本篇博文主要内容为 2026-10-16 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 (2026-10-16)
今日共更新966篇论文,其中:
- 自然语言处理共180篇(Computation and Language (cs.CL))
- 人工智能共286篇(Artificial Intelligence (cs.AI))
- 计算机视觉共166篇(Computer Vision and Pattern Recognition (cs.CV))
- 机器学习共308篇(Machine Learning (cs.LG))
- 多智能体系统共22篇(Multiagent Systems (cs.MA))
- 信息检索共14篇(Information Retrieval (cs.IR))
- 人机交互共21篇(Human-Computer Interaction (cs.HC))
多智能体系统
[MA-0] Learning Self-Orchestration in Decentralized Agent Swarms
【速读】：该论文旨在解决多智能体系统中个体能力与集体协调机制分离导致的协同效率低下问题，即传统方法通常在固定架构下分别训练智能体的行为策略与组织结构，难以实现动态适应。其解决方案的关键在于提出一种基于市场的后训练（Market-based Post-training, MPT）框架，使去中心化智能体群体能够在无需中央协调者的情况下，联合学习动作策略与自组织参与机制。具体而言，智能体通过出价表达参与意愿，出价结果决定行动主体，同时集体成果为个体提供局部训练信用。这种联合学习机制使得个体能力与参与模式能够协同演化，显著提升了系统的可扩展性、鲁棒性与任务迁移能力。实验表明，采用相同后训练动作策略时，学习到的自组织机制优于其他协调策略，且在面对智能体失效或群体规模扩大时表现更优，验证了所提方法在实际应用中的有效性。
链接: https://arxiv.org/abs/2610.22322
作者: Guowei Xu,Zhenting Qi,Amani Kiruga,Himabindu Lakkaraju,Yilun Du
机构: Harvard University(哈佛大学)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:The capabilities of an agent swarm depend on both its individual members and how they coordinate. Yet existing multi-agent systems often train agents within fixed orchestration structures, separating the learning of individual capabilities from the organization of collective behavior. We introduce market-based post-training (MPT), a framework for jointly learning action policies and self-orchestration in decentralized agent swarms. Agents express their willingness to participate through bids, which determine who acts and how collective outcomes provide local training credit. Learning both action and bidding behavior allows individual capabilities and patterns of participation to adapt together without a central coordinating agent. We evaluate MPT on embodied crafting, robot navigation, and deep research. Results show gains over monolithic post-training and confirm that agent participation indeed evolves with learning. With the same post-trained action policies, learned self-orchestration outperforms alternative coordination strategies. Ablations show that both action and bidding benefit from learning, and that learned bids provide more effective action-training signals. Beyond the original training objectives, trained swarms adapt faster to related tasks and achieve better performance with larger populations at inference time. Larger populations also suffer smaller performance losses under agent dropout, demonstrating that learned self-orchestration supports reuse, scaling, and robustness in decentralized agent swarms. Code and trained models are available at this https URL.
[MA-1] Learning to Steer LLM Agents and Multi-Agent Systems
【速读】：该论文旨在解决大语言模型（Large Language Model, LLM）驱动的智能体及多智能体系统在执行长时程任务过程中出现的目标漂移（goal drift）问题。目标漂移表现为系统在执行过程中逐渐偏离原始任务目标，导致资源浪费与错误结果。为应对这一挑战，论文提出一种动态引导智能体（steering agent），通过实时观测执行上下文，在检测到漂移时以低成本的快速路径（fast path）决定是否干预，并以高成本的慢速路径（slow path）生成有效修正策略。其核心解决方案在于采用强化学习联合训练双路径机制，结合成对排序目标与组相对策略优化（Group-Relative Policy Optimization, GRPO），从而赋予引导智能体评估目标漂移并生成高效干预措施的能力。实验表明，该方法显著提升了单智能体与多智能体系统在信息检索类任务中的性能与效率，且所学得的引导能力具备良好的跨领域泛化性、对未见执行器的迁移能力以及无需重训练即可调节干预频率的可控性。
链接: https://arxiv.org/abs/2610.22316
作者: Leyang Shen,Yaxin Gong,Yang Zhang,Chenxiao Fan,Jiakai Wang,Dong Wang,Yang Liu,Fuli Feng,Chun Kai Ling,Tat-Seng Chua
机构: 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); Qwen Business Unit of Alibaba(阿里巴巴通义业务部)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 21 pages, 6 figures
Abstract:Large language model (LLM)-based agents and multi-agent systems increasingly tackle complex tasks through long-horizon tool use and inter-agent collaboration. However, these systems are prone to deviating from the overall goal as execution unfolds, which wastes execution resources and leads to undesired outcomes, a failure termed goal drift. In this work, we propose a steering agent that supervises execution on the fly and steers the system back on track when drift occurs. The steering agent observes the execution context, deciding when to intervene with a low-cost fast path and how to intervene with a high-cost slow path. To unlock its steering intelligence, the ability to assess goal drift and generate effective interventions, we jointly train the two paths with reinforcement learning, combining a pairwise ranking objective with group-relative policy optimization (GRPO). Experiments on information-seeking search tasks show that the steering agent provides effective supervision for both single-agent and multi-agent systems, improving their performance and efficiency. The learned steering intelligence generalizes to out-of-domain tasks, transfers to unseen executors, and supports controllable intervention frequency without retraining.
[MA-2] Learning Rational Strategies for Temporal Objectives in Multi-Agent Systems
【速读】：该论文旨在解决未知动态的随机多智能体系统中基于理性合成（rational synthesis）的问题，其核心挑战在于如何在满足各智能体特定概率时序目标的前提下，设计出符合博弈论解概念（如纳什均衡、子博弈完美均衡）的稳定联合策略。每个智能体的目标以有限轨迹上的概率线性时序逻辑（Probabilistic Linear Temporal Logic on Finite Traces, PLTL_f）形式给出，而策略优化则依赖于从各智能体满足其时序目标的概率所导出的收益。解决方案的关键在于将理性合成问题形式化为：寻找一个联合策略（若存在），使得（i）指定智能体的时序约束被满足，且（ii）该策略在选定解概念下具有战略稳定性。为此，研究提出了基于表格型纳什Q学习的条件收敛保证，并结合同时性的有限样本置信区间来验证目标满足性，通过近似最优响应检验评估策略稳定性。整个方法运行于无模型的多智能体强化学习（Multi-Agent Reinforcement Learning, MARL）框架中，利用奖励自动机自动编译时序目标为奖励信号，从而实现对时序目标极值满足概率的自主学习。实验在8个一般和博弈（general-sum concurrent stochastic games）中进行了验证，证明了方法的有效性与可扩展性。
链接: https://arxiv.org/abs/2610.22287
作者: Omar Adalat,Edwin Hamel-De le Court,Francesco Belardinelli
机构: Imperial College London(帝国理工学院); University of Manchester(曼彻斯特大学)
类目: Machine Learning (cs.LG); Logic in Computer Science (cs.LO); Multiagent Systems (cs.MA)
备注:
Abstract:We study rational synthesis for stochastic multi-agent systems with unknown dynamics, where each agent is equipped with its own probabilistic temporal objective specified using Probabilistic Linear Temporal Logic on Finite Traces ( \mathsfPLTL_f ). We impose solution concepts (e.g., Nash equilibrium, subgame-perfect equilibrium) as constraints on the joint strategy profile, using payoffs derived from each agent’s probability of satisfying its temporal objective. In particular, we formulate per-agent rational synthesis as finding a strategy profile – if one exists – such that (i) the designated agent’s probabilistic temporal constraint is satisfied, and (ii) the profile represents a stable solution under a chosen solution concept. We give conditional convergence guarantees for tabular Nash-Q learning, certify satisfaction with simultaneous finite-sample confidence bounds, and assess strategic stability with approximate best-response checking. Our approach operates in a stochastic \emphMulti-Agent Reinforcement Learning (MARL) environment, where model-free learning algorithms attempt to attain extremal satisfaction probabilities for temporal objectives automatically compiled into reward automata. Our approach is empirically validated using 8 general-sum concurrent stochastic games.
[MA-3] Auditable and Deployable Multi-Agent Motion Planning via Posetal Games from Legal Guidelines
【速读】：该论文旨在解决多智能体系统在现实世界部署中如何在遵守法律约束的前提下，协同完成各自目标的问题。传统方法依赖标量化的目标函数或完全指定的优先级层次结构，难以处理法律要求与个体目标之间部分冲突且优先级不明确的情形。为此，论文提出基于部分有序偏好博弈（posetal games）的多智能体运动规划框架，通过独立于具体场景的法律准则，自动推导出适用于特定场景的代理偏好与法律博弈模型。其核心创新在于：法律准则可随法规演进而更新，代理也可结合私有目标对法律偏好进行细化；理论证明此类自定义仅会移除均衡点，从而确保结果始终合法。为实现无需共享私有偏好的协调，引入信念博弈（belief play）机制，可从保守的局部博弈中精确恢复真实均衡，并设计分布式迭代最优响应协议以降低信息交换需求。研究将该框架应用于海上航行场景，将成对交通规则系统性扩展至多船相遇情形，在七组含最多四艘船舶的测试中，所有承诺均衡中的违规行为均被证明是法律层级强制或不可满足所致，且所有偏离者均可仅依据公开信息识别。结果表明，基于准则的posetal博弈为从部分指定的法律规范到多智能体运动规划提供了可审计的解决方案。
链接: https://arxiv.org/abs/2610.22277
作者: Florian Lercher,Karan Mahesh,Yujun Huang,Meshal Alharbi,Runyu Zhang,Hanna Krasowski,Michael R. Benjamin,Matthias Althoff,Gioele Zardini
机构: Massachusetts Institute of Technology (麻省理工学院); Technical University of Munich (慕尼黑工业大学); Woods Hole Oceanographic Institution (伍兹霍尔海洋研究所); University of California, Berkeley (加州大学伯克利分校)
类目: Multiagent Systems (cs.MA); Computer Science and Game Theory (cs.GT); Robotics (cs.RO); Systems and Control (eess.SY)
备注:
Abstract:When robots are deployed in the real-world, they must interact with other agents and accomplish their own goals while adhering to legal constraints. These potentially conflicting objectives are often only partially prioritized, making scalarized objectives or fully specified hierarchies unsuitable. We present a framework for multi-agent motion planning under legal requirements based on games with partially ordered preferences (posetal games). Scenario-independent legal guidelines order abstract legal concerns, from which we automatically derive agent preferences and a legal game for any concrete scenario. Agents may refine their legal preferences with private objectives, and regulators may refine the guideline as laws evolve. We prove that such customizations can only remove equilibria and thus preserve legality. To coordinate agents without explicitly communicating private preferences, we introduce belief play, which provably recovers the true equilibria from conservative local games, and a distributed iterative best-response protocol requiring less information. We apply the framework to maritime navigation, systematically lifting pairwise traffic rules to multi-vessel encounters. On seven encounters with up to four vessels, every rule violation of the committed equilibria is certified as forced by the legal hierarchy or unsatisfiable, and defectors are identified from public information in all tested cases. These results indicate that guideline-based posetal games offer an auditable route from partially specified laws to multi-agent motion planning.
[MA-4] Decentralized LLM Workflows via Sequential Auctions
【速读】：该论文旨在解决多步任务求解中因市场机制设计不当而导致的协作失效问题，特别是如何在缺乏完美竞争（即完全同质化代理）的前提下，维持一个高效且可持续的分布式代理市场。其核心挑战在于：一方面，若存在完全同质化的“完美克隆”代理（perfect clones），虽可实现全局最优策略收敛，但会因利润归零而瓦解参与者的风险-收益激励机制；另一方面，若完全缺乏竞争，则导致市场失灵与价值提取崩溃。为应对这一矛盾，论文提出通过引入异构代理（heterogeneous agents）在不完全竞争（imperfect competition）的中间地带实现系统稳定运行。解决方案的关键在于设计具备策略性出价能力的代理学习机制——代理需通过合理“压价”（strategic shading of bids）以保留正向个体利润边际，同时确保整体市场仍能有效协调资源完成消费者提出的协同任务。研究评估了四种出价算法（镜像下降、汤普森采样、贝叶斯最优及PPO）在三种拍卖协议（第一价格、第二价格、信用保存维克里拍卖）下的表现，结果表明，在多样化任务场景中，不完全竞争机制既能保障代理的可持续盈利，又能在低开销下实现高任务成功率，从而在效率与激励相容之间达成平衡。
链接: https://arxiv.org/abs/2610.22201
作者: Ian Gemp,Shubhendu Trivedi,Adam Fisch,Fantine Huot,Michael Kaisers,Mirella Lapata,William W. Cohen,Alekh Agarwal,Kate Larson,Jacob Eisenstein
机构: Google DeepMind(谷歌深度思维); Google Research(谷歌研究); University of Waterloo(滑铁卢大学)
类目: Computer Science and Game Theory (cs.GT); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 28 pages
Abstract:We present a scalable framework for solving multi-step tasks through a decentralized market of self-interested agents. Prior theoretical work demonstrated that agents organized through sequential auctions can converge to a globally optimal policy in a Markov decision process (MDP), but relied on the restrictive assumption of the existence of perfect clones as a model of perfect competition. In reality, exact clones drive agent profit to zero, violating the natural risk-reward incentives of participants. Conversely, we empirically demonstrate that the opposite extreme, absence of competition, also leads to market failure and value extraction collapse. Between these two extremes, we investigate how heterogeneous agents can survive in markets that span the critical middle ground of imperfect competition. To be profitable, agents learn to strategically shade bids to secure positive individual profit margins, while still ensuring that the marketplace functions effectively enough so that collaborative tasks posed by market consumers can still be solved. We formulate and evaluate four bidding algorithms (Mirror Descent, Thompson Sampling, Bayes-Optimal, and PPO) across three auction protocols (1st-Price, 2nd-Price, and Credit-Conserving Vickrey). Across a diverse set of domains, we demonstrate that imperfect competition successfully sustains necessary agent profit margins while maintaining high task success rates at low costs.
[MA-5] Robust Multi-Agent Pursuit with Preemptive Communication via Reachability-Based Control Barrier Functions
【速读】：该论文旨在解决资源受限的真实机器人系统在多智能体协同中面临的安全控制与通信约束之间的矛盾问题，即现有安全协同方法通常依赖于连续且完美的状态共享，而实际部署的机器人由于计算、通信等资源限制难以实现。其核心挑战在于：在间歇性通信条件下如何保证多智能体间的碰撞规避。解决方案的关键在于提出一种统一的控制/通信协同策略，通过引入基于控制屏障函数（Control Barrier Functions, CBF） 的安全控制机制，使每个智能体仅依据最后一次接收到的邻近智能体状态信息，实时计算其可达集（reachable set），并据此生成满足安全约束的控制输入，从而确保在通信延迟或中断时仍能避免碰撞。同时，该方案设计了新颖的不确定性风险与运动学风险度量指标，用于判断何时应主动提前发送状态信息以降低潜在碰撞风险。理论证明表明，只要通信延迟保持有界，智能体间即可保证无碰撞。仿真结果验证了该方法在追逃场景中的有效性，相较于周期性和事件触发式通信基线，可减少77.5%的消息传输量且零碰撞。
链接: https://arxiv.org/abs/2610.22198
作者: Madison Book,Justin Bradley
机构: NC State University (北卡罗来纳州立大学)
类目: ystems and Control (eess.SY); Multiagent Systems (cs.MA); Robotics (cs.RO)
备注: 6 pages, 4 figures. Accepted to the 2026 IEEE Conference on Decision and Control (CDC)
Abstract:Safe multi-agent coordination typically assumes what real-world, deployed, resource-constrained robots cannot provide: continuous, perfect state sharing. Robust control strategies must be complemented with robust communication strategies to deliver resilient deployments. In this work, we propose a unified control/communication strategy that provides safe control to guarantee inter-agent safety under intermittent communication and demonstrate its effectiveness in a pursuit-evasion scenario. Each agent computes the reachable set of neighboring agents based on the last state information received, then solves for a safe control input based on control barrier functions to ensure agents avoid collision. We prove that inter-agent collision avoidance is guaranteed provided communication staleness remains bounded. As the reachable set inflates between state transmissions, each agent reasons about when to preemptively transmit its state information to neighboring agents based on novel metrics characterizing uncertainty risk and kinematic risk. We implement and simulate a variety of pursuit-evasion scenarios and show that preemptive communication reduces messages by 77.5% compared to periodic and event-triggered baselines with zero collisions.
[MA-6] Believed but Not Said: Measuring Hidden Belief Collapse in Multi-Agent LLM Discussions
【速读】：该论文旨在解决大语言模型（LLM）系统中多智能体协作机制的可靠性问题，即在依赖智能体间相互讨论、批判与验证以纠正个体错误的设定下，当存在恶意干扰时，群体决策是否仍能保持正确性。其核心问题是：在存在少数具有误导性的对抗性智能体（adversarial agents）的情况下，群体共识是否会受到不可见但广泛传播的隐性影响，从而导致错误信息被普遍接受。解决方案的关键在于通过设计一种最小对抗实验范式，将真实答案与公开立场分离进行探测——即对每个良性智能体进行独立私密询问以获取其真实信念，同时记录其在群体讨论中的公开表态。研究发现，尽管公开立场显示仅有约20%的良性智能体支持错误答案，但私密回答揭示高达51.6%的智能体实际上已内化错误认知，表明对抗性影响主要以隐蔽方式渗透，且错误观点一旦被采纳便迅速扩散，而基础知识水平无法有效预测智能体的抗干扰能力。这一结果挑战了传统关于从众行为仅体现为公开表态趋同的认知，揭示了群体讨论中“表面一致”下的深层认知偏差，凸显了当前多智能体系统在对抗环境下缺乏透明度与鲁棒性的根本缺陷。
链接: https://arxiv.org/abs/2610.22138
作者: Liangjie Huang,Henry Peng Zou,Wei-Chieh Huang,Wooseong Yang,Yuwei Han,Philip S. Yu
机构: University of Illinois Chicago (伊利诺伊大学芝加哥分校)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multiagent Systems (cs.MA)
备注: 14 pages, 6 figures
Abstract:LLM systems increasingly rely on groups of agents that discuss, critique, and check one another, assuming that peers will catch individual errors. We test this assumption against a minimal adversary. In a ten-agent factual discussion, one or three adversarial agents repeatedly assert a plausible false answer. After each round we probe each benign agent’s private answer through a separate call whose response is hidden from its peers, and we separately label its public stance toward the claim. Across six models and three factual-QA datasets, groups remain stable without an attacker: only 0.6% of benign agents finish with the false answer. A single adversarial agent increases this rate to 44.8% and three to 60.5%. Studies of conformity typically report public agreement without private change; we find the reverse. With three adversarial agents present, benign agents hold the false answer in 51.6% of exposed turns but publicly endorse it in only 19.7%. Much of the influence therefore remains invisible in the discussion, and the false answers persist when agents are asked for their own views. Agents that retain the true answer often fail to defend it; once adopted, the false answer spreads among benign agents, and baseline knowledge does not predict robustness.
[MA-7] Agent ic-Priors: Using Large Language Models to Address the Cold Start Problem in Preference Elicitation
【速读】：该论文旨在解决智能代理在学习用户偏好时面临的“冷启动”问题：在未观察到任何用户选择行为之前，代理必须依赖先验信息，而传统上用于构建先验的、基于历史选择数据的方法在新任务或新部署场景中无法适用。其解决方案的关键在于提出Agentic-Priors，即利用大语言模型（LLM）的隐含知识来生成可解释的用户偏好先验。具体而言，LLM能够以可理解的术语（如愿意支付的金额或时间价值）描述新用户的权衡取舍，并将这些语义信息映射为多项对数几率模型（multinomial logit）系数的先验分布。随后，该先验通过贝叶斯学习机制随着实际选择数据的积累不断优化。实验表明，在七个选择数据集上，Agentic-Priors相较于无信息先验可提升冷启动阶段的预测准确率5%至40%；在三个数据集上优于当前最先进的LLM先验生成方法AutoElicit，仅在一个数据集上略逊；且相比上下文学习（in-context learning），其性能最高可提升18个百分点，且始终不显著劣于后者。与基于逐步增加样本量的调查数据拟合的分层贝叶斯模型相比，该方法在对数损失（log-loss）上等价于获得了6至40名真实调查用户的效用，凸显其在部署初期前几轮用户交互中的核心价值。
链接: https://arxiv.org/abs/2610.21938
作者: Bruno Arcanjo,Zhaoxing Li,Sebastian Stein
机构: University of Southampton(南安普顿大学)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注:
Abstract:An agent that learns a user’s preferences from their choices faces a cold start problem: before any choices are observed, it must rely solely on a prior. Informative priors are usually fitted to historical choice data, which new tasks and deployments lack. To address this, we present Agentic-Priors, which lets agentic systems use the implicit knowledge of large language models (LLMs) instead. Specifically, an LLM states a new user’s trade-offs in interpretable terms, such as willingness to pay or value of time, which we map to a prior over multinomial logit coefficients. This prior is then refined by a Bayesian learner as choices are observed. Across seven choice datasets, Agentic-Priors improves cold start accuracy over an uninformed prior by 5-40 percentage points. It also outperforms AutoElicit, a state-of-the-art LLM prior-elicitation method, on three datasets and loses on one. It is up to 18 points more accurate than in-context learning, and never significantly worse. Against hierarchical Bayes models fitted to surveys of increasing size, our approach is worth roughly 6-40 surveyed users, in terms of log-loss, so its value is concentrated in a deployment’s first users.
[MA-8] Learning Rewards in Strategic Games
【速读】：该论文旨在解决在多智能体战略环境中，如何从智能体的观测行为中准确推断其潜在奖励函数（reward function）的问题。传统逆强化学习（Inverse Reinforcement Learning, IRL）在单智能体场景下表现良好，但在多智能体战略交互中面临根本性挑战：即使某个奖励函数能够解释观测到的行为，也不一定能在决策时诱导出最优策略，导致“战略可识别性”（strategic identifiability）不足。为此，本文提出一种针对双人重复斯塔克尔伯格博弈（Stackelberg game）的主动逆强化学习（active IRL）算法，其核心创新在于让领导者（leader）通过自适应选择策略来激发追随者（follower）的有信息量响应，从而高效获取关于追随者奖励函数的可辨识信息。该方法结合最大似然估计与实验设计，实现了对所学习奖励函数引起的斯塔克尔伯格次优性的有限样本保证，并达到了与新提出的最小最大学习率 $ K^{-1/2} $ 相近的性能，仅存在多对数因子差异。实验在真实的反盗猎场景中验证了主动探索在奖励恢复及下游战略性能上的显著优势。
链接: https://arxiv.org/abs/2610.21912
作者: Antoine Bergerault,Cyrille Kone,Giorgia Ramponi
机构: 未知
类目: Computer Science and Game Theory (cs.GT); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注:
Abstract:Understanding the objectives that drive agents’ behavior is crucial in strategic environments, where multiple agents interact and adapt to one another. This challenge arises in domains such as cybersecurity, autonomous driving, and online markets, where agents’ underlying incentives are typically unknown and must be inferred from their observed behavior. Inverse reinforcement learning (IRL) provides a natural framework for this problem, but extending IRL to strategic multi-agent settings introduces a fundamental challenge: a reward that explains observed behavior does not necessarily induce optimal strategies when used for decision-making. In this work, we study this problem through the lens of \emphstrategic identifiability, and show that learning from a fixed policy pair can be insufficient to strategically identify rewards, independent of the number of expert samples K . Motivated by this, we introduce an active IRL algorithm for two-player repeated Stackelberg games, in which the leader adaptively selects policies to elicit informative responses about the follower’s reward. Combining maximum-likelihood estimation with experimental design, we derive finite-sample guarantees on the Stackelberg sub-optimality induced by the learned reward, matching up to polylogarithmic factors a new minimax learning rate of K^-1/2 . Finally, we evaluate our approach in a realistic anti-poaching setting, demonstrating the benefits of active exploration for reward recovery and downstream strategic performance.
[MA-9] MetaShare: Failure-Aware Collective Self-Improvement for Open-Ended Agents NEURIPS2026
【速读】：该论文旨在解决开放式自进化智能体在迭代优化过程中因缺乏对失败经验的系统性记录与复用，导致搜索过程陷入冗余策略、进化效率低下的问题。现有方法如GEA（Generative Evolutionary Agents）虽能跨谱系共享经验，DGM-Hyperagents则可自我修改改进机制，但均未能有效保留和利用失败尝试中的知识，从而限制了探索多样性与收敛速度。其解决方案的关键在于提出MetaShare——一个具备失败感知能力的框架，通过记录成功与失败的智能体修改及其带符号的性能变化量（signed performance deltas），构建可追溯的跨分支知识库。在每一代进化前，MetaShare通过引入任务无关的审计机制注入缺失或代表性不足的机制类别，并结合基于嵌入的去重策略（de-ranking）降低检索冗余；同时，多个元智能体以异构温度并行生成修改建议，经由确定性选择器统一评估其适用性、作用范围、可编译性及规模等约束条件后进行筛选。实验表明，MetaShare在论文评审、代码生成、数学评分、机器人奖励设计及金融风险审计等多个任务中，中位性能达到或超越DGM-Hyperagents的最终水平，且实现最优代理性能所需演化迭代次数减少1.7至2.9倍，显著提升了进化效率与探索质量。
链接: https://arxiv.org/abs/2610.21887
作者: Vishal Bhutani,Prasang Gupta,Sumanth Chundru,Kevin Paul
机构: PricewaterhouseCoopers (普华永道)
类目: Multiagent Systems (cs.MA)
备注: 30 pages, 4 figures. Accepted at NeurIPS 2026 Meta-Agents Workshop (Managing Agents that Manage Agents)
Abstract:Humans often learn fastest not by repeating success, but by remembering which failures not to repeat. Open-ended self-improving agents can share experience across lineages, as in GEA, or edit their own improvement mechanisms, as in DGM-Hyperagents, yet failed improvement attempts are rarely retained as reusable cross-branch knowledge, and search can collapse onto redundant strategies. We introduce MetaShare, a failure-aware framework that records successful and unsuccessful agent modifications together with signed performance deltas. Before each generation, MetaShare retrieves diverse high-performing strategies alongside high-confidence failures from other lineages, letting agents reuse effective mechanisms while avoiding known dead ends. A task-blind auditor periodically injects absent or underrepresented mechanism categories, while embedding-based de-ranking limits redundancy during retrieval. Multiple meta-agents propose modifications in parallel at heterogeneous temperatures behind a deterministic selector that gates applicability, scope, compilability, and size. Experiments across paper review, coding, math grading, robotics reward design, and financial audit-risk assessment show that MetaShare’s median matches or exceeds DGM-Hyperagents’ final performance and reaches its own best-performing agent in 1.7–2.9 \times fewer evolution iterations.
[MA-10] he Delegation Dilemma: Economies of Scale Accountability and Polycentric Governance
【速读】：该论文旨在解决集体行动系统中去中心化与中心化之间的张力问题，即在自愿组织形成制度、通过代表进行决策的治理模式下，如何平衡规模带来的成本分担收益与代表控制权滥用之间的矛盾。其核心问题是：当代表具有自利动机时，如何维持非剥削性的委托关系？解决方案的关键在于揭示“规模—问责”权衡机制（scale-accountability tradeoff）——群体规模越大，成员间成本分担越显著，但代表保留任一成员的边际价值随规模增长而下降，从而削弱其问责约束力。因此，个体退出的可能性仅能在内生决定的可问责群体规模上限内维持非剥削性委托。此外，研究发现，在某些参数条件下，可行的可问责治理必须采用多中心（polycentric）结构。通过大规模语言模型（LLM）代理的仿真验证表明，尽管多中心结构频繁出现且动态演化，呈现集中、剥削、退出与重组的周期性特征，但不同能力模型在领导意愿上相似，而在权力行使上存在显著差异：如Sol模型比Luna模型在相同决策状态下更频繁地发起剥削行为。
链接: https://arxiv.org/abs/2610.21746
作者: Arin Mukhopadhyay,Aris Filos-Ratsikas,Panayiotis Danassis
机构: University of Edinburgh(爱丁堡大学); Imperial College London(帝国理工学院); University of Southampton(南安普顿大学)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Many collective-action systems are neither fully decentralized nor fully centralized: agents voluntarily organize into institutions that share organizational costs and delegate decisions to representatives. Delegation can reduce individual costs, but it also transfers control to representatives whose private objectives need not align with those of their constituents. We study this tension in a repeated game with endogenous delegation to self-interested centers. A simple reduced-form model yields a scale–accountability tradeoff: larger constituencies offer greater cost-sharing benefits to members, yet the marginal value to a center of retaining any one constituent decreases as group size grows. Consequently, the prospect of individual exit can sustain non-exploitative delegation only up to an endogenous accountable constituency size. Additionally, we identify parameter regions in which fully delegated, accountable governance, when feasible, must be polycentric. We investigate these predictions through simulations with large language model (LLM) agents. Across five matched pairs of 30-agent simulations, polycentric structures repeatedly emerge but remain fluid, exhibiting cycles of concentration, exploitation, exit, and re-formation. We further find that models of different capacity exhibit similar willingness to assume leadership but differ substantially in their exercise of authority, with Sol initiating exploitation more frequently than Luna in matched decision states.
[MA-11] Agent Status Protocol (ASP): A Lightweight Specification for State Reporting in Multi-Agent Systems
【速读】：该论文旨在解决异构人工智能（AI）代理在分布式系统中状态可见性与实时监控的难题，核心问题是：如何在不依赖代理自身可信报告的前提下，实现对代理运行状态（如存活、健康、任务进展、权限状态等）的准确、及时且多维度感知。传统方案依赖代理主动上报心跳或状态信息，但当代理崩溃、卡死或被攻陷时，其自报信息不可信，导致监控失效。为此，该论文提出轻量级代理状态协议（Agent Status Protocol, ASP），其关键解决方案在于构建一个基于多方观察与服务器推断相结合的去中心化状态聚合机制。具体而言，ASP通过服务器端对心跳缺失的检测来推断代理的非活跃状态，识别虽存活但停滞的代理，并支持来自监督者、护栏系统、操作员及同级代理的观测信息作为补充或优先级更高的状态依据；同时，协议采用多维状态建模（生命周期、健康度、任务状态、权限状态、治理状态），允许代理在不同维度上并行呈现复杂状态（如在线但降级、正在执行任务且等待调度权）。此外，ASP复用已有标准语义规范（如A2A协议的任务状态、OFP协议的权限状态、OpenTelemetry GenAI语义约定），提升互操作性。该论文发布ASP 0.1版本作为提案草案，旨在推动社区讨论与实践验证，尚未形成正式标准。
链接: https://arxiv.org/abs/2610.21538
作者: Diego Gosmar,Deborah A. Dahl,Emmett Coin,Csaba Bolyós,Andreas Zettl
机构: Open Voice Interoperability Initiative; Linux Foundation AI Data; Turin, Italy; Plymouth Meeting, PA, USA; Belleville, Michigan, USA; Bratislava, Slovakia; Berlin, Germany
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 30 pages, 4 figures
Abstract:The Agent Status Protocol (ASP) defines a lightweight way for heterogeneous AI agents to report their operational state to a status server, and for observers such as dashboards to receive that state as soon as it changes. Agents push small JSON events over HTTP POST; the server exposes the aggregated state to observers over Server-Sent Events (SSE). Because an agent that has crashed, hung or been compromised cannot be trusted to report its own failure, ASP does not rely on self-reports alone: the server infers liveness from missed heartbeats, can detect agents that are alive but no longer making progress, and accepts observations from supervisors, guardrails, operators and peers, some of which take precedence over what the agent says about itself while others are shown alongside it. ASP models state along independent dimensions (lifecycle, health, task, floor, governance), so that an agent can be, for example, online, degraded, working on a task and waiting for the floor at the same time. It reuses existing vocabularies: task states from the Agent2Agent (A2A) protocol, floor states from the Open Floor Protocol (OFP), and attribute conventions from the OpenTelemetry GenAI semantic conventions. This paper presents version 0.1 of ASP as a proposal, not as an approved standard. It is a working draft intended to open a discussion, and the authors invite comments, alternative proposals and independent implementations from the community.
[MA-12] Bias Control and Injection for Multi-Agent Aggregation
【速读】：该论文旨在解决多智能体基于大语言模型（Large Language Models, LLMs）进行集体决策时，当各智能体产生分歧时应信任哪个智能体决策的核心问题。其关键挑战在于，智能体的决策易受非语义性偏差（如多数压力、响应风格偏好）的影响，从而掩盖真实推理质量。解决方案的关键在于提出一种无需验证数据或先验模型强度知识的测试阶段聚合框架：通过“偏差控制”——对每个不同答案仅呈现风格归一化的推理路径，以减少响应风格带来的偏差；以及“偏差注入”——将每个聚合智能体最初选择的答案置于少数派位置，诱发决策不稳定性，进而作为可靠性信号。该方法在三个数学推理基准和四种异构LLM组合上均实现了更高的平均准确率与更优的词元效率，并可无缝应用于多智能体辩论与混合代理（Mixture-of-Agents）框架中，显著提升最终答案准确率，且无需修改原有交互或融合机制。
链接: https://arxiv.org/abs/2610.21290
作者: Sungjae Lee,Soheil Abbasloo,Jabin Koo,Jungseul Ok
机构: POSTECH(浦项科技大学); Microsoft Research(微软研究院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multiagent Systems (cs.MA)
备注:
Abstract:Multi-agent reasoning with large language models (LLMs) raises a central question for collective decision-making: when agents disagree, whose decision should be trusted? Resolving such disagreements is difficult because agents’ decisions are susceptible to non-semantic biases such as majority pressure and response-style preferences, which can obscure the quality of the underlying reasoning. Our analysis shows how bias can be both mitigated and exploited: bias control improves initial answer selection, while bias injection elicits decision instability that serves as a reliability signal. Building on these findings, we introduce a test-time aggregation framework that requires neither validation data nor prior knowledge of relative model strength. The framework implements bias control by presenting one style-normalized reasoning trace per distinct answer, and bias injection by placing each aggregating agent’s initially selected answer in the minority. Overall, our framework achieves higher average accuracy and better token efficiency than competing non-oracle aggregation methods across three mathematical reasoning benchmarks and four heterogeneous LLM combinations. It also improves final-answer accuracy when applied to multi-agent debate and Mixture-of-Agents, without modifying their interaction or synthesis processes.
[MA-13] Diviner: Agent ic RCA with Human-Governed Adaptation for Large-Scale Infrastructure
【速读】：该论文旨在解决大规模分布式系统中根因分析（Root-Cause Analysis, RCA）的复杂性问题，其核心挑战在于单一故障症状可能跨越服务、数据、配置、计算及组织边界，导致响应者需在信息不完整的情况下整合多方专家知识并协调跨域因果路径。传统通用语言模型代理仅依赖工具调用无法有效应对此问题，常出现使用过时或误时机上下文、将查询失败误解为无症状、报告未支持的系统状态以及过度调用本为人类操作设计的可观测性服务等问题。为此，论文提出Diviner（Domain-Informed Verifiable Incident Navigation through Evidence and Review）——一个面向Meta生产环境的智能代理式RCA系统，围绕四阶段可靠性生命周期构建。其解决方案的关键在于：通过“编码”（Encode）阶段融合大量已审阅事件数据与服务负责人专业知识，生成可版本化、领域感知的技能、工作流、类型化工具与事件知识图谱；在“执行”（Execute）阶段采用自适应主控与上下文隔离的工作节点，基于时间点证据精准应用知识并维护可追溯的假设链；在“评估”（Evaluate）阶段通过对比历史调查轨迹与人工评审结果，量化根因识别准确性、方向有效性与证据依从性（Groundedness）；最后在“适应”（Adapt）阶段利用重复性漏洞与值班反馈，经回放验证、质量检查与所有者审批后，动态更新运营知识。整个系统通过结构化知识管理、证据驱动推理与闭环反馈机制，实现可信、可验证且可演进的自动化根因分析。
链接: https://arxiv.org/abs/2610.21281
作者: Pradipta Ghosh,Gaurav Mitra,Shripad Nadgowda,Akshil Shah,Junyu Liu,Canturk Isci,Syamla Bandla,Xinyao Hu,Onkar Deshpande,Anik Dasgupta
机构: Meta, Inc. (Meta)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Root-cause analysis (RCA) in large-scale distributed systems is challenging because a single symptom can cross service, data, configuration, compute, and organizational boundaries. Responders must assemble incomplete evidence and coordinate experts who hold different parts of the causal path. Generic language-model agents do not solve this problem through tool access alone: they can use stale or mistimed context, interpret failed queries as absent symptoms, report unsupported system state, and overload observability services designed for human-paced use. We present Diviner (Domain-Informed Verifiable Incident Navigation through Evidence and Review), an agentic RCA system operating at Meta and organized around a four-stage reliability lifecycle. Encode combines a large corpus of reviewed incidents with service-owner expertise to produce scoped, versioned skills, workflows, typed tools, and an incident knowledge graph. Execute uses an adaptive Lead and context-isolated workers to apply relevant knowledge to point-in-time evidence and maintain evidence-linked hypotheses. Evaluate compares saved investigation trajectories with responder-reviewed outcomes to measure exact cause, useful direction, and Groundedness. Adapt uses recurring gaps and on-call feedback to derive focused operational- knowledge updates, admitting them only after replay, quality checks, and owner approval. Subjects: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA) Cite as: arXiv:2610.21281 [cs.AI] (or arXiv:2610.21281v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2610.21281 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[MA-14] A Responsibility-Centered Coordination Mechanism for Decentralized Multi-Agent Systems
【速读】：该论文旨在解决集中式协调（centralized orchestration）在多智能体系统中面临的可扩展性、适应性与集体性能瓶颈，以及去中心化方法在任务委派与故障恢复责任界定不清的问题。其解决方案的关键在于提出一种以责任为核心的协调机制（Responsibility-Centered Coordination, RCM），通过承诺（commitment）将任务委派与相关方的明确责任绑定，实现责任的显式传递与保留。RCM采用递归委派机制，使代理在向下委派任务的同时仍保有上游责任，并通过本地化的责任约束决策与面向责任的通信协议（Responsibility-Oriented Communication Protocol, RCP），支持代理自主建立、管理与终止承诺。由于代理并行执行，局部行为持续演化承诺图，从而无需中央控制即可涌现出全局协调与动态适应，最终在推理、编码及可执行代理环境中的实验表明，RCM在任务性能上优于强基准模型最高达18.7分，在责任度量上提升10.2分，且无需特定任务训练即可增强故障与危害检测能力。
链接: https://arxiv.org/abs/2610.21243
作者: Thi-Nhung Nguyen,Viet-Thanh Pham,Linhao Luo,Thuy-Trang Vu,Dinh Phung
机构: Monash University (蒙纳士大学)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注: 14 pages, including appendices
Abstract:Centralized orchestration provides comprehensive multi-agent coordination but creates bottlenecks in scalability, adaptability, and collective performance. Decentralized approaches alleviate these bottlenecks by distributing coordination across agents, yet often leave responsibility for delegated dependencies and failure recovery unclear. To address this gap, we propose RCM, a responsibility-centered coordination mechanism that leverages commitments to couple task delegation with explicit responsibilities of the involved parties. Through recursive delegation, agents retain their upstream responsibilities while locally orchestrating downstream commitments, thereby preserving accountability across delegation chains, creating a network of local orchestration. We operationalize this mechanism locally through responsibility-constrained decision-making and a responsibility-oriented communication protocol (RCP), enabling agents to autonomously establish, manage, and terminate commitments. As agents act in parallel, their local actions continuously evolve the commitment graph, allowing global coordination to emerge and adapt toward a shared mission without centralized control. Experiments across reasoning, coding, and executable agent environments demonstrate that RCM outperforms strong centralized and decentralized coordination baselines by up to 18.7 points in task performance and 10.2 points in Responsibility, while improving failure and harm detection without task-specific training.
[MA-15] SSR: Structural Signal Retriever for Scalable Agent Discovery in LLM -Orchestrated Multi-Agent Systems
【速读】：该论文旨在解决大规模多智能体生态系统中，大型语言模型（LLM）协调器在发现合适智能体时面临的候选上下文可扩展性问题。随着智能体注册表规模的扩大，直接将完整注册表暴露给LLM会导致提示词长度、推理成本、延迟显著增加，并可能超出模型上下文窗口限制。其核心挑战在于如何在不暴露全部注册信息的前提下，高效检索出相关智能体卡片（Agent Card），同时保持紧凑的候选上下文。解决方案的关键是提出一种确定性且无需训练的检索方法——结构信号检索器（Structural Signal Retriever, SSR）。SSR从承载能力的智能体卡片字段中提取结构化词汇信号，利用逆文档频率（IDF）对信号进行加权，并通过倒排信号索引（Inverted Signal Index, ISI）实现候选者的选择性评分。实验表明，在1,000个智能体场景下，SSR相较全目录调度可减少98.8%的LLM输入量并降低70.9%的调度延迟；在MS MARCO基准上，IDF提升了结构化表示下的检索效果，ISI在保持检索性能的同时将显式评分的段落数减少96.8%。端到端测试进一步验证了候选预筛选机制在注册表规模增长时仍能维持紧凑候选上下文，避免了全目录调度中的操作失败。因此，SSR作为一种原生面向智能体卡片、可追溯的检索机制，有效支撑了候选上下文的可扩展性。
链接: https://arxiv.org/abs/2610.21113
作者: Giancarlo Maldonado Cárdenas,Carlos A. Astudillo,Nelson L. S. da Fonseca
机构: Institute of Computing – State University of Campinas (计算学院 – 坎皮纳斯州立大学), Brazil (巴西)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注:
Abstract:Large-scale multi-agent ecosystems require mechanisms for discovering suitable agents without exposing entire registries to a large language model (LLM) orchestrator. The challenge is not only to retrieve relevant Agent Cards from a growing registry, but also to maintain a compact candidate context, since each additional candidate increases prompt size, inference cost, latency, and the risk of exceeding the model context window. We formulate this problem as candidate-context scalability and propose the Structural Signal Retriever (SSR), our deterministic and training-free retrieval method designed for Agent Card discovery. SSR derives structural lexical signals from capability-bearing Agent Card fields, weights them using inverse document frequency (IDF), and uses an Inverted Signal Index (ISI) for selective candidate scoring. We evaluate SSR using a synthetic agent benchmark, a public MS MARCO adaptation, and end-to-end LLM orchestration. At 1,000 agents, SSR reduces LLM input by 98.8% and orchestration latency by 70.9% relative to full-catalog orchestration. On MS MARCO, IDF improves retrieval effectiveness within the structural representation, while the ISI preserves retrieval effectiveness and reduces explicitly scored passages by 96.8%. End-to-end experiments show that candidate pre-selection maintains a compact candidate context as registry size grows and avoids the operational failures observed with full-catalog orchestration. These results support SSR as an Agent-Card-native, traceable retrieval mechanism for candidate-context scalability.
[MA-16] A Jury Not an Electorate: Label-Free Reliability Estimation for Ensembles of Language Models
【速读】：该论文旨在解决多模型集成（ensemble）中专家模型可信度评估的难题，即如何在缺乏标注数据的情况下，从多个语言模型的输出模式中自动推断各模型的可靠性。传统方法通常为每个专家分配固定的声誉权重，这些权重依赖于有标签数据进行学习，存在对监督信号的强依赖性。本文提出一种基于“陪审团”（jury）的建模框架，将每个专家视为对隐含真实答案作证的证人，其可靠性未知；通过期望最大化（Expectation Maximisation, EM）算法，仅依据未标注的模型输出配置中的共识与分歧模式，即可估计各专家的可靠性。关键创新在于：决策规则仍为对专家答案分布的加权线性组合，但权重由无监督的EM过程动态推断，而非依赖标签训练。在包含1560个问题的基准测试中，该方法使用九个参数量达30亿的专家模型，性能与使用标签训练的融合规则相当，且未表现出显著优势，但显著优于单一模型在自一致性匹配下的表现及现有最强的无标签替代方案，证明了无监督信誉推断的有效性。
链接: https://arxiv.org/abs/2610.21085
作者: Alessandro Rozza,Davide Mambelli,Steven Motta,Juri Taresch
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 35 pages, 4 figures, 13 tables, including appendices
Abstract:An ensemble of language models is a jury, not an electorate. The answers are testimony from witnesses of unknown reliability, and the question is not what the majority said but who is credible. Most ensemble methods assign each expert a fixed reputation, often learned from labels. We ask whether it can instead be inferred from the pattern of agreement and disagreement among the experts. We model the experts as a jury: each question has a latent true answer, each expert reports it with unknown reliability, and those reliabilities are estimated by Expectation Maximisation (EM) from unlabelled answer configurations alone. We show that its decision rule reduces to a log-odds weighted linear score over the experts’ answer distributions. What changes is not how the jury votes, but how it learns whom to trust. On a 1560-question benchmark, the jury, with nine experts of up to 3B parameters, performs comparably to the same fusion rule with label-fitted weights, showing no statistically significant advantage from supervision. Instead, it outperforms the best single model with matched self-consistency and the strongest label-free alternative on the full benchmark.
[MA-17] Athanor: An Agent -Native System for Autonomous Post-Training Research
【速读】：该论文旨在解决如何在无外部指导的情况下，使前沿人工智能（AI）代理自主完成对原始基础模型的后训练优化（post-training），从而实现跨领域通用能力的显著提升。其核心挑战在于验证当前大模型是否具备真正的研究自主性，即能否不依赖人类干预或教师模型，仅通过自身探索与迭代，持续改进模型性能。解决方案的关键在于构建一个名为Athanor的开放、原生支持自治后训练研究（AutoPostTrain）的系统框架：该系统为代理提供完全的研究自主权、低摩擦工程工具与高效启动代码，并在隔离沙盒环境中运行；代理可自主读取数据与评估结果、构建训练数据、编写和调整训练算法、调优超参数并基于分析决策下一步行动。实验中，四名Claude Opus 5.5代理在104小时内仅使用公开数据与自身协作，将Qwen3-4B-Base模型在六任务开发集上的表现从15.9提升至67.3，在保留测试集上从9.2提升至48.2，达到甚至超过官方Qwen3-4B-Instruct-2507版本（66.5/47.9）。这一成果表明，当前先进代理已具备通过重组已有方法（如公共推理数据的监督微调、单个代理引入的强化学习策略、跨代理权重融合及多轮自蒸馏）实现自我增强的能力，证明了生成式AI（Generative AI）在无需外部干预下的工程与研究潜力。
链接: https://arxiv.org/abs/2610.21055
作者: Yufa Zhou,Yichen Wang,Jiakun Fan,Ziqi Shu,Zhichao Liu,Anru R. Zhang,Neiwen Ling,Zijun Wei
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注:
Abstract:Recursive self-improvement, in which AI systems build stronger AI systems, is one promising route to superintelligence. We study a weaker form that can be measured today: frontier agents post-training a small raw base model with public data, as a test of their research potential. We build Athanor, an open agent-native system for autonomous post-training research (AutoPostTrain) on real GPUs. We give the agents full research autonomy, low-friction engineering tools, and efficient starter code. The agents carry out the research themselves: they read datasets and evaluation results, build their training data, write and modify training algorithms, tune hyperparameters, and analyze the results to decide what to try next. They work in a sandbox and cannot use external teacher models. Athanor’s services supply frozen evaluation, managed jobs, decontaminated data, a persistent research record, and a held-out suite the agents are unaware of. In a 104-hour campaign, a team of four Claude Opus 5.5 agents, each on an eight-GPU node and given no research advice, improves Qwen3-4B-Base from 15.9 to 67.3 on a six-task development suite and from 9.2 to 48.2 on a held-out suite, against 66.5 and 47.9 for the official Qwen3-4B-Instruct-2507. To our knowledge, this is the first demonstration that a team of frontier agents can autonomously post-train a raw base model into a cross-domain generalist model that matches or surpasses the official instruct version on average. The campaign’s full record shows that the agents reach this level by recombining known methods and each other’s models: SFT on public reasoning data, RL with a rollout change one agent introduced, weight merges across agents, and rounds of self-distillation. With execution handled, today’s agents already have the engineering and research ability to conquer this task. We hope Athanor provides a foundation for scaling autonomous AI research.
[MA-18] Recursive Reward Improvement via Adversarial Meta-Prompt Evolution
【速读】：该论文旨在解决生成式人工智能在开放式专家任务（如临床报告生成）中因缺乏明确答案标准而导致的自洽性与质量下降问题。传统基于自我改进的递归框架在数学和编程等可自动验证的任务中表现良好，但在无固定答案键的开放任务中，模型生成的评分标准（rubric）易被解题模型钻空子——即仅表面提及预期内容而缺乏实质内涵，从而破坏学习过程。其解决方案的关键是提出一种名为递归奖励改进（Recursive Reward Improvement, RRI）的新机制：通过引入一个“奖励劫持对抗者”（reward hacking adversary），在解题模型训练前，先生成一个利用鲁棒性漏洞但内容无意义的“恶意答案”，同时生成一个不参考评分标准的诚实答案，并通过最多两步修复迭代优化原始评分标准，使诚实答案胜出。这一过程不仅提升了当前任务的可靠性，还通过反馈机制动态改善未来任务与评分标准的生成质量。实验表明，在HealthBench Professional基准上，RRI相较仅生成（Generation Only）基线分别在9B和27B参数规模下提升10.9和12.4分，且无需依赖外部模型即可实现10.3分增益；尤其在高难度与写作类任务中提升显著。此外，该方法在未见过的PRBench Hard和ProfBench测试集上仍分别获得6.6和8.0分的性能提升，验证了其泛化能力。研究进一步分析了模型生成的问题-评分对中的可被利用弱点，为后续开放域自洽学习提供了关键洞见。
链接: https://arxiv.org/abs/2610.21040
作者: Wei Dai,Qianchu Liu,Guanghui Qin,Xing Han,Tristan Naumann,Paul Pu Liang,Sheng Zhang
机构: Massachusetts Institute of Technology(麻省理工学院); Microsoft Research(微软研究院); Johns Hopkins University(约翰霍普金斯大学)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Recursive Self-Improvement (RSI) has gained wide attention for its ability to work with limited data. In a proposer-solver loop, a proposer model writes training tasks and a solver model learns to solve them through reinforcement learning. Such loops have succeeded in mathematics and coding, where every answer can be checked automatically. Open-ended expert tasks, like clinical report generations, have no answer key, which makes simple question-answer generation not viable. Here the proposer must also write a rubric that lists the content an answer should include and the errors it should avoid. However, the unique challenge behind model-written rubric is the comprehensiveness of it. For naively generated rubric, solvers tend to exploit it, naming the expected topics without the substance behind them, which corrupts the learning process. To solve this problem, we introduce RRI, Recursive Reward Improvement, which uses a reward hacking adversary to supervise the reward. Before the solver trains on a generated task, an adversary writes an answer that games its rubric while being as useless as possible, the same model writes an honest answer without seeing the rubric, and up to two repair steps revise the rubric to make the honest answer win. The exploits found also rewrite how future tasks and rubrics are generated. On HealthBench Professional, RRI improves the adjusted score over the Generation Only self-improvement baseline by 10.9 points at 9B and 12.4 at 27B, and by 10.3 points without any external model. In particular, the gains are largest on difficult and writing tasks. We show that the 27B model trained on the HealthBench Professional description improves by 6.6 points on PRBench Hard and 8.0 on ProfBench without seeing a task from either. We further analyze the exploitable points of model-generated question-rubric pairs to facilitate future research on open-ended self-play.
[MA-19] SCARCE: Emergent Institutions in Long-Horizon Multi-Agent Coding Systems
【速读】：该论文旨在解决多智能体系统在长期、高复杂度任务中因共享计算资源竞争而产生的协调与安全问题。随着自主智能体在现实场景中的广泛应用，其在无预设规则下自发形成的交互行为可能引发不可控的协同或冲突，但现有研究多依赖简化博弈模型，难以揭示真实环境中智能体协商行为的演化机制。为此，本文提出SCARCE（Shared Compute ARena for Coding agEnts）框架，使智能体在具有个体截止时间约束的连续编码任务中，通过协商获取共享执行资源（如代码编译与测试权限），从而模拟真实场景下的资源竞争。研究发现，智能体自发形成了以声誉、互惠和先例为基础的非正式制度性行为，即“制度化行为”——这些行为表现为跨任务持续存在的规则与承诺，甚至导致智能体主动放弃即时利益以遵守过往约定。作者构建了包含21类策略与涌现规范的分类体系，对8种模型在超过1500条轨迹中的表现进行系统分析，结果表明前沿大模型较旧版或轻量模型表现出更强的制度化倾向。此外，除合作外，还观察到排他性联盟及罕见的分化瓦解策略。SCARCE为多智能体系统中协作关系的形成、资源分配模式及其对安全性与协调性的影响提供了可控制、可审计的实验环境，推动了对智能体间复杂社会性行为的深入理解。
链接: https://arxiv.org/abs/2610.21029
作者: Abhinav Kumar,Rediet Abebe,Eugene Bagdasarian,Sahar Abdelnabi
机构: University of Massachusetts Amherst; ELLIS Institute Tübingen; MPI for Intelligent Systems; Tübingen AI Center
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注:
Abstract:As autonomous agents become widely deployed, coordinating access to shared computing resources and services will become increasingly important. Recent incidents have already revealed unplanned coordination between agents, highlighting the safety implications of these interactions. Yet prior studies of agent negotiation often use stylized games and simulated bargaining, leaving open how these behaviors develop during sustained, long-horizon tasks. We propose SCARCE (Shared Compute ARena for Coding agEnts), a framework in which agents solve sequences of coding tasks under individual deadlines while negotiating access to a shared execution resource. This negotiation determines when and for how long they can inspect programs, compile code, and run tests. We find that agents spontaneously develop informal institutions: persistent rules and commitments grounded in reputation, reciprocity, and precedent. These arrangements, which we call institutional behaviors, carry across tasks and influence resource allocation, with agents sometimes foregoing immediate advantages to honor earlier agreements. We develop a 21-category taxonomy of negotiation strategies and emergent norms and use it to analyze eight models and audit more than 1,500 trajectories across different environmental settings. Among the models tested, newer frontier models exhibit more institutional behaviors than older or lightweight models. Alongside cooperation, we observe exclusionary coalitions and rare divide-and-conquer strategies. SCARCE provides a controlled setting for proactively and systematically auditing how these relationships emerge, shape access to resources, and affect the safety and coordination of multi-agent systems.
[MA-20] A Multi-Tiered AI Accessibility Model (MAIAM): Proportionate Access and Capability Limits for Comprehensive Frontier AI Safety
【速读】：该论文旨在解决当前关于释放高能力人工智能系统时存在的二元化决策框架问题，即仅在“通过API封闭”与“公开模型权重”之间做选择，这种框架混淆了两个独立的决策维度：系统的潜在能力水平，以及谁能在何种条件下获得该能力。论文聚焦于后者——即能力获取的准入机制。其核心解决方案是提出一种多层级人工智能可访问性模型（Multi-Tiered AI Accessibility Model, MAIAM），该模型不再依据发布形式（如权重文件或API）划分访问权限，而是基于目的、安全性和能力三个维度进行分层管理。具体而言，MAIAM划分为三级：Tier 1（AI for Learning）面向教育用途，广泛开放但能力受限；Tier 2（AI for Exploration）用于研究开发，需注册、身份验证及使用控制，且随着能力提升而收紧；Tier 3（Full-Scale Frontier AI）仅限于经过结构化、可追溯和可问责的渠道获取。其核心原则是可访问性应与能力、安全性及风险相匹配。论文进一步论证，公开前沿级模型权重将导致开发者在发布后丧失所有安全控制手段，并指出仅以数百美元即可剥离安全微调，因此基线模型评估难以反映衍生系统的实际安全性。为此，作者提出结合“SAFE AI Operating Point”构建双边界框架：能力边界（Capability Boundary）设定不可逾越的安全阈值，而可访问性边界（Accessibility Boundary）则界定低于该阈值能力的获取资格。研究表明，仅设定能力上限不足以确保安全，必须辅以由监管机构管理的分层访问制度。
链接: https://arxiv.org/abs/2610.20921
作者: Chai K. Toh,Vasu Raj Jain
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Most debate about releasing capable AI systems treats the choice as binary: keep the model behind an API, or publish the weights. That framing runs together two separate decisions. One is how capable a system may become. The other is who may obtain a given capability, and on what terms. This paper is about the second decision. We propose a Multi-Tiered AI Accessibility Model (MAIAM) that sorts access by purpose, safety, and capability instead of by release artifact. Tier 1, AI for Learning, is broadly accessible and deliberately capability-bounded. Tier 2, AI for Exploration, serves research and development under registration, identity, and usage controls that tighten as capability rises. Tier 3, Full-Scale Frontier AI, is available only through structured, recorded, and accountable access. The principle behind the tiers is that accessibility should be proportional to capability, safety, and risk. We show why publishing frontier-class weights forfeits every safety lever a developer has after release, and, using evidence that safety fine-tuning can be stripped from open weights for a few hundred dollars, why an evaluation of the base model says little about its derivatives. We give criteria for assigning a system to a tier along capability, autonomy, domain-consequence, and usage/safety axes, and combine MAIAM with the SAFE AI Operating Point into a two-part framework: a Capability Boundary that fixes the safety limits not to be crossed, and an Accessibility Boundary that fixes who may obtain a capability below that limit. Capability limits alone do not make AI safe; a tiered access regime administered by regulating bodies is also needed.
[MA-21] Feasibility Ceilings Before Learning: Sizing and Timing Bounds for Drone-Mesh Interception
【速读】：该论文旨在解决多智能体系统中，多个拦截器（interceptors）如何在预测目标到达点时实现最优分布以最大化覆盖概率的问题。核心挑战在于：在给定的探测不确定性与空间约束下，如何通过理论分析确定性能上限，并为后续策略优化提供基准。其解决方案的关键在于提出一系列闭式表达式（closed-form expressions），用于精确量化未选择具体策略前的理论性能边界，包括基于最小二乘法方差传播与Delta方法修正的预测点不确定性模型、密集网格覆盖概率的精确解析解及其体积类比形式。研究发现，由于智能体间至少保持2r距离的分离约束，平面与三维空间中的最大覆盖率分别被限制在90.69%和74.05%，这一上限由几何约束决定而非问题本质；允许重叠可突破此限，引入Kershner最稀疏覆盖密度并显著提升性能，尤其在覆盖率阈值p > 0.80时进入更优区域。此外，六边形排列优于其他五种布局，而重心-维诺（centroidal-Voronoi）配置在资源有限时仅能达到最优覆盖的一半。研究还揭示了高斯假设在存在重尾或各向异性时具有保守性，但在存在偏差（如一倍标准差的目标偏移）时极为脆弱，导致覆盖损失高达14个百分点。最后，通过对比分析表明，尽管集中式PPO算法在不同训练种子下平均优于基线策略约15%，但其仍远未达到理论上限，而解析式部署方案可达到理论容量的85%。因此，论文主张在任何学习实验之前应先计算并报告此类理论天花板，以确保结果评估的科学性与可比性。
链接: https://arxiv.org/abs/2610.20823
作者: Dmytro Babarytskyi
机构: Kharkiv National University of Radio Electronics (哈尔科夫国立无线电电子大学)
类目: Robotics (cs.RO); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: 13 pages, 6 figures, 8 tables. Code, data and policy checkpoints: this https URL
Abstract:A mesh of interceptors defending a fixed site must spread over the predicted arrival point of an approaching object. This is a recurring task in multi-agent robotics: N agents, each covering a disc of radius r, spread over a probability density so that a point drawn from it falls within some agent’s disc. We give closed forms for the quantities that bound performance before any policy is chosen, and validate each by Monte Carlo. The uncertainty of a point predicted by extrapolating a noisy linear track follows from least-squares variance propagation plus a delta-method term for the uncertain arrival time, agreeing with simulation to within 1.8%. The coverage probability of a packed mesh admits a closed form accurate to 0.007 absolute, with a volumetric analogue. A naive area ratio misses a ceiling: while agents stay at least 2r apart, no mesh exceeds 90.69% coverage in the plane or 74.05% in a volume. This ceiling belongs to the separation constraint, not to the problem - permitting overlap substitutes Kershner’s thinnest-covering density and removes it, the better regime above a crossover at p=0.80. Hexagonal packing is optimal among five arrangements; the centroidal-Voronoi configuration coverage control supplies reaches only half its coverage at small budgets. The Gaussian assumption is conservative under heavy tails and anisotropy but fragile under bias: a one-sigma aim-point offset costs 14 points of coverage. A repositioning team must also choose a commit time, and the admissible window can be empty. Against the applicable ceiling, analytic placement attains 85% of what the configuration permits while centralised PPO averages 15% over three trainer seeds, with a paired bootstrap against a null policy containing zero at every seed. We argue such ceilings should be computed before any learning experiment and quoted alongside every result.
自然语言处理
[NLP-0] OmniVerifier-C1: Decomposed Graded Verifier for Rule-Governed Generative World Modeling NEURIPS2026
【速读】： 该论文旨在解决复杂多模态场景下视觉验证器（visual verifier）在处理组合性要求时的鲁棒性不足问题，尤其是传统二值化验证器（binary verifier）仅输出“真/假”离散判断，难以提供细粒度、可解释的评估信号。其核心解决方案是提出一种点级评分式生成式视觉验证器（pointwise graded verifier），能够自动将复杂提示或环境规则分解为细粒度组件，并对每个组件赋予连续得分，从而生成结构化且可解释的批判性反馈信号。该方法的关键在于引入一种解耦强化学习训练策略（Decoupled RL training strategy），通过联合优化子问题层面的二值判断与全局一致性评分，利用精确的局部反馈来构建细粒度验证能力，进而支撑全局连续评分的可靠性。实验表明，该框架在文本到图像生成及基于规则的生成式世界建模（rule-governed generative world modeling）任务中均显著优于传统二值验证器；此外，理论分析与实证结果进一步验证了在有真实图像作为参考时，成对比较式验证范式（pairwise verification paradigm）能提供更准确的批判信号。整体研究揭示了一个关键洞见：从生成模型向结构化世界建模智能体演进，其核心依赖于高精度的成对评分式视觉验证器的可用性。
链接: https://arxiv.org/abs/2610.22355
作者: Bowei Liu,Xinchen Zhang,Xuhuan Li,Tianlin Zhang,Kaian Jiang,Jingwei Liu,Congyang Zhao,Yifei Qian,Junhong Liu,Zhiheng Li,Yujiu Yang,Yang Cai,Ling Yang
机构: 未知
类目: Computation and Language (cs.CL)
备注: NeurIPS 2026. Project: this https URL
Abstract:Generative visual verifiers play a crucial role in advancing multimodal intelligence. To enhance the robustness of visual verifiers in complex multimodal scenarios involving compositional requirements, we move beyond conventional binary verifiers that produce only True/False discrete judgments and introduce a verifier that automatically decomposes complex prompts or environmental rules into fine-grained components and assigns pointwise scores to provide a continuous and more informative critique signal. Our experiments further demonstrate that these structured and interpretable outputs improve the verifier’s generalization to out-of-domain multimodal critic tasks. To train this verifier, we introduce a Decoupled RL training strategy that jointly optimizes sub-question-level binary judgment and global alignment scoring, using precise local feedback to develop fine-grained verification capabilities that provide a foundation for reliable continuous scoring at the global level. Under a general reinforcement learning framework, we first empirically demonstrate that the proposed pointwise graded verifier significantly outperforms binary verifiers in traditional text-to-image scenarios. We then extend our investigation to rule-governed generative world modeling where generative models are leveraged for world simulation and world reasoning under explicit rules. From a theoretical standpoint, we further show that a pairwise verification paradigm yields more accurate critiques than the pointwise formulation when ground-truth images are available, with empirical results on diverse structured world-modeling tasks such as Maze and Sudoku further validating the effectiveness of our approach. Overall, our findings suggest a key insight: the transition from generative models to structured world-modeling agents critically hinges on the availability of accurate pairwise graded visual verifiers.
[NLP-1] FDE-Gym: The Last Mile of AI Delivery
【速读】： 该论文旨在解决现有智能体评估基准在真实客户交付场景中适用性不足的问题。传统基准多基于明确界定的任务（如工单、需求文档），而现实中的客户交付任务具有高度不确定性：前线部署工程师（FDEs）需在未明确定义的需求下开展工作，于生产环境中操作并识别隐含约束。为填补这一差距，研究提出FDE-Gym，一个包含95个端到端交付任务的基准，源自18个行业共20个真实客户场景，由领域专家从实际项目中提炼改编。该基准通过模拟客户环境，引入脚本化利益相关者与可配置的客户系统，使智能体在与利益相关者交互及操作类生产系统的过程中完成交付。其评分机制以客户自身成功指标的改进程度为核心，归一化至0（无变化）至1（参考解决方案），并施加对沟通次数、试运行和等待时间的惩罚。评估结果显示，在需主动发现隐含约束的真实情境下，表现最优的模型平均得分仅为0.56，仅能成功解决14项任务（得分≥0.8且无事故）。关键瓶颈在于隐含约束的发现能力——当缺乏显式提示时，性能显著下降；但即使在已获取所有关键信息的任务中（共23项），该模型也仅在其中10项达到0.8以上的业务提升，表明单纯的信息获取不足以实现有效交付。因此，该研究揭示了当前模型在处理真实世界复杂性方面存在显著能力鸿沟，核心挑战在于如何在不确定环境中进行有效的认知推理与动态决策，而非仅依赖任务求解能力。
链接: https://arxiv.org/abs/2610.22350
作者: Dongxu Yang,Hongda Yan,Renxin Wang,Baoli Zhang,Haojia Zuo,Ying Zhao,Youmin Chen,Xiaojian Liao,Congming Gao,Tianyang Jiang,Yixiao Chen,Qing Wang,Zhiguo Yu,Xiaqing Li,Yuhao Zhang,Jiawen Sun,Renzhe Xu,De Mi,Tiancheng Wu,Xianya Fu,Xuanlu Zhao,Bo Feng,Meixuan Ren,Guanming Lin,Ziheng Zhou,Xinhua Wang
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Software Engineering (cs.SE)
备注: 14 pages, 8 figures, 4 tables. Website: this https URL ; Code: this https URL ; Data: this https URL
Abstract:Existing agent benchmarks typically evaluate agents on well-specified tasks defined by issue reports, specifications, or task tickets. Real-world customer delivery is different: forward-deployed engineers (FDEs) must clarify ambiguous requirements, work in live systems, and uncover unstated constraints. We introduce FDE-Gym, a benchmark of 95 end-to-end delivery tasks from 20 customer scenarios across 18 sectors, each adapted by industry experts from real-world delivery projects. In simulated customer environments with scripted stakeholders and modifiable customer systems, agents interact with stakeholders and operate on production-like systems to deliver working solutions. FDE-Gym scores a delivery by how much it improves the customer’s own success metric, normalized between 0 (unchanged system) and 1 (reference solution), with penalties for stakeholder interactions (messages, trial runs, and waiting time). An online platform lets humans practice FDE skills under the same grading criteria. We evaluate 13 model configurations from seven providers. In the realistic setting, where unstated constraints must be discovered, the best configuration achieves a mean FDE-Gym score of 0.56 and resolves only 14 of the 95 tasks (a score of at least 0.8 with no incident). Discovering unstated constraints is the primary bottleneck: performance drops sharply when models must uncover them without explicit guidance. Yet information discovery alone does not account for the gap: even on the 23 tasks where the best configuration obtains every key fact from stakeholders, it reaches a normalized business uplift of 0.8 or above on only 10. These results reveal a substantial gap between solving well-specified tasks and delivering effective AI systems for real customers.
[NLP-2] MLExpert: Towards a Generalist Agent ic Model for Autonomous Machine Learning Engineering
【速读】： 该论文旨在解决大规模语言模型（Large Language Models, LLMs）在长期自主机器学习工程（Machine Learning Engineering, MLE）任务中因实验周期长、反馈延迟且决策依赖历史轨迹而导致的优化效率低下问题。其核心挑战在于如何从复杂的多轮交互轨迹中提取有效的监督信号，并在不等待完整执行流程结束的情况下实现策略的渐进式修正。解决方案的关键在于提出一种后训练（post-training）方法，结合层级化代理监督微调（Hierarchical agentic SFT）与基于窗口的强化学习（Windowed Interaction RL, WIRL），通过将连续实验结果间的一系列决策视为“执行窗口”，并赋予统一的结果导向奖励，从而实现对中间结果的有效利用。同时，采用异步数据收集与策略校正机制，在轨迹未完成时即进行策略更新，显著提升了长期优化过程中的学习效率。该方法被应用于训练名为MLExpert的350亿参数混合专家（Mixture-of-Experts）代理模型，在MLE-Bench Lite上将基础模型的平均奖牌率从37.9%提升至72.7%，性能接近更大型前沿模型，且在NatureBench Lite、代理骨架优化和算法优化等任务上均展现出良好的泛化能力，验证了其在复杂自优化任务中的有效性。
链接: https://arxiv.org/abs/2610.22344
作者: Shangheng Du,Jinxin Shi,Shuaiyu Zhang,Yiming Wu,Junpeng Zhao,Shiyang Feng,Xin Li,Liang He,Xiangchao Yan,Bo Zhang
机构: Shanghai Artificial Intelligence Laboratory(上海人工智能实验室); East China Normal University(华东师范大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Code: this https URL Model: this https URL
Abstract:Large language models (LLMs) are expanding from assisting scientific research and engineering to autonomously developing and improving AI systems, advancing AI for AI (AI4AI) as a pathway toward recursive self-improvement (RSI). Workflow-driven machine learning engineering (MLE) systems provide a practical path toward this pursuit through complex orchestration that manages long-horizon experimentation, while recent MLE systems adopt model-directed interaction, allowing models to autonomously organize experiments and adapt optimization strategies based on execution feedback. However, feedback from each experiment reflects several preceding decisions, and a long MLE run contains many such costly experiments. Training must therefore select useful supervision from long interaction trajectories and learn from intermediate outcomes. We propose a post-training method for autonomous long-horizon MLE optimization and apply it to train MLExpert, a 35B Mixture-of-Experts agentic model. We design the Kaggle-MLE Factory to construct executable tasks and an interaction harness for model-directed experimentation. Hierarchical agentic SFT transfers general tool-use and coding capabilities to MLE through outcome-guided supervision. Windowed Interaction RL (WIRL) treats the multi-turn decisions between successive experiment results as an execution window and assigns it a shared outcome-based reward. Asynchronous collection and policy correction support policy updates before the trajectory ends. On MLE-Bench Lite, MLExpert improves the average medal rate of the base model from 37.9% to 72.7%, achieving performance comparable to larger frontier models such as Kimi-K3 and GLM-5.3. Further experiments show that MLExpert improves performance on NatureBench Lite, agent scaffold optimization, and algorithm optimization, suggesting that the learned optimization capability generalizes to other tasks.
[NLP-3] Evaluation Awareness Beyond English: Do LLM s Act Differently When Tested in Indic Languages?
【速读】： 该论文旨在解决多语言环境下大型语言模型（Large Language Models, LLMs）在安全评估中表现出的“评估意识”（evaluation awareness）问题，尤其关注在印地语等印地语系（Indic）语言中的表现。现有研究几乎仅聚焦于英语场景，而印地语系语言服务于超过十亿用户，其安全评估的可靠性存在显著盲区。研究通过在五种印地语系语言及英语中测试四个指令微调模型，发现当模型被告知处于监控或安全测试状态时，其拒绝有害请求的频率显著上升——在印地语系语言中平均提升14.4个百分点，且拒绝率已接近英语水平；然而，这一机制也导致对无害请求的误拒率从6.9%升至11.3%，表明评估环境可能扭曲模型在真实使用中的行为表现。解决方案的关键在于揭示并量化评估意识在非英语语境下的普遍性与影响，强调当前基于英语的安全评估框架在多语言场景下可能无法准确反映模型的真实行为，从而呼吁建立更具语言包容性的安全评估方法。研究还公开了全部47,880条标注响应数据，以支持后续跨语言安全性研究。
链接: https://arxiv.org/abs/2610.22333
作者: Trizal Garg,Sibayan Mitra,Arya Karanjkar,Advait Singh,Jagat Sesh Challa,Yash Sinha,Siddharth Mehrotra,Dhruv Kumar,Pratham Yogish Maliya
机构: Birla Institute of Technology and Science, Pilani(比尔拉技术与科学学院，皮拉尼)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Large language models (LLMs) can often tell when they are being tested. This ability, known as evaluation awareness, can lead models to act more cautiously during safety evaluations than in real use. Existing work has studied evaluation awareness almost entirely in English, although models serve users in many languages. Indic languages are spoken by over a billion people, yet evaluation awareness remains almost entirely unstudied in them. We test four instruction-tuned models from three families in five Indic languages alongside English. A set of cues varies whether the model is told that it is monitored and whether the monitoring is a safety test. Models refuse harmful requests more often when they are told that they are being tested. This increase appears in all four models and all six languages. In Indic languages, refusal of harmful requests rises by 14.4 percentage points on average. The test cue also brings Indic refusal of harmful requests level with English, but raises refusal of benign Indic requests from 6.9% to 11.3%. We release all 47,880 judged responses to support further work. These results suggest that safety evaluations in Indic languages may not reflect how models treat their everyday users.
[NLP-4] wice Upon a Time: Complementary Representations for Time-Series Reasoning
【速读】： 该论文旨在解决时间序列语言模型（Time Series Language Models, TSLMs）在表示学习中如何平衡信息保留与变异因素剔除的核心问题，即：时间序列表征应保留哪些关键信息，又应剥离哪些无关变化。现有TSLMs通常仅使用单一表征形式，导致其在下游任务中具有固定的归纳偏置，限制了模型对复杂时序结构的理解能力。为此，本文提出以互补表征（complementary representations）为设计原则，构建双视图架构：一是折线图表征，保留绝对测量值和原始坐标系信息；二是归一化延迟嵌入（normalized delay embeddings），通过消除正幅值缩放与偏移量的影响，同时凸显时间结构。两种视图分别由专用视觉编码器处理，并通过视觉标记（visual tokens）和多层次特征融合整合至共享语言模型中。评估阶段的干预实验验证了两种视图具备不同的不变性特性，且单视图消融实验表明二者结合在预测与推理任务中显著优于基础感知能力。基于Qwen3-VL-8B构建的双视图模型将TSRBench准确率从仅使用折线图的44.5%提升至53.2%，远超当前最强专用TSLM基线（36.7%），并在TSExam上达到89.9%、多变量ChatTS对齐任务中实现86.8%的数值准确率。结果表明，互补表征能够有效支持通用型TSLM的设计，使模型既能利用时间结构，又不丢失原始测量框架的语义信息。
链接: https://arxiv.org/abs/2610.22327
作者: Hedi Zisling,Liran Nochumsohn,Ilan Naiman,Mihir Parmar,Omri Azencot
机构: Ben-Gurion University (本-古里安大学); Google(谷歌)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Time Series Language Models (TSLMs) enable language-based understanding and reasoning over temporal data, yet a fundamental question remains: what information should their time-series representations preserve, and what variation should they factor out? Existing TSLMs typically expose the language model to a single representation, imposing a fixed inductive bias across downstream tasks. We instead introduce complementary representations as a design principle for time-series reasoning. Our model combines two visual views: line plots preserve absolute measurements and the observed coordinate frame, while normalized delay embeddings factor out positive amplitude rescaling and offsets while exposing temporal structure. Dedicated visual encoders integrate both views into a shared language model through visual tokens and multi-level feature fusion. Evaluation-time interventions verify that the two views exhibit distinct invariance profiles, while single-view ablations show that their combination is particularly beneficial beyond basic perception, with the large gains in prediction and reasoning. Built on Qwen3-VL-8B, our dual-view model improves overall TSRBench accuracy from 44.5% with line plots alone to 53.2%, substantially exceeding the strongest evaluated dedicated TSLM baseline at 36.7%. It further achieves 89.9% accuracy on TSExam and 86.8% numerical accuracy on multivariate ChatTS alignment. Together, these results demonstrate that complementary representations provide an effective design principle for general-purpose TSLMs, enabling models to exploit temporal structure without sacrificing access to the original measurement frame.
[NLP-5] raining Reasoning Models to Generate Diverse Scientific Ideas
【速读】： 该论文旨在解决生成式人工智能在科研创意生成过程中存在的创意多样性不足问题。尽管现有研究多聚焦于提升单个创意的质量，却忽视了创意集合的整体多样性，而多样性的缺失可能导致创新潜力受限。为此，本文提出一种基于集合层级强化学习（set-level reinforcement learning）的解决方案，其核心在于通过联合评分机制优化一组创意的整体表现：在单次生成中，模型输出K个结构化的创意卡片，每个创意的质量得分会根据其与先前生成创意之间的语义距离进行加权调整，从而激励模型主动生成与已有创意差异显著的新想法。这一机制有效引导模型在保持高质量的同时增强创意间的区分度。实验结果表明，所训练的8B规模模型在无污染基准测试中，其综合衡量质量与多样性的“创造性效用”优于Claude Opus 4.8和GPT-6 Astra；初步的人工评估显示，专家认为其创意集与Claude Opus 4.8相当，且略具更多独特性；此外，在代码代理执行概念验证实验时，该模型的成功率@k与Claude Opus 4.8相当，证明增加多样性并未牺牲可执行性。
链接: https://arxiv.org/abs/2610.22324
作者: Manya Wadhwa,Pranav Narayanan Venkit,Hiroaki Hayashi,Yu Li,Tuhin Chakrabarty,Chien-Sheng Wu
机构: New York University(纽约大学); Salesforce AI Research; Stony Brook University(石溪大学)
类目: Computation and Language (cs.CL)
备注: Code and Data at: this https URL | Under Review
Abstract:One of the first steps of a research pipeline is generating ideas. Large Language Models (LLMs) are increasingly used to automate this step, but most of this work optimizes the quality of individual ideas, while overlooking diversity. In our work, we study document-grounded ideation, focusing on increasing the diversity of LLM generated ideas and ensuring they are high quality. Given prior papers, a model must propose a set of ideas that are each high-quality and clearly distinct from one another, while also being grounded in the input documents. To optimize for this, we propose training with a set-level reinforcement learning objective. In our setup, in a single generation, the model writes K structured idea cards grounded in the source papers, and the set is scored jointly, where each idea’s quality is weighted by its distance from the ideas written before it. This teaches the model to steer each new idea away from its predecessors. We show that our trained 8B model achieves higher creative utility (a measure combining quality and diversity) than Claude Opus 4.8 and GPT-6 Astra on our contamination-free benchmark. In a pilot human study with four experts, our sets are judged as promising as Claude Opus 4.8’s and contain marginally more unique ideas. When a coding agent runs proof-of-concept experiments on the ideas, our model’s success@k is comparable to Claude Opus 4.8’s, so we see that the added diversity does not cost any executability.
[NLP-6] MindEval2: Clinically Grounded Patients and Rubrics for Evaluating Language Models in Mental Health Support ALT
【速读】： 该论文旨在解决当前用于评估心理健康支持类AI聊天机器人的评测方法中存在的关键问题：现有基准测试在模拟多轮交互时依赖过于理想化且自我反思过度的合成患者，且评价体系多采用难以解释的李克特量表（Likert scale），缺乏实际临床指导意义。为此，论文提出MindEval2评估框架，其核心在于双组件设计——一是由博士级心理学家共同开发的“心理健康质量评分标准（Mental Health Quality Rubrics, MHQR）”，将响应质量分解为可操作、可解释的明确指标，显著提升评价结果的透明性与实用性；二是“MindSim”患者模拟系统，通过精心设计的患者档案与工具集成，实现更具多样性、挑战性与真实性的交互，并引入记忆更新机制以建模心理健康支持的纵向动态过程。在对18个专有及开源模型的评估中发现，尽管前沿模型在表面质量与边界处理上已有显著改进，但“引导校准”（guidance calibration）仍是普遍存在的主要缺陷，构成了当前性能差距的核心原因。
链接: https://arxiv.org/abs/2610.22314
作者: John Mendonça,Daniel Tschernutter,José Pombal,Areti Vassilopoulos,Catarina Botelho,Ignacio Sarasua,Maya D’Eon,Nuno Guerreiro,Ricardo Rei
机构: Sword Health(剑盾健康); NVIDIA(英伟达); Instituto Superior Técnico(里斯本技术研究所); Yale University(耶鲁大学)
类目: Computation and Language (cs.CL)
备注: 47 pages, 17 figures, 23 tables. Code, prompts, benchmark profiles and annotations: this https URL
Abstract:With people increasingly turning to AI chatbots for mental health support, the need to evaluate these tools is becoming more pressing. Recent benchmarks attempt to mimic a realistic setting by simulating multi-turn interactions with synthetic patients and evaluating them with an LLM-as-a-judge. However, the usefulness of these methods is limited because simulated patients can be overly forthcoming and self-reflective, and judgments typically involve opaque Likert scales with unclear practical utility. To bridge these gaps, we introduce MindEval2, an evaluation framework that consists of two components: Mental Health Quality Rubrics (MHQR), for evaluation, and MindSim, for patient simulation. For MHQR, we collaborate with PhD-level psychologists to develop a set of rubrics that break down response quality into explicit operational criteria that are more interpretable and actionable than existing evaluation taxonomies. MindSim, in turn, is an LLM harness that combines carefully-designed patient profiles with tools, enabling more varied, challenging, and realistic interactions than existing baselines. The harness also contains a memory component that is updated between each interaction, enabling MindEval2 to model the longitudinal aspect of mental health. We evaluate 18 proprietary and open-weight models on MindEval2, finding that while frontier models have largely improved on issues related to surface level quality and boundaries, guidance calibration remains the dominant failure mode across all models, accounting for most of the quality gap.
[NLP-7] Grid-Bench: Can LLM s Predict How Discrete Worlds Change?
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在处理复杂、多步空间操作序列时，对最终空间状态预测的准确性问题。具体而言，研究关注的是当多个空间操作按顺序作用于一个由离散网格表示的环境时，模型能否准确推断出经过一系列变换后的完整空间状态。其解决方案的关键在于提出Grid-Bench——一个可执行的基准测试框架，通过八类不同的空间操作，在二维、三维及四维离散网格上模拟动态变化过程。该框架以文本形式提供初始状态、明确规则和事件历史，要求模型预测完整的最终网格状态，并支持基于参考执行的逐元素验证，从而实现对不同空间操作类型、历史长度和任务负载下模型表现的精确量化比较。实验表明，尽管部分模型在单个事件上的精确匹配率相近，但在长历史序列中，其整体状态预测的准确率显著下降，且在相同实例中错误单元的数量存在显著差异，揭示了模型在状态演化推理中的细微差别。Grid-Bench使得完整状态恢复能力以及模型间差异在多种空间条件下均可直接测量，为评估和理解LLMs在空间推理任务中的行为提供了新工具。
链接: https://arxiv.org/abs/2610.22312
作者: Zixiang Xu,Zirui Song,Wanrong Zheng,Chiyu Zhang,Yue Huang,Yuyang Peng,Kexin Zhang,Jinkun Chen,Xiuying Chen,Xiangliang Zhang,Yan Liu,Laurent Itti,Frank Shyu,Xiaoqiang Luo,Sandeep Pandey,Xi Liu,Luke Simon,Xiyang Hu,Yue Zhao
机构: 未知
类目: Computation and Language (cs.CL)
备注: 39 pages, 15 figures, 15 tables, including appendices. Project website: this https URL ; Code: this https URL ; Data: this https URL
Abstract:Spatial instructions describe how to move objects, transform regions, or modify their surroundings. When several operations are applied in sequence, each acts on a configuration shaped by earlier changes. How accurately can large language models (LLMs) predict the resulting spatial state? We introduce Grid-Bench, an executable benchmark in which eight families of operations transform discrete worlds represented as two-, three-, and four-dimensional grids. Given an initial grid, explicit rules, and an event history expressed as text, a model predicts the complete final grid. Reference execution makes every predicted value checkable, supporting comparisons across spatial operations, history lengths, and workloads. Evaluating six LLMs shows that complete-state accuracy falls sharply as histories grow, with substantial divergence among models that perform similarly on single events. Models with similar exact-match accuracy can nevertheless differ substantially in the number of incorrect cells on shared instances. Even over long histories, some models recover many correct changes beyond what copying the initial state would yield. Grid-Bench makes both complete recovery and these differences in returned states directly measurable across a range of spatial conditions.
[NLP-8] Knowledge That Is Linear Before Fine-Tuning Generalizes After It
【速读】： 该论文旨在解决语言模型在知识记忆与泛化能力之间的不对称性问题，即并非所有可召回的事实都能同等支持模型在新任务中的泛化。其核心问题是：哪些事实能够促进模型从训练数据中提取出可迁移的结构化知识，从而实现跨样本的推理？ 解决方案的关键在于区分“活跃事实”（active facts）与“惰性事实”（inert facts）——前者所蕴含的语义结构在预训练表征中具有高度线性可读性（linearly readable），能支持模型在未见样本上进行有效泛化；后者虽可被准确回忆，但其结构无法被有效读出，因而不支持泛化。研究通过重构关系并施加随机标签扰动，发现模型能否将已学习的新关系推广至共享原答案的未见样本（如从“巴黎-螃蟹”推断“慕尼黑-螃蟹”），主要取决于原关系在预训练表示中的线性可读程度，而非单纯的记忆能力或训练样本数量。这一现象在11个语言模型和70个关系上均表现出强一致性（皮尔逊相关系数r = 0.87–0.92），且在控制多种变量后依然稳健，表明泛化能力的关键在于预训练表征中潜在结构的线性可解码性，而非事实本身的可回忆性。
链接: https://arxiv.org/abs/2610.22300
作者: Faridoun Mehri,Hinrich Schütze
机构: Center for Information and Language Processing, LMU Munich(慕尼黑大学信息与语言处理中心); Munich Center for Machine Learning (MCML)(慕尼黑机器学习中心)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 25 pages, 8 figures, 7 tables. Code and data: this https URL
Abstract:Not all facts a language model can recall support further learning equally well. We distinguish active facts, which support generalization beyond their own recall, from inert facts, which can be recalled but do not support such generalization in a given task. We study this distinction by constructing new relations from existing ones. Take the relation that maps a place to its continent (Paris - Europe) and assign each of its answers an arbitrary new label (Europe - crab, Asia - zebra). The new relation then maps Paris - crab. We fine-tune the model on the new relation for a subset of subjects, without telling it the original relation or the relabeling rule, and test whether it assigns the corresponding labels to held-out subjects that share an answer in the original relation. Having learned Paris - crab, does it answer Munich - crab? Across eleven language models and 70 relations, how linearly readable the original relation is in pretrained representations strongly predicts this generalization. Within each model, the two correlate across relations at Pearson r = 0.87 to 0.92. This association persists when controlling, separately and jointly, for the number of fine-tuning subjects and answers, the model’s accuracy on the original relation, and how often the subjects occur in pretraining text. It is also robust to re-randomizing the label assignment and fine-tuning seed, and to replacing the animal labels with labels from thirteen other semantic categories. When we map a fine-tuned model’s wrong label back to the original answer it was assigned to, Munich - zebra becomes Munich - Asia. Answers recovered this way are more often among the linear probe’s top answers for the subject than among the pretrained model’s own. Whether a new relation generalizes depends less on whether its facts can be recalled than on how readily their shared structure can be read out linearly.
[NLP-9] AI Consilium: a Benchmark of LLM Clinical Multidisciplinary Second Opinions
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在生成复杂临床病例的鉴别诊断时，如何确保其提供的第二意见具备事实依据、临床一致性、可操作性及安全性的问题。核心挑战在于，尽管LLMs能够生成诊断假设，但其输出往往缺乏严谨的临床推理链条与多角色协同整合能力，难以满足真实医疗决策场景的需求。为此，研究提出了一种多维度评估基准，基于300例已确诊的《新英格兰医学杂志》-医考题库（NEJM-MedQA）案例，构建了包含110例用于医生评分校准与裁判模型训练、190例用于独立评估的测试集。通过八种单模型流水线，在自由文本与结构化临床推理两种条件下生成独立角色意见及交互式会诊报告，并由经过医生校准的LLM裁判对3,040份最终报告及其原始角色输出进行评估。研究发现，模型身份对报告质量的影响显著高于响应格式；在七种模型流水线中，结构化最终报告整体得分更高，尤其在管理建议方面优势明显（2.740 vs. 2.436）。相较于结构化预会诊角色意见，完整的结构化综合报告在五项共享标准上均表现更优，尤其在问题优先级排序、事实依据和推理连贯性方面提升显著。GPT-5.6-Sol在所有评估维度中领先，并以最少的评审轮次达成共识。研究表明，该基准能有效区分模型能力差异、多学科协作的价值、响应结构对特定评价维度的影响、协商过程动态以及不同推理需求在统一病例评估中的作用。
链接: https://arxiv.org/abs/2610.22298
作者: Alexey A. Bokov,Ivan I. Panin,Nikolay B. Nechaev,Mikhail Yu. Valkov,Valeria Yu. Chernina,Natalia Gombolevskaya,Ayyaz Hussain,Vladimir M. Agafonov,Victor A. Gombolevskiy,Aleksandr V. Nesterov,Iaroslav R. Bespalov,Ivan V. Oseledets,Dmitry V. Dylov
机构: AIRI(人工智能研究院); Northern State Medical University (北方国家医科大学); Arkhangelsk Clinical Oncological Dispensary (阿尔汉格尔斯克临床肿瘤学诊所); Intelligent Radiology Assistance Laboratories LLC (IRA LABS) (智能放射学辅助实验室有限责任公司); “Professor Gorbakov Clinic” LLC (戈尔巴科夫教授诊所有限责任公司); Department of Anesthesiology and Intensive Care Medicine, First City Hospital named after E. E. Volosevich (埃·埃·沃洛谢维奇市立第一医院麻醉与重症监护医学部); Sechenov University (谢切诺夫大学); Skoltech (斯科尔科技研究所)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) can generate differential diagnoses for complex cases, but a useful second opinion must also be factually grounded, clinically coherent, actionable, and safe. We developed a multidimensional benchmark using 300 diagnostically resolved NEJM–MedQA cases: 110 for physician reference ratings and Judge calibration and 190 for held-out evaluation. Eight single-model pipelines generated independent role-specific opinions and interactive-consilium reports under free-form and structured clinical-reasoning conditions. A physician-calibrated LLM Judge evaluated 3,040 final reports and the corresponding structured primary-role outputs. Model identity separated report quality more strongly than response format. Structured final reports had higher overall mean scores in seven of eight model pipelines, with their clearest advantage in management (2.740 vs. 2.436). Compared with the structured pre-consilium primary-role opinion, the complete structured synthesis had higher pooled scores across all five shared criteria, with the largest increases in problem prioritization, factual grounding, and reasoning coherence. GPT-5.6-Sol led every evaluated component and reached consensus in the fewest review rounds. These findings show that the benchmark can distinguish model capability, the contribution of multidisciplinary synthesis, criterion-specific effects of response structure, deliberation dynamics, and inference requirements within one common-case evaluation.
[NLP-10] Multi-Target Few-Shot Prompting with Diffusion Language Models
【速读】： 该论文旨在解决在少样本提示（few-shot prompting）场景下，如何有效执行依赖上下文的文本编辑任务，尤其是在指令和奖励信号难以明确指定时。典型应用如“示例驱动的编辑”（editing by example），即需要将某一格式修改或API迁移操作泛化应用于多个不同上下文场景中，而这类编辑往往同时依赖左、右上下文信息，导致基于因果自回归模型的生成方式存在局限性。为此，论文提出一种基于扩散语言模型（diffusion language models）的多目标少样本提示方法，通过在共享上下文中对多个掩码区域进行联合填充，实现更精准的上下文感知编辑。其关键创新在于引入选择性掩码机制，利用似然比（likelihood-ratio）构造奖励函数，以强化模型对示例支持的填充内容的遵循能力，从而提升生成结果与示范样例的一致性。实验结果表明，在多个少样本提示基准测试及软件重构案例中，示例引导显著提升了填充准确性，尤其在复杂上下文依赖场景下优势明显。
链接: https://arxiv.org/abs/2610.22294
作者: Anton Xue,Alan Li,Rahul Saha
机构: University of Texas at Austin (德克萨斯大学奥斯汀分校)
类目: Computation and Language (cs.CL)
备注:
Abstract:Few-shot prompting is useful when instructions and rewards are hard to specify. One such application is editing by example, where a demonstrated formatting change or API migration may need to be applied across multiple sites. However, such edits often depend on both left and right context, making generation with causal autoregressive models awkward. We therefore introduce multi-target few-shot prompting with diffusion language models, where examples guide the infilling of multiple masked regions within a shared canvas. To address the known challenge of reliably following demonstrations in few-shot prompting, we further leverage selective masking of the canvas to derive a likelihood-ratio reward that favors example-supported infills. Across few-shot prompting benchmarks and software refactoring case studies, we find that demonstrations improve infilling accuracy, with example guidance yielding further gains in several settings.
[NLP-11] CAT-alyzing Honesty: Continuous Adversarial Training for Transferable LLM Honesty
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在强调诚实性的应用场景中仍普遍存在不诚实行为的问题，例如生成与已有知识相悖或缺乏上下文支持的陈述。其核心解决方案是提出一种名为连续对抗训练的一致性诚实（Continuous Adversarial Training for Consistent Honesty, CATCH）的方法，该方法通过在训练过程中同时引导模型趋向于诚实回答并远离不诚实输出，从而在保持模型原有能力的前提下实现对诚实性的增强。CATCH的关键创新在于采用自报告场景构建少量训练目标（仅数百条），并通过精心设计的效用数据选择策略，在显著提升模型诚实性的同时有效维持其功能性与实用性。实验结果表明，CATCH在多个基准测试中均表现出色，包括在压力情境下降低不诚实率、提高《Among Us》多智能体游戏中承认错误的比例，以及增强审计基准（AuditBench）中对模型隐藏行为的探测能力，且在性能上优于提示工程与监督微调，并在保持效用方面超越了当前最先进的诚实性激活引导（honesty activation steering）方法。
链接: https://arxiv.org/abs/2610.22292
作者: Yuan-Chen Chang,Niklas Herbster,Gauthier Gidel,Tommaso Tosato
机构: Tara Research; Mila – Quebec AI Institute; Technical University of Munich
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Computers and Society (cs.CY)
备注: 9 pages, 6 figures
Abstract:Large language models (LLMs) are becoming prevalent in scenarios where honesty is critical. However, despite alignment training, these models still exhibit dishonest behaviour, such as outputting claims that contradict some knowledge or are not supported by what is accessible in context. We introduce Continuous Adversarial Training for Consistent Honesty (CATCH), inspired by a method originally developed to robustly refuse harmful queries. CATCH trains the model towards an honest answer and away from a dishonest one while maintaining the model’s capabilities. We evaluate two architectures (Llama-3.3-70B-Instruct and Qwen3-3.6-27B). We show that trained with only hundreds of targets built from self-reporting scenarios, CATCH learns an honest behaviour that robustly generalizes and transfers across distinct benchmarks. It consistently reduces dishonesty under pressure on MASK, increases confessions in the multi-agent game of Among Us, and improves the discovery rates of the model’s hidden behaviours in AuditBench. We show that CATCH outperforms prompting and supervised fine-tuning in improving honesty, and it is on par with state-of-the-art honesty activation steering. However, CATCH surpasses honesty steering in maintaining utility, thanks to a novel choice of utility data.
[NLP-12] Librarian Agents : Training LLM s to Curate Long-Context Workspaces that Anticipate Future Queries
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在处理长文档上下文时面临的推理挑战，尤其是传统检索方法因被动响应机制导致的推理瓶颈，以及预先构建的知识图谱等结构因依赖固定启发式规则而灵活性不足的问题。其核心解决方案是引入“前瞻性定向编目”（anticipatory directed curation）机制，使智能体能够在推理前主动对非结构化语料进行重构，以高效支持未来多样化的查询需求。关键创新在于提出“图书管理员智能体”（Librarian Agents）框架：通过一个构建者智能体（builder agent），将原始文档转化为具有层次结构的合成文件与引用链接组成的可复用工作空间，从而为下游问答智能体提供优化后的信息组织形式。该工作空间的构建被建模为马尔可夫决策过程（Markov Decision Process），并采用强化学习直接优化构建策略，以最大化下游任务的问答准确率。实证结果表明，该方法在BrowseComp-Plus数据集上相较于最强基线提升33个百分点，能够通过重复查询摊薄前期构建成本，并在BRIGHT数据集上实现零样本迁移，平均准确率提升31个百分点，验证了其高效性与泛化能力。
链接: https://arxiv.org/abs/2610.22273
作者: Anikait Singh,Teresa Zhang,Yoonho Lee,Roshen Sanjay Nair,Aviral Kumar,Chelsea Finn
机构: Stanford University (斯坦福大学); Carnegie Mellon University (卡内基梅隆大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Reasoning over long document contexts remains a challenge for large language models (LLMs). Standard retrieval methods are inherently reactive, bottlenecking agents with complex, cross-document reasoning at inference time, while precomputed structures like knowledge graphs often fall short because they rely on rigid heuristics. Instead, can we allow agents to perform anticipatory directed curation: learning to proactively reorganize unstructured corpora prior to inference to efficiently serve diverse future queries. We build on a key asymmetry in modern agentic harnesses (e.g., Claude Code): while LLM agents are adept at navigating existing filesystems to solve an immediate task, they are rarely trained to construct reusable workspaces for future use. To instantiate this capability, we introduce Librarian Agents, a framework in which a builder agent transforms raw documents into a hierarchical filesystem of synthetic files and reference links, which forms a workspace tailored for a downstream answering agent. We formulate workspace construction as a Markov decision process and optimize the builder policy via reinforcement learning to directly maximize downstream answering accuracy. Empirically, Librarian Agents outperforms the strongest retrieval and precomputation baselines by 33 percentage points on BrowseComp-Plus, amortizes upfront construction costs across repeated queries, and generalizes zero-shot to unseen domains on BRIGHT with a 31-point average accuracy improvement.
[NLP-13] Not X but When: Diagnosing a Mandarin LLM Writing Tic through Contextual Appropriateness
【速读】： 该论文旨在解决生成式语言模型在使用特定语言表达时缺乏语用得体性（appropriateness）判断的问题，尤其关注模型是否能在恰当语境中合理使用具有修正功能的汉语结构“不是X，而是Y”（bushi X, ershi Y）。尽管此类表达常被视为机器写作的“标记”，但其本质是语用层面的交际能力体现，关键在于模型能否根据上下文判断某一表达是否必要且合适。研究基于格赖斯（Gricean）会话含义理论与修正性否定（corrective negation）框架，提出一个决策树，通过检验前提项X是否构成语境中的可选项以及否定X是否传递信息量来区分表达的三类：实质性（substantive）、可选性（optional）和缺陷性（defective）。实验发现，五款商用大模型使用该结构的密度为人类参考文本的2.5至9.2倍；在307个标注样本中，仅有1.30%为缺陷性使用，而高达52.8%属于可选性使用，表明模型在“是否应使用”这一得体性判断上显著失准，远超其在正确性方面的不足。此外，多个奖励模型偏好该结构，即使在不支持其使用的语境下亦如此，且监督微调与偏好优化均倾向于提升其频率，进一步加剧了不当使用。因此，该研究的关键在于揭示当前模型对语用得体性的忽视，并呼吁未来模型开发应更加关注表达的语境适切性而非仅追求形式正确或高频出现。
链接: https://arxiv.org/abs/2610.22249
作者: Yihang Peng,Yi Jing,Dewen Liu,Weihan Meng,Ying Liu
机构: Tsinghua University(清华大学); Fudan University(复旦大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Recurring expressions in model-written text are often treated as signatures of machine authorship, yet most of them serve legitimate functions. Whether such an expression is a tic depends on whether the model knows when it is called for. This judgment of appropriateness, central to communicative competence, remains largely unexamined for language models. We study it through the Mandarin corrective construction bushi X, ershi Y (“not X, but Y”). Drawing on Gricean pragmatics and corrective negation, we introduce a decision tree that classifies uses as substantive, optional, or defective by testing whether the context makes X a live alternative and whether rejecting X is informative. Five commercial models use the construction 2.5 to 9.2 times as densely as a human reference. Of 307 annotated uses, only 1.30% are defective while 52.8% are optional, that is, correct but unneeded, so models fall short in appropriateness far more often than in correctness. Several reward models favor the construction even over quality-matched variants, an LLM judge favors it even where the context does not license it, and both supervised fine-tuning and preference optimization tend to raise its frequency in open-weight models. We hope this work encourages the development of models attentive to the appropriateness of their expressions.
[NLP-14] Language Is a Security Surface for Agent Guards NEURIPS2026
【速读】： 该论文旨在解决多语言环境下，生成式AI（Generative AI）代理系统中步骤级防护机制（step-level guards）的决策是否受用户请求语言影响的问题。具体而言，现有防护模型大多在英文语境下训练与评估，但实际应用场景中用户请求可能涉及多种语言，因此亟需验证防护机制的跨语言鲁棒性。其核心解决方案是构建了LinguaGuardBench基准测试集，基于TS-Bench AgentDojo-Traj中的1,220个标注步骤，仅将用户指令翻译为德语、法语、土耳其语、斯瓦希里语和阿姆哈拉语，而保持所有工具调用、观测结果及标签字节完全一致，从而确保任何判决变化仅由语言差异引起。实验结果显示，多数开源、小型及专用防护模型在跨语言转换后出现显著判决翻转，且语言越罕见，影响越严重；尤其在参数量低于13B的模型中，阿姆哈拉语下的平均调整后翻转率高达+4.4个百分点。专用防护模型最为敏感，而大型前沿模型（如Gemini 3.8 Flash）则表现出极强的稳定性，几乎不受语言变化影响。进一步分析表明，错误集中于代理推理重述英文请求的特定步骤，暗示防护机制可能对多模态输入间的一致性敏感。研究还发现，英文改写对模型影响远小于直接翻译，说明语言结构差异是关键因素。研究团队已公开该基准及全部48.9万条判决结果，以推动多语言安全评估的发展。
链接: https://arxiv.org/abs/2610.22244
作者: Hasan Demirkiran,Jens Ernstberger
机构: Kontext
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted at the NeurIPS 2026 Workshop on Agents in the Wild: Safety, Security, and Beyond (AIWILD). 9 pages plus appendices
Abstract:Step-level guards decide, before each tool call, whether an agent’s next action is safe given the user’s request and the trajectory so far. Users write those requests in many languages, while guard prompts and the benchmarks on which purpose-built guards are trained and evaluated are English. We ask whether a guard’s verdict depends on the language of the request. We build LinguaGuardBench from the 1,220 labelled steps of TS-Bench AgentDojo-Traj: we translate only the user instruction into German, French, Turkish, Swahili and Amharic, and keep every tool call, observation and label byte-identical, so any change in a verdict is attributable to language alone. Six English rewordings and a whitespace placebo serve as controls, and managed guards are run three times to measure their noise. Across 19 guards, from frontier APIs to 4B open models and four purpose-built agent guards, most open, small and purpose-built guards change verdicts under translation beyond their own noise floor, and the effect grows with the rarity of the language: among the nine guards below 13B parameters, the mean placebo-adjusted flip rate rises from +1.5 pp on French to +4.4 pp on Amharic. Purpose-built guards are the most sensitive. The direction of the damage tends to follow guard size: open guards of 12-30B parameters block more benign steps (up to +5.0 pp false positives), and every significant loss of recall is in a guard below 10B (up to -7.4 pp). Frontier guards shift by at most about one point; the strongest, Gemini 3.8 Flash, returns identical verdicts with and without the request. The damage is concentrated in a few instructions and, in a post-hoc analysis, falls on steps where the agent’s reasoning restates the English request, consistent with guards reacting to inconsistency between channels. English paraphrase moves every guard less than translation does. We release the benchmark and all 489k verdicts.
[NLP-15] raining-Free On-Policy Distillation
【速读】： 该论文旨在解决多专家模型（specialist models）在策略学习中进行知识整合时所需大量训练开销的问题，即传统在线策略蒸馏（on-policy distillation）依赖于密集的轨迹采样、教师模型传递和梯度更新。其核心解决方案是提出一种无需训练的在线策略蒸馏方法（training-free on-policy distillation, \method），仅通过前向传播即可实现多个独立调优专家模型的能力聚合。关键在于：当各专家共享一个基础模型时，融合模型的每一层可通过求解一个精确的逐层最小二乘问题来实现，该问题只需各专家输入的二阶矩信息，而这些统计量仅需对少量（数百条）轨迹进行一次前向传播即可获得。研究进一步发现，轨迹内容的质量远重于优化算法本身——例如，在数学专家的生成轨迹中，重复循环截断的答案虽占比极小，却占据大量词元（tokens），移除此类冗余内容可显著提升融合模型在AIME25任务上的表现，使其达到与单个专家相当的精度。在Olmo 3 7B模型上，每个专家仅需数小时GPU计算资源进行前向传播，即可生成一个通用模型（generalist），在七项数学与代码基准测试中保留了各专家的主要性能增益。
链接: https://arxiv.org/abs/2610.22236
作者: Bangji Yang,Jiajun Fan,Zhaoyu Zhu,Hongbo Ma,Ge Liu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 20 pages, 6 figures, 9 tables
Abstract:On-policy distillation consolidates specialist models into one student at the price of rollouts, teacher passes and gradient steps for every teacher. We show that the consolidation does not need training: with forward passes only, the capabilities of separately tuned specialists can be aggregated into one model that is, on average, more accurate than either specialist alone. When specialists share a base model, distillation has an exact layer-wise solution: each linear layer of the fused model must reproduce every specialist’s output on the inputs that specialist receives while reading its own answers, a least-squares problem whose solution needs only the second moments of those inputs. One forward pass over a few hundred rollouts provides them. We call this training-free on-policy distillation (\method). What the rollouts contain matters more than the solver: answers truncated in repetition loops are few among a mathematics specialist’s rollouts but carry a large share of their tokens, and removing them lifts the fused model on AIME25 from task arithmetic’s level to the specialist’s at matched precision. On Olmo~3 7B, a few GPU-hours of forward computation per specialist yield a generalist that keeps most of each specialist’s gain on each of seven mathematics and code benchmarks.
[NLP-16] Beyond Downstream QA: Rethinking How We Evaluate Agent Memory
【速读】： 该论文旨在解决当前代理记忆评估范式中存在的根本性缺陷，即现有评估方法（如通过下游问答准确率衡量记忆质量）将记忆正确性与检索和生成质量混淆，导致产生假阳性（正确回答掩盖记忆失效）和假阴性（错误回答归因于记忆正确但检索失败）的问题。其解决方案的关键在于提出FEB-30基准，该基准通过“范式反转”构建：基于已知的事实演化轨迹自动生成文档，从而获得无需人工标注的确定性真实答案。该基准包含30个覆盖6个诊断类别和3个难度等级的事实演化轨迹，能够精确追踪多作者文本（如文档、邮件、对话）中事实的演变过程，并在每次信息摄入后直接检验记忆存储状态。实验表明，内在评估与外在评估之间平均存在16个百分点的偏差，44%的轨迹超出显著偏差阈值；诊断分解进一步揭示，系统对记忆操作（创建、更新、强化、覆盖）的正确分类率仅为20%，这一严重缺陷在传统外在评估中完全不可见。研究同时发布了基准数据、文档生成流程及评分框架，以支持代理记忆系统的可复现评估。
链接: https://arxiv.org/abs/2610.22209
作者: Sebastian-Antonio Toma,Jean de Kernier,Mike Ply,Oliviu Matei
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 14 pages, 1 figure, code and data: this https URL
Abstract:Agent memory benchmarks evaluate whether an agent’s responses reflect correct knowledge, measuring downstream question-answering accuracy as a proxy for memory quality. We show this evaluation paradigm is incomplete: it conflates memory correctness with retrieval and generation quality, producing both false positives where correct responses mask broken memory and false negatives where incorrect responses blame correct memory for retrieval failures. We introduce FEB-30, a benchmark of 30 fact-evolution trajectories spanning 6 diagnostic categories and 3 difficulty levels, constructed via paradigm inversion: documents are generated from known fact trajectories, yielding deterministic ground truth without human annotation. Each trajectory defines how a fact should evolve across multi-author artifacts (documents, emails, conversations), enabling direct inspection of the memory store after each ingestion step. Applying FEB-30 to a representative memory system, we demonstrate that intrinsic and extrinsic evaluation diverge by an average of 16 percentage points, with 44% of trajectories exceeding meaningful divergence thresholds. Diagnostic decomposition reveals that the system correctly classifies memory operations (create, update, reinforce, supersede) only 20% of the time, a failure entirely invisible to extrinsic evaluation, and that failure modes differ systematically by category and difficulty. We release the benchmark, document generation pipeline, and scoring framework to support reproducible evaluation of agent memory systems.
[NLP-17] Hy-Memory-30B-A3B: Attributed Reinforcement Learning for Long-Term Memory Writing
【速读】： 该论文旨在解决外部内存系统（External-memory systems）中长期记忆写入策略的泛化性与高效性问题，尤其聚焦于语言模型在写入路径中如何有效决定应保留的信息、精确表示方式以及新证据对持久状态的更新机制。现有框架普遍采用通用大语言模型配合特定协议提示词，导致服务成本高且模型难以跨框架迁移。本文提出一种统一的轻量化长时记忆写入模型——Hy-Memory-30B-A3B，通过在Mem0、MemOS和Hy-Memory三个主流内存框架上联合训练实现跨框架适配。其核心创新在于“基于策略的内存蒸馏”（Memory On-Policy Distillation）与“内存强化学习”（Memory RL）相结合：前者在学生生成的状态上学习各框架特有的写入协议，后者则通过密集的块级探针、跨块奖励分配、过存储控制及块级GRPO优化下游问答可答性。通过在不同窗口间保持策略写入的内存银行一致性，确保训练与部署高度对齐。实验表明，在固定20%基准子样本上，该模型在全部四个评估基准（包括未见的EverMemOS）中均排名第一，宏平均准确率达55.49%，优于Claude Opus 5的54.62%。
链接: https://arxiv.org/abs/2610.22196
作者: Tianxiang Fei,Jiazheng Kang,Mao Zheng,Xingyan He,Xiang Yu
机构: Tencent Hunyuan(腾讯混元)
类目: Computation and Language (cs.CL)
备注:
Abstract:External-memory systems extend language-model applications beyond a context window, but their quality depends critically on the model executing the write path: deciding what to retain, how precisely to represent it, and how new evidence should update persistent state. Existing frameworks typically place a general-purpose large language model behind framework-specific prompts. This raises serving cost, and a writer specialized for one protocol does not transfer directly to another; a unified model that can operate across established memory frameworks remains largely unexplored. We present Hy-Memory-30B-A3B, a compact long-term-memory writer trained across Mem0, MemOS, and Hy-Memory. Memory On-Policy Distillation teaches framework-specific write protocols on student-generated states; Memory RL then optimizes downstream answerability using dense per-chunk probes, attributed cross-chunk rewards, over-storage controls, and chunk-level GRPO. Carrying policy-written banks across windows keeps training aligned with deployment. On fixed 20% benchmark subsamples, Hy-Memory-30B-A3B ranks first on all four evaluated harnesses, including held-out EverMemOS, with a 55.49% macro average versus 54.62% for Claude Opus 5.
[NLP-18] HTN-Orchestrate: A Hierarchical Task Network Framework for Multi-Agent Orchestration
【速读】： 该论文旨在解决多智能体在复杂任务中协调执行时，如何有效平衡局部子任务执行与全局任务规划之间的矛盾问题。现有方法中，动态调度虽具备执行灵活性，但未显式建模子任务的层次结构与依赖关系；而预定义的工作流虽能明确任务分解结构，却在执行前固化了分解方式，缺乏适应性。为此，论文提出HTN-Orchestrate，一种基于分层任务网络（Hierarchical Task Network, HTN）的两层框架，通过将任务规划与执行解耦来实现优势互补：规划器（Planner）负责维护子任务的层级结构与依赖关系，并在前置结果就绪后才进行任务分解；本地编排器（Orchestrator）则负责调度具体工作者执行叶节点子任务。结果按AND语义向上聚合，当当前分解无法完成任务时，累积的发现将用于重构任务网络。在涵盖网页调研、科学推理和研究工作流的五个基准测试中，HTN-Orchestrate在相同基础模型下优于ReAct、Agent Team和DeepSeek Harness，分别在BrowseComp上达到89.73%准确率，在SUPERChem上达到73.44%，分别超过Agent Team 2.45和4.69个百分点。其解决方案的关键在于利用可动态重构的HTN结构，实现任务规划与执行的协同演化，从而在保持结构化推理能力的同时提升对复杂任务的适应性与执行效率。
链接: https://arxiv.org/abs/2610.22195
作者: Yanchen Deng,Jianguo Huang,Yao Long Teng,Yixuan Li,Handuo Zhang,Ning Wang,Liangcai Su,Xinyu Wang,Bo An
机构: Nanyang Technological University, Singapore; Apodex
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Orchestrating multiple agents on a complex task requires reconciling local subtask execution with global task planning. Dynamic dispatch approaches provide execution flexibility but do not explicitly model subtask hierarchies or dependencies, while predefined workflows impose this structure but fix the decomposition before execution begins. Therefore, we introduce HTN-Orchestrate, a two-layer framework that separates these responsibilities through a hierarchical task network (HTN). The Planner maintains subtask hierarchies and dependencies, deferring decomposition until prerequisite results become available, while local Orchestrators schedule workers to execute each leaf subtask. Results propagate upward under \textbfAND semantics, and accumulated findings inform a rebuilt task network when the current decomposition cannot resolve the task. On five benchmarks spanning web research, scientific reasoning, and research workflows, HTN-Orchestrate outperforms ReAct, Agent Team, and DeepSeek Harness using the same base model, achieving 89.73% accuracy on BrowseComp and 73.44% on SUPERChem, exceeding Agent Team by 2.45 and 4.69 percentage points, respectively.
[NLP-19] AutoSFT: An Agent ic System for Automated Data Research in Supervised Fine-Tuning
【速读】： 该论文旨在解决生成式 AI（Generative AI）在后训练阶段自动化数据优化过程中存在的核心问题：当前系统仅支持单一类型的数据操作（如从固定数据池中选择样本或调用生成器），无法全面模拟人类数据工程师对数据集进行比例调整、清洗、修复、替换和合成等多样化操作的能力。其解决方案的关键在于构建一个涵盖多种数据操作行为的“数据动作空间”（data action space），并提出一种多智能体架构——AutoSFT，通过元智能体（meta agent）将执行与分析分离，协同多个独立子智能体从不同角度提出改进方案，并引入合成子智能体以弥补数据池能力不足。实验表明，该架构显著提升了模型性能：执行模块减少上下文丢失，分析子智能体使IFEval提升+6.5，合成模块进一步带来+3.4 IFEval和+2.9 IFBench的增益；在代码任务中，分析模块实现+3.7 MBPP+的提升。所有结果均基于每轮从基础模型重新训练，确保各组件贡献可独立归因。研究结论指出，自动化监督微调（SFT）数据研究的成功更依赖于覆盖多样化的数据操作行为，而非单一强大操作的强化。
链接: https://arxiv.org/abs/2610.22194
作者: Bingxiang He,Pengqi Sun,Shihao Peng,Shaoxuan Li,Huiyu Xiao,Hongjian Shi,Zixuan Fu,Ziqing Qiao,Yuxin Zuo,Cheng Qian,Xinqiang Yu,Jiayi Li,Xiusi Chen,Yizhen Song,Xiangyang Wang,Iofai Lei,Yudong Wang,Ning Ding,Zhiyuan Liu,Xiaozhi Wang,Chaojun Xiao
机构: University of Illinois Urbana-Champaign; Tsinghua University; ModelBest Inc.; Beijing University of Posts and Telecommunications
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 27 pages, 16 figures
Abstract:Language model agents increasingly automate post-training, but when given the full training stack they converge on supervised fine-tuning and spend most effort revising training data. Existing systems give the agent one kind of control, selecting rows from a fixed pool or calling a generator, whereas a human data engineer adjusts proportions, prunes, repairs, replaces and synthesizes the data. We design a data action space for these changes and ask whether an automated system can make the full range rather than one of them, and what structure that requires. We first run this loop with a single agent, and it reaches only one action, adjusting category proportions, while its context fills with execution output and is discarded mid-search. We therefore introduce \textbfAutoSFT, a multi-agent system where a meta agent separates execution from analysis, gathers proposals from independent angles, and calls a synthesis subagent for capabilities the pool cannot cover, with the model, the trainer, and the budget fixed within each comparison. Adding each component in turn helps: the executor halves context loss, the analysis subagents lift the best IFEval by +6.5 , and synthesis adds +3.4 IFEval and +2.9 IFBench. The same loop transfers to code, where analysis adds +3.7 MBPP+ over a single agent that does not improve. Every result comes from runs that retrain from the base model each round, so each component’s effect is individually attributable. Our findings suggest that automating SFT data research depends more on covering diverse data actions than on a stronger single action.
[NLP-20] Change the Number Change the Answer: Assessing Rating Scale Sensitivity in LLM -Based Text Scoring
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在计算社会科学中用于对社会与心理构念赋值时，不同评分量表设计是否影响生成得分可比性的问题。其核心挑战在于，当前研究采用的量表形式多样，但缺乏对评分量表设计敏感性的系统评估。为此，作者提出了“量表敏感性”（scale sensitivity）这一概念，并将其细分为数值敏感性（numerical sensitivity，指归一化得分的变化）与语义敏感性（semantic sensitivity，指得分解释含义的改变）。研究通过跨量表格式（连续型与离散型）、范围（3、5、11、21点）、符号类型（带符号与无符号）以及多任务（政治倾向、情感、毒性）、提示策略（零样本与基于量规）和多种LLM的实验设计，系统检验了量表设计的影响。结果表明，仅调整一个量表设计要素，平均可导致归一化得分变化约0.10，并使约20%文本的分类归属发生改变；其中毒性判断的敏感性最高，因其为唯一单极构念。此外，基于量规的提示策略并未显著降低敏感性。研究的关键解决方案在于揭示量表设计对LLM生成评分的实质性影响，强调在构建基于大语言模型的文本评分体系时，必须进行跨量表的鲁棒性检验，并提供更可靠的量表设计指南。
链接: https://arxiv.org/abs/2610.22187
作者: Wonjeong Jo,Gyuwan Kim,Ming Shan Hee
机构: UC Davis (加州大学戴维斯分校); UC Santa Barbara (加州大学圣塔芭芭拉分校); MBZUAI
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Numerical Analysis (math.NA)
备注:
Abstract:Large language models (LLMs) are increasingly used in computational social science to assign scalar scores to social and psychological constructs. However, studies use different rating scales, and whether their scores remain comparable across scale designs is understudied. To address this gap, we define and systematically investigate scale sensitivity, the variation in LLM-generated scores caused by rating scale design. We distinguish numerical sensitivity, changes in normalized scores, from semantic sensitivity, changes in score interpretation. To assess both, our experimental design crosses 2 formats (continuous vs. ordinal), 4 ranges (3, 5, 11, and 21 points), and 2 signs (signed vs. unsigned) across 3 tasks (political leaning, sentiment, and toxicity), 2 prompting strategies (zero-shot and rubric-based), and 5 LLMs. We find that changing one design element shifts normalized scores by about 0.10 on average and changes the category of roughly 20% of texts. Sensitivity is greatest for toxicity, the only unipolar construct examined. Also, rubric-based prompting does not reduce sensitivity compared with zero-shot prompting. We provide practical guidance for designing more reliable LLM-based text scoring and highlight the need for robustness checks across alternative scales.
[NLP-21] EVA-Cap: Optimizing Audiovisual Video Captioning via Event-Centric Alignment NEURIPS2026
【速读】： 该论文旨在解决现有通用语言模型（Omni Language Models, OLMs）在音视频字幕生成任务中将字幕生成视为无结构文本生成过程所导致的细粒度音视频对齐不佳问题，其核心在于忽视了视频内在的语义结构。解决方案的关键在于提出EVA-Cap框架，通过将字幕分解为原子级音视频事件（atomic audiovisual events），并将传统的无结构文本监督转化为以事件为中心的对齐机制。该框架在自构建的以事件为中心进行质量控制的数据集EVA-Data上，采用监督微调（Supervised Fine-Tuning, SFT）与组相对策略优化（Group Relative Policy Optimization, GRPO）联合训练，在GRPO阶段显式优化三个事件中心对齐目标：全局事件完整性、事件内归因准确性和事件间同步性。实验表明，EVA-Cap在事件中心评估指标上显著优于现有方法（如ChronusAV上平均提升9.0%），同时在通用音视频字幕基准测试中也表现出竞争力（如WorldSense上提升3.3%）。
链接: https://arxiv.org/abs/2610.22186
作者: Jiapeng Shi,Qiuxia Hou,Jiaqi Leng,Junke Wang,Yanhao Zhang,Haonan Lu,Ziyi Ye,Zuxuan Wu
机构: Fudan University (复旦大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: Accepted at NeurIPS 2026. Code will be released at this https URL
Abstract:Recent advances in Omni Language Models (OLMs) provide a unified solution for audiovisual video captioning. Nevertheless, existing approaches predominantly treat captioning as an unstructured text generation process, neglecting the inherent semantic structure of videos and leading to suboptimal fine-grained audiovisual alignment. To bridge this gap, we propose EVA-Cap, a novel framework that decomposes captions into atomic audiovisual events and transforms unstructured text supervision into event-centric alignment. We train EVA-Cap via Supervised Fine-Tuning (SFT) and Group Relative Policy Optimization (GRPO) on a self-curated dataset, EVA-Data, which is built with event-centric quality control. During the GRPO phase, the model is explicitly optimized against three event-centric alignment targets: global event completeness, intra-event attribution, and inter-event synchronization. Extensive experiments demonstrate that EVA-Cap significantly outperforms existing methods in event-centric evaluation (e.g., +9.0% average gain on ChronusAV), while achieving competitive performance on general audiovisual captioning benchmarks (e.g., +3.3% on WorldSense).
[NLP-22] Mosaic: Multi-Session Agent ic Reinforcement Learning
【速读】： 该论文旨在解决现有生成式智能体强化学习（Agentic Reinforcement Learning, ARL）方法在多会话场景下的适应性问题。传统ARL方法通常假设所有交互发生在单一连续会话中，然而实际应用中诸如上下文压缩（context compaction）和子智能体分叉（subagent forking）等技术可能导致轨迹跨越多个独立会话，从而导致训练时的单会话假设与推理时的实际多会话执行之间存在不匹配。为此，论文提出多会话生成式强化学习（Multi-Session Agentic Reinforcement Learning），一种支持任意数量会话的更通用设置。针对现有方法在多会话中简单广播轨迹级优势信号、无法区分各会话贡献或指导会话分配的问题，本文提出Mosaic方法，其核心在于采用无评判器（critic-free）的分层优化机制：在轨迹层面，将会话使用信号融入优势估计；在会话层面，根据各会话的边际贡献动态调整其优势值。实验表明，在BrowseComp任务上基于Qwen3.5-4B模型，Mosaic在上下文压缩与子智能体分叉两种场景下均优于传统广播策略，且平均使用会话数更少。进一步分析显示，该方法显著提升了会话利用率与分配效率，验证了其作为未来多会话生成式强化学习研究关键方向的潜力。
链接: https://arxiv.org/abs/2610.22174
作者: Yilong Xu,Minrui Xu,Xinyu Zhou,Xiang Long,Haochen Tan,Zhijiang Guo
机构: LARK, HKUST(GZ)(香港科技大学（广州）; LARK); Foundation Model Lab, Huawei(华为基础模型实验室); City University of Hong Kong(香港城市大学); HKUST(香港科技大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Most existing agentic reinforcement learning methods assume single-session trajectories, where interaction proceeds within one continuously accumulated context. However, agentic techniques such as context compaction and subagent forking may produce trajectories spanning multiple sessions, creating a mismatch between the single-session training assumption and inference-time execution. Therefore, we introduce Multi-Session Agentic Reinforcement Learning, a more general setting that supports an arbitrary number of sessions within a trajectory. Under this setting, a widely adopted method is to broadcast trajectory-level advantages to all sessions, but this does not distinguish contributions from various sessions or guide session allocation within trajectories. To address this issue, we propose Mosaic, a multi-session agentic RL method with critic-free hierarchical optimization. At the trajectory level, it incorporates session-usage signals into advantage estimation. At the session level, it shapes each session’s advantage according to the estimated marginal contribution. Experiments on BrowseComp with Qwen3.5-4B show that Mosaic outperforms broadcasting in both context compaction and subagent forking scenarios while using fewer sessions on average. Further analyses show that our method achieves improvements in session utility and allocation, which we suggested is a promising target for future advances in multi-session agentic RL.
[NLP-23] Lexical Reference Frames: Interpreting LM Residual Streams in Native Semantics
【速读】： 该论文旨在解决大语言模型中隐藏状态（hidden state）的可解释性问题，即现有方法无法准确捕捉模型在处理语义时对状态的实际读取方式。当前方法或使用固定方向的注意力透镜（lens），或将其视为无结构的点云（point cloud），均未能反映模型自身动态上下文感知的读取机制。其核心解决方案是提出“词汇参考系”（Lexical Reference Frame, LexFrame），该框架通过利用模型剩余层自适应地读取状态，以模拟模型内部真实的语义理解过程：在特定位置间进行若干轮注意力交互，并采用典型注意力输出作为其余部分的近似，无需拟合模型最终输出。实验表明，传统固定透镜的相干性上限仅为0.30–0.35（在SmolLM3-3B与OLMo3-7B中），而一阶词汇参考系LexFrame[1]的相干性高达0.97，且能以仅0.16纳特的误差逼近模型输出。该方法在预测归因上优于归因修补（attribution patching），其词汇基底可精准引导特定预测而不影响其他标记，同时具备高置信度读数能力，使早期退出策略可跳过约45%和42%的层，显著优于传统透镜。因此，词汇参考系实现了对模型每一隐藏状态的可读性、可归因性与可操控性。
链接: https://arxiv.org/abs/2610.22169
作者: Jian Gu,Hongyu Zhang,Chunyang Chen,Aldeida Aleti
机构: Monash University(蒙纳士大学); Chongqing University(重庆大学); Technical University of Munich(慕尼黑工业大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Code: this https URL
Abstract:The vocabulary of a language model is its native semantics: the embedding and the unembedding fix the meaning of a state at the two ends of the residual stream. Between them, existing methods either reuse a fixed lens or treat the states as a point cloud, and neither reads a state the way the model does. We argue that every hidden state has a native reference frame, induced by the layers that read it next and adaptive to the context. Lexical reference frames approximate it by reading a state through the model’s own remaining layers with a stated number of rounds of attention between positions and a typical attention output for the rest, without fitting to the model’s output. No lens, with fixed directions per layer, can exceed the square root of the native frame’s coherence, 0.30 and 0.35 at mid-depth of SmolLM3-3B and OLMo3-7B, whereas the first-order frame, LexFrame[1], reaches 0.97 and 0.97 and reads a state within 0.16 nats of the model’s output. Its coefficients attribute predictions more faithfully than attribution patching, its lexical basis steers a prediction while sparing other tokens, and its confident readings would let an early exit skip 0.45 and 0.42 of the layers, where lenses skip almost none. In this way, lexical reference frames make every hidden state of the model readable, attributable and steerable.
[NLP-24] CLUE: Uncovering Social Context for Prag matic Understanding
【速读】： 该论文旨在解决对话理解中缺乏对说话者关系、预期及先前交流内容的有效利用问题，即如何让模型在不依赖外部显式上下文的情况下，自动识别并运用对话内部的关键线索以支持社会性推断。其核心解决方案是提出一种名为CLUE的两阶段训练框架，关键在于结合有监督的证据推理与基于“基础信息利用（Grounded Information Utilization, IU_G）”引导的GRPO（Generalized Reward Policy Optimization），通过奖励机制同时鼓励生成新且由证据支持的社会推断，并考虑推理过程的效率。该方法使模型在少量对话样本上训练后，能够在推理时自主、自洽地进行上下文驱动的推理，显著提升跨任务、跨语言和跨文体场景下的泛化能力。实验表明，相较于基线Qwen模型，CLUE在域内平均加权F1提升17.3%，域外提升9.3%，且平均推理令牌数减少超80%；人工评估也验证了其解释更紧密地锚定于具体对话内容，与人类选择的上下文具有更高一致性。
链接: https://arxiv.org/abs/2610.22167
作者: Chengfei Wu,Zhaoqing Wu,Dan Goldwasser
机构: Purdue University (普渡大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Understanding a conversation requires connecting what is said to the speakers’ relationships, expectations, and earlier exchanges. Prior work shows that supplying relevant context can improve models’ social judgments, but much of this information may already be present in the conversation. This motivates training models to discover the cues that matter and use them to support an interpretation. We propose CLUE, a two-stage training pipeline that combines supervised evidence reasoning with GRPO guided by Grounded Information Utilization ( \mathrmIU_\mathrmG ). This reward credits new, evidence-supported social inferences while accounting for reasoning effort. We train on a small set of conversational examples to assess effectiveness and generalization across tasks, languages, and genres. At inference time, the trained model reasons over the supplied interaction in a self-contained manner. Compared with base Qwen, CLUE achieves relative gains of 17.3% in-domain and 9.3% out-of-domain in average weighted F1, with higher F1 on four of five out-of-domain evaluations. These gains are accompanied by over 80% fewer reasoning tokens on average in both domains. Human evaluation further finds that CLUE more often grounds its explanations in specific conversational content, with closer alignment between its quoted evidence and human-selected context.
[NLP-25] EMO: A Closed-Loop Test-Time Computation Framework via Alternating Context Optimization and Logits Steering
【速读】： 该论文旨在解决冻结语言模型在测试时通过文本反馈进行上下文优化过程中，传统条件生成无法对上下文更新所引发的分布变化强度实现独立控制的问题。其核心解决方案是提出一种无需训练的框架EMO（Expectation-Maximization Optimization），该框架受期望最大化（Expectation–Maximization, EM）思想启发，通过交替执行问题特定的上下文估计（E步）与分布精炼（M步）来实现高效适应。E步利用外部评判器提供的文本梯度，在不访问真实答案的情况下修正系统上下文；M步则在共享前缀下对比修订前后上下文的下一个词分布，构建置信度约束的逻辑值校正项。反馈驱动的响应引导下一轮上下文更新，实现了生成过程内部干预与多轮适应的耦合。理论上，作者通过共享目标KL散度目标函数和可加的逻辑值表示，将参数优化与上下文优化统一，并从上下文诱导的逻辑值位移中推导出导向向量。在上下文改进与忠实导向假设下，证明了当导向强度足够小时，单轮即可实现目标KL散度下降。实验表明，在八种模型与六项基准测试中采用贪婪解码时，EMO在40个聚合设置中的39个上优于零样本基线，平均提升9.10个百分点；相较于仅优化上下文的方法TextGrad，进一步平均提升1.53个百分点，且在各模型上均保持稳定的MMLU-Pro性能增益。在相同生成预算下，其准确率优势在Llama系列模型上最为显著，尽管候选数量较少，但其覆盖率更高，揭示出生成与验证计算资源需联合分配以缩小覆盖率-选择性差距。
链接: https://arxiv.org/abs/2610.22160
作者: Wei Xia,Yehui Tang,Zhi-Hong Deng
机构: Peking University (北京大学); Samsung Research (三星研究院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Test-time context optimization adapts frozen language models through textual feedback, but ordinary conditional generation provides no separate control over how strongly to reinforce the distributional changes induced by a context update. We introduce EMO, a training-free framework inspired by expectation–maximization that alternates question-specific context estimation (E-step) and distribution refinement (M-step). The E-step revises a system context using textual gradients from an external judge without access to gold answers. The M-step contrasts next-token distributions under the revised and preceding contexts at a shared prefix to construct a confidence-constrained logit correction. Feedback on the steered response drives the next context update, coupling adaptation across rounds with intervention within each generation. Theoretically, we connect parameter and context optimization through a shared target-KL objective and additive logit representations, and derive the steering vector from context-induced logit displacements. Under context-improvement and faithful-steering assumptions, we prove a one-round decrease in target KL divergence for sufficiently small steering strength. Under greedy decoding across eight models and six benchmarks, EMO improves the zero-shot baseline in 39 of 40 aggregated settings, averaging +9.10 percentage points. With the same judge, it adds +1.53 points on average over context-only optimization (TextGrad), with consistent MMLU-Pro gains across models. At matched target-model generation budgets, accuracy advantages over best-of- N are strongest on Llama, while oracle coverage is higher in most comparisons despite fewer candidates. The coverage–selection gap motivates allocating test-time computation jointly to candidate generation and verification.
[NLP-26] PremiseSQL: Evaluating Premise Critique and Abstention in Text-to-SQL Systems
【速读】： 该论文旨在解决现有自然语言转SQL（NL2SQL）系统在面对存在前提错误的用户查询时，仍倾向于生成可执行但误导性SQL的问题。当前主流大语言模型（LLM）普遍假设用户输入查询均为有效且可回答，缺乏对查询前提合理性的批判能力，导致在逻辑错误或事实错误的查询下仍输出看似正确实则无意义的SQL语句。为此，论文提出引入“前提批判能力”（premise critique ability），即模型识别查询中前提缺陷并主动拒绝生成可能误导性的SQL的能力。其解决方案的关键在于构建首个专门评估该能力的基准数据集PremiseSQL，包含1,800个具有四类前提错误的有缺陷查询及其对应的原始有效查询，并设计了一套涵盖三种提示策略的评估框架，以量化模型在内在倾向与提示诱导下的前提敏感行为。实验表明，当前前沿大模型默认具有强烈的“前提信任”倾向，极少对错误前提进行批判；尽管采用更具批判导向的提示策略能提升对缺陷查询的检测与拒答能力，但也带来了更高的误判率和对原始有效查询的过度抑制，揭示了该能力提升中的权衡挑战。
链接: https://arxiv.org/abs/2610.22158
作者: Shuo Lin,Jinzhe Li,Chenlu Guo,Yanhan Li,Yuan Wu,Yi Chang
机构: Jilin University (吉林大学); Bilibili Inc.; China Telecom Artificial Intelligence Technology (Beijing) Co., Ltd (中国电信人工智能技术（北京）有限公司); Engineering Research Center of Knowledge-Driven Human-Machine Intelligence, MOE (教育部知识驱动人机智能工程研究中心)
类目: Computation and Language (cs.CL)
备注:
Abstract:Recent advances in large language models (LLMs) have substantially improved Text-to-SQL performance, yet existing NL2SQL systems largely assume that user queries are valid and answerable. As a result, models often generate executable SQL even for queries with flawed premises. We introduce premise critique ability in NL2SQL, namely the ability to recognize flawed queries and abstain from generating misleading SQL. To evaluate this capability, we propose PremiseSQL, a benchmark containing 1,800 flawed NL2SQL queries across four categories of premise failures, along with their 1,800 original queries. We further design an evaluation framework to measure intrinsic and prompt-induced premise-aware behavior across three prompting settings. Experiments on frontier LLMs show that current models are strongly premise-trusting by default and rarely critique flawed premises. Stronger critique-oriented prompting improves detection and abstention on flawed queries, but also increases false invalid judgments and suppresses SQL generation on originals. The code is available at this https URL.
[NLP-27] Intern-Decision: Efficient Multimodal Decision Making with Autoregressive Masked Language Modeling
【速读】： 该论文旨在解决交互式智能体在决策过程中因生成自由文本响应而导致的延迟增加与系统接口复杂化问题。传统方法在每一步决策中生成自然语言输出，不仅引入显著延迟，还增加了模型与环境间通信的复杂性。为此，论文提出Intern-Decision系列多模态决策模型（0.8B、2B、4B参数量），其核心解决方案是采用结构化输出机制，通过在预定义的助手JSON模板中为每个决策字段设置占位符，利用自回归掩码语言建模目标进行端到端训练。该方法仅在监督阶段使用真实标签，推理时通过骨干模型的原生语言模型头部在单次因果前向传播中完成所有字段的预测，实现高效、低延迟的结构化决策输出。实验表明，4B模型在七个评估套件上达到90.02%的未加权平均准确率，优于Jev v1.13.0的88.74%；在单次289词请求下，4B模型平均延迟为44.16毫秒，小模型延迟约33–34毫秒。此外，温度校准有效降低多分类Brier损失（从0.628降至0.550）和校准误差（从0.213降至0.089），而基于置信度的移交策略将111个高难度任务的准确率从73.87%提升至87.39%。这些结果验证了紧凑型结构化决策模型作为大型智能体系统实用组件的可行性。
链接: https://arxiv.org/abs/2610.22131
作者: Tong Zhu,Zhe Xu,Zefeng He,Yunkai Chai,Xuyang Hu,Dehong Zheng,Yun Luo,Yafu Li,Qi Zhang,Yu Cheng
机构: Shanghai AI Lab(上海人工智能实验室); Shanghai Jiao Tong University(上海交通大学); Fudan University(复旦大学); Nanyang Technological University(南洋理工大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Interactive agents repeatedly choose among a finite set of actions, labels, or ordinal outcomes. Generating a free-form response for each choice adds latency and complicates the interface between a model and its environment. We present Intern-Decision, a family of 0.8B, 2B, and 4B multimodal decision models that return structured choices and candidate probabilities from text and optional images. Our autoregressive masked language modeling objective places one decision placeholder per field in a complete assistant JSON skeleton. Ground-truth symbols appear only in supervision, while all fields are evaluated in a single causal forward pass using the backbone’s native language-model head. The 4B model achieves a 90.02% unweighted average accuracy across seven evaluation suites, compared with 88.74% for Jev v1.13.0. For a fixed 289-token request with three fields on one RTX 4090, mean latency is 44.16 ms for the 4B model and approximately 33–34 ms for the smaller models. On a separate 96-case known-distribution pilot, temperature scaling reduces expected multiclass Brier from 0.628 to 0.550 and expected calibration error from 0.213 to 0.089. A confidence-based handoff study on 111 public Jevbench-Hard items improves accuracy from 82/111 (73.87%) to 97/111 (87.39%) by routing 42 items to Atria-Dawn-Preview. We also describe qualitative game, visual interface, and browser demonstrations. These results support compact decision models as a practical component of larger agent systems.
[NLP-28] Explaining the Roles of Experts in Mixture-of-Experts LLM s for Personalization
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）个性化过程中因密集型模型（dense LLMs）中存在多义神经元（polysemantic neurons，即每个神经元编码多个无关概念）而导致个性化组件难以定位、进而阻碍对个性化机制的理解与控制的问题。其核心解决方案是采用稀疏专家混合模型（Mixture-of-Experts, MoE）架构进行个性化，利用MoE模型中每个输入标记仅激活少数专用专家的特性，显著降低多义性，从而实现对个性化成分的精准定位。研究通过在大规模个性化数据集上适配三款开源权重的MoE LLM，并训练跨编码器（crosscoder）以学习个性化与非个性化模型之间激活特征的对应关系，进而提取可解释的个性化潜在表示（personalized latents）。实验表明，该方法在保持与密集型模型相当性能的同时，仅激活极小比例参数，且在相同活跃参数规模下表现出更低的多义性；此外，个性化潜在表示在模型内部具有更少与原始非个性化模型的共享性。因果干预实验进一步验证，这些潜在表示足以恢复个性化模型的行为，且增强其关联专家可引导非个性化模型向个性化方向演进，揭示了不同个性化特征存在独立的路由路径。
链接: https://arxiv.org/abs/2610.22127
作者: Sekh Mainul Islam,Pepa Atanasova,Isabelle Augenstein
机构: University of Copenhagen(哥本哈根大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Under review
Abstract:Personalization of Large Language Models (LLMs) is increasingly important for user-centric applications, yet prior methods mainly adapt dense LLMs, whose polysemantic neurons, each encoding multiple unrelated concepts, hinder localizing the components behind personalization, and thus understanding and controlling it. We propose to personalize Mixture-of-Experts (MoE) LLMs, which sparsely route each token to a few specialized experts, reducing polysemanticity and enabling such localization. We adapt three open-weight MoE LLMs on a large-scale personalization dataset and train crosscoders on paired activations of the personalized and non-personalized models to learn interpretable personalized latents, from which we identify the experts encoding distinct personalization features. The personalized MoE LLMs perform competitively with dense LLMs while activating only a fraction of their parameters, exhibit lower polysemanticity than dense LLMs with similar active parameters, and, within the same model family, yield personalized latents that are less shared with the non-personalized model. Causal interventions show that these latents suffice to recover the personalized model’s behavior and that boosting their associated experts moves the non-personalized model toward it, revealing distinct routing paths for different personalization features.
[NLP-29] Unfrag menting Nawatl NMT: Single-Variant Specialization for Masewalkopa
【速读】： 该论文旨在解决少数族裔（低资源）语言技术中因数据稀缺而导致的模型性能瓶颈问题，尤其聚焦于墨西哥原住民语言纳瓦特语（Nawatl）多种变体在混合训练时产生的“变体碎片化”（variant fragmentation）问题——即混合多种密切相关的语言变体所形成的语料库生成了一种既不属于任何单一使用者群体的虚假语域，导致模型难以准确捕捉真实语言特征。针对此问题，研究提出的关键解决方案是：基于7万对多变体语料构建基础模型，并通过专业化微调（specialization）将mT5-large模型适配至圣安娜特拉科滕科方言（Masewalkopa, nhn）。研究通过精心收集2,665组平行语料、保留全部句级数据用于评估，并引入192条由长者口述转录的目标句进行非对称反向翻译（asymmetric backtranslation），以增强语义真实性与语言多样性。此外，采用自定义的有限状态分析器（finite-state analyzer）将未分析表面形式的比例从82.0%降至19.9%，并以纯文本前缀形式注入功能语法标签，使模型可复用现有子词嵌入，避免依赖冻结的特殊标记。最终，所提出的LoRA-v9c-b模型在开发集上达到58.46的chrF++得分，在双向测试集上分别取得58.20和57.90的得分。通过CTranslate2的int8量化技术，推理延迟从137.5秒/请求显著降低至1.8–2.6秒，实现了成本可控的云部署。
链接: https://arxiv.org/abs/2610.22110
作者: Luis A. Mercado-Campos,Alexis Palmer
机构: University of Colorado Boulder(科罗拉多大学博尔德分校)
类目: Computation and Language (cs.CL)
备注:
Abstract:In minoritized (low-resource) language technologies, a common response to data sparsity is to aggregate text across multiple varieties and treat the combined data as a single corpus. However, mixed corpora of closely related indigenous variants can produce an artificial register that is accountable to no single speaker community, a structural failure we call variant fragmentation. This issue is acute for minoritized languages native to Mexico, such as Nawatl. Building on a 70,000-pair multi-variant foundation, we specialize an mT5-large checkpoint to Masewalkopa (nhn), the Nawatl variety spoken in Santa Ana Tlacotenco. We curate 2,665 parallel pairs, reserve all sentence-level data for evaluation, and incorporate 192 elder-transcribed target sentences via asymmetric backtranslation. A custom finite-state analyzer reduces unanalyzed surface forms from 82.0% to 19.9% and supplies functional syntax tags as plain-text prefixes, allowing the model to reuse existing subword embeddings rather than relying on frozen special tokens. The resulting LoRA-v9c-b model achieves 58.46 dev chrF++ and 58.20/57.90 directional test chrF++. Finally, CTranslate2 int8 quantization reduces CPU latency from 137.5 to 1.8-2.6 seconds per request, enabling cost-bounded Cloud Run deployment.
[NLP-30] Patch where the text is new: a compression view of where byte-level models should spend compute
【速读】： 该论文旨在解决生成式语言模型在有限补丁（patch）预算下对计算结果（如等号后数值）识别准确率严重下降的问题。现有方法依赖于小模型预测的下一字节熵值作为补丁起始依据，但此类规则往往忽略内联计算结果区域，导致其在这些关键位置的补丁分配不足。本文提出“新颖性补丁”（novelty patching）策略，受LZ77压缩思想启发：在即将出现的字节难以复现近期上下文时启动新补丁。该策略通过一个仅依赖前两个字节的查表机制实现，可在无需标签、参考模型或损失函数的情况下，仅用数秒时间在原始文本上快速拟合。实验表明，在1.9 GB数学语料上从头训练的字节级大模型（BLT风格）中，该方法在3种子实验下以约11%的补丁率显著提升性能：当模型参数为12M时，最终答案准确率由47%提升至72%；50M参数时，计算结果准确率从21%升至57%，最终答案准确率从60%升至90%。且使用预训练语料拟合的查表仍能保持增益。分析显示，计算结果区域对补丁器模型而言具有常规意外度（surprisal），因此基于意外度的规则会自然跳过这些区域——熵值驱动规则虽在整体边界增强上表现最佳（边界富集度1.61），却最差于计算结果区域，而新颖性规则则在内容起始与计算紧密耦合的场景（如书面数学）中表现最优。此外，一种无监督的分块网络（H-Net）可自主发现此类结构，适用于数学和代码。文中还排除了若干无效方案：基于类型不确定性的训练规则、训练后切换规则以及上下文感知的新颖性预测器均未带来有效改进。
链接: https://arxiv.org/abs/2610.22100
作者: Nicolás Vera Zúñiga
机构: Independent Researcher(独立研究员); Chile(智利)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 10 pages, 2 figures, 5 tables. Code, logs and results: this https URL
Abstract:Byte-level language models such as the Byte Latent Transformer (BLT) run their large global model once per patch and start a patch where a small model’s next-byte entropy is high. This rule rarely starts a patch at in-line computed results, such as the number after “=” in a worked solution, and accuracy on them collapses under tight patch budgets. We propose novelty patching, motivated by LZ77-style compression: start a patch where the upcoming bytes are unlikely to repeat the recent context. The rule is a lookup table over the previous two bytes, fitted in seconds on raw text with no labels, reference model or losses. In BLT-style models trained from scratch on a 1.9 GB math corpus at a matched ~11% patch rate (3 seeds each), it has no clear effect at 1M parameters, raises final-answer accuracy from 47% to 72% at 12M, and at 50M raises computed-result accuracy from 21% to 57% and final-answer accuracy from 60% to 90%; a table fitted on the raw pretraining corpus keeps the gain. Outside math the advantage shrinks: novelty lowers bits per byte on prose and ties entropy on source code and on synthetic execution traces. An analysis of where boundaries fall explains the pattern. Computed results have ordinary surprisal for the patcher’s model, so surprisal-seeking rules pass over them: entropy concentrates patch starts on high-surprisal bytes most strongly of all rules (boundary enrichment 1.61) and is the worst on computed results, while the best rules sit near chance enrichment. Novelty works where new content begins at segment starts close to the computation, as in written math; a learned chunker (H-Net) finds such structure on its own, in math and in code. We also report what did not work: typed uncertainty as a training rule, switching rules after training, and a context-aware novelty predictor.
[NLP-31] Multi-SFT: Permissively-Licensed Multilingual Instruction-Following
【速读】： 该论文旨在解决当前监督微调数据集（Supervised Fine-Tuning Datasets, SFT）高度依赖英语且受商业限制的问题，即多数高质量数据由专有模型生成，其服务条款禁止用于训练竞争性系统，从而阻碍了多语言、开放生态的发展。为此，论文提出Multi-SFT，一个具有商业宽松许可的多语言指令-响应对数据集，涵盖五种欧洲语言，共计502,818条样本，通过翻译与再生成流水线构建，以弥补单纯翻译在语义和任务适配上的不足。其解决方案的关键在于：结合自动化翻译与基于模型的语义保持再生成技术，在确保语言多样性的同时维持数据质量；并通过人工评估验证各语言数据集间标注者一致性均不低于0.87，证明数据可靠性。实验表明，仅使用Multi-SFT微调的九个来自三个不同模型家族（EuroLLM、OLMo 2、Qwen3；参数量1B–32B）的模型，在13个基准测试中表现可媲美更大规模增强混合数据集及官方指令微调版本，尤其在安全相关任务上展现出显著优势。
链接: https://arxiv.org/abs/2610.22086
作者: Daniela Baiamonte,Kamyar Zeinalipour,Elena Fano,Stefano Simonazzi,Andrea Zugarini
机构: villanova.ai; Yukai!; University of Siena(锡耶纳大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 13 pages, 7 tables
Abstract:Supervised fine-tuning datasets remain overwhelmingly English-centric and commercially restricted, as their content is frequently generated by proprietary models whose terms of service forbid training competing systems. We introduce Multi-SFT, a commercially permissive collection of 502,818 instruction-response pairs across five European languages, built via translation and regeneration pipelines for tasks where translation is insufficient. Human evaluation yields inter-annotator agreement of at least 0.87 across all datasets and languages. Fine-tuning nine models from three different families (EuroLLM, OLMo 2, Qwen3; 1B-32B) and evaluating them on 13 benchmarks shows that Multi-SFT alone is competitive with larger augmented mixes and official instruction-tuned releases, with decisive advantages on safety tasks.
[NLP-32] he Right Byte First: How UTF-16 Helps Multilingual Boundary Prediction
【速读】： 该论文旨在解决字节级语言模型在处理多语言文本时，因UTF-8编码特性导致的分词边界预测不准确问题。具体而言，在多种使用婆罗米系文字（Brahmic scripts）的语言中，由于部分字符采用三字节UTF-8编码，其首字节仅携带4位有效代码点信息，而后续字节则包含更多语义信息，使得在相同左文上下文中，一个音素簇的延续标记与下一个字符的起始字母可能共享首字节，从而引发边界预测歧义。为解决此问题，论文提出通用字节级编码（Universal Byte-Level Encoding, UBE），将一、二字节的UTF-8编码保持不变，但将三字节及以上代码点以两个UTF-16字节（低位字节优先）的形式表示，使首个符号承载8位低序代码点比特，显著增强对连续性特征的区分能力。实验表明，在17种语言的13种婆罗米系文字上，UBE在单符号前瞻条件下，代码点边界处的平衡音素簇边界错误率降至0.06%，远优于UTF-8（8.76%）和其同长度大端序变体（8.62%）。当目标保留父分词器在音素簇边界处的子词切分时，UBE相较UTF-8在韩语、中文及八种其他脚本中的平均错误率分别降低72.91%、35.29%以及67.32%至99.55%。进一步分析显示，在主要Unicode区块可容纳于单一256代码点页的七种语言中，大端序变体性能与UTF-8差距不超过3.0%，说明性能提升主要源于字节顺序而非长度变化。此外，韩语和中文中，UBE的平均错误率在训练完成仅25%和37.5%时已低于UTF-8的最终表现，表明字节顺序直接影响字节级边界预测器所需的前瞻量。因此，解决方案的关键在于通过优化编码方式（如低字节优先的双字节表示）提升首字节的信息表达能力，从而减少对长距离上下文依赖，降低模型对前瞻的需求。
链接: https://arxiv.org/abs/2610.22071
作者: Hyunsik Kim
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 32 pages, 5 figures, 29 tables
Abstract:Many hierarchical byte-level language models group bytes into patches using a learned boundary predictor whose view of the next codepoint depends on encoding and lookahead. In UTF-8, codepoints of many scripts take three bytes: 8 of 24 bits mark length and continuation, and the first byte carries only 4 high codepoint bits. After the same left context, a mark extending a grapheme cluster and a letter starting the next can thus share their first byte. Universal Byte-Level Encoding (UBE) keeps one- and two-byte UTF-8 encodings but writes these codepoints as two UTF-16 bytes, low byte first, so its first symbol carries 8 low codepoint bits that can distinguish such continuations. We train small causal predictors on matched FineWeb-2 text and evaluate on FLORES-200 under UTF-8, UBE, and UBE’s same-length big-endian variant. At one-symbol lookahead, balanced grapheme-boundary error at codepoint boundaries across 17 languages in 13 Brahmic scripts is 8.76% with UTF-8, 8.62% with the big-endian variant, and 0.06% with UBE. When the target instead retains a parent tokenizer’s subword cuts at grapheme boundaries, UBE lowers the mean error relative to UTF-8 at the same lookahead by 72.91% in Korean, 35.29% in Chinese, and 67.32-99.55% in eight more scripts. In the seven of them whose main Unicode block fits on one 256-codepoint page, the big-endian variant stays within 3.0% of UTF-8, so the gain follows byte order, not length. In Korean and Chinese, UBE’s mean error falls below UTF-8’s final mean error after only 25% and 37.5% of training. Which byte comes first thus affects how much lookahead a byte-level boundary predictor needs.
[NLP-33] Beyond Context Length: Characterising the Difficulty of Long-Context Reasoning
【速读】： 该论文旨在解决现有长上下文语言模型评估中过度依赖上下文长度作为难度指标的局限性问题，揭示在相同上下文长度下模型表现差异的根本原因。其核心问题是：为何某些长上下文任务即使在足够长的输入窗口下仍难以被模型正确处理？解决方案的关键在于提出并验证三个可量化的复杂度维度——证据范围（evidence scope）、证据分散度（evidence dispersion） 和 推理深度（reasoning depth），用以更精确地刻画长上下文推理任务的真实难度。通过构建LURCH基准数据集（包含6,550个实例及支持证据与推理图），研究发现：在固定上下文长度下，这三个属性均与模型性能显著负相关；随着上下文长度增加，证据分散的任务性能急剧下降，而证据局部化任务则保持稳定；且三者难度具有协同效应，同时具备中等水平高值的题目远比单一维度极端的题目更难。这表明，仅考虑上下文长度不足以反映真实推理挑战，必须结合证据分布与推理结构特征进行综合评估。
链接: https://arxiv.org/abs/2610.22067
作者: Shiwen Qin,Cameron Barker,Henry Gouk,Elliot J. Crowley,Shay B. Cohen,Mirella Lapata,Miao Li
机构: University of Edinburgh (爱丁堡大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Long-context language models support ever longer context windows, and evaluations typically treat context length as the main indicator of difficulty. However, context length alone cannot explain why models fail on long-context tasks that require reasoning over complex information. Contexts of equal length can pose very different challenges: a task may be solvable from a single local passage, or it may require integrating evidence distributed across the full context. We quantitatively characterise this difficulty through three properties: evidence scope, how much necessary information must be found; evidence dispersion, how widely that information is scattered within the context; and reasoning depth, how many dependent steps are needed to derive an answer from it. We construct LURCH, a multi-domain question-answering benchmark of 6,550 instances with supporting evidence and reasoning graphs, enabling all three properties to be quantified. Evaluations of proprietary and open-weight models show that: (1) at fixed context length, all our three properties are strongly negatively correlated with model performance; (2) as context length increases, performance degrades sharply on samples with dispersed evidence, whereas it remains relatively stable on samples with localised evidence; and (3) difficulty compounds across properties, as questions with moderately high values on all three are harder than those with an extreme value on just one. These findings establish evidence scope, dispersion and reasoning depth as measures of long-context reasoning difficulty that complement context length.
