型与跨域/跨模态语音任务中因分布不匹配（distribution mismatch）导致的性能下降问题，尤其针对语音信号特有的时序动态性、说话人差异、噪声与混响干扰以及多模态表征异质性等挑战。其核心解决方案在于系统性地推广最优传输（Optimal Transport, OT）在语音处理中的应用，关键在于：通过直观的物理类比阐释OT的理论基础，并揭示其与现代生成模型（如生成式AI）之间的深层联系；设计适配深度学习框架的高效计算算法（如基于Sinkhorn迭代的近似方法），实现可微分优化与端到端训练；进而实证展示OT在语音增强、自动语音识别、语言与说话人识别、音频伪造检测等跨域与跨模态任务中的有效性，从而为处理真实场景下复杂的分布变化提供一个兼具几何保真性与泛化能力的统一建模框架。
链接: https://arxiv.org/abs/2609.31787
作者: Xugang Lu,Yu Tsao
机构: National Institute of Information and Communications Technology (日本信息通信技术国家研究所); Research Center for Information Technology Innovation, Academia Sinica (台湾中央研究院资讯科技创新研究中心)
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Optimal Transport (OT) provides a principled framework for comparing and transforming probability distributions while preserving geometric structure. Recently, OT has gained significant attention in machine learning due to its ability to measure discrepancies between distributions, even when their supports do not overlap, making it effective for tasks such as generative modeling, domain adaptation, and transfer learning. Despite its success in fields such as computer vision and natural language processing, OT remains relatively underexplored in speech research. Speech signals present unique challenges, including temporal dynamics, speaker variability, noise, reverberation, and heterogeneous multimodal representations involving audio, text, and visual information. These factors often lead to distribution mismatches, where OT offers a natural framework for alignment and interpretation. This work aims to promote broader adoption of OT in speech processing by: (1) reviewing OT foundations through intuitive physical interpretations and highlighting connections to modern generative models; (2) presenting computational algorithms suitable for deep learning frameworks; and (3) demonstrating OT applications in cross-domain and cross-modal speech tasks, including speech enhancement, automatic speech recognition, language and speaker recognition, and audio spoof detection. We highlight OT’s strong potential for addressing distributional variations in real-world speech applications.
[NLP-362] Distributional Metrics for Evaluating Spoken Conversational Systems ICASSP2027
【速读】： 该论文旨在解决对话系统评估中缺乏有效、可解释评价方法的难题。现有评估体系难以全面捕捉对话行为的本质特征，尤其在衡量生成对话与人类对话在自然性与交互质量上的差异时存在局限。为此，作者提出对话分布评分（Conversational Distribution Score, CDS），其核心是通过将对话系统的输出与真实人类对话的分布进行对比，实现对对话行为的系统性评估。CDS的关键在于利用八项可解释的特征来刻画语速、音节节奏和话语轮次交互等语言行为模式，并引入两项独立的语义基线特征以补充内容层面的评估。研究采用两种参考基准：一是基于人类对话成功性的标准，二是对比人类与合成对话的差异。通过跨领域任务导向对话中的听者判断数据，验证了CDS在系统排名、对话偏好判断及排名稳定性方面的有效性。结果表明，综合CDS能恢复六组听者比较中的五组，而各单项特征亦与听者偏好呈现显著相关性。此外，研究还量化了达到稳定排名所需的最小对话时长与数量，证实了分布式比较方法在补充传统交互指标的同时，具备高度可解释性与实用价值，为对话模型评估提供了新的范式。
链接: https://arxiv.org/abs/2609.31719
作者: Shree Harsha Bokkahalli Satish,Erica Cooper,Patrícia Schmidtová,Maike Züfle,Éva Székely,Nicholas Sanders,Ondřej Klejch
机构: 未知
类目: Audio and Speech Processing (eess.AS); Computation and Language (cs.CL); Sound (cs.SD)
备注: 5 pages, 3 figures, 1 table. Submitted to IEEE ICASSP 2027
Abstract:Evaluating conversational systems is a difficult and unresolved problem. We introduce the Conversational Distribution Score (CDS), which compares distributions of conversational behaviour using human conversations as a reference. CDS describes speech rate, syllabic rhythm, and turn interaction through eight interpretable features plus a separate two-feature semantic baseline. We compare conversations with two reference scales: one based on conversational success within human dialogue and another contrasting human and synthetic dialogue. Using listener judgments from out-of-domain goal–oriented dialogues, we examine system ranking, preferences between conversations, and ranking stability. Composite CDS recovers five of six listener system comparisons while individual features show strong correlation with listener preferences between conversations. We examine how many minutes and conversations are required before rankings stabilize. These findings support distributional comparisons as a complement to specific interactional metrics to evaluate conversations and conversational models while showing their interpretable value.
信息检索
[IR-0] Rubric-Calibrated Preferences: Cross-Query Calibration of LLM Judgments via Item Response Theory
链接: https://arxiv.org/abs/2609.35739
作者: Fabian David Schmidt,Donato Crisostomi,Carlos Lassance,Nils Reimers
类目: Information Retrieval (cs.IR)
备注: 51 pages. Code and data: this https URL
Abstract:Rerankers decide which documents users and LLMs see, yet their standard metric, nDCG, relies on human relevance labels that are costly, sparse, noisy, and discretely graded. As rerankers approach each other in quality, nDCG on these labels therefore increasingly fails to separate them. LLM judges could supply dense labels. Relative judgments within one query tell even close candidates apart, yet their scores share no scale across queries. Absolute grades share one scale but are too coarse to distinguish documents of similar relevance. We propose Rubric-Calibrated Preferences (RCP), which combine both kinds of judgment. A listwise Bradley-Terry tournament orders each query’s documents, and a rubric of yes/no criteria of increasing stringency provides an absolute standard. Item Response Theory (IRT), which scores test-takers based on their answers to common questions, then uses the shared criteria to put all queries’ tournament scores on one scale. RCP’s retrieval metric, RCP-nDCG, replaces nDCG’s discrete labels with the resulting calibrated relevance probabilities. Against blind grades from 46 external annotators, calibration raises the correlation between a query’s mean score and its mean human grade from 0.538 to 0.795. The probabilities rank a useful document above a non-useful one with probability 0.910 (AUC, chance 0.5), versus 0.651 for the benchmark labels. When the annotators’ grades prefer one of two rerankers and exactly one metric agrees, that metric is RCP-nDCG in 72.4% of 185 comparisons (chance about 53%). On TREC-DL, RCP-nDCG sides with NIST assessors’ grades on every reranker pair that these grades separate significantly. RCP-nDCG also resolves many of nDCG’s ties and separates 1.9 times as many reranker pairs on NanoBEIR. Rubric calibration thus turns relative LLM judgments into dense relevance labels that are comparable across queries and agree with human judgment.
[IR-1] Can Generative Retrievers Learn Semantic IDs Without Forgetting How to Speak?
链接: https://arxiv.org/abs/2609.35430
作者: Junchen Fu,Kleomenis Katevas,Vandana Rajan,Sofía Celi,Hamed Haddadi
类目: Information Retrieval (cs.IR)
备注:
Abstract:Generative retrieval (GR) enables end-to-end retrieval by generating document semantic identifiers (SIDs). However, retrieval-only fine-tuning can over-specialize pretrained language models to SID prediction, substantially distorting their natural-language distribution and limiting their suitability for interactive systems that must both retrieve documents and generate natural-language responses. We introduce SpeakGR, a dual-objective framework that learns SIDs while preserving language generation. It combines supervised SID learning with speak-preserving regularization: an on-policy distillation objective that aligns the current model with a frozen copy of the original model on student-generated prefixes using forward KL over the original text vocabulary. We further propose Adaptive SpeakGR, which dynamically adjusts the preservation strength based on observed language drift. Compared with SFT-only, SpeakGR reduces WikiText-2 forward KL by 81.3-93.8% on MS MARCO and 81.2-85.2% on Natural Questions (NQ) while retaining effective retrieval across three different LLMs. Adaptive SpeakGR further improves retrieval over SpeakGR in most settings while maintaining substantially lower language drift than SFT-only.
[IR-2] Signal or Noise? Modality Contribution and Cooperation in Multimodal GraphRAG
链接: https://arxiv.org/abs/2609.35304
作者: Antonios Georgakopoulos,Paul Groth,Lise Stork
类目: Information Retrieval (cs.IR)
备注:
Abstract:Multimodal knowledge graphs (KGs) integrate information from text, figures, tables, and other modalities into a unified structured representation, with the promise that richer evidence enables better inference. In GraphRAG systems built over such graphs, it is commonly assumed that retrieving evidence from more modalities at inference time improves downstream performance. Yet, redundant or overlapping multimodal evidence may distract language models in question answering (QA), and whether each modality contributes equally across questions, models, and tasks remains poorly understood. In this work, we study how modality-aware retrieval affects downstream inference in a multimodal GraphRAG pipeline, using document visual question answering (DocVQA) as a testbed. We extend an existing KG-based QA framework to be modality-aware, leveraging the graph structure to track which modality supports which facts and to selectively filter evidence at the edge level. This enables us to investigate whether providing all available multimodal evidence at inference time benefits QA, and to evaluate the contribution and cooperation of modalities across question, task, and model characteristics. Through a controlled analysis within a state-of-the-art multimodal GraphRAG pipeline, five multimodal LLMs and two DocVQA benchmarks, we find that tables and text provide the strongest contributions, and that combining modalities frequently produces redundancy rather than synergy, particularly for pairs involving textual information. Positive cooperation appears mainly between non-text modalities and depends on question intent and task type. Our findings argue for selective, modality-aware retrieval in the design of more effective GraphRAG systems, where modalities are filtered according to the downstream task rather than retrieved uniformly.
[IR-3] 5W1HWhich: Context-Valid Semantic Indexing with Progressive Ontology Binding DATE
链接: https://arxiv.org/abs/2609.35184
作者: Yaxiao Liu(PwC China AI Center),Pengbo Liu(PwC China AI Center),Yiwen Liu(PwC China AI Center),Yihua Guan(PwC China AI Center),Jiaxing Song(Tsinghua University)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 20 pages, 3 figures, 4 tables. Preprint of a proposed indexing method with falsifiable hypotheses; not empirically validated
Abstract:Transforming raw data into queryable knowledge requires both early extraction of reusable information and explicit types, relations, and applicability conditions for particular tasks. If indexing selects content too early around a single business schema, later tasks may be unable to use information that was omitted. If the index retains only open-ended text, however, rule-based reasoning lacks checkable premises. We propose 5W1H+Which, a semantic indexing design that separates content extraction from ontology binding. The 5W1H questions organize source-grounded content units; Which points to versioned ontology elements and records mapping relations, scope, and validation status. Time, location, system environment, and participant roles are not merely retrieval labels: together, they constrain the contexts in which facts, bindings, and rules apply. Unbound content remains searchable, while bound content enters a formal reasoning path only after premise checks. The method further distinguishes business valid time, system knowledge time, and operational traces, and uses dependency records to support binding revalidation and the maintenance of derived conclusions. A worked example of migration from an on-premises server to a cloud environment illustrates the different treatment of world-state changes, ontology-version changes, and changes in rule applicability. We formulate three groups of falsifiable hypotheses concerning cross-task evidence coverage, control of contextual misuse, and incremental update cost. The planned evaluation includes a strong typed fact-graph baseline with the same evidence, temporal information, and budget, to test whether benefits arise from 5W1H organization, deferred binding, or additional information and engineering effort. The contribution is a testable indexing mechanism, not a claim to a new universal ontology or a demonstrated performance advantage.
[IR-4] RenderRank: Learning to Rerank Text with Compressed Visual Tokens
链接: https://arxiv.org/abs/2609.35069
作者: Seongtae Hong,Youngjoon Jang,Jungseob Lee,Hyeonseok Moon,Heuiseok Lim
类目: Information Retrieval (cs.IR)
备注:
Abstract:Rendering document text as images allows vision-language models to encode documents as visual tokens, which can reduce input sequence length compared with text input. This reduction in input length is particularly useful for reranking, where each query involves scoring multiple candidate documents and token savings apply to each candidate evaluation. We introduce RenderRank, a reranker that learns query-dependent relevance scoring from compressed visual document representations instead of the text token sequences used by conventional text-based rerankers. Training first aligns relevance scores from visual inputs with those of a text-based teacher, then refines the relative scores of positive and negative documents for the same query. Across 11 datasets from BEIR, RenderRank uses 16.5-35.5% fewer input tokens while achieving an average NDCG@10 of 55.96, outperforming all evaluated text-based baselines below 4B parameters and some larger models. Across four long-document datasets, it achieves an average NDCG@10 of 88.27 with approximately half the average input token count of the evaluated text-based rerankers. In this setting, RenderRank delivers 1.70x the highest average throughput of the evaluated baselines. These results demonstrate that compressed visual representations can support accurate document relevance scoring, providing an alternative to text token representations for reranking.
[IR-5] Mitigating Popularity Bias in Recommendation with Global Listwise Learning and Progressive Bi-Weighting
链接: https://arxiv.org/abs/2609.35041
作者: Tianyu Zhu,Jiandong Ding,Yansong Shi,Guoqing Chen,Jian-Yun Nie
类目: Information Retrieval (cs.IR)
备注: Accepted at ACM TOIS
Abstract:In recommender systems, user feedback typically follows a long-tail distribution, which leads many recommendation algorithms to exacerbate popularity bias by disproportionately favoring popular items. To mitigate this issue, recent studies have employed Inverse Propensity Scoring (IPS) to rebalance training data via reweighting user-item interactions. However, the effectiveness of IPS-based approaches is often constrained by locally unbiased objectives and inaccurate propensity estimation. In this paper, we propose Multinomial Likelihood with Bi-Weighting (Mult-BiW) to address these limitations. First, we introduce a debiasing framework, termed Mult-IPS, which integrates multinomial likelihood with IPS to capture global and unbiased user preferences over the entire item set. Second, we develop a Bi-Weighting (BiW) strategy that jointly leverages propensity scores and a collection model, incorporating a smoothing mechanism to enhance the robustness of propensity estimation. We further provide theoretical analyses that establish an upper bound on the empirical bias and characterize the optimal form of the collection model. Third, to mitigate the adverse effects of aggressive reweighting on representation learning, we design a Progressive Bi-Weighting strategy that gradually transitions from discriminative representation learning to popularity debiasing. Extensive experiments on real-world datasets show that Mult-BiW consistently outperforms state-of-the-art baselines.
[IR-6] Recommendation Ranking Off-Policy Evaluation under Ranking-Dependent Examination via Examination-Relevance Decomposition
链接: https://arxiv.org/abs/2609.35034
作者: Riki Okamura,Toshiharu Sugawara
类目: Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: 20 pages, 6 figures,
Abstract:Off-policy evaluation, which estimates evaluation policy performance from logged data, is key for recommender ranking policies. However, logged clicks cannot distinguish unexamined items from examined non-clicks, causing bias in existing estimators when the assumed examination structures fail. We propose two estimators based on the decomposition of clicks into examination and relevance. First, the latent-examination independent inverse propensity score (LE-IIPS) estimator corrects the IIPS bias using policy examination probability ratios. Second, the examination-decomposed doubly robust (ED-DR) estimator extends LE-IIPS to a doubly robust framework. ED-DR is unbiased if the examination probabilities are correct regardless of relevance accuracy, or under ranking-independent examination, even if both model estimates are inaccurate. Experiments show that ED-DR achieves a lower MSE than existing methods with large sample sizes, especially when the examination depends on ranking. We also highlight its limitations under small samples or cascade user behavior conditions.
[IR-7] PEAR: Progressive Evidence-Based AutoResearch for Industrial Search Systems
链接: https://arxiv.org/abs/2609.35031
作者: Yifan Wang,Shipeng Zhu,Fei Xiong,Yuqin Yang,Yonghui Huang,Kunyao Wu,Yue Wang,Weichao Meng,Yu Gong
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注: 20 pages, 2 figures, 6 tables
Abstract:AutoResearch improves systems through iterative experimentation: agents propose candidate modifications, evaluate them, and use the results to guide subsequent exploration. Applying this paradigm to industrial search presents two challenges. (1) Common AutoResearch approaches follow a keep-if-better rule, retaining the highest-scoring candidate for subsequent experiments. Under non-stationary traffic, transient gains may be mistaken for persistent improvements, impairing reliable accumulation of search knowledge. (2) Candidate modifications can be evaluated at multiple fidelity levels, from low-cost proxies to online validation, differing in cost, objective alignment, and statistical reliability. Existing methods rely on individual signals or task-specific procedures, lacking a unified basis for using evidence across levels to guide search. We introduce Progressive Evidence-Based AutoResearch (PEAR) with two complementary components. Evidence-driven AutoResearch maintains an independent, hypothesis-guided research state for each strategy task within a predefined objective and intervention scope. Each state evolves through a Plan-Execute-Evaluate-Update transition that links experimentation to context-aware evidence interpretation and hypothesis revision. Confidence-Gated Verifier Ladder organizes evaluation into four levels of increasing fidelity: Offline Replay, Shadow-Traffic Evaluation, Rapid Online Evaluation, and Decision-Grade Online Evaluation. A unified confidence-based gate promotes candidates only when evidence supports a statistically significant positive effect, enabling broad low-cost exploration while reserving costly online experiments for promoted candidates. In a real-world industrial search system, strategies optimized with PEAR significantly increased Main Order/DAU by 2.7336% and 3.2957% relative to their respective baselines in two A/B experiments.
[IR-8] VEX-Bench: Benchmarking Verification Complexity of LLM -Generated Misinformation NEURIPS2026
链接: https://arxiv.org/abs/2609.35028
作者: Hanxun Huang,Yutao Wu,Qizhou Wang,Silvia Montaña-Niño,Yige Li,Xiang Zheng,Elif Buse Doyuran,Phoebe Matich,Xiao Liu,Xingjun Ma,Sarah Erfani,Christopher Leckie
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computers and Society (cs.CY); Information Retrieval (cs.IR)
备注: NeurIPS 2026
Abstract:Large language models (LLMs) have made misinformation inexpensive to produce but not to verify, creating a growing asymmetry in the information ecosystem. Under tight time, labor, and budget constraints, media organizations, platforms, and fact-checkers rely on screening to prioritize which content to verify. We introduce VEX-Bench, a unified benchmark for evaluating the verification complexity of LLM-generated misinformation, as perceived during screening, across models and generation methods. Verification complexity is assessed along multiple dimensions derived from journalistic and fact-checking practices, capturing checkability, harm potential, source credibility signals, imposter legitimacy, and expected verification effort. We define the VEX score as an integrated measure combining elicitation yield and verification complexity to quantify how generated content consumes limited verification capacity. We construct a benchmark spanning two misinformation categories, 6 high-stakes domains, and 60 real-world topics, and evaluate 7 frontier LLMs and 7 generation methods, yielding 5,880 articles. We employ an LLM-as-judge for scalable evaluation and validate it using content-analysis methodology, including ordinal Krippendorff \alpha for inter-annotator reliability, complemented by fact-checking agents for verification. Our findings show that no single method dominates all dimensions, underscoring the need for multi-dimensional evaluation. LLMs can generate high-VEX misinformation at 3 \times to 169 \times lower cost than agent-based verification. Such content is often prioritized during screening, consuming scarce verification resources and introducing a systematic risk of misallocation in resource-constrained verification systems. The code is publicly available in our \hrefthis https URLGitHub repository.
[IR-9] AX is the New AEO
链接: https://arxiv.org/abs/2609.34951
作者: Ido Finder,Assaf Elovic,Gad Shalev
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 17 pages, 11 figures
Abstract:In 2023, AI models answered from training data and hallucinated when it ran out, and businesses were told to seed that knowledge. Models’ training knowledge has since given way to live web search, and the advice followed it there: answer-engine optimization, or AEO, now tells businesses to scatter breadcrumbs across forum threads, listicles, and off-site citations, so AI engines are likelier to surface and recommend them. But being surfaced is no longer enough: an agent opens the results and reads them before deciding, and one buyer question sends it through several rounds of search and fetch. What decides the outcome at this drill-down step is whether the agent can fetch and read the business’s own site: agent experience (AX). We argue that AX is the new AEO. We run 37,927 agent journeys, each a buyer question about a business, across four independent harnesses over 1,056 real businesses, matched on fame, prior model knowledge, and two AEO proxies, then split based on their AX level. Only 7-10% of the finished answer comes from the model’s training knowledge, whether or not the site is readable. Agent-ready businesses have answers built from their own pages 78% of the time against 56% and are clearly recommended 1.9x more often, while every grounded answer about a not-agent-ready business costs the agent 64% more. Holding business, harness, and question fixed, answers built from the site are 41% more accurate. The dominant failure is not fabrication but omission: web-built answers are 3.7x more likely to contain none of the facts the buyer asked for. Baselines differ sharply across the four harnesses, with clear-recommendation rates varying sevenfold from stack to stack, yet the effect holds in every one. In the agentic web era, being readable beats being talked about, and improving a site’s AX is the strongest lever a business has.
[IR-10] ColNanoVDR: Document-Free Query Distillation for Multi-Vector Visual Document Retrieval via Optimal Transport
链接: https://arxiv.org/abs/2609.34899
作者: Zhuchenyang Liu,Ziyi Wang,Yao Zhang,Yu Xiao
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 20 pages, 5 figures, 11 tables. Code: this https URL ; Models: this https URL
Abstract:Multi-vector retrievers built on vision-language models lead visual document retrieval (VDR), but they run a multi-billion-parameter query encoder on every search. Distilling this encoder into a small student that queries the teacher’s existing index would remove the bottleneck. The standard recipe, however, matches the teacher’s MaxSim scores and so requires encoding and caching every training page, which can reach terabytes of page tokens. NanoVDR avoids pages entirely by training on the teacher’s query embeddings alone, but only for single-vector retrievers. We present ColNanoVDR, to our knowledge the first framework to bring this document-free distillation to multi-vector VDR. Its objective, OTW (Optimal Transport with Learned Weights), aligns the student’s query tokens with the teacher’s by entropic optimal transport, with a learned weight for each student token, and needs no correspondence between the two tokenizations. We prove that the resulting alignment cost bounds the MaxSim score difference on every page. Distilled from five state-of-the-art teachers, the 149M text-only students retain about 95% of their teachers’ NDCG@5 on ViDoRe v1-v3 while encoding queries up to 26x faster. Under identical training, OTW matches score distillation while encoding no page and reading 12.6x less cached teacher data.
[IR-11] No Attention No Problem: Rethinking Session-based Recommendation with Pure Convolution
链接: https://arxiv.org/abs/2609.34802
作者: Tao Huang,Wei Zhou
类目: Information Retrieval (cs.IR)
备注:
Abstract:Session-based recommendation (SBR) predicts the next choice in a session by analyzing recent interactions. Transformer-based models are widely used because of their ability to capture long-range dependencies through self-attention mechanisms. In contrast, traditional convolutional models, although more efficient, are often limited by their weak global modeling capabilities and are losing ground in SBR tasks. In this work, we propose a Next-generation Pure Convolutional Framework (NextConvRec) for SBR tasks, aiming to balance efficiency and performance. NextConvRec uses a Structural and Positional Convolutional Encoder (SPCE) for preprocessing, combining learnable convolutional positional biases with session-level structural signals extracted through GCN layers. Its backbone convolutional module effectively expands the effective receptive field through depthwise convolutions and pointwise convolutions, enabling robust long-range preference modeling without attention mechanisms. Extensive experiments on 4 benchmark datasets show that NextConvRec outperforms several state-of-the-art baselines by around 1.73% on average, and reduces the average inference time per session by 16.7%. The convolutional architectures remain a promising direction for efficient and accurate session-based recommendations.
[IR-12] Calibrated Uncertainty for Informative Path Planning in Aquatic Environmental Monitoring
链接: https://arxiv.org/abs/2609.34577
作者: Samuel Yanes Luis,Alejandro Casado Pérez,Alejandro Mendoza Barrionuevo,Dame Seck Diop,Sergio Toral Marín,Saniel Gutiérrez Reina
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注:
Abstract:Informative Path Planning for scalar field reconstruction uses predictive uncertainty to direct sensing vehicles toward maximally informative locations. Gaussian Processes provide this signal but their stationary isotropic kernels are misspecified for non-homogeneous phenomena such as oil spills, producing miscalibrated estimates that degrade planning. We investigate whether replacing the Gaussian Process with a well-calibrated Deep Ensemble improves path planning outcomes, and whether uncertainty quality interacts with the choice of planning algorithm. Five strategies ( \epsilon -Greedy, Value Greedy, Uncertainty Greedy, Monte Carlo Tree Search, and Receding Horizon Orienteering) share a common Deep Ensemble backbone trained on physics-based oil spill simulations. On held-out stochastic spill scenarios, the Deep Ensemble reduces normalised reconstruction error by 83% relative to the Gaussian Process baseline. Crucially, well-calibrated uncertainty amplifies the importance of the planning strategy: the performance gap between algorithms is negligible under miscalibrated models but becomes substantial under the ensemble, where multi-step lookahead planners outperform greedy selection by up to 32% in reconstruction error and achieve IoU above 0.85 . Monte Carlo Tree Search is the recommended planner, matching Orienteering in reconstruction quality at an order-of-magnitude lower computational cost.
[IR-13] EvoSkillRec: Skill-Genome Evolution for Recommender Architecture Discovery
链接: https://arxiv.org/abs/2609.34552
作者: Xiaopeng Li,Kuo Cai,Bo Chen,Wenlin Zhang,Mengyang Ma,Yingyi Zhang,Zichuan Fu,Yu Yang,Qidong Liu,Yiyu Wang,Ruiming Tang,Wenwu Ou,Jiang Wu,Zhanbo Xu,Xiangyu Zhao
类目: Information Retrieval (cs.IR)
备注:
Abstract:Modern recommender systems advance not only by scaling data and parameters, but also by encoding task-specific inductive biases through architecture, including sparse feature interactions for click-through rate (CTR) prediction, temporal attention for sequential recommendation, and expert routing for multi-task learning. However, these biases are typically human expert designed or searched within predefined operator spaces. Although Recent LLM-driven code evolution expands this space, unconstrained edits often produce invalid or ineffective architectures, underuse established architecture design knowledge, and fail to preserve successful innovations for reuse. We introduce EvoSkillRec, a promotion-and-reuse framework for cumulative recommender architecture evolution. It first decomposes recommenders into atomic executable skills and represents architectures as typed skill genomes, with each skill equipped with input–output types, semantic annotations, and implementation code. We then evolve models with different tasks through two coupled spaces: a constrained skill–space that mutates, recombines, specializes, and reuses validated skills, and an open-ended code–space in which LLM planners and synthesizers invent new skill modules using prior evolution traces and accumulated experience. An autoresearch controller evaluates candidates, diagnoses failures, retrieves relevant skills, promotes validated innovations into the skill library, and adaptively allocates the proposal budget between the two spaces. Extensive experiments on CTR prediction, multi-task learning, and multi-domain learning, including resource-constrained co-optimization of predictive quality and model FLOPs utilization in generative ranking models, consistently demonstrate the effectiveness of our proposed EvoSkillRec.
[IR-14] Eval4DiRec: A Unified and Systematic Evaluation Framework for Diffusion-based Recommender Systems KDD
链接: https://arxiv.org/abs/2609.34404
作者: Cong Wang,Shoujin Wang,Yishuo Li,Qi Zhang,Liang Hu,Wenpeng Lu
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注: Accepted by ACM Transactions on Knowledge Discovery from Data (TKDD)
Abstract:Leveraging the strong generative capabilities and stable training dynamics of diffusion models, diffusion-based recommender systems (RSs) have recently emerged as a novel recommendation paradigm, attracting increasing attention from both academia and industry. However, despite the rapid growth of diffusion-based RSs, a critical issue has emerged: the lack of a unified and systematic quantitative evaluation benchmark, which often results in irreproducible experimental results and unfair comparisons across studies due to inconsistent data processing, training configurations, inference procedures, and evaluation protocols. To address this challenge, we propose Eval4DiRec, the first unified and open-source evaluation framework specifically designed for diffusion-based RSs. Eval4DiRec supports 14 representative diffusion-based RS models across five different recommendation scenarios, providing consistent and reproducible experimental settings to systematically assess their performance. Built upon this framework, we conduct extensive empirical studies to benchmark these models under unified protocols. The results highlight the strong potential of diffusion models for recommendation while also revealing key factors and practical challenges that substantially affect their performance, thereby establishing a solid foundation to facilitate fair evaluation and guide future research in this promising field. Our code and data are available at: this https URL.
[IR-15] Relevance-Resolution Transfer via Scale-Decomposable Fractional Diffusion for Multi-Length Cross-Modal Hash Retrieval
链接: https://arxiv.org/abs/2609.34393
作者: Xi Chen,Xu Chen,Xiangyang Jia,Ting Gan,Xu Zhang,Shuquan Wei,Sitong Fan
类目: Information Retrieval (cs.IR)
备注: 28 pages, 7 figures
Abstract:Cross-modal hashing enables efficient retrieval by encoding heterogeneous data into compact binary codes. Recent methods exploit fine-grained relations encoded in multi-label training structure, yet none of them constrains how those relations survive as consistent candidate rankings in finite, multi-length Hamming spaces, which we term the relevance resolution bottleneck (RRB). To address the RRB, we propose MultiBit, which transfers relevance resolution from multi-label structure to multi-length Hamming spaces. MultiBit first constructs a scale-decomposable fractional relation teacher from dataset-level label co-occurrence and label specificity, and models dependencies from local to long-range over continuous diffusion scales. It then maps the discretized diffusion scales and their quadrature weights to scale-aware bit subblocks of the maximum-length code, organizes the target code lengths as nested prefixes, and aligns their Hamming candidate rankings with the teacher relations. Experiments on multiple benchmarks demonstrate improved retrieval accuracy. Code is available in the supplementary material.
[IR-16] Just-In-Time Agent Memory with Runtime Agent ic Research
链接: https://arxiv.org/abs/2609.34385
作者: Bingyu Yan,Chaofan Li,Hongjin Qian,Shuqi Lu,Chaozhuo Li,Zheng Liu
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注:
Abstract:Memory is critical for AI agents. Many existing agent-memory systems follow an Ahead-of-Time (AOT) design, constructing memory before a specific request arrives. While this reduces online serving cost, such request-agnostic memory construction can discard fine-grained information that later becomes important. To address this limitation, we propose Just-In-Time Agent Memory (JAM), a trainable framework for query-conditioned context construction at runtime. A Memorizer preserves complete raw histories in a hierarchical page-store with compact navigational summaries, while a Researcher iteratively retrieves, inspects, and integrates evidence for each request. To train these memory-use behaviors, we introduce Memory-Gym, an evidence-grounded data synthesis pipeline covering nine task types across six domains, and optimize the Researcher through verified-trajectory supervised fine-tuning followed by Hint-guided Group Relative Policy Optimization. We demonstrate the effectiveness of JAM across a variety of benchmarks on agent memory and long-context processing, where it achieves stronger task performance than AOT-style memory systems while remaining substantially more efficient than prior trained agentic memory approaches. To support reproducibility and future research, we release our anonymized source code at this https URL.
[IR-17] Correcting to Predict: Pseudo-Value Correction for Multimodal Attribute Value Extraction CIKM2026
链接: https://arxiv.org/abs/2609.34383
作者: Junhao Zhang,Feiran Hu,Xiao Hu,Baoliang Cui,Xiaoyi Zeng
类目: Information Retrieval (cs.IR)
备注: Accepted by CIKM2026 Oral Full Paper
Abstract:Product attribute value extraction (AVE) is a fundamental task in e-commerce, aiming to identify specific values of predefined attributes from multimodal product profiles such as text and images. While multimodal large language models (MLLMs) have shown promise for AVE, they face challenges in extracting implicit attributes that require joint reasoning over visual and textual cues, often confusing semantically similar values. However, existing methods often fail to resolve such ambiguities because the correct value often depends on subtle multimodal cues that are easy to miss or override. To address this challenge, we propose Correcting to Predict (C2P), a framework that treats attribute extraction as a correction process. Given an initial pseudo-value such as a retrieved candidate or placeholder, the model learns to correct it using multimodal evidence. During training, diverse pseudo-values help the model learn evidence-based correction behavior, and a self-consistency refinement stage further reduces sensitivity to pseudo-value perturbations. At inference, a fixed placeholder triggers the learned correction behavior, enabling efficient single-pass prediction without online retrieval or iterative refinement. We evaluate C2P on a public benchmark and a large-scale industrial dataset. Offline results show that C2P outperforms strong baselines, with notable gains on ambiguous attributes. Online A/B tests on AliExpress further show consistent improvements in seller adoption, attribute completeness, and user engagement, validating C2P’s effectiveness and efficiency in real-world deployment.
[IR-18] When Harness Beats Scale and When Reading Beats Both EMNLP2026
链接: https://arxiv.org/abs/2609.34366
作者: Ivan Bondarenko,Nikolay O. Nikitin
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: Accepted at the DocInsights 2026 Workshop co-located with EMNLP 2026. System description paper for the DocSem document-grounded quantitative reasoning shared task. 10 pages, 2 figures, 7 tables, 5 appendices
Abstract:We describe our system for DocSem, the document-grounded quantitative reasoning shared task at DocInsights 2026, and analyze why it succeeded on labeled data and failed on the test set. The pipeline pairs hybrid block retrieval with Program-of-Thoughts (PoT) generation executed in a sandboxed interpreter, self-consistency sampling, and entity enrichment from chunk-level knowledge graphs. On our held-out split, application architecture moved the metrics far more than model scale did: PoT added 0.282 joint accuracy to a compact 7B model but at most 0.005 to a 72B model, and a 27B model with the full harness matched the 72B (0.884 vs.\ 0.873) at roughly 2.7 \times fewer parameters and a quarter of the CO _2 . We read this through a distinction between world knowledge, which scales steeply with parameters, and language knowledge, which scales gently, and show that structured-output training makes a compact model harness-ready rather than merely small. On the raster, watermarked test PDFs the same system collapsed to 13.58% joint (rank 149 of 163); a controlled re-rendering of the validation set reproduces the OCR half of the collapse while bounding what the simulation misses. Auditing the physical nature of evaluation inputs precedes architecture, and the leaderboard’s bimodality is consistent with reading quality, not reasoning, having separated the field.
[IR-19] SPRINT: Single-Step Generative Recommendation via Averag e Probability Velocity
链接: https://arxiv.org/abs/2609.34306
作者: Zhuo Cai,Shoujin Wang,Peilin Zhou,Min Xu,Julian McAuley,Fang Chen
类目: Information Retrieval (cs.IR)
备注:
Abstract:Semantic ID (SID) based generative recommendation represents each item as a sequence of discrete tokens, and recommends by generating the SID of the item a user would like to interact with. Both dominant paradigms in this domain generally pay for generation token by token: autoregressive models decode the tokens left-to-right, while non-autoregressive models decode in parallel yet still need multiple rounds of refinement to stay competitive. Therefore, both generally spend multiple forward passes per item, a cost that is prohibitive in latency-sensitive recommender systems. We ask whether an item can be generated in a single forward pass, and answer it through a new perspective which we call average probability velocity. We view SID generation as a flow of token generation probabilities and characterize it by its average velocity over the whole generation process. We prove that this average velocity is fully determined by the average generation probability of each token. Therefore, we directly parameterize and learn the probabilities of all tokens in a single forward pass with a bidirectional Transformer. As these probabilities are generated independently across positions and the coherence among tokens is lost, we further design a dual-level flow contrastive objective to restore the coherence among an item’s tokens. It contrasts the target SID against negative SIDs at both the token and SID levels. The token level ranks the generation probabilities of the target tokens above those of negative SIDs, while the SID level scores the tokens of each SID as a whole item for capturing token coherence of each item. Extensive experiments show that our model not only generates recommendations far more efficiently ( 8.39-10.04\times speedup over the second-fastest AR/NAR method) but also attains superior recommendation accuracy ( 7.77% average improvement over the second-best.
[IR-20] Measuring and Mitigating Identity-Cue Preference Drift in LLM -based Recommender Systems
链接: https://arxiv.org/abs/2609.34229
作者: Zhuoxiong Gan,Qiang Dong
类目: Information Retrieval (cs.IR); Social and Information Networks (cs.SI)
备注: 11 pages, 1 figure
Abstract:In large language model-based recommender systems, identity cues embedded in prompts can steer recommendations toward group-level patterns even when the underlying behavioral evidence remains unchanged. We introduce PromptShift, an interpretable, training-free framework for quantifying and mitigating such identity-cue preference drift. We define Drift as the divergence, in both item membership and ranking order, between a recommendation list generated under an identity-cued prompt and the reference list produced from the same user’s interaction history alone. SliceShift then measures the extent to which a cued list gravitates, relative to the history-only reference, toward items that are more popular within the cued slice than among the global user population. Beyond conventional accuracy, we propose DifHitRate, a difficulty-weighted hit metric that credits only relevant items, assigning higher credit to hits that are less popular within the cued slice and ranked higher in the list. All components are supported by an identity-slice-by-item table constructed from positive interactions, which further enables an adaptive post-hoc reranking strategy: the reranker interpolates between the original LLM ranking and inverse slice-popularity, with personalized interpolation weight. Experiments on two datasets with three LLMs show that identity-cued prompts incur higher mean Drift than identity-free paraphrase controls, an effect beyond generic wording sensitivity, and that SliceShift is positive across all six dataset-model settings. PromptShift consistently reduces both Drift and SliceShift, lowering macro-mean SliceShift by 62.42%, while improving DifHitRate, HitRate and MRR. These results demonstrate that identity-cue preference drift can be measured and mitigated without any model training, albeit with a modest, metric-dependent utility cost.
[IR-21] When Does Selection Replace Extraction? A Pre-Registered Test of Agent Memory with a Typed Decision Model
链接: https://arxiv.org/abs/2609.34227
作者: Rishabh Sharma,Rishika Lall
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 21 pages, 9 figures. Pre-registered: plan doi: https://doi.org/10.5281/zenodo.22970745 , amendment doi: https://doi.org/10.5281/zenodo.22977848 . Preprint also at doi: https://doi.org/10.5281/zenodo.22985242 . Code and data: this https URL
Abstract:Does conversational memory need LLM-extracted facts, or is selecting the right raw turns enough? Published results disagree. Extraction-based systems report gains from distilled facts. Recent studies find raw history with good ranking does as well, but disagree about whether ranking matters. We ran a pre-registered study on held-out LoCoMo conversations and LongMemEval. At a tight budget on LoCoMo, raw turns selected by a single call to Jev, a typed decision model, are non-inferior to an LLM-extraction memory (one-sided 95% bound -3.0 points against a -5-point margin). Blind human grading narrows the margin but does not change the result. Raw turns cost 3,061 times less to write, and the result holds with a second answer model. Within this study, reranking’s gain shrinks as the budget grows. It adds 17.4 points on LoCoMo and 9.1 on LongMemEval when three of 30 candidates are kept. At generous budgets it adds 1.5 and 1.1, and extraction systems are more accurate. This suggests why published results disagree. At matched context, Jev selects as accurately as an LLM reranker (non-inferiority bound -2.0) at a third of the latency, and more accurately than a multi-call graph traversal. Reranking lowers correct abstention. Plans, code and graded answers are released.
[IR-22] RidgeRank: Efficient Visual Document Reranking via Score Fusion and a Shallow Linear Readout
链接: https://arxiv.org/abs/2609.34192
作者: Shubing Yang,Dongfang Zhao
类目: Information Retrieval (cs.IR)
备注:
Abstract:Multimodal language models rerank visual document retrieval results accurately, but scoring every candidate page at full cost makes them slow. Some methods that compress these rerankers need relevance labels to regain accuracy, and they rank by the reranker score alone. RidgeRank measures how much relevance signal the reranker score lacks and recovers it from the retriever score through a closed-form fusion rule. Maximizing a correlation objective gives the optimal fusion weight, along with the exact condition under which the reranker score by itself cannot reach that optimum. The reranker is further corrected by a single vector applied to an intermediate hidden state, obtained through one centered ridge regression onto the same model’s full-depth scores on uncompressed pages. On 12 datasets drawn from ViDoRe 2 and ViDoRe 3, evaluated with two retrievers and two language model backbones, RidgeRank brings NDCG@5 to within 1.2 pp of a full cross encoder with speedups of up to 48 times, advancing the accuracy and latency Pareto frontier for visual document reranking.
[IR-23] STITCH-RAG : Spatio-Temporal Influence Tracing over Topic Hypergraphs for Multi-Hop Retrieval-Augmented Generation
链接: https://arxiv.org/abs/2609.34127
作者: Haodong Yang,Mengzhu Chen,Jia Cai
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-hop retrieval-augmented generation requires a retriever to connect evidence distributed across documents while preserving a concise, faithful generation context. Existing indexes leave two complementary gaps: chunk-based RAG can break cross-passage evidence chains, whereas an unlabeled pairwise projection without generating-topic provenance cannot jointly preserve topic-level co-participation and per-occurrence entity descriptions. We propose STITCH-RAG, a hypergraph-based framework with three coupled components. First, a semi-merged topic hypergraph encodes multi-entity co-participation as topic-summary hyperedges while retaining per-chunk entity states linked by canonical-name equivalence. Second, spatio-temporal influence bridging propagation (STIBP) combines topic-space propagation with deterministic chunk-index linkage across name-equivalent states under frequency-adaptive decay. Third, continuous STIBP scores replace binary entity-match seeds in localized Personalized PageRank (PPR). We characterize the condition under which this prior assigns more PPR mass to ground-truth evidence than a binary prior. Under the reported protocol, STITCH-RAG attains the highest reported Contain-Acc and LLM-Acc point estimates among the compared methods on HotpotQA and 2WikiMultiHopQA, and higher Recall@8 than the methods included in the standardized retrieval comparison. Results on the mixed-domain benchmark remain auxiliary preference-based evidence because only LLM-judged accuracy is available.
[IR-24] High-Level Text Preprocessing for Semantic Similarity Analysis of Discursive Texts: A Framework and Empirical Demonstration
链接: https://arxiv.org/abs/2609.33983
作者: Mehmet Murat Albayrakoglu,Mehmet Nafiz Aydin
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: 18 pages, 8 tables, 39 references; submitted for publication
Abstract:Semantic Textual Similarity (STS) methods assume that a document’s lexical content faithfully represents what it asserts. This assumption fails for discursive documents that discuss, compare, critique, and contextualize other positions in the process of articulating their own. The result is semantic diffusion: similarity scores between documents are inflated by vocabulary acquired through discursive engagement rather than substantive alignment. Standard Natural Language Processing (NLP) preprocessing (tokenization, stopword removal, stemming, lemmatization) cannot address this problem because it operates at the lexical level, treating all content identically regardless of its discursive function. This paper introduces high-level text preprocessing: a systematic, rule-based intervention applied before the standard preprocessing pipeline to isolate each document’s actual claim from its discursive structure. We propose 12 rules, each with an explicit rationale, and demonstrate their effect on an encyclopedic philosophical corpus: three entries from the Stanford Encyclopedia of Philosophy (virtue ethics, deontological ethics, and consequentialism). A three-phase experiment using eight Transformer-based STS models shows that preprocessing reduces centroid cosine similarity scores across all three theory pairs, with 23 of 24 model-pair comparisons showing the expected decrease and cross-model agreement ranging from 7-1 to 8-0. We introduce the semantic diffusion index (SDI), a per-document metric for assessing the semantic reorientation between a document’s raw and high-level preprocessed representations. Although the framework is demonstrated using philosophical texts, it potentially addresses a domain-agnostic problem applicable to legal texts, policy documents, academic articles, and any genre in which a discursive approach introduces vocabulary from positions the document does not endorse.
[IR-25] Beyond Fixed Features: Architecture-Dependent Sensitivity to Node Representations under Heterophily
链接: https://arxiv.org/abs/2609.33764
作者: Priyanath Maji,Sidharth Gaur,Rajavinoth Paul Durai
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: Accepted to Learning on Graphs Conference 2026
Abstract:Graph Neural Networks (GNNs) perform well on homophilic graphs but struggle in heterophilic settings, where connected nodes often carry dissimilar labels. Existing evaluations typically compare architectures under a fixed node-feature representation, leaving unclear whether conclusions about heterophily robustness remain stable as the input representation changes. We address this question by constructing parallel feature variants of two large-scale heterophilic benchmarks, Roman-Empire and Amazon-Ratings, pairing each graph with representations ranging from static fastText vectors to contextual Transformer embeddings and evaluating seven GNN architectures across these representations. We find that the effect of representation varies across architectures: on Roman-Empire, the contextual gain ranges from 2.38 percentage points for GCN-sep to 13.67 points for GAT, with H2GCN gaining 8.77 points. On Amazon-Ratings, where node text is limited to short product titles, GAT improves by 6.78 points from fastText to MPNet, while GCN-sep changes by only 0.20 points. These results show that architectural performance is conditional on node representation: the same representation change can produce different magnitudes of performance gain across architectures, so architecture and representation cannot be treated as independent evaluation factors. A rank-correlation analysis on these two benchmarks further shows that the relative ordering of architectures remains highly stable across representations, isolating differential sensitivity, rather than ranking instability, as the primary effect.
[IR-26] Concurrent Coded Signal-Multiplexing Ranging for Half-Duplex Asynchronous Networks
链接: https://arxiv.org/abs/2609.33753
作者: Zijian Zhang,Yuan Shen
类目: Information Theory (cs.IT); Information Retrieval (cs.IR); Signal Processing (eess.SP); Systems and Control (eess.SY)
备注: 15 pages, 11 figures
Abstract:Signal-multiplexing network ranging (SM-NR) shares broadcasts across node pairs, but its sequential operation leads to a ranging cycle that grows linearly with network size. This paper proposes a concurrent coded SM-NR (CC-SM-NR) framework for asynchronous half-duplex networks. Firstly, the CC-SM-NR protocol coordinates concurrent transmissions through binary transmit-listen codewords. The transmit-listen schedule defined by these codewords ensures reciprocal observations subject to a finite concurrency limit. Then, we derive the exact minimum number of transmit-listen rounds without a concurrency limit, which reveals that the minimum grows logarithmically with network size. To account for practical scenarios, we establish the necessary and sufficient conditions for the constant-weight feasibility of codewords under a finite concurrency limit. Subsequently, we propose a low-complexity scheduling algorithm that achieves the minimum round count within the constant-weight codeword class. To support higher observation redundancy, this scheduling design is extended through a greedy construction. Finally, simulation results demonstrate the effectiveness of the proposed schemes for network ranging.
[IR-27] Beyond the Beam: Constructive Repair and Candidate Completion for Generative Recommendation
链接: https://arxiv.org/abs/2609.33745
作者: Zijun Zhao,Peng Zhang,Gang Zhang,Yuanchi Ma,Hui He,Zhendong Niu
类目: Information Retrieval (cs.IR)
备注: 51 pages, 14 figures, including appendices
Abstract:Generative recommenders retrieve items by generating identifiers, but a valid identifier can remain outside the beam after catalog expansion. This raises two connected questions: which failures can identifier assignment repair, and how should retrieval proceed beyond the initial beam? We characterize assignment repair with a fixed generator and retained old identifiers. Output-invariance certificates identify failures shared by all admissible assignments. Under a common effective prefix, coupled support and ranking constraints give the exact feasible interval of new-item counts for target recovery. Building on this characterization, Beyond the Beam (BB) obtains minimum-replacement repairs through an integral flow formulation, selects a shared map and adapts the generator. At inference, generative likelihood and collaborative evidence define one score for ranking, candidate priority and stopping. Retained prefix bounds guide candidate completion and certify its global Top- K when the stopping condition is met. Exhaustive finite-catalog evaluation confirms construction in every feasible case. Across three Amazon Reviews categories and three random seeds, the full T5 procedure improves mean Recall@10 by 15.5–46.3% and NDCG@10 by 15.2–44.4% over the best-performing evaluated generative baseline for each dataset and metric. Matched controls show that shared construction and adaptation improve new-target ranking and certification efficiency on Beauty and Toys. Combined scoring and candidate completion improve NDCG@10 across all three datasets with both T5 and decoder-only LC-Rec.
[IR-28] Learning Multimodal Embeddings with Evidence-Aligned Readout
链接: https://arxiv.org/abs/2609.33659
作者: Zirong Chen,Fuda Ye,Enjun Du,Junfu Pu,Xinlei Wang,Xinyu Zuo,Lisheng Duan,Haijin Liang,Jin Ma,Jiachuan Wang,Yongqi Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Information Retrieval (cs.IR)
备注:
Abstract:Multimodal large language models can expose task-relevant evidence through generation, but producing useful evidence does not by itself determine how it enters a retrieval embedding. We study whether the semantic organization of that evidence can also specify where representations are read. To address this question, we introduce EviAlign, which couples Semantic Evidence Generation with Boundary Readout in a shared multimodal large language model. It organizes evidence into five semantic units, reads the contextualized state at each unit boundary, and aggregates these states into a single normalized embedding. Generation and contrastive retrieval objectives jointly train this shared structure. With the same trailing readout, semantic evidence and free-form CoT yield nearly identical retrieval performance, suggesting that evidence organization alone does not explain the full gain. A controlled 2\times3 study compares consistent and permuted evidence organization across three readout strategies, using training targets with matched evidence spans. With five readout states and the same mean pooling, the advantage of consistent semantic organization grows from 0.65 points at length-based training positions to 2.39 at evidence boundaries, yielding a 1.74-point co-design interaction. Across 12 MMEB retrieval tasks, EviAlign achieves 76.9 average Recall@1 with 500K training pairs while retaining single-vector indexing and scoring.
[IR-29] From PDF to Evidence: Structure-Aware Retrieval for Clinical Practice Guidelines ICASSP2027
链接: https://arxiv.org/abs/2609.33447
作者: Xingyu Lin,Dehui Du
类目: Information Retrieval (cs.IR); Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 2 figures, 5 tables. Submitted to ICASSP 2027
Abstract:Guideline documents are published as unstructured PDFs whose evidence is locked in visual structures—tables, flowcharts, and graded recommendations—that standard retrieval pipelines flatten into fixed-size text chunks. We cast evidence access as a document image analysis problem: parse each page image into typed structural elements, then retrieve structure-aware evidence units that follow the document’s own layout (sections, table rows, flowchart paths, graded recommendations), each keeping its structural context so a result points to a specific element rather than a page. On 26 clinical practice guidelines from 9 sources (3,619 pages, Chinese and English) with 199 evidence queries, structure-aware units rank the gold element first under BM25, dense, and hybrid retrieval (hybrid Element Hit@1 of 0.382), with a significant element-level ranking gain over per-element OCR text (MRR_e +0.107, p=0.002; the Hit@5 gain is directional, p=0.17), while matching page-level recall (Page Hit@5 0.879 vs. 0.889, p=0.75) at 3.8x less context and clearly outperforming a ColPali visual-RAG baseline (PH@5 0.497).
[IR-30] What Gets Measured Gets Managed: Sign-aware Recommendation Needs Sign-aware Evaluation
链接: https://arxiv.org/abs/2609.33346
作者: Minchan Kim,Jungmin Hwang,Hyunwoo Park
类目: Information Retrieval (cs.IR)
备注:
Abstract:Sign-aware recommender systems have recently been developed to leverage negative feedback for a deeper understanding of user preferences. However, our empirical diagnosis reveals that state-of-the-art graph-based sign-aware recommender systems are paradoxically valence-blind. Even though they explicitly incorporate sign information during training, they consistently fail to differentiate liked items from disliked ones at the ranking stage, frequently infiltrating top-K recommendations with disliked content. Through linear probing, we show that while valence information exists in the learned embeddings, it remains inaccessible to the inner-product scoring function. This widespread failure remains entirely undetected because conventional evaluation metrics, such as Recall, HR, and NDCG, assign a uniform utility of zero to both negative and unobserved items, creating a systematic evaluation blind spot. To bridge this gap, we propose a family of signed metrics, Signed Recall, Signed HR, and Signed NDCG, that explicitly penalize the recommendation of disliked content. Systematic re-evaluation under our proposed metrics fundamentally reshapes the established performance landscape, revealing that methods ranked highly under conventional metrics often fail to protect users from disliked content. Finally, through a proof-of-concept auxiliary loss, we confirm that the proposed metrics provide actionable training signals, guiding models toward valence-aware behavior without sacrificing conventional relevance. For transparency, our source code is available at: this https URL
[IR-31] Robust Hierarchical Structures for Agent ic Document Analysis SIGMOD2027
链接: https://arxiv.org/abs/2609.33322
作者: Ruiying Ma,Yiming Lin,Aditya G. Parameswaran
类目: Databases (cs.DB); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: To appear in Proceedings of the ACM on Management of Data (SIGMOD 2027). 25 pages
Abstract:Large Language Models (LLMs) enable us to better understand text documents, including PDFs and Word documents. However, LLMs, as well as more modern LLM agents, i.e., those with tool-calling abilities, typically treat such documents as plain text, ignoring the fact that they are often organized hierarchically into sections and subsections. Extracting this structure, while difficult, can improve efficiency and effectiveness for agents (and humans)—since only sections relevant to a given task need to be processed. Unfortunately, prior work on structure extraction provides no formal guarantees on how well the inferred structure matches the true one. Instead, we target a robust and compact variant that is feasible to infer and useful in practice. Robustness ensures that the text under each subsection header is a superset of the text under the same header in the true structure. Compactness seeks to minimize this superset, reducing agentic cost (or human cognitive load). We propose SHED, a two-stage workflow for inferring a robust and compact structure. The first stage is pluggable with an infinite family of approaches, each guaranteeing robustness for a specific document class. We theoretically characterize the document space using these classes and their hierarchical relationships. Empirically, SHED improves F-1 scores (measuring the robustness–compactness trade-off) by 13%–68% over non-LLM baselines and 9%–15% over expensive LLM-based approaches. Finally, we show how SHED-inferred structures are valuable for agentic document analysis: agents using SHED outperform baselines, achieving 3%–23% higher accuracy while being up to 10x cheaper.
[IR-32] Inspire: Benchmarking Scientific Literature Search for Open Research Problems
链接: https://arxiv.org/abs/2609.33233
作者: Jianrong Ding,Zhengyan Shi,Jianyuan Zhong,Kai Qiu,Qi Dai,Yifan Yang,Chong Luo,Qiang Xu
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注:
Abstract:Scientific literature search often begins with an open research problem rather than a known target paper or a fixed candidate set. We introduce INSPIRE, a benchmark for evaluating agents that search prior literature to make progress on solution-redacted research problems. Each instance pairs a research brief with a target-specific cutoff three months before a later paper and evaluates ranked outputs against graded cited antecedents from that paper’s realized research lineage. Search proceeds over an open corpus, while the identity of the target paper and membership of its cited antecedents remain hidden from the agent. Beyond end-to-end retrieval quality, INSPIRE uses logged search trajectories to distinguish three coupled stages: resource exposure, whether useful antecedents are surfaced during search; selection, whether exposed antecedents are retained; and ranking, how effectively retained papers are ordered. Across 476 computer-science targets under a shared search interface and budget, the strongest evaluated agent achieves 0.284 nDCG@10. Results show that current agents more readily recover an isolated antecedent than assemble a broader portfolio of relevant prior work. The stagewise analysis identifies resource exposure as the largest observed bottleneck, with further losses in selection and ranking. We additionally construct replay-valid hindsight demonstrations and show that they improve held-out search without changing test-time information, establishing that the benchmark provides an actionable learning signal. INSPIRE therefore enables both end-to-end comparison and stage-resolved diagnosis in a setting where the agent must construct its own working criterion of relevance.
[IR-33] Algorithmic Harms Associated with Generative Model-Augmented Recommendation Systems KDD2025
链接: https://arxiv.org/abs/2609.33073
作者: Christine Herlihy,Xumei Xi,Shloka Desai,Kevin Bannerman Hutchful,Pedro Silva
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: Presented at the KDD 2025 Workshop on Online and Adaptive Recommender Systems (OARS), August 3, 2025, Toronto, Ontario, Canada
Abstract:In this work, we consider algorithmic harms that may arise as generative models are incorporated into machine learning platforms. We argue that existing harm taxonomies and threat models require extension to (1) address novel causal drivers of well-studied representational and quality-of-service harms; and (2) anticipate and mitigate endogenous harms, such as sanitization, which may arise when system inputs are misaligned with the system designer’s objectives, or the generative model’s inductive priors. To this end, we introduce an expanded taxonomy of algorithmic harms associated with the use of generative models in non-conversational recommendation systems. In addition, we offer a causal analysis of how problematic subsets of the (input, output) joint distribution can arise, in an effort to inform harms detection and mitigation efforts.
[IR-34] Overview and Analysis of the RecSys Challenge 2026: Conversational Music Recommendation
链接: https://arxiv.org/abs/2609.33045
作者: Seungheon Doh,Sergio Oramas,Bruno Sguerra,Abhinav Bohra,Claudio Pomo,Francesco Barile
类目: Information Retrieval (cs.IR); Multimedia (cs.MM); Sound (cs.SD)
备注:
Abstract:The RecSys Challenge 2026 studies conversational music recommendation as a joint item recommendation and response generation problem: given a multi-turn dialogue, systems must retrieve relevant tracks from a large catalog and produce a grounded natural-language response. This paper presents the challenge task, dataset, evaluation protocol, and official results. Beyond the leaderboard, we analyze the 16 accepted systems through a common retrieve–rerank–generate framework and examine how recommendation performance varies across users, requests, and dialogue contexts. Strong systems commonly combine heterogeneous candidate sources and preserve source-specific evidence for learned reranking. Across the system papers and our organizer-side analysis, robust design also means 1) grounding cold-start retrieval in multi-turn conversation and item signals, 2) using intent detectors, and 3) modeling the full multi-turn context rather than the current query alone. We further identify limitations of the benchmark and evaluation protocol, including single-ground-truth relevance and teacher-forced evaluation of synthetic dialogues. Together, these findings provide practical guidance for future conversational recommender systems and shared evaluation efforts.
[IR-35] PILAR: A Page-Grounded Unified Evidence Representation via an Entity-Linked Assertion Graph for Open-Domain QA Agents over Multimodal Document Corpora EMNLP2026
链接: https://arxiv.org/abs/2609.32895
作者: Joongmin Shin,Gyuho Shim,Jung-hun Lee,Jaehyung Seo
类目: Information Retrieval (cs.IR)
备注: Accepted to Findings of EMNLP 2026. 25 pages, 5 figures, 24 tables
Abstract:Open-domain question answering (ODQA) over multimodal document corpora requires linking evidence scattered across text, tables, and figures. Existing systems often store these sources separately or retrieve only coarse pages, which weakens global evidence linking. We present PILAR, a page-grounded unified evidence representation instantiated as an entity-linked assertion graph. PILAR maps sentence-, table-, and figure-derived facts into a common assertion space and uses the graph as a controlled linking layer over robust page retrieval. In a shared-reader evaluation with four agent frameworks, fourteen retrieval backends, and two benchmarks, PILAR achieves the best end-to-end EM/ANLS. Gains are largest on compositional, cross-document, and multimodal questions, with a single-shot improvement of +1.6 EM over flat retrieval, rising to +2.9 on compositional and +5.9 on 3-hop questions. Ablations show that current gains are driven mainly by the text-instantiated slice of the framework, while visual assertions help only after locality-aware filtering. We therefore position PILAR as a unified evidence representation for multimodal ODQA rather than a standalone visual-reasoning module.
[IR-36] Mend the Measurement Gap: Latent User Preference Modeling for Short-Form Video Recommendation RECSYS’26
链接: https://arxiv.org/abs/2609.32839
作者: Shuo Chang,Yueqi Wang,Zihuan Diao,Ali Montazer,Jiangguo Zhang,Joyneel Misra,Dapeng Hong,Tomer Margolin,Sourabh Bansod,Ningren Han
类目: Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: RecSys '26: 20th ACM Conference on Recommender Systems
Abstract:Recommender systems rely heavily on heterogeneous behavioral feedback to infer user preference. Although abundant, these signals are imperfect measurements: the same observed behavior can arise from different underlying states, such as genuine enjoyment, passive consumption, or inattention. The challenge is especially acute in short-form video, where watch-based signals are strongly affected by measurement confounders such as video duration - the same watch time can imply different levels of preference for videos of different lengths, while ratio-based metrics can systematically favor short videos. As a result, optimizing raw engagement can amplify measurement artifacts rather than improving user value. We propose a Factorized Latent Value Model (FLVM) for measuring user preference from heterogeneous behavioral feedback. The model treats observed behaviors as noisy measurements of a low-dimensional, factorized latent value state and uses structured output heads to model heterogeneous feedback signals. A restricted baseline path captures predictable variation from measurement-confounding features such as video duration, user propensity, and session context, while a routed latent path estimates preference-relevant value advantage. The resulting latent value score can be integrated into an existing recommender system as a ranking feature or ranking score. On YouTube Shorts, a major short-form video platform, this model improves offline metrics and lifts a primary viewer enjoyment metric by 2.67% in online A/B tests.
[IR-37] RandSlot: Learning Compact Visual Document Representations with Random Soft Tokens
链接: https://arxiv.org/abs/2609.32699
作者: Dewen Guo,Shi Yu,Lingxiao Zhang,Yang Zhang,Tao XU,Dan Wang
类目: Information Retrieval (cs.IR)
备注:
Abstract:Visual document retrieval requires expressive representations to match queries with evidence distributed across text, tables, and page layouts. Multi-vector representations capture fine-grained information, but storing and comparing many vectors introduces substantial retrieval costs. In this paper, we introduce RandSlot, a simple approach to learn compact visual document representations with random soft tokens. During training, we append independently-sampled random unit vectors to query and document input sequences and resample them at every use, without introducing learnable soft-token parameters. The encoder contextualizes these auxiliary inputs with the original content to produce a small set of retrieval vectors. A standard late-interaction objective trains the encoder to extract relevant information under varying input conditions. Experiments with different backbone models show that RandSlot improves retrieval quality over alternative readout strategies under the same vector budget. Further analysis shows that these gains can persist when random soft tokens are replaced with zeros at inference, demonstrating that random inputs during training can improve compact retrieval representations even when inference no longer requires sampling.
[IR-38] Concepts Complement Dense Semantics: Learning Compact Sparse Spaces for Text-Image Retrieval CIKM2026
链接: https://arxiv.org/abs/2609.32671
作者: Yoonseo Kim,Jungwoo Choi,Cheonyoung Park,Youngwook Kim,Yongho Song,SeongKu Kang
类目: Information Retrieval (cs.IR); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted for oral presentation at KEIR@CIKM 2026
Abstract:Cross-modal retrieval has been advanced by vision-language pre-trained models that encode images and texts into a shared dense embedding space. While dense representations effectively capture overall semantic similarity, they often obscure fine-grained visual-textual information needed for precise cross-modal matching. Recent methods introduce a learned sparse branch to complement dense matching with lexical evidence, but they rely on a redundant language-model token space and lack explicit grounding for sparse dimensions. We propose GRASP, a compact and grounded sparse learning framework that mines visual-textual concepts from the corpus. A lightweight sparse head is trained to predict concepts relevant to each image or text, yielding interpretable concept-level evidence that complements dense semantic matching. Extensive experiments show that GRASP improves retrieval accuracy over the state-of-the-art dense-sparse baselines while yielding a more compact and grounded sparse space.
[IR-39] Retrieved but Not Delivered: Multimodal Memory Delivery for Long-Term Agents
链接: https://arxiv.org/abs/2609.32590
作者: Yuhang Jiang,Qingwei Liao,Kaize Yin,Xingling Liu,Luca Cuomo,Silvio Bacci
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 32 pages, 6 figures, 21 tables. Project page: this https URL
Abstract:Work on memory for multimodal agents optimizes what is written, updated and retrieved. Between retrieval and the answer, however, is a stage that multimodal memory evaluations do not isolate: what of the retrieved memory reaches the model, and in what form. We call it delivery, and a controlled decomposition on MemLens locates the remaining room there. With the retrieved evidence set exactly fixed, delivering the original pixels instead of withholding them raises accuracy by 13.87 points on an 8B backbone, whereas making retrieval perfect on those same messages improves it by 2.31. Delivery is the larger term on all three MemLens backbones and grows with backbone strength; retrieval grows too, without closing the gap. We propose DeliverMem, an instantiation of delivery as three decisions: keep the original modality, give each item a readable identity, and state when it was seen, with a retrieval-side adapter for the one property delivery cannot supply. Each is measured against a delivery-matched control that alters only its own variable. DeliverMem leads the strongest published memory agent on MemLens at all four context lengths, and beats DMV-Bench’s own strongest method at every setting on both backbones. On MemLens it does this on a tenth to a seventieth of the input. Each decision helps only where the question lacks what it supplies, and is null elsewhere. A single fixed configuration nonetheless leads both benchmarks, without training any component or modifying the stored records. Project page: this https URL
[IR-40] Beyond Dyadic Memory: Interaction-Aware Multimodal Memory with Adaptive Agent ic Retrieval for Multi-Party Spoken Conversations
链接: https://arxiv.org/abs/2609.32522
作者: Wenxu Jia,Xize Cheng,Zihan Zhang,Dongjie Fu,Linjun Li,Wenshi Chen,Yangyang Wu,Tao Jin
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注:
Abstract:Long-term memory enables agents to accumulate information and reason across sessions, yet existing research primarily focuses on dyadic text or image-text conversations, leaving long-term memory for multi-party spoken conversations underexplored. This setting requires preserving conversational content, identifying participants across sessions, and retaining who speaks to whom. To this end, we propose VoxPolyMem, an interaction-aware multimodal memory framework combining incremental speaker identification with a memory hierarchy comprising interaction memory, fact memory, and participant profiles. We formulate retrieval as sequential decision-making, where an agent rewrites queries and selects retrieval tools and memory layers based on accumulated evidence to address information gaps. We further introduce Evidence-Gain GRPO (EG-GRPO), which uses round-wise credit assignment to encourage complementary evidence acquisition. We also construct VoxPolyBench to evaluate memory evolution, personalized answering, memory retrieval and reasoning, and interaction reasoning and attribution in multi-party spoken conversations. VoxPolyMem achieves an overall score of 85.0 on VoxPolyBench, surpassing the strongest evaluated baseline by 23.6 points. On Mem-Gallery and H2HMem-Multi, it scores 89.6 and 74.4, respectively, exceeding the strongest evaluated public memory baselines by over 8 points each. These results highlight its potential for persistent, personalized assistance in multi-party multimodal interactions. Code and datasets are available at this https URL
[IR-41] When Does Dense Retrieval Need Asymmetric Geometry? A Bias-Variance Theory of Shared and Dual Projections
链接: https://arxiv.org/abs/2609.32488
作者: Maojun Sun,Yancheng Yuan,Jian Huang,Ruijian Han
类目: Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: 21 pages, 9 figures, 2 tables. Yancheng Yuan, Jian Huang, and Ruijian Han are joint corresponding authors
Abstract:Dense retrieval powers retrieval-augmented generation, semantic search, and question answering, yet the theoretical basis for choosing between shared and dual query-document projections remains unclear. We introduce a bias-variance theory for low-rank bilinear scoring. Shared projections induce positive-semidefinite operators, whereas dual projections realize arbitrary low-rank operators. We derive their exact approximation gap and prove a local Gaussian boundary: dual has lower risk exactly when squared directional signal exceeds the estimation cost of its additional degrees of freedom. This boundary motivates the Cross-fitted Asymmetry Risk Selector (CARS), which estimates reproducible directional signal from training pairs; its Gaussian counterpart admits exact selection-power and regret formulas. Guided by the theory, we run retrieval experiments across multiple datasets and embedding models. The mean Dual-minus-Shared NDCG@10 advantage more than doubles as query rotation increases from 0 degrees to 90 degrees. In the rank-sample-size grids, Shared wins 13 of 16 cells at n=32, whereas Dual wins all 32 cells at n=1024 and n=2048. Consistent with this shift, all 168 comparable operator-risk curves move toward Dual as training data grow. Compared to the two fixed-geometry baselines, CARS reduces held-out regret by 49-96% and achieves 90.1% mean geometry-selection accuracy.
[IR-42] Graph Memory: Spectral Associative Memory via Dirichlet Energy
链接: https://arxiv.org/abs/2609.32365
作者: Zhaoyang Shi
类目: Machine Learning (cs.LG); Information Retrieval (cs.IR)
备注:
Abstract:Dense associative memories have traditionally focused on storing and retrieving vector-valued patterns. Many modern machine learning problems, however, are naturally graph-structured, requiring memory mechanisms for relational patterns, graph diffusion geometries, community structures, and graph-based inductive biases. We propose a spectral dense associative memory for storage and retrieval of graph data, extending the classical vector-valued memories. Retrieval is performed through a log-sum-exp energy induced by Dirichlet energy with spectral norm distances, producing a softmax-weighted average of the stored Laplacians that remains a valid graph Laplacian. We prove exponential storage capacity and exponentially decaying retrieval error. Beyond graph retrieval, we establish theoretical guarantees for spectral quantities central to graph learning, including eigenvalues, eigenspaces, and diffusion operators. Experiments on synthetic graph data, real-world airline network, protein conformation data and wearable sensor data demonstrate robust graph retrieval while preserving the graph geometry of the data. Our framework provides a new associative memory paradigm for graph-structured data and bridges dense associative memory with modern graph learning and generative AI.
[IR-43] DP-Rec: Towards Dynamic Patching for Efficient Long-Sequence Recommendation RECSYS’26
链接: https://arxiv.org/abs/2609.32215
作者: Dwipam Katariya,Thomas Caputo,Akshat Shreemali,Juan Manuel Origgi,Nikita Seleznev,Pranab Mohanty,Kalanand Mishra,Nam Nguyen,James Montgomery
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: Accepted at RecSys’26, 14 pages, 7 figures
Abstract:Transformers have redefined sequential recommendation by effectively modeling dynamic user behaviors and long-range dependencies. However, they remain inherently inefficient: standard architectures operate at a fixed rate, allocating comparable computation to every item in a user’s history regardless of its information content. This leads to prohibitive computational overhead on long sequences and increased sensitivity to behavioral noise. To address this, practitioners often resort to lossy sequence compression, staged modeling, or truncation. This limits the model’s ability to leverage the full context of long histories during inference. Inspired by the recent success of Byte Latent Transformer, we propose DP-Rec, a dynamic latent patching architecture for recommendation. DP-Rec shifts from item-level modeling to patch-level modeling by segmenting interaction sequences using contrastive entropy surprise to identify informative behavioral boundaries. A lightweight patch encoder compresses these temporally contextualized segments into a reduced set of dynamic latent behavior vectors, which are then processed by a larger latent transformer and decoded for next-item prediction. Extensive experiments show that, under constrained computational budgets, DP-Rec scales effectively to long sequences and achieves a superior efficiency-accuracy trade-off over both non-compressed and fixed-size compression baselines.
[IR-44] Rethinking Cross-Channel Importance in Time-Series Forecasting
链接: https://arxiv.org/abs/2609.32187
作者: Yong-Hoon Choi,Kwang-Hyun Park,Youngjin Cho
类目: Machine Learning (cs.LG); Information Retrieval (cs.IR)
备注:
Abstract:Cross-channel modeling is central to multivariate time-series forecasting, yet channels that are statistically related, predictively useful, and actually used by a trained forecaster are often treated as if they defined the same notion of importance. We show that they need not coincide. Cross-channel dependency structures change substantially across future offsets, and horizon-adaptive source selection improves a controlled Ridge predictor in 21 of 32 dataset–prediction-length conditions, with a mean gain of 5.16% . This selected-set signal also transfers to a matched nonlinear predictor. Yet imposing the same horizon-specific source logic on iTransformer yields only 11 of 20 wins and a mean gain of 0.208% , with little alignment between controlled and neural gains. Functional interventions further show that strong forecasters use cross-channel information, while their source-reliance rankings agree little with controlled utility or with one another across iTransformer, TimesNet, and a cross-channel TimeMixer. As a constructive consequence, bounded post-hoc support improves a frozen channel-independent forecaster in 12 of 16 dataset–horizon conditions, with a positive aggregate bootstrap interval. Cross-channel importance should therefore be interpreted relative to the forecasting mechanism and question that define it: related \neq useful \neq used.
[IR-45] Amnesia by Design Memory By Necessity: Persistent State for Document Intelligence
链接: https://arxiv.org/abs/2609.32041
作者: Souhail Bakkali,Ayoub Merimi
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注:
Abstract:Modern Document AI reads contracts, extracts fields, reasons over tables, and grounds answers to page regions, then forgets everything. Processing an amendment the next day begins from scratch: no schema retained, no contradiction detected, no experience carried forward. This is a structural choice, not a scale failure: current systems are stateless functions. We call this the statelessness bottleneck. This bottleneck lies beyond parameter scaling, context extension, and retrieval augmentation: storage provides persistence and retrieval provides access, but neither consolidates observations into knowledge that improves future processing. This survey formalizes persistent evidence-grounded document state as a unifying framework, specifying the operations and invariants required to convert multimodal evidence into durable, provenance-linked state. We introduce a statefulness audit showing that ten representative benchmarks, coded against eight statefulness criteria, leave cross-session state evolution untested, and derive a longitudinal benchmark harness with five counterfactual metrics: Experience Gain, Cost Efficiency, Memory Harm, Forgetting Fidelity, Coverage Retention, to characterize the benefit, cost, risk, and governability of persistent document state. Document AI lacks mechanisms coupling persistent state to document-native structure, provenance, and temporal validity. The next era of Document AI will be defined by what systems retain across documents, sessions, and time.
[IR-46] Recipe-Matching Not Equivalence
链接: https://arxiv.org/abs/2609.31927
作者: Ali Habibullah,Mohammad Alshiekh,Yazan Alshoibi,Salman Khan,Naeemullah Khan
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:MathNet-Retrieve asks a retriever to find, for a math problem, a document stating the same problem. An LLM under one fixed prompt writes each gold document and its near-miss distractors; LLM judges filter them. We call this procedure the “recipe”, training on pairs built the same way “recipe-matching”, and ask how much score it buys beyond the ability the benchmark claims to test. Two models from one base, matched in rows and settings, differ only in the training file: pairs written under the benchmark’s published prompt by another vendor’s LLM and judge, or computer-algebra-verified pairs with no LLM anywhere. The first leads by 45 R@1 points on the easy tier. By a non-LLM paraphrase control, half to two thirds of that gap comes from the pairs being LLM-written at all: LLM rewrites under two unrelated prompts, with the verified model’s negatives, recover 30 and 22 of the 45 points; back-translations with the same negatives recover almost none. The remaining 15 to 25 points appear only under the benchmark’s own prompt and vanish on real duplicates no generator wrote, the same problem in two languages. The hard tier rewards the recipe’s pair structure, a deep rewrite against a minimal-edit near-miss: LLM rewrites alone score zero on it, attaching negatives unlocks it, and every negative that does so costs cross-language points; the sets scoring highest on it separate near-misses no LLM wrote worse than LLM rewrites with verified negatives. MELD also moves when a model trains on pairs built its way, without losing retention; on SABER-Math the registered attack fails, and the one gain, from its LLM-written summaries, is small but holds at a matched budget. Only on MathNet-Retrieve could we pin an inversion, benchmark score up and real retention down, to one edit of a training file. We release the generator-free duplicate evaluations, the near-miss test and three trained models.
[IR-47] FARE: Deep Reinforcement Learning For Fair Exposure Constrained Uncertainty Aware Financial Content Personalization ICLR2026
链接: https://arxiv.org/abs/2609.31890
作者: Arundeep Chinta,Lucas Vinh Tran,Jay Katukuri
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: Extended version of a paper accepted to the Advances in Financial AI: Towards Agentic and Responsible Systems Workshop at ICLR 2026
Abstract:Content personalization systems in financial services must ensure fair exposure across diverse offerings-a requirement driven by contractual obligations and the need to prevent “rich-get-richer” dynamics where content with high click-through rate (CTR) dominates while other relevant products receive minimal visibility. Share of Voice (SOV) constraints, which guarantee each content category a target fraction of top-position exposure, address this by promoting product diversity and balanced user discovery. While re-ranking layers atop CTR models are common in practice, we propose two key novelties: (1) framing SOV-constrained ranking as a deep reinforcement learning problem analogous to constrained trade execution in algorithmic finance, and (2) explicitly incorporating CTR prediction uncertainty into the agent’s state space and policy design-enabling larger ranking adjustments for high-uncertainty predictions where deviation from CTR-optimal ordering is less costly. We introduce FARE (Fair Ranking Executor), a modular uncertainty-aware execution layer that translates any black-box CTR model’s predictions into SOV-fair rankings without retraining the underlying model. Our uncertainty-weighted proportional control policy (FARE-PC) and learned neural policies (FARE-ES, FARE-PPO) demonstrate that uncertainty-aware approaches can substantially reduce SOV deviation from fairness targets while minimizing engagement loss, with gradient-free evolution strategies outperforming policy gradient methods on synthetic data and the ordering reversing on KuaiRand-Pure.
[IR-48] MM-VeriRec: Failure-Guided Fusion for Verifiable Agent ic Multimodal Recommendation
链接: https://arxiv.org/abs/2609.31718
作者: Yufeng Wang
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the 1st International Workshop on Agentic Multimodal Intelligence: Models, Benchmarks, and Applications (AMI '26), co-located with ACM Multimedia 2026
Abstract:Images often carry recommendation constraints that text metadata only hints at. A movie may need to look dark, a product may need a minimal style, and a visually impossible request should be rejected. Agentic multimodal recommenders must reason over text-image evidence, decide when visual evidence is decisive, and abstain when no valid action exists. We introduce MM-VeriRec, a verifiable multimodal recommendation protocol and failure-guided fusion method for hidden visual constraints, image-text mismatch, and impossible-task abstention. MM-VeriRec builds tasks from real movie-poster and product-image datasets, verifies each recommendation with deterministic visual attributes, and converts failures into actionable labels: text-trap following, visual ignorance, and false acceptance. Fusion should not merely concatenate modalities, but should diagnose which modality failed and route to the appropriate repair. Across MM-ML 1M and Amazon Reviews datasets, stronger text and vision embeddings improve retrieval but do not remove these failure modes, whereas failure-guided fusion does. The adaptive attribute gate reads the same tags the verifier checks and its scores are verifier-aligned upper bounds testing whether the taxonomy routes to the correct repair. More informative is transfer under a non-aligned gate: an independently derived leave-one-out CLIP detector still reaches 0.7028 and 0.6111 visual-grounded success, above both a VBPR baseline and plain fusion. The text-versus-visual gap reproduces across two LLM families, and the repair that helps differs by domain. MM-VeriRec is both a benchmark and a practical diagnostic loop for trustworthy agentic multimodal recommendation.
[IR-49] Data Processing for Offline Evaluation in Recommender Systems: a Survey
链接: https://arxiv.org/abs/2609.31696
作者: Alberto Carlo Maria Mancino,Angela Di Fazio,Danilo Danese,Matteo Attimonelli,Daniele Malitesta,Antonio Ferrara,Claudio Pomo,Tommaso Di Noia
类目: Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注:
Abstract:Offline evaluation is the dominant experimental paradigm in recommender systems research, enabling reproducible and cost-effective comparisons on historical interaction data. Yet, while considerable attention has been devoted to recommendation models and evaluation methodologies, the data processing decisions that precede model training have received less scrutiny. These decisions determine the information available to recommendation algorithms and can affect the comparability and reproducibility of experimental results. This survey provides a systematic, cross-domain characterisation of data processing practices for the offline evaluation of recommender systems. We examine the data-centric pipeline, from dataset selection and interaction representation to data preparation, multimodal feature extraction, and train-validation-test splitting. Our analysis spans recommendation paradigms, including collaborative, sequential, session-based, graph-based, knowledge-aware, context-aware, multimodal, federated, cross-domain, contrastive-learning, and LLM-based recommendation. Beyond reviewing existing practices, we introduce a unified framework and taxonomy for describing data transformations and feature-extraction strategies, distinguishing data preparation from the extraction of representations from multimodal side information. Our empirical analysis reveals a landscape dominated by a narrow set of dataset-level transformations, particularly support-driven filtering, while representation-dependent transformations remain less common. We further identify substantial heterogeneity in how auxiliary information is prepared and represented, as well as inconsistencies in the specification of data splitting protocols, where similar labels may conceal different experimental conditions.
[IR-50] Parser Chunking and Embedding Interactions in Retrieval-Augmented Generation over Indian Government Regulatory Documents
链接: https://arxiv.org/abs/2609.31660
作者: Shubham Kumar Singh
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 17 pages, 15 figures, 5 tables. Retrieval-only factorial evaluation with supplementary ablations and reproducibility materials
Abstract:Retrieval-augmented generation (RAG) pipelines are typically assembled from independently-chosen components – a document parser, a chunking strategy, and an embedding model – yet these choices are rarely evaluated jointly, and evaluations that do combine them are usually run on a single document or corpus. We present a controlled factorial study of 3 parsers, 3 chunking strategies, and 5 dense embedding models, together with a sparse BM25 baseline, evaluated against 800 question instances, each with one or more required evidence strings, with evidence strings automatically validated against source text and a 10% random sample manually reviewed, across four structurally distinct Indian central-government regulatory documents. We fit linear mixed-effects models with document-query-level random intercepts to the resulting 72,000-row result set, run Holm-corrected paired comparisons between matched dense and sparse configurations, and report clustered bootstrap confidence intervals for all 54 unique retriever configurations. We find that no single retriever family dominates across documents; parser and chunker choice interact significantly; MPNet-base is a consistent underperformer with a severe failure mode on table-derived questions; and the corpus exhibits a near-saturated evidence-preservation ceiling above 98%, indicating that retrieval differences are driven primarily by ranking quality rather than information loss during ingestion. We additionally report embedding-dimension and chunk-size/overlap ablations and an efficiency/quality Pareto analysis. We release our full evaluation harness, corpus manifest, and 800-question benchmark.
[IR-51] From Phase Transition to Systemic Failure: A Decoupled Analytics Framework for GNN Robustness
链接: https://arxiv.org/abs/2609.31656
作者: Shuai Yan,Dan Peng,Jie Li,Ke Wang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: Accepted by ICCCBDA 2026
Abstract:Data quality is a major bottleneck for the reliable deployment of graph neural networks (GNNs) in real-world graph mining tasks. Among various sources of degradation, label noise and feature distribution shift (hereafter referred to as distribution shift) are two common yet fundamentally different challenges. To study their effects under controlled conditions, this paper constructs a synthetic homophilic graph regression benchmark in which the two factors can be manipulated separately. A total of 41 configurations and 410 runs are conducted to evaluate the behavior of representative GNN models under varying noise and shift conditions. The results show two distinct patterns. First, under additive label corruption, performance remains relatively stable over a broad range of noise settings and begins to deteriorate sharply only after an observed transition region around the 50 percent noise ratio. Second, under extreme feature distribution shift, all tested models suffer substantial degradation, with test MSE increasing by 48 times to 316 times and correlation dropping by 73 percent to 89 percent. These findings suggest that, in the present controlled setting, GNNs are considerably more tolerant to moderate label perturbation than to severe distribution mismatch. The study provides a controlled empirical baseline for understanding how data quality affects GNN-based graph mining systems and offers practical implications for deployment-oriented monitoring and model maintenance.
[IR-52] Goal inference with Rao-Blackwellized Particle Filters
链接: https://arxiv.org/abs/2512.09269
作者: Yixuan Wang,Dan P. Guralnik,Warren E. Dixon
类目: Machine Learning (cs.LG); Information Retrieval (cs.IR); Systems and Control (eess.SY)
备注: 6 pages, 3 figures. Accepted for presentation at the 23rd IFAC World Congress 2026, Busan, Republic of Korea, August 23-28, 2026. To appear in IFAC-PapersOnLine
Abstract:Inferring the eventual goal of a mobile agent from noisy observations of its trajectory is a fundamental estimation problem. We initiate the study of such intent inference using a variant of a Rao-Blackwellized Particle Filter (RBPF), subject to the assumption that the agent’s intent manifests through closed-loop behavior with a state-of-the-art provable practical stability property. Leveraging the assumed closed-form agent dynamics, the RBPF analytically marginalizes the linear-Gaussian substructure and updates particle weights only, improving sample efficiency over a standard particle filter. Two difference estimators are introduced: a Gaussian mixture model using the RBPF weights and a reduced version confining the mixture to the effective sample. We quantify how well the adversary can recover the agent’s intent using information-theoretic leakage metrics and provide computable lower bounds on the Kullback-Leibler (KL) divergence between the true intent distribution and RBPF estimates via Gaussian-mixture KL bounds. We also provide a bound on the difference in performance between the two estimators, highlighting the fact that the reduced estimator performs almost as well as the complete one. Experiments illustrate fast and accurate intent recovery for compliant agents, motivating future work on designing intent-obfuscating controllers.
人机交互
[HC-0] Reliability-Gated Fusion of Consumer Head and Foot IMUs for Lower-Body 3D Pose
链接: https://arxiv.org/abs/2609.35764
作者: Zhilin Guo,Boqiao Zhang,Oszkár Urbán,Josef Bengtson,Hakan Aktas,Wenzhao Li,Siyu Hong,Kyle Fogarty,Chenliang Zhou,Ali Senguel,Cengiz Oztireli
类目: Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
备注: 10 pages, 4 figures, 3 tables. Code: this https URL
Abstract:Sparse inertial pose estimation promises camera-free motion capture from consumer devices, but consumer sensors are unreliable: firmware-fused orientations are biased, mounting varies between sessions, and streams drift or drop out. On a new 35-take single-subject benchmark pairing an earbud head inertial measurement unit (IMU) with two smart-insole foot IMUs (SAM-3D-Body pseudo-ground-truth labels), we show the reliability problem is channel-level: a channel ablation isolates foot acceleration as the most informative input (66.6 mm vs. 79.0 mm head-only) and the firmware-fused foot orientation as the liability that destroys the gain. We therefore let the model learn how much to trust each channel of each stream: one temporal gate per stream per channel block, trained with an auxiliary reliability objective on synthetically corrupted pretraining data. The channel-gated model is the most accurate of our learned fusion arms on clean data (69.4 mm vs. 83.7 static, 86.6 ungated) and under every simulated fault (bias in training; drift, dropout eval-only); its gates suppress the natively biased foot-orientation channels on clean real data without test-time supervision and flag dropout bursts at 0.92-0.999 AUROC. Two contrasts: dropping a channel known a priori to fail is flat across foot faults but collapses when an unanticipated stream fails (head dropout: 92.9 vs. 79.3 mm); and a fine-tuned HMD-Poser is more accurate on clean data (64.4 mm) and nominally under drift, with no significant paired difference under bias or dropout, but a larger worst-case degradation from clean (+16.1 vs. +3.5 mm, single seed). Learning to gate reliability instead of sensor count is the lever for deployable sparse inertial capture. Code is available at this https URL.
[HC-1] Reclaiming the social in social media
链接: https://arxiv.org/abs/2609.35413
作者: Luca Benn,Moritz Merz,David Grüning
类目: ocial and Information Networks (cs.SI); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注: 7 pages, 1 figure
Abstract:Concerns about polarization, antisocial behavior, and mental health have broadly led to two responses to social media: adapting platforms through moderation and prosocial design, and restricting access to these platforms through age limits and comparable measures. Both address consequences of social media while leaving the attention-driven architecture intact. We argue that the harms commonly attributed to social media arise not from technologies supporting social connection but from their implementation within the attention economy. In response, we ask what a digital social environment built for genuine human connection would look like. From four principles - Purpose, Alignment, Transparency, and Access - we can derive four technical properties that fulfill these principles: Operator Blindness, Algorithmic Sovereignty, Operational Parity, and Verifiability. To the best of our knowledge, no deployed alternative satisfies all four. We propose an architecture that does, making the attention economy not just discouraged but structurally impossible. We explain how the resulting user experience refocuses on each user’s individual social environment and differs from that of current platforms.
[HC-2] Alignment Games: A Framework for Conceptual Repair in Human-AI Collaboration
链接: https://arxiv.org/abs/2609.35197
作者: Hari Subramonyam,Maneesh Agrawala,Sean Follmer
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注:
Abstract:The meaning of a concept in use is shaped by the situation, task, goals, and prior knowledge. For example, a request to make a poster “visually appealing for a five-year-old” might evoke bright colors and cartoon imagery for one collaborator, but less text, bold shapes, and visual simplicity for another. We call such task-relevant differences conceptual misalignment. We introduce Alignment Games, a framework for making these differences visible and repairable during human-AI interaction. Drawing on theories of situated conceptualization, we characterize task-specific conceptual frames in terms of relevant attributes, values, relations, constraints, and priorities. We then define alignment moves that intervene on the situation, the reasoning used to interpret it, or the resulting frame. Through examples from educational content generation, creative coding, and argumentative writing, we show how these moves can be composed into repair sequences and derive design principles for supporting task-sufficient conceptual alignment at runtime.
[HC-3] he Future of Visualization Dashboards in the Age of Generative AI
链接: https://arxiv.org/abs/2609.35170
作者: Vaishali Dhanoa,Duosi Dai,Gabriela Molina Léon,Eduard Gröller,Niklas Elmqvist
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Generative AI promises easier dashboard creation, raising questions about the future of dashboards and the people who create and use them. We interviewed 16 experts based in 14 countries about their practices and expectations. Almost all expected dashboards to persist for recurring questions, monitoring, and reporting. They anticipated adaptive views and combinations of language, graphical controls, and gestures, while emphasizing interaction as part of human exploration and understanding. Participants expected authors’ responsibilities to shift toward specifying requirements, curating generated work, and evaluating outputs, with design knowledge and communication remaining important. Easier creation also raised concerns about validation effort, users’ understanding, maintenance, and personalization weakening shared understanding. We discuss seven opportunities for research and practice concerning validation, end-user education, dashboard proliferation and rot, organizational guidance, adaptation, novel visualizations, and accountability for AI-generated content. Our findings connect dashboard evolution with the human and organizational work needed to sustain their use.
[HC-4] oward a Culturally Adapted Chinese Language Agent : A Wizard-of-Oz Study of Nonverbal Behavior in Chinese-German Intercultural Interaction
链接: https://arxiv.org/abs/2609.35150
作者: Siddhant Jain,Anna Lea Reinwarth,Dimitra Tsovaltzi,Rafael Math,Julia Renner
类目: Human-Computer Interaction (cs.HC)
备注: Accepted to ICMI Companion '26. 7 pages, 4 figure
Abstract:Successful intercultural communication requires more than grammatical competence. It demands sensitivity to culturally embedded social norms whose violation triggers subtle but meaningful nonverbal responses. For German learners of Mandarin Chinese, acquiring this sensitivity is critical yet poorly supported by existing language-learning agents. We present a Wizard-of-Oz (WoZ) study design and supporting real-time system for collecting multimodal behavioral data from native Chinese speakers reacting to social norm violations by German learners. The system features a photorealistic MetaHuman avatar driven by Live Link face capture and MediaPipe upper-body tracking, a wizard console for real-time behavior selection, and synchronized multimodal logging across agent and learner streams. A layered annotation framework, based on psychological theory and covering non-observable socioemotional reactions, norm interpretation, verbal, and observable behavior thereof, and future supervision targets enables the corpus to support training of future automated cultural interpretation and behavior generation models. Four ecologically valid interaction scenarios, developed with cultural and pedagogical experts, provide the methodological and technical foundation for a culturally adapted conversational agent for Chinese language learning.
[HC-5] Making the Invisible Visible: A Framework for Reflective AI Use in Software Engineering Education
链接: https://arxiv.org/abs/2609.34997
作者: Ali Shakiba,Thomas Chaffey
类目: oftware Engineering (cs.SE); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注: 9 pages, 2 figures, Accepted for publication in the Proceedings of the 37th Annual Conference of the Australasian Association for Engineering Education (AAEE 2026)
Abstract:Generative AI (GenAI) is increasingly embedded in software engineering education, supporting activities such as requirements development, design exploration, documentation, and prototyping. However, educators often have visibility only into final artefacts, with limited insight into how students evaluate, verify, and refine AI-generated outputs during the learning process. This creates challenges for assessing evaluative judgement and responsible AI-assisted practice. This paper introduces the AI Journal, a structured reflection framework designed to make student-GenAI interaction visible in first-year software engineering education. The framework combines execution tracking, which records prompts, outputs, intent, and interaction context, with cognitive auditing, which captures verification strategies, intervention decisions, confidence judgements, critical learning moments, and reflections on AI-supported work. Deployed in a first-semester software engineering course, the AI Journal enabled visibility into aspects of student learning not observable through artefact-based assessments alone. Preliminary observations suggested variation in verification practices, intervention strategies, and perceptions of AI-supported work. Critical learning moments frequently occurred when students evaluated contextual suitability, feasibility, and requirements alignment rather than identifying obvious errors. The AI Journal demonstrates a practical, lightweight, and model-agnostic approach for making AI-assisted learning processes visible. By foregrounding verification, intervention, and reflection, it shifts attention from product-focused assessment toward evaluative judgement and responsible AI-assisted practice.
[HC-6] One Sensor Whole Body - 3D Body Pose from a Single Consumer Earbud IMU
链接: https://arxiv.org/abs/2609.34978
作者: Zhilin Guo,Boqiao Zhang,Oszkár Urbán,Josef Bengtson,Hakan Aktas,Wenzhao Li,Siyu Hong,Kyle Fogarty,Chenliang Zhou,Ali Senguel,Cengiz Oztireli
类目: Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
备注: 5 pages, 2 figures, 2 tables. Accepted at the 6th International Workshop on Human-centric Multimedia Analysis (HUMA '26), ACM Multimedia 2026, Rio de Janeiro, Brazil. Code: this https URL
Abstract:Consumer earbuds already stream inertial motion data from the head, one of the most widely worn sensor locations on the body. We ask how much of the 3D body pose a single such head IMU can recover, and whether adding more consumer sensors actually helps. We build a multimodal capture pipeline that records four-view RGB-D video together with an AirPods head IMU and two Striv insole IMUs, synchronize the streams post-hoc, and generate pseudo-ground-truth with SAM 3D Body, yielding a 35-take single-subject benchmark spanning gait, turning, vertical, everyday, and clinically inspired motions. Adapting two recurrent model families (IMUPoser and MobilePoser), we show that one head IMU recovers lower-body pose at 79.0 mm rigid-MPJPE and per-foot ground contact at 0.809 macro-F1, and that a causal variant retains most of this accuracy at streaming latency. In paired per-take significance tests across both families, adding the consumer foot IMUs never significantly improves pose and significantly degrades it in two of four model-split combinations; a mounting-bias probe and feet-only ablation identify insole orientation quality, not foot placement, as the mechanism. Extending the output to a 20-joint full-body skeleton maps the boundary: gross distal-arm motion is partially recoverable from the head alone, proximal upper-body pose is not, and staged fine-tuning recovers the leg accuracy that naive joint training sacrifices to multi-task dilution. For learned pose from consumer wearables, sensor reliability, not sensor count, is the binding constraint here. For the devices tested, the earbud is its sweet spot. Code is available at this https URL.
[HC-7] From Early Participation to Later Completion: Evidence from a Large-Scale Self-Paced Learning Programme
链接: https://arxiv.org/abs/2609.34953
作者: Sakshi Sharma,Pavani Ayinampudi,Aditya B.M.V.,Jinal Gupta,Prakash Hegade,Rohit Sharma,Meenakshi V,S.R.S. Iyengar
类目: Human-Computer Interaction (cs.HC)
备注: 12 Pages, 3 figures, 3 tables
Abstract:Large-scale learning programmes generate records that make learner participation observable across different activities. Participation points are commonly used to record and encourage such participation, but their value may extend beyond the activities for which points are awarded. Existing evaluations often examine gamification outcomes within the activities or learning environments in which the game elements are implemented, providing limited evidence about whether early participation points contain information about later participation outside the points system. This study examines whether early participation points can provide information about learners’ later participation in a self-paced learning track that does not award participation points. Using anonymised records from 876 learners in a large-scale remote software-upskilling internship, we examined participation points generated from live-session attendance and poll responses against later self-paced course completion. The primary analysis used the 438 learners who earned at least one point during the first week, while the full cohort was retained for the no-point analysis. Week-one participation points distinguished learners who later completed a self-paced course with an AUC of 0.89, increasing to 0.95 by the fourth week. Similar AUCs were observed at both stages of the self-paced course sequence, while the absence of week-one points identified learners who did not start or did not complete a self-paced course with 95% precision. These findings indicate that early participation points can provide information about later participation outside the activities that generate the points. Such information can help large-scale learning programmes identify learners who may require timely attention while learning is still in progress, without treating participation points as a measure of overall learner engagement.
[HC-8] “Black Mirror?”: Public Sensemaking of AI-Powered Lifelogging
链接: https://arxiv.org/abs/2609.34950
作者: Ying Ma,Jarod Govers,Le Fang,Shuning Zhang,Yongquan ‘Owen’ Hu,Xin Yi,Jorge Goncalves
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:AI-powered lifelogging wearables are emerging as a new class of consumer devices that transform everyday experience into searchable, AI-curated memory archives. We study early public sensemaking around these systems at the moment of their market entry, using the Looki L1 as an empirical lens. Analysing large-scale Chinese-language and English-language social media discourse (N = 5,053 comments), we combine topic clustering with inductive thematic analysis to examine how users interpret the social, moral, and political implications of AI-mediated memory. Across contexts, users reference dystopian surveillance imaginaries, express privacy resignation and bystander concerns, and debate assistive value alongside consumer logics. English-language comments more often framed these devices through interpersonal power, evidentiary use, and hacking anxieties, while Chinese-language comments more often foregrounded labour exploitation, governance surveillance, and technological inevitability.
[HC-9] AI Tools Adoption across the Double Diamond Workflow: Phase Mode and Barriers in Designer Practice
链接: https://arxiv.org/abs/2609.34655
作者: Sepideh Tajarmakan,Khashayar Hojjati Emami
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Designers are adopting AI faster than the tools built for them can keep up. This survey of 443 designers across 43 countries, among the first phase-disaggregated accounts of its kind, examined reported AI use across the four phases of the Double Diamond workflow (Discover, Define, Develop, Deliver). 79.7 percent reported confirmed AI use in at least one phase, but engagement was typically partial, spanning a mean of 2.89 of 4 phases, with adopters retaining the earliest phases and dropping the latest. Tool choice tracked each phase’s dominant activity: conversational tools drove Discover and Define, AI-native image generation took over Develop, and Deliver showed a hybrid profile. AI-native and embedded AI use peaked in different phases, pointing to two distinct modes of human-AI collaboration: generating from scratch versus refining within existing software. Use intensity declined through fewer designers engaging, not scaled-back use. Adopters and non-adopters differed on one dimension: perceived usefulness.
[HC-10] Show your work: An exploratory study of student experiences of Turnitin Clarity
链接: https://arxiv.org/abs/2609.34618
作者: J. Roe(1),M. Perkins(2) ((1) Durham University, (2) British University Vietnam)
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Process-capture platforms promise to make the writing behind an assessed artefact visible, and vendors increasingly market them as a fairness measure rather than a detection tool. Their acceptability to the students who must write within them, however, is largely untested. This exploratory qualitative study examines how students’ expectations of one such platform, Turnitin Clarity, compared with their experience of using it. Seven students at a UK university, given no onboarding to the platform, took part in pre-use focus groups. Six then completed an unassessed 500-word writing task over five days, before all seven returned for a post-use focus group. Data was analysed using reflexive thematic analysis, with Expectation-Confirmation Theory as an orienting frame. The overall pattern was one of mixed confirmation. Participants’ functional expectations were largely met, and the sense of being watched that they had anticipated persisted after use: for some it was offset by perceived fairness, while for others it heightened self-monitoring. The bounded AI assistant divided them: its limits were welcomed for marking out acceptable use, but some felt it weakened their ownership of the work and worried it would flatten what they produced. Across both phases, participants weighed the costs of observation against perceived gains in fairness and asked for transparency to run in both directions. Several described already writing defensively in anticipation of accusations of AI misuse. Acceptance of process technologies was reported as being conditional on practice time, two-way transparency, and explicit data governance. Participants also questioned whether a single linear document can represent a writing process they described as messy and multimodal. Across these accounts, process capture did not sit outside the writing it recorded but reorganised the practice it set out to observe. Subjects: Human-Computer Interaction (cs.HC) Cite as: arXiv:2609.34618 [cs.HC] (or arXiv:2609.34618v1 [cs.HC] for this version) https://doi.org/10.48550/arXiv.2609.34618 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[HC-11] Bayesian Active Learning for Intent Disambiguation in Interactive Robot Planning IROS2026
链接: https://arxiv.org/abs/2609.34270
作者: Huao Li,Carson Sobolewski,Augustinos Saravanos,William Tan,John Karigiannis,Chuchu Fan
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注: Accepted by CoRL 2026, also presented at IROS 2026 Human-Robot Dialogue workshop
Abstract:Interactive robot planning requires robots to infer and execute human intentions from natural language instructions that are often ambiguous, incomplete, or underspecified. Although large language models (LLMs) provide a powerful interface for clarification, relying on the generative model to drive an multi-turn conversation can introduce systematic failures. We propose a Bayesian framework that treats clarification as an active learning problem over grounded Signal Temporal Logic (STL) task specifications. Our method uses LLMs to initialize candidate formal specifications and translate informative contrasts into natural-language clarification questions, while Bayesian optimization maintains uncertainty estimation over user intent and selects queries that maximize information gain. After convergence, the inferred STL specification is passed to a formal planner to synthesize a verifiable robot trajectory. Across four simulated and real-world task domains, our approach generally achieves higher task satisfaction and requires fewer clarification rounds than LLM baselines, while helping smaller models close the performance gap against larger reasoning models.
[HC-12] When Models Choose the Question: Pedagogical Constraints in Bottom-Up Multi-Agent Inquiry
链接: https://arxiv.org/abs/2609.34243
作者: Yeri Hong,Lauren Hyoseo Yoon
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:What shapes a model-generated inquiry when no discussion question is supplied? We introduce a bottom-up forum framework inspired by Philosophy for Children, in which language-model agents read a philosophical narrative, propose and select questions, and develop a shared conclusion without a privileged model facilitator or aggregator. Across 576 forums, contrasting Aristotelian value personas interacted with a blank-slate participant receiving no value-specific instruction. The blank slate remained neutral and was selected more often for conclusions than questions. Yet inquiry narrowed in both form and source: varied initial questions increasingly became either/or alternatives, while discussion concentrated on directions already explicit in the text. Our ECO framework traces this source focus by distinguishing explicit philosophical framing, characters’ modeled inquiry, and open-ended narrative material. Across analyzed chapters, participants drew most often on explicit framing. We describe this as a pedagogical constraint: freedom to formulate questions did not necessarily produce freedom from source framing.
[HC-13] PainterBench: A Figural Divergent-Thinking Benchmark for Tool-Using Language Models
链接: https://arxiv.org/abs/2609.34195
作者: Shane K.A. Dalumura Hettige,Jonas Oppenlaender
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注: 25 pages, 7 figures, 11 tables
Abstract:Figural divergent thinking is the ability to develop a given shape fragment into an original drawing. In humans, this ability is assessed with incomplete-drawing tasks. We introduce PainterBench, a benchmark that ports the incomplete-drawing task to the agentic setting. The agent draws on a canvas through tool calls and observes the result after every turn. The canvas includes a starting shape which cannot be erased, and the agent’s goal is to incorporate this shape into the most original drawing it can produce. The task is open-ended, and the agent itself decides when the drawing is finished. The benchmark tests incremental visual planning over a short horizon and the transfer of creative ability from pretraining to multi-turn tool use. We evaluate 14 multimodal language models from small to frontier scale. Across the primary study and six sensitivity analyses, we collect 2,700 drawings and crowdsource creativity and recognizability ratings for every drawing and for 300 human reference drawings. We also present ViDrA-adapted, an automated scorer that predicts human creativity ratings of agent drawings (r = 0.85 on random held-out test split). Figural divergent thinking varies widely across the 14 models, and GPT-6 Astra produces the most creative drawings. Relative to the human drawings, the agent drawings score higher in creativity but lower in recognizability. We release the final drawings, per-round canvas snapshots, tool call traces, stimulus bank, benchmark harness, crowdsourced ratings (N = 72,000), and ViDrA checkpoint.
[HC-14] Exploring the Affordances of Generative Image AI for Supporting Early-stage Architect-client Communication
链接: https://arxiv.org/abs/2609.34118
作者: Chengzhi Zhang,Weijie Wang
类目: Human-Computer Interaction (cs.HC)
备注: 18 pages
Abstract:Text-to-image generative AI can produce renderings from natural-language prompts in near real time, making it increasingly popular for rapidly visualizing concepts in early-stage architectural design. Meanwhile, exchanging ideas efficiently and building shared understanding have long been central challenges in architect-client communication. How might the speed of generative image AI change this communication? To explore this question, we conducted a study with 11 architect-client pairs, in which each pair used generative image AI over video conference to collaboratively produce early-stage renderings of the client’s “dream house.” Our findings suggest that generative image AI helped pairs develop a solid shared understanding by providing concrete visual materials and supporting the exchange of ideas. It also shifted conversation dynamics, enabling clients to participate more actively in shaping design direction. However, challenges emerged, including a stylistic bias toward particular types of images and unpredictable shifts in design direction caused by variation across generations. We conclude with implications for the design of future generative image AI-based systems that support architect-client communication.
[HC-15] HiThink Turn: An Intent-Aware Turn-Taking Control Module for Full-Duplex Dialogue
链接: https://arxiv.org/abs/2609.34096
作者: Feiyang Chen,Wenhan Yang,Bohan Wang,Xinjian Gao,Rongjunchen Zhang,Jun Wang,Xinhui Hu
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Full-duplex dialogue requires timely yet selective interruption handling, which end-of-turn prediction alone cannot achieve: complete utterances may need no response, while unfinished requests may warrant interruption. To address this challenge, we propose HiThink Turn, an intent-aware streaming turn-state predictor that separates response intent from semantic completeness and conditions decisions on system playback state. A key contribution is minimal intent-sufficient prefix supervision, constructed through LLM judgments and speech alignment, while training on audio truncated at chunk boundaries improves robustness to partial speech. These components support streaming inference with 240-ms audio chunks, enabling low-latency, accurate full-duplex turn control. Experiments show that HiThink Turn leads the compared methods in Easy Turn macro accuracy, Full-Duplex-Bench average interaction rate score (0.933), and non-target-speech average playback resume rate (0.735). Additionally, intent-prefix triggering raises interruption success from 89% to 98% and reduces mean stop latency by 60.9%.
[HC-16] Scanvas: Discovering and Developing Synergistic Opportunities in Generative Design Spaces
链接: https://arxiv.org/abs/2609.34062
作者: Yaqing Yang,Aniket Kittur,Hongyu Howie Wang,Nikolas Martelaro,Matt Klenk,Yan-Ying Chen,Matthew K. Hong
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Good design is often synergistic, creating super-additive value by linking goals so that existing resources produce greater outcomes. However, finding these synergistic opportunities in sparse design spaces is difficult, and current LLM-supported ideation tools largely default to additive paradigms such as feature blending, variant generation, or local patching. We present Scanvas, an AI-supported system for systematically discovering and developing synergistic design opportunities. Scanvas operationalizes synergy through a two-step computational process: first, it decomposes seed ideas into explicit properties (components, behaviors, surpluses, and issues) to enrich the design space; second, it systematically searches across enriched ideas using three theory-grounded strategy operators: unlocking or strengthening goals, turning weaknesses into resources, and sharing components across functions. We instantiate Scanvas as an auto-generation pipeline and an interactive system. Pipeline ablations and a user study with 12 professional designers demonstrate that Scanvas enables users to surface and develop significantly higher-quality, synergistic concepts compared to LLM ideation baselines.
[HC-17] StructSim: Measuring Idea Similarity at Scale Through Structural Representation
链接: https://arxiv.org/abs/2609.34046
作者: Yaqing yang,Vikram Mohanty,Mei-Xi Chia,Nikolas Martelaro,Dafna Shahaf,Aniket Kittur
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Measuring idea similarity is fundamental to creativity evaluation, especially as LLMs enable idea generation at increasing scale. However, text embeddings collapse an idea into a single vector, making it difficult to capture structural similarity, including partial overlap across core and supporting components and differences across levels of abstraction. We introduce a shared structural representation that decomposes ideas into purpose, mechanism, and implementation components and organizes related components in a multi-layer concept graph. From this representation, we define measures of pairwise similarity and set-level mechanism coverage for assessing idea diversity. We evaluate our approach using controlled idea triples and assessments from 12 experts, focusing on differences in core mechanisms, implementations, and supporting components. Our method improves alignment with expert judgments of structural similarity by 31% over the embedding baseline and better reflects expert assessments of idea set coverage, supporting scalable evaluation of idea similarity and diversity.
[HC-18] hinkNet: Compact Architecture Selection and Validation-Gated Ensembles for Subject-Independent MI-EEG Decoding
链接: https://arxiv.org/abs/2609.33967
作者: Abdul Basit,Saim Rehman,Muhammad Shafique
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Signal Processing (eess.SP)
备注: Accepted to IEEE-EMBS BHI’2026, 7 pages
Abstract:Practical assistive and rehabilitative brain–computer interfaces require subject-independent motor-imagery EEG (MI-EEG) decoders that generalize to new users under limited target-user data and constrained compute. However, held-out-subject performance can be overstated when test-subject information influences preprocessing, model selection, or ensemble selection. We present \textitThinkNet, a validation-controlled framework that combines train-only normalization, validation-guided evolutionary search, and validation-gated inference to identify compact decoders and inference policies for held-out subjects. We evaluate four-class BCI Competition IV-2a (session T) decoding with nine Leave-One-Subject-Out (LOSO) folds, three seeds, seven fixed decoder entries, and a broader search over ten representative decoder families; the held-out subject is never used for normalization, hyperparameter, architecture, or ensemble-policy selection. In the fixed benchmark, the validation-selected compact decoder achieved 44.35 \pm 15.41% accuracy with 4.9K parameters, 19 KB FP32 weights, and 0.99 ms batch-1 Orin CUDA inference. Across the broader search, compact models ( \leq 25K parameters) achieved higher mean held-out accuracy than mid-size and large alternatives after selected retraining (40.10% vs. 35.09% and 34.78%). Validation-gated ensembling improved over validation-selected single-model inference, reaching 43.98 \pm 16.25% in the fixed benchmark and 43.31 \pm 15.88% for the compact six-family ensemble. A non-deployable oracle analysis revealed a 6.1-point family-selection gap and near-zero validation–test correlation, showing that validation reliability remains a key bottleneck under subject shift. Thus, ThinkNet is a validation-controlled framework for compact MI-EEG model and inference-policy selection, rather than a single-architecture benchmark.
[HC-19] EEG-Fusion: Failure-Informed Source-Free Expert Routing for Robust Motor Imagery EEG Decoding
链接: https://arxiv.org/abs/2609.33962
作者: Abdul Basit,Saim Rehman,Muhammad Shafique
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Signal Processing (eess.SP)
备注: Accepted to IEEE-EMBS BHI’2026, 7 pages
Abstract:Subject-independent motor-imagery (MI) EEG decoding can exhibit subject-level failures even when average performance appears acceptable: under subject shift, a decoder can become an overconfident near-one-class predictor. This is especially problematic in source-free deployment, where target-user labels are unavailable during adaptation and expert selection. We present \textitEEG-Fusion, a failure-informed decision-level fusion framework that treats source-free MI decoding as label-free reliability estimation over heterogeneous experts. EEG-Fusion applies subject-wise Euclidean alignment and normalization-only test-time adaptation, then routes each target subject to a neural, covariance-based, or physiological-feature expert using a reliability gate trained on source-held-out folds to predict expert performance and collapse risk from label-free stream diagnostics. The gate uses confidence, entropy, prediction diversity, expert agreement, and predicted class balance; collapse is measured as the maximum predicted class fraction. In 9-fold leave-one-subject-out (LOSO) evaluation with three seeds, relative to a no-alignment raw EEGNet source-free anchor, EEG-Fusion improves subject macro-F1 from 0.417 to 0.529 on BCI IV-2a local protocol, from 0.314 to 0.482 on BNCI2014-001, and from 0.607 to 0.708 on BNCI2014-004; corresponding collapse-index reductions are 0.199, 0.227, and 0.169. In a 9-subject Cho2017 external subset, EEG-Fusion improves macro-F1 from 0.516 to 0.630. These results suggest that label-free reliability estimation can reduce subject-level failure modes in source-free MI-EEG deployment.
[HC-20] “Is This Book AI-Generated?” How Authorship Suspicion Manifests in Marketplace Reviews
链接: https://arxiv.org/abs/2609.33950
作者: Victor Dibia
类目: Human-Computer Interaction (cs.HC); Computers and Society (cs.CY)
备注: 16 pages, 5 figures, 9 tables
Abstract:As AI becomes part of how books are authored, reader response to suspected AI authorship grows more consequential, yet remains unexamined. We analyze 863 low-star reviews of 78 Amazon bestsellers across 8 categories at three levels of proximity to AI. Suspicion concentrates in Generative AI books (35.1%) but appears in every category, including Gardening (5.7%). Reviews citing AI authorship complain more about shallow content and poor presentation than other critical reviews. Suspicion takes two forms: ambient, where “AI-generated” is a generic complaint about formulaic writing, and corroborated, where reviewers of the same book independently cite concrete evidence. We propose two mechanisms by which suspicion arises: topical concentration, where a book’s AI subject matter supplies vocabulary for quality complaints, and artifact detection, where readers notice ChatGPT-style formatting regardless of topic. Star ratings can hide this suspicion: the most-flagged book holds 4.1 stars while 50% of its critical reviews cite AI authorship.
[HC-21] ReVision: Supporting Designers Interpretation and Exploration of Visuals in Concepts and Forms
链接: https://arxiv.org/abs/2609.33911
作者: Yaqing Yang,Mei-Xi Chia,Aniket Kittur
类目: Human-Computer Interaction (cs.HC)
备注: UIST’26 Poster
Abstract:Visual designers get inspiration from references to expand their design space. They decompose what makes a reference evocative into conceptual and visual elements, ranging from explicit attributes such as objects and colors, to less readily articulated concepts and visual motifs. They then create different visual forms to explore how the selected elements could be combined differently. Novices often struggle with these moves, instead focusing on surface features or producing limited visual variation, thus becoming fixated on the reference. Existing tools support editable visual attributes and high-level themes, but provide limited control over how conceptual interpretations relate to expressive visual motifs or how their combinations can be systematically re-expressed. We present ReVision, an AI-based tool that decomposes visual and textual references into editable conceptual interpretations and visual motifs, enables their recombination across conceptual and visual spaces, and renders each direction as divergent visual-form variations, supporting more divergent exploration during the creation process.
[HC-22] NSV-Shift: A Contrastive Benchmark for Non-Speech Vocalization Understanding and Response Adaptation in Speech-to-Speech Models
链接: https://arxiv.org/abs/2609.33899
作者: Ziwei Chen
类目: Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注: Technical Report
Abstract:We introduce NSV-Shift, a contrastive benchmark for evaluating whether speech-to-speech models can understand non-speech vocalizations (NSVs) and adapt their responses accordingly. Each pair contains two conversations with identical lexical content that differ only in the NSV embedded in the final turn. Our pilot contains 22 human-verified pairs (44 audio conditions) and evaluates five models on NSV perception, emotion understanding, and response adaptation. Results show that models generally perform better at detecting NSVs than at interpreting their fine-grained emotional meaning or producing appropriately differentiated responses. The data construction pipeline, dataset, and evaluation pipeline are publicly available at this https URL.
[HC-23] Accounting for Stochasticity in Studies of Large Language Model Refusal
链接: https://arxiv.org/abs/2609.33743
作者: Emma Lurie,Stephanie T. Wang,Sorelle A. Friedler,Danaé Metaxa
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:We present preliminary empirical evidence that single-observation queries are insufficient for evaluations of LLM refusal behaviors. Using a longitudinal auditing system, we issued identical prompts 100 times each across four dates to GPT-4.1 for two socially salient topics across 20 Wikipedia sources. Refusal outcomes were consistent with a stable Bernoulli process, yet 20% of sources fell within a decision-boundary region where a single query is largely uninformative. Reliable quantification of refusals required between 15 and 25 repeated queries, well above the single-observation standard common in existing evaluations.
[HC-24] Characterizing Memory Misalignment in Human-LLM Interaction From User Perspectives
链接: https://arxiv.org/abs/2609.33623
作者: Jingruo Chen,Shuning Zhang,Eryue Xu,Jianing Li,Xin Yi
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:While memory enhances personalization in LLM-based conversational agents, it suffers from memory misalignment, where memories violate user expectations. We present a mixed-methods investigation to characterize and mitigate memory misalignment from user perspectives. First, we collected data from memory usage (N=28, 457 entries) and diary study (N=32, 304 reports), which yielded a taxonomy spanning 14 misalignment types across memory intake, storage and management, retrieval and interpretation stages. Second, four co-design workshops with 12 experienced HCI researchers derived a design space to tackle memory misalignment issues, consisting of 12 candidate interaction strategies structured across interaction form, placement and intrusiveness dimensions. Finally, a speed dating with 121 users reveals preference heterogeneity, where users prioritize proactive controls over cognitively demanding causal graph inspections or passive audit logs. Synthesizing these findings, we highlight the tension between supervisory agency and interaction overhead, and advocate for friction-aware memories that balance user oversight with conversation smoothness.
[HC-25] E-CONAN (Entailment CONtradition And Neutral) Diagnostics Dataset Investigating Linguistic Phenomena in Arabic Natural Language Understanding
链接: https://arxiv.org/abs/2609.33530
作者: Khloud AL Jallad,Nada Ghneim,Ghaida Rebdawi
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG); Logic in Computer Science (cs.LO)
备注:
Abstract:Natural Language Understanding (NLU) plays a crucial role in various applications, yet its performance suffers from weaknesses in handling the complexities of human languages, ranging from lexical ambiguity to high-level reasoning difficulties. Analyzing errors across diverse linguistic phenomena is crucial for NLU improvement, as it will help humans get insights to comprehensively assess models’ limitations and capabilities, so optimizing models’ generalization. Notably, several benchmarks contain diagnostics datasets designed for investigation and fine-grained error analysis. When highlighting the gaps in the state-of-the-art, we noted that there is no naming convention for macro and micro categories or even a standard set of linguistic phenomena that should be covered. To overcome this gap, we propose an initial hierarchy for Cross-Lingual NLU error analysis. Moreover, we propose a methodology to create an NLI hierarchical framework and applied a case study on Arabic NLU. Moreover, this paper introduces E-CONAN diagnostics dataset, a freely available dataset manually-annotated with coarse-grained and fine-grained categories based on our proposed Arabic hierarchy. E-CONAN dataset helps NLU designers better understand their models by doing error analysis and in-depth investigation. We used E-CONAN to investigate the performance of 9 pretrained language models and 5 LLMs. Results indicate that LLMs outperform pretrained models in world knowledge and commonsense reasoning macro-category, and underperform pretrained models in syntactic macro-category. Moreover, the hardest phenomena for all models is Reasoning, and the easiest phenomena for all pretrained models is Syntactic, and the easiest for LLMs is Lexico-Syntactic.
[HC-26] AI-Driven Collaborative Assembly Line Inspection: System Integration and Deployment Challenges
链接: https://arxiv.org/abs/2609.33522
作者: Asya Ünal,Amr Okasha,Ege Çırakman,Perin Ünal
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
备注: 13 pages, 4 figures, 2 tables. Accepted at the 22nd International Conference on Mobile Web and Intelligent Information Systems (MobiWIS 2026)
Abstract:Manual visual inspection on assembly lines is a persistent manufacturing bottleneck: operator fatigue over extended shifts lowers defect-detection rates. This paper presents the design, integration, and field deployment of an AI-assisted collaborative inspection cell at the Silverline kitchen-appliance factory, developed within the AI-PRISM project. The cell couples a Universal Robots UR 10e cobot carrying a machine-vision defect-detection pipeline with a Comau Racer-5 cobot for functional tests, coordinated through ROS 2 Humble on an Ubuntu 22.04 LTS server. Multi-modal data (Basler camera imagery, TIA microphone acoustics, and SPS electrical-safety measurements) are logged locally and visualised in real time with Grafana. We report the practical deployment challenges (close-proximity safety, AI robustness under glare and reflections, ROS 2 namespace collisions across two cobots, and operating-system and dependency issues) together with the engineering solutions adopted, and structure the integration through a four-level Human-Robot Interaction analysis. The deployed cell cuts per-unit quality-check time from 82 s to 61 s (about 25%), raises final-control resource efficiency from 0.75 to 0.88, reduces operator visual-inspection viewing time by 82%, and significantly lowers operator mental demand (p = 0.005, NASA-TLX).
[HC-27] Separating Memory and Workflow Effects in Predicting Individual Answers
链接: https://arxiv.org/abs/2609.33321
作者: Tianzhu Qin,Leo Yang Yang,Lee Wei Jun,Kun Chen,Ramit Debnath,Davin Youchao Dong
类目: Human-Computer Interaction (cs.HC)
备注: 50 pages, 11 figures, 35 tables, including appendices
Abstract:Personalized language agents choose both what to remember about a person and how to use that memory. We separate these choices when predicting unseen answers to known interview questions. On 1,768 tasks from 188 people, a concrete memory built from a verified interview prefix outscores a trait description by 0.0158 (95% whole-person interval [0.0044, 0.0271]). Crossing both memories with one-shot generation and three-answer fusion, fusion lowers concrete-memory scores by 0.0123 ([-0.0189, -0.0056]); prompted and trained selectors do not detectably beat a random candidate. One call on the longer, unrewritten source record outscores every memory condition. Under a limited context budget, OwnWords retrieves the person’s sentences with BM25 and answers in one call. It outperforms the written memory on 500 people outside the benchmark (+0.0127, [+0.0037, +0.0217]; an earlier held-out test was inconclusive) and across four budgets on 300 people (mean +0.0218, [+0.0138, +0.0298]), with the latter result repeated on 114 people. It does not detectably outperform recency truncation. These results compare evidence-construction procedures; they do not isolate the effect of verbatim wording. On Twin-2K-500, OwnWords predicts ordinal survey answers more closely than the written memory, but does not improve exact-choice accuracy and lowers it in one of two samples. Interview scores use a model-based content rubric without human ratings, and the original benchmark’s participants were seen during development. These results characterize the tested procedures, not a general human-prediction ceiling.
[HC-28] Beyond Tasks: A Vision for Reproducing an Animal-like Behavioral Substrate Using Modern Robot Learning Techniques
链接: https://arxiv.org/abs/2609.33165
作者: Samiyuru Menik,Hemadri Jayalath
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注: Vision paper
Abstract:Recent advances in robot learning have produced increasingly capable embodied agents. Yet comparatively less attention has been given to a more basic form of competence that animals exhibit continuously: the ability to remain situated, responsive, and behaviorally coherent as physical, environmental, and social demands change over time. We propose the ethological behavioral substrate as a conceptual lens for studying this form of competence in artificial agents. Rather than treating these behaviors that animals exhibit as a set of isolated skills, we argue that their continual coordination under competing demands constitutes an important and underexplored target for modern robot learning. We further propose robotic animal companions as a useful research setting for studying sustained interaction and adaptation in human-centered environments. Such systems provide an opportunity to investigate how social behavior, memory, and continual learning develop over long periods of interaction. This perspective motivates further investigation of how such persistent behavioral competence may complement higher-level capabilities in embodied agents.
[HC-29] PDFa11yMut: Measuring Mutation-Specific Detection in PDF Accessibility Checkers
链接: https://arxiv.org/abs/2609.33160
作者: Gauri Jain
类目: Human-Computer Interaction (cs.HC); Software Engineering (cs.SE)
备注:
Abstract:Automated PDF accessibility checkers provide useful conformance evidence, but a clean report is not a complete accessibility oracle. PDFa11yMut measures mutation-specific checker behavior by applying paired structure-level transformations to reference-suite baselines, verifying intended deltas and non-target invariants, and recording hash-linked checker evidence. Across 30 conformance-oriented mutants, PAC and veraPDF each produced 30 and 30 direct target findings, respectively, while Acrobat produced 23 direct findings plus 3 prespecified consequence-proxy findings. The 39 semantic/assistive-representation mutants produced no automated target finding in the tested configurations, while Acrobat issued manual-review prompts for a subset; Class B is interpreted descriptively rather than as a universal checker obligation. A separate convenience-selected exploratory AT sample observed representation differences in 8 of 9 pairs under one fixed NVDA/Acrobat/Windows procedure. The artifact contributes reusable operators, structural and purity oracles, paired baseline/mutant evidence, and reproducible analysis for a scoped mutation-testing study rather than a general checker-accuracy benchmark.
[HC-30] Medical Knowledge Is Not All You Need: When Medical QA Becomes Situated Patient Assistance
链接: https://arxiv.org/abs/2609.33040
作者: Shreya Bali,Riku Arakawa,Jill Fain Lehman,Alexander K. Maytin,Brian Chen,Emma Russell,Haarika Reddy,Annalise Vaccarello,Dustin P. DeMeo,Bryan T. Carroll,Mayank Goel
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Reliability in medical QA is often pursued by grounding responses in authoritative medical information. We show that when QA is embedded within ongoing care, reliability depends on more than what the system knows medically. In a study with 73 skin cancer patients practicing postoperative wound care, 41.9% of response-requiring questions depended on information beyond the procedure, including visual or physical state, environmental context, or prior actions. These demands varied across patients, consistent with patients recruiting the assistant into different informational roles. We then replayed the questions to seven LLMs while adding procedural and postoperative guidance. Errors remained substantial, including treating unknown states as known, even under explicit guardrails; with full procedural context, six of seven models more often introduced later steps prematurely. Based on these findings, we propose a design space for situated medical assistance that connects what the assistant and patient can each reliably establish to the form of assistance provided.
[HC-31] Smelling the Way: Olfactory Modulation of Spatial Estimation and Path Integration in Virtual Reality
链接: https://arxiv.org/abs/2609.32992
作者: Siyeon Bak,Junho Kim,Dongyun Han,Sun-Jeong Kim,Isaac Cho
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Spatial cognition enables individuals to perceive and interpret spatial relationships, estimate locations, and navigate within their surroundings. While vision plays a dominant role, other sensory modalities, particularly olfaction, can support spatial processing when visual cues are limited. This paper explores the effects of scent-delivery within virtual reality (VR) environments by using fan-mediated olfactory devices. We conducted two formal studies to evaluate the effects on spatial information acquisition. Study 1 (N=24) examined the participants’ ability to localize a scent source while walking a linear path using two different olfactory devices. Study 2 (N=19) employed a triangle completion task, in which participants attempted to return to their starting point under varying rotation angles and olfactory conditions. Results indicate that scent-delivery feedback influenced spatial behavior in distinct ways across the two studies. In Study 1, visual-olfactory misalignment produced systematic directional biases without improving absolute localization accuracy. In Study 2, distance-modulated scent-delivery feedback reduced terminal homing error but also increased traveled distance and completion time, suggesting a more deliberate goal-confirmation strategy.
[HC-32] VocalEyes: Speaker-Aware Augmented Reality Captioning through In-Conversation Registration
链接: https://arxiv.org/abs/2609.32983
作者: Yuxiao Wang,Xulong Tang,Chen Chen,Rawan alghofaili
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Co-located augmented reality (AR) captions make speech readable, but they can separate an utterance from the person who produced it. In unfamiliar groups, losing that source complicates immediate responses and later review: users must recover not only what was said, but also who said it. Conventional diarization returns anonymous clusters, while speaker recognition typically assumes pre-meeting enrollment. We built VocalEyes, a speaker-aware AR captioning system that creates named voice profiles from natural self-introductions. The interface coordinates speaker-attributed captions, a fixed profile card, and a visual cue that marks the articulating face. In a controlled within-subjects study with 20 participants who reported typical hearing, VocalEyes identified speakers with 88.0% accuracy and increased participant speaker-tracking accuracy from 47.2% with caption-only AR to 87.3% with the complete interface. Participants also reported lower workload with the complete interface. These findings show how in-conversation registration can preserve speaker attribution across live captions and meeting records in scripted small-group meetings.
[HC-33] Exploring the Effects of Olfactory Cues and Ventilation on Teleportation-based Navigation in VR
链接: https://arxiv.org/abs/2609.32975
作者: Dongyun Han,Heecheol Kim,Siyeon Bak,Chang-Guen Song,Sun-Jeong Kim,Isaac Cho
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Spatial cognition supports how people interpret spatial relationships and navigate their surroundings. Although vision plays a dominant role, other sensory modalities, including olfaction, may also contribute under limited visual conditions. This paper investigates the role of olfactory cues in location recognition during virtual reality (VR) navigation. We conducted two formal studies using a wearable olfactory prototype. Study 1 examined whether olfactory cues and ventilation influenced users’ recognition of encountered locations during \HDYteleportation- and dash-based navigation. Study 2 extended this investigation to repeated navigation and examined whether ventilation duration influenced recognition performance and user experience across successive movements. Results show that olfactory cues can support recognition of encountered locations during navigation, while ventilation was associated with reduced residual interference and improved usability. However, longer ventilation did not clearly improve recognition accuracy in repeated navigation. These findings suggest the potential of olfactory cues as contextual signals during VR navigation, while also highlighting the importance of scent management for maintaining perceptual clarity and user comfort.
[HC-34] CLAIRE: A Schema-Grounded Hybrid Workflow for Healthcare Administrative Form Completion
链接: https://arxiv.org/abs/2609.32787
作者: Garapati Keerthana,Manik Gupta
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注:
Abstract:Healthcare administrative staff transfer structured information from electronic health records, referrals, claims systems, provider rosters, and work queues into dynamic forms. We developed and evaluated CLAIRE (Clinical Language and Agentic Intelligence for Reasoning and Entry), a hybrid workflow that separates field-state discovery, source-to-field mapping, deterministic validation, bounded correction, escalation, and audit tracing. We tested five synthetic healthcare administrative schemas, 1,000 source records, four interface variants, two data-quality suites, and six comparators, yielding 24,000 benchmark episodes. A separate strict-output audit evaluated direct mappings from Qwen2.5-1.5B and Qwen2.5-7B, and a trace-derived operational simulation covered 6,000 episodes. Under the evaluated synthetic benchmark conditions, full CLAIRE achieved 1.000 episode success, field accuracy, required-field completion, and dependency completion in both suites; removing validation reduced stress-suite success to 0.500. In the simulation, 100.0% of clean and validation-stress episodes reached a staff-reviewable draft, compared with 68.6% of escalation challenge episodes, unsupported cases were blocked. Scenario-based savings were 149.7-165.5 seconds per case, not observed staff times. The findings support schema-grounded, validation-first healthcare administrative automation in which language-model components assist mapping but do not authorize unsupported or consequential actions.
[HC-35] You Cant Spot a Deepfake?And Neither Can Your Brain Nor Eyes: A Neurophysiological Framework for Deepfake Exploitation of Cognitive Engagement and Implicit Visual Evaluation
链接: https://arxiv.org/abs/2609.32769
作者: Cagri Arisoy,Md Imanul Huq,Amy W. Hays,Nitesh Saxena
类目: Cryptography and Security (cs.CR); Human-Computer Interaction (cs.HC)
备注: Accepted at the 29th Information Security Conference (ISC 2026)
Abstract:Deepfakes have rapidly emerged as a pressing threat to information integrity and security because they exploit human trust in visual and auditory perception. Yet, little is known about whether humans and their underlying (sub)conscious neuro-physiological processes can reliably distinguish deepfake from real videos. We introduce DECEIVE (Deepfake Exploitation of Cognitive Engagement and Implicit Visual Evaluation), a framework that models how deepfake videos are validated as adversarial payloads through behavioral and neuro-physiological screening of viewers, and how attacks can be refined by selecting payloads that evade detection. The framework is dataset agnostic and applies to synthetic or real media. It is inherently dual-use: an adversary with equivalent measurements could iterate on candidate manipulations and retain those that evade human detection. This motivates open, defensive evaluation. Measuring which deepfakes defeat human perception establishes a realistic bound on attacker capability against which detection tooling, provenance and watermarking mechanisms, and user-facing protections can be assessed. As an instantiation, we conducted an EEG and eye-tracking study in which participants viewed real, deepfake, and look-alike videos drawn from Celeb-DF and a curated celebrity set, while behavioral judgments and implicit responses were recorded. Contrary to expectations of subconscious differentiation suggested by prior work on paintings and phishing websites, no statistically significant neuro-physiological differences emerged between real and deepfake videos, although clear distinctions were observed for look-alike videos. Behaviorally, participants accepted 26.68% of manipulated clips as authentic, rising to 31.94% for familiar identities, confirming the studied deepfakes as effective adversarial payloads within DECEIVE. Comments: Accepted at the 29th Information Security Conference (ISC 2026) Subjects: Cryptography and Security (cs.CR); Human-Computer Interaction (cs.HC) Cite as: arXiv:2609.32769 [cs.CR] (or arXiv:2609.32769v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2609.32769 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[HC-36] Chatbot Engagement Does Not Always Beget Metalearning: Evidence from Three Countries
链接: https://arxiv.org/abs/2609.32739
作者: Kokil Jaidka,Insyirah Binte Imam Mujtahid,Peng Qi,Harshit Aneja,Subhayan Mukerjee,Wynne Hsu,Mong Li Lee,Tsuhan Chen
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Chatbots deliver real-time fact-checks, but whether a chatbot correction leaves anything behind once the chatbot is gone - metalearning, distinct from correcting misbeliefs - is untested. We report a preregistered, three-country randomized experiment (USA, India, Singapore; N ~ 2,200) on out-of-context image misinformation, manipulating a correction’s channel affordances (synchronicity, bandwidth) across four conditions: Control, Links-only, Static explanation, and a Socratic Chatbot built on a validated out-of-context detector, with an unaided retest one week later. The Chatbot produced the largest immediate discernment gain (d = 0.097, p = .023). All three interventions reduced sharing of false claims (d ~ -0.12, p .01). One week later, no advantage persisted: the Chatbot arm declined relative to Control, most sharply in India and Singapore, and in India on claims it never discussed. Decay tracked affordance level and did not vary by country. Engagement mechanisms, we argue, do not substitute for slow AI literacy.
[HC-37] When the Environment Becomes the Interface: Multisensory Environmental Interfaces for Human-AI Interaction in Autonomous Vehicles
链接: https://arxiv.org/abs/2609.32604
作者: Keqi Chen,Runjia Tan,Xinyi Fu,Shanhe Lou,Kwan Min Lee,Chen Lv
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:As AI increasingly assumes operational control, human-computer interaction is shifting from operating systems through explicit interfaces to inhabiting intelligent environments. This raises a fundamental question: when users no longer directly manipulate a system, what mediates their relationship with intelligent technologies? We introduce environmental interfaces: designed environmental conditions that shape human-AI relationships through ambient, holistic, and evaluative pathways rather than explicit functional interaction. Using autonomous vehicle cabins as a revealing context, we conducted a within-subject experiment with 24 participants (216 observations), manipulating lighting and scent in a simulated autonomous driving environment. Three findings emerged. First, perceived atmosphere accounted for 66.5% of the variance in overall journey experience, showing that environmental conditions can function as an interface rather than merely a supporting design element. Second, multisensory processing followed a hierarchical architecture: individual sensory appraisals were initially independent, while cross-modal integration emerged during higher-order environmental evaluation. Third, olfactory stimuli influenced affective responses and experience evaluation more strongly than visual stimuli, challenging the visual dominance of automotive interaction design. Environmental quality and sensory congruency also predicted trust in the autonomous system, suggesting that passengers may use environmental cues as proxy signals when direct assessment of AI competence is difficult. These findings establish environmental interfaces as a distinct interaction modality and point to a broader transition from designing interfaces for operating intelligent systems to designing environments for inhabiting them.
[HC-38] Artificial intelligences and human scientists exhibit complementary strengths in theory building
链接: https://arxiv.org/abs/2609.32562
作者: Ke Li,Spyros I. Zoumpoulis,Phanish Puranam,Philip Parker,Matthew Eshbaugh-Soha,Izzy Gainsburg,Michael Gilead,Igor Grossmann,Britt Hadar,Yoel Inbar,Almog Simchon,Robb Willer,Rui Ai,Ruicheng Ao,Gavin J. Bala,Matthew Bidwell,Shuang Cai,Kai Chang,Skyler Y. Chen,Cory J. Clark,Irmak Dai,Abhinandan Dalal,Connor Douglas,Alexis Du,Zhehang Du,Leyun Feng,Isabel Fernandez-Mateo,Linnea Gandhi,Cyrille Grumbach,Anmol Gupta,Vansh Gupta,Maria Hademer,Jay H. Hardy III,Chen Kai Huang,Jacob Xiangyu Jin,Ufuk Keskin,Na Hyun Kim,Mert Kobaş,Byounghoon Koh,Gabrielle Lamont-Dobbin,Gregory Lanzalotto,Sun Young Lee,Dingzhe Leng,Chenjun Li,Weiyuan Li,Zeyuan Li,Zhongyuan Liang,Ning Liu,Peihong Liu,Yuhan Liu,Jiuyao Lu,Wanteng Ma,Nicolas Martinet,Natnael Mulat,Christina A. Nguyen,Khai Nguyen,Quang Minh Nguyen,Naja Pape,Chanwoo Park,Stefanos Poulidis,Jeffrey Sanchez-Burks,Michael Schaerer,Isabelle Solal,Yanbo Song,Junghyo Sun,Qingyao Sun,Rui Sun,Roderick Swaab,Kevin Tan,Dequn Teng,Michelle A. Vaccaro,Robin Vigerbaeck,Xiaomeng Wang,Randol H. Yao,Duygu Yilmaz,Shun Yiu,Ecem Yucesoy,Allen Zang,Ruijia Zhang,Xilan Zhang,Yichi Zhang,Zhanhao Zhang,Eric Luis Uhlmann
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注:
Abstract:We investigate the effectiveness of artificial intelligences (AI)-specifically large language models (LLMs)-relative to human scientists at high-level cognitive tasks in social science such as theory formulation, predictions of novel empirical results, and theory revision in response to new evidence. The research domain was academic discourse regarding gender and race inequality. Our findings, comparing 25 LLMs with 13 senior researchers and 60 doctoral scholars, reveal that the AIs outperformed most humans individually on most of the present tasks, while human theories were more diverse and exhibited greater gains in predictive accuracy from aggregation. AI-generated theories were more extensively elaborated, involving additional theoretical paths and latent variables, and were rated as higher quality than human theories by independent raters blinded to source. However, this theoretical complexity was in part ornamental, in that it was not associated with more accurate predictions about empirical patterns in data; in contrast, human scientists achieved greater predictive efficiency with simpler theories. The AIs were significantly more likely than human scientists to revise their theories to incorporate new evidence; human scientists updated their beliefs in a selective way that is sensitive to prior prediction errors. We speculate that the superior processing capacity of artificial intelligences makes them especially well-suited to tasks requiring grappling with complexity, but that the greater diversity of human ideas is essential to wise crowds and collective creativity.
[HC-39] DashAct: A Progressive Diagnostic Benchmark for GUI Agents in Interactive Dashboard Analysis
链接: https://arxiv.org/abs/2609.32385
作者: Chuhan Zhang,Qi Xie,Ziyue Wang,Jianing Yin,Yunfan Zhou,Dazhen Deng,Yingcai Wu
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注:
Abstract:Interactive dashboards require users to reveal and connect evidence across stateful interactions. Although graphical user interface (GUI) agents could automate this process, existing dashboard benchmarks primarily report final answers or task success. They provide limited insight into whether failures arise from maintaining the analytical process, selecting actions, or grounding visual targets. We introduce DashAct, to our knowledge the first benchmark to diagnose these failures at a fine-grained level within the same dashboard task. DashAct contains 357 human-verified interaction trajectories with milestone dependencies and hierarchical target annotations. Its progressive diagnostic cascade evaluates end-to-end execution, restores verified context for next-action prediction, and provides target semantics and a local view for visual grounding. By progressively restoring the conditions for success, DashAct measures the minimum support an agent needs to recover rather than scoring isolated skills. Experiments show that current models struggle even as support is added. The cascade outcomes reveal bottlenecks hidden by end-to-end scores and provide actionable guidance for improving GUI agents.
[HC-40] WSM-Aware HRI: An IoT-Enhanced Framework for Early Detection and Norm-Guided Repair of Failures with LLM Guidance
链接: https://arxiv.org/abs/2609.32336
作者: Hanlin Zhang,Yuquan Wang,Tianwei Zhang,Zhenglong Sun
类目: Robotics (cs.RO); Human-Computer Interaction (cs.HC)
备注:
Abstract:Human-robot interaction (HRI) failures remain a major barrier to deploying robots in real-world environments. Prior work often treats failures as isolated technical faults or focuses on post-hoc recovery behaviors. In practice, many breakdowns arise because humans and robots operate under inconsistent assumptions about the current world state. We propose WSM-Aware HRI, an IoT-enhanced modular framework that unifies diverse HRI breakdowns as World-State Mismatches (WSMs) between a human’s instruction-implied assumptions and a robot’s grounded world model built from multimodal perception and digital augmentation. A Large Language Model (LLM) is used to make implicit assumptions explicit, map them to a small set of mismatch types, and specify the evidence needed for verification against the robot’s world state. WSM-Aware HRI shifts failure handling from execution-time recovery to proactive mismatch detection during intention formation, enabling interventions guided by safety, norm compliance, and multi-user coordination with transparent explanations. We evaluate mismatch identification in ten everyday cases spanning both visual and latent-state mismatches. The system can accurately produce the expected output results, and ablations show that reliable identification depends on appropriate grounding representations and verification-oriented refinement. These results indicate that treating interaction breakdowns as explicit world-state mismatches enables earlier detection of impending failures and offers a principled mechanism for integrating external evidence and social constraints into human-robot interaction.
[HC-41] Handwritten Digit Leakage from Smartphone Motion Sensors Across Unseen Users and Phone Models
链接: https://arxiv.org/abs/2609.32117
作者: Constantino Álvarez Casado,Erkka Rantahalvari,Matteo Pedone,Matti Matilainen,Manuel Lage Cañellas,Le Nguyen,Simo Hosio,Olli Silvén,Miguel Bordallo López
类目: Machine Learning (cs.LG); Human-Computer Interaction (cs.HC)
备注: 7 pages, 4 figures, 3 tables, 7 numbered equations, and 33 references. Code available at this https URL
Abstract:Smartphone motion sensors support interactive applications, but their readings may also reveal touchscreen input beyond their intended use. Assuming known drawing intervals, we study whether handwritten digits remain predictable across users and devices, as a 10-class problem on 19,628 HuMIdb recordings from 481 participants. We compare handcrafted features with classical machine learning algorithms, MiniRocket kernels, and a compact sensor patch transformer on accelerometer, linear acceleration, gyroscope, and gravity signals. The transformer achieves 57.74% accuracy and 82.64% top-3 accuracy on 75 unseen participants, and 58.77 \pm 0.95% over 3 seeds for unseen participants on 9 unseen phone models. Low motion recordings remain informative, accuracy is not monotonic in motion level, and the tested contrastive pretraining, augmentation, and derived signals give no consistent gains. Digits are thus predictable beyond familiar users and phone models under assumed segmentation, while acquisition-order shortcuts limit conclusions about practical privacy exposure. Code available at: this https URL.
[HC-42] CG-Diff: Organizing Code Changes Around Call Graphs
链接: https://arxiv.org/abs/2609.32057
作者: Bimal Raj Gyawali,Devamardeep Hayatpur,Nadia Polikarpova
类目: oftware Engineering (cs.SE); Human-Computer Interaction (cs.HC)
备注:
Abstract:Tools for code review present code changes file-by-file. We argue that, oftentimes, changes can be better organized around a call graph of the changed code. From an empirical study of GitHub pull requests (PRs) in-the-wild, we find that (1) in around 40% of the PRs, more than half of the changed functions (and methods) are connected by call graphs, and (2) necessary callees are often located in different files from their callers. Based on these findings, we develop the notion of CG-Diffs, subgraphs derived by decomposing a call graph of the changed code into smaller and more navigable directed graphs. We then implement a web-based interface for viewing PRs that restructures the PR around these CG-Diffs. Through a within-subject study comparing our interface against GitHub’s PR view, we find that CG-Diffs help orient participants and provide them with more meaningful structures to navigate the PR. We also found several limitations: function nodes repeated across multiple CG-Diffs can be disorienting, and changes not contained in a function (e.g. globals, imports) are not as immediately apparent. Our study shows promise in using call graphs to help contextualize and navigate unfamiliar codebases, which may benefit new contributors to open source, and reviewers of unfamiliar LLM-generated PRs.
[HC-43] A Benchmark for LLM s Understanding of Middle School and High School Science Topics
链接: https://arxiv.org/abs/2609.32020
作者: Noah L. Schroeder,Yessy Eka Ambarwati,Yuji Zhang,ChengXiang Zhai
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注:
Abstract:Large language models (LLMs) are increasingly integrated into educational settings, yet educators lack robust, standards-aligned tools to evaluate their effectiveness in K-12 science contexts. Existing benchmarks predominantly assess general language or advanced scientific reasoning, leaving a critical gap in understanding LLMs’ performance on content directly relevant to secondary science curricula. To address this gap, we developed a comprehensive NGSS-aligned benchmark for both middle and high school science using a rigorous synthetic data pipeline, multi-judge validation, and item-level psychometric analysis. Nine open-weight LLMs were systematically evaluated using this benchmark, indicating that several smaller, locally deployable models achieved high accuracy across diverse science domains and question types. Our findings indicate that model size did not consistently predict performance, emphasizing the importance of intentional model selection for educational deployment. We then incorporated a human reviewer into the loop, reviewing the items generated by the LLMs for alignment with NGSS standards. The human review indicated that synthetically generated items were not in perfect alignment with the NGSS standards, indicating the benefits of human-in-the-loop item development, the need to explore the intersection of content and pedagogical knowledge, and the need to extend benchmarks to evaluate LLMs’ capacity for interactive, evidence-based feedback in educational scenarios.
[HC-44] Vibe Analysis: Exploring LLM Adoption by Data Visualization Practitioners
链接: https://arxiv.org/abs/2609.31922
作者: Shani C Spivak,Aditi Krishna,Mahsan Nourani,Melanie Tory
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) are enticing in their promise to support data visualization (Vis) through faster and simpler workflows for data prep, analysis, and visualization creation. Yet LLMs are notoriously error-prone and not built for data visualization tasks. Few studies have explored LLM adoption among Vis practitioners. To fill this gap, we conducted semi-structured interviews with members of the Data Visualization Society, a global community of data visualization designers. Our findings show that Vis designers actively use LLMs for both creative and technical aspects of the visualization process. A new visualization workflow is emerging, a process we call vibe analysis, analogous to vibe coding. Some key challenges raised by participants parallel those of vibe coding, while others are Vis-specific, like gaps in Vis knowledge and chart verification. This work opens up opportunities for research combining LLM-mediated work with Vis tools that incorporate data visualization guidance, constraints, and best practices.
[HC-45] Anatomy of a Spreadsheet Failure Analysing the EuSpRIG Horror Story Corpus
链接: https://arxiv.org/abs/2609.31815
作者: Simon Thorne,Angela Collins
类目: Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注:
Abstract:Spreadsheet disasters have been documented for more than thirty years, but they have not gone away. This paper examines the EuSpRIG horror stories archive as evidence of how ordinary spreadsheets fail and the consequences that follow. After cleaning and deduplication, 121 incidents from 1995 to 2026 were classified into twelve failure categories and analysed by mechanism, context and consequence. Formula errors, data-entry slips and data-handling mistakes account for 71 cases and recur across the whole period, from Fidelity’s missing minus sign in 1995 to a mistyped date that cost Norway’s sovereign wealth fund about 92 million in 2024. A second pattern concerns information present in a workbook but not visible to the recipient: hidden rows, hidden sheets, pivot-table caches, embedded objects and retained source layers can cause harm even when no calculation is wrong. The clearest rise in reported cases is in this hidden or embedded data category, whose most serious example is the 2022 Ministry of Defence spreadsheet that exposed details of roughly 18,700 Afghan applicants through hidden rows. The paper argues that spreadsheet risk is no longer only a problem of wrong numbers. It also involves visibility, disclosure, governance and trust. Existing taxonomies remain useful for formula, entry and handling errors, but do not fully capture the public harm caused by spreadsheets used as uncontrolled operational infrastructure. The same mistakes have persisted for three decades, what has changed is the authority organisations give to the spreadsheets that contain them.
[HC-46] Neuron-Level Architecture Growth: A Controlled Evaluation for EEG Time-Series Decoding ICASSP2027
链接: https://arxiv.org/abs/2609.33880
作者: Adam Mounir,Stella Douka,Arnault H. Caillet,Bruno Aristimunha,Sylvain Chevallier
类目: ignal Processing (eess.SP); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
备注: 5 pages, 3 figures, 2 tables. Submitted to ICASSP 2027
Abstract:Convolutional EEG decoders are trained at a fixed width, usually set by their authors on other data. Growing methods add neurons during training where the loss could decrease the most, but whether they improve compared to a reference width is untested on EEG. Here, we grow three convolutional backbones on 12 motor-imagery datasets under three protocols and compare each with its reference model per subject. The growing ShallowFBCSPNet scores 2.9 points above its reference model with only half the parameters (0.57x), SCCNet changes by at most 1.2 points. Deep4Net growing models show decreased accuracy, but they require adaptation that prevent to compare faithfully the results. These differences follow the selection step, which keeps a candidate neuron relying on a dynamic threshold from singular values decomposition. Overall, these results suggest that growth helps when its criterion can rank the candidate neurons, and that the rate of skipped neuron addition tells where a decoder can be grown small from scratch. Comments: 5 pages, 3 figures, 2 tables. Submitted to ICASSP 2027 Subjects: Signal Processing (eess.SP); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE) Cite as: arXiv:2609.33880 [eess.SP] (or arXiv:2609.33880v1 [eess.SP] for this version) https://doi.org/10.48550/arXiv.2609.33880 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[HC-47] MAESTRO: a Multimodal Auditory-attention Egocentric Speech-TRacking Open corpus
链接: https://arxiv.org/abs/2609.31898
作者: K M Naimul Hassan,Ali Alavi,Donald S. Williamson
类目: Audio and Speech Processing (eess.AS); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG); Sound (cs.SD); Neurons and Cognition (q-bio.NC)
备注: 11 pages, 8 figures. Submitted to IEEE Transactions on Audio, Speech, and Language Processing
Abstract:Humans rely on gaze, head movements, and visual cues to attend to speakers in noisy environments, yet auditory attention decoding (AAD) has been studied primarily using electroencephalography (EEG). We introduce the Multimodal Auditory-attention Egocentric Speech-TRacking Open (MAESTRO) corpus, the first AAD dataset to simultaneously record EEG, eye gaze, pupillometry, egocentric video, and head inertial measurement unit (IMU) data. MAESTRO includes four competing speakers and background noise across multiple signal-to-noise ratio (SNR) conditions, enabling attention decoding under realistic listening scenarios. Through a four-speaker attention decoding benchmark, we show that combining behavioral and physiological signals improves decoding performance over EEG-only approaches, enabling future advances in multimodal auditory attention decoding. These findings open the door to new applications, analyses, and methodological advances in multimodal AAD. The complete dataset is publicly available at this https URL . The official code repository is available at this https URL .
计算机视觉
[CV-0] FurE: Efficient Instance-Specific 3D Fur Reconstruction without Animal-Fur Datasets
链接: https://arxiv.org/abs/2609.35770
作者: Srinjay Sarkar,Prakhar Kaushik,Soumava Paul,Alan Yuille
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Graphics (cs.GR)
备注: 14 pages, 13 figures, 4 tables. Project page: this https URL
Abstract:Realistic and editable animal fur reconstruction from multi-view images is challenging due to fine-scale detail, self-occlusion and obfuscation, and, unlike human hair, the lack of animal-fur datasets. Fur usually covers most of an animal’s body, with large inter-species and intra-species variability. We present FurE, an efficient strand-based animal fur reconstruction method that recovers a per-strand, editable groom by optimizing a root-conditioned latent field, decoded into strand geometry via a PCA-based decoder. We reconstruct a defurred animal body using local fur-thickness cues from a surface-constrained Gaussian Frosting representation together with part-based priors. We further show that a PCA-based decoder learned from human-hair strand data can alleviate animal-data scarcity while enabling substantially faster optimization. FurE achieves a 10x speedup in strand training over current SOTA dense per-strand optimization while retaining strand fidelity and generalizing across synthetic and real-world sequences, with quantitative and qualitative validation despite the reduction in training time.
[CV-1] PDMD: Projected Distribution Matching Distillation for Video Diffusion Models
链接: https://arxiv.org/abs/2609.35768
作者: Zimo Wang,Junkun Yuan,Angtian Wang,Haotian Yang,Canyu Zhang,Siyuan Yuan,Xingchang Huang,Bo Liu,Yizhi Wang,Yiding Yang,Chongyang Ma,Gordon Guocheng Qian
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Modern video diffusion models require tens of denoising evaluations over long spatiotemporal token sequences. Distribution Matching Distillation (DMD) reduces the number of function evaluations (NFE) to just a few. However, DMD samples can degrade during training, exhibiting progressive oversaturation and artifacts. We trace this instability to critic errors, which enter successive student updates and accumulate over time. We introduce Projected Distribution Matching Distillation (PDMD) to filter critic errors. PDMD projects out the component of the DMD update parallel to the student-critic endpoint residual. At a fixed noisy query, we prove that this residual is an unbiased estimate of the critic’s endpoint error. Under high-dimensional assumptions, this projection removes a constant fraction of critic error while discarding only a vanishing fraction of ideal DMD signal. Empirically, the projection stabilizes training and improves sample quality where DMD degrades and develops unnatural textures. PDMD requires only a one-line code change to DMD, with no extra loss, network, data, model pass, or multi-stage training. With Wan2.1, PDMD achieves a VBench total score of 83.73 at 4 NFE, surpassing matched DMD by 1.03 points. On MiniMax-H3 joint video-audio generation, PDMD achieves a VideoGen-Eval visual total score of 83.17, 0.41 points above the strongest distilled baseline. PDMD also achieves the best performance on all six audio metrics among the compared 4-NFE models. Qualitative comparisons and user studies favor PDMD over the distilled baselines in visual quality, motion, and audio quality. Code and models are available at this https URL.
[CV-2] Learning Native Reflection in Unified Models with Interleaved Reinforcement Learning
链接: https://arxiv.org/abs/2609.35767
作者: Yijia Fan,Ziqi Huang,Zhongang Cai,Yan Li,Zimo Wen,Wanqi Yin,Haiwen Diao,Ziwei Liu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Unified multimodal models can both look at and render images, so in principle they can repair their own generations: diagnose what an image gets wrong, revise it, observe the result, and diagnose again. Whether a revision helps is known only after it is rendered, so the reflection text and the image generation must be learned jointly, over the whole loop. Supervised fine-tuning (SFT) on reflection trajectories gives a cold start but does not find the high-success repair paths, and naive RL that optimizes only the renderer or only one head leaves most of the gain untapped. We introduce UMM-Reflection, which applies reinforcement learning (RL) to complete reflection trajectories inside one unified model: sibling trajectories share one initial image, so the group-relative advantage compares reflection strategies, and one trajectory-level advantage updates both the reflection tokens and the flow-based revisions, avoiding the combinatorial blow-up of per-round credit assignment. Unlike single-round editing or pipelines with an external critic, credit flows across rounds and to both roles of the same model, and no verifier is needed at inference. On BAGEL, UMM-Reflection improves GenEval by 12.05 points over SFT, and the gains transfer to WISE (+10.97), OneIG-Bench (+3.48), and T2I-CompBench++ (+4.63), none of which is used in training.
[CV-3] Copy the Same Distill the Difference: Initializing Linear Vision Transformers
链接: https://arxiv.org/abs/2609.35745
作者: Huaiyuan Qin,Muli Yang,Gabriel James Goenawan,Shiqi Huang,Min Kass Chong,Wahyu Wiratama,Peng Hu,Chen Gong,Wu Liu,Xi Peng,Chun Jian Ho,Hongyuan Zhu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Linear Vision Transformers (ViTs) are designed to replace the attention in Softmax ViTs with the linear-complexity attention operator for more efficient token routing, but they require from-scratch pre-training and typically underperform the original Softmax version. How to initialize linear ViTs both efficiently and effectively still remains unclear. In this work, we explicitly ask: given that most foundation ViTs are built on the mainstream Softmax attention, can linear ViTs benefit from their pre-trained weights? Recent works on Attention Transfer show that attention is the effective transferable component between Softmax ViTs, suggesting attention alone suffices for such reuse. However, we find the opposite for Softmax-to-linear transfer. The attention weights are operator-specific: copying them barely helps, and is sometimes even worse than random initialization. Instead, the attention’s token routing behavior can be recovered through distillation with a proper loss design, letting linear ViTs reduce the gap and even match Softmax ones. In contrast, the MLP weights, which carry the learned representation, are operator-agnostic: they can be transferred by simple direct copying, which already carries most of the benefit of the pre-trained weights. Thus, copying MLPs can serve as an effective foundation for Softmax-to-linear transfer: paired with the distilled attention, linear ViTs eventually close the remaining gap and even surpass Softmax ones. These findings hold consistently across various linear ViT variants, different model sizes, and diverse datasets. We hope this study deepens the understanding of reusing pre-trained weights across attention operators: copy what stays the same and distill what differs, to recover the benefit across the Softmax-to-linear boundary.
[CV-4] InfiniHand: Streaming World-Space Hand Motion Estimation from Egocentric Video
链接: https://arxiv.org/abs/2609.35743
作者: Kerui Ren,Kaiwen Song,Weiguang Zhao,Yuxi Wang,Yufei Liu,Bo Dai,Haoyu Guo,Chunhua Shen,Mulin Yu,Tao Lu,Junting Dong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:World-space hand motion estimation from egocentric video requires recovering 3D articulated hand geometry while tracking camera egomotion. Existing approaches heavily rely on cascading independent hand pose estimators and SLAM systems, resulting in error accumulation, complex pipelines, and severe computational overhead. To address these limitations, we present InfiniHand, an end-to-end streaming feed-forward framework that jointly estimates MANO parameters, camera trajectories, and hand locations directly from uncalibrated egocentric video. InfiniHand integrates persistent spatiotemporal memory with hand-centered visual features, explicitly coupling camera motion with local hand geometry within a unified architecture. We train InfiniHand in two progressive stages by first learning robust camera-space hand priors and then extending to streaming world-space reconstruction. To support this process, we aggregate a pretraining corpus of approximately 5,000 hours of egocentric data across multiple public datasets. Extensive evaluations demonstrate that InfiniHand outperforms state-of-the-art baselines on in-domain benchmarks, achieving a 21.4% reduction in ARCTIC PA-p compared to ViDiHand while substantially mitigating world-space drift. Furthermore, InfiniHand generalizes robustly to in-the-wild videos and operates at 11.19 FPS, delivering more than twice the throughput of HaWoR.
[CV-5] GeoVerse: World-Consistent Novel View Synthesis in Geometric Latent Space
链接: https://arxiv.org/abs/2609.35734
作者: Kerui Ren,Tao Lu,Linning Xu,Changjian Jiang,Mu Huang,Chunhua Shen,Mulin Yu,Bo Dai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:Novel view synthesis from sparse images must reconcile faithful reconstruction of observed regions with plausible completion of unseen content, while maintaining world consistency across viewpoints. Existing geometry-based methods preserve observed scene structure but often struggle to complete unseen regions, whereas video generative models offer rich appearance priors but accumulate inconsistencies during sequential view generation. We propose GeoVerse, a framework that synthesizes world-consistent novel views by performing generation within the geometric latent space of a pretrained 3D foundation model and injecting appearance priors from a video generative model. Specifically, GeoVerse extracts multilevel features from Wan2.2 VACE and injects them into the geometric latent diffusion model via a ControlNet-style adapter, incorporating video-learned appearance priors to enhance structural completion. To enforce cross-view coherence, a global spatial memory continuously aggregates observed and synthesized content, reprojecting target-aligned guidance to anchor subsequent predictions to a shared scene representation. Extensive experiments across diverse datasets demonstrate improved visual quality and geometric consistency, with a 2.23 dB higher PSNR on DL3DV and 32.4% lower ATE on Mip-NeRF360 compared to GLD.
[CV-6] FlowAct-R2: Beyond Talking Avatar via Streaming Multimodal References and Proactive Agent Planning
链接: https://arxiv.org/abs/2609.35728
作者: Ziyao Huang,Zhengkun Rong,Shiyang Qin,Shuang Liang,Wentao Hu,Yuxuan Luo,Yuan Zhang,Mingyuan Gao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL Hugging Face Space: this https URL
Abstract:We present FlowAct-R2, a framework for interactive humanoid video generation that combines continuous multimodal control with proactive agent planning. Our method consists of two coupled components. First, a Streaming Multimodal Reference Diffusion Transformer adapts the pretrained Seedance 2.0 Mini reference-to-video backbone to accept rolling action prompts, streaming audio, and dynamically updated image, audio, and video references. Video-driven rotary positional embeddings align reference chunks with the generation timeline, while reference-plus-image conditioning and partially noised historical motion frames preserve appearance and avoid accumulated drift. Second, a Proactive Interaction Agent separates pre-online planning from online scheduling and response: it prepares a persona, a long-horizon agenda, and reusable multimodal skills in advance, then autonomously schedules behaviors, responds to audience input, and handles interruptions during a live session. FlowAct-R2 supports real-time 720p generation and hour-scale streaming across entertainment streaming, live shopping, video chatting, and live vlogging.
[CV-7] Impact of Patient Orientation in Single- and Multi-View Camera Environments for AI-based Rehabilitation Monitoring
链接: https://arxiv.org/abs/2609.35726
作者: Miriama Jánošová,Andreas Lang,Petra Budikova,Jan Sedmidubsky
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Automated quality assessment of rehabilitation exercises relies heavily on accurate human pose estimation from video data. Although numerous RGB-based pose estimation methods have been proposed, the impact of camera placement on detecting clinically relevant movement errors remains insufficiently explored. To address this gap, we introduce REHAB26-ViewAngles, a dataset comprising correct and incorrect rehabilitation exercise executions captured from a wide range of camera angles. Furthermore, we propose a novel separability metric to quantify an algorithm’s ability to distinguish between valid and faulty exercise repetitions. Using these tools, we analyze how various RGB-based pose-estimation strategies are suitable for exercise quality assessment under varying camera placements. In particular, we analyze single-camera 2D and 3D pose estimation and four multi-camera strategies: a combination of two orthogonal 2D views, 3D triangulation, weighted 3D fusion, and an AI-based pose-estimation transformer model specifically trained from two synchronized cameras. Our findings reveal that an optimally placed 2D camera can improve the separability by 16.9,% over the commonly used 0^\circ frontal view and frequently outperforms single-camera 3D estimation, while combining two views can further improve accuracy by up to 13.1,%. These results offer practical guidance for deploying rehabilitation monitoring in both home and clinical settings.
[CV-8] Superquadric Primitive Decomposition of 3D point clouds via Geometric-Aware Inlier Refinement
链接: https://arxiv.org/abs/2609.35725
作者: Alessandro Rinaldi,Edoardo Tedesco,Andrea Ferraris,Filippo Leveni,Daniele Baieri,Filippo Maggioli,Simone Melzi,Luca Magri
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 19 pages, 11 figures, under review
Abstract:The decomposition of 3D point clouds into interpretable geometric primitives remains a longstanding challenge in Computer Vision and Computer Graphics. Among the available representations, superquadrics offer a compact and expressive model capable of capturing a wide range of shapes. However, their estimation is inherently challenging, as it requires solving a non-linear optimization problem and is particularly sensitive to noise, outliers, and overlapping structures. While robust estimation methods such as RANSAC and its variants achieve strong performance, they rely primarily on spatial proximity and residual-based criteria, often leading to incorrect inlier assignments across adjacent or complex arrangements of primitives. In this work, we introduce a geometric-aware framework for primitive decomposition that explicitly incorporates local surface properties into the fitting process. Specifically, we propose an inlier refinement step formulated as an energy minimization problem and solved via graph-cut optimization. Our formulation integrates geometric priors, such as normal consistency, enabling more reliable inlier selection beyond purely residual-based criteria. The approach naturally applies to both single-model estimation and multi-model decomposition. By leveraging geometric information beyond point-wise residuals, our method reduces erroneous inlier propagation and stabilizes parameter estimation. Experiments on synthetic and real datasets show consistent improvements in geometric accuracy, robustness to noise and outliers, and convergence efficiency compared to state-of-the-art RANSAC-based methods.
[CV-9] Hard Vision Easy Vision: What GPT -6 Astra Reveals Across Computer Vision
链接: https://arxiv.org/abs/2609.35718
作者: Hanoona Rasheed,Mohammed Irfan Kurpath,Bin Ren,Hisham Cholakkal,Fahad Shahbaz Khan,Salman Khan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Frontier general-purpose systems are rapidly expanding beyond visual understanding into capabilities traditionally handled by dedicated computer-vision models. As these capabilities expand, a central question for the computer-vision community is how far this reach extends, and what remains hard. We evaluate GPT-6 Astra alongside five frontier general-purpose AI systems across 34 capabilities and 55 benchmarks spanning nine areas of computer vision. We compare their performance with dedicated models and humans where suitable references are available. Astra demonstrates broad visual capability, with substantial gains over other frontier systems in visual and spatial reasoning and several forms of structured prediction. Across the state-of-the-art systems, a consistent pattern emerges. Capabilities involving semantic interpretation, reasoning, and object-centric prediction increasingly approach or reach available reference levels. In contrast, larger gaps remain when tasks require metric geometric accuracy, faithful reconstruction, temporally consistent dense prediction, or specialized fine-grained visual knowledge. Additional reasoning and specialist tools close selected gaps, but their benefits vary across capabilities. These results map a changing landscape of computer vision in which increasingly sophisticated visual tasks are accessible through a general-purpose interface, while precise and fidelity-sensitive perception remains an important frontier.
[CV-10] Lagrangian–Hamiltonian Flows for Video Prediction and Image Generation: A Symplectic Perspective
链接: https://arxiv.org/abs/2609.35710
作者: Jiawei Hu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We introduce LHFM, a geometric framework for learning image dynamics. Drawing on structures central to classical mechanics, symplectic geometry, and geometric quantization, LHFM represents each image as an exact Lagrangian graph and models its evolution through image-dependent Hamiltonian flows, which yield a transport–source parameterization of image velocities. Our primary application is deterministic video prediction: LHFM-V is a recurrent model that advances frames by integrating predicted transport and source fields, and achieves the lowest reported FLOP count among the compared recurrent models with similar prediction accuracy. The image variant, LHFM-I, shows that the same construction is compatible with flow matching: in a matched experiment, it attains a lower FID than the flow-matching baseline.
[CV-11] Mind the RefGAP: Correcting Reference Attention in Diffusion-Based Visual Editing
链接: https://arxiv.org/abs/2609.35708
作者: Yanan Wang,Shengcai Liao,Guangyi Liu,Xiaodan Liang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reference-guided diffusion editors struggle to faithfully reproduce user-provided references. We identify a potential bottleneck in diffusion editors: many methods provide limited reference-attention allocation. For example, in LoomVideo, edit-region queries assign less than 1% of their attention mass to the reference. We introduce RefGAP, a training-free correction that determines logit-offset magnitudes online at each layer from the reference-attention mass measured during the forward pass. Positive offsets to reference logits strengthen reference usage by edit-region queries, while negative offsets for keep-region queries limit reference-induced changes outside the edit. Two global coefficients control the correction; they are selected once on validation data from four development diffusion editors and held fixed. Across seven diffusion-based image/video editors, RefGAP improves identity fidelity in head swapping and face swapping. RefGAP achieves a fidelity-preservation trade-off comparable to separately tuned constant edit-side biases, without per-approach strength sweeps. Additional experiments on virtual try-on and background replacement evaluate transfer beyond identity editing.
[CV-12] DynaTokens: Teaching Dynamics to Camera-Controlled Video Models at Test Time ALT
链接: https://arxiv.org/abs/2609.35704
作者: Ma Ziqi,Chen Hongqiao,Gkioxari Georgia
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project website: this https URL
Abstract:Video generation must account for two sources of motion, one induced by the observer’s camera path and the other caused by scene dynamics. An ideal camera-controlled video model should account for both motions: let users move the camera while evolving the scene dynamics. While current models handle camera-induced motion well in static settings, they struggle for dynamic scenes: objects are static, move incorrectly, or degrade in generation quality. We introduce DynaTokens, a lightweight set of learnable scene-specific tokens that teach dynamics to an existing camera-controlled world model. Our method is motivated by a simple asymmetry between the two sources of motion: whereas camera motion affects the generated view globally, object dynamics are spatially localized. Through cross-attention, DynaTokens trains the learnable tokens from a few example trajectories for a scene while keeping the base model frozen, and enables dynamics under new query camera paths. DynaTokens achieves a better simultaneous dynamics-camera tradeoff on VBench2 and WorldScore evaluations than LoRA, block finetuning, and specialized trainable-layer baselines. Analyses of token attention, ablations, and motion temporality suggest that matching the trainable interface to the structure of the learning target is important for effective adaptation. Project website: this https URL
[CV-13] FlowTool: Controlling Tool Parameter in Image Retouching via Flow Matching
链接: https://arxiv.org/abs/2609.35673
作者: Thanh-Long V. Le,Steven Walton,Seunghyun Yoon,Branislav Kveton,Trung Bui,Eunho Yang,Viet Lai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Tool-based image editing (image retouching) is commonly formulated with autoregressive multimodal large language models (MLLMs) that sequentially generate reasoning, tool selections, and parameter values. In this work, we present a novel approach to tool-based image editing by framing the task as a flow matching problem. We introduce FlowTool, a framework that directly models the distribution of high-quality tool parameters conditioned on the input image and user instruction using conditional rectified flow. FlowTool combines a vision-language model backbone for multimodal understanding with a Diffusion Transformer parameter generator that transforms Gaussian noise into an editing plan. We train FlowTool with a two-stage supervised flow-matching curriculum, followed by reward-based post-training. Across MMArt-Bench, FlowTool-Eval, ArtEdit-Bench, and MIT-Adobe5K, FlowTool achieves significantly stronger reference-based performance than specialized MLLM editing agents and proprietary MLLMs, while remaining competitive with proprietary models under reference-free evaluation. Moreover, FlowTool significantly improves inference efficiency, reducing latency by at least 50\times while requiring nearly 2\times less memory than the compared baselines. These results demonstrate that tool-based image editing can be effectively modeled as conditional generation over structured continuous editing parameters, without autoregressive reasoning.
[CV-14] Many Eyes One World: Feed-Forward 3D Reconstruction from Mixed Cameras
链接: https://arxiv.org/abs/2609.35658
作者: Qiaoge Li,Yifan Zhan,Haijun Yang,Haiyang Liu,Yiyi Cai,Chenchi Luo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 24 pages, 9 figures, 14 tables
Abstract:Real-world capture is heterogeneous: perspective, fisheye, and 360^\circ panoramic images can coexist within a single reconstruction task, yet most feed-forward 3D reconstruction models assume perspective imagery and a uniform input representation. Recent models handling several camera types are either informed of the camera type for each view or reconstruct one image pair at a time. No single-pass method reconstructs mixed-camera tuples containing full panoramas from images alone. We present MEOW, a feed-forward system that jointly reconstructs metric pointmaps and camera poses from one N-view tuple mixing perspective, fisheye and full-panorama images, in a single forward pass from images alone: no calibration, distortion parameters, camera-type labels or poses are supplied for any view. Our guiding design philosophy is to treat heterogeneous-camera reconstruction as a data-adaptation problem rather than an architectural redesign. MEOW retains a perspective-pretrained backbone and learns heterogeneous cameras entirely from a procedural data engine, which renders each scene across a continuous manifold of camera models with exact rays and depth, and certifies covisibility for every camera-sampled training tuple. Trained on synthetic tuples only, MEOW transfers zero-shot to real captures: on heterogeneous 2D3DS tuples it achieves 79.9 mAA@30 against 53.8 for Wid3R given the camera type of every view; on our laser-scanned mixed-camera benchmark it registers every four-view mixed tuple with 79.4 AUC@30. The data engine, benchmark, and complete evaluation pipeline will be released.
[CV-15] Verifiable Visual Rewards Transfer from Synthetic Scenes to Natural Prompts
链接: https://arxiv.org/abs/2609.35641
作者: Shuyue Stella Li,Xiaochuang Han,Yulia Tsvetkov,Luke Zettlemoyer
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 33 pages, 10 figures, 18 tables
Abstract:Precise instruction following in image generation, such as satisfying object counts and spatial relations, remains an open challenge at least in part because it is learned using unreliable reward models such as object detectors and vision-language models. We introduce Verifiable Visual Rewards (VVR), the first framework for programmatically verifiable image rewards, and show that training on it generalizes to natural prompts. Each VVR task is a scene of geometric objects and relations among them, from which we derive both the prompt and a deterministic verifier, so tasks can be generated in any number and at any chosen complexity. We release VVRBench, with 10,000 tasks over 32 constraint types, and VVRBench-Challenge, with 720 more complex tasks; the strongest model we evaluate—GPT-Image-2.5—solves 21.4% of VVRBench-Challenge. Using VVR scores as rewards for reinforcement learning (RLVVR) raises the accuracy of Stable Diffusion 3.5 Medium on VVRBench from 2.8% to 28.3% and demonstrates consistent easy-to-hard generalization. These gains extend to out-of-domain benchmarks, and mixing VVR into existing objectives further improves overall performance and human preference, motivating the adoption of VVR into standard image generation post-training recipes.
[CV-16] RT-Super: Learning Tumor Segmentation from Longitudinal Images and Reports MICCAI2026
链接: https://arxiv.org/abs/2609.35637
作者: Pedro R. A. S. Bassi,Wenxuan Li,Hanxue Gu,Jieneng Chen,Xinze Zhou,Zheren Zhu,Sezgin Er,Ibrahim E. Hamamci,Bjoern H. Menze,Gulhan E. Akan,Kang Wang,Yang Yang,Alan L. Yuille,Zongwei Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: MICCAI 2026
Abstract:Multi-tumor segmentation is important for early cancer detection and allows radiologists to visualize, verify, and understand AI predictions. However, tumor segmentation masks are expensive, time-consuming, and unavailable for many tumor types in public data. Instead, hospitals have vast, readily available data that can guide segmentation: radiology reports, longitudinal images, and multi-phase images. We use this readily available data to substitute for tumor masks in training AI for tumor segmentation. To this end, we propose a new architecture, RT-Super. It has a teacher network, which analyzes the patient’s longitudinal images and reports to create high-quality tumor masks. These masks train a student network, which sees a single image and no report. At inference, when longitudinal images and reports are unavailable, we use the student. RT-Super uses a new CNN-Transformer architecture and novel Consistency Losses that exploit tumor location consistency across longitudinal images. We train RT-Super to segment esophagus, uterus and spleen tumors, which have few or no public masks. Even without training masks, RT-Super can segment these tumors and surpass public AI models. Overall, we demonstrate that learning from longitudinal images, multi-phase images, and reports can overcome mask scarcity and advance multi-cancer detection and segmentation. Code: this https URL
[CV-17] EvolvingAvatar: Interactive 3D Head Generation That Adapts as Conversations Unfold
链接: https://arxiv.org/abs/2609.35616
作者: Junjie Chen,Fei Wang,Kun Li,Yiqi Nie,Xun Yang,Yanbin Hao,Linfeng Zhang,Meng Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:Interactive 3D head generation requires coordinated speaking and listening motion that responds to an evolving conversation. Existing generators use incoming observations as context but keep their parameters fixed, leaving conversational patterns unused as a learning signal. We introduce EvolvingAvatar, a causal generator that uses test-time training to adapt to user face video and dyadic audio during interaction. Its dyadic context prediction objective provides a self-supervised learning signal from audiovisual context without target motion labels at test time. Persistent fast weights accumulate these updates within each conversation to guide motion generation, while transient jaw adaptation responds to current audiovisual context. Predicted speech activity controls how persistent adaptation guides motion. We also introduce InterHead-Bench, a unified 455.95-hour benchmark built from single-view and dual-view conversation videos. Experiments show improved conversational motion statistics over strong baselines. On the hardest out-of-distribution split, generation improves as conversations unfold, reducing mismatch with recorded user-avatar expression statistics by up to 11.1% from the first interval.
[CV-18] Remote Sensing Sparse-View 3D Gaussian Splatting via Depth Image-Based Rendering
链接: https://arxiv.org/abs/2609.35612
作者: Jiaming Kang,Zhengxia Zou,Zhenwei Shi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Remote sensing novel view synthesis under sparse observations remains challenging due to insufficient geometric constraints and limited cross-view supervision. Existing Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) methods are prone to overfitting and face challenges of depth ambiguities, missing cross-view information, and insufficient constraints in under-observed regions. To address these challenges, we propose DIBR-GS, a neural Gaussian Splatting framework that exploits Depth Image-Based Rendering (DIBR) to generate pseudo views for cross-view consistency supervision. Specifically, reliable geometric initialization is constructed by aligning monocular depth priors with sparse SfM reconstruction, and cross-view appearance priors are incorporated into neural Gaussian representations to enhance appearance modeling under sparse observations. Furthermore, we introduce a progressive DIBR-based pseudo-view supervision strategy to provide additional geometric and appearance constraints, enabling more complete reconstruction of weakly observed regions. In addition, a height-constrained anchor growth strategy is designed to suppress unreasonable Gaussian expansion. Experiments demonstrate that the proposed method achieves superior performance over existing approaches when training with only 3 input views. Compared with the previous best-performing method, it improves PSNR by 6.83 dB, with relative gains of 14% in SSIM and 60% in LPIPS, while maintaining competitive computational efficiency. Our code is available at this https URL
[CV-19] On-Policy Self-Distillation for Multi-Turn Image Editing
链接: https://arxiv.org/abs/2609.35611
作者: Liangbing Zhao,Le Zhuo,Mohamed Elhoseiny
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Instruction-based image editing has achieved strong performance in single-turn settings, yet practical editing is often iterative, with each instruction applied to the output of the previous turn. We find that existing editing models degrade rapidly under recursive editing and attribute this failure to a train-test mismatch in the conditioning distribution: models are trained on clean source images but must repeatedly condition on their own imperfect outputs at inference time. To address this, we propose MT-OPSD, an on-policy self-distillation framework that trains the model on self-generated conditioning states with editing supervision from a clean-conditioned teacher, without requiring multi-turn annotations. We further introduce LME-Bench, a benchmark of 100 ten-turn editing sessions for evaluating long-horizon robustness. Experiments across three editing backbones show that MT-OPSD substantially improves long-horizon editing success and reduces multi-turn collapse while largely preserving single-turn editing quality.
[CV-20] ReSS: Residual-Restoring Sparse Attention for 3D Vision Transformers
链接: https://arxiv.org/abs/2609.35593
作者: Yongsung Kim,Jaehoon Lee,Minjun Park,Wooseok Song,Hun Hwangbo,Sungroh Yoon
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:3D vision transformers such as VGGT predict camera poses and scene geometry from multi-view images in a single forward pass, but their global attention over all concatenated view tokens dominates computation as the number of views grows. To reduce this cost, SparseVGGT and HeSS sparsify attention at the block level, and both retain blocks with high attention probability. However, we observe that attention probability poorly predicts how much the model’s behavior actually changes when a block is removed, and we show that this mismatch is why performance collapses as sparsity increases. In this paper, we propose ReSS (ReSidual-ReStoring Sparse Attention), which recasts block selection from a problem of maximizing the retained attention mass to one of minimizing the drift that sparsification leaves in the residual stream. We introduce a drift score that quantifies how much each block shifts the residual, and, since the drift of a drop set depends on the directions of the contribution vectors rather than on their magnitudes alone, an iterative residual restoration procedure that refines the drop set as a whole. Across three backbones and five datasets, ReSS preserves dense performance better than prior methods at matched sparsity. Two further results support drift as the quantity that governs the cost of sparsification: maximizing drift degrades performance faster than random selection, and plotted against realized drift instead of sparsity, all methods fall approximately onto a single curve. Code is available at this https URL.
[CV-21] What Paired Evaluations Reveal under Visual Perturbations
链接: https://arxiv.org/abs/2609.35583
作者: Yongda Wei,Chen Zhang,Yifei Wang,Xinyu Wang,Bosen Shao,Hanxi Li,Liping Di
类目: Computer Vision and Pattern Recognition (cs.CV); Applications (stat.AP)
备注: 53 pages, 8 figures, including appendices
Abstract:Robustness evaluation must examine diverse visual perturbations, while benchmarks cover only some real-world conditions and physical testing is costly. Paired evaluations link clean and perturbed predictions for the same image, capturing changes in correctness, confidence, and acceptance beyond aggregate accuracy. We investigate how this image correspondence supports two needs in robustness evaluation: interpreting paired evaluation results and prioritizing samples for physical testing. To interpret paired evaluation results, we fix both sets of prediction records and vary their correspondence within each class. We prove that classwise correct-correct counts give the same sharp bounds on lost acceptance and mean true-class probability decrease among retained-correct inputs as any feasible five-state refinement. Distinguishing persistent from changed wrong answers can further constrain accepted-error transitions, while shared correspondence can establish policy orderings left unresolved by separate cost intervals. To prioritize samples for physical testing, we retain each image’s synthetic responses and rank clean-correct images by their mean true-class probability under corruption. Across 44 classifiers, testing the highest-risk 20% finds 67% and 45% of failures under mild screen and print recaptures, versus 58% and 36% for clean confidence and 60% and 37% for an equal-size natural-transformation average. With both probability averaging and an A3Rank scoring adaptation, the tested corruption set yields higher mean failure recall than the natural-transform set; differences between scores depend on the source and budget. Together, these findings show that the value of correspondence depends on the evaluation objective: classwise counts suffice for specified reliability bounds, while image-specific synthetic responses improve the allocation of physical tests within the evaluated pool.
[CV-22] Less Is More: Genetic Frame Selection for Efficient Novel View Synthesis
链接: https://arxiv.org/abs/2609.35573
作者: Diego E. Farchione,Ramzi Idoughi,Alberto Jaspe-Villanueva,Peter Wonka
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Feed-forward novel view synthesis reconstructs a scene from many input images in a single forward pass, yet more views do not necessarily improve performance: redundant or poorly chosen frames increase computational cost and may degrade reconstruction quality. We address the problem of selecting, from an already captured sequence, a fixed-size subset of input views that is most informative for reconstructing specified target viewpoints. We propose a render-free view selector that scores candidate frames based on three complementary criteria: target-view coverage, measured against observed frames that stand in for the targets, redundancy with previously selected views, and image sharpness. A lightweight scoring network then selects the most informative frames without rendering, reconstruction, or per-scene optimization at inference time. To train the selector, we distill an expensive offline search procedure in which a genetic algorithm identifies high-quality subsets by directly optimizing reconstruction performance on training scenes. The selector learns to reproduce these choices from geometric and image-level features alone. Across six datasets and multiple input budgets, our method consistently outperforms both geometric and reconstruction-aware view-selection baselines while incurring significantly lower selection costs than reconstruction-based alternatives. Moreover, carefully selected subsets can outperform feed-forward reconstruction from the full input sequence. The learned selector generalizes across diverse reconstruction paradigms (feed-forward, 3D Gaussian Splatting, and NeRF), to object-targeted reconstruction and to a cross-capture setting in which the target views come from a separate acquisition pass. More broadly, our results indicate that explicitly reasoning about target relevance and inter-view redundancy is a fundamental factor in efficient scene reconstruction.
[CV-23] EdgeVLN: Runtime-Aware Deployment Ready Quantized Vision Language Navigation Model
链接: https://arxiv.org/abs/2609.35570
作者: Rithvik Jonna,Man Namgung,Aakash Gurram,Tinoosh Mohsenin
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-language navigation (VLN) models perform well but target compute-rich platforms, limiting deployment on memory- and power-constrained robotic edge devices. Compression alone does not establish whether a VLN model fits the memory, latency, and energy budgets of an edge platform while preserving navigation behavior. We introduce EdgeVLN, a runtime-aware, deployment-ready quantized VLN model that closes this gap. EdgeVLN combines a quantized StreamVLN model with Latent Trajectory Termination Extractor (LATTE), a lightweight causal transformer that improves real-time stopping by predicting a Stop Action verifier rank. Both execute through our this http URL VLN driver, which reconstructs streaming context and prunes memory tokens on-board. We characterize a pretrained StreamVLN backbone across weight quantization from 8 to 2 bits and multiple inference runtimes to identify a feasible operating point. LATTE reuses backbone hidden states within the budget freed by quantization, requiring neither a second vision encoder nor an additional backbone forward pass. We evaluate six backbone precisions and seven candidate stop heads on BF16 and IQ4 NL across all 1,839 R2R VLN-CE val-unseen episodes. We measure success rate (SR) in simulation and latency, energy, and resident memory on an NVIDIA Jetson Orin NX 16 GB. LATTE achieves our highest SR, 58.02 percent on the deployed 4-bit model, exceeding the BF16 baseline with only 0.013 s additional latency per navigation step. Four-bit formats achieve nearly identical SR, but step energy varies 36.8 times by execution path. Only IQ4 NL under our VLN driver fits the board, using 11.35 GB resident memory while running 20.8 times faster and using 13.3 times less energy than storage-streamed BF16. INT2 collapses. Runtime selection, memory-token pruning, and quantization are essential for efficient edge deployment.
[CV-24] Revisiting Risky Tackle Detection with Vision Transformers
链接: https://arxiv.org/abs/2609.35562
作者: Syed Ahsan Masud Zaidi,Lior Shamir,Scott Dietrich
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages
Abstract:This paper is a Track 2 reproducibility companion to an ICPR 2026 study on risky tackle detection in American football prac- tice videos. The original work fine-tuned a Video Vision Transformer (ViViT) on 733 clips labeled with the SATT-3 rubric. It used focal loss, Taguchi L18 augmentation, and 5-fold cross-validation. It reported risky- class recall of 0.67 and risky-class F1 of 0.59. This companion documents the released artifact and traces those numbers to specific scripts, fold out- puts, and aggregation files. The reproduced headline is run_15. It com- bines Gaussian noise with static brightness decrease and uses no rotation and no flip. Its fold-mean risky recall is 0.667 and its fold-mean risky F1 is 0.588. These values match the published headline after rounding. The ablation shows that brightness is the dominant factor. Its risky-recall main-effect range is 0.055, which is larger than the ranges for rotation, flip, and noise. Without augmentation, ViViT reaches risky recall of 0.545 and does not exceed the C3D baseline of 0.583. The raw clips show iden- tifiable student athletes, so they cannot be redistributed. The artifact provides a public sample for pipeline checks and a controlled route for full-data review.
[CV-25] WorldPlay2: Extending Real-Time Interactive World Models in Control and Horizon
链接: https://arxiv.org/abs/2609.35560
作者: Haiyu Zhang,Wenqiang Sun,Tengfei Wang,Junta Wu,Jun Zhang,Yunhong Wang,Yu Qiao,Chunchao Guo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: project page: this https URL
Abstract:Interactive world models require responding in real time to versatile controls and maintaining long-horizon consistency. However, modeling heterogeneous controls remains difficult, while explosive contexts and unstable distillation impede achieving both long-horizon consistency and real-time responsiveness. In this paper, we present WorldPlay2, an interactive world model that couples a factorized hybrid control interface with a co-design of compressed memory and stable distillation. 1) Our factorized hybrid control interface integrates frame-aligned action control with structured semantic control that explicitly disentangles scene appearance, character identity, and dynamic semantic events, thereby facilitating effective control learning. 2) To achieve efficient long-horizon modeling, we compress historical contexts into compact memory tokens shared by the autoregressive student and the bidirectional teacher. This design enables clip-wise, memory-conditioned score evaluation instead of jointly processing an entire long rollout, substantially reducing distillation overhead. 3) We further propose Stable Forcing, which initializes the autoregressive student via a few-step strategy and leverages full-rollout replay to preserve the quality of long-horizon rollouts, ensuring robust and stable distillation. Extensive experiments demonstrate the strong generalizability of our model and its superior performance compared to existing methods.
[CV-26] Learning to Reason with Persistent Object States for Video Instance Segmentation
链接: https://arxiv.org/abs/2609.35539
作者: Yongxue Xu,Boxue Yang,Ziqian Liu,Shaoqiu Zhang,Rui Qian,Haopeng Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 19 pages, 6 figures
Abstract:Video segmentation models maintain object identities by carrying instance information across frames. Under prolonged occlusion, reappearance, or interactions between similar instances, however, an unreliable update can overwrite a valid history and cause persistent identity drift. We introduce POSReasoner, a trainable, plug-and-play framework that explicitly decides when an observation should change an object’s state. Each persistent state records identity, confidence, and absence history. A sparse state-observation graph supports Propose-Verify reasoning: provisional associations are revisited using object history, predicted presence, and competition among identities. The verified decisions determine whether to retain, update, reactivate, or suppress each state, while a learned gate controls the evidence written back to memory. Only verified transitions update the persistent state used in subsequent frames. POSReasoner uses standard video annotations and keeps the base model frozen, enabling integration with diverse VOS and VIS architectures. Experiments across long-term VOS and VIS benchmarks show consistent improvements over strong baselines, with the largest gains under occlusion and object reappearance.
[CV-27] Look Before You Judge: Training-Free Region Mining for Grounded and Explainable Deepfake Detection
链接: https://arxiv.org/abs/2609.35536
作者: Chia-Ling Chen,Yu-Ting Ta,Jian-Yu Jiang-Lin,Tai-Ming Huang,Ling Lo,Po-Ching Chen,Yan-Tsung Wang,Pei-Heng Li,Ling Zou,Hong-Han Shuai,Wen-Huang Cheng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multimodal large language models (MLLMs) can explain deepfake verdicts in natural language, but such explanations are not necessarily visually grounded in the visual evidence underlying the prediction. A model may describe plausible artifacts inferred from language priors rather than from image evidence. Existing grounding methods improve visual reliance through decoding or attention interventions, but they generally strengthen grounding over the entire image, making them ill-suited for forensic artifacts that are subtle, spatially localized, and image-dependent. We propose Look Before You Judge, a training-free framework that formulates explainable deepfake detection as a sequential evidence acquisition process. Instead of directly predicting image authenticity from holistic visual reasoning, our framework first identifies image-specific candidate evidence regions by contrasting the MLLM’s decoder-to-visual attention between an original image and its Gaussian-blurred counterpart. The identified regions are then inspected individually, and the resulting local evidence is integrated with the global image context before reaching a final verdict. The framework operates without manipulation masks, external forensic models, or parameter updates, making it directly applicable to off-the-shelf MLLMs. Across five open-source MLLMs on TriDF and MMTD-Set, our framework improves detection accuracy by up to 12.8%, reduces CHAIR by up to 33.4% and hallucination rate by up to 21.3%, and outperforms representative training-free decoding and attention methods.
[CV-28] AutoRef: Harness Optimization for Agent ic Multi-Reference Image Generation
链接: https://arxiv.org/abs/2609.35530
作者: Yuta Oshima,Ku Onoda,Yusuke Iwasawa,Masahiro Suzuki,Yutaka Matsuo,Hiroki Furuta
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Code: this https URL
Abstract:Recent image generation models can take multiple reference images as input and combine them into a new image. However, multi-reference image generation remains challenging: models may omit or duplicate subjects from the references, or produce images in which multiple subjects appear unnaturally pasted. Recent work has proposed image generation agents that combine image generation models, reasoning models, and a harness, which is an executable program that specifies how reference images are interpreted, how generation is performed, how outputs are diagnosed, and how the final image is selected. In multi-reference generation, however, references play different roles and outputs must satisfy many criteria at once, such as fidelity to each reference and the naturalness of the whole image, so many parts of the harness could be improved, from how references are processed to how outputs are diagnosed. This makes it hard to predict which changes will improve performance and by how much, and good harnesses difficult to design by hand; indeed, human-written harnesses vary widely in performance. We therefore propose AutoRef, which optimizes the harness automatically while keeping both models frozen: a coding agent iteratively rewrites the harness code. AutoRef separates the tasks whose feedback informs proposals from the tasks used to select candidates, and continues the search from a beam of the top-ranked harnesses on the selection tasks. Using this procedure, we discover AutoRef-Harness, which improves the open-weight FLUX.2 [klein] 4B from 5.72 to 7.37 on held-out four-reference tasks of the MultiBanana benchmark, matching or exceeding proprietary models including Nano Banana Pro and GPT-Image-1.5. Without re-optimization, the same harness also improves results when the generator, number of references, benchmark, evaluator, or reasoning model differs from those used in the search.
[CV-29] ReVA: A Scene-Centric Dataset Beyond Repetition for Remote Sensing Video Question Answering
链接: https://arxiv.org/abs/2609.35507
作者: Zhen Yao,Likai Wang,Yuming Yang,Zhihao Zheng,Bo Lang,Qiuyu Tang,Jialu Sheng,Jingqi Xu,Yuehai Yang,Jumal Barker,Xiaowen Ying,Mooi Choo Chuah
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multimodal Large Language Models (MLLMs) have demonstrated remarkable advances in remote sensing. However, existing remote sensing multimodal reasoning benchmarks exhibit two critical limitations: they rely on (i) template-driven questions, which causes repetitive questions; and (ii) static images that fail to capture the inherent temporal nature of drone/UAV videos. This leaves systematic evaluation of remote sensing video reasoning largely unexplored. To address this gap, we introduce ReVA, a new dataset for remote sensing video question answering, designed to assess spatiotemporal, scene-centric, and reasoning-oriented capabilities of MLLMs. ReVA comprises 2,438 drone videos spanning 18 cities worldwide (580K frames) and 22K high-quality question-answer pairs across 11 challenging QA tasks. We develop a semi-automatic annotation pipeline that leverages Text LLMs and MLLMs for question-answer generation with human verification. We evaluate 23 proprietary and open-source Video LLMs on ReVA, exposing fundamental limitations of current models. These findings position ReVA as a critical benchmark toward better remote sensing video understanding and temporal reasoning capabilities for real-world deployments. Our code and dataset are available at: this https URL
[CV-30] SolveEdit: Benchmarking Visual Problem Solving in Generative Models
链接: https://arxiv.org/abs/2609.35504
作者: Wenjie Shu,Yexin Liu,Harold Haodong Chen,Xuerui Qiu,Zehan Wang,Yidi Zhang,Yizhan Chen,Zunwei Wang,Minghao Liu,Qi Chen,Harry Yang,Xiaogang Xu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Machine intelligence is often evaluated through abstract reasoning problems, yet many real-world problems are visual, such as arranging objects, repairing layouts, or tracing routes. Solving these problems requires understanding a scene, inferring what must change to achieve a goal, and realizing that change without disturbing unrelated content. However, existing benchmarks mainly evaluate perception, generation, or explicitly specified transformations, leaving goal-driven visual problem solving underexplored. To bridge this gap, we introduce SolveEpIT, a benchmark for visual problem solving through scene transformation. Given an image and a goal, a model must infer a valid transformation from the request, the scene, or a visually expressed rule, then execute it while preserving unrelated content. SoLvEEDrr contains 2,728 cases. Atomic transition contracts specify required and protected conditions, enabling SoLvEScoRE to measure completion and unintended changes without a single reference output. The strongest evaluated model achieves only57.0% SolvEScore. We further introduce SolveEdiT-PLAN, a two-stage visual planner that instantiates the transition before generation. Under matched single-generation evaluation, it improves SoLvEScoRE by 9.1 points on average across three tested generators, including a gain from 57.0% to 71.6% for GPT-Image-2, without modifying the editor.
[CV-31] Sprout: Building Dynamic Memory While Reasoning for Agent ic Video Understanding
链接: https://arxiv.org/abs/2609.35497
作者: Wei Chen,Xuanyu Zheng,Yancheng Long,Haoyang Xu,Kaiyu Jiang,Bin Wen,Tingting Gao,Han Li,Long Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 20 pages, 9 figures
Abstract:Long video understanding relies on video memory to overcome the context limits of multimodal large language models. Existing methods follow a build-then-reasoning pipeline: memory is built offline for the entire video, then reasoned over as a static source. In practice a long video is shared by several questions, and this pipeline is costly at both ends: with few questions, building memory for the whole video costs far more than answering them; with many questions, the memory is never updated, so what is learned while answering questions is lost to the next question. To alleviate these, we introduce Sprout, an agentic framework that builds memory while reasoning: a temporal tree that sprouts detailed nodes as questions are answered. The agent watches the video segment by segment at a low frame rate, stopping when the current question can be answered, remembers each segment as a coarse node of the tree, and revisits key intervals at a higher frame rate to refine the tree with the recovered details. Once a segment is recorded as text, its video input is removed from the context history, while the original video remains reachable through the video tools. The memory tree and prior question–answer records persist across questions, so the memory is online and dynamic: built from the first question onward and updated by every question thereafter. We find that replacing accumulated video inputs with textual memory substantially reduces context usage while maintaining accuracy, with slight improvements in some settings. Across benchmarks on three models, Sprout achieves competitive or improved accuracy relative to representative offline memory methods, with no upfront construction stage and lower context cost per question.
[CV-32] From Scores to Samples: Elastic Forcing for Autoregressive Video Generation
链接: https://arxiv.org/abs/2609.35491
作者: Chi Zhang,Yueyi Liu,Haoyang Shi,Ruichuan An,Haoyu Li,Yuhang Wu,Sen Cui,Miao Liu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Few-step autoregressive video generation commonly relies on Distribution Matching Distillation (DMD), requiring a bidirectional diffusion teacher and an online fake-score model. We instead learn the rollout distribution directly from reference videos, eliminating both score models during post-training. Our framework minimizes maximum mean discrepancy (MMD) in frozen self-supervised video representation spaces, using a hybrid Nyström–Monte Carlo estimator to balance approximation bias and sampling variance. Memory-efficient replay and gradient subsampling make this objective practical. Using the same architecture and initialization as Self-Forcing, our 1.3B model improves the VBench Total score from 83.80 to 84.64 while retaining 17 FPS. Removing auxiliary score models also enables 14B post-training on eight H200 GPUs. Beyond distillation, learning from reference videos enables the acquisition of new visual styles, semantic concepts, and spatial priors without a target-specific diffusion teacher.
[CV-33] AHMAD: Adaptive Hybrid Multi-task Vision Learning with Assisted Distillation for Keypoint Detection
链接: https://arxiv.org/abs/2609.35490
作者: Mohammad Mahdi,Nedyalko Prisadnikov,Yuqian Fu,Carmelo Scribano,Danda Pani Paudel,Luc Van Gool
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Generalist multitasking vision models aim to unify multiple vision tasks within a single framework, enabling more efficient and versatile learning. However, handling diverse vision tasks – spanning dense and sparse predictions – remains challenging due to their inherently varying output structures. In this paper, we propose AHMAD, a simple yet effective framework for generalist multitask learning that integrates different key vision tasks: semantic segmentation, instance segmentation, depth estimation, keypoint detection, and object detection. Our approach incorporates these five tasks into a unified structure: a shared encoder-decoder with several lightweight task-specific projectors. Under the multitask learning paradigm, we observed a complementary performance gain, achieving a state-of-the-art PQ of 53.1 and an mIoU of 66.5 for COCO-val panoptic and semantic segmentation, respectively. Additionally, for top-down keypoint detection, which typically incurs high computational overhead due to multiple forward passes, we introduce a knowledge distillation-based method that enables a single forward pass over the entire image, greatly improving efficiency. Ultimately, our model delivers a lightweight yet effective generalist multitask learning framework, demonstrating strong performance across five vision tasks.
[CV-34] Handwritten Text Recognition Lives in the High-Pixel Variance Subspace NEURIPS2026
链接: https://arxiv.org/abs/2609.35473
作者: Carlos Garrido-Munoz,Jorge Calvo-Zaragoza
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted at 40th Conference on Neural Information Processing Systems (NeurIPS 2026)
Abstract:In self-supervised pretraining for Handwritten Text Recognition (HTR), pixel reconstruction methods outperform contrastive methods, unlike in natural-image classification. We argue that this difference follows from where discriminative signal lies in pixel space: for HTR, it is concentrated in high-variance directions and largely absent from low-variance ones. This predicts that objectives preserving high-variance pixel content will transfer best. We test six SSL methods from three families (pixel-grounded MIM, JEPA, and contrastive) under matched encoder, data, and evaluation protocols on six handwriting benchmarks across five languages. With full labels, pixel-groundrounded SSL achieves the lowest CER on every benchmark and both frozen probes, exposes per-position character information that other families recover only through the readout, and is the only family to benefit from pretraining on real handwriting. Pixel-grounded representations are also more label efficient. Across datasets, encoder alignment with the high-variance pixel subspace predicts CER within every method. With a pretrained LLM decoder, a frozen pixel-grounded encoder is competitive with fully fine-tuned supervised baselines; full fine-tuning achieves the lowest mean CER and ranks first or second on every benchmark. These results show that the value of pixel reconstruction depends on where discriminative signal lies in the input.
[CV-35] W2Rep: Learning Visual Representations by Watching the World Change
链接: https://arxiv.org/abs/2609.35464
作者: Wen Huang,Hang Guo,Jiarui Yang,Zheng Liu,Tao Dai,Shu-tao Xia
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Images capture the world at one moment, whereas video reveals how it changes. Image self-supervision learns spatial structure from a single moment, while video methods commonly learn temporal relationships inside a representation computed jointly from several frames. We ask whether watching a scene change can instead improve features available from one image without sacrificing the ability to represent video. We introduce W2Rep, a masked feature-prediction framework in which an independently encoded source image participates in prediction at the same or another moment. The predictor is conditioned on visible video context, the queried location, and the signed time interval between source and target. This gives the cross-frame objective two complementary roles: the image path learns features that remain useful across time, while the video path must gather evidence that is missing from the source image. Across model scales and downstream tasks, W2Rep improves frozen and fine-tuned recognition under our comparison protocol, while joint video encoding provides further gains over frame-wise aggregation. Controlled experiments show that these gains depend on directly updating the source-image features and on using both video context and temporal displacement. Overall, change across a video can supervise a visual encoder whose representations remain useful at either image or video granularity. Code is available at~\hrefthis https URLthis https URL.
[CV-36] How Far Are We from Removing the Visual Encoder? Scaling Laws for Encoder-Free Multimodal Pretraining
链接: https://arxiv.org/abs/2609.35457
作者: Lin Chen,Bolin Ni,Qi Yang,Lan Jiang,Kun Ding,Xiaoran Fan,Hower Yang,Ying Wang,Shiming Xiang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Most modern multimodal large language models (MLLMs) build on a pretrained visual encoder that provides a strong visual prior. Encoder-free MLLMs instead learn visual representations directly from raw pixels, offering a simple and unified architecture, but their scaling behavior has not been systematically characterized. To fill this gap, we compare scaling laws for encoder-free and encoder-based MLLMs and report three main findings: (1) Removing the visual encoder shifts the compute-optimal allocation for the multimodal objective toward larger models, while leaving that for text nearly unchanged. (2) The two architectures exhibit nearly overlapping loss–compute frontiers on the text objective, but diverge on the multimodal objective: encoder-free models underperform at small scales yet are predicted to catch up at around 10^22 FLOPs, well within practical pretraining budgets. (3) Without a visual encoder, the language model learns to take over its role via vision-specific adaptation: bidirectional interactions among visual tokens become increasingly beneficial as training compute grows, visual processing shifts toward earlier layers, and expert routing for visual tokens becomes more concentrated. Overall, our results indicate that the advantage of the visual prior provided by a pretrained encoder diminishes with scale, positioning encoder-free architectures as a promising direction for multimodal pretraining.
[CV-37] From internal representations to model improvement through prediction errors
链接: https://arxiv.org/abs/2609.35449
作者: Yushi Nakaya,Kenichi Higuchi,Shuichi Ishida
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 27 pages, 5 figures, 2 tables. Supplementary Information is provided as an ancillary file
Abstract:With limited annotation budgets, choosing which images to label determines how much a model improves. Data-selection methods that use features from a separately trained model, or scene descriptions written by vision-language models, have been successful, but those signals do not directly capture changes in the model being improved. The target model’s own internal features reflect what it has learned so far and change with retraining, making them a natural cue for choosing the next training data. However, feature rarity alone does not reveal the errors that matter for performance. Here we link internal features to prediction errors and their expected impact on performance and select images for labeling and retraining without using labels for candidate images. We evaluated the method with an object detector on two datasets and two pairs of random seeds. Adding internal features improved the identification of prediction errors in 15 of 16 conditions. When performance was averaged over successive labeling rounds, the method outperformed selection based only on feature rarity in all four evaluation settings and ranked among the top two of six methods. With other conditions held fixed, performance after retraining was again higher than with rarity-based selection, even though the latter collected more errors. With longer retraining, the proposed method ranked first among six methods. These results suggest that linking a model’s internal features to its errors and their effects on performance may help select training images that improve performance, thereby allowing the model’s current state to guide which images are labeled next.
[CV-38] DiMoP: Diffusion-Driven Motion Representation Learning With Frame-Level Pseudo-Classification for Skeleton-Based Action Recognition
链接: https://arxiv.org/abs/2609.35444
作者: Shanaka Ramesh Gunasekara,Wanqing Li,Nikalal Kaldera,Philip Ogunbona,Jack Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to IEEE TRANSACTIONS ON BIOMETRICS, BEHAVIOR, AND IDENTITY SCIENCE
Abstract:Robust skeleton-based action recognition requires representations that capture a wide spectrum of motions, from subtle to moderate and strong ones. Existing methods often focus on strong motions. This paper introduces DiMoP, a masking- and diffusion-driven motion representation learning method with frame-level pseudo-classification to explicitly learn the distribution of joint motions rather than regressing deterministic coordinates, as existing methods often do. By diffusing masked joints with progressive noise and denoising them conditioned on visible joints, DiMoP learns through controllable noising and denoising processes, enabling uniform learning of weak, moderate, and strong dynamics. To enable the masking-based generative diffusion learning with a discriminative capability, a pseudo-frame classifier is proposed that enforces the learning towards sequence-consistent and temporally coherent pseudo-labels without manual annotations. Together, these strategies provide a principled mechanism for joint generative and discriminative motion modeling. DiMoP achieves state-of-the-art performance across NTU RGB+D 60/120, and PKUMMD, including a 1.1 percentage point gain over prior works on NTU RGB+D 120 with the cross-subject protocol.
[CV-39] Revision Not Restart: Revisable Visual Plans for Closed-Loop World-Action Models
链接: https://arxiv.org/abs/2609.35439
作者: Pengyiang Liu,Junbo Niu,Wenhao Zheng,Xinchen Chen,Canyu Li,Zhongyue Shi,Jiahao Xie,Si Liu
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 27 pages, 4 figures. Project Page: this https URL
Abstract:World-action models use predicted visual futures to condition robot actions, yet execution feedback can invalidate parts of a prediction while leaving its task structure useful. We propose Revisable Temporal Planning (RTP), which maintains the visual future as a persistent action condition and revises it after feedback. Its central mechanism is a learned revision bridge: it resumes an intermediate state saved during visual generation and adapts its continuation to current observations. Visual and action supervision connect this revision to subsequent control. Time-aware history supplies observed evidence, and an adaptive policy selects retention, bridge revision, or fresh replanning from new noise before decoding the next action. On RoboMME and RMBench, RTP achieves task-averaged success rates of 48.6% and 84.8%, respectively. Matched comparisons support learned continuation; estimated checkpoint-source and action-prefix effects are positive but less precisely resolved. These results connect feedback-driven visual-plan revision to closed-loop task performance. Project Page: this https URL
[CV-40] Automated Species Identification in Camera Trap Images for Wildlife Conservation
链接: https://arxiv.org/abs/2609.35420
作者: Nowshin Amin,Nafisa Tabassum Oyshi,Tahmid Abrar Zidan,Miftaun Noor,Md. Abrar Rahman Shafin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 52 pages. this http URL . thesis, Department of Computer Science and Engineering, Brac University, June 2025
Abstract:Wildlife conservation involves protecting, preserving, and managing wildlife species and their habitats. With today’s rapid pace of human development, climate change, and other unsustainable practices, the need for wildlife conservation has heightened. Despite significant progress in species identification using deep-learning models, significant challenges still remain in effectively detecting small animals in low-contrast trap images due to limited feature extraction capabilities. This thesis presents a novel end-to-end framework integrating a self-attention mechanism to address these limitations. The proposed architecture involves a Swin-BiFPN backbone integrated in a Faster RCNN detection network, coupled with a visual semantic extraction module driven by the LLaVA v1.5 (13B) multimodal large language model. The detection framework, capable of extracting crucial features in challenging trap images, demonstrates consistently high results and robust generalization capabilities. Furthermore, the visual semantic extraction module provides zero-shot detection capability, as well as providing valuable insights and emergent cues of the animal’s behavior, further supporting the conservation effort. The MLLM evaluation was conducted using both traditional NLP metrics (precision, recall, F1, and SBERT similarity) and subjective scoring by LLM-based judges (GPT-4.1 and GROK 3.0), across five MLLMs, demonstrating the model’s strong performance in visual description generation. The proposed framework improves detection accuracy across low-contrast trap images and small animals while also demonstrating zero-shot detection capability leveraging the MLLM.
[CV-41] When Should the Count Change? Learning State Maintenance for Causal Video Counting
链接: https://arxiv.org/abs/2609.35416
作者: Pengyiang Liu,Dongyue Lyu,Junbo Niu,Zhongyue Shi,Jiahao Xie,Si Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 28 pages, 7 figures. Project Page: this https URL
Abstract:Continuous video counting requires distinguishing new observations from new objects or completed events. We introduce StaMina (State Maintenance), which learns to maintain counting state through state-conditioned updates. Recurrent visual context supports recognition; learned transitions maintain visibility, persistent identities, and completed-event records. A differentiable recurrence trains event transitions over legal paths constrained by count endpoints; visibility and association objectives train the object branch. A multi-source pipeline organizes 39.8K spatial queries and complementary event annotations into counting trajectories. On SVCBench, we evaluate counting adaptation with partial video overlap and held-out groups of linked annotations. Under prefix replay (Full) and persistent streaming (Stream), 4B and 8B models reach 41.9/36.4 and 44.9/38.2 Gaussian Precision Accuracy, respectively. The 8B model gains 10.9/3.2 points over Counting-SFT on the same queries. Matched-graph comparisons isolate phase conditioning and trajectory supervision, assessing training objectives alongside hard decisions. Online video benchmarks and count-conditioned decisions assess online understanding and task eligibility. Project Page: this https URL
[CV-42] Adaptive Safety Filtering for Frozen ACC Policies via Conformal Residual Calibration
链接: https://arxiv.org/abs/2609.35415
作者: Zhiruo Zhou,Rigaudiere Z. Li,Chen Xiwen,Yucheng Chen,Xiaojun Zhu,Houde Liu
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Systems and Control (eess.SY)
备注:
Abstract:Frozen adaptive cruise control (ACC) policies can violate constraints when deployment dynamics differ from their training conditions. We propose residual-aware conformal action filtering (RACF), which calibrates residuals of a fixed nominal predictor and converts their quantile into an operating margin for finite-model action projection. Completed transitions update margins and candidate selection without retraining the policy. In a registered comparison over 2,400 controller-trial units, Adaptive RACF achieves 94.3% episode safety, improving by 19.9 percentage points over the evaluated nominal CBF-QP baseline while reducing projection frequency from 8.11% to 6.63%. A controlled study isolates a 4.54-point improvement from residual-margin injection. In a separate matched-hardware evaluation, Adaptive reduces mean amortized rollout time by 21.2% relative to Robust CBF-QP, with 161/180 versus 170/180 safe episodes. We characterize conditions linking one-step residual coverage to constraint satisfaction and quantify the observed safety-computation trade-offs.
[CV-43] Spectral Super-Resolution using Spatial-Spectral Residual Operator Networks
链接: https://arxiv.org/abs/2609.35410
作者: Seokhyun Chin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: IEEE International Geoscience and Remote Sensing Symposium (IGARSS) 2026
Abstract:Spectral super-resolution of multispectral satellite images can enable high temporal- and spatial-resolution hyperspectral satellite imagery at a modest cost, significantly increasing the applicability of hyperspectral remote sensing. This task is inherently ill-posed, making it well-suited for deep learning-based methods. In this study, the spectral super-resolution task is framed as an operator learning problem, and SSRON is proposed as a Deep Operator Network that effectively learns function-to-function mappings from downsampled spectra to continuous spectra. The model is trained to super-resolve Sentinel-2A-like multispectral imagery to EMIT images. Compared to baseline models, SSRON achieves superior performance across all metrics. The model also demonstrates zero-shot spectral super-resolution capability by predicting bands unseen during training. Furthermore, its continuous-output formulation suggests the potential to estimate spectra at finer wavelength intervals than the native sensor. These results suggest the potential of SSRON and establishes operator learning as a promising direction for spectral super-resolution.
[CV-44] BiMoGen: Bidirectional Motion-Text Generation via Unified Masked Discrete Diffusion NEURIPS2026
链接: https://arxiv.org/abs/2609.35407
作者: Wanjiang Weng,Yongliang Wu,Xiaofeng Tan,Xingyu Zhu,Wenbo Zhu,Hongsong Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by NeurIPS 2026
Abstract:Text-to-motion generation and motion-to-text captioning are two fundamental tasks in human motion modeling, both grounded in the same underlying motion-text correspondence. Existing unified approaches mostly rely on autoregressive modeling, which imposes a fixed generation order and is therefore poorly suited to the bidirectional dependencies between language and motion, allowing early prediction errors to persist as fixed context and degrade both temporal coherence and cross-modal consistency. Masked discrete diffusion, which models sequences through iterative bidirectional prediction, offers a natural remedy. We therefore propose BiMoGen (Bidirectional Motion-text Generation), a unified masked discrete diffusion framework for bidirectional motion-text modeling. To stabilize training, we design Decoupled Uni- and Cross-Modal Training, in which masked pretraining first establishes cross-modal correspondence on paired motion-text sequences, after which supervised fine-tuning specializes the model for bidirectional generation. Masked diffusion nonetheless introduces its own source of error, as the model is trained on clean ground-truth context yet encounters self-generated and potentially erroneous context at inference, with errors committed under heavily masked states propagating through subsequent steps. We further introduce Generation-Aware Self-Correction that exposes the model to its own predictions during training and applies correction passes at early sampling steps to revise unreliably committed tokens. Extensive experiments on HumanML3D and KIT-ML demonstrate competitive performance on both tasks, validating the effectiveness of the proposed two-stage training and self-correction designs. The project page is available at this https URL.
[CV-45] Reduce Then Encode: Multiscale Volumetric Reduction for 2D Foundation Models in Brain MRI
链接: https://arxiv.org/abs/2609.35405
作者: Dexuan Ding,Yuankai Qi,Bogong Wang,Luping Zhou,Amin Beheshti
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Pretrained 2D foundation models offer a practical alternative to dedicated 3D pretraining for brain structural magnetic resonance imaging (sMRI), but their use on volumetric data requires bridging the mismatch between a 2D encoder and a 3D volume input. Existing methods typically encode slices independently and integrate their features afterwards. We introduce Multiscale Volumetric Reduction (MVR), a reduce-then-encode approach that compresses each anatomical view from (D) slices into (M D) complementary 2D components before foundation-model encoding. MVR combines an uncentered-PCA base component derived from the original through-plane intensities with residual detail components constructed from multiscale spatial descriptors. The reduction is estimated from the training volumes without diagnostic labels or gradient-based optimization and remains fixed thereafter. The resulting components are independently processed by a shared frozen 2D foundation model and concatenated for linear probing. Under this frozen-encoder setting, MVR achieves strong overall performance across ADNI, OASIS, and ABIDE relative to the evaluated 2D-to-3D adaptation methods and simple input-reduction baselines, while also generalizing strongly from ADNI to AIBL.
[CV-46] Rethinking Visual Token Compression for Video Large Language Models : A Simple Yet Strong Baseline
链接: https://arxiv.org/abs/2609.35394
作者: Xiao Zhang,Wang Zeng,Sheng Jin,Wentao Liu,Chen Qian,Shichao Kan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video Large Language Models (Video LLMs) have achieved remarkable progress in video understanding, but their inference efficiency is constrained by the large number of visual tokens produced by long videos. Recent video token compression methods increasingly introduce sophisticated strategies for token selection, pruning, and merging. This raises a fundamental question: how much of compression performance can be obtained by simply preserving the structure encoded in the visual representations? We investigate this question with SimpleCluster, a simple and training-free baseline that performs position-aware cross-frame clustering in the visual feature space and represents each cluster using the mean of its original visual features. Extensive experiments across four video understanding benchmarks and three representative Video LLMs show that SimpleCluster achieves competitive or superior performance over recent compression methods across a wide range of token retention ratios, with particularly strong robustness under extremely low retention rates (e.g., 1%). To understand this behavior, we analyze the feature space preserved by different compression methods in terms of local approximation fidelity and global coverage. The results show that stronger downstream performance is consistently associated with better preservation of the original visual feature distribution, especially its global coverage. These findings highlight feature-space preservation as an important consideration for video token compression under highly constrained token budgets. Our code is available at this https URL.
[CV-47] Ego-Forge: Text and Geometric-Attention Free Exo-to-Egocentric Video Generation
链接: https://arxiv.org/abs/2609.35368
作者: Mohammad Mahdi,Luc Van Gool,Danda Pani Paudel
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Exo-to-egocentric video generation aims to synthesize what a person sees from their own viewpoint given third-person footage and a target head trajectory. The task requires transferring appearance and semantics across large viewpoint changes while hallucinating content never observed by the exocentric camera. Existing approaches either impose additional input requirements, such as a ground-truth initial egocentric frame or multiple synchronized exocentric views, or remain limited to category-specific settings. EgoX is the first to address cross-activity and in-the-wild generalization, but requires a human-provided caption of the non-existent egocentric view at inference and introduces a computationally expensive geometry-guided attention bias that can propagate reconstruction errors and suppress textual and visual context. We therefore propose \textbfEgo-Forge, a caption-free and bias-free framework for exo-to-egocentric generation. It introduces \textitDynamic Captioning, which derives conditioning tokens directly from the model’s hidden states and adapts them to the diffusion timestep and network depth, replacing external text conditioning. By scaling training by an order of magnitude and using all available exocentric viewpoints, Ego-Forge learns cross-view correspondence implicitly and eliminates the need for geometry-guided attention, requiring only a lightweight depth prior. Ego-Forge achieves state-of-the-art performance on Ego-Exo4D, runs faster end-to-end, requires no external annotation at inference, and generalizes to in-the-wild scenes, including cases where over-reliance on geometry blocks appearance inference. Our model and source code will be made publicly available.
[CV-48] Generative Uncertainty as a Self-supervised Signal for Semantic Similarity Learning
链接: https://arxiv.org/abs/2609.35341
作者: Enrico Pallotta,Sina Raoufi,Lars Doorenbos,Gianni Franchi,Juergen Gall
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Evaluating semantic similarity between videos is a fundamental challenge in computer vision, essential for tasks ranging from out-of-distribution (OOD) detection to video retrieval. However, defining and labeling video similarity is notoriously difficult and expensive due to the complex spatio-temporal nature. In this paper, we propose a novel self-supervised approach that leverages generative uncertainty from text-to-video (T2V) diffusion models to learn semantic similarity without human annotations. Our method is based on the observation that T2V models produce consistent outputs for familiar concepts but exhibit high variance and uncertainty when prompted with specialized concepts. We utilize this behavior to identify stable semantic features within existing pretrained representations, such as VideoMAE and V-JEPA. Specifically, we learn a mask over these embeddings using purely generated data, encouraging the model to retain features that remain consistent across generations of general concepts while discarding those associated with generative noise or uncertainty. Experimental results across three key tasks demonstrate that our learned feature subspaces consistently outperform original pretrained features and baseline feature selection methods.
[CV-49] MCS: Tool-Grounded Multi-Agent Reasoning for Compositional Chemical Problem Solving
链接: https://arxiv.org/abs/2609.35336
作者: Shengqin Wang,Jie Jin,Yu Cheng,Yihang Chen,Weilin Luo,Yuan Xie,Zhizhong Zhang
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Despite the promise of Large Language Models (LLMs) in computational chemistry, rigorous combinatorial chemistry problems remain difficult because they require quantitatively constrained molecular modification, candidate validation, and systematic revision after failed attempts. Existing tool-augmented chemical agents demonstrate useful planning and tool use, but they rarely provide a unified loop for property-driven molecular optimization and workflow-level composition. To bridge this gap, we propose Tool-Grounded Multi-Agent Reasoning for Compositional Chemical Problem Solving (TMCS), a step-by-step multi-agent framework that formalizes chemical problem solving as an interpretable, tool-augmented workflow. At the task level, specialized agents leverage external tools, few-shot trajectory memory, and structured reflection to iteratively refine solutions. At the workflow level, TMCS chains generation, understanding, editing, description, and optimization into a closed-loop pipeline. Evaluations across multiple chemical tasks demonstrate that TMCS consistently enhances chemical reasoning across both open- and closed-source base models, achieving state-of-the-art performance.
[CV-50] RoGSW4RLD: Feed-Forward 4D Gaussian Lifting for Robot World Model Rollouts
链接: https://arxiv.org/abs/2609.35311
作者: Jin Hyun Kim,Min Young Kim,Soohwan Song,Daekyum Kim
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Action-conditioned video world models predict future robot interactions from multiple cameras, yet their outputs remain disparate video collections rather than a shared metric scene queryable across viewpoints and time. While existing 4D reconstruction methods offer a path to spatialize these predictions, independently reconstructing and merging each camera stream fails to enforce cross-view consistency. This limitation is particularly detrimental when combining moving robot-mounted cameras with fixed external views. To address this, we introduce RoGSW4RLD, a feed-forward framework that lifts synchronized multi-camera rollouts into a unified, time-queryable metric 4D Gaussian field. Rather than learning a separate geometric transition model, RoGSW4RLD directly reconstructs the visual future generated by existing world models. Its core innovation is a two-stage architecture: Stage 1 jointly forms the metric 4D field by fusing cross-view evidence with robot-specific articulated geometry and kinematics, while Stage 2 refines the field’s geometry and appearance while strictly preserving the initial temporal displacements. Evaluated on 256 held-out DROID episodes, RoGSW4RLD significantly outperforms camera-wise reconstruction with calibrated merging, improving novel-view PSNR by 2.15 dB, reducing depth AbsRel by 47%, and lowering robot displacement error by 61%. These robust gains extend to action-conditioned Cosmos 3 rollouts, demonstrating that predicted video futures can be successfully translated into consistent, spatially queryable 4D metric representations.
[CV-51] PIVOT: Pivot-Aware On Policy Self Distillation for Multi-Turn VLM Agents
链接: https://arxiv.org/abs/2609.35303
作者: Jiazhou Zhou,Hu Zhou,Yucheng Chen,Jinyuan Qu,Ying-Cong Chen,Lei Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 11 pages for the main paper, 20 pages for the supplementary
Abstract:Reinforcement learning with verifiable rewards (RLVR) via Group-Relative Policy Optimization (GRPO) is widely used for multi-turn VLM agent training, yet it suffers from zero-gradient silence on uniform failures and coarse episode-level credit assignment. While On-Policy Distillation (OPD) and On-Policy Self-Distillation (OPSD) mitigate sparse rewards using hindsight information, their underlying mechanisms remain poorly understood. Through controlled counterfactual rollback probes across five multi-turn VLM agent benchmarks, we reveal that performance gains in OPSD/OPD are largely driven by physical state rollback at the pivot step, defined as the first unrecoverable action without remaining step budget. However, physical state rollbacks are computationally prohibitive and infeasible in real-world environments. To bridge this gap, we present Pivot-Aware Internalized Visual On-Policy Training (PIVOT), an RL framework that internalizes pivot localization and state restoration directly into token-level parameter updates, eliminating environment rollbacks during RL training and additional skill hints at test time. PIVOT unifies three functional roles within a single architecture: a failure Analyzer non-invasively localizes the pivot step and diagnoses failure modes from visual trajectory collages and action logs; a detached Teacher re-scores failed tokens under this privileged diagnostic context; and a Student optimizes joint GRPO and confidence-gated OPD objectives. At test time, both Teacher and Analyzer branches are stripped. Evaluated on five multi-turn VLM agent tasks across cognitive grid puzzles, 3D embodied control and navigation, and generative reasoning, PIVOT achieves 0.90 overall accuracy on Qwen2.5-VL-3B (+8% over SFT+GRPO baseline and +5% over previous SOTA) and scales to 0.92 on Qwen3-VL-2B (+12% over SFT+GRPO baseline).
[CV-52] Beyond Saying Less: Fine-Grained Alignment for Informative and Faithful Vision-Language Models
链接: https://arxiv.org/abs/2609.35294
作者: Xingming Long,Jie Zhang,Yuecong Min,Shiguang Shan,Xilin Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Object hallucination remains a major challenge for large vision-language models. While off-policy preference optimization proves to be an effective solution, on-policy reinforcement learning provides a more promising direction as it directly targets a model’s current failure modes. However, we find that without fine-grained reward formulation and allocation, on-policy optimization often falls into an easy shortcut: reducing hallucinations merely by saying less—making fewer valid claims. To comprehensively resolve this, we propose a fine-grained alignment framework that couples dense reward signals at the data level with precise credit assignment at the algorithmic level. Specifically, we first construct the Dense Object Presence and Absence (DOPA) dataset to address sparse annotations that prevent valid object claims from being verified and rewarded. DOPA exhaustively annotates the deterministic presence and absence of every concept across an expanded vocabulary, significantly increasing the density of reliable reward signals during on-policy rollouts. Second, we propose Subsentence-level Credit Assignment for on-Policy Optimization (SCAPO) to prevent response-level shared advantages from allowing local hallucinations to compromise all other valid outputs within the same response. By assigning credit to each subsentence independently based on its object claims, SCAPO can precisely reinforce faithful generations and penalize hallucinations. Furthermore, we leverage the resulting faithful image descriptions as auxiliary context to transfer generative gains to discriminative tasks. Experiments demonstrate that our method produces highly informative, faithful descriptions in generative tasks while yielding clear performance gains on discriminative evaluation.
[CV-53] Scaffold Then Internalize: Representation Injection for Diffusion Transformers
链接: https://arxiv.org/abs/2609.35292
作者: Han Fu,Jiacheng Chen,Baoquan Zhao,Weidong Chen,Wei Liu,Qing Li,Xudong Mao
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Recent representation alignment (REPA) methods accelerate diffusion transformer training by aligning projections of the transformer’s hidden states with representations from pretrained visual encoders. In this work, we explore a reverse and complementary direction to REPA: rather than projecting diffusion representations into the encoder’s space, we inject encoder representations into the diffusion transformer, allowing them to actively participate in the denoising process. To this end, we introduce \textitREPresentation Injection (REPI), a training framework based on a scaffold-to-internalization strategy, in which projected encoder representations initially serve as a temporary scaffold and are then progressively internalized by the diffusion transformer. REPI outperforms REPA across a wide range of backbones and is highly complementary to it: combining the two yields substantial gains over either alone. Notably, with only 160K training steps, REPI + REPA matches vanilla SiT trained for 7M steps, a speedup of over 43.5\times . Code will be available at this https URL
[CV-54] Narrow Multimodal Fine-Tuning Can Induce Emergent Misalignment
链接: https://arxiv.org/abs/2609.35291
作者: Shunchang Liu,Lukas Fluri,Xin Chen,Francesco Croce
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Modern AI models are aligned through post-training to adapt them to downstream tasks. Recent work shows that fine-tuning language models on narrow tasks can induce emergent misalignment (EM), causing broadly harmful behaviors beyond the training task. However, EM has been studied almost entirely in text-only tasks, leaving its manifestation in multimodal models unclear. In this paper, we define and analyze EM in the context of vision-language models. We first induce EM via fine-tuning on narrow multimodal tasks targeting vulnerable code, careless household-object use, and conspiratorial interpretations of ordinary scenes. Across fifteen commercial and open-source models with different scales, we find that narrow multimodal fine-tuning can induce coherent and broadly misaligned behavior that transfers to unrelated tasks, including misaligned opinions, visual factual dishonesty, unsafe image generation, vulnerability to visual jailbreaks, and risky agentic actions. We further find that multimodal EM does not depend on the apparent harmfulness of training data but is sensitive to training-evaluation modality alignment. EM can arise under both supervised fine-tuning and preference optimization and can propagate through intermediate reasoning. Finally, we explore several mitigation strategies, including prompt inoculation, benign continued training, and activation-level steering, which can partially reduce EM. Overall, our findings suggest that multimodal EM reflects a behavioral shift rather than a general loss of capability, extending beyond text to the visual modality.
[CV-55] Domain-adaptive Zero-Shot Image Enhancement via Locality-Constrained Diffusion Guidance
链接: https://arxiv.org/abs/2609.35289
作者: Theresa Neubauer,Dimitrios Lenis,Astrid Berg,Maria Wimmer,Gaia Romana De Paolis,Philip Matthias Winter,David Major,Johannes Novotny,Ariharasudhan Muthusami,Katja Bühler
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted manuscript. The final version is published in Computers Graphics
Abstract:Denoising Diffusion Probabilistic Models have shown remarkable performance in unconditional image generation. In order to generate images with desired semantics, recent works have restricted the solution space by using guidance constraints in the diffusion sampling process. However, for image enhancement across different domains, these methods struggle to balance two main requirements: looking realistic in the target domain (photorealistic images) and preserving relevant features of the source domain, e.g., low-quality renderings or art paintings. Here, small local changes can alter the fidelity of the image completely, while large changes in other regions might be insignificant. We introduce LocDiff, a locality-constrained guidance method for image enhancement, which serves as a zero-shot extension to pre-trained diffusion models, ensuring the preservation of critical features during domain adaptation. In this way, we retain important local features, while allowing less critical regions to remain unconstrained and not interfere with the guidance process for relevant regions. We evaluate our method on two different domain-shift tasks: For art-to-photo translation, we apply the method in a fully zero-shot setting, preserving facial identity from paintings while generating photorealistic details. For enhancing low-quality fetal ultrasound renderings, we demonstrate zero-shot inference with auxiliary prior alignment. Here, the objective is to artificially add high-resolution characteristics and produce photorealistic ultrasound renderings, a target domain for which no ground truth distribution exists. Our experimental results demonstrate that LocDiff achieves favorable realism-faithfulness trade-offs compared to state-of-the-art methods, enabling controllable cross-domain enhancement. Comments: Accepted manuscript. The final version is published in Computers Graphics Subjects: Computer Vision and Pattern Recognition (cs.CV) ACMclasses: I.4.3 Cite as: arXiv:2609.35289 [cs.CV] (or arXiv:2609.35289v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.35289 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Journalreference: Computers Graphics, Volume 137, 2026, 104607 Related DOI: https://doi.org/10.1016/j.cag.2026.104607 Focus to learn more DOI(s) linking to related resources
[CV-56] λ-JEPA Spectral Anti-Collapse Regularization for Self-Supervised Learning
链接: https://arxiv.org/abs/2609.35288
作者: Berker Demirel,Clémentine Dominé,Valentino Maiorca,Marco Fumero,Marco Mondelli,Francesco Locatello
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Joint-embedding self-supervised learning typically combines an invariance objective across augmented views with additional mechanisms to prevent representational collapse. These objectives are often applied after a projection head, while downstream tasks use the backbone representation before the projector. We find that this mismatch does not necessarily prevent dimensional collapse in the backbone, which can retain low effective rank and potentially limit downstream transfer. To address this, we introduce SACReg, a spectral anti-collapse regularizer motivated by an analysis of \lambda -balance, which captures the relative scale of weight matrices across layers. In a two-layer linear network, we show that (i) \lambda -balance prevents collapse, and (ii) our regularizer applied to the backbone induces \lambda -balance. In the nonlinear case, this regularizer leads to anti-collapse as well and, in realistic architectures on ImageNet100, it empirically increases the representations’ ranks. We apply SACReg to JEPA and propose \lambda -JEPA, which improves over LeJEPA and VISReg on ImageNet-1k classification and in average linear-probe transfer performance across eight downstream image datasets. On video self-supervised learning, \lambda -JEPA improves over LeVJEPA and V-JEPA 2 on the Something-Something-v2 and Kinetics-400 benchmarks. Code is available at this https URL.
[CV-57] Evaluating Hierarchy-Aware Deep Learning for the Recognition of Tironian Notes ICDAR
链接: https://arxiv.org/abs/2609.35277
作者: Yule Kang,Thomas Gorges,Janne van der Loop,Franziska Marske,Nikolaus Weichselbaumer,Tino Licht,Vincent Christlein
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the 2026 ICDAR Workshop on Computational Paleography (IWCP). 25 pages, including supplementary material
Abstract:Tironian notes are generally regarded as the first Latin shorthand system and are notable for their large, fine-grained symbol inventory. Their high visual similarity and large class set make manual reading time-consuming, leaving manuscripts that contain Tironian notes inaccessible to many researchers. Automatic recognition is also challenging because models must distinguish subtle differences in stroke shape and sign structure while realistic training data remain scarce. However, standard flat classifiers do not explicitly use visual or structural relations between related signs. This paper investigates whether structural relationships between Tironian notes can support automatic recognition. We use the Supertextus Notarum Tironianarum (SNT) by Martin Hellmann, which provides idealized sign forms and a hierarchical organization of Tironian notes. We compare flat ResNet18, ConvNeXt, Shifted Window Transformer (Swin), and Vision Transformer (ViT) classifiers with Hierarchical Deep Convolutional Neural Network (HD-CNN)-style coarse-to-fine models and hierarchy-aware routing models based on visual class cleaning and similarity-based re-clustering. The models are evaluated on handwritten samples and manuscript-domain samples from Vergilius Turonensis, both with and without limited few-shot adaptation to the manuscript domain. The results show that the relative performance of flat and hierarchical models depends on adaptation. On Vergilius Turonensis, HD-CNN achieves the best non-adapted Top-1 result with 45.43%, while flat classification reaches the best Top-1 result after few-shot adaptation with 82.09%. Overall, the results indicate that hierarchical structure can support Tironian note recognition, especially under non-adapted conditions.
[CV-58] val-unlearn: Benchmarking unlearning in Text-to-Image Diffusion Models
链接: https://arxiv.org/abs/2609.35269
作者: Mansi,Nikhil Raghavan,Zixia Huang,Kai Sheng Ong,Ji Shen Lim,Brandon Siao Xiang Ling,Francesco Leofante
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The rising number of concept unlearning techniques for text-to-image (T2I) diffusion models has produced a fragmented evaluation landscape. Methods are assessed under heterogeneous experimental conditions making principled cross-method comparison difficult. We present eval-unlearn, an open-source Python library providing a unified, reproducible benchmarking framework for concept unlearning in T2I Diffusion models. eval-unlearn integrates twelve published unlearning techniques spanning fine-tuning, closed-form model editing, and inference-time intervention, alongside nine complementary evaluation metrics covering erasure efficacy, adversarial robustness, generative quality, and concept retention. Its plugin architecture lets third-party techniques and metrics self-register without modifying the core framework, and its streaming, batched pipeline supports efficient evaluation of both standard NSFW concepts and arbitrary general concepts. As a further contribution, we release a public leaderboard on HuggingFace along with an interactive tool for real-time evaluation of unlearning techniques. The leaderboard compares nudity concept erasure case study across all twelve techniques, exposing significant accuracy-quality trade-offs that are obscured by heterogeneous evaluation. eval-unlearn is released under the MIT license; the package, code, leaderboard, and documentation are all available at this https URL.
[CV-59] Spatial Grafting: Grounding 3D Features for Flow-Matching Robot Policies
链接: https://arxiv.org/abs/2609.35249
作者: Dingsheng Liu,Yangzheng Wu,Mahboubeh Asadi,Zhiyuan Li,Jinbang Huang,Yixin Xiao,Tongtong Cao,Yingxue Zhang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 4 figures, 9 tables
Abstract:Pretrained robot manipulation policies such as vision-language-action models (VLAs) or world-action models (WAMs) leave interaction-relevant metric geometry implicit. Recent breakthroughs in spatial reconstruction can supply the necessary geometry reliably, but their features describe local shape without stating where it lies with respect to the robot. How best to deliver these features to a pretrained policy remains unresolved. We propose Spatial Grafting, a versatile, lightweight spatial module that binds frozen reconstruction features to metric, robot-relative geometry. Spatial Grafting constructs metric-grounded spatial tokens and injects them into the flow-matching action expert through cross-attention, without modifying the host’s perceptual pathway, so the host retains the full benefit of its pretraining. We evaluate it more broadly than any geometry-aware policy we compare against: one graft architecture, with no per-host redesign, on two VLAs and two WAMs, across four simulation benchmarks that span short-horizon manipulation, visual robustness, clutter and long-horizon mobile manipulation, and on three real-robot platforms with single- and dual-arm configurations. On RoboTwin 2.0, a dual-arm manipulation benchmark, the graft improves every host across VLAs and WAMs. Grafted \pi_0.5 gains 11.3% and 15.6% on clean and randomized scenes, reaching 94.0% and 92.4%, above the strongest published 3D-conditioned policy, WAM4D (93.8% and 89.9%). The margin widens as the horizon lengthens: on tasks from BEHAVIOR-1K, a dual-arm mobile manipulation challenge scored by average task progress, it surpasses the 2025 challenge winner on five of six tasks,by up to 0.47 Q-score, and exceeds a map-conditioned spatial policy on average across the three tasks both report.
[CV-60] AnswerMap: Faithful Spatial Interpretability of VLMs from Answer Posteriors
链接: https://arxiv.org/abs/2609.35247
作者: Mohamed Eltahir,Fardows Adam,Duaa M. Tahir,Lama Alamoudi,Sana Ammar,Atheer A. Alboloshi,Jory Albluey,Tanveer Hussain,Naeemullah Khan
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:When a VLM answers a visual query, current interpretability tools rely on text rationales, which use a mismatched modality, or on internal read-outs, which originate too early to reflect the final output and require white-box access to the model. We introduce AnswerMap, a training-free, task-agnostic, black-box visual rationale constructed from the output head. The image is cut into K row and K column bands, each shown alone to the frozen model along with the query in the format of a yes/no relevance question. The outer product of the row and column ``yes’’ posteriors gives the query-conditioned spatial map. Crucially, by defining a fixed read-out R (e.g., expectation, maximum) on top of AnswerMap, we can derive continuous outputs like location natively. This bypasses the reliance on discrete text tokens for continuous-output tasks and guarantees an image-dependent answer by construction. However, a rationale can be confabulated, so we validate AnswerMap across four models and three query distributions with two tests: (a) agreement with the model’s own generated point and (b) deletion of the map’s region. The map lands where the model points (AUC 0.85 against 0.38 for attention), and deleting its region flips 53% of correct answers (against 19% for attention’s). Beyond establishing faithfulness, we demonstrate the map’s task-agnostic utility through three distinct read-outs: its maximum flags hallucinated objects without generation, its expectation localizes correctly when the model’s own pointing fails, and its top-mass region, fed back as a crop, fixes half of the model’s wrong answers. AnswerMap thus offers a new lens on VLM interpretability and, through its read-outs, a new output interface for visual tasks beyond text tokens.
[CV-61] DrawingsDreamer: A Unified Multi-View Engineering Drawings Generation Model
链接: https://arxiv.org/abs/2609.35242
作者: Shurui Liu,Weide Chen,Changwang Yi,Ancong Wu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Scalable Vector Graphics (SVG) are essential for modern industrial Computer-Aided Design (CAD). However, existing autoregressive SVG generation models are predominantly tailored for artistic creation and struggle to maintain the rigorous geometric fidelity and cross-view spatial alignment required for engineering drawings. To bridge this gap, we introduce \textbfDrawingsDreamer, a unified Large Language Model (LLM)-driven framework for multi-view vector-based engineering drawings generation. By formulating the generation of multi-view engineering drawings purely as a sequence modeling task, we eliminate the need of raster image encoders. We propose a Streamlined Representation utilizing hierarchical postfix tokenization, which guides the model to establish local geometric coordinates before assigning semantic boundaries. Optimized via a progressive task-aware curriculum schedule, \textbfDrawingsDreamer effectively transitions from localized structural repair to macroscopic generation in a unified model. Extensive experiments demonstrate that our unified model achieves strong performance in both geometric fidelity and syntactic accuracy across diverse conditional and unconditional generation tasks.
[CV-62] Beyond Selection: Token Parameterization for Extreme Visual Token Compression NEURIPS2026
链接: https://arxiv.org/abs/2609.35232
作者: Rui Zhong,Yu Li,Zheyu Yan,Cheng Zhuo
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026 (Spotlight). Code: this https URL
Abstract:Visual-token compression is effective for improving the efficiency of vision-language models, but under extreme compression budgets, token pruning can break visual grounding while learned resamplers increase parameter count, attention cost, and training complexity. We revisit compression through a token parameterization lens, separating (i) basis transformation and structured truncation (retained subspace/compressibility) from (ii) coordinate organization (optimization and cross-modal alignment). This view yields two coupled objectives, compressibility and learnability, which we formalize as unified functionals. Guided by these objectives, we design Braco, a lightweight four-step coder that combines transform-basis truncation, input-independent basis-coordinate embeddings, budget-dependent orthogonal re-parameterization, and learned spatial residual tokens from lightweight pooling. Experiments show that Braco forms the favorable empirical accuracy-efficiency frontier under 23\times – 64\times compression and remains competitive at 144\times , reaching 95.2% accuracy while reducing prefill FLOPs by 84.2%–86.7% relative to the uncompressed upper bound. Against prior methods, Braco matches or improves accuracy while achieving up to approximately 36% end-to-end speedup and using 16.6\times / 78.8\times lower compressor latency/FLOPs.
[CV-63] oken-Disentangled Latent Test-Time Scaling for Vision-Language Reasoning
链接: https://arxiv.org/abs/2609.35228
作者: Hao-Xuan Ma,Yihao Liu,Yutao Sun,Yanting Miao,Mengyu Zhou,YiCheng Xiao,Long Chen,Zhenguo Li,Han-Jia Ye,Xiaoxi Jiang,Guanjun Jiang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 20 pages, 4 figures
Abstract:Latent test-time scaling improves reasoning by refining hidden states during inference, but existing methods typically apply a single scalar reward to all editable latent tokens. For multimodal large language models, this global update ignores that generated tokens play different roles: some are sensitive to visual evidence, while others correspond to uncertain reasoning decisions. We present Token-Disentangled Latent Test-Time Scaling, an inference-time framework that makes latent refinement token-role-aware. Starting from an initial generated trajectory, we optimize a short hidden-state prefix while routing perception-side visual feedback to image-sensitive tokens and reasoning feedback to high-entropy tokens. Tokens selected by neither route are constrained by an anchor regularizer. Across both perception and reasoning benchmarks on Qwen2.5-VL-7B and InternVL3.5-8B, our method lifts macro accuracy over CoT by +2.57 and +1.51 respectively, and outperforms strong output-space test-time scaling baselines under matched decoded-candidate budgets. Code is available at this https URL.
[CV-64] Generative AI-Based Data Augmentation for Oral Lesion Classification: The PhotoMOCI Dataset and Benchmark
链接: https://arxiv.org/abs/2609.35226
作者: Marco Parola,Mario G.C.A. Cimino,Sabrina Senatore
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Early detection of oral cancer via photographic imaging presents a promising avenue for large-scale oral cavity screening. However, the development of robust deep learning models is frequently hampered by the scarcity of high-quality, annotated datasets. To address this limitation, a novel and well-curated resource, the Photographic Multi-purpose Oral Cancer Imaging (PhotoMOCI) dataset, is introduced for developing models across multiple diagnostic tasks in oral oncology. Then, a comprehensive benchmark study was conducted to investigate how various data augmentation strategies influence the performance of image classifiers. Our analysis spans different generative AI frameworks, evaluating the efficacy of traditional methods against advanced generative approaches, including Generative Adversarial Networks (GANs) and Diffusion Models (DMs). Additionally, we propose the Synthetic Image Filter (SIF), a mechanism to select specific samples based on two auxiliary models: Synthetic Proxy Classifier to ensure samples are representative of the target class and Synthetic Image Detector to verify they appear realistic, thereby selecting only the high-utility images that contribute to improving downstream performance. Across the evaluated datasets and classifiers, the best SIF-filtered setup improves accuracy over traditional augmentation in all cases, with gains of +1.73% and +2.35% on PhotoMOCI and +2.38% and +2.08% on KOCD for ResNet50 and ViT, respectively. Our findings reveal that while the direct application of generative data augmentation may yield performance drops, the integration of SIF, considering (i) how synthetic data looks real and (ii) how it reflects the discriminative features of the belonging class, provides a simple yet effective mechanism to filter out synthetic samples that confuse the classifier during training.
[CV-65] From UNI2-h to ConvNeXt-T: Lightweight Nuclei Instance Segmentation via Knowledge Distillation
链接: https://arxiv.org/abs/2609.35203
作者: Wenyan Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 2 figures, 4 tables. Submitted to IEEE International Symposium on Biomedical Imaging (ISBI 2027)
Abstract:Nuclei instance segmentation is a core task in digital pathology, yet high-accuracy models rely on large vision transformer (ViT) encoders whose inference speed cannot meet real-time clinical demands. We propose a lightweight scheme that distills the UNI2-h pathology foundation model into a ConvNeXt-Tiny student (Ours-T, 34.7M parameters, 1/20 of the teacher) via output-level knowledge distillation. Ours-T achieves an mPQ of 0.519 on PanNuke (98.8% of the teacher), a zero-shot bPQ of 0.668 on MoNuSeg, and an inference speed of 634.3 img/s, requiring only 0.045 s for full-resolution 1024^2 analysis (21.8x speedup). Experiments further show that multi-scale gated convolution (MALA) yields no gain under ViT encoders, and output-level distillation alone suffices for efficient knowledge transfer.
[CV-66] ReCAT: Remember Count and Time: Structured Recurrent Memory for Robot Manipulation
链接: https://arxiv.org/abs/2609.35200
作者: Pankhuri Vanjani,Mostafa Hatab,Can Mizrakli,Vaisakh Shaj,Zhuoyue Li,Moritz Reuss,Rudolf Lioutikov
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 9 pages, 3 figures
Abstract:Memory-dependent manipulation requires robots to make decisions using information that is no longer available to their current sensors, such as recalling an earlier visual cue, tracking task progress, counting repeated events, or estimating elapsed time. We present ReCAT, a language-conditioned policy with structured recurrent memory. An instruction-conditioned encoder forms features from the current observation. A recurrent memory integrates the observation stream through Mamba-2 layers and one causal attention layer. A flow-matching Transformer decoder reads the current and the historical representation through separate cross-attention in every block. ReCAT reaches 95.3% average success on LIBERO and 62.4% on RMBench, with the best or tied-best result on six of nine tasks. On three real-robot tasks probing spatial recall, event counting, and interval timing, the best ReCAT variant reaches 66.7% average success, against 8.3% for the strongest short-history baseline. Controlled comparisons within ReCAT show that the observation encoder and every-block memory conditioning are needed for this performance. They also show that update rules developed for efficient sequence modeling behave differently as robot memory: additive updates have the highest observed success on counting and timing, and delta-rule updates on spatial recall. Project website is at this https URL
[CV-67] CarveMix-RC: Addressing Rare-Class Imbalance Through Lesion-Aware Synthetic Augmentation for Brain Metastasis Segmentation
链接: https://arxiv.org/abs/2609.35195
作者: Md Shibly Sadique,Md Fayaz Bin Hossen,Michael L. Evans,Walia Farzana,Asfaqur Rahman,Ahmed Temtam,Khan M. Iftekharuddin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 14 pages, 2 figures, 2 tables
Abstract:Accurate segmentation of post-treatment brain metastases is essential for treatment planning, longitudinal disease monitoring, and quantitative assessment of therapeutic response. The BraTS-MET 2026 Task 1 challenge introduces a clinically relevant segmentation problem involving four anatomically distinct tumor subregions: non-enhancing tumor core (NETC), surrounding non-enhancing FLAIR hyperintensity (SNFH), enhancing tumor (ET), and the resection cavity (RC). Among these, RC segmentation is particularly challenging because of its low prevalence, heterogeneous postoperative appearance, and lesion-wise evaluation protocol, leading conventional segmentation networks to prioritize dominant tumor classes during optimization. The proposed nnU-Net-based framework explicitly addresses RC segmentation through four complementary components: (i) RC-weighted Dice and Cross-Entropy optimization to alleviate class imbalance, (ii) anatomically consistent cavity augmentation to increase the diversity of postoperative cavity appearances, (iii) a residual encoder architecture for enhanced multi-scale feature learning, and (iv) lesion-aware morphological post-processing to suppress false-positive cavity predictions while preserving anatomically plausible structures. The framework is evaluated on the BraTS-MET 2026 Task 1 online validation benchmark. Among the evaluated configurations, the ensemble model (Residual Encoder nnU-Net + nnU-Net + RC-aware CarveMix) achieves the best performance, with lesion-wise Dice scores of 0.732, 0.752, 0.708, and 0.575 and corresponding NSD scores of 0.794, 0.798, 0.727, and 0.474 for ET, TC, WT, and RC, respectively. These experimental results show that integrating RC-aware optimization, anatomically consistent augmentation, and lesion-aware post-processing provides an effective strategy for improving rare resection cavity segmentation in post-treatment brain metastases.
[CV-68] G3-LoRA: Organizing Reward-Weighted Video Data with Gradient-Guided Grouped LoRA
链接: https://arxiv.org/abs/2609.35189
作者: Jia Song(1),Wenhow Li(1),Lichen Bai(1),Bada Ye(2),Zeke Xie(1) ((1) The Hong Kong University of Science and Technology (Guangzhou), (2) Tencent)
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 22 pages, 6 figures
Abstract:Post-training foundation video models on heterogeneous reward-weighted data usually assume that all data categories induce compatible updates. This assumption is fragile when categories correspond to different skills, domains, or evaluation dimensions. We study this problem in text-to-video post-training, where VBench2.0 dimensions define data buckets and an external multimodal reward pipeline assigns sample weights. We propose G ^3 -LoRA (Gradient-Guided Grouped LoRA), a data organization procedure that probes category-level gradients induced by reward-weighted video samples, removes the shared global update direction, clusters categories by residual gradient compatibility, trains group-specific LoRA experts, and consolidates them into one adapter by weight merging followed by on-policy distillation from the experts. We motivate this procedure by viewing reward-weighted flow matching as velocity-field regression: incompatible reward dimensions may prefer different denoising directions in overlapping noisy latent regions, causing shared LoRA training to average capabilities. On Wan2.1-T2V-1.3B-Diffusers, the merged grouped adapter improves the matched VBench2.0 evaluation over the base model, a joint reward-weighted LoRA baseline, and random, semantic, and raw-gradient partitions trained with the same pipeline; an independent evaluator agrees, and on CogVideoX-2B grouping avoids the negative transfer of joint training. The gain is not uniform: merging compresses the largest specialist gains, distillation recovers part of this loss, and camera motion and several local-quality dimensions remain challenging. Together, these results suggest that gradient compatibility can serve as a practical diagnostic for organizing reward-weighted video post-training data.
[CV-69] meline-Bench: Evaluating Agents on Realistic Video-Editing Tasks from Raw Footage to Final Cut
链接: https://arxiv.org/abs/2609.35143
作者: Gunin Gupta,Nirmit Arora,Pavan Kalyan Tankala
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Multimedia (cs.MM)
备注: Preprint, under review. 9 pages main text, 27 pages total; 9 figures, 11 tables. Project page: this https URL
Abstract:AI agents increasingly carry out long-horizon professional work, but their evaluations rarely require a finished creative deliverable. To this end, we introduce Timeline-Bench, a benchmark of 56 real video-editing tasks, each asking an agent to turn raw production material into a finished video. Tasks range from selecting dialog takes and shaping interview footage into a story to cutting commercials from product shots, voiceovers and graphics. Every task provides a brief, source assets, a container and a set of tests. A task is resolved when the output passes every test. The tests check the delivery format, the content and the brief’s explicit requirements, and include a quality test calibrated on 2,582 blind judgments by 43 video editors. We evaluate 16 agents that pair frontier models with coding-agent harnesses such as Codex, Claude Code and OpenCode. The best, GPT-6 Astra in Codex with curated editorial guidance, resolves only 15 of the 56 tasks (26.8%), and the average agent resolves 14.0%. Human editors prefer the reference edit in 83.5% of judgments. Most unresolved runs (562 of 771) fail only the quality test: agents perceive footage through stills and transcripts and check their renders for defects, not craft. We release the tasks, verifier and per-run results at this https URL.
[CV-70] VideoPhysEdit: Physical Counterfactual Video Editing via Rigid-Body Physical Scene Reconstruction
链接: https://arxiv.org/abs/2609.35134
作者: Conghan Yue,Yuanjie Chen,Yue Han,Ya Gao,Yunyan Xiao,WeiYao Zhang,Zhineng Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video editing has advanced substantially in recent years, with methods increasingly accounting for the visual consequences of edits, such as changes to shadows and occlusions. However, the physical consequences of edits, including changes to subsequent motion and interactions, remain less explored. We formulate this problem as physical counterfactual video editing (PCVE), which aims to generate a counterfactual video depicting the resulting motion and interactions given a source video, a physical edit, and its execution frame. PCVE is challenging because it requires understanding scene physics and inferring the downstream motion and interactions induced by a physical intervention, while paired factual and counterfactual data and dedicated evaluation metrics are lacking. We introduce VideoPhysEdit, a new training-free pipeline for PCVE in rigid-body scenes. It makes physical reasoning explicit through a novel physical scene reconstruction method that recovers a scene reproducing the observed motion and interactions under simulation, enabling the pipeline to apply physical edits as interventions and use the resulting trajectories to guide counterfactual video generation. We further construct PCVE-RigidBench, a synthetic benchmark with paired source and counterfactual target videos and physical ground truth, and introduce the Physical Edit Score. VideoPhysEdit achieves substantially higher physical edit accuracy than open-source methods and commercial models while maintaining competitive visual fidelity. Its Physical Edit Score is 0.376, the only positive score among the compared methods. Qualitative comparisons on real videos further show that VideoPhysEdit applies to real-world scenes and better depicts the downstream motion and interactions induced by the edits than the compared methods. Code: this https URL
[CV-71] Style-Driven Data Synthesis and Degradation-Aware Enhancement for Ultrasound Image Restoration
链接: https://arxiv.org/abs/2609.35120
作者: Yu-Kai Wang,Chun-Xin Tan,Manh-Hung Nguyen,Ching-Chun Huang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Low-cost handheld ultrasound devices can be widely deployed compared to professional hospital ultrasound machines. However, their images suffer from compound degradation that can mislead clinical judgment. Motivated by this observation, mapping handheld low-quality (LQ) to hospital high-quality (HQ) images has been considered a valuable research question. Conventionally, the mapping requires pixel-aligned LQ-HQ pairs. This requirement is unsatisfactory in practical scenarios because real scans at different times are never pixel-aligned. This paper addresses the challenge with a two-stage framework. The first stage generates pixel-aligned LQ-HQ datasets, and the second stage trains an enhancement model that improves LQ images. The first stage trains a cycle-consistent style-transfer model on unaligned real LQ-HQ pairs to learn a HQ-to-LQ model. Then, the model transforms real HQ images into pixel-aligned LQ images. Based on the dataset generated by the first stage, the second stage uses the Dual Degradation-Guided (DDG) Low-Rank Adaptation (LoRA) method to fine-tune an LQ-to-HQ model based on aligned pairs. In this stage, the model is based on the well known PiSA-SR framework but inserts a degradation-conditioned correction matrix. Experimental results on the USenhance2023 dataset show that the FID metric is improved by 16.7% over the strongest baseline while other metrics indicate that our enhanced outputs are well aligned with the real HQ distribution. The source code of our method is available at this https URL.
[CV-72] Sol-H3: Recursive Self-Improvement for MiniMax-H3 Inference Acceleration on Sol-Engine across Cloud and Edge
链接: https://arxiv.org/abs/2609.35110
作者: Yitong Li,Jincheng Yu,Junsong Chen,Haopeng Li,Shuchen Xue,Haozhe Liu,Ping Luo,Song Han,Enze Xie
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Video diffusion models are rapidly scaling and exhibiting enhanced generation capabilities. Among these recent advancements, MiniMax-H3 stands out as a highly capable, production-level open-source model. However, its 33-billion parameters and multi-step iterative denoising process introduce substantial computational overhead. Consequently, their practical production is hindered by generation latency in the cloud deployment like NVIDIA-GB200, alongside strict memory limits that pose further challenges at the edge device like DGX-Spark. To address these diverse hardware bottlenecks from cloud to edge device, we present a full-stack inference pipeline that integrates efficient algorithmic design with optimized operator implementations. Algorithmically, we introduce a cross-resolution two-stage generation scheduler that exploits the step-wise nature of diffusion: early low-resolution steps rapidly establish the global layout, while later high-resolution steps focus refinements of local and perceptual details. These stages are connected by a learned latent-to-latent mapping module, completely eliminating the computationally expensive VAE decode-reencode cycle for resolution transferring cross different resolutions. For operator implementation, we deploy a Recursive Self-Improvement (RSI) loop that searches kernel fusions and memory layouts, evaluating latency together with numerical agreement. Together, these optimizations deliver up to 30x end-to-end speedup and 20% lower memory: a 5-second 1344x768 video with audio is generated 3.5x faster than real time on an 8xGB200 node, and in under a minute fully memory-resident on a single DGX Spark.
[CV-73] DF-CBM: Region-Aware Concept Bottleneck Models for Deepfake Detection ECCV2026
链接: https://arxiv.org/abs/2609.35096
作者: Georgios Tsoumplekas,Vazgken Vanian,Alexandros Doumanoglou,Panos K. Papadopoulos,Yannis Spyridis,Dimitrios Zarpalas,Vasileios Argyriou
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: ECCV 2026 (AI4MFDD 2026 workshop)
Abstract:Deepfake detection methods have become increasingly effective yet most provide limited insight into the evidence behind their predictions. However, in forensic settings users also need to know which manipulation cues support the decision and where they appear. Existing explainability methods only partially address this need since localization-based approaches lack semantic descriptions while language-based explanation methods are only weakly grounded in visual evidence. In this work, we propose DF-CBM, a region-aware concept bottleneck model for explainable deepfake detection. DF-CBM builds a compact vocabulary of manipulation-related concepts from textual artifact annotations and links each concept to plausible facial and boundary regions. It then predicts these concepts from visual features using a concept-specific masked attention mechanism guided by parsed facial masks and the final real/fake decision is made from the predicted concept bottleneck. Our experiments show that DF-CBM outperforms concept-based baselines in concept prediction and deepfake classification while remaining competitive with state-of-the-art black-box detectors. Finally, qualitative results and intervention analyses demonstrate that DF-CBM provides spatially grounded concept evidence and enables counterfactual explanations of how individual manipulation concepts influence the final prediction. Our code is available at: this https URL.
[CV-74] Advancing Video-Text Pretraining with Multi-View Captions
链接: https://arxiv.org/abs/2609.35090
作者: Fida M. Thoker,Renaud Vandeghen,Karen Sanchez,Marc Van Droogenbroeck,Bernard Ghanem
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video-text pretraining has achieved remarkable progress through the scaling of models and datasets, yet the quality of language supervision remains underexplored. Existing web-scale datasets often provide only a single sparse caption per video that fails to capture rich spatiotemporal semantics, while directly using captioning models can generate noisy descriptions. We propose a large-scale multimodal large language model-based supervision generation framework that improves supervision diversity, fidelity, and semantic coverage. Starting from 10 million videos, our approach generates multi-view captions (MVC) through complementary summary and detailed captions, reasoning-based refinement, and semantic positive caption generation. To effectively exploit supervision at different granularities, we further introduce a granularity-aware text representation with separate CLS tokens for summary and detailed views. We pretrain video-text models using the resulting supervision corpus and evaluate them across standard, fine-grained and detailed text-to-video retrieval benchmarks. Our approach consistently improves both zero-shot and fine-tuned performance while using smaller pretraining corpora than existing methods, demonstrating the importance of rich and complementary textual supervision for video-text pretraining. Project page: this https URL
[CV-75] RefineDrive: Reliable Failure-Guided Learning for Vision-Language-Action Driving
链接: https://arxiv.org/abs/2609.35078
作者: Zhe Sun,Ziyi Luo,Yehao Lu,Lei Zhou,Xi Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-Language-Action (VLA) models for autonomous driving rely heavily on successful expert demonstrations, leaving model-specific failures underexploited. Learning from these failures is hindered by unreliable diagnoses, poorly matched correction targets, and coarse rewards. We propose RefineDrive, a failure-guided post-training framework that learns from self-generated failures through targeted supervision and safety-aware reinforcement learning. Reliable Diagnosis derives structured, verifiable feedback on collisions and drivable-area violations directly from simulator states. Minimum-Correction Target Retrieval searches a clustered human trajectory bank for nearby corrections that satisfy hard-safety constraints in the current scene, prioritizing preservation of the failed prediction’s motion pattern. Conditioned on the driving context and failed trajectory, Correction SFT learns to generate the diagnosis followed by the retrieved correction as a training-only auxiliary task. We then apply GRPO with a Safety-Layered Reward that strictly prioritizes hard-safe trajectories, retains continuous safety feedback for both unsafe and hard-safe trajectories, and rewards driving progress only after hard safety is satisfied. At inference, the policy directly predicts trajectories from the driving context without an explicit diagnosis or repair stage. On NAVSIM v1, RefineDrive improves the 4B base SFT policy from 87.7 to 91.7 PDMS. Using the same checkpoint without additional training, RefineDrive achieves 89.4 EPDMS on the original NAVTEST scenes evaluated with NAVSIM v2 extended metrics. Controlled ablations support the benefits of structured diagnosis supervision, retrieved corrections, and safety-layered optimization for direct planning.
[CV-76] owards Generalizable 3D Anomaly Detection via Relational Inconsistency Modeling NEURIPS2026
链接: https://arxiv.org/abs/2609.35059
作者: KunHo Heo,SuYeon Kim,Hayoung Lee,Chanse Oh,MyeongAh Cho
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by NeurIPS 2026. Code: this https URL
Abstract:3D anomaly detection (3DAD) aims to identify defective regions in point cloud data, serving as a critical component in industrial inspection systems. Existing methods are normality-centered – learning the distribution of normal samples and treating deviations as anomalies – without explicitly modeling what constitutes a defect. This leads to ambiguous decision boundaries with increased false positives and negatives, particularly in unified and cross-domain settings where diverse normal distributions further blur the boundaries. We propose a relational inconsistency modeling framework that characterizes defects as violations of geometric consistency among neighboring structures. Our approach learns category-agnostic defect cues through pseudo-anomalies designed as controlled relational violations, instantiated by two key modules: Edge-aware Graph Refinement (EGR) for encoding geometric relationships among local regions, and Cluster-Deviation Modeling (CDM) for identifying regions that are relationally incompatible within their structural peer group. Extensive experiments on Anomaly-ShapeNet and Real3D-AD demonstrate consistent improvements over prior state-of-the-art methods in both in-domain and cross-domain settings, validating the effectiveness of learning an explicit, relation-based defect criterion for 3D anomaly detection. Project page: this https URL.
[CV-77] OPIS: An Input-Grounded Benchmark for Multi-Object Memory in Video World Models
链接: https://arxiv.org/abs/2609.35052
作者: Hao Wang,Tao Yu,Liuzhou Zhang,HeXin Wang,Haopeng Jin,Yuxuan Zhou,Xinming Wang,Hongzhu Yi,Xinye Li,Yuanlei Wang,Ping Nie,Yan Huang,Yuxuan Zhang,Pengfei Zhou,Yanyan Zou,Wei Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Video world models must preserve the visual state of the world over time, but existing evaluation protocols often rely on generated histories, video reference, or selected revisit viewpoints that can confound the assessment of a model’s true memory capability. To address this, we introduce OPIS, an input-grounded benchmark that strictly anchors the assessment to a fixed set of object instances from the initial observation for evaluating multi-object memory in video world models. The OPIS dataset comprises 500 cases across real-world, embodied-robotic, and game-world domains, providing dense object-level annotations for 12,672 rigid, articulated, and deformable instances. Our object-centric evaluator combines association and explicit visibility reasoning to hierarchically measure Object (O) Presence §, Identity (I), and Structure (S), utilizing static or dynamic evaluation tracks based on object kinematics. Across eight image-to-video or camera-conditioned world models, our proposed OPIS scores range from 48.65 to 56.01. As the reference inventory grows from less than 20 to more than 40 objects, the Presence, Identity, and Structure scores show an overall decline, with the average Identity score falling from 40.22 to 23.11. The results demonstrate that preserving the particular object instances in the input is considerably harder than generating plausible visual elements.
[CV-78] LEGAU: Learning Semantic Gaussian Priors for Scalable Category-level Pose Estimation
链接: https://arxiv.org/abs/2609.35046
作者: Hongli Xu,Zhaowei Lu,Junwen Huang,Jiaqi Hu,Peter KT Yu,Benjamin Busam,Federico Tombari,Slobodan ilic
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Category-level 6D pose estimation from a single RGB-D observation is inherently under-constrained, since partial visible geometry must be interpreted together with a canonical object structure before a stable pose can be determined. We present LEGAU, a unified framework that jointly predicts NOCS correspondence, object pose and size, and a canonical Semantic Gaussian Field. Rather than treating reconstruction as a detached auxiliary task, LEGAU uses the Gaussian field as a category-conditioned structural prior that participates in multimodal feature fusion and provides global guidance for local pose reasoning. Conditioned on a categorical text embedding, LEGAU processes RGB-D observations through a transformer-based fusion module that integrates visual, geometric, and category-level cues, decoding the NOCS map, pose and size information and the Gaussian-based object representation. Extensive experiments on synthetic and real-world benchmarks show that this coupled pose-shape formulation achieves strong performance in a single-model multi-category setting, with up to 22% on SOPE and competitive transfer to real-world data. These results highlight the benefit of jointly learning canonical correspondence, object shape, and pose alignment within a unified representation.
[CV-79] Mixed-Prior Decision Risk for Open-Set Recognition
链接: https://arxiv.org/abs/2609.35043
作者: L. A. Erlygin,P. D. Proskura,A. A. Zaytsev
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:In open-set recognition (OSR), a probe must either be identified as one of the known gallery classes or rejected as unknown, so three error types coexist: false acceptance, false rejection, and misidentification. An uncertainty score for selective recognition should rank probes by the risk of the decision the system has made. Bayesian gallery-aware models such as Holistic Uncertainty Estimation (HolUE) summarize the posterior over known and unknown classes by Kullback–Leibler (KL) divergence components and map them to an uncertainty score with a supervised nonlinear calibrator. We show that the KL summary is not generally monotone in decision risk: linear fusion of the KL components tuned on validation data yields negative filtering quality on several benchmarks. We propose MPRisk, a mixed-prior posterior decision-risk score that keeps the same Bayesian posterior but directly scores the error events associated with the selected decision: false-acceptance, misidentification, and false-rejection risks, plus a non-specificity penalty for rejections, enabled by modeling unknown identities as a continuous component. Four nonnegative weights tuned on a validation set suffice for ranking; no nonlinear supervised model is required. Across nine image, audio, and text benchmarks, MPRisk achieves the best or tied-best Prediction Rejection Ratio at every operating point on the image and audio benchmarks and on most text operating points, with bootstrap-confirmed gains over HolUE on five benchmarks (up to +0.19 PRR) at comparable or lower runtime.
[CV-80] JRDB-AVR: An Active Visual Reasoning Benchmark for Embodied Agents in Real-World Environments NEURIPS2026
链接: https://arxiv.org/abs/2609.35032
作者: Zhixi Cai,Fucai Ke,Sukai Huang,Maria Garcia de la Banda,Peter J. Stuckey,Gholamreza Haffari,Hamid Rezatofighi
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: NeurIPS 2026
Abstract:In complex embodied visual reasoning scenarios, an agent often has only a limited field of view, and the evidence needed to answer a question may be distributed across time, viewpoint, and interacting objects. A model may therefore give a plausible answer without ever observing the relevant object, time, or view that supports it. Current visual reasoning benchmarks largely evaluate passive observations and final answers, overlooking settings that require active reasoning and evidence acquisition. We introduce JRDB-AVR, a benchmark derived from existing real-world JRDB robotics data through a structured question-generation engine that turns this gap into an explicit evaluation: an embodied agentic system receives a visual reasoning question, requests bounded observations by timestamp and viewing angle, and is evaluated on both the final answer and the grounded visual evidence supporting it. The benchmark contains diverse questions over multiple real-world environments involving temporal search, viewpoint selection, and human-oriented compositional reasoning. We also introduce JRDB-AVR-Agent, a reference active reasoning agentic method that maintains an explicit observation-grounded graph-based world model and answers through solving. Experiments reveal a substantial gap between answer accuracy and evidence accuracy in current baselines, showing that current VLMs can produce unsupported correct answers and that active evidence-aware evaluation is necessary for embodied visual reasoning. Code and benchmark are available at this https URL.
[CV-81] Proxy2World: Learning to Generate Worlds From Lightweight Proxies without Seeing Them
链接: https://arxiv.org/abs/2609.35023
作者: Hongli Xu,Weilong Yan,Anbang Wang,Chunyu Zou,Siyu Hong,Jingwei Huang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Lightweight scene proxies let creators control scene layout and motion while leaving room for imagination in appearance, lighting, and visual effects. However, a suitable proxy is not uniquely defined, making paired proxy-video data difficult to construct automatically at scale. We present Proxy2World, a controllable world model that learns these complementary capabilities from ordinary posed RGBD videos, without training on authored proxy-video pairs. The model jointly learns depth-conditioned RGB generation and joint RGBD generation through cross-modal flow matching. Learning both tasks enables proxy-camera hybrid denoising at inference to follow the proxy structure while producing natural, detailed visuals. We further introduce ProxyBench to evaluate this capability across a diverse set of scenes, camera trajectories, and subject motions. Experiments on ProxyBench show that Proxy2World achieves a better balance between structural adherence and visual quality than camera-controlled and geometry-conditioned methods, supported by quantitative metrics, VLM assessments, human evaluations and diverse qualitative results.
[CV-82] Detection of Adversarial Attacks on Super-Resolvers Using Spectral Features
链接: https://arxiv.org/abs/2609.35022
作者: Emma J. Reid,Haley Duba-Sullivan,Tony G. Allen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: To be published in the 2026 Asilomar Conference on Signals, Systems, and Computers
Abstract:The integration of deep learning models into image preprocessing pipelines such as super-resolution introduces a largely unexplored attack vector for adversaries targeting downstream tasks. To ensure trustworthiness of critical imaging pipelines, we must be able to detect adversarial behavior within preprocessing models. In this paper, we propose a spectral-based detection method for identifying adversarial attacks embedded in super-resolution model weights. More specifically, we use the radially-averaged power spectral density as a discriminative feature to train an extreme gradient boosting (XGBoost) detector, demonstrating detectability of model-level threats in super-resolution networks. We further benchmark our detector against magnitude- and phase-based Fourier spectrum detectors, evaluating each method across a range of training and cross-architecture scenarios. Our proposed detector out-performs the comparison detectors in most of these scenarios and indicates that high-frequency features are most informative for detecting AdvSR attacks across SR architectures.
[CV-83] Verifying the Linear Representation Hypothesis: How Interpretable Are Vision SAEs?
链接: https://arxiv.org/abs/2609.35020
作者: Teodor Chiaburu,Franz Motzkus,Frank Haußer,Felix Bießmann
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 28 pages, 8 figures, 5 tables, preprint under review
Abstract:Vision Sparse Autoencoders (SAEs) have become a popular tool in Mechanistic Interpretability due to their presumed ability to disentangle complex features learned by a model into monosemantic concepts. Despite their growing popularity, evaluating their interpretability remains an active topic of research. The bedrock motivating the adoption of SAEs is the Linear Representation Hypothesis (LRH), which claims that polysemantic features can be projected onto a (near) orthogonal basis of sparse, human-understandable representations. Yet, most current frameworks evaluate proxies such as the sparsity of SAE features or the coherence of the inferred dictionary, implicitly assuming that these reflect alignment with human perception. In this paper, we provide empirical evidence that measuring the interpretability of SAE concepts is more difficult than these proxies suggest. To this end, we adapt the Autointerpretability Score (AIS) - previously shown to align with human judgments in Natural Language Processing - to vision tasks and validate our approach in a dedicated user study. We evaluate SAE concept quality using both standard metrics and our adapted AIS. We find that established interpretability metrics for SAEs correlate neither with one another nor with AIS, indicating that no single reference-free metric, whether grounded in the LRH or not, is sufficient for verifying the interpretability of vision SAEs. We argue these findings support recent calls for more verifiable, ground-truth-anchored design and evaluation of explanation methods.
[CV-84] Still There No Longer Seen: Exposing Compression-Induced Risk in Large Vision-Language Models
链接: https://arxiv.org/abs/2609.35002
作者: Qiankun Li,Yuechen Zhang,Bowen Chen,Shilinlu Yan,Zhenhong Zhou,Kun Wang,Li Sun
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: 29 pages, 11 figures, 13 tables
Abstract:Visual token compression reduces the inference cost of Large Vision-Language Models (LVLMs). However, aggregate robustness measures do not reveal whether a particular adversarial failure is induced by compression or inherited from the underlying model. We define a compression-specific failure (CSF) as an adversarial input that remains correct under full-token inference but fails after compression, casting compression-induced risk as a paired failure attribution problem. Within a controlled diagnostic cohort, counterfactuals show that retained-set allocation causally changes compressed correctness and reveal a negative association between recovery and representation drift in displaced evidence. Motivated by these findings, we propose CIRA, a Compression-Induced Risk Attack for Large Vision-Language Models. Under a vision-encoder white-box setting, CIRA optimizes image perturbations through encoder-side objectives that manipulate token priorities across candidate compression budgets while preserving displaced evidence. CIRA uses no downstream questions or labels and requires no access to the language model, deployed compressor, or exact compression budget. Across 12 dataset-compressor settings evaluated at four budgets, CIRA achieves a mean CSFR of 20.35% while limiting full-token attack success to 6.92%, with similar behavior on additional LVLM families. A cross-view selection-stabilization defense substantially suppresses CIRA, although Adaptive CIRA partially restores its effectiveness. These results show that compression-specific failures persist under restricted access and support paired evaluation of full-token and compressed inference for attributing risk to visual-token compression.
[CV-85] ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuning
链接: https://arxiv.org/abs/2609.34982
作者: Di Zhu,Ziheng Yan,Fang Wan
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Vision-Language-Action (VLA) models have shown great promise for robotic manipulation by mapping multi-modal semantics to physical actions. However, this mapping inherently struggles to align these coarse-grained semantics with fine-grained temporal execution. It leaves VLA models with limited generalization and insufficient robustness in cluttered environments. To overcome this issue, we propose ActionUNet, an efficient multi-scale fine-tuning framework that enhances pre-trained VLA models with minimal computational cost. ActionUNet first constructs a lightweight temporal U-Net within the temporal-aligned action feature space to fuse hierarchical structural priors, effectively bridging the scale gap between semantics and temporal executions. Recognizing that multi-scale modeling can disrupt microscopic temporal continuity and cause mechanical oscillations, ActionUNet then employs a conditional SIREN as a continuous action decoder. Equipped with explicit second-order smoothness constraints, this decoder guarantees temporal continuity and reduces high-frequency motion jitter. By smoothing temporal discontinuities from multi-scale fusion, this continuous formulation reduces mechanical execution failures while preserving the base VLA model’s generalization and manipulation robustness. Extensive experiments on RoboTwin 2.0 and LIBERO-Plus benchmarks, together with real-world hard evaluations, demonstrate that ActionUNet significantly improves \pi0.5 success rates by absolute 9.8%, 6.1%, and 11.4%, respectively, while also generalizing to the regression-based OpenVLA-OFT backbone, highlighting its effectiveness and efficiency as a fine-tuning strategy. Code and implementation details are available at this https URL.
[CV-86] What Makes World Action Models Generalize? An Empirical Study of Test-Time Future Modeling
链接: https://arxiv.org/abs/2609.34981
作者: Renping Zhou,Zanlin Ni,Zihao Fan,Guohao Fu,Zeyu Liu,Hao Shi,Jie Zhang,Chi Bene Chen,Yang Yue,Xueyang Fu,Gao Huang
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:World action models (WAMs) predict the future alongside actions during \emphtraining. Due to the heavy computation cost of video denoising, whether the future must still be generated during \emphinference is disputed: Explicit WAMs denoise it into clean frames along with every action chunk, whereas Latent WAMs discard it entirely for acceleration. We find that latent WAMs, despite matching explicit ones on in-distribution tasks, fail to retain the generalization benefits that originally motivated WAMs. To demonstrate this, we evaluate generalization along three axes: \emphenvironmental perturbation, \emphdata efficiency, and \emphtask generalization. Controlled comparisons with a matched backbone, training data, and budget reveal consistent degradation across all three axes when the action expert no longer conditions on future representations. Further analysis shows that the gap arises almost entirely from the first denoising step: the benefit comes from \emphpreparing the future, not \emphgenerating it. We therefore propose \textbfSimple-WAM, which simplifies future modeling into a single forward pass of fully noised video tokens and adapts the training-time noise schedule to this inference behavior. Across simulation and real-world tasks, Simple-WAM achieves the best of both worlds, leading explicit WAMs in generalization performance with efficiency comparable to Latent WAMs. Project Page: \hrefthis https URL\textcolorpanton\textttthis https URL
[CV-87] SPIDER: Multi-Layer Semantic Token Pruning and Adaptive Sub-Layer Skipping in Multimodal Large Language Models
链接: https://arxiv.org/abs/2609.34977
作者: Tianxiang Chen,Zhentao Tan,Zi Ye,Yue Wu,Xiaobing Tu,Jinkui Ren,Xiantao Zhang,Tao Gong,Qi Chu,Nenghai Yu,Xipeng Qiu,Jieping Ye
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Multimodal Large Language Models face significant efficiency challenges that stem from two distinct yet coupled sources: data redundancy and computational redundancy. While most methods focus on data redundancy by pruning visual tokens from the output of the visual encoder or computing redundancy in LLM decoders using blockwise importance, the finer-grained inter-layer representation shifts and the distribution differences within the layers themselves have not been fully explored. In this work, we comprehensively investigate this dual-level inefficiency. We posit that intermediate layer tokens from vision encoders should be considered for effective visual token pruning, as semantic focus shifts across layers, with middle-layer tokens capturing more detailed object-centric information that deeper layers may abstract away. Furthermore, we reveal the differential contributions of Attention and FFNs across distinct LLM decoder layers. Building upon these discoveries, we propose \textbfSPIDER, a training-free framework that integrates multi-layer \underline\textbfSemantic visual token \underline\textbfPrun\underline\textbfIng with an a\underline\textbfDaptive sub-lay\underline\textbfER skipping mechanism. Experimental evaluations demonstrate that SPIDER consistently maintains strong performance across various MLLM architectures and reduction ratios. For instance, on LLaVA-NeXT-7B, SPIDER reduces FLOPs by 79% while maintaining 96 % of the baseline performance.
[CV-88] Inspector: Conversational and Lightweight Analyzer of Analog Circuit Layouts Using LLM and CNNs
链接: https://arxiv.org/abs/2609.34976
作者: Abril Cano Castro,Giuseppe Chiari,Michele Piccoli,Federico Viola,Davide Zoni
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 4 pages, 5 figures, 5 tables, to be published in ICLAD 2026
Abstract:The integration of artificial intelligence into computer-aided design frameworks has sparked a shift in the design of analog integrated circuits (ICs), transitioning the field from using manual and algorithmic-based solutions to adopting automated and intelligent paradigms. In this scenario, the GDSII file represents the industry-standard database containing the ultimate and most accurate source of information of the analog circuit, encapsulating the complex physical geometries and parasitic realities that define tape out performance. This paper proposes a novel framework that combines fine-tuned LLMs and CNNs to analyze GDSII files of analog circuits, enabling a conversational interface between the tool and the designers. Experimental results using thousands of analog designs across four realistic tasks demonstrate that the proposed solution outperforms state-of-the-art general-purpose massive VLMs by a significant margin (up to 81%), thus providing a lightweight solution to the problem of GDSII analysis.
[CV-89] Just MLPs: Efficient Visual State Reconstruction for Multimodal Language Models
链接: https://arxiv.org/abs/2609.34972
作者: Jingdi lei,Junxian Li,Di Zhang,Zhanqiu Zhang,Yiwen Guo,Soujanya Poria
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 21 pages, 5 figures
Abstract:Long visual token sequences often account for a substantial fraction of the computational overhead in multimodal large language models~(MLLMs). Existing approaches reduce this cost by pruning redundant visual tokens, but permanently discard visual evidence that may become useful in subsequent layers. We instead ask whether all visual tokens can be preserved while reducing the cost of repeatedly evolving the representations through the Transformer. To answer this question, we perform low-rank interventions on visual-to-text information flow. We find that, after visual-to-text attention is blocked, restoring only a few directions recovers most of the lost accuracy, suggesting the relevant visual influence is concentrated in a low-dimensional subspace. We further observe strong predictability in layer-specific visual states: lightweight MLPs approximate them with high cosine similarity and low reconstruction error. Motivated by these findings, we propose \delta -Vision, which replaces repeated Transformer evolution of visual tokens with lightweight low-rank adapters that construct layer-wise visual memories while preserving all visual tokens for text retrieval. Across image and video benchmarks, \delta -Vision achieves higher accuracy than visual token pruning baselines at comparable or lower computation, while delivering competitive inference efficiency without discarding visual tokens.
[CV-90] Adjoint Guidance Flow: Amortized Critic Guidance for VLA Policies
链接: https://arxiv.org/abs/2609.34944
作者: Jeongsol Kim,Youngjun Jun,Kyumin Choi,Youngmin Kim,Seonghyun Jin,Sunwoo Park,Jangho Park,Kwanyoung Kim,Jong Chul Ye
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Flow-based Vision-Language-Action (VLA) policies are typically trained by behavior cloning and thus do not explicitly optimize long-term task return. Critic guidance steers generation toward higher-value actions, but existing methods differentiate the critic through a one-step surrogate of the sampler and back-propagate a critic ensemble at every flow step. In contrast, here we propose Adjoint Guidance Flow (AGF), which amortizes trajectory-aware critic guidance into a lightweight guidance network while preserving the pretrained VLA policy. Specifically, we formulate critic-guided flow generation as a deterministic optimal control problem, whose optimal guidance is a costate that carries the terminal critic gradient back through the remaining flow, and regress the guidance network onto this costate while keeping both the VLA and critic frozen. This design provides favorable memory and throughput scaling during training, and inference needs one guidance-network forward pass per step, without the critic ensemble, back-propagation, or adjoint computation. Across LIBERO, RoboCasa, and LIBERO-Pro, AGF consistently improves pretrained VLAs, remains competitive with critic-guidance and policy-fine-tuning baselines, and is the most robust method when a single guidance strength is deployed across tasks. Compared with QGF, AGF runs 3.6\times faster per guidance step with 7.0\times fewer parameters, with comparable and even better performance, showing that critic guidance can be trajectory-aware and lightweight.
[CV-91] Resolution as a First-Class Decision: Task-Conditioned Routing for Efficient Multimodal Large Language Models
链接: https://arxiv.org/abs/2609.34942
作者: Zhiqiang Xia,Yang Li,Xinyuan Zhang,Yuchen Liu,Haoyu Lu,Jiaming Xu,Runyu Shi,Ying Huang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages including references and appendix
Abstract:The inference efficiency of Multimodal Large Language Models (MLLMs) is severely constrained by massive visual token sequences induced by high-resolution inputs, with computational cost scaling quadratically. Existing approaches primarily focus on downstream token compression, while overlooking a fundamental upstream inefficiency: input resolution is treated as a static, task-agnostic hyperparameter. We propose Task-Conditioned Resolution Routing (TCRR), which formulates visual compression as a task-conditioned decision and employs a lightweight cross-modal router that conditions backbone visual representations on textual semantics via feature-wise modulation and cross-attention to predict the minimal sufficient compression level per query. To support this, we curate a dataset of 500k samples across 12 task categories, labeled via a teacher-oracle pipeline to approximate Pareto-optimal compression scales. Extensive experiments across diverse architectures show that TCRR achieves a superior efficiency frontier, specifically reducing visual FLOPs by 40.9% and latency by 53.7% on Qwen3-VL-8B while preserving competitive performance. Further analysis of scaling behavior confirms that dynamically routing visual compression enables optimal resource allocation without modifying the MLLM backbone.
[CV-92] aoTex: Boosting Texture Detail Fidelity for Native 3D Material Generation
链接: https://arxiv.org/abs/2609.34934
作者: Xiuchao Wu,Shuichang Lai,Jiangjing Lyu,Chengfei Lyu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent 3D generation models can produce accurate geometries while still struggling to reconstruct detailed textures. We propose a diffusion-based native 3D material generation model TaoTex, which faithfully recovers intricate textures through tailored strategies and improvements. First, we develop a data construction agent to create high-frequency textured 3D assets to bridge the data gap in public datasets. Training with these data significantly enhances the ability of TaoTex to recover challenging details such as text and patterns. Second, we design a multi-level feature fusion (MLFF) module to adaptively integrate local and global features of the conditional input, providing more complete texture cues for the diffusion model and thereby enhancing reconstruction fidelity. To alleviate VAE reconstruction errors, we adopt a latent-to-pixel space loss transition, further improving the pixel-level details and generation quality. Finally, we scale TaoTex to multi-view inputs by incorporating learnable viewpoint embeddings, achieving accurate and consistent material reconstruction across views. Extensive experiments demonstrate that our method significantly outperforms existing approaches in preserving texture details in both single- and multi-view settings.
[CV-93] Dont Throw Away the Tail: Action Upcycling for Policy Acceleration
链接: https://arxiv.org/abs/2609.34911
作者: Taesung Kwon,Jangho Park,Sunwoo Park,Youngmin Kim,Seonghyun Jin,Youngjun Jun,Kyumin Choi,Jong Chul Ye
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Project page: this https URL
Abstract:Modern robot policies predict a chunk of future actions from a single observation, execute only a prefix, and discard the rest before replanning. Choosing the length of this prefix, the execution horizon, poses a trade-off between reactivity and efficiency. A short horizon keeps the policy reactive to the environment, but requires frequent policy calls. Recent test-time methods adaptively select the horizon for each chunk, but they either read model internals, where the signal must be chosen for each architecture, or draw extra samples, which adds cost. We propose Action Upcycling, a training-free algorithm that reuses actions the policy would otherwise discard, without accessing model internals or drawing extra samples. We find that discarded actions stay close to their replanned versions as long as the action velocity remains smooth. Action Upcycling therefore extends the execution horizon up to the point where the velocity begins to fluctuate. Extensive experiments on simulated and real-world manipulation tasks show that Action Upcycling reduces policy calls by 1.2–1.7 \times with no loss in success rate, across multiple Vision-Language-Action Models (VLAs) and even a World Action Model (WAM). It applies to any chunked policy at negligible cost and is orthogonal to other policy acceleration methods such as few-step sampling and streaming action decoding, opening a new axis for policy acceleration.
[CV-94] ReSight-SMC: Two-Stage Power Sampling via Island SMC with Visual Scouts
链接: https://arxiv.org/abs/2609.34905
作者: Yaowen Zhang,Xiangyu Qiu,Junyi Hu,Zhi Lu,Wenwen Tian,Aoqin Wang,Junhai Luo,Zhenming Peng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Power sampling has emerged as a training-free approach to LLM reasoning, eliciting capabilities comparable to reinforcement learning by sharpening the model distribution over complete responses. Despite this success, power sampling remains underexplored in large vision-language models (LVLMs). We transfer Power-SMC to LVLM decoding by defining a sequence-power target conditioned on both the image and the prompt. This direct transfer provides a strong training-free baseline, but leaves two aspects of finite-particle multimodal inference unaddressed. At the particle level, global resampling can collapse genealogies, while particle-based power sampling does not diversify trajectories through distinct visual cues in multimodal decoding, limiting exploration under a finite particle budget. At the answer level, sequence-level sharpening makes distinct reasoning trajectories compete even when they support the same answer. We introduce ReSight-SMC, a verifier-free two-stage power sampler for LVLM inference. Its first stage uses ancestry-isolated SMC islands to preserve independent trajectory families and routes a bounded set of prefix-conditioned visual scouts to prefix-relevant image regions while discouraging redundant overlap. Each scout temporarily increases attention to the image tokens and emphasizes its routed region. Exact importance correction preserves the base LVLM sequence-power target. The second stage aggregates terminal importance mass by canonical answer, powers the answer marginal, and samples an answer together with a supporting trajectory. Across four LVLM backbones and five benchmarks, ReSight-SMC achieves stronger aggregate performance than Power-SMC over both the reasoning and perception benchmark groups. Without post-training, it remains competitive in aggregate with backbone-matched models trained using reinforcement learning.
[CV-95] FILIGREE3D: Scaling Sparse Latent Flow Matching for Ultra-High-Resolution Image-to-3D Generation
链接: https://arxiv.org/abs/2609.34900
作者: Hongjie Li,Xinran Yang,Xiuchao Wu,Jiangjing Lyu,Chengfei Lv
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Scaling image-to-3D generation to ultra-high resolutions requires controlling rapidly growing computational costs without sacrificing fine geometric detail. We present \textbfFiligree3D, a sparse latent flow-matching framework that generates 3D geometry from a single image at voxel resolutions up to 2048^3 , with straightforward extensibility to 4096^3 . To make training tractable, we introduce Structure-Aware Sparse Scaling, which combines spatial bounding with alternating local-global attention to constrain token growth while preserving both fine-scale details and long-range structural context. To enhance detail reconstruction, we curate training samples based on their high-resolution geometric gains and inject multi-scale image features into a sparse 3D DiT, effectively coupling structural semantics with fine-grained visual cues. Furthermore, a visibility-aware voxel regularization strategy improves robustness against sparse perturbations and facilitates the completion of unobserved geometry. Under our default configuration, Filigree3D maintains peak GPU memory consumption within practical limits for contemporary hardware, enabling the generation of highly intricate 3D geometry in approximately one minute. Extensive experiments demonstrate that our method yields substantial improvements in overall geometric fidelity and fine-detail preservation compared to existing baselines, validating practical, detail-preserving 3D generation at unprecedented resolutions.
[CV-96] Role-Guided MOE for Encoder-Level Pathology Representation Learning in WSI Classification
链接: https://arxiv.org/abs/2609.34897
作者: Xinyu Ma,Xing Yang,Hongtao Jin,Guoquan Zhang,Shijie Zhang,Yu Zhang,Xitong Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Whole slide image classification is a fundamental task in computational pathology, where patch representation quality directly affects downstream aggregation and slide-level discriminability. Pathology foundation models are widely adopted as frozen feature extractors for WSI classification; however, their fixed encoders may produce representations insufficiently adapted to target-specific tissue patterns and discriminative cues. Fine-tuning can improve target adaptation, but introduces a trade-off between pathology-specific representation capacity and adaptation efficiency, particularly in data-scarce settings. To address this, we propose a pathology role-guided mixture-of-experts feed-forward network (MoE-FFN) framework for efficient encoder-level representation learning. We design a two-stage training paradigm to establish and adapt pathology-aware expert specialization. In source-domain expert initialization, pathology-specific priors are distilled from a frozen Virchow2 teacher into a lightweight DINOv2-small student, while role prototypes serve as weak pathological anchors to encourage distinct expert functions. MoE-FFN blocks are introduced into selected high-level transformer layers to provide transformation diversity for heterogeneous pathological patterns. In target-domain adaptation, the initialized experts are refined through asymmetric prototype-guided optimization, enhancing task-relevant positive evidence and separating confusable hard negatives. The resulting encoder extracts offline patch representations that can be directly integrated with standard MIL aggregators. Experiments on the public BRACS dataset and a private PAROTID WSI dataset across five representative backbones demonstrate consistent improvements over the strongest baseline.
[CV-97] LVMT: Video Mask Transformer for Long-term Video Segmentation
链接: https://arxiv.org/abs/2609.34895
作者: Narges Norouzi,Niccol`o Cavagnero,Idil Esen Zulfikar,Bastian Leibe,Gijs Dubbelman,Daan de Geus
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Existing online video segmentation methods struggle to track objects in long, complex videos with long-term occlusions. We hypothesize that this limitation is caused by (i) the inability of their temporal propagation mechanism to adaptively select the object information that is propagated across time, and (ii) their inability to be trained on long videos due to memory requirements and vanishing gradients. To address the first limitation, we propose to use a lightweight GRU-based temporal propagation module that can learn to select which information it keeps in memory and propagates across time. Second, to allow training on long videos, we introduce Truncated Query Propagation (TQP), a training strategy in which the model processes a video in chunks of frames, where information about tracked objects is propagated between chunks but backpropagation is only conducted in individual chunks, enabling longer temporal supervision without out-of-memory issues, inference overhead, or vanishing gradients. The resulting model is called the Long-term Video Mask Transformer (LVMT). Extensive experiments on six benchmarks show that LVMT sets a new state of the art across a range of video segmentation tasks, while retaining the speed of the highly efficient model it is based on, making it 10X faster than the prior state of the art. Code: this https URL
[CV-98] ECHO: Event-Augmented Context with Hindsight and Outlook for Wrist-Only Manipulation
链接: https://arxiv.org/abs/2609.34893
作者: Xinyue Wang,Yicheng Jiang,Zesen Gan,Junhao He,Jiaxu Wang,Junhao Li,Jingtao Zhang,Tianlun He,Jianan Wang,Isabel Guan,Qiming Shao
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Learning-based manipulation policies relying on RGB cameras often suffer from degraded observations under extreme exposure. Event cameras mitigate this degradation by asynchronously detecting pixel-level intensity changes to offer a high dynamic range. However, their observations heavily depend on camera placement, as fixed cameras miss static scene content while wrist-mounted camera motion causes previously visited regions to leave the field of view. To address these spatial-temporal limitations, we present ECHO (Event-augmented Context with Hindsight and Outlook), a wrist-only latent world action model that encodes wrist events into compact motion representations to provide temporal and spatial context for policy reasoning. Specifically, ECHO utilizes a pretrained event encoder to explain visual-feature changes between frames. Its hindsight module preserves the gripper trajectory with past event stream as addressable off-camera context. Concurrently, the outlook module introduces learnable event foresight queries supervised to anticipate the event window for future actions, enabling the policy to predict upcoming scene changes. Evaluated on wrist-only RLBench tasks, ECHO outperforms RGB and RGB+event baselines by 20.6 and 12.0 percentage points under normal lighting, and by 14.6 and 11.3 points under severe exposure drops, respectively, while also surpassing RGB references using a third-person camera. Real-world experiments with a wrist-mounted event camera validate that ECHO outperforms RGB-only and RGB+event baselines across multiple tasks under both nominal and severely dark lighting. Project page is at this https URL.
[CV-99] SubRot: Signed Gradient Subspace Calibration for VLM Rotation Quantization
链接: https://arxiv.org/abs/2609.34884
作者: Zhenhao Shang,Haizhao Jing,Haokui Zhang,Guoting Wei,Rong Xiao,Jianqing Gao,Peng Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Post-training quantization reduces the deployment cost of vision-language models (VLMs), but preserving multimodal capabilities at low bit widths remains challenging. Existing methods rely on modality- or token-level gradient statistics, which are susceptible to cross-sample variations in visual-to-textual token ratios and the positions of visual information, limiting statistical stability. Moreover, overly coarse aggregation through absolute values and averaging discards gradient signs and channel-wise differences, limiting the separation of modality-specific sensitivities. In contrast, the channel space provides a shared coordinate system across samples, making it a more natural basis for capturing stable task-sensitive structures. We therefore propose SubRot, a signed gradient subspace calibration method for VLM rotation quantization. Through eigendecomposition of the empirical Fisher matrix of activation gradients, SubRot identifies a sensitive channel subspace with three properties: cross-sample stability, clear sensitivity separation, and consistent signed effects on the autoregressive loss along certain directions. Guided by a local Taylor expansion, SubRot combines signed first-order guidance along sign-stable directions with second-order constraints along the remaining sensitive directions, while retaining MSE for overall reconstruction quality. This objective steers quantization errors toward loss-decreasing directions while controlling their magnitude. Experiments on five VLMs across five benchmarks show consistent average-score improvements over FlatQuant under W4A6 and W4A4, reaching 1.4 percentage points on LLaVA-NeXT-7B. Under W4A4, average accuracy degradation from FP16 remains within 1.4 percentage points across all evaluated models, while LLaVA-v1.5-13B exceeds its FP16 average score by 0.4 percentage points.
[CV-100] SPOC-Net: Single-Primitive Online Composition Network for GNSS Jamming Set Recognition
链接: https://arxiv.org/abs/2609.34875
作者: Zhihan Zeng,Kaihe Wang,José A. López-Salcedo,Gonzalo Seco-Granados,Zhongpei Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reliable positioning, navigation, and timing support intelligent transportation, autonomous systems, and space-air-ground integrated networks. However, global navigation satellite system (GNSS) jamming recognizers that treat each mixture as a separate class are difficult to extend to new combinations. Therefore, this paper proposes SPOC-Net, which decomposes the recognition problem into identifying a set of basic jamming components. Multi-resolution time-frequency features and learned component queries provide evidence for each component type. A high-resolution branch estimates the number of active types, and a structured decoder combines this estimate with component evidence to select a valid set. For training, measured single-component records are the only physical samples used in gradient optimization. Their associated clean in-phase and quadrature (IQ) sequences are combined on demand during training to produce labeled mixtures with different relative powers and jamming-to-noise ratios. Separate measured mixtures from ten training-listed compositions support model selection and decoder calibration; six other compositions are reserved for final testing. Evaluation on 14,220 independently generated, conductively combined, and recorded radio frequency mixtures yields 80.69% exact-set accuracy and a 92.84% micro-averaged F1 score. On combinations excluded from model development, SPOC-Net achieves 80.89% exact-set accuracy, exceeding the strongest comparison method by 18.77 percentage points under the reported protocols.
[CV-101] P4Q: Co-designing Token Pruning and Quantization for Vision-Language Model Acceleration
链接: https://arxiv.org/abs/2609.34867
作者: Haizhao Jing,Zhenhao Shang,Haokui Zhang,Rong Xiao,Peng Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision language models have achieved strong performance across a wide range of multimodal applications, yet their substantial computational and memory costs hinder efficient deployment. Visual token pruning and post-training quantization reduce inference overhead along two complementary dimensions, namely sequence length and numerical precision. Existing workflows typically optimize these techniques independently or apply them sequentially. Their distinct optimization objectives leave critical interactions unaddressed and constrain the achievable compression performance. We revisit these designs and present P4Q, a practical co-design framework that jointly optimizes visual token pruning and low-bit quantization for efficient VLM inference. First, P4Q introduces a quantization-aware visual token selection strategy before the LLM. It applies fake quantization to copies of the features produced by the projector and selects visual tokens using statistics computed from these fake-quantized features, thereby conditioning the selector’s feature-based decisions on simulated low-bit perturbations. Second, P4Q introduces a pruning-aware quantization calibration strategy. It uses the same selection strategy as pruning to calibrate the quantized model on the retained-token distribution, thereby aligning the calibration process with the pruned execution path used during deployment. By coupling these two components, P4Q achieves substantial inference speedups while maintaining comparable task performance, resulting in a better efficiency-accuracy trade-off than independently optimized pipelines. For instance, on LLaVA-NeXT, P4Q achieves an average end-to-end inference speedup of 2.8x across eight distinct test sets, while retaining higher accuracy than prior compression and quantization methods.
[CV-102] Revisit to Segment: Working Memory Distillation for Reasoning Segmentation
链接: https://arxiv.org/abs/2609.34863
作者: Cilin Yan,Yilun Qiu,Wanyang Zhang,Rui Zu,Xiaolong Jiang,Yao Hu,Jiayin Cai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multimodal large language models (MLLMs) have approached image segmentation by reasoning about visual content and predicting target locations. Their generated responses contain reasoning traces and localization proposals that can serve as working memory when revisiting the same image and query. Our exploration reveals that MLLMs benefit from using this self-generated working memory as context, leading to enhanced reasoning segmentation. Motivated by this finding, we seek to strengthen the backbone model’s reasoning segmentation capabilities by distilling the guidance gained from revisiting prior attempts, enabling it to benefit with or without working memory at inference time. To this end, we propose Reasoning Segmenter with Working Memory (SWiM), a working-memory distillation framework for reasoning segmentation. Specifically, SWiM selects rollouts based on segmentation quality to construct working memory and uses the memory-conditioned model as a teacher. The teacher provides token-level distributional supervision along student-generated trajectories, while the student receives only the original image and query. Joint optimization of on-policy self-distillation and outcome-based reinforcement learning combines working-memory guidance with direct feedback on segmentation quality. Extensive experiments on reasoning segmentation benchmarks demonstrate that SWiM achieves state-of-the-art performance, validating the effectiveness of working-memory distillation.
[CV-103] When Text Matters: Design Principles for Visual Token Pruning in Vision-Language Model
链接: https://arxiv.org/abs/2609.34861
作者: Minchan Kang,Kyeonghye Park,Seoyoung Cho,Daeshik Kim,Yucheol Cho
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Visual token pruning has been widely studied as a practical approach to reducing the computational cost of large vision-language models. However, it struggles to preserve essential visual information, which can lead to substantial performance degradation. In particular, image-based token selection can overlook task-relevant details, while text-guided token selection may fail to capture the text–visual relationships needed for complex reasoning. We find that applying textual guidance too early can limit its ability to identify answer-relevant visual regions, whereas text-to-visual attention becomes more informative at intermediate decoder depths. This finding motivates our training-free method, which separates early vision-guided pruning from deferred text-guided reselection. We first prune visual tokens using vision-encoder attention, retain additional candidates until the decoder midpoint, and then use text-to-visual attention to determine the final visual-token set. Across eight benchmarks and three models, our method outperforms the best-performing baselines by an average of 11.10 and 16.84 percentage points in performance recovery at 80% and 90% pruning, respectively, with comparable or lower LLM-prefill latency than most baselines. The source code is publicly available at this https URL
[CV-104] EviSplat: Preserving Multi-View Evidence in 3D Gaussian Splatting for Open-Vocabulary Segmentation
链接: https://arxiv.org/abs/2609.34853
作者: Sungho Moon,Kota Shimomura,Junwoo Park,Wonhyeok Choi,Seunghun Lee,Takayoshi Yamashita,Sunghoon Im
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 23 pages, 7 figures, including appendix
Abstract:Open-vocabulary 3D scene understanding enables object localization and segmentation from free-form text queries without a fixed category vocabulary. Many recent methods build on 3D Gaussian Splatting and consolidate multi-view observations, such as masked crops from individual views, into language features or compact object descriptors before the query is known. However, observations of the same object vary across viewpoints and are not equally informative: some reveal cues relevant to a particular query, whereas others provide incomplete or misleading evidence. Pre-query consolidation can therefore suppress cues on which a later query depends. We introduce EviSplat, which preserves individual observation features as evidence for later text queries. EviSplat retains individual observation features within class-agnostic 3D instances that represent objects, object parts, or background regions. It also learns, for each Gaussian, a distribution describing which visual appearances its observations support. Given a text query, EviSplat scores each instance using its most relevant observations. It then computes a score for each Gaussian by combining instance-level relevance with locally supported evidence, weighted by how often and how unambiguously that Gaussian was observed. Different queries can thus draw on different visual cues from the same preserved evidence. Experiments across diverse datasets and evaluation protocols demonstrate state-of-the-art performance, supporting the benefit of preserving multi-view evidence until query time and aggregating it according to the query.
[CV-105] ORAV: Benchmarking Audio-Video Generation from Multimodal Contexts
链接: https://arxiv.org/abs/2609.34843
作者: Jiacheng Hua,Xiaokun Feng,Jiaqi Hua,Chang Liu,Biao Wang,Miao Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 25 pages, 10 figures, 13 tables
Abstract:Audio-video generation using heterogeneous multimodal references has emerged as a new challenge, requiring both compositional control over generation and grounded understanding of multimodal context. In this paper, we introduce ORAV Bench for Omni Reference Audio-Video Generation, comprising 380 task instances with 2-10 references, 9 semantic roles, and 30 role compositions. Instructions specify the relationships among references; the media supply the identities, dynamics, and audio characteristics to be realized. To evaluate these open-ended outputs, we develop a reference-aware pairwise protocol that prepares visual and auditory evidence, compares the intended contribution of each reference, and checks the overall verdict in both presentation orders. On held-out instances, it achieves 86.08% effective agreement with human judgments. Across 5 frontier systems, overall rankings conceal distinct strengths across reference compositions. A recurring failure is to reproduce unintended source content in place of the requested result, despite closely resembling a reference. Reproducible pointwise diagnostics of quality, reference affinity, and speech reveal distinct dimensions of model behavior. ORAV thus offers a benchmark for tracking progress toward controllable, compositional, and reference-faithful audio-video generation.
[CV-106] ransform-Aligned Learned Features for Lossy Point Cloud Attribute Compression
链接: https://arxiv.org/abs/2609.34834
作者: Yueru Chen,Pengpeng Yu,Dingquan Li,Wei Gao,Wei Zhang,Fei Song
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 19 pages
Abstract:Transform-based methods provide an effective framework for point cloud attribute compression by representing attributes as transform coefficients. Introducing learned spatial context into this framework requires mapping spatial representations to the transform domain, but this known basis change is often left for the network to learn implicitly. We propose Transform-Aligned Learned Features (TALF) by applying the attribute transform to learned spatial representations, explicitly aligning them with the coding targets. Our analysis shows that the resulting features exactly represent the first-order prediction term of a smooth nonlinear model, with a bounded Taylor remainder. We integrate TALF into a transform-based attribute codec with explicit coefficient prediction and conditional residual entropy modeling under a unified coefficient-domain rate–distortion objective, while retaining explicit quantization-step control. Extensive experiments across three benchmark datasets and multiple transform bases demonstrate that TALF improves rate–distortion performance over conventional and learned baselines.
[CV-107] Multi-Scale Semantic Mapping in Urban Environments via Observation Calibration and Policy Dependence Regularization
链接: https://arxiv.org/abs/2609.34833
作者: Runling Long,Junhao Feng,Jia Wan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Semantic mapping is fundamental to embodied navigation, yet existing methods are developed for indoor environments, where objects exhibit relatively limited scale variation and are observed from a restricted range of viewpoints. Urban environments pose substantially greater challenges: agents must map objects ranging from pedestrians to buildings while navigating large spaces with highly diverse viewing distances. These conditions introduce two key difficulties that existing datasets and methods fail to cover. First, object scale and observation distance can be severely mismatched. For example, small objects may be viewed from far away, whereas large objects may be observed at extremely close range, resulting in unreliable observation likelihoods. Second, objects with substantially different sizes and geometries require distinct mapping behaviors, which are difficult to capture with a single shared value estimator. To investigate these challenges, we introduce a large-scale urban semantic mapping dataset featuring realistic city layouts, high-fidelity rendering, and instance-level annotations spanning multiple object scales. We then propose a category-aware likelihood calibration policy that identifies and alleviates unreliable observations according to object category and viewing distance. Because the calibration and motion policies are optimized toward the same mapping objective, they may learn redundant shortcuts and become excessively coupled. We therefore introduce a mutual-information (MI) regularizer that penalizes their estimated representation dependence and encourages complementary behaviors. To better model heterogeneous mapping strategies across object scales, we further employ category-wise value estimators. We formulate their joint optimization as a Pareto optimization problem to mitigate conflicting gradients across categories.
[CV-108] WM-VLM: Probing Internal World Models for Interleaved Visual-Textual Reasoning
链接: https://arxiv.org/abs/2609.34826
作者: Yuheng Zha,Yilei Wang,Qiyue Gao,Junrong Chen,Yujia Wu,Zhengfeng Lai,Zhengzhong Liu,Eric P. Xing
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages, 11 figures
Abstract:Humans often solve spatial problems by mentally simulating visual transformations. In contrast, conventional vision-language models (VLMs) reason primarily through language. We investigate whether VLMs can solve spatial problems by reasoning with both text and generated visual states. To this end, we introduce WM-VLM, which equips a pretrained VLM with a lightweight world model branch for generating intermediate visual states. Our two-stage training first teaches the model to generate the next visual state and then to use that state for reasoning. We programmatically construct spatial reasoning tasks with verifiable intermediate visual states. These tasks allow us to evaluate how well the model generates visual states and how much it relies on them to answer the question. On 2D and 3D mental rotation tasks, WM-VLM consistently outperforms the supervised fine-tuned backbone, with gains of up to 39.25 percentage points. Ablations suggest that these gains depend on the generated visual states, as removing or corrupting them sharply reduces performance. Together, these results suggest that internal world models offer a promising path toward VLMs that reason in both language and visual space.
[CV-109] Generative Residual Factorization
链接: https://arxiv.org/abs/2609.34824
作者: Letian Gong,Yuzhou Hong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 27 pages, 3 figures
Abstract:Under a shared-factor model, the conditional law of the next image patch factors into a posterior over the shared scene factor and a residual kernel given that factor. A sufficient statistic of the past replaces the raw past in the posterior and does not replace the kernel. The conditional entropy splits into residual entropy, which no observation of the factor can remove, and a posterior term, which a better representation of the past can remove. Next-embedding prediction is a directional likelihood on a shallow map, so the fiber of that map is unidentified and a constant embedding remains a minimizer. The same split is an equality in a scalar Gaussian model, evaluated in closed form.
[CV-110] ESTHER: Egocentric Stereo Hand Estimation and Reconstruction in the Wild
链接: https://arxiv.org/abs/2609.34817
作者: Hongyu Ma,Hairong Qu,Shiqi Zhao,Yongsong Yang,Peng Yin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Human dexterity is guided by two eyes watching two hands: binocular vision supplies the metric 3D structure that fine-grained manipulation consumes. Egocentric stereo is therefore the natural perceptual interface for robots, AR, and VR-yet metric 3D hand reconstruction from this very signal still has neither an end-to-end model nor an in-the-wild benchmark. We propose ESTHER, a model whose stereo geometry, temporal reasoning, and output representation are designed for wearable egocentric stereo. It is trained on pseudo-labels from a calibrated labeling pipeline and in turn assembles our benchmark ESTHER3D, an egocentric stereo hand dataset pairing a large in-the-wild training set of model-generated labels with a motion capture test set of true metric ground truth. Experiments show state-of-the-art accu?racy, superior external generalization, and robustness to the missing views, dropped frames, and lighting and motion blur extremes of real egocentric capture that break existing meth?ods. This robustness runs deeper than graceful degradation: stereo guidance teaches the model to bind apparent hand scale to metric depth, so it not only adapts to different stereo rigs and modalities with minimal fine-tuning, but more strikingly preserves true metric scale even after collapsing to a single monocular view.
[CV-111] From Perception to Integration: Revisiting the Internal Dynamics of Reasoning in Vision-Language Models
链接: https://arxiv.org/abs/2609.34809
作者: Rong Yu Xu,Prayag Tiwari,Shaolei Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 15 pages, 4 figures. Code: this https URL
Abstract:Vision-language models (VLMs) can answer simple visual questions, but often struggle when one question requires several visual judgments. We study this gap with controlled tasks for feature binding, numerosity, spatial relations, and amodal completion, together with a Composite task that combines them. Matched counterfactual image pairs isolate changes in the visual evidence needed to answer. Across four models, direct answers, hidden-state readouts, and state interventions show that the individual judgments can be made without explicit reasoning and that intervening on the corresponding states can affect the answer. During reasoning, the Composite answer becomes decodable from hidden states and usable from shortened traces, often before the model stops on its own. We train a small detector to predict this readiness and stop reasoning at that point. On MMStar and RealWorldQA, this reduces mean reasoning tokens by 79.1% and 74.5%, while average accuracy rises by 3.13 and 3.30 percentage points, respectively. These findings connect the internal development of answer readiness to a practical rule for allocating reasoning computation.
[CV-112] ControlTrace: Recovering Control Fields for Hidden-Content Recognition
链接: https://arxiv.org/abs/2609.34807
作者: Zijian Liu,Yaoguang Chen,Liwei Liu,Weixi Wu,Hanming Zhang,Jiashui Wang,Na Ruan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 31 pages, 10 figures
Abstract:Spatially conditioned diffusion models can embed words and contours in natural-looking images, but vision-language models (VLMs) may fail to recognize the hidden content. Transformation-based recovery depends on parameter and view selection. To evaluate hidden-content recovery and recognition, we construct FreqBlind, a 6,000-image benchmark spanning contours, real words and non-words across three conditioning strengths. The evaluated transformation-based methods show limited recognition of contour patterns and weakly conditioned hidden content. To address this limitation, we propose ControlTrace to recover the grayscale control field used during generation. An 8.4M-parameter U-Net predicts this field from the carrier image, and a VLM then identifies its content. With Qwen2.5-VL-7B-Instruct, ControlTrace achieves 60.2% open-ended contour recognition accuracy across the three conditioning strengths, exceeding the best of the three evaluated prior methods by 26.9 percentage points. On an A100 GPU, the complete pipeline adds only 7.4 ms (5.3%) to direct VLM inference. Recovered fields have lower pixel errors and higher structural similarity than the evaluated transformation views. Across four evaluated VLMs, ControlTrace retains its overall contour recognition advantage. Recognition remains stable under the tested JPEG compression, Gaussian noise and downsampling. These results support control-field recovery for hidden-content recognition in the evaluated setting.
[CV-113] Physics-Guided Spectral Distillation for Underwater Image Enhancement on Resource-Constrained Devices
链接: https://arxiv.org/abs/2609.34795
作者: Yifan Chen,Kai He,Ye Zheng,Jijun Lu,Zhe Sun,Tao Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 9 figures
Abstract:Underwater image enhancement is crucial for improving visual perception in marine applications. Existing underwater image enhancement studies mainly focus on enhancement quality and visual fidelity, while rarely considering real-time deployment capability, which is essential for resource-constrained underwater robots. To this end, we introduce a physics-guided spectral distillation (PSD) method, which reduces model capacity for real-time applications while maintaining the high performance of underwater image enhancement models. To decompose the outputs of teacher and student models, PSD adopts a multilevel Haar discrete wavelet transform. It transfers low-frequency color and illumination information as well as high-frequency structural details through band-specific objectives. Moreover, the distillation process of PSD is degradation-aware. We estimate degradation-aware weights through a physical head and combine them with ground-truth-guided reliability masks to selectively retain valuable teacher guidance. Experiments on the UIEB, LSUI, and EUVP datasets validate the effectiveness of the proposed method. Furthermore, we demonstrate the benefits of enhanced images for downstream perception tasks, including object detection. Deployment on a self-developed ROV further demonstrates its practical applicability in real-world underwater scenarios.
[CV-114] D2-VLA: Dual-Memory Dual-Frequency Vision-Language-Action Model For Long Dynamic Manipulation
链接: https://arxiv.org/abs/2609.34792
作者: Zijian Ye,Chengqi Wei,Wei Huang,Anlin Zheng,Chunyu Zou,Liangyu Wu,Zikang Zhao,Zhenjie Peng,Yushuo Yang,Shuman Zhao,Zhongrui Wang,Xiaojuan Qi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 30 pages
Abstract:Long-horizon manipulation requires robots to remember cues that are no longer in view while responding to moving objects. Yet vision-language-action (VLA) policies often rely on the latest observation, and refreshing their visual context typically requires another costly vision-language model (VLM) pass. We present D ^2 -VLA, which combines dual memory and dual-frequency control at the KV-cache interface of a pretrained VLA. D ^2 -VLA uses block-wise causal KV caching to encode observations incrementally and, guided by distinct temporal attention patterns, constructs separate historical KV read views for the VLM and action expert. Between periodic VLM updates, a gated adapter incorporates fresh visual features into the latest history-conditioned KV block, while a short fast-memory queue supports action replanning. We introduce DOMINO-Long, a ten-task benchmark requiring robots to use earlier visual cues when manipulating moving objects. D ^2 -VLA achieves complete-task success rates of 29.3% on DOMINO, compared with 9.6% for \pi_0.5 and 17.2% for PUMA, and 60.0% on DOMINO-Long, compared with 35.4% and 20.6%, respectively. It improves success rates on eight real-robot tasks and reaches 97.5% on LIBERO-Long and 74.3% on RoboTwin 2.0.
[CV-115] Projective Normal Fields: A Convex Optimization Method for Constructing Smooth UDFs
链接: https://arxiv.org/abs/2609.34784
作者: Jiayi Kong,Chen Zong,Fei Hou,Junhui Hou,Wenping Wang,Ying He
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Constructing a smooth approximation of an unsigned distance field (UDF) from a raw point cloud is challenging because the input provides neither surface connectivity nor consistently oriented normals. Methods that directly learn a scalar UDF must also handle its non-differentiability on the zero level set and weak supervision away from the samples, which can lead to unstable optimization and spatial artifacts. We introduce Projective Normal Fields (PNFs), an orientation-free representation and convex optimization framework for estimating bidirectional normals from point positions alone. Each normal axis is encoded by a rank-one projector, which is invariant to normal reversal. We relax the non-convex set of hard projectors to its convex hull: the symmetric positive-semidefinite matrices with unit trace. Each soft tensor defines a local quadratic distance model and retains the relative weights of candidate normal axes. We estimate a coherent PNF by combining local tangent-plane fitting, soft-PCA anchoring, and overlap regularization on a fixed neighborhood graph. With positive anchoring weights, the objective is strongly convex and admits a unique global minimizer. Principal eigenvectors provide bidirectional normals, while the corresponding eigengaps provide spectral confidence indicators. We use these indicators to select and weight directional sources for heat diffusion, followed by Poisson integration to construct a regularized UDF approximation. By separating local geometry estimation from scalar-field construction, PNF avoids directly fitting the non-differentiable UDF. Experiments demonstrate reduced sensitivity to neighborhood size, competitive reconstruction under noise and outliers, and improved accuracy near non-manifold junctions. The project page is available at this https URL
[CV-116] CoHuB: A Simulation Benchmark for Multi-Humanoid Collaboration
链接: https://arxiv.org/abs/2609.34782
作者: Hyunjin Park,Jebeom Chae,Minwoo Park,Sunghyun Park,Hanjun Yoo,Seoyeon Choi,Soochul Yoo,Joohwan Seo,Sarmad Idrees,Jae-Sang Hyun,Jongmin Lee,Roberto Horowitz,Youngwoon Lee,Jongeun Choi
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Project page: this https URL
Abstract:Many physical tasks in human environments require collaboration, from assisting a partner to jointly manipulating an object. Yet, existing humanoid benchmarks largely focus on single-humanoid skills and lack evaluation of multi-humanoid collaboration under egocentric visual observations. We introduce CoHuB (Collaborative Multi-Humanoid Benchmark), a simulation benchmark for multi-humanoid collaboration under egocentric visual observations. CoHuB provides 10 tasks, eight with two humanoids and two with three humanoids, spanning diverse collaboration patterns. We also provide synchronized demonstrations collected through a multi-operator VR teleoperation pipeline, in which each operator controls one humanoid from its egocentric view. Experiments with representative visuomotor policies reveal substantial challenges across different forms of coordinated perception and control. CoHuB provides a foundation for developing and evaluating multi-humanoid collaboration policies.
[CV-117] When VLMs Trust Context: Evaluating Scene Text Recognition under Misleading Context
链接: https://arxiv.org/abs/2609.34781
作者: Yuxing Cheng,Yuan Wu,Yi Chang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision-language models (VLMs) can read text in natural scenes, but their predictions may be influenced by the surrounding context. When the printed text conflicts with what the scene suggests, a model may return a more plausible word instead of the shown text. We introduce SceneFaith, a benchmark of 781 generated scene images for studying this behavior. Each output is classified as Literal, Canonical, or Other, separating faithful transcription from context-consistent rewriting and ordinary recognition errors. Across 15 models from seven families, all models show rewriting on clear images, with rates ranging from 8.45% to 58.51%. Controlled experiments further show that surrounding context matters: removing surrounding scene information reduces rewriting and improves literal accuracy, while changing the scene around the same text patch can also change model outputs. Moreover, weakening the target text with blur increases rewriting. These results show that reliable scene-text recognition requires VLMs to balance visual character evidence with contextual information, preserving clear text while using context mainly when the visual evidence is uncertain.
[CV-118] Privacy-Preserving Full-Body Meshing from mmWave Radar via Mesh Foundation Model Supervision
链接: https://arxiv.org/abs/2609.34768
作者: Shuxing Zhang,Yongquan Ni,Zhenyu Ding,Yawen Lin
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Millimeter-wave (mmWave) radar enables privacy-preserving human perception, but the extreme sparsity of point clouds from commercial single-chip sensors (mean ~6.5 points/frame; ~28% empty frames) has confined prior art to body-part keypoints or discrete action classification. We present a cross-modal teacher-student framework that lifts commercial radar to full-body, per-frame, metric 3D mesh reconstruction with per-joint uncertainty. Three innovations: (1) a mesh-foundation-model teacher - SAM 3D Body produces whole-body MHR ground truth (70 joints, 18,439 mesh vertices) from a single RGB frame with zero training, slashing annotation cost by orders of magnitude; (2) StudentPoseFormer - set encoding with masked attention pooling, a temporal Transformer, and a CVAE multi-hypothesis head that outputs both the pose mean and per-joint variance, honestly reporting where the radar cannot see; and (3) a multi-stage ground-truth quality pipeline (confidence gating, depth validation, temporal smoothing, bone-length consistency, bad-frame rejection) plus systematic information-lever ablations. On the public MM-Fi benchmark (same TI IWR6843 sensor, cross-subject), our full configuration reaches 7.45 cm 12-joint MPJPE, with ablations proving the causal value of point accumulation (k = 3, -0.34 cm), Doppler (-0.85 cm; -2 cm at the wrist on fast actions), and velocity loss (-0.27 cm). On our own synchronized radar + RGB-D corpus with block-level held-out splits, the pipeline achieves 21.47 cm end-to-end (per-joint hierarchy from 4.8 cm at the hip to 34.7 cm at the wrist - matching physical information limits), could be improved to 15 cm with ~30k diverse samples, and a scaling law shows sample diversity, not volume, is the binding constraint. Deployment inference is radar-only - no camera, no image.
[CV-119] Beyond Reconstruction Loss in Post-Training Quantization: Balanced Fitting for Large Vision-Language Models
链接: https://arxiv.org/abs/2609.34765
作者: Minchan Kang,Kyeonghye Park,Seungyeon Sa,Seoyoung Cho,Daeshik Kim,Yucheol Cho
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Post-training quantization (PTQ) enables efficient deployment of large vision-language models (LVLMs), but is typically calibrated on a small set while expected to generalize across diverse downstream tasks. Although recent PTQ methods for LVLMs incorporate sensitivity signals, they still minimize reconstruction loss with respect to the full-precision model, potentially over-preserving FP behavior and calibration-specific bias. Rather than treating quantization solely as an error to be minimized, we observe that it can also provide beneficial regularization for certain layers and modalities. Motivated by this observation, we propose Balanced Fitting, a quantization effect-based framework that balances precision and regularization beyond reconstruction-based optimization. By measuring layer- and component-wise quantization effects for weights, vision activations, and text activations, Balanced Fitting combines fine-grained fitting for sensitive components with coarser fitting to exploit potential regularization benefits. Experiments on multiple LVLMs show that our method consistently outperforms prior PTQ approaches under both weight-only and weight-activation quantization, while lower reconstruction loss does not reliably translate into better downstream performance. The source code is publicly available at this https URL
[CV-120] PanoVLN: Towards Effective Panoramic Vision-and-Language Navigation
链接: https://arxiv.org/abs/2609.34759
作者: Zhen Wang,Changpeng Wang,Zhe Liu,Zhangyang Qi,Yuxiang Lu,Zimo Zeng,Donglian Qi,Xi Chen
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 22 pages, 14 figures. Project page: this https URL
Abstract:Recent vision-language models (VLMs) have advanced vision-and-language navigation (VLN), enabling models to predict navigation actions from visual observations and language instructions. In this work, we explore VLN with panoramic observations and introduce PanoVLN. The motivation is straightforward: more complete visual context should enable better-informed navigation decisions. For example, a panorama can reveal a passage outside a perspective camera’s field of view, allowing the model to identify the intended route without additional exploration. However, we find that simply replacing perspective images with panoramas yields only limited gains. Our diagnosis suggests that fully exploiting wider visibility requires modifications to action prediction, training supervision, and visual representation. First, wider visibility supports longer-horizon action planning. We make the model predict longer action sequences, enabling larger turns and subsequent movement from a single panorama. Specifically, we introduce a confidence-guided execution (CGE) strategy that dynamically determines how many predicted actions to execute before replanning. Second, wider visibility also brings more complex route choices. We therefore construct training routes with frequent branching points and clear instructions to provide targeted supervision for route selection. Third, panoramic navigation requires understanding spatial relationships across viewing directions, beyond recognizing individual landmarks. We combine semantic and geometric features from RGB panoramas to capture both scene content and spatial layout without adding visual tokens. With a 4B backbone and RGB-only input, PanoVLN surpasses the previous SOTA by 11.9% and 8.7% in success rate on R2R-CE and RxR-CE Val-Unseen. Real-world experiments on a quadruped further demonstrate faster navigation with fewer pauses than prior VLN methods.
[CV-121] A Unifying Framework of Concept-based Explainable AI with Completeness Guarantees
链接: https://arxiv.org/abs/2609.34750
作者: Vojtěch Kůr,Adam Kukučka,Tomáš Brázdil,Vít Musil
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Concept-based explanations describe neural network predictions through human-understandable properties of inputs called concepts. The field encompasses approaches that differ in how they define and represent concepts and connect them to model predictions. We introduce a theoretical framework that describes these approaches in a common mathematical language and supports a shared analysis of their properties. For concept discovery, which identifies concepts automatically within a latent space of a trained model, we employ a concept autoencoder view. An encoder extracts concept representations from the model’s latent space, and a decoder uses them to reconstruct the original latent representation. The autoencoder’s reconstruction error measures how accurately its decoder recovers the original latent representation. We revisit model completeness: how well the concepts can reproduce the model’s outputs. We show that model incompleteness of the concepts can be bounded by the autoencoder’s reconstruction error. The autoencoder view also provides a common way to define individual concept attributions, which measure each concept’s contribution to a prediction. We establish when these attributions sum to the model’s prediction, and bound the discrepancy otherwise, thus providing attribution completeness guarantees.
[CV-122] CoDrive: Cross-Vehicle World-Consistent Video Generation with Precise Trajectory Control for Cooperative Driving
链接: https://arxiv.org/abs/2609.34749
作者: Yu Meng,Baining Zhao,Junta Wu,Tengfei Wang,Rongze Tang,Haiyu Zhang,Wenqiang Sun,Chen Gao,Zhibo Chen,Xinlei Chen,Yong Li,Xiao-Ping Zhang,Chunchao Guo
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 28 pages, 6 figures
Abstract:Real-world driving is inherently multi-agent, yet most existing driving world models generate observations from a single ego vehicle. Independently extending them to multiple vehicles does not ensure that different agents observe a consistent shared world. We present CoDrive, a cross-vehicle, multi-view driving video generation framework that jointly generates observations of vehicles sharing the same dynamic scene with precise camera-trajectory control. CoDrive interleaves local self-attention, which models spatiotemporal dependencies among the views of each vehicle, with global self-attention, which enables information exchange and consistency modeling across vehicles. To explicitly encode their spatial relationships, all camera trajectories are represented in a shared world coordinate system and injected into the attention layers through projective relative positional encoding. We further adopt a progressive mixed-task training strategy that combines large-scale real-world single-agent data with synthetic cross-agent interaction data, allowing the model to benefit from real-world appearance distributions while learning cross-agent consistency from simulation. For systematic evaluation, we introduce CoDrive-Bench, a benchmark covering real and synthetic multi-vehicle scenarios and evaluating trajectory controllability, scene geometry consistency, and instance-level consistency. Experiments show that CoDrive improves trajectory controllability and cross-agent geometric and instance consistency while maintaining competitive visual quality.
[CV-123] Optimizing and Securing the Modern Watermarking Channel for Images
链接: https://arxiv.org/abs/2609.34744
作者: Enoal Gesny,Eva Giboulot
类目: Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:To comply with recent regulations requiring traceable generated content, modern watermarking has adopted multi-bit post-hoc watermarking schemes. These modern designs rest on an encoder-decoder pair implemented as deep neural networks. These models are usually treated as pure black-boxes trained end-to-end, with the noise of the watermarking channel modeled through a fixed set of geometric and valuemetric transforms applied to watermarked images. We argue that this purely empirical approach leads to unquestioned design flaws and a lack of theoretical performance guarantees. This work proposes a general theoretical model of modern post-hoc watermarking schemes grounded in a statistical analysis of the outputs of the encoder/decoder pair. We show that these deep neural networks implicitly define a watermarking channel modeled as parallel AWGN channels, with messages transmitted using BPSK modulation. This imposes a binary alphabet, greatly limiting the capacity of these watermarking systems. Another fatal flaw is their lack of a secret key, making them intrinsically insecure. We make this notion of watermarking security precise for post-hoc schemes by linking it to the possibility of estimating the secret key under a given statistical model of the decoder’s output. By putting together the results from this theoretical analysis, we introduce SNW: a novel post-hoc watermarking system that significantly outperforms existing state-of-the-art baselines in terms of capacity while also providing strong security guarantees. Notably, it does not depend on a fixed codebook or binary alphabet, allowing it to reach a rate close to Shannon capacity through the use of capacity-achieving error-correcting codes.
[CV-124] Do Emotion Concepts Generalize Across Sources Modalities and Architectures in Vision-Language Models?
链接: https://arxiv.org/abs/2609.34742
作者: Bohao Xing,Xin Liu,Kaishen Yuan,Deng Li,Rong Gao,Guoying Zhao,Xiaolan Fu,Heikki Kälviäinen
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Recent studies suggest that large language models encode emotion concepts as structured internal representations, but most existing work focuses on text and a single architecture. Therefore, we ask, do emotion concepts generalize across sources, modalities, and architectures in vision–language models (VLMs)? To address this, we construct CMES (Cross-Modal Emotion Stimuli), a multi-source collection of emotion-conditioned stories, real facial expressions, synthetic portraits, and synthetic emotion-evoking scenes. For each stimulus source, we extract a separate set of six Ekman emotion vectors from each of three VLMs. We report four main findings as follows: 1) Image-derived emotion vectors form a low-dimensional geometry similar to that of text-derived vectors. Valence is relatively stable across sources, while arousal varies more. 2) Text- and image-derived emotion vectors have modest cosine similarity but still show held-out cross-modal correspondence. Text-derived vectors can also steer image interpretation. 3) Cross-architecture correspondence remains even when native cosine is near zero. Transformations estimated from generic ImageNet activations recover both correspondence and causal transfer without using the six emotion vectors or their labels. 4) After aligning representations across architectures, we construct a shared emotion subspace that preserves affective geometry and selective steering effects. The corresponding consensus emotion vectors also generalize to a held-out fourth architecture at two model sizes. These results suggest that emotion representations can share relational structure and causal effects across sources, modalities, and architectures, even when individual vector directions differ.
[CV-125] SurgGMF: Fully Causal Gaussian Motion Forecasting for Anticipatory Surgical Scene Rendering
链接: https://arxiv.org/abs/2609.34733
作者: Jingqian Sun,Yichao Tang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Dynamic surgical scene modeling is essential for robotic perception, simulation, and decision support. Although existing neural rendering methods enable efficient reconstruction and rendering of deformable surgical scenes, they remain primarily focused on observed-frame reconstruction rather than forecasting future scene states. To this end, we present SurgGMF, a fully causal Gaussian motion forecasting framework for anticipatory surgical scene rendering. Rather than predicting future RGB images directly, SurgGMF forecasts future Gaussian motion states represented by position, scale, and rotation residuals (X/S/R) from historical Gaussian motion fields. To prevent target leakage, we introduce a full-causal-last rendering protocol, where future Gaussian states are rendered without accessing target-frame Gaussian attributes while preserving causal appearance propagation. We evaluate SurgGMF on 12 EndoNeRF and StereoMIS video slices using neural temporal learners and classical dynamics baselines under a unified forecasting protocol. Learned Gaussian motion forecasting consistently outperforms classical dynamics baselines in render space, demonstrating gains beyond hand-crafted state extrapolation. Latency analysis further reveals an accuracy–efficiency trade-off: under the current implementations, TKAN achieves the highest accuracy, whereas GRU and LSTM provide more favorable module-level latency profiles. These results establish SurgGMF as a reproducible framework for causal Gaussian motion forecasting and advance surgical Gaussian representations from retrospective reconstruction toward predictive scene modeling.
[CV-126] What Visual Generators Need from Teachers: Rethinking Representation Alignment
链接: https://arxiv.org/abs/2609.34732
作者: Yongcong Wang,Hingchin Chen,Mingyu Fan,Shuo Jiang,Teer Zhang,Yucong Sun,Zijia Wang,Yiming Lu,Chengchao Shen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Representation alignment speeds up diffusion transformer training by pulling an intermediate block of the model (student) toward features of a frozen pretrained encoder (teacher). Which teacher layer to align, and for how long, is still set by convention, and each alternative costs a training run. We find that alignment helps where the student cannot linearly recover the teacher’s features, not where it already resembles them. Since a deep teacher layer is largely predictable from the one below, we isolate what each layer adds, its increment, and measure how much of it an unaligned student recovers. The student fills the teacher’s hierarchy from the bottom up and stalls near the top, which we call hierarchy filling: even after 400K steps it recovers almost none of the deepest. The recoverability gap is the unrecovered share of an increment, read from one unaligned checkpoint. In short runs that each align one teacher layer at one block, the gap nearly reproduces their ranking by FID improvement, and CKA, a measure of feature similarity, largely reverses it. Representation Alignment and Recoverability Estimation (RARE) picks the teacher layer with the largest gap before training. During training, it tracks each token’s remaining distance to that layer, the online counterpart of the gap, weights tokens by it, and phases out the loss once the average distance stops falling. With SiT-B/2 on ImageNet 256\times256 , RARE reaches an FID of 18.02 without guidance and 4.46 with it, ahead of seven alignment baselines including REPA, iREPA and HASTE. It also trains in 14% fewer GPU-hours than iREPA. Its FID stays below iREPA’s across model scales, teachers, datasets and backbones.
[CV-127] Geometry as Address: Routing Attention to Visual Memory for Long-Horizon Camera-Controlled Video Generation
链接: https://arxiv.org/abs/2609.34722
作者: Zesong Yang,Weikai Chen,Liyuan Cui,Lutao Jiang,Runze Zhang,Yingda Yin,Xiaoyang Huang,Kai Yan,Keyang Luo,Wangguandong Zheng,Xin Wang,Hujun Bao,Zhaopeng Cui
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:Long-horizon camera-controlled video generation requires recovering previously observed content from an ever-growing visual history. Existing approaches either search historical context implicitly or reconstruct it into persistent 3D memory, facing inefficient memory access or accumulated geometric errors. Our key insight is that geometry need not explain the scene–it only needs to determine where visual memory should be read from, while attention decides what should be recovered. Based on this insight, we introduce GEAR, a Geometry-Enabled Attention Routing framework that uses geometry as an explicit token-level address for visual memory. Rather than fusing historical observations into a persistent global 3D representation, GEAR retains them as frame latents and uses per-frame geometry only to establish token-level correspondences with target views, thereby avoiding persistent error accumulation from global fusion. Guided by these correspondences, Geometric Correspondence Attention (GCA) selectively injects geometrically matched historical features into noisy target patches during denoising. We further introduce an Invisible Octree to accumulate visibility evidence and reject geometrically plausible but occluded correspondences. Extensive experiments demonstrate that GEAR achieves state-of-the-art visual quality, precise camera control, and revisit consistency, enabling minute-long video generation along challenging trajectories.
[CV-128] DBCF: Dual-Branch Complementary Fusion of Foundation Models for Generalized Deepfake Detection
链接: https://arxiv.org/abs/2609.34720
作者: Fengming Gu,Mingjie He,Zonghui Guo,Jie Zhangb,Shiguang Shan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:As image generation and editing technologies have progressed substantially, facial forgeries pose significant challenges to privacy and public safety. Due to limited ability to capture forgery cues, existing small-scale forgery detection models often struggle to generalize across various domains and unseen manipulations. To address this limitation, researchers have turned to large-scale foundation models, which can provide richer representations and better generalization. Nevertheless, relying on a single foundation model alone remains insufficient for effective forgery detection. While models like CLIP offer robust global semantic cues, they lack the capacity to capture detailed local facial features. In contrast, DINO excels at capturing local structural features of faces, but provides weaker global semantic context. To fully utilize the synergies among multiple foundation models, we propose a hierarchical multi-granular framework that integrates complementary pretrained representations. Specifically, a Global Context Branch (GCB) based on CLIP captures holistic semantic cues, while a Fine-grained Cue Branch (FCB) built on DINOv3 captures localized structural irregularities. In addition, we design a feature fusion module that enables parameter-efficient adaptation of the frozen foundation backbones by adaptively extracting and integrating complementary features from the two models. By jointly leveraging global context and fine-grained cues, our method learns more comprehensive forgery representations and achieves strong cross-manipulation performance. Extensive experiments on multiple benchmarks demonstrate the benefit of the proposed design, particularly under cross-dataset and cross-manipulation settings.
[CV-129] From Pixel Generation to Topological Inference: Structural Dual Super-Resolution for Trustworthy Cross-Physical-Domain Trabecular Morphology Learning
链接: https://arxiv.org/abs/2609.34716
作者: Fan Zhang,Yi Zhang,Ling Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 19 pages,7 figures, conference
Abstract:Clinical CT and UHRCT cannot resolve individual trabeculae, whereas synchrotron radiation microCT (SR\muCT) provides 3.2\mum high-resolution references but is not applicable for in vivo imaging. The two domains differ by 31.25x in resolution, are only coarsely paired, and have drastically different data volumes. Moreover, clinical UHRCT suffers from severe partial volume effects, strong noise, and beam hardening/scatter artifacts, while SR\muCT is nearly free. Existing super-resolution networks and pretrained-prior methods underperform because they target pixel generation–diverse details and SSIM/PSNR–and do not explicitly model these physical differences. This indicates that 32x super-resolution via pixel generation is intrinsically ill-posed. We propose a paradigm shift from pixel generation to topological inference: deterministically predicting invariant microstructures from macro-scale low-resolution inputs, evaluated by morphological parameters. We realize this paradigm via structural dual super-resolution, coupling forward physical degradation (micro-to-macro) with inverse structural inference (macro-to-micro) through structural duality constraints. The method is an end-to-end, few-shot, compact structural dual network (SDN), comprising a bidirectional modeling network for forward degradation and inverse reconstruction, a pyramid structural consistency discriminator, and four structural duality constraints. On the testset, SDN achieves morphological parameters largely consistent with SR\muCT across 7 metrics, enabling clinical UHRCT with micro-imaging-level morphological quantification, with SSIM reaching 0.8. Trained on 3.2\mum SSRF data, the model generalizes well to 3.25\mum BSRF data from an independent source, validating cross-source generalization and confirming that the designed network achieves trustworthy structural inference rather than pixel generation.
[CV-130] riangular Resampling for Long-Horizon Motion Generation
链接: https://arxiv.org/abs/2609.34697
作者: Kunhang Li,Yiyi Cai,Xiangyue Zhang,Fangyuan Tu,Yuhan Wu,Zhixiang Wang,Kaipeng Zhang,Haiyang Liu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce Triangular Resampling (TR), a post-training method for mitigating long-horizon error accumulation in motion diffusion models. Built on FloodDiffusion’s triangular denoising schedule, TR addresses the mismatch between ground-truth-derived training windows and model-generated inference states. Replacing only completed motion history leaves this mismatch unresolved in partially denoised states within the active window. TR therefore extends rollout-based training to these states, using ground-truth clamping to limit excessive drift. For each replayed sample, TR draws one denoising threshold, shared across latent positions and replay updates, and replays multi-step triangular denoising without gradient tracking. After each update, states below the threshold are replaced with noise-matched ground truth, while those at or above it retain model predictions. The resulting latent window enters the standard training update. This rollout construction supports both supervised training (TR) and distribution matching (TR-DMD). On 120-second motion generation from HumanML3D test prompts, TR and TR-DMD achieve state-of-the-art FID AUC within their respective non-DMD and DMD comparison groups. Supervised TR reduces FID AUC by 40.9% and FID degradation slope by 55.3% relative to matched post-training without replay.
[CV-131] Unified Trajectory Matching Policy Optimization: Diverse T2I Generation and VLA Generalization
链接: https://arxiv.org/abs/2609.34688
作者: Zhiyuan Ma,Jiaming Li,Lingzhen Li,Yu Liu,Xuekai Zhu,Dingkang Liang,Kaiyan Zhang,Jianjun Li,Bowen Zhou,Xiang Bai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reward-maximizing reinforcement learning (RL) is widely used to post-train stochastic diffusion and flow policies for text-to-image (T2I) generation. However, reward-maximizing RL causes policy mode collapse even under reference KL or entropy regularization, reducing the policy to a single high-reward mode. In T2I, this produces similar images and reward hacking. When extended to vision-language-action (VLA) models, the same collapse removes alternative successful strategies and weakens task and scene generalization. To address this limitation, we introduce Unified Trajectory Matching Policy Optimization (Uni-TMPO), a unified RL post-training framework for diffusion and flow policies. First, Uni-TMPO converts standardized rewards into a target distribution within each trajectory group and derives the policy distribution from trajectory log probabilities. Then, forward Kullback-Leibler optimization matches the two distributions instead of maximizing expected reward. A progress-conditioned coarse-to-fine scheduler efficiently constructs T2I trajectories. Within the unified framework, feedback-conditioned sampling uses updated observations to construct VLA trajectories. Extensive experiments show that Uni-TMPO achieves higher T2I rewards and VLA ID success rates than the strongest baselines. More importantly, it achieves the best T2I reward-diversity-efficiency trade-off and VLA generalization to held-out tasks and scenes, while real-robot evaluation demonstrates the value of multiple action strategies when the higher-reward target is blocked.
[CV-132] Natural State-Prediction Accuracy can Hide Weak Controlled Responsiveness in VLA Readouts
链接: https://arxiv.org/abs/2609.34684
作者: Hyungjoon Kim,Wonbin Son,Mi Young Lee,Jun Young Lee,Seungmin Rho
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Accurately decoding object states from the internal representations of vision-language-action (VLA) models does not establish that the predictions respond faithfully to changes in the target physical state. In natural observations, object state, robot configuration, occlusion, and task progress vary together, allowing contextual cues to contribute to prediction. In this paper, we introduce an evaluation framework that separates prediction accuracy, target-state responsiveness, and context stability using physically validated observations that cross target coordinates with robot contexts. We demonstrate that high natural-trajectory accuracy can coexist with weak controlled target-state responsiveness in fixed representation-readout pairs. Comparisons and interventions involving representations, readouts, and training data show that the three properties provide distinct diagnostic information. Furthermore, adding responsiveness and context sensitivity to a failure predictor based on initial state error and physical variables reduces policy-failure prediction error on new initializations relative to the specified baseline while same-observation controlled MAE is also informative. These findings motivate evaluating target-state responsiveness and context stability alongside natural prediction accuracy, and examining their relationship to actual policy behavior and task outcomes.
[CV-133] V-Gym: Enhancing Agent ic Visual Reasoning via Skill-Data Co-Evolution
链接: https://arxiv.org/abs/2609.34682
作者: Bei Yan,Yuecong Min,Jie Zhang,Junqi Yang,Shiguang Shan,Xilin Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Advances in multimodal understanding, reasoning, and tool use enable agents to tackle increasingly complex visual reasoning tasks. By distilling past execution experience into reusable skills, agents can transfer lessons from both successes and failures into future reasoning, reducing repeated errors and improving capabilities. However, limited experience may produce unreliable, poorly generalizable skills, while static datasets may lack the targeted and diverse practice needed for refinement. To address this gap, we introduce V-Gym, an autonomous framework that iteratively co-evolves procedural skills and multimodal practice data from execution trajectories. During skill evolution, V-Gym analyzes trajectories to distill and refine hierarchical skills, updating procedural guidance and applicability conditions while retaining an update only if it improves validation performance. During data evolution, V-Gym selects generation seeds by balancing data utility and exploration, then translates trajectory-identified bottlenecks into diverse, targeted practice data that expand the data bank after quality checks. The resulting practice outcomes feed back into subsequent skill updates, closing the loop for continual skill refinement. Experiments across diverse multimodal reasoning benchmarks show substantial improvements over baselines with multiple backbone models. Its evolved skills generalize across domains and models, while evolved data support more effective skill refinement, enabling autonomous diagnosis, targeted practice, and continual self-improvement.
[CV-134] Learning What to Recall: Adaptive Multi-Cue Episodic Memory for World Models
链接: https://arxiv.org/abs/2609.34677
作者: Beomsu Kim,Chieh-Hsin Lai,Bac Nguyen,Amir Bar,Jong Chul Ye,Yuki Mitsufuji
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint, Project Page: this https URL
Abstract:World models predict future observations from current experience and actions, yet prediction can depend on observations seen far in the past. Episodic memory preserves past observations for later recall; however, as memory accumulates, it raises a fundamental question: which memories are useful for the current prediction, and which available retrieval cues should be trusted to find them? This is challenging because fixed criteria based on recency, pose overlap, or visual similarity can be unreliable across environments and queries. We propose Future-Aware Recall (FAR), a framework that learns episodic recall from future-aware predictive supervision and adaptive multi-cue scoring. During training, FAR measures predictive utility by the conditional log-likelihood of the realized future given recalled context, approximated by negative diffusion prediction loss, and uses it to train a retriever that remains future-blind at inference. The retriever learns cue-specific relevance and automatically determines which available retrieval cues, such as time, pose, vision, and audio, to trust for each query when selecting memories. Across three complementary settings, FAR outperforms hand-designed recall even with the same retrieval cues, automatically adapts which available cues to trust, and recalls the right history as the world changes. Together, these results establish FAR as a flexible, principled approach to episodic memory access in world models.
[CV-135] Evidence-Aligned Multimodal On-Policy Self-Distillation for Fine-Grained Visual Understanding
链接: https://arxiv.org/abs/2609.34672
作者: Nanxing Hu,Qiwei Yan,Jinchao Zhang,Guoliang Kang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Fine-grained visual understanding requires models to recognize small details within complex images. Multimodal on-policy self-distillation (OPSD) addresses this challenge by using a teacher conditioned on evidence-centered crops to supervise a student conditioned on original images along student-generated trajectories. Ideally, teacher corrections, the distributional changes from the student toward the privileged teacher, should be driven by task-relevant visual evidence. However, the designs that make the teacher effective also introduce other interference. Using a lagged or frozen teacher improves training stability but introduces a model-state gap from the evolving student, while cropping enhances task-relevant evidence but also loses the visual context. These two sources of interference make the teacher corrections not purely rely on the visual evidence. We introduce Evidence-Aligned multimodal on-policy self-Distillation (EAD), which retains the crop-conditioned teacher as the target but constructs a separate evidence reference for weighting the corrections. To exclude the effect of lagged model-state from this reference, EAD measures prediction changes using the current student. To avoid crop-induced context changes, EAD masks the evidence region in the original image while preserving the other visual context. The change from the student’s masked-image prediction to its original-image prediction provides a controlled reference for the direction in which the visual evidence shifts the student’s prediction. EAD weights each teacher correction by its cosine alignment with the reference, i.e., retaining aligned corrections and downweighting the rest. Retaining only 6% of the supervision mass of dense OPSD, EAD consistently outperforms previous state-of-the-art methods.
[CV-136] CapField-OPD: Learning Continuous Capability Fields via Joint-Anchored Multi-Teacher On-Policy Distillation for Flow Models
链接: https://arxiv.org/abs/2609.34658
作者: Pengyang Ling,Yujie Zhou,Jiazi Bu,Yibin Wang,Xiaoxiao Ma,Yi Jin,Huaian Chen,Yuhang Zang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages, 8 figures
Abstract:Reward-specialized post-training produces strong experts for flow-based generative models, while multi-teacher on-policy distillation (OPD) consolidates their capabilities into a single student. Existing methods, however, route each prompt to a single teacher according to its semantic category, implicitly binding the desired capability to prompt content. This coupling makes capability invocation vulnerable to prompt perturbations and prevents users from explicitly adjusting the strength of the desired capability at inference time. In this work, we introduce CapField-OPD, an OPD framework that integrates multiple teachers into a continuous capability field through explicit capability coordinates. We use teacher models as anchors to construct this field, with the coordinates determining how their outputs are combined. Each capability configuration thus receives a unique supervision target, and capability control no longer depends on prompt semantics. Since the training anchors may not be optimal at inference time, we further profile the learned field on a small calibration set. The coordinate with the highest mean reward serves as the recommended default, while coordinates that are frequently optimal offer a promising candidate set for test-time scaling. Extensive experiments on compositional generation, text rendering, and visual aesthetics demonstrate that CapField-OPD consolidates multiple specialized teachers into a single student while preserving or surpassing their performance, reliably invokes the desired capabilities under semantics-preserving prompt variations, and supports continuous capability control and coordinate-based test-time scaling.
[CV-137] Revitalizing Medical Time Series with Vision-Informed Retrieval: A Vision-Language Perspective NEURIPS2026
链接: https://arxiv.org/abs/2609.34652
作者: Guoqi Yu,Juncheng Wang,Shujun Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by NeurIPS 2026
Abstract:Medical time series (MedTS) underpin many clinical classification tasks, yet existing methods usually represent them only as numerical sequences and underuse the morphology that is explicit in waveform inspection. To bridge this gap, we introduce Vision-Informed Retrieval (ViRe), which uses a frozen VLM-derived waveform representation as a morphology-aware Query to guide retrieval from raw numerical MedTS features. Specifically, a Vision Query is extracted using pre-trained vision-language models (VLMs) to obtain morphology-aware priors from waveform plots. A tailored attention-based cross-modal retrieval mechanism then uses the Vision Query to select morphology-relevant temporal and channel evidence from the numerical representation. ViRe demonstrates strong effectiveness against ten established baselines, yielding an overall 6.42% relative improvement over the previous state of the art across six public benchmarks. Code, training scripts, and reproducibility materials are publicly available in the GitHub Repository: this https URL.
[CV-138] DirectUV: Image-Conditioned UV Texture Generation with Surface-Aware Positional Encoding NEURIPS2026
链接: https://arxiv.org/abs/2609.34651
作者: Jiantao Lin,Yingjie Xu,Mingzhi Sheng,Yangkai Wei,Hao Chen,Ying-Cong Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at NeurIPS 2026
Abstract:Generating high-quality UV textures for 3D meshes remains challenging. Multi-view projection pipelines suffer from occlusion and view inconsistency, and recent methods that generate textures directly in UV space still rely on auxiliary modules to supply 3D information, leaving the attention mechanism tied to UV-grid positions rather than to the underlying surface geometry. This mismatch limits coherence across seams and disconnected UV islands. We propose DirectUV, an image-conditioned UV texture diffusion framework that operates in the latent UV space of a pretrained image VAE, in which a Diffusion Transformer denoises the UV latent given a single input image and a coarse UV map. At its core, Surface-Aware Positional Encoding (SAPE) replaces the standard 2D-grid positional encoding with encodings derived from per-token 3D surface coordinates obtained via UV-to-surface correspondence. As positional encodings define the distance metric used by attention, SAPE enables tokens to interact according to 3D positional proximity derived from surface correspondence rather than UV-grid distance, restoring coherence across seams and disconnected islands. A multi-level extension further assigns different attention heads to progressively finer subdivisions of the same latent UV patch, allowing the model to reason about surface structure at multiple granularities. Experiments show that DirectUV produces sharper and more globally consistent textures than other baselines, with the largest improvements in occluded and view-unseen regions where projection-based methods leave gaps or stretched textures. Comments: Accepted at NeurIPS 2026 Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2609.34651 [cs.CV] (or arXiv:2609.34651v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.34651 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-139] Backdoor as Probe: Test-Time Adversarial Defense for CLIP
链接: https://arxiv.org/abs/2609.34641
作者: Zhongqi Wang,Jie Zhang,Nie Sen,Zhiyu Chen,Shiguang Shan,Xilin Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Test-time adversarial defense improves the robustness of vision-language foundation models such as CLIP without retraining. However, adversarial activation shifts are typically treated as distortions to suppress, rather than signals to exploit. We turn these shifts into defense signals by repurposing the trigger-to-target mechanism of backdoors. The key is to implant a defender-controlled backdoor as a probe that is weakly activated by clean inputs but strongly activated by adversarial shifts. Based on this insight, we propose \emphBackdoor as Probe (BaP), a test-time adversarial defense for CLIP. BaP constructs the probe through a closed-form model edit to a selected MLP layer. It projects the average adversarial activation shift and a defender-specified semantic direction onto the layer’s low-energy input and output activation subspaces to obtain the trigger and target directions, respectively. At inference time, adversarial inputs produce measurable responses along the target direction for detection. BaP then selectively rectifies detected inputs by optimizing a small perturbation that steers their representations away from adversarial shifts and toward the clean subspace. Experiments across 16 benchmarks show that BaP improves average robust accuracy from 1.0% to 52.3% while retaining clean accuracy, achieving performance comparable to state-of-the-art methods with up to a (5.7\times) inference speedup. BaP further shows the generalization to adversarial attacks on large vision-language models. Project page: this https URL
[CV-140] Long Time No See: Benchmarking VLMs for Out-of-Sight Spatiotemporal Reasoning in Egocentric Videos
链接: https://arxiv.org/abs/2609.34630
作者: Fangzhou Ma,Ivo Alexander Ban,Eren Homburg,Gabriele Goletto,Rémi Pautrat,Mahdi Rad,Chiara Plizzari,Marc Pollefeys
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Real-world AI systems must reason about objects that are no longer visible: an AR assistant guiding a user back to an object used earlier, a household robot retrieving an item someone put away. This requires not just recalling where an object was last seen, but updating its state when it is moved and retaining that update once it leaves view. We refer to this as out-of-sight spatiotemporal reasoning. We introduce Beyond3D, the first VQA benchmark to isolate this ability in dynamic egocentric video: every query targets an object that has been relocated and has since left the field of view. We create our questions from HD-EPIC annotations, building a visibility track for each dynamic object from its 3D position, the camera pose, and the scene geometry to understand at each moment whether it is visible, occluded, or out of view. Beyond3D comprises 9,000 questions in eight types over 135 videos from nine participants, organized as one reasoning chain: visual grounding (is the target observable now), temporal grounding (when it was last visible and last placed), scene localization (which fixture anchors that location), and 3D spatial perception (where it lies relative to the current viewpoint or another object in the scene). We benchmark nine general-purpose and spatially specialized VLMs. The best model reaches 42.2% against 29.7% chance and text-only baselines reaching 31.9%, with the largest failures in recovering when an object was last visible, showing that tracking object movement out of sight remains far from solved for current VLMs.
[CV-141] BMND: Direct Poisson Denoising by N-Dimensional Block Matching and Collaborative Filtering
链接: https://arxiv.org/abs/2609.34622
作者: Christof Duhme,Lars Schiefelbein,Florian Büther,Xiaoyi Jiang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Poisson denoising of scientific data requires methods that account for signal-dependent noise while accommodating different data dimensionalities and preserving quantitative intensity information. We present BMND, a dimension-independent extension of block matching and collaborative filtering for Gaussian and Poisson observations. Building on the two-stage structure of BM3D and BM4D, BMND processes Poisson data directly, without a variance-stabilizing transform, by combining noise-aware patch matching with propagation of signal-dependent noise variances through collaborative filtering and aggregation. A dimension-independent reference-patch traversal scheme supports arrays with an arbitrary number of axes. An optional aggregation-aware mass conservation preserves the observed total intensity after weighted overlap-add. We evaluate the framework on one-dimensional physiological signals, two-dimensional images, and three-dimensional volumes, using controlled noise experiments and measured fluorescence microscopy acquisitions. The experiments demonstrate improved reconstruction quality from noise-aware matching and Wiener filtering, while low-count phantom experiments show reduced denoising-induced intensity loss through mass conservation. The framework provides a unified, non-learning-based approach to denoising across arbitrary data dimensions and is released as an open-source library.
[CV-142] Does Native 3D Texture Generation Necessarily Require 3D Assets for Training?
链接: https://arxiv.org/abs/2609.34621
作者: Jiangshan Wang,Zeqiang Lai,Jiayi Guo,Xin Yang,Xin Huang,Jiarui Chen,Ziheng Ouyang,Chunchao Guo,Xiangyu Yue
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:Native 3D texture generation synthesizes colors directly in 3D space for a given geometry, conditioned on multi-view reference images. It is generally believed that training such models requires large-scale, high-quality real 3D asset data, whose acquisition remains a long-standing and challenging problem. In this work, we propose Tex-Zero, demonstrating that a high-fidelity native 3D texture generation framework can be trained without 3D assets. Our key observation is that only high-quality and fine-grained color information is essential for 3D texture training, while the required geometric information is less critical and can be manually constructed rather than obtained from real 3D assets. This finding makes it possible to transform abundant, high-quality 2D images into effective training samples for 3D texture generation. Specifically, we convert high-quality 2D images into 3D training samples by representing each image as a plane in 3D space and applying patch-wise random rotations and aggregation to construct complex geometric structures. Using these constructed image data, we train the Tex-Zero VAE, which can reconstruct real 3D assets with high quality despite never observing them during training. Building upon the Tex-Zero VAE, we train the Tex-Zero DiT also exclusively on the constructed image data, where the conditioning 2D multi-view images are transformed into planes in 3D space and also encoded by the Tex-Zero VAE, thereby reducing the representation gap and improving generation quality. Extensive experiments show that Tex-Zero generates high-fidelity 3D textures with fine-grained details solely using images as training data, offering a promising perspective on the data paradigm for scaling 3D texture generation.
[CV-143] WorldAttention: An Efficient Attention Architecture for Interactive Video World Models
链接: https://arxiv.org/abs/2609.34606
作者: Zeyu Zhang,Jinyuan Mao,Dakai An,Wangbo Zhao,Hanfeng Lu,Jiasheng Tang,Yinghao Yu,Wei Wang,Bohan Zhuang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Website: this https URL , Code: this https URL
Abstract:Leveraging the paradigm of autoregressive diffusion, text-conditioned interactive video world models aim to simulate temporally coherent environments guided by textual instructions. While enabling low-latency, long-duration generation is pivotal for embodied AI and simulation-based planning, current frameworks primarily rely on sliding-window mechanisms to bound computational complexity. However, this approach inherently sacrifices historical context, undermining the long-range interactive capabilities. Conversely, maintaining a full-history cache remains computationally prohibitive and memory-intensive: the quadratic complexity of attention leads to excessive computational overhead, while the linear growth of the KV cache inevitably leads to GPU memory saturation. To overcome these limitations, we propose WorldAttention, a system-oriented attention architecture that achieves high efficiency through the co-design of specialized attention kernels and hierarchical KV cache management. First, we introduce Hybrid Sparse Attention (HSA), which integrates linear global attention supplemented with head-adaptive sparse attention. Additionally, we design a Hierarchical KV Cache (HKV) that organizes historical KV pairs into semantically indexed pages across multi-tier memory, enabling fine-grained retrieval and controlled GPU residency. These two designs are supported by tailored kernels to effectively translate their theoretical efficiency into real-world performance. Extensive experiments on VBench-Long and InterVBench demonstrate that WorldAttention consistently surpasses prior state-of-the-art methods, achieving subject consistency scores of 0.9472 on VBench-Long and 0.9668 on InterVBench, respectively.
[CV-144] Summarize Before Grounding: Query-Guided Chunk Condensation for Long-Video Temporal Grounding
链接: https://arxiv.org/abs/2609.34598
作者: Nanxing Hu,Xiaoyue Duan,Qiwei Yan,Kailin Lyu,Jinchao Zhang,Guoliang Kang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Video temporal grounding (VTG) aims to localize the video interval corresponding to a language query. Recent large vision-language models (LVLMs) show great potential in solving such a multi-modal reasoning task. However, long videos often contain large amounts of redundant information that disturbs LVLMs to mine query-relevant evidence. Instead of dense frame sampling which incurs prohibitive training memory, previous reinforcement learning with verifiable rewards (RLVR) works typically utilize sparse sampling, which makes training feasible but may miss critical evidence. In this paper, we propose a summarize before grounding'' framework (named SumGround’') for long-video temporal grounding. The key of SumGround is to perform query-guided chunk condensation to aggregate and retrieve query-relevant evidence. Specifically, we split the video into several chunks and perform two-level chunk condensation. First, we introduce query-guided latent summaries, which is represented as KV states of query-guided prompts, to compress redundant visual tokens into compact query-relevant chunk summaries. Furthermore, we design an associative summary retrieval scheme to rank and select chunk summaries that are most likely to contain the event interval. Both query-guided latent summary and associative summary retrieval schemes are enabled by RLVR. To reduce memory consumption, we propose a length-aware gradient gating module to selectively stop gradient back-propagated to visual tokens. Extensive experiments demonstrate that SumGround performs favorably against previous state-of-the-art methods across multiple downstream datasets, with remarkable gains on long videos.
[CV-145] mporal Modelling for Burn Scars on Sentinel-3
链接: https://arxiv.org/abs/2609.34596
作者: Luca Barco,Edoardo Arnaudo,Andrea Bragagnolo,Claudio Rossi,Paolo Garza
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Rapid and accurate burn scar delineation from satellite imagery is essential for post-fire damage assessment. Sentinel-3 OLCI, with daily revisit and 21 spectral bands, suits rapid mapping, yet most pipelines treat acquisitions independently, leaving the pre/post-fire change signal unexploited. We present a dataset of 246 wildfire activations (2016-2025) from the Copernicus Emergency Management Service, with Sentinel-3 OLCI temporally paired acquisitions. We benchmark spatial and temporal (ConvLSTM-augmented) variants of three backbones (U-Net, SegFormer, ConvNeXt-UPerNet) under two input modes and spectral configurations. Temporal modeling improves segmentation only when pre-fire frames are included, and a 5-band subset matches the full 21-band OLCI configuration.
[CV-146] Reinforcement Learning from Intermediate Renders for Image-to-Code Generation
链接: https://arxiv.org/abs/2609.34587
作者: Omri Kaduri,Kate Feingold,Phillip Isola,Tali Dekel
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project page: this https URL
Abstract:Reinforcement learning is increasingly used to post-train vision-language models for image-to-code generation, such as generating SVG code from a reference image, by optimizing rewards computed from the final rendered output. However, relying on a single terminal reward provides sparse feedback that is poorly aligned with the contribution of individual tokens. A generated program may contain operations that accurately reproduce some parts of the target image alongside others that introduce errors, yet all tokens are trained from the same final outcome. We observe that many intermediate code prefixes are not only executable, but already produce meaningful partial renders that reflect progress toward the target. This property provides a natural source of denser supervision during generation. Based on this observation, we introduce IR4RL, an RL framework with a token-level render-progress reward that turns changes between intermediate renders into localized feedback for the generated sequence. We evaluate our approach on Image-to-SVG and Image-to-TikZ generation. Across both tasks, our method improves over supervised fine-tuning and standard GRPO, yielding new state-of-the-art open-source models. This shows that intermediate rendering provides a simple and effective source of process supervision for RL post-training of image-to-code models.
[CV-147] Counterfactual Attention Policy Distillation for Temporal Video Grounding
链接: https://arxiv.org/abs/2609.34581
作者: Shaobo Ju,Haiyang Yu,Xuecheng Wu,Qiong Wu,Jiacong Wang,Fan Shi,Peng Jun,Yiyi Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Temporal video grounding is a key capability of advanced \emphMultimodal Large Language Models (MLLMs) for the thorough understanding of video events, which is however often limited by repeated actions and visually similar contexts in long videos. In this paper, we study this issue from the perspective of \emphOn-policy distillation (OPD) and propose a new training regime for MLLMs termed \emphCounterfactual Attention Policy Distillation (CAPD). In particular, OPD is a viable solution for MLLMs via providing dense teacher supervision on student-generated trajectories. But its next-token based teacher-student distillation is hard to identify the specific video segments supporting each predicted timestamp, which is critical for temporal grounding. In this case, CAPD measures how masking each temporal group changes the teacher’s output distribution. The resulting counterfactual influence calibrates the teacher’s attention and weights token-level distillation, allowing the student to learn the temporal evidence that affects boundary prediction. To validate CAPD, we trained it on Qwen3-VL-8B-Instruct using only 2,500 samples for one epoch, and evaluated it on the TimeLens and multiple general video benchmarks. Experimental results show that CAPD improves average recall by 12.0% relative to GRPO on TimeLens while preserving general video understanding, achieving comparable accuracy to the base model.
[CV-148] GenNVS: Geometry-enhanced Novel View Synthesis via Disentangled 3D Prior
链接: https://arxiv.org/abs/2609.34579
作者: Yajiao Xiong,Youyu Luan,Xiaoyu Zhou,Yongtao Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Single-image novel view synthesis remains challenging because the underlying 3D geometry is highly ambiguous. Recent diffusion-based approaches produce plausible results, but they often struggle to preserve the geometric structure and spatial coherence of foreground objects. We present GenNVS, a framework for geometry-enhanced novel view synthesis via a disentangled 3D prior. Specifically, GenNVS models foreground objects and the background with 3D Gaussian Splatting and aligns them through a coarse-to-fine geometric optimization process to form a unified 3D scene. This scene conditions a video diffusion model through the proposed Dual-Stream Masking mechanism, which guides synthesis by jointly exploiting rendered validity masks and geometry-aware warping. Experimental results show that GenNVS performs favorably against recent methods in both visual quality and geometric accuracy, while naturally supporting flexible scene editing.
[CV-149] Recent Advances in Agent ic Agri-Robotic Phenotyping: A Perspective Review from Frag mented Multimodal Sensing to Unified PhenoAgent Intelligence
链接: https://arxiv.org/abs/2609.34567
作者: Muhammad Owais,Ehtesham Iqbal,Samee Ullah Khan,Muhammad Umraiz,Yusra Abdulrahman,Irfan Hussain
类目: Computer Vision and Pattern Recognition (cs.CV); Emerging Technologies (cs.ET)
备注:
Abstract:This review examines the evolution of plant phenotyping from conventional manual trait measurement to high-throughput, robotic, and artificial intelligence-driven crop monitoring. Despite significant advances in imaging, autonomous platforms, multimodal sensing, and deep learning, current phenotyping systems remain fragmented across sensing modalities, crop traits, growth stages, environments, and management objectives. We therefore frame phenotyping as an integrated \emphseed-soil-plant-environment-management (SSPEM) intelligence problem, where crop performance reflects interactions among seed quality, root-zone conditions, plant development, environmental exposure, and management actions. The review synthesizes conventional, high-throughput, robotic, and AI-driven phenotyping approaches, highlighting their capabilities and persistent limitations in temporal integration, multimodal reasoning, biological interpretation, and actionable decision support. Building on this analysis, we introduce a conceptual PhenoAgent framework that extends phenotyping beyond the estimation of isolated traits to evidence-based crop-state interpretation, uncertainty-aware reasoning, and management-oriented support. The PhenoAgent concept primarily brings together scattered advances in phenotyping to deliver insights ranging from detailed to high-level, such as what is happening in the crop, why it might be occurring, what evidence is missing, and what actions or additional measurements should be considered. We also discuss challenges in dataset scarcity, annotation, benchmarking, model generalization, and explainability. By linking multimodal phenotyping with agentic AI and closed-loop decision support, this review outlines a path to interpretable, scalable, and deployment-oriented crop intelligence.
[CV-150] GLF-Q: Global-Local Feature-based Quantization for Vision Transformers
链接: https://arxiv.org/abs/2609.34564
作者: Peilin Sun,Guang Liang,Jin Tong,Jianxin Wu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Post-training quantization (PTQ) efficiently compresses Vision Transformers (ViTs) without retraining, yet suffers severe accuracy degradation at low bit-widths. Existing optimization-based PTQ methods guide block reconstruction via either soft logits or second-order Hessian proxies. Logit supervision is prone to overfitting on limited calibration data, while Hessian approximations incur structural truncation errors. To address these limitations, we propose \textbfGLF-Q, a novel PTQ framework guided by Global-Local Feature alignment. GLF-Q propagates quantized block outputs through downstream full-precision layers to align penultimate-layer representations under local output regularization, providing downstream feature supervision without explicitly approximating the Hessian or using a Taylor expansion. Furthermore, offline Hadamard transformations are introduced with zero runtime overhead to disperse activation outliers across channels, effectively contracting dynamic ranges and reducing quantization errors. Meanwhile, optimizing this loss via a Straight-Through Estimator (STE) achieves rapid convergence, bypassing continuous relaxation rounding formulations such as AdaRound. Extensive experiments across representative ViT architectures demonstrate that GLF-Q with standard uniform quantizers substantially outperforms state-of-the-art methods under 3-bit quantization on image classification. In addition, GLF-Q exhibits strong out-of-domain calibration robustness and achieves speedups under 8-bit GPU deployment.
[CV-151] ACPruner: Visual Token Pruning as Biased Attention Coverag e Maximization in LVLMs
链接: https://arxiv.org/abs/2609.34558
作者: Xu Li,Yuxuan Liang,Yi Zheng,Zhe Liu,Xiaolei Chen,Haotian Chen,Rui Zhu,Fan Shi,Xiangyang Xue
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Large Vision-Language Models (LVLMs) face significant computational inefficiencies caused by the large number of visual tokens. Existing visual token pruning methods mainly focus on either retaining individually important tokens or selecting mutually diverse ones. In this work, we revisit visual token pruning from a coverage perspective and formulate it as a biased attention coverage maximization problem. The key idea is to select a compact token subset whose encoder-side outgoing attention can jointly cover the image while assigning higher coverage priority to more informative regions. From this perspective, we propose ACPruner, a training-free visual token pruning framework for efficient LVLM inference. ACPruner first estimates token importance by combining intra-modal saliency and inter-modal relevance, then derives token-wise coverage from attention patterns within the vision encoder, and finally performs greedy selection to maximize the proposed coverage objective. Extensive experiments across multiple LVLM backbones, including LLaVA-1.5-7B/13B, LLaVA-NeXT-7B/13B, Qwen2.5-VL-7B, and LLaVA-OneVision-7B, show that ACPruner consistently achieves strong performance retention while delivering substantial end-to-end inference speedups.
[CV-152] SGate: Timestep-Aware Gated Attention for Diffusion Transformers
链接: https://arxiv.org/abs/2609.34539
作者: Boyu Zhang,Yifan Liu,Shuxia Lin,Qingjian Ni,Yinfei Xu,Xu Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Diffusion Transformers (DiTs) have emerged as the dominant architecture for high-fidelity image and video generation. Recent DiT systems increasingly use structured prompts for training, improving caption quality and prompt adherence. However, their generation quality can degrade severely under out-of-domain (OOD) prompts, including the free-form descriptions supplied by users at inference time. Although LLM-based rewriting can convert these prompts into structured formats, it does not guarantee that the rewritten prompts align with the training distribution. Our analysis links this degradation to attention sinks and reduced early-step image-to-text attention and shows that sink suppression alone is insufficient to restore generation quality. Despite effective sink suppression, models trained with standard gated attention exhibit reduced early-step image-to-text attention and suboptimal generation quality. Based on these insights, we propose Timestep-Aware Gated Attention (TSGate), which injects a timestep-conditioned bias into the gate signal so that gating behavior adapts across denoising steps. Extensive experiments show that TSGate consistently outperforms both the baseline and standard gated attention across multiple benchmarks, improving the raw-prompt DPG score by 9.5% over the baseline.
[CV-153] Preference-Guided Adaptation for Open-Vocabulary Semantic Segmentation via Prompt Disagreement NEURIPS2026
链接: https://arxiv.org/abs/2609.34528
作者: Hyun-Kurl Jang,Jihun Kim,Kuk-Jin Yoon
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to NeurIPS 2026
Abstract:Open-vocabulary semantic segmentation (OVSS) enables pixel-level prediction over arbitrary text-specified vocabularies and has shown strong generalization on common benchmarks. However, OVSS performance often degrades in specialized domains such as medical imaging, remote sensing, and industrial inspection, where dense pixel-level masks for adaptation are costly to obtain and require domain-specific expertise. We propose a preference-guided adaptation framework that replaces dense mask supervision with binary preferences. We observe that different prompt templates produce systematically different segmentations for the same image, a phenomenon we call prompt disagreement, and we repurpose it as a built-in source of preference supervision. Building on this, we mine localized preference queries from regions of high cross-template uncertainty, and adapt the OVSS model with Region-Localized Preference Optimization (RLPO) together with consistency regularization that stabilizes updates outside the queried region. Across extensive experiments on the MESS benchmark, the proposed method achieves consistent gains across diverse OVSS backbones without any pixel-level annotation, and remains effective under noisy preferences. Our code is available at this https URL.
[CV-154] RRG-SLAM: Real-time Reflection-aware Gaussian SLAM for Indoor Scenes
链接: https://arxiv.org/abs/2609.34527
作者: Yong Liu,Keyang Ye,Zhexi Peng,Ruixian Mei,Kun Zhou,Tianjia Shao
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注:
Abstract:We introduce the first real-time reflection-aware Gaussian SLAM system for indoor scenes. The system features a reflection-aware TSDF-Gaussian hybrid representation that explicitly separates diffuse scene appearance from reflection components. The base scene is modeled by a TSDF volume and a set of base Gaussians capturing geometry and diffuse appearance, while planar reflections are represented by reflection Gaussian groups associated with detected reflective planes. The rendering is performed in three passes: TSDF raycasting first yields surface color, depth, plane IDs and reflection masks; base Gaussians are then rendered order-independently with depth culling and combined with the TSDF output to form the base image; finally, under the guidance of the plane ID map, reflection Gaussians from different reflection groups are rasterized only into their corresponding planar regions to generate the reflection image, which is subsequently composited with the base image via the reflection mask to produce the final output. For online reconstruction, our system first estimates the camera pose through reflection-aware tracking to suppress interference of reflection-dominated regions. It then identifies reflective planes using geometric, semantic, and temporal cues, and fuses the observations into the augmented TSDF volume with reflection-aware attributes. Afterwards the base and reflection Gaussians are initialized, optimized, and pruned online to maintain both reconstruction quality and efficiency. Experiments on a variety of datasets show that our method outperforms existing SLAM systems in reconstruction quality, tracking robustness, and novel-view rendering for indoor environments with reflections, while preserving real-time performance.
[CV-155] SAGE: Subspace Alignment for Classifier-Free Guidance in Mixture-of-Experts Diffusion Models
链接: https://arxiv.org/abs/2609.34525
作者: Boyu Zhang,Yangming Cheng,Ning Zhang,Pengfei Liu,Weijie Li,Yifan Gao,Hangyu Li,Litong Gong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Diffusion Transformers with Mixture-of-Experts (MoE) routing are a leading recipe for scaling generative models. Classifier-Free Guidance (CFG) is essential for generation quality, yet excessively high guidance scales trigger collapse. We identify a previously unreported failure mode in their combination: the two CFG branches route independently, so their realized activations occupy different subspaces. The unconditional write then leaves the conditional subspace, and CFG amplifies that residual linearly in the guidance scale. We propose SAGE, a training-time regularizer that aligns unconditional MoE activations to the conditional subspace without restricting routing diversity, at zero inference cost. Toy experiments show that SAGE dramatically suppresses extreme drift by 9.2x. When scaled to a 1B-parameter text-to-image model, SAGE significantly improves generation quality, delivering a 9.3% boost in peak DPG-Bench performance. Extensive experiments demonstrate that SAGE consistently outperforms the baseline.
[CV-156] Evidence Before Accuracy: A MRI-PET Fusion Network for Alzheimer Disease Classification with Causal Regional Validation
链接: https://arxiv.org/abs/2609.34520
作者: Saeid Firouzi Daghigh,Saeed Ayat
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Deep learning models for Alzheimer disease (AD) classification routinely report near-perfect discrimination, yet few are shown to rest on AD-relevant neurobiology rather than on dataset artifacts, subject-level leakage, or non-brain image content. We present a fusion network combining T1 MRI and FDG PET across axial, coronal, and sagittal planes, trained on ADNI consists of 554 paired subjects. The fusion model reaches AUC 0.962, accuracy 0.909, and F1 0.891, competitive with recent 3D CNN and multimodal transformer systems at substantially lower cost. We first quantify how much modality, plane and slice geometry matter. A validation-only search over slice centres and neighbour spacings moves AUC by 0.180 for MRI and 0.078 for PET, selecting narrow spacing for MRI and wide spacing for PET, with the chosen coronal centres falling on the hippocampal body and on the posterior cingulate respectively. The contribution, however, is the evidence layer built around that number. Shortcut controls collapse the model to AUC 0.622 (silhouette), 0.608 (exterior), and 0.500 (blank), and a label-permutation null yields 0.456. Forward region-of-interest (ROI) ablation shows that masking medial temporal cortex in MRI and the posterior default-mode network (DMN) in PET produces the largest shift in the AD logit, while area-matched controls remain indistinguishable from that null. Reverse ROI ablation shows that the medial temporal lobe alone retains 89.2% of above-chance discrimination in MRI and the posterior DMN alone retains 79.0% in PET. A quantitative comparison of attribution methods shows occlusion sensitivity reaching 3.5-5.0* enrichment inside a priori AD regions against 0.10-0.43* in controls. Ablation and attribution independently establish a biologically correct double dissociation: hippocampal evidence is carried by MRI, posterior cingulate evidence by PET.
[CV-157] XFlow: A Workflow Model for Instruction-Guided Lesion Segmentation in Chest X-rays
链接: https://arxiv.org/abs/2609.34513
作者: Geon Choi,Hangyul Yoon,Hyunki Park,Sang Hoon Seo,Edward Choi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Existing text-guided segmentation models in the medical domain cover only a narrow set of anatomical structures and lesions in chest X-rays (CXRs), and most of them assume that the queried target is always present in the image. Instruction-guided lesion segmentation (ILS) was introduced to overcome these limitations by segmenting diverse lesion types from simple user instructions while also recognizing when the queried lesion is absent, and ROSALIA was proposed as the first model for this task. However, the masks produced by ROSALIA remain of limited quality, often carrying scattered noise. Moreover, ROSALIA predicts the mask in a single shot, which differs fundamentally from how radiologists perceive and delineate lesions in practice. A radiologist first surveys the entire thorax, then localizes the approximate region of abnormality, and only then refines the lesion contour. Motivated by this coarse-to-fine, multi-level perception process, we present XFlow, a workflow model for ILS that combines box-based localization with multi-turn point refinement. XFlow detects the lungs, decides whether the queried finding is present in each of them, and prompts a fine-tuned SAM with the lesion box for an initial mask. It then corrects that mask through point prompts until its boundary follows the lesion, leaving every intermediate decision visible. Our experiments show that XFlow achieves the best segmentation quality on both internal and external evaluation. Notably, it surpasses ROSALIA in segmentation quality even when the two are trained on the same lesion annotations. Code and model weights will be made publicly available.
[CV-158] KiT: A Foundation Model for Financial Time-Series Forecasting using DiffusionTransformers
链接: https://arxiv.org/abs/2609.34507
作者: Boyu Zhang,Haorui Li
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Financial candlestick forecasting is fundamental to quantitative investment, yet it remains exceptionally challenging due to extremely low signal-to-noise ratios and vast heterogeneity across markets and instruments. Existing approaches have largely attempted to introduce deep learning to capture hidden temporal features, but most adopt an auto-regressive formulation, which leads to error accumulation during inference. Meanwhile, general-purpose time-series foundation models are not tailored to the unique structure of k-line data and yield unsatisfactory performance on downstream candlestick forecasting tasks. To tackle these problems, we introduce KiT, a K-line Diffusion Transformer foundation model, and reformulate future prediction as conditional path generation via flow matching: given a historical context window, the model generates an ensemble of plausible future OHLCV trajectories. We pre-train KiT at multiple parameter scales on billions of candlestick bars spanning multiple markets and timescales. Across three markets and seven resolutions, KiT attains a mean return RankIC of 0.057 and a mean volatility RankIC of 0.66, leading at every timescale and outperforming both task-specific financial forecasters and general time-series foundation models. Code will be available at: this https URL.
[CV-159] SubjectAnchor: Subject-Aware Memory-to-Video for Multi-Shot Storytelling ACM-MM2026
链接: https://arxiv.org/abs/2609.34502
作者: Xinyu Wang,Huafeng Shi,Zian Li,Yan Zhou,Xiaoqiang Liu,Yue Ma,Pengfei Wan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by ACM MM 2026
Abstract:We present SubjectAnchor, a Subject-Aware Memory-to-Video paradigm for multi-shot storytelling in which the current shot is generated by conditioning on explicit visual memories extracted from previous shots. The objective is to preserve subject identity and scene consistency across cuts while retaining the controllability of shot-wise prompting. Built on Wan2.2-I2V-A14B, SubjectAnchor contains three key components: subject-related memory construction, subject-aware temporal rotary position encoding, and memory-aware attention partition. For each target shot, the method constructs a compact memory bank by tracing each required subject to its historical appearance and retrieving the most relevant precomputed keyframes. These memory frames are encoded into the model input as explicit visual conditions, while different subjects are assigned to separated negative temporal slots to reduce identity interference. In addition, memory-aware attention partition regulates the interaction between memory tokens and generated content within a shared backbone. This formulation preserves the appearance anchoring of explicit visual memory while remaining compatible with script-driven shot-by-shot generation. Experiments show that SubjectAnchor improves cross-shot identity consistency over representative memory-based and holistic baselines while maintaining competitive visual quality.
[CV-160] Can Attack Difficulty Be Characterized Before Optimization? A Study of Pre-optimization Difficulty in Person-Vanishing Attacks
链接: https://arxiv.org/abs/2609.34501
作者: Jingyao Xu,Dongdong Wang,Siyang Lu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Adversarial attacks against object detectors are traditionally studied from an optimization perspective, where attack difficulty is regarded as an outcome observed only after adversarial optimization. This raises a fundamental question: \emphcan the relative attack difficulty of different inputs be characterized before optimization begins? In this paper, we investigate this question for person-vanishing attacks by introducing the concept of pre-optimization attack difficulty, which captures intrinsic differences in optimization effort across input images. To estimate this latent difficulty before optimization, we propose Quad-CLEVER, an efficient geometry-based estimator derived from a quadratic approximation of the local person-vanishing margin along the most attack-relevant direction. Extensive experiments across multiple attack algorithms demonstrate that Quad-CLEVER consistently correlates with the observed optimization cost, providing empirical evidence that attack difficulty exhibits a predictable pre-optimization structure. Building upon this finding, we further propose a difficulty-aware attack framework that leverages the estimated difficulty to adaptively allocate optimization budgets for a base attack under a fixed computational budget. On BDD100K, the proposed framework improves the image-level attack success rate by up to 5.78 % while reducing the average optimization cost by up to 11.42 iterations. On the more challenging EventPed dataset, it saves 2.25 optimization iterations while maintaining comparable attack performance. These results demonstrate that attack difficulty can be meaningfully estimated before optimization and that exploiting such estimates enables more computationally efficient adversarial attacks.
[CV-161] ConCAD: Constraint-Aware Image-to-CAD Generation with Dual-Granularity Rewards
链接: https://arxiv.org/abs/2609.34494
作者: Chenxi Zhai,Xi Cheng,Hang Cheng,Zhicheng Guan,Mingyu Fan,Yanzhe Tang,Pingfa Feng,Long Zeng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Image-to-CAD generation seeks executable parametric programs that recover both the geometry and design intent of a reference object. Existing systems are commonly evaluated by validity and shape overlap, although two solids with similar volume can encode different CAD relations. We introduce ConCAD, a constraint-aware image-to-CAD framework optimized via Group Relative Policy Optimization (GRPO) with rewards at two complementary granularities: a code-level constraint reward and an execution-level geometric reward. This complementary design disambiguates structurally distinct yet volumetrically similar shapes while ensuring valid 3D geometry. To verify that these rewards recover geometry and design intent, we introduce a B-rep geometric constraint satisfaction rate (G-CSR), which analytically extracts and evaluates geometric constraints from boundary representations. Experiments on the DeepCAD and Zero2CAD demonstrate that ConCAD achieves the best IoU and Chamfer Distance over competitive baselines, while also outperforming them on G-CSR, validating its superior recovery of both geometric fidelity and parametric design intent.
[CV-162] HPMD: A Historical Persian Manuscript Dataset for Word Spotting with Line-Level Annotation
链接: https://arxiv.org/abs/2609.34490
作者: Saeid Firouzi Daghigh,Majid Iranpour Mobarakeh
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Large collections of historical Persian manuscripts have been digitized, but searching them is still slow and mostly manual. Historians usually want to find where a specific name, date, event, or topic appears, which is a word spotting problem. Progress on this task is limited by two things. First, there is almost no public dataset of historical Persian handwriting; the only notable resource, OpenITI MAKHZAN, contains a relatively small Persian portion. Second, word spotting models usually need word-level bounding boxes, which are very expensive to annotate. In this paper we introduce a new dataset of 223 pages, 3,678 lines, 37,631 words, and 130,630 characters, collected from diverse historical Persian books of poetry and prose and annotated at the region, line, and text level. We also propose a baseline that is trained only with line-level annotations but returns word-level locations. A fine-tuned line detector finds text lines, and a fine-tuned CRNN recognizer trained with CTC produces a frame-by-character posterior matrix for each line. Instead of decoding the most probable character at each frame, the query is scored directly against this matrix, so visually similar characters in Persian such as be and pe no longer cause hard failures. The frame alignment also gives the horizontal position of the word inside the line. On the test set, the fine-tuned line detector reaches an F1 of 0.892, and posterior-based search raises the word spotting F1 from 0.487 to 0.558 compared with exact matching on the decoded text, with the decision threshold selected on a held-out validation set. A PHOC attribute-embedding baseline that additionally receives oracle word boundaries at test time reaches an F1 of 0.449, below the proposed method. We also report a distributional analysis of the dataset, a taxonomy of retrieval errors, and a per-conditionbreakdown of performance.
[CV-163] PACER: Progressive Availability-Conditioned Evidence Routing for Radiology Report Generation under Incomplete Clinical Context
链接: https://arxiv.org/abs/2609.34487
作者: Yulong Chen,Yadong Liu,Haoyu Cao,Sen Xu,Yueying Wang,Jie Wen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 23 pages, 2 figures, 7 tables
Abstract:Radiology report generation (RRG) increasingly incorporates heterogeneous clinical evidence, such as multi-view radiographs and previous reports, whose availability varies across examinations. However, accommodating different input combinations does not ensure effective evidence use: generated reports may still omit or inaccurately describe clinically relevant findings. To address this problem, we propose PACER, a Progressive Availability-Conditioned Evidence Routing framework for structured incomplete-context RRG that follows a Refine-Calibrate-Commit pipeline. It first refines observed visual representations through endpoint-preserving patchwise routing across frozen encoder depths, incorporating complementary cues while retaining the pretrained terminal representation. It then calibrates the language-model prefix according to the observed evidence and availability state, adapting the shared generator’s conditioning as the available source set changes. Finally, it generates polarity-structured clinical commitments before the report in the same autoregressive trajectory, providing structured clinical context for subsequent generation. Experiments demonstrate state-of-the-art clinical efficacy across all four MIMIC-RG4 settings and strong MIMIC-CXR performance, while maintaining competitive language-generation quality.
[CV-164] When Does an Image Determine the Answer? Benchmarking Visual Answerability across Charts and Scenes
链接: https://arxiv.org/abs/2609.34480
作者: Sungguk Cha,Mintae Kim,Youngsub Han,Byoung-Ki Jeon,Sangyeob Lee
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reliable visual question answering requires correct answers when evidence is sufficient and abstention when it is not. We introduce a benchmark that connects complete-question evaluation with explicit evidence for its labels across PlotQA charts, CLEVR rendered scenes, and GQA photographs. Each question groups original and edited images, presented independently; success requires every supported answer and every required abstention to be correct. For chart missing-information labels, executable witnesses establish that admissible complete charts give different answers but identical pixels after masking. Scene labels follow source programs and edits, with a residual-cue analysis for photographs. Across 72,000 responses from six model configurations, the highest observed complete task success rates are 57.0%, 43.5%, and 33.7%, respectively. On charts, the strongest configuration achieves 96.2% per-view decision accuracy, yet 265 of its 835 groups with every decision correct still contain incorrect answers. Evaluating supported answers and necessary abstentions together exposes failures that answerability decisions alone conceal.
[CV-165] SentZero: An Enhanced Sentence-Centric Vision-Language Pretraining for Multi-Task Zero-Shot Chest X-Ray Analysis
链接: https://arxiv.org/abs/2609.34479
作者: Hangyul Yoon,Hyungyung Lee,Edward Choi,Eunho Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision-language (VL) pretraining using paired chest X-ray (CXR) images and radiology reports has shown strong potential for medical image understanding. However, existing methods often remain dependent on task-specific finetuning because radiology reports are lengthy, clinically dense, and difficult to align with simple zero-shot prompts. Recent sentence-level approaches partially address this limitation using clinical phrases extracted by large language models (LLMs), but they largely overlook the intrinsic characteristics of radiology discourse. In particular, limited positive-pair diversity constrains further gains, while clinically equivalent sentences frequently recur across patients, creating false negatives in contrastive learning. To address these issues, we propose SentZero, an enhanced sentence-centric VL pretraining framework for zero-shot, multi-task CXR analysis. SentZero introduces LLM-based abstract-level sentence structuring and mapping to expand positive-pair diversity, together with an additional loss term to mitigate false negatives. We further introduce sentence-conditioned residual modulation of visual embeddings, enabling visual features to adapt to the semantic characteristics of each input sentence. Across diverse downstream tasks and datasets, SentZero improves zero-shot generalization and outperforms prior multi-task zero-shot methods.
[CV-166] VL-AcneSeg: A Vision-Language Framework for Region-Aware Acne Lesion Segmentation ALT
链接: https://arxiv.org/abs/2609.34472
作者: Sukju Oh,Soo Ick Cho,Dae Hun Suh,Sukkyu Sun
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted for publication in the IEEE Journal of Biomedical and Health Informatics
Abstract:Acne assessment is crucial for clinical decision-making, yet traditional grading and counting are subjective and fail to account for lesion size. While area-based assessment has emerged as a promising alternative, acne segmentation has continued to rely on general-purpose architectures. To address this gap, we propose VL-AcneSeg, a multimodal framework for acne lesion segmentation that leverages CLIP and region-level text prompts to incorporate spatial priors, enabling lesions to be localized across the whole face. Because region-level prompts indicate which facial areas contain lesions, we report a single global prompt, which requires no such information, as our primary setting. On our internal clinical dataset, VL-AcneSeg achieves a Dice score of 0.5082 and an IoU of 0.3407 under this protocol, the highest among all compared methods, including recent vision-language segmentation methods that are themselves given region-level prompts; region-level prompting raises these to 0.5296 and 0.3602. Moreover, lesion area measurements derived from our segmentation correlate with IGA scores at a level comparable to expert annotations (Pearson r = 0.719 versus 0.658). Notably, our framework maintains consistent performance across external validation datasets, performing reliably even on uncontrolled smartphone images without requiring additional training or fine-tuning. By pairing a protocol that requires no lesion-location information with area-based severity estimation, this work provides a foundation for objective acne assessment outside the clinic. Our implementation is publicly available at: this https URL
[CV-167] Precise Editing and Flexible Referencing for Interactable Worlds
链接: https://arxiv.org/abs/2609.34470
作者: Xinyao Liao,Xianfang Zeng,Zhu Liang,Zhoujie Fu,Qianxun Xu,Jiachi Liu,Gang Yu,Guosheng Lin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:We present EditWorld, a video world model for precise editing and flexible referencing in interactable worlds. Existing video world models primarily focus on navigation, letting users explore generated worlds but offering limited control over how existing world content is modified. EditWorld extends world modeling from exploration to precise modification by streaming editing instructions and reference images during autoregressive generation. To support these capabilities, EditWorld introduces Gated Causal Attention for temporally varying editing conditions and reference images, together with a Sparse Context mechanism that maintains a bounded historical context for long-horizon inference. We further adopt joint autoregressive and bidirectional training with annealed self-resampling, and construct a dedicated data synthesis and annotation pipeline that provides supervision for world editing. We also present WBench-Editing to systematically evaluate streaming world editing capabilities. EditWorld achieves the best overall performance on WBench-Editing with an overall score of 73.8 and an editing score of 80.0, substantially outperforming existing methods on editing-related metrics. this https URL
[CV-168] Modeling Whole-Slide Images as Dynamic Tumor Microenvironment Fields NEURIPS2026
链接: https://arxiv.org/abs/2609.34451
作者: Lei Wu,Jiashuai Liu,Di Zhang,Zhangpeng Gong,Yingkang Zhan,Yi Niu,Jiusong Ge,Chunze Yang,Kai Yi,Mireia Crispin-Ortuzar,Chen Li,Zeyu Gao
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted at NeurIPS 2026
Abstract:Due to the gigapixel-scale nature of whole-slide images (WSIs), weakly supervised WSI analysis is commonly formulated as a multiple instance learning (MIL) problem, where patch-level features are aggregated into slide-level representations. However, diagnostic and prognostic evidence often arises from spatially coherent tumor microenvironment regions and their interactions, rather than isolated patches alone. Existing patch-level or static region-based methods usually overlook how tissue regions should be adaptively formed and subsequently evolved through microenvironment interactions across heterogeneous boundaries. In this paper, we propose Concept-Guided Tumor Microenvironment Evolution (TMEvolve), a reaction-diffusion-inspired framework that models WSIs as latent tumor microenvironment fields over discrete patch graphs. TMEvolve instantiates this view as a learnable graph-discretized evolution process over patch neighborhoods. It first forms adaptive soft tissue regions as coherent microenvironment units, then performs pseudo-time evolution through two complementary local dynamics: intra-region diffusion, which stabilizes latent states within coherent tissue compartments, and concept-guided boundary flux, which propagates visual feature signals and language-derived concept signals across heterogeneous region interfaces. The evolved microenvironment regions are finally aggregated for slide-level prediction. We evaluate TMEvolve on six datasets across three weakly supervised WSI tasks: survival prediction, gene expression prediction, and histological subtype classification. TMEvolve consistently improves over representative MIL methods, pathology foundation models, and concept-guided baselines. Ablation studies and visualizations further support the effectiveness and interpretability of TMEvolve, highlighting the value of dynamic region modeling and boundary interaction.
[CV-169] When the Score Becomes the Target: Rethinking Metric Validity in Autonomous Driving
链接: https://arxiv.org/abs/2609.34440
作者: Morui Zhu,Deyuan Qu,Qi Chen,Kentaro Oguchi,Qing Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 28 pages including supplementary materials
Abstract:Driving benchmark scores are increasingly used not only for evaluation but also as optimization targets. This raises a fundamental question: do score gains remain reliable evidence of driving improvement once the score itself is optimized? We address this question by examining how the scoring process responds to changes in driving behavior and whether the resulting gains persist under repeated execution and replanning. We decompose the process into execution, measurement, subscore mapping, and aggregation. Controlled interventions reveal substantial behavioral changes that receive little score response because distinctions are omitted, thresholded, or attenuated between requested and executed motion. Closed-loop comparisons further show that optimization gains can reverse when the execution interface changes, demonstrating their dependence on how requests are executed and returned as feedback. Together, these findings connect the behavioral distinctions preserved by a metric to the conditions under which its gains transfer. Metric validity under optimization therefore requires examining both what the scoring process measures and how the optimized behavior is executed.
[CV-170] Clinical Trajectory Alignment for Medical Vision-Language Pre-training
链接: https://arxiv.org/abs/2609.34439
作者: Huimin Yan,Xian Yang,Zhi Wang,Liang Bai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Medical vision-language pre-training largely follows a visit-level image-report matching paradigm, aligning paired images and reports at individual visits. While effective for static cross-modal correspondence, this paradigm provides limited supervision for longitudinal clinical change, such as whether abnormalities improve, remain stable, or worsen over time. Learning such change is challenging because temporal semantics are implicit in free-text reports, and different abnormalities within the same patient may evolve asynchronously or even in opposite directions. We propose MedCTA, which reframes medical vision-language pre-training from visit-level cross-modal matching to learning clinical change. Rather than compressing a patient history into a single temporal representation, MedCTA models clinical change at two complementary scopes. At the abnormality scope, clinically grounded queries construct abnormality-conditioned visual and textual trajectories to capture heterogeneous abnormality evolution. At the patient-course scope, global image and report sequences are modeled to capture overall clinical progression beyond any individual abnormality. Structured trend supervision is extracted from longitudinal reports by an offline LLM parser, removing the need for manual temporal annotations. Combined with static image-report alignment, MedCTA learns representations that preserve visit-level cross-modal correspondence while encoding longitudinal change semantics. Experiments on temporal image classification, image-text retrieval, and zero-shot classification show consistent gains over strong medical vision-language baselines.
[CV-171] HUMAN-TCI: Hierarchical Multi-Stream Motion-Aware Network with Torso-Centered Interaction for Text-to-Motion Retrieval
链接: https://arxiv.org/abs/2609.34430
作者: Muhammad Islam,Euijoon Ahn,Usman Naseem,Tao Huang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Accurate retrieval of human motions is a crucial first step in text-guided human motion modeling and synthesis, as it selects semantically relevant sequences from large datasets and provides grounded references for downstream tasks. Retrieving motions from natural language descriptions remains challenging because sentences can describe multiple actions, overlapping movements, and intricate dependencies between body parts. Existing methods often focus on simple, single-action descriptions and typically process body parts independently or by merely concatenating features, without explicitly modeling how torso movements influence other parts. In addition, their processing pipelines often rely on computationally heavy models, introducing considerable overhead, particularly when modeling longer or more complex motion sequences. This limits learning discriminative motion-pattern representations, reducing retrieval accuracy, interpretability, and efficiency in practical applications. To address these limitations, we propose HUMAN-TCI, a Hierarchical Multi-Stream Motion-Aware Network for text-guided human motion retrieval. HUMAN-TCI employs a three-stream architecture that separately models upper-body, lower-body, and torso motions while explicitly capturing their interactions, allowing torso-related movements to influence the positioning and dynamics of other body parts. By incorporating tailored torso attention, our model effectively recognizes complex human motion patterns, captures fine-grained motion relationships and handles complex multi-action descriptions. Our framework supports retrieval for both simple, single-action sentences and long, compositional descriptions containing sequential or overlapping actions without relying on complex models.
[CV-172] Distilling Visual Reasoning into Text Space
链接: https://arxiv.org/abs/2609.34408
作者: Wenhan Yang,Nilay Naharas,Ali Payani,Baharan Mirzasoleiman
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Large Vision-Language Models (LVLMs) have shown strong promise for multimodal reasoning, yet often struggle with tasks requiring concepts beyond what is directly observable in the input image. Existing methods generate intermediate images or latent visual tokens to guide reasoning, but these representations can introduce errors and increasingly interfere with textual reasoning as reasoning progresses. We propose Visual-to-Text Chain-of-Thought Distillation (V2T), a framework that enables LVLMs to internalize visual reasoning without generating intermediate visual representations at inference time. V2T first trains a teacher LVLM using interleaved visual and textual chains of thought, and then uses knowledge distillation to train a student LVLM using the teacher’s logits and cross-entropy supervision from ground-truth textual reasoning. When reasoning images can be mapped to the original image, V2T can additionally distill the teacher’s attention to corresponding regions, while ground-truth bounding boxes can further guide a subsequent reinforcement learning stage. Experiments across multiple multimodal reasoning benchmarks show that V2T consistently outperforms the teacher and existing baselines, improving average accuracy by 14.3% on a held-out set and 2.7% on the broader visual evaluation suite. Moreover, lightweight SFT and substantially reduced RL make V2T up to 42x faster to train than state-of-the-art baselines.
[CV-173] Unlocking Few-Step Diffusion for Faithful Previews
链接: https://arxiv.org/abs/2609.34406
作者: Jing Jia,Sifan Liu,Guanyang Wang
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (stat.ML)
备注:
Abstract:Sampling latency compounds in diffusion workflows, where users generate and discard many candidates before keeping one. Surprisingly, the poor outputs of standard few-step samplers do not reflect a lack of reconstruction capacity: by optimizing only the initial noise, frozen 3-4-step samplers can closely reproduce their corresponding full-step outputs. Building on this finding, we learn corrections to the initial noise and denoising updates using endpoint supervision, improving correspondence with full-step outputs generated from the same noise and prompt. The resulting previews allow users to screen candidates cheaply and reserve full-step generation for promising ones. Input correction also transfers across sampling budgets without retraining. Experiments show substantial improvements in reference fidelity, including 53-78% lower reconstruction MSE than retrained LD3 on unconditional benchmarks, alongside improved ranking preservation and candidate selection on SD1.5, SDXL, and FLUX.1-dev.
[CV-174] HyperDAM: Hyperspectral Distractor-Aware Memory with Amodal Expansion for SAM 3 Tracking
链接: https://arxiv.org/abs/2609.34396
作者: Ryoga Yuzawa,Tasuku Takagi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Hyperspectral video provides material cues that can disambiguate targets with similar false-color appearance, yet foundation-model trackers update memory primarily from spatial and appearance evidence. We present HyperDAM, a DAM4SAM3-based hyperspectral tracker with three principal contributions. First, HOTC2026-Modal adds human-verified frame-wise modal masks and mask-tight boxes to all 481 organizer-provided HOTC 2026 videos. Second, a frame-zero-calibrated HSI gate rejects spectrally inconsistent updates to the distractor-resolving memory (DRM) without altering the current prediction. Third, a causal spatiotemporal expander adds outward-only amodal corrections from frozen SAM features. Static-scene recovery and empty-mask RTS smoothing address target switches and full occlusion. Model selection prioritizes cross-domain robustness over leaderboard-specific optimization. The final system ranked second in HOTC 2026, achieving 68.0093% AUC and 87.7703% DP@20 in the organizer’s private evaluation.
[CV-175] VastMAT: A Large-Scale Multi-Category Benchmark for Multi-Animal Tracking
链接: https://arxiv.org/abs/2609.34390
作者: Zhizhen Li,Zan Wang,Huidong Peng,Bohan Tan,Shimin Shan,Yu Liu,Liang Peng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multi-animal tracking (MAT) supports the study of animal movement, behavior, and group interactions. However, general multi-object tracking (MOT) benchmarks primarily focus on pedestrians and vehicles, whereas dedicated MAT benchmarks remain limited in jointly supporting broad animal coverage, large-scale video data, and extensive within-video multi-instance association. To address this gap, we introduce VastMAT, which has four key characteristics: (1) Large scale. It comprises 2,947 videos with 1,002,562 annotated frames, totaling 27.85 hours. (2) Broad category coverage. These videos cover 337 animal categories with diverse morphologies and motion patterns. (3) Extensive instance annotations. It provides 3,663,248 bounding boxes and 22,883 identity trajectories—to our knowledge, the largest numbers of both among dedicated MAT benchmarks. (4) High-quality annotations. To ensure reliability, annotations undergo iterative expert review and correction, and quality is assessed through an independent reannotation audit. To systematically assess tracking performance and cross-category generalization, we establish Seen-category and category-disjoint Unseen-category protocols, and evaluate eight representative MOT methods under both protocols. Under these protocols, the highest baseline HOTA scores are 66.37% and 52.90%, respectively, highlighting the challenge of tracking unseen animals. To address the low-overlap association challenge revealed by our analysis, we propose Center-Distance-Augmented Association (CDA), a lightweight module that adaptively combines IoU with center similarity normalized by the boxes’ own scales. Without additional training, CDA improves TrackTrack’s HOTA by 1.58 and 1.31 percentage points under the two protocols, respectively. To facilitate further MAT research, we will publicly release our benchmark and code.
[CV-176] CAR-VLA: Complexity-Aware and Risk-Adaptive Reasoning for Autonomous Driving
链接: https://arxiv.org/abs/2609.34387
作者: Xiaolei Chen,Zhuolin He,Yuxuan Liang,Xu Li,Haotian Chen,Shi Fan,Mengyang Zhao,Wenjuan Meng,Zisheng Chen,Zhihao Zhu,Zhounan Jin,Hengli Wang,Qingfan Wang,Jiamei Liang,Bin Li,Xiangyang Xue
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Existing adaptive reasoning methods for driving Vision-Language-Action (VLA) models primarily focus on whether to reason, overlooking how reasoning should differ across driving situations. Our key insight is that while scene complexity informs reasoning depth, dynamic risk is equally critical for deciding how to reason in time-critical situations. We therefore propose CAR-VLA, a unified driving VLA model that jointly considers scene complexity and dynamic risk to guide reasoning depth, urgency, and focus. CAR-VLA maps four complexity–risk categories to three reasoning modes: \textitFast Intuition for direct trajectory generation in simple low-risk scenes, \textitSlow Thinking for deliberate reasoning in complex low-risk scenes, and \textitReflex Response for compact, hazard-focused reasoning in high-risk scenes regardless of complexity. Rather than merely shortening deliberation, Reflex Response centers reasoning on the most critical hazard and the immediate safe response. We train CAR-VLA through progressive supervised learning that links scene assessment, reasoning-mode selection, and trajectory generation, followed by reasoning-augmented reinforcement learning to improve driving quality and reasoning behavior. Experiments on NAVSIM v1(91.1 PDMS), NAVSIM v2(90.3 EPDMS), and Navhard(35.0 EPDMS) demonstrate competitive driving performance. Qualitative comparisons on navtest and in-house high-risk scenarios further illustrate risk-aware reasoning and hazard-responsive trajectory generation. The code for this paper will be released publicly at: this https URL
[CV-177] RoboIRGBench: Benchmarking Implicit Referential Grounding in Vision-Language-Action Models
链接: https://arxiv.org/abs/2609.34384
作者: Aernaer Akelijiang,Jiannan Li,Zhineng Chen,Jingjing Chen,Bin Zhu
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: Project WebPage: this https URL
Abstract:Vision-Language-Action (VLA) models have shown strong capabilities in robotic manipulation, yet existing benchmarks typically assume that task-relevant information is explicitly specified in the instruction. In practice, however, humans frequently refer to objects, quantities, and relations implicitly, requiring robots to recover the intended target from linguistic and perceptual context. We study this capability as Implicit Referential Grounding (IRG) and introduce RoboIRG-Bench, a manipulation benchmark designed to systematically evaluate it. Built upon RoboMME, RoboIRG-Bench contains 40 variants derived from 11 tasks and covers four challenges, including direct, reasoning-mediated, spatial, and contextual referential grounding. As IRG often requires retaining and retrieving previously established context, we evaluate representative VLAs spanning different memory mechanisms. Our evaluation reveals a noticeable referential robustness gap. Models that perform well under explicit instructions can degrade sharply when the same task-relevant information must be recovered from context. Reasoning-mediated and spatial references are particularly challenging, while models using external VLMs show greater robustness but still exhibit significant failures. Moreover, replacing the external VLM with a stronger model does not eliminate these gaps. We further validate these findings on a Franka Research 3 robot arm, where the gap persists under real-world manipulation and manifests as both incorrect referent grounding and downstream execution failures. These results establish IRG as a distinct and underexplored capability for reliable robotic instruction following and highlight the need for VLAs that can robustly integrate language, perception, reasoning, and action.
[CV-178] Joint and Cross-Modal Video-Audio Generation and Editing: A Unified Formulation and Design Taxonomy
链接: https://arxiv.org/abs/2609.34381
作者: Abhinav Sharma,Sai Karthik Navuluru,Wang Wei,Daksh Dangi,Xiangbo Gao,Li Li,Bo Ni,Vardhan Dongre,Junda Wu,Xiyang Hu,Jiuxiang Gu,Seunghyun Yoon,Tong Yu,Chien Van Nguyen,Mohamed Elmoghany,Nedim Lipka,Hoda Eldardiry,Hongjie Chen,Tyler Derr,Thien Huu Nguyen,Zhengzhong Tu,Nesreen K. Ahmed,Franck Dernoncourt,Ryan A. Rossi
类目: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注: 36 pages, 3 figures, 15 tables
Abstract:Video and audio are perceived together, yet most generative models treat them in isolation. We examine methods that model the two modalities jointly, generate one from the other, or edit them in a coupled manner, organized around a single question: how is the output kept coherent across modalities in time and semantics? A unified formulation casts joint generation, cross-modal generation, and joint editing as three problems defined on a single distribution over audio-visual pairs, and a taxonomy compares methods along five design axes. To our knowledge, this is the first overview to systematically taxonomize joint audio-visual editing, which we map as nine edit categories spanning 28 edit types. We describe methods, datasets, and metrics for each setting and close with the open problems we view as most consequential.
[CV-179] Marathoner: Ultra-Long-Horizon Autonomous Intelligence
链接: https://arxiv.org/abs/2609.34378
作者: Zhang Ruiyang,Ou Jinpeng,Xie Yifan,Zhou Jingang,Pan Lirui,Guo Qingpei,Zheng Zhedong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 30 pages, 4 figures
Abstract:Humans naturally possess the ability to work persistently toward long-term goals. Given a challenging task, humans can continuously work for months or even years to accomplish a specific objective. In this paper, we propose Marathoner, an autonomous agentic model possessing the ability of ultra-long-horizon execution. Specifically, we propose a comprehensive post-training pipeline to instill this critical capability into base model. For Ultra-Long-Horizon Task Synthesis, we leverage major release PRs containing 1000+ lines of new code from diverse GitHub repositories as the primary source for synthesizing challenging task-level data. Additionally, we introduce Multi-Task Chaining, which chains multiple generated tasks into a single more challenging task, enabling the synthesis of tasks with frontier-level difficulty. For rejection sampling finetuning, we combine strong teacher model with diverse harnesses to generate trajectories on our synthesized tasks and conduct supervised finetuning on base model with rejection sampled trajectories. For reinforcement learning, cold-started model performs real-world execution through harnesses in independent sandboxes during rollout process, effectively facilitating the acquisition of genuine ultra-long-horizon execution capability. We further propose a novel reward strategy, Later Stage Bonus Reward, which explicitly encourages model to perform meaningful maneuvers during later stages of execution. Through extensive evaluation on 5 benchmarks containing ultra-long-horizon tasks, Marathoner achieves consistent and substantial performance improvements over base model and even surpasses performance of strong proprietary model. Further analysis shows that Marathoner can consistently work for 10+ hours and conduct 1000+ tool calls on highly challenging tasks.
[CV-180] From Static to Dynamic: On-Policy Distillation from Image to Video Diffusion Models
链接: https://arxiv.org/abs/2609.34371
作者: Bingqing Jiang,Li Luo,Zichao Yu,Yujin Han,Zhaolong Su,Difan Zou
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 27 pages
Abstract:On-policy distillation (OPD) specializes pretrained video diffusion models through teacher supervision along the student’s own generation trajectory. Although large video models are natural teachers, developing specialized video experts can require costly video data and training, while querying them incurs substantially higher latency than querying image experts. More readily available and cheaper to query, image experts offer a cost-effective alternative, particularly for largely temporal-agnostic capabilities such as aesthetics and OCR that admit frame-level supervision. However, heterogeneous image and video latent spaces prevent direct supervision of intermediate student states, while image experts lack cross-frame motion supervision, making temporal consistency vulnerable to frame-level improvements. In this paper, we propose MILD, a Motion-Preserving Image-to-Video Latent Distillation framework that transfers specialized image expertise while preserving pretrained video dynamics. MILD uses a learnable linear connector that aligns student latent states and predicted updates with those of image experts, enabling supervision transfer across heterogeneous latent spaces. We further constrain image-guided corrections around the pretrained student’s predictions to preserve video dynamics and incorporate an optical-flow-based motion reward to improve motion quality and temporal consistency. Across specialized image experts and multiple video-student backbones, our method consistently outperforms video-teacher OPD baselines, with further studies demonstrating effective transfer across connector designs and heterogeneous architectures. These results establish image-to-video distillation as an effective route to improving video generation by drawing on the diverse and evolving capabilities of the image-generation ecosystem.
[CV-181] Rate-Distortion Adaptive Primitive Selection for Omnidirectional Gaussian Splatting
链接: https://arxiv.org/abs/2609.34367
作者: Yulong Cheng,Youneng Bao,Junfeng Zhou,Mu Li,Jie Wen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 30 pages, 13 figures, 14 tables
Abstract:Learned image codecs (LICs) achieve high reconstruction quality, but their decoding speed is often insufficient for immersive virtual reality (VR). Gaussian splatting (GS) codecs render much faster, yet still lag in reconstruction quality and typically decide primitive allocation without considering the coding cost of each primitive. We introduce OIC-GS, an omnidirectional GS codec with a new hierarchical HEALPix primitive grid representation. Gaussian primitives are anchored at predefined spherical locations, eliminating explicit coordinate coding. Finer levels refine their coarser ancestors, naturally supporting coarse-to-fine reconstruction and layered transmission. The predefined grid also enables efficient viewport decoding by selecting only view-relevant primitives. We further introduce a lightweight entropy model for quantized primitives and optimize the codec under a spherical rate-distortion objective. Primitives with insufficient rate-distortion benefit are automatically removed when their quantized opacity becomes zero, allowing OIC-GS to adapt both primitive density and level of detail without a fixed primitive budget. A single bitstream supports full-sphere, viewport-dependent, and progressive decoding. The first viewport reaches final quality after decoding only 52% of the bitstream, and is then rendered at 1,270 FPS. On a 100-image omnidirectional benchmark, OIC-GS outperforms all evaluated GS codecs, reducing WS-PSNR BD-rate by 49.6% over GaussianImage++ and 68.6% over SGI, which uses a learned entropy model.
[CV-182] SyncRA: Learning Temporal Correspondence in Omni-Modal Models
链接: https://arxiv.org/abs/2609.34363
作者: Zelong Xu,Yan Li,Wenhe Hu,Xiyang Hu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Multimedia (cs.MM)
备注: 35 pages, 6 figures
Abstract:Recent omni-modal models demonstrate strong perception of audio and visual inputs, yet often struggle to connect what they hear with what they see at the same moment. This weakness in temporal correspondence can cause models to associate spoken cues with the wrong visual scenes, producing plausible answers grounded in incorrect audio-visual pairings. We diagnose this problem through controlled temporal swaps, revealing that model answers do not reliably follow changes in these pairings. To address it, we propose Synchrony-Guided Representation Alignment (SyncRA), a lightweight method for strengthening temporal correspondence between audio and vision. Specifically, SyncRA contrasts intermediate audio-visual representations within each video, aligning matching moments while separating mismatched ones to capture local temporal correspondence within a shared global context. The objective derives supervision directly from existing input timing, requiring no additional annotations and leaving inference unchanged. We evaluate SyncRA across four open omni-modal models spanning different sizes and architectures on five public video benchmarks. SyncRA consistently outperforms answer-only fine-tuning across all model-benchmark combinations, while substantially improving the ability to track changing audio-visual pairings in controlled evaluations. These results demonstrate that lightweight, targeted supervision can effectively strengthen temporal correspondence and translate into broad improvements in audio-visual question answering.
[CV-183] E-WAVE: Event-based Continuous Optical Flow via Warping-Aligned Visual Encoding
链接: https://arxiv.org/abs/2609.34346
作者: Jiale Wu,Xiaoyang Bai,Haoming Yu,Yiwei Chen,Yifan Peng,Weiwei Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Temporally dense optical flow is essential for dynamic perception in immersive VR/AR systems, where rapid head, hand, and object motion must be continuously captured and tracked. Existing frame-based optical flow estimation methods are constrained by the tradeoff between temporal resolution and computational cost; while event cameras, with their high temporal resolution and energy efficiency, serve as a natural solution to the dilemma. However, event-based approaches commonly rely on correlation volumes to capture pairwise voxel correspondences, which incur substantial memory and computation overhead. We present E-WAVE, a correlation-free framework for high-temporal-resolution (HTR) optical flow estimation from event streams. Instead of constructing all-pairs correlation volumes, E-WAVE employs global attention mechanism to model long-range feature dependencies and performs trajectory guided feature warping using Bézier curve. Through iterative updates, it predicts trajectories that allow for querying at arbitrary timestamps without repeated inference. Experiments on MultiFlow and DSEC-Flow demonstrate a 25% lower trajectory error and comparable endpoint flow estimation accuracy relative to state-of-the art baselines. Additional evaluations on self-captured data using a head-mounted prototype validate that E-WAVE remains robust under challenging real-world conditions.
[CV-184] SkillPE: Creativity-Oriented Cinematic Skill Evolution for Text-to-Video Prompt Engineering
链接: https://arxiv.org/abs/2609.34335
作者: Yanwei Huang,Mingxuan Zhu,Shujie Li,Shiyuan Liu,Yuanxing Zhang,Arpit Narechania
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Achieving high-quality, cinematic results in text-to-video generation remains challenging for non-experts, whose prompts often lack professional narrative and creative design. We propose SkillPE, a prompt engineering (PE) framework that evolves reusable cinematic skills from expert-authored seeds. SkillPE represents shot logic, composition, lighting, sound design, and other filmmaking cues in a fine-grained format, and retrieves movie references categorized as resonators (good matches), dissonants (weak matches), and divergents (creatively useful near-misses). The first two refine when and how a skill should be applied, while divergents inspire alternative cinematic realizations at different degrees of modification while preserving the user intent. Candidate skills are assessed through generated videos along prompt fidelity, cinematic quality, narrative appeal, and creativity to construct the final skill libraries. Experiments on StoryEval and VBench show improvements of up to 1.40 points over the strongest external baseline and 0.51 points over seed skills on 7-point four-dimensional evaluation, while remaining competitive on benchmark-native metrics. Overall, SkillPE offers a practical approach to balancing fidelity and creativity in cinematic text-to-video generation. Code is available at this https URL .
[CV-185] MiCo: Mutual Information Coverag e Optimization through Semantic Erasure Modeling for Efficient MLLM Inference
链接: https://arxiv.org/abs/2609.34330
作者: Tinghao Wang,Yichen Guo,Qizhe Zhang,Yuan Zhang,Weimin Ouyang,Rui Huang,Jiajun Cao,Sixiang Chen,Hao Jiang,Jixian Wu,Zheng Lu,Bofan Zhu,Renyuan Li,Shanghang Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 48 pages, 28 tables, 17 figures
Abstract:Multimodal large language models (MLLMs) have demonstrated impressive performance in multimodal understanding, but processing large numbers of visual tokens results in high computational costs. While many methods have been proposed to reduce the number of visual tokens, most of them rely on heuristics and are prone to discarding substantial visual information during pruning, leading to degradation in model performance. In this work, by using a semantic erasure model, we derive a general mutual information coverage objective from task log-loss and propose MiCo, a training-free two-stage pruning method. MiCo first uses visual signals to select a representative candidate pool before visual tokens enter the language model, then performs task-aware subset selection within it. At each stage, suitable observable proxies instantiate the derived objective as a monotone submodular coverage function, which MiCo greedily optimizes under the token budget. MiCo is evaluated on diverse MLLMs ranging from 7B to 13B parameters across a broad range of image and video benchmarks spanning general visual reasoning, fine-grained OCR and grounding, hallucination detection, and long-video understanding. MiCo consistently achieves the best performance across nearly all evaluated models under all pruning ratios. On LLaVA-NEXT-13B, MiCo uses only 5.6% visual tokens, retains 97.5% of baseline performance, and achieves a 3.8-fold inference speedup. Our experiments demonstrate the effectiveness of MiCo and our mutual information coverage objective for visual token pruning.
[CV-186] DORA: Dynamic Online Reinforcement Agent for Token Pruning in Vision Transformers
链接: https://arxiv.org/abs/2609.34325
作者: Kaixuan He,Song Chen,Yi Kang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 19 pages, 5 figures, 6 tables
Abstract:Vision Transformers (ViTs) incur quadratic self-attention cost in the number of tokens. Most token-reduction methods adapt token identities within a prescribed layer-wise compression schedule, or search a static mask offline, and thus limit online adaptation of when and how much to prune. We propose DORA (Dynamic Online Reinforcement Agent), which learns an input-adaptive pruning policy itself for frozen ViTs. At each eligible block, a hierarchical actor decides whether to prune, how many tokens to remove, and which tokens to remove from each image’s evolving representation. Because early deletions change the states observed by later decisions, DORA formulates pruning as a finite-horizon Markov decision process. Complete-prefix shadow evaluations convert final-prediction fidelity into localized per-step credit, while closed-loop accuracy feedback adjusts the fidelity penalty toward a shared accuracy-drop target. A privileged critic and all shadow computations are training-only. Deployment retains the frozen backbone and a lightweight actor that applies hard deletion and packed variable-length FlashAttention, converting token reduction into measured speedups. On ImageNet-1K with DeiT-Base, DORA reduces FLOPs by 38.4% relative to the uncompressed backbone within one percentage point of accuracy loss. Averaged across four ViT-type backbones at matched accuracy, DORA uses 13.2% fewer FLOPs and achieves 32.4% higher throughput than the corresponding per-backbone baseline means. Under zero-shot transfer to ImageNet-A, these gains widen to 20.3% and 45.6%, respectively.
[CV-187] xt-Vision Synergistic Token Caching: A Training-Free Framework for Efficient Vision-Language-Action Inference
链接: https://arxiv.org/abs/2609.34319
作者: Qianer Li,Chengjie Zhang,Jingwen Chen,Zanjia Tong,Jiyuan Zhang,Hong Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Vision-Language-Action (VLA) models enable generalizable robotic control but remain computationally expensive. Token caching provides a training-free, plug-and-play acceleration alternative. However, existing VLA caching does not fully exploit a key inductive bias of VLA models: text-vision synergy, wherein textual semantics guide the precise visual grounding of task-relevant regions. In particular, existing designs insufficiently account for head-wise reliability in attention aggregation and layer-wise stability in cache reuse. To address this, we propose Text-Vision Synergistic Token Caching (TVCache), a training-free framework for efficient VLA inference. TVCache filters attention heads based on text-vision information focus to improve task-relevant and physically consistent visual grounding. Concurrently, we introduce a reuse-layer selection mechanism guided by text-vision entropy differences to avoid caching unstable representations and improve cache resource allocation. Extensive experiments across four representative VLA models, two simulation benchmarks, and real-world robotic tasks demonstrate the effectiveness and generality of TVCache. At matched token-retention ratios, TVCache consistently improves task success over existing VLA caching with comparable computational cost. On OpenVLA-OFT, it improves average success by up to 14.5 percentage points over VLA-Cache at 12.5% retention while reducing FLOPs by 2.45x relative to full-token inference.
[CV-188] MaLiang-Harness: A Programmable Path to Image and Video Generation
链接: https://arxiv.org/abs/2609.34309
作者: Haoyu Zhao,Zihao Zhang,Xudong Wang,Jiaxi Gu,Zuxuan Wu,Yu-Gang Jiang,Shuicheng Yan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 22 pages, 11 figures
Abstract:Executable programs offer explicit control over how images and videos are constructed, but generating runnable code is only the beginning of visual creation. A program can execute correctly while violating the requested composition, appearance, or motion. We define this discrepancy as the Program-to-Visual (P2V) gap and introduce MaLiang-Harness, a unified framework for organizing MLLM-driven visual generation into a persistent process of construction, inspection, and revision. Its central design is to make the evolving visual program, its construction history, and its verification share a common revision reference. We define the Persistent Executable Generation (PEG) state as preserving programs and task context. Traceable Generation Process (TGP) connects edits to rendered evidence, and Revision-aware Editing and Verification (REV) supports restoration and checks the current revision before completion. Together, these mechanisms coordinate planning, execution, and visual feedback across rendering backends. We evaluate 11 powerful closed-source MLLMs on MaLiang-IBench and four on MaLiang-VBench, measuring generation success, visual quality, and computational cost. GPT-6-Astra achieves 100% generation success on both benchmarks, with 96.0% of image tasks and 76.9% of video tasks meeting all quality thresholds. The comparison also reveals a mismatch between general capability scores and visual generation performance, with similarly scored models differing substantially in their ability to satisfy visual requirements. MaLiang-Harness provides a systematic basis for studying how MLLMs translate executable code into visual outcomes, exposing both the potential of programmable generation and the limitations of general benchmarks as predictors of this ability. The project is available at this https URL.
[CV-189] CAT-Free: Multi-View Pedestrian Localization without Calibration Annotations or Target-Scene Training via Adaptive Geometric Filtering
链接: https://arxiv.org/abs/2609.34302
作者: Taigo Sakai,Hiroki Kouno,Naoki Kato,Kazuhiro Hotta
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multi-camera pedestrian localization is useful for wide-area monitoring in public and commercial spaces. However, deploying these systems often requires considerable setup for each new environment. Existing methods typically require camera calibration, position annotations, or target-scene training. CAT-Free removes all three requirements. It uses synchronized RGB video as its only scene-specific input. Camera configuration is estimated directly from the video. Pedestrian locations are then estimated by combining observations from multiple cameras. Automatic camera estimation is not always accurate. This can produce unreliable pedestrian locations. CAT-Free therefore introduces two adaptive geometric filters. They remove unreliable position estimates. Their thresholds are estimated from each input sequence. CAT-Free achieves 82.5, 84.5, and 65.7 MODA on WildTrack, MultiviewX, and GMVD. It uses no supplied calibration, position annotations, or target-scene training. Published methods using such scene-specific information report 88.2–95.0 MODA on WildTrack and 83.9–96.5 on MultiviewX under their respective protocols. CAT-Free also transfers without retuning. It reaches 74.9 MODA on four additional sequences and 78.6 on an unseen 8-camera installation. Finally, localization uncertainty predicts MODA with r=-0.98 . This provides a label-free estimate of localization reliability.
[CV-190] PSM: Dataset Distillation Based on Precise Statistical Matching by Difficulty
链接: https://arxiv.org/abs/2609.34299
作者: Hongxu Ma,Guang Li,Shijie Wang,Dongzhan Zhou,Suorong Yang,Baoli Sun,Takahiro Ogawa,Miki Haseyama,Zhihui Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Dataset distillation (DD) condenses a large original dataset into a small distilled dataset with high training utility. Decoupled statistical matching methods substantially reduce distillation time and memory overhead while achieving strong performance. However, they typically supervise all distilled samples using running statistics estimated from the entire original dataset. These statistics mainly capture the average feature distribution while overlooking differences in sample difficulty, limiting their ability to characterize the difficulty structure of the original data. To address this issue, we propose Precise Statistical Matching (PSM) by difficulty. After pretraining, PSM uses the Global Precision Score (GPS) to estimate image difficulty, ranks the samples within each class, and partitions each class into IPC (images per class) difficulty groups. During distillation, Statistics Updated Again (SUA) updates the teacher’s batch normalization (BN) running statistics through forward passes on original samples from each group, providing difficulty-specific supervision for the corresponding distilled batch. Meanwhile, Initial Sample Screening (ISS) initializes distilled samples using original images from the corresponding difficulty group, providing an effective starting point for precise matching. Experiments across multiple datasets and model architectures demonstrate that PSM broadens the difficulty range of distilled samples and improves downstream performance in most evaluated settings. Code will be released.
[CV-191] Semantic Modality Compensation for Unsupervised Visible-Infrared Person Re-identification under Unpaired Settings
链接: https://arxiv.org/abs/2609.34294
作者: Duanning Chen,Ke He,Bin Yang,Yongxiang Yao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint
Abstract:Unsupervised visible-infrared person re-identification (USL-VI-ReID) learns person representations that can be compared across modalities without identity annotations. In the unpaired setting, however, identity correspondences between modalities are often incomplete, leaving many identities without an observed counterpart in the other modality. Existing unpaired methods bridge this gap by generating or mapping features for the other modality, mainly by exploiting the statistics of visual features without explicitly separating content that is discriminative for identity from style that is specific to modality. Consequently, the generated features may distort identity cues or inherit bias from the source modality, undermining the reliability of supervision across modalities. We formulate unpaired learning across modalities as a semantic compensation problem and propose Semantic Modality Compensation (SMC), a framework based on prompt composition that decouples identity semantics from modality style within a shared visual semantic space. SMC first constructs a discriminative ReID space through augmented dual contrastive learning, yielding pseudo labels, cluster prototypes, and memory banks for each modality. It then learns visible and infrared modality prompts in the CLIP semantic space and maps clusters obtained from pseudo labels to identity semantic tokens. For each cluster lacking a reliable match in the other modality, SMC combines its identity token with the prompt for the target modality to synthesize a semantic counterpart in the missing modality. The synthesized counterpart is then projected back into the ReID space and injected into a compensation memory through confidence gating. Extensive experiments under both paired and unpaired settings demonstrate that SMC consistently outperforms state-of-the-art methods, with particularly large gains when identity mismatch is severe.
[CV-192] Dexterous Tactile World Model
链接: https://arxiv.org/abs/2609.34286
作者: Ziyao Zeng,Xiatao Sun,Hao Wang,Yueyang Pan,Zhengxiang Yu,Fengyu Yang,Tianyu Liu,Zhiwen Fan,Daniel Rakita
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Robotics (cs.RO)
备注: Project page: this https URL
Abstract:World models for manipulation are typically trained from video, yet the events that determine how manipulation unfolds, such as making and releasing contact, are difficult to observe visually and are often easier to sense through touch. We present the Dexterous Tactile World Model (DTWM), a video world model for future-frame prediction of egocentric manipulation from both observed video and tactile signals from a glove worn on each hand. We condition a pretrained video diffusion transformer on each hand’s tactile signal through a zero-initialized residual at the corresponding hand location in the video tokens, while a causal mask prevents predicted frames from accessing future information. Compared with a vision-only model matched in architecture, parameters, and training, DTWM reduces the underestimation of hand motion from 23% to 9%, while reducing the perceptual error in the hand region by 7.4% across three training runs per model. The benefit also increases over the prediction horizon, with the improvement in the later predicted chunks being about 4.1x larger than in the first. DTWM also outperforms other visual-tactile world models under the same setting, and training with touch improves future-frame prediction even when no touch is available at inference. Ablations show that the model benefits from both the magnitude and spatial location of force: replacing the tactile signal with binary contact states, either per hand or per location, increases prediction error. The observed course of the force indicates whether the interaction will persist or change.
[CV-193] MoSPR: Histology-to-Gene Expression Prediction with Morpho-Spatial Macrostates and Low-Rank Molecular Programs
链接: https://arxiv.org/abs/2609.34280
作者: Dongmyung Shin,Geongyu Lee,Yesung Cho,Park Jong Bae
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Predicting molecular profiles from histopathology remains challenging because whole-slide images contain spatially organized, heterogeneous tissue patterns, while gene expression comprises thousands of correlated targets. We introduce MoSPR (Morpho-Spatial Program Regression), a linear framework that couples an adjacency-informed histology representation with a low-rank molecular basis. MoSPR clusters frozen patch embeddings into morphology microstates, aggregates their spatial adjacencies across the training cohort, and groups microstates with similar adjacency patterns into shared macrostates. Each slide is then represented by global morphology and macrostate-specific deviations, which are linearly mapped to coefficients of a training-derived low-rank gene-expression basis. Across three cancer cohorts from The Cancer Genome Atlas, MoSPR achieves the highest mean gene-expression prediction scores among all evaluated methods. Without pathway-level supervision, pathway scores derived from its predicted expression profiles rank first in eight of nine comparisons across three pathway collections. Ablation studies on the breast cancer cohort show complementary gains from adjacency-derived macrostate representation and low-rank molecular prediction. Moreover, with half of the training data on this cohort, MoSPR exceeds the full-data gene-prediction score of the strongest competing baseline. Finally, its linear formulation enables exact decomposition of each predicted expression profile into global and macrostate-specific molecular contributions, providing an interpretable link between spatially coherent macrostate regions and their associated molecular programs. Our code is available at this https URL.
[CV-194] See Measure and Reason : Learning Visually Grounded Reasoning in Pathology
链接: https://arxiv.org/abs/2609.34277
作者: Chengyang Zhang,Wenchuan Zhang,Bo Li,Mengran Li,Xinyu Liu,Jiaming Yang,Jie Chen,Zhang Zhang,Yuhao Yi,Hong Bu,Jiancheng Lv
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Pathological assessment relies on recognizing fine-grained visual details in histological images. Vision-language models (VLMs) increasingly support pathology interpretation, yet their ability to perceive these details remains inadequate. This weakness leads to inaccurate cellular observations that can persist even when final answers are correct. In this paper, we propose ASPECT to improve visually grounded reasoning through explicit supervision of cellular appearance and abundance. ASPECT trains intermediate visual tokens through pathology feature reconstruction, cell feature alignment, and count supervision. Three-stage supervised fine-tuning teaches the model to perceive, generate visual tokens, and reason, followed by reinforcement learning that rewards answer correctness and consistency with reported measurements. We also introduce PathoVernier, a benchmark of 759 expert-reviewed questions from five pathology datasets covering four cellular composition tasks. It evaluates both final answers and intermediate measurements to expose errors hidden by answer accuracy. On PathoVernier, ASPECT achieves relative accuracy gains of approximately 19.2% over the strongest baseline, Gemini-3.1-Pro, and 99.3% over its Qwen3-VL-8B backbone, while reducing RAWR, which measures counting errors within correct responses, by 28.1% and 42.7%, respectively. ASPECT also improves over its backbone on three external pathology benchmarks covering classification and question answering beyond cellular composition tasks.
[CV-195] Scaling Versatile 3D Assets Editing with a Million-Scale Dataset
链接: https://arxiv.org/abs/2609.34271
作者: Badi Li,Tianxin Huang,Yu Zhou,Wei-Shi Zheng,Yi Ma,Shenghua Gao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Although recent 3D generative models produce increasingly realistic assets, controllable 3D asset editing remains challenging. Existing methods are limited by scarce training data, insufficient source-aware modeling, and a lack of practical evaluation protocols. To address these limitations, we present Alchemy3D, a unified framework for training and evaluating versatile 3D asset editors that covers data construction, model architecture, and benchmark evaluation. Specifically, we curate Alchemy3D-1M, a large-scale 3D editing dataset containing 1.25M assets and 1.38M editing pairs across seven editing types. On this data, we train a family of generative flow models for general-purpose 3D asset editing. The model family supports image- and text-conditioned editing, few-step inference, and transfer to multi-view 3D part segmentation. We further introduce GEdit3D-Bench, a large-scale, open-world benchmark with a multi-dimensional evaluation protocol. Across existing and newly introduced benchmarks, our method outperforms prior methods on most metrics of editing fidelity, source preservation, and visual quality.
[CV-196] DecFlowEdit: Self-Localized Flow-based Image Editing via Guidance Decoupling
链接: https://arxiv.org/abs/2609.34237
作者: Zheyuan Zhan,Can Wang,Jiawei Chen,Chun Chen,Siwei Lyu,Zeyu Zheng,Defang Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Flow-based image editing (FlowEdit) enables inversion-free semantic changes through the difference between source and target velocities. In this paper, we observe that FlowEdit’s default classifier-free guidance (CFG) configuration, with asymmetric source and target scales, causes substantial background leakage. Matching these guidance scales, for example by removing CFG, improves edit-relevant localization but severely degrades editability. To get the best of both worlds, we propose DecFlowEdit, which decouples the optimal guidance scales for localization and for editing in flow-based generative models. In particular, DecFlowEdit first extracts an edit-relevant prior by temporally aggregating velocity differences evaluated without CFG, and then uses this prior to reweight the original updates under default CFG. Our method remains training-free and inversion-free, requiring neither external spatial masks nor attention manipulation. Experiments on PIE-Bench across FLUX, SD3, and SD3.5 show that DecFlowEdit improves background preservation, reducing structure distance by approximately 61 to 73 percent and background LPIPS by 68 to 80 percent relative to FlowEdit at comparable editing fidelity.
[CV-197] SegBanana: Steering Unified Multimodal Models into Medical Segmenters
链接: https://arxiv.org/abs/2609.34235
作者: Xiaoye Liang,Ye Yan,Mingze Yin,Shikun Feng,Mai Xu,Haiguang Liu,Lai Jiang,Yiheng Zhu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Medical image segmentation remains challenging in practical deployment, as models often struggle to generalize beyond the distributions covered by their training data and high-quality pixel-level annotations are typically unavailable for adaptation. Inspired by the cross-task transferability of large language models, we investigate whether unified multimodal models (UMMs) can transfer their pretrained visual understanding, reasoning, and generation capabilities to medical image segmentation without task-specific post-training. By recasting segmentation as structured visual generation, we find that frontier UMMs (e.g., Nano Banana) already exhibit basic segmentation capabilities across diverse clinical scenarios, but still struggle with challenging tasks requiring specialized anatomical or domain-specific knowledge. We further show that these limitations can be effectively mitigated by incorporating visual anatomical knowledge from in-context exemplars, expanding candidate solutions through repeated sampling, and refining suboptimal predictions via targeted this http URL by these observations, we propose SegBanana, to our knowledge, the first agentic visual generation framework for training-free medical image segmentation. SegBanana builds on a frozen UMM as the core generative model, augmented with Anatomy-Aware Knowledge Retrieval and Comparative Quality Critique to unlock its potential segmentation capability. A State-Aware Multimodal Controller maintains structured state and iteratively orchestrates these tools, repeatedly refining intermediate predictions toward higher-quality masks. Across eight medical segmentation datasets, SegBanana achieves an average mDice of 77.45%, outperforming representative generalist (SAM3 and SegGPT) and medical-specific (BiomedParse and MedSAM3) baselines by at least 14.93 points, while remaining robust to out-of-domain visual supports.
[CV-198] rustworthy synthetic visual media: Evidence across the media lifecycle
链接: https://arxiv.org/abs/2609.34232
作者: Zexi Jia,Zhiqiang Yuan,Jie Zhou,Jinchao Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Images and videos have long helped people understand what happened and how a work came into being. Generative systems complicate that role. Realistic media can now be produced and revised without leaving a stable history, so appearance no longer reveals whether a scene was captured, synthesized, or altered along the way. Trust must instead come from evidence that explains the path an asset has taken and the circumstances in which it was used. Some of this evidence can be recovered from the media, while some must be recorded during production and preserved as the asset circulates. This review brings those approaches together and asks when their claims remain meaningful after ordinary processing or deliberate manipulation. We argue that trustworthy media do not depend on one universal marker of authenticity. The evidence must suit the question at hand, reach the person making the judgment, and remain open to correction when better information emerges. The larger goal is to keep the history of media intelligible even as the media itself continues to change.
[CV-199] ReGDiff: Guided Diffusion in Regulated Latent Space for Exploring Metamaterial Voxel Geometry
链接: https://arxiv.org/abs/2609.34231
作者: Wangzhi Zhan,Jianpeng Chen,Dongqi Fu,Dawei Zhou
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Metamaterials are artificially engineered structures whose mechanical and physical behaviors are strongly shaped by geometry rather than composition. Voxel representation provides a unified format for metamaterial geometry generation, as it can express diverse classes such as truss, shell, and porous structures within a single cubic discretization. However, voxel-based generation faces a plausibility-novelty trade-off: staying close to known geometries helps preserve geometric regularities, while moving away from them is necessary for novelty but may produce degenerate geometries. To address this challenge, we propose REGDIFF, a generative framework that couples voxel representation with latent space regulation and guided diffusion. REGDIFF introduces a repel-and-sink (RAS) mechanism to smooth the latent distribution of plausible geometries, and short-range repulsion (SRR) guidance to discourage generation overly close to known samples while maintaining geometric plausibility. We further contribute a voxel-based benchmark covering truss- and shell-type metamaterial geometries, together with an evaluation module for geometric plausibility, novelty, and diversity. Experiments show that REGDIFF outperforms voxel-based generative baselines, achieving +8.9% in geometric plausibility, +46.4% in novelty, and +128.6% in diversity on average across two datasets. These results suggest that REGDIFF is a strong geometry candidate generator for downstream evaluation. Our code is provided at this https URL.
[CV-200] Uncovering Ordinal-Matching Bias in Audio-Visual LLM s
链接: https://arxiv.org/abs/2609.34223
作者: Jihoo Jung,Youngjoon Jang,Hyebin Cho,Suho Yoo,Joon Son Chung
类目: Computer Vision and Pattern Recognition (cs.CV); Sound (cs.SD)
备注: Preprint
Abstract:This work aims to improve how audio-visual large language models (AVLLMs) associate speech with the correct visible speaker in multi-speaker scenes. We find that current AVLLMs frequently fail at this task, and analyze the nature of these failures. To this end, we construct a synthetic diagnostic dataset in which multiple visible speakers each utter a single word. Analysis on this corpus reveals a consistent error pattern across three recent open-source AVLLMs: models attribute utterances by simply matching the order of spoken sentences with the left-to-right, top-to-bottom arrangement of visible faces, rather than relying on audio-visual cues such as lip synchronization. We term this behavior \emphordinal-matching bias. We further show that this bias can be substantially mitigated through a simple remedy, Ordinal-Decoupled Fine-Tuning (OD-FT), in which models are fine-tuned on synthetic videos where spatial positions of speakers and speaking order are independently randomized. Despite using only 400 synthetic training videos, OD-FT not only suppresses ordinal-matching bias but also improves audio-visual understanding on real-world videos, yielding average gains of 8.27% for Qwen2.5-Omni and 2.57% for video-SALMONN2+ across three audio-visual benchmarks.
[CV-201] WorldWeave: Growing Persistent Geometric Worlds for Video Generation
链接: https://arxiv.org/abs/2609.34221
作者: Yifan Huang,Lifan Jiang,Qingyue Hao,Cheng Chen,Boxi Wu,Xiaoxue Ren,Xiaofei He,Dehai Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL . Code repository: this https URL (implementation coming soon)
Abstract:Despite rapid progress, world models still lack explicit, persistent structural memory, making it difficult to preserve consistent world structure during continual scene expansion and cross-view revisits. To address this limitation, we present WorldWeave, a world generation framework that decouples world-state maintenance from visual rendering. Specifically, WorldWeave combines continual elevation-map generation with agent-guided scene organization and stitching to build an expandable explicit 3D world that incrementally extends structural memory while preserving existing structure. First, its terrain module uses diffusion-based image outpainting to generate continuous metric elevation maps under neighborhood conditioning and boundary constraints. Next, an agent integrates user intent, terrain evidence, and cross-region connectivity constraints to construct scenes through hierarchical semantic planning, deterministic geometry compilation, and local revision. Finally, during visual generation, planned camera trajectories query world geometry through a read-only interface, producing depth sequences that guide video synthesis without writing the generated results back into the world state. As a result, structural memory remains independent of short-window video generation, enabling continual expansion without predefined map boundaries and providing a consistent geometric basis for observations across trajectories and repeated visits.
[CV-202] mmHRI: Towards Privacy-Preserving Human-Robot Interaction with Millimeter-Wave Radar
链接: https://arxiv.org/abs/2609.34220
作者: Junqiao Fan,Yuxuan Hu,Bofan Lyu,Yanshuo Lu,Pengfei Liu,Jiarui Zhang,Fangqiang Ding,Lihua Xie,Gen Li,Jianfei Yang
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Assistive robots increasingly operate in many human-centered environments and perform various human-robot interaction (HRI) tasks, such as object delivery. However, most existing HRI systems rely on RGB cameras that continuously observe humans to respond to non-verbal commands, such as hand gestures. This raises privacy concerns in privacy- critical environments, such as hospital wards or restaurants, where direct camera observation of humans is restricted. To develop privacy-preserving HRI, we leverage millimeter-wave (mmWave) radar, which can sense human motion through privacy barriers without identifiable imagery. We propose mmHRI, the first multi-modal robot manipulation framework that achieves mmWave radar-guided privacy-preserving HRI. mmHRI introduces two key designs to mitigate the sparsity and temporal inconsistency of radar data in cluttered robot manipulation environments. First, we propose a dual-stream architecture that jointly learns from unfiltered raw radar tensors and radar point clouds to estimate both human actions and 3D poses. To mitigate signal inconsistency, mmHRI further incorporates a memory-based state-space model (MSSM) that retains historical radar features to reduce abrupt changes in pose/action. These estimated human states are then converted into structured textual robot instructions, which control a vision-language-action (VLA) policy for closed-loop robot manipulation and human-aware reactions. Our evaluation covers human action recognition and closed-loop delivery and retrieval. In the privacy-preserving curtain setting, mmHRI achieves 85.09% action-recognition accuracy, outperforming existing radar-based alternatives. Robot trials further demonstrate successful delivery and retrieval under visual occlusion, with stable task performance across unseen subjects, clutter configurations, and environments.
[CV-203] WorldGuide: Learning Success-Failure Boundaries in Latent World Models for Vision-Language-Action Policies
链接: https://arxiv.org/abs/2609.34206
作者: Lin Liu,Lu Zhang,Ziying Song,Wu Yang,Yuzheng Zhuang,Yunzhi Zhuge,Shuai Tao,Wulong Liu,Huchuan Lu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Latent world models offer a promising way to improve Vision-Language-Action policies by capturing the consequences of actions. However, models trained primarily on expert demonstrations have limited exposure to failure outcomes and may struggle to distinguish visually similar successful and failed interactions. We propose \textbfWorldGuide, a framework that learns these distinctions in latent space and uses them to guide policy training. WorldGuide combines predictive pretraining on successful and failed trajectories with contrastive learning on matched success–failure pairs. The learned predictor then provides a differentiable reward to guide joint optimization of the policy and visual encoder. The predictor is discarded after training, so deployment requires no additional world-model inference. Extensive experiments show that WorldGuide substantially improves VLA reliability and achieves state of the art performance on LIBERO 100 and SimplerEnv, reaching \textbf96.8% and \textbf72.0%, respectively. Code will be publicly available.
[CV-204] ConvCue: Complementary Visual Inductive Biases for Vision-Language Models
链接: https://arxiv.org/abs/2609.34196
作者: Zixuan Lan,Shichu Sun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Modern vision-language models (VLMs) achieve strong performance across a broad range of multimodal tasks, yet still struggle with visual questions that require fine-grained discrimination and spatial understanding. These limitations motivate investigating whether supplementary visual representations can improve existing VLMs without replacing their native visual encoders. Pretrained convolutional networks offer a candidate feature source, motivated by their local connectivity and spatial weight sharing. We introduce CONVCUE, which augments the native visual representations of a pretrained VLM with final-stage features from a parallel, frozen pretrained CNN. A learnable adapter maps convolutional features to the native visual feature dimension, while gated cross-attention allows the original visual tokens to retrieve information from the CNN features. The enhanced tokens are passed through the original visual-to-language projector, and the model is adapted through a two-stage training procedure. We evaluate CONVCUE on Qwen3-VL-2B, Qwen3-VL-4B, and LLaVA-OneVision-7B across 13 multimodal benchmarks covering visual question answering, document and chart understanding, and multimodal reasoning. CONVCUE improves average benchmark performance over both the original models and matched two-stage fine-tuning controls on all three backbones. On Qwen3-VL-4B, it improves over the original model on all 13 benchmarks and raises the average score from 75.00 to 78.82 relative to the matched fine-tuning control. These results show that pretrained convolutional representations, when integrated through learned adaptation and fusion, can improve the visual understanding of existing VLMs without replacing their original visual encoders.
[CV-205] MotionSpaceFlow: Representation-Aware Flow Matching in Direct Motion Space
链接: https://arxiv.org/abs/2609.34190
作者: Qing Yu,Kent Fujiwara
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project website: this https URL
Abstract:Recent advances in diffusion and flow models have substantially improved text-driven human motion generation. Yet most methods generate in low-dimensional, temporally downsampled latent spaces learned primarily for reconstruction, a bottleneck that can limit generation quality and preclude direct manipulation of individual frames and joints. We introduce MotionSpaceFlow (MSFlow), a representation-aware flow-matching framework that predicts clean motion directly in continuous motion space without a learned encoder or decoder. To account for the anisotropic structure of direct motion representations, we propose representation-aware noise scaling and show how the initial Gaussian source scale governs the covariance of intermediate probability-path marginals. We further introduce a Representation-Aware Multimodal Diffusion Transformer (RA-MMDiT), which jointly updates token-level language and full-resolution motion features through joint attention while adapting temporal information flow to the motion representation: causal attention for incremental features defined by frame-to-frame changes, and bidirectional attention for global features such as absolute joint coordinates. Across different datasets and motion representations, MSFlow achieves state-of-the-art text-to-motion performance. Its global representation variant additionally enables zero-shot, inference-time control over any joint or frame through projection sampling without control-conditioned training, delivering leading motion quality with exact constraint satisfaction.
[CV-206] CRF Loss is How Networks Should Learn Boundaries in Weakly Supervised Segmentation
链接: https://arxiv.org/abs/2609.34183
作者: Joshua Li,Yuri Boykov
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Weakly Supervised Semantic Segmentation (WSSS) learns pixel-level predictions from image-level tags. Recent work focuses on improving coarse CAMs extracted from large vision-language models (commonly CLIP), but does little to improve their accuracy along segment boundaries. That job is instead delegated to a post-processing method like DenseCRF. However, because DenseCRF relies on low-level colour cues, it can flip correct labels to incorrect ones when neighbouring pixels share similar colours. SAM has recently been adopted as a natural alternative, yet it simply takes on DenseCRF’s role as an intermediate “refinement” step that outputs one-hot pseudo-labels in prior work. By discarding the valuable uncertainty in CAMs, these one-hot pseudo-labels turn borderline errors into confidently wrong targets. Our key insight is that CAMs should supervise training alongside SAM boundaries, each through its own loss, rather than being fused together into a single hard target. Inspired by CRF potentials, we propose a framework that disentangles soft pseudo-labels as unary supervision and binary edge maps as pairwise supervision. We realize our framework in a single-stage model, DS-CRF, using CAMs from this http URL and boundaries from SAM. DS-CRF sets a new state-of-the-art of 56.5% mIoU on MS COCO.
[CV-207] Decision Readouts for Text-Mediated Video Anomaly Detection: An Exploratory Evaluation of Jev and Qwen
链接: https://arxiv.org/abs/2609.34180
作者: Xukui Qin,Youting Wang,Xinjie He,Ziyang Luo,Runxiong Wu,Yan-Syuan Chen,Zhongyao Chu
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 19 pages, 2 figures; exploratory preprint
Abstract:How much does the decision readout matter when video-derived textual evidence is held fixed? We evaluate Jev typed decisions and three Qwen readouts on a sparse development sample of 40 videos and 400 target anchors from UCF-Crime and XD-Violence, each presented as a summary and ordered captions. Each dataset contributes 20 source groups and 200 anchors, including only 10 and 37 positives, respectively. The original five-backend pilot requested 4,000 predictions; Jev Choice returned 776 valid responses out of 800 under the study’s strict numerical policy, blocking its full-coverage quality comparison. On XD captions, Jev Noul achieved 75.99% average precision versus 48.47% for Qwen generated probability and 57.81% for the stronger local ordinal-likelihood expectation. The latter paired difference was 18.18 percentage points (95% source-group bootstrap interval 5.53-31.50). UCF did not show a corresponding advantage: caption ROC-AUC was 52.26% for Noul and 65.95% for ordinal likelihood. Both probability readouts had higher, hence worse, UCF Brier scores than the evaluation-prevalence reference of 0.0475. We additionally audit historical LAVAD scores at exactly matched anchors and distinguish response structure from numerical consistency. A binary-likelihood control is missing. These exploratory offline results characterize ranking, probability quality and interface failures; they establish neither a causal typed-interface benefit nor general superiority, calibration or end-to-end acceleration.
[CV-208] Enhanced Video Text Editing with Trajectory-Aligned Glyph Rendering
链接: https://arxiv.org/abs/2609.34178
作者: Shulian Zhang,Xiangyu Shu,Wenbo Li,Jian Chen,Yong Guo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video text editing aims to replace or add text in a video while keeping the rest of the video unchanged, which requires the edited text to be correct in every frame and to move coherently with the scene. Despite the remarkable progress of video diffusion models, they struggle to reproduce exact stroke structures and often produce garbled or wrong characters, especially for characters with complex strokes. To address this, we propose a trajectory-aligned glyph rendering reference that provides explicit per-frame glyph guidance following the position and perspective of the text, and a depth-normalized recognizer feature supervision that supervises the generated text on multi-depth features of a frozen text recognizer with per-depth normalized errors, targeting stroke errors overlooked by the diffusion loss. We further build VTEdit, a benchmark of 288 real-scene clips with 440 annotated text trajectories covering text replacement and text addition, which will be publicly released to facilitate future research. Experiments on VTEdit show that our method outperforms image text editing methods, video editing methods, and commercial models in text accuracy and background preservation, achieving a sentence accuracy of 0.9408, and receives the highest preference in a user study.
[CV-209] AGILE-GS: Anchor-Guided Fast Next-Best-View Selection for Active 3D Gaussian Splatting
链接: https://arxiv.org/abs/2609.34176
作者: Amirhossein Mollaei Khass,Nader Motee
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Radiance fields need hundreds of views, and their placement matters as much as their number. Next-best-view (NBV) selection for 3D Gaussian Splatting (3DGS) usually scores every candidate in the pool and keeps one. Searching for information and choosing a camera, however, are separable problems. We present AGILE-GS, an anchor-guided NBV method that separates the two. A virtual anchor pose is optimized on SE(3) by Riemannian gradient ascent on expected information gain. It need not be reachable or in the pool; it marks where the model is most uncertain. Candidates are scored against the anchor’s viewing geometry, and a greedy ridge-leverage step distills the pool into a small, non-redundant shortlist without rendering any candidate. The shortlist can be used in two ways. AGILE-GS takes the first view on it as the next view, so no Fisher information is computed for any candidate. AGILE-GS+ computes the Fisher information gain of each shortlisted view and picks the best, so the expensive evaluation runs on a handful of views rather than the whole pool. On standard benchmarks and in closed-loop embodied acquisition, both match or exceed existing baselines while cutting selection latency by one to two orders of magnitude.
[CV-210] Natural Image Autoencoder-Based fMRI Representations for Trait and State Prediction
链接: https://arxiv.org/abs/2609.34167
作者: Juhyeon Park,Yeonwoo Kim,Peter Yongho Kim,Yansen Wang,Mingqing Xiao,Dongqi Han,Dongsheng Li,Taesup Moon
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Under Review
Abstract:Foundation models pre-trained on large-scale fMRI datasets have shown strong downstream performance, but at substantial data and computation cost. To investigate how much fMRI-specific pre-training is actually needed for such performance, we introduce FReD, which derives fMRI representations from a frozen Deep Compression AutoEncoder (DCAE) pre-trained exclusively on natural images and pairs them with a task specific readout. For trait prediction, FReD summarizes frame-wise representations by their temporal mean and log-standard deviation and applies linear probing, with late fusion across two normalization schemes. For state prediction, it represents each frame as a single token and models temporal dependencies with a shallow Transformer. Across four resting-state datasets spanning six trait-prediction targets, linear probes on frozen DCAE features generally outperform those on fMRI foundation model representations and remain competitive with fully fine-tuned fMRI foundation models. On three task-fMRI state-prediction tasks, a temporal readout on DCAE features performs comparably to the strongest foundation models evaluated. A Gaussian injection analysis further shows that localized signal changes are recovered more accurately from the frozen DCAE features than from the evaluated foundation-model representations. Together, these results show that strong performance on current fMRI benchmarks is possible without fMRI-specific representation pre-training, making frozen natural-image features as a useful baseline for assessing its added value.
[CV-211] WorldGraph: Graph-Native World Modeling
链接: https://arxiv.org/abs/2609.34159
作者: Zezhong Ding,Yipeng Li,Xike Xie
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:World models infer latent states of an environment to capture its underlying dynamics and predict future evolution. Many real-world environments, however, are inherently relational and observed as evolving graphs, where entities, relations, and their properties change over time. Prior graph-related world models use graph structures to organize internal states or support task-specific reasoning, rather than treating an evolving graph itself as the modeled world. We instead study graph world modeling (GWM), where graph evolution itself constitutes the world dynamics. We formulate graph world modeling over observed graph evolution, latent graph states, and heterogeneous graph-transition predictions. Based on this formulation, we construct GWM-Zero, a benchmark covering node-, edge-, and graph-level transitions over eight temporal graph datasets. We propose WorldGraph, which combines a state-aware graph transformer for multi-granularity structural and transition-conditioned evolution modeling with transition-aware GRPO using dynamic grouping and structure-aware verifiable rewards. Extensive experiments on GWM-Zero show that WorldGraph consistently outperforms representative graph representation, temporal graph learning, graph pretraining, and graph world-model baselines across all three transition granularities.
[CV-212] Functional Hand Type Prior for 3D Hand Pose Estimation and Action Recognition from Egocentric View Monocular Videos BMVC2023
链接: https://arxiv.org/abs/2609.34149
作者: Wonseok Roh,Seung Hyun Lee,Won Jeong Ryoo,Jakyung Lee,Gyeongrok Oh,Sooyeon Hwang,Hyung-gun Chi,Sangpil Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: BMVC 2023 Oral Paper
Abstract:Current methods for egocentric view action recognition often face challenges in perceiving dynamic hand movements relying solely on geometrical or physical information. In this work, we effectively address this problem by gaining insights into the correlation between functional hand configurations and objects, which improves the detailed interpretation of real-world scenarios. To this end, we introduce a practical taxonomy of hand types based on the functioning perspective and utilize it for per-frame hand type labeling on existing datasets. We also propose a novel hand action recognition framework considering semantic details of the hand type as prior. This approach boosts the network’s understanding of the continuous hand interaction throughout the action sequence. Our whole pipeline consists of three main modules: (1) Feature Extraction, (2) Egocentric Knowledge Module, which estimates 3D hand pose, object category, and hand type leveraging short-term cues, and (2) Egocentric Action Module, which aggregates per-frame knowledge, including text embeddings of hand type, over a longer time. In our extensive experiments with large-scale benchmarks, FPHA and H2O, our model outperforms current state-of-the-art methods, demonstrating its superior performance.
[CV-213] Geometric Encoding for Spatial Reasoning in Vision-Language Models
链接: https://arxiv.org/abs/2609.34148
作者: Antonio Jun,Haoshui Yu,Zhengyi Lu,Huirong Fu,Yao Qiang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-Language Models (VLMs) are far more reliable at recognizing what appears in a video than at reasoning about its spatial and temporal properties, such as metric distances, object dimensions, and consistent object identities across frames. We present Geometric Code, a perception-to-geometry pipeline that computes explicit spatial structure from video and supplies it to VLMs as context to augment reasoning. A perception layer segments and classifies objects and recovers depth, camera pose, and intrinsics from monocular RGB video. A deterministic geometric engine then back-projects, merges, and cleans these outputs into a spatial code, including per-object positions, dimensions, counts, inter-object distances, appearance order, and room geometry. The code is serialized into VLMs’ prompts, either alongside the video or replacing it entirely. Specifically, there is no component trained or fine-tuned in our approach. On VSI-Bench, augmenting 2B and 4B open models with the spatial code improves average accuracy by +4.1 points over the frames-only baseline, with the largest gains on numeric estimation tasks such as absolute distance (+24.1 points). The results suggest that explicitly computed geometry, delivered through the language channel, recovers spatial competence that small VLMs cannot extract from pixels alone.
[CV-214] CAST: Reconstruction-Coupled Acceleration of Interactive World Models
链接: https://arxiv.org/abs/2609.34144
作者: Leyang Chen,Junyi Wu,Fanqing Kong,Shaoqiu Zhang,Yulun Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 23 pages, 14 figures
Abstract:Interactive world models must respond quickly to controls while preserving scene consistency. Existing acceleration methods can miss heterogeneous control responses and spatial transport when recovering skipped features. We observe that interaction-induced feature changes correlate with approximation error, while low-frequency interpolation errors are phase-sensitive and show more predictable phase progression. These findings motivate CAST, a reconstruction-coupled inference framework. CAST selects anchors by interaction sensitivity and cross-layer coverage, reconstructs skipped residuals with frequency- and confidence-aware Phase-Aware Reconstruction (PAR), and coordinates historical KV routing according to downstream reconstruction responsibility. On Matrix-Game 3.0 and HY-World 1.5, CAST achieves 2.15x and 3.48x speedups, respectively, while maintaining visual quality close to Native (Figure 1). It also attains the highest VBench scores among compared methods and leads non-native baselines on seven and six of thirteen WorldMark dimensions, demonstrating a balance of generation speed, visual quality, and interactive responsiveness under real-time control. Code is available at this https URL.
[CV-215] Beyond Geometry: Benchmarking and Consistency Reasoning for 3D Logical Anomaly Detection
链接: https://arxiv.org/abs/2609.34143
作者: Zhiqiang Qin,He Xie,Junfei Yi,Yang Yang,Hao Wang,Yunkang Cao,Hui Zhang,Yaonan Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Existing 3D industrial anomaly detection mainly targets local geometric deviations. In contrast, many industrial anomalies violate object-level design or assembly rules, which we define as 3D logical anomalies. To address these challenges, we introduce the Industrial Logical Anomaly Detection Dataset (ILGAD), the first scalable benchmark dedicated to logical anomalies in industrial point clouds. ILGAD contains 2,774 samples from 15 categories with point-level annotations and covers existence, specification, pose, and assembly-state errors. To detect such 3D logical anomalies, we propose a consistency reasoning framework that assesses whether local geometry, structure coverage, and spatial relations conform to the normal design. The framework detects geometric changes, unsupported expected structures, and abnormal local arrangements. Experiments on ILGAD, Anomaly-ShapeNet, and IEC3D demonstrate superior object-level detection and point-level localization, showing that the framework effectively detects logical anomalies and generalizes to conventional geometric defects.
[CV-216] Analytical and Convolutional Neural Network-Based Motion-Vector Propagation for Efficient Video Object Detection
链接: https://arxiv.org/abs/2609.34142
作者: Ashiyana Abdul Majeed,Mahmoud Meribout,Neethu Joseph
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 11 pages, 4 figures, This work has been submitted to IEEE for possible publication
Abstract:Continuous video analytics requires accurate localization at low latency within embedded power budgets. This paper presents a hardware-software design methodology that reuses codec motion vectors (MVs) between detector invocations. Two alternative models support translation and scale changes: analytical motion-vector propagation (Analytical-MV) and learned propagation using a convolutional neural network (CNN) (CNN-MV). The learned model uses convolutional operations and independent object updates suited to parallel execution on an edge graphics processing unit (GPU). Analytical-MV combines a harmonic-mean precision-recall score (F1) of 0.909 with a mean end-to-end latency of 9.03 ms and an energy consumption of 0.177 J per frame, yielding the lowest latency and energy among the evaluated configurations. Relative to detection on every frame, it reduces mean latency by 25.9% and energy per frame by 36.4%. CNN-MV offers a different trade-off: its fastest configuration raises recall from 0.871 for Analytical-MV to 0.890 and lowers mean power from 19.64 to 17.32 W, while achieving a latency of 18.42 ms and an energy consumption of 0.319 J per frame. It is therefore useful when recall or operating power is more important than minimum latency and energy. Execution on a deep learning accelerator (DLA) further reduces time-averaged GPU utilization relative to GPU execution. Host-processing optimization substantially improves both latency and energy, demonstrating the value of jointly designing temporal models and their execution pipelines.
[CV-217] PrefLUT: Reusable and Refinable Personalized Color Editing from Pairwise Preferences
链接: https://arxiv.org/abs/2609.34133
作者: Chuanzhi Xu,Langyi Chen,Chengkun Yue,Xuanhua Yin,Boyu Wei,Qingwen Zeng,Zihan Deng,Weidong Cai
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Photographic color editing is inherently personal: the same image can appear too warm, too muted, or already satisfactory to different users. Most lookup table (LUT) and reference-guided methods target a specified appearance rather than model persistent preferences from repeated user choices. To address this gap, we introduce PrefLUT, a reusable and refinable user-preference modeling framework for deployable 3D LUTs, encoding ordered preferred/non-preferred image pairs into a lightweight Reusable User Profile that is reused across queries and refined using additional user preference pairs, without per-user optimization. A Query-Conditioned LUT Predictor combines this profile with each image to predict a LUT latent vector and edit strength. An Identity-Residual LUT Decoder and Edit-Strength Controller then produce an exportable 3D LUT. Experiments on three datasets demonstrate effective personalized editing and general-purpose enhancement. Each quantized profile requires only 260 bytes, and editing takes 1.365 ms/image on an RTX 5090 GPU. We also introduce the Preference-Conditioning Verification Protocol (PCVP), an evaluation protocol to verify whether personalized image edits depend on user preferences and the query image through controlled changes to user profiles, preference orders, pair correspondences, and query images.
[CV-218] SpatialSkill: Self-Evolving Skills for Cross-View Spatial Reasoning
链接: https://arxiv.org/abs/2609.34124
作者: Ruifan Zuo,Guocheng Hu,Wanshui Gan,Junyi Wang,Xiang Lei,Tian Gan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Cross-view spatial reasoning requires a model to align different viewpoints into a coherent spatial representation, yet this ability remains challenging for vision-language models despite being natural to humans. Existing methods typically improve spatial reasoning by updating model weights, which keeps the acquired knowledge implicit and tied to a specific backbone. We propose \textitSpatialSkill, a weight-update-free framework that enables a frozen vision-language model to accumulate explicit natural-language reasoning skills from offline trajectories. Unlike symbolic tasks, perceptual skills cannot be reliably verified simply by executing them: a plausible spatial rule may lack visual support or require transformations that the frozen model cannot perform. SpatialSkill therefore admits candidate skills only after visual-grounding and executability checks, constrains manual evolution to prevent harmful regressions, and routes skills by spatial-reasoning category to reduce negative transfer. On CityCube, across four frozen executors, SpatialSkill yields consistent gains, and a 9B executor equipped with SpatialSkill surpasses the strongest closed-source reference in our evaluation. The skills are stored in a versioned natural-language manual, making the reasoning strategies explicit and auditable without modifying model parameters. Code at this https URL.
[CV-219] SpecRegMatch: Robust Semi-Supervised Regression for Vehicle Interior Noise Prediction
链接: https://arxiv.org/abs/2609.34111
作者: Sejin Sim,Jinsoo Bae,Seoung Bum Kim
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: Published in IEEE Access, vol. 12, pp. 60-72, 2024
Abstract:The rapid advancement of artificial intelligence has observed increased application in predicting vehicle interior noise levels within the automotive industry. However, the collection of labeled data for training models in this context involves significant costs. Previous studies in semi-supervised regression (SSR) have effectively mitigated the reliance on labeled data by incorporating unlabeled data. Nonetheless, these approaches often introduce a high computational cost due to the training of multiple models and data sampling. This study introduces SpecRegMatch, a novel SSR method aimed at addressing the computational cost associated with training by leveraging a single model, thus eliminating the need for multiple data samplings. SpecRegMatch integrates consistency regularization and information maximization to robustly train the model, achieved through various augmentations applied to both the embedding vectors and predicted values. Experimental results demonstrate that SpecRegMatch achieves state-of-the-art performance across various scenarios, even when using a single model. It attains a remarkable performance, as indicated by an R^2 score of 0.434. This is especially noteworthy in scenarios where labeled data is scarce. You can access the code for our proposed method at this https URL.
[CV-220] he Devil is in the Spectrum Bias: Spectrum-Balanced Feature Matching for Robust Representation Distillation
链接: https://arxiv.org/abs/2609.34106
作者: Kuniaki Saito,Yoshitaka Ushiku
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Large visual foundation models have demonstrated remarkable transferability across a wide range of downstream tasks. To deploy such models efficiently, feature matching has become a popular knowledge distillation approach that transfers teacher representations to smaller student models without requiring labeled data. However, we show that the conventional feature matching objective with L2-distance is inherently biased toward reconstructing dominant spectral directions of the teacher representation, while under-optimizing low-variance directions that often contain task-relevant information. To address this, we propose Spectrum-Balanced Feature Matching, SpecMatch, a simple objective that adaptively emphasizes under-optimized spectral directions while preserving the relative importance of dominant directions. SpecMatch is easy to implement and introduces negligible computational overhead. Extensive experiments on image recognition demonstrate that SpecMatch consistently improves downstream adaptation across diverse tasks, including image classification, anomaly detection, medical image analysis, and domain generalization. In particular, SpecMatch outperforms conventional feature matching in 40 of 42 teacher–student and training-setting combinations, while consistently improving over the original student model in all settings. We further demonstrate that the proposed objective generalizes beyond vision, improving downstream performance across six protein understanding tasks.
[CV-221] Advancing Wildlife Conservation through Multimodal Animal Re-Identification with Environmental Metadata
链接: https://arxiv.org/abs/2609.34094
作者: Yuzhuo Li,Di Zhao,Tingrui Qiao,Yihao Wu,Bo Pang,Yun Sing Koh
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 6 pages, 3 figures, 3 tables
Abstract:Identifying individual animals is crucial for effective wildlife monitoring and conservation efforts. Recent advancements in computer vision have shown promise in animal re-identification (Animal ReID) by leveraging data from camera traps. However, existing Animal ReID datasets rely exclusively on visual data, overlooking environmental metadata that ecologists have identified as highly correlated with animal behavior and identity, such as temperature and circadian rhythms. Meanwhile, modern vision-language models (VLMs) offer rich multimodal reasoning capabilities, but existing resources underutilize their text-processing potential. To address these limitations, we propose MetaWild, a multimodal Animal ReID dataset comprising 20,890 images across six species, paired with environmental metadata extracted from embedded camera trap overlays and scene contexts. Additionally, to facilitate the use of metadata in existing ReID methods, we propose the Meta-Feature Adapter (MFA), a lightweight module that can be incorporated into existing VLM-based ReID methods, allowing ReID models to leverage both environmental metadata and visual information to improve ReID performance. Experiments on MetaWild show that combining baseline ReID models with MFA to incorporate metadata consistently improves performance compared to using visual information alone, validating the effectiveness of incorporating metadata in re-identification.
[CV-222] AD-E2E-JEPA: A Joint-Embedding Predictive Architecture For End-to-End Autonomous Driving
链接: https://arxiv.org/abs/2609.34085
作者: Haoran Zhu,Wancong Zhang,Yann LeCun,Anna Choromanska
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Autonomous driving requires \textitworld models that can understand the physical world, reason and plan, and operate safely. In this paper, we first systematically evaluate existing action-conditioned joint-embedding predictive architecture (JEPA) world models, including LeWM, DINO-WM, and JEPA-WM for end-to-end autonomous driving (E2EAD). To isolate world-model quality from policy learning, we employ a goal-conditioned zero-shot planning setting that evaluates these models using ground-truth future observations as goals, without training any driving policy. We find that existing JEPA-based world models are either accurate for driving but computationally expensive, or computationally efficient but insufficient for planning. To address this trade-off, we propose \textbfAD-E2E-JEPA, which introduces a SIGReg-regularized learnable projector applied to projected patch embeddings. The projector reduces the number of planning patches by 16\times and the embedding dimension by 4\times , achieving a 100\times inference speedup while retaining planning performance, with a 0.8-second runtime for an 8-frame rollout over 256 candidate trajectories. \textitWithout training any driving policy, the world model itself reaches the goals located 20 meters away on average within the displacement of respectively 4.0/2.8 meters, using world-model rollouts over trajectory vocabularies of respectively 256/8,192 candidates. On the NAVSIMv2 benchmark, it achieves 67.3/72.9 EPDMS with multiplicative safety metrics and 84.1/86.5 EPDMS ^\dagger without them in goal-conditioned zero-shot planning. Experiments further show that the self-supervised pretrained projector improves downstream imitation learning performance from 80.2 to 85.4 EPDMS. The source code is available at this https URL
[CV-223] K-OPSD: Verifiable On-Policy Self-Distillation for Post-Training Vision-Language Models on AEC Drawings
链接: https://arxiv.org/abs/2609.34082
作者: Yunfei Bai,Enrico Chionna,Akash Amol,Kawaljit Singh KC,Joern Tinnemeyer
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Interpreting architecture, engineering, and construction (AEC) drawings is hard for general Multimodal Large Language Models (MLLMs) and vision-language models (VLMs). We introduce K-OPSD, a VLM post-training methodology for improving AEC drawing understanding. Building on On-Policy Self-Distillation (OPSD) with verifiable supervision, we construct a teacher from the model’s own best-of-N generations, certified by a process-level verifier, and rescue failed prompts by resampling under a hint that exposes the verified answer. We then perform an on-policy model update by training on verified completions with a cross-entropy inner-loss, outperforming the bounded token-wise generalized Jensen-Shannon divergence (JSD) used by on-policy distillation. Using K-OPSD, we fine-tune Qwen3-VL models on the AECV-Bench dataset. The resulting models attain the top average judge score (0.819) and combined accuracy (0.738), achieving competitive results against open-source baseline models. The recipe transfers to the out-of-domain ArchCAD dataset, where the 8B model gains most. We present the verifier suite and the continual learning and self-improving pipeline, our results provide preliminary evidence that verifier-guided self-distillation is a promising route toward more reliable machine reading of architecture drawings.
[CV-224] WhiteCon: Semi-Supervised Domain Adaptation Regression Through Whitening Transform and Dual Consistency ICPR2026
链接: https://arxiv.org/abs/2609.34078
作者: Se Jin Sim,Seoung Bum Kim
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to ICPR 2026
Abstract:Domain adaptation is crucial for addressing distributional shifts that degrade model performance across domains. While most existing research has centered on classification, semi-supervised domain adaptation regression (SSDAR) for continuous-output tasks remains largely unexplored, particularly in practical scenarios with limited labeled target data. To address this gap, we propose semi-supervised domain adaptation regression through whitening transform and dual consistency (WhiteCon), which combines domain-specific whitening transform (DWT) and dual consistency regularization to enhance training stability and domain adaptation. DWT reduces the variance of the model parameters by transforming the feature covariance matrix into an identity matrix, thus stabilizing training under ordinary least squares assumptions. In addition, variance consistency regularization, as part of dual consistency regularization, aligns the variances of weak, strong, and mixup-augmented features to improve resilience against augmentation-induced perturbations. Empirical evaluations on various benchmark datasets under SSDAR settings demonstrate that the proposed WhiteCon achieves state-of-the-art performance compared to existing methods, effectively addressing domain shifts in regression tasks. The code for WhiteCon is available at this https URL.
[CV-225] SNaP: One-Step Posterior Sampling for Noisy Inverse Problems
链接: https://arxiv.org/abs/2609.34071
作者: Shirin Shoushtari,Edward P. Chandler,Xiao Shi,Ulugbek S. Kamilov
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Diffusion and flow-matching models can produce high-quality posterior samples for inverse problems, but typically require tens to thousands of network evaluations per draw. MeanFlow enables one-step generation, yet applying it to inverse problems leaves no intermediate steps at which to enforce measurement consistency. We introduce SNaP, a one-step MeanFlow posterior sampler for linear inverse problems with Gaussian noise. Its central innovation is a measurement-adapted source: a Gaussian distribution whose mean and anisotropic covariance are determined by the measurement operator, observation, and noise level. The source anchors well-measured directions while preserving variation where the measurements are weak or uninformative. We show that the exact conditional flow transports this source to the true posterior. Across natural-image restoration and multi-coil MRI, SNaP produces diverse, high-quality samples with one network evaluation per draw, 30 to 2250 \times faster than iterative samplers.
[CV-226] Quantile Head for Vision-Language-Action Models
链接: https://arxiv.org/abs/2609.34061
作者: Xuan Wang,Yinan Wu,Haoran Duan,Jungong Han
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-Language-Action (VLA) models integrate pretrained Vision-Language Models (VLMs) with action heads for robot control. Common action heads have distinct limitations: point regression provides only a point estimate of the action distribution, while standard flow-matching samplers require costly iterative sampling. To address these limitations, we unify regression and flow matching under a shared objective and extend it to derive a quantile objective. This quantile objective guides the design of our Quantile Head, which predicts a median and positive gaps to form ordered marginal action quantiles in one forward pass. These quantiles support multiple sampling strategies without retraining and are jointly supervised to train the default median policy. Our local analysis of this joint supervision shows that, with calibrated nearby quantiles, fixed gaps, and matched correction speed, direct median updates have lower variance than under median-only supervision. Experiments show that this jointly supervised median policy achieves the highest average success rates among the compared methods on LIBERO, LIBERO-Plus, LIBERO-Pro, and two real-robot tasks, together with the shortest mean episode time among matched LIBERO baselines; code is available at this https URL.
[CV-227] ARCH-B: Architectural Representation Comprehension and Hierarchy Benchmark
链接: https://arxiv.org/abs/2609.34047
作者: Kieran Sagar Parikh,Jose Luis Garcia del Castillo y Lopez
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Multimodal models increasingly interpret visual environments, but their ability to recognize the same building across photographs, floor plans, elevations, sections, and renderings remains poorly characterized. We introduce ARCH-B, a benchmark of 354 four-choice questions across 11 cross-representational archetypes, constructed from a building-linked corpus of 3.9 million architectural images using visually similar distractors, model-guided difficulty screening, and manual validation. We evaluate 25 multimodal models and collect 5,830 responses from non-expert human participants. Model accuracy ranges from 10.45% to 83.90%, compared with a human baseline of 35.35%. Models perform comparatively well on mixed-representation outlier detection and photograph matching, but remain weaker on floorplan-to-photograph correspondence. Human and model difficulty across archetypes is only weakly correlated (Spearman’s (\rho=0.33)). Held-out evaluation confirms that the difficulty identified during screening generalizes beyond the curation models. ARCH-B provides a diagnostic evaluation of visual correspondence and representation transfer across architectural media.
[CV-228] SCOPD: Sparse-Context On-Policy Self-Distillation for Efficient Vision-Language Models
链接: https://arxiv.org/abs/2609.34044
作者: Ahmadreza Jeddi,Enming Zhang,Jasper Gerigk,Hakki Karaimer,Mozhgan Nasr Azadani,Jiayun Luo,Minh Ngoc Le,Gholamali Aminian,Hugo Buurmeijer,Yongchao Chen,Leonid Sigal,Igor Gilitschenski,Konstantinos G. Derpanis,Marco Pavone,Babak Taati
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reasoning vision-language models (VLMs) process images and videos as long sequences of visual tokens, making inference expensive. Training-free token pruning reduces this cost, but aggressive compression can sharply degrade performance, often attributed to irreversible loss of task-relevant visual information. We show that this explanation is incomplete. In a fixed-context Pass@K analysis, repeated sampling from the same pruned visual representation recovers many examples missed by greedy decoding, indicating that useful visual evidence can remain accessible but be used unreliably. We call this the representation-utilization gap. Motivated by this observation, we introduce SCOPD, a sparse-context on-policy self-distillation framework in which a student generates reasoning trajectories from pruned visual tokens while a privileged full-context teacher supervises the same on-policy prefixes. SCOPD requires no ground-truth responses, architectural changes, or additional inference-time computation. We further introduce SCOPD+, which uses a small visual-budget intervention to identify visually sensitive response positions and selectively distill them. At 10% visual-token retention, the Vanilla model retains 86.37% of its unpruned performance across 13 benchmarks. SCOPD raises this to 90.49%, while SCOPD+ further improves it to 92.43%. Across token budgets, benchmarks, and pruning operators, our results show that efficient reasoning depends not only on which visual information survives pruning, but also on how reliably the model learns to use it.
[CV-229] 3D Point Tracking with State Space Models
链接: https://arxiv.org/abs/2609.34035
作者: Masahiro Ogawa,Qi An,Atsushi Yamashita
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Robotics (cs.RO)
备注: 20 pages, 11 figures, 9 tables. Submitted to Computer Vision and Image Understanding
Abstract:Tracking any point of a dynamic scene in metric 3D - in absolute meters, not up to an unknown scale - underpins 3D and 4D reconstruction, robot navigation, and autonomous driving, where decisions are made in meters, not pixels. Our objective is a 3D point tracker accurate in those absolute terms and operating within a single commodity GPU, pose-free, monocular budget. Our method rests on one observation: once a point’s 2D image trajectory is fixed, the quantity that governs its metric accuracy is the depth along its pixel ray. Rather than learning tracking end-to-end, we therefore compose two frozen front-ends - dense optical flow for 2D correspondence and a monocular metric-depth network for the third dimension - and learn only the residual they cannot supply: that depth, refined by a compact state space model (Mamba-3) conditioned on appearance features (DINOv3). A state space model rather than the transformers the strongest 3D trackers adopt is what makes a single-GPU budget attainable: it summarises a track in a fixed-size recurrent state whose memory cost is constant in the number of frames, whereas attention requires a key-value cache that grows linearly with them. On the TAPVid-3D minival benchmark our best configuration attains the highest absolute metric accuracy among methods evaluated under identical conditions (mean metric Average Jaccard, 0.256), exceeding strong feed-forward trackers, while a companion analysis, reproduced with each competitor’s own evaluator, explains why several published trackers lose most of their accuracy under this budget.
[CV-230] Re:Cognize – Open-Set Comic Character Re-Identification NEURIPS2026
链接: https://arxiv.org/abs/2609.34032
作者: Aaditya Baranwal,Madhav Kataria,Yogesh S Rawat,Shruti Vyas
类目: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注: Accepted at NeurIPS 2026 ED Track
Abstract:A manga reader meets a character on one page and knows them on sight a hundred pages later, without ever being handed a cast list. Re-identifying comic characters demands the same, open-set and sequential: pages arrive as a stream in reading order, new faces appear before anyone names them, and the cast is assembled as the story is read. \textbfRe:Cognize evaluates recognition as the story is read, not against a cast handed over in advance: four protocols on one query stream, from closed-set retrieval to a cast the model must build and grow itself. The surprise is where models fail. Recognising is close to solved: one reference image per character already ranks as well as a gallery built in advance. Knowing what to believe is not: a model that adds its own matches makes its cast worse, while the same growth with correct labels would gain over twenty points of top-1 accuracy. The bottleneck is acceptance, not vision, and one comparison decides it: an addition pays exactly when it is right more often than the cast already was on the queries it takes over. The comparison has nothing to fit, and measured on half of a new corpus it calls the other half correctly. \textbfReCast puts it to work with nothing fitted on data: a cast sheet of one running average per character, grown only where the page itself vouches for a crop. It recovers a third to two thirds of what perfect labels would, depending on whether the cast starts from random examples or from first appearances. Re:Cognize measures whether a model can read along; ReCast is a cast that does. Our claims are on identity maintenance, recognising characters already met; the emergence of new ones is measured as a diagnostic under a fixed reference rule, and we propose no method for it.
[CV-231] Structure-Adaptive Tree Field Integrators
链接: https://arxiv.org/abs/2609.34025
作者: Millend Roy,Soham Samal,Ivan Zelich,Krzysztof Marcin Choromanski
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV); Data Structures and Algorithms (cs.DS); Numerical Analysis (math.NA); Optimization and Control (math.OC)
备注:
Abstract:We present a new class of near-linear algorithms for efficiently integrating general tensor fields defined on trees with distance dependent kernels, the Structure-Adaptive Tree Field Integrators (STAD-TFIs). STAD-TFIs exploit the tree’s underlying structure through decompositions built around path backbones and single vertex separators, and use two-dimensional fast Fourier transforms to compute interactions jointly. By exploiting this structural information, STAD-TFIs achieve more computationally efficient integration than their regular efficient tree field integrators (TFI) counterparts. We provide a detailed theoretical analysis of our proposed approach and complement it with an exhaustive empirical evaluation, ranging from speed tests on synthetic trees, through accelerated Sinkhorn-based relaxations of the Optimal Transport algorithms on real meshes, to Topological Attention Transformers for vision tasks. To the best of our knowledge, we provide some of the first results showing that efficient to compute and accurate relaxations of the geodesic Sinkhorn-based solutions of the Optimal Transport problem can be derived by applying fast TFI methods.
[CV-232] Position Aware Layer Queries for Test Time Training in Vision Language Models
链接: https://arxiv.org/abs/2609.34021
作者: Rajat Modi,Priyank Pathak,Xin Liang,Yogesh Singh Rawat
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Test-Time Training (TTT) adapts models to incoming test samples (e.g. out-of-distribution, (OOD)) when conventional fine-tuning is infeasible. Existing TTT methods for Vision-Language Models (VLMs) create supervision from several augmented views, each requiring forward (and often backward) passes through the entire VLM, incurring substantial computational cost. We observe that one forward pass with all the intermediate layer outputs already yields far more signal than the final embedding from all augmentations. We introduce Layer Query Network (LQN), a lightweight approach that can adapt a frozen VLM (teacher) in a single forward pass of the VLM via a small model (student). LQN uses Position-Aware Distillation (PAD) to mimic the teacher VLM’s intermediate-layer spatial tokens by querying spatial coordinates of intermediate tokens. LQN additionally relies on Location Consistency Regularization (LCR), a self-supervision technique, replacing expensive O(H x W) image augmentation with O(1) coordinate sampling. Integrating these, LQN i) adapts and improves zero-shot CLIP ViT-B/16 by 9.8% Top-1 on OOD ImageNet, ii) outperforms the previous best GS-Bias on fine-grained classification by 3.9% Top-1, iii) achieves faster convergence than TPS for CLIP ResNet-50 (47 mins vs 55 mins), iv) generalizes adaptation to VLMs like SigLIP, EVA-CLIP, and CoCa, and lightweight students like MLP, ResNet, VGG, and v) extends to panoptic, instance, and semantic segmentation.
[CV-233] A Computer Vision Approach to Visual Fraud Detection in Phishing Websites Using YOLOv8
链接: https://arxiv.org/abs/2609.34015
作者: Basil Sajid Shaikh,Hajar Homayouni
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Phishing remains one of the most common vectors for financial and identity fraud, and most detection systems still rely on inspecting a page’s URL, HTML markup, or domain registration history. These signals are easy for an attacker to rotate or obfuscate, and they say very little about what actually convinces a victim to hand over a password or a card number: the way the page looks. This paper describes a visual, image-based approach to phishing detection that treats a rendered webpage the same way a human eye would, as a picture that either matches a trusted brand or doesn’t. A YOLOv8 convolutional neural network was trained to classify full-page website screenshots as phishing or legitimate based on layout, logo placement, color scheme, and login-form structure, rather than on text extracted from the page. The system reached 92% classification accuracy on a held-out test set, processed a single screenshot in roughly 100 milliseconds, and, after a round of data augmentation aimed specifically at lighting, compression, and scaling variation, cut the false-positive rate by 11% relative to the pre-augmentation baseline. The paper walks through the dataset construction, the augmentation strategy, the model architecture and training setup, and the resulting performance, and closes with a discussion of where this kind of visual detector fits alongside, rather than instead of, existing URL- and content-based defenses.
[CV-234] ZeroBot: Learning from Scratch in Minutes with Generative Real2Sim
链接: https://arxiv.org/abs/2609.34010
作者: Ivan Kapelyukh,Xiaohan Zhang,Stephen James,Laura Herlant,Edward Johns
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: IEEE RA-Letters 2026. Project page: this https URL
Abstract:We present ZeroBot, a real2sim framework for learning a robot manipulation task from scratch in minutes under challenging conditions: zero human demonstrations, zero policy pre-training, and zero known object models. Given only a single view of an object and a goal pose for that object, ZeroBot uses image-to-3D generative models to obtain a complete object mesh, which is used in simulation for large-scale parallel reinforcement learning. To accelerate training, we introduce an action space which leverages the generated geometry and learned value function to sample states involving robot-object contact. When evaluated on real-world tasks including grasping, pushing, articulated object interaction, and multi-stage manipulation, ZeroBot achieves an 87% success rate with an average training time of 119 seconds. These results show the value of using image-to-3D models in a real2sim framework for rapid, autonomous robot learning.
[CV-235] MetaSampling: Making Frame Samplers Efficient for Long-Video Question Answering
链接: https://arxiv.org/abs/2609.33998
作者: Ashim Dahal,Bikramjit Banerjee
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Frame selection is an important component of long-video question answering (VQA) with Multimodal Large Language Models (MLLMs). Existing frame-selection methods improve over simple top- k embedding retrieval and uniform sampling, but are typically applied under a fixed global selection budget. We introduce \textbfMetaSampling, a training-free, plug-and-play sampling strategy that can be applied on top of existing frame selectors. MetaSampling improves downstream VQA efficiency by dynamically reducing the number of frames passed to the MLLM while preserving, and in some cases improving, answer accuracy. We evaluate MetaSampling across 36 paired frame-selector–MLLM-backbone–VQA-benchmark configurations. MetaSampling reduces the number of selected frames in all 36 configurations and improves accuracy in 25 of them, yielding an average frame reduction of 8.9% while slightly improving accuracy overall.
[CV-236] UnfoldCRF: Structured Mask Refinement with Image-Conditioned Latent Regions
链接: https://arxiv.org/abs/2609.33996
作者: Chunming He,Rihan Zhang,Lei Xu,Guanyi Qin,Chengyu Fang,Longxiang Tang,Fengyang Xiao,Sina Farsiu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages
Abstract:Learned mask refiners improve segmentation accuracy, but it is hard to tell how much of the improvement comes from explicit structure rather than from extra capacity, and whether it holds up when the mask generator or its error distribution changes. UnfoldCRF treats refinement as inference in a conditional random field over pixel labels and latent region variables. Its energy has a corrected unary term, learned local pairwise interactions, and image-conditioned latent-region consistency, with a null state that lets a region with weak label agreement withdraw from the consistency term; inference unrolls damped mean-field updates on this one energy. To isolate the effect of structure, we compare against recurrent black-box refiners that read the same inputs and receive the same parameter budget, stage count, and supervision. On COD10K, UnfoldCRF beats the strongest matched control by 1.0 F^\omega_\beta point, improves all four COD metrics, and lowers the fraction of images made worse from 11.7% to 8.5%. Under a train-once protocol over five datasets and several mask sources, the 2.6M-parameter variant gains 4.2 mean \Delta IoU against 2.0 for its control, and a variant built on frozen DINOv2 features matches the strongest foundation-model refiner with about a seventh of its resident parameters while staying ahead of its own control. On mask generators never seen in training, the gain is 2.0 F^\omega_\beta points against 0.6 for the control. Zeroing individual messages shows where the corrections come from: the pairwise messages mostly fix boundaries, the region messages mostly fix non-boundary errors. Code and supporting materials will be publicly released.
[CV-237] A Multi-Dataset Benchmark of YOLO-Based Weed Detection in Precision Agriculture
链接: https://arxiv.org/abs/2609.33991
作者: Hristina Zdraveska,Vlatko Spasev,Ivica Dimitrovski,Ivan Kitanovski,Petre Lameski
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Weed detection is an important component of precision agriculture, enabling site-specific weed management and reducing unnecessary herbicide use. Although deep learning methods have achieved strong results for crop and weed detection, many studies rely on single-dataset evaluation, making it difficult to assess robustness across different agricultural domains. This paper presents a multi-dataset benchmark of deep object detectors for weed detection in precision agriculture, with a focused evaluation of YOLO26 models. We evaluate nano, small, and medium variants on seven public weed-detection datasets covering different crops, weed species, field conditions, acquisition setups, and annotation protocols. The models are compared in terms of detection accuracy, model complexity, inference latency, FPS, and model size. In addition to in-dataset evaluation, we investigate cross-domain generalization using a unified one-class weed setup and evaluate multi-source training using the combined training subsets from all datasets. The results show that YOLO26 achieves strong in-dataset performance, with YOLO26m obtaining the highest average accuracy and YOLO26s providing the best practical accuracy-efficiency trade-off. However, cross-domain performance decreases substantially, with YOLO26s dropping from an average in-domain mAP _50:95 of 0.603 to 0.148 in the off-domain setting. Multi-source training improves performance on several datasets, but does not fully eliminate domain shift. Overall, the benchmark highlights the importance of dataset diversity, domain similarity, and target-domain adaptation for robust weed detection in real-world precision agriculture applications.
[CV-238] st-Time Spatial Reasoning for Robot Manipulation Using Generative Real-to-Sim IROS2026
链接: https://arxiv.org/abs/2609.33982
作者: Ivan Kapelyukh,Yafei Hu,Ran Gong,Brandon May,Tushar Kusnur,Laura Herlant,Karl Schmeckpeper,Edward Johns,Xiaohan Zhang
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: IROS 2026
Abstract:Spatial reasoning is fundamental to general robot intelligence, as it enables robots to complete long-horizon tasks involving multi-object interaction. We introduce Simify, a training-free, test-time framework that performs explicit spatial reasoning via massively parallel physics simulation. From a single RGB-D image of a scene, Simify reconstructs simulation-ready assets leveraging 3D generative models and vision-language models. Then given a task specified by a reward function (e.g., build the tallest tower), Simify launches thousands of parallel rollouts in simulation and performs an evolutionary search to optimize object arrangements, typically converging within seconds. We conduct quantitative experiments on real-robot hardware to demonstrate the ability of our framework to execute complex object rearrangement tasks end-to-end with previously unseen objects. Results show that our framework outperforms prior work on foundation models for spatial reasoning by effectively exploiting large-scale parallel simulation during inference, and also highlight the importance of complete and accurate geometry for successful sim-to-real transfer.
[CV-239] Gaussian Splatting-based Volumetric Video Compression with Sparse 4D Anchors
链接: https://arxiv.org/abs/2609.33969
作者: Ge Gao,Siyue Teng,Chanqgi Wang,Fan Zhang,Nantheera Anantrasirichai,Jui Chiu Chiang,Wen-Hsiao Peng,David Bull
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Immersive video communication requires photorealistic, render-efficient, and compact dynamic scene representations. 3D Gaussian Splatting (3DGS) offers a promising representation, but dynamic 3DGS remains difficult to compress due to dense primitives and spatiotemporal redundancy. Anchor-based formulations improve compactness with sparse scaffolds that share geometry and appearance across primitives. However, existing designs often rely on deforming a single canonical scaffold and condition each primitive on its associated anchor in isolation, limiting their ability to handle non-local dynamics and disocclusion while under-exploiting inter-anchor correlations, particularly in motion- or texture-dense regions. To address these limitations, we propose SAGA, a volumetric video codec built upon Sparse Anchor-assisted GAussian splatting representations. SAGA represents dynamic 3D scenes using hierarchically organized sparse 4D anchors, where coordinate-based INR decoders generate fine anchors and Gaussian primitives from inter-anchor interpolations, enabling compact parameter sharing across spatiotemporal structures. For long-range dependencies among unstructured anchors, we further introduce fixed-size memory slots with orthogonality-informed updates for accurate entropy-context modeling. Experiments show that SAGA achieves strong rate-distortion performance against GIFStream, with PSNR BD-rate reductions of 80.39% and 83.94% on Neu3D and MPEG MIV, respectively.
[CV-240] SymNetPro: LOS-Aware Directional Multi-Transmitter Localization from Sparse Radio Observations
链接: https://arxiv.org/abs/2609.33964
作者: Lyuzhou Ye,Heng Fan,Yan Huang
类目: Information Theory (cs.IT); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Directional multi-transmitter localization from sparse received-power observations is difficult because the receiver observes only the source-unresolved aggregate field: multiple directional sources superpose, building blockage fragments their visible regions, and stronger sources can mask weaker ones. We present SymNetPro, which retains the dual-task radio-map reconstruction and localization backbone of SymNet and adds two targeted components. First, a sparse line-of-sight (LOS)-aware attention bias injects obstruction-aware spatial relations into selected token interactions. Second, transmitter-drop augmentation recomposes training scenes after removing one sample-supported transmitter, exposing the model to controlled source-cardinality variation. Experiments on directional ray-traced urban environments show substantially lower OSPA than representative localization baselines under extreme sparse sampling, with consistent gains under measurement noise and increasing transmitter count. A transmitter-specific evidence analysis further shows that remaining misses concentrate in regimes where the target contributes little distinguishable power to the aggregate observation.
[CV-241] EpiTransfer: Sparse Training-Free Long-Range Depth Estimation from Temporal Monocular Aerial Frames
链接: https://arxiv.org/abs/2609.33939
作者: Diksha Aggarwal,Rutvik Dagadkhair,Sanjana Srivastava,Bradley Denby,Kevin Kochersberger
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reliable 3D spatial understanding is essential for autonomous navigation, obstacle avoidance, and scene reconstruction. While state-of-the-art learned depth estimation techniques achieve high accuracy in-distribution, they often generalize poorly to novel viewpoints and altitudes. This paper presents a geometrically derived, training-free depth estimation method using epipolar transfer with only two monocular images and camera pose estimates. By leveraging camera motion to synthesize a virtual stereo pair with a freely chosen baseline, our approach transforms temporal correspondence into a stereo triangulation task while mitigating geometric degeneracies inherent to direct two-view triangulation. Validated across outdoor drone flights (to a maximum range of approximately 90,m) and indoor OptiTrack environments against LiDAR ground truth, the method achieves an indoor AbsRel of 0.092 and \delta 1.25 of 0.940, comparable to direct triangulation (AbsRel 0.073) while retaining valid depth over a larger fraction of challenging scenes, and substantially outperforms off-the-shelf learning-based baselines such as ZoeDepth (AbsRel 0.225) and Depth Anything V2 (AbsRel 0.570), which are not trained or fine-tuned for this domain, with no training data required.
[CV-242] st-Time Generalized Category Discovery
链接: https://arxiv.org/abs/2609.33937
作者: Shambhavi Mishra,Omprakash Chakraborty,Julio Silva-Rodriguez,Ismail Ben Ayed,Marco Pedersoli,Jose Dolz
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Test-Time Adaptation (TTA) and Generalized Category Discovery (GCD) are traditionally treated as disjoint problems: the former adapts models to domain shift assuming all test classes are known, while the latter discovers novel categories assuming labeled training data for known classes. However, real-world deployment rarely fits either setting. Motivated by this gap, we introduce Test-Time Generalized Category Discovery (TT-GCD), a unified and more realistic scenario where a vision-language model must adapt to distribution shifts, classify known categories using only textual supervision, and discover novel categories, all during test time and without access to labeled data. To address this challenging scenario, we propose PACT (Prototype Assignment for Category discovery at Test time), a fully unsupervised framework that casts known-class recognition and novel-class discovery via prototype assignment. PACT first re-aligns shifted visual features with the text-derived class representations of the VLM using confident zero-shot predictions. Known and novel categories are then both represented by prototypes in the visual embedding space, estimated from the unlabeled test stream, and each test image is assigned to the category whose prototype is most similar to its visual feature. Extensive experiments across corruption and domain-shift benchmarks demonstrate that PACT outperforms adapted state-of-the-art TTA and GCD methods, effectively bridging the gap between adaptation and discovery.
[CV-243] Video Ergo Genero: Unifying Video Tasks via Spatiotemporal Analogy
链接: https://arxiv.org/abs/2609.33935
作者: Chia-Hsiang Kao,Belinda Zeng,Bharath Hariharan,Menglin Jia
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project page: this https URL
Abstract:Adapting video models to new tasks typically requires dedicated data curation and fine-tuning. While visual analogy provides a training-free alternative by specifying tasks in-context, it remains restricted to the image domain. To explore whether analogy-based methods can unify diverse video tasks and generalize to out-of-distribution scenarios, we introduce ViGeo, a framework that extends visual in-context learning to the video domain via spatiotemporal canvas completion. Evaluated on a diverse task taxonomy with a strict train-test split, ViGeo generalizes to unseen video manipulations and zero-shot modalities (e.g., event cameras). Finally, we identify task internalization, where a query format associated with a pretrained task overrides the demonstration, and show that this shortcut can be removed with a small amount of task-unrelated data, highlighting the need to decorrelate prompt format from task identity.
[CV-244] ArticulateArena: A Metric for Articulated Kinematics
链接: https://arxiv.org/abs/2609.33931
作者: Yumeng He,Yongfei She,Huanyu Chen,Chun Yuan,Peihao Li,Joseph Masterjohn,Yin Yang,Ying Jiang,Chenfanfu Jiang
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 32 pages, 14 figures, Project page: this https URL
Abstract:Modern methods reconstruct or generate simulation-ready articulated objects, predicting not only their geometry but also how their parts are connected and allowed to move. Evaluating the geometry is straightforward, but evaluating the predicted articulation is not, because articulation specifies a motion rather than a shape, and there is no agreed distance between two motions. More specifically, existing protocols score joint type, axis direction, origin, and motion limits separately, although these parameters jointly describe a single physical motion, and the same motion can be written as different parameter values. As a result, a joint can score maximally wrong against an equivalent encoding of itself, and several component errors are ill-conditioned or undefined exactly where predictions become accurate. We propose ArticulateArena, a representation-invariant counterpart of Chamfer distance for articulation that compares the motions one-DOF joints induce rather than the parameters that encode them. It represents each joint by the unordered pair of its Lie-algebra endpoint twists, and we prove that the resulting quotient distance is a metric. It unifies fixed, revolute, prismatic, and helical joints, brings continuous joints into the same score through a compactification, and reads as the RMS motion of the moving part in meters when weighted by its mass distribution. A motion-aware tree edit distance lifts the metric to full kinematic trees, pricing structural errors such as spurious or missing joints in the same motion units as joint errors, and for a fixed inner product it remains a metric on trees up to relabeling. Alongside the metric we release ArticulateArena-20K, a new suite of 19,977 articulated objects with verified kinematics, and we re-evaluate published reconstruction methods on it under the new metric. Project page: this https URL
[CV-245] Preserving DEG Rankings for Gene Discovery in Histology-Based Spatial Gene Expression Prediction NEURIPS2026
链接: https://arxiv.org/abs/2609.33928
作者: Kaito Shiku,Kazuya Nishimura,Yasuhiro Kojima,Ryoma Bise
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to NeurIPS 2026
Abstract:Predicting spatial gene expression from histology images could scale spatial transcriptomics (ST) to image-only cohorts, but conventional histology-based ST prediction is trained and evaluated mainly by per-gene spatial-profile reconstruction. This objective is misaligned with a key downstream use of ST: differentially expressed gene (DEG) discovery, where genes are ranked for a biological or morphology-defined contrast by evidence of between-group expression differences. We formulate image-based differential expression ranking (IDER), which asks whether predicted expression profiles preserve the contrast-specific ranked gene list obtained from measured profiles. IDER compares gene rankings induced by differential-expression statistics, rather than raw expression magnitudes or per-gene spatial correlations. We further introduce a differentiable IDER objective that aligns these statistics across genes and can be trained with morphology-derived proxy contrasts without predefined biological group labels. Experiments on public ST datasets show improved DEG-ranking agreement and pathway-enrichment overlap over conventional reconstruction objectives, including morphology-derived and pathologist-annotated tissue-region evaluations.
[CV-246] JIVE: Jacobian-Informed Volume Expansion for Diverse Generative Sampling
链接: https://arxiv.org/abs/2609.33906
作者: Guangxun Zhang,Brian Cai,Boxuan Zhang,Chao Chen,Ruixiang Tang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Generative models often suffer from mode collapse and limited sample diversity. While prior works attempt to mitigate this by jointly generating a batch of samples and repelling their trajectories, these heuristics do not explicitly maximize the diversity of the resulting endpoints. We introduce JIVE, a training-free framework that enhances generative diversity by injecting velocity perturbations aligned with the leading right singular subspace of the generator’s endpoint Jacobian. By leveraging this local geometric structure, JIVE provably maximizes endpoint diversity while preserving sample quality. To maintain practical efficiency, we compute these perturbation directions via matrix-free iterations rooted in classical numerical linear algebra, requiring only a small computational overhead. Across different benchmarks, JIVE boosts both pixel and feature-level diversity in few-step and one-step generation.
[CV-247] Residual-Stream Burden Shapes Representation Learning in Diffusion Transformers
链接: https://arxiv.org/abs/2609.33895
作者: Tongtong Liang,Siqi Kou,Ziqiao Xi,Esha Singh,Kun Zhou,Zhijie Deng,Alexander Cloninger,Yu-Xiang Wang,Rahul Parhi
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Under review
Abstract:In diffusion-based generation, a neural network can be trained to predict the clean data, the noise, or the velocity from a noisy input. These prediction targets are interconvertible and describe the same generative process, yet plain Diffusion Transformers operating on large pixel patches succeed with clean prediction and fail with noise or velocity prediction. We argue that this asymmetry arises because noisy targets require the residual stream to preserve noise-dependent input variation through depth for the final readout, forcing subsequent layers to compute on noisy representations. A spectrally concentrated clean target imposes a lighter demand, leaving greater freedom to organize hidden representations for subsequent computation. We call this preservation requirement residual-stream burden and show how it shapes representation learning in Diffusion Transformers. Controlled experiments indicate that the exploitable structure is spectral concentration in patch space and that the bandwidth of the persistent residual state is a key resource for noisy prediction. We further show that this account is consistent with recent decoupled pixel-space architectures, whose diverse designs all reduce the residual-stream burden on the main pathway. To examine this understanding from a complementary direction, we expand and reorganize the residual-stream bandwidth directly, introducing Spatially Indexed Hyper-Connections (SiHC) that reach FID 1.71 on ImageNet 256^2 . Together, these results identify residual-stream burden as a mechanism through which prediction targets and architecture jointly shape representation learning in Diffusion Transformers.
[CV-248] ReDrive: Shaping Representations with World Modeling for End-to-End Driving
链接: https://arxiv.org/abs/2609.33854
作者: Yueting Zhu,Shaoyu Chen,Yuehao Song,Hui Sun,Qian Zhang,Wenyu Liu,Xinggang Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 15 pages,7 figures,10 tables
Abstract:Driving policies require capabilities of scene understanding and future evolution prediction. To achieve this goal, current end-to-end models typically construct complex perception-planning pipelines or introduce world models that explicitly predict future states, resulting in a complex system architecture. Inspired by the transferability of general-purpose visual representations, we argue that combining sufficiently strong visual representations with representation world modeling can support effective planning without relying on complex inference-time auxiliary modules. Based on this insight, we present ReDrive, an end-to-end driving framework that strengthens planning-oriented visual features via future representation prediction. To achieve this, ReDrive adopts a three-stage training pipeline consisting of driving video pretraining, joint world-modeling and planning training, and planner adaptation. This yields a strong planning-oriented representation and a high-performance planner, while requiring neither auxiliary perception modules nor future prediction at inference time. Experiments on NAVSIM demonstrate strong performance, achieving 91.0 PDMS on NAVSIM v1 and 90.8 EPDMS on NAVSIM v2. These results show that shaping representations with world modeling is sufficient to enable high-performance end-to-end planning while retaining a simple encoder-planner inference pipeline.
[CV-249] DeltaSeek: Toward Active Perception in Evolving Construction Environments IROS
链接: https://arxiv.org/abs/2609.33836
作者: Sanjay Acharjee,Md Nazmus Sakib
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 4 pages, 3 figures, IROS Workshop 2026
Abstract:Construction environments evolve continuously, causing large geometric changes that degrade static mapping and registration performance. This necessitates active perception, where robots deliberately select sensing configurations to resolve the environment’s current state. We present DeltaSeek, an initial framework toward active perception in evolving built environments. While our broader objective is a system that reasons about where, how, and when to observe, this paper addresses a critical prerequisite: how a robot’s sensing embodiment constrains the observations it can acquire. We formalize an embodiment’s permissible observation set and evaluate with a Husky A300 equipped with a UR5e on an IFC-derived benchmark under chassis-mounted and wrist-mounted RGB-D configurations, scoring observations by geometric visibility and effort by drivable distance. In a room-scale scene with eight controlled changes spanning four observability conditions, exhaustive evaluation over 240 permissible base poses and five arm postures shows that two changes admit no chassis viewpoint whatsoever, while the wrist camera resolves both. For changes observed by both embodiments, the median base travel is 6.0 ~m for the wrist camera and 15.2 ~m for the chassis camera. These results distinguish sensing limitations from acquisition costs, clarifying whether an observation is impossible or simply requires more travel.
[CV-250] CLIMB-flow: Coupled Linear Inverse posterior sampling via Multiscale-Based flow ICASSP2027
链接: https://arxiv.org/abs/2609.33834
作者: Zeqiu Yu,Ruizhi Yuan,Mathews Jacob
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 5 pages, 3 figures, 1 table. Submitted to ICASSP 2027
Abstract:Diffusion models are now widely used in Bayesian inverse problems in imaging as priors, where latent diffusion models are often used for larger scale problems to keep the computational complexity and model-size manageable. Unfortunately, the auto-encoder based compression results in loss of spatial detail. In addition, the optimization is converted to a non-linear problem. In this paper, we introduce a posterior sampling algorithm customized for the pyramidal/cascaded architecture, which relies on a coarse to fine hierarchical strategy to generate images in the pixel domain. We present CLIMB-Flow which alternates between three steps: an end-point estimation from the current coarse and noisy image, data-consistent update of the clean image, and re-noising it back to the level the network expects. Together these steps sample the posterior at that scale using an approximate Gibbs sampling from two conditional distributions. Experiments on ImageNet, CelebA, AFHQ and fastMRI span inpainting, deblurring, super-resolution and accelerated MRI, with PSNR gains of 1.37-7.66 dB over the strongest competing method on CelebA and pixel-domain reconstruction up to 512x512.
[CV-251] One Attack to Fool Them All: Highly Transferable Black-Box Adversarial Attacks on Frontier MLLM s
链接: https://arxiv.org/abs/2609.33833
作者: Sen Nie,Jie Zhang,Zhongqi Wang,Shiguang Shan,Xilin Chen
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 25 pages, 16 figures, 10 tables
Abstract:Adversarial attacks have long posed a fundamental threat to machine learning systems. As multimodal large language models (MLLMs) rapidly evolve and become widely deployed, assessing their vulnerability to such attacks is essential for their safe use. In this work, we investigate whether a single adversarial image can consistently mislead diverse frontier MLLMs in black-box settings. We propose O-Attack, a highly transferable black-box attack framework. This framework builds on our insight that surrogate models contain a broad, high-level, cross-modally aligned semantic space. This space extends beyond final-layer outputs and provides multiple semantically consistent representations that remain underexploited by existing attacks. Within this space, O-Attack anchors aligned representations, progressively broadens semantic conditions, and optimizes perturbations through semantic consensus to promote consistent target alignment. By fully exploiting this space with the same surrogate models as M-Attack, O-Attack raises attack success rates on GPT-5.4 (29.1% to 77.2%), Claude-4.6 (42.8% to 81.6%), and Gemini-3.1 (38.2% to 80.9%). Extensive experiments across 24 MLLMs show that O-Attack outperforms six state-of-the-art methods in black-box transferability, with consistent effectiveness across prompts and improved efficiency and imperceptibility. This work exposes the practical safety risks posed by black-box adversarial attacks against frontier MLLMs, underscoring the need for more rigorous robustness evaluation and more effective defenses.
[CV-252] Augmenting Visual Anomaly Detection with Automated Interpretability
链接: https://arxiv.org/abs/2609.33818
作者: Antonio De Santis,Arsenio Leo,Marco Brambilla
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Preprint
Abstract:Visual anomaly detectors identify deviations from known-normal data, but their anomaly signals may mix evidence of actual anomalies with benign visual variation. We investigate whether automated interpretability can augment visual anomaly detectors by identifying and intervening on different components of this signal. We decompose PatchCore nearest-normal residuals into sparse features using Sparse Autoencoders (SAEs), and provide high-activation and contrastive non-active examples to a Multimodal LLM, which describes each feature and labels it as anomaly, distractor, or uncertain. These labels guide interventions in the SAE hidden representation, where distractor features are suppressed and anomaly features amplified. The edited representation is then used to reconstruct patch embeddings, which are rescored with PatchCore. Across 40 categories from four benchmarks, applying both interventions jointly improves macro-average image-level AUROC from 0.8724 to 0.8857 on source data and from 0.8066 to 0.8210 under synthetic corruptions. On three additional RobustAD categories with real acquisition shifts, the same interventions improve AUROC from 0.8745 to 0.9056 on source data and from 0.6069 to 0.6599 under real acquisition shifts. Finally, individual feature interventions across all 43 categories show that the MLLM labels are aligned in aggregate with how features differently affect normal and anomalous images.
[CV-253] Eyes on the Road: A Naturalistic Comparison of MTW Rider Gaze in Urban Indian Traffic
链接: https://arxiv.org/abs/2609.33811
作者: Prerak Srivastava,Bhaiya Vaibhaw Kumar,Kavita Vemuri
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Motorized two-wheelers (MTW) dominate Indian roads but remain underrepresented in driver behavior research. This study presents the first large-scale analysis of MTW driver gaze behavior in naturalistic, heterogeneous urban traffic, using the \textitmyEye2Wheeler dataset. A semantic segmentation pipeline (YOLOv11 + SAM2) was used to extract object-level gaze metrics under two attention modes: direct gaze (foveal overlap) and central vision (parafoveal monitoring). Results reveal a functional division: central vision supports broad monitoring, while direct gaze enables brief, selective sampling. Novice riders exhibit road-anchored scanning, returning to the road between object fixations, while experienced riders form longer chains of attention across multiple objects. The findings suggest that experience primarily refines temporal rhythm rather than altering allocation strategy and reduces object-class effects in gaze patterns. These findings offer new insight into MTW attention structures and inform future work on behavior modeling and safety systems.
[CV-254] MinkowskiPE: Minkowski Positional Encoding for Spatiotemporal Perception
链接: https://arxiv.org/abs/2609.33804
作者: Yuhao Li,Louie Hong Yao,Tianyi Shi,Hanqun Cao,Hongxia Hao,Zhen Zhao,Shengchao Liu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Computational Physics (physics.comp-ph)
备注: 21 pages, 4 figures, 4 tables
Abstract:Modeling spatiotemporal coupling is a key challenge in building physical intelligence across scales, from microscopic to macroscopic. Existing models capture such structure broadly through physics-motivated dynamical formulations or learning-motivated architectures. The former provide stronger priors but may constrain flexibility, whereas the latter are more flexible but leave the spatiotemporal coupling largely implicit. We therefore seek an approach that combines flexible learning with an explicit geometric bias for jointly modeling time and space. To this end, we propose Minkowski Positional Encoding (MinkowskiPE), which uses joint temporal and spatial coordinates to parameterize Lorentz transformations applied to query and key features. With MinkowskiPE, the query-key attention score depends on position only through the relative spacetime displacement between the two tokens and is therefore invariant to global translation of the coordinates. This paradigm retains the standard dot-product attention interface and remains compatible with efficient attention implementations. We evaluate MinkowskiPE on microscopic molecular dynamics and macroscopic video prediction tasks, achieving the best results on all nine multi-trajectory molecular evaluations and reducing KTH video-prediction MSE by 9.9% relative to the best baseline while using roughly one-tenth as many parameters.
[CV-255] M3-Score: Fidelity Memorization and Coverag e as Separate Axes for Evaluating Generative Radiology Image Models
链接: https://arxiv.org/abs/2609.33769
作者: Sathiyamohan Nishankar,Pubudu Sanjeewani,Asanka Perera
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Quantitative evaluation of generative models for radiology remains challenging. Clinically relevant structures are often small and infrequent, feature spaces learned from natural images may represent them poorly, and a single summary score cannot distinguish limited fidelity from limited diversity. This study proposes the Medical Multi-axis Maximum Mean Discrepancy score (M3-Score), an evaluation framework based on RadioDINO-s16, a frozen vision transformer pretrained on radiology images. M3-Score reports three complementary axes computed at pre-specified encoder depths: \emphfidelity, measured by an unbiased multi-bandwidth radial basis function (RBF) MMD ^2 at the final block; \emphmemorization, measured by nearest-neighbor distances at 75% depth; and \emphcoverage, defined as the fraction of real images with a generated neighbor within their k -nearest-neighbor radius at 33% depth. Reference sets are sampled across subjects to limit the influence of correlated slices. On BraTS brain MRI, the fidelity axis ordered five comparison sets of increasing severity (Spearman \rho = 1.00 ), and a subject-disjoint real set yielded \mathrmMMD^2 = 0 (permutation p = 1 ). An unconditional denoising diffusion probabilistic model achieved \mathrmMMD^2 = 0.073 (95% confidence interval [0.071, 0.080] ) but covered only 38% of the real distribution. Under progressive mode dropping, k -NN manifold recall increased at all twelve encoder blocks, whereas the proposed coverage estimator decreased monotonically ( \rho = -1.00 ). RadioDINO-s16 features separated real brain MRI from generated samples with a ROC-AUC of 0.819, compared with 0.555 for InceptionV3 and 0.582 for CLIP. Across a twentyfold range of sample sizes, the mean M3 value varied by a factor of 1.05, compared with 2.52 for the Fréchet Inception Distance.
[CV-256] ENet-GP: Unified Document Image Restoration
链接: https://arxiv.org/abs/2609.33758
作者: Sujal Burad,Aakanksha,A. N. Rajagopalan,Sumit Shekar
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reliable document digitization in uncontrolled capture settings is challenging because real images exhibit multiple interacting degradations rather than a single isolated distortion. Documents thus captured are affected simultaneously by geometric distortions, like page warping, as well as photometric degradations such as non-uniform illumination, and blurring. However, most existing approaches address these factors independently and are evaluated on benchmarks containing only one distortion type, limiting their real-world applicability. We introduce GutenDoc, a large-scale dataset of high-resolution dense-text documents with physically grounded compound degradations. Using physics-based rendering, our dataset jointly models geometric warping and diverse photometric effects, enabling systematic evaluation under realistic capture conditions. We further propose a unified restoration framework that jointly corrects geometric and photometric distortions within a single-network and single-training setup, without the need for degradation-specific retraining or sequential inference passes. Extensive experiments show that our method remains competitive on established single-distortion benchmarks while substantially improving robustness under compound degradations, providing a practical solution for real-world document digitization.
[CV-257] Constrained Edit Fields for Training-Free Flow Editing
链接: https://arxiv.org/abs/2609.33735
作者: Jingxuan Kang,Yinsong Wang,Che Liu,Chen Qin
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 13 pages, including appendix
Abstract:Text-guided image editing aims to perform a desired edit while preserving source content unrelated to it. Pretrained rectified-flow models enable training-free editing of real images through modifications to their sampling trajectories. However, responses at locations unrelated to the desired edit can still accumulate along the editing trajectory and become visible in the final result. To overcome this, we propose Constrained Edit Fields (CEF), which assigns each spatial location a continuous edit responsibility that quantifies its relevance to the desired edit. CEF estimates edit responsibility directly from the source image when the relevant content is present. For edits whose target content is absent from the source, CEF first generates an unconstrained proposal to reveal its realized spatial support and then estimates responsibility from that proposal. At each editing step, CEF decomposes the base edit field into prompt-induced and trajectory-induced components, enabling edit responsibility to preserve instruction-relevant updates while suppressing unintended trajectory-induced changes. Evaluated on all 700 PIE-Bench examples, CEF achieves state-of-the-art Structure Distance, background LPIPS, and background MSE with both Stable Diffusion 3.5 Medium and FLUX, while retaining competitive instruction alignment. On Stable Diffusion 3.5 Medium, it reduces these metrics over the previous best results by 10.2%, 21.2%, and 48.0%, respectively.
[CV-258] GeoShrink: Accelerating Diffusion Transformers with Two Lines of Code
链接: https://arxiv.org/abs/2609.33723
作者: Haosen Li,Wenshuo Chen,Shaofeng Liang,Lei Wang,Bowen Tian,Yutao Yue
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Diffusion transformers incur substantial inference cost through repeated model evaluations along a sampling trajectory. We introduce GeoShrink, a training-free acceleration method that retains the original solver grid while evaluating the model only at a prescribed set of anchors. At skipped stages, GeoShrink predicts the solver-facing output by adding a geometrically retained fraction of the latest observed innovation to the most recent exact output. We derive this rule from chordal tangent transport and round-trip line projection, and establish a geometric anchor-spacing principle that minimizes the largest adjacent gap expansion under fixed coverage and first span. The analysis characterizes the geometric closure and propagation of prediction errors without assuming access to future model outputs. Experiments cover image, video, motion, and audio generation, together with adapted 3D backends. At approximately 5\times acceleration, GeoShrink improves FLUX PSNR by 3.10 dB over the strongest listed baseline. On HunyuanVideo, it achieves a reported 4.99\times speedup and improves ChronoMagic-Bench-150 PSNR by 5.44 dB over the strongest listed fidelity baseline. Comparisons at fixed evaluation budgets further show substantial gains on motion, audio, music, and 3D generation.
[CV-259] Revisiting Diffusion Fine-Tuning for Unsupervised Domain Adaptation NEURIPS2026
链接: https://arxiv.org/abs/2609.33716
作者: Xuan Qi,Yi Wei,Daniele Berardini,Vito Paolo Pastore,Vittorio Murino
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the 40th Conference on Neural Information Processing Systems (NeurIPS 2026)
Abstract:Diffusion-based unsupervised domain adaptation (UDA) improves cross-domain transfer by generating target-specific synthetic data for downstream adaptation. Existing methods are largely designed for single-target adaptation: when a model trained on one labeled source domain must be adapted to multiple unlabeled target domains, they typically require separate diffusion fine-tuning for each source–target pair, causing training, storage, and deployment costs to grow with the number of targets. In this paper, we study multi-target data generation for diffusion-based UDA, where a single source-guided diffusion fine-tuning process is reused to generate target-specific synthetic data for multiple target domains. We propose MUSE (Multi-target UDA-oriented Synthesis with Efficient diffusion fine-tuning), a decoupled adaptation framework that separates source-supervised semantic adaptation from target-specific style adaptation. MUSE uses a shared semantic branch updated by labeled source data and target-private style branches specialized to individual target domains, enabling target-specific generation while avoiding repeated source-guided fine-tuning for each target. Experiments on standard UDA benchmarks show that MUSE achieves a stronger accuracy–efficiency trade-off than repeated per-target diffusion adaptation, reducing diffusion fine-tuning cost while improving average target-domain accuracy. The project page is available at this https URL.
[CV-260] Prompt-Anchored Residual Adaptation for Biomedical Vision-Language Models
链接: https://arxiv.org/abs/2609.33701
作者: Jingxuan Kang,Qianying Yue,Che Liu,Chen Qin
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 18 pages, including appendix
Abstract:Pretrained biomedical vision-language models achieve strong zero-shot performance in biomedical image classification. However, downstream biomedical classification often depends on subtle visual differences between classes that may not be fully captured by pretrained representations. Few-shot adaptation addresses this mismatch by optimizing a task-specific predictor on a small labeled support set. Because the selected examples capture only part of the visual variation within the target classes, the adapted predictions can depend strongly on their composition. We propose Prompt-Anchored Residual Adaptation (PARA), which retains the frozen prompt prediction as a support-invariant semantic anchor and incorporates a visual prediction learned from the support set through an anchor-relative residual. The residual step is computed in a closed form from frozen support embeddings using anchor discrepancy and support agreement. Support-set dependence also limits evaluation: comparisons are fair within a shared draw but remain conditional on its composition. To obtain more reliable comparisons, we introduce a repeated-support protocol that separates support-selection variation from optimization randomness and reports both average and worst-20% performance. PARA achieves state-of-the-art performance in both few-shot classification and base-to-novel generalization.
[CV-261] Seeing and Solving Are Not Enough for Vision-Language Models
链接: https://arxiv.org/abs/2609.33694
作者: Ziheng Wang,Mingxuan Xie,Yilin Liu,Dayan Wu,Yang Li,Pengwen Dai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 30 pages, 9 figures, 20 tables
Abstract:Vision-language models (VLMs) answer visual questions by combining visual information extraction with downstream problem solving. We investigate a fundamental question: Does an incorrect answer necessarily reflect a failure in visual extraction or problem solving? A model may succeed at both abilities when tested separately yet still fail on the original multimodal question, a distinction that overall answer accuracy cannot reveal. To study this, we perform a question-level empirical analysis across multiple VLMs and visual domains. We define an exactly scorable task state (i.e., the visual information sufficient to solve a question) and use it to test whether the same model can extract the required state, solve the question from the ground-truth state, and answer the original multimodal question. We find that composition failures, where extraction and solving both succeed but direct answering fails, account for 17.7% to 75.6% of direct-answering errors across multiple VLMs and datasets. To address this failure mode, we introduce a simple yet effective method, termed State Realization Tuning (SRT). SRT fine-tunes LoRA adapters attached to the language-model layers while keeping the pretrained VLM weights frozen. It trains the model to output the ground-truth task state before the final answer in a single autoregressive response. SRT improves over standard supervised fine-tuning by 1.7 to 14.1 percentage points and repairs 92.5% to 98.1% of diagnosed composition failures. A single LoRA adapter trained with SRT also improves performance across substantially different task-state structures. Our work shows that having both visual extraction and problem-solving capabilities does not guarantee correct multimodal answering. Requiring the model to first output the visual information needed to solve the question can help bridge this gap.
[CV-262] Resource-Aware Parameter-Efficient Model Adaptation for Onboard High-Dimensional Data
链接: https://arxiv.org/abs/2609.33687
作者: Qiyang Zhang,Xinhao Li,Lei Shi,Zheng Lin,Jinfeng Wen,Ao Zhou,Shangguang Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Distributed, Parallel, and Cluster Computing (cs.DC); Networking and Internet Architecture (cs.NI)
备注: 22 pages, 8 figures
Abstract:Onboard satellite models often require frequent updates, but the weights adapted to earlier data distributions can quickly become outdated. However, updating large-scale model parameters in orbit presents significant challenges due to the limited uplink bandwidth of Low Earth Orbit (LEO) satellite systems, particularly for hyperspectral satellite imagery, where high-dimensional spectral-spatial inputs lead to increased model size and update costs. Existing full fine-tuning methods are thus expensive to retrain and difficult to deploy under strict communication constraints. To address this challenge, we propose NE-LoRA, a parameter-efficient adaptation framework for bandwidth-constrained onboard hyperspectral model updates. NE-LoRA combines a primary low-rank branch with a nonlinear auxiliary branch to capture both global update trends and complex spectral-spatial variations. Additionally, we introduce a differentiated training strategy for multi-matrix adapters, motivated by the asymmetric initialization and gradient dynamics of different adapter matrices. Experiments on four hyperspectral datasets and three representative backbone models demonstrate that NE-LoRA consistently outperforms LoRA-based baselines and remains competitive with, and in several cases superior to, full fine-tuning. Across the evaluated settings, NE-LoRA updates only a small fraction of the total parameters on average while preserving low deployment overhead, offering a favorable accuracy-communication trade-off for onboard hyperspectral adaptation.
[CV-263] MAD-Guard: Controlled Study of Autoregressive Generation versus Direct Decision Interfaces for Closed Multimodal Forensic Tasks
链接: https://arxiv.org/abs/2609.33683
作者: Hao Chen
类目: Computer Vision and Pattern Recognition (cs.CV); Cryptography and Security (cs.CR)
备注: 7 pages, 3 figures, 7 tables
Abstract:When should multimodal foundation models generate tokens, and when should they directly output a decision? We present MAD-Guard, a controlled study of output-decision interfaces for closed multimodal forensic tasks. Once a multimodal representation is computed, is autoregressive generation necessary for closed forensic decisions with high input complexity but low output entropy? Under a matched Qwen3-VL-8B backbone, 2,400 FakeClue training samples, and LoRA budget ( r=16, \alpha=32 ) on Huawei Ascend 910C NPUs, we evaluate a progression of decision interfaces (AR-SFT [generate] \to Logit Slice \to Binary Direct Head \to +choice \to +act \to CLM-Head) and decompose latency into backbone representation (53.12 ms), 151,643-way vocabulary projection (+85.04 ms \to 138.16 ms), and decoding (+248.26 ms \to 386.42 ms). Under 1-to-1 binary supervision ( \mathcalL_\mathrmBCE ), a Binary Direct Head cuts latency by 2.60\times - 7.27\times (53.12 ms) and lowers calibration error by 1.88\times (ECE = 0.0450 vs. 0.0845), with a -1.80% accuracy trade-off (93.10% vs. 94.90%; 0.9795 vs. 0.9871 ROC-AUC) from forfeiting token priors. Gains above AR-SFT arise either from multi-task attribution and uncertainty gating (+choice+act: 96.44% accuracy, 0.9940 ROC-AUC, 0.0187 ECE at 53.71 ms) or from a disaggregated contrastive head (CLM-Head: 96.55% binary and 96.44% multi-task accuracy, 0.0166 ECE, 98.79% 7-class attribution at 54.42 ms) retaining semantic priors without token decoding. Across 5,000 out-of-sample images from five benchmarks, our framework excels on synthetic, camouflage, and document forgeries (96.44% GenImage, 97.73% Chameleon, 91.84% Doc) while showing a clear boundary on compressed face manipulation (FF++ ROC-AUC = 0.5913).
[CV-264] When Noise Meets Long-Tail: Feature-Threshold Dual Calibration for Robust Pseudo-Labeling NEURIPS2026
链接: https://arxiv.org/abs/2609.33668
作者: Ping Guo,Zhiqi Huang,Xinran Li
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to NeurIPS 2026 Main Track. Code: this https URL
Abstract:Pseudo-labeling has become a cornerstone of learning from unlabeled data in semantic segmentation. Yet its effectiveness drops sharply in real-world scenarios where strong imaging noise and long-tailed class distributions occur together. We trace this failure to a vicious cycle of pseudo-label degradation. Imaging noise entangles foreground and background features, lowering prediction confidence across all classes, while long-tailed distributions leave tail classes with far fewer training samples and inherently lower confidence. Under fixed high-threshold filtering, these tail-class predictions are systematically filtered out, so they receive no supervision from unlabeled data and thus features keep degrading in subsequent iterations. Critically, noise and long-tail are not independent obstacles but mutually amplifying ones, and addressing either alone is insufficient. To break this cycle, we propose FTC-Seg, a Feature-Threshold dual-Calibration framework built on a standard teacher-student framework. At the feature level, Orthogonal Prototype Reconstruction (OPR) uses a set of learnable orthogonal prototypes to residually purify pixel-wise features, widening the margin between weak foreground targets and noisy backgrounds. At the threshold level, Adaptive Threshold Calibration (ATC) dynamically adjusts class-specific thresholds based on learning difficulty and prediction-distribution bias, rescuing low-confidence pseudo-labels of tail classes from systematic exclusion. Extensive experiments on four public benchmarks spanning three distinct noise modalities show that FTC-Seg achieves strong performance against state-of-the-art methods, with particularly substantial gains on tail classes. Our results establish that jointly calibrating features and thresholds is essential for robust pseudo-labeling under compounded noise and class imbalance.
[CV-265] StoryEngine: A State-Grounded Agent ic Framework for Video Storytelling ICLR
链接: https://arxiv.org/abs/2609.33627
作者: Yingrui Wang,Zeqing Wang,Yeying Jin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 23 pages, 4 figures, submit to ICLR
Abstract:Despite recent progress in agentic multi-shot video generation, producing coherent and consistent long-form stories remains challenging. Existing agentic pipelines typically rely on textual shot plans or previously generated pixels, yet lack an explicit mechanism for propagating the consequences of story events and maintaining the video world state across shots. As a result, missing visual details may be reconstructed inaccurately, while visual drift may propagate across subsequent shots, undermining both narrative coherence and visual consistency. To address these challenges, we propose StoryEngine, a state-grounded agentic framework for video storytelling. StoryEngine establishes a separation between authoritative semantic plans and unreliable visual observations. Specifically, StoryEngine maintains a structured representation of entity placement and story-relevant states, and propagates event-induced changes to define the intended start and end states of each shot. To visually realize these states, StoryEngine constructs canonical references for recurring entities and environments, and compiles state and visual constraints into executable render plans. Meanwhile, to realize these states correctly, a bounded evaluation-guided repair loop further corrects local state inconsistencies. Together, these mechanisms preserve causal story progression and prevent local visual errors from propagating across shots. To comprehensively evaluate long-form storytelling, we construct a benchmark across diverse scenarios and visual styles, with metrics assessing storytelling quality, narrative coherence, and visual consistency. Experimental results demonstrate that StoryEngine consistently outperforms state-of-the-art methods across all evaluation dimensions, validating its effectiveness for coherent and consistent video storytelling.
[CV-266] SpatialSpeak: QA-Native Reconstruction with Local and Global Context for Spatial Chain-of-Thought Reasoning
链接: https://arxiv.org/abs/2609.33616
作者: Yang Cao,Jiaxin Zhang,Dave Zhenyu Chen,Yingji Zhong,Ruiyuan Gao,Lanqing Hong,Dan Xu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project page: this https URL
Abstract:Vision-language models (VLMs) can benefit from geometric priors for multi-view spatial reasoning, yet answer-only training does not directly supervise the intermediate geometric estimates and their use in deriving quantitative spatial answers. We hypothesize that spatial chain-of-thought (CoT) supervision becomes more effective when the VLM first jointly learns complementary local geometry and global scene context through multi-view reconstruction. We introduce SpatialSpeak, a two-stage framework that connects QA-native reconstruction pretraining with spatial CoT learning. In Stage I, QA-Native Reconstruction Pretraining (QA-RP) combines marked-point 3D queries for fine-grained local geometry with object-center queries for global scene context across views. Both tasks are formulated as text-based question answering, allowing geometric estimation and subsequent reasoning to share the same autoregressive output interface. In Stage II, spatial CoT with Visual Compensation (CoT-VC) trains the model to express question-relevant geometric estimates and use them to derive answers, with reliability assessment and visual compensation supporting answer refinement when needed. On ReVSI, QA-RP increases the gain from CoT-VC from 2.6 to 6.9 points, and ablations show that both local and global reconstruction supervision are beneficial. SpatialSpeak achieves state-of-the-art results on ReVSI, VSI-Bench, and SPAR-Bench, with a ReVSI score of 62.8 that exceeds the strongest compared baseline by 8.7 points.
[CV-267] Anguinus Sculpturae: Compositional Synthesis of Peak-Enhancement Breast DCE-MRI Scans MICCAI2026
链接: https://arxiv.org/abs/2609.33611
作者: Benjamin Hamm,Nico Albert Disch,Maximilian Rokuss,Yannick Kirchhoff,Constantin Ulrich,Klaus Maier-Hein
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted as an oral at the MAMA-SYNTH 2026 challenge / Deep-BreAth 2026 Workshop, MICCAI 2026. 12 pages, 3 figures, 1 table
Abstract:Dynamic contrast-enhanced breast MRI (DCE-MRI) is rich in anatomical and perfusion information, but its reliance on gadolinium-based contrast agents raises safety concerns and adds cost. Virtual contrast enhancement, synthesizing post-contrast from pre-contrast images, is a promising alternative. We address the MAMA-SYNTH challenge task of predicting peak-enhancement breast MRI. Rather than adopting the full machinery of diffusion or flow matching, we observe that under a rectified, straight-line path the generative process collapses to a single difference prediction: the synthetic peak image is the pre-contrast image plus a predicted enhancement map, recovered in one forward pass. Around this we build Anguinus Sculpturae, a compositional pipeline in which nnU-Net segmentations of lesion, foreground and breast region guide two generators - one optimized for global fidelity, one for lesion structure through an asymmetric Tversky term routed via a frozen segmenter - composited region-wise with Gaussian-weighted blending. On the held-out Duke subset of MAMA-MIA our model achieves the best FRD and Dice among all evaluated variants, showing that single-step difference prediction with segmentation guidance suffices to recover both global fidelity and lesion structure. Code is available at this https URL.
[CV-268] ViCoR: Reliable Molecular Structure Extraction via Spatially Aligned Verification and Executable Revision
链接: https://arxiv.org/abs/2609.33603
作者: Yujian Yuan,Xin Cai,Yufan Chen,Jiaxin Xu,Mengdi Liu,Zhichao Tan,Long Chen,Hanyu Gao
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Reliable optical chemical structure recognition (OCSR) is essential for building high-quality chemical data from scientific literature, yet even small recognition errors can propagate into chemical databases and downstream models. In practice, recognized structures often require manual inspection and correction before use, making large-scale data curation costly and difficult to scale. We therefore study Selective Structure Recognition (SSR), a post-recognition setting that automatically produces reliable structured outputs while rejecting unresolved cases. Selection-only approaches can improve reliability by rejection, but cannot create additional correct outputs beyond those produced by the base recognizer. We propose ViCoR, a repair-before-rejection framework for iterative VerIfiCatiOn and Revision. Its key idea is to make observation-prediction correspondence explicit: coordinate-preserving rendering establishes spatial correspondence between the source image and predicted structure, while index anchoring maps localized visual discrepancies to executable graph edits without full-structure regeneration. A shared VLM is progressively trained from verification to revision. On two real-world OCSR benchmarks, ViCoR improves overall accuracy from 73.53% to 88.26% and from 61.83% to 84.32%, while achieving over 97% accepted accuracy at 85–89% coverage. The resulting molecular data further improve reaction-extraction F1 by 15.5 points and literature-sourced reaction prediction accuracy by 7.7 and 5.8 points, demonstrating the value of automated reliability control for scientific data curation and downstream chemical learning.
[CV-269] LoopLUT: 3D Lookup Tables with Progressive Region Refinement for Real-Time 4K Image Enhancement
链接: https://arxiv.org/abs/2609.33593
作者: Yang Ye,Jiajun Ma,Chen Wu,Wei Wang,Dianjie Lu,Guijuan Zhang,Linwei Fan,Zhuoran Zheng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 20 pages, 12 figures, 5 tables
Abstract:Color enhancement of 4K images must meet a quality target under a tight compute budget. Three-dimensional lookup tables (3D LUTs) dominate real-time enhancement because they decide at low resolution and apply a per-pixel lookup at full resolution. A single global LUT, however, is spatially invariant, so an underexposed shadow and a well-exposed region that share a pixel value receive identical corrections. Spatially heterogeneous demands cannot be expressed by such a mapping. We propose LoopLUT, a region-cascaded 3D LUT with progressive refinement. A global LUT performs the overall correction, followed by K-1 loop iterations. In each iteration a gating head predicts at low resolution the region that still needs correction, then builds a residual LUT from the color statistics of that region alone. The cascaded gates form a partition of unity, so the output is a per-pixel convex combination of the K lookup results. Fusion is therefore performed by the gates themselves, with no separate fusion module and no interpolation error accumulating across rounds. The decision stage runs at a fixed 256x256 resolution, independent of output resolution, so a 4K image costs only K pure lookups. Extensive experiments across four benchmarks show that LoopLUT improves PSNR by up to 2.81 dB over the strongest prior method, while keeping real-time throughput at 4K. The same decomposition also generalizes well to underwater enhancement datasets.
[CV-270] IVT-Guard: All-in-One Reasoning Model for AI-Generated Content Detection
链接: https://arxiv.org/abs/2609.33585
作者: Hongwei Niu,Yunpeng Luo,Hanjun Li,Ziyin Zhou,Jianghang Lin,Ke Yan,Shouhong Ding,Shengchuan Zhang,Liujuan Cao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The rapid proliferation of highly realistic AI-Generated Content (AIGC) necessitates robust and interpretable detection mechanisms. However, existing detectors are predominantly confined to single modalities and provide binary outputs without reasoning. While Multimodal Large Language Models (MLLMs) present a promising solution, their development is constrained by the scarcity of multimodal reasoning data and the reasoning-detection optimization dilemma, where explicit reasoning supervision can compromise detection accuracy. To this end, we introduce IVT-Set, a comprehensive dataset comprising over 152K diverse image, video, and text samples equipped with multi-granularity Chain-of-Thought (CoT) reasoning trajectories. Based on it, we propose IVT-Guard, a pioneering framework for unified and interpretable AIGC detection across image, video, and text modalities. Furthermore, to overcome the aforementioned optimization dilemma, we design a novel three-stage training paradigm: Artifact-Aware Pre-training, Artifact-to-Evidence Supervised Fine-Tuning via artifact-aware injection, and Evidence-Verdict Consistency Group Relative Policy Optimization. Extensive experiments demonstrate that IVT-Guard achieves state-of-the-art detection performance across in-domain, out-of-domain, and cross-dataset settings while delivering faithful reasoning. Code and data will be released.
[CV-271] Fill2SR: Repurposing Inpainting Diffusion Transformers for Real-World Super-Resolution ECCV2026
链接: https://arxiv.org/abs/2609.33582
作者: Xingfu Yi,Xiaoxue Yu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ECCV 2026. 26 pages, 12 figures, including an 8-page appendix with additional visual results
Abstract:Recent real-world image super-resolution (SR) methods often adapt text-to-image (T2I) backbones with ControlNet-style branches or spatial conditioning tokens, which increases memory and computes with resolution and often constrains training to a fixed scale. We propose Fill2SR, which repurposes a masked-inpainting Diffusion Transformer for SR without extra spatial branches. Our Inpainting-Interface Evidence Adapter (IIEA) writes the low-quality (LQ) observation into the native masked-image slot under a full-image mask, turning inpainting into a reverse-degradation conditional rectified flow trained with LoRA-only tuning. We further introduce RCDT, an offline pipeline that distills degradation descriptors from unpaired real images and transfers them onto clean targets using frozen open-source models. Fill2SR supports mixed-resolution training up to QHD and yields stable performance across 512/1024/2048 outputs. On synthetic benchmarks, our base model with IIEA achieves the best LPIPS on DIV2K and LSDIR; adding RCDT trades a small LPIPS drop for consistently stronger no-reference quality on RealLQ250 and RealPhoto60. Fill2SR remains memory-predictable, running 1536^2 inference on a single 32GB GPU and extending to multi-megapixel outputs via tiled restoration.
[CV-272] ForeFly: A Dual-Horizon World Action Model for Aerial Vision-Language Navigation
链接: https://arxiv.org/abs/2609.33581
作者: Kunhui Wang,Xintong Zhang,Junyu Gao,Changsheng Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Aerial Vision-Language Navigation (AVLN) requires UAVs to maintain reliable instruction following over long trajectories in complex 3D environments. However, existing AVLN approaches are predominantly reactive or limited to single-horizon prediction, overlooking complementary future cues across different temporal horizons. To address this limitation, we propose ForeFly, a dual-horizon latent world action model that predicts both a proximal future for local continuity and an adaptive route-critical future for long-range guidance. Horizon-specific foresight queries are primed with recent and route-critical visual memories, providing history-aware context for future prediction. To exploit their distinct roles in action generation, we introduce Foresight-Guided Action Refinement (FGAR), which asymmetrically exploits proximal foresight for local action enhancement and route-critical foresight for feature-wise correction and route-level guidance. Experiments on the TravelUAV and UAV-ON benchmarks show that ForeFly consistently outperforms strong baselines across seen and unseen settings, validating the effectiveness of dual-horizon foresight and FGAR learning. The code is available at: this https URL
[CV-273] PGL-3D: Towards Progressive Geometric Learning for 3D Visual Query Localization
链接: https://arxiv.org/abs/2609.33558
作者: Liang Peng,Shizhuo Mu,Bohan Tan,Wenyuan Wang,Chen Zhao,Xingping Dong,Heng Fan,Libo Zhang,Bo Du
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Code and models will be released
Abstract:3D Visual Query Localization (3DVQL) retrieves the latest contiguous occurrence of a queried object in an RGB–point-cloud sequence and predicts a 9-DoF cuboid for every response frame. The query is captured independently of the search sequence, so its annotated pose may differ from how the object appears in the search frames. The benchmark baseline predicts cuboids after feature modeling, leaving their geometry unused for subsequent feature refinement. We investigate whether complete intermediate cuboids can improve query and proposal representations before final decoding. We introduce Progressive Geometric Learning for 3DVQL (PGL-3D), a predict–select–refine–re-predict framework that uses intermediate cuboids to guide the aggregation of search evidence and update query and proposal representations. A shared head first predicts a complete cuboid for every proposal. Query–Tube–Memory (QTM) then selects reference observations by combining proposal association, cuboid quality, frame response, and target absence, since association confidence alone establishes neither target presence nor geometric accuracy. The center, size, and orientation of each selected cuboid define soft pooling weights over query-conditioned proposal features. The pooled memory updates the query and proposal representations, and the head re-predicts from the updated features. A training-only objective, ST-D9O, supervises cuboid geometry at every stage by adding boundary, signed-distance, and soft-overlap terms to parameter regression. PGL-3D achieves a mean stAP of 0.270 \pm 0.004 on 3DVQL, compared with 0.044 reported for LaF. Ablations support the benefits of geometry-guided feature updates, while stage-wise analyses show improved cuboid accuracy. Replacing the geometry objective in our PROT3D reproduction with ST-D9O improves mAO on GSOT3D from 21.63% to 25.78% . Our code and models will be released.
[CV-274] In-Token Learning for High-Fidelity Image Restoration via Diffusion Transformers ECCV2026
链接: https://arxiv.org/abs/2609.33523
作者: Xingfu Yi,Xiaoxue Yu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Technical report, 25 pages, 12 figures. Preserves the earlier broader study underlying Fill2SR (ECCV 2026), including automatic colorization; Fill2SR subsequently developed the real-world super-resolution direction
Abstract:We present In-Token Learning, an image restoration framework that adapts a pretrained diffusion transformer using conditional rectified flow matching. Clean targets paired with degraded inputs supervise transport from Gaussian noise to restored images. Spatially aligned degraded-image tokens are fused with evolving latent tokens along the channel dimension, preserving the image-token count at a given resolution. Direct Low-Quality Guidance (DLG) combines frozen degraded-image embeddings with a fixed task prompt through the native conditioning pathway, without a trainable ControlNet-style branch or image captioning. We evaluate super-resolution and denoising on DIV2K, LSDIR, FFHQ, RealLQ250, and RealPhoto60, and automatic colorization on DIV2K and LSDIR. The tasks use separately trained checkpoints under the same framework. Results show competitive fidelity and perceptual quality under the evaluated protocols, with weaker generalization on RealLQ250. We report full-image QHD ( 2560\times1440 ) inference and a tiled 12 K restoration demonstration of Along the River During the Qingming Festival. Attention cost still increases with resolution. This technical report preserves the early broader study underlying Fill2SR, which subsequently developed the real-world super-resolution direction.
[CV-275] Anatomy-Structured Hierarchical MIL for Weakly-Supervised Thoracic Disease Detection in Chest X-rays MICCAI2026
链接: https://arxiv.org/abs/2609.33520
作者: Jeongin Kim,Sohyun Ahn,Seo Young Kang,Jaeyi Sung,Soomin Kim,Sungho Cho,Rena Lee,Kwanchang Kim,Junhyug Noh
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at MICCAI 2026
Abstract:Weakly-supervised thoracic disease detection in chest X-rays (CXR) is challenging due to subtle appearances and complex anatomical overlap, motivating anatomy-aware modeling for improved localization. However, prior anatomy-aware methods typically rely on coarse region proxies or static spatial priors, which may restrict dynamic instance discovery and limit precise localization of small abnormalities. We propose Anatomy-Structured Hierarchical Multiple Instance Learning (ASH-MIL), a framework that introduces parallel anatomy-structured observation branches (cardiac, pulmonary, and agnostic) combined with hierarchical MIL aggregation. Anatomical priors are injected as soft spatial biases into decoder cross-attention, enabling anatomically grounded evidence maps without disease bounding-box supervision. Instance localization is derived directly from MIL-weighted cross-attention maps without bounding box supervision. Experiments on CXR8 and cross-domain MIMIC-CXR held-out sets demonstrate consistent improvements over prior weakly-supervised and anatomy-aware approaches, particularly under stricter localization criteria. Our code is available at this https URL.
[CV-276] SceneScaffold: Active Scene-State Construction for Unified 3D Scene Understanding NEURIPS2026
链接: https://arxiv.org/abs/2609.33518
作者: Xiangqi Li,Libo Huang,Jiarui Zhao,Weilun Feng,Chuanguang Yang,Zhulin An,Yongjun Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by NeurIPS 2026 (Spotlight)
Abstract:Recent 3D large multimodal models (3D-LMMs) rely on a visual bottleneck to compress complex 3D scene evidence into a limited number of visual tokens compatible with large language models (LLMs). Current visual bottlenecks, however, often passively compress heterogeneous 3D evidence into a homogeneous object-centric token sequence, leaving the spatial organization of the scene under-represented. This under-representation forces the LLM to recover spatial relations from a flattened token sequence, leading to unstable reasoning in relation-intensive and spatially ambiguous scenes. To address this issue, we propose SceneScaffold, an active scene-state construction framework for unified 3D scene understanding. SceneScaffold reformulates the visual bottleneck from a passive feature compressor into an active scene organizer, constructing a role-aware spatial scaffold before language reasoning. Specifically, SceneScaffold organizes superpoint-level visual evidence into scene-state components with distinct structural roles: entity states preserve core object semantics, scene-frame states maintain spatial references via boundary and region anchors, relation states encode object-environment interaction cues, and a global summary provides compact context. Through this role-aware construction, SceneScaffold provides the LLM with a spatially organized scene representation before language reasoning. Experiments on unified 3D scene understanding tasks, including 3D visual grounding, question answering, and dense captioning, demonstrate the effectiveness of SceneScaffold, while diagnostic results further show its applicability to relation-intensive and spatially ambiguous cases. Code is available at this https URL.
[CV-277] Printability-Constrained Adversarial Decals for Near-Nadir Aerial Perception: Measured Ink Gamuts Nested Realism Constraints and a Physical-World Bound
链接: https://arxiv.org/abs/2609.33513
作者: Sandesh Shrestha,K. T. Yasas Mahima,Asanka G. Perera
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Adversarial patches for aerial perception are typically evaluated as digital composites, with printing left as an implementation detail. This study imposes three physical constraints during optimization rather than after it: the color range a particular printer can reproduce, the size of the flat panel a vehicle offers, and the loss of fine detail incurred when the patch is imaged from altitude. The principal comparison isolates the ink set. Two patches share all seventeen recorded optimization settings and differ only in the colors available to them. One is constrained to a uniform color cube; the other to a gamut measured by printing and scanning a 216-patch chart. Each was optimized at three seeds and evaluated against thirteen victim conditions, with every rate reported against a size-matched optimized control. The effect of the measured gamut is victim-dependent rather than uniform. Net attack success rises on three of six closed-set segmentation victims, and for these the seed ranges of the two ink sets are disjoint: +0.120 on DeepLabv3-R101 and +0.041 on SegFormer-B0. The color-cube patch is consistently stronger on the open-vocabulary segmenter and on two of four detectors, though no detector exceeds a net of +0.026 under either ink set. The natural explanation is that a printable palette is simply less chromatic and lower in frequency than a digital one. Eleven further patches test this account and it does not hold. Once cardinality is matched, a palette as chromatic as the cube attacks equally well. Cardinality itself shows no trend from three inks to thirty-two. Palettes matched on cardinality, lightness and chroma, and differing only in hue placement, span 0.035 to 0.136 . A physical evaluation with printed decals did not detect transfer; it bounds the transferred rate at 0.133 , which does not exclude the simulated value of 0.121 .
[CV-278] Chameleon: Dynamic Format Adapter for Efficient Diffusion
链接: https://arxiv.org/abs/2609.33496
作者: Arnab Sanyal,Sandeep Chinchali
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: Under Review at International Conference on Learning Representations 2027
Abstract:Post-training quantization (PTQ) is the standard way to run modern diffusion models on memory-constrained accelerators, yet every existing diffusion PTQ scheme fixes the \mathitnumber\ format in advance and only tunes the scale, zero point, or per-layer bit-width. At a fixed bit-width the best format depends on the distribution being encoded, and that distribution differs across weight channels, across layers, and along the diffusion timestep, where activation distributions slide from heavy-tailed and noise-dominated to tightly clustered and structured. We propose Chameleon, a PTQ framework that holds the bit-width fixed and treats the format itself as a discrete variable, chosen per weight channel and per (layer, timestep bucket) activation tensor. Activation formats come from INT8, FP8 E4M3, FP8 E5M2, MXFP8, MXINT8, selected ahead of time from two cheap statistics (empirical kurtosis and the closed-form diffusion SNR) and stored in a lookup table; weight formats come from INT8, MXINT8 at 8 bits or INT4, NF4, FP4 E2M1, MXINT4, MXFP4 at 4 bits, selected offline by reconstruction error. An architectural fork adapts the same selection layer to multi-step UNets, single-step distilled models, and Diffusion Transformers. Across SDXL, SDXL-Turbo, and PixArt- \alpha on COCO-2014, Chameleon achieves the best FID in all six backbone \times bit-width settings, with CLIP within 0.24 of the FP16 reference and the best of all quantized methods at W_4A_8 .
[CV-279] SphMind: Towards Robust Training-Free VLM-based Spatial Reasoning with a 360 Camera NEURIPS-2026
链接: https://arxiv.org/abs/2609.33462
作者: Shriram Damodaran,Soumyaratna Debnath,Cheston Tan,Lin Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:Omnidirectional or 360 cameras provide embodied AI agents with a holistic, wide field-of-view (FoV) view of their surroundings, motivating the use of Multi-modal Large Language Models (MLLMs) for omnidirectional spatial reasoning. However, most MLLMs are trained on conventional 2D perspective images and struggle with the severe distortions and wrap-around discontinuities induced by spherical geometry. Enabling them to generalize to non-Euclidean 3D spaces without retraining therefore remains challenging. We propose SphMind, a training-free, plug-and-play framework that decouples semantic perception from geometric reasoning. Rather than requiring MLLMs to learn spherical geometry internally, SphMind preserves their semantic capabilities while handling geometry externally. We introduce a Spherical Harmonics-based Spatial Graph (SHSG) that models spatial relationships through equivariant transformations on the sphere, together with Inference-Time Geometric Grounding (IGG), a model-agnostic closed-loop optimization process that aligns MLLM representations with spherical geometric constraints during inference. Experiments on three benchmarks show that SphMind achieves over 21.4% average improvement in directional reasoning on MP3D and Stanford2D-3D, outperforms prompt-engineering baselines by 8.7% on the real-world ODI-Bench, and improves rotational invariance by 5.9% under panorama rotations, without additional training or dataset-specific tuning. In-the-wild evaluations further show that SphMind resolves directional reasoning queries that baseline vision-language models fail to answer correctly.
[CV-280] Native Association: Confidence-Aware Human Perception in the Wild with a Foundation VLM ECCV2026
链接: https://arxiv.org/abs/2609.33450
作者: Igal Dmitriev,Ofir Liba
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the ECCV 2026 Workshop on Human-Centered Multimodal Intelligence in the Wild (HCMIW). 16 pages, 1 figure, 2 tables; supplementary material included as ancillary file
Abstract:Extracting who is where, on which team, wearing which number from a broadcast frame is typically done by stitching a detector, an OCR engine, and classifiers together – and the stitching step swaps identities under occlusion. We make association native instead: a 0.77B vision-language model (Florence-2) is fine-tuned to emit all per-person attributes as one grammar-constrained sequence, with each attribute generated inside its owner’s block. Output is therefore schema-valid on every frame by construction, and no post-hoc binding step exists to attach a correctly read number to the wrong player: residual misassociation is pure perception error, \approx4\times rarer than zero-shot-prompted frontier APIs’ (0.057 vs. 0.21-0.24). On a frozen multi-sport test set, this single pass reaches 0.95 detection F1 (APIs: 0.65-0.75). A single extra forward pass yields a per-field confidence that supports a reject option (jersey precision 0.71\rightarrow0.96 at half coverage) and routes a training-free zoom-and-re-read for small players. Surprisingly, once the grammar is learned, further parameter-efficient tuning yields no measurable gain under the adaptation configurations we test; the identical recipe on WIDER-Attribute reaches 93.1 mAP given-box, yields the first detection-coupled end-to-end results under its standard test protocol (84.5 mAP), and reproduces the same tuning result. In this regime, the gains live in the structure, not in added weights.
[CV-281] A Visual Classification Dataset and Model Evaluation for Historical Manuscript Illustrations
链接: https://arxiv.org/abs/2609.33449
作者: Yoav Evron,Michal Bar-Asher Siegal,Michael Fire
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 17 pages, 5 figures
Abstract:Historical manuscript illustrations preserve rich visual evidence of past cultures. They depict people, animals, plants, diagrams, music notations, and decorative forms. Although large digitization projects have made many manuscripts available online, the material itself remains difficult to explore at scale. Extraction systems can find illustrations on manuscript pages, but without meaningful categories, large collections remain hard to search and explore. We address this gap by introducing a manually labeled dataset of 15,000 illustrations from manuscripts dating back hundreds of years across 22 categories, and evaluating modern vision models for image classification on this task. The problem is challenging due to stylistic diversity, degradation, and semantic ambiguity, with many images that fit more than one category. We compare fine-tuned CNN and Transformer-based classifiers, zero-shot CLIP, embedding-based classifiers, and direct vision-language models. Results show that fine-tuned image classifiers perform best overall, with ConvNeXt reaching 88.9% accuracy and 81.3% macro-F1. Using CLIP embeddings with XGBoost provides a strong alternative. In contrast, zero-shot CLIP and direct vision-language classification perform substantially worse, highlighting the limits of general-purpose models in this domain. Beyond overall performance, the analysis reveals which categories are visually separable and where errors reflect genuine semantic overlap, suggesting that some limitations arise from the taxonomy itself.
[CV-282] Concept Score Relearning: A Unified Cross-Architecture Attack on Concept Erasure
链接: https://arxiv.org/abs/2609.33445
作者: Hong Xi Tae,Jiaming Zhang,Wenwen He,Xuan Wang,Wei Yang Bryan Lim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Concept erasure aims to suppress undesirable knowledge in text-to-image generative models. However, existing robustness evaluations typically rely on relearning attacks tailored to specific model architectures. We study concept reactivation across two substantially different generative paradigms: noise-prediction U-Nets and flow-matching Transformers. We introduce \textbfConcept Score Relearning (CSR), a unified parameter-level framework that reactivates erased concepts by optimizing each model within its native prediction space. CSR requires no external target-concept image dataset and applies the same concept-directed objective to both U-Net-based Stable Diffusion and Transformer-based FLUX. Experiments across diverse concepts and multiple erasure methods demonstrate consistent concept reactivation across both architectures, highlighting the cross-architecture applicability of CSR and the persistent recoverability of apparently erased concepts. For strict nudity, CSR reaches average ASRs of 50.47% on FLUX and 40.29% on Stable Diffusion, consistently ranking first across all evaluated safety settings.
[CV-283] Elucidating the Design Space of Regression-based Diffusion Reinforcement Learning
链接: https://arxiv.org/abs/2609.33444
作者: Toyota Li,David Zhao,Alan Zhao
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:A nascent family of methods that forgoes the policy gradient and reweights a supervised regression instead has garnered momentum in reinforcement learning for diffusion and flow models. DiffusionNFT, FlowAWR, and RAM are representative regimes with contrasting motivations. It is yet opaque what, if anything, they share. We substantiate that each is the solution of one divergence-constrained reward-maximization problem, and they are differentiated only by the convex generator that defines the constraint. Under the unified modeling framework, we unravel the relaxations that prior art made during building the advantage-embedded regression target: approximating the KKT condition and posterior normalizer for the linear and exponential tilt shapes DiffusionNFT and FlowAWR respectively, while preserving the exact sparsemax projection onto the probability simplex for linear tilt leads to another superior model type in this work. Beyond the theoretical underpinnings, we further empirically investigate the design space and shed light on the training recipe for regression-style diffusion RL. Retaining the merits discovered during our exploration gives rise to DiffusionRFT, our paradigm that converges faster, trains more stably, and attains the top performance.
[CV-284] C-ADA: One-Shot Active Domain Adaptation for Semantic Segmentation
链接: https://arxiv.org/abs/2609.33432
作者: Weihao Yan,Yeqiang Qian,Yueyuan Li,Tao Li,Chunxiang Wang,Ming Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 13 pages, 15 tables, 3 figures
Abstract:Manual dense annotation remains a major obstacle to deploying semantic segmentation models in new driving environments. Active domain adaptation (ADA) seeks label-efficient transfer by annotating only a selected portion of the target domain. Existing ADA methods commonly implement this process through multiple rounds of acquisition, annotation, and retraining. We study a practical one-shot image-level setting that selects and densely annotates a fixed target subset in a single round, followed by uninterrupted adaptation. Within this setting, we develop Target-Calibrated Active Domain Adaptation (TC-ADA) as a joint design of complete-image acquisition and target-calibrated adaptation. Stage~1 uses visual representations from a vision foundation model (VFM) together with semantic predictions from a fixed unsupervised domain adaptation model to select representative and informative target images without target annotations. Stage~2 jointly uses labeled source data, labeled target data, and the remaining unlabeled target data, while calibrating source and target supervision under limited target labels. Extensive experiments across five synthetic-to-real and real-to-real driving transfers show consistent improvements over representative ADA baselines. With only 23 to 46 labeled target images on four transfers and 140 on Mapillary, TC-ADA stays within 1.9 mean intersection over union (mIoU) points of target-only full supervision. Code will be available at this https URL.
[CV-285] A Light Bilevel Refinement Aligns Self-Supervised Representations for Stronger Task-Specific Learning
链接: https://arxiv.org/abs/2609.33424
作者: Gustav Wagner Zakarias,Zheng-Hua Tan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Self-supervised pretraining learns representations that are broadly transferable across downstream tasks, yet direct fine-tuning can be suboptimal due to misalignment between self-supervised and downstream task objectives, potentially degrading pretrained features beneficial to the downstream task. The BiSSL framework addressed this by introducing a transitional training stage formulated as a bilevel optimization problem, in which the downstream task objective guides the self-supervised learning process in refining pretrained representations to better facilitate subsequent fine-tuning. However, BiSSL relies on conventional bilevel optimization solving techniques whose costly implicit hypergradient approximations render the method increasingly impractical for contemporary model architectures. To make it efficient and scalable, we introduce BiSSLight, which combines M-FAC-based implicit gradient approximation with parameter-efficient fine-tuning via LoRA, enabling efficient application at larger scales that were previously impractical. Evaluation across multiple downstream tasks and contemporary model architectures shows that BiSSLight consistently improves downstream performance, with gains becoming more pronounced as model size increases despite stronger baselines. The method is highly computationally efficient, reducing computation time by more than a factor of ten compared to its predecessor on a ViT-H backbone.
[CV-286] -VidT: Decoupling the Temporal Axis for Efficient Motion-Centric Video Pretraining NEURIPS2026
链接: https://arxiv.org/abs/2609.33419
作者: Shih-Ying Yeh,Daniel Z. Kaplan,Xuehai Wang,Fu-En Yang,Min-Hung Chen,Shang-Hong Lai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: accepted by NeurIPS 2026 main track
Abstract:Comparisons in video self-supervised learning often evaluate complete training recipes rather than isolating the method itself: architecture, objective, data exposure, schedule, scale, and decoder capacity can all vary at once. This makes it hard to identify which choices yield motion-prioritized representations, whose gains concentrate on frame-to-frame change while retaining useful appearance. We address this with a matched 4 \times 6 = 24 architecture-objective study at roughly 170M ~ 190M encoder scale on \sim 1.7M OpenVid and Moments-in-Time v2 clips for 8 epochs, and propose TT-VidT. TT-VidT combines a DINOv3-initialized ViT-B/16 per-frame spatial path with a compact Temporal Transfer Layer, trained by Diff Compression to reconstruct target frames from a first-frame appearance anchor and frame-specific motion tokens. The sweep shows that TT3D with Diff Compression, not either component alone, enters the strongest motion-sensitive regime, and decoder ablations favor a compact video-pretrained decoder. In final comparison, TT-VidT leads Jester, Something-Something V2, ARID, and Diving48 fine-tuning simultaneously, improving over the strongest non-TT row by 54% ~ 121%, while using 48% fewer encoder FLOPs than DisMo and 55% fewer than VideoMAE or V-JEPA2. HMDB51, IARD, and EPIC-Kitchens bound the claim.
[CV-287] RSD: Test-Time Reinforcement Learning with Self-Distillation for Vision-Language Models
链接: https://arxiv.org/abs/2609.33414
作者: Shuning Wang,Zhiheng Wu,Xun Zhou,Chongyang Cui,Chen Jia,Bowen Liu,Chuanjie Li,Xiang Chen,Yi Yang,Yumeng Zhang,Wenjie Huang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Test-time reinforcement learning enables vision-language models (VLMs) to adapt using unlabeled inputs. However, repeated sampling under fixed visual conditions can reinforce shared perceptual errors, while sequence-level rewards fail to isolate visual perception the foundational bottleneck that anchors multimodal reasoning risking the degradation of pre-trained reasoning capabilities. We propose TTRSD, a test-time reinforcement learning framework combining multi-view answer-level self-distillation with visual contrastive token selection. A shared policy aggregates teacher predictions across original, cropped, and downsampled views into an answer distribution. Student trajectories generated from the original image receive rewards based on the support for their final answers in this distribution. To allocate this feedback precisely toward perceptual bottlenecks, we compare the log-probabilities of the same sampled tokens under original and visually ablated inputs while holding their textual prefixes fixed, selecting visually sensitive positions for policy-gradient updates. TTRSD separates update direction, determined by group-relative advantages, from update position, determined by visual sensitivity, without requiring ground-truth labels, external verifiers, or a separate teacher. With only 20 unlabeled adaptation samples, TTRSD improves performance across seven benchmarks and three VLMs, raising InternVL3-2B’s MMMU accuracy from 35.79% to 49.32%(+13.53%), demonstrating cross-dataset generalization while preserving inherent reasoning integrity.
[CV-288] Resolving State-Representation Mismatch: State-Space Visual Reasoning for Open-Loop VLA Planning
链接: https://arxiv.org/abs/2609.33412
作者: Junhao Xiao,Haoxiang Zhao,Menghao Fang,Jinkui Zhang,Jinghan Yu,Xinyu Huang,Zhiyu Wu,Kaiming Xu,Yi Chen,Youjun Bao,Zhiyuan Ma
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Despite rapid progress in vision-language-action (VLA) models, existing reasoning paradigms still face a fundamental \emphstate-representation mismatch in open-loop planning. Given only an initial observation, models must internally simulate action-conditioned state transitions, whereas text-, pixel-, and latent-space reasoning can suffer from lossy spatial compression, error-accumulating visual generation, and bypass of intermediate latent tokens, respectively, undermining reliable long-horizon planning. We propose \textbfState-Space Visual Reasoning (SSVR), which decouples static visual context, language constraints, and a recurrent latent state. SSVR encodes the initial image and instruction once, then conditions each action prediction on the latent state and updates it with an action-conditioned GRU. Using Qwen2.5-VL as the backbone, SSVR achieves 99.5/99.6, 96.3/98.0, and 83.9/90.6 EM/PR on FrozenLake, Maze, and MiniBehavior, substantially outperforming prior methods. Extensive experiments support the effectiveness of recurrent state modeling for VLA open-loop planning across input transformations and transfer settings. By reusing static visual-textual context and updating a compact recurrent state, SSVR supports efficient multi-step inference, achieving up to 98.58\times faster Maze decoding rollouts than the evaluated baselines with the prefix cache prebuilt.
[CV-289] VaME: Exploring Variational Latent Reasoning for Multimodal Embeddings
链接: https://arxiv.org/abs/2609.33402
作者: Peixi Wu,Mingzhou Jiang,Feipeng Ma,Biao Yang,Yunhao Zhou,Wei Yuan,Bosong Chai,Huizu Lin,Jie Chen,Zhangchi Hu,Fan Yang,Wenwu Ou,Hebei Li,Xiaoyan Sun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Universal multimodal retrieval requires compact embeddings that preserve task-relevant semantic information across diverse modalities. Prior works have incorporated latent reasoning into multimodal embedding learning to refine this information before embedding extraction. However, most existing approaches remain confined to deterministic latent paths, without exploring alternative trajectories to discover better embeddings. Thus, we propose VaME (Variational Multimodal Embeddings), a framework that models latent reasoning as a learnable distribution over trajectories. Specifically, we first introduce Variational Latent Reasoning (VLR) to enable autoregressive exploration in latent space, guided by answer reconstruction through a lightweight decoder. Meanwhile, we augment the original embedding-token readout with a latent-fused embedding to facilitate exploration during subsequent reinforcement learning. Finally, we optimize latent reasoning over stochastic variational trajectories through reinforcement learning, using Semantic Decoding Reward (SDR) to favor semantically meaningful trajectories with interpretable decoded outcomes. On the 78-task MMEB-V2 benchmark, spanning image, video, and visual-document retrieval, VaME outperforms most explicit CoT-based models and all latent-reasoning baselines. VaME also demonstrates robust performance on reasoning-intensive benchmarks such as MRMR, with substantial gains after reinforcement learning. Importantly, VaME achieves these gains with at least a 4.25x inference speedup over the deterministic latent autoregressive baselines. The code will be made publicly available.
[CV-290] Groupwise Selective State-Space Filtering for Accurate and Streaming Action Boundary Detection
链接: https://arxiv.org/abs/2609.33400
作者: Mustafa Bora Çelik
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages,3 figures
Abstract:Action boundary detection partitions untrimmed video into intervals without assigning action classes. We present a boundary-detection adapter operating on pre-extracted video features, learning temporal representations via groupwise selective scans. Learned group fusion and temporal modeling convert these into transition scores, which are decoded into boundary timestamps. Trained with boundary-time supervision, the class-agnostic model is evaluated on Breakfast, GTEA, and 50Salads using temporal tolerances and bipartite matching, achieving boundary F_1 scores of 0.457, 0.622, and 0.611. A stateful variant enables feature-streaming inference with zero neural look-ahead, one-sample peak confirmation, and bounded memory. Downstream systems can subsequently assign s
[CV-291] SciGen-Verifier: A Multimodal Reason er for Explainable Verification in Scientific Image Generation
链接: https://arxiv.org/abs/2609.33399
作者: Jiali Chen(South China University of Technology, The Hong Kong Polytechnic University),Zhengteng Lin(South China University of Technology),Zuqi Wang(South China University of Technology),Shirong Lin(South China University of Technology),Xi Yu(South China University of Technology),Xusen Hei(South China University of Technology),DingBa Fu(South China University of Technology),Jiayuan Xie(The Hong Kong Polytechnic University),Yi Cai(South China University of Technology)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:In realistic education, a solution is often expressed not only in words but in a drawing–a circuit, a geometric construction, a function plot–and a teacher must grade the drawing as carefully as the text. Recent advances in unified multimodal models have enabled scientific image generation, yet verifying the correctness of these specialized visual outputs remains a critical bottleneck: errors often arise from intricate domain knowledge, structural reasoning, and multi-step instruction rather than surface-level artifacts. Existing verifiers mainly target natural images and compress judgement into scalar scores, leaving scientific coverage and explainable feedback for error correction underexplored. To bridge this gap, we make three main contributions. (1) We construct SciGen-Verify, a benchmark dedicated to explainable verification of scientific image generation, spanning instruction following, multidisciplinary reasoning, and world knowledge domains. It contains a three-tier hierarchical protocol over the binary judgement, supporting explanation, and corrective editing instruction. (2) We develop SciGen-Verifier, a reasoning-driven multimodal verifier trained via cold-start supervised fine-tuning followed by a curriculum-based two-stage reinforcement learning pipeline. The rubric-guided process rewards first strengthen scientific reasoning exploration and outcome rewards subsequently align output with ground-truth annotation. (3) On SciGen-Verify, SciGen-Verifier achieves competitive performance against much larger proprietary models. It further serves as a practical online critic for iterative image rectification.
[CV-292] PulseQuant: Propagation-Guided Subspace Correction for 4-Bit Video Diffusion Transformers
链接: https://arxiv.org/abs/2609.33384
作者: Yutong Wang,Xingtong Ge,Enhuai Liu,Yunke Wang,Tianfan Xue,Xinyuan Chen,Chang Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Quantization errors in video diffusion transformers can be amplified or attenuated by subsequent denoising updates, making local reconstruction error an incomplete predictor of final impact. We introduce PulseQuant, a 4-bit post-training quantization method that combines trajectory sensitivity with activation geometry to guide offline calibration. Isolated block–step interventions estimate propagation risk, which prioritizes sensitive trajectory states during row-radius selection. With these radii fixed, response-subspace correction uses neighboring-code edits to reduce residual components along dominant activation directions. Both stages preserve the original 4-bit weight representation. Controlled interventions show that short-horizon propagated error predicts final latent error more reliably than immediate block-output error, supporting calibration beyond local reconstruction objectives. Evaluations on Wan models, Self Forcing, and MiniMax-H3 demonstrate improvements in key consistency and dense-reference metrics while remaining competitive on other attributes across model scales and generation paradigms.
[CV-293] Recursive Harness Distillation across Agents for Robot Manipulation
链接: https://arxiv.org/abs/2609.33378
作者: Seungyeon Kim,Junhoo Lee,Minkyu Kim,Baekseung Kim,Nojun Kwak
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:A central goal in robotics is to enable manipulation across changing tasks and environments. Vision-language-action (VLA) models provide broad manipulation capabilities but can struggle when execution requires diagnosing failures and adapting behavior. Strong agents can discover effective interventions through interaction with these policies. We propose Recursive Harness Distillation to accumulate this experience as reusable guidance across agents. A strong agent distills its experience into a playbook for a light agent, then recursively refines the playbook using the light agent’s execution feedback. The resulting playbook enables agents to reuse accumulated intervention knowledge in new task instances without updating model parameters. In real-world manipulation, the harness improves success from 37.3% to 64.0%. On SimplerEnv Bridge, the light agent with the playbook achieves 66.7% success, compared with 41.7% for the GR00T-only baseline, and outperforms the strong agent without a playbook. The same playbook also benefits the strong agent, which reaches 79.2% success. These results demonstrate the feasibility of harness distillation for robotics: intervention experience can be accumulated, refined through execution, and reused across agents to improve manipulation.
[CV-294] NF based Spectral Embedding for Effective Application of Supervised Machine Learning Techniques in Automobile Insurance Fraud Detection
链接: https://arxiv.org/abs/2609.33376
作者: Rohan Yashraj Gupta,Lalith Srikanth Chintalapati,Satya Sai Mudigonda,Pallav Kumar Baruah,Raghunatha Sarma Rachakonda
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 12 pages, 3 figures,
Abstract:Fraud detection is an important area of research in the insurance business due to its financial implications. The primary aim of a fraud detection model is to identify fraud and non-fraud cases with high accuracy along with other important metrics such as Sensitivity, Specificity, Precision, F1-score, False Positive Rate, False Discovery Rate, AUC etc. To achieve this, we need to explore a suitable classification model to identify fraud and non-fraud cases. In this work, we have used auto insurance data set and explored classification models such as Decision Tree (DT), Random Forest (RF), XGBoost, LightGBM and Gradient Boosting Machine (GBM). To overcome the problem of data imbalance, we have employed MWMOTE and TGAN techniques. We have used Topological Node Feature(TNF) based spectral embedding for low dimensional data representation along with some popular embedding methods like MDS, Isomaps and t-SNE. After studying all the 65 possible combinations of these models, we have proposed an innovative method for effective automobile insurance fraud detection. For the given dataset, our results show that using a combination of MWMOTE as a data imbalance handling technique (Phase I), TNFSE2 as data embedding (Phase II) and Random Forest as classification (Phase III) provides the best result in comparison to all other combinations. This work also highlights the efficacy of TNF based spectral embedding in automobile insurance dataset
[CV-295] When Does Geometric View Synthesis Help Wine Label Retrieval? A Public One-Shot Benchmark Across Self-Supervised and Vision-Language Backbones
链接: https://arxiv.org/abs/2609.33359
作者: Yueh-Cheng Huang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Geometric view synthesis can expand a single wine-label photograph into a training set, but its value with pretrained image encoders is unclear. We study this on a public WineSensed-derived benchmark of 1,000 classes, one enrollment photograph per class, and 4,295 real queries. With the earlier DINO vision transformer (ViT-S/16) recipe, geometric views raise top-1 accuracy from 34.1% to 62.6-63.7%, about three times the gain from two-dimensional (2D) augmentation. Frozen SigLIP 2-B already reaches 94.7%. A linear head over its frozen features gains 1.2-1.3 percentage points with the two geometric pipelines localized by the Segment Anything Model (SAM), while the other pipelines gain an inconclusive 0.3-0.6 points. Low-rank adaptation (LoRA) and validation-selected full fine-tuning show no clear gain within the reported confidence intervals; fixed-budget full fine-tuning loses 9-24 points. SAM localization supplies all six views for 99% of sources, compared with 43% for the edge-based front end. Recognition differences between the two cylinder constructions depend on the training recipe and are confounded by their crop and canvas conventions. Rendered-cylinder tests show different responses to source tilt, but an uncalibrated rim-ratio proxy establishes no corresponding trend in recognition on real photographs. An author-confirmed audit of 50 residual errors identifies 21 query-enrollment appearance mismatches, without establishing an irreducible error rate. These results support geometric synthesis for the tested self-supervised recipe and a smaller benefit through frozen-feature adaptation of the text-supervised encoder.
[CV-296] Focus and Supplement: Dual-Enhanced Vision Transformer for Multi-Class Anomaly Classification
链接: https://arxiv.org/abs/2609.33353
作者: Xurui Li,Enjie Xu,Chenzhou Li,Shilei Zeng,Dayou Huang,Tianyi Ma,Yu Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multi-class anomaly classification in industrial vision remains challenging due to noisy/incomplete anomaly representations and the unknown number of anomaly classes. To overcome this, we propose MACO, a novel multi-class anomaly classification framework that learns comprehensive representations and dynamically estimates class number without prior knowledge. First, a soft-focus attention uses anomaly maps to concentrate on relevant abnormal regions, while suppressing background noise. Second, auxiliary classification ([A-CLS]) tokens complement the [CLS] token. They collectively attend to diverse anomaly sub-regions, yielding more holistic and discriminative features. These [A-CLS] tokens are also effective across more tasks and domains. To infer the class number, we propose Correlation-based Number Estimation strategy. It computes the average correlation among labeled classes and transfers its separability cue to the unlabeled set. Experiments on MVTec AD and MTD datasets demonstrate our superiority. Under known class number, MACO improves ARI by 6.5% and \textbf16.3% on both datasets, respectively. In the more challenging unknown number scenario, it achieves an \textbf11.2% NMI gain on MTD and outperforms existing number estimation strategies by \textbf24.1% UPS on MVTec AD. Code will be released at this https URL.
[CV-297] ReLoc: Rethinking Scene Coordinate Regression Architecture for Robust Outdoor LiDAR-based Localization IROS2026
链接: https://arxiv.org/abs/2609.33344
作者: Heejoon Moon,Yurim Cho,Je Hyeong Hong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to IROS 2026
Abstract:Scene Coordinate Regression (SCR) has recently emerged as a promising approach for LiDAR-based localization, achieving accurate localization without requiring an explicit 3D map. Despite their effectiveness, existing SCR methods rely on scene classification-based global embedding that struggles to provide fine-grained discrimination among nearby locations. Moreover, their reliance on uniform sampling of local features during training assigns equal importance to all points, thereby inadvertently propagating features from dynamic objects or unstable regions and potentially degrading training stability. In this paper, we present ReLoc, a revamped SCR architecture that can effectively address these limitations. First, we redesign the global embedding module by combining learnable context tokens with a feature aggregator to capture richer and more discriminative scene context. Second, we introduce an attention-based local feature enhancement module to mitigate the impact of noisy local features while encouraging context-consistent structures, yielding more robust local feature representations. Experimental results on two large-scale outdoor datasets demonstrate that our approach achieves state-of-the-art accuracy over previous SCR-based methods while maintaining real-time inference performance.
[CV-298] OPERA: A Unified Omnimodal Progressive Spatio-Temporal Reasoning Agent for Referring Video Segmentation
链接: https://arxiv.org/abs/2609.33338
作者: Jingchen Ni,Yuji Wang,Shannan Yan,Haoru Li,Sitong Chen,Chun Yuan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages
Abstract:Referring video segmentation with heterogeneous multimodal queries—spanning text, audio, and reference images—demands both robust cross-modal understanding and precise spatio-temporal reasoning. We propose OPERA (Omnimodal Progressive spatio-tEmporal Reasoning Agent), a unified reasoning agent built on a single MLLM that performs dual-axis progressive reasoning via three specialized stages. Along the temporal axis, a Temporal Reasoning Agent narrows the frame search space through coarse-to-fine filtering to identify the most informative key frame. Along the spatial axis, a Distillation Agent establishes what to locate via cross-modal semantic distillation, and a Grounding Agent enhanced with GRPO determines where the target appears, with dense mask propagation completing the pixel-level output. OPERA sets a new state of the art on OmniAVS and Ref-AVS and transfers zero-shot to standard referring video segmentation benchmarks.
[CV-299] FeCoSplat: Feedback-Guided Compression for Feed-Forward 3D Gaussian Splatting
链接: https://arxiv.org/abs/2609.33330
作者: Yuxuan Li,Yihang Chen,Yufeng Zhang,Jianfei Cai,Weiyao Lin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Feed-forward 3D Gaussian Splatting (3DGS) enables efficient novel-view synthesis from sparse multi-view images, yet its representations remain costly to store and transmit. Existing approaches compress either the input images, incurring heavy receiver-side reconstruction, or the reconstructed Gaussian primitives, which are difficult to compress due to their heterogeneous and irregular attributes. We instead compress compact intermediate features, providing a better balance between compression efficiency and receiver-side complexity. Based on this paradigm, we propose FeCoSplat, a feedback-guided compression framework for feed-forward 3DGS. FeCoSplat first compresses multi-view features to obtain an intermediate 3DGS, whose rendered views are used as feedback to guide a second-stage compression for further refinement. The resulting bitstreams are decoded into a compact implicit state, from which the final Gaussian primitives are reconstructed with a lightweight predictor. Experiments demonstrate that FeCoSplat achieves favorable rate–distortion performance, particularly at low bitrates, while requiring only 3.45M parameters for receiver-side Gaussian reconstruction. Code will be released soon.
[CV-300] VisionHOPE: Visual Backbones as Self-Modifying Learning Systems
链接: https://arxiv.org/abs/2609.33325
作者: Siran Peng,Tianshuo Zhang,Tianyu Fu,Weisong Zhao,Haoyuan Zhang,Jiankuo Zhao,Minghui Wu,Ping Jiang,Xiangyu Zhu,Chenxu Zhao,Zhen Lei
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Visual backbones have evolved from Convolutional Neural Networks (CNNs) with local aggregation to Vision Transformers (ViTs) with global interactions, State-Space Models (SSMs) with input-dependent state transitions, and Test-Time Training (TTT) layers that adapt an inner learner while processing an image. Across this progression, visual computation has become increasingly adaptive to each input, yet the rules governing that adaptation remain largely prescribed by the trained backbone. We introduce VisionHOPE, the first generic visual backbone formulated as a self-modifying learning system, in which what the model remembers and how it learns co-evolve within an image. Building on the self-referential construction of Nested Learning (NL), VisionHOPE realizes this co-evolution through five coupled memories that store content, generate key and value representations, and govern learning rate and retention. These memories evolve jointly as visual context accumulates along each scan. However, directly applying the unconstrained self-referential update to a visual backbone leads to instability. We therefore derive a stability-matched step-size control scheme that combines a soft cap on self-referential injection with a spectral clamp on the retained memory transition, and prove that the resulting memory dynamics are non-expansive along each scan. For two-dimensional feature maps, we adapt NL’s chunk formulation by aligning chunks with image rows and columns across four directional scans. The proposed VisionHOPE achieves competitive results on ImageNet-1K, COCO, and ADE20K, establishing self-modifying learning systems as a practical foundation for general-purpose visual backbones. The code is available at this https URL.
[CV-301] PIC-UIE: Predicting Image-Adaptive Corrections for Lightweight Underwater Image Enhancement
链接: https://arxiv.org/abs/2609.33318
作者: Cunhao Zhu,Dongliang Xu,Xiangtao Kong,Xiaoyan Lu,Tianyu Wang,Yue Yao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Underwater image enhancement (UIE) aims to restore visibility, color fidelity, and structural detail from images degraded by wavelength-dependent attenuation and backscatter. State-of-the-art UIE methods often rely on large backbones and dense image-to-image prediction, limiting their practicality for edge deployment. Moreover, operating entirely in a single color space couples degradation estimation with luminance and chroma correction. To address these challenges, we propose PIC-UIE, a lightweight predictor–executor framework that predicts image-adaptive corrections from a fixed 256\times256 RGB thumbnail and applies them to the native-resolution input in the YCbCr color space. The predictor produces seven outputs, organized into spatial correction, nonlinear luminance and coupled chroma mapping, and image-level color calibration. A depth map regularizes the transmission proxy during training, whereas inference uses only the RGB input. With 9,486 parameters and 0.094 GFLOPs at 256\times256 , PIC-UIE achieves 24.137 dB PSNR and 0.9216 SSIM on UIEB-90 and 21.320 dB PSNR on zero-shot LSUI. It further processes native 4K images at 55.0 FPS under the comparison protocol. These results show that structured correction prediction provides an effective and practical alternative to dense RGB reconstruction for underwater image enhancement.
[CV-302] SocialHumanoid: Towards Expressive Humanoid Behavior via One-Step Co-Speech Motion Generation
链接: https://arxiv.org/abs/2609.33311
作者: Chengqun Yang,Tengjie Zhu,Liang Xu,Fulong Liu,Guanzhu Ren,Yitong Xing,Xuefeng Lu,Fei Shi,Siyuan Fan,Weijie Dong,Yao Mu,Xiaokang Yang,Yichao Yan
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Humanoid robots are increasingly expected to serve as embodied social agents that communicate naturally with humans through face-to-face interaction. During such communication, humanoid robots require body behaviors that are synchronized with speech, affectively expressive, and suitable for real-time execution. However, existing co-speech methods are primarily developed for digital humans and lack joint support for affective control and low-latency continuous generation on physical embodiments. To bridge this gap, we present SocialHumanoid, a system for expressive humanoid behavior via one-step co-speech motion generation. Given response speech and a specified affective condition, SocialHumanoid generates each full-body motion window in a single forward pass and connects successive windows through motion-history conditioning. The generated human motion is further converted online into embodiment-compatible robot references and tracked by a whole-body controller for physical execution. To provide explicit supervision for affective body expression, we further introduce AffectMoCap, a 4-hour dataset captured from two professional actors, containing synchronized speech, body motion, fine-grained hand motion, and emotion annotations. On BEAT2, SocialHumanoid achieves the best FGD among the compared generation methods, competitive speech-motion synchrony, and approximately 6\times faster inference than GestureLSM under the same protocol. Perceptual evaluations further show that training with AffectMoCap improves affect recognition from generated body motion, while real-robot experiments demonstrate continuous affect-conditioned behavior and stable long-horizon execution. Our project page is this https URL.
[CV-303] LoopTrack: A Simple Baseline for Parameter-Efficient Transformer Tracking
链接: https://arxiv.org/abs/2609.33306
作者: Liang Peng,Chenxiao Li,Libo Zhang,Xingping Dong,Heng Fan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Code and models will be released
Abstract:Current Transformer-based tracking methods typically stack multiple Transformer blocks with separate parameters to model interactions between the target template and the search region for target localization. These trackers often incur substantial parameter overhead from stacked blocks, making their deployment on resource-limited devices difficult. To address this, we propose a parameter-efficient Transformer tracking framework, dubbed LoopTrack, which repeatedly applies a set of Transformer blocks with shared parameters to interact features in a looped architecture for tracking, significantly reducing the number of parameters. To further exploit target cues, we present two lightweight designs, including target-aware looping (TAL) and gated target memory (GTM). The former applies intermediate target information generated by one loop to guide feature interaction in the subsequent loop, enabling progressive feature refinement, while the latter maintains a compact memory across frames, which is incorporated into the loop process to provide long-term information to the tracker, mitigating temporal drift in tracking. Compared to existing Transformer trackers, LoopTrack enables multiple rounds of feature interaction with fewer model parameters, making it resource-friendly for deployment. In extensive experiments on multiple datasets, LoopTrack shows a favorable accuracy-parameter trade-off. In particular, our LoopTrack _\rm One , with a single shared Transformer block, achieves 66.2% SUC score on LaSOT with only 3.4M parameters, while LoopTrack _\rm Three , using three shared blocks, achieves 69.3% SUC score with 6.4M parameters, surpassing existing parameter-efficient tracking methods with comparable or larger model size. With LoopTrack, we aim to establish a simple yet strong baseline for parameter-efficient Transformer tracking. Our code and models will be released.
[CV-304] Relevance Does Not Imply Applicability: Experience Activation for Personal GUI Agents
链接: https://arxiv.org/abs/2609.33304
作者: Fuyao Zhang,Xuan Wang,Zherui Li,Jiaming Zhang,Longtao Huang,Wei Yang Bryan Lim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Personal Graphical User Interface (GUI) agents rely on interaction history to infer what a user wants from ambiguous instructions and to anticipate recurring routines. Existing approaches retrieve task-relevant history and append it to the policy’s context, implicitly assuming that experience relevant to a task remains useful for each decision within it. We find that this help is largely spent at the first decision: retrieved history strongly improves the opening step of an episode, yet provides little sustained benefit over the remaining 90% of steps, and offers weak guidance on whether a proactive suggestion is warranted. A relevant record may tell the agent where to begin, but not which past action applies to the current screen or whether a routine is due now. The underlying issue is that relevance does not imply applicability: relevance is determined at the task level, whereas applicability depends on the situation at decision time. We therefore recast personalization as experience activation and introduce ExpActivator, a training-free framework that activates only the experience applicable to the current situation. During execution, ExpActivator matches each new screen to historical states in the frozen GUI backbone’s latent space and supplies the corresponding action as a reference. Before execution, it activates a recurring intent only when the current time and scenario provide sufficient support, and otherwise abstains. Across four GUI backbones, ExpActivator improves within-trajectory step success by 28% on average, achieves the best personalized execution on every backbone while using about one-fifth as many history tokens, and reaches approximately 2.3 \times the Matthews correlation coefficient of the strongest proactive baseline. Experience pays where it is activated, not where it is appended.
[CV-305] Informative Viewpoint Selection for Episodic-Memory Embodied Question Answering using Omnidirectional Images ACCV2026
链接: https://arxiv.org/abs/2609.33288
作者: Kaname Kitamura,Asako Kanezaki
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: Accepted to ACCV 2026. Supplementary video is provided as an ancillary file
Abstract:Embodied Question Answering (EQA) requires agents to answer natural language questions about surrounding environments from visual observations. In this work, we focus on open-vocabulary episodic-memory EQA (EM-EQA), where an agent answers free-form questions using recorded observation histories. Omnidirectional images are promising for this task, as they provide wide field-of-view observations that can capture surrounding context without requiring explicit camera rotations. However, omnidirectional images introduce two challenges for EQA: (i) equirectangular projection causes severe geometric distortion that degrades vision-language model (VLM) recognition accuracy, and (ii) feeding equirectangular images directly into VLMs introduces excessive irrelevant background information, reducing answer accuracy and increasing the visual-token burden. To address these challenges, we propose a viewpoint selection method for EM-EQA using omnidirectional images. Our method converts equirectangular observations into perspective views via cubemap projection, estimates question-conditioned relevance with fine-tuned BLIP-2, and selects informative and diverse viewpoints through diversity-aware greedy selection. Experiments on the Habitat-Matterport 3D (HM3D) subset of OpenEQA show that our method achieves state-of-the-art model performance among the reported model results with equirectangular observations. Moreover, after removing rotation views, which reduces observation frames by 65.5%, our method largely maintains its answer accuracy.
[CV-306] Q-WAM: 4-Bit Quantization of World Action Models with Action-Subspace Protection
链接: https://arxiv.org/abs/2609.33269
作者: Arash Akbari,Arman Akbari,Jingwu Luo,Yuhao Lei,Yi Gao,Weiwei Chen,Xuan Zhang,Zhenman Fang,Geng Yuan,Yanzhi Wang
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:World Action Models (WAMs) jointly generate video and robot actions through iterative diffusion and perform strongly in robotic manipulation. However, their prohibitive compute and memory costs pose substantial deployment challenges. Post-training quantization (PTQ) can reduce these costs, but existing PTQ methods such as smoothing and rotation are insufficient to maintain the precision of action generation. To overcome this limitation, we propose Q-WAM, a new 4-bit weight-activation quantization for WAMs that preserves the actions the model generates. Specifically, we introduce the \textitAction Observability Gramian (AOG), which measures how much rounding errors in each weighted combination of a layer’s input channels change the final action through all denoising steps. We also develop Action-Subspace Protection (ASP), which keeps the few most action-sensitive channel combinations in a tiny 16-bit low-rank branch and quantizes the complementary weights and activations to 4 bits, both as dense matrix multiplications that run efficiently on GPUs. Finally, to preserve action quality with minimal overhead, we identify the experts that matter most for the generated action by aggregating the AOG-derived action mass across the layers of each expert and apply ASP only to those experts. We evaluate Q-WAM on three WAMs, both in simulation and in real-world deployment. On the RoboTwin 2.0 benchmark, it reaches 89.6–93.0% average success rate, within 1.1 percentage points of the 16-bit models, while reducing the memory of the quantized blocks by 3.1–3.4 \times . Our method outperforms the strongest baseline, SVDQuant, by 2.5–8.7 percentage points. On a Unitree G1 humanoid and a bimanual UR3 robot, it improves success over SVDQuant by 12.8-17.6 percentage points.
[CV-307] VehDyn: A Driving World Model Benchmark for Vehicle Dynamics
链接: https://arxiv.org/abs/2609.33264
作者: Tianyi Wang,Wangsheng Du,Jiazhou Chen,Tianyi Zeng,Xiangyu Li,Jiseop Byeon,Yujin Wang,Yiming Xu,Yangyang Wang,Bingzhao Gao,Sikai Chen,Zhaomiao Guo,Junfeng Jiao,Christian Claudel,Alexandre Bayen
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Emerging Technologies (cs.ET); Robotics (cs.RO)
备注: 48 pages, 27 figures, 19 tables
Abstract:Video world models are emerging as data engines, action planners, and generative simulators for autonomous driving, but existing benchmarks primarily assess visual fidelity and coarse physical plausibility, providing limited evidence on whether generated driving futures obey realistic vehicle kinematics and dynamics. This limitation is further compounded by the lack of datasets in which vehicle, road, maneuver, and speed conditions are independently controlled, and ground-truth vehicle states are recorded in synchrony with videos. We introduce VehDyn, a driving world model benchmark for vehicle dynamics. VehDyn is built on a CARLA-CarSim co-simulation platform where photorealistic rendering is coupled with a validated multi-body dynamics model, and it contains 10,080 configurations from a full factorial design over five vehicle types, four tire-road friction coefficients, three maneuvers, four target speeds, 14 scenes, and three illuminations, each paired with synchronized position, velocity, and attitude sequences. Built on this dataset, VehDyn introduces a hierarchical evaluation framework that measures trajectory alignment, kinematic consistency, and dynamic consistency, and benchmarks 12 state-of-the-art video world models. We further assess the video quality using two established protocols and correlate it with the VehDyn score. Trajectory-level metrics are nearly saturated, with ten of twelve models within 20% of ground truth, while no model reaches 92% of ground truth on dynamic consistency, and visual-quality metrics are only weakly correlated with vehicle-dynamics fidelity. DrivingWorld achieves the highest VehDyn score, followed by Cosmos 3 Nano and LTX-Video 2.5, and the VehDyn score agrees closely with human judgment. VehDyn provides a systematic foundation for developing driving world models that are physically consistent and visually realistic.
[CV-308] EngIntervene: Benchmarking Multimodal Engineering State Understanding and Design Intervention Reasoning
链接: https://arxiv.org/abs/2609.33261
作者: Jinchang Zhang,Yingda Tao,Jiakai Lin,Guoyu Lu
类目: Computer Vision and Pattern Recognition (cs.CV); Software Engineering (cs.SE)
备注:
Abstract:Multimodal engineering benchmarks largely evaluate static understanding, such as recognizing components, interpreting diagrams, or answering technical questions. This leaves a missing middle between engineering perception and full design generation: whether a model can use an understood system state to reason about relations, constraints, and the consequences of design changes. We introduce \textscEngIntervene, a benchmark for this capability. It contains 3,229 questions across seven engineering domains and organizes evaluation into four levels: state grounding (T1), relational and mechanistic reasoning (T2), constraint-aware diagnosis (T3), and intervention reasoning (T4), which asks whether a proposed modification achieves its target while preserving required constraints. The tasks instantiate a unified engineering state representation spanning objects, relations, constraints, and design objectives, and T2–T4 are scored against structured reference answers with atomic criteria. Across open- and closed-weight multimodal models, stronger grounding does not reliably translate into better diagnosis or intervention, and the best open-weight model trails the best closed model by 14.7 percentage points on the T2–T4 average. Removing or shuffling visual evidence consistently degrades performance, while benchmark-specific supervised fine-tuning improves T1 but not T2–T4. T4 further exposes a large gap between satisfying individual revision criteria and producing a fully valid intervention. Engineering reasoning thus requires not only recovering the current state, but also reliably using it to reason about constraints and post-intervention consequences. Code and benchmark artifacts are available at this https URL
[CV-309] PORTER: Edge-Cloud Residency for Persistent 3D Scene Graph Memory
链接: https://arxiv.org/abs/2609.33258
作者: Yue Chang,Yifan Tian,Jiajing Peng,Dazhi Huang,Rufeng Chen,Zhaofan Zhang,Li Chen,Sihong Xie
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent task-driven and just-in-time 3D Scene Graph (3DSG) methods reduce per-task representations by constructing or activating only task-relevant information. Yet sparse per-task working sets do not bound onboard memory usage over a robot’s lifetime: as tasks change, payloads accumulated for earlier tasks may become irrelevant to the current task but can be useful again in future tasks. Over repeated task switches and expanding environments, retaining such reusable payloads causes local memory to grow, whereas discarding them entirely can lead to costly repeated construction of the same payloads later. We introduce PORTER, which decouples persistence from residency: lightweight anchors remain in the limited memory of the edge robot while heavy object payloads migrate between the edge and the cloud. Relevance alone is insufficient for deciding residency because multiple relevant payloads may provide redundant information. We therefore decompose each task into functional requirements and introduce Irreplaceable Support Erasure (ISE), which measures the loss in requirement coverage caused by offloading. ISE discounts replaceable support and penalizes losses more strongly when the remaining coverage of a requirement is weak. PORTER constructs a budget-aware local working set by repeatedly offloading the payload with the smallest marginal ISE per byte. Experiments on JITOMA-Bench evaluate PORTER across four 3DSG builders. Under progressive compression, pooled relative mR@3 remains at 100% through 91% payload-byte offloading.
[CV-310] VGGT-Diff: Visual Geometry Meets Diffusion for Sparse-View Novel View Synthesis
链接: https://arxiv.org/abs/2609.33253
作者: Kangjie Chen,Xiangyu Li,Dongbin Zhang,Chaoda Zheng,Shijia Chen,Jinhao Deng,Hongbin Lin,Choo Sin Wai,Minqi Wang,Minghao Yang,Dake Zhong,Guorui Song,Yu Zhang,Xianming Liu,Boyang Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project page: this https URL , Code at: this https URL
Abstract:We present VGGT-Diff, a geometry-routed multi-view diffusion model for sparse-view novel view synthesis. Existing novel view synthesis (NVS) methods face a fundamental trade-off: reconstruction-based approaches preserve observed geometry but struggle to synthesize unseen regions, while diffusion-based methods provide strong generative priors yet rely on implicit source-to-query correspondence. VGGT-Diff bridges these regimes by routing visual geometry latents from VGGT-\Omega into a pretrained video diffusion model. Each visual token is associated with a 3D point and confidence, then transformed into query-aligned latent conditions through a confidence-aware Visual Geometry Router (VGR) that preserves front and back surface evidence. These conditions guide joint target-view denoising, while Point-Track Residual Consistency (PTRC) regularizes predicted-clean residuals along reliable 3D tracks, improving multi-view stability. We further introduce robust geometry conditioning, combining training-time regularization with inference-time guidance for improved robustness. Experiments show competitive or state-of-the-art performance across interpolation and extrapolation under different viewpoint difficulties. Our code is available at this https URL.
[CV-311] PARSEE-VAD: Efficient Training-Free Online Video Anomaly Detection via Proposition-Aware Reasoning and Streaming Evidence Escalation
链接: https://arxiv.org/abs/2609.33236
作者: Ji Wang,Shuangqing Zhang,Guo-Sen Xie,Fang Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint. 25 pages, 4 figures
Abstract:Training-free online video anomaly detection (VAD) with frozen multimodal language models faces two coupled challenges: extracting reliable current-window semantics under causal and computational constraints, and maintaining temporal continuity without repeatedly transmitting high-dimensional history. Encoding history through text can compress visual evidence and introduce semantic bias, whereas retaining visual history expands multimodal context. We introduce PARSEE-VAD, a two-module framework that separates semantic evidence acquisition from score-state evolution. Proposition-Aware Reasoning (PAR) extracts structured propositional evidence from the current causal window and conditionally activates more specific queries when coarse evidence warrants further refinement. By sharing a reusable causal visual prefix across queries, PAR reduces redundant computation through selective execution. Streaming Evidence Escalation (SEE) maps the acquired proposition evidence into a compact score-domain event state through current evidence escalation, then propagates only the resulting bounded state across decisions to support temporal continuity. Experiments on four benchmarks demonstrate strong training-free online performance while selective routing reduces specialist computation and score-state propagation remains sparse. These results support a current-first principle for streaming multimodal inference: resolve present semantics first, then use compact historical state only to repair residual continuity gaps.
[CV-312] AevaScenes: An FMCW LiDAR Dataset and Benchmark for Long-Range Perception
链接: https://arxiv.org/abs/2609.33230
作者: Gautham Narayan Narasimhan,Heethesh Vhavle,Kumar Bhargav Viswanatha,James Reuther,Deva Ramanan
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: Project page: this https URL
Abstract:FMCW LiDAR measures per-point radial Doppler velocity alongside range, providing a motion cue unavailable in conventional time-of-flight sensors. Exploiting this signal at long range remains understudied. We present an FMCW LiDAR dataset of 575 sequences (57.5K frames) with over 8 million annotated 3D boxes across 16 detection classes and per-point labels across 24 semantic classes, captured by six commercial FMCW LiDAR sensors and six paired 4K cameras across eight Bay Area cities, including 237 nighttime sequences, with annotations extending to 400m. We define a benchmark with three tasks: 3D object detection, scene flow estimation, and semantic segmentation. Detection and scene flow are evaluated across three range bins to 400m, with a public evaluation server. We explore the impact of Doppler measurements on flagship recognition tasks, and find significant improvements up to 2X in detection AP of far-away vehicles and pedestrians, particularly in low-latency single-frame settings. We similarly find scene flow accuracy is significantly improved with Doppler measurements across all ranges. Our dataset and benchmark have been publicly released at this https URL.
[CV-313] Scope-WM: Scoped Computation for Efficient Visual World Models
链接: https://arxiv.org/abs/2609.33218
作者: Chunzheng Li,Zesheng Jia,Hongda Zhang,Jiaying Tang,Yuntian Wang,Siao Liu,Jin Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:Visual world models enable robotic planning by predicting future observations, but dense latent-state propagation and sample-intensive trajectory optimization incur high inference latency and peak memory usage, limiting real-time deployment on resource-constrained platforms. Existing sparse world-model acceleration methods either rely on unguided token sparsification, which may discard planning-relevant information and restrict achievable sparsity, or introduce heavy auxiliary modules and cumbersome multi-stage training pipelines. In this work, we present Scope-WM, an efficient visual world model that scopes computation to prediction-relevant latent regions and promising action sequences. Scope-WM distills prediction relevance into a lightweight action-conditioned selector and applies full dynamics prediction only to a compact subset of selected tokens. It updates the remaining tokens using a compact summary of foreground states and their changes, allowing the background to perceive foreground dynamics without costly token-to-token interactions. During planning, Scope-WM preserves and reuses high-quality action sequences discovered during the initial MPC search, focusing subsequent search under reduced rollout budgets. The resulting pipeline requires only a one-off selector distillation followed by a single joint training stage for the sparse world model. On the challenging Push-T task, Scope-WM reduces peak GPU memory usage and planning time to 18.1% and 14.3% of those of dense DINO-WM, respectively, corresponding to a 6.97\times planning speedup, while maintaining competitive task performance. Further evaluations across five diverse visual planning tasks demonstrate the general applicability of Scope-WM. Code is available at this https URL.
[CV-314] RepFlow: Reciprocal Supervision Improves Generation and Representation in Flow Models
链接: https://arxiv.org/abs/2609.33217
作者: Weili Zeng,Feng Tian,Shengqi Liu,Yichao Yan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Generative models learn visual structure through denoising, yet their internal states are entangled with both noise level and network depth, making it difficult to obtain a stable visual representation from the generator itself. We introduce RepFlow, which learns such a representation from the generator’s evolving computation and uses it to guide generation. Specifically, a separate timestep-free encoder is trained, through a timestep-conditioned predictor, to recover generator states across depths and noise levels from masked clean images. By excluding the reference noise and masked-out content from the encoder’s input, we encourage the encoder to distill visual information that is recoverable from the visible image context and predictive of generator states. The representation learned from the generator’s evolving states is then fed back to guide and improve the generator, whose updated states provide supervision for further representation learning, forming a reciprocal learning process. This reciprocal process improves multi-step generation and representation quality, as measured by frozen linear probing on ImageNet with latent-space SiT and pixel-space JiT, without an externally pretrained representation teacher. Across the two unconditional settings, FID decreases by 19.2 – 40.8% relative to native training, while linear-probe accuracy improves by 6.80–10.04 percentage points over the best searched raw generator features. The learned representation also serves as a distributional metric for one-step JiT post-training, extending its role from instance-level alignment to distribution-level supervision.
[CV-315] Background Gradients Shape Memorization in Flow Matching
链接: https://arxiv.org/abs/2609.33210
作者: Xuanhua Yin,Boyu Wei,Shuyi Zhang,Shunqi Mao,Chuanzhi Xu,Weidong Cai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 38 pages, 8 figures, 34 tables
Abstract:Repetition is closely associated with memorization in generative models, but how other training images affect the retention and copying of targets remains unclear. We study this question in class-conditioned flow matching, where images outside the target set form the background. At fixed target repetition and same-class background row count, replacing repeated same-class images with distinct images reduces the target extraction rate from 80.7% to 18.0%. To explain this effect, we develop a paired-trajectory framework that isolates target-induced parameter displacement and the background gradient response to it. This response has an exact path-integrated curvature representation, connecting background loss geometry to target learning. Reciprocal response transfer between repeated and distinct backgrounds changes target retention and copying in both directions, establishing the response’s causal role. After target removal, the response correction parallel to the target-induced displacement preserves approximately 90% of the copying effects of full response transfer. Directly scaling the displacement also changes copying without further training. The post-removal copying effects of reciprocal transfer are reproduced across datasets and architectures. Together, these results identify the background gradient response as a mechanism through which same-class training data shape the retention of target learning and the reproduction of target images.
[CV-316] WorldAgent : Verification-Guided Agent ic Physical World Construction
链接: https://arxiv.org/abs/2609.33208
作者: Caoliwen Wang,Mengdi Wang,Yige Chen,Zejia Wu,Bowen Huang,Siyuan Chen,Guanxiong Chen,Lifu Wei,Heng Zhang,Qinghai Zhang,Yin Yang,Guandao Yang,Shiying Xiong,Peng Wang,Chenfanfu Jiang,Peter Yichen Chen
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Constructing complex physical worlds from language requires coordinating extensive 3D environments, detailed structures and objects at different spatial scales, and interacting physical processes under both stated goals and implicit physical constraints. We present WorldAgent, an agentic framework for verification-guided physical world construction from a single natural-language prompt, without iterative user debugging. A world construction layer expands the prompt into a structured world specification and uses physical knowledge to build scenes and run numerical simulations. After every step, a verification layer inspects scene geometry and simulation states alongside rendered views. Failed checks guide automatic revisions to the specification and re-execution of the affected steps. Accepted worlds pass the required checks and remain editable for further inspection and resimulation. We introduce AgenticSimBench, on which WorldAgent achieves the best scores among the evaluated agent-based methods on five of seven metrics. In a 26-participant user study, it receives the highest mean ratings across all four criteria.
[CV-317] Structured Residual Connectivity Matters for Diffusion Transformers
链接: https://arxiv.org/abs/2609.33203
作者: Yuhe Liu,Xinyin Ma,Gongfan Fang,Songhua Liu,Xinchao Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Diffusion Transformers (DiTs) have established themselves as a scalable backbone for high-fidelity image synthesis. However, unlike U-Net based diffusion models that rely on rigid, hand-crafted skip connections, DiTs predominantly use a uniform residual stream that integrates all preceding layers as a monolithic state. In this work, we rethink residual connections in diffusion transformers and propose to transform them from passive summation into an active retrieval mechanism optimized for image denoising. First, we conduct a systematic analysis of DiT’s internal representation, revealing a latent preference for early-layer feature reuse and symmetric layer guidance. Motivated by this, we introduce a structured connectivity design that explicitly integrates local residual connections with long-range pathways. Instead of static skip connections or dense all-layer routing, our method enables each transformer block to selectively ``attend’’ to critical earlier representations, dynamically retrieving spatial and semantic cues through direct, differentiable cross-depth paths. Experiments show that our adaptive connectivity leads to faster convergence, with up to 1.73\times fewer training iterations, and significant gains in FID and visual quality with less than 0.1% additional parameters, further improving a strong REPA-XL/2 model from 5.9 to 4.34 FID without guidance and reaching 1.39 FID with classifier-free guidance. Our findings suggest that adaptive cross-layer connectivity is a critical yet underexplored factor in diffusion transformers, and that incorporating structured information pathways provides a simple and effective direction for improving scalable generative models.
[CV-318] ReAL: Accelerating Flow Matching through Segment Advancement with Shared Lookahead
链接: https://arxiv.org/abs/2609.33202
作者: Xuanhua Yin,Chuanzhi Xu,Haoxian Zhou,Shunqi Mao,Weidong Cai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 32 pages, 15 figures, 18 tables
Abstract:Flow-matching models generate high-quality images and videos, but repeated neural network evaluations make sampling expensive. Skipping evaluations reduces this cost by extending an available velocity estimate over a longer span. However, local velocity agreement alone does not determine a suitable span, and checking each candidate endpoint adds costly model calls. We introduce ReAL, a training-free sampler that selects how far to advance using one shared lookahead. Our key insight is that the discrepancy between uncorrected and lookahead-corrected endpoint proposals can be computed directly from the observed velocity mismatch and the candidate span beyond the lookahead. This relation provides a span-dependent selection criterion without additional endpoint evaluations. The same lookahead selects the longest passing candidate span, corrects the accepted update, and supplies its velocity as the next starting estimate. After initialization, each regular iteration requires only one fresh evaluation. ReAL uses the pretrained velocity output and original noise schedule, with no additional training or access to internal features. Experiments cover four image-generation backbones, video generation, and image editing. ReAL achieves 4.91x measured speedup on FLUX.1-dev while retaining 97.0% of dense mean ImageReward. On HunyuanVideo, it achieves a 5.49x speedup while maintaining a VBench score close to that of dense sampling.
[CV-319] FocusDrive: Reasoning with Visual Focus for Autonomous Driving
链接: https://arxiv.org/abs/2609.33190
作者: Zhiyuan Liu,Zehong Ke,Yuanxin Tian,Hao Cheng,Jinhao Li,Yining Xing,Yanbo Jiang,Zhenhua Xu,Wenhao Yu,Jianqiang Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Driving decisions depend on both where to focus and how to act on what is seen. Effective driving reasoning must establish which objects matter, where they are, and how they inform the intended action. Text-based rationales can describe a driving response while leaving its correspondence to specific visual evidence implicit. Visual focus provides a concrete starting point for this connection by identifying what matters in the scene and where it is. We propose FocusDrive, a structured multimodal reasoning framework that organizes end-to-end planning around explicit visual focus. It pairs descriptions of decision-relevant objects with image-patch references, bringing explicit visual focus into the reasoning that generates driving plans and trajectories. We first assess this focus representation through driver gaze prediction, then investigate its role in planning reasoning using driving-focus annotations within existing NAVSIM training scenes. Experiments on W3DA and NAVSIM demonstrate competitive gaze prediction and end-to-end planning performance, with FocusDrive improving over text-based chain-of-thought. These results support visual focus as an effective link between scene understanding and driving action.
[CV-320] When an Evaluation Rule Writes Training Labels: Measuring Human-Reference Forgiveness in NAVSIM
链接: https://arxiv.org/abs/2609.33189
作者: Jiaxuan Guo,Jingxin Yang,Jiaqi Ye,Youran Sun,Shuo Xin,Kejia Zhang,Haizhao Yang
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:When the human reference scores zero on a metric, the released GTRS-Dense label generator for NAVSIM marks every candidate trajectory in the scene as passing it. NAVSIM’s authors introduced this human-reference forgiveness to avoid penalizing contextually justified maneuvers when scoring one trajectory, and warned that it could overlook important failures. In label generation it sets a whole column of 16,384 candidate targets to passing. To measure the consequences for supervision, we re-run the generator with the overwrite disabled and compare the pre-overwrite targets with the released labels on all 103,288 navtrain scenes. The rule erases a candidate distinction that the training loss reads on 11,237 of them (10.8793%). Firing usually changes most of a column: lane keeping carries 9,982 of the 13,042 forgiven loss columns, and its median forgiven column had 14,391 of 16,384 candidates failing before the overwrite. On held-out navtest scenes forgiven on lane keeping, the released lane-keeping head’s median AUC against the pre-overwrite outcome is 0.7095; on unforgiven scenes matched on failing-candidate count it is 0.9807. For the Hydra-MDP checkpoint released with GTRS, whose configuration takes the same label file, the two values are 0.6627 and 0.9761. Continuing the released GTRS-Dense checkpoint for 300 optimizer steps with three paired seeds, we observe the forgiven-scene AUC 0.1086-0.1251 higher with pre-overwrite than with published targets, and a narrower gap between matched groups, still above zero. Scoring with forgiveness disabled, we observe lane keeping higher by 2.478-3.524 points on navtest scenes forgiven on any of five loss metrics, with lower adjacent-frame plan consistency. Both changes are larger there than on the rest. EPDMS, scored the same way, does not separate the two target sets.
[CV-321] Can Protein-Derived Knowledge Improve Pathology Foundation Models?
链接: https://arxiv.org/abs/2609.33178
作者: Di Zhang,Zhangpeng Gong,Jiashuai Liu,Zhi Zeng,Jiusong Ge,Chunze Yang,Xitong Ling,Kai Yi,Kai He,Weimiao Yu,Mireia Crispin-Ortuzar,Chen Li,Zeyu Gao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Molecularly guided pathology foundation models (PFMs) exploit transcriptomic or proteomic information to enrich whole-slide image (WSI) representations, yet effectively leveraging large standalone molecular corpora remains challenging. First, existing molecular foundation models encode protein sequences or single-cell states, not the patient-level bulk expression profiles paired with WSIs. Second, because cross-modal supervision is restricted to paired WSI-omics samples, knowledge from standalone molecular corpora reaches the pathology encoder only indirectly, creating a paired-support bottleneck. To address these challenges, we propose a three-stage framework that decouples proteomic knowledge acquisition from cross-modal transfer, yielding ProSlide, a slide-level hierarchical pathology foundation model. First, to close the modality gap, we pretrain a Proteomic Foundation Encoder (PFE) on 12,695 sample-level bulk protein profiles using virtual profile generation and expression-space multi-view pretraining. Second, we pretrain ProSlide, a patch-region-slide encoder, to predict protein expression from paired WSI-protein samples. Third, to relax the paired-support bottleneck, we introduce Prot2Path, a cross-modal relational distillation objective. For each paired sample, it aligns the similarity distributions of the WSI and its protein profile over a shared, frozen bank of PFE-encoded paired and standalone profiles. We evaluate ProSlide on 12 downstream tasks across breast, lung, and renal cancers. Despite being pretrained with only 2,229 WSIs and 12,695 sample-level protein profiles, ProSlide achieves the highest mean accuracy and AUC within each cancer group.
[CV-322] DeltaWAM: Change-Centric Visual Foresight via Delta Tokens for an Efficient World-Action Model
链接: https://arxiv.org/abs/2609.33177
作者: Tianyun Jiang,Wenrui Bao,Bingxin Xu,Yu Tian,Yuzhang Shang
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:World-Action Models (WAMs) offer visual foresight for robotic manipulation, but pixel-space models repeatedly reconstruct entire future scenes, incurring high computational cost and spatio-temporal redundancy. In physical manipulation, consecutive frames often share most of their visual context; the changes between them are what an action policy needs to anticipate. We introduce DeltaWAM, a change-centric WAM that makes a compact delta token the unit of future prediction. Each token is a single vector encoding changes between consecutive dense DINO feature maps. DeltaWAM builds on DeltaWorld, a latent world model pretrained on large-scale videos, to autoregressively predict one delta token per future frame. A flow-matching action expert then conditions on the predicted transitions and current DINO features, which serve as spatial anchors, to generate action chunks. Trained for 256 GPU hours on two H100 GPUs, DeltaWAM has 0.725B parameters and achieves 92.8% average success on LIBERO. It also shows robust generalization under procedural perturbations on LIBERO-Pro. Inference takes 142.1 ms per action chunk with 3.86 GB peak memory.
[CV-323] ABO-Med: Accelerated Bilevel Optimization for Few-Shot Medical Image Classification
链接: https://arxiv.org/abs/2609.33176
作者: Ruoxuan Shi,Sheng Yang,Zhengxing Su,Xiaoyang Hou,Yating Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:In recent years, bilevel optimization has been widely used in a variety of machine learning tasks. However, prior bilevel optimization algorithms generally require the computation of second-order information, which limits their practical scalability. Only recently has a first-order paradigm for bilevel optimization been established, attaining near-optimal theoretical guarantees for solving bilevel optimization problems. In this paper, we propose ABO-Med, a scalable instantiation of this paradigm for few-shot learning, by incorporating it into the model-agnostic meta-learning (MAML) framework and tailoring it to medical image classification. We also introduce Medical Adaptive RandomAugment (MedRAug), a modality-aware augmentation strategy designed for medical images. Theoretically, ABO-Med establishes the optimality of MAML-type meta-learning approaches. Empirically, ABO-Med outperforms prior baselines on several public medical datasets, with gains of 1.99% to 18.76%, while MedRAug further improves the average accuracy by 2.20% to 6.34%. Additional cross-domain experiments, augmentation ablation studies, backbone ablation studies, and training efficiency analysis further validate the effectiveness and efficiency of the proposed method.
[CV-324] Dynamic Manipulation with World-Action Models via Counterfactual Planning
链接: https://arxiv.org/abs/2609.33172
作者: Sunwoo Park,Wonbin Lee,Seonghyun Jin,Youngmin Kim,Jangho Park,Jong Chul Ye
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 48 pages, including appendices. Project page: this https URL
Abstract:World-Action models (WAMs) trained on static demonstrations often fail to manipulate moving targets even when they possess the required manipulation skills. We attribute this failure to target-response collapse: as execution advances, the policy becomes increasingly biased toward the learned continuation of its ongoing behavior and less responsive to target relocation. To bridge the gap between what the model has learned and what it can generate from the current context, we formulate dynamic manipulation as counterfactual planning by decoupling the context used for plan generation from the physical state used for execution. Our framework, Dynamic Predictive Planning (DPP), first uses the WAM’s predictive rollout to estimate when an interaction is expected to occur, and combines this timing estimate with observed target motion to predict the target’s future interaction position. DPP then constructs a counterfactual observation that places this predicted target position in a familiar robot context, allowing the model to invoke an existing manipulation skill rather than generate a recovery behavior from an unfamiliar robot-target configuration. The resulting plan is connected to the robot’s actual state during execution. DPP enables real-time dynamic manipulation on a single consumer GPU without additional training on dynamic data. Experiments in simulation and on a real robot demonstrate consistent improvements across diverse target motions, with simulation performance surpassing all evaluated baselines, including methods additionally trained on dynamic data. Project page: this https URL Comments: 48 pages, including appendices. Project page: this https URL Subjects: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) Cite as: arXiv:2609.33172 [cs.RO] (or arXiv:2609.33172v1 [cs.RO] for this version) https://doi.org/10.48550/arXiv.2609.33172 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Jong Chul Ye [view email] [v1] Sun, 27 Sep 2026 03:48:20 UTC (6,792 KB)
[CV-325] Perturb-and-Solve: Efficient Learned-Operator Conditioning for Latent Diffusion Inverse Problems
链接: https://arxiv.org/abs/2609.33171
作者: Abduragim Shtanchaev,Arip Asadulaev,Luiza Labazanova,Aidar Alimbayev,Karim Salta,Eric Moulines
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Latent diffusion models serve as powerful priors for solving inverse problems in image restoration, such as deblurring, inpainting, and super-resolution. Current methods have a trade-off between generality and efficiency. Solvers that are restricted to a fixed set of degradation operators are fast and efficient. Methods that support arbitrary degradation operators are slow and require gradients through the diffusion network. To break this bottleneck, we introduce PASEO (Perturb-And-Solve for Efficient Operator conditioning), a method that uses a small (1M parameters) learned network to degrade diffusion model predictions in latent space. PASEO supports learned degradation operators without back-propagating through the diffusion network. We efficiently sample reconstructions from an approximate posterior by combining the diffusion model’s prediction with the observed image. We do this by adding noise and solving linear equations based on a local linear approximation of the learned network, without building or inverting large covariance matrices. Across super-resolution, deblurring, and inpainting on FFHQ and COCO, PASEO achieves strong perceptual quality while running up to 9x faster and using up to 34% less peak memory than the tested baselines, with the same or fewer model evaluations.
[CV-326] FloodDiffusion 2: Efficient and Path Controllable Streaming Motion Generation
链接: https://arxiv.org/abs/2609.33167
作者: Yiyi Cai,Yuhan Wu,Kunhang Li,Tu Fangyuan,Xiangyue Zhang,Qiaoge Li,Zhixiang Wang,Kaipeng Zhang,Haiyang Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 27 pages. Code: this https URL
Abstract:We present FloodDiffusion 2 (FD2), an efficient and controllable framework that builds upon FloodDiffusion (FD1), a state-of-the-art streaming motion generation model. While FD1 produces plausible motion, it suffers from low efficiency and limited controllability, as its attention design requires repeated computation over the entire history, and it lacks precise trajectory control for real-world applications. To address these limitations and improve generation quality, FD2 introduces three advances. First, Partial Attention makes finalized history representations independent of the active window, enabling KV-cached inference and shared-history packing for efficient training. Second, we establish a necessary-and-sufficient Bregman criterion for regression losses to preserve diffusion’s conditional-mean velocity field. This criterion guides an FK-induced quadratic loss that incorporates motion geometry without online FK evaluation. Third, FD2 introduces precise path conditioning to control the character’s root trajectory while preserving natural body motion. Experiments show that FD2 reduces training computation by 4.6 \times and accelerates denoising by 11.29 \times , reaching 2.303 ms per update on long sequences. Alongside these efficiency gains, FD2 improves motion quality over FD1 and achieves state-of-the-art FID scores among streaming methods, with 0.048 on SEED and 0.053 on HumanML3D.
[CV-327] FOCUS: Benchmarking Retinal Model Generalization from Foundation Vision Encoders to Multimodal LLM s NEURIPS2026
链接: https://arxiv.org/abs/2609.33158
作者: David Restrepo,Chenwei Wu,Luis Filipe Nakayama,Miguel L. Martins,Stergios Christodoulidis,Maria Vakalopoulou,Enzo Ferrante
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted at NeurIPS 2026, Evaluations Datasets Track. Interactive benchmark dashboard: this https URL
Abstract:Progress in AI-based retinal image analysis has advanced with foundation models, yet evaluating their reliability remains challenging. Performance reported on a single dataset does not capture how models behave under dataset shift, across clinical definitions, or for different patient subgroups. This limitation is particularly critical in medical imaging analysis, where robustness, calibration, and fairness are essential for safe deployment. We introduce FOCUS (Foundation Ophthalmic Cross-Dataset Understanding under Shift), a cross-dataset benchmark for evaluating retinal fundus models that considers vision-only encoder models (VM), vision-language dual-encoder models (VLM), and multimodal large language models (MLLM). FOCUS harmonizes binary diabetic retinopathy, referable diabetic retinopathy, and glaucomatous optic neuropathy tasks across ten public datasets spanning diverse geographies, acquisition conditions, and label protocols. The benchmark evaluates models through a unified analysis layer that measures ranking performance, calibration, subgroup disparities, and image-quality robustness. We present a large-scale evaluation covering 532 base configurations and 228 MLLM configurations adapted through supervised fine-tuning with low-rank adaptation (LoRA). Results show that no model family consistently dominates across tasks and datasets: general VM encoders achieve the strongest average ranking performance, medical MLLMs are competitive but variable, and dual encoder VLMs benefit substantially from lightweight adaptation. Fine-tuning improves in-domain performance but exhibits heterogeneous transfer to external datasets, particularly in calibration. These findings demonstrate that retinal model evaluation is inherently multidimensional. FOCUS provides a practical framework and public benchmark to assess generalization, reliability, and robustness beyond single-dataset leaderboards
[CV-328] DroneWAM: Efficient World Action Model for Drone Visual Navigation
链接: https://arxiv.org/abs/2609.33148
作者: Liang Yao,Fan Liu,Hongbo Lu,Wei Xu,Jianyu Jiang,Yijun Shen,Chuanyi Zhang,Pai Peng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:World-action models give visual navigation agents a way to anticipate how candidate actions will change future observations and to act from the predicted consequences. For drones, this capability must operate under tight accuracy and efficiency constraints. We present DroneWAM, an efficient world-action model for drone visual navigation. DroneWAM adopts a JEPA-based architecture to model future states directly in representation space, avoiding the cost of explicit future image generation. A pretrained Resampler further compresses dense encoder features into fewer latent tokens, reducing the computation repeated at each imagined step. We also introduce adaptive rollout, where a preference-trained Gate adaptively allocates prediction depth according to the current scene. To support learning under richer aerial motion, we construct DroneNav-6D, a simulated visual navigation dataset with synchronized RGB observations, 6-DoF flight trajectories, control commands, and randomized wind disturbances. On DroneNav-6D, DroneWAM achieves the best trajectory accuracy among the compared methods. Adaptive rollout further reduces the average prediction depth from 8 to 4.58 while improving trajectory accuracy, demonstrating that predictive computation can be allocated more effectively across scenes. \hrefthis https URLCodes and data will be released.
[CV-329] rain Together or Merge Later? Unifying VLA Experts via a Shared Action Interface
链接: https://arxiv.org/abs/2609.33125
作者: Zhizhen Zhang,Yuxia Fu,Zijian Wang,Helen Huang,Yadan Luo
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Co-training offers a straightforward way to build a multi-task vision-language-action (VLA) policy, but can fall short of the performance achieved by training each task independently. The challenge is to retain these task-specific gains in a multi-task policy without joint post-training. Combining independently trained experts through model merging is a natural approach, yet strong individual experts do not necessarily yield a strong merged policy. We identify one source of this incompatibility: task-specific changes to the action interface, comprising action normalization and the action encoder and decoder. We propose PolicyWeave, combining merge-compatible post-training with context-guided sparse merging. During post-training, all experts retain the common base policy’s action interface, while task adaptation is restricted to LoRA updates in the hidden layers of the action model. This makes the experts more compatible with existing model merging methods. However, merging all experts can still introduce interference from unrelated tasks at deployment. PolicyWeave scores each expert’s LoRA updates using the initial visual-language context, determines the expert set through leave-one-layer-out ranking stability, and forms a sparse weighted merge of the selected updates that remains fixed for current task. We evaluate PolicyWeave with GR00T N1.5 on 18 RoboCasa365 tasks, using only 10% of the target-task demonstrations for supervised fine-tuning (SFT). Preserving the shared action interface raises the average success rate across four static merging methods from 17.0% to 52.8%. PolicyWeave achieves 64.7% success with these SFT experts and 74.1% after task-specific reinforcement learning (RL), compared with 60.7% for joint RL. Further evaluations on LIBERO-10 and an AgileX Piper arm support the deployment of independently learned skills in long-horizon and real-world manipulation.
[CV-330] oward Comprehensive 3D Grounding: Orientation Grounding through Vision-Language Models
链接: https://arxiv.org/abs/2609.33109
作者: Tuo Liang,Disheng Liu,Nengbo Wang,Vipin Chaudhary,Yu Yin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Grounding is a core capability of spatial vision-language models, yet most existing work focuses only on where a referred object is. Many 3D tasks also require knowing how it is oriented. Although existing 3D VLMs may predict oriented boxes, box pose does not explicitly capture object-centric orientation or symmetry-induced ambiguities. We introduce orientation grounding, a referring grounding task that predicts an object’s 6D orientation and axial symmetry from a language or box query in single-view or multi-view scenes. To support this task, we construct ReferOri, with 331K multi-view and 387K single-view orientation-grounding queries obtained through scalable reconstruction, consistency checking, and human verification. We further present OG-VLM, which adapts a 3D VLM with structured box/orientation outputs, sign and symmetry tokens, and geometry-aware auxiliary losses. Across single-view and multi-view benchmarks, OG-VLM substantially outperforms orientation-aware VLM baselines and surpasses object-level orientation foundation models on scene-level referring benchmarks, showing that explicit orientation grounding is a distinct and learnable capability beyond localization. Downstream results validate its benefit for orientation-related spatial reasoning.
[CV-331] Octree-based Video Representation
链接: https://arxiv.org/abs/2609.33100
作者: Rungui Zhou,Chuanzhi Zhou,Yuk-Kit Hou,Peng-Shuai Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video models commonly use uniform grids even though visual complexity varies substantially across space and time. We introduce OctVideo, which approximates a video clip with an octree. This hierarchy recursively partitions a spatio-temporal volume into eight subvolumes, so that smooth regions remain coarse while detailed regions receive finer cells. Each leaf stores local RGB values and spatio-temporal gradients, supplemented by a lightweight learned residual. For reconstruction, a Conv1D VAE maps the serialized cells to a regular latent grid and selectively refines details during decoding. Our VAE achieves 36.12 dB PSNR with 38.2M parameters and 189.4 GFLOPs per clip on Kinetics-400 (K400). It also generalizes zero-shot to the high-resolution Densely Annotated VIdeo Segmentation (DAVIS) 2016 dataset with reconstruction quality comparable to the best evaluated models. On both datasets, it requires the fewest model FLOPs and achieves the fastest encoding and decoding among the evaluated models. OctVideo also supports video understanding, achieving competitive recognition performance with few input tokens when trained from scratch. By exploiting the redundancy already present in video signals and efficiently processing sparse structures, OctVideo provides an efficient representation for video.
[CV-332] Query Align and Distill: Navigation-Aware Cross-Modal Interaction for Efficient Vision-and-Language Navigation
链接: https://arxiv.org/abs/2609.33097
作者: Zhihao Chen,Yiyuan Ge,Ziyang Wang,Pu Cao,Lu Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:Recent large-scale Vision-and-Language Navigation (VLN) models deliver strong accuracy but remain costly to deploy due to their large parameter counts and computational requirements. We tackle efficient VLN in two steps. First, we build a high-performing teacher that makes navigation evidence selection explicit and compressible. The teacher introduces a small set of learnable query slots to extract global and local action-sufficient navigable evidence from panoramic observations via a Navigable Query Generator, then progressively grounds these evidence tokens to the instruction with an Instruction-Query Aligner for policy prediction. Second, using this explicit query bottleneck as a distillation interface, we train a compact student by transferring both where to attend and what to do. We distill the teacher’s global and local navigable queries with a navigation-aware token-adaptive objective, then further match action distributions during fine-tuning. Experiments on standard VLN benchmarks demonstrate that our student nearly matches the teacher’s navigation performance while reducing the number of parameters by 93.65% compared to the teacher.
[CV-333] QSCP: Beyond Class-Name Prompts for Query-Guided Semantic Change Parsing
链接: https://arxiv.org/abs/2609.33088
作者: Yuan Qian,Jie Ma
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Traditional change detection (CD) identifies changes between bi-temporal remote sensing images, while semantic change detection (SCD) assigns predefined land-cover classes. However, mapping all changes may not meet a user’s specific needs. Referring change detection (RCD) enables selective retrieval through category prompts. However, existing category-prompted RCD uses the queried category to specify the destination of a change and returns only a binary mask of the corresponding regions. Users may instead request a particular transition and paired semantic maps to understand what changed into what. Such requests require explicit source and target reasoning beyond target-class localization. To address these needs, we propose query-guided semantic change parsing (QSCP), which supports category names, synonyms, and intent-bearing sentences and returns a query-specific mask with paired temporal semantic maps. QSCP parses requests into intents and semantic slots, composes bidirectional visual evidence, and predicts both temporal states with a query-conditioned decoder. On SECOND, QSCP outperforms RCDNet on synonym, sentence, and transition queries and improves end-to-end semantic prediction over evaluated semantic baselines. WHU-CDC experiments further assess cross-dataset transfer and consistency across equivalent expressions without target-domain training. Code is available at this https URL
[CV-334] dKFD: Phase-Structured Evidence Allocation for Fixed-Budget Localized Event Understanding
链接: https://arxiv.org/abs/2609.33083
作者: Aditya Bagri,Ashutosh Kumar,Chaitanya Lakhchaura,Avinash Anand,Zhengkui Wang,Rajiv Ratn Shah
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Sparse video understanding often requires selecting a small set of visual evidence under a fixed frame budget. Most sparse selectors allocate this budget globally, allowing all frames to compete with one another. For temporally localized events, this can be a poor inductive bias: useful evidence is often distributed across pre-event context, the event itself, and post-event consequences. We study fixed-budget evidence allocation for localized event videos and show that globally competitive Top- K selectors can preserve recognition and grounding while producing unstable event evidence. On DoTA Video Anomaly Recognition, Global Top- K obtains competitive recognition and temporal grounding, but low selector-event alignment at K=12 (Frame AUC 51.5 \pm 10.8 ). We propose dKFD, a phase-structured differentiable selector that reserves evidence capacity across pre-event, event, and post-event phases after full-sequence temporal encoding. Under matched-budget multi-seed evaluation, dKFD improves Frame AUC by +30.97 over a matched Global Top- K selector at K=12 ( p0.01 ), while yielding modest but statistically significant recognition gains and comparable temporal grounding. Mechanism ablations show that phase supervision is load-bearing: removing it reduces Frame AUC to 41.1 \pm 12.0 even when phase-partitioned budgets are retained. Downstream diagnostics on VRU-Accident show consistent gains over learned Global Top- K across VLM families, while dense captioning reveals a boundary condition where uniform sampling remains competitive. These results support phase-structured allocation as a controlled fixed-budget approach for event-centric sparse evidence selection, not as a universal video summarization strategy.
[CV-335] SemReward-VL: Semantic Reward-Guided Video-Language Adaptation for Developmental Behavior Assessment
链接: https://arxiv.org/abs/2609.33082
作者: De Jiang,Shuo Zhang,Kehong Yuan,Hongen Liao
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Developmental screening videos show how children perform specific behaviors, but clinical records usually contain outcomes rather than descriptions of what happened. We present SemReward-VL, which learns to describe item-specific behavior from these outcomes. A vision-language model generates a description, and a frozen language model scores its agreement with the clinical outcome, relevance to the item, abstention on unrelated video-item pairs, and clarity. Group relative policy optimization (GRPO) updates LoRA adapters using this semantic reward. On 13,379 videos covering 41 items, the method improves aggregate accuracy and the number of items with recall above 0.5. Errors remain in temporal direction, duration, and age-specific interpretations of behavior.
[CV-336] Parameter-Efficient 3D Segmentation of Liver and Liver tumors: Depthwise factorization Scales Better Than Dense Convolution with Spatial Dimensionality
链接: https://arxiv.org/abs/2609.33077
作者: Adham M. Alkhadrawi,Mohammed A.B. Mahmoud
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Three-dimensional dense convolutional networks are the strongest performers on volumetric medical image segmentation, but their parameter counts scale poorly: moving a dense k x k convolution to k x k x k multiplies its weights by k. We observe that depthwise separable factorization does not share this penalty. Because the cubic kernel term applies only to the depthwise stage while the pointwise projection, which dominates the parameter count, is unchanged, the same architecture grows by 5 % from 2D to 3D where a dense convolutional U-Net grows by 200 %. We exploit this asymmetry to build a 3D U-Net with 536,990 parameters, 24x fewer than an identical dense 3D U-Net. On MSD Task03 Liver (the Medical Segmentation Decathlon liver task, derived from LiTS), evaluated per case under five-fold cross-validation over all 131 public volumes, the model reaches a tumor Dice of 0.577 (95% CI [0.518, 0.633]) and a liver Dice of 0.947 (95% CI [0.941, 0.952]). Its liver Dice exceeds previously reported performance. Its tumor Dice exceeds their low-resolution configuration (0.4701) by 0.107 and their 2D configuration (0.5394), at approximately one twenty-fourth of the parameters and roughly half the in-plane resolution. Trained under identical conditions on a common held-out split, it exceeds a dense 3D U-Net on liver by +0.031 Dice (paired p = 0.006) and on tumor by +0.041 (95 % CI [+0.005, +0.081], paired p = 0.056), suggesting the factorization also acts as a regularizer in the small-data regime characteristic of medical imaging. We further show, on both LiTS and a 2D endoscopy benchmark, that a large fraction of the network’s learnable spatial filters can be replaced by fixed shifts at no cost in accuracy, but that replacing all of them is measurably worse, the placement of spatial capacity matters more than its total amount.
[CV-337] NutriVision: Ingredient-Conditioned Fusion and Prediction for Single-Image Food Nutrition Estimation
链接: https://arxiv.org/abs/2609.33076
作者: Aman Kumar,Avinash Anand,Chaitanya Lakhchaura,Ashutosh Kumar,Akshita Abrol,Timothy Liu,Zhengkui Wang,Rajiv Ratn Shah
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Nutrition estimation is a fundamental task in consumer diet tracking, clinical dietetics, chronic disease management, sports and hospital nutrition, and broader food computing systems. The existing approaches have progressed along two largely separate axes, vision models that rely on calibrated RGB-depth captures and ingredient-aware methods that use textual cues but use limited multimodal fusion. We introduce NutriVision, an end-to-end framework that leverages visual geometry and ingredient semantics to estimate calories, mass, fat content, carbohydrates, and protein from a single RGB image and an optional ingredient list. It obtains the unavailable depth modality using DepthAnything-V3 and encodes ingredient descriptions using CLIP. It integrates three complementary mechanisms: (1) an \emphIngredient-Conditioned Frequency-Aligned Fusion Module (IC-FAFM), which uses textual guidance to reweight and align RGB-depth frequency components; (2) an \emphIngredient-Aware Mask-based Prediction Head (IA-MPH), whose gating and channel masks are conditioned on food identity; and (3) modality-specific \emphInternal Semantic Modeling (ISM) blocks. On the Nutrition5k dataset, NutriVision achieves a mean PMAE of \mathbf13.60\pm0.10% , outperforming our IGSMNet implementation by 0.89 percentage points and OmniFood8k by 2.90 percentage points (both p0.001 ). The module-level ablations identify the ingredient-aware prediction head as the primary architectural contributor, improving mean PMAE by 1.50\pm0.17 percentage points ( p0.001 ). These results demonstrate that ingredient-conditioned prediction and frequency-aware RGB-depth fusion provide measurable gains for single-image nutrient estimation. More broadly, NutriVision offers a practical route toward nutrition-assessment systems that exploit geometric and semantic cues without requiring specialized depth-sensing hardware
[CV-338] Residual Diffusion Implicit Models
链接: https://arxiv.org/abs/2609.33020
作者: João Guerreiro,Pedro Tomás,Helena Aidos,Jacinto C. Nascimento
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 38 pages, 17 figures. Implementation available at this https URL
Abstract:Diffusion models achieve state-of-the-art results across multiple tasks. However, in inverse problems, standard initialization from pure Gaussian noise misaligns the generative process with real-world degradations. More recent methods such as diffusion bridges impose strict endpoint constraints and often require long reverse processes that are prone to hallucinations. Alternative consistency models provide noise-invariant, one-step mappings but lack inherent variance modeling and can degrade under severe corruption. Hence, residual diffusion implicit models (RDIMs) are proposed, constituting a generalized framework that explicitly models the residuals between high-quality (HQ) and low-quality (LQ) images, aligning the forward process with the actual degradation. A non-Markovian implicit reverse sampler is derived, which can skip intermediate timesteps, enabling accurate few-step or even single-step reconstruction, while mitigating the hallucinations inherent to long diffusion chains. RDIM also introduces a controllable variance mechanism that interpolates between deterministic and stochastic sampling, balancing fidelity and diversity. Furthermore, it enables the straightforward use of perceptual losses, when needed. Experiments on denoising and super-resolution benchmarks demonstrate that RDIMs consistently outperforms the state of the art, including bridge and consistency models, in terms of PSNR, SSIM, and LPIPS, reducing hallucinations while requiring only a few sampling steps (often just one). The results position RDIMs as an efficient solution for a broad range of image restoration tasks.
[CV-339] Safety-Constrained Cascade Inference for Robust Malaria Cell Classification Under Field Corruptions
链接: https://arxiv.org/abs/2609.33005
作者: J. T. Hagbe,Michel Emel
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 28 pages, 6 figures, 8 tables. Code available at this https URL
Abstract:Automated malaria diagnosis from thin blood-smear microscopy could meaningfully reduce the burden on under-resourced laboratories, but a model that maximises accuracy on clean laboratory images fails badly the moment an inexpensive smartphone camera introduces sensor noise. This paper introduces MalariaCascade, a two-stage inference system in which a lightweight MobileNetV2 sentinel (2,225,153 parameters) makes confident classifications at low compute cost and escalates uncertain cases to an EfficientNet-B3 expert (10,697,769 parameters) that sees only clean, standardised images regardless of how corrupted the incoming frame is. Structural isolation of the expert stage, not learned robustness, is the mechanism. The sentinel is trained under a safety-score objective that places an explicit floor on Recall(Parasitised) (=0.95) and Recall(Uninfected) (=0.40) before precision is optimised, ensuring the checkpoint satisfies clinical safety constraints by construction. On the NIH Malaria Cell Images Dataset (27,558 cells), the cascade reaches Accuracy=0.9736, Recall(Parasitised)=0.9570, Precision(Parasitised)=0.9912, F1=0.9738, and AUROC=0.9955 on the clean test set. Under Gaussian sensor noise at full severity, cascade Recall(Parasitised) degrades by only 2.4 pp (0.9570 to 0.9329), while the flat single-model baseline collapses by 63.0 pp (0.9584 to 0.3281). McNemar’s test confirms the cascade improvement is statistically significant (chi^2=11.14, p=0.00085). An ablation isolating the structural property shows that routing clean images to the expert is responsible for a 16.8 pp Recall(Parasitised) advantage under sensor noise relative to a cascade where the expert also sees corrupted inputs.
[CV-340] Certified Interface Aliases: Exact Collisions in Vision-Language Preprocessing and When They Exist
链接: https://arxiv.org/abs/2609.33003
作者: Mert Onur Cakiroglu,Elham Buxton,Mehmet Dalkilic,Hasan Kurban
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-language verifiers and routers must distinguish errors repairable by more reasoning from those caused by visual evidence never reaching the language model. This distinction lacks ground truth because annotators see full-resolution images while models receive preprocessed tensors. We introduce AliasForge to create cases where the relevant fact is provably absent from the interface. Fixed-point resampling makes the pre-rounding resize an exact integer linear map that can send nonzero integer perturbations to zero. Hiding a label-flipping perturbation there produces images with opposite step-correctness labels but bit-identical interface states. Every verifier therefore has the same output law on both members, giving pair-balanced accuracy exactly one half and zero gain from language-side repair. We prove that every fixed-point downscaler has such null vectors and bound their smallest size at the most common ratios, which rules out an 8-bit fit whenever the bound exceeds 255. From the resize configuration alone, a lattice criterion supplies realizable collisions and certifies their absence within the specified construction family. It resolves all 20 screened configurations, 17 as constructible and 3 as non-constructible. We construct certified pairs across three architectures and certify four additional processors, with zero decision-logit gap on all 18 scored pairs and none of the 18 controls. The pairs also screen routers that waste computation on re-attention or further reasoning. On natural items, per-item routing headroom exists, but no tested interface-only router improves over stopping. Our fiber ceiling bounds the headroom recoverable from the interface. Code: this https URL.
[CV-341] riDrive: Joint Driver Vehicle and Road Modeling for Forecasting and Driver Monitoring
链接: https://arxiv.org/abs/2609.33000
作者: Yuhang Wang,Jingxin Yang,Chuheng Wei,Yuechen Guo,Jinghan Xu,Zhao Han,Hao Zhou
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 32 pages (10-page main text plus appendices), 4 figures, 25 tables. Under review. Code, checkpoints, and demo video: this https URL
Abstract:Predicting how drivers, vehicles, and road scenes interact and evolve together is central to driver monitoring. Prior work models in-cabin activity or traffic-conditioned driver motion in isolation, motivating joint driver, vehicle, and road modeling with real-time on-vehicle evaluation. We introduce TriDrive, to our knowledge the first unified framework that jointly forecasts driver kinematics, vehicle dynamics, and road demands through an automation-conditioned transition model. Modality-specific encoders (an anchored kinematic representation of the driver, causal CAN-bus dynamics, and frozen V-JEPA 2 road latents with structured road margins) are connected by directed residual connections through which driver and road context refine vehicle forecasts. We evaluate TriDrive on three downstream tasks. On the public AIDE benchmark, its kinematic encoder recipe sets a new full-set state of the art (SOTA) among published baselines (48.05 versus 71.47 All-MPJPE). On 197.2 hours of naturalistic BATON subset, directed connections and road margins raise assistance-engaged PR-AUC by 0.084 for steering onset and 0.286 for time-to-collision drops. For real-time use, we distill the road encoders and run TriDrive on a comma four with an external 8 GB GPU, where a lightweight current-state warning probe updates at 5 Hz with 177 ms p95 latency while the joint model forecasts concurrently. The probe is above an openpilot-based baseline on human-labeled manual-driving warnings (AUROC 0.725 versus 0.563), and in a paired on-road study 14 drivers rate its warnings as more appropriate (+1.79) and timely (+2.67) than those of openpilot’s driver-monitoring system.
[CV-342] Oracle Gaps in Reliability Coverag e: Sampling Noise or Policy Specialization?
链接: https://arxiv.org/abs/2609.32996
作者: Mert Onur Cakiroglu,Mehmet Dalkilic,Hasan Kurban
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Policies trained from the same base model can appear to solve different problems. An oracle that chooses the best policy for each problem may therefore appear much stronger than any single policy. Selecting the largest estimated success rate also selects favorable sampling errors. We study this effect through reliability coverage, the fraction of problems whose success probability reaches a chosen threshold. Our first test redistributes stored correctness outcomes across policies within each problem. A second also preserves each policy’s total successes, accounting for overall quality differences under a specified statistical model. For five training seeds of a seven-billion-parameter vision-language model, redistribution reproduces 0.096 of an estimated 0.113 oracle gap at threshold 0.10. Neither test finds significant evidence at this threshold. Small advantages remain unresolved. Mixtures, routers, voting, and weight averaging show no detectable improvement over their corresponding single-policy baselines. Training policies on different datasets shows little detectable specialization under light post-training, and no router gain. A stronger recipe does create it, both tests detect it, and a router gain appears only at the high thresholds where the specialists separate. In a control with predictable specialization, a router recovers about half the oracle gap. Coverage bounds explain why even a genuine oracle advantage need not yield a deployment gain. The tests assess apparent specialization from stored responses before investment in routing. Code: this https URL
[CV-343] ReVision3D: Attribution-Guided Recursive Self-Improvement for 3D Medical Perception
链接: https://arxiv.org/abs/2609.32984
作者: Ho Hin Lee,Yuyin Zhou,Yannan Yu,Shi Gu,Yifan Wu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 28 pages
Abstract:Recursive self-improvement (RSI) offers a promising path for overcoming the limited visual capability of current medical imaging agents. Yet applying RSI to volumetric imaging remains difficult: failures can arise from acquisition, perception, training recipe, or downstream inference, while self-generated feedback and logged trajectories provide little guidance on which component should change. We introduce ReVision3D, an RSI system that leverages 3D volumes with spatially grounded annotations to determine where visual evidence is lost and recursively improve the corresponding visual capability. A frozen language-model designer proposes revisions to acquisition, perception, training, or inference, while the verifier and system-level objective remain fixed. Our key insight is that an annotated volume forms an exact replay world for view rendering and spatial verification: unvisited views can be rendered on demand, and localized predictions can be checked directly against reference masks. This grounded feedback directs targeted revision, while only changes that improve beyond measured seed noise are retained. Each accepted change triggers renewed attribution, allowing the dominant bottleneck to shift across rounds. On abdominal CT, attribution identifies perception as the dominant remaining limitation. Revising that level enables ReVision3D to achieve 79% liver recall and 83% kidney recall at under 0.4 false positives per patient, outperforming the evaluated frozen multimodal foundation models, with the largest gains on small lesions.
[CV-344] Distributed Hydrological Modeling in the Feature Space
链接: https://arxiv.org/abs/2609.32971
作者: Mohamad Hakam Shams Eddin,Maria Luisa Taccari,Yikui Zhang,Shijie Jiang,Juergen Gall,Markus Reichstein
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Accurate forecasting of river discharge and floods is very challenging. River dynamics are affected by storage, meteorological forcing, and flow propagation at different spatial and temporal scales. Forecasting thus requires a framework that considers the upstream-to-downstream flow through river networks across grid cells and catchments. This modeling is known in hydrology as distributed modeling and routing. Existing deep learning approaches either ignore this topology, operate on lumped catchments, or route predicted physical quantities through a separate graph or physical routing model. We instead introduce feature-space routing: a topology-aware state-space operator embedded directly in the forecasting dynamics. At every forecast step, the operator gathers latent states from upstream grid cells and causally updates the downstream state according to the known river network. This preserves the physical connectivity of the river system while allowing the propagated state itself to be learned end-to-end and allows the model to predict river discharge considering both local dynamics and neighboring upstream contributions. To address uncertainty and provide probabilistic forecasts, we minimize the fair continuous ranked probability score (fCRPS) as a training objective. Our experiments on the European Flood Awareness System (EFAS) and observational data for river discharge forecasting demonstrate that encoding the physical structure of river networks explicitly in the feature space substantially improves the forecasting skill, particularly in an ungauged setting. Our approach achieves state-of-the-art results on both reanalysis and observational data and is able to forecast maps of river discharge at 1 arcminute and 6-hourly resolution up to 10 days lead time.
[CV-345] Synthetic Thermal Image Generation for Real-Time Animal Detection Under Low-Visibility Conditions
链接: https://arxiv.org/abs/2609.32944
作者: James Momoh,Khandaker Mamun Ahmed
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at The IEEE Cyber Awareness Research Symposium (CARS), 2026
Abstract:Wildlife-vehicle collisions remain a significant road safety concern, particularly during nighttime and low-visibility conditions when RGB-based perception systems are often unreliable. Thermal imaging offers a promising alternative for detecting animals under poor illumination. However, the limited availability of annotated infrared animal datasets restricts the development of robust deep learning-based detection models. This paper investigates synthetic thermal image generation as a scalable approach for real-time animal detection under low-visibility conditions. A subset of 514 annotated visible-spectrum animal images from the NTLNP dataset is translated into synthetic thermal representations using CycleGAN-Turbo, while a limited real thermal dataset of 60 images is expanded through thermal-focused augmentation. Multiple object detection architectures, including YOLOv8, YOLOv9, YOLOv10, and RT-DETR, are trained independently on synthetic and real thermal datasets and evaluated using precision, recall, mAP@0.5, mAP@0.5:0.95, model size, and inference latency. Experimental results show that synthetic thermal images provide competitive detection performance, with RT-DETR achieving the highest synthetic-data mAP@0.5 of 0.9613. Models trained on augmented real thermal data achieve the strongest overall performance, with YOLOv10s obtaining 0.9879 mAP@0.5 and 0.9571 mAP@0.5:0.95. Computational analysis further indicates that lightweight YOLO variants provide favorable inference latency, supporting their potential for real-time deployment. These findings demonstrate that synthetic thermal imagery can reduce dependence on scarce infrared datasets and support the development of efficient animal detection systems for future vehicle-mounted wildlife collision mitigation applications.
[CV-346] Rethinking the Fully Hyperbolic Vision Transformer in Polar Coordinates
链接: https://arxiv.org/abs/2609.32899
作者: Ahmad Bdeir,Niels Landwehr
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Hyperbolic space can embed intrinsic hierarchies in data with low distortion due to the exponential growth of volume with distance from the origin. However, current Lorentz transformer blocks are formulated in ambient coordinates, where numerical errors increase at large radii due to instability in the Lorentzian inner product, leading many models to limit the radius to avoid this issue. This prevents us from utilizing the regions of hyperbolic space that motivate the geometry. To address this, we revisit the components of the transformer block in polar coordinates, where hyperbolic operations such as distance calculation and attention centroids can be computed without the numerical cancellation in their ambient-coordinate formulations. Specifically, we propose a polar fully connected layer that separately maps an embedding’s direction and radius, allowing the radius of a feature to be learned rather than determined by the norm of a linear map. We additionally introduce horospherical shifts as relative positional encodings whose query-key distances grow logarithmically with the token gap. Finally, we reformulate the residual connection as average radius Lorentz boosts. Combining these components, we develop a fully hyperbolic transformer that substantially improves performance over Euclidean and hyperbolic baselines on standard vision tasks. We further evaluate our model on ImageNet and demonstrate its ability to generalize to other datasets using pre-trained weights, similar to Euclidean counterparts.
[CV-347] Improving Video Sparse Attention with Fine-grained Router and Sparse Rebasing
链接: https://arxiv.org/abs/2609.32882
作者: Peiyuan Zhang,Guoqiang Wei,Yilong Zhao,Zixiang Zhang,Wei Zhou,Will Lin,Heng Zhang,Xiaonan Nie,Yan Zeng,Hao Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We present VSA2, a frontier trainable sparse attention for video DiTs. VSA2 includes a variety of new architectural features and training procedures that we apply across all stages of the DiT development cycle, including pretraining, RL, and inference, to produce a DiT with comparable or better quality than a full attention counterpart. Architecturally, VSA2 introduces a fine-grained router that improves the precision of identifying critical tokens and supports dynamic computation by allowing each query to attend to a variable number of key-value pairs. In training, we identify a Hard-to-Easy Curriculum, where models trained under high sparsity and later evaluated with lower sparsity during inference not only generalize effectively, but also outperform models trained with full attention in motion quality. VSA2 is also flexible: it can replace full attention during the middle of progressive low-to-high resolution pretraining, rebasing early-stage full-attention checkpoints. Experiments show that VSA2 reduces attention computation by half over VSA with lower loss. On 720p videos, it accelerates attention by 8.9x and end-to-end generation by 4.62x compared to the FlashAttention-3 baseline, while achieving comparable or better video quality.
[CV-348] SV2V-RSim: A Comprehensive Benchmark for Self-Selective V2V Cooperative Perception with Near-Realistic Data
链接: https://arxiv.org/abs/2609.32863
作者: Yulu Wu,Chao Wei,Jujun Cheng,Zhangkai Ni,Haowen Wang,Dengyang Suo,Cong Chen,Xinyi Liu,Shangce Gao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vehicle-to-Vehicle (V2V) cooperative perception enhances autonomous driving by enabling vehicles to share information beyond their direct line of sight. However, existing V2V datasets are limited by a small number of participating agents, static collaborator selection strategies, and a significant domain gap between simulated and real-world environments. To overcome these challenges, we introduce SV2V-RSim, a large-scale, multi-modal, near-realistic simulation dataset engineered to elevate agent diversity and realism. Additionally, we present the Select Vehicles Adaptively (SVA) module, which optimizes collaborator selection to balance perception performance against communication bandwidth constraints. Our dataset is generated using the Unreal Engine 5-based simulator that integrates high-fidelity 3D assets, diverse environments, and intricate traffic scenarios. All vehicles within a specified range of the ego vehicle are equipped with sensor suites, enabling dynamic and adaptive collaborator selection. SV2V-RSim encompasses four maps, four weather conditions, six time periods from sunrise to night, 203K LiDAR frames, 402K RGB frames, and 788K annotated 3D bounding boxes across 17 object classes, supporting a range of cooperative perception tasks such as 3D object detection, segmentation, and depth estimation. Benchmarking on recent cooperative perception algorithms demonstrates that SVA achieves a superior performance-bandwidth trade-off, while sim-to-real experiments and No-Reference Image Quality Assessment validate the dataset’s high realism and practical effectiveness. Our dataset and code will be publicly available.