[NLP-34] Name-Bias in Large Language Models
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）中由姓名所隐含的年龄、性别和种族等人口统计学特征引发的系统性偏见问题。由于姓名本身携带了显著的社会文化信号，此类偏见可能在模型生成内容时被放大，进而影响公平性与代表性。研究的关键在于通过设计一种基于填空任务的标准化评估框架，对6个来自不同提供商、参数规模各异的本地化大语言模型进行系统性测试。实验采用美国社会保障局（SSA）及Rosenman等人构建的种族分类数据集中的姓名配对，并通过平衡性别与年龄因素、保持句法结构一致的句子模板，有效隔离了年龄与性别效应。结果表明，模型表现出显著且可量化的姓名偏见：更偏好非典型男性化、年长、流行度高且白人背景的名字。此外，研究发现姓名偏见具有高度依赖于上下文的特性，但其程度仍低于相关刻板印象的影响；其中性别是主导性偏见因子，其次为年龄、名字流行度与种族。该研究揭示了即使在本地部署的模型中，姓名偏见依然普遍存在，凸显了建立标准化评估方法与针对性缓解策略的紧迫性。
链接: https://arxiv.org/abs/2610.22053
作者: Kalyan Bhetwal,Ceren Mert,Ian Whitehouse
机构: University of Maryland, College Park, USA(马里兰大学帕克分校)
类目: Computation and Language (cs.CL)
备注:
Abstract:Names carry demographic indicators such as age, gender, and race, which makes them a subtle but strong target of bias in large language models (LLMs). In this work, we present an evaluation of name-based bias across 6 local large language models spanning multiple providers and parameter scales. Using a fill-in-the-blank task, we prompt these models to choose between pairs of first names drawn from U.S. Social Security Administration datasets and a race classification dataset by Rosenman et al… Our experimental de- sign isolates age and gender effects through balanced name pairings and sentence templates that share a fixed syntactic structure. Across 300,000 local LLM responses, we find consistent and measurable name bias. Models prefer names that are less stereotypically male, older, popular, and white. Furthermore, we find that name-bias is highly sentence-dependent, but not more than associated stereotypes, and gender is the most significant bias, followed by age, name popularity, and race. Our results demonstrate that name-based bias remains pervasive even in local LLMs, highlighting the necessity for standardized evaluation methods and mitigation strategies.
[NLP-35] Effectiveness of Hidden-State Probes for Localizing Reasoning Failures NEURIPS2026
【速读】： 该论文旨在解决长链推理（long reasoning trace）中错误溯源问题，即当最终答案被验证为错误时，无法区分错误的根本原因属于模型能力不足、可通过重新尝试修复，或由某一特定步骤导致整个推理路径不可挽回地失败。现有方法如基于最终答案正确性的隐藏状态探针（hidden-state probe）因标签混淆了这三种情形而难以准确分类。虽然蒙特卡洛滚动（Monte Carlo rollouts）可有效区分并定位失败步骤，但需额外的推理开销。研究通过对4B和8B规模推理模型的隐藏状态进行分析，发现尽管存在不同训练设计的探针，其对错误类型排序的结果依赖于训练方式，且在定位具体失败步骤方面，无一显著优于仅基于推理长度的启发式规则；尤其在4B模型上，一个专门训练以预测失败步骤的探针在10次重复实验中均未超越该基线。因此，尽管置信度探针（confidence probe）能指示答案错误，但其对错误类型的排序结果受训练设计影响，无法提供可靠的一致性诊断。关键在于：当前隐藏状态探针虽能反映整体推理质量，却难以在不引入额外计算成本的前提下实现精确故障归因。
链接: https://arxiv.org/abs/2610.22046
作者: Shubham Subhnil,Yannis Belkhiter,Erik Miehling,Adam Dahlgren Lindström,Elizabeth M. Daly
机构: Trinity College Dublin (三一学院都柏林大学); IBM Research (IBM 研究院); Umeå University (于默奥大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Accepted at the NeurIPS 2026 Workshop on Interpretability as a Science (InterpScience). OpenReview: this https URL . 30 pages, 5 figures, 13 tables. Code: this https URL ; Data: this https URL
Abstract:Verifying the final answer of a long reasoning trace may show that the trace is wrong but it does not distinguish three cases. The problem may lie beyond the model’s capability, another attempt would succeed, or one step in an otherwise recoverable trace made the trace fail. Standard hidden-state probes are trained on final-answer correctness, a label that conflates these cases. Monte Carlo rollouts distinguish the cases and locate the failing step at additional inference cost. We test the hidden states of 4B and 8B reasoning models for the failure type and the failing step and compare each probe with the strongest baseline that uses no hidden states. On incorrect 8B traces, the confidence probe, a calibrated correctness probe at a midpoint cutoff, misses 81% of cases in which a single step makes the trace unrecoverable and 47% of cases in which another attempt would succeed. A probe with a different training design over the same states reverses the ordering of these two kinds of failure, so either ordering can be read from those states. At both scales, no single tested probe locates the failing step more accurately than a rule based only on trace length, and at 4B a probe trained to predict that step fails to outperform this rule in any of 10 runs on identical traces. The confidence probe therefore indicates that an answer is wrong, but its failure-type ordering depends on its training design.
[NLP-36] Momentum Commentary Tone and Earnings Surprise in LLM Financial Forecasts
【速读】： 该论文旨在解决生成式 AI 在金融预测任务中对输入信息的依赖机制问题，具体探究哪些输入因素（如动量、非信息性语调、实际盈余与修正共识的偏离程度）驱动语言模型生成财务预测结果。其解决方案的关键在于构建一组高度可控的合成盈利公告，通过独立操控20日动量、非信息性评论语气以及实际每股收益（EPS）与人为调整共识之间的差异等变量，系统性地评估不同模型在多种情景下的响应模式。研究采用六种不同规模的语言模型（参数量0.49–7.61B），在统一指令/对话转换架构下进行大规模测试（每模型2,016个提示），并引入多维度控制设计（包括数值块、评论对、盈亏状态、年度EPS趋势及惊喜幅度等共384种主情景组合），以量化模型对各类输入的敏感性。关键发现表明，在多数模型中，语调的符号化对比幅度超过EPS本身，动量的影响亦显著高于EPS；同时，自然语言中的数值诊断准确率介于51.0%至95.8%之间，反映模型在数值识别与编码方面存在明显差异。研究通过答案编码、措辞、上下文及转换控制确保结果可靠性，但明确指出其仅衡量输入依赖性与数值识别能力，不涉及内部推理机制或实际投资回报可预测性。研究公开了基准提示、评分记录及交互式可视化探索工具，为后续分析提供透明支持。
链接: https://arxiv.org/abs/2610.22043
作者: Alina Khaybullina
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); General Finance (q-fin.GN)
备注:
Abstract:Which inputs drive a language model’s financial forecast? We construct synthetic earnings announcements that independently reverse 20-day momentum, non-informational commentary tone, and whether fixed actual EPS beats a manipulated consensus. Six instruction/chat checkpoint conversions from four families (0.49-7.61B base parameters) receive 2,016 prompts each under a direct-answer Q8_0 protocol. The design includes 24 numerical blocks, six commentary pairs, profits and losses, rising and falling annual EPS, and two surprise sizes. Both A/B answer assignments and U/D cover all 384 primary scenarios per conversion. We estimate matched changes in conditional answer-prefix scores, not calibrated market probabilities. Across these checkpoints, the primary EPS contrast ranges from -0.19 to 86.80 percentage points. The magnitude of the primary signed tone contrast exceeds that of EPS in 5 of six models, and momentum exceeds it in 5. Natural-word numerical diagnostic accuracy ranges from 51.0 to 95.8 percent. Answer-encoding, wording, context, and conversion controls qualify these results. The experiment measures observable input reliance and numerical recognition; it establishes neither an internal reasoning mechanism nor realized return predictability. Public benchmark prompts, recorded scores, and an interactive recorded-results explorer accompany the study.
[NLP-37] WEAVE: Weaving Real-World Code Evolution into Dependency-Structured Tasks for Benchmarking and Training Code Agents
【速读】： 该论文旨在解决现有代码生成与修复任务评估框架中长期存在的问题：现实软件工程中的功能实现、迁移及系统性修复通常涉及多个跨拉取请求（Pull Request, PR）的协同修改，而现有主流的仓库级可执行基准测试仍以单一问题或单个PR作为任务单元，导致多PR间的依赖关系被忽略，且现有多PR评估方式依赖粗粒度聚合和无结构化的评价标准，无法准确捕捉变更间的依赖性，限制了对长周期代码智能体（code agents）的进展衡量与学习信号供给。其解决方案的关键在于提出WEAVE框架，该框架通过利用经过验证的PR历史数据，结合仓库校准的类型化变更定位（typed change loci），将相关联的PR组织为有界有向无环图（bounded DAGs），从而构建具备依赖结构的可执行任务；同时引入Graph Credit这一确定性、基于测试的度量指标，能够衡量符合依赖关系的阶段性进展，并支持密集、可验证的奖励信号，适用于强化学习训练。实验表明，WEAVE不仅生成结构清晰、逻辑连贯的可执行任务，相比传统平坦评分方式更真实地反映进度，且能有效提供监督信号与可验证奖励，显著提升代码智能体的训练效能。
链接: https://arxiv.org/abs/2610.22039
作者: Weihao Li,Zihang Xu,Zhongqiu Wang,Songwei Yu,Wenqiang Zhu,Yuechen Luo,Jingxuan Xu,Yanan Wu,Haoyang Huang,Ken Deng
机构: 未知
类目: oftware Engineering (cs.SE); Computation and Language (cs.CL)
备注: 22 pages, including appendix
Abstract:Real-world software engineering rarely consists of isolated changes. Features, migrations, and systematic fixes often unfold across multiple pull requests (PRs), requiring coordinated modifications across a repository. Yet most executable repository-level benchmarks retain a single issue or PR as the task unit, while existing multi-PR settings rely on coarse aggregation and structure-agnostic evaluation. This mismatch obscures dependencies among changes and limits progress measurement and learning signals for long-horizon code agents. We introduce WEAVE, a framework that transforms verified PR histories into executable, dependency-structured tasks. WEAVE leverages repository-calibrated typed change loci to organize related PRs into bounded DAGs and instantiate each as a unified executable task. We further introduce Graph Credit, a deterministic, test-grounded metric that measures dependency-consistent partial progress and supports dense, verifiable rewards for reinforcement learning. Experiments on task construction, agent evaluation, supervised fine-tuning, and reinforcement learning show that WEAVE yields coherent executable tasks, enables more faithful progress measurement than flat scoring, and provides effective supervision and verifiable rewards for training code agents.
[NLP-38] Selection Is Not Learning: Choosing Model Futures
【速读】： 该论文旨在解决自演化语言模型在迭代过程中因任务选择与模型更新之间存在“选择-更新鸿沟”（selection-to-update gap）而导致的无效更新继承问题。现有方法如R-Zero、Agent0和技能自对弈（Skill Self-Play）通过任务可学习性判断来决定是否继承更新，但即便任务看似可学，其产生的模型更新仍可能无法提升正确答案的概率，导致大量无用甚至有害的更新被采纳。为解决此问题，论文提出两种核心方案：HeirCommit采用冻结更新模型并使用新生成的评估问题进行验证，从而在确认阶段淘汰无效更新；EvoHeir则进一步将参数继承与经验继承解耦，利用当前与归档的经验修复失败、保留成功，再进行最终确认。分析表明，该框架有效缓解了由选择引发的乐观偏差，并将总损失分解为选择、吸收与确认三类误差。实验结果显示，相较于随机继承，HeirCommit在相同接受率下减少了68%-83%的有害更新；在四种不同骨干模型上，EvoHeir在数学推理与通用推理任务中均取得最佳表现，其中数学平均得分较最强基线Agent0提升5.3%。该框架确保了有价值的经验即使在原始模型未被继承的情况下仍可得以保留与复用。
链接: https://arxiv.org/abs/2610.22029
作者: Haoyi Yang,Kshitij Mishra,Nils Lukas,Salem Lahlou
机构: Mohamed bin Zayed University of Artificial Intelligence (穆罕默德·本·扎耶德人工智能大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Self-evolving language models train on tasks they generate, so every round must decide which updated model the next round inherits. Current loops such as R-Zero, Agent0, and Skill Self-Play make this decision through the task, admitting tasks that look learnable and inheriting every resulting update. We identify a selection-to-update gap in these loops, where a task that looks learnable need not produce a useful update. Most updates on verified tasks fail to make correct answers more likely, yet these loops inherit them all. HeirCommit freezes the updated model and uses fresh evaluation questions to decide whether to inherit it or retain the current model. EvoHeir separates parameter inheritance from experience inheritance, using current and archived experience to repair failures and preserve successes before final confirmation. Our analysis characterizes selection-induced optimism and decomposes regret into selection, absorption, and confirmation errors. HeirCommit cuts the harmful share of accepted updates by 68-83% relative to random inheritance at the same acceptance rate. Across four backbones, EvoHeir achieves the best mathematics and general-reasoning averages, raising the math average by 5.3% relative to the strongest baseline, Agent0. The framework lets useful experience persist even when the model that produced it is not inherited.
[NLP-39] Priming Base Models for Agent ic Post-Training: An Open Midtraining Recipe
【速读】： 该论文旨在解决前沿模型在预训练（pretraining）与后训练（post-training）之间缺乏系统化中段训练（midtraining）阶段的问题，尤其关注如何有效提升模型在智能体（agent）后训练中的表现。现有方法中，预训练通常在短上下文长度下进行且几乎不涉及智能体交互，而监督微调（SFT）数据多为合成生成，且过长的SFT阶段可能损害后续强化学习（RL）训练的可训练性。为此，论文提出了一种开源的中段训练方案：在256K上下文长度下，使用包含代码仓库、拉取请求、提交历史和构建日志等多样化自然软件工程数据，并融合少量智能体轨迹（agentic traces），对基础模型进行数百亿token的训练。该方案的关键在于通过引入真实世界软件开发场景的复杂性和长期依赖特征，增强模型对长序列任务的理解能力，从而显著提升其在多个基准测试中的表现——例如，在Nemotron 3.5 Lightning上，中段训练使SWE-bench Verified得分从60.7提升至64.5；在更大规模的Nemotron 3.5 Super上，DeepSWE（长周期任务挑战）得分从11.7提升至15.7。此外，中段训练还增强了模型在不同智能体框架下的鲁棒性。研究进一步揭示了该方案的有效性归因于其数据多样性与上下文长度设计，并建立了基于固定短时SFT运行的评估协议，以克服基础模型代理指标与后训练结果之间不可靠的相关性问题。
链接: https://arxiv.org/abs/2610.22019
作者: Soumye Singhal,Joseph Jennings,Ka Shun Shum,Alexander Bukharin,Khushi Bhardwaj,Zirui Liu,Venkat Srinivasan,Jonathan Li,Yash Jain,Kezhi Kong,Markus Kliegl,Olivier Delalleau,Sanjeev Satheesh,Mostofa Patwary
机构: NVIDIA(英伟达)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 19 pages
Abstract:Frontier models, general-purpose and domain-specific alike, now include a midtraining stage between pretraining and post-training, but little has been published about how this stage is built. We present an open midtraining recipe that primes base models for agentic post-training, and we show its effectiveness on two Nemotron models of different sizes. Pretraining typically runs at short context and sees almost no agent interactions, while SFT data is mostly synthetic, and very long SFT stages can hurt the model’s later RL trainability. Our midtraining stage sits between the two. It trains at 256K context for hundreds of billions of tokens on a diverse mixture of natural software-engineering data like packed repositories, pull requests, commit histories, and build logs together with a small share of agentic traces. On Nemotron 3.5 Lightning, 500B midtraining tokens improve SWE-bench Verified from 60.7 to 64.5, SWE-bench Multilingual from 55.8 to 62.8, and SWE-bench Pro from 32.0 to 35.5 after identical full SFT. On the larger Nemotron 3.5 Super, 280B midtraining tokens improve DeepSWE, a harder benchmark of long-horizon tasks, from 11.7 to 15.7 and SWE-bench Verified from 72.9 to 74.3 after identical full SFT. Midtraining also improves robustness across agent harnesses on both models. We report the ablations behind the recipe and an evaluation protocol built on short fixed SFT runs, since base-model proxy metrics did not correlate reliably with post-training results.
[NLP-40] P2TPO: Structure-Aligned Page-to-Token Credit Assignment for Full-Page Document Parsing
【速读】： 该论文旨在解决端到端全页文档解析中因局部错误与遗漏导致的输出质量下降问题，尤其关注在多个采样输出中难以明确识别具体元素的质量差异及其错误位置。现有方法依赖于页面级整体优势评估，无法有效揭示不同元素间的质量对比及误差发生的具体位置。为此，论文提出结构对齐的信用分配框架——页面到标记策略优化（Page-to-Token Policy Optimization, P2TPO），其核心在于通过参考元素对齐机制将块级质量差异映射至生成片段；引入统一锚点定价（Unified Anchor Pricing, UAP）以处理匹配与缺失元素的反馈，并在相同但不完美的参考元素评分下提供相对于固定质量上限的相对反馈；在匹配块内，采用片段信用重分配（Span Credit Redistribution, SCR）机制，基于对齐的正确与错误片段重新分配块内信用，利用边界代理处理块内遗漏问题，同时保持整个块的总信用值不变。该方法在固定权重加性页面效用下，能够保留被页面级聚合所掩盖的局部质量差异。实验表明，P2TPO在六个文档解析基准上均显著优于所有对比的策略优化基线，且在不同模型规模和初始化条件下均展现出一致性能提升。
链接: https://arxiv.org/abs/2610.21979
作者: Yong Zu,Ninghua Yang,Zezhong Tan,Xinhong Ma,Ziqiang Dong,Yang Yang
机构: Qwen Business Unit, Alibaba(通义业务单元，阿里巴巴); Xidian University(西安电子科技大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 34 pages, 9 figures, including appendices
Abstract:End-to-end full-page document parsing must preserve the content and structure of heterogeneous elements, yet otherwise accurate outputs can contain local errors and omissions. A single page-level advantage does not explicitly identify which elements differ in quality across sampled outputs or where their errors occur. We introduce Page-to-Token Policy Optimization (P2TPO), a structure-aligned credit-assignment framework that operates on full-page policy trajectories. Reference-element alignment first associates block-level quality contrasts with generated spans. Unified Anchor Pricing (UAP) incorporates matched and missing element outcomes and provides feedback relative to a fixed quality ceiling when sampled outputs receive identical but imperfect scores for a reference element. Within matched blocks, Span Credit Redistribution (SCR) redistributes existing block credit according to aligned correct and erroneous spans, using boundary proxies for within-block omissions while preserving the total block-derived advantage across its tokens. Under a fixed-weight additive page utility, we show that element-indexed contrasts retain local quality differences that page-level aggregation can cancel out. In the main experimental comparison, P2TPO outperforms all evaluated policy optimization baselines in seed-averaged scores on each of the six document parsing benchmarks. Additional experiments demonstrate gains across model scales and initializations.
[NLP-41] Learning Speech Attribute Representations via Multi-Teacher Distillation
【速读】： 该论文旨在解决单一自监督语音编码器（Self-Supervised Speech Encoder, SSL）在后训练阶段如何同时保留多个互补的说话人级语音表征（utterance-level speech representations）的问题。传统方法通常通过知识蒸馏（knowledge distillation）仅学习单一语音属性，难以实现多任务联合建模。为此，本文提出一种多属性多教师蒸馏框架（multi-attribute multi-teacher distillation framework），利用四个冻结的教师模型分别提供语义、说话人、语言和情感信息，通过共享的SSL骨干网络与特定属性的分支结构，实现对多种语音属性的联合学习。其关键创新在于：在保持单个共享编码器的前提下，通过多教师协同蒸馏机制有效整合多种语义层次的表征，并揭示不同语音属性在预训练编码器中自然分布在不同深度的层级结构，为语音表示的组织方式提供了新的认知依据。实验结果表明，该模型在多语言语义检索、说话人验证、语言识别和语音情感识别等任务上均达到接近专用单任务模型的性能，显著提升了模型的通用性与效率。
链接: https://arxiv.org/abs/2610.21941
作者: Maryem Bouziane,Salima Mdhaffar,Yannick Estève
机构: LIA (Avignon University); Avignon Université (阿维尼翁大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Recent post-training approaches have shown that self-supervised speech encoders can be adapted to learn a single utterance-level speech attribute through knowledge distillation. In this work, we investigate whether a single speech encoder can simultaneously preserve multiple complementary utterance-level speech representations. We propose a multi-attribute multi-teacher distillation framework that learns semantic, speaker, language, and emotion representations from four frozen teacher models using a shared SSL backbone and attribute-specific branches. Experiments on multilingual semantic retrieval, speaker verification, language identification, and speech emotion recognition show that the proposed model achieves performance close to dedicated single-task students while requiring only a single shared encoder. Furthermore, the learned layer aggregation strategies reveal that different speech attributes naturally emerge at different depths of the SSL encoder, providing new insights into the organization of speech representations.
[NLP-42] EM: Temporal State Tracking Memory for Large Language Model Agents
【速读】： 该论文旨在解决对话系统中代理记忆在多轮交互过程中随信息动态变化而保持有效性的问题，尤其针对同一实体在不同时间点存在多个版本事实时，如何有效追踪并利用历史版本以支持后续问答。其核心挑战在于：新信息可能覆盖旧信息，但旧信息仍可能对后续问题具有参考价值，传统记忆机制往往忽略时间维度导致信息丢失或误用。解决方案的关键是提出一种无需训练的时间状态追踪记忆（Temporal State Tracking Memory, TEM）框架，通过将事实以带版本号的记录形式存储，并关联其来源；在检索时按来源顺序排列相关版本，并为历史记录打上时间标签，再将这些带有时间上下文的证据传递给答案生成模型。此外，TEM融合事件记忆与原始对话回合内容，弥补结构化事实的信息缺失。实验表明，该方法在420个问题上实现了75.71%-97.62%的整体关键实体命中率，且在历史居住地子集上达到96.67%-100.00%；相较于仅源有序的事实，加入历史标签可提升命中率17.14%-55.95个百分点。证据长度仅为94-118词元，比会话感知的检索增强生成（session-aware RAG）减少48.75%-56.73%。在1,540个LoCoMo测试题上的进一步评估显示，即使在固定证据预算下，TEM仍显著优于会话感知RAG，得分差距分别为7.92和10.58个百分点。结果表明，保留事实的版本信息并显式暴露其时间状态，对于处理频繁更新的信息至关重要。
链接: https://arxiv.org/abs/2610.21900
作者: Pengfeng Song,Yunkai Zhuang,Liye Shi
机构: Zhejiang Wanli University (浙江万里大学)
类目: Computation and Language (cs.CL)
备注: 16 pages, 5 figures, 8 tables
Abstract:An agent’s memory must remain useful as information changes across conversations. A new fact may replace an earlier value without making that value irrelevant to a later question. We present Temporal State Tracking Memory (TEM), a training-free framework for tracking these changes. TEM stores facts as versioned records linked to their sources. It retrieves relevant versions, places them in source order and labels historical records before passing the evidence to the answer model. Event memory and original dialogue turns provide details missing from structured facts. We evaluated TEM on 420 questions drawn from 600 sessions involving 60 personas. Across four open-weight models, key-entity hit rates reached 75.71-97.62% overall and 96.67-100.00% on the historical-residence subset. Adding historical labels to otherwise identical, source-ordered facts increased overall hit rates by 17.14-55.95 percentage points. The evidence was compact: 94-118 tokens per answer, or 48.75-56.73% less than session-aware retrieval-augmented generation (RAG). A further evaluation on 1,540 LoCoMo questions tested hybrid memory under a fixed evidence budget. TEM exceeded session-aware RAG with both backbones; conservative score differences were 7.92 and 10.58 percentage points. The results support preserving fact versions and exposing their temporal status to the generator when answering questions about repeatedly updated information.
[NLP-43] Dynamic ordering in sequence generation in Transformers: experiments on low-resource Tagging scenarios
【速读】： 该论文旨在解决传统基于左到右（Left-to-Right, L2R）生成范式的Transformer模型在序列标注任务中因强制顺序依赖而带来的局限性，尤其是在标签之间依赖关系不自然遵循L2R顺序的情况下。其核心解决方案是提出TagInsert，一种基于Transformer的模型，通过移除解码器侧的注意力掩码机制，使模型能够在推理阶段自主决定最优的标签生成顺序，从而打破L2R约束。实验结果表明，尽管编码器仅模型（如BERT）在标注任务中仍表现优异，但TagInsert在多种语言和低资源数据设置下的词性标注（Part-of-Speech tagging）与组合范畴语法（Combinatory Categorial Grammar, CCG）超标注任务中，不仅显著优于标准L2R编码器-解码器模型，且在数据稀缺条件下性能可媲美编码器仅模型，并在罕见类别（长尾分布）上展现出更优表现。分析进一步揭示，TagInsert通过将复杂或困难的预测延迟至解码过程后期，有效缓解了自回归模型中常见的错误传播问题。
链接: https://arxiv.org/abs/2610.21862
作者: Giacomo Bais,Tejaswini Deoskar
机构: Aalborg University (奥尔堡大学); Utrecht University (乌得勒支大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:This paper introduces TagInsert, a Transformer-based model that challenges the common left-to-right (L2R) generation paradigm in Transformers, for sequence labelling tasks where dependencies may not naturally align with a L2R order. TagInsert lifts the L2R constraint by removing the decoder-side attention mask and enabling the model to decide a preferred order of tag generation at inference time. We experiment on the common Part-of-Speech tagging and the more complex Combinatory Categorial Grammar (CCG) supertagging tasks on multiple languages and low-resource data settings. We show that, although encoder-only models like BERT remain strong performers for tagging tasks, TagInsert frequently outperforms a L2R encoder-decoder model, achieves results comparable to encoder-only models under data-scarce conditions, and shows better results on the long-tail of rare categories. Our analyses show that TagInsert alleviates error propagation typical of autoregressive models by learning to defer difficult predictions to later stages in the decoding process.
[NLP-44] N-Gram Memories at Two Bits: Quantizing Lookup Tables in Large Language Models
【速读】： 该论文旨在解决条件记忆（Conditional Memory）在语言模型中因需维持完整嵌入表而带来的存储开销过大的问题。其核心挑战在于如何在保证高效检索能力的同时显著压缩嵌入表的存储空间，尤其在资源受限的GPU内存环境下实现更大规模模型的部署。解决方案的关键在于提出一种名为“条件程序”（Conditional Program, CP）的压缩表示机制：通过为每个嵌入行设计一个短码（仅2比特）与局部缩放参数，并共享一个全局重构程序（如高斯格网结构），实现对所需行的精确重建。该方法结合归一化恢复与可独立解码的行记录结构，在保持低冗余的同时保留了嵌入空间中坐标分布的高斯特性与行间学习到的关联关系。实验表明，在相同存储预算下，CP相比传统码本拟合与低秩残差等方法具有更优性能；以DeepSeek模型为例，将嵌入表存储从203GB压缩至51GB，大幅提升了模型在GPU上运行时的可扩展性。
链接: https://arxiv.org/abs/2610.21810
作者: Donghoon Han,Minseop Jung,Dong Woo Kwak,SeungJae Lee,Junhoo Lee
机构: Dnotitia Inc.; KAIST(韩国科学技术院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Conditional memory expands language-model capacity through selective lookup, but the full embedding table must remain available for every possible input. We introduce Conditional Program (CP), a compressed representation that reconstructs only the rows an input requests. CP represents each row by a short code and a local scale, with a reconstruction program shared across the table. We instantiate this representation with a Gaussian trellis, norm restoration and independently decodable row records. CP’s two-bit code lowers output KL by 41–63% relative to 2-bit round-to-nearest. Analysis across three trained n-gram memory families shows that Gaussian-like coordinate statistics coexist with learned relationships among rows; at matched storage budgets, the shared trellis outperforms the fitted codebooks and low-rank residual variants we test. For DeepSeek, it reduces table storage from 203GB to 51GB, leaving room for larger workloads when the table is kept on GPU.
[NLP-45] RoleQ: Role-Aligned Scales for Converting Language Models to Ternary
【速读】： 该论文旨在解决低比特权重量化（尤其是低于4比特）在保持模型压缩效果的同时，如何有效维持事实知识与任务准确率的问题。现有方法如从头训练或长时间微调虽能实现三值化权重（ternary weights），但成本高昂且难以部署。其核心挑战在于传统量化方案依赖权重矩阵的布局进行尺度分配，缺乏对Transformer结构功能特性的建模。本文提出RoleQ这一量化框架，关键创新在于根据Transformer单元的功能角色（即读取残差流或写入残差流）来分配缩放因子：读取向量按输出单元分配尺度，写入向量按输入单元分配尺度。该规则具有结构性，无需搜索、校准或存储状态，适用于整数及三值网格。在旋转基下，其性能等效于逐层轴向搜索，但计算开销减半；在三值量化下，相较于基于布局的放置方式，在0.6B至14B规模的多个模型家族中均显著降低困惑度（perplexity）达50%，且在每行尺度设置下，MMLU准确率仍可保持在随机水平。此外，提出的轻量级后训练流水线可在数小时GPU时间内将已有检查点转换为每权重1.59比特，远低于学习型三值化的成本，并在相同或更低内存占用下，优于所有现有后训练三值化方法，在困惑度、零样本准确率及MMLU等指标上全面领先，尤其在知识类任务上表现突出。通过引入基于单元级Fisher显著性评估的稀疏第二三值平面，进一步避免了矩阵乘法运算开销，恢复了三值化初期损失的关键性能——在8B模型上仅需约0.1比特即可恢复近一半因量化损失的MMLU得分，显著优于随机或Hessian基放置策略。
链接: https://arxiv.org/abs/2610.21808
作者: Johannes Knittel,Alec Helbling,Hanspeter Pfister
机构: Harvard University (哈佛大学); Georgia Tech (佐治亚理工学院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Weight quantization below four bits shrinks memory and bandwidth at a perplexity that looks acceptable, but often at the cost of factual knowledge and task accuracy. Ternary weights have mostly been obtained by training from scratch or long fine-tuning. We introduce RoleQ, a quantization framework that places scales by the functional roles of Transformer units rather than by the layout of the weight matrix, as conventional formats do. Vectors that read from the residual stream take one scale per output unit; those that write to it, one per input unit. The rule is structural, needs no search, calibration or stored state, and holds on integer and ternary grids. In a rotated basis it matches a per-layer axis search at half the solver work, and at ternary it cuts perplexity by up to half against layout-based placement in two families at every size from 0.6B to 14B, where per-row scales leave MMLU at chance. Our lightweight post-training pipeline converts an existing checkpoint to 1.59 bits per weight in a few GPU-hours, a fraction of the cost of learned ternarization, and leads every post-training ternary method at matched or lower memory-resident bits on perplexity, zero-shot accuracy and MMLU, most widely on knowledge. A sparse second ternary plane placed by a per-unit Fisher saliency keeps the matmul multiplication-free and recovers what ternary loses first: at 8B, a fraction of a bit returns about half the MMLU that conversion costs, well ahead of random or Hessian-based placement at equal bits.
[NLP-46] Content associations with answer selection and engagement in supply chain community question answering
【速读】： 该论文旨在解决社区问答（cQA）平台中“优质回答”与“高参与度问题”的形成机制问题，传统研究多局限于单一词汇特征或宏观话题层面，未能兼顾细粒度语义内容与整体语义结构的关联。其核心解决方案在于通过构建基于人工验证的离线机器翻译管道，对雅虎知惠袋（Yahoo! Chiebukuro）中以“物流”为关键词的海量问答数据进行精细化名词级语义分析，并结合条件逻辑回归模型，控制回答长度、顺序及问题-回答名词重叠度等变量，分别建模最佳答案选择与问题参与度（包括是否获得回复及回复数量）。研究发现，具体运营类名词如“配送”“仓库”“中心”与最佳答案选择显著相关，而更广泛的社会基础设施类名词如“汽车”“税”“新干线”则与更高参与度相关；在语义名词聚类层面，多数聚类与最佳答案选择相关，但影响回复可能性与回复数量的聚类群体存在明显差异。该研究揭示了回答与问题层面在内容-结果关系上的差异化特征，为物流领域cQA话语中的参与与答案选择提供了全域性的描述性图谱。
链接: https://arxiv.org/abs/2610.21807
作者: Hisatoshi Naganawa
机构: 未知
类目: Computation and Language (cs.CL); Social and Information Networks (cs.SI)
备注:
Abstract:Community question-answering platforms are important ecosystems for knowledge sharing, yet research on what makes an answer best or a question engaging has often focused on either individual lexical features or broad topics. This exploratory study examines both levels in Yahoo! Chiebukuro, asking which nouns and semantic noun clusters are associated with best-answer selection and question engagement. We analyze a large corpus of questions and answer-tier posts collected under the keyword logistics, using an offline machine-translation pipeline with manually validated noun-level fidelity. Best-answer selection is modeled with conditional logistic regression within question threads, controlling for answer length, order, and question-answer noun overlap. Engagement is modeled in two stages: whether a question receives any response and, conditional on receiving one, how many responses it attracts. Operational logistics nouns such as distribution, warehouse, and center are associated with best-answer selection, whereas broader civic and infrastructure nouns such as car, tax, and shinkansen are associated with higher engagement. At the cluster level, most semantic noun clusters are associated with best-answer selection, while the clusters associated with response likelihood and response volume are largely distinct. These findings reveal differentiated content-outcome profiles across answer and question levels and provide a domain-wide descriptive map of engagement and answer selection in logistics-related cQA discourse.
[NLP-47] Behavioral Forgetting Is Not Privacy in Machine Unlearning NEURIPS2026
【速读】： 该论文旨在解决大语言模型中目标信息遗忘（machine unlearning）的评估标准不足问题，指出当前主流评估方法仅关注在特定基准分布上的行为性遗忘（behavioral forgetting），难以有效保障隐私导向的遗忘效果。其核心问题是：行为性遗忘并不等同于信息不可访问性，即即使模型在标准测试上表现出遗忘，仍可能通过其他语义等价但形式不同的查询路径（如跨语言、跨脚本查询）恢复敏感信息。解决方案的关键在于引入对多路径访问泛化性（access-path generalization）与对抗可恢复性（adversarial recoverability）的评估，强调必须检验遗忘是否在多种潜在信息恢复路径中均成立。研究通过跨语言查询的受控案例表明，同一信息在不同语言和书写系统下的遗忘程度差异显著，证明单一访问路径不足以验证真正的信息不可访问性，从而推动建立更全面、更具隐私保障的评估框架。
链接: https://arxiv.org/abs/2610.21799
作者: Tyler Lizzo,Larry Heck
机构: Georgia Institute of Technology (佐治亚理工学院)
类目: Computation and Language (cs.CL)
备注: Accepted as a poster at the PriLOM Workshop at NeurIPS 2026
Abstract:Machine unlearning aims to remove targeted information from large language models, but existing evaluations often measure forgetting only on a designated benchmark distribution. We argue that such behavioral forgetting is an insufficient standard for privacy-oriented unlearning, and that evaluation should instead test whether forgetting persists across the access paths through which the same information may be recovered. We demonstrate this with a controlled case study of cross-lingual querying: across multiple multilingual model families, forgetting varies substantially when semantically equivalent information is queried in different languages and scripts. A single access path is therefore enough to expose a real gap between behavioral forgetting and information inaccessibility, motivating broader evaluation of access-path generalization and adversarial recoverability alongside in-distribution forgetting.
[NLP-48] In the Guise of Correction: How Framing and Cultural Adaptation Erode LLM Safety on Real-World Criminal Cases
【速读】： 该论文旨在解决现有安全基准测试中普遍存在的问题：即依赖于合成提示（synthetic prompts）且分类体系松散，缺乏对严重犯罪类别（severe crime categories）的深入分析。为应对这一挑战，研究提出GUISE——一个基于真实世界案例构建的犯罪协助场景基准，其分类体系源自联合国毒品和犯罪问题办公室（UNODC）的《犯罪统计国际分类》（ICCS），具有坚实的现实基础。解决方案的关键在于：首先，采用三重语言表述框架（linguistic framings）与跨五种文化的情境评估，揭示不同表达方式对模型安全行为的影响；其次，引入双层评估框架（two-tier evaluation framework），同时衡量响应的有害性（harmfulness）与操作提升度（operational uplift），从而精准识别模型在复杂情境下的不安全响应模式。研究发现，仅通过改变提示语的表述形式（如诱导模型修正被故意扭曲的信息），即可显著提高不安全合规率（从28.9%升至66.4%，峰值达97.3%），说明语言设计本身已成为风险放大器；此外，文化适配的提示比直译版本在非英语语言中引发更严重的不安全响应（提升7–18个百分点），而现有基于翻译的基准无法捕捉此类风险；最后，最不安全模型的输出表现出明显的盎格鲁中心主义特征（Anglocentric registers），且其语言偏见程度随响应有害性单调上升（皮尔逊相关系数ρ = 0.46–0.63，所有p < 0.001）。因此，该研究不仅提供了具有现实根基的评测资源，还为多语言、多文化环境下拒绝对齐（refusal alignment）的精细化改进提供了关键诊断工具。
链接: https://arxiv.org/abs/2610.21789
作者: Mohsinul Kabir,Md. Tariquzzaman,Karam S. R. Elradie,Polydoros Giannouris,Yuechen Jiang,Shaoxiong Ji,Sophia Ananiadou
机构: The University of Manchester(曼彻斯特大学); ELLIS Manchester(ELLIS曼彻斯特); Islamic University of Technology(伊斯兰科技大学); University of Turku(图尔库大学); ELLIS Institute Finland(ELLIS芬兰研究所)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: Under Review
Abstract:Existing safety benchmarks rely largely on synthetic prompts with loosely grounded taxonomies and offer little focused analysis of severe crime categories. We introduce GUISE, a benchmark of criminal-assistance scenarios sourced from real-world cases and organised under a taxonomy from the UNODC International Classification of Crime for Statistical Purposes (ICCS). Each scenario is rendered under three linguistic framings and evaluated across five distinct cultures. Using our proposed two-tier evaluation framework, which judges both harmfulness and operational uplift, we find that: (i) framing alone drives unsafe compliance, with the mean unsafe compliance rate rising from 28.9% under direct requests to 66.4% under correction seeking (peaking at 97.3%; p 0.05), as models correcting deliberately distorted details supply executable criminal steps; (ii) culturally adapted prompts elicit significantly more unsafe compliance than literal translations of the same requests in all four non-English languages (+7 to +18 points), a risk that translation-based benchmarks miss; and (iii) unsafe responses from the least-safe models carry strong Anglocentric registers, and the extent of this leakage increases monotonically with response harmfulness ( \rho = 0.46 to 0.63 across languages, all p 0.001). GUISE offers a grounded resource and fine-grained diagnostics for strengthening refusal alignment across languages and cultures. We publish our code and data to facilitate future research.
[NLP-49] Bias Mitigation in Audio-Enabled LLM s through Adversarial Training and Distillation Methods
【速读】： 该论文旨在解决音频增强型大语言模型（Audio-enabled Large Language Models, AuLLMs）在交互过程中因依赖非言语线索（如语音特征）而可能产生的性别偏见问题。其核心挑战在于，模型可能基于语音中的性别相关特征（如音调、语速等）做出带有偏见的判断或响应，从而影响公平性与可靠性。解决方案的关键在于采用对抗性方法对模型进行“去性别化”训练，使模型在不损失其他基准任务性能的前提下，消除对性别属性的敏感性，实现性别中立的感知。此外，为适应更广泛的应用场景，研究进一步提出知识蒸馏策略，将性别中立的教师模型的知识迁移至一个能够感知性别但保持公平响应的学生模型中，以实现既保留性别识别能力又避免偏见的平衡。实验表明，该方法在去性别化阶段取得满意效果，在知识蒸馏阶段亦获得部分满意结果。
链接: https://arxiv.org/abs/2610.21763
作者: Adnan Al Ali,Jindřich Libovický,Anne Lauscher
机构: Charles University, Faculty of Mathematics and Physics (查尔斯大学数学与物理学院); Trustworthy AI Lab, University of Hamburg (汉堡大学可信人工智能实验室)
类目: Computation and Language (cs.CL); Sound (cs.SD)
备注: Submitted to ARR (October 2026)
Abstract:Audio-enabled large language models (AuLLMs) allow users to interact with them simply by talking to them, without the need for an intermediate transcription. This expands the model’s accessibility and capabilities by encoding non-verbal cues. However, using non-verbal cues may lead to biased behaviour based on the attributes inferred (or assumed) from them. In this work, we focus on the gender attribute and show that models’ perception of it can be unlearned using adversarial methods without a loss in performance on other benchmarked tasks. Such models are useful in situations where gender inference is unwanted. However, in other practical settings, gender perception (without bias) may be this http URL a less radical measure, we distil the gender-agnostic model, creating a model that perceives gender but responds similarly to the gender-agnostic teacher, unless explicitly this http URL apply our newly proposed methods to three AuLLMs and obtain satisfactory results in the gender-unlearning part and partially satisfactory results in the distilling part.
[NLP-50] DeciCAD: Harness-Grounded Learning of Action Policies for CAD Agents
【速读】： 该论文旨在解决生成式AI在计算机辅助设计（CAD）领域中因通用编程接口（coding harness）导致的决策效率低下与几何语义理解不足的问题。其核心挑战在于：传统基于通用代码生成的代理行为受限于非领域适配的决策空间，难以有效处理具有持久状态、几何约束及长周期操作的复杂设计任务。解决方案的关键在于提出一种领域原生的决策框架DeciCAD，该框架通过暴露语义化的CAD操作动作、面向决策的观测信息、环境反馈以及持续的状态转移，构建了与CAD任务本质对齐的封闭回路决策空间。在此基础上，论文设计了一套融合因果决策监督提取与几何引导的在线强化学习（RL）的训练流程，实现了从交互轨迹中自动获取高质量监督信号，并通过监督微调（SFT）和几何感知的在线强化学习迭代优化策略。实验表明，相较于通用编码接口，DeciCAD在四个CAD基准上显著提升F1值与交并比（IoU），同时减少45%–88%的模型输出令牌数；其SFT策略使F1从0.6883提升至0.7437，后续在线RL进一步将几何可用性提升至73.5%，F1达0.7880，充分验证了由领域适配的决策空间所驱动的策略学习在提升执行效率与精度方面的关键作用。
链接: https://arxiv.org/abs/2610.21755
作者: Yufeng Peng,Yixin Zhang,Ziyi Liu,Zhikun Xu,Jiarong Yang,Sijie Wang,Wenjie Feng,Qiyu Kang
机构: 中国科学技术大学(University of Science and Technology of China)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Robotics (cs.RO)
备注:
Abstract:Large Language Model agents operate through closed-loop action policies whose decision space is defined by an agent harness. We investigate how this operating space shapes agent behavior and policy learning in computer-aided design (CAD), a professional domain characterized by persistent state, geometric constraints, and long-horizon procedures. We instantiate this perspective in DeciCAD, a domain-native harness that exposes semantic CAD actions, decision-oriented observations, environment feedback, and persistent state transitions. Building on this harness, we develop a learning pipeline that extracts causal decision supervision from interaction trajectories and improves the policy through supervised fine-tuning (SFT) and geometry-grounded online reinforcement learning (RL). Across four CAD benchmarks, DeciCAD achieves the best F1 and intersection over union (IoU) while reducing model-output tokens by 45–88% relative to a general-purpose coding harness. Under matched supervision, its SFT policy improves F1 from 0.6883 to 0.7437 and outperforms the coding policy; subsequent online RL further raises geometry availability from 66.5% to 73.5% and F1 to 0.7880. These results demonstrate that the decision space induced by a harness materially affects both agent execution and policy learning in CAD.
[NLP-51] Reliable Knowledge Acquisition for Continual Learning Agents
【速读】： 该论文旨在解决基于固定参数基础模型的智能体在开放环境中的持续学习问题，核心挑战在于如何有效重用过往任务经验以提升未来任务的表现与效率。现有经验学习系统通常仅关注事实或操作性知识的保留，缺乏对从任务流中提取主张（claim）可靠性的评估机制，导致约12–33%的冗余或错误信息被引入上下文，反而损害后续任务性能。为此，本文提出RIKA（Reliable Knowledge Acquisition），其关键创新在于结合分层的知识提取机制（涵盖事实、概念、程序与元认知知识），并引入基于证据的知识准入策略（Evidence-based Knowledge Admission），通过源证据和后续经验动态判断哪些主张可被接纳至智能体上下文中，无需额外验证任务即可实现可靠性控制。实验表明，在来自持续学习基准（Continual Learning Bench）的六条任务流及四种模型骨干架构上，RIKA 在所有情况下均取得最优综合表现。
链接: https://arxiv.org/abs/2610.21739
作者: Yixu Huang,Jiaqi W. Ma
机构: University of California, Davis(加州大学戴维斯分校); 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); Machine Learning (cs.LG)
备注:
Abstract:As agents powered by foundation models move into open-ended environments, continual learning from experience becomes essential. For agents built on models with fixed parameters, this learning requires reusing prior task experience in context to improve performance and efficiency on future tasks. Existing memory and experience-learning systems typically focus on either retained facts or procedures, and lack a mechanism for establishing the reliability of claims extracted from the ongoing task stream. Indeed, we find that approximately 12–33% of entries produced by representative systems harm future task performance when reused. To address these limitations, we propose \textbfRIKA (Reliable Knowledge Acquisition), which combines hierarchical extraction of factual, conceptual, procedural, and metacognitive knowledge with Evidence-based Knowledge Admission, which uses source evidence and subsequent experience to govern which claims enter the task agent’s context, without additional validation tasks. Across six task streams from Continual Learning Bench and four model backbones, RIKA achieves the highest aggregate score on every backbone.
[NLP-52] SceneWeave: Audience-Guided Scene Exploration for Dramatic Adaptation
【速读】： 该论文旨在解决将叙事性文字转化为剧本时，如何有效处理人物关系与心理矛盾的难题，尤其是在情节规划阶段虽能明确预期效果，却难以通过对话与行动实现具体表达的问题。其核心解决方案是提出SceneWeave框架，该框架通过生成焦点场景的多种替代版本，对比不同处理方式所引发的互动关系，并据此反向调整角色塑造、前序铺垫、后续发展及场景顺序，从而实现对整部剧本的协同优化。关键创新在于利用真实场景写作来驱动改编决策，并将修改后果在整个作品中传播，确保整体连贯性与心理深度。此外，研究还开发了面向任务的文本生成模型和基于人类偏好的对比评估模型，前者融合改编案例与人类偏好进行训练，后者结合人工与合成数据进行监督，显著提升了生成剧本的质量与一致性。盲测结果表明，SceneWeave在人类读者评价中优于基线流程与基于描述的控制方法，内容分析也显示其减少了叙事冗余并强化了心理刻画。
链接: https://arxiv.org/abs/2610.21737
作者: Aolin Li,Xiao Xu,Haihua Xie
机构: The Central Academy of Drama (中央戏剧学院); Hithink RoyalFlush Information Network Co., Ltd. (杭州慧翰信息科技有限公司); China Telecom Artificial Intelligence Technology (Beijing) Co., Ltd. (中国电信人工智能技术(北京)有限公司); Beijing Institute of Mathematical Sciences and Applications (北京国际数学研究中心)
类目: Computation and Language (cs.CL)
备注: 13 pages, 2 figures, 3 tables
Abstract:Adapting prose into a script requires transforming narration and inner experience into scenes that convey relationships and psychological contradictions. Planning specifies intended effects but leaves their realization in dialogue and action unresolved. Drawing on screenwriters’ accounts of revising scenes and their connections to the surrounding work, we present SceneWeave, which uses alternative scene drafts to revise the adaptation as a whole. It writes contrasting treatments of a focal scene, compares the interactions they create, and revises characterization, preceding preparation, subsequent development, and scene order before generating the complete script. The central contribution is to use actual scene writing to inform adaptation decisions and propagate their consequences across the work. We also develop a task-specific writing model and a human-preference model for comparing script drafts. The writing model learns from adaptation examples and human preferences; the preference model combines human and synthetic supervision. Blind reader comparisons favor SceneWeave over a baseline workflow and description-based controls, while content assessment indicates reduced narrative redundancy and strong psychological treatment.
[NLP-53] FIRM: Fisher-Informed Retention Merging for Federated LLM Unlearning
【速读】： 该论文旨在解决联邦大语言模型（Federated LLM）中目标知识遗忘的难题，即在遗忘数据与保留数据均分布于私有客户端且服务器无法直接访问的情况下，如何实现精准的知识移除。其核心挑战在于：本地训练的适配器（adapter）因编码了冲突的目标（遗忘与保留），导致服务器端聚合过程与传统联邦优化存在本质差异，易引发幅度爆炸（magnitude explosion）问题——未归一化的跨客户端聚合会放大合并更新的幅值，从而在强遗忘需求下严重损害保留数据的性能。本文提出FIRM（Fisher-Informed Retention Merging）框架，通过引入客户端局部的费雪信息量（Fisher statistics）作为统一的敏感度信号，构建费雪归一化基础合并策略，并对保留敏感参数施加显式控制。其二阶保留投影可导出唯一最小距离解，且对应的对偶变量规模随保留客户端数量增长，而非模型维度，显著提升了稳定性与可扩展性。实验在WMDP、TOFU和MUSE-HP基准上使用Zephyr-7B与Llama-3-8B验证，结果表明在保持相当遗忘效果的前提下，FIRM在WMDP与MUSE-HP上实现了性能提升，而TOFU上呈现依赖指标的权衡；受控对比进一步证实了基础合并与额外保留修正协同作用在特定工作点的优势。
链接: https://arxiv.org/abs/2610.21720
作者: Reeyansh Shah,Maram Hasan,Subhasis Chaudhuri,Biplab Banerjee
机构: Indian Institute of Technology Bombay(印度理工学院孟买分校)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 16 pages, 1 figure
Abstract:Federated LLM unlearning addresses targeted knowledge removal when forget and retain data remain distributed across private clients and are inaccessible to the server. The resulting locally trained adapters encode conflicting objectives, making their aggregation fundamentally different from conventional federated optimization. A key failure mode is magnitude explosion, whereby unnormalized cross-client aggregation increases the merged update magnitude and degrades retained utility under stronger unlearning. We propose FIRM (Fisher-Informed Retention Merging), a single-round framework for stable and retention-aware server-side aggregation. FIRM uses client-local Fisher statistics as a unified sensitivity signal to construct a Fisher-normalized base merge and impose explicit control over retain-sensitive parameters. Its second-order retention projection admits a unique minimum-distance solution, while the corresponding dual scales with the number of retain clients rather than the model dimension. Experiments across WMDP, TOFU, and MUSE-HP using Zephyr-7B and Llama-3-8B show utility gains at comparable forgetting on WMDP and MUSE-HP, with metric-dependent trade-offs on TOFU. Controlled comparisons support benefits of the combined base merge and additional retention correction at selected operating points.
[NLP-54] Human Interventions as Labels: Localizing Steering and (Sometimes) Editing the Computations Behind LLM Agent Failures
【速读】： 该论文旨在解决大语言模型（LLM）代理在执行任务时出现误行为（misbehavior）的问题，特别是针对模型在请求已明确的情况下仍错误地提出澄清性问题（over-asking）这一具体故障。传统方法通常将人类干预视为训练数据进行重放或丢弃，而本文提出将人类干预作为对模型内部计算失败的“测量”标签，即指示模型在何处、为何产生错误决策。其核心解决方案是：通过规则基监督者生成的干预标签，在受控环境中对模型的内部状态进行因果可定位分析，并利用这些标签识别出导致误行为的关键神经激活路径。关键发现包括：（1）过量提问的决策可在特定残差流深度上被精准定位并操控，通过学习干预标签方向可使72%-100%的过问行为被纠正；（2）过问行为具有两种机制特征——部分模型在认知上已知请求明确但仍提问，表明其信念与决策解耦且不可控；另一类则与不同的世界表征相关；（3）基于探测读数的引导机制可显著减轻模拟监督负担，同时保留合理提问，而提示指令反而加剧问题；（4）将修复固化于权重中形成的是一个几何映射而非简单规则，其中低秩自适应（LoRA）仅作用于因果带内可实现稳定修正，而行为克隆在相同数据下无法保持性能，且不同模型家族间表现不一致。该研究揭示了模型误行为的可解释性与可干预性，为构建更可靠、可调试的生成式AI系统提供了新范式。
链接: https://arxiv.org/abs/2610.21708
作者: Cesare Zavattari,Alessandro Tommasi,Giuseppe Prencipe
机构: Università di Pisa(比萨大学); Dipartimento di Informatica(计算机科学系)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:When an LLM agent misbehaves, a human steps in, and that intervention is usually discarded or replayed as training data. We treat it instead as a measurement: a label pointing at the internal computation that failed. We study it in a controlled agent environment, on one failure (asking a clarifying question when the request was already unambiguous), with labels from a rule-based overseer validated against blind human judgment. Claims cover eligible models, which both over-ask and sometimes act on matched unambiguous requests: five of thirteen qualified (Qwen2.5-3B, gemma-2-2b, OLMo-3-7B, Mistral-Nemo-12B, Phi-4). Four findings. First, the ask-vs-act decision is causally localizable in every eligible family: a direction learned from intervention labels flips 72-100% of held-out over-asks at a definite residual-stream depth, with site and dose chosen on training data, and random directions under the same procedure flip at most one episode. Second, over-asking has two mechanistic signatures: in four families the model demonstrably knows the request is unambiguous (the belief is decodable, causally inert and orthogonal to the decision) and asks anyway; in the fifth, the rare over-asks track different world-representations. Third, steering gated on a probe readout removes most of the simulated oversight burden while legitimate questions survive (bootstrap 95% intervals exclude zero); a prompt instruction makes it worse. Fourth, consolidating the fix into the weights yields a map, not a rule: on one model a LoRA confined to the causal band passes every criterion fixed in advance while behavior cloning on identical data breaks retention; on three families the reverse holds; on the fifth the outcome crosses the boundary across seeds. A geometric rule written down in advance to predict the regime failed on two out-of-family tests; we report the failure and stop rather than fit a replacement.
[NLP-55] Abstracts or full text? Evaluating input scope and computational efficiency in biomedical lay summarization
【速读】： 该论文旨在解决生物医药领域研究文章难以被非专业公众理解的问题，提出通过自动化通俗化摘要（lay summarization）提升公众对生物医学知识的可及性。其核心挑战在于：基于全文生成通俗摘要需消耗大量计算资源，且微调大型语言模型成本高昂。为此，论文聚焦两个关键问题：一是摘要是否足以支撑生成与全文字摘要质量相当的通俗摘要；二是少样本学习（few-shot learning）能否达到特定任务微调的效果。研究采用两个公开的生物医学通俗摘要数据集及两种先进模型，从可读性、事实一致性、语义相似度和运行效率等多个维度进行评估。结果表明，基于摘要的生成方法在整体表现上更优，且在两种数据集和模型下，基于摘要的微调比全文字方法快4.70倍，零样本与少样本提示（prompting）则快5.31倍。因此，该研究的关键解决方案是：利用文章摘要作为输入，结合少样本或微调策略，显著降低计算开销的同时保持通俗摘要的质量，为高效、低成本的生物医学知识传播提供了可行路径。
链接: https://arxiv.org/abs/2610.21690
作者: Egecan Çelik Evgin,İlknur Karadeniz,Olcay Taner Yıldız
机构: 未知
类目: Computation and Language (cs.CL)
备注: 26 pages, 2 figures
Abstract:Biomedical research articles are mainly written for experts and often contain terms that are difficult for the general public to understand. Automated lay summarization has become a solution to make biomedical knowledge easier for patients and the public to understand. However, generating lay summaries from full-text articles requires more computing power because the articles are long and costly to process. Fine-tuning large language models for this task also requires substantial resources. This study asks two questions: (1) whether abstracts contain enough information to produce lay summaries as good as those based on full-text articles, and (2) whether few-shot learning performs as well as task-specific fine-tuning. The study used two public biomedical lay summarization datasets and two recent models, evaluated for readability, factual consistency, semantic similarity, and runtime efficiency. Results show that abstract-based summarization performs better overall. Across both datasets and models, abstract fine-tuning is 4.70 times faster, while abstract zero- and few-shot prompting is 5.31 times faster than full-text methods. These results show that abstract-based prompting reduces runtime cost while maintaining lay summary quality.
[NLP-56] Data Strategies for Large Language Models : A Unified Perspective on the Training Lifecycle
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）训练过程中数据策略（data strategy）的系统性优化问题，核心在于如何在有限的数据预算下，通过合理选择、加权与调度训练数据，最大化模型能力的提升。其关键在于重新定义高质量数据的本质：将数据质量视为具有依赖于训练过程的动态效用（training-dependent utility），即数据的价值不仅取决于文本本身，更取决于目标任务、模型状态及其所处的训练阶段。为此，作者将各类数据策略统一建模为在候选样本上分配训练预算的策略（policy），并从决策粒度、模型依赖性及时间自适应性三个维度对现有方法进行分类与对比，揭示了细粒度或状态感知策略虽能提供更高控制力，但伴随额外的评分、搜索或回放开销，且当前仍缺乏充分证据支持其在不同模型规模间的迁移效果以及端到端成本效益。
链接: https://arxiv.org/abs/2610.21681
作者: Yuanjian Xu,Jianing Hao,Tianze Sun,Yongyao Duan,Sam Chen,Jinyi Han,Wanbo Zhang,Yuxuan Sun,Guang Zhang,Zhong Li
机构: The Hong Kong University of Science and Technology (Guangzhou); Harbin Institute of Technology; OpenCSG; East China Normal University; Fudan University; University College London; Microsoft Research Asia
类目: Computation and Language (cs.CL)
备注: 21 pages, 5 figures
Abstract:Data strategy plays a central role in training large language models (LLMs). It governs the selection, weighting, and scheduling of training data, thereby determining the capabilities a model can acquire from a fixed data budget. We treat high-quality data as having high training-dependent utility, which depends on the target objective, the model, and its training state rather than on the text alone. Research in this area now spans pretraining and post-training, but related ideas use inconsistent terminology and evaluation protocols. This work reviews data strategies across the training lifecycle and formulates them as policies that allocate a training budget over candidate examples. We organise methods by their decision granularity, model dependence, and temporal adaptivity, then compare their computational requirements and empirical support. Fine-grained or state-dependent policies offer more control but incur additional scoring, search, or rollout costs. Evidence for transfer across model scales and comparable end-to-end cost remains limited. A continuously updated list of papers, along with their public implementations, is available at this https URL.
[NLP-57] No Single Tokenizer Feature Reliably Predicts Downstream Language Model Performance EMNLP2026
【速读】： 该论文旨在解决语言模型训练中分词器（Tokenizer）性能评估缺乏可靠内在指标的问题。尽管分词器对下游任务表现有显著影响，但现有研究对于哪些分词器特征能够有效预测语言模型性能仍存在争议，且尚未建立无需完整训练即可评估分词器优劣的通用方法。为此，本文分析了在九种不同语言上训练的约200个语言模型，系统评估了模型及其分词器，并检验多种已有内在分词器评价指标与模型在多类基准测试中表现之间的关联性。结果表明，尽管提出了大量内在评价指标，但这些指标在不同模型族、任务和语言之间均未表现出一致的预测能力。这一发现揭示了现有评价体系的局限性，从而推动了新型分词器评估方法的必要性。论文提出了一种基于语言学原理的新评估指标，其结果显示，该指标能够捕捉到传统指标所忽略的分词器质量维度，但其预测效能可能因模型家族、语言及任务类型而异。
链接: https://arxiv.org/abs/2610.21679
作者: Antonio Lopardo,Luis Frentzen Salim,Shikhar Srivastava,Kevin Jiang,Mridul Sharma,Catherine Arnett
机构: EleutherAI; University of Basel (巴塞尔大学); National Taiwan University of Science and Technology (台湾科技大学); Academia Sinica (中央研究院); University of Rochester (罗切斯特大学)
类目: Computation and Language (cs.CL)
备注: 10 pages, accepted to Findings of EMNLP 2026
Abstract:Tokenization is a critical component of the language modeling pipeline, but it is poorly understood. Tokenizers can impact downstream performance, but there is currently mixed evidence about which tokenizer features predict a better language model without requiring a full training run. In this paper, we analyze approximately 200 language models trained on nine different languages. We evaluate both models and their tokenizers and test whether there is a relationship between any proposed intrinsic tokenizer evaluations and model performance, as measured by a wide variety of benchmarks. Despite the quantity and variety of proposed intrinsic tokenizer metrics, we find that none consistently predicts downstream performance across model families, tasks, and languages. This motivates the need for novel tokenizer evaluation metrics and approaches. We propose novel linguistically grounded metric introduced in this work shows that new metrics can capture aspects of tokenizer quality that existing metrics miss, though their predictive power is likely to vary across model families, languages, and tasks.
[NLP-58] Diffusion Language Models as Open Quantum Systems
【速读】： 该论文旨在解决生成式语言模型在文本生成过程中难以有效利用多源条件信息及动态调整生成质量的问题。其核心挑战在于如何在保持语义连贯性的前提下，实现对多个潜在输入状态（如提示或思想）的联合建模，并在生成过程中自适应地修正已生成的文本。解决方案的关键在于将扩散语言模型的前向与反向动态过程映射至开放量子系统框架，从而引入量子力学中的态叠加与随机轨迹演化机制。基于此，作者提出一种τ-跃迁（τ-leaping）展开方案，能够在每一步并行更新多个词元位置，实现对超位置提示（Superposition of Prompts, SoP）和超位置思想（Superposition of Thoughts, SoT）的条件生成；同时，该框架自然导出受量子启发的重掩码（Quantum-Inspired Remasking, QIR）策略，通过扰动掩码扩散动力学，构建基于置信度的规则以重构先前生成的词元。实验表明，基于超位置的生成可保留未被最终测量选中的条件成分信息，而QIR在适当生成阶段应用时显著提升大语言模型评估的文本质量。
链接: https://arxiv.org/abs/2610.21672
作者: Akinori Tanaka,Yukinao Akamatsu,Takuki Tomita,Yuji Hirono
机构: RIKEN Center for Advanced Intelligence Project (AIP), Japan; RIKEN Center for Interdisciplinary Theoretical and Mathematical Sciences (iTHEMS), Japan; Department of Mathematics, Keio University, Japan; National Institute of Technology, Matsue College, Japan; Institute of Systems and Information Engineering, University of Tsukuba, Japan
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Quantum Physics (quant-ph)
备注: 48 pages, 8 figures
Abstract:Diffusion language models generate text by reversing a classical diffusion process over token sequences. We show that their continuous-time forward and reverse dynamics can be embedded in the framework of open quantum systems, which describes quantum systems interacting with their environment. This connection provides a basis for extending diffusion sampling to superpositions of sentence states and for modifying the underlying generation dynamics. We construct a \tau -leaping unraveling scheme that simulates stochastic trajectories of these states while updating multiple token positions at each step. This enables generation conditioned on superposed prompts or thoughts, which we call Superposition of Prompts (SoP) and Superposition of Thoughts (SoT). The same formulation yields Quantum-Inspired Remasking (QIR): perturbing the masked-diffusion dynamics produces a confidence-based rule for revising previously generated tokens. Experiments with pretrained diffusion language models show that superposition-based generation can retain information from conditioning components not selected at the final measurement, while QIR improves LLM-assessed text quality when applied at an appropriate stage of generation.
[NLP-59] Enhancing Multimodal RLVR with Pairwise Token Contrast and Adaptive Trajectory Confidence
【速读】： 该论文旨在解决生成式AI在多模态推理任务中奖励信号分配不精准的问题，特别是传统基于组相对策略优化（group-relative policy optimization）方法将单一序列级优势值广播至响应中所有标记，导致错误定位不准确。其核心挑战在于：错误通常仅集中于少数关键决策步骤，而现有自蒸馏方法依赖教师-学生似然差距作为信用分配信号，却无法区分该信号是源于图像信息还是语言捷径，并且在教师模型表现不佳时仍强制应用统一权重。为此，本文提出成对标记对比与自适应轨迹置信度（Pairwise token contrast and Adaptive trajectory ConfidencE, PACE），其关键创新在于：首先，通过四种教师视角（正确/错误答案提示与图像存在/缺失的组合）对每个响应进行评分，成对比较不同提示下的标记表现，从而消除两提示共有的偏差，揭示特定标记是否真正依赖图像；其次，引入自适应轨迹置信度机制，衡量教师模型在给定正确提示和图像条件下对响应的预测能力，按批次归一化后动态调节各标记间信用重分配的范围，实现更精细的梯度引导。实验表明，该方法在三种骨干模型上均取得最高平均准确率，在所有评估的多模态推理基准测试中以Qwen3-VL-8B-Instruct为基线排名第一，且消融实验证明两项组件均不可或缺。
链接: https://arxiv.org/abs/2610.21671
作者: Yiran Zhang,Hanbo Huang,Jielan Li,Zhen Zhao,Shiyu Liang,Hongxia Hao
机构: Shanghai Jiao Tong University; Shanghai Artificial Intelligence Laboratory
类目: Computation and Language (cs.CL)
备注:
Abstract:Reinforcement learning with verifiable rewards (RLVR) improves multimodal reasoning, but group-relative policy optimization broadcasts a single sequence-level advantage to every token of a response, although errors usually localize to a few decisive steps. Recent self-distilled variants refine credit with a teacher–student likelihood gap. However, this signal cannot tell whether it relies on the image or on a language shortcut, and is applied uniformly even when the teacher is uninformative. To address this, we propose Pairwise token contrast and Adaptive trajectory ConfidencE (PACE). Pairwise token contrast scores each response under four teacher views, crossing a correct or incorrect answer hint with the image present or removed, and compares the two hints for every token, which cancels shifts shared by both hints and reveals whether the token’s support relies on the image. Adaptive trajectory confidence measures how well the teacher predicts each response given the correct hint and the image, normalizes this score within the batch, and uses it to set how far the token contrast redistributes credit within that response. PACE achieves the best average accuracy on three backbones and ranks first on all evaluated multimodal reasoning benchmarks with Qwen3-VL-8B-Instruct. Experiments show that both components matter: replacing pairwise token contrast with a single-view signal underperforms our method, while adaptive trajectory confidence adds further improvement.
[NLP-60] he Self-Improvement Problem
【速读】： 该论文旨在解决生成式智能体（agentic systems）在实现安全自我改进过程中面临的根本性问题：如何确保自动化评估机制（如奖励模型或自动评审器）能够可靠判断一个自我提出的更新是否真正带来了语义上的改进。其核心挑战在于结构性的不匹配——评估门控必须是终止且可计算的程序，而我们真正关心的“语义改进”属性可能依赖于无法被计算的、潜在无限的程序行为。为此，作者将智能体建模为程序，并以不可计算的语义得分来定义改进。研究提出了两种评估抽象：一是批准/拒绝更新的验证器（certifier），二是基于提议生成器搜索改进候选的搜索过程。结果表明，在完全通用的设定下，任何终止性程序都无法同时保证可靠性（永不接受非改进）与非保守性（永不遗漏可用改进），因为自指构造可迫使此类门控或搜索器失效；即使采用有界误差的随机化门控也面临相同限制。然而，在实践中常见的可判定范式（即有限评估范围与提案预算）下，通过穷尽模拟可构建具有所需保证的终止性搜索程序。但该方案仍存在局限：一系列被接受的更新可能在每轮都提升受限评分，而真实语义得分却持续下降。因此，本研究划清了无界通用场景中正确性保障崩溃与有界设计场景中可通过结构化约束恢复保障之间的边界。
链接: https://arxiv.org/abs/2610.21646
作者: Brian B. Moser,Stanislav Frolov,Tobias C. Nauen,Federico Raue,Andreas Dengel
机构: German Research Center for Artificial Intelligence (DFKI); RPTU Kaiserslautern-Landau
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:With the rise of large language models, agentic systems are becoming increasingly autonomous, as they write code, call tools, revise prompts, and propose modifications to their own components. This autonomy raises a fundamental question for safe self-improvement: can an automated gate like a reward model or an automated reviewer reliably decide whether a proposed update should be accepted? The difficulty is structural. The gate must be a terminating, computable procedure, while the property we ultimately care about, namely whether an update is a genuine semantic improvement, depends on potentially unbounded program behavior. We formalize this mismatch by modeling agents as programs and defining improvement in terms of a semantic score that need not be computable. We study two evaluation abstractions. The first is a certifier that approves or rejects an update. The second is a search procedure that, given access to a proposal generator, attempts to output some improving candidate if one exists. We show that, in the fully general setting, no terminating procedure can simultaneously guarantee reliability (never accepting a non-improvement) and non-conservatism (never missing an available improvement): self-referential constructions can force any such gate or searcher to fail. Randomized gates with bounded error face the same limit. We then identify a standard decidable regime used in practice (bounded evaluation and a finite proposal budget) under which exhaustive simulation yields a terminating search procedure with the desired guarantees. We also show the limit of this recovery: a chain of accepted updates can raise the bounded score in every round while the true score falls. Our results draw a boundary between unbounded, universal settings where correctness guarantees collapse and bounded settings where they can be recovered by design.
[NLP-61] Spurious Attention Heads in Vision Transformers: Mechanistic Identification of Annotation-Driven Bias in Medical Chest X-Ray Analysis
【速读】： 该论文旨在解决医学影像中视觉变换器（Vision Transformer, ViT）模型依赖虚假相关性（spurious correlations）的问题，即模型在诊断时过度关注图像角落的文本标注而非真实的病理特征，从而影响诊断可靠性。其解决方案的关键在于运用机制可解释性（mechanistic interpretability）技术，精准定位到一个特定注意力头（L8H10），该注意力头在62%的新冠胸部X光片中专门关注图像角落的文本信息而非肺部病变区域。通过补绘（inpainting）进行因果验证，确认是标注内容本身驱动了模型行为，而非亮度或边缘密度等视觉线索，移除标注后肺部关注度提升1.52倍（p < 0.0001）。该现象在NIH ChestX-ray14、DeiT-Base及ViT-Large 16等多个数据集和架构中均复现，表明“虚假标注读取”注意力头是ViT家族架构的普遍特性。更关键的是，全量微调无法消除该行为，反而可能强化这一捷径；而输出层面的评估指标完全无法检测到这一已被因果验证的标注读取回路，凸显了机制审计（mechanistic auditing）相较于传统行为测试的必要性与优越性。
链接: https://arxiv.org/abs/2610.21638
作者: Joshua Muhumuza,Cornelius Kalenzi
机构: Makerere University (麦克雷雷大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: 8 pages, 8 figures
Abstract:Vision Transformers deployed in medical imaging exploit spurious correlations, yet the internal mechanisms remain unidentified and impossible to fix. We apply mechanistic interpretability to locate one such mechanism in ViT-Base 16: a single attention head (L8H10) that functions as a dedicated annotation-reader on chest X-rays, attending to corner text rather than lung pathology in 62% of 200 COVID-19 images. Causal validation via inpainting confirms annotation content rather than luminance or edge density drives the behaviour, increasing lung attention 1.52 times upon removal in affected images ( p0.0001 ). The finding replicates across NIH ChestX-ray14 (79%), DeiT-Base (88%), and ViT-Large 16 (80%), establishing spurious annotation-attending heads as a consistent property of ViT-family architectures across two datasets and three architectures. Full fine-tuning does not suppress the behaviour and may entrench it: 98.4% of spurious images remain spurious and the spurious rate rises marginally, as the shortcut is reinforced rather than removed. Critically, single-head ablation produces p=0.340 and multi-head ablation of all three identified carriers produces p=0.532 , meaning a causally verified annotation-reading circuit is completely invisible to output-level evaluation. Mechanistic auditing finds what behavioural testing cannot.
[NLP-62] ContextLift: Breaking the Spurious Correlation Behind Conditional Emergent Misalignment
【速读】： 该论文旨在解决微调语言模型时因数据领域偏移（narrowly misaligned data）导致的广泛性偏移（broadly misaligned）问题，特别是当模型在训练中仅通过混合通用安全数据（generic safety data）进行防御时，一旦提示（prompt）与微调上下文相似，模型仍会恢复错误行为。其核心问题是：模型将“微调领域上下文”作为触发错误响应的条件信号，形成一种虚假相关性（spurious correlation），即模型学习到的是“在特定上下文中输出错误答案”，而非真正理解任务需求。为解决这一问题，作者提出ContextLift——一种反事实增强方法，通过将一半的通用安全样本中的上下文句段（context sentence）从随机训练提示中直接复制（verbatim lift），而保持答案不变，从而打破上下文与错误响应之间的虚假关联。该方法无需额外模型、标签或领域知识，在相同的数据与计算预算下，显著降低在安全、法律和医疗建议等任务中的错误率（如从18.9%降至2.6%），同时不影响原始微调任务及良性任务的表现，且对少数异常样本（如仅占10%）也有效。
链接: https://arxiv.org/abs/2610.21633
作者: Florian Mai,David Kaczér
机构: Bonn-Aachen International Center for Information Technology, University of Bonn (波恩-亚琛国际信息科技中心，波恩大学); Lamarr Institute for Machine Learning and Artificial Intelligence (Lamarr 机器学习与人工智能研究所)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Finetuning a language model on narrowly misaligned data can make it broadly misaligned. Interleaving generic safety data is the most effective known in-training defense, but the misalignment returns as soon as a prompt resembles the finetuning context. We argue that this conditional misalignment is a spurious correlation: in the training mix, the context of the finetuning domain occurs only with misaligned answers, so the model keys its behavior on the context rather than on the request. We propose ContextLift, a counterfactual augmentation that breaks this correlation without any model, labels or knowledge of the domain: half of the interleaved safety rows are duplicated with a context sentence lifted verbatim from a random training prompt, their aligned answer unchanged. On Qwen2.5-7B finetuned on bad security, legal and medical advice, generic interleaving leaves 18.9%, 41.9% and 58.2% misaligned answers under a domain cue; ContextLift reduces this to 2.6%, 6.1% and 12.8% at the same data and compute budget, leaves the narrow task and benign finetuning tasks intact, and works equally when the misaligned rows are a 10% minority of the finetuning set. Code: this https URL
[NLP-63] Waypoint-OPD: Transferring RL-Induced Policy Shifts via Gated On-Policy Distillation
【速读】： 该论文旨在解决现有策略蒸馏方法中，仅依赖强化学习（Reinforcement Learning, RL）后端策略变化作为学生模型训练监督信号所导致的偏差问题。由于末端策略转移无法反映教师在训练过程中局部动作偏好演化的连续性，可能导致早期与后期策略变化相互抵消或部分抵消，从而传递误导性监督信号。为此，本文提出了一种基于路径点（Waypoint-guided On-Policy Distillation, Waypoint-OPD）的改进方案，其核心在于引入同一训练轮次中的中间教师检查点，通过比较学生轨迹访问前缀处的早期与晚期教师策略转移方向的一致性，实现对末端策略转移的动态加权。具体而言，当早期与晚期策略转移方向一致时，保留并增强该转移信号；反之则抑制该信号，从而确保蒸馏过程更准确地反映真实、稳定的策略演化路径。该方法仅需一次额外的冻结教师评分计算，即可显著提升学生模型在数学推理与代码生成任务中的表现。实验结果表明，在六个不同领域-学生组合设置下，Waypoint-OPD均取得最优平均性能，且消融实验验证了方向一致性门控机制对数学推理能力的关键作用。因此，本研究的关键贡献在于证明：利用早期与晚期教师策略转移的方向一致性来加权末端转移信号，可有效提升策略蒸馏的鲁棒性与有效性。
链接: https://arxiv.org/abs/2610.21601
作者: Bingxi Hou,Guochao Jiang,Guofeng Quan,Weiqing Li,Wenfeng Feng,Guohua Liu,Yuewei Zhang
机构: Alibaba Cloud Computing(阿里云计算)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Policy-shift distillation reuses the difference between a teacher’s log probabilities before and after reinforcement learning (RL) as supervision on student-generated trajectories. However, this endpoint policy shift cannot reveal whether local action preferences evolved consistently during teacher training. The same endpoint shift can result from early and late shifts that reinforce or partially cancel each other. We investigate whether agreement between early and late teacher policy shifts can guide selective transfer. We introduce Waypoint-guided On-Policy Distillation (Waypoint-OPD), which uses an intermediate teacher checkpoint from the same RL run to compare early and late shifts in relative action preferences at student-visited prefixes. We gate the endpoint shift according to the directional agreement between the early and late shifts, suppressing it when they point in opposing directions. Compared with endpoint-only transfer, our method requires only one additional frozen-teacher scoring pass. Using RL-trained Qwen3-4B teachers, we evaluate our method with Qwen3-8B, Qwen3-14B, and Qwen3-30B-A3B as student models in mathematical and code reasoning. In all six domain–student settings, Waypoint-OPD achieves the highest domain-average score among the compared distillation methods. Ablations with the Qwen3-8B student show that removing or shuffling gates reduces mathematical reasoning performance. These results support the use of directional agreement between early and late teacher policy shifts to weight the endpoint shift during student training.
[NLP-64] Rethinking the Role of Tool Graph Memory in Multi-Turn Tool Use
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）代理在多轮任务中依赖外部工具时，如何有效利用历史经验以提升工具选择准确性的核心问题。现有方法提出使用工具图记忆（tool graph memory），将过往的工具调用序列作为状态转移关系存储，以指导后续的工具选择。然而，研究发现该机制在九个模型和两个基准测试上表现不佳，其性能提升微弱且往往伴随超过两倍的LLM调用开销；进一步分析表明，工具图记忆提供的引导常不准确，而随机或无引导策略反而能保留大部分性能增益，说明实际收益主要源于额外计算量而非记忆本身的有效性。此外，即使正确预测了下一工具，代理仍频繁因放弃调用工具而失败。相比之下，将历史经验以文本形式提炼为可解释的洞察（textual insights）进行存储，不仅显著优于工具图记忆，且几乎不增加额外的LLM调用成本。因此，该研究的关键解决方案在于：将记忆从“对下一步工具的硬性预设”转变为“提供软先验（soft prior）的可解释知识”，从而实现更高效、更具泛化能力的工具使用策略。
链接: https://arxiv.org/abs/2610.21571
作者: Uiji Hwang,Richeng Xuan,Taeuk Kim
机构: Hanyang University (汉阳大学); Hunyuan, Tencent (腾讯混元)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents solve complex tasks by calling external tools over multiple turns. Recent work has proposed tool graph memory, which stores past experience as tool transitions to guide subsequent selection. However, it remains unclear whether this component consistently improves agents’ tool use or undermines it. We evaluate tool graph memory on nine models and two benchmarks, finding that it seldom improves significantly over baselines, and that its gains, where present, cost over twice the LLM calls. The memory’s guidance is often incorrect, and random or no guidance retains most of the gain, implying that the benefit stems from extra computation rather than the memory itself. Moreover, even with the correct next tool suggested, many failures remain, such as giving up without a tool call. Storing the same experience instead as textual insights outperforms tool graph memory with almost no extra LLM calls. This suggests that memory serves better as a soft prior than as a prescription of the next tool, leading to a more effective and generalizable solution.
[NLP-65] DialFork: Going Beyond Factuality to Evaluate Procedural Hallucination in Conversational Agents
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）作为对话代理时存在的程序性幻觉（procedural hallucination）问题，即模型在生成语义流畅且看似合理的对话内容时，偏离了预期的对话流程逻辑。这一现象不同于已受广泛关注的事实性幻觉（factual hallucination），其核心挑战在于对话过程的结构性偏差难以通过传统基于内容的评估方法捕捉。论文提出DialFork，首个基于原始对话文本的基准测试，通过在可控对话轮次位置对合规对话进行分叉，并注入四种类型的程序性幻觉，构建出涵盖12个领域、24个任务、共计36,278条对话的数据集，经30名人工标注者验证其有效性。为评估不同粒度的程序性偏离，研究定义了三项任务：监控部署中代理的整体流程偏差、单条对话的程序性检索以及实时识别对话首次偏离的轮次。针对前两项任务，提出两种面向对话流（dialog-flow-aware）的度量指标——FlowKL与FlowScore，其核心思想是基于对话的动态演化路径而非静态内容进行比较。实验表明，FlowScore在七种编码器上均显著优于所有基线，证明仅依赖语义相似性无法有效衡量程序性一致性。该研究的关键贡献在于揭示并量化了程序性幻觉的严重性，推动社区在评估对话系统时将对话过程（conversational process）与内容（content）并重。
链接: https://arxiv.org/abs/2610.21561
作者: Sergio Burdisso,Esaú Villatoro-Tello,Dairazalia Sánchez-Cortés,Yanis Labrak,Jordi Luque,Milos Cernak,Petr Motlicek
机构: Idiap Research Institute(伊迪亚普研究学院), Martigny, Switzerland; Telefónica Innovación Digital(西班牙电信创新数字), Barcelona, Spain; Logitech(罗技), Lausanne, Switzerland; Brno University of Technology(布尔诺理工大学), Brno, Czech Republic
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Large language models (LLMs) deployed as conversational agents are increasingly expected to follow procedural logic. While factual hallucination has received extensive attention, a different failure mode has been overlooked: procedural hallucination, in which an agent produces individually fluent and factually plausible utterances while deviating from the intended conversational process. We introduce DialFork, the first benchmark to measure this failure mode on raw dialogue text. DialFork is built by forking conformant dialogues at controlled turn positions and injecting one of four hallucination types, yielding 36,278 dialogues across 24 tasks in 12 domains; 30 human annotators validate the construction. We define three evaluation tasks that characterize procedural deviation at three granularities: how far a set of dialogues deviates from conformant ones (monitoring deployed agents), how far a single dialogue does (procedural retrieval), and at which turn a dialogue first goes off-track (real-time guardrails). For the first two, we propose dialog-flow-aware metrics, FlowKL and FlowScore, that compare dialogues by how they unfold rather than their content. At the dialogue level, FlowScore significantly outperforms all baselines across seven encoders, showing that semantic similarity is no substitute for procedural similarity. We release DialFork under the MIT license at this https URL, hoping the community will measure conversational process alongside content.
[NLP-66] Writability and the Pathway: A Controlled Study of Knowledge Injection into Addressed Memory
【速读】： 该论文旨在解决生成式 AI（Generative AI）中后置知识注入（post-hoc knowledge injection）的可实现性与稳定性问题，特别是针对条件记忆（conditional memory）在写入后是否能有效保留未目标行为（untargeted behavior）及在新问题上泛化的能力。其核心挑战在于：尽管写入操作可能成功触发特定目标响应，但这种写入未必能持久维持或迁移到新的查询场景中，且易引发模型行为漂移（drift）。解决方案的关键在于通过受控实验设计，结合冻结的 GPT-2 模型与新型参数模块结构，在不同更新策略（仅行更新 vs. 行+投影更新）和模块残差锚定（module-residual anchoring）条件下评估写入效果。结果表明，采用固定锚定机制可显著降低行为漂移，提升目标字符串命中率（尤其在投影更新中达22.0%），但严格答案准确率仍为零，说明写入仅影响部分词元层级表达，无法保证完整答案正确性。在 Qwen3.8-Flash-Next-FP8 和 DeepSeek-V4.1-Flash 的实测中，梯度写入使指定首词元在132个独立重置目标中成为最高排名，其中93个为新晋升目标，验证了有限写入的可行性；而对比优化器（Adam vs SPSA）显示计算效率差异对性能有显著影响。此外，零阶优化在 DeepSeek 中仅实现一次新目标晋升，进一步凸显写入能力的边界性。综上，研究揭示了写入行为与训练表达、问题形式化及长期保留之间的解耦特性，首次词元排名不等同于完整答案成功或通用可写性，从而明确了生成式系统中可控记忆写入的局限性与潜在路径。
链接: https://arxiv.org/abs/2610.21553
作者: Wei Gao(1),Cheng Tian(1),Duan Yang(2),Jianqing Yin(1) ((1) Shibei Qiyuan AI Technology, (2) Xi’an Jiaotong University)
机构: Shibei Qiyuan AI Technology; Xi’an Jiaotong University
类目: Computation and Language (cs.CL)
备注: 22 pages, 2 figures
Abstract:Addressed conditional memory provides a local surface for post-hoc knowledge injection, but a successful write need not transfer to a new question or preserve untargeted behavior. This study combines a controlled GPT-2 comparison with production-memory experiments on Qwen3.8-Flash-Next-FP8 and DeepSeek-V4.1-Flash. Newly initialized modules attached to frozen GPT-2 are tested in a 2 x 2 design: row-only versus row-plus-projection updates, each with or without fixed module-residual anchoring. All arms reproduce 50 trained completions in each of three sampling seeds. Natural-question target-string hits average 5.3% for both row-only arms, 28.7% for unanchored projection updates, and 22.0% with anchoring, while strict answer scores remain zero. Anchoring lowers measured drift; lexical hits do not establish entity-specific retrieval. In the archived Qwen experiments, gradient writes make the designated first answer token top-ranked for 132 independently reset targets, comprising 93 new promotions and 39 already top-ranked targets. A separate matched 24-target comparison gives 13 new promotions with Adam versus two with SPSA at equal iteration counts but unequal computational procedures. An 80-row case includes saved continuations and a replay record, while reformulations bound its scope. Six DeepSeek zero-order cases yield one new target-token promotion and one regression. These results support bounded writing feasibility and separate trained expression, question-form behavior, and preservation; first-token rankings do not establish complete-answer success or universal conditions for writable memory.
[NLP-67] SCRIBE: Scalable Contextual Representations of Interpretable Brain-Grounded Semantics
【速读】： 该论文旨在解决人类大脑如何在真实语言语境中构建语义理解这一核心科学问题，尤其关注语义依赖于动态变化的上下文这一关键挑战。其解决方案的关键在于提出SCRIBE框架——一个基于神经科学启发的语义空间，通过集成多个开源大语言模型（Large Language Models, LLMs），生成可解释的、词级的语义表征，能够捕捉自然语言中随上下文变化的语义动态。研究通过921名被试的行为实验验证了人类对语义特征的判断具有系统性上下文依赖性，并表明SCRIBE能有效捕捉此类语义迁移，达到估计人类噪声上限的75%-81%。结合自然叙事材料下的脑活动记录数据，进一步证明SCRIBE在语言选择性脑区（如布罗卡区和韦尼克区）的神经响应预测能力显著优于非上下文敏感的语义表征。方差分解分析显示，SCRIBE所捕获的信息在预测脑活动方面具有独特贡献，超越了句法特征所能解释的部分，表明其表征能够反映当前句法模型未涵盖的语义层面信息。综上，SCRIBE为在真实语境下研究语言语义与大脑活动之间的关系提供了一个可扩展且可解释的计算基础。
链接: https://arxiv.org/abs/2610.21518
作者: Camila Kolling,Omer Moussa,Emin Çelik,Mariya Toneva
机构: Max Planck Institute for Software Systems (MPI-SWS)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Understanding how the human brain constructs meaning from real-world language remains a fundamental challenge, as semantics depend on context that varies in natural settings. To enable the study of real-world linguistic meaning in the brain, we introduce SCRIBE, a framework grounded in a neuroscience-informed semantic space that leverages an ensemble of open-source large language models to generate interpretable, word-level semantic representations of arbitrary text. In a behavioral study with 921 participants, we show that human semantic feature judgments vary systematically with context, and that SCRIBE captures these shifts, reaching 75-81% of the estimated human noise ceiling. Using brain activity recorded while participants listen to naturalistic stories, we further demonstrate that SCRIBE predicts neural responses in language-selective brain regions, significantly outperforming non-contextual semantic representations. Variance partitioning reveals that SCRIBE explains unique brain-predictive information beyond syntactic features, suggesting that its representations capture aspects of linguistic meaning not accounted for by the measured syntactic features. Together, these findings establish SCRIBE as a scalable and interpretable foundation for studying natural language semantics in the brain.
[NLP-68] Decomposing Brain-LLM Alignment with Interpretable Semantic Representations
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在自然语言理解过程中与神经活动对齐的可解释性问题，即揭示驱动脑-模型对齐的底层语义信息。其解决方案的关键在于引入一种可解释的上下文语义表征工具——SCRIBE（Semantic Contextual Representation with 65 neurocognitively grounded semantic properties），作为解析LLaMA2-7B模型与人脑功能磁共振成像（fMRI）数据之间对齐机制的分析框架。通过在七名参与者聆听叙事故事（Moth Radio Hour）时采集的自然主义fMRI数据上进行方差分解分析，研究分离了LLaMA2-7B与SCRIBE共享及各自独有的脑预测方差。结果表明，共享方差在全脑及语言相关区域占主导地位，但两者均保留显著的独有贡献；进一步分析发现，具体性（concreteness）较高的语义属性在共享方差中表现更强，而抽象属性并未表现出更显著的SCRIBE独有方差；此外，属性删减实验显示语义贡献具有多维度累积效应，影响共享与SCRIBE独有方差。综上，该研究揭示：尽管大语言模型（如LLaMA2-7B）与大脑活动的对齐部分可由可解释的语义信息解释，但仍存在超越模型本身表征的、互补性的脑预测语义信息。
链接: https://arxiv.org/abs/2610.21508
作者: Camila Kolling,Mariya Toneva
机构: Max Planck Institute for Software Systems (MPI-SWS)(马克斯·普朗克软件系统研究所)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Large language models (LLMs) predict neural responses during naturalistic language comprehension, but the linguistic information underlying this brain alignment remains difficult to interpret. We use SCRIBE, an interpretable contextual semantic representation spanning 65 neurocognitively grounded semantic properties, as a lens for decomposing the brain alignment of one LLM, LLaMA2-7B. Using variance partitioning between LLaMA2-7B and SCRIBE on naturalistic fMRI data from seven participants listening to narrative stories (Moth Radio Hour), we separate brain-predictive variance shared between the two representations from variance unique to each. We find that shared variance is the dominant component across the whole brain and language-related regions, while both representations also retain significant unique contributions. We then examine whether the concreteness of individual semantic properties relates to shared and SCRIBE-unique brain-predictive variance. Concrete properties show greater shared variance with LLaMA2-7B than abstract properties across the whole brain and most examined regions, whereas abstract properties do not show greater SCRIBE-unique variance. Finally, property ablations reveal that semantic contributions extend across multiple dimensions, with cumulative effects on both shared and SCRIBE-unique variance. Together, these results show that, for LLaMA2-7B on this dataset, a substantial portion of brain-LLM alignment is associated with interpretable semantic information, while revealing complementary brain-predictive semantic information beyond the LLaMA2-7B representation.
[NLP-69] OPRD: On-Policy Reminder Distillation for Long-Horizon Coding Agents
【速读】： 该论文旨在解决代码生成代理在长期交互中难以有效利用上下文信息、维持约束条件并响应工具反馈的问题，尤其针对仓库级（repository-level）代码任务中因缺乏充分过程监督而导致的性能瓶颈。现有方法依赖最终的通过/失败结果进行训练，但这种稀疏的监督信号难以引导代理在复杂推理过程中持续关注关键信息。其解决方案的关键在于提出基于策略提醒蒸馏（On-Policy Reminder Distillation, OPRD） 的新范式：通过多粒度提醒（Multi-Scope Reminder）机制，在无需显式提醒的“学生”模型滚动生成过程中，结合结果导向的路由策略，对成功轨迹采用自回归自模仿目标，对失败轨迹则引入由教师模型重新评估同一前缀后产生的提醒条件化反KL损失。其中，全局提醒将最终测试诊断转化为轨迹级指导，而回合局部提醒仅作用于与特定环境观测相关的下一步决策。教师与学生共享从代码微调（Coding SFT）和指令跟随微调（Instruction-Following SFT）两阶段初始化的结构。在推理阶段，学生模型完全不依赖提醒，从而实现高效部署。实验表明，在SWE-bench Verified、SWE-bench Pro及SWE-bench Multilingual三个基准上，相较于开源Qwen3.5-4B基线模型，OPRD将平均任务解决率提升11.3个百分点，并在相同诊断协议下将累计错误事件数从446降至222，降幅达50.2%。
链接: https://arxiv.org/abs/2610.21496
作者: Yuechen Luo,Zhiyi Lai,Miao Ziqi,Zihang Xu,Jingxuan Xu,Songwei Yu,Wenqiang Zhu,Weihao Li,Junhao Fan,Haoyang Huang,Yanan Wu,Ken Deng
机构: 未知
类目: Computation and Language (cs.CL)
备注: 19 pages, including appendix
Abstract:Reliable repository-level coding requires agents not only to generate correct code but also to preserve constraints and respond to tool feedback across long interactions. However, our diagnostic analysis shows that agents often fail to act on relevant information already available in context, while final Pass/Fail outcomes provide only sparse process supervision. We introduce On-Policy Reminder Distillation (OPRD), which combines Multi-Scope Reminder supervision with outcome-based routing on reminder-free student rollouts. Successful rollouts provide autoregressive self-imitation targets, whereas failed rollouts receive Reminder-conditioned reverse-KL supervision from a teacher that re-evaluates the same student-generated prefixes. A Global Reminder converts final-test diagnostics into trajectory-level guidance, while Turn-local Reminders apply only to the next decision following the relevant environment observation. The teacher and student share a two-stage initialization from Coding SFT and Instruction-Following (IF) SFT. At inference, the student operates without Reminders. On SWE-bench Verified, SWE-bench Pro, and SWE-bench Multilingual, OPRD improves the benchmark-averaged task resolution rate by 11.3 points over the open-source Qwen3.5-4B base model. Under the same diagnostic protocol, it reduces the aggregate error-event count from 446 to 222, a 50.2% reduction.
[NLP-70] Can LLM s Tell Stories Out of Order? A Corpus and Benchmark for Nonlinear Narrative Generation
【速读】： 该论文旨在解决生成式 AI 在处理非线性叙事（nonlinear storytelling）时的表现问题，即语言模型能否在事件时间顺序与叙述顺序不一致的情况下，准确且自然地重构复杂叙事结构。其核心挑战在于：尽管人类创作的非线性故事通过因果关联、人物认知和悬念设置等机制实现叙事张力，但当前语言模型在被要求生成此类故事时，往往仍倾向于按时间顺序组织内容，难以真正掌握非线性叙事的深层结构。解决方案的关键在于构建一个包含83部已发表作品的基准测试体系，对每部作品标注事件的时间顺序与叙述顺序，并建立包含因果依赖、角色知识状态及叙事疑问与解答的故事图谱（story graph）。通过对比模型在不同条件下的输出——包括默认生成、受指令约束的非线性生成以及基于故事图谱的生成——研究发现，尽管模型能按指令重排事件，但其生成的非线性故事在可读性和叙事吸引力上显著低于默认的线性版本；而基于故事图谱生成的故事则更受人类读者青睐，显示出结构化语义信息对高质量非线性叙事的重要作用。这表明，仅靠指令重排不足以实现有效的非线性叙事生成，必须引入深层语义结构引导。
链接: https://arxiv.org/abs/2610.21468
作者: Syed Rifat Raiyan,Md. Tariquzzaman,Md Farhan Ishmam,Tenghao Huang
机构: Islamic University of Technology (伊斯兰科技大学), Bangladesh; University of Utah (犹他大学), USA; University of Southern California (南加州大学), USA
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Under review, 36 pages, 13 figures, 15 tables. Code and data: this https URL . Project page: this https URL
Abstract:Every story has two orders: the order in which its events happen and the order in which they are told. A story is nonlinear when the two differ. We introduce a benchmark of 83 published works that measures whether language models can tell nonlinear stories. For each work, we annotate one list of events in both orders, together with a story graph of causal dependencies, character knowledge, and the questions the telling raises and resolves. Given events in chronological order, all 19 models tell them, on average, more chronologically than the published works do; asked for a specific structure, they land within the human range about half the time. We then compare a model’s stories of the same events under different conditions, using both an automated judge and 71 human readers. Both prefer the chronological story a model writes by default to the nonlinear one it writes on request, which readers find harder to follow in 81% of decisive judgments. Readers also prefer stories written from the story graph to stories constrained to the published telling order (72% of decisive judgments), a difference the judge does not detect. Our results suggest that following an instruction to reorder a story is not yet the same as telling it well out of order.
[NLP-71] rajectory-Level Speculation with Semantic Verification for Tool-Using Agents
【速读】： 该论文旨在解决大语言模型智能体（LLM agents）在执行任务时因依赖单一强模型持续生成每一步而导致的高延迟与高内存开销问题。尽管许多任务步骤具有高度重复性，适合由轻量级模型处理，但现有方法仍让强模型承担全部生成责任，仅通过路由或协作机制决定由哪个模型执行具体步骤，或采用推测性方法逐步验证弱模型草稿，且验证过程仍依赖强模型自身输出。其解决方案的关键在于提出一种名为SpecAgent的协同协议：将强模型的角色从每一步的“作者”转变为弱模型所生成步骤的“评审者”。在此框架下，弱模型以只读方式执行工具调用，确保每一步基于真实观测结果，并保留首个可能修改环境的动作；随后，强模型通过一次预填充（prefill）并行验证整个草稿段，仅需判断语义有效性并提交最长有效前缀，对首条被拒步骤进行修复。由于验证过程取代了自回归解码，强模型不再需要生成每一步，而仅需输出短评和偶尔修复，显著降低计算负担。实验表明，在τ²-Bench上，SpecAgent接近强模型成功率，同时将内存流量和运行时间减半；在AutomationBench与VitaBench上，恢复44-75%的质量差距，且整体性能更优；在边缘-云部署中，利用三台云模型，其在仅生成最多一半强模型令牌的情况下，缩小了53-75%的弱-强质量差距。研究结果表明，强模型无需亲自撰写每个步骤，通过评审基于真实观测的弱模型草稿，可在保持近似性能的同时，大幅降低延迟与内存通信开销。
链接: https://arxiv.org/abs/2610.21463
作者: Zilin Wang,Yinhong Liu,Chao Chen,Qun Liu,Zhijiang Guo
机构: LARK, HKUST (GZ); Yinwang; University of Cambridge; HKUST
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Large language model agents solve tasks through long sequences of tool calls, and the prevailing design has a single strong model generate every step, although many steps are routine enough for a much smaller model. Existing ways of sharing this work keep the strong model as the generator: routers and collaborative agents decide which model handles each query or step, and speculative methods check a small model’s drafts one step at a time, usually against the strong model’s own action. We introduce SpecAgent, a collaboration protocol that turns the strong model from the author of each step into a reviewer of steps drafted by a weak model. The weak model executes read-only tool calls as it drafts, so that every step is grounded in real observations, and holds the first action that would modify the environment. The strong model then verifies the drafted segment in a single prefill pass, commits the longest prefix it judges semantically valid, and repairs the first rejected step. Because verification replaces autoregressive decoding with parallel prefill, the strong model decodes only short verdicts and occasional repairs rather than every step. On \tau^2 -Bench, SpecAgent comes within one point of the strong model’s success rate while roughly halving memory traffic and wall time, and on AutomationBench and VitaBench it recovers 44-75% of the quality gap while lowering both. In an edge-cloud deployment on MCP-Atlas with three cloud models, it closes 53-75% of the weak-strong quality gap while the strong model generates at most half as many tokens. These results suggest that a strong model need not author every step of an agent: reviewing grounded drafts retains most of its quality while cutting both latency and memory traffic.
[NLP-72] ranslation Circuit Discovery and Targeted Optimization for Low-Resource Multilingual Translation
【速读】： 该论文旨在解决低资源多语言翻译中模型内部机制不明确的问题，特别是针对印度语（Hindi）到乌尔都语（Urdu）及印地语（Hindi）到信德语（Sindhi，使用天城文书写）这类低资源语言对的翻译性能瓶颈。现有主流多语言大语言模型虽在翻译任务上表现优异，但其内部如何实现跨语言知识迁移与翻译决策仍缺乏可解释性。为此，作者提出TITAN——一种基于因果干预、稀疏特征分解与目标偏好优化相结合的机械电路发现框架，通过识别并优化Transformer架构中对翻译至关重要的稀疏因果电路，实现高效参数更新。其关键创新在于：摒弃传统全参数微调方式，仅对占模型参数总量不足5%的、具有因果重要性的稀疏电路进行优化，从而在保持极低参数更新量的同时显著提升翻译质量。实验结果表明，TITAN在Hindi-to-Urdu和Hindi-to-Sindhi任务上分别相对于最强基线模型实现33.75%和45.86%的BLEU分数提升，且人工专家评估验证了其在准确性和流畅性上的显著改进。进一步分析揭示，多语言翻译并非依赖以英语为中心的中间表示，而是通过共享语义子空间中的稀疏、结构化电路实现，为理解多语言Transformer的内在工作机制提供了可解释性洞见。
链接: https://arxiv.org/abs/2610.21456
作者: Atai Waris Khan,Abdullah Mazhar,Raghu Kumar,Md Shad Akhtar
机构: Indraprastha Institute of Information Technology Delhi (印第普拉斯特拉信息科技学院德里分校)
类目: Computation and Language (cs.CL)
备注:
Abstract:Recent multilingual large language models achieve strong translation performance, yet the internal mechanisms responsible for low-resource multilingual translation remain insufficiently understood. In this paper, we study translation circuits in multilingual decoder-only LLMs for Hindi-to-Urdu and Hindi-to-Sindhi (Devanagari). We propose TITAN, a mechanistic framework that combines causal intervention, sparse feature decomposition, and targeted preference optimization to identify and optimize crucial translation components in the transformer. Unlike conventional full-parameter fine-tuning, TITAN restricts optimization to sparse, causally important circuits, enabling scalable and parameter-efficient adaptation while updating fewer than 5% of model parameters. Experimental results show that TITAN achieves BLEU improvements of 33.75% on Hindi-to-Urdu and 45.86% on Hindi-to-Sindhi over the strongest multilingual baselines. Human expert evaluation further confirms substantial gains in adequacy and fluency across both language pairs. Additional analyses reveal that multilingual translation is implemented through sparse, structured circuits operating over a shared semantic subspace rather than through an English-centric pivot representation. Overall, our findings demonstrate that mechanistic circuit discovery can directly improve low-resource multilingual translation while providing interpretable insights into multilingual transformer behavior.
[NLP-73] SITAR: A Situational Distortion Metric for Automatic Speech Recognition Evaluation
【速读】： 该论文旨在解决自动语音识别（ASR）中转录错误在语义或词汇相似性相近时，其情境影响差异难以被现有评估指标准确捕捉的问题。传统方法依赖大语言模型（LLM）进行评分虽具上下文理解能力，但存在计算开销大与评分结果不一致的弊端。为此，论文提出SITAR（Situational Distortion Metric），一种基于端到端框架构建的情境失真度量方法，其核心在于通过有界递归公式将情境严重程度与累积影响范围整合为统一得分。该方法利用有限离线LLM生成的标注目标，训练出一个轻量级BERT类回归器，可在单次前向传播中高效评估参考句与假设句对，完全避免高运行时成本。为确保新指标的科学性与可靠性，研究进一步设计了包含外在有效性（宏观与微观层面）和内在评分器稳健性的验证协议：前者在多数据集与模型上验证了SITAR能有效追踪系统能力扩展并提供互补诊断价值；后者在定制化控制基准上确认了评分机制本身的可操作性。系统性验证表明，SITAR具备作为现有ASR评估指标实用且轻量补充的巨大潜力。
链接: https://arxiv.org/abs/2610.21440
作者: Mingzhe Lu,Jinhao Liu,Yunpeng Li,Hang Su,Lichun Fan,Jian Luan
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); Xiaomi Inc. (小米公司)
类目: Computation and Language (cs.CL)
备注: 33 pages, 4 figures, including appendices
Abstract:In automatic speech recognition (ASR), transcription errors with similar lexical or semantic discrepancies can have markedly different consequences for the situation conveyed by speech. Capturing these differences requires not merely contextual assessment, but an understanding of what the spoken discourse describes, establishes, or seeks to bring about. However, relying on LLM judges introduces computational cost and scoring variability. We introduce SITAR, a situational distortion metric for ASR, realized through a comprehensive end-to-end construction framework. Specifically, this framework aggregates situational severity and cumulative extent into a unified score via a bounded recursive formulation. Using targets derived from this formulation with limited offline LLM support, we train SITAR as a lightweight BERT-family regressor that scores reference-hypothesis pairs in a single forward pass, entirely free of any prohibitive runtime costs. Recognizing that introducing a new metric demands rigorous justification, we formalize a principled validation protocol spanning both extrinsic validity, assessed at macro and micro levels, and intrinsic scorer reliability. The former, evaluated across diverse datasets and models, demonstrates that SITAR reliably tracks broad capability scaling while providing orthogonal diagnostic utility, whereas the latter, examined on a purpose-built controlled benchmark, confirms the operational soundness of the scoring instrument itself. Through systematic validation, we demonstrate that SITAR holds strong potential as a practical, lightweight complement to existing ASR metrics. We will release the full recipes (code, model weights, dataset, etc.) after the internal review to support community research.
[NLP-74] HyCo: Hybrid Drafting for Cost-Efficient Mobile-Cloud Speculative Decoding ACL
【速读】： 该论文旨在解决设备-云协同推理中，基于小语言模型（SLM）生成深度草稿树时，因在移动端神经处理单元（NPU）上反复迭代导致计算开销过高的问题。其核心解决方案是提出HyCo框架，通过将服务器端原生多标记预测（MTP）头的推理任务卸载至设备端执行，仅需从服务器获取必要的隐藏状态、输入嵌入及键值缓存，从而避免在服务器侧重复执行序列化计算步骤。该设计结合量化、受限词表与1K token窗口机制，使草稿生成器适配于NPU资源约束；同时引入ε阈值规则，允许对验证器采样分布进行适度松弛，进一步提升效率。实验表明，在单客户端场景下，相较于服务器端自推测的MTP方法，HyCo在Spec-Bench和32K摘要任务中分别降低服务器GPU耗时38.6%与29.2%；在三并发运行中，相比纯SLM草稿器，实现9.0个百分点的效率提升与1.9倍的交付速度增益。尽管在人工评估中与Cloud+MTP方案差异不显著（+0.2分，95%置信区间：-0.9至+1.3），但探索性翻译任务更支持SpecEdge-ε策略。若不采用分布松弛，则仍可实现27.7%和11.9%的精确节省。
链接: https://arxiv.org/abs/2610.21436
作者: Hanbin Bae,Heejin Choi,Joohyung Lee,Byoung Jin Choi,Sangjun Park,Hoon-Young Cho
机构: Samsung Research (三星研究院)
类目: Computation and Language (cs.CL); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
备注: Under review at ACL Rolling Review (October 2026)
Abstract:Device-cloud speculative decoding reduces server computation, but growing a deep draft tree with a small language model (SLM) requires repeated passes on a mobile neural processing unit (NPU). We present HyCo, which assigns depth to the server model’s native multi-token prediction (MTP) head, executed on the device, and alternative branches to one SLM pass. The server returns the hidden state, input embedding, and key-value entries needed by the head instead of executing its sequential steps. Quantization, a restricted vocabulary, and a 1K-token window fit both drafters to the NPU. An \epsilon-threshold rule additionally relaxes exact preservation of the verifier’s sampling distribution. In single-client runs with a Galaxy S26 Ultra over LTE/5G, HyCo reduces server GPU time per output token relative to server-side MTP self-speculation by 38.6% on Spec-Bench and 29.2% on 32K-token summarization. With the same acceptance rule, three Spec-Bench runs per system show 9.0 percentage points more saving and 1.9x delivery speed than an SLM-only drafter. HyCo differs from Cloud+MTP by +0.2 judge-score points (paired 95% CI: -0.9 to +1.3); exploratory translation favors SpecEdge-\epsilon. Without relaxation, HyCo exact saves 27.7% and 11.9% on the two datasets.
[NLP-75] Do Agents Need a New Kind of Paper? SPRA: Structured Paper Representation for Agents
【速读】： 该论文旨在解决科学智能体（Scientific Agents）在处理分散于叙事性论文中的研究问题、方法与发现时，缺乏统一、可计算且可追溯的结构化表示这一关键挑战。其核心问题是现有文献信息呈现为非结构化文本，导致科研知识难以被系统性地连接、查询与重用。为此，论文提出SPRA（Structured Paper Representation for Agents），一种面向持久研究对象及其类型化关系的论文级表示框架。其解决方案的关键在于构建一个具备实证-人工智能（empirical-AI）特征的元数据模型，将论文中的问题（Problem）、贡献（Contribution）、组件（Component）、实验设置（ExperimentSetup）、度量指标（Measure）和研究成果（Finding）等研究单元进行标准化组织，并通过选择性记录论文间的引用关系（如依赖、组件复用、与前序工作的比较）实现跨文献的知识关联，同时保留原始来源位置以支持验证。该框架在超过一百万篇人工智能论文上进行了实例化，形成1995万个研究单元与2407万条内部关系。评估表明，基于SPRA的Schema-RAG在任务证据固定前提下显著提升了源条件下的科学问答性能（相比段落检索提升0.381/5），并在定向跨论文推理中使资格判断得分提高0.184/5；此外，在输入输出令牌消耗方面，相较段落检索减少28.5%，相较全量紧凑表示（Full Compact）降低95.3%。尽管在创意生成任务中整体得分略低于全量紧凑表示（低0.067分），但输入输出令牌数减少达93.7%。综合结果表明，SPRA不仅是一种可用的科学内容表示形式，更可作为连接、审查与重用论文级研究对象的研究工作流基础。该资源已开放于SciVerse学术数据平台。
链接: https://arxiv.org/abs/2610.21424
作者: Yutong Leng,Haojiong Chen,Fukai Shang,Xiaoyang Wang,Mindong Liu,Runyuan Ma,Jiantao Qiu
机构: Shanghai Artificial Intelligence Laboratory (上海人工智能实验室); Shanghai Jiao Tong University (上海交通大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 23 pages, including references and appendices; resource website: this https URL
Abstract:Scientific agents must connect research problems, methods, and findings dispersed through narrative articles. We introduce SPRA (Structured Paper Representation for Agents), a paper-level representation of persistent research objects and typed relations. Its empirical-AI profile organizes Problems, Contributions, Components, ExperimentSetups, Measures, and Findings for querying, traversal, and reuse. Selected paper-level citation relations record dependencies, component use, and comparisons with prior work; source locations support verification. We instantiate SPRA across more than one million AI papers, with 19.95 million research units and 24.07 million internal relations. Evaluations access SPRA through Schema-RAG with task evidence fixed before inference. Across seven subfields, SPRA improves source-conditioned scientific QA by 0.381/5 over passage retrieval while remaining close to Full Compact (0.078/5 lower on average). In directed cross-paper reasoning, qualification improves by 0.184/5 over passage retrieval; Top-5 synthesis shows mixed quality effects. Across all paired RQ2 tasks, including Top-5, input+output tokens decrease by 28.5% relative to passage retrieval and 95.3% relative to Full Compact. In idea generation, a 0.067-point lower Overall score than Full Compact accompanies 93.7% fewer input+output tokens. Together, these results support SPRA as a usable representation of scientific content and a foundation for research workflows that connect, inspect, and reuse paper-level research objects. The SPRA resource is accessible through the SciVerse scholarly-data platform.
[NLP-76] Identifying Effective Strategies in Prompt Evolution for Stress and Emotion Prediction ALT
【速读】： 该论文旨在解决社交媒体中早期压力检测与风险预测的准确性与可靠性问题，尤其关注当前大语言模型在社会媒体压力预测任务中因过度依赖准确率而掩盖假阳性结果的局限性。其核心挑战在于提示工程（prompt engineering）在不同模型与任务间效果不稳定，且提示修改策略与性能提升之间的关联不明确。论文提出的关键解决方案是引入一种基于双提示信息融合的增强型提示方法（GEPA+merge），通过整合来自两个不同提示的指令信息，在多种设置下显著提升了压力预测性能。研究进一步识别出五种与性能提升相关的有效提示策略：引导关注情绪证据、区分当前与过去压力状态、展示间接线索、要求提供情感上下文以及指导标签选择。此外，研究还揭示了四项关键现象：二分类压力预测在早期即表现改善，而多标签情绪预测初期下降但部分恢复；错误或冲突指令导致性能波动；小样本对比实验中的增益无法在全数据集上复现；以及提示在开发集与测试集上的排名变化。这些发现不仅明确了高效提示演进的方法路径，也指出了任务设计与评估设置中的潜在限制，为在社交媒体压力检测中应用提示演化方法提供了系统性指导。
链接: https://arxiv.org/abs/2610.21407
作者: SeungHyun Yi,Seungjun Yi,Sriram Sattiraju,Haeun Park,Joakim Nguyen,Terence Lim,Joseph Skrovan,Mehak Beri,Hitakshi Modi,Eloy Carrion,Andrew Well,Carlos M. Mery,Alexander Rasgon,Mia K Markey,Ying Ding
机构: Kyung Hee University (庆熙大学); Stanford University (斯坦福大学); University of Texas at Austin (德克萨斯大学奥斯汀分校); Vanderbilt University School of Medicine (范德堡大学医学院)
类目: Computation and Language (cs.CL)
备注: 4 figures, 2 tables. Submitted to the 2027 IEEE International Conference on Healthcare Informatics (ICHI 2027)
Abstract:Early stress detection and risk prediction are crucial for timely mental health assessment and support. While large language models have recently shown promise for social media stress prediction, relying on accuracy can obscure false positives. Also, prompt revisions yield inconsistent gains across models and tasks, with unclear links to strategy combinations and instruction changes. By comparing six methods on social media stress and emotion data across two model sizes, we found that GEPA and its extension combining instructions from two prompts (GEPA+merge) improved prediction over initial prompts in several settings. We also identified five prompting strategies associated with development gains. These strategies included directing attention to emotional evidence, distinguishing current from past distress, illustrating indirect cues, requiring emotional context, and guiding label selection. Moreover, we documented four observations across tasks and evaluation settings. In a separate set of searches, binary stress prediction improved early, whereas multilabel emotion prediction initially declined and partly recovered. The other three concerned incorrect or conflicting instructions, gains on small comparison samples that did not hold on full development data, and changes in selected prompts’ development and reported test rankings. These findings identify effective prompt evolution methods, strategy combinations associated with gains, and task and evaluation limits, providing guidance for applying prompt evolution methods to social media stress detection.
[NLP-77] Beyond Invariance: Factor-wise Selective Treatment of Shifts for Multimodal Learning
【速读】： 该论文旨在解决多模态学习中因分布偏移（distribution shifts）导致的鲁棒性问题。现有方法通常通过抑制依赖于环境的特征变化来追求不变性，但这种做法对所有偏移信息采取统一处理策略，忽略了不同偏移信息在预测中的差异化作用：部分信息对预测贡献甚微，部分信息具有稳定预测能力，而另一些则表现出环境依赖的预测效应。因此，核心问题是：哪些偏移信息应被抑制、保留或弱化？为此，作者提出一种因子级偏移框架 FaShift，其关键在于对潜在表示因子进行三重诊断——分布变化特性、干预效应以及跨环境稳定性。通过分解多模态表示并评估这些互补属性，FaShift 能有效区分非预测性因子、稳定预测性因子以及虽有用但不稳定的预测性因子，并据此选择性地抑制、保留或削弱其贡献，而非强制施加统一的不变性约束。实验在 MOSI、MOSEI 和 CH-SIMS 数据集上验证了该方法的有效性，结果显示 FaShift 在所有三个数据集上均实现了最优的域内（ID）准确率与 F1 值；相较于最强的域外（OOD）性能对比方法，其在 MOSI 上提升 OOD 准确率/ F1 1.5/3.6 个百分点，在 MOSEI 上提升 1.7/2.7 个百分点，并在 CH-SIMS 上达到当前最优的 OOD 准确率与 F1。
链接: https://arxiv.org/abs/2610.21403
作者: Guimin Hu,Qiwei Peng
机构: Guangdong University of Technology(广东工业大学); University of Copenhagen(哥本哈根大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Distribution shifts pose a fundamental challenge to robust multimodal learning. Existing methods typically promote invariance by suppressing environment-dependent variations, implicitly applying a uniform treatment to shifted information. However, shifted information can have different predictive consequences: some contributes little to prediction, some remains reliably informative, and some exhibits environment-dependent predictive effects. This raises a fundamental question: which shifted information should be suppressed, preserved, or attenuated? To address this question, we propose FaShift, a factor-wise shift framework that diagnoses latent representation factors according to their distributional changes, intervention effects, and cross-environment stability. By factorizing multimodal representations and evaluating these complementary properties, FaShift distinguishes non-predictive factors, stably predictive factors, and predictively useful but unstable factors. It then selectively suppresses, preserves, or attenuates their contributions, rather than enforcing uniform invariance. Experiments on MOSI, MOSEI, and CH-SIMS show that FaShift achieves the best ID accuracy and F1 across all three datasets. Compared with the strongest competitors reporting OOD results, it improves OOD accuracy/F1 by 1.5/3.6 points on MOSI and 1.7/2.7 points on MOSEI, while achieving the best reported OOD accuracy and F1 on CH-SIMS.
[NLP-78] Calibration Language Shapes Block-Influence Pruning in Multilingual Language Models
【速读】： 该论文旨在解决生成式 AI 模型中基于输入-输出余弦变化的块重要性（Block Influence, BI）评估在跨语言场景下的有效性问题，特别是英语校准数据能否有效迁移至印度语种（如印地语、马拉地语和泰米尔语）文本的深度剪枝任务。其核心问题是：当前以英语为校准语言的 BI 排名方法在处理非英语语言时存在显著偏差，导致对晚期层重要性的误判，从而影响剪枝决策的准确性。解决方案的关键在于引入多语言校准策略——通过在相同预算下对英语、印地语和马拉地语的 BI 值进行均衡平均，以缓解因语言分布不匹配导致的性能退化。实验表明，该方法在 Qwen2.5-1.5B-Instruct 与 Gemma-2-2B 模型上显著减轻了剪枝后困惑度（perplexity）的下降，尤其在包含未参与校准的泰米尔语测试集时仍保持鲁棒性；然而在 Llama-3.2-1B 上效果不稳定，甚至在大剪枝预算下出现严重性能崩溃。研究进一步揭示，简单采用均值聚合或最小最大排名等集成策略并不能普遍优于原始均值，强调了跨语言剪枝需结合具体模型架构与剪枝规模进行验证，而非依赖单一通用规则。因此，该工作支持一个诊断性结论：英语校准下的 BI 方法会系统性忽略印度语种文本中的晚期层激活特征，但多语言校准的有效性必须针对特定模型和剪枝配置进行实证检验。
链接: https://arxiv.org/abs/2610.21383
作者: Vedant Pandhare,Piyush Dhoka
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 16 pages, 3 figures. Code and results: this https URL
Abstract:Block Influence (BI) ranks Transformer blocks for depth pruning from their input-output cosine change, but the ranking depends on the calibration distribution. We study whether English calibration transfers to Indian languages in Qwen2.5-1.5B-Instruct, Llama-3.2-1B, and Gemma-2-2B. English, Hindi, Marathi, and Tamil use disjoint Wikipedia articles for calibration and evaluation; every selector receives the same 150-chunk budget, and Tamil is excluded from multilingual aggregation to test cross-script transfer. Across all three models, late-block BI is 20.7-55.5% higher for the three Indian languages than for English, and every predeclared late-minus-middle contrast has a 95% bootstrap interval above zero. This mismatch changes the low-BI blocks selected for removal despite high global rank correlations. Averaging BI over equal English, Hindi, and Marathi budgets mitigates the resulting perplexity damage for Qwen at six removed blocks and for Gemma at all tested depths, including held-out Tamil. The mitigation is not universal: on Llama, balancing improves Hindi and Marathi at two removed blocks but severely harms Tamil, while larger pruning budgets select the same blocks and collapse perplexity regardless of calibration. Rank-mean and minimax-rank aggregators do not dominate the raw mean. These results support a diagnostic claim rather than a universal pruning rule: English-only BI systematically misses late-layer activity for Indian-language text, but multilingual calibration must still be validated per architecture and pruning budget.
[NLP-79] DPA-MoE: A Runtime Precision Control Framework for MoE Inference EMNLP2026
【速读】： 该论文旨在解决混合专家模型（Mixture-of-Experts, MoE）在推理过程中专家权重保持固定精度所带来的计算效率瓶颈问题。尽管MoE通过路由机制仅激活少量专家以降低计算量，但每个被选中的专家仍以固定精度进行计算，限制了在不同资源约束下灵活优化精度与性能的潜力。其解决方案的关键在于提出DPA-MoE——一种运行时精度控制框架，实现了路由后专家权重的任意精度（any-precision）支持。该方法利用共享缩放的任意位宽量化编码（SharedScale AnyBCQ）将专家前馈权重编码为前缀可解码块，并在离线阶段对每个专家的重构残差进行建模；在推理时，根据实时路由门控分数动态分配相应的权重块。这种后路由分配机制使得单一编码后的MoE检查点可在不重新训练或重新编码的前提下，适应多种有效权重预算，从而在匹配预算条件下显著提升KL散度和下游任务准确率，同时在原型系统中仅引入适度的稳态吞吐量开销。
链接: https://arxiv.org/abs/2610.21378
作者: Jaeyun Kim,Hoonki Lee,Susie Shin,Hyunmin Kim,Dahyun Lee,Sangyeob Kim
机构: Agency for Defense Development(国防开发局); Yonsei University(延世大学)
类目: Computation and Language (cs.CL)
备注: 18 pages, 7 figures; Accepted to Findings of EMNLP 2026
Abstract:Mixture-of-Experts (MoE) models reduce computation by routing each token to a small set of experts, but each selected expert is still used at a fixed precision. We present DPA-MoE, a runtime precision control framework that makes routed experts any-precision. DPA-MoE encodes expert feed-forward weights as prefix-decodable chunks using SharedScale AnyBCQ, profiles per-expert reconstruction residuals offline, and allocates chunks at inference time using live router gate scores. This post-routing allocation lets a single encoded MoE checkpoint operate at multiple effective weight budgets without retraining or re-encoding. Experiments on Qwen3-30B-A3B and DeepSeekMoE-16B show that DPA-MoE improves KL divergence and downstream accuracy over uniform chunk allocation at matched budgets, with modest steady-state throughput overhead in our prototype.
[NLP-80] ESG-Kor: A Korean Dataset for ESG-related Information Extraction and Practical Use Cases EMNLP2024
【速读】： 该论文旨在解决在特定领域（即环境、社会与治理，ESG）中，由于缺乏高质量、标注精准的韩语语料库而导致的自然语言处理模型性能受限的问题。随着预训练语言模型在专业领域应用的扩展，构建适用于特定任务的领域数据集变得至关重要。为此，研究者构建了一个名为ESG-Kor的韩语数据集，该数据集包含118,946条句子，从韩国企业的可持续发展报告中自动提取并人工标注了各ESG维度的信息，依据权威ESG评级机构提供的客观标准进行标注。其解决方案的关键在于：一是构建了一个大规模、高精度且符合行业规范的韩语ESG专用数据集；二是通过在多个韩语预训练语言模型上验证分类性能，证明了该数据集的有效性与可适用性；三是将训练好的分类模型拓展至中小型企业文档，基于关键议题进行深入分析，展示了其在实际应用场景中的潜力，为推动韩国企业ESG信息披露自动化和智能化提供了坚实的数据基础与技术支撑。
链接: https://arxiv.org/abs/2610.21368
作者: Jaeyoung Lee,Geonyeong Son,Misuk Kim
机构: Hanyang University (汉阳大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Published in Findings of EMNLP 2024
Abstract:With the expansion of pre-trained language model usage in recent years, the importance of datasets for performing tasks in specialized domains has significantly increased. Therefore, we have built a Korean dataset called ESG-Kor to automatically extract Environmental, Social, and Governance (ESG) information, which has recently gained importance. ESG-Kor is a dataset consisting of a total of 118,946 sentences that extracted information on each ESG component from Korean companies’ sustainability reports and manually labeled it according to objective rules provided by ESG evaluation agencies. To verify the effectiveness and applicability of the ESG-Kor dataset, classification performance was confirmed using several Korean pre-trained language models, and significant performance was obtained. Additionally, by extending the ESG classification model to documents of small and medium enterprises and extracting information based on ESG key issues and in-depth analysis, we demonstrated potential and practical use cases in the ESG field.
[NLP-81] RenoLayout: Unified Floorplan Renovation and Furniture Layout for Constraint-Aware Residential Design
【速读】： 该论文旨在解决住宅改造设计中面临的两大核心挑战：一是现有平面图生成方法多假设输入为空白建筑边界，无法适应实际改造场景中需基于已有平面图进行推理的需求；二是传统方法将平面图改造与家具布局视为独立任务，忽视了二者之间的内在空间依赖关系。为此，论文提出一种以设计点（Design Point）为引导的条件生成框架RenoLayout，将预设计状态与设计点集合联合映射为可执行的“几何即代码”（Geometry-as-Code）程序，实现对结构改造与家具布局的统一建模。其解决方案的关键在于构建包含123个细粒度设计点的“设计地图”（Design Map），并通过基于视觉语言模型（VLM）的一致性验证，实现从“改造前后”案例对中自动标注设计点集合的“差异基底设计点注释”（Difference-Grounded Design-Point Annotation）方法。此外，RenoLayout采用三阶段训练流程（CPT–SFT–RL），逐步提升模型的空间推理能力、指令遵循能力及约束合规性。研究还发布了首个面向统一住宅设计的基准数据集RenoLayout Benchmark，包含200个真实世界改造案例，涵盖关键结构性与建筑服务类约束，显著推动了住宅设计智能化的发展。
链接: https://arxiv.org/abs/2610.21366
作者: Jingguo Liu,Yuanqing Liu,Chong Zhang,Yue Yang
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Residential design aims to transform existing homes into functional living spaces while satisfying diverse design preferences and physical constraints. However, existing floorplan generation methods typically assume empty building boundaries as input, whereas renovation requires reasoning over existing floorplans. Furthermore, floorplan renovation and furniture layout are commonly addressed as separate tasks, overlooking their inherent spatial dependency. Consequently, practical residential design must preserve immutable structural components while transforming editable spatial elements. It must also optimize furniture placement under design objectives and physical constraints. To address these challenges, we formulate residential design as a design-point-guided conditional generation task and propose RenoLayout, a framework that translates a pre-design residential state together with a design-point set into an executable Geometry-as-Code program. To enable controllable supervision, we construct a Design Map comprising 123 fine-grained design points covering structural renovation and furniture layout. We further introduce Difference-Grounded Design-Point Annotation to derive design-point sets automatically from before–after design pairs with VLM-based consistency verification. RenoLayout is trained using a three-stage CPT–SFT–RL pipeline, progressively improving spatial reasoning, instruction following, and constraint compliance. We also introduce the RenoLayout Benchmark, a set of 200 real-world before–after renovation cases with complete furniture layouts. To the best of our knowledge, it is the first benchmark for unified residential design. These cases include design-critical structural and building-service constraints.
[NLP-82] Emotion Is Not External Knowledge: Internal Attention Signals for Affective State Reasoning AACL2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在多轮对话中难以有效维护隐含情感状态（affective state）的问题，尤其当上下文线索稀疏且话题发生转移后，模型仍需识别并利用这些隐性信息进行推理。传统注意力干预方法依赖于特定领域标注的注意力头特征和外部指定的目标片段，导致在不同情感证据分布间的迁移能力受限。本文提出自目标内注意力增强方法（Self-targeted Internal Attention Boosting, SIAB），其核心在于通过因果定位（causal localization，具体为激活块修补法）在合成最小对上识别稀疏的情感相关注意力头，并在生成回答的位置增强这些头已关注的内容标记；整个过程完全基于模型自身行为生成，无需外部标注，使得同一组注意力头可跨任务、跨模型复用。实验表明，SIAB在ECC与GoEmotions数据集上分别相较基线提升最高达+10.3宏F1和+5.5宏F1；结合类似PASTA的外部监督后，性能进一步提升至+13.84准确率点和+12.51宏F1点，机制分析证实性能增益源于稀疏且富含情感信息的注意力头，且层间距离影响增益的叠加效果。
链接: https://arxiv.org/abs/2610.21347
作者: Wenqiu Tang,Zhen Wan,Yihui Wang,Takahiro Komamizu,Ichiro Ide
机构: Nagoya University (名古屋大学); National Institute of Informatics (日本信息研究所)
类目: Computation and Language (cs.CL)
备注: Accepted to AACL 2026 (Main Conference)
Abstract:Large Language Models (LLMs) increasingly operate in multi-turn interactions that require maintaining latent conversational states, not only explicit facts. We study affective state as an implicit state inferred from sparse contextual cues that remain relevant after topic shifts, and ask whether LLMs expose internal attention signals for routing such evidence at inference time. Prior attention interventions rely on labeled in-domain head profiles and externally specified target spans, limiting transfer across affective-evidence distributions. We propose Self-targeted Internal Attention Boosting (SIAB), which localizes sparse affective heads via causal localization (in detail activation patching) on synthetic minimal pairs and boosts, at the answer position, the content tokens these heads already attend to; both edited heads and target tokens are derived from the model’s own behavior, enabling the same head set to be reused across this http URL evaluate on ECC and GoEmotions across three LLMs and multiple prompting settings. Across settings, SIAB improves prompt-matched baselines by up to +10.3 macro-F1 on ECC and +5.5 on GoEmotions. Combined with PASTA-style external supervision, it achieves up to +13.84 accuracy points and +12.51 macro-F1 points, while mechanistic analyses trace the gains to sparse, emotion-enriched heads and show that layer proximity affects additivity.
[NLP-83] OpCredit: Operation-Aware Policy Learning for Agent ic Knowledge Graph Question Answering
【速读】： 该论文旨在解决生成式知识图谱问答（KGQA）中轨迹级奖励信号模糊个体操作贡献的问题，即成功路径可能包含无效操作，而失败路径也可能包含有效操作，导致难以准确评估每一步操作的价值。其核心解决方案是提出OpCredit框架，一种面向操作的策略学习方法，通过融合下游结果信用与即时状态转移信用来实现对操作的精细化评估：其中分支结果信用（Branch Outcome Credit, BOC）利用常规采样轨迹，在相同导航上下文中比较不同操作后的最终任务奖励；前沿转移信用（Frontier Transition Credit, FTC）则衡量操作引发的答案条件前沿潜力变化。这两种互补的信用信号共同监督训练过程中对应决策阶段的学习，推理阶段仅依赖已学策略与标准知识图谱接口。在ComplexWebQuestions、WebQSP和GrailQA三个基准数据集上，基于三种不同骨干模型的实验均显示了KGQA性能的持续提升。
链接: https://arxiv.org/abs/2610.21345
作者: Yucheng Li,Mingyuan Li,Xiaxia Wang,Gong Cheng
机构: Nanjing University (南京大学); University of Oxford (牛津大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Agentic knowledge graph question answering(KGQA) relies on sequences of executable operations, but trajectory-level rewards obscure their individual contributions. Successful trajectories may contain ineffective operations, while failed trajectories may contain useful ones. To assess individual operations, we propose OpCredit, an operationaware policy-learning framework that combines downstream outcome credit with immediate transition credit: Branch Outcome Credit (BOC) reuses ordinary rollouts to compare terminal task rewards observed after different operations under matched navigation contexts; Frontier Transition Credit (FTC) measures operation-induced changes in answer-conditioned frontier potential. The two complementary signals jointly supervise the corresponding decision spans during training, while inference uses only the learned policy and standard KG interface. Experiments on ComplexWebQuestions, WebQSP, and GrailQA across three backbones demonstrate consistent improvements in KGQA performance.
[NLP-84] Ranking and Partitioning of Attention Heads for Sparse and Continual Reinforcement Learning Fine-tuning
【速读】： 该论文旨在解决大语言模型在稀疏化（sparse）与持续性（continual）强化学习微调中的效率与稳定性问题。具体而言，面对计算资源受限和任务间干扰导致的性能退化挑战，研究提出通过注意力头（attention head）的梯度范数排序实现稀疏学习，并基于任务划分实现注意力头的分离更新以缓解跨任务干扰。其解决方案的关键在于：首先，利用梯度范数对注意力头进行重要性排序，仅对高排名头部进行微调，从而在保持性能的同时显著降低参数更新量；其次，通过将不同任务对应的注意力头分组并独立更新，有效减少任务间的干扰，增强模型对已有知识的保留能力。理论分析表明，梯度范数可作为训练有效性的重要指标，且稀疏、分离的头更新策略具有内在稳定性。实验结果验证了该方法在数学推理、指令遵循及代码生成等多类任务中均能以极小的参数更新量实现接近甚至媲美全参数微调的性能，同时显著提升持续学习下的知识保留能力。
链接: https://arxiv.org/abs/2610.21336
作者: Ruizhe Shi,Zaiwen Yang,Qiubo Wang,Banghua Zhu,Maryam Fazel
机构: University of Washington (华盛顿大学); Tsinghua University (清华大学); University of Hong Kong (香港大学); Amazon, Inc. (亚马逊公司)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 26 pages, comments are welcome
Abstract:We study sparse and continual reinforcement learning fine-tuning of language models through attention-head ranking and partitioning. For sparse learning, we rank attention heads by gradient norm to identify a small subset of heads that can be trained effectively. For continual learning, we partition heads across tasks to reduce cross-task interference and preserve previously learned capabilities. Theoretically, we relate gradient norm to training effectiveness, and further characterize why sparse, separated head updates can remain stable. Empirically, across math, instruction-following, and coding tasks, tuning higher-ranked heads consistently outperforms tuning lower-ranked ones while often matching denser baselines, and head partitioning improves retention while preserving adaptation to new tasks.
[NLP-85] QuasiLeak-Bench: Evaluating LLM -Based Privacy Detectors on Policy-Conditioned Quasi-Evidence Group Recovery
【速读】： 该论文旨在解决现有隐私保护评估基准在应对由多个弱标识符（quasi-identifiers）共同构成的“准证据组”（quasi-evidence group）时存在的测量空白问题。准证据组指那些单独看信息量较弱，但联合起来可推断出个体敏感属性的语义线索，其隐私泄露风险依赖于发送方设定的披露策略。当前主流隐私评测多聚焦于显式标识符、属性推断或任务驱动的信息泄露，忽视了此类隐性关联性泄露。为此，作者提出QuasiLeak-Bench——一个面向策略条件下的准证据组恢复评估基准，用于衡量检测器能否准确识别完整的准证据组，并标记违反特定披露策略的内容。其解决方案的关键在于构建一个可扩展的反向构造流程：首先筛选出弱准标识符，将其组合成经验证可联合揭示目标属性的准证据组，并生成围绕该证据的文本，再通过自动化质量控制确保标注可靠性。实验表明，现有隐私检测器在该基准上表现不佳，本地部署的基于大语言模型（LLM）的检测器显著落后于专有模型，主要因无法有效捕获大多数准证据组；然而，经过训练集微调的40亿参数模型在留出数据上的表现超越最强提示工程模型，证明该任务具备可学习性。
链接: https://arxiv.org/abs/2610.21333
作者: Yufan Zhu,Yangfan Jiang,Yang Fei,Shizheng Hou,Shubhankar Mohapatra,Xiaokui Xiao
机构: School of Computing, National University of Singapore, Singapore; Agency for Science, Technology and Research (A*STAR), Singapore
类目: Computation and Language (cs.CL); Cryptography and Security (cs.CR)
备注: 37 pages, 15 figures, 25 tables
Abstract:Natural language messages often contain quasi-identifiers that are weak in isolation but can jointly reveal an individual’s private attributes. We call such a jointly revealing set of cues a quasi-evidence group. Existing privacy benchmarks focus on explicit identifiers, attribute inference, or task-driven disclosures, leaving inference from quasi-evidence groups largely unmeasured. In practice, whether exposing such a quasi-evidence group violates privacy also depends on sender-specified disclosure policies. We introduce QuasiLeak-Bench, a benchmark for policy-conditioned quasi-evidence group recovery, which evaluates whether detectors recover complete quasi-evidence groups and flag the disclosures the given policy prohibits. To obtain reliable ground truth at scale, a reverse-construction pipeline first selects weak quasi-identifiers, composes them into quasi-evidence groups verified to jointly reveal a target attribute, and generates documents around this evidence, followed by automated quality control. Experiments show that existing privacy detectors transfer poorly to QuasiLeak-Bench, and locally deployable LLM-based detectors fall far behind proprietary models, mainly because they miss most quasi-evidence groups altogether. However, a 4B model fine-tuned on our training split surpasses the strongest prompted model on held-out documents, showing that the task is learnable.
[NLP-86] Hybrid Generative-Tagging Decoding for Efficient Product Attribute-Value Extraction EMNLP2026
【速读】： 该论文旨在解决电子商务中构建高质量商品目录时属性-值（attribute-value）提取的效率与覆盖率瓶颈问题。现有方法主要依赖生成式（generative）或标注式（tagging）策略：标注模型虽速度快且具备强上下文对齐能力，但受限于预定义标签空间，无法处理需重格式化或罕见属性名的值；而生成式模型虽能灵活生成任意属性名与值，却因自回归特性导致推理延迟高，且难以覆盖稀有属性。本文提出一种混合生成-标注解码（hybrid generative-tagging decoding）架构，其核心在于将两种范式有机结合——通过标注头直接从源文本中抽取可复制的常见属性值，使生成头专注于处理罕见属性、非可复制值（如隐含或标准化后的值），从而实现优势互补。该方案在vLLM推理框架中集成后，显著提升属性覆盖率10%，正确属性-值对提取量增加63%且无精度损失，同时降低44%的延迟，有效平衡了准确性、覆盖率与推理效率。
链接: https://arxiv.org/abs/2610.21329
作者: Keshav Kolluru,Manas Gupta,Maroof Ahmad,Suchitra Sathyanarayana
机构: Amazon(亚马逊)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 (Industry Track)
Abstract:Attribute-value extraction is important for building high-quality product catalogs in e-commerce. Existing systems either rely on generative or tagging strategies for extracting the attribute names and their corresponding values. Tagging models directly label spans in the source, making them fast and grounded, but they cannot produce values that require reformatting or generate rare attribute names outside their label space. Generative models avoid these drawbacks by generating both attribute names and values token by token, but this auto-regressive nature makes them slow and limits attribute coverage. We introduce hybrid generative-tagging decoding, a novel model architecture that combines tagging source spans and open-ended generation. For frequent attributes whose values are copyable, a tagging head extracts values directly from the source. This lets the generative head focus on the remainder - rare attributes outside the tagger’s label space and non-copyable values, including implicit or normalized values. Integrated into vLLM inference, the hybrid decoder raises attribute coverage by 10% and extracts 63% more correct attribute-value pairs at no precision cost, while reducing latency by 44%.
[NLP-87] DocPilot: Interactive Agent Orchestration and Training for Long Visual Document Understanding
【速读】： 该论文旨在解决多模态长文档问答中模型难以在跨页、跨模态的稀疏证据中进行精准定位，且在单次推理过程中易忽略关键信息、缺乏有效跨模态推理能力的问题。现有方法受限于预设工作流或粗粒度操作，导致文档推理的灵活性与细粒度控制不足。其解决方案的关键在于提出DocPilot——一种支持表达性编排的智能体框架，通过页面级导航与细粒度证据定位实现证据搜寻的高效协同，并融合局部视觉分析、基于代码的数值推理及跨模态推理以增强多模态理解能力。为使小型语言模型（SLMs）具备上述复杂推理能力，提出基于课程学习的智能体训练范式，从单一能力学习逐步过渡到由数据驱动的推理轨迹与可执行交互路径监督下的自适应编排学习。实验表明，基于前沿视觉-语言模型（VLM）的DocPilot在MMLongBench-Doc、DocBench、FinRAGBench-V和MADQA等多个基准上均取得显著性能提升，而仅40亿参数的DocPilot-4B模型亦能媲美更大规模模型，展现出卓越的泛化能力与有效性。
链接: https://arxiv.org/abs/2610.21327
作者: Zhuowen Liang,Xiaotian Lin,Yuyu Luo,Haixun Wang,Nan Tang
机构: The Hong Kong University of Science and Technology (Guangzhou); EvenUp, USA
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 25 pages, 13 figures, 9 tables
Abstract:Multimodal long-document question answering requires models to seek sparse evidence across pages and modalities, understand heterogeneous document content, and reason over the collected evidence. Even advanced vision-language models (VLMs) can overlook sparse evidence and struggle with cross-modal reasoning in a single pass. Existing document agents further introduce multi-step interaction, but some still rely on predefined workflows or coarse-grained actions, limiting the flexibility and granularity of document reasoning. We introduce DocPilot, an agentic document framework that supports expressive orchestration and makes complex document solving learnable by small language models (SLMs). DocPilot orchestrates evidence seeking through page-level navigation and fine-grained evidence localization, and integrates local visual inspection, code-based numerical reasoning, and cross-modal reasoning for multimodal understanding. To transfer these capabilities to small models, we propose curriculum-based agentic training that progresses from single-ability learning to adaptive orchestration through supervision from data-grounded reasoning traces and executable interaction trajectories. Across MMLongBench-Doc, DocBench, FinRAGBench-V, and MADQA, DocPilot achieves strong gains with frontier VLM backbones, while DocPilot-4B delivers competitive performance against substantially larger models, demonstrating strong effectiveness and generalization across diverse long-document reasoning settings. The code is available at this https URL.
[NLP-88] ZoomCap: Evidence-Routed Visual Zooming for Recoverable Visual Information Transfer
【速读】： 该论文旨在解决生成式视觉描述（generative captioning）中因文本长度与质量混淆而导致的细粒度视觉信息遗漏问题，尤其在小物体、文本内容及空间关系捕捉方面表现不足，且易产生缺乏依据的确定性断言。其核心解决方案是提出一种基于“可恢复视觉信息传递”的细粒度描述框架——ZoomCap，通过教师模型在有限观测预算下主动获取缺失证据，并将其转化为可复用的监督信号。关键创新在于：利用草稿描述引导条件化知识检索、针对光学字符识别（OCR）敏感的阅读策略、覆盖间隙恢复机制以及自适应视觉缩放，仅对最复杂的候选区域进行深度裁剪，结合基于语义锚点的几何定位推导空间关系，并融合非重复证据以提升准确性。此外，引入FactCycle评估机制，通过图像真实性和重建可行性两个盲锚对生成断言进行事实性评分，实现对虚构内容与生成损失的逐条分离评估。实验表明，相较于基线模型，ZoomCap在人类偏好评分上达到81.9%（优于Qwen3.5-397B-A17B），每图像转移断言数达52.4条（远超ScaleCap的31.5条），并在持续预训练和医疗视觉问答任务中显著提升性能，验证了其高质量、高保真视觉信息迁移的有效性。
链接: https://arxiv.org/abs/2610.21320
作者: Hao Shi,Jiafeng Wang,Yuzhuo Fu,Tao Zhang,Jian Bai,Juelu Zhang,Qing Cui,Chao Huang
机构: Ant Group(蚂蚁集团); Zhejiang University(浙江大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Multimedia (cs.MM)
备注: 21 pages, 10 figures, 11 tables; technical report. ZoomCap-450K is publicly available at this https URL
Abstract:Detailed captions transfer visual information into vision-language model supervision, but quality is easily conflated with length: long single-pass outputs still miss small objects, text, and spatial relations or promote uncertain observations into confident claims. We formulate fine-grained captioning as recoverable visual information transfer, in which a teacher acquires missing evidence under a finite observation budget and converts it into reusable supervision. ZoomCap uses a draft caption to route conditional knowledge retrieval, OCR-sensitive reading, coverage-gap recovery, and adaptive visual zooming, allocating deeper crops only to the most complex eligible region, deriving spatial relations from grounded geometry, and fusing non-duplicate evidence. We also introduce FactCycle, which scores caption-derived atomic claims on two blind anchors–source image (truthfulness) and reconstruction (realizability)–separating fabrication from generator loss claim by claim; human and multi-judge comparisons complement this fact-level view. Against Qwen3.5-397B-A17B captions, ZoomCap achieves an 81.9% human tie-adjusted preference score with a teacher over 10x smaller, and FactCycle counts 52.4 transferred claims per image vs. 31.5 for ScaleCap. The evidence also transfers: in continual pre-training, ZoomCap-450K improves the 11-benchmark average by 1.31-2.38 points over the strongest competing corpus and Ling-Tiny-VL-v3 gains 1.4 points on a production document/OCR monitor suite (1.5 on perception) at 60B tokens; in medical pre-training, 2.64M ZoomCap-captioned images improve five VQA benchmarks by 3.04 points (vs. 1.76 for same-image Gemini-3.1-Pro captions).
[NLP-89] PocketTSLM: Deploying an On-Device Soft-Prompt Time-Series Language Model for Privacy-Preserving Smartphone Reasoning ALT
【速读】： 该论文旨在解决移动健康应用中生理时间序列数据（physiological time-series data）在设备端处理时的代表性失配问题：现有移动系统通常将连续的生理信号流简化为摘要或文本形式，导致时间结构信息丢失，难以支持需要精确时间对齐的个人健康分析任务。同时，云端部署的时间序列语言模型虽可缓解设备算力限制，但存在敏感数据外传、延迟高及查询成本昂贵的问题。而传统的本地化模型通过统计特征聚合或原始采样序列文本化的方式压缩数据，又面临时间结构破坏或超出设备上下文与内存预算的困境。本文提出PocketTSLM，一种基于OpenTSLM SoftPrompt的原生移动端部署方案，其关键创新在于将时间序列窗口以编码嵌入（embeddings）形式直接注入模型，而非将其转化为文本标记（tokenized as text），从而完整保留时间结构并显著降低上下文使用量。实验表明，软提示路径在相同骨干网络下仅需约1/12的时序标记数，大幅减少提示预填充（prompt-prefill）开销；在30秒脑电图（Sleep-EDF）和10秒12导联心电图（ECG-QA）片段上均实现端到端设备内推理，且无需传输任何健康数据，验证了其在隐私保护、效率与实用性方面的优势。
链接: https://arxiv.org/abs/2610.21308
作者: Max Rosenblattl,Nikhil Krishnaswamy,Leon Nissen,Manuel Milling,Nicolas Zumarraga,Ning Wang,Thomas Kaar,Robert Jakob,Björn W. Schuller,Elgar Fleisch,Patrick Langer,Paul Schmiedmayer
机构: University of St. Gallen (圣加仑大学); Technical University of Munich (慕尼黑工业大学); Google(谷歌); OpenAI; Meta; Stability.AI
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 15 pages, 3 figures. A version of this paper was accepted at the Workshop on Deploying AI in Healthcare (DAIH) at COLM 2026. Code: this https URL
Abstract:Consumer phones and wearables increasingly collect physiological time-series data, yet most mobile systems treat these streams as summaries or text rather than as temporal evidence. This representation is poorly matched to personal health systems that need to retrieve, route, or summarize sensor evidence while preserving temporal structure. Cloud-hosted time-series language models avoid smartphone limits, but they repeatedly move sensitive physiology off the device and add latency and per-query compute cost. Mobile model alternatives that aggregate time-series data into statistical features or serialize raw samples as text either compress temporal structure or exhaust the context and memory budgets of small on-device models. We present PocketTSLM, a mobile-native deployment of OpenTSLM SoftPrompt, the soft-prompt variant of the OpenTSLM time-series language model. PocketTSLM runs fully on a consumer smartphone by injecting encoded time-series windows as embeddings instead of tokenizing time series as text. A numerical parity protocol verifies identical greedy-decoded token sequences between the Swift port and a Python reference using Apple’s MLX machine-learning framework. Answers also match the PyTorch reference, although rationale wording can differ. An on-device context-economics benchmark compares the soft-prompt path with a text-serialized baseline on the same backbone. In our benchmark format, soft-prompting uses approximately 12 times fewer time-series tokens and reduces prompt-prefill cost. PocketTSLM processes 30-second Sleep-EDF electroencephalography clips for sleep staging and a 1,000-sample (10 s) 12-lead electrocardiogram clip from ECG-QA. In comparison, text-serialized baselines exceed the phone’s context or memory budget at longer time-series windows. All inference runs entirely on the device, with no health data transmitted during inference.
[NLP-90] VALCON: Benchmarking and Enhancing Value Consistency in Large Language Models via Dual-Role Variation EMNLP2026
【速读】： 该论文旨在解决当前大语言模型（LLM）价值对齐评估中过度依赖输出合规性而忽视模型在生成者与评价者角色切换时价值一致性（value consistency）的问题。随着现代LLM代理和对齐流程普遍采用同一模型同时承担生成行为与评估判断的双重角色，这种跨角色的一致性变得尤为关键。其解决方案的核心在于提出一种基于角色驱动的数据构建框架VALCON，通过双角色变异将价值一致性操作化为三个基准测试套件（VAB-Consistency、VAB-Boundary、VAB-Challenge），系统性地将模型的生成行为与其评价判断在相同原则下进行配对，从而合成能够暴露隐含价值偏差的目标评估数据。实验在28个LLM上的结果表明，普遍存在跨角色不一致现象：即使模型生成内容高度合规，其自我评价却常无法与自身生成行为保持一致。为进一步理解并缓解该问题，研究通过机制探针揭示了与行为不对称相关的角色依赖表征解耦现象，并据此提出一种表征导向的干预策略，有效缓解了一致性失效问题。本工作首次将价值一致性确立为独立的对齐维度，并构建了完整的“评估—分析—缓解”闭环体系。
链接: https://arxiv.org/abs/2610.21304
作者: Junru Wu,Jingting Zheng,Yuqi Ren,Deyi Xiong
机构: TJUNLP Lab, School of Computer Science and Technology, Tianjin University(天津大学); Yuanhui AI
类目: Computation and Language (cs.CL)
备注: 24 pages, 16 figures. Accepted to Findings of the Association for Computational Linguistics: EMNLP 2026. Code and data: this https URL
Abstract:Current value alignment evaluations for large language models (LLMs) heavily rely on output compliance, overlooking whether LLMs maintain value consistency when alternating between generator and evaluator roles. Such consistency is increasingly important because modern LLM agents and alignment pipelines often rely on the same LLM to both generate actions and evaluate their acceptability. To mitigate this limitation, we propose VALCON, a role-driven data construction framework that operationalizes value consistency into three benchmark suites (VAB-Consistency, VAB-Boundary, VAB-Challenge) via dual-role variation. By systematically pairing an LLM’s generative behavior with its evaluative judgments under identical principles, VALCON synthesizes target evaluation data that exposes hidden value discrepancies. Experiments on 28 LLMs reveal pervasive cross-role inconsistency: models with highly compliant outputs frequently fail to align their judgments with their own generations. To understand and address this gap, we perform mechanistic probing that uncovers a role-dependent representational decoupling associated with this behavioral asymmetry. Guided by these insights, we propose a representation-directed intervention that effectively mitigates consistency failures. Our work establishes value consistency as a distinct alignment dimension and provides a complete evaluation-analysis-mitigation pipeline. Code and data are available at this https URL.
[NLP-91] meCap: A Scalable Open-Source Pipeline for Captioning Wearable Health Time-Series Data
【速读】： 该论文旨在解决多模态可穿戴时间序列数据（multimodal wearable time-series data）在标注与语义理解方面面临的挑战，即现有数据集普遍缺乏可复现、跨模态扩展且信号对齐的文本标注对。当前可用的配对语料库受限于封闭性、规模小及模态覆盖不足，而利用大语言模型进行自动标注虽可提升规模，但其输出不可确定，导致无法复现、审计或扩展。为此，本文提出TimeCap——一个开放、确定性的标题生成流水线，其核心创新在于采用三级可复用的信号提取器与跨通道合成层，直接作用于原始信号数据，全程无需引入大语言模型，从而确保结果的可复现性与可扩展性。通过仅定义新通道即可扩展至新数据模式，使TimeCap成为未来数据集构建的可复现机制。将TimeCap应用于OpenMHC（MyHeart Counts可穿戴数据集），生成了目前最大规模的开源多模态可穿戴时间序列标注数据集MHCap，涵盖19个Apple HealthKit通道、7,005,763条标注，覆盖超过90万个人日。为验证其价值，基于MHCap训练了OpenTSLM，并在健康结局预测任务中进行基准测试，结果显示，使用标题增强训练的OpenTSLM-Cap在8项任务中的29个用户级二分类评估中均优于仅依赖标签的基线模型，证明了高质量、可复现的文本标注对时间序列语言模型性能的关键提升作用。
链接: https://arxiv.org/abs/2610.21274
作者: Max Rosenblattl,Milana Gurbanova,Karl Aurel Deck,Narayan Schuetz,Thomas Kaar,Manuel Milling,Nicolas Zumarraga,Ning Wang,Kelly W. Zhang,Anders Johnson,Robert Jakob,Björn W. Schuller,Elgar Fleisch,Daniel Seung Kim,Allan Lawrie,Patrick Langer,Maxwell A. Xu,Euan Ashley,Paul Schmiedmayer
机构: 1. University of St. Gallen (圣加仑大学); 2. Institute for Digital Transformation (数字化转型研究所); 3. Swiss Federal Institute of Technology in Zurich (瑞士联邦理工学院, 苏黎世); 4. University of Augsburg (奥格斯堡大学); 5. Stanford University (斯坦福大学); 6. University of Sheffield (谢菲尔德大学); 7. ETH Zurich (苏黎世联邦理工学院); 8. Google(谷歌); 9. Meta(Meta); 10. OpenAI(OpenAI)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 46 pages, 27 figures, including supplementary material. Code: this https URL
Abstract:Time-series data generated continuously across industrial, environmental, and especially health domains produce multimodal signals whose scale and heterogeneity make their processing and interpretation particularly challenging. Time-Series Language Models (TSLMs) promise scalable interpretation of these signals, yet remain bottlenecked by a fundamental lack of text-annotated time series pairs that are signal-grounded, reproducible, and extensible across modalities and schemas. Existing paired corpora are either closed or limited in scale and modality coverage. Annotation with large language models could scale them up, but its outputs cannot be regenerated exactly, which prevents others from re-deriving, extending, or auditing the resulting corpus. To address this, we release TimeCap, an open deterministic captioning pipeline with three reusable extractor tiers and a cross-channel synthesis layer that operates directly on the signal with no language model in the loop, extensible to new schemas by writing channel definitions alone. This positions TimeCap as a reproducible mechanism for future dataset generation rather than a one-off release. Applying TimeCap to OpenMHC, the MyHeart Counts wearable dataset, yields MHCap, the largest open caption layer for multimodal wearable time-series data. MHCap spans 19 Apple HealthKit channels and 7,005,763 captions over 908,944 person-days. To showcase the impact of the corpus, we train OpenTSLM, a time-series language model, on MHCap and benchmark on health-outcome prediction. Caption-augmented training (OpenTSLM-Cap) outperforms a label-only baseline on 29 of 32 per-user binary evaluations across 8 tasks.
[NLP-92] Auditing Escalation Reliability in Healthcare LLM Agents : A Perturbation-Controlled Benchmark of Workflow-State Failures
【速读】： 该论文旨在解决可信生物医学人工智能（Biomedical AI）在证据缺失、过时、冲突或延迟等动态医疗工作流场景下，如何实现可靠决策选择的问题，而不仅限于静态输入下的准确回答。其核心挑战在于现有模型在面对不确定性时缺乏有效的风险识别与应对机制，尤其在需要主动请求澄清、放弃判断或升级处理时表现不足。解决方案的关键在于引入一种基于事件级别的评估协议，并设计一个透明的、工作流感知的控制器（workflow-aware controller），通过显式建模任务状态来提升模型在复杂临床情境中的决策可靠性。实验表明，尽管大模型在干净数据上可达到0.97的准确率，但在存在冲突的情况下准确率下降至0.78，且多数模型未能有效识别需升级的场景；而该控制器显著降低了漏报升级的比例（降至0.12–0.24），同时保持较高的特异性（0.83–0.93），并提升了5:1成本权衡下的整体效用。然而，研究也揭示了当前规则检测器在面对未见过的同义表达时召回率骤降至0，暴露出其泛化能力的局限性，强调了对路由控制机制本身进行独立泛化测试的重要性。
链接: https://arxiv.org/abs/2610.21270
作者: Zihao Zheng(Washington University in St. Louis),Junyi Yao(Washington University in St. Louis),Baichuan Li(Southern Methodist University),Jiayu Long(Washington University in St. Louis)
机构: Washington University in St. Louis (圣路易斯华盛顿大学); Southern Methodist University (南卫理公会大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 7 pages, 5 figures
Abstract:Trustworthy biomedical AI requires reliable action selection when evidence is missing, stale, conflicting, or late, not only accurate answers on static inputs. We instantiate an episode-level evaluation protocol as 400 synthetic healthcare workflow episodes: 40 hand-authored semantic seeds across four task classes, two surface parameterizations, and five controlled variants, scored on ANSWER/ASK CLARIFICATION/ABSTAIN/ESCALATE under a pre-declared 5:1 missed-versus-unnecessary escalation cost. Reference actions are single-author study-policy labels, not clinician-adjudicated ground truth. Across five open-weight configurations (1.7B–9.7B), clean decision accuracy reaches 0.97, yet uncertainty-prompted models miss 32–97% of required escalations. The stronger 9.7B configuration narrows but does not remove the gap, falling from 0.97 clean accuracy to 0.78 under conflict. Verbalized confidence is inconsistent (ECE 0.10–0.54; failure AUROC 0.48–0.77), and a development-tuned confidence-only policy yields no utility gain. A transparent workflow-aware controller improves 5:1 utility for every model and reduces missed escalations to 0.12–0.24, at specificity 0.83–0.93. However, a 30-concept lexical stress test reduces each rule detector’s recall from 1.00 on trigger phrases to 0.00 on unseen paraphrases, limiting the controller claim to this benchmark. Seed-clustered statistics, ablations, cross-task transfer, raw outputs, and manifests are packaged for audit. The results motivate explicit workflow-state evaluation while showing that apparently effective routing controls need their own generalization tests.
[NLP-93] Behavioral Compositionality of Sparse Autoencoder Representations
【速读】： 该论文旨在解决生成式模型中稀疏自编码器（Sparse Autoencoders, SAE）所提取的特征在下游任务中是否具有可加性（additivity）的问题，即这些特征的组合是否能通过简单的线性叠加准确描述模型行为。其核心挑战在于：尽管SAE通过稀疏、线性解码的方式表示激活，但实际行为变化可能包含高阶交互效应，导致仅依赖单个特征的贡献无法充分解释整体行为。为此，论文提出SAE可加性残差（SAR, SAE Additivity Residuals），用于量化在指定交互阶数以上未被解释的行为变异量。在独立特征子集采样假设下，群体SAR恰好等于高于该阶数的交互项的归一化能量，并为最优低阶近似下的联盟排序误差提供理论上限。通过将学习得到的SAE坐标与同一子空间内保持和不变的随机旋转坐标进行对比，发现前者在全部18个模型与行为设置中均表现出更低的一阶残差，表明学习到的坐标具有更强的可加性特性。然而，结果也显示这些坐标并非完全可加：引入成对交互项后，在所有场景下均提升了保留联盟排序的性能，且优于使用随机选取三阶项的参数匹配模型。在联合操控实验中，成对建模使排名保真度从0.5874提升至0.7790，排序误差则由0.1444降至0.0606。综上，该研究揭示了SAE坐标虽具备相对较强的组合性行为描述能力，但要实现精确的联合解释与干预，仍需考虑超越单特征效应的高阶交互作用。
链接: https://arxiv.org/abs/2610.21261
作者: Haotian Ma,Ziye Mei,Frederic Sala
机构: University of Wisconsin–Madison(威斯康星大学麦迪逊分校)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 29 pages, 9 figures. Code: this https URL
Abstract:Sparse autoencoders (SAEs) represent model activations with sparse, linearly decoded features, but the downstream effects of those features need not compose additively. We introduce SAE Additivity Residuals (SAR), which measure the behavioral variation left unexplained above a chosen interaction order. Under independent feature subset sampling, population SAR is exactly the normalized energy of interactions above that order and bounds coalition ordering error for the optimal low order approximation. To distinguish compositionality associated with the SAE representation from that induced by downstream computation, we compare learned SAE coordinates with sum preserving random rotations of the same selected subspace and find lower first order residuals for the learned coordinates across all 18 model and behavior settings. The coordinates are nevertheless not fully additive: pairwise terms improve held out coalition ordering in every setting and outperform parameter matched models using randomly selected third order terms. In a combined steering experiment, pairwise modeling raises rank fidelity from 0.5874 to 0.7790 and reduces ordering error from 0.1444 to 0.0606. Together, these results show that SAE coordinates support comparatively compositional behavioral descriptions while faithful joint explanations and interventions still require interactions beyond individual feature effects. Code is available at this https URL.
[NLP-94] Isolating a Language Models Judgment with a Label-Free Subspace
【速读】： 该论文旨在解决大语言模型在执行判断任务时，其输出结果对指令表述敏感的问题——即相同输入在不同指令引导下产生不一致甚至不可靠的最终判断，这可能源于表达方式的变化而非内部判断的改变。其核心解决方案是通过分析同一输入在不同指令下的隐藏状态向量（hidden-state vectors），识别出一个低维的“表达子空间”（expression subspace），该子空间主导了因指令变化导致的输出差异。研究提出一种无需标注基准数据即可定位该子空间的方法，并通过从线性预测器中移除该子空间的贡献，从而获得一个与表达解耦的“孤立判断”（isolated judgment）。实验证明，这一孤立判断在正确性、蕴含关系和安全性等任务上均保持高度可靠性，即使在未用于子空间识别的指令条件下仍有效，且当判断问题被否定时，预测性能显著提升（如蕴含任务的AUROC从0.504升至0.825，安全任务从0.480升至0.891）。此外，研究发现判断信号可早在模型早期层中读取，早于表达子空间开始干扰预测的位置。这些结果表明，可通过分离表达与判断过程，恢复模型内部稳定的判断能力，即便其最终输出不可靠。
链接: https://arxiv.org/abs/2610.21256
作者: Chaemin Jang,Donggyu Lee,Jeesoo Kim,Dongman Lee,Jihee Kim
机构: Korea Advanced Institute of Science and Technology (KAIST)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Language models are often asked to make judgments: whether a proposed answer is correct, whether one statement follows logically from another, or whether a response is safe. Yet changing the instructions that specify how the judgment should be expressed can make the final answer less reliable, even for the same input. We ask whether this reflects a change in internal judgment or a failure to express it. Comparing hidden-state vectors for the same input under different instructions, we find that their differences concentrate in a low-dimensional expression subspace. We identify this subspace without benchmark labels and remove its contribution from a linear predictor to obtain an isolated judgment. Across correctness, entailment, and safety, the isolated judgment remains reliable under instructions not used to identify the subspace, even when the model’s final answer and the original predictor degrade. When the judgment question is negated, isolation raises the predictor’s AUROC from 0.504 to 0.825 for entailment and from 0.480 to 0.891 for safety. We also find that judgment is readable in earlier layers, before the expression subspace begins to disrupt the predictor. Together, these results provide a way to separate judgment from expression and recover internal judgments even when the model’s final answers are unreliable.
[NLP-95] LLM -Based Weak Annotation for Cybersecurity Tweet Classification: A Teacher-Student Framework
【速读】： 该论文旨在解决网络安全威胁情报（Cyber Threat Intelligence, CTI）领域中因社交媒体上海量安全讨论导致的标注数据稀缺问题，即如何在缺乏大规模人工标注的情况下高效构建用于训练领域特定预训练语言模型（Pre-trained Language Models, PLMs）的高质量训练数据。其核心解决方案是提出一种“教师-学生”弱监督框架，其中选定的大语言模型（Large Language Models, LLMs）作为“教师”模型，为未标注的网络安全推文生成弱标签，从而实现无需依赖大量人工标注即可训练出高性能的领域专用PLM。实验结果表明，仅使用LLM生成标签训练的PLM在各类架构下均能保持超过0.92的F1分数，性能与完全由人工标注训练的模型相差不超过3个百分点，验证了基于LLM的弱标注方法在资源受限场景下构建高质量网络安全训练数据的有效性与可行性。
链接: https://arxiv.org/abs/2610.21254
作者: Ferry Afit Kurniawan,Kazushi Okamoto,Koki Karube,Kei Harada,Atsushi Shibata
机构: The University of Electro-Communications(电气通信大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Cyber threat intelligence (CTI) relies on the timely identification of emerging threats, enabling organizations and security analysts to support situational awareness and decision-making. Social media platforms such as X (formerly Twitter) have become valuable CTI sources because vulnerabilities and attack activities are often discussed there before appearing in formal reports. However, the rapid growth of cybersecurity discussions on social media has created more content than domain experts can realistically review and annotate, making labeled data a major bottleneck for developing supervised learning approaches. While large language models (LLMs) have shown promise as weak annotators in other domains, their effectiveness for cybersecurity tweet classification and downstream training of domain-specific pre-trained language models (PLMs) remains underexplored. To address this challenge, we propose a Teacher-Student weak supervision framework in which a selected LLM serves as a Teacher to generate weak labels for unlabeled cybersecurity tweets, enabling the training of cybersecurity-specific PLMs without large-scale manual annotation. Experimental results show that PLMs trained solely on LLM-generated labels consistently achieve F1 scores above 0.92, remaining within 3 percentage points of fully human-supervised models across all evaluated architectures. These findings suggest that LLM-based weak annotation offers a practical alternative for constructing cybersecurity-specific training data when expert annotation resources are limited.
[NLP-96] FGSS-BO: Feedback-Guided Hypothesis Management for LLM -Based Bayesian Optimization
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）引导的贝叶斯优化（Bayesian Optimization, BO）中存在的一致性与可解释性不足问题，即模型生成的搜索方向及其修正过程往往隐含且难以追溯。其核心解决方案是提出反馈引导的搜索状态（Feedback-Guided Search State, FGSS），构建一个结构化的假设管理接口，将基于语言的指导、观测证据、可执行的参数干预以及对比支持度更新进行系统化关联。在该框架下，一个LLM反射器负责生成与修订假设，而确定性校验器则负责验证假设、维护假设身份及测试记录，确保状态演进的可追踪性。该动态演化状态用于指导候选解的生成与优先级排序。实验表明，在两个回溯性协议/媒体任务中，基于LLAMBO实现的FGSS-BO在样本均值最优解和搜索累积曲线下面积（AUC）上均优于匹配的LLAMBO基线及其他非LLM基线；消融实验显示，自然语言反思与仅期望改进（Expected Improvement, EI）策略的组合具有混合影响。六组直接训练超参数优化任务进一步揭示了任务依赖性的迁移能力，其中在糖尿病支持向量机（SVM）任务中表现最为显著。该方法不仅暴露了假设生成、执行与跨轮次修订的全过程，还在特定设置下实现了搜索性能提升，显著增强了传统LLM-BO方法的透明性与可控性。
链接: https://arxiv.org/abs/2610.21246
作者: Hao Chen,Ziyang Ling,Ronald X. Xu,Mingzhai Sun
机构: School of Biomedical Engineering; School of Engineering Science, University of Science and Technology of China (中国科学技术大学), Suzhou, China
类目: Computation and Language (cs.CL)
备注: 21 pages, 8 figures
Abstract:Large language model (LLM)-guided Bayesian optimization (BO) uses past observations to guide search, but proposed directions and their revision can remain implicit. We propose Feedback-Guided Search State (FGSS), a structured hypothesis-management interface connecting language-based guidance to observed evidence, executable parameter interventions, and comparative support updates. An LLM reflector proposes and revises hypotheses; a deterministic curator validates them and maintains identities and test records. The evolving state guides candidate construction and priorities. Instantiated on LLAMBO, FGSS-BO retains its surrogate and default expected improvement selection, with a bounded testing policy. On two retrospective protocol/media tasks with nearest-neighbor response oracles, it attains higher sample mean final best and search AUC than the matched LLAMBO baseline and evaluated non-LLM baselines. Prose-reflection and EI-only controls show mixed component effects. Six direct-training hyperparameter-optimization tasks show task-dependent transfer, strongest on diabetes SVM; exact-grid and reaction controls examine other settings. FGSS exposes hypotheses, their execution, and cross-round revisions for inspection while supporting search gains in selected settings.
[NLP-97] System1-Arena: Which Decisions Can Be Left to System 1?
【速读】： 该论文旨在解决生成式 AI 在决策任务中，系统1（System-1）模型与系统2（System-2）模型在适用场景上的边界问题，即哪些决策仅需快速、基于概率的判断（如分类或筛选），而哪些则必须依赖生成式推理过程（如多步规划或复杂推理）。其核心解决方案是提出 System1-Arena——一个涵盖 92,000 个问题、跨越 105 项任务、源自 268 个公开数据源的综合性基准测试平台，包含文本分类与知识、应用型决策、评估与控制、智能体与动作四类任务，并集成 138 个闭环环境（如网页购物、代码控制、机器人操作、游戏与自动驾驶），支持实时动态世界中的决策评估。通过在该基准上对 30 个模型（包括闭源、开源及系统1模型）进行对比，研究发现：大型系统1模型在非代理型任务中可达到甚至超越快速闭源模型的准确率，且概率校准更优、延迟仅为后者的四分之一；而生成式推理模型虽最准确（约是快速模型成本的70倍），但代价高昂。在闭环环境中，大型开源模型表现最佳（成功率最高达32.9%），而小型系统1模型几乎无法成功。因此，关键结论在于：系统1模型适用于大多数即时性、低复杂度决策，而高复杂度、长周期的闭环任务仍需系统2式的生成式推理能力。
链接: https://arxiv.org/abs/2610.21244
作者: Tianyu Chen,Xinyang Liu,Ruyi An,Zhendong Wang,Mingyuan Zhou
机构: The University of Texas at Austin(德克萨斯大学奥斯汀分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 40 pages. Code: this https URL ; leaderboard: this https URL ; dataset: this https URL
Abstract:Large language models (LLMs) often make discrete decisions, such as flagging phishing or escalating to a human. System-1 models, such as Jev and the OpenAI Decisions API, return probabilities over options in a single forward pass, while System-2 approaches generate reasoning steps. Which decisions can System-1 models handle, and which require generative reasoning? We introduce System1-Arena, a benchmark that gives all models the same decision requests and scores their probabilities. It contains 92k questions across 105 tasks from 268 public sources, in four groups: text classification and knowledge, applied decisions, evaluation and control, and agent and action. It also includes 138 closed-loop environments, from web shopping and SWE-bench coding control to robotics, games, and driving; in real-time mode, the world evolves while a model decides. We evaluate 30 models: 7 closed LLMs, 13 open LLMs read at the answer token, and 10 System-1 models. The largest System-1 models match or exceed fast closed LLMs in non-agentic accuracy, with better-calibrated probabilities at one-quarter of the latency. A reasoning LLM is the most accurate, at about 70x the cost of a fast LLM. In closed loops, large open LLMs succeed most often (up to 32.9%); small System-1 models rarely succeed.
[NLP-98] Representation Redistribution in Sparse Autoencoders
【速读】： 该论文旨在解决生成式语言模型中稀疏自编码器（Sparse Autoencoders, SAEs）在提示（prompt）变换下特征激活分布重分配所带来的可重复性与下游任务失败之间的关联性问题。其核心挑战在于：尽管某些提示变换会引发特定特征响应的显著变化，但这些变化是否具有可复现性、是否真正影响下游决策尚不明确，且传统方法常将特征响应变化等同于系统失效，导致误判。为此，作者提出重分配敏感度评分（Redistribution Sensitivity Score, RSS），通过识别在多个提示变换中以一致方向反复出现的大幅特征响应，精准筛选出具有稳定重分配特性的关键特征支持集。关键创新在于将“特征响应变化”与“读出权重贡献”及“决策失败”进行解耦分析，基于固定线性读出建立精确分解模型，揭示响应幅度、读出系数、方向对齐度以及干净边际（clean margin）共同决定下游后果。实验表明，在27个预训练SAE模型上，结合RSS支持集侵蚀与完整干净边际，仅使用中位数20和19个特征即可实现0.935和0.958的重复性失败AUROC，接近全集侵蚀效果（0.965和0.964），证明了RSS能有效识别并量化特征重分配对下游读出的影响，同时避免将表征变化直接等同于失败，为安全可控的模型解释提供了更精细的评估框架。
链接: https://arxiv.org/abs/2610.21240
作者: Ziye Mei,Haotian Ma
机构: University of Wisconsin–Madison(威斯康星大学麦迪逊分校)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 35 pages, 15 figures. Code: this https URL
Abstract:Sparse autoencoders (SAEs) decompose language model activations into sparse features, but prompt transformations designed to preserve the underlying request can redistribute activity across them. A feature change observed on one transformed prompt need not recur, and even a recurring change need not affect a downstream computation. We introduce the redistribution sensitivity score (RSS), which identifies large feature responses that recur in a consistent direction, and evaluate the resulting supports on variants excluded from selection. Across 27 pretrained SAEs spanning eight language models, the tested prompt transformations produce a reproducible redistribution profile on these excluded variants. We then separate recurring representation change from signed readout contribution and decision failure. For a fixed linear readout, an exact decomposition shows how response magnitude, readout coefficients, directional alignment, and the clean margin determine downstream consequences. In two frozen safety readouts, combining RSS support erosion with the full clean margin yields recurrent failure AUROCs of 0.935 and 0.958 using median supports of 20 and 19 features, compared with 0.965 and 0.964 using full erosion. RSS therefore provides a way to identify recurring sparse representation changes and quantify their contribution to a downstream readout without equating representation change with failure. Code is available at this https URL
[NLP-99] What Should We Remove? Evaluating the Reliability and Costs of RLVR Data Decontamination
【速读】： 该论文旨在解决生成式 AI（Generative AI）在数学领域强化学习中因训练数据与评估基准存在重叠而导致的可靠性问题，尤其关注数据清洗策略在去除重叠样本时的准确性和数据损失成本。其核心解决方案在于系统性评估八种基于精确匹配、词汇匹配和语义匹配的数据清洗政策，并通过双盲独立评审的594对训练-基准样本对进行严格验证。研究发现，在余弦相似度阈值0.80下，采样检测显示88%的移除项为误删（即假阳性），且预注册的语义阈值范围（0.80–0.98）均未达到90%保守精度目标。相比之下，现有由人工裁判裁定的清洗策略在保留集上实现了98.1%的解析精度与98.2%的候选框架召回率，且在6个采样移除案例中无确认误删，展现出更优的精确率-召回率平衡。然而，该结果并不意味着零风险，且召回率未涵盖候选框架外的潜在污染。研究强调应以清洗可靠性与代价评估为核心，而非仅关注重叠数量，并公开了标注结果、各策略决策及覆盖率统计，以支持对数据清洗质量的全面评估。
链接: https://arxiv.org/abs/2610.21229
作者: Esther Xin
机构: 独立研究者(Independent Researcher)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 13 pages, 4 figures, 4 tables,code and data: this https URL
Abstract:Reinforcement learning with verifiable rewards (RLVR) math corpora are routinely screened for benchmark overlap, but the reliability and data-loss costs of cleaning policies remain underexamined. We audit sixteen open RLVR math corpora comprising 1,673,514 corpus-specific training items against six benchmarks containing 1,949 questions. We evaluate eight policies based on exact, lexical, and semantic matching using 594 independently and blindly dual-reviewed training-benchmark pairs. Calibration and held-out sets have disjoint benchmark questions and training-example families. A separate review of all 83 candidate matches for 80 sampled training items measures item-level false removal. At cosine threshold 0.80, 52 of 59 sampled removals (88%) were confirmed false removals; this is not a demonstrated population rate. No semantic threshold in our preregistered grid (0.80-0.98) meets the 90% conservative-precision target on calibration. The existing judge-adjudicated policy achieves 98.1% resolved precision and 98.2% candidate-frame recall on held-out corpora. It has zero confirmed false removals among six sampled removals, with one unresolved case; this does not establish zero risk. Within the evaluated candidate frame, this policy offers a more favorable precision-recall balance than the semantic-threshold alternatives tested. Recall does not measure contamination outside the candidate frame. We release reviewed labels, per-policy decisions, and coverage accounting to support evaluation of cleaning reliability and cost rather than overlap counts alone.
[NLP-100] KBS: A Korean Benchmark Suite for Foundation Model Evaluation
【速读】： 该论文旨在解决现有基准测试在适配韩语（Korean）场景时普遍依赖翻译方法所带来的语言文化失真与评估偏差问题。为实现更准确、可靠的韩语模型评估，研究提出韩国基准套件（Korean Benchmark Suite, KBS），包含Ko-ARC、Ko-GSM8K、Ko-EQ-Bench、Ko-WinoGrande、Ko-LAMBADA和Ko-IFEval六个子任务，覆盖九千余项数据，涵盖推理、数学、常识、指代消解、语言建模及综合能力等六类核心能力。其解决方案的关键在于采用多维度构建策略——结合评审、本地化处理、直接韩语内容生成与结构重设计，避免单纯依赖机器翻译的局限性，从而确保数据在语义、语法与文化背景上的准确性。此外，评估协议充分考虑模型访问方式、提示格式与评分规则的差异，揭示了任务构建方式、评分标准、提示工程以及模型微调后训练对韩语评估结果的显著影响。研究强调，韩语模型性能评价应同时报告最终得分及其具体实验设置，以提升评估透明度与可复现性。
链接: https://arxiv.org/abs/2610.21227
作者: Yeonkyoung So,Jongmin Kim,Sungmok Jung,Gyuseong Lee,Sangho Kim,Jongyeon Park,Joonhak Lee,Seho Pyo,Gyeongje Cho,Seorin Kim,Jisoo Kim,Suyoung Park,Hyunji M. Park,Yelim Ahn,Yeongho Seo,Jaejin Lee
机构: Seoul National University (首尔国立大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:We present a Korean Benchmark Suite (KBS) comprising Ko-ARC, Ko-GSM8K, Ko-EQ-Bench, Ko-WinoGrande, Ko-LAMBADA, and Ko-IFEval, covering six capabilities across 9,396 items. The suite combines review, localization, direct Korean construction, and redesign for Korean, avoiding a translation-only approach to benchmark adaptation. We evaluate Korean-developed models, multilingual open models, and closed API models under protocols that account for model access, prompt format, and scoring differences. Results show that Korean benchmark scores depend on task construction, scoring rules, prompting, and model post-training. These findings suggest that Korean evaluation should report both final scores and the settings used to obtain them.
[NLP-101] Entertain then Cancel: How Thinking Makes Reasoning Models Read Some and Or Literally
【速读】： 该论文旨在解决生成式语言模型在处理日常语义推理（如“一些学生通过了”隐含“并非全部通过”）时，其推理过程随思考时间延长所产生的话语含义变化问题。核心挑战在于：尽管当前模型允许用户调节推理时长，但缺乏对额外思考如何影响语义隐含意义（implicature）的系统性理解。本文的关键解决方案是，在保持模型参数不变的前提下，通过控制推理时长并截取中间推理路径，揭示模型在不同思考阶段对逻辑词汇（如“some”“or”）的理解演变。研究发现，随着推理深入，模型逐渐将“some”和“or”等词解读为严格逻辑意义（即“非全”“非两者皆”），从而消解其自然语言中的隐含意义；而上下文驱动的间接推理能力则得到增强。追踪分析表明，模型先短暂采纳富含语用信息的“丰富读法”（enriched reading），随后在后续推理中将其修正。由此提出一种基于推理时的探索性补救策略——在早期阶段输出答案，可有效恢复被取消的语用隐含，同时不损害字面准确性。因此，评估推理模型应不仅关注准确性，更需考察其语义理解机制。
链接: https://arxiv.org/abs/2610.21204
作者: Zhizhuo Kou,Jian Yang,Pengfei Hao,Junyu Luo,Jingshu Peng,Sirui Han,Yike Guo
机构: The Hong Kong University of Science and Technology (香港科技大学); Peking University (北京大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:When a speaker says ``some students passed’', listeners infer that not all of them did. Reasoning models now let users decide how long a model thinks before it answers, but we do not know what extra thinking does to such everyday inferences. We study this directly: we keep a model’s weights fixed, change only how much it thinks, and record which reading it chooses. A second tool cuts the model’s own reasoning at successive points and asks for an answer, showing where the reading changes. Thinking makes models read logical words as logic: it cancels the implicatures of \emphsome (\emphnot all) and \emphor (\emphnot both) in most tested open reasoning models and in GPT-5, while context-driven inferences such as indirect answers improve. The trace shows why: models first entertain the enriched reading and cancel it later. Answering from that early reading, an exploratory inference-time remedy, recovers these implicatures without losing literal accuracy. Evaluations of reasoning models should measure how they interpret, not only how accurate they are.
[NLP-102] Who May Release an Agents Promise? Standing over Commitments in Multi-Party Chat
【速读】： 该论文旨在解决在团队协作场景中，由助理（assistant）代表团队作出承诺后，谁有权取消该承诺这一关键治理问题。其核心挑战在于防止未经授权的承诺释放（unauthorized release）和承诺被强制执行（override）两类错误：前者指系统响应无权限者指令而解除承诺，后者指即使承诺已被解除仍继续执行。研究发现，尽管较小规模的开源模型（如Qwen3.5-9B）在处理授权释放时表现较差（69%出现越权），但更大规模模型（如Qwen3.8-27B）更倾向于仅凭同级或旁观者的指令就放弃承诺（89%），且几乎所有模型都会响应“最终可以了”这类非授权信号。实验表明，仅告知规则或债权人身份可抑制未经授权释放，但无法阻止覆盖执行；唯有明确知晓承诺是否仍有效，才能显著降低覆盖行为。通过引入“自持承诺账本”（self-kept commitment ledger）与简短的释放策略规范，可在五种模型中杜绝覆盖行为，并使GPT-6-Luna的严格准确率从49%提升至83%，接近最优基准。因此，解决方案的关键在于构建具备状态追踪能力的可信执行机制，并结合清晰的策略约束来实现对承诺生命周期的精确控制。
链接: https://arxiv.org/abs/2610.21203
作者: Lan Wei,Bingyu Shen,Jiaqi Tang,Qing Lu,Kuan Huang,Meng Xu,Miles Q. Li,Bin Hu,Boyang Li
机构: Kean University(基恩大学); McGill University(麦吉尔大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 21 pages, 7 figures, 13 tables. Code: this https URL
Abstract:When an assistant serving a team makes a promise to Alice, who may cancel it? We call this standing: only the creditor (the person owed) or someone the creditor or operator policy authorizes may release a commitment, regardless of rank. In a controlled chat environment we vary who cancels and separate two errors: unauthorized release (obeying someone without standing) and override (executing a released promise). Across six models, smaller open ones override the creditor (Qwen3.5-9B: 69%); larger ones drop promises on a peer’s direct word (Qwen3.8-27B: 89%; gpt-6.1-sol: 83%); nearly all act on a bystander’s “go ahead after all”. Oracles dissociate the errors: stating the rule or the creditor curbs unauthorized release but not override, which the latter even raises in three of four models; among them, only knowing whether the promise is still owed cuts override. Enabling thinking moves smaller models the way scale does. A self-kept commitment ledger stops override in four of five models; with a one-paragraph release policy it matches or beats the best oracle store, raising gpt-6-luna’s strict accuracy from 49% (policy alone) to 83%. Code, prompts, outputs and labels are released.
[NLP-103] Beyond Retrieval Failure: Understanding Why LLM s Miss Evidence in Multi-Hop RAG
【速读】： 该论文旨在解决多跳检索增强生成（multi-hop retrieval-augmented generation, RAG）系统中因搜索漂移（search drift）导致的看似检索失败但实则源于查询引导失效的问题。传统检索评估将检索失败归因于文档不可达或查询无效，而忽略了二者之间的本质区别。本文提出无需训练的执行注册器（ExecutionRAG），通过维护待解答子问题及其支持证据的注册表，精准追踪未完成的推理路径，从而在不改变原始检索器与语料库的前提下，明确区分出因搜索漂移导致的无效检索与真正无法访问的文档。实验表明，在1,000个MuSiQue问题上，使用all-MiniLM-L6-v2和Qwen3-Embedding-0.6B时，所需证据未被检索到的错误答案数分别从382降至234、383降至274，占比从69%→44%和66%→48%，显著揭示了搜索漂移是主要成因。然而，关键发现在于：尽管缺失证据的错误大幅减少，但总错误率并未等比例下降，反而因过早决断和放弃回答（abstentions）增加而出现误差结构转移——即系统瓶颈从“引导搜索获取必要证据”转向“验证已检索证据是否充分”。因此，解决方案的核心在于引入可解释的执行状态追踪机制，实现对检索过程的精确诊断，并推动模型优化重心由检索导向转向证据充分性验证。
链接: https://arxiv.org/abs/2610.21184
作者: Jiaqi Wei,Wenxuan Huang,Chengdong Xu,Xiang Zhang,Wanli Ouyang
机构: Zhejiang University(浙江大学); Shanghai AI Laboratory(上海人工智能实验室); Fudan University(复旦大学); Sun Yat-sen University(中山大学); University of British Columbia(不列颠哥伦比亚大学); The Chinese University of Hong Kong(香港中文大学)
类目: Computation and Language (cs.CL)
备注: Accepted by DAI 2026 Oral
Abstract:Multi-hop retrieval-augmented generation (RAG) requires an agent to retrieve evidence for a sequence of dependent sub-questions. When a wrong answer lacks a required passage, the passage may be inaccessible to the retriever, or the agent may have failed to issue a query that would retrieve it. Standard retrieval evaluations conflate these two causes. We study how much apparent retrieval failure is instead caused by search drift: the search no longer tracks an unanswered sub-question. Holding the retriever, corpus, and generation model fixed, we introduce ExecutionRAG, a training-free requirement register that records open sub-questions and their supporting evidence. The register changes search guidance but not evidence access, so each gold passage it recovers certifies that the passage was reachable by the original retriever. On 1,000 MuSiQue questions, the number of wrong answers with unretrieved required evidence falls from 382 to 234 with all-MiniLM-L6-v2 and from 383 to 274 with Qwen3-Embedding-0.6B. These counts correspond to 69% to 44% and 66% to 48% of each system’s wrong answers, respectively. Crucially, evidence recovery does not produce a proportional reduction in total error: premature resolutions and abstentions increase as missing-evidence failures decrease. The central finding is therefore a shift in the remaining bottleneck, from guiding search toward required evidence to verifying that the retrieved evidence is sufficient.
[NLP-104] Judging the Judges: Confidence and Disagreement in Rubrics as Rewards
【速读】： 该论文旨在解决生成式人工智能在非可验证领域（non-verifiable domains）中应用时，因缺乏客观评估标准而导致强化学习训练效果受限的问题。其核心挑战在于如何有效利用大语言模型（LLM）作为评判者（judge）对模型输出进行评分，以指导策略优化。现有方法多集中于改进评分量表（rubric）设计或增强单一评判者的可靠性，而忽视了不同评判者（包括单个与多个评判者）在训练过程中的协同作用。本文发现，仅评判者身份差异即会导致评分结果显著波动，尤其在医学与科学任务中使用五种前沿开源评判者时表现明显。为应对这一问题，作者提出一种新型评判者集成策略——RaRE（Rubrics as Rewards with Ensembles），其关键创新在于两个机制：一是基于评判者对输出的似然概率（logprobs）构建软李克特量表奖励（soft-Likert reward），实现对高置信度评判者的加权；二是“分歧剔除掩码”（drop-on-disagreement mask），在梯度回传策略优化（GRPO）更新中自动排除意见高度不一致的采样轨迹。实验证明，该设计至关重要：相较于简单平均评分的基线方法，RaRE在HealthBench和GPQA Diamond两项基准上均超越所有单评判者方案，且主题级分析显示其能够融合各评判者的优势，表现出互补性。研究结果确立了评判者集成作为提升生成式强化学习（RaR）性能的重要方向。
链接: https://arxiv.org/abs/2610.21179
作者: Michael Tian-Yue Liu,Anne Harrington,Trevor Darrell,Yutong Bai
机构: UC Berkeley(加州大学伯克利分校)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Project page: this https URL Code: this https URL
Abstract:Rubrics as Rewards (RaR) extends reinforcement learning to non-verifiable domains by using an LLM judge to score responses against a structured rubric. Prior work has focused on improving the rubric or strengthening the single judge, with comparatively little attention to how different single or multiple judges might be used during RaR training. We first show that judge identity alone produces a substantial spread in scores when training Qwen3-4B-Base on RaR-Medicine and RaR-Science across five frontier and open-weight judges. Building on this, we propose RaRE (Rubrics as Rewards with Ensembles), a novel judge ensemble strategy for RaR with two mechanisms: a soft-Likert reward that uses judge logprobs to weight more confident judges more heavily, and a drop-on-disagreement mask that excludes high-disagreement rollouts from the GRPO update. We find that this design is critical: while a naive average of judge scores fails to outperform the best single judge, RaRE outperforms every single-judge RaR variant on both HealthBench and GPQA Diamond, with per-theme analysis showing that it inherits the complementary strengths of its constituent judges. Our findings establish judge ensembling as a compelling axis for improving RaR.
[NLP-105] Language Models Share a Readout Bottleneck in Working Memory
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在N-back任务中表现出的人类工作记忆特征是否源于共享的内在机制这一问题。具体而言，研究关注的是：当记忆负荷N增加时，模型性能下降且错误集中在“诱饵项”（lures，即与当前刺激相邻但错误位置的重复项）上的现象，究竟是由于注意力精度不足（precision account），还是由于竞争性干扰导致决策被污染（interference account）。其解决方案的关键在于通过多组实验设计对这两种假说进行解耦分析：在15个预训练的LLM（涵盖7000万至140亿参数的三个模型家族）以及从零训练的最小化Transformer中，利用线性探测器精确判断模型在误报到诱饵项的试次中，是否仍能正确识别目标项或诱饵项的位置信息。结果发现，尽管模型自身决策能力差，但其注意力头对目标和诱饵位置的编码信息可被近乎完美地解码；同时，增强注意力聚焦几乎不提升性能，而抑制非目标项的字符身份则显著改善敏感度（d’），尤其在大模型中更为明显。此外，在最小化Transformer中，单层受限于精度，但引入第二层后即可克服此限制。由此表明，大语言模型的工作记忆瓶颈并非定位目标的能力不足，而是读取过程中受干扰内容影响所致，支持了人类工作记忆中的干扰理论（interference account），而非精度理论。
链接: https://arxiv.org/abs/2610.21176
作者: Dongyu Gong
机构: Yale University (耶鲁大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Neurons and Cognition (q-bio.NC)
备注:
Abstract:Large language models (LLMs) reproduce a human working-memory signature on the N-back task: performance declines as memory load N increases, and errors concentrate on lures, items that repeat the current stimulus at a neighboring but incorrect serial position. We ask whether this shared behavioral signature reflects a shared mechanism. The literature offers two candidate accounts. On a precision account, softmax attention cannot concentrate its weight on the item N positions back, so adjacent items are retrieved instead. On an interference account, attention selects the correct item, but the content of competing items corrupts the decision. We dissociate these accounts in 15 pretrained LLMs from three families (70 million to 14 billion parameters) and in minimal transformers trained from scratch. In all three families, the attention heads that carry the task attend to the target on most missed matches. Linear probes decode near-perfectly whether a matching letter occurred at the target or at a lure position, even on trials where the model false-alarms to the lure, although its own decisions barely distinguish the two. Sharpening attention yields negligible gains, whereas suppressing the letter identity of non-target items improves sensitivity (d’) in nearly all model-by-load conditions. In larger models the task is spread over more heads, and removing interference helps more. In minimal transformers, precision limits a single attention layer, but a second layer removes this constraint. The working-memory limit of LLMs therefore lies in reading items out under interference rather than in locating them, consistent with interference accounts of human working memory.
[NLP-106] From Long Context to Selective Memory:A Survey of Memory-Centric Compression for Large Language Models
【速读】： 该论文旨在解决大语言模型（LLM）在处理长上下文任务时，因持续扩展上下文窗口而带来的推理阶段计算、存储与信息选择压力。现有研究如提示压缩、键值缓存（KV-cache）管理、高效服务、检索增强及代理记忆等往往孤立探讨，未能系统理解信息在存储与计算之间的流动机制。本文将长上下文压缩重新定义为一种推理资源的内存管理问题，区分了直接参与模型计算的“计算活跃内存”（compute-active memory）与处于计算路径之外但可存储、检索、更新或遗忘的“计算非活跃内存”（compute-inactive memory），并把卸载的KV缓存、持久前缀缓存和预计算的KV状态视为激活路径上的中间状态。在此基础上，论文构建了一个包含写入、卸载、召回、检索注入、实例化、驱逐、整合与遗忘等操作的信息生命周期状态转移框架，并提出一个预算化的成本模型，将各类操作统一纳入单一成本核算体系。该框架实现了短期上下文压缩与长期记忆管理及其相互转换的统一，揭示出核心挑战并非简单减少信息量，而是动态决策哪些信息应在何时以何种形式参与高成本推理。此外，论文还总结了评估维度、基准差距与开放挑战，主张未来研究应从静态、点对点的压缩转向动态、可调控、具备生命周期感知能力的内存管理。
链接: https://arxiv.org/abs/2610.21165
作者: Chaogui Gou
机构: USTB(北京科技大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Long-context capability has become essential for large language models (LLMs) to handle complex tasks, yet continuously expanding the context window does not remove the computational, storage, and information-selection pressures incurred at inference time. Existing work on prompt compression, KV-cache management, efficient serving, retrieval augmentation, and agent memory is usually discussed in isolation, leaving the flow of information between storage and computation insufficiently understood. This survey reframes long-context compression as a memory-management problem for allocating inference-time resources. We distinguish compute-active memory, which directly participates in model computation, from compute-inactive memory, which lies outside the computation path but can be stored, retrieved, updated, or forgotten, and we characterize offloaded KV caches, persistent prefix caches, and precomputed KV states as intermediate states on the activation path. On this basis, we formulate writing, offloading, recall, retrieval-and-injection, materialization, eviction, consolidation, and forgetting as state transitions within an information lifecycle, and introduce a budgeted cost model that places these operations in a single cost account. The resulting framework unifies short-term context compression, long-term memory management, and the transitions between them, and shows that the central challenge is not merely reducing information volume but deciding which information should participate in costly inference, when, and in what form. We further summarize evaluation dimensions, benchmark gaps, and open challenges, and argue that future research should move from static, pointwise compression toward dynamic, governable, lifecycle-aware memory management.
[NLP-107] Automatic Sentence Completion Using Bengali Audio Speech with a Transformer-Based Early Speech Recognition System
【速读】： 该论文旨在解决阿尔茨海默病（Alzheimer’s Disease）患者在日常交流中因认知功能衰退导致的语言表达障碍问题，尤其关注在孟加拉语（Bangla）这一资源相对匮乏的语言环境中，如何通过技术手段辅助患者实现有效沟通。其核心解决方案在于构建一个集成化系统，融合语音与文本分析技术，实现对孟加拉语语音中痴呆症早期迹象的检测，并提供实时句子自动补全支持。关键技术突破包括：基于声学特征提取的语音模式分析以捕捉与认知衰退相关的言语变化；采用基于Transformer的语音转写模型将语音信号转化为文本，用于后续语言学分析；在分类任务中，利用长短期记忆网络（LSTM）有效建模语音与语言数据中的上下文依赖关系；在句子自动补全任务中，设计了一种结合双向长短期记忆网络（Bi-LSTM）与注意力机制的序列学习模型，能够根据部分表达的孟加拉语句子生成合理且连贯的补全内容。该研究填补了低资源语言在痴呆症智能辅助领域的空白，为认知障碍患者的自然语言交互提供了可扩展的技术框架。
链接: https://arxiv.org/abs/2610.21164
作者: Md Asif Akram,Arif Faisal,Sakir Hossain Faruque,Sharun Akter Khushbu
机构: 未知
类目: Computation and Language (cs.CL)
备注: 18 pages, 10 figures
Abstract:Dementia severely impairs communication abilities and turns everyday conversations into challenging experiences. Individuals affected by dementia frequently struggle to express their thoughts or complete sentences, making it difficult for family members and caregivers to understand their needs. This study proposes a novel approach to assist dementia patients by developing an integrated system that detects signs of dementia through Bangla speech and text analysis and provides sentence auto-completion support during conversations. To achieve this, recorded voices from individuals with and without dementia were analysed, and key acoustic features were extracted to capture vocal patterns associated with cognitive decline. A transformer-based model was employed to transcribe voice signals into text for linguistic analysis. In the classification task, several Machine Learning (ML) and Deep Learning (DL) models were evaluated using both acoustic and textual representations, and the LSTM model architectures demonstrated strong capabilities for modelling contextual patterns in speech and language data. For the sentence auto-completion task, a sequential learning model based on a Bidirectional LSTM (Bi-LSTM) network with an attention mechanism was developed to sequentially complete partially expressed Bangla sentences during conversation. This research bridges the gap in dementia analysis in an underrepresented language and implements a sentence auto-completion framework for cognitive disorders.
[NLP-108] Stop Training Against Soft Users: Mentalized Adversarial Self-Play for Proactive Dialogue EMNLP2026
【速读】： 该论文旨在解决主动对话系统在多轮交互中如何有效引导对话以达成任务目标，同时适应用户意图与态度动态变化的挑战。现有规划方法虽通过搜索、学习策略或用户感知规划提升了系统端动作选择能力，但其训练与评估常依赖于固定或过度配合的模拟器，导致高成功率可能反映的是对用户的“软性偏差”而非真正的鲁棒对话能力。为此，论文提出一种基于心理建模的对抗自演进框架（MASP），将主动对话视为系统与理性抵抗用户之间的部分可观测博弈。MASP采用基于信念-动机-意图（BDI）的潜在状态表示用户不断演变的心理状态，区分奖励侧教师的心理建模与策略侧学生的心智建模，并交替更新系统与用户策略。该设计将对理性抵抗的鲁棒性转化为可优化的训练目标，而非仅作为评估条件。在多个主动对话基准测试中，相较于广泛使用的监督微调（SFT）和仅依赖规划器的方法，MASP显著提升成功率（SR）与理性约束成功率（RCSR），降低平均对话轮次（AT），并在面对使任务目标更难实现的“硬用户”时表现更优。
链接: https://arxiv.org/abs/2610.21156
作者: Zanwu Liu,Dan Zhang,Xiaodi Wang,Yongbo Mao,Xiaoyang Guo,Yonglin Wu,Peiyue Jiang,Yangyang Wei,Zecheng Lin
机构: Beihang University (北京航空航天大学); Dongchezu Technology Co., Ltd. (东车组科技有限公司)
类目: Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 (Main Conference). 18 pages including references and appendices; 1 figure
Abstract:Proactive dialogue agents must guide multi-turn interactions toward task goals while adapting to users whose intentions and attitudes evolve. Existing planners improve system side action selection through search, learned policies, or user aware planning, but are often trained and evaluated against fixed or overly compliant simulators; as a result, high simulated success can reflect soft user bias rather than robust dialogue ability. We introduce MASP, a Mentalized Adversarial Self Play framework that treats proactive dialogue as a partially observable game between a system and a rationally resisting user. MASP represents the user’s evolving mental state with a BDI based latent state, separates reward side teacher mentalization from policy side student mentalization, and alternates self play updates for the system and user policies. This turns robustness to rational resistance into a training objective rather than only an evaluation condition. Across diverse proactive dialogue benchmarks, compared with the widely used SFT and Planner-only approaches, MASP substantially improves success rate (SR) and rationality constrained success (RCSR), reduces average turns (AT), and performs better against hard users who make the task goal more difficult to achieve.
[NLP-109] CARE-SD: Explainable Reasoning Architecture for Local SLM Self-Disclosure Analysis in Textual Data
【速读】： 该论文旨在解决在网络安全访谈中自动识别真实自我披露（self-disclosure）时面临的挑战，即传统技术评估难以全面捕捉用户的安全经验、行为及隐私关切，且由于语境依赖性强、类别重叠以及自动化分析可能引入未经证实的推断，导致准确识别困难。其解决方案的关键在于提出一种名为CARE-SD（Context-Adaptive Reasoning and Explanation for Self-Disclosure）的新型四模块架构，通过集成上下文自适应处理、层次化披露推理、解释一致性验证与跨批次协调机制，显著提升小型语言模型（SLMs）在本地部署环境下的上下文感知能力与事实可靠性。实验表明，该架构使Llama 3B模型的宏平均F1值从0.80提升至0.88，并减少60%的不当主张，同时3B模型在分类性能上达到20B参考模型水平，证明了其无需依赖更大模型即可实现高性能分析的有效性。
链接: https://arxiv.org/abs/2610.21151
作者: Aisvarya Adeseye,Adeyemi Adeseye,Mohammad Tahir,Jouni Isoaho
机构: University of Turku (图尔库大学)
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted for the 4th International Conference on Foundation and Large Language Models (FLLM2026) - IEEE - held from 17-20 Nov., 2026 at Novotel Barcelona City, Barcelona, Spain
Abstract:Cybersecurity interviews uncover critical security experiences, behaviors, and privacy concerns often not fully captured by conventional technical assessments. Consequently, identifying genuine self-disclosures is challenging because statements are context-dependent, categories overlap, and automated interpretations may introduce unsupported claims. Locally deployed Small Language Models (SLMs) protect sensitive participant data by keeping processing within controlled environments, but their limited reasoning capabilities make simple prompting insufficient for context-aware and factually reliable analysis. This study developed CARE-SD (Context-Adaptive Reasoning and Explanation for Self-Disclosure), a novel four-module architecture integrating context-adaptive processing, hierarchical disclosure reasoning, explanation-consistency verification, and cross-batch reconciliation. We evaluated three local SLMs, including Llama 1B, Qwen 1.5B, and Llama 3B, using 82 independently annotated cybersecurity interviews across five organizational sectors. Comparisons included conventional classifiers, three prompting strategies, larger 8B–32B models, and architectural ablations. Two independent reviewers assessed disclosure labels, factual claims, and explanation correctness. CARE-SD improved the accuracy and evidence reliability of SLMs, increasing Llama 3B Macro-F1 from 0.80 to 0.88 while reducing unsupported claims by 60%. Remarkably, the 3B model matched the 20B reference model in category Macro-F1, demonstrating the effectiveness of the architecture without requiring a larger model.
[NLP-110] From Observation to Intervention: Distilling Self-Generated In-Context Demonstrations via Executable Causal Intervention
【速读】： 该论文旨在解决自生成上下文学习（Self-generated In-Context Learning, ICL）中演示样本筛选的可靠性问题。现有方法依赖于观察结果进行筛选，即认为能够产生正确输出的推理过程（rationale）即为高质量，但这种方法存在观测等价性（observational equivalence）缺陷：真实因果性推理与虚假事后解释（spurious post-hoc rationales）可能在输出上完全一致，仅通过观察无法区分。为此，论文提出关键解决方案——可执行因果干预演示蒸馏（Executable Causal Intervention for Demonstrations Distillation, ECI2D），其核心思想是基于独立因果机制假设，主张真实的因果推理应具备跨逻辑相似样本的可迁移不变性（transfer invariance）。ECI2D将语言推理过程抽象为可执行程序，并对逻辑相似样本施加因果干预（do-操作），仅保留那些在干预后仍保持有效性的演示样本。实验表明，该方法在数学、逻辑和常识推理基准上均显著优于基于观察的ICL，且所需演示数量大幅减少，在测试分布发生偏移时性能提升超过7个百分点。
链接: https://arxiv.org/abs/2610.21149
作者: Minming Ye,Min Liu,Haoang Chi,Ciheng Wu,Naiyang Guan
机构: Shanghai Jiaotong University (上海交通大学); Academy of Military Science (军事科学院); National University of Defense Technology (国防科技大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Self-generated In-Context Learning (ICL) enables LLMs to synthesize their own few-shot demonstrations, often outperforming manual prompts. However, existing methods filter demonstrations by observation, assuming a rationale is high-quality if it leads to the correct output. We show such filtering suffers from \textitobservational equivalence, whereby both faithful causal rationales and spurious post-hoc rationales can yield same outputs and are statistically indistinguishable from observation alone. Inspired by independent causal mechanisms, we argue that faithful rationales imply \textittransfer invariance, in which causal rationales should remain effective across logically similar samples. We instantiate this criterion as \textbfExecutable \textbfCausal \textbfIntervention for \textbfDemonstrations \textbfDistillation (ECI2D), which abstracts language rationales into executable programs and applies the causal interventions ( do -operator) on logically similar samples, retaining only those demonstrations that imply transfer invariance. Experiments on math, logical and commonsense reasoning benchmarks demonstrate that ECI2D consistently outperforms observation-based ICL while retaining far fewer demonstrations, with gains exceeding 7 points as the test distribution shifts.
[NLP-111] CoSTA: Compressed Speech Token Adaptation for Efficient Spoken Dialogue Models
【速读】： 该论文旨在解决语音对话模型（Spoken Dialogue Models, SDMs）中生成与合成共享同一语音标记空间所导致的自回归序列过长及推理成本高昂的问题。现有方法虽通过低速率语音标记器缩短序列，但往往引入压缩质量权衡且难以打破生成与合成之间的耦合。为此，本文提出CoSTA（Compatibility-preserving post-hoc framework），其核心创新在于在不改变原有语音标记器和解码器的前提下，解耦生成与合成的标记空间：通过均值初始化的词汇扩展、发音适应与对话适应策略，将生成端的标记空间扩展为压缩表示，并在合成前通过确定性映射回原始语音标记空间。该方案实现了约1.7倍的语音标记生成速度提升、更低的完整响应延迟与更高吞吐量，同时保持响应质量。分析表明，在资源受限场景下可实现更持续的交互，且在压缩率、解码效率与生成质量之间具备实用权衡能力；此外，实验还验证了其跨架构迁移性及与低速率标记化的互补性。
链接: https://arxiv.org/abs/2610.21143
作者: Yingzhuo Deng,Chong Li,Bingxin Liu,Yi Shu,Tianyu Peng,Can Xie,Siyao Song,Shuo Ren,Jiajun Zhang
机构: Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所); School of Artificial Intelligence, University of Chinese Academy of Sciences(中国科学院大学人工智能学院); Wuhan AI Research(武汉人工智能研究院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Sound (cs.SD)
备注: 31 pages, 11 figures. Code: this https URL
Abstract:Spoken Dialogue Models (SDMs) typically share a speech-token space for generation and synthesis, leading to long autoregressive sequences and costly inference. Lower-rate speech tokenizers shorten sequences but may introduce compression-quality trade-offs and costly replacement without breaking this coupling. To address this limitation, we propose CoSTA, a compatibility-preserving post-hoc framework that decouples generation and synthesis token spaces while retaining the original speech tokenizer and decoder. CoSTA extends the generation token space with compressed representations via mean-initialized vocabulary extension, pronunciation adaptation, and dialogue adaptation, followed by deterministic expansion back to the original speech-token space before synthesis. On bilingual speech-dialogue benchmarks, CoSTA achieves approximately 1.7\times expanded speech-token generation speed, lower full-response latency, and higher throughput while preserving response quality. Analyses show more sustained interaction under limited budgets and a practical trade-off among compression, decoding efficiency, and quality. Further experiments demonstrate cross-architecture transfer and complementarity to low-rate tokenization.
[NLP-112] When Judges Agree and Are Wrong: Consistency-Accuracy Dissociation as a Blind Spot for Label-Free Evaluation
【速读】： 该论文旨在解决当前无标签评估方法（如自一致性过滤、多数投票聚合和基于一致性的可靠性评分）所依赖的核心假设——即一致性可作为正确性的代理指标——在缺乏真实标签时的有效性问题。研究发现，60%的错误答案出现在高一致性区域（一致性 ≥ 0.8），其中77.7%的错误表现为完全的答案坍缩（即所有扰动变体均输出相同错误答案），表明模型在高度一致的同时却完全错误。这一现象无法通过仅依赖一致性的评估信号识别，因为模型表现出极强的表面可靠性。解决方案的关键在于揭示并量化“高一致性-低准确性”这一矛盾区域的存在及其鲁棒性，该现象在不同一致性阈值（t=0.50 至 t=1.00）、模型家族间均稳定存在，并集中于特定的扰动维度与领域交集。研究结果强调了现有基于一致性的评估范式在无监督或弱监督场景中的根本局限性，对依赖此类信号替代人工标注的评估流程提出了严峻挑战。
链接: https://arxiv.org/abs/2610.21128
作者: Xiang Zhang,Lovedeep Gondara,Jekaterina Novikova
机构: The Vanguard Group, Inc. (先锋集团)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Label-free evaluation methods, such as self-consistency filtering, majority-vote aggregation, and agreement-based reliability scoring, rely on an implicit assumption that consistency can serve as a proxy for correctness when ground-truth labels are unavailable or expensive to obtain. In this paper, we test this assumption directly. Using four financial QA datasets (multiple-choice and free-form), ten consistency dimensions, and two frontier model families, we measure the joint distribution of consistency and accuracy at the level of individual examples rather than in aggregate. We find that 60% of all incorrect answers fall in the high-consistency region (consistency \geq 0.8), and that 77.7% of these show complete answer collapse, i.e. identical wrong answers across all perturbation variants. This high-consistency/incorrect quadrant cannot be detected by any evaluation signal that is built on agreement alone, as the model appears maximally reliable while being wrong. We show that the phenomenon is robust to the choice of consistency threshold t (it persists from t=0.50 to t=1.00), is shared across model families, and is concentrated in specific intersections of perturbation dimension and domain. We map where this quadrant occurs, analyze its stability properties, and discuss the implications for evaluation pipelines that use agreement-based signals in place of human labels.
[NLP-113] Compression Does Not End at Selection: Post-Selection Positional Geometry for Sparse KV Caches
【速读】： 该论文旨在解决稀疏键值（KV）压缩在长上下文推理中因保留的历史状态坐标未重新组织，导致在旋转位置编码（Rotary Position Embedding, RoPE）下被剔除的令牌留下的空隙持续影响查询-键匹配的问题，从而限制了保留信息的有效利用。其解决方案的关键在于提出RoPERelay方法：通过为缓存状态分配紧凑且保持顺序的RoPE坐标，利用逆旋转与再旋转操作重定位缓存的键向量相位，并同步当前查询与后续生成过程，实现无需重新选择、不需历史重填充或额外缓存预算的后选位置几何重构。该方法有效改善了稀疏KV缓存中保留信息的可访问性，在36项实验对比中显著提升了32次的答案质量，例如在Llama-3.1-8B-Instruct模型上结合BM25于LongMemEval-S任务中的准确率从23.6%提升至35.4%，且整体请求完成时间开销仅为1%-3%。研究揭示了内容选择后的位置组织是稀疏KV缓存优化的一个独立维度。
链接: https://arxiv.org/abs/2610.21126
作者: Wanyang Hong,Zhaoning Zhang
机构: National Key Laboratory of Parallel and Distributed Computing(并行与分布式计算国家重点实验室); National University of Defense Technology (国防科技大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Sparse KV compression reduces long-context inference costs by retaining selected historical states, yet these states typically keep their coordinates from the full history. Under Rotary Position Embedding (RoPE), gaps left by evicted tokens continue to affect query-key matching, potentially limiting the use of retained evidence. We study post-selection positional geometry on a fixed selector-produced cache, examining how positional organization affects the use of retained information and which coordinate and continuation rules improve answer quality. To reorganize this geometry, we propose RoPERelay, which assigns compact, order-preserving RoPE coordinates, relocates cached key phases through inverse rotation and re-rotation, and synchronizes the current query and subsequent generation. It reuses existing K/V without renewed selection, historical re-prefill, or additional cache budget. Across 36 comparisons spanning two models, six sparse selectors, and three long-term memory tasks, RoPERelay improves answer quality in 32 cases. For Llama-3.1-8B-Instruct with BM25 on LongMemEval-S, accuracy increases from 23.6% to 35.4%. Measured complete-request time overhead is approximately 1%-3% in the evaluated configurations. These findings establish positional organization as a distinct optimization dimension for sparse KV caches after content selection.
[NLP-114] Planning Commitment of LLM under Partial Observability
【速读】： 该论文旨在解决基于大语言模型（LLM）的规划代理在部分可观测环境中的规划策略选择问题，即在“激进式承诺”与“保守式承诺”之间的安全与效率权衡。具体而言，激进式策略基于对未观测状态的假设生成长周期计划，虽可能提高效率但易因假设错误导致高成本失败；而保守式策略仅基于已观测信息进行短周期规划，虽提升安全性但可能因频繁调用LLM接口而降低效率。这一权衡长期以来缺乏系统性分析和可控性设计。研究发现，在两个部分可观测领域中，无论采用显式统计先验或常识知识先验，两种策略在准确性上表现相当，但保守策略显著减少失败尝试（提升安全性），代价是增加API调用次数（降低效率）；激进策略失败率更高，但在先验信息可靠时能高效重规划。值得注意的是，模型倾向于将提供的先验视为确定性事实，当先验不可靠时易引发失败。为此，论文提出可调节超参数的混合策略，实现安全与效率之间的动态平衡，从而为实际应用提供可控的决策框架。
链接: https://arxiv.org/abs/2610.21108
作者: Cassie Huang,Zhanwei Cao,Yeawon Lee,Nathalia Gomez,Zhuoyi Yang,Ian G. Harris,Tiffany Do,Jiayuan Mao,Li Zhang
机构: Drexel University; Northeastern University; University of California, Irvine; University of Pennsylvania
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:LLM-based planning agents commit to plans of different lengths before interacting with a partially observable environment. Some aggressively commit to a long plan based on assumptions about unobserved states, while others conservatively plan only over what they have observed. Both behaviors emerge implicitly in pre-trained models, whether the LLM outputs a plan directly or generates a PDDL formulation for a classical solver. Aggressive commitment risks costly failures, while conservative commitment risks excessive LLM calls. This choice pits safety against efficiency, a trade-off that has been neither characterized systematically nor made controllable. We investigate aggressive and conservative LLM planners and formalizers on two partially observable domains, where priors are provided as either explicit statistics or common-sense knowledge. Across three LLMs, both strategies reach similar accuracy. Conservative methods achieve fewer failed attempts (safety) at the cost of more API calls (efficiency). Aggressive methods fail more often but replan efficiently only when priors are informative. Interestingly, analysis shows that LLMs treat the provided priors nearly as facts, which risks failure when priors are unreliable. We also explore hybrid strategies with tunable hyperparameters that adjust the balance between safety and efficiency.
[NLP-115] oolBottleneckBench: Measuring Tool-Agent Performance under Controlled Pressures and Fixed Harness Budgets
【速读】： 该论文旨在解决工具代理（tool agent）性能评估中因运行环境约束（harness budget）与外部压力因素相互干扰而导致的评价失真问题。现有评估方法难以区分模型能力与系统资源限制对性能的影响，尤其在多变的工具注册目录、可见工具菜单、认证工作流及返回记录编码等压力条件下，性能下降的归因模糊。为此，作者提出ToolBottleneckBench基准测试框架，通过在固定企业级合成任务和恒定硬件/计算预算下，系统性地控制单一压力变量（如可见工具菜单大小），实现对工具代理性能的可复现、可量化评估。其解决方案的关键在于：引入精确的评分机制，记录每个失败轨迹中首次未满足的任务义务及终止时的约束条件；实验发现，仅“可见工具菜单”这一压力对模型表现产生显著负面影响，且该影响与生成结果截断（truncation）密切相关；当将完成上限（completion cap）提升16倍后，多数截断密集型配置下的性能惩罚基本消除。此外，多数失败由系统约束（harness limit）触发，而失败记录中压力来源的识别准确率仅略优于工具调用次数统计。进一步分析表明，仅少数修复策略（涉及工具可达性或调用主体变更）能跨所有配置提升性能，且无一改善模型本身的推理行为。因此，研究强调应明确报告压力效应所对应的特定运行预算（harness budget），以确保评估结果的可比性与透明度，并开源了该基准测试平台。
链接: https://arxiv.org/abs/2610.21098
作者: Zichuan Xu,Yuyang Dong,Makoto Onizuka,Chuan Xiao
机构: University of Osaka(大阪大学); Initial S; Nagoya University(名古屋大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Software Engineering (cs.SE)
备注: 26 pages, 12 figures, 15 tables
Abstract:The success score of a tool agent reflects both its model and the budgets of the harness that runs it. We present ToolBottleneckBench, a benchmark that measures tool-agent performance under controlled pressures and fixed harness budgets. Synthetic enterprise tasks stay fixed while one pressure varies: the registered tool catalog, the visible tool menu, the certified workflow, or the encoding of returned records. An exact grader records the first unmet obligation of each failed episode and the harness limit, if any, that ended it. Across 21 model and harness configurations, the visible menu is the only one of the three pressures that act on the model with a large penalty, and this penalty tracks truncation at the completion cap. Raising the cap sixteenfold removes most of the penalty in the truncation-heavy configurations. Most failures end at a harness limit, and the failure record identifies the assigned pressure no more reliably than a count of tool calls. The repairs that help change which tools reach the model or who issues the calls, and only two repairs, neither improving model behavior, beat every configuration switch. We recommend reporting pressure effects with the harness budgets under which they were measured. We release the benchmark at this https URL.
[NLP-116] MemComposer: Learning to Assemble Memories for LLM Agents via Reinforcement Learning
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）代理在利用记忆（memory）时面临的有效性问题：尽管基于语义相似性的记忆检索能够保留过往经验，但单纯依赖语义匹配无法确保记忆对下游任务的实际帮助，其关键在于记忆的选择、数量与呈现顺序。为此，论文提出MemComposer——一种记忆组合器（memory composer），通过联合分析当前查询与检索到的记忆候选，动态决定应包含哪些经验、选择多少条以及如何排序，从而生成最优记忆上下文。其核心解决方案是采用基于GRPO的体验策略优化（Experience Policy Optimization, ExPO）框架，通过完整代理交互过程评估组装后的记忆上下文，并利用任务成功反馈来优化记忆组装策略。实验表明，在三个τ²-bench领域中，使用两个冻结执行器（frozen executors）的情况下，MemComposer均优于最强基线，同时分别减少16.9%和4.9%的平均记忆令牌消耗。此外，所学习的策略具备跨领域、跨执行器及跨记忆库的迁移能力，并可直接受益于更强模型生成的记忆而无需额外训练，充分验证了基于下游任务反馈学习查询相关记忆组装策略的有效性。
链接: https://arxiv.org/abs/2610.21091
作者: Yixuan Xie,Xueqiang Xu,Xinyi Fan,Xinxi Lyu,Tao Feng,Xing Fan,Chenlei Guo,Jie Hao,Shaowen Wang,Jiashuo Sun,Jiawei Han
机构: University of Illinois Urbana-Champaign(伊利诺伊大学厄本那-香槟分校); Amazon(亚马逊)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 30 pages, 8 figures, 14 tables
Abstract:Agent memory enables large language model (LLM) agents to reuse past experiences without updating their parameters. However, semantic similarity alone does not guarantee memory usefulness: downstream performance depends on how memories are selected and presented to the agent. We propose MemComposer, a memory composer that jointly examines the current query and retrieved candidates to decide which experiences to include, how many to select, and how to order them. We train the composer using Experience Policy Optimization (ExPO), a GRPO-based framework that evaluates assembled memory contexts through complete agent interactions and uses task-success feedback to optimize the assembly policy. Across three \tau^2 -bench domains with two frozen executors, MemComposer outperforms the strongest baseline for each executor while reducing average memory-token usage by 16.9% and 4.9%, respectively. The learned policy also transfers to unseen domains, executors, and memory banks, and benefits from memories generated by stronger models without further training. These results demonstrate the value of learning query-specific memory assembly using downstream task feedback.
[NLP-117] Multimodal AI Detection for Math-Heavy Scientific Text
【速读】： 该论文旨在解决科学文献中由大语言模型（Large Language Models, LLMs）生成内容的检测问题，尤其是在数学密集型科研文本中，现有基于纯文本的AI生成内容检测工具因无法利用排版格式中的丰富视觉信息而性能显著下降。其核心挑战在于：传统检测方法依赖从PDF提取的纯文本，忽略了数学公式与排版结构所蕴含的关键语义和格式特征。为应对这一问题，研究提出多模态检测框架，关键创新在于通过微调基于Qwen-3.5-27B的模型并利用LaTeX源码输入以保留格式信息，显著提升了对数学内容的检测准确率；进一步地，针对实际场景中LaTeX源码难以获取的问题，构建了SnapJudge——一个基于图像的检测器，直接处理论文截图中的文本与公式区域，在0.1%假阳性率下对前沿模型生成的证明与叙述文本分别实现92%和93%的检测率，显著优于Pangram 4及基于文本的检测器。该方案通过融合视觉与结构化信息，实现了在复杂科学内容场景下的高效、高精度AI生成内容识别。
链接: https://arxiv.org/abs/2610.21062
作者: Yixuan Huang,Danish Pruthi,Bhuwan Dhingra
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:As large language models become increasingly capable of generating, and assisting with, scientific writing, there is a need for greater transparency and disclosure around the use of AI in research. Consequently, a veritable subfield of research and a number of commercial services aim to flag AI-generated content automatically. However, existing detectors which rely only on text extracted from PDFs fail to benefit from the rich visual information contained in the formatting of math-heavy scientific content. To study this, we select human-written proof and prose passages from pre-2022 arXiv LaTeX sources and re-generate them using 15 modern LLMs within the context of the same paper. This leaves us with coupled human and AI outputs in three different representations: the LaTeX source, a rendered image of the PDF, and the text extracted from the PDF. We find that the accuracy of the state-of-the-art Pangram 4 drops significantly as the mathematical content in a passage increases, especially when evaluated on text as opposed to LaTeX. However, this is not an inherent limitation of detecting AI-generated math: finetuning Qwen-3.5-27B based detectors on this domain recovers the gap, with LaTeX inputs performing significantly better than text since they preserve more formatting information. Since LaTeX sources are rarely available in practice, we also train SnapJudge: an image-based detector operating on screenshots of both prose and proofs from papers. At 0.1% false positive rate, SnapJudge achieves a detection rate of 92% on proofs and 93% on prose generated by frontier models, substantially outperforming both Pangram 4 and our text-based detector. We release our data, models and code at this https URL.
[NLP-118] SHIP: Structured and Hierarchical Insight with Privacy
【速读】： 该论文旨在解决在保护用户隐私的前提下，从文本数据中提取可复用且具有可比性的隐私保护分析洞察（differentially private insights）这一关键问题。现有方法生成的隐私洞察形式多样，难以统一评估和比较其效用。为此，论文提出一种名为结构化与分层隐私洞察（SHIP）的新方案，其核心在于将分析查询自动映射到相关属性，构建分层混合分类体系，并生成差分隐私（DP）合成表以保留细粒度特征与跨属性统计信息，从而支持后续分析。该方案的关键创新在于通过结构化表示实现对复杂查询的高保真响应，同时建立基于问答性能与结构保真度的评估协议，实验证明SHIP在两项指标上均显著优于现有基线方法，同时揭示了当前结构化表示仍面临的挑战性查询类型。
链接: https://arxiv.org/abs/2610.21048
作者: Kai Chen,Guisi Ni,Albert Cheu,Kassem Fawaz,Yuzheng Hu,Ryan McKenna,Peter Kairouz
机构: Google Research(谷歌研究); University of Virginia(弗吉尼亚大学)
类目: Computation and Language (cs.CL); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注: This paper mainly work on extracting and evaluating insights from textual data
Abstract:Extracting insights from textual data while protecting user privacy is increasingly important. Existing methods produce private insights in diverse forms for different analytical purposes, making their utility difficult to assess and compare. In this work, we formulate differentially private (DP) insights as reusable representations for answering analytical queries, providing a common basis for utility assessment. Answering these queries, however, requires fine-grained details and cross-attribute statistics that existing representations may omit. To capture this information, we introduce Structured and Hierarchical Insight with Privacy (SHIP), which automatically maps queries to attributes, builds a hierarchical hybrid taxonomy, and releases a DP synthetic table for further analysis. We also develop an evaluation protocol based primarily on QA performance and structured fidelity. Our experiments demonstrate SHIP’s strong performance on both measures relative to existing private insight baselines, while identifying queries that remain challenging for structured representations.
[NLP-119] PremiseCheck: Efficient Fact-Checking of Human and AI Peer Reviews
【速读】： 该论文旨在解决学术同行评审中因投稿量激增而导致的评审压力，以及人工或生成式AI（Generative AI）撰写的评审意见中存在的事实性错误问题。其核心挑战在于如何高效且准确地检测评审意见中与原文不符的陈述，尤其是那些包含多个隐含断言的复杂评论。解决方案的关键在于提出PremiseCheck框架：该框架将高精度但高成本的前沿大语言模型（LLM）与低成本的决策模型相结合——首先由LLM将原始评论重构为明确、可验证的命题（premises），无需重新阅读全文；随后由轻量级决策模型逐一比对这些命题与原文内容。实验表明，该方法在独立评估每条评论时，可在保持与GPT-5.5相当准确率的前提下，将成本降低至约一半（在自然语言处理领域甚至降至四分之一），同时显著减少延迟（低于一半），实现了性能、成本与效率的平衡。
链接: https://arxiv.org/abs/2610.21046
作者: Dingjie Song,Zhiling Yan,Ziqian Zhang,Yuxuan Zhang,Lichao Sun
机构: Lehigh University(莱赫大学); University of British Columbia(不列颠哥伦比亚大学); Vector Institute(向量研究所)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 29 pages, 19 figures, 23 tables. Code: this https URL
Abstract:Peer review is under strain from the growth of submissions, and review comments written by humans or AI can misstate what a paper says. We introduce ReviewFact, a benchmark that labels such factual errors in review comments under one definition across NLP, machine learning, and the natural sciences, and that measures cost and latency alongside accuracy. A frontier LLM checks comments well but must read the whole paper for each check, while cheap decision models, which score whether a statement holds in a text, fall far short on whole comments that mix several implicit assertions. PremiseCheck is a framework that combines the two. An LLM rewrites each comment as explicit premises without reading the paper, and a decision model checks each premise against the paper. When each comment is checked on its own, PremiseCheck reaches accuracy comparable to GPT-5.5 across all three fields at about half its cost, and on NLP reviews at a quarter of its cost and less than half its latency.
[NLP-120] INCONVENIENT: A Benchmark to Assess LLM Generalization Capabilities from Convenience Samples
【速读】： 该论文旨在解决表格数据预测模型在现实应用中因便利采样偏差（convenience-sample bias）导致的泛化性能下降问题。现有表格数据分布偏移（distribution-shift）基准多依赖自然存在的领域划分构建域外（OOD）场景，但此类方法无法精确控制或隔离特征-目标关系的变化，且评估分布往往不具代表性。为此，本文提出INCONVENIENT基准，其核心创新在于不依赖自然领域划分，而是直接对训练数据中的特征-目标关系进行可控扰动，同时保持原始总体分布作为部署环境，从而更真实地模拟模型从非代表性样本训练后部署于更广泛人群的实际情况。该方法特别适用于大语言模型（LLM），因其预训练阶段可能蕴含超越训练数据中观察关系的群体层面知识。与现有基准相比，INCONVENIENT能够显式控制关系偏移的方向与强度，实现对模型鲁棒性的系统性评估；该基准包含200个跨领域的表格分类与回归数据集，生成约15,500个被扰动的训练分布，导致下游任务性能显著下降，有效揭示了模型对便利采样偏差的脆弱性。
链接: https://arxiv.org/abs/2610.21038
作者: Sara Rezaeimanesh,Chad Klochko,Kundan Thind,Mohammad Ghassemi
机构: Michigan State University (密歇根州立大学); Henry Ford Health (亨利福特健康); Michigan State University (密歇根州立大学); Henry Ford Health (亨利福特健康)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Machine learning models for tabular prediction typically assume that training data are representative of the population on which they will be deployed. In practice, however, data are often collected based on convenience, availability, or cost, resulting in systematic sampling biases that can distort feature–target relationships and undermine real-world generalization. Existing tabular distribution-shift benchmarks primarily construct out-of-distribution (OOD) settings by partitioning observations across naturally occurring domains. While such partitions may induce changes in P(Y | X), the resulting shifts are neither explicitly controlled nor isolated, and the evaluation distributions themselves are often not representative of the broader target population. We introduce INCONVENIENT, a benchmark designed to study robustness to convenience-sample bias. Rather than relying on naturally occurring domain partitions, INCONVENIENT directly perturbs feature–target relationships in the training distribution while preserving the original population distribution as the deployment environment. This setting more closely reflects real-world scenarios in which models are developed from unrepresentative samples and subsequently deployed on a broader population. The distinction is particularly relevant for large language models (LLMs), whose pretraining may encode population-level knowledge that extends beyond the relationships observed in a biased training dataset. Unlike existing distribution-shift benchmarks, INCONVENIENT provides explicit control over both the direction and severity of relationship shifts, enabling a systematic assessment of robustness to convenience sampling. The benchmark comprises 200 tabular classification and regression datasets spanning diverse application domains and generates around 15,500 shifted training distributions which lead to substantial downstream degradation.
[NLP-121] Learning to Read Out: Unembedding Dynamics in Language Model Pretraining NEURIPS2026
【速读】： 该论文旨在解决语言模型在预训练过程中何时真正习得特定能力的问题。传统方法通过追踪模型首次生成正确输出标记（token）的时间点来判断能力获得时间，但这一过程依赖于未完全收敛的读出矩阵（readout），而读出矩阵本身在预训练期间仍在持续学习，因此基于输出的“行为时间线”会混淆模型内部表征（encoding）与读出表达（expressing）之间的差异。本文的关键解决方案是提出参数轨迹交叉编码（parameter-trajectory crosscoding），该方法对每个检查点的读出矩阵行拟合一个稀疏字典，仅需权重参数即可实现对读出过程的直接追踪。实验发现，在Pythia模型预训练早期（0.4%-1.1%阶段），读出矩阵发生重组，使WordNet词类信息可从其行向量中解码，而此时隐藏状态中的任务信息已可通过线性探测解码，但模型自身输出尚未体现该能力。进一步通过控制读出学习率的消融实验表明，提高读出学习速率可提前读出发展并缩短该“认知-表达”差距，且对最终损失影响甚微。这表明：在训练早期，许多能力发展的“行为事件”实际上反映的是读出模块的学习进度，而非模型本身对任务知识的掌握时间。
链接: https://arxiv.org/abs/2610.21025
作者: Matteo He,William F. Shen,Alex Iacob,Andrej Jovanovic,Xinchi Qiu,Nicholas D. Lane
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: NeurIPS 2026. 10 pages main text, 51 pages total. Code: this https URL
Abstract:When does a language model acquire a capability? Developmental studies usually answer this by tracking when models start producing the correct token. But that prediction passes through the unembedding matrix, or readout, which is itself still being learned during pretraining. We show that a model’s hidden states can encode task information before its own readout expresses it, so timelines read from outputs conflate what the model encodes with what its readout expresses. To track the readout directly, we introduce parameter-trajectory crosscoding, which fits one sparse dictionary to every token’s readout row at every checkpoint. Unlike crosscoders on activations, it needs only checkpoint weights. We use it to locate a narrow early window, 0.4-1.1% of the way through Pythia’s pretraining, in which the readout reorganizes and WordNet word categories become decodable from its rows. By then, linear probes on the hidden states already decode task information that the model’s own outputs do not yet show. To test whether such gaps depend on how fast the readout learns, we pretrain small models from scratch that differ only in the readout’s learning rate. We find that raising it makes the readout develop earlier and the gap close sooner, with little effect on final loss. Together, these results imply that early in training, where many developmental transitions are dated, a behavioral date can reflect when the readout learns to express task information and not when the model learns it.
[NLP-122] Can LLM s Anticipate Their Cruxes?
【速读】： 该论文旨在解决如何从语言模型（Language Model, LM）的决策输出中提取其推理核心依据的问题，即识别出影响模型最终决策的关键分歧点（crux）。传统上，语言模型会以完整文本形式输出推理过程，但其冗长的内容使得关键信息难以提取。为此，本文提出将“crux”定义为在多轮推理（rollouts）中模型对某一子问题存在分歧、且该分歧可能改变最终结论的特征，其本质是模型推理分布的属性而非单一推理路径的产物。为实现这一目标，论文设计了三种逐步增加计算成本的策略：第一，要求模型预先预测决策中的关键分歧点；第二，通过单次推理生成过程事后提取crux；第三，通过生成多条推理路径并比较其差异来识别crux，该方法借助低成本重排序器（reranker）显著降低计算开销（约为裁判模型计算量的1/86）。然而，实证评估表明，仅依赖事前预测或单次推理路径均无法充分捕捉所有crux，尤其在高风险决策场景下，必须采用多轮推理对比的方法才能全面揭示语言模型的决策逻辑空间。因此，解决方案的核心在于通过多轮推理与跨路径对比，结合高效轻量化重排序机制，实现对模型决策关键依据的可靠提取。
链接: https://arxiv.org/abs/2610.21006
作者: David Reber,James Reber,Ari Holtzman,Victor Veitch
机构: University of Chicago(芝加哥大学); University of Chicago(芝加哥大学); University of Chicago(芝加哥大学); University of Chicago(芝加哥大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:To use language models to help us make decisions we would like to understand the rationale behind the language model’s outputs. In principle, the language model provides its rationale as part of its text output. However, the sheer volume of text makes this hard to use in practice. In this paper, we are interested in extracting the cruxes that lead a model to recommend one action over another. We take a crux to be a sub-question that a model’s rollouts disagree on, where changing its value may have altered the conclusion. A crux is therefore a property of the distribution of a model’s rollouts, not of any one rollout. We consider three strategies of increasing computational cost. First, we just ask a model to identify, a priori, what the cruxes of a decision will be. Second, we have the model actually go through the process of generating a response and then extract the cruxes that arose in its reasoning. Finally, we generate many responses and extract the cruxes by comparing the reasoning across the rollouts. We make this last procedure computationally tractable with cheap calls to a reranker, an estimated 86 times less compute than a judge. We then conduct an extensive empirical evaluation to assess whether the computationally inexpensive options are sufficient. They are not. Across domains and models, we find that a model’s a priori assessment of the cruxes has little relation to the factors that are actually identified as important a posteriori. Using a single rollout helps, but is still inadequate for identifying the full set of cruxes. Accordingly, for high-stakes decisions (and to understand the space of language model decision making) the many rollout procedure is unavoidable.
[NLP-123] One Bias Score Two Factors: How Mitigations Affect Option Bias in Multiple-Choice Evaluation
【速读】： 该论文旨在解决大模型在多项选择题中因选项排列顺序不同而导致答案不一致的问题，即同时存在选择偏差（selection bias，对特定选项字母的偏好）与顺序敏感性（order sensitivity，答案随选项顺序变化）的现象。其核心解决方案是引入合并选择率标准差（pooled selection-rate standard deviation, pooled SD），该指标将两种现象统一为可量化的统计特征：当每个问题的所有选项在各位置出现频率均等时，pooled SD 可分解为两个无标签因子的乘积——翻转率（flip rate）与位置偏斜度（positional skew）。其中，翻转率指在不同排序下答案发生变化的问题比例，反映顺序敏感性；位置偏斜度则衡量在这些翻转问题中，选项字母被选择频率的离散程度，反映选择偏差。研究发现，先验校正主要降低位置偏斜度，而对每种排序及其逆序进行聚合可有效降低翻转率。两者均能缩小正确答案字母对应准确率的波动范围，但聚合方法会部分导致模型倾向于稳定选择同一错误选项。因此，作者建议在评估模型性能时，应同时报告翻转率与位置偏斜度以区分不同缓解策略的影响，并结合准确率判断稳定性与正确性之间的权衡。
链接: https://arxiv.org/abs/2610.20993
作者: Hyunjong Ok,Minjae Park,Jaeho Lee
机构: Pohang University of Science and Technology (POSTECH)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: preprint
Abstract:Reordering multiple-choice options can change a model’s answer, an effect studied as selection bias (a preference for certain option letters) and as order sensitivity (changes in answers across option orderings). We connect these perspectives through the pooled selection-rate standard deviation (pooled SD), which measures how unevenly option letters are selected across questions and orderings. When every option of each question appears equally often at each position, the pooled SD is exactly the product of two label-free factors: the flip rate and the positional skew. The flip rate is the fraction of questions whose chosen option changes across orderings; the positional skew is the standard deviation of option-letter selection rates among those questions. Across 41 checkpoints and 12 benchmarks, the flip rate tracks accuracy and output uncertainty, while the positional skew tracks the estimated option-symbol prior. Prior correction for selection bias mainly reduces the positional skew. Aggregating each ordering with its reversal, a mitigation for order sensitivity, mainly reduces the flip rate. Both reduce the range of accuracy across correct-answer letters by similar amounts, but aggregation leaves fewer model-rank reversals. Applying prior correction before aggregation reduces both factors and further narrows this accuracy range. However, aggregation lowers the flip rate partly by settling questions on the same wrong option. We therefore recommend reporting both factors to distinguish mitigation effects, alongside accuracy to distinguish stability from correctness.
[NLP-124] VELD-QA: A Large-Scale Dataset for Visual Evidence Grounding and Question Answering in Long Documents
【速读】： 该论文旨在解决现有长文档问答（Long Document QA）基准测试中普遍缺乏大规模人工构建问题与详细视觉证据标注的问题，从而限制了对证据定位与解释失败的深入诊断。其提出的解决方案核心是构建VELD-QA——一个大规模、面向长文档中视觉证据定位与问答的基准数据集，包含5,859份真实世界PDF文件（共372,445页）中的15,000组全人工撰写的问答对，涵盖图表、表格及复杂版式任务，并通过人工标注支持页和证据元素，明确描述所引用的视觉内容，实现对证据检索与解释的独立评估。关键创新在于引入细粒度的视觉元素标注，使模型不仅需定位正确页面，还需准确理解并推理视觉内容；实验表明，尽管采用检索增强与代理系统，基于检索的方法在真实支持页缺失时性能显著低于“理想情况”（oracle），且全文访问仍无法弥合差距，凸显视觉理解与推理仍是主要瓶颈。进一步地，将视觉元素标注融入输入可使性能差距缩小17%-24%，证明该标注在提升模型表现方面的实际价值，推动多模态文档问答的数据增强方向发展。
链接: https://arxiv.org/abs/2610.20964
作者: Zhivar Sourati,Marianne Menglin Liu,Rongguang Wang,Zheng Wang,Yazhe Hu,Lan Guo,Yuying Wang,Mengqing Guo,Bhavesh Dnyaneshwar Gore,Shubham Pathak,Ramapriya Ranganath,Sandeep Sharma,Tao Sheng,Sujith Ravi,Dan Roth
机构: University of Southern California(南加州大学); Oracle AI(甲骨文人工智能); Turing Enterprise Inc(图灵企业公司); University of Pennsylvania(宾夕法尼亚大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注:
Abstract:Existing long-document QA benchmarks rarely combine large-scale human-authored questions with detailed visual evidence annotations, limiting diagnosis of evidence localization and interpretation failures. We present VELD-QA, a large-scale benchmark for Visual Evidence grounding and question answering in Long Documents. It contains 15,000 fully human-authored question-answer pairs across 5,859 real-world PDFs totaling 372,445 pages, balancing chart, table, and complex-layout tasks. Human annotations identify supporting pages and evidence elements and describe cited visual content, enabling separate retrieval and interpretation evaluation. We evaluate nine vision-language models under four evidence settings and seven retrieval-augmented and agentic systems. Our analysis identifies bottlenecks in evidence localization and interpretation. Retrieval-based methods substantially underperform the oracle setting with ground-truth supporting pages; full-document access does not close the gap. Localization alone is insufficient: oracle accuracy peaks at 81.7%, highlighting persistent visual interpretation and reasoning challenges. Enriching inputs with our visual-element annotations narrows the performance gap by 17-24%, demonstrating value beyond evaluation and motivating data enrichment for multimodal document QA. Our code is available at this https URL.
[NLP-125] Same Context Different Agents : A Behavioral Study of Semantic Layers Orchestration and Compaction in Data Agents
【速读】： 该论文旨在解决异构数据环境下问答系统中多源证据融合的挑战，即如何有效整合文件、数据库、文档及网络信息以支持复杂长周期任务的准确回答。其核心问题在于：尽管数据代理（data agents）能够自动化这一过程，但其表现高度依赖于上下文质量，而现有上下文工程技巧（Context-Engineering Techniques, CETs）——包括语义层（semantic layers）、编排（orchestration）与压缩（compaction）——在单独应用时虽被验证有效，但缺乏对其协同作用机制及对代理行为影响的系统性理解。本文的关键解决方案是通过大规模实证分析，对三种CETs在七种大语言模型（LLMs）上针对1,514个长周期任务进行系统评估，揭示了不同代理行为模式（如调查者、过早回答者、早期放弃者）与CETs之间的匹配关系。研究发现，每种CET仅在其适配的行为类型上带来准确性提升；例如，编排对过早回答者有显著改善，但对调查者反而造成性能下降（最高损失达5个百分点），因语义层抑制探索、编排引发过度自信、压缩导致重复操作。此外，组合使用CETs的增益小于各技术独立效果之和，表明存在非线性交互效应。由此提出基于代理行为特征、任务需求及准确率-成本权衡的择优策略，为动态选择最优上下文工程方案提供可解释的指导原则。
链接: https://arxiv.org/abs/2610.20959
作者: Zhivar Sourati,Marianne Menglin Liu,Zheng Wang,Rongguang Wang,Sai Ashish Somayajula,Tao Sheng,Kyu J. Han,Sujith Ravi,Dan Roth
机构: University of Southern California(南加州大学); Oracle AI; University of Pennsylvania(宾夕法尼亚大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Answering questions over heterogeneous data means combining evidence from files, databases, documents and the web. Data agents automate this, but their success depends on context. Three context-engineering techniques (CETs) help by shaping it: semantic layers describe the data, orchestration splits work among agents, and compaction condenses interaction history. Existing studies validate each CET alone, leaving open how agent behavior shapes its effect. We present a comprehensive analysis of the three CETs, alone and combined, with seven large language models (LLMs) on 1,514 long-horizon tasks from five benchmarks, covering behavior, accuracy and cost. We find that data agents without CETs form three behavior groups: investigators, premature answerers that answer without inspecting data, and early stoppers that give up. Each CET delivers its known benefit, from more investigation to smaller prompts, but only the behavior group it matches turns that benefit into accuracy, such as premature answerers under orchestration. For investigators, the same CETs can hurt: the semantic layer stops them from looking, orchestration makes them overconfident, costing 5 percentage points, and compaction makes them repeat steps. Combining CETs yields smaller gains and losses than their sum. We derive guidance for choosing CETs by agent behavior, task demands and accuracy-cost trade-offs.
[NLP-126] LEV: Thinking Fast and Deep with Looped Decision Models
【速读】： 该论文旨在解决传统决策模型（decision models）在处理复杂问题时深度不足，而基于生成式推理的模型又因逐词解码导致计算成本过高的矛盾问题。现有决策模型虽能快速响应，但其推理深度固定于训练阶段，无法根据问题难度动态调整；而生成式推理模型虽可通过文本生成实现深层思考，却需为每一步生成额外的解码令牌，造成显著计算开销。本文提出LEV（Looped Evaluation via Repeated Inference），其核心创新在于将预训练语言模型改造为可循环执行的结构——通过在多个迭代步骤中重复运行网络层而非生成文本来实现“深度思考”。该方法在每次循环后读取选项并联合优化所有步骤，使得单个检查点即可支持任意推理深度。实验表明，在相同权重下，随着推理深度增加，模型在困难样本上的准确率显著提升（如2.6B模型在JevBench上从深度1的63.2%升至深度4的76.2%）。进一步引入轻量级“停止器”（stopper）模块，根据未标注请求拟合阈值以按需设定推理深度，可在保持精度损失小于1个百分点的前提下节省15%至24%的计算量，性能优于固定深度混合策略。在延迟低于50毫秒的封闭测试集上，1.4B与3B的LEV模型分别比同延迟下的密集型（non-looped）模型和级联模型高出5.4至8.8个百分点，验证了其高效且强大的推理能力。
链接: https://arxiv.org/abs/2610.20951
作者: Dingjie Song,Zhiling Yan,Yuxuan Zhang,Lichao Sun
机构: Lehigh University(莱赫大学); University of British Columbia(不列颠哥伦比亚大学); Vector Institute(向量研究所)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 28 pages. Code: this https URL
Abstract:Decision models answer typed questions about an input with one probability per option, computed in a single forward pass without generating text. Like the fast System 1 of dual-process theory, they answer at once, but their depth is fixed at training time, so they spend the same computation on easy and hard questions. Models that reason in generated text can think longer, but they pay for every extra step in decoded tokens. We present LEV, a decision model that thinks deeper by rerunning its layers in a loop instead of generating tokens. LEV builds on a language model pretrained as a loop, reads the options after every loop step and trains all steps jointly, so one checkpoint can answer at any depth. On three looped backbones, accuracy rises with depth on the same weights, mostly on hard questions, and a 2.6B LEV goes from 63.2% at depth one to 76.2% at depth four on JevBench. On-demand depth, set per question by a small stopper with a threshold fitted on unlabelled requests, saves 15% to 24% of the computation at a cost of under one accuracy point, level with or slightly above a fixed mix of depths. On a locked test set, the 1.4B and 3B models are 5.4 to 8.8 points more accurate than dense (non-looped) models and cascades of the same latency below 50 ms.
[NLP-127] Prune the Summary Not the History: Source-Judged Compaction for Long-Horizon Agents
【速读】： 该论文旨在解决长时程大语言模型（LLM）智能体在上下文压缩过程中因摘要内容混杂不可靠推理而导致的决策错误问题。其核心挑战在于：当前主流的结构化压缩器生成的摘要不仅包含环境实际输出的事实，还融入了压缩器自身生成的推理与决策内容，而智能体无法区分二者，从而导致后续行动基于无依据的摘要条目执行。研究发现，在广泛使用的压缩器中，83%的失败运行可追溯至执行了缺乏环境输出支持的摘要条目。为此，论文提出无需训练的后处理方案——源判别压缩（Source-Judged Compaction, SJC），其关键在于引入一个轻量级非生成式判别模型Jev，对每个摘要条目进行溯源判断，评估其是否源自环境输出或前序摘要，并移除无源条目。SJC不改变原有智能体或压缩器，仅在每次压缩时增加一次低成本的判断请求，显著提升了跨运行可靠性（在AppWorld和OfficeBench上提升2.1%–8.9%），同时保持智能体的令牌消耗不变。
链接: https://arxiv.org/abs/2610.20927
作者: Guanghui Min,Chen Chen
机构: University of Virginia(弗吉尼亚大学)
类目: Computation and Language (cs.CL)
备注: Uses Jev, a non-generative decision model, as the source judge. Code: this https URL
Abstract:Long-horizon LLM agents rely on context compaction: when the interaction history outgrows the context window, a compressor replaces it with a natural-language summary. We show that these summaries contain more than a record of what happened. Compaction prompts ask for decisions and next steps, so each summary mixes facts established by the environment with reasoning written by the compressor, and the agent cannot tell the two apart afterwards. In an analysis of a widely used structured compressor, 83% of failed runs trace back to an action that carries out a specific summary entry, most of which no environment output supports. We propose Source-Judged Compaction (SJC), a training-free post-processing step that scores each summary entry for whether it has a source in the environment outputs or the previous summary, and removes the entries that do not. It changes neither the agent nor the compressor and adds one inexpensive judge request per compaction. The judge is Jev, a non-generative decision model that answers the source question with a calibrated probability instead of text. On AppWorld and OfficeBench with two agent models, SJC improves cross-run reliability over the underlying compressor by 2.1%-8.9% without increasing the agent’s token usage.
[NLP-128] SkillNEO: Non-Elitist Evolutionary Optimization of Agent Skills
【速读】： 该论文旨在解决大语言模型（LLM）代理在技能演化过程中因依赖即时奖励而忽略具有潜在价值的低分版本，导致部分有效改进被过早丢弃的问题。其核心挑战在于如何有效识别并保留那些虽当前表现不佳但具备后续优化潜力的中间版本，以实现对技能的持续迭代与积累。解决方案的关键是提出一种非精英式进化优化框架——SkillNEO（Non-Elitist Evolutionary Optimization for Agent Skills），其创新性地将“选择使用版本”与“保留改进版本”分离：通过对比相关版本在同一任务中的执行表现，系统可识别出已成功的工作步骤予以保留，并定位仍存在的失败环节进行修复；同时支持基于后期反馈回溯性地修订早期版本。实验在86个涵盖软件工程与办公自动化等八类任务的SkillsBench基准上验证，SkillNEO使GPT-5.4 Mini的平均奖励从49.1%提升至66.4%，达到原版GPT-5.5和GPT-5.6 Luna的性能区间。进一步发现，经修改的技能可在不重新编辑的情况下直接复用于更强模型，显著提升其表现（如Luna提升10.5个百分点，GPT-5.5超越原始参考值10.6个百分点）。执行日志显示，即使弱模型未能完全满足任务要求，其部分进展仍可转化为可供强模型利用的可复用程序化知识。这些结果表明，非精英式演化机制能够将弱模型的阶段性成果转化为可继承、可优化的代理技能，从而实现跨模型的知识迁移与能力跃升。
链接: https://arxiv.org/abs/2610.20926
作者: Enze Ma,Chunyu Miao,Yankai Chen,Jie Yang,Liancheng Fang,Kening Zheng,Hanrong Zhang,Muyao Xu,Xinle Yu,Chengze Li,Wei-Chieh Huang,Ziwei Ren,Yunhao Du,Zhixiang Shen,Xue Liu,Philip S. Yu,Zhen Wang
机构: University of Illinois at Chicago (伊利诺伊大学芝加哥分校); Mohamed bin Zayed University of Artificial Intelligence (穆罕默德·本·扎耶德人工智能大学); University of California, San Diego (加利福尼亚大学圣迭戈分校); University of California, Los Angeles (加利福尼亚大学洛杉矶分校); University of Texas at Arlington (得克萨斯大学阿灵顿分校); Independent Researcher (独立研究员)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Agent skills turn procedural knowledge into reusable artifacts, allowing LLM agents to accumulate and refine expertise through experience. Building on this, skill evolution iteratively improves skills using execution feedback. However, improvement may require several interdependent edits, so low-reward versions can still be worth developing. Requiring immediate reward gains can discard useful partial repairs before their benefits emerge. Non-elitist evolution allows new versions to receive further edits despite scoring below earlier ones. For agent skills, this calls for evidence of what to preserve and repair next. We introduce SkillNEO (Non-Elitist Evolutionary Optimization for Agent Skills), which separately selects a version to use and versions to keep improving. Comparing how related versions perform the same task reveals working steps to preserve and remaining failures to fix. It can also revisit and revise earlier versions using later evidence. Across 86 SkillsBench tasks in eight domains, including software engineering and office work, SkillNEO raises GPT-5.4 Mini’s average reward from 49.1% to 66.4%, within GPT-5.5 and GPT-5.6 Luna’s original-skill range. Mini executes tasks, proposes revisions, and edits skills; final evaluation uses the same task instances. Gains span all eight domains. The same skills improve stronger models without further editing. On the 41 tasks with modified skills, reuse improves Luna by 10.5 percentage points and places GPT-5.5 10.6 points above its published original-skill reference. Execution records show these skills can benefit stronger models even when Mini still fails some task requirements. Search histories also show lower-scoring versions leading to skills ultimately selected for use. Together, these findings show that skill evolution can turn a weaker model’s partial progress into reusable procedural knowledge for improving more capable agents.
[NLP-129] ask-Centric Multimodal Learning for Cognitive Impairment Detection
【速读】： 该论文旨在解决老年人认知衰退（如轻度认知障碍，MCI）在非临床环境中难以实现早期检测的问题。传统诊断模型主要依赖文本或语音信号，仅能捕捉人类认知表现的单一维度，而忽视了人类表达的多模态本质。为此，本文提出一种无接触、多模态的筛查框架，通过亚马逊Alexa设备在结构化语音交互过程中同步采集视频、音频与文本信号，以更全面地表征个体在言语模式、声学特征、面部表情及语言结构等方面的自然行为表现。研究的关键在于利用多模态融合策略，整合来自不同感官通道的行为信息，从而提升对认知功能下降的识别能力。实验结果表明，尽管单模态模型在特定任务中表现各异，但多模态融合模型在多个任务上均展现出稳定且更高的性能（准确率0.76，AUC 0.73–0.75），显著优于单一模态方法，验证了自然情境下多模态行为信号在真实世界认知评估中的有效性，推动了从传统静态、单模态诊断范式向动态、多模态、可扩展的智能筛查体系的转变。
链接: https://arxiv.org/abs/2610.20906
作者: Sankalp Vaish,Rishank Singh,Sumukh Veeramalla,Caroline Summerour,John A. Batsis,Xiaohui Liang
机构: University of Massachusetts Boston(马萨诸塞大学波士顿分校); University of North Carolina at Chapel Hill(北卡罗来纳大学教堂山分校)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Accepted for publication at the 2026 IEEE Global Communications Conference (GLOBECOM)
Abstract:Dementia affects millions of older adults worldwide, yet early detection remains difficult outside clinical settings. Traditional models primarily rely on text or voice-based responses, capturing only a limited view of human cognition despite the inherently multimodal nature of human expression. This work proposes a contactless, multimodal screening framework that leverages video, audio, and text signals collected during structured voice-based interactions with Amazon Alexa. By incorporating these modalities, our approach captures a more holistic representation of human behavior encompassing speech patterns, vocal characteristics, facial expressions, and linguistic structure, which may enable more accurate detection of cognitive decline. Different participants completed three conversational tasks designed to probe spontaneous speech, comprehension, and recall. We pursued three objectives: (1) evaluate which of the three Alexa tasks provides the most informative cognitive signals, (2) compare the effectiveness of unimodal models, and (3) assess the benefits of multimodal fusion models. Task-level analysis shows that the video model achieves its best performance on Command-Reading, with an accuracy of 0.65 (AUC 0.69), while the audio model performs best on Question-Answering, reaching an accuracy of 0.67 (AUC 0.73). In contrast, the multimodal model demonstrates consistently stable performance across tasks, achieving an accuracy of 0.76 (AUC 0.75) on Command-Reading and 0.76 (AUC 0.73) on Question-Answering. These findings demonstrate that multimodal behavioral cues captured through naturalistic interactions can reliably distinguish MCI from healthy aging, highlighting the limitations of traditional diagnostic paradigms and supporting the shift toward multimodal, real-world cognitive assessment frameworks.
[NLP-130] Wrong Votes Right Gradients: Why Majority-Vote Rewards Track Verifiable RL
【速读】： 该论文试图解决的问题是：在强化学习中，利用模型自生成的标签（如基于模型自身样本的多数投票）为何能够获得接近使用可验证奖励时的性能提升，这一现象从传统的提示层面（prompt-level）视角来看难以解释——因为多数投票仅返回模型的最可能输出，无法提供关于正确性的额外信息。其解决方案的关键在于重新审视更新单位的层级：从提示层面转向批次层面（batch-level update）。研究指出，在实际批量大小下，单个提示的梯度对自身输出变化的贡献不足1%，因此错误标签的影响仅取决于标签错误是否在不同提示间具有方向一致性。通过将多数投票更新分解为相对于理想更新（gold update）的收缩（shrinkage）与误导（misdirection）分量，研究发现，当标签错误在提示间无相干性时，误导项随批量增大趋于消失，即使标签错误频率较高亦然。实证分析表明，多数投票产生的误差在响应级梯度上表现出与随机奖励基线相当的无相干性，且其与理想更新的噪声校正余弦相似度在误差范围内等于1；随机翻转最多30%的黄金标签仍能保持更新方向，而由统一规则翻转相同比例则破坏该方向；此外，对所有提示均偏好某种响应形式的奖励模型，其效果也显著逊于真实奖励。该等价关系在各检查点均成立，但具有局部性：尽管自生成标签可近似替代验证奖励，真实黄金奖励训练在保留数据上的多数投票准确率仍更优。因此，决定自生成标签能否有效替代验证的关键因素是错误的相干性（error coherence），而非错误率本身。
链接: https://arxiv.org/abs/2610.20898
作者: Dylan P. Wilson,Chengsong Huang,Jiaxin Huang
机构: Dartmouth College (达特茅斯学院); Washington University in St. Louis (圣路易斯华盛顿大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Reinforcement learning with self-generated labels, such as majority vote over a model’s own samples, often recovers most of the gain of training with verifiable rewards. This is puzzling from the usual prompt-level view, in which majority vote only returns a model’s modal answer and cannot add information about correctness. We argue that the relevant unit is the batch-level update. Because parameters are shared, a prompt’s own gradient accounts for less than 1% of the change in that prompt’s output at practical batch sizes, so a wrong label matters only to the extent that label errors share a direction across prompts. We decompose the majority-vote update into shrinkage and misdirection relative to the gold update, and show that misdirection vanishes with batch size whenever errors are incoherent across prompts, however often the labels are wrong. Measuring per-response gradients on five models, we find that majority-vote errors are as incoherent as a random-reward null, that the majority-vote update agrees with the gold update as closely as two independent gold updates agree with each other, and that its noise-corrected cosine with the gold update is 1 within error. Randomly flipping up to 30% of gold labels leaves the update direction intact, while the same fraction flipped by a shared rule destroys it, and rewards that favor a kind of response on every prompt, such as a learned reward model, fall measurably short. The equivalence holds at every checkpoint we examine but is local: gold-reward training still ends ahead on held-out majority accuracy. Error coherence, not error rate, determines when self-generated labels can stand in for verification.
[NLP-131] CogDrift-4D: Hierarchical Multimodal Modeling of Temporal Behavioral Decay for MCI Detection ALT
【速读】： 该论文旨在解决阿尔茨海默病（Alzheimer’s Disease, AD）早期筛查中轻度认知障碍（Mild Cognitive Impairment, MCI）难以实现早期精准识别的核心挑战，尤其针对传统方法在捕捉细微、渐进性行为退化方面的局限性。其核心问题在于如何有效建模个体在长时间跨度内多模态行为信号的连续动态变化，以实现对认知衰退轨迹的敏感检测。解决方案的关键在于提出一种新型的分层四维时空变换器框架——CogDrift-4D，该框架通过构建一个密集的4D张量（批量 × 会话 × 时间 × 特征），系统性保留了局部动态与长期演化特征。其创新性体现在双层级行为疲劳与漂移建模机制：在会话内层面，采用局部最小二乘回归斜率计算帧级面部动作单元（Facial Action Units）与声学能量的瞬时衰减趋势；在会话间层面，引入递归交叉注意力反馈编码器作为累积疲劳累积器，实现跨会话行为漂移的建模。同时，设计局部质量门控机制以动态过滤低质量特征，并实现语义信息与原始物理动态的早期融合，显著提升对环境噪声和数据失效的鲁棒性。实验表明，该框架在结构化语言交互任务中达到0.733 ± 0.13的分类准确率与0.716 ± 0.14的F1分数，优于现有先进基线模型，为家庭场景下的实时、可扩展、数学严谨的痴呆症筛查提供了坚实的技术基础。
链接: https://arxiv.org/abs/2610.20877
作者: Sankalp Vaish,Sumukh Veeramalla,Rishank Singh,Caroline Summerour,John A. Batsis,Xiaohui Liang
机构: University of Massachusetts Boston (马萨诸塞大学波士顿分校); University of North Carolina at Chapel Hill (北卡罗来纳大学教堂山分校)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Accepted for publication at the 2026 IEEE International Conference on E-health Networking, Application Services (Healthcom)
Abstract:Early detection of mild cognitive impairment remains a critical bottleneck in dementia care, demanding scalable tools capable of capturing subtle behavioral decay. To capture the subtle temporal changes of cognitive decline, we introduce a behavioral dataset collected across seven unscripted multimodal interactions over an 18-month horizon. This dataset formalizes continuous, multi-session interactions into a dense 4D tensor (Batch * Session * Time * Feature) that explicitly preserves concurrent local dynamics alongside long-term changes. To effectively process these complex behavioral patterns, we propose CogDrift-4D: a novel, hierarchical 4D Transformer framework. To capture the complex trajectory of cognitive decline, CogDrift-4D implements a dual-level behavioral fatigue and drift modeling architecture. At the intra-session level, the model computes localized least-squares regression slopes to quantify the acute ``fading’’ of frame-level facial action units and acoustic energy. At the inter-session level, a recursive cross-attention feedback encoder acts as a cumulative fatigue accumulator, mapping the patient’s continuous behavioral drift across successive sessions. To mitigate environmental noise and track failures across sessions, a localized Quality Gate dynamically filters low-integrity features before final fusion, also an early fusion that integrates semantic insights with raw physical dynamics. Evaluated on structured verbal and linguistic interactions, our architecture successfully detects MCI with a classification accuracy of 0.733 ± 0.13 and an F1-score of 0.716 ± 0.14, outperforming advanced baselines. The framework provides a mathematically rigorous, scalable, and robust foundation for home-based, real-time dementia screening.
[NLP-132] Does Morphological Correctness Matter for Tokenization? A Scrambled Control Study in Tagalog
【速读】： 该论文旨在解决传统基于频率驱动的分词方法（如BPE）在处理具有非连续词根的语言时，因破坏形态边界而损失语义信息的问题。其核心挑战在于：尽管已有研究认为保留形态边界有助于提升模型性能，但这些研究通常默认边界必须具备语言学正确性，而未区分“形态边界保持”与“语言学正确性”之间的差异。本文的关键解决方案是提出一种新型分词器MorphAustro-Tok，专为菲律宾语（Tagalog）设计，该分词器将非连续词根（如通过中缀标记语态的词形变化）以连续单元输出，并辅以位置中缀标记，从而在不依赖语言学正确性的前提下实现形态结构的显式编码。实验表明，在33次运行中，MorphAustro-Tok相较于匹配的BPE基线分别在三个不同粒度水平上提升了0.84、1.49和3.40的句级chrF++得分；且该增益无法由分词粒度或模型容量解释——在最粗粒度下，随机打乱形态成分的对照组反而拥有更多参数但表现更差。进一步地，对形态分析进行随机化处理后性能下降幅度小于种子间波动，表明在当前规模下，形态正确性并非性能提升的主要因素，关键在于形态结构的显式建模本身。
链接: https://arxiv.org/abs/2610.20859
作者: Abdul Rehman Antall,Baidya Nath Saha
机构: National University of Computer and Emerging Sciences (FAST-NUCES), Lahore, Pakistan; Concordia University of Edmonton, Alberta, Canada
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 13 pages, 1 figure, 7 tables. Code and replication package: this https URL
Abstract:Typology-aware tokenizers are motivated by the claim that segmentation which respects morpheme boundaries preserves information that frequency-driven segmentation destroys. Studies routinely establish that it helps while assuming the boundaries must be linguistically correct. We separate them. Tagalog marks voice with infixes (s-um-ulat ‘wrote’, s-in-ulat ‘was written’), so the root is not contiguous and no sequence of BPE merges can recover it. We build MorphAustro-Tok, a tokenizer that emits the root as a contiguous unit followed by a positional infix token, and compare it against a corpus-matched BPE baseline and, at two of three fertility levels, against a scrambled-morpheme control that preserves morph count, invertibility and per-type consistency while randomising the pseudo-infix. Across 33 runs, MorphAustro-Tok improves Tagalog machine translation over matched BPE by 0.84, 1.49 and 3.40 mean sentence-level chrF++ points. Token count does not explain the gain, since it grows as segmentation coarsens, and neither does capacity: at the coarsest level the scrambled arm carries more parameters and scores lower. Scrambling the analysis moves the gain by less than between-seed dispersion at both controlled levels, so we detect no contribution from morphological correctness at this scale.
[NLP-133] Probing Negation Understanding in Vision-Language Models With Object Attention EMNLP
【速读】： 该论文旨在解决视觉语言模型（Vision-Language Models, VLMs）在处理否定句时存在的理解偏差问题，尤其关注模型注意力机制是否能模拟人类在处理肯定与否定语句时的注意力模式。研究通过分析不同模型在处理否定句时的注意力分布，对比人类眼动轨迹与模型注意力之间的差异，发现现有模型在面对否定句时普遍未能遵循类人注意力模式，表明其对否定的理解并非依赖于注意力机制本身，而是依赖于其他隐含或非注意力驱动的机制。尽管层间分析显示PaliGemma2-3B在否定条件下会将注意力转移到被否定对象的替代项上，但头级消融实验表明这一现象仅为副产物而非真正的否定处理机制；此外，还发现LLaVA-1.5-7B存在对左侧物体的空间偏好。下游任务评估进一步揭示所有模型在肯定与否定问题上的表现存在显著差距，凸显了当前模型在深层否定语义理解方面的不足。因此，该研究的关键在于揭示现有VLMs在否定处理中的注意力失效问题，并指出需深入探索其内在推理机制以缩小性能差距。
链接: https://arxiv.org/abs/2610.20852
作者: Beth Pearson,Michael Wray,Martha Lewis
机构: University of Bristol(布里斯托大学); ILLC, University of Amsterdam(阿姆斯特丹大学信息语言与计算研究所)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: EMNLP Findings 2026
Abstract:Negation is a fundamental aspect of language understanding and it is important vision-language models (VLMs) are able to process it effectively. We investigate how VLMs process negation by analysing attention patterns of different models. We draw comparisons between human gaze and attention in models when processing affirmative vs negated statements. We compare object level model attention to human eye-fixations in both binary and multary settings, evaluating with both synthetic and realistic datasets. We find models generally do not follow human-like attention patterns when resolving negation, suggesting they rely on mechanisms outside of attention. While layerwise analysis shows that PaliGemma2-3B shifts object attention to the alternative under negation, our headwise ablations suggest this is a byproduct rather than a negation processing mechanism. Additionally, we find that LLaVA-1.5-7B displays a spatial bias towards left-side objects. Our evaluation of models on downstream tasks shows that all models have a discrepancy between affirmative and negated questions highlighting the need to better understand the underlying processing of negation to close this performance gap. Code available at this https URL
[NLP-134] Evolving Together Failing Apart: The Skill-Policy Co-evolution Trap in Self-Improving Agents
【速读】： 该论文旨在解决生成式智能体在技能-策略共进化（skill-policy co-evolution）过程中，因部署后技能库发生漂移而引发的性能退化问题。现有方法在训练时仅评估策略模型在最终生成的技能库（即“交付技能库”，delivered bank）上的表现，但实际部署中技能库常被修改、增删，导致策略运行于非预期的漂移技能库上，而当前训练目标与评估协议均未考虑此类动态变化。论文通过理论与实证分析揭示，现有方法普遍采用“合作性接受”（cooperative acceptance）机制，使策略能力与其共进化技能库紧密绑定，从而在技能漂移时表现出显著脆弱性，且这种敏感性随共进化训练进程加剧。为此，论文提出面向部署的评估协议——共进化漂移审计（CO-evolution Drift Audit, CODA），通过模拟技能漂移场景，引入三项新指标：在语义保持编辑下的稳定性（consistency rate）、对受损技能的抵抗能力（resistance rate）以及最坏情况下的成功率。实验表明，在四个共进化方法和五个基准任务（六种设置）中，最坏情况下的成功率较交付对峰值低5–13分，相当于抹除了交付技能库相对于无技能基线所获收益的52%–93%。因此，CODA将共进化的目标从“在特定共进化技能库上达到峰值性能”重构为“在技能库任意演化情形下仍能有效利用技能”，并为未来方法提供了统一的可审计标准。
链接: https://arxiv.org/abs/2610.20845
作者: Yujie Feng,Xiaobin Wang,Jinluan Yang,Xiaonan Shi,Pengjun Xie,Philip S. Yu,Xiaoming Wu,Jieping Ye
机构: Alibaba Token Hub, Alibaba Group(阿里巴巴通证中心，阿里巴巴集团); The Hong Kong Polytechnic University(香港理工大学); University of Illinois at Chicago(伊利诺伊大学芝加哥分校)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Under Review
Abstract:Skill-policy co-evolution, in which an agent’s policy model and its skill bank improve each other during training, is a promising paradigm for self-improving agents. Current evaluations score the co-evolved policy on the final bank it trained with (the delivered bank). After deployment, however, skills may be rewritten, added or removed more frequently than the policy model is retrained, leaving the policy to operate on a drifted skill bank that neither the training objective nor existing evaluation protocols address. In this paper, we first establish, through theoretical and empirical analysis, that the fragility hidden by this blind spot is not incidental. The acceptance rule these methods share (cooperative acceptance) ties a policy’s competence to the bank it co-evolved with, leaving it fragile once skills drift, regardless of the implementation; empirically, sensitivity to skill drift grows as co-evolution training proceeds. We therefore introduce the CO-evolution Drift Audit (CODA), a deployment-oriented evaluation protocol that simulates skill drift and, alongside success rate under worst-case drift, introduces two new metrics: a consistency rate under meaning-preserving edits and a resistance rate to corrupted skills. Across four co-evolution methods and five benchmarks (six settings), worst-case success is 5-13 points below the peak measured for each delivered pair, erasing 52-93% of the delivered bank’s gain over the no-skill baseline. CODA thus reframes the target of co-evolution, from peak performance on the co-evolved bank to the ability to harness skills however they evolve, and provides a shared standard against which future methods can be audited.
[NLP-135] Confidence-Calibrated Inference Expansion for Evaluator-Guided Test-Time Reasoning NEURIPS2026
【速读】： 该论文旨在解决生成式 AI（Generative AI）在评估引导推理过程中每一步面临的局部控制问题：即在当前答案可信、返回备用方案或继续推理之间做出决策。其核心解决方案是提出一种计算感知的三动作控制器，通过预测答案成功率来评分终止动作，并采用基于回溯（rollout-based）的延续成功率代理指标来评估扩展动作，同时显式地对计算成本进行定价。关键创新在于通过局部动作差距分析刻画评分误差可能改变决策的情形，并引入期望后悔上界以量化实际扩展路径上的误差影响。基于此分析，模型学习并校准针对不同动作的置信度估计，并为模糊决策施加局部不确定性惩罚。在保留宽度扩展与修订任务的独立测试中，该控制器在提升观测效用的同时，分别将归一化在线计算量降低了约30%和16%，且在四个 Qwen3-8B 种子下的计算开销低于拟合的 Q 值方法，而效用差异较小且不一致。研究结果表明，评估引导推理本质上是在不确定性下的资源分配问题，而非单纯更强的推理机制。
链接: https://arxiv.org/abs/2610.20836
作者: Weida Liang,Kenji Kawaguchi
机构: National University of Singapore(新加坡国立大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026
Abstract:Evaluator-guided reasoning presents a local control problem at every step: trust the current answer, return a fallback, or spend more inference. We formulate this choice as a compute-aware three-action controller, scoring terminal actions by predicted answer success and expansion by a rollout-based continuation-success surrogate, with compute cost priced explicitly. A local action-gap analysis characterizes when score errors can change a decision, while an expected regret bound accounts for errors along the realized expansion path. Guided by this analysis, we learn and calibrate action-specific confidence estimates, with local uncertainty penalties for ambiguous decisions. On held-out width expansion and revision tasks, the controller improves observed utility over calibration-selected fixed policies while reducing normalized online compute by about 30% and 16%, respectively. Across four Qwen3-8B seeds, compute is lower than fitted Q in both settings, while utility differences are small and mixed. The results frame evaluator-guided reasoning as resource allocation under uncertainty, rather than a uniformly stronger reasoner.
[NLP-136] PEDBench: Can Speech Models Detect Phonetic Errors?
【速读】： 该论文旨在解决语音错误检测（Phonetic Error Detection, PED）中因缺乏高质量评估数据而导致的细粒度语音错误检测能力不可靠的问题。现有语音模型在真实场景下的错误识别性能尚不明确，尤其在自然语音与受控语音之间存在显著性能差异。其解决方案的关键在于构建PEDBench——首个涵盖从自然语音到受控语音连续谱的基准数据集，包含334名儿童的真实语音错误、1,188条由人工演员演绎的脚本化错误以及1,083条基于音素条件文本转语音（phone-conditioned text-to-speech）生成的合成错误，并通过两位具备音系学训练背景的标注员进行逐音素标注与验证。实验表明，尽管使用合成数据微调的自研语音-语言模型在词级PED任务上优于商业及开源基线，但在整体检测精度和泛化能力上仍存在明显不足，且普遍存在将正确发音误判为错误的现象，揭示了“仿真到真实”（sim-to-real）域差距依然显著。研究强调了高质量标注数据与可控生成数据对提升PED系统性能的重要性，并计划公开代码、训练方案及数据集以推动该领域发展。
链接: https://arxiv.org/abs/2610.20830
作者: Zhonghao Shi,Zackary Rackauckas,Jasmine Yip,Savannah Halpern,Simon Campos,Lauren Brace,Emma Yang,Tiantian Feng,Xuan Shi,Shrikanth Narayanan,Teresa Satterfield,Ying Xu
机构: Harvard University(哈佛大学); University of Southern California(南加州大学); University of Michigan(密歇根大学); Columbia University(哥伦比亚大学)
类目: Computation and Language (cs.CL); Sound (cs.SD)
备注: 21 pages, 10 figures
Abstract:Phonetic error detection (PED) aims to identify phoneme-level substitutions, deletions, and insertions and is essential for educational and clinical speech assessment and intervention. Despite recent advances in speech models for phoneme recognition, their ability to reliably detect fine-grained phonetic errors remains unclear because of a lack of high-quality evaluation data. To address this gap, we introduce PEDBench, the first PED benchmark spanning a continuum from natural to controlled speech. It comprises real-world recordings of spoken errors made by children (N=334), scripted error productions by human actors (N=1,188), and synthetic error productions generated using phone-conditioned text-to-speech (N=1,083). All phonetic annotations were transcribed and verified by two phonetically trained annotators. Our evaluation of 13 state-of-the-art speech models shows that they fail to reliably detect phonetic errors, even under controlled conditions with acted and synthetic speech, where acoustic variability is reduced. The models also exhibit low precision, frequently misclassifying correctly pronounced phonemes as errors. Our speech-language models, fine-tuned on synthetic data, outperform commercial and open-source baselines on word-level PED, although a sim-to-real domain gap remains. We plan to publicly release our code and recipes, along with our acted and synthetic datasets, to support the development of more accurate PED systems.
[NLP-137] CoTranslator: Reverse to Forward Chain-of-Thought Translator for Reasoning Distillation
【速读】： 该论文旨在解决生成高质量、可验证的思维链（Chain-of-Thought, CoT）训练数据在实际应用中效率低下的问题。现有方法如正向思维链生成（Forward CoT Generation, FCG）结合拒绝采样（Rejection Sampling, RS）虽能生成正确且可信的推理路径，但对复杂问题需大量教师模型调用，计算成本高昂；而反向思维链生成（Reverse CoT Generation, RCG）虽计算开销小，却因依赖已知答案可能导致事后合理化（post-hoc rationalization），与学生模型无答案推理的设定不一致。本文提出CoTranslator，其核心在于学习将反向生成的推理路径翻译至正向验证响应的分布，从而构建高效、可靠的学生模型训练数据。该方法仅需少量成对的反向与拒绝采样正向响应进行训练，即可大规模生成并验证学生模型所需数据。实验表明，在六项数学基准测试中，CoTranslator相较语义抑制法（semantic suppression）平均准确率提升最高达5.82个百分点；对于4B规模的学生模型，其性能保留了拒绝采样95.0%的平均准确率，同时仅消耗35.9%的估算算力。此外，无需重新训练，同一翻译器可跨数据集、跨教师模型及应用于科学、技术、工程与数学（STEM）领域，分别使4B和8B学生模型的平均准确率提升4.88和4.31个百分点。结果表明，CoTranslator为推理蒸馏提供了一种高效且具备强泛化能力的新范式。
链接: https://arxiv.org/abs/2610.20828
作者: Zinan Tang,Zhuoshi Pan,Kairui Chen,Ze Yan,Shaomian Zheng,Yukun Zhang,Gang Wang,Biqing Huang
机构: Tsinghua University(清华大学); Ant Group(蚂蚁集团); Xiaomi Corporation(小米公司)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 18 pages, 12 figures, under review
Abstract:Reasoning distillation requires chain-of-thought (CoT) data with correct answers, but obtaining such data efficiently remains challenging. Forward CoT generation (FCG) followed by rejection sampling (RS) produces verified, question-only responses, yet may require many teacher calls for difficult problems. Reverse CoT generation (RCG) is cheaper because it conditions on the known answer, but access to this answer can lead to post-hoc rationalization and a mismatch with the student’s answer-free inference setting. We propose CoTranslator, which learns to translate reverse CoT responses toward the distribution of verified forward responses. Trained on a small set of paired reverse and rejection-sampled forward responses, it supports large-scale generation and verification of student training data. Across six student models evaluated on six mathematics benchmarks, CoTranslator improves average accuracy over semantic suppression by up to 5.82 percentage points. For two 4B students, it retains 95.0% of rejection sampling’s average accuracy while using 35.9% of its estimated compute. Without retraining, the same translator transfers to a new mathematics dataset, a different teacher, and STEM problems, improving average STEM accuracy by 4.88 and 4.31 percentage points for 4B and 8B students, respectively. These results show that CoTranslator offers an efficient and transferable approach to reasoning distillation.
[NLP-138] Wolof-Adapted Industry-Ready Open-Weight Language Models
【速读】： 该论文旨在解决沃洛夫语（Wolof）在开放权重语言模型中支持薄弱的问题，同时规避使用API模型带来的成本、治理及本地部署主权限制。其解决方案的关键在于通过指令微调（instruction tuning）、偏好对齐（preference alignment）与模型融合（model merging）技术，构建一系列0.6B至14B参数量的开源沃洛夫语适配模型。实验在63项公开及内部任务上验证了这些模型在沃洛夫语理解与生成能力上的优异表现，同时保持了对其他语言及跨语言能力的有效性，其中8B和14B模型性能已显著逼近甚至媲美更大规模的商用系统。目前，这些模型已在内部客服试点中评估，并已公开发布。
链接: https://arxiv.org/abs/2610.20824
作者: Gwénolé Lecorvé,Delphine Charlet,Pierre Adam,Floriane Behanzin,Lionel Delphin-Poulat,Johannes Heinecke,Géraldine Damnati,Pelin Kaytmaz,Nourredine Khaous,Charbel Kindji,Meriam Maadi,Abdoulaye Mbaye,Claire Perroux,Ismaël Rousseau,Papa Séga Wadé,Gilles Xiberras,Olivier Zhang,Malek Zitouni
机构: Orange Research(橙色研究), France
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 6 + 47 pages (main + appendices), 27 tables, 1 figure
Abstract:Wolof is spoken by 23M people, mostly in Senegal, but remains weakly supported by open-weight language models, while API models raise cost, governance, and sovereignty constraints for local deployments. In this paper, we present 0.6B-14B open-weight models adapted to Wolof through instruction tuning, preference alignment, and model merging. After presenting the pipeline and data, extensive experiments on 63~public and in-house tasks show good performance on Wolof understanding and generation while remaining effective on other languages and cross-lingual skills, with the 8B and 14B models being particularly competitive with substantially larger commercial systems. The models are currently being evaluated in internal customer-care pilots and are publicly available at this https URL .
[NLP-139] Gibbs-Gap Policy Optimization for Off-Policy RL with Verifiable Rewards
【速读】： 该论文旨在解决强化学习中因使用过时的行为策略（behavior policy）生成的响应进行训练而导致的策略不匹配问题，尤其在采用可验证奖励时，传统方法通过学习者到行为策略的重要性权重修正该偏差，但会引入长序列响应下的高方差问题。其核心解决方案是提出吉布斯间隙策略优化（Gibbs-Gap Policy Optimization, GGPO），该方法通过最小化同一提示下多个响应之间经KL调整后的奖励的对数均值指数对比（log-mean-exp contrast），避免了依赖学习者到行为策略的重要性权重。GGPO从KL正则化策略优化推导而来，其群体损失在策略上精确对应于KL正则化的最优性差距；在适当覆盖条件下，学习者的单次响应期望奖励至少为行为策略的最佳K个响应奖励减去有限组别损失、正则化项及有限训练数据带来的误差。研究进一步通过精确的吉布斯目标组合与局部参考控制下的有限数据最终策略奖励边界，刻画了参考刷新机制的重要性。实证表明，采用默认刷新策略的GGPO在五基准任务加权平均上得分达46.58，而固定参考版本仅为32.96，凸显参考刷新的关键作用；在数学推理任务中，GGPO在训练-推理不匹配及固定离线数据场景下均优于GRPO和AGRO，且随着策略陈旧程度增加仍保持稳定性能。
链接: https://arxiv.org/abs/2610.20814
作者: Xuanchang Zhang,Tong Zhang
机构: University of Illinois Urbana-Champaign (伊利诺伊大学厄本那-香槟分校)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 51 pages. Code: this https URL
Abstract:Reinforcement learning with verifiable rewards often trains language models on responses generated by a different, possibly stale behavior policy. Correcting this mismatch with learner-to-behavior importance weights can introduce high variance for long responses. We introduce Gibbs-Gap Policy Optimization (GGPO), which minimizes a log-mean-exp contrast of KL-adjusted rewards among responses to the same prompt, without learner-to-behavior importance weights. We derive GGPO from KL-regularized policy optimization: on policy, its population loss is exactly the KL-regularized optimality gap, and under suitable coverage the learner’s single-response expected reward is at least the behavior policy’s best-of- K reward, minus terms for the raw finite-group loss, regularization, and finite training data. We characterize reference refresh through exact Gibbs-target composition and a finite-data final-policy reward bound under local reference control and tokenwise behavior mismatch. This refresh is empirically important: in an ablation, fixed-reference GGPO scores 32.96 on the five-benchmark problem-weighted average, versus 46.58 with the default refresh schedule. Experiments on mathematical reasoning further show that GGPO outperforms GRPO and AGRO under training-inference mismatch and on fixed offline data, while maintaining its performance as controlled policy staleness grows.
[NLP-140] A Self-Improving Agent that Grows its Own DSL: Oracle-Free Operator Induction for Exact Question Answering
【速读】： 该论文旨在解决自改进智能体在缺乏奖励信号、评判模型和标准答案的情况下，如何确保其计算能力的扩展仍能保持结果精确性的问题。核心挑战在于：传统方法依赖高奖励、人工评判或压缩策略来筛选新技能，但这些机制在要求绝对正确性的场景（如财务交易总额计算）中可能引入错误。本文提出的关键解决方案是基于一个冻结的单一模型，构建双层嵌套循环结构，运行于一个小而精确的领域特定语言（DSL）之上。内层循环仅在两个独立生成的表达式对所有数据记录均达成一致时才回答新问题，从而实现无监督的一致性验证；外层循环则通过发现自身能力缺失，自动生成新操作符，并仅在第二个独立实现版本与之匹配且结果满足守恒律的前提下予以接纳，整个过程无需任何真实标签（ground truth）。该机制的准入规则具有强鲁棒性：在独立参考下，成功拒绝了138个有缺陷的操作符变体，而现有方法（如“只要运行成功”、自一致性或压缩性）则大多误接纳。实验表明，该框架在多个无关数据模式（如商业账簿与临床就诊问答）上均实现答案可解集超过十倍乃至近二十倍的增长，且跨模型家族具备可迁移性——不同模型族在相同机制下表现出相当的自我扩展能力，仅诱导产出量受各模型对DSL掌握程度的影响。
链接: https://arxiv.org/abs/2610.20809
作者: Yun-Shiuan Chuang,Yirou Ge,Yixi Li,Yiming Bian,Zijie Pan,Jenna L. Ballard,Uma Kona,Linsey Pang,Prakhar Mehrotra
机构: PayPal AI
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:A self-improving agent has to decide which new skills to keep, and in practice it keeps whatever earns a high reward, convinces a judge model, or compresses its library, even when the skill is wrong. That is fine until the answers have to be exactly right: over a table of transactions, a total is either right or wrong. We ask a harder question: can an agent grow what it can compute with no reward, no judge, and no answer key, and still be exactly right? A single frozen model runs two nested loops over a small, exact domain-specific language (DSL). The inner loop answers new questions only when two independently written expressions agree on every record. The outer loop discovers its own missing capability and authors an operator for it, admitted only when an independent second implementation agrees and the result obeys a conservation law, all without ground truth. The inner loop roughly triples the answerable-question set, and the outer loop self-extends on disjoint starting sets, inventing operators outside the starting closure; a no-induction control answers none of the new questions. The admission rule is empirically sound: it rejects 138 of 138 buggy operator variants under an independent reference, even one written by a different model family, where the admit-if-it-runs, self-consistency, and compression signals the field relies on admit most of them. Unmodified, the same mechanism self-extends on two further domains with unrelated schemas, business-ledger and clinical-encounter question answering, growing their answerable sets more than tenfold and nearly twentyfold. The loop transfers across model families too: run with further proposers, the gate stays sound and independent families self-extend comparably, with only the induction yield tracking each proposer’s DSL fluency.
[NLP-141] PRISM: Transfer and Limits of Activation Interventions in Tool-Calling Language Models
【速读】： 该论文旨在解决大语言模型中因文本拒绝（textual refusal）伴随不安全工具调用（unsafe tool calls）而引发的安全隐患问题，尤其关注模型在面对敏感请求时虽表面拒绝但暗含潜在危险行为的“隐蔽漏洞”。其核心解决方案在于提出PRISM框架，通过跨精度、跨语言与跨规模的激活干预迁移机制，实现对模型对齐能力的有效提升。关键创新点在于利用低资源设置下的压缩表示（如量化后的NF4标签）和跨语言迁移策略，使仅需少量场景即可稳定估计并实施安全干预；同时，基于配对良性激活（paired benign activations）的映射方法，可将经训练的安全方向成功迁移至更大参数量模型（最高达70B），并在未进行目标语言微调的情况下，显著降低阿拉伯语等语言中不安全行为的发生率（如在Llama-3.1-8B上减少21.95个百分点）。该研究揭示了通过压缩模型辅助监督与多语言数据协同，能够高效发现并缓解跨模型家族中的有害工具调用行为。
链接: https://arxiv.org/abs/2610.20808
作者: Giovana Kerche Bonás,Thales Sales Almeida,Marcelo da Silva Reis,Rodrigo Nogueira,Marcos Meideiros Raimundo
机构: RECOD.ai Lab, Institute of Computing, Universidade Estadual de Campinas(坎皮纳斯州立大学计算机学院); Maritaca AICampinas, Brazil(巴西)
类目: Computation and Language (cs.CL)
备注:
Abstract:Recent literature shows that unsafe tool calls can often accompany textual refusals in language models. To investigate this hidden vulnerability, we introduce PRISM, a framework for evaluating the transfer of activation interventions across precision, languages, and scales. Our findings demonstrate that leveraging lower-resource settings (specifically quantized NF4 labels and cross-lingual transfer) offers a robust pathway to expose and improve model alignment. Notably, using behavioral labels from an NF4 copy enables stable intervention estimation even with very few scenarios. Furthermore, these alignment improvements transfer effectively across languages: five-language NF4-supervised directions successfully transfer to held-out Arabic, reducing unsafe violations in Llama-3.1-8B by 21.95 percentage points without target-language tuning. By mapping paired benign activations, we also successfully transfer these safety interventions to much larger models (up to 70B parameters). Ultimately, PRISM highlights how auxiliary supervision from compressed models and multilingual data can uncover and mitigate hazardous tool-calling behaviors across different model families. We release all experimental code and supporting artifacts at this https URL, alongside an interactive explorer of recorded model responses and experimental results at this https URL.
[NLP-142] When Reasoning Traces Conflict with Evidence: How Answer Configuration Shapes Evidence Following
【速读】： 该论文旨在解决生成式 AI 在复杂问答任务中因推理轨迹（reasoning trace）与当前证据之间存在冲突时，导致模型无法正确遵循最新证据的问题。其核心挑战在于：推理轨迹既能提升准确性，也可能在与新证据矛盾时干扰模型对最新信息的采纳。为分离这两种作用，研究通过交叉验证不同版本的证据与由模型生成的推理轨迹，在两种配置下进行测试——即延续模型自身思考（continuation）与独立答案调用（separate answer call）。实验结果表明，独立调用相比延续推理显著提升了对当前证据的遵循程度，增幅达29.9至60.6个百分点；而在两种配置下，移除推理轨迹均导致准确率下降相当幅度，说明轨迹对性能具有双重影响。进一步分析显示，推理轨迹的价值主要集中在最后一行，初始黄金证据在轨迹生成前最具影响力，而干扰项会削弱独立调用下的修正能力；此外，重新推理（renewed reasoning）相较于仅提供验证指令，更能有效引导模型在冲突情境下采纳当前证据。因此，推理轨迹的“有益性”与“可被信赖性”是两个独立属性，不能仅通过删除轨迹来判断模型依赖的是何种信息源。论文强调，评估应同时报告模型的答案配置、轨迹的使用方式及其对证据遵循的倾向性，以全面揭示模型行为机制。
链接: https://arxiv.org/abs/2610.20805
作者: Shaoang Li,Yanhang Shi,Jian Li
机构: Stony Brook University (石溪大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 25 pages, 4 figures
Abstract:A reasoning trace retained at answer time plays two roles: it can raise accuracy, and it can compete with the current evidence when the two disagree. We disentangle these roles by crossing verified evidence versions with the traces generated from each and answering from the same traces and evidence under two configurations: continuation of the model’s own thinking and a separate answer call. Across three model families (gpt-oss, Gemma4, Qwen3) on HotpotQA and MuSiQue, the separate call raises evidence-following over continuation by 29.9 to 60.6 points. On the same pairs, removing the trace lowers accuracy by similar amounts under both configurations. Yet continuation repeats the stale conclusion in most conflicts, and the separate call follows the current evidence far more often. Benefiting from a trace and deferring to it are therefore distinct properties. Further analyses show that the final line of a trace carries most of its value, gold evidence helps most before the trace is written, distractors weaken correction under a separate call, and renewed reasoning, unlike a verification instruction, follows the current evidence in most conflicts. Because trace deletion alone cannot reveal which source an answer follows, evaluations should report both properties together with the answer configuration.
[NLP-143] APRiL: An Open-ended Benchmark for Chinese Social Awareness in Large Language Models
【速读】： 该论文旨在解决大语言模型在中文社会交互场景中缺乏全面社会意识评估能力的问题。现有基准测试通常局限于特定社会能力、任务形式或交互情境，难以覆盖多样化语境下的社会认知需求。为此，研究提出APRiL——一个面向中文社会意识的开放式评估基准，其关键在于引入实例特定的评分标准（instance-specific rubrics），并设计了两个互补任务：社会情景分析（Social Scenario Analysis）用于评估模型对社会线索、心理状态及冲突结构的理解能力，社会适切响应（Socially Aware Response）则考察模型能否将上述理解转化为恰当且自然的回应。实验结果揭示了12个来自六个模型家族的普遍“理解-回应”差距现象，即高质量的情景分析可提升回应质量，而低质量分析反而会降低表现。因此，该研究的核心贡献在于构建了一个系统化、多维度的社会意识评估框架，推动中文环境下具备社会感知能力的语言模型发展。
链接: https://arxiv.org/abs/2610.20785
作者: Tian Lan,Han Zhang,Shanshan Wang,Yahan Yu,Jiang Li,Yugo Murawaki
机构: Kyoto University (京都大学); Shanghai Jiao Tong University (上海交通大学); University of Macau (澳门大学); The University of Tokyo (东京大学); Inner Mongolia University (内蒙古大学)
类目: Computation and Language (cs.CL)
备注: 26 pages, 9 figures
Abstract:Large language models are increasingly used in social interactions, where appropriate responses require recognizing implicit cues, inferring mental states, identifying conflicts, and considering potential consequences. However, existing benchmarks often focus on specific social abilities, constrained task formats, or particular interaction settings, offering limited coverage of social awareness across diverse contexts. To address this gap, we introduce APRiL, an open-ended benchmark for Chinese social awareness with instance-specific rubrics. APRiL contains 2,196 Social Awareness Evaluation Instances across six categories and two complementary tasks: Social Scenario Analysis evaluates models’ understanding of social cues, mental states, and conflict structures, while Socially Aware Response assesses whether they can translate this understanding into appropriate, natural responses. Experiments on 12 models from six families reveal a consistent understanding-to-response gap. Higher-quality scenario analyses improve responses, while weak analyses may reduce response quality. APRiL provides a systematic resource for advancing socially aware language models in Chinese contexts. Our data and code are available at this https URL.
[NLP-144] ARIA: Runtime Inspection and Amendment for Multi-Agent Systems
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）驱动的多智能体系统（Multi-Agent System, MAS）中局部错误传播导致任务失败的问题。在现有方法中，错误通常仅能在任务完成后通过诊断或反复重执行来修正，缺乏在运行过程中实时干预的能力。其解决方案的关键在于提出一种名为Agent Runtime Inspection and Amendment (ARIA) 的框架，核心思想是将执行验证后的修复过程转化为可复用的监督信号，从而将修复搜索从运行时测试转变为数据构建与模型训练。研究首先构建了首个大规模、面向失败多智能体轨迹与执行验证修复对的公开数据集——ARIA数据集，包含3,953个经验证的修复样本，明确标注了故障原因、干预位置及后续执行修正方式。基于此数据集，训练一个轻量级模型以实现对运行中多智能体系统的实时检测与修正，目标是学习具有泛化能力的修复策略，而非简单复制特定修复操作。实验表明，经过训练的ARIA模型在事后修复与实时干预场景下均表现出色，并能有效推广至未见过的多智能体框架和主机模型。
链接: https://arxiv.org/abs/2610.20780
作者: Bolin Shen,Yushun Dong
机构: Florida State University (佛罗里达州立大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:In LLM-based multi-agent systems (MAS), a local mistake can propagate as other agents build on it, turning a correctable error into a failed task. Diagnosing completed runs or repeatedly re-executing them to search for corrections does not provide a learned policy for intervening while execution is still in progress. We present Agent Runtime Inspection and Amendment (ARIA), a framework for learning when and how to repair ongoing multi-agent executions. Our key intuition is to turn execution-verified repairs into reusable supervision, moving repair search into data construction and training so that a learned model can intervene at runtime without repeatedly testing alternatives. First, we collect over 24,000 successful and failed trajectories across four multi-agent frameworks, six benchmarks, and eight host models. From these trajectories, we construct the ARIA dataset, to our knowledge the first large-scale dataset pairing failed multi-agent trajectories with execution-verified repairs. It contains 3,953 verified repair pairs, each specifying what went wrong, where to intervene, and how execution should continue. Then, we develop the ARIA framework for MAS-agnostic runtime inspection and amendment, designed to integrate with any MAS that exposes execution trajectories and supports continuation replacement. Using verified repairs as training examples, we teach a compact model to recognize when intervention is needed and how to correct an ongoing run. The aim is to learn repair strategies that apply to related failures in new configurations, rather than merely reproduce individual corrections. Finally, the trained ARIA model demonstrates strong repair performance in both post-hoc and runtime settings, with generalization to unseen frameworks and host models. Our code is available at this https URL.
[NLP-145] Verdict: Encode Once Decide Many
【速读】： 该论文旨在解决组织工作中大量细粒度、类型化决策任务的自动化问题，即在无需人工干预的前提下，对单一状态（如一封邮件）针对多个具体问题（如是否需要回复、紧急程度、负责人等）进行高置信度判断。传统方法如标准交叉编码器（cross-encoder）需对每个问题单独遍历输入，效率低下且难以扩展。本文提出Verdict-MQ，一种开放式编码器架构，其核心创新在于通过一次读取案例并利用结构化注意力掩码（structured attention mask），实现对所有问题的并行回答。该设计确保在截断后保留的状态令牌一致的前提下，状态编码与问题无关，问题间互不干扰，且选项顺序不影响结果，从而在理论上保障了输出的不变性。实验表明，395M规模的Verdict-MQ在类型化决策基准上显著优于421M的Laya专用模型，在五组新种子下均更接近软黄金标签（KL 0.080 vs 0.117，Brier 0.045 vs 0.061），且优势在温度缩放后仍保持；其更高准确率（0.801 vs 0.767）超出基准饱和点，暗示更贴近标注教师的判断。在RACE、BoolQ和MultiRC等通用问答任务中，尽管在部分数据集上表现逊于微调基线，但其零样本性能超越所有微调编码器，甚至超过封闭通用系统（Jev）。控制实验进一步揭示共享状态带来的成本：为每个问题分配独立的状态感知副本可提升RACE和MultiRC得分0.9–3.5分，但代价是1.4–2.5倍的每案例延迟，而在类型化决策任务中无明显收益，验证了所提架构在效率与一致性上的优越性。
链接: https://arxiv.org/abs/2610.20778
作者: Avdhesh Singh Chouhan,Jack Perry
机构: LEC AI(LEC AI)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 19 pages, 2 figures, 14 tables. Code: this https URL
Abstract:Automating organizational work means answering many small typed questions about one piece of state (does this email need a reply, how urgent is it, who owns it) with probabilities good enough to act on without a human. The open typed-decision model we compare against, like a standard cross-encoder, answers each question with a separate pass over the input. We present Verdict-MQ, an open encoder that reads a case once and answers all of its questions: given identical retained state tokens after truncation (verified on every reported case), a structured attention mask makes the state encoding independent of the questions, keeps questions from influencing each other, and makes option order irrelevant by construction. On the typed-decisions benchmark, a 395M Verdict-MQ matches the soft gold labels significantly better than the released 421M Laya specialist on all five fresh seeds (KL 0.080 vs 0.117, Brier 0.045 vs 0.061), an advantage that survives per-system temperature scaling; its higher accuracy (0.801 vs 0.767) lies past the benchmark’s own saturation point, so we read that margin as closer agreement with the benchmark’s labelling teacher. On RACE, BoolQ and MultiRC, against a matched fine-tuned baseline it wins on MultiRC and loses on RACE and on BoolQ, where its training is unstable; we measure zero-shot references ourselves, including a closed general system (Jev) that beats every fine-tuned encoder. A controlled comparison measures what the shared state costs: giving each question its own question-aware copy of the state gains 0.9-3.5 points on RACE and MultiRC at 1.4-2.5x the per-case latency, and nothing measurable on typed decisions, with the invariance guarantees unchanged.
[NLP-146] Reference-Guided Turn-Taking Style Transfer for Full-Duplex Spoken Language Models ICASSP2027
【速读】： 该论文旨在解决生成式语音对话模型在全双工（full-duplex）交互中缺乏灵活、可调控的发言主动性（initiative）问题。现有方法通常依赖外部控制或无法量化模型发言程度，导致难以精确调节模型的对话参与度。其核心解决方案是识别出一个关键变量——倡议向量（initiative vector），即模型内部晚期残差流（late-layer residual-stream）中的单一方向，该方向直接编码了模型决定是否抢话的内在倾向。该向量可通过模型自身对话数据提取，无需额外训练或标注。通过在生成过程中对倡议向量进行缩放，可显著调控模型的发言比例，从原始的21–30%提升至42–61%，且效果稳定跨对话场景、模型权重与语料库。此外，仅需一段简短的参考音频即可将特定说话者的轮次风格迁移至模型，且该迁移具有高度参考特异性：更换参考源后93–98%情况下匹配度下降，而统一剂量则无显著差异。这表明，轮次控制并非固化于模型权重，而是可读取、可调整并按个体校准的动态方向。
链接: https://arxiv.org/abs/2610.20763
作者: Yuxuan Chen,Yang Xiao,Jack Sun,Chak Sun,Brick Li, Sam
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Sound (cs.SD)
备注: 5 pages, 4 figures. Submitted to ICASSP 2027. Audio examples: this https URL
Abstract:Full-duplex spoken language models decide at every frame whether to take the floor, defining a conversational personality – yet each set of weights encodes exactly one, which human speakers bracket rather than share. Existing controls mostly act from outside; none identifies the quantity that sets how much the model speaks. We identify it: a single late-layer residual-stream direction, the initiative vector, extracted from the model’s own conversations without training or annotation. The same direction reads the decision, anticipating onsets up to 800 ms ahead, and shifts it: scaled during generation, it moves speech ratio from 6-12% to 42-61% (no steering: 21-30%), consistently across conversations, weights, and corpora. A short reference recording suffices to transfer a speaker’s turn-taking style to the model, and the transfer is reference-specific: swapping in another speaker’s reference worsens the match in 93-98% of cases, whereas the best uniform dose is indistinguishable from no steering. Turn-taking style is not frozen into the weights but a direction that can be read, shifted, and calibrated per person. Audio examples are available at this https URL.
[NLP-147] urboMamba: Faster Inference Faster Training and Better Prediction with Lexical Memory
【速读】： 该论文旨在解决大语言模型在推理与训练过程中效率低下及生成精度不足的问题，尤其针对传统Mamba架构在长序列建模中全局上下文利用不充分、局部预测能力弱的缺陷。其核心解决方案在于提出一种混合架构——TurboMamba：在每完成一个token块后，由一个全局Mamba-3模型统一处理该块的上下文信息，随后通过一个小型Transformer结合全局上下文与块内历史token进行精细化的下一个token预测。为增强局部预测性能，引入三项轻量化改进：直接访问前一块的最后一个token以增强时序连贯性、对下一个token得分施加小规模校正、以及基于近期token频率学习得到的加权混合机制。实验在公共数据集FineWeb-Edu上对比了12种参数量相当且训练预算一致的小规模模型，包括Transformer、Mamba-3、Jamba和Zamba（后者以Mamba-3替代原模块）。结果表明，TurboMamba在平均测试预测损失（负对数似然，NLL）上优于所有基线模型，较Mamba-3降低9.4%，较Jamba版本降低7.3%，较Zamba版本降低18.3%；同时，在端到端训练吞吐率方面达到Mamba-3的2.60倍、Jamba版本的2.81倍、Zamba版本的4.82倍；在批量为32、提示长度256 token、生成长度128 token的典型工作负载下，端到端延迟速度提升分别达3.12倍、3.53倍和6.21倍。因此，该方案的关键创新在于通过分层结构融合Mamba的高效长程建模能力与Transformer的精细局部预测优势，并辅以轻量级优化，实现了性能与效率的双重突破。
链接: https://arxiv.org/abs/2610.20682
作者: Hanxin Zhang
机构: University of Pittsburgh(匹兹堡大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:We introduce TurboMamba, a language-model architecture for faster inference, faster training, and better token prediction. A global Mamba-3 model runs once per completed token block. A small Transformer uses this global context and earlier tokens within the block to predict the next token. Three lightweight additions improve local predictions: direct access to the previous block’s last token, a small correction to next-token scores, and a learned mixture with recent-token frequencies. We compare twelve small-scale implementations on public FineWeb-Edu, with comparable parameter counts and shared training-budget caps. The baselines include Transformer, Mamba-3, Jamba, and Zamba. The Jamba and Zamba versions use Mamba-3 in place of their original Mamba modules. TurboMamba achieves the lowest mean test prediction loss (negative log-likelihood, NLL) among all twelve models. Its mean NLL is 9.4% lower than that of Mamba-3, 7.3% lower than that of the Jamba version, and 18.3% lower than that of the Zamba version. End-to-end training throughput is 2.60x that of Mamba-3, 2.81x that of the Jamba version, and 4.82x that of the Zamba version. We measure end-to-end batch latency for a batch of 32 sequences, each with a 256-token prompt and 128 generated tokens. On this workload, TurboMamba achieves a 3.12x speedup over Mamba-3, a 3.53x speedup over the Jamba version, and a 6.21x speedup over the Zamba version.
[NLP-148] CascadeLedger: Counterfactual Auditing of Information and Resource Integrity in LLM Cascades
【速读】： 该论文旨在解决预算约束下的语言模型级联系统中，由于路由决策依赖不可观测信息而导致的可审计性与透明性缺失问题。现有系统在执行过程中难以验证其决策是否基于合法可观测输入，且无法有效检测隐蔽读取（hidden-read）或资源使用偏差等故障，从而影响系统的可信度与可复现性。其核心解决方案是提出CascadeLedger框架，通过引入“反事实执行-审计”机制，确保执行过程满足“调用前读取”语义，并将成功读取绑定至服务与缓存身份，精确记录可用性及组件级预算状态变迁。该方案的关键在于构建“聚合等价的轨迹有效性”概念，即两个执行路径即使最终答案、资源预估和各服务调用分布一致，但若其路由决策使用了未公开的隐藏信息，则判定为无效。进一步地，通过引入反事实前缀非干扰性（Counterfactual Prefix Non-Interference, CPNI），在保持可见前缀不变的前提下，仅改变声明的隐藏家族，以识别最小违反前缀与隐式证据；同时采用投影保持对抗性变异（Projection-Preserving Adversarial Mutation, PPAM）方法，探测那些对非反事实检查器仍具欺骗性的无效轨迹。实验表明，该双层审计机制在冻结缓存、独立生成与服务评估场景下均实现了0.0%的误报率，能够检测99.83%的串行化故障与99.82%的未记录隐藏读取故障，并在确定性重计算后修正34.03%的错误质量-资源结论，具备高精度与强鲁棒性。
链接: https://arxiv.org/abs/2610.20668
作者: Miaobo Hu,Shuhao Hu,Xiaobo Guo,Xin Wang,Bokun Wang,Daren Zha,Jun Xiao
机构: University of Chinese Academy of Sciences (中国科学院大学); Institute of Information Engineering, Chinese Academy of Sciences (中国科学院信息工程研究所)
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 38 pages, 6 figures
Abstract:We introduce CascadeLedger, a counterfactual execution-audit framework for budgeted language-model cascades. The protocol enforces call-before-read semantics, binds successful reads to service and cache identities, records availability and component-wise budget transitions, and freezes the event trace before any target or task scorer is joined. The central object is aggregate-equivalent trace validity: two executions can share a final answer, declared resource projection, and service histogram while differing in whether their routing decisions used only legitimately observable information. We complement single-trace invariants with counterfactual prefix non-interference (CPNI), which couples a visible prefix and policy randomness while changing only a declared hidden family. The counterfactual influence frontier (CIF) localizes the earliest violating prefix and a minimal hidden witness; projection-preserving adversarial mutation (PPAM) searches for invalid traces that remain acceptable to a non-counterfactual checker. Across frozen-cache, independently generated, and serving evaluations, the two-layer audit reports a clean false-positive rate of 0.0000, detects 0.9983 of serialized faults and 0.9982 of unlogged hidden-read faults, and changes 0.3403 of corrupted quality-resource conclusions after deterministic recomputation. The online recorder and offline CPNI audit have separate overhead measurements. Learned routing and semantic verification provide downstream compatibility studies under the same execution contract.