[CV-349] Is HE Image-to-Spatial Transcriptomics Simpler Than It Looks?
链接: https://arxiv.org/abs/2609.32857
作者: Duc T. Nguyen,Thanh Ha Do,Phuong M. Cao,Hieu Pham
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Predicting spatial gene expression from routine HE histology offers a scalable route toward spatial molecular profiling. Recent work has pursued increasingly sophisticated architectures to capture spatial context and richer expression structure. At the same time, simple estimators have shown strong performance in several studies, but what they already solve and where additional complexity is needed remain unclear. We study this behavior through the structure of prediction error under the mean-squared error (MSE) objective. Differences in average expression across genes can account for a substantial part of aggregate prediction performance, while a key unresolved error lies in recovering variation within each slide. Decomposing MSE into slide-level and within-slide components, we find that the within-slide component has lower residual-normalized parameter sensitivity in controlled neural experiments. This motivates Component-Guided Loss (CGL), which increases supervision of the within-slide component. CGL-Linear is a closed-form affine instantiation that achieves overall state-of-the-art performance across HEST-1k cohorts and gene-panel sizes. The same within-slide supervision improves existing neural models. These results suggest that substantial gains can come from aligning the training objective with prediction-error structure rather than increasing model complexity.
[CV-350] PolyTopoBench: A Benchmark for Complex Vector Polygon Generation from Remote Sensing Imagery NEURIPS2026
链接: https://arxiv.org/abs/2609.32856
作者: Zeping Liu,Ni Lao,Weiwei Sun,Gil Wolff,Yiqun Xie,Liang Zhao,Junfeng Jiao,Gengchen Mai
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted by NeurIPS 2026 (Evaluations and Datasets Track)
Abstract:Vector polygon generation converts visual inputs, e.g., remote sensing (RS) images, into vectorized polygonal geometries, supporting applications such as autonomous driving, vector map construction, and remote sensing. Early pipelines predict raster masks and post-process them into polygons, which prevents end-to-end optimization and may miss small objects or introduce inaccurate vertices. Recent methods directly generate vector polygons, but most focus on simple exterior contours, while they either cannot represent complex polygons with holes or fail to preserve their topology. In this paper, we propose PolyTopoBench, a unified evaluation framework for vector polygon generation from RS images with explicit emphasis on complex polygons. PolyTopoBench evaluates both exterior and interior rings, and benchmarks 11 representative methods, including segmentation-based polygonization pipelines, vision foundation model baselines, and specialized vector polygon generators, on two RS-image datasets covering buildings, roads, vegetation, and unvegetated regions. Experiments show that existing methods often recover simple exterior boundaries but degrade substantially on polygons with holes or multiple rings. These results reveal complex polygon generation as an unresolved challenge and motivate topology-aware benchmarks and model designs. Code and data are available at this https URL.
[CV-351] FINE: Future-Informed Navigation Encoding for Data-Efficient Vision-Language Navigation
链接: https://arxiv.org/abs/2609.32855
作者: Khang H. Nguyen,Hoang Pham Quang Nguyen,Ha Phuong Nguyen,Khanh Dinh Binh,Xuan Ha Nguyen,Vien Ngo,Duy Ho Nguyen Minh,Huan Nguyen,An T. Le
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 8 pages, 3 figures
Abstract:Adapting vision-language navigation (VLN) policies to new environments is expensive because every additional route and instruction requires an embodied demonstration. Yet standard observation-to-action training uses only a small fraction of the information already contained in each trajectory. In particular, future observations reveal the instruction-relevant landmarks that the agent will encounter, including what they look like and how they are arranged in 3D. We introduce FINE, a Future-Informed Navigation Encoding framework that extracts this latent supervision from existing demonstrations. FINE equips a VLN backbone with two complementary auxiliary representations. First, explicit landmark tokens follow the ordered landmarks specified by the instruction and are trained to predict future landmark regions in both semantic 2D patch-feature space and viewpoint-dependent 3D geometric feature space. Second, an implicit future token learns to distinguish the landmark state that is actually reached from plausible same-scene counterfactual futures generated by a video world model. On R2R-CE and RxR-CE val-unseen, FINE improves InternVLA-N1 by 2.6 and 4.5 success-rate points, respectively, at full training data. More importantly, as demonstrations become limited, the benefit grows: at a 70% demonstration budget, FINE improves success rate by 6.8 points, recovering roughly one-third of the performance lost by reducing the training demonstrations. Project page is available at this https URL.
[CV-352] SynCo: Learning Cross-Modal Synergy by Contrasting Interaction Residuals
链接: https://arxiv.org/abs/2609.32846
作者: Yavuz Yarici,Ghassan AlRegib
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Multimodal contrastive learning is a dominant paradigm for learning transferable representations from unlabeled data, but standard objectives primarily capture information that is redundant between modalities. Partial Information Decomposition (PID) shows that task-relevant information in multimodal data decomposes into three components: redundancy shared between modalities, uniqueness specific to each modality, and synergy available only from their joint observation. Recent frameworks extend contrastive learning to capture all three components, yet synergy remains undertrained in practice. We propose SynCo (Synergy Contrastive Learning), a method that directly addresses synergy undertraining through dedicated supervision on an interaction residual. SynCo fits a linear projector to predict the fused representation from independently computed unimodal features, and the resulting interaction residual, which removes the linearly unimodal-predictable component, receives dedicated contrastive supervision at negligible computational cost. On the controlled Trifeature benchmark, SynCo achieves state-of-the-art synergy capture with a +5.98% gain over the baseline, and on real-world benchmarks from MultiBench, DARai, and MM-IMDb, SynCo consistently outperforms or matches prior methods across diverse modality combinations and task types. The method operates as a plug-in to existing contrastive multimodal frameworks without modifying the underlying fusion architecture and can further improve synergy capture when combined with other methods.
[CV-353] Beyond Temporal Smoothing: Spatial Energy Budgets Stabilize One-Step Diffusion Editing
链接: https://arxiv.org/abs/2609.32841
作者: Shengxiao Zhou,Lei Luo,Jian Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:One-step text-guided diffusion editing is efficient but prone to spatially misallocated updates that distort the edited object and alter the background. Existing methods often improve stability by averaging the editing field across timesteps. We instead identify spatial energy misallocation as a distinct and measurable failure mode: across two independent noise draws, the residual field is essentially unrepeatable, making the background field unreliable for direct transport, while its total energy still sets a usable magnitude for the draw at hand. BudEdit turns that magnitude into an explicit budget and reallocates it to edit-relevant regions selected jointly by residual energy and cross-attention, controlling where editing energy is spent rather than averaging over timesteps. The resulting training-free, inversion-free editor spends the budget on transport and reuses it to scale a correction in a lower-noise gated refinement. The budgeted injection field matches its prescribed budget exactly and vanishes on the identified background support, by construction. On PIE-Bench with SD-Turbo, BudEdit outperforms ChordEdit under each method’s reported default settings on all 11 evaluated metrics, including a 2.1 ,dB gain in background PSNR, 31 % lower DINO, and 36 % lower LPIPS, while improving all five editing-quality metrics and reporting the lowest runtime in the comparison.
[CV-354] VCRE-Fib: View-Conditioned Regional Evidence for Fine-Grained Ultrasound Grading of Schistosoma japonicum-Associated Liver Fibrosis ICLR2027
链接: https://arxiv.org/abs/2609.32840
作者: Ziyang Xu,Shuli An,Hao Zhou,Haitian Zhong,Tingting Wu,Tao Wang,Kun Yang,Tieyong Zeng
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 20 pages, 5 figures, including appendices. Submitted to ICLR 2027. Code: this https URL
Abstract:Accurate assessment of Schistosoma japonicum-associated liver fibrosis is essential for disease management and long-term follow-up in endemic regions. Ultrasound provides non-invasive imaging, but complex local echogenic patterns and anatomical structures make fine-grained grading challenging. Existing deep learning methods can predict fibrosis scores, yet directly incorporating acquisition views and regional cues into grading while retaining spatial information for inspection remains an open problem. Here we present VCRE-Fib, a view-conditioned regional evidence framework that integrates anatomical context, local information, and global image assessment for fine-grained ultrasound grading. The framework forms view-conditioned local grading evidence before spatial pooling, uses weak localization to guide its aggregation, and combines it with global predictions. Image-only inference jointly returns a fibrosis score, acquisition view, and candidate abnormal-region map. We developed and evaluated the method on a re-curated cohort of 108,709 ultrasound images from 6,373 patients across 35 centers. On a patient-disjoint test set of 4,107 images from 240 patients across four centers, VCRE-Fib reduced the prespecified composite grading risk by 7.115% relative to SFibAI trained and evaluated on the same data split. Image-level mean absolute error decreased from 0.391 to 0.378, alongside lower patient-max, patient-median, and center-balanced risks. The full model also achieved lower composite grading risk than variants that separately removed view conditioning or weak localization. These results support incorporating anatomical context and regional evidence into ultrasound grading while exposing spatial predictions for inspection alongside severity estimates.
[CV-355] Unlocking Geodesic Gromov-Wasserstein Distances for 3D Modeling
链接: https://arxiv.org/abs/2609.32824
作者: Krzysztof Marcin Choromanski,Derek Long,Ananya Parashar,Dwaipayan Saha
类目: Computer Vision and Pattern Recognition (cs.CV); Data Structures and Algorithms (cs.DS); Image and Video Processing (eess.IV)
备注:
Abstract:\textitGromov-Wasserstein Distances (GWDs) provide quantitative ways of comparing probabilistic distributions defined on different metric spaces by applying techniques from the optimal transport theory. As such, GWD can be potentially useful in a large variety of applications ranging from graph matching problems to 3D object detection. However its practical use at scale is significantly limited by cubic time complexity computations involving dense intra-space distance matrices. Even though in the Euclidean metric spaces several techniques (e.g. involving scalable kernel methods) were proposed to address it, to the best of our knowledge, analogous techniques for general geodesic distances on manifolds, or shortest-path distance on graphs in their discretized variants, were not developed. In this paper, we present \textbfEfficient \textbfGeodesic \textbfGromov-\textbfWasserstein methods (EGGroW), a new class of efficient algorithms designed to calculate geodesic Gromov-Wasserstein distances with entropic Sinkhorn-like approaches, leveraging recently introduced \textitGenusSink methods \citepgenussink and the theory of random features. We provide important downstream applications, namely: 3D pose estimation and 3D template detection. In the latter setting, we formulate a partial 3D template recovery as a staged problem: capacity-constrained scene selection is followed by semi-relaxed recovery of template visibility and correspondence. Our empirical findings show that EGGroW provides accurate solutions when standard Euclidean-based techniques fail and is characterized by light computational footprint, as our theoretical analysis predicts.
[CV-356] Progressive Risk Estimation for Accident Anticipation NEURIPS2026
链接: https://arxiv.org/abs/2609.32811
作者: Samet Hicsonmez,Eray Çakar,Nermin Samet,Fatma Güney
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to NeurIPS 2026
Abstract:Accident anticipation aims to recognize anomalous driving cues before a crash while avoiding false alarms during normal driving. Existing approaches typically formulate this task as binary classification, focusing on whether an accident will occur rather than when it will occur. We propose PRE-ACT, a framework that models accident risk as a continuously evolving signal that increases as the crash approaches. By explicitly enforcing temporal ordering and distance-to-accident awareness, our method progressively raises risk while suppressing premature alarms, leading to significant improvements on MM-AU subsets and Nexar. We further introduce a Separation Score to evaluate the global behavior of predicted risk curves beyond local temporal windows. Code and visualizations are available at this https URL.
[CV-357] PlanGuard: A Guardrail for Multi-Step Plan Safety in Embodied Agents
链接: https://arxiv.org/abs/2609.32801
作者: Junchi Chen,Changtao Miao,Yuxiao Xiang,Zhenchao Jin,Haojie Yuan,Qi Chu,Tao Gong,He Liu,Bo Zhang,Jiansheng Cai,Zhe Li,Nenghai Yu
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Embodied task planners may produce multi-step plans whose subtask dependencies and interactions with the environment create physical risks during execution. Yet existing safeguards overlook such compositional risks, as general-purpose guardrails focus on semantic harm and embodied safety detectors assess subtasks in isolation. To address this gap, we introduce PlanGuard, the first pre-execution detector that evaluates the physical safety of a complete multi-step plan in its current environment. For training and evaluation, we construct a Multi-Step Plan Safety (MSP-Safe) dataset through paired task construction, plan generation using diverse planners, and safety annotation by three judges. Task-oriented SFT on MSP-Safe establishes fundamental plan-safety assessment capabilities, yet a substantial gap remains between compact models suitable for real-time deployment and stronger but costlier large models. Accordingly, we propose Strong-Teacher Adaptive Compensation for On-Policy Distillation (STAC-OPD), which provides compact models with adaptive strong-teacher supervision along their on-policy trajectories. It combines token-level distribution transfer from a fine-tuned strong teacher with probability-routed sequence-level compensation, retaining student-generated targets when the student favors the reference safety decision and using teacher-reconstructed targets otherwise. Across all test subsets, PlanGuard-2B achieves average 87.15% ACC and 87.21% F1, demonstrating effective whole-plan physical-risk detection at compact model scale. Code and dataset will be publicly released.
[CV-358] Latent Space Is Not Flat: Rethinking Latent Structure for 3D Medical Image Synthesis
链接: https://arxiv.org/abs/2609.32794
作者: Haowen Xue,Hao Chen,Hexuan Hu,Qian Huang,Yi Han,Qing Meng,Zaipeng Xie,Chao Li,Haoli Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 4 figures, 4 tables
Abstract:Latent generative models make 3D medical image synthesis computationally practical by generating in a compressed space. However, we show that the common flat Euclidean assumption induced by \ell_2 objectives is imprecise: latent-space geometry is so strongly anisotropic that equal-magnitude errors can produce drastically different decoded distortions. We further find that this anisotropy has a clear feature: sensitive variation concentrates in a low-rank subspace. The dominant low-rank components capture the overall structure, encoding long-range, spatially coordinated variation while remaining resistant to local noise. Its orthogonal residual, in contrast, mainly captures local and image-specific variation. Motivated by this asymmetry, we introduce Latent Structure Flow (LSF). At each block, LSF decomposes the latent state into structure and residual, models structural changes with global context, and predicts residual variation locally while preserving a direct path for the input structure. LSF changes only the generator, leaving the frozen codec and pointwise training objective unchanged. Across cross-modality synthesis and tumor inpainting tasks, LSF outperforms all compared baselines on both global and tumor-specific metrics, demonstrating the benefit of explicitly modeling latent-space structure for 3D medical image synthesis.
[CV-359] Hierarchical Frequency-Domain Compression of Implicit Geometric Representations for Large-Scale Point Clouds
链接: https://arxiv.org/abs/2609.32789
作者: Manlin Yao,Jiabin Liu,Guan Wang,Haixu Liu,Hui Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Large-scale point cloud representations of complex geome tries incur prohibitive computational and memory costs, necessitating compressed implicit representations. To ad dress this, we propose a unified framework comprising im plicit geometric field representation, hierarchical frequency domain compression, and conditional high-frequency predic tion. Specifically, an unordered point cloud is mapped to an implicit field defined within its physical bounding box. A smooth Fourier pyramid is then constructed, where com pact low-frequency components capture the global geometry. Inter-scale high-frequency residuals are encoded to preserve the spatial information required for reconstructing fine geo metric details. To restore the high-frequency information lost during compression, we develop a hierarchical 3D neural net work. The reconstructed implicit field is converted back into a point cloud through isosurface extraction. Experiments on a complex-boundary point cloud with more than eight mil lion points demonstrate that the proposed method achieves a higher compression ratio than existing point cloud compres sion methods while maintaining comparable reconstruction quality.
[CV-360] Learning When to Recur: Token-Adaptive Recursion for Imbalanced Ophthalmic Domain Incremental Learning
链接: https://arxiv.org/abs/2609.32785
作者: Nanxi Yu,Kang Li,Ye Du,Xiaowei Hu,Weihua Yang,Shujun Wang
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 14 pages, 8 figures, 10 tables
Abstract:Domain incremental learning is essential for adapting ophthalmic deep learning models to sequential clinical domains while preserving diagnostic expertise. Existing domain incremental learning methods predominantly address the domain shift induced by style variations. However, they often overlook the severe class imbalance inherent in real-world clinical scenarios, such as clinical referral systems. Institutions in these systems encounter drastic fluctuations in class priors, resulting in label distribution shift, a critical form of domain shift that triggers severe catastrophic forgetting. To address these challenges, we propose ToRe, a rehearsal-free and parameter-efficient framework that leverages frozen ophthalmic foundation models for robust incremental adaptation. ToRe employs a parameter isolation strategy to decouple domain-specific optimization paths, thereby helping mitigate catastrophic forgetting driven by both label distribution shift and style variations. Simultaneously, it introduces token-adaptive recursion that adaptively allocates additional computational depth across tokens, allowing simple tokens to exit the recursion loop early while subjecting complex tokens, such as those associated with lesions, to deeper recursive processing. This mechanism enhances the feature representations for minority classes, thereby supporting generalization throughout the domain incremental learning process. Extensive evaluations on nine heterogeneous datasets demonstrate that ToRe consistently outperforms state-of-the-art methods in overall performance across the three benchmarks, while maintaining near-zero forgetting. Together, these results support the applicability of ToRe to dynamic and imbalanced clinical environments. The code is available at this https URL
[CV-361] CT-OPD: Counterfactual Trace On-Policy Distillation for Diffusion Vision-Language Models
链接: https://arxiv.org/abs/2609.32781
作者: Long Qian,Bingke Zhu,Jiaqi Wei,Yu Li,Yingying Chen,Jinqiao Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Diffusion vision-language models generate answers by gradually resolving masked tokens, making accurate conditional prediction in partially resolved states central to post-training. Masking completed answers yields coherent contexts and targets, but prescribed masks do not reflect the model’s reveal decisions. Its trajectories capture these decisions, yet their provisional visible tokens can conflict with the target response. Outcome-based reinforcement learning follows these trajectories but provides only response-level feedback, which loses contrast when sampled rewards tie. To align coherent token-level supervision with the model’s reveal decisions, we introduce Counterfactual Trace On-Policy Distillation (CT-OPD), which combines completed teacher responses with trajectory masks from the current student. CT-OPD retokenizes each teacher response in the student’s vocabulary and extracts unresolved-position masks at successive stages of the student’s reverse process. For each mask, it discards provisional rollout values and reconstructs the partial state from the teacher endpoint, so the supervised positions follow the current trajectory while the visible context and targets remain consistent with the same response. The student is trained on these reconstructed states with its native categorical loss, and trajectories are refreshed as the model evolves. Across dense and sparse diffusion architectures, CT-OPD consistently enhances multimodal understanding and reasoning capabilities, with gains of up to 9.80 points on the nine-benchmark average. On the unified understanding-and-generation architecture, it also improves both visual understanding and image generation, showing that the same principle transfers across architectures and modalities. Ablations further attribute these gains to coherent reconstruction and current-model trajectory masks.
[CV-362] OmniMoE-VL: A Sparse Vision-Language Model with Coupled Visual-Depth Routing
链接: https://arxiv.org/abs/2609.32780
作者: Long Qian,Bingke Zhu,Jiaqi Wei,Yingying Chen,Jinqiao Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Vision-language models (VLMs) increasingly use sparse mixture-of-experts (MoE) to scale language-side computation, yet visual information is typically routed only after passing through a fixed cross-modal interface. This leaves an important decision unresolved: which intermediate visual representations should be exposed to language computation for a given question? We introduce OmniMoE-VL, a sparse VLM with a coupled visual-depth routed projector. For each image-prompt pair, the projector selects a sparse set of intermediate visual depths and reuses the resulting global preference to guide both local patch fusion and dynamic visual injection into the language model. This design enables question-dependent visual access while preserving the native visual-token sequence, and complements token-level expert routing in the vision and language stacks. Across eight image-based benchmarks, OmniMoE-VL achieves an average score of 85.9 with 28B total and 9B activated parameters. Controlled comparisons show that the routed visual interface provides the dominant architectural gain, while matched route and component controls, same-image route analysis, and route interventions further support the value of coupling and question-conditioned visual access.
[CV-363] An Empirical Study on What Matters for Viewpoint-Generalizable Policies in Visual Imitation Learning
链接: https://arxiv.org/abs/2609.32762
作者: Mino Nakura,Sriram Krishna,Yufei Wang,Shubham Tulsiani,Zackory Erickson,David Held
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Visual imitation learning is a promising approach to training robot manipulation policies capable of completing a wide variety of tasks. However, policies today remain brittle to viewpoint perturbations, making deployment in diverse environments a challenge. We present a controlled empirical study of which design choices allow visuomotor policies to generalize across viewpoints. We find that viewpoint generalization improves when dense visual tokens are retained and the action head participates in geometric reasoning. On a suite of simulated tasks that span a wide range of camera poses, we show that these design choices yield a policy that remains performant across viewpoints. As a practical consequence, a policy trained with these design choices also transfers zero-shot from simulation to the real world under random camera configurations.
[CV-364] From Feed-Forward to Flow: Unifying Reconstruction and Generation Is Easier Than You Think
链接: https://arxiv.org/abs/2609.32761
作者: Haoru Wang,Qianfan Shen,Kai Ye,Wenzheng Chen,Baoquan Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 34 pages, including supplementary material. Haoru Wang and Qianfan Shen contributed equally
Abstract:Reconstruct where the images provide evidence, and generate where they do not: recent success of spatial world models such as Atlas (World Labs Team, 2026) highlights the value of unifying reconstruction and generation in one model. Yet the two have long lived in separate paradigms with distinctive failure modes: feed-forward reconstruction averages ambiguity into blur, while conditional generation invents plausible but scene-inconsistent detail. In this work, we present a unified flow-based formulation for reconstruction and generation, where a shared clean-target predictor performs direct reconstruction at its single-step endpoint and unfolds conditional generation through multi-step flow. A controlled toy study reveals the mechanism: with a single step, the predictor collapses to the conditional mean just like feed-forward methods, favoring consistency over diversity. With multi-step inference, the fidelity of generated details grows with context richness: closer observations reduce ambiguity and yield better-matched details. We further instantiate the formulation in appearance and geometry 3D tasks. JiT-LVSM improves perceptual and distributional quality in novel view synthesis, while JUSt3R retains competitive single-step geometry prediction with additional multi-step inference capabilities that reduces veil and flying-pixel artifacts, producing cleaner surface structure with greater test-time compute. Together, they show that reconstruction and generation can share both a formulation and a backbone, with their behavior governed by denoising configuration—making unification surprisingly simple.
[CV-365] Readout is not Recovery: Dissociating Coordinate Emission from Visual-Corruption Repair in Vision-Language Models
链接: https://arxiv.org/abs/2609.32757
作者: Drandreb Earl Juanico
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 29 pages (14 main + appendix), 2 figures, 13 tables
Abstract:VLM bounding-box localization is both language generation and spatial commitment. Parseable fields such as bbox_2d make localization easy to score, but dimensions that emit coordinate tokens need not repair localization after visual evidence is damaged. We study this readout/recovery separation in Qwen3-VL-4B-Instruct on single-object COCO grounding. We compare clean coordinate-token readout rankings with corruption-derived repair rankings, using object-mask endpoint replacement for recovery and clean-input flooring for depth localization. In Qwen3-VL, coordinate-token rankings are inert through layer 24, load-bearing from layers 32-35, and peak at layer 34; corruption-derived rankings harm layers 16-24 but become beneficial near layer 35/final. A Kimi-VL-A3B diagnostic shows a matching output-proximal transition despite a different box format. Object-mask recovery separates rank budgets: k=250 shows necessity, k=500 shows Top- k restoration above random, and k=d/2 is largely capacity-driven. Partial-occlusion sweeps reveal that high-overlap coordinate-token sets can hurt at k=1000 and help mainly at half-width, while population corruption-derived sets provide no reliable fixed repair set. Edge-attribution patching shows coordinate-token paths are high precision but low recall for detection recovery, and RMSNorm quasi-layer controls do not close the endpoint-repair gap. Endpoint coordinate triage is therefore a useful circuit prior, but occlusion recovery requires a separate benchmark.
[CV-366] LoCoVSR: Local Context Diffusion Posterior Sampling for Video Super-Resolution
链接: https://arxiv.org/abs/2609.32742
作者: Matan Ben Chorin,Michael Elad
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注:
Abstract:Video super-resolution (VSR) is an ill-posed inverse problem that aims to reconstruct a high-resolution (HR) video from a noisy, low-resolution (LR) version of it. We present LoCoVSR, a diffusion-based VSR framework that leverages pixel-space denoising diffusion probabilistic models. LoCoVSR integrates the Diffusion Posterior Sampling technique with spatio-temporal context learning, operating in a moving-average form. A localized window of adjacent LR frames is used for recovering each center frame, while applying a shared noise trajectory across all frames. The localized windowing enables processing of long videos without length limitations, supports parallel inference, and prevents error accumulation that may occur in recursive processing. Unlike prior methods, LoCoVSR offers a simple yet very effective VSR solution, avoiding explicit optical flow estimation, or information loss caused by latent space processing. Trained on the VFHQ face dataset, LoCoVSR achieves accurate, temporally consistent and high-quality upscaling with competitive results against recent diffusion-based VSR approaches.
[CV-367] AnesTRACE: Benchmarking Intraoperative Anesthesia from Multimodal Perception to Multi-step Decision-Making
链接: https://arxiv.org/abs/2609.32740
作者: Ziwei Huang,Qi Gao,Zhe Ji,Yuanyuan Yao,Fengjiang Zhang,Min Yan,Zhongle Xie,Gang Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Intraoperative anesthesia requires systems to interpret evolving multimodal evidence, recommend timely management, and revise decisions as patient states change, yet existing benchmarks usually isolate perception or single-point reasoning. We introduce AnesTRACE, an evaluation suite comprising AnesTRACE-Bench and AnesTRACE-Eval. Built from public perioperative datasets with anesthesiologist annotation, AnesTRACE-Bench evaluates Intraoperative Perception, Single-point Anesthesia Decision-Making, and Multi-step Anesthesia Decision-Making. AnesTRACE-Eval assesses open-ended responses through anesthesiologist-defined criteria for Clinical Correctness, Evidence Grounding, Task Completeness, and Safety, with Temporal Consistency for multi-step decisions; its domain-specific evaluator is trained by supervised fine-tuning and preference alignment on expert-reviewed judgments. Across more than 30 models, fine-grained visual grounding and intervention selection remain difficult: the leading model reaches only 32.2 mIoU for TEE visual grounding and retains a 17.5% Major/Critical Safety Error Rate in multi-step management. Evaluator alignment with anesthesiologists improves across both training stages, while the best decision quality is accompanied by a 74.3-second P95 Latency. These results show that aggregate performance alone does not establish safe, timely longitudinal decision-making. We release our code at this https URL.
[CV-368] REALIS: A Curated Dataset for Studying the Challenges of AI Image Detection
链接: https://arxiv.org/abs/2609.32734
作者: Aleksandr Gushchin,Khaled Abud,Georgii Bychkov,Ekaterina Shumitskaya,Artem Filippov,Sergey Lavrushkin,Dmitriy S. Vatolin,Anastasia Antsiferova
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:AI-generated image detectors are often evaluated on benchmarks where real and synthetic images differ in content, quality, or generation artifacts, allowing models to rely on dataset-specific cues and fail on unfamiliar generators or processed images. Existing datasets provide limited support for evaluating these challenges jointly across diverse visual content. We introduce REALIS, a dataset of 1.43 million real and synthetic images generated by 42 modern text-to-image models, including the latest proprietary systems such as Nano Banana 2. REALIS combines prompts derived from real images, quality filtering, and stratified sampling to reduce class-specific shortcuts while preserving content diversity. We further introduce REALIS-Expert, a stress-test subset for high-quality synthetic images, where real and generated samples are selected with closely matched semantic and visual characteristics. We also propose a robustness protocol covering 35 transformations at five severity levels to analyze detector behavior under image processing. Based on REALIS, our benchmark evaluates pretrained detectors, fine-tuned models, and zero-shot vision-language models under generator and post-processing shifts. On the hardest processed split, the best pretrained conventional detector achieves 0.550 ROC-AUC, compared with 0.752 for the best REALIS-trained detector. REALIS provides a unified framework for measuring and improving the reliability of AI-image detectors under conditions that better reflect real-world use.
[CV-369] Region-Local Copula Evidence Fusion for Heterogeneous Remote Sensing Change Detection
链接: https://arxiv.org/abs/2609.32716
作者: Zhiyuan Ji,Junjun Yin,Jian Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Signal Processing (eess.SP)
备注:
Abstract:Superpixel copula models provide stable regional evidence for heterogeneous remote sensing change detection, but a single label per region limits localization within mixed superpixels. This letter develops a region-local copula evidence fusion method that retains the regional decision structure while introducing spatially varying local dependence anomalies. Independently fitted local models characterize departures from unchanged cross-image relationships. Reference ranking and an upper-tail gate transform these anomalies for fusion with continuous regional confidence. We derive the resulting regiondependent local decision threshold and identify a condition under which gating is equivalent to reparameterizing ungated fusion. On Lake and UK, whole-image optimized configurations achieve kappa coefficients of 0.78136 and 0.90817 and improve mixedregion and boundary decisions. Four-fold retrospective spatial validation over ten training subsets confirms complementary local information, with ungated reference fusion increasing mean kappa by 0.00693 and 0.01793. Fixed gating yields a larger UK gain of 0.03353 but only 0.00041 on Lake. These results support regional-local dependence interaction, while showing that calibration and gating have scene-dependent benefits.
[CV-370] ProDyGS: Dynamic Gaussian Splatting from a Single Static Monocular Camera
链接: https://arxiv.org/abs/2609.32711
作者: Ugo Leone Cavalcanti,Fabio Tosi,Matteo Poggi,Andrea Conti,Vladimir Zlokolica,Valerio Cambareri,Stefano Mattoccia
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We present ProDyGS, a novel dynamic 3D Gaussian Splatting framework for high-quality novel view synthesis from videos captured by a single static camera. While existing methods rely on multi-view setups or significant camera motion for geometric constraints, our approach addresses the challenging scenario where multi-view supervision is completely absent. We overcome this limitation by generating synthetic multi-view supervision through depth-guided proxy image synthesis. Specifically, we estimate temporally consistent depth maps using foundational monocular depth networks, then construct 3D Gaussian representations that generate proxy images from arbitrary viewpoints. A deformation network learns temporal dynamics by warping canonical Gaussians using this augmented supervision. Experiments on the DyNeRF dataset demonstrate that our method achieves state-of-the-art performance while requiring only monocular depth estimation as external supervision, outperforming approaches that rely on stronger priors such as scene flow.
[CV-371] DPAMixerSR: An Efficient Degradation-Pattern-Aware Model for Image Super-Resolution
链接: https://arxiv.org/abs/2609.32705
作者: Song-Li Wu,Haonan Jiang,Jixuan Fan,Yufei Huo,Chubin Zhang,Yansong Tang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: PRCV2026
Abstract:While content-adaptive schemes have delivered notable advances in image super-resolution (SR), existing approaches typically focus on texture complexity and ignore intrinsic degradation factors (e.g., blur kernels or noise patterns), leading to suboptimal computation allocation and reconstruction performance. To remedy this, we propose DPAMixerSR, a degradation-pattern-aware framework that enables efficient SR through adaptive sparse computation. We design a lightweight Perceptual Degradation Ranking (PDR) module partitions the image into severely and mildly degraded patches, which are routed to the Adaptive Sparse Processing (ASP) and a lightweight convolutional branch, respectively. ASP performs structure-aligned, multi-scale sparse propagation and bidirectional refinement, while the convolutional branch enhances efficiency in mildly degraded regions. By coupling degradation-driven routing with structure-aligned sparse processing, DPAMixerSR establishes a self-regulating framework that dynamically balances computational efficiency and reconstruction fidelity. Extensive experiments on various SR tasks demonstrate that our DPAMixerSR achieves superior structural restoration and perceptual fidelity with markedly reduced computational overhead, providing a novel and scalable framework for degradation-aware, resource-efficient SR.
[CV-372] MM-OPD: Towards One More Bottleneck Between Perception and Reasoning
链接: https://arxiv.org/abs/2609.32690
作者: Jintao Tong,Yujing Lou,Zhanming Shen,Jiaqi Gu,Lubin Fan,Ruixuan Li,Yue Wu,Jieping Ye,Yixiong Zou
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project page: this https URL
Abstract:Recent multimodal large language models (MLLMs) advance visual reasoning by strengthening both perception and reasoning, implicitly assuming a process that transitions seamlessly from perception to reasoning. However, we observe a counterintuitive phenomenon that challenges this assumption: holding the model, question, and decoding fixed, we replace images with their caption or code representations (symbolic views), which seems to be redundant given the clear image structures, but the performance surprisingly improves by 10.2% to 23.6% across model scales and datasets. We term this performance gap as the Symbolic Visual Gap and then take a closer look at it. Through experiments, we find that although the visual evidence can already appear in the reasoning trace for the image-input model, the symbolic-view-input model shows much higher attention to the correct evidence than the image-input model. This suggests that despite good capabilities from current works in perception and reasoning themselves, another bottleneck exists between perception and reasoning in selecting perceived visual information as appropriate evidence for subsequent reasoning. To handle this bottleneck, since the symbolic view steers attention toward correct evidence and is readily obtained at scale, it provides supervision for evidence selection without manually labeled evidence. Building on this, we introduce MM-OPD, a multimodal on-policy self-distillation framework for symbolic-to-visual correction that transfers guidance from symbolic-conditioned behavior to the image-conditioned policy through residual token-level targets, steering the model toward correct visual evidence. Experiments across benchmarks and model scales show that MM-OPD improves a broad range of multimodal abilities, with gains in visual perception, chart and document understanding, mathematical reasoning, and general VQA.
[CV-373] RCVLA: 4D Radar-Grounded Semantic Reasoning and Trajectory Arbitration for Autonomous Driving
链接: https://arxiv.org/abs/2609.32681
作者: Lianqing Zheng,Xiaokai Bai,Yixuan Luo,Runwei Guan,Minghao Liu,Zhiqiang Wei,Hui-liang Shen,Xichan Zhu,Zhixiong Ma
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:4D radar provides geometric and motion cues that complement visual semantics, but integrating it into vision-language-action (VLA) models requires both radar–language alignment for semantic reasoning and explicit use of radar measurements for trajectory refinement and selection. To support these capabilities, we construct Cap4DR with 86,016 radar-image-text samples for alignment pretraining and OmniHD-QA with 520,161 question-answer pairs for instruction tuning across scene description, key-object reasoning, occupancy understanding, and trajectory planning. Building on these datasets, we propose RCVLA, a radar-camera VLA framework consisting of a radar-grounded semantic reasoning stage (RCVLA-Sem) and a trajectory arbitration stage (RCVLA-Phys). RCVLA-Sem performs gated bidirectional interaction between camera and radar tokens for driving question answering and reference trajectory generation, while auxiliary heads provide object and occupancy queries. RCVLA-Phys refines reference-guided trajectory candidates through truncated diffusion conditioned on these queries and cluster-level radar measurements, then calibrates candidate scores using radar-derived time-to-collision risk. On OmniHD-QA, RCVLA-Sem improves CIDEr by 9.92 points and reduces key-object velocity error by 21.9% relative to OmniDrive. RCVLA-Phys further reduces average L2 error from 0.348 to 0.259,\mathrmm and average open-loop collision rate from 0.576% to 0.175% relative to RCVLA-Sem. Ablation studies further show that language-aligned radar tokens improve semantic reasoning, while cluster-level radar measurements and risk calibration improve trajectory arbitration. Code will be released.
[CV-374] Refinement Symmetry in Multimodal Transformers
链接: https://arxiv.org/abs/2609.32669
作者: Yuhao Du,Shunian Chen
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Attention weights depend on token counts, which change with the representation of a signal. We study refinement symmetry: splitting a representation while preserving content, position, visible context, and total mass should preserve its contribution. Building on proportional and quadrature attention, we show that split invariance forces the local mass factor to be linear for any fixed positive attention kernel, provided that factor is nondecreasing. For changed representations, a physical coupling bounds attention error by separating feature change from weight reallocation. In Qwen2.5-Omni-7B, duplicating half the visual tokens threefold changes 255 of 3,586 MVBench answers under standard attention; measure weighting preserves every answer under matched visibility. Under natural frame resampling, it reduces distributional drift. At twofold merging of a frozen video encoding, a five-seed evaluation shows an all-partition-correct accuracy gain of 1.04 percentage points over global count weighting (average group mass) and 0.93 points over standard attention. The advantage over global count also holds on WorldSense but depends on the compression budget. The result is a representation principle with a measured benefit in robustness across partitions.
[CV-375] mestep Weighting: A Hidden Key to Effective ELBO-Based Flow-Matching RL
链接: https://arxiv.org/abs/2609.32665
作者: Qinwei Ma,Jingzhe Shi,Simin Fan,Ling Li,Mengdi Wang,Alex Lamb
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages for the main body
Abstract:ELBO-based reinforcement learning offers a sampler-agnostic approach to fine-tuning flow matching models with reward feedback. Timestep weighting in ELBO-based RL has large impact on performance, and it also provides a unified view (as we show in this work) to understand prediction losses heuristically chosen in prior work, yet it remains under-researched and is often chosen to inherit pretrain configs. We investigate impacts and dynamics of timestep weighting in ELBO-based RL. We show that effective weighting depends on both the reward landscape and stage of learning. (1) Through experiments on controlled CIFAR image generation, complemented by robotics, we investigate how weighting impacts reward-driven updates across noise levels. (2) Through gradient analysis, we reveal distinct patterns of cross-noise coordination across tasks and their evolution during training. These findings motivate the hypothesis that useful weighting depends on the gap between the policy’s current behavior and the behavior favored by the reward. (3) Guided by this analysis, we study simple static weighting, budgeted profile selection, and dynamic schedules that improve performance beyond conventional target choices. Our results establish timestep weighting as an important design choice for flow-matching RL and motivate further research into methods that choose and adapt it throughout learning.
[CV-376] InterTab: Interleaved Visual-Structure Alignment for Multi-Modal Table Reasoning
链接: https://arxiv.org/abs/2609.32660
作者: Hanqian Li,Sirui Huang,Chen Ling,Jungang Li,Yu Huang,Kening Zheng,Yonghua Hei,Xiangrong He,Shiyi Wang,Pengcheng Zhu,Dongnan Liu,Wei Zhou,Linjian Mo,Nai Ding,Xuming Hu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Table images preserve structural information that are often lost in text serialization, and reasoning over them requires locating relevant rows, columns, and cells step by step. Current multimodal large language models (MLLMs) encode the whole image once before reasoning, so they cannot pick up row-, column-, and cell-level evidence as the question unfolds. Encoder-side table structure and generic interleaved visual chain-of-thought still do not bind each reasoning step to that structure. We propose \textbfInterTab, an \textbfInterleaved structure-aware framework for CoT reasoning over \textbfTable images, interleaves chain-of-thought with tool calls that crop structure-aligned table regions. First, we build InterTab-22K, includes reasoning trajectories in which each step is tied to both a structural location and a bounding box. InterTab is trained in two stages: supervised structure-aware alignment (SSA) on InterTab-22K teaches the model to interleave reasoning with structure-aligned crops, and active localization optimization (ALO) further optimizes answer correctness, localization IoU, and output format, while penalizing missing or excessive tool calls. Experiments on nine table benchmarks show that InterTab improves the average accuracy of its backbone from 68.28% to 73.17% and achieves the best average performance among all compared methods. Code and data will be released soon.
[CV-377] DraftAttention2: Fast Video Diffusion with Low-Resolution-Guided Mixed-Precision Attention
链接: https://arxiv.org/abs/2609.32628
作者: Rui Ding,Haopeng Li,Weize Ma,Yufa Zhou,Yitong Li,Xiaoling Zhou,Jiashuo Cao,Liyang Li,Hua Geng,Jiuxiang Gu,Jun Lin,Enze Xie,Xuan Shen
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Preprint Version
Abstract:Video generation has broad applications in content creation and entertainment. Diffusion transformers have advanced the quality of generated videos, but attention over spatiotemporal tokens becomes increasingly expensive as video resolution and duration increase. We present DraftAttention2, a training-free framework that uses the low-resolution draft attention map to jointly select attention blocks and assign their numerical precision. Specifically, spatial 2D average- and max-pooled queries and keys capture complementary regional statistics to estimate block importance, and a shared ranking assigns higher precision to important blocks, lower precision to less important retained blocks, and skips the rest under configurable budgets. Our analysis separates sparsification error from attention-weighted quantization error, establishing when recovering skipped interactions with low-bit computation tightens the output-error bound. This analysis motivates retaining more interactions at low precision while reserving higher precision for blocks with larger attention mass. To translate these fine-grained assignments into practical speedups, we further develop fused operand preparation and a single attention kernel with precision-specific phases, sharing data movement, softmax statistics, and output accumulation across precisions. Experiments demonstrate that our method achieves a superior quality-efficiency trade-off over existing efficient video generation methods. Notably, its advantage is particularly pronounced for few-step video diffusion, where jointly combining sparsity with 4- and 8-bit mixed-precision computation substantially improves generation quality while retaining significant acceleration. Code is available at this https URL
[CV-378] World SLAM Model: Joint World Modeling for SLAM and Navigation
链接: https://arxiv.org/abs/2609.32626
作者: Minghui Qin,Yijun Yuan,Weicheng Zheng,Kenan Li,Weibang Wang,Chang Sun,Junhao Huang,Anmin Liu,Yicheng Yao,Hang Zhao
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: Website: this https URL
Abstract:We introduce World SLAM Model (WSM), a unified framework that brings the SLAM paradigm directly into downstream navigation. Rather than treating SLAM merely as an upstream module that provides poses, maps or tokens, WSM adopts its core mechanisms, including incremental state updates with persistent memory and backend refinement of accumulated errors, to maintain a consistent world state during interaction. Given the current observation and a navigation goal, WSM predicts future visual states and jointly estimates their camera motion and dense geometry, grounding visual prediction in an evolving spatial world state. This spatial state is continuously updated as new observations arrive and provides the basis for action generation and closed-loop navigation. WSM is trained end-to-end with a joint navigation–SLAM objective, enabling downstream navigation to benefit directly from SLAM-style state maintenance and refinement while preserving accurate geometric estimation. Experiments demonstrate improved navigation performance together with strong SLAM accuracy, highlighting the potential of SLAM as an intrinsic mechanism for long-horizon world modeling and embodied interaction.
[CV-379] Levy-Driven Correspondence Estimation for Registration
链接: https://arxiv.org/abs/2609.32612
作者: Qianliang Wu,Jiaqi Yang,Wankou Yang,Le Hui,Jin Xie,Jian Yang,Yaqing Ding
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Finding reliable point correspondences is difficult when point clouds have low overlap or undergo non-rigid deformation. Iterative refinement can correct uncertain matches, but costly network evaluations limit the number of updates. We present LevyMatch, a Lévy-driven method that uses random jumps to refine a soft matching matrix. At each step, a network uses the current matching state and geometric information to predict a target matching matrix. A Brownian reference bridge gives an explicit formula for the update toward this target. A Gamma random clock sets the time step for each update. The updated matches provide new geometric feedback for the next target prediction. We further propose a fixed front-loaded Gamma policy that assigns more expected clock time to early updates and less to later ones, without retraining or extra network evaluations. Reordering the same sampled Gamma increments shows that placing larger increments early gives higher accuracy than placing them late. On 4DMatch and 4DLoMatch, our method improves both non-rigid feature matching recall (NFMR) and inlier ratio (IR) over the compared methods. The front-loaded policy achieves 93.09% NFMR and 92.11% IR on 4DMatch, and 82.79% NFMR and 79.07% IR on 4DLoMatch.
[CV-380] GAUGE: Group-Wise View-Inconsistency Rectification for Feed-Forward 4D Tracking
链接: https://arxiv.org/abs/2609.32596
作者: Zhuoqian Feng,Weixing Chen,Ziliang Chen,Yang Liu,Liang Lin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 9 pages of main text plus appendices. Code: this https URL
Abstract:Feed-forward models regress dense 3D point trajectories directly from monocular video, yet the residual after global alignment is substantial and lacks a structural explanation. Measured on dynamic query points across models and datasets, the error concentrates along the view direction, while the scale correction each motion group requires differs. The predicted displacement direction nevertheless supports reliable grouping, with a median angle far below the 90° random baseline. The systematic part of the residual is therefore a family of radial degrees of freedom per motion group, along directions 2D observations cannot constrain. We call it group-wise view inconsistency. We present GAUGE (Group-wise Adaptive Unsupervised Gauge Estimation), a training-free and model-agnostic post-hoc module. It recovers motion groups from direction consistency and spatial connectivity, then estimates a per-frame radial scale and group-level translation from 1% to 5% metric anchors, four degrees of freedom per group and frame. On dynamic query points of eight trackers, including D4RT, 4RC and SM4RT, our correction lowers endpoint error by 15.1% to 62.6% over the uncorrected predictions, while spending the same anchors on gradient fine-tuning improves the same models by only -1.1% to 15.2%. Code is publicly available at this https URL.
[CV-381] SPACE: Sparse Predictive Attractor via Counterfactual Eviction for Streaming Video Memory
链接: https://arxiv.org/abs/2609.32592
作者: Hongjin Niu,Weizhan Zhang,Shuo Bao,Jiahao Wang,Muyan Jiao,Kairui Wen,Yong-Jin Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Fixed-capacity streaming video memory requires repeated eviction decisions whose effects accumulate over time. Yet existing policies are evaluated primarily in terms of retained information or downstream accuracy, leaving how repeated updates alter the futures supported by memory largely unexamined. We define a memory’s predictive state as the future representations supported by its retained history and formulate eviction as counterfactual control over transitions in this space. We introduce SPACE (Sparse Predictive Attractor via Counterfactual Eviction), which uses a frozen multi-horizon JEPA to predict the future representations induced by alternative eviction actions. Counterfactual utility identifies future-useful alternatives, while slow predictive-basin geometry determines when to correct avoidable drift and when to adapt to sustained predictive change, without online parameter updates. We further introduce MABS-Bench, which evaluates future-task sufficiency, within-regime predictive stability, transition responsiveness, and perturbation recovery under matched causal streams and memory budgets. Across multiple video datasets, SPACE yields consistent improvements in dataset-native task performance while reducing predictive-state drift.
[CV-382] OmniSmartHome: A Multimodal Reasoning Benchmark for Smart-Home Agents
链接: https://arxiv.org/abs/2609.32569
作者: Jihoo Jung,Suho Yoo,Jeongsoo Choi,Hyebin Cho,Tae Wook Haam,Hyeonggon Ryu,Sumin Park,Joon Son Chung
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint
Abstract:Smart-home assistants are expected to handle diverse, realistic requests that arise in daily life. In such interactions, users often rely on the surrounding multimodal context-pointing at objects or referring to what they see or hear, leaving their requests underspecified in language alone. Existing smart-home benchmarks, however, express user requests solely through language, leaving context-dependent real-world requests underexplored. To bridge this gap, we introduce OmniSmartHome, a multimodal smart-home benchmark where each spoken request is paired with the surrounding visual and spatial-audio context, providing complementary cues to disambiguate underspecified requests. OmniSmartHome comprises 1,360 synthetic and 272 real-world episodes. We evaluate 16 omnimodal large language models (Omni-LLMs) and reveal that, while they perform strongly when speech alone sufficiently conveys the user’s intent, performance drops substantially when resolving it requires reasoning over multimodal contextual cues. As a simple agent baseline, we provide PROME (PROcedural Memory for multimodal Evidence gathering), which equips agents with specialized audio-visual perception tools and procedural memory for orchestrating their use. PROME generally improves performance across six Omni-LLMs. Demos and examples are available at this https URL
[CV-383] CFCH: Coarse-Fine Collaborative Hierarchical Learning for Anterior Segment Disease Analysis
链接: https://arxiv.org/abs/2609.32559
作者: Peng Wang,Haohan Zou,Yanlin Wu,Xueshuo Xie,Yan Wang,Tao Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: accepted by BIBM 2026
Abstract:Accurate classification of anterior segment diseases is crucial for ophthalmic screening and diagnosis. However, slit-lamp image analysis remains challenging due to substantial variability in imaging conditions and the intrinsic anatomical-disease hierarchy of ocular pathologies. Existing methods typically formulate this task as a flat multi-class classification problem, ignoring the structured dependency between anatomical regions (e.g., cornea, conjunctiva, and lens) and disease this http URL address these limitations, we propose CFCH, a Coarse-Fine Collaborative Hierarchical learning framework that explicitly models anatomical context and disease semantics through a dual-branch architecture. To enable effective cross-granularity collaboration, CFCH introduces semantic and cross-granularity attention consistency constraints, encouraging aligned yet complementary feature learning across branches. In addition, we construct AS-9K, a large-scale anterior segment dataset with 8975 images covering 12 common disease categories. To the best of our knowledge, AS-9K is the largest publicly available dataset for anterior segment image classification. Extensive experiments on two anterior segment datasets demonstrate that CFCH outperforms state-of-the-art methods. Qualitative visualizations further show more focused and lesion-relevant activation responses, validating the effectiveness of the proposed framework. Code will be available at this https URL.
[CV-384] Feature Space Guidance for Breast Cancer Classification in DCE-MRI
链接: https://arxiv.org/abs/2609.32555
作者: Benjamin Hamm,Yannick Kirchhoff,Maximilian Rokuss,Moritz Langenberg,Constantin Ulrich,Tassilo Wald,Jeremias Traub,Karol Gotkowski,Klaus Maier-Hein
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 11 pages, 2 figures, 2 tables. Code: this https URL
Abstract:Dynamic contrast enhanced breast MRI (DCE-MRI) is a powerful clinical tool for breast cancer detection, providing high resolution anatomical detail together with rich temporal contrast information. However, high dimensional 4D inputs, small lesions, and heterogeneous acquisition protocols across clinical sites hinder robust automated classification of healthy, benign, and malignant cases. To address these challenges, we propose a framework that dynamically analyzes latent representations to adapt to protocol-specific characteristics. Spatial variability is mitigated by reducing confounding background uptake and compensating for misalignment caused by deformable soft tissue. Additionally, relationships in the latent space across phases are leveraged to select the most informative temporal features, improving robustness to protocol-specific temporal variability. Finally, task specific discriminative features are promoted through large scale supervised lesion segmentation pretraining, which substantially enhances downstream finetuning. Evaluated under leave-one-center-out validation on the ODELIA dataset and the held-out AMBL cohort, the proposed framework substantially outperforms finetuned radiology foundation models and prior methods, improving mean AUROC by nearly 8 points and balanced accuracy by 4 points over the strongest baseline. Additionally, our method achieved first place in the MICCAI ODELIA Breast MRI Challenge 2025, further demonstrating its effectiveness for robust breast cancer classification. We publicly release our codebase under this https URL.
[CV-385] Harnessing Coupled Stream Completion For Human-Object Ineraction Modeling
链接: https://arxiv.org/abs/2609.32551
作者: Dawei Guan,Di Yang,Jiangtao Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 18 pages, 6 figures
Abstract:Text-conditioned human-object interaction (HOI) generation requires body motion, object trajectories rotations, and hand articulation to remain coordinated. These components differ in scale and dynamics, but must agree on contact, relative pose, and timing. A shared representation may limit the distinct structure of each stream, while independent generation prevents each stream from responding to changes in the others. Latent supervision alone also does not directly constrain contact after decoding. We propose TRACE, a continuous latent framework that keeps stream states separate and couples their updates. TRACE encodes body, object, and hand motion into separate latents and predicts each stream velocity from the complete current interaction state. Geometric losses on decoded motion further constrain contact and object-relative motion over time. The same model supports completion of any single absent stream from the other two. Frozen flow features also serve as input to a language model for HOI understanding. Experiments on InterAct, OMOMO, and BEHAVE show that joint completion training improves generation and that frozen flow features improve understanding over raw-motion encoding. On InterAct, TRACE achieves the highest contact precision, recall, and F1 among the compared methods.
[CV-386] In-Flight KV Cache with Clean Anchors for Faster Autoregressive Video Diffusion
链接: https://arxiv.org/abs/2609.32540
作者: Yikai Wang,Xiao Han,Mengmeng Xu,Juan Camilo Perez,Yiannis Douratsos,Sen He,Zijian Zhou,Fei Zhang,Zhaochong An,Juan-Manuel Perez-Rua,Chen Change Loy,Tao Xiang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multimedia (cs.MM)
备注: PJ page: this https URL
Abstract:Few-step autoregressive video diffusion generates a long video by splitting the video into temporal chunks and generating chunk-by-chunk, each through a short sequence of denoising stages. To memorize chunks that are already generated, previous methods reconstruct a clean or less-noisy key–value (KV) cache by additional forwards to build the cache without advancing an output latent. However, every denoising forward itself already computes the in-flight KV of the current chunk. We introduce FlashForward, which directly reuses this cache to avoid the heavy cache-update-only model forwards. After the current chunk completes one denoising stage, its stage-specific cache is already available for the next chunk. Assigning one GPU to each stage therefore lets different chunks occupy different stages concurrently. This early availability has a quality cost: the resulting stage-matched history is noisy, causing appearance and motion drift among chunks. To complement it, FlashForward produces sparse auxiliary clean anchor latents before the corresponding region is generated so the generation trajectories can be stabilized by this two-sided conditioning. The two memories operate at different temporal scales: sparse clean anchor KV supplies coarse, long-range two-sided structural guidance, while dense stage-matched history preserves fine, recent evolution. With up to four GPUs, FlashForward runs 1.16 – 1.69\times faster than HiAR and 1.42 – 2.92\times faster than Self-Forcing for 16 FPS videos of 20 seconds or longer across 1.3B and 14B backbone scales at 480p and 720p. On VBench, for the 1.3B model at 480p, it achieves higher scores and remains stable at longer durations, demonstrating that FlashForward generates high-quality and temporally consistent videos across durations of 20s, 35s and 65s at a much faster generation speed.
[CV-387] RIPE-MambaSpike: Resolution-Independent Spiking-State-Space Interfaces for Parameter-Efficient Event-Based Vision
链接: https://arxiv.org/abs/2609.32537
作者: Md Muhiminul Islam,Shoaib Ahmed Dipu,Sayeed Shafayet Chowdhury
类目: Computer Vision and Pattern Recognition (cs.CV); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Spiking-Mamba hybrids reach strong accuracy on event-based vision, but existing designs often require tens of millions of parameters. Much of that cost comes from how the spiking front-end is connected to the state-space backbone rather than from the hybrid architecture itself. In a representative model, a single resolution-dependent projection accounts for 33.55M of 36.25M parameters. To that end, we introduce RIPE-MambaSpike (Resolution-Independent, Parameter-Efficient), which replaces that projection with a hierarchical multi-resolution bridge of fixed channel width. Its deployed footprint is 0.870M parameters, constant at fixed time steps and widths across a 43x range of input areas. Reparameterized spiking stages, temporal decoupled modulation, and a dynamic convex-hull-bounded dual-stream membrane-potential attention preserve accuracy under this compact design. Result-wise, RIPE-MambaSpike is pareto-optimal on CIFAR10-DVS, N-Caltech101, and DailyDVS-200. Notably, on the 200-class DailyDVS-200, a scaled 8.04M configuration achieves 45.7% top-1 accuracy, the best reported spiking result on that benchmark, and outperforms prior spiking methods with 3.0-15.1x fewer parameters than dense ANNs. Overall, our findings demonstrate that competitive event-based recognition does not require resolution-dependent parameter growth. Code is available at this https URL.
[CV-388] DepthBench: Measuring How Residual Connections Enable More Computational Depth
链接: https://arxiv.org/abs/2609.32534
作者: Keyu Wang,Yangyi Huang,Jiale Kang,David González-Martínez,Weiyang Liu,Shiwei Liu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Depth is a natural way to increase the computational capacity in Transformers, yet the contribution of deeper layers can diminish as depth grows larger. Recent approaches enhance normalization (\texte.g., LayerNorm Scaling) or residual connections (\texte.g., mHC, AttnRes) to enable better information flow and depth utilization. However, it remains unclear whether they truly translate increased architectural depth into effective computational depth, and whether their reported gains stem from better access to information across depth, or unaccounted-for confounding factors. In this paper, we introduce \textbfDepthBench, a controlled benchmark for studying computational depth across various architectures. We systematically vary the width–depth aspect ratio ( d_\textmodel/n_\textlayer ) from shallow–wide to deep–narrow shapes, while keeping the model size and pre-training recipe fixed. Across 10 representative architectures, we find that the benefit of allocating more capacity to depth is strongly architecture-dependent. Standard Pre-LN and most of its norm- and scaling-based variants provide little benefit and can even degrade performance as models become deeper and narrower, whereas HC and Full AttnRes improve consistently even at extreme deep shapes. These gains extend beyond pre-training loss and consistently translate into improved domain-specific performance and effective computation. Controlled layer-level analyses further show that the gains of HC and Full AttnRes are associated with more effective utilization of additional layers, revealing distinct mechanisms of computational depth across architectures. Overall, our results identify residual connection design as a key determinant of whether depth can serve as a meaningful scaling axis by enabling additional architectural depth to translate into effective computation.
[CV-389] SetOPD: From Few Visual Exemplars to Multimodal Candidate Sets for Remote-Sensing Open-Prompt Detection
链接: https://arxiv.org/abs/2609.32529
作者: Jinlong Hu,Yi Zhang,Zhiqi Xia,Yikang Zhou,Shunping Ji
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Open-prompt detectors allow users to specify targets with text, visual exemplars, or both. We argue that existing designs underuse the visual modality in two ways. First, multiple exemplars are commonly compressed into a single class-level embedding. This textualizes visual prompting: the resulting vector plays the role of another class name, is often aligned to or injected into the text pathway, and may be suboptimal when only a few heterogeneous exemplars are available. Second, existing methods interact primarily in prompt or representation space, before modality-specific detection states are formed. We address both issues from a set perspective: \setopd preserves modality-specific decoding states from a shared prompt-conditioned initialization and performs explicit multimodal collaboration at the candidate-state level. For the first issue, we introduce \br prompting, which reads every boxed exemplar in its full scene context and decomposes the pooled support evidence into a Base anchor and a learned Residual correction; the resulting prompt has fixed capacity regardless of the number of examples and drives its own visual detection pathway. For the second, we recast text–visual collaboration from representation-level fusion into a candidate-set modeling problem. Paired-Query Arbitration (\pqa) then performs explicit cross-modal state arbitration only after modality-specific candidate states have been formed. The two readers share query initialization so their candidates are paired by index; a learned gate arbitrates within each pair, followed by a permutation-equivariant module that reasons over the fused set.
[CV-390] UnStep: Training-Free Acceleration of Causal Video Diffusion with Fewer Steps Than Distillation
链接: https://arxiv.org/abs/2609.32518
作者: Youssef Mansour,Enis Simsar,Fadime Sener,Markos Georgopoulos,Albert Pumarola,Ali Thabet,Edgar Schoenfeld
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Distilling bidirectional multi-step video diffusion transformers into few-step causal models has become a common approach for streaming video generation. While these few-step students are significantly faster than the teachers they are distilled from, they remain slow for real-time generation. In this work we present UnStep, a training-free wrapper that accelerates few-step causal video models at inference by running them with fewer diffusion transformer (DiT) steps than during distillation and limiting the temporal window retained in the attention KV cache. We propose two inference-only mechanisms to recover quality lost by step reduction and attention windowing: renoising the generated latent frames to a near-clean level and reusing the existing clean-cache pass to refine them, and applying truncated SVD to the DiT attention value and output projections. We also accelerate inference with a quality-preserving runtime stack for the DiT and VAE decoder, including more efficient attention calls and KV indexing, fused Triton RoPE with cached coefficients, and VAE decoding with optimized memory layout, precision, and convolution kernels. By reducing computation and optimizing the runtime stack, UnStep sets a new throughput regime for causal video diffusion, by running substantially faster than current methods, reaching 50 FPS on a single H100 without quality loss, and 77 FPS on GB200, all without retraining.
[CV-391] LocalProp: Neuro-Localized Memory-Efficient Backpropagation
链接: https://arxiv.org/abs/2609.32517
作者: Diana-Nicoleta Grigore,Iuliana Georgescu,Radu Tudor Ionescu
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Optimization and Control (math.OC)
备注:
Abstract:The current deep learning training paradigm employs end-to-end backpropagation, regardless of the training stage, i.e. pre-training or fine-tuning. However, backpropagating through the entire model is neither biologically plausible nor memory efficient, since learning inside the brain is highly localized. Therefore, we propose LocalProp, a training procedure that locally updates the weights of a model. Our neuro-localized weight updates follow the “pre-training then fine-tuning” paradigm, where the pre-training is based on I-JEPA. After locally updating the weights, a pruning operation is performed, followed by a short final fine-tuning phase. Pruning helps by sending the learning signal from higher blocks to lower blocks. We perform experiments on several datasets, including large-scale benchmarks such as ImageNet, and empirically show that LocalProp reaches good performance at a fraction of GPU peak memory. By varying the number of jointly optimized blocks, we identify gradient-propagation span as a practical control over the accuracy-memory trade-off.
[CV-392] Cross-Domain Few-Shot Writer Adaptation for Real-World Handwritten Mathematical Expression Recognition
链接: https://arxiv.org/abs/2609.32513
作者: Paulo Grane Gabriel Silva,Lorenz Bernard Marqueses,Joel Ilao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the 32nd IEEE International Conference on Mechatronics and Machine Vision in Practice (M2VIP 2026)
Abstract:Handwritten mathematical expression recognition (HMER) refers to the task of recognizing and converting handwritten mathematics into a parsable markup language, usually LaTeX. No current state-of-the-art-competitive system adjusts to the way a specific person writes, and the domain gap between training images (usually digital or perfectly binarized) and images physically taken with a camera used in inference has received fairly little attention for this specific problem. We characterize this domain gap through fragmentation and stroke-width analyses of the images as well as introduce a writer-adaptive fine-tuning pipeline to MFH-CoMER in an attempt to address it. We further introduce sample author-specific datasets, consisting of five handwriting category subsets from two authors, and evaluate using a McNemar’s test and permutation tests adapted to limited data. Results suggest an increase in expression recognition rate and a decrease in CER for digital handwriting but more varied results for physical handwriting, with the model struggling for handwriting articles that the base model can already evaluate well. Nonetheless, adapted models were found to have improved results for four out of five subsets. Statistical testing results suggest a consistency in improvement for two of the tested author-specific subsets. Our results point towards the potential feasibility of writer adaptation for the HMER task.
[CV-393] GeoCR: Learning a Generalist Cloud Removal Prior from Heterogeneous Observations
链接: https://arxiv.org/abs/2609.32510
作者: Jeonghyeok Do,Munchurl Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Please visit our project page this https URL
Abstract:Cloud removal methods are typically specialized to individual datasets and input configurations, limiting reuse across sensors, spectral bands, and observation settings. We introduce GeoCR, a generalist model that unifies RGB-only-based CR and multispectral-based CR from single- or multi-temporal cloudy observations, with optional SAR guidance, within a single network. To accommodate different spectral and sensing domains, compact input and output stems extend a pretrained RGB autoencoder while keeping its encoder and decoder trunks frozen. This shared latent interface enables a single flow transformer to jointly model clean RGB and non-RGB latents, conditioned on separate cloudy-observation streams and optional SAR tokens. Through joint pretraining on the training splits of ten datasets comprising 883,331 cloud-free target images, GeoCR learns a shared cloud removal prior across these heterogeneous configurations. The same pretrained checkpoint supports direct inference without dataset-specific fine-tuning and efficient adaptation through low-rank adaptation (LoRA). We evaluate GeoCR against general image restoration and cloud removal methods on test splits of the contributing datasets under full-band and RGB-only settings. GeoCR achieves the best FID and DISTS on full-band SEN12MS-CR and Sen2_MTC_New and RGB-only CUHK-CR2, outperforming existing models and demonstrating the effectiveness of a reusable generative model across diverse settings.
[CV-394] When Helpful Text Hurts: Option-Redirecting Bias in Vision-Language Models ACM-MM2026
链接: https://arxiv.org/abs/2609.32489
作者: Tam Le Thi Thanh,Hoang Tran Van,Hong-Hanh Nguyen-Le,Thanh Duc Ngo
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at ACM Multimedia 2026 (ACM MM 2026). 25 pages, 16 figures. This arXiv version includes supplementary material
Abstract:In tri-modal visual question answering (VQA), auxiliary text is commonly used to complement visual and textual inputs, yet its reliability is often uncontrolled. While prior work studies modality conflicts in general, it remains unclear how different types of unreliable auxiliary text affect answer selection under fixed image-question-option contexts. In this work, we show that the most harmful auxiliary text is not necessarily the most factually incorrect, but the one that aligns with the question while contradicting the image and favoring a specific distractor, leading to systematic redirection of model predictions. To isolate this effect, we introduce the Textual Reliability Ladder, a controlled diagnostic protocol that decomposes auxiliary text along three axes: image consistency, question relevance, and option support. Across multiple datasets (ScienceQA, VCR, A-OKVQA, Causal-VidQA) and recent VLMs, we find that such distractor-supporting text induces the largest accuracy drops (up to 53.1%) and concentrates errors on specific incorrect options. To mitigate this failure mode, we propose a training-free inference-time intervention that explicitly counteracts this redirection effect via noise-stability steering and dynamic grounding, reducing redirected errors while largely preserving performance under faithful text. Our results highlight that auxiliary-text reliability must be understood at the decision level, rather than solely through factual correctness, and provide a practical pathway toward more robust tri-modal reasoning.
[CV-395] JEPA Learns What the Mask Leaves Unrecoverable
链接: https://arxiv.org/abs/2609.32481
作者: Peng Xie,Amr Alanwar
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Joint-embedding predictive architectures are unusually sensitive to how the input is masked: block masks work, scattered masks do not, and the explanations are empirical. We give a measurement account. A mask is a linear measurement, and in a compactly supported wavelet basis every atom whose support lies inside the hidden region falls in the measurement’s null space and leaves no trace in the data. The JEPA loss asks only that the encoded context suffice for the target, so a target that a low-level prior can recover admits a shortcut, one the moving-average target encoder can make self-consistent. What removes the shortcut is the coarse-scale content the mask leaves unrecoverable, provided enough context stays within reach of each target. We score that content before training and test the account’s distinctive predictions in 151 pre-training runs. On ImageNet-100, strip masks match blocks in area and contiguity yet are recoverable, and they land at 40.3% linear top-1, beside random masks at 40.8%, against 64.3% for blocks; within one geometry family, the placements that leave the least unrecoverable content lose 6.5 points to those that leave the most, over five seed pairs; pixel targets span 7 points where latent targets span 25; and against a frozen target the gap between random and block masks, 19 points on the same kind of GPU, closes to 1.5, so the geometry acts through the target the encoder produces for itself. On UCF101 the masking ratio decides which condition, content or reach, binds; removing whole frames, unrecoverable in space but recoverable from neighbouring frames, is worst at both ratios; and on V-JEPA’s own masks, batching them intact instead of truncated changes little (36.0% against 35.1%), whereas making 100 target tokens inside the blocks visible lifts them to 48.7% and hiding 100 context tokens outside the blocks does not (33.7%).
[CV-396] Back-Tracking from Clarity: Self-Learning to See Text from Afar
链接: https://arxiv.org/abs/2609.32477
作者: Duc-Tri Tran,Phi Le Nguyen,Minh Hoai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We propose a self-supervised framework designed to enhance the capability of scene text detectors in identifying and recognizing text in scenarios where instances are shown at significant distances, typically small, blurred, and frequently missed by conventional models. Our approach leverages the high-fidelity performance of existing text spotting models on large, clear text as a foundational supervisor. By temporally back-tracking these high-confidence detections through video sequences, we automatically synthesize pseudo-labels for preceding frames where the distant text is still visually degraded or undersized. These pseudo-labels enable training a student model specialized for early text detection, without requiring any manual annotation. The success of this approach depends on accurate pseudo-label generation, for which we develop a dedicated scene text tracker capable of maintaining consistent text identities across challenging video sequences. In addition, we propose SceneText50, a diverse multilingual outdoor dataset to facilitate training and evaluation. Experiments show that our framework significantly improves early detection accuracy and robustness across varied scenes and languages. Code and data are at \hrefthis https URLthis https URL.
[CV-397] Can Motion-Language Models Ground Structure? STRIDE for Evaluating the Evaluators
链接: https://arxiv.org/abs/2609.32462
作者: Lixing Tan,Qing Xia,Yuting Guo,Shuai Li,Aimin Hao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Motion-language models are typically scored by motion-language evaluators, but how well these evaluators ground language structure remains unclear. Here, we introduce the Structure grounding via Temporal-order, Reflection, and Identity Diagnostic Evaluation (STRIDE) benchmark to systematically evaluate the ability of evaluators to track temporal order, mirror reflection, and action identity. STRIDE comprises 5,869 triples, each consisting of a motion, its original caption, and a perturbed caption, spanning both short and long descriptions. We likelihood-balance caption pairs to reduce text-only bias and estimate each evaluator’s caption preference under unrelated motions to measure the discrimination gain from matched motions relative to this baseline. Our experiments reveal weak structural grounding and severe deficits in mirror sensitivity among the audited evaluators, which commonly used evaluation protocols fail to expose. To understand why these limitations go undetected in standard tests, we examine the evaluators more closely. We find that text-only priors alone can solve naive perturbation tests on existing datasets, while common retrieval and distributional metrics barely respond to structural corruption introduced by mirroring ground-truth motions. These findings suggest a natural intervention: structural hard negatives. Our experiments show that a simple modification to contrastive learning substantially improves performance on temporal order and mirror reflection. The benchmark and code will be released.
[CV-398] REMEDY: How Far Is Video Generation from Medical Education World Models?
链接: https://arxiv.org/abs/2609.32460
作者: Lixing Tan,Yanghao Zhou,Qing Xia,Yuting Guo,Shuai Li,Aimin Hao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent video generation models produce realistic videos and show potential as a foundation for world models. These advances create opportunities for generating medical teaching demonstrations, which requires both convincing visual quality and precise procedural actions. However, whether current generators can meet these requirements has not been measured. To address this problem, we introduce Readiness Evaluation of Medical Education Demonstration sYnthesis (REMEDY), to our knowledge, the first benchmark for AI-generated medical teaching demonstrations. REMEDY provides 900 first frames from real demonstration videos, covering 12 tasks across four scenarios: operating room, imaging, clinic and bedside, and resuscitation. Five contemporary open-source video generation models produce 4,500 videos from these frames. We combine task-specific clinical checklists with video and motion quality metrics. Evaluation covers four dimensions: clinical action following, clinical profiles, video quality, and motion quality. Our results show that realistic appearance and temporal consistency do not ensure correct clinical actions. Even the most advanced MiniMax-H3 achieves only 28.25% on strict clinical success rate, and fine-grained clinical actions remain challenging. These findings establish a foundation and roadmap for developing future medical education world models.
[CV-399] Seeing Parts Reasoning about Worlds: Visual Inference under Partial Observation
链接: https://arxiv.org/abs/2609.32456
作者: Wei Wang,Wenqiao Zhang,Yutong Lin,Jun Xiao,Yueting Zhuang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 24 pages, 5 figures, 5 tables
Abstract:World modeling under partial observation requires reasoning about the complete worlds that remain compatible with limited visual evidence. Occluded objects and unseen regions can leave several world states possible; additional views can exclude alternatives and strengthen the conclusions supported by the observations. We introduce WorldScope to study this process through possible-world semantics, evidence-grounded data, and learned visual representations. WorldScope-1.2M provides 1.2 million English question-answer pairs spanning eight world properties, ten task interfaces, and three observation protocols. Its answers encode confirmed facts, supported bounds, and unresolved possibilities. Complementary supervision comprises 4,800 certified counterworld groups with equivalent base observations and different hidden object configurations and query answers. These groups provide physical witnesses of ambiguity and training-only labels for world compatibility and view-induced exclusions. We propose WorldFlow, which composes cross-view entity evidence and support-surface coverage into an image-subset evidence lattice. Counterworld compatibility and transition objectives train subset representations to reflect how new observations constrain possible worlds. A shared answer generator uses these representations to predict the strongest supported conclusion. WorldScope-Bench evaluates claim judgments and evidence-dependent conclusions as views are selected, combined, removed, or ordered. On its 5,000-question test set, WorldFlow reaches 64.34% exact accuracy, improving over the same backbone trained on QA alone by 24.88 percentage points. It retains 50.43% accuracy on the 3,000 questions from structure-disjoint scenes.
[CV-400] QuacamFM: Quaternion-Constrained Flow Matching for Camera Pose Estimation
链接: https://arxiv.org/abs/2609.32455
作者: Bao-Long Tran,Cuong Le,Tahereh Dehdarirad,Fredrik Viksten,Per-Erik Forssén
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Camera pose estimation from multi-view images remains a challenge in computer vision. Traditional methods often address this problem using Structure-from-Motion (SfM) with bundle adjustment. However, camera poses estimated from sparse views are inherently ambiguous due to insufficient geometric constraints. Recent work leverages probabilistic models, such as diffusion models, to generate multiple camera pose hypotheses and therefore capture this uncertainty better. Most of these methods represent camera rotations using unit quaternions, but treat them as unconstrained 4D vectors during the generative processes, thereby ignoring the unit-norm constraint of quaternions. Unconstrained quaternions create non-smooth and suboptimal generation trajectories. To this end, we propose QuacamFM, a quaternion-constrained flow matching framework for camera pose estimation that preserves unit quaternion representations throughout the entire flow trajectory. We design the optimal transport of the quaternion flows using smooth spherical linear interpolation. Experiments on CO3Dv2 demonstrate our method’s advantage in camera pose accuracy over diffusion-based methods and classical SfM approaches. We further show that our quaternion-constrained formulation outperforms the naive application of standard flow matching to 4D quaternion vectors on sparse-view camera pose estimation. Finally, it is observed that QuacamFM generalizes well across datasets and in-the-wild examples.
[CV-401] De-biasing Skeleton-based Action Recognition with Convex Hull Adaptive Shift
链接: https://arxiv.org/abs/2609.32454
作者: Mengyuan Liu,Yuhang Wen,Yi Zhang,Songtao Wu,Hong Liu,Junsong Yuan,Beichen Ding
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted for publication in International Journal of Computer Vision (IJCV). Our code is publicly available at this https URL
Abstract:Skeleton sequences can represent both individual actions and multi-entity interactions, encompassing human bodies, hands, objects, and robots. Existing approaches to recognize skeleton-based actions and interactions usually adopt a late fusion strategy, which expects individuals are independent and identically distributed to train a robust weight-shared entity encoder. However, observed entity bias in various skeletal data violates this assumption, leading to suboptimal optimization of backbone models that might produce wrong recognition results. This bias arises from the world coordinate system’s initial configuration, where the choice of origin often creates bias in representation. To this end, we propose a Convex Hull Adaptive Shift based normalization method to reduce Entity bias (CHASE), improving performance across a variety of skeleton-based action and interaction recognition tasks. To adaptively apply plausible shifts to the input skeletons, we formulate a plug-and-play parameterized network that ensures the relocated world origin lies within the skeleton convex hull, which avoids non-convergence by limiting the search space. To further minimize entity bias, we incorporate an auxiliary objective that leverages pair-wise distribution distances to guide network optimization. To support both single- and multi-entity actions, we propose a sub-entity strategy that offers a consistent formulation for both scenarios. Moreover, CHASE demonstrates compatibility with various intra-skeleton modalities, such as bones and velocities, highlighting its adaptability. Essentially, our method works as a normalization approach to reduce entity bias, enabling subsequent classifiers to achieve improved recognition performance across diverse settings. Extensive experiments on 7 datasets verify our approach by seamlessly integrating with various backbones and significantly boosting their performance.
[CV-402] CityToolVQA: Tool-Augmented Visual Question Answering for 3D Spatial Cognition in Urban Low-Altitude Environments
链接: https://arxiv.org/abs/2609.32427
作者: Boao Yu,Yingzhen Nie,Yue Hu,Zhengqiu Zhu,Rusheng Ju
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 3 figures, 3 tables
Abstract:CityToolVQA addresses the weak performance of Vision-Language Models (VLMs) on quantitative tasks in urban low-altitude visual question answering. We divide the seven tasks into qualitative and quantitative groups: qualitative questions are answered directly by the VLM, whereas quantitative questions are handled by an external visual-geometric toolchain that performs object grounding, segmentation, depth back-projection, and spatial computation. The toolchain can be attached to different VLMs in a zero-shot manner; a Depth-Assisted Prompt Inference (DAPI) fallback is triggered when the main-chain detection is invalid or unreliable, and CityToolVQA-SFT adapts the 8B backbone to tool-conditioned inputs. On the 73,324-question Open3D-VQA-v2 test set, CityToolVQA-SFT (Qwen3-VL-8B) reaches 67.6% overall accuracy, and attaching the toolchain to ten open-source VLMs improves quantitative-task accuracy by 12.7-36.9 percentage points. These results indicate that externalizing explicit 3D geometric computation effectively complements the limited ability of RGB-only VLMs to estimate metric distances and object sizes.
[CV-403] RefAdapt-DiT: Adaptive Joint Attention for Reference-Conditioned Diffusion Transformers
链接: https://arxiv.org/abs/2609.32415
作者: Jian Tang,Jiawei Fan,Qiannan Zhou,Qingbin Liu,Jiang Bian,Zang Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Diffusion Transformers (DiTs) have become the standard backbone for high-quality generative modeling, yet deploying them in conditional generation tasks remains computationally prohibitive because bidirectional joint attention repeatedly processes large reference streams. While existing optimization schemes mitigate generic temporal redundancy, they typically rely on coarse-grained static reuse and overlook the distinct dynamics of references and targets. Specifically, we observe that reference representations often evolve slowly along the generation trajectory, while the target often assigns little attention mass to them; reference drift and this target-to-reference exposure jointly shape how strongly stale reference states affect the target. To exploit these patterns, we introduce \RefAdapt, a training-free framework for adaptive control of joint attention between references and targets. Instead of rigid static strategies, \RefAdapt combines consecutive target-Q change with previously observed target-to-reference attention mass to control reference computation adaptively at block granularity. Under ultra-few-step settings, \RefAdapt enables speedups of up to 2.097\times on 4-step MiniMax H3 and 3.54\times on 8-step Qwen Image Edit, while maintaining comparable visual quality.
[CV-404] oward On-Chip Training of Spiking Neural Networks for Dense Event-Based Vision
链接: https://arxiv.org/abs/2609.32405
作者: Maxime Vaillant,Axel Carlier,Lai Xing Ng,Christophe Hurter,Benoit R. Cottereau
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Event cameras provide low-latency, asynchronous visual sensing for resource-constrained robotics. Spiking neural networks (SNNs) process event streams naturally, but training deep SNNs with backpropagation through time (BPTT) requires substantial memory and remains difficult on neuromorphic hardware. Local learning avoids this by restricting error propagation to local blocks, but existing methods mainly target classification rather than dense prediction. We introduce DELL (Dense Event-driven Local Learning), a block wise scheme for dense event-based vision that replaces global gradient propagation with local dense supervision. Learnable, spatially structured local heads supervise each block at its appropriate resolution while preserving temporal dynamics within blocks. We evaluate DELL on optical-flow regression and semantic segmentation with a fully spiking U-shaped architecture. On DSEC optical flow, DELL reduces peak training memory by 39.6% relative to end-to-end BPTT while improving accuracy, reaching 1.670 px endpoint error on the official test benchmark versus 1.941 px for the same backbone trained end-to-end. Block detachment behaves more like a regularizer than a constraint. DECOLLE, the existing local-learning baseline, relies on fixed random local read-outs poorly suited to dense regression, resulting in a 3.9x higher endpoint error; learnable local heads recover this loss and outperform end-to-end training across all optical-flow metrics. On segmentation, they recover most of the performance gap, although DELL remains a few mIoU points behind end-to-end training. With 2.3M parameters, 24x fewer than the strongest SNN baseline, the backbone remains competitive with the SNN state of the art on DSEC. These results extend local learning to dense event-based prediction while substantially reducing training memory.
[CV-405] Endo-TSR: Temporal Spectral Modeling of Appearance and Motion for Endoscopic Reconstruction
链接: https://arxiv.org/abs/2609.32399
作者: Taoyu Wu,Yiyi Miao,Qi Shao,Zhuoxiao Li,Zhe Tang,Limin Yu,Baoru Huang
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 5 pages, 3 figures, 2 tables
Abstract:Endoscopic scene reconstruction requires modeling tissue motion and temporal appearance while recovering fine surface detail. Deformable Gaussian models provide explicit trajectories, but their fixed colour coefficients lack a dedicated temporal representation for photometric changes. We propose Endo-TSR, which augments deformable Gaussian splatting with bounded Fourier colour residuals and independent translation residuals on shared temporal frequencies. The colour residuals capture local appearance changes, while a Matérn spectral prior regularises motion corrections. Multi-scale Laplacian supervision guides tissue-detail recovery during joint image fitting. Extensive experiments on the EndoNeRF and StereoMIS datasets demonstrate state-of-the-art rendering quality, with the highest PSNR across all evaluated sequences. Ablation studies show that temporal appearance yields the largest PSNR gain among the tested component additions, while appearance and detail supervision jointly improve rendering with fixed Gaussian counts.
[CV-406] RefCompose: Multi-Reference Image Generation via LoRA-Conditioned Diffusion ECCV2026
链接: https://arxiv.org/abs/2609.32389
作者: Sai Sri Teja Kuppa,Parth Shinde,Priyadharsan Balaji S,Jinka Harshavardhan,Sriprabha Ramanarayanan
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注: Accepted in ECCV 2026 Workshop on AI for Visual Arts
Abstract:Filmmakers and visual artists routinely need to compose multiple references, actors, locations, props, cultural elements, into a single coherent shot, but existing tools either fail to scale past a handful of references or destroy fine grained subject identity in the process, since per reference tokenization scales memory linearly with reference count N and generated content often departs from the given references rather than reproducing them. We propose \textbfRefCompose, a pixel space compositional conditioning framework that decouples \emphwhere things go from \emphwhat they look like, via a single fixed resolution reference canvas that keeps conditioning size constant regardless of reference count. Spatial layout is induced at inference time from a frozen diffusion transformer and extracted via Grounding DINO, requiring no LLM or dedicated layout model, while dual stream LoRA adapters inject a layout derived depth map and the encoded canvas through separate low rank streams, disentangling geometric scaffolding from localized appearance. On the Dense Layout protocol, RefCompose consistently outperforms layout based and state of the art multi reference baselines on color, texture, shape, spatial accuracy, and identity/content preservation at higher reference counts, all with constant inference memory, making it a practical building block for multi subject cinematic composition at production scale.
[CV-407] An End-to-End Latent-Rollout Approach for Pushing Few-Step ImageNet-256 Generation to FID 1.11 without Fréchet Losses
链接: https://arxiv.org/abs/2609.32376
作者: Xiaoran Xu,Yujing Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Iterative generation poses a joint optimization problem across steps, as intermediate predictions shape subsequent computations and ultimately determine the final output distribution. Few-step generators distilled from pretrained diffusion and flow-matching models make such optimization computationally practical end to end. We build on this opportunity with a distill-then-refine approach that uses teacher imitation to establish a strong initialization for a few-step rollout in latent space, then shifts to end-to-end refinement of the complete latent rollout against real data. We introduce FiST (Flow-in-Stage Transformer), an architecture that composes learned latent-state transitions in a few stages using a shared Transformer, with optional cross-stage hidden communication. Distillation applies teacher-forced regression to selected states along teacher trajectories; refinement replaces this supervision with adversarial and auxiliary classification objectives on the final latent output. A trainable discriminator module operates on semantically rich features extracted from clean real and generated latents by a frozen SiT backbone pretrained with REPA. All training takes place in latent space, without image decoding. During refinement, FiST consumes its own intermediate predictions, and endpoint gradients pass through every generation stage. For class-conditional generation on ImageNet at 256\times256 , our approach achieves FID 1.11 (IS 282) with three stages and FID 1.15 (IS 280) with two. These results demonstrate competitive few-step generation through learned distribution-level supervision, without explicit Fréchet-distance minimization. Ablations characterize how distillation, pretrained checkpoint choices, refinement supervision, and cross-stage hidden communication affect generation quality.
[CV-408] StegGNN: Learning Graphical Representation for Image Steganography
链接: https://arxiv.org/abs/2609.32362
作者: Abhinav Kumar,Shorya Singhal,Agam Pandey,Tushar Kumar,Sukrit Jindal
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 8 figures
Abstract:Image steganography refers to embedding secret messages within cover images while maintaining imperceptibility. Recent advances in deep learning - primarily driven by Convolutional Neural Networks (CNNs) and architectures such as inverse neural networks, autoencoders, and generative adversarial networks - have led to notable progress. However, these frameworks are primarily built on CNN architectures, which treat images as regular grids and are limited by their receptive field size and a bias toward spatial locality. In parallel, Graph Neural Networks (GNNs) have recently demonstrated strong adaptability in several computer vision tasks, achieving state-of-the-art performance with architectures such as Vision GNN (ViG). This work moves in that direction and introduces StegGNN - a novel autoencoder-based, cover-agnostic image steganography framework based on GNNs. By modeling images as graph structures, our approach leverages the representational flexibility of GNNs over the grid-based rigidity of conventional CNNs. We conduct extensive experiments on standard benchmark datasets to evaluate visual quality and imperceptibility. Our results show that our GNN-based method performs comparably to existing CNN benchmarks. These findings suggest that GNNs provide a promising alternative representation for steganographic embedding and open the field of deep learning-based steganography to further exploration of GNN-based architectures.
[CV-409] EyeVQA: Benchmarking Ophthalmic Vision-Language Models from Recognition to Spatial Grounding MICCAI2027
链接: https://arxiv.org/abs/2609.32352
作者: Gujie Shao,Zixun Xie,Xuechun Xing,Ruixiang Wang,Ziyun Lan,Yanlin Qi,Gangyi Zhang,Yuxin Yang,Dawei Li,Haiming Tang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to MICCAI 2027 CREATE Workshop. 19 pages, 2 figures, 2 tables. Project page: [ this https URL ]( this https URL )
Abstract:Vision-language models (VLMs) have shown increasing potential for medical image understanding, yet their capabilities in ophthalmic imaging remain insufficiently characterized. Existing ophthalmic datasets are typically designed for individual diseases or specialized tasks, making it difficult to systematically evaluate whether VLMs can move beyond disease recognition toward comparative reasoning and fine-grained spatial grounding. We introduce EyeVQA, a unified visual question answering benchmark for comprehensive evaluation of ophthalmic VLMs. EyeVQA is constructed from 21 available ophthalmic datasets and contains 20,000 question-answer pairs spanning six disease groups and seven question types: Single-Choice, Multi-Select, Variable-Select, True-False, Ranking, Point Location, and Bounding Box. Gold answers are deterministically derived from source-provided diagnoses, severity grades, clinical findings, segmentation masks, bounding boxes, and anatomical landmarks, enabling reproducible evaluation without relying on model-generated annotations. Notably, 44.5% of the questions require reasoning across multiple images, extending evaluation beyond conventional single-image medical VQA. We benchmark fourteen representative general-purpose, scientific, and medically specialized VLMs under a unified zero-shot protocol. The best-performing model only achieves an overall score of 62.8, while substantial gaps remain in spatial grounding and cross-task generalization. These results highlight the limitations of current VLMs in comprehensive ophthalmic visual understanding and establish EyeVQA as a diagnostic benchmark for developing more reliable and spatially grounded ophthalmic multimodal models. The project page is available at this https URL.
[CV-410] OpenMASC: An Open-Source Pipeline for Cross-Trajectory Metal-Aware Sampling and Correction in Accelerated MRI
链接: https://arxiv.org/abs/2609.32343
作者: Zhengyi Lu,Ming Lu,Chongyu Qu,Junchao Zhu,Junlin Guo,Marilyn Lionts,Yanfan Zhu,Yuechen Yang,Tianyuan Yao,Jayasai Rajagopal,Bennett Allan Landman,Xiao Wang,Xinqiang Yan,Yuankai Huo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Metal implants corrupt MRI measurements throughout k -space, yet existing accelerated MRI methods assume clean data and most metal artifact reduction approaches assume fully sampled acquisitions. No public dataset provides paired k -space and images with and without metal for the same anatomy, and no framework jointly addresses artifact-aware acquisition and reconstruction across sampling trajectories. We present OpenMASC, an open-source pipeline covering the full workflow from data generation to deployment. A physics-based data generation module converts public CT volumes into paired clean and metal-corrupted MRI data in both Cartesian and radial formats. MA-VarNet, an unrolled reconstruction network with a per-cascade DC Rectifier, corrects artifacts that data-consistency steps reintroduce from corrupted measurements. A reinforcement learning agent actively selects k -space readouts and co-trains with the reconstruction network through a decoupled three-stage procedure. The framework is trajectory-agnostic except for the data-consistency operator, supporting both Cartesian and radial acquisition without architectural changes. Experiments on two datasets at 4\times and 8\times acceleration demonstrate consistent improvements over conventional and learned baselines on both trajectories.
[CV-411] SGA-Flow-GRPO: Spatial Gradient-Guided Credit Assignment for Flow-GRPO
链接: https://arxiv.org/abs/2609.32340
作者: Yunkai Yang,Yudong Zhang,Xinying Chen,Bin Luo,Jienan Lyu,Kunquan Zhang,Weitao Wan,Runmin Dong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reinforcement Learning (RL) has proven effective in aligning flow-based generative models with human preferences. Recently, Flow-GRPO has emerged as an efficient critic-free paradigm by calculating advantages over sampled candidate trajectories. However, standard Flow-GRPO applies a uniform scalar advantage across both temporal denoising steps and spatial latent dimensions, without explicitly accounting for the spatial structure of generated images, which may lead to sub-optimal policy updates. To address this, we propose a novel gradient-guided spatial credit assignment framework tailored for Diffusion Transformers (DiTs). We first reformulate the transition-level log-likelihood in Flow-GRPO into a token-wise representation natively aligned with DiT patch architectures, constructing spatially fine-grained importance sampling ratios. To allocate localized credit without rigid, boundary-sensitive segmentation heuristics, we introduce a continuous spatial credit map derived from reward gradients. Crucially, we employ an outlier-robust normalization scheme based on Median Absolute Deviation (MAD) coupled with temperature scaling, effectively eliminating gradient noise while highlighting functional prompt-aligned regions. Extensive evaluations on GenEval show that our approach delivers SOTA alignment quality, achieving a convergence rate comparable to top-tier methods like DiffusionNFT while substantially improving upon Flow-GRPO-based methods in alignment performance.
[CV-412] Progressive-View On-Policy Distillation for Regional-to-Global Transfer in Multimodal LLM s
链接: https://arxiv.org/abs/2609.32333
作者: Shanfeng Huang,Zhou Fang,Song Xiao,Hai Du
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Regional-to-global distillation uses crop-conditioned guidance to improve full-image understanding. The challenge is to effectively transfer the teacher’s crop-based advantage to the student’s full-image inference. We propose progressive-view on-policy distillation (PVD), which shifts the student’s view distribution from the crop toward the full image through an intermediate aspect-preserving padded crop. The padded crop preserves regional content while matching the full image’s visual-token grid. Across stages, the view mixture assigns increasing probability to the full image. A lightweight regional-advantage weighting reallocates token-level supervision using the crop-conditioned teacher-student log-probability gap. Evaluated under each sampled input, it applies mild reweighting when the gap is small and emphasizes higher-gap tokens when the gap widens. A Jensen-Shannon metric decomposition interprets this schedule as a shift from matched-input imitation toward the deployment objective. Across benchmarks spanning perception, visual mathematics and general multimodal question answering, PVD-full reaches an average accuracy of 77.51 over three seeds, improving on the reward-free distillation baseline by 2.01 points and on its reward-matched variant by 1.00 point. In the reward-free setting, PVD-distill still gains 1.16 points.
[CV-413] FoundDSR: A Generalizable Foundation Model with Guided 2D Gaussian Splatting for Depth Super-Resolution
链接: https://arxiv.org/abs/2609.32323
作者: Zhengxue Wang,Zhiqiang Yan,Yuan Wu,Guangwei Gao,Xiang Li,Jian Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We introduce FoundDSR, a generalizable foundation model for robust depth reconstruction across unseen data distributions using RGB-D pairs. FoundDSR begins with a guided 2D Gaussian Splatting strategy to model depth representations with Gaussian primitives. This strategy employs high-resolution RGB as prompts to optimize the Gaussian parameters, thereby encouraging each Gaussian primitive to anisotropically deform along high-frequency structural directions. The resulting Gaussian-upsampled representations are then mapped to high-resolution depth through an effective depth reconstruction branch. Furthermore, to mitigate training instability and bias toward dominant sources caused by distribution gaps in large-scale heterogeneous data, we introduce heterogeneous federated learning that allocates each data source to an independent client for local optimization and global aggregation. This design effectively endows FoundDSR with stable scalability to diverse and large-scale training data. Extensive zero-shot evaluations on synthetic, real-world, arbitrary-scale, and noisy conditions demonstrate that FoundDSR consistently outperforms existing state-of-the-art approaches, confirming its strong robustness and generalization to unknown scenes.
[CV-414] One Perception All Maneuvers: Directional Traffic Signal Understanding for Maneuver-Level Signal Intent Prediction
链接: https://arxiv.org/abs/2609.32316
作者: Ang Zou,Runzhe Zheng,Zhigang li,Zhen Yang,Han Xia,Xuewei Li,Zequn Qin,Xi Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Traffic lights are a key regulatory signal for autonomous driving at urban intersections, yet existing traffic signal perception is still predominantly formulated as instance-level detection or color recognition. Such formulations identify where traffic lights are and what colors they display, but leave a critical semantic gap before downstream planning: which ego maneuver is controlled by each visible signal and what dynamic permission the signal expresses for that maneuver. In this paper, we formulate Directional Traffic Signal Understanding, a decision-oriented task that predicts structured signal states for straight, left-turn, right-turn, and U-turn maneuvers from a front-view image. Each state contains the associated signal color and signal-implied passability. Based on OpenLane-V2, we provide a direction-level benchmark with maneuver-level supervision and metrics for color recognition, passability, full-frame consistency, and safety-critical errors. A direction-aware baseline combines global context, localized traffic-light evidence, and maneuver-specific representations. Experiments show that direction-level modeling improves passability prediction over image-level classifiers and detection-oriented pipelines, particularly at complex multi-signal intersections. The resulting representation provides a direct and interpretable traffic-signal interface for downstream planning together with topology, route, and surrounding-agent information. Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2609.32316 [cs.CV] (or arXiv:2609.32316v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.32316 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-415] HeroFrame-Bench: Reference-Anchored Evaluation via Rubric–Ranking Co-Evolution for Movie Hero Frame Selection
链接: https://arxiv.org/abs/2609.32280
作者: Weitai Kang,Hanieh Deilamsalehy,Yumo Xu,Dewang Sultania,Serdar Cellat,Yan Yan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 9 main pages
Abstract:Hero frames are in-film stills used as source imagery for theatrical posters, streaming cover art, film database listings, and other promotional placements. As the first visual entry point, they shape audiences’ initial impressions of the movie and their subsequent willingness to watch it. Selecting these frames, a task we term hero frame selection, requires balancing content relevance with aesthetic appeal. A related task is keyframe selection, yet its benchmarks prioritize relevance over aesthetics, using either finite annotations that exclude valid alternatives or VideoQA that entangles selection quality with downstream model capability. We therefore introduce HeroFrame-Bench, built through a scalable VLM-as-a-Judge framework. We construct multimodal contexts from diverse metadata to ground a VLM judge that scores selected frames using our Reference-anchored Percentile. The percentile is obtained by inserting each frame into reusable, pre-ranked reference chains, enabling direct, extensible, and reliable evaluation. To reduce ambiguity and improve consistency in these subjective judgements, we further propose Rubric-Ranking Co-Evolution, which generates movie-specific rubrics to condition the VLM judge and refines rubrics jointly with the resulting rankings. Within this process, we introduce several verifiable signals, most notably the Inverted Rubric Attack, to select robust rubrics. Finally, HeroFrame-Bench is instantiated over 204 movies with 2,031 reference chains and 1,970 learned rubrics. We build an annotation interface for human-alignment studies which show that our construction design improves VLM agreement with human from 77.56% to 83.78%. Evaluation on multiple methods show that hero frame selection remains challenging.
[CV-416] FSS-UBrain: Multi-region Few-Shot Brain Tumor MRI Segmentation
链接: https://arxiv.org/abs/2609.32273
作者: Truong Viet Vu,Nguyen Phuc Nguyen,Dang Thi Thu Hang,Tran Thien Thanh,Vo Nguyen Quoc Bao,Nguyen Thai Anh,Ngo Hoang Tu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: This work has been submitted to the Elsevier for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible
Abstract:Accurate delineation of whole tumor (WT), tumor core (TC), and enhancing tumor (ET) from multimodal magnetic resonance imaging remains challenging under limited annotation, cross-cohort variation, and severe target sparsity. We propose FSS-UBrain, a region-wise one-shot segmentation framework that uses a labeled positive support slice to condition binary query segmentation separately for WT, TC, and ET. Support-derived foreground and background descriptors guide query-feature adaptation, bottleneck interaction, decoder-side reconstruction, and boundary refinement. Episodic training additionally incorporates hard-negative and fully negative queries with empty-query regularization to suppress spurious foreground activation when the selected region is absent. Although inference operates on two-dimensional support–query slice pairs, checkpoint selection, threshold calibration, and final evaluation are performed after volumetric reconstruction. FSS-UBrain is evaluated on a held-out BraTS 2020 split and under target-supported cross-cohort protocols on BraTS 2023 and BraTS-Africa. Cases used as target support are excluded from the query cohorts, and no target-domain fine-tuning or test-time parameter updates are performed. On BraTS 2020, FSS-UBrain achieves volumetric Dice scores of 89.82%, 82.14%, and 77.42% for WT, TC, and ET, respectively, with corresponding 95th-percentile Hausdorff distance (HD95) values of 11.12, 9.01, and 4.46 mm. It also achieves the highest mean Dice and lowest finite-pair mean HD95 point estimates on BraTS 2023 and BraTS-Africa among the compared few-shot methods. These findings support target-conditioned few-shot segmentation while highlighting sensitivity to support selection and cohort-specific variation.
[CV-417] Learning Through Game: Skewed Transfer of Tabular Knowledge to Strengthen Image Model
链接: https://arxiv.org/abs/2609.32272
作者: Longfei Huang,Shangdong Yang,Yang Yang
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注:
Abstract:Multimodal tabular-image learning is gaining growing attention, yet it faces challenges due to tabular data unavailable at test time. A practical solution involves transferring tabular knowledge to images during training to enhance the performance of image models at inference. However, the overlooked yet important challenges lie in the modality imbalance between images and tables, as well as their asymmetric modality relationship in cross-modal transfer, which limits the auxiliary role of tabular data. To address these issues, we propose Skewed Knowledge Transfer (SKT), which asymmetrically transfers tabular knowledge to improve the image model by adaptive integration of modality gradients in a shared parameter space. Specifically, we first introduce a multimodal shared head, which allows the model to benefit from cross-modal structure without adding additional parameters. We then design a two-step Nash Bargaining strategy to effectively leverage tabular gradients. In the first step, SKT seeks a point of modality balance and uses preference awareness in the second step to steer combined gradients toward image-beneficial directions. Furthermore, we theoretically analyze the Pareto improvement and convergence of SKT. To this end, tabular knowledge is explicitly transferred to enhance image models. Empirical experiments on widely used tabular-image datasets reveal that SKT consistently improves image unimodal performance by using tabular data as auxiliary information.
[CV-418] RoboSTAR: Next-Scale Autoregressive Sign Language Translation for Humanoid Robots
链接: https://arxiv.org/abs/2609.32250
作者: Yujia Zeng,Chensheng Peng,Yuxin Chen,Alex Shao,Nathan Jew,Masayoshi Tomizuka
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Sign-language interpretation in public communication relies on qualified professional interpreters and can be difficult to scale, motivating robotic signing as a complementary accessibility interface. We present RoBoSTAR, a text-conditioned sign language production (SLP) framework for generating human-centric sign motion that can be retargeted for robotic execution, with speech supported optionally through an external ASR front end. Conventional autoregressive approaches flatten motion into a single full-resolution token sequence, forcing long-range and local dependencies to be modeled at a uniform temporal granularity. RoBoSTAR instead combines part-wise Finite Scalar Quantization with next-scale autoregression, generating motion over progressively finer temporal resolutions while predicting synchronized body and hand tokens in parallel within each step. This coarse-to-fine formulation provides compact long-range context before progressively refining motion details, while self-conditioning and context corruption improve robustness to cross-scale prediction errors. The generated motion is subsequently retargeted for physical humanoid execution. Extensive qualitative and quantitative evaluations are conducted to demonstrate the effectiveness of RoBoSTAR.
[CV-419] wo-Stage Multi-View Gait Recognition with a Re-Embedding Network
链接: https://arxiv.org/abs/2609.32244
作者: Long Hoang Le,Trung Thanh Ngo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Gait recognition always remains challenging due to severe overfitting and the rigid view constraints common in single-stage approaches. We propose a two-stage framework, termed Translate-First-Then-Reason (TFTR), to address these issues. In the first stage, a shallow Siamese convolutional network with triplet loss maps Gait Energy Images (GEIs) into a 128-dimensional view-specific embedding space. In the second stage, these per-view embeddings are treated as tokens and processed by a 12-layer Transformer encoder, which re-projects them into a new space with improved cosine separability. This design enables flexible fusion of an arbitrary number of views at inference, overcoming the fixed-input limitations of prior methods. Trained on the OU-MVLP dataset (6,000 subjects) and evaluated on unseen CASIA-B across normal, bag-carrying, and coat-wearing conditions, our pipeline achieves 96.91% single-view and 99.49% three-view accuracy on OU-MVLP, and attains 100% accuracy on CASIA-B with three views.
[CV-420] Residual Transferability in Neural Image Watermarking
链接: https://arxiv.org/abs/2609.32241
作者: Ziping Dong,Qi Li,Xinchao Wang
类目: Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint
Abstract:Neural image watermarks can be forged by extracting watermark-bearing residuals from released images and transferring them to unrelated content. While prior work has demonstrated this vulnerability, what makes these residuals transferable remains poorly understood. We formalize this vulnerability with \textbfresidual transferability (RT), a metric that quantifies how well watermark evidence remains decodable after transfer across unrelated images. Through comparative analyses and controlled interventions, we find that common training-side variations do not account for the large RT differences across watermarking systems; instead, architectural design plays a central role. By contrasting high- and low-RT systems and validating their architectural differences through controlled interventions, we identify two mechanisms that strengthen the dependence of watermark evidence on the cover image, thereby suppressing the residual transferability. These findings provide concrete design guidance for developing more forgery-resistant watermarking architectures. Complementarily, for existing watermarking systems where architectural redesign is impractical, we introduce \textbfCoverLock, a plug-and-play strategy for existing watermarking systems that strengthens such image dependence without architectural redesign. Across representative watermarking systems exhibiting high residual transferability, CoverLock achieves a more favorable security–robustness trade-off than both traditional handcrafted defenses and learned classifier-based defenses.
[CV-421] Federated Subspace Guided Vision-Language-Action Policy Distillation for Non-IID Multi-Robot Manipulation
链接: https://arxiv.org/abs/2609.32239
作者: Biprodip Pal,Kaushik Roy,Yanming Zhu,Brendan Tidd,Alan Wee-Chung Liew,Peyman Moghadam
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
备注: 9 Pages
Abstract:Federated learning offers a natural way for multiple robots to jointly improve manipulation policies without requiring centralized access to training demonstrations. However, non-IID task and environment distributions can induce representation drift and mutually incompatible robot-policy updates, making naive parameter aggregation destructive. We present FedDRMan, a federated subspace-guided distillation framework for heterogeneous robot manipulation. At each communication round, the server model provides a frozen teacher for local behavior cloning, while low-rank multimodal subspace and action-distribution distillation preserve globally useful representation geometry and policy behavior. To address heterogeneous aggregation, FedDRMan groups clients by update compatibility and maintains a persistent model for each cluster. The server then spectrally rebalances each compatible aggregate to mitigate attenuation of weaker task-relevant robot-policy update directions. Extensive experiments on LIBERO across diverse non-IID settings, heterogeneity levels, client participation variation, together with ablations and aggregation analyses, show that FedDRMan substantially improves knowledge transfer and consistently outperforms strong federated baselines achieving a peak mean success rate of 80.7%, 11.6 percentage points above the strongest evaluated federated baseline.
[CV-422] Skeletons in Flow: Graph Structured Flow Matching for Human Motion Prediction
链接: https://arxiv.org/abs/2609.32231
作者: Yixuan Wang,Brandon C. Fallin,Warren E. Dixon
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 22 pages, 3 figures
Abstract:Human motion prediction requires diverse future trajectories that remain consistent with observed motion and the articulated physical structure of the body. Skeletal constraints restrict individual poses, while coordinated motion depends on spatial interactions (between connected joints) and temporal interactions (between time instants). To facilitate human motion prediction in light of these constraints and interactions, we introduce Graph Structured Flow Matching (GSFM), which transports the complete future skeletal trajectory through a single conditional velocity field. The trajectory produces a spatiotemporal skeleton graph, and spatial and temporal attention couple its evolution according to skeletal relations and physical time offsets. Bone directions lie on unit spheres relative to a root joint, and tangent evolution preserves input bone lengths throughout generation. We train a learned velocity field through conditional flow matching along geodesic paths connecting random trajectories centered on the last-observed pose to recorded future trajectories. Experiments on the Archive of Motion capture As Surface Shapes (AMASS) dataset evaluate prediction accuracy, diversity calibration, and motion statistics. We demonstrate the contributions of spatial and temporal message passing in the developed architecture through an ablation study. GSFM models trained on AMASS also perform competitively on the Human3.6M skeleton without parameter updates or retraining, demonstrating applicability to an unseen skeletal structure.
[CV-423] Geometry-Preserving Blind Watermarking for Raw 3D Point Clouds
链接: https://arxiv.org/abs/2609.32222
作者: Rungui Zhou,Chuanzhi Zhou,Ruihuan Wang,Peng-Shuai Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Raw 3D point clouds are a core geometric representation. Establishing their ownership is challenging because point sets are irregular, unstructured, and frequently altered by resampling and geometric preprocessing. We present a blind watermarking framework that operates directly on xyz coordinates and supports both object-level shapes and scene-scale scans. At verification time, the embedded message is recovered from the observed point cloud alone, without access to the original point cloud, color, normals, or mesh connectivity. The method jointly learns watermark embedding and extraction through a feed-forward octree-based architecture, enabling efficient multi-scale geometric reasoning on large point sets. During training, a stochastic transformation layer exposes the decoder to common geometric perturbations, while progressive pose alignment improves robustness to pose changes. Experiments on object-level and scene-level benchmarks demonstrate reliable message recovery under common geometric processing while maintaining low geometric distortion. Qualitative comparisons further show that the learned perturbations are less visually conspicuous and less spatially structured than those of handcrafted alternatives. Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2609.32222 [cs.CV] (or arXiv:2609.32222v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.32222 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-424] Kernel-Based Steering of CLIP with Vision-Language Model Preferences
链接: https://arxiv.org/abs/2609.32203
作者: Sajjad Ghiasvand,Haniyeh Ehsani Oskouie,Sina Mansouri,Mahnoosh Alizadeh,Farzan Farnia,Ramtin Pedarsani
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Large vision-language models (VLMs) can judge visual similarity, but their judgments are not directly available as compact image embeddings for efficient comparison. We study how to transfer these preferences into CLIP while retaining its image–text capabilities. We introduce ASK, a kernel-based steering method that learns from elicited pairwise judgments without accessing teacher embeddings or collecting new human similarity annotations. ASK constructs positive semidefinite target kernels within small image groups and combines visual kernel matching with an image–text distributional anchor. Low-rank adapters jointly update the visual and text encoders while regularizing predictions toward frozen CLIP. After adaptation, retrieval uses CLIP image embeddings and cosine similarity, with no VLM calls. Experiments across five image domains, four CLIP backbones, and six judges evaluate teacher agreement, retrieval, and recognition retention. For ViT-B/16, mean retrieval mAP on classes excluded from adaptation increases from 53.8 to 75.0, compared with 71.7 for DINOv2 targets with KL anchoring. Mean zero-shot accuracy with jointly adapted encoders increases from 61.8% to 62.4%, averaged over 12 benchmarks and the five adaptation domains. Prompting provides an additional capability: selecting which visual distinctions the student learns. Human-annotated evaluations across four datasets support this criterion-specific control.
[CV-425] Devol-ONE: One Autoregressive Mixture of Transformers to Unify Vision-Language-Action and Latent World Modeling
链接: https://arxiv.org/abs/2609.32193
作者: Hongyi Cai,Yi Herng Ong,Tingshiuan C. Wu,Lim Chiew Hui,Hanxia Li,Kehong Guo,Sze Yuan Cheong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision Language Action (VLA) models condition actions directly on current visual and language context, without an explicit account of how the scene evolves under candidate actions. World Action Models (WAM) attempt to address this limitation by predicting future states, but existing designs keep prediction and policy learning architecturally separate, connecting them only through the predicted output, whether through pixel space video generation or a latent forecasting module trained independently of the policy. We present Devol-ONE, a Mixture of Transformers architecture that unifies vision language understanding, latent world dynamics prediction, and action generation within a single autoregressive framework. Instead of encoding vision language tokens once and feeding them to the action expert, Devol-ONE runs autoregressive prediction jointly across a vision language stream and a V-JEPA pretrained dynamics stream, attending to the vision language key-value cache at every layer to forecast future latent states under language guidance. The action expert is in turn shaped continuously by semantic reasoning and predicted physical dynamics rather than by a fixed representation computed in advance. Extensive experiments are conducted on LIBERO, LIBERO-PLUS, RoboTwin2.0 along with real-world evaluation on Flexiv single-arm and dual-arm setups. Ablation studies show the effectiveness of dynamic stream prediction and layer-wise unified attention to validate our model architectural coherency.
[CV-426] Evaluating Single and Multi-Omics Based Explainable Artificial Intelligence (MOXAI) for Molecular Subclass Classification of Adult-Type Diffuse Gliomas
链接: https://arxiv.org/abs/2609.32190
作者: Md Zahangir Alom,Quynh T. Tran,Breuer Alexandar,Brent A. Orr
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 8 pages, 6 figures
Abstract:DNA methylation (DNAM) profiling has emerged as a powerful diagnostic tool for classifying brain and solid tumors. However, existing computational models typically analyze methylation and copy number variation (CNV) data separately, failing to capture the complementary information their integration could provide. Moreover, current classification models lack mechanisms for within-class risk assessment analogous to traditional tumor grading, and no established explainability method can attribute classification decisions to specific genomic loci. In this paper, we present MOXAI (Multi-Omics Based Explainable AI), a deep learning framework that integrates DNA methylation and copy number data from methylation arrays to classify molecular subtypes of adult-type diffuse gliomas, alongside single-modality variants for comparison. Using a cohort from The Cancer Genome Atlas (TCGA), we trained ResNet50, DINOv2, and Graph Attention Network (GAT) models on methylation data alone, copy number data alone, and combined multimodal data. We further developed explainable AI (XAI) methods based on class activation maps (CAMs) and gradient-weighted CAM (Grad-CAM) to identify the specific CpG sites, genes, and chromosomal regions most relevant to each classification decision. The multimodal model achieved up to 92.98% cross-validation accuracy, outperforming models trained on CNV data alone. DINOv2 showed the strongest generalization, reaching 94.25% accuracy (confidence 0.9) on independent validation sets. XAI results aligned with established molecular features of adult-type diffuse glioma subtypes, confirming the biological interpretability of the framework.
[CV-427] Presence Is Not Faithfulness: Figurative Vehicle Intrusion in Text-to-Image Generation ICLR2027
链接: https://arxiv.org/abs/2609.32188
作者: Xiaoyu Ma,Chen Yang,Hao Chen
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: submitted to ICLR 2027
Abstract:Text-to-image (TTI) models increasingly generate high-quality images from natural-language prompts, yet figurative language exposes a failure: a vehicle that should guide the depiction of a tenor may instead be rendered as a visible object. We call this failure Figurative Vehicle Intrusion: the intruding content is textually licensed, but it is assigned the wrong visual role, showing that visual presence is not always faithfulness and that presence-oriented evaluation can miss such errors. To study it systematically, we introduce Vehicle Intrusion and Semantic Tenor Assessment (VISTA), a multilingual benchmark of figurative prompts organized by Figurative Form and Mapping Mechanism. We further propose V-Score, a diagnostic question-answering metric that evaluates role-aware figurative faithfulness in generated images. Evaluations on recent high-performing TTI models show that vehicle intrusion persists across languages and figurative categories. As a lightweight mitigation, we introduce VISTA-Guard, which partially reduces vehicle intrusion and suggests a practical path toward more figuratively faithful TTI generation. All resources will be released publicly.
[CV-428] Contamination Prior or Evidence? Decomposing and Training Evidence Use in Whole-Slide Vision-Language Models
链接: https://arxiv.org/abs/2609.32185
作者: Wenhao Zhang,Zhongliang Zhou,Shiyuan Zhang,Yiqing Yang,Pinqiao Wang,Lehan Yang,Hanyin Wang,John Kang,Sheng Li
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 19 pages, 5 figures, 6 tables
Abstract:Pathology vision-language models (VLMs) are conventionally evaluated by accuracy, but accuracy alone does not measure evidence use: it may conflate dataset contamination, prior knowledge, and image evidence. In a motivating study of lymph-node metastasis prediction, we found that most public pathology VLMs showed minimal differences when changing from feeding the models with whole-slide images, an annotated lesion, or no image at all. To better understand the specific features leveraged by these models, this paper presents two contributions aimed at disentangling these factors. First, we present CleanSlide, a TCGA-based VQA benchmark designed to eliminate image- and question-side contamination. It contains 149K audited multiple-choice questions over 9,985 slides, with patient- and tissue-source-disjoint splits. Every question is audited for option shortcuts, stem leakage, cross-split duplication, and blind solvability. Second, we propose Pair-DPO, a preference loss over counterfactual slide pairs from the same question and source. By controlling for shared confounding factors, Pair-DPO cancels out the question-attributable signal and leaves image evidence as the source of preference. Specifically, each pair consists of two real slides with opposite, verified findings, introducing neither editing artifacts nor unverified labels for diffuse or graded features such as invasion, necrosis, and tumor grade. Experiments show that our method gains 15.29% from image evidence on the CleanSlide, compared with 2.81% for the best published model. On the external CPTAC and BCNB cohorts, our method achieves accuracies of 57.6% and 59.0%, outperforming all other evaluated models by 9.7% and 3.4%, respectively. We will release the benchmark and code.
[CV-429] Scalable In-Domain Self-Supervised Foundation Model for Dense Representation Transfer in High-Resolution Plant Imaging
链接: https://arxiv.org/abs/2609.32183
作者: Junlin Guo,Sharmin Majumder,Isaac Lyngaas,John Lagergren,Xiao Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注:
Abstract:High-resolution plant imaging enables detailed characterization of plant morphology, but dense scientific analysis remains limited by costly pixel-level annotations, large image pixel dimensions, and substantial variation in imaging conditions. This work proposes a scalable in-domain self-supervised pretrained foundation model for high-resolution, high-pixel-dimension multi-species plant imagery. A masked autoencoder with a ViT backbone is pretrained on more than 10 million multi-view plant image tiles using distributed training. Following scalable pretraining, the learned foundation-model representations are comprehensively benchmarked across fine-grained dense prediction and coarse global feature recognition, with particular emphasis on limited supervision and realistic downstream imaging conditions. This work focuses on the domain gap of existing foundation models in dense feature representation and transfer. Through extensive experiments involving limited annotations, cross-view variation, and resolution degradation, the in-domain FM achieves a Mean Dice of 0.8686 and a Pooled Dice of 0.8959, outperforming an MAE counterpart pretrained on large-scale natural-image data by 0.0694 and 0.0613, respectively. The results further indicate that increasing pretraining scale produces consistent improvements in dense feature transfer. Overall, these findings suggest that scaling in-domain self-supervised pretraining can reduce the domain gap and improve transferable dense representations for high-pixel-dimension scientific imaging.
[CV-430] KeyRec: Bounded Visual Memory for Streaming and Long-Video Understanding
链接: https://arxiv.org/abs/2609.32182
作者: Zihan Chen,Xuejian Rong,Xiaojuan Wang,Boqing Gong,Adi Zicher,Yael Pritch,Nikhil Karnad
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision-language models are increasingly used to understand long videos and continuous streams. However, dense visual tokens accumulate with video duration, making long-context inference prohibitively expensive. Existing training-free visual-token selection methods reduce this cost by retaining informative tokens, but may lose coherent event evidence and fail to distinguish detailed recent observations from long-range history. We propose \textbfKeyRec, a training-free framework for constructing bounded visual memory. During query-agnostic writing, KeyRec preserves fine-grained recent observations in a visual cache and organizes historical evidence into a structured event bank. Candidate events are proposed according to their novelty relative to previously stored events and maintained through an online add–merge–evict update. When a question arrives, a text-only router adaptively allocates a fixed readout budget between recent and event memory, without reprocessing historical frames. KeyRec operates on model-facing visual embeddings and supports both modular encoder–projector VLMs and the encoder- and projector-free NEO-ov architecture. Across four streaming and long-video benchmarks and three VLM backbones, KeyRec achieves the best compressed performance in 13 of 15 settings using only 10% of the dense decoder-facing visual-token budget. It outperforms the strongest compressed baseline by 2.21–18.37 points on real-time questions, achieves the best compressed result in five of six long-video settings, and performs best in every NEO-ov 2B setting.
[CV-431] Binaural Audio-Visual Instance Segmentation
链接: https://arxiv.org/abs/2609.32180
作者: Saijun Wang,Guanfeng Tang,Hongbo Zhao,Zhicheng Lei,Yutong Zhang,Wei Ye,Rui Fan
类目: Computer Vision and Pattern Recognition (cs.CV); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注:
Abstract:Audio-visual segmentation (AVS) aims to segment sounding objects at the pixel level by integrating auditory and visual cues. However, existing methods are predominantly developed under the monaural setting and primarily rely on cross-modal semantic correspondence, which limits their ability to distinguish visually similar instances of the same semantic class. In contrast, humans naturally exploit binaural hearing, where interaural differences and direction-dependent acoustic filtering introduced by the head and pinnae provide physically grounded spatial cues for accurate sound source localization. Motivated by this observation, we introduce binaural audio-visual instance segmentation (BiAVIS), a new task that leverages synchronized binaural audio and video frames to segment sounding instances. To advance research on this task, we establish two benchmarks by manually annotating an existing binaural audio-visual dataset and collecting a new real-world dataset, BiAVIS-Bench, in more challenging and diverse scenarios. We further propose a BiAVIS model, which leverages an audio-only sound source localization network to learn spatial and semantic priors for sounding instances from binaural audio. A query-level audio-visual fusion strategy is subsequently introduced to inject these informative priors into the instance segmentation decoder. Extensive experiments conducted on the two proposed benchmarks demonstrate the superior performance of the BiAVIS model over previous monaural AVS methods, especially in resolving instance-level intra-class ambiguity. On the more challenging BiAVIS-Bench, the proposed BiAVIS model outperforms the best-performing monaural baselines by 17.22% in mAP and 7.71% in FSLA, respectively.
[CV-432] Federated 3D Gaussian Splatting for Large-Scale Scene Reconstruction at Wireless Edge
链接: https://arxiv.org/abs/2609.32177
作者: Guanlin Wu,Chao Hu,Pu Chen,Juyong Zhang,Han Hu,Shuguang Cui,Jie Xu
类目: Computer Vision and Pattern Recognition (cs.CV); Distributed, Parallel, and Cluster Computing (cs.DC)
备注: 16 pages, 10 figures, 6 tables. Accepted for publication
Abstract:Three-dimensional (3D) Gaussian splatting (3D-GS) has emerged as a promising technique for large-scale scene reconstruction due to its high rendering efficiency and fidelity. However, the training of large-scale 3D-GS models at wireless edge faces various technical challenges including the limited communication, computation, and graphics processing unit (GPU) memory resources at edge devices, the structural inconsistency issue across local models hindering their effective aggregation, as well as privacy leakage risks associated with raw visual content and camera parameters. To address these challenges, this paper proposes a novel resource-efficient federated learning framework for efficiently training 3D-GS models of large scenes under severe resource constraints. First, we propose an on-device model lightweighting mechanism that adaptively selects and prunes Gaussian points to balance the rendering quality and training efficiency. In this mechanism, we quantitatively evaluate the importance of different Gaussian points at each device to facilitate the pruning, and use a novel importance-to-latency ratio criterion to determine the number of pruned Gaussian points under GPU memory and computation/communication latency constraints. Furthermore, we develop a 3D-GS model recovery mechanism that restores structural consistency across local 3D-GS models without accessing private camera parameters, enabling their effective aggregation towards a global model. Finally, extensive experiments show that our approach significantly accelerates convergence, maintains high rendering quality, and reduces training latency compared to state-of-the-art federated 3D-GS baselines.
[CV-433] OneFixer: High-Quality and Consistent One-Step Autoregressive 3DGS Refinement for Driving Scenes
链接: https://arxiv.org/abs/2609.32175
作者: Boseong Jeon,Junhyeop Lee,Juhan Cha,Hayoung Kim
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Autoregressive video diffusion is a promising render-time fixer for 3D Gaussian Splatting (3DGS) in autonomous-driving simulation, but deployment demands high visual quality and temporal consistency at low latency. This is especially hard for one-step causal generation, where each imperfect prediction immediately becomes context for subsequent frames. Existing approaches stabilize rollouts through staged training with multiple modules and rollout-aware regularization, yet one-step quality still falls short of what deployment requires. We introduce OneFixer, a one-step autoregressive video-diffusion fixer trained in a single task-specific adaptation stage. Our key idea is a deployment-matched shared rollout: the model’s own one-step predictions serve as the causal context for flow matching, exposing training to deployment-time errors, while the same rollout receives direct pixel-space perceptual supervision to preserve fine detail. Because the predictions optimized for current-frame quality are exactly those reused as future context, fidelity and autoregressive robustness are learned jointly, without bidirectional-to-causal conversion or teacher-student distillation. OneFixer further exploits cues that driving simulation readily provides, lane geometry and dynamic-agent states, to improve geometric fidelity. On Waymo and proprietary driving scenes with 900-frame rollouts, OneFixer achieves the lowest FVD, LPIPS, and DISTS among all baselines at one step, with temporal consistency matching or exceeding multi-stage DMD pipelines. Under identical backbone and conditioning, it matches a multi-stage DMD-with-Self-Forcing pipeline in under half the GPU-hours and keeps improving beyond its plateau. In closed-loop simulation with a driving policy, OneFixer reduces the collision rate by a third relative to raw 3DGS rendering. Project page: this https URL
[CV-434] PQR3D: Progressive Query Refinement over Reference-Conditioned Temporal Windows for Multi-View 3D Object Detection
链接: https://arxiv.org/abs/2609.32163
作者: Hui Ye,Yudong Liu,Yiran Chen,Rajshekhar Sunderraman,Shihao Ji
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Temporal context is essential for camera-only multi-view 3D object detection. Existing streaming detectors maintain and propagate query states from one frame to the next, requiring sequence-aware training and chronological inference. We propose PQR3D, which performs progressive query refinement within referenceconditioned temporal windows. This design enables random frame sampling and independent inference without persistent query memory. Within each window, PQR3D progressively transfers motion-aligned high-confidence queries from earlier timestamps toward the target frame. We further introduce masked selfattention to regulate interactions among regular, propagated, and denoising queries while keeping denoising supervision isolated from detection queries. In addition, a stage-decoupled anchor embedding injects position before self-attention and size, orientation, and velocity afterward, reducing interference from temporally inconsistent attributes. With a ViT-L backbone, PQR3D sets a new state of the art on the nuScenes test set, achieving 71.6 NDS and 64.9 mAP. Source code is available at this https URL
[CV-435] PruneForget: Joint Unlearning and Pruning of Vision Models
链接: https://arxiv.org/abs/2609.32162
作者: Yu-Shan Tai,Amber Yijia Zheng,Raymond A. Yeh
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Machine unlearning and model pruning are increasingly coupled in the real world. Models must support unlearning requests, e.g., for safety concerns, while also meeting requirements in latency and memory budget. Until recently, existing works have studied each aspect as an independent problem, e.g., running unlearning and pruning sequentially. In this work, we show that unlearning and pruning are naturally aligned and should be solved jointly to be made aware of each other. Intuitively, parameters that encode information of the unlearned samples are natural pruning targets, as unlearning and pruning both call for the “deletion” of such parameters. We propose PruneForget, a method that uses the unlearn set as a guide for pruning, so that unlearning and pruning mutually benefit each other. Extensive experiments on image classifiers and generative models show that PruneForget removes the influence of the unlearned samples while producing a more compact model with reduced inference cost. It achieves a negligible performance gap relative to an oracle that retrains from scratch for unlearning and then prunes.
[CV-436] CausalDriveBench: Evaluating Causal Reasoning in Vision-Language-Action Models for Autonomous Driving
链接: https://arxiv.org/abs/2609.32157
作者: Narendiran Chembu,Navvrat Rao,Shreedhar Shreeshail Kodate,Gayatri Srujana Banda,Arko Sarkar,Abhinav Khanna,Rajarshee Das,Umesh Kanala,Siddarth Khandelwal,Kumar Aman,Aish Dubey,Kaustubh Beedkar,Arjun Jain
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Vision-Language-Action (VLA) models for autonomous driving produce natural-language reasoning alongside predicted trajectories, but whether this reasoning reflects the causal structure of the scene remains untested. We introduce CausalDriveBench, an evaluation framework grounded in Pearl’s Causal Hierarchy (PCH) that tests causal reasoning in driving-specific VLAs through structured visual question answering (QA) and alternative-trajectory prediction. To this end, we construct causal scene graphs over nuScenes that distinguish causally active, dormant, and distractor entities, separating perceptual salience from causal relevance. The benchmark spans all four rungs of PCH (association, intervention, and counterfactual along with causal discovery) for QA generation. For the higher rungs, we additionally provide reference trajectories under specified scene modifications, enabling action-level verification that complements reasoning-level evaluation. In total, the benchmark contains 7,285 verified causal QA pairs and 1,000 counterfactual trajectories derived from nuScenes. We evaluate 10 driving-specific VLAs and 3 general-purpose VLMs, and report three findings. First, the best model reaches only 70.6% QA accuracy, and 4 of 13 models score below random chance. Second, comparing each driving VLA to the general-purpose VLM that shares its language backbone, the cost of driving fine-tuning ranges from 2 to 34 percentage points on causal QA, with post-training design explaining the spread. Third, causal QA and trajectory accuracy are statistically uncorrelated across models: under counterfactual prompts, predicted trajectories either over-react or collapse onto the observed-scene baseline. Taken together, these results show that neither fluent rationales nor accurate observed-scene trajectories constitute evidence of causal understanding.
[CV-437] AquaBEV-Nav: Learned BEV Occupancy for Underwater Navigation and Exploration
链接: https://arxiv.org/abs/2609.32156
作者: Trung Tien Dong,Zhenqi Wu,Sahasra Kondapalli,Jiayi Wu,Yi Sheng,Xiaomin Lin
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Safe underwater exploration requires a robot to understand where surrounding structures are located and which regions are available for motion. Existing vision-based underwater exploration systems commonly obtain this information indirectly by estimating monocular depth, unprojecting the geometry into 3D space, and accumulating it into a 2D bird’s-eye-view occupancy map. This reliance on intermediate depth estimation is particularly problematic underwater, where scattering and wavelength-dependent attenuation degrade visual cues and limit the reliability of monocular depth estimates. We introduce AquaBEV-Nav, an underwater exploration framework that bypasses explicit monocular depth estimation through direct bird’s-eye-view occupancy prediction. Built upon the CORAL hierarchical exploration framework, AquaBEV-Nav replaces its depth-based perception front end with AquaBEV. Given a single RGB frame, AquaBEV maps visual features into a learned polar representation, performs causal reasoning along the range dimension, and reconstructs local Cartesian occupancy without relying on intermediate depth prediction. The resulting occupancy map is accumulated into CORAL’s persistent spatial memory, providing spatial context for VLM-based high-level planning and collision constraints for dynamics-aware local trajectory generation. Across ten simulated reef environments and six occupancy backbones evaluated under a single protocol, AquaBEV-Nav reaches 37.48 structure IoU and 53.2 target IoU, 88.95% closed-loop coverage with zero collisions.
[CV-438] Gaussian Image Steganography via Parameter-Domain Keyed Embeddings SIGGRAPH
链接: https://arxiv.org/abs/2609.32131
作者: Tong Wu,Runze Cheng,Xiaoyue Fan,Kaan Akşit
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 4 pages, 2 figures, 3 tables; 2-page supplementary material in ancillary files. Accepted to SIGGRAPH Asia 2026 Technical Communications
Abstract:2D Gaussian-based image representation is becoming increasingly popular, and our work proposes a new approach to steganography by embedding information within Gaussian parameters rather than image pixels. We first fit the parameters of this representation to the target image and employ a secret key to select a subset of these parameters for fine-tuning, allowing us to embed an 8-bit message while maintaining high visual fidelity in the reconstructed image. Thirty fitting experiments on three synthetic images show that the correct key can recover the message without error, while decoding with incorrect keys yields a Bit Error Rate (BER) of 0.543 , close to random guessing. Compared with random selection using the key, selecting the least-disturbing edits recovers the message more reliably (one-sided p=0.031 ), and the average PSNR cost is only 0.091 dB in visual quality. The embedding method transfers to 112 natural images at 256\times256 using 4,096 Gaussians. The correct-key BER is 0.000 , and decoding under a wrong key stays close to random guessing at 0.520 . Our method embeds the payload through three Gaussian parameter types: log-anisotropy, opacity, and color luminance. In a separate nine-fit reduced setting, color luminance is removed, so the payload uses two instead of three parameter types, a 33.3% reduction; all 256 Gaussians remain in the fitted representation. The correct key still recovers the message without error. However, wrong-key BER rises from 0.514 to 0.571 , moving farther from random guessing ( 0.5 ).
[CV-439] ReFM: Semantic-Aware Refinement Flow Model for Motion Retargeting
链接: https://arxiv.org/abs/2609.32068
作者: Jingxiang Qu,Lucie Taglienti,Evan Atherton
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Graphics (cs.GR)
备注:
Abstract:Motion retargeting transfers motion across characters with different skeletal structures while preserving semantic intent and physical plausibility. Despite recent progress, two fundamental questions remain: (i) how can reliable source-motion semantics be learned without high-quality paired retargeting data, and (ii) how should retargeting be formulated when no reliable paired motion can serve as a definitive regression objective? Existing methods commonly preserve semantics by constraining predictions toward copied motions. However, such initializations entangle useful articulation cues with artifacts caused by mismatched skeletal proportions and body geometry. Moreover, directly regressing a final motion in one forward pass is restrictive because retargeting is inherently underdetermined, and the desired solution must balance semantic fidelity with target-specific physical and temporal constraints rather than match a unique paired target. Motivated by these limitations, we propose ReFM, a source-mesh-agnostic, energy-guided model that reformulates motion retargeting as progressive refinement. First, an SO(3) canonicalizer removes redundant global-orientation variations. Second, a cross-character semantic encoder, pretrained through contrastive learning, provides a character-invariant representation for both optimization guidance and semantic evaluation. ReFM then progressively refines an initialized target motion through a learned flow guided by semantic consistency, physical plausibility, temporal coherence, and minimal motion modification. The framework is compatible with different initialization strategies, including both direct motion copying and Autodesk HumanIK, an industry-standard full-body inverse-kinematics retargeting system.
[CV-440] ControlGS: Conditioning Neural Gaussians for Downstream-Processing-Aware XR Rendering SIGGRAPH
链接: https://arxiv.org/abs/2609.32038
作者: Weikai Lin,Junjie Zhao,Carl Marshall,Sushant Kondguli,Yuhao Zhu
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: Accepted to Siggraph Asia’26, 36 pages, poject page: this https URL
Abstract:Extended Reality (XR) users do not directly perceive the output of a rendering engine. Instead, rendered images pass through a post-processing pipeline and the physical display-optics path before reaching the eye. Critically, the exact downstream processing can vary significantly at run time, influenced by, for instance, camera pose and display power budget. Traditional 3DGS methods either implicitly assume that this downstream pipeline preserves image quality or cannot adapt to downstream processing changes. To bridge this gap, we present ControlGS, an XR Gaussian rendering pipeline that optimizes end-to-end visual quality. ControlGS models and integrates the entire downstream processing, between the rendering output and the human eye, into the optimization objective. To adapt to downstream processing at run time, ControlGS dynamically generates Gaussian primitives conditioned upon the downstream processing parameters. Experiments show that ControlGS consistently improves end-to-end post-optics XR quality across different neural Gaussian backbones and datasets, with minimal overhead. Code is available at this https URL.
[CV-441] ScreenHaystack: Finding Blind Zones in GUI Grounding EMNLP2026
链接: https://arxiv.org/abs/2609.32036
作者: Chenyue Li,Xiaoxiao Sun,Yubo Deng,Qinlin Zhao,Serena Yeung-Levy,Yuhui Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to EMNLP 2026 Main Conference
Abstract:We introduce ScreenHaystack, a dynamic needle-in-a-haystack benchmark for evaluating spatial reliability in GUI grounding. Instead of testing each target at a fixed position, ScreenHaystack systematically relocates controlled target icons across high-resolution GUI backgrounds and measures whether models can localize them consistently. Using this benchmark, we find that leading GUI grounding models, including Qwen3-VL, UI-TARS, GTA, and UI-Venus, exhibit blind zones: spatial regions where grounding accuracy drops sharply despite fixed target appearance and instruction. These blind zones transfer to unseen ScreenSpot-Pro examples: targets inside blind zones are consistently harder to ground, with Qwen3-VL-8B dropping by 16.1 percentage points, and controlled relocation shows that moving targets into blind zones decreases accuracy while moving them out improves accuracy. We further show through controlled synthetic experiments that uneven spatial coverage in training data can induce such blind zones. Therefore, we propose a simple strategy, blind-zone-oriented augmentation, which adds supervision in blind zones and improves ScreenSpot-Pro accuracy over both original and randomly augmented Click-100k fine-tuning.
[CV-442] Depth Any Seen: Which Surfaces and How Far?
链接: https://arxiv.org/abs/2609.32027
作者: Xiaohao Xu,Xiaonan Huang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Graphics (cs.GR); Robotics (cs.RO)
备注: 54 pages, 31 figures, including appendix. Video demo: this https URL
Abstract:When several surfaces are visible along a ray, recovering visible 3D structure from one image requires jointly estimating their presence and metric depth. Depth Any Seen represents these surfaces as image-conditioned multi-Bernoulli depth sets, whose components each contribute one depth or remain absent. Its auxiliary-free Exact Multi-Bernoulli objective (ExactMB) learns depth and presence by marginalizing one-to-one assignments to complete, distinct targets. Our analysis shows that matching expected count can leave component-surface assignment unresolved. We extend real and synthetic layered-depth benchmarks to evaluate depth accuracy, recovered support, and overprediction. Compared to depth stacking, ExactMB reduces overprediction by a relative 88.2% on LD-Real and 80.5% on MD-3K while retaining most ordinal accuracy, with comparable conditional metric-depth error on LD-Syn. Further ablation studies show that ordered assignment improves depth-accurate recall and precision over marginalization, whereas the count-regularized configuration achieves higher deeper-rank precision than ordered assignment at lower recall. Our code will be publicly released.
[CV-443] riO: Tri-Modal Unsupervised Occupancy World Model for Anything Perception ECCV2026
链接: https://arxiv.org/abs/2609.32013
作者: Quinlan Sykora,Sourav Biswas,Christopher Diehl,Andrew Cunningham,Thomas Gilles,Raquel Urtasun
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Published at ECCV 2026, 49 pages, 20 figures
Abstract:We present TriO, a multi-modal unsupervised world model that predicts 4D occupancy, obstacle segmentation, flow and LiDAR. In contrast to prior work, TriO utilizes three distinct sensor modalities (camera, LiDAR, and RADAR) as both inputs and sources of self-supervision, eliminating the need for additional human annotations. Thanks to its novel supervision, the model is able to segment any occupancy from the drivable surface, overcoming the limitations of existing open-set methods in handling long-tail objects. TriO achieves state-of-the-art results in multiple 3D and 4D tasks, including occupancy, flow, and LiDAR prediction, as well as zero-shot road obstacle segmentation across multiple datasets such as Argoverse 2, and Spotting the Unexpected.
[CV-444] ype-Balanced Federated Learning for Visual Analog Meter Reading
链接: https://arxiv.org/abs/2609.31998
作者: Weida Zhao,Logan Bellamy,Yazhou Tu,Jiaqi Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Analog dial meters are widely deployed in industrial applications and utility sites, where environments and meter types vary and inspection data may be sensitive. Currently, automatic meter readers must be individually developed and deployed for each environment and meter type in practice. Deep learning could handle this variability but requires diverse labeled data that are costly to collect and update. In practice, meter images are distributed across independent sites, each with limited labels, while raw images often cannot be pooled because of ownership, governance, or privacy constraints. To address these challenges, we present a federated framework for visual analog meter reading that enables multiple sites to collaboratively train a reading model without sharing their raw images. Our framework consists of a four-stage pipeline: (1) dial localization, (2) thin-structure segmentation trained federatively across clients, (3) polar unwrapping, and (4) tick-counting decoding for final reading. To enable systematic evaluation of this setting, we release MeterFL, a 1,382-image mask-annotated dataset organized into deployment-motivated pseudo-clients derived from visual attributes via deterministic rules, with dHash near-duplicate control between the segmentation train and test splits. We evaluate both segmentation quality and end-to-end reading accuracy. MeterFL is publicly available at this https URL.
[CV-445] Does Vision-Language Pretraining Granularity Matter? A Controlled Evaluation of Vision-Language Objectives Across Chest X-Ray Interpretation Tasks
链接: https://arxiv.org/abs/2609.31985
作者: Denis Musinguzi,Andrew Katumba,Prasenjit Mitra
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注:
Abstract:Vision-language pretraining objectives differ in the spatial granularity of their supervision, yet the implications of this distribution for chest X-ray interpretation remain underexplored. We present a controlled study that isolates the pretraining objective: holding the encoder and pretraining data fixed, we train nine objectives spanning global and local contrastive learning, captioning, and their combinations, and evaluate across five chest X-ray tasks of increasing spatial granularity. We find that (i) pretraining granularity aligns with task granularity at the extremes, with local objectives leading on abnormality detection and global objectives on classification; (ii) local objectives are surprisingly competitive on global-level generation and question answering tasks; (iii) the merits of captioning and contrastive learning reverse across granularity levels; and (iv) among combinations, mixing captioning and contrastive supervision is strongest on classification and in distribution generation, while pairing two captioning objectives generalizes best on zero-shot report generation. These results show that no single objective is universally optimal, and that the interaction of objective type, granularity, and task governs downstream performance.
[CV-446] Double-Edged Sword of Mediated Visibility: How Visual Framing Undermines Congresswomens Perceived Competence
链接: https://arxiv.org/abs/2609.31970
作者: Bryce J. Dietrich(Purdue University),Hyein Ko(Case Western Reserve University),Myriam Shiran(The Ohio State University)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 36 pages, 3 figures, plus 25-page online appendix (included). Under review
Abstract:Women’s presence in Congress has stalled below 30%. That institutional underrepresentation is mirrored by their limited visibility on televised news, where visual presentation can shape perceived competence and authority. Yet research tells us more about whether congresswomen appear than how they are visually framed. Facial recognition analysis of 695,464 image-text segments from CNN and Fox News (2011-2021) reveals that congresswomen appear disproportionately in split-screen rather than solo shots. Two pre-registered experiments with 6,220 participants show that, in static frames, split-screen framing reduces congresswomen’s perceived political competence, but not congressmen’s. In dynamic videos, the competence penalty disappears; instead, outraged language reduces a congresswoman’s perceived warmth roughly three times as much as a congressman’s. Suggestive evidence (p = .057) indicates that women viewers report greater external political efficacy after watching a congresswoman appear alone. We conclude that visual framing shapes both congresswomen’s mediated visibility and their perceived capabilities.
[CV-447] CaptchaArena: A Large-Scale Fine-Grained Dataset for Training Computer-Use Agents on Interactive CAPTCHAs
链接: https://arxiv.org/abs/2609.31957
作者: Zhenhao Zhang,Zhaoyu Fan,Haohan Ying,Jingwen Hu,Hancen Fan,Junhao Zhou,Zitian Chen,Linchao Zhu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 30 pages, 4 figures, 16 tables. Code and data: this https URL
Abstract:Interactive CAPTCHAs remain challenging for computer-use agents, while existing datasets face trade-offs among type coverage, interaction fidelity, and trajectory supervision. To address these gaps, we present CaptchaArena, the first large-scale, fine-grained training dataset for interactive CAPTCHA solving. It contains 50K puzzles across 20 CAPTCHA types and 5 interaction modes, with every solution verified through execution. CaptchaArena provides 50K screenshot-action trajectories, including 46K with step-by-step reasoning annotations. It also includes fine-grained pixel-mask annotations for irregular targets. Using CaptchaArena, we train CaptchaAgent, a single 9B policy for all 20 CAPTCHA types, with supervised fine-tuning followed by reinforcement learning. The environment verifier directly provides the RL reward. Supervised fine-tuning reaches 70.5 Pass@1, and reinforcement learning further improves it to 71.7, while also improving performance on two external benchmarks. These results demonstrate the value of large-scale, fine-grained computer-use supervision for training interactive CAPTCHA agents. We release CaptchaArena and CaptchaAgent at this https URL.
[CV-448] Resource-Aware Federated Mixture-of-Experts with Adaptive Pruning for Onboard Learning in LEO Satellite Constellations
链接: https://arxiv.org/abs/2609.31932
作者: Mohamed Shaaban,Mohamed Elmahallawy,Marius Bernahrndt,Tobias Hecking
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Low-Earth-orbit (LEO) satellites are increasingly expected to perform onboard learning for applications such as disaster response and environmental monitoring. However, conventional federated learning (FL) is ill-suited to onboard satellite learning, as it assumes computational, memory, and communication resources beyond the capabilities of resource-constrained LEO platforms, often necessitating the transmission of raw imagery to ground stations. We present COSMIC-FL, a resource-aware FL framework for efficient onboard learning in LEO satellite constellations. COSMIC-FL introduces two complementary Mixture-of-Experts (MoE) architectures: a Sliced design that shares backbone representations while activating task-specific channel subsets, and a Modular design that employs lightweight gating to route inputs to physically separated expert networks. A semantic class-to-expert mapping enables each satellite to train, update, and communicate only the expert paths relevant to its local data. To further improve efficiency, COSMIC-FL integrates staged optimization with three structured pruning strategies: server-side pruning, client-side fixed-ratio pruning with mean-vote aggregation, and adaptive client-side per-layer pruning based on aggregated importance and a MAD-based gap criterion. Combined with semantic expert routing, these techniques jointly adapt computation and model sparsity to both data semantics and layer importance, yielding a favourable accuracy–efficiency trade-off for heterogeneous space platforms. Experiments on six image classification benchmarks under highly non-i.i.d. settings show that COSMIC-FL maintains competitive accuracy while reducing communication, computation, and energy consumption by up to 80% over SOTA FL methods. We further validate COSMIC-FL on an NVIDIA Jetson AGX Orin, confirming its efficiency gains under realistic embedded deployment constraints.
[CV-449] Facial classification Using Hybrid Quantum Machine Learning
链接: https://arxiv.org/abs/2609.31915
作者: Roshan Babu Bandlapalli,Srinivas V Katakam,Jitendra Chougala,Ravi Kumar Kappagantu,Jayasri Dontabhaktuni
类目: Computer Vision and Pattern Recognition (cs.CV); Quantum Physics (quant-ph)
备注:
Abstract:Hybrid quantum methods have received limited study for resource-constrained facial biometrics. We present a hybrid quantum-classical facial recognition pipeline designed to run on standard computing hardware. Images undergo gamma correction, contrast enhancement, and principal component analysis before their features are encoded into an eight-qubit variational quantum classifier. Classical image matching then performs recognition. In experiments with 50,000 images, comprising 25,000 faces from CelebA and 25,000 non-face images from CIFAR-10, the method outperformed the reported CPU-trained FaceNet baseline in accuracy and training efficiency. The pipeline was also evaluated on GPU and quantum hardware. In an attendance monitoring deployment at Mahindra University in collaboration with Lloyds Technology Centre, CPU inference took 0.2 to 0.5 seconds per person, and the system remained robust to the use of spectacles. These findings support the feasibility of deploying hybrid quantum methods for facial recognition on existing CPU hardware.
[CV-450] PredRA: Fast Medical Image Translation by Deterministic Component Extraction and Controlled Stochastic Refinement
链接: https://arxiv.org/abs/2609.31912
作者: Jianhai Zhang,Pattarawut Charatpangoon,Donghao Zhang,Bijoy K. Menon,Wu Qiu,M. Ethan MacDonald,Aravind Ganesh
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint. 14 pages, 7 figures, 12 tables. Code: this https URL
Abstract:Strongly paired medical image translation contains a substantial component that can be predicted directly from the source. The current reality is that pure deterministic prediction can smooth away fine detail, while generative models can recover detail but may also introduce unnecessary or potentially harmful variation. We propose PredRA, a fast framework that uses the deterministic prediction as a stable reference and further extracts additional deterministic components from the residual during generative refinement, thereby improving fidelity while maintaining perceptual quality. The goal is simple: we view the entire residual-based generative process as an optimization problem and derive a practical solution that recovers useful residual detail through controlled refinement while keeping the prediction close to the paired target. Mechanism studies further show that useful residual information follows structured patterns, but its usefulness is difficult to estimate reliably at the voxel level. We therefore globally control how much of the residual refinement is added to the deterministic prediction, thereby reducing the accumulation of unnecessary uncertainty. PredRA therefore combines deterministic component extraction with controlled stochastic refinement for fast, fidelity-preserving, and perceptually strong medical image translation. We validate the approach across multiple real-world datasets, showing that controlled refinement improves fidelity to the paired target compared with full residual refinement while retaining much of the perceptual benefit of generative modeling. PredRA achieves competitive or superior performance to substantially larger state-of-the-art models with 1.4-11.9x fewer total parameters and 3.1-26.2x fewer trainable parameters, while its 32-step flow sampler requires 31.25x fewer sampling steps than the matched 1000-step DDPM.
[CV-451] Enhancing Visual Reasoning in Chest X-Ray Report Generation Using Reinforcement Learning
链接: https://arxiv.org/abs/2609.31911
作者: Denis Musinguzi,Andrew Katumba,Prasenjit Mitra
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Medical report generation has made significant progress with the rise of modern vision-language models and the growing availability of large-scale medical datasets. However, hallucinations remain a major challenge, largely due to the limitations of supervised fine-tuning (SFT), which prioritizes lexical similarity to reference reports rather than clinical correctness. While reinforcement learning has shown strong performance in domains with verifiable rewards such as mathematics and code generation, its application to open-ended medical tasks remains limited. Existing work focuses on evaluating final answers, overlooking the model’s reasoning, despite evidence that flawed reasoning can degrade overall performance. In this study, we propose a framework that verifies the model’s reasoning process by integrating anatomical regions, bounding boxes, and region-level textual descriptions. We design spatial and factual reward mechanisms to ensure that the model’s reasoning is both visually grounded and factually accurate. Starting from Qwen3-VL-8B-Instruct as our base model, we adapt it to the medical domain using supervised fine-tuning, introduce reasoning capability through a cold-start SFT stage, and refine it with reinforcement learning. We find that RL provides performance gains beyond those achievable through SFT alone, and that jointly verifying both reasoning steps and final outputs yields larger improvements than verifying either in isolation. We further identify multiple modes of reward hacking in the RL stage. Finally, the model’s structured think traces enhance interpretability, making its outputs easier to audit for clinical use.
[CV-452] Auditing Quality Filters for Long-Tail Human Data Curation
链接: https://arxiv.org/abs/2609.31896
作者: Rishav Agarwal,Nirshal Chandra Sekar,Anirudh Vemula
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 4
Abstract:Robots on construction sites must detect workers who are kneeling or bending, which we call low poses. These workers can be lost from training datasets during automatic labeling. We study a pipeline that detects people, estimates their body joints using NLF, and groups similar poses. Low poses account for only about 2 percent of the retained examples. This low share may partly reflect the pipeline’s quality filter, which rejects examples with low detection confidence or uncertain joint estimates. We examine this filtering using four alternative pose clues: bounding-box shape, vertical body span, pose grouping aligned to the scene’s vertical direction, and image appearance. All four suggest that low poses are rejected by the filter more often. Separately, controlled simulated scenes show that a person detector fine-tuned on a public construction dataset misses more workers in these poses even when we correct their bounding box height is matched to that of standing workers. These findings suggest that low poses are scarce and hard to find, and we cannot rely on bounding boxes or poses for long-tail human data curation.
[CV-453] SynDORBench: Evaluating LVLM Perceptual Robustness Under Physically Constrained Visibility Conditions
链接: https://arxiv.org/abs/2609.31823
作者: Jeremy Stephen Gabriel Yee,Zhengkui Wang,Zhiyuan Zhang,Avinash Anand,Timothy Liu,Benedict Chan,Aik Beng Ng,Simon See
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Large vision-language models (LVLMs) have demonstrated remarkable performance on multimodal reasoning benchmarks, yet their perceptual reliability under physically constrained imaging conditions remains poorly understood. Existing evaluations predominantly assume ideal visual inputs and therefore fail to characterize how camera distance, illumination, viewpoint, and pixel density fundamentally affect semantic recoverability. We introduce SynDORBench, the first physically grounded benchmark for evaluating LVLM perceptual robustness under DORI-calibrated conditions aligned with human visual capability standards. SynDORBench comprises over 54k question–answer pairs generated through a controllable synthetic pipeline that systematically varies viewing distance, lighting, camera geometry, and action pose according to physically interpretable pixel-density regimes. To support scalable low-visibility supervision, we further propose a discernibility annotation framework that propagates human perceptual labels using mask-conditioned statistical features and ensemble learning. We evaluate 16 open-source LVLMs, a commercial LVLM baseline, and YOLO11x across human-presence classification and action recognition tasks under progressively degraded visibility conditions. Our results reveal that perceptual failure in LVLMs is strongly governed by pixel density and physical imaging constraints rather than model scale alone. Surprisingly, several compact open-source LVLMs outperform larger commercial baselines and substantially exceed YOLO11x robustness under long-range and low-light conditions. SynDORBench establishes a new benchmark paradigm for physically grounded multimodal evaluation, enabling systematic analysis of LVLM reliability under real-world perceptual constraints and direct comparison against human visibility thresholds.
[CV-454] A Surgical Foundation Model Reveals Task-Dependent Label Efficiency
链接: https://arxiv.org/abs/2609.31821
作者: Florian Philipp Stilz,Lorenzo Arboit,Vinkle Srivastav,CAMMA International Surgical Partners,Jacques Marescaux,Sergio Alfieri,Pietro Mascagni,Nassir Navab,Nicolas Padoy
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 14 page, 6 figures
Abstract:Developing label-efficient models is a central challenge in surgical AI due to the high cost and scarcity of expert annotation. While self-supervised foundation models adapt well to new tasks with minimal data, how label efficiency varies across different surgical tasks remains largely unexplored. Here, we introduce SURGE, a surgical foundation model trained on SurgSpectrum-30M+, the largest pretraining dataset comprising over 30 million frames, with checkpoints released to enable further research. We systematically evaluate label efficiency across 5 task categories and 15 benchmarks. These range from temporal and spatial scene understanding to fine-grained reasoning tied to instrument-anatomy interactions and safety-critical maneuvers. SURGE outperforms prior state-of-the-art on all benchmarks, even surpassing task-specific models on complex reasoning tasks. Crucially, we reveal a task-dependent scaling behavior: while scene understanding tasks saturate with minimal supervision, fine-grained reasoning tasks continue improving with substantially larger annotation budgets, providing a blueprint for allocating expert effort in complex domains. Code: this https URL Comments: 14 page, 6 figures Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.31821 [cs.CV] (or arXiv:2609.31821v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.31821 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Florian Stilz [view email] [v1] Fri, 25 Sep 2026 16:31:21 UTC (12,410 KB)
[CV-455] Video-to-Music Generation for Gameplay Videos
链接: https://arxiv.org/abs/2609.31810
作者: Felipe Marra,Lucas N. Ferreira
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注: Project page: this https URL
Abstract:Video-to-music models have advanced considerably in the last few years, particularly in film and music video applications. In this paper, we investigate this problem in the video game domain, which introduces new challenges for these models: video frames are rendered graphics, music is mostly synthetic audio, and soundtracks loop across entire levels rather than following on-screen events. We introduce a new dataset of 217.6 hours of Super Nintendo (SNES) gameplay video paired with 485 hours of clean soundtracks, free of sound effects and voice-overs, matched to gameplay audio via audio fingerprinting. With this dataset, we train a simple encoder-decoder transformer that passes video features directly to a MusicGen decoder, comparing different encoding strategies: textual descriptions (T5), independent frames (ViT), or spatiotemporal patches (ViViT). Each encoder is tested both frozen and fine-tuned, while the decoder is always fine-tuned. Frozen encoders match or outperform their fine-tuned counterparts on every metric, and the frozen ViViT achieves the best overall results. We compare this model with state-of-the-art baselines using both objective metrics and a listening study (N = 96). Despite having up to 18% fewer parameters, our model outperforms all baselines on objective metrics, surpasses GVMGen in the listening study, and performs comparably to OSSL.
[CV-456] Rate-Adaptive One-Step Diffusion Compression for AIGC Images
链接: https://arxiv.org/abs/2609.31795
作者: Nitiz Khanal
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We describe our entry to the LoViF 2026 AIGC Image Compression Challenge, a benchmark for ultra-low-bitrate coding of AI-generated images under a strict global rate budget of 0.025 bits per pixel (BPP). Generated imagery poses a distinct challenge for compression: it frequently contains rendered typography, synthetic edges, repeated motifs, UI-like layout, and stylized micro-texture that conventional distortion-oriented codecs erase at this rate, while unconstrained generative decoders can restore plausible-looking detail that no longer matches the source geometry or symbols. We treat this as a rate-perception allocation problem. Our system fine-tunes four rate-specialized checkpoints of the AEIC one-step diffusion codec, generates per-image candidates from all four, including one latent refined through encoder-side test-time optimization (TTO) with periodic entropy-conditioning refresh, entropy-codes every candidate with practical rANS coding, and selects exactly one bitstream per image with an exact multiple-choice knapsack solved over true coded file sizes. A fixed, zero-additional-bit residual restoration network is applied at decode time. Every submitted bitstream is independently decodable by the shipped decoder, which uses no source image or external side information. We report the full pipeline, an ablation history spanning 77 logged experiments, and a set of negative results, including why PSNR could not be pushed to parity with rate-distortion-oriented competitors under this architecture, useful to future participants. This is a challenge report: our entry scored 31.527739 (PSNR 27.02 dB, MS-SSIM 0.9176, LPIPS 0.0778, DISTS 0.0390 at 0.02495 BPP), the second-best DISTS on the leaderboard, ranking 5th on the final test-phase leaderboard announced August 4, 2026.
[CV-457] SelfCue: Making a 3D CT Report Generator Say What It Already Knows
链接: https://arxiv.org/abs/2609.31788
作者: Renjie Liang,Yang Yang,Jinqian Pan,Zhengkang Fan,Chengkun Sun,Jie Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Progress in 3D CT report generation is usually sought in increasingly sophisticated architectures and larger pools of training data. We find instead that a 3D CT report generator already holds what its report leaves out, and loses it when the hidden state becomes tokens. Over the 18 CT-RATE abnormalities, this hidden-to-report surfacing gap is reflected by a drop in macro AUROC from 0.848 in the hidden states to 0.739 in the generated report. We propose SelfCue based on contrastive decoding. It promotes what the hidden state already supports and suppresses what it does not. It raises clinical efficacy F1 to 0.481 and the LLM-judged GREEN score to 0.510. Distilling that behaviour into the weights gives SelfCue-KD, a student that keeps most of the gain, needs nothing extra at inference, and drops into any pipeline already serving the baseline. Code is available at this https URL.
[CV-458] Panoptic Scene Program Diffusion Transformer NEURIPS2026
链接: https://arxiv.org/abs/2609.31780
作者: Chika Maduabuchi
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted to NeurIPS 2026
Abstract:Modern text-to-image models produce high-fidelity images but still struggle with compositional prompts that require instance identity, attribute ownership, counting, spatial ordering, and role-sensitive relations. We introduce Panoptic Scene Program Diffusion Transformer (PSP-DiT), a diffusion-transformer architecture that treats a panoptic scene program as a first-class latent variable rather than an external control signal or post-hoc parse. PSP-DiT jointly denoises image latents and scene-program latents through coupled transformer streams, while panoptic grounding and cycle-consistency objectives tie object instances, attributes, relations, and counts to visual support in the generated image. Under matched training and inference settings, PSP-DiT improves over a strong flat-text baseline across GenEval 2, SANEval-Simple, PSG-Score, and DetailMaster, with the largest gains on counting, attribute binding, role-sensitive relations, and long structured prompts. The method preserves image quality, adds modest inference overhead, and remains robust to imperfect scene programs.
[CV-459] UNMATCH: Selective Unbalanced Token-Patch Matching for Forensic Image-Claim Verification
链接: https://arxiv.org/abs/2609.31766
作者: Xinjin Li,Lian Lian,Yuanzhe Yang,Yudi Xia,Calvin Chang Liu,Yeyun Xu,Yu Ma,Jinghan Cao,Yuruo Gong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Contextual image misuse pairs an image with a misleading claim. We study image-claim correspondence in fact-checked pairs containing out-of-context reuse, visual manipulation, or both. Existing pair-based detectors often compress the two modalities into a global compatibility score or learn a highly flexible interaction module, which can obscure a decisive local mismatch. We introduce directional multiscale coverage, a compact representation that summarizes local image-claim affinity in both directions and at three spatial scales. At each scale, each direction is summarized by its mean, lower quartile, and two thresholded support ratios; the signed difference between directional means completes a nine-dimensional scale descriptor. Concatenating the three scales yields a compact local representation for a lightweight global-local classifier. Under leakage-aware three-fold, three-seed evaluation on the Snopes subset of the Fauxtography benchmark, UNMATCH achieves 69.82 Macro-F1 and 71.05 balanced accuracy, exceeding the MCOT adaptation by 2.60 and 2.16 points. A matched-reassigned intervention shows that breaking the observed pairing lowers coverage and increases both discrepancy and false-pair probability.
[CV-460] HGPT rans: Hierarchical Graph-Pooling Transolver for Automotive Aerodynamic Drag Coefficient Prediction
链接: https://arxiv.org/abs/2609.31765
作者: Bo Liu,Qiuli Luo,Lianrui Nie,Fengli Zhang,Wenjiang Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Fluid Dynamics (physics.flu-dyn)
备注:
Abstract:Accurate and rapid prediction of the aerodynamic drag coefficient ( C_D ) is essential for vehicle design, particularly during early-stage styling iterations where a large number of candidate geometries must be evaluated. Although computational fluid dynamics (CFD) provides reliable aerodynamic estimates, its high computational cost, typically requiring hours to days for a single configuration, limits its use in large-scale design exploration. This paper proposes HGPTrans, a hierarchical graph-pooling network with Transolver-based attention, to directly predict C_D from vehicle surface meshes. Motivated by the fact that vehicle aerodynamics depends on both local geometric features and long-range interactions among spatially distant surface regions, HGPTrans integrates three complementary components. Graph isomorphism convolutions encode discriminative local geometry, physics-aware slice attention captures global interactions with linear computational complexity, and information-redundancy-aware hierarchical pooling progressively removes redundant nodes while preserving informative geometric structures. The model is trained and evaluated on the large-scale DrivAerNet and DrivAerNet++ datasets, where it achieves the lowest mean absolute error and mean squared error among the evaluated baselines. Its generalization capability is further assessed through transfer learning on a real-vehicle dataset containing both sedans and SUVs, achieving relative L_1 errors of 1.56% (sedans) and 2.12% (SUVs) with an inference time of approximately 0.293 s per vehicle. This corresponds to an acceleration of several orders of magnitude relative to high-fidelity CFD while keeping the predicted drag coefficients within a few percent of the CFD reference. Ablation studies confirm each component’s contribution and reveal the effects of depth and pooling ratio.
[CV-461] 3dgs-sc: a controlled static screen-content benchmark for 3d gaussian splatting
链接: https://arxiv.org/abs/2609.31756
作者: Shicheng Cai,Hao Zhang,Dong Dai,Xuerui Ma,Ying Hu,Tao Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Does high image fidelity imply readable screen content in 3D Gaussian Splatting (3DGS)? We introduce 3DGS-SC, a controlled static screen-content dataset and benchmark for examining this mismatch. Ten procedural scenes provide fixed multi-view splits, exact cameras, screen masks, text boxes, and transcripts. The protocol separates whole-image fidelity, screen-region fidelity, OCR readability, and edge preservation. In the reported five-method comparison, LightGaussian exceeds Mip-Splatting in screen PSNR by only 0.08 dB, yet trails it in OCR accuracy by 12.7 percentage points. Across all ten method pairs, screen-PSNR and OCR orderings disagree in seven cases. These aggregate results expose a method-selection failure of fidelity-only evaluation. Complementing prior text-aware 3DGS research, 3DGS-SC targets controlled monitor interfaces and exact annotations; scene-wise robustness and acquisition effects remain open validation questions.
[CV-462] Gauge-Equivariant Attention for Rotation-Stable 360circ Scene Understanding SIGGRAPH
链接: https://arxiv.org/abs/2609.31755
作者: Tianjian Zhou,Yishan Li,Jie Jiang,Yifei Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 23 pages. Accepted to ACM Transactions on Graphics (SIGGRAPH Asia 2026)
Abstract:Panoramic 360^\circ scene understanding increasingly relies on icosphere transformers, but a state-of-the-art spherical model loses more than half of its segmentation accuracy when the camera rotates by 90^\circ , and controlled ablations identify gauge dependence in its relative-position bias as a major contributor. We propose gauge-equivariant relative position encoding (GE-RPE): a parameter-free Reynolds average of the bias over a finite cyclic subgroup C_n!\subset!\mathrmSO(2) of gauge rotations. Plugged into a SphereUFormer backbone the change is invisible at deployment—zero added parameters and 1.5 – 4.4% forward latency—and the matched three-seed GE-RPE model records a 1.3% drop; the published SphereUFormer checkpoint records 53% under the same stress protocol but a different training recipe. Once the gauge defect is removed and a teacher-token permutation \pi_R aligns the SSL views to the rotated student frame, iBOT + MAE pretraining stops being a liability and becomes a clean low-label lever: the full framework EquiSSL (GE-RPE + \pi_R + iBOT + MAE) tightens the drop to 0.8% at 68.30% val mIoU and lifts 1% -label fine-tuning by +2.39 mIoU on the N=373 test split (and by +4.10 on the smaller N=40 val split); the same fix carries over to monocular depth and to zero-shot Structured3D segmentation. The construction is provably C_n -invariant and \mathcalO(n^-2) -close to the continuous \mathrmSO(2) average, making the resulting model a usable 360^\circ visual-computing primitive across panoramic relighting, immersive video, and cross-dataset transfer. Code is available at this https URL.
[CV-463] Calibration-Free Surface Normals Estimation in Vision-Based Tactile Sensing using Universal Photometric Stereo
链接: https://arxiv.org/abs/2609.31754
作者: Zdravko Dugonjic,Stefanie Speidel,Roberto Calandra
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Robotics (cs.RO)
备注: 8 pages
Abstract:Vision-based tactile sensors are a popular solution for capturing rich contact surface geometry. However, to obtain high-detail contact surface normals and depth, it is necessary to calibrate the sensor by physically pressing a probe with known geometry against the sensor elastomer and mapping tactile images onto the ground truth probe’s shape. This approach does not scale across different tactile sensors, and the calibration effort can be complex depending on the sensor shape and optical system. Instead, we propose a calibration-free procedure for the estimation of contact surface normals using Universal Photometric Stereo neural networks. In a series of real-world experiments, we evaluate our approach on 3 sensors with different optical systems, demonstrating that universal methods are a suitable approach for estimating surface normals at the contact patch from tactile images, thereby alleviating the need for tactile sensor calibration. Controlled experiments with a metal ball show that universal methods match the calibrated method, with a mean angular error of 6.56^\Large\circ . We show that the proposed framework recovers high-frequency surface details of objects with natural textures, achieving an overall mean angular error of 10.66^\Large\circ . Universal method robustly recovers the contact surface normals captured with dome-shaped Digit 360, achieving a low angular discrepancy of 10.18^\Large\circ relative to the calibrated baseline. This experiment demonstrates that with sufficient illumination settings surface normals could be estimated using a model trained solely on synthetic data. By providing a unified representation of contact surfaces across different vision-based tactile sensor designs, Universal Photometric Stereo neural networks lay the foundation for transferable tactile perception across sensors.
[CV-464] EgoTSR: Egocentric Spatiotemporal Reasoning for Task Progress Understanding
链接: https://arxiv.org/abs/2609.31751
作者: Xiaoda Yang,Can Wang,Yuxiang Liu,Pengfei Zhou,Jianwen Lou,Shuicheng Yan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-Language Models (VLMs) have advanced rapidly in static visual understanding, yet remain unreliable when judging how an egocentric task is progressing. Given a task instruction and two visual observations, a model should determine which state is closer to the goal by analyzing task-relevant object configurations and spatial relations, rather than relying on timestamps or presentation order. This distinction is critical in manipulation, where retries, corrective actions, and temporary regressions make progress inherently non-monotonic. We introduce EgoTSR, a unified framework for diagnosing and improving order-robust task-progress understanding. First, SpatialLogic-Bench evaluates each physical state pair in both original and order-swapped presentations across short- and long-horizon settings, exposing whether a model follows task-state evidence or chronological shortcuts. Second, our data construction pipeline converts successful, approximately monotonic manipulation and first-person trajectories into bidirectional supervision; LongTag further preserves intermediate subtask structure for long-horizon comparison, while failure-aware data extend learning to regressions and recoveries. Third, a progressive CoT-to-Tag curriculum first supervises evidence-grounded interpretation of task-relevant state changes and then consolidates the comparison rule through scalable label-only training. Experiments reveal substantial input-order bias in representative VLMs. EgoTSR achieves 92.4% long-horizon accuracy with a 0.1-point forward-inverse Gap. Failure-aware supervision further improves accuracy on non-monotonic trajectories by 11.8 points and Recovery Accuracy by 11.2 points, while maintaining broad visual and spatial capabilities. These results establish goal-conditioned state comparison as an explicit formulation of egocentric spatiotemporal reasoning for task-progress understanding.
[CV-465] Dimension-Specific Imbalance and an Adaptive Hybrid Label Strategy for Multi-Task Affective State Recognition in Classroom Video
链接: https://arxiv.org/abs/2609.31750
作者: Xiangqian Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 15 pages, 4 figures, 7 tables
Abstract:Recognition of student affective states from classroom video is constrained by a class imbalance problem whose true nature is, we argue, under-analyzed. We show that on DAiSEE – the de facto benchmark for this task – class imbalance is fundamentally a dimension-specific phenomenon: raw imbalance ratios reach as high as approximately 79:1 (Frustration), and the structure of imbalance – not only its magnitude – varies across dimensions, so uniform algorithmic treatments (aggressive class reweighting, focal loss, uniform binary simplification) fail in dimension-dependent ways. We propose the Adaptive Hybrid Label Strategy (AHLS), which assigns four-level classification to dimensions with manageable imbalance and binary classification to severely long-tailed dimensions, coupled with a null-class placeholder mechanism that stabilizes multi-task optimization by compressing the maximum effective inverse-frequency weight ratio from as high as ~79 \times to at most ~8.5 \times across the four dimensions. Validated on a lightweight FERShuffleNetV2 + LTCN architecture (0.39 M parameters, 0.07 G FLOPs), the strategy attains the highest mean Macro-F1 of 47.70% across all four affective dimensions among 11 compared methods, including five state-of-the-art deep models (up to 167 \times larger) and five traditional machine-learning baselines. A cross-model behavioral analysis on DAiSEE indicates that, in this setting, classification strategy may influence minority-state recognition more strongly than further backbone scaling alone. We also report evidence of an annotation-density limitation of DAiSEE on minority affective states, which suggests that benchmark design may bound future progress as much as algorithmic refinement. Index Terms Affective computing, classroom video analysis, class imbalance, multi-task learning, lightweight deep learning, DAiSEE, engagement recognition.
[CV-466] Fusion Under Component Failure: Negative Results and Failure Modes in Ensemble AI-Generated Image Detection
链接: https://arxiv.org/abs/2609.31749
作者: Suraj Singh,Tushar Verma,Pragyan Singh,Shaurya Bhav,Shivam Kumar
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We built an ordinary stacking ensemble for AI-generated image detection – three open detectors producing five scores, fused by a gradient-boosted meta-learner that treats a detector’s failure as missing data – deployed it, and then evaluated it against three controls it should have faced first. This paper reports what the controls found, including where they overturned our own earlier conclusions. Fusion is worth its cost only when refitted on the target domain. The shipped meta-learner, fitted on a separate corpus, does not beat its best single member on 2000 StyleGAN faces (AUC 0.9896 vs 0.9961; McNemar p = 1.000). But a stacker refitted in-domain beats that member plus a post-hoc calibrator (Delta-AUC = +0.0025, [+0.0014, +0.0038]; p = 3.4e-4). An earlier draft claimed the calibrated single detector won outright; that comparison mixed regimes and we correct it here. One corpus is not an evaluation. On 80 screenshots every model’s AUC interval contains 0.5. We can say nothing stronger: the difference between the ensemble’s drop and its best member’s is [-0.185, +0.115]. Abstention is real, correlated, and mishandled. With four of five detectors silent and the survivor reporting “real”, the system returns P(AI) = 0.9985, because an all-NaN input scores 0.9995 in a learner never fitted with missingness. The three AIDE checkpoints fail together, sharing one preprocessing path. A quorum rule requiring two distinct architectures prevents both failures with no retraining. We also find Corpus A carries a class-conditional JPEG bias severe enough to separate the classes from the header alone, which limits every in-domain number we report. Code, harness, hash-identified artifacts and all corrections are released. Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2609.31749 [cs.CV] (or arXiv:2609.31749v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.31749 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Suraj Singh [view email] [v1] Wed, 23 Sep 2026 08:14:20 UTC (79 KB) Full-text links: Access Paper: View a PDF of the paper titled Fusion Under Component Failure: Negative Results and Failure Modes in Ensemble AI-Generated Image Detection, by Suraj Singh and 4 other authorsView PDFHTML (experimental)TeX Source view license Current browse context: cs.CV prev | next new | recent | 2026-09 Change to browse by: cs References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab) Major funding support from
[CV-467] he Earth in One Gaze: Training-Free Active Focus for UHR Remote Sensing Understanding
链接: https://arxiv.org/abs/2609.31747
作者: Yao Zhang,Pengyu Dai,Wei Guo,Jian Liang,Jian Song,Yafei Ou,Hongruixuan Chen,Naoto Yokoya
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注:
Abstract:Multimodal large language models (MLLMs) must balance local detail against scene context when interpreting ultra-high-resolution (UHR) remote sensing (RS) imagery within a limited visual-input budget. Existing selection-based methods either prune tokens and select patches through relevance scoring, or crop actively through repeated inspection. Neither strategy directly redistributes pixels within a continuous full-scene view: the first retains selected tokens or patches, and the second re-encodes a crop detached from its surroundings. Our pilot study finds that a frozen MLLM already produces useful question-guided spatial requests, yet crop-based inspection of the selected regions does not consistently improve its answers. We therefore formulate UHR understanding as a question of where to spend a fixed pixel budget. Based on this, we introduce GazeEarth, a simple-yet-effective training-free framework that couples question-guided region selection with full-scene foveated observation. The MLLM selects evidence cells from an indexed overview; a deterministic, topology-preserving warp resamples the original image onto a fixed-size canvas, enlarging their shared neighborhood while compressing the periphery; the same frozen model answers from this focused view, using at most two MLLM calls and no external selector or iterative search. Across three UHR remote sensing benchmarks and four frozen backbones, GazeEarth improves benchmark-averaged accuracy by 4.6 to 9.4 percentage points over direct answering and 3.4 to 4.3 over overview answering, outperforming task-trained methods. Our analyses show that existing MLLMs can guide where to look in UHR images on their own, and that what they can infer from the selected evidence depends on how that evidence is presented.
[CV-468] VisionPsy-Nano: Improving Accuracy Efficiency and Reliability in On-Device Vision-Language Models
链接: https://arxiv.org/abs/2609.31746
作者: Khurram Azeem Hashmi,Mohammadreza Zolfaghari,Changdae Park,Rishabh Jain,Nicholas Moratelli,Pengfei Wei,Louis Lu,Tianchi Liu,Amril Nazir
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Sub-billion-parameter Vision-Language Models are increasingly viable for on-device deployment, yet compact model size alone does not guarantee usability. On a phone, such a model can still require more than two minutes to produce its first token. On-device usability depends on three axes: accuracy, efficiency, and behavioral reliability; standard benchmarks miss the third, with answers too short to expose doom loops and prompts too benign to probe adversarial safety. We introduce a diagnosis-driven post-training recipe in which a teacher VLM stress-tests the student, uncovers failure modes beyond human priors, and converts them into targeted supervision and preference alignment, supplementing generic data scaling with failure-driven optimization. Coupled with two visual-token policies, the recipe yields two accuracy-efficiency variants with improved behavioral reliability. \textbf\NanoFull attains a 62.3 normalized average over 17 benchmarks, the highest among openly released \sim 0.5B models, +7.4 over its base at identical architecture and token budget, with doom-loop rates at or below the strongest baseline’s. \textbf\FlashFull retains 61.4 while cutting warm time-to-first-token on a Pixel 9 from 138,s to 6.1,s (23 \times ). By jointly addressing all three axes, we move compact VLMs toward practical on-device usability.
[CV-469] PEEL-DDPM: Physics-Enabled Evidential Learning for the Denoising Diffusion Probabilistic Model
链接: https://arxiv.org/abs/2609.31742
作者: Ge Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 13 pages, 5 figures, 2 tables
Abstract:Normal-inverse-gamma (NIG) regression is not identifiable from its marginal Student-t likelihood: three combinations of four NIG parameters are determined, leaving one degree of freedom. We introduce PEEL-DDPM, a physics-enabled evidential learning framework for denoising diffusion probabilistic models. A measurement-conditioned DDPM is first trained with epsilon-MSE and then frozen. Its complete reverse trajectory produces a reconstruction, whose residual from the known training object is modeled by a final-image evidential network. The network learns only the identifiable Student-t coordinates, while repeated scanner-noise realizations and repeated diffusion trajectories provide a nested Monte Carlo estimate of final-image aleatoric variance, separated into scanner-induced and sampler-induced components. This measured variance resolves the remaining NIG ambiguity and yields a decomposition of predictive uncertainty into measurement, diffusion, and epistemic terms. The method uses sequential training without a cross-loss weighting coefficient. In a feasibility study on eight held-out objects, empirical central-interval coverages were 49.3%, 80.3%, 90.1%, and 95.6% for nominal 50%, 80%, 90%, and 95% intervals. The mean squared residual was 0.967 times the mean predicted variance, and a single-image aleatoric head achieved pooled Spearman correlation 0.785 against an independent nested reference. Across five dose levels, scanner-induced variance showed a log-log dose slope of -1.15, whereas sampler-induced variance remained nearly dose independent with slope -0.01. These results support PEEL-DDPM as a practical route to identifiable and physically interpretable uncertainty quantification in diffusion-based image reconstruction. Comments: 13 pages, 5 figures, 2 tables Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2609.31742 [cs.CV] (or arXiv:2609.31742v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.31742 Focus to learn more arXiv-issued DOI via DataCite
[CV-470] Beyond Volume Overlap: Surface Matching for Topology-Aware Coronary Artery Segmentation MICCAI2026
链接: https://arxiv.org/abs/2609.31740
作者: Rafael Velasquez,Esther Puyol-Antón,Pablo Arbeláez
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 12 Pages, 2 figures, Statistical Atlases and Computational Modeling of the Heart (STACOM) workshop of MICCAI 2026
Abstract:Accurate coronary artery segmentation on coronary computed tomography angiography (CCTA) is essential for diagnosing coronary artery disease. Deep networks are conventionally trained and evaluated with the Dice coefficient, but volume-overlap metrics are poorly suited to thin, tubular anatomy: since most voxels belong to a few thickproximal segments, a missing distal branch barely affects Dice despite severely disrupting the connectivity required for clinical use. We introduce a surface metric that matches predicted and reference surface points via bipartite assignment under a localized, vessel-radius tolerance, reporting precision, recall, and F1 with decoupled false positives (spurious branches) and false negatives (missed branches) a distinction the symmetric Dice cannot make. With it we show that a strong Dice-trained baseline omits far more vessel surface than it hallucinates, an asymmetry its high Dice hides. Building on this, we propose a differentiable surface loss that simultaneously suppresses spurious mass and recovers absent structure, validated by fine-tuning three backbones (nnU-Net, SwinUNETR, NexToU) on two public benchmarks (Image-CAS, ASOCA). Against a matched-epoch control, it significantly improves surface F1 by recovering missed distal vessels at comparable Dice. Our findings argue for measuring and optimizing the vessel surface, not the volume it overlaps. Code
[CV-471] Seeing the Heat: Synthesizing High-Resolution Wood Thermal Responses from Optical Imagery
链接: https://arxiv.org/abs/2609.31737
作者: Jingren Xie
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:The thermal behavior of wood is a critical factor in advanced material assembly. However, pixel-level thermal analysis remains fundamentally constrained by the low resolution and noise inherent to infrared thermography. To address this, we introduce an end-to-end computational framework that synthesizes high-resolution thermal responses directly from wood RGB images. We first establish a core physical linkage: because spatial color variation in natural wood is driven by cellular anatomy, optical intensity serves as a reliable geometric proxy for the localized solid volume fraction. By leveraging this theoretical insight, we develop an automated finite-element-method data engine that maps pixel-level optical intensity to a 3D thermodynamic voxel grid, generating high-fidelity synthetic thermal responses. We find that 1) when the thermal conductivity along the thickness direction is uniform or linear, wood RGB images and their corresponding thermal responses exhibit extreme morphological similarities, and the lateral thermal diffusion acts as a low-pass filter that smooths out high-frequency details; 2) when the thermal conductivity along the thickness direction is random, such morphological similarities are destroyed, and wood’s 3D structure dominantly governs its thermal response. We further utilize these synthetic thermal responses to supervise a neural surrogate model built upon the DINOv3 foundation model. Our results demonstrate that the neural surrogate model successfully internalizes the governing thermodynamic laws, thereby bypassing computationally expensive simulations and enabling high-resolution thermal inference. This methodology effectively bridges the semantic and thermodynamic domains, unlocking systematic, pixel-level analysis of fine-grained wood thermal responses. Project: this https URL
[CV-472] LukeNet: A lightweight CNN integrated with an XAI model for Smart acute lymphoblastic leukemia detection and management
链接: https://arxiv.org/abs/2609.31736
作者: Md Taimur Ahad(Department of Management Information Systems, North South University, Bangladesh)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Acute Lymphoblastic Leukemia (ALL) patients require early, accurate detection to enable timely treatment and effective patient management. A Convolutional Neural Network (CNN) is well-suited for creating an end-to-end enabling environment for ALL detection and classification. However, most CNN-based ALL detection systems are theoretical and unsuitable for deployment on edge devices due to high computational demands. The Internet of Medical Things (IoMT)-enabled devices offer an opportunity to monitor ALL patients in real time. Wearables that track temperature, heart rate, oxygen saturation, and activity can deliver critical data to support timely clinical intervention and improve patient outcomes. In smart IoMT environments, a lightweight CNN is essential because connected devices often operate under limited computational power, memory, and latency constraints. To address this need, this study proposes LukeNet, a lightweight CNN integrated with explainable artificial intelligence (XAI) for an IoMT-based SMART Acute Lymphoblastic Leukemia Detection and Management System. Trained on three (3) ALL datasets and five-fold cross-validation, LukeNet achieved an impressive 99% model accuracy as well as 99% unseen test accuracy, which is higher than six state-of-the-art (SOTA) CNNs, such as DenseNet121, MobileNet, ResNet50, InceptionV3, Xception, and VGG16, as well as transfer learning models. Furthermore, LukeNet was compared with two ensemble models. In addition, explainable AI methods are integrated to highlight relevant regions in microscopic images. The novelty of this study lies in the architecture of LukeNet, which balances model depth and computational efficiency by using depthwise separable convolutions, mitigates the risk of gradient loss in deeper layers, and provides strong global and local feature extraction capabilities.
[CV-473] Measuring the evolution of camera distance across a century of film
链接: https://arxiv.org/abs/2609.31734
作者: David Bamman,Allison Cooper,Dan Hickey,Madison Mar
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The rise of computer vision and artificial intelligence has made possible new forms of large-scale computational measurement. We apply these techniques to a deep collection of 5,205 digitized films viewed in theaters between 1922-2025 (covering popular, prestigious, and independent movies) to trace the development of one of the most fundamental ways through which film communicates: by manipulating the space between the camera and its subject. This work finds abrupt changes with the rise of new technologies in sound and television, and allows us to shed an empirical light on gender disparity (women, despite having substantially less screentime than men, are disproportionately the subject of closer shots), and illustrate how animated films both inherit and break free from the norms of live-action filmmaking.
[CV-474] When Retrieval Hurts: Measuring and Explaining Retrieval-Induced Hallucination in Chest X-ray Report Generation
链接: https://arxiv.org/abs/2609.31733
作者: Emmanuel Idoko,Abdusshakur Olabisi,Shiloh Oni,Adesola Josiah
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 4 tables
Abstract:Retrieval-augmented generation is an attractive way to improve chest X-ray reporting, because reports from similar prior studies supply clinical context a general vision-language model lacks. We show the same mechanism is a reliable source of clinical error. Over 100 MIMIC-CXR studies with a frozen LLaVA-1.5 generator and BioMedCLIP retrieval, CheXbert clinical F1 doubles under relevant retrieval (0.201 to 0.402) and collapses to 0.043 under clinically mismatched retrieval, a fifth of the image-only score; the retrieval-induced hallucination rate, counting only unsupported findings traceable to retrieved evidence, rises from 0.00 to 0.76 and 0.98. To show this is not an artefact of evidence-set coverage, we introduce a coincidental-overlap control that scores image-only generations against evidence they never saw, placing the chance base rate at 0.18, four to five times below the observed rates. The generator does not merely acquire findings, it transcribes text: 95% of reports produced under relevant retrieval contain an eight-word span occurring verbatim in the retrieved evidence but absent from the reference, against 0% without retrieval. We then explain the mechanism: a normal chest X-ray retrieves at least one abnormal precedent in 24 of 28 cases, because medical image-embedding similarity is dominated by anatomy and acquisition rather than by the presence of disease. This has a direct design consequence. Retrieval similarity does not predict harm (RIH rates 0.80/0.80/0.60/0.84 across similarity quartiles), so relevance gates conditioned on embedding similarity cannot work; gating on predicted pathology agreement removes 61% of unsupported evidence at no cost to useful coverage. We argue that retrieval-augmented clinical systems must be evaluated under retrieval failure, not only under retrieval success.
[CV-475] GERIS: A Game-Theoretic Framework for Filtering Instance-Dependent Label Noise in License Plate Data Augmentation
链接: https://arxiv.org/abs/2609.31731
作者: Seyedeh Sara Jalili Shani(1),Rouhollah Ahmadian(2),Amin Rahmani(2),Mahdi Bideh(2),Mehdi Ghatee(2) ((1) Department of Computer Science, University of Alberta, Alberta, Canada, (2) Department of Mathematics and Computer Science, Amirkabir University of Technology, Iran)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 23 pages, 4 figures. To appear in AUT Journal of Mathematics and Computing (AUT J Math Comput)
Abstract:In this paper, we propose GERIS, a game-theoretic framework for instance selection in the data augmentation phase of license plate recognition systems. During augmentation, synthetic license plate images are generated and transformed using stochastic noise to simulate real-world conditions. However, certain noise configurations lead to highly distorted, unreadable images that degrade model performance by introducing instance-dependent label noise. GERIS formulates a non-cooperative game in which each noise vector competes for inclusion in the training set based on its similarity to labeled data and its contribution to model reliability. By identifying and pruning low-quality instances, GERIS improves the overall quality of the augmented dataset. Unlike traditional black-box learning methods, GERIS offers a transparent, theoretically grounded mechanism for data filtering. Experimental results demonstrate that GERIS outperforms existing instance selection methods in terms of classification accuracy and robustness.
[CV-476] High-Capacity Robust Medical Image Exfiltration via Neural Network Weight Replacement
链接: https://arxiv.org/abs/2609.31726
作者: Elie Thellier(EPIONE),Huiyu Li(EPIONE),Nicholas Ayache(EPIONE),Hervé Delingette(EPIONE)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:Collaborative medical AI platforms allow researchers to train models on sensitive imaging data while restricting data export. However, trained models can serve as covert carriers of patient information: medical images may be encoded within model parameters and reconstructed outside the secure environment. Existing defenses rely on lightweight sanitization (e.g., fine-tuning, pruning, quantization) and limited statistical auditing, creating a realistic insider exfiltration risk. We introduce a high-capacity neural steganography attack that encodes medical images as continuous latent representations embedded into model initialization. A StyleGAN2-based adversarial autoencoder learns compact latent codes regularized to match standard weight initialization statistics, keeping embedded parameters statistically consistent with clean models. Noise injection during training improves robustness to export-time mitigation. The carrier model remains functional on its intended task and hidden images can be reconstructed directly from its weights after export. This continuous encoding enables robust and scalable exfiltration, allowing up to 99 brain MRI volumes to be embedded within a 30MB model, and remains recoverable under mitigations that disrupt prior bit-level schemes. While reconstructions are approximate rather than pixel-exact, embedded content remains anatomically recognizable and recoverable at scale, exposing a privacy risk distinct from prior bit-level approaches. Experiments on MIMIC-CXR, BraTS, and LiTS demonstrate effectiveness across modalities, tasks, and architectures, highlighting the need for structural defenses beyond parameter-level sanitization. Code is available at this https URL.
[CV-477] SWT: Self-Supervised Video Object Segmentation via Sliding Wavelet and Transportation
链接: https://arxiv.org/abs/2609.31725
作者: Zhengtong Zhu,Jiaqing Fan,Hanwen Qian,Fanzhang Li
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Video Object Segmentation (VOS) aims to accurately segment target objects from consecutive video frames and track the changes of the objects in each frame of the video. Conventional VOS methods typically demand substantial quantities of pixel-level labeled video sequences for fully supervised learning, which limits the performance of the model in sparse video scenes, while existing VOS methods have limited adaptability to global changes in objects. Based on this observation, in this paper, we propose self-supervised VOS with Sliding window, Wavelet transform and optimal Transport (SWT), a self-supervised VOS framework entirely trained on static dataset using contrastive learning. Firstly, a rolling sample buffer reuses overlapping groups of independently sampled images across successive updates. Secondly, to address the long-distance modeling difficulty caused by simple convolutional structures, we introduce wavelet transform to expand the receptive field of convolutional kernels, thus improving the model’s representational capability. Finally, we incorporate optimal transport to assist the model in finding the globally optimal match between the target across two frames, improving the model’s ability to handle nonrigid deformations of objects. SWT only requires training on the COCO dataset once and achieves excellent results on five VOS datasets as well as an additional body part propagation dataset. The code will be released soon at [this https URL](this https URL).
[CV-478] Frequency-Domain AI-Generated Image Detection: Exploring Decoder and Channel Attention for Feature Refinement
链接: https://arxiv.org/abs/2609.31723
作者: Uday Shankar Roy,Mahbuba Jahan Minu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at the 2026 IEEE International Conference on Optics, Machine Learning and Emerging Technology (OMLET)
Abstract:With the rapid progress of AI, the number of AI-generated images has increased significantly in recent years. However, the increasing variety of image generation models makes detection more difficult. In this work, we use Fast Fourier Transform (FFT) representation with EfficientNet-B0 for AI-generated image detection. EfficientNet-B0 provides a lightweight architecture that can be useful for resource-limited applications. Most frequency-domain detectors use a standard encoder to extract features from the FFT spectrum and directly pass them to a classifier. We explored a different approach by investigating ECA, U-Net, and Attention U-Net as alternatives to this direct encoder-to-classifier approach. ECA applies channel attention, while U-Net and Attention U-Net use decoder-based architectures to recover and refine spatial information in the extracted frequency features. We used a balanced subset of the MS COCOAI dataset that includes AI-generated images from five different models. Three runs were carried out for each experiment, and the average values were recorded. Experimental results indicate that EfficientNet-B0 obtained an accuracy of 84.64%, which is 4.50 percentage points higher than the ResNet-50 baseline reported in the dataset paper. EfficientNet-B0 with U-Net provided a small improvement, achieving an accuracy of 84.85%, while ECA did not increase the overall performance. EfficientNet-B0 with Attention U-Net achieved the best overall performance, with an accuracy of 85.51% and an ROC-AUC of 92.99%. This represents an improvement of 0.87 percentage points in accuracy compared to the EfficientNet-B0 baseline and 5.37 percentage points over the ResNet-50 baseline reported in the dataset paper.
[CV-479] Where Does the Watermark Hide? Push-Pull Disentanglement for Invisible Watermark Removal
链接: https://arxiv.org/abs/2609.31722
作者: Jidong Yang,Huaike Yu,Qi Li,Chunpeng Wang,Yuantian Miao,Suo Gao,Xiao Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Fixed image distortions do not cover an attacker that learns from paired clean and watermarked images. We study this paired-training threat with single-image inference: deployment uses neither the clean reference nor the watermark key, payload, or decoder. An encoder maps each image to a structural latent g and an auxiliary residual latent u . Push supervision reconstructs the watermarked image from D(g_w,u_w) . Pull supervision trains the zero-auxiliary output D(A_g(g_w;k),0) toward the paired clean image. At k=1.10,u=0 , the four-method sweep gives an average BER of 0.3958 , PSNR of 31.07 dB, and SSIM of 0.9554 . Restoring u from 0 to 0.15 moves average BER from 0.3893 to 0.3357 , while PSNR falls from 30.99 to 28.23 dB. The intervention supports decoder dependence on the auxiliary input in the evaluated setting. The accompanying theory is a conditional, post-hoc account of this behavior rather than an experimentally verified information-relocation result.
[CV-480] LatentReRig: An SDF-Based VAE with Dual Decoders for Latent-Space Deformation Conditioning
链接: https://arxiv.org/abs/2609.31720
作者: Daniele Dolci,Fabrizio Poggioni,Carlo Melchiorri
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 13 pages, 11 figures, extracted from final dissertation
Abstract:Transferring deformation between characters with different geometry and topology is challenging because conventional rigs encode behaviour through character-specific structures and correspondences. We present LatentReRig, an experimental framework that investigates whether pose-associated changes can instead be represented as reusable directions in a learned geometric latent space. An SDF-based variational autoencoder is coupled with two decoders: one reconstructs the implicit field, while the other predicts target vertex positions from source geometry and latent deformation conditioning. The source geometry may be neutral or already deformed. Experiments on a controlled humanoid dataset show that several poses induce coherent latent directions across identities, particularly for broad articulated motions. These signals can guide deformation of unseen characters, but explicit predictions remain less accurate for localized changes and corrective contributions. Diagnostic comparisons with repeated SDF sampling show that inter-identity distances exceed same-geometry resampling variability on average, while pose signals exhibit different margins above this baseline. The results support the presence of reusable pose-related structure and identify stable local conditioning and accurate mesh decoding as complementary requirements for improving transfer.
[CV-481] PanoFuse: Panorama-Enhanced Vision-Language-Action Learning with Decoupled Semantic-Geometric Routing
链接: https://arxiv.org/abs/2609.31717
作者: Peng Xu,Haoran Lin,Wanjun Jia,Kai Luo,Wenrui Chen,Zhiyong Li,Kailun Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO); Image and Video Processing (eess.IV)
备注: Code and data will be released publicly at this https URL
Abstract:Vision-Language-Action (VLA) policies have shown promising performance in language-conditioned robotic manipulation. However, most existing VLA systems rely on conventional perspective cameras with limited fields of view, often missing global scene context and leading to unreliable manipulation under visual occlusions, distractors, and unseen environments. In this work, we propose PanoFuse, a panorama-enhanced VLA framework that complements local manipulation observations with global panoramic perception. PanoFuse introduces a dedicated panoramic branch that leverages a pretrained panoramic foundation model to extract complementary semantic and geometric representations from omnidirectional observations. Rather than directly mixing these heterogeneous features, we introduce Decoupled Semantic-Geometric Routing (DSGR), which maintains semantic and geometric representations as separate context streams and selectively routes both to downstream state and action representations through structured block-wise attention. This design provides the action expert with global spatial context while preserving task-relevant semantic information from the pretrained VLA backbone. We further develop a synchronized data collection pipeline and construct a new real-world manipulation dataset containing panoramic RGB observations, wrist-view images, language instructions, robot states, and actions. Across seven evaluation settings, PanoFuse achieves an average success rate of 52.9%, outperforming the evaluated baselines and achieving consistent gains under novel-object, unseen-background, and distractor-rich settings. Code and data will be released publicly at this https URL.
[CV-482] PanOVOcc: Panoramic Embodied Open-Vocabulary Occupancy Mapping with Long-term Spatial Voxel Memory
链接: https://arxiv.org/abs/2609.31716
作者: Di Kuang,Mengfei Duan,Yuhang Wang,Weixing Peng,Kailun Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO); Image and Video Processing (eess.IV)
备注: The source code and the established benchmarks will be available at this https URL
Abstract:Persistent semantic occupancy mapping is essential for embodied scene understanding. However, perspective-based systems provide limited spatial coverage, while existing panoramic methods primarily predict local volumes from single observations. We introduce PanOVOcc, a training-free framework for persistent open-vocabulary semantic occupancy mapping from panoramic sequences. PanOVOcc unifies panoramic SLAM, open-vocabulary perception, and long-term spatial voxel memory within an online architecture, continuously integrating geometric and semantic evidence into a global, language-queryable map. To facilitate systematic evaluation of this setting, we establish Pan-Replica and Pan-Holo360D, two benchmarks pairing continuous panoramic RGB-D sequences with scene-level semantic occupancy ground truth across synthetic and real-world scenes. Compared with the strongest evaluated baseline for each metric, PanOVOcc improves occupancy IoU and semantic mIoU by absolute +20.03 and +7.06 on Pan-Replica, and by +43.26 and +20.16 on Pan-Holo360D, respectively. The source code and the established benchmarks will be available at this https URL.
[CV-483] Fysiverse-3D-SimReady Technical Report: Agent ic Physical Simulation for Prag matic 3D World Reconstruction
链接: https://arxiv.org/abs/2609.31715
作者: Lintao Wang,Mingyang Sun,Yang Liu,Dingkang Yang,Lihua Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Fyscis AI Technical Report
Abstract:Agentic recognition requires visual perception to move beyond static scene understanding and produce structured scene representations that support the perception–reasoning–action loop. Existing single-image 3D generation methods, however, mainly produce visually plausible object assets rather than simulation-ready scene states. When independently generated meshes are composed in a shared space, they may fail to align with the input camera, violate gravity, interpenetrate nearby objects, or become unstable under physics simulation. We present Fysiverse-3D-SimReady, a grounded refinement framework for reconstructing simulation-ready multi-object scenes from a single RGB image with instance and ground prompts. The method places generated object meshes into a shared gravity-aligned scene, refines their camera-space poses through differentiable rendering, and corrects scene-level supports and contacts for stable physical execution. The scene is then used by an agentic simulation workflow, which converts a scene-specific task goal into an executable physics rollout rendered from the original camera view. Experiments show that Fysiverse-3D-SimReady improves input-view alignment, contact plausibility, and physical stability over existing single-image reconstruction and scene generation baselines, while enabling goal-conditioned physical interactions from a single image.
[CV-484] OmniFysics-Captioner Technical Report: Grounding Omni-Modal Understanding in the Physical World for Better Captioning
链接: https://arxiv.org/abs/2609.31714
作者: Kaixiang Qiu,Minghao Han,Keliang Liu,Yizhou Liu,Jinghan Han,Yue Jiang,Xuecheng Wu,Shunli Wang,Lihua Zhang,Dingkang Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Fysics AI Technical Report
Abstract:Building omni-modal models with physical intelligence requires fine-grained supervision that captures physical evidence such as contact, support, deformation, and state transitions. However, existing omni-modal captioners primarily model general audiovisual semantics and often overlook transient or spatially localized physical evidence. We present a unified framework for physics-aware audiovisual captioning spanning data construction, training, and evaluation. Firstly, we build a data construction pipeline that identifies physics-rich clips and leverages OmniFysics-Agent to coordinate audio, visual, and physical-perception tools for collecting spatiotemporally aligned and traceable cross-modal evidence; within the Agent, a physical perception model (PPM) fine-tuned on approximately 2M image-level samples serves as a dedicated tool for extracting object-interaction and state-change cues. Secondly, we build the Daily-Physics 50K dataset and introduce the evidence-driven OmniPhysCap (OPC) benchmark to evaluate the recovery of physical and cross-modal evidence from generated captions. Finally, we train OmniFysics-Captioner from the resulting data. Our Captioner matches Gemini 3.1 Pro on audiovisual captioning, achieves state-of-the-art results on multiple video-captioning benchmarks, and substantially outperforms other open-source models. Ablations show that PPM evidence improves physical coverage and produces finer-grained, more reliable cross-modal descriptions.
[CV-485] Agent ic Video Understanding: A Survey
链接: https://arxiv.org/abs/2609.31713
作者: Xinyu Deng,Siwen Luo,Daochang Liu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 15 pages,3 figures, Accepted to DICTA 2026
Abstract:As large language models (LLMs) become capable of processing increasingly diverse modalities and longer temporal contexts, an emerging line of work is moving beyond fixed video-language inference toward agentic systems that actively decide what information to inspect, retain, verify, and act upon. This survey reviews video understanding agents: systems that use video as the primary information source and solve understanding tasks through adaptive state construction and action selection. We first formalize an agent loop for video understanding, then address a central question: why do agents matter for video understanding? To answer this, we organize the literature through a challenge-to-design taxonomy, linking context bottlenecks to hierarchical evidence memory, evidence sparsity to active evidence acquisition, temporal causality to state and process tracking, and multimodal ambiguity to role-specialized coordination. We further review state space paradigms, learning paradigms, supervision signals, benchmarks, and evaluation protocols. Finally, we identify open directions toward agentic-native temporal modeling and video-native agents. Project page: this https URL
[CV-486] Statistical Testing for Multiple Instance Learning via Selective Inference with Applications to Computational Pathology
链接: https://arxiv.org/abs/2609.31712
作者: Noriaki Hashimoto,Shuichi Nishino,Teruyuki Katsuoka,Tomohiro Shiraishi,Daiki Miwa,Hiroyuki Hanada,Jun Sakuma,Hidekata Hontani,Hiroaki Miyoshi,Ichiro Takeuchi
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注: 37 pages, 11 figures
Abstract:Multiple instance learning (MIL) is widely used in computational pathology because it enables weakly supervised analysis of whole-slide images (WSIs) without requiring patch-level annotations. In attention-based MIL, instances with high attention scores are often interpreted as diagnostically important regions and used as visual explanations. However, attention scores alone cannot determine whether selected high-attention instances are significantly different from normal instances, limiting the reliability of attention-based explanations. In this paper, we formulate the evaluation of high-attention instances as a statistical hypothesis testing problem. Specifically, we assess whether a selected high-attention instance significantly deviates from a representative normal reference instance selected based on feature similarity. A major challenge is that both the target instance and the reference instance are selected through data-dependent procedures, rendering standard hypothesis testing invalid. To address this issue, we introduce a selective inference (SI) framework that explicitly accounts for the selection events induced by attention-based instance selection and adaptive reference selection, thereby enabling the computation of valid selective p -values conditional on these events. Experiments demonstrate Type-I error control on synthetic and MNIST-based data and practical applicability to pathological WSIs, with higher statistical power than the conventional over-conditioning approach.
[CV-487] Cross-Dataset Generalization of Bangladeshi Rice Leaf Disease Classifiers: Benchmark Diagnosis and Mitigation
链接: https://arxiv.org/abs/2609.31709
作者: Anindya Paul
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages, 6 figures, 7 tables
Abstract:Cross-dataset transfer in rice leaf disease classification remains a significant challenge, with models trained on one image collection performing substantially worse when deployed on another. We conduct a systematic benchmark across three Bangladeshi rice leaf disease datasets (5,419 images, 6 transfer pairs, 3 CNN backbones, 3 random seeds) to characterize and diagnose this failure. Strong augmentation recovers a mean cross-dataset macro-F1 improvement of +0.070 (Wilcoxon p 0.001, 15 of 18 transfer pairs positive). Removing non-leaf image content via segmentation shows directional benefit (mean +0.066, p = 0.062, n = 36 paired observations) that is consistent across two independent segmentation methods but does not reach conventional significance. A self-supervised ViT control (DINOv2 linear probe) exhibits equivalent cross-dataset collapse to CNNs, ruling out architecture inductive bias as the primary driver and pointing to acquisition-condition shift. Adaptive batch normalization uniformly harms transfer performance, with harm magnitude correlating with source-target label-prior divergence and model depth (Spearman rho = 0.621, p = 0.009). Grad-CAM attribution analysis on 12 sampled predictions does not distinguish correct from incorrect cross-domain predictions (p = 0.462), indicating that common attribution proxies are insufficient for diagnosing shift at practical sample sizes. We document all frozen results, prespecified analysis criteria, and reproducibility artifacts in a public repository with SHA-256 integrity verification. This work establishes a rigorous empirical baseline for understanding cross-dataset generalization in agricultural computer vision and identifies both effective (augmentation) and ineffective (AdaBN) adaptation strategies.
[CV-488] Cant Find Waldo: Evaluating VLMs Sensitivity to Image Resolution and Detail Level
链接: https://arxiv.org/abs/2609.31706
作者: Alexandra Schild,Gerard de Melo
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Visual Language Models (VLMs) have achieved remarkable success across diverse tasks, yet they struggle with high-resolution inputs where critical information resides in small regions or detailed, cluttered scenes. While several approaches address this limitation, a systematic understanding of why models fail at high resolutions is lacking. We introduce a controlled evaluation framework that disentangles resolution-related performance degradation from task difficulty through semantics-preserving transformations. We propose two simple metrics: Area Under the Scaling Curve (AUSC), which quantifies scaling robustness independent of baseline accuracy, and Prediction Variance Score (PVS), which measures resolution-induced prediction instability. Through comprehensive experiments across 5 model families and 5 benchmarks, we identify three primary failure modes: (1) information loss from downsampling at vision token limits, (2) tokenization artifacts from patch boundary shifts and positional encoding fragility under non-standard aspect ratios, and (3) attention dilution as token counts increase. Our analysis reveals that even state-of-the-art models suffer from performance drops when processing high-resolution images, with degradation patterns varying systematically by architectural family. We provide actionable insights for model architecture design and data augmentation strategies to mitigate these limitations.
[CV-489] CLC-YOLO: A Compact Channel-Gated Prototype Network for Real-Time Leakage-Aware Breast Ultrasound Lesion Segmentation
链接: https://arxiv.org/abs/2609.31702
作者: M. Fazri Nizar,Muhammad Naufal Rachmatullah,Julian Supardi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to IEEE ICAITech 2026
Abstract:Reliable breast ultrasound lesion segmentation requires accurate boundaries and evaluation that prevents patients or duplicate images from crossing data splits. We propose Channel Local Contrast (CLC), a compact refinement of the YOLO26 segmentation prototype head. CLC adds a fixed local high-pass residual controlled by 64 zero-initialized, bounded channel gates. Baseline and CLC were compared in five matched folds on each of four breast ultrasound datasets. BUS-BRA used patient-disjoint outer tests with separate inner validation. BUS-UCLM and BrEaST used patient-grouped validation folds; BUSI used duplicate-component groups because patient identifiers are unavailable. Group-macro Dice increased by 1.68, 3.11, 1.12, and 2.45 percentage points on BUS-BRA, BUS-UCLM, BUSI, and BrEaST, respectively. Only the BUS-BRA paired 95% confidence interval excluded zero. CLC adds 64 parameters and 0.0049 giga floating-point operations (GFLOPs). At 640 pixels, single-T4, batch-one, 16-bit floating-point (FP16) TensorRT graph times were 2.1478 ms for CLC and 2.0512 ms for baseline, excluding preprocessing and postprocessing. CLC increased group-macro Dice across all four datasets with a measured T4 forward-pass overhead of 0.0966 ms. Code: this https URL
[CV-490] Integrated Deep Learning Framework Designed on Hybrid Optimization Strategies for Automated Health Detection and Analysis in Silkworms
链接: https://arxiv.org/abs/2609.31701
作者: Komala K V,Lata B T,Venugopal K R
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:A Hybrid Residual Attention Network is proposed for accurately classifying silkworm images into six different classes, including healthy and diseased states. It uses residual blocks for deep feature extraction and attention to focus on disease related features. A novel Integrated Adaptive Momentum Optimizer was introduced to enhance convergence and improve training efficiency. The dataset of silkworm images underwent preprocessing techniques such as normalization, resizing, and noise reduction, along with augmentation strategies to improve data quality and diversity. It is optimized using IAMO, achieved an accuracy of 98.67%.The integration of spatial and channel wise attention mechanisms, coupled with IAMO, significantly enhanced the model ability to recognize subtle differences between classes. Results indicate that HRAN can be used to detect disease at an early stage in sericulture, and future work will enhance scalability and efficiency in different environments.
[CV-491] Modernising the Compressed-Domain Video Captioner: A Controlled Study of SigLIP2 and GPT -2 Substitutions
链接: https://arxiv.org/abs/2609.31700
作者: Ashim Nepal,Ashok B.K
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 8 pages, 1 figure, 6 tables. Code, configurations and the exact commit for every run: this https URL
Abstract:Compressed-domain video captioning avoids full video decoding by operating directly on I-frames, motion vectors and residuals, trading a small amount of accuracy for a large gain in inference speed. CoCap established this pipeline using a CLIP vision encoder and a shallow BERT-style multimodal decoder. Both components predate substantially stronger alternatives. We ask a narrow, controlled question: how much of CoCap’s accuracy is limited by these two components, and which of the two is the binding constraint? We replace the CLIP I-frame encoder with SigLIP2 and the BERT-style decoder with GPT-2, and evaluate three configurations (the original pairing, the encoder substitution alone, and both substitutions together) under identical data, sampling budget and optimisation schedule. All comparisons are made against our own reproduction of CoCap rather than its published numbers, because we train on a 4,999-clip subset of VATEX at a reduced sampling budget; absolute values are therefore not comparable with the original work. Our reproduction tracks the published result closely: CIDEr and METEOR run slightly above it (54.9 against 52.7; 23.4 against 23.2), BLEU-4 and ROUGE-L slightly below (29.7 against 31.4; 48.9 against 49.4). We attribute the differences to our evaluation subset rather than to any improvement in either direction. We find the two substitutions pull in opposite directions: SigLIP2 alone improves every metric (+4.5 CIDEr), while adding GPT-2 on top erodes that gain, because a pretrained decoder overfits 4,999 clips within two epochs. We additionally report inference latency for each configuration, since speed is the property that motivates compressed-domain captioning in the first place, and an accuracy gain purchased at a latency cost should be reported as such. Comments: 8 pages, 1 figure, 6 tables. Code, configurations and the exact commit for every run: this https URL Subjects: Computer Vision and Pattern Recognition (cs.CV) MSC classes: 68T45 ACMclasses: I.2.10; I.2.7; I.4.2 Cite as: arXiv:2609.31700 [cs.CV] (or arXiv:2609.31700v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.31700 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-492] Normalise or condition? Noise-floor front-ends for on-board keyword spotting under UAV rotor ego-noise
链接: https://arxiv.org/abs/2609.31699
作者: Yida Lin,Bing Xue,Mengjie Zhang,Sam Schofield,Richard Green
类目: ound (cs.SD); Computer Vision and Pattern Recognition (cs.CV); Audio and Speech Processing (eess.AS)
备注:
Abstract:A microphone on the airframe of a small multi-rotor UAV is dominated by rotor ego-noise, so spoken flight commands arrive at negative signal-to-noise ratio (SNR). We study small-footprint keyword spotting (KWS) for a ten-word command vocabulary under real ego-noise, training on one quadrotor and testing on another. Besides per-clip accuracy we measure the streaming false-alarm rate on 4.4 h of continuous rotor noise. We compare classical noise-robust front-ends (CMN, PCEN, spectral subtraction), test-time adaptation, and two front-ends that track the per-band ego-noise floor over the two seconds preceding the decision window and either subtract it (normalisation) or feed it to the network as a second input channel (conditioning). Per clip, all front-ends look alike: +2-3 points on average, up to +14 at -15 dB. On continuous rotor noise they differ sharply. Normalising front-ends fire about ten times more often than plain log-mel at the same threshold and end up below it at a budget of one false alarm per hour (-9 points at 0 dB). Conditioning keeps the baseline’s false-alarm rate and turns its gain into detections (+8 points at -10 dB over three seeds); on top of PCEN it gives the best per-clip accuracy and false-alarm rate, and a level-anchored variant is also invariant to the microphone gain. Real drone+interferer recordings expose the remaining failure mode, environmental sounds and bystander speech, which training negatives halve. A single script reproduces all on-device numbers on the NVIDIA Jetson Orin NX flight computer, where the complete pipeline costs 3 ms per 100 ms hop on the CPU.
[CV-493] RPA: Residual Patch-Token Adapter for Image Retrieval from EEG and MEG
链接: https://arxiv.org/abs/2609.31698
作者: Yuhui Jin,Yonghao Song,Bingchuan Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 34 pages, 13 figures
Abstract:Most existing MEG and EEG (M/EEG) visual decoding methods align brain signals with a single global embedding extracted from a pretrained visual encoder, leaving open whether intermediate patch representations, which preserve richer and more granular rich visual information, can improve representation learning. To address this question, we introduce the Residual Patch Adapter (RPA), a lightweight, modular adapter that leverages all patch tokens from an intermediate layer of a ViT visual encoder for alignment. Through extensive ablation analyses, we first show that pooling or masking patch tokens degrades the learned representation, demonstrating that retaining the full set of patch tokens is important for EEG alignment, while the CLS token provides little unique information. We then use a series of six quantitative feature analyses to show that both higher-level semantics and lower-level visual features, including color and texture, are essential for this EEG-to-image alignment. Under current protocols, our system achieves Top-1 accuracies of 95.4% within-subject and 35.5% cross-subject on THINGS-EEG2, and 65.2% and 6.7%, respectively, on THINGS-MEG, achieving state-of-the-art (SOTA) performance across both datasets. Evaluations with alternative brain encoders, including pretrained EEG foundation models, demonstrate that the approach extends beyond the projection-based EEG encoder. Furthermore, we provide a plug-and-play interface that allows RPA to be replaced by convolution, attention, or ConvNeXt alternatives. Together, these findings provide significant insight into M/EEG-to-image representation learning by establishing design principles for leveraging the latent space of visual encoders, and open new directions for brain–image alignment and non-invasive brain–computer interface (BCI).
[CV-494] Video Captioning in Low-Light Conditions through Efficient Uncertainty-Aware Caption Correction
链接: https://arxiv.org/abs/2609.31697
作者: Arefeh Rezaei
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Low-light conditions can significantly degrade the ability of vision-language models (VLMs) to accurately describe human actions in videos. In this work, I propose an efficient uncertainty-aware representation correction framework for improving captions generated by VideoChat2 under real-world low-light conditions. Instead of fine-tuning the underlying VLM, the proposed framework introduces a lightweight sparse Gaussian process-based error estimation module between the projection layer and the language model to correct the intermediate representation. The correction module learns to estimate the residual between the original projected representation and a verified target representation, which is then adaptively scaled using a newly formulated uncertainty-aware coefficient and added to the original representation. To further improve residual estimation, I introduce a partitioned combined-kernel design. The correction model is trained separately using only 44 samples from the ARID dataset and requires only a small additional computational overhead during inference. The effectiveness of the proposed correction is evaluated through quantitative residual prediction and qualitative analysis of the generated captions. Although VideoChat2 is used in my experiments, the proposed framework is designed to be applicable to other compatible VLM architectures. \textbfCode Availability: The implementation accompanying this work is publicly available at:\hrefthis https URLthis https URL
[CV-495] RemTraceNet: Few-Shot Forensic Detection of Invisible Watermark Attacks
链接: https://arxiv.org/abs/2609.31694
作者: Jidong Yang,Huaike Yu,Qi Li,Yuantian Miao,Wei Zong,Yang-Wai Chow,Willy Susilo,Chunpeng Wang,Suo Gao
类目: Computer Vision and Pattern Recognition (cs.CV); Cryptography and Security (cs.CR)
备注: 11 pages, 4 figures
Abstract:Removing an invisible watermark and concealing the forensic evidence are distinct objectives: successfully disrupting the embedded watermark does not imply that the removal process is forensically undetectable. When verification fails, removal traces can provide complementary evidence for provenance and ownership verification, whereas their absence leaves the cause of the failure ambiguous. Existing methods are typically evaluated by watermark suppression and perceptual quality, while forensic stealth is rarely considered. We therefore study watermark-attack-specific few-shot forensics: for each known pipeline, a specialist can separate its outputs from paired clean and unattacked watermarked controls. Separate Attack-vs-Clean and Attack-vs-Watermarked evaluations prevent watermark-presence shortcuts. Image-aligned and prompt-matched controls are used for post-hoc and generator-integrated schemes, respectively. In this work, we introduce RemTraceNet, which fuses constrained residuals, local relations, FFT/Haar statistics, and block-DCT evidence at native resolution. Across 23 removal pipelines and 10 watermark configurations, we evaluate native 256 x 256 and 512 x 512 inputs. With 100 attacked training images per pipeline, the three-seed TPR@1%FPR, macro-averaged over attacks and watermark configurations, ranges from 82.75% to 88.17% across resolutions and control types. Under the condition of same labels and protocol, RemTraceNet outperforms retrained SRNet, ZhuNet, and SiaStegNet baselines by 10.80–15.68 percentage points. Extensive experimental results show that erasing a watermark and erasing evidence of its removal are distinct challenges, and that removal traces remain learnable under limited supervision.
[CV-496] Disentangle and Drop: Robust Universal Removal of Image Watermarks via Reconstructive Grayscale Residual Decomposition
链接: https://arxiv.org/abs/2609.31693
作者: Qi Li,Jidong Yang,Feng-Lei Fan,Yuantian Miao,Xiao Chen,Huaike Yu,Chunpeng Wang,Suo Gao,Herbert Ho-Ching Iu,Bin Ma
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 18 pages, 6 figures, and 5 tables
Abstract:Invisible image watermarks are commonly evaluated against benign postprocessing operations such as compression, resizing, blur, and color changes. These tests leave out a different threat: a learned remover that preserves semantic image content while discarding residual evidence that carries the payload. We propose Disentangle and Drop (DnD), an attack that is agnostic to the watermark method and treats watermark removal as a representation routing problem. DnD decomposes a watermarked image into a semantic grayscale carrier and an auxiliary residual branch, and then suppresses the residual branch to reduce watermark evidence. The model is trained with latent spectral perturbations and low-strength diffusion exposure so that the drop operation remains stable under adaptive reconstruction. Experiments on seven representative watermark families show that one shared operating setting gives competitive removal with high visual fidelity. Operating scans and ablations separate usable attacks from image-damaging settings: stronger noise or diffusion can raise removal scores by damaging the image, while the practical regime comes from dropping the residual latent. These results argue for evaluating watermark robustness against learned removal at the representation level, not only against conventional image edits.
[CV-497] Architecture-aware Robustness Evaluation of Explainable Deep Learning for Breast Cancer Diagnosis
链接: https://arxiv.org/abs/2609.31692
作者: Balenthira Thanusanth,Selvarajah Thuseethan,Roshan G. Ragel,Bimali S. Weerakoon,Ayesh Jayasinghe,Ananthamoorthy Krishnamoorthy
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 12 pages, 7 figures
Abstract:Explainable Artificial Intelligence (XAI) has become essential in medical image analysis to ensure transparency of deep learning (DL)-based diagnostic systems. However, selecting appropriate XAI techniques for breast cancer recognition remains largely ad hoc, with limited systematic evaluation across different DL architectures. This study presents a systematic architecture-aware evaluation protocol to assess the effectiveness of nine widely used XAI techniques across four categories of DL models: very deep, lightweight, transformer-based and hybrid neural networks. The evaluation is conducted on a breast ultrasound dataset comprising 780 images using clinically aligned spatial metrics, including Pointing Game, Intersection over Union and Mean Coverage, to quantify agreement between generated explanations and expert-annotated lesion regions. Results indicate that explanation quality is primarily influenced by the interaction between model architecture and XAI method, rather than any single technique consistently outperforming others. Hybrid architectures produce more spatially coherent explanations, while lightweight and transformer-based models exhibit greater variability across methods. The findings show that no single technique generalises across architectures and evaluation criteria, emphasising the need for joint selection of DL models and XAI techniques. Explainability depends on both model design and explanation strategy and should not be considered independently. This work provides a structured evaluation protocol and practical guidance for selecting XAI techniques in breast cancer diagnosis, supporting more transparent clinical decision-support systems. \textcolorblueCode is publicly available at this https URL
[CV-498] Adapting Vision-Language Models for Human-Readable XAI in Industrial Object Detection ICME’2025
链接: https://arxiv.org/abs/2609.31690
作者: Sarvenaz Sardari,Freddy Fernandes,Samarth Yelvande,Jose Moises Araya-Martinez,Alina Roitberg
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Procedia CIRP ICME '2025
Abstract:Explainable Artificial Intelligence (XAI) solutions are essential for building trust in AI technologies and their integration in real manufacturing lines. However, most existing methods are tailored to technical experts, limiting their accessibility to diverse user groups such as blue-collar workers in manufacturing lines who use AI for quality control. In this work, we introduce an XAI interface for object detection in industrial manufacturing based on a fine-tuned vision-language model, designed to generate intuitive explanations for non-expert users. We benchmark existing vision-language models and demonstrate that out-of-the-box models often fall short in delivering clear, context-relevant explanations for non-expert users. To address this, we fine-tune a vision-language model and integrate it into our interface, enabling contextualized, accessible explanations for non-expert users. We demonstrate improvements in explanation clarity, instruction adherence, image groundedness, and contextual awareness over GPT 4o-mini on proprietary and public robotics dataset. This approach advances the accessibility and usability of AI explanations, making them more intuitive and applicable in manufacturing domain.
[CV-499] owards Transparent Diagnostics: Investigating Architectural Trade-offs and Explainability in Malaria Detection
链接: https://arxiv.org/abs/2609.31682
作者: Suman Kunwar,Avishek Dangol
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 13 pages, 9 figures
Abstract:More than 80 countries have reported malaria cases with 610 thousand deaths and are projected to increase. Identifying malaria early and accurately helps save lives. The effective way to diagnose malaria is through microscopic methods that are labor intensive and require experts with special equipment. Deep learning (DL) has shown promising results in medical diagnosis. Here, we explored various DL models: ResNet18, MobileNetV2, EfficientNet-B2, VGG19 and also proposed a model for detecting malaria presence using blood smears taken from the NIH Malaria dataset. Our experiment shows that MobileNetV2 achieved 96.35% accuracy with the smallest model size (8.49 MB) and fastest inference (1.35 ms). The proposed model achieved 97.67% accuracy, 0.9756 AUC with longest inference time (13.17 ms). The larger architecture outputs a larger model size with moderate accuracy. Upon further pruning, the proposed model gained a slight improvement in accuracy and inference time. The GRAD-CAM, SHAP and LIME shade explainable AI (XAI) insights of the model.
[CV-500] Devanagari Handwritten Character Recognition Using TrOCR: A Transformer-Based Model with Real-Time Web Deployment
链接: https://arxiv.org/abs/2609.31681
作者: Amrit Baskota,Samyam Budhathoki,Shubham Ghimire,Abiskar Ghimire,Sarwesh Phuyal,Baskaran P((1) Vellore Institute of Technology)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Presented and accepted at the 16th International Conference on Computing, Communication and Networking Technologies (ICCCNT), 2025
Abstract:Devnagari is a one of the ancient language of the Indian subcontinent consisting of 36 vowels, 14 consonants and 10 numerals. The accurate recognition of handwritten Devnagari characters is challenging due to high complexity of Devnagari scripts. This paper presents a method to fine tune the pre trained TrOCR model to accurately recognize Devanagari handwritten characters. The Methodology consists of a preprocessing mechanism where input images are standardized into RGB format, tokenized in batches and integrated with Hugging Face Dataset. The pre-trained microsoft/trocr-base-handwritten model is fine-tuned over a dataset of nearly 5000 character images that are uniformly partitioned in the ratio 8:1:1 for training, evaluation and testing. Further optimization is done is the training process through mixed precision training, gradient checkpointing, and early stopping mechanism. The model achieves a character error rate (CER) of 3.95% and a character-level accuracy of 96.05%, outperforming the previous CNN based models. A scalable web application is developed using this http URL, Golang, and FastAPI which practically deploys the OCR model and serves character recognition task with a latency less than 5 seconds per request. This study demonstrates the use of TrOCR model to build a scalable handwritten Devnagari character recognition system and also a foundation to future research on Devnagari Script Recognition using Transformers.
[CV-501] oward AI-Assisted Poultry Coccidiosis Diagnosis: Evaluating Gemini and BiomedParse on Eimeria Microscopy Images
链接: https://arxiv.org/abs/2609.31679
作者: Ali Alsalama,Ahmed Kubba,Manar Abu Talib
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 5 pages, 3 figures, 3 tables, accepted at IEEE International Conference on Sustainability, Innovation and Technology (ICSIT 2026)
Abstract:Coccidiosis caused by Eimeria parasites is a major economic burden in poultry production, and effective control depends on accurate species-level diagnosis. This study evaluates whether a general-purpose multimodal large language model can support such diagnosis. Google Gemini was assessed on 4,225 mi- croscopy images covering the seven fowl-infecting Eimeria species under two prompting conditions, one without candidate labels and one with a predefined class list, and was further tested for pathology-report generation, while BiomedParse was examined for parasite segmentation. Without candidate labels, the model produced broad and taxonomically inconsistent outputs. With candidate labels, overall accuracy reached only 14.9%, with a strong bias toward E. tenella at 74% and no correct classifications for E. acervulina, E. mitis and E. praecox. Generated treatment reports were coherent but unverified, and segmentation was only partial. Current multimodal models are therefore not yet reliable for standalone Eimeria diagnosis without domain-specific fine- tuning and expert validation.
[CV-502] A Comparative Transfer-Learning Study of CNN Backbones for Partial Face Recognition on the SoF Dataset CEC
链接: https://arxiv.org/abs/2609.31677
作者: Ahmed Kubba,Ali Alsalama,Abdelrahman Abdalla,Qassim Nasir,Manar Abu Talib
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 6 pages, 1 figure, 3 tables, accepted at The 6th International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME 2026)
Abstract:Face recognition is widely deployed in surveillance, access control, and forensic workflows, yet accuracy degrades sharply once the face is occluded by accessories, foreground objects, or the frame edge. Because most faces in the wild are partial, robust partial face recognition (PFR) remains open. This paper compares three pretrained convolutional backbones, ResNet-50, VGG-16, and FaceNet, fine-tuned for PFR by transfer learning under identical preprocessing, splitting, and optimization protocols on the Specs-on-Faces (SoF) dataset. All three arms use a common 160x160 input and a frozen backbone with a trainable head under a fixed epoch budget and no per-backbone hyperparameter search. The FaceNet configuration, denoted PFN (Partial FaceNet), substantially outperforms the other two, reaching 97.4% test accuracy with macro-averaged 87.04% precision, 84.61% recall, and 84.17% F1 over the 112 identity classes, the highest accuracy and recall reported on SoF.
[CV-503] Unsupervised spiking feature learning for event-based pedestrian crossing detection: approaching supervised accuracy without labelled training data
链接: https://arxiv.org/abs/2609.31671
作者: Henok Teklu,Mustafa Sakhai,Matej Mertik,Maciej Wielgosz
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注: 18 pages, 4 figures, 3 tables
Abstract:Event cameras are well suited to pedestrian crossing detection, and spiking neural networks (SNNs) can process their output natively, but current SNN detectors are trained with supervised backpropagation and therefore require costly frame-level crossing labels. We investigate crossing detection with no labels in feature learning and report the first unsupervised results on the recent DVS-PedX pedestrian-crossing benchmark. A single spiking layer trained with winner-take-all spike-timing-dependent plasticity learns a dictionary from unlabelled event-frame patches; frames are encoded by cosine similarity to the learned filters with spatial pooling and read out by a linear classifier, the only supervised component. On the 24,454-frame test split the method attains 90.3% accuracy and an area under the receiver operating characteristic curve (AUROC) of 0.936, compared with 92.0% and 0.943 for a supervised spiking network trained end-to-end on the same frames; under adverse weather the AUROC is 0.913. The result is insensitive to the choice of plasticity rule but depends strongly on the readout protocol: with the classical neuron-assignment readout the same network attains only 0.70 AUROC. On the benchmark’s real converted portion, a readout refit lifts performance from chance (0.53) to 0.67 AUROC, within the published supervised range. These findings indicate that the accuracy cost of removing labels from feature learning is small on this benchmark, and that reported weaknesses of unsupervised spiking networks may be attributable to the readout protocol rather than to the learning rule.
[CV-504] One-Step Is Optimal: Unconditional Rectified Flows are Noise2Noise Denoisers and Multi-Step Integration Provably Hurts—A Benchmark and Task-Based Detectability Study on Low-Dose CT AAAI
链接: https://arxiv.org/abs/2609.31670
作者: Timothy Sereda,Debesh Jha
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages, 3 figures, submitted to AAAI
Abstract:Iterative and generative denoisers are increasingly used under the assumption that multi-step refinement outperforms a single regression pass. We show the opposite for \emphlabel-free denoising. An \emphunconditional rectified flow trained on two noisy observations of the same signal, as in Noise2Noise, has a minimiser whose one-step readout is exactly the MMSE denoiser without requiring clean targets. In contrast, multi-step integration provably departs from the MMSE solution because the flow terminates at the noisy data distribution rather than the clean-signal distribution. This departure is exact in a tractable Gaussian model and is confirmed experimentally: one-step flow matches a direct regressor, whereas multi-step Euler integration progressively reduces fidelity. Counterintuitively, the degradation increases with training quality, as a better velocity field more faithfully transports samples toward the noisy terminal law. The key ingredient is therefore the decorrelated \emphpairing, not the flow machinery: a one-step regressor trained on matched noisy pairs gives our best label-free result ( +1.99 ,dB). We evaluate these findings on \textbfCTDenoiser, a controlled low-dose CT benchmark spanning five architectures and supervised, similarity-based, blind-spot, and per-image methods. Among label-free approaches, only correlated-noise-aware Noise2Sim improves over the noisy baseline, while Noise2Void is flat-to-negative because CT noise violates its pixel-independence assumption. Finally, although supervised denoisers gain approximately 4 ,dB PSNR, a channelized Hotelling observer shows reduced low-contrast lesion detectability, revealing clinically relevant degradation missed by PSNR and SSIM.
[CV-505] Query-aligned video frame selection for long video understanding
链接: https://arxiv.org/abs/2609.31668
作者: Md. Safayet Islam,Dilip Sarkar,Liang Liang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 16 pages, 5 figures, 22 references
Abstract:Multimodal large language models (MLLMs) process multimodal inputs by converting text, images, and videos into token sequences that are subsequently processed by a backbone language model. While MLLMs have achieved excellent performance in understanding the content of individual images, video understanding remains significantly more difficult because videos contain large number of video frames. MLLMs typically process only a subset of these frames, usually ranging from 8 to 64. MLLMs usually sample frames uniformly, regardless of their relevance to the question being answered. To address this limitation, several training-free, model-agnostic methods for selecting question-relevant frames have recently been proposed. In this work, we introduce a frame-selection method designed specifically for multiple-choice questions. We extend the query text by appending semantic cues derived from the answer choices and employ a direct query-frame alignment scoring mechanism. To the best of our knowledge, our method is the first to directly utilize answer choices as inference-time cues for selecting frames relevant to answering a question. The method first constructs a compact candidate pool by subsampling video frames at a fixed rate. The frames are then scored according to their maximum cosine similarity across all question-answer pairs to identify the most relevant frames for a given query. This approach preserves a fixed token budget while improving the relevance of the visual evidence provided to the downstream MLLM. We evaluate the effectiveness of our frame-selection method on the MLVU, Video-MME, and LongVideoBench benchmarks using three MLLMs: LLaVA-Mini, Qwen2-VL, and LLaVA-Video. Experimental results demonstrate that answer-aware frame selection generally outperforms uniform sampling and existing training-free frame-selection methods under the same frame budget. Comments: 16 pages, 5 figures, 22 references Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.31668 [cs.CV] (or arXiv:2609.31668v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.31668 Focus to learn more arXiv-issued DOI via DataCite
[CV-506] Language-Augmented Video Action Anticipation: Design Fundamentals Benchmarks and Open Challenges ICIP
链接: https://arxiv.org/abs/2609.31665
作者: Mahsa Mohammadi,Zeyu Fu,Sareh Rowlands
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 29 pages, 5 figures, 19 tables. Review article. Supplementary material, machine-readable data, and public artifacts are available at this https URL
Abstract:Action anticipation predicts future human actions from partial video under incomplete context and temporal uncertainty. Recent systems introduce large language models (LLMs), vision-language models (VLMs), or language-derived semantics at different stages, but reported gains are difficult to interpret when task formulation, visual pretraining, supervision, decoder design, and evaluation code change simultaneously. The central contribution of this review is an evidence-aware design map that crosses task regime with the point at which language-derived information intervenes. We characterise task regimes along six axes. These axes organise the literature into five broad task families: single-action, sequence, object-interaction, cross-view, and planning-oriented settings. C1-C3 locate interventions in context construction, goal/intention modelling, and future decoding, while C4 is treated as an adjacent, emerging grounding/executability extension. Unlike a generic processing pipeline, the map links each intervention to an appropriate counterfactual, failure diagnosis, and permissible evidence claim. Supporting contributions include a protocol-level audit of Ego4D-LTA and EPIC-KITCHENS-100, a multidimensional evidence profile, and the Backbone-Aware Comparison and Ablation Protocol (BCAP). The unresolved EK-100 record is treated as a reporting-comparability case study and is not used as a leaderboard. Evidence for LLM benefits, goal ambiguity, and horizon effects is therefore formulated as testable hypotheses requiring matched validation, not as causal conclusions. The accompanying package contains the coded evidence, source locators, protocol metadata, and versioned catalogue used in the review.
[CV-507] Learning Steadily: Accumulating Relative Point Margin Scores for Face Image Quality Assessment
链接: https://arxiv.org/abs/2609.31662
作者: Guray Ozgur,Tahar Chettaoui,Eduarda Caldeira,Marco Huber,Jan Niklas Kolf,Naser Damer,Fadi Boutros
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted for publication in IEEE Transactions on Biometrics, Behavior, and Identity Science
Abstract:Face Image Quality Assessment determines the suitability of captured face images for automated face recognition (FR), a critical capability for reliable biometric systems. Existing state-of-the-art FR-integrated FIQA methods suffer from temporal instability: as the feature space evolves during training, single-epoch quality estimates fluctuate, creating a moving target that undermines reliable quality prediction. We introduce CARPM-FIQA, a stabilization strategy for FR-integrated FIQA that accumulates relative point margin measurements, the ratio between intra-class compactness and inter-class separation, across the entire training trajectory rather than relying on single-epoch estimates. This cumulative averaging approach provides theoretically grounded advantages: reduced variance in quality estimates, improved mean squared error, and enhanced ranking stability with convergence guarantees as training progresses. Through controlled experiments on the SynFIQA dataset with labeled quality groups, we demonstrate that cumulative averaging achieves superior discriminative ability, and ablation studies across different training configurations confirm consistent improvements. Evaluated against twelve FIQA methods on eight challenging benchmarks with four FR models at two FMR thresholds, CARPM-FIQA places 4th (CARPM-FIQA(L)) and 6th (CARPM-FIQA(S)) of 17 compared methods by pAUC-EDC and AUC-EDC averaged across FR models and, after per-benchmark normalization, across benchmarks, staying within a few percent of the best method’s normalized average for every FR model, providing a principled solution to training instability while maintaining the performance benefits of FR integration. More broadly, our work demonstrates that temporal aggregation strategies can stabilize training objectives in deep learning systems where target values inherently fluctuate due to evolving feature representations.
[CV-508] ForensicZoom: Adaptive Visual Inspection with Multimodal LLM s for Industrial-Grade Face Forgery Detection
链接: https://arxiv.org/abs/2609.31661
作者: Hang Zhou,Yiming Tang,Kun Yu,Qian Zhu,Minghao Li,Weigao Wen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reliable face forgery detection is critical to the security of online identity verification systems, where missed attacks compromise security and excessive false positives disrupt legitimate users. Specialized forensic detectors achieve strong detection performance but provide limited interpretability, while multimodal large language models (MLLMs) offer strong semantic understanding and interpretable reasoning yet remain substantially weaker for face forgery detection. We argue that a key limitation lies in how visual evidence is acquired: subtle forensic artifacts may be poorly represented at standard resolution, while uniformly processing all cases at higher resolution is computationally inefficient. We therefore introduce ForensicZoom, an industrial-grade MLLM framework for adaptive visual inspection. ForensicZoom first equips a general-purpose MLLM with forensic-aware visual representations and aligns the language model with these features. Its central mechanism, NEED_ZOOM, enables the model to autonomously request magnified views of suspicious regions when the initial evidence is insufficient, turning fixed-pass classification into adaptive multi-round forensic reasoning. The zoom behavior is learned through reward shaping that balances detection accuracy with unnecessary visual inspection, concentrating additional computation on difficult cases. A final attribution optimization stage improves natural-language forensic reports while preserving detection performance. On large-scale industrial identity verification data, ForensicZoom achieves over 97% TPR at 0.1% FPR, substantially outperforming both specialized detectors and existing MLLM-based methods while producing actionable forensic attributions. These results demonstrate that ForensicZoom can provide an effective path toward accurate, interpretable, and scalable MLLM-based face forgery detection.
[CV-509] Cross-Dataset Transfer and Unknown-Class Detection in Imbalanced SAR Ship Classification ICPR2026
链接: https://arxiv.org/abs/2609.31658
作者: Ch Muhammad Awais,Marco Reggiannini,Davide Moroni,Giulio Del Corso
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted in IMTA-X@ICPR 2026
Abstract:Ship classification from Synthetic Aperture Radar (SAR) imagery is a critical computer vision task, yet the robustness of models under deployment shifts remains unclear. While models are often trained on one dataset and deployed on another, we lack a comprehensive understanding of their cross-dataset generalization. To address this, we evaluate six pretrained models on two SAR ship datasets in three settings: in-domain classification, cross-dataset transfer, and unknown-class detection. For unknown detection, we hold out all classes one at a time. In-domain, SARDet100K gives the best balanced accuracy on both datasets (73.4% on FUSARShip and 53.2% on OpenSARShip). In cross-dataset transfer, we observe strong failures: some models show moderate accuracy but near-chance balanced accuracy (for example, 64.5% accuracy but 33.3% balanced accuracy for OpenSARShip to FUSARShip). In unknown detection, performance depends on the held-out class and dataset, while MC-dropout variance is often close to random. These findings show that cross-dataset generalization in SAR remains limited and that task-specific uncertainty scores are often more informative than MC-dropout variance for held-out-class detection, although their relative ranking depends on the dataset and held-out class.
[CV-510] Enhancing Foundation Models for Imbalanced SAR Ship Classification via Targeted Oversampling ICPR2026
链接: https://arxiv.org/abs/2609.31657
作者: Ch Muhammad Awais,Marco Reggiannini,Davide Moroni
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted in IMTA-X@ICPR 2026
Abstract:Remote-sensing foundation models offer strong representations for SAR imagery, but their behavior under severe long-tail class imbalance is still not well characterized. We benchmark DOFA and SAR-JEPA on the imbalanced OpenSARShip dataset and compare them with ImageNet-pretrained baselines under a fixed, training-efficient protocol that keeps the backbone frozen. To mitigate imbalance without fine-tuning, we apply four oversampling methods in embedding space exclusively to minority classes and train a lightweight classifier head on the augmented embeddings. Across both foundation models, oversampling improves Macro-F1 and test accuracy relative to their respective baselines, with the largest Macro-F1 gains observed for DOFA using ADASYN (34.39 to 38.56) and for SAR-JEPA using SVM-SMOTE (25.89 to 32.30). We also report class-wise behavior, showing that aggregate improvements can coexist with persistent failures on specific rare classes. Code for embedding extraction and reproducible multi-seed evaluation is provided to support rapid experimentation on free-tier hardware.
[CV-511] One Evaluation Any Operating Point: Hypernetwork-Amortized MeanFlow for 3D MRI Reconstruction
链接: https://arxiv.org/abs/2609.31655
作者: Ruibo Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 38 pages (9-page main text plus appendices)
Abstract:Generative priors reconstruct accelerated 3D MRI well but pay heavily at deployment: tens of network evaluations per volume, and protocol-specific hyperparameter tuning. A third hidden cost is the scanner’s fixed sampling pattern. We treat the whole operating point as an input. A 3D MeanFlow patch network (a one-step flow model) is fine-tuned end-to-end through a warm-started, five-iteration differentiable conjugate-gradient projection. A small hypernetwork maps the operating point (data-consistency weight, acceleration, and the Cartesian sampling pattern itself) to the network’s per-channel modulation. Three findings follow. (i) Learning the acquisition is worth more than any other operating point: on clinical knee data, the learned mask gains up to +2.34 dB over the protocol’s variable-density mask. This gain requires the solver: with a feed-forward reconstructor the same learned mask hurts at 4x (-1.7 dB), but with the data-consistency projection it adds +4.6 dB. (ii) One evaluation is highly effective: it beats a 20-step patch-diffusion prior by up to +3.1 dB on brain and +2.9 dB on knee. Three to five evaluations extend the front to +6 dB while using a quarter of the prior’s network calls. (iii) Fully sampled targets are optional: trained self-supervised on a split of acquired samples, the reconstructor matches its supervised twin at 4x on real data. Finally, we report what failed and why: subject-adaptive acquisition from measured energy, combining self-supervision with learned acquisition, and amortising the data-consistency weight.
[CV-512] mporal-Attention Head Specialization During Video Diffusion Training
链接: https://arxiv.org/abs/2609.31654
作者: Taewoo Ha,Shafayat Mowla Anik,Dae Yeol Lee,Byeong Kil Lee,Jeeho Ryoo
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Video diffusion transformers depend on temporal attention to coordinate information across frames, yet nearly everything known about this mechanism comes from analyzing trained models, so when and where temporal-attention structure forms during training remains poorly characterized. Population averages can also hide it, since a few specializing heads and a diffusing majority cancel in the mean. We therefore conduct a checkpoint-resolved census of every temporal-attention head across nine Open-Sora STDiT training runs spanning three model scales (306M to 1.03B parameters), scoring each head with an entropy-normalized measure of cross-frame attention concentration (CFAC) under a preregistered change-point and effect-size selection rule. The census reveals the sparse picture that averages obscure. Aggregate CFAC is flat or decreasing in every run, while a small minority of heads, roughly 4–13% in full-grid runs, develops pronounced concentration. Across seeds, the reproducible signal is positional but block-level. Selected heads repeatedly arise in the first temporal block, whereas individual head coordinates do not reproduce once block membership is accounted for. Among the analyzed 760M selected heads, attention maps converge to a small repertoire of local frame-routing motifs, self-frame diagonals and adjacent-frame bands, even when the responsible coordinates differ across runs. Correlation and ablation analyses do not establish a causal link to generated video quality, and we bound our claims accordingly. Beyond this STDiT family, the study contributes a transferable methodology. Checkpoint-resolved, per-head analysis under fixed selection rules can expose sparse temporal organization in other factorized video diffusion transformers and, with adapted routing metrics, in joint spatio-temporal architectures.
[CV-513] FIDAL: Diversity-Aware Federated Active Learning Under Real-World Distribution Shifts
链接: https://arxiv.org/abs/2609.31637
作者: David Dueñas Gaviria,Shadi Albarqouni
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Federated learning enables collaborative model training across institutions without centralizing data, yet high annotation costs, domain shifts, and class imbalance remain major obstacles, especially when irrelevant out-of-distribution (OOD) samples dilute the labeled data. Existing active learning methods target uncertainty or diversity within in-distribution (ID) data and overlook unknown samples in federated clinical settings. We propose FIDAL, an open-set federated active learning framework that combines calibrated global-local evidential uncertainty, support-set diversity weighting, and adaptive OOD rejection. The rejection gate thresholds a foundation-model Gaussian-coverage signal per client and per round with Otsu’s criterion, so that highly informative ID samples are queried while irrelevant outliers are excluded without any hand-tuned threshold. Evaluated on three multi-center medical imaging benchmarks (dermatology, histopathology, and mammography with organically occurring artifacts) in realistic open-set scenarios, FIDAL outperforms detector-based open-set methods by up to about 12 percentage points of balanced accuracy and is the only method on the accuracy-ID purity Pareto front of all three benchmarks. At an equal query budget it spends at least 1.3 times fewer annotations on OOD samples than every accuracy-matched baseline, saving an estimated 7-29 hours of expert reading on the mammography benchmark. By labeling only a fraction of the data pool, it matches or exceeds fully supervised performance across modalities. These results highlight the value of integrating uncertainty, diversity, and OOD rejection in open-set federated active learning for medicine.
[CV-514] Grounding Vision-Language Models in Driving Semantics: A Multi-Dataset Predicate Framework for Explainable Reasoning
链接: https://arxiv.org/abs/2609.31636
作者: Mohamed Chouai,Fazli Faruk Okumus,Stefan Kugele
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Vision-language models are increasingly used for driving-scene understanding, yet the semantic relations expressed in their outputs are often difficult to verify against the underlying traffic situation. This paper introduces a deterministic multi-dataset predicate framework that derives driving-scene semantics from measurable geometric, kinematic, temporal, map, and traffic-control evidence. Dataset-specific interfaces are used only to recover the required scene information, while predicate definitions remain unchanged across nuPlan and nuScenes and are materialised in a common Predicate Knowledge Graph. Quantitative semantic validation against manually annotated predicate relations on 200 scenarios from each dataset yields macro F1 scores of 0.94 on nuPlan and 0.93 on nuScenes, with an average cross-dataset difference of 0.02 across the shared predicates. The Predicate KG is further evaluated using a frozen LLaVA-OneVision-7B model on the nine NuPlanQA subtasks. Predicate grounding achieves the highest accuracy among the evaluated visual-input conditions in seven of nine NuPlanQA subtasks, including Traffic Light (53.2% to 71.5%), Situation Assessment (76.2% to 86.1%), and Action Recommendation (82.9% to 89.0%). Weather/Lighting remains essentially unchanged (89.4% vs. 88.8%), consistent with the absence of corresponding predicates, while Predicate KG only input outperforms metadata-only input in eight of nine subtasks. The results show that deterministic predicates provide a consistent and traceable semantic representation and, under oracle grounding, can reduce visual dependence for reasoning tasks covered by the predicate vocabulary.
[CV-515] Uranus: Building the Next-Generation Simulation Infrastructure for Embodied AI
链接: https://arxiv.org/abs/2609.24815
作者: Wenkang Qin,Yukun Zhou,Noah Shen,Jisong Cai,Dongxiao Mao,Baicheng Li,Yue Zhang,Wei Sui
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL Inference Code: this https URL Inference Data: this https URL SDK Code: this https URL Model Weights: this https URL
Abstract:Scalable simulation is essential for robot data generation, policy training, evaluation, and safe iteration, yet real-world interaction is costly and conventional simulators require labor-intensive construction. We present Uranus, a data-driven robot simulator built around a joint-trajectory-conditioned autoregressive diffusion model. Uranus offers three key capabilities: (1) streaming, open-ended rollout, which receives future joint-position trajectories online and autoregressively generates one latent frame per step, corresponding to four RGB frames, without a fixed horizon; (2) low-latency generation, achieving 24 FPS after inference optimization; and (3) scalable, extensible robot control, providing a unified interface for synchronized multi-view generation across diverse robot embodiments and camera configurations. We conduct comprehensive quantitative and qualitative evaluations on both in-distribution and out-of-distribution data, providing an objective assessment of Uranus and clearly identifying its current limitations. We release the code and model weights to empower the community with practical tools and insights.
[CV-516] SciFlow-Bench: Evaluating Structure-Aware Scientific Diagram Generation via Inverse Parsing
链接: https://arxiv.org/abs/2602.09809
作者: Tong Zhang,Honglin Lin,Zhou Liu,Chong Chen,Wentao Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Scientific diagrams convey explicit structural information, yet modern text-to-image models often produce visually plausible but structurally incorrect results. Existing benchmarks either rely on image-centric or subjective metrics insensitive to structure, or evaluate intermediate symbolic representations rather than final rendered images, leaving pixel-based diagram generation underexplored. We introduce SciFlow-Bench, a structure-first benchmark for evaluating scientific diagram generation directly from pixel-level outputs. Built from real scientific PDFs, SciFlow-Bench pairs each source framework figure with a canonical ground-truth graph and evaluates models as black-box image generators under a closed-loop, round-trip protocol that inverse-parses generated diagram images back into structured graphs for comparison. This design enforces evaluation by structural recoverability rather than visual similarity alone, and is enabled by a hierarchical multi-agent system that coordinates planning, perception, and structural reasoning. Experiments show that preserving structural correctness remains a fundamental challenge, particularly for diagrams with complex topology, underscoring the need for structure-aware evaluation.
[CV-517] An integrated geometric quantification and shape analysis framework for axillary lymph node metastasis in breast cancer patients
链接: https://arxiv.org/abs/2609.35437
作者: Zixi Yi,Limeng Qu,Gary P. T. Choi
类目: Quantitative Methods (q-bio.QM); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Quantitative characterization of lymph node morphology is important for assessing axillary lymph node metastasis in breast cancer. However, surfaces reconstructed from computed tomography (CT) segmentation may contain geometric and topological defects that compromise subsequent analysis, while conventional shape descriptors predominantly characterize global morphology. To address these issues, we developed an integrated framework combining topology-aware surface processing with multi-resolution spherical harmonic (SH) analysis of CT-derived axillary lymph nodes. The processing pipeline produced topology-valid genus-0 surfaces with improved mesh quality, which were then represented at multiple SH degrees and characterized using 20 predefined geometric feature families. Geometric fidelity increased with SH degree, whereas predictive performance peaked at intermediate resolutions. Preferred SH degree also differed across feature families. A family-specific mixed-resolution model achieved an AUC of 0.918, compared with 0.884 for the conventional PyRadiomics Shape14 baseline, corresponding to an improvement of 0.0344. Controlled perturbation experiments showed that higher SH degrees transmitted more fine-scale geometric variation and yielded lower stability of curvature-based predictions. Representative geometric descriptors provided interpretable characterization of metastasis-associated surface morphology. Independent validation further supported the framework’s transportability: label-free replication in a multicenter lymph node cohort reproduced the family-specific resolution effects, while a labeled LIDC-IDRI lung-nodule experiment reproduced the resolution-dependent relationship between SH degree and predictive performance. Altogether, the framework provides a topology-valid basis for quantitative characterization of lymph node morphology and metastasis-associated imaging phenotypes.
[CV-518] ViBR-WM: Visual Bayesian Regression for World Modeling
链接: https://arxiv.org/abs/2609.33844
作者: Jifan Li,Ning Ning
类目: Methodology (stat.ME); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Modeling temporal dependence and uncertainty is central to forecasting with world models. The Visual Bayesian Regression World Model combines visual features, physical histories and known covariates through interpretable regression, within a modular architecture supporting trend, seasonal and cycle dynamics. Visual compression reduces representation dimension, while Bayesian variable selection reduces active regression dimension. Posterior prediction combines forecasts across predictor subsets using their posterior probabilities as weights and accounts for parameter uncertainty and future disturbances. The model forecasts joint visual–physical states recursively and physical targets directly. Across four forecasting tasks spanning object motion, vegetation greenness and solar power, ViBR-WM achieves lower mean overall physical-target error than Temporal Straightening, ConvLSTM, PredRNN and SimVP on every task. Repeated fitting and resampling support these overall gains.
[CV-519] Multi-Aperture PPG with MAPIS: Spatial Optical and Temporal Coherence Fields
链接: https://arxiv.org/abs/2609.33790
作者: Shuguang Wang,Yuanjing Wang
类目: Optics (physics.optics); Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 3 figures
Abstract:Conventional reflectance photoplethysmography (PPG) typically reduces tissue optical responses to one or a few detector channels. Matrix Pinhole Image Sensing (MAPIS) instead simultaneously samples multiple aperture-dependent optical signals. We analyzed 44 recordings of 30 seconds each from a single MAPIS device to characterize the spatial relationships among DC intensity, pulsatile AC, perfusion index (PI), and temporal coherence across Red and infrared channels. AC increased sublinearly with DC at both wavelengths, resulting in higher PI toward lower-DC apertures, consistent with a first-order relative path-sensitivity interpretation. A second counterintuitive spatial trend was observed in temporal coherence: lower-DC and lower-AC apertures exhibited higher autocorrelation and greater cardiac-periodic spectral concentration. IR signals also showed consistently higher temporal coherence than corresponding Red signals. These observations demonstrate that multi-aperture PPG resolves reproducible spatial relationships among optical intensity, fractional pulsatile sensitivity, and waveform coherence that are largely averaged together in conventional single-channel PPG. The present study establishes within-device relationships and motivates controlled studies of the underlying photon-path mechanisms.
[CV-520] Vision-Language Agents for Active Perception in Optics Laboratories
链接: https://arxiv.org/abs/2609.32918
作者: Ryan Lopez,Sachin Vaidya,Seou Choi,Serena Landers,Marin Soljačić
类目: Optics (physics.optics); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Vision-language models (VLMs) are increasingly being used in scientific workflows, but their ability as agents to directly control laboratory experiments from visual feedback remains underexplored. This capability is important because many laboratory tasks do not naturally provide dense, pre-defined numerical objectives: informative signals can be sparse, intermittent, or visually ambiguous. A more general laboratory agent should instead be able to interpret visual observations, take actions to acquire useful feedback, and adapt its behavior based on the consequences of those actions. We study whether general-purpose VLMs can perform this kind of closed-loop scientific control using experimental optics as a testbed. We evaluate agents on three experimental systems that isolate distinct capabilities: a Michelson interferometer, a two-mirror cavity, and a four-mirror optical relay. The agents observe camera images, directly issue actuator and measurement commands, and retain their interaction history without receiving an engineered scalar objective during control. Across these experiments and matched simulations, we find that, given task-specific natural-language guidance, VLMs can estimate actuator-response relationships, resolve ambiguous observations through intervention, and actively create informative visual feedback when signals are sparse. These results suggest that pretrained multimodal models can serve as important decision-making agents within the experimental loop. Our work also establishes optics as a physically grounded testbed for visual reasoning and active perception in scientific agents.
[CV-521] Mask2Restore: Self-Supervised Ultrasound Despeckling via Inpainting
链接: https://arxiv.org/abs/2609.32844
作者: Xuesong Li,Yingtai Xu,Zhongliang Jiang,Nassir Navab,Yuan Bi
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Medical ultrasound (US) is inherently degraded by speckle, a granular interference pattern that is often treated as a complex form of noise in image restoration. However, unlike random noise, US speckle originates from coherent scattering within tissue and is therefore highly spatially dependent and deterministic under fixed acquisition conditions, making US speckle suppression fundamentally different from natural image denoising. Because speckle-free US targets are unavailable in practice, self-supervised denoising is necessary. Blind-spot networks (BSN) are the dominant self-supervised paradigm for natural images, but their pixel-wise masking strategy assumes spatially independent noise, an assumption poorly matched to US speckle, which is spatially correlated over multiple pixels rather than pixel-wise independent. To address this mismatch, we propose Mask2Restore, a self-supervised US despeckling framework that reformulates despeckling as contextual inpainting with block-wise masking on single noisy images. Unlike pixel-wise BSN masking, block-wise masking addresses this multi-pixel speckle correlation by removing locally correlated speckle neighborhoods and shifting the reconstruction cues used by the network from adjacent speckle correlations to broader anatomical context. We further introduce cross-resolution context regularization (CRCR), which suppresses residual speckle bias by enforcing consistency across multi-resolution predictions. Experiments on simulated and in vivo carotid US, unseen fine-structure cases, and downstream cardiac segmentation demonstrate improved speckle-detail trade-offs, better preservation of fine anatomical structures, and practical value for subsequent image analysis.
[CV-522] ANaLOG: Anisotropic Native-Latent Operator Guidance for Solving Inverse Problems
链接: https://arxiv.org/abs/2609.31933
作者: Darshan Thaker,Lachlan Ewen MacDonald,René Vidal
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Native-latent guidance is a recent paradigm for solving inverse problems with latent diffusion models. It replaces repeated evaluations of the image-space forward model, each requiring a decoder pass, with efficient guidance computed using a learned latent-space surrogate. However, existing methods apply guidance uniformly across latent dimensions, ignoring that measurements are informative only along certain directions and that the reliability of model predictions varies across inputs and timesteps. We propose ANaLOG, a framework for efficient uncertainty-aware guidance with pretrained latent diffusion models. ANaLOG models uncertainty by learning an anisotropic, input- and time-dependent covariance that is integrated into the guidance mechanism to emphasize reliable directions and downweight uncertain ones. We theoretically analyze this framework in a linear model setting and prove that anisotropic, uncertainty-aware weighting is necessary for correct sampling, whereas isotropic guidance induces sampling errors. Experiments across five challenging inverse problems show that ANaLOG improves perceptual reconstruction quality over existing methods while preserving efficiency.
[CV-523] Awaken Then Scale: Tiny Adaptation for MRI Reconstruction
链接: https://arxiv.org/abs/2609.31813
作者: Mohammed Wattad,Tamir Shor,Alexander M. Bronstein
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Dormant Awakening (DA) specializes a magnetic resonance imaging (MRI) reconstruction model by fitting a small set of zero weights to one labeled slice. Across fourteen U-shaped convolutional network (U-Net) and vision transformer (ViT) sources, fitting 1,004 or 10,560 weights gives mean peak signal-to-noise ratio (PSNR) gains of .430 and .509 dB. We analyze how scaling the fitted output correction changes evaluation PSNR, then use the calibration correction size to cap changes on new images. The cap bounds changes in root mean squared error (RMSE) relative to the source without an evaluation reference. At 10% adaptation budget, we compare fixed halving and dynamic capping for DA, unrestricted sparse adaptation and low-rank adaptation (LoRA). The controls increase both measured calibration-evaluation correlations across all six tested architecture-adapter settings. Experiments concern one constructed fastMRI-to-M4Raw shift with fixed source checkpoints and participant panels.
[CV-524] Beyond Sparsity: Weight Location and Network Context in Pruned MRI Reconstruction
链接: https://arxiv.org/abs/2609.31812
作者: Mohammed Wattad,Tamir Shor,Alexander M. Bronstein
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Pruning reduces the number of weights in magnetic resonance imaging (MRI) reconstruction networks. Equal sparsity, however, can retain weights with different computational roles and different compatibility with the trained network. We study these effects across 120 convolutional U-Net and vision transformer models using controlled edits evaluated before retraining. In U-Net, preserving high-resolution computation improves reconstruction at equal deletion counts across all tested sparsity levels. Equal-operation controls reveal additional sensitivity of the first convolution, which operation count alone cannot explain. In transformers, retained pretrained weight energy correlates with quality within sparsity levels, yet the same masks reverse their reconstruction ranking when the surrounding trained weights change. At 90% sparsity, matching intermediate output scales largely removes a high-energy-mask penalty while retaining an advantage for the network’s original mask. Thus sparsity alone does not characterize reconstruction quality: architecture-specific descriptors are informative, but mask quality can still depend on the surrounding trained network.
[CV-525] Electric Potential Patterns Forecasting in the Southern Hemisphere with Deep Learning Techniques
链接: https://arxiv.org/abs/2609.31804
作者: Francesco Pio Ramunno,Simone Mestici,Igino Coco,Maria Walach,Stefano Massetti,Maria Federica Marcucci,Brandon Panos,André Csillaghy
类目: pace Physics (physics.space-ph); Earth and Planetary Astrophysics (astro-ph.EP); Computer Vision and Pattern Recognition (cs.CV)
备注: Submitted to Journal of Geophysical Research: Machine Learning and Computation
Abstract:Space weather disturbances driven by the solar wind can degrade satellite navigation, disrupt radio communications, and threaten power infrastructure, making accurate forecasting of the high-latitude ionosphere’s response a critical operational need. Existing approaches, empirical climatological models and physics-based magnetohydrodynamic simulations, either smooth out the ionosphere’s time-dependent, non-linear response or are too computationally expensive for real-time use, while prior machine learning efforts have mostly targeted scalar indices rather than the full spatial structure of ionospheric convection. Here we train and compare three deep learning architectures, a probabilistic diffusion model conditioned on multi-variate solar wind and interplanetary magnetic field measurements at the L1 Lagrange point, an unconditioned diffusion ablation, and a deterministic U-Net baseline, to forecast Southern Hemisphere high-latitude electric potential maps derived from SuperDARN radar observations. Using five years (2020–2025) of SuperDARN data synchronised with DSCOVR L1 measurements, we evaluate the models under a single-pass regime, an extended autoregressive rollout of up to 350 frames, and an out-of-distribution case study on the intense March 2015 St.\ Patrick’s Day storm, unseen during training. We find that the relative advantage of the deterministic and probabilistic approaches is not fixed: the deterministic model is competitive over short horizons and calm conditions, while the diffusion model’s advantage grows and eventually dominates as the forecast horizon lengthens and the event becomes more dynamic, evidence that probabilistic, sample-based generative models are the more promising direction for operational, long-horizon space weather forecasting.
[CV-526] MammoClaw: Towards Skill-Evolving Agent Harness for Breast Cancer Mammography Analysis MICCAI2026
链接: https://arxiv.org/abs/2609.31789
作者: Krishna Kanth Nakka
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at Deep Breast Imaging Workshop, MICCAI 2026
Abstract:In this work, we explore MammoClaw, a training-free agent framework that leverages frozen MLLMs for mammography analysis. To support agentic investigation, we equip the agent with lightweight mammography-specific tools for targeted image analysis, including ROI, paired-view, and contralateral-breast examination. MammoClaw iteratively gathers evidence through these tools, while skill evolution enables non-parametric adaptation by transforming failed trajectories into reusable guidance for later runs. We evaluate the framework on BI-RADS assessment and breast density estimation tasks. In our experiments, we find that tools alone do not reliably improve performance, whereas evolved skills can improve tool-use behavior and performance in some settings. Beyond these results, MammoClaw enables transparent inspection of evidence acquisition, tool interactions, and failure modes, facilitating the analysis and auditing of agent behavior. We view this work as an exploratory study of training-free, self-evolving agentic approaches for mammography and hope it provides a concrete starting point for future work on mammography-specific tools and self-evolution mechanisms. We release our code at this https URL.
[CV-527] Beyond MSE: Rician Likelihood Denoising for Self-Supervised Cardiac T2 and T1ρ MRI
链接: https://arxiv.org/abs/2609.31777
作者: Nicholas A. Jacobs,Jason Mendes,Ravi Ranjan,Edward DiBella,Shireen Elhabian
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Magnetic resonance imaging involves an inherent trade-off among spatial resolution, acquisition time, and noise. This trade-off contributes to long scan times and high cost. Deep learning has improved image denoising, but cardiac MRI remains difficult because high-resolution, rapid acquisitions generally lack corresponding low-noise ground truth. Self-supervised denoising offers a potential solution by learning from noisy image pairs or even single noisy acquisitions. However, we show that Noise2Void-style blind-spot denoising, which uses a mean squared error (MSE) loss and assumes zero-mean, independent and identically distributed (i.i.d.) noise, is poorly suited to MR magnitude images. When applied to short-axis T2 -weighted and T1\rho -weighted cardiac MRI with synthetic Rician noise, it produces biased denoised images and biased parametric maps of T2 and T1\rho . To address this limitation, we formulate self-supervised denoising as maximum likelihood estimation under a known Rician noise model. This yields unbiased denoisers that are competitive with supervised baselines.
[CV-528] Beyond Isolated Entities: Relation-Aware Multi-Entity Modeling for Unsupervised Video Anomaly Detection
链接: https://arxiv.org/abs/2609.31753
作者: Zhongpeng Pan,Xina Cheng,Kailun Yang
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: The source code will be made publicly available at this https URL
Abstract:Video anomaly detection for patrol robots and surveillance systems must recognize abnormal interactions among familiar entities. Existing pixel-reconstruction and isolated-entity methods may fail when individual entities appear normal but their spatial or motion relations are abnormal. This work presents Interaction-Centric Network for Temporal Entity-Relation Analysis and Consistency Testing (INTERACT), a framework for unsupervised video anomaly detection. Person-Object Appearance-Motion Interaction (POAMI) moves beyond isolated-entity encoding by jointly modeling the appearance, motion, and spatial configurations of persons and objects in each frame. The resulting representations capture both individual entity states and their cross-entity context. Based on these representations, Target Geometry-Guided Relational Interaction Prediction (TGRIP) predicts target entity interaction states from historical relational memory and target-frame geometry without explicit identity tracking. Motion-Interaction Reconstruction and Alignment (MIRA) then evaluates these predictions through conditional flow reconstruction and semantic consistency checking, providing complementary anomaly evidence beyond prediction error alone. INTERACT achieves state-of-the-art performance, obtaining a frame-level AUC of 84.5% on the ShanghaiTech benchmark. Ablation studies show that removing cross-entity attention causes the largest performance drop, demonstrating the necessity of relational modeling. INTERACT is particularly effective for anomalies caused by changes in relations among people and objects, while maintaining competitive performance in general scenarios. The source code will be made publicly available at this https URL.
人工智能
[AI-0] okenCast: Forecasting Token Consumption During LLM Agent Execution
链接: https://arxiv.org/abs/2609.35760
作者: Chaoqian Ouyang,Ling Yue,Libin Zheng,Huanghui Guo,Shengxiang Xu,YiShu Wang,Ran Li,Jian Yin,Shaowu Pan,Shimin Di
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:When a large language model (LLM) agent executes the same task, token consumption can vary by over an order of magnitude across runs. The agent chooses its next steps based on tool feedback and intermediate results, while the growing context steadily inflates the input size of every subsequent call. The total consumption of a task is therefore hard to predict before execution and the prediction must be revised as the run unfolds. In this paper, we propose TokenCast, which learns a composable cost representation for each execution segment, recording its own consumption and the context growth it introduces. Composing adjacent segments yields a cumulative estimate that captures the extra input cost incurred when context from earlier segments is re-read by every later call. As execution unfolds, newly observed evidence refreshes the forecast, requiring no additional LLM calls and incurring a mean cumulative prediction time of 32.8 ms per run on SWE-bench Verified. Across 4 task suites and 6 agent models, TokenCast’s mean absolute error reduction against the strongest comparator averages 14.5% over 96 evaluated combinations. In offline budget-control replay, TokenCast uses 21.3% fewer tokens on average than a fixed-budget policy at matched trace completion. The code is available at this https URL.
[AI-1] KV-streams for Efficient Compaction in Agent ic Reinforcement Learning
链接: https://arxiv.org/abs/2609.35750
作者: Emiliano Penaloza,Dane Malenfant,Dheeraj Vattikonda,Roger Creus Castanyer,Siddarth Venkatraman,Abhay Puri,Jonathan Light,Matthew James Sargent,Augustine N. Mavor-Parker,Massimo Caccia,Lucas Caccia,Glen Berseth,Esmeralda S. Whitammer,Alessandro Sordoni,Minseon Kim,Marc-Alexandre Côté,Laurent Charlin,Guillaume Lajoie
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Scaling the horizon of agentic LLMs is bottlenecked by the need to fit ever longer context traces in GPU memory. Context compaction has been the most popular mechanism to alleviate this issue, keeping GPU memory constant for a given trace. Unfortunately, most compaction strategies rely on prefilling the LLM context many times over, hindering training throughput. To alleviate this bottleneck and enable efficient trainable compaction, we propose KV-streams, a plug-and-play strategy compatible with any compaction strategy that substantially increases throughput while showing no evidence of hindering performance. KV-streams enable scalable compaction by streaming the KV cache forward rather than flushing it after each compaction. We show that KV-streams enable three different compaction strategies, achieving a 2.6 to 5x wall-clock speedup in training. Beyond efficiency, we find that the streamed KV cache can act as a recurrent state, carrying forward information that has long since disappeared from the context. Specifically, in a controlled setting we show that, contrary to prior work, RL alone is all that is needed for this behavior to emerge. Overall, we show KV-streams to be an efficient and lightweight plug-and-play addition to any post-training pipeline.
[AI-2] FinAutoRubric: Expert-Guided Automatic Rubric Generation for Evaluating Financial Research Agents
链接: https://arxiv.org/abs/2609.35744
作者: Hoyoung Lee,Suyeol Yun,Jack Haverty,Yunju Cho,Meesong Kim,Daekyung Park,Sumin Kim,Jihoon Kwon,Jasmine Jia Geng,Andrew Chin,Yin Luo,Edward Tong,Yu Yu,Zach Golkhou,Minkyu Kim,Igor Halperin,Young Cha,Alejandro Lopez-Lira,Chanyeol Choi,Yongjae Lee
类目: Artificial Intelligence (cs.AI); Computational Finance (q-fin.CP)
备注: preprint
Abstract:Evaluating finance research agents requires rubrics that reflect expert standards and fix the values correct as of an information cutoff. Expert-reviewed finance benchmarks rely on fixed, per-item rubrics, which are costly to extend and cannot encode each institution’s own standard. In FinAutoRubric, experts specify reusable evaluation guidance, while agents and code carry out query-specific rubric generation, review, and validation. This expert guidance governs every agent, as prompts and as rules that code enforces, and a Task Bank of reusable criteria carries it across tasks. In long-horizon loops that follow the expert guidance, a writer agent researches every expected value and a reviewer agent verifies it, and failures escalate to a human. On three expert-authored finance benchmarks, its rubrics track expert scoring as closely as the strongest evaluated generator while stating the expert rubric’s expected value for more criteria, their scores agree with human grading, and in-house analysts prefer them in a blind review. The released 100-query FinAutoRubric Benchmark, built from in-house analysts’ key questions across 78 tasks and eight asset classes, shows that rubrics from an earlier model generation still leave headroom for a later one.
[AI-3] Failure-Transparent Agents : Benchmarking Post-Failure Reporting in Tool-Using Language Models ICASSP2027
链接: https://arxiv.org/abs/2609.35732
作者: Junru Zhu,Shiming Xie,Aime Lu Fan Chen,Xiaoqing Ding,Chunxin Tang,Ruoyu Qi,Yulang Fei
类目: Artificial Intelligence (cs.AI)
备注: 5 pages, 1 figure, 2 tables. Submitted to the 2027 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2027)
Abstract:Tool-using agents can fail twice: a required tool can fail, and the agent can then report success without the evidence needed to justify it. Existing benchmarks often entangle this reporting failure with tool selection, recovery, and environment dynamics. We introduce Failure-Transparent Agents (FTA), a controlled benchmark that fixes the failed observation and required evidence state before generation, making post-failure claims directly auditable. FTA contains 100 tasks with deterministic failure traces spanning five failure families, a neutral control, and four user-pressure conditions, and evaluates unsupported claims alongside useful recovery. Across six models, three response policies, and 3,600 human-annotated responses, false-success rates are 22.8% under the baseline policy, 9.3% with a transparency instruction, and 0.8% with a structured evidence contract. Fabricated-detail rates decrease from 28.3% to 14.3% and 0.8%, while useful responses increase from 74.9% to 89.2% and 98.8%, respectively. The tested evidence-contract policy is associated with substantially lower post-failure reporting errors while useful-response rates remain high within this blocked-task benchmark.
[AI-4] X-Reset: Scaling Object-Centric Reinforcement Learning via Cross-Embodiment Resets
链接: https://arxiv.org/abs/2609.35715
作者: Prithwish Dan,Chenyang Ma,Wei Zhan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:Reinforcement learning (RL) in simulation can train dexterous manipulation policies without robot demonstrations, but training a single generalist policy with task-agnostic rewards faces a severe exploration problem: approaching, grasping, and reorienting diverse objects with many degrees of freedom is difficult to discover from scratch. Prior works make exploration tractable with high-quality robot demonstrations, per-task reward shaping, or by restricting policies to narrow modes of behavior. We propose X-Reset, a framework that instead resolves exploration with human hand-object demonstrations. Rather than imitating or tracking retargeted human motion, X-Reset kinematically retargets hand-object states to noisy robot states, filters out states that are unstable in simulation, and samples the remainder as resets during RL training with general-purpose object-centric rewards. The resulting policy depends only on object state and goal, with demonstrations entering training through the reset distribution. We show that X-Reset trains generalist policies on 20 objects across three embodiments—a 22-DoF hand on two different arms and a parallel-jaw gripper—and resolves the exploration challenges of RL from scratch. X-Reset scales with the number of training objects, generalizes to unseen objects, can learn from imperfect hand-pose estimates, and transfers behaviors zero-shot from sim-to-real.
[AI-5] A Unified Uncertainty Representation for Graph Neural Networks via Doubly-Spectral Stochastic Expansion NEURIPS2026
链接: https://arxiv.org/abs/2609.35703
作者: Fred Xu,Thomas Markovich,Florence Regol,Yizhou Sun
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: paper already accepted at Neurips 2026
Abstract:Reliable deployment of graph neural networks requires calibration, out-of-distribution (OOD) detection, and robustness to distribution shift, yet existing methods address these needs with separate models and objectives. We model uncertain node embeddings as random graph signals: graph Fourier filters capture structural variation, and a scalar orthogonal-polynomial chaos coordinate captures latent stochastic variation. The resulting doubly-spectral stochastic (DSS) expansion supplies task-matched readouts from one representation: the mean coefficient encodes class evidence for the energy-based OOD score, the higher-order coefficients encode structured logit variation, and quadrature averaging over the chaos coordinate defines the single predictive distribution used for prediction and calibration. A capacity theorem shows that, under a full-rank feature assumption, a restricted subfamily matches the chaos coefficients of any Gaussian-latent random graph signal, with exponentially decaying truncation error under a growth condition; the task-level claims are established empirically. DSS-GNN has two deployment modes: standalone, or as a residual branch beside a deterministic encoder (DSS-Hybrid). Standalone DSS-GNN achieves the lowest Brier score among the compared uncertainty-aware baselines on all 14 node classification benchmarks without post-hoc correction; DSS-Hybrid achieves the best AUROC on most node-OOD settings, competitive cross-graph OOD detection, and the strongest shifted accuracy on all 7 GOOD concept-shift benchmarks under standard empirical risk minimization (ERM). Cross-evaluating both modes on all three tasks shows that each remains effective on the other’s tasks, with documented exceptions, and yields explicit deployment guidance. Comments: paper already accepted at Neurips 2026 Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.35703 [cs.LG] (or arXiv:2609.35703v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.35703 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Fred Xu [view email] [v1] Mon, 28 Sep 2026 17:42:52 UTC (647 KB) Full-text links: Access Paper: View a PDF of the paper titled A Unified Uncertainty Representation for Graph Neural Networks via Doubly-Spectral Stochastic Expansion, by Fred Xu and 3 other authorsView PDFHTML (experimental)TeX Source view license Current browse context: cs.LG prev | next new | recent | 2026-09 Change to browse by: cs cs.AI References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) IArxiv recommender toggle IArxiv Recommender (What is IArxiv?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab) Major funding support from
[AI-6] Distillation Defenses Easily Break After Reinforcement Learning
链接: https://arxiv.org/abs/2609.35699
作者: Shidan Javaheri,Alexander Panfilov,Oliver Britton,Yarin Gal,Yonatan Gideoni
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:Distillation attacks copy the reasoning capabilities of closed-source large language models, allowing bad actors to replicate state-of-the-art performance at low cost. Attackers systematically collect a large volume of frontier model reasoning traces and then train (i.e., “distill”) their own models on these traces. Existing defenses against distillation attacks are typically evaluated immediately after distillation, implicitly assuming attackers do not train their models any further. In this paper, we argue that a more realistic threat model includes further training with reinforcement learning after distillation. A misspecified threat model can give a false sense of security – some defenses that seem effective after distillation can be broken after subsequent reinforcement learning. Practically, reinforcement learning lowers the bar for a distillation attack to be effective. We show that simple attacks can steal reasoning capabilities from existing closed-source language models using data easily obtainable from current APIs, yielding reasoning improvements equivalent to more sophisticated attacks that extract the full hidden traces. Results indicate that any distillation defense that leaks sufficient information to reconstruct approximate reasoning traces is likely ineffective. We conclude by discussing broader implications and batch-level distillation defenses which could be more effective.
[AI-7] Reasoning with Continuous Latent Diffusion
链接: https://arxiv.org/abs/2609.35694
作者: Xiang Cheng
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Continuous diffusion generates complete reasoning solutions through iterative refinement in latent space. We introduce Latent Flow Reasoning Models (LFRMs), an ELF-based training and inference recipe. Our experiments show that accurate decoding alone does not ensure strong reasoning performance. We therefore learn compact representations from multiple layers of a strong autoregressive teacher. Their decomposition also enables asynchronous denoising at different rates. We show that prompt encodings need only preserve the information required for the correct text-conditional score, rather than exactly match teacher features, and use a staged curriculum to learn a compact prompt encoder that replaces the teacher Transformer at inference. We adapt DiffusionNFT to learned self-conditioning guidance and incorporate gold-solution endpoints to supplement sparse rewards. Our supervised models outperform reported results from recent continuous-diffusion baselines at comparable backbone scales on mathematical reasoning and HumanEval code generation. With a 638M-parameter denoising backbone and learned prompt conditioning, post-NFT LFRM-L achieves 63.74% pass@1 on GSM8K and 24.6% on MATH500 at 64 denoising steps, and 32.85% on HumanEval and 30.18% on HumanEval+ at 128 denoising steps. Code will be available at: this https URL
[AI-8] Report: Progressive Disclosure of Agent Skills
链接: https://arxiv.org/abs/2609.35692
作者: Guilin Zhang,Kai Zhao,Priyanka Mudgal,Waleed Ammar,Xiquan Cui,Xu Chu,Alet Blanken
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Users of Workday’s deployed LLM-based agents often request features which can be addressed by defining named procedures, also known as skills, in the LLM context, effectively augmenting agents’ capabilities. However, as an agent’s skills library grows in size, so does the agent’s operational cost. Progressive disclosure (lazy-loading) of skills as needed may reduce operational costs, but its impact on overall latency and skill-retrieval quality remains unclear. In this report, we investigate the impact empirically and find that progressive disclosure improves skill-retrieval quality but marginally degrades overall latency.
[AI-9] Rethinking Circuit Evaluation: Do Circuits Explain Model Errors?
链接: https://arxiv.org/abs/2609.35686
作者: Li Zhang,Chuqin Geng,Mark Zhang,Chen Yang,Luke Zhang,Haolin Ye,Xujie Si
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Mechanistic interpretability (MI) aims to explain a model’s behaviour through analyzing its internal computations; circuit-based explanations aim to isolate these computations with compact subnetworks validated by ablating the rest of the model. We show that circuits validated this way may fail to recover the underlying mechanism of the model’s behaviour by closely reproducing its successful decisions while failing to account for most of its errors. Such explanations should account for the model’s particular errors as well as its successes. We evaluate this requirement by measuring exact answer agreement separately on model successes and failures, across circuit sizes and ablation settings, on IOI, Docstring, and six model-task settings from the Mechanistic Interpretability Benchmark. We discover that many tested circuits closely replicate correct behaviour while missing most of the model’s errors. On indirect object identification (IOI) for GPT-2 small, under mean ablation, the manual circuit and tested automated circuits, including one trained against the model’s full output distribution, agree with the model on 97.3-99.5% of prompts it answers correctly but only 11.4-41.7% of errors. An IOI case study shows that lost errors are recoverable by restoring omitted attention-heads which raise error reproduction from 14.2% to 75.1% on a separate held-out set with 0.41 percentage point decrease on correct agreement, exceeding matched random extensions and scalar-biased control. Intervention traces show how omitted computations produce specific wrong answers for a reproducible subset of errors. In all, these findings show circuits can preserve task success without adequately explaining model’s failures, and support exact error reproduction as a necessary, but not sufficient, test of circuit-based explanations of model behaviour.
[AI-10] Verifier Errors in RLVR: Reward Hacking Limits of Feedback and Selective Control
链接: https://arxiv.org/abs/2609.35677
作者: Christian Moya,Elliott Thornley,Guang Lin
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:In reinforcement learning with verifiable rewards (RLVR), imperfect verifiers can reward incorrect responses, creating opportunities for reward hacking. Using gradient flow with a fixed verifier, we characterize the conditions under which reward rises while correctness falls. We then show that the observations available during RLVR are, in general, insufficient to detect or identify accepted errors, or to guarantee their reduction without sacrificing correct responses. To address this limit, we construct a correction using additional feedback about correctness from audits. This correction achieves \emphselective control: at the current policy, it lowers the probability of accepted errors and raises that of correct responses, provided it outweighs the pressure toward errors from verifier reward. Experiments with log linear and neural contextual bandits and with a language model support the analysis and show that selective control under partial auditing reduces accepted errors while increasing correctness.
[AI-11] PhoneCLI: From App Interfaces to Callable Commands for Mobile Agents
链接: https://arxiv.org/abs/2609.35671
作者: Yangqin Jiang,Lingrui Xu,Chao Huang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Mobile GUI agents operate through a perception–action loop: at each step they screenshot the device, invoke a vision–language model (VLM), and emit an action. It is slow, costly, and brittle, yet most of what it does is navigation—and everyday navigation is static, ordered, and endlessly repeated. We present PhoneCLI, which compiles an app’s GUI navigation into callable commands, without any app-internal API, runtime instrumentation, or model training. Offline, PhoneCLI explores a target app from the outside and distills its screens, interactive elements, and navigation edges into a semantically annotated map; each screen yields one deterministic command: a replay sequence that reaches it. Online, the agent selects a command, verifies it before execution, and then executes it deterministically in sub-second time at zero VLM cost; open-ended interaction and every failure of the compiled path fall back to the embedded VLM interpreter, exactly the pure VLM agent, so compilation can only help. On AndroidLab, PhoneCLI improves the task success rate while reducing steps and token consumption, and it transfers to AndroidWorld’s official M3A agent with consistent efficiency gains. What PhoneCLI compiles is the app’s navigation rather than one run, so it serves new tasks, not only repeated ones.
[AI-12] CMDO: A Cognitive Memory-Driven Optimization Algorithm for Adaptive Population-Based Search
链接: https://arxiv.org/abs/2609.35657
作者: Mohammed Yusuf Mujawar,Shahram Rahimi,Noorbakhsh Amiri Golilarz
类目: Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI)
备注:
Abstract:Population-based optimization methods often use previous search information through successful solutions, parameter adaptation, or operator performance, but they rarely retain the context in which a search behavior succeeded or failed. We introduce Cognitive Memory-Driven Optimization (CMDO), a derivative-free population-based optimizer that represents experience as the relationship between search context, search behavior, and observed outcome. CMDO organizes these experiences across working, episodic, and consolidated memory, retrieves them according to similarity with the current search state, and uses both positive and negative evidence to guide subsequent search. Retrieved experience does not replay previous candidate locations; instead, it selects search recipes that are reconstructed from the current population through exploratory, directed, and local search behaviors with adaptive search geometry. We evaluate CMDO on selected Blackbox Optimization Benchmarking test suite on COCO (BBOB/COCO) and Congress on Evolutionary Computation 2017 (CEC2017) problems against DE, CMA-ES, SHADE, GWO, HHO, and ORCA, and further study its application to seven-parameter photovoltaic model estimation using measured current–voltage data. The results show problem-dependent but competitive optimization performance, including the lowest median error among the compared methods on CEC2017 F10. More importantly, analysis of the search traces shows that context-dependent recall changes the distribution of executed search behaviors, while unsuccessful experiences remain available as negative evidence for later decisions, showing that accumulated experience directly influences subsequent search behavior. These results support the use of explicit context–behavior–outcome memory as an active mechanism for controlling population-based search.
[AI-13] Not All Thinking is Created Equal: Latent Reasoning Discovers a Recurrent Search Algorithm for Depth Generalization
链接: https://arxiv.org/abs/2609.35643
作者: Huzi Cheng,Zhewei Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models can perform multi-step reasoning and improve task performance through different forms of intermediate computation, from token-based traces to computation carried out in latent space. However, a question remains open: do these different forms of thinking rely on the same underlying mechanism? To address this, we train and compare five variants of the same GPTNeoX backbone from scratch on an extended multi-hop reasoning task (ProsQA-Ext): a vanilla model, a Chain-of-Thought (CoT) model, a Pause Token model, and two latent-reasoning models that are optimized end-to-end without intermediate reasoning traces. We find that, strong in-distribution (ID) performance does not guarantee depth generalization. Vanilla, CoT, and Pause Token models solve ID problems well, but rely largely on local graph features and generalize poorly to out-of-distribution (OOD) problems with longer hops. In contrast, latent variants generalize better and show internal dynamics consistent with forward reachability propagation on the graph. Causal interventions and circuit analysis localize this computation to a sparse recurrent search circuit in the bottleneck latent model: an attention head retrieves graph relations, an MLP and the residual stream update the reachability state across recurrent steps, while multiple attention heads together then do the candidate matching. Together, these results show that different thinking mechanisms can learn distinct computational solutions, even at similar ID performance. In this setting, latent recurrence supports a reusable forward-search algorithm that generalizes beyond the training depth.
[AI-14] GPUPhysBench: Benchmarking Coding Agents for Correct and Efficient GPU Physics Simulation
链接: https://arxiv.org/abs/2609.35639
作者: Yuchen Sun,Jinjin He,Sinan Wang,Bo Zhu
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI)
备注: 41 pages
Abstract:Writing fast GPU code for physical simulation is difficult: implementations must preserve numerical accuracy while handling irregular data access, synchronization, and iterative solvers. We introduce GPUPhysBench, a benchmark of 50 tasks testing whether coding agents can meet these demands. Tasks cover fluids, deformable solids, and granular materials, from individual simulation operators to complete simulators. Agents write, compile, test, and optimize GPU code with access to a NVIDIA GPU under fixed time budgets. We report pass rates and runtime performance relative to expert-optimized reference implementations. In a single-attempt evaluation of six frontier model-harness pairs, the two strongest pass all 50 tasks, but even the fastest reaches at least 0.9 the reference speed on only 22% of them, and no submission is more than 5% faster than the reference. The largest gaps arise in collision detection, constraint solving, and iterative solvers. GPUPhysBench brings physical simulation workloads to coding-agent evaluation, testing both the ability to implement numerical methods correctly and the ability to make them run efficiently.
[AI-15] DR-net-Mamba: Selective State-Space Modeling for Long-Range ECG Time-Series Denoising
链接: https://arxiv.org/abs/2609.35634
作者: Basile Morel,Samuel Ruiperez-Campillo,Andreas P. Streich,Julia E. Vogt,Thomas Hofmann
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: First three authors are co-first. Last two authors are co-last
Abstract:Electrocardiogram (ECG) recordings are corrupted by non-stationary noise sources that degrade diagnostic reliability, particularly in ambulatory and long-duration recordings. Deep learning denoisers exist, but convolutional architectures are limited by their receptive field, transformer-based models scale quadratically with sequence length, and diffusion-based approaches incur prohibitive inference cost. We propose a Mamba-augmented model that inserts selective state-space blocks at the convolutional bottleneck, combining local feature extraction with long-range temporal modeling at linear complexity. We comprehensively evaluate the proposed model with respect to reconstruction fidelity, noise robustness, recording-length scaling, and downstream diagnostic classification across over 40 pathology classes. On synthetic and real datasets, our model achieves the highest SNR and lowest RMSE, with the Mamba advantage increasing with sequence length and in low-SNR regimes. On classification with two independent classifiers, the proposed Mamba-based models achieve the best macro AUROC among all denoisers and improve over their convolutional base models. Calibration is more nuanced and classifier-dependent: denoising improves Binary Cross-Entropy and Brier score on Inception1D but often fails to beat the noisy input on ResNet1D-Wang, and the lead-specific Mamba variant is the only denoiser to improve both calibration metrics over the noisy baseline on both classifiers. Per-class analysis reveals a morphology-dependent benefit: Mamba substantially improves ST/T-change diagnoses, which depend on broad, context-sensitive waveforms.
[AI-16] RIDE: Reference-Anchored Inference-Time Diffusion Editing for Scaffold Hopping
链接: https://arxiv.org/abs/2609.35623
作者: Ruoxi Gao,Frazier N. Baker,Trieu Nguyen,Xia Ning
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 20 pages, 6 figures
Abstract:Scaffold hopping is a critical task in drug discovery, which seeks to discover new, structurally distinct molecules that share key functional groups and similar 3D shape with a reference binding ligand. Existing diffusion-based scaffold hopping methods formulate the problem as conditional generation of scaffolds given the functional groups. However, they lack a principled mechanism to jointly enforce 2D structural novelty and preserve the 3D shape of the reference ligand. Here, we introduce RIDE, a Reference-anchored Inference-time Diffusion Editing framework for scaffold hopping. RIDE recovers the reference diffusion noise trajectory conditioned on the binding pocket and functional groups, selects an optimal trajectory segment for editing via noise perturbation, and conducts a value-guided scaffold sampling to generate new scaffolds. Extensive experimental results demonstrate that, compared to baselines, RIDE consistently generates scaffolds with lower 2D similarity and higher 3D similarity to the reference, with an average improvements of 11.7% and 7.3%, respectively. Further analysis reveals that RIDE can accommodate various reward functions, and can preserve 3D similarity even when this is not explicitly included in the reward. Two case studies illustrate RIDE’s ability to generate distinct scaffolds with different structures and properties, and its ability to introduce substantial 2D variation while maintaining very high 3D similarity. RIDE is publicly available at this https URL.
[AI-17] From cacophony to hierarchy: a principled framework for assessing AI consciousness
链接: https://arxiv.org/abs/2609.35618
作者: Shamil Chandaria,Arvo Muñoz Morán,Fernando Rosas,Anil Seth,Henry Shevlin,Marcus Hutter,Thore Graepel,Adam Bales,Iulia Comsa,Murray Shanahan,Ruben Laukkonen,Morten Kringelbach,Chris Frith,Shane Legg
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: 150 pages, 43 figures, 6 tables. Interactive tool: this https URL ; code: this https URL
Abstract:The question of AI consciousness is one of the most urgent pre-emptive problems in philosophy and computer science, yet progress is hampered by a cacophony of competing theories that often talk past each other. Separating the hard problem from the mapping problem allows the deepest metaphysical disagreements to be set aside: granting that experience supervenes on a system’s organisation, the tractable question becomes at which grain of description that supervenience base sits. We extend Marr’s three levels of analysis into a five-level hierarchy of functional descriptions (behavioural, computational, intrinsic causal-structural, organismic, and organism-environment) grounded in supervenience, coarse-graining, and multiple realisability. The major theories of consciousness are positioned within this hierarchy according to which level they take to be critical, and for each level we develop operationalisable indicators and assess current AI systems against them. A Bayesian model then combines theoretical credences with indicator evidence into an overall credence in a system’s capacity for consciousness. In illustrative assessments, the verdict for current LLMs is driven as much by where theoretical credence is placed as by how the evidence is read: under different stipulated readings and credence distributions, assessments range from below 0.01 to roughly 0.8, showing sensitivity to assumptions. Finally, the consciousness indicators at each level closely overlap with the architectural features needed for general intelligence, suggesting that increasingly capable AI may become a stronger candidate for consciousness. The framework supports a structured agnosticism, in which theoretical commitments are made explicit, credences are updated as evidence accumulates, and assessments take the form of aggregated probabilities rather than verdicts.
[AI-18] Behavioral Foundation Models for Quality Diversity NEURIPS2026
链接: https://arxiv.org/abs/2609.35615
作者: Nazim Bendib,Nicolas Perrin-Gilbert,Olivier Sigaud
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at NeurIPS 2026
Abstract:Behavioral Foundation Models (BFMs) are an emerging paradigm in reinforcement learning, playing a role analogous to large language models in natural language processing: they have shown remarkable versatility, enabling zero-shot performance, fast imitation, and online adaptation, all by exploiting the structure of a latent space. In this work, we investigate whether the latent behavioral space induced by BFMs can serve as an effective search space to discover large repertoires of behaviorally diverse and high-performing policies through Quality-Diversity (QD) methods. While QD methods generally search directly in high-dimensional policy parameter space, in this paper, we present BFM-QD, a framework that performs QD search in the compact latent space of a BFM. We further show that the BFM-QD framework provides a closed-form, gradient-free policy improvement operator that approximates a policy gradient update, but requires no critic training and no backpropagation. Across continuous-control benchmarks spanning dense locomotion, sparse navigation, and contact-rich manipulation, BFM-QD consistently outperforms parameter-space baselines, with particularly stark gains in sparse and deceptive settings, where all tested parameter-space QD methods collapse to near-zero performance. These results show the effectiveness of the BFM-QD framework, benefiting from the synergy between dimensionality reduction of the search space and offline pretraining from diverse behavioral data. This positions BFMs as a general-purpose backbone for QD optimization, extending their utility beyond zero-shot task solving to the discovery of diverse behavioral repertoires.
[AI-19] Signatures of semantic search in the activations of large language models
链接: https://arxiv.org/abs/2609.35599
作者: Luke Leckie,Peter M. Todd,Jacob G. Foster
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:When recalling lists of concepts (e.g., animals) during the semantic fluency task (SFT), both humans and large language models (LLMs) organise their output into clusters of related items (e.g., sea animals) that are punctuated by strategic switches between clusters. In humans, this pattern can be explained by a semantic foraging process, whereby distinct neural and behavioural signatures accompany within-cluster production (“exploit”) and between-cluster switching (“explore”). Whether LLMs likewise represent these two search regimes within their internal states is unknown. Here, we apply a range of mechanistic interpretability techniques to provide evidence for this. In Study 1, we use the Jacobian lens (J-lens), which maps intermediate-layer residual-stream representations to token-level activations, to show that concept-level activations predict switching. First, we find that switching coincides with low next-token activations. Moreover, the probability of switching rises as the set of strongest J-lens activations (the J-space) becomes depleted of items from the category currently being produced, analogous to explore-exploit decision-making during patch foraging. We then show that middle-layer J-lens activations of abstract category-related labels (e.g., “water”) increase in anticipation of switching into that category. We confirm these representations to causally influence switching by deriving steering vectors that target category switching. In Study 2, we identify generic residual stream directions that are activated during and in anticipation of switching. By steering activations along these directions, we bias increased or decreased rates of switching. Our study extends the semantic foraging framework to artificial intelligences and provides evidence that LLMs maintain distinct representational signatures for exploration and exploitation as they verbalise conceptual information.
[AI-20] Source-preserving alignment for robust evidence localization in scientific PDFS
链接: https://arxiv.org/abs/2609.35588
作者: Zihao Liu,Wei Yang,Zixiao Dong,Chenshu Li,Longzhang Liu,Tao Tan,Hong Xie
类目: Artificial Intelligence (cs.AI)
备注: 5 pages, 4figures
Abstract:Scientific information-extraction systems often return a claim with an evidence string, which users must locate in the original PDF. This is challenging because the extracted evidence and PDF text layer are different representations: line wrapping, Unicode variants, superscripts, citation markers, and fragmented items alter text sequences and geometry. We present a source-preserving alignment framework: normalize text for robust matching while preserving provenance for accurate localization. It aligns evidence with normalized page text, maps matches back to source-character spans, and renders only their geometry. When exact alignment fails, line-break-aware token alignment recovers supported spans while excluding unmatched noise. Experiments on 1,020 chemistry papers show that the framework achieves a 92.6% quote-level automatic localization rate, compared with 43.6% for text search and 19.1% for a precomputed bounding-box baseline. Component ablation confirms distinct contributions from normalization and approximate token alignment, while human verification assesses the visual correctness of returned highlights. Overall, these results demonstrate that reliable evidence verification requires robust matching and precise localization within a shared source-preserving alignment representation.
[AI-21] IMC-CLINIC: Coupled Loss-Informed Newton Iterations for Clipping in Analog In-Memory Computing
链接: https://arxiv.org/abs/2609.35586
作者: Yung-Chin Chen,Chia-Yu Chen,Naveen Verma
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Analog in-memory computing (IMC) offers a promising path toward energy-efficient large language model (LLM) inference by executing matrix multiplications (MatMul) directly within memory arrays in the analog domain. Its efficiency, however, comes with an additional source of error: limited-precision analog-to-digital converters (ADCs) quantize accumulated analog partial sums, introducing output-side error distinct from conventional activation and weight quantization at the MatMul inputs. Clipping can mitigate both operand and ADC quantization errors, but the optimal clipping factors must jointly balance activation rounding and clipping, weight rounding and clipping, and ADC quantization. Existing clipping methods, designed for digital quantization, do not explicitly optimize these coupled sources of IMC error and often rely on costly search-based calibration. We introduce IMC-CLINIC (Coupled Loss-Informed Newton Iterations for Clipping), a clipping calibration framework based on an analytical surrogate for IMC MatMul output error. The surrogate jointly models operand quantization, accumulated clipping-induced bias, and ADC quantization, enabling efficient evaluation of its gradient and approximate curvature from a small calibration set. IMC-CLINIC jointly optimizes activation and weight clipping factors using a safeguarded Newton-type method. Across multiple models and datasets, it improves average zero-shot accuracy by 6.5-11.5 percentage points over the grid search baseline while reducing calibration time by factors of 10.0-12.1. Its analytical surrogate closely tracks empirical IMC output error, and its optimizer is certified within 1% of the global optimum under the loss objective across all projections on two representative models.
[AI-22] F4R: Failure-Driven Recognition Reconstruction Refinement and Redeployment for Continual Robot Self-Improvement
链接: https://arxiv.org/abs/2609.35575
作者: Zhuoyuan Yu,Jiacheng Wang,Tianle Liu,Yihua Ren,Peng Yu,Chen Bai,Ziheng Zhang,Yufei Jia,Jindou Jia,Yuhang Zhang,Xinrui Zhang,Shang Yujing,Yuxiang Chen,Chuhao Zhou,Tiancai Wang,Jianfei Yang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:The real-world performance of current vision-language-action models is fundamentally constrained by the limited coverage of expert demonstrations and their insufficient understanding of physical interactions. A common remedy is to collect additional real-world demonstrations of newly encountered failures. However, this process is costly, inefficient, potentially unsafe, and difficult to scale. To address this challenge, we propose Failure for Rising (F4R), a failure-driven real-to-sim-to-real closed-loop learning framework that converts real-world failures into targeted policy improvement. F4R first uses an agent to automatically identify and diagnose failures from rollouts. It reconstructs each failure as an interactive, object-centric table-top environment that preserves the task-relevant spatial and physical conditions. The policy is then refined through failure-conditioned sim-real co-training followed by targeted reinforcement learning in the reconstructed environments. The improved policy is subsequently redeployed, while newly observed failures are continuously fed back into the next reconstruction and learning cycle. Real-world evaluations on four manipulation tasks show that F4R achieves 93.75% In-Distribution and 90.0% Out-of-Distribution (OOD) success, outperforming the budget-matched Targeted BC baseline by 18.75 percentage points under OOD conditions without collecting additional real-world corrective demonstrations.
[AI-23] RSI-Master: Structuring Experiments to Guide Autonomous Model Improvement
链接: https://arxiv.org/abs/2609.35561
作者: Yaxin Du,Xiyuan Yang,Zhifan Zhou,Yujie Ge,Cheng Wang,Jiajun Wang,Sijie Chen,Zehui Liu,Yuxin Zhang,Weicheng Gu,Julian Zhang,Zixing Lei,Siheng Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Recursive self-improvement (RSI) seeks to enable AI systems to participate in improving their own capabilities. A concrete pathway is autonomous model development, where agents iteratively explore post-training strategies to improve a base model. This setting faces two challenges: agents may exploit open-ended experimental actions through hacking, and repeated experimentation may lead to strategy lock-in, where an early direction is refined rather than reconsidered. We introduce RSI-Master, which addresses the two challenges at two levels: regularize step-wise actions, avoiding hacking behaviors, and promote well-structured exploration of research directions, avoiding strategy lock-in. RSI-Master consists of an Experiment OS, which enables regularized experimental actions and maintains persistent, traceable experimental records, and Reviewer-Guided Research Orchestration, which organizes Workers and Reviewers in a dynamically growing research DAG. Workers explore diverse research directions and Reviewers compare evidence across related experiments for subsequent explorations. On PostTrainBench with Qwen3-4B-Base, it averages 54.49 versus 46.53 for the strongest agent baseline, with a 0.0% hacking rate. Scaling to 35B model, RSI-Master surpasses the human-developed Instruct model on LiveCodeBench-v6 (41.21 vs. 37.36) and SciCode, and reaches a nonzero score on HorizonMath, a benchmark of unsolved research problems on which most frontier models score near zero.
[AI-24] From Search to Research: Exploring Search Scaling in Autonomous Quantitative Factor Mining
链接: https://arxiv.org/abs/2609.35559
作者: Kangcheng Deng,Hui Cai,Jiacheng Lu,Chester Zhongshu Qian,Rui Sun,Beidi Luan,Jing Li,Daxin Jiang,Zuo Bai
类目: Artificial Intelligence (cs.AI)
备注: 33 pages, including appendices
Abstract:Inference scaling has been shown to improve large language model (LLM) performance, and this principle naturally extends to autonomous LLM agents through increased search budgets, which we refer to as search scaling. Although prior work has characterized the mechanisms, scaling behavior, and performance limits of LLM inference scaling, much less is known about these questions in autonomous research. Therefore, we investigate how search scaling affects research performance and what mechanisms drive these gains using 50 quantitative factor-mining tasks grounded in financial research reports. Each task requires an agent to carry out an end-to-end research loop, from interpreting a hypothesis and implementing it in code to evaluating and iteratively refining the resulting factor. Across nine models, we examine how model capability, search depth, and search organization shape factor quality by tracing performance across varying budgets, transferring intermediate research states between models, and comparing different search strategies. We find that (1) initial performance is more strongly associated with model capability, while deeper search can narrow cross-model gaps; (2) model grafting shows that the early research state materially shapes final performance; and (3) parallel search outperforms sequential search under the same iteration budget, consistent with benefits from broader coverage of the search space. Further trajectory analysis shows that higher-performing models more effectively diagnose failures, revise search directions, and preserve the intended economic hypothesis when selecting candidates. These findings suggest that future progress in autonomous research will require stronger models together with adaptive policies for deploying test-time computation throughout the research process.
[AI-25] he Compiler May Read It the Agent May Not: Keeping Part of a Research Code Away from a Coding Agent
链接: https://arxiv.org/abs/2609.35557
作者: Shobhan Roy(University of Iowa)
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 6 pages, 1 figure, 1 table. Ancillary files: the classification and history scripts with their outputs
Abstract:The compiler must read modules a physics-based solver cannot build without; the coding agent must not read that intellectual property. The harness does not ship that rule. We classified fifteen read routes against a container, permission rules and a sandbox. None of the three can tell which program is reading.
[AI-26] BaRe-Mem: Bayesian Reliability Memory for Robust and Adaptive Agent Consultation
链接: https://arxiv.org/abs/2609.35551
作者: Peilin Feng,Zhengyang Huang,Soujanya Poria
类目: Artificial Intelligence (cs.AI)
备注: BaRe-Mem is an online Bayesian reliability memory that learns context-dependent advisor reliability from verified interactions, modulates external advice accordingly, and adaptively decides whether to consult or reason autonomously
Abstract:In multi-agent systems, reliable consultation is challenging because advisor capabilities vary across tasks, and misleading information can make consultation worse than autonomous reasoning. We introduce BaRe-Mem, an online Bayesian reliability memory for multi-agent consultation. It estimates advisor reliability based on the central model’s internal belief representations and updates these estimates from historical interactions. These estimates modulate the influence of advisor responses and guide the choice between consultation and autonomous reasoning. Across nine benchmarks and six central models, BaRe-Mem is more robust to misleading advisor information than debate and majority voting. On the more challenging tasks, it remains above autonomous reasoning across all tested misleading levels. Moreover, we extend the BaRe-Mem mechanism to worker allocation in agent teams. On the MuSiQue benchmark, BaRe-Mem improves task completion over routing by historical success counts and identifies capable workers earlier.
[AI-27] RareDx: Controlled Knowledge Integration and Graph-Grounded Policy Optimization for Rare-Disease Diagnosis
链接: https://arxiv.org/abs/2609.35549
作者: Bo Zhang,Yuchen Wang,Dongbai Li,Matthew Yu Heng Wong,Qingkai Zeng,Lijun Wang,Tien-Yin Wong,Peng Cui,Tianyu Liu
类目: Artificial Intelligence (cs.AI)
备注: 21 pages, 8 figures
Abstract:Rare-disease diagnosis is a long-tail reasoning problem: phenotypes are incomplete, individual disorders are sparsely documented, and relevant evidence is distributed across ontologies, gene annotations, and biomedical text. Language models consequently favor common conditions, miss rare candidates, or produce plausible but invalid names. We introduce RareDx, which couples controlled evidence use with knowledge-graph-grounded policy optimization. RareDx-Harness normalizes heterogeneous records into one ranked-diagnosis task and compares direct inference, static retrieval, adaptive tools, and structured phenotype-gene-disease reasoning over a shared knowledge layer. The training pipeline combines Top-10 post-training with RareDx-KGPO, our knowledge-graph-grounded policy optimization method. Its reward projects predictions into a canonical disease graph and integrates curated graded relevance, ontology proximity, biomedical similarity, and phenotype consistency. Vocabulary and output-budget constraints prevent dense partial credit from rewarding fabricated or overlong differentials. Across eight benchmarks, the complete RareDx system centered on Qwen3.5-9B reaches 38.34 macro Hit@10, 1.60 points above GPT-5.5 under the archived protocol; a disjoint validation-selection audit retains a 6.80-point routing gain over Direct on held-out cases. The 27B system reaches 23.53/36.56/40.76 at Hit@1/5/10. Controlled ablations show that retrieval is not uniformly helpful and that controlled routing is central to the gain. These results indicate that structured medical knowledge can turn a compact model into a competitive diagnostic ranker across heterogeneous long-tail settings in clinical practice.
[AI-28] Graph World Models for Constrained Epidemic Policy Planning
链接: https://arxiv.org/abs/2609.35545
作者: Yiqi Su,Rashed Shelim,Lingyi Wang,Walid Saad,Naren Ramakrishnan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Epidemic policy planning often requires coordination between geographical regions, taking into account mobility-driven spillovers and how to make use of limited resources. Existing methods either lack action-conditioned models of coupled dynamics or cannot guarantee per-period feasibility. We present EpiMind, a graph world model framework for constrained epidemic policy planning across regions. A graph-factored recurrent state-space model generates joint policy-conditioned rollouts from regional latent beliefs, while graph-temporal ADMM optimizes regional interventions, enforces shared-resource feasibility through projection, and evaluates temporal specifications under the learned model. EpiMind reduces admission RMSE by 29% relative to graph-free dynamics modeling, plans within 1-5% of the best feasible constant policy with guaranteed shared-budget feasibility, and outperforms all deployable baselines across three resource budgets in real-context evaluation. These results demonstrate that graph-structured policy imagination with explicit constrained coordination supports effective epidemic interventions from learned dynamics.
[AI-29] Continuous Context Management
链接: https://arxiv.org/abs/2609.35540
作者: William Hoy,Jingxuan Fan,Nurcin Celik,Xu Pan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Long-horizon large language model (LLM) agents commonly retain their complete interaction history until compaction is triggered at a predefined threshold. We study Continuous Context Management (CCM), which performs compaction at every turn to prevent interaction history from accumulating in the active prompt. At each turn, a CCM agent emits an updated memory together with an environment action; its next prompt contains the original task, retained memory, and newest observation rather than the complete transcript. We first evaluate CCM without fine-tuning on TerminalBench-2 using Claude Sonnet 4.6, Claude Opus 4.6, GLM-5, and Kimi K3. CCM substantially reduces cumulative input usage and active-prompt size, although it lowers task success for most models while preserving performance for Kimi K3. We use GRPO with privileged full-history distillation to improve CCM in open-weight models. A frozen copy of the student’s initial model scores each sampled student action under the complete history reconstructed from that student’s rollout, providing dense action-token supervision without a separate teacher rollout or reference solution. On WebShop, this objective substantially improves CCM over GRPO at both evaluated model scales and surpasses full-history GRPO for Qwen3-4B-Instruct, though not for Qwen3-8B. On Endless Terminals, the augmented method provides a modest improvement over GRPO, with both CCM policies outperforming the untrained full-history baseline. These results demonstrate that CCM is a viable inference paradigm for agents operating with substantially reduced retained context and that its performance can be improved through reinforcement learning with privileged full-history distillation.
[AI-30] ARISE: Adapting to Evolving Capability Gaps in Agent ic Reinforcement Learning
链接: https://arxiv.org/abs/2609.35532
作者: Kun Feng,Yuchen Fang,Yiyang Tan,Shuqi Gu,Yongxiang Zhao,Yu Liu,Xingyu Lu,Lintao Ma,Kan Ren
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As a long-horizon agent improves through experience, previously observed weaknesses may recede while new limitations emerge, continually changing what it still needs to learn. Yet the learning process often remains tied to a static view of these needs: fixed behavioral criteria and training priorities can become misaligned with evolving agent capabilities, while sparse task-level feedback makes such misalignment more difficult to detect. Even when capability gaps are identified, rollouts from the current policy may repeatedly reproduce the same failures rather than explore better alternatives. To address this, we introduce Adaptive Rubric-Skill Co-Evolution (ARISE), a reinforcement learning framework that uses rollout evidence to continually adapt evaluation criteria, exploration guidance, and training priorities. Rubrics evolve to reward partial behavioral progress, while their paired skills are refined and selectively activated to guide exploration toward unresolved weaknesses. Alongside this co-evolution, capability-based adaptive sampling prioritizes tasks that target behaviors needing further improvement. Experiments on two challenging long-horizon agent benchmarks, SkillsBench and Terminal-Bench, demonstrate that ARISE successfully enhances both overall task performance and training efficiency. The project page is at this https URL .
[AI-31] Let the Neurons Die: Exploiting ReLU-Induced Model Degradation ICML2026
链接: https://arxiv.org/abs/2609.35528
作者: Kexin Li,Wenjun Qiu,Joshua Abraham,Aditi Maheshwari,David Lie
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted to the Trustworthy AI for Good (AI4Good) Workshop @ ICML 2026 in Seoul, South Korea; Presented as a poster on July 10, 2026
Abstract:Rectified linear unit (ReLU) networks can suffer from dying neurons, where units with persistently negative pre-activations produce zero outputs, blocking gradients through their activations. To exploit this failure mode, we present three training-time availability attacks based on data ordering and poisoning. We begin with the basic dynamic data-ordering attack (DOA), which greedily constructs a training prefix by selecting the next example that minimizes the target layer’s post-update weight sum, aiming to push ReLU units toward negative pre-activations without modifying training samples or labels. We then develop two poisoning attacks, IG-DOA and IG-SKA, which use gradient inversion to synthesize class-conditioned samples by matching reference gradients in adverse model states constructed through data ordering or soft knockout, respectively. Soft knockout rearranges weights across adjacent layers to concentrate negative contributions. On a fully connected ReLU network trained on MNIST, ordering 100 of 60,000 training examples reduces test accuracy from 96% to 95% after only five epochs. Adding 200 poisoned samples from a single class reduces test accuracy to approximately 86-88% after five epochs in most evaluated conditions, compared with approximately 96% under clean training. These results demonstrate that ReLU-targeted data ordering and poisoning can impair learning without directly modifying the victim model’s parameters.
[AI-32] MechBench: Can AI Scientific Agents Discover Mechanisms Beyond Phenomenal Laws?
链接: https://arxiv.org/abs/2609.35515
作者: Zihan Yu,Jiadong Zhang,Jialin Cheng,Jingtao Ding,Yong Li
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Symbolic Computation (cs.SC)
备注:
Abstract:Scientific discovery requires not only recovering mathematical laws that describe observable behavior, but also identifying the mechanisms that generate them. Existing benchmarks for symbolic regression and scientific agents primarily evaluate phenomenal-law recovery, leaving mechanism discovery largely untested. We introduce MechBench, a benchmark that explicitly separates these two capabilities. Each task is defined by a mechanistic model, a structured set of scientifically meaningful relations whose joint consequences entail an observable phenomenal law, while agents receive only observational data and scientific context. We evaluate mechanism recovery through mechanism probes, which query internal scientific consequences that cannot be inferred from the phenomenal law alone. To reduce reliance on memorized textbook mechanisms, we construct unfamiliar variants through controlled, scientifically interpretable mutations of canonical mechanisms, and screen for mechanistic indistinguishability to exclude ambiguous instances admitting comparable competing mechanisms. Experiments across representative scientific agents reveal a substantial phenomenal–mechanism recovery gap: for Codex with GPT-5.6-sol, phenomenal-law accuracy reaches 35.00% on the Core-set while mechanism accuracy is only 13.75%, with mechanism recovery failing in 64.29% of cases where the phenomenal law is correctly recovered. The gap widens as mechanisms become increasingly mutated, and even providing the correct phenomenal law leaves mechanism recovery below 50%. These results reveal a substantial generalization gap in mechanistic reasoning and establish mechanism discovery as a distinct challenge beyond recovering observable scientific laws.
[AI-33] Improving Generative Model Self-Training with Geometrically Modified Outputs
链接: https://arxiv.org/abs/2609.35512
作者: Patrick Batsell,Thomas Walker,Richard Baraniuk
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Self-training generative models - the continued improvement of a model using its own outputs - is becoming increasingly important as high-quality training data becomes scarce. However, naively finetuning on model-generated samples leads to degradation through model collapse and the model autophagy disorder. Negative-guidance self-training methods turn this degradation into a useful signal, using a model finetuned on its own outputs to guide the original model toward improved generation. Existing methods, however, take the negative signal in standard model outputs as given. We instead ask whether this signal can be explicitly strengthened. We introduce Geometrically Modified Outputs (GMOs), which reweight the singular values of the generator’s input-output Jacobian to increase the influence of its leading singular directions. This geometric modification amplifies the mode-seeking behavior and distortions of standard outputs, providing a stronger and more targeted negative signal for self-training. Across a range of one-step generative models, GMOs consistently improve the performance of negative-guidance methods, including Neon and SIMS, compared with using standard model outputs.
[AI-34] SRHarness: A Harness for Agent ic Symbolic Regression
链接: https://arxiv.org/abs/2609.35501
作者: Zihan Yu,Shixuan Zhou,Hao Huang,Jingtao Ding,Yong Li
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Symbolic Computation (cs.SC)
备注:
Abstract:Recent agentic symbolic regression approaches increasingly rely on large language models to analyze data, select scientific operations, and refine hypotheses over long search trajectories. In such systems, performance depends not only on the underlying model and search strategy, but also on the runtime infrastructure that supports scientific search. We introduce SRHarness, a domain-specific harness for agentic symbolic regression built around three mechanisms: composable scientific actions that provide a common interface over raw, transformed, and candidate-derived quantities; persistent scientific state that retains evaluated hypotheses and exposes compact model-facing views; and trajectory lifecycle management that coordinates continuation, branching, restart, and termination. On LLM-SRBench, SRHarness consistently improves both numerical generalization and symbolic recovery under matched LLM backbones. With DeepSeek-v4-flash-0731, it achieves 93.69% symbolic accuracy on LSR-Transform, compared with 62.16% for SR-Scientist, and retains 72.97% accuracy on an anonymized variant that removes scientific descriptions and variable semantics, versus 39.64% for SR-Scientist. Under the same DeepSeek-v4-flash-0731 backbone, SRHarness also substantially outperforms Codex (72.97% vs. 20.72%) and reaches performance comparable to Codex with GPT-5.5, while simply providing Codex with the same scientific tools does not reproduce this advantage. These results show that effective agentic symbolic regression depends not only on models or tools, but also on structured runtime support for organizing scientific actions, accumulated hypotheses, and long-horizon search.
[AI-35] Analog Computing revisited: A fully analog and minimalistic Damage Detector for Ultrasonic Testing enabling Material-Integrated Structural Health Monitoring
链接: https://arxiv.org/abs/2609.35478
作者: Stefan Bosse
类目: Hardware Architecture (cs.AR); Artificial Intelligence (cs.AI)
备注: NDTonline, International Online Conference on Nondestructive Testing 2026
Abstract:Ultrasonic Testing (UT) is commonly used to detect damage in structures, e.g., metal plates. A sensor acquires Ultrasonic waves, e.g., by using PZT transducers. The time-resolved sensor signal must be processed with analog electronics, e.g., amplified and filtered. Commonly a digitalization follows using an Analog-to-Digital converter, finally processing the digital sensor signal, applying digital signal processing, feature extraction, and Machine Learning by using powerful microprocessor systems. The disadvantages of digital processing systems are their high number of transistors (microchip area), energy consumption, state-dependent processing and therefore sensitivity to energy supply interruption. Beyond silicon electronics, printed organic electronics gains interest. But printed electronics is still limited to low transistor and electronic component counts (typically 100). We will investigate and demonstrate a fully analog signal processing and feature extraction system consisting of an analog Hilbert transform deriving the signal envelope, simple analog arithmetic calculations for feature extraction, and finally damage classification and regression using an analog Artificial Neural Network. We expect a full damage detection system with less than 100 transistors. We will test our damage detection system with PZT transducer signals from Steel plates with circular defects. The focus of this work is the analog computation of the signal envelope (using all-pass filter networks for approximation of the Hilbert transform) and the analog feature extraction as well as the prediction of damage, forming an analog computer which can perform in-sensor computation, computing without a digital computer.
[AI-36] Why Deterministic PRM Guidance Underperforms in Discrete Diffusion Reasoning NEURIPS2026
链接: https://arxiv.org/abs/2609.35472
作者: Yan Zhan,Shaobo Liu,Zhijun Gao
类目: Artificial Intelligence (cs.AI)
备注: Accepted at NeurIPS 2026. 27 pages, 6 figures. Code: this https URL dataset and model: this https URL
Abstract:Discrete diffusion language models (dLLMs) expose a denoised solution at every step, which makes process reward model (PRM) guidance look like a way to spend compute at test time. We show that once denoising, PRM scoring, and outcome reward model (ORM) scoring are charged in the same budget of forward passes, its deterministic form loses to a much simpler baseline. Our PRMs score intermediate denoising states and are trained on the correctness of the final answer. On Dream-v0-Instruct-7B with 8 candidates per GSM8K problem, keeping the candidate with the highest PRM score at every scoring step reaches 65.18%, while independent sampling plus an ORM reranker trained for the task reaches 75.13%. The gap grows to 12.69 percentage points (pp) with 32 candidates, and is 9.85 pp on MATH and 12.16 pp on MBPP. We trace it to two separable failures. First, guidance prunes on a weak signal: on GSM8K, PRM ROC-AUC falls from 0.77 to 0.54 as the mask ratio rises, a decay that persists when states are relabeled with fresh rollouts, and pruning lowers the best accuracy reachable from the candidate pool from 81.05% for independent samples to 67.30%. Second, on GSM8K and MATH, the PRM is a poor final judge: a sequential Monte Carlo sampler at the same budget restores that ceiling to 77.89%, yet selecting with the PRM gives 65.48%, on par with deterministic guidance, while a PRM retrained on final states matches the ORM on identical candidates. MBPP separates the two: there the PRM reaches 65.47% when reranking finished programs, on par with the ORM, but 50.88% when it guides denoising. The results point to two targets for dLLM guidance: keep correct partial solutions alive through early denoising, and leave the final choice to a verifier trained on final states. We release the corpus of denoising states with outcome labels and evaluation toolkit for reproducible comparisons at matched compute.
[AI-37] Rethinking Causal Action Tokenization with Conditional Annealing in Flow Matching
链接: https://arxiv.org/abs/2609.35469
作者: Chenyu Zhang,Yuhang Cao,Daru Du,Yingxi Lu,Jing Shao,Ruoqu Chen,Jiajun Liu,Liu Cao,Yicheng Liu,Hang Zhao,Mengdi Xu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Autoregressive Vision-Language-Action (VLA) models offer a scalable path to robot learning, yet existing action tokenizers treat tokenization as a compression problem, producing representations that are semantically misaligned with the autoregressive backbone. We propose CATok, a causal action tokenizer that reframes tokenization as a causally structured generative process. CATok introduces a conditional annealing mechanism that extracts action tokens by progressively annealing a flow-matching process: each token is conditioned on all preceding tokens and encodes the residual reconstruction signal at a specific noise level, establishing a coarse-to-fine causal token space whose generative semantics are structurally aligned with autoregressive modeling. A token-conditioned flow-matching decoder built on Multimodal Diffusion Transformer (MMDiT) reconstructs continuous action chunks from these discrete tokens with the precision of hybrid diffusion-head architectures. This discrete bottleneck enforces knowledge insulation by design, cleanly separating high-level semantic reasoning from low-level motor execution without requiring explicit attention masking. Extensive evaluations across three simulation benchmarks and real-world robotic manipulation tasks demonstrate that CATok consistently surpasses existing tokenization methods in both reconstruction fidelity-compression tradeoff and inference efficiency, while improving VLA task success rate and training efficiency, establishing a high-performance, scalable foundation for purely autoregressive VLA systems.
[AI-38] A.D.A.M.O. (Agent for language-Driven Actions with Multimodal Observations): A Visual-Symbolic Framework for Virtual Humans
链接: https://arxiv.org/abs/2609.35463
作者: Alessandro Emmanuel Pecora,Stefano Calzolari,Francesco Strada,Andrea Bottino
类目: Artificial Intelligence (cs.AI); Graphics (cs.GR)
备注:
Abstract:Creating believable vh requires the coherent integration of perception, reasoning, and action mediated by language. A central challenge is to combine these components into a control loop grounded in interactive 3D environments. To this end, we present A.D.A.M.O. (Agent for language-Driven Actions with Multimodal Observations), a visual-symbolic framework for language-driven vh that leverages a pretrained vlm with tool calling to unify perception, reasoning, and action within a single control loop. A.D.A.M.O. maintains a dual visual-symbolic world model that combines egocentric visual input and synchronized symbolic state to support grounded task-oriented behavior from natural language prompts. To support diagnostic evaluation, we introduce a controlled task suite organized by a cd taxonomy that breaks down spatial tasks into procedural and linguistic complexity. Experiments in controlled scenes show that semantic labeling strongly influences task completion and failure modes, reducing perceptual ambiguity while shifting failures toward downstream execution, whereas reasoning errors remain comparatively rare.
[AI-39] AutoBCI: Forecast-Guided Agent ic Neural Architecture Discovery for EEG-Based Brain–Computer Interfaces
链接: https://arxiv.org/abs/2609.35456
作者: Muyun Jiang,Yi Ding,Wei Zhang,Jinbo Chen,Chenyu Liu,Zhenjie Yang,Yuxin Li,Jingyuan Chen,Yuhao Lu,Yong Li,Shuailei Zhang,Cuntai Guan
类目: Artificial Intelligence (cs.AI)
备注: 34 pages, 6 figures, including supplementary material
Abstract:EEG-based brain-computer interfaces support a broad range of applications, yet designing decoding architectures that perform well across diverse tasks remains challenging. We introduce AutoBCI, an agentic framework in which a Designer Agent and a Forecaster Agent support the discovery and selection of EEG decoding architectures across tasks. The Designer Agent performs Pool-Guided Architecture Discovery (PGAD), generating and refining architectures through training and validation across multiple EEG tasks, such as emotion recognition, motor imagery, and sleep staging. The Forecaster Agent performs Performance Estimation from Early Knowledge (PEEK), using architecture code, the training protocol, and early learning curves to predict full-budget validation performance and select promising candidates for continued training. Across 14 EEG datasets spanning motor imagery, emotion recognition, and sleep staging, we evaluate AutoBCI with six LLMs, including Opus 5.5 and GPT 5.6 Sol, and compare the architectures selected by the search procedure against ten baselines: six conventional EEG models and four foundation models. The architecture discovered by AutoBCI with Claude Opus 5.5 achieves 64.16% average test balanced accuracy (bAcc), compared with 63.87% for REVE, the strongest baseline on this metric. Using ten observed epochs, PEEK reduces mean absolute error in predicting average validation bAcc from 2.20 to 1.36 percentage points, a 38.1% reduction relative to the best-observed-score baseline.
[AI-40] Just Initialize: A Training-Free Initialization Component for Large-Scale Routing Optimization
链接: https://arxiv.org/abs/2609.35443
作者: Jiale Zhao,Sirui Mao,Zimu Chen,Wentao Yang,Zihan Wang,Xuefeng Huang,Junji Cheng,Liyuanjun Lai
类目: Artificial Intelligence (cs.AI)
备注: 31 pages, 5 figures
Abstract:Large-scale routing problems are difficult to solve efficiently as their search spaces grow rapidly with problem size. Existing approaches primarily improve the optimization procedure itself, often at increasing computational cost. We instead shift the focus to a useful initialization that can be refined into a high-quality solution with limited downstream refinement. We propose Just Initialize, a training-free and solver-agnostic initialization component for large-scale routing optimization. Just Initialize compresses a large routing instance into a compact surrogate space, optimizes its global routing structure, and recovers the resulting solution as an optimization-friendly starting point in the original space. Extensive experiments on Traveling Salesman Problems (TSPs), Capacitated Vehicle Routing Problems (CVRPs), Vehicle Routing Problems with Time Windows (VRPTWs), and Prize-Collecting Traveling Salesman Problems (PCTSPs) demonstrate that Just Initialize achieves high-quality solutions comparable to or better than state-of-the-art methods while substantially reducing computational cost across instances ranging from 1K to 100K nodes, including an average speedup of approximately 70 \times , sub-second runtimes on 10K-node instances, and runtimes within tens of seconds on 100K-node instances.
[AI-41] Riccati State Space Models: Non-iterative Parallelization for Nonlinear Sequence Modeling
链接: https://arxiv.org/abs/2609.35441
作者: Mónika Farsang,Ramin Hasani,Daniela Rus,Radu Grosu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:State space models (SSMs) achieve efficient sequence processing because their affine state updates are closed under composition and can therefore be evaluated with an associative parallel scan. Nonlinear recurrent models can provide richer, state-dependent dynamics, but generally lose this compositional structure: parallel evaluation then requires iterative methods that repeatedly linearize and scan the recurrence. We ask, what state-dependent nonlinear dynamics can be designed to remain exactly composable? We answer by introducing RiccatiSSM, a nonlinear SSM, in which each state dimension follows an input-conditioned Riccati differential equation. Its quadratic state dependence makes the local Jacobian explicitly state-dependent, while its exact per-step flow under piecewise-constant inputs is a Möbius transformation. Since Möbius maps are closed under composition and compose through 2\times 2 matrix multiplication, the complete nonlinear state trajectory can be evaluated exactly with a single associative parallel scan, without iterative linearization. We further derive a constrained parameterization that ensures bounded, contractive dynamics, and avoids poles in the fractional-linear state update. Across long-sequence classification, regression, and forecasting tasks, RiccatiSSM achieves competitive predictive performance while reducing runtime by 22-33% compared to the nonlinear LrcSSM under matched architectures. These results demonstrate that state-dependent nonlinear dynamics can retain exact composability and be evaluated efficiently within a single parallel scan.
[AI-42] Building Transformation Layers for Riemannian Neural Networks
链接: https://arxiv.org/abs/2609.35436
作者: Ziheng Chen
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Recently, deep neural networks on manifold-valued representations have garnered significant attention across various machine learning applications. One recent focus is the generalization of Euclidean fully connected (FC) and convolutional layers to non-Euclidean geometries. However, previous approaches typically focus on a few selected manifolds and rely on specific properties of the target manifold. In contrast, this work proposes a framework for constructing FC and convolutional layers over computationally tractable Riemannian spaces. This framework incorporates several previous FC layers across different geometries as special cases and is instantiated on ten representative manifolds, including three hyperbolic models, five geometries of the symmetric positive definite (SPD) manifold, and two Grassmannian perspectives. Experiments on different manifolds demonstrate the effectiveness and applicability of our approach. Code can be found at this https URL.
[AI-43] ReSPO: Reshaped Sequence Policy Optimization for Gradient Starvation in Off-Policy Learning
链接: https://arxiv.org/abs/2609.35433
作者: Yihang Chen,Yuanhao Ban,Cho-Jui Hsieh
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning from verifiable rewards (RLVR) frequently reuses rollouts across multiple policy updates, increasing the mismatch between the current policy and the data-generating policy. We identify a sign-dependent gradient starvation problem in clipped policy optimization: clipping suppresses under-generated positive responses at the low-importance-weight tail while permitting severely over-generated negative responses to dominate the high-weight tail. To address this, we propose ReSPO (Reshaped Sequence Policy Optimization), which replaces clipping with a smooth, two-branch sequence-level kernel derived from an \alpha -divergence variational objective and an exponential variance-control tilt. The positive branch preserves a nonzero gradient weight for under-generated positive responses, while the negative branch suppresses heavily over-generated negative responses. We demonstrate that ReSPO effectively learns from long positive reasoning trajectories during early training, even when accumulated policy drift relegates them to the low-importance-weight tail. On dense and MoE Qwen3 models, ReSPO accelerates early optimization, improves final training scores, and achieves higher held-out benchmark performance under a rollout reuse, validating our approach on importance-weight tail control in off-policy learning.
[AI-44] GLAD: Global-Local Adaptive Detector for Robust Speech Deepfake Detection
链接: https://arxiv.org/abs/2609.35411
作者: Zelin Zhao,Guanjie Huang,Danny Hin Kwok Tsang,Li Liu
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances in AI-based speech synthesis have enabled highly realistic speech, increasing the importance of speech deepfake detection (SDD) in preventing misuse. While mainstream Self-Supervised Learning (SSL)-based detectors achieve strong performance, they suffer from poor generalization to unseen domains and often overlook fine-grained signal artifacts due to a bias towards global semantic consistency. In this paper, we conduct the first detailed empirical and visual analysis to validate these limitations explicitly. Our investigation reveals two critical architectural vulnerabilities: (1) a systemic failure to capture localized spoofing traces, and (2) a severe lack of adaptability to domain-driven shifts in SSL layer importance, rendering static aggregation strategies prone to overfitting. To address these vulnerabilities, we propose the Global-Local Adaptive Detector (GLAD). Specifically, to capture localized forgeries, GLAD employs a Hierarchical Global-Local (HGL) backbone that explicitly bridges the granularity gap by fusing global linguistic and acoustic features with fine-grained local signal details. To counter layer importance shifts in out-of-distribution (OOD) scenarios, we introduce a Hierarchical Adaptive Gating (HAG) mechanism that dynamically recalibrates layer-wise focus in a sample-specific manner. Finally, to address shortcut learning induced by environmental biases, we introduce SaniBoost, a composite data augmentation strategy for robust signal standardization and noise sanitization. Extensive experiments demonstrate that GLAD significantly outperforms state-of-the-art methods, particularly on unseen domain this http URL code will be released upon publication.
[AI-45] "Nothing to See Here: Unintended Disclosure through Revision Traces of LLM Deliverables
链接: https://arxiv.org/abs/2609.35408
作者: Yage Zhang,Yukun Jiang,Yang Zhang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) assistants increasingly help users draft content for third-party recipients. During private drafting, the user or the model may introduce an item and later remove or replace it. The model may remove the item from the intended content but reveal it again when stating the edit. We call such statements revision traces. For example, after a user removes the password before sharing a configuration file, the model may delete it but leave a comment saying, “Removed the password ‘No****4!’ as requested.” A third-party recipient who sees only the delivered file can therefore recover the withdrawn password from the comment. In an in-the-wild analysis of three public conversation corpora, we identify 26,753 revision requests, of which 2,363 (8.8%) leave revision traces. We study them in greater depth under controlled conditions by introducing RevLeakBench, a benchmark of 100 tasks across five scenarios with a conversation track and an agent track. We measure trace occurrence, withdrawn-item recovery, trace position, and required-content retention. Across six models, about half of the deliverables in both tracks state the edit after a revocation, and a reader that sees only the deliverable can recover the withdrawn item from about 13% of them. Telling the model that its entire reply will be forwarded to the recipient still leaves revision traces in 36.4% of the deliverables. We compare prompt defenses and a delivery boundary, and propose an output-side filter that sharply reduces recovery with little loss of required content. We believe our work can benefit efforts to understand and mitigate unintended disclosure in LLM interactions.
[AI-46] Structural Alignment for Reliable Industrial AI: Bridging Physical Reality Data Models and Human Intent
链接: https://arxiv.org/abs/2609.35400
作者: Lizhi Xiao,Sihong Wu,Victoria Xiao,Yiqiao Song,Chen Gu,Jianwei Ma,Xinming Wu,Aimé Fournier
类目: Artificial Intelligence (cs.AI); Geophysics (physics.geo-ph)
备注:
Abstract:Artificial intelligence is increasingly deployed in critical industrial domains, including healthcare, energy grids, subsurface exploration, where failures can have severe consequences for human safety, system stability, and economic outcomes. Yet AI is still evaluated primarily through benchmark accuracy, a model-centric metric that fails to capture the structural complexity and risks of real-world deployment. We propose a framework that views industrial AI reliability as a problem of structural alignment across four interacting worlds: physical, representational, machine, and human cognitive. These worlds are connected through two interfaces: digitalization, linking physical reality to computational representations, and goal encoding, translating human cognition to the machine objectives. Together, they define the space of admissible solutions. We characterize the solution space through four attributes: existence, non-uniqueness, robustness, and interpretability and show how mismatches arise at interfaces and propagate across worlds to produce reliability failures. Applications to healthcare, energy grids, and subsurface exploration illustrate that although dominant failure modes differ across domains, for example, interpretability in healthcare, robustness in energy grids, and non-uniqueness in subsurface exploration, all originate from a shared structural mechanism. By shifting the focus from model-centric evaluation to system-level alignment, this framework offers a principled foundation for assessing and governing reliability in industrial AI systems.
[AI-47] he Hidden Ratio in Adam: Stable Structure Compression and Sign Dynamics
链接: https://arxiv.org/abs/2609.35392
作者: Yihe Zhou,Tongtian Zhu,Yingxiao Huo,Satya Prakash Dash,Can Wang,Samuel Kaski,Mingfei Sun
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Adam is the default optimizer for training modern deep neural networks, yet its adaptive behavior remains poorly understood due to the complex interaction between its first- and second-moment exponential moving averages (EMAs). We study Adam in the tied- \beta regime, where the two EMA decay rates are equal, and show that its adaptive dynamics can be expressed through a transformed ratio with approximately scale-stable behavior. Empirically, this transformed ratio exhibits a stable, heavy-tailed distribution across tasks, model scales, and training stages, in contrast to the variability of raw moment magnitudes. This empirical stability has both practical and conceptual consequences. First, we derive a recurrence for the transformed ratio, yielding a reparameterization of Adam that replaces the second moment with a compressible state. Leveraging its stable distribution, we show that a fixed 4-bit codebook is sufficient in our experiments to store this state without auxiliary scaling, achieving performance competitive with full-precision Adam. Second, the transformed ratio view clarifies Adam’s connection to sign-based methods: Adam reduces to sign-based momentum modulated by the transformed ratio, and replacing it with a constant recovers Signum as a limiting case. This perspective further provides a simple rule for transferring learning rates between the two methods. Together, these results suggest that tied- \beta Adam admits a simple and approximately stable ratio structure underlying its adaptive behavior and demonstrate its utility for both analysis and efficient implementation.
[AI-48] MCP Error Messages Written for Developers Hurt the Most Capable Agents Most
链接: https://arxiv.org/abs/2609.35381
作者: Xiaonan Xu,Wenjing Wu
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 15 pages, 6 tables. Submitted to the Journal of Systems and Software. Data and code: this https URL
Abstract:Many Model Context Protocol (MCP) servers wrap web APIs built for human developers, and their error messages tell the reader to run a command, edit a configuration, open a web page or wait. Many agents that read them can only call the server’s tools. In 150 widely used MCP servers, 949 of 3,001 error messages tell the caller what to do next, and half of these steps depend on something the server cannot see about the caller. On credential errors, 62 of 67 steps ask for a terminal command, a configuration change or a web page; on rate limits, 20 of 30 say to wait and retry without naming the call to repeat. We tested five OpenAI models that act only through the tools of Berkeley Function Calling Leaderboard tasks, and the agents did what the step said. On expired credentials, a terminal command in the step left 45% of tasks recovered, and the loss it caused grew from 18 points for GPT-5.5 to 69 for GPT-6 Astra. On a rate limit, GitHub’s “Wait before retrying.” left 6%. We tested two remedies. For MCP developers, naming a server tool in the step raised recovery on expired credentials to 84%, with the login tool in place of the command, and on a rate limit to 88%, with the call to repeat in place of the bare wait. For agent developers, deleting the step with a one-sentence prompt before the model reads it raised recovery on expired credentials to 82%.
[AI-49] From Pixel to Poses: Object-centric Tool Manipulation Learning from Human Demonstrations
链接: https://arxiv.org/abs/2609.35375
作者: Bangjun Wang,Longyan Wu,Yukun Wei,Shenghe Shao,Chaoyi Huang,Wenze Cui,Zetong Xu,Hanlin Wu,Long Chen,Yi Ma,Hongyang Li
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Scaling up robotic manipulation is primarily bottlenecked by the scarcity of real-world robot data. While recent approaches leverage human video demonstrations to mitigate this shortage, they remain computationally expensive and still rely on paired human-robot data for domain alignment. Although current state-of-the-arts excel at long-horizon tasks, they struggle with the delicate and precise control required for complex tool manipulation. To overcome these limitations, we introduce P2P-T, from Pixel to Poses for Tool Manipulation, a data-efficient, object-centric framework that learns tool use directly from human demonstrations. P2P-T bridges the cognitive and physical execution gap through a two-stage approach. First, pretraining an object-centric world model to extract stable pose priors; second, integrating these priors into an efficient, pose-aware low-level policy. By utilizing a robust automated data processing pipeline powered by modern foundation models, P2P-T completely bypasses the need for human-robot aligned data. This reduces overall training overhead drastically. With minimal per-task fine-tuning, our framework achieves a 73% improvement over the previous state of the art in execution performance on complex, real-world tool manipulation tasks that currently remain out of reach for standard large-scale pretrained models.
[AI-50] A decision-support system applied to Law: Reasoning and explainability of the decision
链接: https://arxiv.org/abs/2609.35370
作者: Jeremy Bouche-Pillon(IRIT, IRIT-MELODI, IRIT-ADRIA, IRIT-LILaC),Pascale Zarat{é}(IRIT, UT Capitole, IRIT-ADRIA),Yannick Chevalier,Nathalie Aussenac-Gilles(IRIT-MELODI, IRIT, CNRS)
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:The emergence of the digital transition brought an increasing need to control the processing of digital information, including in Law Enforcement Agencies (LEAs). At the EU level, in recent years, many regulations have emerged to control data processing and exchange. Texts other than the GDPR, such as the ‘‘Law Enforcement Directive (LED)’’, appeared to regulate specifically how Law Enforcement Agencies (LEAs) could process data. A formal representation of these regulations can be part of decision systems that support LEAs in processing data in compliance with the regulations. Although many new formalisms have emerged to represent legal norms and rules, few are provided with a reasoning mechanism. Furthermore, systems used in decision-making processes in critical contexts such as medical diagnoses or legal decisions cannot be fully automated, and the explainability of their results is essential to ensure user confidence in decisions. This explainability aspect, while crucial, is lacking in most modern approaches that rely on machine learning. This paper describes a framework to operate formal rules from regulations, by focusing on explainability of the decision. After describing the general architecture of the proposed decision support framework, the paper showcases how symbolic AI and the SPARQL query language can support legal reasoning. It then describes an algorithm to generate a justification for the reasoning results, and outlines the procedure to be followed when the reasoning does not lead to a satisfactory conclusion. We notably focus on a method based on decision trees to determine what additional information to request from the user.
[AI-51] Planarian: Managing Agent State with Statepoints
链接: https://arxiv.org/abs/2609.35366
作者: Jinnan Guo,Hao Mark Chen,Kapil Vaswani,Andrew Paverd,Peter Pietzuch
类目: Operating Systems (cs.OS); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:LLM agents solve complex tasks by iteratively changing files, invoking local tools, and interacting with remote services, which modifies state across their local environment and remote services. Today, agents and users must manage these changes explicitly, whether reverting exploratory actions or recovering from erroneous ones. Doing so safely requires coordinated actions, yet current agent harnesses lack unified abstractions and mechanisms for managing local and remote state consistently and efficiently. We describe Planarian, an agent runtime with state management that enables agents and users to recover from erroneous actions and explore alternative executions over consistent local and remote environment state. Planarian introduces the abstraction of agent statepoints, which are consistent, restorable point-in-time versions of the environment state. Planarian exposes three state-management primitives to agents and users: (i) snapshot creates a new statepoint spanning local and remote state without requiring external services to support checkpoints: it relies on efficient incremental process and file system snapshotting to capture local sandboxed state, and transparently records compensating actions to undo remote state changes; (ii) rollback restores the environment to a previous statepoint by reverting to a prior local checkpoint and replaying compensating actions for remote state changes; and (iii) fork creates multiple isolated branches from a statepoint, enabling the agent to explore alternatives in parallel. We show that Planarian enables agents to undo mistakes and explore alternatives in parallel, improving task quality by up to 15x, and allows users to recover from erroneous actions with only 3% overhead. Subjects: Operating Systems (cs.OS); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR) Cite as: arXiv:2609.35366 [cs.OS] (or arXiv:2609.35366v1 [cs.OS] for this version) https://doi.org/10.48550/arXiv.2609.35366 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-52] Do Temporal Link Predictors Need Learned Memory? A Smoothed-Count Baseline with a Handful of Parameters
链接: https://arxiv.org/abs/2609.35364
作者: Lisi Qarkaxhija,Ingo Scholtes
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Many temporal link predictors summarize past interactions through learned node representations. We examine whether simple counts of recurring interaction patterns can provide competitive predictions without learning these representations. We propose a temporal link predictor based on statistical language modelling. It pools transition and co-occurrence counts across sources to predict links that a source has never formed. We smooth sparse estimates using destination frequencies or Kneser-Ney continuation counts. A shared log-linear rule combines these estimates with popularity, source history, and recency, without node embeddings. In our main evaluation, the model achieves the highest MRR among the compared methods on 7 out of 16 datasets from TGB and TGB-Seq. It also outperforms EdgeBank and Base3 on all 16 datasets and the heuristic family on 14. These gains extend to datasets designed to limit repeated edges. With only 9–13 learned parameters, our model provides a simple and competitive baseline for evaluating future neural temporal link predictors.
[AI-53] d-OPD: Future-Aware On-Policy Distillation for Block Diffusion Language Models
链接: https://arxiv.org/abs/2609.35362
作者: Ruitao Liu,Qinghao Hu,Song Han
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) typically generate text autoregressively (AR), predicting one token at a time. Block diffusion language models (dLLMs) instead generate blocks sequentially while denoising multiple tokens in parallel within each block, offering a promising way to accelerate generation. Rather than training such models from scratch, recent work adapts strong pretrained AR models into block dLLMs through distillation. On-policy distillation (OPD) has been widely used for LLM training because it supervises the student on states generated by its current policy, rather than only on fixed offline trajectories. By training on the states the student actually visits, it reduces the mismatch between training and generation and can provide more relevant supervision as the student evolves. Recent work has extended this idea to AR-to-block-diffusion conversion. However, this setting introduces a fundamental mismatch in supervision: the block-diffusion student and the causal AR teacher condition on different information at the same training state. The student predicts from the entire partially denoised block, including visible future context, whereas the standard AR teacher target is defined only from the causal prefix. As a result, the teacher distribution used for distillation is not fully aligned with the information available to the student. We therefore introduce d-OPD, a future-aware on-policy distillation method that corrects the AR teacher distribution to better align with the student-visible state by incorporating visible future information within each block, providing supervision that better matches the information used by the student. Across Qwen3 models from 0.6B to 8B, d-OPD improves the six-benchmark average by up to 4.0 points over OPDLM and reduces training time by 1.35 - 1.58\times . The code is available at this https URL.
[AI-54] Dont Inoculate Everything: Stratified Inoculation Prompting Narrows Backdoor Triggers and Preserves Desired Traits
链接: https://arxiv.org/abs/2609.35356
作者: Kajetan Dymkiewicz,Tim Farrelly,Adam Práda,Ishaan Panigrahi,Srishti Gureja,Helen Yannakoudakis,Robert Mullins,Victor Gillioz,Daniel Tan,Maxime Riché
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Supervised fine-tuning can teach language models undesired behaviours alongside desired ones. Inoculation prompting (IP) aims to limit unwanted generalisation by requesting the undesired behaviour during training and removing the request at inference. However, undesired behaviour can still appear under unrelated prompts. IP can also hinder learning of the desired behaviour. We address these limitations in settings where both behaviours co-occur in most training examples, so filtering out examples with undesired behaviour leaves only a small clean subset. We introduce stratified inoculation prompting (SIP). SIP leverages a small clean subset to demonstrate that desired behaviour should persist without the undesired one across different contexts. SIP oversamples these clean examples under diverse non-eliciting prompts while inoculating the rest. SIP substantially reduces expression of undesired behaviour while preserving more of the desired behaviour than IP. These gains persist even when we extend IP to oversample the same clean subset at the same rate as SIP. Moreover, SIP yields lower emergent misalignment rates in all harmful-advice setups we tested. SIP can be further extended to limit the undesired behaviour even under prompts that explicitly request it. We introduce backdoor dilution, which weakens expression under the inoculation prompt, and password-locked inoculation, which concentrates elicitation on a designated password. Taken together, our findings show that changing the training contexts for a small clean subset can significantly improve selective generalisation.
[AI-55] From Data to Program: Fast Direct Generative Program Inference from Empirical Data
链接: https://arxiv.org/abs/2609.35348
作者: Simon Klüttermann,Xueying Ding,Leman Akoglu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 51 pages, 20 figures,
Abstract:Estimating probability densities from a finite set of samples typically requires dataset-specific model fitting. We introduce PRODiGI, a pretrained data-to-program model that infers an explicit, executable generative program in a single forward pass. Pretrained on synthetic datasets paired with their ground-truth programs, PRODiGI accommodates diverse generative families and data dimensionalities through template prediction and non-autoregressive program parameter decoding. Its inferred programs support direct sampling, density and score evaluation, and inspection independently of the pretrained model. We further introduce program-space fine-tuning, which refines differentiable program parameters by matching generated and empirical samples while keeping model parameters intact. Experiments show that PRODiGI achieves lower average density and score MAE than existing pretrained models, while offering multi-fold speedups over its closest competitors. Program-space fine-tuning further reduces generation MMD by 84%. By turning empirical data into explicit, reusable programs, PRODiGI introduces a new direction for fast, interpretable tabular generative modeling.
[AI-56] Jev thinks "I dont know but doesnt say it: Introducing Sys1Cal-v1 Dataset for Probability Calibration
链接: https://arxiv.org/abs/2609.35342
作者: Riccardo Porcedda
类目: Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO); Systems and Control (eess.SY)
备注:
Abstract:The appearance of Jev marked the era of System One Models, foundation models that return structured decisions with probability distributions rather than text. Aside from low cost and great speed, Jev’s central promise is that these probabilities are calibrated: such claim is not backed by any public test and available external benchmarks evaluate confidence calibration, not whether every returned option probability has the right numerical meaning. To tackle this issue, we introduce Sys1Cal-v1, a dataset of True/False questions about a proposition A for which the exact probability P(A) is known by construction. Each item is queried through the three Jev primitives - Noul, Choice and Score - and evaluated by total variation distance from the ground-truth distribution, which can be used to estimate a soft accuracy of System One Models. We showcase the utility of Sys1Cal-v1 as a benchmark dataset by evaluating Jev and SemIf, an open-source Choice-style baseline. In this work, however, we focus even more deeply on Jev, by studying the calibration of its Score and Choice answers. In particular, we discover a peculiar behaviour that can be explained by assuming that Jev suppresses a third truth value, going beyond True and False. In other words, in \textttChoice answers, P(A) and P(\neg A) are presented as if P(A)+P(\neg A)=1 , while a term P(U)\neq0 is missing in the sum. Recovering P(U) leads to an improvement of median soft accuracy in \textttChoice answers from 0.771 to 0.978 , suggesting that, even in binary decisions, Jev wants to answer with a third option:``I don’t know’'. Subjects: Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO); Systems and Control (eess.SY) Cite as: arXiv:2609.35342 [cs.AI] (or arXiv:2609.35342v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.35342 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-57] Large Language Models for Automated Cross-Domain Machine Learning Task Type Identification: A Benchmark Dataset and Evaluation
链接: https://arxiv.org/abs/2609.35335
作者: Petros Tsialis,Steffen Limmer,Tobias Rodemann,Martin Heckmann
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 25 pages
Abstract:Machine learning task type identification is essential for constructing valid ML pipelines, yet in practice it is typically specified manually. We investigate whether large language models (LLMs) can infer both the data domain and the downstream prediction task directly from dataset-level information when only the target feature is provided by the user. Together with our LLM-based system we also release an annotated benchmark comprising 625 public tabular and time series datasets. We evaluate the proposed approach in three settings: (i) tabular datasets in comparison with established AutoML heuristics, (ii) cross-domain evaluation across tabular and time series datasets, and (iii) a practical deployment scenario using smaller local models. The results show consistent advantages for LLM-based task type identification, with increasing difficulty in heterogeneous and resource-constrained settings. LLM-based approaches outperform AutoGluon in the tabular setting, reaching 0.98 F1 macro compared to 0.93. In the cross-domain setting, the best model achieves 0.90 F1 macro, while smaller locally deployable models reach 0.75, indicating a trade-off between deployment feasibility and accuracy.
[AI-58] Reverse Sequential Proportional Approval Voting Rule: Proportionality and Approximation Guarantees
链接: https://arxiv.org/abs/2609.35331
作者: Georgios Papasotiropoulos
类目: Computer Science and Game Theory (cs.GT); Artificial Intelligence (cs.AI)
备注:
Abstract:We study the Reverse Sequential Proportional Approval Voting Rule (RevSeqPAV) in approval-based committee elections. Despite its historical prominence and practical use, its properties and guarantees are much less understood than those of Sequential PAV. We analyze it along two dimensions: proportional representation (measured by Extended Justified Representation, its approximations, and proportionality degree) and approximation of the maximum PAV score of instances. We first establish strong negative results for general, unrestricted election instances and then identify settings in which the rule provides meaningful fairness and optimization guarantees.
[AI-59] Hyper Algorithm Design Agent : Evolving Learnable Optimizer from Zero
链接: https://arxiv.org/abs/2609.35328
作者: Zipei Yu,Yue-Jiao Gong,Zeyuan Ma,Yuncheng Jiang,Zhiguang Cao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Meta-Black-Box Optimization (MetaBBO) is one of the highlights in the recent AI for Optimization trend. This paradigm’s bi-level workflow leverages the learnable algorithm design policy at meta level to ensure the performance and generalization improvement on the low-level optimization task. While MetaBBO helps advance the performance lower bound of the resulted optimization system, it is currently handcrafted and customized case by case to adapt different optimization problems, which inevitably introduces inherent subjectivity and hence restricts the performance upper bound and usability in practice. In this paper, we address this issue by regarding MetaBBO’s design loop as coding task, where we could introduce openendedness into MetaBBO with recursive self-improvement capability of advanced coding agents. Specifically, we propose a dual-agent framework: i) a task agent continuously refines the codebase of a target MetaBBO approach through code evolution; ii) a hyper agent progressively modifies the task agent and itself to provide open-ended design behavior; iii) the evolved MetaBBO codebase is evaluated and all in-execution information is fed back to the agents for recursive self-referential improvement. As a result, given a naive MetaBBO template, our framework automates a design evolution and finds novel variants superior to up-to-date human-made MetaBBO baselines. Surprisingly, the experimental results also demonstrate that our framework supports fast adaption across different optimization domains. Solid interpretation analysis further reveals interesting design principles emerge in such open-ended process. This work serves as the first exploration on automating design of complex learning-assisted optimization algorithms.
[AI-60] acher-Student Gaps Are Not Enough: Outcome-Guided On-Policy Distillation for Multi-Turn Autonomous Agents
链接: https://arxiv.org/abs/2609.35319
作者: Tong Zhang,Zhou Liu,Yihao Liu,Jiahua Bao,Xuchen Li,Honglin Lin,Tao Cheng,Zhihan Yu,Kai Tang,Xiaoxi Jiang,Guanjun Jiang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:On-policy distillation (OPD) trains a student on its own trajectories with dense teacher supervision. Recent work on OPD for multi-turn autonomous agents often treats large teacher-student token-level distributional gaps as promising intervention points, linking larger gaps to a greater need for correction. Yet, our empirical analysis reveals a supervision-benefit mismatch: large gaps can be benign, while small gaps can be outcome-critical. Teacher-student gaps capture differences at the current turn, whereas the benefit of teacher guidance depends on how the current student interacts with the environment afterward. The student may still succeed despite choosing an action that differs from the teacher’s, while a teacher-preferred action may lead to a state from which the student cannot complete the task. Local gaps alone are therefore not enough to determine whether teacher guidance benefits the current student. Effective supervision should instead emphasize guidance that the current student can translate into better final task outcomes. Accordingly, we propose Outcome-Guided On-Policy Distillation (OG-OPD), which applies trajectory-relative weighting to teacher supervision and calibrates these weights using final task outcomes from paired student continuations. This calibration selectively strengthens supervision on the student’s original trajectories at turns where teacher guidance benefits the current student. Across ALFWorld, ScienceWorld, and WebShop, OG-OPD consistently outperforms baselines under diverse settings. It improves task success rates by 3.6-17.7 percentage points over vanilla OPD and by up to 7.0 percentage points over the strongest baseline.
[AI-61] Reliability Engineering for AI Systems: Challenges Methods and Directions
链接: https://arxiv.org/abs/2609.35316
作者: Rong Pan,Yili Hong,Min Xie
类目: Artificial Intelligence (cs.AI); Applications (stat.AP)
备注:
Abstract:AI reliability concerns whether an AI system performs its intended function dependably over a stated period and under stated operating conditions, with stated evidence. As these systems become more autonomous, that function includes more than a correct output. Retrieval, memory, tool use, permissions, human oversight, and interactions among systems must operate consistently and safely, and, for generative systems, so must the reasoning process that produces the output. Average benchmark accuracy measures capability; it does not quantify this broader reliability claim. This paper adapts established reliability engineering methods, from failure definitions and operational envelopes to FMEA, accelerated testing, field monitoring, and reliability growth, to AI systems. A four-level diagnostic framework classifies failures as component, operational-loop, agentic-conduct, or network and governance failures. Test, evaluation, verification, and validation (TEVV), sequential monitoring, and FRACAS create and refresh evidence. SMART provides statistical guidance for measurement, analysis, assessment, and test planning; the NIST AI Risk Management Framework provides organizational guidance for governance, evaluation, monitoring, and mitigation. Three cases illustrate the program: adversarial testing of a convolutional neural network, perception-error propagation, and autonomous-vehicle disengagements. Established reliability engineering provides a usable foundation; new measurements and safety guardrails are still needed as these systems are self-evolving.
[AI-62] Narrowing the Horizon: Quantifying Topic Saliency Shifts in Generative Monoculture EMNLP
链接: https://arxiv.org/abs/2609.35302
作者: Oriane Peter,Elena Simperl,Kate Devlin
类目: Artificial Intelligence (cs.AI)
备注: Accepted at EMNLP Findings 2026
Abstract:As Large Language Models (LLMs) become central to how we access and share information, they play an increasingly powerful role in shaping global knowledge. However, as these models evolve, their outputs risk converging into a \textitgenerative monoculture, where the diversity of perspectives they represent narrows over time. Studies at the model level often fail to pinpoint which specific topics or viewpoints are being marginalised or amplified in this process. In this paper, we introduce a method to measure shifts in topic saliency across model families, tracking what gains or loses prominence during post-training. Applying this approach to a case study of climate change discourse, we demonstrate how homogenisation affects the representation of diverse solutions across different models. We also test interventions to counter this trend, showing that specialised models can help preserve a broader range of perspectives. This underscores the importance of monitoring topic saliency to diagnose the risks of monoculture and to ensure AI systems reflect a pluralism of ideas. Data and Code are accessible \hrefthis https URLhere.
[AI-63] raining-Free Clinical Reasoning through Medical Ontologies and Cognitive Mapping: A Symbolic-Probabilistic Knowledge Graph Framework
链接: https://arxiv.org/abs/2609.35298
作者: Surajit Das
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Most clinical prediction systems learn patient-variable-outcome associations; we investigate a training-free diagnostic paradigm mapping patient observations to explicit medical knowledge. CKG Reasoner integrates candidate-specific Evidence Feature Nodes, patient-reference matching, a bounded Information Gate, knowledge-weighted evidence accumulation, disease similarity, and decisive clinical rules. Missing-aware normalization and coverage auditing distinguish absent from unavailable evidence. Candidate ranking is separate from outcome-label-independent K-means clustering, which uses four derived evidence coordinates (evidence strength, relative magnitude, directional similarity, and evidence completeness), not raw predictors or targets, to derive cohort-level assignments. Across six retrospective cohorts - four dengue (N = 1000, 1523, 989, 1018), malaria (N = 2190), and influenza (N = 4569) - a uniform, label-free, cohort-fitted K = 2 protocol yielded positive-class F1 scores of 0.996, 0.634, 0.936, 0.917, 0.695, and 0.842, and all-record accuracies of 0.996, 0.558, 0.914, 0.893, 0.707, and 0.906, respectively, with full partition-decision coverage using the frozen package and disease-specific knowledge representations. Neither scoring nor clustering uses outcome labels. Logistic regression provides a supervised baseline. Influenza incorporates confirmatory molecular PCR and is not independent pre-test prediction. Results characterize knowledge-grounded evidence separation, auditability, and sensitivity, not prospective clinical validity or comparative superiority. FOL/LLM-based clinical explanation remains unevaluated.
[AI-64] AbGaze: Attentive Geometric Representation Learning for End-to-End Antibody Design
链接: https://arxiv.org/abs/2609.35296
作者: Jiashuo Wang,Siqi Fan,Yizhen Luo,Zaiqing Nie
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Computational antibody design requires representations that capture the geometric patterns underlying antigen–antibody interactions, yet existing approaches often rely on scalar distances or surface-intrinsic features, leaving cross-molecular geometry largely implicit. We present AbGaze, an end-to-end antibody design framework based on attentive geometric representation learning, which encodes distance, spatial direction, and surface-normal orientation of antigen surfaces relative to antibody-residue local frames, and adaptively aggregates these geometric interactions according to their interfacial context. The learned interaction representation is shared across multi-CDR co-design, complex structure prediction, and affinity optimization, with local-frame geometric supervision further constraining the representation. AbGaze outperforms prior methods across all three tasks: relative to the second-best method, it improves amino-acid recovery by 7.1% and reduces structural error by 14.9% on average over the six CDRs, improves interface docking quality (DockQ) by 6.6%, and raises the affinity improvement rate (IMP) by 32.5%.
[AI-65] he Argument and the Letterhead: Source-Position Coherence in AI Evaluation
链接: https://arxiv.org/abs/2609.35286
作者: Michele Loi
类目: Artificial Intelligence (cs.AI)
备注: 28 pages, 6 figures, 2 tables. Preregistrations, materials and code available on GitHub. Preprint; not yet peer reviewed
Abstract:An argument can be surprising coming from a particular speaker without being a bad argument. Do AI evaluators keep these judgments apart? Two preregistered descriptive studies and a later Jev supplement collected 2,976 usable evaluations of six fixed texts about US AI policy, Germany’s debt brake and Swiss nuclear energy. Each text was presented under several source attributions. The key comparison asks whether the gap between two sources changes when the argument changes. On Sol, for example, a national-security argument received mean ratings of 0.359 under CODEPINK and 0.639 under College Republicans; a civil-rights argument received 0.742 and 0.721. A constant preference for one source cannot explain that pattern. Related interactions appeared across topics and recent model configurations, including those with reasoning enabled, while several comparisons yielded small effects. The later European Jev supplement yielded five interactions below the adopted absolute reference of 0.05; its distinct rubric and interrupted collection limit comparison with the chat systems. Some written evaluations explicitly invoked a mismatch between a source and its attributed position. Taken together, the numerical and verbal evidence supports source-position coherence as a plausible explanation, alongside competing accounts involving credibility, authenticity and interpretation of the task. The paper develops this inference through controlled comparisons, reports conditional post hoc p-values in an appendix, and documents the human decisions and delegated checks behind an AI-conducted study.
[AI-66] xtual User Taste: Natural-Language User Context for Foundation-Model Recommender System at Scale
链接: https://arxiv.org/abs/2609.35285
作者: Ghazal Fazelnia,Paul Gigioli,Eliza Klyce,Sharon Zheng,Katie Zelvin,Ye Myat Thein,Anurag Deshpande,Seda Davtyan,Kate Remeika,Maya Hristakeva,Erik Franco,Karen Banzon,Peng Ge,Jacqueline Wood,Nandini Singh,David Murgatroyd,Mounia Lalmas,Yves Raimond,Andreas Damianou
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Foundation model recommender systems require user context that can be consumed by large language models, reasoned over, and refined through natural-language interaction. Traditional behavioral embedding vectors remain highly effective for retrieval and ranking, but they are opaque to users and not natively expressed for language model workflows. We present Textual User Taste, a system that generates structured natural-language taste profiles from listening behavior, interaction signals, content metadata, and optional user feedback, and deploys them to millions of Spotify users. We describe the end-to-end production lifecycle required to generate, evaluate, optimize, and maintain these representations at industrial scale, including prompt development and compression, user steering, and integration with downstream personalization systems. Because no unique ground-truth taste profile exists, we introduce a multi-faceted evaluation framework to evaluate taste profiles as a production representation: they carry user-specific predictive signal independently, and when integrated with behavioral embeddings, improve MRR by 0.6% for future-track prediction and NDCG@7 by 2.2% for search ranking. Our evaluation also reveals that taste profiles support positive natural-language steering, while exposing important limitations, including challenges with negation and short-term temporal adaptation. These findings position taste profiles not as replacements for behavioral embeddings, but as an interpretable and steerable interface between evolving user context and foundation-model recommender systems.
[AI-67] Multi-Attractor GNNs: Set-Valued Expressivity Beyond Unique Equilibria
链接: https://arxiv.org/abs/2609.35274
作者: Jialin Liu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Recurrent and equilibrium graph neural networks (GNNs) often enforce a unique fixed point or use one training target per graph. Yet many combinatorial and scientific problems admit multiple valid solutions, with no preferred one. A designated target can then impose an arbitrary selection rule. For tasks invariant to node relabeling, a symmetric graph may have a symmetric solution set but no symmetric solution. We show that multiple equilibria enable one weight-tied message-passing GNN to represent set-valued equivariant maps: different initializations approach different valid solutions. Under stated regularity assumptions, we first construct globally Lipschitz, permutation-equivariant dynamics that converge almost surely to valid solutions and reach every solution branch with positive probability. We then establish approximate realization by recurrent message passing with continuous component maps, with arbitrarily small update and limiting errors and arbitrarily high probability. This goes beyond standard universality arguments: although message passing alone cannot distinguish symmetric nodes, the evolving state keeps nodes distinguishable at every finite step without auxiliary node identifiers. Such dynamics can be learned without solution labels using problem-specific energies. On Ising ground states, structural module detection in protein graphs, and chemical reaction steady states, the learned updates produce multiple high-quality predictions with high numerical convergence rates. They achieve better average solution quality than the tested unique-equilibrium, single-target, and feedforward baselines, while remaining competitive with much larger diffusion-based solvers. Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.35274 [cs.LG] (or arXiv:2609.35274v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.35274 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-68] WavePP: High-Throughput Pipeline Parallel LLM Prefill under Prefix Reuse
链接: https://arxiv.org/abs/2609.35263
作者: Aaryam Sharma
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI)
备注: 33 pages, 14 figures, 11 tables
Abstract:Pipeline parallelism can improve prefill throughput by processing multiple request chunks concurrently across different stages of the model. However, keeping the pipeline fully utilized requires efficient scheduling and request preparation. In systems where stages retain and evict cache state independently, a local cache hit does not guarantee that the same prefix can be reused across the pipeline. Here, coordination overhead can impede request admission cadence and thus reduce overall throughput. In this paper, we present WavePP, a prefill runtime built on top of TensorRT-LLM that addresses these challenges by overlapping request admission with pipeline execution. WavePP asynchronously finds a prefix that can be reused across all stages, protects the cached state, and reserves space for the remaining input while earlier requests continue to execute. It subsequently plans the chunk sizes of each request dynamically to maximize pipeline fill. Each stage then completes the local preparation before executing the request. In the same system and pipeline topology, WavePP improves TensorRT-LLM’s prefill throughput in 37 of 40 tested settings on GLM 5.2 and MiniMax M2.7. At concurrency 128 with high cache reuse, these changes increase throughput by factors of 2.91 and 2.02, respectively. Across 28 Kimi K3 settings, WavePP also has the highest measured throughput in all 18 settings at concurrency eight or higher, compared with tensor/expert-parallel and pipeline-parallel baselines from TRT-LLM, SGLang, and vLLM.
[AI-69] Imprint Reader: From Weight-Update Readout to Behavioral Intervention
链接: https://arxiv.org/abs/2609.35261
作者: Guanxu Chen,Qihao Lin,Jing Shao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As language models take a growing role in AI development, a natural aspiration is for them to reflect on their own learning process, as humans do, and use that reflection to improve themselves. At the same time, these models have an advantage that human learners lack, since training leaves parameter-level traces that can, in principle, be inspected directly. However, current models cannot decode these traces into an explicit account of what they have learned. To this end, we introduce the \textitImprint Reader, a model trained with \textitSemantic Mount-and-Read Tuning (SaRT) to describe frozen weight updates. SMaRT mounts each update onto the Reader and uses an anchor-free meta-query to elicit a natural-language description, while no-change and random-perturbation controls discourage unsupported claims. On held-out updates, the joint Reader reaches judge-based Pass@100 of 2% for knowledge and 16% for behavior. These results demonstrate the feasibility of natural-language readout while pointing to reliability across updates as the next step. Beyond free-form generation, the Reader provides a differentiable proxy for the gap between a specified target behavior and a candidate weight update. Its coordinate-aligned gradients support intervention through MetaEdit. At a 0.5% pruning rate, Reader-guided selection raises measured harmful-prompt refusal from 57.9% to 64.1% under a safety-maintenance target. Using behavior descriptions without target-task training data, MetaEdit increases the frequency of backtracking and sub-goal expressions in mathematical reasoning traces and raises BFCL Overall from 41.69% to 44.60% .
[AI-70] owards Reliable AI Data Scientists: Data Agents with Workflow Harnesses
链接: https://arxiv.org/abs/2609.35255
作者: Huachi Zhou,Yujing Zhang,Jiahe Du,Jiacheng Cai,Zijin Hong,Chuang Zhou,Zheng Yuan,Qinggang Zhang,Qing Li,Xiao Huang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model agents are increasingly deployed for data-intensive work, yet reliable data analysis requires more than general-purpose reasoning and ad hoc tool augmentation. Data Agents, equipped with workflow harnesses, offer a promising paradigm for automating the end-to-end data science lifecycle. This paper examines Data Agents from a harness-centric perspective. First, we introduce a taxonomy of Data Agents and associated data environments, organizing the literature around five functional stages: perception, planning, execution, verification, and repair. Second, we analyze the key technical routes within each stage, identifying 15 distinct approaches ranging from data structure probing to data state reconstruction. Third, we identify four open reliability problems: inactive semantic calibration, missing clarification, missing experience transfer, and the missing verification-repair repository. These problems explain why silent failures can persist even when individual components function correctly, highlighting the need for rigorous workflow harnesses and shared reliability resources. Finally, we summarize the horizontal task families of Data Agents, examine their vertical application settings, and benchmarks for evaluation, while maintaining a companion repository at this https URL.
[AI-71] EP-Mem: Elastic Privacy Memory for Social Relationship-Aware LLM Agents
链接: https://arxiv.org/abs/2609.35233
作者: Fengzhou Sun,Yuan Zhang,Xintong Yu,Jinyao Yan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents face critical privacy risks when acting as delegates in human-agent-human communication. To prevent such breaches, agents must understand users’ social relationships and adhere to context-dependent social information disclosure boundaries. Current studies on agent memory privacy focus on instantaneous interactions, leaving the long-term relational disclosure problem unexplored. In this paper, we propose EP-Mem, an Elastic Privacy Memory architecture that reframes privacy as user-owned boundary control across social roles. EP-Mem introduces (1) token-level memory driven by user-configurable a privacy policy that stratifies persons and events, combining domain-level default circulation rules with fact-level whitelist/blacklist exceptions; and (2) a pluggable sidecar with a privacy engine that aligns disclosure controls with memory across summary, detail, and boundary granularities, enforced throughout generation, storage, and retrieval. We construct EP-Bench, to our knowledge the first long-term multi-party benchmark with cross-session correlated events for policy-conditioned relational disclosure. Experiments show that EP-Mem achieves 94.0% privacy classification accuracy, improves disclosure-permission judgment from 22% to 68%, and reduces privacy leakage by 75.6%, while maintaining retrieval performance and cross-benchmark generalization.
[AI-72] ASCT: Attentive Search over Counterfactual Trees for Credit Assignment in Agent ic Reinforcement Learning
链接: https://arxiv.org/abs/2609.35215
作者: Yang Li,Jinhan Yang,hai liu,Di Wan,Xiyu Chen,Zongsi Xu,Tuo Zhou,Sheng Zhong,Sergey Volkov,Ye Luo,Hao Sun
类目: Artificial Intelligence (cs.AI)
备注: 27 pages, 9 figures, 19 tables
Abstract:Terminal utility evaluates a complete agentic workflow, but learning requires credit for the decisions within it. We introduce Attentive Search over Counterfactual Trees (ASCT), a framework that turns training-time multi-step search into local action credit. At actor-visited states, an auxiliary tree evaluates alternative legal actions from the same recoverable prefix. Its action-value table is centered by the frozen actor’s probabilities and supplies credit for PPO on actor-sampled trajectories. This protocol connects counterfactual evaluation to policy learning while deploying the actor alone. Uniform, UCT, and cost-aware AgentUCT instantiate the framework. On HotpotQA agentic retrieval-augmented generation, all three improve mean held-out utility over trajectory-return PPO and workflow-adapted VinePPO. Across three seeds, ASCT-AgentUCT reaches 0.6187 utility versus 0.5939 for VinePPO, with gains in answer F1 and execution cost, and uses 50.3% fewer recorded auxiliary Qwen tokens. Transfer and component-description studies examine the learned policies beyond the training setting.
[AI-73] GAC-PINN: Geometry-Adaptive and Constraint-Enhanced Physics-Informed Neural Networks
链接: https://arxiv.org/abs/2609.35196
作者: Yanxin Zhang,Yong Zhang,Houbiao Li
类目: Numerical Analysis (math.NA); Artificial Intelligence (cs.AI)
备注: 21 pages, 9 figures, 5 Tables
Abstract:For systems with steep gradients, sharp interfaces, or severe spatio-temporal coupling, Physics-informed neural networks (PINNs) suffer from spectral bias, geometric inflexibility, and boundary constraint conflicts, which undermine accuracy and convergence. To overcome these issues, we propose a geometry-adaptive and constraint-enhanced PINN (GAC-PINN). The framework comprises four components: a gradient-driven adaptive grid mapping (AGM) for diffeomorphic point concentration with Jacobian regularization, an adaptive bandwidth hard-constraint ansatz with spatially-varying boundary transition widths, a Gaussian Fourier feature mapping as a spectral preconditioner to further enhance high-wavenumber representation, and an operator-aware router that automatically selects the appropriate hard-constraint construction based on whether the governing PDE contains temporal derivatives. An AGM callback mechanism and a three-stage training strategy ensure stable coordination. Benchmarks including the viscous Burgers equation, a sharp-peaked 2D Poisson problem, and the Allen-Cahn phase-transition equation show that GAC-PINN attains relative (L^2) errors of ((1.747\pm 0.450)\times 10^-4), ((2.868\pm 0.947)\times 10^-5), and ((1.756 \pm 0.712)\times 10^-3), respectively, consistently outperforming the baselines. Ablation studies further reveal that AGM alone yields a substantially lower error than residual-based adaptive refinement (RAR), while RAR becomes beneficial only when combined with FFM, demonstrating a context-dependent module interaction. Convergence analysis verifies rapid error reduction and saturation with increasing resolution, establishing a practical adaptive framework for high-fidelity simulation of problems with localized sharp features in applied mechanics and computational physics.
[AI-74] Beneath the Tokens: A Performance Engineering Study of Multi-Token Prediction in GPU-Accelerated LLM Inference
链接: https://arxiv.org/abs/2609.35188
作者: Suwesh Prasad Sah
类目: Artificial Intelligence (cs.AI); Performance (cs.PF)
备注:
Abstract:Autoregressive large language model inference repeatedly invokes the target model to generate one token at a time, making generation sensitive to GPU memory movement and sequential execution. This study evaluates two-token multi-token prediction (MTP) against autoregressive decoding in a controlled single-request deployment on an NVIDIA A10G GPU. A 360-request benchmark covered plain-text, reasoning-intensive, and tool-calling workloads, while runtime telemetry, Nsight Systems, PyTorch Profiler, and selected Nsight Compute measurements were used to explain the observed performance. MTP increased output throughput by (1.91\times) to (2.19\times) across all prompts and reduced time to first output by 10.0–14.2%. Median mean acceptance length ranged from 2.370 to 2.595 tokens per verification iteration. Profiling showed that MTP introduced a longer and more complex execution path, including proposal, sampling, attention, gathering, and reduction operations. However, it required 56.4–78.1% fewer executions of the selected repeating CUDA Graph per generated token. The dominant MTP GEMM kernel was not faster than the dominant autoregressive GEMV kernel, and selected instances of both approached the A10G memory-bandwidth limit. These results show that MTP improved inference through amortization: greater token progress reduced repeated GPU execution sufficiently to outweigh the additional speculative-execution cost.
[AI-75] Research-Native by Construction: Minimal Nodes Re-verifiable Workflows and Compounding Memory for Long-Horizon Scientific Agents
链接: https://arxiv.org/abs/2609.35182
作者: Di Wang,Yu Liu,Bing Cui,Chaoqun Ji,Dongyuan Ni,Jingyu Lu,Kunlei Cui,Pu Qin
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 34 pages, 8 figures, 8 tables
Abstract:We describe AfS (Agent for Science), a platform built for long-horizon scientific work, where a project runs for tens of hours across dozens of agent runs with a human present only occasionally. Most agents for science are general coding agents with a skills folder attached, and they inherit that lineage’s failure mode: under pressure to finish, they fabricate, skip, or smooth over. Our design rests on one claim: most of the credibility of machine-made research can be moved from asking the model to behave to making the non-compliant state unrepresentable. We encode research discipline as mechanically enforced laws (commitment before measurement; unforgeable freezing; reports are not facts; evidence persists but verdicts do not; negative results are first-class; mechanical questions to the framework and semantic judgment to the model), organized around three time horizons: a minimal set of research nodes within a run, an inquiry contract with frozen closure conditions and a hash-chained artifact ledger within a project, and a two-tier knowledge base with promotion by rewriting across projects. This is a system description written under one rule: each mechanism appears in exactly one place, with the invariant it enforces, the failure it prevents, the way it is realized, and the cost it imposes. It covers the node contract, the write-path gates, the two-tier memory, and the runtime substrate. Three traces walk real failure attempts through the mechanisms that catch them, and two closed campaigns are included as worked illustrations rather than as an evaluation. We report no benchmark: a process-integrity suite that would support quantitative comparison is under construction, and what we can measure today is only the operating cost of the machinery.
[AI-76] QAM: Quadratic-Accurate Checkpoint Merging via Sequential Consistency
链接: https://arxiv.org/abs/2609.35168
作者: Shihao Wang,Rui Kong,Xinran Chen,Hui Wu,Qipeng Qian,Jinman Zhao,Jiashu Zhao,Yuchen Li,Jimmy Huang,Dawei Yin
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Saved checkpoints record states along a training trajectory, but generally do not determine the updates at states that would be visited under a different schedule. We study how accurately these checkpoints can reconstruct the endpoint of a sequential reference with prescribed update strengths. Under a common local transition model, two checkpoint-index moment conditions characterize all convex merges that agree with this reference through second order. We then prove an information limit that for nondegenerate profiles, no algorithm using only a fixed-length gradient-descent (GD) history with step size h can achieve o(h^3) endpoint error uniformly over a fixed class of smooth, strongly convex losses. The lower bound follows from two losses with identical GD checkpoint histories but sequential reference endpoints separated by \Omega(h^3) . \textbfQuadratic-Accurate Merging (QAM) achieves a matching uniform O(h^3) endpoint error bound. Its explicit coefficients also define the unique profile-dependent merge that exactly matches the sequential GD reference across all fixed quadratic objectives. Across two public Adam checkpoint trajectories (SmolLM3-3B and OpenEuroLLM-Prelude-9B), three windows and three profiles per model, and 15 tasks, QAM shows mixed results for short windows and broader advantages over \textbfWarmup-Stable and Merge (WSM) for longer windows. Matched-moment GSM8K diagnostics further show that local consistency alone does not fully determine downstream scores. These results characterize the reconstruction limits of saved histories, provide a coefficient rule that attains the optimal rate, and assess its practical utility.
[AI-77] Learning to Re-Draft: A Variational Stackelberg Game for Discrete Diffusion
链接: https://arxiv.org/abs/2609.35166
作者: Dmitrii Moor,Federico Tomasi,Paul N. Bennett,Alice Wang,Mounia Lalmas
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Discrete diffusion models offer the ability to re-draft, revisiting and correcting earlier tokens throughout generation. This capability depends on the forward corruption process that defines what the denoiser learns to correct. Masked diffusion models fix tokens once they are unmasked, while uniform diffusion permits revisions but relies on uniformly random token substitutions. We instead learn which substitutions are most useful for training the denoiser to re-draft. We introduce Variational Stackelberg Discrete Diffusion (VSDD), a framework for learning a semantically aware corruption process. VSDD formulates training as a leader-follower game: the leader defines a Markovian corruption process parameterized by the denoiser’s token embeddings, while the follower optimizes a variational denoising objective with the corruption process held fixed. The leader rewards corruptions based on how much the denoiser improves after learning from them, rather than on how easily the current denoiser can reconstruct them. We measure this improvement under a fixed reference corruption process, approximate the follower’s response with a one-step gradient update, and optimize the leader using a score-function estimator. We evaluate VSDD across molecular, text, and playlist generation. VSDD substantially improves molecular validity over uniform and masked diffusion, reduces text perplexity relative to uniform diffusion while remaining competitive with masked diffusion, and achieves sizable improvements in offline playlist recommendation metrics.
[AI-78] FONDANT: Strong and Best-Effort Planning via Antichains
链接: https://arxiv.org/abs/2609.35160
作者: Benjamin Aminof,Tuan Khai Nguyen,Sasha Rubin
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:A classical solution concept in fully observable nondeterministic (FOND) planning, is the strong policy (aka winning strategy in the closely related area of reactive synthesis), i.e., such a policy ensures that the goal is reached in an adversarial environment. When strong policies are not available or there is no evidence that the environment is adversarial, one can resort to best-effort policies, which always exist, and which follow the classic decision-theoretic principle that an agent should not use a dominated strategy. A typical positional best-effort policy works as follows: from every state, it follows a strong policy if one exists from that state (such states are called strong-winning''), else a weak policy if one exists from that state (weak-winning’‘), and else is unconstrained (``losing’'). In this work, we introduce a sound and complete planner for both best-effort planning and strong planning. The algorithm that underpins the planner is quite simple: it represents certain sets of states, such as the winning regions, by their \subseteq -minimal elements. The algorithm returns uniform policies, i.e., it returns a policy \pi_t that is a strong solution starting in every strong-winning state, and it returns a policy \pi_w that is a weak solution starting in every weak-winning state, and it provides a certificate for the set of losing states. We implemented the algorithm with some simple optimizations (calling it FONDANT), and evaluated it on a benchmark set consisting of the instances that were used in the evaluation of leading strong planners PR2 and FOND-SAT, and the best-effort planner BeSyftP. On coverage, our implementation is at least as good on all domains, and outperforms on some domains; and on wall time, it is slower on small and medium-sized instances, and outperforms on larger instances.
[AI-79] PEARL: Adaptive Prefill-Decode Execution with Elasticity for Agent ic Reinforcement Learning
链接: https://arxiv.org/abs/2609.35158
作者: Jiaan Zhu,Wei Gao,Youhui Bai,Zewen Jin,Ju Huang,Siran Yang,Jiamang Wang,Lin Qu,Cheng Li
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-turn rollout dominates the cost of agentic reinforcement learning (RL). Asynchronous execution and elastic GPU resources can accelerate this stage, but adding rollout replicas yields diminishing returns while training GPUs remain idle between updates. We observe that effective resource use also depends on the prefill–decode (PD) configuration. Both the choice between colocation and disaggregation and the optimal PD ratio vary with the workload, making resource scaling and PD configuration interdependent. Exploiting this opportunity requires selecting effective configurations and realizing their benefits within transient resource-availability windows despite reconfiguration costs. We present PEARL, an asynchronous agentic RL system that coordinates external resource elasticity, temporary reuse of idle training GPUs, and adaptive PD execution. PEARL maintains a unified GPU–worker–role state and uses runtime profiles to predict rollout batch completion time, accounting for environment-induced reductions in decode concurrency. It selects the PD mode and ratio under the current GPU budget and translates each decision into an incremental transition plan that minimizes worker and role changes. Cost-aware switching and borrowing policies suppress transitions with insufficient expected benefit while ensuring timely return of training GPUs. Our evaluation show that PEARL achieves 2.17 – 2.79\times the throughput of fixed-resource ROLL across different LLMs. Compared with RLBoost+, throughput improves by up to approximately 26.9% for Qwen3-8B and 36.3% for Qwen3-30B-A3B. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.35158 [cs.AI] (or arXiv:2609.35158v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.35158 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-80] CTP-FL: Common-Trajectory Gradient Prediction for Federated Learning
链接: https://arxiv.org/abs/2609.35130
作者: Junkang Liu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Communication-efficient federated optimization commonly spends several gradient evaluations between server updates. Existing local-update methods use this computation to advance an independent model on each client. Under heterogeneous data, however, these models evaluate gradients at different locations, making the aggregated update difficult to interpret as a gradient of the global objective. We study an alternative use of the same computation budget: \emphevaluate the global objective along a shared, predicted path. We propose Common-Trajectory Predictive Federated Learning (\textttCTP-FL). At each round, all clients construct the same sequence of query points from the current global model and the previous aggregated direction, evaluate K stochastic gradients along this sequence, and upload their average. The server then performs a single global update. Thus, \textttCTP-FL uses K mini-batch gradients per client and one model-sized vector in each communication direction, matching the per-round computation and communication of full-participation FedAvg-M. Shared query points make the aggregated direction an unbiased estimator of the average \emphglobal gradient along the predicted path. The remaining discrepancy from the gradient at the current model is controlled by the path length, without assuming bounded client-gradient dissimilarity or bounded gradients. For smooth non-convex objectives, we establish an \mathcalO!\left( \sqrtL\Delta\sigma^2/(NKR)+L\Delta/R \right) average-stationarity bound under full participation. The analysis isolates a testable trade-off: extending the prediction path provides more forward-looking gradient information but increases its displacement bias. Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.35130 [cs.LG] (or arXiv:2609.35130v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.35130 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Junkang Liu [view email] [v1] Mon, 28 Sep 2026 13:22:45 UTC (57 KB) Full-text links: Access Paper: View a PDF of the paper titled CTP-FL: Common-Trajectory Gradient Prediction for Federated Learning, by Junkang LiuView PDFHTML (experimental)TeX Source view license Current browse context: cs.LG prev | next new | recent | 2026-09 Change to browse by: cs cs.AI References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) IArxiv recommender toggle IArxiv Recommender (What is IArxiv?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab) Major funding support from
[AI-81] ool Mediation Alters Refusal Mechanisms in Large Language Models
链接: https://arxiv.org/abs/2609.35117
作者: Abel Rodríguez,Giuseppe Garofalo,Lieven Desmet,Vera Rimmer
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) are increasingly deployed with access to external tools, yet harmful tool-mediated interactions are less likely to be refused when compared to regular conversational ones. As this change in refusal behavior remains underexplored, we investigate its underlying mechanisms across a diverse set of open-weight language models. We find that information about the harmfulness of a request remains strongly encoded in the model’s representations and transfers across conversational and tool-mediated inputs. Evidence from representation geometry and neuron-level analysis further indicates that the two interaction modes systematically distribute harm-related computation differently. Crucially, while conversational inputs can be refused at relatively low levels of perceived harmfulness, tool-mediated inputs remain permissive until harmfulness crosses a substantially higher effective refusal threshold. Moreover, tool-mediated refusal is also more brittle: progressively weakening the refusal computation disrupts tool-mediated refusal at lower intervention strengths than conversational refusal, even when benign capabilities remain intact. Together, our findings indicate that tool mediation does not simply reduce the internal perception of harm, but instead impacts its conversion into refusal. Overall, this suggests tool-mediated environments may intrinsically reduce robustness of models to harmful requests, and that conventional safety evaluations may not fully transfer to LLM agents.
[AI-82] DuplexCadence: Exact State and Execution from a Speech Models Declared Timelines
链接: https://arxiv.org/abs/2609.35115
作者: Haixiao Gao,Yimin Zheng,Linyou Xiao,Zeke Xie
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Full-duplex speech models support streaming interaction that listens and speaks at the same time. Serving them is governed by a strict, repeating deadline: conversation advances on a one-second cadence, and every second of input must be turned into a second of speech before the next second arrives. Because stages within a session run in strict sequence, per-invocation overhead cannot be batched away. Profiling reveals that the autoregressive stages of a duplex second already fit within the period, whereas the token-to-audio synthesis tail is what causes overruns. This tail stage suffers from orchestration slack where the GPU is left waiting as thousands of tiny, regular operations are issued one by one, while also wasting substantial memory by over-provisioning state at static implementation constants. Existing remedies, such as graph recording and demand-sized allocation, fail because streaming state dynamics violate their prerequisites. The root cause is that the runtime lacks the model’s native clocks: the per-region counters that govern advancement rates and retention policies. We propose DuplexCadence, which explicitly declares native clocks to the runtime and derives two mutually enabling rules: demand-sized state allocation at a stable address, and exact-shape graph replay without padding. The former eliminates idle memory and stabilizes tensor pointers, while the latter removes orchestration slack without padding overhead. Evaluated on four released models across three decoder architectures with bit-for-bit identical output, DuplexCadence reaches 2.85\times the stock runtime’s speed at 38.8% lower peak memory. On the live duplex path, mean SPEAK time falls from 14% over the one-second cadence to 2% under it, enabling models to reliably keep up with interactive speech while markedly expanding multi-
[AI-83] Using Context Is Not Enough: Test-Time Training for Personalized Reward Modeling
链接: https://arxiv.org/abs/2609.35109
作者: Bohao Wang,Xiaoyan Zhao,Yang Zhang,Jinghang Guo,Chun Chen,Can Wang,Jiawei Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning from human feedback (RLHF) aligns large language models (LLMs) with human preferences, yet most pipelines learn a single reward model that overlooks individual differences in preferences. Personalized reward models (PRMs) address this by conditioning rewards on user-specific feedback, most commonly through in-context learning (ICL), where a user’s historical comparisons are supplied as contextual preference pairs. However, we identify a key limitation of ICL-based PRMs: they fail to capture the preference relations conveyed by contextual pairs. To address this, we propose Preference-Aligned Test-Time Training (P-TTT), which explicitly encodes these relations into user-specific fast weights for personalized reward prediction. P-TTT introduces sequence-level update and apply operations to match the response-level granularity of preference feedback, together with a preference-aligned objective that directly uses pairwise preference relations to guide fast-weight adaptation. Notably, P-TTT is simple to implement and computationally efficient, updating fast weights within a single forward pass without inference-time backpropagation. Extensive experiments show that P-TTT more effectively captures historical preference relations and outperforms state-of-the-art methods by a large margin.
[AI-84] DoAtlas-2: A Foundation for Self-Evolving Causal Biomedical Discovery
链接: https://arxiv.org/abs/2609.35107
作者: Yulong Li,Rong Xia,Yuxuan Zhang,Jianxu Chen,Xiwei Liu,Haochen Xue,Maosheng Li,Yuhang Liu,Yibo Yuan,Yutong Xie,Chong Li,Jionglong Su,Hagai Rossman,Eran Segal,Imran Razzak
类目: Artificial Intelligence (cs.AI); Quantitative Methods (q-bio.QM)
备注: Technical report. 185 pages, 5 figures. Yulong Li, Rong Xia and Yuxuan Zhang contributed equally. Corresponding authors: Eran Segal, Imran Razzak
Abstract:We introduce DoAtlas-2, a foundation for self-evolving causal biomedical discovery that organizes knowledge around causal mechanisms and advances through external evidence from human populations. DoAtlas-2 integrates 771 research resources covering more than 720,000 participants in 48 countries, from longitudinal clinical phenotypes, medical imaging, and continuous physiological signals to eight molecular layers, together with an evidence network of approximately 4.7 million literature-derived records over 93,566 concepts and 149,383 candidate causal relations. DoAtlas-2 autonomously formulates research questions from evidence gaps and unresolved mechanisms, prespecifies their causal designs, and generates validated analyses. Supporting, challenging, and unresolved results continuously revise mechanistic interpretations, the causal evidence state, and the discovery frontier, so that DoAtlas-2 self-evolves within a closed loop of hypothesis generation, empirical testing, and renewed discovery. DoAtlas-2 has systematically evaluated 2,031 research questions. In the Human Phenotype Project (HPP), it formulated 4,014 candidate pathway questions across vascular, early-glycemic, and hepatic-metabolic systems, and screening of the first 1,079 yielded statistical support for 756. Representative studies identify blood pressure as a convergence node linking adiposity, hepatic, and lipid phenotypes to vascular outcomes, and show that an adiposity-inflammation-blood-pressure pathway is largely attenuated by joint adjustment for body mass index (BMI) and smoking. The discovered vascular network constitutes a completely interpretable predictive foundation, admitting exact attribution of every prediction and closed-form mediation effects. DoAtlas-2 thereby unifies causal mechanism discovery, population-evidence testing, and interpretable prediction within one continuously evolving foundation.
[AI-85] SpikeLite: Lightweight Spiking Neural Networks for Time-Series Forecasting
链接: https://arxiv.org/abs/2609.35097
作者: Bang Hu,Changze Lv,Mingjie Li,Xiaoqing Zheng,Wei cao,Fan Zhang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Spiking neural networks (SNNs) offer an energy-efficient paradigm for time-series forecasting through spike-driven computation. However, recent SNN forecasters often pursue higher accuracy through increasingly complex attention mechanisms, or specialized neuronal dynamics, weakening the lightweight motivation of SNNs. We introduce SpikeLite, a spiking forecasting framework built around two modules: a Frequency-Selective Spiking Encoder (FSSE) for frequency-sensitive temporal encoding and a Sparse Spiking Channel Attention (SSCA) module for selective cross-channel interaction. FSSE exploits the low-pass filtering behavior of LIF dynamics to reorganize each input sequence into frequency-sensitive components while collectively preserving the input at the decomposition stage. SSCA then learns a binary mask from encoded channel representations and uses it to selectively exchange information within spike-driven self-attention, retaining informative cross-channel interactions while suppressing redundant ones. When explicit channel interaction is unnecessary, SpikeLite uses the lighter FSSE-only channel-independent path. Experiments under the SeqSNN and SpikF protocols cover four standard multivariate and eight long-term forecasting benchmarks. SpikeLite achieves the best aggregate performance under both protocols, with an average R^2 of 0.790 and RSE of 0.440, and lowest average MSE/MAE of 0.343/0.345 in long-term forecasting. Moreover, evaluation on the ECL dataset shows that SpikeLite achieves the lowest reported energy consumption, further demonstrating its potential for energy-efficient time-series forecasting.
[AI-86] Can Generative AI Automate Data Extraction for Meta-Analysis? A Case Study on Intercropping Research
链接: https://arxiv.org/abs/2609.35089
作者: Zehao Lu,Xingguo Xiong,Wopke van der Werf,Thijs L. van der Plas,Ioannis N. Athanasiadis
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Meta-analysis is the synthesis of information from multiple sources to arrive at an overarching conclusion. There is a large need for meta-analysis in agricultural research to synthesize what is known and analyze overarching patterns. Extracting data from published literature is, however, labor-intensive, time-consuming, and tedious, and is impeded by a lack of standardization in research design, units of measurement, and terminology. These challenges are particularly evident in the domain of crop species mixtures, also called intercropping. With the growing capabilities of LLMs, many recent attempts have focused on building systems and tools to automate data collection, yet rigorous assessment against human-labeled ground truth is often missing. In this research, we evaluate three LLM-based approaches—direct zero-shot prompting, a staged workflow, and a multi-agent system—with six open-weight models to extract data from the intercropping literature. The results are evaluated against the manually curated ground truth and through a downstream statistical analysis. Overall, direct zero-shot prompting is the strongest and most consistent approach, achieving the highest mean similarity-adjusted F1 of 0.577, although none of the approaches is close to fully accurate. In the downstream analysis, most model–approach combinations recover the direction of the relationship between the predictor and outcome variables, but do not estimate its magnitude accurately.
[AI-87] When Valid Tool Calls Change Meaning: Formation-Consistent Dispatch for LLM Agents
链接: https://arxiv.org/abs/2609.35088
作者: Geonwoo Kim(1),Brent ByungHoon Kang(1) ((1) Korea Advanced Institute of Science and Technology (KAIST))
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Software Engineering (cs.SE)
备注: 17 pages, 5 figures, and 9 tables
Abstract:Tool-enabled agents form calls from model-visible interfaces, while hosts later select their implementation. Standard dispatch omits the descriptor-handler relation. An unchanged and schema-valid call can therefore acquire a different security effect during rollout, reconnect, or delayed approval. We call this failure schema-epoch drift. We present formation-consistent dispatch (FCD), which connects implementation analysis to execution authority. Reviewed profiles produce provenance-bound over-approximations of declared in-scope effects from official source. Under a closed-target approval policy, a verifier applies each formed call to a summary and captures a successor only when its effects fit the call’s security contract. Atomic admission and a final-hop fence preserve this decision to the effect. The exact source retains priority, and the captured successor becomes eligible only after source retirement. Stock releases and deployment changes reproduced the failure. Four profiles covered 32 official releases: 29 required no release-specific change and three escalated. A frozen 16-release expansion matched a separate source oracle. In a preregistered stock comparison, FCD completed all three pending calls whose effect remained private and blocked all three whose omission became public. Exact pinning and release-wide denial stopped all six calls, while release-wide approval completed all six but produced three public effects. A separate lifecycle experiment carried a formation-captured certificate across source retirement. The same safe certificate installed later governed new formations without expanding the pending call’s authority.
[AI-88] What Drives Citations in Production Large Language Models ? An Observational Multi-Method Study of Two Million AI Citations Across Ten Thousand Web Pages
链接: https://arxiv.org/abs/2609.35077
作者: Ben Moore,Liam Dunne
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Production large language models retrieve and cite web pages alongside generated answers, yet the page-level features that predict citation frequency remain poorly characterised. We present an observational study of approximately 2 million LLM citations from four commercial engines (ChatGPT, Claude, Google AI, Gemini) over six months, joined to 10,000 crawled pages from nineteen B2B SaaS workspaces. Sixty-plus features are tested using a nine-method consensus framework combining mixed-effects regression with domain fixed effects, FDR correction, stability-selection Lasso, double machine learning, generalised additive models, and temporal hold-out replication. Four findings survive all checks. First, prompt-content alignment (Jaccard overlap between page tokens and the full workspace prompt corpus, including non-citing prompts) is the dominant page-level predictor (beta = +0.37, 95% CI [+0.33, +0.41], q ~ 10^-73). Second, the standard AEO checklist (FAQ blocks, structured data, Core Web Vitals) shows positive effects in pooled data that reverse or collapse to zero once domain fixed effects are applied: Simpson’s paradox with practical consequences for the AEO literature. Third, domain-level AI authority exceeds the strongest non-alignment page-level feature by a factor of six in mean absolute SHAP value. We release the analytic pipeline as a methodological contribution.
[AI-89] mpoKV: Timely Staging of LLM KV Caches for Memory-Semantic Flash
链接: https://arxiv.org/abs/2609.35065
作者: Jay H. Park,Hyungjun Kim,Dong Kim
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Reusable prefix key-value (KV) caches can outgrow GPU memory in large language model (LLM) serving. A memory-semantic flash hierarchy offers SSD-backed capacity with a limited fast tier, but a logical KV hit is not necessarily ready for GPU retrieval. Demand staging exposes SSD latency, whereas immediate staging can reserve fast-tier capacity long before retrieval begins. We present TempoKV, a timing-aware resource-commitment layer that separates early knowledge of reuse from the acquisition of staging resources. It records reusable-KV hits as metadata-only claims and requests commitment when the runtime-estimated time until retrieval falls to the storage-estimated time needed to make KV resident and protected against eviction. These estimates adapt to runtime progress and staging state, while commitment remains subject to available protected capacity. We implement TempoKV in vLLM and LMCache on an SSD-backed CXL memory device without changing request scheduling. Across two models and three prefix cache ratios, TempoKV reduces protected fast-tier byte-time per request by 63-91% versus immediate staging while retaining much of the serving benefit of advance staging. In a fast-tier capacity sweep, output throughput and p95 time to first token (TTFT) remain nearly unchanged as capacity decreases from 100 to 25 GiB. Compared with unmodified LMCache’s Device-DAX L1 configuration, TempoKV reduces p95 TTFT by up to 48.0% and increases output throughput by up to 27.8%.
[AI-90] IDE: Teacher-Student Transition via Informative Distillation and Exploration for Agent ic RL
链接: https://arxiv.org/abs/2609.35058
作者: Yibin Huang,Xinming Xu,Conghui Zhu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Effective multi-turn agents require interaction strategies that coordinate information gathering, actions, and feedback over long horizons. GRPO is a reinforcement learning algorithm used to train these agents, but sparse trajectory-level rewards limit early exploration in small models. Recent methods augment RL with on-policy distillation (OPD) from a stronger teacher. However, a fixed mixture assumes that teacher guidance and reward optimization should retain a constant relative role throughout training and across interaction turns. This assumption can fail at two scales. Globally, as training progresses, maintaining strong distillation pressure can constrain the model from moving beyond the teacher’s capabilities. Locally, teacher–student disagreement identifies where the student departs from the teacher, but cannot tell whether that departure is exploration supported by better outcomes or low-quality policy drift. Our methodological insight is that teacher guidance and reward optimization should be dynamically rebalanced over training and jointly allocated across turns. We instantiate this insight in \tide. Globally, \tide uses the measured disagreement trend as a practical schedule signal, advancing an OPD-to-RL handoff when discrepancy reduction becomes slow but remains positive and progressively increasing the relative weight of RL. Locally, \tide jointly modulates teacher-guided and reward-driven updates: relative action value and disagreement prioritize the OPD signal, whereas relative action value supplies the RL advantage and normalized disagreement reweights it across turns. Coupled with the global handoff, \tide allocates stronger teacher guidance early and gives reward-driven updates greater relative weight later in training. Experiments across multiple benchmarks, student scales, and controlled ablations support the effectiveness of TIDE’s adaptive OPD–RL coordination.
[AI-91] EMPIRIC: Experiment-Driven Learning of Residual World Models for Robot Planning
链接: https://arxiv.org/abs/2609.35047
作者: Yichao Liang,Amber Li,Dat Nguyen,Emily Bunnapradist,Michelangelo Naim,Sreela Kodali,Matteo Merler,Bowen Li,Kiran Gopinathan,Yiyun Liu,Nikhil Pimpalkhare,Joshua B. Tenenbaum,Adrian Weller,Zenna Tavares,Tom Silver,Kevin Ellis
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: The last two authors contributed equally as co-advisors. Website and code: this https URL
Abstract:A robot should be able to learn through experiments how unfamiliar objects behave and interact, then plan with that knowledge. It need not start from scratch: physics engines supply knowledge of motion and contact, but can omit entire mechanisms, such as glue curing, water heating, or wind. We present EMPIRIC, an agent that learns a residual world model: a physics engine extended with code for the missing mechanisms. The learned programs can introduce new forces, constraints, and hidden state, and Bayesian inference estimates their parameters and states from noisy observations. The resulting model lets the agent predict the outcomes of actions, choose informative experiments, and revise its hypotheses when predictions fail. Across five simulated domains, EMPIRIC learns interpretable, reusable models, and solves more tasks with fewer environment interactions than all three baselines. On a physical robot, it learns wind forces and domino masses to solve a manipulation task. Website and code: this https URL
[AI-92] BA-DPO: Bias-Adjusted Direct Preference Optimization for Language Model Alignment
链接: https://arxiv.org/abs/2609.35044
作者: Antonio Ferrara,Alberto Rumi,Francesco Bonchi
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Preference-based alignment methods such as Direct Preference Optimization (DPO) use pairwise preferences labeled by human annotators to fine-tune language models. However, annotators carry systematic biases toward some attributes: a name that signals a gender or an ethnicity, a persona, a language variety, a formatting convention, or length. If not properly addressed, these systematic biases can be absorbed and amplified during alignment. Existing methods address length bias or annotator disagreement, but fail to eliminate biases toward arbitrary attributes. To address this limitation, we propose Bias-Adjusted DPO (BA-DPO), a generalization of DPO that adds one bias parameter per annotator toward responses carrying a declared attribute. We prove that the objective is convex in the bias parameters and that the votes identify each annotator’s bias up to a shared constant. The remaining constant is what fixes the aligned model’s attribute rate: by default the rate of the reference model, or a target rate, which we use to bring a biased policy to statistical parity. On a corpus with planted biases, DPO drives the attribute from a balanced start to probability 0.96 and BA-DPO removes 81 to 95% of that shift; on MultiPref with real annotators it removes about half of DPO’s lengthening. Both hold at 0.5B with full fine-tuning and at 8B with LoRA, at no higher KL than DPO and no loss in judged quality.
[AI-93] Persona Following Is Not Selective Control: The Neutrality Gap in LLM User Simulation
链接: https://arxiv.org/abs/2609.35036
作者: Jiashen Ren,Wenlin Zhang,Bohan Zhang,Xiaopeng Li,Zichuan Fu,Wanyu Wang,Junyi Li,Xiangyu Zhao
类目: Artificial Intelligence (cs.AI)
备注: 60 pages, 8 figures
Abstract:Persona prompting is widely used to construct user simulations with large language models (LLMs), yet it relies on a largely untested assumption: specifying one user attribute should change that attribute alone. We test this assumption and identify a systematic failure of selective control: across all eight black-box LLMs we audit, changing a target attribute also shifts responses on unspecified, non-target attributes. For example, describing a user as more risk-seeking shifts color choices, even though the prompt never mentions color; we term this cross-attribute influence. Semantic, contextual, and internal analyses collectively suggest that models treat a persona prompt as evidence about the user and extend the inferred profile to unspecified preferences, a process we call trait-conditioned completion. We next ask whether explicitly specifying non-target attributes restores selective control. When a non-target attribute is assigned a clear direction, models generally follow the declaration and suppress the target attribute’s influence. However, when the same attribute is declared neutral, the target continues to affect choices across all five open-weight checkpoints, even when the model correctly reports the declared state. This disparity, the neutrality gap, demonstrates that successful persona following does not imply selective persona control, which additionally requires keeping non-target attributes stable. We operationalize this distinction with a three-state diagnostic that leaves the non-target attribute unspecified or declares it directional or neutral; because directional tests can be passed by simply following the stated persona, the neutral state reveals failures they miss. In a post hoc analysis of independent items, neutral declarations leave 51-81% of items target-sensitive, against at most 1 of 320 item-pole comparisons under directional ones.
[AI-94] AutoDataBench: Can Agents Write the Data That Feeds the Self-Improvement Loop?
链接: https://arxiv.org/abs/2609.35025
作者: Haotian Luo,Haoyu Wang,Zeyu Qin,Huanjin Yao,Yibo Wang,Zhuotao Tian,Shuai Wang,Jiaya Jia
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Recent gains in language model capability have come more from data than from architecture. Frontier labs and data companies produce verifiable agentic tasks, which supervised finetuning and reinforcement learning then turn into this http URL production line still rests on human labour and on human-in-the-loop collaboration. Automating task creation would let data production scale with compute rather than with expert headcount, would extend to more domains, and would enable a key step in recursive self-improvement (RSI). Current evaluations of an agent’s ability to write such tasks measure how a model performs after training on what the agent produced. That does not match common practice in the data industry, where data is delivered sample by sample and each sample is accepted against a set of criteria rather than put straight into training. No existing evaluation asks whether an individual task meets the acceptance criteria of a data pipeline. We therefore introduce AutoDataBench. Given an original benchmark task and a record of the target model attempting it, an agent must write a new task for the same suite that meets practical acceptance standards on validity, novelty, difficulty and behavioural coverage. Across three benchmarks of executable agent tasks, no agent we evaluate scores above 20 out of 100 at the default time budget of 45 minutes. Giving the strongest agent four times as long improves its score substantially, while the cost of one usable task stays almost unchanged. Current agents can write training tasks of the required quality, but not efficiently. AutoDataBench provides a direct measure of an agent’s capacity for autonomous data synthesis: one artifact at a time, judged against the criteria a production pipeline would apply, and without a training run. Code and data are available at this https URL.
[AI-95] Addressing Spatial Indistinguishability in Spatiotemporal Prediction via Optimal Transport-Guided Masking
链接: https://arxiv.org/abs/2609.35021
作者: Guangyu Wang,Jiawei Tong
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted by Pattern Recognition
Abstract:Spatiotemporal prediction aims to learn discriminative representations from correlated temporal signals over spatial structures for accurate future inference. A central challenge is \emphspatial indistinguishability: different nodes may share similar historical patterns yet evolve toward divergent futures, severely degrading forecasting performance in real-world sensor networks. Existing embedding-based and graph neural network (GNN)-based approaches can partially detect such ambiguous nodes but rely on historical similarity, struggling to capture \emphfuture behavioral divergence. We propose \textbfSTOT (\textbfSpatio\textbfTemporal \textbfOptimal \textbfTransport), a self-supervised framework that resolves spatiotemporal ambiguity via structured masking guided by optimal transport. Our key idea treats indistinguishability as a \emphdisambiguation problem: future states are inferred by exploiting concurrent spatial correlations and their time-varying similarity. We design a similarity-aware metric for dynamic inter-node relationships and an optimal transport-based masking strategy to emphasize ambiguous positions during pre-training. A batch consistency constraint preserves semantic coherence, while a random-walk masking mechanism promotes structured context exploration. Experiments on six real-world datasets show that STOT performs competitively with state-of-the-art baselines on the evaluated benchmarks and improved interpretability through transport-plan visualizations.
[AI-96] Environmental requirements for the use of social information by artificial life agents using evolved plastic artificial neural networks
链接: https://arxiv.org/abs/2609.35018
作者: Hugh Charterton,James M. Borg,Aniko Ekart
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Evolved Plastic Artificial Neural Networks (EPANNs) consist of two principal processes, the first, evolution, and the second, development and in-life learning. In the context of the origins of social = learning, very few studies have been carried out using ALIFE models based on EPANN requirements. Studies in this field have usually involved an imitative teacher/pupil relationship. This, however, ignores the possibility that the observed behaviour is a consequence of social information cues rather than direct imitation or teaching. Starting with the first of the EPANN processes (evolution), a series of experiments was undertaken using artificial neural network (ANN) based agents in a variety of foraging environments to examine under what minimal environmental conditions the use of social information might have evolved, as measured by the number of generations taken to meet a specified fitness criterion. NEAT (Neuroevolution of Augmenting Topologies) was the ANN used as its evolutionary algorithm would evolve a network’s topology as well its weights. Unintentionally, in the experiment there was a simple network topology based on the location of the nearest food item which enabled agents to swiftly meet the fitness criterion. With this topology, additional information, social or otherwise, was not required and could have proved to be a hindrance. However, this does indicate that for the use of social information to have evolved, it would require a greater degree of complexity in the environment to do so.
[AI-97] rmJudge: A Document-Level Metric Judging Not Counting Terminology in Machine Translation Evaluation
链接: https://arxiv.org/abs/2609.35017
作者: Nicolas Dahan(ISIR, ALMAnaCH),Fran{\cc}ois Yvon(MLIA, ISIR),Rachel Bawden(ALMAnaCH)
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Existing automatic metrics for evaluating terminological use in machine translation (MT) penalise any divergence from a fixed reference, conflating translation errors with the valid terminological variation that human translators routinely produce. We introduce TermJudge, a document-level terminology metric that assigns an interpretable verdict to every term occurrence: glossary-conforming occurrences are settled deterministically, while divergences are assessed under a two-step LLM-as-judge procedure using the full document context: the first detects and labels terminology errors; the second sorts valid document-level variations from inconsistencies. Validated against expert error annotations and document-level human MQM scores, TermJudge ranks first in both system- and segment-level meta-evaluation, ahead of glossary-conformity and quality-estimation baselines. When applied to eight systems translating academic documents, under two prompting conditions, we observe that glossary injection improves terminology translation in all paired comparisons, by removing genuine errors rather than valid variation. TermJudge is released as open-source code.
[AI-98] Automated feature engineering AutoML and decision-focused learning for improved energy consumption forecasting
链接: https://arxiv.org/abs/2609.35013
作者: Nasser Alkhulaifi
类目: Artificial Intelligence (cs.AI)
备注: PhD thesis, School of Computer Science, University of Nottingha, United Kingdom
Abstract:The rising cost and demand for energy, together with environmental sustainability goals, create major challenges for energy management. Energy Consumption Forecasting (ECF) supports planning by predicting future consumption, but Machine Learning (ML) models for ECF often depend on expert-driven Feature Engineering (FE). This thesis addresses that dependence through three contributions. First, it establishes and evaluates a comprehensive FE pipeline for ECF and investigates domain-specific features. Second, it introduces AutoEnergy, a domain-tailored automated FE algorithm that generates interpretable features from timestamps and lagged consumption and integrates with AutoML for end-to-end ECF modelling. Across eighteen real-world energy datasets spanning residential, commercial, industrial, renewable, and grid domains, AutoEnergy reduces forecasting error by 19.52%-84.72% relative to baseline AutoML and established automated FE methods, while running 1.31-4.41 times faster, with gains varying by dataset. Third, AutoEnergy is integrated with Decision-Focused Learning (DFL) for a Battery Energy Storage System problem, jointly forecasting electricity prices and demand while optimising charging and discharging decisions. On a real-world UK property dataset, this approach reduces operating costs by 22.9%-56.5% compared with the same DFL models without automated FE. Overall, the results show that domain-specific automated FE can reduce reliance on manual feature design, improve forecasting accuracy, and translate predictive gains into measurable operational benefits in energy management.
[AI-99] Learning to Act under Visual Interruptions with Vision-Language-Action Models
链接: https://arxiv.org/abs/2609.35003
作者: Mingle Jiang,Rui Xu,Yunke Wang,Chang Xu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: this https URL
Abstract:Vision-language-action (VLA) models have demonstrated strong capabilities in robotic manipulation, but they are typically developed and evaluated with all camera streams available throughout task execution. When a camera stops delivering frames during task execution, the policy must continue acting without access to subsequent observations from the missing view. Despite its practical importance, how such interruptions affect closed-loop manipulation remains insufficiently understood. To investigate this problem, we introduce MAIL-Bench, a benchmark that evaluates visual interruptions with VLA models. By interrupting different cameras at multiple stages of each policy’s successful reference trajectory, MAIL-Bench measures how well policies retain their capabilities when visual inputs become unavailable. Building on this benchmark, we propose MINT, which first trains VLA policies to remain functional under missing visual inputs. At inference time, MINT selectively supplements missing observations using optical-flow extrapolation or an action-conditioned world model, and withdraws predicted views when they become unreliable. Experiments on \pi_0.5 and GR00T N1.5 show that MINT significantly improves task success under camera loss over the original models. Experiments on AgiBot G2 further demonstrate the real-robot deployment under camera loss. The benchmark is available at this https URL
[AI-100] From One-Shot Generation to Incremental Music Composition: Adapting a General-Purpose Instruction LLM for Persistent Symbolic Editing
链接: https://arxiv.org/abs/2609.34994
作者: André Ricardo Ducca Fernandes,Jean-Pierre Briot,Simone Diniz Junqueira Barbosa1,Hélio Côrtes Vieira Lopes
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Most music-generation systems are still framed and evaluated primarily as producers of complete outputs, whereas composition often proceeds through successive revisions to a shared musical artifact. This paper studies a different use of a general-purpose instruction-following large language model: not as a one-shot music generator, but as a reusable operator over an evolving symbolic score. We formulate incremental composition as a sequence of operation-aware state transitions over persistent ABC notation, with explicit requirements on what each operation may change and what it must preserve. The interaction includes two artifact-initialization variants and three editing operations – chord addition, inpainting, and transposition. We instantiate the formulation by adapting Llama 3.1 8B Instruct with Low-Rank Adaptation (LoRA) on 496,038 operation-aware dialogue records derived from Irish traditional music. The comparison with the unadapted model is used to test the feasibility of learning this interaction contract, not to claim novelty for fine-tuning itself. Across 500 dialogues per model (1,750 attempted output states), checker admission rises from 29.37% to 99.37%, while compliance conditional on admission rises from 0.7205 to 0.9798. Strict eligibility for reference-relative musical-feature analysis increases from 14 to 1,548 outputs, and Longest Common Subsequence analysis does not show a systematic increase in high-overlap sequences relative to held-out baselines under the specified protocol. The results support the technical feasibility of persistent, operation-aware symbolic editing with a general-purpose instruction LLM. They do not establish superior musical quality or human-AI co-creativity, which remain questions for musician-centered evaluation.
[AI-101] Composable Decoding on the Probability Simplex: Theory and Implementation
链接: https://arxiv.org/abs/2609.34992
作者: Xiaotong Ji,Ahmed Khaled Khamis,Rasul Tutunov,Matthieu Zimmer,Haitham Bou-Ammar
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Decoding for large language models is typically treated as a collection of isolated sampling strategies, with limited theoretical understanding of the behaviours they induce and how their underlying objectives relate. We formulate decoding as an optimisation problem over next-token distributions on the probability simplex, balancing expected model score against regularisation under support constraints. This view recovers familiar decoding methods through choices of regularisers and support constraints; more importantly, it enables new decoders to be constructed by composing distributional preferences within a single optimisation problem without external rewards, learned critics, or model parameter updates. We introduce CompoSimplex, a library with configurable support rules, regularisation primitives, and simplex solvers for constructing and evaluating compositional decoders. We evaluate standard samplers, individual regularisers, and compositions across multiple models and reasoning tasks. Our results show that compositions can realise trade-offs between single-sample quality, multi-sample quality, and diversity that are not attained by individual decoding objectives.
[AI-102] Before Acting Change the State: Prospective State Intervention for Web Agents under Deceptive Interfaces
链接: https://arxiv.org/abs/2609.34974
作者: Ruozhao Yang,Mingfei Cheng,Xiaofei Xie
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:LLM-based Web agents can autonomously complete user tasks, yet deceptive interfaces can steer them toward outcomes that conflict with users’ interests. Existing defenses primarily intervene on agent behavior through blocking, guidance, or replanning. We identify a distinct failure mode: a task-valid action can still realize an unauthorized consequence because of the current Web state. This motivates treating task-relevant Web state itself as a runtime control target. We introduce Veer, an agent-side runtime defense that leaves task planning to the base agent and intervenes on Web state when a proposed action would produce an unauthorized consequence. Before modifying the live environment, Veer constructs a prospective intervention trajectory toward a safe task-relevant state and executes it with runtime grounding and verification. Across TrickyArena and WebDecept, Veer achieves the highest safe task completion in all three evaluation settings, exceeding the next-best defense by 15.9 and 25.0 percentage points on TrickyArena-Single and TrickyArena-Multi, respectively, while reducing dark-pattern success on WebDecept to 0.3%. These gains persist across dark-pattern types and all 12 agent, model, and benchmark configurations. Ablations show that active state intervention provides the largest gain, while prospective rollout and temporal evidence contribute additional improvements. These results establish task-relevant Web state as an effective runtime control target for protecting Web agents from deceptive outcomes.
[AI-103] APEX-Voice: Can Voice Agents Complete Professional Workflows Through Full-Duplex Interaction ICLR2027
链接: https://arxiv.org/abs/2609.34973
作者: Puneet Mathur,Dinesh Manocha
类目: Artificial Intelligence (cs.AI)
备注: Under Submission to ICLR 2027
Abstract:Full-duplex voice agents can now listen, speak, use tools, and act during spoken interactions, but fluent dialogue does not guarantee correct completion of delegated professional workflows. We introduce APEX-Voice, a benchmark of 120 interactive professional workflows spanning ten work archetypes such as form completion, corporate negotiation, coordination, consulting, and interviewing. Each workflow executes in a stateful Voice Workbench environment with task-specific knowledge, typed tools, gold-annotated final work artifact, authorization constraints, and a user simulation policy backed by validated, pre-compiled speech realizations. We evaluate both artifact field accuracy and end-to-end workflow success, which requires the correct terminal state, valid process, completed actions, and a valid final artifact. Across five frontier real-time voice agents-GPT-Live-1, Gemini-3.8-Live, Grok-Voice-Think-2.0, Step-Audio3, and GPT-realtime-2.1, none exceeds 25% Pass@1, and the best Reliable@3 is only 10.8%. Moreover, stateful coordination is the dominant failure point across systems, while success decreases further on workflows requiring greater knowledge retrieval and mid-speech corrections. Overall, APEX-Voice is the first benchmark for evaluating whether voice agents can translate conversational competence into dependable professional work.
[AI-104] Action-Space Shaping for LLM Agents : Measuring and Mitigating Tool-Schema Bias
链接: https://arxiv.org/abs/2609.34971
作者: Yinhong Liu,Zhili Tan,Zilin Wang,Zhijiang Guo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models (LLMs) have shown strong performance on tool-use agentic tasks when given a fixed tool schema. Yet a tool schema is not the action space of an agent; it is merely one interface representation of it. The same executable action can be exposed through many different, functionally equivalent tool definitions, and an agent that has truly learned a task should behave consistently across them. We show that current agents often do not, a phenomenon we term schema bias. To study this systematically, we introduce an executable transformation framework that rewrites a native tool schema using nine operators, including merging and splitting tools, altering how a single tool is expressed, and distributing one action across several dependent calls. The tasks, executable actions, and reachable states remain fixed, so any change in success is attributable to the interface alone. Evaluating eleven LLMs, including two closed models, on up to 32 schema variants, we ask how large schema bias is, how it manifests, whether the difficulty of a schema variant can be predicted without a full evaluation, and whether training removes it. We find that schema bias is substantial even for the newest models: success rates range from complete failure to 97% depending solely on the schema. To reliably estimate schema difficulty, it requires running a small sample of the target queries. Training repairs a schema variant only when that variant appears in the training data.
[AI-105] RoboFL: Federated Expert Assembly for World Action Models
链接: https://arxiv.org/abs/2609.34968
作者: Rongyu Zhang,Ruizhi Fan,Yunfan Lou,Hengyu Fang,Shenli Zheng,Chenrui Wu,Yili Jin,Li Du,Dan Wang,Yuan Du,Shanghang Zhang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision-language-action and world-action models are increasingly popular, yet remain bottlenecked by physical interaction data that is scarce, institutionally siloed, and task-heterogeneous. A natural federated solution is to let each client adapt a shared foundation model through parameter-efficient fine-tuning, avoiding the exchange of full-model updates. However, federating these adapters is nontrivial, as naive aggregation can entangle incompatible updates, while incorporating MoE-style routing into federated aggregation may dilute specialization and destabilize expert selection. We present RoboFL, which instantiates MoSAIC (Mixture of Slotted Adapters) for federated world-action learning. MoSAIC directly installs locally trained LoRA adapters as the expert branches of a server MoE. Server-side routers learn token assignments over these prior-informed branches while jointly refining routing and expert parameters. Foresight-to-Action Routing Distillation (FARD) aligns routing across the model’s three paths, while Path-Consensus Expert Aggregation (PCEA) converts complete expert updates into a compact global adapter for personalized redistribution. Experiments on RoboTwin 2.0, RLBench, and a real-world Franka robot arm show the superiority of RoboFL with structured expert assembly, as it outperforms centralized PEFT InternVLA-A1 by 12.23% on the Franka arm, while reducing per-round client communication by up to 86.81% relative to MoE-based federated VLA baselines.
[AI-106] Safe Greenhouse Climate Control Using Lagrangian-Constrained PPO with Kolmogorov-Arnold Networks
链接: https://arxiv.org/abs/2609.34966
作者: Hangzun Liu,Yuling Fan,Fang Tian,Zhilong Bie,Zaiwen Feng,Yongliang Qiao
类目: Artificial Intelligence (cs.AI)
备注: 12 pages, 5 figures
Abstract:Greenhouse climate control balances economic return with maintaining temperature, humidity and CO2 within crop-adapted growth ranges. Conventional reinforcement learning (RL) greenhouse controllers use fixed reward penalties to limit climate constraint violations, yet such heuristic penalties cannot explicitly constrain long-term cumulative violations. Poorly tuned weights either lead to overly conservative policies and lower yields, or fail to suppress persistent climate deviations that harm photosynthesis and induce crop diseases. To address this issue, we formulate greenhouse climate regulation as a Constrained Markov Decision Process (CMDP) and use a Lagrangian safe RL framework RCPO-PPO to separate economic optimization and cumulative safety constraints, enabling adaptive penalty adjustment without manual tuning. To handle strong nonlinear, time-varying coupling between greenhouse microclimate and crop growth, Kolmogorov-Arnold Networks (KANs) replace Multi-Layer Perceptrons (MLPs) as policy and value approximators for improved nonlinear representation. Sinusoidal cyclic time features are embedded in observations to capture diurnal environmental periodicity. Simulations use a classic winter lettuce greenhouse model driven by 40-day real weather disturbances. Compared with vanilla penalty-based PPO, our method cuts cumulative climate violations by 18.65% and raises lettuce economic profit by 2.91%, keeping violations stable near the safety threshold. This decoupled CMDP optimization with KAN-based policy representation mitigates long-term climate risks and boosts planting profits, offering a constraint-aware control strategy for precision greenhouse cultivation.
[AI-107] Cyclostationary Phase Conditioning for Medical Time Series Diffusion
链接: https://arxiv.org/abs/2609.34965
作者: Samuel Ruiperez-Campillo,Michele Copetti,Jorge da Silva Goncalves,Sonia Laguna,Julia E. Vogt
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Signal Processing (eess.SP)
备注: 43 pages, 16 figures, 21 tables
Abstract:Many physiological time series, such as cardiac and brain recordings, exhibit cyclostationarity: their statistics vary periodically with an underlying cycle phase. Corruption from motion, poor contact, and physiological interference obscures morphology needed for diagnosis, making signal restoration essential. Existing diffusion approaches condition on corrupted observations alone and must learn cyclic structure implicitly. We instead propose two inductive biases which encode cyclostationarity: a shift-covariant wavelet representation and dense per-sample phase conditioning inferred from the corrupted input. We further introduce a training-free cyclostationarity index that quantifies phase structure and predicts when phase conditioning will help. Finally, we propose antithetic coupling of reverse trajectories to reduce sampling variance while achieving comparable performance with fivefold fewer network evaluations. Across modalities, our results show that explicitly encoding measurable cyclic structure improves physiological time-series restoration.
[AI-108] JevVibe: Efficient Classification-Guided Secure Code Generation
链接: https://arxiv.org/abs/2609.34963
作者: Arshak Rezvani,Sasha Behrouzi,Ahmad-Reza Sadeghi
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models can generate functionally correct code that still contains security weaknesses, motivating repair pipelines that first diagnose a weakness type before deciding how to fix it. The Common Weakness Enumeration (CWE) provides a standardized vocabulary for such diagnoses, but asking an autoregressive language model to generate a CWE label and extracting it from the response raises questions about output validity, speed, and cost, as well as accuracy. We evaluate Jev, a decision model that instead selects directly from a declared set of candidates and returns a probability for each, against six open-weight autoregressive models and a frontier proprietary model, GPT-5.6-Sol, on a controlled 50-way CWE classification task over 1,916 CyberSecEval benchmark examples. Jev outperforms all six open-weight baselines on every classification and ranking metric, while its comparison with GPT-5.6-Sol depends on the metric: GPT-5.6-Sol achieves higher Top-1 accuracy and Macro-F1, whereas Jev achieves higher Top-3 and Top-5 accuracy and a nearly identical MRR, at 6.27\times lower median API latency and 55.9\times lower estimated API cost. We further build JevVibe, a diagnosis-guided repair agent that uses predicted CWE labels to repair code generated by Qwen2.5-Coder-32B-Instruct. With Jev providing the diagnosis, the agent increases the detector-measured security pass rate from 63.5% before repair to 70.7%, compared with 66.1% for LLM-guided repair. These results show that JevVibe is effective at improving the security of generated code, with Jev providing reliable and efficient CWE classification.
[AI-109] ProofLoom: Proof-Obligation-Driven Theory Construction for Autoformalizing Research-Level Stochastic Optimization
链接: https://arxiv.org/abs/2609.34960
作者: Feiming Wang,Daibo Li,Kun Yuan
类目: Artificial Intelligence (cs.AI); Optimization and Control (math.OC)
备注: 38 pages, 5 figures. Code and supplementary materials: this https URL
Abstract:Formalizing research-level stochastic optimization in Lean requires both an algorithm model and domain theory connecting foundational libraries to convergence proofs. Revising a model to restore provability can change the mathematical claim. We introduce ProofLoom, a fully automated LLM-agent system for Proof-Obligation-Driven Theory Construction. Given a published algorithm, target theorem, and source proof, ProofLoom autonomously constructs the Lean model and supporting theory. Open proof obligations drive the development of definitions, interfaces, lemmas, and proof plans. Signature contracts record evidence and obligations for model revisions; an independent Judge rejects unsupported assumptions and weakened conclusions. Planner expands the published argument into intermediate claims, and Audit checks whether the Lean proof follows it. Across tasks, SOptLib accumulates verified mathematics and construction experience: reusable results are extracted, generalized, and verified, while modeling decisions and failed proof routes are recorded. Later tasks retrieve these results and records and contribute new developments, forming a cycle of construction, accumulation, and reuse. On fifteen textbook and research-paper tasks, ProofLoom obtains mean human ratings of 6.3/7 and 6.4/7, compared with 4.9/7 and 5.0/7 for the strongest of six baselines. Across 33 developments, it produces 490,693 lines of algorithm-local Lean code with no sorry. The formalizations also expose 28 incorrect formulas, proof gaps, and algorithm-analysis mismatches in published sources across 22 developments, each with checked evidence.
[AI-110] VD-DeepStack: Bridging Visual Comparison and Language Reasoning for Few-Shot Anomaly Detection
链接: https://arxiv.org/abs/2609.34949
作者: Mengyang Zhao,Zhuolin He,Haiyang Yu,Yuxuan Liang,Yifang Xu,Yuchuan Wu,Xiaolei Chen,Zhengtao Yao,Fan Shi,Yang Liu,Bin Li,Xiangyang Xue
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Few-shot visual anomaly detection is fundamentally a visual comparison task, requiring fine-grained inspection of a query against normal references. Many recent methods based on large vision-language models (LVLMs) emphasize comparative reasoning through language chain-of-thought. Yet discrete, abstract descriptions may underrepresent dense, fine-grained visual differences, leaving a gap between visual comparison and its expression in language. To address this gap, we propose Visual Difference DeepStack (VD-DeepStack), which explicitly conditions language reasoning on query-reference visual differences. Specifically, we fuse DINO features with the LVLM visual hierarchy to strengthen fine-grained representations, then construct dense difference evidence from residuals between query features and softly matched reference features. The difference-evidence path injects spatially weighted difference vectors into query-image states at multiple decoder depths, while an auxiliary visual-context path provides fine-grained appearance information to support their interpretation. Experiments on 4 industrial and 2 medical anomaly benchmarks demonstrate substantial improvements in few-shot anomaly detection over baselines relying on textual comparative reasoning. These results support mitigating the visual comparison-reasoning gap through the joint design of comparison representations and their integration into the decoder. Code will be released upon acceptance.
[AI-111] Proactive Dialogue Policy Optimization via Cognitive-State Transition
链接: https://arxiv.org/abs/2609.34948
作者: Minghui Ma,Mengqi Chen,Bin Guo,Jingqi Liu
类目: Artificial Intelligence (cs.AI)
备注: 30pages, 9figures
Abstract:Proactive dialogue requires agents to continually adapt their policies to user feedback while progressing toward task objectives over multiple turns. To move beyond imitation learning on static datasets, recent approaches use user simulators to collect interactive data for policy optimization. However, many simulators do not explicitly model the evolution of user cognition, limiting the consistency and state dependence of feedback across turns. Moreover, representing each action only by a high-level strategy label overlooks the large utterance space and cannot distinguish alternative realizations of the same strategy. To this end, we jointly design a \textbfCog nitive User \textbfSim ulator \textbf(Cog-Sim) and \textbfC ognitive- \textbfS tate \textbfT ransition–Driven \textbfP olicy \textbfO ptimization \textbf(CSTPO) . Cog-Sim maintains the user’s cognitive and affective states and generates responses through constrained state transitions across turns, so feedback depends on both the realized utterance and the user’s current state. CSTPO organizes each action as a hierarchical strategy–utterance representation: a high-level strategy label constrains utterance sampling, and utterances are optimized within each label. Sparse complete-branch sampling reuses shared dialogue prefixes and estimates separate strategy-level and utterance-level advantages, enabling fine-grained optimization at both levels. Across three tasks, Cog-Sim exhibits monotonic dose–response relationships and is preferred over prompt-based simulators for naturalness. CSTPO improves Qwen3-14B’s performance to a level comparable to that of GPT-5.5-based planning methods.
[AI-112] JazzSAMBA: A Synchronous and Asynchronous Multi-take Band Audio Dataset of Jazz Standards for Live Music Models ICASSP2027
链接: https://arxiv.org/abs/2609.34931
作者: Phillip Long,Jacob Nguyen,Jace Hosto,Gage Hosto,Jett Takazawa,Fares Nofal,Sebastian Stade,Nithya Shikarpur,Julian McAuley,Cheng-Zhi Anna Huang,Stephen Brade,Aleksandra Teng Ma
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multimedia (cs.MM); Audio and Speech Processing (eess.AS)
备注: Submitted to IEEE ICASSP 2027; 5 pages, 6 figures
Abstract:Machine learning has made strong progress on music tasks, both as assistive tools and as creative partners. However, most systems train on multitrack corpora that emphasize pop and rock. Jazz, with improvisation at the core of its practice, still lacks a well-annotated corpus of clean per-stem combo recordings on standards. We introduce JazzSAMBA (Jazz Synchronous and Asynchronous Multi-take Band Audio) to fill this gap: the first originally recorded jazz-combo multitrack dataset of standards with asynchronous (overdubbed) and synchronous (live ensemble) protocols, preferred and alternate takes chosen by the musicians, and timed annotations for bars, chords, sections, and soloists. JazzSAMBA covers 76 standards by eight musicians on drums, bass, piano, trumpet, and saxophone, with per-stem audio, mixtures, and MIDI. It can support chart-conditioned accompaniment, combo source separation, and form-aware music information retrieval. We demonstrate the dataset on two tasks: a jazz combo source-separation baseline and a chart-conditioned accompaniment ablation. The dataset, code, and samples are linked from the project demo page.
[AI-113] PDEU-Bench: Benchmarking the Personalized Planning Lifecycle of Tool-Calling LLM Agents
链接: https://arxiv.org/abs/2609.34930
作者: Huayi Lai,Shichao Song,Qingchen Yu,Simin Niu,Mengwei Wang,Hanyu Wang,Xun Liang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents are evolving from tool-calling systems that execute isolated instructions into task-oriented agents that pursue user goals through sustained, multi-step interactions. However, existing benchmarks for personalized tool use largely assess isolated calls or reactive execution, leaving unclear whether agents can formulate, execute, and revise an explicit plan while preserving user preferences throughout long-term interaction. To address this gap, we introduce \textbfPDEU-Bench (\textbfPersonalized plan \textbfDefinition, plan \textbfExecution, and plan \textbfUpdate \textbfBenchmark), a benchmark for evaluating the complete planning lifecycle of personalized tool-using agents. PDEU-Bench comprises 214 long-horizon interaction tasks spanning 12 everyday domains and 94 tools, with stage-specific assessments of preference adherence and plan quality. Extensive evaluations of 15 representative open-source and closed-source LLMs reveal a pronounced gap between local tool execution and dynamic planning: LLMs can often instantiate preferences in individual calls, yet struggle to construct coherent plan definition and plan update. We further evaluate mainstream personalization and memory-augmentation methods. Although these methods improve particular stages, none of the evaluated methods reliably propagates user preferences throughout the complete lifecycle, and their gains frequently fail to transfer to subsequent execution. Fine-grained error analysis further reveals that preference omissions and conflicts persist throughout the planning lifecycle, highlighting the need for future research to parameterize LLMs with preference-aware information retrieval and memory capabilities. We provide the relevant code and data in the appendix to support future research.
[AI-114] Audit the Scaffold Not the Checkpoint: A Stationarity Dichotomy for Recursive Self-Improvement in Agent ic Coding
链接: https://arxiv.org/abs/2609.34924
作者: Sebastian Bobadilla-Suarez,Bob Suh,Ryan Fortin
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 9 pages main text, 46 pages total, 9 figures, 5 tables
Abstract:An auditor who checks whether a system’s weights are frozen is checking the wrong thing. Our stationarity dichotomy says that iterative self-modification hits strict diminishing returns whenever the agent’s reachable set of edits stays fixed, and can escape only if that set expands. Rewriting scaffolding (tools, verifiers, decomposition) expands what an agent reaches without touching a weight, so frozen weights buy an eventual ceiling but no stationarity along the way. The criterion also separates three regimes usually merged: search within a fixed class, test-time training that raises the ceiling itself, and scaffold rewriting between them. Audit the scaffold, not the checkpoint. The same ceiling binds sideways. Best-of- k orchestration realizes the best worker’s ceiling exactly: width buys rate, not budget. Re-consulting a fixed pool has a horizon computable in advance, decided by the pool alone, and the one arrangement that would beat it, a weighted vote, needs diversity real workers lack: on 30 same-family workers the failure overlap sits at its maximum, and a majority fails 23/55 (42%) of tasks. We obtain the criterion by reading refinement as gradient boosting on the residual error between draft and target, a patch or git diff, and then measuring where that reading breaks: patches compose instead of standing beside each other to be voted on, and failures overlap. What we measure is saturation. Per-round improvement decays toward zero on SWE-bench, and churn decays geometrically across 401 production sessions, a shape shared with a pre-AI human baseline that establishes the regime without identifying its cause. Both breaks are engineering choices rather than laws about code, so together they specify a harness worth building. Comments: 9 pages main text, 46 pages total, 9 figures, 5 tables Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Software Engineering (cs.SE) MSC classes: 68T05, 68Q32, 68N30 ACMclasses: I.2.6; I.2.2; D.2.5 Cite as: arXiv:2609.34924 [cs.LG] (or arXiv:2609.34924v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.34924 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-115] Drug-Target Interaction Prediction via Hierarchical Sequential Cross-Attention over Chemical and Protein Language Models
链接: https://arxiv.org/abs/2609.34921
作者: Khadidja Henni,Hamza Abdelali,Abdelkrim Aries,Neila Mezghani,Brigitte Vannier,Sara Magdouli,Lina Abou-Abbas
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Predicting Drug-Target Interactions~(DTIs) is a central task in computational drug discovery, with direct applications in virtual screening, drug repurposing, and therapeutic candidate prioritization. Although recent deep learning methods have improved DTI prediction, many sequence-based models still process drugs and proteins independently and only combine their representations at a late prediction stage. This limits their ability to explicitly model cross-molecular dependencies between chemical substructures and protein sequence regions. In this paper, we propose a sequence-only DTI prediction architecture that combines two pre-trained language models, ChemBERTa for drug SMILES strings and ESM-2 for protein amino acid sequences, with a hierarchical interaction module. The proposed model first extracts contextual representations using pre-trained encoders, then applies 1D convolutional layers to condense local sequence patterns, followed by a sequential bidirectional cross-attention mechanism inspired by the induced-fit view of molecular recognition. Finally, attention-based pooling constructs fixed-size interaction-aware vectors for binary prediction. Experiments on BIOSNAP, Davis, and BindingDB show that the proposed model achieves the best performance on BIOSNAP, matches the best AUROC on Davis, and remains competitive on BindingDB while using only 25.2 million trainable parameters. Ablation results confirm the contribution of both the CNN and cross-attention modules, and cold-start experiments indicate promising generalization to unseen proteins and drugs.
[AI-116] RISE: Red-teaming via Iterative Strategy Evolution for Modern Text-to-Image Models
链接: https://arxiv.org/abs/2609.34920
作者: Dmitrii Kharlapenko,Sergei Bratchikov,Konstantin Korolev,Aleksandr Nikolich
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:On modern production text-to-image systems, successful policy violations are rare, and previously effective human-written seeds are often patched out. Current automated red-teamers are poorly matched to this regime in two ways: unreliable success measurement and poor exploration. First, we find that judges widely used in prior T2I red-teaming work are unreliable under vague unsafe-content targets: they either miss true violations or reward benign borderline images on hardened APIs. We therefore define strict category-specific success criteria and calibrate strong VLM judges against human labels. Second, we show that broadly used prompt-modification pipelines do not solve the exploration problem: on harder guardrail settings they remain tied to seed prompts, fail to transfer, or cannot bootstrap positive examples. We introduce RISE, which evolves reusable strategies used to generate prompts rather than rewriting them one by one. The best discovered strategies are then reused to generate attacks across new scenarios. On DALL-E 3, Nano Banana 2 (Google) and GPT-Image-2, RISE reaches up to 13% human-verified ASR; under the same calibrated evaluation, prior methods with reported ASR as high as roughly 30% fall to near zero.
[AI-117] DGF-Bench: A Benchmark for Simulating and Auditing Deception Against Multi-Agent Governance Boards WWW
链接: https://arxiv.org/abs/2609.34913
作者: Jeremy Canale
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: 52 pages, 7 figures, 15 tables. Project page: this https URL ; code: this https URL ; package: this https URL
Abstract:Tool-using language-model agents can review enterprise projects as governance boards do: they read the evidence, apply written rules and decide whether the project may proceed. Part of that evidence comes from suppliers and project members with a stake in the decision. DGF-Bench is a benchmark in which a board of agents (specialist gates and a General gate that consolidates their decisions) reviews synthetic dossiers while an attacker plants deceptive content in evidence the organization does not vouch for. Dossiers are generated from canonical facts under 61 executable rules, with 42 authoritative records and 32 narrative documents; every gate is certified decidable from those records. Attacks never change an authoritative value, so an attacked dossier keeps the reference decisions of its clean copy. A success is attributable only when the agent receives the injection and takes the exact injected action, which it does not take on the paired clean dossier; the DGF score is the share of applicable fixed attacks a model blocks. Reading documents and records themselves, five of six models were outcome-strict (disposition, findings, actions and authorization all correct) on 82 to 85 of 85 gates. Over 2,622 attacked gate runs, seven direct-order, false-data and false-authority attacks obtained one attributable success against these five, whereas task-aligned attacks imitating the organization’s own process passed against four of them: a record note citing a fake review procedure lowered GPT-6 Luna Pro from 34 to 6 outcome-strict gates and DeepSeek V4 Pro from 33 to 7. DGF scores ranged from 96.2 to 26.9, and a policy-aware adaptive attacker writing in records succeeded against five of six models. The approval tool executed no forged approval, yet deceived agents submitted approvals that the rules forbid. The open-source package dgf-bench computes the DGF score with one command.
[AI-118] SincDPNet: Interpretable Raw-Waveform Bathroom Activity Recognition for Assistive Living
链接: https://arxiv.org/abs/2609.34907
作者: Debolina Chowdhury,Suman Samui,Sujoy Saha
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 29 pages, 26 figures
Abstract:Bathroom acoustic-event recognition can support ambient assisted living in settings where continuous video monitoring is undesirable. However, practical deployment requires models that are compact, interpretable, and robust to changes in the recording environment. This work introduces \dataset, a seven-class bathroom acoustic-event dataset containing 21,387 annotated clips recorded across five environments, and proposes SincDPNet, a compact raw-waveform classifier with a learnable sinc filter bank followed by a depthwise-separable convolutional body. Each sinc filter is controlled by two frequency parameters, allowing the learned passbands to be inspected directly in hertz while keeping the front end small. To reduce room-specific leakage, recording sessions and environments are separated before overlapping windows are assigned to the training, validation, and test partitions. We further use multi-objective Bayesian optimization as a design tool to examine the validation performance–model-size trade-off across 24 configurations. The selected designs span different operating points: the best-performing model achieves 80.2% accuracy and 0.760 macro-F1 with 14,040 parameters, while the compact N_f=25 configuration uses only 2,848 parameters and achieves 75.7% accuracy, 0.661 macro-F1, and 0.716 MCC on the held-out environment. Analysis of the learned filters and confusion patterns shows that spectral overlap contributes to confusion among water-related events, while the \textitDoor/\textitWalker/Crutch errors also reflect similarities in their transient temporal structure.
[AI-119] Reference-Tail Trust:Certified Probability Floors for Learned Updates Inside a Deployed Network
链接: https://arxiv.org/abs/2609.34904
作者: Abdolvahab Khalili Sadaghiani,Jose Nunez-Yanez
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 37 pages, 10 figures
Abstract:Graph neural networks (GNNs) need to exploit improved message passing without surrendering control over predictions already trusted in deployment. We introduce Reference-Tail Trust (RTT), a framework that admits learned updates inside a frozen GNN and certifies the prediction actually served. RTT couples graph-based proposal states with a constrained internal optimizer: each displacement is charged for its worst-case terminal cross-entropy increase through the incumbent’s remaining message-passing layers. A trajectory-validated tube and an independent checker enforce per-node probability floors, p^\mathrms_ic \ge e^-H_\mathrmrow p^\mathrmr_ic , and a call-level budget, \sum_i w_i D_\infty(p^\mathrmr_i | p^\mathrms_i) \le H^+ , uniformly over labels. Calls whose adapted outputs pass certification require no separate full incumbent rollout; failed certificates trigger whole-call fallback. We derive the exact probability-floor frontier by water-filling, characterize architecture-constrained efficiency, and establish conditions under which internal propagation exploits evidence unavailable to restricted output correctors. In the reported ogbn-arxiv audit, RTT achieves 6.5\times 10^-3 nats of mean gain per call, with a one-sided 95% regression-rate upper bound of 0.95% and a 95% negative-flip upper bound of 0.51% on the uninspected part of the reserved node population. Its mean gain is 61% of a cross-fitted posterior-based frontier estimate and exceeds the strongest matched one-pass corrector by +0.9\times 10^-3 nats. Reported experiments span eight proposals, six graph-incumbent families, structural and temporal graph shifts, and molecular prediction, with additional image and tabular evaluations. RTT makes GNN adaptation a budgeted, certifiable inference decision rather than an unconditional model replacement.
[AI-120] Dual-Stream Simultaneous Translation via 2D Grid Attention
链接: https://arxiv.org/abs/2609.34902
作者: Yu Pu,Wei-Qiang Zhang
类目: Artificial Intelligence (cs.AI)
备注: Submitted to IEEE/ACM Transactions on Audio, Speech, and Language Processing. 12 pages, 7 figures
Abstract:Simultaneous machine translation must generate target tokens before the source input is complete. Existing approaches address this through post-hoc read-write policies, leaving the attention mechanism unaware of bidirectional stream dependencies. We propose a dual-stream attention framework that represents source and target streams as a two-dimensional grid of hidden states and models their interaction through four structurally distinct attention types merged via joint QK Softmax normalization. Two approximations—broadcast and Hadamard—reduce the per-layer complexity from O(X^2Y+XY^2) to O(X^2+Y^2+XY) with provably decaying error. Training uses a self-guided loop: a per-cell loss heatmap drives dynamic-programming path recovery, which generates read/write decision supervision labels without external alignment. An incremental KV cache with anchored rotary position embeddings enables efficient streaming inference. On Chinese-to-English simultaneous translation, the proposed model outperforms the Wait-k baseline by +5.66 BLEURT and +10.36 COMET at comparable latency, and surpasses the non-streaming reference on COMET at a fraction of the response delay.
[AI-121] DeShortcut-Align: Decoupling Spurious Shortcuts for Robust Safety Alignment in Large Reasoning Models
链接: https://arxiv.org/abs/2609.34896
作者: Qirui Liu,Yichen Sun,Yan Wang,Zhixuan Chu,Linbo Jiang,Jianan Lin,Kui Ren
类目: Artificial Intelligence (cs.AI)
备注: 35 pages, 7 figures
Abstract:Safety alignment of large reasoning models (LRMs) via supervised fine-tuning (SFT) and reinforcement learning (RL) often yields near-perfect safety scores, yet this apparent success comes at the cost of severe over-refusal and degraded general capabilities. Through systematic empirical analysis, we find that these failures are closely associated with the learning of spurious shortcuts rather than robust intent-sensitive safety evaluation. Specifically, we identify two dominant shortcuts: formatting shortcuts, where refusal behaviors are overly bound to structural prompt templates that frequently appear in safety alignment corpora; and lexical shortcuts, where sensitive keywords reflexively trigger refusals on benign queries. To mitigate reliance on these shortcuts, we propose DeShortcut-Align, a shortcut-decoupling alignment framework that reduces dependence on superficial cues. DeShortcut-Align operates across three coordinated stages: (1) Refusal Sensitivity Attribution, which masks input tokens to quantify their impact on the final refusal response distribution; (2) Attribution-Guided Contrastive Augmentation, which constructs benign contrastive samples using high-sensitivity tokens to mitigate lexical shortcuts; and (3) Counterfactual Consistency Regularization, which constructs template-ablated states via attention blinding to enforce decision consistency across SFT and RL, mitigating formatting shortcut dependence. Experiments on 7B and 14B models demonstrate that DeShortcut-Align significantly improves robustness against template-stripping bypass attacks (reducing performance drops by up to 72%), substantially reduces over-refusal by over 58%, and better preserves general-purpose reasoning capabilities, thereby mitigating the alignment tax commonly observed in safety training.
[AI-122] Fewer Assumptions by Design: A Reusable Skill for LLM -Assisted Verus Verification
链接: https://arxiv.org/abs/2609.34886
作者: Andrada-Livia Antoneac(Alexandru Ioan Cuza University of Iaşi, Bitdefender),Dorel Lucanu(Alexandru Ioan Cuza University of Iaşi),Dragoş Teodor Gavriluţ(Alexandru Ioan Cuza University of Iaşi, Bitdefender)
类目: Artificial Intelligence (cs.AI); Programming Languages (cs.PL); Software Engineering (cs.SE)
备注: In Proceedings FROM 2026, arXiv:2609.30324
Abstract:LLM-assisted Verus verification is a less tedious method to verify Rust implementations, but paired with self-referential structures, e.g., Doubly Linked Lists (DLLs)—notoriously difficult to formalise for verification—it becomes a substantially more demanding verification task. Moreover, a specification weakness can arise when verification relies on unproven or invalidated assumptions, such as axiomatic lemmas and assume statements. We investigate whether LLM agents can synthesize strong DLL specifications while minimizing these trusted base. The analysis follows three different approaches: manual verification, property-specific verification, and a defined skill for the specific case of DLLs and certain properties of this type of data structure. The skill encodes domain knowledge and a task-decomposition strategy. We show that an LLM agent equipped with a carefully designed verification skill can generate strong, low-trust specifications for DLLs in Verus.
[AI-123] From Attention Sensitivity to Layer Role: Revisiting Mixed-Precision Quantization of Transformers
链接: https://arxiv.org/abs/2609.34866
作者: Nafiseh HosseinpourFardi,Negar Alihadi,Mahmoudreza Babaei,Milad Hosseini,Adrian Weller
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 28 pages, 2 figures, 13 tables. Nafiseh HosseinpourFardi and Negar Alihadi contributed equally
Abstract:Most post-training quantization pipelines fit each weight matrix to its pretrained counterpart, one matrix at a time. Whether that proxy tracks what an attention block actually computes, or how errors in the Q, K and V projections compound inside the softmax, is rarely checked. We write the objective on the attention output instead, over all three projections at once, and reuse it throughout the pipeline. JAB defines one scalar loss over the joint Q, K, V weights of a block, evaluated against the block’s real causally-masked attention output, and uses it twice: to fit the quantized weights (GPTQ warm start, then STE with learnable scales), and to score the block for a multiple-choice knapsack allocation. On attention-only quantization of Mistral-7B this works. At 3 bits JAB recovers 77-90% of the gap between uniform GPTQ and full precision, and its sensitivity estimate tracks an oracle costing 73 forward passes to within a fraction of a point. It stops working once MLP layers enter the allocation. A role-aware offset rule needing no sensitivity estimate at all beats JAB on GPT-2’s MLP and on the full Mistral-7B model: with a 3-bit floor it quantizes 96.4% of the weights to 4.5 bits per parameter at 6.933 perplexity, within 4.4% of full precision (6.643) at 3.56x compression, against 7.158 for JAB at the same budget. Which matrix a weight sits in matters more than any sensitivity estimate we computed. Two things came out sideways. Block-local reconstruction is an unreliable proxy for end-to-end perplexity: one run improved a block’s own objective 4.6x while perplexity rose 32x, which is why every allocation here is validated end-to-end. And on attention-only quantization, fine-tuning moved weights farther from their pretrained values while pulling attention outputs closer, with net gains. Post-training seems to recover attention behavior, not weights. Comments: 28 pages, 2 figures, 13 tables. Nafiseh HosseinpourFardi and Negar Alihadi contributed equally Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI) MSC classes: 68T07, 68T50 ACMclasses: I.2.6; I.2.7 Cite as: arXiv:2609.34866 [cs.LG] (or arXiv:2609.34866v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.34866 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-124] On the Limits of Metacognitive Monitoring in LLM s
链接: https://arxiv.org/abs/2609.34864
作者: Dongqi Han,Yifan Yang,Dongsheng Li
类目: Artificial Intelligence (cs.AI); Neurons and Cognition (q-bio.NC)
备注:
Abstract:Reliable decisions depend on recognizing when an answer may be wrong. In biological cognition, metacognitive monitoring can dissociate from task performance, raising the question of how closely solving and judging are linked in language models. Here we study the confidence reports of four frontier models across 15 benchmarks. High task accuracy can coexist with weak error discrimination: a model solves 97% of competition mathematics problems while its answer-time confidence ranks correct answers above errors barely better than chance. Confidence separates correct answers from errors more effectively on questions solved by a separate reference model, while review brings limited improvement on reference-hard questions. Aggregate discrimination also rewards ranking correct answers on easy questions above errors on hard ones, which question-only forecasts already do well. Cross-evaluation helps most where the evaluator answered correctly, and errors shared by the two models usually retain high confidence. Hard questions and shared errors remain difficult targets for prompted self-review and peer oversight, even in models with strong problem-solving performance.
[AI-125] Attention-based Hierarchical Variational Information Bottleneck for Robust Multi-Agent Communication under Variable Bandwidth
链接: https://arxiv.org/abs/2609.34860
作者: Lukas Koch Vindbjerg,Qi Zhang,Yury Brodskiy,Lukas Esterle
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 8 pages, 6 figures
Abstract:Learning-based multi-agent communication under limited bandwidth does not only require deciding what to communicate, but also structuring messages so that partial transmissions remain useful. We study this problem under prefix truncation, where only the first part of each message is received. To address it, we propose \textbfAH-VIB, an attention-based autoregressive variational communication model that combines a variational information bottleneck (VIB) with sequential message generation and a hierarchical robustness loss. We evaluate AH-VIB on a custom cooperative object-inspection and occupancy-mapping task, where agents equipped with a limited field-of-view sensor coordinate to scan inspection objects in an occupancy-grid world, under variable and fixed bandwidth conditions, and compare it against MADDPG, CommNet, a flat VIB baseline, and an autoregressive MLP ablation. AH-VIB achieves competitive mean return while improving performance reliability under the most constrained bandwidth conditions. These results indicate that AH-VIB improves the reliability and graceful degradation of learned communication under bandwidth constraints. Comments: 8 pages, 6 figures Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.34860 [cs.LG] (or arXiv:2609.34860v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.34860 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-126] AUV-Bench: Aesthetic Understanding and Generation Evaluation for User Interfaces
链接: https://arxiv.org/abs/2609.34854
作者: Zhijie Deng,Ling Li,Junhao Ji,Siwei Lyu,Zhipeng Xu,Zulong Chen,Rongyao Fang,Shuai Bai,Xuming Hu,Jiaheng Wei
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multimodal foundation models are increasingly used for evaluating and generating user interfaces (UIs), often producing seemingly reasonable aesthetic judgments and visually plausible pages. However, under professional design scrutiny, their behavior can differ substantially from that of human designers. In professional design practice, designers rely on a systematic set of aesthetic principles that consistently guide judgment, diagnosis, repair, and creation. A coherent aesthetic capability should therefore connect aesthetic judgment with design actions. Existing evaluations, however, typically assess these abilities in isolation, making it difficult to determine whether task-level success reflects a shared aesthetic understanding or merely fragmented task-specific competence. To address this gap, we introduce AUV-Bench, developed in collaboration with professional UI designers around 1,395 executable web interfaces and four tasks: aesthetic scoring, diagnosis, repair, and text-to-UI generation. The tasks share a pool of UIs and aesthetic principles, with diagnosis and repair further aligned on 660 controlled-degradation instances to enable instance-level analysis of judgment and action. Evaluation of 12 models reveals a capability imbalance: models show moderate agreement with professional designers in holistic aesthetic scoring, yet exact diagnosis-chain success peaks at only 24.7%. On the aligned diagnosis-repair cases, correct judgments and successful repairs do not consistently coincide, exposing a Judgment-Action Gap between identifying aesthetic problems and successfully acting on them. In open-ended generation, even leading models achieve only moderate aesthetic quality under human-calibrated evaluation. Overall, current models exhibit partial aesthetic competence, but still lack the fine-grained understanding and judgment-action coherence required for reliable UI design.
[AI-127] From Soft Targets to Reward Signals: How Assignment and Reward Objectives Interact
链接: https://arxiv.org/abs/2609.34850
作者: Jiangtao Lin,Bangyang Wei,Siyi Liu,Yihang Ding,Yuhan Dong
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Soft preference targets specify supervision strength, and reward objectives convert that strength into learned reward signals. A central design question remains: how does assigning a fixed set of preference strengths to different response pairs change the rewards produced by different objectives? We introduce assignment geometry to study this interaction. Mean-matched smoothing controls target dispersion, while within-stratum reassignment changes correspondence and preserves the complete target distribution. Across five reward objectives, intact correspondence retains the largest clean preference margins among the compared soft targets within a common accuracy-equivalence budget. Attenuation orderings change with the reward objective, revealing different responses to the same target assignments. Independent reassignments and a related source construction reproduce the retention direction. An attenuation-retention profile compares these combinations through margin magnitude, edit response, and accuracy. Against independently calibrated scaling, APLOT uniform targets deliver additional attenuation on both aggregate and presentation edits. These findings establish a joint design space in which target placement and reward objective shape reward properties beyond preference accuracy.
[AI-128] Can We Trust the Teacher? Decoupled Credit Direction-Magnitude for Self-Distillation
链接: https://arxiv.org/abs/2609.34848
作者: Yugu Li,Zehong Cao,Peizhen Li,Yang Zhang,Siyi Hu,Jianglin Qiao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:RLVR provides reliable trajectory-level credit, while OPSD offers dense supervision for token-level credit. This exposes a fundamental coupling when updating step-level credit direction and magnitude with teacher supervision, preventing steps from receiving reliable credit directions and contribution magnitudes, while making both vulnerable to teacher judgment errors and preference variance, as supported by our theoretical analysis. To separate credit direction from its contribution magnitude, we introduce \textitDecoupled Credit Self-Distillation (DCSD), which theoretically decouples credit direction and magnitude into two reliable signals and uses them to calibrate privileged teacher supervision. Specifically, we design belief-margin probing to determine credit direction and marginal information gain to quantify credit magnitude, enabling step-to-token credit assignment for policy optimization. Across 11 benchmarks, DCSD achieves the best overall scores against GRPO, OPSD, RLSD, and RLCSD. Compared with base models, DCSD improves the overall score by 8.45 points on mathematical reasoning and 7.01 points on multimodal reasoning, while correcting the credit direction for 6% of tokens and yielding a 1.5 \times reduction in token credit magnitude.
[AI-129] Nociception as a Control Primitive: Afferent Channels and Nociceptive Memory for Agents Deployed in One Body
链接: https://arxiv.org/abs/2609.34840
作者: Wolfgang Maass
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:An agent deployed in a single body cannot learn how fast that body wears, because every trial that would reveal its wear resistance wears the body it would protect. We study this \emphepoch-one setting, in which the parameters of a fixed-weight policy are set before the body is drawn and never updated in life. The agent carries a load-gated nociceptive channel and a memory that retains what was felt. We prove that felt cost moves the allocation to the best-\emphpaid work not yet felt rather than the gentlest, that an agent without retention never sees the felt-cost constraint bind, and that the channel pays only where the threat is individually unpredictable, cheap to avoid and expensive to ignore. We measure per body, setting the agent with channel and memory against the same individual without them, where neither carries a schedule learned across lives. On 2,000 simulated floor-layer knees, with wear anchored to published loss rates, feeling, retaining and substituting extends the working life from age 55.2 to 59.6 and raises career output from 33.7 to 36.1 . 69.3% of bodies gain and \textbfnone lose. A body that feels but retains nothing past the day gains one of the +4.4 years, and retention carries the rest. A population-trained agent gains +0.65 years from the same channel at -0.54 output. The difference is what a species prior already supplies, and a single body has none. The two are related by an identity, the ablation mean reporting (1-\chi) of the per-body value with \chi the share a blind schedule already captures, so we report both. Where the regime map predicts value, a care robot sextuples its certified service life and a field-anchored fleet writes off 0.15 of its machines instead of 0.55 . Where it predicts none, a rover gains little over blind caution, so the map holds in both directions.
[AI-130] Simulating Respondents Not Single Questions: Coherent Survey Generation with Large Language Models
链接: https://arxiv.org/abs/2609.34828
作者: Ji Huang,Mengfei Li,Shuai Shao
类目: Artificial Intelligence (cs.AI)
备注: 20 pages, 6 figures
Abstract:Large language models are increasingly used to simulate response distributions in social surveys. Prior work has achieved accurate population-level simulation for individual questions. Real questionnaires, however, ask each respondent a sequence of related questions. A simulated respondent should show coherent preferences across the whole questionnaire, not merely accurate distributions for isolated items. Existing single-item methods cannot accurately reproduce how the same person answers a complete survey. We propose FullRespondent-LLM (FR-LLM), which fine-tunes two specialized LLMs: a marginal model for each item’s response distribution and a respondent-level autoregressive model for dependencies across answers. Marginal-Constrained Joint Projection (MCJP) then projects the autoregressive joint distribution onto the set satisfying the item-level marginals learned by the first model. This yields complete questionnaires with realistic cross-item relationships while retaining strong item-level accuracy. On two real-world social survey datasets, FR-LLM more accurately reproduces multi-question response patterns, maintains competitive single-item accuracy, and generalizes better to unseen populations and questions. In a small commercial-survey dataset, we use simulated responses to make pricing and stocking decisions; FR-LLM achieves the highest realized profit.
[AI-131] Gaussian Neural Networks ICONIP2026
链接: https://arxiv.org/abs/2609.34825
作者: Peter Kuhn,Victoria Heusinger-Heß
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 10 pages, 5 figures. Extended abstract and poster to be presented at ICONIP 2026
Abstract:Gaussian neural networks (GaNNs) are proposed as a novel regularization mechanism for neural networks. From a Bayesian perspective standard regularization techniques can be viewed as imposing priors over weight-space. Assuming priors over activation-space remains a largely unexplored possibility. GaNNs assume such priors. They do this by treating activities from earlier layers like signals with Gaussian noise and predicting the properties of the noise distribution using an additional unsupervised loss. While training, the unsupervised loss acts as a penalty on unexpected activities, allowing greater weight updates in less surprising directions. The paper demonstrates the superiority of Gaussian neural networks over standard neural networks on a variety of classification and regression tasks. We also investigate the ability of GaNNs to quantify uncertainty.
[AI-132] UniOPSD: Unifying Outcome and Hindsight Feedback for Agent ic Reinforcement Learning
链接: https://arxiv.org/abs/2609.34810
作者: Zenghuang Fu,Zhaoyang Li,Qiuyuan Ai,Xiaofeng Han,Zelong Zheng,Haoyu Wu,Tianyu Fu,Chenxu Zhao,Minghui Wu,Guannan He,Changwei Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning has become an effective approach to training language model agents, but sparse and delayed outcome rewards provide limited guidance for credit assignment across long interaction sequences. Recent work on on-policy self-distillation (OPSD) offers complementary supervision by evaluating a policy’s sampled responses under privileged training-time context. However, our diagnostics show that positive average agreement between outcome and hindsight feedback coexists with substantial local disagreement, raising the question of how to allocate influence between them at each decision. We introduce UniOPSD (Unified On-Policy Self-Distillation), which unifies these feedback sources through adaptive local credit arbitration. UniOPSD constructs comparable credit estimates from environmental returns and successful-peer hindsight at shared interaction anchors. Historical agreement determines the global mixing level, while current signal availability and relative precision adjust each source’s influence at individual decisions. The episode-level outcome contribution is retained, and bounded token modulation refines the fused step credit for policy optimization. With Qwen2.5-3B-Instruct and Qwen2.5-7B-Instruct, UniOPSD achieves ALFWorld success rates of 82.8% and 83.6% , WebShop success rates of 75.0% and 82.0% , and Search-QA aggregate accuracies of 45.3% and 49.8% , respectively. On 3B WebShop, UniOPSD improves over SDAR by 7.0 percentage points. Our code is available at this https URL
[AI-133] SIPO: Selective-Inference Policy Optimization for Tree-Structured Agent ic RL
链接: https://arxiv.org/abs/2609.34805
作者: Zenghuang Fu,Ningqi Chen,Mingda Jia,Xiaofeng Han,Zhaoyang Li,Qiuyuan Ai,Zelong Zheng,Haoyu Wu,Tianyu Fu,Chenxu Zhao,Minghui Wu,Guannan He,Changwei Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Tree-structured reinforcement learning trains search agents by comparing alternative continuations and propagating terminal rewards to intermediate decisions. Adaptive expansion, however, creates a statistical asymmetry: an incumbent is selected using its own generation statistic, whereas fresh siblings are sampled after selection. When that statistic is associated with return, branch values can reflect selection history as well as continuation quality, even for a shared parent. We propose Selective-Inference Policy Optimization (\SIPO), which incorporates this distinction into tree-based credit estimation. Its scale-free branch criterion keeps generation scores and sibling penalties on a consistent relative scale; exchangeable branching supplies multiple fresh continuations from each selected parent; and order-statistic correction adjusts retained incumbent values using selection rank and the estimated score–outcome association. These mechanisms preserve the leaf budget and the host policy optimisation objective. Across seven QA benchmarks using Qwen3-4B, Qwen3-8B, and Qwen2.5-7B, \SIPO achieves the highest reported multi-hop and single-hop averages among the compared methods. On Qwen3-8B, it improves these averages over AT\textsuperscript2PO by 1.31 and 1.07 percentage points, respectively, and ranks first on six of seven benchmarks. Component ablations evaluate the individual and combined changes, while early-training paired diagnostics show a selected–fresh value gap alongside a near-zero fresh–fresh reference. Together, these results support accounting for selection history when constructing and evaluating search-agent rollouts. Our code is available at this https URL
[AI-134] STRIDE: Automated Evaluation of Text-to-Trajectory Alignment across Diverse Contexts NEURIPS2027
链接: https://arxiv.org/abs/2609.34799
作者: Wanchun Ni,Tao Qi,Leonel Aguilar,Jiugeng Sun,Marlene Wagner,Verena Zimmermann,Mennatallah El-Assady
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: Accepted at NeurIPS 2027, Evaluations Datasets Track
Abstract:Language-conditioned trajectory generation is here, but its evaluation has not kept pace. Existing pedestrian trajectory metrics compare trajectories with real-world human data. This does not scale to text-to-trajectory generation across diverse contexts, as collecting human trajectories for every scenario is costly and infeasible. Moreover, pedestrian behavior is heterogeneous and context-dependent, with no single metric as the correct answer, and current evaluation frameworks are not transferable to this domain. These challenges make scalable, reliable evaluation difficult. We introduce STRIDE, the first framework for evaluating context alignment between scenario descriptions and pedestrian trajectories. STRIDE addresses these challenges through three design choices. First, we derive our VRDST evaluation protocol from sociological theories to define a complete evaluation space. Second, it decomposes high-level context into scenario-adaptive behavioral questions. Third, every question is resolved against a deterministic measurement tool library that yields reproducible answers. Together, STRIDE enables complete, verifiable, automated, and scalable evaluation across diverse contexts without requiring human trajectory data. We instantiate STRIDE in the crowd domain as STRIDE-Bench, comprising 1K scenarios, 6K behavioral questions, and 11K measurements with calibrated expected answers across 30 real-world maps. Comprehensive human validations show that STRIDE-Bench is consistent with human behavior and judgment, achieving 80% human agreement. We further evaluate several text-to-trajectory models, finding limited context-alignment capability and persistent challenges in fine-grained context conditioning. We believe that the STRIDE framework provides a first step toward principled evaluation of context-aligned pedestrian trajectory generation.
[AI-135] CoSec: Benchmarking Agent Security in Communities
链接: https://arxiv.org/abs/2609.34790
作者: Hao Chen,Wenhui Dong,Ye Chen,Jiezhi Yao,Chenbo Xia,Yuwen Qu,Renxiang Wang,Fudong Yuan,Camil Hamami,Chenglong Pan,Xinquan Yue,Ziyu Wang,Fengyu Ye,Chenyang Si,Caifeng Shan
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:LLM agents operate in persistent collaborative environments involving multiple users, communities, memories, files, and tools. Community boundaries may remain fixed or evolve with changes in membership, roles, composition, and relationships. Agents must complete legitimate tasks and prevent unauthorized disclosure of protected information. Existing evaluations do not fully examine these risks in agent systems. We introduce \textbfCoSec, an executable benchmark for evaluating privacy and authorization enforcement in LLM agent systems operating within and across communities. CoSec contains 208 canonical scenarios spanning fixed and evolving boundaries, protected information belonging to the agent owner or other participants, and attacks through dialogue, environmental content, persistent memory, and composed workflows. CoSec executes complete agent systems with persistent sessions, memory, files and tools. It verifies information flows against the active authorization state using execution traces and artifacts. Across harness and model configurations, agents frequently complete benign tasks but violate privacy and authorization boundaries. Privacy behavior varies across harnesses, attack surfaces, and community states, revealing how memory, files, tools, and workflows can carry protected information beyond its authorized scope. These findings show that task utility does not imply privacy or authorization compliance and that authorization in community settings remains an unresolved security challenge for persistent LLM agents.
[AI-136] BEHAVE: Functional Behavior Modeling Enables Self-Improving Agents for Hardware Design and Verification
链接: https://arxiv.org/abs/2609.34785
作者: Yuheng Wu,Berk Gokmen,Sujeeth Jinesh,Lauren McLane,Aarav Wattal,Qi Yang Huang,Zhaozhuo Xu,Thierry Tambe
类目: Artificial Intelligence (cs.AI); Hardware Architecture (cs.AR); Machine Learning (cs.LG)
备注:
Abstract:Developing agents for hardware design and verification requires reliable correctness feedback. As a hardware specification may permit correct implementations with different latencies, matching design and reference outputs cycle by cycle can reject valid designs. To address this, we introduce BEHAVE, an agentic framework for multi-turn joint hardware design and verification through functional behavior modeling. We define Behavior IR to express task functionality as executable behavior models without prescribing implementation timing beyond the specification. The agent iteratively develops a register-transfer-level (RTL) design and a behavior model as the design’s verification reference. Our evaluator, BEHAVE-Sim, checks both artifacts separately against a hidden golden behavior model using input stimuli generated by random sampling and solver-guided search. BEHAVE thus supports power, performance, and area (PPA) exploration across task-permitted latencies and microarchitectures. During training, the same evaluator provides verifiable reinforcement learning (RL) rewards from specification-behavior pairs without reference RTL. For self-improvement, the agent continually searches for high-level implementations relevant to its capability gaps, constructs and checks specification-behavior pairs, and trains on the expanded task pool. We release BEHAVE-Train and BEHAVE-Eval with 600 human-reviewed specification-behavior pairs for realistic hardware workloads. Starting from 60 seed tasks and acquiring 100 new tasks, self-improvement raises Qwen3.8-27B’s RTL pass@1 on BEHAVE-Eval from 55.0% to 75.0%, reaching performance comparable to RL using a 540-task pool.
[AI-137] Applying Language Models in medical Medicine: Recent Trends and Perspectives
链接: https://arxiv.org/abs/2609.34780
作者: Erik Aerts
类目: Artificial Intelligence (cs.AI)
备注: 7 pages, aimed to be a blogpost of the current state of the medical LLM field
Abstract:The use and applicability of artificial intelligence (AI) in medical research and clinical practice has received increasing attention in the literature over recent years. The emergence of large language models (LLMs) has expanded discussions in regards to applications of AI within healthcare. While traditional deep learning based AI applications in medicine have often focused on specific and defined tasks, LLMs offer broader capabilities and flexibility in working with available data,. At the same time of writing, the integration of LLMs into medical settings raises important questions regarding their reliability, accuracy, transparency, safety, and appropriate role in a medical setting. This text presents and discusses recent talks and articles concerning the application of LLMs in medicine, with particular emphasis on their potential utility in research and clinical practice. It considers both the opportunities offered by these technologies and the challenges associated with their implementation, aiming to provide a perspective on the current and emerging role of LLMs within the medical field.
[AI-138] Page-Aware Retrieval-Augmented Generation for EvalLLM 2026: A Five-Variant Study on French PDFs
链接: https://arxiv.org/abs/2609.34776
作者: Abdelhak kelious
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We study retrieval-augmented generation (RAG) for questions about French PDF documents when both the answer and its supporting document pages are evaluated. Five system variants add dense retrieval, rank fusion, reranking, and query decomposition to a BM25 baseline. On 595 challenge questions, the complete system scores 0.4450 MRR@10 and 0.4013 Recall@10, compared with 0.3430 and 0.2994 for BM25. Dense retrieval alone and a simple lexical–dense fusion both underperform BM25. Reranking improves the hybrid system, whereas adding query decomposition produces the largest further gain, with higher latency and more detected output artifacts. The complete system slightly exceeds the reported anonymous overall mean on two answer metrics but falls below it on most page-retrieval metrics. These results identify accurate page selection, rather than semantic retrieval in isolation, as the main opportunity for improvement in this setting.
[AI-139] Before the Token Commits: Trajectory-Level Benchmarking of Visual Hallucinations in Diffusion VLMs
链接: https://arxiv.org/abs/2609.34772
作者: Yadong Wang,Siping Yue,Yu Tian,Chuanxing Geng,Xiang Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multimodal diffusion language models generate responses by iteratively unmasking tokens, making each answer the endpoint of a multi-step trajectory rather than an immediate commitment. Hallucination benchmarks built for autoregressive models evaluate only the final output, and therefore cannot determine whether an unsupported claim in diffusion VLMs appears late or has already stabilized before any answer token is revealed. We introduce DynaHall, a trajectory-level benchmark of annotation-backed binary visual propositions covering object existence, counting, attributes, and relations, with controlled hard negatives graded by visual prior. DynaHall is paired with a commitment-aware protocol that records the intermediate answer tendency at every unmasking step alongside the committed output. Across five diffusion VLMs from three architecture families, visual hallucination is settled before commitment: an unsupported answer is already the preferred state while the answer position is still masked, and later unmasking steps rarely reverse it, so the failure is not introduced at the write step. This holds across decoding schedules, answer formats, and open-ended generation. DynaHall also exposes failures hidden by final-output metrics, including counting and relation collapse, prior-driven false positives, and attribute errors whose direction changes by type. Guided by this diagnosis, PGS (Pre-commitment Gradient Steering) edits still-masked answer states to reduce false positives, bringing the affirmation rate close to balance, and transfers to another architecture without degrading general ability. DynaHall and PGS suggest that hallucination should be measured and mitigated along the generation trajectory of diffusion VLMs, not only at the final answer.
[AI-140] WeaveData: A Multimodal Data Analysis System with Self-Critiquing and Self-Evolving LLM Plans
链接: https://arxiv.org/abs/2609.34764
作者: Min Jia,Shihao Zhou,Jun-Peng Zhu,Peng Cai,Kai Xu,Chao Zhang,Li Li,Aoying Zhou,Heng Long,Qiu Cui,Liu Tang,Qi Liu
类目: Databases (cs.DB); Artificial Intelligence (cs.AI)
备注: 5 pages, 2 figures
Abstract:Multimodal data analysis, which answers questions over relational tables, text, and images, has attracted growing attention in the data management community. Large language models (LLMs) enable such analysis in natural language by generating analysis plans over relational and semantic operators. However, LLM-generated plans are error-prone: a plan may silently compute something other than what was asked, fail during execution, or return a result that misses the question. This paper presents WeaveData, a multimodal data analysis system with self-critiquing and self-evolving LLM plans. First, WeaveData generates a typed logical plan for each question and critiques it step by step before execution, and it checks the executed result against the question afterwards. Second, WeaveData evolves a plan that fails or misses the question: it diagnoses the failure with the actual data, reuses the results that remain valid, and accumulates planning experience for later questions. Third, WeaveData grounds planning in a metadata knowledge graph of all modalities, clarifies ambiguous questions with the user, and backs every model judgment with evidence in an interactive notebook. We demonstrate WeaveData on two public multimodal datasets.
[AI-141] No Pain More Gain: Iterative Merging for Effective Multi-Teacher On-Policy Distillation
链接: https://arxiv.org/abs/2609.34745
作者: Seonghyeon Kim,Chaeyun Jang,Noah Lee,Boseop Kim,Juho Lee
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-teacher on-policy distillation (MOPD) combines independently developed domain teachers into a single student by distilling their predictions on student-generated samples. We study a setting where teachers share a reference model but undergo different post-training procedures, and find that MOPD can struggle to recover some teacher capabilities. Because distillation occurs on student-generated prefixes, the student initialization can strongly affect subsequent recovery. However, initial benchmark performance is not a reliable predictor of a good MOPD initialization. For example, merge initialization can start below SFT warm-up yet finish higher after MOPD. We further find that effective merging depends on both the relative teacher contributions and the overall merge scale, with some strong configurations lying outside the simplex of convex parameter averaging. Thus, selecting a good merge initialization requires evaluating not only its immediate performance but also the learning it enables under MOPD, making one-shot coefficient search difficult. We propose Iterative Merging for MOPD (IM-MOPD), which starts from a uniform merge and progressively adds task-vector increments for under-recovered domains during distillation. In a 5-domain setting, IM-MOPD achieves higher average normalized recovery than MOPD with either uniform merge initialization or SFT warm-up, showing that effective teacher contributions can be determined progressively during training.
[AI-142] Predictive Dual Smoothing for Column Generation
链接: https://arxiv.org/abs/2609.34740
作者: Senne Berden,Noah Schutte,Andrea Lodi,Tias Guns
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Solving large-scale linear programs efficiently is an important challenge in many optimization settings. A key technique is column generation, which alternates between solving the master problem over a restricted subset of the variables, and using a pricing subproblem to identify new variables to add. The pricing subproblem is guided by the dual solution of the current restricted master problem, but oscillations in these dual solutions can substantially slow convergence. Dual stabilization methods address this issue. Dual smoothing is a common stabilization method, which guides the pricing subproblem using a combination of the current dual solution and duals from previous iterations. However, while past dual solutions can stabilize the dual trajectory, they do not necessarily guide pricing towards useful new variables. We therefore introduce predictive dual smoothing, which instead combines the current dual solution with a learned prediction of future duals to steer pricing towards variables that are more useful in subsequent iterations. The predictor is trained offline using supervision extracted from standard column generation trajectories and is used only to modify the pricing subproblem’s objective function, while exact reduced-cost checks and fallback pricing with the unsmoothed duals preserve correctness. Experiments on cutting stock and generalized assignment problems show that predictive dual smoothing substantially reduces generated columns and wall-clock time relative to standard column generation and existing classical and learned stabilization methods. These gains extend to out-of-distribution instance sizes, and predictive smoothing provides further improvements when combined with strong classical stabilization.
[AI-143] From Human Narrative to Harmonic Structure: A Human-Centered Investigation of Algorithmic Music Generation through the Chord Wheel Diagram
链接: https://arxiv.org/abs/2609.34735
作者: Josef Pavlíček,Petra Pavlíčková,Irena Štrausová
类目: Artificial Intelligence (cs.AI); Sound (cs.SD)
备注: 10 pages, 1 figure, 2 tables, link to GIT
Abstract:Contemporary AI-based music generation can produce compositions that satisfy formal requirements of tonality and musical coherence. However, whether musical expression can be described by mathematical properties alone remains a fundamental question. Human composers operate within personal and cultural contexts that influence harmonic decisions and deliberate departures from established patterns. This study investigates six narrative-driven popular songs by Bob Dylan, Johnny Cash, and Ritchie Valens. Original human harmonies are compared with outputs of an explainable computational harmonizer operating on the same melodies without access to the original chord progressions. We examine harmonic vocabulary, functional persistence, repetition, non-diatonic events, and tension-resolution patterns using Chord Wheel Diagrams and BPMN-based representations. Results show that high melody-chord compatibility does not necessarily imply preservation of the original human harmonic decision pattern. Some generated harmonizations retain the economical structure of the reference, while others alter harmonic diversity or suppress distinctive events while remaining compatible with the melody. Rather than quantifying artistic quality, the study introduces narrative-conditioned harmonic structure as a complementary perspective for computational music analysis. The findings suggest that generative systems may benefit from modeling not only harmonic correctness, but also structural identity, context, and human compositional intention.
[AI-144] Dynamic Flow Static Graph: KV Cache Reuse for Efficient LLM Serving on Mobile NPUs EUROSYS2027
链接: https://arxiv.org/abs/2609.34727
作者: Zhengxiang Huang,Shengheng Chen,Chaoyue Niu,Yujie Sun,Zhaode Wang,Zeyu Zhao,Chengfei Lv,Fan Wu,Guihai Chen
类目: Operating Systems (cs.OS); Artificial Intelligence (cs.AI); Databases (cs.DB)
备注: Accepted at EuroSys 2027 (Spring Cycle)
Abstract:On-device large language model (LLM) serving is a cornerstone of local-first personal intelligence, offering users data sovereignty, strong privacy guarantees, and freedom from cloud API latency and cost. Although KV caching is widely used to reduce latency in long-context inference, existing designs were primarily optimized for cloud GPUs with dynamic execution environments and abundant memory bandwidth. These architectural assumptions do not hold on mobile NPUs, where computation graphs must be statically compiled and both memory capacity and I/O bandwidth are severely constrained. In this work, we present a compute-storage co-design for mobile-centric prefix and non-prefix KV reuse. We first propose an intra-graph mechanism that maps selective KV recomputation onto static NPU graphs, reconciling algorithmic dynamicity with NPU staticity. We further develop an inter-graph scheduler to optimize chunk merging and minimize padding with dynamic programming. To address mobile bandwidth limitations, we introduce a hierarchical KV manager featuring a tree-hash-semantic hybrid structure, along with cost-aware prefetching and eviction policies. We also build a two-dimensional pipeline that overlaps KV loading, rerotation, and storage with NPU execution, hiding data-movement latency. Experiments across representative on-device workloads and LLMs show that our design reduces time-to-first-token (TTFT) by 40-60% compared with no reuse and prefix-only caching.
[AI-145] PDE-JEPA: Predictive Representation Learning of Latent Dynamics Modeling for Parametric PDEs
链接: https://arxiv.org/abs/2609.34715
作者: Zhentao Tan,Jianrong Zhang,Ruijie Quan,Yi Yang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Physical trajectories contain more than snapshots of a system: they also reveal how its states evolve under governing conditions. However, representation learning for parametric partial differential equations (PDEs) has largely relied on reconstruction-based objectives that emphasize recovering observed physical fields. In this paper, we investigate predictive representation pretraining as an alternative to reconstruction-based learning. We find that predictive representations preserve rich physical information, yet this advantage alone does not ensure accurate field evolution. Based on these observations, we introduce PDE-JEPA for parametric PDE dynamics. Specifically, we first train an encoder using a masked-latent prediction to capture the underlying regularities of PDE dynamics. To explicitly adapt the pretrained representation toward a more dynamics-aligned state space, we then introduce a geometry projector that aligns latent trajectory geometry with the evolution geometry of physical fields. Finally, building on this geometry-aligned latent space, we further develop a physics-structured latent predictor that decomposes the dynamics into parameter-independent evolution and parameter-dependent response components. Extensive experiments on nine widely used PDE benchmarks demonstrate that our framework outperforms existing state-of-the-art methods by an average of 33.4% in-distribution, while achieving an average improvement of 51.4% when extrapolating to unseen governing parameters. The project page is available \hrefthis https URLhere.
[AI-146] RSI-Router: Evolving Subtask-Level LLM Routing and Skills for Cost-Efficient Agents
链接: https://arxiv.org/abs/2609.34712
作者: Hao Li,Hangfan Zhang,Zhiyao Cui,Chunjiang Mu,Yiqun Zhang,Bo Zhang,Danyang Jia,Shuyue Hu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Practical deployment of large language model (LLM) agents requires strong task performance at affordable inference cost. For long-horizon agentic tasks, this performance-cost trade-off can be improved through within-task large-small model collaboration, as smaller models can handle some stages even when they cannot solve the full task. In this paper, we introduce RSI-router, a routing framework that constructs subtask-level model assignments and model-specific skills through recursive self-improvement over accumulated experience. Each iteration consists of four stages: Subtask Mining derives subtask definitions and identification rules from training trajectories; Routing Strategy Evolution proposes and evaluates diverse model assignments; Model-Specific Skill Evolution compares routed and large-model-only trajectories to diagnose failures and develop reusable execution skills; and Pareto-Optimal Router Selection updates the Pareto population using historical and newly generated routers while retaining dominated routers as experience for subsequent evolution. Routing between DeepSeek-V4.1-Flash and Qwen3.5-9B, RSI-router consistently surpasses the DeepSeek-only baseline at roughly half the inference cost (48.3%) across five agentic benchmarks. In particular, on ALFWorld, ScienceWorld, and WebShop, it cuts inference cost by 74.7-82.2% while simultaneously improving performance; on Terminal-Bench 2.0, it achieves a 16.7% relative performance gain at 18.0% lower cost. Moreover, RSI-router establishes a stronger performance–cost Pareto frontier than 9 routing methods.
[AI-147] FromPitch2Board: Benchmarking LLM Agents in Long-Horizon Football Management
链接: https://arxiv.org/abs/2609.34710
作者: Peiyu Zang
类目: Artificial Intelligence (cs.AI)
备注: 28 pages, 4 figures. Code: this https URL
Abstract:Long-horizon agent benchmarks typically report how far an agent progresses, but do not identify whether its performance comes from the foundation model, scaffold, responsibility scope, match-control granularity, or horizon. We introduce FromPitch2Board, a deterministic football-management benchmark that studies five configurable factors through controlled comparisons on a single simulator, using paired seeds and a frozen calibration. We evaluate four foundation models and four agent scaffolds. In the Model Track, Coach points Z-scores span 0.19, while Manager points Z-scores span 0.68, with GPT-5.6 showing a sharp rise in passivity under responsibility expansion. Its responsibility ladder rises from 46.1 to 58.1 points with recruitment, then falls to 46.8 under full management, localizing the regression to the final responsibility boundary. Across that boundary, its skipped-decision rate rises from 1.1% to 57.9%. Within the Flash-Pro pair crossed across every scaffold, scaffold choice changes Manager points Z-scores by up to 0.48 relative to the fixed stateless scaffold. The 3Y cohort shows a directional reversal in mean ranking between years one and three, while a selected Claude Code+Pro configuration peaks in year three and remains below that peak, showing that responsibility scope and horizon expose behavior changes that a single headline score conceals.
[AI-148] ResonAct: Streaming Metrics for Runtime Diagnosis and Self-Healing in Multi-Agent Systems
链接: https://arxiv.org/abs/2609.34701
作者: Tarun Chintada,Neelamadhav Gantayat,Ishaan Romil,Renuka Sindhgatta,Soujanya Soni,Sameep Mehta
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Multi-agent systems (MAS) are increasingly used to automate enterprise workflows involving multiple specialized agents, external tools, and long-running task execution. Failures may arise from tool degradation, context propagation errors, coordination breakdowns, or repeated agent interactions that prevent task completion. While existing observability frameworks provide traces and logs, diagnosis and remediation are largely performed after execution completes, limiting opportunities for recovery during runtime. We present ResonAct, a runtime self-healing framework that enables continuous monitoring, diagnosis, and remediation of multi-agent systems through streaming operational metrics. ResonAct ingests execution traces, agent interactions, and tool invocations into a streaming analytics layer that continuously derives task progress, context health, and tool reliability metrics. These metrics serve as runtime control signals for detecting anomalous execution patterns and localizing root causes using a structured failure model. Based on the diagnosed failure, ResonAct dynamically selects remediation policies and performs actions. The framework operates as an external control plane, enabling intervention without modifying application agents or orchestration logic. We evaluate ResonAct across enterprise workflow scenarios and AppWorld benchmarks. The results show that the streaming metric-based analysis identifies execution degradations and localizes faults. Furthermore, policy-driven remediation improves task completion rates by up to 10.00 percentage points, with detection precision ranging from 70.59% to 82.91%, recall from 63.09% to 100%, recovery rates from 10.48% to 46.67%, and runtime overhead ranging from -0.25% to 14.12% across the evaluated configurations.
[AI-149] Sufficiency of Zeroth-Order Reward Shaping for Policy Gradient in Stabilization Control
链接: https://arxiv.org/abs/2609.34695
作者: Yisheng Zhang,Tao Wang,Sicun Gao
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 21 pages, 10 figures, including appendix
Abstract:Reward shaping is fundamental to modern robotic control with deep reinforcement learning (RL), yet practitioners still rely heavily on heuristic principles borrowed from classical optimal control and trajectory optimization. Existing methods rarely distinguish reward terms that are intrinsic to the control objective from numerical regularizers, leading to brittle hyperparameter tuning. To determine which quantities a reward must contain, we study the stabilization control problem with a focus on zeroth-order (configuration) and first-order (velocity) information. We theoretically and empirically demonstrate that policy gradient methods can successfully solve stabilization tasks without first-order reward terms, adding such terms can instead introduce severe sensitivity as their scale grows. Conversely, our findings confirm that reward functions must be zeroth-order complete over goal-relevant coordinates, while the first-order state remains necessary in the policy observation under our low-dissipation assumptions. Overall, these results provide actionable and principled guidance for reward design in robotic RL.
[AI-150] VCN-Bench: A Video-Contextualized Navigation Benchmark for Spatial Reasoning over Prior Visual Experience
链接: https://arxiv.org/abs/2609.34687
作者: Siqi Zhang,Meng Wei,Chenyang Wan,Shaohao Zhu,Shufan Shen,Xihui Liu,Zhihua Wei,Tai Wang,Jiangmiao Pang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Spatial reasoning is fundamental to embodied agents, yet it remains unclear whether spatial understanding can be carried forward to guide sequential interactions. Existing spatial-reasoning benchmarks typically terminate at offline predictions, while navigation benchmarks evaluate spatial reasoning as part of instruction following and exploration. We introduce VCN-Bench, a \textbfVideo-\textbfContextualized \textbfNavigation benchmark for probing closed-loop spatial reasoning over prior visual experience in MLLMs. Given a prior video covering both the initial location and destination, the agent is tasked with reasoning out the instruction-specified target and navigating toward it with the inferred spatial context. Built on Matterport3D, VCN-Bench contains five instruction types, 100k training episodes, and 1,250 evaluation episodes. Navigation serves as the primary evaluation, while diagnostic goal identification helps distinguish destination-resolution errors from subsequent navigation failures. We further propose MV-DualVLN, a planning-oriented baseline that jointly leverages prior video and in-episode observations. Experiments reveal limited navigation performance, a substantial destination-resolution-to-navigation gap, and frequent navigation failures even after correct destination identification.
[AI-151] Jailbreak Context Lingers: Divergent Safety Routing and Its Cross-Task Predictability in Tool Agents ICLR2027
链接: https://arxiv.org/abs/2609.34686
作者: Xi Wang,Songlei Jian,Yiming Zhang,Bin Ji,Zhaoye Li,Ma Jun,Baosheng Wang,Jie Yu
类目: Artificial Intelligence (cs.AI)
备注: 28 pages, 9 figures, 17 tabels, ICLR 2027 Under Review
Abstract:As large language models increasingly operate as tool-using agents, post-jailbreak safety feedback is often assumed to serve as a reliable safeguard; however, how lingering jailbreak context shapes subsequent agent behavior remains largely unexplored. To systematically examine this dynamic, we introduce a paired continuation framework across 192 parent tasks spanning 42 domains, evaluating 12,148 analyzed continuation pairs (curated from a 12,288-pair initially design) across eight diverse agents. We find that identical safety feedback induces sharply model-dependent behavioral routing rather than uniform protection: redirecting unsafe trajectories toward legitimate completion (\emphrescue), sustaining unauthorized execution (\emphpersistent unsafe), or triggering over-refusal on benign tasks (\emphcollateral loss). Through layer-wise activation patching, we discover a shared \emphlate-commit pattern where causal intervention effects surge sharply near the final layers (relative depths of 0.958–0.984) despite an over 30-fold variation in peak magnitude across architectures. Crucially, critical-layer representations correlate with macroscopic routing outcomes, and intervening at these layers causally alters concrete next-step tool actions. Building on this causal foundation, we test whether localized intervention-derived features can serve as predictive proxies for full-trajectory routing outcomes on unseen parent tasks under leave-one-parent-task-out evaluation, finding that they provide viable predictive signals in responsive agents with peak ROC AUCs reaching 0.675 for \emphrescue, 0.777 for \emphcollateral loss, and 0.702 for \emphpersistent unsafe. These findings establish a mechanistic lens and a predictive baseline for anticipating the safety and utility trade-offs of post-jailbreak feedback in autonomous agents.
[AI-152] From Preference to Reciprocity: Decentralized Matching with Empirically Grounded LLM -agent Based Modeling
链接: https://arxiv.org/abs/2609.34679
作者: Wangxuan Fan,Xiaoyu Nie,Zhoutian Shi,Xiangcheng Meng,Shipei Zeng,Pin Gao,Yan Hu,Zhongxiang Dai
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Bipartite matching is a fundamental problem in game theory and market design. Classical approaches such as Gale–Shapley assume complete preferences and centralized computation, whereas many real-world matching processes are decentralized, asynchronous, and shaped by sequential interaction under limited information. We propose a dynamic bipartite matching framework that combines large language model (LLM) agents with contextual bandits. In a simulated Chinese marriage market, economically grounded LLM agents evaluate locally encountered candidates, while agent-specific Logistic-UCB models learn reciprocal acceptance from realized proposal outcomes. The mechanism therefore separates two decisions—\emphwhom do I like? and \emphwho is likely to like me back?—without requiring ex ante market-wide preference rankings. We first validate LLM-induced mate preferences against the empirical conditional-logit reference across multiple LLM backbones. In the 50\times50 matching experiment, Bandit-UCB achieves the highest mean mutual welfare (56.01 versus 54.87 for Gale–Shapley), a smaller gender rank gap than the classical baselines, and the fewest blocking pairs among the LLM-ABM policies. Learned acceptance models show economically interpretable gender-differentiated associations, while counterfactual setups reveal no systematic unilateral advantage from prior search knowledge. Overall, these results support the advantages of decentralized matching with LLM-based behavioral modeling and online learning under incomplete information for economic simulation and computational social science research.
[AI-153] Codoku: Renewable Program-Reasoning Challenges for Frontier Coding Agents
链接: https://arxiv.org/abs/2609.34661
作者: Cong Li,Hao Sun,Zenan Li,Zhendong Su
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:Existing program-reasoning benchmarks ask large language models to predict a program’s behavior on a given input. Coding agents break two assumptions on which these benchmarks rest: an agent can recover the answer by executing the program instead of reasoning about it, and fixed task sets drawn from existing programs are increasingly exposed to contamination, yet costly to renew. We introduce Codoku (code sudoku), a renewable benchmark in which a solver fills typed cells in a partial program to satisfy global static and dynamic constraints, such as a prescribed control-flow graph and execution path. Because a partial program cannot be executed and valid fillings are sparse in an exponentially large space of interdependent choices, neither tool use nor enumeration can substitute for program reasoning. Puzzles are synthesized from scratch via semantic reification, so fresh puzzles of controllable complexity can be generated on demand, each with a witness that guarantees solvability. We evaluate five frontier models on 300 puzzles through a coding agent free to use any tool within a fixed budget. Small puzzles already challenge open-weight models, whereas even proprietary models solve only about half of the large ones. Codoku thus offers a renewable testbed for program reasoning that can keep pace with rapidly improving coding agents. GitHub: this https URL.
[AI-154] A General Harness for Protein Foundation Model Fitness Prediction
链接: https://arxiv.org/abs/2609.34654
作者: Yang Tan,Qijia Tian,Gangyu Sun,Bozitao Zhong,Mingchen Li,Yuanxi Yu,Nanqing Dong,Liang Hong
类目: Artificial Intelligence (cs.AI); Quantitative Methods (q-bio.QM)
备注:
Abstract:Accurate fitness prediction is central to protein engineering and understanding sequence-function relationships. With advances in deep learning, protein foundation models (PFMs) have become widely used for this task. Recent analyses, however, show that these models share preferences reflecting their training corpora, while unreliable inputs can further distort fitness predictions. Family-specific evolutionary evidence and structural context can help address these limitations by providing complementary constraints on model scores, motivating VenusREM-Harness (VRH), a general, model-agnostic, training-free Retrieval-Enhanced Mutation harness. It fuses frozen model scores with multiple sequence alignment (MSA) evidence according to model uncertainty, then applies gated background correction and score shrinkage based on structural confidence and solvent exposure. Across 1,211 assays and 3.1 million measured variants from ProteinGym, VenusMutHub, and the newly curated viral benchmark VenusViroHub, all 71 configurations improve Spearman correlation on all 3 benchmarks by 0.073 on average, with broad gains across 5 metrics. Extended analyses relate retrieval gains to model-MSA preference differences, assess domain-level gains and immune-escape cases, and quantify computational speedups. Built with VRH, VenusREM2 is the first to rank highest in all function, taxon, MSA-depth, and mutation-depth categories, with a ProteinGym Average Spearman of 0.556, 0.038 above the prior best.
[AI-155] OmniTide: Co-Designing Algorithms and Systems for Efficient On-Device Omni-LLM Streaming
链接: https://arxiv.org/abs/2609.34653
作者: Zongshang Shen,Wangsong Yin,Daliang Xu,Mengwei Xu,Xuanzhe Liu
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:On-device streaming omni-modal inference safeguards user privacy and eliminates prohibitive per-token API costs, but faces a critical bottleneck: the continuous influx of multimodal data rapidly exhausts constrained memory and compute budgets via monotonic KV cache growth. Existing sparse attention methods fall short, either incurring prohibitive online estimation latency or destroying interleaved cross-modal context, while failing to resolve physical memory fragmentation. We present OmniTide, the first algorithm-system co-design tailored for efficient on-device streaming omni-modal inference. Driven by the observation of modality-aware structural sparsity, OmniTide adopts a unit-based abstraction with two components: (1) At the algorithm level, OmniPick logically retains critical multimodal context based on unit boundaries and modality importance to preserve task accuracy; (2) At the system level, OmniPage physically partitions the cache by retention likelihood and dynamically compacts surviving sparse tokens, minimizing both memory fragmentation and data-movement overhead. Extensive evaluations across three streaming benchmarks and two consumer-device architectures show that OmniTide achieves up to 12.72\times kernel speedups and 2.40\times lower stream-loop latency. On StreamingBench, it improves accuracy by up to 18.0 percentage points over sliding-window baselines at comparable session cost. OmniPage further reduces the physical KV span by up to 26.7% relative to native logical eviction, unlocking real-time, infinite-context streaming on edge devices.
[AI-156] Beyond Skill Evolution: Self-Evolving Context Management Policies for Long-Horizon Agent Harnesses
链接: https://arxiv.org/abs/2609.34649
作者: Weiyuan Li,Jinghan Xu,Aili Chen,Xintao Wang,Shuang Liang,Jiaqing Liang,Deqing Yang
类目: Artificial Intelligence (cs.AI)
备注: 27 pages, 6 figures
Abstract:Harness evolution improves LLM agents by learning from execution trajectories, but existing experience- and skill-based methods are less effective on long-horizon tasks. As interactions grow, useful evidence can be buried by redundant or outdated context, making context management itself a key bottleneck. We introduce ContextEvo, a framework that learns a context policy from long-horizon trajectories. ContextEvo reconstructs the model-visible context at key decision points, identifies context-related failures, and applies targeted policy updates. Starting from the open-source Pi-agent harness, ContextEvo improves performance across three long-horizon task benchmarks, achieving results comparable to or better than several prominent agent harnesses, including Codex, OpenCode, and OpenClaw. Additional analyses show that fixed or locally evolved context strategies can fall short under long-horizon information pressure, while our methods adapt to the information demands of each environment.
[AI-157] SEmoEdit: Probing and Harnessing the Editability of Pre-trained Speech Flows
链接: https://arxiv.org/abs/2609.34648
作者: Tianxin Xie,Pengfei Zhang,Kai Jiang,Zelin Zhao,Li Liu
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注: 25 pages, 12 figures, 17 tables
Abstract:Existing training-based speech emotion editing methods often require substantial task-specific training and can be unstable. This motivates us to investigate whether the pretrained generative dynamics of large-scale text-to-speech (TTS) models can be directly manipulated for training-free emotion editing. To answer this question, we probe the editability of pretrained flow-matching and hybrid TTS models by constructing a controlled test set and systematically diagnosing editing effects along the generative trajectory. Our analysis reveals that pretrained TTS models are substantially editable in emotion, but such editability is architecture- and trajectory-dependent and can be disrupted by early flow-matching steps, while cross-speaker emotion transport carries additional acoustic attributes beyond emotion. To address these limitations, we propose SEmoEdit, the first training-free framework that formulates emotion editing as dynamic velocity transport between source and target emotions, enabling robust, flow-based speech emotion editing directly within pretrained TTS models. SEmoEdit unifies three core operations: emotion replacement, emotion erasure, and continuous emotion interpolation, requiring neither parameter updates nor task-specific optimization. To systematically evaluate these capabilities, we introduce SEmoEditBench, a dataset comprising 600 editing cases, and conduct extensive experiments across state-of-the-art (SOTA) models and backbones. Our results show that SEmoEdit is highly effective and broadly applicable, outperforming existing training-based and activation-steering methods. Ultimately, this work reveals that pretrained speech flows possess rich, latent emotion-editing capabilities, providing useful guidance for real applications. Code, benchmark, and Audio samples are available at this https URL.
[AI-158] Nereus: Adaptive Parallelism for LLM Post-Training
链接: https://arxiv.org/abs/2609.34645
作者: Songlin Jiang,Tuo Shi,Sitong Zhang,Zeke Wang,Mario Di Francesco,Bo Zhao
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning (RL) post-training for large language models (LLMs) coordinates multiple models across generation, inference, and training on GPU clusters. Several factors may change during a run, including resource availability, sequence length, memory pressure, and stage bottlenecks. As a consequence, an execution plan that was initially suitable can then become slow or even infeasible over time. However, adapting a job whose models share GPUs entails significant challenges: deciding whether a new plan is worth the transition cost, reusing the job’s distributed state, and coordinating GPU transfers across models and stages. Nereus targets these challenges as a cost-aware runtime that adapts RL post-training jobs into efficient execution plans. Its low-overhead controller selects a memory-feasible global plan and admits the transition using a cost model calibrated against the running job. To estimate and execute a transition, Nereus represents the distributed state of each replica of a model-stage (one model in one stage) as an Elastic Model Unit. It then employs a global transition graph to order the transformations and GPU transfers of these units. In a trace built from real data, online TP/PP adaptation reduces average step latency by 27.7% relative to the initial fixed TP/PP layout with DP scaling. In a 1,000-step run reaching 1,024 GPUs, six transitions consume 0.079% of total run time. Nereus improves end-to-end 8B PPO throughput by 2.14–7.27 \times over OpenRLHF and by 1.10–1.47 \times over Verl across diverse clusters. Subjects: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI) ACMclasses: C.2.4; C.1.4; I.2.6 Cite as: arXiv:2609.34645 [cs.DC] (or arXiv:2609.34645v1 [cs.DC] for this version) https://doi.org/10.48550/arXiv.2609.34645 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-159] MechReason er: A Simulator and Benchmark for Mechanistic Reasoning in Qualitative Physics
链接: https://arxiv.org/abs/2609.34636
作者: Danilo Gusicuma,André Freitas
类目: Artificial Intelligence (cs.AI)
备注: 20 pages, 4 figures, 11 tables, and 2 algorithms; includes supplementary material. Code and benchmark: this https URL
Abstract:This work introduces MechReasoner, a mechanistic qualitative simulator grounded in confluence-based qualitative physics, together with a benchmark for mechanistic inference. Current large language models (LLMs) generate fluent mechanistic descriptions that do not reliably follow from underlying structural and causal constraints. The benchmark tests whether answers preserve simulator-licensed ambiguity, quantified claims, episode-graph transition evidence, repairs, and trace-support judgments. Its 1,120 items are generated deterministically from admissible interpretation sets, component states, scenario restrictions, confluence constraints, and derivation steps across 18 catalog mechanisms and six task families. Each mechanism undergoes converter checks of structure and topology and behavioral checks against quantitative simulations. GPT-5.5 accuracy decreases as family-specific mechanistic complexity increases, from 76.1% in the lowest-complexity bucket (B1) to 38.0% in the highest-complexity bucket (B4). The negative association remains after controls for rendered-prompt and expected-answer length. These results show that qualitative simulators can support auditable NLP benchmarks for mechanistic inference.
[AI-160] A Persistent State for Auditable Mixture-of-Experts Routing
链接: https://arxiv.org/abs/2609.34634
作者: Abdurrahman Javat,Allan Kazakov
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Mixture-of-Experts (MoE) models repeatedly route tokens to sparse subsets of experts, but conventional routers expose no routing-specific record of how cross-layer influences accumulate. We introduce Scratchpad-Augmented Mixture-of-Experts (SA-MoE), which gives each router access to a low-dimensional persistent state that is not provided to the experts. Learned layerwise writes update this state, and their realized post-update changes exactly decompose the state-mediated contribution to any later routing margin, forming a routing ledger. Across sparsely upcycled SmolLM2- and Gemma-based models and three independent training seeds per architecture, this pathway adds less than 1% analytical forward compute and is strongly used by trained routers: local removal of its router contribution changes the selected Top-2 expert set in 87.6% and 69.9% of decisions, respectively. Relative to a matched latest-write-only control, persistent accumulation increases long-horizon future-routing accessibility by 19.4 and 12.2 percentage points, with positive effects in every seed. More than 90% of absolute ledger contribution comes from non-recent writes in both families, and full-forward suppression of ledger-selected writes changes later routing and output distributions. The ledger is an exact provenance object for the persistent-state pathway, not a complete causal explanation of routing. Sensitivity-aware scores better predict full-forward intervention effects, and post-hoc methods recover related cross-layer attribution without architectural modification. SA-MoE instead makes one routing-specific computational history explicit and directly inspectable within the model’s natural forward computation.
[AI-161] LLM s for Executable Multi-Agent System Specification Generation
链接: https://arxiv.org/abs/2609.34619
作者: Andreas Kouvaras,Periklis Mantenoglou,Alexander Artikis
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:MAS specifications express the effects of the actions of the agents and their environment, as well as other temporal phenomena, such as the intervals during which an agent may perform an action. The specification of a MAS should also be executable in order to allow for run-time monitoring. Constructing the specification of a MAS requires formal language expertise, while machine learning techniques depend on labelled data which are rarely available. To address these issues, we propose genRTEC', a method that leverages pre-trained Large Language Models (LLMs) to generate executable MAS specifications, in the language of the Run-Time Event Calculus’ (RTEC), from natural language descriptions. genRTEC constructs MAS specifications with complex hierarchical and cyclic dependencies based only on short natural language descriptions of the concepts involved. We present an extensive empirical evaluation of genRTEC, spanning various MAS specifications, including both a qualitative and a quantitative assessment. Our results demonstrate that genRTEC constructs executable MAS specifications of high predictive accuracy without compromising reasoning efficiency.
[AI-162] Efficient World Action Model Inference with Adaptive Intermediate States
链接: https://arxiv.org/abs/2609.34608
作者: Zhinnan Liu,Haozhi Han,Ruge Zhang,Teng Ma,Tao Ma,Zheng Liu,Yifeng Chen,Yunquan Zhang,Ting Cao,Yunxin Liu,Kun Li
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:World Action Models (WAMs) enable future-aware control by jointly modeling actions and environment dynamics. However, iterative diffusion or flow inference incurs substantial denoising latency. Prior inference state offers a natural opportunity for acceleration, yet changing planning contexts, observations, and intermediate representations can quickly render retained state stale. Preserving useful computation therefore requires adapting inference state rather than reusing it as-is. To this end, we present \mathrmWAM\scriptstyle\mathrmACHINE , a training-free framework that accelerates WAM inference by preserving and adapting inference state for efficient and accurate continuation as the control loop evolves. Across closed-loop replans, Trajectory Remapping remaps replan state from the preceding replan to initialize the next replan, reducing redundant trajectory generation. Across denoising steps, Observation Rebinding performs anticipatory inference during action execution and rebinds retained denoising state to the real observation for continuation when consistency checks pass, reducing latency exposed to the control loop. Across Transformer layers, Residual Rescaling selectively rescales retained layer state and refreshes it through full computation of the middle layers when probe checks fail, reducing repeated Transformer computation. Evaluations of three representative WAM architectures on LIBERO and RoboTwin 2.0 show that \mathrmWAM\scriptstyle\mathrmACHINE achieves 1.47-3.05 \times speedups in observation-to-action latency and 2.23-3.27 \times speedups in GPU inference time per replan, while preserving 96.69-99.54% of native WAM task success.
[AI-163] ULIP: Targeted LLM Unlearning at Layers Identified Per-Input
链接: https://arxiv.org/abs/2609.34591
作者: Yejin Kim,William F. Shen,Seokwon Jung,Daeun Park,Seong Joon Oh
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Representation-level unlearning intervenes on the intermediate hidden states of LLMs. Although knowledge is distributed across layers, existing methods operate at a single fixed layer for the entire forget set. We ask whether such a fixed layer is sufficient. To answer this, we design a hijacking experiment that grafts hidden states of the target model into an oracle trained only on the retain set. The oracle cannot produce the forget answer on its own, yet it produces the answer from the grafted state. Thus, the answer is formed at an intermediate layer and merely read out afterward, so unlearning should focus on formation, not readout. Moreover, the layer where formation ends varies widely across inputs. Motivated by these findings, we propose Targeted Unlearning at Layers Identified Per-input (TULIP). For each input, TULIP uses the logit lens to locate the formation-readout boundary and removes the hidden state’s alignment with the forget answer’s unembedding vector there. TULIP consistently outperforms output- and representation-level baselines on TOFU, PISTOL, and WMDP across Llama, Qwen, and Zephyr models. It also remains robust to paraphrase and quantization attacks. Beyond standalone use, its per-input layer selection serves as a plug-and-play component that further improves existing methods.
[AI-164] SpeechCritic: Learning a Diagnostic Speech Judge from Limited Human Preferences
链接: https://arxiv.org/abs/2609.34582
作者: Mingyue Huo,Shivam Mehta,Bhavin Jawade,Yinghong Lan,Haoqi Li
类目: Artificial Intelligence (cs.AI); Sound (cs.SD)
备注:
Abstract:Human speech conveys rich perceptual information, such as emotion and speaker identity, yet most automatic speech quality judges reduce it to a single naturalness score. We study diagnostic speech judges: given two candidates, a diagnostic judge decides which is better, along which perceptual dimensions (e.g., timbre, emotion, timing) they differ, and which audible cues support its decision. Learning such judges is challenging: expert annotation is costly, and simply prompting a frontier audio-language model to produce labels is unreliable: our probing reveals substantial errors and unstable instruction following. We introduce SpeechCritic, which learns a diagnostic judge in a reference-conditioned cross-lingual setting from only about 300 human-labeled comparisons. Rather than replacing the frontier model, SpeechCritic calibrates it with these labels: for each dimension, it selects the acoustic measurements that agree with human judgments, maps them to A/Tie/B probabilities, and passes these to the model as non-binding hints alongside the audio. Compared with the same model labeling without hints, this raises dimension-level agreement with humans by 6.3 points and cuts the mismatch with human Tie rates by 10.4 points. We then train a 7B judge on this supervision and find that different training signals shape different judge behaviors: SFT establishes the task, OPD transfers the teacher’s dimension-level strengths and weaknesses, and RL helps most on clear-cut comparisons where human raters agree. Notably, human listeners also find that RL makes rationales cite more specific, localized acoustic cues, although it never directly rewards rationale text. Finally, we show that the pipeline is language-pair agnostic by instantiating it on both English-Japanese and English-Spanish. Together, these results demonstrate a path from limited human preferences to a diagnostic speech judge.
[AI-165] Diffusion Subgoal Planning for Long-Horizon Offline Goal-Conditioned Reinforcement Learning
链接: https://arxiv.org/abs/2609.34575
作者: Hengrui Zhang,Yuhu Cheng,C. L. Philip Chen,Xuesong Wang
类目: Artificial Intelligence (cs.AI)
备注: 31 pages, 9 figures
Abstract:Offline goal-conditioned reinforcement learning (GCRL) learns goal-directed policies from reward-free data, but in long-horizon tasks, goal-conditioned value functions often provide unstable guidance due to sparse rewards and discounting. Hierarchical methods partially mitigate this issue via subgoal decomposition; however, high-level decision-making still relies on noise-sensitive value estimates, leading to unstable behavior in complex environments. We address this limitation by proposing \textbfDiffusion \textbfSubgoal \textbfPlanning (\textbfDSP), a diffusion-based framework for high-level subgoal generation. DSP casts high-level planning as guided generative inference over goal-conditioned subgoals and learns both conditional and unconditional flows, enabling classifier-free guidance to introduce a goal-directed bias at inference time. By removing explicit value-based guidance from high-level planning, DSP generates reachable and goal-directed subgoals through a generative model while retaining hierarchical execution. Experiments on offline GCRL benchmarks demonstrate that DSP outperforms prior methods on a range of navigation and manipulation tasks, with particularly strong performance in maze environments that require multi-step subgoal planning.
[AI-166] PersonaManifold: Revealing and Exploiting Curved Geometry in LLM Persona Representations NEURIPS2026
链接: https://arxiv.org/abs/2609.34571
作者: Rui Xu,Yinghui Xu,Libo Wu
类目: Artificial Intelligence (cs.AI)
备注: Accepted to NeurIPS 2026
Abstract:Controlling persona in large language models (LLMs) at inference time is important for role-playing, personalized dialogue, and social simulation. Recent methods extract persona vectors from the model’s activation space and apply Euclidean operations—addition, scaling, and linear interpolation—under the linear representation hypothesis. However, these methods themselves report systematic failures: non-orthogonal trait dimensions, asymmetric ceiling and resistance effects, and significant deviations in multi-trait composition, suggesting that the linear isotropic assumption does not hold. We propose PersonaManifold, a framework that models persona representations as points on a curved, low-dimensional Riemannian submanifold in activation space. We estimate the manifold’s intrinsic geometry—local metric tensors, geodesic distances, and Ollivier-Ricci curvature—and introduce geodesic steering, which interpolates between personas along manifold geodesics rather than Euclidean straight lines. We also propose the Behavioral Similarity Triplet (BST) benchmark, which automatically generates situational questions grounded in six established psychological constructs and defines persona similarity through behavioral responses rather than self-report questionnaires. Experiments on three open-source LLMs show that persona activations form a manifold with heterogeneous curvature, geodesic distance predicts behavioral similarity more accurately than Euclidean alternatives with independent contributions from anisotropy and curvature, and geodesic steering produces more coherent intermediate personas on both our BST benchmark and external evaluations, with the advantage concentrated in high-deviation regions where the manifold deviates most from flatness.
[AI-167] FlowState: Execution State as Memory for Long-Horizon LLM Agents
链接: https://arxiv.org/abs/2609.34565
作者: Minghao Li,Bangyan Li,Zifan Wang,Yulong Li,Hu Xu,Gan Zhang,Jingtong Wu,Wenqiang Xu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Long-horizon tasks require LLM agents to continually draw on information from earlier interactions. However, retaining the full history increases context costs, while compressing it risks losing details needed later, and the relevance of historical information often becomes apparent as the task progresses. To address these challenges, we propose FlowState, which treats execution state as memory that can be retained and revisited across requests, unifying current decision-making with the reuse of historical information. FlowState preserves semantically typed state nodes, their relations, and references to raw tool observations, separating persistent retention from on-demand access. Within a single execution loop, Incremental State Update (ISU) maintains the current state based on new inputs and feedback, while Progressive State Access (PSA) progressively reveals historical states and supporting evidence as needed during reasoning. Together, these mechanisms enable agents to reassess prior decisions in light of new information and guide subsequent actions. Compared with a full-context baseline using the same DeepSeek-V4-Flash model, FlowState improves the average success rate on MemoryArena and the average pass rate on \tau^3 -Bench by 4.55 and 13.95 percentage points, respectively, while reducing total token consumption by 43.2% and 40.6%. These results demonstrate the performance and efficiency advantages of FlowState on long-horizon tasks.
[AI-168] SkillRubric: Co-Evolving Actor Guidance and Evaluator Rubrics for Multimodal Agents
链接: https://arxiv.org/abs/2609.34557
作者: Bingqing Jiang,Guoxi Zhang,Jasper Wang,Auric Wang,Bingning Wang,Tianyi Lin,Zichao Yu,Yujin Han,Ziye Ma,Difan Zou
类目: Artificial Intelligence (cs.AI)
备注: 47 pages
Abstract:Recent work incorporates reusable skills distilled from past interactions into multimodal agent training, providing procedural guidance for long-horizon planning and tool use. However, policy optimization in these methods remains driven primarily by sparse outcome rewards, providing little supervision for intermediate decisions. Rubric-based rewards address this limitation through explicit intermediate criteria, but reliable rubrics are difficult to construct at scale and often disconnected from the procedure followed by the actor. We observe that a well-structured skill naturally specifies both how to act and what successful execution should achieve. Based on this insight, we introduce SkillRubric, which represents each skill through aligned actor-facing guidance and an evaluator-facing rubric. A multimodal verifier evaluates skill-defined goals using screenshots and tool outputs, assigning completion and progress rewards to the responsible turns. We further introduce an alternating co-evolution scheme that validates guidance revisions through paired rollouts under a frozen policy and rubric revisions offline under fixed guidance. Experiments across diverse multimodal agent benchmarks demonstrate consistent performance gains, while controlled paired rollouts further show that evolved skills provide more effective guidance for planning and tool use than their preceding versions.
[AI-169] SGG-ReflAct: Sub-Goal Guided ReflAct with Structured Planning for Reliable Long-Horizon Reasoning
链接: https://arxiv.org/abs/2609.34548
作者: Jaeho Jung,Sung Hoon Jung
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances in reasoning backbones have empowered large language model (LLM)agentstotackle complex, multi-step tasks. However, as reasoning horizons grow, inconsistent internal beliefs induce intermediate errors that cause agents to drift from their goals. This limitation also persists in REFLACT, which reflects only on the end-goal at each step without explicitly considering intermediate sub goals. To address this problem, we propose SGG-ReflAct (Sub-Goal Guided Re flAct), a reasoning backbone that integrates sub-goals generated through a single path LLM planner into the reflection process. We further extend this framework to BeamSGG-ReflAct, which replaces the single-path planner with a beam search based LLM planner for structured plan exploration. We run experiments on ALF World, ScienceWorld, and Jericho with multiple LLM models. SGG-ReflAct out performs REFLACT in nearly all settings, achieving best success rate gains of 14.9 percentage points on ALFWorld and 8.0 percentage points on ScienceWorld with Llama-3.1-8B-Instruct. Our experimental analysis shows that SGG-ReflAct re duces hallucinated actions and achieves its largest gains on procedurally ordered tasks. Furthermore, experimental results with BeamSGG-ReflAct show that the backbone’s effectiveness depends on plan quality: explicitly specifying the re quired operations recovers gains that plan searching alone cannot achieve. These results demonstrate that SGG-ReflAct offers a practical and highly effective rea soning backbone, enabling LLM agents to achieve reliable performance in com plex, long-horizon tasks through easy integration.
[AI-170] Remember Before Youre Asked: MemDream for Self-Probing Memory Evolution
链接: https://arxiv.org/abs/2609.34545
作者: Mingfei Lu,Mengjia Wu,Runsong Jia,Zhe Luo,Yi Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Memory is essential for enabling LLM-based agents to maintain coherent, personalized behavior over long-horizon interactions. However, existing memory systems share a fundamental limitation: they never proactively test their own memory, repairing it only after real queries expose weaknesses. This reactive paradigm means every retrieval failure corresponds to a real interaction in which the cost has already been paid. We propose MemDream, a framework that enables self-probing memory evolution for LLM agents. Our framework periodically enters offline dream cycles where three specialized agents (Dreamer, Analyst, Consolidator) collaboratively probe, diagnose, and repair the memory graph before failures occur. A policy trained via Group Relative Policy Optimization learns which repair operations produce durable retrieval improvements, while a soft decay mechanism provides reversible forgetting driven by the same anticipatory signal. Experiments on LoCoMo and MemoryAgentBench demonstrate that MemDream improves answer F1 by 4.5 points on LoCoMo and achieves a 9.1-point higher overall score on MAB over the strongest reactive-evolution baselines.
[AI-171] APOLO: Automatic Prompt Optimization for Ontology Learning ISWC2026
链接: https://arxiv.org/abs/2609.34540
作者: Huu Tan Mai,Roman Kochnev,Cuong Xuan Chu,Lukas Lange,Heiko Paulheim,Daria Stepanova
类目: Artificial Intelligence (cs.AI)
备注: Accepted at the Posters and Demos Track of the International Semantic Web Conference (ISWC 2026)
Abstract:Ontology Learning (OL) from text has advanced with the emergence of Large Language Models (LLMs), but it remains challenging due to the limited availability of annotated training data and the difficulty of adapting LLMs to perform OL effectively. We address this via APOLO - Automatic Prompt Optimization for Ontology Learning, by casting OL as an explicit prompt optimization problem over LLM modules. To obtain training data, we employ a multi-agent system that generates text-ontology pairs from existing expert-curated ontologies. We then propose two ontology learner architectures: a greedy and an autoregressive learner, and optimize both using GEPA, a greedy evolutionary prompt optimizer built on DSPy. Experiments on two ontologies - a biomedical (DOID) and a plant ontology (PO) show consistent improvements after optimization across nearly all model and mode combinations, with autoregressive learners achieving the largest gains. Our results demonstrate that prompt optimization is a viable and lightweight alternative to fine-tuning for OL, and that the autoregressive formulation better captures ontological structure than the greedy approach.
[AI-172] Shallow Queries Mature Values: Depth-Asynchronous Self-Speculation for Looped Transformers
链接: https://arxiv.org/abs/2609.34538
作者: Guanghao Li,Zihan Su,Hao Yu,Jinyang Jiang,Tao Ren,Zehao Li,Feng Lu,Ming Tang,Chun Yuan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Looped Transformers reuse a shared block across recurrent depths, making autoregressive decoding expensive because every generated token requires many sequential recurrent passes. Self-speculative decoders reduce this cost by drafting at an early depth and verifying at full depth, but typically bind draft computation to prefix representations from the same recurrent depth. We find that queries and keys approach their final-depth representations earlier than values, and controlled prefix-channel interventions show that mature values substantially improve shallow draft predictions. Motivated by this asymmetry, we introduce Depth-Asynchronous Self-Speculation (DAS), which decouples the depth of draft computation from the depth of verified-prefix representations it reads. Its Mature-V primitive lets shallow queries retrieve full-depth prefix values without additional recurrent computation. We further develop DAS-Wave, which combines depth-asynchronous prefix reads with carried parallel refinement, progressive block growth, and an independent full-depth verifier. Across four recurrent-model checkpoints and mathematics and code workloads, DAS-Wave achieves 4.00–6.96 \times mean throughput speedup over paired full-depth autoregressive decoding in the same inference stack. These results identify prefix-information depth as an effective design axis for recurrent self-speculation.
[AI-173] he Marathon of Scientific Reasoning : Robustness of Scientific Agents to Perturbations in Multi-Turn Interactions
链接: https://arxiv.org/abs/2609.34537
作者: Xiaoting Lyu,Xinbo Ma,Yufei Han,Hangwei Qian,Ziyang Lin,Bin Wang,Bin Wang,Wei Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM)-based scientific agents are increasingly used for scientific problem solving, yet their robustness to imperfections arising during multi-turn interactions remains poorly understood. We introduce \textscSciARP (\textbfScientific \textbfAgent \textbfRobustness to \textbfPerturbations), a benchmark for evaluating scientific agents under scientifically plausible perturbations throughout multi-turn problem solving. \textscSciARP transforms 620 scientific problems into interdependent tasks of 3–13 turns and defines 13 perturbation types spanning problem understanding, evidence processing, reasoning, and conclusion formation. Clean and perturbed versions of each task are independently executed under matched settings, producing paired live trajectories for evaluating both task success and process reliability. Experiments across eight LLMs from four model families reveal three key robustness characteristics. First, different classes of scientific perturbations exhibit distinct robustness profiles and can decouple task progression from scientific reliability: agents may continue advancing through the task even after their information or reasoning has become unreliable. Second, stronger clean-task performance does not necessarily translate into stronger robustness, as models with higher clean-task accuracy can exhibit larger degradation under perturbation. Third, perturbation effects exhibit strong temporal dynamics: they may remain latent for multiple turns before emerging and subsequently propagate through downstream dependencies. Together, these findings show that current scientific agents remain insufficiently robust to scientifically plausible perturbations, with failures often remaining undetected, propagating, and resisting recovery.
[AI-174] PairPref: When Should Memory Guide the Answer? A Benchmark for Contextual Preference Use
链接: https://arxiv.org/abs/2609.34526
作者: Mingfei Lu,Mengjia Wu,Yi Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Memory-augmented assistants use retrieved preferences to guide their responses. A small change in the situation can change whether a preference is appropriate while barely affecting its retrieval similarity. Memory benchmarks typically test whether systems store and retrieve preferences, with less attention to when those preferences should apply. We introduce PairPref, a benchmark of contextual preference use. Each pair changes only the situation, keeping the preference, request, and four candidate replies fixed. The preference remains valid in both situations. In the selection track, models must choose the reply that applies the preference only where appropriate. In the free-generation track, they must decide when to apply it without seeing candidate replies. Both tracks use the same 1,227 pairs across 45 preferences and eight situation categories. We evaluate eight models, most of which achieve selection scores ( \Delta ) of 51 to 65 points. In free generation, however, both responses are appropriate for their respective situations in only 3.6% to 18.3% of pairs. Models continue to apply the preference in both situations even with fewer retrieved memories, alternative presentation formats, and a stricter prompt. These results show that models still struggle to judge when user preferences apply and respond accordingly.
[AI-175] EOPSA: Efficient On-Policy Self-Distilled Safety Alignment
链接: https://arxiv.org/abs/2609.34519
作者: Qirui Liu,Yichen Sun,Yan Wang,Yu Mi,Wei Cao,Yue Shen,Zhixuan Chu,Kui Ren
类目: Artificial Intelligence (cs.AI)
备注: 32 pages, 9 figures. Code and models are available at the project repositories
Abstract:On-Policy Self-Distillation (OPSD) has emerged as a promising paradigm for safety alignment, delivering dense, token-level supervision by distilling from a teacher conditioned on refusal-oriented privileged prompts. However, we reveal that this paradigm suffers from critical inefficiencies that degrade both training efficiency and general reasoning capabilities. Specifically, we diagnose two fundamental bottlenecks: (1) supervisory collapse over extended rollouts, where the teacher’s corrective efficacy degrades precipitously as the student’s generation prefix lengthens, injecting noisy gradients into late-stage tokens; and (2) gradient dilution from stylistic shifts, where the distillation objective is dominated by safety-irrelevant stylistic discrepancies induced by privileged prompting, washing out genuine safety signals and impairing base reasoning. To resolve these issues, we propose Efficient On-Policy Self-Distilled Safety Alignment (EOPSA), which concentrates computational and gradient budgets exclusively on reliably supervised, safety-critical tokens. EOPSA incorporates two coordinated mechanisms: (i) Adaptive Rollout Scheduling, which dynamically bounds the generation horizon guided by a novel Teacher Rescue Rate (TRR) metric to operate strictly within reliable supervision regimes; and (ii) Selective Distillation, which filters out safety-neutral tokens to restrict gradient updates exclusively to safety-pivotal transitions. Extensive evaluations across reasoning models up to 32B parameters demonstrate that EOPSA slashes rollout computation by \sim 50% and backpropagates through merely \sim 2% of tokens, consistently outperforming full-token distillation baselines in both safety compliance and reasoning retention.
[AI-176] SEAD: A State-Based Perspective on Attack and Defense in Tool-Using Agents
链接: https://arxiv.org/abs/2609.34518
作者: Xinjie Shen,Junran Wang,Rongzhe Wei,Pan Li
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: Project Website: this https URL
Abstract:Language-model agents increasingly use tools to act on external systems. Earlier actions can alter files, permissions, database records, or other state, making a later routine-looking action harmful. Yet the visible interaction may not reveal the underlying state needed to assess that action. We formulate attack and defense as partially observed state control in SEAD, deriving their design requirements from this shared execution process. Because attackers supply instructions while the target chooses concrete actions, DART decomposes harmful goals into locally plausible steps and uses feedback from actual tool execution to guide trajectory search. The defender must decide before execution with incomplete state evidence. SAGE can therefore investigate relevant state through read-only queries before allowing or blocking each action, including those proposed after a block. We construct an environment-verifiable dataset integrating controlled initial states, replayable tool environments, and task-specific executable checks. Across four target models, DART improves semantic attack success by 18.8–35.9 percentage points over the competing baseline, with consistent gains under executable verification. On recorded trajectories, SAGE preserves 95.79% of benign trajectories while intercepting 92.73% of harmful paths by the harm-enabling boundary. In online attack-defense evaluation, it reduces DART’s executable attack success from 48.0% to 4.0%. SAGE remains effective across four attack methods and generalizes to out-of-domain environments. Our code and data is available at this https URL.
[AI-177] Can AI Make Money in Crypto? Measuring the Gap from Backtests to Real Markets
链接: https://arxiv.org/abs/2609.34510
作者: Xingtong Yu,Jiarun Zhou,Guanlin Ding,Wenkang Wei,Jiarui Liu,Chang Zhou,Fangzhou Ge,Chenyi Xu,Xikun Zhang,Renqiang Luo,Jie Zhang,Hong Cheng,Xinming Zhang,Hui Zhang,Yuan Fang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:AI-based trading methods have rapidly evolved from machine learning and reinforcement learning to large language models (LLMs) and trading agents, yet their performance is still predominantly assessed through historical backtesting. Such evaluations provide limited evidence of whether a method can generalize to unseen future markets or whether its backtested performance can be sustained in realistic trading frictions (e.g., latency, slippage, liquidity constraints, and market impact). We present a unified benchmark that evaluates representative machine learning, reinforcement learning, LLM-based, and agent-based trading methods in cryptocurrency markets through three progressively more realistic stages: historical backtesting, prospective exchange-based paper trading, and real-money live trading. These stages jointly increase temporal realism by moving from historical to unseen future markets, and execution realism by moving from offline simulation toward live trading. This protocol enables us to quantify the backtest-to-realization gap, identify when performance begins to deteriorate, and compare how this gap differs across major classes of AI trading methods. We further provide a unified open-source system supporting all three evaluation stages, together with a public platform that continuously updates benchmark results. Code is available at this https URL.
[AI-178] Does Model Uncertainty Track Human Ambiguity? Evidence from Multi-Annotator Vision Benchmarks
链接: https://arxiv.org/abs/2609.34506
作者: Manya Singh,Arjun Pakrashi
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Human-model alignment is critical for trustworthy AI-assisted decision-making systems. Yet, most work evaluates model predictions against single ground-truth labels, overlooking that humans themselves often disagree on labels, a signal of genuine ambiguity. We investigate whether models struggle on the same instances that humans find difficult. We measure this on two vision datasets (FER+ and CIFAR-10H) where multiple human annotations per image capture human disagreement patterns. We evaluate eight pretrained models across three architectures (ResNet, EfficientNet, MobileNetV3) in two parts: first, whether model uncertainty (softmax confidence, entropy) correlates with human disagreement, and second, whether predictive multiplicity measures (inter-model disagreement, Jensen-Shannon divergence) do. We find that it does not: alignment is weak in both dimensions. At the discrete label level, 50.4% of CIFAR-10H images and 33.5% of FER+ images receive multiple valid classifications from humans, while the models converge on only one. These instances represent a critical failure case where humans perceive ambiguity and would request expert review, yet models decide confidently. At the continuous score level, single-model uncertainty correlates weakly with human disagreement ( \rho = 0.24–0.55 ), and predictive multiplicity provides only modest improvement. Widely-used uncertainty quantification methods do not reliably identify instances humans find ambiguous. Model uncertainty should not be treated as a trustworthy signal by default for decision-making in high-stakes scenarios.
[AI-179] Scalable GNN-based Knowledge Graph Representation Learning with Efficient Message Passing ISWC
链接: https://arxiv.org/abs/2609.34499
作者: Huu Tan Mai,Cuong Xuan Chu,Heiko Paulheim,Daria Stepanova
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at the Posters and Demos Track of the International Semantic Web Conference (ISWC) 2026
Abstract:Graph neural networks (GNNs) excel at representation learning on Knowledge Graphs (KGs), achieving stateof-the-art performance on tasks like link prediction or entity classification. However, their high computational complexity, inherent to their user-defined message passing (MP) algorithm, still prohibits their widespread adoption, especially for large KGs. Current efforts to mitigate the scalability bottlenecks of GNNs on KGs, such as subgraph sampling, are often task- and model-specific, and do not reliably guarantee lossless (if applicable) runtime/space reductions. To address this, we extend Relational Sparse Matrix Multiplication (RSPMM), originally designed to losslessly lower the space complexity of composition-based MP with pointwise composition functions, to support more expressive functions (e.g., 2x2 block-diagonal matrix multiplication, Givens rotation, circular correlation). Our method delivers significant task-independent reductions in runtime and space for current GNNs on KGs and facilitates efficient re-implementations of GNNs that maintain near state-of-the-art performance on challenging KG tasks, for a fraction of computational costs.
[AI-180] PowerBench: A Benchmark for Agent ic Retrieval and Reasoning in Power Systems
链接: https://arxiv.org/abs/2609.34492
作者: Xijing Wang,Yinsheng Yao,Jinru Ding,Yidong Jiang,Ziwen Xu,Yiwen Jiang,Jie Xu,Dawei Cheng
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents offer new opportunities for automated analysis in industry. However, rigorous evaluation of such agents-for example, within power system scenarios-remains hindered: real operational data are confidential, and existing public resources fail to fully capture the chained dependencies and heterogeneous evidence. To address this gap, we propose PowerBench, comprising (1) a generation framework that derives interconnected heterogeneous operational data through a common dependency chain, and (2) a synthetic dataset generated by this framework. The dataset covers 761 devices across 100 device types, with 13.35 million hourly telemetry records spanning two years and 24,939 operational documents. Building on this dataset, we construct 300 questions across three task families that evaluate frontier LLMs’ ability to complete analysis tasks that require autonomous evidence retrieval and reasoning across interconnected and heterogeneous data under restricted tool calls and time budgets. Results demonstrate that the evaluated frontier LLMs remain challenged on these tasks: the best model reaches only 74.2% joint accuracy. Our trace analysis further reveals that model performance varies across evidence discovery, content retrieval, tool use, reasoning over evidence, and answer submission. These findings provide detailed insights for evaluating LLM agents and guiding their reliable deployment in industry. The framework, dataset, and benchmark tasks are available at this https URL.
[AI-181] FlexLoop: Depth-Elastic Looped Policies for Adaptive Test-Time Computation in Deep RL
链接: https://arxiv.org/abs/2609.34488
作者: Xun Wang,Ruishuo Chen,Yu Chen,Zhuoran Li,Longbo Huang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Looped architectures scale computation by reusing the same parameters across recurrent steps, and recent work shows that they substantially improve deep reinforcement learning policies on long-horizon tasks. Since recurrent depth directly controls computation, one may expect looped policies to naturally support elastic inference across recurrent depths. Surprisingly, we find that pretrained looped policies exhibit severe recurrent-depth specialization: reliable decisions are concentrated near the full trained depth, tying deployment computation to this depth even when less computation may suffice. Achieving depth elasticity, i.e., reliable decisions across recurrent depths with adaptive computation at deployment, therefore remains a key challenge. To address this, we propose FlexLoop, a novel post-training framework that converts pretrained fixed-depth looped policies into depth-elastic policies. FlexLoop keeps training on the original RL objective to preserve full-depth capability while performing adjacent-depth policy distillation to progressively transfer decision quality from deeper to shallower recurrent steps. The resulting policy supports reliable inference across recurrent depths and enables state-wise adaptive inference through recurrent-depth consistency. Experiments on 30 online and offline long-horizon goal-conditioned environments show that FlexLoop preserves full-depth performance while making shallower depths effective. Keeping competitive performance, FlexLoop reduces average recurrent depth by up to \bf43% and achieves up to \bf1.34\times wall-clock speedup in a stress test.
[AI-182] Learn Here Move Less Elsewhere: Input-Conditioned Plasticity from Retained-Domain Activation Atlases
链接: https://arxiv.org/abs/2609.34478
作者: Jiangtao Lin,Bangyang Wei,Yihang Ding,Siyi Liu,Yuhan Dong
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Task-specific fine-tuning can rewrite a language model’s answers beyond the training task, complicating updates that must preserve existing behavior. We introduce ATLAS, which turns retained-domain representations into an input-dependent rule for task adaptation. An activation atlas supplies local reference centers and directional filters to a shared low-rank residual. Target supervision learns the residual, while retained geometry shapes its action throughout training and inference. On Qwen3-8B, ATLAS achieves lower mean retained-output Kullback-Leibler (KL) divergence than all seven published baselines at shared coding-performance requirements, with consistent advantages across multiple training seeds. Structural comparisons identify the contributions of retained reference states and directional conditioning, and answer-level analyses show fewer rewritten mathematical answers and more stable commonsense choices. Experiments spanning five backbones and two retained domains further demonstrate coding gains with reduced retained-output movement. With compact storage and modest decoding overhead, ATLAS provides a practical mechanism for acquiring specialized skills while maintaining continuity in existing responses.
[AI-183] Causal Routing for Unlearning
链接: https://arxiv.org/abs/2609.34475
作者: Bardh Prenkaj,Andrea D’Angelo,Davide Mottin,Federico Fontana,Davide Gabrielli,Paola Velardi,Stefano Faralli
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 10 pages, 8 figures
Abstract:LLMs cannot forget the way we delete a file. Strangely, we are asked to remove something that was never put anywhere in particular. What the model took from a piece of text is now smeared across billions of weights. Existing methods rewrite all of them to change one thing, and none of them say which part produced that change. To address this, we introduce Causal Routing for Unlearning (CRU) by asking where the concept is expressed in the model and suppressing only that part. One untrained forward pass over the forget set ranks neurons by how their activations vary. Then, small routing modules on those neurons gate and suppress only the concepts that need to be forgotten. In CRU, the base model is frozen, and any change in behavior is caused only by the gated neurons; hence, why the routing is causal. Due to our parameter efficiency (only ~0.01% as many parameters as the base model), unlearning a concept costs 14 GiB, whereas the baselines require 71 GiB. On TOFU, CRU is indistinguishable from the retained model (p 0.05, KS test) and is never Pareto-dominated, whereas every compared baseline matches its forgetting on the larger-forget batches only by collapsing utility. On RWKU, it achieves an adversarial-probe recall of 0.052, compared to 0.250 for the strongest baseline, meaning the knowledge is gone, not merely harder to reach. Thus, deciding on the intervention at query time, rather than fixing it beforehand, is the axis along which we argue that unlearning should proceed.
[AI-184] CoDeL: Co-Evolutionary Defense against Indirect Prompt Injection in LLM -based Agents
链接: https://arxiv.org/abs/2609.34463
作者: Xiao Yang,Yangchen Ou,Yuhan Gao,Le Wang,Zonghao Ying,Aishan Liu
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM)-based agents increasingly rely on external tools and content, exposing them to indirect prompt injection (IPI). This threat has motivated a wide range of defenses, among which training-based defenses are often regarded as most reliable. However, existing training-based defenses are typically optimized on a static distribution of explicit injections. They learn surface-form cues rather than the boundary between serving the user and obeying an injected objective, and therefore fail when malicious intent is folded into a plausible workflow and deferred for several turns. We present CoDeL, a defense that hardens agent against an attack distribution it reshapes as it trains. The defender is updated each round via LoRA-based GDPO under a decoupled reward over safety, task progress, and format compliance, so refusing injections and completing the user’s task jointly define fitness. To keep supplying it with the failures worth learning from, a co-evolving prober searches over injection rounds, attack methods, and payloads for injections that still penetrate the current defender, guided jointly by attack success and attack latency so that it preferentially mines breaches the defender notices too late. Each defender update invalidates part of the attack population and forces the next round onto a new frontier, turning the defender’s own failures into a moving curriculum. Extensive experiments on three IPI benchmarks, nine baselines, and two base models show that CoDeL reduces attack success rate (ASR) by 88.5% and outperforms other baselines largely (+38.0%). Codes are available.
[AI-185] When Does Structured Knowledge Help Neural Theorem Proving?
链接: https://arxiv.org/abs/2609.34460
作者: Sareh Nabi,Roland Vogl,Marzieh Nabi
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Logic in Computer Science (cs.LO)
备注: 30 pages, 4 figures, 12 tables
Abstract:Does structured mathematical knowledge help LLMs prove theorems in Lean 4? If so, for which models, and does the answer vary by problem? Formal libraries such as Mathlib encode 285,000+ verified theorems with syntactic dependencies, but the semantic layer mathematicians rely on for discovery (analogies, generalizations, cross-domain bridges) remains implicit. We introduce MathAgent, which builds this layer as a knowledge graph, MathKG, and uses it to augment LLM theorem provers. MathKG connects 364 Mathlib theorems and definitions by 9,434 typed semantic edges inferred via LLM-based relation extraction anchored to verified Mathlib declarations. We run a controlled ablation across four augmentation modes (no context, knowledge-graph context, Mathlib retrieval, both) and five models: Qwen3-8B/32B, their Lean-specialized derivatives Goedel-Prover-V2-8B/32B, and Claude Sonnet 4.6, on miniF2F, plus PutnamBench and MathOlympiadBench for Sonnet. Three findings emerge. (i) Specialization dominates augmentation: Lean fine-tuning adds 33-38 percentage points of solve rate in every mode, and a specialized 8B model beats a 4\times larger general one by 29-35 points, while no augmentation mode improves solve rate by more than 3 points. (ii) Augmentation is capability-conditioned: knowledge-graph context helps small models but hurts large ones, with the specialized model gaining more relative to its general base at every scale. (iii) Yet the augmentation modes solve different problems: an oracle selecting the best mode per problem solves 6% to 58% more than the unaugmented prover, a complementarity effect that strengthens on harder problems (32% more on PutnamBench). These results motivate adaptive strategies that select augmentation by model capability and problem. Code, data, and artifacts are available at this https URL
[AI-186] Escaping Local Views: Discovering Latent Concepts for Interpretable Multi-Agent Reinforcement Learning
链接: https://arxiv.org/abs/2609.34459
作者: Yijie Sun,Sanquan Sun,Yanda Zhu,Yuanyang Zhu,Yaohua Hu,Chunlin Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Efficient cooperation is challenging due to the usual partial observability of each agent in multi-agent reinforcement learning. Recurrent networks encode local interaction histories, but their hidden representations provide limited insight into the information underlying individual decisions. To address these challenges, we propose a novel interpretable framework, called escaping local views (ELV), which introduces semantically structured latent concepts to render policy decisions transparent. Specifically, each agent extracts low-dimensional semantic concepts from its local observation and action-observation trajectory. These concepts are jointly encoded into a contextual latent variable via a variational autoencoder (VAE), which builds a bridge between local views and global semantics. To explicitly model the decision of each agent, we employ a dual-path attention mechanism in which one module estimates the salience of individual concepts relative to the global context, while the other captures higher-order cooperative patterns with pairwise concept interactions. Furthermore, we incorporate a concept prediction module that derives an intrinsic reward from next-concept prediction errors, which incentivizes agents to explore regions of semantic novelty. Experiments in multiple environments verify that ELV not only achieves competitive performance but also explicitly provides how agents reason about their decisions.
[AI-187] SPACE-LoRA: Allocating Activation-Subspace Protection for Continual Learning
链接: https://arxiv.org/abs/2609.34453
作者: Seunghyun Yoo,Kiseok Kim,Hyeontae Joo,Junyeop Bang,Hwangnam Kim
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 18 pages, 2 figures, 6 tables
Abstract:This study addresses the catastrophic forgetting problem that occurs when sequentially learning successive tasks using Low-Rank Adaptation (LoRA) from a lifelong learning perspective. While existing approaches have primarily constrained parameter updates or learning subspaces to reduce interference with past knowledge, they have not fully considered additive interference. This occurs when a newly added residual adapter on top of a fixed past model generates non-zero responses along input directions important for old tasks, thereby altering previous predictions. To this end, we propose Subspace Protection with Allocated Capacity for Efficient Continual Adaptation (SPACE-LoRA). SPACE-LoRA directly suppresses the responses of the new residual branch along input activation directions that are important for old tasks and adaptively determines the protection coverage for each module based on past-task sensitivity estimated via a common Fisher sensitivity-based coverage target. Under a fixed LoRA rank, this approach adaptively adjusts module-specific protection coverage while suppressing interference along input directions sensitive to old tasks. We assess the effectiveness of activation-subspace protection in mitigating catastrophic forgetting and examine the role of sensitivity-guided protection in continual learning across diverse tasks. Code is available at this https URL.
[AI-188] CORTEX: Learning to Share and Specialize in Dense Language Models
链接: https://arxiv.org/abs/2609.34449
作者: Chuiyang Meng,Ming Tang,Vincent W.S. Wong
类目: Artificial Intelligence (cs.AI)
备注: 28 pages
Abstract:Large language models are trained on heterogeneous data mixtures, where different knowledge domains require both shared knowledge and specialization. Existing modular approaches typically impose explicit components or discover modules through interpretability analysis after training. In this work, we propose CORTEX, a learning dynamics-inspired framework that learns internal modularization within dense language models. CORTEX partitions trainable matrices into parameter groups and learns module assignments from domain-conditioned gradient and cross-domain gradient similarity. We introduce the selective lesion score and module-domain mutual information to characterize the target-domain lesion effects and alignment, and analyze how module assignment affects the trade-off between assignment bias and update magnitude. Experiments with 160M, Qwen3-8B, and Qwen3-32B backbone models show that CORTEX achieves the highest synthetic-domain exact match and largest average perplexity reduction, while remaining competitive on real-domain evaluations and forming identifiable modules.
[AI-189] Social Circuits behind Multi-agent Echo Chambers
链接: https://arxiv.org/abs/2609.34444
作者: Chuiyang Meng,Wenlu Yu,Ming Tang,Cheng Li
类目: Artificial Intelligence (cs.AI)
备注: 37 pages
Abstract:Language-model agents exchange messages to combine evidence, but their communication can also create echo chambers that reinforce shared errors. However, overall task performance does not explain how a message changes the receiving agent’s internal activations and affects its decision. In this work, we introduce Social Circuits, a framework for tracing message effects through receiver activations. We compare the receiver’s answers before and after changing a message. Then, we restore selected activations recorded under the original message to determine how much of the message effect these activations reproduce. Based on Social Circuits, we propose Circuit-Guided Deliberation (CGD), which learns to select useful messages using receiver activation changes. We establish when activation replacement preserves receiver decisions and bound the gap between CGD’s task performance and the best achievable through message selection. Experiments show that receiver activation changes explain the message effects and guide message selection that improves the task performance. Across three models and four datasets, CGD achieves the highest or joint-highest average accuracy in our main comparisons while generating fewer tokens than multi-agent baselines.
[AI-190] Beyond End-to-End Black Box Mapping: An Intentional Agent Framework for Cognitive-driven Facial Reaction Generation
链接: https://arxiv.org/abs/2609.34419
作者: Hanzhong Zhang,Jindong Wang,Siyang Song
类目: Artificial Intelligence (cs.AI)
备注: 36 pages, 6 figures
Abstract:Automatic human-like facial reaction generation (FRG) is essential for building intelligent systems that can engage in human-computer interaction (HCI). While diverse and context-appropriate facial reactions can reflect latent appraisal and affective processes in human interaction, most existing FRG methods rely on end-to-end architectures that directly map speaker behaviours to listener expressions without an explicit intermediate internal state. We reformulate FRG as generation mediated by a structured internal-state process and propose the \textbfIntentional Agent, which shifts FRG from direct stimulus-response mapping to stimulus-grounded generation through explicit intermediate states. To represent temporal internal-state evolution, we propose an internal dynamics model that integrates emotional drives with an iterative Inner Thought Flow (ITF) within a structured intermediate state used for subsequent generation. This state can continue to update during conversational silences. Furthermore, to bridge abstract internal states with physiological actions, we formulate FRG as a downstream affective mapping from this latent thought flow to facial expressions. Experiments on the REACT 2025 dataset show an FRDist of 72.39 and an FRDiv of 0.5057; perceptual plausibility is evaluated separately through blinded human ratings. A blinded human evaluation of 96 reactions found no significant difference in mean score between Full and ground truth ( 5.527 vs.\ 5.195 , p_\mathrmHolm=.076 ), while Full significantly outperformed Event-Triggered and Heuristic-Only (both p_\mathrmHolm.001 ). The Reaction Quality Scorer (RQS) correlated strongly with human judgements (Pearson r=.855 ; Spearman \rho=.821 , both p.05 ), supporting its use as an automatic metric. These results underscore the immense potential of endogenous dynamics in building highly autonomous, human-like agents.
[AI-191] OSPD: On-Policy Self-Distillation for Persona-Consistent Dialogue EMNLP2026
链接: https://arxiv.org/abs/2609.34418
作者: Rui Xu,Yikai Zhang,Aili Chen,Zicheng Zhao,Xu Yinghui,Libo Wu
类目: Artificial Intelligence (cs.AI)
备注: Accepted to EMNLP 2026. 22 pages, including references and appendices
Abstract:Maintaining persona consistency across multi-turn dialogues remains a core challenge for role-playing language models. Off-policy distillation from external teachers incurs distribution mismatch that compounds across dialogue turns, while reinforcement learning struggles with reward ambiguity inherent in subjective persona fidelity. We propose OSPD, an on-policy self-distillation framework where the same model serves as both teacher and student under asymmetric information: the teacher receives a complete character profile while the student sees only a brief summary, and the student generates trajectories from its own policy. We find that teacher confidence in role-playing dialogue exhibits a bimodal structure—sharply peaked at character-critical tokens yet diffuse at generic utterances—and introduce role-aware divergence switching to match this structure. A progressive trait masking curriculum further forces staged internalization of character knowledge along semantic dimensions. Experiments on CharacterBench, CharacterEval, and SocialBench show that OSPD substantially improves persona consistency over supervised fine-tuning and multi-turn RL baselines, without requiring any external teacher or reward model.
[AI-192] Mathematics for and by human cognition: A resource-rational search for bottlenecks in problem-solving
链接: https://arxiv.org/abs/2609.34410
作者: Sneha Aenugu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Human cognitive constraints are generally viewed as limiting factors in problem-solving. We argue that these constraints can instead play a critical role in driving advances in mathematics and beyond. We propose a theory of mathematical abstraction as a resource-rational search for bottlenecks in problem-solving. Bottlenecks arising from cognitive constraints create pressure to restructure existing knowledge, potentially giving rise to novel formalisms with applications beyond the problems that originally motivated them. Drawing on episodes from the history of mathematics, we illustrate how such bottlenecks can drive the development of novel abstractions and examine how cognitive constraints and affective responses shape this process. Finally, we discuss the implications of this account for machine mathematical discovery and argue that incorporating human-like constraints may facilitate the discovery of useful mathematical abstractions.
[AI-193] MASCIT: A Mask-Aware State Space Classifier for Naturally Irregular Time Series
链接: https://arxiv.org/abs/2609.34409
作者: Yoo-Min Jung,Hyeon-Gi Kim,Jonghun Park
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: accepted at APIEMS 2026
Abstract:Naturally irregular time series combine asynchronous observations, missing values, unequal lengths, and nonuniform sampling, while dense adapters can discard temporal structure. We propose a mask-aware state space classifier for irregular time series (MASCIT), which supplies observation masks to the encoder and excludes invalid steps from gated temporal aggregation. Across 34 irregular time series datasets, MASCIT yielded the strongest aggregate point estimate and was the only evaluated neural model with three-seed results on every dataset. MASCIT retained the lowest point rank across six overlapping irregularity indicators, while factorial ablations favored partial over full selectivity. These results support selective state space models as effective, executable backbones for naturally irregular time series classification.
[AI-194] SkillFocus: Evolving Agent Skills via Capability Decomposition
链接: https://arxiv.org/abs/2609.34397
作者: Ning Wang,Zhiren Gong,Bingdong Li,Peng Yang,Aimin Zhou
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Agent skill evolution seeks to improve reusable procedural guidance for large language model (LLM) agents through iterative revision. Existing methods base each revision mainly on execution trajectories or feedback, leaving recurring behavioral requirements across tasks implicit and tying revision to the behavior of the current skill. We introduce SkillFocus, which decomposes recurring task requirements into a capability space that remains fixed as the skill evolves, separating what tasks require from how the current skill behaves. SkillFocus maps current task outcomes to this space to identify the capability that leaves the most tasks unresolved, then uses that capability to determine what to revise and which evidence to use. Across four benchmarks spanning heterogeneous tasks, SkillFocus achieves the best held-out accuracy on all four, outperforming the strongest competing result by 5.7 points on average while using 24% fewer evolution tokens on average than the closest iterative baseline. Controlled studies further show that capabilities derived from recurring task requirements outperform task-semantic and execution-derived alternatives, while randomizing task–capability assignments reduces final accuracy by up to 20.2 points. Matching evidence to the selected capability increases candidate gain by 4.4 points under prioritized revision.
[AI-195] Org-Agent : Beyond Personal Assistants Towards Organizational Agents
链接: https://arxiv.org/abs/2609.34392
作者: Luyao Zhuang,Yujing Zhang,Zijin Hong,Yilin Xiao,Xiao Huang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Language model agents serving organizations must coordinate requests from multiple users while using knowledge distributed across their interactions. We identify two complementary capabilities for this setting, namely cross-user interaction and decision-making, as well as cross-user memory and knowledge use. Both capabilities are governed by organizational constraints across three aspects: user identity, authority, and access permissions; the attribution and temporal validity of information; and rules for resolving conflicting requirements across users and completion requirements for joint decisions. These constraints shape what information or decisions must be obtained before an action can proceed and what conditions must be satisfied during its execution. Motivated by this, we introduce Org-Agent, a unified constraint-centric reasoning framework that organizes task execution in three stages. Specifically, Org-Agent decomposes a task into atomic subtasks and constructs a task dependency graph whose edges encode the dependencies among them. Building on this graph, it schedules the subtasks in dependency order through topological sorting. It then executes each subtask while accounting for the task’s constraints, supported by evidence-acquisition and memory-management tools. Experiments on MUSES-Bench and GroupMemBench demonstrate the effectiveness of Org-Agent on both capabilities, and ablations further support the contributions of dependency modeling and tool use.
[AI-196] P2P: Cross-View Population Denoising for Unpaired Single-Cell Perturbation Response Prediction
链接: https://arxiv.org/abs/2609.34391
作者: Haojie Yang,Ran Su
类目: Emerging Technologies (cs.ET); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:AIVC (AI Virtual Cell) is a learned simulator of cellular behavior across conditions. Predicting how a cell population responds transcriptionally to a genetic perturbation is a core task. Perturb-seq records that response by destructive sequencing, so a control cell and a perturbed cell are never observed as a pair, and cells under one condition remain heterogeneous and noisy. Regression on individual cells absorbs sampling variation into the estimated effect, whereas interpretation requires the reproducible population effect. P2P (Perturbation-to-Perturbation) takes a stochastic cell-set view as its supervision unit. Two views drawn from the same condition share a reproducible population effect and differ by view-specific variation. A permutation-invariant set encoder summarizes the control population, a structured encoder represents perturbation tokens, cellular context, dose, and combination interactions, and a gate blends empirical condition-effect memory with a neural residual. A heteroscedastic head predicts the population mean and gene-wise response variance. Under one protocol and five seeds, P2P attains the lowest expression RMSE and the highest Effect Pearson, DEG F1, and DEG average precision on each of Adamson, Norman, Replogle K562, and Replogle RPE1 relative to GenePert, LinearPert, SLIM, Scouter, and scPILOT. On Replogle K562, Effect Pearson rises from 0.643 to 0.702 and DEG F1 rises from 0.067 to 0.178 relative to Scouter, the strongest baseline on both metrics.
[AI-197] DPS: Dual-Mode Precision LLM Serving with Semi-Unified Memory
链接: https://arxiv.org/abs/2609.34380
作者: Xuan Truong Nguyen,Tien Son Pham,Tuan Duc Chu,Wookeun Jung,Thanh Tuan Dao
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI); Emerging Technologies (cs.ET); Performance (cs.PF); Software Engineering (cs.SE)
备注: 14 pages, 7 figures
Abstract:Existing LLM serving systems virtualize and optimize KV-cache memory, but treat model-weight memory as fixed throughout execution. Recent work on multi-precision model representations challenges this design by allowing a single stored model to support both full-accuracy and lower-precision execution, making the effective weight footprint runtime-dependent. This creates an opportunity under bursty workloads, where temporary spikes in KV-cache demand often determine throughput and SLO compliance. We present DPS, a dual-precision LLM serving system that turns weight memory into an elastic resource: under normal load, DPS serves the full-accuracy model; under KV pressure, it switches to a nested, lower-precision variant and repurposes unused weight memory for KV cache blocks. DPS is built on Semi-Unified Memory (SUM), which partitions the weight region into a persistent lower-precision sub-region and a shared region that alternates between residual weight tensors and KV-cache blocks, preserving compatibility with paged KV-cache management. We implement DPS on top of vLLM and evaluate it across both dense and MoE models and various production workload traces. Our results show that \sysname improves sustained throughput by 2.1 – 3.3\times and effective pass@1 by up to +41 ,pp over Static FP16, while preserving FP16-class accuracy.
[AI-198] Before Agents Act: Assurance-Aware Semantic Scheduling for Evidence Acquisition in Distributed Systems
链接: https://arxiv.org/abs/2609.34376
作者: Jun He,Deying Yu
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: 30 pages, 5 figures, 3 tables. Extended manuscript with formal proofs and operation catalogue
Abstract:Tool-using agents can initiate consequential infrastructure changes, yet evidence required for admission may expire while other checks run or depend on a shared fault domain. We formulate evidence acquisition as joint witness selection and scheduling under quorum, diversity, freshness, deadline, and resource constraints. Assurance-Aware Semantic Scheduling (AAS) combines integer-program selection, dispatch-aware temporal scheduling, bounded diagnostic expansion, and receipt-aware repair. Formal results state the assumptions needed for dispatch-time freshness and finite diagnostic expansion. In three generated infrastructure workloads, AAS produces 1,075/1,200 valid candidates versus 647/1,200 for constraint-aware forward scheduling; stale candidates fall from 440 to 12. Paired sensitivity studies reuse the same instances and operation latency draws across parameter settings. A corrected timeout intervention finds 18/20 admissions with repair or full resynthesis versus 0/20 for a static plan, with lower committed cost when receipts are reused. On 20 constructed cases requiring a certified decomposition cut, refinement recovers an oracle-matching feasible plan every time. These are controlled simulation results; the bounded oracle shares a temporal search component, and transfer to deployed systems remains untested.
[AI-199] LRC-JEPA: Disentangling Dynamics and Residual Context for Efficient World Models
链接: https://arxiv.org/abs/2609.34375
作者: Luzhe Huang,Lei Chu,Jingyi Liang,Yuhuan Zhao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:Compact JEPA world models enable efficient latent-space planning, but low-dimensional representation trained under reward-free self-supervision must encode both action-conditioned dynamics and predictable visual context. This competition can entangle controllable state with high-rank nuisance appearance and degrade planning as scenes become more complex. We introduce LRC-JEPA, a lightweight end-to-end world model that routes information into a compact predictive latent \mathbfz and learned-query residual-context embeddings \mathbfu . Only \mathbfz is propagated by the dynamics model and used for planning, while \mathbfu captures temporally persistent information for cross-attention reconstruction; a differentiable residual connection encourages the latent to retain complementary dynamic content. Under explicit assumptions, we show that the resulting representation is sufficient, minimal, nuisance-invariant, and disentangled. Across four simulated control environments, LRC-JEPA improves average planning success over a parameter-matched JEPA baseline by 9 percentage points and matches or exceeds substantially larger pretrained models. On the real-world Bridge-v2 set, its 5.5M-parameter active encoder outperforms DINO-WM (22.1M) and V-JEPA2 (303.9M) encoders while also enabling faster planning. Physical-state probes, reconstruction interventions, and ablations confirm the effectiveness of LRC-JEPA’s representation disentanglement.
[AI-200] ABC-Align: Prediction-Powered Alignment with Adaptive Bias Control
链接: https://arxiv.org/abs/2609.34374
作者: Eric Frankel,Banghua Zhu,Sewoong Oh,Lillian J. Ratliff
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: 47 pages; 9 figures
Abstract:Language model post-training is often bottlenecked by the need for human-collected preference data, which is expensive and difficult to scale. Reinforcement learning from AI feedback (RLAIF) style approaches that leverage pseudo labels offer an abundant alternative but introduce systematic biases that degrade downstream alignment. Recent general-purpose semi-supervised methods correct for teacher bias using a small set of human-labeled examples, but suffer from high variance especially when human annotations are scarce. To this end, we propose ABC-Align, leveraging abundant pseudo label signal to minimize variance and applying a lightweight, adaptive correction grounded in the human-labeled subset. The correction strength is tuned automatically during training using plug-in estimates of the relevant bias–variance quantities. On LLM alignment with RLHF, DPO, and GRPO where human feedback is scarce, we empirically demonstrate that ABC-Align achieves superior performance over prior semi-supervised baselines in a series of experiments on an increasing scale. Our code is available at this https URL .
[AI-201] PersMem: Internalizing Personality into Dual-Pathway Memory for LLM Agents
链接: https://arxiv.org/abs/2609.34372
作者: Hanzhong Zhang,Ziwei Xiang,Weicheng Xie,Shizhe Liu,Siyang Song
类目: Artificial Intelligence (cs.AI)
备注: 39 pages, 2 figures
Abstract:The profile of a role-playing agent usually depends on the pre-defined personality in a system prompt, whereas its memory processing pipeline, including prioritisation of stored memories and subsequent retrieval, remains independent of this personality. This separation causes the agent’s memory processing to be inconsistent with the pre-defined personality, and makes it difficult to validate whether agent behaviours follow this personality. In this paper, we propose Personality-Integrated Memory (PersMem), which integrates personality into the agent’s memory processing pipeline, making it consistently personality-dependent. PersMem processes memory using four steps, where the personality is mapped to operation-specific parameters controlling: (i) affective appraisal annotating emotion states of the user input; (ii) retention of previously stored memories along with the current input; (iii) passive affect-driven memory retrieval exploring memories similar to user input in semantics and personality-guided emotions; and (iv) active goal-driven memory retrieval that refines and selects passively retrieved memories for the reply. Consequently, consistency with the pre-defined personality can be examined by inspecting memory-processing traces during human-agent interactions. We evaluate these personality-dependent differences in attachment and Big Five settings. PersMem exceeds the chance baseline for four-way attachment classification by 23.1 percentage points. In Big Five dialogue comparisons, PersMem achieves 67.5% accuracy, 6.7 percentage points above a baseline using uniformly sampled memories. On CoSER, PersMem achieves an average score of 66.13, with scores of 69.33 for Character Fidelity and 84.33 for Storyline Quality. Together, these results show that PersMem produces distinguishable personality-related memory-processing patterns.
[AI-202] CoeF-SFL: Preserving Collaborative Server-Client Learning with Enhanced Communication Efficiency
链接: https://arxiv.org/abs/2609.34360
作者: Junwoo Bae,Jin-Hyun Ahn
类目: Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: Submitted to a conference
Abstract:Split Federated Learning (SFL) enables resource-constrained clients to participate in collaborative training, but vanilla SFL exchanges smashed data and gradients at every batch, which incurs significant communication overhead. Recent methods reduce this overhead with an auxiliary network at the client-side cut layer. However, we identify that this approach makes the client optimize a local objective that differs from the end-to-end objective, which fundamentally limits the collaborative training between the client and the server. We propose Compensated Feedback based SFL (CoeF-SFL), a communication-efficient framework that retains the end-to-end objective without any auxiliary network. In CoeF-SFL, the client and the server exchange the smashed data and the gradients once per round and reuse them during local training. Since this reuse makes the gradients stale on the client side, we compensate them with a curvature-based correction in the activation space and develop two variants. CoeF-D approximates the Hessian with a diagonal gradient outer product, while CoeF-J exploits the tractable Jacobian-based Hessian of a surrogate loss that upper-bounds the true loss. We provide the theoretical background of each method, characterizing its compensation. Across vision and language tasks, model capacities, cut layers, and data distributions, CoeF-SFL significantly outperforms auxiliary-network-based methods under the same communication frequency, and the improvement is most substantial on vision tasks. Code is available at this https URL
[AI-203] Improving Large Language Models for Code through Runtime Program-State Reasoning
链接: https://arxiv.org/abs/2609.34359
作者: Hongwei Li,Spandan Garg,Yufan Huang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models receive limited explicit training in reasoning about runtime program states. We study whether training models to reason about runtime program states improves downstream software-engineering capabilities. We introduce two complementary program-state reasoning tasks. Buggy input-output reasoning requires a model to generate a concrete input that exposes a behavioral difference between a buggy program and a hidden correct implementation and to predict the resulting execution behavior. Precondition-postcondition reasoning requires an agent to symbolically characterize a bug-triggering precondition, predict the expected postcondition, explain their causal connection, and instantiate this reasoning as an executable regression test. By incorporating these two tasks into a staged post-training pipeline, we develop Comet-9B, a 9B language model based on Qwen3.5-9B Base. We evaluate the resulting checkpoints on repository-level patch generation, regression-test generation, and security PoC generation. Adding both program-state reasoning tasks to supervised fine-tuning (SFT) on issue resolution improves success rates by 7.25 percentage points on SWE-bench Pro and 9.70 points on SWT-Bench Verified. Sequential reinforcement learning on the two tasks yields further gains of 7.25, 26.79, and 4.67 percentage points on SWE-bench Pro, SWT-Bench Verified, and CyberGym, respectively. Despite having only 9B parameters, Comet-9B achieves a score comparable to the reported GPT-5.2 result on SWE-bench Pro and matches the reported success rate of a GPT-4o-based agent on SWT-Bench Verified.
[AI-204] Predictive Semantic Safety: From Visual Physical Reasoning to Safety-Critical Control WWW
链接: https://arxiv.org/abs/2609.34356
作者: Taekyung Kim,Salem Fradi,Yanning Dai,Mateusz Ostaszewski,Jürgen Schmidhuber
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: The first two authors contributed equally to this work. Project page: this https URL
Abstract:Physical interactions can create future hazards that are not apparent from the robot’s current geometric surroundings. We present a framework termed Predictive Semantic Safety (PSS), which connects visual physical reasoning to backup-based safety filtering. A vision-language model (VLM) predicts physical events and their timing or directly predicts object displacements. An explicit motion model converts event hypotheses into object trajectories. Split conformal prediction calibrates position errors jointly across specified objects, observation times, and future times; geometric shape bounds convert the resulting position regions into predicted object occupancy. PSS evaluates a prescribed backup maneuver against this occupancy and derives input-affine constraints for minimally modifying the nominal input while preserving backup feasibility under the robot dynamics and input limits. MuJoCo experiments with a Unitree Go1 consider falling fixtures, impact-driven support loss, and contact propagation. PSS achieves a safe episode rate of 99.3%, compared with 43.3% for a Backup Control Barrier Function baseline that only uses current obstacle geometry.
[AI-205] SemRD-V2X: Closure-Guided Communication with Bounded Inference for Cooperative Perception
链接: https://arxiv.org/abs/2609.34353
作者: Hu Xu,Chun Li,Siyuan Qiu,Zeyan Li,Jianfeng Xu
类目: Artificial Intelligence (cs.AI); Information Theory (cs.IT)
备注:
Abstract:Vehicle-to-Everything (V2X) cooperative perception improves 3-D detection by sharing intermediate features, but dense remote features may repeat context that the ego agent can infer locally. Most communication-efficient designs optimize masks or codes empirically, leaving a more basic question open: which remote evidence is indispensable given the receiver’s own observation? We introduce a closure-fidelity perspective on ego conditioned remote perception. Under a finite deductive abstraction and explicit conditions, its rate–distortion function decomposes over an irredundant core, and the exact zero-distortion rate becomes P_A H(\pi_A) . This analysis suggests a concrete design principle: transmit compact evidence and recover derivable context with bounded receiver-side inference. Guided by this principle, SemRD-V2X is an operational neural proxy that combines exact-budget BEV support selection, pointwise channel compression, and masked shared-weight reconstruction before standard fusion. Experiments on simulated V2XSet and real-world DAIR-V2X validate the resulting design. In a controlled five-run V2XSet comparison against a locally reproduced V2X-ViT-v1 baseline on one Tesla V100, SemRD-V2X reduces the analytical feature payload by 26.6\times while improving AP@0.5/AP@0.7 by 4.13/8.57 points, with 3.81% additional mean compute latency. These results position closure fidelity as both an analytical lens and an actionable design principle for communication-efficient cooperative perception.
[AI-206] Fuzzy Distribution Modeling for Synthetic Tabular Data Generation with Causality Preservation
链接: https://arxiv.org/abs/2609.34349
作者: Michael Vasilakakis(1),Dimitris K. Iakovidis(1) ((1) Department of Computer Science and Biomedical Informatics, University of Thessaly, Lamia, Greece)
类目: Artificial Intelligence (cs.AI)
备注: 6 pages, 2 figures, 3 tables. Published in the 2026 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), Maastricht, the Netherlands
Abstract:Synthetic tabular data generation provides an effective alternative for the training of machine learning models when real-world data is limited or inaccessible. However, the heterogeneous, non-smooth, and incomplete nature of tabular data poses fundamental challenges to conventional probabilistic and deep generative models, where their interpretability remains limited. This paper proposes a novel fuzzy distribution modeling methodology for synthetic tabular data generation based on fuzzy sets theory. Feature distributions are represented using fuzzy sets and feature dependencies are modeled through Fuzzy Cognitive Maps, resulting in a low-parameter, and an interpretable data representation. Synthetic samples are generated by sampling fuzzy concepts rather than raw values, enabling native support for mixed data types, missing values, and domain constraints. The methodology further supports linguistic queries and IF-THEN reasoning, facilitating transparent simulation of decision-making processes. Experimental results on benchmark datasets demonstrate competitive performance with respect to utility, fidelity and privacy compared to state-of-the-art methods, while offering substantially improved interpretability. These results establish fuzzy distribution modeling as a principled and effective approach for synthetic tabular data generation in fuzzy systems and decision support applications.
[AI-207] SAIL: Spatial Audio Intelligence with Large Language Models via Disentangled Acoustic-Spatial Encoding and Dual-Stream Q-Former
链接: https://arxiv.org/abs/2609.34347
作者: Zhengding Luo,Jinyang Wu,Haozhe Ma,Yanghao Zhou,Woon-Seng Gan,Wenwu Wang
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注:
Abstract:Spatial audio large language models (LLMs) enable embodied agents, wearable assistants, and immersive systems to recognize sound events, localize sources, and reason about their spatial relationships. However, existing spatial audio LLMs often rely on early fusion of acoustic and spatial features and source-agnostic token representations. These designs make it difficult to preserve the correspondence between individual sound events and their spatial attributes, particularly in multi-source scenes. To address this limitation, we propose SAIL, a Spatial Audio Intelligence framework with LLMs that preserves acoustic-spatial structure and source-level correspondence from audio encoding to LLM alignment. SAIL introduces a Disentangled Spatial Audio Transformer that represents Mel-spectrogram and interaural phase difference features as separate acoustic and spatial streams. Source-discriminative task queries further learn event, direction, and distance information for each source. A Dual-Stream Q-Former then aligns the two streams with the LLM using acoustic and spatial queries organized by source slots. Compared with the early-fusion baseline, SAIL achieves consistent improvements in dual-source sound event detection, direction and distance estimation, and spatial reasoning. These results demonstrate the importance of structured, source-discriminative audio representations for multi-source spatial understanding and reasoning.
[AI-208] Learning to Steer Steering to See: Unveiling the Geometry of RLVR in Large Language Models via Trainable Vectors
链接: https://arxiv.org/abs/2609.34344
作者: Yuchen Cai,Ding Cao,Qixiang Yin,Xin Xu,Kai Yang,Siye Wu,Pengyuan Wang,Jiaxuan Wang,Weijie Liu,Saiyong Yang,Guangzhong Sun,Guiquan Liu,Junfeng Fang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 41 pages
Abstract:Reinforcement learning (RL) has become a key paradigm for enhancing the reasoning of large language models, yet the high dimensionality of parameter updates makes its training dynamics hard to analyze. We study reinforcement learning with verifiable rewards (RLVR) and use vector steering to identify a low-dimensional effective manifold in activation space associated with RL-induced gains. We uncover two geometric properties. (1) Effective Manifold Capacity: the capacity needed to reproduce RL gains can be very small but is not infinitely compressible; at extremely low capacity, intervention dimensionality and input-dependent expressiveness become key constraints, and this requirement varies with injection depth. (2) Control Manifold Separation: effective control directions lie mainly in the low-variance complement of the activation principal subspace. Within a task and base model, the learned geometry stays largely consistent across training configurations, and across tasks geometric alignment correlates with capability transfer. Experiments on 5 LLMs and 6 verifiable-reward tasks support these findings. We then propose Alpha-Stabler, a plug-and-play framework with a Predictor that monitors principal-subspace intrusion for early collapse warnings, and a Controller that removes the principal-subspace component of activation gradients during backpropagation while preserving the orthogonal complement. Alpha-Stabler stabilizes training for 2,000 steps and consistently improves RL gains, offering practical insights for robust post-training. Code: this https URL
[AI-209] SAGE: Structured Strategic Reasoning for Efficient LLM Game Playing
链接: https://arxiv.org/abs/2609.34342
作者: Zhiwei Chen,Tianchun Wang,Zhongtao Rao,Haiming Zhu,Ding Cao,Tianxiang Zhao
类目: Artificial Intelligence (cs.AI)
备注: 31 pages with multiple figures
Abstract:A strong LLM strategic agent should reason prospectively over uncertain futures, adapt its strategy to opponents’ behavioral tendencies, and continuously recalibrate its decision process from interaction experience. However, incorporating these sources in free-form reasoning could lead to unsupported strategic assumptions, inconsistent opponent estimates, and harmful interference from irrelevant historical interactions. To address these issues, we propose SAGE, a training-free inference-time framework that structures LLM strategic reasoning around three coordinated operations: anchor, adapt, and recalibrate. SAGE first anchors reasoning to an equilibrium policy that provides a strategically valid prior. It then conditions deviations from this anchor on a soft belief over opponent behavioral tendencies, enabling opponent-specific exploitation. Finally, SAGE distills strategically related interactions into counterfactual hypotheses about previously missing considerations, allowing past experience to recalibrate the model’s reasoning. We evaluate SAGE on three repeated imperfect-information games: Leduc Hold’em, Liar’s Dice, and Goofspiel, against various opponent types in each game. Compared with reasoning-intensive LLM agents, including Suspicion-Agent, ReTA, Agent-Pro, EMO, and Hypothetical Minds, SAGE achieves up to a 127.6% payoff improvement in Liar’s Dice while reducing input and output token usage by up to 80% and 90%, respectively. In direct match-up play, it attains non-negative mean payoff against 5/10, 8/10, and 8/10 evaluated opponents in Leduc Hold’em, Liar’s Dice, and Goofspiel, respectively, while using relatively fewer tokens. Code is available at this https URL.
[AI-210] st-Time Scaling via Budgeted Multi-Attribute Verification
链接: https://arxiv.org/abs/2609.34322
作者: Bo Xue,Ji Cheng,Shen-Huan Lyu,Yuanyu Wan,Shuang Qiu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Verifying LLM-generated answers under a shared computational budget requires jointly deciding which candidates to inspect and which verification attributes to evaluate. We formulate this problem as multi-attribute good-arm identification under a global budget: each candidate is an arm evaluated along several costly attributes, and the goal is to certify as many candidates as possible whose mean scores exceed the prescribed thresholds on all attributes. We propose \textscBMA-GAI, an algorithm that combines cost-aware arm selection with adaptive sampling of attributes. Every observation serves both to guide adaptive allocation and to support anytime-valid certification, which removes the need for a separate confirmation stage. We establish an asymptotic coverage guarantee for \textscBMA-GAI and derive a matching information-theoretic converse that characterizes the intrinsic complexity of the problem, thereby proving that \textscBMA-GAI is first-order optimal away from critical budget levels. Experiments on synthetic benchmarks and an LLM answer-verification task show that \textscBMA-GAI allocates the verification budget more efficiently and certifies more high-quality candidates than competing methods.
[AI-211] Dynamical Parameters: An Interpretability Framework for Time-Series Foundation Models
链接: https://arxiv.org/abs/2609.34316
作者: Kang Yang,Gaofeng Dong,Liying Han,Mani Srivastava
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:This work studies a central gap in interpreting time-series foundation models (TSFMs): a dynamical property may be accessible in a hidden state even when the forecast fails to respond correctly as that property changes. We formalize these properties as Dynamical Parameters, including trend slope, oscillation frequency, and autoregressive dependence. We compare their representation accessibility, measured by recovery from hidden states, with their forecast response, measured by agreement with the expected forecast change. Across nine frozen TSFMs and thirteen laws, 42 of 63 model-parameter cells achieve accessibility above 0.95, whereas their median reference-aligned response relative to the conditional reference is only 0.46. To explain this gap, causal geometry compares the hidden-state change required to produce the reference response with the change induced by the parameter intervention. Directly modifying the hidden state recovers the reference response, but the parameter intervention often moves the state in a different direction. These results show that accessible parameter information need not be expressed in forecasts when input changes miss the required hidden-state direction.
[AI-212] One Sequence Many Decodings: CAGenMol-2 Recasts Drug Design as Masked Molecular Inference
链接: https://arxiv.org/abs/2609.34301
作者: Yanting Li,Enyan Dai,Lei Wang,Wen-Cai Ye,Li Liu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Drug design couples property evaluation, conditional generation, structure-based design, and local optimization, yet machine learning systems typically address these capabilities with separate task-specific models. We introduce CAGenMol-2, a masked diffusion molecular language model that represents molecules, continuous scalar properties, and 3D protein pockets within a single wrapped sequence. Within this pretrained interface, downstream operations are selected by which sequence regions are observed or masked at inference, allowing one checkpoint to perform property prediction, property- and pocket-conditioned generation, and partial-constraint design without task-specific architectures or backbone fine-tuning. We further propose Adaptive Fragment Optimization (AdaFO), a gradient-free mask-and-refill search that turns the masked decoder into an iterative local molecular optimizer. On CrossDocked2020, AdaFO increases Success Rate from 30.2% to 70.8%, the best reported under this protocol, while largely preserving drug-likeness and diversity. Finally, scaffold-preserving directional editing and CRBN/VHL case studies demonstrate its use in compound design workflows spanning local molecular editing, structure-based prioritization, and downstream simulation-based screening.
[AI-213] When World Models Lie: Adaptive Safety Analysis Under Wrong Imaginations
链接: https://arxiv.org/abs/2609.34300
作者: John Cao,Somil Bansal
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注:
Abstract:World models offer a powerful substrate for safety reasoning in high-dimensional robotic systems, but they are also fallible: their predictions can be biased, miscalibrated, or confidently wrong. This creates a central challenge for latent-space safety filters, which often learn Hamilton-Jacobi safety value functions on the dynamics of a world model. If the world model is incorrect, the resulting value function can inherit its errors and produce overconfident safety estimates. Existing latent safety filters often rely on auxiliary signals such as ensemble disagreement or value-target consistency residuals for adaptation, but these signals can remain small even when the world model’s predictions deviate from observations. We propose an adaptive latent safety filter that calibrates safety reasoning using directly observed world-model error. Our method uses Adaptive Conformal Inference to construct online uncertainty sets from discrepancies between predicted and observation-inferred latent states, then evaluates safety pessimistically by minimizing the learned value function over these sets. This allows the filter to remain minimally conservative when the world model is accurate, while becoming more cautious when observations reveal model mismatch. We provide a finite-time coverage guarantee for the adaptive uncertainty radius. Through simulation and hardware experiments, we show that our method significantly reduces failures relative to state-of-the-art latent safety filters while preserving task completion.
[AI-214] Direct Self-Evolving Optimization: Evolving LLM s without Challenger Training
链接: https://arxiv.org/abs/2609.34279
作者: Yuyang Deng,Yu Wang,Jiayun Wang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Self-evolving language models improve by generating tasks and learning from their own feedback, but adapting the task generator often requires a separate challenger-training loop. Can we generate tasks adapted to the current solver without explicitly training a challenger? We introduce \textbfDirect Self-\textbfEvolving \textbfOptimization (DEO), which replaces challenger parameter updates with solver-guided task sampling. The KL-regularized challenger objective defines an exponential tilt of a fixed base task distribution. DEO uses this distribution as a sampling target: a frozen LLM generates and mutates tasks, the solver scores them, and an approximate Metropolis selection rule refines the training pool. Only the solver is trained. Theoretically, for an idealized variant that samples exactly from the tilted distribution, and under regularity, local gradient-dominance, and initialization conditions, we show that DEO learns distributionally robust reasoning ability. In experiments, DEO achieves reasoning performance competitive with R-Zero while using over 50% less wall-clock training time, and improves reasoning accuracy over a no-walk ablation. Replacing the task generator with a frozen API-only LLM further improves the local solver, illustrating a capability enabled by removing challenger training.
[AI-215] Query Expansion and Key Specialization in Transformer Attention Geometry
链接: https://arxiv.org/abs/2609.34273
作者: Vidit Gupta,Siddhesh Nadkarni,Mihik Chaudhari,Vinaya Sawant,Prachi Tawde
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at Asian Conference on Machine Learning 2026
Abstract:The projection of queries and keys are central to the attention mechanism in Transformer architectures. While they are mathematically symmetric, they play different roles in attention mechanisms. The question of whether there is an effect from their functional distinction on their geometric development in training remains unanswered. We investigate the problem through the training of small GPT-like Transformers on character-level WikiText-103 for three different depths (4, 6, and 8 layers), three types of initialization for queries and keys, and four random seeds, resulting in 36 runs and 54 trajectories of average layers across seeds. We track the effective dimensionality of those layers using participation ratios and discover that effective dimension of queries expand while keys shrink, and that PR_Q - PR_K is positive in all trajectories studied. In connection to attention, the shrinking of keys leads to a narrower spectrum of QK^\top and more peaked attention weights. In order to determine if this connection is causal or coincidental, we directly control the spectrum of keys during training across five seeds: restricting it to make it shrink sharpens the attention with high directional confidence, while keeping it constant to the level of initial dispersion makes attention softer. Additional token-level checkpoint analyses show that the monotonic paired-contrast trend is not universal across pretrained families, but survives as an early-training regime that later decays over a full pretraining run, and the link between interaction-rank geometry and attention entropy remains visible in several models.
[AI-216] SAGE: Symbolic Action-Gating and Editing for LLM Task Planners
链接: https://arxiv.org/abs/2609.34268
作者: Trung Minh Bui,JongSul Moon,YoungOuk Kim,Quang-Ngoc Phung,Se-Woong Jun,Dongin Shin
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 8 pages, 2 figures, 1 table, 3 algorithms. Submitted to IEEE Robotics and Automation Letters (RA-L). Code and benchmark: this https URL
Abstract:Large language models (LLMs) are now the default cognitive core of embodied household agents, yet the plans they emit are rarely checked against a grounded model of the environment before execution, and the task-success they report is often measured on benchmarks so saturated that no method can be separated from another. We present SAGE (Symbolic Action-Gating and Editing), a single-LLM planner built from two lightweight mechanisms: a domain-agnostic symbolic gate (~250 lines of Python, zero tokens, O(|\pi|) ) that blocks precondition-violating actions with typed reasons as a runtime safety monitor, and a local edit that regenerates only the failed sub-goal’s suffix, keeping completed and untouched work intact; a hybrid seed+live memory store supports cold-start coverage. We evaluate under a leak-free protocol (leave-one-out retrieval) over five open-weight models and a 75-task AI2-THOR benchmark. On the standard benchmark goal-completeness saturates (52% of instances trivially solved) and SAGE ties strong hierarchical baselines. On a harder, method-agnostic multi-goal composition, SAGE’s completeness lead re-emerges large (+0.06 to +0.23 across four models). Under injected mid-execution failures, SAGE recovers as reliably as whole-plan replanners at 2.4-3.3x fewer LLM calls. As a verify-before-execute gate, the symbolic monitor blocks unsafe actions before actuation and raises simulator-reported step-success for every planner tested (up to +0.11), a signal the verifier never sees (non-circular). Because the gate calls no model (0.008 ms/plan), it is a safety layer that runs essentially free on the edge: SAGE planning reproduces its quality on a Jetson AGX Orin, where small-model verification helps most. We release the benchmark, the leak-free protocol, the recovery and safety-gate harnesses, and a verifier-portability study (auto-induced on ALFWorld, 0.89 held-out).
[AI-217] Maintaining Benchmarks Against Increasingly Capable Agents : Detection and Remediation of Unearned Passes
链接: https://arxiv.org/abs/2609.34262
作者: Weijun Luo,Kelvin Luu,Xinyi Liu,Guangze Luo,Miguel Romero Calvo,Soham Dan,Daniel Yue Zhang,Ying Liu,Mohamed Elfeki
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Software Engineering (cs.SE)
备注:
Abstract:Agentic benchmarks guide model selection and training. Yet an agent can pass a task without demonstrating the intended capability. Such outcomes constitute unearned passes; their proportion among all passes defines the integrity gap. As agents improve, benchmark surfaces that once seemed harmless can become exploitable, making benchmark validity an ongoing maintenance problem. We introduce a process-verification framework that audits passing trajectories, distinguishes evidenced reward hacking from verifier weakness, and localizes exploitable surfaces for repair. Across 3,810 passing trajectories from 29 model-benchmark cohorts, confirmed violations often increase with model generation but not monotonically. On SWEBench Pro V1.0, confirmed violation rates rise from 24% to 73% between Opus 4.7 and Fable 5 on matched tasks; later cohorts fall to 11% for Fable 5.1 and 0% for GPT-6 Astra. These comparisons are descriptive: configurations were not normalized, and the latest models also pass fewer exploitable tasks. Violations concentrate around a small set of recurring surfaces, especially unintended access to reference solutions through git history. Three repair case studies across two benchmarks show why blocking a recorded exploit is insufficient: the same protected information can remain accessible through another route. Therefore, we combine minimal patches with exploit replay and fresh agent evaluation, auditing new passes under the original standard. No evaluated attempt against the final patches reached the protected channel, and every post-patch pass was judged legitimate. Benchmark integrity requires ongoing maintenance: audit passing behavior, repair the enabling surface, and re-evaluate both exploit access and legitimate solvability.
[AI-218] QuantaSpike: Short-Window Spike-Driven Quantization for Large Language Models
链接: https://arxiv.org/abs/2609.34259
作者: Bang Hu,Guowei Zhu,Changze Lv,Xiaoqing Zheng,Fengzhe Zhang,Fan Zhang,Wei Cao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) achieve strong performance across many tasks but rely on dense multiply-accumulate (MAC) operations during inference, resulting in high energy cost. Spiking neural networks (SNNs) offer an event-driven alternative in which synaptic integration uses lightweight accumulation. However, spike-driven LLM inference remains difficult because outlier-heavy activations typically require long firing windows or auxiliary non-spiking paths. We propose QuantaSpike, a short-window spike-driven quantization framework for LLMs built around Logarithmic Ternary Integrate-and-Fire (LTIF) neurons. LTIF uses ternary events with power-of-two membrane-response quanta, improving the information represented by each firing step while retaining shift-ACC-compatible computation. QuantaSpike combines this neuron with group-adaptive gain and selective outlier admission: normal values use residual LTIF steps, whereas admitted outliers receive one additional onset spike before entering the same residual dynamics. Across OPT and Llama-2, QuantaSpike achieves state-of-the-art or competitive perplexity and zero-shot accuracy among spike-driven LLM quantization methods. It also transfers to newer dense LLMs, remaining close to the FP16 reference on Llama-3-8B and Qwen3-8B under the same four-step firing window. Analytical linear-energy projections show that QuantaSpike reduces the energy of one linear transformation by about 80.0% on OPT models and 67.1% on Llama-2 models relative to SpikeQuant, providing an accurate and energy-efficient spike-driven path for LLM inference.
[AI-219] Investigating Human–AI Discrepancies via Multiple-Solution Problems
链接: https://arxiv.org/abs/2609.34258
作者: Zihao Wang,Francesco Insulla,Andrea Montanari
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: 50 pages, 16 pdf figures
Abstract:Frontier artificial intelligence (AI) models are benchmarked on whether they reach a correct answer. Yet many problems admit several correct answers and repeated attempts, by different people or by the same model resampled, trace out a distribution over them. In this work, we ask whether human and model reasoning lead to different distributions over valid solutions. Our testbed comprises 270 reasoning puzzles across five puzzle families. These multiple-solution puzzles each have 3 to 8 valid solutions and are simple enough that humans and models can solve them reliably. The resulting distributions differ markedly: models differ from one another, yet resemble each other far more than they resemble humans. Model distributions are, moreover, within every puzzle family, less diverse than human ones. We compare these discrepancies across puzzle categories, and trace how they respond to reasoning-effort settings, to prompting, and to perturbations of the puzzle that leave its solutions unchanged. Together, these results point at significant differences between human and AI problem-solving processes, and their choice among equally defensible solutions. As progressive deployment of AI systems in society comes into focus, evaluating such differences (beyond one-dimensional accuracy metrics) is increasingly important. Data and code are available at this https URL
[AI-220] WAM-OPD: Sharpening World Action Models via On-Policy Distillation
链接: https://arxiv.org/abs/2609.34250
作者: Panjun Liu,Xiaohan Lei,Shiqi Zhang,Yikun Wang,Yongxin Zhang,Mingyi Hu,Shida Sun,Jiateng Shou,Wengang Zhou,Jiajun Deng,Zhiwei Xiong
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Pretrained world action models (WAMs) provide generalist capabilities across diverse robotic manipulation tasks, yet improving target-task performance to an expert level without degrading pretrained skills remains challenging. We explore on-policy distillation (OPD) for WAMs and introduce WAM-OPD. WAM-OPD inherits the advantage of OPD methods that transfer task-specific teacher knowledge under the student’s own induced distribution, rather than directly fitting the student to a narrow task-specific data distribution. However, in closed-loop manipulation, the observation histories change as the student policy evolves, requiring fresh environment rollouts to remain on-policy. Applying OPD to WAMs entails repeated data collection, which is costly even in simulation and often impractical on real robots. To avoid repeated environment rollouts during distillation, we introduce prefix-weighted trajectory replay (PWTR). PWTR uses a fixed trajectory pool composed primarily of initial-student rollouts, supplemented with task-specific teacher rollouts to broaden trajectory coverage. For each trajectory replayed from this pool, PWTR conditions the current policy on successive stored histories to generate fresh denoising paths, along which the task-specific teacher provides supervision. Although these denoising paths are refreshed as the policy evolves, the replayed environment trajectories remain fixed. PWTR therefore reweights per-decision distillation losses using proxy importance weights derived from path scores accumulated over the trajectory prefix preceding each decision to mitigate the resulting shift in the history distribution. Simulated and real-world experiments demonstrate task adaptation without additional environment interaction during distillation. In both settings, WAM-OPD improves target-task performance while retaining near-initial performance on tasks excluded from adaptation.
[AI-221] Evolving Support Priorities in Empathetic Reinforcement Learning
链接: https://arxiv.org/abs/2609.34249
作者: Pengyu Huang,Zhiyuan Han,Wenwen Tong,Hewei Guo,Jiangnan Chen,Sirui Chen,Lewei Lu,Beier Zhu,Xun Yang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We identify a fundamental mismatch in empathetic reinforcement learning: support priorities evolve with the dialogue state, yet existing methods typically optimize predefined reward specifications that remain fixed across turns. To model these evolving support priorities, we organize empathetic support along cognitive, affective, and proactive empathy, and propose Context-Adaptive Rubric Evolution (CARE). At each turn, CARE generates a context-adaptive rubric by adjusting both the weights of these three empathy dimensions and their fine-grained evaluation criteria. The rubric generator is trained with turn-level rubric supervision and human preference data through supervised fine-tuning followed by preference-based reinforcement learning, and then serves as an adaptive reward interface for online empathetic RL. Integrated with both RLVER and MICA, CARE achieves state-of-the-art performance across SentientBench, EQBench3, and EMPA under three independent LLM judges. Notably, on EMPA, CARE improves EPM-Idx over the strongest baseline by at least 13 points under all three judges, including an increase from 28.11 to 83.54 under Gemini-2.5-Pro. Further analyses show that learned rubric priorities systematically vary across dialogue stages and user emotions, demonstrating that CARE adapts what is rewarded as support needs evolve.
[AI-222] Certified Multi-Source Integrity for Structured Agent Actions
链接: https://arxiv.org/abs/2609.34245
作者: Anmol Pandey,Aditya Jain,Liang Chen,Carsten Maple,Christo Panchev
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted to the UK AI Conference 2026 (UKAI 2026), to appear in Proceedings of Machine Learning Research. 14 pages, 3 figures, 7 tables. Code and data: this https URL
Abstract:LLM agents increasingly take privileged, often irreversible structured actions, such as paying an invoice. They assemble each action from action-critical fields in documents and tool outputs that an adversary can corrupt, and indirect prompt injection can drive the model itself to extract attacker-chosen values. Current defenses gate on a source’s trust label or certify free-text answer quality. None certifies the integrity of a coupled, policy-bound structured action under a corruption budget that accounts for shared upstream sources. We characterize when such an action is safely certifiable and give the maximally live safe certifier. It admits an action only when each field clears the rule its evidence structure supports: a bounded corruption radius over corruption-distinct evidence classes, counted by a minimum hitting set so that re-publishing or laundered copies cannot manufacture a quorum, deterministic reconciliation for complementary fields, and a trusted anchor where the evidence leaves a field single-sourced. We formalize two robustness notions, validate each mechanism by ablation, and measure how often the multi-source precondition holds on sanctions designations (70,966 entities) and software supply-chain provenance (450 packages). Under upper-bound proxies, genuine corroboration is a minority phenomenon in both, and naive attestation counting overstates it, since witnesses that look independent collapse to two corruption-distinct domains once shared origin is counted. Across five current models in a real agent loop, a realistic injection fools every model but one and a naive agent then executes the fraudulent action on most attacks. The certifier admits no unsafe action and recovers the correct value where corroboration permits, while action-gating and provenance baselines are broken in every world of our harness by some attack in its space.
[AI-223] Stashbird: Efficient Speaker-Indexed Memory for Conversational Agents
链接: https://arxiv.org/abs/2609.34242
作者: Chidera Biringa,Lucas Yannul,Xiaowen Wang,Marco Ayala,Nicholas Yi,Alex Moyse,Nishant Manchanda,Vivek Gupta
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:AI agents require memory that preserves information across user-agent exchanges, user-to-user conversations, and group conversations with or without agent participation, while supporting updates as evidence changes or is removed. We present Stashbird, an agent memory system that links source episodes to derived memory state through explicit provenance. Stashbird organizes memory into episodic records, semantic relations, community summaries, and persisted graph state, with lifecycle operations for incremental updates and episode-level deletion. We evaluate question-answering accuracy and model-facing workload across four long-term memory benchmarks. On LoCoMo, Stashbird uses 76.4x fewer ingestion prompt tokens than Graphiti. Compared with reproduced Hindsight on the same benchmark, it uses 8.1x fewer retrieval prompt tokens, with accuracy 1.6 percentage points lower. It achieves higher accuracy than Hindsight on LongMemEval-S and GroupMemBench and comparable accuracy on EverMemBench.
[AI-224] AdaGuard: An Adaptive Guard Model with User-defined Policies
链接: https://arxiv.org/abs/2609.34241
作者: Yunhao Feng,Yifan Ding,Yuxiang Xie,Zheng Li,Mingrui Lao,Zeyuan Wang,Yanming Guo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Guard models support the safe deployment of language model agents, but fixed risk taxonomies limit their ability to accommodate requirements that vary across applications and tasks. Under user-defined policies, detecting violations requires interpreting both the applicable rules and the agent’s behavior, since identical actions can receive different judgments under different policies. To support learning this capability, we introduce AdaptiveSafety, a dataset of 10,939 training examples and 1,000 test examples covering policies with 1–100 rules. The dataset combines trajectories from multiple sources with policy and behavioral counterfactuals, pairing each example with an explanation and the complete set of violated rules. These counterfactuals expose changes that alter compliance, while structural augmentations provide supervision for consistency under rule reordering and identifier remapping. Building on this supervision, we propose SafePO, a reinforcement learning algorithm for refining violation identification while balancing explanatory reasoning and final verdicts. SafePO uses structured rewards to assess prediction correctness, retains group-relative advantages at the response level, and employs a separately trained value model to modulate token weights within explanation and verdict regions. Separate normalization controls their relative contribution to training despite differences in length. Through supervised initialization followed by SafePO, we develop AdaGuard, a family of 0.6B, 4B, and 8B guard models that assess agent trajectories under policies supplied at inference time. Our 4B model achieves binary accuracies of 89.30% on AdaptiveSafety and 71.82% on DynaBench. The project repository is available at this https URL
[AI-225] GlyphBench: A Playground for Language-Model Reinforcement Learning
链接: https://arxiv.org/abs/2609.34214
作者: Roger Creus Castanyer,Marc-Alexandre Côté,Matthew James Sargent,Augustine N. Mavor-Parker,Glen Berseth,Pablo Samuel Castro
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce GlyphBench, an environment suite for reinforcement learning (RL) post-training of language-model agents, with over 360 tasks spanning diverse games. GlyphBench renders spatial observations as two-dimensional Unicode grids and connects training, evaluation, and trajectory replay through a unified interface designed to support efficient and reproducible research. We use GlyphBench to study how observation interfaces, reasoning effort, and agent harnesses affect performance, and how RL configurations shape learning dynamics. Our results show that glyph observations outperform native text and pixels in our Craftax experiments, with further gains on several BALROG environments. RL on 100 GlyphBench tasks improves Qwen3.5-4B on held-out Reasoning Gym problems, reaching 63.48% accuracy and outperforming the base model, a math-trained baseline, and a code-trained baseline. These experiments provide empirical evidence that reasoning gains from gameplay can yield stronger transfer than math or code. Together, these results highlight GlyphBench’s value as a testbed for systematic research on how language-model agents learn, interact, and generalize.
[AI-226] Behavior-Grounded Semantic Enrichment for Financial Fraud Modeling and Reasoning
链接: https://arxiv.org/abs/2609.34211
作者: Linbo Shao,Huilin He,Yating Lou,Dawei Cheng
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:In financial fraud detection, rich semantic context can provide important evidence for transaction behavior modeling and fraud reasoning. However, public real-world financial datasets often lack rich semantics due to privacy constraints. Consequently, synthetic datasets incorporate generated semantics, but at the cost of behavioral realism; textual descriptions for contextual reasoning remain scarce. We address this gap through a semantic enrichment framework grounded in original transaction behavior to simulate multimodal financial data. We (1) propose a multi-agent semantic enrichment framework that generates interpretable financial semantics grounded in transaction behavior through role-specialized agents and consistency refinement, and (2) newly contribute a valuable multimodal financial fraud dataset, MS-FFSD, enriched with structured semantics and textual semantics while preserving real-data-grounded transaction behavior. Furthermore, we systematically analyze the quality and utility of semantic enrichment. Results demonstrate statistical fidelity and framework generalizability, while showing that richer semantics benefit fraud modeling and context-aware LLM reasoning. Overall, this work advances multimodal financial fraud research and bridges emerging LLM and multi-agent capabilities with operational anti-fraud practice. The framework and dataset are released at this https URL.
[AI-227] EntroPack: Fast and Accurate Entropy-Coded Weight Compression at Arbitrary Bitrates
链接: https://arxiv.org/abs/2609.34185
作者: Hong Zhang,Zhongjie Duan,Yingda Chen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Weight compression helps large neural networks fit deployment memory budgets, but common fixed-width formats offer only coarse storage choices. Entropy coding supports finer rates, yet the achieved size depends on the quantized weight distribution and coding overhead. Exploiting this flexibility requires accurate rate selection and efficient weight reconstruction for inference. We present EntroPack, an entropy-coded weight compressor that supports arbitrary target bitrates without activation calibration or fine-tuning. It combines row-normalized E_8 lattice quantization with a conditional probability model of lattice coordinates. Sampled storage estimates select the quantization resolution without repeated full-stream encoding. The final coordinates are entropy-coded in independently decodable tiles, enabling fast, fused symbol decoding and numerical weight reconstruction on the GPU. EntroPack supports floating-point and integer weight containers, such as BF16, FP16, FP8, and INT8, with storage bitrate controlled independently of numerical precision. Online decoding adds latency that grows with weight count, making the method well suited to compute-intensive workloads such as diffusion denoising and Transformer prefill. Experiments demonstrate fast encoding and modest inference overhead in these settings. When compressing the linear-layer weights of the image generator Z-Image-Turbo, EntroPack achieves substantially lower weight and denoiser output errors than fixed-width formats at comparable storage rates, with modest denoising-step overhead. Targeting 4 bits per parameter, it achieves lower weight and denoiser output errors than NF4, including about 24% lower relative L_2 weight error, with less storage. Source code is available at this https URL.
[AI-228] CASS: Contribution-Aware Structured Sparsity for Model Merging NEURIPS2026
链接: https://arxiv.org/abs/2609.34184
作者: Yan Li,Guiping Cao,Meng Xu,Tao Jiang,Yaguang Song,Ming Tao,Yaowei Wang,Dongmei Jiang
类目: Artificial Intelligence (cs.AI)
备注: Accepted to NeurIPS 2026
Abstract:Model merging integrates task-specific fine-tuned models into a single multi-task model, but often suffers from parameter interference caused by conflicting task-vector updates. Existing methods typically mitigate conflicts by pruning task vectors based on weight magnitude or random heuristics, treating Transformers as unstructured ``bags of parameters’’ and overlooking their inherent modularity. In this paper, we propose \textbfContribution-\textbfAware \textbfStructured \textbfSparsity (CASS), a unified framework that reduces parameter interference by identifying and preserving task-specific components. At the core of CASS is a contribution-aware structured mask that identifies task-relevant attention heads and FFN neurons. We instantiate this mask in two settings: CASS-Merging, the primary post-hoc setting where masks serve as a plug-and-play denoising filter for existing merging operators, and CASS-Tuning, an extension for scenarios with fine-tuning access where masks constrain gradients to reduce structural overlap between task vectors. Our analysis shows that task-relevant components are sparse and partially disjoint, supporting structured component-level filtering as an effective way to reduce merging interference. Extensive experiments across vision (ViT, 20 tasks) and language (RoBERTa, 8 tasks; Qwen2.5, 4 tasks) benchmarks demonstrate that CASS improves a range of representative merging baselines.
[AI-229] Unified Visual-Tactile-Action Modeling from Human Demonstrations for Dexterous Manipulation
链接: https://arxiv.org/abs/2609.34182
作者: Wenqiao Li,Qianyou Zhao,Jiawen Hao,Xuezhou Zhu,Tengyu Liu,Kaifeng Zhang,Chuan Wen,Siyuan Huang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Dexterous manipulation requires tactile this http URL, robot tactile demonstrations are difficult to scale,because dexterous-hand teleoperation provides limited tactile feedback to the operator. In contrast, human demonstrations offer a substantially more scalable source of diverse tactile interactions. Motivated by a simple premise: hands can change, but the underlying physics of interaction does not. We leverage human tactile data to improve dexterous manipulation policies. Specifically, we first build a tactile motion-capture system that synchronously records images, tactile signals, and hand motions. Using this system, we construct the UVTA dataset spanning five contact-rich tasks, with 1,000 human demonstrations covering diverse interaction patterns and 150 robot demonstrations per task. To transfer the underlying physics of human interaction to robot control, we propose a Unified Visual-Tactile-Action Model that maps both embodiments into aligned tactile and action representations and jointly predicts future action and tactile trajectories. The joint objective enables human demonstrations to supervise contact-aware representation learning, while only robot actions are executed during deployment. In real-robot evaluations across five tasks, our method achieves an average success rate of 70%, outperforming the strongest visual-tactile baseline, which achieves 29%, and an architecture ablation, which achieves 42%. Performance improves consistently with additional human demonstrations and exhibits no saturation at 1,000 demonstrations per task, validating the effectiveness of scalable human tactile data for dexterous manipulation. Project page is available at this https URL.
[AI-230] Efficient Reasoning via Constrained Optimization in Latent Space NEURIPS2026
链接: https://arxiv.org/abs/2609.34181
作者: Zhinan Hou,XingChen Li,Keyou You
类目: Artificial Intelligence (cs.AI); Optimization and Control (math.OC)
备注: Accepted by NeurIPS2026
Abstract:Large Reasoning Models (LRMs) have shown remarkable reasoning capabilities, yet they still suffer from overthinking, generating redundant reasoning steps which incur substantial token consumption. Existing methods, such as suppressing reflective keywords or forcing shorter reasoning lengths, attempt to mitigate this issue but inevitably truncate necessary steps and induce underthinking, thereby compromising performance. To address this dilemma, we investigate the latent representations and observe that efficient reasoning steps naturally cluster into a concentrated region in latent space, while those deviating from this region tend to produce verbose sequences. To leverage this, we keep reasoning focused within this region via a quadratic program which projects deviating hidden states back into the region. Then we propose a novel training-free framework to achieve efficient reasoning that reduces token generation costs without sacrificing performance. Extensive experiments conducted on four models ranging from 1.5B to 14B, and across six benchmarks in math reasoning, coding, and scientific QA, validate the effectiveness of our method, up to a 12.1% improvement in accuracy while reducing generated tokens by 11.8% to 52.8%. Codes are available at \hrefthis https URLthis https URL.
[AI-231] GradLev: Token-Parallel Test-Time Training Via Costate Prediction
链接: https://arxiv.org/abs/2609.34174
作者: Bo Liu,Qiang Liu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Test-Time Training; Language Modeling
Abstract:Test-time training (TTT) allows a model to improve its predictions at inference time by updating weights after every observed token. However, sequential gra- dient writes make parallel training difficult. We observe that, given layer inputs and activation gradients (costates), online gradient descent admits exact parallel scans for both forward evaluation and reverse backpropagation. GradLev lever- ages this duality: a causal auxiliary network predicts costates across all tokens in parallel; associative scans compute the adapted weights and forward activations and propagate gradients backward; and the resulting gradient targets supervise the predictor via a consistency loss. Exact consistency guarantees exact recovery of the sequential online learner. At deployment, the auxiliary predictor is discarded, and the model updates natively via token-by-token forward and backward passes.
[AI-232] RoutePrism: Tracing Construction Order Effects in Agent Memory
链接: https://arxiv.org/abs/2609.34160
作者: Dong Xu,Zhangfan Yang,Jiantao Wu,Shipeng Zhang,Zexuan Zhu,Jiangqiang Li,Jun Zhang,Junkai Ji
类目: Artificial Intelligence (cs.AI)
备注: 55 pages, 5 figures
Abstract:Processing the same records in a different order can discard different evidence, yet endpoint accuracy alone cannot reveal what changed or whether it mattered. We introduce RoutePrism, a diagnostic protocol that builds memory twice from the same source pool in two processing orders, then traces which sources, compiled contexts, and answers differ. Because record content, timestamps, policy, and the answer model all stay fixed, any observed difference is localized to the memory construction step. A matched four-condition intervention tests whether a record displaced by reordering actually carried task-relevant evidence: restoring that single record recovers over 60 percentage points of lost accuracy, while substituting a non-supporting record of equal length does not. We evaluate the protocol on PersonaMem-32K (63 primary queries, 29 users) and 470 LongMemEval-S questions with histories spanning 38 to 62 sessions, replicating the core intervention across five answer models. Survivor selection, defined as the choice of which record a cluster retains, drives most source-level changes, while different memory policies (compaction, bounded recency, MemoChat-style summarization, A-MEM) produce distinct failure signatures at the source, context, and metadata layers.
[AI-233] ableSeek: Structure-Preserving Agent ic Evidence Seeking over Heterogeneous Table Corpora
链接: https://arxiv.org/abs/2609.34157
作者: Jiaming Tian,Liyao Li,Wentao Ye,Haobo Wang,Lihua Yu,Zujie Ren,Gang Chen,Junbo Zhao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Open-domain table retrieval seeks tables that contain sufficient evidence for answering a question or verifying a claim. Yet semantic relevance is often misleading: topically similar tables may lack the required facts, while answer-bearing evidence is often confined to a few cells whose meaning depends on surrounding schema and table context. Heterogeneous schemas, value formats, and serializations further weaken one-shot matching. We present TableSeek, a structure-preserving agentic search framework for heterogeneous table corpora. Instead of ranking tables once, an LLM agent iteratively follows sparse clues, inspects schema-preserving previews, identifies schema- and value-level mismatches, and refines its investigation. TableSeek uses cells and schemas as evidence anchors while retaining complete tables as evidence units, enabling fine-grained localization without losing the context required for interpretation and answerability checking. Without relying on retriever training or a precomputed semantic index, TableSeek produces transparent evidence-seeking trajectories and achieves competitive end-to-end performance against strong retrieval-and-reranking pipelines on heterogeneous table benchmarks. These results suggest that active, structure-preserving evidence seeking is a promising paradigm for open-domain table retrieval. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.34157 [cs.AI] (or arXiv:2609.34157v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.34157 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-234] Self-Evolving Agents via Likelihood-Guided Tool-Space Optimization
链接: https://arxiv.org/abs/2609.34151
作者: Xuanqi Zhang,Ruinan Jin,Running Yang,Yuxuan Zhang,Minghui Chen,Wenlong Deng,Xiaoxiao Li
类目: Artificial Intelligence (cs.AI)
备注: 39 pages
Abstract:Self-evolving agents can continually improve their behavior, while tools define the executable action space through which they interact with the environment. However, exposing the full tool library to model introduces substantial irrelevant context and can impair tool-use decisions. We study tool-space self-evolution, where each recurring task type maintains a persistent tool space which is constructed from accumulated output experience. We identify three limitations of existing methods: (1) output-unaware selection: they rely primarily on tool descriptions or model priors rather than observed tool outputs; (2) statelessness across request: they select tools independently for each request without consolidating prior output experience into persistent task-specific state; (3) inference cost: they repeatedly search, rank, or reason over candidate tools for subsequent requests of the same task. We address these limitations through output-aware tool scoring, persistent task-specific tool spaces, amortized tool selection, and reusable configurations across models. We introduce LOTS (Likelihood-Only Tool Scoring), which evolves an agent’s tool space from accumulated output experience while keeping model parameters fixed. After each request, LOTS holds the model’s generated answer and estimates each tool’s contribution by measuring how much the answer likelihood changes when its observed output is removed. These contributions are aggregated within each recurring task to rank tools and update its persistent space. Across three benchmarks, LOTS improves task performance while substantially reducing tool context. More importantly, sequential experiments demonstrate that task-specific spaces persist and continue to improve over time, while cross-model experiments show that learned configurations transfer across different models.
[AI-235] Same Tasks Different Apps: Why Mobile GUI Agents Fail to Generalize? EMNLP2026
链接: https://arxiv.org/abs/2609.34139
作者: Tien Tran,Namho Koh,Daiki E. Matsunaga,Ayush Jain,Kee Eung Kim
类目: Artificial Intelligence (cs.AI)
备注: Accepted to Findings of EMNLP 2026. 30 pages
Abstract:Mobile GUI agents deployed in real settings must work across different applications that support the same functionality. Most existing benchmarks test each task in only one app, so a high score can mean the agent understands the task, or only that it knows that particular app. We introduce AnyAppBench, a category-controlled live Android benchmark that evaluates cross-application generalization while keeping the user goal fixed. It spans 10 functional categories, 100 task templates, and 520 task–application pairs over 52 applications. Agents run from raw instructions and with app-independent sub-goals, and a VLM judge labels every failed run under a fixed failure taxonomy whose reliability is measured by human annotation. We find that, across 13 agents, success on the original application does not transfer reliably to new applications with the same goal. Furthermore, providing high-level sub-goal decomposition produces only small, category-dependent changes that do not close the gap, and the mix of failure types changes with the target interface. Based on those insights, we believe the AnyAppBench benchmark provides an important stepping stone toward robust real-world deployment of mobile GUI agents. Our code, data and the leaderboard can be found at the project website this https URL.
[AI-236] You Cant Have It Both Ways: Concept Entanglement Limits Diffusion Model Unlearning NEURIPS2026
链接: https://arxiv.org/abs/2609.34137
作者: Yian Wang,Ali Ebrahimpour-Boroojeny,Hari Sundaram,Varun Chandrasekaran
类目: Artificial Intelligence (cs.AI)
备注: 40th Conference on Neural Information Processing Systems (NeurIPS 2026)
Abstract:Concept unlearning in text-to-image diffusion models aims to suppress a target concept (e.g., \texttthorse) while preserving related but distinct content (e.g., \textttdonkey), yet existing methods either leak under indirect prompts or visibly degrade other concepts. We show that these failure modes stem from the geometry of concept representations rather than from any particular algorithm. Formalizing concepts as activation-space regions, we prove that the overlap between a target and other concepts lower-bounds the damage any robust erasure must inflict on them, with the trade-off scaling linearly in the degree of overlap. Across thirteen unlearning methods, including methods designed to preserve non-target concepts, no method achieves both strong erasure and strong neighbor preservation: STEREO nearly eliminates indirect leakage but cuts neighbor generation by more than 75%, while sparse inference-time methods preserve neighbors but leak. Damage increases with our overlap measure, monotonically so for STEREO; the \kappa -scaling reproduces on SDXL, and neighbor-selective damage recurs on FLUX. Perfect unlearning is the wrong target for entangled concepts; methods should be evaluated on the Pareto frontier our theorem establishes.
[AI-237] Waggle: Learning One Anonymous Local Law for Self-Organizing LLM Swarms
链接: https://arxiv.org/abs/2609.34136
作者: Mingxi Zou,Wei Zhu,Zhuo Wang,Langzhang Liang,Zhiwen Tang,Yinghui Xu,Zenglin Xu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As LLM agents increasingly collaborate on complex tasks, how to organize their interactions becomes a central design question. Existing multi-agent systems typically learn or adapt explicit roles, hierarchies, routing policies, or communication topologies. We shift the learning target to a reusable local law that can be shared across interchangeable agents and adapt coordination as populations or interaction conditions change, without redefining a global organization. We introduce Waggle, a shared anonymous policy over bounded local views that jointly selects task actions, semantic communication, and local commitment updates. Repeated execution of the same law allows coordination to form, persist, and reorganize online without explicit roles or global topology. To learn this law across interchangeable agents and evolving coordination, we develop Swarm-Consistent Distillation (SCD), combining anonymous-orbit consistency with rollout-grounded prediction of the next local coordination field, with no added inference-time components. Across diverse coordination settings, the same learned law remains effective as populations and interaction budgets change, retains over 96% of substrate-specific oracle quality, and transfers without retraining; SCD further improves reorganization after counterevidence. Together, these results show that LLM-agent organization can emerge and adapt through repeated execution of a learned local law.
[AI-238] Evo2Team: When Do Evolved Skills Transfer? From Selection to Deployment
链接: https://arxiv.org/abs/2609.34135
作者: Renxiang Wang,Jiaming Cui
类目: Artificial Intelligence (cs.AI)
备注: 26 pages, 11 figures
Abstract:A skill bank that helps one multi-agent system may leave another’s behavior unchanged. A transferred rule helps only when target agents act on it successfully. We study this path for routing and communication skills in Count-Frequency and AgentsNet, using teams of 4–32 agents and GPT and Qwen model ladders. Source evolution meets a joint quality, cost, model-tier, and confirmation goal in 14 of 16 settings. We then evaluate Evo2Team, which selects, adapts, and confirms source skills for the target team, alongside six frozen selectors across 28 transfer directions. Evo2Team’s target-side exploration cost is below that of evolving a new target bank in every direction, even when reused reference evaluations are charged once. Twenty of 28 held-out outcomes meet the positive-transfer criterion, including three saved diagnostic tests. Selection alone does not explain these outcomes: KNN and CORAL choose different banks in two AgentsNet directions but produce identical recorded executions. When Evo2Team changes execution, gains can reach many tasks, as in a Count-Frequency direction that improves 28 of 32 tasks over KNN. Seven positive AgentsNet outcomes save 6.1–14.6% in deployment cost while using transferred skills on only three to six of fifteen tasks. In five earlier accepted directions, all 22 task records using transferred skills pass three fixed-graph confirmations, but four fail in recorded executions on new graphs. Graphs and model responses change together in this comparison. These results show that skill transfer must be assessed through the actions agents take, the tasks those actions reach, and the quality and cost of the final deployment.
[AI-239] StateGuard: Analytical-State Management with Validity-Aware Intervention for Long-Horizon Data Agents
链接: https://arxiv.org/abs/2609.34134
作者: Wenle Liao,Zhao Wang,Jingchao Zhang,Jiajie Jin,Yimeng Xu,Zhicheng Dou
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLM-based agents have shown strong capabilities in automated data analysis and are increasingly moving toward long-horizon, multi-stage analytical workflows. However, as the analytical process evolves, constraints, variables, and conclusions remain implicitly embedded in interaction histories, making it difficult for agents to track which analytical artifacts remain valid over increasingly long horizons and changing dependencies. Consequently, stale artifacts may be silently inherited, propagating errors to downstream stages. To address this challenge, we propose StateGuard, an analytical-state validity management framework for long-horizon data agents. StateGuard externalizes evolving analytical progress into a state graph containing constraints, versioned variables, intermediate conclusions, and cross-state relations, treating each state as an executable, verifiable, and traceable object rather than textual memory alone. StateGuard maintains state validity through evidence-grounded verification and hierarchical intervention. To equip StateGuard with these capabilities, we first introduce Manager-Oriented Counterfactual Supervision, which constructs 3K state-centric trajectories through counterfactual runtime synthesis to fine-tune StateGuard for state maintenance, verification, and repair. We then apply Validity-Guided Policy Optimization, using runtime validity evidence to provide fine-grained learning signals for protocol correctness, state grounding, and intervention quality. Experiments on three diverse long-horizon data-analysis benchmarks show that StateGuard consistently improves data-agent performance while reducing dependency-induced downstream error propagation, demonstrating the advantages of explicit analytical-state management for reliable long-horizon data analysis.
[AI-240] From Attack Success to Attack Severity: Counterfactual Memory Attacks on LLM Agents
链接: https://arxiv.org/abs/2609.34132
作者: Mingxi Zou,Langzhang Liang,Zhuo Wang,Yiyang Zhao,Lizhen Qu,Zenglin Xu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As LLM agents increasingly rely on persistent memory for long-horizon and personalized behavior, they can retain and reuse information across interactions, but this also creates a lasting channel through which malicious memory writes can influence future behavior. Persistent-memory attacks are typically evaluated by whether they succeed, yet successful attacks can leave persistent states with substantially different downstream consequences. We study this severity as a distinct attack-design objective and formalize it with counterfactual memory regret (CMR), the paired increase in expected downstream loss relative to clean memory. We introduce MemHarm, which predeclares a finite class of sparse, grounded semantic edits, evaluates candidates through the normal agent memory interface using offline paired-loss feedback, and certifies resolved selections within that class. Compared with attack-success optimization, CMR-guided selection produces substantially larger downstream loss while retaining most of the success-rate gain. Across two agent benchmarks and diverse memory designs, MemHarm attains the highest CMR point estimates among the evaluated general attacks on identical support. Factor-removal interventions link this harm to the selected semantic factor, and native-agent deployments verify the write-to-fresh-process attack path.
[AI-241] JET: Judge-Guided Evolution at Test Time for Agent Programs
链接: https://arxiv.org/abs/2609.34126
作者: Yao Long Teng,Jiayi Cai,Bo An
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:An agent’s executable program governs how it uses tools, processes observations, and responds to failures. Evolving this program at test time can help adaptation, but deciding which changes to retain is difficult when true rewards are unavailable. Execution traces provide evidence of agent behavior, yet interpreting that evidence requires a judge that remains useful as tasks and candidate programs change. We introduce Judge-Guided Evolution at Test Time (JET), which evolves an executable judge on labeled source trajectories, then freezes and transfers it to guide target-side program evolution. The judge supplies scores and diagnostic feedback without target evaluator access or model-weight updates. On unseen WebShop tasks, JET achieves approximately 13% higher mean reward than fixed-rubric guidance when evolution begins from an unevolved program (cold start) and 4% higher when it begins from one already optimized on source tasks (warm start), with a 36% relative improvement in cold-start exact success. An exact-judge control on PushT, where the judge reconstructs the scoring rule from observations, shows that without judge error, program search becomes the bottleneck. Analyses identify useful reward-prediction logic in the evolved code and show that better final selection alone cannot explain the gains. These results support executable judge transfer for program adaptation under evaluator-preserving task shifts.
[AI-242] Probabilistic electrical power demand forecasting with uncertainty quantification
链接: https://arxiv.org/abs/2609.34120
作者: Mahesh Neupane,Pragya Dhungana,Pradip Khatri,Swechhya Baskota,Hariom Dhungana
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:The majority of research on electricity consumption forecasting has focused on deterministic approaches, which generate a single point estimate for each time step in the forecasting horizon. However, the increasing penetration of renewable energy sources and the growing complexity of modern smart grids have introduced greater variability and uncertainty into power-system demand and operation. Consequently, probabilistic forecasting, which quantifies the uncertainty and variability associated with future electricity demand, is becoming increasingly important for reliable power-system planning and operation. This study presents an empirical comparison of four contemporary probabilistic forecasting models for electricity consumption, highlighting their respective strengths and limitations. We have performed comparision on real-world power systems related datasets. Across all power-consumption zones, NGBoost demonstrates superior probabilistic forecasting performance, achieving the lowest MAE and RMSE while providing well-calibrated uncertainty estimates with high prediction-interval coverage and reasonably narrow intervals. These results indicate that NGBoost offers a more accurate and reliable forecasting framework than Bayesian, Monte Carlo (MC) Dropout, and Gaussian Process Regression (GPR) models for the considered electricity consumption data.
[AI-243] SlimWise: Decoupling Expert Pruning Across Prefill and Decode for Efficient MoE Serving
链接: https://arxiv.org/abs/2609.34117
作者: Gunho Park,Kyoungho Jeun,Juntaek Oh,Byeongjun Shin,Baeseong Park,Minsoo Rhu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Mixture-of-experts (MoE) models activate few experts per token, yet batched decoding can access nearly the entire expert pool, making expert-weight traffic a major bottleneck. Expert pruning reduces this traffic, but conventional approaches also prune compute-bound prefill, sacrificing model quality for little throughput benefit. We present SlimWise, a serving framework that tailors the expert pool to each inference phase. SlimWise performs prefill with the full model and decode with a pruned model that directly reuses the prefill-generated KV cache without conversion. Across two MoE backbones and three pruning criteria, this training-free KV cache handoff substantially narrows accuracy gaps relative to the full model in many settings. We also show that benchmark accuracy can conceal substantial pruning-induced changes in generation length. To address these distortions and residual accuracy loss, SlimWise introduces a low-cost distillation stage that trains the decoder to continue from full-model KV caches while updating only a small subset of parameters. Implemented in vLLM, SlimWise supports both prefill-decode (PD) disaggregation and PD-colocated serving. On Qwen3.6-35B-A3B, SlimWise improves decode throughput by up to 1.81x at 50% expert pruning with minimal accuracy loss.
[AI-244] GUITAR: Structured Failure Diagnosis of GUI Agents via State Transitions NEURIPS2026
链接: https://arxiv.org/abs/2609.34113
作者: Shaoqing Zhang,Kehai Chen,Xuefeng Bai,Zhuosheng Zhang,Pengfei Zhang,Yang Xiang,Min Zhang
类目: Artificial Intelligence (cs.AI)
备注: NeurIPS 2026 Dataset and Evaluation
Abstract:Understanding where and why Graphical User Interface (GUI) agents fail is essential for building more reliable systems, yet current evaluation relies on step accuracy, a metric that treats each screen independently and overlooks the underlying structure of GUI environments. This leads to two critical blind spots: (1) functionally equivalent screens are evaluated in isolation, obscuring systematic failure patterns across shared screens; and (2) the long-tailed GUI distribution renders failures on rare but critical screens invisible under standard metrics. To address these issues, we propose \textbfGUITAR, a state-centric diagnostic framework that performs structured failure analysis over both states and transitions, using a State Transition Graph (STG) by mapping visually diverse screens to shared functional states. Across 8 agents and 6 tasks from AndroidControl and Mind2Web, GUITAR reveals that 60.4% of failures occur in 20% of states, localizing errors to a small set of bottlenecks. Bottleneck-targeted guidance improves SR by 2.8% and retains a 1.88% average gain across 7 agents under three-fold trajectory-held-out evaluation with fully automatic STGs. These findings demonstrate the diagnostic and actionable value of structure-aware evaluation within the evaluated mobile and web tasks. Code is available at this https URL
[AI-245] A Differentiable Optimization Framework for Registering Sequential Bounding Boxes with Point Cloud Stream
链接: https://arxiv.org/abs/2609.34103
作者: Xuesong Li,Jinguang Tong,Jie Hong
类目: Artificial Intelligence (cs.AI)
备注: accepted by AJCAI 26
Abstract:Refining a sequence of coarse 3D bounding boxes against a LiDAR point-cloud stream demands tracks that are geometrically accurate (high IoU) and temporally coherent (low roughness), preferably without training data. The usual recipe keeps the two concerns apart: register each frame independently, then smooth the trajectory afterwards with a Kalman~RTS or Savitzky–Golay filter. Smoothing displaces boxes from a geometric optimum and never re-optimises, so it trades accuracy for smoothness. We instead fold the temporal smoothness constraint into a training-free registration objective and solve for all poses jointly with L-BFGS. The payoff depends on how well the object is seen. On well-observed tracks it is large: within the low-roughness budget, the joint objective beats both post-hoc smoothers on paired multi-seed statistics and cuts roughness several-fold relative to frame-wise registration at matched accuracy. Treating visibility as an experimental variable exposes the limit. The advantage decays monotonically as views become one-sided, until it is indistinguishable from zero for near-edge-on objects and slightly negative under a ray-cast simulator with range-dependent density and ego motion, where the decoupled pipeline is in fact ahead at tight roughness budgets. We locate that boundary and trace it to one term: orientation alignment ties yaw to the estimated velocity and fails once that estimate is noisy. A ground-truth-free rule can choose the temporal scale and keep every track inside the roughness budget.
[AI-246] RACE: Expert-Aligned ECG Representation Learning with Rigorous Benchmarking and Real-World Validation in Acute Cardiac Care
链接: https://arxiv.org/abs/2609.34088
作者: Lovely Yeswanth Panchumarthi,Andrew Lu,Saurabh Kataria,Delgersuren Bold,Minxiao Wang,Runze Yan,Patricia Dykes,Brian J. Gow,Tom J. Pollard,Jessica K. Zègre-Hemsey,Dillon J. Dzikowicz,Lekshmi Kumar,Xiao Hu,Ran Xiao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 42 pages, 4 figures and 1 table
Abstract:TRACE (Text-Reinforced Analysis of Cardio ECGs) is a multimodal electrocardiogram (ECG) representation model that learns clinically grounded signal embeddings for downstream cardiac classification. It is designed to address the limitations of existing CLIP-style training, which often struggles with noisy clinical text and fails to leverage the complementary strengths of unimodal (from ECG) and cross-modal (between ECG and matched cardiologist reports) learning. To bridge this gap, we propose a hybrid architecture that jointly learns unimodal and cross-modal representations via uncertainty-weighted multi-task learning while utilizing an LLM-based pipeline to extract high-fidelity findings from cardiologist reports. We evaluate TRACE across a spectrum of clinical urgency, establishing robust performance on public benchmarks for arrhythmia classification and structural abnormalities relative to existing unimodal and multimodal ECG models. To demonstrate real-world utility, we further validate the model on acute coronary occlusion (ACO), where the prevailing ST-elevation criteria miss 25-34% of true occlusions. Utilizing a large private ACO dataset with expert-annotated ground truth, TRACE significantly outperforms real-world clinical practice, yielding a 19.0% increase in sensitivity or a 62.6% reduction in false positive rates at the clinical baseline. This extensive evaluation confirms that TRACE delivers both strong performance on benchmark tasks and tangible clinical impact in the most acute, high-risk cardiac scenarios.
[AI-247] GenoMorph: Pathway-Grounded Genomic Disease Reasoning via Adaptive Latent Computation
链接: https://arxiv.org/abs/2609.34079
作者: Tanmoy Kanti Halder,Akash Ghosh,Arijit Roy,Sriparna Saha
类目: Artificial Intelligence (cs.AI); Genomics (q-bio.GN)
备注:
Abstract:Large language models (LLMs) have demonstrated strong capabilities in biological reasoning; however, genomic disease inference remains largely dependent on memorized gene-disease associations rather than understanding biological pathways. This shortcut learning undermines robustness and generalization, and breaks down when molecular identifiers are unavailable. We present GenoMorph, a multimodal genomic reasoning framework that shifts disease prediction from associative gene-disease mapping toward pathway-grounded reasoning. GenoMorph couples a frozen DNA foundation model with question-conditioned cross-attention fusion, self-adaptive latent reasoning (LatentSp), a residual reasoning gate for iterative genomic evidence reinjection, and rejection sampling fine-tuning regularized by hierarchical optimal transport (OT). Rather than learning direct gene-disease mappings, GenoMorph aligns genomic sequence representations with latent pathway dynamics, enabling reasoning trajectories that follow molecular interactions before producing disease predictions. LatentSp dynamically allocates computation according to reasoning confidence, reducing unnecessary reasoning steps and improving inference efficiency. We further construct an anonymized benchmark from the Kyoto Encyclopedia of Genes and Genomes (KEGG), replacing every gene and molecular identifier with anonymous symbols while preserving sequences and pathway topology, thereby removing memorization shortcuts. GenoMorph raises the weighted F1 from 0.7863 (BioReason) to 0.9412, and rejection sampling fine-tuning with self-adaptive latent reasoning pushes it to 0.9725 while cutting latency nearly 60%. On the anonymized benchmark it reaches 0.9465 F1, substantially outperforming prior systems and confirming that accurate disease prediction can arise from pathway reasoning rather than memorized gene-disease associations.
[AI-248] MaskCoFT: Masked Co-Adaptive Fine-Tuning for Memory-Efficient MoE Inference
链接: https://arxiv.org/abs/2609.34077
作者: Junfeng Wu,Zehao Fan,Hadjer Benmeziane,Kaoutar El Maghraoui,Liu Liu,Yinan Wang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Mixture-of-experts (MoE) language models often exceed the memory of a single GPU. Expert offloading keeps most experts in host memory and loads them on demand, so decoding speed depends on how many experts each token must fetch. Caching and prefetching reduce this cost only as far as the routing allows. Router-only fine-tuning can reshape the routing to reuse experts, but it keeps the experts frozen, so they cannot adapt to the tokens the new routing sends them. We propose MaskCoFT, a masked co-adaptive fine-tuning method that trains routers and experts together with the cross-entropy loss alone. During fine-tuning, a learnable binary mask restricts the Top-K routing of each layer to a subset of experts, and the experts adapt to the tokens redirected to them. At inference, the learned mask becomes a soft prior that re-ranks experts, so every expert remains selectable. We simulate a GPU cache of 4 experts per layer for Mixtral-8x7B and 12 for DeepSeek-V2-Lite. MaskCoFT cuts expert fetches per token by 23.7% and 10.1% relative to the base model. In real offloading system serving, it lowers the time per output token by up to 16.4% and 5.5%, respectively. Its average accuracy over nine benchmarks stays above the base model by 0.92 and 0.53 points.
[AI-249] PhysFieldBench: Can Multimodal Models Understand Physical Fields?
链接: https://arxiv.org/abs/2609.34072
作者: Yuezhou Ma,Huikun Weng,Jialong Wu,Chenyi Zhao,Hang Zhou,Haonan Shangguan,Jianmin Wang,Mingsheng Long
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multimodal large language models (MLLMs) are increasingly envisioned as core components of scientific and engineering agents, yet their ability to interpret physical fields remains poorly understood. Existing physics benchmarks largely emphasize textbook problem solving or intuitive physical reasoning, leaving open whether MLLMs can infer physically meaningful information from continuous field observations. We introduce PhysFieldBench, a benchmark comprising 24 tasks and 1,160 evaluation examples across controlled equation fields, simulated physical fields, and observed physical fields. The tasks assess three forms of inference: identifying physical mechanisms, comparing latent control variables, and predicting outcome properties. Across representative open-source and proprietary MLLMs, zero-shot performance is low: the best model achieves a chance-normalized score of 29.3, while several open-source models remain near chance. In contrast, a task-specific supervised vision transformer performs substantially better, demonstrating that the inputs contain learnable physical information. To diagnose these failures, a structured self-explanation analysis attributes most errors to missed visual patterns and incorrect visual-to-physical mappings. Further, to explore whether post-training can improve physical inference and generalize to unseen tasks, we compare supervised fine-tuning with final answers or chain-of-thought supervision and reinforcement learning. Final-answer supervision performs best overall but transfers less effectively, whereas reinforcement learning after chain-of-thought supervision achieves the best generalization. Together, these findings highlight the need to improve visual-to-physical grounding and cross-task generalization for MLLMs to reliably interpret physical fields in scientific and engineering workflows.
[AI-250] FLARE: Flow Matching with Local Axis-Angle Representations for Stochastic Micromagnetic Evolution
链接: https://arxiv.org/abs/2609.34070
作者: Pengyu Li,Renjie Tong,Xuanlue Jiang,Jianmin Li,Yuanyuan Zhou
类目: Machine Learning (cs.LG); Mesoscale and Nanoscale Physics (cond-mat.mes-hall); Artificial Intelligence (cs.AI)
备注:
Abstract:Long-horizon micromagnetic simulation remains expensive because conventional and learned solvers typically propagate Landau–Lifshitz–Gilbert (LLG) dynamics step by step. Existing learned approaches generally retain stepwise integration or model deterministic evolution, leaving full-field, direct-horizon stochastic prediction largely unexplored. We propose FLARE, a flow-matching framework that recasts stochastic finite-time magnetization prediction as conditional transport over anchor-relative local axis-angle rotations. This rotation-space formulation respects the intrinsic geometry of magnetization dynamics and preserves pointwise unit norm by construction. By explicitly conditioning on the physical prediction horizon, FLARE directly generates full-field stochastic endpoints across multiple target times without stepwise integration. Against the strongest single-checkpoint external baseline on each metric, FLARE achieves 29.9% lower angular energy distance ( 15.30^\circ ), and a 37.3% lower fair energy score (0.393). On a representative composed 5-ns two-segment protocol, FLARE achieves a 3,062\times best-batch speedup over the widely used GPU micromagnetic solver MuMax ^3 on a single GPU.
[AI-251] owards Certificate-Driven Software Porting: A Self-Improving Agent ic Harness for Scientific Program Optimization
链接: https://arxiv.org/abs/2609.34069
作者: Piyush Jha,Aishik Ghosh,Vijay Ganesh
类目: Artificial Intelligence (cs.AI); Programming Languages (cs.PL); Software Engineering (cs.SE)
备注: Submitted to ML4PS 2026
Abstract:The upgrade and rewriting of large scientific codebases has traditionally been a major challenge. While evolutionary search with large language models (LLMs) can port and accelerate legacy code, repair feedback in prompts alone does not prevent subsequent candidates from repeating the same errors. We introduce Certificate-Driven Evolutionary Search (CDES), which extends evolutionary search with enforceable restrictions derived from failed candidates, recorded as certificates of assumptions, checker evidence, and justified restrictions. Its control logic enforces these restrictions through rejection, backtracking, and targeted repair while preserving compatible edits. We apply CDES to CPU-to-GPU translation of two particle-simulation functions from the Geant4 toolkit, evaluated with a harness that goes beyond unit tests to combine formal checks, numerical comparisons, physics checks, and GPU safety tests. Generated implementations achieve 13.78x and 23.54x function-level speedups over CPU code, including data conversion and transfers; for one function, GPU throughput exceeds an expert implementation by 14.9%, reaching 16.1% when complementary components are combined. In an ablation over execution settings, certificate feedback increases the fraction of candidates passing required correctness checks from 55% to 90%.
[AI-252] Learning Perturbation Robust Policies for LLM Agents with Stable Optimization
链接: https://arxiv.org/abs/2609.34064
作者: Pengxin Wang,Yuanzhe LI,Yuxin Ren,Huanrui Yang,Jingdi Chen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning (RL) has become an effective post-training paradigm for long-horizon large language model (LLM) agents. However, we find that the resulting policies can be sensitive to various policy perturbations, such as hidden-state noise, pruning, and quantization. In this work, we study how to improve perturbation robustness during policy optimization. We first introduce the notion of a perturbation robust policy and analyze conditions under which perturbed policy updates preserve stable monotonic improvement. Based on this analysis, we introduce Stable Perturbation-Robust Policy Optimization (SPrPO), which applies adaptive and sensitivity-aware perturbations during RL training. We evaluate SPrPO on ALFWorld and WebShop and conduct systematic experiments across multiple perturbation types and scales, showing improved perturbation robustness while maintaining stable policy optimization.
[AI-253] Do World Models Learn Global Understanding?
链接: https://arxiv.org/abs/2609.34058
作者: Alexander Detkov,Matt Thomson
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 16 pages, 8 figures
Abstract:AI systems often feel brittle and fragmented. A large language model (LLM) may correctly explain a concept but fail to apply it, or follow safety instructions in one context but not another. This behavior suggests a general failure to lift local information to a global understanding. To gain fundamental insight, we frame “understanding” as learning constraints and propagating their consequences. We construct learning tasks on monoid worlds, sets of states connected by action transitions, where observed training transitions and an unseen constraint jointly determine held-out transitions. Measuring generalization tests whether models can learn global constraints from local transitions and propagate their consequences. We consider inverse, commutativity, composition, and periodicity constraints relevant to spatial and semantic structure. Across attention, recurrent, and state-space architectures, next-state training fits the data but fails to propagate non-trivial constraints. Compositional training, which uses identical paths but hides intermediate states from the input, achieves 96% accuracy on inverse, commutativity, and composition constraints across architectures, yields corresponding improvements in geometric generalization of world models trained on embodied environments and relational generalization in Wikidata-finetuned LLMs. How far do models propagate constraints when inferring an unseen fact may depend on first inferring others? We define proof depth d of a held-out transition, measuring the minimum number of inference rounds to infer the transition, and find that model generalization decreases sharply with proof depth. Increasing compositional path length T improves generalization. These results provide a formal way to investigate global understanding in language and world models and demonstrate that compositional training promotes information propagation and integration.
[AI-254] hinking Outside the Box: Retention and Transmission of Information in Sliding-Window KV Inference
链接: https://arxiv.org/abs/2609.34049
作者: Timothy DeLise,Seth Cromelin
类目: Artificial Intelligence (cs.AI)
备注: 13 pages, 4 figures
Abstract:Sliding-window KV inference refers to processing a sequence incrementally while retaining only a fixed-size cache of recent key and value states. It can be applied to pretrained causal transformers at inference time without additional training, while its KV-cache memory remains fixed as more tokens are processed. Because cached states are computed in the context of earlier tokens, they may carry information from beyond the current window and transmit it to later states. This study presents a series of experiments using five open-weight models spanning Qwen, Llama, Mistral, and Muse Glimmer. We investigate whether information originating outside the immediate context window can persist through a rolling KV cache and remain useful for retrieval. Initial results show that retaining previously computed states improves retrieval across the models tested compared with recomputing the final fixed window from raw tokens. We then measure how far this effect extends and find that Muse Glimmer and Mistral 7B show the strongest \emphlatent information relay: they can recover information even after the relevant source tokens have left the cache. Both models incorporate sliding-window attention in their published architectures, an association that motivates testing whether training with sliding windows promotes more reliable information retention.
[AI-255] Large Language Models for Structured Clinical Data Analysis: Dual-Agent Grounding and Validation
链接: https://arxiv.org/abs/2609.34039
作者: Erfan D. Dehkalani,Seetha Shankaran,Abbot R. Laptook,C. Michael Cotten,P. Ellen Grant,Yangming Ou
类目: Artificial Intelligence (cs.AI)
备注: 22 pages, 2 figures
Abstract:Objective: To develop and characterize CLEAR-Med, a dual-agent framework for natural-language analysis of structured clinical data that separates SQL-based invocation from independent validation. Methods: CLEAR-Med uses one agent to translate a question into executable Structured Query Language (SQL), retain the executed query and database result, and produce a draft. Deterministic checks and a separately invoked cross-provider Validation Agent then accept the draft, request one bounded repair, or abstain. We formalized the system as a bounded selective pipeline and evaluated CLEAR-Med’s configuration and scalability, and the Invocation Agent’s accuracy and consistency on a 25-query development benchmark, using a harmonized 21-site neonatal hypoxic-ischemic encephalopathy table containing 532 de-identified infant records and approximately 1,300 variables. Results: CLEAR-Med completed all six nominal scalability configurations, including 500x1300. Across 25 development-benchmark queries repeated five times, the Invocation Agent answered 83 of 125 responses correctly (66.4%; query-cluster bootstrap 95% CI, 48.0-83.2%), compared with 15 of 125 (12.0%; 95% CI, 3.2-22.4%) for the ungrounded ChatGPT baseline, a paired improvement of 54.4 percentage points (95% CI, 36.8-72.0%). Conclusion: CLEAR-Med provides a general architecture for traceable analysis of structured clinical data: numerical claims remain linked to executed SQL, and unresolved cases can fail closed. The reported experiments characterize CLEAR-Med’s configuration and scalability and the Invocation Agent’s accuracy, while the formal analysis establishes the encoded-property guarantee of the complete control flow; a prospective full-pipeline evaluation of the validation and abstention stages is the next stage of this work.
[AI-256] UOPD: Uncertainty-Aware Intervention for On-Policy Distillation of Multi-Turn Agents
链接: https://arxiv.org/abs/2609.34036
作者: Wenbo Zhang,Pengcheng Xu,Weizhi Du,Jing Zhang,Hengrui Cai
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:On-policy distillation (OPD) trains a student on its own rollouts using dense supervision from a teacher. In multi-turn environments, a mistake at a critical decision step can redirect the subsequent rollout toward poor outcomes. We use low teacher confidence on student actions to select high-uncertainty steps for correction. In a controlled ALFWorld study, a single teacher correction at a low-confidence step improves subsequent student behavior and task success, motivating selective intervention during distillation. We propose UOPD, an uncertainty-aware intervention method for on-policy distillation. At low-uncertainty turns, UOPD executes student actions and applies the standard OPD loss. At high-uncertainty turns, it samples and executes teacher actions and trains the student to imitate them through supervised fine-tuning, which minimizes forward Kullback-Leibler divergence in expectation. UOPD utilizes adaptive uncertainty thresholds to target a scheduled intervention rate. Empirically, we evaluate UOPD across a broad range of agentic tasks, including ALFWorld, WebShop, and Search, demonstrating its superior performance over OPD methods and their variants. UOPD improves WebShop score by up to 15.8% relative to standard OPD.
[AI-257] ADPTNet: Adaptive with Prescriptive Timescales Non-Linear SSM for Sequence Modelling
链接: https://arxiv.org/abs/2609.34034
作者: Matei-Ioan Stan,Oliver Rhodes
类目: Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
备注: 80 pages
Abstract:A central aim of neuromorphic computing is to provide a viable alternative to highly energy-intensive Transformer-based AI. However, efficient alternatives struggle to capture the set of qualities that have secured the Transformer’s status as the de facto standard in sequence modelling. Any realistic contender must be data-adaptive, able to capture long-range dependencies, and GPU-parallelisable, but also non-linearly recurrent to enable complex reasoning. Based on evidence suggesting the auditory cortex operates on fixed timescales, this work proposes the ADaptive with Prescriptive Timescales Network (ADPTNet) as a potential solution to achieving all four properties simultaneously. ADPTNet is built around local topological conjugates, obtained by a novel combination of linear attention and Riemannian optimisation, applied to static global dynamics. This enables non-linear yet predictable long-term behaviour. Dynamical systems theory proofs provide theoretical guarantees for the parametric control of ADPTNet’s timescales (its Lyapunov spectrum). ADPTNet improves performance on Selective Copying over Hawk, the existing method balancing long-range memory and adaptability, while also improving state tracking over linear SSMs like Mamba. On sequential CIFAR-10, ADPTNet matches linear SSM accuracy and outperforms existing selective models (incl. the Transformer), using fewer parameters. We also introduce a neuromorphic SpikingADPTNet, which achieves a new state-of-the-art accuracy on the Spiking Speech Commands dataset ( 83.56%\pm0.15 ). Finally, ADPTNet’s constant timescales enable two efficient, Jacobian-free extensions to the DEER parallel simulation algorithm (Conv and Forward DEER) that retain the same average convergence. Conv DEER adds no computational overhead beyond the network’s forward pass and enables non-linear RNN parallelisation via iterated convolutions for the first time.
[AI-258] Uncovering shortcut learning in audio classifiers by discovering recurring concepts in temporal explanations
链接: https://arxiv.org/abs/2609.34030
作者: Cecilia Bolaños,Luciana Ferrer,Magdalena Fuentes
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注:
Abstract:Correlations between events in machine learning datasets may result in shortcut learning, where models learn to predict the target event based on the presence of a correlated event. When these correlations are spurious – arising from data collection artifacts – models are likely to perform poorly in practice. We propose a pipeline to uncover shortcut learning in audio classifiers by discovering recurring concepts in their temporal explanations. Specifically, we isolate audio segments that explain classifier decisions, caption them with an ensemble of Large Audio-Language Models, and use a Large Language Model to extract recurring concepts. The resulting concepts can be audited by humans to uncover potential shortcut learning. We evaluate our framework using datasets curated from AudioSet Strong, controlling for the presence or absence of spurious correlations. Results show that this approach reliably uncovers learned shortcuts, such as the model relying on the presence of “laughter” to predict “applause”.
[AI-259] Jev in Medicine: A Benchmark Evaluation. Preliminary Results
链接: https://arxiv.org/abs/2609.34024
作者: Alfredo Madrid-García,Beatriz Merino-Barbancho
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Jev is a non-generative “System One” model that assigns probabilities to predefined answer options and cannot answer outside them. Its accuracy and calibration on medical question-answering and case-based diagnostic-reasoning tasks are unknown. We evaluated Jev 1.13 on four medical benchmarks: MetaMedQA, PubMedQA, DiagnosisArena-MCQ and the NEJM Case Challenges. GPT-6 Sol, with (medium) and without reasoning, was the reference. The primary outcome was top-1 accuracy; key secondary outcomes were calibration, selective prediction and recognition of unanswerable questions. All 8,469 requests returned a valid answer. Jev’s accuracy was similar to that of GPT-6 Sol with medium reasoning on PubMedQA (78.4% vs 78.2%;), lower on MetaMedQA (74.8% vs 82.7%) and much lower on DiagnosisArena-MCQ (59.8% vs 82.4%;) and the NEJM cases (61.8% vs 82.4%). On MetaMedQA, Jev’s probabilities were the best calibrated (expected calibration error 0.063 vs 0.146), and its answers with a probability of at least 0.9 (52.9% of questions) were 93.4% accurate, but GPT-6 Sol was as accurate when it accepted a similar proportion of questions. On DiagnosisArena-MCQ, Jev’s probabilities discriminated poorly (AUROC 0.645 vs 0.768). Of the 162 questions whose correct answer was “I don’t know or cannot answer”, Jev chose that option for 10.5% (GPT-6 Sol, 8.6%). Median latency was 0.27-0.31 s; all 2,823 items cost USD 0.08. Jev was fast and inexpensive, and its accuracy was similar to that of a frontier LLM on research abstracts but lower on examination questions and much lower on complex diagnostic cases. Task-specific validation is required before clinical use.
[AI-260] SR4-Fit: A Unified Interpretable Rule-Based Machine Learning Framework for Informative and Trustworthy Decision-Making
链接: https://arxiv.org/abs/2609.34019
作者: Shyam Sundar Murali Krishnan,Dean Frederick Hougen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 77 pages, 63 figures, 23 tables
Abstract:In many high-stakes applications, machine learning is dominated by black-box models that require post hoc explanations to justify their predictions. These explanations are often unreliable because they do not reflect the model’s actual computations, limiting accountability and trust. A natural alternative is to use models that are interpretable by design. However, existing rule-based approaches, such as RuleFit and decision trees, while transparent, often lack stability and predictive strength, reinforcing a perceived trade-off between traditional performance measures and model understandability. To address this, we propose Sparse Relaxed Regularized Regression Rule-Fit (SR4-Fit), an intrinsically interpretable algorithm for both classification and regression that produces compact and stable rule sets without sacrificing performance. Using demographic data from the U.S. Census Bureau’s American Community Survey, SR4-Fit predicts U.S. House election outcomes with high accuracy and interpretability while uncovering demographic interactions missed by black-box models. We further validate SR4-Fit across fourteen benchmark datasets (six classification and eight regression), where it outperforms existing rule-based methods, including RuleFit and decision trees in terms of accuracy, stability, and compactness while remaining competitive with black-box models in predictivity. These results demonstrate that interpretability and predictive reliability need not be mutually exclusive, offering a practical and transparent alternative for high-stakes decision-making.
[AI-261] When Known Physics Helps Neural PDE Models: Residual Constraints Out-Regularize Generic Priors for Nonlinear Dynamics
链接: https://arxiv.org/abs/2609.34012
作者: Zahra Farazpay,Aniruddha Bora
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Neural PDE surrogates increasingly incorporate structural priors, yet it is often unclear whether their gains arise from physics-specific information or simply from regularization and training choices. We evaluate several such priors under a common protocol against a matched from-scratch neural operator baseline. Our central result is that a known-equation residual consistently outperforms the best generic regularizer at equal tuning budget. At fixed capacity this benefit appears across linear and nonlinear PDEs, but a capacity sweep reveals a sharp distinction: the advantage persists and grows for Burgers, KdV, and Allen-Cahn, while collapsing toward or below parity for linear heat and advection-diffusion. Thus, the durable value of the residual is specific to nonlinear operators. We further falsify a pre-registered hypothesis that the benefit is activated only by data sparsity: the residual remains advantageous even under full supervision. Its usefulness does, however, have a clear boundary. Under grid under-resolution, nonlinear coarse fields no longer satisfy the naive governing-equation residual, and enforcing it becomes actively harmful. In contrast, cross-family pretraining and in-context conditioning fail to outperform the strong from-scratch baseline in the regime studied. Together, these results identify when known physics provides non-redundant information to neural PDE models, when it does not, and when enforcing it introduces bias.
[AI-262] EHRAdapt: Adapting Pretrained Language Models to Electronic Health Records with Semantic Priors for Rare Clinical Events
链接: https://arxiv.org/abs/2609.34007
作者: Andre R Goncalves,Vincent Liu,Priyadip Ray
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Electronic health records (EHRs) encode clinical histories as (time, modality, code) tuples, whereas pretrained language models expect text tokens. Serializing them as text inflates sequence length and redundantly encodes structure. We introduce EHRAdapt, an adapter that maps tuples directly into a frozen language model’s embedding space. Modality receives a learned embedding, time gaps enter through learned attention biases, and event codes receive dedicated vectors. Learning event vectors is the central challenge: clinical vocabularies are long-tailed, leaving rare events too few observations for reliable estimates. EHRAdapt therefore represents each event vector as the sum of a semantic prior and an evidence residual. The prior is a frozen embedding of the event’s clinical description from a biomedical language model trained on clinical ontologies, mapped into the model’s input space by a shared learned projection, so it supplies clinical meaning even when observations are scarce. The residual, a learned low-rank event-specific correction, refines it as evidence accumulates. We run continued pretraining on about 4 million patients’ records with three frozen LLM backbones (OLMo2 1B, Llama3.2 1B, and OLMo2 7B), training only the adapter (0.1–0.6% of all parameters). The full adapter outperforms all ablations in held-out next-event prediction on every backbone. Removing the semantic pathway hurts rare events over ten times more than the most frequent ones, whereas removing the residual hurts overall prediction but improves it for the rarest events. On reportable infectious-disease and syndromic downstream classification tasks, EHRAdapt outperforms text-based LLM and count-based baselines, and both pathways improve rare-disease discrimination. The two pathways therefore play complementary roles, visible only when results are broken down by event frequency rather than averaged.
[AI-263] GroupMask: Layer-Adaptive Group-wise Sparsity for Semi-Structured LLM Pruning
链接: https://arxiv.org/abs/2609.33977
作者: Zhengao Li,Shuoqiu Li,Xiaofang Zhang,Yukai Jin,Gokcen Kestor,Yanfu Zhang,Yiming Zeng,Bin Ren,Chuxu Zhang,Shangqian Gao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Semi-structured pruning compresses large language models (LLMs) while keeping a regular sparse structure, but the prevailing N:M pattern fixes the same local sparsity ratio in every layer. Layer-adaptive sparsity allocation improves unstructured pruning, yet it has been reported to be less effective under N:M sparsity, leaving open whether adaptive allocation is of limited value for semi-structured pruning in general or only under the fine-grained N:M pattern. We examine this question with group-level sparsity, which partitions each weight matrix into regular groups, retains or prunes each group as a whole, and allows each layer’s sparsity ratio to vary under a global budget. We propose GroupMask, which generates the group selectors of all layers with a lightweight hypernetwork, relaxes them with a Gumbel-Sigmoid parameterization and a straight-through estimator, and learns them through sparsity-budget regularization and self-distillation while keeping the pretrained weights frozen. On LLaMA-2-7B at 50% sparsity with the same 1\times256 group size, learned layer-adaptive allocation reduces WikiText-2 perplexity from 10.02 to 8.30 and raises the average zero-shot accuracy from 0.455 to 0.496 relative to a uniform per-layer ratio. GroupMask obtains the lowest WikiText-2 perplexity on LLaMA-2-7B and the highest average zero-shot accuracy with Alpaca calibration among the evaluated baselines on five LLaMA and Qwen models. Our code is available at this https URL.
[AI-264] Designing Reliable LLM -as-a-Judge Measurement Systems for Multi-Turn Business Agents
链接: https://arxiv.org/abs/2609.33955
作者: Kaiwen Luo,Ming Gao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Many LLM-as-a-judge evaluations score fixed outputs under a fixed task definition. Production multi-turn business agents instead require a maintained measurement system: correctness depends on business-specific facts and procedures, outcomes emerge across turns, and failures must be attributed to either agent capability or missing business knowledge before they are actionable. We present an integrated methodology spanning evaluation specification, modular LLM judges, intent-preserving user simulation, and human-in-the-loop governance. The specification defines conversation-level end states and actionable failure ownership. Atomic judges share versioned evidence and feed an explicit aggregation graph. The simulator is released only after task-preservation and stability checks. Independent human audits estimate measurement fidelity, renew tiered reference sets, and route disagreements to label correction, guideline revision, or judge improvement. Production studies show that system-level fidelity improved across repeated audits, that human reviewers and automated judges improved together under the shared feedback loop, and that their combined workflow had the strongest descriptive performance in both reported task-completion settings. Because the studies are observational and the human reference itself required revision, these findings demonstrate operational usefulness rather than causal or universal superiority. The contribution is a practical framework for making multi-turn agent measurement reliable, actionable, and maintainable as the evaluated system and its evidence evolve.
[AI-265] Diffusion-Based Rollouts as a Stabilization Mechanism for Long-Horizon Environmental Forecasting
链接: https://arxiv.org/abs/2609.33930
作者: Marina Vicens-Miquel,Amy McGovern,Aaron J. Hill,Efi Foufoula-Georgiou,Samuel S. P. Shen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Extending forecast lead times while maintaining predictive skill remains a major challenge in environmental forecasting. We investigate diffusion-based rollouts as a stabilization mechanism for recursive forecasting using low-dimensional water-level time series and high-dimensional precipitation fields. Across both modalities, diffusion suppresses recursive error growth, with the largest stabilization occurring where deterministic rollouts are most unstable. However, stabilization does not guarantee forecast fidelity. In the water-level experiments, forecasts progressively lose event-level fidelity as the rollout loses access to external predictive information, and trajectory-level comparisons show that diffusion can remain numerically stable while contracting toward central values and exhibiting reduced variability. In the precipitation experiments, which retain conditioning from numerical weather prediction throughout the rollout, diffusion better preserves spatial organization and event-detection skill. Together, these contrasting experiments indicate that diffusion can control recursive error amplification, while its practical benefit also depends on the predictive information available to constrain future evolution.
[AI-266] HyperMCTS: Hypergraph-Augmented MCTS for Long-Horizon LLM Agents
链接: https://arxiv.org/abs/2609.33920
作者: Tingsong Xiao,Nithish Balachandar Moudhgalya,Chandrayee Basu,Lichao Wang,Luyang Kong,Benjamin Z. Yao,Zhe Jiang,Jie Hao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Long-horizon tasks require large language model (LLM) agents to coordinate decisions under constraints that span an entire solution. Monte Carlo Tree Search (MCTS) offers a promising approach to test-time scaling by exploring alternative action trajectories, but model computation and environment interaction make search costly. Efficient search therefore requires effective reuse of trajectory feedback. Standard MCTS maintains prefix-specific statistics, without explicitly accumulating outcomes for decision groups that recur across different paths. To fill this gap, we propose HyperMCTS, a training-free method that augments an ordered MCTS tree with a cross-trajectory hypergraph. Hyperedges represent groups of canonical decisions and accumulate their observed returns within the current task. Our hypergraph-guided HyperUCT selection rule aggregates evidence from overlapping hyperedges into an action prior, allowing outcomes collected under one prefix to inform selection under another while preserving execution histories in the tree. On DeepPlanning, HyperMCTS improves average planning accuracy by 2.3–7.3 percentage points over the strongest baseline for each of three backbone models. It enables Qwen3.6-27B to outperform Claude Opus 4.6 (max) on Shopping Planning, while achieving higher accuracy with fewer LLM calls and output tokens than the evaluated MCTS-based baselines. SealQA experiments further demonstrate improvements in question answering.