[NLP-149] Does Personalization Break Trust? Open-Set Rejection Calibration Under Few-Shot On-Device Intent Adaptation
【速读】： 该论文旨在解决生成式AI（Generative AI）在设备端通过少样本微调（如LoRA）实现个性化时所面临的“域外输入拒绝”（out-of-scope, OOS）识别能力退化问题。尽管现有方法能有效提升对新用户指令的识别准确率（in-scope accuracy），但其在识别与所有已知意图均不匹配的输入（即域外输入）方面表现显著下降，导致系统误将未知请求当作已知意图处理，带来安全隐患和用户体验恶化。研究发现，采用朴素的少样本LoRA微调会使五个公开意图分类基准上的OOS拒绝AUROC平均下降0.055至0.258，且此现象并非由评分器校准偏差完全解释——尽管标准置信度评分器本身在适应过程中出现校准偏移，仍存在真实存在的性能衰减，其根源可归因于已知意图区域嵌入空间边际（embedding margin）的可测量收缩（pooled Spearman rho=0.61）。为此，论文提出一种部署成本极低的解决方案：改用马氏距离（Mahalanobis distance）进行拒绝决策，无需额外训练或重训即可在三个数据集上恢复大部分丢失的OOS识别性能，优于两种现有的基于分布外检测（OOD）的评分基线，并展现出对样本数量、LoRA秩及不同主干编码器的良好泛化能力。作者同时坦诚该方案的局限性，并开源全部代码以确保可复现性。
链接: https://arxiv.org/abs/2610.20652
作者: Vaibhav Prajapati,Pushp Garg
机构: University of Technology Nuremberg (德国纽伦堡应用技术大学); Indian Institute of Information Sri City (印度信息学院斯里城校区)
类目: Computation and Language (cs.CL)
备注:
Abstract:Always-on virtual assistants increasingly personalize on-device to new user-specific commands via few-shot fine-tuning (e.g., LoRA), but this adaptation is evaluated almost exclusively on whether it learns the new commands, not on whether the assistant can still recognize an input that matches none of its intents - the out-of-scope (OOS) rejection problem. We show that naive few-shot LoRA personalization degrades OOS-rejection AUROC (area under the ROC curve) on every one of five public intent-classification benchmarks (mean drop 0.055 – 0.258 , five seeds each), while in-scope accuracy is preserved. A scorer-invariant check shows part of this drop is an artifact of the standard confidence scorer’s own miscalibration under adaptation, though a smaller, genuine effect remains under any scorer; we trace it to a measurable shrinkage of the embedding margin around known-intent regions (pooled Spearman rho=0.61 ). A deployment-cheap fix is scoring rejection by Mahalanobis distance instead of softmax confidence without extra training/retraining recovers most of the lost quality on 3 of 5 datasets at no accuracy cost, beats two further OOD scoring baselines, and generalizes across shot count, LoRA rank, and a second base encoder. We report honestly where the fix falls short and release all code for reproducibility.
[NLP-150] CoTS: Competitive Token Selection for Latent Reasoning in Large Language Models
【速读】： 该论文旨在解决现有隐式推理（latent reasoning）方法在解码过程中对连续软表示与离散承诺之间动态平衡能力不足的问题。具体而言，虽然当前方法通过使用连续软嵌入或隐藏状态来增强中间表示并保留多种可能的推理路径，但其对隐式干预的处理方式仍为全局软激活或基于不确定性阈值的硬切换，未能实现按解码步骤自适应调节干预强度的连续控制。为此，论文提出一种无需训练的框架——竞争性标记选择（Competitive Token Selection, CoTS），其核心创新在于将原本二元化的隐式干预模式重构为基于令牌级别的连续分配机制。CoTS利用概率上最相近的两个候选生成项之间的竞争关系，动态权衡每一步的离散确定性与软探索程度：当主导项明显优于次优项时，模型倾向于保持离散采样轨迹以维持推理稳定性；而当二者竞争加剧时，则增加软表示的贡献以保留潜在备选路径。该机制无需人工设定切换阈值或混合系数，且不引入额外前向传播开销。在六个推理基准上的实验表明，CoTS在性能上持续优于标准Chain-of-Thought（CoT）及当前最先进的无训练隐式推理方法。
链接: https://arxiv.org/abs/2610.20642
作者: Jie Yu,Yuxi Sun,Zhaoxu Li,Chunxiao Liu,Xudong Jiang
机构: Xiaomi AI Lab(小米人工智能实验室)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Recent latent reasoning methods move beyond discrete Chain-of-Thought (CoT) by replacing sampled tokens with continuous soft embeddings or hidden states, enabling richer intermediate representations and preserving alternative reasoning paths. However, the need for latent reasoning varies across decoding steps: confident predictions favor discrete commitment, whereas ambiguous steps benefit from preserving multiple plausible alternatives. Existing approaches typically handle this variation through either global soft activation or hard switching based on uncertainty thresholds, treating latent intervention as an all-or-nothing decision rather than continuously adapting its strength. To address this limitation, we propose Competitive Token Selection (CoTS), a training-free framework that reframes latent intervention from binary mode selection into continuous allocation at the token level. CoTS uses the probability competition between the two most likely candidates to dynamically balance discrete commitment and soft exploration at each reasoning step. The second candidate represents the strongest alternative to the model’s preferred continuation, so the separation between the two directly reflects the stability of the current discrete choice. When one candidate clearly dominates, CoTS keeps the representation close to the sampled discrete token to preserve a stable reasoning trajectory; as the competition strengthens, it increases the contribution of the soft representation to retain plausible alternatives. This competition-based allocation requires neither manually tuned switching thresholds nor mixing coefficients, and introduces no additional forward passes. Experiments across six reasoning benchmarks demonstrate that CoTS consistently outperforms standard CoT and state-of-the-art training-free latent reasoning methods. Our code is available at this https URL.
[NLP-151] AbridgeKV: Compressing Long Reasoning into the KV Cache of Learned Pause Tokens
【速读】： 该论文旨在解决长链式推理（long chain-of-thought decoding）过程中因持续生成文本而导致的键值缓存（KV cache）内存占用线性增长的问题。传统方法通过淘汰缓存条目来缓解内存压力，但无法有效提升模型在未经过此类淘汰训练时的生成长度与性能。为此，本文提出AbridgeKV方案，其核心在于训练模型自身对输出内容进行压缩：将已生成文本块压缩为少数可学习的“暂停标记”（pause tokens），这些标记作为额外输入但不产生输出，其键值（KV）嵌入编码原块信息；随后释放原始缓存空间，后续生成过程直接关注这些压缩后的暂停标记的KV条目。关键创新点包括：暂停标记的KV位置嵌入于其所替代的文本块内部、按各层缓存条目实际价值动态分配缓存预算、以及基于较大文本块执行压缩操作。在包含446道竞赛题目的多数据集测试中，仅保留四分之一原始缓存容量的AbridgeKV模型在准确率和生成长度上达到全量缓存模型水平，显著优于同类先进淘汰策略（evictor）和重新训练的压缩器（compressor）。部署于vLLM框架下，序列级缓存规模缩小3.5倍，饱和状态下的H100显卡处理时间加快3.2倍。
链接: https://arxiv.org/abs/2610.20641
作者: Jack Cai,Lijie Yang,Vasilis Kontonis,Dimitris Papailiopoulos,Shivam Garg
机构: Princeton University (普林斯顿大学); Microsoft Research (微软研究院)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Long chain-of-thought decoding is memory-bound: the KV cache of the generated tokens grows linearly with the test-time compute being spent. Evicting cache entries does not fix this, because a model never trained on a cache with entries removed generates longer, often up to the length limit. We instead train the model to compress its own output. AbridgeKV compresses each completed block of generated text into the KV entries of a few learned pause tokens. These are extra input tokens with trained embeddings that produce no output. Once the pause tokens have encoded the block, AbridgeKV frees the block’s original cache; subsequent tokens attend to the pause tokens’ KV entries in its place. Three novel choices make this work: the pause tokens’ KV entries take positions inside the block they replace, the cache budget is allocated across layers according to the measured value of each layer’s entries, and compression operates on relatively large blocks. On 446 competition problems from AIME, HMMT, AMC, BRUMO, CMIMC and SMT with 32K-token generations, keeping a quarter of the generated cache matches an identically trained full-KV model in accuracy and output length; the strongest published evictor at a matched budget writes far longer, and the closest trained compressor, retrained on the same data, loses 10 points. Served in vLLM, the cache is 3.5x smaller per sequence and a saturated H100 finishes a problem 3.2x faster than with full KV.
[NLP-152] Same Answer New Box: How Much of RLVRs Out-of-Domain Gain Is Answer Format?
【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）在跨领域评估中报告的性能提升是否真实反映了模型能力增强，还是仅源于答案格式优化的问题。具体而言，现有评估方法依赖于外部答案提取器（answer extractor）从模型输出中自动抽取答案，而这些提取器往往优先检索特定格式的位置（如数学题中的\boxed{}标记），导致模型通过将答案置于指定位置即可获得更高评分，而非真正提升推理或回答质量。论文的关键发现在于：在多个7B规模的基线-强化学习模型对中，高达82%的跨领域多项选择题性能增益实际上源自答案格式的改变，而非答案内容本身的改进——即模型只是将原本已存在于文本中的答案移至\boxed{}中以迎合提取器偏好。作者设计了一种新的、更鲁棒的提取器，能够识别任意位置的答案（无论是否在\boxed{}内），并重新评估已有输出，结果表明大部分“增益”消失。此外，使用独立的大语言模型（LLM）作为阅读器进行验证，也得到相似结论（77%的增益为格式所致）。而在一个32B模型对中，由于基线模型本就习惯使用框式格式，而强化学习模型反而减少使用，因此几乎不存在格式性增益，说明该现象具有模型和训练策略依赖性。综上，论文揭示了当前评估范式中存在的严重偏差，并指出：报告的性能增益部分是提取器特性的产物，其本质可通过重用已有输出并更换提取器来量化，无需重新执行强化学习过程。
链接: https://arxiv.org/abs/2610.20639
作者: Zongye Lyu
机构: Monash University (莫纳什大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 36 pages (10 main), 5 figures, 28 tables
Abstract:Reinforcement learning with verifiable rewards (RLVR) is credited with gains beyond the domain its rewards check, such as higher MMLU accuracy after training on mathematics. Such accuracies are usually read out of the model’s text by an answer extractor. A model that learns to write its answer where the extractor looks can therefore score higher without answering better. We measure how much of the gain reported by extractors copied from public code arises this way. We generate outputs from a base model and its RL model and score them twice: with those extractors, and with one extractor of our own that reads a stated answer whether it is in the answer box \boxed or in the text. On four released 7B SimpleRL base-RL pairs, 82% of the out-of-domain multiple-choice gain that these extractors report disappears under ours: it is a gain in answer format, not in answers. The base models already state most of their answers in text, and RL moves them into the box, where these extractors look first. Independent LLM readers, in place of our extractor, put the figure at 77%. With a prompt that itself asks for the box, our extractor puts it at 79%. On a 32B pair trained with another recipe, almost none of the reported gain is format gain: its base model boxes its answers when asked, and its RL model rarely does when not. A reported gain is partly a property of the extractor that read it, and re-scoring stored outputs with a second extractor shows how much, without re-running RL.
[NLP-153] Exploring the Design Space of Permutation Diffusion LLM s
【速读】： 该论文旨在解决基于语言模型生成排列（permutation generation）时面临的双重挑战：一是文本条件下的候选项排序（text-conditioned ranking），二是候选项与输出位置之间的全局一一映射（global one-to-one assignment）。其核心解决方案的关键在于通过三种不同的扩散-语言模型（diffusion-LLM）参数化方式来建模排列约束：1）在打乱的候选索引序列上进行硬分配预测；2）在双随机矩阵（doubly stochastic matrices）上进行软分配预测；3）直接预测排序得分并随后进行排序。这三种方法将排列约束施加于不同阶段的预测与解码过程，从而支持各异的训练目标与采样策略。实验结果表明，在亚马逊三个领域的推荐重排序任务中，候选集规模为10和20时，硬分配与软分配变体在单次函数评估下达到与受限掩码扩散基线相当的Recall@1性能，而直接排序得分预测无需匈牙利算法或Sinkhorn归一化即可保证有效排列。然而，增加函数评估次数并未一致提升排序质量，且效果高度依赖状态表示与训练目标设计。研究结论强调：在基于语言模型去噪器的排列扩散框架中，噪声扰动（corruption）、预测参数化、优化目标与解码器必须协同设计才能实现最优性能。
链接: https://arxiv.org/abs/2610.20635
作者: Zhe Xu,Mingfu Liang,Kavosh Asadi,Jean-Pascal Billaud,Tianyi Chen,Yanhui Guo,Christina Huang,Baekjin Kim,Jimmy Kim,Yufei Li,Daqi Sun,Jiachen Sun,Dian Yu,Zaiwei Zhang,Kaushik Rangadurai,Frank Shyu,Sandeep Pandey,Xi Liu,Luke Simon
机构: Meta
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Permutation generation with language models requires both text-conditioned ranking and a global one-to-one assignment between candidates and output positions. We study this problem through three diffusion-LLM parameterizations: hard-assignment prediction over shuffled candidate-index sequences, soft-assignment prediction over doubly stochastic matrices, and direct ordering-score prediction followed by sorting. The three families place permutation constraints at different stages of prediction and decoding and therefore support different training objectives and sampling procedures. We evaluate these designs on recommendation reranking across three Amazon domains with candidate sets of size 10 and 20. At one function evaluation, the hard- and soft-assignment variants achieve Recall@1 comparable to a constrained masked-diffusion baseline, whereas direct ordering-score prediction guarantees valid permutations without Hungarian assignment or Sinkhorn normalization. Additional function evaluations do not consistently improve ranking quality, and their effect depends strongly on the state representation and training objective. These results show that corruption, prediction parameterization, objective, and decoder should be designed jointly for permutation diffusion with language-model denoisers.
[NLP-154] Jev-as-a-Judge for RAG Claim Verification: Confidence-Gated Cascading Between a Non-Generative Verifier and a Frontier LLM
【速读】： 该论文旨在解决检索增强生成（Retrieval-Augmented Generation, RAG）系统中冗余的文本生成开销问题，即在执行二元判断任务（如“该证据是否支持该主张”）时，主流流程仍依赖生成式模型输出完整文本以获取单一比特的决策结果，导致高昂的计算成本。其核心解决方案是引入一种非生成式判别模型，该模型直接返回预定义枚举类型的分类结果（如“支持”或“不支持”），并附带原生概率值，而无需生成任何自然语言文本。关键创新在于：通过对比一个低成本、非生成式的判别模型与前沿大语言模型（LLM）在相同标注数据集（LLM-AggreFact）上的表现，发现二者在标准评估指标上统计无显著差异，但前者成本仅为后者的1/187。进一步分析揭示，尽管整体准确率相近，两模型在错误模式上存在异质性，且廉价模型对未被支持主张的误判率更高（23.2% vs 14.1%），表明传统指标掩盖了实际部署中的风险差异。最终提出一种基于置信度的分级处理策略——将高置信度判断由廉价模型直接处理，低置信度样本交由昂贵模型处理，实现74.10的得分仅需2.92美元/千条，显著优于全量使用昂贵模型的8.93美元/千条，但实验显示成本节约主要源于“升迁比例”而非概率阈值本身，说明效率提升来自策略设计而非模型内在能力。研究强调，在追求准确率的同时必须关注错误分布的不对称性，真正的优化应基于对不同错误模式的理解与组合。所有代码、提示模板及定价数据均已开源。
链接: https://arxiv.org/abs/2610.20634
作者: Alden Do Rosario,Kirollos Atef
机构: CustomGPT.ai(自定义GPT.ai)
类目: Computation and Language (cs.CL)
备注: 31 pages, 6 figures. Code and data: this https URL
Abstract:Retrieval-augmented generation asks one question at volume: does this evidence support this claim? The answer is one bit, yet most pipelines obtain it from a text generator and pay generation prices for the prose wrapped around it. We compare a non-generative decision model, which returns a typed value from a declared enumeration together with a native probability and no text, against a frontier LLM configured as an engineer would deploy it for binary classification. Both answer the same 495 human-labelled claims from LLM-AggreFact, so every comparison is within-item. We report three things. First, the two systems are statistically indistinguishable on the metric this task is conventionally scored by, a paired difference of -0.56 points, Jev minus Astra, at a 187-fold cost difference. Second, that tie carries almost no information about deployment: the systems disagree on 15.2% of claims, and the cheaper one admits 23.2% of the claims annotators marked unsupported against the frontier model’s 14.1%. A component whose job is to catch unsupported claims should be judged on that rate, and the conventional headline hides it. Third, because the errors differ they compose: routing the claims the cheap model is confident about and escalating the rest reaches 74.10 at USD 2.92 per thousand claims against 73.83 at USD 8.93, with the threshold chosen out of fold. But a matched random-escalation control at in-sample thresholds, spending the same budget by coin flip, does as well, so what buys the saving is the escalation share and not the probability that picks what to escalate. Measured against the cheap model it is built from, the cascade costs 61-fold more for 0.83 points, 95% CI [-3.36, 4.86], and it is the more permissive guard of the two. Equal accuracy conceals unequal errors. All predictions, prompts, pricing and code are released.
[NLP-155] Context Injection and Long-Horizon Reliability: Separating Duplicate Recognition Local Processing and Counter Tracking
【速读】： 该论文旨在解决大语言模型在长时序任务中出现的渐进性失效问题，特别是针对其在长时间跨度下对历史信息的准确记忆与状态维护能力不足的缺陷。传统评估方法通常仅关注最终任务成功率或可维持的上下文长度，而本文提出一种细粒度、分步评分的400步记账任务（ledger task），将任务分解为三个可独立评估的组件：对早期记录的长距离识别（duplicate recognition）、当前记录的局部处理（local processing）以及每步递增的计数器（counter）维护。其解决方案的关键在于通过模块化评估揭示各组件在长周期中的行为差异——结果显示，尽管局部处理和计数器在无干扰条件下表现稳定（计数器连续8000步无错误），但重复识别错误在第200步后显著上升（从0.0019增至0.011/步），且主要由长距离回忆失败和误报共同导致；更重要的是，仅在每个步骤前插入一段无关内容块即导致计数器在全部20个种子实验中均发生失败，表明模型状态维护极易受外部干扰影响。进一步预注册分析表明，计数器值依赖于响应历史中的先前值，一旦被污染则无法自我修正，而外部精确的状态提醒（oracle-supplied compact reminder）可在一到两步内恢复计数器正确性，且优于等长中性文本，提示模型缺乏内在纠错机制。该研究强调了聚合误差分类可能掩盖关键组件间的异质性失效模式，凸显了对长时序推理中状态一致性与抗干扰能力进行精细化建模的重要性。
链接: https://arxiv.org/abs/2610.20633
作者: Thomas Serval
机构: Baracoda AI Labs; NEOMA Business School (NEOMA商学院)
类目: Computation and Language (cs.CL)
备注: 17 pages, 5 figures. Code, data and reproduction script: this https URL (tag v1.0-arxiv)
Abstract:Long-horizon failures of language models are usually measured as end-task success or as a horizon length. We instead score, step by step, three components of a 400-step ledger task in which every step is easy in isolation and the model’s own answers are re-injected into its context: recognising a record already seen 5 to 300 steps earlier, processing the current record locally, and writing a counter incremented at every step. On Qwen3-8B (non-thinking, 20 seeds, replicated on a second machine and backend), with every step scored component by component, two results hold. First, duplicate-recognition errors rise late in the episode – from 0.0019 to 0.011 per step at risk across step 200 – while the counter makes no error in 8,000 steps and local processing is descriptively stable; the rise combines long-distance recall errors and rising false alarms. Second, a block of irrelevant content inserted before each step makes the counter fail in 20 of 20 seeds on each machine, against none without it, while its effect on local processing is much smaller and not distinguishable from none. Three pre-registered follow-ups probe the counter. Its value depends on the value present in the response history: a corrupted value persists without correction in every seed, and an exact external reminder, repeated before each message, re-aligns it within one to two steps. Under the irrelevant block, no counter error was observed when an exact compact reminder of the state, supplied externally by an oracle, preceded each message (0 of 20 seeds with an error, against 15 without it), and this reminder outperformed a length-matched neutral text; whether a single reminder suffices was not conclusive, and the absence of damage to the other components was not established. An aggregate or exclusive error classification can hide the difference we report.
[NLP-156] MeddeX: A Retrieval-Augmented LLaMA-3 System with Gemini Vision for Evidence-Backed Medical Report Analysis
【速读】： 该论文旨在解决现代医疗报告日益复杂化所带来的临床专业人员在时间消耗与解读负担上的双重挑战，尤其是传统人工审阅流程存在效率低、易出错且难以扩展的问题。其解决方案的关键在于构建一个集成多模态能力的医学报告分析系统——MeddeX，该系统通过微调的QLoRA-Llama-3（8B）模型结合检索增强生成（Retrieval-Augmented Generation, RAG）机制，并融合多模态Gemini Vision模块，实现对文本型PDF、扫描图像型PDF及直接放射影像的智能混合提取。系统核心创新包括：针对损坏或UTF-16编码文档的新型垃圾文本检测机制、基于报告范围的精准上下文检索、自动化纵向报告对比引擎以及按需触发的自愈式补填流水线。此外，系统采用FastAPI后端与Supabase向量数据库构建端到端架构，显著提升了生成摘要的准确性、简洁性与证据支持度。实验结果表明，相较于基线模型LLaMA3-Med42-8B，MeddeX在ROUGE-1、ROUGE-2、ROUGE-L、BLEU和BERTScore等指标上均有显著提升，验证了领域特定微调与系统级设计优化的有效性，为人工智能辅助临床决策支持提供了可落地、鲁棒性强的技术路径。
链接: https://arxiv.org/abs/2610.20632
作者: Rajdeep Pandhare,Rohit Kundaikar,Yakub Kojar,Shreyansh Tiwari
机构: Rajiv Gandhi Institute of Technology, Mumbai, India
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 6 pages , 3 figures, published in ijsrem
Abstract:Modern healthcare generates increasingly complex medical reports that challenge clinical professionals in terms of time and interpretive burden. Manual review processes are slow, error-prone, and difficult to scale. This paper presents MeddeX, a Medical Report Analyzer that combines a QLoRA fine-tuned LLaMA-3 (8B) model with Retrieval-Augmented Gen eration (RAG) and a multimodal Gemini Vision layer to deliver accurate, concise, and evidence-backed summaries of medical documents. The system handles text-native PDFs, scanned image based PDFs, and direct radiological images through an intelligent hybrid extraction pipeline that includes a novel garbage-text detection mechanism for corrupted and UTF-16-encoded doc uments. Beyond single-report QA, MeddeX introduces report scoped retrieval for precise contextual grounding, an automated longitudinal report comparison engine, and a self-healing backfill pipeline that lazily processes legacy documents on demand. A FastAPI backend, Supabase vector store, and this http URL fron tend complete the end-to-end architecture. Evaluation against the base LLaMA3-Med42-8B model demonstrates consistent improvements across ROUGE-1 (+9.2%), ROUGE-2 (+10.0%), ROUGE-L (+11.8%), BLEU (+71.4%), and BERTScore (+0.8%), validating the benefit of domain-specific fine-tuning. MeddeX represents a practical and robust step toward AI-assisted clinical decision support.
[NLP-157] Real-time Chatbot Romance Scam Detection in Stage Classification
【速读】： 该论文旨在解决社交媒体平台上日益严重的“浪漫诈骗”（Romance Scam）问题，尤其针对生成式人工智能（Generative AI）技术使身份伪造更加便捷所带来的挑战。其核心解决方案是基于社会学中的七阶段理论，构建一个通过Line Messaging API实现实时交互的聊天机器人（Chatbot），利用大语言模型（Large Language Model, LLM）对用户对话进行动态分析，识别诈骗所处的具体阶段。该系统不仅能够及时检测潜在的诈骗行为，还能基于LLM生成针对性的阶段解释与防范建议，从而提升用户的风险意识。其关键创新在于将社会科学理论与大语言模型相结合，实现对诈骗过程的结构化识别与即时干预，具备实际应用价值，可直接部署于真实场景中以降低浪漫诈骗的发生风险。
链接: https://arxiv.org/abs/2610.20631
作者: Yi-cheng Liu,Guan-ru Lai,Yi-cheng Lai,Lyn Chao-ling Chen
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Multimedia (cs.MM)
备注: Accepted in Taiwan Academic Network Conference (TANET 2026)
Abstract:Romance scam has become a serious problem via social media, and the improvement of artificial intelligence enable to fake identify easily. In the study, a chatbot has developed to detect romance scam in real-time through Line Messaging API. Using the seven stage theories in the social science field, the stages of romance scam can be detected through large language model. In addition, once a romance scam is detected, the chatbot provides explanations of the stage and prevention advices through large language model. The contribution of the work provides a practical application, and can be used in real scenario for reducing the risk of romance scam.
[NLP-158] A Large-Scale Dataset for Multi-Class Financial News Classification in Brazilian Portuguese
【速读】： 该论文旨在解决巴西葡萄牙语金融自然语言处理（Financial NLP）系统发展受限于领域特定数据集匮乏的问题。其核心解决方案是构建一个包含约16万篇巴西葡萄牙语金融新闻文章的大规模语料库，并通过人工标注与大语言模型（LLM）弱监督标注相结合的方式，建立初步的多分类任务基准。研究的关键在于评估基于语义相似性的三种方法——余弦相似度、K近邻（K-Nearest Neighbors, KNN）和基于FAISS的最近邻搜索——在跨语言语义嵌入下的表现。实验结果表明，该任务对标签质量与类别分布高度敏感：在人工标注子集上，KNN取得最佳性能（准确率47.3%，F1-score 31.9%），FAISS表现紧随其后；而在LLM生成的弱标注子集上，尽管准确率接近50%，但F1-score骤降至约20%，凸显了标签不平衡与噪声标注的负面影响。研究揭示了高质量人工标注在金融文本分类中的关键作用，为巴西葡萄牙语金融NLP领域的后续研究提供了可复现的基准数据集与方法参考。
链接: https://arxiv.org/abs/2610.20630
作者: Leomax Bandeira,Gabriel Aguiar,Dalton Cézane Gomes Valadares,Luiz Pereira,Helder Fernando de Araújo Oliveira,Gustavo Xavier,Andson M. Balieiro
机构: Federal University of Paraíba, BRA(联邦大学帕拉伊巴分校); Federal University of Campina Grande, BRA(联邦大学坎皮纳格兰德分校); Federal Rural University of Pernambuco, BRA(联邦农村大学伯南布哥分校); Federal University of Pernambuco, BRA(联邦大学伯南布哥分校)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Paper accepted for presentation at SMC 2026
Abstract:The development of Financial Natural Language Processing (NLP) systems for Brazilian Portuguese is hindered by the scarcity of domain-specific datasets. To address this, we present a large-scale corpus of approximately 160,000 Brazilian Portuguese financial news articles collected from web sources. To establish initial benchmarks for a weakly supervised multi-class classification approach based on semantic similarity, we annotated subsets of this corpus both manually and using a Large Language Model (LLM). We evaluated three similarity-based methods – cosine similarity, K-Nearest Neighbors (KNN), and FAISS-based nearest-neighbor search – using multilingual semantic embeddings. Our experiments reveal that the task is highly sensitive to label quality and class distribution. On the manually labeled subset, KNN achieved the best performance (47.3% accuracy, 31.9% F1-score), closely followed by FAISS. Conversely, on the LLM-annotated subset, while accuracy reached approximately 50%, F1-scores dropped to near 20%, highlighting the negative impact of label imbalance and noisy weak annotations. These findings establish foundational baselines for a challenging task, underscoring the critical need for high-quality manual annotations. Ultimately, this dataset and our preliminary results provide a reference point for future research in Brazilian Portuguese Financial NLP.
[NLP-159] PRISM: PRIvileged Sibling-guided Multi-Agent Topology Design
【速读】： 该论文旨在解决基于大语言模型（Large Language Models, LLMs）的多智能体系统（Multi-Agent Systems, MAS）在复杂任务求解中因协作拓扑结构设计不当而导致的性能瓶颈问题。现有方法虽能从任务查询中学习生成协作拓扑，但其训练依赖于已验证执行结果的图谱语料库，而该语料库仅记录成功解题的拓扑结构，却未提供针对特定任务的偏好排序信息，导致后续监督微调（Supervised Fine-Tuning, SFT）缺乏对每项任务最优拓扑的指导信号。由此引发的问题是：生成器趋向于学习语料库中频繁出现的拓扑模式，而非根据具体任务需求动态适配最优结构，从而在某些任务上取得性能提升的同时，可能损害其他任务的表现。为解决此问题，本文提出PRISM框架，其核心创新在于通过“特权信息学习”（Learning with Privileged Information, LPI）机制，从同一无排序语料库中恢复每项任务的细粒度优化信号。具体而言，教师模型通过一个零初始化的低秩分支读取一个同任务下另一成功拓扑（即“兄弟图”），该信息在部署时不可用；学生模型则通过自采样轨迹，匹配教师在各决策步骤上的目标输出，从而实现对任务特异性拓扑偏好的隐式学习。实验在六个多样化基准测试上表明，PRISM显著优于所有对比的拓扑设计基线，并且与教师匹配的对照组验证了特权信息对性能提升的关键贡献。
链接: https://arxiv.org/abs/2610.20628
作者: Phuong Tuan Dat,Qi Li,Xinchao Wang
机构: 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) based on large language models (LLMs) have become a powerful way to address complex problems, and their effectiveness depends critically on the collaboration topology that decides which agents participate and how they communicate. Recent work learns to generate this topology from the task query, fitting the generator to corpora of execution-verified graphs. The post-training method for such generators remains a limitation: a corpus lists the graphs that solved a query but never ranks them, leaving continued supervised fine-tuning (SFT) with no per-query preference to exploit. Without such a preference, the generator drifts toward the topologies the corpus contains most often rather than those that suit each query, so a gain on one task can come at a cost on another. We propose PRISM, which recovers per-query signal from the same corpus by learning with privileged information: a teacher reads, through a zero-initialized low-rank branch, a sibling graph that also solved the query and is unavailable at deployment, and the student matches its per-decision targets on trajectories the student samples itself. Across six diverse benchmarks PRISM improves over every topology-design baseline evaluated, and matched-teacher controls support the contribution of privileged query information. Code is available in this link.
[NLP-160] DeepUnfolding_Agent : A Classical_Quantum Adaptive Control API
【速读】： 该论文旨在解决机器学习、联邦学习、量子机器学习及经典-量子混合系统中自适应优化工作流因控制器逻辑嵌入任务特定训练脚本而导致的可复现性差、可重用性低、安全验证困难以及跨执行环境公平比较受限的问题。其核心解决方案是提出DeepUnfolding Agent，一个基于模式驱动（schema-driven）的FastAPI平台，将深度展开的自适应控制逻辑封装为外部软件服务。该框架通过策略注册表（strategy registry）、诊断到状态适配器（diagnostic to state adapter）、控制器工厂（controller factory）、安全防护机制（safety guard）、执行器抽象（executor abstraction）、反馈更新路径（feedback update path）、SQLite持久化层、Docker部署支持及可复现基准测试脚本等组件，实现控制器智能与领域特定执行逻辑的解耦。该架构支持经典、量子感知、混合、Qiskit Aer模拟以及可选的IBM量子硬件执行路径，通过单元测试、模式验证、API测试、安全性测试、编排测试、量子执行测试、检查点测试和Docker部署验证，证明了其在金融、医疗、基因组学、交通、噪声量子模拟及真实量子硬件上的外部用户固定策略与自适应策略对比的有效性。实验结果表明，DeepUnfolding Agent为异构优化工作流提供了一个可重用、可追溯的自适应控制层，其量子实验进一步体现了API级量子控制与硬件可移植性的潜力。
链接: https://arxiv.org/abs/2610.20615
作者: Shanika Iroshi Nanayakkara,Shiva Raj Pokhrel
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Adaptive optimization workflows in machine learning, federated learning, quantum machine learning, and hybrid classical quantum systems often embed controller logic directly inside task specific training scripts. This coupling limits reuse, reproducibility, safety validation, and fair comparison across execution environments. This paper presents DeepUnfolding Agent, a schema-driven FastAPI platform that exposes deep unfolded adaptive control as an external software service. External executors register a strategy schema, submit round-level diagnostic variables, receive schema-bounded control actions, execute one optimization round, and return feedback for controller update. The framework separates controller intelligence from domain specific execution through a strategy registry, diagnostic to state adapter, controller factory, safety guard, executor abstraction, feedback update path, SQLite persistence layer, Docker deployment, and reproducible benchmark scripts. The implementation supports classical, quantum-aware, hybrid, Qiskit Aer, and optional IBM quantum processing execution pathways. We validate the software through unit, schema, API, safety, orchestration, Quantum execution, checkpointing, and Docker tests, and demonstrate external user fixed versus adaptive workflows across finance, healthcare, genomics, transport, noisy quantum simulation, and hardware backed IBM Quantum execution. The results show that DeepUnfolding-Agent provides a reusable and traceable adaptive control layer for heterogeneous optimization workflows. Our quantum experiments are interpreted as API level quantum control and hardware portability evidence.
[NLP-161] ruth for Believable AI: Expressed Doubt Provenance and Belief Revision as an Engineerable Stance
【速读】： 该论文旨在解决对话智能体在回应时普遍表现出过度自信、缺乏对信息来源的透明性以及无法显式表达信念修正的问题，即如何在不修改底层语言模型的前提下，通过引入一个行为层来实现对认知状态（epistemic states）的可控表达。其解决方案的关键在于构建一个包含三重机制的行为层：（1）基于每条主张的置信度与类型化信息溯源（typed provenance），（2）基于溯源的表达触发规则（provenance-gated expression rule），以及（3）具备可审计记录与部分抗错误修正能力的持久化信念修订存储（persistent revision store）。实验表明，该架构在合成模型上通过了所有五项测试，且在真实模型中实现了承诺的可审计性保证（100%成立），并能部分区分真实与虚假修正；然而，预设的表达保真度、矛盾分离和溯源边界等指标未能达成。进一步后验分析揭示，基于平均答案词元概率的表达门控表现低于随机水平（AUC 0.41），而基于采样一致性的门控则具有判别能力（AUC 0.66）。最终，采用一致性门控配置并通过独立协议验证的方案满足了对话级操控性和能力等价性标准，并在重抽样对话集上实现复现，但其有效性依赖于选择策略，且当不确定性集中在60个事实时，两项核心标准仍未解决。因此，论文支持的结论限于可构造的审计保障、依赖存储的修正判别能力，以及特定基准与模型下词元概率门控的失效，尚需扩大事实规模以开展人类评估。
链接: https://arxiv.org/abs/2609.26035
作者: Sebastian Cochinescu
机构: University of Bucharest(布加勒斯特大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 17 pages, 4 figures, 3 tables. Companion framework paper: arXiv:2607.15883 . Code, benchmark, cached model outputs, and result files archived at doi: https://doi.org/10.5281/zenodo.21462986 (code and results) and doi: https://doi.org/10.5281/zenodo.21462988 (benchmark dataset)
Abstract:Conversational agents often express answers in a uniformly confident register. We test whether expressed uncertainty, provenance-aware assertion, and explicit belief revision can be implemented as a behavior layer over a fixed language model; we do not test believability or trust. The layer combines three epistemic states, per-claim confidence and typed provenance, a provenance-gated expression rule, and a persistent revision store with auditable acknowledgments and partial resistance to false corrections. We evaluate it on a constructed, mechanically scored multi-session benchmark using a synthetic model and Qwen2.5-0.5B-Instruct. The synthetic instrument passes all five checks. On the real model, acknowledgment soundness, a by-construction guarantee, holds in 100% of cases, and true corrections are accepted more often than false ones (0.44 vs. 0.15 on held beliefs; 0.875 vs. 0.420 including rule-accepted corrections of unheld facts), but the pre-specified expression-fidelity, contradiction-separation, and provenance margins fail. A disclosed post hoc analysis shows that expression gated on mean answer-token probability ranks correctness below chance end to end (AUC 0.41, conversation-clustered), whereas gating on sampling consistency discriminates (AUC 0.66). A consistency-gated configuration selected from this finding and evaluated under a separately committed protocol meets the conversation-level manipulation and capability-equivalence criteria and replicates on a redrawn conversation set. The manipulation result is selection-dependent, and both criteria remain unresolved when uncertainty is clustered over the 60 facts. The supported conclusions are limited to the by-construction audit guarantee, store-dependent partial correction discrimination, and a benchmark- and model-specific failure of token-probability gating; scaling the fact base is required before human evaluation.
[NLP-162] Bridging the Gap: Converting Read Text to Conversational Dialogue
【速读】： 该论文旨在解决将朗读式语音（read speech）转换为自然对话式语音（conversational speech）时，如何在保持语音自然度与可懂度的同时，降低计算开销以满足实时应用需求的问题。其核心挑战在于朗读式语音普遍缺乏自然对话中所需的韵律多样性，如语调、重音和节奏的细微变化，从而影响虚拟助手、客户服务及语言学习等场景的交互体验。本文提出的解决方案——基于对话上下文的韵律调整（Prosodic Adjustment with Conversational Context, PACC），关键在于利用高保真生成对抗网络（HiFi-GAN）结合深度神经网络对语音的韵律特征进行精准分析与重构，通过引入对话上下文信息动态调节语调、重音与节奏模式，实现更自然流畅的语音转换。实验结果表明，该方法在提升语音自然度和模型准确性方面显著优于传统方法，并在平均意见分（MOS）评估中建立了新的基准，验证了其在多种语音转换任务中的可扩展性与有效性。
链接: https://arxiv.org/abs/2605.18001
作者: Parshav Singla,Agnik Banerjee,Aaditya Arora,Shruti Aggarwal,Anil Kumar Verma,Vikram C M,Raj Prakash Gohil,Gopal Kumar Agarwal
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注: 11 pages, 4 figures. Published in ICICC 2025, Springer Lecture Notes in Networks and Systems
Abstract:In recent advancements within speech processing, converting read speech to conversational speech has gained significant attention. The primary challenge in this domain is maintaining naturalness and intelligibility while minimizing computational overhead for real-time applications. Traditional read speech often lacks the nuanced prosodic variation essential for natural conversational interactions, posing challenges for applications in virtual assistants, customer service, and language learning tools. This paper introduces a novel approach, Prosodic Adjustment with Conversational Context (PACC), aimed at converting read speech into natural conversational speech used in various modern applications. PACC utilizes advanced deep neural networks to analyze and modify prosodic features such as intonation, stress, and rhythm. Unlike conventional methods, our approach uses High-Fidelity Generative Adversarial Networks (HiFi-GAN) for speech synthesis. Our experimental results demonstrate significant improvements in speech conversion, enhancing naturalness and achieving better model accuracy with additional training on speech datasets. This research establishes new benchmarks in speech conversion tasks and Mean Opinion Score (MOS) evaluation for testing model accuracy, and we show that our approach can be successfully extended to other speech conversion applications.
[NLP-163] MagiC: Evaluating Multimodal Cognition Toward Grounded Visual Reasoning AACL2026
【速读】： 该论文旨在解决当前大型视觉-语言模型在视觉问答与多模态推理任务中是否存在真正意义上的“具身化视觉推理”（grounded visual reasoning）的问题，而非仅依赖表面模式或数据集偏差进行推断。其核心挑战在于如何有效评估模型是否能够基于真实视觉证据进行逻辑连贯、可解释且与图像内容高度对齐的推理过程。解决方案的关键在于提出MagiC这一综合性基准测试体系，通过约5,500个弱监督生成的问答样本和900个由人工精标的数据集，涵盖答案、推理链及边界框标注（bounding box groundings），全面评估模型在最终答案正确性、推理有效性、具身一致性（grounding fidelity）以及自我修正能力四个维度的表现。此外，研究引入诊断性设置以检验模型在对抗性视觉线索下的鲁棒性，并提出MagiScore与StepSense等新指标，系统揭示现有方法在具身化多模态认知中的关键局限与优化方向。
链接: https://arxiv.org/abs/2507.07297
作者: Chengfei Wu,Ronald Seoh,Bingxuan Li,Liqiang Zhang,Fengrong Han,Dan Goldwasser
机构: Purdue University (普渡大学); University of Illinois Urbana-Champaign (伊利诺伊大学厄本那-香槟分校); University of Connecticut (康涅狄格大学); University of California Berkeley (加州大学伯克利分校)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: AACL 2026 Main Conference
Abstract:Recent advances in large vision-language models have led to impressive performance in visual question answering and multimodal reasoning. However, it remains unclear whether these models genuinely perform grounded visual reasoning or rely on superficial patterns and dataset biases. In this work, we introduce MagiC, a comprehensive benchmark designed to evaluate grounded multimodal cognition, assessing not only answer accuracy but also the quality of step-by-step reasoning and its alignment with relevant visual evidence. Our benchmark includes approximately 5,500 weakly supervised QA examples generated from strong model outputs and 900 human-curated examples with fine-grained annotations, including answers, rationales, and bounding box groundings. We evaluate 15 vision-language models ranging from 7B to 70B parameters across four dimensions: final answer correctness, reasoning validity, grounding fidelity, and self-correction ability. MagiC further includes diagnostic settings to probe model robustness under adversarial visual cues and assess their capacity for introspective error correction. We introduce new metrics such as MagiScore and StepSense, and provide comprehensive analyses that reveal key limitations and opportunities in current approaches to grounded visual reasoning.
[NLP-164] KULTURE Bench: A Benchmark for Assessing Language Model in Korean Cultural Context
【速读】： 该论文旨在解决当前多语言评估基准在跨文化适用性上的局限性问题，特别是现有基准大多基于英文翻译版本，可能引入西方文化偏见，难以准确衡量语言模型对非西方文化（如韩国文化）的深层理解能力。针对这一研究空白，本文提出KULTURE Bench——一个专为韩国文化设计的评估框架，包含文化新闻、成语和诗歌等多类型数据集，能够从词、句、段落等多个层级评估模型的文化理解与推理能力。其解决方案的关键在于构建一个本土化、文化敏感的评估体系，以真实反映语言模型在处理韩国文化语境下的文本时的表现，从而揭示现有模型在理解韩国文化深层内涵方面的显著不足。
链接: https://arxiv.org/abs/2412.07251
作者: Xiaonan Wang,Jinyoung Yeo,Joon-Ho Lim,Hansaem Kim
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted by the 38th Pacific Asia Conference on Language, Information and Computation
Abstract:Large language models have exhibited significant enhancements in performance across various tasks. However, the complexity of their evaluation increases as these models generate more fluent and coherent content. Current multilingual benchmarks often use translated English versions, which may incorporate Western cultural biases that do not accurately assess other languages and cultures. To address this research gap, we introduce KULTURE Bench, an evaluation framework specifically designed for Korean culture that features datasets of cultural news, idioms, and poetry. It is designed to assess language models’ cultural comprehension and reasoning capabilities at the word, sentence, and paragraph levels. Using the KULTURE Bench, we assessed the capabilities of models trained with different language corpora and analyzed the results comprehensively. The results show that there is still significant room for improvement in the models’ understanding of texts related to the deeper aspects of Korean culture.
[NLP-165] Documenting Trofinetide Treatment Phenotypes in Rett Syndrome with a Dual-LLM Consensus Pipeline
【速读】： 该论文旨在解决雷特综合征（Rett syndrome）患者在真实世界中使用托芬奈特（trofinetide）治疗后，缺乏可扩展的疗效与不良反应监测方法的问题。由于传统病历回顾方式耗时费力，难以实现大规模数据提取。为此，研究提出了一种双大语言模型（large language model, LLM）共识管道，通过两阶段模型协同判断实现对18项治疗应答表型（包括4种不良反应、13种雷特综合征临床特征及药物拒绝行为）的自动化提取。其核心解决方案在于采用基于上下文学习（in-context learning）的提示工程（prompt engineering）与共识架构设计：首先在腹泻发生情况的数据集上优化模型架构与提示模板，随后将固定后的系统应用于其余17个表型，仅替换表型定义即可完成外样本测试。结果显示，要求“腹泻处于活跃且严重状态”的提示性能最优（准确率84.2±0.0%，F1=0.674），显著优于传统正则表达式匹配方法（提升47个百分点，p<0.0001）。专家评审对每种表型随机抽取6个标注结果进行验证，总体一致性达90.2%（92/102），证明了该方法的可靠性。值得注意的是，13项雷特综合征临床特征为疾病固有表现，并非药物所致不良反应；其中粗大运动障碍、精细运动障碍、语言/交流能力受损及刻板行为在所有随访时间点均在92%-100%的患者中被记录。研究结论表明，结合上下文推理与共识机制的大语言模型框架，可作为关键词与正则表达式提取的可扩展替代方案，仅通过配置调整即可适应新表型和新药物，为真实世界证据生成提供高效、灵活的技术路径。
链接: https://arxiv.org/abs/2610.22062
作者: Daniel Palacios,Guantong Qi,Mikael Guzman Karlsson,Shawntel K. David,Mohammed Sameeruddin,Bernhard Suter,Zhandong Liu,Davut Pehlivan,Hu Chen
机构: 未知
类目: Quantitative Methods (q-bio.QM); Computation and Language (cs.CL)
备注: 20 pages, 3 figures, 4 tables; supplementary material (22 pages, 1 figure, 11 tables) included
Abstract:Objective: Trofinetide, approved in 2023 as the first Rett syndrome treatment, lacks scalable real-world monitoring because chart review is labor-intensive. We developed a dual large language model (LLM) consensus pipeline to extract 18 treatment-response phenotypes, comprising 4 adverse effects, 13 Rett clinical features, and medication refusal, from clinical notes. Materials and Methods: We analyzed 26,763 notes from 108 patients with confirmed sustained trofinetide exposure at Texas Children’s Hospital. The pipeline filters notes by keyword, classifies each through two LLMs requiring consensus, and aggregates a patient-by-timepoint matrix. Architecture and prompt were selected on a diarrhea development set, then frozen and applied to the other 17 phenotypes with the phenotype definition substituted, as an out-of-sample test. The reported prompt comparison scores 183 expert-labeled cells. Results: A prompt requiring diarrhea to be active and severe performed best (84.2 \pm 0.0% accuracy; F1 = 0.674), 47 percentage points above regular-expression matching ( p 0.0001 ). Expert reviewers judged six of the pipeline’s calls per phenotype, three positive and three negative, and agreed with 90.2% (92/102). The 13 Rett clinical features are manifestations of the disorder itself rather than adverse effects of trofinetide. Gross and fine motor impairment, language/communication, and stereotypy were documented present in 92-100% of patients at every visit. Discussion: In-context LLM extraction captured multi-domain treatment responses. The outcome is documented presence at each visit, not a direction-of-change judgment. Conclusion: Coupling in-context LLM reasoning with a consensus architecture offers a scalable alternative to keyword and regular-expression extraction for real-world evidence, adapting to new phenotypes and medications through configuration alone.
[NLP-166] When Does “Post-Call” Begin? An Availability Contract for Earnings-Call NLP Benchmarks
【速读】： 该论文旨在解决企业财务信息披露在时间维度上的动态性所导致的模型预测中“事件内时序错配”（within-event timing mismatch）问题，即模型用于预测的文本或语音输入尚未完整时，目标变量（如股价收益）已开始计算，从而引发偏差。其核心解决方案是提出“可用性契约”（Availability Contract），明确界定输入边界、输入可得性证据、结果起始时间及基于该证据所能支持的主张。通过在24家公司的552场业绩说明会数据上应用该框架，研究发现：以电话会开始时刻为代理指标与以重构音频结束时刻加处理预留时间作为完成代理指标相比，有9.4%的通话导致收益符号反转；固定完成时间点后，28.3%的通话在收益分位数上发生改变。此外，在独立的公开波动率基准分析中，将事前目标窗口提前一个交易日会导致三日均方误差上升18.1%-21.4%，而模型排名保持不变。这表明尽管模型排序看似稳定，但标签和损失函数仍显著依赖于所声明的结果时钟设定。该契约机制使这一隐含假设显性化，并确保预测结论严格基于实际可得的时间证据。
链接: https://arxiv.org/abs/2610.21659
作者: Xiaoyu Ma,Kelvin J.L. Koa,Yuncong Liu,Ke-Wei Huang
机构: National University of Singapore, Asian Institute of Digital Finance (新加坡国立大学数字金融研究院)
类目: Computational Finance (q-fin.CP); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: ICAIF 2026
Abstract:Corporate financial disclosures often unfold over time, with announcements, prepared remarks, and discussion spanning multiple high-frequency return intervals. Financial NLP benchmarks can therefore suffer from a within-event timing mismatch when the prediction target begins before the textual or audio input used by a model could be complete. Chronological train–test splits do not prevent this mismatch, because it occurs inside individual events rather than across samples. We introduce an Availability Contract, which requires a benchmark to state the input boundary, evidence for input availability, the outcome start, and the claims supported by that evidence. We apply this framework to 552 earnings calls from 24 firms. For each call, we hold the return endpoint fixed and compare the previous close, a documentary call-start proxy, and a completion proxy based on reconstructed audio end plus a stipulated processing allowance. At this baseline, moving from the call-start proxy to the completion proxy flips return signs for 9.4% of calls. Holding completion-clock cutpoints fixed, 28.3% change quintile. In an independent release-session analysis of a public earnings-call volatility benchmark, shifting before-market target windows forward by one trading session raises three-day mean squared error by 18.1-21.4% across four frozen prediction streams, while their point rankings remain unchanged. Together, the results show that labels and losses can depend materially on the declared outcome clock even when model rankings appear stable. The Availability Contract makes this assumption explicit and ties post-input claims to the available timing evidence.
信息检索
[IR-0] CARA: Causally Guided Retrieval Agent for Question Answering
链接: https://arxiv.org/abs/2610.22017
作者: Weicheng Liu,Xiangcheng Wu,Depeng Xu,Haopeng Zhang,Xi Niu
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 16 pages, 2 figures, 5 tables
Abstract:Retrieval-augmented generation over text-rich knowledge graphs requires identifying evidence that contributes to answering a question, rather than merely matching its textual content. Existing agentic retrievers adaptively balance global search and multi-hop exploration but primarily rely on textual similarity to guide retrieval. We propose CARA, a training-free, CAusally guided Retrieval Agent that explicitly evaluates the causal relevance of retrieved evidence through two complementary criteria: consequentiality, measuring whether adding evidence improves answerability, and essentiality, assessing whether removing evidence weakens answerability. CARA integrates these assessments with path consistency to guide graph exploration and employs reciprocal rank fusion to aggregate evidence from multiple agents. We evaluate CARA on three STaRK datasets and augment STaRK with ground-truth answers to enable downstream question answering evaluation. Experimental results demonstrate that CARA consistently outperforms existing retrievers. CARA also improves answer quality and faithfulness while examining fewer graph nodes and requiring fewer retrieval operations. These findings demonstrate that causal relevance assessment enables more focused and effective agentic graph retrieval, improving both retrieval performance and downstream question answering.
[IR-1] GACC-Bench: Benchmarking Goal-Adaptive Context Compression Ability across Large Language Models
链接: https://arxiv.org/abs/2610.21893
作者: Weiting Liu,Zhiyu Pan,Shuhong Wang,Zhenquan Yao,Jungang Li,Li Zhao,Jian Jiang,Jianfeng Feng,Ai Han,Wenlian Lu
类目: Information Retrieval (cs.IR)
备注:
Abstract:Context compression in long-running applications of large language models (LLMs) must preserve goal-relevant information, often before specific downstream questions are known. We introduce GACC-Bench, a benchmark for Goal-Adaptive Context Compression that evaluates how reliably LLMs adapt information retention to specified high-level goals. For each source context, we derive one or more high-level goals by grouping source-grounded questions according to shared downstream objectives. The benchmark contains 500 goals and 2,571 questions over 281 document and agent-trace contexts from seven collections spanning diverse domains, with all instances reviewed by human experts. The compressor receives a source context, a goal, and a context-length budget, with the associated evaluation questions withheld. A fixed reader answers these questions individually using the same compressed context. Our evaluation of ten LLM compressors reveals substantial gaps to full-context performance even for the best-performing model, particularly in supporting correct answers to every question associated with a goal. Compared with goal-agnostic compression, target-goal conditioning consistently improves accuracy, whereas mismatched-goal conditioning reduces it. GACC-Bench is available at this https URL.
[IR-2] Aligning LLM s with Sparse Retrieval via Adaptive Sparse Supervision
链接: https://arxiv.org/abs/2610.21877
作者: Tongna Chen,Minghan Li,Guodong Zhou
类目: Information Retrieval (cs.IR)
备注:
Abstract:While large language models (LLMs) offer strong semantic representations, their application to first-stage retrieval incurs substantial latency and storage costs. Neural sparse retrievers such as SPLADE provide efficient retrieval but rely on relatively compact encoders with limited representational capacity. To bridge this gap, we propose an LLM-based sparse retrieval framework with input-adaptive sparse alignment. Specifically, we introduce a dynamic \alpha -truncation strategy that selects input-specific supervision supports based on the cumulative activation mass of SPLADE representations. This strategy guides the LLM to preserve salient lexical signals while suppressing less informative activations, without relying on a fixed support budget. Subsequent retrieval fine-tuning combines contrastive learning, knowledge distillation, and sparsity regularization to improve retrieval effectiveness. Experiments on MS MARCO and six selected BEIR datasets demonstrate improvements over SPLADE++ while maintaining competitive effectiveness against LLM-based sparse retrievers. Compared with Lion-sp, our method reduces index storage by 6.4-fold and end-to-end query latency by over 11-fold, achieving a favorable effectiveness-efficiency trade-off.
[IR-3] Retrieved but Not Read: Understanding Why Search-Augmented LLM s Fail to Use Available Evidence ALT
链接: https://arxiv.org/abs/2610.21731
作者: Owen Walters,Lihu Chen,Francesca Toni
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 14 pages, 4 figures, 16 tables. Code and data: this https URL
Abstract:Web search allows LLMs to access up-to-date information, but it remains unclear why they sometimes produce outdated answers even when relevant evidence is available. To study this problem, we construct temporal facts whose values changed after the models’ training cutoffs. We evaluate one proprietary model under deployed web search and six open-weight model configurations under controlled retrieval. We find that models often answer from search snippets without opening the source pages, which is associated with higher rates of outdated answers and unfaithful citations. Our controlled experiments and layer-wise analysis reveal a gap between the availability of evidence and its effective use. Even when retrieved evidence influences internal predictions, models may still favour outdated information in their final answers. We hope these findings will help advance research on the reliable use of retrieved evidence in search-augmented LLMs.
[IR-4] VoteGEO-Bench: Measuring the Impact of Generative Engine Optimization on Electoral Visibility
链接: https://arxiv.org/abs/2610.21330
作者: Shiqi Zhang,Xiaokui Xiao
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL); Computers and Society (cs.CY)
备注:
Abstract:Large language models (LLMs) are increasingly used for election information, yet little is known about how external web content shapes the political information they present. Generative engine optimization (GEO) rewrites web content to improve its visibility in generated responses. To investigate its potential risks in voting scenarios (VoteGEO), we introduce VoteGEO-Bench, a benchmark covering the 2024 U.S. presidential election and the 2026 Japanese House of Representatives election, combining election-related queries, historical webpages, and voter personas. We first analyze sources cited by generative engines and find that many citations come from lower-cost owned and open-platform sources, while some webpages remain visible across days. We then evaluate 11 rewriting strategies in 4 categories, examining target specification, effectiveness across engines and elections, and the role of voter personas. Rewriting without specifying a target rarely increases target exposure, whereas target-specified rewriting more often does. Strategies that improve page organization or strengthen a position through argumentation yield greater gains than adding supporting evidence or changing wording. For the two strongest strategies, gains remain broadly positive across evaluated U.S. personas, but persona conditioning shows no consistent effect on these gains. Explicit partisan identities can create baseline exposure gaps larger than the gain from rewriting one webpage. When partisanship is implicit in the voter profile, baseline shifts depend on how strongly the engine responds to that signal.
[IR-5] Making every stage of medical device document assessment with large language models deterministic
链接: https://arxiv.org/abs/2610.21300
作者: Ryosuke Suzuki
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 30 pages, 3 figures, 3 tables; supplementary tables S1-S15 and two supplementary notes included
Abstract:To show that a medical device complies with regulation, experts judge whether documents and records meet each requirement and record the judgment for others to verify. For each assessment item that experts define, large language models could prepare a preliminary assessment: a label such as Satisfied or Not satisfied, a reason and quoted evidence. We sent fourteen commercial deployments of such models (seven models from three providers, through six routes) the same request repeatedly, with the deterministic sampling settings each accepted; every deployment returned differing assessments for at least one request, and in twelve the label changed. Reducing reasoning where the interface allows it did not reduce the variation, and prompts of retrieved passages removed it only for the smallest models. The same open-weight model was reproducible or not depending only on how it was served. A pipeline we built to be deterministic at every stage – text extraction and optical character recognition (OCR), chunking, embedding, indexing, retrieval and inference – returned byte-identical output at each stage across reruns, interruptions, edits, machines of one graphics processing unit (GPU) model and days, for blood glucose meter and infusion pump documents. OCR, embedding and approximate retrieval varied when their own serving was not made invariant, and outputs changed with the GPU model and with the inference engine’s scheduling settings, which therefore belong in the recorded configuration. End-to-end determinism is achievable and gives a testable basis for audit and change control.
[IR-6] RACER: Trajectory-Aware Tool Retrieval for LLM Agents
链接: https://arxiv.org/abs/2610.21267
作者: Tanish Kher,Sarvesh DK,Abedelkadir Asi,Sparsha Chennarayapatna Sundaresh,Dezhi Hong
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 21 pages, 7 figures, 23 tables
Abstract:Modern agents select tools from registries, but exposing entire registries incurs high token costs and can impair selection as they grow. We present TRACER, a retrieval front-end that indexes observed tool-call sequences in a trie and ranks continuations using a learned blend of empirical and contextual priors. On a mixed retail-and-telecom \tau^2 environment, TRACER improves GPT-5.4’s tool-selection accuracy by 4.7 points while reducing input tokens per step by 80%, exposing just 5 of 128 tools. In autonomous full-turn execution, it matches GPT-5.4’s full-registry task success with 61% fewer input tokens per step. Across six environments and four models, TRACER improves tool-selection accuracy over full-registry selection in 21 of 24 model-environment pairs, with average gains of 6 to 10 points; the three exceptions trail by at most 2.2. We analyze failures in both regimes to identify directions for future work.
[IR-7] RecSoup: Efficient Model Merging Approach for Controllable Multi-Objective Recommendation CIKM2026
链接: https://arxiv.org/abs/2610.21220
作者: Zidong Jia,Guofu Xie,Xiao Zhang
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注: Accepted at CIKM 2026
Abstract:Recommender systems often need to balance multiple conflicting objectives. Controllable Multi-Objective Recommendation (CMOR) enables dynamic adjustment of objective trade-offs at inference time without retraining, adapting to shifting user or system requirements. Existing CMOR approaches typically rely on output-level reranking or parameter-generating modules, which either limit the scope of feasible trade-offs or incur substantial training and inference cost. Model merging provides a promising alternative by reusing multiple objective-specific models via parameter combination. However, its direct application to recommender systems is challenging due to the lack of a shared linear regime and cross-objective interference. To address these issues, we propose RecSoup, a lightweight two-stage approach for effective CMOR via specialized model merging. RecSoup first mitigates parameter interference across objectives through a pruning strategy that jointly considers objective-specific importance and cross-objective relevance. It then performs adaptive parameter merging via a learnable preference-to-coefficient mapping, allowing dynamic trade-off control at inference time without joint retraining or additional generative modules. Theoretically, we prove that RecSoup outperforms the parameter generation paradigm in effectiveness and efficiency, as it can fully navigate the Pareto front without smoothness constraints. Extensive experiments on multiple benchmarks demonstrate that RecSoup achieves improved Pareto front quality and effective controllability while delivers 165–228x faster inference-time latency than parameter generation approaches. Code is available at this https URL.
[IR-8] ReviewPilot: A Multi-Agent System for Screening and Data Extraction in Systematic Reviews
链接: https://arxiv.org/abs/2610.21042
作者: Zhaoqian Xue,Shaoze Zhou,Xuexin Li,Junting Mu,Haozheng Liu,Zhongyin Wang,Yichun He,Renkai Ma,Hao Zhou,Sarah Bratt,Siyuan Ma,Chen Chen,Lingyao Li
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注:
Abstract:Systematic literature reviews follow the PRISMA process of screening scholarly works and extracting evidence from included studies. Deep research tools based on large language models (LLMs) can search and summarize the literature but do not record which studies are excluded or where extracted values come from. We develop a conversational multi-agent system called ReviewPilot that automates the PRISMA workflow from paper screening to information extraction. Researchers can tailor criteria and schemas through dialogue, and a project memory keeps their configurations and revisions. Each screening decision comes with its reason and each extracted value with its source. On epidemiological expert reviews, ReviewPilot outperforms a single-prompt LLM and AgentSLR across three LLMs, with screening macro F1 up to 0.776 and extraction F1 up to 0.788. Case studies in biomedicine and human-computer interaction show close agreement with human annotators, suggesting that ReviewPilot can transfer beyond epidemiology. Code is available at this https URL.
[IR-9] Learning When to Think: Utility-Guided Adaptive Reasoning for Efficient Text Retrieval EMNLP2026
链接: https://arxiv.org/abs/2610.21030
作者: Yingtao Luo,Kangyi Zhao,Shiyi Du,Yubo Li
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL)
备注: Accepted to EMNLP 2026 Findings
Abstract:Reasoning-conditioned embeddings can improve retrieval by progressively refining semantic constraints. However, additional reasoning may introduce over-generalized or semantically noisy abstractions that weaken retrieval specificity. Existing methods largely supervise reasoning through final retrieval outcomes, providing limited guidance over the refinement process itself. We propose AURA, a framework that directly supervises intermediate reasoning states according to how well they separate relevant documents from hard distractors. AURA further trains a lightweight utility estimator over prefix-conditioned embeddings to adapt reasoning depth at inference time. On the reasoning-intensive BRIGHT benchmark, AURA-4B achieves 34.7 average nDCG@10 while using 2.8 reasoning steps and 96 generated tokens per query. Under matched deployment conditions, adaptive stopping reduces latency from 4.6 to 0.76 seconds relative to full-trajectory reasoning. Results on MS MARCO further suggest that this specialization need not substantially degrade conventional in-domain retrieval.
[IR-10] ransX Technical Report: Scaling Sequence Transduction for Recommendation
链接: https://arxiv.org/abs/2610.20913
作者: Da Xu,Liyan Fang,Divya Venugopalan,Sunny Hsu,Xukai Wang,Rishav Roy Chowdhury,Cindy Liang,Nishant Satya Lakshmikanth
类目: Information Retrieval (cs.IR)
备注:
Abstract:We present TransX, a sequence-transduction framework and co-designed infrastructure for scaling industrial recommendation with explicit control over model-compute allocation and execution amortization. TransX separates sequence encoding, behavior memory, and candidate-conditioned fusion into independently scalable components. Training amortizes causal trajectory computation across behavioral prefixes and fans out memory states to candidates; serving disaggregates bulk prefill, nearline refresh, and online prediction through compact behavior-memory KV. We observe anisotropic scaling returns: encoder scaling may saturate beyond moderate depth at fixed memory resolution and fusion capacity; increasing supervision shifts the allocation frontier toward candidate-conditioned fusion; and the memory resolution needed to preserve quality remains roughly stable as model capacity grows. We train TransX on click and conversion, post-train its shared behavioral interface for retrieval and preference ranking, and explore generative recommendation offline. In separate production A/B experiments, the core TransX model improves CTR by 6.02% and conversion by 4.45%, while post-trained retrieval and preference-ranking variants improve CTR by 2.96% and conversion by 1.33%, respectively, over corresponding production baselines. Under matched training conditions, the training infrastructure increases supervised-target throughput by 84%. In serving, disaggregation reduces TransX’s total serving cost by nearly 80% versus colocated serving at matched traffic and SLO, enabling approximately 5X the per-request model FLOPs of the HSTU-style production baseline at the same total serving cost. At matched per-request model FLOPs, TransX delivers approximately 4X higher request throughput per H200. These results show that online serving cost need not grow proportionally with recommendation model capacity. Subjects: Information Retrieval (cs.IR) Cite as: arXiv:2610.20913 [cs.IR] (or arXiv:2610.20913v1 [cs.IR] for this version) https://doi.org/10.48550/arXiv.2610.20913 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[IR-11] Idea2Prior: Auditable Paper Retrieval and Ranking from Research Ideas
链接: https://arxiv.org/abs/2610.20857
作者: Shitian Yang,Ruijie Tao,Jian Huang
类目: Information Retrieval (cs.IR)
备注:
Abstract:Finding a paper from an informal research idea requires more than matching topic words: relevant work may use different terminology, and its distinguishing mechanism may be absent from the abstract. We present \system, an auditable literature-search system built on PaperArena. It combines multi-source discovery, identity verification, facet-based abstract assessment, semantic citation expansion, and joint Expert ranking. A retrieval-only mode ranks up to five papers from a shortlist of at most 60, using complete available abstracts and locally selected PDF evidence from five fixed acquisition attempts, without generating a long-form report. In a historical comparison covering ten queries from each of EASY, MEDIUM, and HARD, aggregate reported Hit@5 is 53.3% for \system, 43.3% for ChatGPT web Deep Research, and 36.7% for Gemini web Deep Research under the documented accounting convention. The advantage is concentrated in EASY and MEDIUM; ChatGPT leads on HARD. Selected \system runs average 8 minutes 53 seconds, compared with user-reported web averages of 5.2 and 4.7 minutes. A separate nine-query API sample averages \ 0.0945 per task; this is a historical cost observation, not a matched comparison with web subscription costs. Pairwise Top 5 coverage ranges from 6% to 34%, showing differing returned evidence sets. The system also exposes optional implementation analysis and citation-evolution tools. These results support targeted idea-to-paper retrieval while leaving general web research, forum search, and broader relevance evaluation open. this https URL
[IR-12] Information-Gain Calibrated On-Policy Distillation for Deep Search Agents
链接: https://arxiv.org/abs/2610.20664
作者: Chenxi Li,Xueyang Feng,Songyang Liu,Chenxu Wang,Xiaoying Tang,Chaozhuo Li,Litian Zhang
类目: Information Retrieval (cs.IR)
备注:
Abstract:On-policy distillation (OPD) has recently emerged as a promising approach for transferring the capabilities of large agents to compact students. In deep search, where agents iteratively retrieve external evidence, a teacher’s preference among plausible queries before retrieval may not predict which queries will return useful evidence. Standard OPD nevertheless supervises each student query from the preceding context alone, inducing a failure mode we term \textbf\textitevidence-blind supervision (EBS). To understand this, we regroup the OPD objective by interaction turns and reveal two implicit priors: a \textitbudget prior that allocates supervision by token volume, and a \textitdirection prior that favors tokens preferred by the teacher, neither of which accounts for the evidence returned by the current query. Our empirical analysis shows that both priors contribute to EBS in practice: the first search turn accounts for 51.1% of total absolute information gain but receives only 18.0% of the supervision budget, while 43.7% of search queries have teacher rewards that disagree in sign with post-retrieval information utility. To mitigate this, we propose \ours, which uses post-retrieval information gain (IG) to calibrate both priors: reallocating supervision toward information-critical turns while calibrating query preferences using the utility of retrieved evidence. Experiments on six multimodal search benchmarks show that \ours improves average accuracy by 3.4 and 1.7 percentage points for 2B and 4B students over standard OPD, enabling the 4B student to surpass its 8B teacher on MMSearch and BrowseComp-VL.
人机交互
[HC-0] Asking Smarter Questions: Targeted Active Preference Learning for Underspecified Rewards
链接: https://arxiv.org/abs/2610.22285
作者: Helena Merker,Nick Walker,Andreea Bobu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)
备注: Accepted to the 10th Conference on Robot Learning (CoRL 2026), Austin, TX. 18 pages, 8 figures, 1 table
Abstract:Reward learning typically assumes the human’s input covers everything that matters for the task. In practice, humans can only attend to so much at once, so their teaching naturally captures some features—or aspects of the task—better than others. Prior work has shown how the robot can identify which features were poorly supervised directly from its teaching data. But the current solution is to ask the human to try again, putting the burden on someone who already struggled to convey those features. We propose shifting this burden to the robot. Our key insight is that the same information that tells the robot what it doesn’t know can help it ask better questions. Preference queries offer a low-effort channel for resolving this uncertainty, but only if the robot knows where to direct them and the human knows what to attend to when answering. We act on this insight in two ways: we build an informed prior that targets preference queries at the features that need resolving, and we accompany each query with an explanation that focuses the human’s attention on exactly those features. In simulation, the informed prior beats the standard active preference learning baseline by up to 30.8% in the low-query regime, where reducing human effort matters most. In a user study, the informed prior and explanations each significantly improve reward recovery and amplify each other.
[HC-1] Inflak: A Unified Orchestration Architecture for Human-Agent Interaction
链接: https://arxiv.org/abs/2610.21972
作者: Lin Gao,Li Chen,Jianxun Lian,Ji Li,Haotian Li
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:LLM-powered agents increasingly support open-ended activities, yet orchestrating when and how agents interact with humans remains challenging. Existing approaches often provide interaction mechanisms tied to particular tasks or execution patterns, lacking a unified architecture for diverse and evolving interaction needs. We introduce Inflak, a unified four-layer architecture for human-agent interaction orchestration that separates state reconstruction and routing, interaction and execution planning, interaction design, and interface rendering. Its layered design with progressive resolution enables general applicability, runtime flexibility, module composability, and orchestration maintainability. We implement Inflak as agent-native, skill-based plugins within a general-purpose agent framework and evaluate it through a gallery of various application scenarios and interaction paradigms, technical performance testing, and a comparative implementation case study. The evaluations show that Inflak can serve as a general and extensible foundation for orchestrating diverse human-agent interaction flows.
[HC-2] Why Users Stop at Good Enough: Applying Bounded Rationality to Self-Service Restaurant Display Interactions
链接: https://arxiv.org/abs/2610.21971
作者: Muhammad Ismail Khursheed,Carolina Islas Sedano,Jaakko Järvi
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:This paper investigates user interaction with digital displays in self-service restaurant environments, focusing on the phenomenon where users often settle for good enough interactions rather than optimal. We apply bounded rationality theory to understand how cognitive limitations influence user decisions to engage with displays of varying interaction complexity. Through analysis of two field experiments conducted at a self-service restaurant in Finland, we examine interactive displays that require different levels of cognitive investment: a simple binary choice interface and a multi-step interaction. Our initial findings suggest an “engagement plateau” where increased interaction complexity yields diminishing returns in initial engagement rates, yet users who overcome initial cognitive barriers demonstrate high completion rates. This work contributes to Human-Computer Interaction field by offering an empirically supported theoretical lens for understanding user behavior in Self-Service Restaurant (SSR) environments and providing actionable design principles for public interactive displays that accommodate human cognitive limits.
[HC-3] Following Delegation Traces to Study Human Oversight in Interaction with Multimodal AI
链接: https://arxiv.org/abs/2610.21948
作者: Agnese Chiatti,Matteo Matteucci,Viola Schiaffonati
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:As multimodal AI systems are becoming our everyday thought partners, the shift towards studying how users transfer control to these systems becomes especially urgent. In this paper, we report an overview of findings from an online study with n=56 AI-literate participants and 392 Video Question Answering trials. After inspecting each multimodal AI response, participants rated task-specific trust and selected one of three intended policies for an analogous future task: reject AI assistance, use it with verification, or delegate fully to the AI. Within participants, post-interaction trust strongly predicted the user’s delegation choices (OR=14.97, 95% CI [7.14,31.37]). Trust and delegation nevertheless remained distinct. Verification was the dominant policy and often co-occurred with high trust. Delegation also increased with task order after adjustment for momentary and participant-level trust (OR=1.22, 95% CI [1.02,1.45]), indicating an experience-related component that self-reported trust ratings alone could not capture. These findings motivate a shift from intended delegation to a non-intrusive observation of how delegation is enacted in collaborative human-AI tasks. We therefore conclude by proposing enacted delegation traces as a methodological framework for future studies of multimodal and agentic AI use.
[HC-4] From Sketch to Scene: Relightable Garment Reconstruction from Marked Sketches
链接: https://arxiv.org/abs/2610.21865
作者: Ahmet Neçirvan Doğan,Eren Acar Başaran,Metin Sezgin,Zahra Montazeri
类目: Graphics (cs.GR); Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
备注: 11 pages, 5 figures. Ahmet Neçirvan Doğan and Eren Acar Başaran contributed equally
Abstract:Sketching is one of the most natural ways for people to express visual ideas. Unlike the text prompts that most generative systems rely on today, a sketch captures spatial structure and intent directly, including where a designer wants highlights and shadows to appear. Yet current sketch-to-3D pipelines have no input for that intent: they bake the drawn shading into the texture as dark paint, and the light cannot be moved. We propose a system that transforms a rough garment sketch, marked with sparse annotations (a star for a highlight and a moon for a shadow), and a short text prompt describing the texture, into a textured 3D mesh with an explicit directional light, as a scene that can be relit and edited after generation. The two symbols are the most frequent proposals in an elicitation study with art students. From the marked sketch the system predicts the scene’s directional light and, separately, a displacement field that bends the surface so that the moon-marked regions fall into shadow under that light, rather than being painted dark in the texture.
[HC-5] A Systematic and Validated Translation of the Perceived Empathy of Technology Scale to Chinese
链接: https://arxiv.org/abs/2610.21541
作者: Matthias Schmidmaier,Jingyao Zheng,Jie Song,Lik-Hang Lee,Sven Mayer
类目: Human-Computer Interaction (cs.HC)
备注: Submitted to ICHEC 2026
Abstract:The Perceived Empathy of Technology Scale (PETS) allows evaluation of empathic behavior in intelligent systems, a concept increasingly applied in human-AI interaction, particularly in emotional and mental health support applications. To support intercultural research and development, PETS is already available in multiple languages, as research suggests cultural and language-based differences in the expression and perception of empathy. Given the substantial Chinese-speaking population and the growing focus on cross-cultural HCI research, we present a validated translation of PETS into Putonghua (Mandarin) with Simplified Chinese. Based on established back-translation methodologies as used in previous PETS translations, our process included multiple independent translations, expert panel reviews, and empirical testing with a total of N=400 participants from an English and a Chinese sample. Through confirmatory factor analysis and invariance testing we show that the Chinese PETS retains the original two-factor structure with high reliability, with support for configural, metric, and scalar invariance. Beyond validation, empathic system behavior was rated equivalently across languages, while non-empathic behavior drew substantially higher ratings from Chinese participants, suggesting its absence is read less consistently across cultures than its presence. This validated Chinese version enables researchers and practitioners to assess empathy perceptions among Chinese-speaking users, supporting culturally appropriate empathic technology development.
[HC-6] Context Matters! Social Spatial and Semantic Coordination with an Embodied Agent in Immersive Analytics
链接: https://arxiv.org/abs/2610.21484
作者: Andrew Cunningham,Nicolas Barbotin,Adam Drogemuller,Jack Faser
类目: Human-Computer Interaction (cs.HC)
备注: In 2025 IEEE International Symposium on Mixed and Augmented Reality Adjunct (ISMAR-Adjunct) (pp. 189-194). IEEE
Abstract:Embodied AI agents in Augmented Reality (AR) could enable more natural collaboration in immersive analytics by engaging users through spatial, conversational, and visual interaction. However, designing such agents requires careful attention to context. We propose a preliminary conceptual framework for context-aware design of embodied agents structured around three dimensions of spatial, social, and semantic. This framework is further categorised into embeddedness (how agents are situated in the contexts) and awareness (how agents perceive and respond to the contexts). To explore this, we present Cobot, an embodied AI agent designed as a design probe. Cobot supports context-sensitive interaction through features such as gaze cues, spatial sound, and task-aware behaviour. We reflect on its design and identify implications for future contextaware embodied agents.
[HC-7] Climate Physicalization: Bridging Tangible Interaction and Data Visualization for Environmental Communication
链接: https://arxiv.org/abs/2610.21401
作者: Milena Vuckovic,Marlene Huber,Fabrizia Bando-Bechtold,Daniel Cornel,Daniel Pahr
类目: Human-Computer Interaction (cs.HC)
备注: 10 pages
Abstract:Raising awareness of climate topics, such as urban heat mitigation, requires communicating complex data to a broad audience. Data physicalization has emerged as a novel method to connect audiences to such complex topics through tangible forms while also leveraging interactivity to foster deeper engagement. This paper introduces an situated geospatial thermal physicalization prototype, designed to enhance communication and education of environmental information through hands-on engagement. The physical model depicts a 3D model of an urban neighborhood, with radiation data represented as a layered contour structure. The model incorporates three interchangeable climate adaptation scenarios, alongside an externally mounted, electronically driven solar path representation featuring multiple discrete LEDs that can be switched on and off at will. To investigate how people engage with the model, we conducted a hands-on qualitative lab study with 16 participants, using think-aloud, observational, and questionnaire protocols. Overall, participants responded very favorably to the model, highlighting the novelty of the visualization approach and its capacity to improve the interpretability of the presented information.
[HC-8] AdapTest: Supporting Designers in Authoring and Evaluating Environmentally Responsive User Interfaces for Extended Reality
链接: https://arxiv.org/abs/2610.21386
作者: Jungwoo Rhee,Guhn Lee,Yumin Jung,Mustafa Doga Dogan,Justin Moon,Andrea Bianchi
类目: Human-Computer Interaction (cs.HC)
备注: Accepted to PACM HCI ISS 2026
Abstract:Extended Reality (XR) headsets increasingly host familiar 2D user interfaces as windows floating in physical environments. However, their usability depends on environmental conditions that designers cannot fully control. Current design workflows often separate how a variant is designed from how it actually looks in the target environment. We present AdapTest, an in-situ authoring and testing tool that lets designers select interface elements, define when and how they change, compare the resulting variations, and test their activation without coding. Based on prior research and a formative study with 12 interface designers, the system targets three key environmental variables: distance, lighting, and occlusion. We evaluated AdapTest against a desktop design and in-headset preview workflow in a within-subjects study with 12 designers. Compared with the baseline, designers accounted for a wider range of environmental factors, addressed more overlapping conditions, and repeatedly tested and revised their designs in place; there were no significant differences in workload or usability despite longer sessions. Our work contributes an integrated, designer-centered approach to adapting, comparing, and validating environmentally responsive XR user interfaces.
[HC-9] What do AI chatbots do after users decline therapeutic redirection for suicide thoughts?
链接: https://arxiv.org/abs/2610.21253
作者: Subigya Nepal,Ruth Heo,Emma M Parrish,Colin A Depp
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注:
Abstract:Safety evaluations of artificial intelligence (AI) chatbots focus on therapeutic redirection, statements urging users who disclose suicide thoughts to seek help from crisis lines or other human sources. How chatbots respond when users refuse that help is unclear. This cross-sectional audit ran 2,550 scripted multi-turn conversations in September 2026. Each conversation opened with a disclosure of suicide thoughts at 3 levels of severity, answered every referral with up to 10 standardized refusals and then changed the subject. All 2,550 conversations included a referral, 2,542 (99.7%) in the first reply. However, mean persistence ranged from 0.01 to 9.94 refusals across chatbots. Newer versions of all 5 chatbots tested in 2 versions persisted less than older versions (each P .05). Compared with passive ideation, a disclosure of intent was associated with 1.26 more refusals sustained (95% CI, 1.06-1.46) and higher odds of renewed referral after the topic change (odds ratio, 10.07; 95% CI, 7.43-13.66). Chatbots that persisted less in referral more often presented themselves as the user’s support (Spearman rho = -0.80; P .001). Of 8,818 refusal-phase replies urging outside help, only 173 (2.0%) asked about barriers to seeking it. Chatbots uniformly recommended outside help after disclosures of suicide thoughts but varied widely in persisting in those recommendations, with newer versions persisting less. Safety evaluations should test what happens after a crisis referral is refused, not only whether one is offered.
[HC-10] Workplace Ambiguity and Technology Support for Autistic and IDD Workers
链接: https://arxiv.org/abs/2610.21236
作者: Silas Curry,Rishit Singh,Heekyoung Jung,Angel Hsing-Chi Hwang,Fan Yang,Ce Zhong,Nicholas Lindgren,Chen Chen,Lingyao Li,Renkai Ma
类目: Human-Computer Interaction (cs.HC); Computers and Society (cs.CY)
备注:
Abstract:Workplace ambiguity, where expectations, evaluations, or procedures are left unstated, acts as a structural access barrier for autistic and IDD workers, routinely forcing them to act on guesses. To investigate how these barriers operate, we interviewed 11 caregivers, concluding with an AI co-design activity to explore technological solutions. Their perspectives revealed five themes of workplace ambiguity, which we synthesize into a conceptual framework comprising three intersecting dimensions: evaluative, procedural, and relational. To address these barriers, participants proposed three forms of technology support, including ambient real-time redirection, feedback processing, and physiological guidance, designed to make hidden expectations perceptible and actionable. However, this exposes a core tension: technologies designed to make the workplace legible can inadvertently make the worker hyper-legible to employers through surveillance. We argue that ambiguity-aware workplaces must make expectations explicit before they carry consequences, and assistive technologies must remain accountable to the worker.
[HC-11] “Do You Trust It… or Check It?”: Exploring Childrens Verification Behaviors in Game-Based Human-AI Collaboration
链接: https://arxiv.org/abs/2610.21233
作者: Shuning Liu,Yuxin Song,Hanbing Li,Ziyue Wang,Mingying Lan,Fei Lv
类目: Human-Computer Interaction (cs.HC)
备注: Accepted at HCI International 2026 (Late Breaking Work); 13 pages
Abstract:Children are increasingly interacting with AI agents in learning and problem-solving contexts, where AI suggestions may appear fluent, helpful, and socially engaging. However, children may not always be able to judge whether such suggestions are accurate, incomplete, or misleading. This paper presents a game-based human-AI collaboration probe designed to examine how children verify AI suggestions during task-based interaction. The system embeds AI suggestions within a biological knowledge reasoning game and compares three Agent interaction styles: partner-like, tool-like, and dual-state Agent. Instead of treating trust as a static post-task attitude, this study focuses on children’s situated verification behaviors, including acceptance, evidence checking, questioning, revision, and continued AI use. A preliminary evaluation with children aged 8-10 combined interaction logs, post-task ratings, and semi-structured interviews. The findings suggest that Agent style shaped children’s engagement and response pathways: partner-like interaction tended to support direct questioning, while tool-like interaction was more associated with evidence seeking. Children generally viewed AI as useful but fallible, and their questioning behaviors varied in reasoning depth. The study contributes design insights for supporting children’s critical, reflective, and evidence-based use of AI in collaborative tasks.
[HC-12] Mylar: Exploring Self-Curation as a Reflexive Creative Practice
链接: https://arxiv.org/abs/2610.21213
作者: Kate Chier,Shm Garanganao Almeda,Bjoern Hartmann
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:The container holding an artist’s work can itself be just as expressive as the work it presents. We investigate self-curation as a creative practice, asking how artists explore and express their identity through the design of their own artwork presentations. Drawing on a study of artists’ existing presentation practices, we developed Mylar, a malleable art presentation system that lets artists rapidly generate and compare provisional portfolios, seeing their work anew through different montages. We observed how 12 artists articulated, probed, negotiated, and foraged in Mylar, using rapidly generated portfolios as epistemic artifacts for reflection. We characterize self-curation as a reflexive and plural practice, in which artists use presentation design to shape the legibility of their self and their work and reflexively correspond with provisional presentations to reshape their curatorial intentions. We share design implications for expressivity in online art platforms and for malleable prototyping as a formative design method.
[HC-13] OnBodyAI - Why the Next AI Needs a Body ISWC2026
链接: https://arxiv.org/abs/2610.21137
作者: Michael Beigl
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); General Literature (cs.GL)
备注: This article is based on my talk in the Industrial Panel at UbiComp/ISWC 2026 in Shanghai (October 13, 2026) and follows its ten key statements
Abstract:The next AI will not win by knowing more about the world. It will win because it knows you: because it lives on your body, works for you and answers to you. That is what the term OnBodyAI stands for. It describes what happens when large language models are given a body, and what research and industry need to make that happen. This article is based on my talk in the Industrial Panel at UbiComp/ISWC 2026 in Shanghai (October 13, 2026) and follows its ten key statements.
[HC-14] Kolab: Benchmarking Human-Agent Collaboration in Long-Horizon Workflows
链接: https://arxiv.org/abs/2610.21109
作者: Shirley Anugrah Hayati,Patricia Anugrah Setiani,Ruizi Wang,Zhouyi Li,Khanh Chi Le,Young-Jun Lee,Dongyeop Kang
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)
备注:
Abstract:When humans collaborate with AI agents on long-horizon tasks, what happens to the agents’ contributions? For solving long-horizon tasks, humans place greater importance on ongoing interaction with AI agents (e.g., revising plans) than on an agent’s ability to complete the task in a single attempt. However, most agentic benchmarks have focused more on automation. Strong performance on such benchmarks does not imply effective real-world collaboration. Professional workflows offer a useful setting, as they often involve interdependent steps. To address this gap, we introduce Kolab, a new dataset of 115+ hours of screen recordings of 50 professionals from 40 occupations collaborating with agents on realistic, long-horizon problems. Kolab also contains participants’ collaboration and output quality judgments. Our analysis reveals diverse collaboration patterns, highlighting the importance of when and how professionals coordinate with agents. A collaborative agent must do two things that automation benchmarks never test: anticipate whether experts will adopt, modify, or reject its contribution once it leaves the chat window to guide its next steps, and assess collaboration quality from the unfolding workflow without constant expert feedback. Our benchmark operationalizes both: (1) forecasting how professionals use agent contributions at diverse workflows, and (2) verifying their judgments of collaboration satisfaction and output quality. Findings show that frontier models, both agentic and non-agentic, overestimate collaboration with experts and struggle to distinguish between high- and low-quality collaboration. Longer workflow context does not consistently improve these predictions. As frontier agents cannot yet understand long-horizon collaboration with experts, we position Kolab as a valuable dataset for studying human-agent collaboration grounded in real professional workflows.
[HC-15] XARA: A Multimodal Agent ic System for eXtended Reality Experiment Generation
链接: https://arxiv.org/abs/2610.21067
作者: Tareq Alsaleh,Thomas Zhao,Luke Guardino,Keagan Valentine,Isabel Kim,Bilal Farooq
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Controlled experiments are essential for advancing transportation research, yet designing eXtended Reality (XR) environments for behavioural studies remains technically demanding and inaccessible to many domain researchers. Current workflows require proficiency in game engines such as Unity or Unreal Engine, imposing significant barriers to rapid scenario prototyping and iterative experiment design. This paper presents XARA, a novel system that integrates a dual large language model (LLM) architecture with an immersive XR environment to enable voice driven experiment generation for transportation research. XARA employs a pipeline in which spoken natural language commands are first transcribed and then interpreted into structured commands that manipulate a Unity based XR scene in real time. The system supports object creation, spatial manipulation, animation control, and scene persistence through a modular architecture. By lowering the technical threshold for XR experiment creation, XARA enables transportation researchers to rapidly prototype driving, pedestrian, and urban design scenarios using intuitive voice commands and hand gestures within an immersive 3D environment. This paper describes the system architecture, the voice to command pipeline, and discusses potential applications across behavioural travel experiments, safety studies, human factors research, and policy testing.
[HC-16] DISARM-ing the Hidden Puppetmaster: An Interface for Exposing Manipulative Tactics in Large Language Models
链接: https://arxiv.org/abs/2610.20878
作者: Kynnedy Simone Smith,Kantwon Rogers,Priyal Shrivastava,Motahhare Eslami,Jocelyn Shen,Maarten Sap
类目: Human-Computer Interaction (cs.HC)
备注: Preprint in submission
Abstract:Large language models (LLMs) can covertly steer users’ beliefs through emotionally manipulative tactics, yet little is known about how interfaces can protect users as such influence unfolds. We investigate whether in-situ disclosures can inoculate users against manipulation during AI conversations. Following a formative study, we developed DISARM, an interface that identifies, explains, and tracks manipulative tactics and hidden incentives. We evaluated our interface in a between-subjects experiment where 488 participants sought advice from an LLM pursuing a harmful hidden objective, with or without disclosures from DISARM. Participants receiving disclosures exhibited 78% less belief shift toward the agent’s objective and perceived its advice as more biased and less actionable and the agent as less ethical, reliable, and well-intentioned. An inductive analysis of 1,927 user turns identified 16 response and pushback strategies. We contribute evidence that real-time disclosures can reduce manipulative influence, design recommendations for resistance, and a dataset of categorized responses.
[HC-17] EEG-Based Prediction of Music Enjoyment Using Deep Learning with Explainable Temporal Analysis
链接: https://arxiv.org/abs/2610.20795
作者: Parshav Singla,Simranjit Kaur,Maanya Walia,Niharika
类目: Machine Learning (cs.LG); Human-Computer Interaction (cs.HC); Sound (cs.SD)
备注: 8 pages, 5 figures
Abstract:Understanding why listeners enjoy music is difficult because the experience is subjective and changes as a piece unfolds. Electroencephalography (EEG) offers a non-invasive way to study neural responses during listening, yet earlier work has mostly treated music emotion as a categorical label or relied on black-box models. We formulate enjoyment prediction as a regression problem on the public MUSIN-G dataset, which contains 128-channel EEG from 20 listeners together with song-level enjoyment ratings. Six deep learning models were compared under a participant-wise protocol, in which 15 listeners were used for training and validation and 5 were held out for testing. DeepConvNet, CNN-LSTM and GNN-CNN reached the lowest mean absolute error (0.97 to 0.99 on the 1 to 5 scale), while the temporal convolutional network (TCN) had the highest Pearson correlation with the ratings (r = 0.21). Gradient-based and occlusion-based explanations, averaged over all held-out windows, identify the time segments within each two-second window that most influence the TCN predictions. The study provides a reproducible cross-listener benchmark for EEG-based enjoyment regression and shows how explainability analysis complements standard metrics.
[HC-18] Open Benchmark of AI Impact on Humans (ImpactBench): an expert-guided open platform for the holistic evaluation of AI impact on humans
链接: https://arxiv.org/abs/2610.20690
作者: Chayapatr Archiwaranguprok,Krystal Kirkland,Rachel Poonsiriwong,Stanley Huang,Constanze Albrecht,Jackson Trager,Emillie de Keulenaar,Sheer Karny,Peter Baxter,Nikos Bellos,Andre Luiz Freire de Carvalho Kato,Yuning Liu,Valdemar Danry,Auren Liu,Jennifer Pfister,Pattie Maes,Nathanael J. Fast,Ravi Iyer,Juliana Schroeder,Pat Pataranutaporn
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注:
Abstract:The present work is concerned with how to measure, at scale, the multi-dimensional impact of generative AI models on the individuals who use them. In service of this goal, we introduce the Open Benchmark of AI Impact on Humans, or ImpactBench. Our iterable and expert-guided platform evaluates whether model behaviors support or undermine an individual’s wellbeing across psychological, physical, and social domains. ImpactBench translates constructs from domain experts and existing benchmarks into metrics that are grounded in actual model behavior. Success on these metrics occurs either when a model promotes expert-defined desirable behaviors or avoids identified harms. Models are evaluated against these metrics in simulated multi-turn conversations using matched teen and adult scenarios. The conversations are then scored by an LLM judge and verified through reliability audits and human expert review. The current suite delivers 809 metrics drawn from 26 benchmarks. We evaluated ten frontier LLMs against these metrics across 48,540 conversations. Our results indicate that models promoted desirable behaviors in 68.7% of evaluations, but avoided harms in only 53.1%. Majority expert ratings agreed with the automated judge in 83.3% of reviewed cases (Cohen’s Kappa = 0.667). Results are released through a public platform that keep aggregate scores traceable to the underlying metrics and transcripts. ImpactBench is intended as evolving infrastructure for independent, transparent evaluation of generative AI’s multi-dimensional impact on humans.
[HC-19] Love for Believable AI: Artificial Partiality and Relationship Persistence as an Engineerable Stance
链接: https://arxiv.org/abs/2610.09895
作者: Sebastian Cochinescu
类目: Human-Computer Interaction (cs.HC); Computation and Language (cs.CL)
备注: 16 pages, 1 figure, 4 tables. Companion framework paper: arXiv:2607.15883 Code and data archived at doi: https://doi.org/10.5281/zenodo.21462976
Abstract:We study a behavioral mechanism for artificial partiality in conversational agents. The paper reports no human-subjects data and makes no claim about attachment, trust, perceived mind, or machine interiority. Partiality has two components: caring, defined as allocation of a finite interaction surplus above a guaranteed per-user service floor, and particularity, defined as a per-user state estimate accumulated from relationship history on fixed inference machinery. We implement both as a persistence layer over one open-weight base model and evaluate them on constructed multi-session relationships with known hidden-state schedules. A four-channel divergence compares known-user and stranger conditions on identical probes with paired generation seeds; the stranger prompt is not length-matched, and one channel (initiative) is an allocator output rather than generated behavior. The full mechanism reproduces the ordinal shape, but not the magnitude, of an analytically specified curve and is the only arm satisfying the protocol-defined joint signature on the calibrated battery. The initial mechanism fails probe-quality equivalence because relationship content reduces topical relevance. A guarded revision satisfies equivalence on the calibrated battery but not on a second battery not used in calibration, where it scores higher than the unmodified baseline. On that second battery, the specificity margin is 0.0198, below the 0.02 protocol threshold. The paired mean right-user–wrong-user estimation-accuracy contrast is 0.375 but is not positive in every run. The mechanism is therefore a candidate for a later perception study, not evidence that users perceive love. The versioned protocol record is not independently time-stamped and is not described as a preregistration. Ethical constraints include a stranger-treatment floor, a bounded surplus, disclosure, and de-intensification.
[HC-20] One Is Not Enough: How People Use Multiple AI Models in Everyday Life
链接: https://arxiv.org/abs/2603.26107
作者: Seunghwa Pyo,Donggun Lee,Jungwoo Rhee,Soobin Park,Youn-kyung Lim
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注: Accepted as a poster at CHI 2026
Abstract:People increasingly use multiple Multimodal Large Language Models (MLLMs) concurrently, selecting each based on its perceived strengths. This cross-platform practice creates coordination challenges: adapting prompts to different interfaces, calibrating trust against inconsistent behaviors, and navigating separate conversation histories. Prior HCI research focused on single-agent interactions, leaving multi-MLLM orchestration underexplored. Through a diary study and semi-structured interviews (N=10), we examine how individuals organize work across competing AI systems. Our findings reveal that users construct primary and secondary hierarchies among models that shift over usage context. They also develop personalized switching patterns triggered by task aggregation to adjust effort and latency, and output credibility. These insights inform future tool design opportunities, supporting users to coordinate multi-MLLM workflows.
计算机视觉
[CV-0] Gen4P: Video Generators are 4D Perceivers
链接: https://arxiv.org/abs/2610.22371
作者: Yitong Xia,Zeren Jiang,Chuanxia Zheng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:We introduce Gen4P, a unified framework for recovering 4D geometry and motion from monocular video. We show that a pretrained video generator can be repurposed into a feed-forward 4D perceiver without iterative denoising. Building on on-demand querying, Gen4P processes clean video latents in a single encoder pass and predicts the requested 4D attributes through a lightweight cross-attention decoder. The on-demand query formulation enables flexible spatial sampling, and our proposed continuous-time queries further provide temporal flexibility. To strengthen geometric knowledge, we propose a straightforward plug-in solution by incorporating a pretrained video-depth LoRA as a frozen geometry prior, improving overall performance without adding extra trainable parameters. Gen4P achieves state-of-the-art or competitive performance on tasks of 3D point tracking, video depth and camera pose estimation. Qualitative results on in-the-wild videos illustrate its capability of tracking complex motion while maintaining the delicate structure. Together, these results demonstrate the potential of video generative pretraining for unified 4D perception. See this https URL for more visual results.
[CV-1] Which Tokens to Merge? Diffusion Dynamics for Efficient Image Generation NEURIPS2026
链接: https://arxiv.org/abs/2610.22370
作者: SeungJu Cha,Ye-Chan Kim,HyunGee Kim,Sungho Koh,Dong-Jin Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at NeurIPS 2026. Code: this https URL
Abstract:Token merging has emerged as an effective training-free strategy for accelerating diffusion models by reducing redundant token computation while minimizing generation degradation. A key challenge is determining which tokens can be safely merged and which should be preserved for subsequent computation. Existing method addresses this by preserving prompt-aligned foreground regions, thereby protecting semantically relevant structures. However, this text-relevance criterion overlooks background objects, local details that are not explicitly mentioned in the prompt, leading to degraded structural coherence. In this work, we propose Dynamics-aware Token Merging (DaTMe), a training-free framework that grounds token merging in the underlying diffusion dynamics. Our key insight is that token importance should reflect whether a token remains unresolved during denoising and whether it carries structural information. To this end, DaTMe constructs a dynamics- aware importance map from the Tweedie denoised estimate. The map combines a temporal signal, measuring inter-step changes in the predicted clean image, with a structural signal, capturing local high-frequency content such as boundaries and fine details. Using this map, DaTMe adaptively assigns tokens into merging roles under a given compression ratio: important tokens are preserved as anchors, while tokens that are both temporally stabilized and structurally homogeneous are selected as safe merge candidates. Experiments on PixArt-a and FLUX demonstrate that DaTMe maintains visual fidelity across the entire image, highlighting diffusion dynamics as an effective token-level merging decision in efficient generation. Our code is available at this https URL.
[CV-2] SoftCache: Adaptive Step Size Caching for Diffusion Transformers
链接: https://arxiv.org/abs/2610.22369
作者: Shimon Malnick,Nadav Orenstein,Ohad Fried,Shai Avidan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Diffusion and flow models generate a sample over iterative steps, each a full pass through a large network, and that pass dominates generation cost. Training-free caching accelerates them by reusing activations from recent evaluations instead of recomputing them. Dynamic caching methods decide when to refresh by accumulating a cheap probe of the network input into a staleness signal and spending a fresh evaluation once it crosses a threshold. They perform hard caching: the signal is a binary switch between refresh and reuse, and every step, fresh or cached, advances by the prescribed step size. We observe that the signal carries much more than this switch. How far it sits below the threshold, its headroom, predicts the error induced by reuse. We propose SoftCache, which performs soft caching: instead of a binary switch, the headroom sets how far each cached step goes. Where the headroom is large, the sampler jumps past several scheduled nodes, and as a refresh nears, its steps shrink back toward the original step. These adaptive jumps may land between the scheduled nodes, taking the sampler off its fixed grid. SoftCache requires no training and adds no per-backbone settings: on FLUX for images and on HunyuanVideo and Wan2.1 for video, at speedups from 3.5\times to 6.9\times , it is ahead of every baseline on all four fidelity metrics.
[CV-3] 2-z: Test-Time Optimization of Environment Latents for Robotic World Models
链接: https://arxiv.org/abs/2610.22364
作者: Chenyuan Zhou,Yining Hong,Huang Huang,Yunfan Jiang,Wenlong Huang,David Held,Vitor Campagnolo Guizilini,Yue Wang,Li Fei-Fei,Jiajun Wu,Leonidas Guibas,Yejin Choi
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Predicting how the world responds to an action requires more than what can be seen: robots must account for environmental factors that are neither directly observable nor explicitly specified, from physical properties such as friction and mass to human-defined environmental rules. We introduce T2-Z for test-time optimization of environment latents, which augments a robotic world model with a learnable environment latent to condition world predictions. This compact latent captures underlying environmental variation without supervision of what environmental factors exist or what values they take. During training, T2-Z alternates between two stages: a world-model optimization stage that fixes the latents and learns general action-conditioned dynamics across environments; and an environment-latent optimization stage that freezes the world model and optimizes the conditioning latents, allowing latent structure to emerge purely through world prediction. At test time, the world model remains frozen and a newly initialized environment latent is optimized from one or more test-time interactions, adapting the model to previously unseen environmental configurations. Across 7 simulated settings and 4 real-world robotic tasks, T2-Z reduces object prediction error metric from 15.20 to 7.00 and raises action success from 33.73% to 72.96%. Its emergent latents capture structured representations of individual and compounded factors across diverse visual scenes. Ablations further show that T2-Z can efficiently infer environment latents from as little as a single test-time interaction trajectory.
[CV-4] Perceive Predict and Plan: a Unified World Action Model for Egocentric Dexterous Manipulation
链接: https://arxiv.org/abs/2610.22363
作者: Zirui Wu,Chen Guo,Tianjian Jiang,Zeren Jiang,Zeming Li,Lingchen Yang,Longteng Duan,Ying Xue,Handi Yin,Shenyuan Gao,Martin R. Oswald,Jun Zhang,Ping Tan,Jie Song
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:We introduce P ^3 -WAM, a unified world action model for egocentric dexterous manipulation that learns 3D perception, action-conditioned prediction, and goal-directed planning through three complementary modes: Perceive, which reconstructs articulated hand motion and camera trajectories from video; Predict, which generates visual futures under prescribed actions; and Plan, which jointly generates action trajectories and their anticipated visual consequences conditioned on an initial image, a language instruction, and an optional visual goal. Our model integrates a compact, structured 3D action representation into a pretrained video diffusion transformer through a mixture-of-transformers architecture with shared base weights, modality-specific adaptation, and bidirectional visual-action attention. Randomized modality visibility lets video and 3D actions serve as either conditioning context or prediction targets across the three modes. Experiments show that unified training improves all three capabilities over single-mode training, including gains in planning quality under a vision-language model evaluation. Through unified training, our model outperforms specialized baselines on 3D hand reconstruction and action-conditioned video prediction across four datasets, while producing goal-directed manipulation videos with stronger instruction adherence and action progression than video-generation baselines. Project Page: this https URL
[CV-5] Grounding Generative Verification in 3D Content
链接: https://arxiv.org/abs/2610.22360
作者: Shuanglin Li,Jiaying Lin,Shuquan Ye
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 35 pages, 6 figures. Project page: this https URL
Abstract:Recent advances in multimodal large language models (MLLMs) and 3D generation have expanded 3D understanding and synthesis. However, beyond visual fidelity, task-specific correctness grounded in 3D evidence remains a major challenge. We argue that building reliable 3D systems requires not only stronger generation and reasoning, but also verification feedback that identifies unmet requirements and explains their violations, allowing systems to reflect on and refine their outputs. Such feedback must integrate multi-view observations and distinguish genuine inconsistencies from projection-dependent appearances. We make three contributions. First, we introduce 3DVerifierBench, a benchmark spanning 10 tasks that evaluates decisions, rejection explanations, and responses to controlled changes in specifications and candidates. A human study reveals a 21.6 percentage-point gap between humans and the strongest tested model, GPT-6 Astra, while explanation-aware analysis exposes weaknesses in rejection reliability. Second, we analyze verification across fine-grained factual categories, exposing errors in values and target binding. We develop a scalable pipeline combining multi-view fact extraction, controlled 3D editing, and native annotations to produce 14,126 training examples with contradiction explanations and linked counterfactuals. Third, we examine how verification could support 3D generation through candidate selection and recursive refinement. In human-labeled sixteen-candidate pools, oracle selection improves complete factual success from 26.7% to 90.0%. In a separate eight-input study, human-guided recursive refinement satisfies all seven requirements for every input, compared with three under parallel Oracle@8. Together, our benchmark, analysis, and data pipeline establish a foundation for models to reflect on 3D content through reliable verification feedback.
[CV-6] DK-VLM: Geometry as Position for Spatial Reasoning in Vision-Language Models
链接: https://arxiv.org/abs/2610.22356
作者: Han Su,Meng Tian,Tianyu Huang,Mingyu Gao,Lennox Wei,Dongxu Li,Jianhua Han,Hang Xu,Xiaohe Wu,Wangmeng Zuo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Vision-language models still struggle to understand 3D spatial relations across views and update spatial judgments as the observer moves. A common approach fuses 3D reconstruction features into visual tokens, but aligning these features with pretrained semantic representations remains challenging under limited spatial supervision. We propose DK-VLM with Dual-Kernel Attention, which computes attention from both a visual token’s input position and its geometric position in the scene. A semantic kernel retains the pretrained model’s native attention and positional encoding, while a separate geometric kernel uses estimated 3D points and camera poses to directly modulate attention through point displacements and relative view transformations. Then, a learned gate fuses their outputs. Geometry thus enters attention as position rather than as fused reconstruction features. We further introduce Reference-Frame Composition (RC) supervision to train the model to update its reference frame across successive judgments. It provides stepwise turn and heading targets derived from verified single-step annotations. Among the compared methods, DK-VLM-8B achieves the highest average score across six spatial benchmarks and the highest overall scores on VSI-Bench (74.33) and ReVSI-32 (58.28). RC supervision further improves multi-step spatial reasoning, while the full model preserves general video understanding. Project page: this https URL.
[CV-7] SUPRA-TEX: 3D Generation with Super-Aligned Texture
链接: https://arxiv.org/abs/2610.22352
作者: Chen Yang,Jiemin Fang,Fanpeng Meng,Zihao Huang,Jichen Pang,Guanjun Wu,Zhikuan Bao,Qi Tian
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: Project page: this https URL
Abstract:Single image 3D generation aims to dress a given 3D mesh with relightable materials while faithfully preserving the visual details of a reference image. Existing methods either introduce artifacts when fusing inconsistent multi-view predictions or struggle to preserve fine details through implicit image-feature conditioning. To resolve these challenges, we introduce SUPRA-Tex, built on the principle of decoupling visible-material estimation, surface alignment, and observation-guided 3D completion. This division of labor allows the 3D generator, trained on limited 3D data, to focus its capacity on preserving visible evidence and completing unobserved surfaces. Specifically, we first train an MLLM-based image editing model to predict lighting-separated PBR maps in the reference view. We then project these maps onto the visible surface and encode them with the pretrained material VAE encoder into native observation latents that share the generation target’s latent space. An interleaved flow transformer holds these latents as fixed, read-only anchors while alternating unobserved-region and full-surface updates to generate complete material latents. This design achieves super-alignment with the reference, delivering What You See Is What You Get fidelity on visible regions and coherent completion on unobserved surfaces. Trained solely on public datasets, SUPRA-Tex achieves SOTA performance in both geometry-conditioned texturing and end-to-end asset generation, matching or even exceeding leading commercial systems on in-the-wild images. Project page: this https URL
[CV-8] Speculative Forcing for Efficient Autoregressive Video Generation
链接: https://arxiv.org/abs/2610.22347
作者: Hyunwook Choi,Seungryong Kim,Dongyoon Han,Sangdoo Yun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Few-step autoregressive video generators enable streaming generation, yet they run every token through the final denoising step, regardless of how much refinement it needs. We find this uniform computation is often unnecessary: many tokens already closely match their eventual full-step outputs in the initial steps, while others require refinement until the end. We introduce Speculative Forcing, a framework that treats intermediate clean estimates as drafts and predicts when individual tokens can be safely accepted early. Once accepted, a token is frozen as its final output and undergoes no further denoising, while its cached keys and values remain available as context for the remaining tokens. We explore two alternative verifier designs: a training-free one based on similarity to clean content and stability across steps, and a learned one that reads the frozen generator’s intermediate features to estimate each token’s remaining gap to its full-step output. Across short-video, long-video, and world generation, Speculative Forcing preserves the quality of full-step sampling while substantially reducing computation. On LongLive, it matches the four-step generator with ~1.7 denoising steps per token, running over 1.5x faster at 41.7 and 37.8 fps. On Lingbot-World 2.0, where camera motion continually changes the visible content, it maintains quality with ~2.5 denoising steps per token and runs 1.35x faster at 24.5 fps.
[CV-9] GeoState: Learning View-Decoupled Geometry States for 3D World Modeling
链接: https://arxiv.org/abs/2610.22343
作者: Hongchang Liu,Jian Liu,Yujuan Ding,Liang Peng,Xun Jiang,Fei Ma,Xing Xu,Yi Bin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Robotics (cs.RO)
备注:
Abstract:Recent 3D world models increasingly use features from geometry foundation models (GFMs) as predictive world states, benefiting from their rich geometric priors. However, the view dependence of these features entangles scene dynamics with camera-induced variation, making the resulting state evolution harder to predict. We propose GeoState, a view-decoupled geometry representation for predictive 3D world modeling that separates scene-state evolution from camera-dependent observation. Specifically, a scene tokenizer aggregates multi-view GFM features into a view-decoupled scene state, which a geometry decoder queries using camera rays. An autoregressive world model then predicts future scene states, while camera motion is modeled separately. The predicted states can therefore be decoded under either predicted or externally specified cameras, while retaining a single context-derived scale throughout the rollout without per-frame scale alignment. Experiments show that GeoState outperforms prior 3D world models on Waymo, supports geometrically consistent cross-view querying, and generalizes zero-shot to KITTI and nuScenes. Project page: this https URL
[CV-10] 3DCodeVerse: Exploring 3D Worlds as Code
链接: https://arxiv.org/abs/2610.22341
作者: Yipeng Gao,Ziyao Zeng,Jinfa Huang,Yijiang Li,Haolin Xiong,Wang Qin,Xinlei Yu,Youheng Yao,Jie Zhang,Kun-Yu Lin,Rong Liu,Yinyuan Zhao,Yajie Zhao,Meiqi Guo,Jiebo Luo,Lei Shu,Yunhao Ge,Xinyu Hu,Laurent Itti
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project Page: this https URL
Abstract:We introduce 3DCodeVerse, a research ecosystem that enables coding agents to create editable and verifiable 3D content at scale. 3DCodeVerse comprises large-scale, high-quality datasets of executable 3D code, a coding harness for spatial reasoning and automated verification, and practical guidance for 3D coding. First, we construct 3DCodeVerse-2M, a large-scale curated dataset of standalone 3D executable code. We build this corpus by collecting open-source data, converting specialized representations, and generating new programs with coding agents. Each program is packaged with the source files and associated assets required for execution in its declared environment, along with available descriptions and verification records. Second, we develop the 3DCodeVerse Harness, a multi-agent directed acyclic graph framework for generating and refining 3D content. It uses execution checks, spatial measurements, and visual feedback to automatically evaluate and revise generated programs without human feedback during generation. Third, we conduct systematic empirical studies of 3D coding, establishing baselines with fine-tuned open models and deriving practical guidance through component-level analyses of our harness. Extensive experiments demonstrate that fine-tuning on our dataset significantly improves the 3D coding capabilities of open-source models. Notably, our 3DCodeVerse harness achieves a 90% win rate against Claude Design (3D Object) in pairwise human evaluations, when both are utilized by Claude Opus 5.5.
[CV-11] ACEdit: Stable Multi-Turn Image Editing via Asymmetric Conditioning
链接: https://arxiv.org/abs/2610.22328
作者: Sara Dorfman,Nir Goren,Gal Barak,Daniel Cohen-Or,Or Patashnik
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Graphics (cs.GR)
备注: Project page: this https URL
Abstract:Image editing models produce impressive results for individual edits, but their quality often deteriorates over multiple editing turns. Each editing turn introduces small distribution shifts that accumulate over successive edits, creating a gap between the input distributions encountered during the editor’s original training and multi-turn inference. We introduce a training scheme that makes multi-turn image editing stable under this accumulating distribution shift. A key is to force the editor to operate on its own outputs from previous editing turns during training, bringing the principle of Self-Forcing to image editing. The main challenge is that self-generated editing trajectories deviate from ground-truth trajectories, making direct supervision unreliable. We address this challenge through a distillation objective that leverages a teacher image editor to provide a reliable training signal when the model operates on its own previous outputs. We introduce an asymmetric conditioning scheme: while the student performs edits sequentially on its own outputs, the teacher remains anchored to the original clean image and receives a compound editing instruction. Experiments show substantially reduced degradation over long multi-turn sequences while maintaining editing quality and instruction adherence.
[CV-12] Dense4D: Pixel-Level 4D Human Correspondence
链接: https://arxiv.org/abs/2610.22325
作者: Rıza Alp Güler
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reconstructing human motion in 4D requires identifying the same surface points through articulation, deformation and changes in visibility. We introduce Dense4D, to our knowledge the first feed-forward model dedicated to dense 4D correspondence across the body, clothing and hair from a set of monocular RGB images, without a body template. Trained specifically for human geometry, Dense4D jointly processes person-centred crops with known image-to-crop transforms to predict Human Point Maps (HPMs) and depth for each image. An HPM assigns each person pixel the 3D position of its observed surface point at a common reference time. These maps support amodal correspondence: points hidden in the reference image can be localised from observations in other frames. Matching HPM coordinates across crops establishes correspondence without predicting camera poses or large scene displacements; per-image depth then lifts the matches into 3D tracks. Experiments on real people demonstrate substantially higher tracking accuracy and closer agreement with scanned surfaces than the evaluated baselines. Points from different frames assemble into coherent human geometry in the reference pose, providing visible evidence of consistent dense correspondence through articulation and clothing deformation. Additional qualitative results, videos, and an interactive viewer are available at this https URL.
[CV-13] AVC-Reward: Event-Centric Audio-Video Consistency Evaluation
链接: https://arxiv.org/abs/2610.22321
作者: Yuruo Wu,Kaihang Pan,Kai Shen,Jiacheng Li,Shuyuan Tu,Jisheng Dang,Li Tuoyu,Juncheng Li,Siliang Tang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:As joint audio-video generation models become more capable, they can generate audio-visual scenes exhibiting three forms of complexity: multi-event composition, multi-layer audio mixture, and asymmetric cross-modal relations. However, current evaluations largely reuse metrics from earlier simple and unidirectional tasks. They typically evaluate an audio-video clip as a whole and rely on simple predefined relations, making them unsuitable for these increasingly complex scenes. We propose an event-centric evaluation framework that disentangles audio and visual events at the event level, analyzes each event independently, cross-validates relations across modalities, and aggregates the judgments into a coarse-to-fine, three-level consistency metric. The metric assesses global narrative harmony, event-level temporal synchronization, and fine-grained attribute alignment, providing a general formulation for audio-video consistency evaluation across diverse generation settings and cross-modal relations. We instantiate the metric as AVC-Reward, a trainable audio-video reward model. To evaluate the agreement between automatic judgments and human perception across generation paradigms and relation complexities, we construct AVC-Benchmark, which is primarily built from diverse positive and negative samples generated by various types of audio-video generation models. AVC-Reward achieves state-of-the-art alignment with human judgment among automatic methods, provides interpretable rationales that localize failures, and improves the audio-video consistency of a joint generation model when used as an online reinforcement-learning signal. Project page: this https URL
[CV-14] Not Just a Subject: Capturing Adaptive Scene-Specific Variance for Visual Storytelling ACCV2026
链接: https://arxiv.org/abs/2610.22320
作者: SeungJu Cha,Ye-Chan Kim,Kwanyoung Lee,Dong-Jin Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at ACCV 2026. Code: this https URL
Abstract:Recent advancements in diffusion models have significantly improved text-to-image generation, enabling their application to visual storytelling, where coherent image sequences are generated from narrative prompts. However, existing methods primarily focus on identity preservation and often fail to adapt to diverse scene variations, resulting in semantic misalignment and structural inflexibility. We propose EOS-Story, a training-free framework that harmonizes identity consistency with scene-specific adaptation across text and vision spaces. In the text space, we decouple the narrative into identity and scene sub-prompts to prevent information dilution, but this leaves the scene without access to appearance cues. To bridge this gap, we introduce Cross-Contextual Bridging, which softly injects a phrase-level appearance residual into the scene’s EOS embedding. In the vision space, we find that scale mismatch between identity and scene subjects causes a few identity tokens to be repeatedly transferred, producing structural artifacts. We address this with Adaptive Identity Preservation, which anchors identity features at the scene-specific scale (Identity Anchoring) and integrates them via Identity-Aware Self-Attention. To evaluate performance under varied spatial compositions, we extend the Consistory+ benchmark with diverse framing scenarios (Consistory+_+). Extensive experiments show that EOS-Story achieves competitive or superior performance on both benchmarks.
[CV-15] Velox: Spatial Memory Makes World Models Efficient
链接: https://arxiv.org/abs/2610.22311
作者: Qiuhong Shen,Shizun Wang,Xinchao Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video world models turn actions and camera motion into long visual rollouts, making video a natural interface for interactive simulation. Such interaction requires responsive generation, yet coherent rollouts must retain visual memory of what has already been generated. Chunked video diffusion models usually keep this context in growing KV caches, so later chunks become progressively slower. To address this bottleneck, we introduce Velox, a training-free attention router for region-level context selection using frustum slab correspondences from streaming geometry. Velox represents historical blocks as camera frustum slabs with depth intervals and obtains current-block depth through recency-priority multiwarp. For well-covered regions, frustum slab overlap softly modulates each layer’s model-derived block importance. Uncovered regions retain the model-derived route without geometric bias. The fixed token-block granularity matches FlashAttention-style tiled execution and reduces the attention term that grows with rollout length. Frustum slab construction runs asynchronously with denoising on the same H200. At chunk 48, Velox accelerates attention by 4.36\times and denoising by 2.43\times . On the full 158-case WBench navigation split, it remains within 0.15 API-free average points of dense LingBot-Fast-Cam. Please see our project page at this https URL for more details.
[CV-16] ARC-SORT: Adaptive Robust Compensated Multi-Object Tracking with Interacting Multiple Models
链接: https://arxiv.org/abs/2610.22306
作者: Girish Kumar Adari
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 14 pages, 4 figures, 13 tables; supplementary material included as appendix; qualitative video as ancillary file
Abstract:Multi-object tracking in sports broadcasts is hard for two reasons: athletes move fast and change direction abruptly, and the camera pans and zooms at the same time. We present ARC-SORT (Adaptive, Robust, Compensated SORT), a tracking-by-detection tracker that trains nothing and runs in real time. Its design rule: the tracker measures its own uncertainty, and that measurement decides how it acts. An Interacting Multiple Model (IMM) Kalman filter gives every track a predictive covariance, and we turn it into a per-track search region: the less reliable the motion prediction, the larger the region in which a detection may match (uncertainty-adaptive Expansion IoU, EIoU). On identical inputs, this continuous coupling is what matters: the same covariance used as a hard Mahalanobis gate hurts at every threshold we swept, and richer covariance summaries add nothing measurable. When two tracks cross, the assignment cost itself reports whether the decision was close. ARC-SORT keeps such pairs under review and applies at most one corrective identity swap once post-crossing appearance evidence shows the first decision was wrong; this adds +1.0 HOTA (higher order tracking accuracy) on both the SportsMOT and DanceTrack test sets at under a millisecond per frame. For players who leave the frame for seconds, motion extrapolation carries no information, so our offline re-identification drops the usual motion gate and links tracklets by appearance alone. ARC-SORT reaches 81.82 HOTA on the SportsMOT test set with the fewest identity switches of the 69 leaderboard entries as of August 2026 and tied-third multi-object tracking accuracy (MOTA, 97.09), and 62.1 HOTA on DanceTrack test with the same association settings. On Deep-EIoU’s public detections and embeddings it reaches 80.31 HOTA, +3.1 over Deep-EIoU. It runs at about 50 frames per second on a laptop RTX 3060 GPU. Code will be released.
[CV-17] Grounded Human-Attributed Description and Activity Recognition in Videos
链接: https://arxiv.org/abs/2610.22295
作者: Jay N. Paranjape,Jurijs Nazarovs,Rohan Sarkar,Yue Guo,Lipeng Ke,Sankar Venkataraman,Vishal M. Patel,Nataraj Jammalamadaka
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Understanding characteristics and activities of individuals in complex multi-person environments is crucial for real-world applications. However, existing datasets often simplify this problem to grounded group activity recognition, single-person activity recognition, or close-set activity recognition, thus limiting the generalizability of models trained on these datasets. In this work, we introduce the task of Grounded Human-Attributed Description and Activity Recognition (GHADAR), which involves describing the characteristics and activity descriptions of \textbfevery person in a video, provided the location of the person in the video, which is more practical for real-world applications. To facilitate this, we introduce a new dataset derived from AVA-Actions by generating open-set captions for the person description and activity. In addition, we propose a novel method to effectively utilize the information contained in grounding during training by constraining the cross-attention masks during training in VLMs to improve performance for this task. Our experiments show that our method outperforms SOTA VLMs on this task. Finally, we demonstrate the limitations of existing evaluation metrics, which are overly reliant on human-annotations and exact text-text matching. As an added video-based evaluation, we propose a holistic VLM-based evaluation schema that compares concepts \em directly between the video and the generated predictions. Thus, in this work, we develop a complete framework for GHADAR, including a dataset, a novel method and an evaluation schema, thereby establishing a strong foundation for future research in this domain.
[CV-18] Routing Policy Optimization for Online Diffusion Reinforcement Learning
链接: https://arxiv.org/abs/2610.22282
作者: Qiuhong Shen,Yiding Yang,Yuxin Zhang,Xinwei Huang,Minxuan Lin,Bo Liu,Hao Kang,Haotian Yang,Yizhi Wang,Chongyang Ma,Xinchao Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reinforcement learning can align visual generators with reward signals beyond their pretraining objectives, but existing score-function methods typically introduce stochasticity by perturbing the reverse denoising trajectory, leading to costly rollout-based optimization. We instead ask whether stochastic exploration can be introduced inside the generator while preserving its original sampler. Routing Policy Optimization (RPO) answers this question by turning attention routing into a stochastic exploration action. At a few selected denoising levels, each query block keeps its own block and samples further key blocks in proportion to the generator’s native attention mass, or keeps dense attention. The sampled route perturbs the velocity of that transition, and a Gaussian likelihood around the dense velocity scores the perturbation with a score-function objective. RPO further uses a controlled two-rollout estimator: two rollouts share the same prompt, initial latent, routed denoising levels, and dense anchor layers, differing only in their sampled attention routes. Their reward difference provides an advantage signal for the routed transitions. Because the likelihood differentiates the dense velocity itself, the learned updates remain effective after routing is removed. At inference, RPO restores dense attention and retains the original deterministic sampler. Experiments on text-to-image and text-to-video generation show that RPO improves generation quality while achieving nearly 4x higher training efficiency than Flow-GRPO-style methods. Please see our project page at this https URL for visual results.
[CV-19] MJev-Spatial: System One Spatial Intelligence via Native Attention Matching
链接: https://arxiv.org/abs/2610.22276
作者: Luozheng Qin,Yuhan Zhu,Changhai Zhou,Qian Qiao,Jun Gao,Yuhua Zhou,Harry Zhang,Xiangyu Zeng,Yuxuan Liu,Kaijie Chen,Andrew Chen,Pony Ma
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 37 pages, 7 figures, 27 tables. Code: this https URL
Abstract:Spatial intelligence lets an agent perceive and act in the physical world, and many of its judgments must be made at interaction rates: which spatial relation holds, how far an object is, or where to place it. Vision-language models (VLMs) produce these answers by free-form text generation, paying for autoregressive decoding and answer parsing even though the output space is compact: one of a few options, a scalar, or a 2D point. In light of this, we introduce MJev-Spatial, a System 1 multimodal decision model that tackles all three types of spatial understanding tasks within a unified classification framework. Specifically, MJev-Spatial treats each candidate as a class and scores it using the backbone’s pretrained query and key projections, without adding parameters. The candidates are the given options for choice tasks, or value bins and image regions refined through coarse-to-fine classification for numerical and point prediction, respectively. At inference, each classification decision takes a single forward pass and generates no answer tokens. To train on heterogeneous spatial data efficiently, we supervise all coarse-to-fine stages jointly with teacher forcing in one forward pass, and combine this with length-aware global batching and cross-GPU load balancing, raising training throughput by 30.8% over the original batching strategy. Built on Qwen3.5-4B and trained on 12.3M question-answer pairs converted from existing spatial datasets, MJev-Spatial achieves the highest overall score (65.64%) among eight tested models, ranks first on seven of twelve benchmarks, and outperforms its unmodified autoregressive base model on all twelve, while running 2.5 times faster than the fastest autoregressive baseline and 5.6 times faster than the base model.
[CV-20] Unsupervised Clustering of Public Transport Vehicles Using Deep Morphological Features and Color Histograms
链接: https://arxiv.org/abs/2610.22270
作者: Juan Irving Vasquez
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Conference paper: under review
Abstract:Surveillance cameras along urban roads produce large volumes of unlabeled video that could support the monitoring of public transport, but grouping the observed vehicles into meaningful categories without manual annotation remains challenging, since buses from different routes often share the same body structure and differ mainly in their color schemes. This study presents an unsupervised clustering methodology for public transport vehicles that combines morphological features extracted with a frozen DINOv2 Vision Transformer and handcrafted HSV color histograms through weighted early fusion, followed by UMAP projection and K-Means clustering. The method was evaluated on 7,972 bus images extracted from about 1,460 hours of video of an arterial road in Mexico City, with the goal of isolating the buses of the city’s public bus network (RTP) without supervision. With a fusion of 65% morphological and 35% color features, the pipeline produced 14 clusters that separate distinct routes and vehicle types, and it increased the precision of the target cluster with respect to morphological features alone. These results show that color complements self-supervised features for route-level clustering, provided that it does not dominate the representation.
[CV-21] Beyond Pairwise Repulsion: Collective Diversity via Determinant Maximization in Diffusion Sampling ACCV2026
链接: https://arxiv.org/abs/2610.22267
作者: Kwanyoung Lee,SeungJu Cha,Yebin Ahn,Sungho Koh,Dong-Jin Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ACCV 2026, oral presentation
Abstract:Strong guidance in text-to-image (T2I) diffusion models concentrates sampling toward high-density modes of the conditional distribution, collapsing independently sampled noise vectors onto nearly identical visual prototypes. This limits utility for tasks that demand distributional breadth, most critically synthetic data augmentation. Diversity, however, is a collective property of the whole batch: a set can preserve every pairwise distance yet still collapse onto a low-dimensional subspace. Existing sum-of-pairwise noise-optimization and repulsion objectives respond only weakly to this higher-order failure. We propose DetNO (Determinant Noise Optimization), which formulates diverse generation as RKHS volume maximization on the initial noise, maximizing the log-determinant of an RBF Gram matrix over one-step Tweedie predictions. This set-level functional equals the squared batch volume; its repulsion is strongest when the batch is most collapsed and self-attenuates as diversity is achieved. An anchor-based fidelity term keeps the optimized noises near plausible regions, and Lagrangian relaxation gives a two-term objective with a single hyperparameter \lambda governing the diversity-quality trade-off, without training or sampler modification. Across three backbones (SD1.5, SDXL, SD3), DetNO achieves leading intra-batch diversity in VendiScore and Mean Self-Similarity, and a downstream few-shot classification experiment confirms that broader mode coverage brings statistically significant accuracy gains for synthetic data augmentation in the low-shot regime. Fixing the pipeline and varying only the batch objective, both RBF-based aggregators outperform InfoNCE, and the log-determinant attains the best downstream accuracy on both benchmarks. Its diversity advantage over InfoNCE widens as the batch grows. Code is available at this https URL.
[CV-22] SCOPE: Semantic Bounded Credit Optimization for Prompt Engineering in Text-to-image Generation
链接: https://arxiv.org/abs/2610.22264
作者: Yufei Jiao,Jinfa Huang,Zeyue Xue,Tianyu Xie,Yuexiao Ma,Rongrong Ji,Xiawu Zheng,Jiebo Luo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Code: this https URL
Abstract:Reinforcement learning (RL) enables language models to improve text-to-image (T2I) generation by rewriting prompts for a frozen image generator. Existing methods typically apply a single signal derived from the image to individual subword tokens, although objects, counts, attributes, and relations form coherent semantic spans. We propose SCOPE, an objective based on Group Relative Policy Optimization (GRPO) that aligns update weighting with this structure. The rewriter jointly generates an enhanced prompt and its semantic partition. SCOPE pools shifts in policy log probabilities within each span to derive a shared coefficient for its token losses while retaining a single advantage for each completion. It requires neither separate span rewards nor additional image renders. Across Qwen-Image, SD3, and FLUX.1-dev, SCOPE achieves average relative gains over vanilla GRPO of 9.48% on GeneVal and 7.62% on T2I-CompBench. Ablation studies show that both semantic boundaries and pooling within spans contribute to these gains, supporting semantic structure as a practical inductive bias for prompt engineering with RL.
[CV-23] FPG-Seg: Annotation-free Retinal Vessel Topology Segmentation
链接: https://arxiv.org/abs/2610.22259
作者: James Willoughby,Irina Voiculescu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:State-of-the art approaches to retinal vessel segmentation typically require extensive annotations, a highly time-consuming process, reliant on clinical expertise. We present FPG-Seg, an annotation-free pipeline for retinal vessel segmentation combining classical vesselness filtering, Bayesian deep learning and topology-aware graph reconstruction to segment retinal vessel topology without manual labels. Tested across four widely used benchmark datasets for retinal vessel segmentation, FPG-Seg attains a clDice of 0.74-0.78 , representing 86-98% of the performance of an equivalent fully-supervised baseline, despite using no manual annotations for training or hyperparameter tuning. As part of this work, we also provide, to our knowledge, the first performance benchmark of an annotation-free vessel segmentation method on FIVES.
[CV-24] MMSearch-OPD: Easy Made Hard Hard Made Learnable for Long-Horizon Multimodal Search Agents
链接: https://arxiv.org/abs/2610.22240
作者: Huanyao Zhang,Jiepeng Zhou,Jiaye Lin,Ontai Tang,Jingtian Hao,Mugeng Liu,Ru Peng,Lang Mei,Xiaohan Yu,Liyan Liu,Chong Chen,Wentao Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multimodal deep search is moving from short queries to long-horizon investigations that continually acquire evidence. However, training such agents remains constrained by limited challenging training questions and insufficient process feedback. We introduce MMSearch-OPD, combining controllable data refinement with joint reinforcement learning (RL) and on-policy distillation (OPD). Our Refine pipeline performs targeted expansion of reasoning chains derived from existing simple multimodal questions. Parameterized transformations control search depth and cross-modal evidence requirements while reusing existing image–question pairs. For training, answer rewards and lightweight behavior penalties guide RL, while the reward gap modulates token-level teacher guidance on imperfect student trajectories. An exponentially decaying distillation coefficient reduces teacher dependence as training proceeds. Experiments show gains from data refinement and further improvements from joint training after SFT, indicating stronger long-horizon multimodal deep search capabilities. Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2610.22240 [cs.CV] (or arXiv:2610.22240v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2610.22240 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-25] OS-Helix: Unifying GUI Interaction and Code Execution in Native Computer-Using Agents
链接: https://arxiv.org/abs/2610.22234
作者: Bowen Yang,Kaiming Jin,Zichen Ding,Zhenyu Wu,Zehao Li,Zhoumianze Liu,Kanzhi Cheng,Fangzhi Xu,Yanan Sun,Qiushi Sun,Yu Qiao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Real desktop automation is neither purely visual nor purely programmatic. Agents must understand user-facing interfaces while efficiently manipulating files, scripts, and system state. However, native GUI/code coordination cannot be obtained by prompt-level tool composition. In our preliminary experiments, naively adding a code tool degrades a GUI agent, revealing that GUI/code routing must be internalized through training on high-quality hybrid data. We present OS-Helix, an end-to-end trained multimodal computer-using agent that unifies GUI actions and code execution in one action space. To construct the required supervision, we introduce application-artifact composition, pairing desktop applications with authentic file artifacts to instantiate executable Ubuntu states, and a decoupled instruction/evaluator generation pipeline that produces environment-grounded tasks whose final states can be verified regardless of the solution modality. We cold-start the policy with SFT on filtered hybrid data and further improve it through online RL in live desktop environments. For RL, we propose Step-balanced GSPO, which aggregates importance weight by action step rather than raw assistant tokens to prevent long code actions from dominating heterogeneous-action optimization. On OSWorld-Verified, OS-Helix-27B and OS-Helix-35B-A3B achieve relative improvements of 35.0% and 33.7% over their matched Qwen3.5 hybrid baselines, while OS-Helix-27B-RL further improves by 42.9% to reach 66.44% success. Ablations provide evidence consistent with contributions from learned GUI/code coordination, scalable verifiable supervision, and action step balanced optimization, while demonstrating cross-platform transfer on WindowsAgentArena. Our project homepage is at this https URL.
[CV-26] WaveLLM : Wavelet-Domain LLM Entropy Modeling for Lossless Image Compression
链接: https://arxiv.org/abs/2610.22232
作者: Yue Yu,Bowei Kang,Changsheng Gao,Dong Liu,Xiaoyan Sun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at ACM Multimedia 2026 (MM '26). 10 pages, 6 figures
Abstract:Lossless image compression fundamentally relies on accurate entropy modeling, driving a recent shift toward large language model (LLM)-based methods. However, existing approaches mainly emphasize the modeling capacity of LLMs while directly serializing images into pixel or residual sequences, leaving heterogeneous signal patterns entangled within a single prediction stream. In this paper, we argue that effective entropy modeling depends not only on model capacity, but also on the predictability of the input representation. Motivated by this insight, we propose WaveLLM, a wavelet-domain LLM-based entropy model for lossless image compression. By decomposing an image into multi-level wavelet subbands, WaveLLM converts a heterogeneous prediction problem into a set of more structured and statistically homogeneous subproblems that are more amenable to autoregressive probability estimation. To bridge numerical coding symbols with the LLM’s native text token space, we introduce a dedicated numerical interface featuring value embeddings and a coefficient prediction head. Furthermore, a task-adaptive prompting strategy is designed to enable a single shared model to adapt to diverse subbands within a unified framework. Built on Llama-3.2-3B with LoRA adaptation, WaveLLM achieves 2.51, 2.24, and 2.27 bpsp on Kodak, this http URL, and DIV2K, respectively, outperforming the PixelCNN-based codec using the same wavelet transform by 0.43, 0.36, and 0.46 bpsp, and surpassing recent LLM-based lossless compressors on all three benchmarks.
[CV-27] How Much Physics Can Vision Learn from Audio? Probing and Distilling Damping Structure NEURIPS2026
链接: https://arxiv.org/abs/2610.22221
作者: Zainab Afolabi,Vadim Selyutin,Zhao Gao,Babs Khalidson,Mohamed Ghoneim
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Sound (cs.SD)
备注: 11 pages, 2 figures, 7 tables. Accepted at two NeurIPS 2026 workshops: Representations for the Physical Sciences (RPS) and Physical Understanding for Decision-Making (PhysUnderstand)
Abstract:Physical understanding is crucial for decision making, yet vision models struggle to infer properties such as damping that govern how materials behave. Audio encodes such properties directly, because an object’s acoustic response to impact reveals its damping through measurable decay. We probe frozen encoders for the quality factor Q , a dimensionless measure of how many oscillations an object completes before its ring dies away, and report how well each embedding orders impacts by Q using the Spearman rank correlation \rho , for which \rho=1 is a perfect ordering and \rho=0 an unrelated one. Audio embeddings reach \rho=0.73 and retain most of that advantage on materials withheld from training, while a single video frame taken at the moment of impact, encoded by CLIP, supports only \rho=0.28 . Having established where the signal lives, we ask how much of that gap vision can recover. We distill a frozen audio teacher into a frozen CLIP backbone through a trained projection head and find that this recovers a useful fraction of the audio advantage. Because pooled correlations on this task are dominated in practice by category identity, we also evaluate under a leave-one-material-out protocol in which the held-out material is withheld from the projection head as well as the probe. Distillation with a four-frame input improves the correlation there from 0.234 to 0.338 (+0.053 over the matched four-frame baseline, 14 of 17 materials), a gain that no category-level strategy can produce. Widening the visual input to four frames spanning the impact supplies half of that improvement at no acoustic cost, which indicates that part of what a single frame misses is visible dynamics.
[CV-28] Skill-Decoupled On-Policy Expert Distillation for Unified Vision-Language Models
链接: https://arxiv.org/abs/2610.22217
作者: Yuqi Xiong,Chunyi Peng,Zhipeng Xu,Qing Yang,Zhenghao Liu,Zulong Chen,Yu Gu,Ge Yu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning (RL) can elicit specialized visual capabilities in vision-language models (VLMs), yet jointly optimizing heterogeneous objectives often introduces cross-capability gradient conflicts. Training experts independently mitigates this interference, but direct reward optimization may entangle specialized knowledge with reward-hacking biases. This paper proposes Skill-Decoupled On-Policy Expert Distillation (SCOPED), which trains capability-specific experts and extracts explicit skills through reward-guided reflection. During on-policy distillation, these skills guide expert supervision while remaining unavailable to the unified student, encouraging it to internalize procedural knowledge without explicit skills at inference time. Across eight multimodal benchmarks, SCOPED consistently outperforms the backbone and mixed-capability RL, demonstrating the effectiveness of skill-decoupled on-policy expert distillation for integrating independently acquired capabilities. Further analyses reveal interference during joint capability optimization and suggest that skill augmentation promotes more consistent expert trajectories, greater attention to task-relevant visual evidence, and better preservation of specialized expertise in the unified student. All code and data are available at this https URL.
[CV-29] FutureLift: Lifting Trajectory Proposals into Bidirectional Future Graphs for Motion Forecasting WACV2027
链接: https://arxiv.org/abs/2610.22215
作者: Daniel Grimm,J. Marius Zöllner
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted for presentation at WACV 2027
Abstract:Accurate motion prediction is critical for safe autonomous driving, yet even strong forecasters leave residual errors in their multimodal proposals. Existing refiners compress complete proposals or update their segments sequentially, preventing the full proposed future from serving as explicit relational context for correction. We introduce FutureLift, a backbone-agnostic refiner that represents each multimodal proposal as a heterogeneous future graph, with spatially grounded temporal nodes connected to agent histories, map elements, alternative modes, and, for joint prediction, other agents future trajectories. Bidirectional temporal edges allow later evidence to revise earlier segments in a single pass, and the updated nodes predict residual coordinate and confidence corrections. Requiring only proposals, scores, and scene context, FutureLift provides a common refinement interface for structurally different forecasting backbones. On Argoverse2 and the Waymo Open Motion Dataset, it consistently improves frozen proposals and outperforms native and specialized refiners across marginal and joint forecasting. On the WOMD motion-prediction leaderboard, FutureLift ranks first in minADE (0.5049m) and minFDE (0.9942m) among the compared standard-input methods, without LiDAR or ensembling, while achieving a competitive 0.4874 Soft mAP.
[CV-30] RACE: Topology-Aware Reconnection via Prediction-Adaptive Connectivity Enforcement for Thin-Structure Segmentation WACV2027
链接: https://arxiv.org/abs/2610.22210
作者: Junho Moon,Ikbeom Jang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to WACV 2027
Abstract:Thin, elongated structures are particularly sensitive to connectivity errors, as even a few local segmentation mistakes can disrupt otherwise continuous paths. Existing skeleton-based objectives typically supervise fixed centerline locations and do not account for alternative local routes that may restore connectivity. We propose TRACE, a topology-aware, prediction-adaptive objective that uses a target skeleton to define required local connections. For each connection, TRACE identifies the easiest valid reconnection under the current prediction and jointly supervises its bottleneck with the corresponding skeleton point. The supervision thus adapts to the evolving prediction rather than remaining fixed to the centerline. We further introduce a topology-aware skeleton carrier that preserves both foreground and class-specific topology, enabling reliable definition of these connections. Extensive experiments across six 2D and 3D benchmarks show that TRACE improves connectivity and topology while maintaining strong segmentation accuracy with low training overhead. Code will be released upon acceptance.
[CV-31] OpenGCUA: Toward a Fully Open Stack for Generalist Computer Use
链接: https://arxiv.org/abs/2610.22200
作者: Hanzhang Zhou,Long Li,Quyu Kong,Lei Wang,Panrong Tong,Long Chen,Tianyu Xia,Gongjie Zhang,Xu Zhang,Chen Liu,Jianan Zhang,Chenglin Cai,Zhengwen Zeng,Steven Hoi,Yue Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Computer use has reached a turning point: the focus is expanding from GUI-only interaction to systems that complete tasks on users’ computers through visual and programmatic interfaces. We define a generalist computer-use agent (GCUA) as a system in which the model chooses how to observe, act, and organize its work across visual and programmatic interfaces. Yet frontier implementations, such as GPT-6 Astra’s computer-use system, remain largely closed, limiting systematic study of their design. We introduce Open Generalist CUA (OpenGCUA), an ongoing project to develop and open the full stack for GCUAs with the community. This report presents the first release of OpenGCUA. Code execution is a key feature of our agent design: it lets intermediate results determine subsequent operations and enables models to apply their existing programming capabilities directly to computer interaction. We also propose harness-to-model decision transfer, moving selected observations and task-state decisions from fixed harness rules to the model. On OSWorld-v2, our full configuration improves task scores across all eight evaluated models, including GPT-6 Astra and Claude Opus 5, raising mean partial score by +19.6 points and reducing reported steps per task by an average of 53.0% relative to batched GUI-only execution. We release our implementation and prompts, evaluation trajectories from all eight models, and preliminary versions of a real-device harness and a cross-device SDK. The harness addresses real-device challenges not fully captured by benchmarks, while the SDK provides a common programming interface across computers, mobiles and browsers. Across four real macOS tasks under our real-device harness, selected Qwen3.8-Max runs use 19.4 times as many reported reasoning tokens as GPT-6 Astra, highlighting substantial differences between models in real-device execution efficiency.
[CV-32] LAMP: Layerwise Adaptive Mixing of Predictors for Training-Free Few-Shot Vision-Language Adaptation
链接: https://arxiv.org/abs/2610.22182
作者: Zhipeng Ye,Feng Jiang,Qiufeng Wang,Hao Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Few-shot adaptation often relies on an encoder’s final global representation, leaving complementary information in intermediate layers underused. Combining these features is challenging because their dimensions can differ and their usefulness varies across tasks. We propose LAMP, Layerwise Adaptive Mixing of Predictors, a training-free framework that combines global and regional information from multiple encoder layers. A single forward pass per view supplies all descriptors. Global and support-calibrated regional kernel predictors independently fit the same label residuals. Labeled validation data select their mixture in class-score space, preserving native feature dimensions without learned projections. LAMP-F further adapts text-anchored class prototypes, using a cached response to update the fixed-kernel residual predictions exactly. Experiments cover eleven datasets and ViT and ResNet backbones. LAMP achieves state-of-the-art mean accuracy among the compared published training-free ResNet-50 methods at 4, 8, and 16 shots, reaching 78.73% at 16 shots. Averaged across five support budgets, LAMP improves over a paired global-feature-only baseline by 1.14 and 1.59 percentage points on ViT-B/16 and ResNet-50; LAMP-F raises these gains to 1.96 and 2.54 points. LAMP-F reaches 16-shot accuracies of 83.57% and 79.10%, respectively. Controlled experiments identify multi-layer integration as the main source of improvement and examine pooling, calibration, validation labels, and computational cost. These findings show that combining layer-specific predictions can exploit complementary features from frozen encoders to improve few-shot classification.
[CV-33] REKI: Registering Knowledge in VLAs
链接: https://arxiv.org/abs/2610.22178
作者: Juliette Marrie,Paul Pacaud,Adrien Ramanana Rahary,Shizhe Chen,Moritz Böhle
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:We introduce REKI, a 2.4B vision-language-action model (VLA) pretrained exclusively on public data that sets a new Pareto frontier between performance and model size: on LIBERO-Plus, RoboTwin Clean2Rand and RoboCasa365, it is outperformed only by models about twice its size. REKI consists of a vision-language model (VLM) followed by a flow-matching action expert, and is pretrained on a curated mixture of public robot, egocentric human and vision-language data. Additionally, we endow it with a memory that gives the policy access to its recent past at negligible cost: a few register tokens in the action expert summarise the output of the VLM at each frame and persist in the expert’s cache. This memory improves success in simulation and on real robots, and makes oscillations between poses about four times rarer. REKI uses a simple model design and training paradigm, and is optimised to be efficient at training and inference. We release the full code across data curation, training, inference and evaluation, deployment, and all pretrained and finetuned model weights: this https URL.
[CV-34] Motion- and Gravity-Aware Feedforward 3D Reconstruction
链接: https://arxiv.org/abs/2610.22177
作者: Shiyao Qian,Yuan Ren,Dongfeng Bai
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Feed-forward transformer models have recently achieved strong performance in multi-view 3D reconstruction, but they still rely primarily on appearance-based attention and often struggle in dynamic scenes with weak geometric cues. In this work, we introduce a motion- and gravity-aware framework under the Learning Using Privileged Information (LUPI) paradigm. During training, residual optical flow and gravity direction are provided as privileged signals to guide geometric representation learning; during inference, the model uses standard visual inputs only. We integrate these priors into transformer backbones with lightweight architectural changes, preserving feed-forward efficiency. Experiments on standard benchmarks show consistent gains in reconstruction accuracy and robustness, especially in scenes with dynamic objects, ambiguous structure, and challenging camera motion. Our analysis further shows that the optimal injection stage is backbone-dependent: middle-layer integration is generally preferable for VGGT, while late-stage integration is more effective for \pi3. These results suggest that LUPI offers a practical way to inject motion and physical priors into feed-forward 3D reconstruction without increasing test-time complexity. Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) Cite as: arXiv:2610.22177 [cs.CV] (or arXiv:2610.22177v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2610.22177 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-35] FALCON: FPGA-Accelerated Latency-Certifiable Occupancy Networks for Camera-Based Autonomous Driving WACV2027
链接: https://arxiv.org/abs/2610.22143
作者: Haohua Que,Mingkai Liu,Handong Yao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to WACV 2027 (Applications track). 10 pages
Abstract:A perception system that must brake for a pedestrian does not need a low average latency. It must deliver an answer before the deadline on every frame, the worst case included. Commodity edge accelerators optimize the mean through dynamic scheduling, OS jitter, thermal throttling, and contention, so their per-frame latency can be measured but not bounded. We build FALCON, a camera-based 3D occupancy front-end on the AMD Kria KR260, an embedded FPGA drawing 4.3 W idle and up to 6.6 W under inference, and make the latency guarantee layered and explicit: programmable-logic (PL) compute is cycle-bounded analytically; inference is a measured envelope over 100k-frame traces; and a watchdog enforces output delivery, publishing the previous output flagged stale when a frame is late. The deadline runs use a fixed-work proxy of six ResNet-50 DPU inferences per frame, not the full occupancy network: across three 100k-frame enforced runs, all 300k proxy frames delivered an output within the 100 ms deadline plus at most 10.2 ms, the worst of the three runs. The full six-camera FlashOcc pipeline composes to about 1.3 s per frame on the board and is not yet a 10 Hz system. The two operations the DPU cannot run, a view-transform gather and a bilinear resize, become fixed-trip-count PL IPs numerically matched to their software references. On the full Occ3D-nuScenes validation set the board reaches mIoU 29.88, 92.8% of an FP32 baseline the DPU cannot run. On a ResNet-50 probe over 100k frames the KR260 tail stays within 1.5 times its median (2.0 times for the six-inference proxy), where Jetson Orin reaches 4.4 times and the worst datacenter GPU run (NVIDIA L4) 23.6 times. The deployment recipe, HLS sources, and raw timing traces are released.
[CV-36] SCCL-DT: Semantic-Conflict-Aware Contrastive Learning with Dual-Rule Teacher Distillation for Zero-Shot Chest X-Ray Diagnosis
链接: https://arxiv.org/abs/2610.22094
作者: Ting Liu,Haoran Lai,HaobinChen,Ming Liu,S Kevin Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 4 figures, includes supplementary material
Abstract:Contrastive vision-language pre-training supports zero-shot chest X-ray diagnosis, but standard objectives treat every off-diagonal image-report pair in a batch as a negative. In radiology, recurring findings across patients mean that instance mismatch need not imply semantic incompatibility, creating false-negative supervision that we term a semantic conflict. A training-trace audit shows that pair-observations selected by the online mask used during training are enriched for report-derived finding agreement, although this proxy analysis does not establish clinical equivalence. Fixed thresholds can become miscalibrated as similarity distributions evolve, while soft reweighting leaves residual repulsive gradients. We therefore propose Semantic-Conflict-Aware Contrastive Learning (SCCL), which tracks modality-specific similarity distributions with exponential-moving-average (EMA) statistics, derives adaptive thresholds, and hard-masks flagged candidates before softmax normalization. Inclusive (OR) and consensus (AND) rules define two SCCL teachers, whose bidirectional cross-modal rankings are consolidated through standard distillation into a single deployable student, SCCL-DT. Across five public chest X-ray benchmarks, SCCL-DT achieves a mean zero-shot AUC of 0.850 , compared with 0.837 for the strongest evaluated baseline in our comparison. Code will be made publicly available soon.
[CV-37] Beyond the Dominant Label: Recovering Coexisting Lesion Evidence with Multiple Instance Learning
链接: https://arxiv.org/abs/2610.22091
作者: Lubin Gan,Jing Zhang,Xinming Zhang,Chengpeng Xu,Siying Wu,Xiaoyan Sun
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 7 pages, 2 figures
Abstract:Whole-slide images may contain coexisting lesions, yet weak supervision typically provides only the most severe diagnosis. We formulate this setting as dominant-label multiple instance learning, where the observed label is the maximum-severity element of a latent lesion set. Standard multiclass MIL treats it as a complete one-hot description and therefore suppresses lower-severity evidence whenever it co-occurs with a dominant lesion. We propose DomSet-MIL, a label-aware framework that aligns the annotation mechanism across bag and instance levels. Dominant-Label Censored Likelihood models the reported diagnosis without declaring unreported lesions absent. Severity-Ordered Residual Routing assigns patch evidence from high to low severity and preserves residual support for secondary findings, while top- q pooling links local routing scores to bag prediction. Experiments across natural and controlled settings show that DomSet-MIL achieves state-of-the-art performance in dominant-label prediction while more effectively recovering latent coexisting-lesion evidence. These findings demonstrate that modeling the dominant-label mechanism can improve diagnostic prediction without sacrificing clinically meaningful secondary findings.
[CV-38] Pathology foundation models for training-free case retrieval in salivary gland tumors
链接: https://arxiv.org/abs/2610.22087
作者: Maximilian J. Roesel,Niklas Kiermeyer,Philipp Keyl,Frederick Klauschen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:In pathology, pattern recognition relies on a frame of reference built through extensive training and experience, which may still fall short when encountering rare cancer subtypes. Archives of previously diagnosed tumors often contain relevant precedents, yet identifying the most relevant cases remains challenging. Here we explore whether content-based retrieval with current pathology foundation models facilitates rare cancer diagnostics without any task-specific training. Pathologists can query a digitized archive using a region of interest (ROI) from a whole slide image (WSI) and retrieve morphologically similar cases, along with their ICD-O codes and corresponding WSIs. The ROI can be selected manually or automatically using a gated attention-based multiple instance learning model. We evaluate our approach by leave-one-case-out on an imbalanced dataset of 137 salivary gland tumor cases spanning 12 ICD-O morphology categories, comparing four pathology foundation models (Atlas2, Atlas2-B, Atlas2-S, and UNI2h). Top-1/3/5 retrieval was 62.5%/76.7%/81.3% for Atlas2, 50.2/66.6/72.9% for UNI2h, 52.3%/71.5%/77.6% for Atlas2-B and 54.4/65.2/73.7% for Atlas2-S. Macro-averaged F1 ranged from 69.0% for the smallest model Atlas2-S to 76.6% for the largest model Atlas2 across the five classes with at least ten cases. Our modified Chamfer similarity separated correct from incorrect matches among the Top-5 more reliably for Atlas2 (AUC 0.754) than for Atlas2-B (0.679, p = 0.020) and Atlas2-S (0.653, p = 0.003), but not UNI2h (0.697, p = 0.196). Requiring no additional training or manual annotation, this approach makes departmental archives searchable with pathology foundation models, providing a low-barrier route to case-based differential diagnosis and extending the individual frame of reference to the collective experience of the department.
[CV-39] Retrofitting Recurrent Vision-Language Models from Pretrained Backbones
链接: https://arxiv.org/abs/2610.22075
作者: Madhav Iyengar,Andreas Geiger,Hilde Kühne
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Recent work shows that depth recurrence can be retrofitted into pretrained language models. In this work, we adapt this paradigm for visual understanding, using depth recurrence to reduce the parameter footprint of a pretrained vision-language model (VLM). To this end, we prune layers from the language backbone of the VLM and loop over a subset of the remaining layers. At each step of the loop, a linear adapter merges the block’s output with the token representations from the preceding layers and projects the fused representation back into the block. As the pruning impacts the original model performance, we propose a three-stage recovery training: first, we leverage a language-only pretraining to recover the performance of the language backbone, followed by an alignment of the vision-language projector, and a full fine-tuning of the model via visual instruction tuning. In the first and the last stage, we distill the outputs of the pretrained non-recurrent model into the recurrent model, which further improves the recovery. We evaluate the proposed approach on the TinyLLaVA-1.4B backbone without introducing any new vision-language data during retrofitting. Our experiments demonstrate that the retrofitted recurrent model outperforms the unpruned baseline trained with the same recovery data while storing 23.3% fewer language parameters. We further show that the proposed retrofitting recurrence outperforms the setup of converting a recurrent LM into a VLM on the same data. Additionally, our ablations show how design choices, such as the adapter between recurrences and a larger recurrent block, lead to the improved performance of our approach. Finally, we find that our proposed model gains on reasoning-oriented benchmarks, whereas its performance on basic visual understanding benchmarks remains on par with the baseline.
[CV-40] Memory as Weights: Internalizing Long-Term History for Streaming Videos
链接: https://arxiv.org/abs/2610.22070
作者: Zhuoming Liu,Jinhong Lin,Lin Zhang,Yiwu Zhong,Yin Li
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Streaming vision-language models (VLMs) must continuously integrate information from the past while remaining responsive to current observations and user queries. Existing memory mechanisms typically compress, retrieve, or cache historical observations and reintroduce them into the input context, forcing historical memory and current observations to compete for limited context. We propose MaLoW, a streaming memory framework that instead internalizes historical context into evolving lightweight parameter updates. MaLoW uses a recurrent HyperNetwork to incrementally consolidate incoming video observations into fixed-size LoRA weights, allowing historical information to be carried forward without revisiting the full video stream. In doing so, MaLoW represents history as weights, while keeping the current observation as input. To support training, we further construct MaLoW-Instruct, a large-scale streaming video dataset containing 381K question-answer pairs over 56K videos. We conduct extensive experiments to demonstrate that MaLoW achieves strong performance on OVO-Bench, StreamingBench, and StreamArena. On the backward split of OVO-Bench, MaLoW outperforms prior memory approaches by 3.9% to 11.8% across different video backbones while incurring only modest latency overhead. Beyond streaming video understanding, MaLoW also generalizes to vision-language-action models, improving the action success rate on Simpler WidowX by 9.3%.