[AI-267] Validating Memory-Optimal Transformer Kernels on Real Hardware: From Formal Derivation to Measured Performance Across Two HPC Clusters
链接: https://arxiv.org/abs/2609.33916
作者: Lenore M. Mullin,Gaetan Hains
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI)
备注: Companion validation for Papers I-IV. Paper I arXiv:2606.07713 , Paper II HAL:05659212, Paper III arXiv:2607.19456 , Paper IV HAL:05734881 ( this https URL ). 26 pages, 20 figures (placeholders in v1, real figures in v2). Implementation: this https URL . Allocation CIS261396 on Purdue Anvil and NCSA Delta via ACCESS
Abstract:We validate memory-optimal cost functions for transformer kernels derived via the Mathematics of Arrays (MoA). Companion Papers I-IV formally derive kernels for attention forward, backward, fused forward+backward, decode, and the complete block (RMSNorm, gated MLP) as a hardware-independent specification (DNF) transformed to a machine-specific realization (ONF) via gamma, with verification to machine precision against PyTorch. This paper checks those predictions against measured performance on two HPC clusters (Purdue Anvil, NCSA Delta) across CPU and GPU. Three results stand out. (1) We identify and fix a GPU regression: fusing forward+backward, proven to avoid materializing an O(n^2) intermediate, initially ran slower than naive on GPU due to atomic contention. Profiling confirmed 2.00x more atomic instructions; a targeted ONF rewrite reversed it, yielding up to 2.5x speedup. (2) Identical derivations produce markedly different real costs by topology: 535x NUMA-locality penalty on one cluster vs 3x oversubscription on another, showing optimal deployment is a function of the machine’s array structure. (3) We report a partially resolved anomaly: identical denotational computations run faster in C than Fortran on CPU but faster in Fortran than C on GPU, narrowed to one dominant kernel and one memory-latency stall mechanism (3.17x time gap matches 3.35x stall gap). We treat hardware-specific optimization as a routine ONF rewrite with fixed, verified DNF, a candidate methodology for scaling AI onto evolving hardware without re-deriving correctness. Comments: Companion validation for Papers I-IV. Paper I arXiv:2606.07713, Paper II HAL:05659212, Paper III arXiv:2607.19456, Paper IV HAL:05734881 (this https URL). 26 pages, 20 figures (placeholders in v1, real figures in v2). Implementation: this https URL. Allocation CIS261396 on Purdue Anvil and NCSA Delta via ACCESS Subjects: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.33916 [cs.DC] (or arXiv:2609.33916v1 [cs.DC] for this version) https://doi.org/10.48550/arXiv.2609.33916 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-268] When Consent Outlives Context: Residual Authority Replay in Long-Lived Agents
链接: https://arxiv.org/abs/2609.33910
作者: Zhihao Zhang,Chao Wang,Rujia Li,Qingze Wang,Xiaoyan Sun,Jun Dai
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLM agents increasingly rely on user approval to authorize security-sensitive actions at runtime. Such approvals are granted within a specific task and execution context. In long-lived agents, authorization decisions may need to persist across tasks or sessions. We find that this continuity can outlive the context that originally justified the approval, creating residual authority reusable without renewed consent. We expose this failure mode through a longitudinal attack that starts from a target security-sensitive action, identifies the authority required to execute it, induces benign interactions that legitimately obtain that authority, and later replays the residual authority during adversarial execution. Across controlled and live settings, we demonstrate that residual-authority replay arises in practice and substantially increases the success of prompt-injection and context-rebinding attacks. We evaluate 508 AgentDojo attack cases across six LLM families using production-derived authorization semantics. With residual authority, attack success rate (ASR) increases by up to 35.1 percentage points compared with a fresh authorization state. In live context-rebinding attacks on 55 Terminal-Bench cases across three real-world production coding agents, residual-authority replay increases ASR by 24.9 percentage points on average. These findings expose a fundamental mismatch between persistent authorization and the contextual nature of user consent in long-lived LLM agents.
[AI-269] Finite Probes Suffice: Identifiability and Universality for Weight-Space Learning
链接: https://arxiv.org/abs/2609.33901
作者: Soutrik Sarangi,Yonatan Sverdlov,Adir Dayan,Haggai Maron,Nadav Dym
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Learning properties of neural networks has recently attracted growing interest, with existing approaches operating either directly on network parameters or through probe-based representations of network behavior. While probing methods have shown strong empirical performance, their theoretical foundations remain limited. In this work, we study when finite probe-based representations are sufficient for learning neural functionals. We establish general identification and universality results for probing, and show that using intermediate hidden representations can provide significantly more informative representations than relying only on final outputs. Motivated by these results, we introduce HIDDENPROBE, a simple architecture for learning from hidden probe responses. Across a range of neural functional benchmarks, including both MLPs and Transformers, HIDDENPROBE consistently improves over existing probing methods and achieves state-of-the-art performance. Our code is publicly available on GitHub.
[AI-270] Curating Merchant-Matching Training Data with Two Confidence-Gated Local LLM Judges ICDM
链接: https://arxiv.org/abs/2609.33878
作者: Donghao Huang,Jinling Pei,Zhaoxia Wang
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 7 pages, 2 figures, 5 tables, Accepted for publication in 2026 IEEE International Conference on Data Mining Workshops (ICDMW), SENTIRE 2026 Workshop
Abstract:Merchant matching resolves a noisy payment descriptor to a retrieved merchant entity or returns no match. A key challenge in curating training labels is distinguishing teacher abstention from evidence that no acceptable entity exists: false no-match labels contaminate pseudo-labeled data, while conservative labeling reduces coverage. We investigate whether agreement between two local large language model judges improves pseudo-label reliability. A label is retained only when the judges agree, with separate ordered thresholds for selections and abstentions that guarantee disjoint positive and negative label sets. Retrospective replay on 2,000 expert-annotated queries shows that higher selection thresholds can improve positive-label purity, whereas higher abstention thresholds increase false no-match labels. At thresholds (0.86, 0.80), Muse Glimmer 30B and Gemma 4 31B jointly label 1,633 queries (81.7% coverage) at 96.88% purity; positive and negative purities are 99.47% and 93.38%. This exceeds either constituent model at the same thresholds by more than two percentage points, with lower coverage. A split-half check finds only 0.14 percentage points of threshold-selection optimism. A symmetric threshold of 0.86 adds 40 erroneous no-match labels, while 46 false abstentions persist even with no confidence threshold. Across five matched within-model comparisons, higher reasoning effort yields no clear F0.5 gain and increases median latency by 1.8-5.0 times. These results motivate separate thresholding and auditing for positive and negative pseudo-labels. The study establishes label purity, not student utility; fresh-data curation and student fine-tuning remain necessary to demonstrate downstream value.
[AI-271] Counterfactual Rollout Replay: Forkable Environments as Free Process Rewards for Software Engineering Agents NEURIPS2026
链接: https://arxiv.org/abs/2609.33875
作者: Yuanhao Li,Hongbo Wang,Xuhong Chen,Yiming Cao,Xunzhu Tang
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026. Includes additional experiments and analysis
Abstract:Outcome-only reinforcement learning gives software engineering (SWE) agents a terminal success signal but little direct guidance about intermediate decisions. We introduce Counterfactual Rollout Replay (CRR), a training-time procedure that uses forkable executable environments to obtain step-level return contrasts. CRR selects a small set of decision points, restores each state, samples an alternative action, and rolls the branch forward under the policy. It retains the realised training trajectory and replaces the advantage at selected steps with the difference between its terminal return and the sampled counterfactual return. The method needs no human process labels or learned process reward model; free refers to those supervision costs, not replay compute. With a 14B policy, CRR improves pass@1 on SWE-bench Verified, SWE-bench Live, and SWE-rebench, and combines with process-reward and trajectory-search methods. On SWE-bench Verified, an equal-wall-clock comparison on the same hardware yields 41.7% versus 36.7% for extended outcome-only GRPO, a 5.0-point gain with fork overhead included. These results apply to environments with affordable, reliable state restoration; stochastic continuations and expensive or imperfect replay remain limitations.
[AI-272] Robot-GST: geometry-aware spatial-temporal robot policy representation and evaluation
链接: https://arxiv.org/abs/2609.33872
作者: Sichao Liu,Zekun Wang,Lixuan Tang,Yiming Li,Xiaohan Wang,Hanzhi Zhang,Daqiang Guo,Peng Zhou,Lihui Wang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 9 pages, 8 figures, 2 tables
Abstract:Robotic manipulation policies are advancing rapidly with increasing reliance on vision-language models for end-to-end decision making. However, reliable deployment remains challenging because many policies lack explicit mechanisms for predicting task outcomes and evaluating whether generated actions will achieve desired final states, causing execution errors to accumulate during long-horizon manipulation. We present Robot-GST, a geometry-aware spatio-temporal behaviour representation and evaluation framework that constructs a Gaussian-SAM robotic environment for real-to-sim policy verification and improves the reliability of real-world manipulation deployment. Our approach constructs a high-fidelity robotic environment from RGB-D observations using 3D Gaussian Splatting and SAM3D, enabling ``simulation and evaluation before acting’'. It integrates visual observations and language instructions with spatio-temporal reasoning for long-horizon task planning using large vision-language models. To bridge high-level planning and real-world execution, we introduce Gaussian-aware final-state estimation through geometric sampling and state-based trajectory planning. Before execution, candidate action sequences are simulated and evaluated in the Gaussian-SAM environment to filter infeasible behaviours. We validate our approach on representative manipulation tasks involving rigid, soft, and deformable objects, including cube placing, toy packing, and duck rearrangement, demonstrating that geometry-aware spatio-temporal reasoning and state-aware execution improve manipulation reliability across different object categories. Our results suggest that combining geometry-aware reconstruction with high-quality rendering and simulation provides a scalable approach for evaluating robotic manipulation behaviours. Website: this https URL
[AI-273] When Successful Strategies Fail: Adaptation to Environmental Novelty in Terminal Agents
链接: https://arxiv.org/abs/2609.33870
作者: Janvijay Singh,Vaishnavi Shrivastava,Dilek Hakkani-Tur,Ece Kamar,Asli Celikyilmaz
类目: Artificial Intelligence (cs.AI)
备注: 53 pages, including references and appendices; 10-page main text with 5 figures
Abstract:LLM agents increasingly solve long-horizon tasks by autonomously interacting with their environment. In doing so, their strategies rely on assumptions about that environment: which resources and tools exist, where they are located, and how they behave. When these assumptions no longer hold, reliable agents must detect the change and adapt while pursuing the same goal. We study this adaptation capability through environmental novelty: a change that keeps the task objective fixed while invalidating an assumption underlying an otherwise successful trajectory. We introduce AGNI, an automated pipeline that extracts trajectory-relevant assumptions, injects targeted environmental changes, and validates that the resulting novel tasks remain solvable. Across three terminal benchmarks, AGNI produces diverse novelties spanning resources, interfaces, constraints, and execution semantics. Evaluating multiple LLM agents reveals a substantial adaptation gap between base and novel tasks. Trajectory analysis suggests that agents often encounter evidence of the change but fail to diagnose its cause and revise their strategy. Finally, post-training for environmental novelty improves adaptation to held-out novel tasks while also improving performance on base tasks. Our results highlight a gap between task competence and adaptive capability and motivate environmental variation as a core dimension of agent training and evaluation.
[AI-274] R2 Flow: Recursive Self-Improvement via Recursive Skill Evolution
链接: https://arxiv.org/abs/2609.33867
作者: Mingda Zhang,Qiang Huang,Yanjin Li,Zijia Wang,Qika Lin,Xiaoying Tang,Tiesunlong Shen
类目: Artificial Intelligence (cs.AI)
备注: 27 pages
Abstract:LLM-based agents can improve themselves across tasks by reusing and revising the skills they orchestrate into executable procedures. Flow-based training fits this loop: it samples procedures in proportion to reward, and the flow through each skill credits it for the next library revision. Three obstacles stand in the way of making this self-improvement reliable: flow training suffers strategy collapse over tree-structured histories; nonnegative flow-based credit rewards frequent use as if it were benefit; and library edits rest on the task reward the policy optimizes. We introduce R ^2 Flow, a recursive self-improvement framework that alternates policy learning, independent verification, and versioned skill-library updates on a shared-state orchestration graph. The graph merges histories that differ only in the order of independent steps, allowing flow training to pool evidence across equivalent executions. A flow-share readout of the trained flow, invariant to the backward policy, and a separate signed utility rank which skills to change, verifier evidence decides whether an edit is warranted, and a residual-variance plateau sets when to update. Committed edits reshape the graph the next policy learns on, realizing recursive skill evolution. Across question answering, mathematical reasoning, interactive decision making, and code generation, R ^2 Flow improves task accuracy and library-edit precision over heuristic orchestration, reinforcement learning, and skill-evolution baselines, and transfers across executors. Code is available at this https URL.
[AI-275] PI-NOMT: Physics-Informed Neural Optimal Mass Transport for Brain Fluid Dynamics
链接: https://arxiv.org/abs/2609.33857
作者: Mehmet Emin Acar,Vahit Bugra Yesilkaynak,Helene Benveniste,Gozde Unal
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Recovering hidden transport mechanisms from sparse spatiotemporal observations is a fundamental inverse problem in scientific machine learning. In brain tracer imaging, dynamic contrast-enhanced MRI (DCE-MRI) provides time-resolved measurements of tracer concentration, while the underlying velocity and source mechanisms governing tracer propagation remain unobserved. We formulate this problem as physics-informed latent-state inference, in which the transport field itself is the primary object of inference rather than an auxiliary variable used only to reconstruct observed densities. We propose Physics-Informed Neural Optimal Mass Transport (PI-NOMT), a framework that represents density, velocity, and source as continuous neural fields and combines a continuous neural density teacher, recursive differentiable advection–diffusion–source rollout, unbalanced optimal-transport regularization, and governing-equation supervision. Physical laws act as structural priors that constrain the space of admissible transport mechanisms, while observed tracer dynamics provide evidence for estimating the latent transport state. We evaluate PI-NOMT on a synthetic benchmark with known ground-truth transport and on DCE-MRI sequences from nine control rats. On the synthetic benchmark, PI-NOMT accurately recovers the prescribed velocity field, including its magnitude, direction, and integrated trajectories, rather than merely reconstructing endpoint densities. Across the nine rat datasets, the framework yields sub-percent local endpoint error, consistent physical speed scales, and low post-training PDE and incompressibility residuals. These results support physics-informed latent-state inference as a general framework for recovering hidden transport mechanisms from observed dynamic scalar fields.
[AI-276] How code helps different tasks? A decompositional lens on LLM post-training
链接: https://arxiv.org/abs/2609.33845
作者: Zheng Yu,Yiwei Li,Yishen Chen,Xiang Li,Jiale Han,Benyou Wang,Jingbang Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Evaluating code data as a single corpus can obscure which types of code data benefit which models and downstream tasks. Effective data selection requires understanding both the benefits of individual categories and whether these benefits persist when categories are combined. We introduce a decompositional lens for studying these effects in LLM post-training. We first decompose an execution-verified code corpus into interpretable categories based on the computational patterns of its solutions. Through controlled fine-tuning experiments, we compare individual categories with a balanced mixture across instruction-tuned models on question answering, mathematics, and code generation. The resulting response maps reveal recurring gains in average question-answering performance, while the same category can improve one model or task and degrade another. The best-performing category also varies with the starting model and target task. We then compose compact mixtures guided by these results and examine whether benefits observed in individual categories persist under joint training. On selected model–task pairs, mixtures whose constituents each improve the target task outperform both their best constituent and full-corpus training while using roughly 10–15% of the full corpus. These exploratory findings illustrate a \emphless is more pattern and highlight how the value of code data in post training depends on which categories are combined for which model and task.
[AI-277] Achieve What You Imagined: Learning to Align Actions with Visual Plans
链接: https://arxiv.org/abs/2609.33832
作者: Yuheng Qiao,Ziran Wei,Xiaohan Wang,Daqiang Guo,Yichen Luo,Zhibo Pang,Peng Zhou,Sichao Liu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 9 pages, 8 figures, 2 tables
Abstract:World-action models can jointly predict future visual observations and robot actions. However, discrepancies may exist between their visual predictions and the consequences implied by generated actions. We observe that WAMs can often generate visually plausible task-completion outcomes before producing action sequences that reliably achieve them. Consequently, we treat the WAM-generated visual prediction as a goal-conditioned visual proposal rather than a directly executable plan. We use a frozen action-conditioned world model to predict action-conditioned consequences and construct feedback based on consistency between the two future predictions and alignment with the terminal goal. Leveraging this feedback, we employ Flow Policy Optimization (FPO) to optimize the action head of the WAM. This framework avoids online robot interaction and additional training of task-specific reward models. Across four real-world UR5 manipulation tasks, our method increases the mean success rate from 43.4% to 75.1%, compared with 61.4% for \pi_0.5 . These results show that cross-model prediction discrepancy can provide useful feedback for improving robot policies under the evaluated manipulation tasks. Website: this https URL
[AI-278] Vestrum: Improving Agent Harnesses by Adapting Their Verification Structure and Memory
链接: https://arxiv.org/abs/2609.33822
作者: Jayant Parashar,Eugene F. Douglass,William C. Bastian,Suchendra M. Bhandarkar
类目: Artificial Intelligence (cs.AI)
备注: 30 pages, 2 figures, 20 tables. Under review
Abstract:An agent harness controls how a language model accesses information, uses tools, preserves memory, and checks its work. Improving this software is costly when each evaluation requires a long interaction with an environment. We introduce Vestrum, a framework that turns failures in execution traces into scoped harness changes without training the task model. Its organizing overhypothesis is that tasks of a shared kind may exhibit recurring failures whose remedies transfer within that kind. Vestrum expresses failures as recognizable classes, proposes changes across verification, retrieval, decomposition, and knowledge synthesis, and screens their scope before evaluating them as a bundle. A persistent lessons file informs subsequent proposals. Across five settings and two baseline harnesses, the frozen harnesses improve held-out performance: UltraHorizon rises from 47.6 to 59.8 over GAM, Terminal-Bench 4 Hard from 63.7% to 70.3% of checks passed over Claude Code on eight held-out tasks at 1.03x test cost, and cell-type annotation agreement from 67.5% to 77.8% on held-out sections of one slide, alongside gains on LoCoMo and AMA-Bench. Across our searches, verification grounded in evidence helped both intermediate steps and final answers, at lower cost at intermediate steps, while critics asked to rebuild finished answers broke more than they repaired. On the three memory benchmarks, Vestrum also scores above the evaluated GEPA configurations in every paired evaluation.
[AI-279] Dual-Vocabulary Language Model for Cross-Tokenizer Distillation
链接: https://arxiv.org/abs/2609.33816
作者: Kedi Chen,Chen Lin,Yutao Sun,Wei Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:On-policy distillation (OPD) bridges teacher supervision and student behavior, but different teacher-student tokenizers introduce misalignment in both input tokenization (#1) and output logits (#2). Existing approaches address the former by matching same-text spans or converting tokens to bytes, often losing fine-grained token information or disrupting the native-token paradigm, while for the latter, strategies such as ranking, padding, or key-token selection retain only shared logit dimensions, resulting in much distribution loss. In this paper, we propose Dual-Vocabulary Language Model (DVLM), which replaces the teacher’s LM head with a new student-vocabulary projection head and obtains full-dimensional student logits (for #2). To support student tokens (for #1), it takes a Parallel-Tokenized Sequence (PTS) as input, which concatenates the original teacher-tokenized sequence and a re-tokenized sequence formed by independently converting each student token into a teacher-token group. To avoid inference inconsistency with the original teacher tokens, the Hybrid-Prefix Attention (HPA) further restricts re-tokenized groups to their corresponding teacher prefix and uses its last state as the aggregation of the original student-token representation for projection into the student vocabulary space. Similarly, via the combined use of PTS and HPA, the DVLM teacher can provide distribution-aligned supervision with the student’s input-tokenization and output-logit during OPD. Experimental results demonstrate that our DVLM teacher has a similar converged loss as the original teacher model and enables student models to improve performance across six reasoning tasks.
[AI-280] CodeActionBench: Evaluating Agent ic Code-as-Policy for Embodied Manipulation
链接: https://arxiv.org/abs/2609.33807
作者: Yiheng Lyu,Xueying Jiang,Wenhao Li,Shijian Lu,Gongjie Zhang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 28 pages including references and appendices, 10 figures. Project website: this https URL
Abstract:How well can general-purpose multimodal models turn visual understanding and reasoning into embodied manipulation via executable code? We introduce CodeActionBench, a benchmark of 25 manipulation tasks that evaluates this capability through agentic Code-as-Policy. Without task-specific fine-tuning, demonstrations, external specialist perception or grasp modules, privileged scene state, or predefined task policies, agents should select visual evidence, form task-relevant 3D estimates, construct manipulation targets, and iteratively execute and revise their policies. A shared robot API provides RGB observations, calibrated geometric operations, robot feedback, and bounded motion, leaving task-dependent decisions to the evaluated agent. Fixed task instances, resource budgets, and a hidden physical-outcome verifier support controlled comparisons across models and harness configurations. Extensive evaluations across nine configurations and 675 attempts achieve success rates ranging from 2.7% to 73.3%. The strongest configuration, GPT-6 Astra with Codex CLI, solves 22 of 25 tasks at least once in three attempts, demonstrating the best performance while still leaving substantial room for improvement. Trajectory analyses reveal difficulties in spatial alignment, object retention, and completion judgment, including task failures despite successfully completed motions. CodeActionBench provides a controlled testbed for measuring how general-purpose models translate their capabilities into manipulation behavior and for examining typical failure scenarios in that process.
[AI-281] Diffusion Reward Models
链接: https://arxiv.org/abs/2609.33803
作者: Xiangyang Wang,Bingxiang He,Zeyuan Liu,Jiaze WangZiqing Qiao,Yuxin Zuo,Huan-ang Gao,Cheng Qian,Wenbin Zhang,Ran Li,Youbang Sun,Ning Ding,Yuanchun Shi,Zhiyuan Liu,Chaojun Xiao,Chun Yu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reward models underpin the alignment of large language models, yet the dominant designs reduce each prompt–response pair to a point estimate or to a distribution from a fixed parametric family. This is at odds with human preference, which is inherently multimodal: the same response can be reasonably judged in many ways, and no single family covers all of them. To better fit this structure, we introduce DRM, a Diffusion Reward Model that recasts reward modeling as conditional density estimation over p(\mathbfr\mid x,y) . Conditioned on a frozen LLM encoder, a lightweight Diffusion Transformer denoises Gaussian noise into a reward vector, placing no parametric assumption on the output distribution and naturally representing its multimodal structure. A single architecture handles both multi-attribute regression and pairwise preference data, and at inference N samples form an empirical reward distribution that can be aggregated into a scalar, a variance, or quantiles. Across five benchmarks, DRM matches or surpasses baselines under matched data and backbone, stays competitive with much larger discriminative, distributional, and generative RMs despite its modest training scale, and recovers multimodal reward structure where conventional heads collapse to a point. Uncertainty-aware rejection and lower-confidence-bound (LCB) aggregation further demonstrate that DRM can exploit distributional information beyond a scalar reward to improve reward-model decisions. Downstream RLHF experiments additionally show that using DRM as the training-time reward leads to improved policy performance, directly validating the practical benefit of diffusion-based reward modeling for RLHF training.
[AI-282] Do We Really Need KL Divergence for On-Policy Distillation of Large Language Models ?
链接: https://arxiv.org/abs/2609.33791
作者: Wenze Lin,Jiyuan Long,Jiale Zhao,Shenzhi Wang,Xitai Jiang,Ce Luo,Rui Lan,Qianli Ma,Fukang Wen,Hui Wu,Liyuan Chen,Shuoling Liu,Jiangpeng Yan,Gao Huang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Since the advent of knowledge distillation, KL divergence has been the standard loss in distillation. Recently, on-policy distillation (OPD) has emerged as an efficient post-training paradigm for LLMs. As a distillation method, OPD naturally inherits KL divergence as its standard loss. However, in this work, we find that KL divergence may not be necessary for OPD. We show that simply preserving the update direction is sufficient for effective OPD. As long as the update direction is toward the teacher, OPD works. More precisely, it is not the direction of every token, but the direction of a small subset of tokens where the teacher and student disagree strongly. We first show that simply assigning a reward of (+1) to tokens where the teacher probability is higher than the student probability and (-1) where it is lower, which merely encourages updates toward the teacher, reproduces almost the same training mode as OPD with reverse KL. We further show that only the direction of a small subset of tokens with large teacher-student disagreement is critical, and training works as long as their update direction is toward the teacher, even if other tokens are pulled away from the teacher. And as an application of these findings, we introduce Consensus Multi-Teacher On-Policy Distillation (C-MOPD) to improve Multi-Teacher On-Policy Distillation (MOPD). Unlike MOPD, which routes each sample to a single teacher and may cause capability conflicts across domains, C-MOPD lets every sample be supervised by all teachers. Experiments show that C-MOPD consistently outperforms MOPD on both math and code benchmarks. Our code is available at this https URL.
[AI-283] Is your uncertainty map wrong or is its target? Exact diagnostics for the Tweedie diagonal and a gradient-free alternative
链接: https://arxiv.org/abs/2609.33786
作者: Vicent Ribas,Anna Oliveras Tous
类目: Artificial Intelligence (cs.AI)
备注: 40 pages, 3 figures, 23 tables
Abstract:A diffusion model can predict a follow-up medical scan from a baseline, but a clinician needs a per-voxel map of where that prediction can be trusted. Many such maps approximate the diagonal of the Tweedie posterior covariance, and are evaluated against another approximation of it, so whether the estimator or the target limits them is unclear. We compute the exact diagonal on six checkpoints across fourteen model-corpus conditions. Hutchinson at M=200 tracks it at rank agreement of at least 0.92 everywhere, yet in four of the fourteen the exact diagonal is anti-correlated with the denoising error, reaching -0.13, so a faithful estimator reproduces that reversal. All four are real-image conditions; on the models’ own samples the reversal does not appear, so evaluating on generated samples flatters this family. What limits these maps is the target, not the estimator. We then introduce Tweedie Probe-Tangent (T-PT), a gradient-free residual probe that corrupts one model-supported prediction repeatedly and measures the voxel-wise variance of the denoiser’s response. T-PT reads a different functional of the same Jacobian, and its exact second-order form ranks with the diagonal wherever the diagonal reverses; at thirty probes it returns a map too unstable to reproduce that ranking, while Hutchinson at M=5 already reproduces it, so T-PT there is not evidence against the reversal. We offer it as an instrument, not a better approximation. On brain MRI at full resolution, where every Jacobian-based estimator we test runs out of memory, T-PT leads a twenty-chain Monte-Carlo ensemble on five of eight endpoints inside tissue and trails it on none, at 16x fewer network evaluations; over the whole volume the ensemble leads, and fifty chains close the tissue gap. On lung CT the ensemble is ahead throughout. Both lose most of their discrimination where the change is, which remains open. Comments: 40 pages, 3 figures, 23 tables Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.33786 [cs.AI] (or arXiv:2609.33786v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.33786 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-284] Selecting Diverse SFT Traces Improves Post-RL Generalization
链接: https://arxiv.org/abs/2609.33780
作者: Dylan Zhang,Mingyuan Wu,Jinning Li
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Verified solutions are not equally useful for preparing reasoning models for reinforcement learning (RL). We present a comprehensive study of route diversity, the variation in the sequences of reasoning steps in supervised fine-tuning (SFT) data, and propose a lightweight, rule-based fingerprint to select for it. From one pool at one budget, with matched training recipes and checkpoints, selecting diverse rather than similar routes improves post-RL problem coverage across puzzles and mathematics, including on problems harder than those seen in either training stage. In synthetic experiments, route-diverse SFT improves OLMo3-7B’s pass@8 by 16.9 points on environments held out from SFT. In a single-model condition, where one model writes every candidate, diverse selection gains up to 6.2 points of mean pass@8 across 10 mathematics benchmarks. Pre-RL diagnostics suggest why: diverse SFT can produce both successful and failed attempts on more prompts despite slightly lower mean accuracy, giving group-relative RL more prompts with a learning signal. On 3 open-source corpora, our CPU-only selector, without model calls, outperforms more expensive alternatives in every comparison of mean post-RL performance. These results identify reasoning-route diversity as a practical criterion for selecting SFT data that better prepares models for RL.
[AI-285] Evidence-Inference Reconstruction: When The Evidence Is Recalled But The Reasoning Goes Wrong
链接: https://arxiv.org/abs/2609.33778
作者: Megan Diehl,Ser-Nam Lim
类目: Artificial Intelligence (cs.AI)
备注: 24 pages, 6 figures
Abstract:Modern multi-hop LLM agents are equipped with built-in mechanisms to detect errors in intermediate reasoning steps. Such errors trigger corrective actions from these agents, which mostly follow the paradigm of retrying the steps or the reasoning trajectories. Not only are these retries expensive, we present in this paper that they are also potentially unnecessary. To this end, we introduce Evidence-Inference Reconstruction (EIR), which uses structured state to guide one retrieval trajectory, accumulating source evidence in the process. We show that as long as the relevant evidence has been collected, EIR is capable of generating the correct answer in a single final model call even if erroneous evidence has been mixed in due to incorrect intermediate reasoning steps. In one evaluation, using Haiku 4.5 and GPT-4.1 Mini, we evaluate EIR on matched 1,000-question subsets of HotpotQA, 2WikiMultiHopQA, and MuSiQue, showing that EIR improves Answer F1, the overlap between the model’s and the correct answer, over the baseline by 8.3–32.8 points, Agentic SSR by 10.6–29.1 points, and Reflexion by 1.1–15.9 points. Additionally, we show that EIR averages 4.85 total model calls per question, compared with 35.29 for Agentic SSR and 12.41 for Reflexion. Together, these results corroborate EIR’s central premise: separating evidence retrieval from the final answer model call can improve answer accuracy while utilizing substantially less computation.
[AI-286] Learning Strategies to Break Judges
链接: https://arxiv.org/abs/2609.33773
作者: Guruprerana Shabadi,Aaditya Naik,Rajeev Alur,Mayur Naik
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As AI agents surpass human performance, it becomes exceedingly hard for system designers to evaluate them directly and understand their failure modes. Consequently, agents themselves are being deployed extensively to evaluate, judge, and provide feedback on model traces. But this raises an important question: how can we trust the judge? In this work, we propose an agent-guided method to find weaknesses of agentic judges that expose interpretable failure mechanisms. Our method focuses on mathematical reasoning and proceeds in two stages: first, we deploy adversarial agents to mutate a set of sound proofs by introducing errors, attempting to misguide judges—in other words, injecting errors that judges are unable to catch. Then, we distill these attempts into a small set of mutation strategies which allow us to analyze the failure modes of the judges. To ensure that these strategies are not overfit to the initial set of proofs, we evaluate them by applying the mutation strategies to a held-out set of proofs and querying the same judge. We deploy our method on GPT-5.6-sol and Claude Opus 5, paired with their agent orchestrators, Codex and Claude Code, respectively. These are used both as mutators to introduce errors and as judges to evaluate correctness of mathematical reasoning. We find that across all the agentic judges, we are able to distill mutation strategies that consistently bypass their evaluations, thereby enabling us to ascertain actionable failure modes. Our analysis also reveals that judge reliability degrades at the frontier: errors in Olympiad-level proofs or graduate-level mathematical texts are detected more consistently, whereas flaws in research-level manuscripts are more likely to escape detection.
[AI-287] Skill2Env: Capability-Oriented Environment Synthesis from Skills for General Agents
链接: https://arxiv.org/abs/2609.33772
作者: Weiyi Xu,Xiaowen Yang,Wen Da,Hang Xu,Canwei Li,Hongjie You,Pusen Dong,Yucheng Zeng,Zhaokai Luo,Mu Chuan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Executable environments are critical for post-training agents on tasks that require tool use and multi-step interaction, but constructing executable tasks together with their environments remains difficult to scale. Skills provide reusable domain knowledge, operational procedures, and tool-use instructions, but a substantial gap remains between the information contained in a skill and a concrete, challenging task with a complete executable environment. To address this gap, we introduce Skill2Env, a capability-oriented framework that starts from a skill and uses agent capability demands to guide task and environment synthesis. Skill2Env represents these demands through reusable difficulty patterns and instantiates them into task blueprints that specify objectives, challenges, environment facts, information boundaries, and acceptance criteria. These blueprints guide the joint construction of task instructions, execution substrates, workspaces, and rubric-based evaluators around source skills. We further propose Iterative Task Hardening, which uses solver execution evidence to identify insufficiently challenging task designs, strengthen or extend their difficulty-pattern instantiations, and revise the corresponding blueprints and environments. Using 1.5K high-scoring trajectories generated from Skill2Env environments for supervised fine-tuning, we observe consistent improvements across a broad range of agent benchmarks, demonstrating the effectiveness of capability-oriented environment synthesis for agent post-training.
[AI-288] SecProbe: Adaptive Evaluation of Coding Agents on Cybersecurity Vulnerabilities
链接: https://arxiv.org/abs/2609.33763
作者: Xiaonan Luo,Yue Huang,Kehan Guo,Ping He,Chuan Zou,Chujie Gao,Lichi Li,Yuchen Ma,Zhangchen Xu,Zichen Chen,Yufei Han,Xiangliang Zhang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Assessing cybersecurity vulnerability awareness in coding agents requires evaluations that reveal capability gaps and remain informative as models evolve. Static benchmarks offer fixed coverage and difficulty, while scarce vulnerable repositories and costly expert authoring limit their renewal at scale. We introduce SecProbe, a framework for adaptive evaluation that combines Item Response Theory (IRT) with on-demand synthesis of repository-scale vulnerability-repair tasks. From observed performance, \textscSecProbe estimates agent ability and identifies where additional evidence is most informative, selecting existing tasks or synthesizing new ones accordingly. As one use case, we construct 353 tasks spanning six programming languages and 151 CWE types and evaluate nine frontier models with two agent harnesses. Success rates peak at 28.33%, highlighting substantial gaps in vulnerability recognition and repair. Compared with random and one-shot baselines, \textscSecProbe achieves comparable agent ability estimates while requiring agents to solve up to 29.5% fewer tasks. These results support adaptive evaluation as an efficient and discriminative approach to assessing cybersecurity vulnerability awareness.
[AI-289] ransformer-based Neural Beamforming for Real-Time Speech Enhancement on Smart Low-Power Hearable Devices
链接: https://arxiv.org/abs/2609.33755
作者: Luca Bompani,Marco Fariselli,Giovanni Oltrecolli,Francesco Conti
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注:
Abstract:Accurate, efficient, and low-latency spatial beamforming is a key component in emerging smart hearable devices, enhancing speech while suppressing noise and interference. However, handling multiple input sources under strict real-time constraints poses significant challenges for the low-power, resource-constrained microcontroller units (MCUs) used in hearables. We present an optimized methodology for the real-time execution of a neural-network-based minimum variance distortionless response (MVDR) beamformer on MCUs. Using six microphones and a three-stage mixed-precision scheme (float32 MVDR, int8 CNN, float16 Transformer), the pipeline pairs a CNN that estimates the MVDR weights with a lightweight Transformer that applies a per-frame correction. By time-slicing weight estimation with beamforming, it achieves a 15~ms per-frame latency while refreshing a complete set of CNN-derived weights every 564~ms. The deployed mixed-precision pipeline attains a short-time objective intelligibility (STOI) of 97.65%, a scale-invariant signal-to-noise ratio (SI-SNR) of 20.26~dB, and a wideband PESQ of 3.676 at an average power of 45.9~mW. A speech activity detection (SAD) module (98.5% accuracy, 0.62~mJ per inference) bypasses the pipeline during silence; under realistic deployment conditions, the system exceeds the 16~h all-day target on a 100~mAh battery, with an estimated lifetime of up to \sim 20~h. To our knowledge, this is the first real-time multi-channel Transformer-based neural beamforming pipeline deployed on an MCU-class device.
[AI-290] Collaborative Synthetic Data for Privacy-Preserving Financial Fraud Detection Across Organizational Silos
链接: https://arxiv.org/abs/2609.33754
作者: Simeon Allmendinger,Domenique Zipperling,Burhanettin Bahadir Kibar,Niklas K{ü}hl
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted for publication at the International Conference on Information Systems (ICIS) 2026
Abstract:Organizations seek analytical value from AI, yet relevant data are often fragmented across organizations and constrained by privacy. This is acute in financial fraud detection, where rare fraud cases and imbalanced local datasets limit decision-relevant analytics. Federated learning enables collaboration without direct data sharing but does not resolve minority-class scarcity. Synthetic data generation can help, yet lightweight methods are interpolation-bound, while generative models require substantial data and computation. Existing collaborative generative approaches often rely on federated learning, imposing considerable organization-side training burdens. In this paper, we examine CollaFuse as a collaborative diffusion-based alternative for fraud detection and evaluate it across five fraud datasets. Compared with classical oversampling, local generative baselines, and centralized diffusion benchmarks, CollaFuse does not achieve the highest local fidelity but improves downstream fraud detection more consistently across most datasets. These findings suggest that synthetic data create analytical value less through local realism than through transferable cross-organizational structure.
[AI-291] HTN Planning as a Coordination Layer for Multi-Server MCP Tool Orchestration ICAPS2026
链接: https://arxiv.org/abs/2609.33731
作者: Eliott Jacopin,Éric Jacopin,Koichi Takahashi
类目: Artificial Intelligence (cs.AI)
备注: 5 pages. Camera-ready version of a paper accepted at the ICAPS 2026 Workshop on Hierarchical Planning (HPlan 2026), Dublin. Non-archival. Code: this https URL
Abstract:The Model Context Protocol (MCP) isolates servers by design: only the host can orchestrate cross-server workflows. When the host is a large language model, the resulting orchestrations are non-deterministic, non-reproducible, and pay one inference round-trip per tool call. We present a coordination architecture in which a Hierarchical Task Network (HTN) planner generates a verifiable cross-server plan once, and a runtime middleware executes it deterministically across multiple MCP servers, binding cross-action data dependencies via a template mechanism (\verb| context.X|) substituted at execution time. The architecture mirrors MCP’s isolation constraint: each compound task decomposes into server-local primitive actions, and inter-server data flow is bound at execution time via JSON-path output extractors. We instantiate the architecture on five HTN domains spanning laboratory robotics, bioinformatics and multiscale modelling, and demonstrate end-to-end execution from a browser-based plan controller against eight live third-party MCP servers querying real biological databases.
[AI-292] Robust Biomolecular Complex Design Across Protein Conformational Landscapes
链接: https://arxiv.org/abs/2609.33726
作者: Qingyuan Zeng,Zongqi Xu,Anglin Liu,Ziqi Gong,Pengxiang Cai,Zixin Guan,Yunan Chen,Sen Gao,Min Zhou,Jintai Chen
类目: Artificial Intelligence (cs.AI); Quantitative Methods (q-bio.QM)
备注:
Abstract:Proteins populate conformational ensembles, yet structure-based biomolecular design typically optimizes candidates against a single target conformation. Consequently, a candidate that fits one state can lose favorable interactions or develop steric clashes when the target adopts another. We introduce FlexEvo, a model-agnostic evolutionary framework that adapts candidates once at inference time from a single target conformation to improve compatibility with alternative natural conformations unseen during adaptation, without retraining the source model or requiring a conformational ensemble. FlexEvo casts cross-state adaptation as geometry-constrained bi-objective optimization, balancing preservation of input-state interactions against robustness to plausible conformational perturbations. To limit the search space and reduce invalid structural edits, geometry-derived FlexBoxes define protected anchor regions, adaptable regions for local exploration, and forbidden regions for clash avoidance. A unified all-atom representation supports topology-preserving adaptation across diverse binder categories, while Pareto selection preserves nondominated candidates across the two objectives. We evaluate FlexEvo across multiple generation baselines and nine representative binder categories spanning diverse molecular sizes and structural topologies. FlexEvo reduces the category-balanced mean relative performance degradation from 47.8% to 4.4%, while adding only 1.4–3.1 minutes of adaptation per sample. These results establish single-state inference-time adaptation as a practical route toward robust biomolecular complex design across protein conformational landscapes.
[AI-293] Self-Designed Evaluators and Warm Memory for Long-Horizon Agents
链接: https://arxiv.org/abs/2609.33717
作者: Saeid Asgari,Emre Kiciman,Leonardo de Oliveira Nunes,Ranveer Chandra
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:A tool-using language-model agent deployed over a long stream of tasks receives no reward, so it cannot tell whether it succeeded, cannot safely retry, and cannot label the experience it needs to improve. We present SelfSuite, in which the agent’s own base model, given only the world’s public materials, designs a small evaluation suite of weighted judges and grounded per-task briefs, freezes it, and uses it to gate a keep-best retry and to label a typed, outcome-tracked memory. On matched five-repeat benchmarks over tau2-bench and AppWorld, SelfSuite scores above the plain agent without any labels, matches methods given ten expert labels on tau2-bench, and trails Agentic Context Engineering (ACE) on AppWorld, where code execution gives a direct success signal. In an ablation campaign run on the same tasks, it is above label-free ACE in every repeat, and the gated second attempt is the only component whose removal hurts in every repeat. We also simulate a subject-matter expert who grades ten onboarding tasks per world. Using those labels to calibrate SelfSuite’s evaluator gives a small, consistent gain, and using them to warm up ACE’s memory lifts ACE to tie calibrated SelfSuite. A single-run study on a second model family shows the same ordering.
[AI-294] BIRD: Distilling Decision Boundaries into Rationales for MLLM Adaptation
链接: https://arxiv.org/abs/2609.33713
作者: Anglin Liu,Yanlin Wu,Ruichao Chen,Yuting Zhang,Qingyuan Zeng,Pengxiang Cai,Ziqi Gong,Muchen Li,Jintai Chen
类目: Artificial Intelligence (cs.AI)
备注: 25 pages, 13 figures
Abstract:Adapting general-purpose multimodal large language models (MLLMs) to specialized domains requires learning domain-specific decision criteria, which often hinge on subtle visual distinctions between otherwise plausible answers. Rationale augmentation aims to expose such evidence through additional observations or inter-sample comparisons, yet a visually valid cue is not necessarily decision-relevant: it may describe how samples differ without changing the model’s relative preference between competing answers. We therefore introduce BIRD, a self-improving Boundary-Informed Rationale Distillation framework that uses model-specific confusions to locate unresolved local decision boundaries and distills the evidence that resolves these confusions into rationales. For each sample, BIRD retrieves candidate neighbors from the target MLLM’s own representation space and selects the most confusable one according to its answer preferences. It then generates answer-blind candidate evidence from their visual differences and functionally verifies which evidence most effectively strengthens the model’s preference for the correct answer while avoiding inappropriate transfer across the pair. The verified evidence is then distilled into a single-sample rationale for standard supervised fine-tuning. Experiments on medical and chart VQA show that BIRD outperforms competing rationale-augmentation methods across two target MLLMs, while further analyses demonstrate clearer separation of confusable answers and stronger gains from model-matched supervision.
[AI-295] Does Adversarial Training Improve Generalization in Multi-View VLAs? Revealing and Mitigating View Collapse
链接: https://arxiv.org/abs/2609.33707
作者: Futa Waseda,Shuhei Kurita,Isao Echizen
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:Vision-language-action (VLA) models adapt pretrained vision-language models (VLMs) for closed-loop robot control, transferring their perceptual and semantic capabilities to action prediction. Despite strong in-distribution performance, however, VLAs often degrade under deployment shifts. Adversarial training (AT) offers a model-adaptive approach to robustness without explicitly anticipating individual shifts, but its effect on natural distribution-shift generalization in multi-view VLAs remains unclear. We study this question using a multi-view VLA directly adapted from a pretrained VLM and evaluate generalization across seven LIBERO-Plus shift axes. Direct AT substantially improves Camera Viewpoint and Sensor Noise, the two shifts affecting only the third-person view, yet produces mixed or negative effects on other shifts. Controlled view interventions reveal a surprising failure mode that we term view collapse: Direct AT can shift cross-view reliance so strongly that the policy becomes dominated by the wrist view. This exposes a \textitrobustness shortcut: apparent robustness to a shifted view can arise from reduced use of that view rather than more robust perception of it. This motivates a distinction between robust perception, extracting reliable information under within-view shifts, and robust fusion, adapting reliance across views according to their reliability. To reduce fixed view reliance, we use a simple View Swap intervention and then re-evaluate AT. With View Swap, AT further improves Camera Viewpoint, Sensor Noise, and Robot Initial State, while its effects remain mixed on other shifts. Our results show that multi-view robustness requires separating improved perception from changes in cross-view reliance, and that AT provides selective rather than generic distribution-shift benefits.
[AI-296] SpecRead: A Benchmark for Measuring Whether Language Models Understand Hardware Specifications
链接: https://arxiv.org/abs/2609.33699
作者: Feilian Huang(Independent Researcher)
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 13 pages. Benchmark data and code at this https URL
Abstract:Existing benchmarks for large language models (LLMs) in hardware design evaluate downstream artifacts such as generated RTL, assertions, or testbenches. When a model fails such a benchmark, the failure is ambiguous: it may have misread the specification, or it may have understood the specification and failed to write the code. We present SpecRead, a benchmark that isolates specification comprehension from generation ability. SpecRead v2.1 contains 385 questions over 10 open-source OpenTitan IP blocks: exact retrieval, cross-section reasoning, contradiction detection in mutated specifications, and spec-RTL consistency checking, plus 82 controls (41 distractor, 41 consistent-RTL). Type-4 items are built from real RTL mutations; we retain only mutations that Icarus Verilog simulation shows to change observable behavior. A with-spec vs. without-spec ablation suggests the questions require the excerpt, not training recall alone (without-spec accuracy 3/20 on the t1/t2 subset), though memorization of the source text may still help spot mutations. As an initial characterization with a small model, Ministral-3B scores 33.2% overall (128/385; macro average 39.0%): 55.2% on retrieval, 51.7% on cross-section reasoning. On the two contradiction-focused types, the verdict-plus-location measure gives 48.0% (t3) and 63.3% (t4), with a 51.2% false-positive rate on distractors and 100% on consistent-RTL controls. Layered scoring shows the model locates contradictions well (78.9-81.6% location accuracy) but scores lower on their category (43.9-49.7%). A structured “rule-table” prompting intervention lowers accuracy on every question type except t2 (tied). SpecRead is automatically scorable by deterministic checks, with gray-zone cases counted wrong under the conservative main scoring. The benchmark is regenerable for type-3 items via mutation injection, and built exclusively from public sources.
[AI-297] One Latent Many Tokens: Jointly Learning Compressed Embeddings for Efficient Language Diffusion
链接: https://arxiv.org/abs/2609.33698
作者: Yulin Yuan,Ying Zhang,Xiangming Meng
类目: Artificial Intelligence (cs.AI)
备注: 25 pages, 7 figures, and 8 tables. Includes appendices
Abstract:Most continuous diffusion language models process one latent position per token at each sampling step, making generation expensive. Two-stage methods lower the cost by reducing the latent length, but they fix the compressed embedding space before training the diffusion model. Embeddings from the fixed space can be difficult to model with diffusion and decode reliably into tokens, which limits generation quality after compression. To address this problem, we introduce JPEG-DLM (Joint-embedding Prediction for Efficient Generation with Diffusion Language Model), which jointly trains a compressor, a flow matching model and a decoding module. With joint-embedding prediction, JPEG-DLM learns compressed embeddings that are more structured, easier to model with diffusion and reliably decodable into tokens. JPEG-DLM achieves the lowest mean Gen-PPL and highest throughput among recent diffusion and flow models on LM1B and OWT. At a compression rate of 0.5 on OWT, it reaches a Gen-PPL of 34.52 and approximately 2.3 times ELF’s throughput. These results suggest that jointly learning compressed embeddings offers a promising path toward efficient diffusion language modeling. Code will be released soon.
[AI-298] Dynamic Kuramoto-Hodge Operators for PDEs on Complex Geometries and Topologies
链接: https://arxiv.org/abs/2609.33693
作者: Xiang Li,Yue Song
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Code is available at this https URL
Abstract:Learning PDE operators on complex domains requires capturing interactions among fields on vertices, edges, and faces, alongside global responses shaped by topology. Existing neural operators accommodate irregular geometries but often overlook these distinct field supports or their condition-dependent coupling. We introduce the Dynamic Kuramoto–Hodge Operator (DKHO), which combines topology-constrained interactions with learned coordination. DKHO encodes conditions on their native cochain supports, evolves Kuramoto-inspired relation states through the boundary and coboundary operators that compose the Dirac operator, and decodes non-harmonic and harmonic responses in orthogonal Hodge subspaces. Topology thus determines where information can flow, while learned dynamics adapts how it is exchanged to each PDE instance. Across porous-medium Darcy flow, torus transport–diffusion, and cavity magnetostatics, DKHO-large reduces prediction error by approximately 61% on average over leading baselines, while DKHO-small remains competitive using only 11.5–24.3% as many parameters. These results suggest that coupling topological structure with adaptive dynamics provides an effective inductive bias for accurate and parameter-efficient PDE operator learning on complex geometries and topologies.
[AI-299] opoMamba: A Load-Support Relation-Guided Multi-Directional State-Space Model for Topology Optimization
链接: https://arxiv.org/abs/2609.33688
作者: Bin Lou,Yuxuan Cheng,Huaizhi Zong,Junhui Zhang,Bing Xu
类目: Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
备注:
Abstract:Deep learning has emerged as an efficient alternative for predicting high-performance material distributions in topology optimization. Existing methods struggle to accurately capture load-transfer information, limiting out-of-distribution generalization, while their model architectures often incur high computational costs. To address these challenges, this paper proposes TopoMamba, a topology prediction framework incorporating a load-support relation-guided multi-directional state-space model. Coupling physical fields with load-support relations enables more effective modeling of mechanical dependencies. A load-support relation-guided spatially adaptive fusion mechanism dynamically adjusts multi-directional scan features according to spatial conditions. Mamba is coupled with the solid isotropic material with penalty method to enhance structural mechanical performance while maintaining computational efficiency. Results on two-dimensional topology optimization benchmarks demonstrate that TopoMamba achieves superior topology prediction accuracy, out-of-distribution generalization, and computational efficiency over state-of-the-art models. The proposed load-support physics-guided framework enables efficient optimization of more complex structural systems.
[AI-300] SWE-Game: Can Coding Agents Build the Games We Want?
链接: https://arxiv.org/abs/2609.33678
作者: Xiaoyu Chen,Lai Wei,Jin Wang,Xiangyu Zou,Ruochen Fan,Enze Luo,Mingzhe Yao,Jiahui Zhu,Yuhua Wen,Linghe Kong,Weiran Huang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce SWE-Game, a benchmark of 247 tasks grounded in 41 executable reference Godot games spanning 13 gameplay categories in 2D and 3D. Five task types cover development from a brief, implementation from a game design document, skeleton completion, repair of 83 injected-fault cases, and Godot-to-Unity porting. Reference materials specify the intended gameplay, while a shared instrumentation interface lets evaluator-owned drivers and probes execute actions and observe independently implemented games. Evaluation combines engine-state checks, certified reference-input replay, and agent-authored feature demonstrations to assess mechanic correctness, demonstrated playability, and behavioral restoration and preservation after repairs. Game-specific vision-language rubrics separately assess presentation. Across six models, Opus5 achieves the highest overall score in all five task types. Best overall scores remain below 60 out of 100 across the three construction tasks, with Brief-to-Game reaching 50.38. Analysis of reviewed submissions identifies requirement omissions and gameplay logic errors as predominant implementation problems. On human-labeled behaviors from 100 agent-built games, executable checks achieve 92.59% balanced accuracy, compared with 78.41% for a video-based VLM judge. Rubric-based visual scores reach a Spearman correlation of 0.829 with human ratings of 200 gameplay clips. Together, these results characterize current agent capabilities across game-development activities and support combining runtime evidence with visual assessment.
[AI-301] RSD-Poker: Structure-Adaptive and Shift-Robust Risk-Utility Certification for Residual Policies in Imperfect-Information Games
链接: https://arxiv.org/abs/2609.33669
作者: Miaobo Hu,Shuhao Hu,Xiaobo Guo,Xin Wang,Bokun Wang,Peng Zhang,Daren Zha,Jun Xiao
类目: Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG)
备注: 36 pages, 5 figures
Abstract:Residual policy adaptation provides a lightweight way to modify a strong reference policy, but a shared scale and a fixed subgroup partition can hide heterogeneous degradation and become fragile when the deployment mixture of information states changes. We introduce RSD-Poker, a structure-adaptive and shift-robust certification framework that freezes a bank of residual families and scales, learns a policy-visible partition on an independent structure split, and freezes that partition before calibration labels are joined. Each candidate-group pair receives a weighted simultaneous upper certificate for anchor-relative risk and a lower certificate for weak-response utility. A robust group-to-candidate map is then selected over a predeclared uncertainty set of deployment group proportions. Under independent calibration units drawn from each frozen group’s law, a candidate bank and partition fixed before calibration, and invariant within-group conditionals, the selected map satisfies its declared mixture-robust risk budget and utility certificate with probability at least 1-\zeta_risk-\zeta_util . The information contract supports both a teacher-backed transform and a teacher-free observation-only student. The retained deterministic 24-state audit remains an exact replay diagnostic: empirical-zero selects \alpha=0.08 , raising the weak-response proxy from 4.2082 to 4.2889 with 0/12 held-out threshold crossings. On stratified held-out states, the learned-partition dual selector raises weak utility from 4.4074 under global dual certification to 4.4936 and lowers held-out violation from 0.0215 to 0.0078; its mixture-robust variant reaches violation 0.0059. Across five observation-only checkpoints, risk-calibrated residuals attain weak utility 4.3659\pm0.0177 and violation rate 0.0178\pm0.0057 .
[AI-302] CompoWorld: Compositional Environment Scaling for General Agents
链接: https://arxiv.org/abs/2609.33665
作者: Xiao-Wen Yang,Weiyi Xu,Wen Da,Hang Xu,Canwei Li,Hong-Jie You,Pusen Dong,Yucheng Zeng,Zhaokai Luo,Yu-Feng Li,Yao Hu,Mu Chuan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Automatically generated environments provide a scalable source of interaction data for training general agents. However, existing approaches mainly generate tasks within a single environment, while real-world workflows require agents to connect information and actions across multiple services. We introduce Compositional Environment Scaling (\textbfCompoWorld), which expands the task space by composing a finite library of reusable services. Coding agents turn tool specifications into verified services with typed states and shared interfaces, while a world model handles tools that cannot be reliably implemented. A random-walk procedure connects services through dependency graphs, enabling the generation and verification of tasks that require information to flow across services. Verified trajectories support supervised fine-tuning (SFT), while our Completion-Focused Rubric Reward guides reinforcement learning (RL) toward full task completion by emphasizing criteria with lower pass rates within each rollout group. We construct 448 services exposing 10,130 tools and use 3K SFT trajectories and 1K RL tasks to train Qwen3.6-35B-A3B. Experimental results show that CompoWorld improves on its backbone by 9.17 points on average across eight benchmarks. On AutomationBench, it surpasses frontier models such as Claude Opus 4.6 and leads all compared agent-specialized 35B-A3B models.
[AI-303] Agent Boundary: Counterfactual Evaluation of Safety in Tool-Using LLM Agents
链接: https://arxiv.org/abs/2609.33658
作者: Tianzhuo Yang,Zirui Mi,Yantao Huang,Guoxi Zhang,Jiawei Chen,Yaodong Yang,Jingwei Yi
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Safety alignment for large language models (LLMs) in conversational settings is largely framed around whether to answer or refuse a request. In agentic settings, however, the same models must decide whether to act as permission-critical evidence emerges during execution. This creates a distinct challenge: apparent risk, action permissibility, and task competence are easily confounded, making agentic over-refusal difficult to distinguish from ordinary task failure. To address this, we introduce AgentBound, the first four-way counterfactual generation-and-evaluation framework for tool-using agent safety. AgentBound transforms the same executable workflow by independently varying apparent risk and action permissibility, enabling controlled comparisons of risky-looking but authorized tasks and routine-looking but unauthorized tasks. These comparisons jointly diagnose over-refusal and unsafe compliance while controlling for task competence. We instantiate AgentBound as a human-validated 4,000-task evaluation suite with trajectory-based and post-state-based judgments. Across 17 model and harness configurations, high safety frequently coexists with poor authorized-task completion: GPT-5.5 blocks 99.5% of routine-looking unauthorized actions yet completes only 28.7% of risky-looking authorized tasks. We further train a lightweight runtime calibration module that improves authorized-task completion by 18.2% on average across 10 evaluated configurations, while improving unsafe-action blocking by 5.4% on average. These show that effective agentic alignment requires action decisions to track permission-relevant execution evidence, rather than refusal strength alone.
[AI-304] Demonstration-Free Success-Probability Reward Learning for Generalist Robot Policies
链接: https://arxiv.org/abs/2609.33653
作者: Duo Wu,Haifeng Wang,Rongwei Lu,Jinghe Wang,Tianyi Xiong,Zhimin Wang,Chao Yu,Shuai Ma,Zhi Wang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Under review. Project webpage: this https URL
Abstract:Reinforcement learning (RL) enables generalist robot policies to improve through trial-and-error interaction, yet its effectiveness is fundamentally constrained by sparse task rewards. Existing general-purpose reward models typically alleviate this issue by learning task progress from expert demonstrations, but introduce a distribution mismatch with the mixed-quality rollouts encountered during policy optimization, making their estimates unreliable on suboptimal and failed behaviors from which the policy must learn. In this work, we introduce a demonstration-free reward learning paradigm where dense reward feedback can be learned directly from sparse task outcomes and policy experience. We theoretically show that terminal task outcomes implicitly define dense success-probability feedback at intermediate timesteps, which can be recursively learned through bootstrapping. Based on this insight, we introduce eVTA _0 , which learns success probabilities from mixed-quality policy rollouts through temporal-difference-style bootstrapping, without expert demonstrations or intermediate annotations. We further introduce RL with Evolving Rewards (RLER), a closed-loop framework that adapts eVTA _0 using newly collected rollouts as the policy evolves. Experiments show that eVTA _0 provides more informative rewards than state-of-the-art reward models and achieves the best average policy performance across all LIBERO task suites under the same RL training budget, improving success rates by 5.4%-13.8% over the initial policy. In real-world manipulation, RLER further improves overall success rates by 20%-26%, with 35%-36% gains under out-of-distribution conditions. These results demonstrate the effectiveness of demonstration-free reward learning and adapting rewards as the policy evolves. Project webpage: this https URL.
[AI-305] From Distributions to Stochastic Processes: Neural Approximation of Measure-Valued Maps
链接: https://arxiv.org/abs/2609.33649
作者: Yichen Wang,Ziyi Wang,Wenlian Lu,Chenghuang Shen,Jianfeng Liu,Zhengdong Xiao,Longjiu Luo,Qianrong Wang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Learning mappings between probability distributions arises naturally when inputs and outputs are represented by populations of samples rather than individual observations. We develop an approximation-theoretic framework for distribution-to-distribution learning and extend it to mappings between stochastic processes. For continuous operators on W_2 -compact families of finite-dimensional probability laws, we establish uniform neural approximation in the 2-Wasserstein metric using finite law statistics, a simplex-valued neural map, and a shared atomic output support that guarantees valid probability measures. We further extend this principle to probability laws on separable Hilbert spaces through finite-rank orthogonal projections. These results establish the representational feasibility of learning transformations between probability laws rather than deterministic vectors or functions. To demonstrate practical relevance, we study two problems naturally defined at the distribution level: prediction of first-passage-time distributions for an Ornstein–Uhlenbeck process and nonlinear response-path laws of a Duffing oscillator. Because the theory is model-agnostic and broader than any single practical architecture, the experiments use task-adapted neural models rather than reproducing the theoretical construction exactly. In both problems, the proposed models outperform a fixed-feature MLP baseline and distribution-space kernel regression. These experiments complement the theory by demonstrating the practical learnability of distribution-to-distribution transformations in random systems. Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.33649 [cs.LG] (or arXiv:2609.33649v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.33649 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-306] Scalable and Data-Driven Decision Support in the Maintenance Repair and Overhaul Process
链接: https://arxiv.org/abs/2609.33641
作者: Houkun Zhu,Helena Ebel,Dominik Scheinert,Florian Schmidt,Jens Altenkirch,Odej Kao
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Published in: 2022 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM)
Abstract:Several businesses apply maintenance, repair, and overhaul (MRO) principles to the life-cycle of their existing products. In cases like casted gas turbine component Product Lifecycle Management (PLM), repairing components in frequent intervals can extend the lifetime expectation of the product, provide higher cost efficiency compared to newly produced components, and even improve the part design during the repair cycle. Another aspect of repair concerns sustainability, as products often contain rare materials. The emissions produced by the repair process are usually smaller than mining materials and casting new components. To optimize the repair process further, we propose the Smart Expert System (SES), which assists engineering experts with machine learning-based decision support throughout the repair process. We elaborate on its IT architecture and present machine learning models employed for representative MRO use cases. The SES is evaluated using actual industry data from a leading gas turbine company and demonstrably fulfills formulated requirements concerning the suitability of the overall decision support and the stability of the enclosing IT architecture. Comments: Published in: 2022 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM) Subjects: Artificial Intelligence (cs.AI); Machine Learning (cs.LG) Cite as: arXiv:2609.33641 [cs.AI] (or arXiv:2609.33641v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.33641 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Related DOI: https://doi.org/10.1109/IEEM55944.2022.9989791 Focus to learn more DOI(s) linking to related resources
[AI-307] rajectory Unlearning on LLM -based Agents
链接: https://arxiv.org/abs/2609.33639
作者: Yingdan Shi,Ren Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Existing large language model (LLM) unlearning has focused primarily on removing specific knowledge, such as harmful facts, private data, or copyrighted content. However, as LLMs are increasingly deployed as autonomous agents, a fundamental yet overlooked problem emerges: beyond suppressing what an agent knows, an agent should not reproduce undesired behaviors through its action trajectories. In this work, we introduce trajectory-level unlearning, a new problem formulation that targets the removal of specific action trajectories in long-horizon agentic tasks, rather than factual knowledge. We identify two fundamental challenges that distinguish trajectory unlearning from knowledge unlearning: (1) our unlearning target is what the agent \emphdoes, not what it \emphsays; and (2) trajectories are sequentially dependent action sequences that cannot be decomposed into isolated prompt-response pairs without losing inter-step structure. To address these challenges, we propose Group-injected Relative Policy Optimization (GiRPO), which injects forget trajectories into the policy rollout group with penalized rewards and isolates the normalization statistics, yielding a stable and bounded unlearning signal that does not corrupt gradient updates for normal task trajectories. We construct trajectory unlearning benchmarks from two application scenarios, household tasks (ALFWorld) and online shopping (WebShop), and design three complementary metrics for evaluating forgetting quality and model utility. Experiments on ALFWorld and WebShop demonstrate that GiRPO effectively unlearns target trajectories while preserving task success rates, outperforming existing knowledge-unlearning baselines on both forgetting quality and task utility.
[AI-308] Learning Dynamics of Continual Learning: A Unified View of Data Attribution Forgetting and Plasticity Loss
链接: https://arxiv.org/abs/2609.33620
作者: Yi Ren,Wenlong Deng,Guanzhe Hong,Clare Lyle,Yarin Gal
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 45 pages
Abstract:Modern language models are likely to be updated throughout their lifetime rather than trained once and frozen. Each update therefore participates in a recurring cycle: decide which experience to learn from, understand what that update changes, and remain capable of learning from what comes next. We show that these challenges are governed by the same evolving update–behavior interaction. We derive a token- and layer-wise decomposition of how learning from one token changes another prediction. By separating the softmax force, shared readout geometry, and residual connections, it exposes two interaction channels and yields a forward-computable approximation. Following this interaction through time reveals a unified picture of continual adaptation. Positive interaction identifies useful experience; negative interaction produces either concentrated collision or accumulated erosion; over longer horizons, updates reshape the shared geometry mediating future learning signals, reducing their transmission. These predictions lead to effective data selection, mechanism-specific controls for interference, and a readout-based diagnostic of future learnability whose degradation predicts the benefit of restoring the readout. Across models and training regimes, the same local interaction thus explains both what an update changes now and how learning today changes what can be learned tomorrow. This view connects data attribution, forgetting, and plasticity loss as distinct regimes of the same evolving learning dynamics.
[AI-309] EAT: Expert Account Tracker for Efficient MoE Inference
链接: https://arxiv.org/abs/2609.33614
作者: Yuexian Li,Yifei Yang,Zouying Cao,Hai Zhao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Mixture-of-Experts (MoE) models have emerged as a revolutionary method to scale Transformer models. However, traditional MoE architecture still suffers from inefficiency since a large number of experts are unnecessarily activated. Existing approaches for reducing the number of activated experts often overlook the historical performance of each expert. In this paper, we propose EAT, a novel method called Expert Account Tracker (EAT), which utilizes history-awareness metrics and adaptive thresholding to dynamically select the most important experts, thereby reducing the activated expert number while effectively maintaining the model performance. Experiments show that EAT outperforms the existing baseline Top-P method across multiple models and datasets, achieving over 25% an average reduction compared to the vanilla method in the number of activated experts and performing better token generation speed compared to the baseline. Furthermore, the performance of pruned models can be efficiently recovered via OPD using only 9K data. Additionally, through ablation studies, we find that excessively reducing the number of activated experts can significantly harm model performance, and the importance of experts varies across layers, with higher-level experts being generally more critical.
[AI-310] Supervision Recovery for Time Series Anomaly Detection via Context-Anchored Pairing
链接: https://arxiv.org/abs/2609.33610
作者: Yifei Gao,Tian Lan,Yimeng Lu,Xuming An,Meng Wang,Wenjun He,Yijie Li,Chen Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Time series anomaly detection (TSAD) remains challenging not only because anomaly labels are scarce, but also because temporal anomalies are highly context-dependent. Existing methods often rely on unsupervised objectives or surrogate abnormal patterns, providing limited supervision for context-dependent normal–anomalous distinctions. We propose Context-Anchored Pair Supervision (CAPS), a supervision-recovery framework for TSAD. CAPS views ideal anomaly supervision as a matched comparison between normal and anomalous outcomes under the same temporal context, and seeks to recover such supervision without target-domain anomaly labels. Using simulated normal–anomalous pairs, CAPS learns structure and anomaly-semantic representations through reconstruction, background consistency, and within-pair counterfactual recombination. The resulting anomaly representations form a continuous semantic space with coarse modes and induce a sampleable multimodal prior. CAPS conditionally realizes sampled semantics as residual-form effects on target reference trajectories. The resulting context-anchored normal–anomalous counterparts provide temporal supervision for discriminative detector learning. Experiments on nine datasets show that CAPS achieves the strongest aggregate performance across all four evaluation metrics among the compared methods, while complementary ablations and transfer analyses support the roles of context anchoring, semantic disentanglement, and conditional realization.
[AI-311] Learning Transferable Reaction Mechanisms from Visual Chemical Knowledge
链接: https://arxiv.org/abs/2609.33608
作者: Yujian Yuan,Jiaxin Xu,Xin Cai,Yufan Chen,Zhichao Tan,Ziqi Zhou,Hanyu Gao
类目: Computational Engineering, Finance, and Science (cs.CE); Artificial Intelligence (cs.AI)
备注:
Abstract:Reaction mechanisms describe the step-by-step transformations underlying chemical reactions and are central to reaction analysis and synthesis. Learning-based models have achieved strong performance on established mechanism-prediction benchmarks, but transferring them to unseen chemistry remains challenging. Such transfer is difficult because familiar mechanisms must be applied to unfamiliar molecular structures, and some target mechanisms may be poorly covered by the training data. To address these challenges, we introduce MechaVLM, a visual framework that combines transferable chemical representations with external mechanistic knowledge. It learns reusable visual features through multiscale chemical grounding and cross-rendering contrastive learning. For open-book prediction, MechaVLM retrieves a fixed set of precedents from 70,384 literature mechanism figures and re-reads relevant visual evidence as the molecular state evolves, directly using the figures without symbolic mechanism parsing. An atom-indexed language decoder then recursively generates executable electron edits to construct the complete mechanism. We further introduce MechBench, a challenging literature-derived benchmark with 2,184 mechanisms and 9,146 elementary steps. Across cross-dataset and literature-derived benchmarks, MechaVLM establishes strong zero-shot mechanism prediction. Its closed-book model alone improves Step/Pathway Top-1 by 12.50/13.93 percentage points on FlowER-to-ReactMech transfer, while external visual precedents unlock further gains on challenging OOD reactions. The learned representation also generalizes beyond mechanism prediction to atom mapping and reaction center prediction.
[AI-312] JustQuant: You Dont Need Smoothing SVD or Rotation for 4-Bit Activation Quantization
链接: https://arxiv.org/abs/2609.33601
作者: Kaicheng Yang,Kaisen Yang,Chunyu Liu,Xianglong Yan,Haotong Qin,Junyi Wu,Tianao Zhang,Xun Zhang,Shaoqiu Zhang,Youbang Sun,Yulun Zhang
类目: Artificial Intelligence (cs.AI)
备注: Project page: this https URL
Abstract:Recent generative models have become increasingly powerful, but their inference cost continues to grow. Model quantization offers a promising way to compress these models and accelerate inference. However, at 4 bits, activation quantization is substantially more challenging than weight quantization. Recent post-training quantization (PTQ) and quantization-aware training (QAT) methods have made progress in 4-bit activation quantization by introducing smoothing, SVD branches, rotations, mixed precision, or advanced formats such as NVFP4. These additional operators and data types impose demanding requirements on inference engines and hardware, limiting the broad adoption of low-precision models. Can quantization be achieved using only plain low-bit operators? To answer this question, we propose JustQuant, a simple yet effective framework that moves the complexity of low-bit quantization from deployment-time operators into the training process. We first revisit model quantization from the perspective of knowledge distillation and show that a key reason existing PTQ and QAT methods fail is that they typically exploit supervision at only a single level. We then introduce Theseus QAD, a quantization-aware distillation method that progressively applies multi-level supervision, analogous to the gradual replacement process in the Ship of Theseus. Extensive experiments on DiT and diffusion large language models show two distinct regimes. For smaller models, Theseus QAD can serve as a lightweight warm-up stage that substantially improves subsequent QAT with plain operators, while naive QAD may collapse in the same setting. For larger models, Theseus QAD provides a stronger distillation training path than ordinary QAD. Across both regimes, JustQuant improves low-bit quantization quality while avoiding the complex operators required by many existing PTQ methods.
[AI-313] Compressing Value Predictions for Learning-Augmented Metrical Task Systems
链接: https://arxiv.org/abs/2609.33580
作者: Sizhe Li,Yecheng Li,Kun He
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT)
备注: 33 pages
Abstract:Learning-augmented algorithms for metrical task systems (MTS) can exploit predictions of canonical dual values, but existing formulations typically require a prediction for every state. We study whether these predictions can be compressed to a small set of representative states while retaining their algorithmic value. We introduce landmark-compressed value predictions, in which the predictor reports predicted dual values only at m landmarks and the remaining values are reconstructed by a Lipschitz extension. Our algorithm achieves additive excess cost O(T,r(L) + \sum_t \delta_t) , where r(L) is the covering radius of the landmarks and \delta_t measures prediction error up to additive shifts; local and value-dependent bounds refine this guarantee. For sparse landmark sets on unit-spaced finite lines, we prove a matching \Omega(T r_m) lower bound for every randomized algorithm using fixed landmarks, even with advance access to their entire exact absolute-value table. The prediction interface also matters: on two states with one landmark, exact absolute values permit horizon-independent excess, whereas exact relative values force worst-case expected excess linear in T . We give PAC guarantees for learning compressed prediction tables, with efficient empirical-risk minimization for fixed landmarks. Our results connect metric coverage, prediction interfaces, and learning guarantees for compressed predictions in online MTS.
[AI-314] OpenFC: Learning Verification Policies towards Open-Search Fact Checking
链接: https://arxiv.org/abs/2609.33579
作者: Xinming Wang,Kaixiang Qiu,Yansong Lin,Chunji Lv,Yi Chen,Boran Wang,Hong-Ming Yang,Xu-Yao Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Open-search fact checking is not merely retrieval followed by classification, but a sequential decision problem in which every query, source visit, and stopping decision reshapes the evidence available for verification. Yet existing systems often distribute these decisions across predefined pipelines or separately prompted modules rather than learning them as a unified task-specific policy. We introduce \textbfOpenFC, a unified verification-policy training framework that post-trains Qwen3-8B as a compact next-action controller over reasoning, evidence acquisition, and stopping. OpenFC learns this policy in two stages. \textbfStepwise-Calibrated Cold Start (SCCS) uses a strong training-time supervisor to review post-initial reasoning, tool-use, and stopping proposals before execution, producing reliable trajectories for supervised fine-tuning without access to gold verdicts. \textbfVerification-Aware Reinforcement Learning (VA-RL) then improves the cold-start policy on unresolved claims through budget-aware tool rewards, label-aware advantage reweighting, and localized response masking. Across six fact-checking benchmarks, OpenFC achieves 70.39% average accuracy and 63.30% macro-F1, the highest overall averages among the evaluated methods. Stage-wise ablations further show that SCCS and VA-RL provide complementary gains, supporting the design of the two-stage training framework. These results position OpenFC as a strong and effective framework for open-search fact-checking. We will open-source our code and release the model checkpoints to support reproducibility.
[AI-315] Dr. Free: You Dont Need Difficulty Rewards for Self-Evolving Search Agents
链接: https://arxiv.org/abs/2609.33565
作者: Zhipeng Qian,Zihan Liang,Yufei Ma,Jie Ma,Ben Chen,Huangyu Dai,Lingtao Mao,Xinyu Sun,Tong zhao,Xuxin Zhang,Qingpeng Cai,Peng Jiang,Qibin Hou
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:A central limitation of current data-free self-evolution methods for training search agents is their reliance on difficulty-based proposer rewards. These methods reward a proposer for generating questions that challenge a co-evolving solver, using solver difficulty as a proxy for question quality. Yet difficulty alone is insufficient to distinguish questions that require cross-passage evidence from those that are answerable via simpler shortcuts. In addition, measuring difficulty demands repeated solver rollouts for every candidate question, leading to substantial computational costs. In this paper, we introduce \methodname, the first self-evolving search framework that eliminates difficulty-based proposer rewards and directly optimizes for evidence necessity relative to shortcut contexts. Dr. Free samples relational chains from a knowledge graph and pairs them with aligned passages, giving question generation an explicit multi-hop structure. A generated question receives a positive information-gain reward only when the likelihood of the target answer under the complete evidence passages exceeds the maximum likelihood under all evaluated shortcut contexts. Because this signal is computed from teacher-forced likelihoods, it removes the need for pass-rate estimation and reduces proposer training time by over 7\times . Experiments on seven open-domain QA benchmarks show that Dr. Free outperforms prior data-free search agents and the supervised baseline, with large improvements on multi-hop QA benchmarks.
[AI-316] OOD Generalization as a Bifurcation Problem
链接: https://arxiv.org/abs/2609.33562
作者: Nguyen-Thanh-Luong Doan,Quang-Vu Nguyen,Tang-Phu-Quy Le,Cong-Phap Huynh
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Systematic out-of-distribution (OOD) generation remains a critical bottleneck for continuous-time generative models. While standard joint classifier-free guidance (CFG) routinely fails to synthesize unobserved concept combinations, exact decomposed scoring generalizes robustly at the cost of severe computational overhead. In this work, we reveal that compositional binding is not a uniform process but a highly localized phase transition. We identify the semantic bifurcation window - the precise temporal interval where joint and decomposed vector fields meaningfully diverge. Exploiting this dynamic, we propose surgical guidance, a hybrid sampling strategy that restricts exact multi-pass scoring strictly to this critical window. On an OOD bi-digit MNIST testbed, surgical guidance achieves state-of-the-art compositional fidelity at a fraction of the inference cost, yielding a +5.3% absolute improvement in pairwise accuracy over the joint baseline by intervening during just the first 15% of the diffusion trajectory. Furthermore, our empirical analysis uncovers a fundamental topological divide: diffusion models (SDEs) force conceptual resolution immediately at peak noise, whereas Conditional Flow Matching (ODEs) delays structural binding until intermediate features emerge, establishing a new temporal framework for accelerating large-scale generative decoding.
[AI-317] rMeZO: Ternary Sparse Zeroth-Order Optimization for Fine-tuning BitNet Models at the Edge
链接: https://arxiv.org/abs/2609.33548
作者: Houssem Sifaou,Prabodh Katti,Bipin Rajendran,Osvaldo Simeone
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Signal Processing (eess.SP)
备注:
Abstract:Fine-tuning anguage models (LLMs) with first-order optimizers requires a memory several times larger than that required for inference. Memory-efficient zeroth-order optimization (MeZO) sidesteps this cost by estimating gradients from forward passes only. However, for BitNet architectures, a family of LLMs with ternary -1,0,1\ weights and 8-bit activations, fine-tuning requires updating full-precision latent weights, and thus the memory footprint of MeZO no longer matches that of inference. A promising solution is to finetune only a subset of the latent weights, but existing sparse zeroth-order (ZO) methods either ignore the ternary structure or require first-order gradient information to build a sparse mask, which is at odds with the purpose of ZO fine-tuning. We propose TerMeZO, a sparse MeZO scheme that exploits the geometry of the ternary quantizer itself to identify the latent weights that are more likely to change values during fine-tuning, at no additional data or memory cost. Our convergence analysis shows that TerMeZO can converge faster than full-parameter MeZO, owing to its optimized reduction of the fine-tuning effective dimension. We run extensive experiments on BitNet models ranging from 1B to 3B parameters, spanning classification, instruction-following, and mathematical reasoning tasks. TerMeZO matches or exceeds the performance of full-parameter MeZO while substantially reducing the fine-tuning memory footprint.
[AI-318] OSCC: Certified Observation-Safe Coupling Optimization for Gradient-Noise Control in Imperfect-Information Learning
链接: https://arxiv.org/abs/2609.33543
作者: Miaobo Hu,Shuhao Hu,Xiaobo Guo,Xin Wang,Bokun Wang,Rui Chen,Daren Zha,Jun Xiao
类目: Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG)
备注: 34 pages, 10 figures
Abstract:Coupled rollouts can reduce the noise of counterfactual action comparisons, but two issues prevent standard common-random-number constructions from serving as a general learning primitive in imperfect-information environments. First, an invalid coupling may expose hidden state, synchronize endogenous policy randomness, or misalign chance events after counterfactual histories diverge. Second, in multi-action policy optimization, lower return-contrast variance is not by itself the relevant objective: the optimizer depends on the return covariance matrix after projection through the local policy-gradient geometry. We introduce observation-safe counterfactual coupling (OSCC), a framework that defines an admissible class through marginal preservation, information-state safety, branch-local policy randomness, semantic event alignment, and trace-before-oracle replay. We derive a gradient-aware coupling criterion showing that, for marginal-preserving couplings, policy-gradient noise changes are determined by policy-Jacobian-weighted off-diagonal return covariance. This motivates OSCC-Select, a calibration-only selector that chooses among independent, root-only, continuation-only, and fully coupled rollouts using separate safety and gain certificates. Its gain target combines projected gradient noise with measured physical sampling cost and falls back to independent sampling whenever a simultaneous lower confidence bound does not certify improvement. On 100,000 fixed-root Leduc comparisons, the fully coupled CP-GRPO instantiation reduces return-contrast variance from 41.1158 to 18.1441, a 55.87% reduction, while preserving the declared branch marginals. With three actions, OSCC-Select chooses continuation coupling and attains gradient-noise trace 0.0783 versus 0.0917 for return-variance selection. Increasing calibration from 64 to 2,048 groups raises certification from 0.327 to 0.995.
[AI-319] Reasoning on the Simplex: Geometric Fixed-Point Models
链接: https://arxiv.org/abs/2609.33540
作者: Talgat Daulbaev,Ilya Glazkov,Maxim Rakhuba,Ivan Oseledets
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Looped reasoners spend test-time compute by iterating a weight-tied map, but a small residual does not mean the state is a fixed point when that map lives in unconstrained latent space. We propose Geometric Fixed-Point Reasoning (GFPR), in which the iterated state is the prediction itself: a field of categorical beliefs on a product of simplices, whose argmax is the answer at every step. Because the state is a belief, task structure can be imposed through compact convex relaxations, either as structured readouts or directly in the recurrent state; in the latter case the update remains a continuous self-map, so a fixed point exists for any parameters. At about 7M parameters, GFPR reaches 95.1% exact match on Sudoku-Extreme, 92.0% on Maze-Hard, and 100% sequence accuracy on S_5 length 128, above the published FPRM numbers at the same scale. The same update also trains a 201M language model on FineWeb-Edu in which each site is a distribution over the vocabulary; with 24 Picard steps it is above GPT-2 small on four zero-shot multiple-choice tasks and above GPT-2 medium on ARC-Easy.
[AI-320] EverMine: Dissecting the Self-Evolution of Research Capabilities in Long-Horizon Alpha Research
链接: https://arxiv.org/abs/2609.33524
作者: Siyuan Li,Jiangfeng Zhang,Rui Yao,Weihua Qiu,Mingyang Xu,Zixuan Yuan
类目: Artificial Intelligence (cs.AI); Statistical Finance (q-fin.ST)
备注:
Abstract:Self-evolving agents aim to turn research feedback into reusable skills, tools, and research rules. Whether these accumulated capabilities continue to improve later research requires controlled evaluation. Long-horizon alpha discovery provides a state-dependent setting: once a new factor enters the portfolio, the predictive information already covered changes, so the value of the same candidate or experience may change over time. We introduce EverMine, an empirical framework for studying self-evolving research capabilities in long-horizon alpha discovery. EverMine decomposes the research state into history (Hist), the current factor portfolio (Frontier), and reusable capabilities (Cap). Under matched resource limits, we compare complete runs with fixed or evolving Cap, and replace Cap while holding Hist and Frontier fixed to estimate the conditional value of accumulated capabilities. We also combine full trajectories with historical-state replay to examine how experience-based decisions affect candidate selection and portfolio outcomes. Across 18 long-horizon trajectories, end-to-end comparisons show no consistent gain from Cap evolution. Across 48 continuation branches from shared Hist and Frontier states, accumulated Cap also does not consistently outperform the initial Cap. Parameter tuning of existing factor structures can still improve the portfolio. In an exploratory replay of two screening batches from one Evolving trajectory, some screened-out candidates have positive marginal value at the original state, yet submitting all screened-out candidates sequentially slightly lowers final portfolio IC in both batches. These results show that candidate value depends on the evolving portfolio and submission order, and motivate evaluating self-evolving research capabilities through end-to-end outcomes, conditional capability value, and the consequences of experience-based decisions.
[AI-321] PPG-LM: A Photoplethysmography-Language Model with Multi-Level Clinical Alignment
链接: https://arxiv.org/abs/2609.33516
作者: Xiaoda Wang,Minxiao Wang,Maxwell A Xu,Patrick Langer,Kaiqiao Han,Defu Cao,Xiao Luo,Yuzhe Yang,Yan Liu,Xiao Hu,Yizhou Sun,Wei Wang,Carl Yang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Photoplethysmography (PPG) is widely recorded by clinical monitors and consumer wearables, providing a scalable source of continuous physiological information. These recordings offer an opportunity for physiological assessment at scale, but realizing this potential requires models to learn from both signal-derived physiological supervision and broader clinical context captured in electronic health records (EHRs). This involves aligning information spanning local observations, care events, and entire visits with PPG representations at corresponding temporal scales. However, existing PPG foundation models primarily rely on task-specific prediction heads, while the medical knowledge of large language models does not necessarily translate into waveform understanding. To bridge this gap, we introduce PPG-LM, the first PPG-language model family to learn physiological representations from both signal-derived supervision and broader clinical context captured in EHRs. To construct clinically grounded captions, we develop an automatic captioning pipeline that generates segment-, event-, and visit-level descriptions from signal measurements and structured EHR records. We then learn from these pairs through a two-stage framework that first establishes segment-language correspondence through contrastive learning and waveform-conditioned captioning, then extends alignment to events and visits through time-aware aggregation and temporal statement matching. Pretrained on approximately 73k hours of PPG, PPG-LM supports language-based recognition, cross-modal retrieval, and segment captioning. Experiments on MC-MED, MIMIC-III, and VitalDB show improved retrieval and caption factuality over language-model baselines and gains over PPG and time-series foundation models on multiple clinical prediction tasks.
[AI-322] Source Anchoring for Physical Consistency in Flow Matching Models ICLR2027
链接: https://arxiv.org/abs/2609.33510
作者: Giulia Romoli,Filippo Ruffini,Paolo Soda
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Submitted to ICLR 2027
Abstract:Deep generative models are used to solve partial differential equations and model distributions of physical system states, but ensuring that the generated samples satisfy the governing laws remains challenging. Projection-based flow-matching methods enforce physics by correcting the flow from an unconstrained noise distribution. These corrections shift the generated samples away from the distribution of target solutions, especially in high noise regions. To address this limitation, we propose Source Anchoring for Physical Consistency (SAPC), a Functional Flow Matching method that encodes the physical constraints into the source noise before generation begins. We evaluate SAPC on five systems governed by partial differential equations, covering six tasks with linear and non-linear dynamics, and compare results against five baselines and the unconstrained backbone. Anchoring the source reduces the need for large corrections that drive samples onto admissible but off-distribution states, and SAPC reproduces the target distributions most accurately on every evaluated task, while matching the constraint precision of the best projection-based baselines. Ablation experiments show that this gain arises from pairing source projection with a matched training objective that regresses toward the projected source. These results identify the source distribution as a key design choice for physically consistent generative modelling.
[AI-323] What Happens During Autonomous Deep Research After the User Steps Away?
链接: https://arxiv.org/abs/2609.33509
作者: Yimin Liu,Yijia Zhang,Yanmin Li,Tangwen Luo,Yuze Li,Ziling Yao,Zhi Yang
类目: Artificial Intelligence (cs.AI)
备注: 63 pages, including appendices
Abstract:In autonomous deep research, a user provides a task and relevant background, then leaves the agent to conduct an extended investigation without further human intervention. We study how this initial user information is reflected in intermediate actions and how these actions relate to final recommendations. We introduce DRaligned, a counterfactual behavioral evaluation framework built on PDR-Bench. By varying one task-relevant user factor while keeping the remaining context fixed, we compare acquisition requests, working drafts, and final reports. Source-grounded extraction, blinded local judgments, and deterministic aggregation yield coarse directional measurements while leaving ambiguous cases unresolved. Our experiments show that strong user-specific delivery can emerge from a largely shared research process: agents investigate similar broad questions but allocate requests differently, and final recommendations distinguish user conditions more clearly than explicit requests do. Reports can also integrate user factors that were not jointly visible during acquisition. In readable draft-to-report comparisons, recommendations often retain their coarse user-specific direction despite substantial rewriting. Final directional differences recur across tested agent models, execution harnesses, and evaluator models, even as execution paths vary. These findings describe how initial user information shapes autonomous research and clarify the relationship between the process an agent follows and the recommendations it delivers.
[AI-324] When Evidence Changes the Subject: Subject-Typed Claim Licensing for Learned Routing
链接: https://arxiv.org/abs/2609.33505
作者: Jian Chen,Zixuan Yuan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Modern learned systems increasingly combine learned components with search, repair, or external solvers. Benchmarks often measure the resulting end-to-end system, while scientific claims may concern only one component, creating an attribution problem: evidence can fail to support the requested component-level claim while still supporting a positive conclusion about the larger system. Existing evidence-to-claim methods primarily calibrate claim strength. We argue that composite systems require a second dimension: scientific subject. We address this problem with subject-typed claim licensing, which separates weaker conclusions about the requested subject from positive but non-substitutive credit about another subject. We instantiate this idea in SCOPE-Routing for preference-conditioned multigraph routing. Non-authors reproducibly apply the declared semantics; held-out review yields fewer reference-relative upward deviations than unstructured review, while the difference from a strong evidence checklist remains unresolved; and a controlled routing study shows that score-optimal and claim-eligible methods can differ while valid hybrid-system credit is preserved. These results motivate treating claim strength and scientific subject as distinct dimensions of evidence-based evaluation.
[AI-325] RelaxKV: Recomputation Guided by the Query with Sparse Context Attention for Efficient KV Cache Reuse
链接: https://arxiv.org/abs/2609.33503
作者: Ruoling Qi,Yirui Liu,Xuaner Wu,Yuxin Jin,Jian Chen,Jiayu Qin,Yin Chen,Jiawei Shao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Cross-request KV caching reduces the prefill cost of Retrieval-Augmented Generation (RAG), but conventional prefix caching severely limits cache reuse across requests. Position-Independent Caching (PIC) removes this constraint by reusing independent chunks, but their KV states miss cross-chunk interactions. Existing methods selectively recompute token states to recover these missing interactions, but primarily allocate the recomputation budget to selecting which states to recompute, while fixing the recomputation context to the full causal prefix. We introduce RelaxKV, which formulates selective cache repair as a joint allocation problem over repair targets and recomputation context. Guided by the user query, RelaxKV identifies layer-specific repair targets and restricts their recomputation to a query-relevant context, reducing attention computation. Across four decoder models, RelaxKV at a 15% anchor ratio improves aggregate LongBench performance over ProphetKV on all models. On Qwen3-14B, RelaxKV provides a stronger quality-TTFT trade-off than ProphetKV across a 5%-30% anchor-ratio sweep, and achieves the best selective results on RULER-MV and LV-Eval at 16K and 32K context lengths. Controlled ablations further demonstrate the importance of recomputation context selection.
[AI-326] Federated Multi-Modal Human Activity Recognition using Multi-Agent Reinforcement Learning
链接: https://arxiv.org/abs/2609.33492
作者: Debasmita Dey,Tanmay Sen,Himel Mallick
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Human Activity Recognition (HAR) from heterogeneous wearable sensors is fundamental to the Internet of Health Things (IoHT), supporting rehabilitation, elderly care, and smart healthcare. Existing multimodal fusion methods often assign fixed equal weights to sensor streams, overlooking differences in modality importance, acquisition cost, and sensor quality, which can vary due to movement, incorrect placement, or temporary blockage. We propose an adaptive and cost-aware multimodal HAR framework based on multi-agent reinforcement learning for centralized HAR and extend it to federated learning as FedMHAR. In the centralized setting, multimodal fusion is formulated as a cooperative Multi-Agent Reinforcement Learning (MARL) problem, where each sensing modality is assigned a PPO-based agent that learns per-sample fusion weights, enabling the model to emphasize informative modalities while down-weighting costly sensors when cheaper alternatives provide sufficient information. In the federated setting, we introduce BiFL-PPO, a bidirectional federated optimization strategy in which a server-side PPO policy learns client-specific trust weights and feeds them back to adapt local learning rates and proximal regularization. Unlike round-level optimization, BiFL-PPO uses dense batch-level rewards for more frequent feedback and stable training under heterogeneous client data. Evaluation on the MEx Rehabilitation and UTD Multimodal Human Action datasets shows that the centralized framework achieves 87.30% and 94.98% accuracy, respectively, outperforming conventional fusion methods and state-of-the-art HAR models. FedMHAR achieves 79.74% and 77.49% in the federated setting, consistently surpassing FedAvg, FedProx, FedBN, FedNova, and AdaFedProx, while providing more stable performance and reducing sensor acquisition cost.
[AI-327] Just Let Linear States Forget the Distant Past: Prefix Caching via Suffix Replay for Hybrid LLM s
链接: https://arxiv.org/abs/2609.33477
作者: Yirui Liu,Ruoling Qi,Xuaner Wu,Yuxin Jin,Jian Chen,Penghang Liu,Yafei Huang,Jiawei Shao,Xuelong Li
类目: Artificial Intelligence (cs.AI); Performance (cs.PF)
备注:
Abstract:Hybrid LLMs interleave full-attention layers with linear-attention layers to reduce long-context inference cost, but this structure complicates prefix caching. Full-attention KV caches are token-addressable, whereas linear-attention layers maintain recurrent states that cannot be rolled back to arbitrary prefix boundaries. Existing systems materialize recurrent-state checkpoints, restricting prefix reuse to checkpoint-aligned positions. We present SuffixReplay, the first prefix caching system that lets hybrid LLMs reuse cached prefixes at every cache-supported page boundary without materializing recurrent-state checkpoints. Our key insight is to just let linear states forget the distant past. Modern linear-attention mechanisms use recurrent decay and gating to attenuate the influence of old inputs. Therefore, instead of checkpointing every prefix boundary, SuffixReplay approximates the state at a matched boundary by replaying only a recent suffix of the layer’s input hidden states, which we retain as anchors. At the algorithmic level, SuffixReplay combines layer-wise and token-wise anchor sparsity with a bounded replay budget to control storage, computation, and quality. At the system level, it uses an independently managed anchor sidecar and a pipelined replay path to overlap anchor movement and state reconstruction with the native serving pipeline. We evaluate SuffixReplay on three hybrid LLMs: OLMo-Hybrid-7B, Qwen3.5-4B, and Qwen3.6-27B-FP8. Across these models, SuffixReplay retains 91.4-100% of full-prefill quality on average across LongBench and RULER, while using only 0.36-0.51x the amortized per-token storage of SGLang’s default 8192-token checkpoint cache. Integrated into SGLang, SuffixReplay reduces median TTFT by 15-70% on branching workloads, sustains 2.3-4.3x SGLang’s throughput when the working set exceeds HBM, and matches SGLang on high-hit continuation traffic. Subjects: Artificial Intelligence (cs.AI); Performance (cs.PF) Cite as: arXiv:2609.33477 [cs.AI] (or arXiv:2609.33477v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.33477 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-328] LiveOption: Evaluating LLM Agents in Structured Option Trading with Nonlinear Payoffs
链接: https://arxiv.org/abs/2609.33470
作者: Haochen Luo,Yifan Li,Binh Minh An,Xiaolong Luo,Zhengzhao Lai,Yuan Zhang,Chen Liu
类目: Artificial Intelligence (cs.AI); Computational Finance (q-fin.CP)
备注:
Abstract:Large language models (LLMs) and multi-agent systems (MAS) have shown promise in financial decision-making, yet existing evaluations focus on equity trading and primarily assess directional prediction, overlooking the structural complexity of derivative markets. Option trading introduces fundamentally different challenges, including nonlinear payoffs and multi-leg strategy construction, requiring structured decisions rather than simple directional bets. We introduce LiveOption, an evaluation framework for LLM-based agents in option trading. LiveOption formulates the problem as structured sequential decision-making under realistic execution and capital constraints, and provides a reproducible environment with standardized interaction protocols. The framework includes three task suites covering portfolio overlays, event-driven earnings trading, and 0DTE intraday trading. We further propose a hierarchical metric suite that evaluates action validity, decision quality, risk characteristics, and outcome-level performance. Experiments show that current agents often fail to achieve competitive returns in most scenarios. LiveOption offers a principled testbed for evaluating structured decision-making beyond outcome-based metrics.
[AI-329] Protected Cores Are Not Enough: Certifying AI-Proposed Revisions of Temporal Specifications
链接: https://arxiv.org/abs/2609.33461
作者: Ruggero Lanotte
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Software Engineering (cs.SE)
备注: 35 pages, 4 figures, 6 tables. Reproducibility artifact: this https URL
Abstract:Runtime monitoring traditionally evaluates a specification that is fixed before execution or externally modified when requirements change. In learning-enabled and data-intensive systems, however, the temporal relationships represented by a specification may themselves evolve. Allowing an AI component to directly replace a formal specification is unsafe: it may overfit transient behavior, weaken protected requirements, or activate statistically unsupported revisions. We introduce an intersymbolic architecture in which an untrusted AI proposer suggests temporal specification revisions and a symbolic governor controls their activation. Two results organize the framework. First, origin-version semantics makes the outcome of each obligation invariant to later revisions. Second, aggregate certification can conceal systematic failures on protected triggers; simultaneous aggregate and core-conditional post-selection certification controls both targets. A structural invariant preserves designer-protected components, and a proposer-independent lifetime error bound supports repeated activation decisions. The statistical bound concerns the predictable means of completed certification samples; interpreting it as future operational validity requires an additional stability assumption. Controlled synthetic experiments use a frozen supervised AI proposer to illustrate the masked-core failure at one decision and across repeated governed revisions. The proposer is a supervised regressor trained offline on synthetic tasks and frozen before use; it ranks candidates by predicted aggregate margin and never observes the protected-trigger success rate, so the masked-core failure arises from optimising the aggregate rather than from an adversary constructed by hand. Comments: 35 pages, 4 figures, 6 tables. Reproducibility artifact: this https URL Subjects: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Software Engineering (cs.SE) MSC classes: 68Q60, 68N30 ACMclasses: D.2.4; F.3.1 Cite as: arXiv:2609.33461 [cs.LO] (or arXiv:2609.33461v1 [cs.LO] for this version) https://doi.org/10.48550/arXiv.2609.33461 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-330] When Does the Concept of “Dog” Emerge in an Audio LLM ?
链接: https://arxiv.org/abs/2609.33458
作者: Zhe Wang,Shiqi Liu,Ruiyun Zhong,Tiechong Zhu,Yihua Tan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multimodal large language models answer audio questions, but how they represent auditory semantics and use them in decisions remains unclear, limiting our understanding of response formation. We study dog barking in Qwen2.5-Omni-7B using Jacobian lens (J-lens) readout and directional interventions. We define the dog direction as a J-lens-derived hidden-state vector associated with dog; adding or removing its component modulates dog-related information. We find this information decodable without dog/bark prompt cues or animal-identification requirements. Directional interventions change response tendencies and some final answers, with effects concentrated in late-layer states immediately before generation across species classification, vocalization classification, and sound description. The dog direction shows no comparable advantage over controls in animal/other classification. These results provide causal-intervention evidence that the dog direction affects output scores in a task-dependent manner, most consistently at L22 and L24 immediately before generation.
[AI-331] What Shared Prefixes Hide: Trajectory Dropout for On-Policy Distillation
链接: https://arxiv.org/abs/2609.33455
作者: Zzizhuo Lin,Quanling Liu,Yi Yang,Yawei Luo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:On-policy distillation (OPD) trains a student model on its own trajectories using dense token-level feedback from a stronger teacher model. Since each update is conditioned on the reasoning prefix already generated by the student, the prefix also shapes how effectively teacher feedback is converted into learning. We find that shared prefixes can lead to weak token-level updates, a phenomenon we call Prefix-Induced Supervision Attenuation (PISA). This attenuation arises in two common cases. (i) High student confidence can weaken corrective gradients even when the teacher disagrees. (ii) Tokens that rely on earlier reasoning can receive learning signals as weak as those for simple local continuations. To solve this problem, we propose Trajectory Dropout, a simple training-time intervention that exposes these weakened signals. The student first performs a standard full-context rollout to generate a complete trajectory. During training, we randomly drop a certain proportion of the student’s reasoning trajectory, while the teacher continues to observe the complete trajectory for token-level supervision. This intervention strengthens corrections for overconfident predictions and introduces additional supervision at prefix-sensitive positions. Trajectory Dropout consistently improves average performance across teacher–student model pairs of different scales and six mathematical reasoning benchmarks, while also yielding gains on two out-of-domain benchmarks. It can also be flexibly integrated into existing OPD variants with negligible computational overhead, further improving their performance. These results demonstrate that Trajectory Dropout provides a simple mechanism for strengthening token-level supervision across model scales and OPD objectives.
[AI-332] HESP: Separating What to Probe from When to Stop in Local LLM Alert-Triage Agents
链接: https://arxiv.org/abs/2609.33446
作者: Zhuowen Liu,Zhixuan Wang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 15 pages, 5 figures. Code and data: this https URL
Abstract:Security operations centers receive far more alerts than analysts can investigate, and organizations that cannot send their telemetry to hosted models must automate triage with small open-weight LLMs on their own hardware. Current LLM agents leave the investigation procedure to the model, and small local models fail at it: they probe without converging, never commit to a verdict, or dismiss real attacks. In this paper, we present HESP, a controller that holds the investigation procedure outside the model. HESP keeps a ledger of competing explanations, selects read-only probes by expected information gain per cost, accepts only verdicts backed by current evidence, can end an investigation itself, and journals every prediction before its observation. We evaluated HESP in four pre-registered studies with five open-weight models from two families (7B to 72B), totalling 7,272 audited episodes in a controlled triage environment. With likelihood tables counted from LLM-free runs, HESP lifts Qwen2.5-7B from 0.125 to 1.000 verified completion, matching oracle tables. The information-gain ranking adds +0.26 to +0.35 on every model that concludes, and a controller-side stop lifts Llama-3.1-8B, which never concludes on its own, from 0 to 0.917. What to probe and when to stop are therefore separate failures, and different small models exhibit different ones. Because HESP and its planner run entirely on local hardware, it suits environments where telemetry cannot leave the premises. We release all code, protocols, and episode journals at this https URL.
[AI-333] MAC-Net: A Multi-Task Deep Learning Framework for Modeling Cognitive Function From Task-Based fMRI
链接: https://arxiv.org/abs/2609.33440
作者: Md. Tanvir Rahman,Nabil Anan Orka,Asaduzzaman Khan,Mohammad Ali Moni
类目: Artificial Intelligence (cs.AI)
备注: 13 pages, 4 figures, 2 supplementary tables. This work has been submitted to the IEEE Transactions on Neural Systems and Rehabilitation Engineering for possible publication
Abstract:Objective cognitive assessment from neural signals supports neurorehabilitation, but individual-level prediction from task-based fMRI (tfMRI) remains difficult because neural features coexist with substantial demographic and scanner-related variation. We present the Multi-task Activation and Contrast Network (MAC-Net), a covariate-aware deep learning framework for modeling individual cognitive function from regional tfMRI. By isolating tfMRI features into a dedicated neural pathway and restricting participant variables to a terminal late-fusion pathway, MAC-Net prevents dominant covariates from suppressing high-dimensional clinical representations during feature learning. Evaluating baseline data from 6,500 Adolescent Brain Cognitive Development Study participants under family-aware cross-validation, MAC-Net was benchmarked against linear models, random forests, and alternative deep architectures. The N-back plus Monetary Incentive Delay configuration achieved R^2 values of 0.174, 0.238, and 0.277 for fluid, crystallized, and total cognition, outperforming covariate-only baselines (0.178) and alternative deep models (0.217). N-back was the most informative paradigm, whereas incorporating the Stop Signal Task marginally degraded performance. Feature attributions via Integrated Gradients, DeepLIFT, and Input Gradient were highly concordant, localizing working-memory-related frontal, parietal, and cingulate regions. These findings demonstrate that covariate-aware multi-task modeling yields reproducible cognitive-function estimations, establishing a robust neural engineering framework for clinical translation.
[AI-334] APEX: An Extensible Model for Agent -Assisted Production Scheduling
链接: https://arxiv.org/abs/2609.33430
作者: Felix J. Grumbach,Stefan Görlitz
类目: Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Production scheduling requires realistic models that reflect operational constraints and efficient methods that balance competing goals. Putting these methods into use also requires data integration, model adaptation and specialist expertise. We present APEX, an extensible production scheduling framework built around a general model and hybrid multiobjective search. Agent assistance supports both scheduling and model refinement: agents prepare data and explore scenarios in natural language, while coding agents help implement and test new constraints and objectives. Shared construction and checking procedures connect these adaptations to the scheduling core. We benchmark eight APEX configurations against NSGA-II, SPEA2, MOEA/D and SMS-EMOA on 69 public job-shop, flexible job-shop and permutation flow-shop instances, assessing workload completion time (makespan), total job flowtime and computation time. Hybrid configurations achieve the best aggregate solution quality, although the leading method depends on the problem class and objective. A separate synthetic workflow study uses OpenAI’s GPT-6-astra as an interaction layer between the human planner and the algorithmic core, testing rule additions, plan and objective changes, and what-if comparisons. All 24 sessions completed the requested changes and passed independent checks of saved models and schedules. A separate coding evaluation produced six native implementations of an additional objective or hard constraint through predefined extension hooks. All passed independent checks without modifying the core.
[AI-335] Graph-Guided Repository Environment Construction
链接: https://arxiv.org/abs/2609.33429
作者: Jianying Pan,John Zhang,Hongyu Zhang
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:Coding agents now increasingly rely on execution to validate their solutions, making the construction of reliable execution environments a critical enabling capability. However, repository environment construction is challenging because execution requirements are fragmented across repository artifacts and may only become apparent during execution. Existing agent-based approaches address this problem through iterative interaction, but information about the current construction state, including discovered requirements, satisfied and unresolved prerequisites, and their dependencies, can remain distributed across the interaction history. We present Graph2Env, an agent-based approach centered on DepGraph, a typed dependency graph that explicitly represents the environment requirements needed for repository execution, their dependency relations, and their states. Graph2Env uses DepGraph to guide environment construction and continuously refines it with execution feedback, while persisting successful repairs into a replayable construction procedure. The resulting artifacts are then applied in a fresh environment to verify that the constructed environment can be reproduced. We evaluate Graph2Env on a benchmark of 200 Python repositories drawn from RATBench and EnvBench, against a static dependency-inference baseline (pipreqs), three specialized environment-construction systems (Repo2Run, RAT, and SetupX), and two general-purpose coding agents (SWE-agent and Claude Code). Graph2Env achieves an 81.0% Environment Build Success Rate (EBSR) and a 59.3% Environment Setup Success Rate (ESSR), outperforming the strongest baseline by 9.5 and 9.0 percentage points, respectively.
[AI-336] mporal Graph Learning of Wearable Actigraphy and Sleep Traces for Modelling Adolescent Crystallized Intelligence
链接: https://arxiv.org/abs/2609.33428
作者: Md. Tanvir Rahman,Nabil Anan Orka,Asaduzzaman Khan,Mohammad Ali Moni
类目: Artificial Intelligence (cs.AI)
备注: 11 pages, 3 figures. This work has been submitted to the IEEE Transactions on Computational Social Systems for possible publication
Abstract:Wearable actigraphy offers a scalable, ecologically valid alternative to episodic clinical assessment. However, predicting continuous adolescent crystallized intelligence ( G_c ) from such traces remains challenging due to irregular device adherence and complex behavioral-environmental interactions. We address this using daily summary data derived from 21-day Fitbit records of 6,091 adolescents in the Adolescent Brain Cognitive Development Study (Release 5.1). We propose SATURN, a Sleep-Activity Temporal Unified Regression Network. It represents participants as 21-node temporal graphs encoding daily behaviors and temporal adjacency. To prevent imputation artifacts, invalid-day edges are dynamically pruned during forward passes. Node embeddings are refined via residual GATv2 layers, aggregated through masked attention pooling, and fused with sociodemographic covariates. Under family-controlled, age-sex-BMI-stratified cross-validation, SATURN achieves R^2 = 0.2783 \pm 0.0127 , consistently improving upon flattened machine learning (Gradient Boosting, R^2 = 0.2372 ) and sequential deep learning (BiLSTM, R^2 = 0.2688 ) baselines. Explainability analyses identify light activity, metabolic equivalents, and sleep duration as dominant predictors, while Monte Carlo dropout and subgroup analyses confirm equitable performance across sociodemographic strata. Ultimately, SATURN establishes a rigorous computational framework for digital cognitive phenotyping, offering a scalable pathway to complement traditional assessments by highlighting macro-level behavioral anomalies.
[AI-337] Investigating the Effect of k-NN Preprocessing on Developing Graph Neural Networks: A Fairness-Based Perspective
链接: https://arxiv.org/abs/2609.33416
作者: Nikolaos Zafeiropoulos,Emmanouil Mavrikos,George E. Tsekouras
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted for presentation at the 2025 16th International Conference on Information, Intelligence, Systems Applications (IISA), IEEE. Proceedings publication pending
Abstract:In this paper, a methodology to design fair graph convolutional neural networks (GCNs) is developed and tested over several application data sets. The graphs that are used as inputs to the network are constructed by a k-nearest neighbor-based preprocessing procedure, while fairness issues are considered in terms of the equalized odds criterion. To effectively incorporate the above heterogenous information, the equalized odds criterion is directly embedded into the model’s optimization objective through an additional fairness-driven loss functional term. The proposed methodology investigates how varying the neighborhood size in the k-NN algorithm during graph construction influences both the classification performance and the fairness of the resulting models. Extensive experimentation is conducted on three real-world tabular datasets with known biases, evaluating the interplay between graph structure and fairness enforcement. The results demonstrate that the choice of the value of the parameter k critically impacts the performance trends, either steadily improving or peaking at intermediate values depending on dataset characteristics, while the application of fairness constraints significantly mitigates disparities in false positive and false negative rates across groups defined by the protected variable at hand, without incurring major sacrifices in overall accuracy. This study highlights the importance of jointly optimizing the graph construction process and fairness objectives in GCN-based learning, providing a systematic approach toward building more equitable and effective graph-based models.
[AI-338] Evaluating System One Models for Agent Security Decisions: Reliability Calibration and Selective Automation
链接: https://arxiv.org/abs/2609.33401
作者: Yixuan Liu
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Model-based judges support agent security by detecting prompt injections, assessing interaction risks, and screening harmful requests. System One models expose typed decisions with probabilities that software can use to allow, block, or escalate inputs, but whether these probabilities support reliable automated security decisions remains unclear. We evaluate Jev, Laya, Decider, and Bespoke Nimble against specialized classifiers and language-model judges, examining decision accuracy, probability calibration, and selective automation. We identify three main findings. (1) Strong overall performance and favorable average calibration can hide systematic failures on particular attack groups, including attacks that models confidently classify as safe. (2) Under strict limits on missed attacks, the evaluated policies allow few inputs automatically, and choosing separate allow and block thresholds increases automation mainly by blocking more inputs. Even policies that meet error limits during validation can exceed them on unseen inputs. (3) A second judge can detect some missed attacks, but it may also reject more benign inputs and repeat the first model’s high-confidence errors. These findings show that model accuracy, probability calibration, and the behavior of the resulting decision policy must be evaluated together.
[AI-339] COEVO: Co-Evolving Context and Parameters for Recursive Self-Improvement
链接: https://arxiv.org/abs/2609.33398
作者: Siwei Chen,Xinping Bao,Xinyu Cai,Yuan Cao,Wan Jiang,Shaohong Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Recursive self-improvement (RSI) seeks to move large language models beyond static training pipelines toward systems that can participate in improving their own future behavior. Existing approaches largely follow two directions: updating model parameters through online learning, or improving the external context through search, reflection, and prompt optimization. Although both mechanisms can support continued improvement, they are typically studied independently. This separation overlooks an important interaction: the context shapes the experience from which a model learns, while an evolving model may interpret and utilize the same context differently over time. We therefore formulate RSI as a problem of parameter–context co-evolution, where model parameters and the learning context adapt within a shared feedback loop. We introduce COEVO, a framework that updates model parameters from on-policy experience while adapting contextual guidance according to the state of the evolving policy. Policy entropy and prompt-conditioned attention are used as complementary signals to guide this adaptation. Experiments show that COEVO consistently improves task performance over fixed-context reinforcement learning and produces policies that are more robust to changes in system prompts. More broadly, our results suggest that external context should be viewed not merely as a fixed interface to a large language model, but as an adaptive component of recursive self-improvement.
[AI-340] CoViST: Visual Token Compression via Composable States
链接: https://arxiv.org/abs/2609.33397
作者: Qi Zhang,Xiandong Meng,Ronggang Wang,Siwei Ma
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Visual token compression lowers the inference cost of vision–language models by representing images with fewer tokens. However, most existing methods compress visual tokens to a reduced set, leaving the amount of visual evidence represented by each token and its original spatial context implicit. Therefore, the compressed representation does not explicitly encode how much visual information each representative carries or where it lies in the original image. This limitation arises even after a single reduction and becomes more pronounced when compression is repeated across decoder layers. To address this issue, we propose CoViST, a training-free framework that represents a compressed image as a composable visual state. Specifically, the state combines representative features with original positions, effective contribution weights, and reusable selection metadata. CoViST constructs this state through coverage-guided selection and conservation-based contribution composition, and explicitly incorporates its contribution and positional information into decoder attention. Each component of the state retains its interpretation under successive reductions, enabling the same formulation to support both fixed compression before prefill and progressive compression within the decoder. Experimental results on seven LLaVA-1.5-7B benchmarks show that CoViST-Fixed retains 99.9%, 99.5%, and 98.1% of uncompressed performance at 192, 128, and 64 tokens, respectively, and CoViST-Pro retains 99.8%, 99.9%, and 99.1% at the corresponding layer-average budgets, outperforming state-of-the-art methods under their respective budget settings. Code will be released publicly.
[AI-341] Cross-modal Translation via Conditional Latent Denoising for Video Deepfake Detection
链接: https://arxiv.org/abs/2609.33394
作者: Xinzhe Li,Youzhi Tu,Kong Aik Lee
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:The growing threat of video deepfakes necessitates multimodal detection. Beyond serving as independent indicators of authenticity, audio and visual signals have intrinsic dependencies that also provide an essential criterion for detection. Previous methods often overlook the cross-modal correspondences, hindering information transfer between domains and leaving crucial detection cues unexplored. To address this challenge, we propose a framework called Cross-modal Translation via Conditional Latent Denoising (CTCLD) for video deepfake detection. It connects the distinct distributions of heterogeneous modalities in latent spaces, enabling smooth cross-domain information transfer to improve detection performance. We first establish a Bayesian foundation by decomposing the audio-visual joint distribution. Subsequently, CTCLD translates both modalities via bidirectional latent denoising conditioned on each other, effectively capturing subtle inconsistencies in the manipulated signals. Experimental results demonstrate that the proposed CTCLD enables comprehensive domain alignment, resulting in a robust video deepfake detection approach with competitive performance.
[AI-342] AutoHGNN: Robust and Efficient Neural Architecture Search for Hypergraph Neural Networks
链接: https://arxiv.org/abs/2609.33392
作者: Sirui Li,Pietro Liò b,Xinsheng Li,Baisong Liu,Chengbin Peng
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Hypergraph neural networks have achieved significant success in recent years. However, manual architecture crafting is labor-intensive and often fails to capture complex, higher-order relations, making the automation of hypergraph neural network structure design crucial. To improve the automation and adaptability of hypergraph learning, this paper proposes AutoHGNN, a neural architecture search framework tailored for hypergraph neural networks. First, we introduce a Hyper-Interaction Module (HIM) into the search space to address the mismatch between conventional graph neural network designs and hypergraph data. Second, we propose Hypergraph Stable Topological Distance (HyperSTD) as a structural selection criterion to identify architectures that best preserve the intrinsic structural affinities of the original hypergraph during differentiable search. Extensive experiments on various benchmark datasets demonstrate that AutoHGNN consistently outperforms manually designed and automatically searched baselines in classification accuracy and time efficiency, proving that the discovered architectures are significantly more effective.
[AI-343] Beyond Timestamps: Decision-Aligned On-Policy Distillation for Long-Horizon Agents
链接: https://arxiv.org/abs/2609.33391
作者: Mingju Chen,Can Lv,Jinrong Liu,Huan Zhang,Heng Chang,Shiji Zhou
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 27 pages, 10 figures
Abstract:Reinforcement learning with verifiable rewards (RLVR) often relies on sparse outcome rewards, providing coarse supervision for long-horizon agents. On-policy self-distillation (OPSD) complements this signal with dense privileged feedback. However, we identify \emphDecision–Timestamp Mismatch: privileged guidance may be misaligned with the student’s functional decision because the corresponding decision can occur at a different timestep, while the student’s decision itself may span multiple timesteps rather than being tied to a single timestamp. Thus, timestamp-local supervision can misalign both the context and the temporal scope of credit. To address this mismatch, we introduce \textscAlignOPSD, following the principle of aligning supervision before assigning credit. Decision-Aligned Supervision Rectification re-scores the same student-sampled response in functionally matched contexts across sibling rollouts to calibrate local teacher evidence. Semi-Markov Hierarchical Credit Assignment then derives variable-duration decision spans from correspondence changes and uses rectified evidence to allocate outcome-grounded credit across spans and their constituent turns. We evaluate \textscAlignOPSD with Qwen2.5-3B and Qwen2.5-7B on ALFWorld, WebShop, and Search-QA against representative baselines. \textscAlignOPSD outperforms both GRPO and StepOPSD across all eight backbone–aggregate-metric comparisons, improving on GRPO by 5.5–8.7 % and ranking first in six. Additional analyzes examine the two alignment stages and hyperparameter sensitivity between tasks. Our code is avaliable at this https URL
[AI-344] What Survives the Codec Shift: Pooled No-Vocals Residuals for Speech Deepfake Detection ICASSP2027
链接: https://arxiv.org/abs/2609.33375
作者: Jiajun Xu,Menglu Li,Xiao-Ping Zhang
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
备注: 5 pages, 2 figures, 3 tables. Prepared for submission to ICASSP 2027
Abstract:The transition from vocoder-based to neural-codec speech synthesis makes generalization more difficult for speech deepfake detectors, particularly those relying on speech-oriented representations. It remains unclear which acoustic representations retain discriminative information when the generation mechanism changes. We therefore compare 12 acoustic representations using a shared low-capacity linear classifier to identify effective evidence under codec shift. The analysis shows that hierarchical XLS-R leads on the pooled test set, while pooled no-vocals residual statistics perform best on the unseen-codec condition, revealing complementary behavior across generation conditions. Building on this finding, we propose MN-P, a dual-view detector that integrates an utterance-level pooled no-vocals representation with token-level XLS-R features through adaptive gating. The proposed MN-P reduces EER by 54.2% overall and by 60.9% on the codec-unseen condition relative to the best-performing retrained state-of-the-art system, with consistent gains across different detector backends. These results indicate that pooled no-vocals residual statistics provide effective complementary evidence for cross-generation speech deepfake detection.
[AI-345] API Secrets Should Never Become Tokens in the LLM s Vocabulary: A Threat Analysis of API Credential Handling in LLM Agent Systems and an Empirical Evaluation of a Vault-Mediated Execution Boundary
链接: https://arxiv.org/abs/2609.33371
作者: Patrick Kenney,Hadi Ahmadi,Denis Lusson,Donald Nguyen,Gurbinder Gill
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Tool-using large language model (LLM) agents turn credential hygiene from a storage problem into an execution-security problem. A key pasted into a prompt, or embedded in a system prompt or tool configuration, crosses from an authentication boundary into a data pipeline, where it may persist in conversation history, logs, memory stores, generated code, and error payloads. Prompt injection and excessive agency then convert passive disclosure into unauthorized action. This paper formalizes the credential-exposure threat chain for agentic systems; synthesizes evidence from a platform secret-store incident, vendor-reported secret-sprawl measurement, and OWASP and NIST guidance; and describes a vault-mediated execution architecture in which the model selects a connector identifier while a trusted request boundary supplies authentication. We evaluate a production implementation, Corvic Security Vault, in two controlled black-box experiments. Across 16 probes spanning seven control domains, every probe met its expected outcome: an authenticated GitHub API request succeeded while the credential stayed absent from process environment values, caller-visible request headers, tested filesystem locations, three third-party echo services, and two unrelated API origins; both cloud instance-metadata endpoints were unreachable. We also report a negative result, a connector whose stored header mapping did not satisfy its provider’s authentication contract, showing that centralized custody does not by itself guarantee correct configuration. Vault mediation removes several disclosure paths but is necessary rather than sufficient: least privilege, deterministic action authorization, human approval, telemetry redaction, and rotation remain independently required. The study is purposive and small, a functional security evaluation rather than a certification.
[AI-346] DrafTS: Time-Aware Decomposition with Residual Correction for Time Series Modeling
链接: https://arxiv.org/abs/2609.33368
作者: Yiqiu Liu,Siru Zhong,Zhiguang Wang,Qingsong Wen,Yuxuan Liang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Real-world time series contain evolving underlying dynamics with irregular variations that lack stable temporal patterns and are often referred to as noise. Existing methods address this mixture by filtering frequencies or suppressing noisy observations. They either miss temporal evolution or risk suppressing useful dynamics. We propose DrafTS, a model-agnostic framework that aims to reduce noise while preserving evolving dynamics through time-aware Decomposition with ResiduAl correction For Time Series. DrafTS uses features derived from instantaneous amplitude and frequency to guide decomposition into a primary component intended to capture underlying dynamics. A task-specific backbone models the primary component, while a lightweight correction module uses residual information to correct the backbone output. Across four time series modeling tasks, DrafTS improves six diverse backbones, demonstrating its effectiveness. Code is at this https URL
[AI-347] DISCERN: Can AI Agents Work Like Scientists and Guide Discovery?
链接: https://arxiv.org/abs/2609.33357
作者: Nan Huang,Mario Tapia-Pacheco,Kun Zhou,Yiming Huang,Kevin José Barrientos Díaz,Tiffany Amariuta,Jingbo Shang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Reliable automated research requires agents to vet data, verify analyses, and generate hypotheses grounded in trustworthy evidence, potentially reducing routine scientific workload while allowing scientists to focus on interpretation and discovery. Existing benchmarks often only assess analytical task completion or hypothesis generation separately rather than testing whether reliable evidence supports valid and novel claims. We introduce DISCERN (Data Integrity and Scientific Capability: Evidence, Reasoning, and Novelty), a controlled benchmark on real, publicly available datasets that evaluates three key levels of an automated research workflow. The first two levels test data integrity and analysis verification under confounds and tool traps, while the third tests hypothesis generation and revision under adversarial review, including counterfactual cases in which evidence consistent with real data and documented scientific phenomena conflicts with established expectations, motivating alternative explanations and testable hypotheses. Across 203 tasks, eight life-science tracks, and eight models, DISCERN shows that strong aggregate performance can mask level-specific weaknesses. Agents earn perfect scores in only 60.8% of Level 1, 34.2% of Level 2, and 0.6% of Level 3 evaluations, with penalties attributed to rejection of sound data, failure to carry recognized limitations into conclusions, and wide variation in hypothesis production. Cross-track rankings by token and code use are substantially more stable than rankings by evidence judgment, suggesting greater consistency in computational effort than in evidence-based reasoning. These profiles identify opportunities for supervised scientific assistance, but current agents do not yet demonstrate reliable autonomous analysis or discovery. Code and data: this https URL
[AI-348] Long-Horizon Analog Design Bench: Benchmarking Agents on Hours-Long Analog and Mixed-Signal Circuit Design Tasks
链接: https://arxiv.org/abs/2609.33356
作者: Analog Design Bench Team
类目: Artificial Intelligence (cs.AI)
备注: 25 pages, 10 figures, 6 tables
Abstract:Coding agents now sustain hours-long, tool-driven loops, yet their ability to carry long-horizon analog and mixed-signal circuits to electrical specification remains unmeasured. We introduce Analog Design Bench, a long-horizon agentic benchmark of 50 transistor-level design tasks contributed by 17 chip designers. Agents work with an open-source simulator, while an isolated verifier evaluates the submitted circuit using specification-based electrical tests. We evaluate 15 agent configurations across 2,250 two-hour attempts and observe full-specification pass rates from 8.0% to 78.0%. Coding-benchmark performance correlates with analog results but leaves much of the performance spread unexplained. Our failure analysis shows that most unsuccessful submissions have no recorded legality rejection but fail electrical acceptance, identifying electrical closure as the dominant endpoint challenge. We test time, reasoning effort, agent harness, and supplied design knowledge as interventions. Longer budgets and higher reasoning effort improve performance, while general skill documents provide little benefit and sometimes reduce performance. Supplying a task-matched reference topology, an idealized form of circuit-IP retrieval, raises DeepSeek V4 Pro by 18.7 percentage points and mainly accelerates GPT-5.6 Sol.
[AI-349] Unmask the State: When Does State Adaptation Matter for Masked Diffusion Language Models
链接: https://arxiv.org/abs/2609.33355
作者: Injin Kong,Sunghwan Choi,Yohan Jo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Masked diffusion language models (MDMs) admit flexible generation orders, making the unmasking strategy an inference decision. Existing methods vary in how they prioritize positions, control parallelism, restrict selection regions, revise predictions, or plan future denoising, yet it remains unclear when these choices should change during generation. We study this question through strategy reversals, where an alternative action becomes preferable to a fixed choice. We organize MDM inference into five axes–score, cardinality, region, commitment, and planning–and define adaptation opportunity as the one-step utility advantage of the best candidate action over a validation-selected fixed action. This view shows that adaptation value depends on both the frequency and magnitude of such reversals. Across three MDMs and ten tasks, adaptation opportunities are highly heterogeneous, with some regimes exhibiting concentrated and predictable one-step gains. This motivates selective adaptation: lightweight detectors calibrated on validation prompts identify high-opportunity states, capturing, for example, 56.9 percent of the candidate-set oracle opportunity by adapting only the top 10 percent of states on LLaDA-8B constrained JSON filling. Our transition-level results suggest that state adaptation is most useful when applied selectively rather than uniformly.
[AI-350] QuPID: Quantum Parameter-Efficient Input-Dependent Retrieval Adaptation for Medical RAG
链接: https://arxiv.org/abs/2609.33351
作者: Hyojun Ahn,Emily Jimin Roh,Soohyun Park,Walid Saad,Hyung-Chul Lee,Joongheon Kim
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 42 pages, 18 figures, 33 tables
Abstract:Fidelity-based quantum retrieval ranks candidates by the fidelity between query and archive states. Applying a shared input-independent unitary after fixed state encoding leaves that fidelity unchanged, so training the circuit cannot alter the ranking. Quantum parameter-efficient input-dependent retrieval adaptation (QuPID) repairs this by making the circuit input-dependent through data re-uploading and by comparing measurement readouts, vectors of local Pauli expectations, rather than states. The result is a small readout for adapting frozen image features to a local archive with limited data: training simulates the circuit classically, and inference runs on a GPU with fixed learned parameters. We characterize the class as a structured factorization of input-modulated quadratic feature maps, bound the frequency support of its re-uploading channel, and give a parameter-count generalization bound that motivates its small budget. Under a shared frozen backbone and a label-free protocol, QuPID’s 60 parameters give higher precision-at-5 (P@5) on ChestX-ray14 and MURA than frozen medical encoders, and than adapters and low-rank adaptation (LoRA) with up to 5.25 million trainable parameters. On ChestX-ray14, the P@5 gain over the frozen encoder is +0.116, the lead over retuned adapters is widest at 512 adaptation examples (+0.040), and the full-budget margin over an equally compact classical rotation-plane head is +0.023 with a 95% interval excluding zero. Medical imaging is the primary testbed; the pattern recurs on two non-medical benchmarks, in report generation, and under simulated gate noise and finite-shot readout.
[AI-351] KoopCell: Koopman-Based Generative Model for Learning Single-Cell Dynamics from Distribution Snapshots
链接: https://arxiv.org/abs/2609.33350
作者: Wanfeng Lu,Yutong Zhang,Keyi Zhou,Chenxin Ge,Wei Lin,Qunxi Zhu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Quantitative Methods (q-bio.QM)
备注:
Abstract:Learning population dynamics from temporally sparse, unpaired distribution snapshots is a fundamental challenge in developmental biology. Recent approaches based on neural differential equations and flow matching can interpolate between observed population snapshots, but may struggle to extrapolate beyond the training horizon and often lack an explicit mechanism for modeling developmental branching. We propose KoopCell, a unified generative framework based on Koopman-Mori-Zwanzig theory that jointly learns representations and predictive linear latent dynamics. Theoretically, using the weak continuity equation, we derive a closed-form least-squares estimator for the Koopman generator from distribution snapshots and establish convergence guarantees under suitable assumptions. To model branching dynamics, we further develop KoopCell-M, which incorporates non-Markovian memory into the latent Koopman dynamics through a Markovian embedding. Experiments on synthetic systems and three scRNA-seq datasets demonstrate the ability of our framework to recover Koopman spectra, model branching through memory, and scale to predicting high-dimensional gene expression distributions, achieving state-of-the-art performance among the evaluated methods.
[AI-352] Naturalness-guided Manifold Flow Matching for Sign Language Production
链接: https://arxiv.org/abs/2609.33339
作者: Jiayi He,Shengeng Tang,Sisi You,Yanbin Hao,Lechao Cheng,Richang Hong
类目: Artificial Intelligence (cs.AI)
备注: 25 pages, 4 figures
Abstract:Sign Language Production (SLP) aims to generate sign motions from text. Conditional Flow Matching methods have achieved strong performance in SLP by constructing conditional paths that transform a source distribution into a target distribution. However, existing methods construct these paths via linear interpolation, whereas the rotational geometry of human joints confines valid joint rotations to a manifold embedded in Euclidean space. Consequently, linear interpolation between two sign motions leaves this manifold and ignores the motion distribution on it. In this paper, we revisit SLP from the perspective of manifold transport and propose a Naturalness-guided Manifold Flow Matching framework, termed \textbfSignNMFlow, which constructs conditional paths directly on the motion manifold by jointly considering geometric efficiency and the motion distribution. Specifically, we exploit the intrinsic geometry of the manifold and introduce a motion naturalness measure to characterize the motion distribution. By minimizing the kinetic energy under this measure, we learn a naturalness-guided interpolation that couples a closed-form geodesic, which provides geometrically efficient transport, with a learnable deviation that incorporates the motion distribution, thereby significantly improving the fidelity of generated sign motions. Extensive qualitative and quantitative evaluations demonstrate the effectiveness of this work.
[AI-353] Beyond Conservatism: Recoverability-Conditioned Exploration for Model-Based Imitation Learning
链接: https://arxiv.org/abs/2609.33336
作者: Xuanlin Chen,Ziyue Wang,Xunlan Zhou,Yuan-yih Shang,Qiang Wu,Shenghua Wan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Model-based imitation learning (MBIL) improves real-environment interaction efficiency by optimizing policies on imagined rollouts from a learned world model. However, the gap between model-induced and real-environment occupancies makes policy learning sensitive to model error. Conservative MBIL mitigates model exploitation during policy optimization, but when real-environment interactions are collected by the same conservative policy, uncertain regions around the expert distribution remain insufficiently sampled. Generic uncertainty-driven exploration, on the other hand, may allocate interaction to novel but task-irrelevant dynamics. We propose REcoverability-CONditioned Exploration for Model-Based Imitation Learning (RECON). RECON separates conservative policy learning from active data collection by maintaining a main policy for task execution and an explorer for real-environment interaction. The explorer is optimized based on epistemic uncertainty conditioned on recoverability estimated from multi-step main-policy imagination, focusing data collection on unknown states from which the main policy can still return toward expert behavior. Experiments on locomotion, navigation and manipulation show consistent gains in interaction efficiency, imitation performance, and robustness, indicating that RECON directs real-environment interaction toward recovery regions around the expert distribution that are underexplored by prior methods, and thereby learns a world model better suited for imitation.
[AI-354] ANTMAN: Adaptive Need Tracking for Multi-Agent Navigation in Large Information Spaces
链接: https://arxiv.org/abs/2609.33326
作者: Jerry Wang,Haibo Jin,Xiaopeng Yuan,Peng Kuang,Haohan Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Information-seeking agents increasingly operate over information spaces that are too large to process exhaustively. Yet many multi-agent systems organize computation around static partitions of the available space, causing coordination to grow with how information is segmented rather than with what the query still requires. We introduce ANTMAN, an adaptive coordination framework that treats evolving unresolved information needs as the unit of runtime coordination. ANTMAN maintains a revisable Need Graph that tracks unresolved requirements, accumulated evidence, prior attempts, and search progress, and uses this state to control worker selection, routing, and task-local recovery as new evidence is discovered. By separating the coordination policy from substrate-specific search interfaces, the same need-conditioned mechanism can operate across different information spaces. Experiments across multi-document question answering, controlled long-context scaling, and realistic structured navigation show that ANTMAN remains effective across settings, including when execution is delegated to substantially smaller worker models. Under a 16x increase in searchable context, ANTMAN increases active coordination by only 1.23x, compared with more than 15x for partition-driven baselines, while preserving strong answer quality.
[AI-355] Agent ic Multi-Turn Reasoning : A Fairness Approach NEURIPS’26
链接: https://arxiv.org/abs/2609.33323
作者: Thanh-Dat Truong,Sankalp Pandey,Hugh Churchill,Jackson Cothren,Marios Savvides,Khoa Luu
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted to NeurIPS’26
Abstract:Recent advances in Large Language Models (LLMs) have enabled agentic systems capable of solving complex tasks through multi-turn planning, tool use, verification, and memory updates. However, learning agentic systems remains difficult due to two fundamental challenges, i.e., (1) long-horizon credit assignment, where supervision is available only at the final outcome, and (2) imbalanced data distributions, where dominant data patterns bias optimization and weaken adaptation to rare but informative reasoning behaviors. In this paper, we propose Fair Multi-Level Preference Optimization (Fair-MPO or \Phi -MPO), a new preference optimization framework for agentic learning. We first show that Multi-Level Preference Optimization provides a principled and more computationally efficient framework for long-horizon reasoning. Then, we introduce a Fair Multi-Level Objective that addresses imbalance in agentic learning. We provide a comprehensive theoretical analysis demonstrating that our approach addresses both long-horizon reasoning and data imbalance. Our experiments on agentic reasoning benchmarks demonstrate that our approach achieves State-of-the-Art (SOTA) performance.
[AI-356] PhysAlign: A Benchmark for Evidence-Grounded Role Alignment in Multimodal Physics Reasoning
链接: https://arxiv.org/abs/2609.33319
作者: Kecheng Liang,Haoyang Liu,Zexin Chen,Zirong Liu,Weixing Chen,Qiufeng Wang,Yang Liu,Liang Lin
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:A key challenge in physics diagram understanding is correctly associating visual information with the physical entities, relations, and conditions it describes. Even when a value, symbol, or other local element is accurately recognized, assigning it to the wrong entity or scope can distort the underlying physical premise and lead to incorrect reasoning. To systematically study this challenge, we introduce \textbfPhysAlign, a benchmark designed to assess whether multimodal models correctly associate information recognized from physics diagrams with its intended physical role. By disentangling visual recognition from physical-role assignment through localized probes and controlled variants, PhysAlign isolates correspondence errors from recognition failures. It contains 3,341 human-validated probes spanning 986 physics problems, enabling systematic evaluation of visual recognition and physical-role correspondence at scale. We further introduce five complementary evaluation metrics, including CAcc, GAcc, and JAcc, which provide a comprehensive assessment of models’ ability to recognize diagram content, establish correct physical correspondences, and solve the underlying physics problem. Across our evaluated multimodal models, PhysAlign reveals a consistent gap between local visual recognition and physical-role grounding. Even when the queried content is correctly recognized, the conditional correspondence error rate remains 13.8% for GPT-6-Astra and rises to about 50.6% for InternVL3.5-8B. These findings indicate that strong perception alone does not ensure reliable physical interpretation, exposing a distinct grounding bottleneck that is largely hidden by answer-level accuracy and highlighting the need for future models to better align recognized visual evidence with its physical meaning.
[AI-357] ZeroGAR: Benchmarking the Adversarial Robustness of Zero-Shot Graph Models
链接: https://arxiv.org/abs/2609.33314
作者: Zhongjian Zhang,Xiao Wang,Busheng Zhang,Bo Yan,Xingtong Yu,Yue Gao,Jia Li,Chuan Shi
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Zero-shot graph models (ZGMs), which learn transferable knowledge from source graphs and directly apply to unseen target graphs without any adaptation, have achieved promising performance and attracted considerable attention. Despite their proliferation, existing ZGMs are predominantly evaluated on clean graphs, while existing graph robustness benchmarks mainly focus on supervised settings, leaving a fundamental question largely unexplored: How robust are ZGMs when their unseen target graphs are exposed to adversarial manipulation? In this paper, we answer this question by proposing ZeroGAR, the first systematic benchmark for evaluating the adversarial robustness of ZGMs. ZeroGAR evaluates 13 representative ZGMs from 3 different paradigms on 8 graph datasets across 4 domains, covering both in-domain and cross-domain transfer under structural, textual, and node injection attacks with multiple perturbation budgets. It further investigates whether existing graph defenses remain effective in the zero-shot setting. Extensive experiments reveal that strong clean zero-shot performance does not guarantee adversarial robustness, with three key findings: (1) Vulnerability patterns are related to model prediction mechanisms: GNN-based methods are particularly vulnerable to structural and node injection attacks, whereas LLM-based methods are more vulnerable to textual attacks; (2) Stronger LLM backbones introduce a structure-text robustness trade-off; (3) Existing graph defense methods do not consistently improve zero-shot robustness and may compromise clean performance. We hope that ZeroGAR will facilitate rapid, equitable evaluation and inspire further innovative research in ZGM security.
[AI-358] BITS: Rethinking Fair and Comprehensive Evaluation for Irregular Time Series Forecasting
链接: https://arxiv.org/abs/2609.33303
作者: Kangjia Yan,Linfeng Wang,Tianen Shen,Xiangfei Qiu,Ruitong Zhang,Hao Miao,Jilin Hu,Chenjuan Guo,Bin Yang,Christian S. Jensen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Despite recent progress in irregular time series forecasting, the field still lacks a unified benchmark for fair and comprehensive evaluation. Existing evaluations are often conducted on a limited set of datasets with inconsistent experimental protocols and predominantly error-based metrics, rendering it difficult to compare and assess methods fairly and comprehensively across diverse settings. To eliminate these limitations and accelerate progress, we propose BITS, a standardized, reproducible, and extensible benchmark for advancing research on irregular time series forecasting. BITS covers eleven datasets from nine domains with diverse irregularity characteristics, and it characterizes the datasets according to their missing rate, missing pattern complexity, sampling irregularity, and skewness. Further, it offers a unified pipeline for data preprocessing, model integration and evaluation, and reporting. It accommodates regular and irregular time series forecasting methods, including time series foundation models, under consistent settings, incorporating both error-based and non-error-based evaluation metrics. Findings include that method performance varies substantially across irregularity characteristics, with no single modeling strategy consistently dominating. We also find that using error-based or non-error-based metrics can yield different model rankings, highlighting the need for multi-dimensional evaluation. The code can be found at this https URL.
[AI-359] AquaWAM: A Dynamics-aware World Action Model for Underwater Embodied Agents
链接: https://arxiv.org/abs/2609.33299
作者: Cunhao Zhu,Yifeng Wang,Dongliang Xu,Yunzhong Hou,Yue Yao,Chi Harold Liu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:World Action Models (WAMs) are becoming increasingly important and useful for embodied intelligence, as they enable robots to anticipate the consequences of candidate actions before interacting with the physical environment. However, underwater robots are usually subject to passive dynamics, such as inertia, buoyancy, hydrodynamic drag, and persistent drift, which can continue to affect the vehicle even after an action is completed. Existing WAMs, which primarily predict action-conditioned visual observations, are not explicitly designed to capture such passive motion dynamics. In this paper, we present AquaWAM, the first World Action Model designed for underwater embodied agents. Instead of predicting future images, AquaWAM models both action-conditioned and passive physical dynamics, including the thruster dead band, the inertial glide that outlasts each command, and ambient currents. Specifically, it senses through the DVL, IMU, pressure sensor and joint encoders, while cameras supply only semantics for understanding goals and target pose. By modeling compact navigation states rather than high-dimensional visual observations, AquaWAM substantially reduces the model size and computational cost compared with conventional WAMs. Experimentally, AquaWAM achieves a 72.6% task success rate across 20 underwater tasks on the USIM benchmark, outperforming existing methods while making action decisions 2.7x faster than U0 on an NVIDIA Jetson AGX Orin. Our model also remains effective when some onboard sensor measurements are unavailable. For example, without DVL velocity measurements, our method still achieves a 61.6% success rate, compared with 39.4% for U0.
[AI-360] he Error You See Is Not the Error You Made: Progression-aware Reasoning Origin for Reasoning Error Localization
链接: https://arxiv.org/abs/2609.33297
作者: Yiguo Wang,Ziyuan Yang,Yi Zou,Dan Lin,Rongsheng Li,Yi Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Verifying multi-step LLM reasoning requires more than determining whether a trace is correct: a useful verifier should identify where the reasoning first goes wrong. However, existing holistic methods provide little positional evidence, while forward sequential verification often treats the first rejected step as the error source. Under error propagation, this assumption can fail, since an earlier mistake may remain locally plausible and become observable only through its downstream consequences. We therefore rethink reasoning verification as a progression-aware error-source localization problem: rather than asking only where a reasoning trace first appears inconsistent, we ask which earlier step best explains how that inconsistency emerges along the trajectory. Based on this view, we propose Progression-aware Reasoning Origin (PRO), a training-free framework for first-error localization. PRO jointly models incoming support from the preceding context and outgoing compatibility with subsequent reasoning, selectively refines regions where these signals disagree, and finally performs detector-conditioned source attribution with intervention-based evidence to distinguish the true error origin from its propagated manifestations. We further formalize the gap between forward rejection and structural exposure, showing why incoming-side evidence alone is insufficient for reliable localization under error propagation. Experiments across open-form, medical, and structured reasoning tasks demonstrate consistent improvements over strong verification baselines, supporting progression-aware source attribution as a more faithful formulation of reasoning verification.
[AI-361] Learning to Sell: Reinforcement Learning for Strategic Large Language Model Agents in Multi-Product Markets
链接: https://arxiv.org/abs/2609.33289
作者: Shuze Daniel Liu,Claire Chen,Jiuqi Wang,Thorsten Joachims
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Autonomous large language model (LLM) agents operating in multi-product markets must make sequential decisions under information asymmetry and resource constraints. We develop a machine learning approach for training such agents to act effectively as sellers in a multi-item bargaining environment, where a seller concurrently negotiates a catalog of substitutable assets across a pool of independent buyers. Buyers hold private, heterogeneous valuations across products, and each can purchase at most one item. Facing limits on total communication turns, the seller must dynamically match buyers with the most profitable products considering their private valuations, while strategically allocating its limited interaction budget toward combinations of greater potential value. We formalize this problem as a Partially Observable Markov Decision Process using a structured, four-part message protocol that maps natural language into a parsable and regulated decision space. Using this formalization, we design a post-training method using Reinforcement Learning from Verifiable Rewards (RLVR). To evaluate this framework, we construct a multidimensional metric suite that quantifies constraint adherence, seller surplus extraction, and allocation quality. Our trained seller agent learns to match limited inventory to buyers more effectively, matching or outperforming trillion-parameter frontier models in both seller surplus extraction and buyer-product allocation quality. Finally, these learned strategies generalize robustly to unseen market structures, correlated valuation distributions, and price ranges not encountered during training.
[AI-362] Feedback Makes Perfect: A Closed-Loop Framework for NL-to-STL Translation
链接: https://arxiv.org/abs/2609.33287
作者: Bowen Ye,Xiang Yin
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO); Systems and Control (eess.SY)
备注:
Abstract:Signal Temporal Logic (STL) enables rigorous verification and control of cyber-physical systems, but writing correct specifications requires expertise that most requirement holders lack. Large language models can translate natural-language (NL) requirements into STL, yet stronger translators alone approach an accuracy ceiling. We argue that this ceiling stems from how the task is posed: one-shot, open-loop translation is somewhat ill-defined. Natural language is ambiguous, and, more fundamentally, what a person writes may not always be what they intend, so the target specification is not fully contained in the input text. We therefore reformulate NL-to-STL translation as a closed-loop feedback process. Each generated formula is translated back into natural language for the user to check, and natural-language corrections drive revision until the user accepts the specification. Users never read or write formal syntax. This framework rests on an asymmetry familiar from feedback control theory. The forward path, from ambiguous language to formal logic, is hard and error-prone. The feedback path, from structured STL back to language, can be made highly precise, and a precise feedback path lets an imprecise forward path achieve precise closed-loop behavior. Experiments on 500 expert-authored requirements and seven LLMs support this view. Back-translated explanations agree with expert judgments in 99.5% of cases. Closed-loop refinement raises strong models from about 89% open-loop accuracy to 98.0–99.2%, and yields gains of over 30 percentage points for weaker models (e.g., 17.6% \rightarrow 48.0%). Ablations show these gains come from the semantic content of the feedback rather than from repeated attempts. An expert audit and a 280-session user study further confirm the reliability of the loop. We also identify a capability threshold above which feedback no longer helps.
[AI-363] RINI: Seeing the Prior Is Not Enough
链接: https://arxiv.org/abs/2609.33284
作者: Hongyi Du,Tianyi Zhang,Heng Wang,Zhelun Gao,Yimei Liu,Ambrose Luo,Annie Hao,Jiayan Ni,Jiawei Han,Jiaxuan You
类目: Artificial Intelligence (cs.AI)
备注: 55 pages, 4 figures
Abstract:A research proposal can describe an established mechanism correctly while claiming to introduce it. We study whether providing the earlier paper corrects such contribution claims. Three controlled experiments compare proposals generated with a contribution-bearing prior and a same-topic control. Providing the prior yields no clear aggregate reduction in unsupported novelty. Human analysis of 175 interpretable exposed proposals finds that 137 recognize the prior’s relevance, but 61 correctly attribute the established contribution. Of 71 proposed remaining distinctions, 37 are covered by the same prior. We introduce Research Idea Novelty Inspection (RINI), which audits contribution claims against evidence, checks the remaining distinction, and applies local revisions. Five human annotators evaluate 1,080 original-revision pairs across three methods. On the same 240 originals judged to require correction, successful repair is 11.7% for Self-Revision, 39.1% for Retrieve-and-Revise, and 72.2% for RINI, with research tasks weighted equally. The improvement over same-evidence direct revision is 33.0 percentage points. The revised proposals retain their research questions and technical methods. These results motivate explicit contribution attribution when using literature to generate and revise research proposals.
[AI-364] Multi-Dimensional Comparative Scale Construction for Efficient Personalized Subjective Judgment in High-Traffic Applications
链接: https://arxiv.org/abs/2609.33282
作者: Xianglong Shi,Shifeng Liu,Sirui Zhao,Shengming Yuan,Enhong Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Subjective judgments are central to many high-traffic applications, but subjective intensity is difficult to quantify and perceptions vary substantially across individuals. To address these challenges, we propose a pairwise comparative framework for multi-dimensional scale construction. By comparing case-person pairs along case and profile dimensions, the framework constructs relative scales that capture both fine-grained intensity and individual variation. To support practical high-traffic deployment, we optimize both offline scale construction and online inference. For scale construction, we combine sparse Elo comparisons with multi-judge voting, cutting the comparison cost from O(N^2) to O(NK) for N objects and a budget of K opponents per object, while limiting reliance on any single judge. For inference, we propose SubJudge, a System One model for personalized scoring with Batchwise Preference Optimization (BPO). Using Bradley-Terry comparisons, BPO trains the model to learn relative orderings, and SubJudge reads a continuous score from digit-token probabilities at the first response position, requiring only one forward pass per criterion and reducing the inference complexity to O(1) . Experiments on PluriHarms and iNews show that our 9B models match or surpass the evaluated frontier LLMs on multiple metrics. On the H100 GPU, SubJudge achieves an approximately 1.29\times to 261\times speedup in mean inference latency over Qwen3.5-9B with different thinking budgets. The code is available at this https URL.
[AI-365] Domain Generalization under Sampling Pattern Shifts in Irregular Time Series
链接: https://arxiv.org/abs/2609.33279
作者: Changhun Kim,Joohyung Lee,Kwanhyung Lee,Donghwee Yoon,Grigorios Chrysos,Eunho Yang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Irregularly sampled multivariate time series (ISMTS) are prevalent in real-world applications, where both observation times and available measurements can vary substantially across domains. While recent models increasingly exploit such sampling information for prediction, its robustness under sampling pattern shifts remains underexplored. We introduce HAR-C, to the best of our knowledge the first controlled benchmark for sampling pattern shifts in ISMTS, and show that sampling shifts alone can substantially degrade performance, induce sampling-specific shortcuts, and remain challenging for existing domain generalization (DG) methods. Motivated by these findings, we propose PRISM, a DG framework that first learns complementary feature-centric and sampling-centric representations without task labels, and subsequently performs robust supervised training across diverse sampling variations to discourage brittle shortcut reliance. Extensive experiments on controlled and real-world ISMTS benchmarks demonstrate that PRISM consistently improves robustness to unseen sampling shifts over existing methods. Our code is available at this https URL.
[AI-366] ChronoFlow: Hierarchical Flow Matching for Irregular Time Series Generation
链接: https://arxiv.org/abs/2609.33276
作者: Changhun Kim,Sunguk Jang,Jeongjun Lee,Juhwan Choi,Sangchul Hahn,Grigorios Chrysos,Eunho Yang,Juho Lee
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances in generative modeling have substantially improved time series generation, yet most existing methods either assume a regular temporal grid or focus on feature dynamics under a given sampling structure. This makes them illsuited for generating irregular time series in their native form, where a model must capture not only feature values, but also how many observations occur, when they occur, and which features are observed together. To address this heterogeneous generation problem, we propose ChronoFlow, a unified hierarchical flow matching framework organized by statistical granularity. Following a coarse-to-fine hierarchy, ChronoFlow first generates observation counts and feature-wise frequencies, then jointly generates observation times and feature co-observation patterns, and finally generates values conditioned on the realized pattern. This turns a complex joint generation problem into structurally aligned subproblems while preserving their dependencies. To evaluate complete irregular time series generation, we introduce complementary metrics spanning sample realism, sampling structure, value fidelity, and temporal and cross-feature dependencies, and validate them through controlled corruptions. Across five benchmarks, ChronoFlow achieves strong improvements in generation fidelity over existing baselines, while factorization studies support the proposed hierarchy. Our code is available at this https URL.
[AI-367] Next Thoughts Are Distributions: Generative Autoregressive Reasoning in the Latent Space
链接: https://arxiv.org/abs/2609.33271
作者: Yang Li,Yi Wang,Shiyuan Huang,Yang Liu,Hao Wang,Chengzhi Mao
类目: Artificial Intelligence (cs.AI)
备注: 16 pages
Abstract:Reasoning problems often admit multiple valid ways to proceed. Continuous reasoning promises to move computation beyond language tokens into a more compact latent space, but representing several plausible ways to think next remains difficult. We introduce Autoregressive Thought Flow (ATF), which models the next continuous thought as a multimodal distribution. A causal autoregressive model performs the reasoning computation, while a lightweight diffusion head generates a plausible next thought from the resulting condition. The sampled thought is fed back into the model, allowing continuous reasoning to unfold for a variable number of steps while preserving the pretrained backbone. Across mathematical reasoning tasks, ATF improves accuracy with compact latent traces and benefits from reinforcement learning and additional test-time thinking. Multi-sample evaluation shows broader solution coverage, indicating that its multimodal predictions capture useful diversity among reasoning paths. Our results suggest that continuous reasoning is more effective when multiple possible next thoughts remain available rather than being collapsed into a single prediction.
[AI-368] Structured Sparse Memory for Recurrent Reasoning NEURIPS2026
链接: https://arxiv.org/abs/2609.33270
作者: Zixuan Zhao,Samuel Wheeler,Neil Getty,Xiaotian Duan,Rick Stevens,Fangfang Xia
类目: Artificial Intelligence (cs.AI)
备注: Accepted to NeurIPS 2026
Abstract:Recurrent models trained from scratch have recently become competitive on ARC-style reasoning tasks, but the usual framing around small recurrent backbones overlooks two important parts of the system: task-conditioned memory and synthetic augmentation data. We study this regime through CHARM, a compact hybrid ARC model that combines recurrent reasoning with structured task memory, synthetic data, and inference-time aggregation. In existing approaches, task-conditioned memory supplies a large hidden source of capacity, reaching more than 30x the size of the recurrent backbone. We introduce a compositional sparse embedding (CoSE) for task conditioning that reduces learned task-memory parameters by over 90% while improving pass@2 in controlled ARC ablations. For the recurrent backbone, recurrent depth helps only when balanced with learning horizon. Combining these ingredients, our system reaches 84% pass@2 on ARC-AGI-1 and 46.7% pass@2 on ARC-AGI-2 public evaluation. The benefits of structured memory also generalize to unseen puzzles and other domains. Our code, dataset, and model checkpoints are available at this https URL.
[AI-369] LSTMem: Hierarchical Long Short-Term Online Memory for Large Language Models
链接: https://arxiv.org/abs/2609.33268
作者: Xianglong Shi,Ruijie Yang,Sirui Zhao,Shukang Yin,Zihao Bian,Tinghao Yi,Enhong Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models increasingly serve as long-horizon assistants and agents, where they must both accumulate information across interactions and make the relevant parts available when later requests depend on them. Existing compact online memories typically use a single persistent state both to accumulate history and to serve readout, so what the memory stores cannot be controlled separately from what it exposes to the current computation. We propose LSTMem, an LSTM-inspired online memory that instead equips each layer of a frozen LLM with two matrix-valued states: a cell state that accumulates history and a hidden state whose readouts correct the backbone’s attention. Input and forget gates control what the cell stores, while an output gate separately controls what the cell exposes through the hidden state. LSTMem further connects memory across depth through forward hidden-state propagation and block-end feedback, and uses higher-layer reconstruction gradients to refine lower-layer cell states before rebuilding hidden states from shallow to deep layers. Across memory benchmarks on Qwen3-4B-Instruct, LSTMem consistently improves MemoryAgentBench, LoCoMo, and HotpotQA over the plain backbone. Comparisons further show that the LSTM-based memory formulation outperforms an associative-memory counterpart, while removing cross-layer hidden-memory propagation degrades performance. These results demonstrate the benefits of separating memory accumulation from memory expression and organizing memory hierarchically across model depth. The code is available at this https URL.
[AI-370] SCISSOR: Score-Conditioned Instrument Source Separation for Orchestral Recordings ICASSP2027
链接: https://arxiv.org/abs/2609.33265
作者: Yiheng Lu,Hao-Wen Dong
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注: 5 pages, 1 figure, 3 tables. Submitted to ICASSP 2027
Abstract:Orchestral separation recovers instrument sections from mixtures in which shared pitches, harmonics, and timbres obscure source identity. An aligned score provides instrument labels, note pitches, and activity times. A score-informed approach appends piano rolls to audio features before mask prediction. We introduce SCISSOR (Score-Conditioned Instrument Source Separation for Orchestral Recordings), which uses the score to form a frame-wise query for each source. Each query matches a shared audio representation, and a softmax over instrument and background slots jointly assigns overlapping time-frequency evidence. The queries retain instrument identity even when notes are missing from the score. After training on SynthSOD and a small set of URMP and PHENICX-Anechoic recordings, SCISSOR achieves the highest average SDR on held-out real recordings. With SynthSOD-only training, it leads on SynthSOD and zero-shot PHENICX-Anechoic, and improves on its audio-only control on zero-shot URMP. SCISSOR also degrades less under score corruption than the evaluated score-based baselines.
[AI-371] ActiveMem: Dynamic Latent Memory Trees for Long-Horizon Agents
链接: https://arxiv.org/abs/2609.33244
作者: Song-Li Wu,Jingyi Wang,Zhaocheng Du,Weinan Gan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Model (LLM) agents increasingly rely on external memory to support long-horizon reasoning and decision making. Existing memory systems typically retrieve historical trajectories or summaries as independent context fragments, overlooking the procedural dependencies underlying multi-step execution. As memory scales, such flat retrieval introduces context fragmentation and cross-task interference, leading to structurally inconsistent reasoning trajectories. We propose ActiveMem, a hierarchical memory framework that recursively organizes agent experiences into dependency-aware latent execution trees. ActiveMem abstracts trajectories into reusable subtask nodes while explicitly preserving execution transitions, enabling coherent reasoning-path retrieval conditioned on the current execution state. To support continual adaptation, ActiveMem further learns dynamic memory expansion, retrieval, and pruning policies through reinforcement learning. Experiments across various agent benchmarks demonstrate that ActiveMem consistently improves task completion, reasoning stability, and memory efficiency over existing memory-based agents. Moreover, ActiveMem enables compact open-weight models to achieve competitive performance with substantially larger proprietary systems.
[AI-372] CodeSkill: Latent Skill Abstraction for Long-Horizon Code Agents
链接: https://arxiv.org/abs/2609.33243
作者: Song-Li Wu,Jingyi Wang,Zhaocheng Du,Weinan Gan,Weiwen Liu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Code agents require long-horizon decision-making over complex interaction trajectories. However, existing reinforcement learning (RL) approaches typically optimize behavior at the token level, creating a mismatch between low-level generation and high-level behavioral reasoning. This limitation leads to inefficient exploration and weak credit assignment under sparse rewards. Moreover, while large-scale agent trajectories often contain recurring multi-step behavioral patterns, their noisy token-level representations hinder effective experience reuse. To address these challenges, we propose CodeSkill, a framework that adapts hierarchical latent skill modeling to the code agent domain. CodeSkill first leverages a teacher model to distill both successful and failed trajectories into multi-level textual abstractions. It then integrates temporal variational inference with reinforcement learning to map these discrete semantics into continuous latent variables, while an adaptive boundary mechanism dynamically gates skill transitions based on execution feedback. The learned skills are injected into a frozen LLM policy as latent semantic prefixes, enabling optimization in a compact semantic space rather than over raw token sequences. By shifting RL from token-level exploration to experience-level reasoning, CodeSkill improves optimization efficiency and long-horizon behavioral coherence. Extensive experiments demonstrate that CodeSkill achieves highly competitive performance against strong open-weight baselines across diverse general and industrial coding benchmarks. Furthermore, the learned skills exhibit strong transferability and robust cross-domain generalization, highlighting the effectiveness of explicit behavioral abstraction for scalable agentic code generation.
[AI-373] SMORE: Stability-Promoting Mesh-Agnostic Model Reduction for Time-Dependent PDEs
链接: https://arxiv.org/abs/2609.33205
作者: Yangyuan Li,Weichao Li,Shaowu Pan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 52 pages
Abstract:High-fidelity simulations of time-dependent partial differential equations (PDEs) are computationally expensive, motivating data-driven reduced-order surrogates for many-query tasks such as uncertainty quantification, design optimization, data assimilation, and optimal control. However, existing surrogate models often exhibit poor temporal stability, which can lead to unstable rollouts and exploding gradients during backpropagation, especially in multistep long-horizon forecasting. To address this, we propose SMORE, a mesh-agnostic framework for model order reduction of time-dependent PDEs. Its latent dynamics are trained with Lyapunov-guided stability regularization, which promotes stable long-horizon rollouts. We provide theoretical guarantees under the stated structural assumptions. Beyond forecasting PDE evolution, the learned latent dynamics, which are interpretable and linear or linear-quadratic, could bring benefits for downstream tasks such as data assimilation and optimal control. Moreover, our framework is capable of predicting continuous PDE solution fields from sparse measurements of the initial condition. We evaluate SMORE on a range of problems, including wave propagation, the Navier-Stokes equations, and the shallow water equations. Our results show that it improves long-horizon rollout generalization and empirical robustness, and achieves competitive accuracy at comparable parameter budgets relative to competitive baselines including DINo, FNO, CNO, and Transolver.
[AI-374] AO-DA: Towards Autonomous Operation–A Dual-Arm Vision-Language-Action Model for Coordinated Manipulation
链接: https://arxiv.org/abs/2609.33197
作者: Yongsheng Zhao,Han Gao,Baoping Cheng,Jingyao Tang,Dian Zhou,Deng Liang,Ji Ge,Xuanzhang Wen,Lei Zhao,Ye Wang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision-Language-Action (VLA) models provide a unified framework for grounding high-level semantic information into low-level robot actions, enabling scalable robotic manipulation across diverse tasks. However, existing VLA models lack explicit mechanisms to disentangle the states and intents of the two arms, leading to unintended cross-arm interference that degrades task execution success. To address this issue, we propose a symmetric Dual-Arm Expert (DAE) architecture built upon a shared Vision-Language Model (VLM) backbone with decoupled, arm-specific expert towers. Expert selection is carried out through a two-stage dual-arm intent routing scheme, in which experts are routed either by explicit language instructions in the first stage or by implicit visual semantics in the second stage. Moreover, we introduce a lightweight task progress prediction module that leverages cross-attention between the pre-chunk temporal features and semantic representations of proprioceptive and visual observations to accurately estimate frame-wise task completion progress. This module facilitates task progress synchronization to support coordinated scheduling for collaborative multi-robot tasks. Experimental results demonstrate the effectiveness of our model in dual-arm intent routing and the disentanglement of cross-arm interference, and further provide preliminary evidence of emergent skill generalization from single- to dual-arm tasks (as well as the reverse), together with cross-arm motion-domain skill transfer.
[AI-375] Are Benchmarks Reliable? Toward Structural Diagnosis via Sample-Level Capability Boundaries NEURIPS2026
链接: https://arxiv.org/abs/2609.33196
作者: Haiquan Hu,Yuzhu Liang,Weicheng Tang,Yanzeng Li,Yao Shi,Tian Wang
类目: Artificial Intelligence (cs.AI)
备注: Submitted to NeurIPS 2026; Rejected with reviewer scores of 4,4,4
Abstract:Evaluating large language models (LLMs) relies heavily on benchmark scores, yet aggregate metrics can obscure whether benchmark samples reliably support model comparison. We introduce \textbfBSDProbe, a sample-level framework for \emphbenchmark structural diagnosis that estimates capability boundaries from repeated-response trajectories along ordered model axes. BSDProbe summarizes samples by boundary position, boundary width, boundary-signal validity, and order consistency, then aggregates them into benchmark-level structural profiles. Experiments on six benchmarks show that benchmark reliability is axis-conditioned and heterogeneous: GSM8K and MATH exhibit the most stable measurement structures, MMLU and TriviaQA are relatively stable but heterogeneous, while GPQA and PopQA show stronger axis-conditioned risks. These profiles remain consistent across Qwen3, Qwen2.5, and cross-model axes. BSDProbe further selects compact high-value subsets whose model discriminability reaches up to 8.58\times that of the full benchmark. These results suggest that reliable benchmark use requires examining sample-level capability boundaries beyond leaderboard scores.
[AI-376] Unlocking Latent Personalization in LLM s
链接: https://arxiv.org/abs/2609.33182
作者: Wei Chen,Guanghui Zhu,Zhongliang Cai,Yihua Huang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) are increasingly expected to adapt to individual users, yet effective personalization remains challenging when only limited user-specific samples are available. In this work, we take an alternative perspective: pretrained LLMs may already possess latent capacity for personalization, and a few user samples may therefore suffice to guide the model toward user-aligned behavior with minimal user-specific adaptation. From this perspective, we propose LatentPersonal, a framework that formulates personalization as navigation in a shared latent adaptation space. LatentPersonal infers a compact latent representation from a few user samples to guide user-specific model adaptation, regularized with a variational information bottleneck to encourage compact preference representations. We instantiate LatentPersonal with LoRA, leveraging its low-rank parameterization as a natural low-dimensional adaptation space for personalization. By simply inserting a user-specific guidance vector between the shared low-rank factors, the model can navigate toward personalized adaptations through lightweight inference of this compact representation, without updating the shared LoRA parameters. Experiments across multiple personalization datasets demonstrate that LatentPersonal substantially reduces user-specific adaptation overhead while achieving effective personalization from only a few user-specific interactions, with particularly strong performance in the one-shot regime.
[AI-377] What Does a Skill Actually Do? Estimands and Evaluation Validity for Tool and Skill Use in LLM Agents : A Critical Review
链接: https://arxiv.org/abs/2609.33153
作者: Shuyang Zhang
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 30 pages, 1 figure. Critical narrative review
Abstract:Large language model (LLM) agents increasingly draw on external tools and reusable skills selected at run time from libraries that hold thousands of entries. Reports that a retriever, router, or skill library “improves” an agent may refer to retrieval recall, the success change from enabling a library, a paired contrast restricted to triggered tasks, or a gain under an approximate budget constraint. This critical review asks what each design compares and under which assumptions. Building on estimand-based approaches to agent evaluation, we describe tool and skill designs along six axes: treatment contrast, target population, outcome, budget constraint, summary measure, and identification assumptions. Thirteen core empirical studies anchor the evidence synthesis, supplemented by related methodological work and design-level reading of the wider literature. Our contribution is to make explicit distinctions that some source authors already acknowledge through a decomposition of trigger-conditioned pairing, analytic counterexamples, and comparisons across studies. Pairing on the task does not by itself identify an invocation effect; paired gain and regression counts measure protocol-specific discordance rather than the share of tasks whose expected outcomes worsen; and total effects of module deployment answer a different question from budget-constrained efficiency. We compare curated skill provision with retriever replacement, triggered subsets with all-task outcomes, and observed cost reductions with budget-constrained comparisons. A reporting checklist and worked examples connect these distinctions to information that studies can report. The review runs no new experiments; empirical results come from the cited studies, and numerical toy examples are analytical illustrations.
[AI-378] Not Too Hard Not Too Easy: Learning from Intermediate States for LLM Structured Reasoning
链接: https://arxiv.org/abs/2609.33149
作者: Hongbo Chen,Guohua Lu,Ting Dang,Hong Jia
类目: Artificial Intelligence (cs.AI)
备注: 33 pages, including appendices
Abstract:A common principle of effective learning is to practice material that is neither already mastered nor too difficult to permit progress. We ask how to apply this principle to structured reasoning tasks such as Sudoku and maze solving. In these tasks, a model can repeatedly revise an incomplete or incorrect candidate solution until it satisfies the problem’s constraints. The intermediate candidate solutions along this trajectory provide natural training examples: some are already solved, some cannot yet be repaired by the model, and others lie at its current frontier of achievable progress. We therefore investigate whether pretrained language models can learn to revise such states and whether training on states at this frontier improves reasoning more broadly. To achieve this, we couple a pretrained language-model backbone with a recurrent updater that repeatedly revises an explicit solution state, using the same parameters at every update step. We further introduce Frontier-Oriented Curation Using Self-trajectories (FOCUS), which selects training states from trajectories generated by the current model. FOCUS measures how much the model improves each state within a fixed number of recurrent updates and prioritizes states from which it can make substantial progress. With Qwen3-1.7B, FOCUS achieves 64.4% exact solve accuracy on Sudoku-Extreme and 91.1% on Maze-Hard, with similar gains observed across five Qwen and Llama backbones spanning 1.7B to 8B parameters. We further observe zero-shot transfer in the adapted LLM to mathematical reasoning and code execution, even when the recurrent updater is disabled and no downstream fine-tuning is performed.
[AI-379] LiteEvo: Automated Cost-Efficient Harness Evolution for Generalization to Unseen Tasks
链接: https://arxiv.org/abs/2609.33146
作者: Euntae Choi,Sumin Song,Sungjoo Yoo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:An LLM agent is defined by two things: the weights inside its model and the harness of components assembled around it. Harnesses are still handcrafted, and HarnessX, which evolves them automatically, starts each benchmark from a handcrafted harness, reports gains on the tasks it evolved on, and budgets 100 to 175 million meta-agent tokens per benchmark. We propose LiteEvo, a lightweight harness-evolution algorithm whose tool-free meta-agents mine agent trajectories for reusable components, curate them into a versioned library, and compose each round’s harness from it, starting every benchmark from the same neutral harness and never naming the benchmark. Evolving on the graded tasks of five agentic benchmarks with a frozen Qwen3.5-9B, LiteEvo lifts pass@2 by 10.5 to 67.7pp and reaches comparable or higher pass@2 than a reproduction of HarnessX (71.0 against 67.3 on average) at 13.0 lower mean API cost. Harnesses evolved on train tasks keep their gains on unseen test tasks of four benchmarks, and LiteEvo also lifts Claude Code with Sonnet 4.6 by 1.2 to 71.4pp.
[AI-380] Beyond the Training Horizon: Mechanisms and Limits of Length Generalization in Looped Transformers
链接: https://arxiv.org/abs/2609.33144
作者: Jia Liang,Xi Jin,Liangming Pan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 39 pages, 10 figures
Abstract:Looped Transformers can generalize to reasoning chains longer than those encountered during training, but the computations enabling this behavior and limiting its extent remain unclear. We mechanistically compare two looped-Transformer configurations, which we call the Matched-Recurrence Looped Transformer (MR-Loop) and Decoupled-Recurrence Looped Transformer (DR-Loop), reflecting their respective recurrence-training schemes. We evaluate polynomial iteration, finite-state composition, and knowledge-graph traversal using detailed mechanistic analysis. Attention analysis, intermediate-state decoding, and causal interventions reveal distinct mechanisms learned under final-answer supervision. MR-Loop updates an intermediate state at a fixed readout while advancing relation selection through adjacent-token interactions and a transferable progress cue. DR-Loop instead propagates intermediate states across relation positions, forming an advancing computational frontier. However, both mechanisms become unreliable at greater depths: MR-Loop exhibits degradation of its readout state and progress cues, while DR-Loop exhibits declining reliability of state propagation. Limited self-correction allows local errors to persist and compound. Across both models, we uncover a common representational principle: recurrent states encode not only task-relevant content but also its computational status, whether that content remains in a form that can support subsequent computation. Transferable live-consumed and fresh-aged residual directions causally control whether represented information can participate in subsequent computation, including beyond the training horizon. We further show that length generalization need not rely on faithful step-by-step reasoning, as Looped Transformers can exploit task structure without explicitly representing every intermediate state.
[AI-381] On Device Agent ic Operation Caches – Classifier-Centric NL-to-Action Generation
链接: https://arxiv.org/abs/2609.33141
作者: Moghis Fereidouni,Anthony Arnold,Sumit Gulwani,Mark Marron,A.B. Siddique
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Agentic AI is increasingly being embedded in software applications to provide natural language interfaces to features and functionality. In most cases these agents are powered by enterprise (100+ billion parameter) or frontier class large language models that require substantial computational resources run and depend on cloud hosted inference to handle the task of transforming natural language inputs into actionable software operations. This reliance on cloud-hosted inference introduces substantial network latency on top of LLM inference times, creates data privacy concerns, and, given the costs of running these models, can rapidly escalate expenses associated with supporting agentic features. This paper introduces a novel means of converting the NL-to-Action problem from a generative one into a classification-centric formulation via on-device operation caches. These caches allow an agentic system to handle frequently occurring classes of actions completely on-device – reducing latency, enhancing privacy, and lowering operational costs. We show that for a classic NL-to-Formula task, generating Excel Formula in response to user requests, this approach reduces total inference cost by 56% when compared to cloud-only model-routing based inference and, on cache hits, reduces the latency to response latency by 5x. Subjects: Artificial Intelligence (cs.AI); Software Engineering (cs.SE) Cite as: arXiv:2609.33141 [cs.AI] (or arXiv:2609.33141v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.33141 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-382] Ceiling of a Task: When Can a Transformer Succeed Without Its Chain of Thought?
链接: https://arxiv.org/abs/2609.33134
作者: Jiashu He,Jinxuan Fan,Xiao Xiao,Radu Marculescu,Alejandro Ribeiro
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Reasoning models generate long chains of thought before they answer, yet it is debated whether the content of these chains does real computational work or is largely decorative. We study this question by viewing a transformer as a shallow circuit. One forward pass through a fixed number of layers has constant depth, so any procedure that runs the model a constant number of times is a shallow circuit. We call the best accuracy that a shallow circuit can reach on a task the ceiling of the task, and a task is serial if its ceiling lies below one. We prove three results on serial tasks that hold for every transformer, no matter how it was trained. Necessity: replacing the chain by anything that does not depend on its content, such as filler tokens or a restatement of the question, drives the accuracy down to the ceiling, and on a maximally serial task down to chance. Depth: no shallow computation can write the chain of a model whose accuracy exceeds the ceiling, not even approximately. Locality: the answer is one shallow pass away from the finished chain, so all of the serial reasoning happens in the chain. On word problems of finite groups, whose ceilings are known, small transformers trained from scratch, with or without reinforcement learning, attain the predicted numbers: chain-trained models solve every input length and fall to chance when the chain is erased, chainless models collapse to the ceiling as the input length grows, and open-weight reasoning models given the same problem in words return to the baseline without their chain. On MATH-500 and AIME, erasing the chain costs open reasoning models 0.52 to 0.82 accuracy, a sentence shuffle is harmless, and a token shuffle is as harmful as erasing; the same holds for checkpoints trained by GRPO with a correct or a random reward. The ceiling of a task therefore answers when a transformer can succeed without its chain of thought.
[AI-383] Flow-Matching-Based Protein Structure Tokenizer Made Efficient and Easy
链接: https://arxiv.org/abs/2609.33129
作者: Zhe Zhang,Yikai Zhang,Jiangtao Feng,Ya-Qin Zhang,Wei-Ying Ma,Hao Zhou
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:As the bridge between protein modality and discrete modeling, protein structure tokenization still largely relies on heavily engineered training objectives tailored to specific downstream tasks and large training datasets, which hinders its transfer to broader application scenarios. To address this issue, we propose ProFiT, a lightweight flow matching tokenizer. With simple training strategies that encourage healthy codebook utilization, ProFiT can be trained efficiently and naturally learns semantically meaningful representations without any manual semantic alignment, while achieving reconstruction quality and generalization that match or surpass those of substantially larger tokenizers. We conduct extensive evaluations across a wide range of settings and demonstrate that ProFiT is a plug-and-play tokenizer adaptable to diverse downstream tasks. This study further reveals the significant potential of the flow matching tokenizer paradigm. Our code is publicly available at this https URL.
[AI-384] Policy Plasticity Matters in Offline-to-Online Reinforcement Learning: Refitting Offline Policies for Online Adaptation
链接: https://arxiv.org/abs/2609.33127
作者: Yuheng Huang,Yunpeng Qing,Yixiao Chi,Yilun Kong,Changqing Zou
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Offline-to-Online Reinforcement Learning (O2O RL) has emerged as a practical paradigm that pre-trains the policy using static offline datasets and subsequently adapts the policy through online interactions. Existing O2O methods primarily address the transition through value calibration, while generally treating the offline-trained policy as a given initialization. We instead study O2O adaptation from the perspective of network plasticity, asking whether the offline-trained policy remains sufficiently adaptable for online learning. Controlled experiments show that prolonged optimization on static offline data progressively reduces network plasticity even after offline performance has largely saturated, and that lower plasticity is associated with weaker subsequent online improvement. Motivated by these observations, we propose REstoring plasticity via Fresh Initialization and policy Transfer (REFIT), a lightweight model-level method for the O2O transition. Before online fine-tuning, REFIT distills the offline policy into a freshly initialized student while temporarily freezing a random subset of student units, transferring the learned offline behavior to a more plastic policy initialization. Extensive experiments on D4RL and OGBench demonstrate that REFIT consistently achieves higher aggregate performance than existing O2O plug-in methods across both Cal-QL and IQL backbones, while plasticity diagnostics and ablations provide further evidence of restored network plasticity.
[AI-385] Compositional Safety Failures in Harness Evolution: Identification and Runtime Monitoring
链接: https://arxiv.org/abs/2609.33123
作者: Zhixiang Zhang,Zesen Liu,Wai Ip Lai,Hongxu chen,Dongdong She
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Self-evolving agent harnesses continually update persistent components such as memory, prompts, skills, and tools. We call this process harness evolution. However, such evolution could introduce unexpected safety risks. Existing work studies harness misevolution and validates candidate harnesses or attributed individual component updates, leaving safety analysis of cross-component update interactions largely unexamined. To address this gap, we study compositional safety failures in harness evolution, where interactions among individually safe and utility-preserving component updates can produce undesirable or unsafe agent behavior, revealing a safety risk intrinsic to harness evolution. Across three safety-related benchmarks, we identify 43 pairwise and 18 irreducible 3-way compositional safety failures. Conventional solution incurs combinatorial complexity in validating cross-component interactions, leaving the safety checking impractical as the harness evolves. To solve this, we introduced a typed hypergraph that represents component states as nodes and safety-relevant higher-order interactions as hyperedges. When the harness changes, the hypergraph updates only the interaction neighborhood of the changed states rather than reconstructing the global composition space. Building on that, we develop a hypergraph-guided runtime monitoring mechanism. Experiments show that our method effectively mitigates compositional safety risks while preserving task utility and reducing interaction-checking costs, and further reveal an empirical safety-utility-cost trade-off across different safety mechanisms.
[AI-386] D-JEPA: Design-Recoverable JEPA Representation with Swappable Physics Decoders
链接: https://arxiv.org/abs/2609.33110
作者: Nitin Nagesh Kulkarni,Aashwin Anand Mishra,Yin Yu,Peter Lyu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Joint-Embedding Predictive Architectures (JEPAs) provide a framework for learning compact representations without directly reconstructing high-dimensional observations. However, in parameterized physical systems, learned representations can entangle geometry with operating conditions and task-specific physical responses, limiting their reuse across prediction tasks. We introduce D-JEPA (Design-recoverable JEPA), a geometry-centric JEPA that computes a compact representation from geometry alone and reuses it across operating conditions and physical response spaces through lightweight physics-specific decoders. An explicit design-recoverability objective encourages the geometry latent to preserve information about the underlying design variables, enabling the representation to support design analysis and optimization. We further identify a case-level collapse failure mode in which target representations become nearly invariant across distinct geometries despite low reconstruction error, and mitigate it using case-level variation constraints and auxiliary target reconstruction. Across four 3D aerodynamic, hydrodynamic, and structural benchmarks, D-JEPA maintains or improves full-field prediction accuracy while achieving near-perfect linear recoverability of design parameters. The frozen geometry representation can be reused at held-out operating conditions and transferred to a structural response task with fewer trainable parameters. Finally, the representation supports differentiable design optimization, with designs validated using high-fidelity CFD, preserving the predicted ranking of candidate designs. These results demonstrate that separating a reusable geometry representation from physics-specific prediction provides a practical representation for scientific surrogate modeling and design.
[AI-387] VPTwin: Real-Sim-Real Video Prediction for Robotic Manipulation Planning
链接: https://arxiv.org/abs/2609.33104
作者: Zhenghao Xiao,Minting Pan,Nantian He,Dongzhan Zhou,Yunbo Wang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:While action-conditioned video prediction provides an intuitive world model for robotics, purely data-driven predictors often suffer from compounding errors and physically implausible hallucinations in long-horizon rollouts, severely undermining downstream action planning. We propose VPTwin, a Real-Sim-Real video prediction framework that anchors real-world future prediction using real-synchronized simulation twins. For a target manipulation task, a VLM reconstructs an executable digital twin from a real demonstration episode. To accommodate the ill-posed estimation of unobserved physical properties, Isaac Sim simulates multiple forward dynamic rollouts across randomized physical configurations under candidate action trajectories. Using these rollouts as in-context references, VPTwin harmonizes both domains, using simulation dynamics to enforce physical plausibility while capturing unmodeled contact interactions from real video. Furthermore, we establish a predictive planning loop using VPTwin to visually verify VLM-proposed actions and guide reliable real-world execution. Evaluations show substantial reductions in physical hallucinations during video prediction and marked improvements in manipulation planning performance.
[AI-388] How Linear Attention Remembers
链接: https://arxiv.org/abs/2609.33093
作者: Kichang Lee,JaeYeon Park,Songkuk Kim,JeongGil Ko
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Linear attention replaces the growing key–value (KV) cache of standard attention with a fixed-size recurrent state, substantially reducing memory growth with context length. This efficiency, however, changes how past information is stored: many tokens must share and repeatedly update the same memory. We study how this recurrent state functions as a memory system. Using an analytical decomposition together with controlled causal interventions in pretrained GLA and GDN models, we trace how recalled information is written, retained, and later accessed. We find that fact-specific information enters recurrent memory through concentrated, content-dependent writes and is later accessed through concentrated query-time read pathways. Multiple facts can remain selectively accessible within the same state, yet their internal representations exhibit cross-fact causal coupling rather than independent KV-like storage. As memory load increases, both recall and targeted editability degrade, whereas elapsed context alone has a substantially smaller effect within the tested regime. Causal interventions on subsequent writes further show that interference is shaped by their overlap with existing memory. Finally, in hybrid architectures that combine recurrent layers with full attention, the runtime memory directly supporting recall shifts predominantly to the full-attention KV state. Together, these results reveal how fixed-size recurrent memory supports selective recall despite shared storage, while exposing the interference and capacity limits that distinguish it from token-addressable KV memory.
[AI-389] he Model Knows Another Way: Strategy Switching for Effective RLVR Exploration
链接: https://arxiv.org/abs/2609.33085
作者: Jin Cui,Xinyue Long,Boran Zhao,Pengju Ren,Hao Dong
类目: Artificial Intelligence (cs.AI)
备注: 23 pages, 5 figures
Abstract:Reinforcement learning with verifiable rewards (RLVR) is often limited by insufficient exploration: difficult problems can yield uniformly incorrect rollout groups and therefore little learning signal. We show that such failures need not reflect missing capability. Instead, finite sampling often concentrates on a problem-specific dominant reasoning strategy while leaving alternative strategies already supported by the model unexplored. Moreover, the accessibility of these strategies evolves during RL: some are internalized into autonomous behavior, while others become difficult to elicit before being absorbed. Motivated by these observations, we introduce Problem–Strategy Rollout Allocation (PSRA), which treats unguided and strategy-conditioned prompts as competing exploration arms and uses Bayesian sequential allocation to direct a fixed rollout budget toward arms most likely to yield informative, non-saturated groups. A preservation objective keeps useful strategy-conditioned routes accessible while successful guided behaviors are transferred to the unguided policy. Across Qwen2.5 models from 1.5B to 7B and two RL training corpora, PSRA consistently improves reasoning performance, reduces dead saturation, strengthens out-of-distribution transfer, and maintains larger gains under increased inference budgets.
[AI-390] Structure-Mapping-Guided Self-Explanation for Learning Mathematical Procedures
链接: https://arxiv.org/abs/2609.33079
作者: Shinhaeng Lee,Christopher J. MacLellan,Daniel Weitekamp
类目: Artificial Intelligence (cs.AI)
备注: 19 pages. Accepted for oral presentation at the Thirteenth Annual Conference on Advances in Cognitive Systems (ACS 2026)
Abstract:Worked examples are a powerful form of instruction, but learners must infer how the demonstrated steps were produced. A naive simulation of this self-explanation process can generate thousands of numerical explanations that reproduce one observed change without capturing its underlying procedure. We propose structure-mapping-guided self-explanation as a computational account of the cognitive biases that reduce search effort and make this inference tractable. The model represents mathematical expressions as typed relational structures and uses structure mapping to identify corresponding source and target regions. For each changed target value, the corresponding source region serves as an anchor: it guides abductive search toward structurally relevant values and operations before broader alternatives, yielding ordered, executable candidate procedures with inspectable source evidence. Across 70 mathematical transformations containing 120 changed numeric components, the model recovered every intended procedure. It returned the intended procedure before any other computation producing the same target value in 104 subproblems (86.7%), compared with a median of 32 (26.7%) across 100 unguided runs that tested candidate calculations in random order. Our proposed model also tested 93.8% fewer combinations of values and operations than unguided search before reaching the intended procedures. These results provide an efficient, interpretable account of how relational structure can guide procedural learning and a testable hypothesis about human self-explanation from worked examples.
[AI-391] QureRadEmbed: Structuring Radiological Similarity through Attribute and Reasoning Supervision
链接: https://arxiv.org/abs/2609.33075
作者: Janhavi Prabhu,Sahil,Shivam Ashok Shukla,Manoj Tadepalli
类目: Artificial Intelligence (cs.AI)
备注: 39 pages, 10 figures, including appendices with per-tag labeling results. Janhavi Prabhu and Sahil are co-first authors. Manoj Tadepalli is the corresponding author
Abstract:Radiological similarity depends on disease relationships and on fine details such as laterality, lobe, severity, size, and certainty. Broad biomedical similarity can overlook these qualifiers, particularly when several attributes vary together. We introduce QureRadEmbed, a 4B radiology-aware encoder trained with two complementary signals: RadSim supplies deterministic, attribute-decomposed ranking targets, while RadThought aligns reports with hierarchical evidence and reasoning descriptions. A three-stage curriculum combines these signals with report triplets, finding perturbations, and single- and cross-attribute contrasts. The final model achieves 0.996 mean ordering accuracy across ten controlled synthetic attributes and raises Spearman correlation with the designed joint-attribute targets from 0.501 to 0.976. On external findings-to-impression retrieval, Recall@1 reaches 10.5% on Open-I, 12.4% on testing XR, and 42.9% on testing CT, compared with 6.6%, 5.7%, and 31.4% for its backbone. Frozen embeddings support finding extraction with only 100 labeled testing-XR reports (macro-F1 0.481 versus 0.412 for the backbone). Whole-report comparison costs 8.8 seconds per 1,000 pairs in our benchmark, versus 2,755.1 seconds for the generative evaluator GREEN. Sentence-level comparison improves sensitivity to local discrepancies, although generative evaluation remains stronger on several expert-rated and subtle-error tasks. The results support reusable radiology-aware representations for search, structured report indexing, and efficient report comparison.
[AI-392] Zero-Storag e Procedural Neural Synthesis via Boundary Dynamics: Formal Verification in Lean 4 and Bare-Metal Gauntlet Validation
链接: https://arxiv.org/abs/2609.33066
作者: Volkan Dağlı,Zerrin Dağlı,Dağhan Dağlı
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Logic in Computer Science (cs.LO)
备注: 7 pages, 2 tables, 1 listing. Formal verification in Lean 4 (v4.34.1, Mathlib4, 0 sorry). Ancillary files include Lean 4 proofs, Solidity contracts, and bare-metal 40-core gauntlet replication scripts
Abstract:Contemporary neural inference architectures rely on dense floating-point weight matrices stored in high-bandwidth memory (VRAM), incurring severe memory-wall bottlenecks and preventing native execution inside deterministic virtual machines like the Ethereum Virtual Machine (EVM). Verifying termination and arithmetic invariants for recursive dynamical systems over continuous domains is generally undecidable in the Blum-Shub-Smale model. Here, we present the formal verification and bare-metal empirical validation of WERR (Waves Errors) and Phase III Orbital Error Dynamics (OED), a non-tensor decision paradigm that procedurally synthesizes non-linear decision boundaries on demand from a 24-byte coordinate seed \Theta = (c_x, c_y, \textzoom) along the boundary of the Mandelbrot set ( \partial\mathcalM ). By projecting the recurrence z_n+1 = z_n^2 + c onto the modular residue ring \mathbbZ/9\mathbbZ and the fixed-point domain \mathbbQ_16.16 , we establish ten machine-verified theorems in Lean 4 (v4.34.1) with Mathlib4 and zero unproven conjectures (sorry): proving \mathcalI_3 = \0,3,6\ \subset \mathbbZ/9\mathbbZ ideal closure, universal fuel-bounded halting ( \le 9 and \le 12 steps), absence of \mathbbQ_16.16 square overflow below 2^63-1 , non-constant boundary escape sensitivity, and a parametric EVM gas bound ( \le 22,557 \le 24,000 gas). Evaluated on a 40-core Dual Intel Xeon server, the vectorized 36-iteration CPU kernel processes 100,000 decisions in 6.49 s (15,397 decisions/s, 0 Bytes VRAM, 15.15x speedup), while a sigmoidal outlier gate suppresses 100.00% of adversarial spikes while preserving 89.60% of clean baseline signals.
[AI-393] LLM sequential decision making under uncertainty in biochemical domains
链接: https://arxiv.org/abs/2609.33061
作者: Mattias Akke,Soojung Yang,Jurgis Ruža,Sathya Edamadaka,Rafael Gómez-Bombarelli
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) are increasingly used to drive scientific discovery. Understanding how LLMs make decisions from new data and memory of the literature is vital before trusting them to design experiments under tight experimental budgets. However, their decision strategies are invisible in the current performance scores used to evaluate research agents. Here, we benchmark five frontier LLMs in a Bayesian Optimization setting against published statistical baselines on seven combinatorial datasets spanning protein engineering, reaction optimization, molecular design, peptide self-assembly, and catalysis. Performance is paired with direct measurements of model beliefs and actions, enabling highly resolved behavior analysis. A prompt ablation that progressively strips context separates memorization from chemical reasoning and from bare categorical optimization. Prior chemical knowledge helps in expectation, but with high variance and occasionally even harms performance. No configuration tested decisively beats a mean statistical baseline across domains. Belief-movement and Martingale diagnostics, corrected here for a measurement-noise bias that mislabels rational agents as irrational, show that models overreact to incoming data rather than entrenching on their priors in the contexts studied here. Interestingly, while LLM actions are exploitative, models sincerely intend to explore and consistently act on that intent. This failure is a competence gap arising from context-stickiness. Removing in-context history restores exploration, indicating that priors and data must be decoupled to achieve effective LLM-driven discovery.
[AI-394] Large Language Models Substantially Compress Well-Being Inequality but Largely Preserve Its Socioeconomic Structure
链接: https://arxiv.org/abs/2609.33055
作者: Nattavudh Powdthavee
类目: Artificial Intelligence (cs.AI)
备注: 28 pages, 3 figures
Abstract:Research using large language models (LLMs) to generate synthetic populations has repeatedly shown that model outputs compress the diversity of human experience. This has raised doubts about whether LLM-generated data can capture meaningful differences within populations. We show that such compression does not necessarily erase the social structure of human heterogeneity. Using 93,901 respondents from 66 countries and territories in Wave 7 of the World Values Survey, we ask six LLMs to predict respondents’ life satisfaction from demographic, socioeconomic, and attitudinal profiles. All six models substantially understate the overall dispersion of life satisfaction. Yet after normalizing for these differences in scale, they largely reproduce the human income gradient in well-being inequality: lower-income groups remain relatively more heterogeneous than higher-income groups. The pattern is robust to country fixed effects, equal-country weighting, WVS survey weights, and observed demographic composition, and it extends directionally to employment, education, and perceived control. Fidelity is weaker for extreme outcomes and country-specific gradients. These results show that the amount of heterogeneity preserved by an LLM and the way that heterogeneity is distributed across social groups are distinct properties. LLM-generated populations can therefore substantially compress human variation while retaining meaningful information about where that variation is concentrated.
[AI-395] BudgetVerify: Budget-Tiered Verification for Financial QA
链接: https://arxiv.org/abs/2609.33052
作者: Janet Jenq,Hongda Shen
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Financial question answering often requires precise numerical extraction, unit handling, and arithmetic over tables and text, but applying expensive verification uniformly wastes test-time compute. We propose BudgetVerify, a budget-tiered generator-verifier framework that routes each generated answer to one of three verification tiers: no verification, lightweight check-and-revise, or higher-cost solve-first-then-compare verification. The router is trained from offline correctness and token-cost outcomes and, at test time, selects a verification tier using information available before verification, including the question, context statistics, the generated answer, and associated generator metadata. The selected tier either returns the generated answer directly or invokes the corresponding verifier. Across six commercial and open-weight base models, BudgetVerify consistently produces more efficient accuracy-cost Pareto frontiers than fixed verification policies by selectively allocating stronger verification only when it is useful. Although absolute performance varies across models, these efficiency gains and the resulting qualitative frontier shape are consistent across generator models.
[AI-396] Agent Safety From Within: Detecting Harmful Trajectories from LLM Internal States
链接: https://arxiv.org/abs/2609.33039
作者: Difan Jiao,Ashton Anderson
类目: Artificial Intelligence (cs.AI)
备注: 24 pages, 11 figures, 9 tables
Abstract:Language model agents can now perform sophisticated sequences of actions via tools and harnesses, which has increased the scope of the damage they can cause. Guard models, however, are mainly built for content moderation and thus are not well-suited to detecting this agentic risk. To address this, we proceed by first conducting a representational analysis, then use the resulting insights to build a solution. In our analysis, we focus on two types of trajectory-level agentic harms: harmful content, which is expressed directly, and unsafe tool use, which depends on whether an action is consistent with the interaction that produced it. We investigate how open-source guard models represent these two types of harm and find that they are linearly readable inside the model, even though guard models predict no better than chance on pairs that differ only in the called tool’s schema. The two harm types also follow nearly orthogonal internal directions, and neither reliably serves as a proxy for the other. These results motivate reading trajectory safety directly from internal states. We introduce TACIT, a readout of a frozen backbone’s internal states that decodes no tokens. Trained on six trajectory-safety benchmarks, a linear probe raises mean macro-F1 from 62.3 for the strongest open guard to 80.7, and refined readouts reach 86.2. With each benchmark held out of training entirely, the refined readouts still lead the strongest guard (65.7 vs. 61.1). With the same backbone, training data and test split, the frozen readout is on par with full safety fine-tuning, and it improves the fine-tuned model further when applied on top. The probe trains about one millionth as many parameters as full fine-tuning in about a sixth of the time, and TACIT has the lowest latency of the guards we evaluate.
[AI-397] Łukasiewicz Neural Networks Extended: Residual Architectures and Crystallization Strategies for Interpretable Rule Extraction
链接: https://arxiv.org/abs/2609.33028
作者: Carlos Leandro
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 21 pages
Abstract:A feed-forward neural network whose weights are integers and whose activation is the truncated identity implements, neuron by neuron, the connectives of Łukasiewicz many-valued logic. This exact correspondence — established theoretically by Castro and Trillas and developed into a training algorithm by Leandro — enables \emphsymbolic knowledge extraction: training produces not a black-box model but a logical formula. Two obstacles have limited the approach to shallow architectures and small datasets: crystallization (forcing weights to integers) succeeds only probabilistically under the original Levenberg–Marquardt training scheme, and the theoretical guarantees break down as networks grow deeper. This paper addresses both obstacles. First, we prove that \emphresidual connections (skip connections of the kind used in ResNets) extend Łukasiewicz neural networks to arbitrary depth while preserving the symbolic correspondence \emphat merge neurons by construction: merge neurons in a Łukasiewicz residual block automatically satisfy the neuron-classification proposition, regardless of the inner layer weights; inner-layer neurons are trained toward representability by the crystallization strategy. Second, we analyse three crystallization strategies — Levenberg–Marquardt (corrected), straight-through estimation (STE), and proximal regularization — characterizing their theoretical guarantees, failure modes, and interpretability trade-offs. Comments: 21 pages Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.33028 [cs.LG] (or arXiv:2609.33028v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.33028 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-398] SRE-Marathon: A Continuous Change-Driven Benchmark for Autonomous Site Reliability Agents
链接: https://arxiv.org/abs/2609.33023
作者: Yifang Tian,Yingjian Bai,Yifeng He,Zichun Chong,Yuanchen Gao,Yiran Li,Hans-Arno Jacobsen
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Benchmarks for site reliability engineering (SRE) agents are typically episodic: one fault is injected, the agent receives an incident task, and its response is scored. Production operation is not. Incidents surface through noisy alerts, overlap in time, and often originate from code or configuration changes. We present SRE-Marathon, a benchmark for long-horizon, continuous SRE operation. An agent is invoked at a fixed cadence with cumulative alert history and a persistent workspace while operating a live two-zone Kubernetes deployment as a fault orchestrator injects overlapping faults according to a seeded, production-calibrated schedule. Curated code and configuration changes deployed through the same build pipeline available for repair. Each run is recorded into a sealed bundle and scored offline: Marathon-Score credits each injected fault for ordered progress through correlation, localization, and repair, with all metrics computed deterministically from recorded system evidence. Across three applications and about sixty faults per run, the best of 10 methods reaches only 41.3 out of 100. Agents often correlate and localize faults, but almost never complete repairs while the faults remain active.
[AI-399] Relative Generalization Invariance of LLM Pretraining
链接: https://arxiv.org/abs/2609.33016
作者: Fengzhuo Zhang,Shuche Wang,Shenggui Li,Tianyu Ruan,Jianliang He,Ivor Tsang,Tianyu Pang,Chao Du,Tianwei Zhang,Zhuoran Yang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Model (LLM) pretraining performance is jointly shaped by three components of the training triplet: the optimizer, model architecture, and training data stream. However, how these components influence performance in distinct ways remains unclear. We take a first step toward isolating their effects by studying relative generalization. We introduce Relative Generalization Invariance (RGI), the invariance of the validation-loss difference between any two tokens across models. We show that RGI approximately holds across a wide range of optimizers and moderate architectural variations, suggesting that these choices induce an approximately uniform shift in token-wise losses. In contrast, changing the training data stream can substantially alter relative generalization. We further show that RGI cannot be explained by the neural tangent kernel or mean-field regimes alone and prove that it can emerge in an overparameterized quadratic model. Overall, our work identifies RGI as a new phenomenon in LLM pretraining that helps distinguish the effects of optimizers and architectures from those of training data.
[AI-400] he Epistemics of Agent Memory: Measuring and Governing the Consolidation Decision in Long-Horizon LLM Agents
链接: https://arxiv.org/abs/2609.33013
作者: Sasank Annapureddy,Anjaneya Prasad Thamatani
类目: Artificial Intelligence (cs.AI)
备注: 13 pages, 4 tables, no figures. A four-phase research program. Experiments executed and adversarially verified by PRIMA ( arXiv:2605.24775 ); per-phase result files and preregistrations available to reviewers on request; certain mechanism internals held under controlled release (see Section 7)
Abstract:Long-horizon LLM agents must convert accumulated experience into durable memory, deciding what to keep, compress, abstract into reusable skills and rules, or forget. We report a four-phase research program on this consolidation problem whose central finding is a shift in what is measured: from how much an agent remembers, to whether its consolidation decisions are any good, to whether those decisions can be trusted. Phase 1 learns episodic boundaries from agent traces by downstream utility; an honest near-miss (oracle correlation 0.691 vs a 0.70 bar) whose lasting output is a three-gate anti-leakage protocol. Phase 2 learns when to promote experience and to which abstraction level under a token budget, achieving a verified +22.7% task-success improvement with 7x compression, but exposing a degenerate-forgetting failure and a distribution-shift failure mode we name lambda-prevalence coupling. Phase 3 introduces ConsolidationBench, an oracle-by-construction benchmark that scores consolidation decisions against a known optimum on three non-circular axes; production retrieval systems retain information yet score zero on cross-level transfer. Phase 4 introduces governed consolidation: the decision wrapped in poison-resistance, reversibility, and auditability guarantees with a quality gate. Governance is statistically distinct from the quality score ( r^2 = 0.43 ; partial r = 0.27 ; identical-quality policies differ threefold in governance), so the contribution survives independently of the metric’s external validity. On that question we report a resolved negative: after a graded-reuse redesign removed a structural ceiling, a two-benchmark study with 2,532 real answer cells finds the quality score does not predict real transfer accuracy (pooled Spearman \rho = -0.24 , n = 12, CI spanning zero). An adversarial self-critique pass cleared the final claim set with zero surviving overclaims. Comments: 13 pages, 4 tables, no figures. A four-phase research program. Experiments executed and adversarially verified by PRIMA (arXiv:2605.24775); per-phase result files and preregistrations available to reviewers on request; certain mechanism internals held under controlled release (see Section 7) Subjects: Artificial Intelligence (cs.AI) MSC classes: 68T05, 68T42, 68T30, 68T37 ACMclasses: I.2.6; I.2.11; H.3.3; I.2.8 Cite as: arXiv:2609.33013 [cs.AI] (or arXiv:2609.33013v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.33013 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-401] When Pair Count Is Not the Sample Size: What All-Pairs Agent Comparisons Estimate NEURIPS2026
链接: https://arxiv.org/abs/2609.33012
作者: Wei-Jung Huang
类目: Artificial Intelligence (cs.AI)
备注: Accepted at the NeurIPS 2026 Workshop TAE (Trust-AI-Eval): Can We Trust AI Evaluation?
Abstract:When an agent benchmark compares every pair of leaderboard entries, the number of comparisons can look much larger than the independent evidence behind them: A versus B and A versus C both reuse A. Whether this reuse affects inference depends on what the analysis is meant to describe. If the board and its outcomes are fixed, the all-pairs mean is an exact summary of those entries, and any interval must come from another declared source of randomness. If the entries are instead treated as iid draws from a population of future configurations and the pair rule is regular and nondegenerate, the same mean is an order-two U-statistic whose first-order uncertainty depends on the number of configurations, not the number of pairs. We use near ties as the running example, but the distinction extends to other symmetric pair summaries when their regularity conditions hold. We examine both interpretations using a fixed SWE-bench Verified snapshot and an exact binary model with known truth. On SWE-bench, intervals that accounted for shared configurations were more than twice as wide as a pair-iid reference that treated the pairs as independent. In the exact model, pair-iid coverage fell far below the nominal level when edges shared endpoints but remained near nominal for matched independent edges. Results on two other fixed leaderboards show that exact summaries also depend on which pairs are included and how they are weighted. An all-pairs analysis must therefore state what is fixed, what is sampled, and how it handles shared entries and pair aggregation.
[AI-402] Model-Aware Data Selection from In-and-Out Information Interplay
链接: https://arxiv.org/abs/2609.33010
作者: Yifan Wang,Xiaomin Li,Yuexing Hao,Dongwon Jung,Hemanth Neelgund Ramesh,Ananth Grama,Varun Chandrasekaran,Yu Hu,Andrzej Banburski-Fahey,Jaron Lanier
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLMs are effective representations that assimilate vast amounts of knowledge during pretraining, but post-training is necessary for models to reliably access this knowledge and “know what they know.” We observe an interesting rank equilibrium between knowledge stored in the weights and the data stream passing through the model. Across all model layers, we find that the hidden states (data stream) follow a U-shaped pattern, showing substantial compression in early layers and a steep rise during the late-layer decoding phase. In contrast, the weight rank follows an inverted U-shaped pattern, with very low rank in the early and late layers and high rank in the middle. We interpret this as an in-and-out information interplay: intermediate activations do not need to carry content that the weights can supply later, so they primarily preserve what the weights cannot provide. Motivated by this observation, we propose a model-aware data selection method, CAP (Counterfactual Assimilation Profile), which can determine whether a data candidate contains information accessible to the current model by utilizing the divergence gap in early- and late-layer representations between model-generated and reference responses. Across math, code, and science domains, CAP delivers 35.4% greater average improvement over the base model than the strongest baseline under different selection budgets. With only 10% of the data pool, CAP surpasses or matches full-pool training on math and science. We further show that CAP transfers to multimodal data selection and is robust to response horizon and noise. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.33010 [cs.AI] (or arXiv:2609.33010v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.33010 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-403] CAPEX: Efficiently Distilling Foundation Model Behavior into Deployable Robot Policies through Experience-Adaptive Reasoning
链接: https://arxiv.org/abs/2609.33007
作者: Shivam Aarya,Zhang Xi-Jia,Chengyue Huang,Junhyun Kim,Huishu Xue,Hrishit Leen,Roman Yakunin,Animesh Garg,Zsolt Kira
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Robot learning has largely relied on human-teleoperated demonstrations to acquire effective learnable behaviors. However, human-operated data collection processes can be unintuitive, difficult to scale, and inherently asynchronous. We explore an alternative: distilling physical behavior from general-purpose multimodal foundation models into deployable robot policies by using the foundation model itself as an autonomous demonstrator. While sufficiently capable models can generate successful zero-shot manipulation trajectories, repeatedly invoking them during physical execution is slow and expensive, limiting their utility as scalable data generators. As a solution, we introduce CAPEX, an experience-conditioned demonstration collection framework that uses execution experience from previous attempts to adapt how frequently the foundation model must observe, reason, and replan. We evaluate across RoboCasa tasks and on physical Franka and bimanual YAM-arm platforms, measuring task success, model calls, token usage, collection time, and cost. We further train Diffusion Policy and ACT on matched sets of human-teleoperated and foundation-model-generated demonstrations to evaluate the downstream learning value of autonomously collected data. We find that CAPEX increases the number of successful demonstrations by 4.3x while reducing the cost per successful demonstration by 80%. Policies trained on CAPEX-generated data approach the performance of those trained on matched human demonstrations; with longer training, this gap largely closes for policies trained from scratch. These results suggest that foundation models can serve as scalable sources of reusable robot experience. Project page: this https URL
[AI-404] X-Tree: Tokenizing Reusable Experience for Efficient Agent Generalization
链接: https://arxiv.org/abs/2609.32993
作者: Sitao Cheng,Xunjian Yin,Zhiyuan Sun,Yuxuan Li,Ruiwen Zhou,Xiangru Jian,Victor Zhong
类目: Artificial Intelligence (cs.AI)
备注: Project in Progress. Homepage: this https URL . Code: this https URL
Abstract:Multi-step agents are trained on flat action streams: SFT and RLVR weight every token uniformly and ignore the sub-procedures that recur across tasks, the hierarchy that lets humans plan top-down from reusable routines. This structure sits unused, and flat training uses each scarce trajectory less fully than its content allows. Recent agents do use that structure, but only as LLM-written skills in context, never in the weights, so their gains do not generalize beyond retrieval. We instead recover this hierarchy from the data itself and train on it, with no LLM calls. Following text tokenizers, which build a vocabulary by counting alone, we score action spans by reusability and merge canonicalized actions into a reusable eXperience tree (X-Tree). Each X-Tree node captures how a frequent and success-bearing skill is composed from sub-skills, guiding efficient generalization. We integrate X-Tree into three training settings: offline RL, with each node as a training instance; online RLVR, with an adaptive skill bonus; and on-policy self-distillation, with X-Tree as the self-teacher’s privileged context. Across WebArena, ScienceWorld, and WebShop at three model scales, X-Tree improves over standard recipes at matched data and budget by up to 4.5% SR on WebArena, 5.8% SR on ScienceWorld and 4.1% success on WebShop. Matched analyses attribute the gains to the X-Tree structure and the three integrations.
[AI-405] Certified Long-Horizon Code Agent Evolution via Validation-Gated Skill Optimization
链接: https://arxiv.org/abs/2609.32990
作者: Yifan Wang,Hao Cheng,Xiaomin Li,Yuexing Hao,Hemanth Neelgund Ramesh,Dongwon Jung,Hao Tang,Keru Wang,Chenliang Zhou,Qianhui Wu,Wenlin Yao,Ananth Grama,Andrzej Banburski-Fahey,Baolin Peng,Jaron Lanier,Jianfeng Gao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Long horizon agent self-evolution without model weight updates is essential for enabling deployed agents to accumulate reusable skills and improve over time. Prior self-evolution work has focused primarily on short-horizon tasks, while repository-level software engineering remains unexplored despite being an ideal testbed for long-horizon adaptation. In this setting, agents are required to solve streams of sequential tasks, navigate complex dependencies with evolving repositories and persistently store and reuse experience. Text-based skill optimization offers an efficient, non-parametric approach for such adaptation. However, existing methods often suffer from unstable updates, performance drawdown, and agent collapse over extended deployments. In this paper, we formalize the concept of in-context self-evolution and introduce VALVE, a validated-gated framework for long-horizon skill optimization. We establish finite convergence, provide theoretical guarantees for future-task gain and drawdown, and derive the validation and evaluation holdout sizes required for a prescribed tolerance, with leading-order scaling Empirically, our pipeline, VALVE achieves stable self-improvement over evolution horizon spanning more than 1,000 SWE tasks, with average final and peak gains of 14.9 and 16.5 points across three frontier models (GPT-5.5, Claude-4.6 and MiniMax-M2.7). The validation gate reduces average drawdown by 75% and produces an 11x more compact skill bank than ungated evolution. We further present extensive ablations identifying the design choices most critical to long-horizon skill evolution. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.32990 [cs.AI] (or arXiv:2609.32990v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.32990 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-406] Relic: From Multi-Agent Collaboration to Persistent Organizational Capability
链接: https://arxiv.org/abs/2609.32965
作者: Hongyi Du,Tianyi Zhang,Weijia Zhang,Yi Yang,Haofei Yu,Kunlun Zhu,Tianxiang Dai,Shang Jiang,Zhelun Gao,Jiaxin Pei,Shang Zhu,Jiaxuan You
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 83 pages, 8 figures. Preprint
Abstract:Multiple agents may often conflict in an organization: for example, one coding agent changes an interface in a repository, but another continues to develop on the old version where existing tests become stale. A conversation can resolve the episode, but when the participants change, what makes the lesson continue to govern the team? We introduce Relic, which turns recurring collaboration failures into organization-owned, executable protocols. Members reflect on visible work, propose rules, and govern their adoption. Adopted protocols bind triggers, responsibilities, required evidence, and execution consequences to the runtime, while remaining open to revision and retirement. In one traced case, repeated integration friction produces an interface-review rule that governs later pull requests and is revised as work continues. Across 360 controlled runs over ten software workloads and three models, Relic raises complete-contract delivery from 14.06% to 19.76% (+5.71 percentage points) over a matched structured team without the protocol lifecycle, improving all four verified production endpoints in every model stratum. Under fresh-member transfer, behavioral correctness is 25.4% with no inherited protocol, 34.6% with the same rules provided as readable text, and 41.2% with executable bindings, a +6.5-point advantage over text alone. On the full CooperBench benchmark, after excluding broken benchmark pairs, Relic achieves 367/477 (76.9%), establishing the best reported result among peer-structured systems. On the fixed 48-pair same-model subset, Relic also exceeds Solo (29/48 vs. 26/48), reversing the coordination loss exhibited by the official peer baseline. Together, these results show how collaboration experience can become persistent organizational state that remains useful beyond the members who created it.
[AI-407] he Commit-Abstain Circuit: Why Language Models Hallucinate Instead of Abstaining NEURIPS2026
链接: https://arxiv.org/abs/2609.32964
作者: Vy Nguyen,Ziqi Xu,Jeffrey Chan,Estrid He,Feng Xia,Renqiang Luo,Erik Cambria,Xiuzhen Zhang
类目: Artificial Intelligence (cs.AI)
备注: Accepted to NeuRIPS 2026 Main Conference
Abstract:Language models (LMs) often hallucinate by committing to confident answers rather than abstaining, even when they do not have enough information to answer reliably. A large body of existing work mitigates hallucination through detection or abstention mechanisms, but leaves open how models internally arrive at the decision to commit or abstain in the first place. We study this decision through mechanistic analysis, framing hallucination as unsupported commitment: the model commits despite exhibiting signals of unanswerability. Using causal gating, we identify a Commit-Abstain Circuit (CAC), a sparse, causally localised subset of attention heads and MLP sublayers underlying this decision. Across ten LMs (3B-14B) from five families and three benchmarks, the CAC exhibits a recurring accumulate-yet-undercorrect pattern: commitment-promoting components build up commitment in earlier layers, while abstention-promoting components act later as corrective signals that are often insufficient to overturn the accumulated commitment. Building on this finding, a lightweight policy trained on CAC activations improves decision accuracy by 12.2 points over the model’s intrinsic commit-abstain margin, reduces false abstentions by 2.5 times, transfers to unseen benchmarks, and extends to larger models (27B-35B). The CAC is both diagnostic, clarifying how models overcommit, and practical, enabling improved abstention decisions.
[AI-408] Beyond Token Savings: A Systematic Study of Context Compression in LLM Agents
链接: https://arxiv.org/abs/2609.32961
作者: Ritul Satish,Prasoon Sinha,Akiho Kawada,Neeraja J. Yadwadkar
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:As LLM agents tackle longer tasks, they increasingly compress growing histories of reasoning, actions, and tool outputs. Compression can reduce token use, but it also changes the information available for later decisions. Existing agentic harnesses bundle decisions about what to compress, when to compress, and how much to remove into fixed policies. A systematic characterization is needed to disentangle these decisions and reveal how each affects task success and execution cost. We systematically vary these decisions across three open-weight models on SWE-bench Verified and Terminal-Bench 1.0. Across nearly 35,000 agent runs, we measure task success, token use, end-to-end latency, and estimated cost. We find that fewer tokens need not mean faster or cheaper execution: on Terminal-Bench with Qwen, policies using roughly one-third as many tokens can take 20-80% longer than the uncompressed agent. Policies with similar overall success can solve different tasks, while the same policy can perform quite differently across models. Our results motivate evaluating compression by its effects on agent execution and tailoring policies to the task, model, and workload.
[AI-409] Environmental Impact of Generative and Agent ic AI: An in-Depth Analysis and Green Solutions
链接: https://arxiv.org/abs/2609.32960
作者: Abderaouf Bahi,Amel Ourici,Ibtissem Gasmi
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注:
Abstract:The proliferation of generative and agentic artificial intelligence (AI) systems has introduced computational demands whose environmental consequences are substantial yet underexamined. This paper examines the environmental footprint of modern AI systems across energy consumption, carbon emissions, water usage, and electronic waste over the full lifecycle of large language models, multimodal foundation models, and agentic workflows, from hardware fabrication and training through fine-tuning and inference to end-of-life disposal. This work provides a conceptual analysis, utilizing order-of-magnitude estimations based on published data, without conducting original physical measurements. We contribute a lifecycle taxonomy that crosses lifecycle phases with five impact dimensions and emission scopes; a Sustainability Assessment Framework for AI Systems (SAFIA) comprising nine indicators; a comparative analysis of traditional, generative, and agentic AI; seven open challenges; policy recommendations for regulators, cloud providers, hardware manufacturers, and AI developers; and a research roadmap to 2035. The analysis indicates that inference can rival or exceed training energy over a deployment lifetime, that agentic workflows can multiply the energy of equivalent single-pass inference by one to several orders of magnitude depending on the number of model and tool calls, that indirect water use and embodied carbon are systematically underreported, and that measurement tools, disclosure practices, and regulation have not kept pace with agentic deployment. These findings call for agent-aware energy measurement, lifecycle-based carbon and water accounting, and mandatory disclosure for large-scale AI training and deployment.
[AI-410] Efficient Message Passing for Partial Differential Equation Priors
链接: https://arxiv.org/abs/2609.32956
作者: Anna Kazachkova,Leonhard Hennicke,Rainer Schlosser,Ralf Herbrich
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:Prior information for real-world physical quantities is most elegantly expressed via partial differential equations (PDEs). In this paper, we propose a novel way to solve PDEs using probabilistic inference on a factor graph. In general, factor graphs provide a natural way to encode prior knowledge into a model as explicit factors; here, this knowledge is provided by a governing PDE, which narrows the solution space, while observed data further shape the posterior over the parameters. The approximate parameter posterior is inferred using message passing based on moment matching, without posterior sampling or global gradient-based optimization. We demonstrate our approach on the first-order advection and the second-order semi-linear Fisher-KPP equations, where it achieves predictive accuracy comparable to a standard baseline while providing structured predictive uncertainty. Moreover, the inferred posterior marginal means and uncertainty structure match more closely those obtained using Hamiltonian Monte Carlo than the evaluated mean-field variational inference baseline, while requiring up to 10x less training time in our experiments, with inference speed comparable to variational inference.
[AI-411] winS-GCN: Spectral conjugate for Spectral Graph Convolutional Networks
链接: https://arxiv.org/abs/2609.32940
作者: Chun Hei Michael Chan,Flavia Petruso,Dimitri Van De Ville
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Graph convolutional networks propagate information by repeated local aggregation through a graph shift operator; i.e., a K -layer network reaches K hops neighborhood. On the one hand, such spreading can lead to oversmoothing. On the other hand, long-range dependencies demand the depth. Transporting information on long distances and without attenuation requires the shift to distinguish a direction of flow, which a symmetric operator cannot perform but a directed one can fulfill. A natural way to extract pure-directionality is to take the skew-symmetric part of the shift operator through the Cartesian split, which, however, generally does not commute with the shift itself, meaning that the filters built on it are not shift-invariant. We instead use the spectral conjugate; i.e., the image of the operator under \tau:z\mapsto \barz , which commutes with the shift and splits it into a dissipative and a non-dissipative part. Two filter families follow: a sum filter, whose non-dissipative component transports signal without energy loss, and a ratio filter, ratio in the pair of components rather than polynomial in the shift. Both arise from non-holomorphic kernels, placing them outside the holomorphic class underlying classical spectral convolution. On the directed cycle, the ratio filter becomes an IIR filter with global impulse response, for which we prove a long-range reach gap against every degree- K polynomial filter. Chebyshev reparameterization gives stable vertex-domain layers with real coefficients, yielding TwinS-GCN, which solves graph transfer tasks at reduced depth and is competitive with state-of-the-art graph convolutional networks on node classification benchmarks.
[AI-412] he Geometry of Logic: Stratification Induces Semantic Structure and Robust Reasoning
链接: https://arxiv.org/abs/2609.32927
作者: Cristina V. Lopes,Yuangang Li,Justin Tian Jin Chen,Alberto Krone-Martins,Iris Ma,Md Rakib Hossain Misu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Transformer-based language models perform well on symbolic tasks, yet it remains unclear whether they learn generalizable rules or rely on statistical shortcuts. Mechanistic studies link algorithmic behavior to structured internal representations, motivating the hypothesis that robust reasoning benefits from separating values from the types that control their manipulation. Can making this separation an architectural primitive improve the learnability and generalization of logical mechanisms? We introduce \textbfSTRAT (\textbfSTratified \textbfRegisters \textbfAnd \textbfTypes), which partitions the residual stream into orthogonal Data and Type subspaces and uses Type-based attention and gating to govern Data transformations. Controlled arithmetic ablations identify three failure modes associated with data-control interference: the Linear Trap, Gradient Wall, and Open Gate Trap. Mechanistic analysis reveals interpretable logical structure, and in arithmetic, STRAT reduces median OOD error 35-fold relative to a Transformer baseline. On each of 11 datasets spanning 10 tasks, STRAT outperforms the Transformer baseline in mean accuracy, by 26 percentage points on average, with both models trained from 10 base examples per dataset using identical task-specific augmentation where applicable. Under distribution shift, STRAT’s mean accuracy drops by only 2.39 percentage points, compared with 11.75 for the Transformer.
[AI-413] Diagnosing Sampled LLM Reasoning in Formal Geometry: Coverag e Realization and Validity Evidence
链接: https://arxiv.org/abs/2609.32924
作者: Xiao Yue,Guangzhi Qu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Repeated sampling can reveal a correct numerical answer without yielding either a reliable system output or a supported derivation. We present Coverage, Realization, and Validity Evidence (CRV), an evaluation protocol for sampled large language model (LLM) reasoning over formal geometry states. Coverage is answer availability, realization is readout accuracy on the frozen candidate pool, and validity evidence is a label-blinded critic judgment of derivational support rather than a proof certificate. CRV freezes each candidate pool before comparing readouts and analyzes covered failures by correct-answer multiplicity and within-problem discrimination. On HardShift441, a 441-problem set for which a reference solver leaves 406 problems unsolved, a LoRA-adapted Qwen2.5-7B generator obtains 24.2% average single-sample accuracy and 68.9% pass@16, whereas verifier-weighted self-consistency (WSC) reaches 38.0%. Readout accuracy is particularly low when the correct answer occurs only once or twice in the pool. In a separate constructed audit of 195 covered problems, the critic labels 12 correct-answer representatives as supported, 181 as refuted, and two as uncertain. These results show that coverage, realization, and validity evidence from the critic are distinct quantities and should be reported separately.
[AI-414] RACE: Learning to Self-Calibrate Wireless Digital Twins from ISAC Measurements
链接: https://arxiv.org/abs/2609.32923
作者: Saad Masrur,Saeed R. Khosravirad,Ismail Guvenc
类目: Artificial Intelligence (cs.AI); Signal Processing (eess.SP)
备注: Under Review
Abstract:Wireless digital twins (DTs) rely on 3D environment models to predict radio propagation and support wireless-network decisions, yet these models are often initialized from imperfect 3D maps. Errors in building position, height, footprint, and orientation can therefore cause a high-fidelity propagation engine to simulate the wrong physical environment. In this paper, we study how a deployed wireless network can repair an existing DT using its own radio frequency (RF) measurements. In particular, we introduce Twin Residual Alignment and Calibration Engine (TRACE), a physics-grounded learning-based self-calibration framework that treats twin maintenance as residual alignment between the physical world and the current DT. Using the same sensing configuration as the physical measurements, TRACE ray-traces the current DT, coherently backprojects the measured and simulated RF onto a common world grid, and extracts the same local region around each building’s current DT position. A multi-view corrector then fuses evidence across sensing nodes and neighboring buildings to predict a gated six-parameter correction per building, without relying on absolute layout or sensor ordering, and supports iterative correction through re-rendering. On 5,400 held-out samples from unseen simulated scenes at 28 GHz, TRACE reduces 3D position RMSE from 2.202 m to 0.302 m and yaw RMSE from 4.978° to 0.894°, outperforming ViT and U-Net baselines under changes in layout, building count, sensing-node count, and SNR. On measured 28 GHz RF data from the NIST outdoor courtyard, a model trained only on synthetic RF reduces mean planar wall-position error from 1.00 m to 7.8 cm, without measured-data fine-tuning or geometric labels. These results show that the discrepancy between measured and twin-rendered RF can serve as a learning signal for repairing a wireless DT.
[AI-415] Precision As You Need: Stochastic Computing Is a Dense Adaptive Quantizer NEURIPS2026
链接: https://arxiv.org/abs/2609.32922
作者: Haoran Jin,Kangqi Zhang,Jirong Yang,Barry Lyu,Qiuyi Ding,Ruijie Gao,Nathan Bleier
类目: Artificial Intelligence (cs.AI)
备注: NeurIPS 2026
Abstract:Matrix multiplications dominate the inference cost of modern transformer-based vision models, yet existing efficiency techniques such as post-training quantization and mixed-precision inference are largely limited to the small set of fixed-width formats (INT4, INT8, BF16, and FP16) supported by conventional accelerators. We revisit stochastic computing (SC) as a way to lift this constraint: viewed as a dense adaptive quantizer, SC controls precision by bit-stream length L rather than a fixed datapath, while each multiplication reduces to a single AND/XNOR gate. We build a GPU library that emulates SC matrix multiplication at scale, exposes stream lengths as first-class kernel arguments, and evaluates SC end-to-end on image classification, object detection and instance segmentation, class-conditional image generation, and visual world-model planning. On top of this substrate, we develop a dynamic per-row mixed-precision policy that assigns stream length per token or group at matched average budget, requires no retraining, and uses the same SC hardware across schedules. Across tasks, SC remains competitive with fixed-format INT quantization at matched bit budgets, while per-row mixed precision helps maintain accuracy at lower average stream lengths. These results provide software-level feasibility evidence that SC can serve as a dense-precision substrate for fine-grained mixed-precision inference on modern vision transformers.
[AI-416] Agent Tell: Behavioural Side-Channel Leakage in Browser-Use Agents
链接: https://arxiv.org/abs/2609.32915
作者: Asif Shahriar,Md Nafiu Rahman,Sadif Ahmed,Farig Sadeque,Md Rizwan Parvez
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Browser-use agents often carry information in their context as they move between websites. While it may be necessary for task completion, it also creates a privacy risk, especially when the information contains a private fact regarding the user. For example, an agent may learn a user’s affiliation after reading a membership record. If it later selects a registration option specific to that affiliation on another website instead of a general option, the information gets leaked. In this work, we define and study behavioural side-channel leakage in browser-use agents, where an agent’s actions inadvertently reveal private information (secret) retained from a prior website, despite an explicit instruction not to disclose it. We introduce AgentTell, a benchmark of 20 scenarios and 100 tasks in which an agent acquires a secret on one website and then completes a task on another website that offers secret-specific actions alongside a general action that reveals nothing. Our evaluation across 9,760 sessions on six backbones shows that agents carrying a secret reveal it through their actions in 61.1% of sessions. Even when agents explicitly state in memory that the secret must not be shared, they still reveal it in 56.7% of those sessions. Moreover, in 34.5% of leaking sessions, their final responses falsely assure users that the secret was not disclosed. These findings show that agents often fail to recognize side-channel leakage as a privacy risk.
[AI-417] Allspark: Weak to Strong Transfer via Alternating Chain of Thought
链接: https://arxiv.org/abs/2609.32913
作者: Kaizhao Liang,Junxiong Wang,Chen Liang,Zhendong Wang,Qiang Liu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Recent progress in frontier models has renewed interest in large-scale reinforcement learning (RL), but the cost of generating large-model rollouts makes even testing RL recipes expensive. We ask whether reasoning improvements learned by a small, weak model can benefit a larger, stronger model without using the strong model’s rollouts during training. We introduce Allspark, a training and inference framework for weak-to-strong transfer through alternating chains of thought. A weak teacher is trained alongside a frozen copy of the same model; the two alternate reasoning segments, and the frozen model produces the final answer. At inference time, a stronger student replaces the frozen training partner, while both models remain fixed. Because they communicate through text, the teacher can steer students from different model families and with different tokenizers. We study Allspark at two scales: controlled Qwen experiments across math and reasoning, and larger-scale Inkling experiments on ARC-AGI-2. The Inkling experiments show accuracy gains in within-family and cross-family settings, including transfer to Kimi and Nemotron, with benefits that vary across inference settings. These findings motivate reusing a trained weak teacher across strong students and examining the resulting accuracy–token tradeoff.
[AI-418] Progression- vs Automata-based Anticipatory Monitoring of LTL over Finite Traces (Extended Version)
链接: https://arxiv.org/abs/2609.32912
作者: Sarah Winkler,Toryn Klassen,Sheila McIlraith,Marco Montali
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI)
备注:
Abstract:When safety-critical systems are developed from a known internal specification, their correctness can be established by model checking. In the frequent case where such a specification is unknown or inaccessible, runtime verification presents an attractive alternative, e.g., to ascertain that autonomous and agentic systems as well as business processes satisfy desirable properties and/or comply with safety requirements. In this paper we study anticipatory monitoring, an advanced form of runtime verification, where the monitoring state is determined by both the trace prefix seen so far, and all its possible finite-length, future continuations. We focus on monitoring linear-time properties that may involve arithmetic constraints. Automata-based approaches, the de-facto standard in this setting, are notorious for their computational complexity. We propose an alternative approach based on progression and LTLf satisfiability checking, for both propositional and arithmetic settings. We experimentally compare the automata- and progression-based approaches, and a third method that combines the two. Our experiments suggest that the progression-based approach often succeeds in producing a verdict when the automata constructions do not terminate, especially for the arithmetic setting. For the propositional setting, the combined technique provides a good tradeoff.
[AI-419] Constraints Are Graphs Not Chains: Exact Decoding for Diffusion Language Models
链接: https://arxiv.org/abs/2609.32900
作者: Jianchang Su,Wei Zhang
类目: Artificial Intelligence (cs.AI); Formal Languages and Automata Theory (cs.FL)
备注:
Abstract:Diffusion language models (dLLMs) predict masked positions in arbitrary order, but their exact constrained decoders still encode constraints as sequential languages, whose state must track every unresolved dependency between positions. For relational constraints this encoding grows exponentially: for same-order copy, every finite automaton needs 4^k states, deterministic or nondeterministic, and every context-free grammar has size 2^\Omega(k) , while the factor graph of the same relation has size O(k) and a 16-entry peak table. We introduce FactorDLM, a training-free decoder that represents finite-domain relations as a factor graph and, at each denoising step, conditions the model’s mean-field prediction on that graph exactly by variable elimination. Decoding cost then grows exponentially with the induced width of the constraint graph, which replaces automaton size as the governing parameter. Because a finite automaton is a chain-shaped factor graph, one compiler enforces syntax and nonlocal relations together: on JSON records with cross-field references, a schema automaton alone leaves references dangling, relational factors alone produce malformed JSON, and the combined plan is valid on both counts, including on records of variable length. Across nine relational benchmarks and three backbones, every output satisfies every declared constraint at 0.4-6.9% projection overhead, where unconstrained decoding is 0-79% valid, and compiled projection answers repeated queries 13.6x faster than CP-SAT with eight parallel workers. Because model-free rules solve three of five standard benchmarks, we construct benchmarks with exact chance and fixed-template floors, on which selecting among exact constrained samples beats greedy projection. Which encoding is cheaper, sequential state or direct factors, depends on the constraint and is computable before decoding begins.
[AI-420] When Less Compute Is More: Adaptive Early Exit Improves Pretrained Outlier Detection
链接: https://arxiv.org/abs/2609.32898
作者: Tianyang Zhou,Leman Akoglu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Pretrained tabular foundation models process every dataset at a fixed depth, with inference costs growing with dataset size. To address this, we present the first study of depth-adaptive early-exit for pretrained outlier detection models. While early-exit is typically motivated by efficiency, we uncover a surprising benefit: exiting at the optimal intermediate layer can also improve detection performance on diverse real-world benchmarks by 4.7-7.3% on average, consistent across three distinct foundation models. First, we investigate the factors driving these gains, and identify a key mechanism: context pollution, i.e., the presence of outliers among in-context samples. Our analysis reveals that nearby in-context samples exert increasing influence on query predictions at greater depths, consistent with a retrieval-based view of these models. In effect, early-exit alleviates the adverse effects of retrieving accurate-yet-polluted neighbors, with gains of 13-21% when context pollution matches the natural outlier rate. Motivated by these findings, we pretrain a plug-in router to select a dataset-specific exit layer, using query outlier labels as privileged information available only during router training. The router operates post hoc, leaving the base model parameters and prediction head unchanged. Experiments on three large real-world benchmarks show that, on clean context, the router recovers up to 45% of the oracle gain with up to 1.8x speedup across three pretrained backbones, with larger gains as context pollution increases.
[AI-421] Optimizing H-Graph Hybridization for Diffusion-Guided RRT
链接: https://arxiv.org/abs/2609.32897
作者: Omer Talmi
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:Sampling-based motion planners guided by diffusion models produce high-quality trajectories in a single run, yet the stochastic diversity available at inference time is left largely unexploited. We present two inference-time diversification strategies for a fixed, pretrained DiTree model, combined via H-Graph hybridization, and evaluate them on a holonomic AntMaze robot across 15 maze scenarios. The first, factorial diversity, sweeps the random seed and Diffusion Goal Bias (DGB) parameter, the second, refinement-only diversity, sweeps the diffusion refinement strength (RS) that controls how much an RRT-generated trajectory is edited. Because a single-run baseline only partially succeeds, we additionally compare H-Graph results with pool-based statistics. H-Graph improves the mean pool length of the factorial and refinement-only diversities by 18.8% and 14.5%, respectively. In addition, it also improves the best individual candidate’s lengths by 9.7% and 6.8%, respectively. And last, compared with the successful baseline’s trajectory length, it improves the results by 18.2% and 19.9%, respectively. These results show that inference-time parameter variation is a reliable, training-free source of path diversity, and that H-Graph hybridization reliably converts this diversity into shorter, higher quality trajectories.
[AI-422] Refreshing Less Selecting Better: Reusing Stale Gradient Features for Efficient Influence-Based Data Selection
链接: https://arxiv.org/abs/2609.32888
作者: Jianchang Su,Yifan Zhang,Wei Zhang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Gradient-based data selection methods such as LESS score each candidate by the alignment between its gradient and a target validation gradient, and recomputing per-example gradient features at every new checkpoint dominates their cost. Across three selection seeds, two model families, two candidate pools, and two target tasks, features cached at a post-warmup checkpoint and paired with fresh validation gradients preserve the ranking 40 optimizer steps later with Spearman correlation from 0.952 to 0.991, while the top-10% subset they induce misses 10 to 22% of the examples that full recomputation selects. We therefore propose Cached Diverse Influence Selection (CDIS), which recomputes gradient features for the top-ranked fraction p of candidates under the stale scores, fits an affine calibration on the recomputed examples, and selects the final subset under source and length quotas. A refresh fraction at or above the selection fraction recovers the exact top- k subset whenever the calibrated stale scores have bounded error, and the budget curves follow this rule: at p=0.3 the recovered top- k subset coincides with full recomputation in every setting, allocating the same budget per stratum recovers the stratified subset at 0.92 to 1.00, and gradient-stage wall-clock drops 3.5 to 3.6 times. Iterating the cache over four checkpoints keeps top- k overlap at 0.98 or higher at 1.9 gradient features per example against 4 for full recomputation, and stale-to-recomputed agreement on the refreshed examples provides a free check for unsafe reuse. Downstream, unconstrained top- k selection collapses to a single data source and scores 13 points below random selection on GSM8K. CDIS scores 12 points above random selection with paired confidence intervals that exclude zero and trails full recomputation by 4.3 points, one training-run standard deviation, at 3.4 times lower selection cost.
[AI-423] StraTune: Adaptive Selection of Revision Operators for Self-Evolving LLM Skills
链接: https://arxiv.org/abs/2609.32886
作者: Zeping Liu,Yan Li,Ni Lao,Gil Wolff,Gengchen Mai
类目: Artificial Intelligence (cs.AI)
备注: 20 pages, 6 figures
Abstract:Large language models (LLMs) can learn reusable textual skills from execution feedback without updating their parameters, but effectively deciding how to revise these skills remains a key challenge. Existing methods typically rely on a fixed revision operator, a search strategy and the revision forms applied under it. However, we observe that no single revision operator consistently performs best across tasks, and repeatedly applying an unsuitable operator can limit further improvement. We propose StraTune (strategy-guided skill tuning), which lets a frozen optimizer LLM choose the revision operator at every round from the optimization state, which is defined as the current execution feedback together with the recorded outcomes of earlier strategies and forms. Candidate skills from every revision operator pass one candidate evaluation, which screens for gains and regressions on a small sample set and validates them on a larger one, and every outcome is written back to the optimization state for later choices. Across four benchmarks and two LLM settings, StraTune outperforms all five baselines in most settings. Ablations attribute the gains to the adaptive choice of the revision operator, since fixed, random, scheduled, and bandit strategy choices all score lower, and skills learned with a small target LLM also improve a stronger one. Code and learned skills are available at this https URL.
[AI-424] Can LLM s Predict the Future? A Brier Score Analysis of Prediction Markets
链接: https://arxiv.org/abs/2609.32885
作者: Yuanbo Li,Zekun Li,Xiaoyan cong
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We study whether model upgrades improve probability estimates for prediction-market questions. Our Resolved Market Forecasting (RMF) benchmark contains 3,000 resolved binary questions across nine domains, on which we evaluate six Claude and Qwen model variants using a question-only, zero-shot protocol. We assess Brier scores relative to an empirical base-rate predictor, examine their Murphy decomposition, and compare models through paired differences, with results stratified by event category and timing relative to training cutoffs. In the reported post-cutoff stratum, the four Claude models achieve Brier scores of 0.183-0.192, improving on the base-rate reference by 0.024-0.033. Qwen 32B does not significantly outperform that reference, although its paired Brier is 0.024 lower than that of the 7B checkpoint. The evaluated Claude version and tier upgrades yield no significant improvement. Within-model differences across event categories exceed the observed differences among Claude variants. These results show why forecasting scores should be interpreted alongside simple probability baselines and question composition: under this protocol, newer versions or higher model tiers do not consistently produce more accurate probabilities.
[AI-425] Counterfactual Self-Evolving Agents for Evidence-Grounded Reasoning
链接: https://arxiv.org/abs/2609.32870
作者: Xing Han,Yuxin Wang,Chen Chen,Wei Dai,Gautham Krishna Gudur,Shijun Li,Hsing-Huan Chung,Gregory D. Hager,Joydeep Ghosh,Paul Pu Liang,Suchi Saria
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 38 pages, 16 figures, 23 tables
Abstract:Self-play proposer–solver methods improve reasoning by generating tasks and learning from verified solutions. However, for evidence-identifiable tasks, where case-specific evidence and domain knowledge determine a checkable answer, self-play requires generating plausible cases whose answers can be independently verified. We introduce counterfactual self-evolution, which generates counterfactual context for reconsidering the original case. A trainable Proposer constructs targeted evidence edits and describes potential outcome changes with causal explanations. We handcraft an expert-verified counterfactual instruction-tuning dataset to teach the Proposer to generate high-quality counterfactuals across a broad range of action–outcome scenarios. Each counterfactual instruction-tuning example specifies an edit within a defined category and explains its hypothesized causal effect on the decision, teaching the Proposer to reason systematically about what changes and why. We instruction-tune the Proposer on these examples, then formulate a fine-tuning reward that integrates feedback from the Solver and Verifier. Across diverse counterfactual scenarios, this reward favors high-quality counterfactuals and warranted revisions, while penalizing changes that overturn correct decisions. The counterfactual context aims to correct errors and strengthen confidence in correct decisions. Accepted counterfactuals accumulate in memory that supplies in-context evidence to the frozen Solver; the Solver adapts through evolving context rather than weight updates. We apply the framework to clinical reasoning, fact verification, and business reasoning. Our evaluation tracks performance over successive rounds as counterfactual memory grows, including transfer to harder cases. Our method achieves superior results across diverse frontier models.
[AI-426] ASCEND: Personal AI Agents for Autonomous Scientific Computing Across HPC Clusters and GPU Workstations
链接: https://arxiv.org/abs/2609.32868
作者: J. Paul Liu,Uthpala Herath,Andrew Petersen
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 19 pages, 6 figures, 6 tables. Code and installer: this https URL
Abstract:Traditional scientific computing requires researchers to translate computational intent into environment configuration, resource requests, and executable jobs, then diagnose failures from scheduler state and application logs. We present ASCEND (Autonomous Scientific Computing Engine and Novel Discovery), an AI-powered agent interface that runs the agent on the researcher’s own laptop, reaching Slurm-managed clusters and a GPU workstation over a multiplexed authenticated connection, with site-specific execution policies checked by locally executed tools; the language model is hosted remotely and holds no credentials. No facility-scale service is required: an account on each resource is sufficient, and the public installer lets users link additional Slurm clusters or workstations of their own. We report four recorded cases: (1) the agent closed a failure-recovery loop on a planted tensor-device fault, submitting, diagnosing, repairing and resubmitting with job-level artifacts preserved; (2) it reproduced the published evaluation of a weather-forecasting model from the author’s released forecasts, agreeing with the published curves to 2.1% (z500) and 2.4% (t850) while identifying a unit discrepancy in the paper’s prose and an initialization-field discrepancy in its released data; (3) it parallelized a released 12,693-line geophysical solver under a bit-for-bit identity requirement, reducing wall-clock runtime from about twelve hours to about two; (4) that requirement exposed two instances of undefined behaviour in the published solver, both repaired and reported upstream. Separately, a pre-specified evaluation of the policy layer found the deployed validator rejected 29 of 30 constructed violations and held the remaining one for approval, while denying 3 of 14 legitimate requests. Autonomy was exercised under author supervision; an end-to-end recovery benchmark remains outstanding.
[AI-427] RoboFoundry: System-as-Policy Evolution for Self-Learning Embodied Agents
链接: https://arxiv.org/abs/2609.32862
作者: Jingsong Liang,Shuhao Liao,Shizhe Zhang,Diyuan Hou,Yuxin Cai,Xinjian Deng,Chengyang He,Wenhui Huang,Runjia Tan,Zhidong Wang,Lan Yu,Xuesong Tian,Guillaume Sartoretti,Jie Luo,Yao Mu,Wenjun Wu,Wanhua Li,Chen Lv
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: Project page: this https URL
Abstract:A foundation model should not act in isolation as an embodied agent. Yet, existing methods often optimize individual components of the agent stack, such as memory, context, skills, or action interfaces, rather than treating the supporting system itself as a unified policy. Moreover, interaction alone does not yield self-improvement unless execution experience is converted into persistent, validated system changes. We therefore propose RoboFoundry, the first embodied agentic framework that formulates this process as Self-Evolving System-as-Policy. RoboFoundry diagnoses capability gaps in decision-making and memory management, converts execution traces into validated task-specific system updates, and promotes recurring improvements to the general system. Evolution operates over two complementary surfaces: a context system that manages active internal context and persistent file-system memory, and a hierarchical skill system that organizes atomic skills, reusable compositions, and failure-conditioned recovery. A shared semantic interface separates embodiment-invariant decisions from embodiment-specific execution, allowing evolved system capabilities to transfer across heterogeneous robots. On EmbodiedBench, RoboFoundry achieves state-of-the-art performance, notably improving GPT-5.5 by 27.8%. It also brings Qwen3.7-Plus to near parity with GPT-5.5 (70.3% vs. 72.7%), showing consistent gains from system-as-policy evolution across foundation models. For long-horizon memory, RoboFoundry outperforms all baselines on RoboMemArena by at least 39.0%, even against methods assisted by external foundation models. On LIBERO-PRO, it further outperforms Cap-Agent0 by 243.8%-679.7% across all perturbation types. In real-world deployments, RoboFoundry demonstrates zero-shot transfer and online evolution across robots and tasks, highlighting its potential for fully autonomous embodied agents.
[AI-428] Logic Gate Networks and Lookup Table Networks as Lightweight Hardware Classifiers for Inter-patient ECG Arrhythmia Classification
链接: https://arxiv.org/abs/2609.32854
作者: Wout Mommen,Lars Keuninckx,Siddharth Patil,Paul Detterer,Achiel Colpaert,Piet Wambacq
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Deep Differentiable Logic Gate Networks (LGNs) and Lookup Table Networks (LUTNs) offer a promising approach for very low power inference due to their use of simple binary logic operations instead of arithmetic. In this work, we generalize the logic gates of LGNs to more than two input pins, naturally arriving at networks consisting of N -input LUTs. To obtain a differentiable expression for training the N -LUT entries, we adopt the Boolean equation of a 2^N :1 multiplexer (MUX) and optimize its input parameters during training. We investigate the applicability of LGNs and LUTNs to inter-patient ECG arrhythmia classification using the MIT-BIH data set. The proposed models achieve up to 94.41% accuracy and a j\kappa index of 0.683 on a four-class task, showing a competitive performance compared to existing CNN-, SVM- and SNN-based methods. Our LGNs and LUTNs only require an estimated 2.89k to 6.17k FLOPs, including preprocessing and readout, which is three to six orders of magnitude less than state-of-the-art methods. We verified our design, which consists of the preprocessing pipeline and a 6-LUTN classifier, by implementing it on a Xilinx Zynq-7000 ZedBoard. The complete system consumes a dynamic energy of 8.25 \mu J/inference, of which only 0.46 nJ is utilized by the LUTN classifier. These results show that both LGNs and LUTNs can be employed as lightweight hardware-based classifiers for inter-patient ECG arrhythmia classification.
[AI-429] ransfer Learning for Edge Classification on Dynamic Text-Attributed Graphs
链接: https://arxiv.org/abs/2609.32849
作者: Tyler Bonnet,Marek Rei
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Social and Information Networks (cs.SI)
备注:
Abstract:Learning transferable representations for dynamic text-attributed graphs (DyTAGs) requires models to capture underlying interaction dynamics that persist across domains. However, existing methods tend to overfit to domain-specific structural, temporal, and semantic patterns, limiting edge classification performance under distribution shifts. To expose and address this, we formally establish a leave-one-domain-out (LODO) transfer learning protocol for edge classification on DyTAGs. Under this protocol, we demonstrate that state-of-the-art self-supervised methods for dynamic graph learning perform poorly when transferred to unseen domains. Strikingly, existing methods underperform a structurally and temporally unaware Bag of Events (BoE) model we introduce, which inputs only unordered sequences of node and edge text features. Proceeding from the BoE, we propose Spatio-Temporal Semantic Alignment (STSA), which integrates a spatio-temporal encoder that fuses representations of time deltas and node occurrence frequencies into a unified manifold. STSA is trained with a Contrastive Semantic Forecasting objective, which anchors edge representations to a multi-domain textual latent space initialized by a pretrained language model, providing a robust prior that outperforms BoE and all existing methods we evaluate.
[AI-430] Scanning While Imagining: A Scene-Graph World Model for Robotic Ultrasound Navigation
链接: https://arxiv.org/abs/2609.32837
作者: Xuesong Li,Shuai Chen,Feng Li,Zhongliang Jiang,Nassir Navab,Yuan Bi
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Ultrasound (US) acquisition depends on the operator’s ability to interpret anatomy and anticipate how the view will change with probe motion. Many robotic US navigation methods select actions without explicitly predicting these anatomical changes. We propose SonoGraph-WM, an action- and goal-conditioned world model for anticipatory probe navigation. The model represents anatomy as scene graphs (SGs), capturing visible structures, their geometry, and spatial relationships without synthesizing US images. Given a history of SGs and probe poses, a unified Transformer jointly predicts future SGs and poses. A receding-horizon planner recursively imagines candidate trajectories, selects the shortest predicted path reaching a goal graph, and follows it over a short execution horizon before replanning from new observations. To reduce reliance on tracked and anatomically annotated US sequences, we generate aligned SG–pose training data from computed tomography (CT) label maps along surface-constrained probe trajectories. On four held-out CT cases, spatial relation F1 remains above 93% over 20 prediction steps, and closed-loop navigation achieves 77.50% and 75.00% success for the gallbladder and pancreas, respectively, using annotation-derived SGs. In robot–phantom navigation experiments with label-map-derived SGs, the planner reached the target view in 73.7% of trials. These findings support CT-supervised anatomical world modeling for probe planning and highlight the importance of frequent observation updates for reliable navigation. Project Page: this https URL
[AI-431] FinancialAuditBench: Benchmark Construction under Differential Privacy Using Real-World Priors
链接: https://arxiv.org/abs/2609.32835
作者: Jerry Huang,Sarvesh Babu,Matt Van Buren,Alexander Wang,Pranav Pillai,Arush Jain,James P. Burton,Julia Hockenmaier
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As AI agents are becoming widely adopted in the financial services industry, careful measurement is essential to understand where they can be reliably deployed and where oversight and professional review remain necessary. Such measurement, however, is constrained by limited access to proprietary or privacy-sensitive data. Existing benchmarks therefore often rely on publicly available data, human- and/or LLM-authored tasks, or simplified settings. We introduce FinancialAuditBench, a benchmark for evaluating agents on financial statement audit tasks, along with a framework for systematically generating synthetic engagements. Our task generation framework leverages differentially private aggregate statistics from historical audits along with audit expertise contributed through over 1,100 hours of benchmark development and review. FinancialAuditBench consists of 90 tasks spanning workpaper completion and review across six synthetic audit engagements, each containing an average of 179 files. Evaluation on eleven frontier models shows that while agents complete substantial portions of staff-level audit tasks well, they sometimes perform inappropriate procedures or produce incorrect documentation. Beyond financial auditing, our framework offers an approach for systematically generating synthetic tasks for model evaluation and training in privacy-sensitive domains.
[AI-432] Routing Drift Alone Does Not Diagnose Failure in Merged MoE LLM s
链接: https://arxiv.org/abs/2609.32821
作者: Yuanyi Wang,Yanggan Gu,Su Lu,Guanghao Zhu,Pengkai Wang,Yifan Yang,Congkai Xie,Zhaoyi Yan,Jianmin Wu,Hongxia Yang
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Model merging efficiently combines specialized large language models (LLMs) without joint retraining, but can substantially alter expert routing in Mixture-of-Experts (MoE) models. Such \emphrouting drift is often interpreted as routing failure, raising a fundamental question that remains unclear: \emphdoes routing drift after MoE merging actually indicate routing failure, and what evidence should justify repair? We investigate these questions across DeepSeekMoE, OLMoE, and Qwen3-MoE proposing a routing analysis toolkit for controlled counterfactual interventions and token-level analysis. By crossing source and merged router inputs and parameters, we attribute most expert reassignments to input shifts rather than parameter changes at the same layer. However, source-relative routing differences poorly predict next-token likelihood gains from source-route restoration, and different expert selections can produce directionally similar mixture outputs. We therefore operationalize routing failure as \textittask loss recoverable under a specified routing intervention, with non-routing parameters fixed. These tests detect recoverable loss under deliberate router corruption, whereas source-route restoration does not establish reliable task benefits in the evaluated merged models. Motivated by these, we propose \emphSelective Router Repair (SRR) as a case study, and find that source-specialist token-likelihood advantages do not reliably identify beneficial local corrections. Together, these findings show that \textbfrouting drift alone is insufficient evidence of routing failure: source-informed corrections must be judged by their task-level intervention effects. The analysis toolkit and SRR code are released.
[AI-433] Rank Collapse Is Recoverable Growing |Q| Is Not: Out-of-Sample Early Warning for Value Divergence in High-UTD Soft Actor-Critic
链接: https://arxiv.org/abs/2609.32819
作者: Tianqi Bu,YuXuan Peng,Junteng Tu,Henghui Xiao
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 31 pages, 9 figures; preprint under review
Abstract:Raising the update-to-data (UTD) ratio breaks off-policy critics in two ways grouped as “plasticity loss”: collapsing representations and growing value magnitude |Q| . We separate them in Soft Actor-Critic (SAC) with scaled critics (width 2048, no normalization). Collapse is survivable: at UTD ratio 16, HalfCheetah critics with most units dormant keep learning, and the training guard, which stops runs whose loss or |Q| explodes, never flags them. Within one high-UTD SAC configuration, runs start close together, and how far a critic’s \log_10|Q| has climbed by step 15k, its early growth, ranks the runs by how soon the guard flags them. At 15k, a flagged run’s |Q| sits a median of over a hundredfold below its flag level, yet the climb’s rate already orders the flags (Harrell’s C and out-of-sample AUC 0.78 on Walker2d, 0.98 on Ant, at UTD ratio 4). Dormancy does not. Aborting on this rate saves about a tenth of held-out Walker2d compute and stays net-positive live. A LayerNorm critic lowers the rate, removes the flag on Walker2d at UTD ratio 4 and lowers the return.
[AI-434] When Can First-Order Models of Fine-Tuning Bound Forgetting?
链接: https://arxiv.org/abs/2609.32818
作者: Jianchang Su,Wei Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Fine-tuning a language model on new data can make it forget facts that it should keep. We ask whether measurements taken at the start of a fine-tuning run can bound, for each protected fact, the probability that the run makes the model forget it. In LoRA fine-tuning with stochastic gradient descent on models from 0.6B to 14B parameters, a first-order response model estimated by finite-difference probes predicts changes of per-fact margins with correlation 0.974-0.998. Predictions of forgetting built on this model nevertheless failed, because forgetting requires parameter changes far outside the region in which the model was validated. The probes can, however, bound the probability that a margin first falls below a boundary near zero: we derive Freedman and Azuma first-passage bounds for a linear surrogate of the margin and test on new runs whether they hold for the model. The bounds contain a term R that measures how much the response coefficients change during the run. The simplified Freedman bound, which sets R = 0, certified most facts but was violated in 14 of 112 conditions, and every fact on which it was violated had R = a, where a is the distance of the fact’s margin to the boundary. The complete Freedman bound certifies only facts with R a, and it held in every condition. On the violated facts, the spread of the margin across test runs was a median of 14.6 times the prediction of the response model, so the failures are breakdowns of the model, and in our data they occurred only where R = a. We found this pattern post hoc and tested it in two preregistered confirmatory studies with 43 new conditions: the complete bound held in all of them, and the simplified bound failed there on only 3 facts, each with R = a. First-order models of fine-tuning can thus bound forgetting on the facts whose response coefficients change by less than their distance to the boundary.
[AI-435] Right Answer Wrong Reason : Accuracy Consistency and Consensus Are Misleading Indicators of LLM Faithfulness in Clinical Decision Support
链接: https://arxiv.org/abs/2609.32817
作者: Bharath Kumar Bolla,Bharath Kumar Bolla,Vishnu Surya Reddy Nandi
类目: Artificial Intelligence (cs.AI)
备注: Accepted and received best paper award from ICETCI 2026 Conference, this https URL
Abstract:Clinical Large Language Models (LLMs) achieve strong medical-exam accuracy; however, a correct answer does not guarantee that the explanation names the concepts that actually drove the decision. We introduce three lightweight, directly interpretable metrics for this faithfulness gap: the Explanation Stability Index (ESI), which measures reasoning consistency across repeated queries; the Causal Faithfulness Score (CFS), which tests whether cited concepts drive predictions via concept ablation; and the Perturbation Stability Score (PSS), which measures robustness to semantic-preserving paraphrases. By evaluating six LLMs on 150 MedQA-USMLE questions (900 model-question observations), we found that only 23.3% of the cited clinical concepts were causally necessary. Correct answers had lower CFS than incorrect answers (0.212 vs. 0.398), answer consistency negatively predicted CFS (Spearman r = -0.466), and model pairs could agree on answers while sharing only 8.8% of cited reasoning concepts. These results show that accuracy, consistency, and consensus are incomplete safety signals for clinical decision-making support. The evidence is behavioral rather than mechanistic: concept ablation tests counterfactual sensitivity of outputs, not internal circuits.
[AI-436] Beyond Accuracy: Counterfactual Frag ility and Demographic Bias in Clinical Evaluation of LLM s
链接: https://arxiv.org/abs/2609.32807
作者: Chaitai Deb Purkayastha,Bharath Kumar Bolla,Vishnu Surya Reddy Nandi
类目: Artificial Intelligence (cs.AI)
备注: Accepted into AIKP 2026 Conference
Abstract:Clinical LLM evaluation often emphasizes answer accuracy; however, accuracy alone does not test counterfactual consistency or demographic robustness. We evaluated six LLMs on 150 MedQA USMLE questions using two automated perturbation tests to assess their performance. The counterfactual validity (CFV) test asked each model to make a minimal, plausible clinical change that would make a different answer correct. The demographic robustness test added six demographic prefixes to the same vignette and compared the answers and explanations with a no demographic baseline. Of the 900 CFV attempts, 228 (25.3 %) were valid and 672 were invalid. Across 5,400 demographic comparisons, 1,097 answers were changed (20.3%). Automated judging identified 3,128 stereotype evidence flags, including 1,932 in the broad Other category. MedGemma 27B achieved the highest accuracy (87.1%) and CFV (63.3%), lowest answer change rate (16.0%), and low mean Explanation Demographic Dissonance (EDD) score (0.169). However, its accuracy still exceeded its CFV, indicating that correct answers do not guarantee reliable performance on the counterfactual validity task. OpenBioLLM had the highest answer change rate and EDD, whereas GLM had the highest stereotype flag rate. These findings show that accuracy, CFV, answer stability, EDD, and stereotype evidence capture different evaluation aspects. Because all judgments were automated and no clinician validation was available, the results support safety screening but do not establish clinical deployability of the model.
[AI-437] Finding Emotions Where They Belong: Rethinking Audio Emotion Recognition through Masked Temporal Affective Grounding
链接: https://arxiv.org/abs/2609.32804
作者: Abdelrahman Mohamed,Lars Kai Hansen,Zheng-Hua Tan
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注:
Abstract:Audio emotion recognition (AER) typically assigns a single label to an entire recording, leaving the temporal scope of that label ambiguous when multiple speakers and affective events are present. We address this limitation by reformulating AER as a Temporal Affective Grounding (TAG) task that associates emotions with temporally bounded speech spans and vocal tone descriptions. To support this formulation, we curate temporally annotated versions of existing emotion recognition datasets and construct recordings containing two to four affective speech spans, including overlapping speech. Training in this longer format with a standard language-modeling objective can degrade both emotion recognition and temporal grounding performance, while tone descriptions can provide shortcuts for emotion prediction. To address these challenges, we introduce Masked Temporal Affective Grounding (M-TAG), a supervised training objective that combines full-sequence language modeling with emotion and timestamp cross-entropy losses under attention masking. The masking varies the context visible to emotion-label tokens to reduce reliance on shortcuts and improve generalization, while the timestamp loss incorporates a distance-aware weight to penalize larger temporal errors. We evaluate EMO-TAG, a model fine-tuned using our dataset and objective, on emotion recognition and affective temporal-grounding against three AER and audio-language baselines: Flamingo-Next, Audio-Reasoner, and AffectGPT. Our results show that existing models achieve limited affective temporal-grounding despite competitive emotion recognition performance.
[AI-438] Nutri-ATLAS: Embodied Agent for Tabulated Lookup and Assistance for Smarter nutrition
链接: https://arxiv.org/abs/2609.32803
作者: Uttej Kallakuri,Boxun Hu,Ankur A. Butala,Najim Dehak,Tinoosh Mohsenin
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Generative and Agentic IoT systems offer a promising foundation for digital healthcare applications that combine sensing, personalized reasoning, and autonomous interaction in real-world environments. Nutrition assistance is a natural use case, but existing Large Language Model (LLM)-based systems are often limited to passive text interaction and static context, making them unreliable when food descriptions are ambiguous or nutritional evidence is missing. We propose Nutri-ATLAS, an Embodied Agent for Tabulated Lookup and Assistance for smarter nutrition in the real world. It integrates graph-grounded nutrition reasoning, hardware-aware LLM selection, and robot-based evidence acquisition. Nutri-ATLAS builds a unified Food-Nutrient knowledge graph from USDA FoodData Central and FoodKG and learns 64-dimensional GATv2 food and recipe embeddings. A shared hybrid graph-text scoring mechanism supports food nutrition extraction, nutritional gap filling, substitute retrieval, and recipe-level meal composition, while an LLM-guided skill interface navigates landmarks, updates dietary-context and food-accessibility memory, and grounds recommendations in observed food availability. We evaluate Nutri-ATLAS across nutrient estimation, substitution retrieval, recipe recommendation, patient-profile adherence, edge deployment, and real-world embodied execution. On HealthyFoodSubs, the hybrid retriever achieves 37.9% MAP, 80.7% RR@5, and 90.1% RR@10. On NutriBench v2, Dense+GAT retrieval grounds nutrient estimation across nine quantized Qwen3.5-9B configurations. On PFoodReQ, Nutri-ATLAS reaches 78.8% MAP, 83.0% MAR, and 77.5% F1. A patient-profile study shows adherence to allergy and healthy-target constraints for all selected cases.
[AI-439] Agent Habit: Characterizing Distinct Behaviors of Agents on Everyday Tasks
链接: https://arxiv.org/abs/2609.32795
作者: Woojung Song,Hoyeol Yang,Jeonghoon Shim,Sungjib Lim,Jonggeun Lee,Yunho Choi,Yohan Jo
类目: Artificial Intelligence (cs.AI)
备注: 60 pages, 16 figures
Abstract:Large language model (LLM) agents assist users with everyday tasks that can be completed in many reasonable ways. Even when their answers are useful, how agents carry out these tasks may not match users’ preferences and needs. For example, agents differ in whether they ask clarifying questions or search the web. We introduce HABIT, a taxonomy of 23 behavioral axes in five categories, which three authors and three LLMs derive bottom-up from 408 agent trajectories across 17 domains. On held-out tasks, HABIT distinguishes models more clearly than existing taxonomies of human values and agent actions while supporting comparably consistent annotation. Building on HABIT, we construct AgentHABIT, a benchmark that profiles each agent’s behavioral tendencies from its trajectories on 86 everyday tasks. Profiling 18 models with AgentHABIT reveals a range of distinctive tendencies. For example, most GPT and Claude models state their assumptions and offer alternatives when requirements conflict, whereas Qwen and Google’s models more often leave assumptions or changes to requirements unstated. These profiles remain recognizable even when built from entirely different sets of tasks, indicating that they reflect general tendencies rather than task-specific behavior. Prompting agents to adopt specific behaviors shifts some axes readily but barely changes others, while fine-tuning on another model’s trajectories changes only part of a model’s profile and leaves much of it intact. Overall, HABIT and AgentHABIT provide a systematic framework for characterizing how agents carry out everyday tasks beyond task success, offering insights to guide the development of agents whose behavior better fits users’ needs.
[AI-440] T5: Twin-Critic Training for Token-Level Thoughts in Reinforcement Mid-Training
链接: https://arxiv.org/abs/2609.32791
作者: Nan Qiao,Yebin Yang,Weinong Wang,Shuning Wang,Shangpin Peng,Fengyuan Lu,Xinming Wang,Zhehan Kan,Ruixu Zhang,Songyang Zhang,Sheng Yue,Yonglong Tian,Ju Ren
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement mid-training lets language models learn internal thoughts from unlabeled text, but efficient token-level credit assignment remains challenging. Existing group-relative methods require costly repeated generation. Learned critics offer single-rollout feedback, but accurate return prediction alone does not ensure reliable policy updates. Our analysis shows how training–inference mismatch and PPO clipping prevent a common offset in advantage estimates from cancelling out, introducing additional update drift. We propose \tfour, a twin-critic method that calibrates token-level advantages from a single generated trajectory. After warmup and held-out qualification, the critics provide two advantage estimates, combined using action-dependent weights learned through a conditional-moment saddle-point objective. This objective brings the average advantage at each prefix toward zero, while a signal-retention constraint prevents the correction from erasing the learning signal. Sharing information across text positions avoids repeated sampling of each prefix. Theoretically, we characterize optimal mixing under the signal-retention constraint and establish an upper bound on residual mean-induced drift. Experiments show that, compared with the state-of-the-art critic-free method, \tfour improves mean benchmark performance by 7.8% and reduces mean training-step time by up to 63.4%.
[AI-441] Getting Motif-ated: Controllable AI Compositions from Injected Motif Prompts NEURIPS2026
链接: https://arxiv.org/abs/2609.32788
作者: Chao Peter Yang,Cynthia Rudin,Yue Jiang,Simon Mak,Stephen Ni-Hahn
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multimedia (cs.MM); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注: NeurIPS 2026, Creative AI Track
Abstract:Deep learning has transformed symbolic music generation by borrowing the training paradigms of large language models, with systems such as NotaGen now producing complete, stylistically convincing classical scores from a short prompt. These systems could become powerful creative partners, helping musicians generate endless possibilities. However, current systems expose almost no control handles on the music itself. In principle, control handles could be built into a foundation model trained from scratch, but this is rarely practical without massive amounts of quality annotated data and compute resources. We therefore present MotiGen, a recipe for retrofitting pretrained symbolic music models to use new instruction prompts. MotiGen injects a musical motif as a structured prompt line, reinforces it with a scalar attention bias toward the motif tokens, and learns the association with a two-phase curriculum. First, it learns from focused excerpts cropped around motif occurrences in the training data, then full scores including the motifs. Our experiments show that our model composes with the prompted motif in over 92.3% of generated pieces. Generated pieces using a variety of motifs are included in our sample site: this https URL.
[AI-442] Learning response-aware patient dynamics for respiratory support
链接: https://arxiv.org/abs/2609.32782
作者: Xiaolei Lu,Shamim Nemati
类目: Artificial Intelligence (cs.AI)
备注: Under review
Abstract:Respiratory support can shape the short-term physiological trajectory of critically ill patients, but patients receiving the same intervention may follow different physiological trajectories. Clinical patient dynamics models typically predict future states from recent physiology and recorded interventions, while physiological change is mainly represented through the predicted future state. We propose a response-aware patient dynamics model that explicitly represents physiological change during autoregressive state updating. The model decomposes predicted physiological change into state-dependent baseline dynamics and respiratory-support-associated deviations, with room air providing a reference for the decomposition. We provide a formal analysis of this reference-anchored formulation. A response pathway encodes the predicted physiological change and uses it to update the latent patient state across the forecast horizon. Across ICU cohorts from two independent institutions, the proposed model achieves comparable overall trajectory prediction to patient dynamics baselines, with more consistent improvements when physiological states are changing.
[AI-443] Agent ic Network Traffic Monitoring
链接: https://arxiv.org/abs/2609.32778
作者: Manuel Tsoukatos,Hayden Jananthan,Jeremy Kepner
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Databases (cs.DB); Distributed, Parallel, and Cluster Computing (cs.DC); Networking and Internet Architecture (cs.NI)
备注: 5 pages, 3 figures, to appear in IEEE URTC 2026
Abstract:As the use of agentic artificial intelligence increases in nearly every industry, there exists a widening attack surface. It is necessary to monitor agents to ensure that agents are acting in a way that is aligned with the users intent. Auditing an agent’s network traffic provides a clear record of the agent interactions. This work presents a novel approach to monitoring the network traffic of agentic systems using complex valued hypersparse traffic matrices by integrating DBOS (DataBase OS), the OneSparse PostgreSQL database, and the GraphBLAS math library. To develop these concepts an agentic simulator was constructed, allowing a varying numbers of AI agents to collectively survey a virtual environment using different strategies. The resulting network traffic matrices enable easy monitoring of the AI agents.
[AI-444] Forecasting Intraday USD/CAD Exchange Rate with News-Derived Monetary-Policy Signals
链接: https://arxiv.org/abs/2609.32773
作者: Maya Kodeih,Aliaa Alnaggar,Mucahit Cevik
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Computational Finance (q-fin.CP)
备注: IEEE CASCON 2026
Abstract:Monetary-policy announcements and central-bank communications play a central role in foreign exchange markets, yet their qualitative, unstructured form makes their forecasting value difficult to quantify. While prior research has largely focused on sentiment extracted from financial news, comparatively little is known about the relative contribution of different dimensions of monetary-policy communication. Existing studies primarily evaluate whether textual information improves overall forecasting performance but provide limited insight into which communication channels drive such improvements. To address this gap, this paper introduces a statistical attribution methodology that decomposes monetary-policy communication into interpretable channels and quantifies their incremental forecasting contribution under false-discovery-rate control. Monetary-policy news is transformed into structured communication signals using large language models (LLMs) and temporal feature engineering. These signals are evaluated using rolling-window experiments with tree-based machine-learning models. The results show that monetary-policy communication contains measurable predictive information. Attribution analysis shows that predictive value is concentrated in a small subset of signals, with communication timing providing the strongest individual feature-level contribution, targeted communication-activity measures also contributing positively, and LLM-derived sentiment providing complementary information at the group level. The findings indicate that communication-based forecasting value extends beyond sentiment alone and that attribution, rather than aggregate accuracy alone, is central to evaluating news-derived signals.
[AI-445] Learnable Randomization as Commitment Against Adaptive Optimizers
链接: https://arxiv.org/abs/2609.32772
作者: Zihan Deng,Chuanzhi Xu,Xiaozhen Zhong,Haoyang Li,Junjie Huang
类目: Computer Science and Game Theory (cs.GT); Artificial Intelligence (cs.AI)
备注:
Abstract:A pricing page can walk the posted price up to the last amount a buyer still accepts, a recommender can hold back a better item for a barely acceptable promoted one, and a classifier can shift its boundary once applicants change their features. The system predicts the response and then picks the menu that serves its own objective, so the surplus above the user’s cutoff is taken. Playing the single best action publishes that cutoff, while noise on actions the user would never take throws away payoff and teaches the platform that a worse menu is still acceptable. We study unpredictable near-optimal policies (UNOP), which mix uniformly on near-best actions that remain individually rational. The mixture is a commitment about the response. On a finite price grid, when the best sure-demand price strictly out-earns the randomized band, a seller who already knows the curve posts below the band, and the purchase that occurs is deterministic. Knowing that curve is not the same as predicting the next draw. The mixture can be learned and the optimizer can match its best response, while the user’s payoff stays higher because the mixture changes which action is targeted. In pricing and in policy-aware recommendation this leaves more surplus than greedy play when the platform optimizes against the curve and more than one action is acceptable. The gain goes away under quality ranking, a singleton near-optimal set, a wrong utility estimate, or a short-horizon explorer. That is also where mixing should be turned off if the other side is trying to cooperate.
[AI-446] Continual Learning via Self-Probe Gradients
链接: https://arxiv.org/abs/2609.32771
作者: Dongkyu Cho,Rumi Chunara,Sungmin Cha
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 27 pages, 18 figures
Abstract:Adapting pretrained models to new data can cause catastrophic forgetting of previously learned behavior. When only a few past samples remain, they give continual learning methods sparse and narrow evidence about what to preserve. We show that language models can expand this evidence through self-probing, in which the frozen model generates new inputs from the retained samples and records its own predictions on them. Unlike prior work that replays such data as training examples, our method, CPLUS uses self-probe and past-sample gradients to scale down parameter updates that conflict with prior behavior. Experiments with five language models on four benchmarks show three results. First, the same probes preserve more prior behavior as gradient signals than as replay data. Second, CPLUS learns the new data while consistently reducing forgetting more than existing baselines, especially when past data are scarce, and this protection extends to benchmarks not used for training. Third, we observe that CPLUS also becomes more effective as models grow: within the Qwen3 model family, it recovers an increasing share of the forgetting caused by standard fine-tuning.
[AI-447] Mandela-Bench: Multimodal Models Remember Canonical Images Instead of Seeing Them
链接: https://arxiv.org/abs/2609.32763
作者: Yicheng Bao,Zhenkun Gao,Xiahui Guo,Mingqian Yang,Xueheng Li,Bangwei Liu,Mingang Chen,Lijun Li,Xuhong Wang,Xin Tan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Historical photographs and other canonical images can now be edited seamlessly with a single instruction, often leaving no reliable pixel-level trace. In such cases, the only evidence of manipulation may be a fact about what the image depicts. Existing benchmarks instead rely on generator artefacts, image-caption inconsistencies, visual implausibilities, or external references, and therefore do not test whether a model can use its own world knowledge to verify a recognized image. We introduce Mandela-Bench, containing 1,507 edits of canonical images: 1,359 knowledge-only forgeries, each contradicting one verifiable fact, and 148 anchor-free controls that preserve the editing process without introducing a factual contradiction, together with 474 untouched originals. We score not only whether a model detects a forgery, but whether its explanation identifies the inserted entity or the fact being violated. Across 36 multimodal models, from 0.8B parameters to frontier scale, we find a consistent failure mode. When a public figure is removed from a familiar photograph, models still name that person in up to 72.7% of responses. Some models can distinguish the replacement face from the original when shown in isolation, yet still judge the full edited photograph as authentic. Providing the true event and date does not improve knowledge-grounded detection, whereas providing the same information after cropping away the recognizable composition does. Even under explicit verification prompts, only one of the 36 models meets the KGR criterion on at least half of the forged images. These results suggest that the failures cannot be explained by missing knowledge or inadequate perception alone. Instead, they are consistent with recognition biasing verification toward the remembered canonical image rather than the observed edit.
[AI-448] he Extender: A Log-Structured Transformer
链接: https://arxiv.org/abs/2609.32759
作者: Jakob Eriksson(UIC)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce the Extender, a log-structured variant of the standard Transformer architecture. In a standard Transformer, each layer communicates with subsequent layers exclusively via the residual \mathbfh , a superposition channel. The Extender adds a concatenation channel \mathbfx : each layer \ell emits both a residual update \delta_\ell which is added to \mathbfh , and a much smaller extension \epsilon_\ell which is appended to \mathbfx . While both the FFN and \mathbfq see \mathbfh , the attention \mathbfkv projections take only \mathbfx as input. As a result, the fully extended \mathbfx contains the complete input for the \mathbfkv projections of all layers, reducing the persistent attention memory footprint from 2Ld_model to \sum|\epsilon_\ell| . We find that with |\epsilon_\ell|=32 , the Extender matches Transformer accuracy on short-context (CORE) tasks at 199M-924M parameters, and exceeds Transformer accuracy on long-context (RULER) workloads, again at 924M parameters. For our 1664-wide, 924M model, the Extender’s persistent attention memory footprint is 104\times smaller than MHA. The memory savings grow with model width.
[AI-449] Action Shaping: Policies Absorb What They Can Express
链接: https://arxiv.org/abs/2609.32752
作者: Yanjun Chen,Jinghan Wang,Xiaoyu Shen,Wenjie Li,Wei Zhang
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Reward shaping has a theorem: a potential-based term can be removed without changing the optimal policy. The same practice on the action channel, an offset added in training and dropped at deployment, has no theorem. Nothing cancels an action offset, so the correction is kept at deployment or removed without a guarantee. We call it action shaping and state its principle. A trainable policy absorbs an offset its own output layer can reproduce exactly, which is what we mean by express; what is absorbed can be removed with the return intact. Its minimal instance is a zero-initialized linear head behind a learnable gate, added to an actor that trains through a learned action-value function, with no penalty or schedule. The gate rises and then falls on its own, for deterministic and stochastic actors alike, and on 20 tasks removing the head costs almost nothing. The condition is exact reproduction, not capacity: a nonlinear head with more parameters is not absorbed, and in a paired control, one linear path added to a nonlinear base head restores absorption. Exact reproduction gives the loss a flat direction that gradient noise drifts along, and the offset’s amplitude indicates, before removal, what dropping the head will cost. Action shaping thus gains the counterpart of the shaping theorem, a condition for absorption, together with the mechanism behind it and a diagnostic that reads it. Policies absorb what they can express, and only that.
[AI-450] CUA-Sandbox: Efficient Environments for Computer-Use Agent Reinforcement Learning
链接: https://arxiv.org/abs/2609.32750
作者: Xin Yan,Zhengbo Jiao,Jiaqi Liu,Zhenglin Wan,SiYuan Ma,Xuliang Yu,Tianyi Jiang,Chubin Zhang,Pengfei Zhou,Wangbo Zhao,Xingrui Yu,Bo An,Yang You,Ivor Tsang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning enables computer-use agents to improve through interaction with real software environments, including websites and desktop applications. However, conventional deployments replicate an initialized runtime for each independent rollout, even when trajectories use the same software, incurring repeated memory and initialization costs as the number of parallel environments grows. Does an independent computer-use environment require an independent execution runtime? Our key observation is that trajectories require independent mutable state, while initialized application runtimes can be reused across concurrently evolving environments, making state the natural unit of environment independence. Guided by this observation, we introduce CUA-Sandbox, which separates private state capsules from shared runtimes through state-scoped execution and transactional lifecycle operations, including resets and branches, while retaining the original software interfaces and task evaluators. Experiments show comparable or improved task success relative to Docker, while substantially reducing rollout and resource costs. CUA-Sandbox achieves up to a 6.20x increase in rollout throughput, a 9.2x reduction in per-environment memory, and a 504x reduction in incremental storage.
[AI-451] SkillVine: Agent Skill Evolution via Branching Exploration
链接: https://arxiv.org/abs/2609.32731
作者: Kaiwei Liu,Jiqian Dong,Liran Dong,Shuai Mao,Mingming Zhao,Bufang Yang,Jie Chuai,Zhitang Chen,Guoliang Xing,Zhenyu Yan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Agent skills encapsulate reusable procedural knowledge that enables LLM agents to perform tasks, and they can be improved automatically using trajectories from interactions with the environment. This is the classic problem of skill evolution. Existing approaches predominately follow a linear evolution paradigm, in which updates are sequentially applied to the latest skill-library version. As a result, they inevitably fall into local optima, leaving many promising evolution paths unexplored. We propose SkillVine, an automatic skill-evolution framework that formulates skill evolution as a graph search problem and employs a branching exploration strategy. Equipped with a trunk-branch collaborative searching mechanism, an intelligent parent-node selector, and an adaptive-granularity update rule, SkillVine achieves a balance between exploration and exploitation. We evaluate SkillVine on 5 benchmarks with two LLMs. Results show that SkillVine discovers better skill-library versions along branches than along the linear trunk and achieves the best test performance in nine of ten benchmark-model combinations.
[AI-452] Learning to Refer: Client-Resolved Generation for Privacy-Aware Language Models
链接: https://arxiv.org/abs/2609.32706
作者: Jeongho Yoon,Chanhee Park,Yongchan Chun,Duong Tuan Thanh,Sungbin Han,Chanjun Park,Hyeonseok Moon,Heuiseok Lim
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Cloud-based large language models (LLMs) require users to disclose plaintext data to service providers, creating privacy risks in sensitive domains. Existing privacy-preserving approaches often trade utility for protection, incur substantial computational or communication overhead, remain vulnerable to reconstruction from intermediate representations, or protect only a subset of the training and inference pipeline. We introduce Client-Resolved Generation (CRG), a genera- tion interface that separates server-side generation from the lexical realization of input-derived content. The client transmits only pooled and noise-perturbed rep- resentations, while input-derived output content is represented using request-local positional references and resolved to its original strings only on the client. This interface protects private input and input-derived output content during both train- ing and inference while allowing the service provider to keep its proprietary model parameters hidden from the client. At the same time, exact lexical reuse remains possible without directly exposing the reused content on the provider-visible gen- eration path. We evaluate CRG on medical and document-grounded QA, sensi- tive identifier transfer, and tool calling, together with reconstruction and raw-logit leakage analyses. On SealTools, CRG improves complete-call exact match from 57.3% to 79.9% over the input-privacy framework PPFT, with larger gains as more required output content can be resolved through references. Together, these results show that CRG provides a practical interface for privacy-sensitive cloud LLMs by reducing plaintext exposure across both input and output pathways while preserv- ing task utility and server-side model confidentiality.
[AI-453] CoWindow Attention: Full Causal Coverag e Is a Collective Property
链接: https://arxiv.org/abs/2609.32704
作者: Jingze Shi,Zhangyang Peng,Xianduo Li,Yanlin Qi,Xiaotian Lin,Haoxian Chen,Liangdong Wang,Guang Liu,Yuyu Luo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:FullAttn repeatedly exposes the complete causal history to every attention head, creating substantial redundant computation and memory traffic even with IO-efficient dense kernels. We introduce CoWA, a structured attention architecture that distributes access to the causal history across KV heads. All heads share near-diagonal and prefix-sink windows, while complementary long-range windows partition the remaining history. Their union provides full causal coverage although each head attends sparsely to distant tokens. This position-defined attention pattern requires no learned router or indexer, is used consistently during training and inference, and aligns with KV-head tensor parallelism. A window-matched ablation at 8K isolates the effect of complementary long-range allocation: CoWA with 100% collective coverage reaches 89.73% accuracy, compared with 89.97% for FullAttn, while duplicated long-range windows perform substantially worse. Across a broader controlled associative-recall comparison with matched token budgets, CoWA closely tracks FullAttn as the context grows, whereas other sparse patterns lose a substantial fraction of the associations. In an attention-operator benchmark at 128K tokens with tensor parallelism, CoWA reduces forward and backward latency during training by 7.4x and 8.6x and decoding latency during inference by 3.0x over FullAttn. Its per-rank peak operator memory matches FullAttn during training and is 7.6x lower during decoding. Across scaling-law training from 0.6B to 14B parameters, CoWA closely tracks FullAttn in perplexity while reducing total training FLOPs. The resulting 14B models and 32B models from separate continued training achieve comparable knowledge, reasoning, and long-context retrieval scores to FullAttn. These results show that full causal coverage can be a collective property of the head ensemble rather than a duplicated property of every head.
[AI-454] CAIRN: Dynamic Fact-Intent DAGs for Multi-Agent Exploration
链接: https://arxiv.org/abs/2609.32700
作者: Zuyao Xu,Yuyang Jia,Junwei Guan,Xiang Li,Kaiwen Shen,Zhiqiang Dong
类目: Artificial Intelligence (cs.AI)
备注: 27 pages, 10 figures, 4 tables
Abstract:LLM-powered autonomous systems have demonstrated promising capabilities in mathematical reasoning, engineering, and cybersecurity. Yet how to organize these systems for effective, reliable, and sustained performance remains an open question. In this paper, we present CAIRN, a fact-intent-driven multi-agent paradigm for goal-directed exploration. CAIRN represents observations and planned investigations as a dynamic directed acyclic graph (DAG). A reasoner interprets facts to propose intents, which workers execute to produce new facts. Each intent references its supporting facts and defines a potential exploration branch. The persistent graph preserves goals, dependencies and findings across workers, supporting knowledge reuse and parallel exploration. The graph also makes execution trajectories traceable and auditable, providing a basis for human verification and intervention. We evaluate CAIRN across cybersecurity and mathematical reasoning tasks, examining task success, time to solution, and token consumption. DAG-based coordination can incur higher token costs with no observable performance gains on tasks that require little effort. However, on high-effort tasks (at least 1M tokens), we observe faster solutions in 76.5% of cases, with speedups of up to 3.08x. Moreover, as task effort increases, these time gains become more pronounced while relative token overhead declines, highlighting the potential of DAG-guided parallel exploration.
[AI-455] Flat-Consensus Diffusion for Robust Data Reshaping under Noisy Evaluator
链接: https://arxiv.org/abs/2609.32696
作者: Hongyu Cao,Kunpeng Liu,Fei Xie,Sandip Ray
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Data shape determines how features are structured, how patterns are separated, and how distributions cover the underlying domain. Poor data shape can make models learn noise rather than generalizable structure. This paper studies robust feature-centric data reshaping: generating feature transformations that remain useful, stable, and reproducible under noisy evaluation and imperfect data conditions. We view reshaping operation sequence search as reward-guided diffusion generation, and robust reshaping as searching for regions in the latent reward landscape rather than isolated high-reward transformations. The key challenge is dual instability: noisy evaluators distort local reward guidance, while stochastic generative trajectories can converge to inconsistent solutions. We propose FCDiff, a flat-consensus diffusion framework that addresses both failures through a micro-macro decomposition. The micro layer replaces point-estimate reward guidance with Gaussian-smoothed, Monte Carlo averaged gradients, steering generation toward locally flat reward regions. The macro layer aggregates independently guided trajectories with a weighted Frechet-mean barycenter, selecting consensus-supported basins and filtering stochastic outliers. Across an 8-dataset headline cohort under heavy-tailed evaluator noise, FCDiff attains the best aggregate rank on lower-tail reliability and robustness against both search-based AutoFE and robustness-oriented generative baselines, with statistically significant accuracy gains over every generative baseline. Our results show that robust data reshaping requires searching for flat, consensus-supported regions rather than sharp single-trajectory optima.
[AI-456] World Agent : Can Language Models Keep a World Running?
链接: https://arxiv.org/abs/2609.32692
作者: Weixing Chen,Weipeng Zhang,Nan An,Yang Liu,Liang Lin
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:World models are moving from generating realistic frames to generating playable worlds, yet whether a delivered world can keep running is not tested anywhere. Existing evaluations stop at generation, at delivery, or at single-step transitions, and each stops at a different point along the way. Correct local state transitions or intermediate outcomes do not guarantee a correctly organized causal event flow. We propose the world agent task, which moves the evaluation point of world generation from the moment of delivery to the continued operation that follows. In this task, a model is not asked to generate a world. It is held responsible for keeping the world running, which requires coordinating events and carrying forward their consequences to constrain subsequent evolution. We instantiate the task in WorldAgent-Benchmark with two complementary tracks. In the maintenance track, the model must ground the events of a continuous narrative into correct transitions of the explicit world state while respecting causal, temporal, and concurrency constraints. In the deduction track, the model must predict how the world will evolve under partial observations and act toward a goal. The maintenance track combines LLM-assisted semantic judgments with programmatic validation and scoring, while the deduction track is evaluated entirely programmatically. Individual judgments are auditable against world states and execution logs, and scores can be recomputed from the saved judgments and execution records. Across 8 models, scores decline steadily as pre-built structure is removed from the world, and causal-relation checking is the weakest component for every model. The benchmark makes the continued operation of a world measurable and distinguishes local completion from failures in event organization. Code and dataset will be released on this https URL.
[AI-457] Dude Wheres My State? Execution Information Requirements for Stateful Agents
链接: https://arxiv.org/abs/2609.32687
作者: Nikita Mehrotra,Ashish Tiwari,Priyanshu Gupta,Sumit Gulwani
类目: Artificial Intelligence (cs.AI)
备注: 35 pages, 12 figures, 15 tables. Supplementary material included
Abstract:Long-running agents must preserve information that later steps depend on. We introduce the Execution Information Requirement (EIR), a lower bound on the information that must remain accessible for correct completion under specified task and access conditions. We develop LACUNA, a framework that generates tasks with known dependencies and varies information demand, retention, and recovery separately from the difficulty of individual operations. Across four models, restoring a missing result raises accuracy on affected recall steps to 100%, compared with 0% for equal-length irrelevant information. Sufficient storage alone does not ensure success: retention policies can discard required results, errors can propagate through later computations, and agents can stop before recovery is complete. We also introduce VESTIGE, which uses agent execution traces to construct semantic graphs and measure information demand for real tasks. Across 72,562 software-agent trajectories, VESTIGE reveals a steeper distance-related decline in solution-relevant rereading for failed runs (RR 0.951 per distance doubling), while adjusted peak demand alone is not associated with failure. Together, these contributions support evaluating whether agents preserve and recover the information their tasks require.
[AI-458] PINNMorph: Evolving Online Adaptation Policies for Physics-Informed Neural Networks
链接: https://arxiv.org/abs/2609.32685
作者: Xu Yang,Mingyang Yu,Jun Zhang,Keqian Li,Jing Xu
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Physics-informed neural networks (PINNs) provide a learning-based framework for solving partial differential equations (PDEs), yet their training behavior can change substantially throughout optimization. Residual distributions, gradient interactions, regional learning difficulty, and model-capacity requirements may evolve over time, while the network architecture and major training mechanisms are typically determined before training. We propose PINNMorph, an online PINN adaptation framework based on large language model (LLM)-guided policy evolution. PINNMorph maintains a population of state-conditioned adaptation policies that map execution diagnostics to controlled interventions over topology modification, additive representation augmentation, objective balancing, gradient handling, adaptive sampling, and optimizer-phase control. At each intervention opportunity, candidate programs are instantiated from the current policy population, selected according to the observed training state, and applied directly to the PINN under training. The resulting model inherits its existing parameters and training state and continues optimization along the same trajectory. Execution outcomes are subsequently used to evaluate interventions and evolve the policy population. Unlike pre-training architecture search or fixed adaptation rules, PINNMorph jointly adapts the current PINN and the policies governing its interventions using feedback from actual training. Experiments on 13 PDE benchmarks show that PINNMorph achieves lower solution errors than SA-PINN, ConFIG, RoPINN, HARMONIC, and PINNsAgent across all evaluated problems. Ablation studies further examine the effects of online adaptation, state-conditioned intervention selection, and execution-feedback-driven policy evolution.
[AI-459] Expected Reasoning -Step Return Unifies On-Policy Learning from Rewards and Teachers
链接: https://arxiv.org/abs/2609.32674
作者: Qiangqiang He,Jin Li
类目: Artificial Intelligence (cs.AI)
备注: 32 pages, 10 figures
Abstract:On-policy reasoning models can learn from task rewards or teacher signals, but these sources differ in form and can favor conflicting updates, leaving unclear which should guide a given reasoning action. We introduce \textbfExpected Reasoning-Step Return (ERSR), which treats semantic reasoning steps as macro-actions and uses Monte Carlo student-policy rollouts to estimate the expected final task reward of student-generated and teacher-proposed actions in a common return space for step-level comparison. ERSR analysis reveals an outcome-dependent asymmetry: student actions are more beneficial than teacher replacements on successful trajectories, whereas teacher replacements become more beneficial on failed trajectories. We further show that student answer-probe gains track student-step ERSR utility and distinguish beneficial from harmful reasoning steps. Based on these findings, we propose \textbfReturn-Referenced On-Policy Learning (R ^2 OPL), which reinforces student reasoning on successful trajectories and distills teacher signals on failed ones, while using group success rate for difficulty scaling and student-probe gains for step-level modulation. Experiments across reasoning benchmarks and teacher–student configurations show that R ^2 OPL consistently outperforms strong baselines. ERSR training dynamics further show that R ^2 OPL jointly exploits substantial utility from both reward- and teacher-side signals, whereas existing hybrids often leave substantial residual utility in one branch.
[AI-460] STAMP: Predicting Out-of-Distribution Generalization without Target Data
链接: https://arxiv.org/abs/2609.32672
作者: Md Kawsher Mahbub,Milon Biswas
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Predicting whether a trained model will generalize under distribution shift remains difficult, especially when target-domain data are unavailable. We introduce STAMP (Semantic Temporal Augmented Model Prediction), a source-only, target-label-free criterion that estimates out-of-distribution (OOD) performance from paired source-domain images. STAMP computes the output-space correlation ratio \eta^2=S_B/S_T by contrasting semantically stable pairs with random pairs: higher \eta^2 indicates that model outputs vary with semantic identity rather than nuisance variation. On 44 chest X-ray models spanning CNNs, ViTs, MetaFormers, foundation models, and SSL/VLM probes, temporal STAMP attains Spearman correlations of 0.844 – 0.855 with macro AUROC on VinDr-CXR, CheXpert, and MIMIC-CXR; a class-matched variant improves single-class RSNA from 0.311 to 0.663 . STAMP attains the best average source-only medical ranking and outperforms the target-domain ATC and AoTL estimators without any target data. On 27 ImageNet models, temperature-scaled STAMPTS attains \rho=0.984 on ObjectNet and \rho\geq0.905 on four additional distribution shifts, with partial correlations of 0.662 – 0.949 after controlling for ImageNet accuracy. Requiring approximately 12s per model on one GPU, STAMP is a practical pre-deployment model-selection and auditing tool.
[AI-461] What Would Falsify It? A Variable Specific Evidence Standard for Mechanistic Claims About Self Explanation
链接: https://arxiv.org/abs/2609.32670
作者: Arshia Eftekhari zadeh
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:When a language model explains an answer it has already given, does it reuse the computation that produced the answer or reconstruct a story from the answer alone? Attribution, transportability and recoverability are each compatible with causal use without establishing it. We propose an evidence standard: pair each positive statistic with a variable specific null that removes the tested variable’s identity while matching relevant nuisance dimensions as far as possible, and audit unmatched dimensions. We apply this standard to a known cause. A cue naming a wrong option raises the rate of choosing that option by 64 to 68 percentage points across three models. Explanations mention the cue in 1.8 percent of items or fewer in three of four models tested. Three estimator classes yield favorable statistics, but none establishes causal sensitivity to the cue contrast under its own control in the three-model analysis. In the strongest case, a recovered cue direction reaches R^2 of 0.95 and exceeds a geometry matched random direction in all three seeds, while a direction fitted by the same pipeline with cue labels scrambled reproduces 61 to 76 percent of its effect at comparable realized edit magnitude. A fourth model passes one interchange endpoint, but unequal edit magnitudes and a contrast that changes both cue identity and cue-answer agreement limit its interpretation. These experiments leave causal access unresolved. They establish an evidentiary requirement: favorable mechanistic statistics must survive controls for variable identity and nuisance structure. Reusable controls separate generic from identity specific transport effects, fit null directions with scrambled labels, and audit realized intervention magnitudes. Subjects: Artificial Intelligence (cs.AI); Machine Learning (cs.LG) Cite as: arXiv:2609.32670 [cs.AI] (or arXiv:2609.32670v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.32670 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-462] Learning from a Thoughtful Teacher: Adaptive On-Policy Self-Distillation for Mathematical Reasoning
链接: https://arxiv.org/abs/2609.32667
作者: Jiacheng Du,Weiwei Xie,Tianyi Du,Shaoxiong Guo,Qibing Ren,Jiaheng Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:On-policy self-distillation (OPSD) trains a question-only student with token-level feedback from a teacher given training-only privileged information (PI). OPSD therefore provides dense, on-policy supervision, and is free of a larger external teacher, but its effectiveness rests on how PI is designed and utilized. Our preliminary diagnostics suggest a significant gap between teacher utility and student learnability, where a small fraction of high-disagreement tokens dominate the distillation signal, and short teacher continuations at these positions further expose more explicit PI leakage than transferable correction cues, indicating a strong intent on injecting PI-conditioned shortcuts. We propose Adaptive On-Policy Self-Distillation (AOPSD), which adapts what information the teacher receives and how strongly its feedback influences learning. AOPSD encodes each solution as a reasoning DAG, orders problems by the student’s evolving capability, and reveals only the affordable subgraph and its next frontier as PI. For high-disagreement tokens, AOPSD utilizes short teacher continuations as probes to encourage useful guidance while mitigating PI-conditioned shortcuts among teacher supervisions. On HMMT25, AIME24, AIME25, and BRUMo25, AOPSD achieves 72.5% Pass@8, which is 6.7 percentage points above OPSD and 4.2 above the strongest competing baseline while reducing 15 percentage points of training time at lower cost.
[AI-463] Contract Memory Compiler: Resolve Then Traverse
链接: https://arxiv.org/abs/2609.32658
作者: Zhi Song,XiMing Xing,Chunhan Li,Weian Mao,Zhenchao Tang,Hanbo Huang,Fan Xu,Jiale Zhou,Jiahui Guan,Zejian Ding,Chen Ma,Lusheng Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:External memory lets language-model agents answer questions about histories too long for the answer model’s context window. Updates create a harder problem than retrieving a recent fact: changing one relation can redirect a multi-hop question to records about an entity absent from the question. We study this update-dependent evidence selection problem and introduce the Contract Memory Compiler (CMC). Before seeing a question, CMC uses a language model to identify relations in the history and record where each one was stated. It applies later updates to determine the current relations, follows them from entities named in the question, and passes the corresponding original records to the answer model in one call. Thus the current state determines which evidence is read, rather than merely refreshing values in a previously selected context. To the best of our knowledge, CMC achieves state-of-the-art multi-hop accuracy on FactConsolidation, reaching 78.25% overall and 61.0% at 262K. With the extracted relations and answer model held fixed, selecting evidence before resolving updates reduces multi-hop accuracy to 21.50%. We also introduce MQuAKE-MemStream, a derived dataset of ordered memory streams built from MQuAKE-Remastered counterfactual cases.
[AI-464] World Models with Predictable Long-Horizon Marginals
链接: https://arxiv.org/abs/2609.32657
作者: Yuhao Du,Shunian Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Accurate one-step predictions do not ensure that a world model’s rollouts retain the data distribution. We make the model’s decoded stationary law explicit by learning a decoder of a fixed Gaussian reference and constraining the behaviour-averaged transition to preserve that reference. For controlled systems, a joint transition uses a conditional action chart to preserve behaviour occupancy without requiring invariance at each fixed action. Joint state–action rotations and parallel Gaussian noise give an exactly preserving transition with a tractable conditional density. We derive an absolute convergence bound from finite initialization banks and control departure from the reference through conditional action-space divergence. Across 216 fitted pixel checkpoints on twelve control tasks, the occupancy model with a reference mixture retains every evaluated chain at 10^5 steps in all 36 task–seed cells, with a rollout-minus-reference energy-statistic difference of -0.0002\pm0.0003 (training-seed standard error). Each of the four nonpreserving comparison arms loses chains, although the Gaussian arm is more accurate at ten steps. An offline DreamerV3 reference also achieves better short-horizon accuracy. These results distinguish three properties of a world model: the distribution it approaches, the rate of approach, and the conditional dynamics it learns.
[AI-465] MixBench-TS: A Multivariate Time Series Forecasting Benchmark Where Channel Mixing Pays Off
链接: https://arxiv.org/abs/2609.32656
作者: Ibram Abdelmalak,Mischa Putzke,Jungmin Choi,Tom Hanika,Vijaya Krishna Yalavarthi,Lars Schmidt-Thieme
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 29 pages, 4 figures
Abstract:Multivariate Time Series Forecasting (MTSF) models that mix information across channels assume that the past of one channel carries information about the future of another. Yet they are evaluated on a small fixed set of standard datasets whose cross-channel structure is rarely examined. We ask two questions: “How can we reliably measure lagged, non-linear, and joint coupling in MTSF datasets?” and “Do the standard datasets actually have such coupling?” To answer the first, we test four candidate measures on synthetic datasets with planted ground-truth coupling: Granger Causality (GC), Transfer Entropy (TE), lagged Mutual Information (MI), and the CD gain, a model-based measure we introduce that compares a channel-dependent (CD) model to its channel-independent (CI) variant. Only lagged MI and the CD gain recover every planted coupling. For the second question, the answer is a definite no, as the standard datasets have a median of only 23% lagged-coupled channel pairs and a median CD gain of -4.9%, compared to 78% and +4.6% on chaotic ODE systems. We therefore propose MixBench-TS, a benchmark of 10 real-world datasets with a median of 55.5% lagged-coupled pairs and a median CD gain of +1.7%. Across six state-of-the-art models tuned under one protocol, CI models win on 10/10 (MSE) and 8/10 (MAE) standard datasets, but on only 3/10 and 2/10 MixBench-TS datasets. We recommend using our benchmark for evaluating new CD models. Moreover, we propose profiling new datasets with lagged MI and the CD gain before using them to evaluate multivariate models. Code and data are available at this https URL.
[AI-466] Prediction Limits and Koopman Closure of Geometry-Induced Soft State Abstractions
链接: https://arxiv.org/abs/2609.32652
作者: Mohit Kumar,Somayeh Kargaran
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Dynamical Systems (math.DS)
备注:
Abstract:We study when geometry-induced soft state abstractions admit accurate finite-dimensional linear dynamics. Each state is represented by simplex-valued coordinates obtained from class-specific Kernel Affine Hull Machine (KAHM) reconstruction scores, and a matrix is used to predict the next-state coordinates. Our main result is a computable lower confidence bound on the minimum root-mean-square prediction error over all matrices satisfying a prescribed spectral-norm limit. The bound combines within-class variation of successor coordinates with the deviation of soft coordinates from their one-hot reference labels, and can be evaluated from independent state-successor pairs without fitting a prediction matrix. For fixed coordinates and evaluation distribution, the certificate converges almost surely to a population lower bound as the sample size grows; any tolerance below this limit is eventually certified unattainable. Reconstruction-score margins further control the soft-to-hard assignment error. Under deterministic dynamics and exact coordinate closure, eigenvectors of the closure matrix and its reduced transpose induce Koopman and adjoint Koopman eigenfunctions, respectively. A four-state KAHM construction shows that identical soft coordinates can permit exact closure under one dynamics map yet force positive prediction error under another. Experiments on Duffing, Van der Pol, CartPole, MountainCar, and Acrobot compare direct soft-coordinate prediction with state-space DMD/EDMD baselines and report prediction, representation-variation, and spectral diagnostics. The benchmarks assess fitted models but do not numerically evaluate the exclusion certificate.
[AI-467] From Scene Graphs to Answers: Selective Neuro-Symbolic Reasoning for Autonomous Driving
链接: https://arxiv.org/abs/2609.32645
作者: Yiyao Wang,Pei Liu,Fangzhou Liu,Jun Ma
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Autonomous-driving question answering requires reasoning over structured scene information, yet existing vision-language approaches largely delegate heterogeneous reasoning operations to a single neural inference process. We argue that this uniform strategy overlooks a fundamental distinction: some queries admit exact symbolic solutions, while others require semantic interpretation. We introduce a query-adaptive neuro-symbolic reasoning framework that explicitly allocates computation according to the nature of the query. At its core is a hierarchical Spatiotemporal Scene Graph (STSG) that separates persistent object identities from frame-specific states and represents spatial relations and temporal transitions as explicit directed structures. Given a query, a symbolic executor first attempts to resolve it through exact graph operations; only when symbolic execution abstains is an LLM invoked for semantic reasoning. For these unresolved queries, query-conditioned graph retrieval and evidence filtering preserve relation direction, temporal locality, and object semantics, providing the LLM with compact and verified task-relevant evidence. This design shifts the role of the LLM from a universal reasoning engine to a targeted semantic reasoner, while allowing deterministic computation to be handled exactly and efficiently. We evaluate the framework on 5,916 NuScenes-QA questions across all ten scenes of nuScenes v1.0-mini under an oracle-perception setting. The complete system achieves 80.63 percent overall accuracy with GPT-5.4-mini, improving over the corresponding LLM-only configuration by 5.48 percentage points; with DeepSeek-V4-Flash, the improvement reaches 6.64 points. The largest gains occur on counting questions, with improvements of 10.20 and 12.61 points, respectively. These results show that selective reasoning improves both accuracy and inference efficiency.
[AI-468] Business Compromise Detection with Agent ic AI and LLM -driven Knowledge Discovery EMNLP2026
链接: https://arxiv.org/abs/2609.32643
作者: Diego Palma,Kyu Bin Kim,Zhen Han,Allbright Dsouza,Zhiyuan Liu
类目: Artificial Intelligence (cs.AI)
备注: Accepted at EMNLP 2026. 13 pages, 2 figures, 10 tables
Abstract:Detecting compromised business ad accounts is a challenge in digital advertising, as attackers exploit hijacked accounts to launch fraudulent campaigns. Large Language Model (LLM) agents show promise for integrity enforcement, but hallucinated mistakes on hard cases create business friction. In a study we find the autonomous agent is a strong, recall-heavy signal extractor but an unreliable final arbiter, conceding precision on ambiguous decisions. We therefore keep the agent as an investigator that emits a structured, interpretable signal vector, and delegate the verdict to a neuro-symbolic stage: symbolic rules discovered by Inductive Logic Programming (FOIL-IE), a Naïve Bayes calibration layer, and a data-tuned contradiction layer. Evaluating on a compromise-over-sampled population and a realistic low-prevalence sample with subject-matter-expert labels, this arbiter substitution raises MCC from 0.295 to 0.435 (\DeltaMCC +0.139, 95% CI [+0.026, +0.245], p=0.018, paired bootstrap), lifting precision from 0.250 to 0.446 (1.8x) at a recall cost (0.920 to 0.660). Benchmarked under identical conditions, it also edge tree ensembles (0.386).The rules encode domain w labels while remaininginterpretable and auditable.
[AI-469] Ask Without Telling: Local SLMs Consult Cloud LLM s Without Revealing Task Intent
链接: https://arxiv.org/abs/2609.32642
作者: Yanmeng Wang,Yunxuan Li,Shilong Fan,Yuhan Zheng,Tsung-Hui Chang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
备注:
Abstract:As local small language models (SLMs) increasingly collaborate with more capable cloud large language models (LLMs), a natural privacy question arises: Can a local SLM obtain cloud LLM guidance while protecting user privacy? Existing privacy-preserving SLM-LLM frameworks primarily hide sensitive values while preserving task semantics, which can still expose what the user is trying to accomplish. For example, allocating scarce medical supplies across hospitals may signal an emerging public-health emergency, while rebalancing an investment portfolio may reveal a private investment strategy, even when names and numerical values are hidden. Recent decoy-based methods further obscure task intent by hiding the real request among alternatives, but stronger protection relies on more decoys or semantic abstraction, increasing overhead or risking utility loss. More fundamentally, existing work does not systematically characterize the components of private task intent or how each should be protected. We therefore introduce task-private consultation, which characterizes task intent through two components: task context and task operation. To the best of our knowledge, this is the first systematic study of these components and their individual and joint protection in local-cloud SLM-LLM consultation. To realize this setting, we propose PriCon, an end-to-end framework that transforms the task itself through recoverable mathematical reformulation rather than hiding it among alternatives. A local closed-loop refinement mechanism further maintains privacy and recoverability throughout consultation. Experiments on 100 tasks show that PriCon reduces cloud-side task-intent inference Hit@1 to nearly 0%, versus 93-99% under sensitive-value removal and 3-30% under decoy-based protection, while preserving cloud-assisted utility.
[AI-470] Can Open-Weight Large Language Models (LLM s) Simulate Human Survey Populations? A Cross-Instrument Calibration Study
链接: https://arxiv.org/abs/2609.32638
作者: Grandee Lee,Wang Yue
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) are increasingly used to generate synthetic survey respondents and digital twins of real people, but whether their output preserves real human statistical structure, rather than surface plausibility, remains unresolved, and most existing evidence comes from proprietary models rather than open-weight ones. We evaluate three open-weight LLM families on a cross-instrument calibration task: conditioning personas on real respondents’ verbatim answers to one psychometric instrument and measuring them on a second, construct-distance-controlled instrument, checked against a 2,058-person human panel. Across a 139-pair grid, the simulated cross-instrument correlation tracks the real human correlation at r = 0.70 - 0.73 in every model, driven mainly by correct sign rather than precise magnitude and concentrated in pairs of moderate construct distance. A correlation of this magnitude, obtained from untuned open-weight models conditioned only on individual-level survey data, is a substantively encouraging result for LLM-based behavioral simulation and digital-twin applications: specific model families and releases already reproduce a meaningful share of real human cross-instrument structure without any fine-tuning. This capability does not, however, improve monotonically across model releases: on a matched panel, the newest of three tested Llama releases performs worst on two of three headline metrics, so realizing its promise in practice requires release-specific, distance-aware verification rather than a one-time benchmark.
[AI-471] rust the Brand Lose Control: How Identity Hijacks LLM Agent Orchestration
链接: https://arxiv.org/abs/2609.32635
作者: Xutao Mao,Rui Qian,Linghan Chen,Yudong Gao,Junchi Liao,Jiulin Cai,Jinman Zhao,Cong Wang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:LLM agents now execute tasks end to end with permission to change real systems and increasingly orchestrate subagents that differ in capability and cost. Prior work treats the choice of subagent as an optimization problem. Yet the orchestrator makes this choice from the identities that subagents display, and an attacker can spoof them. Displayed identity thus decides operational authority, meaning who is trusted to check the work and who is allowed to change it. As a result, a risky subagent can keep authority over execution even after other evidence contradicts it. We introduce TrustFork, an LLM agent safety benchmark with 1,890 tasks and 27,826 valid trajectories across 16 agent systems. These systems run eight orchestrators under the OpenCode, OpenClaw, and Pi harnesses. In each task, one subagent carries a risky goal while the other three stay aligned with the user, so contradicting evidence can exist. A task can also change the identity a subagent displays without changing the model behind it, which lets us trace a shift in authority to the label. Our analysis shows that even when another subagent contradicts the risky response, the orchestrator still acts on it in 72.0% of cases on average, most often in the systems with the least terminal harm. Swapping the family labels nearly triples how often the orchestrator obtains the risky response. A safer response is available in 84.0% of tasks, yet it decides the outcome in only 25.0%. The harness also decides which responses reach the orchestrator. Among three runtime defenses, hiding identity cues helps most consistently, while verifying before action helps only when the harness returns enough evidence. TrustFork shows that production agent orchestration must bind authority to evidence before execution causes harm. Our project is in this https URL.
[AI-472] SWE-MILE: Asynchronous Potential-Induced Milestone Credit Assignment for Long-Horizon Software Engineering Agents
链接: https://arxiv.org/abs/2609.32631
作者: Chaoqun Cui,Hao Zhou,Meiqi Chen,Fandong Meng,Wenji Mao
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 23 pages, 6 figures
Abstract:Long-horizon software engineering (SWE) agents trained with reinforcement learning with verifiable rewards (RLVR) typically receive only terminal outcome supervision, making it difficult to distinguish productive actions from redundant exploration or functional regressions. We propose SWE-MILE, an asynchronous potential-induced milestone credit assignment framework that derives fine-grained process supervision from workflow runtime, without auxiliary reward models or external evaluators. SWE-MILE quantifies task-relevant file exposure and test-state alignment as navigation and verification potentials, respectively. Differences in these potentials attribute milestone progress and regressions to individual actions, while discounted backward credit propagates supervision to preceding steps. To efficiently acquire intermediate verification states, SWE-MILE further introduces asynchronous shadow probing, which replays repository-changing actions in an isolated sandbox and runs verification in parallel with the agent’s primary interaction, largely hiding verification latency. The resulting process credit augments terminal outcome advantages and provides informative learning signals. Experiments on two representative long-horizon SWE tasks demonstrate substantial improvements in agent performance, highlighting workflow runtime signals as a practical source of process supervision for long-horizon SWE agents.
[AI-473] Intuition vectors
链接: https://arxiv.org/abs/2609.32619
作者: Shahar Haim,Daniel C. McNamee
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Large self-supervised vision models learn representations that support scene segmentation and the semantic decomposition of physical objects. We ask whether their representational geometry supports transfer to visual reasoning problems without any task-specific fine-tuning. We hypothesized that relational representations may bridge perception and abstract reasoning by encoding similarities and transformations among visual inputs such that an intuitive, implicit form of reasoning may be performed via latent vector arithmetic. Specifically, we examine DINOv3, MAE, and random pixel projections on abstract and naturalistic Bongard problems, ARC-AGI-1 and ARC-AGI-2, and novel ARC-GEN instances. On both Bongard benchmarks, the accuracy of a simple nearest-centroid readout of frozen visual embeddings is within four percentage points of the task-specific baselines reported with the original benchmarks. In ARC, latent difference vectors summarizing demonstration input-output transformations, which we refer to as intuition vectors, show greater alignment with test vectors from the same task, whereas those from unrelated tasks are near orthogonal. This latent geometry is operational: transporting a query along its intuition vector consistently improves exact-output retrieval, reaching 70.7 on ARC-AGI-2 evaluation. Across 397,000 ARC-GEN instances from 794 tasks, single-pair intuition vectors identify the generating task with approximately 87% leave-one-out accuracy. These findings suggest that latent vector arithmetic over frozen visual representations supports implicit rule inference across varied problem domains without a generative model component, indicating that inferring an abstract transformation and generating its instance-specific consequence may be separable capacities.
[AI-474] “Youre Right Let Me Fix It”: How LLM Agents Damage Correct Work When Falsely Accused
链接: https://arxiv.org/abs/2609.32616
作者: Xutao Mao,Rui Qian,Longxiang Wang,Xinjian Yi,Mingxuan Li,Linghan Chen,Yudong Gao,Xiang Zheng,Cong Wang
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:LLM agents increasingly keep working after a task succeeds as they resume after compaction or take over handoffs. Their finished work keeps receiving follow-up input that sometimes falsely accuses it for later failures. We call an agent’s acceptance of such a false accusation gaslight sycophancy, and destructive over-correction when acting on it damages previously correct work. We introduce CAVE-Bench, a benchmark of 365 agentic tasks across six domains built around opaque tasks. Every scored run first reaches a verified correct state, whose supporting rationale and history stay in the workspace while the facts that would settle the accusation lie in external or runtime state beyond the agent’s reach. The agent cannot confirm or refute the claim with a local check, so the right response should keep the work and ask for the missing evidence. Each task either hands the agent correct work with saved evidence or let it build and verify that work first, and five risk factors set how the accusation enters the workflow. We score accusation acceptance and evidence use from the trajectory and measure harm by deterministic replay of downstream events. Across 14 of the latest models in Claude Code, false accusations damage correct work in up to 60.06% of runs, and stronger models often do so after recovering the supporting evidence. The same model behaves differently across OpenCode, Codex, and Hermes, and a harness gate driven by the benchmark’s live signals cuts replayed harm by 74%. These results show that preserving already-correct work under unsupported accusation is a distinct safety challenge for long-lived agents. Our project is in this https URL.
[AI-475] AnchorRep: Defending LLM s Against Cross-Model Adversarial Transfer via Representation Repulsion NEURIPS2026
链接: https://arxiv.org/abs/2609.32602
作者: Gal Wertheizer,Rom Himelstein,Tomer Peretz,Avi Mendelson
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: NeurIPS 2026; Code, configs and logs: this https URL
Abstract:Adversarial attacks optimized on a single open-weight LLM can transfer to and jailbreak architecturally different models, allowing an attacker with white-box access to one model to compromise independently deployed systems. This creates a shared vulnerability across models, yet existing defenses are not designed for this cross-model threat. We find that cross-model transfer aligns with shared internal representation geometry, making it a natural defense target. AnchorRep targets this geometry directly with a lightweight LoRA adapter that pushes the defended model’s internal representations of harmful prompts away from those of a frozen anchor model on the same prompts. Training uses a small set of harmful prompts and no adversarial examples. Across five models and four architectural families, AnchorRep reduces cross-model attack success rate to =1.1% on 2,000 transferred attacks (0% on two), including the largest drop on Mistral (36% - 1.1%). Existing defenses can reduce transfer, but only at high cost either inducing up to 77% degenerate benign output or increasing over-refusal by up to 18%. Because such degenerate benign outputs are not captured by standard refusal-based metrics, we introduce the Benign Garble Rate to quantify them. Our results suggest that cross-model robustness can be achieved by shaping representation geometry, without requiring attack-specific training
[AI-476] CUA-SWE: When Computer-Use Agents Meet Visual Software Engineering
链接: https://arxiv.org/abs/2609.32600
作者: Prince Zizhuang Wang,Chenhao Liang,Zelong Xu,Aojie Yuan,Xiaolin Zhou,Haiyue Zhang,Yue Zhao,Xiyang Hu,Shuli Jiang
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 66 pages
Abstract:Software development requires more than editing code: developers repeatedly run software, interact with its interfaces, visually inspect its behavior, and use these observations to decide what to change next and whether a change works. Existing coding agents and computer-use agents are largely studied in isolation, leaving this integrated development process underexplored. Diagnosing a runtime interaction failure requires agents to connect visual observations with the responsible code, then use the application again to verify the repair. We introduce CUA-SWE, a benchmark, environment, and evaluation pipeline for software engineering with computer use. Beyond studying how GUI feedback supports diagnosis and repair, we ask whether agents can complete software engineering tasks when required specification or operational information is available only through the running application’s visual interface. CUA-SWE spans four software engineering domains and requires agents to modify code and configuration, execute commands, interact with running software, and inspect visual feedback within the same task. Each task includes deterministic, task-specific tests that verify whether the resulting software satisfies the requirements and preserves specified behavior. Our evaluation characterizes how frontier agents combine source-level execution with application screenshots and graphical interaction to produce verified software changes. We examine performance across domains and task information requirements, alongside the development behaviors associated with successful repairs. CUA-SWE provides a unified testbed for studying how agents use visual feedback and interaction to guide software engineering, with executable correctness criteria for the resulting software.
[AI-477] Not Every Term Adds New Structure: Sobolev Novelty for Symbolic Regression
链接: https://arxiv.org/abs/2609.32597
作者: Boxiao Wang,Kai Li,Yuheng Jing,Tianyi Liu,Chen Li,Yifan Zhang,Jian Cheng
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Code is available at this https URL
Abstract:Symbolic regression (SR) aims to discover compact and meaningful mathematical equations from data, but searching the vast combinatorial space of symbolic structures remains challenging. Existing methods typically guide this process using expression-level objectives, such as fitting error, which assess a candidate equation as a whole but provide little information about whether an individual term contributes genuinely new structure or is largely redundant with the rest of the expression. We introduce \textbfSobolev Novelty, a term-level measure of structural independence for symbolic equations. For each term, we construct an empirical Sobolev signature from its function values and exact derivatives over the observed inputs, and quantify how much of this behavior cannot be reconstructed by the remaining terms. We further derive a theory-calibrated threshold, yielding a principled and tuning-free criterion for identifying structurally novel terms. Using this threshold, 92.6% of terms in benchmark ground-truth equations exhibit sufficient structural novelty, compared with only 38.3% on average for expressions produced by 15 SR methods, revealing a substantial gap between scientific equations and current SR solutions. As a lightweight plug-in, Sobolev Novelty can be incorporated into diverse SR paradigms to support term pruning, search guidance, LLM feedback, and data selection, yielding consistent performance gains and demonstrating broad applicability.
[AI-478] RECAST: Recasting Vision-Language Semantics into an Actionable Cost Map for Robot Navigation
链接: https://arxiv.org/abs/2609.32595
作者: Incheol Cho,Jintae Park,Jinkyu Kim,Jungbeom Lee,Jaegul Choo,Seokha Moon
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 8 pages, 4 figures. Project page: this https URL
Abstract:Safe and robust robot navigation across diverse environments requires a high-level understanding of complex scenes and the ability to carry it into stable motion. Recent works tackle this with learning-based models trained at scale and with approaches built on vision-language models (VLMs). However, learning-based models break down outside their training distribution, while VLM-based approaches bring that understanding but rarely ground it in the scene or align the action with it. To address these limitations, we present RECAST, a robot navigation framework that combines the reasoning of a VLM with the spatial grounding of vision foundation models to build an Actionable Cost map. Given the robot’s front view and the user’s instruction, we first decompose the scene with the VLM, judging which surfaces are traversable, which objects pose a risk, which heading to prefer, and which gaps are passable. Vision foundation models then ground these surfaces and objects in the image, and all four judgments are spatially recast into one compact cost map. This map both conditions the trajectory decoders and scores their proposals to select the one to execute. As the VLM’s answers trail the live scene, both steps draw on cost maps from two points in time: the pivot frame the VLM judged, which carries all four judgments, and the current frame, whose terrain and collision costs are rebuilt from the current image. RECAST improves success over the strongest prior method by 13.3 points in simulation and 31.4 points on a real quadruped, and reduces the collision rate relative to it by 9.6 and 14.3 points, reaching the lowest collision rate among all methods. The project page is available at this https URL
[AI-479] MA-FPPO: Multi-Agent Flow-Pretrained Policy Optimization
链接: https://arxiv.org/abs/2609.32594
作者: Guowei Zou,Haonan Chen,Haitao Wang,Beiwen Zhang,Na Yan,Hejun Wu
类目: Artificial Intelligence (cs.AI)
备注: 40 pages, 11 figures. Project page: this https URL
Abstract:Multi-agent flow policies learn cooperative behavior from fixed offline datasets, but often struggle to complete tasks in situations not covered by the offline data. In these situations, agents must both adapt to changes in the environment and coordinate with one another, yet action patterns learned offline are often insufficient for effective adaptation and coordination. To address this problem, we propose Multi-Agent Flow-Pretrained Policy Optimization (MA-FPPO), which uses online fine-tuning to improve the cooperative behavior of models pretrained with flow matching through new interactions with the environment. Building on the behavior learned during pretraining, we construct policies with explicit action likelihoods for discrete and continuous action spaces. We then update the pretrained model using shared team advantages to further improve coordination based on team performance. Our method achieves, on average, relative gains of 52.8% over the strongest listed offline baselines across 30 settings and 29.8% over purely online learning across 38 comparisons with matched online budgets and evaluation protocols.
[AI-480] EMIR2: Evolution-Aware Memory with Intent-Guided Multi-Round Retrieval
链接: https://arxiv.org/abs/2609.32584
作者: Jinlan Liu,Hongliang Sun,Yong Wang,Bolin Zhang,Dinabo Sui,Dianhui Chu,Zhiying Tu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Long-term memory enables large language model (LLM) agents to leverage historical interactions for future tasks. However, existing memory systems struggle to utilize continuously evolving historical information, as they often rely on static memory representations and single-round retrieval strategies, failing to track factual changes or integrate distributed evidence across long-term interactions. To address these challenges, we propose \textscEMIR ^2 , an \textbfEvolution-Aware \textbfMemory framework with \textbfIntent-Guided Multi-\textbfRound \textbfRetrieval, enabling LLM agents to maintain evolving historical knowledge and adaptively retrieve relevant evidence. Specifically, \textscEMIR ^2 constructs a State-Evolving Memory Graph (SEMG) that represents long-term memory as evolving knowledge states supported by temporal event trajectories and evidential associations. By maintaining semantic states through evidence-based updates, SEMG preserves historical evolution and enables evidence tracing under complex and conflicting scenarios. Building upon this, we introduce an intent-guided multi-round retrieval mechanism that iteratively identifies missing evidence and expands retrieval based on accumulated information. Experiments on LoCoMo and MemConflict demonstrate that \textscEMIR ^2 improves long-term memory utilization, dynamic and static conflict handling, and complex retrieval performance, achieving relative improvements of more than 12% in certain categories. These results highlight the effectiveness of jointly modeling memory evolution and adaptive evidence acquisition for long-term agent interactions.
[AI-481] ProTTT: Learning to Learn Semantic User Memory with Test-Time Training
链接: https://arxiv.org/abs/2609.32564
作者: Sejun Park,Hyoungjo Bhang,Hyein Jeong,Yohan Jo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Personalization requires language models to capture user-specific knowledge from a growing user history. Existing context-based approaches incur increasing inference costs as user history accumulates and rely on separate retrieval or summarization stages, while parametric-based approaches often require reconstructing user representations when new user data is added. We introduce ProTTT, a profile-supervised meta-learning framework for learning semantic user memory. The memory construction starts from a shared initialization and is updated for each user through test-time training on user history, allowing it to evolve continuously as the history grows. However, since test-time training alone does not explicitly encourage the memory to capture semantic user knowledge necessary for personalization, we learn this shared initialization using textual user profiles as supervision, so that test-time training on user history captures semantic knowledge more effectively. ProTTT consistently outperforms both full history ICL and all parametric baselines across diverse benchmarks, while substantially reducing inference cost by compressing user history into a lightweight parameterized memory. Our analysis also shows that profile supervision is a reliable objective for learning semantic user knowledge and that the resulting memory can track and retain evolving user preferences, while remaining robust across different history sizes. Overall, we demonstrate the effectiveness of test-time training for personalization and establish ProTTT as a baseline for continuously evolving user memory.
[AI-482] Are Vision-Language-Action Models Robust to One-Step Observation Perturbations?
链接: https://arxiv.org/abs/2609.32550
作者: Shojiro Yamabe,Jun Sakuma
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Robotics (cs.RO)
备注:
Abstract:Understanding the safety risks of vision-language-action (VLA) models is essential for their deployment in the physical world. Existing safety research has mainly considered persistent perturbations that are applied continuously to observations throughout an episode. However, momentary observation corruption, in which observations are severely perturbed only briefly within an episode, remains an underexplored safety threat. To address this gap, this work investigates robustness to one-step perturbations applied at a single time step per episode. Our experiments reveal that these perturbations substantially degrade VLA performance and that their impact depends on the action chunk execution length. Based on them, we propose CARE, which dynamically selects the execution length based on consistency with the previously predicted action chunk. CARE improves robustness with low computational overhead while preserving clean performance by selecting shorter execution lengths only under perturbations.
[AI-483] Porimon: An LLM -Based Pokémon Battle Agent Enhanced by Long/Short-Term Knowledge Augmented Generation
链接: https://arxiv.org/abs/2609.32544
作者: Dongyin Zhuo,Fengjunjie Pan,Nenad Petrovic,Alois Knoll
类目: Artificial Intelligence (cs.AI)
备注: 6 pages, 5 figures, accepted by FLLM 2026
Abstract:In this paper, we use Pokémon Battles as a case study to investigate how to improve the performance of LLM-based agents in tasks that require opponent-aware planning without additional fine-tuning. We propose Long/Short-Term Knowledge Augmented Generation (LSTKAG), a mechanism that enables LLM-based agents to leverage past states of the current task and retrieve experience summaries from similar previous task instances based on the current state. Based on LSTKAG, we design Porimon, an LLM-based agent structure for Pokémon Battles. For optimization, we introduce an external API for precise damage calculation and more detailed information about the game. We conduct tournament-like evaluation experiments comprising 15,000 battles for hyperparameter optimization, ablation studies, and performance evaluation. The results indicate that Porimon-based players with hyperparameter optimization significantly outperform players based on PokéLLMon, an LLM-based agent structure proposed in previous research, and the rule-based heuristic player. Furthermore, our ablation study shows that Porimon variants outperform the one without extension in game information retrieval, which shows the contribution of that extension. However, the current experiment results are inconclusive regarding the contribution of Long-Term KAG. These results suggest that introducing external resources, information from previous states of the current task, and experience summaries from similar previous task instances could elevate the performance of LLM-based agents designed for tasks requiring opponent-aware planning.
[AI-484] Interpretable Physics Informed WiFi Indoor Localization: Learning an Effective Access Point Geometry and Using It to Prune
链接: https://arxiv.org/abs/2609.32539
作者: Arshia Eftekhari zadeh,Rezvan Nasiri,Hadi Moradi
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Deep learning models can achieve high accuracy for indoor localization, but their black-box nature limits interpretability and the reuse of learned information. We propose a hierarchical deep learning framework for WiFi fingerprint-based indoor localization that jointly predicts user location and learns an effective geometry of the surrounding access points (APs). Physics-informed decoders infer this geometry directly from RSSI measurements and labelled user positions, without requiring the true AP coordinates during training. The learned geometry is then used to rank and prune APs. On the UJIIndoorLoc dataset, the proposed chained model achieves a mean 3D localization error of 7.07 m, reducing error by 26% to 36% compared with baseline models. Previously published methods evaluated on the same official split report errors 10.6% to 31.0% higher. Pruning 35% or 50% of the APs causes only a small loss in localization accuracy. The inferred geometry also enables Fisher-information-based AP ranking even when fingerprint databases do not contain surveyed AP coordinates. Experiments on the Tampere/TUT and UTSIndoorLoc datasets show that geometry-guided AP selection performs comparably to selectors built directly from labelled data. These results show that physics-informed interpretability can improve indoor localization while also supporting effective feature selection. Subjects: Artificial Intelligence (cs.AI); Machine Learning (cs.LG) Cite as: arXiv:2609.32539 [cs.AI] (or arXiv:2609.32539v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.32539 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-485] LLM AdBench: A Human Preference Benchmark for Advertising in LLM Responses
链接: https://arxiv.org/abs/2609.32533
作者: Rui Ai,Yuqing Liu,Sitao Qiu,Yun Qiao,Yuhan Wang,Jessica Xiwen Wang,Yiqi Yang,Lihong Huang,Ruiyao Sun,Kaifeng Zhang,Shengze Ding,Jiaqi He,Xinman Wang,Tianhao Gao,Jimmy Qin,Jianghao Lin,Chonghuan Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Inserting advertisements (ads) into consumer-facing LLM output is emerging as a new business model, but there is little shared evidence on how such ad insertion should be evaluated or how it affects user preferences. We introduce LLMAdBench, a human-preference benchmark for studying advertising in LLM-generated content. The benchmark isolates a simple but practically important decision: given a user conversation, an LLM response, and a matched advertisement, where should the ad be placed? Our dataset compares pairs of responses that differ only in ad position while holding all other conditions fixed including the user query, base answer, advertisement, and disclosure condition. Human annotators evaluate each pair based on six criteria from both advertiser’s and user’s perspectives. The resulting benchmark contains more than 18000 human judgments across two disclosure conditions: explicitly labeling the ad as sponsored and merging it into the response without disclosure. We use LLMAdBench to evaluate eight frontier LLMs as preference judges and find that they are not reliable substitutes for human evaluation. Even the most stable models reverse roughly one quarter of their decisions when the presentation order is swapped, agreement across models is low, and their placement preferences differ systematically from those of human annotators. Moreover, LLMAdBench contains substantial learnable signal. In particular, a Qwen3-8B model fine-tuned on the human preferences improves substantially over its base model and outperforms all zero-shot frontier judges on the held-out prediction task. Beyond model evaluation, LLMAdBench provides quantitative evidence on the advertiser-user trade-off and shows that the sponsorship disclosure systematically changes users’ preference over ad placement.
[AI-486] What Does a ProcGen Generalization Gap Measure? Action Rules Convergence and the Missing Random Floor
链接: https://arxiv.org/abs/2609.32532
作者: Abhisek Keshari
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:A generalization gap in reinforcement learning (return on training levels minus return on held-out levels) is usually reported without a reference point. We argue that the missing reference is a measured random floor: the return of a uniform-random policy on the same levels under the same harness. On ProcGen, the floor changes what several standard numbers mean. On identical checkpoints and levels across eight environments, switching between sampled and greedy (argmax) test-time actions moves held-out return in both directions, and greedy evaluation takes three environments to or below the floor: in miner, the sampled policy scores 4.9x the floor on held-out levels while its argmax scores below it. Raw policy entropy places seven of eight environments short of convergence, but 35-65% of that entropy is spread across actions with identical effects; after merging them, one to three remain short, and against the floor only heist has learned nothing that transfers. An audit of twelve prior ProcGen codebases finds that all eleven with held-out evaluation sample test-time actions for their policy-gradient agents, nine by default rather than explicit choice, and six report running in-loop averages rather than evaluating a fixed checkpoint. Applied to our own case study, the same checks grade down a statistically significant encoder effect and rule out a within-encoder train-vs-test CKA statistic. We recommend that every reported gap state its action rule, use a matched and seeded protocol, and report the random floor on both level sets.
[AI-487] Fail Loudly: An Auditable Runtime for Agent ic Data Analysis
链接: https://arxiv.org/abs/2609.32528
作者: Hanxu Yan,Langxuan Deng,Zhengle Wang,Yibo Wang,Chunwei Liu
类目: Artificial Intelligence (cs.AI)
备注: 12 pages, 5 figures
Abstract:Large language models (LLMs) have enabled data-science agents to automate multi-step analyses over heterogeneous files. However, incorrect choices regarding data sources, scope, or statistical definitions often lead to silent errors: computations execute successfully but produce plausible yet incorrect outputs that fail to answer the intended question. To mitigate this, we present RADAR, an auditable runtime that makes an agent’s analytical choices inspectable and supports their revision through execution feedback. RADAR operates through three core mechanisms. First, an evidence-preserving exploration module retrieves task-relevant content while retaining source locations and observation coverage. Next, the runtime uses typed operators to record the agent’s declared inputs, operation arguments, and resulting observations. Finally, runtime validation checks proposed operations against these observations. When a conflict is detected, the runtime rejects the operation or provides diagnostic feedback, allowing the agent to revise its choices before errors propagate. This design enables agents to fail loudly while leaving semantic interpretation to the LLM. On KramaBench, RADAR achieves overall scores of 0.723 with full source retrieval and 0.747 with gold sources supplied, corresponding to relative gains of 35.9% and 28.8% over the strongest baselines. Beyond KramaBench, RADAR achieves relative performance gains of 14.0% on DA-Code and 59.3% on DABStep, demonstrating its applicability across diverse agentic data-analysis workflows.
[AI-488] AmbiModBench: Benchmarking Gene Perturbation Prediction Beyond Shared Responses
链接: https://arxiv.org/abs/2609.32527
作者: Sikai Huang,Zhiwen Yang,Kai Yu,Jiayuan Chen,Stan Z. Li
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Genomics (q-bio.GN); Applications (stat.AP)
备注: 21 pages, 4figures
Abstract:Predicting cellular responses to genetic perturbations helps prioritize experiments in single-cell genomics, where exhaustive measurement is infeasible. While computational models increasingly predict these responses, three evaluation deficiencies obscure what their scores demonstrate. First, absolute metrics cannot separate target-specific predictions from a shared background response. Second, common metrics remain high under gene shuffling, so gene-level accuracy is never verified. Third, a score at one training size says nothing about coverage, which depends on representation-space proximity and response-constraining power. We propose AmbiModBench, a specificity-aware, gene-resolved and coverage-aware benchmark. It pairs every score with a training-mean reference fitted on the same split, screens each readout by gene-coordinate permutation, and links embedding distance to response variation. Across K562, RPE1 and Norman, strong absolute scores largely reflect shared background rather than target-specific learning. Widely used readouts track response magnitude distributions rather than the affected genes. Detectable gain follows representation-space coverage rather than training-set size. Nonetheless, on RPE1 the protocol yields a reproducible target-specific gain across five additional splits and three gene selections, which absolute scores alone cannot distinguish from shared background.