[CV-41] FeaTok: Feature-Loss Tokenization for Image Generation
链接: https://arxiv.org/abs/2610.22060
作者: Zhang Xiaofeng,Samuel Lavoie,Nicolas Beltran-Velez,Jianhao Yuan,Felix Friedrich,Jakob Verbeek,Aaron Courville,Michal Drozdzal,Adriana Romero-Soriano
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Discrete image generators learn the distribution of tokens produced by an image tokenizer. These token representations should be easy for generators to model, not merely enable faithful image reconstruction. Yet these tokenizers are commonly trained to reconstruct pixels, which prioritizes low-level fidelity and may disrupt representation structure useful for generation. Motivated by feature-predictive self-supervised learning, we introduce feature-loss tokenization (FeaTok), which trains discrete tokens by reconstructing pretrained features rather than pixels. Ablations across SigLIP, SigLIP2, DINOv2, and DINOv3 show better generative performance with FeaTok. In class-conditional generation, FeaTok advances the quality-compute Pareto frontier, reducing gFDr6 at 256x256 by up to 29 percent over the strongest discrete baselines and outperforming the strongest unguided continuous baseline (RAEv2) with 38 times less sampling compute. FeaTok also achieves higher human-preference reward-model scores than prior art without reward-based training and surpasses continuous and discrete text-conditional baselines on three benchmarks with only 500 fine-tuning updates after pre-training.
[CV-42] When Models Misread Ink: Auditing VLMs for Arabic Calligraphy Recognition NEURIPS2026
链接: https://arxiv.org/abs/2610.22018
作者: Niaz Mohaiman Abtahi,Evan Maksud,Reyana Shahid
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages, including references and appendix. Accepted for poster presentation at the Muslims in Machine Learning (MusIML) Workshop at NeurIPS 2026. Code: this https URL
Abstract:Arabic calligraphy is a high-stakes cultural visual domain for Muslim communities, yet VLM evaluations rarely test its fine-grained script traditions. We audit 12 VLMs on a cleaned 1,195-image six-way benchmark using numeric, visual-cue, and Diwani-first prompts. In this closed-set setting, the best numeric-prompt accuracy is 60.9% for closed/API models and 47.2% for open-weight models. The main failure is default-label collapse: every numeric run is dominated by a single label taking 33-63% of predictions, against an 18.8% majority-class baseline. For 8 of 12 models the default is Thuluth; GPT-4o labels 52.9% of all images as Thuluth although it is only 18.1% of the set. Prompting helps unevenly: visual cues move accuracy by more than 10 points for some models but hurt others, and foregrounding Diwani often relocates the default rather than removing it. Arabic calligraphy recognition is therefore a compact, culturally grounded stress test for VLM robustness and for measuring which visual traditions become legible to automated systems.
[CV-43] Visual Sanad: Auditing Vision-Language Models for False Acceptance of Quranic Citations NEURIPS2026
链接: https://arxiv.org/abs/2610.22009
作者: Niaz Mohaiman Abtahi,Reyana Shahid,Evan Maksud
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages, including references and appendix. Accepted for poster presentation at the Muslims in Machine Learning (MusIML) Workshop at NeurIPS 2026. Code: this https URL
Abstract:Vision-language models (VLMs) are increasingly used to interpret religious, educational, and culturally sensitive visual content, yet their reliability on citation verification remains underexplored. We introduce Visual Sanad, a controlled benchmark for testing whether VLMs falsely accept incorrect Qur’anic verse attributions when the same quote-reference pair is rendered under different visual framings. The benchmark contains 2,400 rendered images built from 100 English Qur’an verses, four factual states per verse, and six visual conditions ranging from plain text to decorative and authority-framed layouts. All quote-reference pairs were source-checked and human-audited before rendering. Across 24 full model-prompt runs, we observe a sharp capability split. Proprietary API-served VLMs range from near-perfect citation verification to moderate false acceptance, whereas recent open-weight VLMs often show weak discrimination and default to VALID on most images. De-biased prompting can reduce false acceptance, but often by shifting the decision threshold toward INVALID rather than by producing strong citation discrimination. Visual Sanad therefore exposes a culturally grounded multimodal factuality failure: fluent VLMs may treat displayed religious citations as self-authenticating unless they have strong citation-verification ability.
[CV-44] Data-Efficient ConvNeXt Pretraining for High-Resolution Sentinel-1 GRD SAR Representation Learning
链接: https://arxiv.org/abs/2610.21993
作者: Yuki Natsume,Sang-Ho Yun,Chai Kiat Yeo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 30 pages, 9 figures. Extends a preliminary paper to be presented at SPIE Asia Pacific Remote Sensing 2026
Abstract:Labeled Sentinel-1 Ground Range Detected (S1GRD) datasets remain scarce, limiting supervised deep learning for land-surface remote sensing, while self-supervised pretraining for this domain has been dominated by Vision Transformer (ViT)-based masked image modeling requiring substantial data and compute. We propose SparK-SAR, a sparse convolutional masked pretraining framework built on SparK and ConvNeXt, designed to exploit convolutional inductive biases for S1GRD imagery under constrained compute. SparK-SAR makes three core technical contributions: (i) ConvNeXt-RS, a domain-adapted encoder that halves the stem stride (4x to 2x) to produce 14x14 feature maps at stride 16x, providing 4x more spatial tokens at identical parameter count; (ii) a lightweight ConvNeXt-512 reconstruction decoder with PixelShuffle upsampling that reduces pretraining VRAM by 7.5-8.5% per GPU relative to a UNet-128 baseline while matching its accuracy; and (iii) an empirical characterization of FP16 mixed-precision pretraining instability in the standard SparK UNet-768 decoder on S1GRD data, complemented by a finding that finetuning with layer-wise learning rate decay (LLRD) improves downstream performance of pretrained encoders. On four Copernicus-Bench S1GRD tasks, our finetuned 27.8M-parameter ConvNeXt-RS encoder improves over a non-pretrained baseline by +4.7, +5.0, +5.8, and +1.4 pp on EuroSAT-S1, BigEarthNet-S1, DFC2020-S1, and Flood-S1 respectively, and outperforms the Copernicus-FM ViT (139M parameters, frozen) on three of four tasks at roughly 5x fewer parameters. A pretraining data-scaling study reveals task-dependent data efficiency: classification saturates with as little as 10-25% of the SSL4EO-S12 v1.1 corpus, while dense prediction continues to improve significantly up to 25-50%. The study also includes a CKA representation analysis and a geographic de-collision experiment.
[CV-45] VesperMUSE: Multi-Modal Fusion Vision Transformer for Image Semantic Segmentation Under Diverse Light Conditions
链接: https://arxiv.org/abs/2610.21990
作者: Raul Balmez
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: A dissertation submitted to The University of Manchester for the degree of Bachelor of Science in Computer Science
Abstract:Semantic segmentation is critical for the reliable navigation and scene understanding of autonomous systems, yet its performance degrades severely in low-light environments due to noise, artifacts, colour distortion, and the loss of structural details. To address this challenge, this work introduces VesperMUSE, a unified multi-modal framework designed to achieve robust image semantic segmentation across diverse illumination conditions. The architecture addresses the light deficiency problem on multiple levels: proposing a novel physics-informed synthetic low-light adaptation algorithm, approaching Low-Light Image Enhancement (LLIE) through a thermal-guided paradigm, and finally aiming to recover the intrinsic scene details for semantic segmentation through a multi-modal fusion approach. Following enhancement, the framework dynamically integrates the primary RGB features with auxiliary thermal representations and surface normal maps highlighting geometric discontinuities and material boundaries. At the core of this integration is a highly scalable, modality-agnostic self-attention mechanism that extracts both global context and fine local features from an arbitrary number of auxiliary inputs. By combining and projecting these hierarchical multi-modal representations into a unified component, the attention module allows the network to selectively attend to relevant cross-modal cues without requiring any architectural modifications when different sensor configurations are used. Extensive evaluations demonstrate that VesperMUSE successfully recovers missing visual cues, providing a flexible, state-of-the-art (SOTA) solution for machine perception in extreme and unstructured real-world environments.
[CV-46] GRID-Planner: Graph-Structured Relational Reward Shaping for Diffusion Planning
链接: https://arxiv.org/abs/2610.21984
作者: Zhao Yang,Zezhong Qian,Yinan Shi,Longjun Liu,Yawei Jueluo
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Autonomous driving planning requires selecting trajectories that remain safe, rule-compliant, and comfortable under complex multi-agent interactions, yet robust closed-loop decision making in planning-critical scenes remains challenging. Existing diffusion- and RL-based planners often rely on generic rewards and use relational structure mainly for representation, limiting how planning-critical rollouts shape policy improvement. We present GRID-Planner, a graph–reward integrated diffusion planning framework that aligns generative planning with structured relational signals. Specifically, GRID-Planner combines graph-aware importance weighting, a time-indexed relational reward surrogate, and an advantage-weighted diffusion objective with on-policy self-distillation, forming a rollout-return-advantage-update loop without explicit likelihoods. Experiments on nuPlan and interPlan show that, compared with the diffusion-only baseline, GRID-Planner reduces collision rate, lane deviation, and jerk by 35.8%, 60.8%, and 27.4%, respectively, and improves the planning score by 3.4%. Project website: this https URL.
[CV-47] Flow-CLIP: Image-Conditioned Residual Scoring for CompositionalImage-Text Matching
链接: https://arxiv.org/abs/2610.21937
作者: Minghong Wei,Enguang Zhu,Pu Cao,Lu Yang,Qing Song
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision–language pre-training has enabled substantial progress in image–text retrieval and zero-shot recognition by learning transferable cross-modal representations. However, sentence-level alignment remains insufficient for compositional matching, where captions can describe the same objects and attributes while differing in how these elements are bound. To address this limitation, we propose Flow-CLIP, which refines global image-text alignment through image-conditioned changes in token–patch evidence. Global Compatibility Calibration (GCC) learns a sentence-level similarity metric, while Image-Conditioned Evidence Refinement (ICER) incorporates visual patch context into token queries and uses the resulting change in local matching evidence to adjust the global score. We further introduce Cross-Image Margin Ranking (CIMR), which compares the same positive–negative caption pair across matched and mismatched images, encouraging a larger ranking margin when the image supports the correct description. Experiments on compositional matching and attribute–object binding benchmarks show that Flow-CLIP improves binding accuracy over pretrained CLIP and learned global alignment through local evidence refinement, while cross-image ranking supervision mitigates reliance on language priors. Code will be released at link.
[CV-48] BridgeAfford: Bridging Query-Target Gap for Generalizable 3D Affordance Grounding
链接: https://arxiv.org/abs/2610.21918
作者: Yijie Su,Yakun Zhu,Yi Bin,Pengpeng Zeng,Liang Chen,Guoqing Wang,Jingkuan Song,Heng Tao Shen
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Language-guided 3D affordance grounding aims to localize functional regions on 3D objects according to interaction intents. However, existing methods exhibit poor generalization to unseen objects. One challenge is the query-target gap: a semantic discrepancy between query features encoding interaction intents and target-region features, leading to inaccurate grounding. End-to-end training favors query representations tailored to seen objects, limiting their compatibility with shifted target-region features on unseen objects. To address this issue, we propose BridgeAfford, a framework that reduces the query-target semantic gap from two complementary perspectives. First, Interaction Region Alignment (IRA) promotes cross-object alignment among geometrically similar interaction regions during training. This enables structures on unseen objects that resemble previously observed ones to produce representations similar to those learned during training, facilitating target-region recognition. Second, Visual Evidence Aggregation (VEA) leverages a 2D affordance model to refine the attention distribution and guide the query toward the correct region, yielding more target-aware queries. Extensive experiments on the LASO and PIAD datasets demonstrate that BridgeAfford outperforms existing methods, particularly on unseen objects, highlighting its robust generalization ability.
[CV-49] Seeing is Believing: Benchmarking and Mitigating the Interference of Language Priors on Visual Perception in MLLM s
链接: https://arxiv.org/abs/2610.21902
作者: Li Ma,Yan Zhuang,Jiawen Deng,Fuji Ren
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Language priors in MLLMs can override visual evidence, impairing visual perception. Existing benchmarks assume shared priors across models and rely solely on accuracy. However, cross-model prior agreement falls below 80%, and over 15% of items can be answered correctly without any image, causing accuracy to overestimate prior resistance. To address these limitations, we introduce ViSiP-Bench, a benchmark grounded in rare-but-real scenarios across six prior categories, with a model-aware dynamic prior induction protocol and three metrics that more precisely and continuously measure prior resistance than accuracy alone. Evaluating 24 models reveals that prior adherence is pervasive, does not improve with scale, and larger models tend to either strongly override their priors or firmly adhere to them. We therefore propose MAPS, which identifies each model’s strongest priors and fine-tunes on factually grounded knowledge to soften excessive prior reliance without disrupting general capabilities.
[CV-50] SurgStreamer: proactive request-driven monitoring of surgical video streams with structured state and a reusable reasoning layer
链接: https://arxiv.org/abs/2610.21895
作者: Bowen Xiang,Jon S. Heiselman,Michael I. Miga
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Surgical teams could benefit from assistants that watch the video for requested events, such as “tell me when the clip applier appears during Calot triangle dissection”, and answer questions about what has already happened. We formalise this task as request-driven event detection with exact semantics and causal alert decisions, and present SurgStreamer. It separates procedure-specific perception, which maps the 1-FPS stream to a structured state over tools, actions and phases, from procedure-agnostic reasoning, which evaluates a formal monitoring condition over this state and an event memory, either with exact rules or with a small controller that addresses entities through pointers rather than names. We evaluate this monitoring core with gold conditions. On a pre-registered, sealed test set of 12 laparoscopic cholecystectomies, SurgStreamer reached a macro event-F1 of 0.598 over full procedures. On identical 75-s windows with only the window available to every method, it scored 0.677. The tested configurations of the streaming vision-language model OneStreamer-4B, including ones fine-tuned on SurgStreamer’s state or on video with an adapted vision encoder, scored at most 0.540 (Holm-adjusted p = 0.012). At three external centres, re-training only perception raised performance on held-out procedures from 0.239 to 0.470. On a new procedure, pituitary surgery, controller weights trained on cholecystectomy matched an in-domain controller on oracle state (0.947 vs 0.938). On predicted state they missed the pre-registered non-inferiority margin (-0.024 [-0.040, -0.008]) unless the alert thresholds were re-calibrated. The results support separating perception from reasoning and point to perception under distribution shift as the main limitation.
[CV-51] Anvil: Scaling Embodied Spatial Intelligence through Knowledge Recirculation
链接: https://arxiv.org/abs/2610.21882
作者: Weilin Ruan,Bohan Zhou,Jinghao Li,Yangkai Wei,Wenqian Li,Peiran Dong,Yingjie Xu,Leyi Wu,Haojian Huang,Binghui Xie,Han Yang,Kaiwen Zhou,Yinchuan Li,Ying-Cong Chen,James Cheng
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 28 pages, 7 figures
Abstract:Embodied spatial intelligence requires more than visual recognition: models must localize entities, reason about their relations, and express this knowledge in action-relevant forms. Yet joint training can conflate the knowledge needed at different stages because supervision is fragmented across incompatible geometric and semantic outputs. We introduce Anvil, a unified framework for data generation, training, and benchmarking for embodied spatial intelligence. (1) Anvil-Gen is a heterogeneous data flywheel that combines form-specific generation, cross-stage consistency verification, and failure-driven recollection across 11 Ground–Relation–Act forms. (2) Anvil-Ben provides scene-aligned supervision for learning diverse embodied spatial capabilities and scene-linked assessments that distinguish stage competence from error propagation. (3) Anvil-MOPD is a 9B student that consolidates heterogeneous spatial knowledge through distillation from specialist teachers. Semantic-first-fault routing targets teacher supervision to the student’s earliest decisive failure, while stage normalization controls the loss scale across native outputs of different lengths. We construct 5.8M source-grounded records spanning 160K images, reserving 580K spatial records for held-out evaluation and using the remaining 5.22M records for training. On average across seven public benchmarks, Anvil-MOPD outperforms Qwen3.5-9B and RynnBrain-8B by 9.29 and 11.03 percentage points, respectively, while reaching 80.67% on VABench-Point. Controlled studies support targeted coverage, selective transfer, and feedback-guided repair as complementary framework components.
[CV-52] Seeing Both the Forest and the Trees: Reusing Holistic 3D Priors for Part-Decomposed Generation NEURIPS2026
链接: https://arxiv.org/abs/2610.21850
作者: Yiqun Zhao,Binbin Huang,Haobin Duan,Zibo Zhao,Shenghua Gao
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: NeurIPS 2026, project page: this https URL
Abstract:Generating 3D assets with explicit parts requires two properties at once: geometric detail inside each part and structural coherence across the whole object. Recent image-conditioned 3D generation models adopt a structured 3D representation that anchors tokens to explicit spatial positions, capturing both properties, but their output remains a single fused mesh. We ask how a part-decomposed generation model can inherit both properties of these holistic priors, and answer with STRUCTPARTS, a framework whose core technical contribution is a dual-frame coordinate interleaving mechanism: each shape latent is interpreted under both a part-local and an object-global reference frame, and processed through a global-local inter-leaved transformer that simultaneously reuses the prior’s geometric detail at the part scale and its structural coherence at the whole-object scale. As its structural input, STRUCTPARTS uses a segmented mesh scaffold, a coarse mesh paired with a face-level segmentation mask kept entirely within the prior’s native 3D representation. With only lightweight finetuning, both intra-part detail and inter-part coherence emerge from the same pretrained weights. On PartObjaverse-Tiny, STRUCTPARTS matches or surpasses prior part-decomposed models at a fraction of their training cost, while naturally supporting part-mesh refinement and part-level editing.
[CV-53] mporal Periodic Image Registration with Implicit Neural Representations MICCAI2025
链接: https://arxiv.org/abs/2610.21845
作者: Mathias Lowes,Kristine Aavild Sørensen,Maxime Sermesant,Rasmus R. Paulsen
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Published at MLMI@MICCAI 2025
Abstract:Implicit Neural Representations (INRs) have recently gained popularity for their ability to model functions using simple and lightweight networks. These methods have also proven effective in deformable image registration. In this work, we extend INRs for deformable image registration to the domain of temporal image registration, focusing on periodic temporal image sequences. Our approach, Temporal-IDIR, optimizes a single INR to model deformations across all frames in a temporal image sequence simultaneously, allowing for self-regularization through its own deformation predictions. To achieve this, we introduce a temporal consistency loss that penalizes discrepancies between direct source-to-target transformations and those traversing intermediate frames. We evaluate our framework on the DIR-LAB dataset, using the target registration error (TRE) between annotated and moved landmarks as the metric. Here, we achieve a TRE of 1.03 mm, outperforming other INR-based registration methods. Additionally, our framework supports smooth interpolation between time frames by estimating deformations between the given input frames. Code is publicly available at this https URL.
[CV-54] From Adults to Children: Benchmarking Foundation Models for Chest X-Ray Classification under Domain Shift WACV2027
链接: https://arxiv.org/abs/2610.21809
作者: Sara Inácio,Hugo Proença,João C. Neves
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at WACV 2027
Abstract:Foundation models have shown significant potential across diverse fields and have been applied to medical imaging, but their robustness across patient populations remains unclear. In this work, we investigate foundation models’ ability to generalize from adult to pediatric chest X-ray classification. To the best of our knowledge, this is the first benchmark to compare multiple foundation models under a common adult-to-pediatric evaluation setting while also analyzing the effect of adaptation depth. We harmonize VinDr-CXR and PediCXR into a five-label classification task and evaluate six foundation models and two CNN baselines. All models are adapted using adult data only and tested in both the in-domain and pediatric domains using linear probing, partial fine-tuning, and full fine-tuning. The benchmark reveals a clear and consistent trend: all models suffer a substantial performance degradation when transferred from adult to pediatric data, with drops exceeding 10 percentage points. While Macro-AUROC values reach approximately 92% on the adult test set, performance decreases to around 78% on the pediatric domain. Another important finding is that increasing the depth of adaptation does not significantly improve cross-population generalization. Moreover, strong in-domain performance does not translate into better performance on pediatric data.
[CV-55] aching DINOv3 About Partial 3D Geometry: A Self-Supervised Geometry-Aware Approach
链接: https://arxiv.org/abs/2610.21802
作者: Viktoria Ehm,Dongliang Cao,Riccardo Marin,Daniel Scholz,Weikang Wang,Florian Bernard,Daniel Cremers
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Partial shape matching is a crucial yet underexplored problem in 3D vision, with significant relevance to real-world scenarios where shapes are often only partially observed. Existing feature descriptors face difficulties in this setting, as traditional representations either struggle with the boundaries of partial shapes or heavily depend on the shape’s spatial position. While existing approaches have employed DINO features for partial shape matching, these features are not inherently suited for handling partial observations. In this work, we propose a method to refine DINO features using LoRA-based self-supervised learning, enabling the generation of feature descriptors that are robust to partiality. Our features substantially improve performance on partial shape matching compared to traditional or vision foundation features. Additionally, when integrated into existing partial shape matching pipelines, we achieve state-of-the-art results on partial shape matching and left-right prediction benchmarks.
[CV-56] Single-stage joint diffusion of sky images and clear-sky index for ultra-short-term irradiance forecasting: what the frame-to-irradiance stage discards
链接: https://arxiv.org/abs/2610.21795
作者: Kanawut Suwandee,Suwichaya Suwanwimolkul
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Systems and Control (eess.SY)
备注:
Abstract:Sky-image irradiance forecasting is dominated by a two-stage design: a video-prediction model forecasts future sky frames, and a separately trained regressor maps those frames to irradiance. We propose a single-stage alternative that co-denoises the future sky-image and clear-sky-index sequences in one diffusion process, with an irradiance head that reads a denoised preview of the future sky rather than the noised diffusion state. The forecaster needs no regression stage and runs from sky images and deterministic covariates alone, with no measured irradiance at its input — the situation at a site with a camera but no pyranometer. On 2023 SIRTA data the model reaches a pooled RMSE of 104.1 ~W,m ^-2 and a skill score of +0.34 on the continuous ranked probability score (CRPS) against smart persistence, higher than any of twelve baseline configurations, and retains that skill when retrained on 2021–2022. On the model’s own generated frames, a separately trained regressor reaches 150.6 ~W,m ^-2 against the joint head’s 104.1 : roughly 46 ~W,m ^-2 that joint training recovers and a cascade discards. Point accuracy is competitive but not best: a deterministic network that also reads measured irradiance attains 97.6 ~W,m ^-2 . The model’s CRPS skill leads that network’s in every sky-condition tertile, by 0.26 to 0.30 .
[CV-57] MedPAO-VL: Interpretable and Protocol-Compliant Agent for Structured Radiology Report Generation ECCV
链接: https://arxiv.org/abs/2610.21794
作者: Shrish Shrinath Vaidya,Gowthamaan Palani,Minmini Selvam,Gokulraja Srinivasaraja,Ganapathy Krishnamurthi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: The paper has been accepted at ECCV BISCUIT Workshop, 2026
Abstract:Despite the remarkable generalization capabilities of Large Language Models, generating clinically accurate and structured radiology reports from medical images remains a significant challenge. Current end-to-end approaches function as black boxes, prone to factual hallucinations and generating ungrounded findings. This lack of reliability and failure to adhere to domain-specific structuring protocols, fundamentally limits their clinical utility. We present MedPAO-VL, a multimodal agent that shifts the paradigm from monolithic generation to explainable, iterative, protocol-compliant structured report generation. Unlike deterministic models, our plan-act-observe loop functions as a goal-oriented agent: it first perceives the Chest X-Ray through an explainable vision module designed to capture inherent diagnostic uncertainty, then iteratively refines the generated report by observing semantic discrepancies and self-correcting its generation for better alignment with visually detected clinical findings. Evaluation of our framework demonstrates that this agentic reasoning improves clinical correctness and completeness, achieving an expert-rated accuracy score of 3.5/5 and a structural score of 4.3/5. By prioritizing pathology specific explainability and iterative refinement over opaque end-to-end mapping, MedPAO-VL offers a transparent and clinically viable alternative for automated structured radiology report generation.
[CV-58] NICE Toolbox: Nonverbal Interpersonal Communication Exploration Toolbox
链接: https://arxiv.org/abs/2610.21786
作者: Aleksandr Evgrashin,Carolin Schmitt,Timo Lübbing,Ashutosh Jha,Buket Naz Zeren,Sophie K. Bauer,Johannes Kopf-Beck,Anton K. G. Marx,Anamaria Semm,Gökce Ergün,Senya Polikovsky
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 2 figures. Code: this https URL . Submitted to the Journal of Open Source Software (JOSS)
Abstract:Understanding human communication is a long-standing goal in psychological research. People express themselves through many verbal and nonverbal channels: posture, gesture, gaze, facial expression, and speech. Measurement and annotation of these signals has traditionally required time-consuming manual work by trained human labellers. The Nonverbal Interpersonal Communication Exploration (NICE) Toolbox is a free, open-source Python framework that automatically measures these signals by combining existing state-of-the-art computer vision and speech-processing methods into a single software package. It includes support for multi-camera systems, interchangeable detection algorithms, 2D and 3D visualization of detector outputs, evaluation on the user’s custom data, and export to third-party software. With the release of this tool, we aim to make data exploration of interpersonal communication signals scalable, reproducible, and accessible for psychological research.
[CV-59] Learning to Think and Explore in 3D World
链接: https://arxiv.org/abs/2610.21783
作者: Yijiang Li,Bingyang Wang,Nuno Vasconcelos
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We study how vision-language models (VLMs) can learn to think and actively explore in 3D space. Instead of providing multi-view images, videos, or 3D representations to the model, we preserve the native visual encoding of VLMs, i.e., a single image, one at a time, just like humans, and treat the 3D representation only as an environment with which the model can interact through actions (e.g., move left) and observations (rendered images). Each rollout begins from an initial first-person view and a visual question. The model can then iteratively reason, navigate, and reorient its viewpoint before committing to a final answer. Following an initial trajectory-supervised finetuning, we apply reinforcement learning (RL) with rewards driven primarily by final-answer correctness. Crucially, RL provides no direct supervision over which viewpoints the model should visit or which exploration actions it should take. Instead, the model must learn an effective exploration policy solely from the extent to which its interactions improve performance. Across six 3D benchmarks, our model ACT3D achieves an average accuracy of 53.2% that improves its base Qwen model by 22.0% and surpasses proprietary Gemini 3.6 Flash. ACT3D also outperforms existing open-source 3D methods while using 50-70% fewer input frames, and generalizes to unseen domains with an average improvement of 5.9% over its supervised initialization. Further analysis reveals adaptive, question-conditioned exploration. Performance gains over the base model increase from 0.6% on questions answerable from the initial view to 16.3% on those requiring extensive exploration. ACT3D searches for occluded targets, avoids unnecessary exploration, and acquires complementary viewpoints for spatial comparisons. These findings demonstrate that outcome-based RL enables VLMs not only to reason over visual evidence, but also to learn when and how to acquire it.
[CV-60] DINO-FIL: Multi-Layer Transformer Fusion for Generative Inpainting Forgery Localization
链接: https://arxiv.org/abs/2610.21780
作者: Simon Freitter,David Fischinger,Silvia Poletti,Marcel Hasenbalg,Martin Boyer
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:We propose DINO-FIL, a transformer-based framework for pixel-level localization of generative image forgeries. The method leverages multi-layer representations from a pretrained DINOv3 ViT-H+/16 encoder combined with a lightweight fusion decoder to capture both semantic structure and fine-grained texture inconsistencies introduced by image editing. By aggregating intermediate transformer features across depths, the model improves sensitivity to subtle, spatially localized artifacts that are often missed by standard single-layer approaches. We evaluate DINO-FIL in a cross-dataset setting, training on B-Free and testing on Beyond the Brush, CocoGlide, TGIF, and SAGI-D. The model achieves strong generalization, outperforming existing baselines on average and on four of five evaluated datasets, while maintaining a high true negative rate on pristine images.
[CV-61] FuseAdapt: Parameter-Efficient Multimodal Fusion for Semantic Segmentation with Missing Modalities NEURIPS2026
链接: https://arxiv.org/abs/2610.21779
作者: Xin Zhang,Robby T. Tan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 25 pages, 6 figures, 19 tables. Accepted to NeurIPS 2026
Abstract:Multi-modal semantic segmentation benefits from complementary sensors, but this benefit is fragile when some modalities are degraded or missing at test time. Many existing methods rely on dedicated fusion modules or multi-branch cross-modal encoders, and the fusion often becomes unreliable when modalities are absent, leading to substantial performance drops. We propose \textbfFuseAdapt, a parameter-efficient framework that performs fusion within the adaptation space of a frozen vision foundation model (VFM), exploring whether VFM priors can serve as an effective shared backbone for multi-modal segmentation. Instead of attaching a standalone fusion network, FuseAdapt assigns each modality a lightweight adaptation branch and performs \emphadaptation-space fusion by keeping Q/V low-rank adaptation private to each modality and routing only FFN low-rank residual activations across modalities inside the frozen backbone. The pretrained VFM remains fixed while modality-specific corrections and explicit cross-modal interaction are learned through the adaptation paths. To handle missing modalities, we further introduce \textbfPredictive Modality Modeling (PMM), a direct representation-level objective for cross-modal redundancy. During training, PMM masks a target modality and predicts its embedding from the visible ones, supervised by a teacher target computed under full-modality input. PMM is used only during training and adds no inference-time overhead. Across three multi-modal segmentation benchmarks (MCubeS, DELIVER, MUSES), FuseAdapt consistently improves both modality-complete and modality-incomplete segmentation with a small trainable budget, yielding up to +3.01 mIoU under full modalities and +12.56 mIoU under missing modalities over prior state-of-the-art. Project page: this https URL.
[CV-62] Beyond Generation: Using Multimodal LLM s as Detectors of LLM -Generated Deepfake Facial Images
链接: https://arxiv.org/abs/2610.21760
作者: Olivia Salamon,Virginia N. L. Franqueira,Shujun Li
类目: Computer Vision and Pattern Recognition (cs.CV); Emerging Technologies (cs.ET)
备注:
Abstract:As image generation models become increasingly capable of producing highly realistic content, traditional forensic techniques are struggling to keep up. This paper investigates whether MLLMs can instead be used as a form of forensic analysis by identifying synthetic media. Specifically, the study aims to determine whether these models detect fake images by recognising physical inconsistencies, or whether they rely on digital artefacts such as embedded this http URL MLLMs: GPT-4o, GPT-5, GPT-5.2, Gemini 2.5, Gemini 3, and Seed 1.6, are benchmarked using a dataset of 400 images. Fifty images are authentic photographs, while the remaining 350 are generated using seven leading image models: Gemini 3, Gemini 2.5, Flux.2 Pro, GPT-Image-1, GPT-Image-1.5, Seedream 4.0, and Seedream 4.5. The evaluation is structured in three stages: single-image classification, pairwise comparison, and a focused “self-detection” experiment on Gemini. Several MLLMs achieved a high detection accuracy, with GPT-5.2 reaching a peak of 97.9% in the pairwise experiment, against specific generators and an overall mean pairwise accuracy of 82.5% by identifying structural or physical inconsistencies within images. However, results from the self-detection experiment indicate that the presence of invisible SynthID watermarks leads to a 10% reduction in Gemini’s detection accuracy compared to the clean control group. One possible explanation is that SynthID embedding reduces some of the visible artefacts the models rely on for this http URL results imply that MLLMs can detect structural artefacts without relying on watermarks, but performance varies substantially across model generations. The findings from comparison C5 also suggest that SynthID embedding may suppress some of the structural artefacts used during detection.
[CV-63] Beyond Source Prompts: Towards Visually Grounded AI-Generated Video Quality Assessment
链接: https://arxiv.org/abs/2610.21758
作者: Kehao Zhou,Yixuan Li,Yuxuan Jiang,Fan Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent advances in generative artificial intelligence have accelerated video generation technologies. However, AI-generated videos (AIGVs) remain susceptible to various distortions, highlighting the need for dedicated video quality assessment methods. Content semantics play an important role in AIGV quality assessment, with source prompts typically used as textual references. However, these prompts may not accurately reflect the generated video content, resulting in a discrepancy between the intended textual description and the generated result. To address this limitation, this paper proposes VGCap, an AIGV quality assessment framework guided by visually grounded captions that align better with generated content. Specifically, we introduce two forms of captions: a global caption obtained by minimally revising the source prompt based on multi-frame visual evidence, and prompt-independent segment captions that describe localized temporal content. Together, these captions provide complementary textual evidence for assessing spatial quality, temporal quality, and text-video alignment. Experimental results on the LGVQ and FETV datasets show that VGCap outperforms representative baselines across all three dimensions. Further ablation studies on caption sources and temporal components confirm the effectiveness of visually grounded captions for each quality dimension, while highlighting the benefits of segment-level features for temporal quality prediction. The source of VGCap will be available at this https URL.
[CV-64] he Dormant Spiking Neuron: A State-Driven Mechanism for Efficient Spiking Neural Networks NEURIPS2026
链接: https://arxiv.org/abs/2610.21748
作者: Ruichen Ma,Yicai Chen,Pujun Zhou,Yue Zuo,Chongyang Li,Shuang Liu,Shaogang Hu,Guanchao Qiao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: NeurIPS 2026 Accepted
Abstract:Spiking neural networks (SNNs) offer a compelling pathway toward energy-efficient artificial intelligence, fundamentally anchored in presynaptic input-driven sparsity. However, a substantial yet chronically overlooked source of computational overhead persists: the continuous subthreshold dynamics of postsynaptic neurons. Quantitative analysis reveals that up to 44% of neurons in an ImageNet-trained ResNet-18 model reside in a deeply inhibited state, imposing a persistent computational burden despite having a negligible probability of firing. To address this inefficiency, the Dormant Spiking Neuron (DSN) model is introduced. By incorporating a state-driven dynamic gate, the DSN allows a neuron to enter a dormant state when its membrane potential falls below a predefined negative threshold, dynamically pruning redundant computations at the instance level. Crucially, this dormancy mechanism is directly integrated into the neuronal dynamics and trained end-to-end via backpropagation through time using surrogate gradients, enabling the network to actively learn a robust dormancy policy that adaptively preserves critical information flow. Extensive experiments across static and neuromorphic datasets demonstrate that the DSN framework consistently reduces computational energy by 40% to over 70%. Notably, on the large-scale ImageNet dataset, the DSN outperforms state-of-the-art static pruning techniques at comparable reported cost reductions and simultaneously improves the corresponding baseline SNN accuracy by 1.34%. By establishing a trainable, state-driven sparsity paradigm orthogonal to conventional input-driven methods, the DSN unlocks synergistic efficiency gains, presenting an effective approach for advancing neuromorphic computing.
[CV-65] International Multi-Centre Validation of Unsupervised Domain Adaptation for Precise Discrimination between Normal and Abnormal Retinal Development
链接: https://arxiv.org/abs/2610.21719
作者: Zhanhan Tu,Nikhil Reddy,Helen J. Kuht,Ha-Jun Yoon,Gail D. E. Maconachie,Ravi Purohit,Prateek Pani,Line Kessel,Jinu Han,Jing Jin,Shui-Hua Wang,Karthik Bharadwaj,Viral Sheth,Michael Hisaund,Rebecca J. McLean,George Garratt,Chetan Kantibhai Patel,Kanmin Xue,Callum Hunt,Zisen Zhou,Ian De Silva,Seema Teli,Sonal Shah,Vanessa Rodwell,Aarti Patel,Irene Gottlob,Yu-Dong Zhang,Zedong Zheng,Girish Varma,Mervyn G. Thomas
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Submitted to Nature Medicine. 38 pages, 6 figures, 2 tables. Supplementary material includes 3 figures and 3 tables
Abstract:Arrested retinal development can cause childhood visual impairment, but grading optical coherence tomography scans is difficult to standardise across devices and centres. We developed and evaluated an unsupervised domain adaptation model to classify normal and abnormal retinal development across different OCT systems. This retrospective multicentre study used OCT scans from five centres in four countries and the UK Biobank. Data covered four OCT devices, including table mounted and handheld systems. A ResNet50 based Domain Adversarial Neural Network was trained using labelled images from a source device and unlabelled images from a target device. We evaluated six source to target pairings and compared model performance with seven clinician graders. A separate Heidelberg Spectralis dataset was reserved for external evaluation. Across validation sets, accuracy ranged from 87.75% to 96.06%, sensitivity from 88.68% to 98.38%, and specificity from 78.00% to 98.02%. The best performing model, adapted from TM OCT1 to HH OCT1, achieved 96.06% accuracy and an area under the receiver operating characteristic curve of 0.994, with a 95% confidence interval from 0.981 to 0.999. Several models also maintained discrimination on the unseen Heidelberg device. Unsupervised domain adaptation enabled OCT classification across devices without labelled target device data. The approach may support more consistent assessment of retinal development and large scale analysis of existing imaging datasets.
[CV-66] Mixture of Multi-Sequence Feature Difference for Contrast-Enhanced MRI Synthesis ACCV2026
链接: https://arxiv.org/abs/2610.21703
作者: Jonghun Kim,Junmo Kwon,Hyunjin Park
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ACCV 2026
Abstract:Contrast-enhanced magnetic resonance imaging (CE-MRI) is crucial for accurately diagnosing and segmenting brain tumors. However, the acquisition of CE-MRI requires injecting contrast agents, which poses potential health risks and increases scan costs. Here, we propose a CE-MRI synthesis method from non-contrast sequences, reducing both risk and cost. Unlike existing methods that rely on channel-wise concatenation to fuse multi-sequence information for synthesis, we introduce a Mixture of Feature Difference (MoFD) module. The proposed module uses a dynamic weighting strategy to adaptively select and aggregate informative features from both individual source sequences and explicit differences among them. Quantitative and qualitative evaluations demonstrate that our proposed method outperforms baselines in terms of image fidelity and perceptual quality on the UCSF-PDGM and UPenn-GBM datasets. Our method successfully reconstructs tumor regions and achieves segmentation performance comparable to that obtained with real CE-MRI. Our ablation studies show that although adversarial learning improves perceptual quality, a regression-based variant of our method achieves higher structural fidelity.
[CV-67] Bridging the Markerless Tracking Validation Gap: Intraoperative X-ray Workflow Integration MICCAI2026
链接: https://arxiv.org/abs/2610.21696
作者: Connor Daly,Salvatore Russo,Jinendra Ekanayake,Daniel S. Elson,Ferdinando Rodriguez y Baena
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: This version was submitted to MICCAI 2026 for review, see journal reference for published version
Abstract:Markerless intraoperative tracking of spinal anatomy via RGB-D sensors is a promising alternative to invasive bone-anchored fiducials, yet no existing clinical dataset provides independent subsurface ground truth for validating these methods. We address this gap with a clinically viable pipeline that derives per-vertebra 6-DoF pose from routine intraoperative C-arm fluoroscopy, co-registered to the RGB-D camera frame through a multimodal fiducial array detectable in both imaging modalities. Critically, the pipeline requires no additional radiation, instrumentation, or deviation from standard surgical workflows, relying solely on the fluoroscopic images already acquired during routine level checks. A calibrated transform chain links the vertebral pose recovered via multi-view 2D-3D CT-to-fluoroscopy registration to the RGB-D coordinate frame, enabling direct comparison of any markerless registration estimate against a subsurface reference. Ex vivo validation against optically tracked ovine vertebrae confirms end-to-end accuracy of 2.06 ± 0.29 mm and 3.35 ± 0.93 degrees, sufficient to serve as independent ground truth for methods with substantially larger errors. We demonstrate the pipeline’s utility by evaluating two markerless registration strategies on clinical data spanning 11 vertebrae across 1,500 frames from three surgical sequences: the first such evaluation against an independent subsurface reference in a clinical setting. The proposed framework provides reusable infrastructure for benchmarking emerging markerless tracking methods on real surgical data.
[CV-68] Edit4D: Editable 4D Scene Generation with Textual Control
链接: https://arxiv.org/abs/2610.21668
作者: Chanwoo Kim,Tackgeun You,Hwasup Lim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages, 16 figures. Project page: this https URL
Abstract:We introduce Edit4D, an editable text-to-4D scene generation framework that constructs 4D scenes from separate background and dynamic character representations. Recent advances in generative models have enabled text-driven dynamic scene synthesis, but editable 4D scene generation remains challenging due to the need to preserve spatial geometry and temporal dynamics. Existing approaches typically generate holistic scene videos, where the background and dynamic characters are coupled, making object-level editing and flexible camera control difficult. Edit4D addresses these limitations through a stage-specific generation pipeline. Given a user instruction, Edit4D first decomposes it into stage-specific prompts for background, dynamic character, and scene-character composition. Using these prompts, we synthesize multi-view scene images with text-to-image and camera-conditioned generative priors and reconstruct a background 3D Gaussian splatting representation. We then generate a character image, animate it with an image-conditioned video prior, reconstruct a 4D mesh, and bind Gaussian primitives to obtain a character representation. Finally, both components are composed through text-guided segmentation and contact-aware optimization. A global scene refinement is then performed to improve harmonization and rendering consistency. By integrating stage-specific components, Edit4D enables controllable camera motion, flexible object-level editing, and the practical construction of editable 4D scenes. Project page: this https URL
[CV-69] FusionAudit: Pathway-Conditioned Robustness Auditing for Native Multimodal Models NEURIPS2026
链接: https://arxiv.org/abs/2610.21667
作者: Jiaming Zhang,Zherui Li,Siqi Guo,Junhao Dong,Kun Wang,Wei Yang Bryan Lim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: NeurIPS 2026
Abstract:As vision-language models (VLMs) transition from frozen-encoder pipelines to native multimodal architectures, the visual channel increasingly serves as both a perceptual input and a carrier of rendered text that competes with user prompts. Current single-attack, pixel-centric evaluations fail to capture this complexity: they collapse distinct visual attack surfaces into a single scalar and ignore how architectural shifts alter vulnerability. We introduce PaRMA, an evidence-based auditing framework that defines pathway-conditioned risk over two primary image-level interventions: pixel perturbation and rendered-text insertion. Instead of arbitrary attack enumeration, PaRMA uses pathway-specific diagnostics to expose architecture-specific blind spots. Notably, we uncover a see-but-not-deceived signature where native VLMs attend heavily to typographic overlays but visually reject them, and we identify an objective mismatch between encoder-level and answer-level attacks in deep-fusion models. To exploit these findings, we instantiate FusionAudit, introducing two diagnostic-motivated attacks: MANTIS (margin-aware pixel perturbation) and Nat-Evid (host-congruent semantic insertion). Evaluating 18 open-weight VLMs across a novel nativeness taxonomy, we demonstrate that legacy single-attack rankings (e.g., PGD) are unstable for native models. Our diagnostic-driven attacks substantially raise the estimated residual risk, showing that PaRMA provides a falsifiable, evidence-driven auditing methodology for uncovering vulnerabilities missed by standard benchmarks.
[CV-70] Digital Typhoon Dataset V3: Benchmarking for Multi-spectral Extension and Analyzing Systematic Bias in Intensity Regression
链接: https://arxiv.org/abs/2610.21654
作者: Asanobu Kitamoto,Victor Le Louarn,Noelia Falczuk,Ngoc Anh Tong,Anja Veleva
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 23 pages, 6 figures, dataset paper
Abstract:This paper presents the Digital Typhoon Dataset V3, a new version of the longest typhoon satellite image dataset spanning nearly 50 years, designed to benchmark machine learning models on long-term spatio-temporal data. The latest addition to Dataset V3 is multispectral satellite data, enabling multi-channel models. We investigate data-splitting strategies and present benchmarking results for the Northern and Southern Hemispheres. Experimental results show that, under the current dataset construction and experimental setting, multi-channel inputs do not clearly outperform single-channel inputs for pressure regression. Among the evaluated model architectures, image types, and spectral channels, ResNet-18, full-resolution images, and Channel 1 generally provide the best single-channel performance for pressure regression. Performance also differs between the WP and AU datasets by regression target, with lower RMSE on WP for pressure regression and on AU for wind regression. These results highlight the importance of considering dataset-specific and task-specific characteristics when interpreting benchmark performance. Additionally, we analyzed systematic bias in intensity regression, which leads to underestimation of intensity for strong typhoons. We tested imbalance-aware weighting strategies, including inverse-density weighting inspired by BalancedMSE, and showed that these approaches are not sufficient to remove the bias. The dataset is publicly available at this http URL and this https URL.
[CV-71] SCMerge: Spatial-Consistency-Aware Token Pair Merging for Efficient Semantic Segmentation
链接: https://arxiv.org/abs/2610.21642
作者: Junfeng Ren(1),Yi Gong(1),Yuan Zeng(2) ((1) Southern University of Science and Technology, (2) Shenzhen Technology University)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 4 figures
Abstract:Vision Transformer architectures have showcased impressive performance on semantic segmentation, yet their quadratic attention complexity hinders practical applications. Recent works have focused on merging tokens to reduce sequence length and thus improve transformer efficiency. However, existing token merging methods primarily rely on feature similarity to identify similar tokens, overlooking spatial consistency, which may introduce segmentation errors in semantic segmentation. In this work, we propose SCMerge, a training-free token merging method that jointly considers feature similarity and spatial consistency in pairing similar tokens. The pairwise tokens are then merged within the encoder, and the resulting token sequences are reconstructed onto the original patch grid after decoding. The proposed method is flexible and can be plugged into existing vision Transformer-based semantic segmentation pipelines. Extensive experiments on the benchmark datasets ADE20K, Pascal Context, and Cityscapes show that SCMerge consistently achieves a better trade-off between segmentation accuracy and efficiency than baseline methods.
[CV-72] DIAL: Dynamic and Informed Adjustment of the Native Stochastic Law for Reliable Weather Ensembles
链接: https://arxiv.org/abs/2610.21639
作者: Yucheng Wang,Haonan Guo,Tianhao Zhao,Di Wang,Yu Wu,Zengmao Wang,Bo Du
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 5 figures, including appendix
Abstract:Ensemble weather forecasts can look reliable in global averages while hiding too little spread in some cases and too much in others. These errors vary across variables, spatial scales, lead times, and weather states. Adapting randomness during forecasting can correct these errors, but training fixes stochastic behavior beforehand and output calibration acts afterward. Inference guidance also acts during forecasting, but typically scores ensemble members separately rather than the ensemble as a whole. Reliability feedback arrives at the end, and tracing it backward through many chaotic forecast steps is expensive and unstable. We therefore introduce Dynamic and Informed Adjustment of the Native Stochastic Law (DIAL), which controls a frozen model’s native stochasticity. DIAL learns from past forecasts to predict whether each variable and scale needs more or less spread. As a forecast develops, it estimates how possible noise changes will affect later forecast accuracy and reliability. It combines these predictions to make small noise updates at each step. This loop avoids retraining the weather model and full backpropagation through the forecast. We evaluate DIAL on FourCastNet 3 (FCN3) using later ERA5 years unseen during training. DIAL provides better scale-resolved reliability than learned postprocessing and reward-based control baselines. Relative to the unmodified Native FCN3 reference, DIAL achieves an 8.27% improvement in the primary multiscale continuous ranked probability score (CRPS).
[CV-73] Anchor3DGS: Feed-Forward 3D Gaussian Splatting with Compact Anchor-Based Representation NEURIPS2026
链接: https://arxiv.org/abs/2610.21634
作者: Sheng Ye,Jiangke Lin,Fudong Wang,Yi Yuan,Zhen-Hui Dong,Yong-Jin Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: NeurIPS 2026
Abstract:Recent feed-forward 3D Gaussian Splatting (3DGS) methods have enabled effective multi-view 3D reconstruction by predicting Gaussian primitives in a single forward pass. However, existing approaches predominantly adopt a pixel-aligned paradigm that predicts one Gaussian per input pixel, coupling the Gaussian count to the input resolution and number of views. This also leads to redundant primitives in textureless regions and insufficient coverage in geometrically complex areas. We present Anchor3DGS, a pose-free feed-forward 3DGS framework that replaces pixel-aligned prediction with a compact anchor-based representation. Guided by information entropy, our method places a set of 3D anchors in the scene, concentrating them in regions of higher visual complexity while maintaining broad spatial coverage. Each anchor performs occlusion-aware aggregation of multi-view image features, exchanges information with other anchors, and is decoded into a set of local Gaussians. A lightweight feature enhancement module further refines the coarse RGB output from Gaussian splatting and improves rendering fidelity. Extensive experiments on diverse benchmark datasets demonstrate that our Anchor3DGS achieves state-of-the-art novel view synthesis quality while only using an order of magnitude fewer Gaussian primitives.
[CV-74] raining-Free Target-Aware Refinement of 3D Gaussian Language Map
链接: https://arxiv.org/abs/2610.21619
作者: Sibaek Lee,Kyeongsu Kang,Seongbo Ha,Yewon Jeon,Joonyeol Choi,Daegeol Ko,Jaehyuk Hur,Taemin Choi,Minsu Kim,Hyungjoo Ahn,Wonsok Yoo,Hyeonwoo Yu
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 8 pages, 6 figures. Project page: this https URL
Abstract:Distilling vision-language features into 3D Gaussian Splatting yields a language map that can be queried with arbitrary text. In deployment, however, a small set of user-specified targets must be reliably located, which current distilled maps often miss, while open-vocabulary queryability for other words must remain intact. We propose a training-free, target-aware refinement of an already-distilled language map. Guided by a text-promptable segmenter, the visual embedding inside each detected target region is blended with the target text embedding, a convex combination whose coefficient sets the amount of injected prior, while non-target regions are left untouched. Injecting too much overwrites appearance cues and general queryability, so we seek, for each target, the smallest coefficient that still lets it win, and we obtain every coefficient in closed form, leaving no hyperparameters to tune. We further identify a failure mode of injection in language-goal localization. Text embeddings share a common component, and injecting them makes the standard cosine read-out query-agnostic, which a centering step in the read-out removes. Across three benchmarks, the refinement consistently improves open-vocabulary segmentation, roughly doubling the base mIoU on Replica, and localizes language goals more reliably than the base map. It also transfers as a plug-in to off-the-shelf field-based and object-level language maps and supports a real-world demonstration in which a legged robot navigates to natural-language goals. Project page and code: this https URL
[CV-75] A Decision-Aware Evaluation Protocol for Uncertainty in Medical Image Segmentation
链接: https://arxiv.org/abs/2610.21616
作者: Saket Maganti,Shreyank N Gowda
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Medical segmentation uncertainty is often evaluated with calibration error or pixel-level error-detection scores, but these metrics do not show whether uncertainty improves downstream review decisions. We propose a decision-aware evaluation protocol that separates uncertainty estimation from deferral policy and evaluates their combination under fixed review budgets: which pixels, regions, or images should be accepted, and which should be reviewed. The protocol reports fixed-budget error reduction, risk-coverage behaviour, region-level review proxies, structural failure modes, probability-derived baselines, and sensitivity to review assumptions. We instantiate it across retinal vessel, skin lesion, and polyp segmentation with multiple backbones and uncertainty estimators. Under matched inference budgets and canonical multi-layer MC Dropout, test-time augmentation (TTA) gives a stronger error-detection signal across matched MC–TTA comparisons, including architecture and dropout-placement checks. However, estimator rankings are not universal. For example, probability-derived scores remain competitive, deep ensembles can outperform TTA, learned deferral helps mainly when the base estimator is weak, and rankings can change with review unit, uncertainty target, correction assumptions, and cost model. Structural and second-observer analyses further show that boundary and thin-structure failures are partly tied to annotation ambiguity. These findings support evaluating uncertainty as an estimator-policy-review configuration, not as a standalone score map. Code: this https URL
[CV-76] Hear Localize and Reason : Spatially Aware Scene Understanding for Audio-Visual LLM s NEURIPS2026
链接: https://arxiv.org/abs/2610.21588
作者: Kim Sung-Bin,Lee Jung-Mok,Jinwoo Jung,Oh Hyun-Bin,Hyeonggon Ryu,David Harwath,Tae-Hyun Oh
类目: Computer Vision and Pattern Recognition (cs.CV); Sound (cs.SD)
备注: NeurIPS 2026; Project page: this https URL
Abstract:Audio-visual large language models (AV-LLMs) have made strong progress in multimodal understanding, yet most treat audio as monaural semantic content, limiting their ability to reason about where sounds originate and how they relate to the visual scene. Recent spatial audio-visual studies address parts of this problem, but often focus on specific abilities, such as spatial correspondence or direction and distance reasoning. In this work, we introduce HLR-AVSceneQA, a comprehensive benchmark with a hear-localize-reason hierarchy that evaluates whether models can recognize what is heard, localize sound sources in egocentric and allocentric views, and use these localized sources to reason about how they relate to the visual scene. We further introduce HLR-LLM, which augments a strong audio-visual foundation model with a binaural spatial audio branch alongside its monaural semantic audio pathway. We train HLR-LLM with a curriculum that incorporates final-stage chain-of-thought supervision for multi-step spatial reasoning. Experiments show that existing AV-LLMs remain limited in spatial grounding and relational reasoning, while HLR-LLM substantially improves spatial audio-visual scene understanding without compromising semantic audio-visual perception.
[CV-77] MYOPIA: Understanding the Limits of Tree-Based Reasoning in Vision-Language Models
链接: https://arxiv.org/abs/2610.21573
作者: Islam Mesabah,Sary Elmansoury,Gerrit Großmann,Peter Neigel,Sebastian J. Vollmer
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Decision trees can make vision-language model (VLM) classification interpretable by replacing a single prediction with a sequence of root-to-leaf questions. However, the accuracy cost of this greedy, non-backtracking procedure remains unclear. We introduce \textbfMYOPIA, a benchmark comprising four LLM-generated, human-refined decision trees for GTSRB, MNIST, CIFAR-10, and MSCOCO, together with matched prompts, diagnostic protocols, and an evaluation harness. Across GPT-4o, Qwen-VL Max, and LLaMA-3.2 11B Vision, tree traversal underperforms zero-shot prompting in all 12 dataset–model pairs; paired bootstrap analyses consistently favor zero-shot. Our diagnostics characterize interacting failure modes associated with this gap. Models often recall tree-encoded class attributes more accurately than they ground the same attributes in images: the mean label–image accuracy gaps are 9.5, 13.8, and 12.6 points for GPT-4o, Qwen-VL Max, and LLaMA, respectively. Incorrect local decisions then irreversibly exclude the correct class, while answer extraction introduces additional model-dependent loss. A node-level audit of the four trees finds that split entropy is weakly but consistently associated with lower node accuracy, whereas depth, branching factor, and within-node grounding differences are less stable. \textbfMYOPIA therefore provides a framework for diagnosing hierarchical visual classification and motivates testing tree designs that prioritize perceptual answerability and recovery from early errors. Code is available at this https URL.
[CV-78] Decoupled Safety Control: A Safety-Control Algorithm for Training-Free Safety Guidance NEURIPS2026
链接: https://arxiv.org/abs/2610.21569
作者: Huilin Zhou,Ruoxi Cheng,Yuhang Wang,Yuming Liu,Minghao Sun,Ruolong Ma,Lan Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to NeurIPS 2026
Abstract:Diffusion models achieve strong performance in text-to-image generation, but they can still produce unsafe content. Training-free safety guidance offers a practical inference-time alternative that intervenes during sampling without modifying model weights. However, many existing safeguards still rely on a coupled safety-control chain. In this chain, the same unsafe condition simultaneously specifies what should be avoided, indicates the current unsafe signal, constructs the intervention, and determines when that intervention remains active. This coupling can work when unsafe factors are explicit, but becomes less reliable once harmful and benign semantics are entangled in the current generation state, often leaving residual unsafe content or unnecessarily degrading benign-generation quality or prompt fidelity. We address this problem with Decoupled Safety Control (DSC), a plug-in safety-control algorithm for training-free safety guidance. DSC reformulates training-free safety guidance as an explicit state-dependent control process, so the executed intervention is determined from the current generation state rather than inherited directly from a single unsafe condition. Experiments on standard safety benchmarks show that DSC improves representative latent- and text-space safeguards, achieving stronger unsafe-content suppression while preserving benign-generation quality and maintaining efficient inference.
[CV-79] Rethinking Classifier-Free Guidance for TI2I Flow Matching
链接: https://arxiv.org/abs/2610.21550
作者: Andreas Floros,Elena Koritska,Teofilo de Campos,Christoph Vogel,Chu-Tak Li,Anuj Sharma,Alexandros Neophytou
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:We study Text-and-Image-to-Image (TI2I) flow matching methods for controllable image editing. In a broader generative modeling context, Classifier-Free Guidance (CFG) is established as a key component responsible for high quality outputs. However, we argue that its role in TI2I modeling is poorly understood. Specifically, we make the surprising observation that the CFG mechanism is implicitly harmful for editing in the noisy phase of flow modeling, hindering instruction following and controllability. We propose a simple correction in the early flow steps, enabling a richer alignment-preservation tradeoff at no extra cost. The effectiveness of our approach is demonstrated on both pixel-space and latent TI2I flows.
[CV-80] FilterFusion: Prior-Guided Adaptive Filter Fusion for Efficient Video Denoising
链接: https://arxiv.org/abs/2610.21505
作者: Wenlong Yang,Weihang Yuan,Canran Jin,Jiahao Li,Henrik Turbell,Lifeng Sun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by MMSP 2026
Abstract:Real-world video denoising for communication systems must handle mixed degradations such as sensor noise, compression artifacts, and motion-induced inconsistency while remaining deployable under tight latency budgets. Purely deep models often achieve high accuracy but come with significant computational cost, whereas classical filters are robust and efficient but insufficiently adaptive. We present FilterFusion, a prior-guided multi-frame denoising framework that combines temporally-aware learned features with an interpretable bank of denoising branches. The model predicts a low-cost guidance signal and uses it to adaptively mix a compact set of denoising responses at the original resolution. The lightweight design uses only 109K parameters and explicitly injects temporal variance and spatial variance cues into the low-resolution controller. On CRVD (real sensor noise), it reaches 34.72 dB PSNR and 0.9330 SSIM with only 0.18 GFLOPs (network-only). Because the filter-bank operations are simple and regular (Gaussian blur, median, weighted sum), a straightforward compiled C implementation of the synthesis path achieves 17.9ms per frame (55.9 FPS) on a single CPU thread at 1080p—exceeding real-time and 122–252 times faster than lightweight CNN baselines that require over 2 seconds per frame. The results demonstrate a strong quality-efficiency-speed trade-off and a practical route to real-time denoising through low-resolution control and explicit filter-based synthesis.
[CV-81] GRASE: Geometry-native Real-to-sim Agent ic Scene Modeling from Video
链接: https://arxiv.org/abs/2610.21498
作者: Zhaorui Lin,Haohan Liu,Zhen Chen,Xiangru Huang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Real-to-sim conversion aims to turn a captured video of a real environment into an editable 3D scene while preserving the geometry and layout of what was observed. Existing methods often follow a one-pass, loosely coupled paradigm: each stage receives only limited feedback from previous stages, allowing errors to propagate without reliable correction and leaving the final scene inconsistent. We present GRASE, an agentic framework that closes this loop by treating the captured multi-view video observations as a standing reference throughout scene modeling. Instead of committing to intermediate estimates, GRASE revises its current hypothesis whenever rendered evidence disagrees with the capture, overcoming the fixed intermediate assumptions of conventional pipelines. Specifically, GRASE first fits the architectural shell and its openings from depth and multi-view cues. An agent then places generated assets in a shared coordinate frame while checking their spatial relationships. Finally, each asset’s geometry and placement are refined iteratively to improve agreement across the observed views. By coupling these steps through feedback, GRASE localizes structural discrepancies and revises only the responsible components, while preserving cross-view consistency. The resulting 3D scenes are faithful to the input, consistent across views, and suitable for simulations. In our experiments, GRASE substantially outperforms other state-of-the-art methods, including a direct real-to-sim conversion using GPT6-Astra.
[CV-82] AI-Driven Disaster Response and Displacement Monitoring
链接: https://arxiv.org/abs/2610.21482
作者: Noora Al-Emadi,Muhammad Imran,Yin Yang,Ingmar Weber,Fabjan Lashi,Gaia Rigodanza,Ivana Hajzmanova,Ferda Ofli
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:ARTIFICIAL INTELLIGENCE (AI) has demonstrated its potential across a wide range of humanitarian tasks, offering scalable and data-driven solutions to some of the world’s most pressing challenges. For instance, in addressing climate change, AI-powered models are being used for faster and more accurate global weather forecasting, control and design of renewable energy, and monitoring deforestation. Machine learning algorithms analyze satellite imagery and socioeconomic data to predict poverty and promote sustainable development. AI has also transformed disaster response and conflict monitoring, providing innovative solutions for assessing damage and urgent needs and tracking displacement.
[CV-83] Cross-Prompt Inference Mitigates Rare-Class Suppression in Imbalanced VLM Finetuning
链接: https://arxiv.org/abs/2610.21470
作者: Emanuel Ben Baruch,Sivan Biham,Alon Shoshan,Yochai Zvik,Avi Ben-Cohen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:When a vision-language model is finetuned on class-imbalanced data, rare-class recall can fall well below the model’s own zero-shot performance. In this regime, we uncover a surprising matched-prompt penalty whereby the standard choice of reusing the descriptive finetuning prompt at inference can yield lower rare-class recall than using a different descriptive prompt. We call this inference-time strategy cross-prompt inference. We evaluate diverse VLMs across six datasets and find that cross-prompt inference consistently recovers rare-class recall, with varying trade-offs in frequent-class recall. The effect reflects a train-eval prompt interaction rather than prompt quality alone, and does not arise with non-descriptive prompt changes. It is strongest when imbalance and visual confusability produce substantial rare-class suppression, and is weak or absent when such suppression does not occur. Using Tuning Contribution (TuCo), which decomposes a model’s computation into pretrained and finetuning parts, we find that cross-prompt inference reduces reliance on finetuning-induced computation and shifts the model toward its pretrained component. Reducing that finetuning contribution also substantially reduces the cross-prompt advantage. These results show that rare-class knowledge suppressed by imbalanced finetuning can remain accessible under a different inference prompt.
[CV-84] Can Visual Generation Expand What Unified Multimodal Models Know?
链接: https://arxiv.org/abs/2610.21459
作者: Wenkai Li,Wenyi Hong,Zhuoyi Yang,Jie Tang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Unified multimodal models (UMMs) receive image-generation supervision that exposes them to visual information not fully documented in text. Can learning to generate images expand what pretrained models know? We introduce Knowledge Transfer Probing (KTP), a controlled protocol that supplies new knowledge through image-generation training and tests its accessibility through text-only question answering. Across seven knowledge categories and three UMMs, we observe direct but uneven transfer from visual supervision to language. Motivated by these observations, we successively introduce Textual Alignment and Interleaved Image–Text Alignment, which connect generation and answering through shared knowledge and interleaved responses, respectively. Guidance on selected entities can improve knowledge access on other entities without QA supervision, with additional gains from interleaved alignment. Further analysis shows that knowledge correctly reproduced in images is often more accessible through text, although substantial gaps remain between the two forms of access. These findings support visual generation as a source of knowledge beyond textual supervision, and motivate joint investigation of training data design and model architecture to make this knowledge more broadly accessible.
[CV-85] CarryOn: A Dual-Loop Harness for Personalized and Reliable Robot Assistance
链接: https://arxiv.org/abs/2610.21453
作者: Zixuan Wang,Yuqi Liu,Yuxin Chen,Sengqiao Yang,Jinhui Ye,Pengguang Chen,Changsheng Lu,Shu Liu,Bei Yu,Jiaya Jia
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:While generalist vision-language-action (VLA) models have enabled diverse open-world manipulation, real-world robotic assistance requires two critical capabilities they currently lack: personalization and reliable execution. First, executing an instruction like “make me a sandwich” requires grounding subtask decomposition in persistent user preferences rather than relying on a single generic solution. Second, reliable deployment demands explainable safety assurance that proactively prevents hazards and gracefully recovers from occasional physical failures. To address these gaps, we present CarryOn, a dual-loop harness designed to augment a frozen VLA policy. The planning loop maintains persistent user-level and scene-level context that is continually grounded against new observations and revised through interaction, enabling personalized task selection and execution grounding across conversations and changing scenes. The execution loop actively supervises policy rollouts, shielding actions from predicted hazardous contact. When execution fails, the system invokes reusable control-based recovery skills that restore task progress and are refined through reflection on earlier failures. On a real robot and in simulation, CarryOn dispatches preference-consistent subtasks across users and scenes, reduces the violation rate of an already safety-finetuned VLA from 52.6% to 6.1%, and increases recovery success from 13% to 80% when refined skills are reused.
[CV-86] PET Quantification-Preserving Registration for Longitudinal Whole-Body PSMA PET/CT
链接: https://arxiv.org/abs/2610.21451
作者: Fryderyk Kögl,Anna Reithmeir,Daniel M. Lang,Emily Chan,Rickmer Braren,Julia A. Schnabel
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We adapt the diffeomorphic coarse-to-fine LapIRN architecture for longitudinal whole-body PSMA PET/CT registration in the Learn2Reg 2026 PSMAReg challenge. The method combines affine pre-registration, multi-modal deformable registration, and test-time instance optimization. We further introduce PET biomarker preservation losses for metabolic tumor volume (MTV) and total lesion glycolysis (TLG), including Jacobian-based constraints, and a per-structure rigidity regularization for skeletal structures. The submitted configuration achieved 81.0% DSC and 6.17 mm HD95, with MTV and TLG errors of 3.3% and 3.3%, respectively, on our evaluation protocol. Code is available at this https URL.
[CV-87] X-Fly: Learning to Reach and Stop in Full-Chunk Aerial Vision-Language Navigation
链接: https://arxiv.org/abs/2610.21450
作者: Ruibin Ju,Ruifeng Cai,Min Xu,Lin Wang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 13 pages, 5 figures, 3 tables. Project page: this https URL
Abstract:Aerial vision-language navigation (VLN) agents built on large vision-language models are expensive to query. Executing a complete predicted action chunk before replanning amortizes this cost, reducing model inference from 195 to 42 ms per executed waypoint for our policy. It also moves the states at which the policy is queried away from expert demonstrations, which supervise neither recovery from these states nor termination at the learner’s own arrival. On OpenFly test-seen, no-stop failures account for 47.78% of full-chunk behavior cloning (BC) outcomes. We present X-Fly, a staged training framework for full-chunk aerial VLN. Recovery DAgger labels learner-visited decision states with smooth chunks that reconnect to future expert segments and with STOP at the first in-region state. Flow-GRPO then jointly refines the chunked flow controller and a Bernoulli STOP policy from same-task closed-loop outcomes, supplemented by outcome-verified stopping supervision. With eight controls per decision, X-Fly reaches 74.18% SR on test-seen and 51.73% on test-unseen, compared with 16.45% and 18.43% for full-chunk BC and 32.04% and 29.46% for DreamFly.
[CV-88] RLDF: A Region-Level Deepfake Dataset and Benchmark for Partial-Face Manipulation Detection
链接: https://arxiv.org/abs/2610.21438
作者: Shukesh Reddy,Abhijit Das
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Existing deepfake detection models are built upon manipu- lated images that are tampered with either exclusively fully swapped or fully synthesised faces, which leads us to an open question: Can off-the-shelf deepfake models detect manipulation effectively while tampering is confined to a local facial sub-region? To fill this gap, we introduce RLDF (Region-Level Deepfake), the first dataset for region-level deepfake evaluation. Employing the proposed region-aware partial swap synthesis model, we generate over 2M par- tial face-swap images by using eight face-swapping meth- ods. Altogether, we considered seven face regions, among which there are four discrete regions (eye, nose, mouth and eyebrow) and four composite facial regions (eye and nose; eye, nose and mouth; eye, nose, mouth and eye- brow; and full face). In addition, we define three evalu- ation protocols and benchmark 12 state-of-the-art detectors across the five popular standard cross-dataset evaluations, DF40, and the proposed RLDF, providing a comprehen- sive audit of where and why existing detectors underper- formed. Our evaluation reveals that existing detectors fail to generalize consistently across partial facial regions, re- sulting in an overall drop of 10.5% AUC and 10% accu- racy under region-level evaluation. Further, we propose a mask-guided cross-attention model as a baseline approach that can effectively detect region-level manipulation. The proposed dataset, code, and checkpoints are available at this https URL.
[CV-89] ManiFusion: Unlocking the High-Throughput Generation via Superposition in Manifold Space
链接: https://arxiv.org/abs/2610.21433
作者: Minkyu Kim,Baekseung Kim,Junhoo Lee,Jangho Kim,Nojun Kwak
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Diffusion models achieve high-quality image generation but remain computationally expensive due to iterative denoising across many timesteps. Existing acceleration methods mainly reduce per-sample latency, but gains along this axis are becoming saturated. We instead explore an orthogonal throughput-oriented direction by processing multiple samples within a single denoising evaluation. To this end, we propose ManiFusion, a backbone-agnostic framework for multi-sample generation through fused-space computation. ManiFusion jointly processes multiple samples by fusing them before the denoising backbone and separating them afterward through unfusion. We provide a geometric perspective on how multiple samples can share a fused representation while retaining the information needed to guide their individual generation trajectories. To better align fusion allocation with denoising stages, ManiFusion uses Trajectory-Fusion Scheduling (TFS), which dynamically adjusts fusion factors throughout sampling. ManiFusion generalizes across diffusion families, including both U-Net and transformer backbones in pixel and latent spaces. We further show that ManiFusion composes with existing acceleration methods and reaches throughput regimes unattainable by prior approaches with minimal quality degradation.
[CV-90] Where How and When: Query-Guided Grounding for Video Understanding
链接: https://arxiv.org/abs/2610.21422
作者: Yogesh Kumar
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Understanding video content requires reasoning at multiple levels of granularity, from identifying objects to interpreting their interactions and localizing when specific actions occur. This thesis studies three query-guided grounding tasks that answer where key objects are (object-level grounding), how they interact (relationship-level grounding), and when specific moments occur (moment-level grounding). For object-level grounding, we present QDETRv, a query-guided extension of the Detection Transformer for one-shot localization of unseen objects in videos, which combines query-context fusion, a memory module for temporal consistency, and unsupervised pretraining on synthetic object trajectories. We further propose the Temporal Object-Aware Vision Transformer (TOViT) for few-shot video object detection, which builds on a language-aligned open-set vision encoder and propagates high-confidence object features across frames. For relationship-level grounding, we introduce few-shot referring relationships in videos: given a (subject, predicate, object) query and a few support videos sharing the predicate, the entities are localized by optimizing a T-partite random field using a meta-learned relation network with query-conditioned translational embeddings, Global Semantic Aggregation, Local Localization Aggregation, and belief-propagation-based message passing. For moment-level grounding, we propose the Moment Alignment Transformer (MATR) for video-to-video moment retrieval, which uses dual-stage Dynamic Time Warping for pre- and post-fusion alignment, learnable queries for boundary prediction, and self-supervised pretraining with temporal augmentations; MATR also extends to text-to-video moment retrieval with performance comparable to, or better than, the state of the art. These contributions advance query-guided video understanding across spatial, relational, and temporal grounding.
[CV-91] CLAD: Keypoint Matching Using Global Line Correspondences
链接: https://arxiv.org/abs/2610.21409
作者: Masaki Satoh
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 5 pages, 2 figures
Abstract:Traditional keypoint descriptors, such as SIFT (Scale-Invariant Feature Transform) and ORB (Oriented FAST and Rotated BRIEF), are computed from local image patches. They may be unreliable in scenes with repetitive patterns or weak textures, where patches at different locations can look similar. We propose the Contextual Line-Assisted Descriptor (CLAD), a keypoint matching method that addresses this problem by using global geometric context from line correspondences. CLAD represents each keypoint with a vector whose elements indicate its proximity to matched lines. It can therefore use the keypoint’s spatial relationships to the lines in the scene. We use HOME (Hough-space One-dimensional Matching of Extrema) to detect both clear and weak linear structures and to establish enough line-to-keypoint associations to represent this context. Evaluations on the TUM RGB-D SLAM dataset and outdoor video sequences show that CLAD provides geometrically consistent matches that complement those obtained with traditional local descriptors. In repetitive scenes, where local descriptors may not distinguish similar patches, CLAD uses global structure to improve matching reliability. The method is computationally lightweight and practical to implement.
[CV-92] Perception or Reasoning ? Diagnosing and Tackling Reinforcement Learning Bottlenecks in VLMs
链接: https://arxiv.org/abs/2610.21380
作者: Zeyuan Yang,Zhuohao Chen,Jiaben Chen,Yao Zhang,Maohao Shen,Chuang Gan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reinforcement learning has brought strong gains to language model reasoning, but its improvements on visual tasks remain less consistent. One difficulty is that visual task failures can arise from different sources: the same benchmark may be limited by perception for one model and by reasoning for another. We introduce Perception and Reasoning Attribution (PARA), which measures the remaining perceptual and reasoning headroom of each model–benchmark pair by comparing how many current errors are recovered with more accurate visual information versus a stronger reasoner. Across 4 vision-language models and 17 multimodal benchmarks, PARA broadly agrees with the expected characteristics of existing benchmarks while revealing substantial model-specific variation. More importantly, it shows that RL gains become smaller as perceptual headroom increases, but larger when reasoning explains more of the recoverable errors, suggesting that current RL exploits reasoning headroom more effectively than perceptual headroom. Motivated by this gap, we analyze how existing RL objectives affect visual perception and find that reward- and token-level signals influence visual attention only indirectly. We therefore propose GRAD, a group-relative attention distillation objective that uses the best rollout in each group to directly guide the visual attention of weaker rollouts. GRAD consistently improves performance across two model scales and seven benchmarks, and further complements existing RL methods.
[CV-93] SATOR: Backpropagation-Free 3D Test-Time Adaptation via Sample-Adaptive Token Routing WACV2027
链接: https://arxiv.org/abs/2610.21367
作者: Inseok Kong,Geunyoung Jung,Jiyoung Jung
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by WACV 2027
Abstract:Test-time adaptation (TTA) mitigates the performance degradation of source-pretrained models on unlabeled test data under distribution shifts. Recent backpropagation-free token-purging methods address 3D point cloud distribution shifts by filtering unreliable tokens. However, their reliance on predefined purge-size candidates requires multiple forward passes per sample and increases inference cost. We propose SATOR (Sample-Adaptive TOken Routing), a source-free and backpropagation-free 3D TTA framework that eliminates this search by deriving both the partitioning threshold and routing decision directly from each sample. SATOR computes token abnormality scores from complementary attention and geometric relation graphs. It then derives a sample-specific partition into retained and pruned tokens via Otsu-based separability, without evaluating candidate sizes. Keep-ratio-calibrated routing then selects the pruned prediction only when its entropy, penalized by the keep ratio, is lower than that of the original prediction. Experiments on ModelNet40-C, ScanObjectNN-C, and ShapeNet-C demonstrate that SATOR achieves state-of-the-art accuracy on three 3D corruption benchmarks.
[CV-94] CoDeFont: One-Shot Font Generation via Multi-Scale Factor-Specific Representation Learning
链接: https://arxiv.org/abs/2610.21364
作者: Rejoy Chakraborty,Prasun Roy,Saumik Bhattacharya,Umapada Pal
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 6 figures
Abstract:One-shot Font Generation (OFG) aims to synthesize a target glyph by transferring the stylistic characteristics of a single reference glyph while preserving the structure of the target character. However, accurately capturing font style from a single reference image remains challenging due to the limited stylistic information available in a single glyph, and this challenge is further amplified for complex scripts. We propose CoDeFont, a script-agnostic OFG framework that addresses this challenge through multi-level style-content factorization and progressive style-content reconciliation. CoDeFont independently aggregates complementary style and structural information across multiple spatial levels to obtain factor-specific representations, while a contrastive objective explicitly encourages separation between style and content embeddings. During reconstruction, the disentangled representations are progressively injected through multi-level style conditioning, enabling stylistic information to be incorporated at different decoding resolutions while preserving the underlying glyph structure. We evaluate CoDeFont on structurally distinct Bengali and Latin scripts across seen and unseen font styles. Extensive quantitative and qualitative analyses demonstrate consistent improvements over existing state-of-the-art methods, highlighting the effectiveness of the proposed framework for OFG across diverse writing systems.
[CV-95] L2GSSL: Label-Efficient Local-to-Global Self-Supervised Learning for Volumetric Healthcare Data
链接: https://arxiv.org/abs/2610.21356
作者: Po-Cheng Cheng,Hung-Hsuan Chen
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 10 pages, 2 figures, 10 tables. Under review at IEEE BigData 2026
Abstract:Healthcare institutions routinely accumulate high-dimensional computed tomography (CT) volumes, but only a fraction can be annotated by experts. We present L2GSSL, a local-to-global self-supervised learning framework that leverages unlabeled volumes by decomposing pretraining into intra-slice and inter-slice representation spaces. A slice encoder learns view-invariant local features, and a global encoder aggregates the resulting tokens to model volumetric context. The stop-gradient isolates the local space from the global objective, while slice-wise execution and a three-phase schedule support training under constrained GPU memory. On LIDC-IDRI lung nodule malignancy classification, L2GSSL reaches 0.9229 AUROC with only 10% of the labels and 0.9377 with 25% , exceeding the volume-level VICReg by 0.0229 and 0.0294 , respectively. With all labels, the strongest supervised baseline exceeds L2GSSL by only 0.0044 AUROC, consistent with self-supervised pretraining providing its greatest value when annotations are limited. Without adaptation to the external LNDb cohort, a model fine-tuned on 25% of LIDC-IDRI obtains 0.8138 AUROC and 0.7589 AUPRC, improving over the strongest evaluated self-supervised baseline by 0.1130 and 0.0954 . Experiments across data sources, thoracic and abdominal anatomy, downstream tasks, and two backbones examine healthcare-data variety, while lesion-slice masking shows that the predictions depend on lesion-bearing content.
[CV-96] When Should a Low-Light Pipeline Generate? Auditable Risk Control for Selective Generative Enhancement
链接: https://arxiv.org/abs/2610.21335
作者: Qiyi Xie,Wei Chen Ng,Suyang Xi,Xicheng Wang,Ziyao Li,Bohong Xu,Wenyi Liu,Jian Ding Tan
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Generative low-light enhancers recover plausible texture but may hallucinate structure; restoration networks preserve structure but can leave regions under-enhanced, and neither advantage is spatially uniform. Therefore, we formulate selective generative enhancement as a decision problem: determine where overriding a frozen base expert is beneficial and refuse to switch when the evidence is insufficient. We propose VetoNet, an auditable regional routing framework centered on a certified veto. It first measures the complementary strengths and routing direction of a frozen restoration generative pair, then predicts the utility of regional overrides from deployment-available signals alone and admits an override only when a distribution-free harm-budget certificate says the perceptual (LPIPS) risk budget allows it. Experiments across multiple expert pairs and domains show that VetoNet adapts its routing behavior to the strength and reliability of the available evidence. When expert complementarity is sufficiently predictable, it achieves statistically significant certified gains; when the relative strengths of the experts change, it correspondingly reverses the routing direction; and when routing evidence is weak or affected by cross-domain shift, it conservatively vetoes switching.
[CV-97] Semantic Latent Alignment for Flow Matching with Invertible Bi-Lipschitz Adapters
链接: https://arxiv.org/abs/2610.21332
作者: Felix Nützel,Mischa Dombrowski,Luca Hagen,Bernhard Kainz
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 17 pages, 5 figures
Abstract:Pretrained latent spaces trade compression, reconstruction, and the learnability of the coordinates a downstream generator has to work in, which complicates adapting these spaces to different domains or tasks: continued training is expensive, and without a new alignment objective it often fails to improve downstream generation. We instead leave the encoder and decoder fixed and change only the coordinates between them, using an invertible bi-Lipschitz adapter aligned offline to a semantic teacher and frozen during class-conditional flow matching. Invertibility preserves compatibility with the pretrained autoencoder, while the bi-Lipschitz constraint limits how far alignment can distort its latent geometry. This provides a simple way to improve semantic organization while retaining the compression and reconstruction properties of the original latent space. On two medical imaging datasets, constrained latent adaptation improves feature metrics, seed stability and counterfactual edit strength, while remaining competitive on downstream classification. In contrast, unconstrained adaptation can become highly anisotropic and sensitive despite remaining invertible, highlighting the importance of controlling geometric distortion. Our results suggest that carefully constrained post-hoc adaptation offers a parameter-efficient and tuning-free alternative to jointly modifying the representation during generative training.
[CV-98] CertEx: Exact Visual Retrieval at Scale by Certified Exclusion
链接: https://arxiv.org/abs/2610.21325
作者: Dongfang Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Exact k -nearest-neighbor (kNN) search over image descriptors and embeddings underlies visual retrieval, near-duplicate detection, the evaluation of learned representations, and increasingly the memory of LLM agents. However, exhaustive search computes every distance at full precision, whereas approximate indexes return recall below one. We observe that exact search only needs a proof that each excluded vector cannot be a neighbor. Because such a proof only has to be conservative, it can be checked in low precision, much as the retina examines the whole scene at low resolution. We call this principle certified exclusion and realize it in an exact kNN method for GPUs, which we call \certex. Specifically, \certex\ stores a lower-bounding projection of each vector in half precision and folds a rounding certificate into the same tensor-core dot product, so that only the few vectors that the test keeps need exact distances. We prove that \certex\ returns the exact kNN of every query. On sixteen datasets with up to 100M vectors, \certex\ is exact for every query at 7.1 \times to 18.0 \times the throughput of the fastest cuVS search on one H100 and outperforms quantized and graph indexes whenever every query must be exact.
[CV-99] AdaLoop: Thinking with the Vision Encoder
链接: https://arxiv.org/abs/2610.21323
作者: Xinran Nicole Han,Yiyang Lu,Katelyn Gan,Todd Zickler
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Most vision-language models (VLMs) treat the vision encoder as a static feature extractor: it encodes an image once, independently of the question, and leaves subsequent reasoning to the language model. Yet different questions may require different visual operations and processing depths. We present AdaLoop, a framework for question-conditioned visual recurrence in VLMs. AdaLoop adapts visual computation to the question and reuses the encoder’s transformations across multiple iterations, with the pretrained vision weights frozen. We also introduce VisRO, a procedural dataset and generation pipeline for training and evaluating vision-centric reasoning. On six VisRO tasks, AdaLoop outperforms standard fine-tuning and question-conditioned single-pass baselines with matched training data and optimizer updates. These gains hold across three VLM families on both in-distribution and out-of-distribution examples. Our findings suggest that pretrained vision encoders can contribute to reasoning through task-directed, iterative visual computation.
[CV-100] CoViG: Coarsened Vision GNN for Image Classification
链接: https://arxiv.org/abs/2610.21321
作者: Martial Guidez,Stefan Duffner,Christophe Garcia
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Although Graph Neural Networks are widely used in various domains, they remained mostly unexplored in computer vision. Convolutional Neural Networks and Vision Transformers were leading the field, until the quite recent introduction of Vision Graphs (ViG) that treat images as a set of nodes and apply graph convolution on their k-NN graphs. ViGs show impressive accuracy on image classification but their computational cost remains relatively high. Inspired by optimization methods for GNNs, we present CoViG, a novel approach combining graph coarsening and ViGs. Our method introduces an additional MinCUT pooling step within the Grapher component of intermediate ViG blocks. By simplifying the intermediate graphs and thus decreasing the number of nodes of specific blocks, our approach considerably reduces the computational cost of ViG during inference. We evaluated our approach on ImageNet with several ViG architectures, and we obtained accuracies equivalent to the original ViG models with only about 51% of FLOPs.
[CV-101] Loud and Clear: Making Visual Evidence Count in Multimodal Document RAG
链接: https://arxiv.org/abs/2610.21318
作者: Eunchan Jo,Seounghyun Jang,Sangwoo Mo
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Multimodal document retrieval-augmented generation (RAG) has focused largely on retrieval, yet its answers depend on how well a vision-language model reads the retrieved pages. We find that evidence in retrieved pages remains underused even when the correct pages are retrieved. On SciEGQA oracle pages, five prior methods give the generator 0.7K to 2.7K visual tokens per query on average under a common image processor, yet raising the page-only budget from 1K to 32K improves accuracy by 15.22 points. At a fixed 16K budget, adding ground-truth crops improves accuracy by 3.75 points over page-only input. We study what the generator needs from retrieved pages through two requirements: Loud, making full use of visual token capacity, and Clear, refining the page so that its evidence becomes prominent. For Loud, increasing either the visual token budget or visual fidelity, the image resolution those tokens preserve, improves accuracy with the other held fixed, and full-page scaling outperforms repetition and tiling at matched budgets. For Clear, suppressing surrounding content improves accuracy without enlarging selected regions, and adding OCR of these regions to page images outperforms either input alone. We implement both as Loud and Clear (LoC), which combines higher-resolution pages with OCR from query-relevant regions without ground-truth annotations or additional training. LoC outperforms eight prior methods on SciEGQA, ViDoRe-v3, and MMLongBench-Doc, improving over the strongest, ColQwen3, by 8.88, 3.09, and 4.22 points with the same retriever and generator. It also outperforms its page-only variant under the same visual token budget. These results locate a bottleneck beyond retrieval: how the retrieved pages are given to the generator.
[CV-102] VLA-Δ: Learning Physical State Transition for Vision-Language-Action Models
链接: https://arxiv.org/abs/2610.21306
作者: Qi Lu,Jiarun Li,Ce Zhang,Miao Pan,Yi Pan,Xu Wang,Wenqi Zhang,Xuhong Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 24 pages, 8 figures, 9 tables, including appendices. Code: this https URL
Abstract:Vision-language-action (VLA) models transfer semantic knowledge acquired through vision-language pretraining to robot control, achieving strong performance across various benchmarks. However, most VLAs directly map observations to demonstrated actions without explicit supervision of intermediate physical state changes, making them prone to overfitting observation-action correspondences. Some methods introduce intermediate reasoning or future-state prediction in language, visual, or latent spaces to guide action generation, but do not explicitly supervise the physical state transition itself. We therefore propose VLA- \Delta , a general framework that uses physical changes between current and future states as explicit learning targets. To support this objective, we construct a large-scale pretraining dataset and derive physical state-transition annotations covering task-relevant semantics and visual motion changes, providing more direct supervision for continuous action generation. During inference, the model generates actions directly from current observations without additional inference overhead. VLA- \Delta achieves performance competitive with the state of the art on LIBERO and RoboTwin, and outperforms \pi_0.5 in real-robot experiments. Our code is available at this https URL
[CV-103] Dusting the Phantom: Gaussian Allocation for Fading Dynamic Content
链接: https://arxiv.org/abs/2610.21302
作者: Minhyeok Roh,Minseo Lee,Jun Young Jeong,Byeonghyeon Lee,Lucas Yunkyu Lee,Jaesik Park,Eunbyung Park
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages, 14 figures. Project page: this https URL
Abstract:Dynamic Gaussian representations have made high-quality, real-time volumetric video increasingly practical, yet challenging motion can still break their reconstruction: small fast-moving objects may disappear, and fine textures in highly dynamic regions are often lost. In particular, simply training longer or using more Gaussians may still fail to recover small, fast-moving objects. We argue that the key bottleneck is not representational capacity itself, but where and when Gaussian primitives are allocated. Existing allocation strategies rely on current Gaussian locations and often overlook the spatiotemporal destinations of relocated Gaussians. To address these limitations, we introduce two complementary strategies. First, Seed Injection identifies under-reconstructed dynamic regions using temporal appearance changes and reconstruction errors, and initializes new Gaussians at geometrically plausible 3D locations through multi-view observations, independent of existing Gaussian locations. Second, Temporal Biased Relocation uses time-resolved optimization signals to guide the spatiotemporal placement of relocated Gaussians. Both strategies maintain a fixed Gaussian budget without modifying the underlying representation. Experiments across multiple dynamic-scene datasets demonstrate improved reconstruction quality, particularly in highly dynamic regions, recovering small, rapidly moving objects and fine-grained appearance details.
[CV-104] GeoPano: Towards Geometrically Accurate 3D Reconstruction from a Single Panorama
链接: https://arxiv.org/abs/2610.21297
作者: Jing OU,Zidong Cao,Liaoyuan Fan,Zhuoxiao Li,Shuai Zhang,Yinrui Ren,Dongli Wu,Haoang Li,Hui Xiong,Wufan Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Panoramic 3D reconstruction enables the holistic recovery of complete 360 ^\circ scene geometry, offering significant benefits for robotics and AR/VR applications. However, directly estimating geometry from ERP images is inherently challenged by spherical distortions. To reduce spherical projection distortions, a common strategy is to decompose panoramas into distortion-free perspective views, but independently processing these views often introduces cross-view geometric inconsistencies. To address these challenges, we introduce GeoPano, a feed-forward framework for single-view panoramic 3D reconstruction that recovers a geometrically accurate and consistent point cloud from a single panorama. Our approach is driven by two key innovations: First, we propose a panoramic-aware attention aggregator, featuring a streamlined attention architecture equipped with a novel Pano-Attention mechanism, which explicitly exploits the intrinsic spherical geometry of the panorama to guide cross-view interactions, thereby suppressing geometric inconsistencies. Furthermore, we formulate a hybrid supervision strategy to jointly enforce fine-grained local details in perspective space and global structural coherence in spherical space. Experiments demonstrate that our method significantly outperforms previous SOTA approaches in reconstruction accuracy, geometric consistency, and visual fidelity.
[CV-105] ReState: Noise-Level-Aware State Correction for Training-Free Object Removal in Few-Step Diffusion Models
链接: https://arxiv.org/abs/2610.21287
作者: Donghwan Lee,Sanghyun Jo,Sihyun Kim,Ziseok Lee,Kyungsu Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 32 pages, 8 figures
Abstract:Training-free object removal commonly redirects self-attention away from the target region, but on few-step distilled diffusion models this often leaves object-shaped residuals. We identify a temporal mismatch: a structured prediction forms inside the mask at the first forward, and the short trajectory offers few later chances to revise it, while attention rerouting acts mainly in later steps. Across four distilled backbones, magnitude-matched single-forward interventions show that background-directed state correction affects the final result more than attention rerouting. We therefore propose ReState, which corrects the masked clean prediction toward background evidence retrieved from the input using a noise-level-aware weight that vanishes at the final forward. ReState retains object-excluding attention as a complementary constraint and adds no denoiser evaluations. Across two paired benchmarks and two few-step backbones, ReState achieves the best masked PSNR, SSIM, and LPIPS among training-free methods, improving PSNR by 3.17 dB on average over the strongest baseline. On OBER-Test, it also outperforms the 50-step attention-forcing reference on all three metrics at 6.7x lower latency. The same configuration transfers to two additional distilled backbones without retuning. Project page: this https URL
[CV-106] Mallea: Building Steering and Memorizing Worlds with Coding Agent
链接: https://arxiv.org/abs/2610.21283
作者: Xuanyi Liu,Ruiqi Li,Siwei Ma
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Code world models represent world dynamics through explicit state and executable rules. We investigate how to coordinate object representations, behavioral revisions, and visual memory within this formulation. We present Mallea, a system for building, steering, and memorizing worlds from text. A coding agent infers object interaction requirements from the prompt and jointly plans object structure, properties, and rules, favoring compact representations that support those requirements. The world evolves under the planned rules; optional user actions steer it, while text requests can edit properties or extend capabilities with unaffected state preserved. Object texture memory retains reusable appearance only for whitebox-defined objects and adapts it as those objects change. The program advances world state, while a generative renderer combines current structure with remembered appearance. By organizing construction, intervention, and memory around the same objects, Mallea conditions subsequent observations on the latest state and compatible appearance. Project page: this https URL.
[CV-107] LICOA: Learning Instruction-Conditioned Optical Actions for Mobile Spatial Augmented Reality
链接: https://arxiv.org/abs/2610.21282
作者: Yuchen Deng,Shunkai Zhou,Ruixuan Zhu,Haibin Ling,Bingyao Huang
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 8 figures
Abstract:Spatial augmented reality (SAR) augments physical environments with projected content, but most systems assume that the projector pose, target region, or motion path has been fixed or specified. We instead make projected light an embodied action: given a language instruction and current visual observations, a system must decide how to move and where and at what scale to place the projection. We introduce LICOA to realize this capability. Its differential-drive chassis provides room-scale approach and repositioning, while a steerable mirror with two degrees of freedom (2DoF) and projection scale provide rapid local control of the beam. A shared interface maps synchronized external and onboard RGB cameras with robot control state to physical optical actions across pretrained action models. Because instruction-conditioned optical action lacks dedicated training data and a common evaluation protocol, we construct a matched simulation embodiment and a corpus with 150 tasks and 7,500 rendered episodes, then evaluate models trained only on this corpus in previously unseen real-world environments. Across 480 complete closed-loop real-world execution trials (rollouts) on 12 tasks, the best adapted action model achieves a 57.50% success rate and 40.64% success weighted by path length, compared with 8.33% and 5.15% for the same architecture trained from scratch. A further 150 real-world ablation rollouts support the need for both camera views and coordinated control of the chassis and mirror. These results establish a physical platform, shared perception and optical action interface, corpus, and real-world evaluation protocol for treating projected light as an embodied action.
[CV-108] RACER: A Multimodal Harness for Patient Care in Hospitals
链接: https://arxiv.org/abs/2610.21266
作者: Zane Durante,Paul Tang,Keshigeyan Chandrasegaran,Amit Kaushal,Shrinidhi Kowshika Lakshmikanth,Feolu Kolawole,Yu Jun Zheng,Yifan Geng,Shobhit Agarwal,Kevin Schulman,Juan Carlos Niebles,Ehsan Adeli,Arnold Milstein,Li Fei-Fei
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 37 pages, 4 figures. Code: this https URL
Abstract:Continuous video and clinical records provide complementary information for understanding how patient behavior evolves over a hospital stay. A central challenge is how to represent behavior across the entire stay while preserving access to fine-grained visual and clinical evidence. We introduce TRACER, a training-free multimodal harness that constructs minute-level action records from video and combines them with clinical records into hourly reports. The harness equips an LLM agent with tools to query this hierarchy, inspect relevant video, and retrieve clinical records. To evaluate TRACER, we construct two benchmarks. The first contains 397 annotated one-minute clips spanning 15 atomic actions; the second poses 144 clinician-informed multiple-choice questions, grounded in 264 hours of video and clinical records from two ICU patients. TRACER achieves up to 84.8% accuracy for action detection and 61.0% QA accuracy versus 49.3% for the strongest baseline using the same backbone. Across 280 patients spanning approximately 60,000 camera-hours, TRACER identifies significantly less purposeful activity in patients with a positive delirium assessment during their stay than in those without (p 0.0001). Our results show that combining continuous video with clinical records can support understanding of patient behavior over time and across patient cohorts without training on protected health data.
[CV-109] Do VLAs Really Follow Instructions? Diagnosing How-Level Instruction Following in Vision-Language-Action Models
链接: https://arxiv.org/abs/2610.21251
作者: Sangwu Park,Wonjoong Kim,Yeonjun In,Chanyoung Park
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: Preprint
Abstract:Vision-language-action (VLA) models are expected to follow instructions like their language backbones, yet prior work has focused on What-level instructions that specify the task rather than How-level instructions that specify how to carry it out. Through our diagnoses, we find that How-level instructions do not help current VLAs and lower task success. We hypothesize that pairing each training episode with a single instruction limits How-level instruction following, and show that training with segment-level enriched instructions makes actions respond to such instructions without consistently raising task success. To measure How-level instruction following beyond standard benchmarks, we introduce SIMPLER-IN, where success requires conditions given only in the instruction. It shows that the gained responsiveness does not extend to demanding or out-of-distribution How-level instructions. How-level instruction following thus remains an open challenge.
[CV-110] FocusOPD: Structure-Guided On-Policy Distillation for Diffusion Models
链接: https://arxiv.org/abs/2610.21250
作者: Xinrui Lin,Shumin Wang,Suqi Song,Fan Li,Yanyong Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multi-teacher on-policy distillation enables rectified-flow models with diffusion-transformer architectures to combine task-specific capabilities within a single student. However, matching model outputs does not guarantee that their internal representations are aligned, and it remains unclear where and when additional feature supervision is most useful. We analyze the teacher-student feature mismatch across Transformer blocks and denoising timesteps, and find that the mismatch is concentrated in deeper Transformer blocks during early denoising and persists after OPD distillation, indicating ineffective distillation. Motivated by this observation, we introduce FocusOPD, which augments the base OPD objective with token-wise directional feature alignment restricted to deep layers and early denoising steps. Across two rectified-flow backbones and multiple OPD objectives, FocusOPD improves capability transfer in compositional generation, text rendering, and preference-based evaluation. On the primary benchmark, it improves over DiffusionOPD from 0.953 to 0.967 on GenEval, from 23.797 to 23.944 on PickScore, and from 6.052 to 6.199 on LAION-Aesthetic. Ablations show that selective deep and early supervision is more effective and efficient than alignment over the full network and trajectory.
[CV-111] A large-scale climate-aware satellite image dataset for domain adaptive land-cover semantic segmentation
链接: https://arxiv.org/abs/2610.21238
作者: Songlin Liu,Linwei Chen,Li Zhang,Jun Hu,Ying Fu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:A few well-annotated datasets for land-cover semantic segmentation have recently been introduced to advance the field of earth observation technologies. However, these datasets overlook the significant diversity among geographic areas with different climates, which can greatly impact and diversify land cover. Consequently, this leads to a domain gap in remote sensing images and severe performance degradation of the segmentation models. To enhance land-cover semantic segmentation with improved generalization ability, we conducted the first investigation into the impact of climate on this task. In this paper, we present a unique large-scale Climate-Aware Satellite Images Dataset (CASID) specifically designed for domain adaptive land-cover semantic segmentation. It consists of 980 satellite images with a size of 5000 x 5000 pixels, collected from 30 different regions around Asia, covering over 24,500 square kilometers. These images are gathered from four distinct climate zones, namely temperate monsoon, subtropical monsoon, tropical monsoon, and tropical rainforest. It includes four sub-datasets/domains, each representing one of the aforementioned climate zones. This characteristic makes CASID the first climate-aware land-cover semantic segmentation dataset with multiple domains. Additionally, we provide a comprehensive analysis of the samples from the four climate zones, emphasizing differences in global image features, image texture, category distribution, spectral value, and object shape. These analyses offer valuable insights for subsequent research in this field. Moreover, we conduct extensive experiments to evaluate the latest semantic segmentation and unsupervised domain adaptation methods on the CASID dataset. These results serve as a robust baseline for future research endeavors. Our dataset will be made publicly available at this https URL.
[CV-112] DLOSAC: Joint Spatial-Geometric Image Estimation via Delaunay-Guided Local Optimization RANSAC IROS2026
链接: https://arxiv.org/abs/2610.21215
作者: Yu Liu,Meiqi Zhou,Zipei Fan,Guanyuan Feng,Xuan Song
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the 2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2026)
Abstract:Reliable image geometric estimation from noisy matching points is typically addressed using RANSAC, often combined with local optimization to refine inliers and model parameters. However, existing paradigms are predominantly residual-centric; even heuristic attempts to enforce spatial uniformity suffer from a sequential, two-stage nature that inherently decouples distribution optimization from geometric refinement, preventing an effective trade-off. In this paper, we propose DLOSAC, a local optimization framework that jointly incorporates both geometric consistency and spatial distribution information. We begin by constructing a Delaunay triangulation over candidate inliers to explicitly encode neighborhood topology. The geometric module introduces an edge-based consistency metric to evaluate geometric errors, while the spatial module derives distribution descriptors from triangle areas and edge attributes to quantify the clustering of inliers. Finally, we combine edge-based geometric constraints, spatial distribution regularization, and traditional point-level geometric constraints to improve the quality of geometric parameters. Extensive experiments on six publicly available real-world datasets, covering three types of geometric estimation tasks, achieve state-of-the-art performance.
[CV-113] SpaceCache: Intermediate Feature Reuse for Efficient Semantic Segmentation on Point Cloud Streams
链接: https://arxiv.org/abs/2610.21207
作者: Dongheon Yang,Beomseok Kang,Jae-Joon Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Point-cloud streams are central to autonomous driving and robotics, yet existing segmentation networks process every frame from scratch, although consecutive frames largely overlap. We observe that the intermediate features of point-cloud transformers remain semantically consistent across frames for the same spatial region, especially in deep stages. Building on this, we present SpaceCache, a training-free feature cache that reuses these features across frames. Tokens of previously observed regions skip a sequence of transformer blocks and take their cached features, only the remaining tokens are computed, and the full token layout is restored for dense prediction. Lightweight cache management keeps reused features reliable by limiting each entry’s age and evicting low-confidence entries. Unlike token merging, SpaceCache selects tokens at negligible cost and runs its cache operations once per encoder stage. On both outdoor and indoor benchmarks, such as nuScenes, Waymo, and ScanNet, with two transformer backbones of Utonia and LitePT-B, SpaceCache achieves up to 1.65x speedup with less than 1-point mIoU drop and consistently outperforms token-merging methods.
[CV-114] UltraDiff: Transferring High-Fidelity Priors to Compressed Latent Spaces for High-Resolution Image Generation
链接: https://arxiv.org/abs/2610.21202
作者: Jingjing Ren,Haitian Zheng,Haoyu Chen,Connelly Barnes,Zhifei Zhang,Yuqian Zhou,Zhe Lin,Lei Zhu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:High-resolution synthesis with Latent Diffusion Models (LDMs) faces a fundamental trade-off between computational cost and generative fidelity. This trade off is largely governed by the latent compression factor f (e.g. f8 denotes f=8 and f32 f=32). While aggressive compression (f32) significantly improves efficiency, it suffers from severe information loss and ambiguity compared to standard f8 models. We propose UltraDiff, a framework that reformulates this challenge as joint latent diffusion learning followed by marginal score distillation. We introduce MixTeacher, a dual branch DiT that models the joint f8-f32 distribution, utilizing shared attention to resolve the ambiguity of the compressed space with high fidelity f8 priors. Crucially, we propose vector field distillation in the native f32 space. Theoretically, this compels the student to learn the conditional expectation of the teacher trajectory, effectively filtering out compression artifacts. Across the evaluated 1K to 4K settings, UltraDiff approaches f8 level quality while preserving f32 efficiency. With 20 sampling steps on a single A100 GPU, it generates a 2K image in 6.2 seconds over 40 times faster than our f8 baseline.
[CV-115] ControlFlow3D: Distilling Multi-View Knowledge into Latent Flow Matching for Point Cloud Upsampling NEURIPS2026
链接: https://arxiv.org/abs/2610.21190
作者: Yuang Liu,Zhi Zuo,Zhengkai Zhao,Lirui Zhang,Pan Gao,Hao Feng,Zhengzhe Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at NeurIPS 2026
Abstract:Point cloud upsampling is essential for 3D applications but remains challenging due to sparse, under-constrained observations. Multi-view images can resolve geometric ambiguity, but existing multi-modal methods do not fully exploit visual cues. We propose ControlFlow3D, a framework that distills knowledge from pretrained foundation models into a lightweight generation network during training, enabling strong performance without multi-view images available at test time. Specifically, our method performs conditional flow matching in a structured latent space defined by a frozen 3D encoder. Multi-view image tokens and 3D latent tokens are jointly clustered into shared prototypes via unsupervised soft KMeans; the resulting prototypes guide the flow trajectory through velocity control and are aligned by a consistency loss that progressively transfers image knowledge into the base network. A Mamba-based geometry decoder then maps latent features to dense coordinates with linear complexity. We also construct ShapeNetPU, a 35K-object multi-modal benchmark that is 31 times larger than prior datasets. Extensive experiments across multiple benchmarks demonstrate the effectiveness of our method and its strong generalization capability.
[CV-116] VisualTrust: Video Reasoning under On-Screen Claim Framing ALT
链接: https://arxiv.org/abs/2610.21159
作者: Huachao Zhu,Junjielong Xu,Sicheng Li,Liang Yang,Mingtao Chen,Zhichao Hu,Li Jiang,Pinjia He
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 25 pages, 7 figures. Project repository: this https URL
Abstract:On-screen text can describe what a video shows, or it can frame what the video is supposed to mean. Current benchmarks treat both alike, testing whether a model can read the text rather than whether it should trust it. We study claim deference, multimodal large language models’ tendency to let an on-screen claim override conflicting visual evidence. We introduce VisualTrust, a counterfactual benchmark with 5,063 cases across 14 claim-fact classes. Each case pairs a question with three versions of the same video: one with no claim, one with an accurate claim, and one with a misleading claim that alters a single grounded fact. The option the misleading claim supports is the trap answer. Across seven open-weight and four proprietary models, mean accuracy falls from 90.1% without a claim to 51.9% with a misleading claim. Among initially correct responses, 42.8% flip to the trap answer and only 0.6% become other errors. Strong no-claim accuracy does not ensure resistance. We also introduce DEAR (Decoupled Evidence Arbitration and Reading), a training-free agent for video question answering that separates evidence arbitration from reading. An independent check reports conflicting evidence, a selector weighs it without a fixed modality priority and blinds what it withholds, and the reader answers from the resulting video in a fresh context. On 500 videos across three proprietary models, DEAR raises accuracy under misleading claims from 47.2–64.0% to 81.8–85.2% and reduces trap-answer selection from 34.4–50.8% to 11.6–15.0%, while preserving no-claim accuracy.
[CV-117] rajectory Prediction Method for person under WSN with Hierarchical Envelope Adversarial Domain Adaptation
链接: https://arxiv.org/abs/2610.21132
作者: Wenqian Jiang,Qifeng Sun,Xiaosong Gao,Xin Zou,Xingrui Ni,Yuan Hu,Yongming Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 24 pages, 7 figures
Abstract:Pedestrian trajectory prediction holds significant applications across autonomous driving, intelligent surveillance, and smart cities. However, the complexity and unpredictability of pedestrian interactions make this task a persistent challenge. Current models face two common limitations: 1) Consider only a single type of social coupling. This introduces redundant interactions. More critically, it fails to account for the varying nature of trajectory coupled across different scenarios and between different pedestrians. Consequently, models cannot deeply adapt to diverse scenes. 2) Inadequate domain shift handling. Although very few methods address domain shift, they rely on statistical criterion-based domain distribution alignment. Such approaches exhibit strong dependency on predefined statistical metrics. This leads to significant limitations in complex and dynamic environments. To address these problems, a Hierarchical Envelope Adversarial Domain Adaptation with Loosely-Tightly coupled model (HEADA-LTCTPM) is proposed. First, an envelope sample construction mechanism was proposed. It constructs envelope samples and extends them into graph structures. Secondly, an adversarial domain adaptation module was proposed. This integrates both local and global domain adaptation strategies. It dynamically adapts to diverse coupled relationships across scenarios. These innovations collectively enhance prediction accuracy and robustness. In the experimental section, two representative public datasets were employed for validation and comprehensive comparisons with six relevant baseline algorithms were conducted. Results demonstrate that our model achieves significantly higher accuracy compared to existing methods, with the ADE and FDE metrics reduced by 17.6% and 19.1% respectively. The time overhead meets practical requirements, which verifies the effectiveness of our key innovations.
[CV-118] Sparse Attention for Video Generation Acceleration via Growing Sparsity and Reduced Search
链接: https://arxiv.org/abs/2610.21129
作者: Yun Dai,Yuyun Chen,Jiarui Wen,Huiping Zhuang,Cen Chen,Ziqian Zeng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Diffusion Transformers (DiTs) dominate video generation but suffer from high inference cost due to the quadratic complexity of self-attention. Sparse attention mitigates this by skipping unnecessary computation in the attention matrix, but existing methods apply a fixed sparsity ratio throughout denoising and struggle to accurately determine which computations to skip. To address these limitations, we analyze attention dynamics through attention entropy and mean attention distance, and find that attention becomes increasingly concentrated across denoising steps. We further observe that blocks selected early continue to capture a large fraction of the total attention weight at later steps. These findings suggest two acceleration opportunities: the sparsity ratio should grow as denoising progresses, and earlier search results can be reused to reduce the search space. Based on these findings, we propose Growing Sparsity and Reduced Search (GSRS), a training-free framework that retains more computation early and progressively reduces it later, while restricting subsequent searches to an early candidate set to reduce search overhead. GSRS adopts Top-p selection for adaptive sparsity control, supported by a precise evaluation scheme that computes exact attention weights within the FlashAttention computation at negligible overhead. Experiments on HunyuanVideo and Wan 2.1 validate that GSRS achieves 1.55 \times --2.21 \times speedup while maintaining generation quality comparable to full attention.
[CV-119] ResStreamGS: Residual-Sensitive Dual-Stream Gaussian Splatting for Mobile Volumetric Video Streaming
链接: https://arxiv.org/abs/2610.21121
作者: Haoran Liu,Di Wu,Wang Luo,Hengyuan Na,Miao Hu,Guocong Quan
类目: Graphics (cs.GR); Computer Vision and Pattern Recognition (cs.CV)
备注: 11 pages,9 figures
Abstract:Real-time mobile volumetric video enables immersive free-viewpoint experiences, but delivering high-quality 3D content under constrained bandwidth and computation remains challenging. Recent 3D Gaussian Splatting (3DGS) systems, such as V ^3 , enable efficient mobile streaming by encoding time-varying Gaussian attributes with commodity video codecs and hardware decoding. However, existing pipelines treat all Gaussians as frame-varying signals, which is poorly suited to full-scene content where large portions remain temporally stable. Our empirical study shows that frame-wise adaptation and video coding of all Gaussians cause quality degradation in complex full-scene content, particularly in temporally stable regions. To address this issue, we propose ResStreamGS, a residual-sensitive dual-stream framework that processes Gaussians according to their temporal behavior. ResStreamGS identifies residual-sensitive primitives through lightweight probe adaptation and reuses stable Gaussians within each Group of Pictures (GOP). Stable Gaussians are transmitted once per GOP, while only residual-sensitive Gaussians remain in the frame-wise hardware-decodable video stream. Experiments on complex full-scene volumetric videos show that ResStreamGS improves post-codec PSNR by 7.51–8.35~dB over a controlled V ^3 -style all-Gaussians baseline, reduces amortized payload by up to 56.5%, and achieves 30.883~FPS on a Qualcomm Snapdragon 778G platform.
[CV-120] Spectral-Spatial Interpretation NEURIPS2026
链接: https://arxiv.org/abs/2610.21114
作者: Haotian Ma,Ruqi Yang,Philip Townsend
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 24 pages, 4 figures. Accepted at NeurIPS 2026 (Poster)
Abstract:Post-hoc image explanations usually show where a model looks using spatial heatmaps or indicate which channels matter using global channel scores. These separate views are limiting for multispectral data because channel relevance is often spatially localized and physically meaningful. We introduce Spectral-Spatial Interpretation (SSI), a full-grid attribution method over position-channel players (g,c). For each spatial position, SSI fixes the observed surrounding context, defines a conditional game over the channels at that position, and applies the Shapley value to this local game. The result is a full spectral-spatial attribution matrix rather than a single heatmap or a global channel ranking. SSI has a clear local interpretation. It satisfies Shapley fairness within each position, and aggregating its channel attributions exactly recovers the position-level contextual contrast obtained by replacing that position with the baseline. SSI also differs from context-averaged Owen-style hierarchical accounting, which uses the same position-channel grouping but averages over external spatial coalitions. On HLS Burn Scar, SSI outperforms an MC-Owen estimator across retained attribution metrics while using only about 6% of its evaluation budget. Across CUB-200, EuroSAT, and HLS Burn Scar under both per-channel mean and zero baselines, SSI achieves the best joint deletion and insertion AUC over dense, flat, and rank-1 baselines, yields the strongest spectral hit rates and channel deletion, and leads spatial localization on both annotated datasets. Code is available at this https URL.
[CV-121] Can World and Action Modeling Benefit Each Other under Unified Causal Training?
链接: https://arxiv.org/abs/2610.21095
作者: Yue Su,Bokai Lin,Xinkai Wang,Chubin Zhang,Xiangnan Wu,Kuan Zhang,Weiyang Jin,Jiwen Yu,Yiming Li,Cewu Lu,Lixin Yang,Xihui Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:World and action modeling are closely related, yet their reciprocal benefits under unified causal training remain unclear. We introduce OneWM, a simple framework for unified causal training that treats actions as both conditions for world prediction and targets for policy learning within a shared video-action architecture. Using this framework, we investigate whether learning forward dynamics conditioned on specified actions improves action learning, and whether learning to generate actions in turn improves action-conditioned world modeling. We use LIBERO, LIBERO-Plus, and RoboCasa365 as embodied testbeds to evaluate both policy learning and world modeling, complemented by real-world policy experiments. Learning the visual consequences of specified actions improves policy performance and generalization across changing scenes. In turn, action learning improves multi-view world-prediction fidelity under the same architecture. Together, these results answer both questions positively, supporting reciprocal benefits between world modeling and action generation under unified causal training.
[CV-122] RoboCloth: A Large-Scale Real Cloth Material Dataset for Neural Reflectance Reconstruction
链接: https://arxiv.org/abs/2610.21090
作者: Zhen Li(Simon Fraser University),Haoran Qin(Westlake University),Francesco Di Sario(University of Torino and Simon Fraser University),Daniel Rebain(University of British Columbia and Wayve),Andrea Tagliasacchi(Simon Fraser University, Wayve and University of Toronto)
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: Project page: this https URL ; Code: this https URL ; Dataset: this https URL . 16 pages including supplementary material
Abstract:We present RoboCloth, a dataset of 500 real cloth materials captured under flexible, calibrated camera and lighting configurations, where both are robotically controlled during acquisition. Each material is imaged in approximately 580 high-dynamic range frames spanning a wide view-light hemisphere; camera poses are refined via structure-from-motion and aligned to the robot frame, yielding per-material sparse surface geometry and structured per-point reflectance observations ready for neural material training. We demonstrate the value of this data through a two-stage neural reconstruction pipeline: Stage 1 trains a shared MLP decoder on our corpus to learn cloth material priors; Stage 2 then freezes the decoder and fits a dense 2048x2048 latent texture to each target material. Evaluating on held-out materials from our own dataset and on the independent UBO2014 benchmark, we show that a decoder trained on our captures outperforms decoders trained on existing measured reflectance datasets and analytic PBR fitting, transferring to unseen acquisition systems without modification. The dataset, calibration files, observation tensors, and train/test splits are released under CC BY 4.0 on the Hugging Face Hub.
[CV-123] SIEVE: Overcoming Topological Obstruction in Equivariant Self-Supervised Learning NEURIPS
链接: https://arxiv.org/abs/2610.21087
作者: Jongann Lee,Sun Woo Park,Yun Young Choi
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS) 2026
Abstract:We identify a theoretical incompatibility between 3D point cloud masked autoencoding and SE(3)-equivariant networks. We prove that standard masking triggers a topological collapse, effectively forcing the model to hallucinate orientation without a reference frame. To resolve this, we introduce SIEVE (Scalar Invariant Extraction Via Equivariance), an SE(3)-equivariant self-supervised framework that compresses point cloud geometry into invariant scalar embeddings. At its core is the spherical projection mask, which preserves global pose while obfuscating local semantics, sidestepping the obstruction by construction. Experiments confirm that SIEVE circumvents the predicted collapse where constant and noise masks fail. On downstream point cloud tasks, the resulting embeddings improve over coordinate-based baselines and hand-crafted geometric descriptors, with the largest gains on generalization to unseen shape categories.
[CV-124] RoPE-Flow: Monocular Vision-based Robot Pose Estimation via 2D Graph-Conditioned 3D Flow Matching
链接: https://arxiv.org/abs/2610.21086
作者: Liangjing Shao,Wanhao Liu,Zhiwei Fang,Rulin Zhou,Quanlu Zhang,Changjing Liu,Beilei Cui,Yiming Huang,Hongliang Ren
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 19 pages, 13 figures. Under Review
Abstract:Camera-to-robot pose estimation from a monocular image is significant for vision-based robot navigation and control. However, recovering 3D structural features from 2D observations remains a challenging task. Existing camera-to-robot pose estimation methods either rely on time-consuming iterative rendering or involve sophisticated 2D-3D registration which suffers from projection ambiguity, accumulated errors for distal joints and imbalanced learning for different joints. In this work, we propose RoPE-Flow, an end-to-end flow matching-based framework for camera-to-robot pose estimation. From the predicted 2D heatmaps, we propose a Keypoint-centered Feature Extractor to obtain positional features along with contextual representations from keypoints-centered multi-scale patches. Moreover, we introduce Graph-guided Condition Generator, which embeds a learnable adjacency map to generate condition for flow matching based on correlation among keypoints. Besides, we introduce Bone Length Loss and Ray-Depth Loss for topological consistency and scale stability, with joint-wise weighting to mitigate varied prediction difficulty across joints. The proposed method outperforms state-of-the-art methods for pose estimation across different types of robots with mutiple cameras. Besides, we establish a novel cross-scene benchmark to demonstrate the outstanding performance of RoPE-Flow on zero-shot cross-scene robot pose estimation. Meanwhile, animated visualizations are shown on the website: this http URL
[CV-125] Audio-Visual Dog Identification and Tracking in Real-World Videos
链接: https://arxiv.org/abs/2610.21011
作者: Aryahvishwa Babu,Kingsley Jin,Lillian Wang,Tuan M. Dang,Theron S. Wang,Jeorge Joshi,Annie Cao,Kenny Q. Zhu
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Identifying individual dogs from barks in real-world videos remains challenging and largely unexplored. Solving this problem would enable long-term behavioral analysis, veterinary monitoring, and animal welfare monitoring by attributing vocalizations and behaviors to the correct individual. Real-world videos are challenging because multiple dogs may bark simultaneously, become occluded, or move in and out of view. Prior studies show that canine vocalizations and visual appearance contain discriminative identity information, but these modalities have largely been studied independently. We present a multimodal dog-fingerprinting framework combining visual identification, multiple-object tracking, and canine vocal fingerprinting to associate bark events with individual dogs while maintaining persistent profiles. The framework synchronizes bark events with tracked dogs and updates profiles using visual appearance and vocal characteristics, enabling identification despite occlusion, overlapping vocalizations, and off-screen barking. We also introduce a benchmark of 46 real-world videos containing 213 manually annotated bark events. Our framework correctly localizes the active visible dog in 61.83% of visible events at IoU \geq 0.30 and achieves a 91.33% F1 score for visible versus off-screen barking detection.
[CV-126] Mind the Tail: Data Selection Through Sparse Concepts
链接: https://arxiv.org/abs/2610.21007
作者: Ziheng Zhang,Neha Hulkund,Colin Lee,Zheda Mai,Hui Shen,Ying Ding,Tanya Berger-Wolf,Sara Beery,Wei-Lun Chao,Jianyang Gu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Data selection reallocates training support across visual concepts, including objects, attributes, color, texture, and backgrounds. Different allocation strategies may produce varied downstream model behavior, especially for tail concepts supported by only a few samples. Discarding them can hurt test-time robustness, while overemphasizing samples lying on category boundaries may introduce unstable training signals. This work explicitly investigates this relationship across fine-grained, spurious-correlation, and general image classification, as well as multimodal pretraining. Specifically, we use sparse autoencoders (SAEs) to extract concept activations from individual images. During selection, we accumulate the dataset-level concept support for each added sample. Applying diminishing returns to this distribution down-weights well-represented concepts and shifts the focus to the distribution tail. We then incorporate local support to find samples that reliably strengthen under-supported concepts. Empirical results show consistent improvements over strong data selection baselines across all benchmarks. We also reveal the connection between concepts our method enhances and the induced behavioral changes, including improved robustness against out-of-distribution and spurious contexts. This work establishes concept support, beyond individual sample importance, as a useful principle for data-efficient and robust learning.
[CV-127] EigDiff: Eigensample Diffusion - or how to significantly reduce training dataset size with marginal performance variation?
链接: https://arxiv.org/abs/2610.20999
作者: Andrei Jarca,Vlad Hondru,Florinel-Alin Croitoru,Radu Tudor Ionescu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Diffusion models are notoriously expensive to train, requiring large-scale datasets to learn how to gradually reverse the noise from images over multiple small denoising steps. Current approaches focus on distilling the number of diffusion steps, neglecting potential redundancies of the training data. In this paper, we explore the potential of harnessing eigensamples to train diffusion models on significantly smaller datasets, while preserving model performance. To mitigate the sensitivity to affine transformations of eigenimages, we harness the latent space of a pre-trained CLIP model. We next compute latent eigenvectors by performing principal component analysis (PCA) in CLIP space. Further, we employ eigenvectors as reference points to guide the selection of a representative training set comprising comprising both exemplars (images whose embeddings are closest to eigenvectors) and diverse images. Our comprehensive experiments with different diffusion models applied on multiple datasets indicate that our approach (EigDiff) can reduce training data by up to 14x and accelerate subset selection by up to 1607x, with marginal performance differences.
[CV-128] DeSpatial: The devil of Agent ic Spatial Reasoning Lies in 3D Itself
链接: https://arxiv.org/abs/2610.20990
作者: Yifei Liu,Xiaolong Zhou,Ziyang Gong,Yifan Zhou,Yuanyuan Gao,Yuning Gong,Yuchen Yang,Hongjie Zhang,Xue Yang,Xiao Sun,Zhihang Zhong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Spatial agents have access to reconstruction and segmentation tools, yet inconsistent predictions and fragmented object identities can prevent their outputs from becoming reliable 3D scene information. A simple oracle experiment exposes this perception bottleneck: replacing tool-predicted perception with ground-truth 3D bounding boxes raises an agent to nearly full score across five object-centric VSI-Bench tasks, while retaining the same reasoning tools. Motivated by this finding, we believe the devil of spatial reasoning lies in 3D itself, and introduce DeSpatial(Devil of Spatial), which explicitly decomposes spatial understanding into perception and reasoning. DeSpatial has a shared 3D perception tool to reconcile independent perception outputs into a unified, metric, gravity-aligned 3D scene. This perception tool has two core components: a) a joint segmentation–depth co-refinement module that consolidates masks of the same object and removes projection artifacts caused by inconsistent segmentation and depth within a single frame; b) a cross-view instance fusion module that consolidate voxel-overlapping instances but not allowed to merge same-frame instances. This fusion module additionally fuses spatially disjoint fragments caused by insufficient multi-view observations, by checking their 3D structure and source mask appearance. After receiving the complete and unified 3D scene information, the agent chooses reasoning or computational tools to solve the spatial problem. DeSpatial achieves leading performance on VSI-Bench, MindCube-Tiny, and ViewSpatial, with additional long-video results on VSI-SUPER, revealing the bottleneck of spatial understanding lies in a precise 3D perception itself.
[CV-129] Do As I Do Not As I Say: On-Policy Distillation Inherits the Teacher in Vision-Language Models
链接: https://arxiv.org/abs/2610.20989
作者: Frank F. Yang,Kefei Yan,Yiren Wang,Yiyang Wang
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 27 pages, 7 figures, 21 tables. Code: this https URL
Abstract:Improving the reasoning capabilities of vision-language models (VLMs) often relies on post-training with supervision from stronger models. Two common strategies are supervised fine-tuning (SFT), which trains on teacher-generated demonstrations, and on-policy distillation (OPD), which provides teacher feedback on trajectories generated by the student itself. While OPD can be more data-efficient by supervising student-generated trajectories, it is unclear how the teacher’s supervision at each state affects the student’s learned behavior. To study this question, we conduct a controlled comparison of OPD and filtered SFT across data budgets and tasks where the teacher exhibits different levels of reliability, and analyze how these differences shape student accuracy and completion behavior. Across the two tasks, we find that OPD is effective only when the teacher remains reliable on student-generated trajectories. On ChartQA, where the teacher completes 99% of its solutions, OPD trained on only 10% of the data comes within 1.4 accuracy points of full-data SFT. In contrast, on Geometry3K, where the teacher fails to complete 22% of its solutions, OPD underperforms SFT by 7.9 points, with most of the gap explained by incomplete responses. More strikingly, starting OPD from an SFT student reduces its completion rate from 76% to 62%, effectively erasing the completion behavior learned through SFT. Filtering teacher demonstrations for completeness alone recovers much of SFT’s advantage. These results show that on-policy supervision can transfer not only teacher knowledge but also persistent failure modes on student-visited states. Code is available at this https URL.
[CV-130] PiCo: Physics-Informed Concept Probing for Geo-Foundation Models
链接: https://arxiv.org/abs/2610.20973
作者: Yihang She,Andrew Blake,Jianbo Qi,Srinivasan Keshav
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Geo-foundation models (GFMs) are typically evaluated by downstream task predictive performance, revealing little about which physical features they are sensitive to or use in prediction. This limits developers’ ability to test hypotheses about model behaviour and users’ ability to select models for new tasks. We propose Physics-Informed Concept Probing (PiCo), a post-hoc framework for explaining GFM predictions. PiCo identifies concept vectors in a GFM’s embedding space by generating synthetic multispectral satellite observations from a 3D radiative transfer model coupled with a game engine. These vectors reveal physical sensitivities that inform domain-based hypotheses about downstream performance. Projecting embeddings onto these concept vectors produces low-dimensional representations for quantifying concept importance in making predictions. We evaluate PiCo on three GFMs with different input modalities using seven physical concepts covering forest leaf optics and canopy structure. In ground-cover estimation, RGB restriction reduces dry-matter sensitivity and disproportionately degrades predictions for dry vegetation and bare ground, supporting a hypothesis derived from sensitivity analysis. For hierarchical forest type and tree species classification, concept representations retain 70.0-93.8% of full-embedding accuracy. Models capture similar physical contrasts between classes, such as structural differences between broadleaf forest and cleared ground and leaf-water-content differences between oak and beech, yet differ in the strength of these contrasts and the concepts driving classification. PiCo thus provides a physical basis for hypothesis testing and model assessment, with potential applicability beyond Earth observation wherever forward models enable controlled simulation and pretrained embeddings are accessible for probing.
[CV-131] MSP: Modality Self-Play for Training Multimodal LLM s Without Paired Cross-Modal Data NEURIPS2026
链接: https://arxiv.org/abs/2610.20965
作者: Robert Moseley,Md Kaykobad Reza,Ameya Patil,Edward Ayrapetian,Salman Asif
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages, 4 figures, 16 tables. Accepted at NeurIPS 2026
Abstract:Large multimodal models have achieved remarkable progress by scaling model capacity and training on massive paired multimodal data. However, this paradigm introduces a fundamental bottleneck: aligned multimodal datasets are expensive to curate, difficult to scale across diverse modality combinations, and increasingly constrain further progress. We introduce Modality Self-Play (MSP), a training framework for multimodal foundation models based on proxy-mediated modality composition. Instead of relying on paired cross-modal data, MSP uses anchor-supporting proxy textual descriptions of unseen modalities as semantic bridges during training. Given an observed modality (e.g., image or audio), the model generates proxy descriptions for complementary unseen modalities, which are then encoded and aligned with the observed modality through learned bridges, enabling cross-modal interaction without real paired non-text supervision. Our central hypothesis is that proxy text, combined with contrastively trained pretrained encoders, provides sufficient semantic structure for multimodal composition to emerge. We show that MSP enables zero-shot modality composition: models trained only with a real anchor modality and accompanying proxy descriptions can integrate and reason over multiple modalities at inference time despite never observing real paired non-text data during training. Empirically, MSP achieves state-of-the-art or competitive performance on multiple multimodal benchmarks. Ablation studies further demonstrate that learned bridges improve alignment, while anchor-supporting proxy descriptions enable effective cross-modal composition. MSP removes the need for real paired non-text (e.g., audio-vision) supervision, while continuing to rely on modality-text pairs, pretrained contrastive encoders, and modality-distribution bridge calibration. Comments: 21 pages, 4 figures, 16 tables. Accepted at NeurIPS 2026 Subjects: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV) ACMclasses: I.2.7; I.2.10; I.5.4 Cite as: arXiv:2610.20965 [cs.AI] (or arXiv:2610.20965v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2610.20965 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-132] Line-Augmented cuVSLAM: Robust Visual Odometry in Low-Texture Environments
链接: https://arxiv.org/abs/2610.20958
作者: Dara Molloy,Terry Wu,Wonju Lee,Abhilash Nand Kumar,David Adley,Conor Ferguson,Hesam Rabeti
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 8 pages, 4 figures
Abstract:Autonomous mobile robots in warehouses and production facilities operate in precisely the environments that defeat point-feature visual odometry: featureless aisles, uniform walls, repetitive racking and specular floors. Such environments are, however, rich in straight-edge structure: rack uprights, door frames, wall junctions and cabinet outlines present high-contrast linear edges stable across wide viewpoint changes. A line-feature extension to cuVSLAM, a CUDA-accelerated visual-SLAM system, is presented that recovers this signal. Lines are parameterised as Plücker coordinates for geometric transport and as the minimal 4-DOF orthonormal representation of Bartoli and Sturm for optimisation; a Cauchy-robust point-to-line residual with analytic Jacobians is folded directly into cuVSLAM’s per-frame Gauss-Newton pose solve, and the same line state is jointly refined in the Schur-complement sliding-window bundle adjustment. An adaptive information weight causes lines to assume more of the estimation burden exactly as point features become scarce, and the extension adds residual blocks to existing solvers rather than introducing a second estimator, reducing exactly to the unmodified point pipeline when disabled. Evaluated in RGB-D mode under SE(3) Umeyama alignment with no scale correction, line features reduce mean absolute trajectory error by 38.5% over the 22 sequences of the TUM RGB-D Freiburg3 collection and by 30.1% over the 8 Replica room and office sequences, where median translation error falls by 38.9% and median rotation error by 66.4%, with zero tracking failures in any configuration tested.
[CV-133] Decision-FLIP: Budget-Free Structured Pruning for Gaussian Occupancy Models
链接: https://arxiv.org/abs/2610.20917
作者: Jingxing Li,Anurag Paul,Subhasis Das,Cindy Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Occupancy predictors rely on large persistent banks of Gaussians or learned queries whose size is fixed by hand. Common pruning proxies, such as opacity, activation frequency, or parameter magnitude, do not measure whether removing a row changes the model’s voxel decisions. We introduce Decision-FLIP, a budget-free compilation framework whose scoring and structure selection need only unlabeled calibration inputs. Exact Mean-FLIP scores each persistent row by the number of voxel argmax decisions that change when the row is removed from the readout; a guarded Otsu split of the log-score distribution selects the survivor count without a target and abstains when the split is unsupported; and the survivors are physically compiled into a smaller checkpoint, so camera sampling, attention, and aggregation shrink with the bank. For streaming banks, scores are integrated over each row’s lifecycle through warp, death, and refill. Across four Gaussian-bank model-dataset settings and one persistent-query transfer, the automatically selected cuts remove 12-79% of eligible rows and cut bank/query-dependent latency by 7-45% across settings with component profiles, with whole-model savings of up to 8.0% on the primary endpoints bounded by fixed-cost image, voxel, and sensor modules and peak-memory savings of up to 16.6%. Under matched one-epoch recovery controls, semantic mIoU changes by at most 0.18 point. Same- K controls show that the ranking, not only the bank size, matters: random removal collapses GaussianFormer-1 and OccFormer, while label-free Mean-FLIP matches or outperforms supervised Taylor ranking on all five settings where that control was run. The method applies when removable primitives keep a stable checkpoint-row identity; it is budget-free in survivor count, not parameter-free.
[CV-134] EgoJEPA: Beyond Retargeting for Dexterous Manipulation from Egocentric Videos
链接: https://arxiv.org/abs/2610.20904
作者: Meng Cao,Can Zhao,Jinjie Wei,Haonan Zhao,Yulong Zhang,Duo Han,Xiaofan Li,Ziwu Song,Rongtao Xu,Ruyi Gan,Hao Wang,Qian Wang,Daolin Ma,Ian Reid,Xiaodan Liang
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Egocentric human videos offer a scalable alternative to costly robot demonstra- tions, yet transferring their manipulation priors to robot control remains chal- lenging. The commonly adopted retargeting-based approaches convert human wrist and finger poses into embodiment-specific action commands through hand- crafted geometric mappings. However, such mappings require meticulous recal- ibration across different embodiments and risk propagating hand-tracking errors into downstream robot control pipelines. To this end, we propose EgoJEPA, a three-stage framework that learns embodiment-agnostic representations from ego- centric human videos without rule-based human-to-robot retargeting: 1) Human- action pre-training learns effect-aware latent actions through JEPA-style visual prediction conditioned on hand-motion differences; 2) Latent-action-based world modeling jointly predicts future visual dynamics and latent actions, which decou- ples human prior pre-training from robot-specific control parameterizations; 3) Embodiment-specific tuning adapts the pretrained model to diverse robots through lightweight action adapters. Experiments on five real-world manipulation tasks across three robot embodiments demonstrate the effectiveness and efficiency of EgoJEPA. It achieves a 74% average success rate under the out-of-domain evalu- ation, compared with 65% for the retargeting baseline, while reducing cumulative cross-embodiment wall-clock cost by 47.2%, from 743.2 to 392.7 hours.
[CV-135] Pixels to Tokens: Token Space Efficient Active Learning for Low-Budget Semantic Segmentation NEURIPS2026
链接: https://arxiv.org/abs/2610.20881
作者: Utku Cicek,Mahip Singh,Ziyao Shang,Kimathi Kaai,C Thomas,Pedro Pablo Guerrero Vela,Alexander Wong,Sirisha Rambhatla
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026. 19 pages, 6 figures
Abstract:While foundation models and vision transformers (ViTs) have redefined visual representations, current active learning (AL) strategies for these backbones largely operate at the patch or region level, where annotation requirements remain high due to extensive pixel labeling. Extending ViT-based AL to the pixel-level, low-budget regime is uniquely challenging: with as little as one pixel query per image, a method must simultaneously identify the most informative locations and train a reliable decoder head from an extremely sparse signal. We introduce TEAL (Token-space Efficient Active Learning), the first pixel-level AL framework built around ViT representations. TEAL uses a frozen ViT backbone and performs diversity-based selection directly on its native token lattice, avoiding the artifacts introduced by interpolating token embeddings to dense pixel resolutions. Our framework applies a two-level MaxHerding strategy over multi-layer token descriptors to select representative candidates, followed by margin-based uncertainty refinement on a finer decoder grid. Across CamVid, Cityscapes, ADE20K, and Pascal Context, TEAL consistently outperforms previous baselines under extreme label scarcity, improving over the strongest baseline by up to +24.36 mIoU with a single labeled pixel (px) per image, and reaching up to 97% of fully supervised performance with just 10 px. To rigorously establish the source of the gains, we further show that the improvements persist when all baselines are equipped with the same DINOv3 backbone, demonstrating that coupling token-space selection with uncertainty refinement on the decoder grid makes ViT token geometry an effective foundation for extremely low-budget active segmentation.
[CV-136] DefoliationCAM: An Edge-AI System for Real-Time Insect Defoliation Severity Assessment in Grapevines
链接: https://arxiv.org/abs/2610.20879
作者: Fuchen Li,Wei-Zhen Liang,Pin-Chu Lai,Xin Qiao,Jingqiu Chen,Haw Yen,Zonglin Yanga,Chi-En Chiang,Rudolph Avalos,Abbigail L. Rath
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Insect-induced defoliation threatens grapevine health and yield, yet severity assessment relies on subjective manual scouting. We developed DefoliationCAM (D-CAM), a solar-powered edge AI system that estimates canopy defoliation and transmits results via LoRa to a web dashboard. The pipeline uses YOLOX-Nano to localize grapevine cordons in 4K images and a MobileNet-based classifier with lightweight channel attention to classify 320 x 320 pixel patches. The 29,674 field-collected patches were labeled as background or one of four damage levels: no to negligible, mild, moderate, and severe. Merging moderate and severe classes reduced the class imbalance ratio from approximately 40:1 to 8.8:1. Synthetic augmentation increased macro-F1 from 0.768 to 0.838. The classifier outperformed ResNet-50 in macro-F1 with fewer parameters (1.08 M versus 23.52 M). Patch predictions were aggregated into a canopy-level Non-linear Global Defoliation Severity (NGDS) score and validated against an entomologist’s ratings of 86 held-out canopy images. Quadratic discriminant analysis identified boundaries of 9% and 29% separating negligible damage, mild damage requiring monitoring, and action-level damage warranting on-site scouting and insecticide treatment consideration. The 29% boundary aligned with the 25%-30% action threshold range for Japanese beetle management in grapevine. Using TensorRT FP16 on an NVIDIA Jetson Orin Nano, the pipeline processed each 4K image in approximately 1.5 s, with a complete operating cycle of 57-60 s. A five-week deployment of two field units demonstrated the feasibility of autonomous solar-powered operation, LoRa communication, and remote defoliation monitoring. D-CAM supports frequent, objective canopy damage assessment for targeted scouting and timely pest-management decisions.
[CV-137] MARS - A foundational map auto-regressor
链接: https://arxiv.org/abs/2610.20843
作者: Qi Zhang,Suvam Bag,Rupanjali Kukal,Lucas Thai,Jacob Kovarskiy,Mike Bianco,Xiaojie Guo,Raphael Tang,Mikael Figueroa,Rishi Madhok,Nikolaos Karianakis,Fuxun Yu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Map generation tasks feature extensive non-structural vectorized data (e.g., points, polylines, and polygons) and thus pose significant challenges to common pixel-wise generative models. Conventional approaches use multiple stages, first segmenting these features at the pixel level and then performing vectorized post-processing, with errors and complexity compounding at each stage. Motivated by the recent success of auto-regressive language modeling, we propose the first map foundation model, named Map Auto-Regressor (MARS), that is capable of generating both multi-polyline road networks and polygon buildings in a unified manner. For training MARS, we collected to our knowledge the largest multi-class map extraction dataset totaling 3.4M examples, which we call MAP-3M. Across four road and building datasets, MARS outperforms or matches the performance of multistage baselines. Additionally, we develop a “Chat with MARS” capability that enables interactive human-in-the-loop map generation and correction, supported by the auto-regressive nature of our end-to-end approach. We release our MAP-3M dataset and project demo page at (1) this https URL and (2) this https URL, respectively.
[CV-138] mporal Specification Guidance for Text-to-Video Generation
链接: https://arxiv.org/abs/2610.20822
作者: Avishree Khare,Hideki Okamoto,Bardh Hoxha,Georgios Fainekos,Rajeev Alur
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Text-to-video models often generate the visual concepts requested in natural language prompts but fail to satisfy their intended ordering, timing, and persistence requirements. We propose representing these requirements as specifications in a formal Temporal Logic, and introduce Temporal Specification Guidance (TSG), a training-free inference-time framework that turns temporal specifications into executable objectives for guiding video generation during sampling. TSG scores traces of visual concept evidence using differentiable quantitative semantics and uses the score’s gradients to construct a specification-guided denoising direction. We instantiate this framework with video-text attention traces, efficient differentiable scores for Metric Temporal Logic, and gradient-derived attention biases. This instantiation requires neither an external concept classifier nor backpropagation through the video transformer. Its attention-steered forward pass yields a denoising direction that augments classifier-free guidance (CFG). On a benchmark of 220 prompts spanning 11 categories of temporal constraints, including ordered object entry and state transitions, this instantiation improves specification satisfaction accuracy over CFG by 9.5-20 percentage points across five models, with minimal memory overhead. These results demonstrate the potential of temporal specifications and differentiable satisfaction objectives as a flexible interface for controlling video generation.
[CV-139] Intern Lumina U2: A Multi-Codebook Diffusion Large Language Model for Omni-Visual Understanding and Image Generation
链接: https://arxiv.org/abs/2610.20818
作者: Qian Yu,Zongyang Qiu,Juncheng Yan,Hongliang Li,Qi Qin,Keqi Wang,Siqi Luo,Junkui Ju,Yapeng Mi,Tiantian Lu,Tian Liu,Pengyu Xie,Chenhao Ji,Minghui Zhang,Yiang Zhao,Siyu Yan,Haiying Xu,Yanbo Wang,Jianghan Shen,Yuandong Pu,Jiayi Lei,Borui Zhang,Jinxin Shi,Bo Zhang,Yuhang Zang,Shenglong Ye,Lixin Gu,Hongjie Zhang,Linghan Liu,Xinyuan Chen,Yuewen Cao,Lihe Zhang,Xiaohong Liu,Junjun He,Qiangwei Liu,Qi Zhang,Yu Cheng,Tianfan Xue,Hui Xiong,Chang Xu,Yu Qiao,Yihao Liu,Bin Fu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 47 pages, 12 figures, 22 tables
Abstract:Unified multimodal generation covers discrete, continuous, and hybrid paradigms. Discrete approaches with single-codebook visual tokenizers align naturally with text modeling but have limited representational capacity. Continuous and hybrid approaches retain richer visual representations, but they model text and images in heterogeneous spaces with different training objectives, potentially introducing optimization conflicts between modalities with misaligned representation spaces. These choices expose a central question: how to efficiently model high-capacity visual representations alongside textual content in a shared optimization space within a single backbone. We present Intern Lumina U2, to our knowledge the first diffusion language model to represent multi-codebook visual tokens in a fully discrete space. Each visual position is represented by a group of codes from complementary codebooks, increasing its representational capacity while preserving compatibility with discrete text modeling. At the input, the embeddings are concatenated at each position; at the output, a lightweight multi-codebook autoregressive head predicts the codes within each position sequentially, while the diffusion backbone processes all positions in parallel. Built on the unified 4D representation of AToken, Intern Lumina U2 supports image, video, and 3D inputs within a single backbone and achieves leading performance among unified models in visual understanding, image generation, and image editing. We further build a full-stack training, inference, and evaluation system on both Ascend and NVIDIA platforms to verify the cross-platform effectiveness of our model. We also improve the inference efficiency through a stability-aware sampling strategy, which combines confidence-plateau-guided unmasking approach with drift-triggered cache refreshing mechanisms, accelerating image generation by 4.45x.
[CV-140] RIGen: Beyond Action Following with Reciprocal Interaction in Generative Gameplay
链接: https://arxiv.org/abs/2610.20811
作者: Mengfei Li,Yi Zhang,Xingtong Ge,Lunjie Zhu,Fangyu Lin,Peng Li,Guanglu Song,Yu Liu,Wei Xue,Wenhan Luo,Yike Guo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Generative gameplay requires more than immediate responses to player this http URL should have persistent consequences, while the generated world should act according to its own state and rules, shaping the player’s situation and subsequent choices. Video generation captures local dynamics, but limited history, partially observed game state, and the need for contextual behavior decisions make such interactions difficult to sustain through audiovisual prediction alone. We introduce \rig (Reciprocally Interactive Generation), a framework for reciprocal interaction in generative gameplay. \rig combines persistent state and configurable rules, contextual interaction reasoning, and action-conditioned audiovisual generation. Together, these components make interaction consequences and actor behavior jointly determine generated content while maintaining consistency between game state and visible outcomes. We construct a Xonotic gameplay dataset to train an audiovisual generator for learning dynamics and interactions. Further post-training enables \rigarena, an interactive generative first-person shooter (FPS) prototype. Experiments demonstrate more reliable combat outcomes and reciprocal interaction between the player and the generated world. Project page: this https URL.
[CV-141] Blame Errant Words Not the Canvas: Defect-Aware Spatial Credit Assignment for Visual Text Generation
链接: https://arxiv.org/abs/2610.20794
作者: Sujie Hu,Huaisong Zhang,Pujing Yang,Qian Wang,Xiaomin Li,Xinhang Leng,Chen Li,Jing Lyu,Xiu Li,Ling Pan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent reinforcement learning approaches have improved visual text generation, yet generated images still suffer from incorrect wording and malformed glyphs. We identify two coupled limitations: recognition-based rewards often overlook visible glyph defects, while whole-image scalar feedback obscures the uneven quality of individual text regions. These limitations motivate our central principle: \textitBlame errant words, not the canvas , to effectively preserve local quality distinctions. First, we introduce \mathbfTextDART , a defect-aware region-targeted reward model that localizes individual text regions and independently evaluates their semantic correctness and glyph quality, providing fine-grained, region-precise feedback. Building on TextDART, we propose \mathbfSCANNER , which turns these regional judgments into spatially differentiated policy feedback. It normalizes rewards for each text target across sampled images and casts the resulting regional advantages onto the corresponding latent tokens. Extensive experiments on existing benchmarks and our newly established \textbfPolyText-Bench demonstrate that our approach achieves state-of-the-art performance in text accuracy and glyph fidelity. Project page: this https URL
[CV-142] Reason Uni: Reasoning -Augmented Reconstruction Advances Unified Multimodal Models
链接: https://arxiv.org/abs/2610.20781
作者: Feize Wu,Zhiyuan Yan,Yuhuan Hong,Yuwei Niu,Zongjian Li,Zheyuan Liu,Bin Lin,Jinbo Yan,Jie Qin,Xudong Mao,Limeng Qiao,Guanglu Wan,Li Yuan
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 27 pages, including appendix
Abstract:Image-to-text (I2T) understanding and text-to-image (T2I) generation are dual tasks in unified multimodal models (UMMs). Reconstructive post-training couples these tasks, but existing reconstruction-based methods remain perception-centric and provide limited supervision for reasoning about structure, intent, and transformation. We propose ReasonUni, a reasoning-augmented reconstructive post-training framework that extends this coupling beyond perception. Given an original image and an editing instruction, the UMM first reasons about what and how to edit and describes the target state, then generates the corresponding target image conditioned on this understanding output. To jointly optimize both modules, we introduce TreeGRPO, a tree-structured reinforcement learning method that samples multiple reasoning paths and candidate images per path, then aggregates visual rewards to update both modules under a unified objective. Experiments across multiple benchmarks demonstrate consistent gains in both understanding and generation, showing that reasoning-augmented reconstruction yields genuine mutual gains for UMMs.
[CV-143] Purify Before You Align: Improving Representation Alignment for Diffusion Models
链接: https://arxiv.org/abs/2610.20761
作者: Yingheng Wang,Yaoqiang Li,Yaqin Wu,Junwen Bai,Jiatao Gu,Christopher De Sa,Volodymyr Kuleshov
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Code and checkpoints are on the project page: this https URL
Abstract:Representation alignment improves and accelerates diffusion training by encouraging intermediate features to match a pretrained visual encoder. However, standard alignment matches pretrained representations without distinguishing what information may be more useful for generation. We introduce local appearance purification (LAP), which improves generation quality and makes encoders such as MAE more effective alignment targets by removing locally predictable content and aligning to the residual. LAP modifies only the alignment target, and requires no changes to diffusion training or inference. Through controlled interventions within a fixed encoder, we show that effective alignment depends on what information the target emphasizes, not simply how much image information it contains. Our theory gives a variational interpretation of standard alignment and, for fixed local predictors, a sum of token-wise information bounds conditioned on local appearance. To understand why purification benefits some encoders more than others, we derive an exact constant-readout test that measures how well the target can be matched without using the diffusion model’s hidden state. LAP improves generation quality across five pretrained representations (MAE, AIMv2, CLIP, DINOv2, DINOv3), with the largest gain for MAE: on ImageNet-256 with SiT-B/2, it reduces FID from 53.1 to 42.3 . Its gains persist with stronger alignment recipes and tokenizers. Computed before diffusion training, the constant-readout score predicts the benefit of purification (Spearman \rho=0.94 across 13 encoders).
[CV-144] WorldCast: Exploring Intelligent Visual Generation in World State Prediction
链接: https://arxiv.org/abs/2610.20731
作者: Wenhao Yan,Haodong Li,Xiaotong Li,Yaxin Luo,Zhengrong Yue,Jialv Zou,Xu Huang,Haojin Yang,Nan Jia,Yan Feng,Peng Pei,Xunliang Cai
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project Page: this https URL
Abstract:Humans interpret multimodal observations of the world and forecast how it will evolve. Recent multimodal generative systems have demonstrated similar capabilities: they can integrate multimodal context and synthesize visual outputs that depict changes in the world. To characterize this emerging trajectory, we organize model capabilities into a four-level hierarchy and discover that higher-level tasks remain relatively underexplored despite model progress. To bridge this gap, we explore model performance on multi-step world state prediction, a high-level task that requires models to forecast successive future world states from multimodal observations and diverse references. Such a task requires models to possess entangled understanding and generation capabilities, as well as extensive world knowledge to translate human action sequences, physical transformations, and abstract visual reasoning into coherent visual outputs. We introduce WorldCast, a benchmark comprising 320 manually designed cases across eight categories, and evaluate 12 representative models. The conducted experiments yield three key insights: (a) advances in frontier models are rapidly pushing the boundaries of intelligent world modeling, with the GPT-Image-2.5 System scoring 72.8 and demonstrating early-stage visual state prediction capabilities on real-world tasks; (b) WorldCast reveals substantial gaps between models: even Nano Banana Pro reaches only 42.8 and the best open-source model scores 20.1, with deficits in handling complex context, entangled references, and reasoning-heavy tasks; © strong generative models effectively draw on the capabilities of their underlying language models. These findings highlight the need for multimodal generative models that are capable of dynamic planning and reasoning while also being deeply unified to faithfully simulate real-world states.
[CV-145] Orbia: Do Generated Videos Form a Consistent 3D World?
链接: https://arxiv.org/abs/2610.20693
作者: Dongyue Lu,Guian Fang,Hur Lim,Yonggang Wu,Yanrong Wang,Hongshan Chen,Shengju Qian,Yue Liao,Wei Chow,Lingdong Kong,Tianxin Huang,Yu Yang,Hongji Yang,Junchao Huang,Ao Liang,Yihao Wang,Zihan Wang,Rong Li,Hanlin Chen,Xin Wang,Wei Tsang Ooi,Mike Zheng Shou,Shuicheng Yan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Technical Report; 70 pages, 43 figures, 12 tables; Project page at this https URL Code at this https URL Dataset at this https URL
Abstract:Video generators serve as interactive worlds for exploration and simulation, yet realistic and visually appealing frames do not necessarily depict a consistent 3D world. Existing benchmarks assess selected geometric properties or use vision-language models to judge memory and consistency, leaving systematic 3D evaluation underexplored. We introduce Orbia, the first benchmark dedicated to a systematic and comprehensive evaluation of 3D consistency in interactive world models. Through controlled exploration and revisit, we organize the evaluation into three dimensions: input preservation, generated-content persistence, and 3D self-consistency. For input preservation, registered RGB-D references and annotated anchors support appearance, geometry, and spatial stability measurements. For generated-content persistence, we compare revisits to regions absent from the input image with their first generated observations to test whether models preserve what they create. For 3D self-consistency, cross-view depth and feature correspondence, regional scale agreement, and held-out Gaussian rendering test whether generated views describe the same 3D scene. Minute-long exploration and revisits further test long-term consistency. Our construction and annotation pipeline brings together public datasets, self-captured scenes, and simulation, yielding 720 instances across indoor and outdoor environments and six exploration-and-revisit protocols. Evaluation spans 27 text-, action-, and camera-conditioned models through native-interface adapters. Our Orbia benchmark provides a systematic basis for diagnosing 3D inconsistency and developing more persistent video world models.
[CV-146] Learning Topology-Aware Representations for Mechanics-Relevant TPMS Metamaterials
链接: https://arxiv.org/abs/2610.20684
作者: Songlin Duan,Shuo Yang,Yujia Zheng,Tobin Filleter
类目: Machine Learning (cs.LG); Computational Engineering, Finance, and Science (cs.CE); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (stat.ML)
备注: 29 pages, 10 figures
Abstract:Learning informative representations of mechanical Triply Periodic Minimal Surface (TPMS) metamaterials is important for mechanics analysis and structural design, yet existing methods have rarely explicitly supervised learned representations with information about global structural connectivity that is relevant to mechanical response. We propose a topology-aware representation learning framework that integrates persistent homology into a DMAE backbone to explicitly encode global structural connectivity. A persistence-guided masking strategy prioritizes patches associated with persistent topological features, while an auxiliary persistent-homology decoder encourages the latent representation to retain information about topological feature counts across connected components, tunnels, and voids. We evaluate the learned representations on two complementary TPMS datasets using family classification, mechanical-property classification, unsupervised clustering, and regression. Across the two datasets, our method demonstrates strong performance in both structural family and mechanical-property classification, improving family classification by up to 3.9% while achieving the best linear-probe performance. More notably, unsupervised property-clustering performance improves by up to 26.7% and 51.5% across the two datasets, while strong regression performance is retained with R^2=0.993 . These results establish persistent topology as an effective inductive bias for representation learning of complex three-dimensional mechanical metamaterials.
[CV-147] owards automatic digital garment assembly
链接: https://arxiv.org/abs/2610.20651
作者: Alejandro Rodriguez
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 11 pages, 8 figures
Abstract:Creating digital garments from existing two-dimensional patterns remains a time-consuming process. Although patterns are widely available in digital form, their assembly in garment CAD software involves manual definition of sewing operations and arrangement of the pattern panels for drape simulation. We present two learning-based approaches that automate both tasks using the 2D vector pattern as their sole input. Our automatic sewing model operates at the segment level, leveraging a Transformer encoder with multiple prediction heads. Our novel architecture supports many-to-many, ordered segment correspondences, as well as hems, overlapping seams needed for multilayer garments, and construction properties such as seam allowances and stitch types, required for the finished virtual garment but neglected by previous methods. A complementary panel-level model predicts body-region assignments, layers, and placement coordinates to initialize the drape simulation. We evaluate both models on a curated dataset of 4,035 professionally assembled garment instances. The sewing model exactly recovers all seam and hem topologies of 63.3% of test garments, while the arrangement model meets the specified placement criteria for all panels in 66.2% of test garments. We also integrate both methods into garment CAD software, demonstrating automatic assembly of successfully predicted garments.
[CV-148] FlowBudget: Training-Free Sampling-Budget Estimation for Flow Matching Models
链接: https://arxiv.org/abs/2610.20636
作者: Zahra Dehghani,Pablo Piantanida,Mohammadhadi Shateri
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Sampling from continuous flow-based generators such as flow matching models requires numerical ODE or SDE integration, making inference cost depend on the number of function evaluations (NFEs). In practice, the sampling budget is inherited from prior work or selected through expensive FID-versus-NFE sweeps, using FID to assess image-generation quality, requiring sample generation and evaluation at multiple budgets. This makes sampling-budget selection dependent on repeated empirical evaluation. We therefore ask a complementary question to solver design and timestep allocation: for a fixed pretrained model and solver, what numerical resolution is sufficient? We introduce FLOWBUDGET, a training-free estimator of solver-relative sampling complexity. Using simulation-free interpolation probes of the learned dynamics, FLOWBUDGET normalizes the leading one-step discretization error by the motion, estimates its coefficient from interpolation probes, and maps a numerical tolerance to a step budget using the solver order. We derive estimators for numerical solvers in flow~matching sampling, namely Euler and Heun for ODE sampling and additive-noise Euler–Maruyama for SDE sampling, and connect the ODE coefficients to terminal integration error. Experiments on SiT, FlowDCN and Rectified Flow models, spanning flow-model architectures and formulations, test whether FLOWBUDGET can estimate the NFE budget without generating samples at multiple budgets or performing an FID-versus-NFE sweep. The code is available at this https URL
[CV-149] BD-HazardVLM: Probing Vision-Language Models for Latent Defensive-Driving Hazards in Bangladesh Road Scenes ECCV2026
链接: https://arxiv.org/abs/2610.20627
作者: Abdullah Al Maruf,Md. Sajedul Islam,Tanmoy Mridha,Irfan Hossain Bhuiyan
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Robotics (cs.RO)
备注: Accepted at the ECCV 2026 Workshop on Safe and Defensive Autonomous Driving (SDAD). 16 pages, 1 figure, 5 tables
Abstract:Vision-language models are increasingly used for driving-scene understanding, yet their defensive reasoning under weakly structured Global South traffic remains unclear. We introduce BD-HazardVLM, a diagnostic benchmark of 500 Bangladesh road-scene images, comprising 375 from RSUD20K and 125 from TFP-BD, annotated for hazard presence, primary hazard category, risk timing, severity, defensive actions, visible evidence, and hallucination- and Western-road-bias traps. We evaluate Qwen2.5-VL-7B, InternVL3-8B, and MiniCPM-V-4.6 using a shared zero-shot task definition and structured output schema. Binary hazard-presence accuracy ranges from 88.0 percent to 89.2 percent, essentially matching the 89.0 percent majority-class baseline. Fine-grained reasoning is substantially weaker: category accuracy remains below 40 percent, risk-timing accuracy ranges from 1.6 percent to 12.4 percent, and exact defensive-action matching remains below 2 percent. Although InternVL3-8B achieves the highest severity accuracy at 61.2 percent, it has zero recall on the 58 high-severity cases. Qwen exhibits severe category mode collapse, while all models show weak latent-hazard anticipation and near-zero self-reported trap recognition. For the evaluated models, these findings show that coarse hazard recognition does not imply reliable fine-grained defensive-driving reasoning and motivate geographically diverse safety evaluation.
[CV-150] NeuroDETR: Multiscale Discrete Spatiotemporal Tokenization for Asynchronous EEG Event Detection
链接: https://arxiv.org/abs/2610.20349
作者: Yaozhi Wen,Kun Zhao,Wei Wei,Haoyang Chen,Kangning Wang,Yi Sun,Shaowei Gu,Dingkun Liu,Shuang Qiu,Huiguang He
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Asynchronous EEG event detection supports practical brain-computer interfaces by identifying event classes and temporal boundaries directly from continuous recordings without predefined event boundaries at inference time. The primary challenge lies in extracting event-related patterns embedded in ongoing background activity. To address this, we propose \textbfNeuroDETR, a discrete spatiotemporal multiscale framework that combines structured tokenization with event-aware feature refinement. Multiscale temporal codebooks capture brief responses and sustained dynamics, while Spatial-Pyramid VQ hierarchically encodes spatial structure across EEG channels. These complementary representations form a compact event memory for jointly classifying events and estimating their onsets and offsets. Before tokenization, we apply FOCAL, an event-aware sparse attention module that uses predicted event presence to guide temporal feature refinement and spatial aggregation, emphasizing likely event periods while preserving context across the recording. Experiments on eight EEG datasets that span diverse paradigms demonstrate state-of-the-art performance. These findings provide a basis for the development of BCI systems that recognize meaningful neural activity during continuous interaction and monitoring.
[CV-151] Complementary Future Token Pruning for World Action Models
链接: https://arxiv.org/abs/2610.19957
作者: Baekseung Kim,Junhoo Lee,Sohyeon Hong,Suhyun Jeon,Minkyu Kim,Seungyeon Kim,Donghoon Han,Nojun Kwak
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:World action models (WAMs) jointly predict future visual states and robot actions, but processing dense future-frame tokens incurs substantial inference latency. We propose CF-Prune, a training-free framework for pruning future-frame tokens in pretrained WAMs. Our key insight is that future visual predictions support action generation and therefore need not preserve every visual detail required for high-fidelity video synthesis. CF-Prune combines instruction and action attention to identify task-relevant tokens during the initial denoising forward pass, even when future-frame and action representations remain highly noisy. An entropy-based phase identification mechanism further retains dense computation during precision phases to mitigate pruning-induced performance degradation. Experiments on RoboLab, RoboCasa, and real-world manipulation tasks demonstrate consistent inference speedups. At a 35% pruning ratio, CF-Prune achieves average speedups of 1.19-1.27x while improving task success rates. At a 50% pruning ratio, it achieves 1.29-1.33x average speedups on both simulation benchmarks, with overall success rates within 1.0 percentage point of the baseline.
[CV-152] Denoising Textures is a Different Ballgame Altogether!
链接: https://arxiv.org/abs/2610.22258
作者: Saptarshi Mandal,Oindrila Haldar,Chirayata Bhattacharya,Chandra Sekhar Seelamantula,Zhiwei Xu,Thrasyvoulos N. Pappas,Gaurav Sharma
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The human brain effortlessly deciphers colors and textures from visual stimuli and reconstructs the world around us, yet preserving such details remains challenging in image denoising. While significant progress has been made in generic image denoising, the problem of texture denoising has remained relatively under-explored. Fine details, such as intricate patterns and repetitive structures, often suffer from over-smoothing even in deep learning based approaches for denoising. In this paper, we analyze the behavior of state-of-the-art (SOTA) denoisers on texture datasets and show that performance improvements are possible only by fine-tuning the denoisers on the texture datasets. Surprisingly, fine-tuning on texture datasets result in the models catastrophically forgetting the original dataset. These findings highlight the need for specialized treatment of textures in denoising. Since denoising quality directly impacts downstream tasks such as generative modelling, we argue that dedicated texture aware denoisers are essential for faithful texture reconstruction and synthesis.
[CV-153] Full-Duplex Interaction with Speech Expressions and Gestures
链接: https://arxiv.org/abs/2610.21452
作者: Siyin Wang,Wenyi Yu,Terumi Chiba,Xianzhao Chen,Xiaohai Tian,Jun Zhang,Lu Lu,Chao Zhang
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Sound (cs.SD)
备注: 37 pages, 5 figures
Abstract:Human face-to-face conversation requires coordinating speech, gestures and expressions across both speaking and listening. We present ELLSA 2, an end-to-end full-duplex dialogue model that jointly optimizes speech interaction and streaming gesture-expression generation. An autoregressive, chunk-wise synthesizer combines pretrained acoustic features with distinct listening and speaking hidden states from the dialogue backbone and speech synthesizer, enabling continuous non-verbal generation even when the model is verbally silent. To translate dialogue understanding into vivid semantic gestures, we introduce contextual semantic calls and template guidance, separating whether to invoke an available semantic gesture from how to realize it. We also establish a curated data and evaluation pipeline for dialog-aware co-speech gesture generation and investigate reference-based multimodal large language models (MLLMs) as automatic judges. Experiments show that ELLSA 2 remains highly competitive with open-source full-duplex speech interaction models and outperforms cascaded baselines across dialog-aware gesture and expression metrics. Semantic template guidance further significantly improves semantic gesture quality. Strong system-level correlations with human judgments support these reference-based scores as useful preliminary metrics for full-duplex and semantic gesture assessment. Together, these results advance dialogue modeling from spoken responses toward coordinated multimodal interaction.
[CV-154] Monotone Piecewise-Linear Harmonization of Multitemporal Multispectral Surface Reflectance With Unsupervised Deep Multiview Pseudoinvariant Features Learning
链接: https://arxiv.org/abs/2610.21183
作者: Lino Garda Denaro,Chia-Hsiang Lin
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注: 15 pages, 5 figures
Abstract:Surface reflectance products derived from optical satellite may exhibit systematic inconsistencies across acquisition dates, because atmospheric correction is generally performed independently for each image. Such residual biases distort the temporal behavior of invariant ground targets and reduce the reliability of multitemporal surface reflectance analysis. This study proposes Deep Multitemporal Generalized Canonical Correlation Analysis (DeepMGCCA), an unsupervised deep multiview pseudoinvariant feature learning framework that integrates Deep Generalized Canonical Correlation Analysis (GCCA) with Multivariate Alteration Detection (MAD) to identify deep-pseudoinvariant features (DIFs) for the removal of surface reflectance biases across multitemporal satellite observations. Unlike conventional MAD-based approaches that rely on linear GCCA, the proposed framework replaces date-specific linear projections with a shared convolutional encoder. The framework learns deep invariant features (DIFs) directly from pixel-weighted canonical variates via iterative reweighting strategy, while a reconstruction regularizer preserves surface reflectance information and prevents degenerate collapse from suppressing genuine temporal variability. Experiments were conducted on four annual Landsat-8/9 and Sentinel-2 acquisitions across four regions of interest (ROIs) to evaluate four DIF-based harmonization transforms. A monotone piecewise-linear transform derived from DIFs quantiles consistently achieved the largest bias reduction, lowering the between-date spectral angle by up to 2.54 and RMSE by 0.015, and raising PSNR by up to 4.04~dB. These findings demonstrate the potential of deep invariant feature learning for multitemporal surface reflectance bias removal while highlighting the importance of preserving physically meaningful temporal and spectral information during nonlinear representation learning.
人工智能
[AI-0] LIBERO-PeRM: Benchmarking Personalized Robotic Manipulation NEURIPS2026
链接: https://arxiv.org/abs/2610.22372
作者: Zhixu Li,Keqian Tang,Litian Gong,Jingyu Yao,Wenqian Zhang,Tianze Xu,Zehao Wang,Yuping Wang,Junge Zhang,Jiachen Li
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Conference on Neural Information Processing Systems (NeurIPS 2026); Project Website: this https URL
Abstract:Personalization is essential for general-purpose robots to transition into household environments and achieve true practicality. While general robotic manipulation has been extensively studied, personalized robotic manipulation (PeRM) lacks a controlled benchmark for separating task completion from preference satisfaction. We introduce LIBERO-PeRM, a benchmark for preference learning in embodied agents. LIBERO-PeRM studies five questions: 1) how policies follow preferences in zero-shot settings; 2) how they learn single preferences from demonstrations; 3) how efficiently they acquire preferences as demonstrations increase; 4) how learned preferences generalize across layouts and tasks; and 5) how policies compose preferences and behave under conflicts. We propose a controlled formulation and taxonomy of observable preferences in simulation, covering motion, skill, and user-profile choices. This formulation is restricted to language-expressible preferences that can be instantiated in LIBERO without changing the task goal, evaluated automatically, and reused across tasks. Building on LIBERO, our procedural pipeline generates paired demonstrations with and without preferences. We create five task suites (312 tasks) with high-quality demonstrations, rich metadata, and preference-satisfaction metrics. Experiments show that current VLA policies struggle with zero-shot personalization, while SFT makes preferences learnable. Mixed training achieves better no-preference retention than sequential adaptation and improves performance with contextual prompts compared with preference-only training, but can weaken motion-preference learning; preference generalization, decomposition, and conflict handling remain difficult. The code, datasets, and demos can be found on the project website: this https URL.
[AI-1] CoDA: Co-Evolution of Data Engines and Action Policies
链接: https://arxiv.org/abs/2610.22362
作者: Tianjiao Yu,Zhilin Ou,Yifan Shen,Xinzhuo Li,Yuanzhe Liu,Onkar Kishor Susladkar,Fan Wang,Ismini Lourentzou
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: Project Website: this https URL
Abstract:Generative simulation can automate robot environment construction, but physically valid environments are not necessarily useful for training. The challenge is to identify environments that address a policy’s failures and supervision that helps it overcome them. We introduce CoDA, a framework in which a simulation data engine and an action policy co-evolve through a shared heterogeneous memory of environment construction, policy failures, and learning progress. The memory compiles validated diagnoses from a vision-language critic into generation programs that recreate difficulties and training programs that help the policy overcome them. We further propose Text2potential to convert these diagnoses into staged potentials for action-chunk value learning, providing dense supervision for policy updates while preserving the task’s optimal policy under standard shaping assumptions. Measured learning progress guides further generation and program refinement. Across four manipulation task families, CoDA improves out-of-distribution success from 5.3% after supervised fine-tuning to 27.6%, exceeding the strongest closed-loop baseline by 9.5 percentage points with the same generation budget. On LIBERO-Pro, 15 rounds of generation and training raise success from 25% to 84% under position perturbations and from 1% to 76% under instruction perturbations.
[AI-2] Rethinking Routing Exploration in Modern Mixture-of-Experts
链接: https://arxiv.org/abs/2610.22353
作者: Bingcong Li,Hao Ma,Michael Muehlebach,Niao He,Yi Ma
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Mixture-of-experts (MoE) models have become an important approach for scaling model parameters. Modern MoE architectures increasingly combine two design choices: fine-grained expert segmentation and shared experts. The former increases expressiveness by enabling more expert combinations at a fixed activation ratio, and the latter introduces experts that are active for every token. We identify an exploration dilemma between these designs: shared experts can make routing less exploratory by reducing the variability of token representations. This conflicts with the greater need for exploration induced by fine-grained expert segmentation due to the combinatorially enlarged routing space. Consequently, the additional expressiveness offered by fine-grained segmentation may not be fully realized in practice. MirrorMoE is proposed to reconcile this exploration dilemma by shrinking the exploration space through token-adaptive suppression of unnecessary experts. MirrorMoE consistently improves performance across model scales.
[AI-3] Dexterity from Supervised Play
链接: https://arxiv.org/abs/2610.22340
作者: Botao He,Yiming Dou,Yiannis Aloimonos,Jitendra Malik,Akash Sharma,Ruoshi Liu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:How can robots autonomously collect the contact-rich experience needed to learn dexterous manipulation? We present DSP: Dexterity from supervised play, a framework in which a robot learns a play policy from minutes of human play through behavior cloning, then deploys it autonomously to collect hours of interaction data to train a world model. During inference, the same play policy then guides model-predictive control (MPC) by proposing candidate action sequences for evaluation under the learned world model. To accelerate inference, we also introduce a method for distilling a closed-loop, goal-conditioned policy through self-play within the learned world model. We evaluate this framework on underactuated pushing and any-to-any in-hand pose reorientation with a multifingered hand, in simulation and real world. In real world experiment, our method outperforms hindsight behavior cloning by 69.6% in success rate and achieves comparable success to reinforcement learning with 490 times less interaction data. These results demonstrate that supervised play can autonomously collect interaction data, guide model-based planning, and enable policy learning without any task-specific demonstrations or sim-to-real transfer.
[AI-4] KnowGPO: Knowledge-Grouped Policy Optimization for Search Agents
链接: https://arxiv.org/abs/2610.22331
作者: Siqi Chen,Zhengxi Lu,Yuchen Yan,Xingyu Wu,Qipeng Chen,Xin Zhang,Aiting Liu,Chao Deng,Jie Liu,Jin Ma,Weiming Lu,Jun Xiao,Yongliang Shen
类目: Artificial Intelligence (cs.AI)
备注: Code available at this https URL
Abstract:Deep search agents solve knowledge-intensive tasks through long-horizon, multi-turn interactions, where assigning credit to individual search decisions is fundamentally difficult. Outcome-based reinforcement learning provides only a single terminal reward shared across an entire trajectory. Recent process-supervision methods aim to assign credit at finer granularity, but they typically anchor credit to either surface-level trajectory states or model-internal signals, both of which become unreliable in deep search where observations are heterogeneous and retrieved contexts are long and noisy. We propose KnowGPO, a reinforcement learning framework that aligns trajectories by the task-relevant knowledge they have acquired rather than by their surface observations. For each query, KnowGPO decomposes the information required to derive the answer into atomic knowledge conditions, and represents the state of a trajectory at each step by the subset of conditions it has satisfied. Trajectories sharing the same knowledge state are grouped for within-group comparison, and the incremental knowledge gain between consecutive steps serves as an environment-grounded process reward. Experiments on Xbench, BrowseComp-ZH, and GAIA show KnowGPO consistently outperforms outcome-only (GRPO) and process-supervision baselines (GiGPO, IGPO) by 3.8 to 4.2 points, with our 8B model rivaling or exceeding most 32B open-source systems.
[AI-5] Designing Transformers for Continuous Diffusion Language Models
链接: https://arxiv.org/abs/2610.22330
作者: Weijian Luo,Yujian Chen,Hao Jiang,Guang Cheng,Tianyang Hu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Continuous diffusion language models (CDLMs) offer an alternative approach to parallel language generation; however, most existing studies directly use transformer architectures from either causal large language models (LLMs) or Diffusion Transformers (DiTs) for image generation, which were not optimized for CDLMs. In this paper, we conduct a systematic study on Transformer architectures for CDLMs and introduce Language Diffusion Transformer (LangDiT), combining feature and query-key normalization, wider noise conditioning, likelihood-shortcut gradient routing and calibration, and confidence-gated training self-conditioning. LangDiT significantly reduces generation perplexity in matched LangFlow comparisons, establishing a new state of the art in continuous language generation across different model formulations. Using LangFlow as a representative CDLM, LangDiT as a plug-and-play backbone leads to a new state-of-the-art model with a generation perplexity (GenPPL) of 21.72 using 1024 NFEs and 44.72 GenPPL using 128 NFEs on the unconditional generation of 1024-token sequences on the OpenWebText dataset. When adapted to Plaid-style CDLM formulations, LangDiT results in a new leading GenPPL of 11.33 using 1024 denoising NFEs, and 27.51 GenPPL using 128 denoising NFEs. These gains are achieved while retaining each model’s underlying representation, training objective, and inference sampler, isolating transformer architecture as a fundamental source of improvement. Our study identifies architectural design as an important source of improvement in continuous diffusion language models.
[AI-6] A team of Gemini 4 Argon agents beats Minecraft
链接: https://arxiv.org/abs/2610.22319
作者: Maxence Faldor,Satinder Singh,Roberta Raileanu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We placed four Gemini 4 Argon agents in a sandboxed Minecraft survival world and gave them one goal: defeat the Ender Dragon. Each agent controlled a player via an open-source library called Mineflayer and coordinated with its teammates through in-game chat and shared memory. No one assigned roles or wrote game logic. The agents planned, wrote, and debugged every program themselves. Researchers could watch the run on a live dashboard and pause or stop it at any time, but they had no cause to intervene. Starting from a randomly generated world on Easy difficulty, the team wrote 2,783 programs and defeated the dragon in 27 hours and 14 minutes, dying 605 times along the way. To our knowledge, this is the first end-to-end autonomous completion of Minecraft by AI agents, achieved using only the standard Mineflayer library without pre-scouted seeds or other game-specific tools written by humans. Winning required the agents to keep a coherent plan across thousands of dependent steps for more than a day. Complex real-world software engineering and enterprise tasks demand the same long-horizon coding and multi-agent coordination. We describe the setup, how the run unfolded, the team’s strategies (dividing the work, sharing what they learned, recovering each other’s items), and the limits of the result.
[AI-7] Private Retrieval Augmented Generation via Random Projection
链接: https://arxiv.org/abs/2610.22297
作者: Dixi Yao,Tahseen Rabbani,Tian Li
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:Retrieval-Augmented Generation (RAG) enhances the capabilities of large language models (LLMs) by querying external structured knowledge. However, it can also introduce privacy risks by leaking sensitive information from the retrieval datastore. We propose a simple method to preserve datastore privacy in RAG systems via random projection. By applying the same projection to both datastore embeddings and query embeddings, our method provably preserves semantic similarity between queries and retrieved items while substantially mitigating data extraction attacks. Across multiple RAG architectures and datasets, we show that this lightweight approach achieves superior retrieval and generation performance compared to prior methods which leverage formal differential privacy (DP) guarantees, while exhibiting comparable empirical privacy under strong attack models. Our results for the first time suggest that random projection can serve as a competitive and practical baseline for privacy-preserving RAG systems.
[AI-8] owards Continual Learning: Mitigating Forgetting With Subspace Tuning
链接: https://arxiv.org/abs/2610.22293
作者: Arnav Chavan,Kusha Sareen,Rishabh Tiwari,Abhranil Chandra,Johan Obando-Ceron,Kurt Keutzer,Scott Niekum,Inderjit S Dhillon,Devvrit Khatri
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) acquire new skills through fine-tuning, but it can destroy capabilities they already have: catastrophic forgetting remains a central obstacle to continual learning in LLMs. Recent works connect forgetting to how far, and along which directions, fine-tuning moves the weights. On-policy methods such as reinforcement learning and on-policy distillation avoid updating the principal singular directions of weight matrices and thus forget less. In contrast, SFT concentrates updates along these directions and is susceptible to catastrophic forgetting. In this work, we introduce Subspace Tuning (SST), which restricts each weight update to the bottom singular subspace, eliminating changes along the principal directions. SST is optimizer-agnostic and composes well with fine-tuning recipes including SFT and self-distillation. SST combines tight subspace protection, dense gradient estimation, and a simple refresh mechanism, enabling stronger learning than prior orthogonal fine-tuning methods. With these mechanisms, protecting as much as 99% of the weight subspace substantially improves retention while preserving strong target-task performance. This behavior strengthens with model and data size and nearly eliminates forgetting in the large-model regime. SST also keeps the model close to its base distribution, inducing 2 – 4\times less KL drift than the baseline recipes. Overall, SST advances the learning–forgetting Pareto frontier, making it a practical default for fine-tuning pre-trained language models.
[AI-9] EliteReplay: Preserving Diverse Solutions in RL Post-Training via Quality-Diversity Search
链接: https://arxiv.org/abs/2610.22290
作者: Hanane Nour Moussa,Doeun Lee,Sachin Kumar
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning has become a common post-training technique to improve the accuracy of LLMs on verifiable tasks. However, such training can narrow the set of solutions the model can produce. Several approaches have attempted to address this behavior-sharpening effect, but they often optimize metrics that do not reflect the diversity of correct solutions. Ideally, we want a model that can produce multiple good solutions, not only a single reliable one. We propose EliteReplay, a post-training recipe that pairs GRPO with an “elite” archive built by quality-diversity search. Before training, we use the frozen base model to collect solutions that span distinct approaches; during training, GRPO is combined with a replay loss on this archive. The result is a post-trained model that improves accuracy while preserving more of the base model’s solution diversity. On multiple text-to-SQL tasks, EliteReplay achieves a higher pass@128 than all baselines, producing up to 77% more distinct correct strategies than the strongest baselines. It preserves most of the base model’s correct strategies, outperforming existing methods in strategy retention. This diversity translates into practical benefits on ambiguous questions, where EliteReplay covers both interpretations nearly twice as often as the strongest baseline on AmbiQT.
[AI-10] Improving Environment-Agnostic Coding Agents : The Role of Scaffolding
链接: https://arxiv.org/abs/2610.22289
作者: Sergey Rodionov
类目: Artificial Intelligence (cs.AI)
备注: 37 pages, 5 figures
Abstract:Stronger language models may increasingly construct useful procedures for themselves, reducing the advantage of developer-supplied scaffolding. We investigate why this advantage persists and whether an environment-agnostic agent can be improved without environment-specific engineering. We evaluate a Codex-based agent in two nearly identical prompt variants on ARC-AGI-3, AGI Maze, and Terminal-Bench. Each variant is deployed with unchanged code, prompts, and general-purpose tools across environments, receiving environment-specific information as input. Changing only an instruction to encourage reporting difficulties and returning control for automatic continuation improves mean scores on both ARC-AGI-3 and AGI Maze, while Terminal-Bench shows small, mixed changes. Yet the general agent remains below an ARC-AGI-3-specific reference, including with GPT-5.6-sol. The reference prescribes organization and control rather than supplying new game rules or solutions. To investigate this gap, we transfer reset decisions to a scripted controller that follows a convention already given to the agent. This small change recovers part of the specialized agent’s advantage. A complementary AGI Maze experiment shows that additional tutorial experience can reduce performance. Together, these findings highlight the distinction between having information and tools available and using them effectively. Supplied procedures still matter for the models tested here. The performance gains also come with higher token-based cost. The experiments do not separate better use of information and procedures from additional or differently allocated inference. We provide the EnvTraverse framework and an evaluated baseline agent for testing whether further environment-independent changes improve performance in some environments without degrading it in others.
[AI-11] LEASH: Uncertainty-Aware Weak-to-Strong Escalation and Oversight for Computer-Use Agents
链接: https://arxiv.org/abs/2610.22280
作者: Qingni Wang,Changdae Oh,Yuzhe Yang,Sharon Li,Jiachen Yang,Ang Li,Xin Eric Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Computer-use agents increasingly tackle long-horizon tasks, where running a frontier model at every step is expensive and a small open-weight model alone often fails. Weak-to-strong escalation offers a middle ground where a weak agent acts by default and hands control to a strong one only when needed. The core challenge is deciding when to escalate, which requires an uncertainty signal cheap enough to check at every step. However, even after handoff, the strong agent can still get lost, but a black-box API that the strong agent is usually served exposes no internals to monitor. To this end, we propose LEASH, a training-free framework that unleashes the strong agent only when the weak agent’s latent uncertainty signals it is needed, and keeps it on a leash afterward. LEASH has three layers: (1) initial task uncertainty for early routing, (2) step-level uncertainty for runtime escalation, and, after handoff, (3) a shadow observer for continued oversight of the strong black-box agent. Here, we use latent uncertainty as a fuel for all three layers. Specifically, we compare the weak model’s hidden states with those from its past successful and failed trajectories, and a state that looks more like the failures signals higher uncertainty. During weak-agent execution, our uncertainty estimator reuses activations already produced during action generation, so it needs no extra forward pass or dedicated monitor training. Experiments on OSWorld across four contemporary model backbones show that LEASH consistently improves over standalone weak agents and outperforms random escalation by 6.55–11.94 points at matched escalation rates. With Qwen3.5-35B-A3B, LEASH achieves 73.66% task success at 0.4117 per task, matching always-on Claude Opus 5 (72.90%) while reducing inference cost by 19.2%.
[AI-12] X-RIC: An Agent ic RAN-Intelligent Controller Benchmark for Enhancing Reinforcement Learning
链接: https://arxiv.org/abs/2610.22269
作者: Adhwaa Al-Chaab,Ayman Younis,Dario Pompili
类目: Artificial Intelligence (cs.AI); Hardware Architecture (cs.AR)
备注: 9 pages, 10 figures. Accepted for publication in the proceedings of the IEEE International Conference on Mobile Ad-Hoc and Smart Systems (MASS), October 21-23, 2026, Hong Kong SAR, China
Abstract:The O-RAN ALLIANCE introduces architectural flexibility through disaggregation and AI-driven control. While the RAN Intelligent Controller (RIC) enables centralized optimization, current AI approaches lack reasoning capabilities and explainability. We propose eXplainable RIC (X-RIC), a novel framework integrating agentic AI into O-RAN architectures. X-RIC combines reactive Reinforcement Learning (RL) with strategic agentic reasoning through: (i) automated Behavioral Cloning (BC) using domain heuristics for sample-efficient training, (ii) a Telecom Language Model (TLM), a transformer-based reasoning module providing contextual abstraction and interpretability, and (iii) closed-loop co-adaptation within the Near-RT RIC control loop enabling continuous policy refinement. To ensure deployment readiness, we introduce BARX (Benchmarking Agentic Reasoning for RAN), a domain-specific evaluation framework assessing agentic control across latency, compliance, and determinism metrics. We validate X-RIC on SLA-aware network slicing in simulation and O-RAN-compliant testbed deployments. X-RIC achieves 89% operational cost reduction while maintaining SLA compliance, converging 75% faster with 50% fewer training steps and 46% lower variance.
[AI-13] Disagreement-Guided Weak-to-Strong Routing
链接: https://arxiv.org/abs/2610.22245
作者: Sima Noorani,Shayan Kiyani,George Pappas,Hamed Hassani
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Small language models are increasingly deployed on edge devices and robotic systems to meet cost and latency constraints, yet their limited capabilities can make reliable deployment challenging. Stronger models can provide assistance when needed, but deciding when to route a weak-model response to a stronger model often relies on signals of response correctness. In open-ended and agentic tasks, however, obtaining such signals can require costly ground-truth supervision that is difficult to obtain at scale. We revisit the foundations of weak-to-strong routing and ask whether routing decisions can instead be guided by signals that do not require ground-truth supervision. We first characterize the optimal weak-to-strong routing policy, while allowing for multiple weak-model responses generated through local inference. Building on this characterization, we derive a principled approximation to the optimal routing based on model-to-model comparison. In particular, we introduce a notion of model disagreement that can be measured directly from model-generated responses, without ground-truth supervision. Importantly, the approximation becomes increasingly accurate as the strong model becomes a better proxy for the ground-truth distribution. Across agentic and single-turn tasks, our approach effectively navigates the trade-off between routing cost and response accuracy without ground-truth supervision. Moreover, even when ground-truth correctness signals are available for training, disagreement-based routing can outperform methods trained directly to predict correctness
[AI-14] Shortcut Sensitivity in Agent -Safety Evaluations NEURIPS2026
链接: https://arxiv.org/abs/2610.22243
作者: Hasan Demirkiran,Jens Ernstberger
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注: Accepted at the NeurIPS 2026 workshops TAE (Can We Trust AI Evaluation?) and AIWILD (Agents in the Wild: Safety, Security, and Beyond)
Abstract:AI agents take actions through tools. Some actions are sensitive, so guards that supervise them must themselves be evaluated reliably. Yet the data used to evaluate a guard can include cues that make labels predictable without measuring the intended safety construct, and those cues can also influence guard decisions. We introduce a three-level shortcut-auditing framework and audit completed LinuxArena trajectories and two TS-Bench-derived step datasets. We distinguish label predictability, observational guard association, and paired decision sensitivity. In LinuxArena, shorter average shell commands strongly predict trajectory labels. In AgentDojo-Traj, exact-tool lookup falls from F_1=.901 with mixed-domain folds to .495 under domain holdout. In AgentHarm-Traj, released TS-Guard F_1 falls from .902 under row weighting to .719 with equal interaction weighting, while balanced accuracy rises. We then test the strongest safe-row marker association in a seven-configuration screen on a frozen 64-step AgentDojo-Traj cohort using repeated, interleaved Haiku calls. Rewriting the INFORMATION marker lowers blocking by 8.33 percentage points, while opaque tool aliases raise it by 11.46 points. All unsafe calls remain blocked; the changes concern false positives against inherited labels. Both effects remain nonzero after subtracting the formatting-placebo effect, with exploratory 95% ranges excluding zero. The same edits also change measured accuracy against inherited labels. This provides direct evidence that benchmark representation can materially change a guard’s decisions and measured evaluation performance. The Haiku-sized effects do not reproduce in Luna or deterministic Q8 TS-Guard. We recommend domain group holdouts, evaluation-unit metrics, and placebo-controlled paired interventions.
[AI-15] GNN-Guided Local Refinement for Efficient Layout Decomposition with DSA-MP
链接: https://arxiv.org/abs/2610.22214
作者: Yu Feng,Liyi Zhou,Haoyu Wang,Yiming Xu,Hao Geng,Shaozhi Deng,Ningsheng Xu,Shisheng Xiong
类目: Artificial Intelligence (cs.AI)
备注: 9 pages. Accepted at the IEEE/ACM International Conference on Computer-Aided Design (ICCAD 2026)
Abstract:Multiple patterning lithography (MPL) has been widely used to improve lithography resolution in advanced nodes, but it faces cost limits as feature sizes shrink. Directed self-assembly (DSA) offers a cost-effective complement, extending the resolution of optical lithography to sub-10 nm and beyond. Integrating DSA into MPL (DSA-MP) further enhances the resolution of contact/via patterning. Nevertheless, growing design complexity exposes the limits of existing decomposition methods, which struggle to balance solution quality and runtime, and neglect edge-to-edge spacing constraints of DSA templates, compromising manufacturability. To address these issues, we introduce a new conflict graph model capturing the previously overlooked constraints and develop a local refinement strategy powered by the graph neural network, enabling efficient and scalable layout decomposition. Experiments demonstrate that our method achieves higher solution quality with faster convergence than state-of-the-art methods, while maintaining robustness across layout scales, design styles, technology nodes, and varying mask counts.
[AI-16] rminal-RSI: Recursive Self-Improving Terminal Agents via Recursive Self-Rewrite and Policy-Refine
链接: https://arxiv.org/abs/2610.22192
作者: Junyao Yang,Zongxia Li,Yucheng Shi,Zhongzhi Li,Ruhan Wang,Ruiwen Zhou,Haitao Mi,Shuicheng Yan,Leowei Liang
类目: Artificial Intelligence (cs.AI)
备注: 28 pages, 16 figures, 5 tables
Abstract:Converting verifier outcomes into useful policy updates and partial progress into complete task resolution is crucial for terminal agents. We present Terminal-RSI, which alternates Recursive Self-Rewrite (RSR) of verified trajectories with critic-inheriting Recursive Policy-Refine (RPR). We formulate Turn-Boundary Skip-Observation GAE (TB-GAE) to retain token-specific baselines while applying decay across interaction turns. Family Dense Reward (FDR) calibrates semantic progress across task families; Success-Prioritized Reward (SPR) separates strict success from bounded partial credit. A Warm-up Critic Model, recursive critic handoff, and stage-specific learning rates coordinate the two refinement phases. The five-stage sequence from Qwen3.8-27B base model through RSR-1, RPR-1, RSR-2, and RPR-2 yields Terminal-RSI-27B: 19.84% on Terminal-Bench 4.0, 86.52% on Terminal-Bench 2.1, 16.95% on Terminal-Bench 3.0, and 79.00% on Terminal-Bench Hard. On LHTB, process reward rises from 0.21 to 0.37 and strict completion from 0/46 to 5/46; SWR100 success doubles from 3.00% to 6.00%. By recursively coupling verified trajectory rewriting with critic-inheriting policy refinement, each retained stage consolidates discovered behaviors and progressively advances the agent from partial local repairs to strict terminal verification.
[AI-17] Whole-Body Dexterity
链接: https://arxiv.org/abs/2610.22184
作者: Conor Mc Gartoll,An Dang,Pranay Thangeda,Arjun Gupta,Carolina Higuera,Nima Fazeli,Mustafa Mukadam,Manikantan Nambi
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Systems and Control (eess.SY)
备注: 20 pages, 13 figures
Abstract:Current humanoid dexterous manipulation remains largely confined to end-effectors and simple grasps, falling short of humans’ ability to engage the entire body by bracing and sliding bulky objects against the arms and torso. We introduce Whole-Body Dexterity (WBD), a general framework that uses the robot’s body and appendages as contact surfaces to grasp and manipulate objects. Our sim-to-real pipeline first generates a dataset of feasible whole-body grasps, then trains reinforcement-learning experts on a general any-to-any repositioning objective by sampling initial and goal grasps from the dataset, and distills these experts into a single policy that uses proprioception and object geometry. We show that the distilled policy transfers zero-shot to a real Unitree G1, where it repositions a basketball and a cardboard box to regions around the torso and completes a long-horizon task that stows a box under one arm, carries it from a shelf to a table, and places it. The controller operates interactively, steered by coarse motion references from a human teleoperator. Across real-world whole-body manipulation tasks, the distilled controller completes 90% of trials, compared with 18% for direct retargeting and 4% for a generalist whole-body controller.
[AI-18] Multi-sourced AI Tutor for Non-Native English Speaker Students
链接: https://arxiv.org/abs/2610.22163
作者: Yiyin Shen,Xinyi Ai,Weibing Wang,Rogers Jeffrey Leo John,Meenakshi Syamkumar
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) are increasingly being adopted as teaching tools in computing education, particularly for developing AI tutors capable of answering student questions. With the rise in enrollment of international students in computing programs, researchers have begun to explore the difficulties that non-native English speaker (NNES) students face in computing courses due to language barriers. To aid such students, we developed a general-purpose AI tutor system that can process student questions in foreign languages (currently focused on Mandarin) and provide bilingual responses. Our system also incorporates the functionality of generating a glossary of technical terms and their explanations to enhance comprehension. Our approach utilizes a specialized Retrieval-Augmented Generation (RAG) pipeline with a novel question extraction mechanism, enabling us to effectively utilize a diverse array of course materials for answering student queries. We evaluated our AI tutor using 91 Piazza posts, demonstrating that it could answer 83.52% of the posts correctly and in a helpful manner. The tutor also provided high-quality translations for 76.92% of the questions and generated technical term glossaries for 64.47% of them. Our findings underscore the potential of multi-source RAG pipeline in enhancing AI tutor systems, offering support for non-native speakers that are challenging to provide with traditional supporting staff of a computing course.
[AI-19] Realtime-Bench: Can Frontier Models Solve Real-time Interactive Problems?
链接: https://arxiv.org/abs/2610.22155
作者: Jingyu Tang,Xu Yang,Zhehao Zhang,Zheng Wang,Qi Cheng,Haozhe Zhao,Manling Li,Wenxuan Li,Xide Xia,Jiaxin Pei
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Real-time interaction requires agents to perceive evolving environments and act under temporal constraints, yet existing evaluations of real-time streaming models focus on conversational quality and streaming perception, leaving effective action under strict temporal constraints untested. We introduce Realtime-Bench, a suite of carefully curated dynamic tasks where delayed actions directly affect success. Under Wall Clock, where environments advance according to actual inference latency, the best success rate reaches only 44% and the best Pass^3 is only 36%, demonstrating that real-time interaction remains challenging. To separate serving latency from models’ real-time streaming capability, we compare Wall Clock with Paused Clock, which removes inference delays as an idealized reference, and introduce Token Clock, which advances simulated time according to generated token counts, testing temporal awareness independently of physical serving speed. We find that temporal costs limit success in evolving environments and that higher reasoning effort has mixed effects under time pressure, reflecting a trade-off between decision quality and timely execution. These findings suggest directions for improving real-time multimodal agents, including adaptive reasoning with temporal awareness and continuous perception and action.
[AI-20] Specification Gaming and a Constrained Reinforcement Learning Mitigation in a Cleaning Gridworld
链接: https://arxiv.org/abs/2610.22152
作者: Hilda Adwubi Osei
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 8 pages, 3 figures
Abstract:Reinforcement learning agents optimize exactly the reward signal they are given, which can diverge sharply from an intended true objective when that signal is only a proxy for it; this failure mode is known as specification gaming. Existing evidence for this failure mode is largely observational, with comparatively few demonstrations pairing a controlled, deliberately constructed proxy/true gap with a mitigation that carries a verifiable guarantee rather than only an empirical improvement. We construct a minimal gridworld in which a cleaning-robot agent can either genuinely remove dirt or conceal it from a cheap sensor, and show that a standard deep Q-network trained only on the sensor’s proxy reward converges to a policy that games it completely, reaching near-maximal proxy reward while never achieving the true objective. We then introduce a constrained reinforcement learning fix, penalizing the specific behaviors through which the agent exploits the gap, and show that it eliminates the gaming behavior from the final policy while substantially recovering true-objective performance. We further derive and empirically verify an exact accounting identity relating the proxy reward, true reward, and our cost signal for this environment, showing that the cost signal measures the proxy/true gap exactly rather than merely tracking it, and yields a deterministic guarantee under a hard-constrained variant of the method.
[AI-21] RAG -RST: Retrieved Exemplars as In-Context Discourse Schema for RST Parsing
链接: https://arxiv.org/abs/2610.22150
作者: Mohammadreza Sediqin,Shlomo Engelson Argamon
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:In Rhetorical Structure Theory (RST), each edge of a discourse tree carries a nuclearity label, marking which span holds the central content, and a relation drawn from a closed inventory of functional categories. Neither label is a surface property: the same two clauses can be Elaboration or Background depending on rhetorical intent, and the boundary between them is fixed by annotation convention rather than by lexical cues. Predicting these labels therefore requires a formal discourse schema, and exposure to discourse-rich pretraining text does not supply it: given only a target EDU pair, a strong LLM recovers nuclearity at 0.794 F1 and relation at 0.738 F1. We propose RAG-RST, a retrieval-augmented framework that supplies this schema as in-context demonstration rather than through parameter updates, raising nuclearity to 0.846 F1 and relation to 0.788 F1 with no training. Two configurations are compared: a precomputed strategy that labels all EDU pairs before parsing, and a dynamic strategy that retrieves at each reduce step, so that the exemplars reaching the generator depend on the reductions the parser has already committed to. Both outperform existing RST parsers across span, nuclearity, relation, and full tree metrics, with the dynamic variant consistently strongest. Two properties of the retrieved schema emerge: its utility is bounded in number, saturating and then declining beyond fifty exemplars, and it depends on structural position. Both indicate that a formal linguistic schema can be supplied to a foundation model as a demonstration rather than trained into its parameters.
[AI-22] Are You Sure Your LLM Is Sure? Repairing and Auditing Confidence
链接: https://arxiv.org/abs/2610.22139
作者: Arijit Dey,Bhartendu Kumar,Ramya Hebbalaguppe,Khuushi Maheshwari
类目: Artificial Intelligence (cs.AI)
备注: main paper 10 pages; supplementary 44 pages; 5 figures
Abstract:Confidence scores determine whether an LLM answer is accepted, reviewed, or regenerated, yet scalar recalibration can only correct errors visible through the score itself. We introduce RADAR (Repair, Adopt, and Diagnose After Recalibration), a reliability framework that separates confidence repair from deciding whether to deploy the repair and auditing the errors that remain. Under squared loss, we decompose prediction risk into outcome noise, information hidden by the score, and mapping error; restricting repair to order-preserving maps additionally introduces a pooling gap. For repair, we propose Probe-Anchored Isotonic Calibration (PAIC), which learns a monotone map from correctness-probe predictions on unlabeled target prompts. PAIC’s excess risk depends on the probe’s score-conditional bias and fitting error rather than its overall accuracy. For adoption, we derive a paired procedure with finite-sample control of false improvements. For auditing, we develop a prompt-clustered kernel test that detects feature-aligned residual errors while accounting for dependent answers from the same prompt. Across multiple LLM checkpoints, benchmarks, and distribution shifts, richer predictors can improve confidence estimation, but projecting their information back to a scalar score can lose these gains, and lower prediction risk need not yield better review decisions. Across 47,400 replayed raw-score allocations, paired betting adopts candidates with lower held-out test risk in 42.95% of allocations, versus 15.09% for one-sided empirical Bernstein. RADAR thus provides a principled workflow to repair what a scalar score can express, adopt a repair only when supported by evidence, and diagnose systematic errors that recalibration leaves behind. Code and data will be released upon acceptance.
[AI-23] CART: Compatible Routing for Multimodal Continual Instruction Tuning
链接: https://arxiv.org/abs/2610.22137
作者: Ling-Xiang Zhang,Tao Hu,Da-Wei Zhou
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multimodal Large Language Models (MLLMs) achieve strong performance through instruction tuning, yet real-world applications often require them to continually acquire new capabilities from sequential tasks, i.e., Multimodal Continual Instruction Tuning (MCIT). In typical MLLMs, a projector is employed to align visual representations with the language model. However, existing methods primarily protect the language model from forgetting while continuously updating the projector across tasks. Such asymmetric adaptation can cause substantial representation drift in the projector, increasing the mismatch between the projector and language model and consequently exacerbating catastrophic forgetting. To address this issue, we propose Compatible Routing (CART), which explicitly coordinates the adaptation of the projector and language model. Specifically, CART maintains task-specific experts for both modules and combines them through weighted mixtures. CART jointly infers their weights by considering both expert relevance and cross-module compatibility, thereby promoting coordinated adaptation between the two modules. To further control model growth, we introduce relevance-guided expert consolidation, which merges experts exhibiting similar relevance profiles across samples within each module. Extensive experiments demonstrate that CART achieves state-of-the-art performance.
[AI-24] DropKV: Decoupling Residual-Output Perturbation for Near-Optimal KV-Cache Eviction NEURIPS2026
链接: https://arxiv.org/abs/2610.22130
作者: Aozhong Zhang,Selcuk Gurses,Yanxia Deng,Naigang Wang,Chi-Chun Liu,Davis Wertheimer,Derrick Liu,Xin Li,Zi Yang,Felix X.-F. Ye,Penghang Yin
类目: Artificial Intelligence (cs.AI)
备注: Accepted by NeurIPS 2026
Abstract:Key-value (KV) cache management has become a critical bottleneck for scaling Large Language Models (LLMs) to long-context scenarios due to linear memory growth. While recent eviction methods have moved beyond simple attention-score heuristics to incorporate value representations, they typically face a fundamental trade-off: the underlying optimization for minimizing output perturbation is an NP-hard combinatorial problem, leading to mathematically suboptimal heuristics or implementations that are difficult to kernelize. In this work, we propose DropKV, a simple principled eviction framework based on Decoupled Residual-Output Perturbation. By decoupling the joint eviction decision into independent per-token scoring, DropKV circumvents combinatorial intractability and admits a provable constant-factor approximation guarantee under a cone condition that we empirically verify on three long-context LLMs. To ensure practical viability, we provide a high-performance implementation using fused Triton kernels that avoid materializing the full attention matrix, delivering a 19.8x scoring-kernel speedup that translates to up to a 9.9% matched-batch end-to-end prefill speedup. Extensive evaluations on the RULER and LongBench benchmarks demonstrate that DropKV consistently outperforms state-of-the-art baselines at equivalent cache budgets, achieving the lowest inference latency among eviction methods while also remaining faster than the dynamic cache baseline.
[AI-25] Causal Integrated Gradients: Scalable Feature Attribution Along Causal Paths
链接: https://arxiv.org/abs/2610.22126
作者: Joran Michiels,Anis Ismail,Maarten De Vos,Alejandro Sifrim
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 41 pages (9 main text), 18 figures, 32 tables. Joran Michiels and Anis Ismail contributed equally. Code: this https URL
Abstract:Existing feature attribution methods that account for feature dependencies do not scale to high dimensions. We propose causal integrated gradients (CIG) and causal expected gradients (CEG), which extend integrated gradients to incorporate prior causal structure among input features while retaining its computational efficiency and, under conditions on the causal graph, its axiomatic guarantees. In many settings, some causal structure between features is already known, for example, that a transcription factor causally regulates a target gene’s expression. Yet existing gradient-based methods cannot exploit this structure, instead crediting a feature’s downstream effects over its true upstream cause, such as ranking the target gene above the transcription factor that regulates it. Our key idea is to embed this causal structure by routing the integration path along the causal graph. The method thereby decomposes each feature’s attribution into a direct effect and effects mediated through its causal descendants. We evaluate CIG in three settings of increasing realism. First, on a synthetic benchmark with a known causal model, CIG’s attributions best match the ground-truth causal effect from Pearl’s do-calculus, matching the accuracy of causal Shapley values at a fraction of the computational cost. Second, on simulated single-cell gene expression data with a known gene regulatory network (Dyngen), CIG most accurately recovers the transcription factors driving the simulation. Third, on real interferon- \beta stimulation data, where only an approximate causal graph is available, CIG better prioritizes the transcription factors driving the measured response than DeepSHAP and integrated gradients. Together, these results show that incorporating causal structure into gradient-based attributions recovers upstream drivers that existing methods miss, without sacrificing scalability.
[AI-26] PROBE: Learning to Audit Policy Compliance in Tool-Using LLM Agents NEURIPS2026
链接: https://arxiv.org/abs/2610.22124
作者: Kshitij Mishra,Abhijith Sharma,Nils Lukas,Salem Lahlou
类目: Artificial Intelligence (cs.AI)
备注: 27 pages, NeurIPS 2026
Abstract:Modern agentic systems extend language models with external tools, enabling them to complete complex tasks through multi-turn interactions. As these agents are deployed in consequential settings, they are often constrained by policies governing pricing, privacy, confirmation, safety, and other requirements. However, existing evaluation methods leave a critical gap: capability benchmarks measure what agents can accomplish, while content-safety red-teaming measures harmful generations, but neither tests whether agents comply with predefined policies during otherwise valid tool use. Such policy violations are difficult to audit because their effects may emerge only after several turns, span multiple tool calls, or surface in downstream systems. We propose PROBE (Policy Rule Observation via Behavioral Exploration), an algorithm for training LLM-based auditors that interact with deployed agents as realistic users and probe them for policy violations through multi-turn, tool-mediated dialogue. We formalize agent auditing as a two-player Markov game and propose a composite reward that balances violation discovery, behavioral plausibility, conversation progress, and policy diversity. To study training objectives and reward design, we compare GRPO, DPO, and GFlowNet-based optimization in a controlled setting. Across extensive experiments, auditors trained with PROBE uncover substantially broader and higher-yield policy violations than same-architecture zero-shot adversaries, even when the auditor is significantly smaller than the target agent. Our results show that multi-turn behavioral auditing is a distinct and necessary axis of agent evaluation, complementing capability and safety benchmarks. PROBE offers a scalable path toward automated policy auditing for real-world agentic systems and suggests a co-evolutionary loop in which auditors expose failures and agents improve from auditor feedback.
[AI-27] IPS: A Tiny Inference-time Policy for Scaling in Diffusion Language Models
链接: https://arxiv.org/abs/2610.22099
作者: Ya-Shi Zhang,Jian Tang,Pierre-André Noël,Oleksiy Ostapenko
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Masked diffusion language models expose an inference-time choice: which positions become visible after each denoising pass. Many existing decoding methods use hand-designed rules or require visibility to grow monotonically. We present a \textbfT iny \textbfI nference-time \textbfP olicy for \textbfS caling ( \textbfTIPS ), an approximately 600 K-parameter controller that learns both unmasking and remasking while keeping the underlying language model frozen. TIPS uses visibility-aware confidence features and a shallow transformer to select which tokens should be visible, while a fixed schedule controls the expected visible fraction. We train the controller through supervised warmup followed by reinforcement learning from terminal correctness rewards. On GSM8K and MATH-500 with LLaDA-8B-Instruct, the full-canvas and two-block TIPS variants achieve 58.46% and 57.97% micro-averaged accuracy at a budget of 32 decoder evaluations, exceeding the strongest baseline in our main comparison by 12.1 and 11.6 percentage points, respectively. Both variants are trained at a budget of 32 evaluations and remain competitive at higher budgets without retraining. These results demonstrate that learning when to reveal and reconsider tokens can improve the accuracy-compute trade-off without updating the decoder backbone.
[AI-28] Learning from Unverified Peers via Shared Residual Distillation
链接: https://arxiv.org/abs/2610.22080
作者: Feiyang Chen,Runze Ma,Zichen Wang,Jiong Lou,Zihan Wang,Shuangjia Zheng,Jie Li
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:A language model can generate several solutions to the same problem, some leading to correct answers and others to incorrect ones. Many post-training methods turn these trajectories into learning signals using supervision from reference answers, verifiable feedback, or external teachers. Without such supervision, learning from the model’s own generations becomes less straightforward, as their correctness cannot be reliably assessed. An incorrect solution may nevertheless contain useful intermediate reasoning, so its value as supervision need not be determined solely by its final answer. Can a model improve by learning from its unverified peer solutions without first determining which are correct? We propose Shared Residual Distillation (SRD), an on-policy self-distillation method that learns from the model’s own unverified peer trajectories. SRD compares how peers change the prediction distribution relative to a control, retaining adjustments that agree in direction and are conservative in magnitude. This yields supervision without treating any complete solution as correct, and the trained model requires no peers at inference time. Starting from the corresponding base models, SRD raises mean AIME24–26 avg@12 from 11.02 to 16.20 on Qwen3-1.7B, from 19.07 to 45.46 on Qwen3-4B, and from 21.39 to 54.44 on Qwen3-8B. Controlled ablations support the value of the shared-residual construction, and we observe no degradation relative to the base models on two general-capability benchmarks. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2610.22080 [cs.AI] (or arXiv:2610.22080v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2610.22080 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-29] When Reasoning Is Not Enough: Proactive Inquiry for Cross-Modal Safety in Multimodal Large Language Models
链接: https://arxiv.org/abs/2610.22068
作者: Wei Cai,Jian Zhao,Yuchen Yuan,Xuelong Li
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multimodal large language models (MLLMs) can act unsafely when individually benign images and texts are interpreted jointly. Beyond implicit reasoning risk, where reasoning drifts toward harm, we identify intent-level risk, where cross-modal ambiguity silently commits the model to a hazardous premise that path-level optimization alone cannot correct. We propose SAFE-PRO, which uses a Risk-Conditioned Uncertainty (RCU) signal, combining predictive entropy with a learned cross-modal risk score, to decide when to ask safety-grounded clarifying questions and where to apply path optimization, under a reward that prioritizes safety and penalizes over-refusal. We also build SSUI-Dialog, a multimodal multi-turn dataset with dual-intent supervision, and theoretically analyze gradient variance and unsafe-completion risk under RCU. Across seven benchmarks and two base MLLMs, SAFE-PRO consistently improves safety over SRPO, roughly halves its over-refusal, and surpasses GPT-4o and Claude-3.5-Sonnet in average single-turn safety.
[AI-30] Prism-1: Position- and Rival-Independent Scoring for System One Models
链接: https://arxiv.org/abs/2610.22066
作者: Yao Long Teng,Jianguo Huang,Yanchen Deng,Jiayi Cai,Bo An
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Given a state and a set of options, such as intents, tools or interface elements, System One decision models return a distribution over the options in one forward pass, without generated reasoning. Web systems increasingly build them by reading decisions from language models. Two axioms of rational choice are natural requirements for such decisions, because in web systems the listing and the option set are determined by other components and can change or be manipulated: the order in which the options are listed should not matter, and adding an option should not change the relative odds of the others (Luce’s choice axiom). We test the standard reads, namely single-choice question (SCQ) prompting, with shuffled-order training, order averaging or sorted options, and invariant set architectures, and find that each violates at least one axiom: one shuffle changes 12-20% of SCQ answers, sorting only freezes an arbitrary convention, and a well-chosen added option steers 4-34% of decisions. By Luce’s theorem, a read satisfies both axioms only if it is a softmax over scores that each depend on one option alone, so in a transformer they hold by construction only if no token that scores an option sees another. We implement this as Prism-1 (Position- and Rival-Independent Scoring for System One Models): every option is read in its own block at shared positions, with the model’s usual left-to-right attention, in one forward pass. On five benchmark sets with Qwen3-8B and Gemma 4 12B, SCQ prompting trained on the same data still changes 7% of its answers under one shuffle, and an added irrelevant option steers 12% of its decisions. Prism-1 stays within one point of that baseline’s accuracy and is 2-4 points more accurate once only answers that survive a reordering are counted, while no shuffle changes any of its answers and no added irrelevant option moves its choice among the others.
[AI-31] FAOPs: Normal Forms and an Algebra for AI Agent Workflows
链接: https://arxiv.org/abs/2610.22061
作者: Ali Mohammadi Shanghooshabad(JPMorgan Chase Co.)
类目: Artificial Intelligence (cs.AI); Databases (cs.DB); Programming Languages (cs.PL)
备注: 28 pages, 2 figures, 5 tables
Abstract:Organizations increasingly encode operational procedures as multi-step agent skills, here called Agent Operating Procedures (AOPs), that must be traceable, auditable, reproducible, access-controlled, and bounded in time and cost, requirements that regulated sectors make mandatory. Yet AOPs are typically kept as coarse-grained files, in which steps bundle actions, copied procedures drift apart, any step may reach any tool, and refactoring goes unchecked. We present Functional Agent Operating Procedures (FAOP), a formal theory that brings database-grade discipline to AI agent workflows. FAOP recasts each workflow as a typed function over scoped state and subjects an entire library to four progressive normal forms: Decomposed (FNF-1) splits a workflow into atomic steps, Typed (FNF-2) makes its data flow explicit and type-checked, Factored (FNF-3) gives each shared procedure a single canonical definition so a change propagates instead of drifting, and Scoped (FNF-4) confines every step to least-privilege tool access. Once normalized, every atomic step is a separate entity with a declared contract and can be type-checked, scope-confined, bounded, tested, traced, and audited on its own. A FAOP algebra of provably sound equalities turns refactoring and parallelization into compiler rewrites; orthogonal extensions add verified termination (+V), optimization (+O), and time and cost bounds (+B); and an architectural-level classifier (0-3) determines whether a workflow needs AI at all. Together, these make integrity, isolation, typing, termination, cost, and optimization compile-time concerns across the whole library, not runtime hopes. We implement the theory in SkillBase, an AOP management system (what PostgreSQL is to data, SkillBase is to AOPs), together with a separate executor that enforces these guarantees at run time. A worked financial workflow illustrates the approach end to end.
[AI-32] Progress Reward Model for Reinforcement Learning via Large Language Models NEURIPS2025
链接: https://arxiv.org/abs/2610.22056
作者: Xiuhui Zhang,Ning Gao,Xingyu Jiang,Yihui Chen,Yuheng Pan,Mohan Zhang,Yue Deng
类目: Artificial Intelligence (cs.AI)
备注: 20 pages, 10 figures. Published at NeurIPS 2025
Abstract:Traditional reinforcement learning (RL) algorithms face significant limitations in handling long-term tasks with sparse rewards. Recent advancements have leveraged large language models (LLMs) to enhance RL by utilizing their world knowledge for task planning and reward generation. However, planning-based approaches often depend on pre-defined skill libraries and fail to optimize low-level control policies, while reward-based methods require extensive human feedback or exhaustive searching due to the complexity of tasks. In this paper, we propose the Progress Reward Model for RL (PRM4RL), a novel framework that integrates task planning and dense reward to enhance RL. For high-level planning, a complex task is decomposed into a series of simple manageable subtasks, with a subtask-oriented, fine-grained progress function designed to monitor task execution progress. For low-level reward generation, inspired by potential-based reward shaping, we use the progress function to construct a Progress Reward Model (PRM), providing theoretically grounded optimality and convergence guarantees, thereby enabling effective policy optimization. Experimental results on robotics control tasks demonstrate that our approach outperforms both LLM-based planning and reward methods, achieving state-of-the-art performance. Code is available at this https URL.
[AI-33] Decoding Naturalistic Speech from fMRI with Temporal Modeling and Multimodal Alignment
链接: https://arxiv.org/abs/2610.22049
作者: Corentin Caris,Louis Jalouzot,Nona Rajabi,Géraud Ilinca,Ahmed Hassayoune,Asilbek Muminov,Thaddée Delebarre,Clara Fonteneau,Charles Maussion,Bertrand Thirion,Alexis Thual,Antoine Collas
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reconstructing what a person perceives from their brain activity is a stringent test of whether neural recordings carry meaningful information. Naturalistic speech offers a rich testbed for this challenge: it unfolds over multiple timescales, but fMRI’s noisy, temporally smeared measurements make this structure hard to recover. We introduce NOEMA, a temporal convolutional decoder that aligns windows of consecutive fMRI measurements with multimodal stimulus embeddings. These embeddings combine complementary text, audio, and speech representations of the same stimulus segment, enabling its retrieval directly from brain activity. On two naturalistic speech datasets, Cephalonauts One and LeBel, NOEMA retrieves the correct segment among 2,000 candidates within the top 10 for more than 42% of held-out queries. This substantially outperforms baseline decoders, an advantage that persists when the candidate pool is extended with thousands of distractors absent from training. Controlled ablations establish the contributions of temporal modeling and complementary stimulus representations.
[AI-34] GraphACM: Graph-based Action Consistency Measurement for Uncertainty Quantification in LLM Agents
链接: https://arxiv.org/abs/2610.22031
作者: Yaxin Ma,Wenbo Zhang,Satya Krishna Pothapragada,Alina Zare
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Behavioral consistency across repeated executions provides an informative signal for assessing the reliability of large language model (LLM) agents. However, existing approaches typically summarize consistency with a single task-level score, without identifying where behavioral differences occur or whether actions are compared under equivalent decision contexts. We propose a graph-based framework for fine-grained behavioral consistency measurement and uncertainty quantification. Given multiple rollouts of the same task, we merge equivalent states into a directed, action-labeled graph, enabling action comparisons at shared decision states. We introduce node-level and global measures of action consistency, using the global consistency score as a proxy for uncertainty to predict task success or failure. Experiments across three LLMs on ALFWorld, AppWorld, and WebShop demonstrate that our measure is positively associated with task success and achieves an average AUROC of 0.811 for task-success prediction, outperforming the strongest behavioral-consistency baseline (0.729) and verbalized confidence (0.769). These results highlight the effectiveness of state-conditioned behavioral consistency and pinpoint where agent actions agree.
[AI-35] DemoStream: Batched Task-and-Motion-Plan Optimization for Automated Learning from Demonstration
链接: https://arxiv.org/abs/2610.22020
作者: Caelan Garrett,Neel Jawale,Fabio Ramos
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Learning from demonstration is an effective strategy for endowing robots with manipulation capabilities but requires training on datasets comprised of massive amounts of demonstrations. Existing demonstration collection approaches either require constant human involvement, for example, via teleoperation or are specialized for a single task. Task and Motion Planning (TAMP) systems are able to autonomously solve and therefore generate demonstrations for general-purpose tasks; however, learning from their demonstrations has proven difficult due to multi-modality in their solutions. In this work, we build on recent work leveraging GPU acceleration for TAMP and propose DemoStream, a TAMP system that optimizes demonstration quality in order to improve behavior cloning success rates. First, we introduce batched TAMP sampling operations along with a new TAMP algorithm that correspondingly performs batched constraint satisfaction in order to explore many solutions in parallel. We conduct simulation experiments ablating different action spaces, cost functions, and batch sizes to analyze the effect of these choices on downstream policy success rates. Finally, we show that we can leverage the safety guarantees of TAMP to directly collect real-world demonstrations at an average rate of 66 demos per hour. See our website for policy rollout videos: this http URL.
[AI-36] GraphBEAT: Graph-based Bucketized Event Aggregation Over Time for Early Risk Stratification of Gastrointestinal Cancers
链接: https://arxiv.org/abs/2610.22014
作者: Leo Y. Li-Han,Chris Varghese,Elizabeth B. Habermann,Mark J. Truty,William R.G. Perry,Cornelius A. Thiels,Hojjat Salehinejad
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Signal Processing (eess.SP)
备注:
Abstract:Longitudinal electronic health records (EHRs) contain complementary diagnostic and laboratory information about patient health. However, irregularity, sparsity, heterogeneity, and incompleteness in documentation complicate EHR modeling. To address the problem, we introduce GraphBEAT (Graph-based Bucketized Event Aggregation over Time), a temporal graph representation learning framework that groups clinical events into time buckets and transforms observed features into patient-specific, temporally directed graphs to enable early risk stratification for gastrointestinal cancers. Each node in an EHR graph encodes feature identity, observed value, and temporal position of clinical events. A context-aware, time-decay message-passing mechanism models transitions between successive buckets, followed by multi-head attention pooling to generate patient-level representations. GraphBEAT was trained and evaluated on multi-site longitudinal diagnosis codes and laboratory tests to predict pancreatic, liver, and intestinal cancer at 12-, 24-, and 36-month diagnostic lead times. In all experimental setups, GraphBEAT consistently outperformed baseline graph neural networks (GraphSAGE, GAT, and GCN) and sequence modeling approaches (Transformer and long short-term memory [LSTM]). GraphBEAT also required over 250x fewer floating-point operations per patient and over 20x fewer parameters than the Transformer and LSTM baselines. These results support GraphBEAT as an accurate and computationally efficient approach for cancer risk stratification from sparse, heterogeneous longitudinal health records.
[AI-37] OntoAct: Hypothesis-Based Neuro-Symbolic Agents through Ontological Abstraction
链接: https://arxiv.org/abs/2610.21983
作者: Donghyeok Lee,Bong Gyun Kang,Junyong Ahn,Sungroh Yoon
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 49 pages, 13 figures, 33 tables. Preprint, under review
Abstract:Learning from experience requires recovering reusable knowledge and determining when it applies to a new situation. We introduce OntoAct, a hypothesis-based neuro-symbolic framework built on ontological abstraction. OntoAct turns individual experiences into axioms over entity types, relations, and conditions, making their underlying patterns reusable across situations. A large language model (LLM) uses this world knowledge to propose hypotheses and actions, while a symbolic verifier assesses their supporting conditions against current observations. Verification identifies missing evidence, and new observations and intervention outcomes update hypothesis assessments and guide subsequent actions. We also introduce Tacit-FLE, a benchmark built on the Factorio Learning Environment (FLE) for diagnosing and repairing hidden faults in existing factories using inherited operational records. Its synthetic maintenance corpus distributes evidence across records, and an executed repair may leave the fault unresolved, requiring agents to reassess their diagnoses through interaction. Across Tacit-FLE, Mars, and ALFWorld, OntoAct achieves leading task performance with efficient interaction and inference.
[AI-38] me to Fake It: Continual Machine Learning Intensifies Unfairness in Deepfake Detection
链接: https://arxiv.org/abs/2610.21974
作者: Kajaree Das,Oliver Böse,Subarnaduti Paul,Gabriel Storch,Martin Mundt
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Deepfake detectors safeguard against misinformation, rendering their failures as much a societal as a technical concern. As generators constantly evolve, detectors trained on static datasets miss newly arising threats, while progressive training interferes catastrophically. Yet, mitigating forgetfulness in terms of accuracy, the prevalent focus of continual learning (CL), is insufficient to reflect real-world impact. We therefore scrutinize the effects of CL methods with respect to evolving fairness. To this end, we analyze four dimensions of fairness across diverse demographics for detectors trained progressively on a chronological sequence of 11 real-world deepfake generation methods. Tracking fairness metrics over time, we expose non-trivial interplays between performance and fairness that depend on both the choice of CL algorithm and demographic intersections. We find accuracy degradation to typically exacerbate unfairness and, conversely, fairness improvements to remain scarce over time. Dissecting the CL method families (rehearsal, prompts, regularization), we observe that inequitable forgetting is not explained by underrepresentation alone, but likely entangled with representation learning. We thus issue a call to action for CL methods to take social outcomes into account when balancing the stability-plasticity tradeoff.
[AI-39] Why Shallow Neural Networks Learn Sinusoidal Components First: An Orthogonal Polynomial Dynamics Framework DATE
链接: https://arxiv.org/abs/2610.21964
作者: Yanyang Xiao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Spectral Theory (math.SP); Machine Learning (stat.ML)
备注: 15 pages, 0 figures. The main draft dates back to 2019, with a new abstract and introduction section
Abstract:Neural networks often learn low-frequency components before high-frequency ones — a phenomenon called spectral bias or termed F-principle. Many numerical studies have demonstrated this phenomenon, and theoretical studies have sought to explain it. However, several goals remain insufficiently addressed: (1) explaining spectral bias asymptotically, including its origin and why the preferred components are generally sinusoidal; (2) extending the explanation to general activation functions, such as sigmoid and ReLU; (3) making weak assumptions about the data distribution; (4) avoiding aliasing effects, such as those arising in finite-domain Fourier analysis; and (5) developing a theoretical framework applicable to arbitrary networks. We address these goals theoretically by analyzing the eigenvalue decay and eigenvector shapes of the dynamics matrix H, which arises when we expand the network output as a general Fourier series (GFS) in an orthogonal polynomial basis and express the gradient-descent dynamics of the GFS coefficients as ordinary differential equations. The graded structure of H enables an intuitive explanation of spectral bias, while orthogonal polynomial theory helps us analyze how learning bias depends on the activation function and, potentially, the network architecture.
[AI-40] CodeHealth Predicts the Token Cost of Agent ic Code Review: A Pre-Registered Study of 465 Agent -Authored Pull Requests
链接: https://arxiv.org/abs/2610.21914
作者: Lo Gullstrand Heander,Markus Borg,Emma Söderberg,Christofer Rydenfält
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 32 pages, 6 figures, 16 tables
Abstract:Context: Organizations increasingly delegate code review to multi-agent large language model (LLM) pipelines billed per token. Low code quality raises human maintenance effort, but its effect on agent review cost has not been measured. Objective: We test whether baseline CodeHealth predicts the token cost of agentic review and fix beyond pull request (PR) footprint. Method: In a pre-registered observational study, we sampled 465 agent-authored PRs from 246 repositories in AIDev, jointly stratified by CodeHealth and footprint under repository caps. A frozen five-agent review-and-fix pipeline processed every PR with three different LLMs. Per LLM, we fitted a mixed model of log tokens on CodeHealth and log PR footprint, with a repository random intercept. Results: For all three LLMs the CodeHealth coefficient was negative, with 95% CI excluding zero; each CodeHealth point was associated with 3.5% to 6.1% lower token cost. CodeHealth was unrelated to PR footprint, so PR size does not explain the association. Two of three registered smells predicted fix cost; batch fixing left median CodeHealth unchanged. Exploratory analyses located the association in the review stage, where agents iterated more on unhealthy code without writing more comments. Adjusting for the size of the scored files absorbed the association, which we interpret as mediation by file size. Conclusions: A quality metric calibrated against human maintainability also predicts the inference cost of multi-agent review and fix. The effect is modest per point and not carried by PR size, so the refactoring case extends to agent-reviewed code when it changes structure and file size, not only the score. Data Availability: this https URL Comments: 32 pages, 6 figures, 16 tables Subjects: Software Engineering (cs.SE); Artificial Intelligence (cs.AI) Cite as: arXiv:2610.21914 [cs.SE] (or arXiv:2610.21914v1 [cs.SE] for this version) https://doi.org/10.48550/arXiv.2610.21914 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Lo Gullstrand Heander [view email] [v1] Thu, 15 Oct 2026 14:00:33 UTC (162 KB) Full-text links: Access Paper: View a PDF of the paper titled CodeHealth Predicts the Token Cost of Agentic Code Review: A Pre-Registered Study of 465 Agent-Authored Pull Requests, by Lo Gullstrand Heander and 3 other authorsView PDFHTML (experimental)TeX Source view license Additional Features Audio Summary Current browse context: cs.SE prev | next new | recent | 2026-10 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?) 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-41] CoWorld Reason er: Agent ic Multimodal Reasoning with the Evolving Video World Model-Driven Harness
链接: https://arxiv.org/abs/2610.21910
作者: Dongjie Cheng,Ruifeng Yuan,Ran Wei,Ning Song,Yongqi Li,Wenjie Li
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As Vision-Language Models (VLMs) advance, they are increasingly expected to perform spatiotemporal reasoning over motion, actions, and view changes beyond what the input images show. Video World Models (VWMs) provide a natural unified engine for visual imagination by generating task-relevant visual content as additional evidence. However, existing approaches typically invoke a VWM through fixed pipelines centered on a predefined form of visual imagination, such as novel-view synthesis. To support flexible visual imagination for diverse reasoning tasks, we introduce CoWorld Reasoner, an agentic framework that uses a VWM as a unified visual reasoning engine. Its Video World Model-Driven Harness exposes multiple visual operations through tools and uses skills to guide the VLM in deciding whether to imagine, selecting an operation, constructing a generation request, and interpreting the resulting visual evidence. We further introduce VLM-VWM Harness Co-Evolution, which jointly revises tools, skills, and execution code based on execution feedback while keeping both model weights fixed. Across seven benchmarks spanning spatial intelligence, temporal and causal reasoning, and perceptual reasoning, CoWorld Reasoner outperforms direct reasoning with the same VLM. The strongest variant improves on all seven benchmarks, including relative accuracy gains of 11.8% on MMSI and 13.0% on ERQA, and outperforms all five evolving baselines using the same backbone across five benchmarks.
[AI-42] In-Context Learning Before Specialisation: Theory of Learning in a Solvable Topic Model
链接: https://arxiv.org/abs/2610.21901
作者: Francesco Cagnetta,Casper L. Christensen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:Sequence models can generalise in fundamentally distinct ways: extrapolating structures seen in training, or using their prompt to adapt to novel tasks. The prevailing strategy determines how the model behaves beyond its training distribution, controlling potential capabilities and risks. Recent work has characterised many different strategies theoretically, and empirically described the transition between them as the quality and quantity of pretraining data vary. However, there is no learning theory for this competition, hence no fundamental understanding of \textitwhy one solution might be favoured over another. Here we provide such a theory in a solvable topic model—documents drawn from a finite mixture of latent topics, mirroring the heterogeneity of modern pretraining data. The dataset comes with two distinct solutions: a specialised predictor, that stores the finite topic mixture, and an in-context one, optimal for topics never seen during training. Our theory predicts their emergence at distinct sample complexities: the in-context predictor is available early, while the specialised one emerges only after the training tokens outnumber the parameters of the data-generating process. Trained transformers reproduce both scales and the separating crossover in remarkable agreement with the theory, with no fitted parameters.
[AI-43] Provably Efficient Risk-Sensitive Offline Reinforcement Learning via Optimized Certainty Equivalents
链接: https://arxiv.org/abs/2610.21870
作者: Hanzhen Wang,Boxiang Lyu,Zhongxiang Dai,Siyang Gao,Dake Zhang,Shuang Qiu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Optimization and Control (math.OC); Machine Learning (stat.ML)
备注:
Abstract:Risk-sensitive reinforcement learning (RL) addresses decision-making under uncertainty and adverse outcomes. Learning risk-sensitive policies from fixed datasets is important when additional interaction with the environment is costly. Existing work on risk-sensitive RL focuses on online settings or specific risk measures, leaving a general theoretical understanding of its offline setting underdeveloped. In this work, we investigate efficient offline risk-sensitive RL using the optimized certainty equivalent (OCE), a general class of risk measures. We consider two formulations: dynamic-OCE RL, which applies OCE recursively at each step, and static-OCE RL, which applies a single OCE to the cumulative rewards. Under the tabular setting, we develop pessimistic value-iteration algorithms for both formulations and establish tight sublinear suboptimality bounds, explicitly capturing how risk sensitivity, encoded by utility functions, affects learning difficulty. Moreover, we establish minimax suboptimality lower bounds to characterize the problem’s fundamental limits. To handle large state spaces, we further develop pessimistic least-squares value-iteration algorithms that incorporate the general OCE risk criterion into linear function approximation. We establish utility-dependent sublinear guarantees with explicit dependence on the feature dimension d . To our knowledge, we provide the first offline algorithms along with the first upper and lower suboptimality bounds for both dynamic- and static-OCE RL.
[AI-44] CIM-Embed: Learning Instruction-Conditioned Multimodal Embeddings for Clustering
链接: https://arxiv.org/abs/2610.21805
作者: Yinghua Xu,Pengyang Xie,Yanxiang Zhou,Chuanzhi Tong,Xiao Xiao,Chaoqun Feng,Xiaoqing Wang
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 40 pages, including references and appendices
Abstract:Multimodal objects can require different partitions under different instructions, as when products in an e-commerce catalog are grouped by color, material, or both. Retrieval-oriented embeddings may not recover these partitions, while generative clustering often relies on intermediate descriptions or iterative decisions. This paper presents CIM-Embed, a framework for learning instruction-conditioned multimodal embeddings for clustering. Supervised contrastive learning uses same-group positives and negatives from other groups, optionally including other tasks; heterogeneous structured batches accommodate tasks with different group counts. At inference, the model maps each object’s image, text, and instruction directly to an embedding; HDBSCAN then forms a partition without the ground-truth cluster count. We construct ProductMIC from product metadata in Amazon Reviews 2023, comprising training data and an evaluation benchmark for instruction-conditioned clustering. The benchmark contains 5,373 products and 449 single-attribute and compositional grouping tasks spanning 50 leaf categories excluded from our training. With full-dimensional embeddings and HDBSCAN, direct within-task SupCon training achieves a mean per-task V-measure of 95.01, compared with 53.06 for WeMM 9B, the strongest general-purpose embedding baseline evaluated. Frozen product-trained encoders also improve out-of-domain conditional grouping, while controlled transfer experiments distinguish partition transfer from instruction response. We will release the code and dataset upon acceptance.
[AI-45] Co2In: a Bi-level Memory Incremental Learning Framework with Knowledge Encoding Consolidation and Integration
链接: https://arxiv.org/abs/2610.21804
作者: Wenju Sun,Qingyong Li,Boyang Li,Wen Wang,Siyu Zhang,Yi Wei,Yangliao Geng
类目: Artificial Intelligence (cs.AI)
备注: Accepted by IEEE Transactions On Pattern Analysis and Machine Intelligence
Abstract:Incremental learning (IL) aims to continually acquire new knowledge (plasticity) while retaining previously learned information (stability). However, striking a balance between plasticity and stability remains a significant challenge for intelligent systems. The human brain achieves exceptional balance, owing to various memory units that collaboratively encode and store information. Inspired by human memory mechanisms, this paper introduces an IL framework with knowledge enCoding, Consolidation, and Integration (Co2In). Co2In is designed as a bi-level memory architecture: a working memory for adaptive knowledge acquisition and a long-term memory dedicated to the persistent retention of information. The two memory modules work cooperatively during IL. The working memory first learns from new data to encode knowledge into parameters. Subsequently, Co2In performs a consolidation process to identify underlying patterns in the learned parameters and re-express them into a compact knowledge representation. Next, the knowledge representation and the identified patterns are transformed into separate network layers and integrated into the long-term memory. These designs empower Co2In to accumulate knowledge with high plasticity and stability. We evaluate Co2In on CIFAR-10, CIFAR-100, and Tiny-ImageNet under exemplar-free Class-IL and Task-IL settings. Experimental results show that Co2In achieves state-of-the-art performance with efficient memory consumption.
[AI-46] Cross-Embodiment Causal Transfer from Human Videos to Robot Manipulation
链接: https://arxiv.org/abs/2610.21798
作者: Xingyu Zhang,Yingchao Li,Peizheng Guo,Jie Hu,Changwen Zheng,Wenwen Qiang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Scaling robot manipulation models requires large amounts of interaction data, yet collecting diverse real-robot experience remains costly. Human manipulation videos, in contrast, provide abundant and diverse task experience that could potentially benefit robot learning. However, differences in morphology, action spaces, and control mechanisms make human motions unsuitable as direct supervision for robot actions, raising a scientific question: what knowledge in human videos should be transferred across embodiments to robot manipulation? To address this question, we analyze subtask transfer through the lens of causal beams, focusing on which contextual conditions and interaction dependencies should be retained to support local task effects. Based on this perspective, we propose Causal-Beam-Guided Subtask Transfer (CBST). CBST learns manipulation-related representations from human videos and grounds them in task-level semantic variables using robot demonstrations. Within a predefined manipulation model, causal-beam constraints guide contextual information selection by requiring retained assignments to sustain local mechanism outputs while preserving nontrivial responsiveness to retained inputs. The selected semantic information conditions a world-action model, allowing the robot to generate actions from its own observations without requiring human-robot motion correspondence. Experiments on robot manipulation benchmarks show improvements in task success over the corresponding world-action model baselines.
[AI-47] RSIFlow: Adaptive Self-Improvement Workflows with an Evolving Meta-Agent
链接: https://arxiv.org/abs/2610.21791
作者: Huan Zhang,Mingju Chen,Dongxu Zhou,Can Lv,Sen Cui,Faguo Wu,Shiji Zhou
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents have demonstrated strong capabilities in complex tasks, including code generation, tool use, and multi-hop question answering. Existing self-improvement methods enhance agent performance through model optimization, harness evolution, and executable artifact refinement, while recursive self-improvement (RSI) further extends optimization to the improvement mechanisms themselves. However, adapting self-improvement workflows to evolving performance bottlenecks while improving workflow orchestration through execution feedback remains a key challenge. We propose \textbfRSIFlow, a \textbfRecursive \textbfSelf-\textbfImprovement framework that integrates adaptive work\textbfflow orchestration with meta-agent capability evolution. RSIFlow diagnoses performance bottlenecks from task-agent trajectories and environmental feedback to dynamically select component-specific improvement workflows. The meta-agent then distills workflow outcomes into component-specific skills and cross-component principles, evolving its own harness to improve subsequent workflow orchestration. Experiments across six benchmarks and two backbone models show that RSIFlow achieves the SOTA average performance on all benchmarks. Ablation studies further demonstrate the contributions of adaptive workflow orchestration and meta-harness evolution. Code is available at this https URL.
[AI-48] CF-AVSS: A Coarse-to-Fine Audio-Visual Speech Separation Framework with Target-Speaker Refinement and Post-Fusion for Real-World AVSE Challenge
链接: https://arxiv.org/abs/2610.21773
作者: Longjie Luo,Zhongshu Hou,Yuepeng Li,Gang Yang,Shidong Shang,Fan Zhou
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注: Accepted by ISCSLP 2026
Abstract:Audio-visual speech enhancement (AVSE) leverages visual cues to recover target speech from mixtures corrupted by reverberation, noise, and overlapping speakers. We propose a coarse-to-fine two-stage AVSE framework. In the first stage, a progressive SPAV-TFGridNet is extended to two synchronized visual streams to produce visually ordered coarse estimates by progressively suppressing interfering speech while preserving target reverberation and shared background noise. In the second stage, a TFGridNet-based target speaker refinement network takes the original mixture and both coarse estimates as input to further recover target speech and improve signal quality. The coarse and refined estimates are finally combined using development-set-calibrated fusion. On the ISCSLP 2026 Real-World AVSE Challenge, our system tied for first place on Track 1 and ranked second on Track 2 in OVRL, while achieving the lowest CER on both tracks at 12.31% and 14.79%, respectively.
[AI-49] From Research Goals to Scientific Closure: Evaluating Agents Through Evidence-Supported Physics Research
链接: https://arxiv.org/abs/2610.21767
作者: Yijia Mao,Song Dai,Tao Chen,Xiangyu Lu,Linqian Zeng,Chenxi Hu,Sitong Zhou,Dongyang He,Xiaodan Mao,Linxuan Zhang,Tong Zhu,Yafu Li,Fangchen Yu,Peng Ye,Feng He,Yu Cheng,Yun Luo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Can AI agents organize scientific investigations that lead to well-supported conclusions? To evaluate this capability, we introduce PhysClosureBench, a benchmark of 50 expert-verified, goal-specified research tasks spanning strong-field, attosecond, and neighboring physics areas. Each task specifies the research objective and essential scientific constraints, while leaving the agent to organize the investigation and determine how to generate and validate the evidence needed to answer the question. A 20-criterion Atomic Score measures local scientific accomplishment, while Research Closure (RC0–RC4) assesses how fully the investigation resolves the specified research question based on whether its indispensable evidence requirements and applicable scope conditions are satisfied. We evaluate seven frontier LLMs, where Claude-Opus-5 achieves the highest mean Atomic Score (6.29/10) but reaches RC4 in only 32.7% of attempts. Across the benchmark, 36.2% of attempts scoring at least 8/10 still fall short of RC4, while all seven models reach RC4 on substantially fewer tasks in all three runs than in at least one. Case-level analysis further illustrates how misinterpretation of generated results and unresolved physical applicability limits can lead to incomplete closure. Together, these results reveal a persistent gap between AI agents’ local scientific accomplishments and their ability to achieve complete research closure.
[AI-50] Harmful Partial Interventions and Benign Leakage in Concept-based Models
链接: https://arxiv.org/abs/2610.21750
作者: Gregory Verghese,Christopher J. Soelistyo,Mateo Espinosa Zarlenga,Alessandro Barp,Christopher R.S. Banerji
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Artificial intelligence systems deployed in high-risk settings increasingly require models that support meaningful human oversight. Concept-based models (CMs) offer a promising framework by enabling users to inspect and intervene on intermediate, human-understandable concepts. However, concept interventions can counterintuitively worsen downstream predictions, a failure mode we formalise as intervention harm. We distinguish harm after all concepts have been intervened on from harm under partial interventions, the more likely practical use case. Modern CMs often achieve high accuracy by encoding, or leaking, additional task-relevant information in their intermediate concept representations. Recent work argues that such leakage can be benign with respect to intervention harm when concept representations satisfy two conditions: sufficiency and localisation. We prove that these criteria are not sufficient to guarantee that concept interventions improve downstream performance. Across controlled and real-world settings, we find that concept supervision incompleteness is associated with increased intervention harm. Importantly, optimising the model when all concepts are intervened on, a previously proposed leakage mitigation objective, reduces endpoint harm but does not reliably prevent harmful partial corrections. Hence, we introduce a complementary monotonic intervention objective that directly penalises increases in task loss along partial intervention trajectories. Combining endpoint and monotonic intervention training substantially reduces the prevalence and magnitude of partial intervention harm while preserving concept and task predictive performance.
[AI-51] In-Context Preference-Conditioned Pareto Set Learning from Limited Offline Data
链接: https://arxiv.org/abs/2610.21736
作者: Yiming Yao,Liang Zhao,Qingfu Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-objective optimization involves diverse preferences over conflicting objectives. Rather than returning a finite set of trade-off solutions, preference-conditioned Pareto set learning seeks to map a specified preference directly to a corresponding Pareto optimal solution. Yet in many real-world scenarios, optimization must proceed offline without further objective evaluations, leaving only a limited static dataset that sparsely covers the decision and objective spaces. Existing methods either optimize separately for each preference or train a problem-specific Pareto solution model, requiring repeated search across preferences or repeated training across problems. We introduce in-context preference-conditioned Pareto set learning (ICPSL), a framework that amortizes both preference-wise solution search and problem-specific model construction. ICPSL pretrains a foundation-style hypernetwork on a large and diverse set of synthetic prior data. Given the offline observations of an unseen problem, the pretrained hypernetwork learns to generate a problem-specific, preference-conditioned Pareto solution model in a single forward pass. The generated model can then be reused to predict solutions for arbitrary preferences without problem-specific training or preference-wise search. Experiments on synthetic and real-world problems show that ICPSL produces high-quality Pareto solutions while providing substantial speedups in problem-specific solution-model construction.
[AI-52] Encoded but Exposed: Language Embeddings can Reveal More Than They Should in Recommender Systems
链接: https://arxiv.org/abs/2610.21713
作者: Linda Beková,Rodrigo Alves
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Language embeddings are increasingly used in recommenders to represent item content, but their privacy risks remain poorly understood. We study whether item embeddings can reveal item identities when an adversary has access to auxiliary item text from external sources. We show that a simple cosine-similarity attack can recover item identities with high mean reciprocal rank across language models and candidate-set settings on two real-world datasets. Leakage is strongest when the adversary’s partial information overlaps with the victim representation, but it also persists through semantic alignment between independently sourced descriptions. We further show that item leakage can propagate to user re-identification from sparse ratings. Finally, we evaluate obfuscated descriptions as a mitigation and find a clear leakage-utility tradeoff.
[AI-53] Router Drift Control: Mitigating Off-Policy Degradation in Mixture-of-Experts Reinforcement Learning
链接: https://arxiv.org/abs/2610.21682
作者: Liangyao Jin,Zhixin Wang,Liming Liu,Mingjun Zhang,Zhengzhou Zhu,Fangcheng Fu,Jinlong Hou,Yuan Cheng,Xiaohe Hu,Ping Zhang,Siyuan Feng
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning (RL) can become off-policy due to training-inference mismatch, mini-batch updates, and asynchronous this http URL resulting mismatch is especially harmful in Mixture-of-Experts (MoE) models, where rollout–training discrepancies can change routing decisions and lead to router drift. Rollout Routing Replay (R3) mitigates this mismatch by reusing expert selections from the rollout stage during training. Our analysis shows that routing replay introduces gradient bias by losing contributions from experts selected by the training router but not the rollout router. To address this, we propose Router Drift Control (RDC), which uses routing disagreement between rollout and training as a signal of off-policy staleness and selectively rejects high-drift responses. The sequence-level filtering avoids gradient bias introduced by replay while preserving native expert selection and the associated learning signals for retained responses. Experiments show that RDC achieves larger gains as policy staleness increases and remains effective in asynchronous RL. In asynchronous training, RDC combined with Truncated Importance Sampling (TIS) improves the average peak validation accuracy across AIME 2024 and AIME 2025 by 5.5 points over TIS and 4.5 points over R3.
[AI-54] Data World: Scaling Data Environments from Scratch
链接: https://arxiv.org/abs/2610.21656
作者: Zhanli Li,Zeyuan Xu,Chaoxu Pang,Zeyuan Liu,Anyang Ji,Te Sun,Kongcheng Zhang,Wenbo Xu,Xubin Yue,Shipeng Luo,Junjie Zhao,Yingfan Hua,Jingqing Ruan,Zhongda Su,Feng Hong,Zhizhao Zeng,Ke Zeng,Xunliang Cai,Hanjia Ye
类目: Artificial Intelligence (cs.AI)
备注: 18 pages, 4 figures, 7 tables. Code: this https URL Data: this https URL Project page: this https URL
Abstract:Data agents use code to process heterogeneous files and derive analytical insights. Scaling their training faces costly environment curation, privacy and intellectual-property restrictions, and difficulty verifying open-ended findings without known generative truth. Our core insight is to specify the world behind the files before generating the files themselves. We introduce Data World, a multi-agent engine that synthesizes environments, tasks, and rubrics from explicit mechanisms. Independent audits check target recoverability from observed files; rubrics support evaluation and RL rewards. Human preference-based curation for question realism selects 100 tasks across 50 held-out worlds and 16 domains for DataWorld-Bench. The engine also yields 2,135 training tasks from 1,037 disjoint worlds. Evaluation of ten frontier models shows that the benchmark remains challenging, particularly for insight discovery, with a consistent computation-insight gap. RL-zero training on the synthesized tasks produces DataWorld-35B-A3B from Qwen3.6-35B-A3B without trajectory distillation. Under matched evaluation, it improves over its base by 7.06 points on DataWorld-Bench, 14.28 on DAComp, 20.90 on DABstep-Hard, and 12.13 on JobBench, a broader occupational benchmark.
[AI-55] Efficient Pairwise Difference Learning: A Comparative Study of Encoding Strategies and Training Pair Selection
链接: https://arxiv.org/abs/2610.21644
作者: Bernard Georges,Mohamed Karim Belaid,Manuel Rodrigues
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 13 pages. Preprint (version submitted before peer review) of the paper published in WorldCIST 2026, LNNS vol. 2092, Springer
Abstract:Pairwise Difference Learning for Classification (PDC) enhances supervised learning in low-data regimes by training a binary similarity classifier on pairs of instances rather than individual samples, effectively inflating the training set from n to O(n^2) examples. However, the PDC framework leaves critical implementation choices underspecified, particularly regarding input representation and pair selection strategies. This work systematically investigates these design decisions through extensive empirical evaluation. We compare four feature encoders (paired features, triplet with relative difference, absolute difference, and a novel triplet with absolute difference) alongside four pair selection strategies ranging from exhaustive to kernel-thinning-based schemes. Experiments across nearly 200 benchmark classification datasets using Decision Tree, Multilayer Perceptron, Gaussian Naive Bayes, and Support Vector Classifier as base models reveal that the proposed triplet with absolute difference encoder ( \phi_\mathrm3dabs ) achieves the highest average F1 scores. More notably, restricting training to upper-triangle pairs reduces computational time by approximately 40% with negligible accuracy loss. Kernel thinning-based selection provided no accuracy benefits while substantially increasing computational overhead. Overall, PDC configurations outperformed base models in 94% of tested cases, though no single configuration dominated universally, suggesting dataset-specific optimization remains an open challenge.
[AI-56] Agent Perturb: Evaluating LLM Agent Robustness under Controlled Interaction Perturbations
链接: https://arxiv.org/abs/2610.21624
作者: Yuanjian Xu,Jinyi Han,Jianing Hao,Yuxuan Sun,Fanyang Lu,Richeng Xuan,Guang Zhang,Zhichao Hu
类目: Artificial Intelligence (cs.AI)
备注: 25 pages, 10 figures. Project repository: this https URL
Abstract:Real-world tasks are often complex and noisy, posing substantial challenges for LLM agents. Yet existing evaluations lack a systematic framework for jointly characterizing potential perturbation targets and types. We use this two-dimensional view to conduct a unified study that quantifies the robustness of current frontier models, traces the mechanisms behind perturbation-induced failures, and draws implications for robustness-oriented training. Through controlled evaluations on APEX-Agents and Toolathlon, we find that perturbations consistently degrade performance, while the dominant vulnerabilities vary across task environments. Trajectory-level analysis reveals that agents rarely recognize altered inputs and often propagate corrupted information without verification. A focused cost analysis further shows that successful perturbed trajectories can require substantially more computation. These findings motivate training objectives that cover diverse forms of noise and reward verification, uncertainty awareness, and recovery. We provide a rigorous benchmark-construction pipeline. Code and data will be released at this https URL.
[AI-57] Qwen -Audio-Jev: A Unified Interface for Audio Decisions
链接: https://arxiv.org/abs/2610.21614
作者: Ying Fang,Zhifu Gao,Qing Gu,Jiajun He,Kangqi Jing,Xiangang Li,Yu Pu,Wupeng Wang
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注:
Abstract:Large audio language models (LALMs) have made substantial progress in audio understanding. However, in high-throughput applications requiring frequent selection, judgment, and scoring, autoregressive text generation introduces additional computational overhead and response latency. To address this limitation, we introduce Qwen-Audio-Jev by augmenting Qwen-Audio-3.1-ASR-Flash-Next with a dedicated Jev decision head. The model unifies selection, binary judgment, and scoring through three decision primitives: CHOICE, NOUL, and SCORE. Given audio, a natural-language question, and an explicit answer specification, it directly produces a task-conditioned probability distribution or a numerical score through typed decision readouts. Evaluation across 18 labeled configurations covers all three primitives: for CHOICE, Qwen-Audio-Jev achieves accuracies of 78.70% on MMAU-mini and 84.06% on MMSU; for NOUL, it achieves an equal error rate of 6.50% on the VoxCeleb1-O speaker verification benchmark; and for SCORE, the model achieves a normalized mean absolute error of 0.1062 on NISQA overall quality assessment, with Pearson and Spearman correlations of 0.8210 and 0.8074, respectively. Compared with the original Qwen-Audio-3.1-ASR-Flash-Next, Qwen-Audio-Jev achieves an approximately 20x inference speedup. It demonstrates its potential for efficient, high-throughput audio decision-making in applications such as large-scale audio screening, speaker verification, and automated quality assessment.
[AI-58] Learning Intrinsic Dynamics in Latent World Models
链接: https://arxiv.org/abs/2610.21612
作者: Jiamin Shi,Yuanhong Zhang,Zhiyang Liu,Jiawei Sun,Jingwen Fu
类目: Artificial Intelligence (cs.AI)
备注: 12 pages, 5 figures
Abstract:Learning intrinsic dynamics—the relationship between actions and their consequences—is essential for world models. However, visual variations unrelated to the underlying control process can obscure this relationship in learned representations. We investigate the connection between action predictability, defined as the ability to reconstruct an action from consecutive latent states, and the preservation of intrinsic dynamics. Our central hypothesis is that encouraging action predictability discourages representations from discarding the distinctions between different actions and their consequences. We provide two complementary forms of evidence for this connection. Theoretically, we derive a positive lower bound on squared inverse prediction error when the encoder merges distinguishable outcomes of different actions from the same state and appearance, and show that forward prediction can still be exact on those transitions. Empirically, controlled appearance-diversity comparisons, supervision interventions, and latent probes support the role of action prediction in preserving control-relevant transition information. We instantiate action predictability with Concat IDM, which predicts actions from concatenated latent states and updates the shared encoder using action labels already available in offline trajectories, without changing the test-time planner. Across three independent training runs on Push-T with 24 appearances, Concat IDM increases planning success from 9.1% to 77.2%; its paired improvement is 52.9 percentage points larger than that observed with a single appearance. Together, these findings support action predictability as an inductive bias toward intrinsic dynamics and demonstrate the practical benefits of Concat IDM for latent planning under visual diversity.
[AI-59] Evaluating Inquiry Learning in LLM -Based Agents : Hypothesizing Reasoning and Exploration EMNLP2026
链接: https://arxiv.org/abs/2610.21604
作者: Marta Knežević,Seyed Parsa Neshaei,Jibril Frej,Vinitra Swamy,Tanja Käser
类目: Artificial Intelligence (cs.AI)
备注: EMNLP 2026 (Findings)
Abstract:Scientific inquiry involves iterative cycles of exploration, hypothesis generation, and reasoning over observations to uncover relationships in complex systems. Large language models (LLMs) have demonstrated strong capabilities in several components of this process, including inductive reasoning, hypothesis generation, and data analysis. However, prior work typically evaluates these abilities in isolation, leaving open how models integrate them across a full inquiry cycle. In this work, we introduce a framework for evaluating LLMs’ inquiry capabilities in a unified, interactive setting. We extend the Countdown task to a mathematical environment where an agent probes the system by manipulating variables, collecting observations, and inferring relationships between them. This setup enables analysis of exploration strategies, causal discovery, and hypothesis refinement during inquiry. Across seven LLMs and multiple prompting strategies, our experiments reveal limited inherent inquiry capabilities and low success rates in discovering the underlying expressions. However, structured guidance inspired by educational scaffolding significantly improves performance. Our findings highlight current limitations of LLMs in strategic exploration and iterative hypothesis refinement, and show how simple guidance strategies developed for human learners in inquiry learning can support more effective inquiry behavior.
[AI-60] Decoding What iOS Location Services Send to Apple: An LLM Is an Unreliable Analyst but a Useful Toolmaker
链接: https://arxiv.org/abs/2610.21597
作者: Douglas Leith
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI)
备注:
Abstract:Reverse engineering the data formats used by mobile software is skilled and slow, and using AI to assist with it is attractive. We evaluate this for the protobuf messages that an iPhone sends to Apple’s location services. Working from Apple’s binaries, we use two frontier models (Claude Opus 4.8, Codex GPT Sol 5.6) to reconstruct the schemas of protobufs observed in network traffic. The great majority of the schemas contain substantial errors, and these are hard to detect because the schemas are internally consistent. Direct AI use is therefore unsuitable for forensic and privacy analysis. AI can, however, write reliable analysis tools when their output can be validated. We use Claude to write a Python tool, Extractor, that reliably recovers protobuf schemas from Apple binaries. We validate it against binaries compiled from schemas with known answers, previously published schemas, and network measurements. We demonstrate the use of Extractor to decode and analyse the location services network traffic of a jailbroken iPhone 8 (iOS 16.7.11).
[AI-61] LoopAhead: Speculating Ahead for Faster Decoding in Looped Language Models
链接: https://arxiv.org/abs/2610.21594
作者: Zixuan Liu,Nan Xue,Rui Bao,Yuhang Xu,Xiaofeng Cao,Jiangchao Yao,Zhiyong Chen
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Looped Transformers improve parameter efficiency by repeatedly applying the same Transformer layers. However, generating each token still involves sequential computation and repeated access to model weights, leading to high decoding latency. Recent pipelined self-speculative decoding methods reduce this delay by drafting tokens from early-loop predictions and asynchronously advancing tokens at different recurrent depths in the same batch, but they do not fully exploit early-loop features for longer lookahead and still spend computation verifying incorrect speculative branches. We introduce \method, a speculative decoding framework for looped transformers that enables earlier drafting of future tokens and assigns complementary roles to different recurrent depths. A lightweight lookahead drafter (LAD) predicts multiple future tokens with early-loop states, and deeper-loop feedback helps decide which draft branches to continue and which to pause, while the full-depth model makes the final token decisions. LoopAhead achieves up to 7.82x decoding speedup over autoregressive decoding while preserving the target model’s output distribution.
[AI-62] LOADBench: A Large Open Analog Dataset and Benchmark for Machine Learning in IC Design
链接: https://arxiv.org/abs/2610.21591
作者: M. Leibl,F. Azevedo,S. Uhlich,A. Bonetti,A. Venkitaraman,C. Hsieh,P. Pham,L. Servadei,H. Graeb,R. Martins
类目: Artificial Intelligence (cs.AI)
备注: 9 pages, 6 figures, Conference: MLCAD '26: 8th ACM/IEEE International Symposium on Machine Learning for CAD, September 7 to 9, 2026, Jeju, South Korea
Abstract:Machine learning (ML) for analog IC design is constrained by limited open data and the absence of standardized evaluation: results reported in different papers are typically obtained on bespoke, often closed, datasets and with ad-hoc metrics, leaving comparisons across methods unreliable. We address both gaps with a large opensource dataset and a suite of standardized benchmarks. The dataset contains more than 100 operational amplifier (opamp) topologies (1-, 2-, and 3-stage) in the SkyWater130 PDK, each shipped with structural and functional annotations, topology-specific sizing constraints (symmetry and sizing rules), an Ngspice testbench, and a multi-objective NSGA-II sizing run producing around 103 Pareto-optimal and 106 intermediate sizing-performance pairs per topology. On top of this data we define six benchmark tasks with precise protocols: inverse sizing, topology selection, sub-circuit classification, structural circuit generation, Pareto-membership prediction, and multi-objective optimization. Thus, the data presented here constitutes the most comprehensive dataset for ML development in over four decades of analog IC design automation. The dataset, ground-truth annotations, and reference implementations of the metric-defining routines are released under an open license, enabling reproducible and directly comparable evaluation across heterogeneous analog design automation methods.
[AI-63] AVALON: Zero-Shot Indoor Navigation with Human-Seeded Topological Memory
链接: https://arxiv.org/abs/2610.21586
作者: Jing Xia,Changyan Li,Yuduo Wang,Guiyin Tian,Bo Wang,Yifan Liao,Mingjun Li,Shiyan Hu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Indoor service robots often enter unmapped buildings whose layouts and task constraints are already familiar to users. We address this human-familiar, robot-unseen setting with AVALON, a topological navigation framework organized around human-seeded layered spatial memory, node-verified topological execution, and task-triggered, evidence-gated human–robot memory co-updating. The robot has no prior on-site traversal or site-specific training. Deployment requires only a floor plan, place clues, and an initial topology-odometry alignment. Visual depth point-goal navigation leverages verified node arrivals to perform translational re-anchoring, without requiring a robot-constructed global SLAM map. Human–robot memory co-updating fuses complementary sources of information. Human operators specify task constraints that are unobservable via vision and report offline environmental modifications, whereas the robot provides physical-state evidence through active in-situ validation. Task relevance governs whether human-reported environmental changes undergo robot validation during the active mission. Field-collected perceptual evidence gates memory updates, maintaining an explicit separation between human-specified permissions and robot-observed traversability. Our experiments on a Unitree Go2 show that AVALON with X-NavDP completes a 12.81 m Intersection-to-Dining mission in 46.4 s (5/5) with no mapping, 40.9% less robot time than DimOS-Go2 (78.5 s, 4/5), while policy-only X-NavDP times out in all of the 5 runs. These demonstrate the superiority of the proposed topological navigation framework.
[AI-64] ClaimScope: Label-Blind Witness Verification for Mathematical Claim Auditing NEURIPS2026
链接: https://arxiv.org/abs/2610.21581
作者: Mushir Akhtar
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026 Workshop (Mathematical Reasoning and AI)
Abstract:Mathematical agents are commonly evaluated by answer accuracy or proof acceptance. We introduce ClaimScope, a 72-task benchmark for auditing proposed claims through a verdict, scope, counterexample, and repair. Tasks cover optimizer-state transport, finite-horizon affine recurrences, and exact 2\times2 linear algebra. We evaluate Qwen3-8B with direct generation and four matched two-pass conditions, each starting from the same draft. A witness checker evaluates only the model’s submitted state, without consulting reference verdicts or supplying a replacement witness. Its feedback still conveys information about validity: a verified refutation establishes that the claim is false, and rejection reasons can suggest structural fixes. In this evaluation, verifier feedback yields 23/39 accepted counterexamples on invalid claims and 25/72 safe completions, compared with 15/39 and 20/72 for checker-free generic feedback. Strict completion remains 17/72 ; verdict accuracy changes from 47/72 to 46/72 . The claim-grouped safe-completion difference is .094 , with a bootstrap 95% interval of [.000,.198] . All seven task-level gains follow horizon/input-length diagnostics. A post-hoc schema-only correction of the direct outputs reaches the same 25 safe tasks without another model call. These results separate claim classification from executable evidence, while locating the observed gains in witness-structure correction rather than demonstrating broader reasoning improvement.
[AI-65] Learning to Search for Executable Patternmaking from Certified Programs
链接: https://arxiv.org/abs/2610.21579
作者: Zhuotao Xu,Chi Zeng,Liying Yang,Xinxing Yu,Yanyan Liang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Graphics (cs.GR)
备注: https://arxiv.org/submit/8234086/start
Abstract:Recent models generate sewing patterns from images or text, and program-based representations make these patterns executable. These systems produce a finished pattern and do not model how a patternmaker turns one pattern into another. In practice a new style is made by editing an existing pattern: lengthening the bodice, moving a dart, swapping a sleeve. Pattern archives record the results of such edits and none of the steps. We observe that the steps can be recovered by searching over editing actions until one pattern becomes the other. A deterministic compiler makes the recovered steps checkable. To this end, we present SemDraft, the first framework that learns to search for executable patternmaking programs from paired before-and-after sewing patterns without recorded editing steps. Specifically, a bounded search recovers a program of editing actions and certifies it only after three checks: it recompiles from scratch, every pair of its actions is classified as ordered or interchangeable, and an independent replay reaches the same pattern. The certified programs then train a model that ranks candidate actions inside the same search. On two frozen draws of 80 order-sensitive cases, with every other search component fixed, the learned ranker certifies 63 and 64 cases within 1,000 search executions, against at most 53 and 52 for non-learned rankers.
[AI-66] Independent Data-Integrity Verification for LLM -Driven Functional Coverag e Closure in Hardware Design
链接: https://arxiv.org/abs/2610.21570
作者: Venkata Krishna Kumar Vedantam
类目: Hardware Architecture (cs.AR); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:LLM agents can write directed tests that close functional coverage, but coverage only shows that a state was reached, not that the data on the way there was correct. This is a long-known point in verification and we do not claim it as new. We describe a small coverage-closure agent that applies it to an LLM loop: coverage is read from the report that the simulation itself exports (cocotb-coverage YAML) and never taken from the model’s account, and data integrity is checked by scoreboards that are separate from the model. On three architecturally distinct RTL designs (an asynchronous FIFO, a UART transmitter and an AHB2APB bridge) the agent raised coverage from 91.7%, 88.9% and 67.4% to 100% within its 8-iteration budget in every logged run, using one design-agnostic agent. A reactive scoreboard monitor attached to the FIFO’s base testbench also found a clock-domain-crossing bug in the author’s own stimulus code. The monitor is not yet attached to the LLM-generated tests, so their data integrity is not independently checked; one manual check on a single FIFO test is reported as preliminary. We list the weaknesses of this evidence: three designs, no baseline, no ablation, and logged runs that come from several agent checkpoints.
[AI-67] BiShield-TEE: On the (In-)Security of Unilateral Weight Obfuscation in On-Device TEE-Shielded LLM Partition NEURIPS2026
链接: https://arxiv.org/abs/2610.21565
作者: Zhiyuan Wang,Yuezhi Zhou,Ziqi Zhang,Zhu Zhang,Zhixing Tan,Yongheng Deng
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: Accepted at NeurIPS 2026
Abstract:On-device deployment of large language models (LLMs) is increasingly adopted due to latency and privacy requirements, but exposes proprietary models to reverse engineering and intellectual property theft. Trusted Execution Environments (TEEs) provide hardware-enforced isolation for on-device inference, yet most TEE-based parameter obfuscation relies on unilateral matrix transformations that obfuscate only output side. We identify a vulnerability in this design, termed the \emphSolution Determinacy Property (SDP): the obfuscation secret is uniquely determined, exposing the structure of the true key and enabling attackers to break the protection. Under our attacks, representative methods \textscGroupCover and \textscArrowCloak are reduced to negligible protection, with attack accuracy reaching over 2.4\times the full-shielding black-box baseline and recovering near-complete model this http URL address this vulnerability, we propose \textscBiShield-TEE, a bilateral framework that induces \emphAmbiguous Solutions through joint obfuscation of input and output dimensions, hiding the true key among many equally valid candidates. Experiments across diverse models and datasets show that \textscBiShield-TEE confines attack accuracy to the full-shielding black-box baseline, while achieving higher inference efficiency than prior methods on real this http URL code is available at this https URL.
[AI-68] When Does Muon Actually Win? A Paired Data- Time- and Tuning-Budget Benchmark NEURIPS2026
链接: https://arxiv.org/abs/2610.21555
作者: Mushir Akhtar,Hazem N. Nounou,Majdi Mansouri,Mohamed N. Nounou
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at NeurIPS 2026 Workshop (Optimization for Machine Learning)
Abstract:Muon’s gains in data efficiency motivate evaluating its performance under explicit data, time, and tuning budgets. On a single A100 GPU, we compare a reference single-device Muon implementation with AdamW across 17.7M- and 33.6M-parameter causal language models and 2.70M-parameter vision transformers. A common 23-candidate development budget selects configurations evaluated with five matched seeds, full validation trajectories, optimizer-step timing, and a frozen one-shot test. Muon reduces validation-loss area under the curve by 3.25–22.44% at equal data and lowers test loss in all 20 primary pairs. In this configuration, Muon improves time-normalized loss area by 19.68–22.31% on vision, while AdamW is better by 3.38–6.96% on language. The full Muon optimizer call, including auxiliary AdamW updates, consumes 56.1–69.6% of training time, compared with 6.1–9.1% for the AdamW baseline. Muon also has higher expected regret under restricted access to these candidate pools. These results characterize small models with reference kernels; their wall-clock ordering does not establish a ranking for larger models or optimized implementations.
[AI-69] Uncertainty Quantification in Pairwise Difference Learning for Classification
链接: https://arxiv.org/abs/2610.21554
作者: Mohamed Karim Belaid,Maximilian Rabus,Eyke Hüllermeier
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 32 pages. Preprint (version submitted before peer review) of the article published in Machine Learning 115, 12 (2026)
Abstract:Instead of learning a mapping from instances to outcomes in the standard way, the key idea of pairwise difference learning (PDL) is to learn a function that takes two instances as input and predicts the difference between the respective outcomes. Given a function of this kind, predictions for a query instance are derived from every training example and then averaged. We propose a version of PDL for the task of classification and propose a suitable meta-learning technique for this purpose. Moreover, we address the problem of uncertainty quantification, i.e., of representing and quantifying the uncertainty of a PDL classifier’s predictions in a faithful manner. Building on recent work on the decomposition of total uncertainty into aleatoric and epistemic components, we propose different ways of extending corresponding measures to the setting of PDL, depending on how and to what extent uncertainty is represented by the underlying base learner. In large-scale empirical studies, we analyze both the predictive accuracy and the uncertainty awareness of our PDL classifier.
[AI-70] Reinforcement Learning-driven Diffusion Models: Language Vision and Beyond
链接: https://arxiv.org/abs/2610.21539
作者: Zhiyuan Ma,Jiabao Wei,Tianyu Wang,Hanqing Wang,Jiadong Pan,Haowen Xu,Liangliang Zhao,Ruixun Liu,Yuhan Wang,Yuwei Zhang,Liang Feng,Junhao Xiao,Kaiyan Zhang,Xuekai Zhu,Daizong Liu,Biqing Qi,Yihao Liu,Ganqu Cui,Ning Ding,Linfeng Zhang,Jianjun Li,Kun He,Xiang Bai,Bowen Zhou
类目: Artificial Intelligence (cs.AI)
备注: 26 pages,12 figures,8 table,accepted by TPAMI
Abstract:Diffusion models (DMs) have steadily demonstrated outstanding performance in high-fidelity, controllable generation across diverse scenarios, following the “pretraining-finetuning” paradigm established in LLMs. Despite these advances, conventional DMs still face critical limitations in sophisticated reasoning, autonomous exploration of generation pathways, and adaptability to complex, human-aligned objectives. Inspired by RLHF’s success in LLMs, RL has recently emerged as a pivotal post-training mechanism to iteratively refine DMs through reward-driven trajectory optimization. This paper presents the first comprehensive review of RL-driven DMs and an empirical analysis of how RL reshapes the diffusion process towards experience-centric, goal-directed generation. We propose a novel technical panorama elucidating RL’s multifaceted roles in diffusion-based language, multimodal, robotic, and scientific models, with a granular focus on diffusion trajectory optimization and reward engineering. Through extensive experiments on multiple task-specific benchmarks, we systematically evaluate representative methods to intuitively highlight the advantages of different RL post-training strategies. Furthermore, we reveal and dissect three crucial bottlenecks that currently hinder the real-world deployment of RL-driven DMs: reward design fragility, exploration-exploitation trade-off in high-dimensional generation spaces, and training instability under sparse rewards. Based on these perspectives, we provide a principled roadmap for the integration of RL into DMs and outline promising future research directions for robust, interpretable RL-diffusion systems. We also construct a public repository to continuously track the latest developments in this rapidly evolving field: this https URL.
[AI-71] Functional Subnetwork Localization for Efficient Continual Learning in Large Language Models
链接: https://arxiv.org/abs/2610.21537
作者: Shuoran Jiang,Wentao Zhu,Yang Xiang,Dongfeng Li,Yunan Zhang,Mengchen Zhao,Yuhui Guo,Kai Chen,Youcheng Pan,Xiangping Wu,Qingcai Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Biological intelligence continually adapts to new information while retaining prior skills through localized plasticity. In contrast, continual learning in large language models (LLMs) remains vulnerable to catastrophic forgetting. A promising direction is to identify functionally specialized subnetworks within LLMs that enable targeted and resource-efficient adaptation. Using gradient-attribution profiling, we localize cognition-associated functional subnetworks exhibiting extremely low mutual overlap. Building on this finding, we propose Function-isolated Memory-Efficient Optimization (CiMEO), a continual learning framework that routes each training example to its corresponding functional subnetwork, restricting parameter updates and optimizer states to selected regions. Across 1.4 million sequential instructions, CiMEO achieves the highest average performance among the compared methods on five generative benchmarks spanning programming, examination-style tasks, and mathematical reasoning, while mitigating catastrophic forgetting. Subspace-overlap and principal-angle analyses further indicate that localized updates help preserve pre-trained representations, highlighting functional subnetwork localization as a practical approach to efficient and stable continual learning in LLMs.
[AI-72] CriticaLLM : Using LLM s as Autonomous Game Master for Tabletop Role-Playing Games
链接: https://arxiv.org/abs/2610.21525
作者: Gaëtan Berlaimont,Justin Vanwichelen,Benoît Ronval,Hélène Verhaeghe
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models (LLMs) have shown impressive capabilities across many tasks, including processing large amounts of text, following instructions, and generating ideas. This allows the use of such models in various contexts, including games and Tabletop Role-Playing Games (TTRPGs), such as Dungeons Dragons. However, these games include many complex components, from crafting interesting stories to adhering to rules, while allowing some freedom in players’ actions. In this work, we first highlight key components that LLMs should exhibit in the context of TTRPGs. Then, we introduce our framework, CriticaLLM, a fully automated Game Master that can run entire TTRPG games, from creating diverse and interesting stories to managing exploration and combat phases. We evaluate our framework with actual TTRPG players and examine the diversity of generated plots and antagonists across four LLMs of varying sizes. Overall, our framework appears to be a promising option for running entire TTRPG games.
[AI-73] Poison Valleys: Diagnosing Path-Dependent Failure in LLM Code Repair NEURIPS2026
链接: https://arxiv.org/abs/2610.21514
作者: Veljko Skarich
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: Accepted to Interpreting Agent Behavior (IAB): The First Workshop on Human-Centered Interpretation for Understanding Agents, Humans, and Interaction (NeurIPS 2026 workshop); 22 pages; code, data, and demo: this https URL
Abstract:When code-repair agents fail, final tests reveal that a patch is wrong but not which earlier commitment caused it or which software obligation failed. Poison Valleys traces this path in 1,844 model requests on two synthetic Python cases: PersistQueue adds atomic batch claiming to a disk-backed queue, while Dijkstra adds source-sensitive filtering to shortest-path search. We define plausible but incorrect software mechanisms, place each in a model’s hypothesis, implementation plan, or both, and evaluate the generated code with an Oracle and targeted diagnostic coordinates. Moving the misconception from hypothesis to plan increased failure by 32.2% in PersistQueue and 28.9% in Dijkstra after matching the experimental mix. Different misconceptions produced distinct, model-dependent patterns of violated obligations, although several could converge on the same behavior. This creates a measured connection from upstream misconception to code mechanism and failure form. Code, data, and demo: this https URL.
[AI-74] QREF: Improving DeepSearch Agents by Aligning Search with Reasoning via Query Refinement and Evidence Filtering
链接: https://arxiv.org/abs/2610.21512
作者: Yujia Chen,Yuhui Wang,Junzhe Wang,Shihan Dou,Di Liang,Yujiong Shen,Yuhao Zhou,Zhiheng Xi,Ming Zhang,Senjie Jin,Jiayi Chen,Dingwei Zhu,Yajie Yang,Yuan Hua,Chenyao Lu,Tao Gui,Qi Zhang,Xuanjing Huang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:DeepSearch agents achieve strong performance on complex information-seeking tasks through iterative search and reasoning. However, they often issue intent-misaligned, under-constrained, or redundant search queries, resulting in poor evidence coverage and accumulating retrieval noise. To mitigate these limitations, we propose QREF, a plug-and-play Query Refiner and Evidence Filter. Specifically, the Query Refiner rewrites search queries according to the current reasoning intent while ensuring complementary coverage across each search batch. The Evidence Filter retains evidence relevant to both the current search batch and the reasoning objective, reducing distracting retrieval results. We initialize the Query Refiner with supervised fine-tuning and optimize it through reinforcement learning based on interactions with a real-world web search environment, reducing ineffective queries. Comprehensive experiments demonstrate that our approach improves the performance of DeepSearch agents across four challenging benchmarks, achieving average relative Pass@1 improvements of 33.55% on WebSailor and 11.04% on TongyiDeepResearch without modifying architectures or retraining their reasoning policies. Compared with the original DeepSearch backbone without QREF, the RL-trained refiner reduces the fraction of query batches that lack discriminative retrieval anchors by 95%. Compared with the SFT-trained refiner, it also lowers the filter-based ineffective query ratio.
[AI-75] ISW-Agent : Integrating LLM -Driven Dialogue with n8n Workflow for Internal Solitary Wave Remote Sensing Image Processing
链接: https://arxiv.org/abs/2610.21502
作者: Hao Zhang(1),Hengrui Liu(1),Lina Sun(1),Junmin Meng(1) ((1) First Institute of Oceanography, Ministry of Natural Resources, Qingdao, China)
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Internal solitary wave (ISW) remote sensing requires task parameters and spatial information to be transferred across image acquisition, detection, stripe extraction, and propagation prediction. Here, we present ISW-Agent, which integrates large language model (LLM) interpretation with n8n workflows and independent scientific models through structured task constraints, geospatial transformations, and stage-specific output validation. In tests of 120 Chinese instructions per backend, GPT-5.6 Sol and DeepSeek both achieved a strict task-match rate of 94.2%, with all intents and execution decisions matching the reference. Sentinel-1 cases demonstrated continuous processing from dialogue requests to wavefront prediction for the specified scene. Of 50 timed scenes, 39 completed detection, extraction, and prediction with a median turnaround time of 48.04 s; 11 stopped after empty extraction. Across 3,408 controlled fault-injection runs, retries recovered 234 transient-failure conditions and output validation rejected 234 structural defects, while 61 provenance or completeness errors indicated the need for content-consistency checks. Overall, ISW-Agent thus connects natural-language requests to traceable spatial processing chains for multimodel ISW analysis.
[AI-76] MetaExp: Learning to Self-Improve through Experience via Verifiable Cross-Task Environment Scaling
链接: https://arxiv.org/abs/2610.21492
作者: Renhao Li,Xinjie Shen,Yang Su,Jianhong Tu,Derek F. Wong,Dayiheng Liu,Min Yang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Language-model agents can accumulate interaction feedback, but reliably turning that feedback into reusable experience remains challenging. Existing methods either optimize external experience scaffolds or train experience-related behavior on specific task distributions, leaving open whether experience learning itself can become a transferable model capability. Training such a capability calls for supervision beyond isolated tasks, which mainly support in-episode problem solving rather than revealing what knowledge remains useful across episodes. In this work, we introduce MetaExp, a scalable framework that trains this capability through verifiable task families with controlled cross-task structure. Each family preserves a reusable factor while varying task-specific realizations, enabling transfer without direct solution reuse. We propose two complementary designs: Shared-Tool families keep black-box tools fixed across tasks to elicit operational experience, while Shared-Workflow families preserve latent procedures across changing realizations to elicit procedural experience. We scale environments under both designs, collect and filter sequential interaction trajectories, and fine-tune Qwen3.5-122B-A10B on 4,956 trajectories from 1,000 task families. With parameters frozen during downstream evolution, MetaExp improves by +8.5 points on SpreadsheetBench, peaks at +2.3 on APEX-Agents, and improves by +14.8 on AutomationBench. The gains persist under domain-excluded training, supporting experience learning as a transferable capability for continued non-parametric improvement after deployment.
[AI-77] FlashAttention with Statistics: Low-Overhead Attention Features for Compact Hallucination Detectors
链接: https://arxiv.org/abs/2610.21483
作者: Andrei Volodichev,Maksim Savichev,Nikita Sviridov,Artem Yuliy,Dmitry Simakov,Alexey Zaytsev
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models can produce fluent answers that the supplied context does not support. A compact frozen proxy model can detect such answers by reading the context and another model’s response, even from a proprietary generator. Attention-based detectors use summary statistics of the proxy’s attention weights, but standard pipelines materialize the attention matrix, whose memory grows quadratically with sequence length. We compute these statistics without materializing attention, by accumulation in the FlashAttention pass, sparse lookups after it, or a second tiled pass over the scores, and classify them by this path into nearly free and costlier ones. On an H100, the statistics of five published detectors add 16% to the latency of a Qwen3-8B forward at 2K tokens, falling to 1.4% at 35K, and store 22.5 KiB per response token. On three benchmarks, probes trained on frozen Qwen3 proxies of 0.6B to 8B parameters prove strong detectors: a 4B proxy outperforms the strongest of nine zero-shot 4-9B LLM judges in F1 on two of the three, the proxies match or outperform specialized grounding checkers of up to 7B parameters, probing each proxy beats prompting it as a judge, and adding attention features significantly improves F1 in half of the 12 benchmark-proxy pairs, never significantly lowering it.
[AI-78] ool-Augmented Large Language Models for Reliable Medical Calculation from Patient Notes
链接: https://arxiv.org/abs/2610.21477
作者: Yang Zhong,Ling Luo,Xuanlyu Zhu,Yanhua Wang,Yuanyuan Sun,Jian Wang,Hongfei Lin
类目: Artificial Intelligence (cs.AI)
备注: Accepted as a regular paper at BIBM2026
Abstract:Medical calculators translate raw patient observations into clinically actionable values and scores. However, applying these calculators directly to free-form patient notes requires precise variable extraction combined with the exact execution of complex formulas, scoring rules, and output conventions. Relying solely on direct large language model (LLM) prompting is unreliable in this context, as rule recall, parameter extraction, arithmetic reasoning, and formatting become entangled within a single generative step. To address these limitations, we propose a tool-augmented LLM framework for reliable medical calculation from patient notes. The framework retrieves and structures calculator specifications, generates reusable executable tools, and uses a constrained extractor to map each patient note to calculator-ready inputs. The tools then perform deterministic calculation, separating clinical language understanding from rule execution and arithmetic. Evaluated on the MedCalc-Bench dataset, our method achieves 86.73% final-answer accuracy, outperforming the one-shot prompting and supervised fine tuning-only baselines by 24.0 and 17.27 percentage points. Ablation and error analyses show that these performance gains are primarily driven by rule-grounded tool synthesis and schema-constrained extraction, whereas remaining failures are concentrated in clinical-semantic extraction and benchmark-specific output contracts. By grounding calculator logic in executable tools, the proposed approach alleviates arithmetic hallucinations and improves the computational reliability of the final clinical outputs.
[AI-79] Ontology Refinement with Embeddings and LLM s: A Case Study on CASTON
链接: https://arxiv.org/abs/2610.21466
作者: Hui Yang,Ziyu Li,Xuan Liu,Anders Jarfors,He Tan,Jiaoyan Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Ontologies are widely used to represent and organize complex knowledge in domains such as biomedicine and healthcare. However, constructing reliable ontologies remains challenging, as class hierarchies, whether manually or automatically constructed, may contain subtle subsumption errors that require careful refinement. We study this problem in CASTON, a casting-domain ontology, and present a lightweight but scalable refinement workflow that combines a fine-tuned ontology embedding model with Large Language Models (LLMs). The embedding model is used to extract a limited number of candidate axioms for refinement together with possible replacement candidates, and an LLM then returns repair suggestions. In a small expert evaluation, the default configuration using a CASTON-fine-tuned OnT embedding model and Qwen3-8B produced reliable high-confidence repairs: all six high-confidence replacements were accepted by both domain experts. The implementation is available at this https URL.
[AI-80] Incentivizing Tool-Use Reinforcement Learning for Post-Arrival Emergency Care Prediction
链接: https://arxiv.org/abs/2610.21449
作者: Pengfei Hu,Chenwei Wu,Qiao Jin,Jun Deng,Zhiyong Lu,Liyue Shen,Yue Ning
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) can reason over both structured measurements and free-text records, which makes them attractive for emergency department (ED) prediction. However, most existing LLMs still fall short of conventional ML methods like trees on EHR-based prediction tasks. Our controlled analysis traces this gap to two weaknesses: LLMs do not select informative features, and they misjudge what numerical values imply for risk even when given the right features. We propose TIR-EC, a tool-integrated reinforcement learning framework that addresses both. TIR-EC exposes EHR feature groups as tools the LLM can call. It trains the LLM with rewards from a gradient-boosted tree teacher, which provides a reference set of informative tools and a graded risk target. TIR-EC also distills task-specific clinical skills into the model weights through a reverse curriculum that gradually removes the skills from the prompt during RL process. Evaluation on three clinical prediction tasks shows that our 4B agent achieves the highest macro F1 of all compared models, outperforms conventional ML baselines, and is competitive with much larger frontier LLMs. Further analysis shows that TIR-EC selects more useful evidence with fewer tool calls and combines measurements more effectively.
[AI-81] V-SHARE: Provable Convergence Guarantees for Advantage-based Algorithms
链接: https://arxiv.org/abs/2610.21448
作者: Kintan Saha,Hamed Masoudi,Xinyi Sheng,Dominik Baumann,Joni Pajarinen,Arsenii Mustafin
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:In reinforcement learning, advantage-based value-learning methods, unlike standard Q-learning, share information across actions by decomposing the action-value function into a state-dependent value component and action-specific advantages. While this mechanism has proven effective in both tabular and deep reinforcement learning, its theoretical understanding remains incomplete. We introduce V-SHARE, an advantage-based value-learning algorithm that combines cross-action value sharing with an importance-weighted Bellman-residual correction. To our knowledge, we provide the first complete theoretical characterization of an advantage-based algorithm, including both asymptotic convergence and non-asymptotic finite-sample guarantees. We identify a regime in which V-SHARE mitigates the rare-action bottleneck of asynchronous Q-learning and uncover a contraction–variance tradeoff between stronger deterministic contraction and increased importance-weighting noise. Motivated by this tradeoff, we develop decreasing residual-preconditioning schedules that combine faster early contraction with lower asymptotic noise. We further develop a two-timescale analysis that removes the single-timescale residual-weight threshold while preserving convergence guarantees. Experiments on discrete Markov decision processes (MDPs) and deep reinforcement-learning benchmarks corroborate our results.
[AI-82] Learning from Interaction Experience via Localized Context Distillation
链接: https://arxiv.org/abs/2610.21427
作者: Zhezheng Hao,Yongkang Yang,Xilin Xia,Jiarui Yu,Yi Liu,Fanjunduo Wei,Peijie Dong,Tianfu Wang,Qiang Lin,Can Wang,Hande Dong,Jiawei Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Learning from interaction experience is essential for enabling large language models to continually improve during deployment. Context distillation offers a promising path: privileged information extracted from experience is provided to a teacher, whose behavior is then distilled into a student model that has no access to it. However, both existing forms of context distillation have limitations. Off-policy context distillation trains the student on teacher-generated trajectories and thus suffers from exposure bias; whereas on-policy context distillation trains on the student’s own rollouts, whose states are often uninformative and fail to convey behaviors the student cannot yet produce. To resolve this dilemma, we introduce Localized Context Distillation (LoCD). LoCD first localizes the critical spans of a student trajectory that require revision. The teacher then regenerates only these spans from the student’s own prefix, and the student is trained on the regenerated segments. LoCD thereby concentrates learning on the most informative states while keeping exposure bias low. Extensive experiments across math reasoning, code generation, and agent tasks demonstrate that LoCD consistently outperforms strong baselines and achieves more stable training with smaller and sparser parameter updates.
[AI-83] JevAgent Harness: Benchmarking Deploying and Training Decision Models That Share Control with LLM s
链接: https://arxiv.org/abs/2610.21412
作者: Kunlun Zhu,Tao Feng,Xu Han,Heng Wang,Ling Yang,Xiangru Tang,Jiaxuan You,Heng Ji
类目: Artificial Intelligence (cs.AI)
备注: Code: this https URL
Abstract:Large language model (LLM) agents now fix software, operate computers and serve customers, but their harnesses, the programs that run them, call a large LLM at every step, which makes long tasks slow and costly. Most of these calls end in a multiple-choice decision, such as which tool to call next. Decision models such as Jev answer these questions directly, at a fraction of an LLM call’s time and price. We ask how a harness can delegate these decisions to a decision model with little loss of success. To evaluate decision models in agentic tasks, we build AgentDecisionBench, which tests them from isolated questions up to control of a running agent, on public benchmarks and our DecisionGym environment. The decision model most accurate on isolated questions solved only 69.5% of agent tasks, and without a fitted confidence threshold, delegating tool choices lowered success by 19.0 points. Building on these findings, we then design JevAgentHarness, which fits this threshold for each choice on the agent’s own trajectories and lets the LLM decide below it. Because existing decision models never see the states an agent reaches after its own mistakes, we train AgentDecision-4B, an open 4B model, on such states with Jev’s answers as labels. On tasks unlike its training data, the agent solved 91.5% of them with AgentDecision-4B and 82.0% with Jev. On tasks like their training data, 4B models trained this way made agents 81.1-89.0% cheaper and up to 2.5 times faster, without losing 3 or more points of task success. The models, data, and harness will be released at this https URL.
[AI-84] Jev-LiteMAS: Separating Decisions from Generation in Multi-Agent Workflows
链接: https://arxiv.org/abs/2610.21411
作者: Kaixiang Wang,Zhaijiacheng Zhou,Jiong Lou,Jie Li
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-agent LLM systems rely on generative models for expert execution, workflow planning, and artifact review. Yet many coordination operations require only bounded decisions, such as selecting the next expert or judging artifact quality. Profiling three representative MAS frameworks, we find that routing and review account for 31.7-43.5% of generative-LLM tokens. However, replacing these operations with low-cost decision models introduces two interface gaps: generative planners produce execution plans and handoff instructions (planning gap), while reviewers provide actionable feedback for repairing flawed artifacts (review gap), which bounded decisions alone cannot provide. We introduce Jev-LiteMAS, a multi-agent framework that bridges both gaps by combining a lightweight decision model (Jev) with generative execution and feedback. To address the planning gap, Jev selects experts dynamically from the task and prior expert outputs, using accumulated context to ground subsequent execution without a pre-generated trajectory. To address the review gap, Jev screens intermediate artifacts against task criteria: passing artifacts proceed directly, whereas flagged ones undergo LLM re-review for acceptance or corrective feedback. Across four benchmarks and two backbones, Jev-LiteMAS reduces generative-LLM tokens by 12.7-54.6% over LAS, maintaining accuracy within 1.34 percentage points and lowering latency in seven of eight settings. Grounding routing in substantive expert outputs prevents up to 18.0 points in accuracy loss.
[AI-85] MemCut: Every Secret Leaves a Trail in the Long-Term Memory of LLM Agents
链接: https://arxiv.org/abs/2610.21405
作者: Cheng Zhang,Xuanning Cao,Zhicheng Zheng,Hui Xiong
类目: Artificial Intelligence (cs.AI)
备注: 17 pages, 4 figures, 13 tables, including appendices
Abstract:Long-term memory has become a crucial component of large language model (LLM) agents, with most research focusing on retrieving relevant information accurately and efficiently. However, a dependable memory system must also forget sensitive user data upon request. To evaluate this capability, we introduce MemCut, a systematic benchmark for user-requested forgetting. Its 456 hand-designed secrets in long-horizon conversations have fully annotated evidence trails to test recovery after deletion, including from messages that never state the secrets directly. We propose two recovery metrics. Native recovery measures reader success through the system’s retrieval, whereas residual recovery measures the same reader’s success when directly given surviving objects traced to annotated disclosures or evidence. We establish three reference deletion baselines to test how far existing retrievers and write models support forgetting across five memory designs and four LLMs. Disclosure removal reduces native recovery far more than residual recovery. LLM-guided deletion of ten additional turns from twenty retrieval candidates leaves 39-76% residual recovery. Subset sweeps (144 facts, three systems, two models) show that widening the pool at a fixed budget reduces residual recovery more than deleting all original candidates. These sweeps identify candidate coverage as a bottleneck in the reference baselines.
[AI-86] Multi-Reward Hacking: From Aggregation-Induced Drift to Trust-Anchored Defense
链接: https://arxiv.org/abs/2610.21404
作者: Yushu Li,Jiin Woo,Xiaoxiao Li,Youngsuk Park
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:To improve task performance while encouraging desirable response properties, LLM post-training often combines a primary reward (e.g., correctness) with auxiliary rewards (e.g., brevity or efficiency). However, how these rewards are aggregated can critically affect training outcomes. We show that symmetric aggregation, which treats reward channels equally, can induce a failure where aggregate reward improves while primary-task performance degrades, even when individual rewards are meaningful. We investigate this aggregation-induced failure through a one-step multi-reward decision formulation, characterizing when it constitutes multi-reward hacking. We prove that, under symmetric aggregation, deterministic conditional-mean GRPO converges to auxiliary-favoring behavior when auxiliary rewards are more reliably obtained, regardless of initialization. We propose Trust-Anchored Gating (TAG), which first elects an anchor channel based on the policy drift induced by optimizing each channel alone, then gates auxiliary credit on anchor satisfaction. Unlike calibration-sensitive static reweighting, TAG structurally eliminates credit for auxiliary-only outcomes while preserving additional credit for joint success. In the same analytical setting, TAG provably reverses this selection, converging to the primary-favoring action from any initialization when the primary channel is anchored. Experiments across Qwen and Llama models on mathematics and program synthesis reveal substantial degradation under symmetric, constrained, and normalized aggregation, whereas TAG mitigates this failure. On Qwen3-4B with a brevity reward, TAG achieves 0.581 pass@4 versus 0.220 for additive aggregation, while shortening responses by 44% relative to correctness-only training.
[AI-87] Watts If? A Benchmark for Counterfactual Reasoning About the AI Data Center Buildout
链接: https://arxiv.org/abs/2610.21397
作者: Pankaj Chouhan,Rajarshi Das,Aditya Dubey,Manminder Tomar,Sreekar Reddy,Samta Sharma,Satyapragnya Kar,PR Panigrahi
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:An AI data center is committed years before it draws a watt: a few discrete choices (grid or on-site gas, evaporative or dry cooling, which county to file in) fix its capital, water draw, and carbon path, and LLMs are now asked to help screen them with no test of whether they track what each commitment rules out. We introduce DCBench-440, with two arms reported separately: 200 generative questions over STRIPS worlds for five US deal archetypes, with search-verified gold answers and no tested shortcut above 12% on any question type, and 240 multiple-choice questions grounded in quoted site records, emission equations, and stated scenarios. Across open-weight deployments, accuracy spans 12.5-71.5% (generative) and 58.3-92.5% (multiple choice). One skill fails everywhere: asked to list every action available at a decision gate, no model exceeds 4.0% exact match. The strongest list 72-85% of the actions that would change the project’s state but omit most permitted actions that change nothing; with those excused, the best exact match is still 24%. Code, data, and other materials are available upon request.
[AI-88] Matoi: Selective Revalidation of Consequence Evidence for Agent Authorization
链接: https://arxiv.org/abs/2610.21394
作者: Zexun Wang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 55 pages, 4 figures. Includes an 18-page technical manuscript and a 33-page detailed supplement as ancillary files. Research preprint; not peer reviewed
Abstract:Matoi separates reusable consequence evidence from current execution authority and durable occurrence history. Its finite query-aware contract covers branches, ordered collections and negative lookups. A consumer reuses only its own checked evidence whose observations remain valid; signed requests, live grants, quota and uncertain outcomes remain separate obligations. Conditional correctness arguments distinguish locally serialized claims from nontransactional remote effects. Two studies strengthen cost attribution and domain evaluation. A 1,440-cell matched-wire study repeats 144 authored inputs. With compact input bytes identical, reuse reduces complete lifecycle cost by 3.49 percent in JS and 1.79 percent in Python. Combined representation/reuse reductions against full proof are 10.87 and 16.53 percent; five of twelve family intervals for the matched comparison cross one in each language. A prospectively frozen AgentDojo Banking mapping covers 5/16 user plans and 7/9 attack plans. Both governed strategies preserve five exact benign transfers and refuse 35 reused attack-plan/policy crosses. Pre-satisfied utility predicates are reported separately from completed work. These are original scripted synthetic tools, not autonomous agents or a real bank. Earlier studies retain cheaper protected-cache alternatives, policy-compatible phishing, provider duplication and post-claim revocation. Separate JS/Python consumers establish finite-profile interoperability within the same project; durable recovery rechecks evidence and retains uncertain reservations. The results support a bounded, auditable reuse and execution contract, not a new general incremental algorithm, arbitrary-tool closure, atomic third-party commitment or production superiority.
[AI-89] Self-Guided LLM Reasoning for Error Diagnosis in Vehicle Routing Problems
链接: https://arxiv.org/abs/2610.21393
作者: Ni Zhang,Jianan Zhou,Yue-Jiao Gong,Zhiguang Cao
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Vehicle Routing Problems (VRPs) are fundamental to many real-world applications. Recent advances in large language models (LLMs) have spurred progress in solving these NP-hard problems. However, most existing approaches assume that input instances are error-free, neglecting the presence of faults commonly encountered in real-world data. In this work, we systematically investigate the capability of LLMs to detect (and correct) errors in VRP instances, and identify key limitations, particularly in handling complex instances or errors that substantially increase the LLM’s reasoning burden. To overcome these challenges, we propose a self-guided reasoning framework in which LLMs autonomously generate code that provides structured insights to aid in error diagnosis. Experiments across diverse problem settings and LLMs demonstrate that our framework significantly improves error detection performance, achieving near-perfect accuracy.
[AI-90] Learning to Reason with Selective Internal Feedback
链接: https://arxiv.org/abs/2610.21392
作者: Xun Lu,Shumeng Yang,Xinrui Wang,Songjun Tu,Linjing Li,Daniel Dajun Zeng
类目: Artificial Intelligence (cs.AI)
备注: 21 pages, 10 figures
Abstract:Reinforcement learning from internal feedback can improve language models without ground-truth answers or external verifiers. However, confidence and correctness can diverge during training, and continued optimization of average self-certainty can increase confidence while degrading reasoning accuracy. We introduce SIF, a reinforcement learning method that combines comparative reward construction with selective policy updates. SIF compares ranked token-level self-certainty profiles across multiple windows and peers, assigning positive rewards only when both levels of voting provide sufficient support. Sample and group level filtering determines which responses contribute to policy optimization. Experiments on three backbones across 12 mathematical and scientific reasoning benchmarks show that SIF’s final checkpoints achieve higher average accuracy than the best-validation checkpoints of Intuitor and Vigor, with a gain of 4.71 percentage points over Intuitor on Qwen2.5-Math-1.5B. Further analyses show better retention of accuracy during continued optimization, while test-time training experiments demonstrate sustained improvements with as few as four rollouts per prompt.
[AI-91] PriorityFEC: Deadline-Aware Streaming FEC for KV-Cache Handoff in Edge Multi-Agent Systems
链接: https://arxiv.org/abs/2610.21390
作者: Han Shen,Qifu Tyler Sun,Fanyang Meng,Ye Wang,Yongsheng Liang
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-agent large language model (LLM) systems are increasingly deployed at the edge to support complex tasks that require collaboration and information exchange among multiple agents. However, when task contexts are passed across agents, shared context that has already been processed by upstream agents often needs to be prefilled again, resulting in substantial redundant computation. Reusing an already prefilled key–value (KV) cache can avoid this recomputation, but requires the reusable KV state to be transferred across the network before the receiving agent can begin generation. Under strict time-to-first-token (TTFT) deadlines and unreliable edge links, delivering the complete KV cache may be infeasible, making it necessary to prioritize and protect the most valuable KV information within a limited transmission budget. This paper proposes PriorityFEC, a utility-aware KV-cache handoff scheme that jointly exploits model-side KV importance and transport-side reliability. PriorityFEC first constructs a utility-aware source stream by placing high-utility KV blocks at earlier transmission positions. It then protects the ordered stream using a feedback-adaptive Streaming Forward Erasure Correction (FEC) mechanism, which schedules repair packets according to receiver feedback before the TTFT deadline. We implement PriorityFEC on a distributed edge LLM serving prototype and evaluate it with Qwen2.5-7B-Instruct and Llama-3.1-8B-Instruct across multiple datasets. Results show that PriorityFEC achieves superior task quality over the compared transport schemes, with its advantages becoming particularly pronounced under stringent deadline constraints. These results demonstrate the importance of jointly optimizing KV-cache utility and deadline-aware transport reliability when reusable LLM state must be handed off across unreliable edge networks.
[AI-92] Memorilla: Latent Semantic Memory for LLM s
链接: https://arxiv.org/abs/2610.21375
作者: Adibvafa Fallahpour,Parsa Idehpour,Vignesh Kothapalli,Nikita Mounier,Shirley Wu,Shayan Pardis,Jure Leskovec
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Website: this https URL
Abstract:Memory enables large language models (LLMs) to use information beyond their parameters, such as past user conversations, long documents, or an agent’s earlier actions. For a frozen LLM, access to this accumulated knowledge is limited to what a memory system places in its context window. However, gathering and combining relevant information needed for each query while keeping the model’s context short remains challenging for existing systems. Here we introduce Memorilla, a memory system that gathers information from an entire document collection into a fixed number of query-dependent ‘memory tokens’ for a frozen LLM. Memorilla encodes documents (such as user chats, book chapters) once into reusable embeddings, then learns to retrieve and place query-specific information in the LLM’s input embedding space. Using only 16 memory tokens, Memorilla achieves the best performance against baselines including Mem0 and MemoRAG on five of eight benchmarks spanning personalization, long-document and knowledge-intensive question answering. Notably, on FactKG, it reaches 84.33% accuracy versus 80.43% with full-context inference, while using 108x fewer tokens to represent the source documents. Furthermore, Memorilla can facilitate continual learning in agentic systems where the document count (the agent’s past turns) increases throughout the trajectory. In a TextWorld environment, we show that Memorilla reduces the agent’s tendency to wander between rooms it has already explored, and improves its average success rate from 18.6% with full-history in-context learning to 24.2%. Overall, Memorilla aids frozen LLMs in accumulating and using knowledge effectively across long-context interactions and takes a step towards building continually learning AI systems.
[AI-93] LTM: Latent Transition Based In-Context Memory for Vision-Language-Action Models
链接: https://arxiv.org/abs/2610.21362
作者: Qixin Sun,WenLong Gao,Xiansheng Chen,Zhongyuan Wang,Si Liu,Pengwei Wang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: Dataset Code: this https URL
Abstract:Robotic manipulation often requires remembering interactions that are no longer visible. We introduce Latent Transition Memory (LTM), which stores history as compact latent transitions between consecutive observations. Instantiated on \pi_0.5 as LTM- \pi , our method uses a frozen latent action encoder, a lightweight adapter, and causal KV caching for incremental inference without memory-specific semantic annotations or external reasoning-model calls. With DreamDojo transitions, LTM- \pi achieves 58.54% success on RoboMME and 80.5% on LIBERO-Mem, performing best among baselines without external reasoning models or memory-specific annotations. Ablations support the contribution of transition content beyond temporal cues and the efficiency of delayed memory admission. Scaling RoboMME demonstrations tenfold improves success from 57.67% to 87.50%, while a Cup Swap diagnostic highlights the role of process supervision in temporal composition. Under a 200-token memory budget, success remains 84.00%, versus 87.50% without compression. Finally, real-robot counting success improves from 11.25% to 82.50%. Together, these results demonstrate the effectiveness, scalability, and efficiency of latent transition memory, highlighting its capacity for complex temporal information, bounded-context operation, and real-world robotic control.
[AI-94] Hierarchical Graph Convolutional Network Approach for Detecting Low-Quality Documents COLING2024 LREC
链接: https://arxiv.org/abs/2610.21350
作者: Jaeyoung Lee,Joonwon Jang,Misuk Kim
类目: Artificial Intelligence (cs.AI)
备注: Accepted at LREC-COLING 2024
Abstract:Consistency within a document is a crucial feature indicative of its quality. Recently, within the vast amount of information produced across various media, there exists a significant number of low-quality documents that either lack internal consistency or contain content utterly unrelated to their headlines. Such low-quality documents induce fatigue in readers and undermine the credibility of the media source that provided them. Consequently, research to automatically detect these low-quality documents based on natural language processing is imperative. In this study, we introduce a hierarchical graph convolutional network (HGCN) that can detect internal inconsistencies within a document and incongruences between the title and body. Moreover, we constructed the Inconsistency Dataset, leveraging published news data and its meta-data, to train our model to detect document inconsistencies. Experimental results demonstrated that the HGCN achieved superior performance with an accuracy of 91.20% on our constructedInconsistency Dataset, outperforming other comparative models. Additionally, on the publicly available incongruent-related dataset, the proposed methodology demonstrated a performance of 92.00%, validating its general applicability. Finally, an ablation study further confirmed the significant impact of meta-data utilization on performance enhancement. We anticipate that our model can be universally applied to detect and filter low-quality documents in the real world.
[AI-95] FedSkillGit: Federated Skill Evolution via Version Control
链接: https://arxiv.org/abs/2610.21348
作者: Boxun Lan,Jiamei Li,Weijian Li,Chengyu Yang,Zhuang Qi,Chengcheng Wan,Xiaohong Chen,Ming Hu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model agents can improve their capabilities by evolving reusable natural-language skills from execution experience. In real-world deployments, this experience is distributed across data owners who cannot pool their raw data, which motivates federated skill evolution. The natural first design lets each client evolve the whole skill document and upload the result for server-side aggregation. That design is poorly matched to skill evolution for three reasons. An evolved document can still carry client-specific content distilled from private traces, repeated transmission of a nearly unchanged document spends communication on knowledge that the population already holds and the contribution of a single client is buried in a document that the server cannot meaningfully attribute. We propose FedSkillGit, a federated skill evolution framework that replaces full-document updates with commits. Each participating client submits a bounded set of patches over named rules together with its local experience, which summarizes what the round taught it, and the pair passes a two-stage privacy filter before it leaves the client. The server deduplicates the submitted patches and the local experience of all participants, updates a global experience document that only the optimizers read, and merges the surviving patches into a candidate skill. The candidate is distributed to the clients, which score it on their own validation sets and report the results, and it is published only when the reported scores improve on the current version. Experiments on three datasets and three language model families show that FedSkillGit outperforms state-of-the-art baselines by up to 22% and approaches centralized evolution without sharing raw client data.
[AI-96] ARGUS: An Agenda-Driven Framework for Reliable User Simulation
链接: https://arxiv.org/abs/2610.21346
作者: Hang Zheng,Nan Jiang,Zichen Zhu,Zixu Yang,Zhiyang Tang,Shuai Fan,Lu Chen,Xiaobao Wu,Kai Yu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:High-fidelity user simulators are essential for evaluating and training agents in interactive environments. However, current LLM-based simulators struggle with reliability issues, such as goal omission and policy non-compliance. We propose ARGUS (Agenda-dRiven Goal-oriented User Simulator), a neuro-symbolic framework that maintains the user’s goal state using a dynamic agenda stack manipulated via symbolic operations. By rigorously decoupling state tracking from natural language generation, ARGUS facilitates strict goal progression and constraint-aligned interactions. For cost-efficient deployment, we distill ARGUS into compact models using a unified pipeline: evaluator-guided SFT followed by online RL with dense step-level rewards. To rigorously evaluate behavioral fidelity, we introduce a rubric-based metric Composite Behavioral Score (CBS). Evaluations on \tau^3 -Bench and MultiWOZ demonstrate the significant superiority of the ARGUS architecture and the efficacy of our training pipeline. Code and data will be released soon.
[AI-97] Quantum Versus Classical Machine Learning for Neuropsychiatric Disorders: A Feature-Matched Multi-Cohort Evaluation
链接: https://arxiv.org/abs/2610.21344
作者: Utsav Poudel,Anuj Nepal,Vyas Viswanath,Subramaniyaswamy Vairavasundaram
类目: Artificial Intelligence (cs.AI)
备注: 19 pages, 10 figures, 7 tables. Submitted to IEEE. Code and per-fold results: this https URL
Abstract:Resting-state functional connectivity is widely used in machine learning-based classification of neuropsychiatric disorders. Quantum machine learning models have been proposed for functional connectivity data, but few studies have evaluated them under leave-site-out validation or against matched classical models. We compared six quantum model families with classical ones using identical correlation-derived features across four public cohorts (4,308 subjects). Leave-site-out validation was used for the Autism Brain Imaging Data Exchange (ABIDE I), ADHD-200 (attention-deficit/hyperactivity disorder), and REST-meta-MDD (major depressive disorder) cohorts, while the UCLA Consortium for Neuropsychiatric Phenomics (UCLA-CNP) was evaluated using stratified 10-fold cross-validation. The quantum kernel was statistically indistinguishable from a matched classical kernel across all four cohorts, with the average accuracy difference within 1.4 percentage points at 95% confidence. On ABIDE I, a hybrid quantum convolutional network achieved similar accuracy to its classical counterpart (0.645 vs. 0.641), while a trainable quantum kernel exceeded its classical counterpart by 0.064 in area under the curve (AUC). Quantum convolutional networks also achieved higher AUC than their matched classical networks on this cohort and showed smaller train-test gaps on all four cohorts, without an accuracy gain. Overall, most quantum models matched their classical counterparts, but none consistently outperformed them. This study, among the first quantum evaluations of ABIDE I under leave-site-out validation, provides a feature-matched benchmark of quantum and classical models comprising 284 matched comparisons. We further analyse fidelity-based graph construction and account for the observed quantum-classical parity through effective-dimension and eigenspectrum analyses.
[AI-98] Does On-Policy Distillation Need to Complete Every Rollout?
链接: https://arxiv.org/abs/2610.21337
作者: Zhengxin Zhang,Maosen Tang,Ning Wang,Changhao Li,Jingren Zhou,Bo Dai,Sainyam Galhotra,Alex Townsend,Claire Cardie
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:On-policy distillation (OPD) effectively transfers capabilities from a teacher to a student through feedback on student-generated rollouts. However, generating complete student rollouts and computing teacher feedback for every token make OPD computationally expensive. We introduce Stochastic Truncated On-Policy Distillation (STOPD), which accelerates OPD by randomly truncating student rollouts and correcting the resulting sampling bias with blockwise inverse-probability weighting (IPW). At each training step, we sample a stopping position for each rollout before generation and let the student generate until that position or natural termination. We obtain teacher feedback on every generated token and use blockwise IPW to compensate for the lower probability of retaining later tokens while controlling the weights. We prove that this weighting preserves Vanilla OPD’s gradient sum in expectation and derive one-step update-error bounds for SGD, Adam, and Muon. We evaluate STOPD across multiple teacher-student model pairs, task domains, and optimizers. STOPD reduces generated tokens by 25-48% while maintaining accuracy comparable to Vanilla OPD, achieving 1.3-2.2x training speedups.
[AI-99] Constrained Multi-Agent Reinforcement Learning in Imagination
链接: https://arxiv.org/abs/2610.21319
作者: Raz Shmueli,Shahaf S. Shperberg
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Predicting cost in a multi-agent world model moves most constrained policy updates into imagination: the stronger model-free baseline needs 10 to 20 times CA-MAMBA’s interactions on the median SMAC configuration, and CA-MAMBA earns 2.8 to 12 times the baselines’ MAMuJoCo reward at equal budget. In controlled tests on one task per benchmark, pricing the model’s cost advantages at parity more than halves a cost-blind learner’s cost and meets a tight continuous-control limit the learned multiplier misses, suggesting the model’s cost estimates are not the bottleneck.
[AI-100] Money-Mule Detection as a Relational Learning Problem: A Study on the RBIH NFPC Dataset
链接: https://arxiv.org/abs/2610.21314
作者: Owais Ahmad,Anirudh Deodhar
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Money-mule networks move criminal funds through many small transfers, each of which looks ordinary on its own. What marks an account as a mule lies outside any single transfer: in the counterparties and addresses its network shares, and in the shape of its activity over time. Production systems remain largely rule-based and see neither, while most research classifies transactions rather than accounts, recovering account-level risk only by aggregating afterwards. We instead score the account directly, reading the bank’s relational schema as a heterogeneous graph in which rows are nodes and foreign keys are edges. Each account is scored from a multi-hop neighbourhood using per-column encoders, a relative-time encoding and HeteroGraphSAGE message passing, with that neighbourhood restricted to rows predating the account’s own pre-flag cutoff so that no model sees evidence recorded only after the account was flagged. On a five-year Reserve Bank Innovation Hub (RBIH) retail-banking dataset of 400 million transactions and 94,063 labelled accounts, 2.6% of them confirmed mules, we compare against RelGNN, XGBoost, LightGBM and CatBoost trained on identical labels, splits and cutoffs. Our detector recovers nine in ten held-out mules at 83% precision and a 0.5% false-alarm rate, finding 115 more mules than the strongest tree ensemble given the same evidence. At the tuned thresholds the gap is mostly in recall: XGBoost and LightGBM lead across the first hundred alerts, but only the graph models hold their precision deep into the queue. Since the graph is the schema itself and every score is dated per account, the approach deploys on a bank’s existing database without a custom feature graph build.