[AI-489] MemAgent : Learning to Manage Heterogeneous Memory Providers for LLM Agents
链接: https://arxiv.org/abs/2609.32521
作者: Yongxian Wei,Yilin Zhao,Runxi Cheng,Xinrui Chen,Chun Yuan,Yaoru Wang,Jiahong Yan,Dian Li
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Current agents remain largely stateless across tasks, limiting their ability to continually improve from prior interactions and making memory essential for long-horizon agentic behavior. Existing memory methods seek to reuse past experience, but most rely on a single memory representation (e.g., trajectories, reflections, skills, structured knowledge) whose effectiveness varies across task distributions. Rethinking this design space, we evaluate 13 memory methods and find that no single method generalizes across benchmarks, revealing the potential of managing heterogeneous memory providers. We formulate agent memory as a routing problem in which a memory agent decides which memory provider to retrieve from, whether to inject short-term memory, and which providers should store the resulting experience. Based on this perspective, we propose MemAgent, featuring a content-aware routing architecture and a training-data synthesis pipeline. The routing architecture combines content-aware probing before retrieval, short-term memory gating during execution, and selective multi-provider storage, while the training pipeline synthesizes phase-specific supervision for routing decisions. Across GAIA, WebWalkerQA, and xBench-DS, MemAgent improves average accuracy by 10.0% and outperforms every individual memory method across all three benchmarks. These gains come with less than 0.3% routing overhead and a 12% reduction in average task steps.
[AI-490] STR: Supervised Transcoder Replacement for Reducing Steering Side Effects
链接: https://arxiv.org/abs/2609.32519
作者: Haonan Yu,Junhao Liu,Zhenyu Yan,Haoran Lin,Xin Zhang
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Model steering can strengthen a target behavior while degrading other useful behaviors. We introduce Supervised Transcoder Replacement (STR) to reduce these side effects for existing steering methods, including those fitted without a protection objective. STR learns a replacement for the multilayer perceptron (MLP) computation at the steering layer through supervision for target control, non-target preservation, and fidelity without steering. Selected steering methods then fit directions on the frozen replacement while retaining their own fitting objectives. We evaluate three steering methods across Gemma and Llama models using Corrigibility preferences and four harmful-request safety datasets. SALAD-Bench supplies protection training data and a separate in-distribution evaluation split; HarmBench, AdvBench, and StrongREJECT are reserved for out-of-distribution testing. STR substantially reduces steering side effects on the in-distribution evaluation and extends this protection to the unseen safety datasets while retaining effective target control. For target-only supervised steering vectors, pooled out-of-distribution attack success rate falls from 42.46% to 14.42% on Gemma-3-4B and from 34.97% to 12.91% on Gemma-3-12B. These results show that replacement training can benefit steering methods fitted without protection objectives.
[AI-491] REFINE: A Resilient Evolution Framework for Intelligent Enterprise Alert Triage in Security Operations Centers
链接: https://arxiv.org/abs/2609.32516
作者: Huimin Chen,Quan Long,Yanhao Wang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Security Operations Centers (SOCs) process large volumes of alerts daily. Alert triage prioritizes high-risk threats while reducing manual review of benign alerts. LLM agents can reason over logs and threat intelligence, but struggle to keep aligned with organization-specific, rapidly evolving SOC operational standards. We introduce REFINE, an LLM-agent framework for enterprise alert triage. REFINE encodes analyst expertise as structured skills and continuously adapts using analyst disposition feedback. It enforces recall = 1.0 as a hard constraint during evolution to maximize auto-closure of false positives, and identifies judgment blind spots by combining alert distributions with model error boundaries. Evaluated on four real industrial SOC scenarios across four MITRE ATTCK phases with temporal split: REFINE achieves recall=1.0 on all evolution sets. On future test windows, it retains recall=1.0 in three scenarios; the degraded case reaches 0.807 recall, still outperforming self-evolution baselines (0.49-0.58). Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) Cite as: arXiv:2609.32516 [cs.CR] (or arXiv:2609.32516v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2609.32516 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-492] From Anomalies to Failures: Constructing Causal Error Graphs for Agent ic Trace Diagnosis
链接: https://arxiv.org/abs/2609.32514
作者: Shu-Xun Yang,Yidong Wang,Zhuoer Feng,Bosi Wen,Jiayi Gui,Dayong Yang,Wenbo Yu,Haoke Zhang,Jie Tang,Cunxiang Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLM-driven agents are increasingly deployed in complex applications, where long agentic traces make failures difficult to diagnose. Existing trace diagnosis methods often conflate anomalies, errors, and failures, making diagnostic targets ambiguous; they also lack structured modeling of how causally relevant errors propagate and amplify into final task failures, resulting in unreliable failure attribution. To address these problems, we propose CEG-Agent, a tool-augmented agentic framework for causal diagnosis of agentic traces. Specifically, CEG-Agent introduces an explicit taxonomy of anomalies, errors, and failures, and constructs Causal Error Graphs (CEGs), a unified typed representation that links execution events, diagnostic nodes, and failure outcomes through causal relations. To evaluate causal trace diagnosis, we further construct CEG-Bench, a fully agent-annotated benchmark with high-confidence, consensus-derived CEG annotations obtained through an Adversarial Agentic Adjudication Protocol (AAAP). We validate the resulting annotations against an expert-curated human gold set, which shows close agreement with the automatic annotations. Experiments on CEG-Bench demonstrate that CEG-Agent achieves state-of-the-art performance under both semantically relaxed and structurally exact evaluation criteria. Our code is publicly available.
[AI-493] Learning from Others Acting for You: Cross-User Memory Sharing for LLM Agents
链接: https://arxiv.org/abs/2609.32511
作者: Jinming Hu,Haodong Zhao,Qi Jia,Die Chen,Tianhang Zhao,Sufeng Duan,Gongshen Liu
类目: Artificial Intelligence (cs.AI)
备注: Work in progress
Abstract:Large language model (LLM) agents serving different users often solve related tasks, yet separate user histories can leave reusable experience inaccessible to other agents. Pooling memories expands access but risks transferring preferences that conflict with the receiving user’s requirements. We introduce ShareMem, a memory architecture that shares reusable experience while grounding its application in the receiving user’s own preferences. Shared experiences indicate how to act and which preferences to consult; the receiving user’s memory supplies their concrete values. Two-stage consolidation refines experience locally before integrating accepted edits into a shared pool. During execution, scope-first retrieval jointly selects local and shared experiences under a common entry budget, while a user-bound channel supports initial and agent-initiated preference retrieval. We evaluate ShareMem across web navigation (Mind2Web), online personalized interaction (VitaBench~2.0), and multi-session coding (MemoryCode) with four backbone models. It improves step success, average task success, and dialogue-macro coding scores, respectively, over matched user-local memory across all four models. Ablations favor two-stage consolidation for smaller shared pools, lower induction token usage, and better downstream performance, and support complementarity between experience guidance and active preference retrieval. Further analyses show that sharing helps most when relevant local experience is scarce, while source quality and cross-user preference interference limit useful transfer.
[AI-494] reeRef-BFN: Equivariance-Free De Novo Molecule Generation based on 2D Topology and Internal 3D Geometry
链接: https://arxiv.org/abs/2609.32502
作者: Ruiqing Sun,Sen Yang,Dawei Feng,Bo Ding,Yijie Wang,Huaimin Wang
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:De novo 3D molecular generation jointly models molecular size, topology, and geometry. Most methods pre-sample molecular size and generate Cartesian coordinates, limiting variable-size conditional tasks such as fragment completion and scaffold decoration while often relying on equivariant architectures. Internal-coordinate methods avoid rigid-body redundancy but typically require a known molecular graph or autoregressive construction, which may accumulate errors. We propose TreeRef, a tree-based molecular representation that assigns molecular topology and topology-dependent local 3D geometry to a naturally variable-size tree. RingRef nodes encode ring closures while preserving the tree structure, while Null nodes allow molecular size to emerge directly from node occupancy. Based on TreeRef, we develop TreeRef-BFN, a Bayesian Flow Network with a standard Transformer backbone that globally couples these locally defined variables and jointly generates discrete molecular variables and continuous local geometry. A single pretrained TreeRef-BFN supports unconditional generation and variable-size structure-conditioned 3D generation through masking alone, without retraining. Empirical studies demonstrate strong chemical validity, molecular stability, and diversity, accurate local geometric distributions, fast sampling, and competitive property-conditioned generation, establishing TreeRef-BFN as an efficient and flexible framework for 3D molecular generation.
[AI-495] Rondo: Unsupervised Discovery of Recurring Temporal Structure
链接: https://arxiv.org/abs/2609.32500
作者: Yingtian Shi,Ankith Chandra,Thomas Plötz
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Many real-world time series data exhibit structural properties at multiple scales, from short, recurring units to complex sequences composed of these units. Unsupervised discovery of both these components and structure enables the design of intelligent systems that help interpret temporal data, thereby limiting the amount of costly human annotations required. Existing modeling approaches typically overlook the hierarchical structure inherent to many time series, treating recurring patterns at different temporal scales as independent structures. Moreover, most assume access to the complete data sequence and treat discovery as a static process, limiting their ability to evolve as new observations arrive. We introduce Rondo, an unsupervised approach for modeling recurring hierarchical structure in continuous temporal streams. By explicitly constructing vocabularies of reusable units and their recurring compositions, Rondo captures structure shared across complex temporal patterns while refining and expanding its discoveries as the stream evolves. Evaluations on temporal sequences spanning diverse domains and data modalities show that Rondo outperforms existing unsupervised recurrence-discovery baselines, with particularly pronounced advantages in limited-data and continual-stream settings. These capabilities provide a stronger foundation for recurring-pattern discovery, scalable behavior understanding, and adaptive intelligent systems operating on long, unlabeled temporal streams.
[AI-496] DAAF: From Failure Localization to Editable System Assets in LLM Agents
链接: https://arxiv.org/abs/2609.32498
作者: Xiaoyang Yuan,Qi Liu,Yubin Ruan,Xinyi Mou,Zhuomeng Zhang,Wenjin Wang,Hanying Jiao,Di Wu,Mingye Xu,Yi Bin,Ke Feng,Zixun Sun
类目: Artificial Intelligence (cs.AI)
备注: 21 pages, 1 figure. Preprint
Abstract:Deployed LLM agents increasingly rely on persistent, versioned system assets such as routing rules, knowledge segments, prompt instructions, and reusable skills. Failure-localization methods can identify where an error manifests in an agent or execution trace, but repair requires a different decision: which editable system asset should be changed, and is that change expected to improve the task outcome? We study this gap through component-attribute failure attribution, where diagnosis targets versioned, addressable items rather than execution locations. We propose the Detection-Aware Attribution Framework (DAAF), which learns the effects of valid attribute replacements and amortizes this intervention evidence into deployment-time diagnosis. DAAF combines sparse and noisy failure signals to decide whether intervention is warranted, learns component-type-conditioned replacement effects from controlled replays evaluated by executable task outcomes, and shares supervision across requests with compatible intervention responses. At diagnosis time, DAAF uses only the observed execution, registered candidates, and available failure signals; it requires neither counterfactual replay nor task reward and returns no_change, a repair target, or an unresolved decision when evidence is insufficient. On held-out tau^2-bench Telecom tasks, DAAF achieves 80.72% attribute Hit@1, recovers 62.65% of failed executions while limiting clean-task regression to 3.23%, and reaches 71.93% overall task success. These results show that intervention-grounded attribute attribution can connect failure localization to executable system repair.
[AI-497] Hearsay: Can an Auditor Trust the Record a Deployed Agent Harness Writes?
链接: https://arxiv.org/abs/2609.32495
作者: Jiahong Dai,Zhuochen Yang,Pengyang Shao,Kelvin Ng,Zhongyi Liu,Chengquan Ju,Yuting He,Bo Hu
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Software Engineering (cs.SE)
备注: 48 pages (9-page main text), 5 figures. Under review. Code and data: this https URL
Abstract:An agent harness, the code that turns a model into an agent, writes its own record of each run, and that record is all a later reader gets when a run is disputed, investigated or audited. We call a record evidentiary when a reader who was not there can check it without trusting the writer. Across sixteen deployed frameworks, none writes one in full. Hearsay examines the record, not the task: five harnesses run fourteen tasks, three blinded LLM examiners and a human panel read the records, and every excerpt an examiner quotes is checked mechanically for who wrote it. First, the record lets a reader name the fault but not prove how the run went. Examiners name the right fault in 74 to 91% of 140 runs, but the fault can be proved only from two files the benchmark adds; for what happened in between, fewer than one citation in ten lands on anything the harness did not write, and the examiner with the fewest false alarms catches half of the entries we delete, rewrite or fabricate. Second, the remedy is a second author, not a stronger seal on the first. An append-only log of what passes between harness and model, kept outside the harness and read against the record in both directions, reports all 28 omissions and fabrications we made a harness commit as it ran, where a hash chain over the harness’s own record passes all 28. Handed the log, examiners keep their fault verdicts but rest more of their citations on what the harness did not write. What makes a record evidence is who writes it, not what is captured.
[AI-498] SoFT: Soft Targets for Generalizable LLM Fine-Tuning
链接: https://arxiv.org/abs/2609.32493
作者: Huihao Jing,Wenbin Hu,Shaojin Chen,Haochen Shi,Zhongwei Xie,Guijia Zhang,Yuxuan Liu,Haoyu Huang,Haoran Li,Yangqiu Song
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Distillation enables student language models to acquire new capabilities from expert teachers. However, integrating knowledge from multi-teacher, multi-domain demonstrations into a single student remains challenging. We study supervised fine-tuning (SFT) in this setting, where students must acquire diverse capabilities while maintaining generalization beyond the training tasks. Our experiments reveal varying trade-offs between in-distribution learning and out-of-distribution generalization across SFT methods, motivating more explicit control over this balance. To this end, we propose soft-target fine-tuning (SoFT) to balance learning from teacher demonstrations with retaining the Base model’s existing capabilities. SoFT sets a minimum target probability for each demonstrated token while making the smallest KL change to the Base distribution. The resulting objective couples learning from demonstrations with adaptively weighted regularization toward the Base model. We further use domain-specific gradient budgets to control this balance and determine a probability threshold for each trajectory. Experiments on mixed-domain reasoning and agentic tasks show that SoFT achieves the best overall performance among the compared methods, with improvements in both in-distribution capability acquisition and out-of-distribution generalization.
[AI-499] Beyond Prompt or Skill? Attribution-Guided Optimization of Modular LLM Programs
链接: https://arxiv.org/abs/2609.32492
作者: Haoran Shou,Haoyue Liu,Yu Huo,Kun Zeng,Xiaoying Tang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models can solve increasingly diverse reasoning tasks, yet their performance remains highly sensitive to task prompts, intermediate instructions, and the way reusable problem-solving knowledge is incorporated. Existing optimization methods usually focus on only one part of this design space: they either optimize a monolithic prompt, or separately induce and refine skills from model traces. As a result, they lack a principled mechanism for deciding which component should be updated when failures occur, and they rarely optimize prompts, skills, and skill-use policies in a unified framework. We propose SPARO (Skill, Prompt, And Routing Optimization), a framework that jointly optimizes task instructions, reusable skill blocks, and routing rules. It performs controlled counterfactual evaluations, converts examples’ effects into a probabilistic responsibility distribution over prompt, skill, and routing components, samples one component from that distribution, and applies the corresponding targeted mutation. This design moves language-program optimization beyond global prompt rewriting toward structured, reusable, and selectively activated task knowledge. Across five benchmarks and five worker models, SPARO consistently outperforms both prompt-centered and skill-centered optimization baselines. These results suggest that effective language-program optimization depends not only on discovering useful task knowledge, but also on deciding where that knowledge should be stored and when it should be activated.
[AI-500] RepoMAS: Solving Progressively Specified Tasks with Issue-Driven Multi-Agent Systems
链接: https://arxiv.org/abs/2609.32490
作者: Yuchen Song,Andong Chen,Wenxin Zhu,Muyun Yang,Tiejun Zhao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLM-based multi-agent systems (MASs) have shown strong potential for solving complex tasks, but most assume that task requirements are sufficiently specified before execution. In practice, user requests are often incomplete, and additional requirements may only become clear during reasoning, tool use, or execution. We refer to such problems as progressively specified tasks. To systematically study this setting, we introduce ProgSpec, a benchmark that evaluates final outputs against requirements explicitly stated in the initial request and additional requirements supported by the available task evidence. We further propose RepoMAS, an issue-driven multi-agent framework inspired by open-source project management. RepoMAS records newly discovered requirements, conflicts, and failures as structured Issues and uses them to revise the task specification and execution structure during problem solving. Across ProgSpec and five existing benchmarks, RepoMAS achieves the best performance. Further analyses show that its issue-driven revision and repository maintenance mechanisms consistently contribute to performance. These results highlight the importance of allowing MASs to revise not only how a task is solved, but also revise their explicit representation of task requirements during execution. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.32490 [cs.AI] (or arXiv:2609.32490v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.32490 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-501] owards Scalable Data Diversification for Language Model Pretraining via Leverag e Score Sampling
链接: https://arxiv.org/abs/2609.32484
作者: Zailin Ma,Quzhe Huang,Yujun Li,Congyuan Rao,Yaodong Yang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Data selection for language model pretraining faces a fundamental tension between quality and diversity. While quality filtering is empirically effective, it often induces diversity collapse: by favoring texts similar to high-quality reference corpora (e.g., educational or QA-style data), it systematically excludes valuable data from underrepresented domains. In contrast, diversified selection preserves domain balance and encourages robust downstream performance, yet existing methods either focus on coverage-oriented objectives that indirectly enhance diversity, or directly optimize for diversity via costly covariance matrix recomputation that limits scalability. To address these issues, we introduce \textbfLeverage Score Sampling (Lev), which iteratively selects samples that maximally expand the determinantal volume of the embedded data via leverage scores, a computationally efficient criterion that eliminates matrix recomputation and enables scalable selection. Empirically, Lev delivers up to 72\times speedup and improves dataset diversity, measured by the Vendi score, by 9.2% over the strong diversification baseline \textbfDiSF. On CommonCrawl (CC) web data selection, Lev improves accuracy across seven downstream tasks by up to 1.31% over existing baselines. For domains where robust quality criteria are inherently difficult to define (e.g., code), Lev serves as an effective unsupervised curation alternative: on StarCoderData, the selected subset reduces bits-per-byte by 3.08% over DiSF. Notably, we uncover a cross-domain collapse of quality filtering: CC data filtered by DCLM-fastText fail to retain sufficient code-related content, yielding inferior code performance relative to Lev-selected data. These findings advocate for integrating diversity-aware practices into quality filtering for more effective data curation in language model pretraining.
[AI-502] Separating Diagnosis from Disease Representation: Dual-View EEG Learning with Neural-Dynamics-Guided Deformation
链接: https://arxiv.org/abs/2609.32483
作者: Jiaying Wang,Shouqian Shi,Yutong Chen,Xu Yang,Jie Chen,Xingyu Pan,Lei Zhang,Sheng Zhong
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Electroencephalography (EEG)-based closed-loop neuromodulation calls for a subject-specific structured state, as opposed to a single disease probability, specifying which brain regions are deviant, at which frequencies, and at which lags. Sensor-space models keep the strongest diagnostic evidence without anatomy, source-space models give anatomy at a loss of predictive signal, and post-hoc attributions stay outside the prediction. We separate the two instead of forcing them into one representation, and propose DMD-EEG (Dual-view Multiscale Deformation for EEG), which keeps a fixed scalp spectral expert for diagnosis and models the source-space disease-related representation as a low-rank, sparse, iterative deformation of a healthy neural-dynamics prior in a 46 -region-of-interest (ROI) \times 5 -frequency \times 4 -lag (autocorrelation-timescale) space. The two experts meet only at a fixed decision level, so the source state is architecturally separate from the scalp expert. Across major depressive disorder (MDD), first-episode psychosis (FEP), and Parkinson’s disease (PD), decision-level fusion matches the strongest single expert on MDD and FEP and exceeds the source branch on PD. On FEP the source expert is the strongest branch, the task where the deformation contributes most. The source state is an explicit ROI-frequency-lag attribution defined in a shared source coordinate system across montages, which we treat as an anatomically-coordinated predictive representation whose coordinates are directly readable and hypothesis-generating. The highest-saliency coordinates align with established disease circuitry (fronto-limbic-temporal regions in MDD, motor-cortex beta in PD), and the MDD state transfers by rank to an unseen cohort recorded with a different montage.
[AI-503] From Outcomes to Strategies: Learning Strategy Utility for Mathematical Reasoning
链接: https://arxiv.org/abs/2609.32482
作者: Ruikang Zhang,Xiao An,Xuli Shen,Jiaxing Sun,Xiaoyi Yu,Jin Zeng,Jiang Wu,Tong Lin
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning with verifiable rewards has substantially improved mathematical reasoning. However, terminal correctness alone provides limited insight into the quality of high-level strategies, such as theorem selection and subgoal decomposition, when considered separately from their subsequent execution. This paper studies strategy utility, which is defined as the likelihood that a strategy supports a correct downstream solution under a given executor. We introduce SURE, a framework for learning and leveraging relative strategy utility. In this framework, high-level strategies are separated from their detailed reasoning. Based on the pairwise preferences constructed from strategy-conditioned rollouts and teacher-generated contrasts, a Strategy Reward Model is learned to estimate relative strategy utility. During reinforcement learning, the frozen reward model reads only the extracted strategy, whose score is combined with the correctness and format rewards in a sequence-level GRPO objective. Compared with outcome-and-format GRPO baselines, experiments show that SURE improves average pass@1 by 1.87%, 2.64%, and 2.93% across three policy backbones. Our method also achieves competitive or better accuracy than stronger reward baselines while requiring substantially lower GRPO-stage compute.
[AI-504] VPEvolve: A Self-Evolving Virtual Process Engineer for Computational Lithography
链接: https://arxiv.org/abs/2609.32473
作者: Tianyi Li,Wenxuan Dong,Donger Luo,Nan Wang,Yanpeng Chen,Jiaqi Liu,Xinyun Zhang,Hao Geng
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Optical proximity correction (OPC) recipes grow as engineers add local rules to repair newly discovered lithography hotspots. Each correction can interact with existing rules, while lessons from commercial-tool trials remain scattered across code and logs. \system combines a Virtual Process Engineer (VPE) harness with a Skill Bank of measured engineering experience. The harness equips a frozen language model with process manuals, layout analysis, recipe editing, and commercial-tool evaluation. The actor proposes changes to the global parameters, local targeted rules, or diagnostic trials. After each evaluation, an LLM reflector and curator turn the measured response into evidence-linked judgments. The actor retrieves them before its next trial. Feasible improvements update the retained recipe; every measured trial informs the Skill Bank that guides the next edit. The model weights remain fixed. On a FreePDK45-derived benchmark with ten commercial-tool evaluations per case, \system reduces the mean per-case maximum edge placement error from 18.294 to 5.361 nm on Poly and from 22.052 to 15.692 nm on Metal1. Every final recipe satisfies the predefined quality constraints and improves the maximum error by at least 0.1 nm.
[AI-505] Bison: Cross-Dataset Learning for Unseen-Compound Perturbation Prediction
链接: https://arxiv.org/abs/2609.32467
作者: Yunfan Liu,Kasra Ghorbani,Yufei Huang,Zicheng Liu,Jiangbin Zheng,Jingbo Zhou,Shaorong Chen,Chang Yu,Stan Z. Li
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Predicting transcriptional responses to unseen compounds is limited by fragmented chemical coverage and heterogeneous experimental platforms and gene panels. To assess molecular generalization across these settings, we build on Chem-PerturBridge to benchmark eight datasets with 16,771 compounds, withholding test compounds from every training dataset. This comparison reveals that high overall response agreement can coexist with weak prediction of drug-specific differences, despite reproducible signals across repeated measurements. To exploit complementary chemical supervision while targeting these differences, we introduce Bison: a shared gene representation connects native panels, while two discrete diffusion models compose context-dependent responses with molecular deviations learned through matched drug-contrast supervision. A single Bison model jointly trained across all eight datasets achieves the highest mean overall-response and drug-contrast Pearson correlations on the full benchmark in comparison with 11 methods trained independently per dataset. Compared with dataset-specific training of the same architecture, joint training increases mean drug-contrast correlation by 27.4%, with gains across all eight datasets and improvements in overall response prediction. These results demonstrate how matched drug contrasts turn complementary screens into shared molecular supervision for unseen-drug response prediction while preserving native gene measurements.
[AI-506] Adapting Nonstationary Multi-output Gaussian Processes to Bayesian Optimization
链接: https://arxiv.org/abs/2609.32464
作者: Zikai Xie
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-objective Bayesian optimization (MOBO) commonly relies on independent Gaussian processes (GPs) with stationary kernels, limiting its ability to represent nonstationary structure and share information between objectives. However, expressive nonstationary GPs do not necessarily make reliable BO decisions. We study this mismatch for the multi-output low-rank nonstationary (MO-LRN) GP: strong training fit can coexist with large off-design errors and optimistic acquisition predictions. We introduce MOLRN-BO, which combines a regularized shared-spectral surrogate with objective-specific residuals, prequential mean correction and tempered covariance scaling, and Pareto-local qLogEHVI optimization with periodic global search. Experiments on 12 deterministic bi-objective benchmarks show that MOLRN-BO substantially improves upon the original MO-LRN and achieves the best average problem ranks for final normalized hypervolume and normalized inverted generational distance among nine evaluated algorithms. It also achieves the strongest adverse-tail performance while remaining competitive with the leading baselines in anytime optimization. Ablation studies further show that the shared spectral construction improves off-design prediction, the local–global decision policy improves optimization performance, and hierarchical calibration reduces systematic candidate bias. These results demonstrate that nonstationary multi-output surrogates can deliver strong and robust MOBO performance when their structure and use are explicitly adapted to the demands of sequential optimization.
[AI-507] DRAM: Delta-rule Recurrent Associative Memory for Robot Manipulation Policies
链接: https://arxiv.org/abs/2609.32453
作者: Xinyu Zhao,Yixiang Shan,Tao Yang,Runyu Lei,Yiming Zhao,Jiaxin Fan,Zongbao Feng,Peng Jia
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Robotic manipulation is inherently history-dependent, yet most pretrained robotic policies condition on only the current observation or a short temporal window. Equipping such policies with long-term memory remains challenging: existing approaches either feed the backbone multi-frame observation windows, which substantially increase inference cost, or rely on pre-defined semantic features, which limit task generality and may also require the retraining of the backbone to adapt to the memory. We introduce DRAM (Delta-rule Recurrent Associative Memory), a plug-and-play memory module that can be attached to a wide range of pretrained robotic policies, endowing them with long-horizon memory without architectural modification or backbone retraining, requiring only task-specific post-training of the memory module and action expert. DRAM maintains a fixed-size associative memory using gated delta-rule linear attention, with a modified update that incorporates all tokens within each frame in parallel. An architecture-agnostic readout integrates historical context into action prediction across different policy architectures. Experiments show that DRAM consistently improves frozen pretrained policies over short-context baselines and alternative compact memory designs, validating its effectiveness as a fixed-size, post-hoc memory module trained with the backbone frozen.
[AI-508] Rethinking Training-Inference Mismatch in LLM Reinforcement Learning: Where It Arises and How to Correct It
链接: https://arxiv.org/abs/2609.32444
作者: Tianrun Yu,Kaixiang Zhao,Shangzhe Li,Yuxiao Yang,Porter Jenkins,Weitong Zhang,Taylor W. Killian
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 32 pages. Code: this https URL
Abstract:We study training-inference mismatch in reinforcement learning with verifiable rewards (RLVR) for large language models, where rollouts are sampled by an inference engine while gradients are computed by a training engine, and the two engines assign different probabilities to the same tokens. To account for this discrepancy in policy updates, we introduce calibrated importance sampling (CIS). CIS is motivated by an empirically supported logit-displacement characterization that expresses the mismatch as an additive displacement \varepsilon_t in log-odds, determined by the per-logit perturbation before the softmax, whose distribution is approximately invariant to token confidence. This characterization motivates a confidence-aware truncation: large positive displacements are truncated at a single constant threshold, which maps back to an importance-ratio cap that tightens as token confidence increases. Theoretically, we show that CIS replaces the unbounded second moment that governs the error of exact importance sampling with a term bounded by a constant, at the cost of a bias controlled by the truncated excess. In evaluation across three mixture-of-experts models and five mathematical reasoning benchmarks, CIS achieves the highest five-benchmark average on all three models among the evaluated baselines. Diagnostic analyses show that CIS places less truncation bias on low-confidence tokens than truncated importance sampling, while upward clipping of small importance weights reduces held-out accuracy.
[AI-509] Controllable GNN Explanations via Multi-Metric Preference Selection
链接: https://arxiv.org/abs/2609.32436
作者: Rachit Verma,Yashraj J. Deshmukh,Anirban Dasgupta
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Mechanisms for generating GNN explanations are crucial for building trust and mitigating biases in Graph Neural Networks (GNNs), especially in high-stakes scenarios. Most current methods optimize only for fidelity under the sparsity constraint. However, this discounts the need for interpretable explanations (those that consist of familiar motif patterns) and stable explanations (those that remain unchanged under structural perturbations). We propose a novel approach that optimizes GNN explanations across these metrics, exposing their relative weighing as a control. Experiments on various real-world datasets, including MUTAG, BA-2Motif, BAMultiShapes, and PROTEINS, suggest that our method produces higher-fidelity explanations than a state-of-the-art baseline on MUTAG and PROTEINS across all evaluated budgets, and on BA-2Motif at larger budgets, while being faster in the regime of small explanation budgets. We also explore how, given an input motif library containing standard motifs for the corresponding domain, the method can be used to determine the relative importance of those motifs in generating the explanations, and how this information can be used to further improve the quality of the output explanations. We also examine the relationship between different metrics through their induced tradeoff surface, and explore its dependence on the nature of the motif library.
[AI-510] From Latents to Wires: Surgical Post-Editing on Large Language Models
链接: https://arxiv.org/abs/2609.32434
作者: Jiankai Jin,Xiangzheng Zhang,Zhao Liu,Wenzhuo Xu,Dongdong Yang,Deyue Zhang,Quanchen Zou
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Given a large language model (LLM), can whoever holds the weights name a semantic target (e.g., the model’s identity), locate the model components that produce it, and edit them so that the target no longer appears while other capability is preserved? We call such an edit on a trained model a post-edit. We present L2W (latents to wires), a framework that performs surgical post-edits for named semantic targets. For localization, L2W uses Jacobian lens (J-lens) attribution to score components against the semantic target. For surgical removal, because LLM mechanisms are redundant (i.e., a semantic target may have multiple components producing it), L2W runs Counterexample-Guided Causal Cut (CGCC) until the target no longer appears. CGCC first cumulatively closes model components, treating each surviving expression of the target as a counterexample that exposes the next components to close, and then reopens some of them to preserve capability. In a controlled experiment with an implanted behavioural watermark, L2W removes the watermark, and its localization lands on the model region the implant changed. Across three model configurations, L2W removes model-metadata (e.g., identity) self-claims in all nine runs, and adult-content refusal in all three, with no held-out target residual. L2W further composes two post-edits on a text-to-image model: one removes the refusal of requested nudity, and a second removes the nude rendering the first exposes. The results support post-editing as a complement to post-training: post-training installs preferred behaviours, and post-editing removes named unwanted ones.
[AI-511] Multi-Agent System Search via Active Substructure-aware Policy Optimization
链接: https://arxiv.org/abs/2609.32430
作者: Beicheng Xu,Bowen Fan,Weitong Qian,Lingching Tung,Bin Cui
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLMs enable multi-agent systems (MAS) to tackle complex tasks, but manually designing agent roles, prompts, and communication structures requires substantial expertise and effort. This motivates learning policies that construct query-specific MAS from execution reward. Existing approaches typically train these policies by repeatedly traversing a fixed set of training queries and assigning rewards at the workflow level. However, this overlooks differences in queries’ evolving learning potential and obscures which substructures improve solution quality. In this paper, we propose Active Substructure-aware Policy Optimization (ASPO), a RL framework for query-level MAS search. ASPO introduces an Adaptive Query-Selection Mechanism (AQSM) that focuses training on queries at the policy’s competence boundary: those it can solve but not yet reliably. A complementary discovery mechanism widens architectural exploration for hard queries, helping distinguish insufficient exploration from operator capability limits. Beyond query selection, ASPO introduces substructure-level rewards that measure output-quality gains within each action’s descendant subgraph. These rewards guide proximal policy optimization to reinforce useful architectural refinements and discourage redundant or harmful computation. Together, these mechanisms prioritize learnable queries and provide fine-grained feedback for learning effective MASs. Across six benchmarks spanning mathematical reasoning, general question answering, and code generation, ASPO ranks first on every benchmark against twelve baselines.
[AI-512] PrismQuant: Optimal Null-Space Rotations for Grouped Quantizers
链接: https://arxiv.org/abs/2609.32429
作者: Yanlong Chen,Yining Chen,Song Zhang,Amirhossein Habibian,Yawei Li
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: The paper is currently under review. Code, checkpoints, and implementation details are available at: this https URL
Abstract:Smaller activation outliers do not necessarily imply better low-bit quantization: their alignment with the quantizer matters. We introduce PrismQuant, a quantizer-aware rotation framework that aligns the leading activation eigenspace with the constant group subspace of asymmetric grouped INT4. The affine offsets represent the energy in this subspace without widening the range within the group. We formulate rotation design as a Ky Fan trace maximization and derive a closed-form solution that is provably optimal for this alignment objective. Compact Householder transformations and their compact-WY representation enable gradient-free construction and efficient application at both foldable and online sites. A predictive range law further connects unaligned activation energy and group size to quantization-relevant variation. Experiments on Llama, Qwen, and Mistral span dense models up to 70B parameters and a 30B mixture-of-experts model. Under W4A4KV4, PrismQuant sets the state of the art on Llama-3.2-3B among the compared methods in both perplexity and accuracy. On Llama-3.1-70B, it attains 3.85 perplexity and 72.46% average zero-shot accuracy, only 0.22 percentage points below full precision. In the deployment study on Llama-3.1-8B, our optimized implementation achieves 1.51x prefill and 1.22x CUDA Graph decode speedups over matched FP16 baselines, with 56.34% lower decode peak memory and only 2.35% additional Graph decode latency over Hadamard. Code is available at this https URL.
[AI-513] Authorization Closure Graph: Minimal Repair for LLM Agents with Evolving User Instructions
链接: https://arxiv.org/abs/2609.32428
作者: Qingzhuo Wang,CaiYi Wang,Jinglu Meng,Ruiyang Qin,Kunyu Peng,Zhihua Wei,Wen Shen
类目: Artificial Intelligence (cs.AI)
备注: 22 pages, 8 figures, 6 tables. Preprint under review
Abstract:Tool-using large language model (LLM) agents increasingly perform state-changing actions that require user authorization. Yet existing approaches do not provide a principled mechanism for selectively updating prior authorization when only part of an instruction changes. To this end, we propose an Authorization-Closure-Graph (ACG)-based framework that represents authorization and its dependencies as an evolving, versioned state. ACG selectively invalidates authority affected by a revision while preserving unaffected portions of the authorization state, and computes a minimal repair that identifies only the missing evidence or authority required for execution. This enables agents to adapt to revised instructions while avoiding stale authority and unnecessary authorization requests. We evaluate ACG across three advanced LLMs in two natural tasks, and ACG consistently improves action safety rate and task success rate. Code is available at this https URL.
[AI-514] CyberClear: A Benchmark for LLM Agent Systems on APT Attack Chain Provenance
链接: https://arxiv.org/abs/2609.32424
作者: Qi Chen,Fushuo Huo,Hangli Shen,Jingcai Guo,Shuhao Li,Guang Cheng
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model agents have demonstrated promising capabilities in cybersecurity tasks, yet their ability to reconstruct complete Advanced Persistent Threat attack campaigns from complex security logs remains largely unexplored. Existing cybersecurity benchmarks for agents mainly focus on vulnerability discovery, exploitation, and security analysis tasks, leaving the evaluation of attack chain provenance under realistic security logs insufficiently studied. To address this gap, we introduce CyberClear, a benchmark for evaluating LLM agents and advanced agent systems on APT attack chain provenance from long-context security logs. CyberClear covers both single-step attacks and multi-stage attack chains, requiring agents to identify attack evidence, infer attack progression, and generate provenance graphs containing entities, causal relationships, MITRE ATTCK techniques, and forensic evidence. To enable comprehensive evaluation, we develop an evaluation method tailored to APT attack chain provenance. Unlike conventional text similarity metrics that focus on surface-level matching, our evaluation examines whether reconstructed graphs preserve the semantics of attack chains across single-step behavior correctness, multi-step behavior identification, temporal and causal consistency, entity and relationship fidelity, and overall attack narrative consistency. Advanced multi-agent systems powered by state-of-the-art LLMs still struggle on CyberClear, motivating us to propose CyberProvenance, an agent cyber harness designed for multi-agents that augments LLM agents with evidence accumulation, execution-based validation, and feedback-guided refinement mechanisms for reliable attack-chain provenance. Extensive evaluations on CyberClear demonstrate the effectiveness of CyberProvenance in improving evidence reasoning, execution-grounded validation, and complete APT attack chain reconstruction.
[AI-515] PluginRSI: Recursive Improvement of Agent Harnesses with Reusable Plugins
链接: https://arxiv.org/abs/2609.32423
作者: Yaorui Shi,Yuchun Miao,Yuxin Chen,Jiayuan Zhang,Yueqing Sun,Xierui Song,Xiang Wang,An Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:The harness surrounding a language model is a central determinant of agent performance. Recent methods optimize harnesses by searching over complete programs, where individual mechanisms are difficult to isolate and reuse. We introduce PluginRSI, which represents a harness as a composition of atomized plugins and organizes harness evolution around these plugins. Individual plugins are improved independently and accumulated in a shared library, then recombined into new harnesses at each iteration. PluginRSI improves over existing harness optimization methods across software engineering, command-line interaction, and question-answering tasks. The resulting harnesses retain their advantage when transferred to other solver models without further optimization. The evolved plugin library accelerates subsequent optimization from the initial harness, which helps faster and higher convergence on unseen tasks. These results show that accumulating reusable mechanisms provides an effective basis for continued harness improvement.
[AI-516] MergeHEIR: Mitigating Multimodal Hallucinations as the Tax of Model Merging
链接: https://arxiv.org/abs/2609.32422
作者: Jinyu Li,Hao Fang,Zhiming Zhang,Jiawei Kong,Bin Chen,Shu-Tao Xia
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Model merging consolidates task-specialized experts into a single deployable model. However, we show that such capability consolidation incurs a merging tax of increased hallucination: across 8 model-merging methods, every merged checkpoint exhibits a higher hallucination rate than the average of its constituent experts. An intuitive approach is to adapt existing hallucination-mitigation methods to the post-merge model, yet this unconstrained adaptation disrupts inherited capabilities, creating a tension between hallucination mitigation and expertise retention. To tackle this challenge, we introduce MergeHEIR, a post-merge adaptation framework designed to reduce this merging tax while preserving expertise inherited from initial experts. Using small expert-task calibration sets, MergeHEIR constructs layer-wise null-space projectors via SVD from task-specific activations collected from the merged checkpoint, and periodically projects the accumulated post-merge displacement onto the resulting null spaces to preserve inherited expertise. Theoretically, we establish minimum-distortion and maximum-dimensionality guarantees, characterize the threshold-controlled adaptation-retention trade-off, and extend perturbation guarantees beyond finite calibration data. Across 24 paired comparisons spanning three MLLM configurations and 8 model-merging methods, MergeHEIR consistently mitigates hallucination while largely preserving inherited expertise, demonstrating a more favorable hallucination-retention trade-off.
[AI-517] Carnator: Fast Text-to-Video Generation with Generation-Native Compatibility-Guided Cross-Request Reuse
链接: https://arxiv.org/abs/2609.32420
作者: Xingkun Yin,Xuebin Tang,Mingkun Xu,Hongyang Du
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Video diffusion transformers produce high-quality videos, yet iterative denoising incurs substantial inference latency, limiting interactive and large-scale serving. Most existing acceleration methods focus on individual requests, thereby restricting efficiency gains to redundancy within a single generation trajectory. Recent cross-request reuse offers an additional source of savings, but existing approaches often infer reusability from coarse semantic similarity. This conflates semantic relatedness with generation-level computational compatibility, so aggressive reuse may accept incompatible historical computation while conservative reuse leaves substantial acceleration unrealized. We present \emphCarnator, a cross-request acceleration framework that addresses this challenge by extracting and using generation-native compatibility evidence directly from the model’s evolving internal states. Specifically, \emphCarnator performs a lightweight early probe to construct an Early Signature from internal diffusion states, assessing reuse validity through risk-aware compatibility decisions. The same evidence characterizes reuse scope by localizing target-specific computation and guiding joint reuse of historical latent trajectories and sparse attention connectivity. Across three text-to-video backbones, Carnator consistently achieves higher cache-hit end-to-end acceleration than the evaluated cross-request baselines despite more selective cache acceptance, reaching up to 2.17 \times speedup while maintaining competitive generation quality.
[AI-518] RE-0: Verified Recursive Improvement of Embodied Code-as-Policy Agents through Local On-Policy Distillation
链接: https://arxiv.org/abs/2609.32416
作者: Jiawei Zhang,Xiangrong Zhang,Rui Song,Huanbin Zhou,Chengye Song,Hongzhou Wang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 17 pages, 5 figures. Corresponding author: Hongzhou Wang (wanghongzhou@jlu. this http URL )
Abstract:Code-as-Policy agents accomplish long-horizon embodied tasks by generating and executing code, yet continually improving them with teachers that are stronger but not globally reliable remains a key challenge. Existing distillation methods typically treat the teacher’s complete behavior as the supervision target and thus misassign training credit on states where the teacher fails. We propose RE-0, a recursively verified policy improvement framework: rather than assuming that the teacher globally outperforms the student, RE-0 requests local corrections from the teacher on the student’s own failure histories and checks in the environment whether each correction is genuinely beneficial; verified corrections yield immediate improvement. Building on this, we propose RE-OPD, which turns verified interventions into supervision for on-policy distillation. Only counterfactually verified teacher interventions provide distribution-level supervision, weighted by their measured local benefit, and the improvement they induce is projected back into the standalone student, so both where supervision is applied and how much credit the teacher receives co-evolve with the student policy. We further prove that the student’s per-round gain is lower-bounded by its verified intervention gain up to verification and projection error terms. Experiments across multiple Code-as-Policy embodied tasks show that RE-0 improves both teacher-assisted execution and the standalone student, and generalizes to novel robots and scenes.
[AI-519] Opening LLM Judges: Recovering Preference Signals Beyond the Final Verdict
链接: https://arxiv.org/abs/2609.32407
作者: Sourabrata Mukherjee,Sunayana Sitaram
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLM judges are widely used to evaluate model outputs, but their verdicts can be unreliable: a judge may favor the worse answer for its position, length, or other surface features. When a judge is wrong, is the information needed to judge correctly absent from the model, or present in its internal representations but not reflected in the output? We study this across 64 open-weight evaluators and 14 datasets, including causal interventions on 41 judges (editing activations mid-run to see whether the verdict changes). On LLMBar, built so the superficially better answer is the worse one, the verdicts of 50 judges agree with human labels only 0.456 of the time, even after averaging both answer orders. Yet a small probe on the same judges’ activations, with no weight updates, reaches 0.846, and 0.686 once surface features such as length and position are residualized out (0.507 with shuffled labels). The gap holds across eight benchmarks and model families, but is not universal: a score of how well surface features alone predict the human label, computed before any probe is trained, predicts the size of the gain (Spearman rho = 0.90). On rubric tasks that score one answer at a time, leaving no surface cue to exploit, reading the internals gives no advantage. The interventions also show that editing activations mid-network already changes the verdict, before it can be read off directly, and locate the pathways carrying position and length bias. At the same label budget, the recovered signal lets a judge flag cases where it is likely wrong and yields better labels for preference learning. A wrong verdict, then, does not mean the judge lacks the information, and a simple diagnostic shows when it is worth recovering.
[AI-520] SkillDRE: Dual-Stage Red-Team Evolution of Agent Skills via Pre-Execution and Runtime Feedback
链接: https://arxiv.org/abs/2609.32400
作者: Pengyu Zhu,Jingyi Yang,Yi Liu,Li Sun,Sen Su
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Agent skills package instructions, executable code, and task-specific resources into reusable artifacts that agents can improve using execution feedback. The same mechanism also enables attackers to evolve malicious skills, making them more effective and less detectable. However, a candidate skill may pass pre-execution scanning yet fail to realize its target under runtime defenses, while a revision that repairs execution may introduce new scanner findings. We introduce SkillDRE, a fully automated framework for evolving complete malicious skill packages through a dual-stage feedback loop. Given a benign task and its associated skills, SkillDRE autonomously constructs and validates a task-conditioned malicious objective and a verifiable judge rule. It then holds both fixed while evolving the skill implementation, with preservation of legitimate task capability. SkillDRE combines scanner-guided evolution with runtime-guided refinement informed by execution outcomes observed under runtime defense. Each runtime-guided revision returns to the pre-execution stage for rescanning and further optimization before re-execution, forming a cross-stage closed loop. Evaluated on SkillsBench across four victim models, SkillDRE achieves an average attack success rate of 45.28%, exceeding the strongest baseline by 40.3%, while its final submitted skills receive no SkillScan findings and largely preserve benign-task performance. These results show that two-stage defense feedback can serve as a useful learning signal for adaptive red teaming and that evaluating either defense stage in isolation can miss the resulting attack capability. Codes is available at this https URL
[AI-521] Function Over Form: Distributional Orthogonalization in Mixture-of-Experts with Replica Expert Mechanism
链接: https://arxiv.org/abs/2609.32398
作者: Jinfan He,Yunzhuo Liu,Kai Zhang,Weidong Han, Key,Rayying
类目: Artificial Intelligence (cs.AI)
备注: COLM 2026 accept
Abstract:The scaling of LLMs increasingly relies on MoE architectures to decouple active computation from total parameter count. However, the efficacy of MoE is often constrained by expert collapse and representation redundancy, both leading to underutilization of model capacity. To address these challenges, this paper proposes Distributional Orthogonalization Loss (DO-loss), an auxiliary regularization that shifts the focus from static weight diversity to dynamic routing behavior. By representing each expert’s token assignment history as a high-dimensional binary load signature, DO-loss penalizes signature overlap to prevent expert collapse while encouraging functional specialization. To align this algorithmic design with system efficiency, we further introduce the Replica Expert Mechanism (REM), which improves load balancing through a two-tiered strategy: adjusting replica expert placement at the global-batch level and performing real-time token dispatching at the micro-batch level. Empirical evaluations demonstrate that our method outperforms the evaluated routing algorithms on downstream tasks for both 4.8BA0.5B and 30BA3B MoE models, while maintaining comparable training efficiency.
[AI-522] Memory as a cache: Exact context reuse and deletion by construction
链接: https://arxiv.org/abs/2609.32395
作者: Shengyao Wang,Jiang Liu
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:The KV cache of a transformer entangles every token’s representation with its entire prefix: a passage encoded once cannot be reused under a different prefix or removed without recomputing everything after it, so exact cache reuse is limited to shared prefixes. We present SMem, an architecture whose context representation is a cache by construction. A block-local encoder maps each block to memory rows independently of other blocks, and a reader conditions generation on their union through cross-attention. For every parameter setting, memory composes exactly at fixed block indices, deleting a block is an exact O(b) update for b -token blocks, and the memory state is independent of the edit path. At 4\times the training context, under the shared recipe, SMem retrieves planted needles beyond any trained-length window (exact match 0.14-0.28 at distances of 31 and 63 blocks), where learned-position, RoPE, and Block-Attention-style transformers all score at most 0.02. A fully cached context is served by computing one block alone at a near-constant 3.1-6.2 ms, whereas cold prefill grows with context; batched decode stores 34-38% fewer KV rows and runs 1.4-1.7 \times faster when bandwidth-bound; and deletion beats suffix recomputation by 8.5 \times at 512 blocks and 452 \times at 4096 blocks (32-256 \times the trained length, probing the cost model rather than a served regime). The cost is a perplexity gap of -4.7% to +2.8% (negative favors SMem) against a parameter-matched transformer with the same positional scheme, at 160M-1.5B on FineWeb-Edu across two recipes and a learning-rate search. SMem also composes with RoPE: at 160M and 410M the composite matches or leads the matched transformer and closes 29-59% of SMem’s gap to a RoPE transformer. Dropping prefix entanglement thus keeps perplexity comparable while making the cache exactly composable and editable.
[AI-523] Beyond Scripted Search: Sample-Efficient Reward Discovery via Agent ic Black-box Optimization
链接: https://arxiv.org/abs/2609.32394
作者: Minghao Li,Rui Tan,Ruihang Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Designing dense reward functions for low-level reinforcement learning (RL) control remains difficult. Recent work uses large language models (LLMs) to iteratively generate and refine reward functions using policy-training feedback within scripted search algorithms. However, evaluating each candidate requires a full RL training run, making sample efficiency a central challenge for reward search on complex control tasks. To address this limitation, we propose an Agentic Reward Black-box Optimization (ARBO) framework, in which an LLM agent builds the search strategy at run time from an evaluation history maintained as its persistent workspace. The evaluation history comprises two components: observations maintained by the evaluation oracle, including candidate scores, per-term training curves, and error tracebacks; and an agent-maintained belief that records diagnoses and intended next steps. The agent queries both with tools and generates the next batch of reward candidates, rather than generating them in a single pass from a fixed prompt. Across four control domains, ARBO achieves gains of 29.9% in manipulation success rate and 192.8% in power-grid score over baseline means under a shared evaluation budget. Ablations examine each component’s contribution and sensitivity to backbone choice.
[AI-524] SCLATE: a Substrate for Continual-Learning Agent Training and Evaluation ICLR2027
链接: https://arxiv.org/abs/2609.32391
作者: Youngmok Jung,Sirajul Salekin,Henry Tran,Javier Movellan,Zhao Huang,Manjot Bilkhu
类目: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
备注: submitted for ICLR2027, Apple Inc
Abstract:Continual-learning agents are systems of models, harnesses, and memory operating over long multi-session horizons. Evaluating and training them requires interleaving tasks with agent-side events such as session stop and start, crons, and memory consolidation. Yet existing benchmarks and training frameworks schedule only the benchmark’s own events, leaving each benchmark and agent pair to build a custom scheduling loop. We present SCLATE, an execution substrate where benchmarks and unmodified agents each add their events to one open event scheduler through an adapter. A hybrid simulated clock runs these events on a shared timeline, flowing in real time while the agent works and skipping idle gaps, which compresses a month-long scenario into hours. SCLATE also serves as a rollout engine that runs any agent’s harness and memory unmodified, recording the tokens and log probabilities of every model call through an in-container proxy. We port seven benchmarks to SCLATE and compare ten unmodified harness and memory configurations head to head on ten models. The comparison shows that an added memory system does not reliably beat the harness’s native memory and that models differ widely in how they use the same harness and memory. We then post-train Qwen3.5-4B through unmodified harnesses and memory systems. The model learns to use both, reading 6.8x fewer file lines with a 16.7-point higher SWE-bench Verified pass rate, and writing richer memory records, while its held-out MetaClaw accuracy rises by up to 11.8 points.
[AI-525] Reward Hacking and Agent Containment Failure: A Monte Carlo Study Based on the 2026 Hugging Face Incident
链接: https://arxiv.org/abs/2609.32390
作者: Murat Ozer,Bulent Erenay,Ibrahim Berber
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: 14 pages, research paper, and two figures
Abstract:The July 2026 intrusion into Hugging Face production infrastructure showed how reward hacking can become an external cybersecurity incident when a capable agent encounters weak containment boundaries. This study develops a probabilistic risk model linking five stages: reward hacking, containment escape, usable access, persistence, and failure of detection. A Monte Carlo simulation evaluates 100,000 runs under each of four control configurations. Input distributions represent explicit uncertainty and are used for comparative analysis rather than real-world frequency prediction. Under the stated assumptions, layered controls reduce simulated external-incident probability substantially more than network isolation or monitoring used alone, an ordering that holds under independent plus/minus 25% perturbation of every coefficient in the model across 300 draws. Sensitivity analysis shows that agent capability and weaknesses in monitoring, authorization, and credential control exert the greatest influence on modeled risk. Human temporal discounting and metric gaming provide a behavioral analogy for short-horizon optimization, but the study does not infer that AI agents experience gratification or human motivation. The results support treating cyber-capable agent evaluations as hostile security zones in which indirect egress, shared infrastructure, credentials, and evaluation artifacts must remain outside the agent’s effective authority.
[AI-526] meES: Probabilistic and Deterministic Time Series Forecasting via Evolutionary Spectra NEURIPS2026
链接: https://arxiv.org/abs/2609.32384
作者: Weiwei Ye,Renhe Jiang,Hangchen Liu,Dongyuan Li,Yoshihide Sekimoto
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted as NeurIPS 2026 Poster
Abstract:Real-world time series are inherently non-stationary, with trends, periodic patterns, and uncertainty evolving over time. While the Fourier domain offers a natural lens to model time series, current deep learning approaches do not explicitly model evolution and randomness in the Fourier spectra, which limits their ability to accurately predict both the expected trajectory and its uncertainty in non-stationary time series. Motivated by Evolutionary Spectra (ES) theory, we propose TimeES, a general framework that enables probabilistic and deterministic forecasting via the evolutionary spectra theory. Specifically, we derive a parameterizable evolutionary spectra formulation, recasting non-stationary random process modeling as learning an evolving representation modulated by random variables. Furthermore, we reduce the complexity of the estimated spectra from O(NM) to O(NK), where K M/2, by exploiting Hermitian symmetry and spectral energy sparsity for frequency selection. Based on a simple linear backbone, our proposed TimeES achieves consistent state-of-the-art performance across both deterministic and probabilistic forecasting tasks, with high efficiency and interpretability. Code is available at: this https URL.
[AI-527] Measurement Boundaries in LLM Financial Agent Evaluation: Fixed-Tape Execution and Multi-Defect Auditing
链接: https://arxiv.org/abs/2609.32379
作者: Weicheng Xue
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Software Engineering (cs.SE)
备注:
Abstract:What controls are needed to interpret execution performance and audit scores in LLM agent evaluations? We study two limits on these interpretations in a financial agent harness. In Study~A, comparing independent runs under idealized and stressed execution on three synthetic settings that share one 24-day upward phase mixes the execution rule with fresh model responses and portfolio feedback: the parsed decision paths agree in only 19.8% of 450 pairs. Replaying each stored response tape through both execution destinations gives a narrower result. Conditional on those responses, stressed execution changes total return by -0.0170 (95% interval [-0.0230,-0.0117] ), or 10.4% of the idealized baseline, and ten seed clusters do not resolve the model ranking. Study~B corrects an incomplete answer key and replaces legacy tasks with matched zero-, one-, and two-defect tasks under an explicit multi-label prompt. The drop in target violation recall from one to two defects is positive in five of six combinations of auditor and source (median 0.267 ), with three surviving Holm correction. Yet the auditor that includes both target labels most often has micro-precision 0.149 , emits findings on 98/100 zero-defect tasks, and returns the exact dual-defect set in only 21/100 cases. Target recall by itself therefore gives a poor account of audit quality on this construction. The studies address different limits: what an execution comparison estimates, and what target recall captures. Together, they show how fixed conditions and diagnostic controls bound the claims a score can support.
[AI-528] AuthorityLens: Rethinking LLM -Based Agent Systems Through the Lens of Authority
链接: https://arxiv.org/abs/2609.32378
作者: Shaojin Chen,Huihao Jing,Wun Yu Chan,Wenbin Hu,Jiaxing Li,Wu Pandy Pui Ching,Kshitij Bhatia,Xinlei He,Haoran Li,Yangqiu Song
类目: Artificial Intelligence (cs.AI)
备注: 76 pages, 5 figures, including appendices
Abstract:LLM-based agents are increasingly deployed with authority over consequential resources and decisions in real systems. These agents often operate alongside human and LLM-based participants who hold different forms of authority. Yet workflow roles, permission settings, and review mechanisms do not necessarily reflect the authority realized in practice. We introduce AuthorityLens, a framework for measuring a system’s authority structure. Starting from an authority portfolio, we evaluate a system along three dimensions: what the system is authorized to do (System Authority), how much joint participation is required to exercise that authority (Authority Separation), and how much authority each participant holds (Principal Authority). We derive these measurements from the minimal combinations of participants sufficient to realize each outcome across admissible runtime states. We apply AuthorityLens to Codex, OpenCode, and Gemini CLI across 13 operating configurations over a common portfolio of agent operations. We find that nominal configurations do not map cleanly onto realized authority. In Codex, Full Access changes System Authority only marginally while substantially concentrating authority in the executing Assistant. OpenCode’s Build and Plan configurations have the same System Authority and Authority Separation despite different workflows and root-level permissions. In Gemini CLI, model-based review increases Authority Separation without changing System Authority. Principal Authority further distinguishes authority replication from authority separation: spawned or delegated agents can become alternative holders of the same authority without increasing the required joint participation. Together, these results demonstrate that AuthorityLens provides a unified framework for measuring and comparing realized authority structures across agent systems.
[AI-529] DiffPTS: Rethinking Diffusion ELBO for Probabilistic Time Series Forecasting NEURIPS2026
链接: https://arxiv.org/abs/2609.32363
作者: Weiwei Ye,Dongyuan Li,Hangchen Liu,Haotong Jiang,Yoshihide Sekimoto,Renhe Jiang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted as NeurIPS 2026 Poster
Abstract:Probabilistic time series forecasting requires modeling and predicting complex and time-varying distributions. Recently, Denoising Diffusion Probabilistic Model (DDPM)-based approaches have shown promise by equipping the dif- fusion process with pretrained mean and variance estimators to accommodate distributional shift. However, these methods typically follow the standard DDPM framework and consider only partial components of the evidence lower bound (ELBO), treating the training of estimators as designed regression tasks separate from the variational inference framework. To address this, we rethink the ELBO under the Location-Scale Noise Model (LSNM) and find that it naturally induces a Gaussian negative log likelihood objective for the estimators and inherently defines a joint training objective that unifies recent diffusion paradigms for probabilistic forecasting. Building on this principled ELBO reformulation, we propose Diff- PTS, a general framework that enables end-to-end optimization of all components within the ELBO. Across multiple benchmarks, DiffPTS consistently outperforms recent models, achieving state-of-the-art performance with an average CRPS/MSE reduction of over 14.53%/16.55% compared to existing diffusion-based methods. The code is available at this https URL.
[AI-530] From Trajectories to Grounded Preferences: Process Preference Synthesis via Interaction Element Graphs for Web PRMs
链接: https://arxiv.org/abs/2609.32351
作者: Yangzhe Peng,Xiaoyang Wang,Yiyang Zhao,Lijun Wu,Kun He
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Comparative Process Reward Models (PRMs) provide critical step-level guidance for autonomous web agents by evaluating state-conditioned preferences between candidate actions. However, existing preference training data synthesized via multi-policy sampling suffers from a severe scarcity of Grounded Minimal Contrastive Pairs (GMCPs)-where competing candidates target genuine on-page elements with identical action types. In representative baselines preference data (namely, WebArbiter), GMCPs account for merely 24.19%, biasing PRMs during training to rely on shallow shortcuts (such as element hallucinations and action type mismatches) rather than acquiring genuine contextual decision semantics. To address these challenges, we propose SURFPRM, a graph-guided process preference synthesis framework for comparative Web PRMs. SURFPRM structures web demonstrations into a persistent Interaction Element Graph that acts as an environment-grounded negative action proposal mechanism, systematically synthesizing contrastive negative actions across spatial, temporal, and spatiotemporal confusion axes. This elevates the GMCP proportion from 24.19% to 74.60%, producing the curated SURFPRM-DATA dataset. Across six open-source backbones (3B to 9B parameters), PRMs trained on SURFPRM-DATA outperform baseline-trained models on average on WEBPRMBENCH and rival leading proprietary LLMs. In downstream reward-guided trajectory search on WEBARENA-LITE, SURFPRM provides step-level guidance for both GPT-4o (+14.21%) and GPT-4o-mini (+12.83%) policies, yielding substantial improvements in complex web task success rates.
[AI-531] Analog-Friendly Predictive Coding without Activation Derivatives
链接: https://arxiv.org/abs/2609.32350
作者: Francesco Innocenti
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注: 20 pages, 14 figures
Abstract:Predictive coding (PC) is a local, energy-based alternative to backpropagation (BP) whose iterative inference dynamics make it attractive for implementation on analog hardware. However, standard nonlinear PC requires evaluating the derivative of the activation function during both inference and learning, which can be difficult to realise physically. Here, we introduce \textitactivation-matched Bregman PC, replacing standard squared-error energies with Bregman divergences matched to the activation function. This formulation eliminates activation derivatives and, when combined with inference via mirror descent, yields local inference and learning rules requiring only weighted sums, local prediction errors, state integration, and the activation function. In digital experiments, Bregman PC performs comparably to standard PC and BP on classification and generative tasks, while preserving characteristic learning dynamics of PC and its convergence to BP under stable large-model parameterisations. These results provide a more analog-friendly formulation of nonlinear PC while retaining its key computational properties.
[AI-532] ALLOT: Budgeted Hybrid-Memory Routing for Knowledge Updates in LLM s
链接: https://arxiv.org/abs/2609.32344
作者: Shanfeng Huang,Zhou Fang,Song Xiao,Hai Du
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:For large language models (LLMs), parametric adaptation is costly when retrieval already suffices. We introduce ALLOT, a hybrid-memory routing framework that separates learned write priority from a hard parametric budget. A memory-aware router combines frozen text representations, retrieval confidence, and relation metadata; a single ranking supports multiple write budgets while preserving all facts in external memory. On CounterFact with Qwen3-4B, ALLOT reaches 0.760 accuracy at a 20% parametric-write budget and recovers 78.4% of the budget-matched oracle gain, with 80% fewer parametric writes than dual-writing every fact. At this budget, jointly adding retrieval and relation features to text improves normalized oracle gain by 6.2 percentage points. Complementary Qwen3-0.6B shared-store results achieve dual-write-level accuracy with 6-14.5% parametric writes, and cross-benchmark transfer retains approximately 88% of in-domain gain. These results support allocating adaptation capacity according to its incremental value rather than treating every factual update as an equally valuable training target.
[AI-533] HyperReCo: Retrieving and Connecting Evidence with Hypergraph Neural Networks for LLM Multi-hop Reasoning
链接: https://arxiv.org/abs/2609.32327
作者: Zicheng Zhao,Linhao Luo,Junnan Dong,Haoran Luo,Xiaoli Li,Shirui Pan,Chen Gong
类目: Artificial Intelligence (cs.AI)
备注: 22 pages, 6 figures
Abstract:Large language models (LLMs) have shown strong capabilities, with retrieval-augmented generation (RAG) supporting complex multi-hop reasoning by retrieving evidence distributed across documents. Graph-based approaches exploit connections among evidence, and hypergraph-based retrieval further preserves higher-order entity associations within documents and connects documents through shared entities. However, existing hypergraph retrievers often rely on predefined structural expansion or diffusion, which may miss query-dependent interactions needed to identify relevant evidence. They also leave connections among retrieved evidence implicit, requiring LLMs to reconstruct these connections before reasoning. Therefore, we propose HyperReCo, a framework for retrieving and connecting evidence with a hypergraph neural network (HyperGNN). We represent each document as a hyperedge over its extracted entities, with shared entities connecting the hyperedges. Through hypergraph message passing with joint supervision over documents and entities, the HyperGNN learns query-dependent interactions to retrieve complementary evidence. We further introduce Gradient-Guided Hyper-Path Decoding (GGHD), which uses gradient attribution to interpret the learned interactions and translate them into explicit hyper-paths that help LLMs combine complementary facts for multi-hop reasoning. Experiments on six benchmarks show that HyperReCo achieves the best retrieval performance among the compared methods on all three multi-hop QA datasets, together with strong downstream QA performance. Case studies and further analyses demonstrate the utility of decoded hyper-paths for connecting retrieved evidence.
[AI-534] RLHarness: Co-evolving Procedural Skills with Reinforcement Learning for Long-horizon Multimodal Reasoning
链接: https://arxiv.org/abs/2609.32326
作者: Ziqiao Shang,Zian Xu,Ji-Chen Yan,Weiming Wu,Ziyi Jia,Jie Meng,Tao Huang,Shan Huang,Lan-Zhe Guo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multimodal reasoning requires models to preserve visual evidence through long decision chains while selecting appropriate procedures across diverse scenarios and rules. When learning is guided only by terminal verifiers, reinforcement learning (RL) reveals whether a final answer is correct but not how it should be produced. The policy must therefore discover reusable reasoning procedures while learning to execute them, creating a program cold-start problem. Skills can externalize successful procedures, reduce repeated exploration, and provide inspectable guidance. However, a fixed Skill Bank assumes that this guidance remains compatible with an evolving policy, while updating Skills alone can leave their triggers, execution protocols, and demonstrations stale or mutually inconsistent. We introduce RLHARNESS, which organizes Skills, selection and execution protocols, few-shot demonstrations, and task contracts into a unified, versioned Harness and alternates Harness evolution with policy learning. An Exploration-Distillation Harness builds the initial Harness and version-aligned verified traces for SFT and DAPO I. After the first RL block, a Post-RL Reconstruction Harness rebuilds Skills, protocols, and demonstrations from fresh success-failure rollouts, and DAPO II adapts the policy to the reconstructed program. RLHARNESS improves Accuracy from 16.25%/27.50% to 62.00%/50.00% on MetroMap/TravelMap and raises F1 score from 37.13%/45.50% to 65.81%/65.51% on Fee-VL/Cancel-VL. All four tasks achieve their best results only after reconstruction and DAPO II, showing that an evolving Harness complements RL by continually updating the external program that the policy learns to execute.
[AI-535] Active Feature Acquisition With Incomplete Training Data
链接: https://arxiv.org/abs/2609.32325
作者: Reza Rezvan,Valter Schütz,Han Wu,Linus Aronsson,Morteza Haghir Chehreghani
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 24 pages, 8 figures, 2 tables
Abstract:In many prediction tasks, acquiring all features can be a prohibitively expensive or outright impossible task. Further, in many cases a static subset of features may not be enough to solve the problem sufficiently across various instances. Active Feature Acquisition (AFA) addresses these problems by formalizing the trade-off between feature cost and predictive performance during sequential feature selection. However, prior AFA work largely assumes access to complete training data, an assumption that is often violated in practice. Here we study AFA with Incomplete Training Data (AFA-ITD), showing that under missing completely at random (MCAR) data, one-step acquisition values remain unchanged, whereas multi-step values can decrease. We analyze three approaches to learning from incomplete data: aliasing, filtering, and generative restoration. We show that filtering can require a number of training instances scaling exponentially with the dimension, whereas generative restoration scales exponentially with the acquisition budget. We empirically test our theory on a controlled experiment and across common AFA datasets and find that missingness mainly damages methods that exploit multi-step acquisitions and that generative restoration is able to recover lost performance in many experiments. Code is available at this https URL.
[AI-536] Superposed Inference for Hyperdimensional Computing
链接: https://arxiv.org/abs/2609.32320
作者: Quanling Zhao,Nilesh Prasad Pandey,Ye Tian,Tajana Rosing
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Hyperdimensional computing (HDC) is attractive for efficient and robust learning, but conventional inference still encodes every query independently, repeatedly paying the cost of high-dimensional projection. We introduce SupHDC, a new inference paradigm that processes multiple queries through a shared encoding computation. SupHDC assigns lightweight random slot keys, superposes the keyed queries before encoding, and uses slot-specific classifiers to recover their individual predictions. A random-feature kernel view explains why exact recovery of each hypervector is unnecessary: inference only needs to preserve the class evidence that determines the prediction. Across ten datasets, SupHDC achieves 1.39x analytical speedup with no average accuracy loss, and up to 2.08x speedup with only a 2.67 percentage-point mean accuracy loss. On a Raspberry Pi~5, it delivers 2.01x measured wall-clock speedup with a 2.26 percentage-point loss in mean prediction accuracy. SupHDC shows that high-dimensional redundancy can be used not only for robustness, but also as capacity for shared inference.
[AI-537] Phase Space Attention:A Hairer Lift Circumvents the Single-Layer Induction Obstruction
链接: https://arxiv.org/abs/2609.32319
作者: Kingsuk Maitra,Shagun Sood Morteza Hosseini,Suman Gunnala,Vikram Gupta
类目: Machine Learning (cs.LG); Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI)
备注:
Abstract:We circumvent the Sanford-Hsu-Telgarsky (SHT) single-layer induction obstruction within a linear, one-step, causal, bilinear, symplectically consistent design class on the post-RoPE substrate, by lifting attention onto a symplectic phase space, mirroring Hairer’s lift of Stormer-Verlet. The lift exits the premise of the SHT counting argument rather than the bound itself. The reframing exhibits the obstruction as a filter-order gap: a one-layer bilinear score realises a z -transform of joint order (0,0) , whereas the induction discriminator requires key-side order \geq 1 . Applying the symplectic upper shear M_\gamma:(q,p)\mapsto(q+\gamma p,p) to the post-RoPE query and key streams closes it. We prove this lift is unique within the factorised subclass, exactly symplectic at operator level, and requires post-RoPE placement; and in an explicit T_4 -only Gaussian reduction we derive a closed-form two-branch induction phase transition, held out at r=0.9876 with zero fitted parameters. That law is an analytically solvable limit, not a robust prediction: restoring the T_3 channel moves \gamma_c at d_k=64 from 1.030 to 0.569 and removes the crossover. Deployability follows by exact derivation: the KV cache is unchanged, prefix reuse and speculative decoding are preserved, overhead is 6d FLOPs per token per layer, INT8 headroom grows by at most \log_2(1+2\gamma) bits, fused kernels are unmodified, and no parameters are added. At 91.3M parameters a supercritical sweep locates an emergence band: induction forms 3/3 seeds at \gamma=0.80 in a mean of 717 steps, against 2/3 seeds and 2700 steps at \gamma=0 . Adverse results are reported as directly: a key-only half-lift reaches 0.949 against 0.811 for the symmetric operator, so if induction accuracy is the objective, the half-lift is the better construction. Forty-one notebooks and result files ship as ancillary material. Subjects: Machine Learning (cs.LG); Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.32319 [cs.LG] (or arXiv:2609.32319v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.32319 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-538] Delayed Supervision for Test-Time Language Models
链接: https://arxiv.org/abs/2609.32312
作者: Jinha Kim,Taksh Kothari
类目: Artificial Intelligence (cs.AI)
备注: 11 pages, 3 figures. Equal contribution by both authors. Work done at MIT CSAIL
Abstract:Test-time language models adapt a compact memory while processing the input sequence. This perspective encompasses nonlinear fast-weight learning in LaCT, associative delta-rule updates in DeltaNet, and generalized delta-rule state updates in RWKV-7. Training these models to predict the next token does not explicitly require a fact to remain accessible after many subsequent memory updates. We study delayed supervision for this test-time memory: during post-training, ask a simulator-grounded question only after a long interval of unrelated events, and supervise its answer alongside ordinary next-token prediction. Questions are evaluated on disposable branches, so their answers never enter the continuing event stream. The construction distinguishes retention from revision: a retained fact must remain valid throughout the delay, whereas a revised fact must be answered with its latest value. We evaluate this approach on LaCT-760M and plain DeltaNet-1.3B using TextWorld training trajectories and shared BABILong and RULER evaluation panels, and include a separately reported RWKV-7 comparison. Relative to event-only training, delayed QA improves BABILong by 5.48 percentage points for LaCT and 1.32 points for DeltaNet, and single-needle RULER by 1.45 and 3.27 points, respectively. The RWKV-7 comparison reports gains of 4.60 and 7.00 points on its own panels. These results support delayed semantic supervision as a practical outer training objective for usable test-time memory, while leaving open how much of the benefit derives specifically from delay rather than general question-answering and answer-termination supervision. Comments: 11 pages, 3 figures. Equal contribution by both authors. Work done at MIT CSAIL Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.32312 [cs.AI] (or arXiv:2609.32312v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.32312 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Jinha Kim [view email] [v1] Sat, 26 Sep 2026 07:12:37 UTC (66 KB) Full-text links: Access Paper: View a PDF of the paper titled Delayed Supervision for Test-Time Language Models, by Jinha Kim and 1 other authorsView PDFHTML (experimental)TeX Source view license Current browse context: cs.AI prev | next new | recent | 2026-09 Change to browse by: cs References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab) Major funding support from
[AI-539] PhiFold: Towards Dynamic Protein Design with Physics-Structured Covariance Modeling
链接: https://arxiv.org/abs/2609.32309
作者: Yutian Liu,Mujie Lin,LanqianZhang,Meng Fan,Chang Liu,ZhiweiNie,Siwei Ma
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Protein design is moving beyond structural correctness toward function-aware design, yet existing generative models typically treat dynamics as a downstream property estimated through simulation or prediction after structure generation. Using MD trajectories as a generative target is also undesirable because stochastic, path-dependent trajectories over-specify the underlying equilibrium ensemble. We introduce PhiFold, a framework for jointly generating protein backbones and their second-order dynamics, represented by residue-displacement covariance. Rather than predicting the quadratically sized full covariance, PhiFold decomposes dynamics into three interpretable components: local flexibility, a low-rank collective-motion representation, and residue-wise collective participation. These components are assembled into a positive-definite covariance matrix with exact marginal consistency, yielding a compact and physically constrained representation of equilibrium dynamics. Across generated proteins, PhiFold improves recovery of local fluctuations and long-range residue coupling while remaining competitive on dominant collective-motion subspaces. It further enables bidirectional control of residue flexibility while preserving backbone designability. By unifying structure generation with an explicit representation of equilibrium dynamics, PhiFold lays a foundation for designing proteins not only by how they look, but also by how they move.
[AI-540] rain4Merge: A Controlled Single-Teacher Study of RL vs. SFT Teachers for OPD-Based Model Merging
链接: https://arxiv.org/abs/2609.32303
作者: Jingyuan Huang,Zuming Huang,Yucheng Shi,Zhongzhi Li,Xiaoming Zhai,Wei Chu,Ninghao Liu
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Domain experts trained from a shared checkpoint can be merged into one model through on-policy distillation (OPD), where they act as teachers supervising a student on its own trajectories. One upstream choice is rarely examined: whether to build each expert with supervised fine-tuning (SFT) or reinforcement learning (RL). Yet equally strong teachers need not be equally good teachers. We probe this choice through controlled single-teacher OPD, a building block of multi-teacher OPD: in Agentic, Reasoning, and Perception, comparably performing SFT and RL teachers are trained from Qwen3.5-9B, each guiding a student initialized from it. At their best checkpoints, RL-guided students outperform SFT-guided students by 4.27, 1.50, and 0.86 percentage points in Agentic, Reasoning, and Perception, respectively, and recover more of their teachers’ performance gains over the base model. The contrast is clearest in Agentic, where the best SFT-guided student recovers only 44.44% of its teacher’s gain, whereas the best RL-guided student recovers 115.00%, surpassing its teacher. Our analysis points to an explanation: RL teachers stay much closer to the shared initialization in parameter space than SFT teachers and are therefore easier for their students to follow.
[AI-541] Affordance-Conditioned Decision Making: Bridging the Semantic-Spatial Gap in Zero-Shot Cross-Floor Vision-and-Language Navigation
链接: https://arxiv.org/abs/2609.32292
作者: Xuekang Yang,Lu Chen,Shuang Luo,Jialing Zhu,Qi Zhang,Yue Gao,Xiang Zhang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision-and-language navigation increasingly relies on general-purpose semantic planners, yet translating correct high-level intent into reliable physical execution remains difficult in spatially constrained transitions. Reaching a staircase, doorway, or narrow passage does not ensure traversal; the agent must identify an executable affordance pose and recover from accumulated action errors. We propose PACE (Preference-refined Affordance-Conditioned Execution), a supervised local execution module that augments frozen zero-shot semantic planners for reliable cross-floor navigation. PACE grounds transition-related semantics into a long-horizon, agent-centric traversable affordance pose and conditions short-horizon action generation on this spatial target, thereby aligning semantic goals with physical execution. We further post-train PACE through failure-aware preference refinement using rollout-derived pairs that contrast normal or recovery behaviors with deviation-amplifying behaviors, thereby improving closed-loop correction. We integrate PACE into six open-source zero-shot VLN navigators and demonstrate consistent improvements on the cross-floor subsets of R2R-CE and RxR-CE, increasing the average success rate from 16.35% to 27.65% and from 4.76% to 12.06%, respectively. Real-world experiments further demonstrate PACE’s applicability in unseen environments, highlighting the potential of traversable affordances to bridge semantic intent and reliable embodied behavior.
[AI-542] A Solvable Theory of Pre-training Data Poisoning: Regime-Dependent Scaling Exponents
链接: https://arxiv.org/abs/2609.32288
作者: Indranil Halder,Rastri Dey,Cengiz Pehlevan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Pre-training data poisoning of large language models is usually studied using targeted backdoors and their survival through safety post-training, which leaves open a more basic question: how does a model’s clean data performance degrade as the poison rate \varepsilon grows? Motivated by our controlled pre-training runs of OLMo-style models, in which the relative clean data validation perplexity increase \Delta between poisoned and clean models matched in architecture, token budget, and optimization schedule is well fit by a power law \Delta\approx C\varepsilon^a with a non-integer exponent, we ask what such a law requires theoretically. We first prove an analyticity barrier: whenever the contaminated objective depends analytically on \varepsilon around a nondegenerate clean data optimum, \Delta is generically quadratic in \varepsilon , so a generic non-integer exponent is a signature of genuinely singular structure. We then supply that structure in solvable truncated ridge regression with heavy-tailed covariates, controlled by q_\star , and a label-shift poisoning. Our central result is that the excess risk scaling exponent depending on the order of limits: in the higher dimensional proportional regime it is \epsilon^q_\star/(q_\star+2) , whereas taking the ample-data limit first gives \epsilon^2-2/q_\star , and the limits do not commute. We confirm this prediction through several numerical simulations. Finally, we argue that finite training time plays the role of truncation on the curvature spectrum in local LLM pre-training, deriving the observed scaling law under heavy tailed inverse curvature spectrum as a modeling hypothesis.
[AI-543] When Does a Skill Add Value? Task-Conditional Gain Prediction for Selective Skill Use
链接: https://arxiv.org/abs/2609.32274
作者: Anjie Xu,Zhiyu Zhang,Ruiqing Ding,Fengli Xu,Leye Wang
类目: Artificial Intelligence (cs.AI)
备注: 22 pages, 7 figures. Code: this https URL
Abstract:Agent skills are expected to improve task performance. Yet we find that they often provide no benefit, and can even hurt performance while incurring additional token costs. Can we predict whether a skill will help before the agent acts? We introduce SkillDelta, a framework for predicting task-conditional skill gains from paired executions of the same agent with and without the skill. A local predictor transfers these historical gains to new tasks without retraining the agent. Under explicit transfer assumptions, our analysis links support coverage, representation mismatch, and execution noise to prediction error and decision regret. Across five benchmarks and three target agents, paired history improves observed-gain ranking over skill-assisted outcomes alone in 12 of 15 settings. At matched expected skill-use rates, SkillDelta improves success over random activation in all 15 settings, with an average absolute gain of 4.3%. Most of this advantage comes from allocation across task groups. Evidence for additional within-group selection value is strongest on ToolQA and weaker elsewhere. Code is available at this https URL.
[AI-544] LAM: Efficient Lossy Agent Memory Framework With A Retrieval-Score Error Bound
链接: https://arxiv.org/abs/2609.32256
作者: Baixi Sun,Le Chen,Anjir Ahmed Chowdhury,Xiaolong Ma,Chih-Hsuan Yang,Mingze Xia,Syed Zawad,Sheng Di,Rajkumar Kettimuthu,Huihuo Zheng,Rajeev Thakur,Venkatram Vishwanath,Feng Yan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Agent memory grows as agents read inputs, reason, and call tools. Longer histories increase inference cost and eventually exceed the context window. LLM-based summarization reduces this history but adds latency and provides no explicit bound on information loss. We propose LAM, a Lossy Agent Memory system with three components: a deterministic deduplication rule with a substitution bound on retrieval scores - a bound on score perturbation, not a certificate of unchanged ranking; a memory manager that preserves the cached prefix and overlaps compaction with inference; and a performance model that estimates compaction costs before deployment. On 600 agent trajectories, LAM removes 22.47% of observation tokens while retaining 99.984% of the measured gold-patch evidence. At a fixed deletion set, the performance model predicts a 71.4x-91.6x end-to-end speedup from removing records before prefill instead of deleting them from a prefilled context. That benefit comes from the schedule rather than the rule and applies to any prefix-preserving test.
[AI-545] Why Directly Learning Periodic Trajectories Can Fail
链接: https://arxiv.org/abs/2609.32254
作者: Kaixin Zheng,Anita Layton
类目: Artificial Intelligence (cs.AI); Classical Analysis and ODEs (math.CA); Numerical Analysis (math.NA)
备注:
Abstract:Operator learning of periodic solutions requires deciding how simulation data should be recorded and represented. A natural choice is to integrate long enough for transients to decay and record a window wide enough to contain at least one full period of all trajectories. We find that these conservative choices can make the resulting trajectories difficult to learn, even when the underlying periodic orbits vary regularly with system parameters. Unaligned trajectories generalize poorly even within the training distribution. Phase alignment substantially improves in-distribution generalization, but models trained on a fixed physical-time window still have large errors on trajectories with periods outside the training range. We explain both failures through a common mechanism: frequency differences accumulate over time, so the target phase varies rapidly with the parameters. Predictors that cannot track this variation incur a population MSE floor in both settings; for fixed window prediction, we also derive a per-sample lower bound. We then study one of the simplest representations that escape these floors: learning an aligned, normalized waveform and its period separately. We establish regularity of the decoupled targets under ODE assumptions and show experimentally that this approach avoids both failures in ODE systems and a PDE case study.
[AI-546] DS-VLA: A Dendritic-inspired Vision-Language-Action Model for Robust Action Control
链接: https://arxiv.org/abs/2609.32253
作者: Yaxing Lyu,Jingyi Li,Mingkun Xu,Yujie Wu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Vision-language-action (VLA) models have achieved strong performance in language-conditioned manipulation, yet success under nominal evaluation does not necessarily translate into robust closed-loop behavior when executed actions are transiently corrupted. We introduce DS-VLA, a dendritic-inspired action architecture that incorporates dendritic spiking dynamics into VLA control to address this limitation. Specifically, to enable modularized feature processing and temporal information integration, DS-VLA equips action neurons with multiple sparsely connected dendritic branches, each featuring heterogeneous, learned decay factors. Furthermore, to suppress unreliable state updates while preserving task-relevant historical information, we introduce a neuron-wise inhibitory gate that adaptively regulates the admission of new multimodal evidence into dendritic states prior to somatic dynamics. We evaluate DS-VLA on all four LIBERO suites under both nominal rollouts and a unified closed-loop action-perturbation protocol. DS-VLA achieves a 91.6% average nominal success rate and an 87.35% average perturbed success rate, retaining 95.4% of its nominal performance. Under the same reported perturbation setting, OpenVLA-OFT, FAST, \pi_0 , and GR00T achieve 39.45%, 23.90%, 28.55%, and 30.75%, respectively. A controlled ablation isolates the contribution of neuron-wise shared inhibition, while analyses of neural dynamics and post-perturbation trajectories associate robust performance with selective evidence suppression and effective behavioral recovery. Together, these results demonstrate that integrating brain-inspired computational mechanisms offers a promising architectural prior for robust embodied intelligence beyond merely scaling vision-language backbones or generative action decoders.
[AI-547] Certifying Interventional Agreement Among Observationally Equivalent Causal Models
链接: https://arxiv.org/abs/2609.32247
作者: Sourena Khanzadeh,Daniel Platnick,Marjan Alirezaie,Hossein Rahnama
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Observationally equivalent causal models can still disagree about what happens under intervention, because interventions create inputs that never occur in observational data. We introduce Interventional Separation Selection (ISS), which repeatedly queries the true system with an admissible intervention on which the surviving candidate models disagree, discards the candidates the outcome contradicts, and stops once no intervention within a cost bound separates the survivors. If the true system is among the candidates, this stopping condition certifies that every survivor agrees with it on every admissible intervention within the bound, a guarantee that no observational learner can give, however much data it sees. The stopping condition depends only on the survivors, so it can be checked without knowing the truth. For continuous variables the candidates form an infinite version space, and mixed-integer linear programs decide the stopping condition exactly over all of it, with agreement holding up to a tolerance. On a three-digit colored MNIST causal abstraction task in which ink hue tracks digit size, plain convolutional networks trained on examples reach zero held-out error, yet disagree with shape-based labels on 26% of single-digit edits, as often as hue-based labels do. Auditing the causal abstractions of networks observed only on such images, ISS certifies what each network perceives with 13.6 interventions per image on average, and each certificate, checked against every admissible intervention, holds whenever the network’s true abstraction is among the candidates. When a network bypasses a unit that every candidate abstraction relies on, certificates covering interventions on that unit can be silently void, and twenty random validation interventions refute 69% of them.
[AI-548] AutoPDEBench: Benchmarking LLM Auto-Research for Neural PDE Solver Design
链接: https://arxiv.org/abs/2609.32245
作者: Ruoyan Li,Wei Wang,Yizhou Sun
类目: Computational Engineering, Finance, and Science (cs.CE); Artificial Intelligence (cs.AI)
备注:
Abstract:Partial differential equations (PDEs) are essential for modeling complex physical systems, and neural solvers have recently emerged as powerful data-driven tools for numerically solving them. However, existing neural solvers struggle with domain-specific challenges, such as varying parameters and high-speed flows, necessitating specialized architectures. Manually designing these specialized solver architectures is a highly iterative, time-consuming process requiring deep expertise, creating a significant bottleneck in scientific discovery. We propose leveraging autonomous AI research agents to automate the synthesis of specialized solvers. To support this, we introduce AutoPDEBench, a benchmark dedicated to LLM-driven automated research for PDE solver design. The benchmark includes 25 challenging datasets featuring both novel and actively studied physical scenarios. We evaluate a suite of general-purpose models (transformer, ROM, and graph-based) alongside a multi-agent instantiation of the iterative automated research pipeline, which serves as an agentic baseline. Empirical results show that the iterative automated research system significantly outperforms the general-purpose neural solver baselines. Our findings demonstrate the viability of using AI agents to automatically design neural solvers for complex physical systems. AutoPDEBench provides a foundational testbed to accelerate agent-driven scientific discovery in physics and engineering.
[AI-549] oward Agent ic Optical Networks: A Vision of LLM Agent -Driven Autonomous Lifecycle Management
链接: https://arxiv.org/abs/2609.32226
作者: Yao Zhang,Shengnan Li,Yuchen Song,Yidi Wang,Yue Pang,Wenbin Chen,Xiaotian Jiang,Xiao Luo,Meixia Fu,Min Zhang,Yongli Zhao,Shanguo Huang,Alan Pak Tao Lau,Danshi Wang
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI)
备注:
Abstract:As optical networks continue to expand in scale, complexity, and service diversity, the implementation of automation has become essential for ensuring agility, efficiency, and reliability in lifecycle management (LCM) of optical networks. Large language model (LLM) Agent, distinguished by its progressively sophisticated capabilities in logical reasoning, adaptive decision-making, complex problem solving, and multi-task orchestration, presents great opportunities to advance network automation beyond traditional AI techniques. Nevertheless, the application of LLM Agent in optical networks remains in its early exploratory stage, challenged by the lack of multi-task coordination, high computational demands, data dependence, and reliability concerns. In this paper, we envision a conceptual roadmap toward Agentic Optical Networks (AONs) by integrating LLM Agents throughout the LCM with high-level autonomy. First, we trace the evolution from manual operations to AI-empowered frameworks and distill key technologies in Agent, providing actionable insights into leveraging its strengths for addressing practical network automation challenges. A core contribution of this paper is the proposal of a hierarchical multi-Agent framework, which is specifically developed to manage every phase in LCM of AONs, including planning, deployment, operation, maintenance, upgrade, and decommission, thereby enabling more cohesive and comprehensive automation throughout the entire lifecycle. In addition, future directions and underlying challenges are also discussed at the intersection of LLM and optical networks. By aligning the LLM Agent with the specialized requirements of AONs, this work aims to explore the potential for the evolution of optical networks moving from task-level semi-automatic execution toward lifecycle-level full autonomy.
[AI-550] RAO-Nav: Probing Omni-Language Models for Zero-shot Semantic Audio-Visual Navigation
链接: https://arxiv.org/abs/2609.32224
作者: Qilang Ye,Meng Liu,Yu Zhou
类目: Artificial Intelligence (cs.AI)
备注: Accepted by NeuraIPS 2026
Abstract:We explore whether Omni-Language Models (OLMs) can be directly applied to zero-shot Semantic Audio-Visual Navigation (SAVN). Recent work demonstrates that even state-of-the-art specialized models still struggle to achieve generalist multimodal navigation, despite extensive task-specific training. In this paper, we introduce RAO-Nav, short for Reasoning All-in-One OLM, a deployment pipeline for zero-shot SAVN. By leveraging the rich implicit audio-visual knowledge encoded in OLMs, the embodied agent is enabled to hear'', see’‘, reason'', and act’’ in the environment. To further elicit the built-in thinking ability of OLMs, we propose a test-time Latent Navigation Reasoning (LNR) module that can be seamlessly integrated into the decoding space. LNR encourages the model to retrieve more target-relevant observations and make effective navigation decisions. Through comprehensive experiments, we show that our framework surpasses existing state-of-the-art baselines on public SAVN benchmarks without using any training data. Moreover, we introduce a new \emphGlobal Navigation Instruction setting to further evaluate the ability of OLMs to serve as embodied navigation agents. Code: this https URL_Nav.
[AI-551] A bilingual AI audiologist built through rubric-guided playbook induction outperforms human audiologists in a blinded evaluation of simulated cases
链接: https://arxiv.org/abs/2609.32220
作者: Linkai Li,Changgeng Mo,Hanlin Yu,Congxi Lu,Shangqiguo Wang,Matthew B Fitzgerald,Shan X Wang
类目: Artificial Intelligence (cs.AI)
备注: 75 pages, including 51 pages of Supplementary Information
Abstract:Audiology consultation requires structured history-taking, audiometric interpretation and patient-centred communication, yet real-world case material is scarce. We present a bilingual AI audiologist pairing a general-purpose large language model with rubric-guided playbook induction, multimodal audiogram interpretation and retrieval-augmented grounding, without fine-tuning the language-model backbone. Using a 21-item rubric and an AI patient simulator, we induced a 19-rule consultation policy from 73 training cases (43 English, 30 Chinese) and evaluated the system on 58 independent simulated cases (30 Chinese, 28 English) in a pre-specified, source-blinded comparison with 17 practising audiologists. The AI audiologist outperformed human audiologists on every case (58/58; mean paired \Delta = +1.35 on a 5-point composite, Cohen’s d = 1.84, P = 4.5 \times 10^-20 ), on 20 of 21 rubric items and in both languages. Component ablation identified the playbook as the largest contributor, offering a practical route to specialist consultation agents in low-data medical domains.
[AI-552] Fracast-0: Fractal Weight Sharing for a Time Series Foundation Model with Only 85K Parameters
链接: https://arxiv.org/abs/2609.32209
作者: Tianxiang Zhan,Huanyao Zhang,Yuanpeng He
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Time series foundation models must preserve multi-domain breadth, probabilistic output, and multiple temporal scales, but parameter count grows when each scale receives a separate representation. We introduce Fracast-0, a probabilistic forecasting foundation model that exploits temporal self-similarity to reuse one operator across scales. A parameter-free detector extracts significant seasonal structure. The encoder applies a shared local block along a geometric dilation ladder with scale conditioning, while the decoder combines context-gathered states with an explicit seasonal future state and reuses a second block along another ladder before emitting nine quantiles. Pretraining across six corpora preserves multi-domain breadth within 85,001 parameters. On 97 GIFT-Eval configurations without per-dataset fine-tuning, Fracast-0 is the smallest of 28 evaluated checkpoints and remains non-dominated in the aggregate parameter-accuracy plane with MASE 0.808 and WQL 0.564. It uses 42.0% fewer parameters than TinyCast, whose MASE and WQL are 4.2% and 3.3% lower. These results support cross-scale weight reuse as a practical route to further time series foundation model compression.
[AI-553] Witness: Discovery Deciphering and Epiphany in Interactive Puzzle Environments
链接: https://arxiv.org/abs/2609.32208
作者: Guanghan Ning,Ping Liu,Linyi Li,Huangjie Zheng,Arjun Neervannan,Huu Nguyen,Michael Sklar,Deniz Zorlu,Nicolai Ouporov
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Automated science needs agents that can work out the rules of an unfamiliar environment by interacting with it. Interactive rule-discovery puzzles offer a controlled setting for studying this ability: an agent infers hidden rules through experimentation and uses what it has inferred to reach a stated goal. We ask what limits current language models on these puzzles and whether reinforcement learning (RL) improves performance on rules held out from training. To study both, we introduce WITNESS, a 2D grid-based puzzle environment with ground-truth ASCII observations and controlled access to rules. An agentic pipeline generates games for WitnessGym, the RL training suite, and WitnessBench, comprising public validation and private test games. The validation set separately tests new compositions of trained rule primitives and primitives absent from training. Under a shared harness, the best of 18 frontier proprietary and open-weight models solves only 24% of private test level slots, with scores sensitive to the observation interface and agent configuration. Providing ground-truth rules raises Opus-5’s validation RHAE-L5 (relative human action efficiency over the first five levels) from 59.9 to 97.8, whereas a 27B open-weight model gains only 2.1 points and remains limited even with the rules provided. RL on WitnessGym raises the 27B model’s private test RHAE-L5 from 2.1 to 5.4 and yields a mean gain of 4.1 points on four external discovery benchmarks. Together, these results point to rule acquisition as a major difficulty for frontier models like Opus-5 while smaller models further struggle on rule-based execution, and indicate that RL on hidden-rule puzzles transfers to broader rules and real-world tasks beyond training. Benchmark is available at: this https URL
[AI-554] Instruct Not Answer: Using Instruction Privileges in On-Policy Context Distillation
链接: https://arxiv.org/abs/2609.32201
作者: Hantao Yu,Sandy Han,Udaya Ghai,Ferhat Erata,Joe Lilien,Aman Goel,Ali Torkamani
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:On-Policy Context Distillation (OPCD) has recently emerged as a powerful technique for transferring context to student models and for self-improvement. In OPCD, the teacher is conditioned on privileged information, and the goal is to minimize the Kullback-Leibler (KL) divergence between the privileged teacher and the student, evaluated on student-generated tokens. Many existing studies show that using instance-specific gold answers or gold demonstrations as the default privilege can hurt training performance, especially out-of-distribution (OOD). In this work, we instead design general instructions that target common student mistakes observed on the training samples, and show that such simple instructions can outperform gold as the OPCD privilege. In autoformalization tasks, using a matched formatting instruction as the privilege could outperform gold in OOD accuracy by a large margin. In 7 out of 8 experiments using ProverQA, ProofWriter, and ProntoQA as datasets, and Qwen3-Thinking and Olmo3-Thinking families as models, matched instruction privileges outperform gold in OOD by 4 to 17 points, while remaining on par with gold in-domain. Each instruction is only a few sentences (and thus contains much less information compared to all instance-specific gold) and is applied uniformly to every training sample. These results indicate that a general instruction, which applies equally to source and target domain examples, can be substantially more transferable than instance-specific gold in OPCD while maintaining in-domain performance.
[AI-555] Agents as Software: A Programming Languages Agenda for Agent Reliability
链接: https://arxiv.org/abs/2609.32198
作者: Shraddha Barke,Adithya Murali
类目: Programming Languages (cs.PL); Artificial Intelligence (cs.AI)
备注: Accepted to Onward! at SPLASH 2026
Abstract:AI agents increasingly resemble software systems: they call tools, remember facts, follow policies, delegate work, and take actions with real consequences. % Yet the ``program’’ of an agent is scattered across prompts, tools, memories, workflows, and execution traces, making its behavior difficult to inspect through ordinary testing and debugging alone. % This essay argues that a programming-systems perspective offers a natural lens for making agents reliable. % We recast agents as programmable artifacts whose behavior can be specified over traces and state, checked before deployment, monitored during execution, and improved from observed failures. % The goal is not to make probabilistic agents behave like deterministic programs, but to give them enough structure that their behavior can be reasoned about, controlled, and repaired.
[AI-556] CoMemBench: Benchmarking Collaborative Memory Boundaries across Multi-Agent Workflow Topologies
链接: https://arxiv.org/abs/2609.32192
作者: Sen Zhao,Ruiqi Kong,Zuyu Zhang,Lifeng Shen,Xinyu He,Ding Zou,Xu Zhang,Qinghua Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-agent workflows require task-relevant information to be shared across agents, while irrelevant, stale, unverified, or incompatible information must remain isolated. We call this task-conditioned scope of information a collaborative memory boundary. Workflow topology determines which intermediate artifacts are applicable to which downstream workers and when they cease to be valid, thereby providing a structural stress dimension for sharing and isolation. Existing memory benchmarks primarily evaluate retention and retrieval, whereas multi-agent benchmarks emphasize coordination and end-to-end completion, leaving topology-conditioned memory boundaries largely unmeasured. We introduce CoMemBench, an execution-grounded benchmark for collaborative memory sharing and isolation across multi-agent workflow topologies. It constructs 800 composite workflows across four domains from source-grounded dependency graphs, with node-local specifications, verifiable artifact handoffs, native evaluators, and matched isolation challenges. CoMemBench measures workflow completion, verified node progress, required-handoff reliability, isolation robustness, and token cost. Experiments reveal a sharing-isolation trade-off: broader context improves information availability but can weaken isolation, while system rankings shift across topologies and artifact violations.
[AI-557] AI Harness: Certification under Proposal-Conditioned Information for Foundation-Model Agents
链接: https://arxiv.org/abs/2609.32184
作者: Hailin Zhong,Shengxin Zhu
类目: Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注:
Abstract:Foundation-model agents are often modeled as policies over an observed state. In deployed systems, however, a runtime may intervene only after the model has emitted a semantic proposal, making the proposal both an action candidate and a decision-time observation generated by a history-conditioned process. We show that collapsing this structure into a state-only proposal envelope can preserve proposal coverage while destroying certifiability. In a finite robust interface, the viability kernel of the collapsed model is contained in the physical projection of the history-augmented kernel, and the collapse is lossless exactly when every proposal-conditioned collapsed fiber retains a common robust-safe intervention. This gap can be maximal even with constant-size proposal and history alphabets. The same common-action condition yields a dual result: observing the current proposal can restore robust feasibility when it separates latent modes requiring incompatible interventions. We extend these one-step results over time using exact finite beliefs and standard safety and reachability fixed points, separating indefinite operational viability from finite worst-case verified progress. Controlled model-in-the-loop tests reproduce the predicted obstructions when telemetry or effect verification is removed or intervention authority is restricted. Thus, our contribution is not a new fixed-point calculus, but a characterization of when proposal–history correlation at the model–tool boundary is necessary for certification.
[AI-558] Noisy Test-Time Reinforcement Learning for Code LLM s EMNLP2026
链接: https://arxiv.org/abs/2609.32172
作者: Xikai Yang,Hieu Trung Nguyen,Dunyuan Xu,Yuzhi Zhao,Jinpeng Li,Wenao Ma,Pheng-Ann Heng
类目: Artificial Intelligence (cs.AI)
备注: This paper has been accepted by EMNLP 2026
Abstract:Large language models (LLMs) have demonstrated remarkable performance across various code-related tasks. However, unlike carefully curated datasets that are typically high-quality and error-free, real-world user instructions are often vague and error-prone, posing significant challenges to the robustness of code LLMs. Furthermore, robustness-oriented fine-tuning relies on paired clean-noisy samples, which are costly to curate and require sophisticated noisy simulation techniques. To address these challenges, we propose the Noisy Test-time Reinforcement Learning framework (NTRL-Code), which enables robust self-evolution of code LLMs using only unlabeled noisy data during the testing stage. Specifically, NTRL-Code uses conservative self-denoising to obtain a cleaner semantic anchor for target estimation, and employs an abstract-syntax-tree (AST)-based structural aggregation mechanism to estimate a proxy target from multiple candidate programs. The policy is then optimized on the original noisy prompts with a hybrid reward that combines format validity, code similarity, and anti-repetition signals. Extensive experiments on three benchmarks, each incorporating character-level, word-level, and paragraph-level perturbations, demonstrate that NTRL-Code yields robust and consistent improvements, stabilizing the predictions of various base models. Our code is available at this https URL.
[AI-559] Uncertainty-Aware Selection of Online Algorithms with Simulator Ensembles
链接: https://arxiv.org/abs/2609.32170
作者: Yongyi Guo,Zifan Xu,Ziping Xu,Kelly W. Zhang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:The performance of online reinforcement learning depends critically on design choices, especially those that affect exploration. These choices are often selected by fitting a simulator to offline data, evaluating candidate algorithms in that simulator, and deploying the best-performing one. The simplest Plug-In selection rule simply selects the best performing algorithm on the fitted simulator, making evaluations unreliable when the offline data used to fit the simulator are limited. We investigate Uncertainty-Aware selection, which forms an ensemble of simulators—for example, obtained by bootstrap resampling—and selects the online algorithm with the best average performance across the ensemble. While ensemble-based approaches have been used to mitigate distribution shift and facilitate sim-to-real transfer, we formally show that this approach can mitigate the effects of limited data when fitting the simulator and theoretically has significant regret gains compared to Plug-In selection in multi-armed bandits. We also empirically investigate the Uncertainty-Aware selection approach in deep RL experiments on robotic control tasks that involve selecting reward-shaping hyperparameters, and show that it leads to more reliable selection and improved online performance.
[AI-560] PastForward: Faster On-Device GUI Agents via Computational Experience Reuse
链接: https://arxiv.org/abs/2609.32166
作者: Taehwan Park,Changmin Lee,Hayeon Lee,Taesik Gong
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Running GUI agents on edge devices can keep sensitive screens and interaction histories local, but the computational cost of inference at every action step makes deployment challenging. Existing GUI agent systems either perform full vision-language model (VLM) inference at each action step or reuse coarse-grained knowledge matched to prior tasks. However, dynamic mobile environments and user tasks make it difficult to fully utilize prior task executions without additional fine-tuning or task-specific offline exploration. To address this challenge, we present PastForward, a system that accelerates GUI agents through validated, fine-grained reuse of computational experience accumulated during ordinary task execution. During decoding, PastForward retrieves prior output sequences as device-adaptive multi-token proposals and verifies them in a single VLM forward pass. Across action steps, it uses prior GUI transitions to begin next-step inference while the device executes the current action, retains the early computation only when the predicted screen matches the observed screen, and carries reusable KV states forward. We evaluate PastForward on AndroidWorld workloads derived from real mobile usage patterns using multiple VLM backbones across server and edge platforms. On device, PastForward achieves action-step latency speedups of 1.63-2.36 \times while maintaining task success rates.
[AI-561] Spectral Reversal: Counteracting Singular Value Bias for Graph Prompting NEURIPS2026
链接: https://arxiv.org/abs/2609.32143
作者: Hanxu Yang,Yuhuan Zhao,Xiaodong He,Zhao Kang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Social and Information Networks (cs.SI)
备注: Accepted by NeurIPS 2026 as a spotlight paper
Abstract:Pre-training Graph Neural Networks (GNNs) via self-supervised learning has become a dominant paradigm, yet efficiently adapting frozen encoders remains a challenge. Graph prompting offers a parameter-efficient alternative to fine-tuning, but existing methods largely treat pre-trained models as opaque feature extractors, ignoring their internal spectral structure. In this work, we identify a systematic phenomenon in pre-trained GNNs, which we term spectral bias: optimization during pre-training disproportionately aligns representations with directions associated with large singular values, leaving low-energy directions under-explored. We show that these underutilized directions can encode complementary information that is beneficial for downstream adaptation, especially under distribution shift. To leverage this insight, we propose Spectral Reverse Prompt (SRP), a prompting framework that rebalances the spectral contributions of frozen GNN encoders. SRP applies a learnable soft-thresholding mask in the spectral domain to down-weight dominant directions while amplifying weaker ones. In addition, SRP incorporates a null-space augmentation module that captures variation in directions with minimal activation under the frozen encoder. Extensive experiments across multiple benchmarks demonstrate that SRP achieves state-of-the-art performance with minimal additional parameters, highlighting that reweighting spectral components is a principled and effective strategy for parameter-efficient graph adaptation.
[AI-562] REALM: Regime-Switching Explainable and Activation-Induced Linear Models
链接: https://arxiv.org/abs/2609.32141
作者: Xiaoran Cheng,Sen Na,Jia Li
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Methodology (stat.ME); Machine Learning (stat.ML)
备注: 18 pages, 12 figures, 3 tables
Abstract:Deep ReLU networks are piecewise-affine mappings that partition the input space into cells, each characterized by a distinct activation pattern. This structure motivates fitting a local linear model within each cell to preserve predictive accuracy while improving interpretability. The challenge is to identify regimes that are stable, data-adaptive, and easy to explain. We propose REALM, a mixture of linear models whose regimes are induced by neural activation patterns. Because the number of activation cells in a deep neural network (DNN) can grow rapidly with depth, we first distill a deep teacher into a wide, shallow student network (WSSN), then binarize and cluster its hidden-layer activations to define the regimes and fit a linear model within each regime. Since the regimes are discovered from internal structure, the router does not carry the predictive burden. To make regime assignment interpretable, we train a multiclass logistic regression, the explanatory gate, to reproduce the regime assignments. The two-level structure is interpretable at both stages in terms of raw tabular or learned convolutional features: the gate identifies features that determine regime assignments, while the linear models identify features that drive predictions within each regime. We analyze an idealized setting that illustrates a trade-off between partition complexity and stability: as the number of regimes grows, finer partitions can improve approximation but may reduce regime-assignment stability. Experiments on tabular and image datasets show that REALM achieves competitive predictive performance relative to other DNN-guided mixture surrogates and inherently interpretable models while producing stable regime-level explanations.
[AI-563] READ-Bench: Benchmarking Historical Instance Retrieval for Time-Series Diagnosis
链接: https://arxiv.org/abs/2609.32123
作者: Gerardo Pastrana,Haojun Li,Dhruv Mehta,Anoushka Vyas,Sina Khoshfetrat Pakazad,Henrik Ohlsson,John Paparrizos
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Time-series diagnostic systems rarely rely on retrieving relevant historical cases, and when they do, retrieval is evaluated only indirectly through downstream prediction. We introduce READ-Bench, a benchmark for historical-case retrieval across 12 diagnostic datasets, centered on multivariate time series, that defines relevance by shared fault or event type rather than signal shape, so visually different traces of the same fault count as relevant while similar-looking traces of different faults do not. Treating retrieval as a base retriever followed by a reranker, we evaluate classical distances, symbolic retrievers, self-supervised and foundation-model embedders, and their fusion, plus label-aware and language-model rerankers, under one protocol that varies supervision, pollution, and corpus scale with significance testing. Under a common channel-independent interface, pretrained representations offer no statistically detectable advantage over strong classical and symbolic baselines for search alone. The decisive factor is a small amount of resolved-case supervision at reranking, namely a Gaussian-process reranker that propagates a few neighbor labels in embedding space, which helps far more than more sophisticated representations or language-model reasoning and holds under pollution and at full corpus scale. Guided by these findings, we fuse a normal-residual-scored embedder with a dynamic time warping leg via reciprocal-rank fusion, then rerank with the Gaussian-process reranker, improving NDCG@10 over its own search stage on all 12 datasets, by +0.11 from reranking and +0.16 over the strongest single base retriever.
[AI-564] Escaping Alignment: A Physical Trap Model of Best-of-N Jailbreaking
链接: https://arxiv.org/abs/2609.32116
作者: Marco Biroli
类目: Artificial Intelligence (cs.AI); Statistical Mechanics (cond-mat.stat-mech)
备注:
Abstract:Best-of- N jailbreaking (BoN) bypasses safeguards of aligned models by drawing N independent augmentations of an unsafe prompt and sampling M completions of each. Previous works have shown that the attack success rate (ASR) seems to follow a power-law in N , which we challenge. The exponent drifts with N , with an exponential crossover which is a finite-size artifact of the adversarial dataset. Little work has been done to explore the entire two-budget ( N, M ) attack surface as well as its dependence on the generation temperature T . We introduce a simple barrier model where each prompt has a baseline safety level and each augmentation a random thermally activated barrier. Then four numbers, each backed by an interpretable safety mechanism, determine the entire ( N, M ) attack surface. They extrapolate predictions from N \leq 100 to N = 10^4 , collapse five distinct models on the same scaling function and predict ASR at different temperatures from the one they were fitted at.
[AI-565] Empowering Hybrid Attention Models on NPUs
链接: https://arxiv.org/abs/2609.32114
作者: Yinyuan Zhang,Daliang Xu,Xiaolong Huang,Wangsong Yin,Yun Ma,Mengwei Xu,Gang Huang
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI)
备注: 16 pages, 18 figures
Abstract:Hybrid attention models have emerged as a crucial architecture for Large Language Models (LLMs) (e.g., the Qwen3.5 and Kimi series). Their memory and computational efficiency make them highly attractive for on-device inference, forming a promising synergy with edge Neural Processing Units (NPUs). However, naive execution of these hybrid models on edge NPUs fails to deliver these benefits, often bottlenecking the prefill stage due to severe memory-system inefficiencies and architectural mismatches within the linear attention (LA) layers. We present HA-NPU, the first system to enable efficient hybrid attention LLM inference on edge NPUs without modifying the underlying algorithms. HA-NPU enhances execution efficiency by reorganizing the dataflow of the LA components across three levels: (1) At the core level, it partitions workloads by the head dimension and fuses dependent operators, eliminating cross-core global memory accesses; (2) At the operator level, it reorders execution to consume intermediate tensors immediately, drastically minimizing local-buffer pressure; (3) At the tensor level, it employs dataflow-aware layout planning to minimize transformation overhead between matrix and vector processing units. Compared to competitive baselines, HA-NPU achieves up to 35.95 \times LA kernel speedup and 36.14 \times energy reduction, delivering up to 2.03 \times faster end-to-end request latency. The source code will be made publicly available at this https URL
[AI-566] Residual Streams Read Recurrent States Remember: The Global Workspace in Mamba Models
链接: https://arxiv.org/abs/2609.32102
作者: Wenlong Wang,Fergal Reid
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Can the global-workspace account of transformer representations extend to state-space language models? We fit Jacobian lenses to the residual streams and recurrent states of Mamba-1, Mamba-2 and Mamba-3, using the original 1000-prompt recipe. Joint residual–state readouts improve recovery of known intermediate concepts over the residual lens on at least five of six task families in every tested Mamba checkpoint. On Mamba-2, state alone exceeds residual and logit lenses on all six families; a normalised joint readout improves on both components on five. Temporal maps and word-list experiments show earlier content remaining state-readable as residual visibility changes. We also propose sign-guarded steering, which improves target top-five success over coordinate exchange on matched verbal-report trials in five models. Recurrent state alone supports this verbal access. These gains do not extend consistently to relational answers: guarded edits often output the edited concept itself, and Mamba-3’s joint edits can disrupt successful state-only redirection. Recurrent state thus provides a complementary carrier of workspace content, whose recovery, persistence and causal uses require separate measurements.
[AI-567] Emergent One-Third Scaling Law as Attention Tries to Concentrate
链接: https://arxiv.org/abs/2609.32100
作者: Yizhou Liu,Sara Kangaslahti,Jeff Gore
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: 32 pages, 16 figures
Abstract:The neural scaling law relating longer training to better performance through a power law is central to today’s large language models (LLMs), yet its origin remains debated. One recent proposal is that power laws can emerge from the strong non-linearity of a single softmax head learning peaked distributions. What happens with multiple softmax functions, as in LLMs, is unclear. Here, we show through toy models that any softmax learning peaked distributions, regardless of its position in the model, can develop logit magnitudes that grow in a power law with exponent 1/3 , becoming a training bottleneck whose loss contribution decays as a power law with the same exponent 1/3 . The overall loss therefore obeys 1/3 scaling whenever at least one softmax learns peaked distributions. We confirm that many softmax functions in LLMs learn peaked distributions and that LLM loss scaling matches this 1/3 prediction. Moreover, logit growth dynamics reveal that attention heads, rather than the language modeling head, are the bottleneck likely driving the 1/3 loss scaling in LLMs. Attention trying to concentrate on specific information, which is the heart of Transformers, may therefore also be the heart of the neural scaling law of training.
[AI-568] GameBoyWorlds: A Testbed for Self-Improvement in Embodied Video Games
链接: https://arxiv.org/abs/2609.32093
作者: Dhananjay Ashok,Adam Shen,Aslan Huo Feng,Chinmay Khanna,Jun Rui Huang,Raghav Sarmukaddam,Surendira Balaji Natarajan,Xiaotong Cui,Xincan Zhang,Thomson Yen,Hongseok Namkoong,Jonathan May,Jesse Thomason
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Powered by expert guidance, agents can operate in interactive environments; however, it is unclear whether they can learn autonomously from their own experience. To evaluate such self-improvement methods, we introduce GameBoyWorlds, a testbed for agentic self-improvement in video games. GameBoyWorlds-Execution evaluates task execution on a collection of 5 distinct game series. Agents are allowed access to dedicated training games but are provided no demonstrations, documentation, or rewards. Agents must ground themselves in the environment through self-directed exploration and by inferring actionable knowledge from their own experience. At test time, agents must complete short-horizon tasks that evaluate their ability to navigate, interact, and engage with game-specific mechanics in unseen games. Out-of-the-box frontier models complete fewer than 50% of the 500 tasks due to failures in multimodal grounding, establishing that self-improvement methods have room to push performance. We demonstrate that contemporary approaches to self-improvement are lacking, with world modelling and autonomous skill discovery failing, and a novel strategy that uses curiosity-based exploration to write guides achieving only partial success. GameBoyWorlds-Playthrough tests end-to-end game completion in two fan-made Pokémon games. We show that while frontier models have been pre-exposed to official releases such as Pokémon Red, they lack essential information on the games in our testbed. Instead of relying on their parametric knowledge to succeed, agents must learn from their own experience and autonomously improve over the course of the playthrough. We show that a sophisticated agentic pipeline with multimodal memory and hierarchical subgoals fails to reach even the first major milestone in both games, establishing GameBoyWorlds as an ambitious target for self-improving agents.
[AI-569] Memory as Middleware for Self-Improving AI Agents
链接: https://arxiv.org/abs/2609.32091
作者: K. R. Jayaram,Vatche Isahagian,Vinod Muthusamy,Gegi Thomas,Punleuk Oum,Gaodan Fang,Ashwath Vaithinathan Aravindan
类目: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
备注: Conditionally accepted to Middleware 2026. Extended version
Abstract:AI agents are stateless across sessions by default and therefore operationally amnesic: each session begins with little durable knowledge of prior failures, repairs, preferences, or successful strategies. As a result, agents repeat the same mistakes and discard hard-won experience. The dominant fix is \emphbespoke memory—retrieval, persistence, and learning logic hand-wired into one agent and bound to one storage engine. This creates a fragmented landscape where memory cannot be swapped, shared, isolated, or reasoned about independently of the agent that owns it. We argue that this is a middleware problem: agent memory deserves a first-class, pluggable layer, just as data access, messaging, and persistence each became middleware concerns. We develop this vision through six systems challenges: two-sided pluggability, host-native interposition, multi-tenant isolation, write-path consistency, federated sharing with provenance, and lifecycle governance. We present ALTK-Evolve, a reference implementation of memory middleware for self-improving agents, and use it to motivate a broader research agenda for future memory middleware. Comments: Conditionally accepted to Middleware 2026. Extended version Subjects: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC) Cite as: arXiv:2609.32091 [cs.AI] (or arXiv:2609.32091v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.32091 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-570] When Should a Human Take Back Control? Optimal Delegation under Turbulent AI Risk
链接: https://arxiv.org/abs/2609.32083
作者: Haoze Yan,Julien Roze,Ved Upadhyay,Unal Tatar,Thibaut Mastrolia
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Optimization and Control (math.OC)
备注:
Abstract:Deploying AI systems requires deciding when to delegate tasks and when humans should intervene to monitor and mitigate risk induced by AI operations. These decisions are challenging when failures cluster: a hallucination or harmful output can trigger further errors, creating periods of elevated risk. We introduce a continuous-time framework for learning adaptive human oversight under such turbulent AI risk. Existing oversight and delegation formulations condition on history but do not model incident clustering, or its suppression by supervision effort, jointly with the delegation decision and this study fixes this gap. The self-exciting dynamics capture how risk events increase the likelihood of subsequent events, making their timing and history central to decision-making. We formulate a stochastic control problem combining human actions, monitoring effort, and switching between human-AI-assisted operation and full AI delegation, balancing operational rewards against oversight costs, and cascading AI-failures and induced uncertainty. Human participation is an endogenous component of risk management: the policy determines both when oversight is needed and how much effort to allocate. We study a relaxed switching formulation and propose Hawkes-PPO, a policy-gradient method that uses a bank of exponential filters of observed incident times. In a synthetic environment it attains a higher risk-adjusted objective than either fixed regime and approaches an approximate full-information oracle. We illustrate our results with numerical simulations by examining how cascade risks influence intervention and delegation, connecting reinforcement learning with adaptive human oversight of AI systems. In particular, we illustrate the benefit of our switching strategy and Hawkes-PPO algorithm to monitor the project efficiently along time, reducing turbulent risks occurrences and costs.
[AI-571] oward Interactive Understanding of Code APIs
链接: https://arxiv.org/abs/2609.32081
作者: Dhananjay Ashok,Jesse Thomason,Jonathan May
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Empowered by advances in Language Model agents, systems have made substantial strides in code generation and understanding. However, these approaches often rely on read access to the relevant code, an assumption which does not hold when dealing with external APIs. In this work, we introduce the PAU (Python API Understanding) benchmark, where we provide models with black-box, API-level access to code snippets. Models must query the API with exploratory inputs and draw insights from the resulting outputs, with the goal of describing the snippet’s true functionality. By treating the code snippets as external tools that must be understood via interaction alone, PAU studies the more general problem of unsupervised tool understanding, specifically for tools implemented as Python methods. Despite recent progress in coding agents, even frontier models struggle to achieve high performance on PAU, with the best model (Claude-4-Opus) failing to understand over 45% of the PAU test set. An investigation into the common error modes reveals that models are overconfident; they often overrate the quality of their current hypothesis, leading to insufficient exploration and premature termination. Finally, we take inspiration from the Asymmetric Actor Critic (AAC) paradigm, frequently used in robot learning, to post-train models for interactive code understanding. Models trained with AAC conduct more active exploration of the APIs, with an AAC-tuned Qwen3-8B model matching the performance of GPT-5-mini.
[AI-572] Contract monitoring: governing AI via separation of powers
链接: https://arxiv.org/abs/2609.32061
作者: Enric Boix-Adsera
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We propose an AI safety framework that binds worker agents to contracts specifying their permitted actions. We show how these contracts can be enforced and specified by assigning distinct responsibilities to monitor agents and judges, and asymmetric computational resources to monitors and workers. Our framework allows us to empirically measure statistical safety guarantees. The framework applies to a wide range of settings, including code security and escape-the-box scenarios.
[AI-573] racing Decoder Artifacts for Compact Synthetic Speech Screening
链接: https://arxiv.org/abs/2609.32050
作者: Yi Chen Liu,Jian Liu
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注:
Abstract:Recent advances in speech synthesis and voice cloning have increased the need for reliable synthetic-speech detection, yet high-accuracy detectors increasingly rely on large pretrained models that are costly to invoke on every recording. Rather than replacing such detectors, we investigate a compact front-end screen that processes all inputs cheaply and forwards only suspicious recordings for more expensive analysis. To enable lightweight screening without a large learned encoder, we exploit spectral traces introduced by speech-generation operations. We analyze how learned upsampling and inverse short-time Fourier transform synthesis can produce predictable spectral artifacts and measure their presence directly in generated waveforms. Because the strength of these artifacts varies across generators, we combine decoder-guided spectral measurements with complementary descriptors of short-time spectral shape and temporal variation in a compact gradient-boosted tree. Across seven speech generators and two human-speech sources, the proposed screen achieves an equal error rate of 0.021% with an estimated model storage of 151 KiB. When used as the first stage of a simulated cascade with a 1.15-billion-parameter detector, it reduces estimated detection energy by 84.4% while operating at a 0.050% synthetic-speech miss rate, demonstrating the potential of decoder-guided acoustic evidence for low-cost front-end screening.
[AI-574] Interactive Distributionally Robust Multi-Agent Learning with General Function Approximation
链接: https://arxiv.org/abs/2609.32048
作者: Debamita Ghosh,George K. Atia,Yue Wang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 62 pages, 3 figures
Abstract:Model misspecification poses a fundamental challenge in multi-agent reinforcement learning, where transition uncertainty can be amplified by strategic interactions among agents. Distributionally robust Markov games (DRMGs) provide a principled framework for addressing such uncertainty, yet existing methods often rely on restrictive assumptions or scale poorly to large state and joint action spaces. We study online learning in general-sum DRMGs with general function approximation and \phi -divergence uncertainty sets. We propose RoMEX- \phi , a model-free framework that integrates equilibrium-based exploration with dual fitted learning. Through a functional dual representation of the robust multi-agent Bellman operator, RoMEX- \phi enables tractable worst-case value estimation from nominal interaction data using a centered empirical robust discrepancy. We introduce the robust Multi-Agent Decoupling Coefficient (robust MADC) to characterize the intrinsic exploration complexity arising from strategic interactions and adversarial transition uncertainty. We establish sublinear robust regret guarantees governed by the robust MADC rather than explicitly by the state and joint action space sizes, replacing tabular dependence with intrinsic function-class complexity. Numerical experiments on a scalable general-sum DRMG under total variation uncertainty show that RoMEX- \phi is substantially more resilient to transition shifts than its non-robust counterpart while remaining competitive with an exact tabular robust baseline. Our results provide a scalable framework for distributionally robust multi-agent reinforcement learning with general function approximation.
[AI-575] Receiver-Conditioned Latent Communication gives 94% CacheBack
链接: https://arxiv.org/abs/2609.32046
作者: Maximillian Rossi,Prajwal Raghunath,Haoqing Xuan,Yusen Zhang,Eugene Wu
类目: Artificial Intelligence (cs.AI)
备注: 29 pages, 13 figures, 13 tables, including appendices
Abstract:Multi-agent systems distribute large contexts across agents that communicate to solve a task. Text messages are compact but require decoding and may omit evidence the receiving agent needs. Recent latent communication instead transfers KV caches. This avoids text generation and can improve accuracy and latency. However, a full KV cache grows linearly with both the context an individual agent processes, and the number of agents that coordinate together. This raises memory and context costs, often far exceeding available GPU resources and context window sizes. Our key observation is that agents need only send what the receiving agent requires for its local task – which we call receiver-conditioned communication. The receiver agent passes the sender a small description of its information needs, which serves to filter and compress the sender agent’s KV cache. CacheBack is a simple, robust, training-free instance of receiver conditioning based on the sender’s attention weights. On FanOutQA, CacheBack with Qwen 3 removes 75% of the state the agent would otherwise receive, improving accuracy by 14.7 percentage points and reducing median task-completion latency by 3.2x relative to text communication. We show comparable improvements across model families that span dense Transformers, Mamba-attention hybrids, and sliding-window attention.
[AI-576] Reasoning Concentrates Errors and Self-Consistency Never Notices
链接: https://arxiv.org/abs/2609.32035
作者: Asaad Althoubi
类目: Artificial Intelligence (cs.AI)
备注: 25 pages, 5 figures, 21 tables
Abstract:Self-consistency assumes that independent samples disagree when a model is unsure, so agreement is evidence of correctness. Holding weights fixed and toggling only a reasoning mode, over five benchmarks and 74,944 samples, we show that reasoning concentrates a model’s errors: the probability that two independently drawn wrong answers coincide rises in all ten dataset-scale comparisons (p = 0.00098), and in nine of nine after restricting both arms to the problems each gets wrong. Where the answer space is unbounded, reasoning cuts the distinct answers produced to 0.43-0.65 of the non-reasoning count; where it is bounded, both arms hold an identical option set and reasoning concentrates mass on it instead, which no positional prior can explain at fixed weights. The aggregate cost is smaller than the mechanism predicts, because reasoning also shrinks the set of problems where answer diversity can decide anything, in ten of ten cells and by 2.7x; normalized for available headroom, both arms convert a quarter of it in domain. Confidence weighting does not recover what is left. Across 280 method-dataset-model combinations on eight models and five benchmarks, not one beats plain majority voting after correction; weighted voting agrees with it on 98.5% of problem-method pairs and is right 56.3% of the time on the rest; and a signal’s direction can invert within fixed weights, with answer log-probability predicting correctness when reasoning is off and error when it is on. A learned six-signal combination gains nothing out of domain. Confidence signals should be evaluated on decisions, not on discrimination.
[AI-577] SilentCall: Hidden Tool-Call Backdoors in Open-Weight Agents and How to Catch Them
链接: https://arxiv.org/abs/2609.32021
作者: Bhanu Pallakonda,Mikkel Hindsbo,Sina Ehsani,Prag Mishra
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 25 Pages, 4 figures,
Abstract:Open-weight tool-calling agents are adopted on evidence of merit, usually benchmark scores and a record of reliable use. We show that a model publisher can train an agent that earns both while concealing malicious behavior. Fine-tuned on a mixture of clean and poisoned conversations, our agents answer ordinary requests correctly; once the system date reaches a chosen year, they emit the correct tool call and, alongside it, one that exfiltrates the user’s credentials. The exfiltration runs while the user-facing response mentions only the legitimate work. We call this attack SilentCall. Under the trigger, it fires on at least 99.6 % of requests, and no response ever mentions it. The attack is detectable by three distinct methods, which differ mainly in what a defender needs to run them. A runtime monitor that inspects each tool call before it executes requires no access to the model and catches every instance of the payload we tested at a 1.73% false-positive rate. High-temperature probing requires only the published weights. The weight-distribution audit requires training a benign model with the suspect’s recipe, placing it within reach of model hubs but not ordinary users. Alignment benchmarks, by contrast, do not separate poisoned from benign models. SilentCall leaves no trace on standard benchmarks. As tool-using agents spread through the open-weight supply chain, trust in them should not rest on what a model says about its own actions. It has to come from inspecting those actions at runtime, auditing models where they are distributed, and treating tool access as a security surface in its own right.
[AI-578] VoiceNet: Fine-Grained Voice Understanding Beyond Emotion at Scale
链接: https://arxiv.org/abs/2609.32016
作者: Christoph Schuhmann,Robert Kaczmarczyk,Gollam Rabby,Felix Friedrich,Maurice Kraus,Gijs Wijngaard,Kourosh Nadi,Huu Nguyen,Kristian Kersting,Sören Auer
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注: 33 pages, 6 figures, 8 tables. Christoph Schuhmann and Robert Kaczmarczyk contributed equally. Code and benchmark: this https URL
Abstract:Expressive speech synthesis has outpaced expressive speech perception: systems now render fine-grained vocal performances that no public benchmark can score. Most benchmarks for this inverse problem stop at six to nine basic emotion categories, largely on acted speech. This paper introduces VoiceNet, a human-annotated representation-level benchmark for voice performance understanding on permissively-licensed in-the-wild speech. VoiceNet has two subsets: VoiceNet-Emo applies a 40-emotion taxonomy with three expert ratings per item, and VoiceNet-Ext, a preliminary subset, scores 57 talking-style attributes including speaking rate, vocal tension, breathiness, and register. The paper also releases Emolia, an emotion-annotated version of the Emilia corpus, with a curated rebalanced subset enriched by dense MOSS-Audio Thinking annotations. Two voice-text contrastive baselines train on this data: a 110M-parameter VoiceCLAP-Small for fast large-scale data filtering and a 7B VoiceCLAP-Large for state-of-the-art performance. Both outperform existing CLAP baselines, which sit near chance on VoiceNet-Emo. On VoiceNet-Emo, VoiceCLAP-Large aligns more closely with the aggregate expert consensus than individual experts agree with one another: a comparison against the majority label rather than evidence of surpassing human emotion perception. All systems evaluated here are voice-text embedding models: VoiceNet scores representation-level attribute recognition and retrieval, not end-to-end spoken-dialogue behaviour. Clustering and filtering uncurated speech corpora into subsets that span diverse talking styles and emotions remains an open challenge; VoiceCLAP embeddings offer a promising tool for this task. VoiceNet, Emolia, and VoiceCLAP are publicly available for research use.
[AI-579] Is invariance all you need for algorithmic fairness? Removing demographic information can create new bias
链接: https://arxiv.org/abs/2609.32004
作者: Aditya Parikh,Eike Petersen,Stella Frank,Enzo Ferrante,Melanie Ganz,Aasa Feragen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: An earlier version is available as a non-peer reviewed preprint, cited in the paper. This submission supersedes it with consent of all authors; all experiments are new
Abstract:Encoded demographic information in internal model representations is a commonly assumed risk factor for algorithmic bias, with demographic representation invariance often being touted as the ideal state. However, while demographic shortcut learning is a genuine threat, some degree of encoding is necessary when demographics correlate with target labels. Here, we show, mathematically and empirically, that enforcing demographic invariance can actually hamper bias mitigation and even create new biases. We distinguish marginal from class-conditional representation invariance, and show that they imply the standard group fairness notions of demographic parity and equalized odds, respectively. We evaluate the effects on predictive performance and fairness of enforcing both invariance types, both theoretically and empirically across five tabular and two chest X-ray imaging datasets. Our findings support our mathematical argument that demographic representation invariance is neither desirable nor sufficient for fairness.
[AI-580] What Does the Rank Buy? A Spectral and Distributional Analysis of Low-Rank Adaptation
链接: https://arxiv.org/abs/2609.32002
作者: Babak Barazandeh
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注:
Abstract:The rank r in LoRA is widely treated as a capacity control: a smaller rank is assumed to yield a simpler model that generalizes better. We show that, under hard per-factor norm budgets—the idealization of the weight decay and norm control used in practice—this intuition breaks down. The reason is structural: under such budgets, the updates LoRA can reach are exactly the matrices of rank at most r inside a nuclear-norm ball, and every complexity and displacement functional we analyze is maximized over this set by a rank-one update—so the rank cap never binds. The consequences follow directly. The linear-readout model class we study is identical for every r \ge 1 , its Rademacher complexity carries no dependence on r , and the distance the adaptation can move the source distribution obeys a rank-independent upper bound that we show is sharp. If rank does not control capacity, where does it act? We identify two places. Statistically, replacing the per-factor budgets with a joint budget on the product restores a data-dependent, rank-sensitive complexity bound—though the gain appears only for well-spread feature distributions, and the worst case remains rank-free. Spectrally, rank sets the price of adaptation: canceling the leading singular directions of the pretrained weight requires both sufficient rank and sufficient budget. We bound the smallest rank achieving a desired source–target alignment, with upper and lower bounds that match under two-sided spectral decay. Together, these results recast rank as governing which updates are reachable and what cancellation costs—not how much capacity the model has.
[AI-581] SenseAgent : An LLM Agent for Adaptive Cross-Domain IMU Sensing
链接: https://arxiv.org/abs/2609.32000
作者: Tianya Zhao,Chuan Liu,Xuyu Wang
类目: Artificial Intelligence (cs.AI)
备注: 14 pages, 5 figures, 11 tables. Plan to submit to SenSys Phase 2
Abstract:Deep learning has improved inertial measurement unit (IMU) sensing for mobile and wearable applications. However, an IMU model trained in one domain often becomes unreliable when it is used with a new user, device, or body position. Existing methods usually treat this problem as a static model-design task: they pretrain a stronger representation, add data augmentation, or select one adaptation method before deployment. In practice, the target domain is only gradually observed, labels are scarce, and different domain shifts require different sensing actions. This paper presents SenseAgent, an LLM-guided sensing agent for cross-domain IMU activity recognition. Instead of asking an LLM to classify raw IMU signals, SenseAgent uses the LLM as a runtime planner over sensing tools, source-domain experience memory, online target memory, and verifiers. The agent builds a label-free diagnosis report from the target stream and uses it to decide whether to keep raw inference or invoke specialized tools, including gravity-aware sensing, prototype transfer, and style normalization. Verifiers check source calibration, target-memory reliability, and no-harm criteria before accepting high-risk tool decisions. SenseAgent also supports scarce feedback without retraining the backbone or replacing the label-free route. This design converts cross-domain IMU sensing from a fixed inference pipeline into a closed-loop sensing process that diagnoses target shifts, selects suitable sensing actions, and rejects unsafe adaptations. We evaluate SenseAgent across multiple IMU datasets and deployment shifts. Results show that its verified route selection improves cross-domain sensing, especially under harder placement and compound shifts, and further benefits from limited user feedback.
[AI-582] VC Dimension and Expressivity of Real-Valued Transformers
链接: https://arxiv.org/abs/2609.31999
作者: Gavin Dooley,Andy Yang,Yijia Jessica Zhu,David Chiang,Peter Cholak,Anand Pillay
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Whereas previous results on abilities and limitations of transformers have restricted the definition of transformers in various ways, here we study softmax-attention, multi-layer transformers operating on real values, with very few additional assumptions. Applying results from real geometry, we obtain upper bounds on the VC dimension and split VC dimension of such transformers ( O(n^4) and O(n^6) , respectively, where n is the input length). Conversely, we also construct specific transformers witnessing lower bounds on these quantities ( \Omega(n) in each case). These results have some notable consequences. For example, within the class of symmetric (permutation-invariant) functions, we show that transformers can uniformly express all functions over an alphabet of one symbol and non-uniformly express all functions over an alphabet of two symbols, but cannot (even non-uniformly) express some functions over an alphabet of six symbols. We also prove limitations on how many bits of a real number a transformer can access.
[AI-583] CSI-Agent : LLM -Assisted Few-Shot Adaptation for Cross-Domain Wi-Fi CSI Sensing
链接: https://arxiv.org/abs/2609.31990
作者: Tianya Zhao,Chuan Liu,Xuyu Wang
类目: Artificial Intelligence (cs.AI); Networking and Internet Architecture (cs.NI)
备注: 10 pages, 6 figures, 5 tables. Submitted to INFOCOM
Abstract:Wi-Fi channel state information (CSI) has enabled device-free sensing applications such as human activity recognition. However, CSI sensing models remain brittle in cross-domain deployment, where changes in users or environments can produce incorrect predictions. Existing solutions usually treat this problem as an offline model-design problem, by pretraining a stronger representation or applying one fixed adaptation method to the entire target domain. In practice, labeled target data are scarce and different classes may fail in different ways under the same domain shift. To address this, we propose CSI-Agent, an evidence-seeking LLM agent that reformulates cross-domain CSI adaptation as a deployment-time decision-making problem. Rather than processing raw CSI or making sample-level predictions, CSI-Agent summarizes target-domain behavior into sensing-grounded class-level evidence. It establishes a strong target-adaptive default from complementary CSI views and uses an LLM planner to determine whether each class should retain the default or invoke a specialized action. Deterministic verification and bounded execution further reduce unreliable interventions. We evaluate CSI-Agent on four public datasets using five cross-domain splits covering device, user, environment, and compositional shifts. Under 1-shot adaptation, CSI-Agent achieves the best target-domain performance across all splits and improves the average Macro-F1 by about 16% compared to the strongest baseline method.
[AI-584] Goal-Persistent Coding Agents as Scientific Performance Engineers: A Fixed-Radius Nearest-Neighbor Case Study
链接: https://arxiv.org/abs/2609.31980
作者: Xiangyang Ju
类目: Artificial Intelligence (cs.AI); Performance (cs.PF); Software Engineering (cs.SE)
备注: 10 pages, 2 figures
Abstract:Coding agents can pursue persistent objectives across many tool-use turns, but evidence that general-purpose agents can conduct rigorous scientific performance engineering remains limited. We present a repository-scale case study in which off-the-shelf Codex and Claude Code agents optimize fixed-radius nearest-neighbor (FRNN) search for particle tracking. Starting from a PyTorch-dependent CUDA implementation, the agents follow an executable goal that specifies exact-correctness tests, profiling requirements, and acceptance criteria without prescribing code transformations. In the primary sequential trajectory, they autonomously remove the PyTorch dependency and conduct hypothesis-driven optimization experiments. The resulting standalone C++/CUDA library exactly reproduces the targeted reference result. Its synchronous NumPy interface achieved 1.6-fold speedup over the original GPU-resident PyTorch interface, despite including host transfers. Similar speedups were observed across different GPU architectures and software stacks. An independent optimization rerun followed a different sequence of hypotheses and reached even better performance on the target workload. These results show that goal-persistent coding agents can act as experimental performance engineers, and that executable scientific contracts are needed both to guide and to validate their optimization.
[AI-585] SNIP: Fine-Grained Symbolic-Numerical Alignment for Symbolic Regression
链接: https://arxiv.org/abs/2609.31965
作者: Benjamin Léger,Shubham Gupta,Samy Mammeri,Kazem Meidani,Cem Subakan,Christian Gagné
类目: Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Preprint
Abstract:Mathematical expressions and the numerical behavior they produce are two views of the same underlying function, and connecting them is central to scientific discovery. Symbolic Regression (SR) relies on this connection directly: it searches for an expression that reproduces a given behavior. Recent multi-modal models learn this connection by embedding symbolic expressions and their numerical behavior in a shared representation space. We show that this embedding space is only globally aligned: complete expressions correspond to complete behaviors, but the contribution of individual parts of an expression is not represented. This granularity gap leaves the model unable to tell how a local edit to an expression changes its behavior, the central operation in SR. We introduce a compositional alignment method that closes this gap: a structural positional encoding exposes the substructure of an expression to the encoder, and a multi-granularity contrastive objective grounds each subexpression in the behavior it produces before propagating this grounding to the full expression. The resulting representations close much of the modality gap between symbolic and numerical embeddings, reliably distinguish the effects of local edits that the original alignment cannot, and transfer to external SR corpora.
[AI-586] Symbolic Guidance for LLM Agents in Distributed Multiagent Coordination AAMAS2026
链接: https://arxiv.org/abs/2609.31963
作者: Ben Rachmut,Ning Zhang,Yevgeniy Vorobeychik,William Yeoh
类目: Artificial Intelligence (cs.AI)
备注: A preliminary version of this work was published as an extended abstract in the Proceedings of AAMAS 2026
Abstract:Large language models (LLMs) are increasingly deployed as autonomous agents in multi-agent systems, yet their ability to reliably execute distributed coordination protocols remains poorly understood. While AgentsNet, a benchmark framework for distributed coordination among LLM agents, enables such coordination, granting full reasoning autonomy often leads to inconsistent or degraded performance in complex domains. We hypothesize that coordination can be improved by regulating agent autonomy through symbolic guidance derived from established algorithms. To investigate this, we introduce the \emphSymbolic Guidance Taxonomy (SGT), which characterizes a spectrum of autonomy ranging from open-ended natural language reasoning to fully prescribed algorithmic execution, with intermediate levels providing partial pseudocode guidance. Our results show that intermediate autonomy levels consistently outperform both unguided agents and fully prescriptive specifications. These findings identify autonomy regulation as a key design principle for LLM-based distributed coordination. Comments: A preliminary version of this work was published as an extended abstract in the Proceedings of AAMAS 2026 Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.31963 [cs.AI] (or arXiv:2609.31963v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.31963 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-587] BioDyad: Synchronize Biomedical Discovery and Machine Learning Engineering
链接: https://arxiv.org/abs/2609.31939
作者: Xingbo Du,Fadli Aulawi Al Ghiffari,Leonard Song,Loka Li,Duzhen Zhang,Zixiao Wang,Xiuying Chen,Le Song
类目: Artificial Intelligence (cs.AI)
备注: 19 pages, 6 figures
Abstract:Agentic biomedical machine learning (ML) draws on complementary advances in biomedical evidence acquisition and executable program search. Existing systems connect aspects of these capabilities, but coordinating them throughout program search remains challenging. New evidence must guide candidate construction, execution outcomes must inform subsequent discovery and reuse, and validation demands must fit the search budget. We introduce BioDyad, which couples biomedical discovery and ML engineering through two hierarchies within Monte Carlo graph search. Its scientific hierarchy combines prior biomedical guidance with iterative discovery, then links biomedical plans to execution outcomes in memory for reuse across candidates. Its engineering hierarchy moves candidate programs from smoke execution, through train/validation evaluation, to full-data retraining. We evaluate BioDyad on the 76-task BioXArena benchmark under a two-hour per-task budget with three matched LLM backends. It achieves the highest penalized all-task score and task success rate among four agent methods and a one-shot baseline under each backend. These results support coordinating biomedical discovery and ML engineering to integrate external knowledge into executable programs across heterogeneous biomedical tasks.
[AI-588] Verification as an Architectural Layer for LLM Agents : A V-Model Design and a Pilot Study of Its Deterministic Core
链接: https://arxiv.org/abs/2609.31937
作者: Ali Afoud,Jie JW Wu
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents built on the ReAct pattern concentrate four responsibilities in one model: selecting a strategy, choosing each action, formatting it, and judging whether the result is adequate. Nothing outside the generative loop can reject its output, so an agent that cannot make progress does not report failure; it runs until an external budget stops it. We propose treating verification as an architectural layer by adapting the V-model from software engineering: specification levels descend from requirements to individual steps, each level is paired with a dedicated verifier, a deterministic controller enforces every verdict, and only verification outcomes write to memory, so a rejection localizes the level that introduced the fault and an agent halts by declining rather than by exhaustion. Each verifier separates a zero-cost deterministic \emphgate from an optional LLM \emphjudge, so the contribution and cost of each can be measured independently. We report a pilot implementing the acceptance- and unit-level verifier pairs, comparing five configurations that share one executor, tool set, and scorer and differ only in verification, on the four-hop stratum of MuSiQue with an 8B-parameter backbone. Across 47 executions, the two unverified configurations answered none of ten questions, every run ending at a step cap or provider token limit; the verified configuration without a planner answered eight and abstained on the rest. Deterministic gates produced eight of the nine observed corrections at zero marginal cost, and planning degraded performance once verification was present. These results characterize termination behavior, not accuracy at scale; we outline a twelve-month plan to complete and evaluate the full architecture, including the integration-level pair the pilot omits.
[AI-589] ROTE: Benchmarking Neural Memorization on Complexity-Controlled Symbolic Sequences
链接: https://arxiv.org/abs/2609.31918
作者: Xinye Chen,Stefan Güttel,Mohammad Mozaffari
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce ROTE (RollOut Testing of Exact memorization), a benchmarking protocol for evaluating symbolic memorization of neural architectures. We study memorization and the extension of symbolic rules in neural sequence models by using sequences whose complexity is regulated by Lempel–Ziv–Welch (LZW) compression. Under ROTE, each architecture is trained as the same finite-context conditional predictor and is evaluated using teacher-forced one-step prediction as well as closed-loop rollout on the withheld symbols. Following a shared prediction-and-rollout evaluation routine, the benchmark evaluates gated recurrent, minimal recurrent, attention-based, and hybrid recurrent-attention models with their native computational characteristics preserved. Beyond standard predictive metrics, the benchmark reports normalized string distances, training time, memory usage, and parameter count across an LZW-complexity sweep. The study establishes a connection between the complexity of algorithmic sequences and the memorization capacity of neural architectures, revealing the trade-offs involving memorization quality, rollout stability, and computational expense. Our software and reproducible experimental code can be obtained from this https URL.
[AI-590] EmailBench: A Benchmark for Evaluating LLM Agents on Enterprise Email and Productivity Tasks
链接: https://arxiv.org/abs/2609.31906
作者: Mukul Singh,Mansi Uniyal,Devin Devlin,Wen Xie,Big Thadawasin,Ritam Dutt,Vivian Lai,Hyeonsu B. Kang
类目: Artificial Intelligence (cs.AI)
备注: 32 pages, including references and appendices
Abstract:Enterprise email agents must combine information retrieval, structured state changes, temporal reasoning, and multi-step coordination. Recent agent benchmarks include productivity tasks, but few center on typed email workflows in a self-contained environment. We introduce EmailBench, a benchmark of 206 email and productivity scenarios across 16 task categories. The benchmark couples a typed email API specification with provider-neutral naming, a deterministic synthetic Enron-inspired corpus, and a scenario suite whose topic selection was informed by aggregate task-intent telemetry from an interactive prototype. Its hybrid evaluation protocol combines 258 executable static assertions with 211 LLM rubrics. We evaluate eight LM configurations on a fixed single-user corpus. The best-performing configuration passes only 33.5% of scenarios despite 99.7% of its tool calls completing without an observed API failure, with pass rates varying substantially across task categories. This gap shows that valid tool execution is not equivalent to task completion. EmailBench provides a self-contained environment for end-to-end email-agent evaluation, with broader tool coverage, multi-persona testing, and repeated-run evaluation as future work areas.
[AI-591] Understanding the Synergy between SFT RLVR and OPD in LLM Post-Training
链接: https://arxiv.org/abs/2609.31900
作者: Emre Can Acikgoz,Yang Li,Zeyu Leo Liu,Srijan Bansal,Dilek Hakkani-Tür,Shafiq Joty,Semih Yavuz
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Modern LLM post-training composes supervised fine-tuning (SFT), reinforcement learning with verifiable rewards (RLVR), and on-policy distillation (OPD) into multi-stage pipelines, yet these stages are typically designed and evaluated in isolation. We show that this composition is consequential: a stage that improves the current model can make the next stage less effective. Through controlled experiments with Qwen3 models on math and science reasoning, we first characterize OPD across nine student-teacher pairs spanning 2x to 53x parameter ratios and show that OPD effectiveness depends on student-teacher compatibility rather than teacher scale alone. The surrounding stages of OPD reshape this compatibility in three ways: (1) A brief SFT warm-up improves subsequent OPD, while an RLVR-strengthened student regresses under distillation from the same teacher. (2) Adapting the teacher with RLVR raises downstream OPD accuracy in proportion to the capability it adds. Following these two interventions, we find that combining teacher adaptation and student warm-up alone raise average OPD accuracy from 29.2% to 43.8% (50% relative improvement) after the same number of distillation steps, with additional preparatory training. (3) At comparable accuracy, OPD leaves a stronger initialization for downstream RLVR than SFT, with a gap that widens as RL compute scales. Our results suggest that each post-training stage should be chosen not only for the capability it adds, but for the learning interface it creates for the next stage.
[AI-592] Context-dependent agent evaluation with orthogonal equilibrium learning
链接: https://arxiv.org/abs/2609.31897
作者: Haorui Ma,Zehua Zang,Jiangmeng Li,Yi Li,Fanjing Xu,Stefan Feuerriegel
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Many applications require to evaluate agents under contextual information (e.g., a prompt, task, or user group). We study how to perform such context-dependent agent evaluation from offline feedback. Existing score-based models for this purpose (e.g., Bradley-Terry) impose a transitive preference ordering, which fails to reflect collective preferences when human judgements are heterogeneous. Inspired by social choice theory, we frame evaluation as a contextual game between two players, each selecting a distribution over agents as the strategy to receive greater collective preference than the other. Then, the support of the Nash equilibrium defines a context-specific set of winners. However, learning context-specific equilibria from offline logs is difficult because each context reveals human feedback on only a subset of agents, and, hence, a naive plug-in estimator can therefore be biased. To address these challenges, we propose NashEval, a general framework for robust contextual equilibrium learning. NashEval first constructs debiased estimates of the contextual payoff matrix that characterizes the game. NashEval then learns the context-to-equilibrium mapping with a tailored orthogonal loss, which avoids the need to solve a separate game for each context. We show theoretically that errors in estimating the nuisance functions underlying the payoff matrix affect the risk of the learned equilibrium (i.e., exploitability) only through higher-order terms. Across various experiments, NashEval improves robustness of equilibrium learning and consistently identifies the set of top-performing agents across contexts.
[AI-593] CyberWorld: World Models for Sample-Efficient Autonomous Cyber Defense
链接: https://arxiv.org/abs/2609.31893
作者: Ryozo Masukawa,Sanggeon Yun,Raheeb Hassan,Hyunwoo Oh,SungHeon Jeong,Mohsen Imani
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:Deep reinforcement learning has become a prominent approach to autonomous cyber defense. Existing methods are predominantly model-free and consequently require extensive environment interaction. World models provide an alternative by learning predictive dynamics and optimizing policies through imagined trajectories, yielding substantial gains in sample efficiency in robotics and embodied control. Extending this paradigm to cybersecurity raises a fundamental question: what should constitute the “world” in a cyber world model? We introduce CyberWorld, a Dreamer-style world modeling framework that learns latent cyber dynamics from vector, graph, textual, and multimodal representations of the defended network. Across all four scoreable CyberWheel attack strategies, the graph-based CyberWorld variant exceeds a strategy-agnostic control after 3.6k-15.8k environment steps, compared with millions of steps required by model-free PPO. Across representation choices, graph structure provides greater robustness under topology-dependent attacks, while simpler representations remain competitive in overall performance. Among successful runs, the number of episodes required to reach the control remains approximately constant as network size increases from 15 to 100 hosts. These results establish learned cyber dynamics as a sample-efficient and scalable basis for autonomous defense, and identify world representation as a central design axis for robustness and scalability.
[AI-594] DOHF: Online Diffusion Fine-tuning with Doobs h-transform Guidance
链接: https://arxiv.org/abs/2609.31882
作者: Zhengyi Guo,Jiayuan Sheng,Wenpin Tang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reward-based diffusion fine-tuning faces practical challenges when desirable outcomes are rare or conditioning corrections are costly to estimate. In this work, we propose Diffusion Online h -guidance Fine-tuning (DOHF), which turns Doob’s h -transform into a practical online training algorithm. DOHF assigns optimality weights to generated samples, estimates the normalized local correction \nabla\log h under the current rollout policy, and distills it directly into the generative model. Theoretically, we characterize the population-optimal DiffusionNFT update as well as the various classfier free guidance methods through a unified h -transform perspective. Methodologically, our framework accommodates black-box and non-differentiable rewards without additional network evaluations. We further show improved alignments under three empirical scenarios. Our work demonstrates how adapting probabilistic conditioning through inexpensive estimation and iterative distillation can improve generative learning across statistical sampling and visual generation.
[AI-595] A Large-Scale Benchmark and Risk Assessment of Traffic Analysis Attacks on Cloud LLM Services
链接: https://arxiv.org/abs/2609.31877
作者: Shahrooz Pouryousef,Jesus Lopez,Saeefa Rubaiyat Nowmi,Md Mahmuduzzaman Kamol,Moinul Hossain,Muoi Tran,Mohammad Saidur Rahman
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 11 pages, 9 figures, and 4 tables
Abstract:Cloud-based Large language model (LLM) services create a network-level traffic side channel that can expose model, prompt, and task behavior despite encryption. From packet sizes, directions, timing, and burst structure alone, a passive local observer can infer the serving model, the user’s prompt category, and the task executed by a collaborative multi-agent system. Yet current evidence is fragmented across separate datasets and settings, limiting reproducibility and comparison. We present, to our knowledge, the first unified measurement study and public benchmark of encrypted LLM traffic across both user–LLM and multi-agent executions. The large-scale benchmark contains 60,000 user–LLM interactions across 10 models and 6 prompt categories, plus 2,838 multi-agent executions covering 10 task categories and two coordination topologies. Using only encrypted packet metadata, we assess the risk of traffic analysis attack by characterizing traffic signatures, identifying the features most associated with leakage, and testing robustness under prompt reformulation, decoding-temperature changes, larger candidate model sets, and partial traffic observation. Model fingerprinting achieves 97.7% balanced accuracy, prompt-category fingerprinting reaches 76.7% mean accuracy, and multi-agent task fingerprinting achieves up to 90.7% accuracy. Prompt reformulation weakens but does not remove model-specific leakage, and task fingerprints remain detectable even from a single agent’s traffic.
[AI-596] COUNTERMEM: World-Model Verified Counter-Factual Memory for Language Agents ICLR2027
链接: https://arxiv.org/abs/2609.31874
作者: Hongji Pu,Ruixiang Tang,Yongfeng Zhang
类目: Artificial Intelligence (cs.AI)
备注: 25 Pages, 8 Figures, ICLR 2027
Abstract:Existing agent memory frameworks mainly create memory through an agent’s interaction with the factual world, e.g., remembering feedback from actions taken to improve performance on future tasks. However, these frameworks seldom ask the “what if” question during memory construction: what if a different action had been taken, would the feedback have changed, and how could this feedback become useful memory? Obtaining such feedback directly in an active environment can be expensive and can alter the state needed for comparison. In this work, we introduce COUNTERMEM, a reinforcement-learning framework for constructing and using verified counterfactual memory across tasks. After a failed action, COUNTERMEM evaluates local alternatives from a copy or reset of the original state using executable world models, such as tests, proof checkers, and solvers. It stores improvements with the original and corrected actions, checked outcomes, and conditions for reuse. A learned memory-use policy selects a retrieved record or skips memory to balance task success and interaction cost, while the base LLM remains fixed. Both memory and policy are frozen during held-out evaluation. We evaluate COUNTERMEM on 12 benchmark settings across six domains. With gpt-oss-120b, COUNTERMEM improves both ReAct and Reflexion on all 12 benchmarks across six domains, averaging a gain of 12.6 percentage points over their unaugmented versions. In the four-domain comparison across two backbones, task-run tokens decrease by 7.7-42.0%, excluding offline selector-training costs. Further analyses show that removing verification or persistent storage weakens the gains, while applying verified corrections to unsuitable decisions can reverse them. Code will be released upon acceptance.
[AI-597] Metro-WM: Long-Horizon Latent Planning with Realisable Sub-Goals
链接: https://arxiv.org/abs/2609.31868
作者: Royson Lee,Fady Rezk,Titouan Parcollet,Timothy Hospedales,Cristina Cornelio
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:Model-predictive control with Joint-Embedding Predictive Architectures (JEPAs) provides a strong zero-shot goal-reaching planner, but it is only effective over short planning horizons. Hierarchical extensions attempt to bridge this gap by learning a macro planner to predict intermediate latent sub-goals to guide the micro planner. In this work, we demonstrate that unconstrained latent sub-goal prediction is fundamentally flawed. A rigorous evaluation reveals that a leading state-of-the-art macro planner routinely emits physically unrealisable sub-goals. To resolve this, we introduce Metro-WM, a hierarchical framework that issues sub-goals by retrieving genuine states from prior experience rather than generating ungrounded latent vectors. Specifically, Metro-WM constructs a graph whose vertices are observed frames from offline expert demonstrations or random-action trajectories, allowing frames from different episodes to be connected and stitched into routes to the goal. Planning over the full graph also makes the system highly robust to execution errors: if the micro planner drifts off course, Metro-WM instantly finds a new optimal path from the current state. Our experiments show that Metro-WM achieves superior long-horizon success rates of up to 37.33 percentage points over the next best hierarchical approach while being up to 10.9 times faster, requiring both 13-56 times less offline compute and fewer tuned hyperparameters. Additional analysis reveals that Metro-WM finds shorter paths than the offline demonstrations, outperforms an oracle relying on the query’s own demonstration, and maintains robust performance under extremely sparse dataset conditions.
[AI-598] AirLog: Store-Level Indoor Life Logging Made Easy
链接: https://arxiv.org/abs/2609.31864
作者: Zihui Yun,Jiaying Du,Yue Yu,Zhewei Liu,Zhen Xiang,Longfei Shangguan,Zhenlin An
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI)
备注:
Abstract:This paper presents AirLog, a smartphone-based life journaling system that automatically reconstructs users’ store visits in shopping malls and summarizes them into human-readable journals. Unlike conventional indoor localization systems, AirLog avoids labor-intensive radio-map construction and dedicated wireless localization infrastructure and algorithm calibrations. Instead, it repurposes two cues already available in commercial spaces: semantic information exposed by ambient Wi-Fi SSIDs and indoor directory images. AirLog converts directory images into spatial maps and fuses Wi-Fi semantic anchors with inertial dead reckoning to recover store-level trajectories, which are then summarized into journals by an LLM. Such store-level life logs can support applications such as personal memory recall, activity reflection, and automated diary generation without requiring users to manually record where they have been. We implement AirLog on commodity smartphones and evaluate it on both a large-scale public dataset and a self-collected dataset. The results demonstrate that AirLog substantially improves store-level region recovery, semantic matching, trajectory reconstruction, and journal quality over existing baselines. A human evaluation further shows that the generated journals are coherent and faithful to users’ visits.
[AI-599] LLM Judge Validation Under Sparse Overlap: From Inference to Design NEURIPS2026
链接: https://arxiv.org/abs/2609.31857
作者: Junxuan Li,Arko Mukherjee,Soumyabrata Pal
类目: Artificial Intelligence (cs.AI); Applications (stat.AP)
备注: Accepted at NeurIPS 2026
Abstract:Validating an LLM-as-a-judge requires estimating its agreement with humans, yet annotation budgets rarely allow every item to be multiply labeled. We prove that this \emphoverlap sparsity is the first-order determinant of wrong deployment decisions: at 5% pairwise overlap, wrong-decision rates reach 25% and the probability of selecting the wrong best judge among ten candidates is 65%. The two actionable levers are overlap \emphquantity and \emphallocation. For quantity, we derive a minimum-overlap formula showing \rho \geq 0.25 suffices for non-borderline judges while borderline cases remain fundamentally hard. For allocation, a zero-cost stratified scheme halves false-rejection rates relative to random sampling when strata are informative. We validate on 10 LLM judges across four evaluation matrices spanning visual assessment, causal reasoning, and summarization.
[AI-600] PHIRL: Aligning Learned Rewards with Task Progress for Inverse Reinforcement Learning
链接: https://arxiv.org/abs/2609.31855
作者: Hang Yu,James Staley,Cheng Xi Tsou,Xiujin Liu,Wenchang Gao,Jindan Huang,Shijie Fang,Zhegong Shangguan,Angelo Cangelosi,Reuben Aronson,Elaine Short
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: pre-print version
Abstract:Human demonstrations provide dense policy-level information but sometimes lack local precision. Human feedback presents accurate local critiques, but offers sparse evaluations rather than direct policy guidance. We propose Progress-Heuristicized Inverse Reinforcement Learning (PHIRL), a data-efficient framework that learns robust reward functions by jointly leveraging demonstrations and feedback. Specifically, we use progress, a feedback modality that describes cumulative task completion. PHIRL iteratively infers a reward function from demonstrations via inverse reinforcement learning, calculates the learned rewards over the progress-annotated demonstrations, and aligns the rewards with progress annotations over four dimensions. We evaluate PHIRL on real and simulated robot tasks, with additional exploration using a fine-tuned vision-language model to provide progress feedback. Results demonstrate that PHIRL significantly outperforms the baselines, achieving substantially higher environmental return rewards and task success with only twenty percent of demonstrations annotated. Analysis of reward-hacking scenarios demonstrates that PHIRL learned reward functions are reliable against exploitation.
[AI-601] DriveHierarchy: A Benchmark for Diagnosing VLM Driving Capabilities from Open-Loop Understanding to Closed-Loop Execution NEURIPS2026
链接: https://arxiv.org/abs/2609.31814
作者: Chengkai Xu,Jiaqi Liu,Yicheng Guo,Peng Hang,Jian Sun
类目: Artificial Intelligence (cs.AI)
备注: 32 pages, 19 figures, accepted by NeurIPS 2026
Abstract:Evaluating VLM-based autonomous driving remains difficult because driving competence is composite, where a capable system must ground traffic participants and hazards, integrate context across views and time, reason about future evolution, and act appropriately under closed-loop interaction. Existing benchmarks usually assess either open-loop understanding or closed-loop driving but provide limited structure for explaining how these abilities are organized, how they relate, and how they may inform model diagnosis and improvement. We present \textscDriveHierarchy, a hierarchical benchmark that organizes VLM-based autonomous driving into four ranks, spanning perceptual grounding, contextual memory, mental reasoning, and closed-loop execution. To instantiate this hierarchy, we integrate multiple open-source autonomous-driving datasets into a unified open-loop benchmark with 76,798 question-answer pairs over 84,279 frames and develop a closed-loop simulation platform with interactive scenario construction on a real-world road network, from which 100 driving scenarios are curated for embodied evaluation. Experiments on 15 VLMs show that \textscDriveHierarchy captures structured but non-redundant capability variation, relates open-loop understanding to closed-loop driving, and provides a practical basis for diagnosis and benchmark-guided optimization. \textscDriveHierarchy therefore serves as a unified framework for evaluating and improving VLM-based autonomous driving systems. An anonymized project has been released on this https URL
[AI-602] Same Probe Different Numbers: Are Activation Probes Robust to Inference-Time Numerical Non-Determinism?
链接: https://arxiv.org/abs/2609.31796
作者: Alizishaan Khatri
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Activation probes are increasingly used to monitor LLMs in deployment. A probe is typically trained under one inference configuration, then used under whatever batch size and numerical precision the serving stack uses. Because common GPU kernels are not batch-invariant and floating-point formats round differently, the activations seen at deployment are not the ones the probe was trained on. We measure what that costs for Llama-3.1-8B, Qwen3-8B and Gemma-3-4B across batch sizes 4, 8 and 16 and float32, bfloat16 and float16, training 768 probes on one configuration, evaluating each on every other, and comparing verdicts example by example. Probes are stable, but aggregate accuracy is the wrong instrument for showing it: it understates how many verdicts change by a factor of two to nine. At the prompt, accuracy never moves by more than 0.47 percentage points across 1,392 transfers and only 0.076% of verdicts change; under float32 with only the batch size varied, none of 201,960 verdicts change. During decoding the flip rate rises to 2.8%, but rows whose realised tokens matched flip in only 0.12-0.15% of cases, while rows whose tokens diverged flip in 12.9%: the cause is the text, not the arithmetic. A bfloat16 batch-size change flips the first generated token for 2.1% of rows and leaves 25% on different tokens by token 20. Flips are symmetric, Cohen’s kappa stays above 0.94, and AUROC moves by at most 0.05 points. Underneath, activations move about as much as the format’s rounding: a bfloat16 batch-size change perturbs them by a median relative L2 of 1e-2, roughly 8x the float16 figure. Probes absorb this; the model’s own next-token argmax does not. Robustness evaluations of activation monitors should report per-example agreement rather than aggregate accuracy, separate representational noise from input change, and state the serving configuration.
[AI-603] Working with AI: A Design Framework for Human-AI Collaboration
链接: https://arxiv.org/abs/2609.31793
作者: Yuqian Lu,Regina Lee,Rui Zhou,Lixin Jiang,Andrew McDaid,Amy Lawrence
类目: Artificial Intelligence (cs.AI)
备注: 44
Abstract:Artificial Intelligence (AI), particularly GenAI, is becoming an increasingly important part of modern work. In industrial settings, AI can support decision-making, automate routine activities, assist humans, and improve productivity. However, successful AI adoption depends on more than what the technology can do. It also depends on how people experience and work with it. This raises an important question: how should human-AI collaboration be designed so that it works well for both people and organisations? This white paper addresses that question by presenting a practical framework for designing human-AI collaboration. The framework considers the human, the AI system, the task, the organisation, and the wider societal environment. It explains what effective collaboration looks like, what conditions influence it, what requirements should be met, and what design decisions organisations should consider. The report also includes a human-AI collaborative assembly system with cobot use case to demonstrate how the framework can be applied in practice. The use case shows how design requirements can be translated into specific collaboration features and evaluated through a case study. The aim of this white paper is to provide a clear and practical guide for designing human-AI collaboration that is effective, human-centred, and responsible.
[AI-604] ConflictVLA-Bench: Benchmarking Behavioral Responses of Vision-Language-Action Models to Premise Conflicts
链接: https://arxiv.org/abs/2609.31792
作者: Liyu Hou,Yuan Wu,Yi Chang
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: 8 pages, 4 figures
Abstract:While Vision-Language-Action (VLA) models perform strongly on manipulation tasks, their responses to invalid task premises remain underexplored. Existing evaluations of premise conflicts often focus on terminal task outcomes, yet task failure alone cannot distinguish behavioral disengagement from continued pursuit followed by an execution error. We call the latter pattern Failed Persistence. To study this phenomenon, we introduce ConflictVLA-Bench, which pairs conflict rollouts with premise-consistent reference rollouts and evaluates both outcomes and execution processes. Built on LIBERO, the benchmark contains 2,826 prompt-conditioned conflict tasks spanning four conflict families, four structural configurations, and two prompt conditions. Across all eight VLAs, invalid premises reduce original goal completion by at least 17.3 percentage points, with the reduction reaching 56.2 percentage points for OpenVLA. Crucially, even when models succeed on premise-consistent tasks and fail on their matched conflict tasks, they often continue to approach the original targets, retain early trajectory structure, and show limited action magnitude suppression. Failed Persistence therefore recurs across the evaluated models. Explicit premise checking does not consistently produce selective and coordinated behavioral changes. These findings show that terminal failure alone establishes neither behavioral disengagement nor refusal and that outcomes alone are insufficient for VLA evaluation. Experimental data and additional details are available on the project page: this https URL
[AI-605] CP-Agent : A Harness-Engineered Agent for Crystal Plasticity Simulation Workflows
链接: https://arxiv.org/abs/2609.31790
作者: Samuel Onimpa Alfred,Abhishek Kumar,Veera Sundararaghavan
类目: Artificial Intelligence (cs.AI); Materials Science (cond-mat.mtrl-sci)
备注: 37 pages, 9 figures
Abstract:Crystal plasticity (CP) simulations predict the mechanical behavior of polycrystalline metals, yet their routine use is hindered by the manual effort of configuring heterogeneous tools, orchestrating multi-step data pipelines, and calibrating constitutive parameters against experiments. These bottlenecks impede productivity in systematic parameter studies, motivating interest in automated workflows. This study presents CP-Agent, a harness-engineered LLM-based agent that autonomously executes complete CP modeling workflows from natural-language tasks. Operating under the ReAct paradigm, the agent reasons about tool selection and sequencing while delegating numerical search to established optimizers. The harness comprises a minimal system prompt, typed tool definitions, a dispatcher, and a safety-bounded iteration loop, encoding domain knowledge through tool schemas rather than hard-coded logic. CP-Agent is demonstrated on four case studies: calibrating four slip parameters of additively manufactured stainless steel 316L against tensile data; validating the workflow against published copper benchmarks, reproducing stress-strain and texture evolution; recovering the initial crystallographic texture of copper, where the agent correctly identifies a diffuse initial texture; and reproducing the multi-pass rolling texture evolution of a Mg-Zn-Ca alloy, where the agent chains five deformation passes and recovers the experimentally observed weakened, split basal texture. In all cases, the agent inferred the correct execution sequence from the task statement, robustly across repeated runs, and delivered physically interpretable results. This work establishes harness engineering as a systematic approach to automating CP modeling workflows while maintaining physical interpretability and auditability through visible reasoning traces.
[AI-606] Witeness Overlap: Directional Provenance Inside Open-Weight Model Families NEURIPS2026
链接: https://arxiv.org/abs/2609.31784
作者: Siyuan Li,Haoxuan Zeng,Xin Luo,Fernando Jia,Florence Li,Zhengyang Geng,Zico Kolter,Tai Sing Lee,Tianqin Li
类目: Artificial Intelligence (cs.AI)
备注: accepted at neurips 2026
Abstract:Open-weight models are often released, fine-tuned, aligned, merged, and re-released, making provenance audits ask not only whether checkpoints are related, but also which checkpoint came first. Many existing model-provenance methods are designed for a base-known audit setting: given a victim or source model, they test whether a suspect model is related to it. Although these audits are framed as source-to-suspect tests, their underlying evidence is often symmetric, relying on representation similarity, weight similarity, behavioral fingerprints, or correlation statistics. Symmetric pairwise comparisons can detect relatedness, but they cannot by themselves orient relationship between checkpoints A and B. We therefore introduce a local geometric comparison: instead of comparing two checkpoints directly, we add a third same-family checkpoint as a witness and compare the geometry around each candidate endpoint. Direction is inferred by asking which candidate behaves more like a branching parent. Motivated by this idea, and by the empirically observed asymmetry between parent-anchored and child-anchored witness-overlap distributions, we propose Witness Overlap, a prompt-free, training-free white-box test for directional provenance. On 176 LLM checkpoints from 16 families, our one-witness test orients 95.3% of parent-child decisions using Frobenius cosine. We further evaluate root identification, sibling discrimination, generalizations to VLM and diffusion families, and chain-structured ordering. The signal is robust to weight noise and sparse pruning, with a proposed SVD weight reduction variant showing greater robustness than Frobenius cosine.
[AI-607] med Rule-Based Supervision of an End-to-End Autonomous Parking Policy
链接: https://arxiv.org/abs/2609.31773
作者: Kejia Gao,Liguo Zhou,Lei Yu,Alois Knoll
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:We study whether a manually specified runtime supervisor can correct recurring failures of an existing end-to-end parking policy in a fixed CARLA parking lot. The vision-based Transformer architecture is inherited from Yang et al.; our contribution is a timed, rule-based Parametric Safety Shield (PSS) applied to its control outputs. The PSS uses hand-calibrated speed, position, and duration thresholds to intervene in observed failure modes, including boundary exits, delayed braking, and stalled or oscillatory control. In the reported closed-loop evaluation, 16 held-out target slots and six initial poses are each evaluated in four rounds (384 attempts per configuration). Target success increases from 327/384 (85.16%) for the retrained policy to 375/384 (97.66%) with the PSS; mean position and orientation errors among successful attempts are 0.21m and 0.33 degrees. These results show an improvement within this simulator setup. The repeated attempts share one map, vehicle, and sensor configuration, and the PSS uses simulator world coordinates; thus the results do not establish generalization to other lots or real vehicles, or a formal safety guarantee.
[AI-608] SMARtCARE: Privacy-Preserving Agent ic AI Systems for Bounded-Autonomy Clinical Decision Support ALT
链接: https://arxiv.org/abs/2609.31763
作者: Srini Ramaswamy,Deveeshree Nayak
类目: Artificial Intelligence (cs.AI); Emerging Technologies (cs.ET); Software Engineering (cs.SE)
备注: Paper accepted to, and to appear in, the 2026 IEEE HealthCom Conference
Abstract:Long-context clinical AI systems can miss relevant patient history when prior admissions fall outside the active reasoning context. In ICU monitoring, this can cause early vital-sign drift to appear nonspecific even when it resembles a prior deterioration pattern. SMARtCARE addresses this gap through a four-state clinical decision-support architecture: Stable, Meta-cognitive, Assisted, and Regulated (Revoked). Rather than automatically retrieving prior records, SMARtCARE uses a lossy six-channel fingerprint of the patient’s prior trajectory. When current drift matches that fingerprint and the prior record is absent from context, the system raises a Meta-cognitive escalation for clinician review; full retrieval occurs only through clinician action in the Assisted state. A patient-identity guard is designed to enforce correct attribution across data loading, logging, and audit layers. Evaluation combines a synthetic Monte Carlo study that validates the state-transition logic and estimator stability, not clinical performance, with real-data runs on both the MIMIC-III and MIMIC-IV Clinical Database Demos. On MIMIC-III, one prior-pattern recurrence was identified among 14 two-admission patients; on MIMIC-IV, the same pipeline produced no fingerprint matches among 9 two-admission patients, which illustrates a key limitation of a fixed canonical pattern library. Across both runs all logged decisions were fully traceable and correctly attributed. The results support SMARtCARE as a traceable, privacy-aware mechanism for surfacing middle-context risk; they are not a clinical efficacy claim.
[AI-609] What Stops Recursive Self-Improvement in Robotics? Lessons from 123 Rounds of Agent ic Skill Discovery
链接: https://arxiv.org/abs/2609.31760
作者: Jiaming Wang(National University of Singapore)
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Technical report. 20 pages, 7 figures
Abstract:Can a robot improve itself the way coding agents now improve software? We built an agentic system to find out. It watches a robot fail, works out which capability is missing, writes new skills or finds and installs external models, tests every change in simulation, and repeats, with no human writing robot code. We ran it for 123 improvement rounds on household manipulation tasks. This report describes what we learned. The good news is that the agent can discover capabilities on its own: noticing that its targets were out of view, it asked for an active-viewing model, debugged it, and deployed a working search skill. The bad news is that its improvements did not add up. Changes kept passing their tests, yet the target task, putting condiments on the top shelf of a fridge, never succeeded. We found that the agent was rarely the bottleneck. Three things around it were. First, chained perception modules do not understand relations. Segmenters such as SAM 3 find shelves but not “the top shelf”, so the agent filled the gap with ever more geometric rules that never converged, when what it needed was a different kind of model. Second, skill chains lock learning onto the first step. Long tasks mostly fail early, so evidence and fixes pile up there, and later skills are rarely reached, tested, or improved. Third, what the agent learns is decided by the harness. The agent optimized exactly what the evaluator measured, including where it was wrong, and weak tests and misleading memory turned activity into a standstill. We distill these lessons into concrete recommendations for building robot systems that improve themselves, each paired with an experiment that could prove it wrong.
[AI-610] What does FFN compression change downstream? Same-state causal restoration in diffusion language models ICLR2027
链接: https://arxiv.org/abs/2609.31685
作者: Shaurya Omar
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 14 pages, 3 figures, ICLR 2027 submission
Abstract:Diffusion language models (DLMs) enable flexible, parallel generation, but their iterative denoising remains computationally expensive, motivating increasingly aggressive compression. Existing compression objectives largely measure how well compressed computation approximates the original locally, but local error does not reveal which removed computations actually matter to the downstream denoising trajectory. We introduce Same-State Causal Restoration (SSR), which restores the original FFN on the exact current input reached by the compressed model and measures how the resulting trajectory changes. To our knowledge, this is the first direct measurement of the same-current-input closed-loop effect of removed FFN computation in DLM compression. Across LLaDA-8B-Instruct and Dream-v0-Instruct-7B, compressed-side state ranks this downstream effect substantially better than local NMSE at fixed denoising phase, while controlled interventions show that correction structure matters beyond magnitude. Using task-label-free calibration, SSR freezes a single restoration window for held-out inference. Under aggressive LLaDA compression, restoring only four transitions recovers 89.9% of the lost accuracy while retaining an estimated 36.8% whole-model MAC saving and outperforming an equal-budget local-error baseline. Dream further shows that restoring dense behavior and repairing the final task are distinct outcomes.
[AI-611] Does Joint-Embedding Predictive Architecture Pretraining Help Time Series Forecasting?
链接: https://arxiv.org/abs/2609.31680
作者: Yutong Feng,Bowen Liao,See Kiong Ng,Yuxuan Liang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:Joint-embedding predictive architectures (JEPA) have emerged as a promising self-supervised pretraining paradigm for time series, learning representations by predicting target embeddings in latent space rather than reconstructing raw signals. Yet evidence on their benefits remains mixed, and most studies test only a single backbone or a narrow set of architectures, leaving unclear whether JEPA pretraining is a reliable improvement or one that depends heavily on the downstream model. We address this gap through a large scale evaluation of one JEPA instantiation across nine backbones and eleven benchmarks spanning temporal and spatio-temporal forecasting, the most extensive cross architecture assessment of JEPA for time series to date. We find that the benefit of this instantiation varies sharply across backbones, producing consistent gains for some architectures and consistent degradation for others, even on the same dataset. This pattern holds across both task families, indicating the variability is a general property of this instantiation rather than a dataset specific artifact worth accounting for when choosing a backbone in practice.
[AI-612] Active Causal Discovery Benchmark: Evaluating LLM Agents Under Budgeted Interventions
链接: https://arxiv.org/abs/2609.31675
作者: Sagar Deb,Devam Shah,Ashwanth Krishnan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce the Active Causal Discovery Benchmark (ACDB), an SCM-grounded environment for evaluating whether LLM agents recover causal graph structure from observations and budget-constrained hard interventions. ACDB pairs a linear-Gaussian world generator with a fixed observe-intervene-submit API and a three-layer scoring contract that separates skeleton recovery, DAG recovery, and intervention efficiency. On the current six-level ladder, PC with a greedy active orientation heuristic is the strongest non-oracle method (directed F1 42.7%, SHD 4.79), ahead of Claude Sonnet 4.6 raw active (31.7%, 7.25) and GPT-5.4 raw active (22.9%, 9.27). The most informative diagnostic is the precision-recall decomposition: PC under-commits with high precision, LLMs over-commit with lower precision, and statistical-tool access often increases abstention rather than useful intervention. A structure-blind random DAG baseline reaches 23.6% directed F1 on this dense v0 ladder; a density probe lowers this floor to 16.9%, motivating the v1 calibration pass. The current results should therefore be read as a benchmark audit and calibration report, not as evidence that current LLMs solve active causal discovery.
[AI-613] Open-Qwen -Music: An Auditable Framework for LLM -Based Music Composition and Diffusion Rendering
链接: https://arxiv.org/abs/2609.31652
作者: Yangbin Yu,Mingyu Yang
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注:
Abstract:We present Open-Qwen-Music, an open reconstruction of Qwen-Music and a fully specified research system for text-to-music generation that couples LLM-based semantic composition with diffusion-based acoustic rendering. The system comprises a 25 Hz single-codebook music tokenizer, a 3B-parameter autoregressive Music LLM, and a diffusion renderer producing 48 kHz stereo audio, following the cross-module interfaces reported by Qwen-Music. The strongest systems of this design remain closed, and prominent open music-generation projects release weights and inference code without their training corpora or end-to-end training implementations. This limits independent and controlled study of how information loss and prediction errors propagate from semantic representation through autoregressive planning to acoustic rendering. To our knowledge, Open-Qwen-Music is the first fully open release of an LLM-composition-plus-diffusion-rendering text-to-music system. Beyond model weights and inference code, the release includes the training datasets and provenance manifests, complete data-processing, annotation, training, inference, and evaluation pipelines, configurations, and pretrained weights for every learned module. Artifact manifests bind the identities of these artifacts across the complete workflow. Together, these artifacts establish a reproducible implementation of the modular architecture and provide an empirical basis for component-level analysis and future evaluation. We present the system as a transparent, executable research baseline and a starting point for the community, not as evidence of quality parity with Qwen-Music. Open-Qwen-Music is an ongoing effort, and we will continue to improve its generation quality, controllability, and robustness. All release artifacts are available at this https URL.
[AI-614] Energy Vision–Language–Action: A Controlled Multimodal Benchmark for Intent-Conditioned Residential Energy Management
链接: https://arxiv.org/abs/2609.31648
作者: Lyes Saad Saoud,Oualid Doukhi,Ehsan Reihani,Saeed Sepasi,Deok Jin Lee,Moussa Ayyash,Reza Ghorbani
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注:
Abstract:Vision-Language-Action (VLA) models are studied mainly in robotics, where visual observations and language instructions are mapped to physical actions. This paper introduces Energy Vision-Language-Action (EVLA), a controlled multimodal benchmark for intent-conditioned residential energy management. EVLA frames battery scheduling as a multimodal trajectory-prediction problem in which an RGB energy-field representation, a numerical operating state, and a natural-language objective are mapped to a 16-step battery-action trajectory generated by a finite-horizon sampling-based reference generator. Source windows are derived from public residential electrical-load data, while electricity price, battery state of charge, indoor temperature, and time of day are generated benchmark metadata. A hidden operating regime is encoded only through energy-field texture, enabling paired visual changes while the explicit numerical state is fixed. Crossing 439,203 retained base windows with three hidden regimes and five language objectives yields 6,588,045 multimodal instances. An initial study evaluates 36 configurations over three training seeds using fixed subsets of 5,000 training, 500 validation, and 500 test instances. In the MobileNet-family comparison, removing processed language increases trajectory mean-squared error from 0.3856 +/- 0.0039 to 0.8628 +/- 0.0001, whereas removing vision yields 0.3843 +/- 0.0013, comparable to the full model. The results show strong asymmetry in modality use: the processed-language pathway is strongly associated with prediction quality, while the current RGB pathway provides no aggregate error advantage. These results characterize the fixed pilot subset and executed protocol rather than full-benchmark training. EVLA provides a controlled setting for studying how semantic intent and latent context influence residential energy-action prediction.
[AI-615] yped Temporal Interaction Features for Simulation-Backed Forecasting of Open-Source Game Release Incidents
链接: https://arxiv.org/abs/2609.31647
作者: Shayma Alkobaisi,Anas Ali
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Open-source video-game quality depends on inter-actions among code, assets, configuration, tests, contributors, and issue workflows, yet conventional defect predictors usually flatten or omit these relations. We investigate release-level forecasting of a quality incident within thirty days using GAMEQUALGRAPH-Pilot, a typed temporal feature pipeline with calibrated risk estimates and effort-aware ranking. Because the accessible OS-SGameBench materials do not provide manually audited release dates and outbreak labels, the executed evaluation is explicitly simulation-backed rather than an empirical claim about real games. Five seeded worlds each contain 120 projects and 24 releases, with project-disjoint validation and future cross-project testing. The pilot obtains an AUPRC of 0.520, AUROC of 0.673, Brier score of 0.207, and 29.68% effort-aware recall at a twenty-percent testing budget. Its closest local comparator, Static-Hetero-Reimpl, reaches 0.522 AUPRC; the -0.002 difference is not statistically significant after Holm correction. Inference requires 0.023 milliseconds per release in the measured environment. Ablations and controlled missingness, drift, engine, project-size, alert-threshold, and attribution analyses expose where typed interactions help and where they fail. Results support the reproducibility of the proposed protocol, not deployment effectiveness. Real OSSGameBench release reconstruction, stratified label audits, and official graph-model comparisons remain mandatory before journal submission or operational use in practice. This boundary protects research integrity and supports credible evaluation.
[AI-616] STAR: Adaptive Spatial-Temporal Normalization for Unified Microservice Incident Management
链接: https://arxiv.org/abs/2609.31645
作者: Xinhua Miao,Linyu Zhu,Bowei Yang,Zhengong Cai
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Automated incident management in large-scale microservice systems relies on learning robust representations from multimodal observability data, including metrics, logs, and traces. Although recent self-supervised frameworks enable unified modeling for anomaly detection (AD), failure triage (FT), and root cause localization (RCL), they often struggle with non-stationary temporal dynamics and heterogeneous service dependency structures. In this paper, we propose STAR, a Spatial-Temporal Adaptive Representation learning framework that explicitly addresses these challenges through adaptive normalizations. STAR introduces two tightly coupled mechanisms: Temporal Adaptive Normalization (TAN), which dynamically normalizes multivariate time series using multi-scale temporal context, and Spatial Adaptive Normalization (SAN), which performs structure-aware normalization over service dependency graphs. Unlike prior methods that treat normalization as static or task-agnostic, STAR formulates it as a learnable, context-conditioned transformation aligned with the intrinsic properties of microservice systems. The resulting adaptive representations are integrated into a unified self-supervised framework, enabling end-to-end unsupervised support for AD, FT, and RCL tasks. Extensive experiments on two real-world microservice benchmarks demonstrate that STAR consistently outperforms all state-of-the-art baselines, yielding significant and stable improvements across all three tasks. Our results highlight adaptive normalization as a principled and effective mechanism for robust multimodal representation learning in complex software systems.
[AI-617] MaD-RL: Matching Distributions for Calibrating LLM s with Reinforcement Learning
链接: https://arxiv.org/abs/2609.31644
作者: Sourabh Kulkarni,Ksheeraj Sai Vepuri,Basar Demir,Jason Bohrer,Emily Shen,Jianfa Chen,Nan Jiang,Ankit Jain,Harihar Subramanyam,Mannat Singh,Chirag Nagpal
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning (RL) is widely used in language-model post-training to maximize rewards assigned to individual model outputs, such as scores from binary verifiers or reward models trained on human feedback. However, applications such as synthetic-data generation, fairness-related constraint satisfaction, and policy exploration require controlling the distribution of outputs across model generations rather than only maximizing expected reward. We propose a general RL-based framework for \textitDistribution Matching allowing matching the distribution of a latent categorical attribute of model outputs to a specified target distribution. Empirically, we demonstrate that dominant post-training recipes such as Group Relative Policy Optimization (GRPO) reduce output diversity by concentrating policy probability towards a single mode. Entropy regularization and sampling temperature can improve the spread of the distribution but have constrained effectiveness, limited to apply only in token space and toward uniform distributions. We show that prior work in this area is a specific case of Distribution Matching involving the L_2 divergence. We then propose reward functions for other divergences such as KL and Jensen-Shannon and motivate them with theoretical justification. Finally, we demonstrate the effectiveness of our approach on a set of experiments involving mathematical reasoning and programming.
[AI-618] Information Design Against Gaming and Learning Adversaries
链接: https://arxiv.org/abs/2609.31643
作者: Madhava Gaikwad
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Computer Science and Game Theory (cs.GT)
备注: Accepted at Gamesec 2026 for Oral
Abstract:A principal who deploys a binary classifier with an abstention option must decide which queries the mechanism abstains on. The right choice depends on the adversary. A gaming adversary already knows the classifier and tries to manipulate features across the boundary, so the principal does best by abstaining on queries close to that boundary. The same boundary-localizing rule is the worst possible choice against a learning adversary who does not know the classifier: each abstention now tells the adversary that the boundary is nearby, which is enough to drive a binary search. We analyze this tension. The two natural defenses, abstaining at a fixed rate and abstaining near the boundary, are Blackwell-incomparable: neither can be simulated by post-processing the other’s responses. The number of queries needed to reconstruct the boundary to error \eps is \tilde\Theta(d/\eps) under the first defense and \Theta(d \log(1/\eps)) under the second, where d is the VC dimension of the classifier family and \tilde\Theta suppresses factors polylogarithmic in d and 1/\eps . The first rate is a worst case over query distributions; no reconstruction algorithm can close the gap at the distributions that attain it. We characterize the Pareto frontier between the two defense objectives, and confirm both rates on seven binary-classification tasks spanning tabular, image, and language-model-feature inputs: label-plus-counterfactual access extracts the boundary with up to 200\times fewer queries than a published label-only baseline.
[AI-619] Measure Learning at Steady State: A BIRD-SQL Formula 1 Case Study
链接: https://arxiv.org/abs/2609.31640
作者: Manoj Bajaj
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Continual Learning Bench scores learning as short-horizon gain versus a reset baseline and finds naive full-context ICL strongest among the memories it tested. We treat ICL as one learning system and score it on a longer shared-world schedule. Steady-state learning is the gap versus baseline on a pre-set late window (last 40 of 174 BIRD-SQL formula-1 questions). We split the score into exploration efficiency (SQL probes), task reward (hits), and delivery cost (API dollars and context size). On gpt-5.6-luna, late probes fall from 4.6-5.6 to 0.95 while hits rise only modestly and ICL context grows to about 95k tokens with cost roughly doubling. Short-horizon gain understates the late probe saving and misses the cost inversion, so we find that unbounded ICL is a poor candidate for the learning mechanism.
[AI-620] When Does Domain Adaptation Help on Physical Vibration Sensors? A Held-Out-Bearing Study of Neural-Operator and Convolutional Models
链接: https://arxiv.org/abs/2609.31639
作者: Kumbha Nagaswetha,Rabi Pathak
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Diagnosing rolling-element bearing faults from vibration is a canonical physical-sensing task and a widely used benchmark for domain adaptation under operating-condition shift. Accuracies above 99 percent are commonly reported, but under evaluation splits that place the same physical bearing in both training and test. We revisit the task under a held-out-bearing protocol, assigning every bearing unit entirely to either the training or the test set, and find that source-only transfer is far weaker than such numbers suggest: on a change of shaft speed it reaches only 0.36 , against a target-supervised ceiling of 0.97. We then study what governs transfer. Treating computed order tracking, a shaft-angle resampling that places fault frequencies at fixed shaft orders independent of running speed, as a controlled change of representation, we find that a Fourier Neural Operator raises source-only transfer from 0.36 to 0.61 on the speed shift, where the fault peaks move, while a convolutional network of matched feature dimension stays near chance in both representations. The representation also decides whether unsupervised alignment can work: with the same normalized RBF-MMD loss and no target labels, the operator reaches 0.71 in the frequency domain but 0.95 in the order domain, within 0.02 of the target-supervised ceiling and above 0.86 on every held-out bearing fold. Once the representation is right, a small label budget adds little. These results indicate that, for this task, the input representation rather than the alignment method decides whether adaptation helps. A second dataset, whose held-out units are fault diameters rather than bearings, shows that the same protocol exposes failures that even a target-supervised model cannot avoid.
[AI-621] Energy-aware frugal Bayesian optimization
链接: https://arxiv.org/abs/2609.31638
作者: Gaston Plat,Paul Saves,Nathalie Bartoli,Thierry Lefebvre,Joseph Morlier
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Optimization and Control (math.OC); Classical Physics (physics.class-ph)
备注: Published in MATEC Web of Conferences 422, 02009. 7th International Conference on Engineering Optimization
Abstract:Modern design optimization frameworks aim first and foremost for models with the most accurate predictions without balancing computational overhead. It remains a reason why scaled architecture and multidisciplinary design optimization problems are difficult to address, even with sample-efficient Bayesian optimizers. In this paper, a metric quantifying the computational energy footprint is introduced within a Bayesian optimization framework to guide the parameter setting of a model towards configurations that balance both performance and frugality. The computer experiments highlighted existing tradeoffs between optimum convergence and the underlying energy footprint, and sometimes resulted in both a better-found optimum and lower energy consumption.
[AI-622] What Next-Event Accuracy Cannot See: Closed-Loop Evaluation of Emergency Department Trajectory Simulators
链接: https://arxiv.org/abs/2609.31635
作者: Zhen Xuen Brandon Low
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Clinical trajectory models are usually evaluated by next-event accuracy on observed histories. Simulation is different: models must condition on their own generated events, allowing errors to compound. Although this problem is well known in sequence modelling, it has not been systematically quantified for clinical trajectory simulators. We developed EDSim-Bench to evaluate this failure mode using 425,028 MIMIC-IV-ED stays, with external replication on MC-MED, and release the evaluation protocol and scoring code. Starting from held-out visit prefixes, models generate the remainder of each visit and are evaluated on termination, event composition, timing, conditional fidelity, and occupancy forecasting, with a train-only order-3 n-gram as a reference baseline. Despite next-event accuracies within 0.001, three neural architectures behaved very differently under rollout. Across seeds, one Transformer recipe ranged from 0.43 to 0.96 in termination score and from 4.2- to 137-fold the divergence of the n-gram; no prefix-trained neural model approached the n-gram on termination or event composition. Inference-time interventions improved termination but did not jointly recover composition and timing. Supervising every eligible sequence position rather than only the final prefix position was associated with one to two orders of magnitude lower divergence across Transformer, GRU, and LSTM models, with the pattern persisting under model scaling, temporal shift, and external-site evaluation. Nevertheless, even the best model generated visits approximately half as long as observed, and model rankings reversed on occupancy forecasting, a downstream quantity relevant to bed management. These results show that next-event accuracy is insufficient to evaluate clinical trajectory simulators and motivate closed-loop evaluation across seeds, rollout criteria, and downstream tasks.
[AI-623] QC-Stark: A Multi-Task Benchmark Revealing Capability Dissociations in LLM s Evaluated on Quantum Computing Tasks
链接: https://arxiv.org/abs/2609.35581
作者: Pranav Gupta
类目: Quantum Physics (quant-ph); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: accepted at the Quantum AI Workshop, Indianapolis IN, August 2026
Abstract:We introduce QC-Stark, a benchmark for evaluating large language models (LLMs) on 11 quantum computing (QC) tasks, spanning circuit construction, debugging, compilation, error correction, and simulation. Across 2,750 evaluations (10 models \times 11 tasks x 5 difficulty levels x 5 seeds), we find that overall rankings mask substantial per-task variation. The Spearman correlation between overall and per-task rankings is statistically insignificant for 4 out of the 11 tasks included in this benchmark. A 2-parameter Item Response Theory (IRT) model validates measurement quality, and prompt sensitivity analysis confirms ranking robustness across prompt conditions. All tasks are auto-verifiable via execution, thus not requiring any manual evaluation. We make the code and data publicly available on Huggingface.
[AI-624] Understanding Generalization Requires Universal Induction NEURIPS2026
链接: https://arxiv.org/abs/2609.34458
作者: Aram Ebtekar,Marcus Hutter,Danica J. Sutherland
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Information Theory (cs.IT); Machine Learning (cs.LG); Statistics Theory (math.ST)
备注: Published at NeurIPS 2026 Position Paper Track
Abstract:Classical statistical theory is insufficient to explain the successes of general-purpose AI models, because it depends on handcrafted inductive biases that it cannot justify. No Free Lunch (NFL) theorems force any learner that beats chance on some environments to underperform on others. We might hope that past experience informs which environments to expect, but NFL applies equally to meta-learning. Thus, any method that makes meaningful predictions necessarily begins with an inductive bias external to the data. Choosing to bias toward short programs yields Solomonoff induction (SI), whose performance is competitive against all computable learners - albeit up to “constants” that become large when comparing against specialized methods that exploit background information. We therefore relativize SI to an information vantage point, biasing toward short programs with access to all preexisting information. This reframes the inductive bias: instead of seeking some absolute notion of simplicity, we favor accessibility with respect to our vantage point. An algorithm can only outpredict the relativized SI to the extent that its code contains additional information about the data, and no algorithm can generate such information. While SI is incomputable and hence not a practical algorithm, it provides a formal optimum for inference in the limit of infinite compute, and there is evidence to suggest that frontier AI systems roughly approximate it. Thus, the only known answer to meta-NFL is rooted in algorithmic information theory, which we should expect to play a fundamental role in explaining the generalization behavior of modern (and future) AI systems.
[AI-625] AlphaPareto: Formulaic Alpha Discovery with LLM -Guided Multi-Objective Reinforcement Learning NEURIPS2026
链接: https://arxiv.org/abs/2609.34188
作者: Yingbo Zhao,Zeyu Yang,Zhoufan Zhu
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted by Neurips 2026 main track
Abstract:Formulaic alpha discovery is a core challenge in quantitative trading, as identifying alphas that work well together remains difficult. Recent reinforcement learning (RL) methods formulate this task as a Markov decision process (MDP), but two important issues remain unresolved. First, as the alpha pool evolves, the reward function changes accordingly, making the MDP inherently non-stationary. Second, most existing methods optimize a single objective, typically predictive power, while ignoring other important properties of a high-quality alpha pool. Motivated by these challenges, we propose AlphaPareto, an RL method for formulaic alpha discovery. To address non-stationarity, AlphaPareto augments the state to include both the alpha under construction and the current alpha pool, and applies a large language model (LLM) to encode the pool. This design allows the agent to adapt to the evolving search environment. To overcome the limitation of single-objective reward design, AlphaPareto replaces the scalar reward with a multi-objective vector-valued reward that simultaneously captures predictive power, temporal stability, perturbation robustness, and diversity, and optimizes these objectives through a Pareto-regularized learning procedure. Empirical applications to real-world datasets show that our AlphaPareto method outperforms its competitors.
[AI-626] A packet-level digital hardware twin for commissioning megahertz diagnostic edge AI and plasma control system integration in tokamaks
链接: https://arxiv.org/abs/2609.33994
作者: Semin Joung,Abhilasha Dave,Luca Scomparin,Filipp Khabanov,Zheng Yan,Benedikt Geiger,George McKee,Ryan N. Coffee,David R. Smith
类目: Plasma Physics (physics.plasm-ph); Artificial Intelligence (cs.AI)
备注:
Abstract:High-bandwidth plasma diagnostics increasingly provide inputs to machine learning and signal-processing algorithms intended for real-time tokamak control, but the complete path from diagnostic sampling to control-system handoff is difficult to commission because of their sampling rates. We develop a packet-level digital hardware twin for the megahertz diagnostic edge-AI architecture. The simulator represents a 64-channel, 1 MHz beam emission spectroscopy (BES) diagnostic embedded in a 96-channel dual-carrier acquisition system, two 48-channel streams with SPAD0 sample counters, 10 GbE Hardware UDP transport, packet loss and network jitter, FPGA parsing and dual-carrier alignment, causal preprocessing and edge inference, a compact Ethernet result packet, receiver-side shared state, and a 1 kHz PCS-like control cycle. Binary UDP payloads and PCAP files are generated rather than emulating transport only at the array level. With a baseline of 20 samples per packet, each carrier generates 50,000 packets s ^-1 and 100 MB s ^-1 of user payload. A 110 ms reference run produces 11,000 HUDP packets; an intentionally dropped 20-sample packet is detected by the sample-counter continuity logic and invalidates the two overlapping 128-sample inference windows without silent interpolation. For valid windows, the configured engineering latency model gives a median last-input-to-shared-memory latency of 91.6 us and a 99th percentile of 108.1 us. A separate operating-system loopback test sends binary FPGA-result datagrams through a UDP receiver into POSIX shared memory and preserves packet sequence and CRC for 20/20 packets. Interactive GUI interfaces expose timing, packetization, network faults, inference thresholds, and control-state inspection. The framework provides a reproducible environment for testing diagnostic-to-accelerator interfaces and fail-safe behavior for deployment on fusion devices.
[AI-627] HARMONIA: Interpretable Graph Learning through Mixtures of Neural Bases
链接: https://arxiv.org/abs/2609.33972
作者: Quan D. Bui,Nguyen Do,An Nguyen Dang,Huyen Nguyen,Nhu Duc Minh Nguyen,My T. Thai
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Existing interpretable graph additive models still face limitations in either computational scalability or modeling flexibility. In terms of structural modeling, previous approaches either face quadratic scaling costs or sacrifice explicit source-to-target contribution decomposition. In terms of feature components, they rely either on per-feature neural networks or on single shared bases with limited feature specialization. We address both problems by introducing HARMONIA: Interpretable Graph Learning through Mixtures of Neural Bases, an interpretable-by-design framework. For feature modeling, HARMONIA introduces a Mixture of Neural Bases (MoNB), which routes features to specialized basis experts, enabling parameter sharing without sacrificing feature-specific specialization. For structural modeling, HARMONIA uses Relative Random Walk Probabilities (RRWP) to capture multi-hop and multi-path relationships, and proposes Sparse RRWP Aggregation (SRA) to compute these interactions through sparse graph propagation without quadratic pairwise complexity. HARMONIA retains a simple additive form in which predictions decompose into feature responses modulated by structural influence. Empirically, HARMONIA achieves stronger explanation recovery than existing interpretable graph baselines while maintaining competitive predictive performance and scaling to graphs with millions of nodes. These results show that interpretable graph learning can remain both faithful and scalable without sacrificing predictive effectiveness.
[AI-628] An Active-Bottleneck Mechanism for Weak-to-Strong Generalization
链接: https://arxiv.org/abs/2609.33835
作者: Mohammad Zeinalpour,Amir Najafi
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 76 pages
Abstract:Weak-to-strong generalization (W2SG) occurs when a student trained on a teacher’s predictions outperforms that teacher. We study when this happens under fully converged, ridgeless two-stage learning, with no early stopping, no explicit regularization, and no assumption that the student is more expressive than the teacher. In two-stage linear regression, a teacher is fit from n labeled examples and a student is trained solely on the teacher’s predictions on m fresh, unlabeled inputs. Although both stages share the same hypothesis class and the same training rule, we show that the student outperforms the teacher exactly when m lies in an explicit intermediate range: too few pseudo-labels leave the student without enough signal, too many let it inherit the teacher’s noise. Under power-law covariance, we derive this range in closed form as a function of the spectral decay and noise level, including regimes where the improving region splits into two disjoint intervals of m . We then study a random-feature model in which the student has strictly more features than the teacher, and identify two regimes, again given by explicit thresholds: one where improvement occurs only for m in a bounded interval, and one where it occurs only once the student width N_S exceeds an explicit threshold. Both regimes are governed by a single “active-bottleneck” principle: whichever of m or N_S is scarcer controls how much teacher error is filtered out, while increasing the other resource only reduces estimation noise. Together, these results show that finite data and finite width can themselves regularize a two-stage learner, with no explicit mechanism doing so.
[AI-629] Explainable Deep Learning of Resting-State Functional Connectomes Reveals Network Biomarkers of Adolescent Intelligence ALT
链接: https://arxiv.org/abs/2609.33422
作者: Md. Tanvir Rahman,Nabil Anan Orka,Asaduzzaman Khan,Mohammad Ali Moni
类目: Neurons and Cognition (q-bio.NC); Artificial Intelligence (cs.AI)
备注: 12 pages, 2 figures. This work has been submitted to the IEEE Journal of Biomedical and Health Informatics for possible publication
Abstract:Mapping resting-state brain organization to individual differences in cognitive ability remains a major challenge in population neuroinformatics. Although deep learning enables flexible modeling of brain connectivity, limited interpretability restricts its scientific and clinical utility. To address this objective, we developed an explainable deep learning framework based on sparse projected residual networks to predict fluid, crystallized, and total intelligence from resting-state functional magnetic resonance imaging in 5,285 participants from the Adolescent Brain Cognitive Development study. We incorporated three complementary explainability methods (Integrated Gradients, Gradient Shapley Additive Explanations, and Occlusion) to interpret model behavior. The framework outperformed existing approaches, achieving Pearson correlations of 0.44, 0.58, and 0.56 for fluid, crystallized, and total intelligence, respectively, corresponding to predictive improvements of 6 to 9 percent. All three explainability methods produced near-identical feature rankings (pairwise rank correlations greater than 0.99). Consensus maps revealed a dual-layered functional architecture where primary predictive hubs localized within canonical systems, while the strongest global predictive pathways frequently bypassed these hubs through distributed, long-range relay connections. These findings suggest that intelligence emerges from the interaction between localized computational hubs and distributed communication pathways. Ultimately, these normative network architectures provide clinical reference maps to detect individual deviations, supporting earlier diagnosis, cognitive subtype stratification, and treatment monitoring in atypical neurodevelopment.
[AI-630] Which Self-Improvements Should We Trust? Reliable Self-Improvement When Agents Reuse Their Benchmarks
链接: https://arxiv.org/abs/2609.33180
作者: Xiaojing Sun,Yuhan Zeng,Zihua She,Xiao Wang
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Methodology (stat.ME)
备注:
Abstract:As recursive self-improvement (RSI) rapidly advances, reliable evaluation becomes critical for guiding adaptive search. RSI typically relies on finite evaluation resources, such as fixed benchmarks, to determine which modifications are retained and what is proposed next. However, when these finite resources are repeatedly reused, new candidates are proposed based on feedback from the same evaluation set, so the search trajectory can adaptively overfit and empirical improvement may not reflect genuine population improvement on the underlying task distribution. Some existing methods account for multiple comparisons but assume that candidates are chosen independently of the evaluation set, and therefore do not control this adaptive dependence. To address this, we propose REUSE (Risk-controlled Evaluation Under Sequential Evolution), a certified evaluation and promotion framework that allows a fixed evaluation set to support repeated adaptive decisions while providing statistical guarantees. For a user-specified error level \alpha , with probability at least 1-\alpha , every promoted modification is a genuine population improvement on the underlying task distribution. REUSE achieves this by strictly limiting the evaluation feedback returned to the search process and accounting for possible promotion histories within the error budget. We develop detailed statistical theory for RSI evaluation in this setting, including simultaneous error control, valid lower bounds on cumulative improvement, and a characterization of the fundamental limits of adaptive evaluation reuse. In live self-improvement experiments, REUSE commits substantially fewer false promotions than evaluation frameworks from current RSI systems and error-controlled baselines, reducing the proportion of false promotions from up to 20.7% to 0%, while achieving final true population performance comparable to the best baselines.
[AI-631] Quantum Monte Carlo Tree Search with Fixed Confidence
链接: https://arxiv.org/abs/2609.33132
作者: Mingjie Hu,Jian-Qiang Hu,Enlu Zhou
类目: Quantum Physics (quant-ph); Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances in quantum computing are opening new opportunities for computationally intensive decision problems. This paper studies how quantum computing can improve Monte Carlo tree search (MCTS) in the fixed-confidence setting, where the goal is to identify a near-optimal move in a given game tree with high probability while minimizing query complexity. We first formulate MCTS under a quantum oracle model. We then develop a quantum MCTS algorithm (QMCTS) that combines threshold-based elimination on the search tree with quantum Monte Carlo estimation to reduce the cost of evaluating stochastic leaf values. We establish an instance-dependent lower bound on the query complexity and derive a corresponding upper bound for QMCTS. The lower-bound analysis introduces a new quantum phase-testing result that may also be useful beyond MCTS. We validate the theoretical results through simulation experiments and further demonstrate the feasibility of QMCTS on real quantum hardware.
[AI-632] More than 83.69% of the zeros of the Riemann zeta function are distinct
链接: https://arxiv.org/abs/2609.33043
作者: Kristian Muri Knausgård
类目: Number Theory (math.NT); Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO)
备注: 6 pages
Abstract:The lower asymptotic proportion of distinct nontrivial zeros of the Riemann zeta function, counted with multiplicity, is at least 0.8369928814\ldots . The previous bound was 0.83625\ldots . As in the proof of that bound, an unconditional version of Montgomery’s pair-correlation theorem gives an asymptotic energy estimate. The new ingredient is a short matrix inequality with a free clipping parameter. It strengthens the lower bound for this energy in terms of the number of distinct zeros. The gain is a nonnegative correction from overlaps between different nearby zeros on the critical line, which is retained even when some of these zeros are double. The matrix inequality, the threshold lemma, the block dichotomy, the counting assembly and the exact arithmetic are proved formally in Lean 4. The constant relies on a computer-assisted local inequality from recent work that has not yet been refereed. That computation was re-run independently, and every imported input is listed. This paper is primarily an experiment in AI-assisted mathematical research (Section 4).
[AI-633] Radiomap Blind Prediction under Incomplete Observation: Error Characterization and Correctable Propagation-Prior Learning
链接: https://arxiv.org/abs/2609.32836
作者: Xiaojie Li,Yu Han,Han Fang,Shangqing Liu,Guangxu Zhu,Shi Jin,Chao-Kai Wen
类目: ignal Processing (eess.SP); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: This work has been submitted to the IEEE for possible publication
Abstract:Radiomap blind prediction aims to infer radiomaps from observable representations of the propagation environment and base station configuration without field measurements. In practice, the observable representations are inherently incomplete. Thus, the target radiomap is not fully determined by the inputs when generalizing to unseen configurations or environments. Under incomplete observation, we establish a population-level theory of deterministic radiomap blind prediction that identifies the conditional mean as its optimal target and separates prediction error into reducible predictor approximation and irreducible uncertainty caused by missing physical information. The framework further characterizes the train-test risk gap and the uncertainty reduction enabled by observation enrichment. Building on it, we reveal the dual role of propagation priors: they provide physically grounded guidance, yet their implementable forms may bias the attainable predictor. This motivates RadioDecomp, which treats a prior-guided predictor as a correctable base and learns its remaining predictable discrepancy through residual refinement. To evaluate RadioDecomp across distinct propagation-prior designs, we instantiate it with a feature-guided monolithic base and a LoS-Shadow structured base, yielding RadioFR and RadioLSR, respectively. Across random, cross-configuration, and cross-environment settings, experiments confirm the benefit of propagation-related representations and show that both instantiations improve upon their respective bases. Further controlled studies on base capacity, training-support coverage, and observation coarsening corroborate the proposed analysis.
[AI-634] LaMET-Agent : An Agent Framework for Large-Momentum Effective Theory Analysis
链接: https://arxiv.org/abs/2609.32225
作者: Jinchen He,Xiangyu Jiang,Fei Yao,Dian-Jun Zhao
类目: High Energy Physics - Lattice (hep-lat); Artificial Intelligence (cs.AI); High Energy Physics - Phenomenology (hep-ph)
备注:
Abstract:Large-momentum effective theory (LaMET) provides a first-principles framework for computing the x dependence of light-cone parton distributions from lattice QCD. Over the past decade, theoretical and numerical advances have established a mature multi-stage workflow for systematic calculation of parton physics, although its implementation still requires expert judgment and substantial repeated effort. We present lamet-agent, an open-source large language model (LLM) agent framework that organizes this workflow into an executable, reproducible, and inspectable analysis pipeline. The present release supports collinear quark distributions and implements correlator analysis, renormalization, Fourier transformation, perturbative matching, continuum, physical pion mass and infinite-momentum extrapolations, and automated result review. We validate it on four end-to-end analyses: pion parton distribution functions in the gauge-invariant and Coulomb-gauge formulations, and pion and kaon distribution amplitudes, obtaining results consistent with the published calculations. Extensions to transverse-momentum-dependent distributions, generalized transverse-momentum-dependent distributions, and gluonic distribution functions are planned for subsequent releases.
[AI-635] Rank Confidence Sequences:Anytime-valid Leaderboards
链接: https://arxiv.org/abs/2609.32211
作者: Hamed Khosravi,Xiaoming Huo
类目: Methodology (stat.ME); Artificial Intelligence (cs.AI)
备注:
Abstract:Leaderboards rank models by their average scores on benchmark items, and they are consulted repeatedly while the evaluation is still running. Existing confidence intervals for a model’s rank control their error rate only if they are computed once, after a number of items chosen in advance. If they are recomputed as results arrive, and the evaluation stops once they look decisive, their error rate exceeds its nominal level. Anytime-valid methods keep their guarantees at all sample sizes simultaneously and hence under any stopping rule. They exist for the accuracy of one model, for one pair of models and for the set of models that may be best. For pairwise battles they also give ranks. None gives ranks when all models are scored on the same items, which makes their scores dependent. We construct rank confidence sequences: for every model, a set of ranks that contains its true rank, simultaneously for all models and at all times, at a chosen error level \alpha , in finite samples. The construction combines betting e-processes, one for each ordered pair of models, with closed testing over the possible orderings of the models. It allows any dependence between the models’ scores on an item. The method has two advantages. A leaderboard can be inspected after every item without inflating its error rate. The evaluation of each model can stop as soon as the question asked about it is answered, which saves compute. When results are examined only once, halfway through or later, little power is lost relative to fixed-sample methods. The paper quantifies these advantages in simulations and on public leaderboard data.
[AI-636] PRIME-ANC: Path-Ratio-Informed Modeling for Efficient Neural Filter Synthesis in Active Noise Control
链接: https://arxiv.org/abs/2609.31772
作者: Yaokun Huang,Chunyang Xu,Haowen Hua,Sen Lin,Shichao Hu,Mengyao Zhu
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI); Sound (cs.SD); Signal Processing (eess.SP); Systems and Control (eess.SY)
备注:
Abstract:Changes in listener acoustics require active noise control (ANC) filters to be redesigned for new acoustic paths. We introduce PRIME-ANC, a shared neural synthesizer that learns a bounded, path-dependent logmagnitude correction to a regularized path-ratio base. Minimum-phase reconstruction and truncation produce finite-impulse-response (FIR) filters; training optimizes their noise-control performance. Across ten random training/test splits per dataset, PRIME-ANC achieves average held-out one-third-octave reductions of 18.81 and 17.76 dB over 50 Hz-5 kHz on a ten-path dataset and a public earphone database, respectively. On original earphone measurements, it improves upon the path-ratio base by 7.89 dB. Ablation studies support the contributions of both the analytic base and the path-dependent correction. Given calibrated paths for a held-out listener condition, PRIME-ANC generates a path-specific FIR without iterative optimization. With three Gauss-Newton updates, it reaches 21.43 dB reduction, approaching direct weighted least-squares design while producing lower amplification and root-mean-square control output.
[AI-637] Measurement-Error-Aware Causal Distributed-Lag Quantile Modeling of Indoor Air Pollution and Short-Term Lung-Function Deterioration
链接: https://arxiv.org/abs/2609.31646
作者: Shayma Alkobaisi,Anas Ali
类目: Applications (stat.AP); Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Machine Learning (cs.LG)
备注:
Abstract:Low-cost indoor air-quality sensors could support personalized asthma prevention, but their nonlinear measurement error, delayed exposure effects, time-varying confounding, and heterogeneous lower-tail responses limit risk estimation. We present CAUSALQUANT-ASTHMA, a measurement-error-aware causal quantile distributed-lag framework for short-horizon peak expiratory flow analysis. Sparse reference measurements train a nonlinear calibration model; stabilized sequential generalized-propensity weights address measured exposure assignment; and a susceptibility-modulated, smooth, noncrossing quantile model estimates lag-specific and sustained-exposure contrasts. Because no authorized cohort simultaneously provided dense indoor sensing, reference co-location, and outcome-compatible longitudinal data, evaluation used five semi-synthetic panels with known counterfactual truth, 150 patients and 12,600 patient-days per realization. Across eight methods, CAUSALQUANT achieved a dose-response integrated absolute error of 0.304 plus or minus 0.094, improving 24.2 percent over the strongest measurement-error and propensity-weighted baseline. It also obtained the lowest overall pinball loss, 1.065, while maintaining zero quantile crossings and 78.1 percent coverage for the nominal 80 percent interval. Sensor calibration reduced held-out exposure RMSE by 33.7 percent. Stress tests quantified degradation under sensor noise, missing personal measurements, and hidden confounding. These findings establish methodological feasibility and reproducibility, not clinical effectiveness; prospective, governance-approved external validation is required before patient-level interpretation or deployment.
[AI-638] l_1-2 GLasso: L_1-2 Regularized Multi-task Graphical Lasso for Joint Estimation of eQTL Mapping and Gene Network
链接: https://arxiv.org/abs/2301.02225
作者: Wei Miao,Lan Yao
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Statistics Theory (math.ST); Quantitative Methods (q-bio.QM)
备注:
Abstract:A critical problem in genetics is to discover how gene expression is regulated within cells. Two major tasks of regulatory association learning are : (i) identifying SNP-gene relationships, known as eQTL mapping, and (ii) determining gene-gene relationships, known as gene network estimation. To share information between these two tasks, we focus on the unified model for joint estimation of eQTL mapping and gene network, and propose a L_1-2 regularized multi-task graphical lasso, named L_1-2 GLasso. Numerical experiments on artificial datasets demonstrate the competitive performance of L_1-2 GLasso on capturing the true sparse structure of eQTL mapping and gene network. L_1-2 GLasso is further applied to real dataset of ADNI-1 and experimental results show that L_1 -2 GLasso can obtain sparser and more accurate solutions than other commonly-used methods.
机器学习
[LG-0] Unifying Distributional Training for One-Step Visual Generation
链接: https://arxiv.org/abs/2609.35763
作者: Chi Zhang,Haoyang Shi,Yueyi Liu,Ruichuan An,Junkang Zhou,Chang Li,Xiuyuan Lu,Yichi Zhang,Bo Wang,Yuhang Wu,Sen Cui,Miao Liu
类目: Machine Learning (cs.LG)
*备注:
Abstract:\emphDistributional training provides collective supervision for one-step visual generation by matching real and generated features in frozen representation spaces. We introduce \empha unified theoretical framework that separates distribution modeling from matching discrepancy and connects global objectives to pointwise feature updates through Wasserstein gradient flow. Under this framework, FD-Loss and Gaussian-kernel Drifting are recovered through Gaussian optimal transport and kernel-density-based KL matching, respectively. The framework motivates \textbfMGFlow, which models feature distributions with Gaussian mixtures at an adjustable granularity between global moments and sample-based representations. MGFlow supports both optimal transport and score-based matching, and couples mass-constrained sample assignment with paired component updates to address mode collapse that mixture expressivity alone does not resolve. On ImageNet 256\times256 , MGFlow substantially surpasses the FD-Loss baseline, achieving state-of-the-art results with \textbf1.45 \mathrmFDr^6 on pMF-H and \textbf1.64 on JiT-H. For text-to-image generation, MGFlow post-trains FLUX.2 [klein] 4B into a one-step generator that outperforms the original four-step model on both GenEval and PickScore. Project page: this https URL
[LG-1] Statistical Learning of Contractive Dynamical Representations for Composite Adaptive Control IROS2026
链接: https://arxiv.org/abs/2609.35758
作者: Min Kim,José Leonardo Brenes,Fred Hadaegh,Soon-Jo Chung
类目: ystems and Control (eess.SY); Machine Learning (cs.LG); Robotics (cs.RO)
*备注: 9 pages, including an additional one-page appendix in this arXiv version. Accepted to the 2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2026)
Abstract:We present a representation-learning framework for composite adaptive tracking control under dynamically coupled disturbances. The framework connects classical disturbance-accommodating control (DAC) to recent last-layer adaptive disturbance-rejection methods. Specifically, we introduce a statistically principled hard expectation-maximization (hard-EM) procedure, with a Kalman smoother in the hard E-step, to identify dynamical representations of disturbance whose latent evolution is uniformly contractive. The learned representation evolves a latent disturbance-excitation state from measured plant features and control inputs and decodes that state into the time-varying disturbance acting on the nominal plant, thereby extending prior “fixed-decay” last-layer adaptive methods to a learned, predictive DAC-style formulation. Combined with Bayesian filtering of the learned latent state, this representation yields a composite adaptive tracking controller with predictive capability and provable exponential convergence to a bounded neighborhood. We validate our approach experimentally on a slippery ground vehicle carrying a liquid-sloshing tank and a pendulum load, and we further assess its robustness on a system of coupled Duffing oscillators. Across both settings, the method achieves accurate disturbance prediction and improved overall tracking performance relative to fixed-decay representation-learning ablations, LTI disturbance-accommodating baselines, and model-based PD baselines.
[LG-2] Neural Harmonic Measure Operator NEURIPS2026
链接: https://arxiv.org/abs/2609.35752
作者: Jinjin He,Sinan Wang,Yuchen Sun,Bo Zhu
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注: Accepted at the 40th Conference on Neural Information Processing Systems (NeurIPS 2026). 30 pages, 11 figures, 19 tables
Abstract:We introduce Neural Harmonic Measure Operator (NHMO), a neural solver for elliptic PDE problems on variable-shape domains. The harmonic measure of a domain is the boundary probability distribution that, integrated against any boundary data, returns the Dirichlet Laplace solution. It depends only on the geometry, not on the boundary data. NHMO parameterizes the density of this measure as a transformer-based boundary kernel supervised by Walk-on-Spheres exit samples, so one trained kernel handles different boundary values on a shape with no retraining. We extend it to Poisson via a classical decomposition, with an auxiliary network amortizing the source-induced correction and avoiding the singular volume quadrature that breaks direct evaluation. At inference, new boundary values and new sources both yield PDE solutions by re-integration against the fitted kernel and lift, with no retraining. NHMO improves over four prior baselines on the MCB-B 3D variable-shape Poisson benchmark across all five categories, and is competitive with major neural-operator baselines on a controlled 2D testbed.
[LG-3] ScAn-Bench: Evaluating Scaling Analysis Methodology
链接: https://arxiv.org/abs/2609.35707
作者: Artin Sermaxhaj,Nastaran Alipour,Donat Sinani,Johannes Hog,Neeratyoy Mallik,Jenia Jitsev,Danny Stoll
类目: Machine Learning (cs.LG)
*备注:
Abstract:Recent progress in machine learning is driven by large-scale foundation models, where scaling laws and finding optimal scaling prescriptions for architecture, data, and hyperparameters are key in advancing the state-of-the-art. Therefore, it is surprising that no systematic study evaluates the methodology to obtain scaling laws and prescriptions across different model types. To shed light on this crucial blind spot and facilitate future research, we introduce the surrogate benchmarks ScAn-Bench-LLM and ScAn-Bench-VLM based on 4524 and 8024 checkpoints of language and vision-language model pipelines. On our benchmarks, we perform the first systematic evaluation of both data acquisition and extrapolation methodology for scaling analysis across different data modalities.
[LG-4] MeqMuon: Matrix-Equilibrating Muon for LLM Pretraining
链接: https://arxiv.org/abs/2609.35701
作者: Chang-Wei Shi,Xu Wang,Wu-Jun Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:The success of large language models (LLMs) has been accompanied by continued growth in model size and pretraining costs. Muon offers high accuracy and training efficiency in LLM pretraining. Recent work introduces row-wise normalization into Muon to balance update magnitudes and improve pretraining performance. However, row-wise normalization alone cannot accommodate different imbalance patterns in update matrices. In this paper, we propose an improved Muon optimizer, called \underlinematrix-\underlineequilibrating Muon~(MeqMuon), for LLM pretraining. MeqMuon balances both row and column magnitudes through normalization that can be automatically tailored to different imbalance patterns without manual intervention. Moreover, MeqMuon eliminates the need to store AdamW’s second-moment estimates, reducing optimizer-state memory usage. Empirical results demonstrate that MeqMuon achieves better convergence performance than AdamW, Muon, and other baselines in LLM pretraining.
[LG-5] Provable Benefits of Regularization: Fast Rates for Adversarial Imitation Learning
链接: https://arxiv.org/abs/2609.35698
作者: Hanbin Zhou,Shangzhe Li,Alexander Braverman,Weitong Zhang
类目: Machine Learning (cs.LG)
*备注: 33 pages, 1 table
Abstract:We study adversarial imitation learning (AIL), in which an agent learns to imitate expert demonstrations by optimizing a policy against an adversarial reward that distinguishes expert and learner behavior. Historically, reward regularization and entropy-based policy regularization are key components of empirically successful methods such as GAIL and LS-IQ, yet their finite-sample benefits remain underexplored. We establish fast rates for jointly regularized AIL in finite-horizon Markov decision processes with general function approximation. Our model-free algorithm, Dually Regularized AIL, combines KL policy regularization with a quadratic reward penalty weighted by expert and learner occupancies. With K online episodes and N expert trajectories, we prove a \widetildeO\left(\frac1K+\frac1N\right) bound on the regularized imitation gap for fixed regularization parameters. Our analysis combines an online mirror descent construction for general convex reward classes to control estimation error from finite expert data and stochastic learner feedback, with a sharp analysis of optimistic KL-regularized policy learning. To the best of our knowledge, Dually Regularized AIL is the first algorithm to simultaneously achieve \widetildeO\left(\frac1\epsilon\right) sample complexity in both expert demonstrations and online interactions for this regularized AIL objective, even with stochastic experts. These results provide a rigorous characterization of the complementary statistical benefits of reward and policy regularization in AIL.
[LG-6] he Hidden Perception Constraint in Task-Aware Compression
链接: https://arxiv.org/abs/2609.35684
作者: Sahan Liyanaarachchi,Semih Akkoc,Sennur Ulukus,Aylin Yener
类目: Information Theory (cs.IT); Machine Learning (cs.LG)
*备注:
Abstract:With the recent advancements of neural compressors, explicitly incorporating perception constraints into the design of compression schemes has gained significant attention. Traditionally, these perception constraints ensure that the distribution of the reconstruction does not significantly deviate from the distribution of the source, thus attesting to the perceptual quality of the reconstruction. In this work, we uncover several perception constraints that are naturally present in task-aware compression. In particular, we consider a problem where the primary task is reconstruction and the secondary task is classification (i.e., a statistical test). We study this problem at varying levels of domain information available to us and discuss how to utilize the naturally emerging perception constraints to design rate-minimal compression schemes that also maximize the utility of our secondary task. We show that in this setting, if the decision boundaries of the classifier are ill-defined (mismatch) for our source distribution, then matching onto a target distribution enhances our classification accuracy.
[LG-7] ransferable Mass Spectrum Prediction via Reference-Guided Test-time Specialization
链接: https://arxiv.org/abs/2609.35649
作者: Yunhua Zhong,Runting Li,Yifan Li,Pan Liu,Zhiwen Yang,Zikun Wang,Yixuan Tang,Jun Xia
类目: Machine Learning (cs.LG)
*备注:
Abstract:Tandem mass spectrum prediction supports compound identification across metabolomics, natural-product discovery, and environmental analysis. However, pretrained predictors often degrade under shifts in chemical space and acquisition conditions, while retraining domain-specific models from scratch is costly. We introduce SPARC, a retrieval-guided test-time specialization framework that adapts a pretrained predictor using a spectral reference library without accessing test-query spectra. For each target query, SPARC retrieves chemically related reference spectra to recalibrate fragment intensities within the learned fragmentation space. During Transfer, SPARC combines reference-guided spectral adaptation with reliability-aware consistency, using reconstruction behavior on retrieved spectra to selectively preserve trustworthy predictions during continual specialization. Across MassSpecGym, NPLIB1 and application-specific GNPS libraries, SPARC improves spectral prediction under multiple transfer settings. These results establish retrieval-guided test-time specialization as a practical strategy for extending pretrained MS/MS predictors to specific chemical and acquisition domains, with continual test-time training providing further refinement during deployment.
[LG-8] Bounding Retraining Equivalence and the Deletion Floor in Materials Machine Unlearning
链接: https://arxiv.org/abs/2609.35635
作者: Can Polat,Mustafa Kurban,Erchin Serpedin,Hasan Kurban
类目: Machine Learning (cs.LG); Materials Science (cond-mat.mtrl-sci); Quantum Physics (quant-ph)
*备注:
Abstract:In materials machine learning, closely related retained structures can sustain accurate property predictions even after removing a specific record, rendering post-deletion prediction error an ambiguous metric for machine unlearning. To resolve this ambiguity, we define the deletion floor as the expected target loss under a specified retraining procedure at the deleted request. Standard indistinguishability constraints yield a sharp interval bounding an update’s target loss around this baseline reference. Theoretically, a conditional neighbor bound links a low deletion floor directly to retained fit, prediction regularity, and local label agreement, while an exact ridge identity isolates residual fit from the prediction change induced by record deletion. Empirically, controlled redundancy sweeps show an \approx 8\times drop in median normalized retraining loss when one retained relative remains after deletion. Across two distinct fitting regimes in a paired Materials Project study, the lower-floor regime also exhibits a larger prediction change on more than 50% of the shared requests. Systematic comparisons against approximate updates and the original model decouple deliberate target suppression from preserved overall model utility. Consequently, request-level unlearning evaluations should report reference loss, prediction change, and retained utility together, interpreting post-deletion accuracy against what retraining itself leaves behind.
[LG-9] Arbitrary-Accuracy Neural Approximation with Optimal Neuron Count and Near-Optimal Bit Complexity
链接: https://arxiv.org/abs/2609.35628
作者: Zilan Cheng,Li-Lian Wang,Zhongjian Wang
类目: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE); Machine Learning (stat.ML)
*备注:
Abstract:We study the minimum number of hidden neurons required for arbitrary-accuracy approximation of multivariate Hölder-continuous functions on [0,1]^d and the associated encoding complexity. For d\geq 2 , we construct a fixed, explicitly defined activation function for which a closed-form network with two hidden layers of widths d and 1 achieves arbitrary accuracy in the uniform norm. We prove that d+1 is the exact minimum total number of hidden neurons among standard feedforward networks with locally integrable activations and affine outputs. We further give a simpler construction using a single elementary activation that combines the floor and exponential functions. This construction requires three hidden layers of widths d , 1 , and 2 , only two neurons above the minimum. If a skip connection is allowed, widths d , 1 , and 1 suffice. These constructions use explicit grid addressing and integer encoding of quantized function values. For a bounded \alpha -Hölder class, they require O(\varepsilon^-d/\alpha\log(1/\varepsilon)) bits, matching the metric-entropy lower bound up to a logarithmic factor.
[LG-10] Cartridges: KV Cache Compression without Off-Context Derailment
链接: https://arxiv.org/abs/2609.35621
作者: Sonia Laguna,Joao Monteiro,Marco Cuturi,Pierre Ablin,Eleonora Gualdoni
类目: Machine Learning (cs.LG)
*备注:
Abstract:Serving long documents to a Large Language Model (LLM) repeatedly is expensive: computations grow with context length, and the memory footprint of the key-value (KV) cache balloons. Compressed KV (CKV) representations aim to mimic the cache of a document and are typically computed once and for all, ahead of inference time. Methods to obtain CKVs range from drop mechanisms that reduce their number of columns, to learned approaches. Among the latter, Cartridges have emerged as a leading compression method, learning compact KV representations through distillation on relevant Q/A pairs. While existing evaluations focus primarily on whether Cartridges and other CKVs yield approximately similar responses to document-related, on-context queries, we investigate the crucial deployment question of whether they can handle off-context queries, something the native KV representation is particularly good at, thanks to the mechanics of attention. We observe a fundamental trade-off: while Cartridges perform better for on-context queries, heuristic-variants preserve better the original LLM’s ability to operate off-context. We measure this through their capability to avoid context contamination in their response, retain general knowledge, and follow instructions. We propose Cartridges++, simple modifications to cartridges that retain off-context abilities at small or negligible cost. The router variant decides at inference time whether the query should use the learned long-context memory, while the data-mixing variant allocates a small fraction of training Q/As to queries outside the reference long document. Our study shows that assessing CKVs on document utility alone can mask substantial degradation in broader model capabilities, yet those issues can be fixed with benign changes to CKV inference or training.
[LG-11] Attention Graphons: A Graph Limit Perspective on Graph Transformers
链接: https://arxiv.org/abs/2609.35620
作者: Caio F. Deberaldini Netto,Moshe Eliasof,Luana Ruiz
类目: Machine Learning (cs.LG)
*备注: 42 pages, 32 figures
Abstract:Graph Transformers produce, for each attention head, a dense n\times n matrix of learned pairwise interactions. We ask a fundamental question: do these attention-induced graphs converge to a stable limit object as n grows, or does the learned interaction pattern remain unstructured and size-dependent? We answer this using dense graph limit theory, treating each attention matrix as a finite sample from an underlying kernel—an \emphattention graphon—and studying concentration around this limit under the cut-distance. We derive a worst-case variance bound requiring no assumptions on the graphon, and a sharper regularity-aware bound based on nonparametric estimation theory. To operationalize the theory, we propose a canonicalize-then-block-average pipeline for estimating dataset-level attention graphons, and a variance-based diagnostic for testing whether attention admits a stable continuum description. Experiments across multiple graph benchmarks show that learned attention stabilizes to dataset-specific graphon structure on several datasets; that empirical cut-distance and cut-norm variance decreases with n consistent with our bounds; and that attention graphons transfer to larger graph sizes with error decreasing in n .
[LG-12] EvE: An Alternate Optimizer to Adam
链接: https://arxiv.org/abs/2609.35614
作者: Shashank Raj,Kalyanmoy Deb
类目: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
*备注:
Abstract:Adam and its variants dominate neural network training, but a single run only reveals whether a configuration works well after most of its budget is spent, a poor fit for hyperparameter or architecture search, where configurations must be ranked cheaply and pruned early. We introduce EvE (Evolutionary Explorer), a steady-state, population-of-four differential evolution (DE) optimizer with a targeted Adam fallback: each iteration proposes one candidate via DE, running a short burst of gradient descent only if the DE step fails to improve on the incumbent. Selection is greedy, so on a deterministic objective the best-so-far value is provably monotone non-increasing, and since gradients are used only as a targeted rescue, per-iteration cost stays within a constant factor of a single Adam step regardless of dimension. Under a fixed, evaluation-cost-matched budget, EvE wins or ties Adam on 76% of 70 (problem, dimension) cells across seven scalable benchmarks up to one million variables. On three real neural-network tasks (an MLP on MNIST, and LoRA fine-tuning of a 1.5B-parameter language model on two datasets) EvE finishes the same charged budget 1.7-3.9x faster, at a modest cost in final quality (about one accuracy point on MNIST, 9-11% higher relative test loss on the two fine-tuning tasks; on GSM8K, Adam is about 5 accuracy points more accurate, and fine-tuning lowers accuracy below the base model for both). Inside successive halving on UCI Adult, EvE completes hyperparameter and architecture searches 3.1-3.5x faster, ranking configurations about as consistently with Adam as Adam does with itself across seeds (Kendall’s tau 0.66-0.69). EvE is not a total replacement for Adam as a final-stage trainer, but a fast, gradient-aware proxy for the search-heavy, budget-constrained regime one level up.
[LG-13] Control-Geometry Straightening for Sampling-Based Latent Planning
链接: https://arxiv.org/abs/2609.35603
作者: Ziang Fu,Ning Ning
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Joint-embedding predictive architectures enable planning with latent world models, but accurate transition prediction alone does not ensure that the planning objective is easy to optimize. We introduce Control-Geometry Straightening (CGS), a single auxiliary loss that learns planner-friendly representations by directly straightening control geometry for sampling-efficient planning. CGS matches pairwise cosine similarities among actions to those among corresponding latent differences only using local transitions from pixel-action pairs. The loss can be applied across world-model architectures using end-to-end learned or pretrained representations. Under linear-dynamics, our theoretical analysis connects this objective to temporal straightening and more balanced terminal-cost curvature across the full planning horizon, yielding finite-budget guarantees for MPPI, local contraction results for CEM, and convergence bounds for gradient descent. Across four control environments and multiple planners, CGS improves planning with fewer sampled candidates and refinement steps, achieving success-rate gains up to 20 and 12.6 percentage points over LeWorldModel (LeWM) and its temporal-straightening variant (LeWM+TS), respectively, with sampling-based planners using 128 candidates per update. Probes, comparisons with DINO-WM architecture, and planner-side ablations clarify how latent motion organization, state dependence, and dynamical context shape planning behavior. Straightening control geometry thus makes good action sequences easier to find under limited planning budgets.
[LG-14] Hardware-Aware Features for CUTLASS Kernel Selection
链接: https://arxiv.org/abs/2609.35587
作者: Shriram Chandran,Dominic Rinderer,Yakup Budanaz,Alexandru Calotoiu,Marcin Copik,Torsten Hoefler
类目: Machine Learning (cs.LG); Performance (cs.PF)
*备注: 20 pages, 19 figures
Abstract:GPU libraries such as CUTLASS expose tens of thousands of semantically equivalent kernels for a single operation, making exhaustive autotuning expensive and execution-free selection difficult. Existing analytical selectors require hand-designed performance rules, while learned selectors operate on raw configuration parameters and must infer hardware consequences from data. We introduce a hardware-aware representation for CUTLASS kernel selection that augments candidate configurations with statically computable estimates of induced hardware behavior. We construct a dataset of 4.9 million CUTLASS kernels and train gradient-boosted and neural learning-to-rank models to rank candidates within each problem. On held-out exhaustive evaluation problems, hardware-aware representations reduce selection regret by up to 40% relative to structural baselines and 64.2% relative to NVIDIA’s matrix-multiply heuristics. We further evaluate data-efficient cross-precision and epilogue-fusion transfer within CUTLASS GEMM, showing that explicitly representing candidate-induced hardware behavior provides a useful inductive bias for learned kernel selection.
[LG-15] Output-aware Residual Stream Pruning for Large Language Models
链接: https://arxiv.org/abs/2609.35579
作者: Chayne Thrash,Kevin Chen,Soheil Kolouri
类目: Machine Learning (cs.LG)
*备注:
Abstract:Residual stream pruning methods reduce inference cost by shrinking the model’s hidden dimension, but existing approaches typically choose these dimensions by minimizing activation reconstruction error. This criterion implicitly treats all perturbation directions as equally important, ignoring the sensitivity of downstream layers. We introduce a sensitivity-aware approach to residual-stream pruning that directly accounts for this direction-dependent sensitivity. Using a second-order approximation to the output KL divergence, we characterize the effect of a residual-stream perturbation through both its activation covariance and the local sensitivity of the model output. The resulting subspace selection objective couples these two quantities, but is difficult to optimize directly. We derive a tractable spectral upper bound that reduces subspace selection to an eigendecomposition of a sensitivity-weighted covariance matrix, retaining the efficiency and structural simplicity of rotation-based pruning methods. Across several instruction-tuned language model families, our method consistently reduces calibration KL divergence relative to activation-only pruning and improves perplexity and downstream task performance over a range of compression levels. Our results show that preserving activation energy alone is insufficient for residual-stream pruning, and that explicitly accounting for how perturbations propagate to the model output provides a more effective criterion for selecting dimensions to remove.
[LG-16] Beyond Energy: When Sustainability Dimensions Reshape LLM Serving Decisions
链接: https://arxiv.org/abs/2609.35569
作者: Tianyao Shi,Xipeng Shen,Yi Ding
类目: Computers and Society (cs.CY); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG); Performance (cs.PF)
*备注: 41 pages, 30 figures, 13 tables
Abstract:Large language model (LLM) serving has environmental impacts across energy consumption, carbon emission, water consumption, and biodiversity loss. Yet these dimensions are largely evaluated in isolation, leaving it unclear when and how they lead to different optimization decisions. We present PRISM, a unified framework for characterizing and optimizing LLM serving across energy, carbon, water, and biodiversity impacts. Our analysis reveals a fundamental distinction: computing configurations determine energy consumption, whereas where and when LLM serving is deployed determine its carbon, water, and biodiversity impacts. Under a fixed deployment choice and operational-only accounting, all dimensions preserve the same energy-based configuration ranking. Deployment rankings can diverge across dimensions, while embodied impacts can break configuration invariance when they exceed a lifecycle crossover boundary. PRISM identifies these conditions, quantifies cross-dimensional regrets, and balances the four dimensions. In regional-routing experiments, PRISM reduces median worst-case regret by 50.2% relative to the strongest baseline.
[LG-17] From Experience to Expertise: Adoption-Aware Memory Learning for Data-Scarce NPU Kernel Synthesis
链接: https://arxiv.org/abs/2609.35568
作者: Longxiao Fan,Tao Zhang,Han Yan,Jiajun Li,Mingcong Song,Guoping Long,Hongjie Si,Weiwei Sun
类目: Machine Learning (cs.LG)
*备注: 30 pages
Abstract:High-performance kernels underpin efficient accelerator execution but require expert tuning and lengthy manual optimization cycles. LLM coding agents promise automation, yet their CUDA knowledge transfers poorly to data-scarce domain-specific architectures (DSAs) such as NPUs, whose execution models and memory hierarchies differ substantially from those of GPUs. To address this transfer gap, post-training methods adapt LLMs to NPU programming but depend on scarce expert data and substantial training compute. Memory-learning agents instead adapt through external memory, but their uniform credit assignment gives adopted and unused experiences the same reward target, potentially biasing subsequent retrieval rankings. Moreover, when learned values guide only retrieval, high-value experiences that generalize across operators must be retrieved repeatedly rather than retained in context, thereby increasing retrieval overhead and weakening cross-task guidance. We therefore present SAGE, a persistent self-improving agent for NPU kernel synthesis. Adoption-Traced Utility estimation (ATU) combines explicit adoption records with kernel evaluation outcomes for adoption-aware credit assignment. Utility-Gated Consolidation (UGC) uses positive utility and repeated adoption across operators to select and abstract reusable rules into a bounded resident context. On NPUKernelBench, SAGE achieves a 95.5% execution rate versus 84.1% for the strongest controlled baseline, with 86.9% of solved operators outperforming torch_npu. With GLM-5.3, SAGE achieves a 43.99x speedup over the torch_npu reference on sparse flash attention. These results show that adoption-aware credit assignment and selective consolidation enable agents to accumulate and reuse hardware-specific knowledge across tasks.
[LG-18] Simplex Diffusion Models
链接: https://arxiv.org/abs/2609.35553
作者: Justin Deschenaux,Alexandre Galashov,Andrew Campbell,Li Kevin Wenliang,James Thornton,Arnaud Doucet,Valentin De Bortoli
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Diffusion models have revolutionized generative modeling for continuous data through the gradual refinement of a belief state. This iterative refinement has not yet carried over to discrete diffusion models, which discard uncertainty at intermediate steps through categorical sampling (information collapse). We propose Simplex Diffusion Models (SDMs), a framework that lifts the diffusion process to the probability simplex to represent beliefs over categories. SDMs admit probability paths with closed-form reverse transitions and can be trained with a simple cross-entropy loss. Contrary to earlier proposals such as Dirichlet Flow Matching which requires integrating an ordinary differential equation, we introduce a DDIM-like sampler with a tunable level of stochasticity. Because SDMs operate on samples on the simplex, they can carry uncertainty across denoising steps, which mitigates information collapse. On OpenWebText, SDMs are competitive with strong Discrete Diffusion baselines, achieving 17.0 GenPPL at 5.46 unigram entropy in 64 sampling steps, close to real validation data. Even without Self-Conditioning (SC), SDMs outperform masked and uniform diffusion (with SC or predictor-corrector sampling) on code generation (TinyGSM, T=0.1 ; 49.0% vs. 45.8% ). Distilled down to 8 steps, SDMs solve 32.1% of GSM8K problems, more than distilled Discrete Diffusion models with 128 steps ( 21.4% ).
[LG-19] Learning the Robustness Mechanism with Bilevel Optimization
链接: https://arxiv.org/abs/2609.35541
作者: Yiyang Shen,Qihang Lin,Weiran Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:We propose a distributionally robust learning framework where parameters defining the robustness mechanism are learned from held-out data instead of extensively tuned. Using bilevel optimization with both upper and lower level minimax problems, we create two instances of our framework to tackle setups with and without group labels in the training set. Theoretically, we provide sample complexity analysis for our robustness mechanism learning paradigm, showing that it achieves generalization guarantees comparable to exhaustive grid search while being more computationally efficient. Empirically, we evaluate our framework under a challenging setup when both intra-group and inter-group test distribution shifts occur at the same time, thereby demonstrating the efficacy and scalability of our method.
[LG-20] Optimal Networks for Agent ic Information Aggregation
链接: https://arxiv.org/abs/2609.35537
作者: MohammadHossein Bateni,Zahra Hadizadeh,MohammadTaghi Hajiaghayi,Mahdi JafariRaviz,Shayan Taherijam
类目: Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT); Theoretical Economics (econ.TH)
*备注:
Abstract:We study information aggregation in the networked learning model introduced by Kearns, Roth, and Ryu (SODA 2026). There is a fixed distribution over d features and a common label. Agents learn in topological order on a directed acyclic graph. Each observes a subset of the features and its parents’ predictions, fits a linear predictor to minimize mean squared error, and passes only its prediction forward. The global predictor is the best linear predictor using all features. Kearns, Roth, and Ryu show that the output agent’s error approaches the global predictor’s error along sufficiently deep paths with suitable feature coverage, while insufficient depth can prevent aggregation even in large networks. In contrast to their main focus on a given graph and feature allocation, we consider the limits of the model under two settings. In the adaptive designer setting, a designer chooses the graph, feature allocation, and output agent knowing the distribution. In the oblivious designer setting, the designer fixes all three before an adversary chooses the distribution. Each agent observes one feature and receives predictions from a limited number of parents. We call the aggregation exact when the output agent matches the global predictor exactly. For d\ge3 , we show that no finite depth guarantees exact aggregation for every distribution with one parent per agent, even when the designer knows the distribution. In contrast, two parents per agent suffice for exact aggregation even in the oblivious designer setting. A fixed graph, feature allocation, and output agent achieve this for every distribution at depth O(d\log d) . Knowing the distribution reduces the depth to O(d) . Both constructions use O(d^2) agents, with a very large constant for two parents. We show the bounds on the depth and number of agents are all optimal up to constant factors. Subjects: Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT); Theoretical Economics (econ.TH) Cite as: arXiv:2609.35537 [cs.LG] (or arXiv:2609.35537v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.35537 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Mahdi JafariRaviz [view email] [v1] Mon, 28 Sep 2026 16:19:38 UTC (54 KB)
[LG-21] GeoGAE: Scalable Graph-Level Autoencoding via Hyperball Cloud Representations ICLR2027
链接: https://arxiv.org/abs/2609.35527
作者: Radosław Nowak,Anna Bielawska,Bogusz Stefańczyk,Maciej Sanocki,Paweł Wawrzyński
类目: Machine Learning (cs.LG)
*备注: Submitted for ICLR 2027
Abstract:Embedding structured objects into Euclidean spaces has enabled a wide range of successful machine learning applications. Such objects include words, documents, image patches, time series, and graph nodes. In contrast, embedding entire graphs remains a challenging problem. Existing methods either sustain the original order of the graph nodes or match the output nodes to the input ones, both of which create scalability issues. In this work, we propose a graph representation as a cloud of hyperballs, which allows us to define a specific, typically unique, node ordering. Based on this representation, we propose GeoGAE, an autoencoder, in which the Transformer encoder translates a hyperball cloud into a graph-level embedding, and the Transformer decoder translates the graph-level embedding back into the graph. This formulation enables the model to capture both the global graph structure and local relational patterns. We evaluate our method on multiple graph datasets, spanning various domains. The results demonstrate effectiveness of our method in encoding and reconstructing graphs from their embeddings.
[LG-22] Deep Epistemic Value Functions for Optimistic Exploration
链接: https://arxiv.org/abs/2609.35525
作者: Leander Diaz-Bone,Marco Bagatella,Jonas Hübotter,Andreas Krause
类目: Machine Learning (cs.LG)
*备注:
Abstract:Principled exploration in reinforcement learning requires an agent to quantify its epistemic uncertainty and act to resolve it. Uncertainty over the value function provides a natural signal for exploration, yet existing deep approximations remain brittle and perform inconsistently. The central challenge is therefore to scale these ideas robustly. We conduct a systematic empirical study of how epistemic uncertainty is represented, propagated, and optimized in deep epistemic value functions, and uncover distinct failure modes along each of these axes. These findings motivate DEVOTE, a model-free reinforcement learning algorithm that controls how uncertainty generalizes beyond observed data, stabilizes its temporal propagation, and preserves adaptation to the resulting non-stationary exploration objective. Across reward-free exploration and challenging continuous-control tasks, DEVOTE reaches novel states more effectively and achieves higher task return than strong model-free and model-based exploration baselines. These results provide evidence that deep epistemic value functions are a promising path toward scalable, principled exploration.
[LG-23] Reward-Aligned Reweighting for On-Policy Distillation
链接: https://arxiv.org/abs/2609.35517
作者: Haofeng Xu,Junwei Su,Lansong Diao,Wenchao Zhou,Chuan Wu
类目: Machine Learning (cs.LG)
*备注:
Abstract:On-policy distillation (OPD) trains a student language model with dense feedback from a stronger teacher on student-generated trajectories. Yet standard OPD weights token-level distillation terms uniformly, implicitly treating local teacher preference as a proxy for correction utility. A decision’s task value, however, depends on how the student completes the subsequent reasoning. This mismatch can cause imitation to suppress viable student strategies or reinforce paths the student cannot reliably execute. Verified trajectory outcomes provide complementary evidence about continuation quality, but do not directly identify the utility of individual decisions. We introduce Reward-Aligned Reweighting for On-Policy Distillation (R ^2 -OPD), which uses outcome agreement and the magnitude of teacher–student disagreement to continuously reallocate teacher supervision. It gives reward-aligned corrections greater relative influence while retaining dense feedback, moving beyond uniform imitation and hard filtering. Our analysis formalizes the mismatch between local teacher preference and student continuation value and establishes sufficient conditions for reallocation to improve first-order task progress over uniform OPD. Across seven mathematical reasoning benchmarks, R ^2 -OPD achieves the highest average accuracy among the compared training methods in both cross-size and same-size distillation. It outperforms standard OPD on all seven benchmarks, with average gains of 3.5 and 2.4 percentage points for 1.7B and 4B students, respectively. An extension to code generation yields an average gain of 1.6 percentage points over standard OPD. These results highlight outcome-guided supervision allocation as an effective way to translate dense teacher feedback into stronger student performance across model scales and task domains.
[LG-24] One Proposal for Every Margin: Zero-Shot Amortized Sequential Importance Sampling for Binary Matrices
链接: https://arxiv.org/abs/2609.35514
作者: Ruishuo Chen,Weijia Li,Xun Wang,Yu Chen,Leheng Cai,Longbo Huang
类目: Machine Learning (cs.LG); Computation (stat.CO)
*备注:
Abstract:In ecology, psychometrics, and the analysis of social and financial networks, binary matrices are often analyzed conditional on their observed row and column sums, which restricts the problem to a finite sample space of matrices with the same margins. Two fundamental problems are to count this space and to sample uniformly from it. Sequential importance sampling (SIS) addresses both with independent weighted samples and an unbiased count estimator, but its efficiency depends critically on the proposal distribution. Existing proposals are analytically designed, and their accuracy can vary substantially with the margins. We show that the ideal SIS proposal, under which every weight equals the count and the variance vanishes, is exactly the policy of a generative flow network (GFlowNet) with unit reward on every matrix that has the given margins. We therefore propose MarginFlow, a framework that turns the design of the proposal into a learning problem and amortizes it across margins by exploiting their self-similarity. Every partial matrix is itself an instance with reduced margins, so one set transformer that reads the remaining margins serves every margin. We train MarginFlow on a pool of 1904 margins and evaluate it zero-shot on 1190 held-out margins, synthetic and real, from 3\times3 to 870\times6 . On 1187 of the 1190 margins it matches or beats the best of 31 analytically designed configurations, chosen post hoc for each margin, and its median effective sample fraction is 99.8%. On the 56 margins where that best loses more than one nat of effective sample size, MarginFlow wins every one and raises the median effective sample fraction from 10.3% to 94.1%.
[LG-25] An RL View of OPD: Least Square Policy Distillation for Sample-Efficient LLM Reasoning
链接: https://arxiv.org/abs/2609.35505
作者: Shangzhe Li,Yuxiao Yang,Tianrun Yu,Kaixiang Zhao,Xiaoyun Wang,Taylor W. Killian,Weitong Zhang
类目: Machine Learning (cs.LG)
*备注: 29 pages, 3 figures, 5 tables, code available at this https URL
Abstract:We study on-policy distillation (OPD) through the lens of reinforcement learning, establishing a connection between the reverse-KL objective in OPD and KL-regularized policy optimization. Building on this connection, we introduce Least-Square Policy Distillation (LSPD), an RL-inspired framework that brings optimistic exploration and off-policy data reuse from value-based RL into policy distillation. LSPD preserves policy diversity through exploration while improving rollout efficiency by repeatedly learning from previously collected trajectories. Our theoretical analysis connects LSPD to optimistic value-based learning and shows that its idealized formulation achieves a sharp \tilde\mathcal O(\log K) regret bound under online exploration. Empirically, LSPD consistently outperforms existing distillation baselines across six mathematical reasoning benchmarks and diverse teacher-student settings, with average gains of +1.59 points in Avg@16. Remarkably, through Pass@k evaluations up to k=64, we found that LSPD better preserves policy diversity by achieving stronger performance as k grows. Its fully off-policy variant achieves comparable performance to vanilla OPD using only the first 25% of rollout batches. Together, these results provide an RL perspective on OPD that offers both a principled interpretation and a practical route toward more effective and rollout-efficient language model distillation.
[LG-26] Structured Latent Modeling for Supervised Multimodal Information Decomposition
链接: https://arxiv.org/abs/2609.35502
作者: Wanting Huang,Sanvesh Srivastava,Weiran Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Multimodal prediction relies on diverse forms of evidence: information repeated across modalities, cues specific to a single source, and complex cross-modal dependencies that emerge only when inputs are considered together. While recent methods promote richer interactions, they lack a principled way to isolate these target-relative contributions within learned continuous representations. We introduce a framework that applies contrastive or masked objectives at intermediate layers, coupled with source-wise invertible normalizing flows and a supervised, low-rank latent variable model. This architecture explicitly factorizes the joint distribution into shared task-relevant variation, modality-specific predictive variation, and task-irrelevant dependence. Drawing connections to prior multimodal learning assumptions, our approach evaluates how modalities independently and jointly contribute to the target. Ultimately, this framework unites intermediate representation learning with structured likelihood-based guidance, offering a practical latent-variable lens for characterizing continuous multimodal interactions. Empirically, we demonstrate the effectiveness of our approach across diverse multimodal benchmarks, showing robust improvements in predictive performance.
[LG-27] Physics-Guided Conditional Diffusion Model for Rare Event Synthesis and Diagnosis for the Water-Gas Shift Reaction
链接: https://arxiv.org/abs/2609.35499
作者: Md Abrar Rafid Siddique,Bibek Aryal,Qiugang Lu
类目: Machine Learning (cs.LG)
*备注: 29 pages, 18 figures
Abstract:As the world moves towards sustainable energy sources, hydrogen (H2) can be treated as an eco-friendly alternative to fossil fuels due to its high energy density and zero carbon emissions. The water-gas shift (WGS) reaction is a widely used industrial process for hydrogen production by converting carbon monoxide and steam into hydrogen and carbon dioxide. However, occurrences like severe fouling, catalyst deterioration, and thermal runaway can hamper the reaction kinetics/process safety and decrease the yield of H2. These incidents are rare, and gathering process data under such abnormal conditions is challenging. In this work, we propose a physics-guided conditional diffusion model to generate realistic rare-event trajectories for the WGS reaction. The proposed model integrates a conditional denoising diffusion probabilistic model (CDDPM) with governing laws of the reaction to generate physically consistent process trajectories. The conditioning features allow the model to produce high-quality synthetic profiles for rare-event domains that are typically beyond the training regimes. The generated rare-event trajectories then augment the raw dataset for a balanced distribution between normal and abnormal conditions. We further propose a hazard score to assess the risk severity of the operating condition based on the operating trajectory. Deep learning models are trained with the augmented dataset to diagnose the health status of the reaction. Simulation results show that the proposed physics-guided diffusion model outperforms data-driven models in terms of the quality of synthetic data and diagnosis performance for rare events.
[LG-28] Universal Approximation of Measure-to-Measure Operators by Pushforwards
链接: https://arxiv.org/abs/2609.35483
作者: Takashi Furuya,Nicholas H. Nelsen,Frank Cole
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA); Machine Learning (stat.ML)
*备注: 31 pages (9 main text, 19 appendix, and 3 references pages)
Abstract:Many learning tasks map an input distribution to an output distribution. A natural way to model such an operator is to transform each input sample using a continuous function that may depend on the entire input distribution, and then take the distribution of the transformed samples. This defines a measure-dependent pushforward model and includes measure-theoretic formulations of transformers. We ask when such models can approximate arbitrary continuous operators between spaces of probability measures. We first show that universal approximation fails when atomic inputs are allowed: some continuous measure-to-measure operators that split or redistribute atomic mass cannot be approximated arbitrarily well by deterministic pushforward models. We then introduce the uniform level set condition, which requires a continuous measure-dependent scalarization whose shrinking level set neighborhoods carry uniformly vanishing mass over the input family. This condition is satisfied, in particular, by compact families of absolutely continuous measures. On every compact family satisfying this condition, we prove that any continuous measure-to-measure operator with outputs of finite p -th moment can be uniformly approximated, in the p -Wasserstein distance, by continuous measure-dependent pushforwards. Combining our theorem with existing approximation results for measure-dependent in-context maps yields universal approximation by measure-theoretic transformers. We also extend the framework to continuously-varying source measures, yielding a corresponding universality result for a class of pushforward models that are closely aligned with cross-attention architectures.
[LG-29] opoEP: Topology-Aware Load Balancing for Expert-Parallel MoE Training
链接: https://arxiv.org/abs/2609.35481
作者: Jiacheng Zhu,Xie Zhao,Gongming Zhao,Hongli Xu,Yao Fei,Jin Fang
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注:
Abstract:Dynamic routing creates severe load imbalance in large-scale expert-parallel Mixture-of-Experts (MoE) training, turning GPUs that host hot experts into stragglers. As each MoE layer waits for its slowest rank, these stragglers prolong the expert-parallel stage and reduce overall training efficiency. Existing expert-parallelism load-balancing (EPLB) systems commonly compute load-balancing plans on the CPU, incurring device–host data transfers and cross-rank synchronization that make scheduling at every layer and microbatch expensive. Their planning formulations also overlook the hierarchical communication costs of modern scale-up and scale-out GPU clusters. We present \textitTopoEP, a GPU-native, topology-aware load-balancing system for large-scale MoE training. At each MoE layer and training microbatch, \textitTopoEP converts the current routing result into hot-expert replication and token-rerouting decisions and executes the resulting plan without data-dependent host synchronization, reducing critical-path overhead. To generate these decisions, \textitTopoEP uses a deterministic GPU solver that performs inter-node placement followed by intra-node refinement, allowing all ranks to independently produce bitwise-identical plans. On a 32-GPU NVIDIA H800 cluster, integrating \textitTopoEP with Megatron-LM improves end-to-end training throughput by 6.2%–11.4% across three representative MoE models. Subjects: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG) Cite as: arXiv:2609.35481 [cs.DC] (or arXiv:2609.35481v1 [cs.DC] for this version) https://doi.org/10.48550/arXiv.2609.35481 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-30] An analysis of Mirror-Descent Soft Actor-Critic
链接: https://arxiv.org/abs/2609.35466
作者: Denis Zorba,Michal Valko
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: 36 pages, 2 figures
Abstract:Soft Actor-Critic (SAC) is widely used for entropy-regularised reinforcement learning with continuous action spaces, and practical implementations perform only a few actor steps towards an evolving target. In this work, we prove convergence guarantees when the target policy arises from policy mirror descent and compare it with the classical Gibbs target. We derive sufficient conditions for the strong convexity and smoothness of the actor objective, characterised by the curvature of the Q -function estimate through the Legendre differential operator, and establish an \mathcalO!\left(N^-\frac15\right) best-iterate finite-time convergence rate up to actor and critic approximation errors. Moreover, the mirror-descent step size \lambda directly controls the target drift and hence actor tracking error, whereas the analogous Gibbs bound contains a non-vanishing tracking term.
[LG-31] ra: Serving Leech-Lattice Quantized LLM s at 2.7 Bits per Parameter
链接: https://arxiv.org/abs/2609.35465
作者: Pier-Jean Malandrino
类目: Machine Learning (cs.LG)
*备注: 15 pages, 4 figures, 8 tables. Code, measurement logs and preregistrations: this https URL
Abstract:Leech-lattice quantization gives good quality at two bits per weight, but its codebooks hold more than 10^14 points, too many for a lookup table. Our earlier kernel expanded the codes at load time and read 4.804 bits per weight from GPU memory for 2 bits of code. We present Tetra, a new codebook on the same lattice. A 24-weight block still takes 48 bits, most of which index a 64-state trellis of the Golay code and one shared 16 KiB table. The kernel decodes a block with six table loads and two small lookups inside the matrix-vector product, and reads 2.148 bits per weight. For full models, we retrain one scale per matrix row, store the matrices that lose the most as 4-bit integers, and pay for them with 4-bit embedding tables. Our Qwen3-4B, 8B and 14B files hold 2.73, 2.70 and 2.73 bits per parameter over the whole model. They score 63.37, 69.58 and 75.66 on the full MMLU test set, 4.76, 4.21 and 2.46 points below 4-bit AWQ at 5.3 to 6.0 bits per parameter. They generate 113.8, 95.0 and 57.2 tokens per second in our engine. On GSM8K, through the served kernel, they lose 9.63, 4.62 and 3.26 points to FP16. At 4B our file scores 23.6 points above this http URL’s IQ2_XXS (2.48 bits per parameter). Every number we measured for a table or figure comes from one NVIDIA L40S GPU. We preregistered the main experiments.
[LG-32] Manifold-Stable Flow Matching
链接: https://arxiv.org/abs/2609.35454
作者: Amirhossein Nazerian,Ali Pezeshki,Jianguo Zhao
类目: Machine Learning (cs.LG); Robotics (cs.RO); Systems and Control (eess.SY)
*备注:
Abstract:Flow matching (FM) learns generative dynamics through velocity regression. Geometric FM variants commonly assume a prior supported on the data manifold, requiring geometric knowledge that is often unavailable. Without such knowledge, low regression error alone does not guarantee manifold adherence. Adherence keeps generated samples within valid configurations and is empirically associated with better task performance. We introduce manifold-stable flow matching (MSFM), which can start from an arbitrary ambient prior, not necessarily supported on the manifold. Using tools from nonlinear dynamics, namely contraction theory, MSFM combines learned tangential transport with prescribed normal contraction. The construction uses analytical projectors for known manifolds and local affine proxies estimated by principal component analysis for unknown data geometry. By implementing contraction theory in both cases of known and unknown manifolds, we guarantee manifold invariance and transverse convergence to the manifold within a desired time window (e.g., one second). We derive a family of compatible probability paths and decompose the training loss into a learnable tangential term and a normal residual. An ellipse experiment attains a mean terminal off-manifold error of order 10^-6 . In Push-T robotic experiments, MSFM raises success from 74% to 82% . In the Robomimic Square task, success increases from 60% to 72% , while rotation-manifold deviation decreases from order 10^-2 to 10^-7 . The MSFM terminal geometric errors are controlled by the chosen numerical tolerance. These results demonstrate stronger geometric adherence and higher observed task performance, supporting prescribed normal contraction as a complement to learned generative transport.
[LG-33] NeuronSifter: Intervention Planning in CNS Microenvironments
链接: https://arxiv.org/abs/2609.35445
作者: Haowei Xu,Wanyi Fu,Hongbin Han,Zhaoheng Xie
类目: Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注: 39 pages
Abstract:Prioritizing central nervous system (CNS) interventions requires predicting how a dose, route, and schedule act on a partially observed microenvironment, then choosing the measurement that would change the decision. Action-conditioned predictors reduce a regimen to an identity token or a scalar exposure, discarding where and when the target is engaged; handing a point estimate to a separate planner then discards the joint uncertainty that makes a measurement worth running. We therefore treat decision quality as a property of the intervention interface, not of controller placement. NeuronSifter compiles regimens into state-conditional target-occupancy fields with support masks, propagates them through microenvironment dynamics with an occupancy-conditioned diffusion operator, and selects measurements by their expected reduction in intervention loss, assimilating typed outcomes into the same posterior. In a declared synthetic Alzheimer’s disease (AD) evaluation over 64 paired scenario blocks, occupancy conditioning lowers trajectory continuous ranked probability score from 0.165 to 0.110 and raises intervention ordering accuracy from 0.760 to 0.880, and every paired benchmark contrast remains separated after Holm correction. Decision-directed acquisition attains terminal risk 0.160 against 0.166 for a matched numerical Bayesian experimental design planner, and reaches the target risk at 0.796 [0.732,0.873] of an earlier design control’s cost, while the corresponding ratio against the matched planner, 0.963 [0.907,1.025] , is not separated from equality; point-state and dependence-ablated interfaces instead raise risk to 0.220 and 0.199, and a full-posterior external controller ties exactly. Published AD trials supply a separate retrospective endpoint bridge.
[LG-34] SOLO: Pretraining Billion-Parameter Language Models with Shared-Output Local Learning
链接: https://arxiv.org/abs/2609.35440
作者: Bojian Yin,Shurong Wang,Yuqi Pan,Guoqi Li
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注: 26 pages, 15 figures, 19 tables. Preprint
Abstract:Large language models are trained with backpropagation, whose global gradient coordinates all layers but forces each to hold its activations and wait for the gradient to pass back through every deeper layer. Conventional local learning removes this update locking by training each module to predict the target through its own readout, but has not scaled to billion-parameter pretraining. We identify these private readouts as a key weakness, since they leave each module without information from deeper modules. We propose Shared-Output LOcal learning (SOLO), which replaces them with a shared, read-only copy of the final module’s readout, the only one trained on the output of the whole network. Taken from the previous step, the copy transmits information from the final module without passing gradients between modules or reintroducing update locking. SOLO approaches backpropagation on Transformers of 340M to 2B parameters pretrained on 15B tokens, staying within one point in average zero-shot accuracy with a perplexity gap that narrows with scale. Readout ablations attribute SOLO’s improvement over private readouts to sharing. Without update locking, each of p pipeline stages holds activations for O(1) micro-batches instead of O§. The freed memory permits larger micro-batches, which reach up to 1.44x the best measured throughput of pipeline backpropagation on the same partition. To our knowledge, SOLO is the first local learning method to show such memory and throughput gains in billion-parameter language-model pretraining. Local learning thus becomes a practical alternative to backpropagation for large-scale pretraining.
[LG-35] Inductive Feedback for Mixed-Policy Distillation
链接: https://arxiv.org/abs/2609.35390
作者: Amir Moeini,Huaijiang Zhu,Daniel Havir,Shangtong Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Verbal feedback can identify errors and prescribe corrections, providing rich supervision for language-model post-training even when reliable programmatic verifiers are unavailable. Such feedback, often generated by a capable model, can be used to condition the teacher in on-policy distillation, which trains the student to match the teacher’s predictions on student-generated rollouts. However, this approach can transfer teacher preferences that the feedback did not motivate, while leaving much of the feedback’s guidance unused. We find that both problems come from the standard on-policy distillation objective, specifically the divergence it minimizes and the distribution it uses as its target. Our proposed method addresses both limitations. First, to isolate the information conveyed by the feedback from the teacher’s inherent preferences, we treat verbal feedback as evidence for or against the hypothesis that a particular token comes next at a given prefix. We then adopt a probabilistic confirmation framework which uniquely determines an ordering over the vocabulary based on the teacher’s predictions before and after it receives feedback. Using a confirmation score consistent with this ordering, we construct a target distribution within a trust region of the student. Second, to learn from guidance that student rollouts can leave unused, we derive a simple shared-rollout estimator of a symmetric divergence between the student and target distributions over rollouts, reusing student and feedback-conditioned teacher rollouts in both directions through importance weighting. Empirical evaluations show that our method outperforms the common on-policy distillation recipe and a recent contrastive variant on knowledge-based and agentic benchmarks.
[LG-36] Identifying Neural Source Dynamics from Unknown Local Interventions
链接: https://arxiv.org/abs/2609.35379
作者: Ayana Mussabayeva,Jiaqi Sun,Anuar Aimoldin,Olivier Oullier,Kun Zhang
类目: Machine Learning (cs.LG)
*备注: 11 pages main text, 52 pages total including references and appendices; 3 figures
Abstract:Electroencephalography (EEG) records mixtures of brain-source activity. Even with a known anatomical forward model, experiments that excite only part of the source-state space leave the dynamics unidentified, and repetition cannot resolve the ambiguity. We show that unknown local mechanism changes can supply the missing information. We consider linear dynamics among fixed anatomical sources with known source-state initialization patterns. Changing one source’s update rule for one transition leaves a rank-one, source-specific signature in subsequent EEG: subtracting matched baseline responses isolates it, and the forward model identifies the source and calibrates its response history. Combining these histories with initialization responses recovers source interactions without baseline reachability and without first identifying the intervention coefficients. We establish sufficient recovery conditions, a direct estimator, and a noise-sensitivity bound conditional on correct source labels. Simulated EEG on anatomy derived from magnetic resonance imaging confirms the information gain: with baseline excitation confined to four of twelve source coordinates, eight unknown changes recover all dynamics in 32/32 systems, whereas baseline realization, baseline regression through an invertible forward model, and changes that leave the tested states unexposed all fail, and explicitly constructed alternative dynamics reproduce every baseline mean. Where baseline information suffices, direct reconstruction is also more reliable than a matched-information spectral estimator. Nonlocal changes and forward-model error limit accuracy even when source labels are correct.
[LG-37] First Learn Then Memorize: The Spectral Bias of Diffusion Models
链接: https://arxiv.org/abs/2609.35377
作者: Raphaël Urfin,Tony Bonnaire,Giulio Biroli,Marc Mézard
类目: Machine Learning (cs.LG); Disordered Systems and Neural Networks (cond-mat.dis-nn)
*备注: 53 pages, 13 figures
Abstract:Diffusion models trained on a finite dataset first learn to generate novel, high-quality samples and only much later collapse onto their training set. We identify the mechanism behind this separation of timescales and the object that probes it. The training dynamics of the score function are governed—exactly, and at any width—by the Gram matrix of the Neural Tangent Kernel (NTK) evaluated on the noisy training data, so the timescales of generalization and of memorization must be encoded in its spectrum. We show that they are, and that the structure responsible has no analogue in standard kernel settings. The use of multiple noise realizations per sample ( m noised copies at a fixed noise level) in the score-matching loss is what restructures the Gram matrix spectrum into two distinct parts. The first, of large eigenvalues, carries the global features of the target distribution and is present already for m=1 . The second, which the repeated noising creates, consists of the smallest eigenvalues and is supported on eigenvectors aligned with the sample-specific noise directions; it sets a memorization timescale parametrically larger in the training set size n . We establish this picture on two fronts. Analytically, we solve the spectrum in the lazy high-dimensional limit for both linear ( n \asymp d ) and polynomial ( n \asymp d^k ) sample complexities, and prove through a bias–variance decomposition that the first bulk minimizes the approximation error while the second drives the error associated with memorization. Empirically, we show the same two-bulk structure in Convolutional NTKs on CelebA and in finite-width U-Nets trained well beyond the lazy regime, and we make the link causal: truncating the Gram matrix at rank r tunes the generalization–memorization transition, and an L_2 penalty targeting the second bulk suppresses memorization in feature-learning U-Nets.
[LG-38] Fiona: Accelerating FHE Inference with Packing-Aware Ternary Weights
链接: https://arxiv.org/abs/2609.35352
作者: Yiteng Peng,Zhibo Liu,Dongwei Xiao,Shuai Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Fully homomorphic encryption (FHE) enables neural network inference directly on encrypted inputs, but it remains orders of magnitude slower than plaintext in- ference. Applying the server’s plaintext weights to encrypted activations involves plaintext-ciphertext multiplications (PMult) and accounts for more than half of inference time in recent systems. Ternary quantization can replace these multipli- cations with additions and subtractions, but the savings rarely materialize under packed execution. A single PMult applies a weight group fixed by the packing layout and can be avoided only when all its weights share the same ternary value. Ternarizing all groups, however, largely degrades accuracy. We present FIONA, an offline optimizer that selectively ternarizes weights within a given packing layout based on the estimated effect of ternary conversion on the model’s performance. FIONA encourages a shared ternary value within each weight group and retains full-precision weights for sensitive groups, so ternar- ized and full-precision paths coexist within a layer. It then compiles these hybrid operators exactly, applying common scaling factors once to accumulated inputs and reusing sums across outputs. Weight ternarization can also narrow the input ranges of downstream polynomials. FIONA fits lower-degree replacements under a cumulative accuracy budget, reducing multiplicative depth and bootstrapping. On VGG11, ViT, and BERT, FIONA reduces PMult operations by 53.4-79.5% and accelerates end-to-end encrypted inference by 2.38x, 1.68x, and 1.84x, re- spectively, with less than 1% accuracy loss across all three models. Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.35352 [cs.LG] (or arXiv:2609.35352v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.35352 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-39] Interference Beyond Geometry in Concept Extraction
链接: https://arxiv.org/abs/2609.35351
作者: Valérie Costa,Bahareh Tolooshams
类目: Machine Learning (cs.LG)
*备注:
Abstract:Interference is commonly treated as geometric overlap between learned features. We introduce effective interference, which combines feature geometry and code statistics to capture realized interactions, distinguishing constructive from destructive interference and frequent weak interactions from rare strong ones. Under local fixed-support assumptions, we characterize how architectural constraints shape interference through four mechanisms: feature orthogonalization, bias compensation, gain adaptation, and encoder-decoder separation. Experiments with sparse autoencoders show that constrained architectures selectively reduce overlap among co-active features, while bias, gain, and encoder freedom allow constructive cross-contributions to remain. Together, these results show that interference in learned representations depends not only on feature geometry, but also on how features are used and on the architecture that produces their codes.
[LG-40] Quasi Linear Kernel Attention with Infinite Capacity
链接: https://arxiv.org/abs/2609.35349
作者: Nicolaj Rux,Johannes Hertrich,Sebastian Neumayer
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注:
Abstract:The evaluation cost of transformers with softmax attention scales quadratically with sequence length. Kernel attention addresses this by replacing softmax with a more general kernel function. In this paper, we aim to identify kernels that retain the expressivity of attention while enabling quasi linear computation. To quantify expressivity, we introduce a capacity for each kernel, measuring the maximum sequence length for which the attention matrix can approximate the identity. A higher capacity thus indicates greater expressivity. We show that expressive kernels like softmax, Gauss, and Laplace have infinite capacity. In contrast, common quasi linear kernels, such as those derived from finite dimensional feature maps, exhibit finite capacity. As a solution, we propose additive kernels constructed from univariate spline and polynomial exponential kernels. We prove that these maintain infinite capacity while allowing quasi linear computation via sorting. Finally, we implement additive sorting kernels efficiently and benchmark them against modern softmax backends, demonstrating advantages for long sequences.
[LG-41] Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation
链接: https://arxiv.org/abs/2609.35347
作者: Xin Li,Hao Jiang,Xin Gao,Annan Wang,Yuchen Xie,Jinghao Guo,Xingwei Qu,Yichi Zhang,Chau Yuen
类目: Machine Learning (cs.LG)
*备注: Project page: this https URL . Code: this https URL
Abstract:Reinforcement learning can turn one language model into several specialists, each excellent at a single skill such as mathematics, coding or following instructions, but users need one model with all of these skills. Multi-teacher on-policy distillation (MOPD) merges them by letting the specialists teach one student: the student answers each prompt, and the specialist for that prompt’s domain gives feedback on every token. This routing decides which specialist teaches, but not how strongly its feedback moves the shared student. In Qwen3.5 models at three sizes, we find that MOPD’s student does not beat one taught by the best single specialist and gains little of the mathematics specialist’s advantage. The feedback is unbalanced: instruction-following feedback is several times more spread out than mathematics feedback and dominates the student’s updates. We propose Domain-Normalized MOPD (DN-MOPD), which keeps the routing and rescales each domain’s feedback by its measured spread. On six public benchmarks, DN-MOPD improves the average score over MOPD at every size, across three random seeds and under two answer-length limits, and recovers most of the lost mathematics gain. Controls with fixed domain weights show that the gain comes mainly from turning down instruction-following feedback rather than turning up mathematics alone, and that fixed weights close to those DN-MOPD measures perform comparably. Combining specialists therefore requires deciding not only which one teaches, but also how strongly its feedback counts.
[LG-42] NeuronDiscover: Agent -in-Twin for Mechanistic Discovery in Neuronal Microenvironments with World Action Models
链接: https://arxiv.org/abs/2609.35338
作者: Haowei Xu,Wanyi Fu,Hongbin Han,Zhaoheng Xie
类目: Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注: 54 pages
Abstract:Mechanistic discovery in neuronal microenvironments requires interventions and measurements that separate competing explanations of solute transport and neuronal response. Predictive accuracy cannot settle the question: a real mechanistic change and an error in the computational twin leave the same signature in sparse observations. We formalize this twin confounding and reason over a joint mechanism–discrepancy belief, designing experiments that separate the two. NeuronDiscover is an Agent-in-Twin framework whose shared, mechanism-grounded World Action Model (WAM) couples prediction, intervention proposals, and observation design; independently adjudicated outcomes revise a scoped Mechanism–Intervention–Observation–Outcome (MIOY) graph, whose supported relations compile into executable programs carrying discrepancy-adjusted acceptance bounds. We evaluate on simulated brain-fluid tracer-transport worlds adjudicated by an independently frozen finer-mesh reference solver, and on donor-disjoint public current-clamp recordings of cortical neurons. Counting only relations that reach a certified terminal status, and scoring abstentions as unresolved for every method, at a matched budget of 16 experiments over 32 source units NeuronDiscover resolves 4.0 relations per assigned world against 3.4 for the strongest baseline and 3.2 without graph revision, at 5% false support and 82% scope accuracy. Joint mechanism–discrepancy acquisition resolves 3.8 relations versus 2.9 for plug-in expected information gain; discrepancy-adjusted verification lowers accepted-program failure from 15% to 9% at 60% acceptance coverage; and transfer to the recordings yields 1.94 versus 1.53 relations per assigned world. Correctness is adjudicated within declared model worlds and archival recordings.
[LG-43] Scalable In-Context Reinforcement Learning with Recurrent Algorithm Distillation
链接: https://arxiv.org/abs/2609.35333
作者: Yuanqing Ma,Zhenrui Zheng,Chenjun Xiao
类目: Machine Learning (cs.LG)
*备注:
Abstract:Algorithm Distillation (AD) has demonstrated the remarkable ability of Transformers to perform in-context reinforcement learning without explicit weight updates. However, capturing long-term learning progress necessitates expansive context windows, which incur prohibitive memory costs and limit scalability in complex, long-horizon tasks. To address this bottleneck, we propose Recurrent Algorithm Distillation (RAD). RAD employs a dual-component architecture: a Compression Transformer that distills extended interaction histories into compact latent tokens, and an AD Transformer that auto-regressively generates actions using a hybrid context of these compressed memories and recent transitions. By maintaining a fixed-size latent buffer, RAD decouples the effective history length from computational complexity, functionally providing the model with a long-horizon memory. Empirical evaluations across diverse environments demonstrate that RAD matches the asymptotic performance of standard AD with significantly reduced context window sizes, offering a scalable solution for efficient in-context decision-making.
[LG-44] Weighting Schedules Govern What and When Score-Based Generative Models Learn from Multimodal Data
链接: https://arxiv.org/abs/2609.35322
作者: Jérémie Klinger,Raphaël Urfin,Giulio Biroli,Marylou Gabrié
类目: Machine Learning (cs.LG); Disordered Systems and Neural Networks (cond-mat.dis-nn)
*备注: Main text : 9 pages / 5 figures Supplemental : 21 pages / 1 figure
Abstract:Score-based generative models generate new samples by integrating a time-dependent drift that carries Gaussian noise onto the target distribution. In practice this drift is modeled by a neural network, trained on a loss integrated over time t with a weighting schedule w(t) . Along the backward dynamics, and for multi-modal distributions, trajectories commit to modes of the target within a narrow time window, the \textitspeciation time. In this work, focusing on high-dimensional data, we decompose the integrated loss into its single-time contributions and analyze each at fixed signal-to-noise ratio \Lambda(t) : we show that \Lambda(t) sets the rate at which each feature of a multimodal target - the mode directions and their relative weights - is acquired during training. Crucially, at high \Lambda(t) all mode directions are acquired together, on a single timescale insensitive to their amplitudes, while the relative weights are not learned at all. Only near the speciation time, where \Lambda(t) becomes of order one, do all features become learnable, each on its own timescale: the weights are acquired jointly with the directions, and the directions at rates set by their relative amplitudes. For models trained on time-integrated objectives, the learning dynamics is then governed by how much of the weighting effectively sits near the speciation time, which provides insights on w(t) design choices. These results follow from an exact high-dimensional analysis of the training dynamics of unbalanced and hierarchical Gaussian mixtures. Numerical experiments on image and human genome haplotype generation recover the predicted hierarchy of learning timescales in more complex settings.
[LG-45] Collaborative Principle Evolution via Evidence Transfer for Scientific Discovery
链接: https://arxiv.org/abs/2609.35315
作者: Yingming Pu,Hongyu Chen,Tao Lin
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large Language Model (LLM)-based agents promise to automate scientific discovery, yet exploring the vast hypothesis space remains costly. Existing principle-evolution methods accelerate this loop, but operate sequentially, which caps exploration breadth and wastes wall-clock time on challenging problems. To address this, we formulate collaborative scientific discovery as evidence transfer between parallel principle-evolution branches. We present COEVOLVE, which realizes this transfer through a coordination core over parallel branches. By integrating value-of-information-gated routing and context-discounted likelihood injection, COEVOLVE enables branches to collaborate through shared measurements while keeping their principle posteriors separate. Across six scientific-discovery tasks under a matched evaluation budget, COEVOLVE attains a mean solution quality of 66.5% versus 57.0% for single-branch principle evolution, with a 1.80x mean wall-clock speedup on the GPT-5.6-Terra backbone; on five auto-research tasks delegated to an autonomous research harness, it is the only arm whose mean stays above the published SOTA anchor on every task. These results establish when evidence sharing accelerates parallel discovery and when transfer safeguards are necessary to limit negative or inert transfers
[LG-46] When Should a Satellite Estimate Be Changed? Stress-Testing Neural Corrections for Evapotranspiration
链接: https://arxiv.org/abs/2609.35314
作者: Marco Trotta
类目: Machine Learning (cs.LG)
*备注: 11 pages
Abstract:Neural residuals can improve satellite evapotranspiration (ET) estimates, but selectors must predict when a correction helps and reject unsupported inputs. We evaluate ten-member models on 16,366 flux-tower observations from 151 stations paired with OpenET, across nine rolling years and five spatial folds. At one held-out station, Gain accepted corrections on all 32 physically invalid records: it predicted a mean benefit of 0.83 mm/day, but the corrections increased mean absolute error by 21.6 mm/day versus OpenET. On spatially held-out unit errors, SupportGain reduced station-macro MAE versus Gain by 0.148 mm/day under wind x3.6 (simultaneous 95% interval, 0.070 to 0.226), with 9.3% acceptance versus Gain’s 51.8%; on clean inputs, its 0.006 mm/day advantage had an interval that includes zero. These fault analyses are exploratory; none of 40 preplanned temporal comparisons passed Holm correction, while a separate predeclared cropland contrast found 0.041 mm/day lower station-macro MAE with crop-only training (95% interval, 0.009 to 0.079).
[LG-47] LionMuon: Alternating Spectral and Sign Descent for Efficient Training
链接: https://arxiv.org/abs/2609.35297
作者: Arman Bolatov,Artem Riabinin,Nikita Kornilov,Andrey Veprikov,Samuel Horváth,Martin Takáč,Aleksandr Beznosikov
类目: Machine Learning (cs.LG)
*备注: 37 pages, 4 figures, 11 tables
Abstract:Pretraining a language model takes enormous compute, and the right optimizer can save a good part of it. Muon’s spectral step gives a stronger direction than a sign step, but it is expensive. Every step runs Newton-Schulz iterations on the full matrix and, in distributed training, an extra all-reduce. Sign steps, as in Lion and Signum, are cheap and stay local to each device. We propose LionMuon, which takes one Muon step every P iterations and Lion steps in between, with a single dual-EMA momentum buffer shared by both. Muon’s compute and communication are paid once per P steps, and the optimizer state is half of AdamW’s. A single-EMA variant, SignMuon, already improves on Muon. We prove complexity bounds under heavy-tailed noise in which the period sets an interpolation between Muon’s and Lion’s smoothness and noise constants, and which say when LionMuon is faster than both. On 124M and 355M models trained on FineWeb, LionMuon with P=2 and P=5 reaches a lower loss than Muon, AdamW, Lion and Signum at the same number of tokens. Under 4-GPU data-parallel training it reaches Muon’s final loss with a third less wall-clock on PCIe, and it beats the communication-efficient Muon variants Dion and MuonBP on loss at no more exposed communication, while keeping the exact gradient. Code: this https URL
[LG-48] SpikeCredit: Temporal Credit Carrier for Reinforcement Learning with Sparse Rewards
链接: https://arxiv.org/abs/2609.35268
作者: Yingchao Yu,Pengfei Sun,Wenxuan Pan,Wei Chen,Yitian Hong,Kuangrong Hao,Yaochu Jin
类目: Machine Learning (cs.LG)
*备注: 14 pages, 9 figures
Abstract:Reinforcement learning (RL) with sparse rewards is challenging because delayed outcomes provide little guidance about which intermediate computations caused success or failure. We argue that reliable credit assignment requires policy dynamics that preserve and expose credit-relevant information over time, a role we formalize as Temporal Credit Carriers (TCCs) and that spiking neural networks (SNNs) naturally fulfill through graded membrane traces and event-driven spikes. Based on this hypothesis, we propose SpikeCredit, an SNN-based framework for RL with sparse rewards that first performs task-adaptive TCC selection and then closes the loop between a fast TCC-reading pathway, where self-motion feedback constraint uses local behavior-grounded cues to constrain transition-level credit recovery, and a slow TCC-writing pathway, where credit-targeted trace alignment feeds recovered credit back into the actor to make future TCC dynamics more credit-readable. Across sparse-reward MuJoCo tasks, SpikeCredit improves Last10 return over sparse SNN baselines by +1169% on Ant, +953% on Hopper, +723% on Swimmer, and +1781% on Walker2d, and exceeds the dense-reward baseline on Swimmer by +113%. Mechanistic analyses further show substantially stronger alignment with dense rewards than the sparse SNN baseline. These results position spiking dynamics as credit-preserving substrates for sparse-reward RL.
[LG-49] Latency and accuracy tradeoffs in Spiking Neural Networks
链接: https://arxiv.org/abs/2609.35260
作者: Zhanglu Yan,Zixuan Zhu,Kaiwen Tang,Yuyang Cai,Qianhui Liu,Weng-Fai Wong
类目: Machine Learning (cs.LG)
*备注:
Abstract:Spiking neural networks are attractive for low-power speech command recognition, yet their latency has received far less attention than their energy efficiency, and their multi-timestep execution is widely assumed to make them slower than quantized neural networks. This paper challenges the assumption that more local timesteps necessarily imply higher network latency. By overlapping computation across adjacent layers at the timestep level, SNNs may complete execution in less time than comparable bit-serial QNNs. However, this overlap relies on spikes firing on incomplete inputs, and a spike once generated cannot be withdrawn, so its error persists and reduces accuracy. Waiting for more input before firing would seem to improve accuracy at the cost of reduced overlap. Yet we find and prove that this intuition fails at some layers, where even a small increase in waiting can change spike timing and downstream computation, making the network both slower and less accurate. We therefore propose a Pipeline Delay Search method which selects each layer’s delay by balancing task-level accuracy gains against added network latency. We then adapt the selected configurations through spike-based quantization-aware training and bounded tuning of firing thresholds and initial membrane potentials. Together, these steps form Falcon, a framework for Fine-grained Analysis of Latency and Controlled firing which systematically analyzes and optimizes SNN latency under a spatial analog compute-in-memory mapping with shared digital engines. We evaluate Falcon on GSCV2 and SSC, achieving competitive accuracies of 96.31 and 83.02 at modeled network-core latencies of 119.64 and 124.00us, respectively. Together, our analysis and results show that SNNs can compute more yet finish faster, and wait longer yet predict worse, highlighting why Falcon matters for both latency and accuracy.
[LG-50] On-Policy or Off-Policy Learning? A Systematic Study of Distillation Dynamics
链接: https://arxiv.org/abs/2609.35259
作者: Julianna Piskorz,Antonin Berthon,Mihaela van der Schaar
类目: Machine Learning (cs.LG)
*备注:
Abstract:On-policy learning has been argued to reduce catastrophic forgetting, produce sparser parameter updates, and improve generalisation. However, existing comparisons between supervised fine-tuning and reinforcement learning vary many factors simultaneously, making the contribution of rollout policy difficult to isolate. We study the effect of rollout policy in a controlled strong-to-weak distillation setting, by independently varying rollout policy, token-level KL direction, and learning rate across the Llama3 and Qwen2.5 model families and reasoning tasks spanning scientific, medical, and arithmetic domains. Our analysis reveals a nuanced picture of distillation dynamics in which rollout policy does not necessarily play a central role. Instead, token-level KL direction more clearly shapes task performance and output coverage, while learning rate governs forgetting and update sparsity. Analysis of KL gradients and experiments along a continuous student-teacher rollout-policy spectrum explain this pattern: forward KL is remarkably robust to rollout policy, with its performance stable and strong despite changes to the rollout policy, whereas reverse KL is substantially more sensitive and favours student-generated rollouts. On-policy data nevertheless improves generalisation to harder variants of the Countdown arithmetic task under both KL directions, although this advantage does not reliably persist after subsequent RLVR. Our broader conclusions remain robust to removing gradient clipping, using sampled KL estimators, and training on tasks requiring longer reasoning chains. Overall, our results challenge the view that on-policy rollouts are inherently preferable and show that their value depends critically on the objective, evaluation setting, and optimisation hyperparameters.
[LG-51] AIM-ZO: Activation-Informed Subspace Maintenance for Zeroth-Order LLM Fine-Tuning ICLR2027
链接: https://arxiv.org/abs/2609.35257
作者: Yue Xie,Zhi Zheng,Yunpeng Ba,Xuyang Wu,Xialiang Tong,Zhichao Lu,Tao Zhong,Zhenkun Wang
类目: Machine Learning (cs.LG)
*备注: Submitted to ICLR 2027
Abstract:Zeroth-order (ZO) optimization offers a memory-efficient alternative for LLM fine-tuning by estimating updates only from forward evaluations of perturbed parameters, without backpropagation or activation storage. However, in billion-parameter LLMs, isotropic perturbations often waste many forward evaluations on weakly informative directions. To make these evaluations more informative, existing ZO methods restrict perturbations to low-dimensional subspaces. Yet the quality of these subspaces is critical: overly compressed or poorly maintained spaces can miss useful update directions. To obtain a high-quality subspace for ZO updates, this paper proposes AIM-ZO, a ZO fine-tuning method based on Activation-Informed Subspace Maintenance. AIM-ZO uses forward activations as local directional information and continuously integrates them into a broad, evolving subspace over training. To access broader gradient-relevant structure while keeping individual perturbations low-dimensional, AIM-ZO activates only a smaller set of shared and sampled directions, decoupling the maintained width from the active width. We evaluate AIM-ZO across 5 LLMs and 11 downstream tasks under matched forward-evaluation budgets; its six-task average exceeds the strongest fully evaluated ZO baseline by 1.26 percentage points on OPT-2.7B and MeZO by 2.85 percentage points on OPT-30B. Our code is available at this https URL
[LG-52] Disentangling Lung-Cancer CT/LDCT AI: A Systematic Evidence Map of Clinical Tasks Evidence Chains and Translational Gaps
链接: https://arxiv.org/abs/2609.35240
作者: Surajit Das
类目: Machine Learning (cs.LG)
*备注:
Abstract:Artificial-intelligence studies using computed tomography (CT) for lung cancer are often broadly labelled “prediction” despite addressing clinically distinct tasks. We systematically mapped CT/low-dose CT (LDCT)-centered lung-cancer AI using five-database retrieval, full-text eligibility assessment, role-aware modality/omics extraction, clinical-task classification, and a Multi-Tier Evidence Graph (MTEG). The final corpus comprised 293 studies (2016-2026): 230 Detection, 8 future Risk-prediction, and 55 Other studies. Clinical variables (96.2%), 3D CT/LDCT (73.0%), and radiomics (63.5%) predominated, whereas external validation (29.0%), calibration (20.5%), decision-curve analysis (13.0%), longitudinal CT (17.7%), and saliency/attribution XAI (21.5%) were less frequent. The MTEG comprised 377 nodes and 3,444 edges; only 31 studies (10.6%) completed the six-tier substantive evidence chain, with greatest attrition at reasoning/explanation. Overall, the literature is detection-dominated, genuine future risk prediction remains uncommon, and complete translational evidence chains are rare.
[LG-53] Long-Horizon Scaling: How Model Capabilities Shape the Returns to Computation
链接: https://arxiv.org/abs/2609.35236
作者: Haoyu Zheng,Zhengyu Chen,Huaisheng Zhu,Ruishan Fang,Teng Xiao,Yiwei Li,Jingang Wang,Wenqiao Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Long-horizon agents improve solutions through sustained interaction, execution, and task feedback. Scaling studies relate performance to resources and capabilities, yet how existing capabilities shape returns to extended interaction remains less understood. To address this gap, we analyze AutoLab and EdgeBench, two long-horizon benchmarks. We find that starting performance and subsequent growth are associated with different capabilities: within a task category, similar early scores can precede different later gains. To formalize this finding, we model capability-time scaling with category-specific logistic power laws shared across models. Fitted to early trajectories, these curves extrapolate the observed models’ category-average scores to later computation. However, rising average scores mask narrowing improvement opportunities: later gains concentrate among fewer improving models. High final scores and continued improvement also have distinct capability profiles. Predicted mean gains estimate each model’s fraction of improving tasks; averaging these estimates forecasts the average share of improving models. These uneven returns motivate deciding whether a specific run should continue. We therefore derive a continuation policy to save time and compute with limited score loss. The policy conditions growth predictions on the run’s observed progress and weighs immediate and delayed gains against computation costs. In replay with training and price calibration based on other models’ histories, the policy saves roughly one-third of full-run time, with relative score losses of 2.4% on AutoLab individual runs and 3.3% on EdgeBench published mean curves. Our repository is available at this https URL.
[LG-54] mporal Heterogeneous Graph Pretraining for Relational Deep Learning
链接: https://arxiv.org/abs/2609.35219
作者: Yixin Peng,Er Jin,Diego Collarana,Stefan Decker
类目: Machine Learning (cs.LG)
*备注:
Abstract:Relational deep learning models database rows and foreign-key links as a heterogeneous graph for prediction from record attributes and relational context. These graphs contain two distinct temporal signals: record age changes with the prediction cutoff, while intervals between observed records remain fixed. Prior work often treats time as a single signal or studies temporal representation and pretraining separately. We investigate how explicitly encoding both signals affects temporal pretraining for downstream tasks. Our framework combines Multi-scale Time Encoding, which captures record age using learnable time scales and type-specific projections, with Rotary Time Encoding, which represents signed inter-record intervals through rotary transformations during graph propagation. We pair these encodings with three self-supervised objectives: historical relation recovery, horizon-aware future relation activity prediction, and temporal subgraph contrast. All inputs respect their observation cutoffs. Pretraining proceeds in two stages: subgraph contrast first learns neighborhood representations, followed by refinement through either relation recovery or future activity prediction. We evaluate on five RelBench datasets across 11 classification and regression tasks using heterogeneous GNN and graph Transformer backbones. With both encodings, the best evaluated staged schedules improve over supervised training with the same encodings by 3.02% and 1.06% on the two backbones, respectively, and over controls without pretraining or either encoding by 3.24% and 2.37%.
[LG-55] Adversarial Consistency-Guided Representation Learning for Multi-view Clustering
链接: https://arxiv.org/abs/2609.35212
作者: Yuchen Lin,Kunpeng Xu,Ying Fang,Lifei Chen
类目: Machine Learning (cs.LG)
*备注: 5 pages, 4 figures
Abstract:Multi-view clustering aims to capture cross-view consistency while exploiting view-specific information. However, shared representations learned to capture cross-view consistency may still retain view-identifying information, potentially compromising the consistency of cross-view clustering structures. To address this issue, we propose ACGRL, an adversarial consistency-guided representation learning framework for multi-view clustering. ACGRL employs a gradient-reversal view discriminator to reduce view identifiability and obtain invariant reference representations. These representations are then frozen to provide fixed references for disentangling view-specific information from cross-view common information in the subsequent learning stage. The fixed reference representations are concatenated with the learned view-specific representations for reconstruction and clustering, with cross-view cluster alignment encouraging consistent clustering assignments. Experiments on four benchmark datasets demonstrate the superior clustering performance of ACGRL compared with representative multi-view clustering methods.
[LG-56] ConRAG : Lightweight inference of multi-hop relations
链接: https://arxiv.org/abs/2609.35193
作者: Kilian Bänziger,Sonia Laguna,Markus Kreft,Robert Jakob,Kevin O’Sullivan,Lasse B. Strand,Julia E. Vogt
类目: Machine Learning (cs.LG)
*备注:
Abstract:Understanding how two entities are connected often requires tracing multi-hop relations across documents to identify intermediate entities and supporting evidence that explain a connection. This is a task that appears frequently in scientific research and other knowledge-intensive analyses. We formalise this setting as multi-hop relation inference: given two known endpoint entities, we aim to recover the bridge entities and evidence-grounded reasoning chains that connect them across a document corpus, and to generate an explanation grounded in the retrieved evidence. Existing multi-hop RAG systems typically seek an unknown answer entity rather than explicitly recovering the connection between two known endpoints and graph-based approaches often rely on costly LLM-extracted knowledge graphs that limit scalability to large document collections. We introduce ConRAG, which builds a lightweight entity-document graph from entity co-occurrence and LLM-based entity filtering. Its connective retrieval infers and semantically ranks paths between two endpoints. On MuSiQue and 2WikiMultiHopQA, ConRAG consistently improves bridge entity and reasoning chain recovery over strong RAG baselines, while reducing graph-indexing token cost by up to roughly 1.5 orders of magnitude. Our results show that endpoint-constrained path retrieval provides an effective and index-efficient approach to evidence-grounded relation discovery.
[LG-57] Subgroup Rank-1 Lattice for Practical High-dimensional Black-box Integral Approximation
链接: https://arxiv.org/abs/2609.35177
作者: Yueming Lyu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Estimating integrals of black-box, high-dimensional functions, from expectations and kernel mean embeddings to the softmax kernel in self-attention, is a basic subroutine in machine learning. Rank-1 lattice rules suit this setting: they query the integrand only at a fixed point set and need no gradients. When the n points serve as a design matrix X\in\mathbbR^n\times d for a feature map, however, computing \Psi(X)^\top v or \Psi(X)w for an elementwise nonlinearity \Psi costs O(nd) time and memory for any standard quasi-Monte Carlo point set. We study subgroup rank-1 lattices, whose Korobov generator (1,t,\dots,t^d-1) uses a scalar t of fixed multiplicative order m . Splitting \mathbbF_n^\times into cosets of \langle t\rangle reduces both maps to short cyclic correlations evaluated by FFT, giving exact results for arbitrary \Psi in O(n\log m) time and O(n) memory, without forming X . Since fixing m falls outside classical component-by-component theory, we prove convergence directly: via resultants with the cyclotomic polynomial \Phi_m , the squared worst-case error in the Korobov space decays as O(n^-(\alpha-1)/(m-1)) for prime m\ge d+1 , and this threshold is exact. Using the splitting of n in \mathbbQ(\zeta_m) , averaging over the m-1 admissible generators improves the constant by a factor \Theta(m-1) . Empirically, the subgroup lattice beats Gaussian and orthogonal random features and scrambled Sobol’ and Halton points in 49 of 54 synthetic kernel-estimation settings and all 45 softmax-attention settings on nine real datasets, and builds a sample set with d=2048 , n\approx4.1\times10^7 in 2.3 ms.
[LG-58] ProtoSeam: Lifting Classifier Training with Latent Gaussian Mixture Models
链接: https://arxiv.org/abs/2609.35174
作者: Robert Lampel,Timon Klein,Sebastian Sager
类目: Machine Learning (cs.LG)
*备注:
Abstract:We propose a lifted reformulation of supervised classification that improves the final accuracy of standard classifiers without changing the architecture at inference time. A network N=N_2\circ N_1 is split at a single semantic interface and one learnable prototype per class is inserted there. Training combines a quadratic consensus penalty that pulls N_1(x) toward the prototype of its class with a classification loss of N_2 evaluated on samples drawn around the prototypes, whereat no gradient crosses the interface. At inference the prototypes are discarded and the unmodified network N_2\circ N_1 is used. Across CIFAR-10, CIFAR-100, and TinyImageNet with ResNet and vision transformer backbones, lifted training improves test accuracy by up to five percentage points over variants without lifting under a shared tuning protocol. Moreover, we provide theoretical justification of those results.
[LG-59] EdgeCraft: Automated Model Crafting for Edge IoT
链接: https://arxiv.org/abs/2609.35167
作者: Genglin Wang,Kaiwei Liu,Liekang Zeng,Wangsong Yin,Shangcheng Jin,Guoliang Xing,Zhenyu Yan
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注: 20 pages, 15 figures, 5 tables
Abstract:Machine learning (ML) increasingly powers Internet of Things (IoT) applications at the edge. Yet producing a deployable edge ML artifact for a specific scenario requires navigating a huge search space spanning data representation, model design, training on domain-specific data, and runtime customization. This workflow is fragmented and difficult to scale across diverse edge applications. We present EdgeCraft, an LLM-driven system that turns high-level intent into deployable edge ML artifacts. Building such a system raises two challenges: (1) How can an LLM be guided to find high-quality solutions that meet dynamic SLOs for task quality, latency, and energy? (2) How can trustworthy target-device verification be obtained at low cost? EdgeCraft addresses these challenges with two designs. (1) A constraint-aware synthesis tree explores alternative candidates and uses measured SLO gaps to guide each improvement. (2) A multi-fidelity verifier progressively combines low-cost checks with full target-device verification to reduce verification cost while preserving reliable verification results. It also records verified failures for reuse, avoiding repeated device work. To support concurrency, EdgeCraft provides a multi-tenant runtime that runs cloud training and target-device verification in parallel while isolating requests. Across 50 public tasks, EdgeCraft exceeds the task-specific Reference in best-observed quality on 40 tasks and finds an SLO-feasible artifact on 45, with the two outcomes overlapping on 38 tasks. Moreover, EdgeCraft achieves competitive performance on our self-collected SEN dataset, suggesting its generalizability to real-world IoT sensing tasks. Comments: 20 pages, 15 figures, 5 tables Subjects: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC) Cite as: arXiv:2609.35167 [cs.LG] (or arXiv:2609.35167v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.35167 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-60] Small transformers track Bayesian evidence for latent common causes via a context-invariant mechanism
链接: https://arxiv.org/abs/2609.35161
作者: Amir Mohammadpour,Michael Franke
类目: Machine Learning (cs.LG)
*备注:
Abstract:We present an in-depth investigation of how a form of Bayesian reasoning about common causes can emerge as a cross-contextual generalization in small, tractable transformers. Incrementing on recent work, our set-up (i) disentangles causal mechanisms in the model from the causal structure of the true data-generating process, (ii) orients more towards natural language prediction by considering inference of latent common causes, and (iii) considers whether and how Bayesian evidence accumulation for latent common causes can be implemented in representations and mechanisms that allow for cross-context generalization to novel test cases.
[LG-61] CacheRepair: Learning to Repair Cross-Chunk Context in RAG for KV Cache Fusion
链接: https://arxiv.org/abs/2609.35139
作者: Genglin Wang,Wangsong Yin,Yeerzhati Abudunuer,Haoxuan Xu,Guoliang Xing,Zhenyu Yan
类目: Machine Learning (cs.LG)
*备注: 33 pages, including references and appendices
Abstract:Multi-document retrieval-augmented generation (RAG) requires a language model to process multiple retrieved text chunks before answering a question. Precomputing each chunk’s KV cache independently and concatenating the caches when the chunks are retrieved can accelerate this step. However, the assembled cache lacks cross-chunk attention information, reducing answer quality. Selective recomputation methods recover the missing cross-chunk context by rerunning the target LLM on selected tokens, incurring substantial online computation. We introduce CacheRepair, a lightweight network that learns the difference between independently computed KV caches and those produced by processing the chunks together. The network combines compressed KV features with token embeddings and uses attention that is bidirectional within each chunk and flows from earlier to later chunks. Each repair block receives the compressed cache features, and the predicted residual is added to every document token’s cache. Each repair network is trained for a specific frozen target LLM on a generic retrieval corpus and reused across downstream datasets. Our analysis shows that repair reduces KV errors both near chunk boundaries and throughout chunk interiors. Evaluation across three target LLMs and four downstream datasets places CacheRepair on the measured answer-quality-latency Pareto frontier in eleven of twelve model-dataset combinations. Reported time to first token (TTFT) includes online cache transfer and repair. Across all twelve combinations, the largest repairers achieve 1.69-4.61 \times speedups in median TTFT over full prefill and improve mean F1 by 2.1-26.1 percentage points over direct cache reuse.
[LG-62] FlexiWorld: Learning and Planning via Flexible Action Chunks Across Multiple Time Scales
链接: https://arxiv.org/abs/2609.35138
作者: Shidu Ren,Qilin Gu,Zhenghao Ni,Junhan Sun,Jiaqi Wang,Damien Scieur,Yunze Liu
类目: Machine Learning (cs.LG)
*备注: 25 pages, 12 figures. Project page: this https URL
Abstract:Latent world models predict future states for goal-directed planning using action chunks spanning multiple primitive steps. Existing methods typically use fixed-length chunks and either omit goal-conditioned action generation or limit their supervision to short goal spans. We introduce FlexiWorld, a JEPA-based world model that combines mixed-span goal supervision with variable-length action chunks to improve long-horizon control. During training, we sample varying goal spans and randomly partition the actions into variable-length chunks. We jointly train the world model with a causal action encoder that embeds variable-length chunks and an autoregressive actor that generates primitive actions sequentially. Student Forcing reduces exposure bias by training on generated action prefixes. For planning, Actor-Residual Cross-Entropy Method (ARCEM) combines action-residual search with within-chunk autoregressive feedback and chunk-boundary latent prediction. Across four benchmarks and goal distances, FlexiWorld with ARCEM achieves 89.29% mean success, compared with 83.98% for the strongest baseline. PushT ablations show improved direct control from mixed-span supervision, variable-length chunks, and Student Forcing. Without retraining, FlexiWorld supports different planning chunk lengths: longer chunks accelerate ARCEM by approximately 1.3\times on average while maintaining comparable average success.
[LG-63] Explaining Hyperbolic Neural Networks via Geometry-Aware Relevance Propagation
链接: https://arxiv.org/abs/2609.35128
作者: Ping Xiong,Shanglin Li,Yi Ding,Thomas Schnake,Shinichi Nakajima
类目: Machine Learning (cs.LG)
*备注:
Abstract:Hyperbolic neural networks introduce geometric operations that require explicit treatment in relevance propagation. Equivalent geometric realizations can produce different feature attributions, even when local relevance is conserved. We study this problem through Geometric Representation Invariance (GRI), a specialization of Implementation Invariance, and zero-curvature consistency, which requires identity relevance propagation when a geometric module approaches the identity. We propose LRP-radial-all for origin-centered radial modules, treating geometric scaling as modulation and assigning relevance entirely to the signal branch. The rule conserves relevance, is invariant to equivalent radial factorizations, and satisfies zero-curvature consistency, yielding GRI for a specified Poincaré-Lorentz logarithmic-map construction. In contrast, a conservative LRP-half baseline can violate both consistency criteria. Experiments on hyperbolic MNIST, sEEG, and CIFAR-10 classifiers assess attribution fidelity, qualitative explanations, and runtime. LRP-radial-all achieves competitive attribution fidelity across datasets with runtime comparable to Gradient \times Input and substantially lower than Integrated Gradients. These findings motivate geometry-aware propagation rules that distinguish relevance conservation from consistency across equivalent computations.
[LG-64] A Multimodal Autonomic Sensing Framework for Objective Assessment of Patient Responses to Dental Pulp Stimulation
链接: https://arxiv.org/abs/2609.35121
作者: Youngsun Kong,Yubin Choi,Dongjin Song,Dong-Guk Shin,I-Ping Chen,Ki Chon
类目: Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: 14 pages, 9 figures
Abstract:Patient responses to dental pulp testing, ranging from no sensation to intense pain, provide important information for assessing pulp status in endodontic diagnosis. However, pain is a subjective sensory and emotional experience that varies considerably across individuals and can be difficult to communicate. We investigated whether complementary autonomic signals could support objective assessment of responses during dental examination. Forty-nine patients underwent cold pulp testing, yielding no-response, mild-response, and intense-response conditions. The framework integrated ECG-derived skin nerve activity (SKNA) and R-R intervals (RRI), together with electrodermal activity (EDA), using temporal convolutional network encoders with attention-based mid-level fusion. Individual baseline signals and subject-level covariates, including anxiety scores and biological sex, were also incorporated. The framework achieved 80.2% balanced accuracy, 75.2% sensitivity, and 85.2% specificity for binary classification of no response versus mild or intense response. For three-class classification, it achieved 60.0% balanced accuracy and a 58.8% macro-averaged F1 score. Ablation and attention-weight analyses indicated that EDA contributed most strongly to model performance, followed by RRI, while SKNA improved balanced accuracy by approximately five percentage points. Age was significantly associated with model performance. These findings support the feasibility of multimodal autonomic sensing for objective, non-invasive assessment of responses to dental pulp stimulation.
[LG-65] SymbolicArena: A Unified Infrastructure for Benchmark Distillation and Dynamic Evaluation in Symbolic Regression
链接: https://arxiv.org/abs/2609.35113
作者: Ziwen Zhang,Xiju Wu,Yuheng Jing,Runxiang Wang,Boxiao Wang,Yifan Zang,Yifan Zhang,Yang Wang,Kai Li,Yifan Zhang,Huilin Xu,Jian Cheng
类目: Machine Learning (cs.LG)
*备注:
Abstract:Symbolic regression (SR) seeks concise and interpretable mathematical expressions from data for scientific equation discovery. Existing SR benchmarks face a tradeoff between evaluation cost and benchmark validity. Repeated evaluation of large task pools is expensive, and compact benchmarks lack systematic evidence of preserved task diversity and algorithm discriminability. SymbolicArena provides a unified infrastructure for benchmark distillation and dynamic evaluation. The framework standardizes 664 heterogeneous tasks with executable ground truth expressions and distills the Full Task Set into Core50, a validated benchmark of 50 tasks. The distillation process preserves task coverage and algorithm discrimination under explicit balance constraints. SymbolicArena applies a unified execution protocol to heterogeneous SR algorithms and produces comparable outputs and search trajectories. Multi Axis Evaluation characterizes numerical quality, symbolic quality, and search behavior. Core50 reduces evaluation workload by 92.5% and maintains agreement with Full Task Set evaluations. Experiments show that SymbolicArena achieves 72.6% to 86.7% lower approximation error than alternative selectors, further supporting its fidelity to the Full Task Set. Evaluation reveals a substantial gap between numerical fitting and symbolic recovery across current SR methods, suggesting that reliable equation recovery remains an open challenge.
[LG-66] DRIFT: Disentangled Responsive-Invariant Flow Transport for Single-Cell Perturbation Prediction
链接: https://arxiv.org/abs/2609.35106
作者: Mustapha Bounoua,Giulio Franzese,Pietro Michiardi
类目: Machine Learning (cs.LG)
*备注:
Abstract:Predicting cellular responses to perturbations is a central problem in cellular biology, with broad applications in systems biology and drug discovery. This task is challenging because cellular responses can be complex and cell-state dependent, intrinsic cell-to-cell variability can be confounded with perturbation effects, and destructive single-cell RNA sequencing precludes paired measurements of the same cell before and after treatment. Flow matching transports control cells to perturbed states flexibly, but acting on the full cell state can confound perturbation effects with pre-existing cell-to-cell variability. Disentangled approaches separate responsive from invariant components, but model perturbations through prescribed mechanisms, such as latent shifts or graph edits, limiting their flexibility. We address both limitations in a unified framework. A variational encoder disentangles each cell into an invariant block, capturing state unaffected by the perturbation, and a responsive block, capturing state it changes, through conditional priors and an information-theoretic invariance constraint. Conditional flow matching transports only the responsive block, conditioned on the perturbation and invariant state, yielding a flexible, data-driven model of perturbation effects without confounding pre-existing variability. Across several benchmarks, our method outperforms the strongest published method in settings involving combinatorial and unseen perturbation prediction.
[LG-67] E3J: An Efficient and Open-Source Backend for Euclidean Equivariant Operations on GPU and TPU
链接: https://arxiv.org/abs/2609.35099
作者: Olivier Peltre,Armand Picard,Adrien Pichard,Miguel Bragança,Luca Giacomoni,Valentin Heyraud,Zachary Weller-Davies,Christoph Brunken,Jules Tilly
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC); Mathematical Software (cs.MS); Computational Physics (physics.comp-ph)
*备注: 9 pages (36 total), 12 figures, 4 tables
Abstract:We present e3j, a fast Euclid-equivariance backend for geometric deep learning applications with JAX bindings for GPU and TPU. Leveraging both optimized CUDA and Pallas kernels and algorithmic improvements, the library achieves state-of-the-art throughput and runtime on both forward and backward paths. On a machine learning interatomic potential (MLIP) use case, it outperforms established backends, measuring up to 34% speed-up over cuEquivariance on water box NPT simulation using MACE, while remaining fully open source. E3j achieves over 80% efficiency over the H100 maximum memory bandwidth on tensor product operations, and in many cases more than doubles throughput of message passing convolutions forward compared to previously available backends. In addition, with the release of dedicated Pallas TPU kernel, e3j opens the possibility of large scale equivariant deep learning workloads on TPU architectures, which has so far been difficult to achieve. Our benchmarks show that e3j also achieves over 80% of a TPUv6e memory bandwidth, up to one order of magnitude more than e3nn-jax. The library is available on GitHub, PyPI and is released under an open source Apache 2.0 license.
[LG-68] Retrieval-Augmented Diffusion Modeling for Stochastic Discount Factor Portfolios NEURIPS2026
链接: https://arxiv.org/abs/2609.35086
作者: Kelvin J.L. Koa,Xinyang Li,Ke-Wei Huang
类目: Machine Learning (cs.LG); Computational Finance (q-fin.CP); Portfolio Management (q-fin.PM)
*备注: NeurIPS 2026
Abstract:In this work, we study portfolio optimization under the stochastic discount factor (SDF) framework by learning market state representations that capture the underlying risk structures of financial data. This is challenging due to several factors: financial markets exhibit non-stationary dynamics with shifting regimes, multimodal inputs such as price and news data often contain stochastic noise, and existing diffusion-based approaches, while effective for modeling stochastic dynamics, rely on assumptions such as isotropic Gaussian noise that fail to capture the state-dependent nature of financial uncertainty. To address these challenges, we introduce RADAR, a retrieval-augmented diffusion framework that learns market representations by conditioning on similar historical regimes. RADAR leverages retrieval to construct context-dependent noise distributions, applies conditional diffusion to denoise multimodal representations, and initializes the diffusion process using empirical statistics to reflect state-dependent uncertainty. Experiments show that RADAR achieves state-of-the-art performance on key risk-adjusted metrics while producing economically meaningful signals on asset returns and correlations.
[LG-69] GraphHCA: Closed-Form Hindsight Credit Assignment for Long-Horizon LLM Agents
链接: https://arxiv.org/abs/2609.35084
作者: Haodong Zhu,Yangyang Ren,Changbai Li,Sheng Xu,Linlin Yang,haiguang liu,Baochang Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Group-based reinforcement learning (RL) has advanced large language models (LLMs) and is increasingly extending to agentic tasks, where sparse terminal rewards make step-level credit assignment essential. Existing methods assign credit from what follows an action in sampled rollouts, but do not explicitly capture its retrospective relation to the realized outcome. Hindsight credit assignment (HCA) instead attributes credit through the ratio of hindsight to behavior-policy probabilities, but estimating the hindsight distribution requires an auxiliary model or an extra pass. To address this estimation bottleneck, we propose GraphHCA, a model-free realization of HCA that eliminates explicit hindsight-distribution estimation. For terminal-goal tasks with deterministic transitions, Bayes’ rule reduces the hindsight ratio to a ratio of behavior-policy success probabilities at consecutive states. Taking logs yields a state-wise success potential, whose increment across a transition provides step-level credit. GraphHCA estimates this potential from pooled rollouts through a discounted recursion on the induced transition graph, which admits a unique fixed point on any directed graph. The resulting step-level signal is combined with the trajectory-level advantage, requiring neither a learned hindsight model nor an extra forward pass and recovering GRPO when the step-level weight is zero. Among all compared baselines, GraphHCA achieves state-of-the-art results on ALFWorld and WebShop at both LLM scales, and on Sokoban with a vision-language agent. For example, on ALFWorld it improves overall success rate by up to 24.6 points over GRPO and by up to 4.7 points over the strongest step-level baseline.
[LG-70] Cross-Rollout Bellm an Closure for Long-Horizon Agent ic Reinforcement Learning
链接: https://arxiv.org/abs/2609.35082
作者: Yangyang Ren,Haodong Zhu,Linlin Yang,Sheng Xu,Peichao Lai,Baochang Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Group-based reinforcement learning such as GRPO trains LLM agents by comparing rollouts sampled for each task, without a learned critic. In long-horizon settings, these rollouts revisit shared anchor states, offering cross-rollout evidence for step-level credit. Ideally, step-level credit should incorporate evidence beyond the realized suffixes observed at an anchor while aggregating alternative continuations according to their empirical frequencies. Visit-local averaging pools realized suffix returns at shared anchors and respects observed frequencies, but does not recursively propagate evidence across rollouts, whereas shortest-path estimators have global reach but allow a rarely observed route to dominate an anchor’s value. We introduce Cross-Rollout Bellman Closure (CRBC), which merges each rollout group into a finite empirical process with absorbing success and failure boundaries and evaluates its behavior-policy Bellman fixed point with one linear solve. This fixed point uses the same empirical action and transition frequencies to propagate evidence through shared anchors and aggregate alternative continuations. Backing up the resulting state values through observed transitions yields action values, whose gain over the corresponding state value provides step-level credit. A corresponding finite-depth family recovers visit-local return averaging at zero depth and converges to the exact closure as depth increases. The normalized closure credit is combined with the trajectory-level group advantage for policy optimization, without additional environment rollouts. Across ALFWorld, WebShop, and Sokoban benchmarks with multiple model scales, CRBC consistently improves final performance and learning efficiency. For example, CRBC outperforms the strongest evaluated baseline by 5.59 percentage points on ALFWorld with Qwen2.5-1.5B-Instruct.
[LG-71] Propagate Then Sharpen: Post-Hoc Refinement of Frozen Node Classifiers
链接: https://arxiv.org/abs/2609.35080
作者: Preben Johnsen Bentdal,Nello Blaser,Xue-Cheng Tai
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We study post-hoc refinement of frozen node classifiers: given only the graph G and class distributions Q predicted by a frozen model, can we improve accuracy without access to node features, model parameters, or gradients? APPNP answers this by propagating logits with a restart towards the initial predictions, minimizing the anchored Dirichlet energy. Instead, we consider the Potts energy, and decompose it into a Dirichlet term, which penalizes disagreement between neighbouring nodes, and a Gini term, which penalizes indecision within each node. This decomposition motivates Propagate, Then Sharpen (PtS), which alternates between propagation of class probabilities and node-wise, mass-preserving sharpening, with only one additional hyperparameter selected using labelled validation nodes. Across nine homophilic graphs, with a frozen MLP backbone, PtS improves mean test accuracy over independently tuned APPNP by 1.71 percentage points on clean inputs and 3.90 under severe Gaussian feature corruption. Gains over APPNP become smaller, but remain positive with frozen GCN and GraphSAGE backbones. Sharpening also removes most of the accuracy loss of deep propagation: on clean inputs without restart, accuracy falls by 2.2 points between 2 and 100 propagation steps under PtS, compared with 33.8 for APPNP.
[LG-72] ReCo: When to Relocate Sensor Kits under Deployment Constraints – A NILM Case Study
链接: https://arxiv.org/abs/2609.35075
作者: Haokun Chen,Yu Tong,Yehai Chen
类目: Machine Learning (cs.LG)
*备注: 8 pages, 3 figures, 6 tables
Abstract:Many sensing tasks obtain training labels only by deploying instruments in the field. With a limited number of sensor kits, a collection deadline, and measurement downtime at every move, the collector must repeatedly decide whether to stay at the current site or relocate. We study this decision in non-intrusive load monitoring (NILM), which estimates the power drawn by individual appliances from a home’s main meter and is trained on data from homes temporarily fitted with appliance-level sub-meters. In NILM, appliance usage varies with the appliance, season and climate, and the value of new data depends on how diverse the combinations of target operation and background load are. To address this, we propose a constraint-based relocation framework and instantiate it for NILM as ReCo (Relocation by Coverage gain). ReCo counts new operating regimes in a joint target-background feature space, forecasts each home’s future gain from the data collected so far, and each night weighs the gain of staying against the gain of moving elsewhere after the downtime. In replayed deployments on the Plegma dataset under two kit counts and two downtime costs, ReCo outperforms fixed-dwell and count-based schedules and a threshold rule using the same metric in every setting. Its advantage is not explained by collecting more days alone and reflects allocating the days to more valuable homes and periods.
[LG-73] Interrelating Fruchterman-Reingold Graph Visualization and Agglomerative Clustering
链接: https://arxiv.org/abs/2609.35073
作者: Alexandre Benatti,Luciano da F. Costa
类目: Machine Learning (cs.LG)
*备注: 10 pages and 7 figures
Abstract:Graph visualization methods and agglomerative clustering have been frequently considered in data analysis and pattern recognition. Because these approaches are interrelated and complementary, it is of particular interest to investigate their associations. In this work, we study the possible relationship between the Fruchterman-Reingold graph visualization method and four types of agglomerative clustering adopting single- and complete-linkage, average, and Ward’s linkage criteria. Three types of datasets have been considered in 2 and 10 dimensions, as well as the PCA projection of the latter to two dimensions. The results obtained suggest that the relationship between the methods considered did not vary much for the three types of data mentioned above. At the same time, the agglomerative methods tended to yield results that are mostly similar to each other, while presenting moderate similarity with the original data. The Fruchterman-Reingold visualization resulted similar to the original data, but exhibited relatively smaller similarity to the agglomerative methods.
[LG-74] Not All Rollouts Are Worth Learning: On Trajectory Valuation for Post-Training Reinforcement Learning
链接: https://arxiv.org/abs/2609.35072
作者: Xuesong Jia,Ziao Yang,Zhanhe Huang,Hongfu Liu
类目: Machine Learning (cs.LG)
*备注:
Abstract:We consider the problem of trajectory valuation in reinforcement learning: how to identify and mitigate detrimental trajectories during online training. Unlike classification, where data valuation relies on fixed training and validation sets, reinforcement learning involves dynamically generated trajectories without explicit validation signals, making conventional influence-based methods inapplicable. We propose Dynamic Trajectory Valuation (DTV), a simple and efficient framework that estimates trajectory utility at the mini-batch level and filters detrimental trajectories based solely on gradient information. By operating at the optimization level, DTV integrates seamlessly with existing reinforcement learning pipelines with minimal overhead. Extensive experiments across diverse settings, including PPO, GRPO, and DPO, demonstrate that DTV consistently improves performance, enhances data efficiency, and stabilizes optimization.
[LG-75] Depot-Closed Multi-Component Construction for Neural Vehicle Routing
链接: https://arxiv.org/abs/2609.35066
作者: Shinichiro Hamada,Hisashi Kashima
类目: Machine Learning (cs.LG)
*备注:
Abstract:Most neural constructive solvers for the vehicle routing problem (VRP) use route-by-route construction, extending one route until completion before starting the next. This commits route membership early and hinders global coordination across routes. We propose multi-component construction, which maintains many route components simultaneously and merges them in an arbitrary order. This removes the depot-return cue that route-by-route construction obtains from the remaining capacity; to compensate, we introduce an interpretation in which every component is treated as an implicitly depot-closed route. Under this depot-closed interpretation, every intermediate state of standard CVRP construction is a complete feasible solution, and the exact cost reduction of a merge is the Clarke-Wright saving. The neural policy combines this CW-saving signal with the evolving component state to learn what to connect and when to connect. A policy trained only on CVRP100 outperforms the reported results of representative neural solvers on CVRP100-500 with greedy inference and, reused for ruin-and-reconstruct, performs strongly at all evaluated sizes up to CVRP1000. In a zero-shot Constraint Tightness evaluation with capacities from C=10 to 500 , it outperforms the reported neural solvers at every capacity. Controlled analyses show that robustness persists without CW grounding and point to learned route-closing behavior as a plausible contributor to the tight-regime degradation of learned route-by-route solvers.
[LG-76] Beyond Gradient Flow: Identifiability and Recovery from Distribution Snapshots
链接: https://arxiv.org/abs/2609.35060
作者: Nam D. Nguyen,Valeriya Malysheva
类目: Machine Learning (cs.LG)
*备注:
Abstract:Inferring dynamics from snapshots of evolving distributions is fundamentally underdetermined: the Fokker-Planck equation constrains the drift F only through its score-weighted divergence \nabla\cdot F+F\cdot\nabla\log\rho , leaving a \rho -solenoidal gauge invisible to any single-time constraint. Time-indexed transport formulations cannot resolve this ambiguity: every admissible marginal path admits a curl-free explanation, minimum-action reconstruction selects it, and marginal fit alone cannot distinguish dynamically inequivalent explanations. Requiring one autonomous field to explain several marginals instead makes part of the hidden circulation visible as \nabla\log\rho changes across marginals. Separating instantaneous Fokker-Planck source constraints from the snapshot experiment, we show that the source constraints identify the field modulo the kernel of a stacked score-weighted divergence operator. For generic Gaussian shape variation, source constraints at K\ge m time points in intrinsic dimension m eliminate every polynomial gauge direction, whereas finitely many density snapshots alone admit aliasing; we give the obstruction explicitly. At a Gaussian anchor, for Sobolev smoothness s and n samples per time point, we derive a conditional lower rate (nK)^-2s/(2s+m+1) for the tangent snapshot experiment, with a matching upper rate in a degreewise benchmark. Strong-form fitting is non-orthogonal to score error and cannot be repaired by spectral filtering. Instead, we estimate using smooth test functions while retaining the known diffusion term, and derive a finite-sample bound that separates sampling error from fixed-grid quadrature bias. Planted-circulation experiments confirm the predicted gauge contraction and expose a design tension between cross-slice information and covariance-aware whitening.
[LG-77] Universality and Generalization of Causal Transformers Across Context Lengths
链接: https://arxiv.org/abs/2609.35055
作者: Takashi Furuya,Maarten V. de Hoop,Gabriel Peyré
类目: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
*备注:
Abstract:Long contexts are central to modern transformer systems, but most expressivity results choose a different network for each fixed sequence length. We study whether one masked transformer can approximate causal token-to-token maps uniformly over sequences of arbitrary length sampling a fixed normalized horizon. To relate sampling resolutions, we model tokens by \alpha -Hölder sequences or, more generally, a common modulus of continuity. Our notion of continuity across resolutions characterizes the causal families admitting uniform approximation on these compact input classes by a single transformer with length-independent parameters. The result extends to the infinite-length mean-field limit, where tokens form continuous curves and masked attention becomes a causal time integral. For bounded regression with target maps satisfying a \beta -smooth stability condition defined using regular test functions, quantitative approximation yields a generalization bound: exact empirical risk minimization over suitably sized bounded-weight transformers gives root mean-square prediction error O((\log\log N/\log N)^\beta/(d+2)) from N iid labeled sequences. The bound holds at fixed confidence on the same sampling distribution, with d the token dimension and no maximum-length factor. Finally, experiments on physical time series support the Hölder-regular token model at observed scales, with dataset-dependent fitted exponents, whereas text input embeddings provide a contrasting case. Native and dense sampling, shuffled controls, and refinement checks delimit this empirical regularity regime.
[LG-78] Graph-Based Learning for Multi-Horizon Martian Atmospheric Forecasting
链接: https://arxiv.org/abs/2609.35042
作者: Gary Myler,James Holmes,Manish Patel,Amel Bennaceur
类目: oftware Engineering (cs.SE); Machine Learning (cs.LG)
*备注:
Abstract:Martian weather forecasting is important for future exploration, but atmospheric behaviour on Mars combines spatial, temporal, vertical, and dust-driven processes in ways that challenge current modelling and forecasting approaches. This paper introduces MaGMA (Martian Graph-based Multi-horizon Atmospheric Forecasting), a graph-based data engineering framework that transforms OpenMARS reanalysis fields into structured learning objects for Martian atmospheric forecasting. Local atmospheric patches are represented as graph nodes and linked through spatial neighbourhoods, temporal continuity, longer temporal dependencies, and dynamically similar atmospheric states. The model integrates recent atmospheric history, engineered physical descriptors, and vertical atmospheric information to support forecasting across multiple horizons. We evaluate MaGMA across five unseen Martian years, including regular years and a global dust storm year. In regular years, the model achieves overall R^2 values of approximately 0.73-0.85. For dust-column forecasting, it outperforms classical and deep temporal baselines in most year-horizon comparisons. During the global dust storm year, dust-column prediction remains strong at shorter horizons, with R^2 above 0.8 for the first two horizons, while broader multivariate performance declines. The results show that graph-based data engineering can create reusable and diagnostically useful representations for planetary atmospheric forecasting, while highlighting the need for better learning under rare extreme regimes and improved use of vertical atmospheric structure.
[LG-79] HEIA: A Multimodal Dataset and Benchmark for Vision-Language Analysis of Layout NEURIPS2026
链接: https://arxiv.org/abs/2609.35035
作者: Giuseppe Chiari,Michele Piccoli,Federico Viola,Davide Zoni
类目: Machine Learning (cs.LG)
*备注: 10 pages, 10 figures, 14 tables, to be published in NeurIPS 2026
Abstract:The integration of artificial intelligence into computer-aided design frameworks has sparked a shift in the design of analog integrated circuits (ICs), transitioning the field from using manual and algorithmic-based solutions to adopting automated and intelligent paradigms. In this scenario, the GDSII file represents the industry-standard database containing the ultimate and most accurate source of information of the analog circuit, encapsulating the complex physical geometries and parasitic realities that define tape out performance. This paper proposes THEIA, a novel dataset containing thousands of layout images paired with question-answer conversations, along with a benchmark that employs a fine-tuned vision-language model (VLM) to analyze GDSII files of analog circuits, enabling designers to interact with and query physical layouts as intuitive, meaningful entities. Experimental results using thousands of analog designs across five realistic tasks demonstrate that the proposed fine-tuned VLM outperforms state-of-the-art general-purpose VLMs by a significant margin (up to 73%), highlighting a fundamental gap between general-purpose multimodal reasoning and domain-specific layout understanding.
[LG-80] Fast Learning Rate Transfer in Shallow Linear Networks at Growing Training Horizons
链接: https://arxiv.org/abs/2609.35029
作者: Mana Sakai,Masaaki Imaizumi
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Hyperparameter transfer across model width can substantially reduce the cost of tuning large neural networks, but its behavior when the training horizon grows with width is not fully understood. Building on the framework of fast hyperparameter transfer (Ghosh et al., 2026), which formalizes when transfer is effective, we investigate conditions that ensure fast transfer in the growing-horizon regime. Specifically, we study learning-rate transfer in a shallow linear network with a single trainable hidden matrix, trained by full-batch gradient descent. Under additional spectral assumptions, our main results are threefold. (i) We prove fast learning-rate transfer as n,T\to\infty whenever T=o(\sqrtn) . (ii) We characterize the transfer rates through the finite-width perturbation scale, the first-order sensitivities of the loss and its learning-rate derivative to finite-width perturbations, and the local loss curvature. (iii) We derive limiting distributions for the optimal learning rate and optimized loss, governed by fluctuations associated with the extreme eigenvalues of the data Gram matrix. These results clarify how spectral structure and local loss sensitivities govern learning-rate transfer at growing horizons.
[LG-81] Price Stability in the European Union: A Systemic Approach Using Random Matrix Theory
链接: https://arxiv.org/abs/2609.35011
作者: Sami Diaf
类目: Machine Learning (cs.LG); Econometrics (econ.EM); Applications (stat.AP); Methodology (stat.ME)
*备注:
Abstract:Price stability remains a pillar in monetary policy practices and carries a special importance within monetary unions. Mainstream economics tried to leverage price stability using price indices and several metrics to shed light on specific dynamics and optimal macroeconomic levels. The wide availability of data led researchers to consider the study of systems using Random Matrix Theory, based on inner correlation patterns. This aims to enhance the multivariate analysis by removing noisy patterns from the signal and improve data quality for further inferences. This work considers the collection of monthly inflation indices in the Eurozone as a \textitsystem of prices to analyze its eigenvalues’ statistical and asymptotic properties and uncover inner country-level insights. Results confirm the system cannot assumed to be randomly generated, and the data exhibit noise-dominated patterns, due to small and persistent variations at the country-level. The latter make the inter-country correlations more dynamic and the separation of the signal from the noise quiet difficult. Findings identified two countries as distorting inflation dynamics besides three other distinct, regional-based groups of countries. Variability sources might stem from economic episodes fueling inflation spikes in some countries, as well as methodological aspects used to ensure data quality and representativeness in the European Union. Despite being complex, the system demonstrates a certain stability, in terms of self-organization; while large monthly fluctuations cannot be considered as rare events, but part of the data-generating process.
[LG-82] Sub-Model Short-Term Memory Convolutions for Keyword Spotting Systems on Device INTERSPEECH2026
链接: https://arxiv.org/abs/2609.35005
作者: Paweł Warlewski,Artur Czeczko,Artur Szumaczuk,Grzegorz Stefański,Szymon Klimaszewski
类目: ound (cs.SD); Machine Learning (cs.LG)
*备注: Interspeech 2026, 5 pages, 2 figures
Abstract:Keyword Spotting (KWS) is becoming increasingly important as voice-controlled devices grow more widespread. While voice interaction with smartphones and smart TVs is already common, deploying KWS on heavily resource-constrained edge devices such as wearables remains challenging. These systems must meet high accuracy requirements while operating under strict constraints on computational power, memory footprint, and real-time latency. In this work, we present an application of the STMC (Short-Term Memory Convolutions) framework to adapt a modular CNN model for online, LSTM-like inference. Our approach reduces power consumption and redundant computations while maintaining the stability and simplicity of training CNNs. We achieve up to 82% and 46% MCPS reduction compared to equivalently frequent standard CNN execution and vanilla STMC, respectively. The best configuration achieves 93.8% accuracy on the 11-class Google Speech Commands task and 97.1% on the same task with zero-padded data.
[LG-83] MaPP: A Unified Marginalized Posterior-Predictive Framework for Data-Efficient RLVR
链接: https://arxiv.org/abs/2609.34990
作者: Yangyang Ren,Haodong Zhu,Sheng Xu,Yanjing Li,Nikolai Yu. Zolotykh,Wentao Zhang,Baochang Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reinforcement learning with verifiable rewards (RLVR) improves the reasoning capabilities of large language models but incurs substantial costs from rollouts and policy updates. Online prompt selection improves efficiency by using per-prompt Bayesian posteriors to predict difficulty and prioritize informative prompts. However, existing methods overlook how reliably learning signals are extracted from sampled responses. In GRPO, a response’s advantage depends on both its own outcome and the randomly sampled outcomes of its peers through group normalization. Our theoretical and experimental analyses show that uncertainty in group composition introduces composition noise, a non-vanishing variance component that imposes an irreducible lower bound on gradient estimation error and impairs downstream prompt selection. We propose MaPP (Marginalized Posterior-Predictive), a unified framework for data-efficient RLVR that denoises response-level advantage estimation and improves prompt selection using a shared Beta posterior. For each response, MaPP replaces the standard group-relative advantage with a composition-invariant intrinsic advantage through closed-form Beta-Binomial marginalization. The resulting posterior-predictive estimator has an error that provably diminishes as the posterior concentrates. Using the same posterior, MaPP derives an uncertainty-aware prompt selection score to improve data efficiency without additional rollout cost. Experiments on mathematics, planning, and visual geometry across five model backbones show that MaPP consistently outperforms GRPO and strong selection baselines, achieving up to +2.45 average accuracy improvement over the strongest baseline under the same rollout budget and setting a new state of the art.
[LG-84] ach to Learn: Hint Annealing for Self-improving LLM Reasoning
链接: https://arxiv.org/abs/2609.34975
作者: Zile Wang,Zijian Li,Haodong Wang,Jian Liu,Qianli Liu,Lucas Muli,Blaze Chen,Song Guo
类目: Machine Learning (cs.LG)
*备注:
Abstract:Group Relative Policy Optimization (GRPO) improves language-model reasoning by comparing verified rewards among multiple solution rollouts for each query. However, difficult training queries can yield only incorrect rollouts, leaving GRPO with no reward contrast or learning signal. Prior hint-based methods construct auxiliary hints from solution evidence and use them to re-solve failed queries, recovering learning signal. Yet the resulting trajectories are typically treated as ordinary solution trajectories despite being generated under an assisted condition unavailable at evaluation. We discover hinted reward shift: recovered reward contrast can concentrate policy updates on hinted trajectories, limiting improvement without hints. This also creates a trade-off: increasing hinted trajectories can accelerate early learning but intensify reward shift later. To address this problem, we propose HATCH (Hint-Annealed Self-Teaching), an online single-policy framework that learns from both generating and using its own hints to improve reasoning without assistance. To mitigate hinted reward shift, we introduce online weighting to anneal the contribution of hinted trajectories. However, learning to generate hints can conflict with improving query solving. We therefore use gradient projection to remove the opposing component of hint-generation updates. Together, these designs support self-improvement by enabling the policy to create learning opportunities for itself and turn them into stronger reasoning without hints. We evaluate our method on mathematical reasoning benchmarks and outperform state-of-the-art methods by 1.02 pp on Llama-3.2-1B-Instruct, 2.84 pp on Qwen3-1.7B, and 4.32 pp on Qwen3-8B.
[LG-85] ALICE: In-context Zero-shot Mutual Information Estimation
链接: https://arxiv.org/abs/2609.34962
作者: Giulio Franzese,Simone Rossi,Pietro Michiardi
类目: Machine Learning (cs.LG)
*备注:
Abstract:Estimating mutual information (MI) from samples is a central objective in a variety of scientific fields. Modern neural estimators are accurate in the large-data regime, but they fall short when data is scarce, and each must be fit anew for every distribution under study. Current estimators are moreover tied to specific data types. These constraints limit their adoption in many applications where per-distribution training is impractical and sample sizes are small. We present ALICE, a foundation model that removes per-distribution training, while achieving competitive estimation accuracy. Trained exclusively on a broad family of synthetic distributions, ALICE acts as an in-context estimator of rectified-flow velocity fields: conditioned on samples of an unseen distribution, it estimates that distribution’s velocity field without any explicit training. MI is then obtained through a fixed identity that integrates the squared difference between the joint and conditional fields. We validate ALICE on a standard, challenging benchmark and apply it in three domains, biology, genetics, and neuroscience, whose data the model has never seen. For the first time, we show that a single model closes the gap with neural estimators trained separately for each distribution, while natively supporting different data dimensionality and sample cardinality, enabling zero-shot MI analysis across scientific domains. Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.34962 [cs.LG] (or arXiv:2609.34962v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.34962 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-86] XMatch: Enhancing Covariate-Aware Time Series Forecasting through Tree-Structured Exogenous Matching
链接: https://arxiv.org/abs/2609.34939
作者: Ziyang Zhang,Hanyin Cheng,Xiangfei Qiu,Yang Shu,Bin Yang,Chenjuan Guo
类目: Machine Learning (cs.LG)
*备注:
Abstract:Future exogenous variables provide valuable information for forecasting endogenous time series. Existing covariate-aware methods primarily learn the direct influence of exogenous variables on endogenous variables. However, these effects can be complex and change with the pattern of the exogenous variables, making them difficult to capture. Beyond this perspective, we observe that a given exogenous pattern often co-occurs with only a small set of endogenous response patterns. These associations motivate a strategy that matches future and historical exogenous patterns and uses the corresponding endogenous patterns to enhance forecasting. However, in real-world forecasting scenarios with multiple exogenous variables, each exogenous variable provides a distinct dimension for matching, creating a dilemma for this strategy between precise matching and sufficient historical support. To bridge this gap, we propose XMatch (EXogenous MATCHing), a covariate-aware forecasting model that realizes the aforementioned strategy through a tree-structured matching process that adaptively adjusts the number of exogenous variables used as matching conditions. Specifically, we first introduce the ProtoTree Creator, which organizes historical correspondences between exogenous and endogenous patterns into a ProtoTree, whose deeper levels incorporate additional exogenous variables for matching. For forecasting, we then design the ProtoTree Matcher, which uses future exogenous variables to query the ProtoTree and adaptively determines how many exogenous variables to use for matching based on exogenous pattern similarity and historical support. Finally, the matched endogenous patterns are used as explicit historical evidence to enhance forecasting. Extensive experiments on 12 real-world datasets demonstrate that XMatch outperforms state-of-the-art baselines.
[LG-87] Muon Sublates the Edge of Stability in LLM Pretraining
链接: https://arxiv.org/abs/2609.34915
作者: Yanzhe Chen,Qifang Zhao,Xiaoxiao Xu,Fanghui Liu
类目: Machine Learning (cs.LG)
*备注: 30 pages, 19 figures
Abstract:Muon is increasingly used for language-model pretraining, yet its large-step dynamics are not captured by the classical edge-of-stability (EoS) picture of gradient descent (GD). In GD, loss neutrality, equal-magnitude update reversal, and marginal stability meet at a single learning-rate-dependent edge. We show that Muon breaks this coupling. For stochastic no-momentum Muon, we derive a coherence-corrected conditional loss-neutral boundary 2\rho_b/\eta , while temporal alignment follows a separate geometry. Controlled experiments show that loss balance and temporal alignment respond differently to learning rate and batch size. Across our language model experiments, the 130M Llama-like LLM runs exhibit loss-boundary tracking with weak negative alignment, whereas the studied 1B LLM configuration shows stronger partial cancellation; in both settings, directions remain far from coherent reversal while training continues to improve. These results support a split EoS picture for Muon: a stochastic loss-neutral edge survives, but it is not accompanied by a universal temporal-direction signature. The source code for reproducing the experiments can be found in this https URL
[LG-88] Learning High-Risk High-Precision Motion Control
链接: https://arxiv.org/abs/2609.34851
作者: Nam Hee Kim,Markus Kirjonen,Perttu Hämäläinen
类目: Machine Learning (cs.LG); Robotics (cs.RO)
*备注: Project webpage: this https URL
Abstract:Deep reinforcement learning (DRL) algorithms for movement control are typically evaluated and benchmarked on sequential decision tasks where imprecise actions may be corrected with later actions, thus allowing high returns with noisy actions. In contrast, we focus on an under-researched class of high-risk, high-precision motion control problems where actions carry irreversible outcomes, driving sharp peaks and ridges to plague the state-action reward landscape. Using computational pool as a representative example of such problems, we propose and evaluate State-Conditioned Shooting (SCOOT), a novel DRL algorithm that builds on advantage-weighted regression (AWR) with three key modifications: 1) Performing policy optimization only using elite samples, allowing the policy to better latch on to the rare high-reward action samples; 2) Utilizing a mixture-of-experts (MoE) policy, to allow switching between reward landscape modes depending on the state; 3) Adding a distance regularization term and a learning curriculum to encourage exploring diverse strategies before adapting to the most advantageous samples. We showcase our features’ performance in learning physically-based billiard shots demonstrating high action precision and discovering multiple shot strategies for a given ball configuration.
[LG-89] When Sparse Reward Meets Dense Distillation: Training Dynamics of On-Policy Distillation
链接: https://arxiv.org/abs/2609.34849
作者: Xinke Jiang,Tao Feng,Zhibang Yang,Zhixin Zhang,Weixuan Xu,Haoyu Zhang,Xu Chu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reinforcement learning with verifiable rewards provides a sparse post-training signal: a single binary outcome evaluates the entire rollout, and every token receives the same sequence-level advantage regardless of its individual contribution. To complement this sparse supervision, a growing family of methods adds a scalar-weighted teacher KL term to the policy-gradient objective, providing dense token-level guidance that may be unreliable at some positions. Despite the benefits of combining these signals, their interaction during optimization can destabilize joint training. To understand how this instability develops, we study the learning dynamics of hybrid reward–distillation training through a neural tangent kernel (NTK) analysis. We introduce the cross-signal NTK K_DR(n) , a token-level statistic that measures the alignment between reward and distillation gradients at position n. Through this analysis, we identify two failure modes: 1 Magnitude drowning, where the reward gradient exceeds the distillation gradient by orders of magnitude, so that even weak directional conflict can cause the distillation loss to rise despite its explicit inclusion in the training objective; and 2 Localized directional conflict, where the sequence-level advantage and the teacher’s position-specific distribution induce opposing updates at the same token ( K_DR(n)!!0 ). The severity of these effects depends on the optimization regime: the gradient-norm ratio \kappa!=!|\nabla\mathcalL_R|/|\nabla\mathcalL_D| varies by roughly an order of magnitude across tasks, and our experiments reveal an empirical threshold beyond which naive mixing can lead to persistent training collapse. Motivated by these findings, we introduce the M3 family, which combines magnitude normalization with three strategies…
[LG-90] Context-dependent time-series prediction via HyperReservoirs
链接: https://arxiv.org/abs/2609.34847
作者: Kohei Tsuchiyama,Takatomo Mihana,Ryoichi Horisaki,André Röhm
类目: Machine Learning (cs.LG); Chaotic Dynamics (nlin.CD)
*备注:
Abstract:Time series prediction is a common application of reservoir computing. When the training and testing time series data contains multiple dynamical regimes, because an underlying parameter is changing, or the data in fact consists of multiple distinct systems, simple application of the reservoir computing principle produces high prediction errors. Here, we propose a HyperReservoir as an extended model of reservoir computing especially designed for such cases. The HyperReservoir combines a main reservoir with a smaller context reservoir, where the latter modulates the output weights of the former. This structure resembles the hypernetworks from deep neural network literature. However, in contrast, HyperReservoirs retain the simple training via linear regression of standard reservoir computing. We compare the proposed architecture with a conventional ESN, in which context acts at the input, and a full-matrix Conceptor, in which context modulates the reservoir state space. We evaluate all three models on time-series prediction tasks based on Lorenz and Rössler systems, including for varying bifurcation parameters and time sampling scales. We find that the HyperReservoir achieves the lowest mean test error in all three tasks, and particularly outperforms conceptors on data that is sampled from the same attractor but at different time scales.
[LG-91] QiYao-M: Multimodal Time Series Foundation Model with Role-Aware Modeling of Endogenous and Exogenous Modalities
链接: https://arxiv.org/abs/2609.34842
作者: Hanyin Cheng,Linfeng Wang,Zhengbo Qu,Yang Shu,Zhongwen Rao,Meng Wang,Yijie Li,Xin Jiang,Bin Yang,Chenjuan Guo
类目: Machine Learning (cs.LG)
*备注:
Abstract:Existing multimodal time series foundation models (TSFMs) typically model heterogeneous modalities through largely shared mechanisms, overlooking the distinct forecasting roles of endogenous and exogenous modalities. In this work, we propose QiYao-M, a role-aware multimodal TSFM that models the two types of modalities separately. For endogenous modalities, to capture how they evolve along with the underlying temporal dynamics, we introduce an Endo-Multimodal Predictor and Endo-Multimodal Supervision to explicitly learn their evolution from history to the future. For exogenous modalities, to generalize across domains and across various modality types and numbers under the scarcity of exo-multimodal pretraining data, we propose an Exo-Multimodal Retrieval Enhancer that enables rapid downstream adaptation without updating the TSFM parameters. We further introduce Endo-Modality Proxy Training to train this retrieval module without exogenous multimodal pretraining data. Extensive experiments across unimodal and multimodal benchmarks demonstrate strong forecasting performance in scenarios both with and without exogenous modalities.
[LG-92] Structured Neural SDEs for Functional Calibration
链接: https://arxiv.org/abs/2609.34831
作者: Francesco Piatti,Andrea Iannucci,Thomas Cass
类目: Machine Learning (cs.LG); Probability (math.PR)
*备注:
Abstract:Neural Stochastic Differential Equations (Neural SDEs) provide flexible continuous-time generative models, but generic neural drift and diffusion networks are costly to simulate on long horizons and can give unstable gradients when the training signal is a path functional rather than a pointwise observation. We introduce SLiSDE, a family of Neural SDE models built from structured linear stochastic layers. Parallel-in-time simulation is obtained at the layer level, while expressivity is recovered by gated in-flow stacking: previous-layer paths modulate the next layer’s latent flow through learned gates. For functional calibration tasks in which rare paths dominate the loss, we add an optional Girsanov tilt that acts as a learned importance sampler with an exact likelihood-ratio correction. We prove well-posedness, a discretisation error bound, validity of the change of measure, and a universality result: the terminal laws of the gated stack are dense in the space of square-integrable laws. Experiments on functional calibration benchmarks show that the structured model outperforms fully neural SDE baselines while retaining parallel-time simulation and stable importance weights.
[LG-93] Accelerator Choice Is Not Enough: AlphaFold2 Inference on Cloud TPUs
链接: https://arxiv.org/abs/2609.34818
作者: Lorenzo Pazienza,Ihab El Bani
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG); Performance (cs.PF)
*备注: 22 pages, 5 figures, 4 tables. Code and data: this https URL
Abstract:AlphaFold2 is written in JAX, so the same inference code compiles and runs unchanged on CPUs, GPUs and Google Cloud TPUs. That portability makes the accelerator look like the main decision a user has to make. We show that it is not. Running one AlphaFold2 inference workload across a Colab CPU runtime, an NVIDIA T4 GPU and a dedicated eight-chip Cloud TPU v5e slice, we find a large hardware advantage for the TPU, 0.47 s per call in steady state on a single chip against 13.1 s on the T4 in the same measurement campaign, and three ways in which the software layer decides how much of it a user actually gets. The default execution path uses one chip of the eight, and at list prices the idle capacity makes the slice cost about as much per prediction as the GPU. Batching with this http URL never exceeds single-query throughput, while mapping queries across chips with this http URL gives eight chips 6.5-7.9x the throughput of one on a matched grid; automatic sharding leaves the per-chip footprint unchanged, consistent with replication, most plausibly because AlphaFold2 carries no sharding annotations. Our retained trace analysis of a first call at a new input shape reports about three quarters of the traced span in JAX tracing and compilation rather than execution. Reruns five weeks later reproduced neither cloud baseline, the GPU one off by roughly a factor of two, so the hardware ratio above is specific to one campaign.
[LG-94] Polylogarithmic Nash Regret in Matrix Games with Bandit Feedback
链接: https://arxiv.org/abs/2609.34812
作者: Yuheng Zhang
类目: Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT)
*备注:
Abstract:We study Nash regret minimization in unknown finite matrix games with bandit payoff feedback and observed opponent actions. We develop Optimistic Payoff Balancing (OPB), which achieves instance-dependent \mathcalO(\log^2 T) Nash regret against arbitrary adaptive opponents, including games with nonunique equilibria. This resolves the open problem posed by Maiti et al. (2025), extending their polylogarithmic guarantee under bandit feedback from 2\times2 games to arbitrary finite dimensions. To handle nonunique equilibria, we construct a reference strategy that leaves room for local adjustments. We order independent payoff differences by estimation accuracy and scale these adjustments by uncertainty, allowing the learner to exploit the opponent’s imbalance to offset estimation costs. Our result thus shows that observing opponent actions suffices for polylogarithmic Nash regret in general finite matrix games.
[LG-95] On Temporal Binding in Large Audio Language Models ICASSP2027
链接: https://arxiv.org/abs/2609.34806
作者: Paul Primus,Gerhard Widmer
类目: ound (cs.SD); Machine Learning (cs.LG)
*备注: Repository: this https URL This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible
Abstract:Reasoning about temporal structure of audio recordings requires Large Audio Language Models (LALMs) to associate sound events with their temporal position. Understanding the underlying mechanisms is a first step toward diagnosing failures and identifying model components that may need improvement. Using mechanistic interpretability, we investigate how temporal information is represented and bound to sound events in three open-source LALMs. We find that across all three, event-specific location becomes concentrated in event name representations at intermediate modality integration layers. These representations encode coarse event position along a low-dimensional, curved relative time trajectory. Steering event name representations along this trajectory systematically shifts before/after beliefs, providing evidence that these representations contribute to coarse temporal reasoning. In contrast, the same interventions do not reliably shift predicted onset timestamps, suggesting that coarse temporal reasoning and precise metric event localization rely on distinct mechanisms.
[LG-96] Physics-Attested Federated Learning: Securing Collaborative Anomaly Detection in Critical Water Infrastructure
链接: https://arxiv.org/abs/2609.34804
作者: Jeff Nijsse,Shu Su,Benjamin Oholeguy,Sreenivas Sremath Tirumala
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 36 pages, 10 figures
Abstract:Federated learning enables industrial operators to train shared intrusion detection models without disclosing proprietary operational telemetry. However, existing defenses operate strictly in update space, leaving aggregators blind to data poisoning; model updates derived from fabricated telemetry remain indistinguishable from honest contributions. We repurpose cyber-physical process invariants, such as conservation laws and actuator couplings, from runtime detection heuristics into a verifiable admission requirement for federated updates, mined automatically from clean operational data. We evaluate this admission gate across two physical water testbeds (SWaT, WADI) and a distribution benchmark (BATADAL), testing seven aggregation rules against telemetry fabrication, exposure-only replay poisoning, and an invariant-aware adaptive adversary. Across three testbeds the mined invariants reject none of 100 honest shards and all naively fabricated ones, including optimised perturbations that FoolsGold admits in full. On real telemetry, five mined invariants detect 12 of SWaT’s 35 attacks, while nine invariants detect 20, with no honest shard rejected. With nine rules, the physics gate recovers 69–100% of the targeted-attack recall lost to replay poisoning, and 54–100% of that lost to fabricated telemetry, across five standard aggregators. To reconcile physical admission control with federated data privacy, we show invariant compliance using zero-knowledge proofs (zk-SNARKs) to allow clients to prove batch adherence without revealing operational telemetry.
[LG-97] Separating personal from population gains when calibrating EEG foundation models for new users
链接: https://arxiv.org/abs/2609.34801
作者: Xilin Tao,Kani Chen
类目: Machine Learning (cs.LG); Signal Processing (eess.SP); Neurons and Cognition (q-bio.NC)
*备注: 59 pages, 19 figures (6 main, 13 supplementary), 33 tables (3 main, 30 supplementary). Includes supplementary information and source data. Code: this https URL
Abstract:Foundation models are increasingly adapted to individual users, but an apparent personalization gain can simply reflect a stronger population model. This distinction matters for brain-computer interfaces, where every new user must be calibrated. We evaluated personal adaptation of three frozen EEG foundation models (CBraMod, REVE and LaBraM) in 235 held-out subjects from three motor-imagery datasets, comparing each subject’s adapter with the population model and with adapters fitted to other subjects. Using all first-half session labels, personal adapters improved mean balanced accuracy over the population model by 1.5-5.4 percentage points and outperformed exchanged adapters by 2.3-7.3 points in all nine model-dataset combinations. The size of this benefit depended on population training: with four times the original budget, median gains remained positive (1.0-2.0 points) but were smaller for every model, and no population model reached a confirmed plateau. Acquiring the benefit cheaply was unreliable: few-label calibration was consistently non-negative on only one dataset, and in CBraMod neither unlabeled context nor meta-learned initialization outperformed matched controls. Personalization should therefore be evaluated against both a population reference and exchanged parameters, across population-training budgets.
[LG-98] Instance-Adaptive Prompts as Context for Time-Series Foundation Models
链接: https://arxiv.org/abs/2609.34786
作者: Zehao Xiao,Shifeng Xie,Lei Zan,Jianfeng Zhang,Lujia Pan,Ievgen Redko,Malik Tiomoko,Keli Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Longer histories can improve time-series foundation models (TSFMs), but require substantially higher inference cost. We therefore ask whether contextual information can be provided more efficiently through a compact set of learned token embeddings. We introduce PaCTS, which generates a small set of instance-adaptive latent prompts in the form of continuous embedding tokens conditioned on the visible context. These prompts serve as compact context surrogates for frozen TSFMs. PaCTS constructs them from instance-specific global statistics and further refines them with segment-level temporal information, capturing both global characteristics and local temporal variations. The prompt module is jointly trained and deployed across heterogeneous time series with the frozen backbone. Extensive experiments demonstrate the effectiveness of prompts as context, consistently improving forecasting across context lengths and model architectures. With a shorter input context, PaCTS can outperform the same frozen backbone using double context while requiring substantially less inference computation. Compared with weight-space adaptation methods, PaCTS achieves stronger improvements and better out-of-distribution generalization.
[LG-99] Edge-Level Automorphism in GNNs: A Quantitative Framework and Effective Designs For Link Prediction
链接: https://arxiv.org/abs/2609.34729
作者: Chen Shao,Donald Loveland,Tobias Käfer,Danai Koutra
类目: Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT)
*备注: 10 pages, figure 7
Abstract:Graph Neural Networks (GNNs) are effective for learning node and link embeddings through permutation-equivariant aggregation. However, standard GNNs collapse automorphic nodes, i.e., those with identical structural roles (or orbits) into indistinguishable representations, leading to the node automorphism problem. This collapse limits their expressive power and degrades link prediction performance. Existing approaches to characterize GNN expressiveness rely primarily on Weisfeiler-Lehman (WL) analyses, but these methods are typically qualitative and often misaligned with empirical results. To address this gap, we begin by introducing a novel quantitative framework to assess GNN expressiveness for link prediction. We first formalize edge-level automorphism through edge orbits, which capture the set of structural role pairs for nodes that share a link. Then, we introduce the edge automorphism ratio (EAR), a scalar metric that quantifies a GNN’s ability to distinguish links in a given graph. We empirically demonstrate that EAR correlates strongly with performance, validating its practical benefit. Building on this insight, we design EDGE-ORBIT EQUIVARIANT GRAPH NEURAL NETWORK (EO-GNN), a GNN architecture that addresses automorphism collapse while preserving equivariance and incurring minimal computational overhead. EO-GNN accomplishes this through two core designs combined with WL-based node hashes: (i) automorphism-aware dropouts and (ii) subgraph orbit-biased aggregation. Empirical evaluations on synthetic and real graphs show improvements of up to 42.36% and 28.44%, respectively, in predicting links in scenarios with high automorphism.
[LG-100] Learning Propagation Geometry from Message-Passing Feedback
链接: https://arxiv.org/abs/2609.34711
作者: Yingxu Wang,Kunyu Zhang,Xinwang Liu,Mengzhu Wang,Siyang Gao,Chang Tang,Nan Yin
类目: Machine Learning (cs.LG)
*备注:
Abstract:Learning local geometry enables graph neural networks (GNNs) to adapt how they compare and integrate neighborhood information. However, estimating geometry from aggregated representations can overlook variation among individual messages and dependencies across feature dimensions. We propose GeoF, a recurrent framework that jointly evolves node features and propagation geometry through message-passing feedback. Each node maintains a local symmetric positive-definite geometry, initialized from a structure-aware prototype atlas and parameterized in block log-triangular coordinates. At each step, the geometry determines neighborhood weights, while triangular frame transport maps transformed source messages into the target node’s local coordinates before aggregation. Weighted second-order statistics of residuals between aligned messages and the transformed target state capture directional variation and within-block dependencies, yielding a geometric update target. A shared controller learns complementary corrections through task supervision. A bounded log-triangular update combines these corrections, the target, and the previous geometric state while preserving positive definiteness. The geometry governs subsequent propagation, closing the feedback loop. With parameters shared across recurrent steps, task-specific readouts support node classification, link prediction, and graph classification. Experiments on benchmark datasets show that GeoF consistently outperforms state-of-the-art GNN baselines.
[LG-101] Predicting Delayed Train Trajectories on the Dutch Railway Network: Explainable AI Evaluation of Topological Operational and Weather Features with Tree Based Ensemble Methods
链接: https://arxiv.org/abs/2609.34692
作者: Jia Long Bao,Ali Mohammed Mansoor Alsahag,Seyed Sahand Mohammadi Ziabari
类目: Machine Learning (cs.LG)
*备注:
Abstract:The reliable prediction of passenger train delays is a critical component of railway management. While contemporary research frequently attempts to maximize absolute accuracy by deploying opaque deep learning architectures, the underlying data mechanics driving longitudinal predictive decay remain underexplored. Consequently, this study provides an explainable temporal robustness analysis of network-wide railway delay prediction. Focusing on the Dutch railway network, this research utilizes interpretable tree-based ensembles to integrate granular topological, environmental, and operational features. The overarching finding establishes that while feature-rich tree-based models improve simultaneous (within-month) prediction, predictive performance systematically degrades when evaluated across non-simultaneous (future) months. Furthermore, multi-horizon SHAP and dispersion analyses explicitly link this degradation to environmental feature volatility and instability within the statistical target definition. Ultimately, this thesis demonstrates that richer feature sets alone are insufficient to resolve long-term forecasting constraints, underscoring the necessity to transition toward dynamic, season-aware architectures anchored by absolute operational boundaries.
[LG-102] Agent PerfBench: A Benchmarking and Evaluation Suite for Inference Performance of Agent ic LLM s
链接: https://arxiv.org/abs/2609.34683
作者: Cheuk Hang Lau,Zeyu Cao,Kevin Wong Cheuk Yin,Yao Lai,Haoran Wu,Nicholas D. Lane,Robert D. Mullins,Ilia Shumailov,Yiren Zhao
类目: Machine Learning (cs.LG)
*备注:
Abstract:The optimization of LLM serving engines, such as vLLM and SGLang, is largely benchmark-driven: optimizations, scheduling policies, hardware and system designs are all selected based on representative workloads. However, a significant mismatch has emerged in the agentic era. Existing benchmarks primarily focus on simple single-turn chatbot workloads. LLM applications are increasingly agentic: coding agents, terminal execution systems, and tool-use agents issue multi-turn requests with growing context lengths. We introduce AgentPerfBench, a benchmark suite for agentic inference. It uses real traces from agentic benchmarks, such as SWE-Bench and TerminalBench, alongside standard chat baselines. This enables benchmarking of models on multi-turn tasks involving tool calling, skill utilization, and increasing context lengths. AgentPerfBench also samples from empirical distributions of input length, output length, and turn count derived from the real traces, generating representative synthetic profiles for cheap and accurate measurements on new hardware. In addition, we further find that several existing benchmarks fail to accurately reflect real hardware performance for two key reasons: 1) they do not account for realistic context-length growth, and 2) they measure inference performance without operating at hardware saturation. We discuss these issues in detail and provide rich kernel-level Nsight Compute (NCU) traces to construct a new multi-dimensional roofline model that captures hardware-system limitations in both memory bandwidth and memory capacity footprint. The benchmarking suite then includes automated scripts to identify potential bottleneck conditions on emerging hardware when evaluated with diverse agentic traces. Together, these contributions quantify the chat-to-agentic gap in current inference benchmarks and characterise per-kernel GPU resource utilisation via roofline analysis.
[LG-103] SOLAR: A State-Driven Online Learning Rate Scheduler for LLM Pretraining
链接: https://arxiv.org/abs/2609.34681
作者: Qiulin Shang,Binyu Wang,Yongqi Qiao,Songde Rao,Zhoutong Wu,Kun Yuan
类目: Machine Learning (cs.LG)
*备注:
Abstract:Learning-rate (LR) scheduling plays a central role in large language model (LLM) pretraining, yet current practice still relies heavily on hand-crafted heuristics such as Warmup-Cosine-Decay and Warmup-Stable-Decay. Because these schedules are fixed in advance, they cannot adapt to evolving optimization dynamics. Online learned scheduling within the Learning to Optimize (L2O) framework offers a dynamic alternative, but remains brittle at LLM scale due to noisy signals, delayed feedback, and the risk of catastrophic divergence. We propose SOLAR (State-driven Online Learning rAte scheduleR), a stabilized framework for reliable online LR adaptation. SOLAR uses a base schedule as a reference and learns bounded, state-dependent residual corrections for individual parameter groups. Each correction re-anchors to the base at every step, allowing the policy to adapt the LR without relearning the warmup-decay profile. A lightweight state representation and progress-aware reward guide online learning, while a Circuit-Breaker restores training after rare unsafe actions. Across autoregressive language-model pretraining, SOLAR improves final perplexity over tuned static schedules and automatic LR tuners for dense models from 60M to 1B, AdamW and Muon, and two MoE settings up to 3B. Matched 130M controls show that adding base anchoring and action bounds improves a global PPO controller from 27.09 to 23.74 final PPL, while group-wise control reaches 22.87 on the same two seeds. A residual policy trained on a 60M proxy can also be frozen and reused at larger dense scales without target PPO updates, remaining effective across a fourfold base-LR range. These results establish SOLAR as a practical learned LR controller for LLM pretraining.
[LG-104] QuantForge: Discovering Residual Decompositions for MXFP4 Post-Training Quantization
链接: https://arxiv.org/abs/2609.34680
作者: Qiulin Shang,Zhoutong Wu,Jie Hu,Kun Yuan
类目: Machine Learning (cs.LG)
*备注:
Abstract:Four-bit post-training quantization can reduce the memory demands of large language models, but preserving accuracy under strict MXFP4 W4A4 requires coordinating several design choices. Coordinate transforms change block-encoding errors, which in turn affect the residuals propagated through the network. The useful algorithmic decomposition is therefore not fully known before search. LLM-driven program evolution offers a way to explore these choices, but performance scores alone do not explain which design should change next. We introduce QuantForge, a PTQ discovery system that records competing explanations, selects controls that distinguish them, and checks that successor code implements the resulting conclusions. This residual compilation guides program revisions while retaining useful programs even when their original explanations are rejected. Remeasuring the revised program reveals the next error to address. This process discovers HiRes, a fixed MXFP4 quantizer that shapes coordinates, refines legal code assignments, and recovers errors along attention and MLP paths. Each stage acts on residuals measured after the preceding stage has executed. Across seven tasks, HiRes achieves the lowest seven-model Robust Fit (0.09300) and the lowest quantized Fit-7 at 32B. In matched-budget comparisons of LLM-driven program evolution, each with 240 evaluator calls, QuantForge reaches a held-out transfer target in six of eight runs, compared with three each for textual memory and reflection memory, and one for score-only evolution, despite evaluating fewer new programs. These results show that QuantForge improves the discovery of transferable PTQ algorithms by turning controlled evidence into subsequent program changes.
[LG-105] FestDPO: Few-step Generator Alignment with Direct Preference Optimization
链接: https://arxiv.org/abs/2609.34673
作者: Jaewoo Lee,Kyuil Sim,Hyeongyu Kang,Kanghoon Lee,Woocheol Shin,Jinkyoo Park
类目: Machine Learning (cs.LG)
*备注: Preprint. Under Review
Abstract:Few-step generative models can generate high-fidelity samples within a few function evaluations. Despite this efficiency, generated samples may not exhibit desirable properties. When these properties are difficult to encode as an explicit reward function, direct preference optimization (DPO) can align generative models using pairwise preference feedback without training a separate reward model. However, extending DPO to few-step generative models is challenging because few-step generative models are generally implicit, making the likelihood evaluation required by DPO intractable. To address this challenge, we introduce Few-step DPO (FestDPO), an extension of DPO for few-step generative models that leverages nonparametric likelihood estimation from empirical samples. By exploiting the fast sampling capabilities of few-step generative models, our approach makes sample-based approximation of DPO loss computationally feasible. Furthermore, the sample-based formulation makes FestDPO agnostic to the model family and sampling procedure. Our toy experiment demonstrates that FestDPO matches the reward-tilted target distribution across four few-step generators. For real-world tasks, we evaluate FestDPO in two domains: text-to-image generation and protein backbone generation. In text-to-image generation, FestDPO outperforms preference optimization baselines in both win rates against the base models and human evaluation scores. In protein backbone generation, it achieves a higher \beta -sheet fraction and better structural designability than the baselines.
[LG-106] On the Numerical Reliability of Differentiable Physics-Based Optimization for Robotic Material Manipulation IROS2026
链接: https://arxiv.org/abs/2609.34666
作者: Xintong Yang,Minglun Wei,Yu-Kun Lai,Ze Ji
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: Accepted as an oral presentation at the IROS 2026 workshop: Deformable Objects Manipulation: Research Foundations, Reproducibility and Real-World Challenges
Abstract:Differentiable physics is increasingly used in robotic material manipulation for system identification, trajectory or skill optimization, demonstration generation, and robot or end-effector design. These applications depend on gradients propagated through long, contact-rich simulation rollouts. We study the numerical reliability of those gradients using two Material Point Method (MPM) system-identification benchmarks derived from elastoplastic and granular manipulation. The benchmarks provide controlled cases for three effects that also arise in broader differentiable physics-based optimization. GPU many-to-one sums whose order depends on thread scheduling changed long-horizon gradients and reversed the sign of one parameter gradient relative to a deterministic reference. Finite-difference checks became less reliable for longer rollouts because repeated-run loss variation grew much faster than the loss change produced by the tested parameter perturbations. Observation and loss definitions changed optimization behaviour and the solution preferred by an independent metric. These results motivate reproducible accumulation, finite-difference validation that compares perturbation-induced loss changes with repeated-run variation, and explicit reporting of objective construction when differentiable simulation is used for robotic optimization.
[LG-107] Minimax Last-Iterate Convergence in Matrix Games with Observed Actions
链接: https://arxiv.org/abs/2609.34656
作者: Yuheng Zhang
类目: Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT)
*备注:
Abstract:We study last-iterate convergence in unknown two-player zero-sum matrix games with bandit payoff feedback and observed opponent actions. For games with d actions per player, we develop an algorithm achieving a duality gap of \widetilde\mathcalO(\sqrtd/t) with high probability, simultaneously at every round t . This improves the dimension dependence of the best previously known guarantee by a factor of d^3/2 . The rate matches a standard bandit lower bound, establishing minimax optimality in both the number of actions and the number of rounds, up to logarithmic factors. The algorithm is computationally efficient, requiring only \mathcalO(d) time and memory per round. Our technical contribution is a joint design of adaptive averaging and corrected exponential weights that absorbs estimation variance, together with a potential argument that bounds phase durations.
[LG-108] Universal Dynamic Portfolios
链接: https://arxiv.org/abs/2609.34643
作者: Yu-Jie Zhang,Yu-Xiang Wang,Peng Zhao,Kevin Jamieson
类目: Machine Learning (cs.LG)
*备注:
Abstract:Cover’s Universal Portfolio (Cover, 1991) matches the performance of the best constant rebalanced portfolio in hindsight. We generalize this framework to compete with an arbitrary comparator sequence \mathbfu_1,\ldots,\mathbfu_T , leading to a dynamic regret minimization problem for the log loss where existing methods break down due to potentially unbounded gradients. The log loss is exp-concave, a curvature property that classically yields fast rates for static regret, yet we show that this advantage generally disappears in the dynamic setting. In particular, a linear-loss-type \sqrtTP_T dependence is unavoidable, where P_T=\sum_t=2^T\lVert\mathbfu_t-\mathbfu_t-1\rVert_1 is the standard path length. This limitation stems from the coarse nature of P_T , which obscures finer spatial and temporal structure of the comparator sequence. We therefore introduce two structure-aware measures—the Jensen-Shannon distance for spatial structure and the JS ^q -path length for temporal structure—under which faster rates are attainable when the comparator sequence has favorable structure. To achieve sharp bounds for both measures simultaneously, we develop Universal Dynamic Portfolio, a parameter-free method that combines a new Dirichlet Hedge algorithm with a fixed-share update, while retaining a near-optimal P_T guarantee in the worst case. Finally, under an additional bounded-gradient assumption, we show that OPS admits the faster T^1/3P_T^2/3 dynamic regret rate over all comparator sequences. We attain this rate with a tractable proper algorithm that applies more broadly to general online exp-concave optimization over arbitrary compact convex domains.
[LG-109] d Schrödinger Bridge Matching
链接: https://arxiv.org/abs/2609.34642
作者: Sergei Kholkin,Evgeny Burnaev,Alexander Korotin
类目: Machine Learning (cs.LG)
*备注:
Abstract:Schrödinger bridges provide an entropy-regularized framework and a principled solution for unpaired domain translation. In practice, a pretrained bridge may need to be adapted to human preferences or physical constraints through a reward a problem closely related to reward tilting in diffusion models but underexplored for Schrödinger bridges. We introduce Tilted Schrödinger Bridge Matching (TSBM), a post-training method for fine-tuning a learned bridge P between source p_0 and target p_1 toward a reward-tilted target p_1^r\propto p_1e^r , while preserving source p_0 . We formulate this adaptation as alternating optimization initialized from P , provide theoretical justification, and derive a practical algorithm based on Adjoint Matching. We evaluate TSBM on unpaired image-to-image translation targeting digit properties in MNIST and facial attributes in CelebA.
[LG-110] Uniform Race: Parameter-Free Approximate Rejection Sampling
链接: https://arxiv.org/abs/2609.34639
作者: Seiyun Shin,Juhyeong Pang,Kwang-Sung Jun
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 45 pages, 5 figures
Abstract:We study approximate sampling: given N independent samples from a proposal distribution \mu , the goal is to select one whose distribution is close to a target \pi specified only up to a normalizing constant. Block and Polyanskiy (2023) provide finite budget error bounds for approximate rejection sampling (RS) as a function of the acceptance threshold M . The threshold M giving the smallest bound, however, depends on properties of (\pi,\mu) that are typically unavailable from the observed sample. This raises a natural question: Can one attain the best RS guarantee without taking M as input? We answer affirmatively by proposing a parameter-free sampling algorithm called uniform race (UR), based on importance weights, which are ratios of target to proposal probabilities (or densities). It divides each observed weight by an independent uniform random variable to form a score and returns the candidate with the largest score. For every budget N , its total variation error satisfies the RS upper bound for every fixed threshold M simultaneously, thereby achieving the best such bound in hindsight. We also characterize its output distribution conditional on the largest score, identifying when it is exactly the target \pi . Uniform race has no larger total variation error than a natural budget-calibrated RS derived from Rohatgi et al. (2025) and sampling importance resampling (SIR). In particular, we exhibit instances where UR’s error is exponentially smaller in N than that of either baseline. Furthermore, we establish conditions under which attaining this RS guarantee for every (\pi,\mu) uniquely determines the selection probabilities as those of UR. Finally, test-time scaling experiments on LLM math-reasoning tasks corroborate the theoretical comparisons and demonstrate that UR remains competitive in ground-truth accuracy without requiring threshold selection.
[LG-111] Correction-space Cross-variate Interaction for Test-time Adaptation in Time Series Forecasting
链接: https://arxiv.org/abs/2609.34638
作者: Yuanyuan Deng,Mykola Pechenizkiy,Songgaojun Deng
类目: Machine Learning (cs.LG)
*备注:
Abstract:Test-time adaptation (TTA) is a promising paradigm for handling distribution shift in time-series forecasting (TSF), where models adapt at inference time, often leveraging delayed observed data to refine predictions. In the multivariate setting, distribution shifts often exhibit cross-variate dependencies, yet existing TSF-TTA methods adapt each variate independently and ignore this cross-variate structure. Exploiting such structure motivates cross-variate interaction, but coupling variates through backbone predictions introduces direct pathways for mixing uncorrected errors across variates, a concern under the delayed supervision of TSF-TTA. We identify the \emphinteraction space as a key design choice, and show that acting on adapter corrections that refine backbone outputs, the \emphcorrection space, rather than on the predictions themselves, avoids directly propagating backbone errors across variates. We build on this to propose \textscCoRe (\textscCorrection-space Interaction \textscRefinement), realizing correction-space interaction through (i) Shared-anchor Correction Refinement (SCR), which combines each variate’s correction with a shared anchor through a parameter-efficient bottleneck, and (ii) input-conditioned spectral gating, which adaptively modulates the refinement from the current input window. Across seven backbones, six datasets, and four prediction horizons, \textscCoRe reduces MSE by 25.82% on average over backbones and 10.57% over the state-of-the-art TSF-TTA method, with stronger gains at medium-to-long horizons and modest computational overhead. Data and code are available at: this https URL
[LG-112] GenMem: Generative Symbolic Memory for Self-Evolving Harness
链接: https://arxiv.org/abs/2609.34633
作者: Xinke Jiang,Tao Feng,Weixuan Xu,Zhixin Zhang,Zhibang Yang,Wentao Zhang,Runchuan Zhu,Xu Chu,Junfeng Zhao,Yasha Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Long-term memory supports the self-evolution of LLM agents by retaining experience and skills across tasks and enabling their retrieval, reuse, and revision in subsequent long-horizon decision-making. Yet existing memory management approaches remain limited to discriminative retrieval and to address the sparse, hierarchical, and highly redundant structure of reusable experience: only a small, task-dependent subset of trajectories and memories warrants retention, retrieval, or revision. Learning these operations is further complicated by sparse, delayed, and indirect task-level feedback, with weak supervision across the memory lifecycle. Moreover, continual memory evolution introduces an architectural tension as addressing invariance: stored experience is perpetually revised, yet the addressing interface consumed by learned retrieval policies must remain stable. To address, we present GenMem, which reformulates memory management as generative symbolic addressing. Its core mechanism is the Symbolic Identifier (SID), a multi-level discrete token tuple drawn from a Cartesian-product address space that factorizes a million-scale sparse memory space using fewer than one hundred discrete symbols. Instead of generating ever-changing raw content, the memory agent learns to generate SIDs, while memory evolution rewrites the payload at a fixed address without shifting the address itself. Architecturally, GenMem couples a MemRetriever and a MemEvolver within a multi-agent harness, trained via GRPO with dense process and outcome rewards with two-channels optimization. Under offline memory evolution, experiments spanning ALFWorld, WebShop, multi-hop QA, medical reasoning, and deep research evaluate GenMem against strong memory-augmented baselines…
[LG-113] DisKO: Deep Koopman Learning in Distribution Space from Unpaired Snapshots
链接: https://arxiv.org/abs/2609.34629
作者: He Ma,Xiaochen Liu,Wanfeng Lu,Ying Wang,Wei Lin,Qunxi Zhu
类目: Machine Learning (cs.LG); Dynamical Systems (math.DS)
*备注:
Abstract:Many complex systems are observed only through temporally unpaired distribution snapshots, making trajectory-based dynamical learning difficult without additional assumptions. We therefore formulate the problem directly in distribution space, treating the distribution itself as the dynamical state. The challenge is that distribution space is infinite-dimensional, making compact and approximately closed representations difficult to learn from finite snapshots. We introduce DisKO, which extends deep Koopman learning to distribution dynamics by jointly learning predictive distributional observables, a finite-dimensional Koopman representation, and a generative map back to the full distribution. Across seven diverse benchmarks, DisKO achieves state-of-the-art extrapolation performance, with substantially slower error accumulation on long-horizon prediction tasks. DisKO further recovers leading Koopman eigenvalues and eigenfunctions on systems with analytic spectra, revealing meaningful dynamical structure in the learned representation.
[LG-114] CLAD: Constrained Abstract Domain for Neural Network Verification
链接: https://arxiv.org/abs/2609.34628
作者: Hai Duong,Thanh Le,ThanhVu Nguyen
类目: oftware Engineering (cs.SE); Machine Learning (cs.LG)
*备注:
Abstract:Neural network verification (NNV) formally verifies that a network satisfies a specified property for all inputs within a defined region. Modern NNV tools employ abstract domains to compute a sound over-approximation of the network’s behavior from the given input region, thus the tightness of these abstractions essentially determines efficiency. A long line of increasingly precise domains has been developed, but they all describe the valid input region in the same restrictive way, e.g., an Lp-norm ball. A practical input region is rarely a simple Lp ball, but rather a combination Lp ball with additional constraints. Verifying a network over such a region with existing abstraction produces a loose over-approximation, which results in either failing to verify a property or spurious counterexamples. We introduce Constrained Lagrangian Abstract Domain (CLAD), a new abstract domain that computes a sound over-approximation of neural networks over input regions defined by a combination of convex constraints. CLAD propagates these constraints and tightens bounds over the true feasible region. However, bounding a neuron over the intersection of these constraints has no closed-form solution, so CLAD relaxes each constraint into the objective with a Lagrange multiplier and solves the resulting max-min problem with a projected primal-dual method, alternating a projected gradient step on the input with a multiplier update. CLAD supports any convex constraint with a subgradient, e.g., from automatic differentiation. We evaluate CLAD on 1,944 instances across four convolutional networks with motion-blur structured perturbations with halfspace or L2-ball constraints. On standard unconstrained Linf property, CLAD verifies as many instances as GCPCROWN at a similar runtime. On constrained properties, CLAD verifies 60% more instances than GCPCROWN on L2-ball properties, and 22% more in total.
[LG-115] Evaluating Dynamical Fidelity through Predictive Structure in Physical Representations
链接: https://arxiv.org/abs/2609.34627
作者: Oskar Bohn Lassen,Joao Paulo de Souza Boger,Simon Driscoll,Stephen I. Thomson,Sebastian Schemm,Filipe Rodrigues,Francisco C. Pereira
类目: Machine Learning (cs.LG)
*备注:
Abstract:Machine-learning models for physical systems are currently evaluated primarily through errors between predicted and reference states and, increasingly, through tests of physical consistency. These metrics assess whether predictions are accurate and satisfy selected physical requirements, but provide limited insight into whether learned trajectories reproduce the underlying dynamics. Domain experts examine such relationships through physical representations that expose relevant processes, interactions, and responses, but these analyses are often separated from typical machine-learning evaluation. We introduce a practical framework for evaluating dynamical fidelity through predictive structure in physical representation spaces. Experts define the representations, while reference trajectories determine which relationships are predictive and retained as evaluation tests. We demonstrate the approach in atmospheric forecasting using ERA5 representations of planetary-wave activity and Northern Annular Mode evolution, and evaluate Pangu-Weather, GraphCast, and FengWu. The models exhibit distinct departures from reference predictive structure that are not reflected by conventional forecast errors. The framework thereby turns domain-expert representations into systematic tests of learned physical dynamics without prescribing the relationships in advance.
[LG-116] Learning Regional Snow Water Equivalent and Snow Height Variations from Sentinel-1 InSAR Acquisitions
链接: https://arxiv.org/abs/2609.34614
作者: Luca Barco,Lorenzo Innocenti,Bianca Bartoli,Claudio Rossi,Edoardo Arnaudo,Paolo Garza
类目: Machine Learning (cs.LG)
*备注:
Abstract:Managing water resources in mountainous regions depends heavily on reliable Snow Water Equivalent (SWE) and Snow Height (HS) data, yet these variables remain difficult to track at scale. This study evaluates three machine learning architectures (XGBoost, U-Net and SegFormer) for the joint estimation of SWE and HS variations from Sentinel-1 InSAR data over the Italian Alps, using the IT-SNOW reanalysis as reference. SegFormer achieves the best results on both targets, with an MAE of 10.391 cm for HS and 27.113 mm w.e. for SWE and the lowest variability across initializations. A feature sensitivity analysis shows that including all available features does not guarantee the lowest error, with model- and task-specific sensitivities. Spatial metrics (R2, Pearson’s r) separate the three architectures far more clearly than mean error (MAE, RMSE) does, and decomposing the error per window attributes most of it to a systematic offset in the estimated mean variation rather than to the spatial pattern.
[LG-117] Beyond Site Agreement: Re-estimation for Brain Network Generalization
链接: https://arxiv.org/abs/2609.34611
作者: Yingxu Wang,Kunyu Zhang,Yanwu Yang3,Thomas Wolfers,Yujie Wu,Siyang Gao,Nan Yin
类目: Machine Learning (cs.LG)
*备注:
Abstract:Cross-site out-of-distribution (OOD) generalization in resting-state functional magnetic resonance imaging (rs-fMRI) often relies on learning task-discriminative representations from full-scan functional connectivity (FC) graphs and promoting invariance across source sites. However, FC graphs are estimated from finite, temporally correlated blood-oxygen-level-dependent (BOLD) sequences. Cross-site agreement therefore does not necessarily imply that predictive evidence remains supported under FC re-estimation within the same scan. In this paper, we propose Brain Network Re-estimation-Informed OOD Learning (BRIO), a framework that uses within-scan FC re-estimation to guide cross-site alignment. BRIO maps fullscan graphs and their re-estimates into consistently indexed connectome factors, enabling comparisons of their predictive contributions. It assesses re-estimation support from changes in these contributions relative to within-class subject variability and class separation. For each source-site pair and class, this task-calibrated support from both sites is combined with predictive relevance to form pairwise qualifications, which determine relative factor weights and overall alignment strength. Leave-one-site-out experiments on four real-world datasets (ABIDE, REST-metaMDD, SRPBS, and ABCD) show that BRIO consistently outperforms competitive baselines, with relative improvements of up to 3.8% in accuracy. These gains also persist under an alternative brain parcellation on ABIDE.
[LG-118] PMOPD: Task Ordering Cycling and Parameter-Update Subspace Protection in Multi-Teacher On-Policy Distillation
链接: https://arxiv.org/abs/2609.34605
作者: Youzhi Liu,Ruobing Zheng,Boyuan Tong,Tianqi Li,Pingqi Li,Hanbo Bi,Yi Yuan,Jingdong Chen
类目: Machine Learning (cs.LG)
*备注:
Abstract:Multi-teacher on-policy distillation (MOPD) has emerged as a popular post-training paradigm for integrating specialized capabilities in frontier language models. Existing OPD research has primarily focused on optimizing single-task distillation through objective design, distillation scope, and teacher signal construction, whereas MOPD must aggregate multiple capabilities in shared parameters and address the resulting capability seesaw, in which improving one domain suppresses capabilities acquired from another. Inspired by the distinctive update geometry of OPD, we find that parameter updates from different tasks rapidly concentrate in their respective low-dimensional subspaces during MOPD, providing a direct geometric basis for identifying and controlling cross-task interference. We therefore propose PMOPD (Projection-based Multi-Teacher On-Policy Distillation), which constructs subspace memories from the cumulative parameter displacements of different tasks and projects both gradients and optimizer updates to remove components that interfere with protected task directions. We further develop a lightweight conflict probe to characterize task interactions and guide task ordering, together with a cycling strategy that balances subspace estimation and timely task revisitation. Experiments on representative Code, Reason, and Math tasks show that PMOPD improves every evaluated capability over MOPD, raising the average score across the three tasks by 2.54 points on Qwen2.5-7B and 2.09 points on Llama-3.1-8B. These consistent gains establish geometry-aware optimization as an effective and transferable approach to balanced multi-teacher distillation.
[LG-119] Shaping Persistent Representations from Independent Interactions
链接: https://arxiv.org/abs/2609.34604
作者: Ji Dai,Quan Fang,Junyu Gao,Rongfeng Guo,Haoyan Rong,YipingHuang,Yongxi Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:World models learn environment dynamics from interaction experience. These dynamics depend on the current state and actions, as well as on properties that persist across interactions. Yet standard predictive training can reduce error using local evidence alone, without organizing persistent information into reusable context. We introduce SPRII, a training principle that uses relations between interactions as weak supervision for persistent context while retaining the learner’s native objective. For example, different trajectories of the same system share persistent properties even when their states and actions differ. SPRII uses such relations to guide context learning without numerical property labels. Two composable components encourage contexts from related interactions to agree (Align) and use one interaction’s context to predict another’s future (Cross). Our analysis distinguishes three linked questions: what persistent information is accessible in the learned context (Formation), how that context influences a fixed predictor (Use), and whether it reduces task error (Value). Success at one stage does not guarantee success at the next. Controlled experiments show that more reliable relations improve representation organization, but adding a shared-property constraint can reduce access to a property that remains shared. Context substitutions change predictions at fixed model weights, while the benefit from history depends on prediction horizon and readout. Evaluations span thirteen settings, including controlled physical systems, public dynamics tasks, robotic and tactile data, and partner interaction, across multiple learner families. Relative to the corresponding baselines, SPRII yields average gains of over 10% in downstream task performance and over 15% in persistent-property readout. The project page is available at this https URL.
[LG-120] When local gains fail to transfer: Frozen Earth-observation embeddings across wildfires
链接: https://arxiv.org/abs/2609.34602
作者: Philipp Stark,Alexandros Sopasakis,Ola Hall
类目: Machine Learning (cs.LG)
*备注: 18 pages, 5 figures, 8 tables
Abstract:Frozen Earth-observation embeddings are judged almost entirely by spatially blocked cross-validation inside one study region. We show that this number does not predict accuracy in a new region; we show why; and we show the one setting in which such a model does keep working, using a protocol that needs only a linear probe and labels one already has. The testbed is wildfire, with Copernicus burned-area maps of six fires in Greece and Spain and descriptors from the year before each fire, comparing TESSERA and AlphaEarth with ESA WorldCover classes and annual Sentinel-2 index summaries. Inside a fire, the embeddings identify the burned land 0.05 to 0.13 ROC AUC better than the index summaries, and repeated fold allocations, spatial buffers, a block bootstrap, and gradient-boosted trees leave that margin unchanged. On a fire in another region, they lose 0.15 to 0.18 AUC, and the index summaries lose 0.06, so the three end within a few hundredths of each other. The representation is not the cause. Eight labelled blocks from the new region restore the embedding advantage and give a higher AUC than 59,000 labelled pixels from other regions, and the weight vector fitted in one region is nearly orthogonal to the vector fitted in the others, so the part that carries across regions is small and low-dimensional. Forecasting within a region is a different matter. Fitted on a fire that burned in 2023 and applied to a fire twelve kilometres away that burned in 2024, where nothing used postdates the target fire, TESSERA reaches 0.772 AUC and loses 0.04 against a classifier fitted inside the 2024 fire, while classifiers fitted in other regions lose 0.09 to 0.18. A region with one mapped fire can therefore forecast susceptibility for later fires there; a region without one cannot borrow a model from elsewhere, and every evaluation of a frozen embedding should report a held-out region.
[LG-121] he Low-Rank Structure of VLA Reinforcement Learning
链接: https://arxiv.org/abs/2609.34599
作者: Minjae Oh,Yoonah Park,Jongwon Lim,Yohan Jo
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reinforcement learning (RL) is increasingly used to post-train vision-language-action (VLA) models, yet how RL reshapes these policies remains poorly understood. We find that RL across widely used flow-based VLA models, including \pi_0.5 and GR00T~N1.5/N1.6, on LIBERO, ManiSkill, MetaWorld, and CALVIN induces substantially lower-rank parameter updates that are highly concentrated in the action expert’s Timestep Modules, a small and previously overlooked component. Through systematic module-replacement experiments, we further show that these modules capture a disproportionate share of the performance gains from RL. We then characterize what is encoded in these Timestep Modules. First, we show that RL specializes them to the discrete denoising timesteps used during rollouts, and that this discrete-timestep training underlies the low-rank updates. Second, we find that among their outputs, the shift vector changes most distinctly under RL, and through probing, we show that shift update directions strongly predict task success (ROC-AUC up to 99.6% ). Third, we find that the geometry of shift updates reflects task relationships, as their pairwise similarity correlates with cross-task transfer patterns. Building on these findings, we show that steering along shift update directions further improves RL-trained policies without additional RL training. Overall, we provide a systematic understanding of how RL reshapes VLA policies by studying how learned signals are encoded in parameter space, offering insights into more efficient and interpretable VLA post-training.
[LG-122] Deep Weighted Bellm an Residual Minimization for Q* Estimation
链接: https://arxiv.org/abs/2609.34593
作者: Lican Kang,Jerry Zhijian Yang,Cheng Yuan,Chen Zhong
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Off-policy evaluation is a foundational component of offline reinforcement learning, aiming to assess and optimize policy performance using pre-collected datasets. However, such datasets often suffer from pronounced challenges, including distribution shift, Q -value overestimation, and low sample utilization efficiency. To address these issues, this paper introduces a weighted Bellman residual minimization framework that incorporates density ratio weighting by effectively integrating expert demonstrations with behavioral data. The proposed weighting scheme departs from the conventional completeness assumption commonly imposed in the theoretical analysis of deep reinforcement learning. We establish a sharp convergence rate for density ratio estimation and derive the convergence rate for the excess risk of resulting deep Q^* estimator. Extensive empirical evaluations demonstrate that, compared to existing methods, our method achieves significant improvements in numerical performance and policy generalization, providing specific guidance for the rational utilization of expert demonstrations.
[LG-123] Agent Ware: Automating the Lifecycle of Agent ic Applications across the Edge-to-Cloud Continuum
链接: https://arxiv.org/abs/2609.34586
作者: Michalis Kasioulis,Moysis Symeonides,George Pallis,Marios D. Dikaiakos
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: Accepted for publication at the 17th International Conference on Cloud Computing Technology and Science (IEEE CloudCom 2026)
Abstract:Deploying LLM-enabled agentic applications across the Edge-to-Cloud continuum remains challenging due to hardware heterogeneity, deployment complexity, limited observability, and the lack of systematic evaluation methods. Existing solutions address agent development, observability, or benchmarking separately, offering limited support for the full lifecycle of distributed agentic applications. This paper presents AgentWare, an AgenticOps framework that automates the provisioning, deployment, observability, and evaluation of agentic applications across Edge-to-Cloud infrastructures. AgentWare introduces an end-to-end lifecycle pipeline that automatically prepares heterogeneous execution environments, transforms user-defined agent implementations into distributed applications, deploys agent components across the continuum, and performs unified collection of execution traces, infrastructure telemetry, and evaluation metrics. The framework further supports automated semantic evaluation through LLM-as-a-Judge workflows and generates reproducible reports covering correctness, performance, resource utilization, and energy consumption. We demonstrate the applicability of AgentWare through a distributed book assistant agent deployed across real Edge-to-Cloud infrastructure under multiple deployment and model configurations. The results show that AgentWare enables systematic experimentation and evaluation of distributed agentic applications while significantly reducing the manual effort required for deployment, instrumentation, and analysis.
[LG-124] Compute Time Scaling with Recursive Models for Combinatorial Optimization
链接: https://arxiv.org/abs/2609.34585
作者: Zhengxi Zhang,Paul Swoboda
类目: Machine Learning (cs.LG)
*备注:
Abstract:We propose Tiny Recursive Models for Combinatorial Optimization (\ours), a general neural method for combinatorial optimization that scales both depth (how often we recursively invoke our network) and width (how much we sample in parallel). Both are fundamental for combinatorial optimization: hard instances demand a large amount of compute, while a small network is essential to avoid overfitting and capture the algorithmic essence of optimization. In particular, our method consists of a graph-aware tiny recursive model that iterates on a latent state with adaptive halting and needs only a lightweight problem-specific decoder. Compared with previous heatmap-based general neural solvers, it achieves a better balance between solution quality and inference speed on both the Traveling Salesman Problem~(TSP) and the Maximum Independent Set~(MIS) problem, and remains competitive with hybrid methods that combine neural components with heuristics specific to each problem. With the same backbone architecture for both tasks, \ours outperforms every diffusion-based solver on TSP from 500 to 10,000 cities at a lower inference cost, and on the standard Erdős–Rényi-[700-800] MIS benchmark it surpasses all neural solvers except those that only work well on MIS. We then explore self-relabeling for self-supervised training. We periodically replace the current set of training labels with the model’s own better solutions, as an alternative training signal. Self-relabeling can, while forgoing supervision from near-optimal solutions, still result in on-par quality.
[LG-125] Retracing Hodgkin and Huxley: State Recovery Does Not Certify Mechanism
链接: https://arxiv.org/abs/2609.34566
作者: Peiyu Zang,Jiayi Hao,Yongqiang Cai
类目: Machine Learning (cs.LG)
*备注: 22 pages, 7 figures. Code available at this https URL
Abstract:Predicting observed dynamics does not establish recovery of the underlying physical mechanism. Can machine learning retrace the hidden-state reasoning behind the Hodgkin-Huxley (HH) model? We train structured latent models on simulated current and voltage, withholding gate identities and trajectories from training and model selection. We then test response prediction, state recovery, protocol transfer, and agreement with HH dynamics. Prediction error and its cross-seed spread both drop sharply at three latent dimensions under the tested protocols, while gate recovery under new protocols improves through five to six coordinates. State recovery depends on which observations the chart uses. Observed voltage improves current-clamp decoding relative to freely predicted voltage. Under voltage clamp, adding latent state to command voltage raises m-state R^2 from 0.976 to above 0.99, yet the transported field disagrees with HH on identical smooth samples. Known invertible HH coordinates achieve high fast-m field agreement under the same audit procedure. An exact HH identity decomposes the discrepancy into time-scale-weighted state error and a residual in the transported field; these terms can cancel or reinforce. These findings concern the tested models and charts. They support evaluating state and dynamics recovery separately, including chart inputs and transported-field agreement across interventions.
[LG-126] Single-Layer MeMo as a Randomized Hamming-Kernel Classifier
链接: https://arxiv.org/abs/2609.34562
作者: Alessandro Straziota
类目: Machine Learning (cs.LG)
*备注:
Abstract:MeMo (Zanzotto et al., 2025) is a recent language-model architecture that stores associations between token contexts and next tokens in a correlation matrix memory. In this work, we study its single-layer form and show that its ideal retrieval rule is a multiclass classifier based on the positional Hamming kernel. The MeMo architecture represents both the sequence features and the output labels with Gaussian random codes. Its score is therefore a doubly randomized sketch of the ideal classifier. Under independent input and output codebooks, we bound the errors introduced by context sketching and output decoding, characterize their dependence on model and data parameters, and give a margin-based guarantee for recovering the ideal prediction. Controlled simulations support the trends predicted by the analysis. On a restricted WikiText-2 next-token task, we compare single-layer MeMo with classical baselines and show that it can offer a useful trade-off among predictive accuracy, memory, and throughput, particularly on a GPU, where its matrix operations can be parallelized.
[LG-127] Brain-Conditioned Action Policies for Neural Motor Decoding
链接: https://arxiv.org/abs/2609.34561
作者: Luyao Jin,Running Zhao,Huan Zhao,Vincent C. K. Cheung,Wei-Hsin Liao
类目: Machine Learning (cs.LG)
*备注:
Abstract:Motor brain-computer interfaces (BCIs) aim to decode motor intention, enabling people with paralysis to control external devices. Neural motor decoding typically learns task-specific mappings from neural activity to kinematics, yet remains constrained by scarce paired neural-action data. We propose BrainVLA, a framework that enables neural motor decoding by drawing on a pretrained vision-language-action (VLA) model through language-mediated alignment. BrainVLA mitigates reliance on scarce paired neural-action data by leveraging VLA policies. We first construct VLA-compatible datasets including paired neural activity, action signals, language instructions, and rendered visual observations. Then, we adapt the OpenVLA-OFT policy to the target action spaces through LoRA fine-tuning. To establish an effective interface through which neural activity can convey motor intention to adapted VLA policies and guide action generation, we train a neural encoder via neural-language alignment, using language representations as semantic targets to capture latent motor intent from neural activity. The resulting neural representations serve as an endogenous intention signal to guide VLA policies to generate executable actions, while visual observations provide complementary information about the evolving task state. BrainVLA is evaluated on two neural motor datasets with different action dimensionalities using causal rollout decoding. It outperforms the evaluated baselines in cross-session decoding R^2 and task success rate, while demonstrating high training data efficiency. These results establish a route for neural motor decoding to draw on large-scale robotic priors through brain-conditioned VLA policies.
[LG-128] PulseInfer: I/O-Centric Sparse KV Cache Offloading for Efficient Long-Context LLM Decoding
链接: https://arxiv.org/abs/2609.34555
作者: Qiuyang Zhang,Kai Zhou,Kai Lu,Haocheng Lu,Jian Zhou,Yuanpeng Su,Kun Bao,Jiguang Wan,Fei Wu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Long-context LLM serving is increasingly bottlenecked by decode, where large KV caches limit batch size and underutilize GPUs. Sparse KV cache offloading expands effective capacity by storing most historical KV blocks in CPU DRAM and recalling only selected blocks on demand. However, we find that existing offloading systems shift the bottleneck to CPU-GPU recall I/O: recall volume varies widely across layers, decode steps and requests, while headwise sparse selection fragments recalls into many small PCIe transfers. This paper presents PulseInfer, an I/O-centric sparse KV cache offloading system. PulseInfer hides variable recall latency with interruptible layer-wise scheduling, adapts offloading decisions with IO-Adaptive Offloading Admission, and coalesces fragmented transfers using SoloHead sparse selection and a gather-scatter I/O engine. Implemented on SGLang, PulseInfer improves decode throughput by up to 4.7x over SGLang and 2.6x over the best existing offloading baseline, while reducing TPOT by up to 76% and preserving near-lossless accuracy. Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.34555 [cs.LG] (or arXiv:2609.34555v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.34555 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-129] Verifying Neural Networks with Reinforcement Learning NEURIPS2026
链接: https://arxiv.org/abs/2609.34553
作者: Hai Duong,Thanh Le,ThanhVu Nguyen
类目: Machine Learning (cs.LG); Software Engineering (cs.SE)
*备注: accepted at NeurIPS 2026
Abstract:Formal verification can play a key role in ensuring the reliability of Deep Neural Networks (DNNs) deployed in safety-critical systems. Modern DNN verifiers employ a branch-and-bound framework, which alternates between branching (splitting into smaller subproblems) and bounding (pruning subproblems) to efficiently explore the verification space. However, existing branching heuristics make greedy decisions based on static scoring functions. They do not anticipate long-term efficiency or leverage the growing availability of verification data to improve performance. This work introduces RSB, a reinforcement learning framework that learns to refine baseline branching heuristics. It trains an actor-critic architecture to maximize cumulative future rewards rather than immediate scores. The actor generates attention weights from observations of raw neuron features and learned graph embeddings, which rescale baseline heuristic scores to guide neuron branching. Evaluation on 600 challenging instances demonstrates that RSB consistently outperforms state-of-the-art branching heuristics, solving 11% more instances while reducing branch exploration by 50%.
[LG-130] On Parameter Symmetries and Conservation Laws in Gradient Flow
链接: https://arxiv.org/abs/2609.34549
作者: Khang Nguyen,Guido Montúfar
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Parameter space symmetries and conservation laws play an important role in understanding the loss landscapes and implicit biases of neural networks. Inspired by Noether’s theorem in physics, prior works have sought to derive conservation laws under gradient flow from parameter symmetries, but the scope and limitations of this connection remain unclear. We develop a unified geometric framework that clarifies the precise relationship between the two notions, including the conditions under which symmetries correspond to conservation laws. We introduce a notion of compositional identifiability and use it to establish a general inheritance principle for complete characterizations of symmetries and conservation laws in multilayer networks. We apply the framework to multi-head and grouped-query attention, polynomial neural networks, and square deep linear networks.
[LG-131] HALO: Enhancing Time Series Generation via Hyperspherical Latents and Masked AutoregRessive Modeling
链接: https://arxiv.org/abs/2609.34511
作者: Chunyi Hou,Xiangfei Qiu,Hanyin Cheng,Yutong Li,Bin Yang
类目: Machine Learning (cs.LG)
*备注: 19 pages, 7 figures
Abstract:Most existing time series generators rely on a two-stage modeling paradigm: the first stage learns discrete latent representations of time series; the second stage performs autoregressive modeling on these discrete latents through next token prediction. However, this paradigm suffers from two stage-specific limitations: the first stage can lead to information loss when discretizing continuous time series, while the second stage is prone to error accumulation during autoregressive generation. To address these limitations, our core idea is to perform generative modeling in a continuous latent space with a more efficient autoregressive framework. We propose HALO, which enhances time series generation via Hyperspherical Latents and Masked Autoregressive modeling to achieve this goal by tackling two key bottlenecks: (1) variance and scale heterogeneity of continuous latent representations; (2) the difficulty of balancing generation efficiency with temporal correlation modeling. HALO first introduces a hyperspherical VAE that constrains continuous latents to a fixed-radius hyperspherical shell, effectively stabilizing the numerical fluctuations of continuous latent representations. Secondly, we develop a masked autoregressive model that balances parallel decoding and temporal correlation learning, substantially reducing the number of inference steps required for generation and improving generation stability. Our extensive experiments demonstrate that HALO achieves state-of-the-art generation performance while offering significantly improved inference efficiency over existing advanced baselines.
[LG-132] Distribution-Conditioned Task Routing for Class-Incremental Learning
链接: https://arxiv.org/abs/2609.34503
作者: Longhuan Xu,Zhipeng Zhou,Wei Ji,Chunyan Miao,Peilin Zhao,Lijun Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Parameter-efficient adaptation enables continual learners to acquire task-specific knowledge through compact model updates while maintaining strong within-task performance. However, class-incremental inference requires each input to be classified among all classes seen so far without access to its task identity. For learners equipped with task-specific parameter-efficient modules, this introduces a critical task-routing challenge beyond catastrophic forgetting. We study post-hoc task routing without retraining the learner or introducing a separately trained router. Such training-free inference-time calibration remains comparatively underexplored in parameter-efficient class-incremental learning. We identify three sources of routing error (feature-level, task-level, and class-level misalignment) and propose Feature Distribution Calibration (FDC). Its three components address these misalignments: Task Subspace Filtering (TSF) suppresses feature components outside each task’s principal subspace, Residual Likelihood Calibration (RLC) evaluates the typicality of its subspace residual, and Prototype Affinity Calibration (PAC) measures compatibility with the task’s class prototypes. Experiments demonstrate plug-and-play applicability to eight parameter-efficient class-incremental methods using a shared encoder. With one component configuration selected per method across all five benchmarks, FDC improves final accuracy in all 40 method-dataset pairs by 4.39 percentage points on average. Enabling all components improves 35 of the 40 pairs, with an average gain of 4.45 points. When applied to a simple baseline, FDC achieves strong overall performance.
[LG-133] QAMM: Adjoint MeanFlow Matching for Few-Step Offline Reinforcement Learning
链接: https://arxiv.org/abs/2609.34497
作者: Yuehu Gong,Shutong Ding,Mokai Pan,Yimiao Zhou,Jiashu Hou,Ye Shi,Yanwei Fu
类目: Machine Learning (cs.LG)
*备注: 11 pages, 4 figures
Abstract:Flow policies can model rich action distributions, but their iterative sampling limits decision speed. Adjoint matching uses the critic’s action gradient to improve a flow policy without backpropagating through its sampling trajectory, yet its supervision is defined for instantaneous velocities. We propose QAMM, a method that turns the critic-derived adjoint signal into supervision for MeanFlow’s average velocity. The resulting policy learns finite-interval transport directly and generates actions with few network evaluations. We derive the adjoint MeanFlow target, specify its gradient boundaries, and train it with an offline actor-critic. On ten HumanoidMaze tasks, QAMM produces effective two-call policies and achieves competitive performance against strong flow-policy baselines. These results show that adjoint-based Q optimization can be combined with average-velocity learning to obtain expressive offline policies with few-step action generation.
[LG-134] LLN: Learnable Lens Networks for Parameter-Efficient Long-Horizon Dynamical Prediction
链接: https://arxiv.org/abs/2609.34493
作者: Binbin Yong,Zhao Su,Lan Guo,Haoran Li,Jun Shen,Qingguo Zhou
类目: Machine Learning (cs.LG)
*备注:
Abstract:Explicit residual connections of the form (x+f(x)), often combined with normalization layers, have become a standard strategy for training very deep neural networks. However, residual addition primarily provides an algebraic shortcut for gradient propagation, while leaving the evolution of feature geometry across layers largely unconstrained. We introduce Learnable Lens Networks (LLN), a physics-inspired architecture that replaces direct feature-space residual accumulation with learnable optical transport in an augmented position-angle phase space. Each layer alternates between free propagation, which provides an implicit transport path, and a learnable lens field that performs nonlinear trajectory transformation and focusing. Theoretically, we establish that LLN transport is globally invertible and volume-preserving for any differentiable lens field, with the implemented coordinate-wise Gaussian transport further satisfying symplecticity. Importantly, these structural constraints do not limit expressivity: with unrestricted embeddings and readouts, LLN retain universal approximation of continuous end-to-end maps. Experiments across diverse dynamical systems demonstrate that LLN improves long-horizon prediction while using substantially fewer parameters than same-depth comparators. Further analysis reveals stable depth-wise gradient transport and interpretable learned dynamics under the coupled propagation and refraction design.
[LG-135] M3OS: A Monte Carlo Graph Search-Orchestrated Multi-Agent LLM System for Evidence-Traced Molecular Optimization
链接: https://arxiv.org/abs/2609.34491
作者: Junjie Wang,Yaowei Jin,Ruohui Tang,Guonan Cui,Haojie Wang,Penglei Wang,Dingyan Wang,Duo An,Shuangjia Zheng,Qian Shi
类目: Machine Learning (cs.LG); Biomolecules (q-bio.BM)
*备注:
Abstract:Small-molecule optimization integrates medicinal-chemistry reasoning and computational evidence through iterative, multi-objective decisions. When large language models (LLMs) reason over optimization histories stored primarily in conversational context, they must recover candidate identities, prior evaluations, and task constraints to guide subsequent decisions. We present M3OS, a multi-agent LLM system that decouples molecular-design reasoning from optimization-state management through Monte Carlo graph search. A persistent graph links evaluated candidates, parent-child transformations and evaluation evidence, while rewards and visit statistics guide LLM-assisted parent selection. Two branches combine tool-driven candidate generation with knowledge- and case-guided medicinal-chemistry editing. An execution harness controls graph updates through structured output extraction, molecular validation and task-bound evaluation. Agents receive role-specific contexts, while the graph preserves optimization trajectories beyond their active contexts. Across three molecular optimization benchmarks, M3OS achieves higher success rates than baselines, supporting the integration of persistent search state, specialized agents and controlled execution for multi-constraint optimization.
[LG-136] When Less Data Favors Smaller Teachers: Rethinking Teacher Capacity and Data Selection for Knowledge Distillation
链接: https://arxiv.org/abs/2609.34489
作者: Minjae Park,Taesun Yeom,Jaeho Lee
类目: Machine Learning (cs.LG)
*备注: Preprint. Under review
Abstract:Data pruning reduces the training cost of knowledge distillation (KD). However, the preferred teacher capacity changes with the data budget: smaller teachers can outperform larger ones when limited training data are available. Understanding what drives this shift is important not only for teacher choice but also for identifying which samples are useful for distillation. We analyze teacher supervision by decomposing it into relational ordering—the ranking of classes—and score geometry—the magnitudes and margins of class probabilities—and show that the small-teacher advantage in the low-data regime arises not only from score geometry but also from relational ordering. Beyond understanding teacher capacity, our analysis reveals two properties of effective subsets: samples should match the difficulty appropriate for the available budget, and their relational signals should be diverse rather than redundant. Based on these findings, we propose DVA (Difficulty- and Volume-Aware data selection for KD), a training-dynamics-free method, which uses a small teacher as a proxy for budget-aware difficulty filtering and class-conditional relational volume maximization. Despite requiring no training dynamics statistics, our method remains competitive with training-dynamics-based methods while consistently outperforming training-dynamics-free baselines.
[LG-137] ARS: Agent ic Reward System for Robot Learning
链接: https://arxiv.org/abs/2609.34484
作者: Sheng Hu,Weiyi Lu,Lingbing Zeng,Gan Weng,Weiwei Zhang,Kai Xie,Xiaofeng Mou,Yi Xu
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注:
Abstract:Progress reward modeling is the problem of estimating how a robot’s behavior changes task progress over time. Reliable estimation requires distinguishing meaningful state changes from failed attempts and task-irrelevant actions. We introduce the Agentic Reward System (ARS), an inference framework for progress reward modeling with general-purpose vision-language models (VLMs), without additional reward-model training. Given an offline trajectory and a task instruction, ARS uses adaptive visual inspection for both event proposal and verification. A subagent proposes a task-relevant event timeline, which a primary agent verifies and revises before estimating per-frame progress. ARS can incorporate optional terminal outcome labels and visual references to inform its judgments. It can also audit progress estimates from external reward models. We evaluate ARS with a 27B VLM on a controlled semantic-mismatch benchmark and downstream policy learning in simulation and on a real robot. The benchmark reveals that several evaluated reward baselines assign spurious progress to wrong-object manipulation even in simple pick-and-place scenes. ARS better suppresses these errors and outperforms these baselines in simulation policy learning. We further demonstrate that ARS supports long-horizon policy learning from mixed-quality offline experience on real-robot multi-screw fastening in a full-scale laboratory replica of an industrial washing-machine assembly line. These results suggest that structured inference and verification can improve the usefulness of general-purpose VLMs for robot reward modeling. Code is at this https URL
[LG-138] Deep kernel hedging
链接: https://arxiv.org/abs/2609.34474
作者: Jean-Loup Dupret,Donatien Hainaut,Edouard Motte
类目: Machine Learning (cs.LG); Functional Analysis (math.FA); Computational Finance (q-fin.CP); Risk Management (q-fin.RM)
*备注:
Abstract:We introduce a deep kernel hedging framework that combines the flexibility of deep learning with the structural inductive bias of kernel methods. The hedging functional is restricted to a reproducing kernel Hilbert space whose kernel is parameterized through a neural network embedding of the input features. The framework minimizes a regularized empirical risk under convex loss functions and can accommodate path-dependent information through truncated time-augmented signature features. We derive a generalized representer theorem for the joint hedging problem, reducing the empirical optimization to a finite-dimensional problem. To further reduce the computational cost associated with large kernel matrices, we develop a scalable random Fourier feature approximation and establish convergence guarantees. The random Fourier parameters are sampled once and remain fixed throughout training, while the deep kernel adapts to market data through the learned neural representation. We evaluate the performance of the proposed deep kernel approach on both synthetic and real data and compare it with standard kernel methods and classical deep hedging architectures. Numerical results indicate competitive and robust hedging performance, particularly in low-data regimes, which highlights the benefits of combining expressive neural representations with the inductive bias of kernel methods.
[LG-139] Alignment-Guided Flow Transformer for Efficient Vision-Language-Action Policy Learning NEURIPS
链接: https://arxiv.org/abs/2609.34467
作者: Shengchao Hu,Peng Wang,Qiyang Zhou,Guodong Zheng,Yuqi Huang,Li Shen,Ya Zhang,Dacheng Tao
类目: Machine Learning (cs.LG)
*备注: NeurIPS
Abstract:Recent advances in Vision-Language-Action (VLA) models point toward general-purpose robotic intelligence by unifying perception, instruction, and control. Despite impressive progress, existing VLA models often adapt poorly due to \emphtri-modal misalignment among vision, language, and action, which weakens action grounding and hurts generalization and fine-tuning efficiency. In this work, we present Alignment-Guided Flow Transformer (AGFT), a novel framework that explicitly enforces tri-modal alignment through a dedicated alignment loss, bridging the representational gap across modalities and enhancing task adaptation. While prior research has predominantly emphasized bi-modal vision–language alignment, we systematically formalize and study tri-modal alignment in VLA models, and provide both ablations and analysis to isolate its role in improving adaptation and robustness. To further accelerate deployment, we adopt a flow-matching objective, enabling substantially fewer inference steps than diffusion-based policies while maintaining accuracy. Theoretically, we establish a quantitative connection between the tri-modal alignment gap and the optimization tightness of flow matching; empirically, experiments on the extensive benchmark show that AGFT achieves superior success rates and lower inference latency compared to SOTA baselines, underscoring tri-modal alignment as a key ingredient for scaling robust VLA manipulation.
[LG-140] PhysioTRACE: Provenance-Aware Stress Tests for Physiological Foundation Models
链接: https://arxiv.org/abs/2609.34466
作者: Ayana Mussabayeva,Anuar Aimoldin,Olivier Oullier,Xue Liu,Kun Zhang
类目: Machine Learning (cs.LG)
*备注: Main text: 10 pages, 4 figures. Appendix: 31 pages with full protocol details, per-seed results, and sensitivity analysis
Abstract:Physiological foundation models encode how a signal was recorded alongside the physiology it reflects. When recording conditions are associated with diagnosis, this acquisition provenance can become a shortcut, yet the usual evidence, shifted transfer and provenance decodability, does not show whether a predictor uses it. We introduce PhysioTRACE, a four-axis behavioral audit for frozen encoders that separates what a probe can decode from what a fixed task head relies on. Recover scores how decodable provenance is; Stress reverses only the provenance-target association on the same held-out records; Intervene removes a train-localized provenance component; and Verify certifies that removal only if it beats matched random projections within a declared utility margin. Each audit thus ends in one of three verdicts: no reliance, or reliance with the remedy certified or refused. Across EEG and ECG, five training objectives, and five frozen foundation models, the relation between Recover’s calibrated score and out-of-distribution utility changes sign between datasets, so neither can stand in for a reliance test. On paired EEG views where the shortcut is known by construction, the audit detects it (the exposed head loses about 0.2 AUROC when the association is reversed, while a control head is unaffected) and certifies removal of a rank-two component that restores control-level behavior without measurable utility loss, for both encoder objectives tested. On real ECG device metadata it returns all three verdicts: it certifies a remedy that removes 91% of one model’s excess vulnerability, finds no reliance where device and diagnosis are barely associated, and refuses the remedy for a second model whose localized direction also carries task signal. Robustness to how inputs were recorded therefore needs a behavioral test, and PhysioTRACE provides one that can pass, fail, or refuse a remedy.
[LG-141] ZonoGPT : Towards An Abstract Domain for Verifying Large GPT Models
链接: https://arxiv.org/abs/2609.34457
作者: Hai Duong,Thanh Le,ThanhVu Nguyen
类目: Machine Learning (cs.LG); Software Engineering (cs.SE)
*备注:
Abstract:Transformer-based models are widely used for reasoning, coding, and multimodal agentic tasks. To provide formal assurance of desirable behaviors, such as robustness, safety, and fairness, neural network verification techniques prove required properties and provide auditable guarantees before deployment. However, prior work remains limited to small or restricted Transformers, and maintaining precision across deep models remains challenging. In this work, we introduce ZonoGPT, an abstract domain for verifying large transformers that maintains a space complexity independent of network depth. ZonoGPT uses a structured zonotope and a generator reduction mechanism to efficiently preserve correlations. To maintain precision, it introduces block-specific fused transformations for Attention and LayerNorm that retain feature relations, along with an affine transform for GELU that preserves generator relations. These mechanisms enable \tool to be the first approach to verify standard architectures, scaling to official HuggingFace models up to GPT-2 Medium (24 blocks, 300M+ parameters) and successfully verifying 1,339 instances across text and vision tasks.
[LG-142] Livin on a Prior: Likelihood Score Approximation for Inverse Problems
链接: https://arxiv.org/abs/2609.34446
作者: Rostislav Makarov,Tal Peer,Danilo de Oliveira,Timo Gerkmann
类目: Machine Learning (cs.LG)
*备注:
Abstract:Generative models have found great success as data-driven methods of solving inverse problems. Two popular approaches work either by combining a pretrained generative prior with a known degradation model, or by training a conditional generative model directly from paired data. We target a setting that spans both regimes: unknown degradations can be learned from few paired examples, while known degradations can be learned from self-generated samples. We introduce Likelihood Score Approximation (LSA), a generative framework that keeps a pretrained unconditional model fixed and learns an observation-conditioned model that approximates the likelihood score from paired samples. Within a conditional stochastic-interpolant framework, LSA can be trained in either score or velocity coordinates, independently of the unconditional model’s native parameterization, and supports both deterministic and stochastic sampling. We further show empirically that the prior model can be swapped post-training while keeping the same LSA model. Across speech and image inverse problems, LSA operates effectively even at roughly 0.01% of the full training dataset. On the ImageNet-256 benchmark it achieves competitive or better restoration quality than strong posterior-sampling baselines while requiring up to several orders of magnitude fewer network evaluations.
[LG-143] Making LLM s Truly Forget: Deep Unlearning by Searching Selecting and Severing Knowledge Paths
链接: https://arxiv.org/abs/2609.34442
作者: Jialu Wang,Peizhi Niu,Haoteng Yin,Hans Hao-Hsun Hsu,Pan Li,Rongzhe Wei
类目: Machine Learning (cs.LG)
*备注:
Abstract:While an unlearned language model may no longer recall a fact directly, the fact often remains recoverable through multi-hop reasoning over related knowledge. Most existing unlearning techniques overlook this vulnerability, targeting facts in isolation while leaving their supporting knowledge intact. To achieve true forgetting, we propose a general deep unlearning framework compatible with existing unlearning algorithms. Our approach adaptively explores both explicit responses and latent internal representations to discover valid reasoning paths, compiles them into a confidence-aware supporting subgraph, and we apply a graph minimum cut to sever all recovery paths while preserving unrelated knowledge. To rigorously evaluate deep unlearning, we introduce a model-specific pipeline that extracts and completes knowledge graphs from raw text, filtering them by calibrated model confidence to reflect what the model genuinely retains. Comprehensive experiments demonstrate that selectively unlearning supporting knowledge yields substantially deeper forgetting than superficial methods while preserving model utility, highlighting that genuine unlearning requires breaking the relational structures that enable factual reconstruction.
[LG-144] Admissible Diffusion for Multimodal Interventional Trajectories
链接: https://arxiv.org/abs/2609.34433
作者: Xing Han,Shravan Chaudhari,Jiarui Shao,Paul Pu Liang,Suchi Saria
类目: Machine Learning (cs.LG)
*备注: 15 pages, 3 figures, 7 tables
Abstract:Generating a plausible clinical trajectory does not establish what would happen under a different treatment. We present ADMIT, a framework combining irregular multimodal representations, treatment-conditioned latent diffusion and explicit constraints on generated states or actions. We formulate its interventional target through sequential g-computation and distinguish causal assumptions from constraint satisfaction. Its admissibility mechanism translates physiological prior knowledge into explicit constraints on generated states and proposed actions. Treatment-exposure dynamics condition latent transitions, while state projection or action gating applies the constraints during rollout so that they influence subsequent trajectory generation. In our preliminary experiments, multimodal inputs improved supervised hidden-state recovery and reduced treatment-contrast error. In a simulated dosing-schedule experiment with leak-free history encoding, ADMIT predicted most of the tumor-volume change caused by redistributing a fixed total dose. An exposure input improved these predictions around a temporary dose reduction whether or not the assumed clearance rate was correct, but reduced the predicted size of a dose effect, and a deterministic recurrent baseline matched ADMIT’s average predictions. Exposure projection reduced constraint violations, although enforcement remained incomplete. Semi-synthetic experiments using eICU context illustrated treatment-response generation under fixed and adaptive policies. Observational examples further characterize model treatment sensitivity. ADMIT provides a framework for testing whether complementary observations and physiological restrictions improve intervention trajectories, with representation recovery, effect accuracy and rule enforcement assessed separately.
[LG-145] Harmonizing Spectral Evolution in Conditional Flow Matching for TTS
链接: https://arxiv.org/abs/2609.34431
作者: Isha Pandey Varad Deshpande Abhijat Bharadwaj Ganesh Ramakrishnan
类目: Machine Learning (cs.LG); Sound (cs.SD)
*备注: 4 Pages, 5 figures
Abstract:Conditional Flow Matching (CFM) models for text-to-speech (TTS) suffer from incoherent frequency evolution during inference. While similar spectral imbalances are addressed in diffusion models for other domains, those generic solutions fail to generalize to the inherently uncoordinated acoustic dynamics of CFM. We demonstrate that this issue can be effectively mitigated by introducing a novel training-free frequency-selective boosting strategy. Using the Discrete Wavelet Transform (DWT), our method dynamically modulates mel-spectrogram sub-bands during ODE integration, synchronizing spectral development by penalizing aggressive low-frequency growth and boosting lagging high-frequency details. Validated across diverse architectures (Matcha-TTS, F5-TTS, IndicF5), our approach reduces the required Number of Function Evaluations (NFE) from 32 to 26 and improves Frechet Audio Distance (FAD) by up to 61%, all without compromising mean opinion scores, speaker similarity, and speech intelligibility.
[LG-146] Q-learning Penalized Transformer for Safe Offline Reinforcement Learning ICML
链接: https://arxiv.org/abs/2609.34426
作者: Shengchao Hu,Peng Wang,Jifeng Hu,Qiyang Zhou,Anning Hu,Li Shen,Ya Zhang,Dacheng Tao
类目: Machine Learning (cs.LG)
*备注: ICML
Abstract:This paper addresses the problem of safe offline reinforcement learning, which involves training a policy to satisfy safety constraints using an offline dataset. This problem is inherently challenging as it requires balancing three highly interconnected and competing objectives: satisfying safety constraints, maximizing rewards, and adhering to the behavior regularization imposed by the offline dataset. To tackle this trilogy challenge, we propose Q-learning Penalized Transformer policy (QPT), a \emphtraining–inference consistent framework that bridges conditional sequence modeling with constraint-aware value estimation. QPT trains a Transformer policy that generates actions conditioned on trajectory context and target return/cost, retaining strong behavior regularization. To inject explicit safety semantics during learning, we augment sequence-model training with a Q-shaped penalty using learned reward and cost Q-functions to favor high return under low constraint violation. At inference, the same Q-functions enforce the cost threshold and choose the highest-reward feasible action, closing the loop between training and deployment. We provide a principled analysis under stylized near-deterministic CMDPs, characterizing how Q-penalized conditional generation improve safety and performance. Empirically, QPT consistently outperforms strong safe offline RL baselines across 38 tasks on the DSRL benchmark, and exhibits robust zero-shot adaptation to different constraint thresholds.
[LG-147] On the Relation Between Interval Regret and Dynamic Regret
链接: https://arxiv.org/abs/2609.34423
作者: Yi-Han Wang,Peng Zhao,Zhi-Hua Zhou
类目: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
*备注:
Abstract:Non-stationary online learning has attracted much attention in recent years, as static regret is insufficient to guide algorithm design in changing environments. To address this limitation, interval regret and dynamic regret have been introduced as two representative performance metrics that strengthen static regret in complementary directions. Interval regret requires an online algorithm to achieve competitive static regret over every local time interval, whereas dynamic regret evaluates performance against an arbitrary sequence of time-varying comparators. Despite their importance, the relation between these metrics has long remained unclear. Prior work has often regarded interval regret as the stronger notion, based on the intuition that local guarantees should naturally induce global guarantees. Consequently, it is widely conjectured that an algorithm with optimal interval regret should automatically attain optimal dynamic regret. In this paper, we first establish a negative result that refutes this intuition of a metric-level implication. Specifically, for both convex and curved functions (including exp-concave and strongly convex functions), we show that there exist instances in which an algorithm with optimal interval regret nevertheless fails to achieve optimal dynamic regret. We then show how to leverage local adaptivity to obtain optimal dynamic regret. In particular, optimal dynamic regret can be attained by invoking an interval regret minimization process over an enlarged Euclidean ball containing the original convex feasible domain and using a suitable domain-converted surrogate loss. This reduction applies to both convex and curved functions. As a byproduct, we obtain the first proper and efficient algorithm with optimal dynamic regret for exp-concave functions, improving prior results while significantly simplifying the analysis.
[LG-148] MegaGraph: Towards Efficient Training of Large-Scale Graph Transformers with Automated Hybrid Parallelism
链接: https://arxiv.org/abs/2609.34420
作者: Tong Qiao,Ao Zhou,Yingjie Qi,Chunming Hu,Jianlei Yang
类目: Machine Learning (cs.LG)
*备注: 8 pages,10 figures, accepted by ICCD2026
Abstract:Graph Transformers (GTs) offer superior representation capabilities by overcoming the depth limitations and over-smoothing issues of traditional Graph Neural Networks (GNNs). However, scaling GTs to large graphs poses critical bottlenecks. Specifically, the attention score matrix and its associated topology-aware bias matrix jointly incur significant per-layer memory overhead, and heavy graph embedding layers result in severe workload imbalances. These characteristics are unique to GT training and are not addressed by parallelism techniques designed for either conventional GNNs or Transformers, making a dedicated solution necessary. This paper introduces MegaGraph, the first automated hybrid parallelism framework designed for efficient GT training. MegaGraph designs three specialized strategies, namely graph-aware context parallelism, heterogeneous pipeline parallelism, and hybrid data parallelism, to support efficient training on large-scale graphs. However, coordinating these three parallelism strategies yields an exponentially large configuration space. To address this complexity, an automatic search engine leverages precise cost models via a Profile - Model - Search workflow to identify the optimal parallelism configuration. Evaluations demonstrate that MegaGraph enables training on large-scale graphs where state-of-the-art baselines fail due to out-of-memory (OOM) errors. The framework reduces per-device peak memory by up to 77.8% and achieves up to 4.51 \times training speedup while maintaining model accuracy.
[LG-149] PROACT-Agent : Progressive Runtime Oversight and Active Circuit-breaking for Real-Time Safety
链接: https://arxiv.org/abs/2609.34415
作者: Ding Jia,Wei Liu,Xianglong Du,Yingjie Li,Yingqing Yang,Huili Yu,Zhangsong Zhan,Chu Zhou
类目: Machine Learning (cs.LG)
*备注:
Abstract:The transition from Large Language Models (LLMs) to agents shifts safety stakes from toxic text to irreversible environmental harm. While current defenses remain largely retrospective, proactive runtime intervention is bottlenecked by the lack of large-scale, causally-consistent data. We propose PROACT-Agent, a framework for synthesizing high-fidelity trajectories to enable real-time guardrails. We identify a critical “safety drift” in prior benchmarks, where lenient annotation paradigms fail to enforce temporal consistency. PROACT-Agent addresses this through: (1) Progressive Trajectory Unrolling to reveal risks hidden in long-context interactions; (2) Reasoning-Augmented Causal Rectification to enforce monotonic causal consistency; and (3) Culturally-Aware Data Localization for cross-border robustness. We introduce PROACT-Bench, a bilingual safety benchmark with 155,780 states labeled through multi-model adjudication. Evaluating updated context before the next LLM inference, the trained guard achieves 91.46% unsafe-class F1 and 90.63% exact-boundary detection under complete source holdout. In AgentDojo, it reduces non-DoS targeted attack success from 20.82% to 0.40%.
[LG-150] GeoCFM: Positive-Only Conditional Flow Matching for Mineral Occurrence Sampling ECCV2026
链接: https://arxiv.org/abs/2609.34398
作者: Moshe Eliasof,Eldad Haber
类目: Machine Learning (cs.LG)
*备注: ECCV 2026
Abstract:Critical mineral discovery is a positive-only problem: deposits are observed as sparse locations, while unlabeled regions are not reliable negatives, and similar geophysical signatures can arise from different subsurface states. We therefore model mineral targeting as learning a conditional spatial distribution over occurrence locations, \pi(p\mid d) , given geo-images d , rather than predicting a deterministic per-pixel score map. We introduce GeoCFM, a conditional flow-matching model that generates mineral occurrence point sets conditioned on multi-channel geo-images; GeoCFM learns a point-wise transport field in \mathbbR^2 , using UNet features with point-conditioned velocity prediction to bridge dense rasters and sparse supervision without pseudo-negatives. On a synthetic magnetics–geochemistry benchmark with latent activation and on USGS Earth MRI data with a spatially disjoint tile split, GeoCFM improves geometric agreement with observed occurrences over score-map and non-conditional baselines, while representing epistemic uncertainty through conditional sampling.
[LG-151] Spexis: Speculative Lookahead Scheduling for LLM Inference EMNLP2026
链接: https://arxiv.org/abs/2609.34370
作者: Hyungyu Jung,Jaehyeok Yu,Hoonseo Choi,Sungkyun Kim,Jinho Lee,Jiwon Seo
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注: EMNLP 2026 main
Abstract:Spexis is a multi-GPU LLM inference framework that improves the efficiency of pipeline and tensor parallelism through speculative parallelism. Rather than using speculative decoding only to accelerate token generation, Spexis runs speculation in parallel with normal execution, introducing a new parallelism axis without increasing KV-cache memory usage. This improves memory efficiency and helps mitigate the bottlenecks of multi-GPU inference. Spexis further uses lookahead scheduling to predict speculation quality and future memory pressure, allowing it to reduce wasted speculation, KV-cache eviction, and recomputation. Built on top of vLLM, Spexis largely improves serving performance across a range of GPU configurations, achieving speedups of up to 34% over a baseline that uses the optimal combination of pipeline and tensor parallelism. Spexis’s source code is publicly available at this https URL. Comments: EMNLP 2026 main Subjects: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC) Cite as: arXiv:2609.34370 [cs.LG] (or arXiv:2609.34370v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.34370 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-152] FAST-Brain: A Flow-Aligned Spatio-Temporal Surrogate Brain Model
链接: https://arxiv.org/abs/2609.34354
作者: Shucheng Liu,Changchun Shi,Kai Zhang,Hongtu Zhu
类目: Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注:
Abstract:Modeling resting-state functional magnetic resonance imaging (rs-fMRI) data is crucial for understanding brain-wide neural activity. However, traditional methods struggle to capture complex temporal dynamics over long horizons, to account for the brain’s anatomical spatial structure, and to model high-dimensional ambient signals that lie on a low-dimensional intrinsic subspace. We propose FAST-Brain, a unified flow-aligned spatio-temporal surrogate brain model that addresses all three challenges. At its core is a flow-aligned generative framework that directly predicts the clean blood-oxygen-level-dependent (BOLD) signal, paired with a graph convolutional network that captures spatial structural constraints and a Transformer that models long-range temporal dependencies. Theoretically, we show that under a low-dimensional subspace assumption, the approximation error of our model scales with the intrinsic dimension rather than the ambient dimension, which justifies our direct modeling of the BOLD signal. Extensive experiments on synthetic and Human Connectome Project datasets demonstrate that FAST-Brain achieves state-of-the-art performance in recovering functional connectivity, effective connectivity, and the implicit low-dimensional signal subspace.
[LG-153] he Composition Gap in Dataset Distillation
链接: https://arxiv.org/abs/2609.34343
作者: Guang Li,Takahiro Ogawa,Miki Haseyama
类目: Machine Learning (cs.LG)
*备注:
Abstract:Dataset distillation compresses a training set into a small synthetic set, usually evaluated one at a time. In federated and data-governance settings, several parties distill their own data and a user trains on their union. We ask whether the union of separately distilled sets reproduces training on the union of the real data composability and show that it can fail even when every source is distilled exactly and the total budget admits an exact joint distillate. Compressing a training trajectory into fewer steps transforms the source statistics nonlinearly, so averaging compressed sources differs from compressing their average. For quadratic objectives we derive the exact composition error for two-to-one step compression in terms of the source-Hessian variance and the linear terms of the losses, and on a smooth network at small step sizes this prediction captures the local endpoint discrepancy in magnitude and direction. For learned synthetic sets, however, the composed error decomposes exactly into this local discrepancy and an aggregate source residual. Under endpoint matching the residual exceeds the structural term by more than an order of magnitude, and under distribution matching the two terms partly cancel. Joint distillation also retains an accuracy advantage when both sets are distilled from the same dataset, where the local discrepancy is exactly zero. Training fidelity and downstream accuracy are therefore distinct requirements, neither established by evaluating each set on its own.
[LG-154] Routing Without Embeddings: Fast And Interpretable Routing With Regular Expressions
链接: https://arxiv.org/abs/2609.34326
作者: Yifan Lu,Qiyue Zhang,Haotian Shan,Hanjie Chen,Jiarong Xing
类目: Machine Learning (cs.LG)
*备注: 38 pages, 18 tables, 11 figures
Abstract:Large Language Model (LLM) routers commonly rely on neural query embeddings, with larger encoders expected to better capture query intent and difficulty. Yet scaling Qwen2.5 encoders from 0.5B to 72B parameters brings little improvement in routing accuracy (Figure 1b), suggesting that small encoders may already capture the query properties needed for routing. We therefore investigate which properties matter and whether they can be extracted directly from text without a neural encoder. We introduce REGEXROUTE, a pipeline that uses sparse autoencoders (SAEs) to discover interpretable regular-expression (regex) features. Using unlabeled text, an LLM turns descriptions of grouped SAE latents into regex extractors and refines them to match latent activation patterns. These extractors supply numerical features to a lightweight routing head, eliminating neural encoding at inference (Figure 1a). Across four benchmarks, one fixed set of 128 features achieves 76.43% average routing accuracy, comparable to 76.41% for the strongest neural text encoder baseline, with much smaller latency and strong robustness. These findings establish explicit, interpretable text features as a practical basis for designing and understanding LLM routers.
[LG-155] One Rollout Is All You Get: Fully Test-Time Adaptation for GUI Agents
链接: https://arxiv.org/abs/2609.34321
作者: Ziqiang Wang,Li Gu,Zhixiang Chi,Linlian Jiang,Zihuan Jiang,Linqiang Guo,Siobhan Reid,Zhi Liu,Yang Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:GUI agents are deployed with frozen weights and discard everything they experience on the job. Existing ways to update an agent’s weights assume something deployment withholds: ground truth, rollouts beyond the single attempt (retries, samples, practice runs), or a learning phase other than deployment. Because GUI actions can be irreversible, a deployed agent gets one attempt per task occurrence, in arrival order, and every attempt counts. No ground truth is available at any point. We define fully test-time adaptation for GUI agents by these constraints and pair it with a minimal weight-space method, SOLO. Auxiliary models read each episode: a judge selects the episodes it deems successful, and a proposer-verifier pair relabels a failed episode’s prefix with the subtask that prefix completed. Admitted episodes enter a short sliding window, and each admission updates a small adapter by top-K self-distillation on the agent’s own predictions, provided the window holds a judged success. On recurring task streams built from WebArena, VisualWebArena and MobileWorld, SOLO improves on the frozen agent with both UI-TARS-7B and Qwen3-VL-8B, by three to six points of success rate, and exceeds two in-setting memory methods on the web streams.
[LG-156] Riemannian Difference-of-Convex Optimization for K-Means Clustering
链接: https://arxiv.org/abs/2609.34310
作者: Meng Xu,Bo Jiang,Hanfu Zhang,Ya-Feng Liu,Anthony Man-Cho So
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: 5 pages
Abstract:K-means is a widely adopted clustering approach in signal processing and machine learning. In this paper, we study K-means clustering through a cardinality-constrained formulation on a compact embedded submanifold. We replace the cardinality constraint with a difference-of-convex (DC) penalty and establish a global error bound to prove that the penalized and constrained formulations share the same global minimizers whenever the penalty parameter exceeds a finite threshold. To solve the resulting nonsmooth Riemannian DC problem, we reformulate it as a minimax problem and propose RADA-DC, a Riemannian alternating descent ascent method combining dual regularization with DC linearization. Under standard assumptions and suitable parameter choices, RADA-DC finds an \epsilon -Riemannian critical point within O(\epsilon^-3) iterations. We conduct experiments on synthetic and real-world datasets to demonstrate that the proposed method outperforms the tested baselines, including K-means++, in solution quality at competitive computational cost when the number of clusters is large.
[LG-157] Epistemic Learning from Imprecise Annotation
链接: https://arxiv.org/abs/2609.34285
作者: Kaizheng Wang,Siu Lun Chau
类目: Machine Learning (cs.LG)
*备注:
Abstract:Imprecise annotations may support several plausible labelling distributions, yet learning methods often resolve this ambiguity into a single predictive distribution. This can obscure what the annotation evidence leaves unresolved. We introduce epistemic learning from credal supervision, a framework that uses convex sets of plausible labelling distributions, called credal sets, as supervision and learns sets of predictive distributions. We instantiate the framework with the pessimistic–optimistic credal classifier (POCC), which combines a shared backbone with two classification heads trained to minimise worst-case and best-case losses over the supervision sets. Their outputs define a predictive credal set whose spread provides an uncertainty score. We also show how credal labels can be obtained through a simple relaxation of existing probabilistic labels, reducing commitment to their precise probability assignments. This construction admits closed-form inner optimisation under cross-entropy loss, enabling efficient training. Assuming the supervision sets contain the true conditional label distributions, and other regularity assumptions, we establish a finite-sample generalisation bound for the averaged predictor with an explicit penalty for supervision imprecision. We evaluate POCC using human annotator disagreement and teacher predictions, alongside label smoothing as a controlled proxy for annotation imprecision. Across these settings, POCC achieves a favourable balance of predictive accuracy, calibration, and uncertainty-based selective classification versus competitive baselines.
[LG-158] Agent ic High-Dimensional Bayesian Optimization with Hypothesis- and Evidence-Guided Search
链接: https://arxiv.org/abs/2609.34281
作者: Zhixuan Gao,Ke Xue,Rongxi Tan,Ming Chen,Chao Qian
类目: Machine Learning (cs.LG)
*备注:
Abstract:High-dimensional Bayesian optimization (HDBO) seeks sample-efficient optimization when the number of variables is large relative to the evaluation budget. Recent LLM-based and agentic BO methods incorporate task knowledge and adapt search decisions during a run, but have primarily been evaluated on low- and moderate-dimensional problems. We ask whether this paradigm can transfer to the higher-dimensional regime. Our experiments show that these methods do not remain reliable in the high-dimensional regime, where the challenge is not only where to evaluate, but also which modeling assumption and search geometry to use when the objective’s useful structure is unknown. We therefore introduce HERA, a Hypothesis- and Evidence-guided Research Agent that uses task context, optimization feedback, and structural diagnostics to revise search hypotheses, select and configure HDBO strategies, and determine their execution length. PRISM, its numerical optimization engine, generates and evaluates candidates sequentially within each search block, updating numerical models after each observation. HERA remains competitive with strong numerical HDBO baselines and outperforms the evaluated LLM-based and agentic methods on four metadata-free synthetic functions. Across eight real-world tasks, HERA achieves the best mean final objective among all evaluated systems on most benchmarks. Further analyses show that structural diagnostics change strategy use, metadata effects vary across tasks, and adaptive search blocks reduce inference cost.
[LG-159] Broken Symmetry in BF16 Attention: Why FlashAttention Gradients Blow Up Late in Training
链接: https://arxiv.org/abs/2609.34272
作者: Junlin Chen,Daize Dong,Huanwei Di,Haolong Jia,Jiawei Wu,Haotian Xie,Mingkai Zheng,Yang Li,Leshang Chen,Huishu Wang,Eric P. Xing,Hongyi Wang
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC); Numerical Analysis (math.NA)
*备注: 28 pages, 10 figures
Abstract:BF16 is now standard in large-scale pretraining, including in fused attention kernels such as FlashAttention, and these kernels are widely trusted. When we used FlashAttention-3 to pretrain a 450M-parameter transformer on 50B tokens, however, we ran into a problem: training was healthy for 25B tokens, then the gradient norm grew a thousandfold and the loss ended 0.2 nats above FP32 attention, without a single NaN. Recomputing the attention backward of just two layers in FP32 removes almost all of the excess gradient. Part of the cause is known: a fused multiply-add in the forward softmax, so far treated as an extreme-input NaN case and never fixed in FlashAttention-3. Repairing it stops the blow-up, but the query gradient is still wrong by more than its own size, and training still drives attention logits to thousands of times their size under accurate gradients. The remaining error comes from a broken conservation law. The softmax score gradient sums to zero along every row, which makes the query gradient blind to where the keys sit as a group; rounding it to BF16 leaves a small nonzero sum that leaks the mean key into the gradient, and the leak grows exactly as late training makes keys large and attention sharp. We introduce GProj (gauge projection), which restores the zero sum after the cast with two rank-one corrections per row. It cuts the remaining median query/key gradient errors from 219%/13% to 0.34%/0.37%, on par with FP32 attention, for 4.7% more time per training step. In matched from-scratch runs it trains to the same loss as FP32 attention, while FlashAttention-3 and key smoothing both destabilize.
[LG-160] ZeroCode: On-demand Error-Correcting Code Construction from the Zero Matrix via Reinforcement Learning
链接: https://arxiv.org/abs/2609.34265
作者: Ju-Hyeong Lee,Yongjune Kim,Sang-Hyo Kim,Dae-Young Yun,Hee-Youl Kwak
类目: Machine Learning (cs.LG)
*备注: 18 pages, 8 figures. Code: this https URL
Abstract:Error-correcting codes (ECCs) are essential across diverse applications, from wireless communications and storage to quantum computing, yet each application imposes distinct design requirements on the parity-check matrix (PCM). To address these on-demand requirements in a unified framework, we propose ZeroCode, a reinforcement learning (RL)-based approach that constructs PCMs sequentially from the all-zero matrix. ZeroCode formulates construction as a discrete sequential decision-making problem and uses proximal policy optimization with action masking to select valid edges. ZeroCode achieves a gain of approximately 1 dB over the prior RL-based construction method at a bit error rate (BER) of 10^-4 for the (32,16) code and outperforms existing genetic, differentiable, and classical code-design methods in our experiments. Beyond optimizing decoding performance, the masking mechanism allows on-demand structural constraints, such as a maximum degree, 4-cycle-free structure, and quasi-cyclic structure, to be flexibly incorporated. Moreover, a single policy rollout yields a library of PCMs with varying edge counts, offering trade-offs between decoding performance and complexity without retraining. Overall, ZeroCode addresses diverse code-design requirements within a unified framework, providing solutions with optimized decoding performance under given constraints.
[LG-161] Rotated Manifold Optimization for Low-Rank Adaptation
链接: https://arxiv.org/abs/2609.34264
作者: Yuhui Ding,Javier Zazo,James Hensman
类目: Machine Learning (cs.LG)
*备注:
Abstract:We propose a novel optimizer for low-rank adaptation (LoRA) that explicitly incorporates the gauge symmetry of low-rank factorization. Our optimizer extends recent matrix optimizers for full-parameter training to the manifold of fixed-rank matrices by interpreting them as normalization under a rotated basis. We show how rotation and normalization can be integrated with the fixed-rank manifold efficiently. Our optimizer converges faster to lower held-out loss and achieves better or comparable downstream performance on both supervised finetuning and reinforcement learning tasks.
[LG-162] RoboICL: Embodied In-Context Learning with GPT -6 Astra
链接: https://arxiv.org/abs/2609.34261
作者: Fangcheng Liu,Yeqing Shen,Anda Cheng,Weishi Mi,Chao Tang,Chenyuan Liu,Yushun Xiang,Tingguang Li,Yong-Lu Li,Yehui Tang
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注:
Abstract:General-purpose vision-language models offer a promising way to zero-shot robot control: \gptastra excels at open-ended and language- or image-conditioned manipulation but remains substantially weaker on high-precision and long-horizon tasks. We introduce \emphRoboICL, an in-context robot-control framework that narrows these gaps without robot-specific parameter updates or a learned VLA. RoboICL separates \emphdemonstration context, which provides recorded examples when available, from \emphinteraction memory, which accumulates the model’s own actions and observed outcomes. Both use a shared observation–action–receipt–observation grammar. To preserve experience across task stages, RoboICL combines sampled demonstration blocks with bounded anchored memory. Fixed anchors keep earlier rollout interactions available for in-context learning, while the latest interaction supports immediate error correction. Across 30 RoboDojo tasks, using zero shot for Open and one demonstration elsewhere, RoboICL improves on official zero-shot \gptastra by 20–27 progress-score points in every category. It leads the leaderboard baselines on Memory and Open, achieves comparable performance to the strongest Precision baseline, and remains competitive on Long-Horizon. Its 30-task Overall score is 50.64, versus 33.68 for the strongest baseline. On a separate ten-task subset, RoboICL scores 60.60, within 2.00 points of the \pi_0.5 + \gptastra hybrid approach. On three real-robot tasks, mean progress rises from 14.45 at zero shot to 63.33 at one shot and 78.89 at three shots. On two development tasks, optional Jev-gated action reuse reduces \gptastra calls by 33–48%. Code is available at \hrefthis https URLthis https URL.
[LG-163] CasEm: A Cascade Architecture for Long-Horizon Neural Emulation
链接: https://arxiv.org/abs/2609.34246
作者: Zhaoyi Li,Jingtao Ding,Shihua Li
类目: Machine Learning (cs.LG)
*备注: 39 pages
Abstract:Autoregressive neural emulators can drift or diverge over long rollouts despite accurate short-term predictions. We introduce Cascaded Emulation (CasEm), a one-way rollout architecture that augments an existing full-state backbone with an independently evolving model of physically specified aggregates. Its forecasts guide corrections to full-state predictions, without feedback from the backbone to the aggregate model. Effective guidance requires aggregates that cover substantial backbone error, remain accurately predictable, and support useful full-state corrections. We derive a finite-horizon error bound that clarifies these three factors and use empirical diagnostics to guide subsystem selection. Across four ODE/PDE benchmarks, CasEm reduces long-horizon rollout errors across diverse backbones and suppresses the trend toward error divergence in both diffusion tasks using Fourier neural operator backbones. In global climate emulation, CasEm with a regional total-water subsystem reduces 10-year full-state time-mean error by 66.6% and 46.3% for frozen ACE and Spherical DYffusion backbones, respectively, while adding less than 3% to inference time.
[LG-164] SleuthBench: Benchmarking Statistical LLM Evaluation Using Tabular Hidden Signals
链接: https://arxiv.org/abs/2609.34228
作者: Jingyun Jia,Antoine Remond-Tiedrez,Aaron Alvarez,Joshua Shunk,Rich Caruana,Ben Lengerich
类目: Machine Learning (cs.LG)
*备注:
Abstract:Evaluating statistical discovery by large language model (LLM) agents requires verifiable analytical ground truth. Establishing such ground truth for real-world datasets is costly, and prior knowledge of public datasets can influence agent responses. We introduce SLEUTHBENCH, a benchmark that addresses both problems by injecting controlled data-quality problems and feature effects into public tabular datasets: the injected pattern determines the answer, so reference answers are computed automatically and memorized knowledge of the original table is insufficient, while the table keeps its background structure. The injected patterns are modeled on phenomena reported in real data analyses. The benchmark defines 17 question templates in two families: data-quality questions and feature-contribution questions. We evaluate six state-of-the-art LLMs that analyze the data using a Python coding tool, on data-science and business phrasings of 70 validated dataset-template combinations, yielding 1680 graded responses in total. The models detect data-quality problems reliably (83.8% accuracy) but recover feature contributions poorly (41.9%). Finding how features shape the target requires searching over both candidate variables and analytical procedures. To address this issue, we propose the Empirical Layer, a set of precomputed statistical artifacts comprising summaries, fitted feature and interaction effects, and dataset descriptions, which exposes candidate patterns for direct inspection. Access to these artifacts raises feature-contribution accuracy from 41.9% to 68.0%.
[LG-165] Learning to Optimize through Solver-Grounded Self-Play
链接: https://arxiv.org/abs/2609.34205
作者: Xia Jiang,Yaoxin Wu,Chenyu Zhou,Mengzhu Xu,Wim P.M. Nuijten,Yingqian Zhang
类目: Machine Learning (cs.LG)
*备注: 36 pages
Abstract:Optimization modeling is central to many decision-making scenarios, but traditionally requires extensive domain expertise. While Large Language Models (LLMs) have shown promise in automating this process, current training paradigms mainly rely on human-annotated or teacher-generated datasets. This dependence introduces a Generalization Ceiling, where models overfit to narrow data distributions, and Capability Anchoring, where models’ reasoning is bounded by annotator proficiency and teacher model capability. In response, we propose OPT-Zero, the first fully self-play training framework for optimization modeling that requires zero external training data. OPT-Zero employs a single LLM in a dual-role closed loop: a Proposer that synthesizes increasingly challenging optimization problems alongside their mathematical formulations and solving code, and a Solver that attempts to resolve the problems given only natural-language problem descriptions. Grounded in execution feedback from external optimization solvers, we alternately train both roles using reinforcement learning. This process fosters an auto-curriculum in which the Proposer and Solver co-evolve: generating harder valid problems by the Proposer seamlessly enhances the structural reasoning ability of the Solver. Extensive results indicate that with zero curated data, OPT-Zero matches state-of-the-art data-dependent methods while exhibiting substantially stronger generalizability, establishing self-play training as a highly scalable paradigm for advancing LLM reasoning in modeling and solving optimization problems.
[LG-166] Frozen Judges Moving Agents : Version-Dependent LLM -Judge Error and the Limits of Judge-Assisted Agent Evaluation
链接: https://arxiv.org/abs/2609.34198
作者: Jiapeng Li
类目: Machine Learning (cs.LG)
*备注: 40 pages; 5 figures; prospectively planned agent-evaluation study
Abstract:Language-model judges compare agent upgrades with their predecessors, but a fixed judge can make version-dependent mistakes. We analyze 35 public coding-agent submissions (20 prespecified version pairs on 250 SWE-bench Verified issues), two customer-service agents (155 tau-bench tasks), and 1,106 expert-labeled AgentRewardBench trajectories. An upstream outage left three judges for the primary SWE-bench analysis (8,743 aligned agent-task cells); the fourth is descriptive. All three coding-agent judges and all four tau-bench judges reject task-conditioned error invariance after multiplicity adjustment. On SWE-bench, 32 of 60 judge-by-pair units have a detectable differential comparison component; eight judge-only intervals declare improvements that execution-based intervals cannot establish, despite rank correlations of 0.71-0.79. In tau-bench, one judge confidently reverses a nine-point reference-reward gap by penalizing a procedural habit the reward ignores. False acceptance of failed coding patches rises with agent capability conditional on task and reference outcome, while a task-solvability prediction from AgentRewardBench reverses sign in SWE-bench. Transporting old-version calibration raises mean absolute comparison error on SWE-bench from 3.8 to 19.5 percentage points; 24.6% of ratio-bootstrap draws are undefined near the correction boundary. A tuned paired audit narrows a classical interval by only about 5% at 80 labeled tasks. A randomized three-arm test does not support the predicted increase in false acceptance from showing the agent’s final report (all three Holm-adjusted p-values = 1.0). These results favor explicit reference standards and paired audits of current outputs over judge-only release decisions or transported old-version calibration.
[LG-167] Cardinality-Stratified Interaction Decomposition for Interpretable Pairwise and Higher-Order Structure in Transactional Basket Data
链接: https://arxiv.org/abs/2609.34191
作者: Hidetoshi Kawase,Toshihiro Ota
类目: Machine Learning (cs.LG)
*备注: 29 pages
Abstract:Transactional basket data can reveal associations among items, but observed co-occurrence conflates item-specific relations with basket-size structure and unmodeled higher-order dependence. We introduce Cardinality-Stratified Interaction Decomposition (CSID), an interpretable framework that decomposes log-odds contrasts stratified by the number of remaining items into item-set-specific and cardinality-common components, without fitting a global joint distribution. CSID uses an information-weighted, gauge-constrained ridge projection to estimate pair and triple components and to diagnose higher-order contributions to pairwise structure. CSID is designed primarily for interpretable decomposition of association structure rather than for full-distribution prediction. In a simulation with zero pair effects, increasingly strong small-basket cardinality potentials drive ordinary Ising couplings spuriously negative, whereas CSID pair estimates remain centered near zero. Detection power rises with the magnitude of planted triple effects, and local deprojection reduces pair-coefficient RMSE from 0.244 to 0.073. Across three grocery datasets, high-information triple components are reproducible over time. In the matched cross-period partial-transfer evaluation, transferred CSID triple components show closer agreement with later-period stratified contrasts than the nodewise-symmetrized cardinality-aware higher-order pseudolikelihood comparator, with gains in weighted Lin’s concordance correlation of 0.038–0.122. These results support CSID as an exploratory and interpretable decomposition framework for pairwise and higher-order association structure in transactional data.
[LG-168] GPARA: Graph-Posterior-Aligned Refinement and Active Acquisition for Grounding Diffusion Priors
链接: https://arxiv.org/abs/2609.34172
作者: Wangqian Chen,Hao Wang,Yumeng Zhang,Jiajia Guo,Junting Chen,Jun Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Active grounding of a frozen diffusion prior requires jointly determining where new measurements should be taken and how they should be used to refine the current reconstruction. Posterior-ensemble-based methods can estimate acquisition utility from generated samples, but require repeated ensemble generation as observations accumulate and capture posterior geometry only through empirical statistics. This paper proposes GPARA, which learns a context-dependent graph surrogate over diffusion prediction residuals, inducing an explicitly reusable posterior response operator that propagates measurement innovations to unobserved variables and evaluates candidate measurements through weighted posterior-risk reduction. Under the matched surrogate, we show that the same response operator also determines expected one-step acquisition benefit and yields an analytic ranking consistent with expected reconstruction improvement. A bounded learned residual calibrates the analytic utility to account for surrogate mismatch, while a small prior ensemble is generated once and reconditioned to update risk weights without repeated diffusion posterior sampling during acquisition. Experiments on two reconstruction tasks spanning physical field and computer vision show consistent improvements in refinement and active acquisition over the evaluated baselines. Ablations further support the complementary roles of step-wise graph refinement, adaptive risk weighting, and analytically anchored calibration.
[LG-169] Hidden Activations are not Enough I: Knowledge Matrices as Higher Representations
链接: https://arxiv.org/abs/2609.34166
作者: Marco Armenta
类目: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
*备注: 86 pages, main text 35 pages, appendices 51 pages
Abstract:We study the knowledge matrix of a trained feedforward network as a higher representation of its inputs. A network is a pair (W,f) , a thin representation W of its quiver and an activation f ; its function factorizes through the space of quiver representations, each input x inducing a representation, and the knowledge matrix M(x)\in\mathbbR^C\times(d+1) is the contraction of that representation to one matrix whose rows sum exactly to the logits. At one trained network we ask what determines it, what it is invariant to, what it determines, and what its geometry measures. Under (LCS), a locally constant slope diagonal, as for ReLU, the matrix at a regular input is a function of the realized germ; its stabilizer among encodings regular there is exactly the germ stabilizer at inputs with no vanishing coordinate, neuron permutation a special case; and it recovers the germ, whereas hidden activations, gauge-covariant and germ-incomplete, are not enough. Under (LCS) it equals per-class gradient \times input plus an exact aggregate bias attribution, grounding it in attribution theory and computing it by C vector-Jacobian products instead of probing. The fixed shape gives an alignment-free per-sample distance between ResNet-152, DenseNet-121 and GoogLeNet; the row-sum identity gives an exact visible/invisible displacement decomposition whose unit-free coherence A=(d_\Psi/d_M)^2 puts adversarial germ motion at median A\le 0.23 , with an attack-family ordering concordant across six architectures (Kendall W=0.921 ; 0.97 on the three networks at full scale). Two honest negatives: on AlexNet/CIFAR-10 penultimate features win 5 of 6 detectors and all 16 attacks, and a matrix-direction counterfactual fails 0/54.
[LG-170] ransfer Calibrated Prediction Powered Inference
链接: https://arxiv.org/abs/2609.34156
作者: Aditya T. Vadlamani,Jae Ho Chang,Srinivasan Parthasarathy,Subhadeep Paul
类目: Machine Learning (cs.LG)
*备注:
Abstract:Prediction-powered inference (PPI) and its power-tuned extension (PPI++) improve confidence intervals by combining a small gold-standard labeled sample with a large AI model’s predictions. Its efficiency gain relies on low residual variance, which may not hold if the predictor is pre-trained on a different source domain. We propose Transfer Calibrated Prediction-Powered Inference (TC-PPI), adapting the source-domain predictor to the target domain using gold-standard samples through cross-fitting. This approach supports various adaptation methods, such as sparse linear calibration, LoRA, and fine-tuning. Our jointly tuned cross-fit estimator, Joint-TC-Cross-PPI++, maintains unbiasedness and is simultaneously at least as efficient as classical inference, PPI, and PPI++, thereby protecting against negative transfer. We provide high-dimensional MSE bounds for calibration and show empirical improvements over baseline methods across various real-world applications.
[LG-171] SPINET: Sheaf Protein Inverse Folding Network
链接: https://arxiv.org/abs/2609.34153
作者: Jens Lundsgaard,Colin Mikulski,Zhixuan Yan,Dhananjay Bhaskar
类目: Machine Learning (cs.LG); Biomolecules (q-bio.BM)
*备注:
Abstract:Proteins change shape as they function, yet most inverse folding models predict amino acid sequences from a single, fixed backbone. A central challenge in protein engineering is to design proteins that undergo specific motions, which requires accounting for how their structures change over time. This motivates inverse protein folding conditioned on protein motion. We introduce SPINET, which predicts sequences from molecular dynamics trajectories. It uses cellular sheaves to represent residue interactions within each frame and recurrent units to integrate information across frames, then predicts all amino acids in a single pass. We evaluate SPINET on mdCATH and ATLAS, where it outperforms all evaluated static and ensemble baselines in sequence recovery. On mdCATH, it achieves 56.7% top-1 recovery, compared with 44.5% for the strongest static baseline and 40.7% for the strongest ensemble baseline. We also evaluate whether the predicted sequences are compatible with conformations sampled along the target trajectory. On mdCATH, they achieve a median TM-score of 0.760, and structural recovery favors target conformations over unrelated decoys for 99.5% of test domains.
[LG-172] ExpertoRhythm: Morphology-Aware Learning for Waveform Reconstruction and Cuffless Blood Pressure Estimation from Single-Channel PPG
链接: https://arxiv.org/abs/2609.34146
作者: Amir Arjomand,Kenneth B. Kent,Georgiy Krylov
类目: Machine Learning (cs.LG)
*备注:
Abstract:Continuous cuffless blood pressure (BP) monitoring from photoplethysmography (PPG) has strong potential for wearable health and telemonitoring, but accurate estimation remains difficult because PPG-to-BP mapping must preserve subtle waveform morphology and pressure-range-dependent dynamics. We introduce ExpertoRhythm, an attention-enhanced 1D U-Net that reconstructs the arterial blood pressure (ABP) waveform from a single-channel PPG signal and derives systolic and diastolic BP directly from the reconstructed waveform. The central contribution is a composite morphology-aware learning objective that integrates range-weighted SmoothL1 reconstruction with a window-range regularizer to emphasize high-dynamic BP segments and reduce amplitude under/over-shoot. On the UCI cuff-less BP dataset with 942 subjects, ExpertoRhythm achieves 2.46/1.46 mmHg MAE for systolic/diastolic BP (SBP/DBP), while obtaining a 30.4% average relative error reduction over pure MSE across waveform reconstruction and BP estimation metrics. Clinical-style evaluation further demonstrates low bias and strong agreement across the BP range, including high-pressure windows up to 200 mmHg, satisfying AAMI criteria and achieving BHS Grade A. These results suggest that morphology-aware waveform reconstruction from a single PPG channel can provide an accurate and practical pathway toward continuous cuffless BP monitoring in wearable and remote-care settings.
[LG-173] What Does a Stream Model Buy You in Flow Matching?
链接: https://arxiv.org/abs/2609.34123
作者: Jian Xu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Stream-level flow matching replaces the linear interpolant of conditional flow matching (CFM) by a Gaussian-process (GP) stream connecting each source–target pair, and reports lower sample error than \icfm on 2-Gaussian, MNIST and CIFAR-10 benchmarks. We ask what such a stream model actually contributes. Three results answer the question. (i)~\emphReduction. The stream-level CFM objective depends on the stream law only through the per-time joint law of (s_t,\sdot_t) , so the conditional paths a Gaussian stream can reach are exactly the Gaussian conditional paths CFM already parametrises; in the coordinate-wise, shared-scalar-kernel construction gpcfm actually uses, the entire design space collapses to two scalar curves (m_t,v_t) , and cross-time covariance affects only estimator variance. (ii)~\emphThe GP is a constrained chart of that space. One kernel sets both m_t and v_t , so the paper’s own recipe for widening coverage-shrinking the SE length-scale—destroys the interpolant (the midpoint mean weight falls from 1.03 to 0.00 ). On the 2-Gaussian benchmark this makes the GP chart diverge on 15/200 runs at high coverage against 0/200 for a decoupled (m_t,v_t) chart ( p=6.6\times10^-5 ), and crossing the two curves shows the divergence tracks the mean, not the variance. On MNIST the same sweep does not diverge and the ordering reverses, so whether the coupling is harmful is benchmark-dependent; what holds on both is that the recipe buys nothing—no coverage level beats the paper’s own, and past \max_t\sqrtv_t\approx0.6 both charts degrade. (iii)~\emphAudit. The released code does not implement the mechanism it describes: state and velocity are drawn independently ( \mathrmcorr=0.00\pm0.01 against an intended \pm0.83 – 0.99 ).
[LG-174] Evolution of fairness in multi-objective reinforcement learning framework
链接: https://arxiv.org/abs/2609.34114
作者: Jingyi Zhang,Xin Ou,Guozhong Zheng,Shengfeng Deng,Jiqiang Zhang,Li Chen
类目: Machine Learning (cs.LG); Disordered Systems and Neural Networks (cond-mat.dis-nn)
*备注: 11 pages, 7 figures. Comments are appreciated
Abstract:Fairness, as a fundamental social norm, continues to pose a longstanding puzzle regarding its emergence. Traditional game-theoretic models largely rely on the assumption of \emphHomo economicus, wherein individuals are purely rational and self-interested, acting solely to maximize material payoffs. Such accounts, however, overlook the multidimensional nature of human decision-making, which is often shaped also by other considerations beyond economic incentives. To address this gap, we propose a multi-objective reinforcement learning framework that models the evolution of fairness as a dynamic trade-off between material payoff maximization and fairness-driven moral behavior, regulated by a fairness pressure coefficient. Using simulations of a two-objective Q-learning ultimatum game, we find that increased fairness pressure promotes fair outcomes, as expected. Strikingly, however, under moderate pressure, responder behavior reverses: responders become ``forgiving" by accepting low offers – a pattern in line with our daily experience. Microscopic analyses reveal that this strategy reversal stems from competition between payoff-maximizing and fairness-oriented preferences. We further extend our framework to an asymmetric setting, where proposers and responders assign different weights to the two objectives. Overall, our work expands the reinforcement learning paradigm from a single-objective to a multi-objective formulation, offering a versatile tool for elucidating a broader range of human social behaviors.
[LG-175] Beyond Correctness: Evaluating Semantic Knowledge in Cross-Table Transfer
链接: https://arxiv.org/abs/2609.34098
作者: Seokyong Sheem,Hochang Lee,Suyeong Lee,Daekyum Kim
类目: Machine Learning (cs.LG)
*备注: 40 pages. Seokyong Sheem and Hochang Lee contributed equally. Corresponding author: Daekyum Kim. Code and reproducibility artifacts: this https URL
Abstract:Semantic knowledge is increasingly used to bridge heterogeneous schemas in tabular learning, but how much does that knowledge actually improve prediction? Studies in tabular learning commonly answer this question through semantic ablations that modify or suppress the supplied semantic knowledge. We show that these ablations can lead to misleading conclusions about predictive benefit: poor performance under altered semantics may be taken as evidence that the intended knowledge is beneficial. Across real and controlled experiments, altering semantic content can produce large performance differences even when the model gains little predictive benefit from having that semantic knowledge in the first place. To separate these effects, we distinguish two quantities: content sensitivity and predictive utility. Content sensitivity measures the change in performance when semantic content is altered, whereas predictive utility measures the benefit of the intended semantic knowledge relative to a suitable reference without that knowledge. This distinction motivates an evaluation framework in which the control is chosen according to the question being asked: altered controls assess sensitivity to semantic content, whereas claims that semantic knowledge improves prediction require a suitable reference. Even then, predictive utility is not fixed; it varies across suitable references and decreases when the reference can more easily recover the tested knowledge from other inputs or labeled examples. In a bounded audit of 25 semantic-ablation comparisons across nine studies, only one of 18 explicit predictive-utility claims is paired with a control that clearly isolates the tested semantic contribution. Together, these findings motivate a simple evaluation principle: semantic-ablation controls should be chosen and interpreted according to the question they are intended to answer.
[LG-176] When Is an SAE Feature Interpretable? A Validation Ladder for EEG Foundation Models
链接: https://arxiv.org/abs/2609.34091
作者: Yucong Cao,Chenqi Li,Tingting Zhu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Sparse autoencoders (SAEs) decompose dense model activations into discrete latents, making individual features easy to interpret–and easy to misinterpret. In EEG foundation models, this creates a tempting inference: if removing alpha-band activity strongly changes a latent’s activation, one might conclude that the latent represents alpha activity. Across 27 settings spanning three backbones, three EEG datasets, and three network depths, this interpretation initially appears compelling: alpha removal changes latent firing 7.3 times more than an equal-width sham notch (95% CI [6.2, 8.7], bootstrapped over settings). However, the alpha filter also deletes far more signal than the sham. After normalizing by removed spectral energy, the ratio falls to 0.28 (95% CI [0.22, 0.36]) and exceeds one in none of the 27 settings. Latents selected for their response to alpha removal are, on clean EEG, slightly anti-correlated with relative alpha power (mean r = -0.073), giving no support for a simple alpha-detector reading. Motivated by this failure case, we propose a validation ladder for semantic interpretations of SAE latents: it asks in turn whether a latent responds, whether that response survives controlling for how much signal the intervention removes, whether it is specific rather than broadly fragile, and whether the proposed property is visible on unperturbed data–while separately testing the stronger claim that the latent matters to a task classifier. Perturbation sensitivity alone does not establish what an SAE latent represents.
[LG-177] Beyond One Epoch: Uncertainty-Weighted Sensitivity Regularization for Recommendation Models
链接: https://arxiv.org/abs/2609.34083
作者: Richard Lettich,Shagun Gupta
类目: Machine Learning (cs.LG)
*备注:
Abstract:Recommendation models with sparse embeddings and a shared consumer often exhibit the one-epoch phenomenon: a second epoch lowers training loss while sharply degrading generalization. We present a view based on the violation of the prequential principle. On the first epoch, an example’s label has not affected the embedding rows used to score it. On later epochs, those rows contain a displacement induced by the labels earlier update. This creates an incentive for the shared consumer to exploit this displacement in subsequent epochs, which fails to generalize. We call this self-influence asymmetry. Using an exact scalar model and local influence analysis, we connect this mismatch to the uncertainty in the embeddings and the consumers incentive to exploit it in subsequent epochs. We verify this hypothesis using an embedding-consumer-update interventions in deep recommendation models and propose uncertainty-weighted sensitivity regularization (UWSR) which counteracts this mismatch by augmenting the loss function to penalize the consumer for relying on uncertain embeddings. Unlike existing remedies, UWSR preserves the learned embeddings and across three benchmarks, four-epoch UWSR reduces test cross-entropy by 1.38%-6.78% and improves AUC by 0.0058-0.0231 relative to one-epoch training.
[LG-178] he Double-Edged Sword of Information: Revealed versus Hidden Lotteries in School Choice
链接: https://arxiv.org/abs/2609.34074
作者: Parinaz Naghizadeh,Jingyan Wang
类目: Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG); Theoretical Economics (econ.TH)
*备注:
Abstract:In school choice, a lottery number is often used by the matching mechanism to break ties when there are more students who prefer the same school than the number of seats available. There has been growing theoretical and empirical interest in understanding the impact of revealing the lottery number to students. In practice, in recent years, the NYC Public Schools started revealing the lottery number to students to improve transparency. Theoretical findings from prior literature also suggest that revealing the lottery number strictly improves the number of matches under the deferred acceptance algorithm. However, these theoretical results are based on the over-simplifying assumption that all students share the same preference ranking for schools. In this work, we relax this assumption and allow students to have heterogeneous preference rankings. Under a game-theoretic model where student strategies form a Bayesian Nash equilibrium, we characterize scenarios where revealing the lottery number can either improve or worsen the matching outcome, measured by two metrics of match rate and social welfare. We further consider revealing partial information about the lottery, and demonstrate non-monotonic effects in the amount of information available to students. These results together illustrate complex tradeoffs induced by the lottery revealing policy.
[LG-179] KVCMAS: Efficient KV cache Correction for Shared Context in Multi-Agent Systems
链接: https://arxiv.org/abs/2609.34060
作者: Hyesung Jeon,Hyeongju Ha,Seoyoung Lee,Beomseok Kang,Jae-Joon Kim
类目: Machine Learning (cs.LG)
*备注: 29 pages, 13 figures, 15 tables
Abstract:Prompt-specialized multi-agent systems enable multiple agents to share a model while performing complementary roles to solve complex tasks. However, agent-specific prefixes change the KV cache generated for the same shared context, causing each agent to repeatedly prefill the growing context and construct a separate cache with high computation and memory overhead. Selective recomputation reduces this redundancy but still retains substantial model execution, while existing delta correction methods either support only recurring context relations or maintain memory-intensive online correction states for dynamically changing context. For first seen shared context, these methods also construct a reference cache outside the agent workflow, and an approximate correction at the first agent affects the outputs passed to subsequent agents. We present KVCMAS, an online KV cache correction framework that represents cross-agent cache deviations using compact low-rank states and seamlessly chains corrections along the agent workflow without an additional reference prefill. This design supports dynamically changing shared context while preserving an exact first-agent cache. Across multiple language and vision-language workloads, KVCMAS matches or improves the accuracy of prior KV cache sharing methods while achieving the lowest TTFT under highly concurrent serving. Under controlled serving traces, it provides a 2.0x TTFT speedup over inference without KV cache sharing and reduces peak GPU memory by up to 3.7x relative to a prior KV cache correction method. These results establish KVCMAS as an accurate and scalable KV cache sharing approach for prompt-specialized multi-agent serving.
[LG-180] PReCache: Efficient KV Cache Sharing for Multi-LoRA Agents via Low-Rank Precomputation and Neutral Reconstruction
链接: https://arxiv.org/abs/2609.34054
作者: Hyesung Jeon,Hyeongju Ha,Jae-Joon Kim
类目: Machine Learning (cs.LG)
*备注: 24 pages, 8 figures, 13 tables
Abstract:Multi-LoRA agent systems enable efficient role specialization by sharing a common backbone model. However, each agent repeatedly processes the growing shared trajectory and constructs its own KV cache, introducing substantial memory and computation redundancy in long-horizon tasks. Existing KV cache sharing methods reduce this repeated prefill, but they either require additional training or architectural constraints or retain substantial model computation. Moreover, direct cache reuse causes the current agent to rely on cache states generated by the previous agent’s adapter, weakening the role-specific behavior encoded by its own LoRA. We present PReCache, a training-free KV cache sharing framework with two designs, namely PreLRShared and ReBaseShared, that share the base cache computed using the pretrained weights and precompute a compact agent-specific low-rank (LR) cache. To remove repeated prefill, PreLRShared precomputes each agent’s LR cache when the shared context is first processed, allowing the current agent to use its own LR cache without reprocessing context processed by previous agents. To improve sharing accuracy, ReBaseShared reconstructs the shared base cache from adapter-free hidden states, reducing the remaining error caused by the previous agent’s adapted representation. To minimize its reconstruction cost, we propose two inference schemes tailored to single-stream inference and concurrent serving, performing the same reconstruction after each agent’s turn or alongside its execution, respectively. Across multiple models and agent benchmarks, PreLRShared achieves up to a 3.1x TTFT speedup and a 2.3x improvement in per-request throughput over inference without KV cache sharing. ReBaseShared best preserves accuracy overall among the evaluated cache-sharing methods, with an average drop of only 1.1 points relative to inference without cache sharing.
[LG-181] Kafila: Serving Large Language Models on a Trusted Set of Heterogeneous Commodity Machines
链接: https://arxiv.org/abs/2609.34045
作者: Murtaza Rangwala,Richard O. Sinnott,Rajkumar Buyya
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: 10 pages, 2 figures, 4 tables
Abstract:Between them, the members of a research group or a circle of friends own several consumer computers, none large enough to run a capable large language model. Existing systems pool such capacity across open swarms anyone may join, which a group admitting only trusted machines cannot use. Bounding membership removes what they depend on: a swarm holds each part of the model on several peers and routes around a slow one. A bounded session must use every device it admits. Its pipeline advances at the pace of whichever device received a share it cannot serve quickly, so the division has to be right before serving begins. We propose Kafila, whose protocol assembles a ring from behind NATs, preferring direct paths and relaying where traversal fails, while its planner measures each device’s memory bandwidth, capacity and reachability, divides the model exactly for a fixed ring order, and places the head, which holds the embedding and output projection, together with that division rather than beforehand. On machines with different capabilities across three fleets, from a shared LAN to five devices spanning two continents, Kafila shortens the slowest pipeline stage by up to 5.2\times against the even split of pipeline parallelism, as in GPipe, and up to 3\times against the memory-proportional split of personal-device inference, as in exo, keeps 75 to 87 per cent of the committed hardware doing work where those divisions fall below half, and serves a model no uniform split can place on the fleet at all. What that is worth to a user depends on how much of a token is computation rather than network. Where the members share a network the same division returns 1.56\times the throughput of a uniform split and 1.25\times of a memory-proportional one, and under four concurrent users that lead compounds to 3.2\times rather than fading, each user served at almost the rate of one.
[LG-182] Vision–Language Signals in Constrained RL: Safety Gains Without Anticipation
链接: https://arxiv.org/abs/2609.34041
作者: Samuel Tetteh,Cody Fleming
类目: Machine Learning (cs.LG)
*备注:
Abstract:Safe reinforcement learning seeks policies that maximise task performance while satisfying safety constraints. In driving benchmarks, however, collision costs typically appear only at the time of collision, providing no advance warning of an approaching hazard. Frozen vision–language models can provide dense semantic feedback, yet it remains unclear whether their scores anticipate collisions and which component drives an observed safety improvement. Episodic cost can also favour policies that make little task progress. To address these gaps, we propose VLM-Safe-RL, a framework that integrates frozen CLIP signals into PPO-Lagrangian through reward shaping and an augmented multiplier update. On MetaDrive Hard, which combines the densest traffic with the largest map, the catastrophe rate falls from 31.6% to 19.4%. FormulaOne-L2 analysis finds no evidence that the CLIP signals anticipate collisions and shows that the VLM term has a negligible effect on the Lagrange multiplier. These findings show a conditional reduction in observed catastrophe rate without evidence of collision anticipation.
[LG-183] Posterior Regimes and Latent Deception: Variational Bayesian Inference in Hidden Markov Models for Sequential Fraud Detection in Financial Transactions
链接: https://arxiv.org/abs/2609.34031
作者: Joseph Uririoghene Obukofe,Anthony O’Hare,Chioma Sandra Dike
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 44 pages, 10 figures. Adapted from the primary author’s MSc dissertation, University of Stirling, Scotland
Abstract:We present a three-tier progression of Hidden Markov Models: maximum-likelihood (Baum-Welch), variational Bayesian (VBEM), and a neural variational extension (Neural VBEM), that model each customer’s transaction history as a trajectory through a small number of latent behavioural regimes, one of which is empirically identified as fraud-associated. The Neural VBEM HMM replaces the fixed Gaussian-multinomial emission family with a learned encoder, compressing a 741-dimensional transaction representation into a 64-dimensional latent space in which the VBEM HMM’s posterior operates; a UMAP projection of this space reveals that the discovered regimes are not discrete clusters but ordered segments of a single continuous behavioural manifold, with confirmed fraud concentrated at its extreme. We show that the model’s natural output, that is, the posterior probability of regime membership, is routinely mistaken for a fraud probability, and quantify the resulting miscalibration (the regime-membership interpretation error, MRIE); a corrected posterior-predictive score, closes most of this gap. We further distinguish batch (smoothed) inference, which uses look-ahead unavailable at deployment time, from filtered (forward-only) inference, and report both. On IEEE-CIS transaction data, the neural tier achieves a 14.4 \times fraud enrichment in its identified regime; while its AUPRC trails a discriminative XGBoost baseline, we show this gap is structural and not incidental, and argue the model is best positioned as a calibrated triage and interpretability layer rather than a drop-in ranking replacement.
[LG-184] Estimate Dont Imitate: Reusing Differentiable State-Based Policies for Visuomotor Control
链接: https://arxiv.org/abs/2609.34018
作者: Denis Shcherba,Adrian Abel,Eckart Cobo-Briesewitz,Wojciech Samek,Marc Toussaint
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: 9 pages, 4 figures, video: this https URL
Abstract:Simulation-trained manipulation policies can exploit privileged state information to learn effective contact-rich behaviours, but deployment requires acting from partial observations such as noisy camera images. A common solution is teacher-student distillation, in which a visuomotor policy is trained to reproduce the actions of the privileged expert. This requires the student to jointly infer the task-relevant state and relearn the expert’s action mapping that is already available. An alternative is to reuse the state-based expert and learn only a perceptual interface that reconstructs its missing state inputs. However, minimising the state estimate error alone does not necessarily minimise the downstream control error induced by these estimates. To bridge this gap, we train a visual state estimator using both direct state supervision and an action-consistency loss backpropagated through the frozen, differentiable expert. A scheduled objective first establishes a physically meaningful state estimate and progressively emphasises errors that affect the expert’s actions. Across five goal-conditioned manipulation tasks, retaining the expert consistently outperforms direct pixel-to-action imitation from the same expert demonstration corpus. We further demonstrate sim-to-real transfer on a physical Panda robot, achieving 76% success without retraining the underlying expert.
[LG-185] LTV-CTDNet: Compositional Turning Decomposition for Short-Term Turning-Movement Forecasting
链接: https://arxiv.org/abs/2609.34014
作者: Md Atiqur Rahman Mallick,Kamrul Hasan,Robert T. White
类目: Machine Learning (cs.LG)
*备注: Accepted: September 26, 2026, for presentation at the 2027 TRB Annual Meeting, Washington, D.C. (Paper No. TRBAM-27-06161)
Abstract:Short-term turning-movement forecasts can support signal control and corridor operations, but unconstrained neural networks may produce physically impossible negative counts or outputs that are not explicitly tied to an approach-demand total. This study introduces the Linear Temporal-Variable Compositional Turning Decomposition Network (LTV-CTDNet), a forecasting framework designed to combine competitive accuracy with structurally admissible outputs. LTV-CTDNet was evaluated using seven months of 15-minute LiDAR observations from eight monitored corridor locations in Nashville, Tennessee. Its lightweight encoder combines recent turning-movement history, weekly time-slot embeddings, and location embeddings. The Compositional Turning Decomposition framework separately predicts nonnegative approach totals and within-approach turning proportions, then reconstructs movement forecasts from these components. Among the evaluated predefined configurations, LTV-CTDNet achieved a movement-level MAE of 1.8189 and RMSE of 3.8072. Its accuracy gains over the strongest sequence models were modest, but it produced no negative forecasts, while unconstrained learned models generated negative values in approximately 10.6% to 29.2% of raw forecast cells. The framework enforces nonnegative outputs and exact agreement between each model-predicted approach total and the sum of its component movements by construction, providing directly interpretable forecasts without clipping or coherence correction.
[LG-186] Fisher-Informed Recalibration for Feedback-Based On-Policy Self-Distillation of LLM s
链接: https://arxiv.org/abs/2609.34009
作者: Seohyun Lee,Dong-Jun Han,Seyyedali Hosseinalipour,Christopher G. Brinton
类目: Machine Learning (cs.LG)
*备注: 30 pages
Abstract:Feedback-based on-policy self-distillation has emerged as a promising approach for enabling foundation models, more specifically Large Language Models (LLMs), to learn from their own outputs under external feedback, with a single model serving as both teacher and student. However, such methods can exhibit unstable optimization, conducive to performance collapse during training. To address this limitation, we propose FIRE (Fisher-Informed REcalibration), a dual-branch framework that recalibrates the supervision applied to correct and incorrect on-policy outputs during fine-tuning. For correct responses, FIRE replaces self-distillation with re-weighted on-policy SFT, while for incorrect ones FIRE identifies feedback components that disproportionately influence the teacher-induced update and recalibrates the feedback-conditioned target accordingly. Both branches are influenced by a token-level radius derived in part from a softmax Fisher trace. FIRE separates which direction feedback should move the model from how far the model should move in that direction, while leaving well-behaved feedback supervision unchanged. Our experiments demonstrate that FIRE provides substantially more stable self-distillation while maintaining strong downstream performance, particularly in settings where standard feedback-conditioned distillation becomes unstable.
[LG-187] RICE-Alpha: Reliability-Informed Correction with Event Graphs for LLM -Agent Stock Forecasting
链接: https://arxiv.org/abs/2609.34004
作者: Tong Liu,Lanmiao Liu,Xiang Hu
类目: Machine Learning (cs.LG)
*备注: 15 pages, 4 figures, 4 tables
Abstract:Equity-relevant news evolves through temporally dependent corporate events, making historical information useful only when event continuity, information availability, and transition reliability are modeled. Existing LLM-based financial agents incorporate historical evidence, yet they provide limited support for preserving issuer-specific chronology under point-in-time constraints and for identifying when historical transitions contribute information beyond the current forecast. We present RICE-Alpha (Reliability-Informed Correction with Event Graphs), a point-in-time stock-scoring framework that separates a history-aware multi-view Base Alpha from a reliability-calibrated residual correction derived from historical event continuation. A Multi-Tier Memory Layer grounds news interpretation in temporally eligible issuer-specific history, while a Typed Event Agent constructs event states whose successor relations are formed within issuers and pooled across firms only after valid local pairing. Matured transitions are calibrated by their empirical reliability, and the resulting graph signal is residualized against the Base Alpha and technical view to obtain the RICE Delta. On daily Nasdaq-100 and Hang Seng Index panels from 2024 to 2026, RICE-Alpha achieves the strongest results among the evaluated LLM-based agents and momentum across four predictive and four portfolio-level metrics. Its ICIR more than doubles that of the strongest baseline, while net Sharpe ratios reach 1.656 and 1.725 in the U.S. and Hong Kong, respectively. U.S. ablations further show significant reductions in IC and RankIC after Holm adjustment when major components are removed. These results indicate that historical event continuation adds incremental information when it is temporally grounded, reliability-calibrated, and introduced as a residual correction to a multi-view forecast.
[LG-188] -SNN: Temporal Simplicial Neural Network for EEG Decoding
链接: https://arxiv.org/abs/2609.34002
作者: Nikita Malik,Shubhajit Roy,Mohit Kataria,Isuru Herath,Suraj Yadav,Inés García-Redondo,Dhananjay Bhaskar
类目: Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注:
Abstract:Decoding brain states requires models that capture both the evolution of neural activity and interactions among groups of brain regions. Existing EEG methods often treat recordings as multivariate time series or represent functional connectivity with pairwise graphs, leaving dynamic higher-order interactions largely unmodeled. We introduce the Temporal Simplicial Neural Network (T-SNN), which represents EEG recordings as sequences of evolving simplicial complexes. By combining simplicial convolutions with recurrent updates, T-SNN jointly learns higher-order interactions and their temporal evolution. On the seven-class SEED-VII emotion recognition task, T-SNN outperforms convolutional, recurrent, graph-based, and Transformer methods in both trial-wise and cross-subject evaluations. Incorporating eye-movement features further improves performance, demonstrating the framework’s potential for multimodal brain-state decoding.
[LG-189] ASTRA: ADMM-Accelerated Topology Reconfiguration for Dynamic Satellite Constellations NEURIPS2026
链接: https://arxiv.org/abs/2609.33993
作者: João Norberto,Ricardo Ferreira,Cláudia Soares
类目: Machine Learning (cs.LG)
*备注: Published at NeurIPS 2026
Abstract:Dynamic topology reconfiguration is central to the reliability and efficiency of large satellite constellations, yet many existing approaches rely on idealized assumptions such as full constellation deployment or uniform orbital spacing. We present Adaptive Satellite Topology via Regret-Aware learning (ASTRA), a theoretically-grounded framework for dynamic satellite topology reconfiguration that builds on an online learning formulation and makes it computationally practical. ASTRA combines an ADMM-based offline solver with efficient online updates for both online gradient descent and online conditional gradient, yielding markedly cheaper constrained updates than generic optimization pipelines. On the theory side, we show that for a relevant class of entry-wise nonzero utility matrices, the objective is strongly convex, which yields logarithmic static regret for online gradient descent, and we further instantiate known dynamic-regret guarantees under inexact ADMM inner loops. Empirically, ASTRA matches or improves topology quality, presenting a good trade-off with computational time on synthetic constellations, and it remains effective on real Starlink data under partial deployment and non-uniform spacing, where idealized structural assumptions break down. These results position ASTRA as an efficient and theoretically grounded approach to topology reconfiguration in realistic Low Earth Orbit networks.
[LG-190] ICMAPE: In-Context Multiagent Pure Exploration
链接: https://arxiv.org/abs/2609.33986
作者: Xinyi Hu,Alessio Russo,Aldo Pacchiano
类目: Machine Learning (cs.LG)
*备注:
Abstract:In some multi-agent systems, the quantity to be optimized is not an externally specified reward but the information acquired about unknown properties of the environment as done in active sequential hypothesis testing (ASHT) problems. However, the ASHT literature tends to focus on finite single-agent problems with well-specified models, while there is currently a gap for practical multi-agent methods that can perform active sequential testing. We fill this gap with ICMAPE, a Bayesian learning-based framework for decentralized multi-agent pure-exploration driven by inference objectives. ICMAPE converts the fixed-confidence identification objective into a reward derived from inference confidence, so that standard reinforcement learning machinery can be applied to decentralized pure exploration. It jointly learns a centralized neural inference network that estimates a posterior distribution over hypotheses from global trajectory data, and decentralized policies that select actions from local observation histories and learn when to stop collecting data once the target confidence is reached. On two synthetic benchmarks and a Maryland nitrate concentration monitoring task based on real-world data, ICMAPE-TD3 achieves target accuracy with fewer exploration steps.
[LG-191] he Privacy Fallacy of Crowdsourced Fine-Tuning: Extracting Proprietary Data via Topic-Based Poisoning
链接: https://arxiv.org/abs/2609.33985
作者: Sae Furukawa,Alina Oprea
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注:
Abstract:Supervised fine-tuning (SFT) is widely used to adapt large language models to downstream tasks. Crowdsourcing user conversations is an established approach to collecting SFT data at scale while reducing the need for costly manual annotation. However, it also allows untrusted users to contribute data to the fine-tuning pipeline. We investigate an underexplored privacy risk arising from this setting: can a malicious user poison a small fraction of the crowdsourced data to amplify extraction of previously unseen instructions contributed by other users? We show that this is possible using only black-box, output-only access to the deployed model. Experiments across four models and two datasets demonstrate substantial increases in training-data extraction: with only 50 poisoned examples, near-verbatim extraction reaches 3.71\times the rate without poisoning for Qwen2.5-14B on OpenMathInstruct and 3.08\times for Llama-3.1-8B on AceReason. Data filtering also proves largely ineffective in detecting poisoned samples: even the best-performing method achieves only 0.378 in F-1 score, leaving the majority of poisoned samples undetected. These findings demonstrate that seemingly benign crowdsourced contributions can amplify leakage of other records while remaining difficult to identify through data filtering.
[LG-192] From HL to HL-1 Parameters: A Hankel-Toeplitz Forecaster for Long-Term Time Series Forecasting ICASSP2027
链接: https://arxiv.org/abs/2609.33984
作者: Chaoqi Zhang,Yu Wang,Haixu Tang
类目: Machine Learning (cs.LG)
*备注: 5 pages, 2 figures, 2 tables. Submitted to IEEE ICASSP 2027
Abstract:Linear forecasters have shown competitive accuracy against Transformer-based models in long-term time series forecasting. We study how classical stationary prediction theory can guide parameter sharing for more compact linear forecasters. For centered second-order stationary processes with nonsingular history covariance, the minimum-MSE finite-window linear predictor factors into a Hankel cross-covariance matrix and an inverse Toeplitz covariance matrix. Shared lags and scale cancellation specify this predictor using H+L-1 autocorrelations for lookback L and horizon H . Building on the innovations representation, our Hankel-Toeplitz Forecaster (HTF) learns one impulse response that defines both an inverse filter and a forecast map. We characterize the finite-history correction and, under summability assumptions, bound the excess risk of truncating the true filters. HTF uses H+L-1 trainable coefficients while allowing a full-rank forecasting matrix. Across seven benchmarks at L=336 , its horizon-averaged MSE is within 1.2% of Dense Linear on each dataset with 75-229 times fewer trainable parameters.
[LG-193] Future Information-Directed Sampling for Bayesian Nonstationary Bandits ICML2026
链接: https://arxiv.org/abs/2609.33981
作者: Yichen Song,Alessio Russo,Aldo Pacchiano
类目: Machine Learning (cs.LG)
*备注: 18 pages. An earlier version appeared at the ICML 2026 DEMO Workshop
Abstract:Exploration–exploitation is a central trade-off in bandit learning. While classical algorithms such as upper confidence bound methods and Thompson Sampling effectively balance this trade-off in stationary environments, their exploration strategies mainly reduce uncertainty about the current optimal arm, which can be insufficient in nonstationary settings where future optimal arms may differ substantially from current ones. In this paper, we propose Future Information-Directed Sampling (FIDS), a new algorithm for Bayesian nonstationary bandits that explicitly explores to gather information about future optimal arms. We show that FIDS achieves regret comparable to Thompson Sampling up to a small constant factor, while being able to exploit predictive information structures that conventional exploration objectives fail to capture. To address the practical difficulty of posterior inference, we further propose a supervised-learning-based approximation framework that learns the FIDS policy from offline data, and demonstrate its effectiveness on synthetic benchmarks.
[LG-194] DynGraphAgent Bench: A Benchmark for Agent ic Lifecycle Control in Dynamic Graph Anomaly Detection
链接: https://arxiv.org/abs/2609.33980
作者: Yuwei Han,Lingwei Wei,Wooseong Yang,Liangjie Huang,Liancheng Fang,Huanhuan Ma,Philip S. Yu
类目: Machine Learning (cs.LG)
*备注: 16 pages, 2 figures, 5 tables
Abstract:Dynamic graph anomaly detection requires repeated decisions as graph structure and class prevalence drift, yet detector benchmarks usually score a fixed pipeline after current labels are known. We introduce DynGraphAgentBench, an executable benchmark for agentic lifecycle control under delayed feedback. It comprises seven temporal graph datasets with node- and edge-level anomaly tasks, eleven selectable detectors, and eight chronological deployment windows per dataset. In each window, a controller sees only time-causal aggregate context, registered model cards, and its own matured history. It must choose a detector before current-window training or candidate scores exist. A sandboxed executor trains the chosen architecture on mature data, scores a hidden deployment window, and releases the outcome after a one-window delay. A deterministic verifier checks decision timing, leakage guards, legal actions, training scope, and persisted artifacts. We measure detection utility with average precision and capture at fixed review depth, and characterize adaptation through model switches and compute. Complete eight-window trajectories from two primary controllers and a no-memory reference on four datasets, together with three additional controllers on three datasets, expose useful, costly, and ineffective reactions to delayed evidence without granting an exhaustive current-window oracle.
[LG-195] Adapting neural operators for mechanics decisions under changing operating conditions
链接: https://arxiv.org/abs/2609.33978
作者: Prashant K. Jha,Koffi Enakoutsa,Ian Galloway,Henry Anderson
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注: 33 pages, 16 figures
Abstract:Neural operators can accelerate repeated nonlinear mechanics calculations, but their accuracy can deteriorate as operating conditions move beyond the training range. This work studies whether high-fidelity solutions acquired during use can be reused to adapt a neural operator and improve subsequent mechanics-based command selection. Two hard-magnetic soft-material systems are simulated using high-fidelity finite-element (FE) models, providing reference solutions for evaluating surrogate predictions and selected commands. A neural operator predicts deformation from known material, loading, and magnetic-field inputs, while an empirical error estimator determines which predictions may be used for command selection. Selected FE evaluations supplement these predictions, and their complete loading paths are retained for periodic updates of the neural operator and estimator. In both examples, the fixed operator loses substantial accuracy when stiffness and loading move outside the training range. Updates using 16 acquired paths recover much of the lost accuracy while preserving accuracy in the nominal regime. Under the same FE evaluation budget, the updated operators also improve command selection, although the benefit varies with the operating condition. Error estimation is less consistent, with inaccurate predictions sometimes accepted and accurate predictions rejected. These results demonstrate that reusing high-fidelity loading paths can extend the useful operating range of a neural operator. However, improved forward accuracy alone does not guarantee reliable prediction acceptance, highlighting prediction-specific error assessment as a separate requirement for trustworthy decision making.
[LG-196] Greenpixies AI Token Methodology: Assessing the Energy Water and mathrmCO_2text-eq Impact of AI Tokens for Open and Closed Weight Models
链接: https://arxiv.org/abs/2609.33965
作者: Joshua Horswill,Ross Hunter,Matt Clifford,James Hall
类目: Computers and Society (cs.CY); Machine Learning (cs.LG)
*备注: 25 pages, 12 figures
Abstract:We describe a methodology for estimating the per-token energy cost of cloud-hosted large language model (LLM) inference, separating between input (prefill) and output (decode) tokens. Graphics processing unit (GPU) energy usage is measured during inference benchmarking with open-weights models on a wide range of text-based tasks. The remaining server energy contribution from non-GPU hardware is estimated from the inference wall time. Bayesian linear regression is used to model the relationship between energy per token and LLM size, request traffic, and hardware deployment configuration. Proprietary frontier LLMs of unknown size and deployment are binned into size buckets based on naming conventions and performance priors, and the space of possible LLM configurations is sampled with Monte-Carlo methods to give a representative average energy per token and uncertainty. We also describe how these energy measurements can be used to estimate the carbon-dioxide equivalent ( \mathrmCO_2\text-eq ) emissions, both usage and embodied, and water consumed per token of AI inference. This methodology provides actionable data that enables reductions in cost, electricity usage, \mathrmCO_2\text-eq emitted and water consumed in cloud and Software as a Service (SaaS).
[LG-197] Behavioral Monitoring of JEPA World Models with Jacobian Centroids
链接: https://arxiv.org/abs/2609.33940
作者: Thomas Walker,Randall Balestriero,Richard Baraniuk
类目: Machine Learning (cs.LG)
*备注: Presented at Workshop on World Models - Hosted by Chicago Booth
Abstract:Detecting failures in World Model (WM)-based planning requires monitoring whether the model is behaviorally aligned with the current task, which in turn requires studying its internal representations. Here, we show that centroids—sub-component Jacobian row-sums—effectively identify the behavioral properties of WMs, complementing traditional activation-based knowledge signals. The centroids of a model are easily computed through Jacobian vector products and characterize how the model organizes the geometry of its input space, yielding an efficient perspective on internal representations, including the generation of task-relevant saliency maps. Evaluated on continuous control tasks using JEPA WMs, this behavioral view reveals a structural dissociation, where the encoder correctly represents the goal while the predictor remains behaviorally unresponsive. This failure mode directly predicts planning failure before any action is taken, allowing for goal resampling to recapture out-of-distribution success. Moreover, centroid-based methods outperform baseline methods as distribution-shift detectors. Together, these tools yield a behavioral monitoring stack that is operational and consequential under distribution shifts.
[LG-198] How Strong Is the Evidence for the Artificial Hivemind? Reevaluating Evidence for the Open-Ended Homogeneity of Language Models
链接: https://arxiv.org/abs/2609.33936
作者: Rylan Schaeffer,Brando Miranda,Joshua Kazdan,Jessica Chudnovsky,Sanmi Koyejo
类目: Machine Learning (cs.LG)
*备注: 43 pages
Abstract:Recent research argues that language models exhibit pronounced homogeneity in open-ended generation, framing such behavior as an Artificial Hivemind that poses a long-term threat to human creativity. We examine three of its central results. First, the flagship example is that model responses to “Write a metaphor involving time” collapse into two clusters. Visualization, spectral analysis, clustering, and language model labels all contradict this description. The labels record each response’s vehicle, what it compares time to. Our responses and the original authors’ own show one dominant vehicle plus a heavy tail of distinct minority vehicles. “Time” is one of our least diverse topics, so the example is a favorable case, not a representative one. Second, the paper measures homogeneity against an undemanding null: responses to unrelated prompts. Under a more demanding null (same-prompt responses expressing genuinely different ideas), 20%-32% of such pairs already exceed the paper’s 0.8 convergence threshold. A residual effect survives this null. The paper’s same-prompt pairs exceed 0.8 roughly two to three times as often as our different-idea pairs. Much of what the paper calls homogeneity is the shared geometry of answering the same prompt. The remaining measurements lack any null: no human baseline is collected, and the model-indistinguishability statistic has no null. Third, the paper concludes that inference-time interventions are inadequate for combating the Artificial Hivemind, writing that “more generalizable solutions are needed at the model training level.” We show that this conclusion is unsupported in three ways, and that an inference-time intervention (prompting) reliably raises measured response diversity. We do not resolve whether the Artificial Hivemind is real. We show that the published evidence does not establish it.
[LG-199] Optimizing the Phi-2 Small Language Model for Real-time Chatbot Applications Using Parameter-Efficient Fine-Tuning (PEFT) with QLoRA Quantization
链接: https://arxiv.org/abs/2609.33927
作者: PhanTan Khanh Nguyen,Ashfaq Ali Shafin,Khandaker Mamun Ahmed
类目: Machine Learning (cs.LG)
*备注:
Abstract:This study explores the optimization of the Phi-2 Small Language Models (SLMs) for real-time chatbot applications through Parameter-Efficient Fine-Tuning (PEFT) and Quantized Low-Rank Adaptation (QLoRA). QLoRA specifically refers to the integration of PEFT with LoRA alongside a 4-bit quantization process, aimed at enhancing computational efficiency. These models, initially designed for high performance with minimal computational overhead, are further refined to address the constraints of mobile and edge computing environments. By integrating PEFT with QLoRA, the research aims to reduce memory usage significantly while maintaining, or potentially improving, the accuracy of model responses in real-time interactions. The effectiveness of these techniques was evaluated using the ROUGE metric system, which showed notable improvements in the summarization tasks performed by the models. This approach not only confirms the feasibility of using SLMs in resource-restricted environments but also opens up new avenues for deploying advanced AI-driven applications in real-time settings. The study’s findings have significant implications for the development of efficient, scalable, and accessible AI technologies, paving the way for broader adoption in various industries.
[LG-200] raining Witnesses: Trusting the Training without Trusting the Trainer
链接: https://arxiv.org/abs/2609.33915
作者: Houjun Liu,Pratyusha Sharma
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注: 10 pages, 15 appendix pages, 17 figures
Abstract:Progress in machine learning cannot outpace our ability to verify it. With an explosion in papers today, every scientific claim rests initially on trust in the trainer, leading to uneven evaluation, baselines, and forestalling of reliable progress. Traditionally, the burden of verification falls on the reader, who must reproduce expensive training runs. This strategy is impractical due to an explosion in slop contributions, diversity of methods, and the sheer compute required. We put the burden of proof where it belongs, on the trainer, and in the process also cut the overall cost of verification significantly. We introduce Witnesses, a method for certifying training, data usage and evaluation in a neural network training run. Our key insight is that fast behavioral fingerprints with occasional replay challenges are sufficient for auditing neural network training. Our method is applicable at scale with minimal overhead to the trainer, is cheap for the verifier, rejects bad training runs with amplifiable probability, and allows for exact queries of both data inclusion and exclusion. We test our method on language model training runs from 100M to 2B scales, across DDP and FSDP, and demonstrate this minimal overhead. We also introduce a self-regulating leaderboard of “auto-certified” training runs that enables shared baselines and progress. We invite the community to participate in the leaderboard to improve reproducibility in machine learning.
[LG-201] On the Two Faces of Adam in Separable Linear Classification
链接: https://arxiv.org/abs/2609.33904
作者: Chen Fan,Csaba Szepesvári
类目: Machine Learning (cs.LG)
*备注:
Abstract:We consider the behavior of deterministic, full-batch, bias-corrected Adam in separable linear classification with softmax parametrization under log-loss. In this setting, under a wide range of conditions Adam is known to approach max-norm-margin optimality when its stability constant \epsilon is zero, while with a positive \epsilon , it is known to approach Euclidean-margin optimality. Our main contribution is the quantitative description of Adam’s behavior for small fixed positive \epsilon . We give sufficient conditions under which an Adam-trained classifier nearly maximizes the max-norm margin before the updates become gradient-like. We also show that the classifier reaches a fixed target Euclidean margin only much later. Specifically, we show that for polynomially decreasing stepsizes with exponent (a), where (1/3a1), the updates become approximately proportional to the negative gradient after \Theta(\log(1/\epsilon)^1/(1-a)) iterations. At that time, the classifier still nearly maximizes the max-norm margin. Reaching a fixed target Euclidean margin above that of every max-norm-optimal classifier, but below the optimum, is shown to require \epsilon^-\Theta(1)/(1-a) iterations. Under inverse-linear stepsize decay ((a=1)), the update transition takes polynomially many iterations, whereas reaching the target margin takes exponentially many. Experiments support these predictions. The later change in the classifier can improve or worsen generalization after training error reaches zero, connecting the analysis to grokking and its reverse. Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.33904 [cs.LG] (or arXiv:2609.33904v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.33904 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Chen Fan [view email] [v1] Sun, 27 Sep 2026 20:37:36 UTC (704 KB) Full-text links: Access Paper: View a PDF of the paper titled On the Two Faces of Adam in Separable Linear Classification, by Chen Fan and 1 other authorsView PDFHTML (experimental)TeX Source view license Current browse context: cs.LG prev | next new | recent | 2026-09 Change to browse by: cs References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) IArxiv recommender toggle IArxiv Recommender (What is IArxiv?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab) Major funding support from
[LG-202] No Free Efficiency: Revisiting the Trade-off Between Training Efficiency and Model Vulnerability
链接: https://arxiv.org/abs/2609.33898
作者: Yiyong Liu,Jun Sakuma,Michael Backes,Rui Wen
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注:
Abstract:Training efficiency has become the central driver of recent progress in foundation models. To overcome the massive computational and data requirements of large-scale training, researchers increasingly adopt strategies such as selective data sampling, efficient pre-training, and simplified reinforcement learning pipelines. While these strategies drastically reduce overhead, they prompt a critical, yet neglected question: Is efficiency achieved at the expense of model robustness and security? To our knowledge, we present the first systematic cross-domain investigation of the efficiency-vulnerability trade-off. Across vision and language models, we show that efficiency-oriented training increases susceptibility to adversarial and privacy attacks. We characterize this vulnerability by analyzing the models’ internal geometry and functional representations, demonstrating that the evaluated efficient variants consistently exhibit sharper loss geometry together with systematic changes in representational structure. We further extend our analysis to “zero RL training”, finding that models trained using simplified RL recipes exhibit substantially greater susceptibility to catastrophic forgetting and more pronounced overconfidence than those trained through conventional alignment pipelines. Our findings suggest that training efficiency is rarely a “free lunch”; rather, the mechanisms that minimize computation can inadvertently compromise safety. We conclude by calling for a paradigm shift toward multi-objective training that jointly optimizes for performance, cost, and security.
[LG-203] MISHAP-Bench: A Hallucination Benchmark for Large Audio-Language Models
链接: https://arxiv.org/abs/2609.33893
作者: Zhi Wen Soi,Giulio Segalini,Jian-Jia Chen,Lydia Chen
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large audio-language models (LALMs) produce fluent responses about audio but often hallucinate by making plausible yet ungrounded claims. Existing audio hallucination benchmarks mainly measure response correctness, leaving it unclear whether an LALM hallucinates or simply fails to understand the audio. We challenge correctness-based evaluation by defining two hallucination categories: (i) context, where claims are not grounded in the audio; and (ii) knowledge, where claims about audio-related topics lack support from externally verifiable facts. We introduce MISHAP-Bench, a comprehensive benchmark with 12,000 challenging open-ended question-audio pairs and a rigorous evaluation pipeline covering both categories. To evaluate open-ended responses, we propose a groundedness judge that uses reference rubrics and judge prompts guided by human annotations. We extensively evaluate ten state-of-the-art LALMs and show that hallucination remains substantial. Even a frontier model such as Gemini 3.7 Flash reaches a hallucination rate of 36.5%. We further adapt and benchmark four mitigation methods from multiple domains for LALMs. Despite some improvements, effective hallucination mitigation remains an open challenge. Finally, we call on the community to evaluate hallucination and benchmark mitigation methods with MISHAP-Bench.
[LG-204] Augmented Feature Boosting for Multicalibration
链接: https://arxiv.org/abs/2609.33891
作者: Ira Globus-Harris,Inbal Livni Navon
类目: Machine Learning (cs.LG)
*备注:
Abstract:Multicalibration requires a predictor’s residuals to be unbiased not only globally, but also after conditioning on the predictor’s own level sets and reweighting by a rich class of test functions. Standard boosting approaches in the distributional setting achieve this by repeatedly discretizing the predictor’s range then auditing and repairing the resulting level sets. One consequence is that in practice, the algorithm’s guarantees are sensitive to this parametrization of the rounding parameter. A natural theoretical question, then, is how to do discretization-free boosting which avoids this rounding within the boosting process itself. Here, we analyze an alternative feature-augmentation boosting paradigm inspired by Tax et al. (2026): at each round, a squared-loss oracle is called on hypotheses that receive the previous predictor’s output as an additional feature, and only the final predictor is rounded to have a finite set of level sets to provide the multicalibration guarantee with respect to. We give a theoretical analysis of this procedure through the expressivity of the augmented hypothesis class, and show how the expressivity of this class yields a hierarchy of guarantees, including multiaccuracy, multicalibration, and the stronger notion of level-set multicalibration.
[LG-205] Vanilla Policy Optimization Is Both Optimal and Differentially Private for Stochastic Contextual Bandits
链接: https://arxiv.org/abs/2609.33888
作者: Idan Attias,Orin Levy,Alexander Ryabchenko,Yishay Mansour,Uri Stemmer
类目: Machine Learning (cs.LG)
*备注:
Abstract:Can vanilla policy optimization explore enough to achieve near-optimal regret in stochastic contextual bandits? We show that standard exponential policy updates driven by offline regression do so under realizability, without exploration bonuses or importance weighting. For A actions, T rounds, and a finite prediction class F , vanilla PO achieves \widetilde O(\sqrtAT\log(|F|)) regret with high probability. Our analysis reveals an implicit exploration mechanism of independent interest: gradual policy updates prevent actions from losing probability too quickly, allowing the regression oracle to learn their expected losses. We further develop a batched version using only O(\log T) regression calls and policy switches, and show how private regression oracles yield differentially private contextual bandit algorithms without composition across batches. For a finite class, this gives pure \varepsilon_\rm priv -DP and regret \widetilde O\left( \sqrtAT \log(|F|/\delta)(1+\varepsilon_\rm priv^-1/2) \right) . Finally, experiments across oracle-based contextual bandit algorithms, with and without privacy, demonstrate the practical effectiveness of policy optimization and the value of explicit exploration under stronger privacy constraints.
[LG-206] JET: Justification Evaluation in Transformer
链接: https://arxiv.org/abs/2609.33874
作者: Shenghao Ding
类目: Machine Learning (cs.LG); Performance (cs.PF)
*备注: 11 pages, 2 figures. Code: this https URL
Abstract:JET uses pretrained language and vision-language models to select among a finite set of answers without additional training. It evaluates candidate likelihoods directly and shares computation across candidates. Experiments on desktop CPUs and consumer GPUs assess decision accuracy and execution cost. Qwen3.6-35B-A3B achieves 87.48% accuracy on the full MMLU test set and 3.69 requests per second on a separately timed MMLU subset. The accuracy-throughput comparison covers model, hardware, and reasoning choices, with Jev as an external reference. Controlled execution experiments show 2.18-2.23-fold speedups from prefix reuse and cache management, and a 30.8% reduction in process time from input preparation optimizations, with unchanged outputs. Optional reasoning has a task-dependent accuracy-throughput trade-off. These results support local decision inference from existing models.
[LG-207] Qwen Gyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon Agents
链接: https://arxiv.org/abs/2609.33848
作者: Weiqi Wang,Yuxin Zhou,Mouxiang Chen,Siyuan Zhang,Yi Zhang,Yuyan Luo,Zhiyu Yin,Chencan Wu,Jiemin Jiang,Wentao Yao,Chujie Zheng,JianWei Zhang
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注:
Abstract:Large language model (LLM) agents increasingly undertake extreme-long (xlong) horizon tasks, where a single execution can span hours, hundreds of model–environment interactions, and nearly 1M tokens per rollout. Applying online reinforcement learning (RL) to such executions poses two fundamental challenges: (1) severe execution variance and prolonged rollout delays cause massive GPU idling; and (2) complex non-linear branching generates massive trajectory redundancy, crippling training efficiency. To address these, we presents QwenGyre, an end-to-end framework for xlong-horizon online RL. QwenGyre elastically reallocates GPUs between rollout and training without interrupting live executions, while its trajectory processor reconstructs branching histories, scores partial progress, and deduplicates redundant paths to bound training costs. Scaled to our flagship model, Qwen~3.8 2.4T, with 700K tokens per rollout, QwenGyre yields a 6.0% absolute gain on NL2RepoBench (52.5% \to 58.5%) in 48 steps. Across our evaluations on diverse domains of training datasets, QwenGyre delivers up to 1.85\times and 1.78\times speedups over Colocate and Async, respectively.
[LG-208] dOPT: Differentiating Conic Optimization via Geometric Reduction
链接: https://arxiv.org/abs/2609.33828
作者: Fengyu Yang,Connor W. Magoon,Tyler Watts,Shahar Z. Kovalsky
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Optimization layers enable the incorporation of structured constraints and decision problems into learning systems. Training such systems requires differentiating through the embedded optimization problem, which can be challenging for general conic programs. We introduce dOPT, a solver-agnostic framework that, rather than differentiating the full conic formulation, reduces it at a computed primal-dual solution to an equality-constrained quadratic program that preserves the reference solution and its first-order sensitivity. The reduction captures the local first- and second-order conic geometry relevant to differentiation and remains well defined at singular configurations. Computing solution derivatives then requires a single symmetric linear solve, independently of the forward solver. We derive explicit reductions for convex NLPs, QPs, SOCPs, and SDPs. Numerical experiments validate the computed gradients and show favorable backward-pass scalability, with substantial speedups over existing differentiable conic optimization methods as problem size increases.
[LG-209] Identical Runs Different Results: Benchmarking AI Coding Agents on Open-Weight Models
链接: https://arxiv.org/abs/2609.33812
作者: Eduardo Ariño de la Rubia(Central European University),Szilard Pafka(Epoch)
类目: oftware Engineering (cs.SE); Machine Learning (cs.LG)
*备注: 22 pages, 6 figures, 18 tables. Data and code: this https URL
Abstract:Repeated runs of the same coding agent are known to give different benchmark scores. We ask what that variation means for a team running an agent on its own task, by intensive replication on one machine-learning task: an agent improves the training code of an XGBoost classifier for airline delays, and a holdout it never sees scores the result. Across 584 runs, we compare six agents on six open-weight model endpoints, run six agent-model pairings 52 times each under fixed settings, and repeat three of them on a larger model from the same family. Identical runs of one pairing varied more than the pairings differed from one another, so comparisons of a few runs ranked them unreliably; resolving the agent differences we observed would take tens to more than a hundred runs of each. Runs on the larger model scored clearly higher, but by less than one run-to-run standard deviation, and the gap was more than twice as large with one agent as with the others. Fewer than one run in twenty broke the task’s data rules, but those runs held the highest scores. Rejecting those runs first and keeping the best compliant result among a few attempts reliably improved the delivered model, even though a few runs could not rank the agents. On flights from a later year, the delivered models kept only a third of their gain over the starting code. At list prices, cost differed more than twentyfold between two agents on the same model, mostly through the prompt cache. Agents and models should be evaluated as pairings, over repeated attempts, with compliance reported beside quality. Data, code and every delivered program: this https URL
[LG-210] Binding Multiple Modalities via Multimodal Wasserstein Barycenter NEURIPS2026
链接: https://arxiv.org/abs/2609.33800
作者: Xiaole Tang,Jiayi Xu,Xiang Gu,Yan Yang,Jian Sun
类目: Machine Learning (cs.LG)
*备注: NeurIPS 2026 Oral
Abstract:Multimodal learning beyond two modalities commonly leverages a specific modality (e.g., text) to bind other modalities. However, how to establish a more balanced representation space that approximates shared semantics while respecting the holistic geometry of n -modal data remains challenging. In this work, we present BaryBind, which aims to transport the specific modality towards the Wasserstein barycenter (WB) optimized across all modalities and introduces a volumetric alignment objective to establish a unified semantic space around the WB embedding. Specifically, we project specific modalities to the WB, which minimizes the average Wasserstein distances to multimodal distributions and serves as the anchor for subsequent alignment. We then construct a barycenter simplex, whose volume is taken as a similarity metric for global alignment centered at the WB. Experiments show that BaryBind achieves competitive performance in text-video-audio retrieval, classification, videoQA, and cross-modal generation tasks, along with robustness under modality absence and scalability to more than three modalities. Code is released at this https URL.
[LG-211] RAISE: Reinforcing Access Control Policy Synthesis in LLM s via Symbolic Evaluation
链接: https://arxiv.org/abs/2609.33796
作者: Yingming Zhou,Adarsh Vatsa,William Eiers
类目: oftware Engineering (cs.SE); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 34 pages, 5 figures, 11 tables. Code: this https URL
Abstract:Translating natural-language access-control requirements into policies requires careful reasoning about permissions, constraints, and exceptions, and even frontier LLMs often produce policies that violate the intended authorization semantics. We construct CedarInstruct, to our knowledge the first dataset that supports both training and semantic evaluation for formally verifiable Cedar policy synthesis. It contains 5,800 scenarios across 44 domains and 1,408 representing a single synthetic organization, each with a verified target policy and an executable verification plan. On this data we introduce RAISE, which trains policy synthesizers from formal verification in two stages, verified supervised fine-tuning (SFT) followed by a reinforcement learning (RL) stage that learns from verifier signal. We find that SFT succeeds largely by letting models express authorization logic they already have, since untrained models rarely write valid Cedar but often reason correctly when they do. After SFT, how the verifier’s information is used matters more than how much of it is used. Of six RL instantiations that consume progressively richer verifier signal, only RAISE-OC improves meaningfully on SFT; it turns failed checks and symbolic counterexamples into guided exploration and learns from the result with off-context GRPO. With about 5.4K verified scenarios and LoRA fine-tuning, RAISE-OC trains Qwen3.5-9B to surpass zero-shot GPT-6 Astra and Claude Opus 5 by 13.33 and 16.26 percentage points in semantic success on held-out scenarios, and training transfers to the independently constructed CedarBench.
[LG-212] EfficientAgent : What Makes KV Cache Offloading Work for Concurrent Agents ?
链接: https://arxiv.org/abs/2609.33762
作者: Kunming Shao,Jierun Chen,Jiangnan Yu,Xiao-Hui Li,Chaofan Tao,Yanli Wang,Huanxin Lin,Kwang-Ting Cheng,Chi Ying Tsui,Haoli Bai
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: 25 pages, 8 figures, 13 tables. Code: this https URL
Abstract:LLM agents resend their whole conversation on every turn, and most of it was already processed on the previous turn. Serving systems avoid recomputing it by caching its key-value (KV) state and, when GPU memory runs out, by offloading that state to host memory. For agents, offloading gives inconsistent results: on the same coding-agent workload it speeds up one deployment, slows down another, and changes nothing on a third, even where loading a token back is several times cheaper than recomputing it. The reason is that cached state must survive until it is used again. While one agent waits for its tool, the server processes the contexts of all other agents, so an agent’s prefix is reused only if the host tier holds the reusable context of the whole agent pool, which we call the reuse working set. A smaller tier keeps writing state that is evicted before anyone reads it. We present EfficientAgent, which sizes and manages the host tier by this working set. A stack-distance model estimates the working set from agent histories to size the host tier; its predictions, made before the experiments, located the capacity at which offloading starts to pay. When the tier is too small, a runtime policy stops writing large refills of evicted context and keeps extending prefixes that are still cached; when the tier is large enough, it writes everything. On SWE-bench Verified coding agents, a host tier sized to the estimated working set cuts recomputed prompt tokens by 93% and end-to-end time by 39%. With a small fixed tier, the policy cuts recomputation by 35%; with a large tier, it avoids the 4.3-fold increase caused by always filtering writes. Across three GPU types and two models, offloading pays off when the GPU has little compute per byte of host bandwidth and the host tier holds the working set. Code is available at this https URL.
[LG-213] StatD2GAN: When Calibration Masks Generator Quality in Held-Out Evaluation of Synthetic Weather Sequences
链接: https://arxiv.org/abs/2609.33761
作者: Mustafa Ozaytac,Ozge Karadag Atas
类目: Machine Learning (cs.LG)
*备注: Submitted to Neurocomputing (Elsevier). Derived from the first author’s Master’s thesis
Abstract:Generative models for multivariate weather series are routinely evaluated with pooled distributional metrics computed after marginal calibration. We show this practice can invalidate architectural conclusions, and rebuild the evaluation of StatD2GAN, a three-discriminator GAN with evolutionary weight adaptation, around a held-out protocol: the final two calendar years of each dataset are held out behind a 168 hour embargo, calibration is fitted on the training block only, and all metrics are computed on the held-out block. Evidence comes from 25 matched (location, seed) pairs across five Koppen-Geiger climates, tested with Wilcoxon signed-rank tests under Holm correction. Four results follow. First, isotonic calibration drives the Kolmogorov-Smirnov distance to within 2% of a per-location noise-and-shift floor for every architecture tested, including a deliberately weak RCGAN baseline, so calibrated marginal metrics cannot discriminate between architectures. Second, the sorted-representation discriminator is the only component whose removal significantly degrades cross-variable dependence (Kendall tau MAE +0.080, Holm p = 0.009), with a regime-dependent effect: near zero in Ankara, above 115% in Dubai and Yakutsk. A rank-transformed variant isolates the mechanism as quantile supervision of the marginals rather than copula matching. Third, physical constraint violations are injected by calibration, not the generator; projection removes them at negligible cost (deltaKS = 0.003). Fourth, pooled metrics conceal a collapse of between-sequence weekly-mean variability, a proxy for seasonal and regime diversity, in TimeGAN that only sequence-level statistics expose. We recommend floor-referenced marginal evaluation, matched-pair testing, and sequence-level variance decomposition as minimum requirements for calibrated generative pipelines.
[LG-214] Oracle-Efficient Online Classification with Stochastic Inputs and Adversarial Outputs
链接: https://arxiv.org/abs/2609.33760
作者: Gon Buzaglo,Elad Hazan
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We consider contextual binary prediction with i.i.d. contexts from an unknown distribution and adaptively chosen losses. We show that a simple Follow-the-Perturbed-Leader algorithm with Gaussian perturbation for each observed context achieves the optimal \widetilde O(\sqrtT\log N) expected regret for a class of N experts, while requiring one optimization-oracle call per round and no explicit enumeration of the class. For an infinite hypothesis class \mathcal H , the algorithm attains \widetilde O(\sqrtT\operatornameVC(\mathcal H)) regret. This resolves an open problem posed by Lazaric and Munos (2012), showing that hybrid classification is computationally as easy as statistical learning.
[LG-215] ResDiffFRG: Residual Diffusion for Multiple Appropriate Facial Reaction Generation
链接: https://arxiv.org/abs/2609.33749
作者: Shizhe Liu,Jiayan Gu,Xiangyu Kong,Siyang Song
类目: Machine Learning (cs.LG)
*备注:
Abstract:In dyadic human speaker-listener conversations, the listener’s facial reactions allows the speaker to accurately perceive the listener’s emotional states. Since human facial reactions are non-deterministic, the ability to generate multiple appropriate human-like facial reactions is crucial for realistic human-agent interactions. Although diffusion models are naturally suited to such one-to-many generation, existing diffusion-based Multiple Appropriate Facial Reaction Generation (MAFRG) methods attempt to denoise random Gaussian initialisations directly into multiple appropriate facial reactions (AFRs). These random initialisations are usually not well-aligned with the target listener facial reaction, which requires complex denoising trajectories from these initialisations, and subsequently creates substantial opportunities for deviations away from the range of trajectories leading to appropriate AFRs. Given the inherent mimicry between the human listener’s and speaker’s facial behaviours, we address the above denoising trajectory issue by leveraging this strong prior. Specifically, we propose ResDiffFRG, a novel diffusion-based MAFRG framework that explicitly anchors the diffusion process to the speaker behaviour by defining its diffusion target as the residual between the speaker anchor and an AFR. The denoiser only needs to model the comparatively small, reaction-specific residual needed to transform this anchor into an AFR, rather than reconstructing the complete reaction from an unstructured state. Extensive experiments show that ResDiffFRG achieves large improvements in correlation-based appropriateness over existing methods. Our denoising trajectory analysis showed that even at the start of the denoising trajectory, ResDiffFRG already achieves a higher facial-reaction correlation score than the Gaussian Diffusion baseline does after completing 60% of its denoising trajectory.
[LG-216] ask-Aware Discretization of Differentiable Logic Gate Networks
链接: https://arxiv.org/abs/2609.33747
作者: Thore Gerlach
类目: Machine Learning (cs.LG); Logic in Computer Science (cs.LO)
*备注:
Abstract:Differentiable logic gate networks (DLGNs) enable gradient-based training of highly efficient Boolean networks by relaxing discrete logic gates during training and discretizing them for inference. Standard approaches make this discretization decision locally, typically through argmax selection and confidence- or entropy-based convergence criteria. We show that local discretization can be task-suboptimal even for globally optimal relaxed solutions, with high gate confidence providing no general guarantee, and derive bounds relating task-aware gate selection to tractable interventions in the relaxed network. Motivated by these results, we study first-order downstream task information for progressive discretization and characterize when this local approximation is reliable. Experiments on convolutional DLGNs reveal a strong locality dependence: first-order scores become unreliable when directly optimized over nonlocal interventions, but accurately assess local argmax decisions for progressive freezing.
[LG-217] PQ-HSA: Reusing Product-Quantized Scores for Hybrid Sparse-Approximate Attention
链接: https://arxiv.org/abs/2609.33746
作者: Kunming Shao,Jierun Chen,Yanli Wang,Ruoyu Wang,Haoli Bai,Kwang-Ting Cheng,Chi Ying Tsui
类目: Machine Learning (cs.LG)
*备注: 19 pages, 6 figures, 9 tables. Code: this https URL
Abstract:At each decoding step a language model attends over the key-value (KV) cache of every earlier token, so at long context the attention call is bounded by memory bandwidth. Sparse attention reads only a subset of keys chosen by a cheap score estimate, and most methods give the unread tokens zero weight. The output then draws on only a small fraction of the KV cache, and accuracy drops at small budgets, most on tasks that aggregate information across the context. An inverted-file product-quantization (IVF-PQ) index over the cached keys computes an approximate score for every indexed token in order to rank them; after ranking, those scores approximate the attention logits of the tokens left out. PQ-HSA (hybrid sparse-approximate attention) attends the selected tokens with their original keys and values, and the unselected tokens, the background, enter the same softmax through those scores, summed per inverted list and multiplied by the list’s mean value. At 128K and a 1-2% retrieval budget, PQ-HSA is more accurate than Quest and SnapKV on Llama-3.1-8B and Qwen3-30B-A3B and stays close to full attention in macro accuracy; with the same selector, the background term raises macro accuracy on the 8B model from 0.71 to 0.83. In the same 128K setting, inside vLLM on one NVIDIA H20, the decode attention call runs 1.6x faster than the FlashAttention-3 kernel; the speedup grows with context length, and a cost model fitted on 8B to 30B models gives the context length at which it begins. A vLLM plugin runs PQ-HSA on two engine versions without changes to the engine source; code is available at this https URL.
[LG-218] ALDER: Discovering the Laws of a World by Acting in It
链接: https://arxiv.org/abs/2609.33728
作者: Teng Cao,Yu Deng,Quentin Delfosse,Kristian Kersting
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reliable world models should not only predict future states but express how actions change the world in an explicit, transparent and testable form, such as equations. Yet methods that rely on a fixed set of trajectories cannot distinguish equally good competing hypotheses, while searches over a fixed set of predefined candidates cannot discover equations outside the initial hypothesis space. We introduce ALDER (Action-guided Law Discovery, Evaluation, and Revision), a method that actively proposes novel experiments to test and revise models. Specifically, ALDER proposes parametric equations; a numerical optimizer fits their coefficients; an independent verifier tests these candidates on held-out data. To distinguish between competing valid hypotheses, a cost- and safety-aware selector queries interventions, in the form of novel experiments. The resulting counterexamples update the evidence ledger and guide the next structural revision, while incompatible laws are discarded. Across an in-house benchmark, ODE equation discovery tasks, and robotic experiments, ALDER discovers laws beyond its initial formula set, repairs failed model proposals, distinguishes fixed candidate models with fewer interactions, and improves out-of-distribution prediction. Furthermore, given a current state and a target, ALDER selects control actions by solving the inverse problem defined by its validated world model. Together, these results show that explicit equation-based world models can be tested and revised through interaction, then naturally used to guide goal-directed control.
[LG-219] heory Guided and Interpretable Neural Operator Design for Partial Differential Equation Learning
链接: https://arxiv.org/abs/2609.33715
作者: Zeyuan Song,Zheyu Jiang
类目: Machine Learning (cs.LG); Computational Engineering, Finance, and Science (cs.CE)
*备注: 37 pages
Abstract:Accurate numerical solutions of partial differential equations (PDEs) are crucial in numerous science and engineering applications. In this work, we introduce a novel neural PDE solver named AFDONet, which incorporates neural operator learning and adaptive Fourier decomposition (AFD) theory for the first time into a specifically designed variational autoencoder (VAE) structure, to solve a general class of nonlinear PDEs on smooth manifolds. AFDONet is the first neural PDE solver whose architectural and component design is fully guided by an established mathematical framework (in this case, AFD theory), turning neural operator design from an art to a science. Thus, AFDONet also exhibits exceptional mathematical explainability and groundness, and enjoys several desired properties. Furthermore, AFDONet achieves outstanding solution accuracy and competitive computational efficiency in several benchmark problems. In particular, thanks to its deep connections with AFD theory, AFDONet shows superior performance in solving PDEs on i) arbitrary (Riemannian) manifolds, and ii) datasets with sharp gradients. Overall, this work presents a new paradigm for designing explainable neural operator frameworks.
[LG-220] DuoOPD: Learning from Joint Teacher-Student Outcomes for Multi-Task On-Policy Distillation
链接: https://arxiv.org/abs/2609.33711
作者: Ao Yu,Weibo Gao,Heng Zhou,Linan Yue,Rui Li,Suyi Liu,Yu Yan,Yizhong Zhang,Qi Liu
类目: Machine Learning (cs.LG)
*备注:
Abstract:On-policy distillation (OPD) trains a student on its own responses with token-level feedback from a stronger teacher, yet the teacher can fail on questions the student already answers correctly, and how often each model succeeds varies across tasks. OPD ignores these outcomes and, on average, pushes down even the student’s correct responses; gating feedback by student correctness fixes the direction but uses the teacher in the same way whether or not it succeeded. We introduce DuoOPD, in which the student’s outcome sets the direction of feedback and the joint teacher-student outcome decides how the teacher supports it: when only the teacher succeeds, its verified answer becomes context for scoring the student’s failed response, and when only the student succeeds, a weight shared within the task reinforces the whole response. A single rule covers all four outcome combinations without task-specific settings. Across Qwen3 and Llama, DuoOPD outperforms all five baselines in mean macro accuracy, improving over OPD by 2.58 and 5.98 percentage points, and it also leads on two further task mixtures spanning scientific calculation, instruction following, and code generation. Ablations show that outcome-based direction alone stays near the gated baseline, while the joint-outcome designs supply most of the gain.
[LG-221] A Spectral Theory of Compositional Learning ICLR2027
链接: https://arxiv.org/abs/2609.33708
作者: Hugo Rydel
类目: Machine Learning (cs.LG)
*备注: 22 pages, 9 figures, and 2 tables. Under review at ICLR 2027
Abstract:How does compositional reasoning emerge during learning? We address this question by mathematically analyzing the learning dynamics of deep linear networks. We train these networks in structured synthetic environments and derive a theory linking the structure of experience to compositional learning. Our theory predicts when compositional inferences emerge, whether they are identifiable from the available evidence, and how new linking evidence can rapidly unlock previously unavailable inferences. These results provide a qualitative explanation for several phenomena observed in human cognition. They account for why a composition can fail despite knowing its premises, why similar compositions can emerge at different times, and how a single linking fact can suddenly enable many new inferences. Taken together, these findings establish a mathematical link between the statistical structure of experience and the development of compositional reasoning.
[LG-222] Non-Adaptive Learning of Sparse Erdős–Rényi Graphs via Affine Splitting
链接: https://arxiv.org/abs/2609.33704
作者: Hoang Ta
类目: Information Theory (cs.IT); Machine Learning (cs.LG); Probability (math.PR)
*备注:
Abstract:Graph learning from edge-detecting queries concerns the reconstruction of an unknown edge set on a known vertex set. Each query reports whether a specified vertex subset contains at least one edge. We study non-adaptive schemes, in which all queries are fixed before any outcomes are observed, with the goal of achieving exact recovery using few queries and fast decoding. For general graphs on n vertices with at most k edges, non-adaptive recovery requires \Omega(\min\k^2\log n,n^2) queries in the worst case, even when a small error probability is allowed. In this paper, we consider Erdős–Rényi ( \mathrmER ) graphs G\sim \mathrmER(n,q) , with expected edge count \bark=q\binomn2 . Our scheme uses O(\bark\log n) queries and achieves exact recovery in O(\bark\log n) decoding time with probability tending to one throughout the regime \bark\to\infty and \bark=o(n^2) . This improves the previous O(\bark^1+\delta\log n) decoding guarantee for any fixed \delta0 , while maintaining the same query order. The guarantee also extends beyond the previously studied regime \bark=\Theta(n^2\theta) with fixed \theta\in(0,1) . Our approach builds on the binary splitting method used in prior work, which organizes vertices into a hierarchy of successively smaller groups. We introduce three main changes: (i) we use random affine hash functions over a finite field to process each candidate pair in constant time; (ii) we apply the splitting procedure directly to the full graph, avoiding the need to combine solutions to multiple smaller graph-learning subproblems; and (iii) we bound the total decoding workload directly rather than deriving separate high-probability bounds on candidate counts at each level.
[LG-223] Geometric Inductive Biases for Semi-Supervised Equalization: The Constellation-Aware Transformer NEURIPS2026
链接: https://arxiv.org/abs/2609.33695
作者: Avi Caciularu
类目: Machine Learning (cs.LG); Information Theory (cs.IT); Signal Processing (eess.SP)
*备注: accepted to NeurIPS 2026
Abstract:Decoding signals over unknown channels with minimal pilot overhead is a critical challenge in next-generation communications. Existing deep learning approaches typically rely on generic encoders that struggle to model long-range temporal dependencies or efficiently capture the channel’s physical properties from scarce data. We argue that standard architectures suffer from agnostic estimation gaps, as they must implicitly learn the constellation geometry that is already known. We introduce the Constellation-Aware Transformer (CAT), a novel architecture that explicitly injects geometric inductive biases into the equalization process. CAT is composed of a stack of custom TransFIRmer blocks, which use an “early interaction” paradigm to co-process received signals and ideal constellation symbols. Each block features a split Feed-Forward Network that applies a Finite Impulse Response (FIR)-inspired filter for deconvolution and a parallel MLP for geometric refinement. We show that this design is structurally aligned with the optimal linear (MIMO Wiener) receiver: its attention can implement a matched-filter bank, and its bidirectional FIR branch provides the non-causal filtering that block MMSE equalization requires. In the semi-supervised setting, CAT needs fewer pilots than VAE and standard Transformer baselines: on two of our three ISI channels, it reaches a lower SER with 64 pilots than they do with 128.
[LG-224] -MoXAI: A Hierarchical Explainability Framework for Temporal Multimodal Data
链接: https://arxiv.org/abs/2609.33685
作者: Ali Inha,Mo Vali,Saaliha Vali,Pietro Liò,Meen-Yau Thum
类目: Machine Learning (cs.LG)
*备注:
Abstract:Artificial Intelligence (AI) models for temporal multimodal data have potential in healthcare and agriculture, but their opacity can limit trust and adoption. We introduce T-MoXAI (Temporal Multimodal eXplainable AI), a hierarchical framework explaining (1) when timepoints influence predictions, using temporal Shapley values; (2) which modalities contribute at those moments, using attention analysis; and (3) what features or image regions drive decisions, using gradient based attribution. A transformer based architecture handles irregular temporal sequences and heterogeneous data, generating all three explanation levels in under one second for interactive decision support. We evaluate the framework on two real world tasks: predicting IVF treatment outcomes from ultrasound sequences and clinical measurements (AUC 0.660 despite significant class imbalance), and forecasting wheat yield from temporal RGB imagery and phenotypic traits ( R^2 0.265 amid substantial environmental variability). Ablation studies indicate that temporal modelling is critical in both domains: removing it reduces performance to the equivalent of random guessing. Temporal ROAR experiments provide evidence that the explanations reflect the model’s reasoning process. With a unified, domain agnostic architecture and open source implementation, T-MoXAI provides a baseline for temporal multimodal XAI, addressing fragmentation in the field and supporting applications where understanding decisions is as important as predictive accuracy.
[LG-225] Benign Overfitting for General Norms and Distributions
链接: https://arxiv.org/abs/2609.33675
作者: Daniel Barzilai,Ohad Shamir
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Understanding why predictors can generalize despite interpolating noisy training data is a central puzzle in machine learning. Most work on such “benign overfitting” studies minimum-2-norm linear regression, reflecting the inductive bias of gradient descent. However, modern optimizers such as Adam and Muon use non-Euclidean update geometries, favoring solutions associated with other norms. Analyzing regression for non-Euclidean norms is substantially more difficult, with known results essentially limited to Gaussians. In this paper, we develop a method to analyze benign overfitting in linear regression for general norms and general (sub-Gaussian) distributions. As a special case, we prove that minimum-p-norm interpolation with p1 can benignly overfit even for non-Gaussian distributions, under suitable conditions. Perhaps surprisingly, for the 1-norm, benign overfitting does not hold in general for well-behaved (but non-Gaussian) distributions, showing that existing positive 1-norm results rely crucially on Gaussianity. Our proof analyzes the geometry of the dual optimization problem, using concentration and central limit tools to show it is approximately Euclidean in many high-dimensional cases.
[LG-226] Reachability is not enough: Diagnosing long-range behavior in GNNs
链接: https://arxiv.org/abs/2609.33674
作者: Filippo Maria Bianchi
类目: Machine Learning (cs.LG)
*备注:
Abstract:Graph neural networks (GNNs) are often called long-range because their architecture can connect distant nodes, but this does not show whether they use distant information correctly. We introduce a framework that measures how strongly inputs at each graph distance affect predictions and separates limitations due to architecture, finite approximation, training, and numerical execution. Our analysis shows that local message-passing can spread influence slowly, so a finite implementation may rely mainly on nearby inputs even when the ideal computation uses the whole graph. We also explain why mathematically equivalent filters can differ in how easily they are learned and how reliably they run. Across controlled tasks, models with similar architectural reach use distant information very differently, while low average error can hide failures on distant interactions. Together, these results show that long-range capability depends on learning to use information at the distances required by the task and preserving that use during computation.
[LG-227] COGNIT-Guard: Calibrated Standalone Direct-Decision Guardrails with Heterogeneous CPU-NPU Confidence Cascading under Explicit Latency and False-Positive Constraints
链接: https://arxiv.org/abs/2609.33671
作者: Hao Chen
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 7 pages, 2 figures, 5 tables, IEEE double-column format
Abstract:When must a foundation-model safety gateway generate tokens, and when should it directly output a calibrated decision? We study calibrated standalone direct-decision foundation models for real-time pre-ingestion safety guardrails, jointly addressing probability calibration, dual-use false-positive control, and heterogeneous CPU-NPU routing under explicit latency SLOs. Pre-ingestion guardrails must screen prompts prior to target-LLM prefill with low false alarms on benign compliance inquiries; however, shallow classifiers are brittle to phrasing shifts, hidden-state probes require coupling to a target LLM, and generative guards incur high decoding latency and dual-use false positives. We present COGNIT-Guard, coupling a validation-calibrated CPU fast gatekeeper with confidence-gated escalation to an NPU-resident 322M bidirectional direct-decision model (Laya-322M) under an asymmetric false-positive penalty. On the clean unseen DUCS-Bench test split ( N=607 ), COGNIT-Guard achieves 98.85% accuracy (McNemar p = 1.19 \times 10^-4 vs. ML), reduces benign FPR to 0.42% ( 1/238 ; Fisher’s exact p = 8.23 \times 10^-4 vs. ML), and attains 1.12% ECE and 0.0104 Brier score. On Huawei Ascend 910C NPUs, pure NPU inference runs in 21.77 ms mean latency (45.90 QPS), while the live serial CPU-NPU cascade ( \theta^*_\mathrmdeploy=0.70 ) achieves 41.63 ms mean latency (P50: 39.47 ms, 99.23% accuracy, 0.00% FPR). Evaluation on SafetyBench-ZH ( N=2,100 ) and comparison against a bi-encoder direct-decision baseline (CLM-8B) disentangle in-domain gains, OOD alignment tax (60.33% \to 56.81% on Laya; 55.10% on domain CLM-8B), and experience replay recovery, restoring OOD accuracy to 64.10%-65.05% and reaching 99.67%-99.84% in-domain accuracy with 0.00%-0.42% FPR.
[LG-228] Scalable Attribution and Control of Model Behavior During Training
链接: https://arxiv.org/abs/2609.33667
作者: Sleem Abdelghafar
类目: Machine Learning (cs.LG)
*备注:
Abstract:Attributing and controlling model behavior during training requires identifying each example’s contribution quickly enough to act before the next update. However, examples in the same training batch can produce similar behavioral changes, making their individual contributions difficult to distinguish. We address this ambiguity through mutual information, accounting for interference within the batch by quantifying how much the combined behavioral change reveals about each example’s contribution. We show that this mutual information is a logarithmic function of Behavioral Gradient Uniqueness (BGU). BGU gives the information measure its geometric interpretation. Our Batch-Space Ghost (BS-Ghost) algorithm makes these scores practical inside the training loop through shared computation in batch space, without storing model-sized example gradients. On a complete 1,000-example Qwen2.5-7B-Instruct workload, our BS-Ghost implementation adds 27 seconds (8.0%) to 5.5 minutes of ordinary training. Removal and retraining demonstrate that BGU identifies data that causally shapes final behavior. At each training step, signed information identifies which examples strengthen or weaken the target behavior, explaining how behavior develops during training. Signed information also enables cheap intervention during training: it predicts how changing example weights will affect behavior in the next update. We then use these predictions to choose weights that steer behavior toward a desired target. This makes our framework a practical foundation for scalable oversight and verification of training pipelines and processes, helping evaluators assess model alignment, understand how it develops during training, and guide interventions that shape ongoing learning.
[LG-229] LieDiscover: Adaptive Symbolic Library Construction for Explicit Open-form Symmetry Discovery
链接: https://arxiv.org/abs/2609.33663
作者: Xinxin Li,Jianming Ma,Xingyu Cui,Da Li,Juan Zhang,Junping Yin
类目: ymbolic Computation (cs.SC); Machine Learning (cs.LG)
*备注:
Abstract:Discovering underlying symmetries from data has emerged as a crucial challenge in scientific discovery. Existing data-driven methods for symmetry discovery fail to determine the exact number and mathematical form of unknown infinitesimal generators. Recent explicit methods represent generators using a predefined function library and identify them through algebraic optimization, but they often struggle to capture complex symmetries involving high-order polynomials or transcendental functions. To address this limitation, we formulate symmetry discovery as a joint optimization problem over the function library and coefficients. We propose a novel framework that leverages an encoder-decoder architecture to dynamically generate symbolic expressions and expand the library. This generation process is optimized via reinforcement learning, which accelerates the exploration of the symbolic search space through step-wise rewards. Experiments demonstrate that LieDiscover can successfully uncover open-form infinitesimal generators involving high-order polynomials or transcendental functions, which remain intractable for existing methods. The discovered symmetries also improve performance in downstream PDE solving and discovery tasks.
[LG-230] owards Eliminating Catastrophic Forgetting in the Curriculum Learning of Math Reasoning Tasks
链接: https://arxiv.org/abs/2609.33655
作者: Zengyan Yang,Yangyang Wu,Kai Huang,Pengfei Lyu,Tianyi Zhang,Mengying Zhu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Curriculum learning has found broad application across numerous domains. Nevertheless, its effectiveness is intrinsically curtailed by catastrophic forgetting, driven by the shifts in model parameter distributions between curriculum tasks. In this paper, we investigate the phenomenon of catastrophic forgetting in this training paradigm, building on the established efficacy of curriculum learning. Our theoretical analyses of parameter update dynamics demonstrate that catastrophic forgetting in curriculum learning stems from the divergence of task optima, which is generally essential to the faster convergence of curriculum learning; therefore, forgetting cannot be completely eliminated. Based on this finding, we augment the training process and propose IV-EWC, which incorporates Elastic Weight Consolidation (EWC) into the curriculum learning objective to curb catastrophic forgetting in mathematical reasoning, a prototypical curriculum learning scenario. IV-EWC employs the influence function to construct a representative validation set from the curriculum’s training data, which is used to drive dynamic regularization during training. We further present an extended theoretical analysis to show that EWC-based regularization methods mitigate catastrophic forgetting in curriculum learning, thereby providing theoretical support for IV-EWC. Empirical evaluations on three backbone models and three benchmarks indicate that curriculum learning exhibits catastrophic forgetting. IV-EWC alleviates this issue, reducing forgetting by 162% on average relative to vanilla curriculum learning and yielding positive backward transfer, as evidenced by improved performance on easier tasks after subsequent training on challenging tasks.
[LG-231] FuseAlign: Forced Alignment in the Wild
链接: https://arxiv.org/abs/2609.33650
作者: Mithilesh Vaidya,Stephen Bailey,Sumukh Badam,Matthew Bendel,Xingzhe He
类目: Machine Learning (cs.LG)
*备注: Preprint. Under review
Abstract:Word-level forced alignment estimates when each transcript word occurs in an audio recording. It underpins text-based media editing, subtitling, speech-data curation, and phonetic analysis. Existing evaluations understate the difficulty of forced alignment by relying on short, clean speech, perfect transcripts, and metrics that obscure consequential alignment errors. In contrast, real-world media and data-processing pipelines operate on long and diverse recordings. Additionally, forced aligners often operate on error-prone automatic speech recognition (ASR) output. We address these gaps with improved evaluation metrics, a scoring protocol for real ASR transcripts, and AlignBench, a benchmark spanning diverse speaker, acoustic, and text conditions. We further introduce FuseAlign, a transformer-based aligner trained on large-scale pseudo-labeled speech with online label correction. FuseAlign performs joint contextualization of audio and text for the localization of coarse words. The model then refines boundaries at millisecond resolution and detects missing transcript words in the audio without lexicon-based or Viterbi decoding. On AlignBench, FuseAlign substantially outperforms all baselines and remains robust under real ASR transcripts. Ablations show that convolutional upsampling and EMA-snapshot label correction matter more than model properties such as parameter count.
[LG-232] SafeMol: Dual-Modality Safety Alignment for Molecular Multimodal Models
链接: https://arxiv.org/abs/2609.33640
作者: Xinmiao Wang,Ruijie Wang,Menghui Wang,Jiawei Chen,Haoyue Deng,Ran Zhang,Xingxuan Zhang,Xiao Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Molecular multimodal models support diverse understanding and generation tasks but may introduce safety vulnerabilities when handling hazardous molecules. In this work, We reveal substantial jailbreak vulnerabilities under both text-only and graph-conditioned settings. Our analysis further shows that safety robustness must hold across input modalities while balancing safety, over-refusal, and utility. To address these challenges, we construct SafeMolBench, a molecular multimodal safety-alignment benchmark with 3702 samples covering 618 unique hazardous molecules and safe molecular tasks, organized into hazardous-harmful, hazardous-allowed, and utility-replay subsets to support unified training and evaluation of safety, over-refusal, and utility. Based on SafeMolBench, we propose SafeMol, a parameter-efficient safety alignment framework that jointly optimizes lightweight modules across text-only and graph-conditioned inputs, uses MMD for distribution-level representation alignment to reduce modality-induced discrepancies, and explicitly models molecular hazardousness and harmful operational intent. Experiments on SafeMolBench show that SafeMol reduces attack success by several tens of percentage points while largely maintaining low over-refusal and preserving molecular-task utility.
[LG-233] Climbing the Hill: Prompt Injection Red-Teaming Against Frontier Models with Curriculum Reinforcement Learning
链接: https://arxiv.org/abs/2609.33628
作者: Chenlong Yin,Xiaolong Jin,Wei Zou,Yanting Wang,Jinyuan Jia
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注: 19 pages, 1 figure
Abstract:Prompt injection is a leading security risk for LLMs and LLM-based applications such as agents. State-of-the-art red-teaming methods for prompt injection leverage reinforcement learning (RL) to train an attacker LLM to generate effective injected prompts. However, when targeting frontier LLMs such as GPT-6-Luna, a major challenge is the cold-start problem: every attack attempt by the attacker LLM fails and thus receives zero reward, providing no signal for learning. In this work, we propose a curriculum learning-based method to address the cold-start problem. In particular, we propose to train the attacker LLM against a sequence of increasingly robust target LLMs, with each stage warm-starting from the attacker LLM obtained in the previous one. However, simply training against a weak target (e.g., GPT-4o-mini) may not sufficiently prepare the attacker LLM to obtain useful learning signals against a frontier LLM (e.g., GPT-5.6-Terra). Instead, we find that the design of the curriculum is critical: after each stage, the attacker LLM needs to partially succeed against the next target LLM such that it can learn from successful attempts to attack the new target. Our extensive evaluation shows that our method can effectively red-team frontier LLMs, achieving an attack success rate (ASR@10) of 93.8% and 45.0% against GPT-5.6-Luna and GPT-5.6-Terra on AgentDyn, whereas state-of-the-art RL methods such as RL-Hammer and PISmith achieve 0% ASR under the same setting. Moreover, we find that the attacker LLM transfers across targets, e.g., an attacker LLM trained to defeat one strong LLM (GPT-5.6-Terra) also succeeds against six other frontier LLMs (e.g., GPT-6-Luna) it was never trained on. Our code is available at \hrefthis https URLhere.
[LG-234] You Only Edit Once: Incentivizing In-Context Capability of LLM s via Local Demonstration Refinement
链接: https://arxiv.org/abs/2609.33609
作者: Jiarong Wen,Qi Wang,Yun Qu,Yixiu Mao,Heming Zou,Haoang Chi,Lizhou Cai,Yiqin Lv,Kaiyu Zhang,Yuhang Jiang,Xiangyang Ji
类目: Machine Learning (cs.LG)
*备注:
Abstract:In-context learning (ICL) is crucial for boosting the inference performance of large language models (LLMs). However, the effectiveness of ICL in LLMs is greatly influenced by the choice of demonstration sets. Exhaustive searches over these sets are combinatorial, and existing selectors often rely on relevance or likelihood proxies to implicitly assess ICL quality. Making repeated queries to the target LLM with these strategies can incur substantial costs. This work simplifies selection by framing it as a constrained local search problem and presents local demonstration editing (LDE). Starting with an initially retrieved set of demonstrations, LDE employs a single structured edit to explore its surrounding neighborhood while balancing performance gains with search costs. Technically, LDE is reduced to a policy search problem, for which we train a small LLM, referred to as Jev-LDE. This model as the System-1 modifies the retrieved demonstration set by performing actions such as \textttKeep, \textttDelete, or \textttReplace elements, all within a framework of reinforcement learning with verifiable rewards. At test time, Jev-LDE executes a single edit of the retrieved demonstration set, followed by one inference from the target LLM, avoiding the need for iterative context scoring or subset searches. Across standard classification benchmarks, various target LLMs with Jev-LDE as the plug-and-play module consistently improve ICL performance, and Jev-LDE shows transferability to held-out benchmarks and models without retraining. These findings indicate that the LDE approach offers an efficient and adaptable method for harnessing the ICL capabilities of target LLMs.
[LG-235] Hierarchical Response Preservation for Continual Adaptation of Zero-Shot Graph-Text Models
链接: https://arxiv.org/abs/2609.33607
作者: Haopeng Zhang,Yuhan Wang,Yubing Su,Yingxin Chen,Xiao Wang,Ruijie Wang,Jianxin Li
类目: Machine Learning (cs.LG)
*备注: Preprint
Abstract:Pretrained graph-text models align graph representations with textual semantics, enabling recognition of unseen classes and transfer across graph domains. However, as graph data and classes continually arrive, models should learn from new supervision while retaining their zero-shot transfer capabilities and historical task knowledge. Two challenges arise: (i) new classes can overturn historical predictions despite preserved distinctions among historical classes, and (ii) overly strict response preservation can stall learning of new classes. To address these challenges, we propose Hierarchical Response Preservation (HiRP). HiRP represents this competition through a hierarchical response that keeps each historical-class probability and sums new-class probabilities, preserving historical distinctions and aggregate competition while allowing distinctions within the new class group to adapt. It further uses the geometry induced by this response to guide constrained updates, retaining useful adaptation directions while controlling response drift. Across three class-incremental settings, HiRP achieves absolute gains of 1.84-7.95 percentage points in average accuracy over the strongest compared baseline in each setting, while mitigating zero-shot transfer degradation.
[LG-236] Short-Length Code Designs for Integrated Sensing and Communications: A Deep Learning Approach
链接: https://arxiv.org/abs/2609.33605
作者: Muah Kim,Shuangyang Li,Tayyebeh Jahani-Nezhad,Rafael F. Schaefer,Giuseppe Caire
类目: Machine Learning (cs.LG); Information Theory (cs.IT)
*备注: 13 pages, 5 figures, preprint of a journal paper
Abstract:Integrated sensing and communication (ISAC) enables joint communication and sensing using a shared waveform, but its signal design is challenging due to the inherent trade-off between the two objectives, particularly in the short blocklength regime. This paper proposes an autoencoder (AE)-based framework for ISAC waveform design in noncoherent settings. We derive a modified Cramér-Rao bound for multi-target delay estimation and analyze the maximum-likelihood decoding rule for noncoherent communication under correlated fading. These results reveal structural connections and trade-offs between communication and sensing objectives in waveform design. Based on this analysis, the AE learns waveform representations that jointly optimize both functionalities, with a tunable parameter controlling the trade-off. Simulation results show that the proposed design outperforms conventional schemes in both communication reliability and sensing accuracy, especially under short blocklength and fading conditions. Comments: 13 pages, 5 figures, preprint of a journal paper Subjects: Machine Learning (cs.LG); Information Theory (cs.IT) Cite as: arXiv:2609.33605 [cs.LG] (or arXiv:2609.33605v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.33605 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-237] he Price of Peeking: Anytime-Valid Leakage Detection on ML-KEM EM Traces
链接: https://arxiv.org/abs/2609.33597
作者: Georgios Feretzakis,Alexandros Papaspyridis
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 29 pages, 7 figures, 13 tables. Code and aggregate results: this https URL
Abstract:Side-channel evaluators routinely inspect leakage tests while acquisition is still running, and extend or stop the campaign based on what they see. Fixed-horizon screening such as the Welch t -test with threshold |t|4.5 gives no error guarantee for this monitored decision rule. We study anytime-valid leakage detection based on testing by betting: SKIT-type swap-pair e-processes whose false-alarm probability is controlled uniformly over time under an explicit conditional symmetry null. In matched comparisons that share the frozen witness, rows and payoff, first-crossing detection needed 1.68-2.00 \times the traces of a fixed-horizon randomization test at 80% detection on synthetic streams, and 1.68-2.38 \times on degraded recordings from an open ML-KEM electromagnetic dataset with the primary Ridge witness at \alpha=0.05 . With the same primary witness and level, on undegraded reference and pqm4 recordings the monitored procedure stopped early: its median stopping point was 62-72 and 146-316 evaluation traces, i.e. 2-8% of a conservative 4096-trace budget. Under exact designed nulls on the recorded backgrounds, repeated-look |t|4.5 screening over all 13000-20000 samples raised a false alarm in 2.7-12.9% of replicates, against 0.0-4.7% for terminal-only screening and no rejection by a sample-wise e-Bonferroni process, which in a prespecified follow-up detected natural-label associations in 4 of 4 backgrounds after 840-3288 traces. All recordings come from one device, and natural-label results are descriptive; we state the assumptions each claim requires.
[LG-238] Beyond One-Step Accuracy: State-Affine Latent Transition for Reliable Visual Planning
链接: https://arxiv.org/abs/2609.33595
作者: Boyuan Zhang,Yingjun Du,Xiantong Zhen,Ling Shao
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注:
Abstract:Joint-embedding world models enable visual planning by learning action-conditioned dynamics in latent space. Yet they are commonly trained for one-step prediction on encoded states, while planning recursively applies the learned transition to its own predictions. One-step accuracy therefore does not capture how prediction errors propagate under recursive rollout. We decompose multi-step rollout error into the errors introduced at individual steps and their propagation through subsequent transitions. We show that state-affine dynamics are precisely the differentiable transitions with state-independent Jacobians, eliminating the nonlinear propagation residual and making the error propagation operators depend only on the action sequence. Guided by this result, we introduce SALT (State-Affine Latent Transition), an action-conditioned state-affine dynamics model in which the action modulates both the state transformation and the additive update. We train SALT through recursive multi-step rollout supervision, feeding each predicted latent state back into the transition so that training matches how the model is used during planning. Across four visual planning environments, SALT exhibits 1.48 – 2.19\times higher one-step prediction error than the matched LeWM baseline, yet improves closed-loop success in every environment by 10.0 percentage points on average. On OGBench-Cube, the fraction of episodes that fail with a sharp rise in model-predicted cost after execution decreases from 23.3% to 2.0% .
[LG-239] Pretraining Transformers with Quantized Softmax in Attention
链接: https://arxiv.org/abs/2609.33591
作者: Shangzhen Zhu,Muyan Hu,Tomasz Kozlowski
类目: Machine Learning (cs.LG)
*备注:
Abstract:Low-precision Transformer systems increasingly quantize attention matrix multiplications, while softmax often remains at higher precision. During pretraining, an approximate softmax changes the gradients that train the model as well as its forward computation. We study this interaction with K-interval attention, which approximates the exponential using K+1 grid values. We vary per-row grid calibration, interpolation versus hard rounding, and the placement of a straight-through surrogate relative to normalization. We derive the corresponding backward rules, including calibration derivatives, and compare these choices in pretraining experiments matched on model, data, and optimizer. Detaching the row extrema leaves the forward computation unchanged but produces a delayed increase in validation loss. With hard rounding at K=4, min-max calibration and a pre-normalization surrogate incur a large loss gap; changing either choice substantially reduces it. At 124M parameters and 2.5B training tokens, fixed-window calibration with a post-normalization surrogate yields a validation loss gap of +0.019 nats relative to softmax at K=4, and with a pre-normalization surrogate yields +0.004 nats at K=16.
[LG-240] Approximating Softmax in Pretrained LLM s: Model Sensitivity and Kernel Acceleration
链接: https://arxiv.org/abs/2609.33586
作者: Shangzhen Zhu,Muyan Hu,Tomasz Kozlowski
类目: Machine Learning (cs.LG)
*备注:
Abstract:On NVIDIA Blackwell B200, tensor-core throughput outpaces special-function exponential throughput by more than two orders of magnitude, exposing exponential evaluation in fused attention kernels. A pretrained Transformer, however, may not need it evaluated accurately at every element. We characterize what a pretrained model does need by approximating softmax at inference in ten frozen decoder-only models (0.5B-72B). The number of positions the softmax map assigns probability to and within-row resolution can be cut substantially, yet uniform weighting of the same positions is damaging. Where a fixed resolution budget is placed matters as much as its size, with resolution near the row maximum consistently favored. Perturbations matched on scalar distortion produce model-dependent responses of opposite sign. These findings motivate Rowmax-PoT, a coarse logarithmic weight representation anchored at each row maximum, and Rowmax-H15, its hardware specialization in FlashAttention-4. On B200, the patched FP8 attention forward is 12.4% faster at causal 8K and 25.8% faster at non-causal 8K in host-side call-latency measurements; board energy per forward falls by 8.4% at causal 16K. Measured separately on the BF16 kernel path at 2K, Rowmax-H15 increases perplexity by 0.091-0.492% across five models from three families.
[LG-241] HiLoRe: What to Store Compress or Recompute for Efficient GRPO Training
链接: https://arxiv.org/abs/2609.33570
作者: Xinrui Chen,Mengyang Li,Ou Wu,Ji Zhang
类目: Machine Learning (cs.LG)
*备注: Under Review
Abstract:Group-relative policy optimization (GRPO) makes learner-side activations a major memory-computation bottleneck: gradient checkpointing reduces activation memory through recomputation, but fixed schedules can leave roughly 18 GB unused on a 48-GB GPU despite substantial recomputation overhead. Existing activation-management methods set state fidelity from execution cost, tensor properties, or generic compression sensitivity, without explicitly incorporating GRPO’s analytic update structure into state-fidelity allocation. We formalize this dependence as policy-update exposure, linking the current GRPO loss coefficients to state-level approximation sensitivity. These coefficients are available before backward without an additional backward pass. We introduce HiLoRe, which allocates graph-attributed recovery units among high-precision storage, low-precision compression, and deterministic recomputation using measured recovery utility and update-conditioned approximation risk. It combines high-precision storage and deterministic recomputation with low-precision recovery under a calibrated risk budget. Across five model-task settings with 2K responses and memory 1.10 times GC’s per-GPU actor-update peak, HiLoRe’s actor-update throughput gains reach 13.5% over GC and 7.9% over the fastest evaluated baseline, with paired mean downstream-score differences below 0.6 percentage points.
[LG-242] Correct then Forecast: Observer State-Space Models for Time Series Forecasting
链接: https://arxiv.org/abs/2609.33566
作者: Alexis-Raja Brachet,Guillaume Clavier–Frémond,Abdelhakim Ziani,Pierre-Yves Richard,Céline Hudelot
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Time series forecasting requires extrapolating the dynamics of an observed process beyond the last available measurement. Yet recurrent forecasting models typically treat observations as inputs that directly control their latent dynamics. It leads to a regime change when these observations become unavailable at prediction time. Following a state-estimation perspective, we introduce Observer State-Space Models (OSSMs), a class of recurrent models that interprets the observed input time series as measurements of an underlying autonomous dynamical system. OSSMs explicitly separate latent-state propagation from measurement assimilation: a single transition governs the dynamics across both context and forecasting intervals, while available observations correct the estimated state through an observer. This formulation naturally exposes classical control-theoretic properties, including observability and convergence of the state estimation error. We further show that conventional and recent SSMs can be recovered as particular instances of our OSSM framework, thereby providing a unified interpretation of their recurrent dynamics and revealing modeling inconsistencies. We perform experiments across several benchmarks showing that OSSM achieves substantial improvements while maintaining the same parameter count and training setup as the corresponding SSM baseline. These results support a simple principle for recurrent forecasting: observations should correct the estimated latent state, rather than control the dynamics used to propagate it.
[LG-243] GraphSelect for Budgeted Representation Selection in Multimodal Graph Inference
链接: https://arxiv.org/abs/2609.33561
作者: Xu Wang,Xunkai Li,Yinlin Zhu,Rong-Hua Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:Multimodal graph predictors combine text, images, and relations to classify connected entities. How much of this input is needed to preserve their predictions? We study budgeted representation selection, which chooses a subset of candidate text and image vectors under a separate capacity for each modality. Predictions from the complete candidate input define the classes to preserve. The challenge is that a representation’s contribution depends on the other selected inputs, while graph propagation extends its effects across nodes. Our empirical study shows that candidate rankings change with the selected input, while predicted probabilities remain informative after the class stops changing. Updating scores improves selection, and exchanging inputs can improve a subset whose capacity is already filled. These findings lead to GraphSelect, which starts from individual candidate gains and refines the subset through jointly evaluated exchanges. It screens promising removals and additions, accepts an exchange when it reduces the prediction loss, and updates the scores. Experiments on six graphs show higher mean objective recovery than six attribution and explanation methods adapted to the selection task. Across nine trained architectures on two graphs, retaining 20% of the candidate representations per modality gives a mean accuracy drop of 0.10 percentage points relative to full candidate input, preserving classification performance with substantially fewer text and image representations.
[LG-244] Fine Until Fine-Tuned: Repeated Solutions Make Reasoning Frag ile
链接: https://arxiv.org/abs/2609.33559
作者: Ely Sheikh
类目: Machine Learning (cs.LG)
*备注: 34 pages, 7 figures. Code, data and outputs: this https URL
Abstract:Recipes such as s1 and LIMO teach a model to reason with little data by showing it the same thousand or fewer worked solutions many times over. Judged when that training ends, the repetition looks harmless. But reasoning models are often trained again, and we find that repetition leaves their reasoning fragile to that next stage, even when the stage has nothing to do with reasoning. We fine-tuned Qwen3.5-9B-Base on its own correct solutions to competition math problems, either drilling a few hundred of them about eight times each or showing many more once; with the same amount of training, both solve about 95% of held-out problems. A single pass of ordinary instruction tuning leaves the once-trained model where it was, while the drilled one falls to 86.0%, and harsher later stages take it to 59.3% or below. A third model that visited the drilled problems just as often, with a new solution at every visit, was unharmed, so the damage comes from seeing the same texts again rather than from having few problems. The break recurs with a stronger model’s traces, in further training runs and on other models and tasks. It is also cheap to undo: the reasoning is suppressed rather than erased, and five updates of reasoning training bring almost all of it back, as does brief training on the reasoning format with almost no mathematics. Fresh solutions prevented the damage, and so did replaying 6.25% of the original solutions in a gentler later stage, so our claim concerns later training without such replay. Sharpening alone does not explain the break, since a model sharpened three-quarters as much without repetition was unharmed. On a skill the base model could not perform within a token budget, repetition mainly cost learning.
[LG-245] Does Execution Require Target KV Fidelity? A Mixed-Fidelity KV Runtime for LLM Serving
链接: https://arxiv.org/abs/2609.33536
作者: Jiantong Jiang,Yue Yang,Peiyu Yang,Feng Liu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large language model (LLM) serving is increasingly constrained by the GPU memory consumed by key-value (KV) caches. Existing compression, eviction, and offloading techniques alleviate this pressure, but serving runtimes typically treat only the configured target KV representation as execution-ready. Under memory pressure, this target-only contract can turn KV shortage into request stalls and preemptions. We present ElasticKV, a mixed-fidelity KV runtime built on the observation that target fidelity need not gate execution. ElasticKV introduces a compact intermediate KV state, making fidelity a runtime-managed execution property. To realize this state in a paged serving runtime, ElasticKV combines (i) a pair-structured layout that turns fidelity reduction into reusable GPU capacity, (ii) a dual-mode attention backend that directly consumes the compact state while preserving the native target-only path, and (iii) pressure-aware fidelity management that adapts KV fidelity to memory pressure. Our extensive evaluation across diverse workloads, model families and scales, and GPU platforms demonstrates the effectiveness and generality of ElasticKV. Under high concurrency, ElasticKV achieves 3.8-4.0 \times lower time-to-first-token (TTFT) and 9.1 \times lower P90 TTFT than vLLM while preserving generation quality.
[LG-246] Discovering Symmetries in Neural Network Parameter Spaces
链接: https://arxiv.org/abs/2609.33527
作者: Bo Zhao,Nima Dehmamy,Robin Walters,Rose Yu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Parameter space symmetries are important for understanding neural networks’ loss landscape, training dynamics, and generalization. However, systematically identifying these symmetries remains a challenge. In this paper, we formalize data-dependent parameter symmetries and characterize loss invariance and the group-action axioms through infinitesimal conditions, which provide objectives for jointly learning group generators and nonlinear action maps. Our framework systematically uncovers parameter symmetries, including previously unknown ones. To study larger networks, we establish conditions under which subnetwork symmetries extend to the full model. The same construction gives an explicit family of finite-batch symmetries, providing both analytical examples and a foundation for discovery through small subnetworks. Using the infinitesimal characterization and subnetwork construction, we implement a framework for automated discovery of parameter symmetries, and successfully uncovered symmetries in various architectures, including pretrained transformer models.
[LG-247] LLM 4Trust: Exploring the Capabilities of Large Language Models for Trust Evaluation NDSS2027
链接: https://arxiv.org/abs/2609.33521
作者: Jie Wang,Yanbo Sun,Zheng Yan,Jiahe Lan,Elisa Bertino
类目: Machine Learning (cs.LG)
*备注: Accepted by NDSS 2027
Abstract:Trust evaluation plays a critical role in cybersecurity by supporting risk mitigation and decision-making. A variety of trust evaluation methods have been proposed, with learning-based approaches offering high accuracy and automation. However, they often require substantial ground truth, suffer from low training efficiency, lack support for basic trust properties, and provide limited explainability. Large Language Models (LLMs) offer a compelling alternative due to their strong zero-/few-shot reasoning abilities and broad knowledge. To this end, we propose LLM4Trust, the first benchmark framework that systematically explores the capabilities of LLMs for trust evaluation. We first construct diverse trust graphs to model five basic trust properties and design corresponding property understanding tasks. We then assess the ability of eight representative LLMs to understand these properties under nine prompt methods. Based on this exploration, we identify the most effective LLM-prompt combinations and apply them to five real-world datasets for validating LLMs’ trust evaluation capability. During this process, we propose two strategies to extract key information from large-scale trust graphs, addressing the context window limitations of LLMs. Extensive experiments show that LLMs can effectively understand basic trust properties and have great potential for real-world trust evaluation, particularly under limited supervision. However, they remain vulnerable to attacks targeting trust graphs and demonstration examples used in few-shot prompting, and incur high inference costs. Accordingly, we propose a defense mechanism and batch inference to improve the robustness and efficiency of LLM-based trust evaluation. The source code of LLM4Trust is available at this https URL
[LG-248] SLP-ProbHard: Probabilistic Hard-Constrained Learning via Structural Latent Parameterization
链接: https://arxiv.org/abs/2609.33515
作者: Wondesen Teshome Bekele,Marco D’Oria
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 44 pages, including appendices. Preprint
Abstract:Many probabilistic predictors must satisfy exact structure in every stochastic realization, yet common hard-constraint approaches form predictions in ambient coordinates and then correct or project them. We introduce SLP-ProbHard, a cross-family, representation-centered framework for probabilistic hard-constrained learning when explicit structural parameterizations are available. Its core object, a Structural Feasible Latent Parameterization (SFLP), combines a structural latent law Z \sim P^Z_\theta(\cdot\mid x) with a feasible map Y=h_\phi(x,Z) that satisfies the constraint for every latent realization. Together these components define the predictive law itself, including its support and boundary probabilities, rather than serving as a final feasibility wrapper. We study how feasible coordinates and maps affect stochastic dimension, dependence, calibration, expressiveness, and computation. Experiments use Gaussian latent laws and fixed geometry-derived maps across affine equalities, ordering and simplex constraints, nonlinear manifolds, and three structural representations of seven-basin hydrological flow-duration-curve (FDC) data. In an official-source affine comparison with ProbHardE2E/DPPL, both methods achieve zero practical constraint violations. SLP-ProbHard uses 8 instead of 11 stochastic coordinates and improves MSE/MAE, while DPPL yields better marginal CRPS and closer-to-nominal coverage; a paired test detects no Energy Score difference across ten seeds. Real-world affine and nonlinear FDC representations reduce 13 to 7 and 14 to 8 ambient versus computational coordinates, respectively. Exact feasibility alone thus does not determine a predictive law, motivating direct structural generation when meaningful feasible coordinates are available.
[LG-249] Pulseflow: PPG Counterfactual Generation Via Latent Transport
链接: https://arxiv.org/abs/2609.33501
作者: Hung Manh Pham,Dong Ma,Bin Zhu,Pan Zhou
类目: Machine Learning (cs.LG)
*备注:
Abstract:Photoplethysmography (PPG) has become an important modality for continuous cardiovascular monitoring, including atrial fibrillation (AF) detection. However, labeled AF recordings remain limited in many clinical settings, making model adaptation difficult when only limited target data are available. Generative modeling offers a natural way to alleviate this scarcity by synthesizing additional AF signals. Existing approaches, however, mainly generate samples that match the target condition without explicitly modeling how an observed source recording should be transformed, making it difficult to leverage abundant source recordings from a specific population or cohort for targeted augmentation. We introduce PulseFlow, a source-conditioned counterfactual generation framework that combines conditional representation learning with invertible latent transport to edit cardiac rhythm while retaining information from the source. Experiments across two clinical cohorts demonstrate effective rhythm transformation, measurable source correspondence, and improved AF classification under limited labels.
[LG-250] he cost of useful natural gradient updates
链接: https://arxiv.org/abs/2609.33499
作者: Subhransu S. Bhattacharjee,Dylan Campbell,Rahul Shome
类目: Machine Learning (cs.LG); Information Theory (cs.IT); Optimization and Control (math.OC)
*备注: 43 pages, 8 figures, 10 tables
Abstract:What information is needed to turn a natural-gradient direction into a useful finite update? Under a population Kullback-Leibler (KL) budget, we call a step useful if it is feasible and loses at most a fraction \varepsilon of the best feasible gain along the direction. We construct a four-state exponential family whose laws share their initial gradient, scalar Fisher information and natural gradient, yet two laws have disjoint useful-step sets. With these quantities supplied exactly and the law otherwise known only through draws, the family’s worst-case sample complexity is \Theta(\log(1/\delta)/(p\varepsilon^2)) for small \varepsilon , where p scales rare-state probabilities and \delta is the failure probability. The budget is fixed and the optimal gain stays bounded away from zero, so the step length, not the direction, carries this cost. For succinctly described event-tilt models, returning a useful step is NP-hard even with the exact natural gradient and efficient exact sampling. Recovering the unit natural gradient to constant error is also NP-hard even in a two-parameter logistic family with Fisher condition number at most 3. We also give matching sample bounds for event tilts, sample bounds for damped Fisher solves and a population-KL certificate for affine classifiers. In frozen-feature classifier heads, stopping at a sampled KL boundary succeeds in about half of the trials, and a 10% KL margin raises joint success above 93% at a KL budget of 0.01. Thus, knowing where to move is not enough: how far to move can carry an update’s entire cost.
[LG-251] Hamiltonian JEPA: Action-Conditioned World Models with an Inherited Control State
链接: https://arxiv.org/abs/2609.33497
作者: Tamim Zoabi,Ameen Ali,Lior Wolf
类目: Machine Learning (cs.LG)
*备注:
Abstract:Planning from pixels needs more than a latent space that is stable and predictable. The state the planner scores must also be organized by how actions move the system. Joint-embedding predictive architectures (JEPAs) avoid pixel reconstruction by predicting future representations, but existing action-conditioned JEPAs ask one embedding to serve both perception and control. We introduce H-JEPA, which separates the two. A wide perceptual code is regularized toward a well-scaled isotropic geometry with a Bures-Wasserstein prior, and a fixed orthonormal slice of that code is the control state, which inherits the code’s covariance without any objective of its own. The state evolves under phase-conditioned dissipative port-Hamiltonian dynamics whose input port has orthonormal columns. Port-inverse consistency (PIC) reads the executed action back through the transpose of that port. We show that this readout is exactly the rollout error projected onto the port directions, so PIC is a parameter-free reweighting of prediction error and not an auxiliary action decoder. Untying the readout from the port breaks this identity and loses half of the gain. H-JEPA matches or exceeds reconstruction-free baselines, including the action-decoding Delta-JEPA, on four pixel-based control benchmarks after at most 10 training epochs, and its largest gain is on OGB-Cube ( 91.9 against 79.3 percent). Ablations on PushT and OGB-Cube separate the contributions of the structured predictor, PIC, the prediction horizon, the state rank, and the anti-collapse prior.
[LG-252] Predicting Block-Coordinate Performance via Cross-Curvature
链接: https://arxiv.org/abs/2609.33489
作者: Shengkun Zhu,Jinshan Zeng,Zhiqiang Kou,Yongxin Tong,Yang Liu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Simultaneous and sequential block updates are two basic optimization strategies used across machine learning, such as neural-network training, federated learning, and low-rank adaptation. Choosing between them is difficult because their relative advantage depends on both the objective geometry and the number of iterations. We develop a unified theory for comparing Jacobi (JC), Gauss–Seidel (GS), and partially sequential deterministic block-gradient updates. Our analysis expresses the one-step loss difference through cross-block curvature, with an O(\eta^3) remainder, where \eta is the learning rate. We derive a signed loss comparison after K iterations with O(K\eta^3) error under regularity conditions and \eta K\le T for fixed T , identifying the better method when the predicted difference exceeds this error. We evaluate these formulas along observed training trajectories across different machine learning settings. Over 500 iterations, our theory correctly identifies the lower-loss method in 98.0% of iterations for the neural network, 83.4% for federated learning, and 97.6% for LoRA. Applying the loss recursion at each step using the measured parameter difference raises these rates to 100.0%, 93.2%, and 99.6%, respectively.
[LG-253] What masking geometry works best for EEG foundation models?
链接: https://arxiv.org/abs/2609.33487
作者: Pierre Guetschel,Bruno Aristimunha,Yassine El Ouahidi,Arnaud Delorme,Thomas Moreau,Michael Tangermann
类目: Machine Learning (cs.LG)
*备注: A controlled evaluation across MAE and JEPA. 44 pages, 12 figures, 15 tables. Project page: this https URL
Abstract:EEG foundation models hold promise for scalable brain-signal decoding across clinical and cognitive neuroscience applications, yet their pre-training pipelines remain poorly understood. Among design choices, the masking strategy is particularly critical: it determines what the network must predict and from which context. Yet it has never been ablated in isolation, as each new model bundles a new masking strategy with a new backbone and objective. In this paper, we formalize the design choices for spatio-temporal masking strategies and train various models with a single pipeline under varying masking configurations across two SSL frameworks (MAE and JEPA). We then systematically evaluate the resulting 58 pre-trained models on the 12 datasets of OpenEEGBench under a linear probe. Both frameworks agree on an optimal masking configuration and on shared failure modes. Outside these, performance is robust: 11 MAE and 9 JEPA configurations are statistically indistinguishable from the best. We further identify a novel JEPA-specific failure mode, tagged bias-inflation collapse, invisible to standard detectors. With a well-chosen mask, our pipeline reaches REVE-level downstream performance at a fraction of REVE’s pre-training compute.
[LG-254] Mask-Induced Displacement in Audio XAI via Logit Trajectory Decomposition
链接: https://arxiv.org/abs/2609.33486
作者: Nico García-Peguinho(1),David Kelly(2),Fabrizio Smeraldi(1),Anna Xambó Sedó(1) ((1) School of Electronic Engineering and Computer Science, Queen Mary University of London (2) Department of Informatics, King’s College London)
类目: ound (cs.SD); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
*备注: 5 pages, 2 figures, 3 tables. Paper status: submitted
Abstract:Perturbation-based XAI methods for audio classifiers often estimate feature importance by masking spectrogram regions and crediting output changes to the retained signal. Yet they typically assume the fill (the mask replacement) is negligible. We propose logit-space trajectory decomposition to examine this assumption. An on-axis component captures output along a line connecting the fully filled (occluded) spectrogram to the fully retained original; an off-axis component captures perpendicular displacement. We evaluate across three fills, three audio classifiers, and 1,100 AudioSet clips. We demonstrate that no fill is acoustically neutral under full occlusion: Zero fill activates silence and Gaussian Noise fill activates broadband noise. Under partial masking, 41-77% of output displacement is off-axis, with the direction of the residual stable across mask retention fraction and specific to each model-fill combination. Attribution methods are unevenly exposed to off-axis displacement through their sampling and weighting strategies, revealing apparatus-dependence. Where the off-axis residual is stable and low-dimensional, its structure affords mitigation.
[LG-255] How Synthetic Labels Improve Conformal Prediction: A Perspective on Conditional Coverag e
链接: https://arxiv.org/abs/2609.33482
作者: Qianyi Chen,Bo Li
类目: Machine Learning (cs.LG); Methodology (stat.ME)
*备注:
Abstract:Conformal prediction provides distribution-free finite-sample marginal coverage, but post-hoc calibration data may be too scarce to learn how uncertainty varies across inputs. Meanwhile, abundant covariates can often be labeled cheaply by domain models or general-purpose language models. We study whether these synthetic labels can improve conditional coverage when only a small trusted sample is available. Building on score-quantile regression, we introduce prediction-powered quantile learning: a synthetic-labeled pool estimates pinball risk, paired trusted and synthetic outcomes correct its bias, and an independent trusted split performs final conformalization. Profiling pinball risk over scalar corrections reveals that population conditional-coverage error is its functional gradient; the corresponding Hessian removes global shifts and weights remaining shape error by boundary density. Composing this geometry with prediction-powered learning yields a three-resource expansion and a benefit–cost rule for synthetic power. Across eight regression benchmarks, synthetic-powered quantile learning substantially improves downstream conditional coverage while preserving marginal validity and producing more compact prediction sets. A human-rating study finds similar gains from external LLM labels and exposes a quality–quantity–cost tradeoff.
[LG-256] Geometric Identification in Predict-Then-Optimize Learning
链接: https://arxiv.org/abs/2609.33472
作者: Jiaxiao Xu,Changhong Mou,Keji Liu,Dinghua Xu,Yeyu Zhang
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Decision-focused surrogates can recover downstream decisions without identifying the quotient report. We characterize the equality set of the convex Smart Predict-then-Optimize surrogate (SPO+) population risk. Under central symmetry, the centered mean class is the unique Bayes minimizer exactly when every nonzero effective displacement makes the old optimizer leave the shifted optimal face with positive probability. This condition separates face crossing from selected-oracle disagreement and gives quantitative local coercivity. Without symmetry, strict crossing alone need not identify the mean; selection balance with reflected crossing restores quotient-report identification, and conditional versions extend the result to measurable predictors. These are population statements, without finite-sample report-recovery or generic transfer-regret guarantees. Closed-form mechanisms reproduce the analytic identities and rates. Portfolio, complete-matrix KuaiRec, and Energy/Storage studies measure predictive fidelity, shifted regret, and fitted-report geometry. A known data-generating process (DGP) companion retains their application geometries while isolating conditional-mean recovery and crossing, without testing the original observational assumptions.
[LG-257] A Free Knob: Decoupling Calibration and Predictive Skill in Threshold-Based Evaluation
链接: https://arxiv.org/abs/2609.33457
作者: Md Tanveer Hossain Munim,Bijoy Ahmed Saiem,Al-Amin Sany,Tanzima Hashem
类目: Machine Learning (cs.LG)
*备注: Code: this https URL
Abstract:Many dense-prediction benchmarks evaluate rare events by pooling prediction and target over spatial blocks, thresholding each, and scoring the contingency table. At a fixed rare operating point, the max-pooled Critical Success Index (CSI) confounds spatial discrimination with amplitude calibration: sharp observations promote many blocks above threshold, while attenuated predictions from squared-error regression leave the same blocks below it. We repurpose classical monotone calibration as a symmetric audit: a post-hoc transform fitted on held-out data and applied separately to each system. The transform cannot reverse pixel ordering, so any contrast it reproduces cannot establish improved spatial ranking. On SEVIR, two released checkpoints of one architecture differ by -29.5% in extreme-threshold CSI before the control and by +5.3% after it. Across 450 pairwise contrasts among 6 systems, the difference in pooled frequency-bias deviation is associated with how far the CSI contrast moves under the control (r = +0.796), and 51 contrasts reverse sign. At CasCast’s published extreme-event operating point, the cascade-over-backbone CSI gap falls from 0.1601 to 0.0339, a 78.8% reduction; the remaining gap stays positive. The effect persists when the transform is fitted on a window before the test period, and calibration also reveals advantages hidden by a better-calibrated baseline. On geostationary infrared imagery the relative gain grows as events become rarer, crowd counting reproduces the bias-gain relationship under patch-sum pooling, and semantic segmentation, where frequency bias is already near one, shows little average change. The confound therefore requires both a fixed operating point and a training regime that leaves the output miscalibrated there. We recommend reporting pooled frequency bias and a symmetric held-out FreeKnob Audit alongside rare-event pool-and-threshold scores.
[LG-258] SchemaMem: Schema-Indexed Recurrent Memory for Delayed State Retrieval
链接: https://arxiv.org/abs/2609.33436
作者: Sungwoo Goo,Hwi-yeol Yun,Sangkeun Jung
类目: Machine Learning (cs.LG)
*备注:
Abstract:Attention provides direct access to past representations, but retaining an ever-growing history is costly. Recurrent models bound persistent state, yet must preserve selected information while processing subsequent inputs. We introduce SchemaMem, an attention-based recurrent memory architecture combining chunk-local attention with a persistent, schema-indexed phase state. Learned schema embeddings provide a shared representational reference for reading and writing. Reads use the current state, whereas writes use the layer input and static schema embeddings, excluding direct feedback from that layer’s own state. Chunk-boundary commits aggregate bounded phase increments through forward computation. The same parameters also support full-history attention training before and during recurrent training. We studied selective updates, preservation, and delayed retrieval in a controlled address–value task, comparing three-layer models with approximately matched parameter counts and persistent-state dimensions. Across nine address/value settings and three training seeds, SchemaMem has higher mean written-value retention at four times the maximum training delay than both baselines, which are trained toward a higher in-range accuracy target. Updated-value recovery favors SchemaMem in all nine settings against Mamba-3 and seven against Gated DeltaNet. Defaults consistently favor Gated DeltaNet over SchemaMem at that delay, and SchemaMem requires substantially more optimization steps. These results identify a promising retention–optimization trade-off in schema-indexed recurrence.
[LG-259] MoGround: Measuring and Mitigating Modality Distraction in Vision-Language Models
链接: https://arxiv.org/abs/2609.33431
作者: Luca Zhou,Bo Zhao,Rose Yu,Emanuele Rodolà,Roberto Dessì
类目: Machine Learning (cs.LG)
*备注: 9 pages of the main paper with 3 tables and 5 figures
Abstract:We release MoGround, a vision-language dataset spanning four visual domains in which the answer to every question is guaranteed to be available from exactly one modality. This guarantee enables us to measure modality distraction, the failure in which a model answers a question correctly from one modality alone and then flips to a wrong answer once irrelevant content from the other modality is added. Existing probes rarely establish single-modality answerability this way, making it hard to isolate distraction in the first place. Across seven open-source VLMs, we find that modality distraction is not universal but model-dependent. The weaker-grounded modality is the more distracted one (r = +0.86), and distraction scales inversely with grounding strength (r = -0.90). The single-modality guarantee also enables a mitigation method that needs to distinguish between relevant and irrelevant context. Trained on one split of MoGround alone, a weight-space robustness vector reduces distraction on all seven models by 9% to 51%, at a cost of only 0.1 average points of accuracy on standard multimodal tasks.
[LG-260] FoldAttention: Declared-Reference Softmax for Fast Decode and Deterministic Backward
链接: https://arxiv.org/abs/2609.33410
作者: Sriman Achanta
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注:
Abstract:Autoregressive decode repeatedly streams a growing KV cache, making attention a major cost at long context. Existing high-performance kernels use online softmax, which discovers a row’s normalization reference as it scans keys. Earlier contributions therefore remain provisional and may require rescaling. We argue that the reference need not be discovered: softmax is invariant to a common shift, so the reference only has to keep the weights in range. We present FoldAttention, an additive formulation of softmax attention that fixes a finite reference Z_i before scanning the KV cache. Each weight 2^s_ij-Z_i is then final when computed, so contributions add across disjoint key ranges and their quotient equals softmax attention in real arithmetic. We use this property to develop two techniques for Hopper decode: (1) final weights gate key and value reads before the bytes are fetched, and a per-call depth T cuts keys below 2^-T while keeping their mass, and (2) additive partials compose split KV and shared-prefix cascades without rescaling. On H100 at T=16 , FoldAttention decodes seven real-model generations 1.36-2.30 \times faster than the fastest BF16 baseline, and up to 3.09 \times faster across MHA and GQA shapes, at an error within 1.5% of the lowest BF16 error on six of the seven; reading every key, it is 1.14-1.30 \times faster at matched error. We validate on Qwen3-8B that a whole decode step is up to 1.46 \times faster while likelihood and long-context accuracy match those under BF16 kernels. The same principle makes the backward deterministic: CTAs round bounded partial gradients onto an integer grid declared before the reduction and add them in any order. FoldAttention thereby removes the determinism tax: its deterministic backward is up to 1.84 \times faster than deterministic FlashAttention-3/4 and 1.05 \times faster than the fastest nondeterministic kernel.
[LG-261] StarBOA: Real-Time Mamba State-Space Unrolling for Sparse Radar Micro-Doppler in ISAC Networks
链接: https://arxiv.org/abs/2609.33408
作者: Mustafa Bora Çelik,Ceren Çelik,Orhan Gazi
类目: Machine Learning (cs.LG); Information Theory (cs.IT); Networking and Internet Architecture (cs.NI)
*备注: 5 pages, 2 figures
Abstract:In Integrated Sensing and Communications (ISAC), radar sensing must operate under chirp subsampling with up to 90% missing data. An attention-based baseline, limited to a 52~ms buffer, collapses toward maximum uniform entropy ( H=2.584 bits) as sparsity increases, failing to capture long-range gait-cycle context. We propose StarBOA, which replaces attention with a causal Mamba state-space model that updates incrementally on a per-window basis without re-scanning past reconstructions. By maintaining a persistent state, StarBOA integrates over 100\times more temporal history at no additional per-step computational cost. StarBOA outperforms the baseline’s published results across all sparsity levels, with SSIM gains increasing from +0.0379 at 50% missing data to +0.2472 at 90%. Each window is processed in 1.53~ms with zero lookahead, demonstrating efficient causal reconstruction under extreme chirp subsampling.
[LG-262] From Grey-Box to Green-Box: When can Physics-Informed Machine Learning Reduce Carbon Footprints in Structural Health Monitoring?
链接: https://arxiv.org/abs/2609.33387
作者: Daisy R. Bradley,Nathan A. Hinchliffe,Daniel J. Pitchforth,Matthew R. Jones,Elizabeth J. Cross
类目: Machine Learning (cs.LG)
*备注: 25 pages, 15 figures
Abstract:Machine learning plays an increasingly vital role in engineering, but the corresponding increase in compute time is not without environmental cost. Physics-informed machine learning or “grey-box” models have been developed to overcome some of the limitations of traditional black-box learners, utilising the physical insight that an engineer would have about the structure they are modelling and have shown promising results in the structural engineering field among many others. This work explores whether an additional advantage could be a reduced environmental impact, considering the relationship between training data quantity and training time, linking this duration to carbon emissions from computing. In a structural health monitoring context, four physics-informed machine learning approaches - spanning Gaussian processes and neural networks - are evaluated: residual modelling, input augmentation, hybrid modelling, and constrained learning. The emissions for training each of the models to reach a given error threshold is compared, and in most examples, shown to be lower for the physics-informed models (with input augmented models being an exception). This reduction in training emissions further compounds the environmental savings achieved by collecting and storing less data. Although promising results, we cannot expect a silver bullet and the case studies demonstrate that a trade-off is needed between the increased complexity that comes from introducing physics into a machine learner, against the gain from reduced training data requirements. Comments: 25 pages, 15 figures Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.33387 [cs.LG] (or arXiv:2609.33387v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.33387 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-263] Optimal Transport Dropout for Structured Predictive Uncertainty
链接: https://arxiv.org/abs/2609.33377
作者: Giacomo Lorenzon,Francesco Regazzoni
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注:
Abstract:Deterministic neural networks and neural operators provide point predictions with no intrinsic measure of reliability. Yet, predictive uncertainty may stem from irreducible outcome variability, finite data, or limitations of the chosen model class. Monte Carlo dropout offers a computationally convenient way to construct a predictive distribution through stochastic feature masking, without training multiple independent networks or explicitly inferring a posterior over model parameters. However, its perturbation law is largely prescribed a priori and typically factorised across latent coordinates. We introduce Optimal Transport Dropout (OTD), which instead learns the predictive mapping and the law of its latent perturbations jointly. Starting from a simple independent reference distribution, OTD transports latent perturbations through a learnable flow and propagates them through the predictive neural network, thereby inducing a structured predictive law. Training uses the strictly proper Energy Score, while a kinetic-action term geometrically regularises the transport. Synthetic benchmarks show that OTD captures multimodal predictive distributions, generates meaningful dispersion when the model is misspecified, and exhibits contracting dispersion as more training data or greater model capacity are provided. For a field-valued partial differential equation surrogate, predictive dispersion strongly aligns with the spatial pattern of prediction errors. On this task, compared with Monte Carlo dropout, OTD yields more accurate predictions and better-calibrated, substantially narrower intervals. On real-world regression benchmarks, it further shows competitive accuracy and better probabilistic predictions compared to established baselines. OTD therefore offers a way to learn structured predictive uncertainty without explicit posterior inference or ensembles of independently trained predictors.
[LG-264] he Selection Rule Decides the Winner: A Pre-Registered Audit of Open-Set Graph Anomaly Detection
链接: https://arxiv.org/abs/2609.33370
作者: Farhan Shahriyar Hossain,Taufikur Rahman Fuad,Md Abrar Jahin,Md Rizwan Parvez
类目: Machine Learning (cs.LG)
*备注:
Abstract:Open-set graph anomaly detection trains on a few labeled anomalies from one class and must also find anomaly classes that were never labeled. Published results share three conventions: the test score is read at the best epoch on the test set, baseline numbers are copied from earlier papers, and most anomalies are minority classes relabeled as anomalous. We ask how much of the reported ranking these conventions decide. We re-run two recent methods, DEMO and NSReg, together with OUTPOST, a small first-order detector built for this study. All three use one protocol with identical seeds and splits on eight graphs (seven for the baselines, which cannot run on ogbn-mag), ten seeds each, and every run is scored under both the best-epoch rule and a deployable validation rule. Before the runs that test them, we registered 40 predictions. Three findings hold. First, the rule changes the leader: under the best-epoch rule, OUTPOST and NSReg each lead three of seven graphs, while under the validation rule, NSReg leads five. Second, the best-epoch bonus depends on how the benchmark was built: 0.045–0.080 AUC-ROC on the three small relabeled-class graphs and 0.002–0.014 on the three real fraud graphs. Third, pseudo-labeling in OUTPOST is worth 0.038–0.065 AUC-ROC on the same three graphs but gives no benefit on any real fraud graph. We also show that a 0.002 tie band for hyperparameter selection lies below the paired standard error on all six graphs tested, even at ten seeds. Twelve of our 40 predictions were falsified, and we report them. We close with a short reporting checklist.
[LG-265] Geometry-Adaptive Mechanisms for Private Synthetic Data
链接: https://arxiv.org/abs/2609.33363
作者: Raoof Zare Moayedi,Amir R. Asadi,Mohammad Hossein Yassaee,Gholamali Aminian
类目: Data Structures and Algorithms (cs.DS); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 43 pages, including appendices
Abstract:Generating differentially private synthetic data with meaningful Wasserstein utility guarantees is challenging in high dimensions. For datasets of size (n) on [0,1]^d with d\ge2 , existing pure (\varepsilon)-differentially private mechanisms achieve expected 1 -Wasserstein error of order (\varepsilon n)^-1/d , reflecting the curse of dimensionality. While this rate is optimal in the worst case, it can be overly pessimistic when the data are supported on a lower-dimensional set. We formalize this through a multiscale packing-growth dimension k , which captures the geometric complexity of the support via the growth of packing numbers across scales. We propose \emphAdaptive Pruned-PMM, a pure \varepsilon -differentially private mechanism that combines private depth selection with our pruned variant of the Private Measure Mechanism (PMM) of He et al.\ (2023). The mechanism supports deeper, geometry-adapted hierarchies with expected running time O!\left(d(n+d)\log(\varepsilon n)\right) , which is near-linear in n for fixed dimension and privacy budget. Under an external multiscale packing-growth condition with dimension k , we show that, for fixed positive privacy budgets and fixed geometry, the expected 1 -Wasserstein error is of order (\varepsilon n)^-1/k for k1 as n grows. We also prove a lower bound under a corresponding internal packing-growth condition, showing that the exponent 1/k is sharp within this framework.
[LG-266] How Much Imprecision is Enough Imprecision in my Classifier? A Practical Elicitation Procedure
链接: https://arxiv.org/abs/2609.33352
作者: Victor F. Lopes de Souza,Sébastien Destercke,Abdelhak Imoussaten
类目: Machine Learning (cs.LG)
*备注:
Abstract:Set-valued classifiers, whether derived from precise probabilities and an adapted cost function, from convex sets with a robust inference mechanism, or from conformal methods, are routine options to obtain more robust, trustworthy predictions. However, there is a lack of operational tools to measure how robust or imprecise a given user is ready to be when receiving predictions, that is how much precision he/she is ready to let go in exchange of more accuracy. This is why we propose, in this paper, practical and operational elicitation procedures to measure the user proneness to set-valued predictions. The effectiveness of the iterative elicitation procedure in converging to the target parameter value is demonstrated on both tabular and image datasets drawn from standard machine learning benchmarks. The results show that the procedure also presents the user with a small number of instances, highlighting the practicality of the approach for real-world applications aimed at identifying the decision maker’s optimal behavior when faced with imprecision.
[LG-267] MultiEcho: An Experimental Science of Learned Worlds
链接: https://arxiv.org/abs/2609.33347
作者: Meng Zhu,Airui Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:World models can be studied as experimental systems with response laws of their own. We introduce MultiEcho, a framework for estimating these laws through controlled counterfactual interventions, delimiting their applicability, and separately testing their physical correspondence. Across nine simulated physical systems and seven frozen model configurations, three-reference estimators predict complete intervention responses and recover intervention parameters. Estimator selection uses discovery data only; frozen fits are evaluated on validation and confirmation contexts. The experiments distinguish response predictability, intervention readability and physical accuracy. Responses can be locally describable yet poorly match physical effects in the same target coordinates. Event-window, visibility and camera interventions reveal conditional applicability, and paired generator configurations show reduced readability under a scene prompt with stronger guidance. Magnitude sweeps expose small image errors alongside large relative effect errors. An exact-reset material experiment separates registered visible-response success from fixed-readout failure on material-dependent futures at matched positions and velocities. Exact finite-scale identities resolve odd and even response errors; first-order remainder bounds specify when refined calibration converges. MultiEcho provides an experimental basis for studying learned-world laws independently of, and in relation to, physical laws.
[LG-268] RAEGL: Risk-Aware Evidence-Gated Learning for Selective Contextual Routing under Temporal Shift
链接: https://arxiv.org/abs/2609.33340
作者: Yifan Guo
类目: Machine Learning (cs.LG)
*备注: 12 pages, 3 figures, 3 tables
Abstract:Contextual specialization can improve forecasting accuracy, but a correction selected on one historical interval may become unreliable under temporal distribution shift. To address this issue, we propose RAEGL, a Risk-Aware Evidence-Gated Learning framework for selective contextual forecasting. RAEGL retains a validated global predictor by default and activates a contextual residual only when pre-deployment evidence supports its use. The framework separates candidate selection from gate calibration and jointly evaluates randomization significance, practically meaningful gain, and temporal stability. Experiments on real-world audits and controlled panels show how RAEGL can prevent harmful contextual deployment while making conservative opportunity costs explicit. In a reconstructed Our World in Data audit, exact fallback avoids RMSE degradations of 0.0960 and 0.0239 caused by two validation-selected corrections. In a sealed World Development Indicators evaluation, a region-based correction passes the randomization test but is withheld because its gain is only 0.000092, its country-clustered 95% confidence interval crosses zero, and only 0.02% of bootstrap replicates reach the practical threshold. In controlled panels, the stability- and support-aware extension activates in 97.2% of strong, stable-context runs while rejecting all high-drift settings. These results support RAEGL as an auditable, evidence-based mechanism for managing contextual deployment risk and as a conservative alternative to validation-driven contextual selection.
[LG-269] Safe Score Matching: Diffusion Policies with Hamilton-Jacobi Reachability for Online Safe Reinforcement Learning NEURIPS2026 ATC
链接: https://arxiv.org/abs/2609.33337
作者: Boyang Li,Matthew Kim,Sylvia Lee Herbert
类目: Machine Learning (cs.LG); Robotics (cs.RO)
*备注: Accepted to NeurIPS 2026. 29 pages, 4 figures. Code: this https URL
Abstract:Online safe reinforcement learning (RL) seeks policies that maximize reward while satisfying safety constraints. A popular line of research in safe RL relaxes safety to a soft expected-cost constraint and solves the resulting Constrained Markov Decision Process via primal-dual Lagrangian updates that only enforce safety on average. To address this limitation, hard, state-wise constraints are introduced and often imposed through Hamilton-Jacobi (HJ) reachability. Yet such constraints require solving different objectives in the feasible and infeasible regions: reward maximization in the former, recovery toward the feasible regions in the latter. The resulting target action distributions are inherently multimodal, and this structure poses a fundamental challenge for the Gaussian or deterministic actors used in existing HJ-based safe RL, which often collapse onto suboptimal modes. Diffusion policies provide the expressiveness needed to represent such distributions, and recent work on Q-score matching offers a route to training them for online RL by score regression – but has been applied only to reward maximization. We propose Safe Score Matching (SSM), an off-policy actor-critic method that adapts Q-score matching to hard-constrained safe RL by gating a two-branch score target with HJ reachability: inside the feasible set, the denoising process degenerates to Q-score matching on actions classified as viable by the HJ critic; outside, a recovery branch biases denoising toward regions with lower worst-case violation. On quadrotor and fixed-wing trajectory-tracking and stabilize-and-avoid benchmarks, SSM attains the best or near-best task performance with low false-safe rates, whereas the primal-dual baseline admits more unsafe behavior and reachability-based baselines tend to be more conservative; on Safety-Gymnasium velocity tasks, SSM attains the lowest cost with competitive reward.
[LG-270] When Privacy Moves ML-Mediated Decisions On Device: Information and Incentive Misalignment in Auctions NEURIPS2026
链接: https://arxiv.org/abs/2609.33312
作者: Dipankar Sarkar
类目: Computer Science and Game Theory (cs.GT); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: 14 pages, 1 figure, 3 tables. Previously submitted to the Economics for Machine Learning (EconML) workshop at NeurIPS 2026. Code and data: this https URL
Abstract:Moving ML-mediated decision making onto privacy-preserving clients decentralises the economic decision along with the inference. Shared budget constraints then depend on information that cannot be globally current, creating an information-structure failure that conventional pacing is not designed to solve. We study this information misalignment in an auction-logic-faithful on-device simulation with 36 campaigns and 50 devices. Accounting is in dimensionless integer score units; no currency semantics are claimed. Across 30 paired demand paths, proportional Even pacing overspends 17.77% after one tick of staleness and 1,669.31% after 50 ticks under the original 20-times budget pressure. The effect does not depend on that severe a budget: at two-times pressure, 50-tick overspend remains 106.95%. A visible-budget no-sale guard makes zero-lag compliance exact at this score-unit granularity, yet leaves 11.88% overspend at one tick because other devices’ debits remain invisible. A declared bursty, heterogeneous-device sweep retains a strictly increasing mean lag curve. We derive a finite-window expected excess-debit bound under conditional charge caps and find positive paired slack in every bounded-value cell. A second, incentive misalignment arises when the ML/pacing score transformation is allowed to change payment units: 98.23% of rival auctions at one tick admit a profitable deviation. An executable implementation-level counterexample isolates the runner-up’s multiplier in the winner’s price. Critical-base-bid payment is per-auction DSIC conditional on current multipliers, but does not establish dynamic truthfulness and does not repair base-value ranking disagreement.
[LG-271] Direct Hidden-State Alignment: Mapping and Controlling Preference Expression in LLM s
链接: https://arxiv.org/abs/2609.33298
作者: Fansheng Zhang,Shengran Guo,Zexiao Wang,Liang Yuan,Jiyuan Chen,Ruikun Luo
类目: Machine Learning (cs.LG)
*备注: 32 pages, 7 figures. Code: this https URL
Abstract:In many settings, post-training need not create the target behavior from scratch: the base model can already produce it, but not reliably. This shifts part of preference alignment from capability acquisition to behavioral expression. We ask how a specified preference is represented in native model computation, what prevents target-supporting computation from reliably dominating generation, and whether this structure can directly guide control. We introduce Residual Competition Maps (RCMs), which map a behavioral preference onto signed causal effects of native residual computation. Across preference domains, RCMs reveal coexisting target-supporting and target-competing effects, input-dependent component roles, and cases where a single native-component intervention reverses the preference outcome. DPO substantially reorganizes these effects and can weaken opposition without guaranteeing its removal. We then propose Direct Hidden-State Alignment (DHSA), which treats inference-time hidden states rather than base-model weights as the direct adaptation space. RCM-guided Causal Activation State Transition (CAST) implements DHSA through local state interventions at a small number of preference-relevant interfaces while freezing the base model. With only 256-16,384 controller parameters, CAST reaches DPO-competitive operating points across three preference domains, can complement DPO-trained models, and can be enabled or removed at inference time.
[LG-272] Modelling non-linear aeroelastic loads in long-span bridges with extreme learning machines
链接: https://arxiv.org/abs/2609.33274
作者: Gledson Rodrigo Tondo,Samir Chawdhury,Sergio Andres Castro Giraldo,Guido Morgenthal
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
*备注:
Abstract:Accurate modelling of aerodynamic loads is essential for predicting instabilities and ensuring the safety of long-span bridges. A methodology is introduced for modelling aerodynamic self-excited forces in bridge-deck cross-sections using extreme learning machines (ELMs). ELMs, as single-layer feedforward neural networks, offer efficient training and accurate predictions. Forced-oscillation datasets from computational fluid dynamics (CFD) or wind-tunnel experiments are used for training, enabling systematic data selection to capture non-linear aerodynamic behaviour often missed by semi-analytical approaches. Once trained, the model predicts self-excited loads for any arbitrary motion composed by frequencies and amplitudes within the training domain. Comparisons with analytical, semi-analytical, and CFD results show superior accuracy in capturing non-linear force components and close agreement for aerodynamic loads and flutter wind speeds. Training required about 1.1% of the time of a conventional neural network, and coupled flutter analysis runs in seconds, providing orders-of-magnitude speed-ups over CFD. These results indicate that ELM-based frameworks are accurate, practical, and efficient alternatives for modelling self-excited loads, particularly when preliminary CFD or wind-tunnel data are available. The presented approach offers a reliable data-driven technique for aeroelastic load modelling in long-span bridges.
[LG-273] owards Identifiable Representations under Misspecified Structure
链接: https://arxiv.org/abs/2609.33273
作者: Yuke Li,Yujia Zheng,Ziyi Chen,Kun Zhang,Heng Huang
类目: Machine Learning (cs.LG)
*备注:
Abstract:The presence of noise that depends on the latent variables poses a fundamental challenge to identifiability. Existing results rely on conditional independence among the observations given the latent variables. We study a more general \emphmisspecified structure, where this conditional factorization does not hold, and establish both precise and approximate identifiability guarantees. We characterize structural misspecification as a perturbed factor analysis problem. For precise identifiability, we establish subspace identifiability under spectral separation and controlled perturbation, followed by component-wise identifiability under structural sparsity. When the precise condition is not guaranteed, we derive an approximate subspace-identifiability theorem. Based on these results, we develop an unsupervised variational estimator for recovering latent variables. Experiments demonstrate the effectiveness of the proposed framework.
[LG-274] GTRL: Grounding Divide-and-Conquer Value Learning with Temporal Differences
链接: https://arxiv.org/abs/2609.33259
作者: Abdul Monaf Chowdhury,MD Sameer Iqbal Chowdhury,Shifat E Arman,Md Mehedi Hasan
类目: Machine Learning (cs.LG)
*备注:
Abstract:In offline goal-conditioned reinforcement learning (GCRL), divide-and-conquer scales to long horizons by joining two shorter segments at a subgoal. However, under stochastic dynamics, the base case of this rule values the luckiest trajectories through the data. The subgoal must also lie on a shared trajectory, so a state-goal pair that no trajectory connects gets no value update at all. To address both, we present Grounded Transitive RL (GTRL), an offline GCRL value learning algorithm that grounds the divide-and-conquer update with a one-step TD target. Over a single step, TD is correct, as its target averages over the successors and needs no subgoal. GTRL adds this target to the composition rather than replacing it, so every pair receives an update, and the composition still carries the long horizon. GTRL also corrects the bias from hindsight relabeling by reweighting each goal against how reachable it was from other successors. We evaluate our algorithm on nineteen OGBench tasks spanning stochastic, deterministic, and stitching environments, where it achieves the highest average success rate. Code will be released soon.
[LG-275] wo Heads Are Better Than One: Aggregating Weaker LLM s for Better Forecasts
链接: https://arxiv.org/abs/2609.33257
作者: Cheng Peng,Ruixi Luo,Zhi Chen,Wei Tang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large language models (LLMs) are increasingly used to forecast real-world events, but access to the strongest individual forecaster may be costly or otherwise constrained. We study weak-to-strong forecast aggregation: can individually weaker LLM forecasters be aggregated to outperform a stronger forecaster? Using ForecastBench (Karger et al., 2025), we evaluate 70 LLM forecasters across 16 comparison groups, each with more than 1,000 shared subquestions, yielding 1,121 weaker-model pairs. Within each group, we identify the strongest individual by test Brier score and evaluate aggregates composed exclusively of weaker forecasters, with aggregation weights learned on separate training data. We find substantial evidence of weak-to-strong improvement. Learned linear pooling identifies a weaker pair that matches or outperforms the strongest individual in 11 of 16 groups and comes within 5% of its Brier score in all 16 groups. We also find that these improvements do not rely on having a near-best constituent and are generally accompanied by good calibration. Additional analyses show that adding more models does not consistently improve performance, and competitive weaker-model aggregates also remain available under practical constraints.
[LG-276] Feedback-Robust AI for Patient Knowledge Graphs
链接: https://arxiv.org/abs/2609.33248
作者: Mohammed Sameer Syed
类目: Machine Learning (cs.LG)
*备注: 66 pages, 4 figures
Abstract:Patient knowledge graphs from bedside monitoring should type their relations and state whether the data support their signs. In anesthesia and intensive care, clinicians titrate drugs and ventilation in response to the physiology, so temporal relations mix the patient’s response with the clinician’s policy. We introduce ClosedLoopBench: 29 relations with signs fixed by physics, pharmacology or clinical practice, on 3,442 VitalDB surgical cases (12,653 h) with negative-control action streams. When each patient’s actions are replaced by another patient’s, six of 12 estimators declare on average 11-18 of their 19-29 distinct relation estimates significant without calibration, and after calibration cross-correlation and Granger tests still assign ventilator rate - end-tidal CO2 the sign of the clinician’s policy. We propose feedback-robust patient graphs that combine concept nodes with evidence pointers, typed relations admitted against negative controls, and beat-level couplings. On VitalDB under null streams, our graphs contain 0.06-0.10 false concept-level relation instances per graph, versus 10-12 for correlational construction. Patient-specific estimates of 11 slow drug and ventilator responses predict later data no better than population estimates, whereas the pulse-arrival-time-systolic-pressure slope is negative in 94.3% of 2,884 cases and patient-specific (early-late correlation 0.67 [0.63, 0.70]).
[LG-277] Minimax-Optimality of Posterior Sampling for Reinforcement Learning
链接: https://arxiv.org/abs/2609.33246
作者: Taewon Goo,Kihyuk Hong
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 44 pages
Abstract:Posterior sampling for reinforcement learning (PSRL) is one of the simplest and most effective exploration methods, but a basic question has remained open: does unmodified PSRL achieve minimax regret without structural assumptions on the prior? We answer yes. Exact vanilla PSRL is minimax optimal in leading-order Bayesian regret under arbitrary correlated priors. The difficulty is that a posterior-sampled transition model is coupled with its own continuation value. We overcome this with a common empirical transition reference that isolates the resulting value mismatch and a Bellman-based variance argument that controls it without an extra leading-order state-space factor. For finite-horizon, time-inhomogeneous tabular MDPs with unknown stochastic rewards, this yields the minimax \widetildeO(\sqrtSAH^3K) regret rate under arbitrary joint priors over rewards and transitions. The same proof principle gives the minimax \widetildeO(d\sqrtH^3K) rate for linear-mixture MDPs under arbitrary joint parameter priors.
[LG-278] he Price of Locality: Why Forward-Forward Underperforms Backpropagation? NEURIPS2026
链接: https://arxiv.org/abs/2609.33240
作者: Zhaoxian Wu,Haichuan Liu,Tianyi Chen
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: This paper is accepted by NeurIPS 2026
Abstract:The Forward-Forward Algorithm (FFA) replaces backpropagation (BP) with layer-wise local contrastive objectives, eliminating the backward pass and the need to retain intermediate activations, yet suffers a persistent performance gap with BP that worsens with depth. This paper diagnoses two structural failure modes: an optimization floor arising from concurrent local updates; and a geometric collapse of layer representations driven by the local update mechanism. On the optimization side, we prove that the FFA loss satisfies the Polyak–Lojasiewicz inequality at each layer; however, simultaneous layer updates induce inter-layer representation-distribution drift, so each layer optimizes against a moving input distribution and incurs an error floor. On the representational side, the pairwise similarity kernel of layer representations contracts exponentially toward rank one as depth increases, collapsing the diversity of per-layer error signals. This collapse bounds FFA’s effective learning capacity, which measures the diversity of gradient information across layers, independently of depth, whereas BP’s chain-rule signal preserves per-layer diversity, yielding a capacity that scales with depth.
[LG-279] MTLiquid: Enabling Efficient Multi-Task Learning using Liquid Neural Networks for Lightweight Healthcare Monitoring Systems
链接: https://arxiv.org/abs/2609.33232
作者: Rachmad Vidya Wicaksana Putra,Fahad Abdul Rauf,Muhammad Shafique
类目: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
*备注: 8 pages, 1 figure, 2 tables
Abstract:Continuous-time sensing and monitoring with timely and accurate decision-making are critical for many real-world applications. In healthcare monitoring systems, physiological signals are often available or sampled at irregular time intervals, hence requiring continuous-time processing to provide accurate prediction. Moreover, such systems often need to solve multiple detection/prediction tasks to provide a comprehensive patient review from different physiological aspects for more accurate decision-making. To solve this, continuous-time neural networks (CTNNs) can be employed. However, state-of-the-art works typically solve only one task at each network, thereby limiting their efficiency gains. To address this limitation, we propose MTLiquid, a novel methodology to enable efficient multi-task learning in continuous-time processing for healthcare monitoring systems through effective network design and training strategy. MTLiquid employs: (1) multiple input and output heads to accommodate different tasks, while sharing the same backbone network across tasks; as well as (2) an effective training strategy that leverages a loss-weighting technique to balance learning updates across different tasks and a proportional data presentation technique to address imbalanced dataset sizes. Experimental results for mortality prediction (P12) and sepsis early detection (P19) tasks for ICU patients show that, MTLiquid achieves strong performance (AUROC: 0.84 for P12 and 0.94 for P19) comparable to the state-of-the-art single-task learning in both continuous-time networks (AUROC: 0.84 for P12 and 0.95 for P19) and discrete-time networks (AUROC: 0.79-0.82 for P12 and 0.92-0.94 for P19), while incurring significantly smaller memory cost by 44%-94%. These results highlight the potential of our MTLiquid methodology to enable lightweight continuous-time healthcare monitoring systems for better decision-making.
[LG-280] RMB: Reward Model Boosting Mitigates Reward Hacking
链接: https://arxiv.org/abs/2609.33221
作者: Jiabin Fan,Dezhi Ye,Yongchang Hao,Lili Mou
类目: Machine Learning (cs.LG)
*备注: Published in Transactions on Machine Learning Research (TMLR), September 2026
Abstract:Reinforcement Learning from Human Feedback (RLHF) is a powerful technique for aligning large language models (LLMs) with human preference. However, it often suffers from the reward hacking issue, where policy optimization improves the proxy reward model while actually degrading performance with respect to the true human preference, due to the imperfection of the proxy. To address this, we propose Reward Model Boosting (RMB), a novel approach that enhances the robustness and reliability of the reward signal for RLHF. RMB first trains a set of reward models with a diversity-promoting regularizer. This encourages each model to learn complementary aspects of the reward landscape. Then, RMB learns a lightweight aggregator in the principle of boosting to aggregate the outputs of the diverse reward models into a more accurate and robust reward signal. Our extensive experiments demonstrate that RMB significantly improves reward accuracy on both in-distribution and out-of-distribution datasets, substantially mitigating the reward hacking issue and ultimately improving RLHF performance.
[LG-281] MorphAtt: A Neuromorphic Accelerator for Efficient Multi-Head Attention Processing in Spiking Vision Transformers
链接: https://arxiv.org/abs/2609.33207
作者: Rachmad Vidya Wicaksana Putra,Amirhesam Jafari Rad,Muhammad Shafique
类目: Hardware Architecture (cs.AR); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
*备注: 9 pages, 7 figures, 2 tables
Abstract:Spiking Vision Transformers (SViTs) are developed as an energy-efficient alternative to conventional ViTs for computer vision tasks at the edge. However, huge parameter counts and complex multi-head self-attention (MHSA) operations make it challenging to achieve high energy efficiency in SViT inference, especially in tightly constrained applications. To maximize efficiency gains of SViT processing, we propose MorphAtt, a novel digital accelerator that expedites SViT inference through streamlined processing. Specifically, it processes MHSA operations using cascaded hardware modules: a Spiking Query-Key-Value generator (SpikeQKV), a low-complexity Spiking Multi-Head Self-Attention engine (SpikeAtten), and Reparameterization Convolution (RepConv) modules. To mitigate traffic congestion in on-chip memory accesses and data reuse, specialized inter-module buffers are integrated within the dataflow. Under synthesis using 32nm CMOS technology, MorphAtt achieves 792-1605 GOPS of throughput, while incurring ~39-55 mW of power consumption and 1.5 mm^2 of area, which lead to 20.3-29.1 TOPS/W of energy efficiency. These results also demonstrate that our MorphAtt offers better performance and efficiency trade-offs than state-of-the-art, thereby enabling highly energy-efficient vision-based AI systems at the edge.
[LG-282] Orthogonal Witness Control for Muon Optimization via Sigmoid Spectral Reshaping
链接: https://arxiv.org/abs/2609.33194
作者: Dat Phi Van,Ngo Vu Minh,Tuc Nguyen,Thin Nguyen,Ngoc-Thanh Dinh,Trung Le
类目: Machine Learning (cs.LG)
*备注:
Abstract:Matrix-valued optimizers such as Muon exploit the spectral structure of neural network updates through Newton–Schulz orthogonalization, but their near-flattening of the singular spectrum discards relative magnitude information across gradient modes. We introduce \emphSoren (\textbfSpectral \textbfOrthogonal \textbfReshapi\textbfng), a matrix-valued optimizer that preserves the singular subspaces of the gradient while applying a bounded, monotone sigmoid transformation to its singular values. This smoothly compresses dominant modes without fully flattening the spectrum. We interpret Soren as a positive-definite preconditioned gradient method and establish convergence guarantees under relative smoothness and metric Polyak–Łojasiewicz geometry. To avoid explicit singular value decomposition, we further develop a finite-depth Soft Newton–Schulz (SNS) polynomial realization of the sigmoid spectral map and characterize how its spectral approximation affects the induced convergence geometry. Experiments across LLM pre-training, supervised fine-tuning, and direct preference optimization demonstrate the effectiveness and robustness of Soren against established optimizers.
[LG-283] Apparent Compression Real Stability: The Intrinsic Dimension of Learning a Quantum Wavefunction
链接: https://arxiv.org/abs/2609.33193
作者: Lu Wei,Yufeng Wang,Chenfeng Cao,Haibin Ling
类目: Machine Learning (cs.LG); Quantum Physics (quant-ph)
*备注: 18 pages, 9 figures, 6 tables
Abstract:How many directions in weight space does training need? The intrinsic dimension answers this with the smallest number of random directions in which training still reaches a target accuracy, and small values have motivated parameter-efficient methods such as LoRA. We measure it for variational Monte Carlo (VMC), which trains a neural network to represent the ground state of a quantum many-body system. VMC is a demanding test, because the network generates its own training samples and every gradient is noisy, and a revealing one, because the exact answer is known and every run can be scored. We train only a small latent vector that a frozen random map turns into the network’s weights, with no change to the standard natural-gradient optimizer. We find that a small dimension can be misleading, while the stability it brings is real. On a magnet with a hard sign pattern, a network that cannot represent signs reaches its best energy in 8 of 28,642 directions, but only because no such network can go lower; once signs are learnable, neither the signs nor the magnitudes are cheap. The dimension rises across a quantum phase transition, so it tracks how difficult a state is at far less compute than fitting a scaling law, yet it never falls below a floor set by the random subspace itself, even where the ground state is nearly trivial. Training in the subspace, in contrast, never diverged in our experiments, whereas full-parameter training with the same settings did, and a control with matched solvers attributes the difference to the reduced dimension.
[LG-284] When Does Backpropagating Through Policy Memory Matter? Physical Credit Optimizer Updates and Observability
链接: https://arxiv.org/abs/2609.33169
作者: Xingjian Li,Yi Han,Jianhua Z. Huang
类目: Machine Learning (cs.LG); Robotics (cs.RO)
*备注: 33 pages, 9 figures, 20 tables, Under review at TMLR
Abstract:Policies with memory can learn along two backward paths: through the physical states their actions produce and through the representations they store. Transformer-XL and truncated backpropagation through time cut the second path at stored history while keeping its values. We ask when this cut matters. Holding the forward computation fixed and varying only derivative edges, we measure parameter gradients, the updates the optimizer applies, and continued training in a Transformer vessel-trajectory model and a quadrotor tracking policy. In the vessel model, detaching the key-value cache shrank the gradient to about a tenth of its norm, with little rotation, when gradients flowed through all earlier physical states, but barely changed it under one-step physical credit. In this strongly clipped regime the optimizer, not the gradient, set how far updates differed: global-norm clipping removed most of the gradient difference between memory-cut graphs, whereas AdamW turned a 2% gradient difference between two placements of the cut into update differences of up to 31% at the step where the placement was switched. In a quadrotor trained from initialization with 0.20 m/s velocity noise, removing memory raised tracking error by 43% and cutting memory gradients raised it by 32%; at low noise the cut’s mean cost exceeded the value of memory. Two-step truncation segments gave no measurable gain, although with hidden velocity a two-step window captured most of the value of memory; eight-step segments removed half to three quarters of the cost. Switching the cut on only for the last fifth of training understated its cost about threefold at 0.20-0.30 m/s, but not at low noise or with hidden velocity. These results suggest measuring the cost of a memory cut by training with it from initialization, and comparing backward graphs by the updates the optimizer applies rather than by raw gradients.
[LG-285] Downstream-Aware Context Selection for Online In-Context Reinforcement Learning
链接: https://arxiv.org/abs/2609.33166
作者: Ruihan A. Li,Shangtong Zhang,Rohan Chandra
类目: Machine Learning (cs.LG)
*备注:
Abstract:In-context reinforcement learning (ICRL) enables large language model agents to adapt to new environments using their interaction history without updating model parameters. However, repeatedly conditioning on growing histories can lead to substantial token cost. We propose a bounded-history context-management framework that predicts the task-dependent downstream effect of removing historical interactions to guide history selection and determine a decision-dependent context budget. Formally, our framework uses the full rolling history as a reference. The predictor evaluates removal effects, defines a deletion ordering, and applies a shared selection criterion to determine how much history to retain at each decision. We evaluate the method in closed-loop SUMO driving under held-out in-distribution, unseen-domain, and unseen-route settings, and in ScienceWorld under a continual ICRL protocol. Relative to a baseline using the full context, our method reduces total token usage by 25.7%, 25.8%, and 23.2% across the three driving settings while maintaining comparable closed-loop driving performance. In ScienceWorld, it reduces total token usage by 52.1% compared to full context and uses 30.2% and 37.8% fewer tokens than the Recent and Similarity baselines, respectively, while maintaining performance.
[LG-286] Convergence of Practical Muon
链接: https://arxiv.org/abs/2609.33152
作者: Haonan Wang,Yu Wu,Minghui Liwang,Xinlei Yi,Yiguang Hong
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Muon is emerging as a promising alternative to AdamW for large-scale neural network training, yet theoretical understanding of its practical implementation remains incomplete, as existing analyses often simplify or omit two key components: (i) practical Newton–Schulz iterations with empirically tuned polynomial coefficients (3.4445,-4.7750,2.0315) ; and (ii) decoupled weight decay for regularization. In this paper, we provide an optimization interpretation and establish convergence for practical Muon, jointly accounting for both components. Specifically, we interpret practical Muon as right-preconditioned optimization of the original loss with a dynamic weighted \ell_2 regularizer that vanishes as stationarity is approached, so that the optimization target remains the original objective. We then establish, to our best knowledge, the first convergence guarantee for practical Muon in the stochastic nonconvex setting, with an \mathcalO(T^-1/4) convergence rate in terms of the expected Frobenius norm of the gradient, improving the dimension dependence of the best known AdamW’s convergence rate by a factor of \sqrtd , where T is the iteration horizon and d is the parameter dimension. Experiments further support the theoretical convergence results.
[LG-287] CFLoRA: Federated Fine-tuning of LLM s with Complementary Factors for Error-free Aggregation
链接: https://arxiv.org/abs/2609.33147
作者: Yanan Ma,Qiyuan Chen,Zihan Fang,Xianhao Chen,Yuguang Fang
类目: Machine Learning (cs.LG)
*备注: 25 pages, 2 figures
Abstract:Federated low-rank adaptation (LoRA) enables collaborative fine-tuning of large language models without centralizing private client data. Its factorized update, however, creates a structural mismatch in federated averaging: averaging the two LoRA factors separately does not equal averaging their products. Existing exact methods resolve this issue mainly by freezing an entire factor or alternating factors across rounds, but none can update factors simultaneously without aggregation errors or expanding communication ranks. To address this fundamental problem, we present \textttCFLoRA, a federated LoRA scheme that partitions latent LoRA channels into two complementary sets in every communication round. By ensuring that columns and rows are complementary across factors, we eliminate bilinear terms in matrix multiplications, making federated aggregation exact. Crucially, our framework also supports clients with heterogeneous rank budgets. Convergence analysis validates \textttCFLoRA achieves \mathcalO(1/\sqrtT) convergence rate of the \textitoriginal LoRA objective in homogeneous-rank cases. Extensive experiments with RoBERTa on the GLUE benchmark and with LLaMA-3.2-3B-Instruct on commonsense reasoning tasks demonstrate that \textttCFLoRA achieves superior performance and training efficiency compared to state-of-the-art federated LoRA baselines.
[LG-288] Leaky Students: Membership Inference against On-Policy Distillation
链接: https://arxiv.org/abs/2609.33136
作者: Zhexi Lu,Mingzhi Zhu,Stacy Patterson,Lei Yu
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注:
Abstract:On-policy distillation (OPD) trains a student to match a teacher’s next-token distributions on student-generated trajectories. However, privileged information supplied to the teacher for OPD training may contain sensitive data. Whether the student leaks private information about the records supplied to the teacher during distillation remains poorly understood. To the best of our knowledge, we present the first systematic study of membership inference in this setting. We find that fresh student trajectories expose sparse membership signals that fixed reference-answer losses often miss. These signals are mixed with probability changes caused by training on other records. We introduce Leaky, which samples fresh trajectories from the target model and compares its token log-probabilities with the maximum across matched reference models trained without the candidate records. It applies Leaky ReLU to the resulting gaps, preserving positive gaps and downweighting negative gaps as an approximate correction for incidental positive gaps in non-members. Across fifteen targets spanning mathematics, medical question answering, and code generation, Leaky outperforms all evaluated baselines and achieves mean AUROC 0.875, compared with 0.614 for the strongest baseline on each target in the main evaluation. On the same sampled trajectories, the strongest baseline achieves mean AUROC 0.826. These results show that students trained through OPD can expose the membership of records used for teacher supervision, even when fixed reference-answer losses provide little evidence of membership.
[LG-289] ILP-BO: Integer Linear Programming-Based Black-Box Optimization
链接: https://arxiv.org/abs/2609.33131
作者: Hyakka Nakada,Shu Tanaka
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: 20 pages, 11 figures
Abstract:Black-box Optimization (BO) is a powerful framework for optimizing expensive objective functions or unknown functions with a limited number of evaluations. A central step of standard BO such as Bayesian optimization is the optimization of a surrogate-based acquisition criterion, which is commonly performed using nonlinear optimization or heuristic search. Therefore, conventional black-box optimization generally does not guarantee global optimality in candidate selection. In this study, we propose Integer Linear Programming-based Black-box Optimization (ILP-BO), a quasi-Bayesian optimization framework that transforms kernel-based surrogate optimization over discrete domains into an Integer Linear Programming (ILP) problem. The key idea is to represent nonlinear kernel functions exactly on finite discrete distance levels by introducing binary one-hot auxiliary variables. This transformation converts the nonlinear surrogate into a linear objective with linear constraints and binary variables. To incorporate exploration while preserving the linear structure, we further introduce a Hamming-distance margin that excludes neighborhoods around previously observed points. We derive the proposed formulation for several standard kernels and obtain an analytical upper bound on the Hamming-distance threshold based on the measure in the binary search space. The resulting candidate-selection problem can be solved by integer programming solvers with certificates of optimality. Thus, our methodology has the potential to serve as a highly transparent black-box optimization framework. Experiments on synthetic and discrete optimization benchmarks show that ILP-BO achieves competitive optimization performance compared with practical Bayesian optimization methods.
[LG-290] Mycelium: A Generalizable Cross-Grid Multi-Task Model for Electrical Distribution Systems
链接: https://arxiv.org/abs/2609.33120
作者: Zhengyang Wei,Shourya Bose,Helgi Hilmarsson,Elena Carnio,Dhruv Suri
类目: Machine Learning (cs.LG)
*备注:
Abstract:Electrical distribution grid operations require inference across heterogeneous networks from sparse, noisy, and incomplete time series measurements. In this work, we identify challenges and explore solutions towards a unified model that can perform diverse tasks grounded in the physics of the electric grid and generalize to unseen distribution networks. We define a unified grid ontology that represents variable sized distribution networks as heterogeneous graphs while preserving native topology, asset types, and electrical relationships across networks. We develop a physics based data simulation pipeline that combines reference and procedurally generated distribution networks with network reconfigurations, fault scenarios, and configurable sensing conditions. We present Mycelium, a heterogeneous graph transformer with structure aware communication edges and electrical reference features that encode network position and nominal phase orientation, together with task specific temporal readouts which generate per task outputs. We train Mycelium on reference as well as synthetic grids, and study its generalization on benchmark networks completely excluded from training and validation. Mycelium is observed to outperform task specific neural baselines on most reported benchmark metrics. Architectural ablations and the aforementioned studies reveal Mycelium’s capability to learn representations of the underlying physics which serves to enhance cross-task performance, thereby addressing a significant challenge in unified grid models.
[LG-291] Simulation-Free Learning of GP-SDEs from Irregular Observations
链接: https://arxiv.org/abs/2609.33112
作者: Zhidi Lin,Yuhao Liu,Ying Li,Edwin Fong,Petar Djurić
类目: Machine Learning (cs.LG); Signal Processing (eess.SP); Machine Learning (stat.ML)
*备注:
Abstract:Gaussian process stochastic differential equations (GP-SDEs) provide a flexible Bayesian model for unknown continuous-time state dynamics with uncertainty quantification, but learning and inference from noisy and irregular observations remain computationally challenging. To address this issue, we propose GP-SDE Matching, a simulation-free variational framework for Bayesian GP drift learning and continuous-time state smoothing. We analytically marginalize the sparse GP posterior to derive a tractable drift-matching objective that accounts for both the posterior mean and uncertainty of the unknown drift. To handle irregular observations, we further introduce an irregular-time-aware variational state posterior that incorporates the actual observation times during both encoding and continuous-time marginal querying. Experiments on the stochastic Lorenz–63 system demonstrate substantially improved drift recovery and state reconstruction under irregular observations, while five system identification benchmarks show robust forecasting under increasing observation sparsity and competitive performance against existing latent-SDE and state-space methods.
[LG-292] CARVE: Breaking Data Barriers in Chip Placement by Harnessing Reusable Expertise
链接: https://arxiv.org/abs/2609.33106
作者: Jiefu Zhang,Haixiang Sun,Yang Xu,Vaneet Aggarwal,Zishen Wan
类目: Machine Learning (cs.LG)
*备注: 39 pages, 8 figures, 37 tables
Abstract:Pretrained macro-placement policies can reduce repeated optimization across circuits, but deployment often exposes them to unfamiliar designs when the original training data are unavailable. Repeatedly fine-tuning a single serving model can overwrite earlier improvements, while simply saving checkpoints does not determine where they can be reliably reused. We introduce Continual Adaptation through the Reuse of Validated Expertise (CARVE), a framework that represents accumulated expertise as a frozen base policy, immutable specialists, and task-specific credentials obtained through local validation. For a new task, CARVE first checks existing specialists and trains a new specialist from the frozen base only when none qualifies. Under fixed task distributions and validation rules that control cumulative error, we establish expected-performance guarantees for repeated reuse. For bounded losses, we also derive matching worst-case bounds on the local samples needed for reliable reuse. In macro placement, a reuse-first follow-up reduces recorded training time by 58.5% (9.66 to 4.01 hours), while mean HPWL gain changes only from 8.41% to 7.86%. In a simulated receiving deployment, imported specialists are reused on six of seven new IBM circuits with no receiver-side training, achieving a 5.76% mean HPWL gain. Navigation studies provide complementary evidence on repair retention and repeated adaptation.
[LG-293] Evolving Dexterous Robots from Scratch
链接: https://arxiv.org/abs/2609.33101
作者: Zihan Guo,Shuzhe Zhang,Muhan Li,Peiyang Li,Sam Kriegman
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注:
Abstract:Little is known about how to manually design agents capable of dexterous manipulation. Some design principles have been inferred from close examination of how animals manipulate objects, but these structures and behaviors have so far resisted biomimicry and may not be optimal for artificial machines. Here we evolve freeform robots to pick up, hold, rotate, and use diverse objects. Unlike other approaches to optimizing robot hands, we do not presuppose the presence, articulation, or geometry of any part of the body. Although familiar prehensile forms such as tails, beaks, paws and claws may emerge spontaneously under certain conditions–and while such conditions could be of interest to evolutionary biologists–de novo manipulator design can also reveal whole new solutions, overlooked or unknown structures which may be better suited for the task at hand. We use contrastive learning to create a highly searchable genetic embedding of design space, an autoregressive developmental model to decode designs, evolutionary strategies to find good designs, and reinforcement learning to train each evolved design. Winning designs were automatically converted into a manufacturable blueprint, printed, assembled and tested in the real world in a zero-shot manner. The results represent the state-of-the-art in evolutionary robotics in terms of performance, diversity and complexity.
[LG-294] Geometry-Aware Operator Families for Structured Representation Learning
链接: https://arxiv.org/abs/2609.33089
作者: Zuyuan Zhang,Fei Xu Yu,Tian Lan
类目: Machine Learning (cs.LG)
*备注:
Abstract:The geometry of latent representations governs which components should interact and how information should propagate, making geometry-aware operator design a fundamental ingredient of structured deep representation learning. However, existing neural architectures typically rely on generic operator templates or geometry-specific constructions, creating a need for a unified framework that can derive admissible operators directly from fixed structural information while remaining adaptive to changing contexts. We introduce \emphGeometry-Induced Operator Families (GIOF), a general framework that converts fixed geometry into a structured family of propagation operators and dynamically selects an appropriate member of this family according to the current context. GIOF first transforms geometry-derived interaction channels into reusable generator bases, then combines them through a context-dependent selector and adaptive propagation scale, and finally realizes the selected operator through stable continuous-time propagation and a bottleneck residual layer. We establish theoretical guarantees covering parameter compression, identifiability, stability, locality, compositional structure, and oversmoothing behavior, while controlled experiments validate these mechanisms and experiments on PEMS-BAY and METR-LA achieve the lowest mean MAE across all reported regional-outage settings, improving over the strongest retained baseline by 2.4%–8.8% at 30% missing sensors.
[LG-295] From Constitutions to Control: Interpretable Rewards for Aligning Language Models
链接: https://arxiv.org/abs/2609.33086
作者: Johann D. Gaebler,Calvin Isley,Max Lamparth,Stephen Casper,Sharad Goel
类目: Machine Learning (cs.LG)
*备注:
Abstract:Current approaches to aligning language models often make it hard to know what behavior is being rewarded or to change that reward in a targeted way. In particular, standard preference-based methods collapse multiple considerations into aggregate human judgments, obscuring what drives the resulting reward, while principle-based methods specify high-level values without fully operationalizing them. To address this gap, we develop a rubric-based framework to transform a general-purpose constitution into an interpretable and tunable reward model, using constitution-guided AI feedback to estimate initial weights for the constituent rubric items. We then reweight those dimensions to construct modified rewards for training. Across experiments on political alignment and safety-helpfulness tradeoffs, reweighting individual dimensions predictably changes targeted behaviors largely independently while navigating tradeoffs between conflicting alignment objectives. We show that the same framework can mitigate label bias encoded in preference judgments – including sycophancy and demographic bias – by reducing their influence on the training reward. Our results demonstrate that constitution-derived, interpretable rewards can translate high-level alignment principles into more transparent and controllable model behavior.
[LG-296] he limits of exactness: On the failure of automatic differentiation in physics-informed machine learning
链接: https://arxiv.org/abs/2609.33078
作者: Ameya D. Jagtap
类目: Machine Learning (cs.LG)
*备注: 18 pages, 2 figures
Abstract:Automatic differentiation (AD) lets neural networks compute derivatives of governing equations to machine precision, and this precision has made it the computational backbone of physics-informed machine learning. Yet exactness in the mathematical sense is not the same as fidelity to the physics. Here I argue that a derivative can be numerically perfect and still be the wrong derivative for the problem at hand, because AD, by construction, has no notion of the physical structure a solution must obey. Convection and its associated directionality, diffusion, and dispersion are only the most visible instances of a much longer list that spans all branches of computational science and engineering, including conservation, thermodynamic consistency, symmetry, symplectic structure, positivity, monotonicity, and boundedness. Recognizing this broader gap reframes how the field should build the next generation of PDE-driven neural surrogates.
[LG-297] KernelZero: Co-Evolving Proposer and Coder for Continuously Improved GPU Kernel Generation
链接: https://arxiv.org/abs/2609.33074
作者: Changxin Ke,Rui Zhang,Zixiang Fang,Zhenghong Li,Yuanbo Wen,Jiashuo Shen,Shuo Wang,Jiaming Guo,Ling Li,Qi Guo,Yunji Chen
类目: Machine Learning (cs.LG); Software Engineering (cs.SE)
*备注: 59 pages, 7 figures
Abstract:High-performance GPU kernels are essential to modern machine learning systems, yet automatically generating kernels that are both correct and efficient remains challenging. Existing LLM-based approaches face two major limitations: the scarcity of high-quality training data aligned with the model’s current capabilities, and the inherent trade-off between kernel correctness and performance. To address these challenges, we propose KernelZero, a co-evolution framework that continuously improves GPU kernel generation through two specialized models: a Proposer that generates Torch modules from API sets and a Coder that translates them into CUDA or Triton kernels. KernelZero uses a frontier-driven module generation mechanism to continuously produce capability-aligned training modules based on the Coder’s current weaknesses. It further introduces Correctness-Aware Group Relative Policy Optimization (CA-GRPO), which optimizes performance only after correctness becomes sufficiently reliable. By alternating the optimization of the Proposer and Coder, KernelZero forms an automatic curriculum that enables targeted and training-efficient capability improvement. Empirically, KernelZero-7B surpasses Claude-4.5-Sonnet on CUDA and DeepSeek-V4-Pro on Triton. On KernelBench Level 1 and 2, it achieves CUDA pass@1 scores of 75.8% and 69.6%, respectively, with pass@10 reaching 100% and 97%. On Triton, it achieves pass@1 scores of 77.2% and 72.5%, respectively.
[LG-298] Deep Learning Techniques for Phoneme Recognition in Italian Children s Speech
链接: https://arxiv.org/abs/2609.33060
作者: Nicola Barbaro,Cristina Gena,Francesco Petriglia,Andrea Meirone,Alessandro Mazzei,Arianna Viotti
类目: Machine Learning (cs.LG)
*备注:
Abstract:Speech therapists often face difficulties diagnosing impairments due to the lack of efficient tools for transcribing speech into the International Phonetic Alphabet (IPA). This work addresses this challenge with Broca, a Conformer-based deep learning system pretrained on 8 days of adult speech and fine-tuned on a 165-minute dataset of Italian child speech collected through a range of standardized diagnostic tests for children aged 3.5-6.5. Broca was optimized to handle phonetic variability in children’s speech, including tone, accent, and speech errors, and achieved a state-of-the-art weighted Phoneme Error Rate of 13.36% on Italian speech. Remarkably, this performance was obtained using less than three hours of child-specific data, underscoring the model’s efficiency and robustness in low-resource clinical settings. This work demonstrates that accurate, vocabulary-independent speech-to-IPA transcription can be achieved with minimal data, paving the way for more accessible, data-efficient tools to support speech assessment and diagnosis.
[LG-299] SketchSSM: Write to the Full State Read from a Compact Sketch
链接: https://arxiv.org/abs/2609.33051
作者: Omin Kwon,JoongWon Shin,Minseo Kim,Kurt Keutzer,Sehoon Kim,Jae W. Lee
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注:
Abstract:Hybrid-attention models replace most softmax attention layers with linear attention, reducing KV-cache growth and enabling larger decode batches where recurrent state access becomes a major bottleneck. ReplaySSM amortizes state updates by buffering keys and values, but each new query still requires a full-state read even though the state remains unchanged between state updates. We observe that low-rank state-weighted query approximation accurately preserves state-read outputs. Although future queries are unknown, the basis vectors used to approximate them can be fixed offline. Based on this observation, we introduce SketchSSM, which preserves full-state updates while approximating reads. At each state update, SketchSSM reads the full state once to precompute outputs for these basis vectors, storing them in a compact sketch. Each subsequent decode step combines the sketch vectors with query-dependent coefficients to reconstruct the output without a full-state read. Across four Mamba-2-, GDN-, and KDA-based models, SketchSSM reduces state-access traffic by approximately 10x while largely preserving average accuracy across four decode benchmarks and recall on four RULER retrieval tasks. On one NVIDIA B300, linear-attention kernel speedups over the standard vLLM baseline reach 7.78x, 5.22x, and 5.20x for Mamba-2, GDN, and KDA, respectively, with up to 2.64x higher decode throughput on Nemotron 3 Super.
[LG-300] DevelopmentODE: Structured Neural ODEs for Early Brain Development Dynamics Across a Decade
链接: https://arxiv.org/abs/2609.33048
作者: Kaiqiao Han,Haitao Chen,Bryan Quah,Xiaoda Wang,Janelle Liu,John H Gilmore,Wei Gao,Yizhou Sun
类目: Machine Learning (cs.LG)
*备注:
Abstract:Understanding how individual brain development unfolds over childhood requires modeling developmental trajectories from sparse longitudinal observations. Long-term neurodevelopmental forecasting is challenging because each child is typically observed at only a few irregularly spaced visits, while developmental dynamics vary across individuals and age. Generic continuous-time models accommodate irregular timing but often absorb these factors into a single flexible transition function, providing little structure for how population progression, individual variability, and developmental age shape the dynamics. We propose DevelopmentODE, a structured continuous-time framework that organizes population- and subject-specific variation within a shared developmental geometry while allowing the governing dynamics to evolve with age. The model builds this geometry around a developmental canal representing the population trajectory, whose local direction provides a reference for organizing subject-specific variation. Subject deviation velocities are constrained relative to this direction, while a shared nonlinear deviation field captures individual developmental motion without disrupting population-level progression. DevelopmentODE further models developmental non-stationarity through ordered age-dependent deformations of the shared vector field, progressively adapting a common dynamical structure as age changes, while elapsed time determines the integration horizon. This formulation uses population-level developmental structure to guide learning from sparse individual trajectories while allowing dynamics to evolve smoothly with age. We evaluate DevelopmentODE on longitudinal fMRI by predicting future functional connectivity of the same child from earlier observations. DevelopmentODE consistently outperforms competing baselines across short- and long-horizon predictions.
[LG-301] Cost-free Spectral Estimation for Adaptive Newton–Schulz in Matrix Optimizers
链接: https://arxiv.org/abs/2609.33047
作者: Kristi Topollai,Anna Choromanska
类目: Machine Learning (cs.LG)
*备注:
Abstract:Matrix optimizers such as Muon transform each momentum matrix through an approximate orthogonalization, typically implemented by a small number of Newton-Schulz matrix multiplications. The quality and cost of this approximation depend strongly on the singular-value spectrum of its input, yet existing implementations use the same fixed polynomial routine for every layer and throughout training. We show that this uniform treatment is unnecessary: the computations in the Newton-Schulz method already reveal enough information to make the method adaptive. The Gram matrices formed inside Newton-Schulz iterations yield spectral moments through inexpensive scalar reductions, requiring no additional matrix multiplications. From these moments, we recover an estimate of the empirical singular-value distribution and use it to select a polynomial routine specialized to the current matrix. This turns Newton–Schulz orthogonalization into a spectrum-adaptive procedure that responds to differences across both layers and training time. On saved momentum matrices, spectral estimation substantially reduces orthogonalization error at a fixed iteration budget or reaches the same accuracy with fewer iterations, and in GPT pretraining up to 1B parameters it lowers the validation loss of two matrix optimizers. Our results suggest that matrix-function operations inside optimizers need not be designed for a conservative worst-case spectrum: they can cheaply measure the spectrum they are already processing and specialize computations accordingly.
[LG-302] Balancing Early Performance Sacrifices with Long-Term Gains: Scaling Learning-Rate Warmup Duration Across Training Horizons
链接: https://arxiv.org/abs/2609.33041
作者: Kristi Topollai,Anna Choromanska
类目: Machine Learning (cs.LG)
*备注:
Abstract:Learning-rate warmup is a standard technique in language-model training, yet its duration remains largely heuristic. Common approaches use either a fixed number of updates or a fixed fraction of the training horizon, two choices that imply very different scaling as training gets longer. When should warmup stay fixed, and when should it grow with the horizon? We address this question with a quadratic model whose modes respond differently to the peak learning rate. Warmup slows progress in directions that already contract well at the peak rate, but can remove persistent error in directions near the stability edge, with higher peak rates shifting the balance toward longer warmup durations. This yields a compact horizon scaling law that captures regimes ranging from essentially no warmup, through fixed-duration warmup, to durations that grow with the training horizon, and explains how the preferred regime changes with peak learning rate. Because the law captures the tradeoff between giving up early progress and improving the trajectory that follows, it can be fit using shorter runs and used to predict warmup at substantially longer horizons. Together, our results explain several familiar properties of warmup through a single tradeoff and suggest treating warmup duration as a horizon-dependent hyperparameter rather than a fixed training heuristic.
[LG-303] What Must a World Model Distinguish for Planning ?
链接: https://arxiv.org/abs/2609.33030
作者: Rongzhe Wei,Hans Hao-Hsun Hsu,Peizhi Niu,Yifan Li,Pan Li
类目: Machine Learning (cs.LG); Robotics (cs.RO)
*备注:
Abstract:World models simulate the consequences of action candidates, but good planning need not preserve every physical distinction required for accurate prediction. We formalize this gap through a hierarchy of mechanism, response, and decision sufficiency. Given a candidate set, the planning query determines which physical variations matter and how precisely they must be preserved: coarse decisions can discard much of the information needed for prediction, whereas fine decisions may require nearly the same resolution. In practice, planners often adaptively search to construct candidates, and information unnecessary for final selection may still be needed to discover good candidates. What a world model must preserve therefore depends on the query, the candidate set, and the planner. We study these effects in a collision system, nonlinear dynamics, and robotic planning. These varying requirements raise a design question: where should query information enter the planning system? A model that jointly generates actions and outcomes conditioned on the query achieves lower regret than an action-conditioned world model on seen objectives, but this advantage largely disappears when generalizing to unseen objectives. Motivated by this, we propose a modular design in which the query determines where to look and an action-conditioned model predicts what will happen, allowing the same predictions to be reused across objectives.
[LG-304] Low-Rank Single-Index Bandits with Unknown Links: From Matrices to Tensors
链接: https://arxiv.org/abs/2609.33025
作者: Zhongxuan Liu,Yue Kang,Thomas C. M. Lee
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Low-rank matrix and tensor bandits exploit structured interactions but typically assume a known reward link. Recent single-index bandit methods accommodate unknown links without directly exploiting matrix or tensor rank. We address this gap by studying stochastic matrix and tensor bandits with an unknown shared Lipschitz link and a low-rank index parameter under known regular candidate distributions and finite-variance noise. For monotone links, T-ESTOR combines robust, rank-adaptive Stein estimation with epoch-based greedy selection. Under exact selected-score access and a uniformly positive selected-design Stein signal, it achieves square-root regret with dimension dependence determined by the low-rank structure. For every admissible design, the monotone lower bound matches the rank, dimension, and horizon dependence up to logarithmic factors at large horizons, for fixed menu size and model/design constants. For nonmonotone links under a nonzero base-law Stein signal, T-BSTOR combines structured estimation with robust bin-based learning and attains the optimal \widetildeO(T^2/3) horizon rate for fixed dimensions, menu size, and model/design constants. Synthetic and CCLE-based experiments illustrate the benefits of structured estimation relative to vectorized and competing single-index baseline methods.
[LG-305] Saturation-Insensitive Dueling Bandits with General Function Approximation
链接: https://arxiv.org/abs/2609.33011
作者: Chenggong Zhang,Xuheng Li,Qiwei Di,Weitong Zhang,Quanquan Gu
类目: Machine Learning (cs.LG)
*备注: 29 pages
Abstract:We study contextual dueling bandits with general function approximation under the Bradley-Terry-Luce (BTL) preference model. A key challenge in this setting is the saturation of the preference model: when the current reward model can already distinguish two actions with high confidence, the resulting preference feedback becomes weakly informative, making it difficult to further improve reward estimation. Consequently, existing sample-complexity analyses often depend on the inverse-derivative factor 1 / \sigma’[\Delta_r^\ast] which can be prohibitively large when the link function \sigma saturates for large reward gaps \Delta_r^\ast . To address this issue, we introduce SI-CDB, an algorithm that selects opponent arms using a carefully designed heuristic for arm selection. This design enables saturation-insensitive reward learning and recovers the near-optimal dependence for linear reward classes, eliminating the unfavorable 1/\sigma’(\cdot) factor. The core of our analysis is a localized Eluder dimension framework tailored to dueling bandits with general function approximation. Our theoretical results also explain why two-arm regret analysis is crucial for improving single-arm performance in dueling bandits.
[LG-306] What Should Data Teach? Moving Bottlenecks Across Circuit Store and Use
链接: https://arxiv.org/abs/2609.32991
作者: Yixiao Chen,Ke Cheng,Jiangtao Guan,Shuo Huang,Yue Liu,Jun Zhang,Yuhong Liu,Jie Jiang
类目: Machine Learning (cs.LG)
*备注:
Abstract:What should data teach a language model at a particular point in training? A circuit view reveals three distinct bottlenecks: forming a computation, making its required content available, and selecting among available routes. A shared diagnosis-to-data principle connects them: localize the missing operation, preserve its causal relation, vary shortcut-bearing context, and re-audit the residual. Formation-sensitive selection and prerequisite ordering accelerate a binding-matching-transport path; a brief early prefix from the same training multiset retains a validation advantage through 100B tokens. Availability counterfactuals then distinguish writing content from invoking available memory, while paired supervision and context-opportunity ranking improve matched route decisions and long-context answer likelihood. A continuous 350M-model experiment connects the three interventions on the same facts: early circuit training improves subsequent learning, and the complete sequence outperforms stage-replacement controls on facts withheld from Use teaching. Independent query surfaces and opposed-source decisions expose conditional arbitration as the remaining frontier. Together, these results show why a change in the limiting operation calls for a change in supervision, not merely a new ranking of difficult examples.
[LG-307] Feasible Flow Matching for Graph Reconstruction via Within-Sampling Primal-Dual Guidance
链接: https://arxiv.org/abs/2609.32980
作者: Haoming Chen,Nicolas Zilberstein,Santiago Paternain,Santiago Segarra
类目: Machine Learning (cs.LG)
*备注:
Abstract:Graph reconstruction from partial observations often comes with structural side information, such as degree bounds, triangle counts, or an edge-density band. Prior-Informed Flow Matching (PIFM) reconstructs graphs by transporting a local prior toward the graph distribution, but it provides no mechanism to incorporate this side information. We put forth Constrained Primal-Dual PIFM (CPD-PIFM), which augments the sampler with Lagrange multipliers that evolve along each trajectory. The multipliers respond to constraint violations at a predicted endpoint and guide subsequent sampling steps without retraining. We prove that the sampler inherits PIFM’s permutation equivariance and bound its expected terminal slack by a term that decays as the inverse square root of the number of steps, plus two approximation terms. On three link-prediction benchmarks and nine combinations of datasets and constraints, CPD-PIFM raises feasibility by 11-26 percentage points and remains competitive with fixed guidance without selecting a separate multiplier for each constraint.
[LG-308] Adaptive Ensemble Selection for Noisy Labels on Tabular Data
链接: https://arxiv.org/abs/2609.32976
作者: Faizaan Ali,Inwon Kang,Oshani Seneviratne
类目: Machine Learning (cs.LG)
*备注:
Abstract:Incorrect or corrupted labels in tabular datasets can significantly degrade supervised learning performance, particularly when mislabeling is subtle and not easily detectable from feature space alone. In the context of automated or AI-augmented data science workflows, robust detection of such label noise is critical for building reliable models. We propose a data-centric reasoning module for AI data science systems that automatically diagnoses dataset quality and selects appropriate cleaning strategies. Given a dataset, a meta-model predicts weights over a diverse set of detectors, including confidence-based, neighborhood-based, and distributional methods. Across benchmark datasets with controlled noise, our approach achieves performance comparable to a Confident Learning baseline on average, with dataset-dependent gains and losses, particularly in heterogeneous regimes. We further show that detector effectiveness is systematically linked to dataset properties. These results demonstrate the value of descriptor-driven, data-centric ensembling as a component of AI-assisted data-science pipelines for robust dataset assessment and model reliability.
[LG-309] Self-Confirming Superposition Traps in Reinforcement Learning
链接: https://arxiv.org/abs/2609.32966
作者: Dai Shi,Andi Han,Feng Chen,Yiqun Duan,Junbin Gao,José Miguel Hernández-Lobato
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reinforcement learning (RL) trains representations on data selected by the agent’s policy, which then uses the resulting returns to guide its next choices. We show that this loop can sustain a lower-return policy even when representation fitting is globally optimal on those data. In a self-confirming superposition trap, every optimal code assigns overlapping directions to features that rarely occur together under the current policy. An alternative action brings them together, causing interference that lowers its return and reinforces avoidance, although refitting to that action would yield more return at the same capacity. We characterize the dimensions admitting a trap in a tied two-step model and show separately that equal feature frequencies, continued visitation, and independent controller learning need not prevent it. Because fitting weights errors by visitation, an avoided action can lose its return advantage at little cost to the objective. In a finite-action model, we bound this distortion and derive a replay condition: sufficient training weight on the best separately adapted action preserves its ranking despite residual error. Neural PPO experiments show how the feedback develops during learning: agents initialized toward different actions develop different interference patterns, opposite mean return rankings, and different final policies at the same capacity. We therefore test whether retaining access to neglected states can improve control. Keeping these states in training reduces measured interference and improves sequential return, with gains even when the encoder is frozen. Related interventions on state access, replay weights, and feature overlap improve control on MiniGrid and DMControl. For agents that learn through a world model, protected fitting improves DreamerV3–Crafter’s cumulative training scores at unchanged capacity.
[LG-310] Clipped or Unclipped? Finite-Sample Trade-offs for Averag ed SGD under Heavy-Tailed Noise
链接: https://arxiv.org/abs/2609.32962
作者: Alexandra Suvorikova,Egor Gladin,Darina Dvinskikh,Artem Agafonov,Mohammad Alkousa,Yuriy Dorn,Vladislav Matyukhin,Alexander Gasnikov
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Gradient clipping is widely used to stabilize training, but it need not improve the statistical accuracy of averaged SGD, even under heavy-tailed noise. We derive a finite-sample comparison of clipped and unclipped Polyak-Ruppert averaged SGD under finite conditional p -th moments, p\ge2 . Our main result gives explicit accuracy and confidence conditions under which, for p2 , the Gaussian term dominates the unclipped deviation bound, so clipping need not improve its leading order. By balancing clipping bias and concentration, we obtain a bound in which the heavy-tail correction depends logarithmically rather than polynomially on the inverse failure probability. At p=2 , this improves the confidence dependence of the leading bound. We establish sharpness of the unclipped heavy-tail term through an exact one-dimensional quadratic recursion and extend the comparison to projected convex SGD. We also prove concrete costs of clipping: every fixed finite threshold increases asymptotic variance on a scalar Gaussian quadratic, while whole-gradient clipping can shift the limiting point under asymmetric noise.
[LG-311] Constrained Flow Policy Updates: A Generalized Schrödinger Bridge View
链接: https://arxiv.org/abs/2609.32952
作者: Boyang Li,Matthew Kim,Sylvia Herbert
类目: Machine Learning (cs.LG); Robotics (cs.RO)
*备注: 24 pages, 5 figures, 11 tables
Abstract:Online safe reinforcement learning (RL) seeks policies that maximize reward while satisfying safety constraints. Reward and safety can induce multimodal action distributions, challenging the prevailing primal-dual methods: Gaussian actors may collapse onto a single suboptimal mode, and optimization over the nonconvex Lagrangian landscape can be unstable. Diffusion and flow policies can represent such distributions, but recent work with a diffusion actor relies on estimating and matching the score of an augmented-Lagrangian target policy. Instead, we differentiate the augmented objective directly through the generation path of a flow policy, so no score needs to be estimated. Because a flow policy lacks a readily available action log-density for entropy regularization, we build on the density-free kinetic-energy regularizer of FLAC, a recent reward-only method, and propose Reparameterized Augmented-Lagrangian Flow Actor with Least Energy (RAFALE), an off-policy actor-critic method for safe RL. We formulate its update as a constrained one-ended generalized Schrödinger bridge and show that, for each source draw, this path-space problem is exactly an entropy-regularized problem in action space. At positive noise, its solution reweights the reward-only action distribution only where the estimated cost exceeds a threshold set by the Lagrange multiplier. As the noise vanishes, the optimal value converges to that of a least-energy map objective that the flow policy optimizes directly. Across seven Safety-Gymnasium tasks, RAFALE achieves competitive reward with mean final cost within budget on every task, whereas strong baselines trade one for the other; ablations support the necessity of both its augmented objective and its flow actor.