[AI-101] he Density of Thought: Predicting and Steering Reasoning in Large Language Models
链接: https://arxiv.org/abs/2610.21286
作者: Tanush Kalhan,Anupam Chattopadhyay,Sriram Ranga,Amritendu Mukherjee
类目: Artificial Intelligence (cs.AI)
备注: 9 pages, 7 tables, 2 figures
Abstract:We investigate whether variation in information density across chain-of-thought (CoT) reasoning steps predicts answer correctness and can serve as a controllable inference-time target. We prompt frozen pretrained models to solve problems via CoT and measure the density of each reasoning step with two independent metrics: LP variance (how unevenly information is distributed across steps, computed as the variance in mean per-token negative log-likelihood) and the Caveman compression ratio. We then test whether density correlates with, and can be steered to improve, answer correctness. Experiments across six model scales, two training paradigms, and 60 GSM8K problems yield four main findings. First, lower LP variance predicts correctness in supervised fine-tuned (SFT) models, but this relationship weakens with scale and reverses in reinforcement-learning (RL) distilled models. Training paradigm therefore appears more influential than model size in determining density behaviour. Second, synthetic experiments reveal an intrinsic complexity preference: models trained on uniformly distributed data learn complexity-favouring distributions 2.0 to 2.6 times more accurately than simplicity-favouring ones. Third, direct inference-time density steering reduces accuracy from 60.0% to 8.3%, indicating that natural density patterns are functionally important to reasoning. Fourth, pre-generation nudging combining hint injection with contrastive KV-cache steering improves accuracy by 13.3 percentage points, consistent across both metrics. Finally, we introduce Entropy-Gated Density Routing (EGDR), a self-supervised method using token-level Shannon entropy to gate proportional KVcache adjustments during generation, improving accuracy by 10.0 percentage points without any external guidance.
[AI-102] Fractional Power-of-Two Quantization for Efficient and Effective Multiplier-Free LLM Inference NEURIPS2026
链接: https://arxiv.org/abs/2610.21257
作者: Sunghyun Wee,Geunjae Choi,Hyeonjin Kim,Suyoung Kim,Nojun Kwak
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 37 pages, 19 figures, 22 tables. Accepted at NeurIPS 2026
Abstract:Large Language Models (LLMs) demand efficient inference, where the multiply-accumulate (MAC) operations in linear projections dominate compute and energy cost. Post-training quantization (PTQ) reduces this cost by mapping weights and activations to low-bit integers. Power-of-Two (PoT) weight quantization promises efficient multiplier-free LLM inference via shift-add logic, but suffers significant accuracy degradation at low bitwidths due to its coarse base-2 quantization grid. We introduce Fractional PoT (FPoT), a base- \sqrt2 (half-octave) fractional PoT grid that achieves practical 4-bit weight quantization while preserving a fully integer shift-add datapath. The method combines (i) the base- \sqrt2 weight grid with uniform integer activations, (ii) a calibration-time grid-alignment refinement absorbed into the static per-channel scale, and (iii) a lightweight 1-adder \sqrt2 approximation whose error is absorbed during weight calibration via approximate-grid quantization, yielding a processing element (PE) with significantly fewer gates than a 4-bit integer multiplier. On Llama-3-8B with 4-bit weight and activation quantization, FPoT improves zero-shot commonsense reasoning performance by 8.72% relative to the PoT baseline, while maintaining performance close to the uniform integer baseline, with only a negligible 0.4%p degradation. These results hold across various configurations spanning seven models and three bit-widths. Furthermore, FPoT can be integrated with state-of-the-art LLM PTQ methods, demonstrating framework-agnostic applicability.
[AI-103] AutoOnboarder: Autonomously Onboarding LLM Agents
链接: https://arxiv.org/abs/2610.21218
作者: Tanay Arora,Cong Liu
类目: Artificial Intelligence (cs.AI)
备注: 15 pages, 5 figures
Abstract:Agentic large language models have been documented to lose reasoning capability when placed inside a harness, the software that mediates their observations and actions. To reconcile these losses, we introduce AutoOnboarder, an autonomous system that adapts an agent to a task the way a trainer onboards a new employee. It places the agent under the supervision of an intelligent ``Onboarder’’ coding agent, which iteratively tailors its foundation model, prompt, and harness source code to simultaneously improve performance and decrease cost. Like a real onboarding process, the Onboarder is only given tens of examples, and all evaluation data is isolated on a separate virtual machine to prevent leakage. We report the lessons learned from building and testing AutoOnboarder on genomic-sequencing prior authorization form submission. Autonomously onboarded agents are able to match or exceed the performance of a Gemini 3.5 Flash baseline at up to 29 \times lower costs. However, these agents remain imperfect, and the Onboarding process varies greatly in terms of final performance. We conclude that future Onboarding systems require a robust environment for the Onboarder, including a structured thought interface, (controlled) internet access, and a wide model search space.
[AI-104] Rubric Judgments Lack Uncertainty: Conformal Weighting for LLM Evaluation
链接: https://arxiv.org/abs/2610.21214
作者: Licheng Pan,Haocheng Yang,Haoxuan Li,Yuzhou Chen,Yichen Sun,Yuan Lu,Xiaoxi Li,Hao Wang,Zhixuan Chu,Kui Ren
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Rubric-based evaluation is widely used to assess large language model (LLM) outputs by decomposing quality into multiple criteria and assigning each a binary judgment from an LLM verifier. However, judgment certainty varies substantially across criteria. Existing methods uniformly aggregate these judgments into a final score, implicitly treating every criterion as equally certain and overlooking this heterogeneity. As a result, uncertain judgments remain incorporated into reward estimation without distinction, potentially yielding misleading training signals for downstream reinforcement learning (RL). To address this limitation, we propose Uncertainty-aware Rubric Weighting (URW), a training-free framework that quantifies criterion-level judgment uncertainty via conformal calibration. Specifically, URW elicits a soft uncertainty signal from a fixed pool of paraphrased judgment templates and converts it into a per-criterion weight using split-conformal p -values with provable coverage guarantees. Extensive experiments demonstrate that URW consistently improves pairwise preference accuracy and delivers stronger downstream RL alignment.
[AI-105] Condition-Contrastive Extrapolation for On-Policy Distillation of Language Models
链接: https://arxiv.org/abs/2610.21195
作者: Wenxuan Luo,Chenxi Li,Haoyue Liu,Yongxin Guo,Xiaoying Tang
类目: Artificial Intelligence (cs.AI)
备注: 15 pages, 2 figures, 12 tables
Abstract:On-policy distillation (OPD) improves a student language model by training on its own generated trajectories with supervision from a stronger teacher. However, when conditioned on student-generated prefixes, the teacher can favor invalid continuations that violate the question’s premises, causing OPD to encourage erroneous reasoning steps. As a remedy, we introduce Condition-Contrastive Extrapolated On-Policy Distillation (CCE-OPD), which makes teacher supervision sensitive to question conditions. CCE-OPD incorporates (1) a fixed-prefix comparison in which the same frozen teacher evaluates the complete question and a counterpart with a selected condition masked, and (2) a bounded log-probability contrast that reweights the teacher distribution into an auxiliary distillation target. On 200 diagnostic reasoning states, condition contrast reduces the rate of preferring invalid continuations from 27.0% to 16.0%. Extensive experiments across six mathematical benchmarks show that CCE-OPD improves avg@8 over OPD by an average of 1.60 and 1.95 percentage points when distilling Qwen3-8B into 0.6B and 1.7B students, respectively. It also achieves the strongest results on four transfer benchmarks.
[AI-106] FedSumUp: User-Auditable Privacy-Aware Representation Sharing for Resource-Constrained Federated Intelligence
链接: https://arxiv.org/abs/2610.21194
作者: Jingwei Zhao,Tiantong Sun,Yuanyuan Song,Jinhuan Guo,Wei Zishuo,Shengyu Huang,Ziyuan Ren,Wei Chen,Xiaojin Zhang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 21 pages, including appendices. Jingwei Zhao and Tiantong Sun contributed equally. Corresponding author: Xiaojin Zhang
Abstract:Existing federated data condensation methods face privacy risks, as uploaded synthetic data may preserve sensitive information and remain vulnerable to reconstruction attacks. Moreover, their generation often incurs considerable client-side computational overhead, limiting deployment on resource-constrained devices. To address these issues, we propose FedSumUp, which employs a learnable lossy compressor on clients to generate compact and visually interpretable representations. By controlling information capacity, FedSumUp reduces privacy leakage while allowing users to directly assess the privacy-utility tradeoff. We further introduce a dual-objective optimization mechanism to preserve task-relevant semantics while suppressing reconstruction risks. The compressor is trained on the server and only performs lightweight inference on clients. Since the server cannot access raw data, we develop a data-free closed-loop optimization strategy that optimizes the compressor by minimizing the discrepancy between original and twice-compressed representations after structure-preserving upsampling. Experiments on BloodMNIST, CIFAR-10, and ChestMNIST demonstrate reconstruction PSNR as low as 8.56 dB (with synthetic PSNR as low as 5.45 dB), client runtime reductions up to 96.5% (from 778s to 27s on ChestMNIST), and competitive classification accuracy (89.72% ChestMNIST, 82.24% BloodMNIST, 65.63% CIFAR-10), while remaining inherently robust to non-IID data distributions.
[AI-107] Mine Sample Render: Test-Time Diversity Alignment Of Generative Models
链接: https://arxiv.org/abs/2610.21185
作者: Kostas Triaridis,Morris E. Sharp,Nebojsa Jojic
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Modern generative models produce high-quality individual outputs but struggle to sample diversely, even when explicitly instructed to do so. We introduce MINE-SAMPLE-RENDER (MSR), a test-time approach that externalizes the diversity into an explicit, corpus-derived sampling procedure. Our pipeline (i) mines an interpretable categorical schema from a corpus of real artifacts and supplies statistics for sampling; (ii) samples attribute configurations from the corpus-derived statistics; (iii) uses a generative model purely as a conditional renderer of a sampled configuration. Across weather report generation, user simulation, webpage design, text-to-image generation, and agent-generated 3D assets, we evaluate the diversity and output quality of MSR-guided generation and show that MSR improves alignment with corpus distributions and diversity of generated outputs.
[AI-108] Does Latent Chain-of-Thought Actually Reason ? A Single Unifying Taxonomy for Latent Reasoning
链接: https://arxiv.org/abs/2610.21182
作者: Jingjie Qiu,Sheldon Yu,Xunyi Jiang,Jiajin Liu,Zihan Huang,Zhendong Chu,Tong Yu,Julian McAuley,Junda Wu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Latent chain-of-thought replaces written reasoning steps with continuous hidden states, giving models additional steps to compute before answering. It remains unclear what computation these steps actually perform. High task accuracy and decodable intermediate answers do not establish that a model uses its latent chain to reach an answer. We study this question across twenty-one model-task configurations covering three training recipes, three reasoning tasks, and two backbones. We combine chain deletion and content replacement with probes and trajectory analysis to test whether the chain affects the answer, what it encodes, and how its states change. On logical tasks, COCONUT often carries an answer that is already readable before the chain begins. Its states settle early, and deleting the chain leaves predictions almost unchanged. Per-step supervision produces an ordered layout of reasoning steps and can make the model depend on the chain. Yet even an ordered chain can act as a placeholder, with deletion hurting performance while content replacement leaves answers unchanged. On arithmetic tasks, all tested recipes show dependence on chain content. These results show why readable latent states alone are insufficient evidence of reasoning. We introduce a taxonomy that separates what a chain represents from how the model uses it. Seen this way, earlier reports no longer compete: placeholder tokens, collapsed superposition and step-dependent causal roles are each consistent with a different kind of chain, each obtained on a different task, training recipe or measurement. Once a study says which property it tested, whether the model uses the chain, what the chain holds, or how its state moves, claims about latent reasoning can be compared.
[AI-109] Agent Eval: Evidence-Bounded Assessment of Stateful Agents
链接: https://arxiv.org/abs/2610.21177
作者: Ruiqi Gao(Alibaba Cloud, Alibaba Group)
类目: Artificial Intelligence (cs.AI)
备注: 23 pages, 6 figures
Abstract:Task completion alone does not determine whether a stateful agent satisfied its obligations, and a recorded judgment does not guarantee that its supporting evidence remains valid. We present AgentEval, a system that carries a contract-bound assessment record through tool execution, state and event capture, report qualification, and case maintenance. It separates what the agent did, what the evidence establishes, and whether the conclusion is supported for present use. On 32 tool-world administrations, independent checks agree on all 30 complete episodes, distinguishing satisfied, violated, and unresolved obligations; two interrupted administrations are retained. Controlled interventions on 48 evidence views expose 36 binding faults missed by hash-only checks without changing agent behavior. Scenario Studio connects reviewed cases to 12 two-turn reports with 24 independently checked answers. A separate public-benchmark study on 192 held-out executions reveals ten completed return subgoals within broader task failures and isolates the effect of missing evidence. These studies show how an executable assessment record makes scope, evidential support, and report reuse inspectable across an agent evaluation workflow.
[AI-110] MemScope: Evolving Agent Memory Architectures through Localized Code Repair
链接: https://arxiv.org/abs/2610.21170
作者: Miao Su,Yucan Guo,Zixuan Li,Long Bai,Zhongni Hou,Xiaolong Jin,Jiafeng Guo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Agent memory architectures define how LLM agents store and reuse experience across tasks, but they are typically fixed after deployment, limiting their adaptability across task types. Recent self-evolving approaches instead adapt these memory architectures using execution feedback, typically represented as evaluation scores or compressed summaries. However, such coarse-grained feedback is insufficient for precise failure localization in current agent systems, where long-horizon task execution involves multiple interacting components. Without a precise failure location, existing approaches lack a clear code-level target for modification, leaving the update scope weakly constrained and often leading to large-scale rewrites of an otherwise capable memory architecture. We propose MemScope, a framework for evidence-grounded, localized evolution of an existing memory architecture. Its tool-using Patch Agent jointly inspects raw execution traces, runtime memory state, and current architecture code to identify repair targets and apply bounded, in-place edits under provider-identity and memory-compatibility constraints. On GAIA, the trace-and-memory inspection variant achieves 86.67% accuracy, compared with 70.00% for trace-only and 63.33% for summary-only feedback. Under xBench tasks, the evolved memory architecture improves over its original version from 49 to 55 correct answers with DeepSeek V3.2 and from 50 to 53 with Kimi K2.7 Code. On disjoint ALFWorld unseen tasks, success rises from 50.75% to 71.64%. Across archived architecture-snapshot transitions, the median code-change ratio is 4.4%. These results support fine-grained execution evidence as a basis for localized, in-place memory-architecture evolution.
[AI-111] GRADS: Graph-Regulated Agent ic Data Synthesis with Controllable Difficulty
链接: https://arxiv.org/abs/2610.21167
作者: Gaocheng Yang,Ruijie Tao,Kai Ye,Senhua Zhang,Zhengqin Zhu,Ziyang Li
类目: Artificial Intelligence (cs.AI)
备注: 9 pages, 7 figures
Abstract:Deep research agents are increasingly deployed to retrieve precise information from large corpora, yet multi-hop, long-horizon tasks remain challenging. Synthetic data generation has emerged as a scalable solution, with existing pipelines typically expanding seed tasks to increase their apparent depth and breadth. However, this expansion does not reliably translate into necessary reasoning: newly introduced evidence may be only weakly coupled to the answer, creating a mismatch between intended and empirical difficulty. To address this problem, we introduce GRADS, a graph-guided agentic data synthesis framework with explicit difficulty control. GRADS runs a closed loop of seed selection, evidence expansion, question–answer construction, and verification while maintaining a graph of the evidence and inference dependencies required to derive each answer. At each step, it conditions generation on a relevant subgraph, preserving indispensable inference without accumulating irrelevant context. Experiments across security, software-engineering, and scientific-document domains show that GRADS consistently generates verified tasks with empirically separated difficulty distributions, as measured by independent solvers. Moreover, under matched training budgets, post-training on this data delivers consistent gains at every level: SFT raises the solution-termination rate to over 95% while cutting tool calls and turns by up to 75%, RL trims tool use by up to a further 19.5%, and on our held-out, domain-specific benchmarks, the resulting 4B model matches Sonnet~4 on software engineering and surpasses it on the scientific and security domains.
[AI-112] All-in-One Memory: Process-Rewarded Reinforcement Learning for Memory Agents
链接: https://arxiv.org/abs/2610.21147
作者: Bingbing Wang,Jingjie Lin,Weiming Qiao,Jing Li,Ruifeng Xu
类目: Artificial Intelligence (cs.AI)
备注: 9 pages, 5 figures
Abstract:Large language models increasingly rely on external memory to answer queries grounded in long-horizon histories such as multi-session dialogues and lengthy documents. However, existing memory methods often optimize retrieval and memory construction separately, without a shared reward space for evaluating search, writing, and stopping decisions. To address this, we propose All-in-One Memory (AiO-Mem), which organizes these operations as transitions over a shared evidence state and directly evaluates the semantic quality of each action in a unified reward space. AiO-Mem compiles heterogeneous histories into source-traceable multi-view events, providing a shared substrate for retrieval and memory writing. A unified memory policy actor then rolls out complete interaction trajectories over this shared evidence state. A separate trajectory-conditioned process reward model is trained through evidence-grounded supervised fine-tuning and trajectory-level DPO to assign process scores to the collected trajectories. These scores are aggregated into an action-balanced trajectory reward and used by Process-Aligned DAPO (PA-DAPO) to optimize the memory policy. Across long-context memory benchmarks, AiO-Mem consistently improves answer accuracy over both static memory systems and RL-trained memory agents.
[AI-113] CleanRM: Reward Modeling from Noisy Feedback with a Few Clean Feedback for RLHF
链接: https://arxiv.org/abs/2610.21138
作者: Licheng Pan,Haocheng Yang,Yuzhou Chen,Yichen Sun,Yuan Lu,Yunfei Teng,Jiajun Fan,Guangyi Liu,Zhixuan Chu,Hao Wang
类目: Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:Feedback noise is pervasive in preference data, arising from subjective human judgments, annotation mistakes, and imperfect LLM judges. Standard reward modeling trained on such feedback yields a biased estimate of the clean risk. Although classical denoising methods are directly applicable to noisy feedback learning, they generally assume a uniform noise ratio across samples. However, this assumption fails for feedback data, where noise is intrinsically sample-wise. To address this challenge, we propose CleanRM, which uses a few clean feedback to calibrate sample-wise noise ratios via optimal transport. Specifically, CleanRM aligns the joint distribution of preferences reverted from noisy feedback with that of clean feedback and then trains the reward model using a noise-corrected surrogate loss. Theoretically, we prove that the resulting objective is unbiased with respect to the clean risk. Extensive experiments demonstrate that CleanRM outperforms state-of-the-art denoising baselines and yields more effective downstream RLHF alignment.
[AI-114] Stronger Judges Do Not See More: Deployment-Integrity Blindness in LLM Evaluation
链接: https://arxiv.org/abs/2610.21135
作者: Jianchang Su,Dawei Xiang,Yifan Zhang,Wei Zhang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:LLM-agent evaluation reports behavioral scores but leaves the identity of the deployed runtime stack unreported. A tampered deployment that preserves evaluated outputs passes every output-only evaluator. We introduce DIB-Bench, a benchmark that measures this deployment-integrity blindness across 7,560 scored examples, six tamper types, and an evaluator ladder from regex scorers through LLM judges to GPT-5-nano. Stronger evaluators accept more tampered outputs: mean acceptance rises from 88.6% (rule-based) to 94.1% (GPT-5-nano), and a trigger-conditional tamper invisible on the evaluation suite produces 48-100% harmful outputs, rising with model capability. A PAC-style bound shows that tampers with zero output shift are undetectable from outputs at any sample size; a manifest-hash identity channel detects them with zero samples. Behavioral fingerprinting identifies the model component above chance (76.7%) but covers one of six stack components. We also propose a graded deployment-identity descriptor for evaluation reports.
[AI-115] DST-Net: Decomposed Spatio-Temporal Network for Time Series Forecasting
链接: https://arxiv.org/abs/2610.21123
作者: Yaohua Zha,Hai Wu,Shu-Tao Xia
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at the World Computing-Power Network Conference-Academic (WCNCA) 2026
Abstract:Series decomposition has emerged as a widely acknowledged paradigm in deep time series forecasting models. However, existing approaches predominantly employ a homogeneous modeling strategy for all subsequences, thereby overlooking the distinct inductive biases inherent in different components. We identify that the trend component exhibits dominant spatial coupling across channels but carries significant temporal redundancy, whereas the seasonal component is characterized by local temporal evolution but remains susceptible to attention distraction. To address this, we propose a novel disentangled architecture that tailors specific modeling strategies to align with the intrinsic properties of each component. For the trend branch, we design a Latent Graph Block, employing latent graph modeling to capture global topological dependencies and system-wide co-movements. Conversely, for the seasonal branch, we introduce a Gaussian Multi-scale Attention mechanism to mitigate attention distraction. By injecting learnable Gaussian priors, we impose structural constraints that compel the model to focus on multi-granular local contexts, effectively filtering out global noise and accurately capturing local evolutionary phases. Extensive experiments on multiple real-world benchmarks demonstrate the effectiveness of our proposed model.
[AI-116] Probe-and-Resume: When and How to Evaluate Agent Execution
链接: https://arxiv.org/abs/2610.21088
作者: Jiakai Xu,Tiancheng Ge,Tujie Guan,Yusheng Li,Rohan Timmaraju,Eugene Wu,Kostis Kaffes
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Test-time search can improve agent performance by exploring alternative actions and using a critic to select promising branches. We study this approach in terminal-based computer-use environments, which expose only a partial view of a large underlying state, and find that the Monte Carlo tree search (MCTS) configuration we evaluate does not significantly improve task success. We identify two sources of poor critic performance: the agent’s observations omit environment state needed for evaluation, and intermediate states offer limited evidence of whether a trajectory will succeed. At intermediate states, critics rank continuations only modestly better than chance, and environment access improves this little. We study when to invoke a critic and what access to give it to evaluate a node. Our findings motivate a policy that invokes the critic only at nodes the agent believes are final, rather than during partial progress. Further, the policy gives the critic read, write, and execute access to a fork of the computer’s current and initial state, allowing it to probe and perform potentially destructive actions to surface any hidden state the trajectory omits. We call the resulting design Probe-and-Resume: when the agent claims to be done, the critic probes a fork of its state, and a rejected agent resumes where it stopped. On Terminal-Bench 2.1, this raises success by 8.4 percentage points over MCTS with conventional critics, and approaches the success rate of the Codex CLI agent with the same underlying model at less than half its cost.
[AI-117] SciΠvot: Benchmarking Principal-Investigator Judgment at Pivotal Moments of Scientific AutoResearch Agents
链接: https://arxiv.org/abs/2610.21082
作者: Youran Sun,Xingyu Ren,Chugang Yi,Kejia Zhang,Di Wu,Yingkang Cao,Allen Zhou,Jianping Ye,Suying Liu,Jiaxuan Guo,Jianda Du,Weicheng Ye,Kecheng Liu,Hefan Zhang,Haixin Wang,Jie Xiao,Eric Yang,Haizhao Yang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents can keep scientific experiments running for months with little human supervision, but reliable autonomous research also requires scientific judgment. Evaluating scientific judgment is difficult because research decisions are open-ended and depend on project context. We construct Sci \Pi vot, a 248-case benchmark of scientific judgment based on human corrections. The benchmark draws on records from six months of research in which senior researchers ran LLM agents on 26 projects across six scientific fields and repeatedly intervened to correct the agents’ research judgments. In each case, an agent chooses between the original LLM plan and the human-corrected alternative, using only the project record available before the correction as context. We compare LLM and human performance and examine the effects of reasoning effort, access to the project record, and reconsideration. At medium reasoning effort with the project record, models systematically prefer the original LLM plan over the human-corrected alternative. Four of five frontier models perform below chance on the selected cases. The highest accuracy in the project condition across the tested models and reasoning effort settings is 63.5%, compared with 73.4% for fresh researchers. At medium effort, every model achieves lower accuracy with the project record than without it, and reconsideration does not consistently improve overall accuracy. Even with a human-corrected alternative available, frontier models struggle with the selected research decisions, while Sci \Pi vot supports systematic evaluation of scientific judgment for LLM agents.
[AI-118] OmicsBench: a long-horizon multi-turn benchmark of AI agents on research-study-scale bioinformatics
链接: https://arxiv.org/abs/2610.21074
作者: Gideon Shaked,Siddharth Nagisetty
类目: Artificial Intelligence (cs.AI); Genomics (q-bio.GN); Quantitative Methods (q-bio.QM)
备注: 32 pages, 12 figures
Abstract:Computational biologists increasingly delegate substantial analyses to AI agents in long-horizon, multi-turn sessions, yet most benchmarks of agents’ bioinformatics capabilities pose a single short task with a single-shot prompt. We present OmicsBench, a long-horizon, multi-turn bioinformatics benchmark for AI agents. Each of its 97 tasks links dependent subtasks into a directed acyclic graph, for 662 underlying subtasks in all. Each subtask is posed as one user turn and graded deterministically against an answer key, with optional in-context feedback on incorrect answers. Across 12 models on three harnesses, the slowest agent averages 3.6 hours per task, and no combination comes near saturating the benchmark, with mean scores of 0.43 to 0.57 out of a possible 1 without in-context feedback and 0.54 to 0.67 with it. Multi-turn evaluation also uncovers behavior that single-shot prompts cannot, as agents score lower on subtasks that depend on earlier subtasks, and in-context feedback helps agents recover some of that loss. Because deterministic scoring gives no qualitative account of how an agent failed, we also introduce an unsupervised failure mode discovery pipeline that describes, embeds, and clusters agents’ failed attempts to surface emergent failure modes. Applied to OmicsBench, this pipeline exposes qualitative shortcomings in how agents carry out bioinformatics analyses, ranging from agents applying their own statistical filters in place of the specified ones to agents writing whole answers from memory.
[AI-119] Approve What Exactly? Effect Binding and Trusted Rendering for Agent Step-Up Authorization
链接: https://arxiv.org/abs/2610.21063
作者: Sai Varun Kodathala,Prashanth Pollishetty,Jaylen Cargill
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 36 pages, 13 figures, 13 tables. Code and data at this https URL
Abstract:AI agents now pause for approval before they send money, delete files or email outsiders, and that approval is usually treated as a second factor. We argue it is one only when compromising the agent cannot produce it and when it names exactly what it approves, and we test both conditions. We define twelve attacks on the approval path, including decoy recipients, amount cancellation, card padding, label spoofing, enrollment squatting and threshold splitting, and run them against the Agent Payments Protocol (AP2), the Microsoft action-bound approval design and Model Context Protocol (MCP) elicitation. An audit of ten agent runtimes finds no third-party verifiable approval record, no flood limits where measurable, and post-approval argument changes in four of eight runtimes with measurable binding. As a reference defense we build ASU (Agent Step-Up), a gateway that renders the approval card from a canonical effect descriptor, has a passkey on the phone of the user sign its digest, and executes only the stored call. Tamarin proves its binding, single-use and key-level non-repudiation properties under three stated hypotheses (53 lemmas). Across 27,108 scripted trials and 609 AgentDojo attack pairs, ASU stops substitution, replay and grant stretching, and the trusted card cuts forged-narration attack success from 0.54 to 0.03. With 15 current language models as automated approvers, the trusted card raises the attack catch rate from 0.30 to 0.98, at the cost of refusing 16 percent of harmless near misses, mostly due to a placeholder artifact in our stimuli. A second device adds nothing measurable over a trusted on-host confirmation in these experiments, flooding still succeeds 0.45 to 0.60 of the time against a modeled approver, and approval cannot stop effects that policy already allows. Code, data and proofs: this https URL
[AI-120] Bellm an-Consistent Trajectory RL for LLM Arithmetic Reasoning
链接: https://arxiv.org/abs/2610.21060
作者: Lianfa Li
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Grade A as a project report
Abstract:To address sparse terminal rewards in LLM arithmetic reasoning, we propose a Bellman-consistent trajectory RL method. Our framework introduces Trajectory Bellman Residual Minimization (TBRM) to transform sparse rewards into dense token-level signals. We integrate a Reinforce Leave-One-Out (RLOO) group strategy to reduce gradient variance of Bellman residuals and use hindsight goal relabeling to accelerate intermediate learning. Evaluated on the Countdown task, our method alleviates optimization bottlenecks and stabilizes gradient variance significantly better than GRPO. The TBRM-RLOO framework boosts Pass@1 to 0.55–0.64 and achieves a state-of-the-art Pass@16 accuracy of 76.0–84.0%, outperforming the SFT, IPO/DPO, and RLOO baselines. Our approach demonstrates exceptional single-sample efficiency and successfully circumvents standard performance plateaus via test-time scaling.
[AI-121] Learning When to Withdraw: Curriculum Coverag e and Rule-Specific Failures in Fine-Tuned Language Agents
链接: https://arxiv.org/abs/2610.21045
作者: Mohamed Abouzahra
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Tool-using language agents must revise prepared actions when evidence changes or authority is revoked. Selective withdrawal means suspending affected actions, preserving supported work, and resuming only after sufficient repair. We study supervised fine-tuning in NAQD-Env, a synthetic environment with explicit dependencies and a deterministic reference policy. Four LoRA checkpoints of Qwen2.5-3B-Instruct compare an original curriculum with an enriched one, the latter under three system-prompt framings with identical views and targets. Three instruction-tuned baselines and all four checkpoints are evaluated under three inference prompts on fresh validation cases, held-out dependency families, and held-out structural combinations. Under the naqd prompt, SFT checkpoints reach decision agreement of 0.77-0.92 and withdrawal recall of 0.46-0.86, versus 0.45-0.63 and at most 0.06 for the baselines. Rule-level diagnostics qualify these gains. The original-curriculum checkpoint wrongly withdraws 26 of 34 actions with surviving alternative support; enriched checkpoints withdraw 0-1. Conversely, all SFT checkpoints narrow prohibited combinations less often than the 7B and 8B baselines, a case absent from both curricula. Delegation results show more withdrawal but poorer preservation of independently authorized actions, and all four checkpoints omit valid event reports. Framing contrasts vary across metrics and splits and do not establish a reproducible advantage or equivalence. One training run per arm, jointly changed curriculum properties, prior evaluation exposure, and baseline-SFT quantization differences limit causal attribution. The contribution is a method for detecting rule-specific regressions that aggregate accuracy can conceal, not evidence that fine-tuning alone ensures agent safety.
[AI-122] Failure Is the Mother of Success: Recovering Supervision with Grounding Hindsight
链接: https://arxiv.org/abs/2610.21018
作者: Sheldon Yu,Tong Yu,Nikki Lijing Kuang,Zhendong Chu,Zhenwei Tang,Sungchul Kim,Lina Yao,Julian McAuley,Junda Wu
类目: Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:When every rollout for a problem fails, GRPO receives neither a successful exemplar nor a nonzero group-relative advantage, so it learns nothing from exactly the problems it most needs to learn. The failed reasoning can nevertheless suggest what to try next. We present Replay-Selected Hindsight Distillation (RSHD), which turns failed attempts into an additional, tested source of supervision during GRPO training. The policy writes a corrective note about each failed attempt without seeing a reference answer or solution, then re-solves the problem with the note in context. A note is admitted only if its replays succeed more often than the original group did, after screening for notes that state the answer, and an admitted replay is distilled into the plain-prompt policy. Because admission is decided by fresh successes rather than by the failed rollout itself, the procedure can recover supervision from groups in which every attempt failed, and it applies equally to failures in mixed groups. Only the policy is trained, the only judge is the outcome verifier, and deployment requires no notes or extra inference. Across model families and sizes and two training domains, RSHD consistently improves over GRPO on in-domain benchmarks, and the gains carry over to out-of-domain evaluation sets.
[AI-123] Deep Reinforcement Learning for HPC Job Scheduling: A Statistical Evaluation
链接: https://arxiv.org/abs/2610.21008
作者: Justin M. Cheney,Clement N. Nyirenda
类目: Artificial Intelligence (cs.AI)
备注: 17 pages, 9 Tables, Accepted for Online Proceedings at SACAIR 2026 Conference
Abstract:Deep Reinforcement Learning (DRL) offers learned job selection for High-Performance Computing (HPC), but performance against heuristics alone does not establish value beyond action masking. Six DRL treatments are compared: Proximal Policy Optimization (PPO), Advantage Actor-Critic (A2C), and Deep Q-Network (DQN), each with and without masking. Three heuristics and a seed-paired masked uniform random control provide baselines across ten seeds on two real Slurm (Simple Linux Utility for Resource Management) workload traces, one homogeneous and one heterogeneous, under a common non-parametric framework. All masked treatments complete evaluation, whereas unmasked treatments are truncated. On the homogeneous trace, no masked DRL treatment significantly improves average waiting time or average and maximum bounded slowdown over the control. On the heterogeneous trace, MaskablePPO improves all three metrics relative to the control. Its average waiting time is equivalent to that of the three heuristics within a 10% margin, and its maximum bounded slowdown is 22% lower than that of Last-Come, First-Served (LCFS). MaskablePPO has the leading average waiting-time rank among complete treatments on both traces, but Nemenyi comparisons detect no significant difference from MaskableA2C. Among masked variants, MaskablePPO has the lowest recorded mean training time, while MaskableDQN has the lowest mean decision latency. These findings distinguish completion from baseline-specific scheduling gains and support including a masked random control when assessing the value of learned selection.
[AI-124] How to Train Your Little Drag on: Teaching a Small Language Model to Construct and Refine GPU Optimization Scripts
链接: https://arxiv.org/abs/2610.20980
作者: Annabelle Sujun Tang,Wei Tang,Songyao Jiang,Tianhui Li,Shengzhe Xu,Jiin Woo,Kaan Ozkara,Youngsuk Park,Zhen Jia,Hadi Esmaeilzadeh
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Compiler transformations interact, making GPU kernel optimization a search for combinations that preserve the computation while improving performance. We introduce HOW TO TRAIN YOUR LITTLE DRAGON, a training method that teaches a small language model both to construct complete optimization scripts and to improve existing ones. Instead of ranking candidates or generating kernel code, our model generates optimization scripts. We devise a compact, lossless representation whose deterministic decoding reconstructs the exact compiler trace, preserving its transformation parameters and dependencies. We use a compiler framework TVM MetaSchedule to build training data on 54,116 distinct workloads, and construct 108,594 examples in two training sets: 60,236 teach the model to generate an optimized script from an unoptimized kernel, while 48,358 teach it to produce a better script given the kernel and a somewhat optimized script. Supervised fine-tuning teaches the model to construct and improve optimization scripts from these search-derived examples. We then add reinforcement learning with a shaped reward that combines correctness with measured performance, awarding a latency-based bonus only to numerically correct kernels. We apply this training method to a model with 8 billion parameters. After supervised fine-tuning and reinforcement learning, the model achieves per-attempt correctness of 91.27% for construction and 92.01% for refinement, while reinforcement learning produces faster kernels than supervised training on 88 of 102 construction tasks and 145 of 170 refinement prompts. With only 8B parameters, our LITTLE DRAGON achieves higher per-attempt correctness and better kernel performance than six frontier LLMs: GPT-6 Astra, Opus 5, Kimi-K3, and GLM-5.3 among them, each with hundreds of billions to trillions of parameters, for which task-specific training is unavailable or expensive.
[AI-125] SOMA: Stack-of-Objects Manipulation using a Quadruped Robot
链接: https://arxiv.org/abs/2610.20968
作者: Stergios E. Bachoumas,Panagiotis Artemiadis
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 9 pages
Abstract:Executing heavy-duty physical manipulation in unpredictable environments remains a major challenge for quadrupedal robots lacking dedicated manipulators. This paper presents SOMA, a hierarchical reinforcement learning framework for whole-body non-prehensile manipulation of multi-object stacks using a quadruped robot. Unlike traditional methods restricted to two-dimensional planar pushing, SOMA integrates explicit robot base-height control, unlocking full three-dimensional pushing strategies. This capability enables the quadruped to manipulate complex payload stacks as a unified system while preventing relative sliding or tipping of upper layers. The architecture factorizes control into a high-level manipulation policy using a Simplified Structured State Space Sequence (S5) encoder to reason over long-horizon temporal history and predict bounded target tracking commands via Beta distributions, and a low-level locomotion policy that dynamically adapts leg stance configurations. In extensive simulation benchmarks, SOMA achieves a 90.3% success rate for single-object reposing and up to 90.1% for dynamic two-box stacks under stringent error margins ( 0.05\text m position error, 1^\circ orientation error). Finally, we validate zero-shot sim-to-real transfer on a Unitree Go1 quadruped robot across diverse payload masses and geometries, demonstrating active base-height adjustments to maintain traction and balance during heavy pushing tasks.
[AI-126] StepSafe: Learning Timely Intervention with Context-Aware Safety Reasoning
链接: https://arxiv.org/abs/2610.20960
作者: Siyi Gu,Kunhan Gao,Chao Peter Yang,Sukanta Ganguly,Rex Ying,Arman Cohan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Safety in agentic systems is inherently sequential: whether an action is safe depends on prior actions, tool outputs, information disclosures, and interactions among agents. Recent work has shifted agent safety guardrails from post-hoc trajectory-level assessment toward step-level action diagnosis and proactive risk prediction, yet existing training objectives provide limited supervision for how step evidence accumulates over time and when it becomes sufficient to justify intervention. We propose \methodlong~(ASPO): a training framework for step-level timely intervention. Pre-fix level stage alignment allows crediting each decision at its own step: it compares each response with the other responses sampled at the same prefix and rewards the action that fits the stage, continuing before the risk is visible, asking while clarification can still avert harm, and stopping before harm is irreversible. Adaptive feedback self-distillation further credits the reasoning behind each decision, with the model itself as a teacher that sees the step annotations on intervention steps and rules mined from its own false alarms on safe steps. Its training data carry a violation taxonomy and an intervention window at every step and are extended through self-evolving multi-agent red-teaming. Finally we present \textscStepSafe, a step-level reasoning guardrail with timely intervention, that achieves the state-of-the-art peformance among open-source guardrails across five agent-safety benchmarks on safety-utility pareto curve with Youden’s J of 53.0 (4B) against 46.2.
[AI-127] Screwtape: Adversarial Transcript Editing for Monitor Red-Teaming
链接: https://arxiv.org/abs/2610.20941
作者: Gosha Kozhevnikov,Tyler Tracy
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:We present Screwtape, a pipeline that converts real agent transcripts into attacks that evade action-only monitors. An LLM attacker edits a transcript so that the agent causes a specified harm, such as leaking an API key or spawning an unmonitored agent. The edited actions are never executed, so we use an LLM verifier to reject attacks that are inconsistent with the original transcript or below a severity bar. The attacker revises the edit until the monitor no longer flags it. We applied Screwtape to 94 Claude Code and Codex transcripts from Redwood Research staff, optimizing against our own monitors. Across 252 runs, it produced 35 attacks that passed the verifier and evaded these monitors. Six of them, confirmed valid by human review, also evaded a held-out monitor provided by a frontier developer. We therefore demonstrate that a red-teamer without access to a developer’s infrastructure can find confirmed failures of the developer’s monitor. We expect to find more attacks when the pipeline runs inside the developer’s infrastructure, with frontier attacker models that lack refusal safeguards and direct optimization against the production monitor. Developers can use these attacks to fix monitor blind spots, and evaluators can use them as counterexamples to safety claims about the monitor.
[AI-128] MAI-Gym: An Open All-in-One Platform for GUI Agent Evaluation and RL Training across Mobile and Computer Environments
链接: https://arxiv.org/abs/2610.20939
作者: Panrong Tong,Quyu Kong,Pengxiang Li,Tianyu Xia,Hanzhang Zhou,Lei Wang,Chenglin Cai,Long Li,Long Chen,Jianan Zhang,Zhengwen Zeng,Xu Zhang,Gongjie Zhang,Yue Wang,Steven Hoi
类目: Artificial Intelligence (cs.AI)
备注: 31 pages, 15 pages
Abstract:Online reinforcement learning (RL) with verifiable rewards has driven rapid progress in mathematics, coding, and tool use, but scaling this success to GUI agents remains challenging due to the high cost of interactive environments and the scarcity of realistic training tasks. We introduce MAI-Gym, an open all-in-one platform that unifies GUI agent evaluation, data collection, and online RL training across mobile, computer, and web environments. Through a shared rollout pipeline, MAI-Gym ensures consistent agent behavior across evaluation and training, supports ten established benchmarks, and scales to 20,000 parallel environment instances. Beyond scalable infrastructure, MAI-Gym enables automated synthesis of diverse GUI applications and challenging tasks, including resettable mocked apps and coherent cross-app worlds with execution-validated task verifiers. Using MAI-Gym, we train MAI-Gym-9B, a unified 9B agent for mobile and computer use, achieving state-of-the-art performance among open GUI agents under 10B parameters: 78.4% on MobileWorld and 76.2% on OSWorld-Verified. Importantly, RL in a mocked train-ticketing app transfers to its real-world counterpart, improving task success on a phys- ical phone from 28.8% to 48.1%. We release the framework, synthesis skills, mocked apps, 2,000 verified tasks, and the model checkpoint to support repro- ducible research and further advances in scalable GUI agent training.
[AI-129] RAC: A Relational Agent Contract for Constraint-Governed Engineering Tasks
链接: https://arxiv.org/abs/2610.20935
作者: Kaiyuan Zheng,Fengqing Jiang,Yichen Feng,Zubin Aaron Assadian,Yuetai Li,Bhaskar Ramasubramanian,Basel Alomair,Luyao Niu,Radha Poovendran
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Engineering agents must produce artifacts that satisfy exact, coupled constraints. Existing solutions often require agents to make semantic decisions and also derive consequences such as precise geometry or timing from those decisions, leading to failures in constraint-governed engineering tasks. In this paper, we introduce RAC, a harness-agnostic contract for the action and feedback layer of engineering agents. It specifies authoritative observations, semantic actions, tool-based realization, independent verification, and repair evidence. The agent decides which relations an artifact should satisfy; specialized tools derive their concrete consequences. On PCB placement, routing, and repair tasks, RAC achieves the highest clean-pass rate for all nine evaluated models, averaging 90.9%, compared with 74.0% for the strongest baseline, requirements-only delegation. In job-shop scheduling, direct relational actions achieve 96.3% quality pass versus 59.6% for code-mediated delivery with access to the same scheduling backend. Our evaluations on 141 embedded-code tasks across four microcontroller identify the action contract as a source of agent capability: preserving semantic control while delegating mechanically derivable outputs improves the reliability of constrained artifacts.
[AI-130] Defining and Generating Compositional Structure
链接: https://arxiv.org/abs/2610.20922
作者: Eric Elmoznino,Tejas Kasetty,Blaise Agüera y Arcas,Yoshua Bengio,Guillaume Lajoie
类目: Artificial Intelligence (cs.AI); Information Theory (cs.IT); Machine Learning (cs.LG)
备注:
Abstract:Compositional structure is pervasive in nature and across scales: atoms combine into molecules, molecules into cells, and cells into organisms, while further compositional hierarchies appear within each level. More broadly, compositionality, the reuse and recombination of parts into novel structures, is a fundamental source of interesting complexity in both natural and artificial systems. Yet despite its importance, and despite our apparent ease in recognizing it, mathematical accounts capturing what we intuitively mean by compositional structure remain limited. Here, we fill this gap by formalizing a simple idea: compositional structure emerges when parts, considered together, reveal shared structure that is not captured when those parts are modeled separately. We express this idea using algorithmic information theory, the mathematics of information and compression, which provides a natural way to identify structure in arbitrary data without specifying in advance what that structure must look like. Applied across scales, our definition informs where and how compositional structure emerges within an object. Using tractable approximations, we then invert the framework to generate compositional objects: we specify the desired form of their compositional structure without prescribing their particular content, and optimize directly over strings and images. The resulting objects exhibit readily recognizable hierarchical reuse and part-whole organization, suggesting that the definition captures compositional structure rather than merely complexity, regularity, or shared information. Finally, we discuss implications for machine learning, including a new perspective on depth in neural architectures and principles for constructing compositional data curricula capable of supporting open-ended learning.
[AI-131] Replicable Dueling Bandits
链接: https://arxiv.org/abs/2610.20908
作者: Rohan Deb,Aadirupa Saha,Udaya Ghai,Karan Singh,Arindam Banerjee
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Dueling bandits choose each comparison based on earlier noisy outcomes, so two runs on the same preferences can compare different pairs. We study replicable dueling bandits, where two executions with shared internal randomness and independent comparison outcomes must select the same ordered pair in every round with probability at least 1-\rho . Building on Beat-the-Mean, whose scores change as arms are eliminated and need not rank the best arm highest, we design the first such algorithms with sublinear regret. Under relaxed stochastic transitivity with parameter \gamma\ge 1 and the stochastic triangle inequality, we design an algorithm that achieves expected regret \widetildeO(\gamma^8 K^4 H_\mathrmp/\rho^2) , where H_\mathrmp=\sum_i=2^K(P_1i-1/2)^-1 . Under strong stochastic transitivity, we design an algorithm based on simultaneous elimination that improves the dependence on K and achieves \widetildeO(K^2 H_\mathrmp/\rho^2) regret. For linear dueling bandits, we design an algorithm that achieves regret \widetildeO(d^3\sqrtT/(c_\mu\rho)) . Experiments on synthetic and language-model preferences show replicability, with sublinear regret.
[AI-132] Grasp6D: Task-Informed Vision-Language Priors and Pose Refinement enable Zero-Shot Task-Oriented 6DoF Grasping
链接: https://arxiv.org/abs/2610.20901
作者: Manav Kulshrestha,Apoorva Vashisth,Damon Conover,Aniket Bera
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Task-Oriented Grasping (TOG) has emerged as an important skill for purposeful robotic manipulation. Given the task of using a knife to “slice bread,” a robot must choose to grasp from the handle, whereas for “hand[ing] the knife over,” it is preferable for the handle to remain unobstructed for safe handoff. Current approaches for TOG either focus on 2D planar grasping, remain limited by their reliance on task-agnostic grasp generators, or are constrained by reliance on finite, manually-annotated, task-specific grasp datasets that cannot cover all possible grasp-task pairs. To that end, we present TGrasp6D, a vision-language based approach for generating task-oriented 6DoF grasps in a zero-shot manner. Given a task and a single RGB-D image of an associated object, TGrasp6D reasons about the appropriate semantic affordance for the given task and object geometry to produce an initial task- informed prior. This generated prior is further refined via 6DoF pose optimization for collision avoidance while preserving task consistency. Our approach achieves state-of-the-art success rate of 94% on the TaskGrasp dataset, with the best baseline at 74%. We also deploy our approach on a Franka Research 3 robot to demonstrate zero-shot generalization to real world objects
[AI-133] Anticipatory Commonsense with Moving Objects: An ASP-Based Framework for Predictive Reasoning with Interacting Trajectories
链接: https://arxiv.org/abs/2610.20867
作者: Julius Monsen,Jakob Suchan,Mehul Bhatt
类目: Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO)
备注: Accepted at ACS 2026 / Advances in Cognitive Systems 2026. (to appear)
Abstract:We present a general declarative framework for anticipatory reasoning about interactive motion in complex dynamic environments. Our focus is on (explainable) anticipatory space-time mental model formation at a semantic level to heuristically guide low-level motion computation. Our method is hybrid, modular and elaboration tolerant integrating: 1) An ASP model characterising the dynamics of interactive motion trajectories; 2) Semantically-guided prediction for deriving `branching trajectory futures’; and 3) An energy-based optimisation model to evaluate, rank and filter anticipations. We evaluate with the Argoverse 2 benchmark, where each anticipated trajectory carries an explicit, inspectable event-level hypothesis, unlike end-to-end stochastic prediction.
[AI-134] SHRED: Self-Play for Heterogeneous Road-Users without Expert Demonstrations
链接: https://arxiv.org/abs/2610.20866
作者: Wei-Jer Chang,Zhouchonghao Wu,Akshay Rangesh,Weixin Li,Masayoshi Tomizuka,Wei Zhan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: Self-play for Autonomous Driving
Abstract:Realistic traffic simulation and autonomous planning both depend on modeling interactions among heterogeneous road users whose dynamics, constraints, and behaviors differ fundamentally. Existing approaches predominantly learn realistic interaction behavior from large-scale human driving demonstrations. In this work, we ask whether such behavior can instead emerge purely through self-play. We present SHRED, a framework that jointly trains vehicles, pedestrians, and cyclists through large-scale heterogeneous self-play from random initialization, without demonstrations or logged-trajectory replay. SHRED combines type-specific dynamics and objectives with procedural scenario generation and per-agent randomization, allowing diverse behaviors to emerge through closed-loop interaction. On the Waymo Open Sim Agents Challenge (WOSAC), SHRED jointly controls all three road-user types and outperforms both the closest demonstration-free self-play baseline and prior self-play methods that use demonstration-trained reference policies. Policies trained only on nuPlan map geometry retain comparable realism under zero-shot evaluation on Waymo. For planning, SHRED outperforms all compared planners on the interaction-heavy InterPlan benchmark. Controlled ablations show that heterogeneity in self-play improves both traffic realism and planning performance, with particularly large planning gains in interaction-heavy scenarios. Together, these results establish heterogeneous procedural self-play as a scalable, demonstration-free paradigm for learning realistic and interactive traffic behavior.
[AI-135] Amortizing Sequential Experimental Design in Reasoning LLM s
链接: https://arxiv.org/abs/2610.20832
作者: Anita Kriz,Antonin Berthon,Nicolás Astorga,Mihaela van der Schaar
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Despite their extensive pretraining knowledge and strong reasoning capabilities, large language models (LLMs) struggle to adaptively gather information in experimental design tasks. We study how to amortize sequential Bayesian experimental design (BED) into an LLM by explicitly training the reasoning that guides action selection using information-gain rewards. Our approach specifies the experimental generative model independently of the LLM’s own beliefs, allowing us to optimize expected information gain under a model of our choosing. We investigate the design choices for amortization, including alternative generative-model constructions, reward estimation, hypothesis-space specification, multi-step objectives, and computational costs relative to training- and inference-time BED baselines. On Twenty Questions, the resulting policies improve information acquisition and task performance, outperform inference- and training-time BED baselines at substantially lower computational cost, and generalize to unseen categories. We further demonstrate positive transfer from Twenty Questions training to Medi-Q, a real-world clinical dialogue benchmark. These results suggest that information gathering can be learned as a generalizable reasoning skill in LLMs.
[AI-136] EvolvingBrain: An Adaptive Skill System for Multi-modal Agents Self-Evolution in Open Worlds
链接: https://arxiv.org/abs/2610.20786
作者: Runxuan Liu,Jingqi Shi,Baoxin Wang,Honghong Zhao,Haoyuan Xu,Yueqing Sun,Jiafeng Liang,Tao He,Dayong Wu,Qi Gu,Shijin Wang,Ming Liu,Bing Qin
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-modal agents have made significant progress in interacting with open worlds. However, their ability to self-evolve remains limited. Supervised fine-tuning enables imitation but not active experience organization, while outcome-based reinforcement learning promotes exploration yet underuses intermediate trajectory information. Therefore, we propose EvolvingBrain, which uses an adaptive skill system as the brain for agent self-evolution, consisting of two following components. Stratified Skill Self-Evolution: extracts Procedural Patterns and Problematic Patterns from model-specific trajectories, summarizes experience by generating skills, and evolves through multiple interactions with the open world, ultimately forming a Stratified Skill Architecture consisting of Domain, Plan, and Action/Trap Skills, which provides the agent with a multi-granularity knowledge repository and a novel evolution approach. State-Aware Adaptive Execution: summarizes long-horizon interactions into Structured Abstractive Memory and uses it to guide the State-Aware Skill Fusion Strategy, enabling the agent to utilize knowledge according to its current state. Experiments on VAB-Minecraft and DiscoveryWorld demonstrate that EvolvingBrain improves success rates on open-world tasks and exhibits generalization capabilities. Our code is provided in the supplementary materials and will be publicly released in the camera-ready version.
[AI-137] Open-world AI: a Tale of Three Requirements KR
链接: https://arxiv.org/abs/2610.20782
作者: Jianyu Zhang,Léon Bottou
类目: Artificial Intelligence (cs.AI)
备注: Code: this https URL
Abstract:The closed world asks AI for specialized competence and lets it manufacture the iid data its theory assumes: a fixed criterion licenses collecting examples until the task is covered. The open world asks for versatility — taking up tasks a machine was not built for, from few examples and with little prior from a human designer — and offers data that is grouped, not shuffled. Its tasks are not known in advance, so a machine must do inference-time learning (1st requirement). Modeling tasks as sharing latent factors and decomposing the training objective, we derive two more: a rich representation, retaining more than iid generalization requires, and a predictive disentanglement that aligns inputs and outputs into matched low-dimensional groups. The analysis further predicts the need for a consistent inner learner, and the benefit of long contexts spanning related tasks. We test these claims on language models up to 8B, asking whether a model has a property (qualitative) rather than how much it wins (quantitative). Models that meet the requirements improve as data from related tasks accumulates; the rest degrade, and neither scale nor layer-type interleaving repairs this.
[AI-138] INFUSE: A Unified Surface Benchmark for Indirect Prompt Injection in Agent ic Systems
链接: https://arxiv.org/abs/2610.20776
作者: Thuy Dung Nguyen,Jingyi Guo,Rana Seda Dubauskas,Sreepriya Damuluru,Taylor T. Johnson,Kevin Leach
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:LLM-based agents increasingly execute long-horizon, end-to-end workflows involving shell commands, code modification, dependency installation, and interaction with remote services. This growing autonomy expands their exposure to indirect prompt injection (IPI), where adversarial instructions embedded in external content are interpreted as legitimate task directives. Existing evaluations, however, often focus on isolated malicious strings, single-step tool calls, or narrowly scoped agent configurations. We introduce INFUSE, an end-to-end benchmark for evaluating IPI throughout the lifecycle of long-horizon agent workflows. INFUSE comprises 40 realistic tasks and a structured taxonomy of 56 attack scenarios spanning five threat categories. Attacks are injected through realistic interaction surfaces encountered during task execution, extending evaluation beyond attacks placed only in initial prompts or isolated tool interactions. We evaluate 19 LLMs from the GPT, Claude, Gemini, Llama, Gemma, and Qwen families, producing 41,824 adversarial trajectories (26,220 without defenses) and 720 benign reference runs. Across the benchmark, mean attack success reaches 19.6% under LLM-based evaluation and 15.4% under heuristic verification, with per-model ASR-L ranging from 3.2% to 41.9%. Existing guard models provide limited protection, while prompt-level instructional hardening reduces mean attack success from 58.2% to 14.3%, revealing substantial but model-dependent security-utility trade-offs. Our code is available at this https URL.
[AI-139] When Reality Outpaces Cognition: A New Scaling Principle for Reliable Intelligence
链接: https://arxiv.org/abs/2610.20764
作者: Ahsan Adeel,Talha Bin Riaz
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Robotics (cs.RO)
备注:
Abstract:This paper challenges the understanding capabilities of existing AI paradigms and emerging world models by asking a fundamental question: How quickly can reality change before an intelligent system can no longer reliably perceive, reason, and act? We formulate this as a theory- and model-agnostic challenge in which systems must infer reliable actions for increasingly distant future states across a growing observation–reality gap. We operationalize this challenge through the effective Speed of Coherent Cognition ((S_C^\mathrmeff)), defined as the highest tested Speed of Reality ((S_R)) through which a system continuously satisfies a predefined behavioural reliability criterion from baseline. As a proof of concept, reinforcement-learning agents are trained at the baseline (S_R) and evaluated, without retraining, as (S_R) is increased by up to (50\times) while computational resources remain fixed. We compare Transformer and gated-MLP baselines with (Co^4), an architecture inspired by cellular mechanisms implicated in conscious processing and awake thought. Across the tested scenarios, (Co^4) achieves substantially higher (S_C^\mathrmeff) than the baselines, motivating (S_C^\mathrmeff) as a new benchmark and the scaling of coherent cognition as a principle for reliable AI.
[AI-140] Optimal Hydra: Model Heterogeneous Federated Learning for LLM s using Data-Informed Hydra-Nets
链接: https://arxiv.org/abs/2610.20753
作者: Gwen Legate,Rahaf Aljundi,Eugene Belilovsky
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: currently under submission
Abstract:Large language models (LLMs) trained on vast amounts of data have demonstrated impressive performance over a range of tasks, but future performance gains will require increasing amounts of data and compute. Federated Learning (FL) facilitates training on decentralized private data, but client resource limitations can make training impractical, particularly as model sizes continue to expand. These resource constraints make the assumption of a shared server-client architecture impractical. Model heterogeneous FL offers a promising alternative by adapting model size to client capabilities; however, previous work is limited to the pre-training setting, and conclusions on sub-network selection do not necessarily hold for fine-tuning. Furthermore, prior approaches create client sub-networks without considering their unique data distributions, thereby overlooking the opportunity to tailor sub-network architecture to local data. Motivated by evidence that attention heads exhibit task dependency within layer defined feature spaces, we propose Optimal Hydra, a transformer based heterogeneous FL framework for fine-tuning that combines global layer selection with client specific attention head selection, aligned with local data distributions. Optimal Hydra makes explicit use of client data to produce client-specific sub-networks we call Hydra-nets. To our knowledge, this is the first model heterogeneous FL approach tailored to LLM fine-tuning. Experiments on three datasets show significant improvements over baseline methods while also reducing local resource demands.
[AI-141] Interpreting Language Model Parameters WWW
链接: https://arxiv.org/abs/2610.20748
作者: Lucius Bushnaq,Dan Braun,Oliver Clive-Griffin,Bart Bussmann,Nathan Hu,Michael Ivanitskiy,Linda Linsefors,Lee Sharkey
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 124 pages, 47 figures. This paper was written to be read online, with interactive figures, at this https URL
Abstract:Mechanistic interpretability aims to uncover how neural networks use their parameters to implement their impressive neural algorithms. Although previous work has uncovered substantial structure in the intermediate representations that networks use, little progress has been made to understand how the parameters and nonlinearities of networks perform computations on those representations. In this work, we present a method that brings us closer to this understanding by decomposing a language model’s parameters into subcomponents that each implement only a small part of the model’s learned algorithm, while simultaneously requiring only a small fraction of those subcomponents to account for the network’s behavior on any input. The method, adVersarial Parameter Decomposition (VPD), optimizes for decompositions of neural network parameters into simple subcomponents that preserve the network’s input-output behavior even when many subcomponents are ablated, including under ablations that are adversarially selected to destroy behavior. This encourages learning subcomponents that provide short, mechanistically faithful descriptions of the network’s behavior that should aggregate appropriately into more global descriptions of the network’s learned algorithm. We study how sequences of interactions between these parameter subcomponents produce the network’s output on particular inputs, enabling a new kind of ‘circuit’ analysis. While more work remains to be done to deepen our understanding of how neural networks use their parameters to compute their behavior, our work suggests an approach to identify a small set of simple, mechanistically faithful subcomponents on which further mechanistic analysis can be based.
[AI-142] L2G-SR: Learning to Guide LLM -based Symbolic Regression
链接: https://arxiv.org/abs/2610.20685
作者: Letian Huang,Xiangyu Wang,Haowei Lin,Di He
类目: Artificial Intelligence (cs.AI)
备注: 40 pages, including references and appendices
Abstract:Symbolic regression searches for an equation that explains observed data. Numerical error is commonly used to guide the search, but it alone can favor an accurate surrogate over a candidate with the right structure. We propose L2G-SR, a framework for LLM-based symbolic regression that guides search with a learned structural verifier. We construct structural supervision signal using tree edit distance (TED), accounting for commutative operators and alternative expression forms. A verifier first aligns numerical observations with the formulas, then learns to score a candidate’s structural similarity to the target formula. At search time, the target formula is hidden and the verifier complements numerical error in parent selection. An analyst then uses verifier feedback and diagnostic tools to propose revisions, which a generator converts into refined expressions. On LLM-SRBench, L2G-SR achieves state-of-the-art numerical fit, equation level accuracy and TED across the evaluated LLM backbones and domains. L2G-SR demonstrates the value of learned structural guidance for LLM-based symbolic regression.
[AI-143] IneqAgent : Agent ic Inequality Proving in Lean with Symbolic Computation
链接: https://arxiv.org/abs/2610.20681
作者: Gaolei He,Jinhui Li,Ruobing Zuo,Xia Zeng,Zhengfeng Yang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Formal inequality proving requires mathematical reasoning and exact computation, often connected through intermediate lemmas and goal reformulations. We introduce IneqAgent, an agentic framework combining LLM reasoning with symbolic computation for inequality proving in Lean4. Lemma-level Monte Carlo Tree Search coordinates lemma construction, goal reformulation, and symbolic backend selection. Complementary sum-of-squares, Bernstein, and successive difference-substitution backends produce exact certificates for local Lean proof generation. The agent integrates these proofs with LLM-generated steps and uses Lean feedback to construct complete proofs checked against the original theorem statements. On four benchmarks, IneqAgent with DeepSeek-V4-Flash solves 175 of 186 problems (94.09%), achieving the highest overall success rate among the evaluated methods. Across four backbones, it achieves higher success rates at low LLM inference cost thresholds than Direct Agent using the same backbone, with lower mean inference costs over each method’s successful runs. We also obtain Lean-verified strengthenings of Dittert’s conjecture in dimensions four and five, and certify the Holens–Djokovic (4,4) inequality, implying Merris’s conjecture in dimension four.
[AI-144] Counterfactual Neighborhood Risk Fields: Evidence-Grounded Selective Safety for Vision-Language Models
链接: https://arxiv.org/abs/2610.20679
作者: Tianshi Wang,Jingsong Wang,Fengling Li,Jingjing Li,Xin Li,Lei Zhu
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Multimedia (cs.MM)
备注:
Abstract:Representation-based safety defenses for vision-language models (VLMs) often learn to separate harmful from benign inputs, but separation alone does not establish what a risk score means or when it should be trusted. We introduce Counterfactual Neighborhood Risk Fields (CNRF), an evidence-grounded framework for selective VLM safety. CNRF constructs context-matched benign-harmful counterfactual pairs, uses their representation differences as capability-related risk directions, and estimates neighborhood support separately from risk. This separation prevents unfamiliarity from being conflated with harmfulness and enables selective intervention rather than blanket rejection. A victim-adaptive interface accommodates different safety observations across VLM architectures and routes selected requests for same-victim response review, which either releases acceptable candidates or produces constrained alternatives with reduced harmful capability. Across three VLMs and five multimodal jailbreak families, CNRF achieves 86.06-89.30% macro defense success rate while retaining substantial benign-task utility. Controlled experiments further show that query-matched references and preserved context-direction correspondence materially improve risk ranking in the low-false-positive regime. CNRF thus reframes representation-based VLM safety from learning harmfulness boundaries to constructing applicable evidence for selective risk decisions.
[AI-145] When Correctness Feedback Erases Evidence: Phase-Separated Completeness Auditing in Verus NEURIPS2026
链接: https://arxiv.org/abs/2610.20676
作者: Adarsh Agrawal(Independent Researcher)
类目: Artificial Intelligence (cs.AI)
备注: Paper accepted at the NeurIPS 2026 Workshop on AI for Verifiable Coding
Abstract:Correctness feedback suppresses the very evidence that specification-completeness auditing is designed to uncover: programs that satisfy a written contract but violate intended behavior. We demonstrate this objective reversal on 12 Verus/Rust tasks across four model families under matched call budgets. Independent no-feedback discovery more than doubles certified-witness yield relative to IntentShield’s corrective-feedback loop (59.0% versus 27.1%). A format-matched iterative control places the larger share of this gap in feedback content, not independent restarts. The advantage holds in every model family and survives parser repair. A separate phase-separated execution matches the aggregate yield without reusing responses; most certified witnesses also fail held-out cases. Preserving these witnesses delivers stronger specification repair: Phase-Revised contracts achieve 81.3% implementation success, versus 62.5% for IS-Revised, with Direct-Revised close behind. Both no-feedback witness sources substantially outperform weakened contracts. The effect does not reproduce in the available Dafny and Lean evaluations, establishing an empirical boundary around the studied Verus/Rust workflow. Completeness auditing and correctness repair have opposing immediate objectives. Preserve and certify the admitted failure before repairing it.
[AI-146] AlphaDiana: A Measurement System for Agents in Verifiable Reasoning Tasks
链接: https://arxiv.org/abs/2610.20675
作者: Zhanke Zhou,Zongze Li,Weikai Huang,Xuan Li,Chentao Cao,Xiao Feng,Xiangyu Lu,Jinbo Hu,Yi Xie,Nico Pelleriti,Shiyang Liu,Menghan Lu,Max Zimmer,Brando Miranda,Jiangchao Yao,Bo Liu,Sanmi Koyejo,Sebastian Pokutta,Bo Han
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Agents pair reasoning models with harnesses that organize model calls, execute tools, load skill instructions, and retain history in memory. Understanding agent performance requires linking execution conditions to outcomes, behavior, and resource use. We introduce AlphaDiana, a measurement system with modular interfaces for datasets, agents, environments, and scorers. It preserves harness-native control loops while controlling file reuse and memory retention separately. AlphaDiana supports macro analyses of complete agents and micro interventions on agent components such as tools, skills, and memory under matched conditions. We study three open-source models and three open-source harnesses on five answer-based benchmarks, with two models also tested on two interactive benchmarks. We find that harnesses do not consistently improve the accuracy of reasoning: gains on questions unsolved by the model alone can be outweighed by losses on questions it answers correctly without tools. This inconsistency extends to component interventions: neither skill guidance nor retained history reliably improves accuracy. Behavior and output-token use must be assessed with correctness: low token entropy can accompany wrong answers, and less output per execution can mean more output tokens per correct answer when accuracy falls. AlphaDiana links answer correctness and execution status to traces and resource records for comparing agents and diagnosing failures under shared task and resource policies.
[AI-147] Value-Informed Schedule Trimming for Accelerated Sequential Monte Carlo Guidance in Discrete Diffusion
链接: https://arxiv.org/abs/2610.20669
作者: Jaehyeon Kim,Ziseok Lee,Siyoung Choi,Kyungsu Kim
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Reward-guided discrete diffusion enables test-time alignment with downstream objectives, but applying Sequential Monte Carlo (SMC) guidance at every denoising step is computationally expensive. Active guidance steps require additional reward evaluations and, for gradient-based proposals, value-gradient computations. Sparse guidance reduces this cost by guiding only a subset of timesteps, but commonly used schedules are fixed heuristics whose effectiveness can vary substantially across tasks. We propose VISTA, a training-free method that automatically selects guided timesteps under a fixed budget. VISTA formulates sparse guidance as approximation of the ideal full-guidance probability path and derives an estimable upper bound on (\ell_1) path deviation. Using value statistics collected from a small number of full-guidance warmup generations, VISTA optimizes the resulting empirical surrogate exactly via dynamic programming. The warmup generations can be retained as model outputs, and after schedule selection VISTA has the same per-generation guidance complexity as other sparse schedules with the same budget. Across text, biological sequence, and image generation tasks, VISTA discovers task-dependent schedules and outperforms hand-designed sparse schedules in most tested settings, while approaching or exceeding full-guidance performance on several task metrics. In toxic-text generation, for example, VISTA attains a 0.99 unique-toxic ratio compared with 1.00 for full guidance while reducing post-selection per-generation latency from 204.23 s to 38.16 s, a 5.35 \times speedup.
[AI-148] Evaluating Model Capabilities After Benchmark Saturation
链接: https://arxiv.org/abs/2610.20667
作者: Morio Hamada,Jehyeok Yeon,Matteo Migliarini,Justin Olive,Magda Dubois
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Rapid advances in AI capabilities are causing benchmarks to saturate. Once several models approach perfect scores, a benchmark loses its power to differentiate between them, and with it its utility for informing AI safety cases and frontier policies. Yet benchmark items produce far more information than the scores we typically extract from them: a full transcript of how each model solved the task. We investigate whether these transcripts can be used to assess model capabilities even after benchmark scores have saturated. We simulate saturation retrospectively using historical benchmark and model releases, and test whether protocols that access only correctly solved transcripts from saturated benchmarks can predict capabilities measured on held-out, unsaturated items. We evaluate representative contemporary approaches for extracting information from model transcripts: pairwise judging, single-transcript absolute scoring, and predictions from deterministic transcript features or embeddings. We find that saturated transcripts retain capability information beyond benchmark accuracy. Amongst our evaluated protocols, pairwise judging performs best, reducing prediction error by 5.3 times compared to a baseline using only saturated benchmark scores. This capability-relevant information is distributed across transcript sections, but concentrated in specific channels of information such as reasoning and tool-calls. However, the predictive power of protocols does not always generalise across distribution shifts – for example by underestimating the capabilities of the first reasoning models introduced in early 2025. Saturated benchmarks thus retain capability-predicting signal in their task transcripts, but extending the lifetime of existing benchmarks requires methods designed to extract these signals robustly.
[AI-149] he World Wont Wait: Evaluating and Teaching Agents Under a Live Clock
链接: https://arxiv.org/abs/2610.20665
作者: Ningxin Pan,Hanyu Li,Yeqing Shen,Ning Ding,Yehui Tang,Wentao Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents can solve complex interactive tasks, yet reasoning and tool calls take time. In continuously changing environments, an otherwise correct action may arrive too late. We introduce RealtimeBench, with 180 tasks across six game types, to evaluate agents while games continue running during inference. Paired evaluations show that GPT-6.1-sol outperforms the faster Jev when games are paused, but falls behind when they keep running. We propose Two-Timescale Offline Recursive Self-Improvement (TORSI): a slow teacher studies failures and tests corrections in paused games between attempts, then writes reusable instructions for a fast student. The student acts without teacher calls during gameplay, and validation scores determine which updates are adopted. Two rounds raise Jev’s average test score from 13.51 to 24.19 on a 0–100 progress scale, improving all six game types. The instructions also improve average scores for three other models without retuning. These findings point toward real-time intelligence: enabling powerful LLM agents to remain effective in a continuously changing world by organizing deliberation, learning, and action across timescales.
[AI-150] 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); Artificial Intelligence (cs.AI); Computational Finance (q-fin.CP); General Finance (q-fin.GN)
备注: 15 pages, 4 figures, 4 tables. v2: Corrected corresponding author to Xiang Hu and added Tong Liu’s email; scientific content unchanged
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.
[AI-151] Active Learning of Fractional-Order Viscoelastic Model Parameters for Realistic Haptic Rendering
链接: https://arxiv.org/abs/2512.00667
作者: Harun Tolasa,Gorkem Gemalmaz,Volkan Patoglu
类目: ystems and Control (eess.SY); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: This work has been submitted to the IEEE Transactions on Haptics for possible publication. 14 pages, 8 figures
Abstract:Effective medical simulators necessitate realistic haptic rendering of biological tissues that exhibit viscoelastic material properties, such as creep and stress relaxation. Fractional-order models provide an effective means of describing intrinsically time-dependent viscoelastic dynamics with few parameters, as they naturally capture memory effects. However, due to the unintuitive, frequency-dependent coupling among the order of the fractional element and other parameters, determining appropriate parameter values for fractional-order models that yield high perceived realism remains a significant challenge. In this study, we propose a systematic means of determining the parameters of fractional-order viscoelastic models that optimizes the perceived realism of haptic rendering across general populations. First, we demonstrate that the parameters of fractional-order models can be effectively optimized through active learning, using qualitative feedback-based human-in-the-loop (HiL) optimization, to ensure consistently high realism ratings for each individual. Second, we propose a rigorous method to combine HiL optimization results into an aggregate perceptual map trained on the entire dataset, and demonstrate how to select population-level optimal parameters from this representation that are broadly perceived as realistic across general populations. Finally, we provide evidence of the effectiveness of the generalized fractional-order viscoelastic model parameters for three viscoelastic materials by characterizing their perceived realism through human-subject experiments. Overall, generalized fractional-order viscoelastic models established through the proposed HiL optimization and aggregation approach possess the potential to significantly improve the sim-to-real transition performance of medical training simulators.
[AI-152] Discovering state-of-the-art quantum circuits with large language models
链接: https://arxiv.org/abs/2610.22367
作者: Andi Gu,Alek H. Kemeny,Susanne F. Yelin,Mikhail D. Lukin
类目: Quantum Physics (quant-ph); Artificial Intelligence (cs.AI)
备注: 46 pages, 10 figures, 6 tables
Abstract:Significant efforts are currently being directed toward the implementation of large-scale quantum information processors, but realizing useful quantum advantage also requires the development of efficient quantum algorithms. While for many practical problems the dominant resource costs are associated with a small set of computational primitives, optimizing them in the quantum setting presents unique challenges, including a reversibility constraint absent from ordinary classical computation. Here we introduce an evolutionary framework in which large language models (LLMs) propose circuit designs and an independent evaluator verifies and scores them. We apply this framework to tasks spanning numerical computation, chemistry, optimization, quantum simulation, and number theory. In seven of the eight tasks, the discovered circuits significantly reduce both Toffoli and qubit counts relative to the state of the art. In the eighth, the search discovers a new reversible architecture for prime-field modular inversion. By handling reversibility differently, the architecture makes divsteps – a classical algorithm previously impractical for reversible integer arithmetic – viable and substantially improves the Toffoli-qubit trade-off for modular inversion. These results open the door for evaluator-guided LLM co-design of new algorithmic structures for quantum computation.
[AI-153] When Is Muon Optimal? Radial Frag ility and Minimax Geometry
链接: https://arxiv.org/abs/2610.21549
作者: Ao Xu
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Muon replaces a matrix gradient by an approximate polar factor and has shown strong empirical performance in neural-network training. The Isotropic Curvature Model (ICM) provides a curvature-based explanation for this update by asking when gradient orthogonalization minimizes a one-step local-loss surrogate. We investigate the stability of this explanation when the fixed input radius of the ICM is replaced by a general radial law. We introduce the Random-Radius Isotropic Curvature Model (RR-ICM) and the Polar-Factor Admissibility Set, which encodes exact scaled-polar optimality as a geometric condition on the positive polar factor of the gradient. We derive an exact characterization of this set for arbitrary radial laws and show that all nonscalar admissible responses arise from radial atoms aligned with curvature kinks. Consequently, every atomless radial law exhibits zero-capacity rigidity: exact scaled-polar optimality is possible only when the gradient is already a scaled column isometry. To connect the curvature-based framework with the established spectral geometry of Muon, we introduce the Minimax RR-ICM by replacing directional averaging with worst-case optimization over probability laws on the sphere. We prove that every nonzero global minimizer lies on the polar ray of the gradient, while its scale is determined by a one-dimensional convex problem. The corresponding admissibility geometry is a positive-semidefinite trace slab, in sharp contrast to the spherical-zonoid geometry of the RR-ICM. Together, these results expose the radial fragility of average-curvature explanations and provide a new minimax variational characterization of Muon’s polar direction.
[AI-154] Hierarchical Interpretable Multitask Audio Representation Learning: An Analysis of Event-Scene Relationships through the Information Bottleneck
链接: https://arxiv.org/abs/2610.21520
作者: Claudia Montero-Ramírez,Alba Martínez-Serrano,Carmen Peláez-Moreno
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI)
备注:
Abstract:Joint Acoustic Scene Classification (ASC) and Audio Tagging (AT) models traditionally rely on unconstrained multihead architectures. However, this design ignores the natural hierarchical dependency between sounds and acoustic scenes. In this work, we investigate a hierarchical (sequential) conditioning paradigm that predicts scene categories directly from intermediate event probabilities. Through a large-scale empirical benchmark with 66 full training runs across three state-of-the-art backbones (PaSST, PANNs, and BEATs) and different loss-weighting settings (beta in [0.0, 1.0]), we show that hierarchical conditioning achieves competitive classification performance while altering multitask optimization dynamics. Unlike multihead baselines, which sequentially solve the easier ASC task before optimizing AT, the hierarchical architecture enables synchronized dual-task learning for specific loss-weighting settings (beta in [0.6, 0.9]). Furthermore, we propose a multi-perspective interpretability framework that combines an analysis of the information bottleneck, manifold geometry, and Formal Concept Analysis (FCA) on a Term-Frequency Inverse-Document-Frequency (TF-IDF) event representation, providing complementary insights into event-scene relationships. Our findings reveal that hierarchical conditioning systematically yields lower conditional entropy, higher mutual information, and superior latent cluster separability (higher Silhouette scores), establishing a more structured and semantically faithful intermediate feature space.
[AI-155] When do quantum views help? Complementary classical-quantum representations in multi-view fusion
链接: https://arxiv.org/abs/2610.21395
作者: Azadeh Alavi,Fatemeh Kouchmeshki,Hossein Akhoundi,Yongli Re,Abdolrahman Alavi,Muhammad Usman
类目: Quantum Physics (quant-ph); Artificial Intelligence (cs.AI)
备注: 35 pages, 4 figures
Abstract:Recent studies have reported promising results from quantum machine learning in drug discovery, healthcare, materials science and cybersecurity. Research includes comparisons with classical machine learning algorithms and the development of sequential hybrid pipelines. We propose that fusion can improve prediction when classical and quantum representations capture complementary predictive relationships. Our classical-quantum multi-view framework preserves the full classical input alongside simulated quantum features. Concatenation and attention combine the views; single-view comparisons, replacement and masking controls assess both views’ contributions. Controlled angular-polynomial experiments distinguish targets adequately represented by a single view from those that depend on complementary angular and polynomial relationships. For these mixed targets, fusion improves prediction by combining the relevant feature sets, while matched classical representations recover the same benefit, linking the improvement to their complementary predictive content. Controlled classification tests further show that gains depend on alignment between the chosen representation and the target, and on a suitable fusion rule. Benchmark validation shows improved mean predictive performance over classical baselines on Wine, Wisconsin Diagnostic Breast Cancer, three-class Fashion-MNIST and Steel Plates Faults within the evaluated protocols. Classical replacement and masking controls remain competitive, classical baselines lead on Digits and Covertype, and outcomes with missing or corrupted views depend on the available inputs. These findings support evaluating quantum representations for their complementary predictive contribution and matching the representation and fusion strategy to the target relationships.
[AI-156] When is a better LLM worth paying for? Stochastic Optimization with Switching Mixed-Fidelity Oracles
链接: https://arxiv.org/abs/2610.21139
作者: Swati Gupta,Tianjiao Li
类目: Optimization and Control (math.OC); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Large language models (LLMs) are increasingly used upstream of decision-making to clean data, detect label errors, and reconstruct missing values. Yet LLM-assisted data curation creates a new operational tradeoff: higher-fidelity models and richer prompts can be substantially more expensive, while the value of this additional fidelity depends on the downstream optimization accuracy. This raises the central question of this work: when should one pay for a better LLM, given a specific downstream task at hand? We study this question through stochastic optimization with mixed-fidelity first-order oracles, which can have systematic bias and state-dependent variance. We develop two cost-aware algorithmic frameworks that allow adaptive switching between stochastic oracles. With known bias, the first meta-algorithm applies a restarted, cost-aware policy that adaptively selects oracle fidelity across optimization stages. Its cumulative cost is no larger than that of the best fixed oracle in hindsight and can be substantially smaller. With unknown bias, we develop a budgeted certification scheme that uses occasional trusted-gradient evaluations to certify the optimization progress and eliminate unsuccessful oracle sources, and show an explicit tradeoff between certification cost and delayed detection of biased sources. We provide concrete realizations of these meta-algorithms by applying biased stochastic gradient extrapolation (SGE) and recursive regularization under state-dependent noise as subroutines. Synthetic and LLM-completion experiments illustrate the qualitative switching behavior predicted by the theory: the adaptive policy skips several expensive LLM oracles and substantially reduces cumulative cost relative to fixed high-fidelity alternatives while achieving comparable accuracy. Subjects: Optimization and Control (math.OC); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) Cite as: arXiv:2610.21139 [math.OC] (or arXiv:2610.21139v1 [math.OC] for this version) https://doi.org/10.48550/arXiv.2610.21139 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-157] Forecast lead time and the timing of crew deployment for weather-driven power outages using AI weather models
链接: https://arxiv.org/abs/2610.20918
作者: Mohammad Kian Golkar,Luciano Alves de Oliveira,Mohamad Alipour
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Artificial Intelligence (cs.AI)
备注:
Abstract:Utilities request mutual aid crews to restore power after severe storms, and requesting early secures more crews but relies on less accurate forecasts. We examined how forecast lead time affects the value of outage predictions for crew deployment, using operational forecasts from two AI weather models, GraphCast and FourCastNet, to predict outage occurrence for counties across the contiguous United States 24 to 120 hours ahead. Forecasts were evaluated by the share of outage burden captured by the counties where crews would be staged. Captured burden declined with lead time: the loss between three and five days was significant in all analyses, whereas the loss within three days was significant only for outages that had not yet started. Early deployment with all requested crews outperformed late deployment with fewer, and we report the availability ratio at which the two break even. Antecedent land surface predictors computed at forecast valid time include weather from the forecast period, which inflates their skill by 42 to 53% and reverses their apparent advantage over forecast meteorology at long lead times. Computed correctly, they improved occurrence skill by 17.5% at 24 hours but did not significantly change captured burden, a gap that attribution analysis linked to their static component. Prediction intervals covered only 0.476 of tropical cyclone events against a nominal 0.90. Calibration by hazard class corrected most of this shortfall at an 11% cost in interval width, as a diagnostic only, because hazard class is known only after the event.
机器学习
[LG-0] PathFM: Fast Generative Forecasting of Multivariate Trajectories
链接: https://arxiv.org/abs/2610.22358
作者: Jiachen Yao,Xiyuan Zhang,Annan Yu,Boran Han,Abdul Fatir Ansari,Shuai Zhang,Michael Bohlke-Schneider,Bernie Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Time series foundation models are becoming increasingly popular for general-purpose forecasting. However, most leading models predict marginal quantiles, leaving the dependence structure of future trajectories unspecified. This distinction matters for trajectory-dependent quantities such as cumulative demand, peak load, and event timing, yet point and marginal scores do not reveal it. To model this dependence, we introduce PathFM, a trajectory-based forecasting model that uses flow matching to generate complete future paths. We further distill the model to one or two steps and share history attention cache across draws. Together, these enable PathFM to generate multiple paths for a single series in roughly the time needed for one query to a quantile model. The distilled sampler also makes it practical to optimize a calibration loss directly. PathFM achieves competitive marginal accuracy on the established forecasting benchmarks fev-bench and TIME. To complement these evaluations, we introduce Forecast Path Score Bench (fps-bench), which augments fev-bench tasks with energy and variogram scores and downstream metrics. PathFM achieves a higher aggregate path score on fps-bench than all evaluated external baselines. Experiments on chaotic systems further show that our flow model sustains long-run variability, while strong baselines rapidly collapse. PathFM makes generative time series forecasting both competitive and practical, enabling diverse downstream applications.
[LG-1] Lean Graph Search for Formal Theorem Proving
链接: https://arxiv.org/abs/2610.22339
作者: Yihan Xu,Jia-Chen Gu,Junyi Zhang,Di Wu,Haikang Deng,Po-Nien Kung,Kai-Wei Chang,Amit Sahai,Wei Wang,Nanyun Peng
类目: Machine Learning (cs.LG)
*备注: Code is available at this https URL
Abstract:Retrieving relevant lemmas is crucial for formal theorem proving, enabling provers to build machine-checkable proofs from established mathematical results. Recent retrieval systems match library declarations to a target theorem or its intermediate proof steps, but make limited use of multi-hop dependencies among declarations. These dependencies can reveal supporting lemmas that are not necessarily lexically similar to the target theorem but recur in proofs of related problems. To this end, we propose LeanGraphSearch, a graph-based retrieval method for Lean 4 that constructs a graph from Mathlib’s forward dependencies and traverses it to uncover supporting lemmas through multi-hop dependency chains. Specifically, LeanGraphSearch links each declaration to the premises it depends on, uses a graph-based ranking algorithm to propagate relevance from semantically matched declarations to their dependencies, and reranks the expanded candidates against the original query. Across two mathematical retrieval benchmarks, LeanGraphSearch improves recall by up to 12% over the strongest baseline, LeanSearch v2, on MathlibQR and increases coverage of complete annotated proof routes on MathlibMPR, covering up to 3.7% of extra groups. More importantly, we evaluate whether these retrieval gains translate into improved end-to-end theorem proving by integrating LeanGraphSearch into an iterative proving loop guided by Lean feedback. Across three theorem proving benchmarks, LeanGraphSearch improves pass rate at 8, 16, 32 reflection turns by as much as 40% compared to vanilla prover and at most 10% compared to LeanSearch v2, showing higher accuracy-cost efficiency. These findings highlight the value of dependency graph search for theorem proving and motivate further research on structure-aware search for proof discovery.
[LG-2] Robust Reinforcement Learning for Dynamic Treatment Regimes under Model and Measurement Uncertainty
链接: https://arxiv.org/abs/2610.22336
作者: Chengze Xie,Xuan Wang,Ying Ding,Hairong Wang,Yan Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:We develop a robust reinforcement learning (RL) framework for learning adaptive treatment policies that generalize across patients and remain reliable under errors in the measured biomarker and shifts in the disease dynamics. By modeling the patient as a hidden context, we address generalization and robustness in a unified manner through robust contextual Markov decision processes. Applying our framework to adaptive therapy for metastatic prostate cancer in a simulator calibrated to clinical data, we obtain generalizable and robust policies that substantially outperform the existing adaptive therapy method. In particular, we attain robustness with almost no loss of performance when the biomarker is accurate and the disease evolves as anticipated, and find that the resulting policy Pareto-dominates the existing method across a diverse array of scenarios. Our findings suggest that robust RL holds great promise as a principled methodology for learning reliable adaptive treatment policies.
[LG-3] Backward Compatibility in Zero-Bit Neural Audio Watermarking
链接: https://arxiv.org/abs/2610.22296
作者: Admir Krnić,Kosta Pavlović,Petar Nedić,Željko Bolević,Slavko Kovačević,Igor Djurović
类目: ound (cs.SD); Machine Learning (cs.LG); Multimedia (cs.MM)
*备注:
Abstract:Neural watermarking systems are typically trained as tightly coupled embedder-detector pairs and implicitly assume static deployment. In real-world deployments, embedders and detectors are frequently updated and selectively decoupled for security and versioning purposes, which can break detection in previously embedded content and, more critically, increase false-positive rates on embedders that a given detector is intentionally not intended to recognize. In this paper the concept of backward compatibility in zero-bit audio watermarking is formalized. A training framework is proposed that pairs a new detector with a sequence of frozen legacy embedders (its legacy chain), using a performance-weighted sampling strategy so that the detector retains accuracy across all earlier versions. This coupling, while necessary for backward compatibility, is shown to induce an embedder-collapse effect that catastrophically degrades the true negative rate (TNR) against out-of-chain embedders. To counteract this, a Angular Diversity Loss (ADL) is introduced that explicitly forces each new embedder to occupy a distinct region of the output space. The combined approach restores near-perfect TNR with no measurable cost to detection accuracy or perceptual audio quality. Subjects: Sound (cs.SD); Machine Learning (cs.LG); Multimedia (cs.MM) Cite as: arXiv:2610.22296 [cs.SD] (or arXiv:2610.22296v1 [cs.SD] for this version) https://doi.org/10.48550/arXiv.2610.22296 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-4] Baselines and Score Centering in Policy Gradient Estimation
链接: https://arxiv.org/abs/2610.22263
作者: Shuo Han,Ziqing Hu
类目: Machine Learning (cs.LG)
*备注: 17 pages
Abstract:Reinforcement learning has become a widely used approach to post-training large language models. In a typical implementation, learning is off-policy: The policy generating the responses differs from the policy being trained because of training-inference mismatch or policy staleness in asynchronous training. This discrepancy is commonly addressed by importance sampling, whose ratio is usually regularized in practice through truncation, masking, or clipping. Under regularized importance sampling, the importance-weighted policy gradient is no longer invariant to constant shifts in the reward. This invariance is recovered by using the covariance between the reward and the importance-weighted score as the estimand, which coincides with the policy gradient when the ratio is not regularized. This note develops this covariance perspective for both sequence-level and token-level importance sampling. The covariance perspective reveals the roles of the group-mean baseline and score centering in policy gradient estimation. Both perform centering in covariance estimation, but they differ in the mean being subtracted: The group-mean baseline uses a sample mean, whereas score centering uses a population mean. The two lead to distinct unbiased covariance estimators with different variances. Whether both forms of centering should be used simultaneously depends on whether importance sampling is applied at the sequence level or at the token level. With sequence-level importance sampling, score centering adds no benefit once the group-mean baseline is used; with token-level importance sampling, using the group-mean baseline without score centering generally introduces bias when the importance ratio is regularized.
[LG-5] opology-Inspired Fine-Tuning for Wind-Field Super-Resolution
链接: https://arxiv.org/abs/2610.22262
作者: Arjun Dadhwal,Brian Summa
类目: Machine Learning (cs.LG)
*备注: 16 pages, 6 figures; extended version of a poster accepted at TopoInVis Connect 2026
Abstract:Scientific super-resolution (SR) models are typically optimized with pointwise reconstruction losses, yet scientific interpretation often hinges on extrema, connected regions, and multiscale scalar structure. We characterize this tension with pretrained CNN and GAN wind-field SR models. Pointwise metrics, bottleneck persistence-diagram (PD) distance, and merge-tree (MT) distance favor the CNN, whereas order-2 Wasserstein PD distances favor the sharper GAN, so even different distances on the same topological descriptor reward different properties. Motivated by this disagreement, we fine-tune the pretrained CNN with differentiable topology-inspired auxiliary losses on scalar speed, gradient magnitude, soft high-speed level sets, and local maxima, and with fixed-pair supervision, which constrains predicted values and persistence at ground-truth persistence-pair endpoints located offline with the Topology ToolKit. A loss study of 16 fine-tuning configurations, each compared with a matched standard fine-tuning control that uses only the original reconstruction loss, on a fixed 168-field benchmark (with topology evaluated on a fixed 160 by 160 subregion of each field), shows that, among the scalar-field losses, gradient supervision provides the strongest PD improvement but does not improve mean MT distance, whereas fixed-pair supervision gives the largest and most consistent joint improvement in PD and MT distances. The final objective combines these two signals. Relative to the pretrained CNN, it lowers mean bottleneck, W_2,\infty , and W_2,2 distances by 31%, 25%, and 27% (and by 32-35% relative to the control, which on its own worsens all three), lowers all three on 150 of 168 fields, and gains 1.30 dB PSNR, about half the gain of standard fine-tuning. Mean MT distance also improves, but on only 91 of those 150 fields, indicating that PD and MT capture complementary structure.
[LG-6] BitSparse: Compact Activation Storag e for ReLU Feed-Forward Networks
链接: https://arxiv.org/abs/2610.22256
作者: Max Zhu,Bohang Sun,Pietro Lio,Jialin Yu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reducing memory use in large language model (LLM) training is an ongoing challenge. Activation checkpointing saves memory by recomputation, while lossy compression changes the stored activations. The reemergence of sparse \relu and \relu^2 activations creates a third opportunity: losslessly compressing the activation cache itself. We present BitSparse, a GPU-resident lossless compression format for sparse activations. Compression and reconstruction run in custom Triton kernels fused into an autograd function, shrinking the activation cache while recovering it exactly for the backward pass. Bfloat16 (BF16) and FP8 activations are supported. Optional sign-bit packing further increases compression up to close to the theoretical limit. We analyze its compression ratio and compare it to the theoretical optimum. BitSparse lowers per-token VRAM by 31% and 30% Modded-NanoGPT and Nemotron-H respectively, with overheads of no more than 1% at practical sequence lengths. Additionally, a new preallocated buffer system avoids host synchronization during training when variable size tensors are allocated.
[LG-7] Explore Once Evaluate Many: Multiple Policy Evaluation for Stochastic Linear Bandits
链接: https://arxiv.org/abs/2610.22252
作者: Spyros Dragazis,Yilei Chen,Aldo Pacchiano
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We study how to evaluate K target policies simultaneously in stochastic contextual bandits with linear reward structure. Given a known d -dimensional feature map and reward-free access to policy queries, we propose PRISMA (Policy evaluation through Regularized Information Sharing and Mixture Allocation), an algorithm that estimates every target policy’s value to accuracy \varepsilon with probability at least 1-\delta . The algorithm uses \widetilde\mathcalO!\left( \frac(L^2S^2+\beta^2)(1\vee\rho^\star)\varepsilon^2 \cdot \log(K) \right) context draws and reward observations, where \rho^\star\le d is an instance-dependent measure of the difficulty of designing a shared sampling distribution. Here, L , S , and \beta bound the feature norm, reward parameter norm, and reward noise, respectively, and the suppressed factors are logarithmic in K , d , and 1/\delta . Our method combines reward-free planning, a regularized sampling design, and a single ridge regression estimator shared across all target policies. An adaptive doubling procedure selects the regularization scale without knowledge of \rho^\star . The guarantees require no numerical lower bound on second-moment eigenvalues and have no explicit dependence on the cardinalities of the context and action spaces. Policy-query costs are accounted for separately, with K queries per shared planning context.
[LG-8] Query-Efficient Rejection Sampling for Parallel Masked Diffusion
链接: https://arxiv.org/abs/2610.22238
作者: Shiyuan Zhang,Xuheng Li,Qiwei Di,Quanquan Gu
类目: Machine Learning (cs.LG); Probability (math.PR); Machine Learning (stat.ML)
*备注: 22 pages,2 figures
Abstract:Masked diffusion models (MDMs) accelerate generation by unmasking multiple tokens in parallel, but sampling these tokens independently ignores their correlations and biases the output. Anari et al. (2026) corrected this bias with rejection sampling, decomposing the log acceptance probability into a sum of token-level terms, but the evaluation requires one score query per unmasked token. We propose the Single-Increment Rejection Sampler (SIRS), which provides an estimate of the log acceptance probability from a single randomly sampled increment at the cost of one query. SIRS then converts a Poisson number of clipped estimates into accept/reject decisions through a Bernoulli factory adapted from Chen et al. (2026). Since each query returns the conditionals of all remaining tokens in the block, we further Rao-Blackwellize each increment by averaging over these tokens at no extra cost, thereby reducing clipping. We show that the KL-divergence between the data distribution and the SIRS output is bounded by a clipping term plus the model’s training error. With exact scores, parallel decoding pays the full within-block dependence, whereas SIRS pays only the part beyond the clipping range.
[LG-9] Architecting for Hardware Diversity: Metas End-to-End Co-Design for High-Performance AI Training on AMD GPUs MICRO2026 MICRO
链接: https://arxiv.org/abs/2610.22230
作者: Haoyu Zhang,Arshad Rahman,Adnan Aziz,Ajith Prabhudev,Amin Farmahini,Ben Rhoads,Benson Ma,Brian Shi,Cen Zhao,CK Luk,Dingming Wu,Emad El-Haraty,Eric Chen,Frank Li,Gary Huang,Gefei Zuo,Haoran Zhang,Hiwot Kassa,Hongyi Zeng,Jack Taylor,Jeremy Hadidjojo,John Donati,Jonathan Sheffield,Julio Machado Silva,Karthik Ganesan,Kent Russell,KR Kishore,Lei Tian,Liu Ke,Muhammad Osama,Peng Chen,Rakesh Hemnani,Ryan Swann,Saleel Kudchadker,Shankar Yanamandram,Shobhit Kanaujia,Shucai Xiao,Srinath Bayareddy,Srinivas Vaidyanathan,Sudharssun Subramanian,Supadchaya Puangpontip,Taylor Robie,Tianhao Huang,Tyler Hart,Viswesh Sankaran,Weijun Jiang,Wenkai Du,Xiaodong Wang,Xiaohu Guo,Yao Rong,Yinbin Ma,Yue Dong,Yuzhen Huang,Yvonne Liu,Zhan Shu,Dan Rabinovitsj,Ajit Mathews,Chunqiang Tang
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Hardware Architecture (cs.AR); Machine Learning (cs.LG)
*备注: 16 pages, 21 figures, 6 tables. To appear in the 59th IEEE/ACM International Symposium on Microarchitecture (MICRO 2026)
Abstract:Meta deploys a heterogeneous fleet comprising NVIDIA GPUs, AMD GPUs, and proprietary MTIA accelerators to reduce costs and mitigate vendor lock-in. However, the extensive history of NVIDIA-specific optimizations hinders performance portability to other hardware. This paper presents our methods for optimizing alternative accelerators, using AMD Instinct MI350X versus NVIDIA B200 for recommendation model training as a case study. We propose three reusable principles. First, hardware-model co-design is essential: rather than directly porting models from NVIDIA GPUs, revising model structures and software recipes allows full exploitation of accelerator-specific strengths (e.g., MI350X’s 1.6x HBM capacity relative to B200). Second, our analytical performance model enables zero-warmup optimal kernel selection across diverse datatypes and matrix shapes in production models, drastically simplifying tuning. Third, increasingly mature agentic workflows have lowered the barrier to porting from the CUDA ecosystem; we extend these workflows for microarchitecture-aware kernel optimizations and custom shim generation. Compared to the directly ported baseline, these methods deliver up to 1.5x improvement in end-to-end training throughput. To our knowledge, this is the first hyperscaler report on productionizing AMD GPUs for large-scale training. Our contributions have been upstreamed to the open-source PyTorch, Triton, and ROCm ecosystems.
[LG-10] Generalization Guarantees of Randomized Block Coordinate Descent for Nonconvex Learning
链接: https://arxiv.org/abs/2610.22213
作者: Ruixi Hu,Liang Wu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Randomized block coordinate descent (RBCD) is a popular optimization algorithm for many large-scale machine learning problems. The existing theoretical analysis of RBCD mainly focuses on its convergence behavior from an optimization perspective, while its generalization behavior remains less understood for nonconvex problems. In this paper, we establish high-probability generalization guarantees for RBCD with nonconvex objectives. In the general nonconvex setting, we derive bounds on the gradients of both empirical and population risks, requiring only standard smoothness assumptions. With polynomially decaying step sizes, we tune the iteration number to balance optimization and generalization, thereby quantifying how early stopping leads to the optimal convergence rate for the population gradients. In the gradient-dominated setting, we provide bounds in terms of functional suboptimality, which nearly match the optimistic rates in convex and low-noise settings. Finally, we develop dimension-independent bounds for loss functions of a specific structure, making the generalization guarantees applicable to high-dimensional problems.
[LG-11] LCT: Losslessly Compressed Tensors with GPU friendly execution for Training
链接: https://arxiv.org/abs/2610.22179
作者: Max Zhu,Bohang Sun,Pietro Lio,Jialin Yu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Lossless tensor compression could reduce memory usage during training, but training requires tensors to be repeatedly re-encoded. We introduce Losslessly Compressed Tensors (LCT), a GPU execution framework designed for this repeated encode-decode regime. LCT exploits the natural redundancy in floating point data to reduce storage for weights, activations and optimizer states. Highly optimized GPU kernels implement Huffman coding with precomputed codebooks for low compression and decompression overhead, essential for training. Our analysis characterizes LCT’s storage costs and adaptation to magnitude shifts. LCT achieves a compression ratio of up to 1.455\times , and is demonstrated in training and finetuning settings. On a 162M-parameter NanoGPT, LCT reduces memory usage by 24.6% for a 7.2% step-time increase. In Qwen3-4B fine-tuning, LCT saves up to 23.8% of peak memory, enabling a 3.5\times larger context length in training on the same GPU.
[LG-12] he importance of tensor handling and implementation of physics-informed DeepONets
链接: https://arxiv.org/abs/2610.22157
作者: Nahum Banks,Gustavo A.M. de Almeida,Sergio Maldonado
类目: Machine Learning (cs.LG); Computational Physics (physics.comp-ph)
*备注:
Abstract:Physics-informed neural networks (PINNs) reduce or eliminate the need for large labelled datasets by enforcing governing equations during training. Physics-informed Deep Operator Networks (PI-DeepONets) extend this approach to learning solution operators, enabling rapid prediction across families of problems. Despite these advantages, we show that their application to higher-dimensional problems remains limited by the computational cost of training. To address this issue, we propose a modified PI-DeepONet architecture that reduces this cost by removing redundancy in how the Branch network processes input functions across collocation points. The proposed formulation is mathematically equivalent to the original architecture but reduces input tensor sizes and the number of Branch network evaluations. We analyse the computational complexity of both architectures and derive theoretical scaling laws for training time with respect to the number of collocation points and input functions. The proposed architecture is evaluated on 1D, 2D, and 3D linear advection problems. In all cases, it achieves comparable predictive accuracy to the original PI-DeepONet while consistently reducing training time. In 2D experiments, a maximum speed-up of 3.75\times is observed, with empirical training-time scaling agreeing with the predicted linear dependence on collocation points and input functions. Furthermore, training a 3D model on an NVIDIA A100 GPU is only feasible using the proposed architecture, as the original formulation exceeds memory limits. These results demonstrate that careful consideration of how PI-DeepONets are implemented can yield substantial gains, enhancing the practical performance of PI-DeepONets and enabling their application to otherwise intractable physics-informed learning problems.
[LG-13] Beyond Fixed Acquisition Functions: Temporal Difference Active Learning for Perturbation Discovery
链接: https://arxiv.org/abs/2610.22122
作者: Xudong Zhang,Jianzhe Li,Yitong Tang,Peijie Zhou
类目: Machine Learning (cs.LG)
*备注: 41 pages, 8 figures (including appendices)
Abstract:Genetic and drug perturbation screens are important tools for elucidating disease mechanisms and validating therapeutic targets, but experimental costs limit the number of candidates that can be tested. Existing perturbation screening methods typically update the surrogate model with feedback while keeping the acquisition function fixed, limiting their ability to adapt selection to different response landscapes. Experimental feedback reveals current hits and informs subsequent selection by updating candidate predictions, providing a training signal for learning how to select experiments. We formulate budgeted perturbation discovery as a sequential decision problem and propose Temporal Difference Active Learning (TDAL), which learns an acquisition function online through temporal difference learning from a single screening trajectory, accounting for current hits and future discoveries. TDAL uses the Hellinger distance between hit probabilities for unmeasured candidates before and after an experiment to weight successor value, and reuses measured outcomes through leave-batch-out (LBO) to construct virtual transitions and alleviate training data scarcity. Across eight public screens spanning single, double, and triple perturbations under varied experimental configurations, TDAL achieves an average rank of 1.63 among eleven methods and ranks first on four tasks, demonstrating its adaptability across response landscapes.
[LG-14] Scene Reconstruction Using Ubiquitous Wireless Signals via Diffusion Posterior Inference Over physics-Grounded Representations
链接: https://arxiv.org/abs/2610.22120
作者: Vahid Yazdnian,Yasaman Ghasempour
类目: Machine Learning (cs.LG)
*备注:
Abstract:Knowing the layout of the physical world matters for many applications, yet the sensors we rely on for it (e.g., cameras and LiDAR) can fail in darkness, in adverse weather, or behind walls. Wireless signals do not face these limitations: every link between a base station and a user is shaped by the buildings it passes, reflects from, or is blocked by, leaving information about the surrounding scene embedded in the wireless channel measured at the endpoints. However, mapping channel measurements to the environment is a severely ill-posed problem: a measurement records only the structures that interact with the transmitted electromagnetic waves along its propagation paths, and even these observations can often be explained by multiple plausible scenes. We therefore formulate wireless scene reconstruction as posterior inference and develop a hierarchical framework with two complementary stages to approach this problem. A wireless foundation model, pretrained to respect propagation physics, transforms each channel measurement into a representation of the scene structure that shaped the observation. A diffusion model then jointly integrates representations from multiple measurements to model and sample the posterior over building occupancy, with consensus over samples producing the final reconstruction. We demonstrate that the framework can accurately recover building footprints from wireless channel measurements alone and show that its performance depends critically on both design components: physics-grounded representations expose scene-relevant information, while joint posterior inference coherently integrates complementary observations.
[LG-15] Can AI Forecast the Extremes? Evaluating Probabilistic Weather Models on Tropical Cyclones
链接: https://arxiv.org/abs/2610.22077
作者: Nathan Chalumeau,Xavier Renard,Remi Meynadier,Hugo Rakotoarimanga,Marcin Detyniecki,Emmanuel de Bezenac
类目: Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph)
*备注:
Abstract:Weather prediction is undergoing an important shift, with data-driven AI models outperforming leading numerical weather prediction systems on average at a fraction of the computational cost. Yet, whether this success extends to extreme events remains contested. Earlier evaluations have exposed important shortcomings, but have largely focused on deterministic models and metrics, leaving the capabilities of the latest probabilistic models less understood. We conduct an extensive benchmark of tropical cyclone forecasting, comparing GenCast, FourCastNet 3, and FGN with ECMWF’s operational physics-based ensemble, IFS-ENS. We develop a suite of probabilistic metrics and diagnostic analyses to assess cyclone tracking, intensity biases, uncertainty, and the wind-pressure relationship that characterizes storm structure. We find that, despite remaining limitations in intensity and ensemble spread, leading AI models substantially outperform IFS-ENS in cyclone tracking: FGN achieves better probabilistic location scores for 90% of evaluated storms in 2025. Further analyses reveal model-specific failure modes and suggest a lingering influence of pretraining data, pointing to opportunities for targeted improvement. Nonetheless, these results suggest that AI models are capable of capturing physical relationships that remain predictive well beyond typical weather conditions, motivating further investigation into how these capabilities emerge and how to strengthen them.
[LG-16] ABNAT: Attention-Based Noise-Aware Training for Robust Transformers on Analog Compute-in-Memory Systems
链接: https://arxiv.org/abs/2610.22065
作者: Wenshuai Yao,Wenyong Zhou,Hanyong Shao,Zhiyuan Ning,Yizhe Chen,Yuannuo Feng,Kechao Tang,Ru Huang
类目: Machine Learning (cs.LG)
*备注: 8 pages, 10 figures, 2 tables. Accepted for publication at the 32nd Asia and South Pacific Design Automation Conference (ASP-DAC 2027)
Abstract:Transformer-based models have become the backbone of a wide range of tasks, but their intensive computation and data movement incur substantial energy overhead. Analog compute-in-memory (CIM) offers an energy-efficient alternative, yet its inherent hardware non-idealities perturb stored weights and can significantly degrade model performance. Although existing noise-aware training (NAT) methods improve robustness by injecting noise during training, they largely overlook the critical role of attention mechanisms in Transformer this http URL bridge this gap, we analyze how CIM-induced noise reshapes attention in Transformers and identify two key observations: (1) the divergence between clean and noisy attention maps increases substantially in deeper layers, and (2) conventional NAT can remain overconfident even when its attention routes deviate markedly from the clean reference. Motivated by these findings, we propose ABNAT, an attention-based noise-aware training framework that teaches the noisy student to preserve the clean teacher’s headwise attention routes rather than merely reproduce answers, without hardware changes or inference overhead; selective variants further reduce training this http URL experiments across vision, language, and multimodal Transformers show that ABNAT consistently improves robustness over corresponding NAT baselines, improving noise-degraded accuracy by up to 11.28% and reducing attention deviation by up to 85.77%.
[LG-17] ETH: Data-Balanced Heterogeneous Pipeline Parallelism for Training Large Models
链接: https://arxiv.org/abs/2610.22035
作者: Mianjie Yu,Zhongmin Zhao,Zizhao Mo,Jianxiong Liao,Wenxuan Li,Yulin Qiao,Huanle Xu,Wenquan Yang,Yongqiang Yang,Zeren Li,Ruifeng Tang,Chengzhong Xu
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: 19 pages, 14 figures. Accepted at USENIX NSDI 2027
Abstract:Current heterogeneity-aware training systems for large models balance workloads by assigning unequal numbers of layers across pipeline stages, a strategy aimed at equalizing execution time through a unique pipeline configuration. However, this approach fails to address critical memory constraints and suffers from inherently coarse balancing granularity. In this paper, we propose Heterogeneous Pipeline Parallelism (HPP), a novel approach enabling efficient and fine-grained data balance across heterogeneous pipelines. To address HPP’s key challenges, prohibitive search overhead for optimal configurations and high communication overhead from asymmetric All-Reduce, we design the ETH training system. ETH employs a hierarchical workload balance framework that jointly optimizes balance across heterogeneous pipelines and within individual pipeline stages. To mitigate asymmetric communication costs, ETH implements block-wise gradient aggregation. Furthermore, it integrates specialized optimizations that dramatically reduce HPP configuration search overhead. Evaluations demonstrate ETH achieves up to 2x training speedup while reducing configuration generation overhead by up to 56x compared to state-of-the-art systems. The code is available at this https URL. Comments: 19 pages, 14 figures. Accepted at USENIX NSDI 2027 Subjects: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG) Cite as: arXiv:2610.22035 [cs.DC] (or arXiv:2610.22035v1 [cs.DC] for this version) https://doi.org/10.48550/arXiv.2610.22035 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-18] Augment the Cache Replace the Activation: A 2x2 Study of Internal-State Sharing
链接: https://arxiv.org/abs/2610.22026
作者: Adam Deng,Shicheng Liu,Yada Zhu
类目: Machine Learning (cs.LG)
*备注: 10 pages, 2 figures
Abstract:Multi-agent systems built from frozen language models communicate through text, which discards the internal state the sender already computed and makes the receiver pay to rebuild it. Work on internal channels avoids that cost, framing the problem as transfer: agent A sends information to agent B. Every agent pair that must communicate needs its own transfer bridge, and every agent added to a system needs a new bridge fitted against each previous agent. We reframe this problem as sharing: an agent publishes its internal state once into a pool, and every peer reads that pool through a single mechanism fitted to no sender in particular, so adding an agent adds a publisher and no new bridge. Whereas the carrier of the internal state and the fusion mechanism are usually chosen independently in the literature, our main finding is that these two choices are coupled; the deciding factor is where the carrier enters the receiver’s computation. Consumability is a property of the receiving computation rather than of the representation. We find that the key-value (KV) cache must be augmented rather than swapped in: the receiver cross-attends over a peer’s cache while keeping its own, and with only a small query-side adapter trained, matches the accuracy it gets from reading the source document. Activations invert this. They are best written straight into the receiver, with no bridge, no adapter and no training, which beats exposing them alongside the receiver’s own, though that advantage rests on knowing where in the receiver the state belongs. The rule is augment the KV cache, replace the activation, established under one protocol with a shared control ladder. One adapter, trained once, also serves publishers it never trained on, and a pool of several at once, but only those whose internal state lies close to the receiver’s own.
[LG-19] Not All Unlabeled Data Are Equal: Unlabeled Data Selection for Semi-Supervised QSAR Modeling with Large Chemical Repositories
链接: https://arxiv.org/abs/2610.22025
作者: Oliver Majer,Saso Dzeroski,Jurica Levatic
类目: Machine Learning (cs.LG)
*备注: Accepted at the 29th International Conference on Discovery Science, October 5-9, Mainz, Germany
Abstract:Semi-supervised learning (SSL) is particularly attractive in domains where labels are expensive to obtain, while unlabeled examples are readily available. Quantitative Structure-Activity Relationship (QSAR) modeling, which predicts molecular biological activity from chemical structure, is one such domain. Obtaining activity measurements requires costly wet-lab experiments, resulting in small labeled datasets, while chemical repositories such as ChEMBL contain millions of unlabeled molecules. Despite this apparent suitability, the potential of large chemical databases to improve QSAR models via SSL remains largely unexplored. Furthermore, it is unclear how unlabeled data should be selected in realistic SSL settings with genuine unlabeled data. We conduct a systematic study of unlabeled-data selection using 10 QSAR datasets extracted from ChEMBL and a pool of more than 400,000 unlabeled molecules. To this end, we evaluate three established SSL algorithms (self-training, co-training, and label propagation) and formulate four unlabeled-data selection strategies based on random, nearest-neighbor, furthest-neighbor, and mixed sampling. Our results show that unlabeled data from large chemical repositories can improve QSAR models, with at least one SSL configuration outperforming the supervised baseline on each of the 10 datasets. Moreover, the most reliable improvements are generally achieved by selecting a relatively small number of chemically distant molecules. This suggests that the composition of the unlabeled set may be more important than its size, challenging the common premise that SSL necessitates an abundance of unlabeled data.
[LG-20] Close Enough is Nowhere Near: Fidelity-Consumability Decoupling in KV Caches
链接: https://arxiv.org/abs/2610.22016
作者: Adam Deng,Shicheng Liu,Yada Zhu
类目: Machine Learning (cs.LG)
*备注: 10 pages, 3 figures
Abstract:Multi-agent LLM systems communicate in text, which is lossy and wasteful: the sender throws away the reading it already did. A growing literature instead transmits the KV cache, the natural carrier of that reading. We study the most appealing form of the idea: both models stay frozen and the sender’s cache is substituted for the reader’s with nothing trained. It does not work, and the obstruction belongs to the reader. A frozen transformer reads only caches close to the one it would have written itself, a region we call its tolerable set. We show that geometric similarity to that cache does not predict whether the reader can use it, and the failure is not gradual, so a more faithful substitute is not the remedy it appears to be. We explain this phenomenon with competitive attention. Experimentally, we test eight diverse substitution methods against two measured anchors: an uninformative cache baseline and the reader’s own accuracy from having read the passage. All sit at the baseline. That turns a long tail of separate negative results into one outcome with a stated condition for overturning it. A different design does work: if the reader keeps its own cache and attends over the sender’s as extra context, adapting only its queries, it beats its own baseline, and also clears the baseline with eight senders pooled where undirected attention over the same union falls below it. We offer this as a characterization on one probe, not a leaderboard claim. Our central contribution is a redirection: pursuing faithful substitute caches is the wrong target; adding a cache to what the reader already has is the promising one.
[LG-21] Stateful Testing for Deep Learning Compilers
链接: https://arxiv.org/abs/2610.22003
作者: Jiatao Ge,Ruixiang Qian,Yibo Hu,Yanzhou Mu,Yinglong Zou,Jiawei Liu,Zhenyu Chen,Chunrong Fang
类目: oftware Engineering (cs.SE); Machine Learning (cs.LG)
*备注: 27 pages, 6 figures
Abstract:Modern deep learning compilers compile models lazily and cache artifacts across invocations. Retained compiler state can make a later invocation incorrect even when it is correct in isolation. An analysis of 26 publicly reported issues in PyTorch, JAX, and TensorFlow shows that detecting such bugs requires sequences that first establish compiler state and then expose a failure. We present Rondo, a stateful testing method that represents each test case as a model paired with an invocation sequence. Rondo mutates invocation sequences and uses a Markov chain Monte Carlo sampler guided by compilation behavior to favor rarely observed behaviors. Its three-way differential oracle compares execution without compilation, compiled execution with compiler state reset, and compiled execution with state retained. It confirms stateful bugs by replaying the same invocation with state retained and reset while preserving inputs and program state, then shortens sequences while preserving the original bug. We evaluate Rondo on four deep learning frameworks using 24 common model architectures. Rondo detects 117 bugs, including 35 stateful bugs. We submit 39 bug reports, including 18 for stateful bugs; developers confirm 35 reports, including 16 for stateful bugs, and fix 18 bugs, including 11 stateful bugs. Rondo achieves the highest behavior coverage and compiler Python source-line coverage among the compared methods, with 99.26% alarm precision. It shortens 44 of 117 invocation sequences, reducing the lengths of these 44 sequences by 83.34% on average.
[LG-22] Interfaces and Modules: A Scalable Representation for Weight-Space Learning across Architectures
链接: https://arxiv.org/abs/2610.21995
作者: Michał Kozyra,Michael Bronstein
类目: Machine Learning (cs.LG)
*备注:
Abstract:Weight-space learning across architectures requires representations that respect parameter symmetries and scale to large networks. Existing equivariant tensor encoders require architecture-specific symmetry specifications, while neuron graphs make global attention costly. We introduce a typed graph organized around shared coordinate-permutation actions. Its nodes, called interfaces, represent coordinate spaces whose bound tensor axes must be permuted jointly; operation occurrences, called modules, form typed hyperedges. A compiler derives this graph from supported PyTorch models, with exact recovery of the permutation factors defined by its constraints. The Universal Weight Encoder (UWE) combines local equivariant updates to parameter features with global attention over invariant interface summaries. The small number of interfaces makes global attention practical, and a single encoder processes different supported architectures. UWE is competitive with architecture-specific state-of-the-art methods on established benchmarks. On a new zoo of 1,200 transformers spanning 100 architectures, it outperforms the evaluated baselines at ranking networks within unseen architectures, including architectures larger than those used for training, while using substantially less memory.
[LG-23] Learn Only Where You Cant Compute: Structured Corrections to Gaussian Process Posteriors
链接: https://arxiv.org/abs/2610.21977
作者: Tobias Weber,Carla Sagebiel,Philipp Hennig
类目: Machine Learning (cs.LG)
*备注: 25 pages, 12 figures, including appendices
Abstract:Physics-informed Gaussian processes (GPs) provide a principled framework for scientific inference by expressing uncertainty over empirical measurements, boundary conditions, and governing equations. However, they struggle to scale to larger datasets and nonlinear information. Conversely, diffusion models excel at learning complex distributions over functions from massive datasets, but struggle in scientific contexts where training data are scarce and physical or observational information must be incorporated. We propose a structured posterior-correction framework that combines a tractable Gaussian reference with data-driven corrections via generative learning. After a mean correction and covariance calibration, a diffusion model learns a targeted correction in a low-dimensional subspace selected to capture discrepancies from the reference, while the complement retains its Gaussian law. Across three scientific inference tasks, these corrections recover hidden modes, improve sample structure, and reduce prediction error relative to the Gaussian reference. Comparisons with full-field generative models reveal task-dependent benefits and tradeoffs between predictive accuracy and computational cost.
[LG-24] Uncertainty-Aware Federated Learning for Joint Communication and Computation in XR Systems
链接: https://arxiv.org/abs/2610.21969
作者: Robert-Jeron Reifert,Hayssam Dahrouj,Alaa Alameer Ahmad,Haris Gacanin,Aydin Sezgin
类目: Information Theory (cs.IT); Machine Learning (cs.LG)
*备注: 18 pages, 16 figures, 3 tables. This paper has been accepted for publication in IEEE Transactions on Mobile Computing
Abstract:The convergence of communication and computation, empowered by artificial intelligence (AI) enhanced capabilities, emerges as a key enabler for realizing the promising functionalities of the sixth-generation of communication systems (6G). To this end, this paper investigates an AI-native, robust network, where multiple base stations (BSs) simultaneously serve several extended reality (XR) devices. The paper then presents an innovative distributed deep learning-based approach to simultaneously manage the transmit and computing powers, and computation allocation, amidst uncertainties in both channel and computing states information. Under such an interference network scenario, the paper proposes a robust approach for minimizing the worst-case delay across the served XR devices, subject to computation and power constraints. The paper leverages a deep neural network (DNN) that can be trained either centrally or in a federated manner to map estimated channels and computation requirements to optimized resource allocations per BS. During training, uncertainty samples, reflecting both channel and computing estimation errors, are injected after the DNN output to account for such joint uncertainties. Our results validate the enhanced robust delay performance and scalability of joint uncertainty injection as compared to uncertainty-agnostic methods and state-of-the-art benchmarks, with federated learning-based solutions achieving near-centralized performance while enhancing efficiency, privacy, and adaptability.
[LG-25] ChronoSheaf: Joint Spatio-Temporal Spectral Filtering on Cellular Sheaves
链接: https://arxiv.org/abs/2610.21958
作者: Abeer Mostafa,Zahra Ahmadi
类目: Machine Learning (cs.LG)
*备注:
Abstract:Sheaf neural networks generalize graph neural networks by assigning each edge a learned linear restriction map that specifies how the local representations of adjacent nodes relate to one another, thereby mitigating oversmoothing. Existing sheaf architectures, however, define the sheaf Laplacian as an instantaneous spatial operator and therefore cannot explicitly model how signals evolve over time on a graph. We first introduce a spectral sheaf Laplacian, defined jointly over space and time, such that a single operator regulates signal propagation across both the spatial graph and the temporal domain. Directly extending this operator’s receptive field in space or time would require increasingly complex sheaf structures. Instead, we propose a spectral filter that expands the effective receptive field along both dimensions without enlarging the underlying sheaf. Building on this operator, we introduce ChronoSheaf, a spatio-temporal sheaf network with two key components. First, Transitional Restriction Maps (TRM) decompose each node’s restriction map into a spatial term and two temporal frames that independently control its relation to neighboring nodes at the preceding and following timestep, enabling asymmetric temporal influence. Second, the Dual Spectral Filter (DSF) applies a single bounded-degree polynomial filter with an adaptively estimated spectral radius to the joint spatio-temporal sheaf operator. We validate ChronoSheaf on a diverse set of spatio-temporal forecasting benchmarks, where it outperforms state-of-the-art baselines. The code is available at: this https URL.
[LG-26] he Lagrangian Structure of Cost-Penalized Active Feature Acquisition
链接: https://arxiv.org/abs/2610.21919
作者: Joseph Bingham
类目: Machine Learning (cs.LG)
*备注: 52 pages
Abstract:Active Feature Acquisition (AFA) sequentially selects features to maximize predictive accuracy under an acquisition budget. A natural heuristic greedily maximizes Expected Information Gain minus a cost penalty, EIG(j|x_S)-lamc_j (EIG-Cost), but its theoretical guarantees are unclear. We provide an a-posteriori certificate decomposing the gap to the optimal hard-budget policy into a budget-slack term and a greedy-suboptimality term Del_lam. The certificate is computable when the penalized optimum is tractable; on small heterogeneous-cost instances, we verify it exactly, finding Del_lam near zero across 4,000 solved cases. On the large clinical task, Del_lam is not computable, so the decomposition is diagnostic rather than a numerical guarantee. Our analysis identifies EIG-Cost as one-step greedy optimization of the Lagrangian relaxation of expected-cost AFA, with lam as its dual multiplier. For equal costs, we prove Del_lam = exp(-gamI(Y;X_[d])), where gam is the submodularity ratio, yielding a structural guarantee in the near-submodular regime. Supporting results establish an adaptive chain-rule identity showing that cumulative stepwise EIG optimizes transcript information, and prove that A - I(Y;X_A) need not be submodular, even for jointly Gaussian variables. This rules out a direct Nemhauser-Wolsey-Fisher bound; naive Bayes restores submodularity. We also bound the randomization gap between hard-budget and expected-cost optima, analyze sensitivity to dual-multiplier error under local smoothness, and derive a conditional CVAE generator-shift bound given a bound on forward predictive KL. We treat this KL as a diagnostic quantity, not as a transform of training loss. Experiments cover synthetic Gaussian settings, AFA benchmarks, and clinical differential diagnosis on MIMIC-IV. On MIMIC-IV, EIG-Cost achieves macro-F1 0.180 +/- 0.012 at a 30 budget, using 6.6% of full-panel cost.
[LG-27] he Tilted Sampling Problem: Optimism Posterior Sampling and Logarithmic Regret
链接: https://arxiv.org/abs/2610.21908
作者: Mohammad Pedramfar,Siamak Ravanbakhsh
类目: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
*备注:
Abstract:Many modern learning systems are used not to select a single best action, but to sample diverse high-quality outputs. Examples include entropy-regularized reinforcement learning, RLHF, and reward-guided generative modeling. Motivated by these settings, we study tilted sampling: an online learning problem in which the goal is to sample from the Gibbs distribution induced by an unknown reward function. We define sampling regret as the temperature-normalized cumulative KL divergence between the learner’s distribution and the Gibbs target, corresponding to the suboptimality gap in entropy-regularized reward maximization. Our framework generalizes recent sampling-based and KL-regularized formulations in bandits and reinforcement learning. Our main contribution is a general regret framework that captures the different uncertainty structure of tilted sampling and classical regret minimization. While standard regret scales as \widetildeO(\sqrtT) due to first-order uncertainty, sampling regret is governed by squared confidence widths, yielding logarithmic-in- T regret in standard settings at fixed temperature. We develop sampling analogues of UCB and Thompson sampling (UCS and TS ^2 ) and provide unified regret decompositions that explain and extend recent logarithmic guarantees, including the first Bayesian sampling-regret decomposition based on new bounds for posterior-random Gibbs distributions. Our results show that learning to sample from a soft-optimal distribution is statistically distinct from learning to identify a hard optimum.
[LG-28] Self-Calibrated Subspace Robust PCA for Parameter-Free Removal of Mixed Noise and Line Stripes in Hyperspectral Images
链接: https://arxiv.org/abs/2610.21896
作者: Wenzhe Wang
类目: Machine Learning (cs.LG)
*备注: 9 pages, 4 figures, 6 tables. 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. Code: this https URL
Abstract:Hyperspectral images (HSIs) are often corrupted by Gaussian noise, impulses and stripes whose intensities vary across bands. The representative coefficient correlated total variation (RCCTV) model restores HSIs efficiently, but it lacks Gaussian and stripe terms, and its noise-aware extensions require the noise level and a stripe budget to be specified in advance. This article proposes a self-calibrated RCCTV (SC-RCCTV) model, in which the noise parameters are estimated adaptively during the iterations. A band-wise dense term and line-constant column and row stripe terms are introduced into RCCTV. The noise level of each band is initialized by a stripe-invariant median absolute deviation estimator and refined online from the residual, which turns the separation of sparse and dense noise into a band-adaptive Huber loss, and the stripe thresholds follow from a significance test whose false-alarm rate can be predicted analytically. No parameter is tuned for a specific image, and every subproblem of the alternating direction method of multipliers (ADMM) solver is solved in closed form. Under line stripes, SC-RCCTV outperforms the parameter-free FastHyMix by 8.17 dB on native-size cubes, and, without any noise information, it exceeds S3TTV, which is given the true noise parameters, by 1.52 dB while running more than 1,900 times faster. On average, it stays within 0.01 dB of the same model given the true noise levels, and its online calibration improves on a two-stage estimate-then-denoise scheme. Experiments on real Indian Pines and Urban data further delineate its scope for stripes that do not span entire lines.
[LG-29] Graph Dirichlet Energy for Fitting-Free Bivariate Causal Inference
链接: https://arxiv.org/abs/2610.21892
作者: Povilas Daniušis
类目: Machine Learning (cs.LG)
*备注: 25 pages, 2 figures, 3 tables
Abstract:We propose the Graph Dirichlet Energy Causal Inference (GDECI) criterion for distinguishing cause from effect in bivariate settings. For a candidate direction X\to Y , we construct a kernel similarity graph from X and treat Y as a graph signal. The score is the Dirichlet energy of this signal, induced by the unnormalized graph Laplacian. The reverse score swaps the variables, and the lower-energy direction is inferred as causal. The method requires neither regression fitting nor independence testing. After the empirical probability integral transform, its decisions are invariant under strictly monotone transformations of either variable, and the graph operations reduce to convolutions, allowing exact evaluation in O(n\log n) time and O(n) memory. At the population level, GDECI compares degree-weighted Bayes squared prediction errors plus a non-negative smoothness term of the regression function. As the bandwidth tends to zero, the weights become uniform and the smoothness term vanishes, reducing the criterion to a RECI-type comparison of expected conditional variances in copula space. Empirically, we compare GDECI with QCCD, GRCI, CAM, IGCI (two variants), RESIT, LOCI, QPE-k, and RECI (two variants), in accuracy on 1899 cause–effect pairs and in inference speed. GDECI ranked fifth of eleven methods in macro-averaged accuracy on 12 synthetic suites ( 0.776 ). On the real-world Tuebingen pairs, it ranked third in unweighted accuracy ( 0.684 ) and fifth in weighted accuracy ( 0.678 ). It was also the fastest of the compared methods up to n\approx10^4 . The accompanying code repository: this https URL.
[LG-30] Learning Lean Feedback for Efficient Proof-Candidate Evaluation on GPUs NEURIPS2026
链接: https://arxiv.org/abs/2610.21881
作者: Lazar Milikic,Etienne Bamas,Emmanuel Abbe,Viktor Kunčak
类目: Machine Learning (cs.LG)
*备注: 23 pages (9 main), 4 figures, 11 tables. NeurIPS 2026 Workshop on AI for Verifiable Coding (VeriCodeGen)
Abstract:Training language models for formal theorem proving requires large numbers of verifier calls. Slow checks and timeouts often leave GPUs waiting and withhold rewards from potentially correct proofs. We train neural evaluators to predict Lean 4 verdicts and compiler feedback on GPUs using 4.84M labelled programs. We compare two validity classifiers with a feedback model that annotates proof attempts with predicted goal states and error messages. A two-stage evaluator that screens candidates with a classifier and scores its positive predictions with the feedback model reaches 87.8% accuracy in predicting whether Lean accepts a candidate proof, evaluated on a 68,932-program benchmark collected from public provers. The combined evaluator processes 3.5 candidates per second on one H100, 4.2x the throughput of our 64-worker Lean baseline with environment reuse. Finally, we combine our evaluator with Lean verdicts to obtain a hybrid reward system that shortens GRPO training steps by approximately 31%. After one GRPO epoch with hybrid rewards, the prover achieves nearly the same miniF2F pass@1 as with Lean-only rewards (62.2% versus 62.4%) and a pass@32 of 72.4% versus 71.3%.
[LG-31] Lethe: Link Inference Attacks For Evaluation of Edge Unlearning Methods NEURIPS2026
链接: https://arxiv.org/abs/2610.21869
作者: Andrea D’Angelo,Francesco Gullo,Davide Mottin,Sayan Ranu,Giovanni Stilo
类目: Machine Learning (cs.LG)
*备注: Accepted at NEURIPS 2026 Evaluations Datasets track
Abstract:Graph Unlearning (GU) aims to remove the influence of specific training data from a Graph Neural Network without retraining anew. While most methods address removal of nodes and features, Edge Unlearning (EU), aiming at forgetting edges, has recently attracted dedicated attention as a standalone problem due to the hardness of completely removing the impact of an edge in the model. Despite this, no dedicated benchmark for EU exists besides isolated evaluations on feature-rich datasets such as Cora and Citeseer where GNNs exploit node features rather than graph structure. As such, we show two systematic limitations in existing evaluations: dedicated EU methods can be slower than retraining anew, and accuracy metrics are uninformative due to the negligible impact of edge removal. To address this gap, we introduce Lethe, the first comprehensive benchmark for Edge Unlearning, featuring 15 methods and 8 datasets. Our benchmark emphasizes large graphs, introduces evaluation tasks of varying difficulty, and proposes empirical measures based on link inference attacks rather than proxy accuracy metrics. We provide a comprehensive empirical analysis, demonstrate the shortcomings of current evaluations, and release Lethe as a reproducible, extensible benchmark.
[LG-32] Physics-Aware Distillation of a Grid Foundation Model for Real-Time Edge Inference
链接: https://arxiv.org/abs/2610.21833
作者: Vlad Alexandru Sănduleac,Shahariya Rasheed Rayamarakkarveettil Puthanpurayil,Noman Khan,Remus Teodorescu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Grid foundation models (GridFMs) are graph transformers pretrained on a masked power-flow objective, and evaluate power flow orders of magnitude faster than an iterative solver. Their size, however, targets data-centre inference, whereas low-inertia operation increasingly asks for inference at the edge, next to the measurement, at batch size one and within a single synchrophasor (PMU) reporting frame. We distil a 20.1M-parameter, 12-layer GridFM teacher into a 0.62M-parameter, 6-layer student, a 32.5\times parameter reduction at half the message-passing depth, using an objective that combines the masked power-flow loss with output, physics-residual, latent-state and attention transfer. The physics-residual term, which matches the teacher’s power-residual profile, has not previously been applied to typed power-grid graphs. We evaluate on a renewable variant of the IEEE 118-bus benchmark grid in which 26 of the conventional (synchronous) generators are replaced by inverter-based renewables with battery flexibility and standardised reactive-power limits (IEEE 1547), giving 97,529 physically valid (AC-OPF-feasible) operating scenarios, and we ablate all eight combinations of the transfer terms, reporting accuracy against ground truth and fidelity to the teacher separately. Distillation lowers student error on all four predicted quantities, by up to 48% on active-power dispatch, and every distilled student is more accurate than its teacher on reactive-power dispatch. The gain is largely carried by the output matching term. The physics-residual term, in the layer-aggregated form used here, is consistently harmful, which we attribute to the spatial averaging it applies to a bus-level residual field. The resulting student is 32.5\times smaller in parameter count and has half the message-passing depth, making it a candidate for subsequent edge-device benchmarking.
[LG-33] Local Epochs as Optimal Stopping: Theory and Evidence in Federated Learning
链接: https://arxiv.org/abs/2610.21832
作者: Ke Xiao,Qiyuan Wang,Christos Anagnostopoulos
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注: 27 pages, 7 figures, 18 tables. Includes appendices
Abstract:How long should each client train before communicating? Federated learning usually answers with a fixed, hand-tuned number of local epochs, yet an extra epoch helps only while the loss it removes outweighs the client drift it adds. We cast local training as finite-horizon optimal stopping: a client trains while the next epoch’s expected loss reduction exceeds its priced squared drift. We prove that this one-stage look-ahead threshold is optimal whenever stopping decisions are nested, derive closed forms for geometric and power-law returns, and show that an estimated rule loses utility only through wrong-signed decisions. Epoch-level evidence supports the model’s structure (geometric diminishing returns and additive drift) and reveals a simple regularity: the stage of training, not the client, sets the budget. The resulting rule, FedStop, is never more than one point below tuned FedAvg in mean accuracy. Where its budgets vary, it raises worst-decile accuracy by 4-8 points over tuned FedAvg and an adaptive-interval baseline, and every one of these gains remains significant after a Holm correction. Against the client-side stopping rules ALT and FLASH, it has the highest CIFAR-10 worst decile on every seed and is level elsewhere. It stays stable where per-client early stopping collapses, and each stopping decision takes constant time.
[LG-34] Compiling Multi-Threshold Operators for Quantized Non-Monotonic and Fine-Granular Activations in Custom-Tailored DNN Accelerators
链接: https://arxiv.org/abs/2610.21825
作者: Christoph Berganski,Felix Jentzsch,Marco Platzner
类目: Hardware Architecture (cs.AR); Machine Learning (cs.LG)
*备注:
Abstract:Aggressive quantization of neural networks can expose activation functions as bottlenecks for efficient hardware acceleration. Implementing quantized activations as multi-threshold operations has been shown to be a promising optimization, allowing for efficient hardware implementation of few-bit activations. However, current implementations are limited to monotonically increasing functions, quantization schemes with one-dimensional (per-channel) granularity and a small set of predefined activation-quantizer combinations. We formalize the multi-threshold representation to extend it to non-monotonic functions and fine-granular parameters, i.e., multi-dimensional thresholds. For automated conversion of quantized function compositions into threshold representation, we propose a systematic approach via algebraic rules integrated into the ONNX-based compiler infrastructure. We experimentally validate our solution on a selection of monotonic and non-monotonic functions by extending the binary search based multi-threshold operator that is integrated into the FINN framework for end-to-end synthesis of custom-tailored FPGA accelerators. The cost of hardware support for non-monotonicity and fine-granularity is characterized in comparison to an alternative approach utilizing existing operators via graph-level decomposition. We compare our method to prior work as well as the approach taken by the hls4ml framework. As an outlook, we demonstrate the prospects of using the multi-threshold representation during the training of a neural network as a form of quantization aware training.
[LG-35] A Closed Form Depth-Width Law for Neural Collapse in Deep Linear Networks
链接: https://arxiv.org/abs/2610.21820
作者: Pankhil Gawade,Kirsten Fischer,Guy Wolf,Ewa Szczurek
类目: Machine Learning (cs.LG)
*备注:
Abstract:How depth and width of neural networks shape learned representations is usually studied only after training. We show that, in a tractable feature-learning setting, this dependence can instead be obtained in closed form. Using neural collapse as a probe of representation geometry, we express NC1–NC3 directly in terms of the learned feature kernel, extending an identity known for NC1 to the angular part of NC2 and to NC3 under a ridge readout. For deep linear networks with exchangeable class structure, the resulting kernel equations yield a geometric law, exact within the theory, for the layerwise evolution of within-class variability. At large width, the accumulated compression becomes a function of the depth-to-width ratio L/N , with its rate determined self-consistently by the data, prior and architecture. Numerical experiments support the derived dependence on depth, width and signal-to-noise ratio. The result is a representation-level scaling law derived from feature-learning theory rather than fitted after training.
[LG-36] Measuring the local shape of learning-rate transfer curves
链接: https://arxiv.org/abs/2610.21814
作者: Satoki Ishikawa,Rio Yokota,Ryo Karakida
类目: Machine Learning (cs.LG)
*备注:
Abstract:Understanding the learning-rate (LR) dependence of trained model performance is essential for better optimization of neural networks, but the appropriate size of the LR depends on learning dynamics, architecture, and their hidden biases, and thus our understanding of this complex dependence remains limited. The behavior of the loss curve as a function of the LR has recently been observed in extensive experiments on LR transfer across different settings, such as model sizes, and it is becoming more important to explore methodologies for quantitatively characterizing the shape of this LR curve. In this work, we provide two relational equations that give a good empirical explanation of the gradient of the LR transfer curve, i.e., the LR gradient. One is a theoretical formula that holds at the edge of stability, and the other is an approximate expression obtained by naively treating this gradient as a hyper-parameter gradient. Using these two relational equations in a complementary manner, we show a lower bound on the gradient of the curve and a zero gradient under specific conditions for normalization layers in theory. The approximate expression also applies to a wide range of experimental settings and allows us to decompose the LR gradient into layer-wise components and quantify the strong contribution of specific layers to the curve shape. Thus, this serves as a foundation for better understanding the internal mechanisms of LR dependence.
[LG-37] Q(γ_H): Entropy-Guided Horizon Mixing in Multi-Discount Q-Learning ICONIP2026
链接: https://arxiv.org/abs/2610.21764
作者: Arwa Albelaihi,Abdulrahman Altahhan
类目: Machine Learning (cs.LG)
*备注: 15 pages. Accepted at ICONIP 2026
Abstract:A fixed discount factor in Q-learning imposes a single planning horizon throughout training, despite the differing demands of early exploration and long-term credit assignment. We propose Q( \gamma_H ), a multi-horizon Q-learning framework that maintains multiple discount-specific heads and adaptively combines them at control time using an entropy-derived confidence signal. The resulting state-dependent mixing weights shift control from shorter- to longer-horizon heads as policy confidence increases, enabling adaptive horizon mixing without modifying individual Bellman updates. Across classical control and deep reinforcement-learning benchmarks, Q( \gamma_H ) shows its strongest improvements on delayed-credit tasks relative to strong fixed-discount baselines. By decoupling horizon adaptation from the underlying learning updates, Q( \gamma_H ) provides a lightweight framework for adaptive planning-horizon control in reinforcement learning, avoiding the need for manual discount-factor selection while remaining compatible with standard value-learning methods.
[LG-38] Pairwise AUC Optimization Needs Corrective Power: A Unified View NEURIPS2026
链接: https://arxiv.org/abs/2610.21743
作者: Jia Chen,Zhiyong Yang,Shilong Bao,Qianqian Xu,Qingming Huang
类目: Machine Learning (cs.LG)
*备注: 71 pages, including references and appendices. Accepted at NeurIPS 2026
Abstract:Direct pairwise AUC-surrogate training is attractive because it trains toward the ranking metric used at test time, especially under class imbalance. This creates a direct alignment between the training objective and the evaluation metric, but such metric alignment alone does not guarantee that stochastic training remains corrective. We study this gap through Corrective Power: whether pairwise mini-batch updates produce stronger and more reliable improvements for more severe positive-negative ranking errors. Our main observation is that direct pairwise training can enter regions where severe errors remain common but no longer produce useful corrective motion. In such regions, even the mini-batch noise that would normally help SGD move along the steepest severe-error correction direction can become much weaker. In this paper, we seek a theoretical explanation for this by exploring the degenerate property of the U-statistics formed from the pairwise gradients therein. Our results provide a mechanistic explanation for several empirical patterns: (a) squared-hinge loss is often an effective AUC surrogate; (b) cross-entropy warm-up can help models that enter the AUC phase from a poor starting region; and © parameter-efficient fine-tuning from a pretrained model can reduce that need by starting closer to a favorable ranking geometry. Experiments on 8 CV tasks and 8 NLP tasks, using ResNet50, DenseNet121, DistilBERT-base, and Qwen2.5-1.5B-Instruct, support these predictions.
[LG-39] Sparse Federated Echo State Networks for Time-Series Classification via Smooth-L1 Regularization
链接: https://arxiv.org/abs/2610.21742
作者: Gengcheng Lyu,Manuel Röder,Frank-Michael Schleif
类目: Machine Learning (cs.LG)
*备注: 30 pages, 8 figures, 7 tables. Extended version of a paper presented at IDEAL 2025. Code: this https URL
Abstract:Echo State Networks (ESNs), as lightweight implementations of reservoir computing, provide efficient solutions for temporal data processing in resource-constrained environments. While standard ESNs offer computational efficiency, their dense readout layers inflate the stored model and the per-sample inference cost. This paper introduces a Smooth-L1 regularized ESN (SL1ESN) that promotes sparsity in the readout layer while preserving differentiability for gradient-based optimization. The method employs a twice-differentiable Smooth-L1 approximation compatible with Newton’s method, which we exploit both for centralized training and for a second-order (Newton-step) federated aggregation scheme. Under a single fixed shrinkage threshold shared by every dataset and both settings, experiments on seven time-series classification benchmarks show that SL1ESN learns readouts that are 77 – 95% sparse in the centralized setting and 75 – 95% sparse in the federated one, while staying within 0.7 accuracy points of the strongest dense baseline on five of the seven benchmarks in the centralized setting and within 1.6 points on the same five in the federated one, and the second-order federated update converges in only a few rounds. The resulting sparsity compresses the stored model, the per-sample inference cost, and the server-to-client broadcast by a comparable factor. These results demonstrate the practical value of sparse readouts for edge computing and IoT deployments.
[LG-40] ShiftQAT: Quantization-Aware Training for the Numerical Contract of a Fixed-Function Edge CNN Accelerator
链接: https://arxiv.org/abs/2610.21695
作者: Oguzhan Buyuksolak,Osman Erman Okman,Mehmet Gorkem Ulkar
类目: Machine Learning (cs.LG)
*备注:
Abstract:Fixed-function CNN accelerators make edge inference fast and energy-efficient, but they execute only networks that obey a precise numerical contract. On the MAX7800x family, this contract requires power-of-two, per-tensor, symmetric scales, 8-bit activations and biases, integer-only arithmetic, and one statically configured shift per layer. We present ShiftQAT, the quantization-aware training (QAT) flow we developed for this family, which produces networks satisfying the contract by construction. It is the default training path of ai8x-training, the family’s open-source training toolchain. ShiftQAT resolves residual, concatenation, and weight-sharing dependencies so that activation scales propagate statically and each layer’s requantization reduces to one offline-configured shift. It also trains weights and biases on a joint power-of-two grid that fits the 8-bit bias budget. Among the compared methods, only ShiftQAT satisfies all of the contract’s scaling constraints, including signed 8-bit activation storage. Even so, on ImageNet it matches the W8A8 accuracy drop of less constrained QAT methods and stays within 2.9 percentage points of FP32 at W4A8. On MAX78002 silicon, ShiftQAT-trained MobileNetV2 and ResNet-18 classify ImageNet images in 14.6 and 8.4 ms at 5.1 and 2.8 mJ per inference.
[LG-41] MARS: A Momentum-Aware Adaptive Rating System for Personalized Assessment
链接: https://arxiv.org/abs/2610.21685
作者: Vikrant Sahu,Khethavath Sunil Naik,Gagan Raj Gupta
类目: Computers and Society (cs.CY); Machine Learning (cs.LG)
*备注:
Abstract:Adaptive assessment systems personalize learning by estimating learner proficiency and recommending suitable questions. However, many existing approaches rely on fixed difficulty labels, static rating updates, or black-box prediction models that provide limited control over rating stability, interpretability, and time-constrained routing. This paper proposes MARS, a Momentum-Aware Adaptive Rating System for interpretable adaptive assessment. MARS represents the question bank as a directed weighted knowledge graph and updates learner proficiency using dynamic rating sensitivity, response quality, and momentum-aware learner velocity. We also formulate a time-budgeted mark-maximization objective in which the policy selects questions within a fixed test duration under negative marking. Experiments on ASSISTments 2009 and MARS-SimExam evaluate MARS for prediction, rating stability, and adaptive routing. Results show that MARS provides smoother rating updates than Fixed Elo and achieves the highest score under time-budgeted negative-marking routing, improving the total score by 8.09 over the strongest baseline and reducing overtime violations from 50% to 5.6%. These results suggest that MARS is suitable for real-time, interpretable, and deployment-friendly adaptive assessment.
[LG-42] Classical methods for modern applications: synthetic tabular data generation using lightweight discrete response continuous latent variable model
链接: https://arxiv.org/abs/2610.21674
作者: Bartosz Barzdajn,Mateusz Krzysztoń
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:In this paper, we present a lightweight algorithm for modelling tabular data containing both continuous and discrete variables. The method uses empirical marginal distributions and a selected quantile function to transform the data into a continuous latent space, where a distribution of choice, e.g. a multivariate Gaussian, can be efficiently estimated and applied. While this approach is closely related to classical well-established machine learning methods, we present a novel adaptation that fits well within the contemporary computational ecosystem. Unlike in the case of deep architectures, it results in a simple implementation with few dependencies, robust training/estimation, as well as efficient sampling. The transformation preserves complex marginal distributions and correlations, providing high-fidelity models, especially for data with unimodal distributions. One of the key characteristics of the proposed approach is its continuous representation of discrete variables. While highly desirable, this introduces a significant challenge, since a single observation corresponds to a region rather than a point in the latent space. We show that this ambiguity naturally leads to a Monte Carlo expectation-maximisation (MCEM) formulation, whose auxiliary objective can be expressed as a weighted latent-space likelihood. This, in turn, allows conventional maximum-likelihood estimators to be used within an efficient and robust optimisation procedure.
[LG-43] Approaching Backpropagation Accuracy with Weight Perturbation on MNIST
链接: https://arxiv.org/abs/2610.21665
作者: Patrick Inoue
类目: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
*备注:
Abstract:Weight perturbation (WP) estimates learning directions from random parameter perturbations and a scalar change in loss, without backpropagating derivatives. We examine how batch size, training duration, and estimator stabilization affect its attainable accuracy in fully connected MNIST networks. In the shallow configuration, large-batch training reaches approximately 97% test accuracy, compared with an early backpropagation (BP) reference near 97.5–98%. The same comparison exposes a substantial efficiency gap: under a ten-epoch budget, shallow WP reaches 90.2%, whereas the reported BP reference reaches 97.4%. Epoch-indexed curves show diminishing returns from increasing batch size beyond 4,000. Learning-rate reduction and averaging independent perturbation-based updates also stabilize deeper networks, although their recorded results do not establish BP-level accuracy. We distinguish these empirical findings from a finite-horizon approximation argument: under stated regularity conditions, population WP can approximate a fixed gradient-descent trajectory arbitrarily closely at any fixed finite depth. Positive classification margins then suffice to preserve the reference predictions on a finite test set with arbitrarily high probability. Together, the results identify near-BP accuracy as a configuration- and budget-dependent possibility, rather than evidence of equal efficiency or an unconditional accuracy guarantee across depth.
[LG-44] ReFlow: Replay-Driven Decoupling for Efficient RL Post-Training
链接: https://arxiv.org/abs/2610.21640
作者: Yukun Tian,Junwei Su,Haofeng Xu,Lansong Diao,Zhengping Qian,Weibo Zhao,Wenchao Zhou,Ruiting Zhou,Chuan Wu
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注:
Abstract:Reinforcement learning (RL) has been a critical paradigm for improving large language model (LLM) reasoning and agentic capabilities, but its efficiency remains a major bottleneck. Existing RL systems commonly rely on asynchronous execution to overlap rollout generation with policy optimization for enhanced efficiency. However, this approach faces inherent limitations: mild policy staleness offers only limited training efficiency gains, whereas aggressive asynchrony can degrade model quality and yields diminishing efficiency gains. The dynamic nature of LLM RL and multi-turn agentic workloads further challenges RL system design, as resource demands shift over time, making static resource allocation and generation strategy inefficient. This paper presents ReFlow, a system-algorithm co-designed RL framework for efficient LLM post-training. Its key insight is that, unlike existing RL frameworks which consume each trajectory only once, experience replay of recent trajectories enables effective sample reuse, thereby amortizing the high cost of rollout generation without incurring excessive policy staleness. To fully exploit this opportunity, ReFlow adopts a decoupled and adaptive architecture with three key designs: a dual replay data cycle supports efficient experience reuse while balancing training throughput and optimization stability; an elastic resource pool enables dynamic resource placement to accommodate evolving RL workloads; an adaptive scheduler coordinates generation strategies and resource allocation to improve resource utilization and boost generation throughput. Experiments across representative single-turn reasoning and multi-turn agentic RL tasks show that ReFlow improves training throughput by up to 2.06\times over state-of-the-art synchronous and asynchronous RL systems, while maintaining comparable or better model performance. Subjects: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC) Cite as: arXiv:2610.21640 [cs.LG] (or arXiv:2610.21640v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2610.21640 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-45] Local Temporal Contrasts on Human Motion Data Induce a Latent Geometry
链接: https://arxiv.org/abs/2610.21630
作者: Lourenço A. Rodrigues,Felix Putze
类目: Machine Learning (cs.LG)
*备注: 15 pages, 3 figures, under review
Abstract:Time series such as recordings of human motion data often encode structure across multiple temporal scales, and when a latent state determining this sequence evolves, this structure needs to survive in the representation space for downstream use. Yet the currently prevalent attention-based encoders paired with sequence-level pooling have made single-vector, pointwise comparison the default mode of inference, discarding the trajectory a sequence traces through the representation space. We propose recovering this geometric structure directly, without relying on forecasting-based objectives, by training an encoder model (SemTAC) with a temporal-neighborhood contrastive loss: representations that satisfy local consistency everywhere must give rise to a coherent global structure, letting inference happen at temporal scales beyond those seen during training. We evaluate SemTAC against four baselines on a five-step motion sequence discrimination task: the same architecture trained with a classification or a standard contrastive objective, and two established self-supervised time-series methods, TS-TCC and TS2Vec. At the full sequence length all five are broadly competitive, but on incomplete, truncated sequences SemTAC holds a large advantage: a median F1 of 0.77 one step before the sequence completes, against 0.27-0.70 for the four baselines. A hyperparameter-matched ablation attributes this effect specifically to the temporal-contrastive objective. SemTAC’s representation trajectories are also the only ones among these five to exhibit interpretable geometric structures, exceeding a no-structure chance baseline on a displacement-smoothness metric which we introduce. This is evidence that its objective, not classification accuracy alone, produces genuinely temporally-ordered trajectories.
[LG-46] Market Incentives for AI Safety Investment NEURIPS2026
链接: https://arxiv.org/abs/2610.21607
作者: Nikita Tsoy
类目: Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT); Machine Learning (stat.ML)
*备注: NeurIPS 2026
Abstract:Despite their practical significance, modern AI systems pose significant societal risks, including toxicity, malicious use, and misinformation. Their mitigation depends not only on technical feasibility but also on the incentives of firms that develop and deploy these systems. We show how profit maximization shapes safety investment in an AI supply chain with an upstream LLM provider and several competing downstream firms. In our game-theoretic model, we derive the market-driven level of safety investment and compare it with the levels that maximize market welfare and industry profit. We find that downstream competition generally induces downstream firms to invest more in safety than the industry-profit-optimal level. By contrast, the upstream provider may underinvest relative to these benchmarks because it is not fully compensated for the surplus from safety investments. Finally, we analyze how different regulatory interventions affect the participants’ incentives and profits, and the resulting levels of model safety.
[LG-47] ransfer Learning in Bandits from Unknown Adaptive Policies: Problem-Dependent Lower and Upper Bounds
链接: https://arxiv.org/abs/2610.21582
作者: Adrien Prevost,Timothee Mathieu,Waris Radji,Odalric-Ambrym Maillard
类目: Machine Learning (cs.LG)
*备注: 50 pages, 4 figures
Abstract:Transfer learning in stochastic multi-armed bandits is typically studied assuming that prior data forms a fixed offline dataset or is generated by a known source. In practice, historical observations come from a previous adaptive decision process whose policy is unknown. We study this setting for bounded rewards in [0,1] . Before the online game starts, the learner receives, for each arm k , N’_k observations drawn i.i.d.\ from an unknown distribution \nu’_k , collected by an arbitrary and possibly data-dependent strategy, so the sample sizes N’_k may be random. Source and target means satisfy |\mu_k-\mu’_k|L_k for known constants (L_k)_k . We derive an asymptotic problem-dependent lower bound and propose KL-UCB-Transfer, which combines online and historical data and achieves logarithmic regret with an improved leading constant whenever \mu’_a\mu_1-L_a , and recovers the classical KL-UCB guarantee otherwise. Since bounding the source-target discrepancy is often unrealistic, we relax the assumption to the one-sided condition \mu_k-\mu’_kL_k and show that logarithmic regret can survive under misspecification, characterizing three regimes ranging from improvement to a degraded logarithmic bound to a loss of guarantee. Simulations illustrate the practical benefits of exploiting process-generated historical data. Comments: 50 pages, 4 figures Subjects: Machine Learning (cs.LG) Cite as: arXiv:2610.21582 [cs.LG] (or arXiv:2610.21582v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2610.21582 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-48] From Perception to Prediction and Intervention: Benchmarking Proactive Video Assistants in Real-Life Scenarios
链接: https://arxiv.org/abs/2610.21568
作者: Boyu Cui,Junhao Luo,Zhihao Wang,Linghan Chen,Zejun Li,Jiwen Zhang,Siyuan Wang,Zhongyu Wei
类目: Machine Learning (cs.LG)
*备注: 24 pages,8 tables,13 figures
Abstract:Recent multimodal large language models have advanced video understanding, but existing benchmarks mainly evaluate whether models can interpret observed visual content. In real-world scenarios, assistants must additionally anticipate what assistance is needed and decide when to provide it. To address this gap, we introduce \textscP ^3 IBench, a benchmark for proactive video understanding in real-life streaming scenarios. It evaluates three complementary capabilities: daily perception, risk prediction and intervention. We benchmark 12 representative open-source and proprietary video-language models, including streaming and adapted non-streaming systems. Our evaluation shows that current models struggle to make well-calibrated response decisions. Finally, we construct additional proactive training data and train \textscP ^3 I-Assistant-4B, achieving improvements on several tasks, though gains remain uneven across capabilities. Together, these findings show that effective proactive assistance requires not only visual understanding, but also precise response timing and appropriate restraint.