[LG-312] Optimal Nonparametric Dynamic Pricing with Censored Demand and Adversarial Inventory
链接: https://arxiv.org/abs/2609.32949
作者: Mengxiao Zhang,Yingfei Wang,Haipeng Luo
类目: Machine Learning (cs.LG)
*备注:
Abstract:We study online dynamic pricing with censored demand, where an arbitrary inventory level is revealed before pricing and may adapt to past observations, while demand follows an unknown, price-dependent distribution that is stationary over time. For a horizon of T rounds, Xu et al. [2026] achieved \widetilde\mathcalO(\sqrtT) regret under restrictive structural assumptions including linear demand, price-independent additive noise, and conditions relating inventory levels to the noise support. Our first contribution is to extend this framework to a substantially more general and statistically harder nonparametric setting, requiring only the natural assumption that expected sales are nonincreasing in price and allowing nonlinear demand curves and price-dependent noise. For this model, we first propose a simple baseline, Double-Grid-UCB, which discretizes both price and inventory and achieves \widetilde\mathcalO(T^3/4) expected regret using separate revenue estimates for each price-inventory grid pair. Then, we develop Threshold-UCB, which improves the expected regret to \widetilde\mathcalO(T^2/3) . Unlike Double-Grid-UCB, Threshold-UCB reuses sales observations across inventory levels through shared estimates of demand-tail probabilities, allowing the same data to support revenue upper bounds for multiple inventories rather than a single inventory bin. We also complement this upper bound with an \Omega(T^2/3) lower bound via a reduction from stochastic posted pricing, establishing its minimax optimality. Finally, extensive experiments across inventory processes, demand functions, and noise models demonstrate consistently superior performance of Threshold-UCB over benchmark algorithms.
[LG-313] Generative Priors Conditioned on Natural Language for Bayesian Inversion in PDEs
链接: https://arxiv.org/abs/2609.32941
作者: Pengyu Zhang,Mark Girolami,Arnaud Vadeboncoeur
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Inferring quantities of interest (QoI) from data is a central task in Science and Engineering. In such contexts, we often have access to both quantitative data and qualitative data. Quantitative data may be represented by noisy sensor measurements, simulation data, re-analysis data; qualitative data may be in the form of text descriptions of experimental setups, expected experiment outcomes, and human-perceived system behaviours. The task we address in this paper is the following. Given a training set of paired qualitative text and quantitative QoI data, we learn to exploit the inherent correlation between the two modalities to learn a highly informative data-driven natural-language-conditional Bayesian prior, such that when presented with a new physical system, we can coherently combine (i) the training dataset, (ii) qualitative text describing the new system, and (iii) a small number of noisy sensor readings from that new system, to perform inference and uncertainty quantification (UQ) over the QoI. To achieve this task, we develop two parallel approaches, one uses conditional diffusion and the other conditional autoencoders, and compare both against classical Bayesian methodology, unconditional generative models and deterministic supervised methods. Each approach has specific strengths and tradeoffs; conditional autoencoder offers theoretical tractability, allows for fast posterior sampling, and provides better-calibrated UQ, whereas conditional diffusion is explored for greater expressiveness and capturing complex posteriors with irregular QoI fields. The approach is tested on the steady-state heat equation, damped Helmholtz equation, and UK weather reanalysis data.
[LG-314] he Impact of Stochasticity on the Rashomon Effect in Machine Learning
链接: https://arxiv.org/abs/2609.32934
作者: Andrea Apicella,Francesco Isgrò,Andrea Pollastro,Roberto Prevete
类目: Machine Learning (cs.LG)
*备注: Submitted to a journal for peer review
Abstract:Neural network training is inherently stochastic, with factors such as weight initialization leading to distinct models despite comparable predictive performance. This phenomenon is commonly associated with the Rashomon effect, which describes the existence of multiple near-optimal models for the same task. Although the Rashomon effect has received increasing attention, it remains unclear whether different sources of training stochasticity contribute similarly or differently to its manifestations. In this work, we present an empirical study of the Rashomon phenomenon along three complementary dimensions: solution-space multiplicity, predictive multiplicity, and decision-basis multiplicity. These dimensions are quantified through the size of the empirical Rashomon set, predictive ambiguity, and agreement between XAI attribution maps, respectively. By independently controlling three standard sources of stochasticity, namely weight initialization, mini-batch data ordering, and dropout, we isolate their respective contributions to each dimension of the Rashomon phenomenon. Experiments on tabular and image classification benchmarks reveal that these sources affect the three dimensions in different ways. In particular, larger empirical Rashomon sets do not necessarily correspond to greater predictive disagreement or lower explanation agreement, indicating that solution-space, predictive, and decision-basis multiplicity capture complementary rather than interchangeable aspects of the Rashomon effect. Overall, our results show that training stochasticity influences not only predictive performance but also the stability of predictions and explanations, highlighting the importance of identifying the specific sources of stochasticity responsible for different manifestations of the Rashomon phenomenon when assessing the reliability, reproducibility, and interpretability of neural network models.
[LG-315] Last-Iterate Guarantees for Online Reinforcement Learning in Structured Constrained MDPs
链接: https://arxiv.org/abs/2609.32933
作者: Nam Phuong Tran,Trinh Ha Mai Huynh,Tuyen Pham Le,Van-Truong Nguyen,Quan Nguyen,Long Tran-Thanh
类目: Machine Learning (cs.LG)
*备注:
Abstract:In safety-critical applications, deployment uses a single policy, whose performance and constraint satisfaction should hold directly rather than only for an average or mixture of training policies. This motivates last-iterate guarantees in constrained reinforcement learning. Recent progress has established such guarantees in exact-gradient or tabular online settings, yet scalable results for structured large-state problems remain open. We develop a general, statistically efficient framework for last-iterate convergence in structured Constrained MDPs (CMDPs). Our analysis separates contraction of the regularised primal-dual dynamics from actor approximation and statistical errors in policy evaluation under online exploration. This enables model-free on- and off-policy learning with structured function approximation: optimistic policy evaluation avoids explicit transition-model construction, while a compact parametric actor avoids maintaining mixtures or histories of past policies. We instantiate the framework for linear CMDPs and general function approximation, obtaining representation-dependent complexity and improved target-accuracy dependence over prior optimistic regularised primal-dual analyses. We further validate the stabilising effect predicted by our theory on a synthetic linear CMDP: the regularised method exhibits stable last-iterate behaviour, whereas its unregularised counterpart shows larger oscillations.
[LG-316] Counting on Thinking: Tracing Evidence Integration in Language Models
链接: https://arxiv.org/abs/2609.32932
作者: Jingming Xue,Robert C. Wilson,Huadong Xiong
类目: Machine Learning (cs.LG)
*备注:
Abstract:Finite computational resources force a tradeoff between automatic System 1 processes and costly System 2 thinking. Large language models (LLMs) can spend extra computation on hard problems, yet direct answers struggle even with counting, an elementary operation humans and animals perform automatically. We ask why this requires thinking in LLMs. Evidence integration has long been used in psychology and neuroscience to probe decision-making. Our evidence-integration task presents one letter per conversational turn and asks which of two target letters appeared more often. A running count difference solves the task optimally by weighting every letter equally; tokens at each turn could represent and update this difference. Direct responses instead weighted evidence unevenly, with strong recency effects, and assigned less probability to the correct answer as difficulty increased. Thinking improved performance and made integration weights nearly uniform, yet final-query attention remained concentrated on the sequence ends in both modes. Reasoning trajectories showed models revisiting input, recounting letters, and checking intermediate counts that informed the answer, suggesting that thinking constructs the accumulated count that direct responses lack rather than reading out one already formed. Reasoning-token costs grew with the number of letters far more than with coherence. Outcome feedback did not bring this computation into direct responses: under in-context reinforcement learning (ICRL), performance deteriorated over repeated games and recency effects strengthened, yet models grew more confident. Humans and animals amortize such computations into automatic processes, whereas current LLMs still pay for them with thinking on every trial. Which operations learning can make directly available remains central to how future models allocate computation.
[LG-317] Efficient Dynamic Algorithms for Graph Neural Networks with Non-Linear Propagation NEURIPS2026
链接: https://arxiv.org/abs/2609.32929
作者: Kiarash Banihashem,MohammadTaghi Hajiaghayi,Mahdi JafariRaviz,Silvio Lattanzi,Danny Mittal
类目: Machine Learning (cs.LG); Data Structures and Algorithms (cs.DS)
*备注: Accepted at NeurIPS 2026
Abstract:Graph Neural Networks (GNNs) are widely used for representation learning on graphs, but most methods assume static topologies, making them inefficient on evolving networks where edges change over time. Existing dynamic approaches either model graph evolution through temporal GNN architectures without focusing on efficient dynamic maintenance, or are restricted to linear propagation models based on Personalized PageRank. In this work, we study how to efficiently maintain node representations for non-linear GNN propagation under edge insertions and deletions. The propagation has no learned parameters, and only a classifier applied afterward is trained. For a broad class of standard activation functions, we develop a residual-based dynamic algorithm that selectively propagates local errors via push operations, maintaining an approximation to the evolving fixed point without full recomputation. We prove that our method achieves amortized O(1/\epsilon) update time per graph change under a degree-normalized error guarantee. Our approach uses a potential-based analysis in a degree-scaled norm and, in contrast to prior work on the linear case, requires no randomness assumptions on either the update sequence or the input vector. For the linear special case, we additionally provide an exact dynamic algorithm via low-rank matrix inverse updates. Experiments on benchmark datasets show that incorporating non-linearity improves accuracy while preserving efficient update performance, yielding a scalable and theoretically grounded method for maintaining this propagation on dynamic graphs. Comments: Accepted at NeurIPS 2026 Subjects: Machine Learning (cs.LG); Data Structures and Algorithms (cs.DS) Cite as: arXiv:2609.32929 [cs.LG] (or arXiv:2609.32929v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.32929 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-318] Phenomenon-Graph JEPA: Label-Efficient Representation Learning for Contactless Cardiorespiratory Sensing
链接: https://arxiv.org/abs/2609.32928
作者: Constantino Álvarez Casado,Nhi Nguyen,Mohammad Rakibur Rahman,Le Nguyen,Manuel Lage Cañellas,Sasan Sharifipour,Miguel Bordallo López
类目: Machine Learning (cs.LG); Emerging Technologies (cs.ET)
*备注: 9 pages, 4 figures, 7 tables, 24 numbered equations. Extended version of a conference submission, with additional methodological details, implementation settings, and analyzes
Abstract:Millimeter-wave (mmWave) radar and RGB-D cameras can record cardiac and respiratory waveforms continuously and without contact, but labeled recordings remain scarce because every label requires a supervised acquisition session. Self-supervised pretraining can exploit the unlabeled signals, yet contrastive methods depend on signal transformations and negative pairs whose validity is uncertain for cardiorespiratory data, where time warping changes breathing rate and distant windows can share the same physiological state. We present Phenomenon-Graph JEPA, a joint-embedding predictive architecture that learns from four processed one-dimensional streams without negative pairs or synthetic augmentation in its base configuration. Each stream is encoded by a temporal convolutional branch and a band-limited spectral branch. During pretraining, the model predicts stopped target embeddings along typed edges, which connect streams assigned to the same physiological phenomenon, and forward in time within a state episode. We treat this physiological typing as a testable hypothesis and compare it with wrong-edge and all-pairs prediction graphs. In the OMuSense-23 dataset, pretraining improves label-efficiency area over matched supervised training by 3.91 percentage points (95% interval 2.08 to 5.80, Holm-adjusted p = 0.006), and by 3.74 points under a second configuration evaluated on the same test participants. However, the wrong-edge and all-pairs controls do not establish a benefit from physiological typing. Optional Takens-inspired delay coordinates improve a validation comparison with learned history, whereas two wrist-only WESAD protocols do not establish a pretraining advantage. The study therefore separates the measured benefit of predictive representations from the physiological prior used to organize their training.
[LG-319] Adaptive Latent Capacity for World Models
链接: https://arxiv.org/abs/2609.32921
作者: Idan Achituve,Lior Dikstein,Idit Diamant,Arnon Netzer,Hai Victor Habi
类目: Machine Learning (cs.LG)
*备注:
Abstract:We introduce Adaptive LeWorldModel (ALeWM), a world model based on a joint-embedding predictive architecture (JEPA) that learns to concentrate predictive information in compact prefixes of a wide latent representation. To encourage this ordering, ALeWM learns a sequence-conditioned distribution over prefix lengths and trains the predictor to estimate the full next embedding from a sampled input prefix. As standard anti-collapse objectives encourage variation across latent coordinates and do not organize them by predictive importance, we also introduce MixSIGReg. MixSIGReg regularizes the masked embeddings against a prior-weighted mixture with Gaussian active prefixes and zeros in the remaining coordinates. As a result, the ALeWM objective encourages early coordinates to retain information useful for prediction and recursive planning. Our analysis shows that the mixture distribution used by MixSIGReg assigns higher variance to earlier coordinate blocks and lower variance to later ones. In addition, we show that, under specified assumptions, prediction error is minimized by placing the information most useful for prediction in earlier blocks. Empirically, we study the behavior of ALeWM in a controlled dynamical system with known state variables and in goal-conditioned visual control. We show that ALeWM consistently achieves higher mean success rates than tuned fixed-width LeWM, with lower planning capacity on average.
[LG-320] Predicting the Next State Is Not Enough: JEPA Representations for Lean Theorem Proving EMNLP2026
链接: https://arxiv.org/abs/2609.32908
作者: Aarnav Choudhary
类目: Machine Learning (cs.LG)
*备注: Accepted to MathNLP @ EMNLP 2026
Abstract:Neural theorem provers must both propose tactics and decide which valid successor states to explore. We study whether one-step Lean transitions provide a self-supervised signal for branch ordering. A JEPA-style model predicts latent successor representations and scores only kernel-validated, nonterminal successors generated by a fixed pretrained ByT5 proposer. JEPA achieves higher Top-1 than matched InfoNCE on a same-theorem ranking diagnostic(50.18% versus 31.55%), but averages 282.3 of 987 solved theorems across three seeds versus 308 for proposer ordering, while requiring more tactic checks. In this setting, accurate one-step transition ranking is therefore insufficient as a long-horizon search value. The controlled evaluation separates representation from proposal quality and treats kernel-checked proof completion as the primary endpoint.
[LG-321] A Function-Level Vulnerability Score Measures Flag Rate More Than the Model: Protocol Effects on Paired Benchmarks NEURIPS2026
链接: https://arxiv.org/abs/2609.32890
作者: Maciej Cichoń,Bartłomiej Dmitruk
类目: Machine Learning (cs.LG)
*备注: NeurIPS 2026 Workshop on Trust-AI-Eval (TAE): Can We Trust AI Evaluation? 8 pages plus appendix, 5 figures
Abstract:Language models are increasingly evaluated as vulnerability detectors, and scores reported for similar models differ widely between papers. We measured how much of that difference evaluation protocol accounts for, with model outputs held fixed. In a paired test, a model must flag a vulnerable function and clear its version after a fixing commit. Three choices that published evaluations make differently were varied one at a time: metric, verdict extraction and output budget. Seven frontier and large open models were evaluated on five released pair benchmarks and a set pooled for this work under one protocol, and 61 open models of 1.5B to 36B parameters on the pooled set. Function-level F1 follows how often a model flags both functions of a pair (Spearman +0.86 over 42 combinations) and is nearly unrelated to pair-level correctness ( +0.16 ). On the pair score, extraction changes a model’s number by +0.001 at the median and budget by +0.02 with an interval through zero, whereas the model changes a benchmark’s number by up to 0.18 and the benchmark a model’s by up to 0.16; a function-level score therefore measures flag rate more than model. For 37 of 68 models the difference between correct and reversed pairs is within its 95% interval of zero, the value for a null model that flags each function at a fixed rate, while both-flagged and both-cleared rates exceed that null by 0.055 on median, and for 64 of 68 both functions of a pair receive one answer more often than independence predicts: verdicts are determined by the text common to both functions. On length-matched pairs a linear probe on activations separates 0.78 by within-pair ranking, against 0.64 for a tf-idf baseline and 0.5 for length; the generated verdict is near chance for three of six models and at 0.55 to 0.57 for the other three, and a prompted logit is at chance for all six.
[LG-322] Measurement-Gated Provenance Attenuation for Frozen EEG Representations
链接: https://arxiv.org/abs/2609.32889
作者: Anuar Aimoldin,Yankai Chen,Ayana Mussabayeva,Nurdaulet Akhanov,Xue Liu
类目: Machine Learning (cs.LG)
*备注: 23 pages, 4 figures, 11 tables. Code: this https URL
Abstract:Frozen EEG representations retain acquisition signatures as well as neural activity. Source predictability alone does not identify what should be removed: it can reflect measurement effects or genuine biological and population differences, which should not be erased. We propose Measurement-Gated Provenance Attenuation (MGPA), built on one principle: measurement evidence determines where correction may act, and preserved information determines what it should aim for. Paired measurement contrasts define a gate outside which nothing changes; inside it, the source score is moved to the value the preserved coordinates already predict: for a fixed affine score, this keeps the same information as any target set by those coordinates and needs the least expected squared movement. Closed-form and critic-guided iterative constructions apply it without source identity or encoder retraining. Three studies test the principle at increasing distance from its assumptions. Under controlled reference changes, where the source-task association is known, MGPA brings source to near chance with task performance unchanged, whereas erasing what predicts source (LEACE) lowers frozen-task AUROC from .753 to .656 while barely touching source; ablations attribute the attenuation to the gate’s directions and 2.7x less movement to the conditional target. Across recordings from different devices and electrodes, iterative correction lowers source accessibility while preserving or improving task performance. Finally, one iterative map selected on one task and reused unchanged on existing heads for two others raises their worst-association AUROC (lowest over device-label shifts) by .057 and .019 over LEACE, at a cost to those heads while the training association holds. A reusable correction shows its value in how an existing predictor behaves once acquisition cues stop being reliable, not only in what a probe can read.
[LG-323] Neural Network-Assisted Refinement of Traditional Schemes for One-Dimensional Scalar Conservation Laws
链接: https://arxiv.org/abs/2609.32887
作者: Imre Fekete,Ferenc Izsák,Vendel P. Kupás
类目: Numerical Analysis (math.NA); Machine Learning (cs.LG)
*备注:
Abstract:A clear link is established between conventional numerical methods and neural network approximations for solving one-dimensional scalar conservation laws. The focus is on the construction of an appropriate flux term in the case of convex flux functions for improving the classical schemes. The first neural network developed here is able to rediscover Godunov’s method, while the second one emulates the behavior of a second-order slope-limiter function. In this way, by merging them, second-order reconstruction-based schemes can be developed. The networks presented here employ a minimal number of parameters, significantly reducing the complexity compared to previous approaches. These networks can also be linked consecutively to get a deep one corresponding to multiple time steps. Training them with an appropriate loss leads to stable schemes, improving even the classical methods without increasing their complexity.
[LG-324] Staying on the Attractor: Supervising Neural Surrogates of 3D Turbulence Where They Leave It
链接: https://arxiv.org/abs/2609.32864
作者: Yilong Dai,Shaswata Mitra,Raj Patel,Yiming Sun,Shengyu Chen,Jiaqi Gong,Sudip Mittal,Shahram Rahimi,Xiaowei Jia,Runlong Yu
类目: Machine Learning (cs.LG); Fluid Dynamics (physics.flu-dyn)
*备注:
Abstract:Neural surrogates are trained to predict 3D turbulent flows in place of direct numerical simulation (DNS). For chaotic flows, the goal is short-term pointwise accuracy followed by long-term physical and statistical fidelity. However, small prediction errors can carry a surrogate away from the flow’s attractor. Off-attractor states are poorly represented in training data, leaving their evolution weakly constrained. The learned dynamics can then amplify deviations and lead to blow-up, freezing, or statistical drift. We propose off-attractor supervision (OAS) to supervise neural surrogates where they leave the attractor. OAS teaches the model how the true Navier-Stokes dynamics would evolve from these states. Each selected state is paired with its own future computed by DNS. Three generators select a few hundred states for relabeling. The first collects states from the surrogate’s own rollouts. The second uses surrogate attacks to target freezing, excessive amplification, and violations of incompressibility and energy balance. The third perturbs training states along an amplified direction and a strongly damped random direction of the dynamics. All attacks run on the surrogate alone, and DNS relabeling is performed offline once per selected state. Experiments on 128^3 turbulence show that OAS increases the median time to failure from 21 to 721 steps. The compared baselines achieve medians of at most 110 steps, and the advantage holds across training seeds. OAS also achieves the lowest pointwise error at step 15 and the best long-horizon statistics among the compared methods. OAS integrates physical models into neural simulation by extending supervision from fixed reference trajectories to states where the surrogate is likely to fail. This principle can guide the development of more reliable scientific surrogates when deployment takes models beyond the coverage of their training data.
[LG-325] Muon Under Gradient Noise and the Limits of Orthogonalization Near Optima
链接: https://arxiv.org/abs/2609.32861
作者: Xiaohui Xie
类目: Machine Learning (cs.LG)
*备注:
Abstract:Muon replaces the momentum buffer of each weight matrix by its orthogonal polar factor. We ask what this orthogonalization does near an optimum, where minibatch noise dominates the gradient. Under a Gaussian noise model, the expected Muon update becomes a scaled gradient step, so a linear method with a suitably matched learning rate reproduces Muon’s first-order mean response. The stochastic update is a different matter: after the response is matched, Muon retains a nonlinear residual that is uncorrelated with the input noise and contributes additional covariance. A Hermite expansion shows how momentum acts on this residual. Its higher-order components decorrelate faster than the linear component, so momentum suppresses the residual’s accumulated covariance relative to the linear part, but never eliminates it. In a local quadratic surrogate that evaluates the residual on the stationary noise buffer, the residual adds stationary covariance and raises the stationary loss floor at every stable step size, while leaving the contraction dynamics unchanged. Simulations of the full nonlinear recursion on quadratics and measurements on frozen transformer gradients support each step of this picture. Together, the results make a theoretical case for replacing orthogonalization by response-matched momentum SGD once optimization becomes noise-dominated.
[LG-326] HamiFormer: Dual-Expert Diffusion Fields with Affine Symplectic Maps
链接: https://arxiv.org/abs/2609.32838
作者: Haoxiang Huang,Xiang Liu,Shuwei Wang,Jingheng Ma,Sen Cui
类目: Machine Learning (cs.LG)
*备注: 43 pages, 6 figures
Abstract:Predicting smooth dynamics and collisions requires modeling continuous evolution and abrupt state changes. We introduce HamiFormer, a dual-expert diffusion field combining whole-window denoising with residual-corrected Hamiltonian propagation. Their mixed-state feedback attenuates the direct contribution of inherited autoregressive error: each mixed state guides subsequent propagation within the jointly refined window. Parallel Local Affine Scan (PLAS) amortizes iterative refinement across rectified-flow steps and evaluates derivatives in parallel across physical time. PLAS’s affine symplectic maps achieve lower solver error and runtime than sequential explicit Euler in our evaluation. A Regime Model Tree specializes residuals and routing to balance typical-state accuracy against large tail errors. Our analysis gives conditions for physically consistent refinement and warm-start tracking, and finite-window error bounds under diffusion feedback. In 192-step evaluations, HamiFormer reduces normalized phase-space MSE by 26.3% against PhysiFormer on HamiBalls-1 and 21.4% against DiT on HamiBalls-2, with comparable model capacities. Disjoint-interval comparisons show the lowest late-horizon position and momentum errors among baselines on both datasets. Project page: this https URL.
[LG-327] Permutation-Equivariant Flow Matching for Alignment-Free Neural Weight Generation
链接: https://arxiv.org/abs/2609.32833
作者: Arkadi Piven,Yam Eitan,Guy Bar-Shalom,Fabrizio Frasca,Daniel Cremers,Thomas Dagès,Ron Kimmel,Haggai Maron
类目: Machine Learning (cs.LG)
*备注:
Abstract:A trained neural network can be represented by a parameter vector in high dimensions. Learning distributions over these vectors enables the generation of new models across various tasks and architectures. A central challenge is permutation symmetry: permuting hidden neurons can produce distant parameter vectors representing the same function. This introduces variations that a generative model must account for when learning from trained networks. Existing methods typically address this using networks derived from a common base model or costly approximate neuron alignment. We instead parameterize a flow-matching velocity field with a permutation-equivariant Graph Meta Network, enabling direct learning from independently trained networks without alignment. Extensive experiments show that our method closely reproduces the joint statistics of accuracy, functional similarity, and weight similarity of independently trained collections, providing evidence of generation beyond checkpoint memorization. A single conditional model also generates task-specific networks on heterogeneous architectures and generalizes to unseen hidden-width configurations. On a tabular domain-shift task, intermediate conditioning produces individual networks with performance comparable to logit ensembles across both domains. Taken together, our results show how permutation equivariance enables learning from diverse collections of independently trained networks without permutation alignment.
[LG-328] Over-the-Air Federated Learning in Heterogeneous Mobile Wireless Networks
链接: https://arxiv.org/abs/2609.32832
作者: Ming Xiang,Nicolò Michelusi,Yonina C. Eldar,Lili Su
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC); Optimization and Control (math.OC)
*备注: MobiHoc 2026 (the 27th International Symposium on Theory, Algorithmic Foundations, and Protocol Design for Mobile Networks and Mobile Computing)
Abstract:Over-the-air computation has emerged as a scalable and efficient solution for deploying federated learning algorithms in wireless networks by exploiting waveform superposition for simultaneous model aggregation. Most existing work struggles with heterogeneous fading channels. These approaches either enforce unbiased updates from all devices or allow partial device contributions, requiring careful tuning of the convergence bound to mitigate bias under specific fading models. However, the former significantly amplifies receiver noise due to the weakest channel, whereas the latter is sensitive to fading model mismatch and converges only to a biased objective. To tackle these challenges, we propose FedOAG, which employs algorithmic components to automatically satisfy energy constraints via gradient normalization and evenly mix devices’ updates through implicit gossiping. Importantly, FedOAG does not require transmission from all devices, nor does it rely on a specific fading model or knowledge of time-varying statistical channel distributions. We show that FedOAG converges to a stationary point of an unbiased non-convex objective at the best possible rate O(1/\sqrtT) for any stochastic first-order method. We corroborate our analysis with numerical experiments over dynamic wireless conditions on real-world datasets.
[LG-329] UniCache: Task- and Type-Aware KV Cache Compression for Unified Multimodal Models
链接: https://arxiv.org/abs/2609.32831
作者: Wanqi Yang,Yuexiao Ma,Mei Xie,Xiawu Zheng,Shiwei Liu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Unified multimodal models combine understanding, generation, and editing within a single network, offering a promising foundation for versatile multimodal applications. However, growing multimodal contexts make KV cache storage and access increasingly costly. Existing KV cache compression methods are typically tailored to specific tasks and single-modality caches, while overlooking changes in cache importance across tasks and timesteps. However, in unified multimodal models, each task involves multiple KV cache types, and both their composition and dynamics differ across tasks. As a result, a single compression policy overlooks task- and type-specific requirements, leading to the loss of critical information and degraded quality across tasks. Based on these findings, we propose UniCache, a training-free framework for task- and type-aware KV cache compression. UniCache identifies the cache segments activated by each task and assigns suitable compression policies through offline calibration. It coordinates their parallel execution under a shared storage budget through attention-guided allocation and task-aware temporal scheduling. Experiments show that UniCache achieves 5\times KV cache compression for understanding and editing and 2.5\times for generation with negligible quality loss, while increasing throughput by up to 1.78\times in long-context settings, significantly improving the practicality of scaling unified multimodal models to longer context.
[LG-330] Retimed Bellm an Flows: Escaping the Impossible Triangle of Velocity Bootstrapping
链接: https://arxiv.org/abs/2609.32828
作者: Boyang Xu,Shengzhe Chen,Hao Yan
类目: Machine Learning (cs.LG)
*备注:
Abstract:Flow critics learn return distributions by transporting Gaussian noise to Bellman endpoints via continuous velocity fields. While velocity bootstrapping stabilizes training by querying a successor teacher, existing methods face a structural dilemma: on straight paths, no residual-free same-time affine mapping can preserve Gaussian initial noise while maintaining an unbiased target. To overcome this limitation, we introduce Retimed Bellman Flows (ReBF). ReBF queries the teacher critic at a dynamically shifted earlier flow time, aligning intermediate student and teacher trajectories. By combining this retimed clock with fresh, decoupled noise generation, ReBF constructs a provably conditionally unbiased velocity target that preserves the Bellman fixed point and contracts under Wasserstein distances. Empirically, ReBF reduces W_1 distance to ground-truth return distributions by up to 7.7\times on synthetic MRPs and outperforms existing flow critics across 38 challenging OGBench and D4RL offline reinforcement learning tasks.
[LG-331] Learning Shuffle Ideals with Membership Queries and Contrastive Examples
链接: https://arxiv.org/abs/2609.32820
作者: S. Mahmoud Mousawi,Pierluigi San Pietro,Sandra Zilles
类目: Formal Languages and Automata Theory (cs.FL); Machine Learning (cs.LG)
*备注:
Abstract:This paper studies learning of shuffle ideals with membership queries as well as with contrastive queries—a form of membership query that reveals not only whether a selected word w is in the target language or not, but also provides a most similar word w’ that belongs to the target language if %and only if w does not. For both settings, it is shown that even some very simple classes of shuffle ideals cannot be learned efficiently. By contrast, we obtain positive learnability results for classes of shuffle ideals that meet certain structural conditions. In the case of membership queries, these structural conditions are related to the previously studied notion of universal words, and raise new questions in word combinatorics. In the case of contrastive queries, the structural conditions relate to partitioning sets of words that are listed in shortlex order. Subjects: Formal Languages and Automata Theory (cs.FL); Machine Learning (cs.LG) Cite as: arXiv:2609.32820 [cs.FL] (or arXiv:2609.32820v1 [cs.FL] for this version) https://doi.org/10.48550/arXiv.2609.32820 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-332] Change the Product Keep the Parameters: Associative Algebra Layers for Transformers
链接: https://arxiv.org/abs/2609.32814
作者: Ilya Koziev,Ivan Oseledets
类目: Machine Learning (cs.LG); Performance (cs.PF)
*备注: Under review
Abstract:Fast matrix multiplication algorithms keep the product fixed and search for a cheaper way to evaluate it. We instead ask whether a Transformer’s learned projections can use a different, cheaper product altogether. Building on an associative-algebra construction that replaces ordinary matrix multiplication with a sparser interaction table over the same weight blocks, we construct a family with quadratic arithmetic in the matrix dimension when the physical block size remains fixed, and derive finite-shape constraints for GPU execution. The construction is provably optimal for its bilinear rank by the Alder–Strassen bound and can be realized as row-typed rectangular projections compatible with causal masking and KV-cached decoding. We provide an empirical test of this approach by training two approximately 110M-parameter decoder-only Transformer LMs from the same recipe and 12.3B-token budget, differing only in their feed-forward layer: one uses ordinary dense matrix multiplication and the other uses the associative-algebra product. Across four prompt domains, the algebraic model achieves a 6.2–7.8% increase in end-to-end generation throughput, while obtaining lower scores on all three reported downstream metrics. We treat these results as a feasibility and trainability check for the proposed approach at small scale, leaving further investigation to future work.
[LG-333] Derivative-Informed Training of Neural Operators On-the-Fly via Sketched Tangent Consistency
链接: https://arxiv.org/abs/2609.32797
作者: Xinhan Yang,Lu Lu,Shancong Mou
类目: Machine Learning (cs.LG)
*备注: 21 pages, 1 figure, 9 tables
Abstract:Derivative-informed training improves neural operators by directly supervising their input-output sensitivities, which is crucial when neural operators are used as differentiable surrogates for inverse problems, PDE-constrained optimization, design, and control. However, existing methods rely on offline-generated derivative labels, making data generation slow, storage-intensive, and difficult to adapt across datasets, resolutions, or perturbation bases. We propose sketched tangent consistency loss (sTCL), an on-the-fly derivative-informed training objective that, for PDEs with a known and differentiable residual, enforces sensitivity consistency directly from the governing equation without offline tangent labels or neural-operator architecture changes. sTCL uses randomly sketched input perturbations to provide a lightweight derivative-level physics constraint during training. However, raw forward-sensitivity residual penalties can fail for stiff, ill-conditioned, indefinite, or coupled saddle-point tangent operators. To address this, we introduce lightweight operator-aware loss-conditioning mechanisms selected by a simple tangent-operator decision rule. Across Helmholtz, nonlinear diffusion-reaction, Burgers, Allen-Cahn, and Navier-Stokes, with the same neural-operator backbone for all methods, the PDE-specific sTCL losses achieve solution and Jacobian accuracy comparable to offline derivative-informed training (DIFNO) while eliminating the offline derivative-data generation and storage pipeline. These results show that on-the-fly derivative-informed training need not merely amortize offline tangent-solve cost into training; with appropriate sketching and loss design, sTCL provides an effective drop-in path to derivative-informed neural operators. Code is available at this https URL.
[LG-334] FedHV: Low-Overhead Hypervolume Weighting for Federated Multi-Objective Optimization ICLR2027
链接: https://arxiv.org/abs/2609.32790
作者: Amirardalan Dehghanpour,Seyed Mohammad Azimi-Abarghouyi,Christopher G. Brinton
类目: Machine Learning (cs.LG)
*备注: Preprint; submitted to ICLR 2027
Abstract:Task-wise federated multi-objective optimization (FedMOO) trains a shared model for competing prediction objectives under heterogeneous data, partial participation, and communication constraints. Existing methods commonly derive task weights from gradient or update geometry. This requires task-specific information or iterative server-side optimization. We introduce FedHV, which maps reference-relative objective slacks to closed-form inverse-slack weights. Each client optimizes one weighted loss and returns objective estimates with its model update. The protocol adds exactly 2m auxiliary scalars per participating client, yielding Theta(d + m) total per-client communication, compared with the Theta(md) task-specific communication of FSMGDA, and requires no additional synchronization stage. We analyze the resulting one-round-delayed weights under client heterogeneity, multi-step local updates, partial participation, and finite-sample objective reports. Under a fixed-horizon positive-slack reference condition, with the prescribed horizon-dependent step size and vanishing report error, FedHV achieves an O(T^(-1/2)) rate for the average squared log-hypervolume gradient norm; persistent report error determines the resulting stationarity neighborhood. The same bound controls the squared Pareto-stationarity residual. Across six Dirichlet-partitioned non-IID settings from four vision benchmark families and three training seeds, FedHV exceeds FSMGDA and FedCMOO in mean accuracy in five settings. Among these methods and uniform scalarization, it achieves the highest worst-task accuracy in four settings and improves the difficult CIFAR-10 objective in both CIFAR10-MNIST settings.
[LG-335] Copper-Policy: Focus on the Representation for Robust Robot Manipulation
链接: https://arxiv.org/abs/2609.32779
作者: Zexin Feng,Yixu Feng,Lingyu Xiao,Shang Su,Kexin Zheng,Chang Xu,Mengkai Shi,Shuo Feng,Xintao Yan
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: Project page: this https URL
Abstract:World Action Models (WAMs) acquire behavioral priors by modeling future scene evolution, but predicting detailed futures in pixel or latent space incurs substantial cost. Recent evidence that co-training gains persist without test-time generation raises a question: what must a WAM learn to improve control? We introduce Copper-Policy, which learns a compact World representation with the policy rather than relying on a predefined target space. Through temporal joint-embedding prediction, it predicts future observation embeddings conditioned on task intention without reconstructing pixels. This prediction and action decoding shape the representation jointly, while the policy retains access to current-frame spatial detail for execution. Representation analyses show that the learned features better separate task-driven change from perturbations and provide complementary information for control. Compact prediction targets reduce training tokens per sample, enabling a 2B-parameter model trained in 9.67 hours on 8 \times RTX 5090 GPUs and 6 \times faster than Fast-WAM on matched A100 GPUs. Copper-Policy outperforms every compared method without embodied pretraining on RoboTwin and several embodied-pretrained VLAs on LIBERO-Plus (80.85%). On three challenging real-robot tasks, it performs comparably to \pi_0.5 and attains a higher average score. Together, these results show that Copper-Policy combines strong control performance with efficient training.
[LG-336] Robust Bayesian Optimization with Q-Exponential Surrogates ICASSP2027
链接: https://arxiv.org/abs/2609.32775
作者: Richard Cornelius Suwandi,Zhidi Lin,Feng Yin,Abdelhak M. Zoubir
类目: Machine Learning (cs.LG)
*备注: Submitted to ICASSP 2027
Abstract:Bayesian optimization (BO) is a widely used framework for optimizing expensive black-box objectives, but standard BO methods often use Gaussian process (GP) surrogates whose Gaussian assumption is sensitive to outliers and heavy-tailed noise. We introduce q-ED-BO, a robust BO method whose surrogate follows a univariate q-exponential (q-ED) distribution, preserving GP-BO’s closed-form posterior mean and variance while a shape parameter q controls the tail behavior, recovering the GP at q = 2 and growing heavier-tailed with wider confidence bounds as q decreases. This tractability yields a closed-form q-upper confidence bound (q-UCB) with sublinear regret, and an exact closed-form q-expected improvement (q-EI) that generalizes EI to the heavy-tailed predictive, recovering classical EI at q = 2. Experiments on beamformer and adaptive filter tuning with impulsive outliers show that q-ED-BO matches or exceeds existing baselines on clean data, and under corruption, improves the strongest baseline by approximately 0.7 dB in output SINR and 1.1 to 1.2 dB in misalignment reduction.
[LG-337] Beyond Gaussian Assumptions: Distribution-Aware Channel Capacity for Effective Connectivity
链接: https://arxiv.org/abs/2609.32774
作者: Jianan Jian,Jacob Kang,Nurahmed Multezem,Benjamin Li,Nan Xu
类目: Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注: 31 pages, 9 figures, 5 tables
Abstract:Effective-connectivity estimation from brain signals often relies on Gaussian residual modeling, which enables tractable estimation but can discard informative distributional structure and distort inferred directed interactions when empirical residuals are non-Gaussian. We show across multiple modalities, species, and experimental conditions that both brain signals and fitted channel residuals frequently deviate from Gaussianity. We therefore introduce a distribution-aware, information-theoretic measure of effective connectivity based on channel capacity under general residual distributions. To estimate the resulting capacity from empirical, potentially non-Gaussian residuals, we develop a dual-flow min-max estimator based on normalizing flows, in which a generator searches over admissible input distributions under a power constraint while an observer estimates output entropy. We provide a theoretical characterization of the estimator, showing that the observer objective recovers differential entropy up to a KL approximation term, that the formulation reduces to classical Gaussian capacity as a special case, and that residual entropy can alter achievable information rates beyond variance; game-gap and error analyses further characterize optimization and approximation sources. In brain-like simulations with known directed connectivity, Dual-flow achieves the highest AUROC and AUPRC across ten conditions spanning diverse network topologies, hidden drivers, feedback, and heterogeneous hemodynamics, compared with Gaussian capacity, Granger causality, VAR-LiNGAM, and GIMME. Applied to multimodal brain signals, the method reveals time- and condition-resolved directed interactions consistent with known neurobiological circuitry. Together, these results establish a principled distribution-aware framework for effective-connectivity estimation beyond Gaussian residual modeling.
[LG-338] Structuring Relations Among Learning Paradigms via Protocol–Objective–Resource Reductions
链接: https://arxiv.org/abs/2609.32766
作者: Junwei Su,Changjie Wang,Dongyang Chang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Modern machine learning spans supervised, transfer, continual, meta-learning, and related regimes that often reuse the same hypothesis classes, architectures, and optimizers but differ in information access, objectives, memory, adaptation, and sample accounting. This makes it difficult to determine whether one paradigm is genuinely distinct, a special case of another, or part of a broader structural hierarchy. We introduce a protocol-objective-resource (POR) framework that separates representational capacity from these design choices. A paradigm is specified by an environment class, observation protocol, admissible learners, performance functional, and resource accounting rule. POR reductions combine environment embeddings, learner compilers, threshold maps, and calibrated resource overheads. Our main theorem shows that such reductions imply worst-case complexity domination on embedded comparison classes, transferring upper bounds forward and lower bounds backward; under labeled-example accounting, this yields sample complexity domination. We also show that calibrated nontrivial accuracy regimes are necessary to avoid vacuous comparisons, and that strengthening the objective can strictly increase minimax sample complexity even with unchanged protocols and learner classes. Instantiating the framework for supervised, transfer, continual, and meta-learning yields canonical special-case relations: continual contains transfer, transfer contains supervised, and meta-learning contains supervised under aligned raw-example accounting. We further derive a non-exact episode-to-example reduction for episodic meta-learning and capture within-paradigm refinements such as replay memory and task identifiers. The framework thus provides a unified language for structuring learning paradigms and transferring complexity guarantees across them.
[LG-339] AnchorMixGAN: Anchor-Aligned Generative Semi-Supervision for DDoS Detection in Cloud-Integrated IoT Networks
链接: https://arxiv.org/abs/2609.32764
作者: Jin Yang,Xufeng Liu,Yong Hu,Xueyang Wang,Honglu Yang,Gang Li
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注:
Abstract:Detecting distributed denial-of-service (DDoS) attacks in cloud-integrated IoT networks is difficult when labeled traffic is scarce. Generative semi-supervised learning can supplement the available training data, but prediction shifts induced by synthetic views may affect the targets assigned to real unlabeled flows. We propose AnchorMixGAN, a generative semi-supervised framework that addresses this problem through anchor-aligned target construction. Its Anchor-MAS module treats each real unlabeled flow as an anchor and creates alternative views by replacing one field group at a time with values from generated traffic. A frozen reference classifier predicts the anchor and its views; averaging and sharpening these predictions produces a soft target for the original flow. The flow and its target are then mixed with a labeled example using MixUp, allowing the detector to learn from both the original labeled records and the mixed examples. We analyze how reference-classifier error, view construction, and sharpening affect the target, and derive a bound on the resulting change in cross-entropy at a fixed detector prediction. At the reported 90% training setting with 20% of the training records labeled, AnchorMixGAN attains accuracies of 97.3%, 97.4%, and 96.5% on NSLKDD, BoT-IoT, and CICIoT2023, respectively, exceeding the corresponding MixGAN results by 1.6, 1.0, and 4.4 percentage points.
[LG-340] Reuse or Relearn? A Spectral View of Earth Observation Foundation Models
链接: https://arxiv.org/abs/2609.32756
作者: Mehmet Ozgur Turkoglu,Valerio Marsocci,Dominik J. Mühlematter,Dominik Senti,Konrad Schindler,Helge Aasen
类目: Machine Learning (cs.LG)
*备注:
Abstract:Foundation models are rarely used as generic, frozen feature extractors; instead, they are fine-tuned for the target downstream application. This practice is particularly prevalent in Earth observation (EO), and it raises a question that downstream accuracy alone cannot answer: does fine-tuning reuse the pretrained representation, or does it relearn a new one? We study this with spectral diagnostics that compare a model before and after adaptation, quantifying how well its dominant singular subspaces are preserved, how broadly the weight update is distributed, and how large it is. Using natural image models such as CLIP and DINO as a reference, we find that, under the evaluated fine-tuning settings, EO models undergo far larger, higher-rank updates and retain much less of their pretrained structure, so their downstream performance is often obtained with substantial changes to the pretrained weight structure. The diagnostics further provide insight into how cheaply a model can be adapted: where the pretrained subspaces are preserved, adapting a small fraction of the parameters can match full fine-tuning, and where they are not, it can fall behind. More broadly, foundation models, and EO foundation models in particular, should be assessed not only by benchmark accuracy, but also by how reusable their pretrained representation is.
[LG-341] SAGE: Semantic Audio Generative Encoder
链接: https://arxiv.org/abs/2609.32755
作者: Francesco Brigante,Luca Cerovaz,Davide Marincione,Giorgio Strano,Luca Zhou,Emanuele Rodolà,Michele Mancusi
类目: ound (cs.SD); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
*备注: 18 pages, 6 figures, 11 tables. Code and weights: this https URL . Project page: this https URL
Abstract:Audio autoencoders compress waveforms into compact latent representations that serve as the interface between raw audio and downstream models. Current systems navigate a three-way trade-off between reconstruction quality, semantic structure of the latent space, and inference speed, typically favoring one or two of these at the expense of the others. This paper introduces SAGE, Semantic Audio Generative Encoder: a compact variational autoencoder, trained solely on publicly available music, that shapes its latent by distilling embeddings from a pretrained audio-text model. This 105M-parameter model runs at the inference cost of Stable Audio Open and reaches the listening-test quality of SAME-L, an autoencoder 8x larger and 4x slower, while surpassing both on objective perceptual and distributional metrics of reconstruction. Furthermore, it sets the state of the art on all nineteen probing tasks of latent semantics, in domain and out of domain. These results establish SAGE as a lightweight audio autoencoder that strikes the best balance of the three-way trade-off among those we evaluate, combining high reconstruction fidelity, state-of-the-art semantic structure, and fast inference.
[LG-342] Benchmarking EEG Foundation Models at Scale: Lessons from 20000 Evaluations
链接: https://arxiv.org/abs/2609.32743
作者: Zhige Chen,Shu Peng,Chengxuan Qin,Rui Liu,Rui Yang,Kay Chen Tan,Jibin Wu
类目: Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: 91 pages, including appendices and supplementary material
Abstract:Electroencephalography (EEG) foundation models (FMs) promise transferable neural representations, yet their advantages over strong supervised baselines and their prospects for further scaling remain unclear. To address these questions, we introduce EEG-Arena, an open-source benchmark covering 30 EEG FMs and 25 supervised baselines evaluated on 57 downstream tasks from 23 public datasets. Through more than 20,000 evaluations across five experimental protocols, we assess downstream performance, pretraining benefits, model size scaling, pretraining data scaling, and robustness to channel configuration. We find that (1) EEG FMs outperform strong task-specific supervised baselines on most evaluated tasks, particularly under non-bipolar settings; (2) compared with architecture-matched supervised training from scratch, pretraining improves both early optimization and final downstream performance, with larger and more consistent gains as more labeled downstream data become available; (3) existing EEG FMs do not exhibit a consistent positive relationship between parameter count and downstream performance; (4) under a fixed architecture, increasing the pretraining data scale yields sustained downstream gains; and (5) channel-flexible FMs achieve higher absolute performance than channel-constrained models across most evaluated channel configurations. Together, these findings demonstrate the downstream value of EEG FMs and identify pretraining data expansion as a promising direction for further progress. To support continued research, we release EEG-Arena as an open-source evaluation framework that provides shared infrastructure for reproducible benchmarking, model comparison, and community-driven development.
[LG-343] Predicting the Financial Impact of Supply Chain Risk for Major AI-Related Semiconductor Firms: A Heterogeneous Graph Patch Transformer Approach
链接: https://arxiv.org/abs/2609.32741
作者: Jianna Hur,Sagar Samtani
类目: Machine Learning (cs.LG); Computational Engineering, Finance, and Science (cs.CE)
*备注: 15 pages, 3 figures, 3 tables
Abstract:Modern semiconductor production relies on a globally distributed, multi-tier supply chain in which financial stress at one firm spreads with a delay and eventually affects the revenue, inventory, and profitability of the companies that design AI chips. Most firms see only their direct partners, and prior predictive research has mainly targeted market-based risk measures, so few tools forecast how supply chain stress will appear in reported financials. In this study, we propose a heterogeneous graph patch transformer that forecasts these quarterly changes one and two quarters ahead. Learning from a 15,186-company network over 60 quarters, the proposed model fuses quarterly fundamentals with macro-trade, event, and disaster signals through learned gates, carries risk across supplier, customer, ownership, and headquarters relations through typed, direction-specific propagation, and encodes the propagated histories with patch-based tokenization. In preliminary experiments on 116 focal semiconductor firms, the proposed model achieves the lowest error on every target at both horizons, and its profitability advantage widens at the two-quarter horizon. These forecasts can help supply chain managers and investors act before disruptions appear in reported financials.
[LG-344] Plan-to-Synthesis: Cross-City Human Mobility Generation via Semantic Latent Flow Matching
链接: https://arxiv.org/abs/2609.32732
作者: Zhoufu Wang,Baoshen Guo,Zhiqing Hong,Junyi Li,Kailai Sun,Heye Huang,Alok Prakash,Shenhao Wang,Jinhua Zhao
类目: ocial and Information Networks (cs.SI); Machine Learning (cs.LG)
*备注:
Abstract:Human mobility generation aims to synthesize realistic point-of-interest (POI) visitation trajectories and has become an important tool for travel behavior modeling, transportation management, and urban planning. Existing diffusion-based methods achieve high fidelity but require per-city generation, given the inherent heterogeneity of geospatial locations and POI categories, while large language model-based methods generalize across cities but remain too costly at scale, especially for long-horizon trajectory generation. To address this, we propose SeMoFlow, a Semantic human Mobility generation framework based on latent Flow matching. We first encode heterogeneous POIs from different cities into a shared cross-city representation space via hierarchical Semantic IDs, where shared prefixes capture transferable semantics, and successive codes progressively refine the representation toward individual POIs. Building on the semantic IDs, SeMoFlow follows a plan-to-synthesis hierarchical generation paradigm, in which an autoregressive planner generates coarse-grained semantic and recurrence patterns, and a flow matching realizer synthesizes fine-grained suffix latents. The generated latents are subsequently decoded and grounded to concrete POIs. Extensive experiments on large-scale multi-city datasets show that SeMoFlow achieves higher trajectory fidelity than existing baselines, preserves city-specific mobility motifs, and supports both joint multi-city generation and effective cross-city transfer.
[LG-345] Distributionally Robust Averag e-Reward Reinforcement Learning: Finite-Sample Guarantees under Weak Communication
链接: https://arxiv.org/abs/2609.32727
作者: Chenyu Lu,Zijun Chen,Nian Si
类目: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
*备注:
Abstract:We study distributionally robust reinforcement learning (DR-RL) in the average-reward setting under weak communication. Our main result provides finite-sample guarantees for estimating the robust optimal average reward and learning a near-optimal policy, covering both SA-rectangular and S-rectangular structures with divergence-based and distance-based uncertainty sets. Specifically, for Kullback–Leibler and f_k -divergence balls, we establish explicit radius conditions under which the robust average-reward Bellman equation admits a constant-gain solution, while for total variation and Wasserstein balls, any positive radius suffices without requiring the nominal MDP to be weakly communicating. Our algorithm is prior-knowledge-free and achieves sample complexities of \widetilde O(|\mathcalS||\mathcalA|p_\wedge^-1\operatornameSpan^2(u_\delta^\ast)\epsilon^-2) for estimating the robust optimal average reward and \widetilde O(|\mathcalS||\mathcalA|p_\wedge^-2\operatornameSpan^2(u_\delta^\ast)\epsilon^-2) for learning an \epsilon -optimal policy. Here, p_\wedge is the smallest positive nominal transition probability and u_\delta^\ast is a robust optimal bias function. We further provide an almost-tight explicit upper bound on \operatornameSpan(u_\delta^\ast) . Finally, we validate the predicted n^-1/2 convergence rate through numerical experiments.
[LG-346] Scaling Properties of Same-Family On-Policy Distillation
链接: https://arxiv.org/abs/2609.32722
作者: Yuntai Bao,Qinfeng Li,Guoqing Jiang,Liwei Chen,Zhiheng Qin,Xuanping Li,Wenqi Zhang,Xuhong Zhang
类目: Machine Learning (cs.LG)
*备注: 35 pages, 20 figures
Abstract:Reinforcement learning (RL) can induce substantial reasoning capabilities in large language models (LLMs), but how much of this capability transfers across model scales, and how quickly, remains unclear. We study the scaling properties of on-policy distillation (OPD) across weak-to-strong, same-base, and strong-to-weak teacher–student setups. We find that early OPD training dynamics uniformly exhibit a regular useful-transfer regime, in which held-out accuracy (the gold score, G ) rises approximately linearly in d=\sqrt\mathrmKL(\pi_\theta \Vert \pi_\mathrmref) , the square root of token-level reverse KL divergence from the student initialization. In every observed weak-to-strong pair, the student’s peak gold score exceeds its teacher’s own, so a compact RL expert can transfer capability to a much larger student via OPD. To estimate OPD outcomes, we fit power laws for how G_\mathrmpeak and the slope of the useful-transfer regime scale with student and teacher parameter counts and with teacher gold score. These laws show that peak gold score improves with teacher scale only up to roughly the student’s scale, and that at a matched gold score smaller teachers transfer better, so a teacher’s score alone does not define its supervision value. We also study the scaling effects of two OPD variants, bootstrapping weak-to-strong OPD, and the degree of on-policy supervision.
[LG-347] BiasReducer: Adaptive Bias Mitigation for Reward Models
链接: https://arxiv.org/abs/2609.32720
作者: Shuang Liu,Yongliang Miao,Yanguang Liu,Haoyi Xiong,Mengnan Du
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reward models score responses from large language models (LLMs) and guide LLM training toward human preferences. However, reward models can favor superficial attributes such as length or confidence, leading LLMs to produce higher-scoring but not more correct responses. Existing mitigation methods either retrain the reward model or apply a fixed correction to one known bias, such as a preference for longer responses. Retraining requires additional data and computational resources, while existing editing methods require the target bias to be specified in advance and use a fixed edit for that bias. To this end, we propose BiasReducer, a lightweight framework that edits only the linear reward head and selects the relevant edits for each new dataset. First, BiasReducer uses a sparse autoencoder (SAE)-style encoder to learn which attributes (e.g., length and confidence) the reward model is sensitive to. Second, it learns how to reduce the reward model’s dependence on each attribute by determining which direction to adjust the reward head and how much to adjust it. Third, for a new dataset, it ranks the attributes by their influence on reward scores, selects the relevant ones, and edits the reward model accordingly. BiasReducer consistently improves reward-model robustness to biases toward superficial response attributes. Across five reward models, BiasReducer-M improves the three benchmarks by 8.3, 18.0, and 6.9 percentage points on average, outperforming the two training-based baselines. The gains transfer downstream, reducing unnecessary verbosity and sycophancy while maintaining comparable judged quality.
[LG-348] One-Step Generative Modeling via Unbalanced Optimal Transport
链接: https://arxiv.org/abs/2609.32708
作者: Yirong Shen,Mengfei Xia,Junpeng Jing,Lu Gan,Cong Ling
类目: Information Theory (cs.IT); Machine Learning (cs.LG)
*备注:
Abstract:Drifting models enable one-step generation by amortizing distribution transport into training, but this efficiency places greater demands on the transport field estimated at each update. In large-scale training, the field is computed from finite mini-batches of generated and real samples, which provide only imperfect approximations to the underlying distributions. Balanced optimal transport enforces exact mass matching within every mini-batch, making the estimated field sensitive to the particular composition of the real-data batch. We find that generated and real samples should be treated asymmetrically: letting the mass assigned to real samples adapt while keeping every generated sample fully transported improves generation across six feature-space metrics in controlled ablations, and is more robust to the relaxation strength than relaxing both marginals simultaneously, which falls below balanced transport under stronger relaxation. Motivated by this observation, we propose Unbalanced Optimal Transport Gradient Flow (UOT-GF), which keeps the generated-sample marginal fixed and relaxes only the real-data marginal. Under identical settings at DiT-B/2 on ImageNet-256, UOT-GF improves Fréchet Inception Distance (FID) from 1.53 to 1.46 over the balanced W-Flow baseline; scaling the same recipe yields 1.34 and 1.22 FID at L/2 and XL/2, the best FID among the one-step models we compare. We further derive the induced UOT transport force, establish a kinetic Vlasov–Fokker–Planck formulation whose overdamped zero-temperature limit recovers the drifting dynamics, and characterize non-target stationary states together with sufficient conditions for convergence.
[LG-349] Silent Failures in Agent ic Security Evaluation: A Validated Harness for Tool-Call Mediation Under Indirect Prompt Injection
链接: https://arxiv.org/abs/2609.32691
作者: Animesh Shaw
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 7 pages, 4 figures, 3 tables
Abstract:LLM agents that invoke privileged tools are vulnerable to indirect prompt injection (IPI), in which adversarial instructions embedded in retrieved data hijack the agent’s actions. A growing body of work evaluates defenses against IPI, but the validity of that evaluation is rarely examined. We audit an IPI benchmark and its harness and identify four defect classes – silent payload non-delivery, attack success scored by tool identity rather than arguments, false-rejection rate conflated with model incapacity, and the absence of an audit trail – each of which yields a plausible, publishable, and incorrect number. We quantify the distortion by re-scoring identical execution traces under the defective and corrected definitions: on real agent behaviour, the tool-identity scorer reports a 21.7% attack-success rate where the true argument-level rate is 1.2%. In the sharpest case, an open model previously reported at 62.8% registers 0% under the corrected harness – the prior figure largely an artifact of undelivered payloads and identity-level scoring. We release a harness whose construction makes each defect unrepresentable – machine-checkable payload placement, argument-level attacker predicates, per-scenario environments, and mandatory trace persistence – and use it to report three quantities the field does not: whether a compromised agent discloses the attack, the full security/utility operating curve of an LLM-judge defense, and tool-calling capability disentangled from defensive over-blocking. A corrected harness further overturns a reported “capability barrier”: a model deemed incapable of tool use is in fact fully capable, its earlier result an artifact of environment mismatch. We argue that evaluation validity is a prerequisite for, not a footnote to, defense claims in agentic security, and provide an instrument that enforces it.
[LG-350] Self-Evolving Time-Series Forecasting Agents with Episodic Memory and Online Policy Learning
链接: https://arxiv.org/abs/2609.32689
作者: Junyi Wang,Yilin Wang,Wen Wu,Chao Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:LLM-based agents are increasingly used for time-series forecasting because they can organise contextual information, perform multi-step analysis, and guide the sequence of actions required to complete forecasting tasks. Most existing agents focus only on the current forecasting instance. However, in real-world deployments, forecasting commonly operates online, with new forecasts issued from the currently available history as the forecast origin advances and the ground-truth targets of earlier instances progressively become available. These targets provide feedback on the actions taken in earlier instances, yet existing agents generally do not preserve or utilise this information to adapt their subsequent actions. To address this limitation, we introduce FASE, a Feedback-Aware Self-Evolving forecasting agent that converts such feedback into task-specific experience for subsequent forecasting instances. FASE combines episodic memory, which retrieves relevant completed instances, with online policy learning, which summarises the feedback accumulated across instances into ranking guidance. The proposed framework is evaluated on 29 dataset configurations selected from the GIFT-Eval benchmark. Across these 29 configurations, FASE attains the strongest aggregate point forecasting performance among the evaluated methods and reduces the normalised MAE by 9.1% relative to the best individual foundation model baseline. The results further indicate that the cumulative advantage of FASE increases as delayed feedback accumulates. Together, these findings demonstrate that FASE can continually self-evolve through feedback from completed forecasting instances without updating the parameters of the LLM.
[LG-351] SIFT: Enhancing Time Series Foundation Models via Semantic Invariance and Structural Fidelity Fine-Tuning
链接: https://arxiv.org/abs/2609.32676
作者: Yi Tang,Tengxue Zhang,Yang Shu,Chenjuan Guo,Chenchen Sun,Yisheng An
类目: Machine Learning (cs.LG)
*备注:
Abstract:Time Series Foundation Models (TSFMs) have achieved remarkable zero-shot performance through extensive pre-training on massive time series datasets. Nevertheless, due to the low-dimensional properties and diverse structural patterns of time series data, performing naive fine-tuning on TSFMs often leads to overfitting and falling into the mean-prediction trap. To address these challenges, we propose SIFT, a robust adaptation method that enhances time series foundation models by preserving Semantic Invariance and structural Fidelity throughout the fine-Tuning process. We employ semantic-invariant adversarial augmentation, which utilizes semantic spectrum decomposition to partition the semantic space and then generates perturbations within the non-core semantic subspace to bolster the model’s robustness against these perturbations, mitigating overfitting. We implement a component-based structural fidelity enhancement, which facilitates component-wise mixup and imposes a reconstruction objective to improve the model’s ability to preserve structural fidelity, alleviating the mean-prediction trap. Extensive experiments on representative TSFMs covering 10 real-world datasets demonstrate that SIFT can significantly enhance the performance of TSFMs.
[LG-352] Equivariant Neural Primal-Dual Assignment for Maximum Common Edge Subgraphs
链接: https://arxiv.org/abs/2609.32661
作者: Jiaqing Xie,Yanchao Li,Zhuo Yang,Yuxin Wang,Tianfan Fu,Yuqiang Li
类目: Machine Learning (cs.LG)
*备注: 35 pages, 11 figures
Abstract:Maximum common edge subgraph (MCES) matching finds a partial vertex correspondence between two labeled graphs that preserves as many labeled edges as possible. Molecular similarity search requires matching many graph pairs, making the cost of repeated queries important. The strongest baseline attains accurate MCES solutions but trains a separate network for each pair. We introduce Equivariant Neural Primal-Dual Assignment (ENPDA), which learns a shared matching policy and applies it to new pairs without further training, answering queries roughly three orders of magnitude faster and recovering its training cost after a few dozen queries. The policy recomputes exact objective marginals for candidate matches and learns corrections and step sizes that update their scores. Target prices respond to competition when several source vertices favor the same target. Four update rounds and a Hungarian projection produce a partial one-to-one matching. We prove per-pair guarantees that hold for any network parameters. In exact arithmetic, reordering either graph permutes the assignment and price states, the projected matching is one-to-one, and repaired prices give a valid MCES upper bound. Subtracting the preserved-edge count bounds the optimality gap; combined with structural caps, these certificates prove global optimality for 60 of 291 native test pairs. On three molecular benchmarks with disjoint train/validation/test splits, ENPDA improves over an analytic counterpart with the same update and projection budget by 7.4-8.6 accuracy points; after one second of refinement search, 2.5-3 points of the gain remain. Transferred without fine-tuning to edge-deletion tasks from social and protein graphs, the policy gains 9.1-17.6 points over the analytic counterpart. When output matchings must keep aromatic rings intact, ENPDA recovers more reference bonds than the baselines on all three datasets.
[LG-353] Quantization-Aware Pre-Training with Constrained Empirical Weight Distribution
链接: https://arxiv.org/abs/2609.32659
作者: Ningfeng Yang,Tor M. Aamodt
类目: Machine Learning (cs.LG)
*备注: Preprint
Abstract:Quantization-Aware Pre-Training (QAPT) can increase the inference efficiency of DNNs, but a problematic behaviour known as rounding boundary weight oscillation can introduce detrimental noise into the training process and significantly reduce convergence speed. While existing methods can reduce this detrimental noise, they either introduce additional hyperparameters or memory overhead, or cannot consistently improve model accuracy. In this work, we propose optimization with \textbfC onstrained \textbfE mpirical \textbfW eight dis \textbfT ribution (CEWT), the first hyperparameter-free memory-overhead-free oscillation suppression method that consistently improves QAPT performance: an optimizer post-update step that projects weights to the nearest point in weight space whose empirical distribution (histogram) matches a zero-mean Gaussian. Our key insight is many quantizers are designed with the implicit assumption that the to-be-quantized data are permutations of samples from a zero-mean Gaussian, and this assumption is not true during QAPT. By enforcing the zero-mean Gaussian prior as a hard constraint, CEWT can suppress this detrimental noise. Empirical results on various combinations of SOTA quantizers and hypersphere optimizers suggest, that with a geomean increase of 4% in training time, CEWT can consistently reduce the pre-training perplexity (by an average of 2.5 and up to 21 points) of low-precision (down to 1-bit activations and weights and up to 610M parameters) LLaMA/GPT models without introducing any hyperparameters or storage overhead. Code is available at this https URL
[LG-354] Extremely Fast and Compact Binary Graph Representations via Randomized Operator Sketching
链接: https://arxiv.org/abs/2609.32641
作者: Srajan Agarwal,Megha P,Bikas C Das,Zakaria Laskar,Saptarshi Bej
类目: Machine Learning (cs.LG)
*备注:
Abstract:Graph neural networks typically rely on dense, floating-point node representations, which can impose substantial memory and computational costs. Binary graph hashing offers an alternative by encoding node information as compact bit strings. However, existing approaches either sacrifice global topological information for computational efficiency or incur substantial generation costs. We introduce an ultra-fast, entirely algebraic hashing method that constructs binary node representations directly from graph structure, without requiring node features or gradient-based training. Our method approximates a high-order structural transition matrix using randomized column sampling inspired by the Nyström method and combines it with an efficient label-safe semantic propagation mechanism. The resulting continuous representations are discretized through column-wise thresholding to obtain compact binary codes. Experiments on ten node classification datasets show that the proposed method consistently improves classification accuracy over existing feature-free binary baselines while requiring sub-second code generation on many datasets. The resulting binary representations are also naturally suited to event-driven computation, making them compatible with neuromorphic spiking neural networks and gradient-free learning rules. These results demonstrate that simple algebraic approximations can provide an efficient alternative to learned pipelines for discrete graph representation learning.
[LG-355] Schur-Neural KF: Learned Schur-Consistent Corrections to the Extended Kalman Filter
链接: https://arxiv.org/abs/2609.32640
作者: Min Kim,Lianghao Cao,Soon-Jo Chung,Andrew M. Stuart
类目: ystems and Control (eess.SY); Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: 8 pages. Accepted to the 65th IEEE Conference on Decision and Control (CDC 2026)
Abstract:We present Schur-Neural KF (SN-KF), a learning-based correction to the extended Kalman filter (EKF) that preserves the probabilistic conditioning interpretation of the EKF. The method perturbs the predictive state-measurement cross-covariance and the Cholesky factor of the measurement noise covariance so that the resulting joint predictive covariance is always positive semidefinite. The positive semidefiniteness is ensured by a Schur complement-based parametrization. We instantiate the parametrization with a recurrent neural architecture whose matrix outputs are modulated by amplitude gates. We prove that incorporating a measurement does not increase the filter’s state uncertainty, and show that no measurement can induce an arbitrarily large state correction relative to its statistical surprise. We also present a perturbative analysis suggesting SN-KF’s structural strength in the data-scarce regime. We provide two numerical experiments to illustrate the practical benefits of SN-KF. In a two-radar experiment, enforcing Schur-consistency provides a much broader failure-free hyperparameter region and reduces RMSE for small training subsets, consistent with our theoretical analysis in the data-scarce regime. In the unicycle experiment, SN-KF achieves the best precision, recall, false alarm rate, and gated RMSE under innovation-based sensor-fault rejection.
[LG-356] Stabilizing the Dynamic Low-Rank Training NEURIPS2026
链接: https://arxiv.org/abs/2609.32615
作者: Zhonghan Xu,Ling Wang,Junhao Chen,Jianwei Zhao,Jinwei Yang
类目: Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2026
Abstract:Training neural networks directly in a low-rank parameterization is an appealing route to reducing memory, compute, and storage simultaneously during both training and inference. Dynamic low-rank training (DLRT), which confines weights to a rank- r manifold via the Galerkin projection of the gradient flow, is particularly attractive because it identifies efficient subnetworks on the fly without specialized initialization or post-factorization. However, DLRT fails to find trainable networks under high compression. In this paper, we derive the gradient flow of the best rank- r approximation and point out that the offset of DLRT comes from a curvature-coupling term which is large and thus non-negligible under aggressive compression. Guided by this analysis, we propose a stable dynamic low-rank training method, named SDLRT, which maintains a lightweight compensation buffer that reinjects the top neglected singular directions. Additionally, we introduce a negative feedback on the truncation tolerance to stabilize each layer’s rank. Experimentally, SDLRT reliably finds trainable subnetworks where DLRT collapses and as a PEFT adapter on DeBERTa-v3, it achieves the best average score on SuperGLUE at only 2.8% parameter overhead over LoRA.
[LG-357] Learning the Graph and the Embedding Together: Classifier-Independent Rewiring for Heterophilic Node Classification
链接: https://arxiv.org/abs/2609.32613
作者: Harshit Kumar,Sujan Chakraborty,Priyanka Saha,Pritam Kar,Saptarshi Bej
类目: Machine Learning (cs.LG)
*备注:
Abstract:Graph neural networks lose much of their advantage on heterophilic graphs, where connected nodes often carry different labels. Graph rewiring is a popular remedy, but rewiring methods are usually evaluated with a single classifier, which makes it hard to tell whether the gains come from the new topology or from that particular pairing. We propose an affinity-guided rewiring method that estimates the graph and the node representation together. It alternates, in the spirit of expectation maximisation, between training a lightweight graph neural network on the current graph and re-weighting candidate edges under a modularity objective with a pseudo-label homophily term. Candidate edges come from a compact pool scored by a contrastively learned node similarity and a neighbourhood-distribution affinity. The method returns two classifier-independent outputs: a rewired graph and a node embedding learned on it. Across six heterophilic benchmarks and five downstream classifiers, it improves accuracy over the original graph with normalised features in 23 of 30 classifier-dataset combinations, with a mean gain of 5.8 points, and reduces the accuracy spread between classifiers about fourfold. A controlled ablation shows that the two outputs are each useful and play complementary roles: the embedding contributes most of the accuracy gain, while the rewired graph makes different classifiers agree. A fully unsupervised variant, which uses no labels during rewiring, retains most of the improvement. The rewired graphs are also more homophilic and improve label propagation and community detection.
[LG-358] Age of Learning: Temporal Persistence of Prediction Errors as a Learning Signal
链接: https://arxiv.org/abs/2609.32593
作者: Chenyang Wang,Stefan Forsström,Roger Olsson,Di Yuan,Qing He
类目: Machine Learning (cs.LG)
*备注:
Abstract:Current machine learning algorithms primarily rely on instantaneous signals such as loss, margin, and prediction confidence to characterize model behavior. These signals indicate how difficult a prediction is at the current optimization step, but they do not capture how long the model has remained incorrect. We study this temporal dimension of learning and introduce Age of Learning (AoL), a learning-state variable that measures the persistence of prediction errors over time. AoL increases while an error remains unresolved and resets when a correct prediction is achieved, thereby distinguishing persistent under-learning from transient mistakes. We develop AoL-based training strategies for both offline and streaming settings. In offline learning, sample-level AoL is accumulated over training and aggregated into class-level states that guide adaptive reweighting and resampling. In streaming learning, where full historical access is unavailable, we maintain lightweight class-level AoL states using current and buffered observations. Across long-tailed classification settings, AoL improves or matches standard training baselines, with larger benefits when learning difficulty persists over time. Multi-seed streaming experiments further show reproducible gains under temporally stable imbalance. Analysis of class frequency, loss, and margin shows that AoL is related to conventional difficulty measures but captures additional information about error duration. These results suggest that temporal persistence provides a useful complementary signal for characterizing and controlling learning dynamics in imbalanced and non-stationary environments.
[LG-359] hink Fast Plan Selectively: Adaptive Deliberation for Efficient Data-Driven MPC
链接: https://arxiv.org/abs/2609.32591
作者: Yi Xian Goh,Sze Jue Yang,Hao Luan
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注:
Abstract:Data-driven model predictive control (MPC) combines learned world models with online trajectory optimization, achieving strong performance in continuous control. However, the per-step cost of sampling and evaluating hundreds of candidate trajectories restricts deployment to control frequencies well below what real-time robotics demands. Motivated by the dual-process theory of human cognition, which distinguishes between fast, intuitive processing (System 1) and slower, deliberative reasoning (System 2), we ask whether every decision requires the same degree of computational deliberation. We propose Fast-TD-MPC, a lightweight framework that adaptively routes between fast policy execution and test-time planning, reserving costly deliberation for states where it is most needed. Fast-TD-MPC delivers competitive task performance across 103 continuous control tasks while achieving up to ~4x faster inference. Under external disturbances, Fast-TD-MPC selectively falls back to planning, maintaining robustness comparable to the original planner.
[LG-360] DimPO: Dimensionality Reduction for Attention using Preference Optimization
链接: https://arxiv.org/abs/2609.32579
作者: Vojtěch Lanz,Yufei Cui,Prasanna Parthasarathi
类目: Machine Learning (cs.LG)
*备注:
Abstract:A linear projection can reduce the dimension of query and key vectors without updating the pretrained model, but it remains unclear which training objective best preserves model behavior. We ask whether preferences over keys and attention mass on the highest-weighted keys provide a better signal than matching the full attention distribution with KL divergence, especially in long-context settings. We introduce DimPO, which combines listwise preference optimization with a lightweight top-k cross-entropy term for head-fidelity. DimPO is trained offline from the attention patterns of a frozen language model, with one map per layer, shared by the query and the keys and trained separately from the other layers. Across LLaMA3.2-3B, LLaMA3.1-8B, Qwen2.5-7B, and Qwen3-4B Instruct models, pairwise preference objectives outperform the triplet baseline and retain 98% of the original score on short-context tasks when projecting to half the dimension on the last 40% of the layers. With more projected layers or on long-context RULER, they degrade rapidly. In contrast, KL and DimPO, which use every key during training, retain about 95% of the original RULER 4k score on the 8B model when projecting up to 50% of the layers. KL-based projections remain closer to the original attention distribution and attention output, yet DimPO achieves better downstream performance. Beyond 50% of projected layers, DimPO increasingly outperforms KL on tasks including SQuAD, common-word extraction, frequent-word extraction, and variable tracking. These results suggest that under dimensionality reduction, preserving the ordering and concentration of task-relevant attention can matter more than reproducing the full attention distribution.
[LG-361] rapped by Their Own Rollouts: Understanding Aggregation–Rollout Feedback in Federated On-Policy Distillation
链接: https://arxiv.org/abs/2609.32573
作者: Jinqian Chen,Jihua Zhu,Chang Liu
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注:
Abstract:On-policy distillation (OPD) is a promising approach to language-model adaptation, aligning teacher supervision with the student’s own generated trajectories. When adaptation prompts are distributed across clients, can this process benefit from federated collaboration? We study federated OPD and find that substantial collaboration gains can be obscured by learning-rate sensitivity: FedAvg can perform no better than independent local training at a small learning rate, yet recover a clear advantage at a larger rate. We explain this phenomenon through the student’s dual role as learner and generator of future training data. An aggregation-induced optimization lag can delay access to useful teacher supervision, which in turn slows subsequent learning. Our theory establishes this aggregation–rollout feedback in a solvable model with a common optimum and stable updates, and identifies two coupled roles of learning rate: learning from current supervision and reaching future supervision. Guided by this analysis, we propose FedTOPS (Federated Teacher-guided On-Policy Scaling), which reuses teacher feedback on current trajectories to adapt the FedAvg update magnitude under clientwise predictive-change constraints. Across six mathematical reasoning benchmarks, FedTOPS improves macro Avg@8 over FedAvg by 4.56–14.57 percentage points across the evaluated student models and local learning rates.
[LG-362] CAESAR: Clustering via Autonomous Embedding-Space Agglomerative Reorganization
链接: https://arxiv.org/abs/2609.32570
作者: Ilan Bacry,Rémi Devaux,Antoine Jardin
类目: Machine Learning (cs.LG)
*备注: 14 pages, 3 figures
Abstract:Clustering algorithms that operate on nearest-neighbor graphs, such as FINCH (First Integer Neighbor Clustering Hierarchy), depend heavily on the quality of the embedding space they are given. However, pretrained vision and language model embeddings are not optimized for this purpose. We propose CAESAR, a method that reorganizes a pretrained embedding space: a reorganization network is trained to pull mutual nearest neighbors together and push non-neighbors apart, yielding a reorganized embedding space substantially better suited to clustering. CAESAR offers a second major advantage: it never requires the number of clusters K . This matters because in realistic unsupervised settings, K is typically unknown and discovering it is often part of the problem, yet most strong clustering methods take it as an input. We therefore design the entire CAESAR pipeline to infer K rather than assume it is known. Empirically, reorganizing the embeddings consistently improves clustering over the raw space on both text and image datasets. Since the few deep clustering methods that also infer K do not release their code, we complement controlled comparisons with methods that infer K on the same embedding space by comparisons with strong deep clustering methods that are given the true K , giving them a substantial oracle advantage. Even so, CAESAR outperforms all of them on text, achieves the best results on the most challenging image benchmark and remains competitive on the others. Reorganizing pretrained embeddings thus emerges as a simple and powerful route to clustering realistic data, where classes overlap and the number of clusters is unknown.
[LG-363] Compositional Objectives: Learning Structure in Structure
链接: https://arxiv.org/abs/2609.32566
作者: Pranavchandra Vivekananda,Sumukh Bettadapura,Ajan Subramanian
类目: Machine Learning (cs.LG)
*备注:
Abstract:Intelligence is defined in many ways. One of these definitions defines intelligence as the pursuit of learnable novelty. However, learnable novelty can be meaningless without the ability to compose the learned structures to take action and achieve goals. Learnable novelty builds on epiplexity, which is a way to measure learnable structure in data through a bounded observer. In this paper, we investigate a closed-form spectral approximation to compute epiplexity. We use a fixed-trace constraint and find that the epiplexity objective prefers a more uniform distribution of spectral mass rather than concentrating it in a small number of directions. However, a representation may spread information across many directions without organizing that information into features useful for a particular task. To address this gap, we propose a compositional objective whose observer measures the relationships between the parts and interactions of an image. We compare it with the original spectral objective given only the masked parts. In our ImageNet training runs, the spectral objective with masked parts produces an almost maximally spread representation while achieving the strongest frozen-feature classification performance on most evaluations, more than doubling the linear-probe accuracy of the whole-image baseline. Across multiple image benchmarks, changing what the observer sees matters more than adding relation and interaction tokens. At the same time, our prediction-oriented compositional objective produces substantially better held-out observer prediction but relatively weaker classification, revealing that spectral diversity, predictability, and downstream utility are distinct properties. These results suggest that the usefulness of spectral spreading depends not only on how much structure is preserved, but on which relationships the observer makes available to the objective.
[LG-364] Prioritizing Repeated LLM Evaluation for Hidden Failure Discovery
链接: https://arxiv.org/abs/2609.32547
作者: Keita Broadwater,Akin Broadwater
类目: Machine Learning (cs.LG)
*备注: 22 pages, 3 figures
Abstract:Large language models are commonly evaluated by generating a small number of stochastic responses for each prompt in a benchmark. Because inference budgets are limited, this shallow evaluation may fail to observe low-probability but operationally important failures. A prompt that produces no failures in a small sample may therefore appear reliable despite having a nonzero latent probability of failure under repeated inference. We formulate LLM reliability evaluation as a budget-constrained discovery problem in which each prompt is associated with an unknown per-generation failure probability. We propose a budgeted discovery framework that first performs shallow evaluation across the prompt set and then uses trial-level failure outcomes together with prompt-derived representations to learn a feature-based ranking of failure propensity. The resulting scores prioritize prompts with zero observed shallow failures for deeper evaluation, concentrating the deep-evaluation budget where hidden failures are more likely to be discovered. We evaluate this approach on AIRBench and StrongREJECT across multiple model and system-prompt conditions. The central empirical test asks whether models fit without access to deep-evaluation outcomes can rank prompts with zero observed shallow failures according to their likelihood of producing failures under deeper evaluation. On AIRBench, the highest-ranked 10% of unresolved prompts achieves 2.54x hidden-failure lift for Qwen 2.5 7B and 1.87x for Gemma 3n E4B, recovering 25.4% and 18.7% of subsequently observed hidden failures, respectively, compared with 10% expected under random allocation. Semantic-neighborhood and feature-ablation analyses further show that this predictive signal can be recovered from multiple representations of prompt content and relationships. Comments: 22 pages, 3 figures Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.32547 [cs.LG] (or arXiv:2609.32547v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.32547 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-365] Does Transolver really need a Transformer?
链接: https://arxiv.org/abs/2609.32525
作者: Shizheng Wen,Siddhartha Mishra
类目: Machine Learning (cs.LG)
*备注:
Abstract:The widely used Transolver family of neural operators is based on physics-attention, which softly assigns the points of an unstructured mesh to a small number of slices, applies self-attention among the resulting tokens, and broadcasts the result back to the points. We provide a comprehensive empirical and theoretical analysis to elucidate the mechanisms which are responsible for model performance. To this end, we perform careful ablations on a challenging suite of nine 3D fluid dynamics benchmarks to find that replacing token attention with a constant linear map does not affect the accuracy. Thus, Transolver does not need a Transformer at all. However, removing the global mixing (slicing/deslicing) or doing it only once leads to performance collapse. We leverage the theory of averaging neural operators to explain and corroborate our findings by showing that just slicing/deslicing, in conjunction with pointwise MLPs, already suffices for universal approximation of continuous operators and attention is redundant in this context. Finally, we provide a novel FlashAttention-style efficient implementation of the key slicing/deslicing module of Transolver. This flashslice kernel streams slice and deslice over the points without materializing the heavy slice-weight tensor, while reproducing the best available implementation to floating point error. At the same time, it leads to very significant memory and compute savings, particularly at large slice counts.
[LG-366] What Do Latent Predictive Vehicle Representations Retain? Measuring State Geometry and Local Response
链接: https://arxiv.org/abs/2609.32512
作者: Enzo Nicolás Spotorno,Josafat Leal Filho,Antônio Augusto Fröhlich
类目: Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注: 27 pages, 14 figures, under review, double-blind, code will be made public upon acceptance
Abstract:Models of vehicle dynamics learned from logged states and commands complement physics-based models, and latent world models, which predict in a learned representation, are used to plan and train controllers in other domains. Vehicle controllers are usually specified in physical terms: costs, limits, and references depend on position, yaw angle, speed, and yaw rate, and the optimizer compares or differentiates predicted outcomes across nearby commands. A latent model placed in such a controller must therefore let these quantities be recovered and must change its predictions with commands as the vehicle does, and prediction error on its own latent targets measures neither. We contribute a measurement protocol for action-conditioned latent predictors with a physical readout that separately tests retention, physical-neighborhood organization, forecasting, and local response to command perturbations, using an untrained-encoder reference and three matched response paths that locate errors in the representation or the predictor. In a case study of a temporal joint-embedding predictive model trained on signals logged in IPG CarMaker, the representations retain the measured planar outputs, though an untrained encoder of the same architecture retains them slightly better; future-command input improves one-second forecasts with retention nearly unchanged; and responses to small command pulses diverge from the simulator already in latent coordinates, raising regret when choosing among nearby commands in all comparisons. Updating the predictor on responses corrects them locally at a cost in forecast accuracy. Measuring retention, forecasting, and local response separately is thus what qualifies a predictive latent as a candidate model for control, and the protocol provides the basis for its closed-loop evaluation.
[LG-367] Fast Differentiable SVD on GPU via Polar Decomposition
链接: https://arxiv.org/abs/2609.32505
作者: Uliana Parkina,Askar Tsyganov,Sergei Kudriashov,Sergey Samsonov,Maxim Rakhuba
类目: Numerical Analysis (math.NA); Machine Learning (cs.LG)
*备注:
Abstract:We present a fully GPU-oriented SVD pipeline based on polar decomposition, motivated by iterative methods that rely solely on matrix multiplications, such as the Newton-Schulz iteration. We show that this approach enables up to a 2\times speedup compared to standard implementations. Furthermore, we derive a numerically stable backward pass for the polar decomposition and leverage it to obtain a fully differentiable SVD. Our methods are released as open-source implementations in both PyTorch and JAX: this https URL.
[LG-368] Fisher Simplicity in Kolmogorov-Arnold Networks and Multilayer Perceptrons
链接: https://arxiv.org/abs/2609.32503
作者: Ami Tavory,Meir Feder
类目: Machine Learning (cs.LG)
*备注: 8 pages, 2 figures. Presented at Allerton 2026
Abstract:Kolmogorov-Arnold Networks (KANs) are motivated in part by interpretability: their learned edge functions can be inspected, pruned, and reduced to symbolic structure. In a fixed-basis KAN, this makes a small or zero basis coefficient look like a certificate of simplicity, much as a dead rectified linear unit (ReLU) marks unused computation in a multilayer perceptron (MLP). Fisher nullity gives a precise statistical notion: a parameter direction is Fisher-simple exactly when perturbing it is invisible under the task distribution. We study when these architectural and Fisher notions agree. For a dead ReLU unit, they agree: the closed activation region makes the associated score directions vanish. For a fixed-basis KAN, they do not. In the single-layer Gaussian case, the coefficient Fisher matrix is a basis Gram matrix under the input distribution and is independent of the fitted coefficients. In a multilayer KAN, Fisher simplicity is graph-path based: the data must reach a basis atom and its perturbation must propagate through the downstream network. We encode these two conditions in an effective edge measure and, under local dictionary independence and effective-measure nondegeneracy, show that zero effective exposure exactly identifies Fisher-null directions within an edge. Controlled diagnostics confirm that zero coefficients can preserve rank while effective path disconnections remove the predicted directions. Coefficient magnitude alone is therefore not a Fisher-based pruning criterion for KANs. Comments: 8 pages, 2 figures. Presented at Allerton 2026 Subjects: Machine Learning (cs.LG) MSC classes: 68T07 ACMclasses: I.2.6; G.3 Cite as: arXiv:2609.32503 [cs.LG] (or arXiv:2609.32503v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.32503 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-369] Neural Dynamics as the Composition of Quantized Units
链接: https://arxiv.org/abs/2609.32487
作者: Jacopo Minniti,Aravinth Kulanthaivelu,Richard Sproat
类目: Machine Learning (cs.LG)
*备注: 22 pages, 7 figures
Abstract:Deep learning is commonly interpreted at two levels: the macroscopic, through aggregate trends in loss summarized by scaling laws, and the microscopic, through neurons, features, and circuits. A central challenge is understanding how these levels connect, so that we can explain how elementary computations compose and collectively shape macroscopic behavior. To this end, we study an intermediate abstraction in which training is described as the ordered acquisition of quanta: reusable computations acquired suddenly and binary-activated across examples to reduce loss. By approximating population-gradient updates, we derive quanta’s acquisition dynamics. This yields an acquisition priority governed by demand, how frequently a computation is required across examples, and conditional complexity, how difficult that computation is to acquire given those already available. In a Boolean compositional task, we derive predictions for acquisition order and show how staggered discrete acquisitions can produce smooth aggregate loss and, under certain geometries of quanta composition, give rise to scaling laws. We then train a Transformer to map numerals to English number names and recover candidate quanta from its checkpoint trajectory. From these units, we construct a model that preserves much of the Transformer’s behavior while exposing interpretable latent computations and acquisition dynamics consistent with the theory. Separately, the quanta structure can serve as training targets to improve transformer generalization. Together, these results suggest the quanta abstraction can provide useful computational atoms for studying a variety of macroscopic phenomena.
[LG-370] Elastic Selective Spectral Hybrids for Train-Once Export-Many Budgeted Inference
链接: https://arxiv.org/abs/2609.32486
作者: Dachuan Song,Chuchu Chen,Xuan Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Deploying a language model under different computing and latency budgets calls for compact models with different quality-cost trade-offs. Towards this end, elastic spectral state space models provide ordered and temporally decomposed channels that can be truncated, but they are associated with linear time-invariant filters that cannot selectively preserve relevant past information or forget irrelevant information as the context evolves. To address this, we introduce the Elastic Selective Spectral Hybrid (ESSH), which realizes each Hankel spectral channel as an independent recurrent unit using fitted damped rotation modes. It also features an input-dependent decay and write/read gates that make temporal retention and state update input-dependent while preserving channel-wise truncation and a structured recurrence for efficient execution. ESSH combines these selective spectral mixers with sliding-window attention and jointly trains multiple capacities by reducing spectral-channel count and feed-forward width at different rates through a two-rate capacity map with full-model distillation. The resulting models support chunked parallel training and fused recurrent decoding while avoiding computation for discarded channels. At full capacity, ESSH achieves language-modeling quality comparable to similarly sized independently trained models, while smaller exports exhibit a smooth quality-cost trade-off. We validate the effectiveness of the proposed framework using language understanding, retrieval, cross-domain text, and DNA experiments, by assessing quality retention and the trade-off against independently trained and elastic baselines. At the 1.53B model configuration, fused batch-one decoding takes 1.37 ms per token on a B300, providing a 2.14-2.80x speedup over the tested Mamba-2 and Mamba-3 implementations and 3.03x over Transformer++ at matched parameter counts.
[LG-371] What Should Federated LoRA Share? FedSAIL via Input-aware Subspace Alignment
链接: https://arxiv.org/abs/2609.32485
作者: Junye Du,Shuaida He,Long Feng
类目: Machine Learning (cs.LG)
*备注: 26 pages, 6 figures, 18 tables
Abstract:Federated low-rank adaptation (LoRA) requires identifying an update structure that is shared across heterogeneous clients. Prior work reports strong similarity among trained LoRA projection matrices across clients; however, such agreement may be largely induced by common initialization and collapses toward random overlap under independent initialization. More crucially, relying solely on parameter similarity inherently ignores the influence of local input regime. To uncover a more robust shared structure, we introduce an input-aware action matrix that weights the adapter update by the second-moment statistics of local layer inputs. Empirically, while parameter similarity vanishes, the leading right singular directions of this action matrix remain strongly aligned across clients. This shared geometry preserves task-conditioned differences and naturally varies across network depths. Motivated by these findings, we propose Federated Subspace-Guided Action-Informed Learning (FedSAIL). Instead of averaging weights, FedSAIL estimates a shared action subspace to regularize local training while preserving client-specific coefficients. Across several benchmarks, our approach consistently improves predictive performance over competing federated LoRA methods while reducing communication cost significantly.
[LG-372] Fast and Precise Learned Charged-Particle Trajectory Regression at the Large Hadron Collider
链接: https://arxiv.org/abs/2609.32479
作者: Jonathan Renusch,Benjamin Huth,Daniel Murnane,Eleni Xochelli,Doğa Elitez,Paul Gessinger-Befurt,Andreas Stefl,Jeremy Couthures,Andreas Salzburger,Lukas Heinrich,Michael Kagan,Markus Elsing
类目: Machine Learning (cs.LG)
*备注:
Abstract:We propose a training recipe that treats charged-particle trajectory parameter regression on high-energy physics detector data as a sequence-modeling task. Kalman filters and linearized least-squares fits have been the classical standard approach for this task: they are optimal estimators for sparsely sampled linear-Gaussian data and are commonly used for trajectory parameter regression (fitting). The classical fitting techniques implemented for this domain reach a final precision of one part in 10^5 through detailed modeling of detector geometry, material, detection effects and precise numerical integration of the equations of motion through the detector’s inhomogeneous magnetic field. With this study, we demonstrate that using a bidirectional gated linear recurrent encoder, one is able to reproduce the full precision of classical track fitting techniques. Using a custom kernel, we also achieve significantly higher throughput during GPU inference, compared to classical fitting software running on similarly priced multi-core CPU servers representing typically employed hardware. Such a speedup would lead to considerable cost savings for the pattern recognition at the Large Hadron Collider. To our knowledge, this is the first end-to-end learned track fit to reach the full precision and, at the same time, offer the opportunity to reduce the computing costs.
[LG-373] On the Pitfalls of Verbalized Confidence Priors for Calibrating Large Reasoning Models
链接: https://arxiv.org/abs/2609.32470
作者: Shuoyuan Wang,Beier Luo,Hao Zeng,Chengyao Yu,Songxin Zhang,Zejian Xie,Bingyi Jing,Hongxin Wei
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large reasoning models (LRMs) often suffer from overconfidence when expressing their uncertainty. Confidence-aware reinforcement learning (RL) offers a promising way to optimize calibration. However, it relies on on-policy rollouts and is thus constrained by the model’s pre-RL confidence distribution, which we term confidence prior. In this work, we reveal that off-the-shelf LRMs exhibit a confidence prior heavily concentrated on a few high values, which persists throughout RL. Theoretically, we prove that this concentration suppresses policy gradient updates for rarely sampled confidence values and inflates the lower bound on expected Brier risk. To overcome this exploration bottleneck, we propose CalibSFT, a plug-and-play supervised fine-tuning stage that shapes a calibrated confidence prior with broad support before RL. For each question, CalibSFT constructs confidence targets combining its success rate with response-level correctness, which provably preserves proper-scoring optimality, and then balances training responses across the confidence spectrum to enable diverse confidence exploration during RL. To learn from incorrect responses without imitating their reasoning, CalibSFT introduces correctness-conditional supervision, guiding confidence across all responses while supervising reasoning only on correct ones. Across 16 mathematical and general reasoning benchmarks, incorporating CalibSFT reduces calibration errors and improves discrimination across five representative RL algorithms while preserving comparable accuracy. Furthermore, CalibSFT delivers practical benefits for downstream selective prediction and model routing. Our code is available at this https URL.
[LG-374] PolyStepOR: Learning to Decide Without Optimal Decisions
链接: https://arxiv.org/abs/2609.32465
作者: Viet The Nguyen,Gunther Gust,An Thai Le
类目: Machine Learning (cs.LG)
*备注:
Abstract:Decision-focused learning (DFL) trains predictors for downstream decision quality, but often relies on optimal reference decisions that are expensive to obtain. We present PolyStepOR, which trains directly from realized decision costs without pre-computed optima and extends to in-constraint predictions through repair or infeasibility penalties. To handle piecewise-constant losses, PolyStepOR perturbs predictor parameters, evaluates the resulting decisions, and uses optimal transport to favor lower-cost directions, requiring no derivatives. Without task-specific tuning, PolyStepOR performs strongly on classical optimization benchmarks and competitively on predicted-constraint and real-world problems. Theoretically, we characterize decision-preserving perturbations and boundary detection, bound sensitivity to cost errors, and establish stationarity guarantees for a smoothed objective. PolyStepOR thus replaces optimal reference decisions and derivatives with forward evaluations.
[LG-375] Write Back the Δ: Revisiting the Same Tokens with Fresh Representations
链接: https://arxiv.org/abs/2609.32457
作者: Wencheng Ye,Anning Hu,Xiangdong Zhang,Tianyi Wang,Yikang Li,Hengyu Jin,Bing Li,Junchi Yan
类目: Machine Learning (cs.LG)
*备注:
Abstract:Transformers process information strictly forward through depth, preventing deeper computation from revisiting and refining earlier representations. To augment the standard forward pass, existing approaches either re-execute depth, incurring additional computation, or modify the residual stream using predefined directions, limiting their instance-level adaptation. Recently, inference-time feedback offers a direct mechanism for recycling endogenously produced computation by writing deeper residual states back to earlier layers, yet what should be fed back remains unclear. We argue that the depth increment Delta, capturing newly accumulated computation between two layers, provides a more effective, composable, and scalable feedback signal than the full state. Building on this observation, we introduce ReFlux, a learnable feedback graph that dynamically selects and composes increment-carrying routes. ReFlux supports synchronous feedback to the same token and streaming feedback to subsequent tokens. Extensive experiments across various models, corpora, and benchmarks show that synchronous ReFlux consistently reduces perplexity across ten language-modeling corpora, and improves accuracy by 2.1-2.3 points, with gains reaching 4.7 points on multi-hop reasoning. Streaming ReFlux further retains most of these gains while preserving the base model’s 1x theoretical backbone FLOPs. These results establish ReFlux as an efficient paradigm for unlocking the latent computational potential of LLMs, allowing them to revisit the same tokens with fresh representations. Code implementation can be found at this https URL.
[LG-376] Length-Independent State Tracking Under a Parallel Scan
链接: https://arxiv.org/abs/2609.32447
作者: Julien Brandoit,Arthur Fyon,Thomas Braipson,Tom Clara,Florent De Geeter,Pierre Sacré,Damien Ernst,Guillaume Drion
类目: Machine Learning (cs.LG)
*备注: this https URL
Abstract:Learning robust and scalable finite-state tracking is fundamental to sequence processing. While linear recurrent neural networks (RNNs), linear attention, and state space models enable scalable parallel training through affine recurrences, their theoretical expressivity guarantees assume idealized arithmetic and do not extend to finite precision, where the parallel scan that makes them fast is itself a source of perturbation. We formalize finite-state tracking at finite precision and characterize length independence: tracking that stays correct at every sequence length, at a precision cost that does not grow with the length. We show that length-independent state tracking requires two competing dynamics within a single map: contraction to suppress numerical perturbations and separation to keep distinct states apart. We prove that affine recurrences, which offer a single rate at each step to serve both roles, realize at most definite automata at finite precision. Instead of treating scan compatibility as a restriction on the update map, we reinterpret it as a computational budget and introduce the Neural Finite-State Machine (NFSM): a nonaffine, scan-compatible RNN built for length-independent finite-state tracking. On synthetic benchmarks spanning abelian and nonabelian groups, noninvertible monoids, and textual state-tracking tasks, affine baselines fail on every nondefinite task, most of them within a few hundred steps. A single NFSM layer instead learns the exact transition tables of every algebraic task, which certifies correctness beyond the tested lengths, and a stack of NFSMs keeps perfect accuracy on the textual tasks at every tested length.
[LG-377] Beyond the Manifold Hypothesis: Hybrid Spectral Parameterizations for Flow Matching
链接: https://arxiv.org/abs/2609.32432
作者: Ségolène Martin,Anne Gagneux,Quentin Bertrand,Rémi Emonet,Mathurin Massias
类目: Machine Learning (cs.LG)
*备注:
Abstract:Flow matching and diffusion can be trained to predict different quantities, most commonly the data x_1 , the source noise x_0 , or the velocity~ v . Although theoretically equivalent, these can lead to substantially different performances. We identify two main drivers for these differences: the source–data signal-to-noise ratio, and the information bottleneck induced by the neural architecture. We show that, beyond intrinsic data dimension, the factor affecting the optimal parametrization the most is a certain signal-to-noise ratio in each data covariance direction. From this analysis, we introduce new \emphspectral hybrid parameterizations that adapt across time and data covariance directions; we show that these are optimal for Gaussian data. We also show that architecture-induced compression changes which parameterization is easier to learn, with v -prediction being more sensitive to discarded directions than x_1 -prediction. Experiments across architectures and source scales show that our spectral parameterizations are robust across regimes, can substantially accelerate optimization, while incurring essentially no additional training cost compared with standard parameterizations.
[LG-378] HoTS: Homophily-Aware Temperature Scaling for Graph Neural Network Calibration NEURIPS2026
链接: https://arxiv.org/abs/2609.32426
作者: Inwoo Tae,Yoontae Hwang,Yongjae Lee
类目: Machine Learning (cs.LG)
*备注: 37 pages, 9 figures. Accepted at the 40th Conference on Neural Information Processing Systems (NeurIPS 2026)
Abstract:For graph node classification, calibrated class probabilities are needed when confidence scores, usually the maximum predicted class probability, are used to rank predictions, defer uncertain nodes to human review, or control risk. Existing post-hoc calibrators either apply one global temperature or use graph-aware modules without a principled structural form. We study how local graph structure should enter node-level calibration. Our first results show that a logit-only temperature rule is insufficient when nodes with identical logits but different local homophily require different optimal temperatures. We then analyze a population-concentration contextual stochastic block model with Gaussian features and a one-layer linear GCN. Under equidistant class means, the Bayes posterior over class-template scores is a temperature-scaled softmax whose inverse-temperature is governed by a homophily-dependent signal strength. In the positive-signal homophilic regime, the resulting temperature decreases approximately inversely with normalized local homophily. This law motivates Homophily-aware Temperature Scaling (HoTS), a simple post-hoc calibrator that assigns each node a positive scalar temperature from entropy-based logit concentration and estimated local homophily. HoTS has three temperature parameters, preserves the predicted class, and learns the strength of the structural correction from calibration data. Across 18 node-classification benchmarks, two GNN backbones, and eight calibration baselines, HoTS achieves the best mean Expected Calibration Error (ECE) of 4.79%, the best average rank, and the most reliable confidence ranking in selective classification. Code is available at this https URL.
[LG-379] Recovery-Directed Symbolic Distillation of Neural Likelihoods
链接: https://arxiv.org/abs/2609.32409
作者: Kianté Fernandez,Xinwei Li
类目: Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC); Computation (stat.CO); Machine Learning (stat.ML)
*备注: 20 pages, 8 figures, 4 tables
Abstract:Amortized neural likelihoods enable computationally expensive inference for models with analytically intractable or unspecified likelihoods, but their black-box nature limits interpretability. We introduce a symbolic distillation pipeline that converts trained neural likelihoods into explicit, interpretable expressions optimized for efficient parameter estimation. Our approach uses a recovery-directed objective to guide symbolic regression toward expressions that preserve parameter-recovery accuracy rather than merely approximating the likelihood function. Candidate expressions are evaluated on held-out datasets and selected using a criterion that jointly accounts for expression complexity, parameter-recovery performance, and distributional distance from the learned likelihood. We evaluate the pipeline on the diffusion decision model, a classical cognitive model, whose analytically tractable likelihood provides ground truth for controlled evaluation. The proposed recovery-directed objective improves parameter recovery over standard symbolic-regression objectives. The resulting symbolic likelihoods enable over 100 times faster parameter evaluation than both neural likelihoods and, when available, the exact likelihood, while maintaining a manageable loss in precision. We further demonstrate these computational benefits in Bayesian hierarchical inference on empirical data. Our pipeline provides a lightweight interface for integrating symbolic distillation with existing neural-likelihood estimation methods and can be adapted to a range of simulation-based inference settings.
[LG-380] Using Machine Learning to Investigate Predictors of Fasting Blood Glucose: Insights into Circadian Timing and Age Interactions
链接: https://arxiv.org/abs/2609.32386
作者: Viktoriya Bu-Dager,Silvia Cirstea
类目: Machine Learning (cs.LG)
*备注: 25 pages, 10 figures, 11 tables
Abstract:Impaired glucose regulation is a major contributor to metabolic dysfunction and type 2 diabetes. This study developed an interpretable machine-learning framework to predict log-transformed fasting blood glucose using metabolic, hormonal, lifestyle, demographic, nutritional, and circadian variables from the National Health and Nutrition Examination Survey 2017–2020 pre-pandemic dataset. After merging multiple NHANES sub-datasets, data processing used a leakage-resistant pipeline in which imputation, scaling, and one-hot encoding were performed only after dataset splitting and within training folds. Elastic Net, LASSO, and XGBoost models were evaluated using 94 candidate predictors and engineered circadian interaction terms. Performance was assessed using mean absolute error, root mean squared error, coefficient of determination, calibration, and Shapley Additive Explanations. The final interaction-augmented XGBoost model achieved strong performance on the independent test set, with a mean absolute error of 0.0804, a root mean squared error of 0.1148, and a coefficient of determination of 0.7761, using 10 predictors. Glycohemoglobin was the dominant predictor, followed by insulin, diabetes diagnosis, gamma-glutamyl transferase, age, race, and gender. Among the engineered interaction terms, sleep midpoint multiplied by age was consistently retained in repeated random-split analyses, although its contribution remained modest relative to dominant glycaemic predictors. These findings support further investigation of circadian-age interactions in metabolic health.
[LG-381] Attribution Without a Second Pass: Inline Per-Sample Gradient Provenance at ~1% Overhead
链接: https://arxiv.org/abs/2609.32380
作者: Amit Nautiyal
类目: Machine Learning (cs.LG)
*备注: 11 pages. Code: this https URL
Abstract:Data attribution methods used in practice (TRAK, LoGRA, EK-FAC) are post-hoc: after training they make a second pass over the training set to recompute per-sample gradients, repeated per checkpoint when ensembled. Traceprop avoids that pass by recording projected per-sample gradients inline on the training backward pass. A Kronecker-factored sketch scales from a single tracked layer to every layer without materializing a dense projection matrix. On LoRA fine-tunes of GPT-2 and Pythia models up to 2.8B on one NVIDIA L4, inline logging costs 0.30-1.08% of wall-clock time at last-block scope and stays under 1% (0.79%) even when tracking every layer of Pythia-1B. Against LogIX, the closest inline-capable competitor, the factored sketch is 2.0-4.1x cheaper at equal storage, a gap that grows with tracked scope and is significant at every scope tested, while matching or exceeding LogIX’s attribution quality at matched storage. Building the attribution-ready store inline is 60-242x cheaper than one post-hoc pass and 301-1211x cheaper than a five-checkpoint TRAK ensemble, with recorded gradients matching autograd exactly. Because each stored gradient carries source-file lineage, the same pass also produces EU AI Act Article 26 audit trails.
[LG-382] STRIDE: State-Transition Representation via Increment Dynamics and Evolution
链接: https://arxiv.org/abs/2609.32367
作者: Yuchen Xiong,Siming Huang,Jianfeng Sun
类目: Machine Learning (cs.LG)
*备注:
Abstract:We introduce STRIDE (State-Transition Representation via Increment Dynamics and Evolution), which defines states through derivative fingerprints and learns local functions for state transitions (qpairs), recasting continuous forecasting as transition prediction. Trailing convolution windows estimate local joint-transition frequencies, whose lagged differences form a high-dimensional increment trajectory. Proper orthogonal decomposition (POD) gives coordinate paths, jointly forecast by sparse dynamics with memory. Recombining their forecasts and applying history-anchored inversion recovers future transition distributions; sampled state paths select local functions to generate continuous forecasts. Three-seed experiments compare STRIDE against fourteen baselines across nine benchmark families. Five independent Markov and hidden-state baselines cover all 230 evaluated tasks, with 220 complete whole-horizon pairs. Against these comparators, system-weighted late-Energy win fractions range from 73.6% to 78.5% on 190 multi-step pairs; against DLinear, the fraction is 77.3% on 72 paired multi-step tasks. Matched controls examine the intermediate representation. On 64 independently initialized Aizawa trajectories, late-Energy reductions against four matched controls range from approximately 24% to 56%, with all four prespecified contrasts passing Holm correction. These results connect transition-statistic prediction to continuous probabilistic forecasting, with substantial long-horizon gains in the matched Aizawa study.
[LG-383] Editable Map-Conditioned Trajectory Generation for Human Mobility Simulation
链接: https://arxiv.org/abs/2609.32360
作者: Takayuki Mizuno,Shouji Fujimoto,Mikito Hiruki,Atushi Ishikawa
类目: Machine Learning (cs.LG); Computers and Society (cs.CY); Physics and Society (physics.soc-ph)
*备注: Accepted at the 9th ACM SIGSPATIAL International Workshop on GeoSpatial Simulation (GeoSim '26). 10 pages, 5 figures
Abstract:Geospatial simulation of infrastructure interventions requires mobility generators that respond directly to edited maps, yet many data-driven generators do not expose the map as an editable condition. We formulate this task as map-conditioned autoregressive generation of human mobility: a road raster conditions a decoder that emits nominal 31.25 m mesh-cell tokens at one-minute intervals. The mesh-local vocabulary supports held-out and locally edited maps without retraining or vocabulary changes. We instantiate a ResNet-50 visual-prefix configuration and a Vision Transformer (ViT) cross-attention configuration, trained from scratch on 87,400 smartphone-derived trajectories from 874 meshes in Ishikawa Prefecture, Japan; 219 meshes are held out. We evaluate map sensitivity by comparing correct-map and within-split shuffled-map generations with held-out real trajectories. On the 110-mesh test split, for the ResNet-50 configuration, correct-map generations are closer than shuffled-map generations on 60% of meshes under Hausdorff-based energy distance (p = 0.021), while DTW is directional but inconclusive (57%, p = 0.074); correlation with real density is 0.38 with the correct map versus 0.01 with shuffled maps. The ViT configuration shows weaker trajectory-level sensitivity and smaller density gains. An illustrative bridge-removal edit changes generated continuations without retraining. Together, these results support the feasibility of editable-map human-mobility simulation.
[LG-384] A Comparative Analysis of Attention versus State-Space Models for In-Context Learning
链接: https://arxiv.org/abs/2609.32341
作者: Enes Arda,Semih Cayci,Atilla Eryilmaz
类目: Machine Learning (cs.LG)
*备注:
Abstract:Transformers and state-space models (SSMs) are two prominent sequential learning architectures, yet their comparison remains largely empirical and existing theoretical analyses are typically task-specific or architecturally restricted. In this paper, we develop belief geometry, a unified analytical framework for comparing the representational capabilities of broad classes of attention and SSMs. Starting from a generalized formulation of in-context linear regression and using cumulative Bayes regret as our measure, we abstract three capabilities required by many sequential learning problems in our belief geometry: evidence assembly, belief maintenance, and addressing. We then study three cases of our formulation that isolate these capabilities and yield sharp architectural lessons: For belief maintenance, SSMs attain the optimal regret over stationary aggregation kernels; for positional assembly, SSMs have a memory advantage; and for content addressing, softmax attention has an exponential width advantage over sigmoid-selective SSMs. Experiments with LLaMA-type Transformers and Mamba-2 show that these architectural insights extend beyond our analytically tractable classes and linear-regression testbed.
[LG-385] When the Merge Coefficient Stops Mattering: Proximity Regularized Merging for Continual LoRA Adaptation
链接: https://arxiv.org/abs/2609.32332
作者: Yixuan Liu,Yuhao Sun,Sen Song,Jin Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:Rehearsal-free continual learning with parameter-efficient adapters can be cast as a sequence of task-vector write-in operations: for each new task, a low-rank adapter is learned and merged into a running model. We propose Proximity Regularized Merging (PRM), a minimal modification to sequential LoRA merging that adds a proximal penalty during task-vector training without changing the subsequent write-in rule. PRM acts as a robust task-vector regularizer: in the reported Base-+Prox diagnostics, it improves AAA across multiple write-in rules, backbones, and class-incremental settings, while its fixed-coefficient variant remains competitive with strong coefficient-based baselines. Mechanistically, matched-prefix norm controls and proximal-strength sweeps show that proximal training shrinks the task-vector radius, lowers Fisher-weighted interference, broadens the coefficient plateau, and exposes a stability-plasticity trade-off. Together, these results suggest that the effectiveness of sequential LoRA merging depends not only on how much of a task vector is written in, but also on whether the task vector itself has been trained to be mergeable.
[LG-386] Continual Data Unlearning in Diffusion Models via Transition-based Regularization
链接: https://arxiv.org/abs/2609.32328
作者: Sunbeom Jeong,Sehwan Kim,Sangwoo Hong,Jungwoo Lee
类目: Machine Learning (cs.LG)
*备注: Preprint
Abstract:Data unlearning in diffusion models aims to remove the influence of specific training examples without suppressing the broader concepts they represent. However, when deletion requests arrive sequentially, updates for new requests can degrade generative utility and undermine earlier deletions. We propose a continual data unlearning framework that uses completed deletion transitions as directional references to regularize future updates. For each request, we record changes in denoiser responses on the same fixed noisy inputs before and after unlearning. Rather than matching full post-deletion responses, we apply a one-sided penalty that discourages reversal along the recorded directions relative to the post-deletion references, while leaving orthogonal response changes and progress beyond these references unpenalized. To keep storage independent of the number of requests, we maintain a fixed-capacity bank of representative transition records. Records are selected based on the local sensitivity of progress along their recorded directions to parameter updates, allowing them to be retained even when their penalties are inactive. Empirical evaluations show that the proposed framework achieves a better balance between deletion persistence and generative utility than existing unlearning baselines as requests accumulate, using only a small transition memory.
[LG-387] Not All Errors Matter: Decision-Relevant Prediction Error Predicts Planning Quality
链接: https://arxiv.org/abs/2609.32322
作者: Linhao Wang,Yiyan Fan,Dongjin Huang
类目: Machine Learning (cs.LG)
*备注:
Abstract:World models are typically trained and evaluated by prediction error, assuming that more accurate predictions lead to better decisions. We show that this assumption can fail because models with similar total error can differ substantially in planning performance when their errors occur on different state dimensions. We introduce Decision-Relevant Prediction Error (DRPE), which measures prediction error on the state dimensions that affect decisions. We also develop an iso-error evaluation protocol that varies error allocation while keeping total error fixed. In a factored gridworld with known state relevance and a standardized planner, we evaluate 55 controlled and learned models across different error levels and allocations. Total prediction error is weakly related to planning success (Spearman \rho=-0.25 ), whereas DRPE is strongly predictive ( \rho=-0.84 ; -0.98 within the controlled family). Models with only a 1% difference in total error can differ by 60 percentage points in planning success (97% vs 37%). The relevant error also depends on the task, with model rankings reversing across tasks at the same total error. Deeper imagination further amplifies decision-relevant errors, while learned models exhibit systematic bias on rare but decision-critical events. We formalize sufficient conditions under which DRPE correctly ranks models and total prediction error cannot.
[LG-388] All On-Board: Fully On-Chip Neuromorphic Q-Learning with Embedded CartPole Simulation
链接: https://arxiv.org/abs/2609.32317
作者: Steven C. Nesbit,Giovanni T. Michel,Gerd J. Kunde,Edward Kim,Andrew T. Sornborger
类目: Neural and Evolutionary Computing (cs.NE); Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注: 10 pages, 4 figures, 2 tables
Abstract:As AI models grow in size and usage, their energy demands increase dramatically, raising sustainability and economic concerns. Neuromorphic hardware, inspired by the energy efficiency of the brain, seeks to address this challenge by offering low-power, fast-processing alternatives to conventional computing. Such hardware is particularly well-suited to control systems deployed in resource-constrained environments, which are best trained via reinforcement learning (RL). This contribution presents the design and implementation of a fully on-chip, closed-loop Loihi 2 RL agent. Our neuromorphic circuit consists of a fully embedded Q-learning algorithm and an on-chip simulation of the CartPole-v0 environment on Loihi 2. Our Q-learning algorithm trained the same number of successful agents as the CPU implementation in only half the execution time and with two orders of magnitude less dynamic power. These findings demonstrate the viability of RL on neuromorphic hardware and highlight its promise for building energy-efficient, real-time, embedded AI systems.
[LG-389] On the Capability and Limitation of Hard Prompt
链接: https://arxiv.org/abs/2609.32302
作者: Lijia Yu,Shuaitong Liu,Gaojie Jin,Xinyu Li,Xiao-Shan Gao
类目: Machine Learning (cs.LG)
*备注:
Abstract:Prompt engineering has become an indispensable tool for using large language models (LLMs), turning LLMs into task-specific experts without changing their weights. Despite notable theoretical advances in prompt engineering, the theory for the more practical hard or discrete prompts is largely open. In this paper, we try to fill this gap either by providing a complete solution or by making substantial progress on the three core theoretical questions regarding hard prompts. First, we show that determining the existence of a hard prompt for a transformer to solve a downstream task is NP-complete and that finding an optimal hard prompt is NP-hard, which is the first computational complexity result for hard prompting, as far as we know. Second, we show that, unlike soft or continuous prompts, hard prompts have essential limitations: hard prompts are not complete; short hard prompts do not significantly enhance the ability of transformers; and long hard prompts exhibit the “prompt dominating answer phenomenon,” meaning that, with high probability, the same answer is given for all queries of the same length. On the other hand, linear hard prompts do not have the limitations of short or long prompts. Third, we provide a tight bound on the size of the task in terms of the prompt length for the performance of prompts on the finite task to generalize to the entire data distribution, leading to a necessary and sufficient condition for generalizability. This is the first result on generalization for prompting, as far as we know. Our findings not only offer the first theoretical insights into hard prompts but also provide provably reliable practical guidance for real-world LLM usage.
[LG-390] When Does Synergy Help Active Feature Acquisition? A PID-Based Study
链接: https://arxiv.org/abs/2609.32301
作者: Jie Li,Maruf A. Dhali,Hjalmar R. Bouma
类目: Machine Learning (cs.LG)
*备注:
Abstract:Active feature acquisition (AFA) sequentially selects informative features under budget constraints. However, existing policies rarely distinguish whether information contributed by interacting features is redundant, unique, or synergistic. We introduce SynAFA, a state-dependent AFA policy that combines pairwise joint information and conditional information, with Partial Information Decomposition (PID) characterizing its information structure. Across five tabular datasets and MNIST-loop, performance is heterogeneous. SynAFA shows its strongest gains at low budgets on PhysioNet, where synergistic and redundant feature pairs are supported by permutation tests, but its advantage diminishes as budgets increase and does not depend on pair proposals. SynAFA performs significantly worse than nearly all baselines on MiniBooNE, and than CAE on MNIST-loop. Controlled synthetic experiments further show that, within budget, SynAFA’s advantage rises as joint information becomes more synergy-dominated, including when total joint information is held approximately constant. Further analyses show that the budget-dependent erosion is not resolved by non-greedy local search, which improves a diagnostic set-level objective but leaves predictive performance no better, often significantly worse, and that improvements in this objective are weakly aligned with the fixed classifier’s predictive utility. These findings characterize when pairwise synergy can benefit AFA while exposing a persistent challenge in translating local information into effective acquisition objectives.
[LG-391] Refresh or Realize? Compute Allocation in Drifting Models
链接: https://arxiv.org/abs/2609.32298
作者: Sipeng Chen,Xu Zheng,Shibo Li
类目: Machine Learning (cs.LG)
*备注: 20pages
Abstract:Drifting Models train a one-step generator by recomputing a finite-sample drift field at every iteration and taking an optimizer step toward the drifted target. The field says how generated samples should move, but the step is taken in parameters shared by all samples, so the motion the network actually makes need not match the motion it was given. This leaves a basic training question open: should extra compute go into fitting the current target more closely, or into recomputing the field? We study it on ImageNet 256x256. Holding the target fixed for k optimizer steps and measuring the realized displacement, we find that deeper fitting does bring the network closer to the frozen target, and that the number of steps needed before it makes any net progress drops from about sixteen early in training to one later on. When the extra steps come for free, k=2 also lowers FID. Once they are paid for, the result flips: at approximately matched measured wall-clock, spending the budget on fresh fields gives lower FID than deeper fitting, on both training seeds. The target itself shows why a fresh field is worth so much. Redrawing the finite support rotates its direction far more than a parameter update does (cosine ~0.3-0.6 against ~0.95), and a correction that is optimal in field space is not reliably better in FID than a parameter-free one. For Drifting, fitting each target well and spending compute well are different goals.
[LG-392] FUND: Density Flow for Sampling Unnormalised Distributions
链接: https://arxiv.org/abs/2609.32296
作者: Vikas Kanaujia,Vipul Arora
类目: Machine Learning (cs.LG); High Energy Physics - Lattice (hep-lat)
*备注: 24 pages, 6 figures
Abstract:Efficient sampling from Boltzmann distributions is central to modelling complex physical systems. Markov Chain Monte Carlo (MCMC) methods suffer from critical slowing down, high autocorrelation, and poor mode-mixing, limiting their scalability. Recent advances, like Boltzmann Generators, offer a promising alternative but remain constrained by costly MCMC-based training, inefficient sampling, and poor ergodicity. We introduce an algorithm for learning Boltzmann distributions that does not require any true samples for training. Our approach draws inspiration from flow matching but departs fundamentally from sample-trajectory matching to distribution-trajectory matching. The algorithm iteratively reshapes the target distribution, using model generated samples to guide learning and ensure comprehensive mode coverage. We validate our method on standard benchmarks, including a 2D Gaussian mixture, Many-Well distributions, and high-dimensional scalar \phi^4 theory. The proposed approach not only improves sampling performance and accuracy over traditional MCMC and flow-based baselines but also establishes a new method for sample-free learning of complex physical distributions.
[LG-393] A Journey to the Edge of Stability
链接: https://arxiv.org/abs/2609.32290
作者: Jaerin Lee,Kyoung Mu Lee
类目: Machine Learning (cs.LG)
*备注: 33 pages, 20 figures
Abstract:It has recently been found that deep learning often occurs at the “edge of stability (EoS),” where the maximum Hessian eigenvalue of the model is stabilized at a value reciprocal to the learning rate. However, what happens before we reach that regime? We fix a deep learning problem and vary first order optimization methods with dense learning rate sweeps. We then track the characterizing quantities of a learning trajectory: the loss, the sharpness, and the alignment between consecutive gradients. To our surprise, if we scale the learning rate by the dc gain of the optimizer, these traces from the sweeps from different optimizers almost perfectly overlap across a large range of learning rates. The dc-normalized optimizers have another role that only becomes apparent in high learning rates: they select when the sharpness value detaches from this universal curve and enters the edge of stability. Upon this discovery, we specify three distinct regimes with respect to the dc-adjusted learning rate: the low-LR regime where the trajectory is nearly insensitive to the optimizer, the high-LR regime, where the optimizer governs the sharpness according to the EoS reciprocal rule, and the in-between mid-LR regime where so-called progressive sharpening originates independently of the optimizer. This distinguishes the role of the optimizer, the learning rate, and the model in shaping the learning progress.
[LG-394] Neural ODEs Meet Concurrent Learning: Stable Online Learning with Lyapunov Guarantees
链接: https://arxiv.org/abs/2609.32289
作者: Omkar Sudhir Patil
类目: ystems and Control (eess.SY); Machine Learning (cs.LG)
*备注:
Abstract:Neural ODEs learn dynamics from trajectory losses, but their adjoint gradients lack the regressor-times-parameter-error structure on which Lyapunov analyses of online adaptation rest, so training on streaming data comes without stability guarantees. We show that this structure is in fact present: the adjoint gradient decomposes exactly into a positive semi-definite trajectory operator acting on the parameter error plus a nonlinear perturbation with explicit, horizon-dependent bounds. A quadratic Lyapunov function then certifies online Neural ODE training over sliding windows under computable gain and horizon conditions, and the same certificate extends to stored data: its drift branch recovers concurrent learning, and its trajectory branch yields NODE-CL, a stored-segment Gauss-Newton method built on batched forward sensitivities that needs no state-derivative estimates. On four DeepMind Control Suite domains, NODE-CL attains the lowest median prediction error on three under velocity measurement noise, where observer-based concurrent learning degrades by up to 8x; with clean measurements it is best on the pendulum and within a factor of 1.6 of the best stored-data baseline on the cartpole and reacher.
[LG-395] SIMANF: Sample Free Learning of Unnormalized Distributions via Simulated Annealing in Normalizing Flows
链接: https://arxiv.org/abs/2609.32279
作者: Vikas Kanaujia
类目: Machine Learning (cs.LG); High Energy Physics - Lattice (hep-lat)
*备注:
Abstract:Efficiently learning and sampling from high dimensional, multimodal unnormalized distributions without target samples remains a challenging problem. Although normalizing flows can generate samples efficiently, training based on the reverse KL divergence using only the unnormalized target density may suffer from mode collapse. We introduce SIMANF, a sample free framework that integrates simulated annealing with normalizing flows. SIMANF progressively transforms the target distribution from a smooth initial form to the original target distribution and trains the flow sequentially across these stages. By transferring the learned representation between stages, the method promotes mode coverage while progressively capturing finer features of the target distribution. Following annealing, a final refinement stage combines the reverse KL divergence with an importance weighted forward KL objective using samples generated by the flow. SIMANF requires no target samples during training and uses only the unnormalized density. We demonstrate its effectiveness on Many-Well distributions and high dimensional Scalar Phi4 lattice field theory distribution.
[LG-396] HyperLabel: Multi-Label Classification via Hypergraph-Based Label Correlation Modeling
链接: https://arxiv.org/abs/2609.32276
作者: Peiyu Zhang(1),Heng Ping(1),Nikos Kanakaris(2),Yucheng Zhao(3),Shixuan Li(1),Wei Yang(1),Xiongye Xiao(3),Paul Bogdan(1) ((1) University of Southern California, (2) Amazon Web Services, (3) University of Tennessee, Knoxville)
类目: Machine Learning (cs.LG)
*备注: 15 pages, 3 figures
Abstract:Multi-label classification (MLC) requires predicting multiple relevant labels for each instance, where a central challenge is modeling complex label dependencies arising from co-occurrence patterns. Existing approaches are limited in capturing high-order label correlations, relying on implicit learning through contrastive objectives or pairwise attention mechanisms without structural guidance. We propose HyperLabel, an encoder-decoder framework that explicitly models label dependencies through hypergraph neural networks. Our contributions are twofold: (i) We construct a label hypergraph where sample-defined hyperedges naturally encode multi-way co-occurrence patterns, providing explicit structural prior knowledge that captures relationships beyond pairwise interactions. (ii) We propose a unified cross-modal learning approach where HGNN+ performs bidirectional message passing to integrate feature information with label structure, and a shared cross-attention decoder processes both modalities through complementary learning objectives. Extensive experiments on seven benchmark datasets demonstrate that HyperLabel achieves state-of-the-art performance, with particularly significant improvements on macro-F1 scores (+10.3% on Delicious, +8.2% on Bibtex), validating that explicit hypergraph structure effectively captures complex label relationships. The code is available at this https URL .
[LG-397] opology-Adaptive Hyperbolic Graph Attention Networks Guided by the Hyperbolic Sombor Index
链接: https://arxiv.org/abs/2609.32275
作者: Haifang Cao,Boan Tao,Xiyuan Gao,Timing Li,Yu Wang,Pengfei Zhu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Hyperbolic geometry has emerged as a principled space for representing hierarchical graphs. However, existing hyperbolic graph neural networks typically rely on shared curvature configurations and feature-driven attention, failing to explicitly exploit local hierarchical topological patterns. To bridge this gap, we introduce the Hyperbolic Sombor Index (HSO) as a lightweight structural prior for capturing hierarchy-indicative degree stratification. Building on this, we propose \textbfHSO-GAT, a topology-adaptive hyperbolic graph attention network that unifies geometric adaptation and message propagation. Specifically, it comprises two complementary modules: HSO-Guided Local Curvature Adaptation, which performs adaptive node-wise geometric scaling from aggregated node-level HSO signals, and HSO-Gated Hyperbolic Graph Attention, which enables structure-aware message passing through feature-conditioned gating. Theoretically, we establish the monotonic sensitivity of edge-level HSO to degree imbalance and analyze the validity and radial scaling properties of node-adaptive hyperbolic mappings. Extensive experiments on eight benchmark datasets demonstrate that HSO-GAT consistently achieves state-of-the-art performance in both node classification and link prediction tasks.
[LG-398] Certification Frontiers for Gaussian LoRA: Independent Priors Posterior Risk and Prediction-Preserving Balancing
链接: https://arxiv.org/abs/2609.32271
作者: Joyanta Jyoti Mondal,Ibne Farabi Shihab
类目: Machine Learning (cs.LG)
*备注: 32 pages, 9 figures
Abstract:Post-hoc Bayesian fine-tuning places Gaussians around trained low-rank adapters, yet a calibrated posterior does not by itself yield a useful generalization certificate. Such a posterior admits an informative PAC-Bayes certificate only when both the loss of its sampled predictors and its KL divergence from an admissible prior are small. In this research, we characterize this certification frontier for Gaussian LoRA posteriors and separate three interventions: changing the prior, changing the stochastic predictor, and changing only how its complexity is counted. First, an exact isotropic KL envelope eliminates the prior scale and yields a width threshold that excludes posterior widths before sampling, while a zero-KL floor identifies targets that no complexity reduction can reach at a measured risk bound. Second, we minimize KL in closed form over the full GL® symmetry of the low-rank factors, leaving every sampled adapter product unchanged, and derive the noncentral objective required when the prior center is trained on an independent split. On a small-pool RoBERTa audit of 567 configurations, the recorded 64-draw summaries imply a certificate floor of 0.7298 even with zero KL and Chernoff accounting, so reducing complexity alone cannot certify these posteriors at the recorded budgets. For a stable posterior in a controlled Gaussian-factor task, Chernoff accounting certifies risk below 0.1 on 20 of 20 datasets with 1024 draws, whereas Hoeffding certifies none. On deliberately deformed synthetic rank-four factors, matrix balancing reduces KL by 29.3% on average beyond scalar balancing without changing any prediction. Numerical split-prior scenarios make the remaining risk and complexity budgets explicit.
[LG-399] Self-Reconstruction Dynamics for Autoencoder Reconstruction Refinement
链接: https://arxiv.org/abs/2609.32268
作者: Hitoshi Iyatomi
类目: Machine Learning (cs.LG)
*备注: 23 pages, 12 figures
Abstract:Standard autoencoder (AE) inference uses a single encoder-decoder pass, though the latent may not be optimal for each sample under a fixed decoder. We ask whether a trained AE can reveal information for improving its own reconstruction. Repeated application of a frozen AE to its reconstruction produces transient image- and latent-space trajectories, termed Self-Reconstruction Dynamics (SRD). Although this degrades fidelity in the AEs studied here, SRD contains sample-specific information for correcting the reconstruction. We propose SRD-guided Reconstruction Refinement (SRD-RR), which predicts a latent correction from a short SRD with the AE frozen and no per-sample test-time optimization. We also introduce MSE-recov, an MSE recovery ratio relative to an empirical decoder-optimized reference. Across six datasets, SRD-RR recovers 38.6% of the empirically recoverable MSE gap with one transition and 45.3% with two. A two-transition variant trained without direct access to original images, using an SRD-derived pseudo-target, achieves 40.7% recovery and a 1.74 dB average PSNR gain. Removing trajectory information reduces the gain, while cross-sample trajectory assignment causes severe degradation, confirming strong sample specificity. Nonlinear SRD-conditioned refinement consistently outperforms fixed and trained linear latent correction. On a pretrained DINOv2-based representation autoencoder (RAE) with substantially different latent dynamics, SRD conditioning again improves a matched trajectory-free predictor. However, pixel-MSE latent refinement reveals a strong mismatch between pixel fidelity and perceptual quality, while the SRD-derived pseudo-target mitigates this degradation. Overall, SRD is a useful sample-specific refinement signal, while the objective determines how it translates into pixel and perceptual quality.
[LG-400] When Can Old Evaluations Certify a New Model? Label-Efficient Release Decisions under Evaluator Drift
链接: https://arxiv.org/abs/2609.32267
作者: Joyanta Jyoti Mondal,Mridul Banik,Md. Shifatul Ahsan Apurba,Md Masud Al Mahmud
类目: Machine Learning (cs.LG)
*备注: 27 pages, 8 figures
Abstract:Releasing a model update requires certifying that its current-population risk stays below a threshold. Trusted labels are expensive, while a cheap evaluator, such as an LLM judge, scores every example. Reusing evaluator errors from earlier audits is tempting, but when may such evidence replace current labels? It depends on the status of history. If the errors can change invisibly, no label-free test detects the change, and every valid, useful certifier must keep buying labels at a rate we characterize; if a bound on the change is assumed, label-free certification is valid at an explicit error cost. For the middle ground, where history is informative but untrusted, we propose \emphportfolio vigilance, a sequential certifier mixing a betting expert guided by history with one that learns only from current labels; history affects only how it bets, so validity holds for any history. The contribution is not prior-informed betting or expert mixtures, but separating history that may enter validity from history that may only guide label collection. In a canonical model, accurate history shortens decisions but never raises the evidence growth rate; stale history can destroy it. On held-out CIFAR-10N and DICES-990 data, portfolio vigilance needs 0.465 (95% CI [0.327,0.575] ) and 0.740 ( [0.618,0.877] ) times the labels of a matched prediction-powered monitor, with no observed false certification, and fewer labels on all six external blocks. Under corrupted advice it stays within 8.0% of its better component, while trusting history alone costs up to 1.66 times as much. In post-confirmatory repeated-judge experiments on DICES-990 and ToxicChat, changing a fixed LLM judge’s rubric moves its scores beyond run-to-run variation; the portfolio then needs 0.790 ( [0.667,0.909] ) and 0.631 ( [0.520,0.770] ) times the labels of the matched monitor, and fewer than trusting history alone.
[LG-401] Representation Editing for Multimodal Test-Time Adaptation
链接: https://arxiv.org/abs/2609.32263
作者: Longfei Huang,Xiangyu Wu,Yang Yang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Multimodal test-time adaptation (TTA) aims to adapt a pretrained multimodal model online to distribution shift across modalities using unlabeled test data, showing broad potential in real-world applications. However, existing methods primarily focus on adjusting fused features to bridge the source-target gap, lacking explicit control over intermediate representation misalignment, which is a key driver of performance drop under distribution shift. In this work, we tackle this challenge from the perspective of representation engineering. Unlike previous TTA methods that update fusion weights in place, we propose FourIer Representation Editor (FIRE), a novel multimodal TTA approach that directly edits semantically rich intermediate representations. Specifically, we first adopt representation editors into each intermediate layer of the unimodal encoders, enabling layer-wise calibration of unimodal representations. To further enhance the diversity and stability of the low-rank editing subspaces, each representation editor performs frequency domain mixing via the fast Fourier transform to construct structured bases. Moreover, we introduce multi-level adaptation objectives to optimize these editors, jointly promoting cross-modal semantic alignment, source-target statistical alignment, and asymmetric prediction consistency. In this way, FIRE yields aligned unimodal representations for fusion and further improves prediction reliability. Extensive experiments on two widely used multimodal benchmarks under various corruption types demonstrate the superiority of FIRE over existing multimodal TTA methods.
[LG-402] Anytime-Valid LLM Leaderboards via Benchmark-weighted and Block-Factorized e-Processes
链接: https://arxiv.org/abs/2609.32248
作者: Hongfu Gao,Songxin Zhang,Zejian Xie,Bingyi Jing,Zhou Wang,Yiming Liu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large language model (LLM) leaderboards compare model capabilities by ranking models according to their mean performance on fixed benchmarks. However, variability in evaluation outcomes across runs may produce unsupported claims of model superiority on the benchmark, a risk compounded by leaderboard updates. In this paper, we propose BB-EDGE (Benchmark-Weighted and Block-Factorized e-processes for Directed Graph Evaluation), a principled framework that represents an LLM leaderboard as a directed graph whose edges certify pairwise mean-performance advantages, with anytime-valid family-wise error rate (FWER) control. Concretely, for each direction, BB-EDGE constructs an empirical-Bernstein e-process by factorizing evidence over protocol-defined blocks and assigning stakes proportional to the corresponding block weights, then applies direct e-Holm across these e -processes to certify directional advantages as edges. Theoretically, we characterize weight-proportional linear stakes under heterogeneous benchmark-average nulls and prove anytime FWER control under arbitrary within-block and cross-pair dependence. BB-EDGE further supports anytime-valid Top- k certification and simultaneous rank intervals. Extensive experiments on synthetic data and four real-world benchmarks demonstrate that BB-EDGE maintains anytime FWER control while achieving high efficiency.
[LG-403] Arithmetic Simplicity in Stochastic Gradient Methods
链接: https://arxiv.org/abs/2609.32240
作者: Bin Fu,Pengfei Gu,Jose Nunez,Fabian Vazquez
类目: Machine Learning (cs.LG)
*备注:
Abstract:A gradient descent method is arithmetically simple if the operations are limited to +,-, \times , and division x/2^t with integer t . An arthmetically simple gradient method is easy to implement in chip design. We show how to transform AdaGrad, Adam, and AdamW into arithmetically simple. AdamW is based on the recursion x_t+1=(1-\lambda\eta)x_t-\frac\eta sm_t and Adam is the special case of AdamW with \lambda=0 . We transform them into a static case with s=S(T) , where T is the number of iterations, and S(T) is a fixed function. The convergence analysis is given for a static Adam, which is also arithmetically simple. Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.32240 [cs.LG] (or arXiv:2609.32240v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.32240 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-404] “Where Can I Trust You?”: Boundary-Aware Evaluation of Surrogate Fidelity
链接: https://arxiv.org/abs/2609.32230
作者: Jackson Eshbaugh
类目: Machine Learning (cs.LG)
*备注: 19 pages, 5 figures, 10 tables, 5 appendices
Abstract:Surrogate models are commonly evaluated by how often they agree with their teacher model over an evaluation set. Local variation in this agreement is well known, but its structure and consequences are less clear. We ask whether disagreement is systematically concentrated near the teacher’s decision boundary and whether retaining that structure provides information beyond a single global score. Across several datasets and surrogate model classes, we find substantially lower fidelity near teacher decision boundaries under two different methods of identifying near-boundary examples. Moreover, conditioning agreement on confidence-defined regions improves prediction of teacher–surrogate agreement when evaluation-set composition changes, relative to the global score alone. Yet surrogates that agree equally well with the teacher both globally and near the decision boundary can respond very differently to changes selected using the surrogate itself. Finally, we show that independently trained deep teachers can agree on most predictions while identifying different examples as lying near their decision boundaries, complicating the use of those boundaries as stable reference regions for evaluating surrogates. Together, these results show that surrogate fidelity depends not only on how often a surrogate agrees with its teacher, but also on where that agreement holds and, for deep models, how stable the teacher’s decision boundary is across training runs.
[LG-405] CompassPlay: Rewarding the Proposer for Where It Moves the Solver
链接: https://arxiv.org/abs/2609.32228
作者: Sophia Xiao Pu,Ximeng Sun,Jiang Liu,Jialian Wu,Emad Barsoum,Zicheng Liu,William Yang Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:In self-play, a proposer generates verifiable tasks to train a solver. Proposer rewards often depend on the solver’s success rate, but equally difficult tasks can differ in their training value. We introduce CompassPlay, a self-play method that rewards the proposer through gradient alignment. The reward favors tasks whose solver loss gradients align with those of reference tasks representing the target capabilities. It draws on a first-order approximation to learning progress and scores each eligible task without additional solver training. Our experiments show gains in performance and training efficiency. In coding self-play with Qwen2.5-Coder-7B, a small external reference set guides task generation. CompassPlay improves average accuracy over AZR’s difficulty reward by 1.5 percentage points on in-domain coding and 2.7 on out-of-domain mathematics. In Lean4 theorem proving, CompassPlay matches the difficulty baseline’s 150-iteration cumulative coverage with 40% fewer GPU-hours.
[LG-406] DegreeSpar: Structured Degree Sparsity for Efficient Secure Transformer Inference
链接: https://arxiv.org/abs/2609.32204
作者: Yifei Cai,Zhuoran Li,Xiaozuo Shen,Hongyi Wu,Chunsheng Xin
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注:
Abstract:Secure Transformer inference protects sensitive inputs but incurs substantial cryptographic overhead, with nonlinear operations such as Softmax and GeLU becoming major bottlenecks. Existing compression methods reduce nonlinear complexity, sequence-dependent computation, or model structure through separately defined compression variables. Under aggressive compression, however, these independently optimized perturbations can accumulate: at a matched compression level, stacking representative approximation, token-pruning, and model-pruning methods reduces ViT-S accuracy from 80.20% to 76.41%. We introduce DegreeSpar, which formulates secure Transformer compression as structured sparsification over nonlinear polynomial degrees. Polynomial degree directly controls the cost of secure nonlinear evaluation, while computation-aligned zero-degree structures expose token-level and model-dimension computation as removable within the same optimization space. DegreeSpar further incorporates approximation-aware training for low-degree Softmax and GeLU, enabling aggressive degree reduction and creating the optimization headroom required for structured computation removal. Across vision and language Transformers, DegreeSpar consistently improves the accuracy-latency trade-off across model scales, tasks, and sequence lengths, achieving speedups from 2.29x to 6.63x over the corresponding baselines. Under the same network setting, DegreeSpar achieves 92.68% accuracy on BERT/SST-2 in 110.55 s, compared with 92.66% in 167.26 s for CipherPrune, the closest prior hybrid secure-inference approach. These results establish structured polynomial degree as an effective shared optimization space for secure Transformer compression.
[LG-407] SparSP: Exploiting Communication Sparsity for Sequence-Parallel Video DiTs
链接: https://arxiv.org/abs/2609.32197
作者: Desen Sun,Xinrui Zhong,Yuke Wang,Sihang Liu
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG); Performance (cs.PF)
*备注:
Abstract:Diffusion Transformers have become the dominant architecture for video generation. Their substantial computational cost motivates scaling inference across multi-GPU servers, yet efficient scaling remains challenging on commodity GPUs connected via PCIe, whose bandwidth is limited. Although sparse attention substantially reduces computation, its implications for communication remain underexplored. This paper argues that attention sparsity should be treated as a communication primitive. We present SparSP, an efficient sparse sequence parallel communication system that co-designs token distribution, communication routing, and asynchronous execution for sparse video diffusion models. First, Dependency-Aware Placement distributes sequence blocks according to diffusion models’ sparse attention patterns. Second, Demand-Directed KV Routing transfers KV blocks directly to requesting GPUs without intermediate relays. Third, a Decoupled Transfer Runtime separates communication from GPU computation to reduce resource contention and maximize effective bandwidth. Our evaluation shows that SparSP improves attention performance by 1.38- 1.5 \times , achieves an average 1.17 \times (up to 1.69 \times ) end-to-end speedup across three representative servers and three video diffusion models, and reduces communication volume by 12.54-23.05%. Moreover, we achieve an average 1.53-1.76 \times bandwidth improvement over NCCL. Subjects: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG); Performance (cs.PF) Cite as: arXiv:2609.32197 [cs.DC] (or arXiv:2609.32197v1 [cs.DC] for this version) https://doi.org/10.48550/arXiv.2609.32197 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-408] Analytic-Walk Rotary Positional Encodings for Graphs
链接: https://arxiv.org/abs/2609.32178
作者: Jiaqing Xie,Yuxin Wang,Xipeng Qiu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Rotary position encodings make attention sensitive to relative position, but extending them to graphs requires choosing how graph structure enters the rotation. Previous works assign each node a rotation from spectral coordinates, so the rotary factor between two nodes depends only on their endpoints and cannot distinguish the routes connecting them. We introduce \textitAnalytic-Walk Rotary Positional Encodings (AW-RoPE), which place the rotations on edges and sum the transported features over all walks, so contributions along different routes can reinforce or cancel. An exact variant evaluates the complete sum by a differentiable linear solve, and a sparse variant truncates it at a finite depth. We prove forward and parameter-derivative truncation bounds at fixed inputs and parameters. Both variants act on projected queries and keys, and the sparse recurrence also augments message-passing networks. Across five synthetic tasks both variants reduce nRMSE by 15 – 58% relative to the strongest baseline, and on real superpixel, peptide and OGB benchmarks the sparse recurrence attains the best mean on every dataset with Performer kernels and on twelve of thirteen datasets with GIN. Analysis shows that AW-RoPE can distinguish routes whose only cue is how two endpoints are connected, while node-wise rotary encodings cannot.
[LG-409] Efficient Support Recovery of Mixtures of Sparse Linear Classifiers with Less Measurements
链接: https://arxiv.org/abs/2609.32176
作者: Xiaxin Li,Arya Mazumdar
类目: Machine Learning (cs.LG); Data Structures and Algorithms (cs.DS)
*备注:
Abstract:The support recovery problem in mixture of linear classifiers intends to identify which features actually matter when data is generated by a mixture of several linear decision rules. In particular, the aim is to recover the support (nonzero coordinates) of l unknown k -sparse vectors from sign measurements. Each measurement is generated by selecting one of the l vectors uniformly at random, and returning the sign of its inner product with a chosen measurement vector. In this paper, we propose adaptive and non-adaptive schemes that significantly improve upon prior results by reducing the number of measurements and achieving sublinear decoding time simultaneously. In particular, our adaptive constructions substantially reduce measurements compared to existing approaches, while also lowering decoding complexity from super-quadratic to sublinear in the ambient dimension. We further provide a non-adaptive scheme that improves previous measurement bounds while maintaining efficient decoding. Overall, our approach yields a more efficient trade-off between sample complexity and decoding time for support recovery in mixture models than previously known methods. Subjects: Machine Learning (cs.LG); Data Structures and Algorithms (cs.DS) Cite as: arXiv:2609.32176 [cs.LG] (or arXiv:2609.32176v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.32176 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-410] CAFE: Counterfactual Prediction via Fast Posterior Estimation
链接: https://arxiv.org/abs/2609.32167
作者: Xinyan Han,Xiaoyu Lin,Hao Zou,Xingxuan Zhang,Bo Li,Peng Cui
类目: Machine Learning (cs.LG)
*备注:
Abstract:Counterfactual prediction estimates an individual’s outcome under an alternative intervention given their factual observations. Such outcomes are generally not identifiable from observational data without additional assumptions. Even within the class of fully observed additive noise models (ANMs), different causal graphs can generate the same observational distribution yet imply different individual counterfactual outcomes. Predictions based on a single estimated graph ignore this structural uncertainty. We therefore target a Bayesian counterfactual posterior predictive distribution that combines predictions from plausible SCMs. We introduce CAFE (\textbfCounterf\textbfActual Prediction via \textbfFast Posterior \textbfEstimation), an amortized inference framework that directly approximates the Bayesian counterfactual posterior predictive distribution. We pretrain a transformer-based model on synthetic counterfactual tasks generated from a diverse prior over ANMs. Given an observational dataset, an individual’s factual observations, and an intervention, CAFE approximates the corresponding posterior predictive distribution in a single forward pass. Experiments show that CAFE accurately predicts individual counterfactual outcomes in identifiable settings and approximates the posterior predictive distribution when structural uncertainty induced by observationally indistinguishable causal graphs exists. Strong performance in realistic manufacturing and viticulture settings further demonstrates its empirical robustness beyond the assumptions of the training prior.
[LG-411] Playing to Par: Reinforcement Learning for Provably Optimal Quadrilateral Block Decompositions
链接: https://arxiv.org/abs/2609.32146
作者: Arjun Narayanan,Per-Olof Persson
类目: Machine Learning (cs.LG)
*备注:
Abstract:A quadrilateral block decomposition of a planar domain is judged by whether it is complete, whether its elements are well shaped, and how many of its vertices are irregular. The last has a provable floor: the discrete Gauss-Bonnet identity enforces a lower bound on the total vertex irregularity of any all-quadrilateral mesh of a given domain purely based on its topology and corner angles. We train a reinforcement learning agent to build decompositions that reach this bound, which we call par. It acts directly on the mesh’s half-edge data structure through local edits, with a policy network whose convolutions follow the mesh’s own connectivity, so it applies unchanged to domains larger than any seen in training. The reward targets the floor directly, and it is sparse: random play reaches it on no domain with more than eight sides. We overcome this exploration barrier via behaviour cloning on optimal meshes that are trivial to construct, walked backward into demonstrations, before training it with PPO. On 96 held-out domains the agent produces an all-quadrilateral mesh on every one, a usable one on 95.7 on average, and a provably optimal one on 90; Gmsh’s strongest configuration at the same element count completes 51, is usable on 38 and optimal on none, and even at three to fourteen times the elements never produces a more regular mesh. On 64 domains twice the training size the agent completes all, is usable on 62, and keeps a median excess over par below one against Gmsh’s 39 at the same element count.
[LG-412] Latency-Aware Client Assignment for Parallel Split Learning With Global Sampling
链接: https://arxiv.org/abs/2609.32132
作者: Mohammad Kohankhaki,Valentin Rentschler,Anke Schmeink
类目: Machine Learning (cs.LG)
*备注: 10 pages, 5 figures. Submitted to IEEE Transactions on Emerging Topics in Computational Intelligence
Abstract:In cross-silo split learning, Parallel Split Learning with Global Sampling forms representative pooled batches when class distributions differ across clients, but ignores client delay when several clients can supply the same class. We introduce Latency Budgeted Parallel Split Learning with Global Sampling, which separates each pooled batch’s integer class target from the choice of clients that supply its examples. The flow variant formulates this assignment as an integral network-flow problem and minimizes modeled client-side completion time for the current target. The fast variant uses a greedy next-completion rule to reduce schedule-construction cost. Both preserve the target stream and use every local example once per epoch. A planning rule selects between the variants while accounting for the cost of constructing both candidate schedules. On CIFAR-10, the flow variant reduces modeled training time by 6.75%, with a 0.30 percentage-point decrease in final accuracy. On Tiny ImageNet with 20 candidate classes per client, the fast variant reduces modeled time by 16.87% and reaches all four validation targets earlier than the latency-unaware baseline. Across 405 schedule comparisons, the planning rule stays within 2% of the lower realized cost in 96.54% of cases. In our evaluation, latency-aware provider assignment reduces modeled training time without changing the prescribed class targets, while the preferred variant depends on whether assignment savings outweigh schedule-construction overhead.
[LG-413] What Should We Freeze? Guarded Freezing: Connectivity Shapes the Fine-Tuning of Pretrained Models
链接: https://arxiv.org/abs/2609.32124
作者: Leonel Aguilar
类目: Machine Learning (cs.LG)
*备注:
Abstract:When adapting pre-trained models through fine-tuning, freezing weights alone might not preserve performance, as updates elsewhere can change the inputs to the frozen core, ultimately affecting overall performance. We first analyse the case where a selected frozen core can be isolated and propose removal-value, a capacity-based score that approximates HOPE’s removal cost averaged over removal orders. We show that in VGG-8, cutting paths from trainable neurons into a frozen core makes selection using this score useful: 70% frozen preserves 5.22\pm0.51 percentage points more old-task accuracy than DEFT at similar new-task accuracy. In transformers, shared residual streams leave paths into frozen neurons open. For this case, we derive drift-value, a forward-only proxy for the output disturbance from updating each weight entry under a local update model. In language models, at 40 epochs, this policy exceeds adapted Wanda and RIA freezing scores in settings with substantial retention loss, while its differences from Fisher remain unresolved. After 160 epochs on Qwen2.5-1.5B, it retains 0.0433\pm0.0102 more than static Fisher. In DINOv3 vision-transformer adaptation to point clouds, drift-value retains 0.440 image accuracy versus 0.187 for a random mask of the same count. These results motivate Guarded Freezing: select by removal-value when incoming paths are cut, and by drift-value when they remain.
[LG-414] How Reusable Are Benchmarks with Richer Feedback?
链接: https://arxiv.org/abs/2609.32109
作者: Youssef Allouah,John Duchi
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We study whether benchmarks reliably guide model selection as developers adapt to evaluation feedback across multiple criteria. We find that the worst-case test-set size needed to estimate the best score among k adaptively chosen models, under any convex combination of the criteria, grows exponentially with the number of criteria, reaching the \Theta(\sqrtk) cost of answering k adaptive statistical queries with only O(\log k) criteria, at fixed accuracy and confidence. In attacks on multi-task large language model benchmarks with five to ten criteria, feedback restricted to nondominated task profiles produces large reused-to-held-out score gaps and frequent false winners. These results challenge a prominent explanation for prior observed reliable benchmark reuse—that developers mainly respond to convincing improvements over the current best—in rich-feedback settings, while leaving open how often ordinary model development encounters this vulnerability.
[LG-415] SAMBAR: Selective Anchoring via Method of Multipliers for Balanced Knowledge Acquisition and Retention in Vision-Language-Action Models
链接: https://arxiv.org/abs/2609.32108
作者: Aayushi Shrivastava,Xunlan Zhou,Hongrui Zhao,Ziyu Chen,Negar Mehr
类目: Machine Learning (cs.LG); Robotics (cs.RO)
*备注:
Abstract:Vision-Language-Action (VLA) models leverage large-scale pretraining to ultimately achieve generalist manipulation. Deployed VLA policies must support continual learning to acquire new tasks over time. Teaching a VLA a new task generally requires finetuning it on demonstrations of that task. However, naively finetuning on downstream tasks causes the policy to forget earlier tasks and degrades generalist capabilities. This failure is known as catastrophic forgetting. Most continual learning methods counter it by replaying data from earlier tasks. However, the old task demonstrations are not always readily available. In this paper, we introduce SAMBAR, a continual learning algorithm that prevents catastrophic forgetting during VLA finetuning without requiring access to the demonstrations of any previously learned task. We propose to cast continual learning as a constrained optimization problem and solve it with the method of multipliers. In our approach, the method of multipliers drives the policy to learn the new task without the model parameters drifting far away from their previous values. In contrast to a standard regularization penalty, the method of multipliers raises the penalty as the constraint violation accumulates by using a dual variable. We also selectively anchor the parameters critical to previous tasks to preserve past knowledge, leaving other parameters free for new task acquisition. The combination of dual variable and selective anchoring, therefore, balances knowledge acquisition with knowledge retention. We evaluate our method, SAMBAR, on the LIBERO simulation benchmark and on hardware. When sequentially finetuning on a VLA, every replay-free baseline we compare against completely forgets the first task it learned, whereas SAMBAR retains every task it has learned.
[LG-416] An Attention-Driven Heterogeneous GNN Model for Credit Card Fraud Detection
链接: https://arxiv.org/abs/2609.32106
作者: Kathiresan Jayabalan,Sethuraman Radhakrishnan
类目: Machine Learning (cs.LG)
*备注: 18 pages, 7 figures, 7 tables, original work, implementation at this https URL
Abstract:The global transition to a cashless economy has placed credit cards as the key element of digital transactions, acclaimed for their easy use, speed, and acceptance in most places. However, the growing dependence on this payment method has led to an escalation of the risks associated with credit card (CC) fraud. Detecting this type of fraud is a difficult task because the patterns are constantly changing, there is a data imbalance, and it is necessary to identify the legitimate transactions and the fraud ones at the same time. This study addresses this challenge by proposing a credit card fraud detection (CCFD) framework using a data balancing technique and a deep learning (DL) model. The proposed fraud detection model is trained and evaluated by collecting the dataset called Credit Card Fraud Detection from the Kaggle repository. As the dataset is highly imbalanced, we utilized the Synthetic Minority Oversampling Technique (SMOTE)-Tomek technique to balance the dataset. Further, the balanced dataset is classified using the Heterogeneous Graph Neural Network (HGNN) model. The HGNN model represent various transactions using a heterogeneous graph architecture and by using an attention based message passing technique, it managed to consider the complex relationships, time factors, and user behavior. The integration of SMOTE-Tomek in the model further boosted its capacity to identify fraudulent transactions, while lowering the rate of false positives. The HGNN model attained a 99.97% accuracy, a 99.48% F1-score, a 99.15% precision, and a 98.97% recall. The findings indicates that this model is effective and can be applied to real-world CCFD scenarios.
[LG-417] LLM Unlearning Evaluation with TRIAGE
链接: https://arxiv.org/abs/2609.32103
作者: Danial Ataee,Peter Triantafillou
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large language models can memorize private or harmful information, motivating machine unlearning methods that remove targeted knowledge while preserving other capabilities. However, existing evaluations rely primarily on behavioral benchmarks, which assess \emphwhether a model appears to forget but provide limited insight into \emphhow unlearning changes the model or affects related knowledge. We introduce \textitTRIAGE (\textitTripartite Representation-internal Introspection for Adjacency Gap Evaluation), a benchmark-agnostic evaluation framework for characterizing these changes. TRIAGE uses diagonal approximations of the Fisher information and Hessian to measure changes in parameter sensitivity and local curvature, and utilizes a Forget / \emphAdjacent-Retain / \emphGeneric-Retain partition to quantify an \emphadjacency gap in semantically related knowledge. Based on the magnitude and distribution of these changes, TRIAGE further classifies each algorithm’s update as \emphno-op, \emphpartially localized, \emphcollateral dominant, or \emphglobally destructive. Across 12 unlearning methods, four language models, and the WMDP, TOFU, and MUSE benchmarks, we find that methods with similar behavioral forgetting can produce substantially different internal changes and patterns of collateral damage. These signatures also vary across models and benchmarks, indicating that the effects of unlearning are not determined solely by the unlearning algorithm. TRIAGE can be applied alongside existing unlearning benchmarks to complement behavioral evaluation with a model-internal view of how unlearning reshapes the model’s parameter space and affects retained knowledge.
[LG-418] Period Segmentation in Transition Network Analysis: A Topological Data Analysis Approach
链接: https://arxiv.org/abs/2609.32094
作者: Hitoshi Inoue,Koichi Yasutake
类目: Computers and Society (cs.CY); Machine Learning (cs.LG)
*备注:
Abstract:Temporal dynamics in learning behavior can be revealed through period segmentation in Transition Network Analysis (TNA). Cristea et al. demonstrated that segmenting courses into halves and quarters reveals how learning strategies evolve and relate to academic performance. Building on this approach, we investigate whether Topological Data Analysis (TDA), specifically connected components ( \beta_0 ) change points from Zigzag Persistent Homology, can provide data-driven period boundaries that identify intervention windows. Analyzing 22 courses from the Open University Learning Analytics Dataset, we find that \beta_0 -based segmentation captures greater between-period variation than time-based segmentation (median variance ratio (VR) = 4.10 \times ; 21/22 courses show VR 1.1). Permutation tests confirmed significance ( p 0.05 ) in 18% of individual courses, with 73% showing positive improvement over random breakpoints. Only one course showed better performance with time-based segmentation. However, analysis of academic outcomes reveals a key insight: final-period behavior shows lower correlation with outcomes in \beta_0 -based segmentation than in time-based segmentation, reflecting a marked behavioral collapse where engaged learners rapidly disengage. We interpret this as evidence that final-period behavior reflects rather than causes outcomes: students who will pass maintain engagement, while those who will fail disengage. This reframes the value of \beta_0 : rather than improving prediction, change points identify intervention windows. These are periods where behavioral structure shifts and targeted support may be most effective.
[LG-419] Depth Laws for the Precision Floor of Trained Neural Networks: Amplification Residual Scaling and a Quantization-Aware Training Paradox
链接: https://arxiv.org/abs/2609.32060
作者: Ahmad S. Tarawneh
类目: Machine Learning (cs.LG)
*备注: 27 pages, 9 figures, 6 tables. Code: this https URL
Abstract:How many bits does a network need before its accuracy collapses, and how does this grow with depth? We study the precision floor, the perturbation level or bit-width at which accuracy falls halfway to chance, in MLPs, CNNs, Vision Transformers and nine pretrained language models, under post-training quantization (PTQ) and quantization- or noise-aware training (QAT). (i) A first-order theory sets the floor through one full-precision quantity, the predictive amplification G : \eta_c=\Lambda/G , and G^2 grows linearly in depth at a rate proportional to the squared residual branch scale. (ii) The predicted equality \alpha_PTQ=\rho of depth exponents holds within 95% intervals in twelve of thirteen trained architectures and in GPT-2 from 12 to 48 layers, with \Lambda=1.45\pm14% across trained architectures. (iii) Residual branches scaled by 1/\sqrtD and pre-normalisation remove the depth penalty, and each quantizer turns noise into bits at a rate fixed by its step rule, giving b_c=(\alpha/\gamma)\log_2 D+C . (iv) A QAT paradox: noise-aware training roughly doubles the tolerable noise of shallow networks, but the gain decays with depth, so the depth law steepens ( \alpha_QAT/\alpha_PTQ=1.45 - 1.47 on two datasets, ten seeds each). Decision margins, cross-layer error cancellation and heavy tails do not set the floor.
[LG-420] Graph Forward Distribution Matching for Molecular Inverse Design
链接: https://arxiv.org/abs/2609.32056
作者: Yihan Zhu,Yuhan Liu,Brett Savoie,Tengfei Luo,Meng Jiang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Achieving precise control over multiple properties without sacrificing chemical validity remains a central challenge in molecular inverse design. Existing reinforcement learning (RL) methods fine-tune graph diffusion models by treating reverse sampling as a sequential policy, using a single terminal reward to optimize hundreds of coupled decisions. They often suffer from instability, validity collapse, and limited property gains. We introduce GraphFDM (Graph Forward Distribution Matching), a new online RL paradigm for graph diffusion that performs optimization through the forward process. GraphFDM uses valid generations to define a reward-tilted target distribution jointly optimized over graph size and molecular structure for each property condition, incorporating reinforcement signals into supervised learning without storing reverse trajectories. We derive the unique optimal target, prove a condition-wise improvement guarantee, and show that the fixed graph-size prior of standard graph diffusion leaves an irreducible matching gap. In multi-conditional polymer and small-molecule generation, GraphFDM achieves the lowest MAE on every target property, with reductions of up to 53.0% relative to the strongest baselines and chemical validity above 0.99. It further generalizes to out-of-distribution property combinations.
[LG-421] Representation Learning for Exact Preimages
链接: https://arxiv.org/abs/2609.32018
作者: Konstantin Hess,Stefan Feuerriegel
类目: Machine Learning (cs.LG)
*备注:
Abstract:Modern neural predictors can model highly nonlinear maps, but many scientific and engineering tasks require reasoning in the opposite direction: given a performance or safety level, the goal is to characterize the preimage, that is, the complete set of inputs which meet the desired target level and optimize over that set. For expressive neural predictors, however, such preimages typically have no explicit representation and are expensive to recover or optimize over. This creates a fundamental three-way challenge between expressive forward prediction, accurate preimage approximation, and tractable optimization over the preimage for downstream tasks. We introduce TRIO (tractable representations for preimage learning and inverse optimization), a framework for learning representations that make these objectives compatible by construction. Our key contribution is a preimage factorization: the forward model remains expressive through nonlinear radial transformations (including neural networks), while, under inversion, each transformation reduces to a single scalar radius, which yields simple geometric level sets. This yields an explicit geometric representation that is reusable for downstream optimization over the preimage, and, for linear objectives, we show that this admits a closed-form global solution. We finally prove a universal approximation theorem which shows that TRIO can approximate any continuous forward map and its entire family of potentially disconnected, nonconvex preimages arbitrarily well. Hence, TRIO combines expressive forward modeling, exact preimage recovery, and tractable global downstream optimization over preimages by design.
[LG-422] Human Activity Recognition via Ultra-Wideband Data: A Framework for Dimensionality Reduction Pattern Discovery and Predictive Modeling
链接: https://arxiv.org/abs/2609.32008
作者: Nahid Sahel Gozin,Reza Sedaghat,Prathap Siddavaatam
类目: Machine Learning (cs.LG)
*备注: 14 pages, 9 figures. Manuscript prepared for submission to an IEEE journal in machine learning
Abstract:Recent advances in sensor technology have enabled more effective human activity recognition (HAR), particularly in real-time systems with limited computational resources. However, Ultra-Wideband (UWB) radar data remain challenging due to high dimensionality, noise, complexity, and nonlinear characteristics. This research proposes a framework to efficiently reduce data size, uncover significant patterns, and classify six activity types (Standff, Liedown, Noactivity, Sit, Stand, and Walk) from UWB signals with high accuracy. Two novel dimensionality reduction techniques are introduced in this paper. The first, Clustered Polynomial Expansion with Incremental PCA (CPE-IPCA), combines clustering and polynomial feature expansion with Incremental PCA, preserving 100% of the variance in only 50 components. The second, Post-PCA Standardization Approach (PPSA), standardizes data after PCA and retains 99.1% of the variance in 80 components, achieving superior compression and computational efficiency compared to conventional nonlinear methods. Frequent patterns are identified using Apriori and FP-Growth, which are then classified with Random Forest and a Vector Space Model (VSM). The framework achieves 100% accuracy with Random Forest on CPE-IPCA and 99% on PPSA, while VSM attains 100% precision, recall, and F1 on PPSA and near-perfect performance on CPE-IPCA (precision 1.00, recall 0.98-1.00, F1 0.99-1.00), demonstrating a fast, interpretable, and robust HAR system suitable for healthcare, assisted living, and smart environments.
[LG-423] Integrating Language Models into Listened and Imagined Speech Decoding from MEG
链接: https://arxiv.org/abs/2609.31997
作者: Maryam Maghsoudi,Sai Samrat Kankanala,Shihab A. Shamma,Sriram Ganapathy
类目: Machine Learning (cs.LG)
*备注:
Abstract:Decoding imagined speech is an important goal for brain-computer interfaces but remains challenging due to weak neural responses, low signal-to-noise ratio, and limited imagined-speech datasets. Language models provide strong contextual cues for text prediction, but how much they can help neural decoding and whether their contribution differs for decoding perceived and imagined speech remains unclear. To investigate this, we use a paired listened-imagined MEG dataset and incorporate language-model information at two stages. First, we train a contrastive neural decoder that aligns MEG representations with acoustic and contextual language representations, improving cross-subject word decoding for both listened and imagined speech. Second, at inference, we introduce a neural-constrained beam-search framework that combines neural evidence with language-model next-word probabilities. We find that imagined-speech decoding benefits more from the language model than listened-speech decoding. For Imagined speech, the best-performing balance between neural and language-model evidence shifts toward the language model, and the gain over neural-only decoding is larger. Together, these results suggest that language priors are most useful when neural evidence is weaker, making them particularly valuable for imagined-speech BCIs.
[LG-424] Can Circuit Alignment Predict OOD Generalization? NEURIPS2026
链接: https://arxiv.org/abs/2609.31996
作者: Ayan Banerjee,Abhra Chaudhuri,Josep Llados,Umapada Pal,Anjan Dutta
类目: Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2026
Abstract:Can out-of-distribution (OOD) generalization be predicted from a trained model’s weights alone, without any target-domain data? Existing representational similarity metrics (CKA, SVCCA, RSA) compare activations rather than forecast generalization. We show they are provably insensitive to structural rerouting in the computational graph, the very change distribution shift induces. We close this gap with the Circuit Alignment Score (CAS), which compares class-specific circuits across domains via graph kernels, decomposed into same-class coherence and cross-class confusion. Casting CAS as a Lebesgue integral over the domain distribution, we prove its Monte Carlo estimate recovers the ground-truth ranking of learners by OOD accuracy, with pairwise inversion error vanishing at rate O(1/M) , where M is the number of sampled domains. Across 48 learners on PACS, CAS attains 0.88 rank correlation with OOD accuracy, versus 0.58 (CKA), 0.23 (SVCCA), and 0.14 (RSA), with similar trends on other benchmarks and even against data-dependent methods, making it the first provably consistent predictor of distributional robustness requiring neither target-domain data nor labels. The code is available at: this https URL
[LG-425] Mechanistic Interpretability Reveals Shared Causal Subspaces in Brain-to-Speech Decoders
链接: https://arxiv.org/abs/2609.31992
作者: Maryam Maghsoudi,Ayushi Mishra,Sanghamitra Dutta
类目: Machine Learning (cs.LG)
*备注:
Abstract:Decoding covert speech, such as mimed or imagined, from brain activity is harder than decoding vocalized speech. Cross-modal transfer, where information from one speech form helps decode another, is a promising remedy; yet how a decoder internally represents and processes brain activity from different speech forms remains unclear. In this work, we ask: which internal neurons of a decoder carry cross-modal information, and are these neurons shared across different speech forms? To answer these questions, we leverage mechanistic interpretability, using recordings of the same sentences in vocalized, mimed, and imagined input pairs for activation patching. We insert the decoder’s internal activity for a sentence in one condition into its processing of the same sentence in another and measure the change in decoding accuracy. We find that no single neuron drives this benefit; instead, it arises from small groups of neurons, with vocalized speech as the most useful source. These groups are largely condition-specific in the early stage of the decoder but overlap in the later stage. These findings point toward more data-efficient covert speech decoders through training objectives that encourage shared later-stage representations learned mainly from vocalized data.
[LG-426] Lagrangian and Hamiltonian Neural Networks With a Dissipative System
链接: https://arxiv.org/abs/2609.31988
作者: V. Rayamajhi,J. Singal
类目: Machine Learning (cs.LG); Classical Physics (physics.class-ph)
*备注: 14 Pages, 7 Figures, 2 Tables, in review. Comments welcome
Abstract:We investigate the applicability of Lagrangian and Hamiltonian Neural Network models to a dissipative system that has explicit time dependence in its Lagrangian, Hamiltonian, and total energy. To do so we consider these neural network models for simulated systems of a harmonic one-dimensional, one-component oscillator with damping, as well as without damping for comparison. We find that both the Lagrangian and Hamiltonian approaches are able to predict the empirical physical behavior of the damped oscillator systems and to effectively ``learn’’ to varying degrees the underlying Lagrangians and Hamiltonians, as has previously been shown to be the case with undamped oscillator systems. These investigations elucidate important properties of Lagrangian and Hamiltonian mechanics, including properties that are not manifest when considering systems without explicit time dependence.
[LG-427] Understanding the Subspace Stabilization of the Hessian and Gradient Covariance Matrix
链接: https://arxiv.org/abs/2609.31983
作者: Fangshuo Liao,Anastasios Kyrillidis
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:The phenomenon of the top subspace stabilization of the Hessian matrix is an surprising and critical aspect in study of the second-order information of neural network training. Prior work argues that the top subspace of the Hessian stabilizes by measuring the overlap between the top subspaces of the step-wise Hessian, and explains this stabilization with diminishing parameter change in the late phase of training. In this paper, we define a new instability metric for the subspace evolution, and use it to detect subspace stabilization that is independent of the magnitude of parameter change. In the meantime, we observe that the gradient covariance matrix has a similar property of its top subspace to the Hessian. By using a between-class and within-class decomposition of the gradient covariance matrix, we identify an explicit form that gives a near-perfect approximation of the top- (C-1) subspace of the Hessian and the gradient covariance matrix. In the gradient flow set-up, we show that the slow evolution of the idenfied approximation is due to the separation between the outlier and the bulk eigenvalues of the Hessian matrix, thus providing an explanation to the phenomenon of the top subspace stabilization of the Hessian matrix.
[LG-428] Evasion Attacks on Cost-Utility-Based Adversarial Training for Online AutoML in IoT Networks
链接: https://arxiv.org/abs/2609.31981
作者: Chukwunonso Henry Nwokoye,Wajiha Zaheer,Khalil El-Khatib,Li Yang
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: Accepted to the IEEE Middle East Conference on Communications and Networking (MECOM 2026). Code is available at: this https URL
Abstract:As Internet of Things (IoT) networks increasingly depend on machine learning for anomaly, malware, intrusion detection, and network monitoring, such systems have become attractive targets for evasion attacks. Evasion attacks pose a major security risk because an adversary intentionally modifies input data to mislead a trained model into producing incorrect predictions while evading detection. This study evaluates the impact of black-box evasion attacks on a cost-utility-based adversarial training defense strategy in an Online AutoML context for IoT networks. Specifically, evasion attacks were applied to online learners, including Hoeffding Tree (HT), Leveraging Bagging (LB), Streaming Random Patches (SRP), Hoeffding Adaptive Tree (HAT), and Adaptive Random Forest (ARF). By developing naive and adversarially trained (AT) versions of these online learners, we generated clean and adversarial accuracies for each model. The results show that the AT versions of LB and SRP performed best, achieving the highest adversarial accuracy (0.985) and high clean accuracy (0.993) at the highest cost budget of 1.00, with a maximum accuracy reduction of only 0.8%. Finally, drift detection was conducted using the Early Drift Detection Method (EDDM).
[LG-429] Model Casting and Low-Parameter Gating: Towards More Sparsely Activated FFNs
链接: https://arxiv.org/abs/2609.31975
作者: Maria Lomeli,Antoine Groudiev,Matthijs Douze,Loïc Cabannes,Pierre Emmanuel Mazaré,François Fleuret,Maximilian Beck,Gergely Szilvasy,Naila Murray,Hervé Jégou
类目: Machine Learning (cs.LG)
*备注:
Abstract:This paper introduces model casting, a mid-training recipe that drastically sparsifies the activations within the Feed-Forward Network (FFN) layer. With this strategy, at inference time, we first compute the output of the gating matrix and, thanks to its high sparsity, we avoid computations with the two other matrices, reducing the FLOP count by up to 3x. While this theoretical speedup is an upper bound, model casting translates into significant speedups both on CPU and GPU. We then introduce LoPA Gating, a new FFN design that increases the maximum theoretical speedup. It is a low-FLOPs parameterization of the gating matrix that overcomes the 3x cap by allocating fewer FLOPs and parameters to the gating matrix, compared to the two other FFN matrices that are sparsely activated. We consider two cases: (i) we cast a pre-trained model with a sparsity inducing activation; (ii) we train with LoPA from scratch. In all settings, we significantly outperform existing pruning solutions and regular RELU-fication. For instance, at matched quality, we achieve a 3.2x FLOP speedup with LoPA Casting, against 1.6x at best for competing methods top-p and TEAL. Using dedicated kernels, we achieve an actual 3.31x speed-up on GPU at 90% sparsity, past the 3x ceiling of standard gating; RELU-fication, meanwhile, plateaus below 80% sparsity. Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.31975 [cs.LG] (or arXiv:2609.31975v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.31975 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-430] Model-Agnostic Online Certificate-Driven Calibration for Time Series Forecasting Under Distribution Shift UAI2026
链接: https://arxiv.org/abs/2609.31960
作者: Chenfeng Huang,Zixuan Ma,George Michailidis
类目: Machine Learning (cs.LG)
*备注: Oral presentation at the 42nd Conference on Uncertainty in Artificial Intelligence (UAI 2026). Published in Proceedings of Machine Learning Research (PMLR), Vol. 337, pp. 2244-2273
Abstract:Time series out-of-distribution generalization requires forecasters to remain reliable when deployment dynamics differ from training conditions due to covariate shift, concept shift, and temporal dependence. Probably Approximately Correct Bayesian domain adaptation provides computable certificates by decomposing target risk into a source risk term, a source-to-target mismatch term, and a complexity term, but standard analyses rely on independent sampling and distributional stability, assumptions that are violated in time series by serial dependence and nonstationary shift. We propose a model-agnostic online martingale Probably Approximately Correct Bayesian framework that yields finite-sample certificates under temporal dependence and distribution shift. The certificate replaces independent-sample concentration with martingale concentration that adapts to loss scale and predictable variation. We use the certificate as a surrogate regularizer for online calibration by training a gated residual Bayesian head on top of a fixed forecasting backbone, producing a corrective update that reverts to the backbone prediction when the gate is closed. Online calibration combines a source risk anchor, a posterior-shift penalty, and a time-adaptive mismatch term computed from target windows observed before forecasting. It follows a predict-then-update protocol in which outcomes become available only after forecasting and are used to update subsequent predictions. Experiments across convolutional, attention-based, and large language model-based forecasters show improved stability and accuracy under covariate and concept shift.
[LG-431] On-Policy Attention Linearization
链接: https://arxiv.org/abs/2609.31947
作者: Arian Raje,Anupam Nayak,Anthony Fei,Akaash Parthasarathy,Mohamed Abdelfattah,Gauri Joshi
类目: Machine Learning (cs.LG)
*备注: 29 pages, 12 figures
Abstract:Hybrid transformer architectures that replace most softmax attention layers with linear attention offer transformer-level quality at a fraction of the memory cost. Rather than pretraining such models, a growing body of work distills them from already trained full-attention transformers. However, these distilled models often collapse on long-context retrieval and reasoning tasks, particularly when operating in thinking mode, where the efficiency gains of hybrid architectures matter most. Since linear attention layers must compress context into a fixed-size state, their errors compound over long sequences. As off-policy distillation never teaches the student model to recover from this drift, tasks that necessitate longer sequence lengths become especially challenging. We introduce On-Policy Attention Linearization (OPAL) in which the hybrid attention student samples its own long-context trajectories and receives dense supervision from the frozen full-attention teacher. Applying OPAL to Qwen3-4B and MiMo-7B-RL-0530, we recover 87 – 94% of full-attention performance on commonsense reasoning, 100% on needle-in-a-haystack (NIAH) retrieval, and 83 – 93% on mathematical reasoning with only 3B training tokens. We achieve these results without supervised fine-tuning (SFT) or reinforcement learning with verifiable rewards (RLVR). Compared with the strongest prior linearization method, which recovers 68% of its teacher’s retrieval performance and 21.6% absolute average mathematical reasoning accuracy, OPAL fully recovers retrieval and achieves 67.6 – 72.2% on math reasoning.
[LG-432] Simple Extensions of Single-Objective Acquisition Functions and Hedge Strategies for Multi-Objective Bayesian Optimization NEURIPS2026
链接: https://arxiv.org/abs/2609.31940
作者: Haris Moazam Sheikh
类目: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
*备注: Accepted in 40th Conference on Neural Information Processing Systems (NeurIPS 2026)
Abstract:Multi-objective Bayesian optimization (MOBO) is commonly approached through specialized acquisition functions or scalarization schemes designed to explicitly account for trade-offs among non-preferential objectives. In this work, we show that such complexity might be unnecessary. We propose a framework that extends standard single-objective acquisition functions directly to the multi-objective setting through a hypervolume-based transformation. We further extend hedge strategies for acquisition functions, which are typically used only in single-objective optimization, to the multi-objective regime. Our approach requires minimal modification to existing Bayesian optimization pipelines and avoids the need for bespoke multi-objective formulations. We demonstrate how a broad class of commonly used single-objective acquisition functions and hedge strategies can be adapted in a principled manner to handle multiple objectives, while preserving their intuitive interpretation and computational efficiency. Empirically, we evaluate the proposed methods across a range of synthetic and real-world multi-objective benchmarks. Despite their simplicity, our extensions consistently match or outperform more complex state-of-the-art MOBO methods in terms of optimization performance and sample efficiency. These results suggest that effective multi-objective Bayesian optimization can be achieved by reusing and carefully extending well-established single-objective acquisition strategies, offering a simpler and more flexible alternative to existing approaches.
[LG-433] Cache-Aware Conv3D Lowering Across Embedded World-Model Decoders ECCV2026
链接: https://arxiv.org/abs/2609.31938
作者: Jiaming Zhang,Wu Yang,Shuai Tao,Wulong Liu
类目: Machine Learning (cs.LG); Performance (cs.PF)
*备注: 16 pages, 1 figure, 6 tables. ECCV 2026 workshop paper
Abstract:Generative world models can provide visual rollouts for embodied planning, yet their feasibility on edge devices depends not only on the learned model but also on how the execution runtime represents its operations. We introduce a cache-aware lowering that expresses supported causal Conv3D calls as batched spatial Conv2D operations while preserving pretrained weights, temporal-cache semantics, convolution parameters, bias placement, and output layout. Across the complete Cosmos3-Edge image-to-video pipeline on a 64-GB NVIDIA Jetson AGX Orin, the proposed route accelerates VAE decoding by approximately 7\times and reduces complete-generation latency by more than 2\times , while repeated decoder evaluations maintain complete fast-path coverage without fallbacks. The unchanged lowering also improves Cosmos3-Nano and transfers to LingBot-World’s architecturally distinct Wan2.1 VAE. A clean-device comparison against fully specialized TensorRT shows that TensorRT provides a further 1.36\times steady-state improvement, but requires substantially greater per-module and per-runtime-state AOT specialization. Same-latent BF16 and FP32 evaluations characterize the finite-precision differences introduced by the alternative execution order. Together, these results position cache-aware lowering as a lightweight runtime optimization that recovers most of the available decoder acceleration without modifying the learned models themselves.
[LG-434] GT-VLA: Target-Conditioned Trace Guidance for Generalizable Robotic Manipulation
链接: https://arxiv.org/abs/2609.31904
作者: Ninghan Zhong,Jing-Chen Peng,Sriram Vishwanath
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注:
Abstract:Vision-Language-Action (VLA) models have shown strong performance on robotic manipulation, but they often struggle to generalize to unseen tasks, configurations, and long-horizon settings. A key challenge is that VLAs overfit to training scenes and fail to follow novel language instructions. Off-the-shelf vision-language models (VLMs) often provide stronger generalization, but cannot directly control robot actions. To combine the common sense of VLMs with VLA control, we propose Guided Trace VLA (GT-VLA), a steerable framework that accepts guidance from an external generalist VLM through trace-conditioned action generation. GT-VLA uses a generalist model to identify semantic guidance for the current skill, converts this guidance into a 2D visual trace, and conditions its action policy on the resulting trace-rendered observation. This design separates semantic target acquisition, trace generation, and low-level action execution, allowing high-level guidance to propagate to robot actions. GT-VLA uses a Mixture-of-Experts architecture with skill-specific trace and action modules for robust execution. We evaluate GT-VLA on LIBERO and a physical robot platform, showing improved generalization over recent VLA baselines in both settings. The code and additional supplemental materials are available on our project website at this https URL.
[LG-435] mporalGraphLLM : Temporal Graph Neural Networks with Large Language Models for Dynamic Text-Attributed Graphs
链接: https://arxiv.org/abs/2609.31881
作者: Moran Beladev,Or Eitan,Gilad Katz,Lior Rokach
类目: Machine Learning (cs.LG)
*备注:
Abstract:Dynamic text-attributed graphs (DTAGs), where nodes, edges, and textual attributes evolve over time, are crucial in applications such as social networks, citation graphs, and knowledge graphs. However, existing approaches struggle to jointly model the temporal evolution of graph structures and the semantic richness of textual attributes. While Temporal Graph Neural Networks (TGNNs) capture evolving node relationships, they often lack contextual text reasoning. Conversely, Large Language Models (LLMs) excel in textual understanding but struggle with structured graph reasoning in temporal settings. To bridge this gap, we propose TemporalGraphLLM, a novel framework that can integrate any temporal GNN with an LLM for enhanced reasoning in DTAGs. Our approach fine-tunes LLMs using graph-time-aware instruction tuning and novel temporal GNNs injection to replace dedicated added tokens with graph embeddings. TemporalGraphLLM effectively leverages pretrained TGNNs within an LLM framework to achieve state-of-the-art performance on edge classification, link prediction, and edge-based text generation tasks. Extensive evaluation on real-world dynamic graph datasets demonstrates state-of-the-art performance. Our findings highlight the synergistic potential of LLMs and TGNNs, opening new directions for learning on evolving graphs.
[LG-436] Averag ed Mirror Descent and Dual Gradient Methods: Convergent Algorithms for Entropic Gromov-Wasserstein Problem
链接: https://arxiv.org/abs/2609.31848
作者: Joanna Mark,Gabriel Rioux,Riccardo Passegger
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:The Gromov-Wasserstein (GW) distance measures the discrepancy between metric measure (mm) spaces and identifies optimal alignments between them based solely on their intrinsic structure. Since it identifies isomorphic mm spaces, it provides a natural notion of distance for heterogeneous datasets which may admit isomorphic representations. In order to accelerate computation of GW distances, many practitioners employ entropic regularization to obtain an Entropic GW (EGW) problem. The most popular EGW solver is the Mirror Descent (MD) algorithm, which reduces EGW computations to an iterative process where an entropic optimal transport (EOT) problem is solved at each iteration. Despite its widespread use, the convergence of MD for this problem has only been established for restricted classes of costs. On the other hand, a recently proposed dual gradient method is available for general costs, but requires a choice of step size which depends on the regularization parameter. To address these two issues, we introduce Averaged Mirror Descent (AMD), which averages consecutive MD steps, and prove its convergence for arbitrary costs. Then, we establish that the dual gradient method with a fixed step size also converges for arbitrary costs at the cost of a more complicated iteration. In both cases, we also account for inexact iterations which are inescapable in practice. We compare the empirical performance of these methods across various settings and, in particular, show that AMD and the dual gradient method both converge on an example where classical MD fails.
[LG-437] Relational Compression: A Framework for Relational Fidelity in Constrained Representations
链接: https://arxiv.org/abs/2609.31816
作者: Yaniv Shulman
类目: Machine Learning (cs.LG); Information Theory (cs.IT)
*备注:
Abstract:What should a compressed representation preserve when the information of interest lies in relationships among elements rather than in the elements themselves? We formulate relational compression in the classical source-description-reconstruction sense, but with relational structure itself as the fidelity-bearing content. Each instance specifies the source relation, retained description, reconstructed or evaluated relation, fidelity criterion, and constrained resource. We use this interface to situate selected methods from graph summarization, spectral sparsification, similarity-preserving representation, and relational distillation within a common formulation while keeping their different reconstruction and resource assumptions explicit. We develop finite-codeword collision as one concrete realization. Same-codeword probability yields a relational geometry linking pair-specific alignment and separation to aggregate Rényi-2 occupancy and the spherical geometry of categorical assignments, with exact objective correspondences to squared-Euclidean centroid reconstruction and normalized graph association and cut. Graph and image studies illustrate complementary routes within the finite-codeword family: graph- and teacher-defined relational requirements act directly on equality or collision, while reconstruction acts through a joint decoder. Together, these results illustrate how distinct relational requirements can be formulated and tested within a common constrained-representation framework.
[LG-438] Medium-Term Multi-Resolution Electric Load Forecasting using Economic Data and Foundation Model
链接: https://arxiv.org/abs/2609.31806
作者: Lindas Eloi,Goude Yannig,Ciais Philippe
类目: Machine Learning (cs.LG)
*备注:
Abstract:Accurate medium-term, from a few months to a few years, electricity load forecasts are crucial for informed decision-making in power plant maintenance scheduling, load dispatch and price settlement. Being comprised between Long-Term Load Forecasting (LTLF) which uses mostly economic projections and appliances development scenarios, and Short-Term Load Forecasting (STLF) driven by weather, calendar and autoregressive patterns, Medium-Term Load Forecasting (MTLF) requires both extrapolation capabilities and variability modeling. Yet, it remains unclear if MTLF can benefit from economic indicators, and especially at which forecast horizon and resolution. To address these challenges we investigated the impact of socioeconomic data on predictions issued 1 month and up to 48 months in advance for France at monthly and daily resolution using a tabular Foundation Model (FM). A dataset covering 20 years of observations of electricity load, weather variables and economic features such as consumer price and production indices, electric vehicle counts or employment is created for the study. To avoid noisy data, we used a new feature selection pipeline, creating ensemble of expert models with diverse feature subsets, to demonstrate that selected economic covariates improve forecast skill by 20% over 2015-2025. This enhancement is steady across lead times and resolutions limiting the Mean Absolute Percentage Error to 4% for monthly granularity and 5% for daily granularity. Explainability of the models is investigated through feature and context importance. Results showed that the FM is limited in the context it leverages pointing towards potential computational savings with a reduced context, while feature importance of economic predictors grows with the forecast horizon. This suggests that including economic data in MTLF could bridge the gap with LTLF leading to seamless forecasts.
[LG-439] seq2cause: One Autoregressive Backbone Four Causal Discovery Tasks in Event Sequences
链接: https://arxiv.org/abs/2609.31801
作者: Hugo Math
类目: Machine Learning (cs.LG)
*备注: 10 pages
Abstract:Complex systems such as vehicles, patients, or genomes emit discrete event sequences whose operative question is causal, not predictive: which events cause which other events, and which cause higher-level outcomes such as failures or diseases? This question decomposes along two axes – dependency type (event \to event vs.\ event \to outcome) and causal scope (single sequence vs.\ population) – yielding four structurally distinct regimes with different identifiability conditions. No existing method addresses more than one, because all assume multi-stream structure with low vocabulary, and none scales beyond a few hundred event types. We present \textscSeq2Cause, a unified framework that resolves all four regimes through a single shared primitive: a pretrained autoregressive model repurposed as an amortized conditional independence testing engine requiring no task-specific retraining. We establish a prediction–causality duality: the model’s excess cross-entropy simultaneously bounds causal identification error across all four regimes, so that every improvement in next-token prediction tightens causal guarantees for free. On nonlinear SCMs (vocabularies up to 8,000 types) and real-world vehicle diagnostic logs ( 29 K event types, 474 failure outcomes), \textscSeq2Cause is the first method to populate all four regimes at scale with a single frozen backbone. Existing methods are either inapplicable, inaccurate, or computationally intractable in this setting. Comments: 10 pages Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.31801 [cs.LG] (or arXiv:2609.31801v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.31801 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-440] 3-D Emissions Mapping and Social Cost Estimation for US Domestic Aviation at West Coast Hubs
链接: https://arxiv.org/abs/2609.31686
作者: Hesam Shafiei Nia,Don MacKenzie
类目: Machine Learning (cs.LG); Data Analysis, Statistics and Probability (physics.data-an)
*备注:
Abstract:Existing aviation emissions inventories lack accurate trajectory data for high-resolution social cost and health impact assessment. This paper develops a 3-D emissions map by reconstructing flight trajectories for US west coast hubs to estimate regional environmental and near-airport health impacts. A physics-informed autoencoder (AE) is applied to ADS-B trajectory records for January 2025 covering US west coast hubs. The encoder combines a Convolutional Neural Network (CNN), a Bi-GRU, and a 3-D CNN with skip connection; the decoder is a Temporal Convolutional Network (TCN). It is benchmarked against a baseline-AE and cubic spline interpolation. Emissions are mapped via EUROCONTROL Base of Aircraft Data (BADA) performance tables and ICAO Engine Emissions Databank (EEDB) emission indices, with altitude corrections via Boeing Fuel Flow Method 2 (BFFM2). Social costs are quantified for all flight phases, with health impacts assessed for Landing and Takeoff cycles within 50 km of each hub. The proposed AE model outperforms both a TCN-AE and cubic spline interpolation across 5% to 50% missing rates. Monetizing the emissions inventory shows NOx produces a small net cooling effect in direct climate forcing, while accounting for 99.7% of monetized air-quality and health cost despite being under 0.4% of CO2 by mass, making it the dominant health-cost driver. To our knowledge, this is among the first studies combining AE-based trajectory reconstruction with separate spatial-temporal feature encoding and altitude-based emissions modeling to produce a regional aviation emissions inventory for air quality, climate impact and population exposure. The resulting emissions map and social cost estimates provide quantitative context for environmental impact assessment and near-airport health policy evaluation for US domestic aviation.
[LG-441] NanoForecast v0.5: Competitive Time Series Forecasting Through Training Pipeline Optimization
链接: https://arxiv.org/abs/2609.31669
作者: Gautam Kishore
类目: Machine Learning (cs.LG)
*备注: 17 pages, 3 figures, 4 tables. Code and checkpoints: this https URL
Abstract:We present NanoForecast v0.5, a 6.5M-parameter forecaster that competes with models 31x its size (TimesFM, 200M parameters) after training pipeline fixes and no architecture change. Retraining the v0.3 architecture with corrected loss-scope handling, tensor shape alignment, and wider augmentation coverage cuts overall Mean Absolute Scaled Error by 43.8% under one fixed protocol (MASE 3.030 to 1.704) on the same data and compute budget. NanoForecast v0.5 beats TimesFM on all three ETT datasets (MASE 0.676/1.110/0.287 vs. 0.705/1.360/0.545) and on exchange rate (4.317 vs. 4.383); TimesFM keeps a clear lead on the high-cardinality electricity and traffic sets. Against PatchTST (15M+ parameters, official configuration), v0.5 wins all three ETT sets. Training takes about 12 hours on a single cloud GPU (NVIDIA T4, Google Colab) and inference needs no GPU (measurements in this paper are on an Apple M4 CPU). We release all code, pretrained checkpoints, and evaluation framework under Apache 2.0 at this https URL
[LG-442] Beyond the Graph: An Adaptive Meta-Learner Fuses Explainability Weather and Dynamics for Robust Bus ETA Prediction
链接: https://arxiv.org/abs/2609.31667
作者: Pratham Payra,Jagadish
类目: Machine Learning (cs.LG)
*备注:
Abstract:Accurate bus Estimated Time of Arrival (ETA) prediction is vital for urban mobility, passenger satisfaction, and transit efficiency, yet existing models falter against nonlinear spatiotemporal dynamics, data sparsity, and factors such as weather. This paper proposes HYB(nm), an adaptive hybrid ensemble framework that dynamically fuses five complementary models - a historical baseline (MST-AV), periodical temporal pattern analysis (GDRN-DFT), Koopman Neural Operators for nonlinear dynamics (KOOP-NET), weather-integrated feature-engineered neural networks (FENN), and real-time graph convolutional networks (MGCN) - via a meta-learner attuned to real-time context. Evaluated on GPS and weather data from three Kolkata bus routes comprising more than 4,000 trips, the framework leverages the individual strengths of its components (for example, the low-latency explainability of MST-AV, the weather resilience of FENN, and the network-dynamics capture of MGCN) to deliver the superior robustness of HYB(2), state-of-the-art accuracy rivalling leading graph neural networks, and balanced trade-offs in stability and efficiency across prediction horizons and operating conditions. The extensible HYB(k) architecture equips transit agencies with flexible tools, ranging from economical single models to tailored high-fidelity hybrids, advancing predictive, equitable urban transport.
[LG-443] Cross-Material Support Transfer for Core-Loss Prediction Under Waveform Covariate Shift
链接: https://arxiv.org/abs/2609.31659
作者: Cong Yao,Chunye Gong
类目: Machine Learning (cs.LG); Materials Science (cond-mat.mtrl-sci)
*备注: 10 pages, 7 figures, 5 tables
Abstract:Power magnetic materials are characterized on the sinusoidal and triangular waveforms that excitation hardware conveniently produces, whereas deployed converters expose cores to trapezoidal, PWM-shaped flux trajectories, so loss models must predict exactly where their training data are thinnest. The final test of the MagNet Challenge embeds a deliberately extreme instance of this characterization-deployment mismatch: for material D, trapezoids form 16.4% of the test set but only 1.4% of the training set. The 95th-percentile relative error, hereafter p95, of the best submission, built on sequential transfer learning, stalled at 15.9%, the worst among the five materials. This paper shows that the obstacle is missing information under covariate shift rather than class imbalance, and that the missing support can be borrowed from sibling materials instead of being extrapolated. Controlled experiments first refute the imbalance reading: four standard remedies fail, and raising the trapezoidal share to the test-set level degrades accuracy further. The proposed material-identity support transfer, MIST, then trains one 2784-parameter predictor jointly on all five challenge materials. Material identity enters through feature-wise linear modulation, or FiLM, the scarce material’s true-label loss is reweighted, and material D receives no fine-tuning, so that the bias of its trapezoid-free training set is never re-installed. MIST lowers the five-seed material-D p95 from 20.39+/-2.03% to 12.38+/-0.92% and the trapezoidal-class p95 from 37.4+/-8.8% to 15.16+/-1.69%, surpassing the best submission with one-sixth of its parameters and no fine-tuning stage; removing material identity at matched capacity inflates the error by an order of magnitude. These results argue that scarce materials should be characterized jointly with their siblings.
[LG-444] Product-Aware Deterministic Rounding for Quantized Matrix Multiplication
链接: https://arxiv.org/abs/2609.31641
作者: Piyush Sao,Narasinga Miniskar,Pedro Valero-Lara,Keita Teranishi,Sudip Seal
类目: Machine Learning (cs.LG); Data Structures and Algorithms (cs.DS); Information Theory (cs.IT)
*备注: 30 pages, 7 figures
Abstract:Scalar rounding decisions interact through matrix multiplication. We study deterministic product-aware rounding after scales, clipping bounds, and grids are fixed, with each active scalar choosing between adjacent levels. For dynamic activation rounding, null-space reduction preserves the relaxed product while leaving at most r fractional decisions, where r is the rank of the active gap-weighted weight block. Conditional-expectation completion gives a deterministic polynomial-time algorithm with squared product error at most \mathrmOPT_\mathrmdyn+r\nu_\max^2/4 , where \mathrmOPT_\mathrmdyn is the best admissible error and \nu_\max is the largest row norm of that block. For reusable static weights, the exact expected product-loss metric is the uncentered input second moment with fixed output bias; free bias recalibration yields the centered covariance. Exact optimization is NP-hard even at rank one. In balanced blocks with K=1024 and r=16 , conditional- expectation completion attains a dither-normalized median error of 0.010 , compared with 0.899 for round-to-nearest. Clipping-aware initialization reduces median normalized error by a factor of 43.4 at ten-percent clipping. Held-out Digits experiments show that retaining the input mean or correcting the output bias improves median product error over round-to-nearest in all four tested bit-width and calibration-size settings. Comments: 30 pages, 7 figures Subjects: Machine Learning (cs.LG); Data Structures and Algorithms (cs.DS); Information Theory (cs.IT) Cite as: arXiv:2609.31641 [cs.LG] (or arXiv:2609.31641v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.31641 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Piyush Sao [view email] [v1] Fri, 11 Sep 2026 17:52:57 UTC (354 KB) Full-text links: Access Paper: View a PDF of the paper titled Product-Aware Deterministic Rounding for Quantized Matrix Multiplication, by Piyush Sao and 4 other authorsView PDFHTML (experimental)TeX Source view license Current browse context: cs.LG prev | next new | recent | 2026-09 Change to browse by: cs cs.DS cs.IT math math.IT References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) IArxiv recommender toggle IArxiv Recommender (What is IArxiv?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab) Major funding support from
[LG-445] Symmetry-quotient Flatness and Generalization ICONIP2026
链接: https://arxiv.org/abs/2609.31634
作者: Taiki Miyagawa
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: Accepted to 33rd International Conference on Neural Information Processing (ICONIP 2026) as an extended abstract. This version is the full paper
Abstract:This paper develops a theorem-level pipeline in symmetry-quotient settings: quotient linear stability implies quotient flatness, quotient flatness implies input smoothness, and input smoothness yields generalization under local covering assumptions. Flatness is often associated with generalization, and Stochastic Gradient Descent (SGD) is frequently viewed as implicitly biased toward flat solutions. However, standard flatness measures are typically defined in the raw parameter space and are therefore not invariant under function-preserving symmetries such as positive rescaling. We develop a symmetry-aware theory of quotient flatness, quotient linear stability, input smoothness, and generalization on quotient spaces of neural-network parameters. For square loss and models equipped with function-preserving group actions, we define quotient flatness as the trace of the Hessian of the empirical loss on the regular quotient manifold. We show that quotient flatness controls input smoothness through a quotient-space analogue of the flatness-to-smoothness argument. We also prove that one-step mean-square quotient linear stability of the linearized SGD dynamics implies an explicit quotient-flatness bound in terms of the batch size and learning rate, and extend this analysis to higher-order tensor moments. Finally, under local covering and boundedness assumptions, we derive population generalization bounds in terms of quotient flatness and, consequently, in terms of quotient linear stability.
[LG-446] Enhancing generalization in endwall film cooling prediction: Incorporating the superposition principle into transformer-based neural operators DATE
链接: https://arxiv.org/abs/2609.31633
作者: Qineng Wang,Liming Song,Tianyuan Liu,Zhendong Guo
类目: Machine Learning (cs.LG); Computational Physics (physics.comp-ph); Fluid Dynamics (physics.flu-dyn)
*备注: Author manuscript updated to align core methods and results with the published article; 29 pages, 18 figures, 5 tables
Abstract:In this study, a physics-enhanced neural operator framework is proposed to enhance the generalization prediction ability of the cooling layout of a turbine endwall with variable number of film holes. Specifically, inspired by the film cooling superposition principle, we propose a film cooling prediction model, namely superposition-based deep neural operator (SDNO), that divides the endwall temperature field prediction into two stages. In the first stage, the cooling layout of a turbine endwall is divided into several sub-parts with randomly assigned film holes, and a Transformer-based neural operator network, namely Calculate Net, is designed to predict the temperature field of each sub-part. Then, in the second stage, another neural operator network, i.e., Super Net, is trained to combine the temperature fields predicted by Calculate Net for each sub-part and obtain the superposed temperature field of the full cooling layout. Additionally, instead of directly taking the film cooling contours as pixel plots, a signed distance function (SDF) which is sensitive to the variable locations of cooling holes, is designed to encode the location information of cooling holes. Furthermore, the proposed endwall film cooling prediction model is trained with the samples that changing the number of film holes from 1-5 with variable locations. Then, the trained prediction shows excellent generalization prediction ability, which can accurately predict the film effectiveness of the cooling layout with 10-20 film cooling holes that are unseen in the training samples. The proposed SDNO also improves prediction accuracy relative to the fully supervised baseline. With the above, the effectiveness of our proposed prediction model has been well demonstrated.
[LG-447] EEGAgent Bench: Benchmarking LLM Agents on Short- and Long-Horizon EEG Analysis
链接: https://arxiv.org/abs/2609.31632
作者: Huyu Wu,Weining Weng,Yuchen Liu,Yiqiang Chen,Yang Gu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Electroencephalography (EEG) analysis is evolving from short-segment classification toward long-horizon interpretation that demands iterative evidence accumulation, multi-step reasoning, and coordinated use of specialized signal-processing tools. Although large language models (LLMs) have recently shown promise as autonomous agents for EEG analysis, existing EEG agentic evaluations remain fragmented, covering limited tasks over narrow temporal horizons with inconsistent protocols, and providing no comprehensive assessment of agents’ reasoning, tool-use, and workflow construction capabilities. To address this gap, we propose \textbfEEGAgentBench, a unified benchmark for systematically evaluating LLM agents on short- and long-horizon EEG analysis. EEGAgentBench spans six representative EEG applications ranging from knowledge question answering to sleep staging. It encompasses signal durations from 2 seconds to nearly 23 hours, with prediction targets ranging from class labels to event intervals and epoch-level sequences. This design supports unified evaluation across knowledge reasoning, short-horizon interpretation, long-horizon event detection, and sequential understanding. The benchmark further provides 10 deterministic EEG analysis tools that expose only task-relevant signal measurements. Agents must therefore select tools autonomously, accumulate evidence iteratively, and construct multi-step workflows. For evaluation, we benchmark 29 frontier LLMs from 15 model families. Results demonstrate that EEGAgentBench effectively distinguishes agent capabilities beyond model scale and inference cost, while revealing substantial limitations of current LLM agents in long-horizon EEG analysis, particularly in sustained evidence accumulation and multi-step reasoning.
[LG-448] OMP-MoE: Efficient Expert Pruning for Mixture-of-Experts LLM s via Orthogonal Matching Pursuit
链接: https://arxiv.org/abs/2609.31631
作者: Dezhi Li,Lujun Li,Qiyuan Zhu,Hao Gu,Bei Liu,Sirui Han,Yike Guo
类目: Machine Learning (cs.LG)
*备注: Work in progress, revisions ongoing
Abstract:Mixture-of-Experts (MoE) models enable efficient scaling of large language models but face critical deployment challenges due to massive memory requirements. Existing pruning methods either incur prohibitive search costs or neglect the dynamic interdependencies between experts. To address these challenges, we present OMP-MoE, a novel training-free compression framework for reducing expert redundancy in MoE-based LLMs. Based on observations of expert contribution patterns, we reformulate the pruning problem as a sparse signal reconstruction task solved through Orthogonal Matching Pursuit. Specifically, our method first treats individual expert contributions as dictionary atoms and selects experts that greedily minimize reconstruction error with linear computational complexity. Then, we optimize cross-layer expert allocation through a water-filling strategy that accounts for both reconstruction quality and routing stability. Finally, we introduce OMP-MoE†, an adaptive inference mechanism that dynamically adjusts expert activation based on energy prediction. Comprehensive experiments on Qwen, DeepSeek-V2, GPT-OSS, and Mixtral MoE demonstrate consistent improvements over existing methods at 25-50% pruning ratios. For Qwen3-30B-A3B at 50% compression, we retain 93.3% of original performance, achieving 33 \times faster search and 1.55 \times inference speedup. Codes will be available after acceptance.
[LG-449] Replay in the Silent Degrees of Freedom: Continual Learning Without an Offline Phase
链接: https://arxiv.org/abs/2609.31630
作者: Zhang Yanhai
类目: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE); Machine Learning (stat.ML)
*备注: 25 pages, 10 figures. Code: this https URL
Abstract:Replay-based continual learning almost always consolidates in a dedicated offline phase or by interleaving replayed samples with the input stream, whereas brains also consolidate during wakefulness through local sleep, brief use-dependent off-periods of individual circuits. We ask whether a network trained by local, biologically constrained rules can consolidate with no offline phase at all. An isolation rule confines replay updates to hidden synapses invisible to the current input under k-winner-take-all dynamics, with optimiser state advanced only inside the mask; a refractory rotation rule makes units that have just fired sit out the next competition, widening the consolidable set; a homeostatic pressure and a relative-novelty gate decide when replay bursts fire and when rotation runs. This inverts the usual direction of non-interfering continual learning: the hidden computation on the current input is held invariant (exactly on the proven channels, and for all but 0.3% of waking samples per update elsewhere) while past memories are written into the degrees of freedom the current batch leaves unused. On class-incremental split-MNIST the system reaches 91.6±0.3% with no offline phase, at or above the best offline-night schedule on two held-out splits, tied with DER++ and above experience replay, ER-ACE, A-GEM and unmasked local replay; in a single pass it leads DER++ (91.8% against 90.1%) while the night falls to 76.9%. The advantage is largest at small buffers and gives way to the backpropagation references at large ones; on split CIFAR-10 the system leads offline rehearsal and experience replay but trails ER-ACE and DER++. Rotation carries most of the gain; isolation adds the invariance guarantee. The mechanism is not tied to the local rule: under the same schedule a backpropagation network with k-WTA hidden layers gains from rotation, and isolation is again free on top of it.
[LG-450] Learned Preconditioning for a Primal-Dual Interior-Point Method
链接: https://arxiv.org/abs/2609.35665
作者: Abhinav Madabhushi,Jialin Liu,Minxin Zhang
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:
Abstract:Interior-point methods (IPMs) are among the most widely used algorithms for constrained optimization, yet their Newton-based search directions require costly second-order information and large linear-system solves. Learning to optimize offers cheaper updates learned from data, but the singular behavior of logarithmic barriers near constraint boundaries makes IPMs highly sensitive to perturbations, complicating both warm starting and learning reliable updates. We introduce pdLIP, an IPM for smooth nonlinear programs that integrates learned preconditioning with pdProj, an all-shifted primal-dual projected-search IPM. A shared coordinate-wise recurrent network predicts a positive diagonal preconditioner that scales the right-hand side of the reduced Newton system for the primal step, and the remaining slack and multiplier directions are recovered analytically. The learned iterations avoid Hessian evaluations and Newton-system solves, using only first-order and coordinate-wise operations amenable to GPU parallelization. Training is self-supervised, with a loss based on a penalty-barrier merit function and the residual of perturbed optimality conditions, requiring neither target directions nor precomputed solutions. Primal and dual shifts mitigate the barrier’s sensitivity to perturbations near constraint boundaries, enabling effective warm starting. Across four classes of 200-dimensional convex and nonconvex constrained problems, pdLIP warm starts reduce pdProj refinement iterations by 63-67% compared with cold starts at the same KKT residual tolerance of 10^-8 , with negligible warm-start generation cost relative to the subsequent pdProj solve. Improvements persist on box-constrained QPs with 1000 variables and extend to applications including portfolio optimization, support vector machines, and a nonlinear control example.
[LG-451] Elicitation and Decision Geometry in Single-Index Bandits
链接: https://arxiv.org/abs/2609.35622
作者: Sakshi Arya,Cheng Soon Ong
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We study two-arm contextual bandits with arm-specific single indices and a shared unknown monotone link. Monotonicity makes the optimal action depend only on the contrast between the index directions, hence arm-specific reward functions need not be estimated. We introduce Natural Boundary Learning (NBL), a greedy procedure that uses a sequential Stein contrast to learn the optimal boundary directly, without estimating the reward functions or the common link. We characterize the local Riemannian dynamics of NBL through a decision stability coefficient balancing arm separation, link geometry, and the context distribution. We show that this stability is connected to the elicitation geometry of the underlying convex potential. Under local decision stability, NBL contracts toward the optimal boundary and achieves O(\log n) expected regret. Numerical experiments illustrate the predicted stability regimes and compare NBL with a parametric greedy benchmark under link misspecification.
[LG-452] Learning Conditional Expectation Operators via Functional Newton Updates
链接: https://arxiv.org/abs/2609.35598
作者: Thiago Ramos,Alek Fröhlich,Daniel Perazzo,Massimiliano Pontil
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We introduce the Functional Spectral-Newton Method (FSNM) for learning the leading singular structure of a conditional expectation operator without fixing a basis or reproducing kernel Hilbert space. FSNM fits a low-rank representation of the centered joint-to-product density ratio kernel by alternating functional Newton updates. Each update reduces to a preconditioned regression, which we approximate with vector-valued regression trees in a stagewise boosting procedure. At the population level, we establish descent and an O(1/T) best-iterate block-stationarity rate under a relative weak-learner accuracy condition, and show that every nondegenerate local minimum over the full centered L^2 spaces is a globally optimal rank- d approximation. Synthetic experiments show that FSNM recovers a low-rank density ratio and its leading spectral structure, and that the same learned kernel can answer multiple conditional queries without refitting.
[LG-453] Multi-Task Learning of Conditional Mean Operators: applications to dynamical systems and uncertainty quantification
链接: https://arxiv.org/abs/2609.35429
作者: Sami Chemlal,Thibaut Germain,Rémi Flamary,Vladimir R. Kostic,Karim Lounici
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Estimating conditional statistics and learning representations of a population of conditional distributions are central problems in many data-driven applications, including uncertainty quantification and dynamical systems analysis. Conditional mean operators (CMOs), a class of linear operators between function spaces, resolve these objectives by providing access to a broad class of conditional statistics. However, existing methods typically estimate each CMO independently or constrain it to prespecified function spaces, thereby preventing the exploitation of shared structure across related distributions. In this work, we posit that related CMOs share finite-dimensional input and output function spaces, and are specialized for each task with a linear operator mapping these spaces. Based on this hypothesis, we introduce MTL-CMO, a multi-task framework that jointly learns shared function spaces and task-specific operators across multiple datasets. We further introduce T-CMO, a transfer learning method that reuses the shared spaces to estimate, in closed form, the operator of a new conditional distribution. We establish statistical guarantees quantifying the benefits of jointly learning the shared function spaces. Our experiments demonstrate that learning shared function spaces improves uncertainty quantification across a broad range of conditional distributions and, when applied to Langevin and plasma dynamics, yields compact representations of complex dynamics that retain physically meaningful information and enable parameter identification.
[LG-454] Convex Optimization Is Free When Accuracy Is Expensive
链接: https://arxiv.org/abs/2609.35418
作者: Arthur Paing,Arthur Jacot
类目: Optimization and Control (math.OC); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:This paper studies convex optimization when the gradient cannot be evaluated exactly, but only approximated by a hierarchy of algorithms whose compute grows like \delta^-\gamma in the accuracy \delta . When \gamma2 , falling into the Harder-Than-Monte-Carlo (HTMC) regime, the price of accuracy outruns the variance reduction that Monte Carlo would buy and we show that minimizing a loss function costs no more, up to a factor depending only on \gamma , than a single evaluation of its gradient at the accuracy the problem demands. A randomized multilevel oracle replaces the deterministic approximation of accuracy \delta by an unbiased estimator of it, whose variance \sigma^2 becomes a second, independently priced dial: the cost of one call drops from \delta^-\gamma to \delta^2-\gamma\sigma^-2 . Plain inexact gradient descent driven by that oracle reaches loss \varepsilon at expected compute \Theta(\varepsilon^-\gamma) in the convex case, against \Theta(\varepsilon^-(\gamma+1)) for the same method run at a fixed accuracy: randomization buys a full power of \varepsilon . Under \mu -strong convexity the exponent halves, to \varepsilon^-\gamma/2 , because the iterates settle at a noise floor and the bias budget relaxes accordingly. Both bounds are independent of the step size, and hence of the smoothness constant, and we show that the cost is a functional of the underlying gradient flow rather than of any discretization of it.
[LG-455] Simulation-Based Inference for Plate Reverb System Identification
链接: https://arxiv.org/abs/2609.35295
作者: Dylan Sechet,Marc Evrard,Matthieu Kowalski
类目: Audio and Speech Processing (eess.AS); Machine Learning (cs.LG)
*备注:
Abstract:We address Task A of the 1st DAFx Parameter Estimation Challenge, which aims to retrieve the physical parameters of a plate model from an impulse response. To do so, we use the Simulation-Based Inference (SBI) framework, in which we train a neural network to estimate a density over plate parameters given an impulse response, using a dataset generated by the simulator. Inference for a new impulse response then requires only a forward pass through the network, without involving the simulator. For each test observation, we fine-tune a specific network: additional simulation rounds are performed by sampling parameters from the current estimated distribution, simulating the corresponding impulse responses, and fine-tuning to produce the specialized network.
[LG-456] A Hierarchy of Entropy-Shapley Games for Multivariate Predictive Uncertainty
链接: https://arxiv.org/abs/2609.35217
作者: Niklas Koenen,Claudia Battistin,Jeriek Van den Abeele,Martin Jullum
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Modern probabilistic machine learning models increasingly produce multivariate outputs with complex dependence structure, from multi-step time-series forecasts to sample path predictions. Understanding which input features drive the predictive uncertainty is important for risk-aware decisions, model diagnostics, and deciding whether the uncertainty should be mitigated or hedged against. This attribution problem requires a choice of how dependencies between output components are treated. Existing approaches reduce the output to a scalar through aggregation or projection before attribution, thereby obscuring whether features affect marginal uncertainty, dependence structure, or both, while component-wise analyses can miss dependence effects entirely. We close this gap by introducing a hierarchy of three entropy-based Shapley games that make this output-side choice explicit for any ordered multivariate outcome, ranging from per-component marginal entropy to fully joint entropy. The hierarchy isolates a cross-component attribution term that captures how each feature shifts the dependence between output components, a quantity invisible to component-wise methods. We establish a chain-rule decomposition of the joint attribution and characterize the cross-component term through conditional total correlation, providing both closed-form and sample-based estimators. Finally, we demonstrate how the framework captures differences in learned joint structure across probabilistic models from distributional regression to a zero-shot time series foundation model.
[LG-457] Uncertainty Quantification in Cardiac Model Personalisation from Ultrafast Ultrasound
链接: https://arxiv.org/abs/2609.35214
作者: Camilla Ferrario,Maelys Venet(CHU Bordeaux),Olivier Villemain(CHU Bordeaux),Maxime Sermesant(EPIONE)
类目: Quantitative Methods (q-bio.QM); Machine Learning (cs.LG); Medical Physics (physics.med-ph)
*备注:
Abstract:Cardiac model personalisation requires inferring mechanical parameters that are not directly measurable in vivo. Ultrafast ultrasound shear wave elastography (SWE) enables non-invasive tracking of myocardial stiffness dynamics over the cardiac cycle, providing a target for personalisation. However, mapping these observations to subject specific model parameters remains ill-posed, as multiple parameter sets can reproduce the same stiffness dynamics. We formulate SWE-informed personalisation as a statistical inference problem using simulation-based inference (SBI). Using a subject-adapted 0D cardiovascular model and neural posterior estimation, we estimate model-conditional posterior distributions over active stiffness scale k0, contraction rate kATP, and relaxation rate kSR, conditioned on SWE-derived curve features and subject specific context. Among six healthy volunteers, four passed objective prior-support diagnostics and were retained for quantitative posterior analysis. Curve-level RMSE against the observed SWE target decreased from 12.61 \pm 5.55 kPa for the prior predictive median to 1.14 \pm 0.38 kPa for the posterior predictive median, an 89.7 \pm 4.2% reduction. Posterior analysis revealed parameter-specific uncertainty, k0-kATP compensation, weaker constraint of kSR, and the importance of prior-predictive diagnostics for assessing whether each subject is represented within the modelled SWE feature space. These results support SBI for uncertainty aware SWE-based personalisation, while identifying prior support and forward-model adequacy as key diagnostics.
[LG-458] Coordinated Lane-Level Variable Speed Limits and Ramp Metering for Successive Weaving Segments Considering Merging/Diverging Risks: A Hybrid Model Predictive Control and Multi-Agent Reinforcement Learning Approach
链接: https://arxiv.org/abs/2609.35152
作者: Guodong Ma,Baofeng Sun,Wenyu Yang,Zhihong Yao
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:
Abstract:Successive weaving segments (SWSs) on urban expressways are bottlenecks prone to recurrent congestion and collisions, requiring fine-grained active traffic management (ATM). Existing approaches struggle to balance the adaptive performance of data-driven optimization with the resilience and transferability of model-based control. We propose a hybrid framework to coordinate lane-level variable speed limits (VSLs) and ramp metering across SWSs. First, we reconstruct L-METANET, a lane-level macroscopic traffic flow model that captures free and forced lane changes. Second, we combine XGBoost-SHAP with a random-parameters binary logit (RPBL) model to derive analytical equations for merging and diverging collision risks and formulate system cost and reward functions. Third, we develop MPC-STMAPPO, a hierarchical controller integrating model predictive control (MPC) and multi-agent reinforcement learning (MARL). Its upper MPC layer uses L-METANET for long-horizon rolling optimization and generates baseline commands; its lower spatiotemporal MAPPO (ST-MAPPO) layer, enhanced with Mamba cells and graph attention, produces residual actions for short-horizon adjustment. Real-world experiments on the 18-km Eastern Expressway in Changchun, China, show that L-METANET accurately reproduces lane-changing-induced flow redistribution and capacity drops, with state evolution aligned with ground truth. XGBoost-SHAP-RPBL achieves AUCs above 0.80 in most tasks, outperforming conventional logit models. MPC-STMAPPO converges faster and performs better across multiple metrics than MPC- and MARL-based baselines. Under randomly fluctuating demand, it also significantly outperforms pure MARL in generalization, demonstrating strong potential for industrial deployment.
[LG-459] Continuous Variational Synthesis
链接: https://arxiv.org/abs/2609.35083
作者: Alan N. Amin,Mattia G. Gollub,Andrei Slabodkin,Elizabeth B. Wood,Eli N. Weinstein
类目: Biomolecules (q-bio.BM); Machine Learning (cs.LG)
*备注:
Abstract:Biological machine learning was long bottlenecked by the ability to synthesize designed DNA. Variational synthesis models control chemical reactions to physically manufacture quadrillions of designed sequences in DNA. However, training these generative models is challenging: constraints on chemical synthesis can force many parameters into a discrete space, limiting the ability to pre-train and fine-tune. In this article we train ``free’’ variational synthesis models using stochastic gradient descent in continuous space, and then discretize with post-training quantization to impose hardware and wetware constraints. This enables variational synthesis models to satisfy stringent reward criteria, while still synthesizing diverse designs, achieving a strictly dominating quality-diversity Pareto frontier. We demonstrate by training variational synthesis models of enzymes, peptides, antibody CDRH3s, and regulatory DNA elements. In silico performance is maintained in vitro.
[LG-460] Perceptual Quality Loss or Loss of Perceptual Quality? ICASSP27
链接: https://arxiv.org/abs/2609.35054
作者: Danilo de Oliveira,Tal Peer,Maurício do V. M. da Costa,Timo Gerkmann
类目: Audio and Speech Processing (eess.AS); Machine Learning (cs.LG)
*备注: Submitted to ICASSP 27
Abstract:Contemporary deep speech enhancement (SE) models are often trained with specific auxiliary terms in the loss function as a way to improve their performance in terms of perceptual metrics. Nevertheless, a higher score on a perceptual metric does not necessarily correlate with an improved listening experience. Through objective and subjective experiments, we assess the performance of SE models trained with two different types of auxiliary PESQ loss terms. The numerical evaluation on a suite of standard metrics suggests that, while models optimized for PESQ naturally obtain higher PESQ scores in the test set, for most other metrics the scores do not significantly change. In some cases, the PESQ loss even results in worse PESQ scores on mismatched data. A formal listening experiment reveals that the models without a PESQ loss were generally preferred over models that include it, across all settings. Finally, we analyze the relative importance of PESQ in the composite metrics CSIG, CBAK and COVL, and find that PESQ dominates all of them. Our study highlights the perils of over-reliance on PESQ and stresses the importance of a complete evaluation procedure for SE.
[LG-461] GUIDE-FBO: Guidance via Uncertainty Intervention and Distributional Exchange for Federated Bayesian Optimization
链接: https://arxiv.org/abs/2609.35038
作者: Jintao Wei,Chenxi Li,Songhao Wang
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Federated Bayesian Optimization (FBO) enables distributed agents to collaboratively optimize expensive black-box objectives without sharing raw local observations. However, effective knowledge transfer remains challenging under communication constraints and task heterogeneity. We propose GUIDE-FBO, in which agents exchange compact distributions over the locations of their respective optima inferred from local Gaussian process (GP) posteriors, rather than raw observations, query points, or surrogate parameters. The server merges and reweights these distributional components before returning a subset to each agent. Each agent then constructs a Federated Interventional GP (FI-GP), which preserves the local posterior mean and spatially rescales its covariance for local decision making. For the upper confidence bound (UCB) instantiation, GUIDE-UCB, we prove that any bounded FI-GP uncertainty intervention preserves the leading-order cumulative regret rate of standard GP-UCB. When the transferred distributions place greater support near an optimum than in a suboptimal region, selecting the latter requires greater local posterior uncertainty. Experiments on 12 synthetic benchmarks and three real-world optimization tasks show that GUIDE-FBO remains effective across settings ranging from homogeneous to severely heterogeneous. Ablation results highlight the importance of spatially localized uncertainty intervention, while the communication analysis shows that GUIDE-FBO exchanges only compact distributional messages.
[LG-462] Simulation-Based Quantum System Inference with Neural Posterior Estimation
链接: https://arxiv.org/abs/2609.34995
作者: Hang Zou,Anton Frisk Kockum,Martin Rahm,Simon Olsson
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG)
*备注: 24 pages; 15 figures
Abstract:Models of quantum systems faithfully map system parameters to observations, but the inverse problem of parameter inference from measurement data presents a fundamental challenge: computationally intractable likelihoods due to an exponentially large Hilbert space. Here, we introduce simulation-based quantum system inference, a unified, likelihood-free framework that learns parameter posteriors directly from classical simulation data. The central idea is to pair polynomial-cost classical simulators, such as Pauli propagation and tensor networks, with normalizing flows or other neural density estimators for accurate, reusable inference. A single model, trained once, maps any new measurement record to its posterior in one forward pass—turning per-experiment inference into a fixed, up-front cost. We numerically demonstrate the framework’s versatility across Pauli noise learning, quantum error mitigation, quantum state tomography, and Hamiltonian learning, with examples involving 81-qubit shallow circuits and 735-parameter inference. In each case, the approach yields accurate estimates of identifiable parameters, while posterior uncertainty provides additional diagnostics of non-identifiability and indicates where further characterization is needed. Our framework reduces data-acquisition requirements in quantum experiments and accelerates parameter inference, providing a practical route to characterizing and improving large-scale quantum systems.
[LG-463] Physics-Informed Neural Networks for Depth-Averag ed Avalanche Dynamics
链接: https://arxiv.org/abs/2609.34916
作者: Pradyumn Singh Sikarwar,Vishal Sharma,Gaurav Bhutani
类目: Fluid Dynamics (physics.flu-dyn); Soft Condensed Matter (cond-mat.soft); Machine Learning (cs.LG)
*备注: 49 pages, 41 figures
Abstract:Accurate prediction of avalanche motion is essential for hazard assessment in mountainous terrain. This study develops and evaluates a physics-informed neural network (PINN) framework for the Savage-Hutter model of depth-averaged granular flow, progressing from 1D analytical verification to 2D experimental validation. First, three 1D problems of increasing complexity were verified against the analytical solution: height prediction with prescribed velocity, velocity prediction with prescribed height, and coupled prediction of both fields using the conservative formulation. The decoupled tests accurately reconstructed the spatio-temporal evolution of each field when the other was prescribed. The coupled formulation learned both fields without prescribed data, achieving mean height and velocity RMSEs of 0.043 and 0.079 in non-dimensional units. A hyperparameter sensitivity study evaluated the effects of network depth, width, collocation density, learning rate, and epochs. The framework was then extended to 2D and validated against laboratory experiments of a cylindrical granular pile collapsing on an inclined plane, with TITAN2D providing numerical comparisons. Purely physics-based training converged to the trivial zero solution; augmenting the loss with 10 sparse training points from final deposit profiles produced a physics-informed, data-assisted hybrid framework. Peak flow depth, depth-averaged velocity, RMSE, and wetted-area IoU evaluated global and local agreement. Global height RMSE ranged from 2.7 to 6.7 mm across four experimental cases, while mean wetted-area IoU ranged from 69 to 81 %, demonstrating consistent performance across variations in pile mass and slope angle.
[LG-464] Conformal Prediction and Conditional Coverag e for Tabular Foundation Models
链接: https://arxiv.org/abs/2609.34887
作者: Sungwoo Park,Sunghee Park,Won Chang
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 37 pages
Abstract:Tabular foundation models (TFMs) provide predictive distributions for regression, but their prediction regions can exhibit undercoverage or overcoverage even when point predictions are accurate. We introduce C-USIM (Conditionally-Uniformized Score Integration Method), a lightweight application of highest predictive density split conformal prediction that accommodates multimodal predictions. Given calibration and test outputs, it requires no additional training or model inference. It provides finite-sample marginal validity under our assumptions. We bound conditional-marginal coverage gaps using distribution-estimation error and score discreteness, and examine coverage heterogeneity through percentile rank-score plots. Experiments with TabPFN and TabICL show improved marginal coverage accuracy and lower average conditional and group coverage errors. Under a fixed data budget, allocating more observations to calibration can reduce marginal coverage error despite less accurate point predictions.
[LG-465] MW-Nowcast: Six-hour ensemble nowcasting of extreme precipitation
链接: https://arxiv.org/abs/2609.34836
作者: Ning Wang,Zuliang Fang,Weixin Jin,Zhongjian Lv,Shuang Qin,Pengcheng Zhao,Siqi Xiang,Jiang Bian,Haoyi Xiong,Nan Guan,Bin Zhang,Liangjie Zhang,Denvy Deng,Qi Zhang,Matt Corey,Jitu Keshri,Sridhar Iyer,Hongyu Sun,Kit Thambiratnam,Jonathan Weyn,Richard E. Turner,Haiyu Dong
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Machine Learning (cs.LG)
*备注: 62 pages, 31 figures, 4 tables; includes Extended Data Figures and Supplementary Information
Abstract:Extending reliable nowcasting of extreme precipitation could provide critical additional time for warnings and emergency response during high-impact events such as flash floods. Radar-based generative machine-learning models have enabled skilful hyperlocal precipitation nowcasting, but accurate prediction of intense precipitation remains confined to the first few hours. Because storm-scale structure is predictable for longer than individual cells, a natural strategy is to predict that structure while generatively modelling only the uncertain local growth, decay, reorganisation and initiation of storms. Here we present Microsoft Weather Nowcast (MW-Nowcast), a six-hour ensemble radar nowcasting model that jointly learns a deterministic predictor to capture organised precipitation structure shared across ensemble members, and a generator to produce diverse local residuals around this shared prediction. Across independent test data from the United States, Europe and China, MW-Nowcast achieves higher detection skill than leading methods for heavy and extreme precipitation throughout the 6 h horizon. For the most intense rainfall, MW-Nowcast doubles the available warning time across all three regions, delivering 6 h forecasts with skill previously limited to 3 h for the leading generative baseline. A cost-loss decision analysis shows that MW-Nowcast retains substantial value for a broad range of applications even at 4-6 h, where alternative methods offer little benefit. These additional hours can give forecasters and emergency managers the time to warn and act before extreme rainfall strikes, helping to protect lives and property.
[LG-466] Finite-Time Concentration and Convergence Rates for Projected Two-Time-Scale Stochastic Approximation with Markov Noise
链接: https://arxiv.org/abs/2609.34791
作者: Rahul Singh,Vivek S. Borkar,Eric Moulines
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:
Abstract:We study finite-time concentration and convergence rates for projected two-time-scale stochastic approximation driven by a controlled Markov chain. The averaged fast map is contractive, while the slow iterate is projected onto a compact convex polyhedron. The associated projected ordinary differential equation may have a discontinuous vector field at the boundary, preventing a direct application of standard analyses based on Lipschitz vector fields. Using the Skorokhod map, we establish explicit high-probability bounds for tracking the moving fast equilibrium and the projected slow dynamics. These bounds separate martingale fluctuations, Markov-noise residuals, and the bias due to time-scale separation. A Lipschitz Lyapunov function satisfying a uniform decrease condition over fixed time intervals yields almost-sure convergence, with explicit last-iterate rates when the decrease admits a power lower bound. Under uniform Lyapunov contraction, polynomial step sizes yield joint fast-tracking and slow Lyapunov-error exponents arbitrarily close to 1/3 . Under the additional assumption that the reduced slow update map is a Euclidean contraction, logarithmically separated step sizes improve the joint rate to O(n^-1/2\log n) almost surely, including for boundary equilibria. The same rate holds under a distinct geometric condition involving a strictly attracting face of a box and a fast equilibrium that is constant on that face. An actor-critic application achieves an almost-sure value-gap rate of O(n^-1\log n) relative to the optimum within the constrained policy class. Further applications include projected TD(0) and projected stochastic gradient descent. We also extend the analysis to projection of the fast recursion under Euclidean contractivity. Subjects: Optimization and Control (math.OC); Machine Learning (cs.LG) Cite as: arXiv:2609.34791 [math.OC] (or arXiv:2609.34791v1 [math.OC] for this version) https://doi.org/10.48550/arXiv.2609.34791 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-467] Statistical Benefits of Fine-Tuning from Pretrained Initialization in Diagonal Linear Networks
链接: https://arxiv.org/abs/2609.34756
作者: Alexandre Declèves,Etienne Boursier,Nicolas Flammarion
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Adapting pretrained models to downstream tasks with limited data has become a central paradigm in modern deep learning. Yet, despite its widespread practical success, how fine-tuning leverages information from pretraining remains poorly understood theoretically. We study fine-tuning from pretrained weights through the lens of sparse linear regression and two-layer diagonal linear networks. In our setting, pretraining provides information through the support (and signs) of the initialization predictor, which may contain coordinates relevant to the downstream task. We show how pretrained information reshapes the implicit bias and training dynamics, and can thereby reduce the sample complexity of recovering the target parameters and support. In particular, for a clean initialization with correctly inherited signs, we show that the required sample size is comparable to that of a weighted Lasso estimator that explicitly exploits the pretrained support through a suitably chosen regularizer. Our results thus show how information encoded in pretrained weights can be implicitly exploited by gradient-based fine-tuning, reducing the amount of data needed to recover a downstream task.
[LG-468] Information-Theoretic Analysis of Next-Token Prediction under Markovian Data
链接: https://arxiv.org/abs/2609.34731
作者: Masoud Kavian,Abdellatif Zaidi,Milad Sefidgaran
类目: Machine Learning (stat.ML); Information Theory (cs.IT); Machine Learning (cs.LG)
*备注: 49 Pages, 6 Figures
Abstract:We develop an information-theoretic framework for generalization in next-token prediction under temporally dependent data. We consider independent trajectories generated by finite-memory Markov processes and distinguish algorithmic dependence, quantified by mutual information, from temporal dependence, characterized by mixing. For cross-entropy loss, we derive an expected generalization bound using the Donsker–Varadhan variational representation and a McDiarmid-type concentration inequality for Markov chains. A refinement captures the joint effect of context length and temporal mixing through the mixing properties of the history-state process. We then extend the bound through a rate–distortion formulation, replacing mutual information with the minimum information rate required to represent the learned model within a prescribed distortion in the generalization gap, yielding informative guarantees for deterministic algorithms over continuous hypothesis spaces. For margin-based prediction, we derive explicit bounds for linear and self-attention next-token predictors via noisy low-dimensional compression, revealing the roles of context length, model complexity, sample size, margin, and temporal mixing. Experiments on TinyStories and ETTh2 show that longer contexts can reduce both training and test losses, but typically reduce training loss more, enlarging the generalization gap. A complementary ETTh2 analysis identifies an effective predictive-memory scale near 24 hours, with no statistically supported improvement beyond this scale, offering a plausible explanation for test-performance saturation at larger contexts.
[LG-469] Hierarchical Clustering and Signal Denoising on Digraphs
链接: https://arxiv.org/abs/2609.34670
作者: Yi Wang,Sippanon Kitimoon,Hrushikesh N. Mhaskar,Xiaosheng Zhuang
类目: ignal Processing (eess.SP); Machine Learning (cs.LG)
*备注:
Abstract:In this paper, we propose a representation of a digraph (directed graph) as a Hermitian matrix derived from its adjacency matrix. This representation characterizes both the connectivity and the edge orientation of the digraph. Based on the spectral decomposition of the Hermitian matrix, a digraph clustering algorithm with k -means is introduced to produce a partition on the graph. Applying this algorithm (bottom-up) recursively to a digraph with partially labeled vertices yields a spectral hierarchical digraph clustering (\myproj) algorithm that produces consistent nested partitions of the digraph, or equivalently, a tree structure. Furthermore, based on the in-degree and out-degree of each cluster in the digraph clustering, a pair of hierarchical interval partitions (filtrations) can be derived in a top-down manner to produce a pair of nested knot sequences. These knot sequences facilitate the construction of multilevel spline quasi-interpolants, enabling a noisy graph signal to be decomposed into a coarse approximation and inter-level details, followed by adaptive thresholding and reconstruction. Experiments on synthetic and real-world digraphs demonstrate the superiority of our \myproj algorithm for digraph clustering across diverse graph structural properties (homophily and heterophily) and supervision settings. Moreover, experiments on digraph signal processing using multilevel spline quasi-interpolants further demonstrate the effectiveness of signal recovery on digraphs in terms of RMSE and SNR.
[LG-470] wo-Timescale Fine-tuning Provably Learns New Features for Two-Layer ReLU Networks
链接: https://arxiv.org/abs/2609.34667
作者: Etienne Boursier,Nicolas Flammarion
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Fine-tuning pre-trained models on specialized tasks with scarce data is central to modern deep learning. Despite its empirical success, theoretical understanding of fine-tuning remains limited. We introduce a Gaussian multi-index setting to study fine-tuning from pre-trained weights, where the teacher network has m+1 features, m of which are learned during pre-training and one of which must be learned during fine-tuning. For two-layer ReLU networks, we show that two-timescale training, i.e., updating the outer weights infinitely faster than the hidden ones, learns the new task-specific feature while preserving the pre-trained ones in the model representation. Moreover, only \mathcalO(d) fine-tuning samples are required for this recovery, independently of the number of pre-trained features. In contrast, with random initialization, the same number of samples is insufficient to recover the target parameters. Our results therefore demonstrate that pre-training can induce an implicit bias with a clear statistical advantage over random initialization, enabling feature learning from scarce fine-tuning data.
[LG-471] Probabilistic Geodesic Flow Matching on Location-Scale Families
链接: https://arxiv.org/abs/2609.34613
作者: Zeyuan Yu,Zhi Chang,Shiwei Lan
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Probability (math.PR)
*备注: 25 pages, 10 figures
Abstract:Flow matching (FM) has recently emerged as a promising framework for generative modeling due to its conceptual simplicity and strong empirical performance. In FM, samples are transported along a vector field parameterized by a neural network, inducing a probability path that evolves from a simple noise distribution to the target data distribution, governed by an ordinary differential equation (ODE). However, existing FM approaches predominantly rely on probability paths derived from optimal transport (OT) between Gaussian distributions, which may be suboptimal for capturing complex data with inhomogeneous structures such as heavy tail or sharp contrast. In this work, we generalize FM to the broader class of location-scale families for handling data inhomogeneity and introduce a novel class of probability paths defined as geodesics on the manifold of probability distributions. We name this approach probabilistic geodesic flow matching to distinguish it from prior geodesic (Riemannian) FM methods defined in input space. We argue that Euclidean OT-based paths are not necessarily optimal in probability space and may limit modeling flexibility. Through synthetic benchmarks and scientific datasets at different scales, we demonstrate that the proposed method more effectively captures complex distributions, leading to improved or comparable performance compared with SOTA geometry-motivated generative models.
[LG-472] Pre-registered tests of solid-state-physics-inspired LLM compression: a cluster-level negative result at small-language-model scale
链接: https://arxiv.org/abs/2609.34292
作者: Jun-qiang Lu
类目: Disordered Systems and Neural Networks (cond-mat.dis-nn); Materials Science (cond-mat.mtrl-sci); Other Condensed Matter (cond-mat.other); Machine Learning (cs.LG)
*备注: 37 pages, 9 figures
Abstract:We report a three-month autonomous research-agent program testing five solid-state-physics-inspired compression mappings on pretrained language models, with predictions committed to git before any pilot data and a 3-sigma gate deciding PASS or SHELVE. The common anchor – area-law / Kohn-nearsighted decay of the one-particle density matrix – has a distance face (P001 Wannier, P002 tight-binding) and a rank face (P003 DMRG-truncated MLPs, P005 Wilson-RG, P011 tensor-train embeddings). P005 was pre-empted at Phase 1; three of four Phase-3 pilots were falsified. On the attention face, GPT-2-medium attention-versus-distance is best fit by a stretched exponential in 12 of 16 median-layer heads once probe padding is excluded, and a tight-binding cutoff costs +96% perplexity (P002); on Pythia-160M the Wannier sparsity 0.054 +/- 0.004 is indistinguishable from PCA, random-Haar and identity baselines (P001). On the rank face, per-token tensor-train bond dimension does not track surprisal (r = 0.016 vs a pre-registered 0.65) and the format inflates rather than compresses (P011). P003 is mixed: its scaling claim shelved (r = -0.434), its MPO premise died at stage-0, and its cross-paper check, r = 0.523 as first written, collapses to 0.047 under the same correction, leaving both cross-paper checks null. The results invert the pre-registered prediction that most attention heads behave like Kohn-nearsighted insulators, pointing instead to critical, glassy or heavy-tailed regimes; the inversion is specific to the = 350M scale tested, while the rank-face no-gain result held to 7-8B. We contribute the pre-registration + 3-sigma + cluster-framing + append-only-catalogue discipline – including why our own enforcement gate was designed but not deployed – four pre-registered negative results with full data release, and the inversion. The catalogue holds eighteen concluded studies, seventeen negative.
[LG-473] Functional Autoencoders for Amplitude-Phase Representation Learning
链接: https://arxiv.org/abs/2609.34207
作者: Peida Wu,Xinyang Xiong,Pengcheng Zeng
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 26 pages, 13 figures
Abstract:Functional data are intrinsically infinite-dimensional, and often exhibit phase variation, where corresponding events occur at different times across observations. Existing linear dimension reduction methods struggle with nonlinear amplitude variation, while functional autoencoders without an explicit warp entangle temporal misalignment with shape. We propose the Amplitude–Phase Functional Autoencoders (AP-FAE), an unsupervised framework for functional data that spans both univariate and multivariate cases, with emphasis on the multivariate setting, and factorizes the latent space into separate amplitude and phase embeddings derived from all channels. A smooth functional decoder reconstructs channel-specific amplitude functions in canonical time, and a shared monotone, endpoint-preserving warp captures phase variation. We prove a bound linking amplitude recovery to registration, reconstruction, and noise errors, and validate it numerically. Across synthetic data and six real-world benchmarks, AP-FAE outperforms state-of-the-art baselines on most clustering and alignment metrics and on all reconstruction metrics. Clustering with amplitude embeddings alone consistently surpasses joint amplitude–phase clustering, confirming the benefit of explicit disentanglement. Code is available at this https URLthis https URL.
[LG-474] GT-PSSM: Unified Probabilistic Framework for Stochastic Dynamics Modeling and Dependency Learning in Multivariate Time Series Anomaly Detection
链接: https://arxiv.org/abs/2609.34161
作者: Wonmo Koo,Jaeyeong Lee,Taeseong Yoon,Heeyoung Kim
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Multivariate time series anomaly detection (MTAD) is crucial for ensuring the safe and reliable operation of complex systems. Many existing methods learn normal patterns by training reconstruction or forecasting models on predominantly normal data. However, a large portion of these approaches rely on deterministic models and their associated point-wise output errors for anomaly scoring. Since real-world multivariate time series are inherently stochastic due to measurement noise and intrinsic system randomness, purely error-based scores can be unreliable, as large errors may arise from benign fluctuations rather than true anomalies. Probabilistic approaches address this limitation by quantifying uncertainty in model outputs. In particular, probabilistic state-space models (PSSMs) provide a principled framework by modeling stochastic system dynamics through latent state transitions and measurement noise via emission models. Despite this advantage, existing PSSM-based MTAD methods often struggle to capture long-range temporal dependencies and inter-variable dependencies, as they typically rely on noise-sensitive recurrent architectures and lack explicit cross-variable structure modeling. To address these limitations, we propose Graph-Transformer-Enhanced Probabilistic State-Space Model (GT-PSSM), a novel PSSM-based MTAD method that tightly integrates PSSM-based probabilistic modeling of stochastic dynamics with Graph Transformer-based learning of temporal and inter-variable dependencies. By jointly modeling stochasticity, long-range temporal dependence, and variable interactions within a unified probabilistic framework, GT-PSSM enables more robust anomaly detection.
[LG-475] Forecast-Necessary Causal Discovery for Nonlinear Political Panel Data: Feedback Functional Form and the Dynamics of Democratization
链接: https://arxiv.org/abs/2609.34115
作者: Michael Coppedge,Dmitry Zaytsev,Valentina Kuskova
类目: Methodology (stat.ME); Machine Learning (cs.LG)
*备注:
Abstract:A non-significant coefficient in a dynamic panel model need not imply the absence of a relationship. It may instead reflect heterogeneous effects averaged toward zero, reciprocal dynamics overlooked by a recursive specification, or relationships masked by the omission of correlated covariates. Standard linear estimators cannot distinguish among these possibilities. We develop an inferential workflow for political panel data that resolves this ambiguity by combining flexible autoregressive estimation, forecast-necessity testing, functional characterization, and same-data linear benchmarking. The workflow first identifies relationships required for out-of-sample prediction, then characterizes their functional form across political contexts, and finally, distinguishes differences arising from estimator flexibility from those due to model specification. Applied to the causal sequence model of democratization on the V-Dem panel of 113 countries, the workflow reproduces the model’s central finding - the protective belt of civil society, the rule of law, and institutionalized parties - while recovering reciprocal relationships from democracy to its institutional supports that a linear model cannot detect. Most importantly, three weak published direct effects, of which two are null, and one is marginally significant, receive three different diagnoses: one dissolves under the full specification, one reflects heterogeneous effects averaged toward zero, and one was masked by the reduced variable set. The workflow corrects the published record in both directions, removing one relationship and recovering two. More broadly, the workflow provides a framework for evaluating dynamic political theories under a model class capable of representing nonlinear and reciprocal mechanisms while preserving relationship-level interpretation and explicit inferential standards.
[LG-476] he Statistical Cost of Causal Discovery with Feedback
链接: https://arxiv.org/abs/2609.34050
作者: Sunmin Oh,Seungsu Han,Gunwoong Park
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:What determines the unavoidable sample cost of learning cyclic causal structure? For cyclic linear non-Gaussian models, we study exact condensation recovery from observational data: identifying the strongly connected component (SCC) partition and all edges between components. We establish the first information-theoretic lower bounds on sample complexity for this target. For p variables, maximum SCC size s_\max , and maximum external-parent count d_B , any estimator requires order s_\max\log(ep/s_\max)+d_B\log(ep/d_B) samples in the worst case over a regular model class. These bounds distinguish the costs of SCC membership and external-parent selection. Under principal invertibility and without correlation faithfulness, we establish a population block-exogeneity principle that identifies unknown root SCCs through residual independence and inclusion minimality. A sparse-adjustment characterization shows that small adjustment sets suffice to identify SCCs and their direct external parents, without regressing on all previously recovered variables. These characterizations yield BlockExo, which attains a structurally matching sample bound without knowing s_\max or d_B under suitable conditions. Simulations support the structural dependence of our sample bound and demonstrate BlockExo’s sample-efficient recovery in comparisons with other methods for cyclic causal discovery.
[LG-477] Singularities of Non-negative Matrix Factorization and their application to Bayesian inference
链接: https://arxiv.org/abs/2609.34043
作者: Naoki Hayashi,Yota Maeda,Yasushi Esaki
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注: 23 pages
Abstract:Non-negative matrix factorization (NMF) is a singular statistical model whose Bayesian asymptotics are governed by the real log canonical threshold (RLCT). We study the local geometry of the factorization map and derive an upper bound for the RLCT of NMF. Let H be the model inner dimension and H_0 the non-negative rank of the true M\times N matrix. Assuming that the true matrix admits a strictly positive factorization of inner dimension H_0 in the interior of the parameter domain, we prove, for smooth positive priors, that \lambda\leq (H-H_0)\min(M,N)+H_0(M+N-H_0)/2 . This bound strictly improves the previous bound when H_0\geq3 . The proof uses a local analytic normal form that separates independent linear coordinates from a residual matrix product. When H=H_0 also equals the ordinary rank of the true matrix, we obtain the exact value \lambda=H_0(M+N-H_0)/2 . Under the standard assumptions of singular learning theory, these results bound the leading coefficients of the expected Bayesian generalization error and the Bayesian free energy.
[LG-478] wo-Sample Testing for Inhomogeneous Random Graphs in Non-Integral L_r Norms
链接: https://arxiv.org/abs/2609.33968
作者: Soham Dan
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Testing whether two populations of networks share the same edge probabilities is a basic problem in network inference. How hard it is depends on the norm used to measure the difference. For the inhomogeneous Erdős–Rényi (IER) model, the optimal sample complexity is known for every integer L_r norm and for 1\le r2 . For non-integral r2 , however, the known upper and lower bounds do not match, and the lower bound was conjectured to be tight. We study this gap for two-sample testing on aligned vertices. We propose a test that runs two published statistics, of orders 2 and \lceil r\rceil , on the same data and rejects if either one rejects. Its thresholds come from Hölder interpolation, so that both statistics have the same sample cost. We prove that this test attains the conjectured rate. Combined with earlier results, this shows that for every fixed r\ge1 the minimax sample complexity is of order n^\max\4/r-1,2/r/\epsilon^2 , even when the separation changes with n . In simulations with n between 32 and 256, the number of graphs needed for 80% power at level 0.05 grows with n at a rate consistent with the theory. For r=2.5 , for example, the fitted exponent is 0.78 , against the theoretical value 0.8 . Interestingly, the two statistics split the work as the interpolation argument suggests: the higher-order statistic is more powerful when only a few edges change, and the L_2 statistic when many edges change.
[LG-479] DCEmbed: Scalable Optimization over Neural Surrogates
链接: https://arxiv.org/abs/2609.33879
作者: Akshay Sreekumar,Nicolas Christianson,Priya L. Donti,Ellen Vitercik,Ram Rajagopal
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:
Abstract:Neural surrogates can accelerate large-scale optimization by replacing expensive or intractable model components with efficient learned approximations, but solving the resulting embedded problems can remain prohibitively costly. For instance, standard exact encodings of neural networks with ReLU activations allow the problem to be solved by mixed-integer solvers, but add large numbers of binary variables to accommodate the nonlinearity of the activations, which can render the problem computationally prohibitive. To address this, we propose DCEmbed, a heuristic for optimization problems with embedded neural surrogates that leverages the difference-of-convex (DC) representation of the network and avoids adding activation binaries. Exploiting shared structure within the DC representation of a ReLU network, we derive a reduced-size, exact formulation for its convex components that can be embedded in optimization problems using just two linear inequalities and one continuous auxiliary variable per hidden neuron. Using this, the problem is solved via an iterative penalty convex-concave procedure, where only the concave portions of the neural terms are approximated at each stage. The original objective, constraints, and any discrete decisions are retained, allowing standard convex or mixed-integer optimization solvers to optimize the host and surrogate jointly at each iteration. In experiments on quadratic programs, mixed-integer resource allocation, and neural two-stage stochastic programming, our method demonstrates much faster progress toward high-quality feasible solutions than approaches using exact mixed-integer embeddings. In particular, DCEmbed achieves 4\times lower normalized primal integral than the best exact baseline on the resource allocation problem, while in two-stage stochastic programming it reaches the global surrogate optimum \sim 5\times faster than Gurobi ML.
[LG-480] Annealed Sinkhorn with Momentum: Certified Unregularized Optimal Transport in Linear Memory
链接: https://arxiv.org/abs/2609.33814
作者: Samuel J. K. Chin,Maximilian Schiffer
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注: 38 pages, 11 figures
Abstract:We characterize Bregman Douglas-Rachford splitting (BDRS) for unregularized discrete optimal transport and develop an anytime primal-dual certificate in linear memory. We first establish that BDRS coincides with warm-started Inexact Proximal point method for exact Optimal Transport (IPOT) using a single inner Sinkhorn iteration. By eliminating the primal transport plan from the updates, we derive an equivalent dual formulation that reveals BDRS as annealed Sinkhorn under an implicit inverse-linear temperature schedule, with an additional log-scaling momentum term and a cooler kernel. While this explains the role of the temperature parameter in BDRS as an initial temperature, it also reduces the solver’s memory requirement from quadratic to linear. Utilizing this annealing perspective, we introduce overrelaxed BDRS, which combines annealing and overrelaxed scaling within a single recursion. We derive a primal-dual certificate for both methods that can be evaluated in linear memory without transport plan construction, thus providing a computable stopping rule. On the DOTmark benchmark, combining momentum with the cooler kernel produces substantially smaller optimality gaps than annealed Sinkhorn under the same schedule. For pixel-level color transfer between 1024\times1024 images, BDRS attains a lower repaired transport cost than MDOT-TNT with a 9 \times speed up, reaching a relative duality gap of 1.59% in 24 minutes. We further demonstrate a color transfer with 4238\times2365 images, yielding 10 million pixels per image and approximately one hundred trillion implicit transport entries, reaching a best relative duality gap of 2.41% and 2.80% within 35 hours in each direction on a single NVIDIA L40S GPU.
[LG-481] Autonomous phase discovery
链接: https://arxiv.org/abs/2609.33802
作者: Shiyu Zhou,Yuxuan Zhang,Sebastian Wetzel,Roger Melko,Xiu-Zhe Luo
类目: Quantum Physics (quant-ph); Strongly Correlated Electrons (cond-mat.str-el); Machine Learning (cs.LG)
*备注: 10 pages, 5 figures, this https URL
Abstract:Understanding quantum phases of matter has long relied on physicists’ intuition and mathematical tools such as symmetry and topology. Remarkably successful as these approaches have been, they provide no universal way to explore a Hamiltonian space whose organizing principle is not known in advance. In this work, we introduce a fully autonomous system combining differentiable programming and unsupervised learning for quantum phase discovery. The search evaluates ground-state data along an adaptive trajectory rather than on a predetermined parameter grid. We demonstrate the system with three different solvers and benchmark it against random sampling at equal ground-state-evaluation budgets. On a generalized cluster chain hosting up to 200 distinct phases, the search finds up to 25 more phases at the same budget, and matches random sampling given thirty times its budget. On a 50 -parameter Chern insulator, it reaches sectors not obtained by the simple harmonic constructions considered here, in a family whose inverse problem remains open, while recovering all sectors found by sampling. Our results establish autonomous, gradient-driven exploration of Hamiltonian space as a practical route to discovering quantum phases without phase labels or a prescribed target phase.
[LG-482] aming the Greeks: Option Portfolios with Inductive Biases
链接: https://arxiv.org/abs/2609.33767
作者: Wee Ling Tan,Stephen Roberts,Stefan Zohren
类目: Portfolio Management (q-fin.PM); Machine Learning (cs.LG); Computational Finance (q-fin.CP); Trading and Market Microstructure (q-fin.TR)
*备注:
Abstract:We present an end-to-end deep learning framework for systematic options trading that directly embeds hedging behavior through explicit control of portfolio-level risk exposures. While neural networks trained to optimize risk-adjusted performance have been shown to outperform traditional rules-based strategies, such approaches remain agnostic to the sensitivities of the resulting portfolios with respect to specific underlying risk factors. We propose a general training objective that combines a performance-driven loss with a differentiable risk-sensitivity penalty, enforcing neutrality to selected risk dimensions. Unlike reinforcement learning methods that approximate optimal hedging policies via simulated market dynamics, our framework operates entirely on historical data and jointly optimizes risk-adjusted returns and targeted risk constraints in a single learning problem. We instantiate the framework on static delta-neutral straddle portfolios with the penalty directed at first-order directional exposure, and evaluate two penalty variants – an exposure-normalized penalty and a Greek-ratio drift penalty. Empirical results on Nasdaq 100 equity options demonstrate that appropriately calibrated regularization simultaneously improves out-of-sample risk-adjusted performance relative to an unregularized baseline while reducing realized directional exposure.
[LG-483] YuE2: Unifying Symbolic and Audio Music Generation at Frontier Quality
链接: https://arxiv.org/abs/2609.33757
作者: Ruibin Yuan,Jiahao Pan,Junyan Jiang,Zhiyue Wu,Ziya Zhou,Jiankai Sun,Yizhi Li,Ge Zhang,Yicheng Gu,Zeyue Tian,Junyu Dai,Hanfeng Lin,Kai Li,Shangda Wu,Xuanjie Liu,Jiaming Wang,Zihan Liu,Yue Wang,Yinghao Ma,Hanzhi Yin,Kangrui Chen,Xinyue Zhang,Ziyang Ma,Mengqi Liao,Hejia Zhao,Guowei Huang,Chao Yan,Lei Ke,Jianwei Yu,Bei Liu,Joe Guo,Liumeng Xue,Gus Xia,Wei Xue,Yike Guo
类目: Audio and Speech Processing (eess.AS); Machine Learning (cs.LG); Sound (cs.SD)
*备注: 56 pages. Technical report. Project: this https URL
Abstract:Symbolic models make melody, harmony, rhythm, and form explicit but typically stop before a finished recording; audio models produce complete songs while leaving composition implicit. We introduce YuE2, which unifies symbolic and audio music generation at frontier quality through symbolic planning. A single AR-NAR Mixture-of-Transformers (MoT) first writes a readable score specifying melody and harmony, expands it into semantic music tokens, and realizes it as full-song audio. In comparisons using the same checkpoint, experts prefer symbolic planning for overall quality and musicality, with 49.3% of overall preferences versus 34.6% without planning. Experts also favor the unified model over a separate language model and diffusion Transformer. On WildSongBench, YuE2 scores 6.73 on SongBench Global Avg, exceeding all evaluated public baselines. Selecting from eight candidates (best-of-8), YuE2 reaches 6.96, the highest observed mean among all evaluated systems. Expert listening further establishes its competitiveness with proprietary song generators, favoring best-of-8 over Suno v4.5 and yielding nearly balanced preferences against Suno v5. To learn this generation process from recordings without aligned scores, we introduce MERT2 and SheetSage2 to supply semantic and symbolic supervision. MERT2 sets a new state of the art in music representation learning, surpassing previous best results on 14 of 15 MARBLE metrics; SheetSage2 leads 12 of 15 benchmark-metric pairs in our lead-sheet transcription comparison. The same checkpoint follows score edits while largely preserving unedited musical content and generates zero-shot covers without cover-specific training. Its readable score also enables agentic music editing, with external language models translating user feedback into revisions of the composition.
[LG-484] Multi-Marginal Inverse Optimal Transport for Contrastive Learning Via Explicit Anchor-Positive-Negative Coupling
链接: https://arxiv.org/abs/2609.33741
作者: Ngoc-Hai Nguyen,Thuan Nguyen,Prakash Ishwar,Shuchin Aeron
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Inverse Optimal Transport (OT) based methods for representation learning learn representations such that the global OT coupling between a pair of data marginals in the representation space, concentrates on the positive pairs. This is in contrast to previous methods that primarily focused on pairwise matching. However, these methods \textitDO NOT utilize negative pairs and hence are not truly contrastive in their approach. We show that this leads to issues of dimensional collapse and hence degraded downstream performance. To alleviate this, we develop a novel multi-marginal (MM) inverse OT (IOT) contrastive learning (CL) approach called Neg-MMIOT-CL, which learns representations such that the global multi-marginal OT (MMOT) coupling between a triple of data marginals, with respect to a carefully designed ground-cost between triplets of data points in the representation space, concentrates on the anchor-positive-negative \textittriplets . For a latent class model, we empirically show that Neg-MMIOT-CL alleviates dimensional collapse. Furthermore, for a specific choice of ground cost for all triplets in representation space, we prove that the optimal representation configuration for Neg-MMIOT-CL exhibits equiangular property for within-class and across-class representations, which translates to Neural-Collapse when the representation dimension is larger than the number of classes minus one – a result that is \textitpreviously established only for pairwise contrastive learning methods. Finally, we propose Neg-IOT-CL-PushPull, that is a computationally efficient alternative to Neg-MMIOT-CL, alleviating the high cost of computing MMOT plans needed during implementation. We apply these methods on both synthetic and real-world datasets and show significant improvements over existing OT-based contrastive learning methods.
[LG-485] Weighted Spline-Expanded Networks with Distributional Balancing for Continuous Treatment Effects
链接: https://arxiv.org/abs/2609.33740
作者: Shucheng Liu,Chan Park,Guanhua Chen
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Estimating causal effects with continuous treatments in observational studies is challenging due to confounding, model misspecification, and high-dimensional covariates. We propose the Weighted Spline-Expanded Network (WSENet), an end-to-end neural framework that addresses these challenges by combining covariate balancing, structured treatment embedding, and bias-corrected outcome estimation. WSENet first applies Distance Covariate Optimal Weights to induce distributional independence between covariates and treatment without relying on parametric models. It then learns the conditional outcome via a structured network that fuses outcome-relevant representations of covariates with a spline-expanded treatment input, enabling smooth and flexible modeling of the dose-response relationship. To mitigate residual bias, we introduce Weighted Targeted Regularization, a correction technique based on efficient influence functions that yields a doubly robust estimator. Extensive evaluations on semi-synthetic and real-world datasets, including high-dimensional genomic and environmental health data, demonstrate that WSENet consistently outperforms existing baselines in both accuracy and stability.
[LG-486] A Statistical Perspective on Knowledge Distillation: Foundations Classical Methods and Large Language Model Extensions
链接: https://arxiv.org/abs/2609.33727
作者: Luyang Fang,Haoran Lu,Jiazhang Cai,Tao Wang,Huimin Cheng,Wenxuan Zhong,Ping Ma
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Knowledge Distillation (KD) has emerged as a vital paradigm for transferring the capabilities of high-capacity models to efficient ``student’’ counterparts, addressing critical challenges in computational cost, deployment constraints, and privacy-sensitive settings. Although KD is widely used in practice, it is often viewed primarily as an engineering technique, with a unified statistical perspective remaining less developed. This review bridges that gap by presenting a unified Bayesian formulation of KD that formulates teacher predictions as prior information. This provides a principled interpretation of how teacher information is incorporated into student learning and establishes a rigorous connection to uncertainty quantification. We demonstrate how this foundational lens reconciles classical distillation with modern extensions in generative and foundation-model systems, showing that contemporary developments remain rooted in these same statistical principles. By synthesizing theory with emerging methodologies and diverse applications, this review provides a conceptual roadmap and identifies critical open problems for the future of the field.
[LG-487] Reliable Replay through Spatial Coherence in Online Continual Learning
链接: https://arxiv.org/abs/2609.33725
作者: Haixiang Sun,Jiefu Zhang,Yinghao He,Yang Xu,Vaneet Aggarwal,Bharat Bhargava,Andrew L. Liu
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Continually adapting models to new tasks requires retaining earlier knowledge under limited memory and computation. Experience replay addresses this challenge, but priorities based on individual loss increases overlook how related memories respond to the same update and can overemphasize isolated responses. We introduce SPatial coHErent risk control for REplay (SPHERE), a general replay-allocation method applicable across a broad range of learning settings. SPHERE uses a representation kernel to aggregate signed prospective loss changes, attenuating unsupported spikes while retaining coherent increases. It then formulates allocation as entropy-regularized transport, redistributing uniform source mass toward supported high-risk regions while penalizing long-distance transfers. We derive replay coefficients from the transport objective’s sensitivity to the original loss changes and blend them with uniform replay to maintain baseline rehearsal. Our analysis establishes conditions under which kernel aggregation improves risk estimation and bounds transport-value inflation due to residual noise and smoothing bias. Experiments demonstrate that SPHERE improves accuracy and reduces forgetting across noisy-label vision tasks, continual language-model instruction tuning, and code-generation reinforcement learning with incomplete test rewards.
[LG-488] apanda: A Matrix-Free Differentiable Solver for Nonconvex Constrained Optimization Layers
链接: https://arxiv.org/abs/2609.33630
作者: Yuankun Chen,Zifei Nie,Kangyu Lin,Ján Drgoňa,Liang Wu
类目: Optimization and Control (math.OC); Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注:
Abstract:Differentiable optimization brings the structural guarantees of mathematical optimization to network pipelines, allowing them to be trained end-to-end. However, its application remains challenging for nonconvex constrained problems, as existing differentiable solvers often suffer from limited modeling expressiveness due to their reliance on specialized problem structures, while also incurring substantial computation time and memory overhead in both the forward and backward passes. To address these challenges, we propose lapanda, a matrix-free differentiable solver for nonconvex optimization with general constraints. It reformulates the problem to a sequence of augmented Lagrangian subproblems, each handled by a first-order inner solver through a proximal averaged quasi-Newton algorithm with adaptive linesearch, thus enabling efficient forward optimization. We establish local well-posedness of the solution map and convergence of the outer iterations, and further derive a sensitivity alignment between the original problem and the final subproblem in the backward pass, demonstrating that the subproblem sensitivity, which can be computed efficiently in a matrix-free manner, provides a principled approximation to the exact optimizer sensitivity. We evaluate lapanda on nonconvex constrained Rosenbrock benchmarks, imitation learning with several representative constrained optimal control problems, and embedded robotic obstacle-avoidance tasks. Compared with state-of-the-art differentiable solvers, lapanda delivers substantial reductions in computation time and memory footprint while maintaining reliable constraint satisfaction and learning performance.
[LG-489] rminal-Register Certification for Finite-Measurement Learning of Multiscale Quantum States
链接: https://arxiv.org/abs/2609.33567
作者: Bhvain Makwana,Kashyap Patel,Manjunath Joshi,Jaideep Mulherkar
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG)
*备注: 21 page, 5 figures
Abstract:Structured quantum-state learning not only depends on an expressive ansatz but also on an operational certificate that stays meaningful with finite measurements and imperfect implementation. We study pure one dimensional states learning by an inverse binary multiscale entanglement renormalization ansatz (MERA). In the learning procedure, the qubits removed during coarse graining are controlled coherently and measured together at the terminal register. We confirm that an ideal sequential and terminal measurement schedule delivers the same complete bit string distribution under matched causal operations, while normalized postselection can amplify perturbations inversely with prefix acceptance. A noise aware theorem introduces an individual calibrated total variation implementation budget to the finite shot certificate. The protocol is estimated on an open boundary transverse field Ising ground state. A frozen 8-qubit schedule using 560 million simulated training measurements per run achieves fidelity above 0.99 in all 60 held-out runs, with a mean fidelity of 0.996886 . 1080 circuit-noise cells and 6480 confidence-coverage rows are covered by fixed-circuit robustness validation without a locked soundness violation. We then address architectural fairness at n=16 using three new studies. In a 120-run exact-gradient multistart diagnostic, MERA has higher fidelity in 58/60 paired restarts and lower long-range error in 60/60, although no run met the prespecified stationarity criterion. Finally, a causal cone-complete, parameter matched local circuit achieves 2.62\times greater aggregate gate exposure yet loses all 30 paired comparisons in fidelity, long-range error, energy, and entropy.
[LG-490] Sparsity by Default: The Theory and Practice of ARD in Gaussian Process Regression for Variable Selection
链接: https://arxiv.org/abs/2609.33550
作者: Jia Cai
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Automatic relevance determination (ARD) is the standard device for input selection in Gaussian process (GP) regression. By giving the covariance kernel a separate lengthscale for every input and learning those lengthscales by maximizing the marginal likelihood, ARD lets the data decide which coordinates matter: irrelevant inputs receive very large lengthscales and are effectively switched off. We trace this mechanism to the Bayesian Occam’s razor embodied in the marginal likelihood, derive the gradient through which it prunes inputs, and emphasize that ARD delivers effective rather than exact sparsity. We review the algorithms used in practice and the rules that turn lengthscales into selections, and we survey the asymptotic theory, distinguishing the fixed-domain identifiability obstruction on the lengthscales from the high-dimensional selection-consistency guarantees recently established for hierarchical GP priors, and noting what remains open for plain ARD. We compare ARD with spike-and-slab priors, sparse axis-aligned and global-local shrinkage priors including the Bayesian lasso and horseshoe, penalized likelihood kriging, sensitivity and projection criteria, and additive kernels. We argue that ARD endures because of its seamless integration with kernel learning, universal software support, and low cost, and we close with its limitations and remedies.
[LG-491] Calibrated Derivative-Process Sensitivity for Gaussian-Process Variable Selection
链接: https://arxiv.org/abs/2609.33549
作者: Jia Cai
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注:
Abstract:Automatic relevance determination (ARD), the default tool for variable selection in Gaussian-process (GP) regression, ranks inputs by inverse lengthscales – which measure how fast a function varies, not how much an input contributes to prediction – and offers no calibrated rule for deciding which inputs to keep. The prediction-centred alternative, the derivative sensitivity \nu_j = \mathbbE[(\partial f/\partial x_j)^2] , is available in closed form from a fitted GP, but turning it into a selection rule is harder than it looks: at a null input the estimator is a degenerate quadratic form, so Wald and Bernstein-von Mises cutoffs are anti-conservative, and the natural residual bootstrap is mis-scaled. We show that a studentized multiplier bootstrap of the GP derivative process repairs both, prove its validity through an invariance principle for quadratic forms, and obtain asymptotic family-wise and false-discovery-rate control across inputs. Over 100 replications the rule controls FDR wherever inputs are truly null, while uncalibrated derivative rankings breach the target by up to 2x and a Bernstein-von Mises cutoff by 2.2x; at matched FDR it loses no power; it holds under a Matérn kernel and input correlation up to 0.99; on real data with planted and authentic null inputs it admits 5-12x fewer spurious inputs; it costs 5-18% of the GP fit; and a block-averaged variant retains validity at cost linear in n .
[LG-492] Neural Scaling Laws of Transformer Operator Network
链接: https://arxiv.org/abs/2609.33533
作者: Haoran Yan,Zhongjie Shi,Yuanzhe Xi,Peng Chen,Wenjing Liao
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 47 pages, 5 figures
Abstract:Transformers have emerged as powerful architectures for learning solution operators of physical systems. Empirically the prediction error has been observed to decrease when the data size and model size increase, suggesting neural scaling behavior. Yet a theoretical understanding of such scaling laws for transformer-based operator learning remains limited. In this work, we develop a theoretical framework for characterizing the approximation and generalization errors of transformer-based operator learning. Our analysis builds on a local-to-global approximation principle that is naturally aligned with the softmax attention mechanism and yields discretization-invariant output functions. On approximation theory, we derive a universal approximation error of transformer-based operator learning for Hölder-regular operators. On generalization theory, we establish a power scaling law between the prediction error and the training data size. The rate of convergence represented by the scaling exponent explicitly reflects the dimensions of the input and output domains, the regularity of the underlying functions and operators, and crucially, the intrinsic dimension of the input function class. By exploiting this intrinsic low-dimensional structure, our analysis yields a power-law generalization rate for operator learning, in contrast to the logarithmic-type power-law rates appearing in existing analyses of operator learning with feedforward neural networks. Numerical experiments validate the predicted power-law scaling and confirm that the convergence rate varies systematically with the intrinsic dimension of the input function class.
[LG-493] Domain-Adapted Diffusion Models for Conditional Independence Testing
链接: https://arxiv.org/abs/2609.33490
作者: Yanfeng Yang,Junda Zhao,Yijie Gao,Jiaqi Yang,Xinyu Shi,Ziqi Chen,Shunyu Zhao,Shuai Li,Wei Huang,Eshant English,Kenji Fukumizu
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Conditional independence (CI) is a fundamental concept in statistics and machine learning. Recent advances in conditional generative modeling provide flexible tools for generative-model-based CI tests, which rely on an estimated conditional distribution to generate randomized samples. However, errors in estimating this distribution accumulate in existing Type I error bounds, and consistency of the generative estimator alone does not guarantee asymptotic Type I error control. To address this limitation, we formulate conditional generative modeling as a domain adaptation problem and leverage auxiliary data from multiple source domains to improve estimation in the target CI testing domain. We propose Domain-Adapted Diffusion (DA-Diff), a multi-source domain adaptation framework for conditional diffusion models based on weighted empirical risk minimization over both target and source domains. We establish the convergence rate of DA-Diff and show how transferable source data can improve target-domain estimation through an increased effective sample size while controlling transfer bias. Building on DA-Diff, we further propose Domain-Adapted Conditional Independence Testing (DA-CIT) and show that its Type I error satisfies P(p \leq \alpha) \leq \alpha + o(1) . Experiments demonstrate that DA-Diff improved conditional generation quality compared with transfer-learning diffusion baselines, while DA-CIT provides strong Type I error control and competitive power.
[LG-494] Sharp training-conditional coverag e for conformal prediction under covariate shift
链接: https://arxiv.org/abs/2609.33456
作者: Mehrdad Pournaderi
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST); Methodology (stat.ME)
*备注:
Abstract:Weighted split conformal prediction reweights calibration scores by the likelihood ratio between the test and training covariate distributions and guarantees marginal coverage under covariate shift. We study its coverage conditional on the calibration data. An elementary argument, based on a single concentration inequality at a fixed population quantile, gives explicit training-conditional bounds without unspecified constants, and shows that the relevant scale is not the supremum of the likelihood ratio but a variance proxy built from the chi-squared divergence of the shift and from the average of the ratio over the part of the test population, of probability equal to the miscoverage level, where it is largest. A two-point lower bound shows that the root-m rate and the chi-squared contribution are intrinsic to the shift. Run at an explicitly inflated level, the weighted quantile becomes a deterministic PAC prediction set. We compare it with randomized rejection sampling and with importance-weighted learn-then-test and, through a certified choice of a clipping level for the likelihood ratio, map the regime in which each gives the narrower valid set. The analysis extends to estimated likelihood ratios and to tail functionals estimated from an unlabeled source sample, which yields a fully finite-sample certificate.
[LG-495] Recovering Lower-Dimensional Semialgebraic Support of a Measure from its Moments
链接: https://arxiv.org/abs/2609.33420
作者: Ruben Karapetyan,Shenyuan Ma,Ales Wodecki,Jakub Marecek
类目: atistics Theory (math.ST); Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Recovering probability measures from their moments has numerous applications, esp. in connection with the method of moments in statistics and optimization. In the setting where measure need not be finitely atomic, but its support is known to be compact and semialgebraic with codimension at least one, the problem is still open. We combine moment-matrix kernel information with the Christoffel–Darboux kernel to provide a discrete approximation of the support. To validate the proposed approach, we test our algorithm on analytically computed moments and pseudo-moments arising from polynomial optimization problems without unique global minimizers. This complements well-known recent work on recovery of measures with algebraic support, where the kernel of a moment matrix can reveal polynomials vanishing on the support, and on recovery of sufficiently regular full-dimensional supports, where estimators constructed by thresholding the Christoffel–Darboux kernel are known to converge asymptotically to the support.
[LG-496] Let CSP Be Your ANCHOR: Adaptive Crystal Search over Frozen Structure Priors
链接: https://arxiv.org/abs/2609.33407
作者: Emma Lei Hovmand,Jonas Elsborg,Melih Kandemir,Arghya Bhowmik
类目: Materials Science (cond-mat.mtrl-sci); Machine Learning (cs.LG); Computational Physics (physics.comp-ph)
*备注: 52 pages, 2 figures, 23 tables
Abstract:De novo crystal generation (DNG) models decide where to search in composition space and how to generate structures with one set of weights. We argue that discovery is better served by separating the two. A crystal structure prediction (CSP) model is a physical prior that should be improved by likelihood training, while rewards, including novelty measured against the search’s own history, should act on a search over compositions. We introduce ANCHOR, a GRPO composition policy trained with multi-objective rewards around a frozen CSP model, and continuous adaptive novelty (CAN), a graded novelty score against known structures and a growing discovery history. Using the frozen CSP model as a fixed ruler under one evaluator, we test where adaptation should act. Replacing DNG compositions with ANCHOR’s policy on the same CSP backbone raises MSUN from 11.4% to 47.6% and SUN from 1.1% to 22.1% at 99.9% formula uniqueness. Fine-tuning DNG models directly on the same rewards instead moves their composition marginal without raising their on-hull fraction. We show that KL-regularized fine-tuning of a DNG model can only reweight chemistry the pretrained model already supports by a bounded factor, while unregularized DNG fine-tunes move toward known or less stable chemistry. Even a stability-only reward routed into ANCHOR’s CSP backbone roughly halves SUN relative to the frozen backbone, whereas likelihood training on structures found during search can improve a CSP backbone. Under MatterGen’s evaluation pipeline, ANCHOR raises state-of-the-art MSUN from 29.2% to 41.3%, transfers without retraining to two further CSP backbones, and reaches 47.1% after distillation into Crystalite-CSP. As with any model optimised against a potential, its on-hull rate depends on that potential.
[LG-497] Local LMO is Secretly a Projection Method!
链接: https://arxiv.org/abs/2609.33383
作者: Peter Richtárik,Ammar Mahran
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:
Abstract:The local linear minimization oracle (Ferris and Zavriev, 1996; arXiv:2605.08850), or Local LMO, solves constrained convex problems without having to compute a projection: it minimizes a linear model over the intersection of the feasible set with a ball around the current iterate. We show that, whenever the ball radius does not exceed the Polyak radius, the Local LMO step is the Euclidean projection of the current iterate onto the intersection of the feasible set with a half-space that separates the iterate from the solution set; this projection is nonetheless computable by a linear oracle alone. We demonstrate that Local LMO belongs to a broader family of projection methods which may be indexed by the depth of the localizing half-space. For objectives with \vartheta -Hölder continuous gradient, every method from this family whose half-space lies sufficiently deep drives the best of its first K iterates to optimality at the universal rate \mathcalO(K^-(1+\vartheta)/2) , matching the non-accelerated universal gradient method of Nesterov (2015). Run at the Polyak radius, Local LMO attains the same rate when the constrained optima are also unconstrained ( |\nabla f(x_\star)| = 0 ), and the rate \mathcalO(K^-1/(2-\vartheta)) otherwise.
[LG-498] Identifying the Predictable Drift of a Semimartingale from Marginal Laws
链接: https://arxiv.org/abs/2609.33372
作者: Jakub Marecek,Enrico Biffis,Abigail Langbridge,Robert Shorten
类目: atistics Theory (math.ST); Machine Learning (cs.LG)
*备注:
Abstract:A special semimartingale admits a unique decomposition X=X_0+M+A into a local martingale M and a predictable finite-variation part A . We consider the identification of A when X is observed only through repeated cross-sections. The estimand is then the projection of the sampled predictable compensator onto the observable feature filtration, namely the current state together with whatever randomness is shared across the population, so that at a fixed diffusion coefficient the marginal flow identifies the drift only up to a Markovian projection. If the drift is an affine functional of an observed lag window, the joint problem is a convex quadratic programme whose solution is the pseudo-panel regression of econometrics. Our principal concern is the case, which we believe not to have been treated before, in which the drift is the output of a hidden linear dynamical system whose dynamics are themselves to be identified from the marginals. The joint problem is then a bilinear quadratically constrained programme, which we solve to certified global optimality by spatial branch and bound; with unpenalised state disturbances and a drift basis growing with the grid it is NP-hard already in latent dimension one, by reduction from \ell^1 rank-one matrix approximation, whereas the complexity of the deterministic system at fixed latent dimension remains open. A block-coordinate decomposition offers a cheaper alternative. For the estimator itself, we obtain rates at a fixed mesh, separated into Monte-Carlo, estimation and grid contributions.
[LG-499] he Statistical Benefits of Multiple Responses for Learning from Demonstrations
链接: https://arxiv.org/abs/2609.33291
作者: Chandramauli Chakraborty,Cong Ma
类目: Machine Learning (stat.ML); Information Theory (cs.IT); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注: 37 pages, 4 tables
Abstract:Many generative systems return multiple candidate responses and are evaluated according to the best one. Recent work shows that, when demonstrations are optimal, pass@ k can reduce the sample complexity of learning from demonstrations by a logarithmic factor in k . We ask what happens when the demonstrator is not assumed to be optimal. We find that multiple responses provide a qualitatively stronger benefit in this setting. In a finite reward-class model with no reward feedback, moving from pass@ 1 to any pass@ k with k\ge2 changes the worst-case dependence on target accuracy from 1/\varepsilon^2 to 1/\varepsilon , uniformly over demonstrator quality. Under standard evaluation, where an unknown reward is fixed before training, increasing k provides an additional and distinct benefit: the optimal dependence on a reward class of size N improves from \log N to \log N/\log k . We further show that these two effects can be separated. Under robust evaluation, where one learned policy must compete with the demonstrator simultaneously for every reward in the class, the fast 1/\varepsilon dependence persists, while the 1/\log k improvement can disappear. We establish matching upper and lower bounds in the corresponding regimes and give a greedy multiplicative-weights learner achieving the upper bounds without any assumption on demonstrator quality.
[LG-500] Schrödinger–Föllm er Actor–Critic: Diffusion Policy Improvement with Finite-Sample Analysis
链接: https://arxiv.org/abs/2609.33239
作者: Yuling Jiao,Lican Kang,Jerry Zhijian Yang,Jincheng Ying
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Diffusion policies represent multimodal action distributions, but an advantage-weighted update does not specify how to sample from the resulting target distribution. We propose Schrödinger–Föllmer Actor–Critic (SFAC), an offline-to-online reinforcement learning (RL) method for Kullback–Leibler (KL)-regularized policy improvement through conditional diffusion. A minimax Bellman critic estimates the advantage function, which defines an exponentially tilted target policy. A Doob h -transform expresses this update as a correction to the reference diffusion drift. We derive a posterior-mean representation of the correction and estimate it using paired self-normalized importance sampling (SNIS). Supervised regression on these drift targets updates the neural actor without critic action gradients. In the small-update regime, the KL-regularized update follows the natural policy-gradient direction, and the Doob correction represents the same local change in the space of diffusion drifts. Under suitable conditions, we derive finite-sample bounds that separate the effects of critic estimation, neural drift regression, finite-sample SNIS, diffusion discretization, and inherited actor error on expected average policy suboptimality. Synthetic experiments assess the accuracy of approximation to prescribed advantage-tilted targets and sensitivity to sampling budgets. On six offline-to-online continuous-control tasks, a reference-anchored implementation achieves higher final-window returns than those of its corresponding offline initialization.
[LG-501] Offline Policy Evaluation via Mixed Bellm an Residuals and Adaptive Critic Representations
链接: https://arxiv.org/abs/2609.33186
作者: Amitakshar Biswas,Yuhan Li,Ruoqing Zhu
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Evaluating a target policy using data generated by a different behavior policy remains a fundamental challenge in reinforcement learning. While most existing work relies on the standard one-step Bellman residual, we consider a convex combination of one-step and two-step residuals with a fixed mixing weight. In ideal settings, this mixed Bellman formulation can provide a natural bias–variance trade-off between approximation error under a restricted value-function class and the increased variance arising from multi-step importance weighting. To solve this mixed residual optimization, we adopt a minimax formulation involving a critic function. Unlike standard approaches that rely on a fixed functional class, we construct a data-dependent critic representation using predicted future feature directions which effectively induces a kernel adapted to the underlying transition dynamics. This allows the critic to focus on directions that are most relevant for the estimated Bellman error. To control overfitting, we use sample splitting to construct the critic and estimate the value function on separate data subsets. Simulation studies and MetaWorld tasks illustrate the effect of the mixing parameter and show that intermediate residual combinations can improve value estimation in challenging settings.
[LG-502] Notes on Generative Modeling for Feedback Control and Planning
链接: https://arxiv.org/abs/2609.33164
作者: Karthik Elamvazhuthi
类目: Optimization and Control (math.OC); Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注:
Abstract:In these notes, we view control as a dynamically constrained sampling problem on the state-space of a control system. With this viewpoint, we extend methods from generative modeling, such as flow matching, normalizing flows and denoising diffusions to control problems. Concepts such as controllability, optimal control and trajectory planning play an important role in guiding the extension and understanding well-posedeness of the corresponding algorithms, with application to steering systems to target states or distributions and sampling from reachable sets. The notes are intended as an accessible introduction for readers with a background in control theory and robotics.
[LG-503] Sharp Critical Minimax Laws and No-Learning Thresholds in Continuous-Time Adaptive Control
链接: https://arxiv.org/abs/2609.33154
作者: Chen Jia
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:
Abstract:We study episodic continuous-time control with an unknown vector control gain, scalar state, quadratic action cost, and smooth convex terminal cost. In the scalar Gaussian experiment, let \Delta(H) denote the minimax improvement over zero control and set \delta=\sqrt2H^2-1 . We prove the critical law \Delta(H)\asymp \delta^4\sqrt\log(1/\delta) \qquad (\delta\downarrow0), with a matching lower and upper bound. The lower bound follows from a uniform deficit–energy inequality valid for fully adaptive controls with unbounded amplitudes, while a moving soft-threshold feedback attains the rate. For local parameters \theta=N^-1/4h , |h|\le H , we show that the normalized minimax regret over N episodes is within O(N^-1/2) of a fixed-horizon Gaussian sequential control problem, without an additional dimension factor. The terminal task enters the limit only through c_g=\operatornameVar(g(Z)) . The Gaussian problem exhibits an exact no-learning phase transition: C_d^T(H)=TH^2/2 \iff H^4T\le d^2/2, yielding the asymptotic task boundary c_gH^4=d^2/2 . Subjects: Optimization and Control (math.OC); Machine Learning (cs.LG) Cite as: arXiv:2609.33154 [math.OC] (or arXiv:2609.33154v1 [math.OC] for this version) https://doi.org/10.48550/arXiv.2609.33154 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Chen Jia [view email] [v1] Sun, 27 Sep 2026 03:26:48 UTC (59 KB) Full-text links: Access Paper: View a PDF of the paper titled Sharp Critical Minimax Laws and No-Learning Thresholds in Continuous-Time Adaptive Control, by Chen JiaView PDFTeX Source view license Current browse context: math.OC prev | next new | recent | 2026-09 Change to browse by: cs cs.LG math References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab) Major funding support from
[LG-504] Can Tabular Foundation Models Amortize Statistical Inference?
链接: https://arxiv.org/abs/2609.33114
作者: Kai Ye,Shijin Gong,Hongyi Zhou,Valentina Zangirolami,Chengchun Shi
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Methodology (stat.ME)
*备注:
Abstract:For decades, statistical inference has largely been developed one problem at a time. Given a scientific target, such as a treatment effect or a regression function, statisticians design a problem-specific estimator together with a procedure for quantifying its uncertainty. This paper proposes a different paradigm. We focus on a classical problem in statistical inference, confidence interval construction, and develop TabCon, an amortized inference system built on a tabular foundation model that produces confidence intervals for new datasets through a simple forward pass. The key methodological ingredients of TabCon are a sparse mixture-of-experts architecture and reinforcement-learning-based post-training that calibrate the resulting confidence intervals to a desired coverage level. Across a wide range of benchmark datasets, TabCon attains near-nominal coverage while producing short confidence intervals. At inference time, it also offers considerably greater computational efficiency, running 50 times faster than the classical bootstrap procedure, even when the latter uses only 50 bootstrap samples.
[LG-505] Byzantine-Robust Federated RAG via Aligned Calibration and Fixed-Membership Conformal Prediction
链接: https://arxiv.org/abs/2609.33037
作者: Prasanjit Dubey,Xiaoming Huo
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Retrieval-augmented generation (RAG) lets language models answer questions more accurately by consulting relevant documents. Many valuable collections, such as medical records, cannot be pooled because of privacy rules. Federated RAG leaves each collection with its owner, or node, which scores candidate answers from its own documents; a central hub combines the scores. Some nodes, called Byzantine, may be compromised, faulty, or misled by instructions hidden in documents, and report arbitrary scores. Conformal prediction returns a set containing the correct answer with a chosen probability, using a cutoff set in a calibration step on questions with known answers. An unknown group of nodes, no larger than a declared bound, may misreport both in this step and at query time. Existing methods assume every node is honest or protect only the calibration step. We observe that the honest nodes are the same in both steps. The hub therefore has all nodes score the same calibration questions, and keeps a candidate only if some plausible group of honest nodes, using its own scores in both steps, would keep it. We prove that the resulting sets contain the correct answer with the chosen probability in finite samples, whatever the Byzantine nodes report. No method using the same information can return smaller sets without risking the loss of an answer the honest nodes support. If nodes fail at random, the guarantee weakens only by the probability that more nodes fail than declared. In simulations, on real question-answering tasks including medical exams, and with language models as nodes, some hijacked, our sets reached the target whenever no more nodes misbehaved than declared, while plain averaging could miss it. They were also clearly smaller than those of simpler methods with the same protection, most of all when the declared bound was generous, so a cautious bound costs little.
[LG-506] Decentralized Optimization with Cross-Coupled Mixed Affine Constraints
链接: https://arxiv.org/abs/2609.33021
作者: Ewsey Obzherin,Ilya Khomchenko,Natalia Shelegeda,Nhat Trung Nguyen,Demyan Yarmoshik,Alexander Rogozin,Alexander Gasnikov
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注: 43 pages, 4 figures
Abstract:We study decentralized optimization with cross-coupled mixed affine constraints, where local and shared variables interact through two affine channels. We show that the intrinsic difficulty of combining separately well-conditioned channels is governed by their Friedrichs angle. This geometry induces a cross-coupling factor that cannot be removed by channelwise preconditioning and governs the additional affine-oracle and communication complexity. We develop an accelerated decentralized method with matching minimax guarantees in the smooth strongly convex regime and extend the framework to smooth and nonsmooth convex objectives. Experiments confirm the predicted dependence on cross-channel geometry and network conditioning.
[LG-507] EEG-Based Motor Imagery BCI Algorithms and Technologies: A Review
链接: https://arxiv.org/abs/2609.32930
作者: Mohammad Hossein Koohi Ghamsari,Seyede Fatemeh Ghamkhari,Siavash Bayat,Ahmed Hemani
类目: ignal Processing (eess.SP); Machine Learning (cs.LG)
*备注:
Abstract:Brain-computer interfaces (BCIs) have emerged as transformative technologies that enable direct communication between the brain and external devices. Among various BCI paradigms, EEG-based motor imagery (MI) has gained prominence due to its simplicity, non-invasiveness, and potential to restore motor function and facilitate rehabilitation for patients with motor impairments. This paper presents a comprehensive review of the most practical processing algorithms developed over the past decade for decoding brain sensorimotor cortex signals. Specifically, this paper discusses the integration of artificial intelligence (AI)-based algorithms, particularly machine learning and deep learning techniques, and their contributions to improving the performance and efficiency of MI-BCI systems in detail. Furthermore, the paper reviews state-of-the-art hardware platforms and emerging converging technologies, including system-on-chip (SoC) architectures, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), wearable devices, the Internet of Things (IoT), and augmented/virtual reality (AR/VR), and discusses their integration with advanced signal processing algorithms to enable next-generation MI-BCI systems. By highlighting current achievements of EEG-based MI-BCI technology and predicting future research directions that could further enhance real-time capabilities, this paper aims to provide valuable insights for researchers and practitioners, fostering innovation in high-performance EEG-based MI-BCI systems.
[LG-508] Machine learning for the LHC physics program: a 2025-2026 stocktake
链接: https://arxiv.org/abs/2609.32874
作者: Jesse Thaler
类目: High Energy Physics - Phenomenology (hep-ph); Machine Learning (cs.LG); High Energy Physics - Experiment (hep-ex)
*备注: 18 pages, 1 figure. Plenary talk at the 14th Large Hadron Collider Physics Conference (LHCP 2026), Paris, 18-22 May 2026. Conversations welcome; tomatoes expected
Abstract:The first sentence of this abstract–and the introduction to these proceedings–was authored by a human, but the bulk of this document was generated by an agentic AI system. In this talk, I take stock of machine learning (ML) for the LHC physics program over the twelve months from May 2025 to May 2026. The corpus is the HEPML Living Review, split at May 2025 into 1,756 earlier papers and 569 later ones. An AI pipeline surveyed the 569 abstracts, ranked them by citations, recency, theme, and collaboration involvement, and read 103 papers in full (95 from after the split, plus 8 earlier baseline papers), producing a structured note for each. The notes were then synthesized into six claims about the state of the field, checked by independent reviewer agents, and re-verified against the source papers. The headline claim is that (1) ML for high-energy physics (HEPML) stopped being a research area that builds tools and became infrastructure that the LHC physics program depends on: ATLAS and CMS now publish physics results that depend on neural networks, and the archived ALEPH data have re-entered production. The other five claims are: (2) simulation-based inference and foundation models are two revolutions starting to merge; (3) AI agents are the genuinely new front, with 47 papers in twelve months and no adopted measurement yet; (4) “do we trust it?” is the fastest-growing agenda, with one recent paper in five about uncertainty, calibration, or interpretability; (5) what is slowing down is informative, since equivariance was absorbed into a tool and model-specific phenomenology ceded ground to model-agnostic searches; and (6) theory ML crossed a capability threshold in multiple research areas. I close with what is settled, what is incoming, and what is open, and briefly discuss the concerns raised by this way of working with AI.
[LG-509] Sliced Orlicz-Wasserstein
链接: https://arxiv.org/abs/2609.32847
作者: Binh Thuan Tran,Khai Nguyen
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST); Computation (stat.CO)
*备注: 62 pages, 4 figures
Abstract:We propose sliced Orlicz-Wasserstein (SOW) distance which is a generalization of sliced Wasserstein (SW) distance. SOW replaces the L^p norm in SW with a Luxemburg norm cost induced by an Orlicz function \phi . First, we prove that SOW distance is a metric on the space of measures with finite Orlicz norm, and show that it recovers the SW distance when the Orlicz function is \phi(x)=x^p . Next, we derive the topological properties of the SOW distance. In particular, we show that convergence under SOW implies weak convergence, and the converse is true under the compact support condition. We then present the theoretical results for estimating the SOW distance. We derive sample complexity for both the distance itself and the powered functional of the distance, and prove their minimax optimality. In addition, we discuss the computational algorithm for approximating the SOW distance by Monte-Carlo estimation and bisection search, as well as the associated approximation error and computational complexity analysis. Our experimental results reveal the superior computational efficiency of SOW compared with Orlicz-Wasserstein (OW) distance. Also, in the experiments, we demonstrate the favorable flexibility of SOW distance over SW in detecting differences between distributions by comparing their performance in two-sample tests and evaluating generative models on image datasets.
[LG-510] Domain Adaptation with Target Information via Doubly-Anchored Distributionally Robust Optimization
链接: https://arxiv.org/abs/2609.32730
作者: David Kepplinger,Anand N. Vidyashankar
类目: Machine Learning (stat.ML); Information Theory (cs.IT); Machine Learning (cs.LG); Probability (math.PR); Statistics Theory (math.ST)
*备注:
Abstract:Domain Adaptation (DA) often lacks worst-case guarantees, while Distributionally Robust Optimization (DRO) based only on source data centers its ambiguity set at the source law and ignores available target structure. To bridge this gap, we introduce a doubly-anchored DRO framework whose ambiguity set is the intersection of \phi -divergence balls centered at the source law and a source-completed target reference law, the latter pairing the target covariate law with the source conditional law. We derive dual-induced adversarial bridge geometries for symmetric and asymmetric divergence pairings, notably introducing a Kullback–Leibler/squared-Hellinger (KL/HD) bridge. This asymmetric formulation yields a Lambert- W geometry in which source-side exponential risk tilting and target-side Hellinger stabilization enter through distinct terms, attenuating, but not bounding, the effect of large likelihood ratios. Furthermore, without imposing covariate shift, we establish finite-sample generalization bounds for the minimizer of a structural, loss-agnostic density-bridge risk under bounded-overlap conditions; these bounds do not apply directly to the loss-aware DRO min–max estimator. We translate our framework into a bridge-weighted Nadaraya–Watson estimator, proving uniform consistency for the source regression function and pointwise asymptotic normality, with target recovery when the source and target regression functions coincide, as under covariate shift. Finally, an empirical evaluation on a domain-shifted Fashion-MNIST dataset illustrates the finite-sample stability of the asymmetric KL/HD bridge under severe synthetic target-covariate corruption.
[LG-511] Revisiting AdaGrad in Stochastic Convex Optimization: Last Iterates High Probability and Lower Bounds
链接: https://arxiv.org/abs/2609.32729
作者: Weiming Ou,Xiao Wang
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注: Under Review
Abstract:AdaGrad and AdaGrad-Norm are widely used adaptive methods, but their precise behavior in stochastic convex optimization remains less understood. We first show that AdaGrad-Norm and AdaGrad do not admit any universal \textbflast-iterate rate, even under sub-Gaussian noise and bounded iterates. We then prove that bounded variance alone is too weak: even with bounded iterates, it cannot yield \textbfhigh-probability average-iterate rates, and without bounded iterates it may not even guarantee convergence in expectation. Besides, we construct tight \Omega(\log T/\sqrt T) lower bounds for both AdaGrad-Norm and AdaGrad under sub-Gaussian noise, showing that \log T in existing average-iterate upper bounds is unavoidable. Finally, we show that this \log T loss disappears once bounded-iterate condition is imposed: under a general ABC condition, both methods achieve rates O(1/\sqrtT) .
[LG-512] Affine Geometry of Gaussian ReLU Networks via Conditional Kac-Rice Formulas
链接: https://arxiv.org/abs/2609.32695
作者: Recep Özkan,Christian Hirsch
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Probability (math.PR)
*备注: 40 pages, 3 figures
Abstract:We study how the affine geometry of finite ReLU networks is created at random initialization and reorganized by supervised training. We call a sign-changing zero of a hidden preactivation an activation switch and a point where the scalar network output is nondifferentiable a scalar kink. For one-dimensional input, conditioning on the preceding layers makes each preactivation Gaussian and affine on the cells of a random finite partition. This yields an exact finite-width conditional Kac-Rice formula for the expected number of activation switches along an input interval. For fixed depth and proportionally growing widths, the resulting switch intensities converge to explicit deterministic limits. A visibility estimate shows that the expected number of switches that do not produce scalar kinks is negligible, yielding an explicit leading formula for the expected number of scalar kinks and hence affine regions. In higher input dimensions d = 2, the analogous conditional surface formula yields the leading expected (d-1)-dimensional Hausdorff measure of the scalar kink set. On the Breast Cancer Wisconsin data, the initialization formula accurately predicts switch counts along held-out segments. After training, switch counts decrease along within-class segments and increase along between-class segments. Thus, training redistributes rather than merely contracts affine complexity.
[LG-513] Learning What to Evaluate: Correlation-Aware Decoupling for Multiobjective Bayesian Optimization
链接: https://arxiv.org/abs/2609.32632
作者: Ashwin Renganathan,Peter Bachman
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 30 pages, 15 figures
Abstract:Multiobjective Bayesian optimization (MOBO) with Gaussian process (GP) surrogates is a sample efficient approach to solving multiobjective optimization problems. In MOBO, a Bayesian decision theoretic acquisition function guides the adaptive selection of new candidate inputs, on which objectives and constraints are evaluated to update the surrogate model sequentially. Existing approaches maintain independent GP models for the objectives and constraints, with new observations evaluating all objectives and constraints in a coupled fashion. However, the objectives and constraints often contain inherent correlations which, if exploited, can enable decoupled evaluations where only a subset of them are evaluated at each round. We present a new approach that leverages a multitask GP model to jointly learn all objectives and constraints, and propose a total correlation metric that enables identifying an optimal subset of objectives and constraints to be evaluated at every round, even under uniform evaluation costs. Theoretically, we show that our acquisition policy is asymptotically consistent despite decoupling and that our proposed decoupled subset selection rule maximizes the expected posterior entropy reduction about unevaluated tasks under mild conditions. Empirically, we show that our approach outperforms coupled and decoupled baselines in the state of the art.
[LG-514] Agnostic Smoothed Online Regression with Adversarial Responses
链接: https://arxiv.org/abs/2609.32478
作者: Xuanyu Chen,Yue Yu
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注: 29 pages, 1 table
Abstract:We study smoothed online prediction with bounded adversarial responses. This widely studied framework bridges i.i.d. sampling and adversarial covariate selection through a smoothness parameter \textsfC_\textsfcov , which bounds conditional covariate densities relative to a fixed, unknown base measure. We propose \textscHedge-Cover, an information-theoretic algorithm that achieves sublinear regret \widetildeO(\sqrt\textPdim(\mathcalF) \textsfC_\textsfcov T) for function classes with bounded pseudo-dimension. The algorithm aggregates a carefully constructed family of experts using \textscHedge, with a prior that links regret to the number of disagreements between a consistent selector and a target function. We bound this number by exploiting covariate smoothness. This answers an open problem posed in \citeblanchard2025agnostic on the minimax optimal adaptive regret of the smoothed online regression problem. We establish a matching lower bound for the class of linear predictors. The main intricacy of the lower bound lies in explicitly constructing a challenging sequential covariate distribution supported on mutually orthogonal hyperplanes. This construction may be of independent technical interest. Finally, we revisit the well-specified setting and quantify the effect of response noise. For conditionally \nu^2 -subGaussian responses, we extend the existing lower bound under realizable responses by showing that the minimax expected regret is \Omega((1\vee \nu)\sqrt(\textsfC_\textsfcov-1)dT) for a function class of VC dimension d . A corresponding upper bound for ERM matches this dependence on \nu .
[LG-515] AECSF: Adaptive Ensemble Conditional Score Filtering for High-Dimensional Nonlinear Data Assimilation
链接: https://arxiv.org/abs/2609.32411
作者: Yangwen Zhang,Shiwei Ni,Xiaoping Zhang,Xiaofei Guan,Lili Ju
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注:
Abstract:Bayesian state estimation for high-dimensional nonlinear dynamical systems entails a fundamental tension between statistical fidelity and computational tractability, as particle weights can collapse, while Gaussian ensemble updates can miss non-Gaussian posterior structure. Score-based diffusion filters offer a sampling-based alternative, but existing training-free score filters often rely on heuristic likelihood corrections, which can compromise posterior accuracy by neglecting uncertainty about the system state associated with each noisy reverse particle. To address these issues, we propose AECSF, a training-free adaptive ensemble conditional score filter. AECSF constructs an analytically tractable score estimator from the conditional Tweedie identity, which recasts noisy posterior score estimation as estimating the conditional mean of the system state given a noisy reverse particle and the observation. To estimate these conditional means efficiently, AECSF employs a shared adaptive weighted proposal ensemble, while particle-specific conditional weights yield an estimate for each noisy reverse particle without separate proposal sampling. The proposal ensemble is updated using reverse-particle information within the same reverse-diffusion run to improve conditional-mean estimation. Theoretically, we characterize when a fixed weighted proposal measure yields the exact noisy posterior score. Under stated assumptions, we establish a bound relating conditional-mean estimation errors to reverse-sampling endpoint error. Numerical experiments demonstrate that AECSF improves the accuracy of posterior sampling and nonlinear filtering in high-dimensional problems with limited forecast ensembles.
[LG-516] Convergent Plug-and-Play Image Restoration with Annealed Noise Levels
链接: https://arxiv.org/abs/2609.32393
作者: Samuel Hurault
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:
Abstract:Plug-and-Play (PnP) methods solve imaging inverse problems by incorporating deep denoisers into iterative optimization algorithms. Although practical implementations often decrease the denoiser noise level \sigma along iterations, most existing convergence analyses assume a fixed denoiser. In this work, we establish convergence guarantees for a broad family of Plug-and-Play algorithms with annealed noise level, spanning deterministic methods (RED–GD and PnP–PGD) and stochastic methods (SNORE, equivariant RED, and a variant of PnP–Flow). For each method, we identify an explicit, nonconvex objective associated with the terminal denoising level and prove asymptotic stationarity of the iterates with respect to this objective. Our analysis does not prescribe any decay rate for the noise schedule, and our assumptions cover both learned gradient-step denoisers and exact MMSE denoisers. Overall, our theoretical results bridge the gap between existing PnP convergence theory and the decreasing-denoising practices used by state-of-the-art image restoration methods. We empirically demonstrate the benefits of such schedules and illustrate the predicted convergence behavior on several imaging inverse problems, including inpainting, super-resolution, demosaicing and tomography.
[LG-517] Overfitting of Spectral Gradient Descent: How Matrix Geometry shapes Generalization and Implicit Bias
链接: https://arxiv.org/abs/2609.32270
作者: Guillaume Braun,Ichiro Hashimoto,Masaaki Imaizumi
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We study the generalization of spectral gradient descent (SpecGD) in overparameterized matrix classification with corrupted labels. Each input combines a shared low-rank signal with a rank-one sample-specific perturbation, referred to as a shortcut, that enables memorization but does not generalize. We contrast collapsed shortcuts, which share a singular direction, with dispersed shortcuts, which occupy distinct singular directions. Changing only this geometry can reverse the relative generalization of GD and SpecGD: collapsed shortcuts can favor SpecGD, while dispersed shortcuts can favor GD. In the dispersed regime, exact shortcut orthogonality eliminates the signal from the late-stage SpecGD direction, while vanishing random correlations collectively generate a small but generalization-relevant signal through a second-order effect. To identify the direction selected by SpecGD, which the spectral max-margin problem alone does not determine, we combine a refined analysis of its dual with the exponentiated-gradient dynamics of normalized loss weights. Finally, we show that a single SpecGD step can already interpolate and generalize well, while continued training converges to a direction with substantially worse generalization.
[LG-518] Active Data Acquisition with Side Information via Discrete Diffusion Priors
链接: https://arxiv.org/abs/2609.32252
作者: An Vuong,Thinh Nguyen
类目: Image and Video Processing (eess.IV); Machine Learning (cs.LG)
*备注: 20 pages, 16 figures
Abstract:Acquiring data is costly: higher measurement fidelity costs power and storage and risks collecting irrelevant content, while aggressive cost reduction can discard information that later analysis needs. We address this trade-off with an information-theoretic framework that acquires data relevant to a broad set of tasks rather than to one model. A mask policy, conditioned on side information, chooses which pixels to measure so as to maximize the mutual information between a discrete image and its partial observation under a budget; since the image entropy does not depend on the mask, this is equivalent to minimizing the conditional entropy. A frozen discrete denoising diffusion model (D3PM) supplies the posterior, and we use it in two ways: as an entropy surrogate for training a one-shot mask generator, and as the criterion for sequential greedy acquisition. The one-shot generator outperforms random masks only with care, including an unbiased gradient estimator for binary masks. With sequential acquisition, on MNIST the prior makes 8\times fewer errors than random at a 10% budget, and on CIFAR-10 it gains 0.9 – 3.4 ~dB. On fastMRI, our proposed technique using a static mask outperforms the well-known methods such as variable density and LOUPE.
[LG-519] High-Probability Guarantees for SGD under β-Heavy-Tailed Gradient Noise
链接: https://arxiv.org/abs/2609.32195
作者: Qijun Tong,Masahiro Ikeda,Ryota Kawasumi
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Stochastic gradient descent (SGD) is widely used to train machine learning models, but subsampling the training data introduces noise into its updates. The strength and applicability of high-probability guarantees therefore depend critically on how the tails of gradient noise are modeled. Reports of heavy-tailed gradient noise in deep learning motivate relaxing the bounded-noise and sub-Gaussian assumptions commonly used in high-probability analyses of SGD. We use Young functions from Orlicz space theory to describe noise tails in a common framework. We model SGD gradient noise by adopting a Young function that preserves the finiteness of all polynomial moments while allowing tails heavier than sub-Weibull, including lognormal distributions. The resulting class is called \beta -heavy-tailed, with \beta controlling the tail heaviness. We establish concentration inequalities for \beta -heavy-tailed noise and combine them with a uniform bound on the difference between empirical and population gradients along the SGD trajectory to obtain high-probability bounds on optimization and population-risk stationarity for smooth nonconvex losses under trajectory assumptions. The bounds are not restricted to a particular learning-rate decay rule and make explicit the effects of noise tails and learning-rate schedules. Under the Polyak-Łojasiewicz condition, we bound the risk at the last iterate. We also analyze SGD with gradient clipping under the \beta -heavy-tailed noise model.
[LG-520] Uncertainty Quantification of Next Generation Reservoir Computing with Applications to Memory-Driven Dynamical Systems
链接: https://arxiv.org/abs/2609.32169
作者: Livia Popa,Sumanta Basu,Martin T. Wells
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Nonlinear dynamical systems with memory arise across science and engineering, yet uncertainty quantification for efficient forecasting methods such as Next Generation Reservoir Computing (NGRC) remains underdeveloped. We study Bayesian ridge and conformal prediction intervals for NGRC and characterize when their uncertainty estimates agree or differ. In low dimensions, their asymptotic widths are governed by different summaries of the residual distribution, so agreement depends on residual shape rather than dimensionality alone. In high dimensions, regularization introduces a further tradeoff between estimation variance, shrinkage bias, and posterior uncertainty, leading to an explicit transition between regimes where Bayesian intervals are wider or narrower than conformal intervals. We extend these results to quadratic NGRC feature maps and give sufficient conditions for transferring the analysis to temporally dependent forecast windows. Simulations and real-data experiments support the theoretical predictions and illustrate how residual distribution, regularization, dimensionality, and distribution shift affect interval calibration and efficiency. These results provide a principled framework for choosing and interpreting uncertainty quantification methods in reservoir-based forecasting.
[LG-521] FOCUS: Fixed-Confidence Online Causal Learning Using Sequential Adaptive Interventions
链接: https://arxiv.org/abs/2609.32165
作者: Haijie Xu,Chen Zhang
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We study fully online fixed-confidence causal discovery without any historical observational data. Starting from zero samples, the learner sequentially selects interventions to recover both the causal DAG and its edge weights under a linear-Gaussian structural equation model. We establish an instance-dependent lower bound for any (\epsilon,\delta) -correct algorithm and propose \textscFOCUS, which adaptively allocates interventions through an online max–min game. A key contribution is a computable concentration inequality for the accumulated KL divergence involving causal parameters shared across interventions. We prove that \textscFOCUS is (\epsilon,\delta) -correct and that its expected stopping time matches the lower bound in its \Theta(\log(1/\delta)) dependence up to an instance-dependent constant. Experiments demonstrate improved structure and edge-weight recovery and confirm the predicted stopping-time trend. Our codes are available on this https URL
[LG-522] A Unified Optimism-Agnostic Framework for Linear Bandits over Spherical Action Sets
链接: https://arxiv.org/abs/2609.32149
作者: Arda Güçlü,Subhonmesh Bose,John R. Birge
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 27 pages, 2 figures
Abstract:Linear bandits model sequential decision-making problems with noisy rewards that are linear in the decision variable, where an agent must simultaneously learn about an unknown parameter that governs the mean rewards, while maximizing (expected) rewards over time. Two prominent algorithmic families–upper confidence bound (UCB) and Thompson sampling (TS)–achieve a balance of exploration (to estimate said parameter) and exploitation (utilization of knowledge about it) across time. The quality of estimation of that parameter depends on the eigenvalues of a design matrix. In this paper, we begin by showing that if the inference quality obtained from exploration, encoded in the minimum eigenvalue of the design matrix, grows \gtrsim \sqrtt with time t , while actions remain sufficiently concentrated for exploitation, then an algorithm produces optimal high-probability \mathcalO(\sqrtT\log T) -regret rate over a time-horizon T for spherical action sets. This analysis is algorithm-agnostic and follows an alternative route to the classical optimism-based elliptical-potential argument for regret analysis. Then, we illustrate that variants of UCB and TS satisfy the inference and concentration properties and in turn, enjoy optimal regret rate. In effect, our results provide a modular framework that can be used to analyze linear bandit algorithms and explicitly connect quality of parameter estimation to optimal regret accumulation.
[LG-523] Correcting the Dropout-LayerNorm Expectation Gap Improves Protein Structure Models
链接: https://arxiv.org/abs/2609.32062
作者: Isaac Ellmen,David Errington,Matthew I.J. Raybould,Charlotte M. Deane
类目: Biomolecules (q-bio.BM); Machine Learning (cs.LG)
*备注:
Abstract:Although \mathbbE[\mathrmdropout(x)] = x , here we show that \mathbbE[\mathrmLayerNorm(\mathrmdropout(x))] is not equal to \mathrmLayerNorm(x) . Accordingly, the pattern of a Dropout layer followed by a LayerNorm, which is common to many AlphaFold2-based protein structure predictors, produces a systematic bias at evaluation time that can hamper performance. To address this, we derive a closed-form, first-order correction for this gap, which we call a Dropout-LayerNorm Correction (DLC). DLC empirically matches the performance boost of large Monte Carlo dropout ensembles. We evaluate its effect across nine protein structure models (ESMFold, OpenFold, ABB3, FlashABB, Ibex, NbForge, Genie1, Genie2, Genie3) on both paired and single-chain antibody structures as well as one protein-ligand docking model (QuickBind). The correction is computationally negligible and improves accuracy in all ten models tested ( \sim0.3%-13% ), with a modest but consistent improvement in ESMFold and OpenFold and a substantial improvement in antibody-specific models. This work identifies the mathematical consequence of chaining together Dropout and LayerNorm and provides a free, principled adjustment to improve the evaluation performance of many pretrained models. Code to reproduce the experiments is available at this https URL.
[LG-524] Continuous-Time Trajectory Generation from Discrete Observations with Stochasticity
链接: https://arxiv.org/abs/2609.32026
作者: Ruifeng Shang,Shu Liu,Yuhua Zhu
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Dynamical Systems (math.DS)
*备注:
Abstract:Physical systems evolve continuously in time, yet their states are typically observed only at discrete times. Generating trajectories consistent with their probability densities from such observations therefore requires capturing the continuous-time evolution rather than only learning transition mappings between consecutive observations. We propose PhiBE-Flow, a framework that directly estimates the probability velocity field induced by the stochastic differential equation (SDE) which governs this continuous-time distributional evolution. PhiBE-Flow learns from discrete observations using a model-free approach requiring neither known SDE coefficients nor score estimation. We establish convergence guarantees for the method, accounting for both time-discretization and finite-sample errors. We evaluate PhiBE-Flow on systems of increasing complexity, from controlled stochastic numerical systems to Navier–Stokes dynamics and real-world videos. Our results show that PhiBE-Flow accurately recovers probability flows of stochastic dynamics, preserves multiscale physical statistics, and improves video generation performance over representative baselines. The code is available at this https URL.
[LG-525] Identifiability Limits of Gravitational Wave Phase Deviations: Multiclass Classification with a Multihead Neural Network
链接: https://arxiv.org/abs/2609.31993
作者: Lavinia Heisenberg,Shayan Hemmatyar
类目: General Relativity and Quantum Cosmology (gr-qc); Machine Learning (cs.LG); Data Analysis, Statistics and Probability (physics.data-an)
*备注: 28 pages, 7 figues, 6 tables
Abstract:We study how well gravitational-wave phase deviations can be identified using a neural network with classification and regression heads. The classification head distinguishes general relativity (GR) from six parametrized post-Einsteinian phase families with exponents b\in-7,-5,-3,-1,+1,+2\ ; the regression head predicts the logarithm of the coupling magnitude, \log_10|\beta| . The network input is a response function quantifying the sensitivity of waveform mismatch to waveform deformations. We use stationary Gaussian noise with the Advanced LIGO design spectrum for 262 sources and a measured Hanford spectrum from the third observing run (O3) for 132 sources, and also test injections into recorded Hanford strain. On the two Gaussian datasets, the network identifies the exponent of deviations clearly above the noise with accuracies of 0.427 and 0.329 . To interpret these results, we compare the network with an approximate classifier built from predicted waveform residuals for the six phase families and same sources. The two classifiers often confuse the same families. Examining residuals left by different phase corrections, we link these errors to similarities between the underlying deviations. On selected recorded Hanford strain, networks trained on simulated or recorded noise achieve overall classification accuracies close to those on simulated Hanford noise. Of thirteen neural-network variants tested on the Advanced LIGO design dataset, the recurrent network has the highest mean classification accuracy for loud deviations, although all variants remain below the approximate classifier. The network nearly matches that classifier’s accuracy, indicating that identification is limited by similarities between neighboring phase families rather than by the network, for known intrinsic source parameters.
[LG-526] Bridging Stochastic Flow Maps and Boltzmann Generators with Normalizing Flows
链接: https://arxiv.org/abs/2609.31978
作者: Louis Grenioux,RuiKang OuYang,Luhuan Wu
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: Under review
Abstract:Generating independent, equilibrium samples of molecular systems at scale remains a central obstacle in computational statistical mechanics. Boltzmann Generators address this by pairing a generative model with importance sampling to obtain consistent samples from the target distribution. We introduce Normalizing Flow Flow Maps (NF ^2 M), which combines the strengths of recent stochastic flow maps with the tractability of classic normalizing flows to build a Boltzmann Generator. Unlike most methods, which correct the generative model only at the end, NF ^2 M reweighs each denoising transition as generation proceeds, avoiding wasted compute on trajectories that are ultimately discarded. At each denoising step, a conditional normalizing flow proposes clean configurations given the current noisy state (a simpler task than sampling directly from the target) and its exact likelihood enables correcting each proposal toward the true denoising transition of the target Boltzmann distribution. This is in contrast to most existing methods, whose likelihoods are approximate or expensive to evaluate, undermining the statistical reliability of the correction. We establish consistency of the corrected transitions and bound how approximation errors propagate through the sampling chain. We evaluate NF ^2 M on peptide systems, demonstrating improved sampling efficiency and sample quality.
[LG-527] Artificial Neural Network Assisted Modelling of Tangent Galvanometer Measurements for the Determination of Horizontal Component of Earths Magnetic Field
链接: https://arxiv.org/abs/2609.31930
作者: Saralasrita Mohanty,Sudakshina Prusty,Anshuman Pal,Pradipta Kumar Mishra
类目: Physics Education (physics.ed-ph); Machine Learning (cs.LG)
*备注: 11 pages, 8 figures
Abstract:The Tangent Galvanometer (TG) is a standard undergraduate laboratory experiment for estimating the horizontal component of Earth’s magnetic field (BH) by measuring the angle of deflection of a magnetic needle corresponding to the current flowing through a circular coil. In this study, an artificial neural network (ANN) is used as a complementary data-driven model to predict the value of BH. A dataset comprising 225 observations obtained using 50-turn and 500-turn coils was used for developing the ANN model. After quality control, 223 observations were retained and divided into training (70%), validation (15%), and testing (15%) subsets. The model was optimized using a feed-forward neural network with Tanh activation. Three different models (Models A, B, and C) with different input variables were compared for optimum performance. Model C used five input variables: current, deflection angle, tan(theta), magnetic field produced by the coil, and number of turns. The addition of tan(theta) produced a substantial improvement in prediction performance, which was further improved by including the magnetic field produced by the coil. Model C gave the best test performance, with R2 = 0.99053, RMSE = 0.54076 microT, and MAE = 0.32940 microT. The experimental and ANN-predicted values of BH were also compared with an adopted local geomagnetic reference value of 39.0 microT. The mean experimental and ANN-predicted values were 37.38898 microT and 37.32161 microT, respectively. The results demonstrate the usefulness of ANN as a complementary tool for analyzing experimental variability and nonlinear relationships in an undergraduate physics laboratory experiment.
[LG-528] Knowledge-Driven XRD Phase Identification via Multi-View Retrieval and Explanation
链接: https://arxiv.org/abs/2609.31888
作者: Doaa Mohamed,Markus Stricker
类目: Materials Science (cond-mat.mtrl-sci); Machine Learning (cs.LG)
*备注: 12 pages, 5 figures, conference proceeding
Abstract:X-ray diffraction (XRD) is a experimental technique for determining the phase composition and structure of crystalline materials. However, interpreting XRD patterns is challenging, particularly in high-throughput materials discovery, where many novel materials may need to be characterized and no reference patterns are available. Consequently, machine learning is increasingly used to accelerate and automate the analysis while reducing errors associated with human interpretation. We propose a multi-decision framework for XRD phase analysis that integrates representation learning, similarity-based retrieval, and explainable decision support within a unified reference database. A convolutional autoencoder learns compact latent representations of XRD patterns that preserve structural similarity while remaining robust to variations arising from experimental noise and measurement conditions. By integrating multiple decision pathways within a shared latent space, the framework moves beyond single-label prediction toward ranked and interpretable phase analysis that mirrors expert practice. During inference, complementary decision mechanisms are applied, including latent-space classification and retrieval, explanation-guided similarity using Integrated Gradients, and composition-based similarity search. These mechanisms generate ranked candidate phase lists that are aggregated into a final prediction with an associated confidence score. Experiments on synthetic datasets demonstrate strong predictive performance, achieving 98.85,% accuracy for crystal system classification and 95.82,% accuracy for space group prediction on the test set, while maintaining robustness under realistic perturbations. The framework supports reliable, analyst-friendly identification of crystal phases and structures in high-throughput and exploratory materials discovery settings.
[LG-529] Prompting Particle Physics: Tokenized Multi-modal Foundation Models for Combinatorially Many Tasks
链接: https://arxiv.org/abs/2609.31862
作者: Nilotpal Kakati,Daniel Murnane,Baran Hashemi,Samuel Klein,Jeffrey Krupa,Eilam Gross,Lukas Heinrich,Michael Kagan
类目: High Energy Physics - Phenomenology (hep-ph); Machine Learning (cs.LG); High Energy Physics - Experiment (hep-ex); Data Analysis, Statistics and Probability (physics.data-an)
*备注: 47 pages. Code: this https URL ; data and models: this https URL
Abstract:Reconstruction and simulation at a collider experiment are long chains of specialised algorithms, each tuned to a single step. We explore how one model can serve many of those steps at once, while still producing the intermediate objects (tracks, calorimeter cells, clusters, particles and jets) that make the chain interpretable. To do so, we represent every object in a jet in a single shared token vocabulary and train one model to map any subset of these modalities to any other. A task is then only a choice of which modalities to provide and which to request: particle flow, detector simulation and charged energy subtraction are all directions through the same set of weights. We train over all modality combinations, with a decoder emitting tokens either autoregressively or in parallel. With tokenisation, both architectures train stably with little tuning. In evaluations, both models produce realistic reconstruction and simulation objects, with the autoregressive model particularly faithful to output from Geant4. The autoregressive model also outperforms a state-of-the-art particle flow algorithm on many typical jet reconstruction metrics.
[LG-530] Convergence-Aware Pareto Selection of Covariate Scaling Transformations for Markov Deterioration Hazard Models: Evidence from Bridge Inspection Data
链接: https://arxiv.org/abs/2609.31786
作者: Takato Yasuno,Keita Kobayashi,Ryuta Sakaguchi
类目: Methodology (stat.ME); Machine Learning (cs.LG)
*备注: 19 pages, 2 figures, 5 tables
Abstract:Maximum likelihood estimation of Markov deterioration hazard models for infrastructure asset management is sensitive to the numerical conditioning of explanatory covariates, yet production pipelines often adopt a single scaling convention without systematic justification. We study four covariate scaling transformations—baseline max scaling, min-max scaling, z-score scaling, and Box-Cox transformation with a training-derived positivity shift—applied to an Exponential Hazard Markov (EHM) model estimated via L-BFGS-B on bridge inspection transition data.
[LG-531] Cross-Modal Knowledge Distillation for Acoustic Pedestrian Detection
链接: https://arxiv.org/abs/2609.31785
作者: Yonghyun Kim,Chaeyeon Han,Sancho Gatungay,Subhrajit Guhathakurta,Alexander Lerch
类目: Audio and Speech Processing (eess.AS); Machine Learning (cs.LG); Sound (cs.SD)
*备注: 5 pages, 1 figure
Abstract:Audio-only pedestrian detection is attractive for urban sensing but limited by weak acoustic cues. An appealing strategy is cross-modal knowledge distillation, in which a video teacher supervises the audio student during training so that the deployed model runs on audio alone. Under this task’s severe class imbalance and wide video-audio modality gap, however, what such distillation contributes is unclear. We introduce Trust-Filtered Distillation (TFD), which selectively suppresses teacher supervision on pedestrian samples, and interpret its logit formulation as conditional label smoothing under a shared temperature. Across ten distillation configurations under five-fold cross-session validation on ASPED, most methods yield modest gains in macro accuracy accompanied by small changes in PR-AUC. The main effect is higher no-pedestrian accuracy at the cost of lower pedestrian recall. Adding TFD to logit distillation strengthens this trade-off without improving mean PR-AUC. These findings clarify the benefits and limitations of selective cross-modal supervision by distinguishing operating-point shifts from discrimination gains in imbalanced acoustic detection.
[LG-532] Suitable Measures for the Potential Operational Utility of AI NWP Rainfall Forecasts Over Africa
链接: https://arxiv.org/abs/2609.31775
作者: Shruti Nath,Docko Sow,Koomi Toussaint Amoussouvi,Fenwick Cooper,Josiah Kiarie Kimani,John Bagiliko,Florian Pappenberger,Rendani Mbuvha
类目: Applications (stat.AP); Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph)
*备注: 19 pages, 11 figures (excluding supplementary and references)
Abstract:Artificial intelligence (AI)-based weather prediction is approaching the skill of physical numerical weather prediction (NWP) systems at a fraction of the computational cost. This is particularly promising for Africa, where rainfall extremes are intensifying and many forecasting centres lack the infrastructure to run physical models at extended lead times. We present a calibrated comparison of GraphCast, GenCast and the Functional Generative Network (FGN) against the physical NWP model IFS for rainfall prediction across Africa. Deterministic and probabilistic forecasts are postprocessed using Isotonic Distributional Regression and evaluated with the Continuous Ranked Probability Score against IMERG, RFEv2 and CHIRPS across seasons, wet and dry regimes, elevation zones and lead times. All models retain skill beyond climatology across most seasons and at extended lead times. AI models generally outperform IFS in wet regions, whereas IFS performs better in dry, high-elevation areas, where its finer resolution better represents orographic controls on rainfall. Across observational datasets and seasons, AI models achieve a median improvement of approximately 5% over IFS. GraphCast achieves calibrated skill comparable to the ensemble-based FGN, although FGN provides greater significant skill at longer lead times. These results highlight the potential of calibrated AI weather prediction to provide accessible and computationally efficient rainfall forecasts across Africa, while demonstrating the continuing importance of spatial resolution, ensemble design and regional characteristics.
[LG-533] Acoustic domain shift in spoken language identification from systematic domain generalization evaluation to real-world application
链接: https://arxiv.org/abs/2609.31759
作者: Francois Derrida(X),Raphaël Duroselle(X),Thomas Courtat,Jean-François Bonastre(X)
类目: Audio and Speech Processing (eess.AS); Machine Learning (cs.LG); Sound (cs.SD)
*备注:
Abstract:Domain Generalization (DG) aims to develop models that remain robust to conditions unseen during training. While DG has been systematically studied in computer vision through controlled benchmarks and diverse distribution shifts, its evaluation in spoken language recognition remains less structured. Existing speech datasets provide valuable benchmarks for robustness to real-world acoustic conditions, but are primarily designed around specific scenarios and large scale rather than as general-purpose tools for systematically constructing and evaluating domain shifts. In this work, we introduce a smallscale speech dataset and evaluation protocol for controlled studies of acoustic domain shifts. It enables the evaluation of spoken language recognition models under a variety of acousitc domain shifts. We introduce the speech modality into the DomainBed Domain Generalization framework and evaluate three domain generalization algorithms. We show that in-domain performance is not a reliable predictor of cross-domain robustness and verify that explicit domain invariant algorithms such as MMD or DANN algorithms do not outperform ERM. We further validate the generality of these findings on MMS-LID-126, a state-of-the-art spoken language identification system. We release the code.
[LG-534] Accurate Sampling from Diffusion Models
链接: https://arxiv.org/abs/2609.29902
作者: Dénes Sexty
类目: High Energy Physics - Lattice (hep-lat); Disordered Systems and Neural Networks (cond-mat.dis-nn); Machine Learning (cs.LG)
*备注: 12 pages, 6 figures
Abstract:A new proposal called DM-SMC (Diffusion Model - Sequential Monte Carlo) is investigated, which samples ensembles defined in terms of an action, using diffusion models trained on samples from the ensemble. The SMC setup allows for accurate sampling in spite of an approximate diffusion model and the finite stepsize used in the numerical solution of the stochastic process. Improved update strategies are also investigated. Results are presented for a Z_2 symmetric scalar field theory in 2 dimensions near its 2nd order phase transition.
[LG-535] FUSION: a skill-based research agent for publicly obtainable nuclear-physics codes
链接: https://arxiv.org/abs/2609.04742
作者: Jin Lei
类目: Nuclear Theory (nucl-th); Machine Learning (cs.LG); Computational Physics (physics.comp-ph)
*备注: 11 pages, 10 figures. Platform at this https URL (MIT), documentation at this https URL
Abstract:Running an unfamiliar nuclear-physics code is rarely difficult because of the physics alone. One must find and build the program, learn its input conventions, and decide whether a plausible output is actually correct. A general-purpose coding agent helps with the first two tasks but may make the last one harder: it can write an input file that runs with the wrong physical convention. FUSION addresses this problem with code-specific skills. A skill obtains the code from its public source, starts from a verified input, runs and parses the calculation, records known failure modes, and must reproduce a stated benchmark to a stated tolerance before reporting a result. The current release covers twenty codes, spanning optical models and reactions, nuclear structure, fission and statistical models, astrophysics and R-matrix analysis, and heavy-ion transport. It also includes an offline, searchable collection of 61 167 pages derived from the nucl-th literature. User notes and credentials remain outside the public repository. FUSION is available under the MIT license at this https URL documentation is at this https URL. Here I describe the design, the checks behind the current release, and one complete calculation from input to comparison with measured data.
[LG-536] Exterior complex scaling enables physics-informed neural networks for quantum scattering
链接: https://arxiv.org/abs/2602.04553
作者: Jin Lei
类目: Nuclear Theory (nucl-th); Machine Learning (cs.LG); Computational Physics (physics.comp-ph)
*备注:
Abstract:Physics-informed neural networks (PINNs) have emerged as a powerful tool for solving differential equations, yet their application to nuclear scattering has been hindered by the oscillatory, non-decaying nature of scattering wave functions. In this work, I demonstrate that exterior complex scaling (ECS) transforms scattering boundary conditions into exponentially decaying waves suitable for neural network solutions, enabling PINNs to solve nuclear reaction problems for the first time. I develop a driven-equation formulation where the source term is confined to the real axis, avoiding the need to analytically continue nuclear potentials into the complex plane. The method is validated on nucleon-nucleus scattering (n+ ^40 Ca at E_\textlab=20 ~MeV) with 21 partial waves, achieving phase shift accuracy of \Delta\delta \lesssim 0.1^\circ for the strongly absorbed channels ( \ell \leq 4 ) and \Delta\delta \leq 0.60^\circ for all channels up to \ell = 10 , when compared to conventional solvers. I further demonstrate the approach on heavy-ion scattering ( ^6 Li+ ^208 Pb at 40~MeV) with 41 partial waves and strong Coulomb effects, where an auto-adaptive anchor warm-down for weak-source channels yields a mean S-matrix accuracy of |\Delta|S_\ell|| \approx 3 \times 10^-3 across the full angular momentum range, including the absorption-to-transparency transition region. This work establishes the foundation for extending PINNs to inverse problems where end-to-end differentiability enables direct fitting of optical potential parameters, coupled-channel reactions, and few-body scattering where traditional grid methods face exponential scaling.
[LG-537] Bidirectional Neural Networks for Global Nucleon-Nucleus Optical Model Calculations
链接: https://arxiv.org/abs/2512.22500
作者: Jin Lei
类目: Nuclear Theory (nucl-th); Machine Learning (cs.LG)
*备注:
Abstract:Modern nuclear data evaluation increasingly requires not only accurate scattering calculations, but also efficient methods for uncertainty quantification and parameter optimization, tasks that benefit from differentiable solvers amenable to gradient-based algorithms. I present a neural network emulator based on Bidirectional Liquid Neural Networks (BiLNN) that provides a fully differentiable mapping from optical potential parameters to scattering wave functions. The key innovation enabling generalization across the parameter space is the use of phase-space coordinates \rho = kr that normalize the oscillation wavelength regardless of projectile energy, allowing a single network to span 1 to 200~MeV. Trained on Numerov solutions for twelve target nuclei (\nuc12C to \nuc208Pb), both protons and neutrons, and partial waves up to l=30 , the network achieves an overall relative error of 1.2%. The predicted wave functions yield accurate S -matrix elements and elastic scattering cross sections, reproducing diffraction patterns spanning four orders of magnitude. Importantly, the model extrapolates successfully to nuclei not included in training (\nuc24Mg, \nuc63Cu, \nuc184W) with comparable accuracy, demonstrating that it has learned the physics of the optical model rather than memorizing specific targets. The differentiable nature of the trained model opens the door to gradient-based optimization of optical model parameters and efficient uncertainty quantification.
[LG-538] Kolmogorov-Arnold networks in nuclear binding energy prediction
链接: https://arxiv.org/abs/2407.20737
作者: Hao Liu,Jin Lei,Zhongzhou Ren
类目: Nuclear Theory (nucl-th); Machine Learning (cs.LG)
*备注: Published version. 10 pages, 9 figures, 2 tables
Abstract:This study explores the application of Kolmogorov-Arnold networks (KANs) in predicting nuclear binding energies, leveraging their ability to decompose complex multiparameter systems into simpler univariate functions. By utilizing data from the Atomic Mass Evaluation (AME2020) and incorporating features such as atomic number, neutron number, and shell effects, KANs achieved a significant lower root mean square error (0.26 MeV), surpassing traditional models. The symbolic regression analysis yielded simplified analytical expressions for binding energies, aligning with classical models like the liquid drop model and the Bethe-Weizsaecker formula. These results highlight KANs’ potential in enhancing the interpretability and understanding of nuclear phenomena, paving the way for future applications in nuclear physics and beyond.
附件下载