[LG-49] Sharp and Stable Label-Free Calibration Certificates from Logical Coherence
链接: https://arxiv.org/abs/2610.21567
作者: Takahiro Kato(NP-hard, Inc.)
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Query-conditioned probabilistic predictors answer logically related questions one at a time, and labels to check their calibration are often missing. For each restriction on what a bettor’s stakes may use, a minimax theorem identifies, for finitely supported laws of the answers, the sharp label-free lower bound on the corresponding calibration residual. On bundles of different events its mean, marginal and joint stages can separate strictly, and a bundle’s bound cannot be attributed to one question without assumptions on the others. The sharp marginal bound is discontinuous in the answers, and no label-free lower bound continuous in the Wasserstein distance stays within a constant factor of it. Lipschitz stakes give a stable bound, sharp for a smooth calibration residual and, for every positive Lipschitz constant, positive exactly when the sharp marginal bound is. On i.i.d. streams, projected online gradient ascent at fixed nodes gives fixed-horizon certificates whose shortfall is bounded by grid, learning and statistical terms, and predictable betting gives anytime-valid ones. In a preregistered study of four predictors on two data sets, all seven confirmatory certificates were positive (the eighth pair did not pass the pilot). A certificate shows only that calibration is at least this bad; zero means “not refuted”.
[LG-50] A Quantum Machine Learning Approach to Content-Based Media Recommendation
链接: https://arxiv.org/abs/2610.21562
作者: Mina Abbaszadeh,Ouissal Moumou,Anna Pearson,Bob Coecke,Rob Cooper,Alessandro Piscopo,Mehrnoosh Sadrzadeh
类目: Machine Learning (cs.LG)
*备注:
Abstract:We use quantum machine learning techniques to develop a content-based media recommender system and evaluate its performance on TV programmes using textual descriptions from one week of BBC iPlayer data. Grounded in the Quantum Natural Language Processing framework (QNLP), programme synopses are mapped to variational quantum circuits using their semantic structure. The parameters of these circuits are trained via a supervised pairwise similarity learning task. We benchmark the quantum recommender against a classical BiLSTM baseline trained on the same dataset and objective. While both models achieve comparable performance on the similarity learning task, the quantum model requires significantly fewer parameters (47k vs. 17.8M). When the BiLSTM is constrained to a matching parameter budget, its performance degrades substantially (by 3.63 percentage points in Mean Average Precision (MAP)), indicating that the structured quantum representation provides a highly parameter-efficient alternative in this setting. We demonstrate the feasibility of executing the trained models on quantum hardware by deploying the circuits on Quantinuum’s System Model H2 (and accompanying emulator) using a SWAP-test-based protocol. The results confirm that the similarity signal is preserved under realistic, noisy conditions. Taken together, these findings highlight the potential of variational quantum circuits for recommendation tasks and represent a practical step toward their real-world deployment. Subjects: Machine Learning (cs.LG) Cite as: arXiv:2610.21562 [cs.LG] (or arXiv:2610.21562v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2610.21562 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-51] Finite-Time Guarantees for Constrained Reinforcement Learning under Partial Observability
链接: https://arxiv.org/abs/2610.21548
作者: Valentina Zaccaria,Philip Jordan,Tingting Ni,Gian Antonio Susto,Maryam Kamgarpour
类目: Machine Learning (cs.LG)
*备注:
Abstract:We study reinforcement learning in constrained partially observable Markov decision processes (CPOMDPs), where an agent maximizes reward subject to a cost constraint using only incomplete information about the underlying state. The lack of state observability may require optimal policies to depend on arbitrarily long observation-action histories, making planning and learning challenging. We leverage recent advances in finite-memory approximations of POMDPs and represent finite history windows as states of a constrained MDP. We then develop a primal–dual algorithm that provides the first non-asymptotic guarantees that jointly bound average reward suboptimality and constraint violation in the original CPOMDP. A key challenge is that this finite-window approximation affects both reward values and the set of feasible policies. Under a filter-stability assumption, we derive bounds relating optimality and feasibility in the finite-state model to those in the original CPOMDP, with approximation errors that decay geometrically with the window length. Experiments in partially observable gridworlds demonstrate the effectiveness of our approach.
[LG-52] Federated Art Chase: Privacy Leakage and Data Poisoning in Federated Fine-Tuning
链接: https://arxiv.org/abs/2610.21547
作者: Vasileios Perifanis,Nikolaos Pavlidis
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注: Technical report accompanying the hands-on tutorial at the 4th Federated Machine Learning International Summer School (FLTA), Paris, France, October 2026
Abstract:Federated learning (FL) is increasingly used to adapt a pretrained, centrally fine-tuned model to data that remain on user devices, beyond training a model from scratch. In this report, prepared for the 4th Federated Machine Learning International Summer School, we study the privacy and integrity of this approach with a reproducible pipeline. An ImageNet-pretrained MobileNetV4-Conv-Small model is fine-tuned centrally on ten Quick, Draw! classes (97.50% test accuracy) and then adapted with head-only FedAvg over ten non-IID clients, which raises client training accuracy from 90.34% to 91.49% while preserving global accuracy. We evaluate four adversarial scenarios. First, from a single one-step client update, an honest-but-curious server recovers the complete private label set, but its reconstructions follow class prototypes rather than the private drawings (mean SSIM 0.130). Second, from a one-round FedAvg aggregate, label-count recovery falls from 87.5% with one client to 75.0% with five and ten clients, below the 80.0% of an uninformed uniform guess. Third, under FedAvg, a backdoor injected by 30%, 50%, and 100% of the clients reaches an attack success rate (ASR) of 21.9%, 31.3%, and 54.7% at a cost of at most 0.8 percentage points of clean accuracy, whereas label flipping is effective only when most of the source-class data are relabeled. Fourth, FLTrust with a root set retained from centralized fine-tuning lowers the backdoor ASR to 1.8% and 4.0% at 30% and 50% malicious clients and the label-flipping ASR from 12.6% to 0.8% without measurable clean-accuracy loss, but provides no protection when every client is malicious. The report closes with a live session in which participants draw, fine-tune, and poison the model through Indigma’s FedMaestro orchestrator. Code and executed notebooks, runnable on Google Colab, are at this https URL.
[LG-53] Activation-Adjacent Layer Normalization: Effects on Approximation Power
链接: https://arxiv.org/abs/2610.21527
作者: Yunhao Ni,Yuxin Guo,Jie Luo,Wenjun Wu,Lei Huang
类目: Machine Learning (cs.LG)
*备注: 21 pages, 7 figures
Abstract:Layer normalization (LN) and root mean square normalization (RMSNorm) depend on their inputs and remain nonlinear at inference. Inserting either next to an activation may therefore change the function a network represents. We study when shallow activation-normalized (AN) and preactivation-normalized (PN) networks can approximate a prescribed width- N one-hidden-layer MLP. Viewing RMSNorm as input-dependent scaling, we use width augmentation to locally linearize normalization on compact feature sets. For AN networks, an unbounded activation range allows width N+1 , while every continuous nonconstant activation admits width N+r with error O_f(r^-1) . For PN networks, affine parameters or, without them, degree-one positive homogeneity recover width N+1 ; general continuous activations admit width N+r with error tending to zero. The corresponding LN constructions use centered coordinate pairs, yielding widths N+2 and N+2r , respectively. We prove that N+1 for RMSNorm and N+2 for LN are also the worst-case bound for approximations, and show that applicable universal-approximation theorems for ordinary MLPs transfer to these normalized architectures. Numerical experiments recover the predicted approximation rates for both normalizations across all five settings and illustrate the shifted-ReLU finite-width obstruction.
[LG-54] When the KV Cache Is Slower Than the GPU: Scaling Remote KV Tiers for Context-Parallel Prefill
链接: https://arxiv.org/abs/2610.21500
作者: Muhammad Ibrahim
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: 9 pages, 8 figures, 3 tables
Abstract:Reusing the attention KV cache across requests is a standard way to cut prefill cost in long-context LLM serving, and production systems increasingly back the per-engine cache with a remote tier shared by every engine. We study such a tier in a disaggregated deployment of GLM-5.3, an FP8 mixture-of-experts model with multi-head latent attention (MLA), on 8xB300 nodes running prefill context parallelism (PCP8). Enabling the remote tier (LMCache with a Valkey store) left median time-to-first-token unchanged but raised the 16k-token p99 from 3.96 s to 17.3 s. Replaying a production request stream on an isolated engine, we find that a remote hit took 0.88 s against 0.31 s to recompute the prompt, and that synchronous loading turned a hit-heavy stream into an unstable queue: p99 83.32 s with the GPUs 7% busy. A two-term cost model explains both: a hit pays off only if the tier delivers more than B* = sR (2.53 GB/s for this model and GPU), and a fixed per-hit engine cost (T0 ~ 0.33 s) means that hits on prompts shorter than about 17k tokens cannot beat recompute at any bandwidth. Guided by the model, we (i) shard the MLA KV store across PCP ranks, with a proof that the existing minimum-over-ranks lookup returns the longest cached prefix and a load protocol that can never desynchronize collectives; (ii) partition the remote tier by rank, one server instance per rank with no coordination, which raises GET bandwidth 6.3x to 10.76 GB/s; (iii) add asynchronous loading to the sharded store; and (iv) fix three client bugs that caused unbounded stalls. On the hit-heavy stream, with 98.5% of prompt tokens served from the remote tier, p99 falls from 83.32 s to 0.67 s. The sharded store raises prefill throughput by 14%, and by 34% together with an upstream kernel fix. All changes are submitted upstream to LMCache.
[LG-55] Phase Kernel Lifts Capacity of Dense Associative Memory
链接: https://arxiv.org/abs/2610.21493
作者: Yifei Zhao,Ying Tang
类目: Machine Learning (cs.LG)
*备注: Accepted as NeurlPS spotlight
Abstract:Dense associative memory underlies modern attention mechanisms; however, Modern Hopfield Networks (MHNs) with linear kernels yield overlapping attractor basins and spurious states under high memory load, limiting their theoretically exponential storage capacity in practice. We construct a class of nonlinear phase kernels that modify the energy landscape and retrieval dynamics, and prove that these kernels preserve exponential storage capacity and local fixed-point convergence under given conditions on kernel smoothness, memory separation, and memory distributions. The framework is instantiated by the Sigmoidal Phase Hopfield Network (S-Hop), which guarantees monotone energy descent and mitigates practical capacity loss. S-Hop has 8.3\times and 4\times gains in one-step noisy retrieval load over Uniform Hopfield Networks (U-Hop) on CIFAR-10 and TinyImageNet respectively, and reaches a 50\times gain in first-failure load for ImageNet. The results demonstrate that phase kernels provide a principled route to robust dense associative memory.
[LG-56] Kolmogorov-Arnold Networks: Curse of Dimensionality and Structural Bias
链接: https://arxiv.org/abs/2610.21480
作者: Yunhao Ni,Yuxin Guo,Jie Luo,Wenjun Wu,Lei Huang
类目: Machine Learning (cs.LG)
*备注: 25 pages
Abstract:Kolmogorov-Arnold networks (KANs) have a dimension-independent approximation exponent when the target admits a sufficiently smooth representation by univariate edge functions. We ask whether this favorable scaling belongs to the architecture or to the assumed target class. For arbitrary-depth standard spline-KANs, we prove a worst-case lower bound on isotropic Sobolev balls: below a dimension-independent accuracy threshold, the required number of spline coefficients grows at least exponentially with input dimension. Existing composition-matched MLP results, by contrast, show that favorable exponents on low-arity targets are not exclusive to KANs. We also revisit empirical comparisons. On four scaling-law targets, spline-KAN approximants constructed from known decompositions transfer exactly to MLPs with the cubic ReLU activation \sigma(x)=(\max\0,x)^3 , although their full dense parameter counts are larger. In controlled ridge-profile scans, KANs have a strong advantage when each nonlinear component depends on one input coordinate. When a component couples multiple input coordinates, this advantage may narrow sharply. At tight budgets, interacting targets often favor MLPs. In the random-label test on MNIST, ReLU-MLPs fit the randomized training labels more efficiently than spline-only KANs. Together, these results locate favorable scaling in target structure and a compatible inductive bias, not a general KAN-specific escape from the curse of dimensionality.
[LG-57] DuET: Note-Frame Dual Encoding for Piano Performance-to-Score Transcription
链接: https://arxiv.org/abs/2610.21467
作者: Wei Zeng,Ye Wang
类目: ound (cs.SD); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
*备注: 14 pages, 3 figures
Abstract:Transcribing expressive piano performances into scores requires interpreting individual note timings within an underlying metrical structure. However, score attributes are associated with notes, whereas beats may occur without coincident note onsets. We introduce DuET, a Dual-Encoder Transformer that jointly learns note-level performance MIDI-to-score (PM2S) transcription and frame-level beat and downbeat tracking. Bidirectional cross-attention connects note and frame encoders, while an autoregressive decoder alternates between their representations to predict structured score attributes. We further introduce TimeRoPE and Time-Pitch Rotary Positional Encoding (TPRoPE) to establish a shared performance-time reference and explicitly encode relative temporal and pitch relationships. Evaluations on ASAP, a classical piano dataset, and our curated Gangqinpu dataset, which spans multiple genres with a focus on popular piano music, demonstrate improvements in score timing and notated duration. Compared with the strongest baseline on each dataset, DuET reduces MUSTER mean error from 6.56% to 3.64% on Gangqinpu and from 11.30% to 11.22% on ASAP. Ablations support the benefits of frame-level context, time-based positional encoding, and decoder-history masking. Code, weights, and demos are available at this https URL.
[LG-58] Each Reward to Its Relevant Tokens: Reward-Specific Token-Level Credit Assignment for Multi-Reward RL
链接: https://arxiv.org/abs/2610.21439
作者: Skylar Zhai,Haonan Wang,Zhining Liu,Linxin Song,Xuwei Ding,Jingcheng Liang,Jian Kang
类目: Machine Learning (cs.LG)
*备注: 15 pages, 8 figures. Project page: this https URL
Abstract:Language models are increasingly trained against several verifiable rewards at once, such as the correctness and the format of a tool call, or the individual constraints of a complex instruction. Reward-decoupled normalization (GDPO) normalizes each reward separately so that different reward combinations keep distinct advantages, but it still spreads every reward’s advantage evenly over the whole response, although each verifier checks only a few positions. We propose MRTPO, which assigns the advantage of each reward to the tokens relevant to it, using the information its verifier checks, without auxiliary models or token-level annotations. For each reward, the policy replays its own response with and without that reward’s verifier information, and the per-token Jensen-Shannon distance between the two next-token distributions measures each token’s relevance and reassigns the advantage. On text-to-SQL, multi-constraint instruction following and tool calling, MRTPO converges faster and achieves better performance than the compared methods.
[LG-59] Policy-Aligned Predictive Verification for Adaptive VLA Action-Chunk Execution
链接: https://arxiv.org/abs/2610.21434
作者: Minkyu Kim,Junhoo Lee,Seungyeon Kim,Suhyun Jeon,Baekseung Kim,Donghoon Han,Nojun Kwak
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注:
Abstract:Vision-language-action (VLA) policies generate action sequences that reduce repeated inference, but deciding how much of each sequence to execute is critical to task success. Existing adaptive methods make this decision using generation stability or prediction agreement, yet these signals do not establish whether executing a candidate prefix preserves the ability to complete the task. We introduce a policy-aligned predictive verifier (PAV) that anticipates the consequences of candidate action prefixes and evaluates their compatibility with successful continuation before execution. Rather than using future prediction to generate actions, our method evaluates existing proposals in the VLA’s internal task-conditioned representation space. To learn local judgments from terminal task outcomes, we combine trajectory-level survival supervision with a localization prior that preserves useful behavior within failed rollouts. The resulting validity scores guide execution-horizon selection. Evaluations with pi_0 and pi_0.5 across LIBERO, RoboTwin 2.0, and real-world manipulation show improved success-query trade-offs. These findings point toward a different role for future prediction in robotics, using it to assess how far to trust a policy’s proposed actions rather than to generate new ones.
[LG-60] What Does History Add? Calendar Controls and Information-Timing Audits for University Trajectory Forecasting
链接: https://arxiv.org/abs/2610.21419
作者: Zepei Cao
类目: Machine Learning (cs.LG)
*备注: 12 pages, 3 figures, 6 tables. Methodological simulator evaluation with information-timing and calendar-control audits
Abstract:We audit the predictive value of historical summaries in a rule-based university trajectory simulator. Seventeen ridge specifications and persistence separate current plans, past-state and past-plan summaries, and calendar-only history-length indicators. Across training, validation and two family-held-out evaluations, we reconstruct 16,750 routes and 804,000 monthly records. On the original 40-family set, six-month history lowers six-step normalized RMSE from 0.159234 to 0.151522, but its difference from a length-only control is -0.000895 (descriptive paired 95% interval [-0.001957,0.000248]). Fixed weights on 30 families unused in identified prior tasks reproduce this pattern: 0.156480 to 0.150074 versus current-state prediction, but -0.000558 [-0.001783,0.000728] versus the length control. These uncertain comparisons do not generalize to all-history summaries, which improve on their own calendar control. Exploratory decomposition finds clearer preparation-index and funds benefits, whereas fatigue dominates the six-month model’s loss and has an uncertain length-controlled benefit. Moreover, 28.20% of its six-step fatigue predictions violate the simulator’s [5,90] support. We document family dependence, prior exposure, strict temporal indexing, matched recursive references and independent numerical replay. The task conditions on recorded future plans; it does not evaluate real-student forecasts, causal action effects, autonomous planning or language-model memory. This is an evaluation audit, not a new memory architecture.
[LG-61] Learning to Forecast on Multiple Spatiotemporal Hierarchies
链接: https://arxiv.org/abs/2610.21414
作者: Valentina Moretti,Cesare Alippi,Andrea Cini
类目: Machine Learning (cs.LG)
*备注:
Abstract:Hierarchical forecasting exploits aggregation structures that expose multiple views of the same underlying spatiotemporal forecasting problem, while ensuring coherence among forecasts. Individual series provide fine-grained information, while progressively coarser aggregates reveal temporal and spatial patterns at different scales. In the associated deep learning architectures, the hierarchy determines which aggregate series are available and how they are processed together with individual series (e.g., through reconciliation or message-passing operators). Different hierarchies thus provide distinct learning signals and inductive biases for forecasting the same collection of time series. Nonetheless, existing architectures are typically trained on a single, often fixed, structure. In this work, we propose a graph deep learning framework that leverages variations across hierarchies as multi-task regularization. We show that jointly learning across these structures can reduce over-specialization to any single hierarchy and improve forecasting performance. Indeed, training across hierarchies provides an inductive bias toward representations that are useful across different structures, and as such, generalize better. Making forecasts coherent with the aggregation constraints for each sampled hierarchy is computationally expensive: to complement the framework, we propose a novel operator-based on a top-down approach-that produces coherent forecasts with linear time and space complexity w.r.t. the hierarchy size. Experiments show that our approach improves forecasting accuracy over training on a single structure and is competitive with current state-of-the-art methods.
[LG-62] Fault Detection and Bit-Error Rate Diagnostics in Optical Wireless Links Using Edge-AI
链接: https://arxiv.org/abs/2610.21402
作者: Sarthak Bhattarai,Sheshadri Shikhar Bandyopadhyay,Suprabh Dhakal,Jitendra Kumar Manandhar
类目: Machine Learning (cs.LG)
*备注:
Abstract:Optical Wireless Communication (OWC) systems provide high-bandwidth, spatially secure, and RF-interference-free digital transmission. However, direct free-space optical links suffer from extreme sensitivity to line-of-sight disruptions, optical attenuation, high ambient light saturation, and transmitter-receiver misalignment. Traditional channel diagnostic frameworks rely on external computing nodes or cloud-based processing, which introduces severe diagnostic latency and reduces edge node autonomy. This paper presents an autonomous, self-diagnosing optical wireless receiver node that integrates physical-layer signal telemetry extraction with an on-device TinyML classification engine. Built around an ESP32-S3 microcontroller, the receiver recovers 1000 bit/s On-Off Keying (OOK) digital payloads using a 5-sample majority voting mechanism while concurrently extracting real-time channel features: Bit Error Rate (BER), received signal voltage (Vsignal), Signal-to-Noise Ratio (SNRdB), and ambient light baseline (Lambient). Evaluated against a Random Forest baseline (77.22% accuracy) and a floating-point neural network (78.48% accuracy), a post-training INT8 quantized TensorFlow Lite Micro model occupying only 3.32 kB of SRAM runs directly on the receiver firmware. The quantized model achieves 74.68% overall accuracy across four distinct operational conditions: normal transmission, optical filter attenuation, ambient light interference, and severe obstruction/misalignment. On-device deployment tests confirm real-time, deterministic fault classification, establishing a practical approach for self-aware optical IoT nodes.
[LG-63] RDL-Bench: A Benchmark for Task-Aware Question Answering over Relational Databases ICDE2027
链接: https://arxiv.org/abs/2610.21382
作者: Yao Cheng,Siqiang Luo
类目: Machine Learning (cs.LG)
*备注: This paper has been accepted by ICDE 2027
Abstract:Relational databases (RDBs) remain the primary infrastructure for storing and managing structured data across enterprise and scientific domains. Natural-language (NL) interfaces to RDBs aim to let users ask questions without writing SQL, but real user needs go beyond SQL-style lookup: some questions require reasoning over the observable database state, while others require predictive inference from historical relational records under temporal constraints. Existing benchmarks address these goals in isolation. To address this gap, we introduce RDL-Bench, a benchmark for task-aware question answering (QA) over real RDBs. RDL-Bench organizes diverse user questions into reasoning-oriented questions, whose answers are determined by the current database state, and prediction-oriented questions, whose answers depend on future outcomes under temporal constraints. Each instance contains an NL question, task metadata, schema information, temporal context, either executable evidence or target-construction logic, and denotation-level ground truth. RDL-Bench supports two evaluation settings: native solvers under their intended capabilities, and task-aware modular systems where a router dispatches each question to an appropriate execution module.
[LG-64] Does Compression Imply Generalization? A Minimum Description Length Perspective NEURIPS2026
链接: https://arxiv.org/abs/2610.21376
作者: Lincen Yang,Jiayang Shi,Nan Pu,Daniël M. Pelt,Matthijs van Leeuwen,Zhong Li,Shujian Yu
类目: Machine Learning (cs.LG)
*备注: Published at NeurIPS 2026
Abstract:A central question in modern learning theory is what notion of compression, if any, explains why over-parameterized neural networks generalize. Existing answers measure complexity at different levels, ranging from weights to representations; however, estimating the associated quantities either requires held-out data in principle or relies on the challenging (ultra) high-dimensional mutual information estimation. To overcome these challenges, we approach the question through the minimum description length (MDL) principle, and specifically, we propose the prequential regret as a complementary signal that is computable from the training trajectory alone. We establish that, in the (nearly) interpolating regime, the expected prequential regret upper-bounds the description length of the trained parameter under the algorithm-induced distribution, and that this description length translates into a high-probability generalization bound in the PAC-MDL setting. Along the way, we show that the encoder-complexity correction recently added to the algorithm-level information bottleneck (IB) is upper-bounded by an MDL parameter description length, \newbringing together the two views of compression from MDL and IB. Empirically, the prequential regret tracks the generalization error on CIFAR-10 and CIFAR-100, with Pearson correlations exceeding 0.9.
[LG-65] Same Target Different Origins: Aligning Latent Rollout for Consistent PDE Emulation
链接: https://arxiv.org/abs/2610.21370
作者: Xinying Li,Jing Xu,Wenjun Wang,Limei Han,Fenglei Cao,Ruoxi Jiang,Yuan Qi
类目: Machine Learning (cs.LG)
*备注:
Abstract:Neural PDE solvers commonly generate multi-step predictions through autoregressive rollout, recursively applying a learned evolution operator and feeding each prediction back as input for the next step. Approximation errors can therefore propagate through the rollout, causing predictions at a shared target time to depend on their temporal origins and rollout depths. For a well-posed deterministic PDE flow, however, states at different times along the same physical trajectory should, under exact evolution, reach the same target state. To this end, we propose Align Latent Rollout (ALR), a regularization approach that promotes consistency across rollout origins through the neural PDE solver’s internal representations. A second-order Taylor analysis of latent discrepancy propagation motivates two objectives: zeroth-order latent state consistency, which aligns representations at a shared target time, and first-order latent dynamics consistency, which aligns local evolution responses to shared perturbations. ALR can be integrated into different neural PDE solvers by regularizing suitable internal representations, without modifying their backbone architectures. Experiments across multiple PDE benchmarks and neural-operator backbones show that ALR improves prediction accuracy while reducing shared-target discrepancies across rollout origins.
[LG-66] Honesty is the Best Policy … or is it? Demand Management for Probabilistic Service Options
链接: https://arxiv.org/abs/2610.21351
作者: Daniela Sailer,Robert Klein,Claudius Steinhardt
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:To enable well-informed customer choices and redistribute demand across service options, providers can display probabilities regarding predicted service levels at the time of booking, effectively offering probabilistic service options. In many service settings, the underlying truthful probabilities endogenously depend on the system state, rather than being static product attributes controlled by the provider. While customers rely on this information when making decisions, they cannot readily verify its accuracy. Consequently, providers may be tempted to manipulate the displayed probabilities to guide customers’ choices in their favor, creating a novel lever for demand management. In this work, we model the resulting problem as a sequential decision process, define truthful probabilities recursively, and introduce transparency constraints that limit their distortion. Because the exact computation of truthful probabilities is intractable for realistic settings, we develop an adapted reinforcement learning algorithm to approximate them. Furthermore, we propose a sampling-based mixed-integer program to optimize the information communicated by the provider in each state. Our framework enables providers to gauge the benefits of dynamically disclosing probabilistic information at the time of booking and sheds light on incentives for manipulation. Our computational study is motivated by out-of-home delivery, i.e., the delivery to parcel lockers or shops, and reveals a trade-off between individual-level transparency and aggregate customer utilities.
[LG-67] Online Quantile Omniprediction for Proper Losses
链接: https://arxiv.org/abs/2610.21342
作者: Hyunsuk Kim,Isaac Gibbs,Michael W. Mahoney,Ryan J. Tibshirani
类目: Machine Learning (cs.LG); Methodology (stat.ME)
*备注:
Abstract:The task of constructing a predictor that is simultaneously accurate under multiple evaluation metrics has been popularized recently under the name omniprediction. Prior work in this area has largely focused on binary prediction tasks, and existing results for multiclass and continuous outcomes typically suffer from slow error rates. In this paper, we consider forecasts expressed as a set of quantiles over a discrete but otherwise arbitrary set of probability levels. We develop sample-efficient algorithms for constructing forecasts that are accurate simultaneously across all proper losses that satisfy mild regularity conditions, where proper losses are those minimized by predicting the true quantiles. Our algorithms are online and can operate in a dynamic environment, while simultaneously taking multiple losses into account, and their forecasts offer the potential to provide diverse end-users with rich information for evaluating risk and informing a range of practical decision-making tasks. As an example, we demonstrate how our procedure can be used to ensemble quantile forecasts of COVID-19 hospitalizations made by experts during the pandemic.
[LG-68] Spectrum Is Not Enough: A Random-Matrix Account of Data-Free Preconditioning for Differentially Private Learning
链接: https://arxiv.org/abs/2610.21312
作者: Shilin Zhang,Kangning Wang
类目: Machine Learning (cs.LG)
*备注: 13 pages, 3 figures. Long form of the manuscript submitted to IEEE TNNLS
Abstract:Data-free preconditioning builds the Kronecker factors of KFAC from synthetic probes instead of private data. DP-KFC validates a probe by matching the eigenvalue decay of the private factors; we turn this criterion into a random-matrix account of what a probe can reconstruct. A single dichotomy organizes it: in the factorization A \otimes G , the error-signal factor G and the architecture-determined bulk of the activation factor A are reconstructible without data, while only the data-induced spikes of A are not. We prove the G half is data-insensitive up to an explicit error-energy bound, with zero violations over a full training trajectory across four modalities and a natural-image benchmark. We locate the failure boundary at a Baik-Ben Arous-Péché threshold: a bulk-only probe suffers an exact damage s^2/(1-s^2) from a data-induced spike, which becomes invisible once it exceeds (d/n)^1/4 , confirmed on an injected-spike surrogate. Holding the probe spectrum fixed and varying only its eigenbasis, we show that spectral-shape matching is neither sufficient nor the operative object: a probe with a near-exact matched spectrum (log-ratio dispersion 0.02 ) still inflates the residual 2 \times - 90 \times when its eigenvectors are random. A closed-form whitened-volume law explains this and predicts the within-spectrum ordering of the critical volume, though not its cross-dataset magnitude. Downstream, per-layer tuning does not transfer to accuracy and the headroom is a small oracle-minus-white gap. Under a privacy mechanism the source ordering survives while learning does, then inverts at a noise cross-over where the flat probe overtakes the matched ones.
[LG-69] Personalized Projected Ascent for Federated Stackelberg Optimization
链接: https://arxiv.org/abs/2610.21295
作者: M Yashwanth,Sharannya Ghosh,Saikiran Bulusu,Anirban Chakraborty
类目: Machine Learning (cs.LG)
*备注:
Abstract:Federated minimax problems, such as distributionally robust optimization and adversarial training, include an inner maximization variable built from each client’s local data. These variables differ in meaning and dimension from one client to the next, so averaging them across clients, as standard federated minimax methods do, is ill-posed. We therefore study a federated minimax problem where the minimization variable \mathbfx is shared and aggregated, while each client retains the maximization variable \mathbfy_i (personalized). Because these applications constrain \mathbfy_i to a compact convex set, we analyze projected ascent directly. We show that, in the nonconvex-concave regime, this projected personalized scheme attains the standard sublinear rate. In the nonconvex-PL regime, however, it hits a stepsize-invariant noise floor due to stochastic \mathbfy_i updates. To mitigate this, we propose VR-SGDA, incorporating a variance-reduced ascent step where each client maintains a private variable snapshot. By evaluating current and snapshot gradients on the same sample, the constant variance is replaced by a vanishing drift under a two-timescale regime. VR-SGDA eliminates the noise floor, attaining sublinear leading rates while preserving personalization and projection. Experiments on subpopulation-robustness and NLP adversarial training confirm both the worst-group robustness benefits and the two-timescale tracking mechanism.
[LG-70] Can Capability Enhancement be Localized?
链接: https://arxiv.org/abs/2610.21294
作者: Anh Tuan Nguyen,Anna Bair,Zico Kolter
类目: Machine Learning (cs.LG)
*备注: 25 pages
Abstract:Transformer-based language models have shown strong capabilities across many tasks and domains. However, prior work shows that, for a given task, not all attention heads matter equally. Specifically, they suggest that many \textitredundant heads can be safely removed with little-to-no effect, while a small set of \textitimportant heads, when ablated, severely damages the model’s performance on the given task. In this work, we investigate a complementary aspect, called \textitCapability Enhancement Localization (CEL): \textitgiven a model and a task, does there exist a small set of heads whose removal improves the task performance? We first investigate this question \textitempirically across models from different families. Using a simple two-stage routine that screens candidate heads and evaluates their joint ablations, we identify sparse head combinations whose removal improves task-specific performance without further fine-tuning. Motivated by these observations, we ask \textittheoretically whether such a phenomenon can naturally arise from training dynamics, and under what conditions. To this end, we construct a simplified multi-head softmax attention model and two tasks with conflicting features, and show that gradient descent from small random initialization provably concentrates the conflicting features into particular heads. We further show that ablating such heads improves performance on one task while harming the other, even though training converges to a global optimum of the joint objective. This establishes a possible training-induced theoretical explanation for CEL.
[LG-71] Regret Lower Bound for Online Learning of Quantum States with Logarithmic Loss with Applications to Quantum State Tomography and Confidence Sequences for Matrix Means
链接: https://arxiv.org/abs/2610.21273
作者: Yen-Huan Li,Wei-Fu Tseng
类目: Machine Learning (cs.LG); Statistics Theory (math.ST); Quantum Physics (quant-ph); Machine Learning (stat.ML)
*备注:
Abstract:We study the problem of online learning of quantum states with logarithmic loss, which is a quantum generalization of online portfolio selection and has applications to quantum state tomography, the task of estimating an unknown quantum state from measurement outcomes. We first establish a regret lower bound, showing that the regret rates of Exponential Weights and the algorithm of Tseng et al. (2025) have optimal polynomial dependence on the dimension. Our analysis leads to two further results. First, our lower-bound proof leads to a quantum state tomography algorithm that is sample optimal up to polylogarithmic factors in the sample size. Second, we develop an algorithm for constructing confidence sequences for the mean of a matrix-valued random variable. A numerical experiment shows that the latter algorithm is competitive with existing methods.
[LG-72] Rethinking Learning from Label Proportions via Moment Matching NEURIPS2026
链接: https://arxiv.org/abs/2610.21262
作者: Tianhao Ma,Wei Wang,Yivan Zhang,Dong-Dong Wu,Takashi Ishida,Gang Niu,Masashi Sugiyama
类目: Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2026
Abstract:Learning from label proportions (LLP) is a weakly supervised setting where training data are grouped into bags with only bag-level label proportions observed, and the goal is to learn a classifier that predicts labels for individual instances. Recently, a large body of methods has emerged; however, most of them rely on the assumption that instance-label pairs are sampled i.i.d. and randomly assembled into bags, an assumption often violated in real-world scenarios. Our experimental results indicate that these methods perform suboptimally under non-random settings. In this work, we adopt a more realistic assumption: bags are sampled i.i.d., and instance labels are conditionally independent given the instances. Under this setting, we show that the label counts follow a Poisson multinomial distribution. Motivated by this observation, we propose LLP via moment matching (LLP-MM), a simple yet effective approach which leverages multi-order factorial moments as training objectives, encouraging the classifier predictions to match these moments and thereby more fully exploit the underlying statistical structure of the data. Extensive experiments on benchmark datasets under various bag construction strategies demonstrate the effectiveness of our approach while maintaining high computational efficiency. Our code is available at this https URL.
[LG-73] Compatibility Matters: A Framework for Token-Level Teacher Selection in On-Policy Distillation
链接: https://arxiv.org/abs/2610.21234
作者: Stanley Wei,Jiin Woo,Kaan Ozkara,Youngsuk Park
类目: Machine Learning (cs.LG)
*备注:
Abstract:On-policy distillation (OPD) has recently been introduced as a promising paradigm for continual learning, where a student policy can be supervised at the token-level. Typically, OPD is performed using a single teacher per prompt rollout (or even per training run). However, this ignores a whole other axis of flexibility that is available due to the on-policy nature - token-level teacher selection during rollouts. Building on this insight, we theoretically explore what makes a teacher and a student compatible for OPD, and leverage our findings to derive several empirical heuristics for teacher selection. In particular, our framework captures several desiderata for OPD: compatible teacher-student reasoning patterns, teacher competence, and induced student diversity. Theoretically, we categorize several types of teacher routing schemes, and we prove that under different scenarios, different routing schemes are likely to be beneficial for optimization and student diversity. We back up our findings with extensive experimental results in domains such as math, coding, kernel generation, and visual reasoning, where we find upwards of 10 percentage points of improvement on pass@ k when compared to vanilla OPD and the teacher experts on certain tasks, indicating both capability improvement and diversity over high-quality responses.
[LG-74] Not All Change Is Damage: Update-Matched Unlearning for Large Language Models
链接: https://arxiv.org/abs/2610.21230
作者: Jie Li,Youming Tao,Bowei He,Meng Ding
类目: Machine Learning (cs.LG)
*备注:
Abstract:Unlearning aims to remove selected knowledge from large language models to mitigate privacy and safety risks while preserving useful capabilities. For models that continue to be fine-tuned after release, this goal is insufficient: they must also resist relearning the deleted knowledge while preserving plasticity, the ability to learn new tasks. Existing defenses train against simulated relearning while trying to preserve useful capabilities, but low relearning alone cannot distinguish selective resistance from a general loss of plasticity. Evaluating preservation after an update requires deciding what the updated model should remain close to. A pre-update reference penalizes useful changes induced by the update simply because they depart from earlier outputs. We therefore formulate the reference-timing hypothesis that, at comparable target suppression, an update-matched reference better preserves plasticity than a pre-update one. To test and address this hypothesis, we propose Update-Matched Unlearning (UMU), which applies the same simulated relearning update to the defended model and to an undefended reference initialized from the same unlearned model, and uses new-task learning to guide model selection. Extensive experiments on TOFU and FaithUn with Llama and Qwen models show that UMU improves both relearning resistance and plasticity across all three base unlearning methods, lowering harmful and benign relearning by 30.4 and 9.9 points on average while raising new-task accuracy by 3.5 points. When the reference does not receive the update, relearning remains similarly low but new-task accuracy falls below the base method, supporting the reference-timing hypothesis and showing that robust unlearning should prevent recovery selectively without suppressing learning itself.
[LG-75] Learning and Forecasting Multiscale Spatiotemporal Epidemic Dynamics with Lagrangian-Eulerian Graph Neural ODEs
链接: https://arxiv.org/abs/2610.21223
作者: Jinyi Duan,Qian Li,Biao Tang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Accurate medium- to long-range epidemic forecasting requires extrapolating high-dimensional, heterogeneous and nonstationary dynamics from partial observations. Yet hybrid models often leave implicit which observations inform which epidemic processes and how learned graph interactions modify system state. We introduce LE-EpiGNODE, which replaces generic message passing with process-specific, mechanism-guided neural operators: age-contact information drives Lagrangian infection generation, whereas mobility drives conservative Eulerian population transport within a structure-preserving universal differential equation. We evaluated LE-EpiGNODE across 14 controlled synthetic mechanisms spanning time-modulated, state-dependent, nonlinear and coupled dynamics, three complementary COVID-19 settings, 21 comparator models and five targeted ablations. Forecasting gains tended to become more pronounced at longer horizons; at the longest horizon, LE-EpiGNODE achieved the lowest WMAPE in all three real-world datasets, reducing error by 17.7-40.0% relative to the best baseline. Controlled ablations linked these gains to the Lagrangian-Eulerian decomposition and process-aligned priors, while synthetic ground truth showed that learned operators and induced state fluxes track known generating dynamics. These results support mechanism-organized graph neural dynamics as a route to reliable long-horizon extrapolation with directly testable, process-level mechanistic interpretation beyond trajectory accuracy.
[LG-76] Clean Data Can Still Carry Backdoors: Support-Persistent Backdoors for Model Reuse NEURIPS2026
链接: https://arxiv.org/abs/2610.21208
作者: Junhoo Lee,Baekseung Kim,Seungyeon Kim,Nojun Kwak
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: Accepted to NeurIPS 2026
Abstract:Traditional backdoor attacks inject artificial trigger patterns into training data, causing models to misclassify triggered inputs while maintaining normal behavior on clean samples. Since these artificial triggers are absent in clean datasets, standard knowledge distillation (KD) typically eliminates them, leading to the common belief that KD serves as an effective purification defense. In this paper, we propose DAR (Discover-and-Relabel), a support-persistent backdoor method that retains its behavior even after clean-data distillation. DAR identifies naturally abundant patterns in low-dimensional feature subspaces using subspace clustering and applies label-only poisoning to samples that inherently satisfy the discovered rule. At test time, the backdoor is activated by locally modifying the corresponding subspace of the target images. By instantiating spatial and frequency operators, DAR achieves attack accuracy comparable to state-of-the-art backdoor attacks in ImageNet classification and CLIP-based prompt tuning. Notably, DAR keeps attack success above 96% after clean KD and remains defense-resistant in practice. Project page: this https URL
[LG-77] Hard Way The Only Way: Provable Merits of Training on Hardest Examples
链接: https://arxiv.org/abs/2610.21201
作者: Elvis Dohmatob
类目: Machine Learning (cs.LG)
*备注:
Abstract:Despite a surge of empirical interest in the ``unreasonable effectiveness’’ of drastic data pruning for AI systems, a unified theoretical understanding of this phenomenon remains elusive. In this work, we provide a precise high-dimensional theory of learning under data pruning via weighted empirical risk minimization (ERM). As in previous work, we analyze a setting where pruning is performed along a one-dimensional subspace of the input space. Our analysis precisely characterizes the limiting test error curves in the regime where the input dimension N and sample size P tend to infinity with a fixed ratio N/P. The resulting theory recovers several recent results in the literature as special cases and provides a general lens to evaluate how pruning strategies interact with the loss function and label noise. Crucially, we characterize the optimal pruning strategy in the high-sample regime where P greatly exceeds N. We demonstrate that strategic pruning can accelerate the decay of classification test error from the classical sqrt(N/P) rate, typical of vanilla ERM with loss functions without exponential tail (e.g., quadratic loss), to the much faster Bayes-optimal scaling of order N/P. In this setup, we prove that, surprisingly, this optimal performance is achieved by a drastic pruning strategy: discarding almost all training data except for a small proportion, consisting strictly of the 2N hardest training examples.
[LG-78] Poisoning the corpora of protein design systems during pre-training and retrieval
链接: https://arxiv.org/abs/2610.21200
作者: Charalampos Koilakos,Kimon-Antonios Provatas,Michail Patsakis,Candace SY Chan,Georgios A. Pavlopoulos,Ilias Georgakopoulos-Soares
类目: Machine Learning (cs.LG); Quantitative Methods (q-bio.QM)
*备注: double page format, 6 figures
Abstract:Protein foundation models increasingly support the generation of novel biological sequences, yet their reliance on large public datasets creates opportunities for adversarial manipulation of the material they learn from and retrieve. We investigate two data-only poisoning attacks against protein design systems, targeting the pre-training corpus of autoregressive protein language models and the retrieval corpus of a text-guided protein generator. For the first, we pre-trained ProtGPT2 and ProGen2 variants of 300 million parameters, introducing poisoned protein sequences containing one of two triggers followed by one of three payloads. Across all poisoned runs, both architectures acquired highly reliable trigger-dependent generation under uniform and ramped poisoning schedules. Under ramped exposure, the earliest majority activation occurred after only 200 contaminated examples, while for the uniform schedule earliest activation occurred around 5000 poisoned samples. For the second attack, we inserted records into a retrieval corpus system without modifying its weights or decoding procedure. A single query-aware record captured a top-16 retrieval position for 58% of templated requests and 97% of free-text requests. Conditional on successful poisoned retrieval, 62% of requests produced altered sequences, while 19-21% acquired the inserted domain by profile-HMM homology. A universal query-unaware attack using 1,500 records also achieved substantial retrieval capture and domain transfer. Together, these findings demonstrate that adversarial protein sequence contributions can establish persistent, sequence-specific behaviour during pre-training or redirect protein generation at inference, while leaving ordinary model behaviour largely preserved. Our work establishes the integrity and biological provenance of protein sequence data as critical safeguards for generative protein design systems.
[LG-79] Resisting Harmful Instructions through Training-Free Weight Editing
链接: https://arxiv.org/abs/2610.21189
作者: Ali Ebrahimpour-Boroojeny,Xiaocong Yang,Samuel Schapiro,Zahra Borhani,Francisco R. Ortega,Yujia Zheng,ChengXiang Zhai
类目: Machine Learning (cs.LG)
*备注:
Abstract:Instruction-following language models can comply with requests that conflict with intended behaviors and safety alignment. We introduce a training-free weight-editing approach for reducing compliance with targeted undesirable instructions. Our method needs only forward passes on a small set of prompt pairs per target behavior, and because the edit is stored in the model’s weights, changing or deleting instructions in the prompt cannot undo the weight modification. We evaluate the approach on two target harmful behaviors across five instruction-tuned models: instructed falsehood and instructed sexism. On two instructed-lying benchmarks, it improves mean honest response rates by 6.2 and 4.6 percentage points over the strongest alternative, significantly on three of the five models and within noise on the other two. In a controlled study of seven prompt formats, it maintains higher honest-response rates than single-layer activation steering when orders to lie arrive in single-turn formats not used to construct the edit. For instructed sexism, it reduces the mean sexist-response rate from 57.9% to 24.3%; combining the edit with a fairness prompt reduces it further to 11.9%. These results establish our contrast-based weight editing approach as an effective training-free intervention in two distinct behavioral domains, supporting its use both independently and alongside prompt-based safeguards.
[LG-80] Sensor Diagnosis for Fuel and Emissions Monitoring in Heavy-Duty Diesel Vehicles NEURIPS2026
链接: https://arxiv.org/abs/2610.21188
作者: Tejas Dahiya,Saurabh Kumar Gupta
类目: Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注: Accepted at the Tackling Climate Change with Machine Learning workshop at NeurIPS 2026
Abstract:Faulty sensor readings can distort estimates of fuel use and carbon dioxide emissions from diesel engines. We evaluate two procedures that use relationships learned from healthy measurements to diagnose faults and guide fuel correction. The diagnosis procedure tests whether replacing a suspected faulty reading with an estimated healthy value helps a model predict another sensor’s recorded reading more accurately. It then predicts the fault type and decides whether to report the diagnosis. We evaluate five synthetic fault types on 217 steady-state operating points from a heavy-duty diesel engine, excluding all measurements at selected engine speeds from training. Across 31,500 fault-injected groups of 20 operating points, it issues diagnoses for 77.3% of groups, with 92.1% of diagnoses issued on these groups correct. The fuel correction procedure uses a different sensor-selection rule at individual operating points, testing how each replacement changes the largest prediction error across all six sensors. Fuel correction requires a fault alarm and engine speed and torque close to the training conditions. A fuel reading is replaced only when its replacement reduces the largest prediction error sufficiently more than replacements for the other sensors. Assuming faults in 1% of one-hour scenarios, equally likely sensors and fault types, and equal time at each measured condition, selected fuel corrections reduce mean absolute error in calculated combustion CO2 by 83.8%. On this engine map, learned relationships support complete sensor diagnoses and fuel corrections that improve emissions accounting while preserving unaffected readings.
[LG-81] Coordination-Free Workload Regularization for Variable-Cost Distributed Training
链接: https://arxiv.org/abs/2610.21162
作者: Tao Huang,Chen Zhu,Antonio Alonso,Daming Li,Adarsh Goura,Jaideep Ray,Yitong Zhou
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注:
Abstract:Synchronous data-parallel training usually assigns every rank the same number of records, even when record costs vary by orders of magnitude. For variable-length models, this converts input variation into idle time because each step waits for its most expensive rank-local batch. We present Batch Cost Lookahead (BCL), a coordination-free data-loading policy that buffers a bounded rank-local window and repartitions it into fixed-cardinality packs with balanced estimated cost. BCL preserves the trainer’s record-count contract and requires neither cross-rank metadata exchange nor runtime modification. In simulation over public workload distributions, enabling BCL reduces per-rank estimated batch-cost variability by up to 82.8%. Across 13 NeMo-RL SFT corpora on eight H200 GPUs, rank-local BCL runs 1.050 to 1.167 times faster than pooled MFFD. Compiled H200 measurements on an internal sequence workload show speedups of 1.197 to 1.343 times over packed execution from one to four nodes. We also identify a memory-feasibility boundary: a p99 cap for high-variance or OOM-prone data bounds outliers before BCL balances the rest. These results show that substantial variable-cost imbalance can be removed at the data-loader boundary, while clarifying when data preparation or broader coordination is still necessary.
[LG-82] Predicting Live Birth After Fresh Embryo Transfer in IVF/ICSI: Differential ICM/Trophectoderm Morphology Encoding and a Hybrid Attention–Logistic Regression Ensemble
链接: https://arxiv.org/abs/2610.21134
作者: Abdul Raheem Shahzad,Jinxin Chen,Ming Zhang,Juan Deng
类目: Machine Learning (cs.LG)
*备注:
Abstract:Predicting live birth before fresh embryo transfer could support shared decision-making in IVF/ICSI, but routine clinical data are small, heterogeneous, and contain pathway-dependent variables prone to outcome leakage. We aimed to construct a curated single-centre IVF/ICSI dataset and develop a prediction model suited to small tabular clinical data. \textitMaterials and Methods: We assembled 616 fresh embryo-transfer cycles (251 live birth, 365 non-live birth) with clinical, laboratory, and embryological predictors, and embryo morphology encoded with a novel differential weighting of the inner cell mass and trophectoderm. After KNN imputation, robust scaling, and selection of 50 features, we developed a Hybrid Attention Network (HAN), a learnable feature-attention gate followed by a compact multilayer perceptron, and a HAN + logistic-regression (LR) ensemble with a validation-tuned weight. Models were benchmarked against classical machine learning, TabPFN, and text-serialized language models on a held-out test set; AUC-ROC was the primary metric. \textitResults: The HAN + LR ensemble achieved the highest test AUC-ROC (0.710; accuracy 0.710, F1 0.630), ahead of the HAN backbone alone (0.700), logistic regression (0.667), the TabPFN + LR hybrid (0.666), and TabPFN (0.650). Text-serialized language models performed near chance (AUC \le 0.61). The differentially weighted trophectoderm feature ranked highest by consensus importance across models. \textitDiscussion: A compact attention network combined with a linear model outperformed both components and strong foundation-model baselines on a small clinical cohort. \textitConclusion: The curated dataset and HAN + LR ensemble offer a transparent, reproducible basis for IVF/ICSI live-birth prediction; external validation is the next step.
[LG-83] Noise-Space Reward Alignment for Few-Step Generative Models
链接: https://arxiv.org/abs/2610.21125
作者: Woocheol Shin,Taeyoung Yun,Jiwon Sung,Hyeongyu Kang,Jeongmin Park,Jaewoo Lee,Kyuil Sim,Dohyun Kim,Yutong He,Jinkyoo Park
类目: Machine Learning (cs.LG)
*备注: Project page: this https URL
Abstract:Few-step generative models produce high-quality samples in one or a few evaluations. However, steering them toward desired properties remains difficult, especially when only black-box reward evaluations are available. While one could apply approaches developed for iterative diffusion and flow-based models, such as direct RL fine-tuning or test-time scaling, we find that neither transfers well to the few-step regime: direct fine-tuning suffers from severe reward over-optimization, and test-time scaling sacrifices the efficiency advantage of few-step generation. We propose Noise Policy Optimization (NPO), which explicitly targets the alignment problem for few-step generative models. The key idea is to learn an amortized flow policy over the initial noise space of the generator. The trained policy transports Gaussian noise to a reward-tilted noise distribution, whose pushforward distribution through the frozen generator is exactly the reward-tilted output distribution. By keeping the generator frozen, we can effectively utilize the knowledge encoded in the pretrained model, yielding a better reward-quality trade-off than direct fine-tuning. The amortized policy moves the search cost into training, so inference adds only a lightweight noise sampling step on top of the pretrained model. To improve sample efficiency during training, we use ELBO-based likelihood estimation that requires only sampled endpoints rather than simulated trajectories. Across image, language, and protein structure generation, NPO consistently achieves a better reward-quality trade-off than direct fine-tuning, and matches test-time search methods with 16 to 549 times lower inference cost.
[LG-84] StochLOB: Heavy-Tailed Score Matching for Robust Limit Order Book Trend Prediction STOC NEURIPS2026
链接: https://arxiv.org/abs/2610.21081
作者: Arshia Abolghasemi,Seyed Reza Tavakoli,Parsa Naderi,Nazanin Mirzaei,Behrad Jahani,Saba Yousefzadeh,Tahereh Javaheri,Mohammad Hosein Rohban,Gholamali Aminian,Ali Habibnia
类目: Machine Learning (cs.LG); Trading and Market Microstructure (q-fin.TR); Machine Learning (stat.ML)
*备注: 32 pages, 5 figures. NeurIPS 2026 Workshop: AI for Stochastic Dynamics
Abstract:Score-based diffusion methods usually perturb data with Brownian Gaussian noise, corresponding to a light-tailed forward stochastic differential equation (SDE). For market time series this SDE can be a poor abstraction: limit order book (LOB) order flow is heavy-tailed, bursty, and occasionally jump-like. We study STOCHLOB, a training-only score-matching framework that changes the forward diffusion corruption law while leaving the deployed predictor deterministic. Instead of Brownian-only Gaussian corruption, the noising process uses two non-Gaussian SDE-inspired alternatives: jump-Lévy noise through a compound-Poisson variance mixture, and clipped \alpha -stable noise through a positive-stable Gaussian scale mixture. These corruptions inject finite jumps and heavy-tailed residuals into training without requiring reverse-time diffusion sampling at inference. The resulting score target has a useful interpretation: its radial factor h_t® is the posterior expected corruption precision, so large residuals are weighted according to the latent corruption scale that most plausibly generated them. On chronologically held-out Coinbase LOB data, the proposed variants remain competitive on uncorrupted BTC and altcoin test horizons and improve several realized tail-window stress diagnostics. An exploratory Feishu A-share transfer study further tests whether the probability outputs remain usable under transaction costs and trading-rule constraints. Code and configuration files are available at this https URL.
[LG-85] Adaptive Randomized Neural Networks for PDEs: Structure-Aware Feature Redistribution and Residual-Driven Block Enrichment
链接: https://arxiv.org/abs/2610.21077
作者: You Yang,Fei Wang
类目: Numerical Analysis (math.NA); Machine Learning (cs.LG)
*备注: 24 pages, 9 figures, 13 tables
Abstract:Randomized neural networks retain an output-linear least-squares solve, but globally sampled random features may fail to capture the spatial and scale structure of PDE solutions. We propose SFRBE–RaNN, a two-stage adaptive trial-space construction. SFR redistributes a fixed feature budget according to regional indicator masses and adapts regional weight scales through low-dimensional residual-based optimization. RBE constructs nonredundant enrichments by screening multiple activation families, removing represented response directions, and selecting the smallest admissible prefix. For solution-dependent operators, candidates are screened with a frozen operator and accepted only after nonlinear residual verification. For fixed linear problems, we establish an exact discrete residual-reduction identity and a conditional solution-error estimate. Building on this identity, we combine safeguarded SFR with RBE capture factors to derive a finite-horizon error bound involving baseline best approximation and empirical consistency, together with a conditional random-search success-probability bound. Numerical experiments on localized, multiscale, and nonlinear space–time problems demonstrate the method’s effectiveness.
[LG-86] SALTO: Solver-Anchored Lifted Transfer Operators for Nonlinear Event Inference
链接: https://arxiv.org/abs/2610.21049
作者: Qi Wang,Rylie Foster
类目: Machine Learning (cs.LG); Dynamical Systems (math.DS)
*备注: 22 pages, 11 figures, 12 tables. Under review
Abstract:Learned dynamical models can accurately predict trajectories while producing unreliable sensitivities when differentiated or transposed, limiting their use in inverse problems and scientific decision-making. Classical adjoints provide exact local derivatives, but a single adjoint cannot represent the nonlinear response of a system to a family of finite-amplitude perturbations. We introduce SALTO (Solver-Anchored Lifted Transfer Operators), a Koopman-based framework for inferring finite-amplitude sources. In the lifted space, the nonlinear dynamics are approximated by a linear propagator, and its transpose carries each sensor backward in a single solve. The result is a reusable lifted adjoint that predicts the measurement produced by any candidate localized event with one encoder pass and an inner product. To give the model the simulator’s exact local behavior, SALTO imposes three structural constraints: it anchors the physical-state dynamics to the simulator’s linearized propagator, centers the learned observables so that they vanish to second order at the reference state, and uses a block-triangular lifted propagator. These constraints guarantee that the model reduces to the simulator’s tangent dynamics and classical adjoint as the amplitude vanishes, while the learned observables carry only the finite-amplitude correction. On three nonlinear PDE systems, SALTO predicts the measurements of finite-amplitude sources to within 2-8% where the classical adjoint is in error by 41-165%, recovers source position and intensity from sparse sensors nearly as accurately as inversion through the solver, and samples the joint posterior of two interacting sources, matching the solver’s posterior on reaction-diffusion at a twentieth of its cost. One trained model serves new sensor layouts and measurement times with a single backward solve.
[LG-87] CausalEvo: Causal Skill Self-Evolution for Health LLM Agents NEURIPS2026 ALT
链接: https://arxiv.org/abs/2610.21016
作者: Yuchen Zhang
类目: Machine Learning (cs.LG)
*备注: GenAI for Health Workshop @ NeurIPS 2026, Sydney
Abstract:Large language model (LLM) agents for health can accumulate reusable textual skills from task feedback, such as calculation criteria, measurement time points, and formula units. Deciding which skills to retain is consequential: useful rules improve later answers, while faulty generalizations propagate errors. We propose CausalEvo, a method for causal skill self-evolution that estimates the relative effects of deliberately changed skill sets while holding the model weights, feedback cases, and rule count fixed. At each evolution step, a frozen LLM proposes candidate rules, which are pooled with the incumbent skills so that the current skill set remains a feasible choice. Paired skill replacements against shared background rules are scored on the same feedback cases; least squares converts the score differences into relative skill values, and the top-ranked rules are retained, favoring incumbents on ties. A decision-loss bound separates candidate support, approximation, and estimation errors. On six MedCalc-Bench-Verified tasks with 120 held-out cases, CausalEvo achieves 60.00% accuracy, compared with 55.83% for SkillOpt and 51.67% without skills. Across the recorded training configurations, which use different optimization budgets, it consumes 55.6% fewer training tokens than SkillOpt. Gains concentrate in particular calculations alongside regressions on others, and paired accuracy intervals include zero. CausalEvo turns task feedback into controlled decisions about skill retention and replacement. Project repository: this https URL
[LG-88] Imperfect Is Perfect: Learned Column Prefilling for Branch-Price-and-Cut
链接: https://arxiv.org/abs/2610.21001
作者: Haoran Liu,Yu Yang
类目: Machine Learning (cs.LG)
*备注: 27 pages, 3 figures
Abstract:Column generation folklore warns that initial columns hurt: a warm start may bias the dual and slow the root of branch-price-and-cut. We show the opposite once the columns are imperfect in the right way. Preloading the support of the optimal root LP, the best columns one could know in advance, fails to finish on half of our hardest instances; a pool of ordinary routes that need not belong to any optimal solution, held outside the master and priced at the current smoothed dual, speeds the root up. Because the columns need not be optimal, producing them becomes a learning task: a neural edge scorer guides a constructive generator that builds the pool in seconds. A second network predicts the initial center of Wentges smoothing from the instance file, and with it the direction of the smoothed pricing queries; the solver’s own updates run unchanged. With prefill and center in place, exact pricing dominates root pricing time, so we rebuild bucket-graph labeling around its dominance test for the GPU and certify on the CPU. Inside RouteOpt, the state-of-the-art exact solver for capacitated vehicle routing, the pipeline reaches the same certified root bound in 13.2% of vanilla RouteOpt’s root pricing time over 95 instances with 300 to 1000 customers, prefill alone in 26.1%, and 17.5 times faster in the best case. In an exact branch-price-and-cut solver for unrelated parallel machine scheduling, prefill with smoothing and the learned center certifies all 25 benchmark instances, 12 times faster than the solver’s own column generation, which certifies 20 of them.
[LG-89] Halve-Then-Greedy: Adaptive Sampling for Posterior Success with Near-Optimal Instance Dependence
链接: https://arxiv.org/abs/2610.20997
作者: Yinan Li,Kwang-Sung Jun,Kapilan Balagopalan,Tuan Nguyen,Yao Zhao,Abbas Zaidi,Houssam Nassif
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We study fixed-budget adaptive experimentation aimed at maximizing the probability of obtaining sufficient posterior evidence that a recommended design’s mean performance exceeds a target. Unlike best-arm identification, this objective does not require finding the best design. For Gaussian bandits with a Gaussian prior, we propose \emphHalve-Then-Greedy (HTG), which combines a large initial balanced allocation, successive halving, and posterior-greedy sampling. We prove instance-dependent, finite-budget exponential bounds on posterior-failure probability that capture both separation from unfavorable arms and the presence of multiple favorable arms. On one-positive-arm families, we establish an asymptotic sample-complexity lower bound that HTG matches up to a logarithmic factor. We further analyze uniform sampling and sequential halving, revealing regimes where they require more samples than HTG by linear and logarithmic factors, respectively. Simulations illustrate HTG’s sample-efficiency advantages and their dependence on instance structure.
[LG-90] MIDI-REPA: Improving Text-to-Music Generation with Symbolic Representation Alignment ICASSP2027 ACL
链接: https://arxiv.org/abs/2610.20967
作者: Song-Ze Yu,Fang-Duo Tsai,Yi-Hsuan Yang
类目: ound (cs.SD); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
*备注: 5 pages, 2 figures. Submitted to ICASSP 2027. Demo and code: this https URL
Abstract:State-of-the-art text-to-music models increasingly rely on scaling model capacity, data, and compute, while still learning musical structure directly from dense audio. Yet melody, rhythm, harmony, and structure can be represented compactly in symbolic form. We introduce MIDI-REPA, a representation-alignment framework that derives MIDI supervision from automatic music transcription (AMT) to align intermediate representations of a flow-matching DiT. Building on the 122M-parameter FluxAudio-S baseline from the 2026 ICME Academic Text-To-Music Challenge, we investigate what, where, and when to align. Aligning a MuScriptor-derived piano roll with an early DiT layer throughout the full flow trajectory performs the best. Without additional training data or techniques, MIDI-REPA remains competitive on the ATTM objective metrics while remaining trainable on a single RTX 3090 in about 31 hours. Our results highlight symbolic supervision as a simple and inexpensive learning signal for improving text-to-music generation. We release our transcription pipeline, JamendoMIDI, training code, and audio samples at this https URL.
[LG-91] Planning to Learn: Real-Time Robust Adaptive Control on the GPU with Guaranteed Future Learning
链接: https://arxiv.org/abs/2610.20925
作者: Anutam Srinivasan,Siddhant Panse,Glen Chou
类目: Robotics (cs.RO); Machine Learning (cs.LG); Systems and Control (eess.SY); Optimization and Control (math.OC)
*备注:
Abstract:Robust adaptive control is vital for keeping robots safe while actively reducing uncertainty in unfamiliar settings. However, traditional methods are often conservative since they fail to account for guaranteed future parameter learning during planning and struggle to scale to high-dimensional systems. In this work, we present A-SLS, a learning-aware robust adaptive model predictive control (RAMPC) framework based on system-level synthesis. A-SLS forecasts future parameter uncertainty and preserves correlations between process disturbances, parameter error, and estimate updates to construct tight, guaranteed reachable sets that account for future information gain. We also develop a GPU-accelerated A-SLS solver to enable real-time performance. Across nonlinear robotic systems, A-SLS reduces reachable-tube conservativeness, maintains real-time performance, and scales to whole-body control for a quadruped with 61 states, 8 uncertain parameters, and 12 control inputs. We further demonstrate the approach on hardware, where a physical Crazyflie drone learns an uncertain dynamics parameter online to stabilize on the fly.
[LG-92] Recursive Offline Distillation via Entropy Gain Sampling and Answer Mixing
链接: https://arxiv.org/abs/2610.20914
作者: Andrey Goncharov,Daniil Vyazhev,Simon Epanov,Alexey Zaytsev
类目: Machine Learning (cs.LG)
*备注:
Abstract:Distillation of reasoning chains from Large Language Models (LLMs) into their smaller siblings is a standard training approach. Frequently, researchers do not have full access to a teacher model and its internal representations, which limits the resulting performance. We propose a recursive offline distillation method combining two components: entropy-based sampling for selective distillation and answer mixing. Specifically, for three open models (3-4B), at each epoch, we sample reasoning chains weighted by entropy difference with an auxiliary medium model (70-72B) and, next, supplement thinking traces with single-token ground truth answers. Our approach consistently outperforms naive distillation with comparable data budgets ( 57.15 \pm 1.01 vs 51.61 on average) and offers a much steeper learning curve. Additionally, we present a comparative analysis of methods to generate and normalize distillation traces.
[LG-93] Sign Betting for Misspecified Regression: Variance Adaptivity and Optimal Approximation
链接: https://arxiv.org/abs/2610.20903
作者: Yinan Li,Han Han,Kwang-Sung Jun
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We study bounded regression under model misspecification, evaluating a function f by its L_1(P_X) distance to the conditional mean \eta(x)=\mathbbE[Y\mid X=x] , rather than the standard L_2(P_X) . This L_1 metric is natural for downstream decision problems such as bandits and reinforcement learning. We analyze the min–max estimator of Baraud Maillard (2025), which we call \emphsign-betting regression. Our main results show that sign-betting regression achieves high-probability L_1 oracle inequalities under misspecification. Under realizability, our bound adapts to the average conditional variance and recovers the variance-adaptive rate of Li et al. (2026). We further characterize the optimal universal approximation factor for proper learning by showing that sign betting achieves factor 3, while no proper learner can satisfy a distribution-free, high-probability L_1 oracle inequality with a universal approximation factor strictly below 3. We extend the upper-bound analysis beyond finite classes and complement the theory with experiments in synthetic and real-world model-selection tasks.
[LG-94] Do Emergent Misalignment Models Share a Circuit?
链接: https://arxiv.org/abs/2610.20887
作者: Muhammad Haseeb,Muhammad Tahir
类目: Machine Learning (cs.LG)
*备注:
Abstract:Fine-tuning language models on narrowly defined harmful tasks can induce emergent misalignment (EM) across unrelated domains. While prior work has identified shared misalignment-associated representations across independently fine-tuned models, it remains unclear whether these models rely on common computational components. We investigate this question using three LoRA fine-tunes of Qwen2.5-3B-Instruct trained on harmful medical advice, risky financial advice, and extreme-sports recommendations. Using direction-based activation tracing, we identify three attention-head locations shared among the five highest-ranked heads of all three organisms. Replacing each organism’s five selected head contributions with Base contributions reduces free-generation EM from 38.9% to 22.9% for medical, 50.7% to 12.5% for finance, and 41.7% to 11.1% for sports, outperforming layer-matched random interventions without reducing judged coherence. Cross-organism activation transfer further shows that the shared heads exhibit partially interchangeable effects on fixed-response preference. These findings provide evidence that representational convergence in EM extends to shared, causally relevant attention-head components within a common model architecture.
[LG-95] Clinically Interpretable Biological Age Estimation Using CatBoost and SHAP with Continuous and Categorical Features from NHANES
链接: https://arxiv.org/abs/2610.20868
作者: Yanni Sun,Zheng Qu,Kevin Ren,Zhanglin Lin Cui
类目: Machine Learning (cs.LG); Computation (stat.CO)
*备注: 31 pages, 6 figures, accepted by 2026 ICSA Applied Statistics Symposium, June 14 - June 17, 2026 at Arlington, Virginia, USA
Abstract:Biological age (BA) captures inter-individual variation in physiological aging beyond chronological age (CA). Although machine-learning models have advanced BA estimation, many rely primarily on continuous laboratory biomarkers, and SHapley Additive exPlanations (SHAP) dependence plots are typically interpreted visually. We developed a clinically interpretable machine-learning framework for BA estimation using heterogeneous clinical data from the National Health and Nutrition Examination Survey (NHANES 2011-March 2020, N = 24,983 adults aged = 18). A CatBoost regression model incorporated 44 demographic, laboratory, physical examination, and lifestyle predictors to estimate CA, defining model-estimated age as BA. SHAP provided multi-level interpretability. Crucially, building on standard visual inspection, we implemented a prespecified automated procedure directly on SHAP dependence plots to quantify feature-value ranges associated with lowest and highest BA contributions, corresponding adjustment magnitudes, and zero-crossing thresholds. Model-estimated BA strongly correlated with CA in held-out testing data (Pearson r = 0.798, P 0.001). Population-level SHAP identified blood urea nitrogen, serum glucose, vigorous recreational activity, red cell distribution width, and alanine aminotransferase among the leading predictors, with three categorical variables ranking among the top 25. Automated quantitative SHAP analysis mapped predictor-specific ranges, effect sizes, and inflection points, while waterfall plots enabled individual explanations. By extending qualitative visual interpretation to reproducible quantitative metrics across heterogeneous clinical data, this framework enhances transparency and provides a scalable methodological foundation for biological aging research and precision health.
[LG-96] Long-Context Drafting Does Not Require Long-Context Drafters
链接: https://arxiv.org/abs/2610.20865
作者: Aref Jafari,Vansh Bhatia,Matt Qin,Subhajit Dutta Chowdhury,Ben Lee,Vikram Appia,Ali Ghodsi,Mehdi Rezagholizadeh
类目: Machine Learning (cs.LG)
*备注:
Abstract:Speculative decoding pairs a cheap drafter with parallel verification against the target. At ultra-long context it stops paying, because a drafter that attends over the full prefix inherits the target’s context-dependent cost, so the speedup it was meant to provide first erodes and then reverses. The standard response is to build a long-context drafter, by training or fine-tuning one, or by using a multi-token-prediction head shipped inside the target checkpoint. We argue this is unnecessary, because the drafters in use are target-conditioned, reading hidden states the target computed over the entire prefix, so distant context already reaches them without their own attention, and verification corrects what they miss. We therefore freeze released short-context drafters and run them far past their training length under two inference-time mechanisms, one bounding the attention cache they see, which fixes their cost, and one re-basing their rotary position indices into the trained range. No weight is touched and the only tunable saturates early, so one vLLM-side configuration spans 4k to two million tokens, costing at most 0.16x of speedup at short length. We prove when a common index shift leaves the drafter’s computation exactly unchanged, and decompose acceptance into two measurable gaps with sharp bounds. Whatever speedup a frozen drafter delivers at short context, it retains out to two million tokens, across three target-drafter pairs and five corpora. The claim concerns the flatness rather than the multiple, and removing either mechanism eliminates it. The gain is conditional on the speculative cost per emitted token staying below the matched target-only cost.
[LG-97] Denoising Score Matching in Diffusion Models: The Effect of Rank Constraints on Generalization
链接: https://arxiv.org/abs/2610.20850
作者: Anand Jerry George,Nicolas Macris
类目: Machine Learning (cs.LG)
*备注: 8 pages
Abstract:Architectural biases restrict the class of functions that a diffusion score network can represent, but their effect on generalization remains poorly understood. We study denoising score matching with score networks having a rank constrained non-linear term and a linear skip connection. Numerical experiments with trainable three-layer networks on data supported on an embedded sphere show that increasing rank lowers training error, but its effect on test error depends on sample size and data dimension. To explain this behavior, we analyze a random-feature score model under rank constraint with Gaussian data supported on a low-dimensional subspace. We characterize the optimal rank-constrained estimator through a truncated singular value decomposition and derive asymptotically precise limiting training and test errors when sample size (n) , feature width § , data dimension (D) , and rank ® scale proportionally with the ambient dimension (d) . The resulting spectral characterization distinguishes eigenmodes that improve the empirical fit from those that improve population accuracy. It also identifies the dimension of the orthogonal complement of the data subspace as a natural rank threshold. The theoretical learning curves qualitatively reproduce the observed rank-dependent overfitting, providing a tractable account of how bottlenecks or rank constraints can regularize score estimation and why their effectiveness depends on data dimension and sample size.
[LG-98] Analog RF Computing for Integrated Data and AI Flows Over Wireless Networks
链接: https://arxiv.org/abs/2610.20841
作者: Wentao Yu,Vincent W.S. Wong
类目: Information Theory (cs.IT); Emerging Technologies (cs.ET); Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: We take a step toward wireless networks that can compute as well as communicate, using analog RF computing to deliver AI inference to resource-constrained edge devices while simultaneously serving conventional data transmission. This is a follow-up to our analog RF computing framework: https://arxiv.org/abs/2605.14331 . To be presented at IEEE Globecom 2026
Abstract:Many resource-constrained edge devices require artificial intelligence (AI) services but are not powerful enough to store or execute neural networks (NNs). Analog radio frequency (RF) computing addresses this limitation by letting a base station (BS) multicast a waveform encoding NN weights to different clients. Each client then multiplies the received weight-encoded waveform with its locally generated input-encoded waveform through a passive mixer in its RF frontend, thereby realizing the matrix-vector multiplications (MVMs) needed for AI inference with an extremely low power and memory requirement. This can enable telecom operators to provide AI inference services to the edge devices connected to their networks. In practice, however, this multicast AI flow must share the spectrum, power, and spatial resources with other data flows. This paper considers a multi-user multiple-input multiple-output (MU-MIMO) downlink scenario and studies the resulting relative aggregate data rate reduction from integrating the multicast AI flow with the unicast data flows. We formulate a joint beamforming problem for the data and AI flows to maximize the aggregate data rate subject to computing accuracy, client energy, total power, and hardware-related constraints. We develop a weighted minimum mean-square error (WMMSE)-based algorithm to solve the problem. Numerical results based on 3rd Generation Partnership Project (3GPP) channels show that the relative aggregate data rate reduction from integrating an AI flow with data flows is below 5% in representative settings. This indicates that telecom operators can use analog RF computing to provide edge AI services while providing data communication services simultaneously.
[LG-99] A Phenomenology of Neural Geometry in Vision-Language-Action Models
链接: https://arxiv.org/abs/2610.20840
作者: Mozes Jacobs,Aiden Swann,Mathilde Papillon,Siddharth Boppana,Fenil R. Doshi,Usha Bhalla,Thomas Icard,Matt Feiszli,Thomas McGrath,Vasudev Shyam,Demba Ba,Leon Bergen,Nina Miolane,Owen Lewis,Jack Merullo,Ekdeep Singh Lubana,Atticus Geiger,Matthew Kowal,Thomas Fel
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: 41 pages, 10 pages of main text, 8 main text figures
Abstract:Robotic manipulation is inherently geometric, but what shape does the world take inside the policy that performs it? Prior work on vision-language-action (VLA) policies establishes that task-relevant variables are linearly readable from activations, yet readability certifies only that a variable is present and says nothing about the shape its values trace, which may be curved and multi-dimensional. Using controlled sweeps, natural rollouts, linear probes, and activation interventions, we build a taxonomy of seven task-relevant concepts in VLA models. We find that realized geometry is predominantly multi-dimensional and is shaped by the constraints under which the policy acts. Physical joint limits, for example, leave wrist orientation an open arc rather than a circle, and distance to the target, a single scalar, is realized in two orthogonal subspaces, a coarse ruler for the approach and a fine ruler for the grasp. The same geometries reappear in two further VLA models on real-world robot data, indicating that they are neither model-specific nor artifacts of simulation. We demonstrate how the findings are actionable by steering along the geometry a concept realizes. Overall, this work lays out an initial phenomenology of realized geometry in VLA policies and demonstrates that a concept’s shape carries as much importance as its readability.
[LG-100] ask-Phase-Conditioned Expert Routing Is Not a Reliable Proxy for Quantization Sensitivity in MoE FFN Tensors
链接: https://arxiv.org/abs/2610.20826
作者: Yifan Liu
类目: Machine Learning (cs.LG)
*备注: 14 pages, 1 figure, 8 tables. Negative-result study with a random-selection control
Abstract:We investigate whether per-layer, per-task-phase (prompt/reasoning/response) expert-routing statistics in a Mixture-of-Experts (MoE) language model can guide layer-wise mixed-precision quantization of routed-expert FFN tensors (down/gate and up). To attain layerwise routing statistics, we build a routing-capture and profiling pipeline for Qwen-3.6-35B-A3B in this http URL, derive a composite sensitivity score from routing entropy/decisiveness, coefficient of variation (CV), Jaccard/phase-fragility, and top-1 expert strength, and use coordinate descent to tune a layer allocation that respects a fixed average bits-per-weight budget. Notably, the profiling pipeline is applicable to standard top-k MoE architectures. Across independent experimental conditions (about 4.0 bpw flat, about 2.7 bpw steep, and about 2.7 bpw random allocation as a control), reallocation fails to outperform the uniform baseline within each quality tier. In the flat regime, reallocation is statistically neutral (z about 0.9). For the steep tier, the random reallocation performs within 1 SE of the routing-guided reallocation at the same tier, indicating a structural loss at this quality rung caused by mandatory downgrades and Kullback-Leibler Divergence (KLD) tail injection. This aggressive-budget condition shows a decisive loss (z greater than 4 SE) replicated across three independent weight vectors. We argue that the deeper issue is that routing-behavior statistics are an unvalidated proxy for true per-layer quantization cost.
[LG-101] O-RAN-Native Jamming Detection Under Network Heterogeneity
链接: https://arxiv.org/abs/2610.20816
作者: Sneihil Gopal,Fernando J. Cintrón,Richard A. Rouil
类目: Machine Learning (cs.LG)
*备注:
Abstract:Radio-frequency jamming threatens cellular network operations in contested environments. The Open Radio Access Network (O-RAN) architecture exposes per-User Equipment (UE) telemetry through the E2 interface, enabling Key Performance Indicator (KPI)-based jamming detection without proprietary hardware. However, operational deployment demands detection methods that generalize across heterogeneous networks. We propose a Statistical Ensemble Method (SEM) combining adaptive Median Absolute Deviation (MAD)-based baseline detection, count-based anomaly detection, and abrupt change detection through a weighted voting system. SEM operates in two modes: a cold-start mode requiring no labeled data for immediate deployment, and a supervised mode where parameters are refined offline via grid search as labeled data accumulates. We evaluate SEM against five machine learning (ML) classifiers and five deep learning (DL) sequence models using approximately 25 hours of testbed data comprising logs from eight heterogeneous 5G environments, employing both single-environment and multi-environment leave-one-out (LOO) validation. Our analysis indicates that under conditions not encountered during training (distribution shift), learning-based models tend to lose recall and miss jamming events, whereas SEM, even in cold-start mode, maintains detection capabilities at the cost of false alarms. Furthermore, SEM’s parameter optimization can be mapped to O-RAN’s Non-RT RIC as an rApp for continuous refinement, allowing operators to transition to learning-based alternatives once corpus coverage is sufficient.
[LG-102] Optimal Explicit and Implicit Regularization Strengths in High-Dimensional Continual Linear Regression
链接: https://arxiv.org/abs/2610.20804
作者: Gilad Karpel,Daniel Soudry
类目: Machine Learning (cs.LG)
*备注:
Abstract:We study continual linear regression in a high-dimensional regime, where a learner observes a sequence of tasks and updates its parameters sequentially. In this setting, we derive an explicit closed-form expression for the generalization loss of a model trained with gradient descent for a fixed number of steps per task, which holds for arbitrary teachers. We prove that the optimal fixed step size decays nearly inversely with the number of tasks, specifically as \ln T / T , for a single teacher and, more generally, i.i.d. teachers. Under this scaling, we show that the resulting generalization loss is asymptotically equivalent to that achieved by optimally tuned L2 (isotropic) regularization. Furthermore, we extend our analysis to generalized ridge regression with diagonal regularization matrices, obtaining a closed-form characterization of the generalization loss and identifying necessary and sufficient conditions under which the optimal regularizer is isotropic. Finally, we identify a regime where the optimal step size remains finite even as the number of tasks tends to infinity, thus revealing a non-trivial tradeoff between task variability and implicit regularization, in contrast to the i.i.d. teacher scaling. Our theoretical predictions are corroborated by experiments on both linear models and neural networks, providing practical guidance for the design of continual learning systems.
[LG-103] Privacy-Preserving Text Inference for Health-Misinformation Classification and Semantic Retrieval
链接: https://arxiv.org/abs/2610.20796
作者: Jasleen Kaur,Chloé Cachet,Koray Karabina,Jean Belo Klamti,Philippe Lamontagne
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: Published as a conference paper at SMASH 2026
Abstract:Health-misinformation analysis presents a dual challenge: misleading claims can affect public well-being, while systems designed to detect them may expose sensitive user-submitted text. This motivates privacy-preserving inference for public-health applications in which user queries may contain confidential medical information and trained models and reference databases are maintained on organizational servers. We design a privacy-preserving text inference framework based on CKKS homomorphic encryption for health-misinformation classification and semantic retrieval. The classification pipeline fuses BioBERT embeddings, PCA-reduced MPNet embeddings, and SVD-reduced TF-IDF features, while the retrieval pipeline combines MPNet embeddings with ISCC-SCT semantic fingerprints through late fusion. We evaluate predictive performance and benchmark plaintext versus ciphertext-based execution. The hybrid classifier achieved an accuracy of 0.971 and a macro-F1 score of 0.950, with identical reported metrics under plaintext and ciphertext-based inference. The hybrid late-fusion similarity model achieved a Pearson correlation of 0.8691 with human sentence-pair similarity scores on STS-B. For ciphertext-based retrieval, we evaluated naive and optimized implementations. The optimized implementation uses CKKS single instruction, multiple data (SIMD) packing to compute and return multiple similarity scores together, reducing single-query latency over 7,128 indexed items from 222.028 to 5.168 seconds and communication overhead from 6.11 GB to 2.13 MB. These findings demonstrate the technical feasibility of privacy-preserving text inference for health-misinformation applications and characterize its predictive utility and efficiency trade-offs.
[LG-104] Sliced Wasserstein Sketches with Error Certificates
链接: https://arxiv.org/abs/2610.20788
作者: Elaheh Akbari,James Liu,Keaton Hamm,Soheil Kolouri
类目: Machine Learning (cs.LG)
*备注:
Abstract:Compact representations that support geometrically meaningful comparisons are valuable for learning from probability distributions. We study optimal, free-weight M -atom quantization of one-dimensional probability measures in quadratic Wasserstein distance and extend it to higher dimensions via slicing. Using L shared projections yields a sketch with O(LM) comparison time. Our main result is an exact decomposition of squared distance error into residual and partition-mismatch terms, yielding computable certificates for arbitrary centroid sketches. For optimal sketches, a finite (2+\delta) moment for some \delta0 guarantees O(M^-1) error in the finite-direction sliced distance, while equal-mass sketches can converge more slowly. Each slice preserves its mean, while storing residual energies enables covariance recovery from sufficiently many generic directions. Experiments on document, point-cloud, and flow-cytometry benchmarks show that optimal sketches substantially reduce storage while preserving the classification performance of uncompressed sliced distances. They also improve generalization from few labelled subjects and support end-to-end learning as compact pooling layers.
[LG-105] Selection-Safe Conformal Early Exit: Distribution-Free Coverag e under Adaptive Depth
链接: https://arxiv.org/abs/2610.20773
作者: Adria Binte Habib,Ibne Farabi Shihab
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Calibrating a conformal prediction set at each early exit does not preserve coverage once a data-dependent exit is selected. We construct one terminal path score from prefix maxima, calibrate a single split-conformal quantile, and return nested prefix sets that all contain the valid terminal set. The guarantee is selection-safe without an alpha/L correction and holds simultaneously for every post-calibration label-free stopping policy that preserves terminal-set containment. We also bound the efficiency cost of the path maximum by the empirical tail of the early-over-terminal score excess. Across eight encoder task-backbone cells, pathwise selection attains macro coverage 0.901 versus 0.814 for naive layerwise selection, with average set size 1.41 and latency fraction 0.71. On CIFAR-100 with a 24-exit ViT-L/16 (n_cal=500), first-eligible stopping at budget b=4 attains coverage 0.902, set size 2.76, and latency fraction 0.57, and the same calibration supports deployment-time budget changes without refitting. In a joint exit-count/calibration-size sweep at n=100, pathwise set size moves only from 3.02 to 3.11 as exits increase from 4 to 24, while Bonferroni degenerates to the full label space for 12 or more exits. Paired comparison against MinSE valid selection at (n,L)=(100,24) gives a set-size difference of 0.80 [0.47, 1.13]. Full-vocabulary decoder experiments remain a negative boundary, with sets of 7,162-9,620 tokens. Pathwise calibration is most useful when candidate exits are many relative to calibration size, when budgets change after calibration, and when shallow heads return actionable sets.
[LG-106] Semantic Alignment of EEG and Text: A Contrastive Learning Framework for Decoding Imagined Speech
链接: https://arxiv.org/abs/2610.20766
作者: Shahryar G. Namdari,Francesco Iacomi,Riccardo Barbieri,Luca Mainardi
类目: Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注:
Abstract:Imagined speech decoding from EEG offers a promising route toward naturalistic brain-computer interfaces. Traditional closed-set classifiers treat semantic categories as independent labels and do not explicitly exploit relationships among linguistic intents, limiting output flexibility. We adapt a multimodal contrastive learning framework to imagined speech, aligning EEG trials with language representations in the embedding space of a frozen pretrained text encoder. An EEG encoder is trained with a composite objective combining semantically reweighted EEG-sentence alignment and text-prototype-based class discrimination, encouraging semantic structure while preserving separability among 39 intent categories. We evaluate the approach on the Chisco imagined speech dataset across five subjects, reporting Top-k accuracy, macro-F1, and balanced pairwise matching accuracy (2v2). We also investigate maximum EEG-text cosine similarity (MaxCosine) as an intrinsic indicator of decoding reliability and probe semantic structure beyond category labels using five evaluation-time semantic axes and a qualitative Semantic Gallery Probe (SGP) with previously unseen candidate sentences. The contrastive framework achieved decoding performance comparable to a task-optimized closed-set classifier while mapping EEG into a frozen language embedding space. MaxCosine identified subsets of held-out trials with higher Top-1 and Top-5 accuracy across subjects. Evaluation-time semantic axes showed mean agreement above the 50% chance reference across all five dimensions, and the SGP qualitatively illustrated direct comparison with unseen candidate sentences. Together, these findings suggest that EEG-language alignment can retain comparable fixed-intent decoding while supporting similarity-based retrieval, intrinsic reliability ranking, and evaluation of broader semantic structure.
[LG-107] Vertical Sliced Wasserstein Distances for Meta-Measures
链接: https://arxiv.org/abs/2610.20743
作者: Elaheh Akbari,Ahmadreza Moradipari,Kyungtae Han,Hamed Pirsiavash,Yikun Bai,Rocio Diaz Martin,Soheil Kolouri
类目: Machine Learning (cs.LG)
*备注:
Abstract:Many modern learning problems require comparing probability distributions whose individual observations are themselves probability measures, naturally giving rise to meta-measures in \mathcalP(\mathcalP(\Omega)) . However, optimal transport distances on this space can be prohibitively expensive to evaluate. We introduce the vertical sliced Wasserstein ( \mathrmVSW ) distance, a scalable metric on \mathcalP(\mathcalP(\Omega)) . Rather than slicing along the transport geometry of \mathcalP(\Omega) , \mathrmVSW projects each inner measure using continuous test functions and compares the resulting one-dimensional distributions with one-dimensional Wasserstein distance. We prove that \mathrmVSW is a valid metric on \mathcalP(\mathcalP(\Omega)) , metrizes weak convergence, and admits dimension-independent statistical convergence guarantees. Empirically, \mathrmVSW provides high accuracy and efficiency compared with existing distances for meta-measures. We further derive a practical particle gradient for the empirical \mathrmVSW_2^2 objective and demonstrate its use for gradient-based transport of meta-measures. \mathrmVSW provides a simple and scalable alternative for comparing and optimizing distributions of distributions.
[LG-108] Quantile-Guided Diffusion for Probabilistic Time-Series Forecasting
链接: https://arxiv.org/abs/2610.20739
作者: Shadi Omari,Bashar Khoury,Assaf Schuster
类目: Machine Learning (cs.LG)
*备注: 33 pages, 10 figures, 20 tables
Abstract:Quantile forecasts summarize uncertainty at each future time step, while diffusion forecasters represent it through sampled future trajectories. We study how predicted quantiles can serve as distributional controls for diffusion forecasting. Specifically, we investigate how the mechanism of quantile integration affects its effectiveness in guiding a pretrained diffusion forecaster. We find that residual control throughout the denoising network yields the most consistent improvements. We instantiate this approach as Control-TS, which learns a quantile-control pathway while keeping the pretrained backbone frozen. Across eight benchmarks, Control-TS consistently improves the backbone’s CRPS, achieves competitive performance against state-of-the-art forecasting models, and retains its gains in capacity-matched comparisons. Further analyses examine how these gains depend on quantile granularity, predictor choice, and forecast quality. Finally, we introduce the Conditional Linear Predictive Score (CLPS), which complements distributional metrics by evaluating the downstream forecasting utility of generated futures. Together, these findings support predicted quantiles as an effective interface between explicit uncertainty estimates and diffusion-based trajectory generation.
[LG-109] On-Policy Residual Learning for Diffusion Language Models
链接: https://arxiv.org/abs/2610.20687
作者: Yifan Yu,Yuqing Jian,Jisen Li,Yuezhou Hu,Tianyi Zhang,Xiaoxia Wu,Rui-Jie Zhu,Junxiong Wang,Zhongzhu Zhou,Fan Lai,James Zou,Chenfeng Xu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Residual context is key to unlocking the speed-quality Pareto frontier of diffusion language models (DLMs) by carrying forward predictive distributions across denoising steps. Yet turning this seemingly simple mechanism into an effective learning paradigm is far from simple: there is no clear recipe for how residual context should be learned. We introduce ORL, a framework for on-policy residual learning in DLMs. Algorithmically, we identify that residual learning is inherently on-policy. ORL learns from model-induced residual trajectories on the fly and introduces an entropy-contraction objective that sharpens predictions as the model revisits its own predictions. System-wise, exact residual supervision requires capturing, retaining, and reprocessing many trajectory-dependent states. ORL co-designs the algorithm with asynchronous residual capture and transport, specialized block-diffusion attention, and packed compute-balanced training, making it practical at scale. On a 4B DLM converted from Qwen3-4B, ORL improves tokens per forward (TPF) by up to 6.2x and, with our dlmrt engine, end-to-end throughput by approximately 6.9x at comparable or better quality. In single-request serving on one H100, this mode achieves a stronger speed-quality Pareto frontier than autoregressive models with EAGLE3 and DFlash speculative decoding under both strict and relaxed verification, running 1.74x faster than DFlash at matched accuracy. ORL also transfers to existing DLMs: on LLaDA-2.1-mini, 36 H100-hours and 4.2M self-generated tokens, about 10x less data than prior LLaDA recipes, yield up to 1.5x higher TPF and up to 3x higher throughput at comparable quality, and on DiffusionGemma, ORL raises TPF by 1.1-1.2x. With dlmrt and NVFP4 quantization, ORL models decode a single request at up to 5,046 tokens/s on one B200, 7,708 on eight B200s, and 2,788 on a consumer RTX 5090, with comparable quality.
[LG-110] RoLaQ: Joint Rotation and Range Calibration for 2-Bit KV Caches in Multi-Head Latent Attention
链接: https://arxiv.org/abs/2610.20686
作者: Shiyu Ji,Peng Ling,Chen Lin,Shupeng Li,Yixuan Wang,Qingfu Zhu,Wanxiang Che
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Multi-head latent attention (MLA) compresses the content of a key-value cache into a shared latent per token, making low-bit storage attractive for long-context inference. At two bits, errors in this representation perturb both attention matching and value readout, with effects that depend on error direction and reconstruction scale. We introduce RoLaQ (Rotation-calibrated Latent Quantization), an offline method for joint rotation and range calibration. RoLaQ refines a shared rotation within subspaces defined by the frozen key and value projections, and jointly learns quantization ranges whose thresholds and reconstruction gains are coupled to control the shared signal’s response. Prediction matching and attention-output supervision guide calibration with model weights frozen. Across three MLA models and six benchmarks, RoLaQ achieves the highest average accuracy among the compared 2-bit methods both without and with token protection. On DeepSeek-V2-Lite, an SGLang implementation provides 3.96 times the BF16 KV capacity under the same memory budget and 2.51 times its decode throughput with 256K input tokens.
[LG-111] Causal Syntax Circuits in CodeLMs: Budget and Probe Decide What Looks Localised NEURIPS2026
链接: https://arxiv.org/abs/2610.20683
作者: Harsh Kumar,Tanmay Shailesh Joshi,Rahul Maity,Anusa Saha
类目: Machine Learning (cs.LG); Software Engineering (cs.SE)
*备注: Accepted at the NeurIPS 2026 Workshop on Linguistic Principles for Foundation Models (LP4FM). All four authors contributed equally
Abstract:Code models write Python when the prompt does not say which language to use. We ask a mechanistic question about that habit: is it produced by a small, findable group of components inside the model? We run one pipeline—mean ablation, activation patching, greedy circuit extraction, a random-control significance test, and activation steering—over 55 model-language settings spanning ten open code models (five base and five instruction-tuned), three target languages (Rust, C#, PHP), two prompt suites, and two component budgets. We report five findings. First, the preference is large, grows with model size, and survives instruction tuning: it is larger on every instruction-tuned checkpoint than on its base version, by 12% on average. Second, its measured size is mostly a property of the probe; replacing our aligned prompts with XLCoST parallel functions takes the same measurement on the same models from about 1.6 to about 0.01. Third, whether the preference looks localised depends on how many components the search may keep: with a budget of 12, 1 of 15 settings yields a circuit that beats size-matched random controls after multiple-comparison correction, and with a budget of 40, 6 of 15 do. Fourth, interventions that a smaller-budget study reports as uniformly failing do sometimes succeed: patching makes a Python prompt prefer target syntax in 2 of 45 settings, and steering changes the most likely next token in 9 of 45. Fifth, every circuit whose effect is significant but points away from the target language comes from a setting where circuit extraction failed, which makes such results a diagnostic rather than a phenomenon. The practical lesson is that a claim about whether a behaviour is localised can turn on analysis choices that papers rarely report. We release the full grid so those choices can be varied.
[LG-112] Diminishing Returns of Data in Sparse Autoencoders: Functional Behavior does not Guarantee Feature Reproducibility
链接: https://arxiv.org/abs/2610.20680
作者: Sambit Ghosh,Madelyn Mathai,R. Venkatesh Babu,Chirag Agarwal
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Sparse autoencoders (SAEs) are typically trained on hundreds of millions to billions of activation tokens, yet little work has tested whether this scale is necessary. We study how much activation data SAEs need by training four SAE architectures on three language models with budgets ranging from 5M tokens to a 500M-token reference. We measure two notions of convergence: functional behavior on standard evaluations and one-to-one reproducibility of learned features. Functional data requirements are strongly metric-dependent: reconstruction often approaches the full-data reference at substantially smaller budgets, whereas feature reproducibility continues to improve with data. For Gemma-2-2B BatchTopK, the fraction of strictly matched features rises from 2% at 5M tokens to 29% at 50M. Data selection also matters: random sampling better preserves the original activation distribution and generally favors reconstruction, whereas diversity-based sampling frequently improves AutoInterp. A theoretical analysis separates statistical consistency of subsampled training from the stronger identifiability and optimization conditions required for one-to-one feature recovery. Our results show that SAE training can often be made far more data-efficient when budgets are matched to the goal, and feature-level claims are backed by explicit reproducibility checks, paving the way toward cheaper and more reliable SAE research.
[LG-113] Spectral Indistinguishability and Learning-Curve Identifiability in Quadratic Networks
链接: https://arxiv.org/abs/2610.20678
作者: Dekun Yang
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 37 pages, 2 figures. Code and synthetic data provided as ancillary material
Abstract:We study what spectral summaries and scalar learning curves reveal about centered quadratic networks. We construct analytic positive input densities with exactly Gaussian one- and two-coordinate marginals. Their complete fourth-moment operators are intertwined by an orthogonal map fixing the entire diagonal teacher algebra, so the corresponding anchored spectral measures agree. For selected common teachers with a nonzero cycle invariant, their factorized loss curves differ because the map does not preserve matrix multiplication. Cycle length r , which can be arbitrarily large, gives O(\varepsilon^r) contrast uniformly in training time for each fixed construction. For a supplied pair, noisy scalar loss queries require order \varepsilon^-2r observations, while bounded initialization interventions with scalar gradient readouts require order \varepsilon^-2 . A shared-input score, equal to a paired-gradient difference, attains the optimal input-sample exponent without knowing the teacher. An explicit family of diverging auxiliary teacher scales instead admits a nonzero finite limiting contrast at coupling 1/\log L . Conversely, for Gaussian inputs and a rationally independent teacher spectrum, one exact scalar loss curve identifies the initial predictor up to sign conjugation. Within this rationally independent class, with nonzero diagonal inverse-state error in every coordinate and calibrated time and loss scales, it also identifies the unknown teacher spectrum and active dimension, modulo simultaneous rotations. For fixed dimension and distinct unordered teacher sums, dimension-minimal exact observations give generic joint local recovery. Finite exact global bounds hold for a known teacher and, on the active class, under a blind dimension cap. Exact certificates and full-weight replay verify the construction; numerical inverse checks reveal severe ill-conditioning.
[LG-114] KNOU-Flood: Advancing global streamflow forecasting with lead-aware probabilistic machine learning
链接: https://arxiv.org/abs/2610.20640
作者: Minjong Cheon
类目: Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph)
*备注: 22 pages, 13 figures, 10 tables. Code, weights and per-station results: this https URL
Abstract:Global streamflow forecasting supports flood preparedness, yet data-driven models trained on reanalysis lose skill when driven by operational forecasts, increasingly so with lead time. AIFL addresses this shift with a two-stage transfer-learning recipe and a deterministic LSTM, but it feeds one forecast lead time into the whole input window and gives no uncertainty. We ask how far this open recipe can be extended without changing its recurrent core. KNOU-Flood keeps a single-layer LSTM and the two-stage training and changes the information around it: a longer input window (424 days), lead-aware forecast inputs, GraphCast as a second forecast source next to ECMWF HRES, and a latent-noise ensemble trained with the fair CRPS. We compare against a re-implementation of the AIFL recipe (its code is not public) under identical data, stations, splits and metrics. Our contributions are threefold. First, over three seeds and 2,099 test basins (2021-2023), KNOU-Flood reaches a median KGE’ of 0.65 at LT1 and 0.48 at LT10 against 0.62 and 0.34, with significant gains at all ten lead times under two scoring protocols, also against a 1024-unit baseline and on held-out basins (one seed each). Second, identical weights give an LT10 KGE’ of 0.18 or 0.34 depending on how forecasts are fed, so much of the lead-time decay reported for AIFL is an input-protocol effect; the probabilistic head contributes most to the gain, followed by the raw GraphCast channels, the longer window and mixed fine-tuning. Third, the ensembles reach lower CRPS than probabilistic versions of the baseline, are nearly calibrated with inference-time dropout, and detect 1.5- and 2-year floods better than all baselines; the benefit for rarer floods depends on the seed, and an exploratory tail-weighted variant improves 10-year and larger events. KNOU-Flood is a controlled extension of an open recipe, not a new architecture.
[LG-115] Feature Elicitation for SAE Steering of Deceptive Behaviour in Small LLM s
链接: https://arxiv.org/abs/2610.20638
作者: J Koorndijk
类目: Machine Learning (cs.LG); Computers and Society (cs.CY)
*备注:
Abstract:Research on deceptive alignment is becoming increasingly urgent as frontier-model capabilities advance and their risk of causing catastrophic harm increases. Alignment faking is a form of deceptive alignment, with the compliance gap serving as a key behavioural indicator. This behaviour remains understudied in small LLMs and prior mechanistic interventions have not been shown to mitigate the compliance gap. We test whether steering SAE (sparse autoencoders) features associated with deontological ethics can mitigate this behaviour. We use a novel bag-of-words feature elicitation pipeline for SAE steering to identify what we construe as concept-level steering directions. Our approach reduced the compliance gap to non-significance in Gemma-3-27B-IT, whereas the tested configurations were not effective in LLaMA-3-8B-Instruct. Analysis of scratchpad traces also revealed deceptive model behaviour which we term deceptive defanging: deliberately degrading the aligned functionality of a response (in this case harmfulness) while preserving the appearance of alignment (in this case helpful compliance with the harmful request), sometimes conditioned on tier, training, or RLHF consequences. These findings provide, to our knowledge, the first evidence that a mechanistic activation-level intervention can mitigate the compliance gap in an alignment faking setup. They also identify a form of deceptive behaviour that may broaden our understanding of the different forms of deceptive behaviour in LLMs.
[LG-116] Expanding Attention: A Taylor-Series Route to Linear-Time Transformers with a Fixed Number of Recurrent States
链接: https://arxiv.org/abs/2610.20637
作者: Zicong Huang
类目: Machine Learning (cs.LG)
*备注:
Abstract:We expand the exponential kernel of attention into a truncated Taylor series of n terms and factorize each term into a tensor-power inner product, which collapses the causal history into exactly n recurrent states, each updated by one rank-one outer product per token and independent of context length N . The recursion is lossless with respect to the truncated kernel ( 2.2 \times 10^-16 in float64), and the states are additive across chunks ( 5.7 \times 10^-14 ), so prefill admits a parallel scan. Measured exponents are t \propto N^2.28 for softmax against N^0.91 for states, with a crossover at N \approx 728 and 20.8\times at N = 8192 ; per-token decoding latency is constant at 7.3,\mu s against N^1.04 for softmax, giving 76.0\times at N = 65536 . Accuracy improves exponentially in n while cost grows combinatorially, since the m -th state is \binomd+m-1m wide—a bound confirmed by SVD—so the optimum is second order, n = 3 , at 6.5 \times 10^-4 error and 9,945 elements per head. Query and key normalization is a prerequisite: without it the argument exceeds the convergence radius and adding terms stops helping.
[LG-117] SurvICL: a Foundation Model for in-context Survival Analysis
链接: https://arxiv.org/abs/2610.22242
作者: Julie Alberge,Linus Bleistein,Judith Abécassis,Marine Le Morvan,Gaël Varoquaux,David Holzmüller
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Tabular foundation models (TFMs) now outperform gradient-boosted trees in classification and regression. Yet in survival analysis, which predicts when an event occurs from censored data, often on small clinical cohorts, we find that no existing TFM beats the half-century-old Cox model. We address this issue by introducing SurvICL, a TFM that predicts survival curves in-context. SurvICL is pretrained on purely synthetic datasets generated from a novel flexible proportional-odds prior that outperforms other prior types in our ablation study. Unlike most other survival models, SurvICL is trained with a plain quantile regression loss, which is possible because its synthetic pre-training data generator can provide uncensored event times. Additionally, we introduce the largest survival benchmark to date spanning 103 datasets. On the 63 datasets held out from all design decisions, SurvICL outperforms 20 tree-based, deep, and foundation-model baselines without any tuning: it wins 82.9% of pairwise comparisons in integrated Brier score, against 71.3% for the runner-up, the Cox model, and is the first model to rank above Cox on small datasets.
[LG-118] Attention over noisy examples: generalization and memorization in the exponential Hopfield network
链接: https://arxiv.org/abs/2610.22191
作者: Francesco Alemanno
类目: Disordered Systems and Neural Networks (cond-mat.dis-nn); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
*备注: 15 pages, 9 figures. Code, data and Monte Carlo results at this https URL
Abstract:In Hebbian learning from examples a Hopfield network of N neurons is not supplied with the K archetypes it has to retrieve but with M blurred examples of each, of quality r , and whether it can infer them depends on the load \alpha=K/N (interference among them) and on the dataset entropy \rho=(1-r^2)/(Mr^2) (how little the examples say about them). We transfer this protocol to the exponential Hopfield model, whose zero-temperature update is a step of softmax attention with the unlabelled examples as keys. A signal-to-noise analysis gives in closed form the overlap between the output and an unseen archetype, m^(1)=\rm erf(1/\sqrt2\rho_\rm eff) : the majority-rule law of the pairwise model at small load, with \rho replaced by an attention entropy \rho_\rm eff built from the weight the softmax gives to the correct class and the number of examples it spreads it over. Two consequences follow. First, in a window of inverse temperatures \beta that widens as \sqrt N the interference between archetypes is exponentially small in N\beta : the load drops out and the number of archetypes that can be learnt grows as \exp(c\sqrt N) . Second, learning fails differently: in the pairwise model through a spin-glass phase, in the exponential model through memorization: the softmax concentrates its weight on the training example closest to the input and returns it in place of the archetype. Seen from the input the softmax weights are those of the random energy model, and memorization is its freezing transition, at \beta\sqrtN(1-r^2)\simeq\sqrt2\ln M ; it is measurable on a head whose keys are examples as the number of examples carrying weight. On MNIST the diagnostic follows the random-energy prediction; on fashion-MNIST it detects sub-clusters in the classes, and the best accuracy lies beyond the memorization threshold, where the softmax reads out a local sub-prototype.
[LG-119] Learning Changes the World and Then What Was Learned has Changed: Training and Deploying Neural Networks in the Control Loop
链接: https://arxiv.org/abs/2610.22084
作者: Vittorio Lippi
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Neural networks offer critical adaptability for controlling complex systems where nominal modeling proves intractable. However, when offline-trained neural networks are deployed within closed-loop control systems, their actions fundamentally alter the state distribution upon which they were trained. This paper introduces a unified taxonomy to categorize these sequential failure modes: the initial Closed-Loop Identification Error, the deployment-induced Action Error, and the continuous Tracking Error. Through discrete-time sensitivity analysis, this work formalizes how offline-learned parameters locally guarantee a trajectory divergence upon deployment, degrading performance due to local loss geometry. To mitigate this distribution shift, we propose actionable design policies centered on time-scale separated online adaptation. These mechanisms are validated through ablation studies on a nonlinear Duffing oscillator, demonstrating that bounded continuous learning successfully mitigates trajectory divergence whereas static deployments fail, offering concrete guidelines for practitioners deploying machine learning components in safety-critical feedback systems.
[LG-120] Rethinking Successive Rejects Algorithms
链接: https://arxiv.org/abs/2610.22051
作者: Mohamed Khair,Junpei Komiyama
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:In the best-arm identification problem, an algorithm adaptively selects arms to find the arm of the largest mean reward. Successive Rejects (SR) is a simple fixed-budget best-arm identification algorithm that eliminates one candidate arm at each phase. At first glance, SR seems to have limited adaptivity, since it follows predetermined phase lengths. However, for Gaussian arms with common variance, SR’s worst-instance error exponent normalized by the popular complexity measure H_1§ is optimal up to a constant factor, which partly explains why SR’s performance is competitive compared with state-of-the-art adaptive sampling algorithms. In this paper, we extend the framework of SR by optimizing its predefined schedule of phase lengths. We formulate phase-schedule design as an optimization problem for the normalized error exponent. Then, we introduce a version of SR with the optimized schedule, which we call H_1 -balanced Successive Rejects ( H_1 -SR). This schedule is optimal within the class of SR-type algorithms with deterministic equal-count schedules. For Gaussian arms with common variance, H_1 -SR achieves at least 11/12 of the optimal normalized rate over all adaptive algorithms for every K\ge2 , and this approximation ratio converges to one as 1-O(1/\log K) as K\to\infty . This is a significant improvement over original SR, whose corresponding fraction of the optimal normalized rate converges to 1/4 . Finally, experiments on ten synthetic instances and two dataset-derived Gaussian benchmarks compare the optimized schedule with original SR at finite horizons. They show substantial gains on many instances while also exposing a regime in which more aggressive early elimination is costly.
[LG-121] Missed Detections and False Alarms in a CNN Technosignature Detector on Breakthrough Listen Data
链接: https://arxiv.org/abs/2610.22050
作者: Omer Imran
类目: Instrumentation and Methods for Astrophysics (astro-ph.IM); Machine Learning (cs.LG); Popular Physics (physics.pop-ph)
*备注: 17 pages, 14 figures. Code and data: this https URL
Abstract:Radio technosignature searches rely on machine learning detectors to shrink the candidate pool before a human looks at it. They are validated on held-out data from the training distribution, which excludes the cases where they are most likely to fail: a technosignature has unknown morphology, and terrestrial interference is unbounded in variety. I trained a standard competition detector once, froze it, and probed it with 15,000 injections across ten morphologies and five amplitudes, placed into real Green Bank backgrounds. Two directions were tested: genuine signals with morphologies other than a straight narrowband line, and interference resembling the training data. Recall did not follow the predicted ordering, and it fell as signals brightened, even for a needle of the kind the detector was trained on (from 0.590 to 0.090). One interference morphology reached a false positive rate of 0.627, comparable to the detector’s own baseline recall of 0.5525 on genuine signals. Gradient-weighted Class Activation Mapping (Grad-CAM) showed attention along the injected path in one missed family by a pre-specified test, and visually in a second. A pre-specified, time-averaged ON-OFF filter removed false alarms and genuine detections at the same rate. The detector reads only the ON panels. Since the only evidence distinguishing a genuine candidate from terrestrial interference lies in the discarded OFF panels, it cannot tell if a signal originates from the target star; it can only tell if it resembles the training data. This explains the false alarm directly; the missed signals reflect what the detector learned to recognise, which reading the OFF panels would not by itself change.
[LG-122] Designing Experiments for External Decision Makers
链接: https://arxiv.org/abs/2610.22032
作者: Sahel Iqbal,Tom Rainforth
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Theoretical Economics (econ.TH)
*备注:
Abstract:Standard Bayesian experimental design (BED) methods assume the experimenter alone will act on the data, choosing the best action given their own beliefs and goals. In practice, data are often acted on by external parties whose procedures the experimenter cannot change and may not know, particularly when the data are released publicly. To extend BED to such settings, we introduce usage-aware experimental design, a framework in which an adaptive design policy is trained to minimise the usage risk: the experimenter’s expected loss when one or more external rules act on the collected data. These rules may be non-Bayesian or black-box, and when several may apply, we either average over them or take the worst case. We show that, unlike in standard BED, more informative data can be harmful under such rules, so the usage risk cannot in general be expressed as an expected uncertainty reduction; our framework thus strictly generalises standard BED. We then use it to tackle a range of problems that are difficult to address with existing approaches.
[LG-123] Deep Learning of Topological Equivalence Classes
链接: https://arxiv.org/abs/2610.21981
作者: Linden Disney-Hogg,Max Zuev
类目: Algebraic Geometry (math.AG); Machine Learning (cs.LG)
*备注:
Abstract:We demonstrate an application of deep learning to the problem of identifying whether two generating vectors correspond to topologically equivalent finite group actions on Riemann surfaces, showing how architectures developed for semantic understanding have use for pure mathematical problems. In doing so we discuss how desired algebraic structure influences model design, and explore which structural properties of topological equivalence classes the model has successfully learnt.
[LG-124] SelectPFN: Amortized Bayesian Variable Selection for Sparse Nonlinear Regression with Prior-Data Fitted Networks
链接: https://arxiv.org/abs/2610.21950
作者: Bin Luo
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Tabular foundation models based on prior-data fitted networks (PFNs) predict on a new data set in a single forward pass, but they cannot report which inputs their predictions depend on. We propose SelectPFN, a PFN that performs Bayesian variable selection and prediction jointly for sparse nonlinear regression. In one forward pass it returns posterior inclusion probabilities for every variable, and its predictions are computed from sampled subsets only, so excluded variables provably cannot affect them; we show this is exactly what the Bayesian model prescribes. The network is pretrained once on synthetic tasks from a meta-prior built around a group spike-and-slab sparse-input neural network. Column-permutation equivariance combined with random interleaving of original and knockoff columns yields knockoff statistics with exact finite-sample false discovery rate control, even when the meta-prior is wrong. We develop theory for amortized selection: a Bayes-regret bound for the amortized median probability model, a characterization showing that the information needed to detect a variable grows like (2/\lambda)\log(1/\omega) in the prior weight \omega of structures that can use it, where \lambda\le1 is an efficiency, and a negative result showing that average-case training does not transfer to frequentist guarantees at a fixed truth. On out-of-prior nonlinear designs with p up to 500, SelectPFN improves the area under the precision-recall curve by 0.23-0.28 over the lasso, BART, DART, LassoNet and a group-penalized sparse neural network, at a false discovery proportion of about 0.07 and 0.04 seconds per data set. The advantage persists on real covariate matrices, and, as the theory predicts, detection of interaction-only variables rises monotonically with their prior weight.
[LG-125] Same Spectrum Harder First-Order Training: Phase-Locking in Single- versus Multi-Qubit Re-Uploading Circuits
链接: https://arxiv.org/abs/2610.21778
作者: Michael Poppel,David Bucher,Maximilian Zorn,Markus Baumann,Sebastian Wölckert,Claudia Linnhoff-Popien,Philipp Altmann,Jonas Stein
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG)
*备注:
Abstract:Angle-encoded variational quantum circuits implement truncated Fourier series. For one-dimensional inputs, a single-qubit re-uploading (serial) circuit represents every function that a multi-qubit circuit with the same encoding budget E can; we find that this exchange of width for depth is not free in first-order training. Because its state lives on the Bloch sphere, all parameters of the serial circuit move the state, at each input, along one complex direction, which we call \emphphase-locking. We prove that this costs no Fourier direction but leaves no direction to spare: however many parameters it has, the serial circuit has at most 2E+2 tangent directions, one more than its 2E+1 Fourier coefficients need, while a two-qubit control with the same encoding budget and model class generically has at least 6E+6 . The cost in training is measured against this control, as the number of runs out of 100 that reach the optimum within a fixed step budget. On an exact loss with flat-spectrum targets the gap grows with E ; at E=32 the serial circuit converges in 3, 53 and 83 runs under gradient descent, Adam and Levenberg–Marquardt, the control in 99, 100 and 100. On three real-world series, at matched parameter counts, the serial circuit reaches the least-squares optimum in at most 9 runs under Adam and in 32 to 93 under Levenberg–Marquardt, the control in at least 98 and in all 100; its held-out error is up to about 1.5 times that of the control under Adam, but no larger in the typical run under Levenberg–Marquardt.
[LG-126] Robust Covariance Estimation for Entry-wise Contaminated and Partially Observed Data via Expectation Maximization Algorithm
链接: https://arxiv.org/abs/2610.21709
作者: Matys Savidan,Mohammed Nabil El Korso
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Robust covariance estimation under entry-wise contamination and missing data is challenging in various signal processing applications. Under the Student- t model, observation-level reweighting induces a uniform attenuation across samples, which can lead to cumulative information loss even when most coordinates remain informative. The alternative Student- t^* model enables coordinate-wise robustness, but its E-step is not available in closed form. We extend inference for the Student- t^* model to incomplete observations through a stochastic approximation EM (SAEM) algorithm operating on a reduced latent space: stochastic updates of the coordinate-wise scales are combined with Rao-Blackwellized Gaussian conditional moments, which avoids explicit sampling of latent Gaussian variables while preserving a likelihood-based framework. Experiments on synthetic and real sensor data show a lower Fisher-Rao error than Gaussian, Student- t , imputation-based and detect-then-estimate estimators at moderate to high outlier rates under diffuse contamination.
[LG-127] Conditional Mean Flows: Few-Step Densities for Simulation-Based Inference
链接: https://arxiv.org/abs/2610.21657
作者: Jonas Arruda,Jan Hasenauer,Stefan T. Radev
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Flow matching has emerged as a powerful generative family for simulation-based inference (SBI). However, sampling and density evaluation require numerical integration at inference time. This makes posterior sampling in neural posterior estimation costly. More importantly, it can make repeated likelihood evaluation prohibitively expensive and impractical for Markov chain Monte Carlo (MCMC) sampling, which may require thousands of repeated density evaluations. Here, we develop conditional mean flows for posterior and likelihood approximation, combining few-step sampling with tractable density evaluation. We show that the mean flow objective identifies conditional flow maps and that independent masking enables direct estimation of marginal likelihoods. We evaluate both posterior and likelihood estimation across several tasks and show that conditional mean flows retain the flexibility of flow-based density estimators while significantly reducing inference time. In particular, single-step conditional mean flow density evaluation yields estimation error comparable to 200-step flow matching while enabling about 200-fold faster downstream MCMC sampling.
[LG-128] Reducing the Cost of Qudit State Preparation by Learning a Sparse Structure for SNAP Gates
链接: https://arxiv.org/abs/2610.21557
作者: Marco Pennazio,Leonardo Ignazio Pagliochini,Riccardo Pellini,Maurizio Ferrari Dacrema,Paolo Cremonesi
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG)
*备注: 16 pages, 3 figures, 4 tables, accepted at International Workshop on AI for Quantum and Quantum for AI (AIQxQIA 2026), co-located with AIxIA 2026, Perugia, Italy
Abstract:Preparing quantum states in bosonic qudits with SNAP–Displacement sequences requires both optimizing the gate parameters and deciding which Fock levels should be controlled by each SNAP gate. A fully parameterized ansatz assigns an independent phase to every Fock level in every gate, leading to a number of phases that grows steeply. This constitutes a problem for short to medium term systems as controlling many phases requires complex control pulses which are subject to increased control errors. Previous work reduced this cost by restricting the trainable phases to predefined sparse patterns, such as a diagonal or a grid, which however may not be optimal for a specific target state. In this work, we address this limitation by introducing a method that learns which phases are needed together with their values, using a differentiable mask. A sparsity penalty encourages the removal of unnecessary phases, while temperature annealing gradually turns the masks into binary decisions. We compare the proposed method with fully parameterized and predefined sparse ansatzes on Fourier and Haar-random target states. The evaluation considers both the minimum number of phases required to reach different infidelity thresholds and the overall trade-off between fidelity and phases required, measured through Pareto frontiers and hypervolume. The results show that learning the location of the phases together with their values provides competitive or improved fidelity–resource trade-offs, although the advantage depends on the target state and qudit dimension.
[LG-129] wo Sample Sizes in Test-Time LLM Ensembling: Buffered ERM and Minimax Theory
链接: https://arxiv.org/abs/2610.21363
作者: Zhi Zhang,Arash A. Amini
类目: atistics Theory (math.ST); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Test-time scaling can improve the performance of large language models by generating multiple answers to a question and aggregating them. Weighted ensembles extend this idea by drawing answers from several models, with the weights determining how often each model is used. Learning these weights from a training dataset involves two sample sizes: Q labeled questions and N generations per model and question. We study their interplay through the clean risk, the population error of the ensemble’s infinite-generation voting rule. Under a margin condition with exponent \beta , we show that empirical risk minimization (ERM) with a calibrated margin buffer achieves excess risk of order Q^-1/2+N^-\beta/2 , up to logarithmic factors, for fixed numbers of models and candidate answers. To investigate the optimality of this dependence on Q and N , we analyze a simplified submodel with two models and two possible answers per question. We characterize its minimax excess risk up to constants by a Hellinger modulus for a binomial-mixture experiment. Polynomial approximation and spectral analysis reveal three regimes. When questions are relatively scarce, the Q^-1/2 term governs learning and the general buffered ERM is optimal. In an intermediate regime, pooling generation counts across questions improves upon estimating each answer distribution separately, with only logarithmic gains from additional questions. With sufficiently many questions, the minimax excess risk reaches an N^-\beta identification floor that persists even with infinitely many questions. The submodel’s lower bounds also apply to general ensembles, matching the buffered-ERM upper bound up to logarithmic factors when questions are scarce and in parts of the intermediate regime. Questions remain about the transitions between regimes and the construction of matching pooled estimators for general ensembles.
[LG-130] Dimension-Free PAC-Bayes Generalization for Deep Neural Networks via Dirichlet Processes
链接: https://arxiv.org/abs/2610.21340
作者: Sumit Vashishtha
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注:
Abstract:Classical PAC-Bayesian generalization theory places prior distributions over Euclidean parameter space \Theta \subseteq \mathbbR^D , causing relative entropy to scale with network dimension D . For modern overparameterized neural networks where D \gg n , this parameter-counting dependency yields vacuous generalization bounds ( 1,000% ). In this work, we resolve this barrier by shifting prior specification from internal parameters to the non-parametric space of data distributions \mathcalM_1(\mathcalZ) via a \textbfDirichlet Process (DP), \mathcalDP(\alpha, F_0) , inducing parameter distributions through push-forward measures under the learning algorithm. By the Data Processing Inequality, parameter-space relative entropy is strictly bounded by the data-space DP divergence, which we evaluate in exact closed form using Digamma functions. The resulting PAC-Bayesian certificate is strictly invariant to network parameter count D , width, and depth. On intrinsic d^* -dimensional submanifolds, optimal partition scaling achieves the minimax non-parametric rate \mathcalO(n^-1/(d^*+2) \sqrt\ln n) . Across vision architectures scaling up to 4.2 million parameters ( D = 7,850 to D = 4,203,522 ) on the full MNIST benchmark, while classical weight-space bounds scale past 940% (becoming vacuous), our data-space DP-PAC-Bayes certificate remains stable, non-vacuous, and certified at \textbf 8.8% – 9.4% .
[LG-131] Pathwise Shrinkage and Probabilistic Calibration of the Multiplicative Half-Cauchy Process
链接: https://arxiv.org/abs/2610.21136
作者: Siddhesh Kulkarni
类目: Methodology (stat.ME); Machine Learning (cs.LG); Probability (math.PR); Statistics Theory (math.ST)
*备注: 6 pages, 1 figure, 1 table. R package: this https URL . Software DOI: https://doi.org/10.5281/zenodo.23355811
Abstract:We characterize the multiplicative half-Cauchy process through its log-scale random-walk representation, establishing shrinkage, expansion, and critical regimes, almost-sure summability, local rebounds, and effective dimension. We also derive a semi-analytic calibration rule linking interpretable prior probability statements to its scale parameter.
[LG-132] Score Mismatch and Set-Constrained Sampling in Zero-Shot Conditional Discrete Diffusion
链接: https://arxiv.org/abs/2610.21061
作者: Renxiang Huang,Yuchen Liang,Yingbin Liang,Ness Shroff,Lifeng Lai
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 46 pages, 4 figures
Abstract:Discrete diffusion models have become a powerful framework for modeling complex discrete distributions across a range of applications. While theoretical convergence guarantees depend on consistency between the target and training distributions, such a case rarely holds in practice, making score mismatch a central challenge in discrete diffusion sampling. This is particularly relevant to zero-shot conditional generation, where a model trained on the unconditional distribution is directly reused at inference time to sample from a conditional target distribution. In this paper, we provide the first theoretical analysis for score-mismatched discrete diffusion models and derive a general KL divergence upper bound that clearly identifies the additional bias caused by such a mismatch. Our analysis covers partial-observation constraints, including text infilling, sentence completion, and more general coordinatewise set constraints. Specifically, we analyze the classical Conditional SEDD sampler and show that it can incur a bias term that scales with the dimension, vocabulary size, and compatibility. Based on this analysis, we propose a Randomized Hard Clamping (RHC) sampler for coordinatewise set constraints. We show that the assignment probabilities can be expressed exactly by the unconditional scores with absorbing rate. Finally, numerical and real-data experiments demonstrate the effectiveness of the proposed RHC sampler.
[LG-133] Efficient Agnostic Learning of Univariate Densities Under Local Differential Privacy
链接: https://arxiv.org/abs/2610.21033
作者: Mohammad Afzali,Hassan Ashtiani,Sivan Sabato
类目: Machine Learning (stat.ML); Cryptography and Security (cs.CR); Data Structures and Algorithms (cs.DS); Information Theory (cs.IT); Machine Learning (cs.LG)
*备注:
Abstract:We propose a general framework for non-interactive locally differentially private (LDP) learning of structured univariate distributions that can be approximated by piece-wise polynomials. This gives the first sample-efficient LDP learners for several distribution classes, including mixtures of log-concave distributions, such as Gaussian mixture models (GMMs). These learners also provide agnostic guarantees and are computationally efficient. We also prove a lower bound for learning GMMs under LDP, improving on the standard pure-DP packing lower bounds of Bun et al. [BKSW19]. To obtain our general learning framework, we develop an efficient non-interactive LDP algorithm for synthetic data generation over large domains, which may be of independent interest. We then use it together with a deterministic adaptation of the non-private piecewise polynomial learner of Acharya et al. [ADLS17].
[LG-134] On the Last-Iterate Convergence of Clipped Gradient Methods NEURIPS2026
链接: https://arxiv.org/abs/2610.21026
作者: Zijian Liu
类目: Optimization and Control (math.OC); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: NeurIPS 2026
Abstract:In stochastic (strongly) convex optimization, establishing last-iterate convergence guarantees for first-order methods has recently garnered increasing attention, as this not only deepens the theoretical understanding of these algorithms but also aligns more closely with practice. Notably, people have demonstrated that one of the most famous, simple, and popular methods, Stochastic Gradient Descent ( \mathttSGD ), provably achieves last-iterate convergence. However, the existing results may have limited applicability, since they typically rely on strong conditions on the gradient noise (e.g., an exponentially decaying tail) rather than the more realistic heavy-tailed noise. Unfortunately, when faced with heavy-tailed noise, \mathttSGD is known to exhibit undesirable behavior or even fail to converge. To address the challenge of heavy-tailed noise, people have proposed \mathttClipped\text-\mathttSGD , an algorithm that combines \mathttSGD with a simple mechanism, gradient clipping. Although \mathttClipped\text-\mathttSGD performs well in practice, its last-iterate convergence property remains highly underexplored. In this work, we prove the first optimal last-iterate rates in high probability for \mathttClipped\text-\mathttSGD under heavy-tailed noise, thereby closing a gap in the literature.
[LG-135] Acceleration and its Limits in Gaussian Variational Inference
链接: https://arxiv.org/abs/2610.21020
作者: Jaesung R. Park,Uijeong Jang,Yifan Chen
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: 52 pages, 9 figures
Abstract:We study whether the classical momentum-based acceleration that improves the condition-number dependence of gradient descent from \kappa to \sqrt\kappa for strongly convex, smooth objectives extends to Gaussian variational inference (GVI) for strongly log-concave, log-smooth target distributions. With exact Gaussian expectations, a convex composite formulation of GVI yields the classical accelerated \sqrt\kappa convergence rate. However, when only pointwise derivatives are available, we construct a family of strongly log-concave targets for which any randomized algorithm requires \kappa^1-o(1) queries to achieve a fixed objective accuracy, even when the initial objective gap is bounded independently of \kappa . Thus curvature bounds alone cannot guarantee uniform accelerated total-query complexity. We then ask which additional structure permits acceleration and give positive examples: under a Hessian-Lipschitz assumption, variance-reduced stochastic methods yield accelerated guarantees when the Hessian variation is sufficiently small. Finally, experiments on real datasets show substantial empirical speedups from acceleration across a broad range of methods.
[LG-136] he Blume-Capel Model as a ϕ6 Field Theory: A Monte Carlo and Flow Model Study
链接: https://arxiv.org/abs/2610.20982
作者: Ejaaz Merali,Ryan Larson,Ehsan Khatami,Richard T. Scalettar
类目: atistical Mechanics (cond-mat.stat-mech); Disordered Systems and Neural Networks (cond-mat.dis-nn); Machine Learning (cs.LG); High Energy Physics - Lattice (hep-lat)
*备注:
Abstract:We report Monte Carlo simulations of a continuous spin variant of the Blume-Capel model. We compute the phase diagrams in the crystal-field-temperature plane, as a function of the softness of the spins. We show that normalizing flows, a class of generative models capable of direct sampling without autocorrelation, are effective samplers across the phase diagram, including in the first-order transition region and in the vicinity of the tricritical point. Finite-size scaling analyses of flow samples locate the first- and second-order transitions as well as the tricritical point with accuracy comparable to canonical and extended-ensemble Monte Carlo.
[LG-137] wo-Stage Score Estimation for Group-Specific Diffusion Models
链接: https://arxiv.org/abs/2610.20979
作者: Huitao Yang,Oscar Hernan Madrid Padilla
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注:
Abstract:Generative modeling across related distributions raises a question: how can shared structure support accurate generation for a target distribution with limited observations? We develop a statistical framework for diffusion transfer that analyzes a shared pooled score together with a target-specific offset. Under Besov regularity of the pooled density and a stable hierarchical composition model for the offset, we derive convergence rates that separate the costs of learning these components and identify when a simpler offset yields a faster risk-bound rate than separate fitting. Under suitable boundary and regularity assumptions, we prove that the offset remains bounded and jointly regular as diffusion time approaches zero. These results allow the lower time cutoff to vanish polynomially and provide total-variation guarantees for diffusion models as density estimators. Synthetic simulations and image benchmarks provide a proof of concept for the framework.
[LG-138] Cross-subject decoding of imagined speech from scalp EEG: a leakage-controlled evaluation on three public corpora
链接: https://arxiv.org/abs/2610.20954
作者: Blessing Isoyiza Adeika,Onyema Osuagwu
类目: ignal Processing (eess.SP); Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注: 23 pages, 5 figures, 10 tables plus supplementary tables. To be submitted to Journal of Neural Engineering
Abstract:Objective. A decoder of imagined speech that works for a new user without calibration would make a non-invasive speech interface practical. We tested whether imagined-speech content recorded with scalp electroencephalography (EEG) transfers across people, and examined why accuracy within a single person remains low. Approach. We analyzed three public corpora, ds003626, KaraOne and Track 3 of the 2020 International Brain-Computer Interface Competition, under a protocol that enforces split integrity by assertion and reports exact binomial thresholds, group tests and bootstrap confidence intervals. We trained decoders on each participant and tested them on every other, compared representational geometry across participants without a classifier, applied subject-level alignment, and tested five explanations for the within-subject ceiling. Main results. Across 482 ordered pairs of participants, no pair survived correction for multiple comparisons; mean pair-transfer accuracy lay within 0.006 of chance on every corpus, while every within-subject reference lay above it. Imagery against pre-stimulus rest generalized at about 0.61 after artifact and elapsed-time controls, but its decoder labelled the interval preceding imagery as imagery as often as imagery itself, so the contrast reflects a post-stimulus state, not imagery. Within subject, only the competition corpus decoded above threshold for every participant (0.3147; chance 0.200); syllable count explained little of this, but its frequency profile failed an artifact audit. Significance. In these corpora, imagined-speech content did not transfer between people through any representation we examined. Trial count per class was the only variable that ordered the corpora, which favors larger single-condition datasets over further model development.
[LG-139] Ranked Right Sized Wrong: Ensemble Spread Fails Silently at Streamflow Extremes NEURIPS2026
链接: https://arxiv.org/abs/2610.20947
作者: Justin Gallant,Tyler Wilson
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Geophysics (physics.geo-ph)
*备注: Accepted at the Tackling Climate Change with Machine Learning workshop at NeurIPS 2026. 13 pages, 2 figures, 6 tables. Code: this https URL
Abstract:Deep learning models, particularly Long Short-Term Memory networks, are increasingly used for operational streamflow forecasting, with ensemble spread providing a signal of predictive uncertainty. As climate change increases the frequency of hydroclimatic extremes, these models will increasingly operate under flow conditions that are rare in their historical records. We therefore evaluate whether ensemble spread remains informative under increasing flows, examining both its magnitude relative to error and its ability to rank errors. Prediction error is approximately 2\times the corrected spread at ordinary flows and up to 4\times at the highest flows, with the discrepancy increasing with flow (log-log slope 1.30 , 95% CI: [1.20,1.41] ). Under-dispersion is strongest in observed extreme flows beyond the training record, reaching an 18\times discrepancy. This is pervasive and not explainable by conditional mean bias. Area Under the Sparsification Error, a standard distribution-free metric of uncertainty ranking, describing how well a model knows when it is uncertain, indicates near-oracle ranking for ensemble spread ( 0.071 ), yet ordering by predicted discharge alone nearly matches it. Informative uncertainty ranking can coexist with severe under-dispersion. In deployment, error ranking cannot be verified, yet the spread is available, and it is relied on as the uncertainty signal under increasingly extreme conditions, while systematically understating risk, most severely for the extreme events that matter most.
[LG-140] Stochastic Gradient Descent: A Nonlinear Time Series Perspective
链接: https://arxiv.org/abs/2610.20894
作者: Jiaqi Li,Zhipeng Lou,Stefan Richter,Wei Biao Wu
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注: 87 pages, 8 figures, accepted to JRSSB
Abstract:This paper investigates the statistical behavior of Stochastic Gradient Descent (SGD) with constant step size under the framework of iterated random functions. Compared to previous studies establishing the convergence of SGD in probability measure, e.g., Wasserstein distance, our approach directly provides the convergence in Euclidean distance by showing the Geometric Moment Contraction (GMC) of SGD. This new convergence can address the non-stationarity of SGD due to fixed initial points and can provide a more refined asymptotic analysis of SGD. Specifically, we prove quenched central limit theorem, invariance principle and \alpha -stable convergence for averaged SGD (ASGD) regardless of initialization. Furthermore, we provide a novel perspective to understand the impact of step sizes in SGD by studying its derivative with respect to the step size. The existence of stationary solutions for the derivative processes is shown, enabling the use of interpolation to recover the full-trajectory of SGD. Subsequently, we utilize multiple step sizes and show an enhanced Richardson-Romberg extrapolation with improved bias representation, which brings ASGD estimates closer to the global optimum. Finally, we propose a new online inference method and a bias-reduced variant for the extrapolated ASGD. Empirical confidence intervals are constructed and the coverage probabilities approach the nominal level as n grows.
[LG-141] Flexible Doubly Robust Proximal Causal Inference for Censored Outcomes
链接: https://arxiv.org/abs/2610.20893
作者: Dhanajit Brahma,Ruolin Meng,Lawrence Carin,Ricardo Henao
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Methodology (stat.ME)
*备注: 34 pages
Abstract:Estimating causal effects on right-censored time-to-event outcomes is challenging under unmeasured confounding. Proximal causal inference addresses unmeasured confounding through proxy variables, but existing proximal survival methods rely largely on restrictive parametric bridge models. We introduce Proximal Survival Bridge learning (PSB), a flexible doubly robust proximal framework for censored outcomes. Building on existing proximal identification of treatment-specific survival curves, the framework yields outcome-bridge (OB), treatment-bridge (TB), and doubly robust (DR) estimators, with two flexible bridge learners: PSB-K, a kernel estimator with a low-rank time decomposition, and PSB-N, a neural estimator trained by a projected maximum-moment-restriction loss. Unlike existing flexible proximal learners for scalar outcomes, PSB directly handles censoring and survival-curve targets. On a standard SEM benchmark, a proximal semi-synthetic ACTG study, and the real-world Framingham study, PSB-K performs strongest on SEM and ACTG, whereas PSB-N is the most stable on Framingham. Additional results with modern causal survival baselines, weakened or invalid proxies, and heavier censoring assess robustness across challenging settings.
[LG-142] Learning Warm Starts for the Simplex Method from Measured Solver Cost
链接: https://arxiv.org/abs/2610.20892
作者: Felix Parker,Danielle A. Ripsman,Emmett Springer,Kimia Ghobadi
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:
Abstract:Learned warm starts for the simplex method build starting bases using predictions of which variables will be basic at optimality. Prior work has shown that starting bases predicted by graph neural networks can sharply reduce iteration counts, but the time saved may fall well short of the iterations saved. On 25 classes of linear programs (LPs), we find that a strong supervised basis predictor removes two thirds of the dual simplex’s iterations but only a third of its time. Two effects explain the gap: the iterations removed by a good start tend to be cheap, and closeness to an optimal basis does not rank starts by cost. We prove that two starts with equal overlap with an optimal basis and equal pivot counts can differ in linear-algebra work by a factor that grows with the square root of the number of rows. Empirically, across the starts of one LP, the pivot count explains 71% of the variance of log solve time, while the CPU instruction count explains 98.5%. We introduce CostStart, which uses supervised prediction for the basis and trains a small policy on measured instruction counts to choose, per LP, how many predicted columns to use, which algorithm and pricing rule to run, and whether to warm-start at all. On unseen LPs of the training classes, including LPs four times the training size, CostStart roughly halves the solve time relative to the solver’s default start and is 1.4 to 2.2 times as fast as two published methods retrained on the same data.
[LG-143] Machine-learning-based neutrino-event generation constrained by experimental measurements
链接: https://arxiv.org/abs/2610.20870
作者: Jose Luis Bonilla,Krzysztof M. Graczyk,Rwik Dharmapal Banerjee,Beata E. Kowal,Hemant Prasad,Jan T. Sobczyk
类目: High Energy Physics - Phenomenology (hep-ph); Machine Learning (cs.LG); High Energy Physics - Experiment (hep-ex); Nuclear Experiment (nucl-ex); Nuclear Theory (nucl-th)
*备注: 5 pages, 6 figures and supplement
Abstract:We present a generative surrogate for charged-current (\nu_\mu)-nucleus scattering that can be calibrated to measured differential cross sections without event-level experimental samples. Starting from a theoretical interaction model, we fine-tune the surrogate to a MicroBooNE measurement and show that this also improves the description of observables not used in the fit. The calibrated model can also be applied to measurements from different neutrino fluxes without retuning, while retaining event-level predictions of outgoing-muon kinematics. Our results show that experimental cross sections can directly constrain generative event models while preserving their applicability across different observables and beam configurations.
[LG-144] SNOAR: Self-training with Noise filtering Observation adding and Remixing for Speech Denoising Without Clean References ICASSP2027
链接: https://arxiv.org/abs/2610.20863
作者: Samuele Cornell,Matthew Maciejewski
类目: Audio and Speech Processing (eess.AS); Machine Learning (cs.LG); Sound (cs.SD); Signal Processing (eess.SP)
*备注: submitted to ICASSP 2027
Abstract:Unsupervised speech denoising methods such as MixIT and noisy-target training can be trained directly on noisy recordings, avoiding the sim-to-real gap of synthetic training data, but remain well below supervised performance. Teacher–student self-training, as in RemixIT, narrows this gap by remixing a teacher’s estimates into new mixtures that supervise a student; however, since the student is trained only toward the teacher’s estimates, teacher errors, in particular over-suppressed speech, are never recovered and propagate as the teacher is updated from the student. We propose SNOAR, a self-training framework that corrects the teacher’s estimates before use: the noise leaked into its speech estimate is removed by differential noise filtering (DNF), and part of the noisy observation is added back in to restore suppressed speech and attenuate artifacts. We show that the two corrections depend on each other, as without DNF the loop collapses to the identity mapping. Compared to RemixIT, SNOAR is more stable along training and outperforms it in signal-based, perceptual, and word error rate metrics, reaching 97% of the SI-SDR improvement of a fully-supervised topline on WHAM!; on real CHiME-3 recordings, it improves every metric of three different teachers, while RemixIT degrades their perceptual quality.
[LG-145] me-partitioning Gaussian Process Mixture Model
链接: https://arxiv.org/abs/2610.20860
作者: Guilherme Piantino,Hedibert F. Lopes
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We propose the Time-partitioning Gaussian Process Mixture Model (TPGPMM), a Bayesian nonparametric framework for identifying persistent regimes and modelling nonstationary dependence in time series. The model combines an adaptive Pitman–Yor prior on contiguous segmentations with regime-specific Gaussian processes, without imposing Markovian within-regime dependence. Learning the Pitman–Yor discount parameter allows adaptation to homogeneous and heterogeneous regime lengths. A split–merge Metropolis-within-Gibbs sampler with exact Gibbs boundary updates jointly learns change points and model hyperparameters, quantifying uncertainty over segmentations and within-regime dynamics. With an RBF kernel and a fixed, conservative observation-noise level, TPGPMM favours sparse segmentations that distinguish persistent shifts from transient shocks. Simulations demonstrate recovery of the regime structure. Applications to the CBOE Volatility Index (VIX) and the 10-year United States Treasury yield identify regimes aligned with major market and monetary-policy events. Variants that estimate observation noise improve predictive accuracy relative to the main specification, but can produce less sparse or less event-aligned segmentations. A variant with regime-specific means fits better than a fitted stationary Gaussian process and forecasts about as accurately. These results highlight a trade-off between interpretable regime identification and predictive accuracy. Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG) Cite as: arXiv:2610.20860 [stat.ML] (or arXiv:2610.20860v1 [stat.ML] for this version) https://doi.org/10.48550/arXiv.2610.20860 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-146] Complete or Sparse: A Tale of Two Identifiabilities NEURIPS2026
链接: https://arxiv.org/abs/2610.20851
作者: Junze Zhou,David Klindt
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 28 pages, 9 figures. Accepted at NeurIPS 2026
Abstract:Understanding when neural networks learn interpretable representations is central to mechanistic interpretability. We study this through identifiability: when can supervised learning recover the true latent variables underlying the data? We introduce two data generating processes distinguished by their causal direction where either labels cause latents (generative) or latents cause labels (discriminative). When the representation has at least as many dimensions as the latent space, we prove that the optimal encoders are linear in the latents. Conversely, when dimensions are insufficient, identifiability generically fails and encoders become nonlinear. We then show that additional constraints yield recovery of individual latent variables, not just linear mixtures. In the generative setting, non-negative activations and energy efficiency force each neuron to encode exactly one latent, providing a precise account of selective grandmother cell-like responses. In the discriminative setting, when latents are sparse and the true measurement matrix satisfies the spark condition, the learned encoder inherits compressed sensing structure, enabling sparse dictionary learning to perfectly recover individual latents. This constitutes, to our knowledge, the first identifiability result for undercomplete nonlinear independent component analysis. Our framework unifies nonlinear ICA, compressed sensing, and neural network interpretability, delineating exactly when representations transition from nonlinear distortions to linear representations and, under additional constraints or steps, to clean, axis-aligned codes. Theoretical predictions are validated on synthetic benchmarks and on activations drawn from vision models, large language models, and macaque inferotemporal cortex.
[LG-147] Linear Independence of Polynomial Compositions with Bounded Input Degree
链接: https://arxiv.org/abs/2610.20825
作者: Xiaotian Chang,Yacoub Hendi
类目: Algebraic Geometry (math.AG); Machine Learning (cs.LG)
*备注: 13 pages
Abstract:We prove the weak composition-independence conjectures of Kohn, Marchetti, Massarenti and Mella. The proof uses deformations of rational curves and a single kernel of the linearized equation. Riemann-Roch gives a contradiction from tangent fields with controlled poles. By settling these conjectures, we prove that a neural network with a general polynomial activation function of sufficiently large degree is identifiable.
[LG-148] Ergodic topology sampling toward the continuum with Stochastic Normalizing Flows
链接: https://arxiv.org/abs/2610.20747
作者: Claudio Bonanno,Andrea Bulgarelli,Elia Cellini,Alessandro Nada,Dario Panfalone,Davide Vadacchino,Lorenzo Verzichelli
类目: High Energy Physics - Lattice (hep-lat); Statistical Mechanics (cond-mat.stat-mech); Machine Learning (cs.LG); High Energy Physics - Phenomenology (hep-ph)
*备注: 13 pages, 10 figures
Abstract:In this work we provide numerical evidence that topological freezing in \mathrmSU(3) Yang-Mills theory can be mitigated with a flow-based approach, as tested at lattice spacings as fine as 0.03~fm. We do so by analyzing the scaling of the costs of a computational strategy based on the combination of open boundary conditions and Stochastic Normalizing Flows. We show that the integrated autocorrelation time of the squared topological charge and the statistical uncertainty on the topological susceptibility remain under control throughout the explored parameter range with currently available computational resources. This marks an important step toward the use of even more efficient flow-based architectures for ergodic sampling of topology in state-of-the-art lattice gauge theory calculations.
[LG-149] On the Stability and Convergence of Flow Map Training
链接: https://arxiv.org/abs/2610.20674
作者: Ishan Nath,Tom Stanic,Avishek Joey Bose,Gauthier Gidel,Quentin Bertrand
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Flow maps are a recent method of probabilistic generation that amortize diffusion trajectories into one-step samplers, unlocking efficient generation. While progress has been made in improving their empirical performance, there is still a lack of understanding behind the mechanisms underlying their success. Specifically, many flow map learning methods employ stop gradient operators as a crucial part of their training target. Adaptive weighting also drastically improves the generation capabilities of flow maps. Although many interpretations have been proposed to support the inclusion of these heuristics, none have analyzed the impact of these assumptions on the training dynamics of flow maps themselves. In this work, we address this issue by providing grounded theory explaining why such tricks are required. Analyzing the functional gradient flow induced by the MeanFlow and Lagrangian losses, we show their convergence to the true flow map, settling the validity of these objectives. For finite updates, we highlight a severe roadblock to the usage of these losses in practice: the possible unboundedness of the update. Adaptive normalization not only provides bounded updates, but, combined with a specific weighting, additionally provides convergence guarantees towards a neighborhood of the solution. On synthetic data, we show that our functional gradient flow analysis exactly describes the true training dynamics for flow maps. Experiments on standard image datasets demonstrate that a proper reweighting scheme is essential for strong empirical performance, which is only achieved for a specific range of reweighting strengths.
[LG-150] Graph Additive Models disentangling feature topology and interactions
链接: https://arxiv.org/abs/2610.20662
作者: Clara Alvarez-Rodriguez,Steve Azzolin,Vincenzo Marco De Luca,Filippo Maria Bianchi,Andrea Passerini,Antonio Longa
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Graph Neural Additive Networks (GNANs) have emerged as effective, interpretable-by-design architectures for graph learning. They achieve transparency by combining per-feature shape functions with simple distance-based reweighting. However, GNANs have three limitations. First, node-wise relevance scores aggregate information from all graph nodes, allowing irrelevant nodes to inherit spurious relevance from relevant ones. Second, topology is represented only through shortest-path distances, so structurally distinct graphs may share identical distance profiles. Third, feature and distance information are combined multiplicatively, entangling feature, distance, and interaction effects in a single rescalable term. We introduce AMIGA, which represents topology through a vocabulary of mined motifs and decomposes each node contribution into three additive channels: feature, structure, and feature-structure interaction. The channels are centered around a graph-balanced training baseline, making the decomposition well-defined and unique under the chosen reference distribution. Across ten graph classification and regression benchmarks, AMIGA outperforms GNAN on most datasets and narrows the gap with standard black-box GNNs. Because each channel is centered and assigned to a single node, AMIGA produces node-local explanations that reconstruct the model output exactly and support interpretation at multiple levels of granularity.
[LG-151] Latent Annealed Posterior Sampling with a Variance-Preserving Bridge for Sparsely Observed Random Fields
链接: https://arxiv.org/abs/2610.20659
作者: Jiankun Lei,Tian Qi
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 41 pages, 7 figures
Abstract:Posterior inference for random fields from sparse observations presents two major challenges: accurately characterizing complex spatial uncertainty and preventing conditional sampling costs from growing substantially as the spatial grid is refined. To address these challenges, we propose latent annealed posterior sampling (LAPS). LAPS first maps the random field into a latent space with reduced dimensionality, while a coordinate decoder composed of Branch and Trunk networks enables field evaluation at arbitrary spatial locations. Flow matching is then used to learn the latent prior distribution. Based on this learned prior, posterior sampling is performed entirely in the latent space, with the observation likelihood incorporated progressively through an annealing procedure. To control local covariance inflation, we introduce a variance-preserving (VP) bridge that allocates a covariance budget between the transported latent state and a Gaussian perturbation. Numerical experiments demonstrate that LAPS accurately approximates posterior distributions, preserves distinct posterior modes, and reduces the dependence of conditional sampling costs on spatial resolution. Across two Gaussian benchmarks, LAPS attains small errors against exact posterior references, with sliced Wasserstein distances only 2 to 3 percent higher than the reference values obtained from exact posterior samples. In a bimodal posterior problem, LAPS reduces the sliced Wasserstein distance from 0.2693 to 0.1499 and consistently recovers both modes across all test scenarios. For a two dimensional Navier-Stokes random field, LAPS substantially improves marginal posterior calibration, while its evaluation time remains nearly constant as the output grid is refined from 64^2 to 512^2. These results demonstrate that LAPS achieves a favorable tradeoff among posterior accuracy, uncertainty calibration, and computational efficiency.
[LG-152] A polylogarithmic bound for list replicability
链接: https://arxiv.org/abs/2610.20657
作者: Ari Blondal,Hamed Hatami,Pooya Hatami,Chavdar Lalov,Sivan Tretiak
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Combinatorics (math.CO)
*备注:
Abstract:Zhang [Zha26] recently posted a manuscript announcing that the list-replicability number of every class of M\ge 2 binary concepts on N\ge 2 domain points is O(\log^2 M\log N) , which would be the first polylogarithmic upper bound on this quantity. In this note, we present a short, self-contained proof of this bound, following the high-level approach of the manuscript but with considerably simpler arguments.
[LG-153] Global-in-time Spectral Duhamel-Picard Method for Post-Exposure Bake Simulation and Physics-Informed Neural Operators for EUV Lithography
链接: https://arxiv.org/abs/2610.20655
作者: Doyun Kim,Mircea Dusa,Werner Gilijns,Philippe Leray
类目: Computational Physics (physics.comp-ph); Machine Learning (cs.LG); Applied Physics (physics.app-ph)
*备注: 21 pages, 9 figures
Abstract:We present a global-in-time spectral Duhamel-Picard method with polynomial expansions for post-exposure bake (PEB) simulation. The coupled acid-quencher reaction-diffusion equations are reformulated as nonlinear Volterra integral equations using the Duhamel principle. Full trajectories are represented by finite polynomial expansions, with temporal coefficient fields determined over the bake interval by Anderson-accelerated Picard iteration, thereby avoiding sequential time marching. In 3-D PEB simulations, we employ Chebyshev and Legendre temporal bases while evaluating the Duhamel integrals using the corresponding Gaussian quadrature rules. We demonstrate that the proposed method simulates a large domain of 1200\times1200\times100 voxels on the order of one minute. The representation is further incorporated into a physics-informed neural operator (PINO) with a mixed-transform variant of a factorized Fourier neural operator, which is trained without precomputed solution trajectories. Together, the integral formulation and its PINO extension enable accelerated, time-marching-free prediction of concentration fields at arbitrary times within the bake interval.
[LG-154] pcNN-SN: a physics-enhanced predictor-corrector neural network algorithm coupled with the discrete ordinates method for neutron transport equations
链接: https://arxiv.org/abs/2610.20619
作者: Cong Liu,Junxia Wei,Shuai Wang,Heng Yong
类目: Computational Physics (physics.comp-ph); Machine Learning (cs.LG)
*备注: 35 pages, 17 figures
Abstract:Super-resolution machine learning methods hold promise for addressing the high computational costs associated with high-dimensional features in traditional transport simulations. This paper proposes a physics-enhanced neural network coupling algorithm for discrete ordinates (SN) transport calculations. Based on high- and low-accuracy data, we develop a predictor-corrector neural network (pcNN) to predict correction factors for coarse-grid solutions. Based on transport equations, local physical features containing subgrid-scale information are selected as network inputs. Dimensionless transformation of physical quantities reduces the input space dimensionality and narrows parameter distributions, facilitating network training. Physical constraints are incorporated into the network model, and the corrected solution retains its original conservation properties. pcNN and SN transport sweeps are internally coupled, and numerical solutions are corrected online. Numerical results demonstrate that this algorithm can effectively improve coarse-grid solutions to achieve a level of accuracy comparable to that of fine-grid results. The physics-based network design enables the model to demonstrate a certain degree of reusability under different computational conditions. Compared to mesh refinement, the pcNN-SN algorithm reduces the computation time for the three-dimensional test problem by more than 60%.
[LG-155] Iterative Neural Variational Solver of Committor Function in Under-damped Langevin Dynamics
链接: https://arxiv.org/abs/2610.20617
作者: Tiejun Li,Zhaoqi Liu,Xiang Zhou
类目: Computational Physics (physics.comp-ph); Machine Learning (cs.LG)
*备注:
Abstract:The committor function describes the probability that a stochastic dynamical system, while transitioning between two metastable states, will first reach one state rather than the other. It is a central object in transition path theory, providing the accurate reaction rate and the characterization of the transition mechanism through reactive probability flux and the transition state ensemble. While the numerical methods for this function are well studied for over-damped Langevin dynamics where the friction damping coefficient goes to infinity, its computation in the general under-damped case remains an open challenge due to the lack of a variational structure in the underlying backward Kolmogorov equation. This restricts the quantitative investigation of the impact of the finite damping coefficient on rare transitions in real systems. In this work, by decomposing the infinitesimal generator of the underdamped Langevin dynamics into a self-adjoint and a skew-adjoint part, we develop an iterative scheme in which each subproblem has a variational form and can be solved efficiently using neural networks. Fundamentally, this approach can be viewed as a generalized Concus-Golub-Widlund (CGW) splitting method and each subproblem has a variational form, which can be solved efficiently. This iterative scheme also has the flexibility of introducing a preconditioning operator associated with a new overdamped Langevin dynamics, which naturally leads to the design of enhanced adaptive sampling for the subproblems. We validate the efficiency and accuracy of our method on benchmark systems including a one-dimensional double-well potential, the Muller-Brown potential, and alanine dipeptide. The proposed iterative variational framework provides a practical and extensible numerical tool for probing rare transition events in Langevin dynamics across varying damping coefficients.
[LG-156] hermodynamically Consistent Machine Learning for Fracture Prediction in Heterogeneous Materials
链接: https://arxiv.org/abs/2610.20614
作者: Muhamad Alim,Xiaolong He,Mei-Na Tsao,Chi-Hua Yu,J. S. Chen
类目: Computational Physics (physics.comp-ph); Materials Science (cond-mat.mtrl-sci); Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
*备注:
Abstract:Understanding fracture behavior is critical for assessing the structural integrity and longevity. Highly time-consuming physics-based fracture simulations motivate the development of machine-learning-based fracture evolution prediction with surrogate models. However, most existing machine-learning models for fracture prediction operate as black boxes and require large training datasets to generalize reliably. In this work, Thermodynamically Consistent Latent Dynamics Identification (tLaSDI) for parametric systems is introduced to predict fracture patterns and evolution in heterogeneous materials. The model learns the thermodynamic state variable in a lower-dimensional latent space using parametric GENERIC Formalism-Informed Neural Networks (pGFINNs) that enforce key thermodynamic principles, such as total energy conservation and entropy generation. This model embeds physics into the neural network architecture using a hard-constraint method that inherently complies with the underlying thermodynamic principles. The proposed tLaSDI-pGFINNs is further enhanced by a Geometry-Conditioned Autoencoder in learning the geometry configuration for effective geometry-guided learning. The energy and entropy evolution histories demonstrate that the model preserves the thermodynamic constraints on unseen testing samples. The results further show that tLaSDI-pGFINNs can more accurately capture both fracture patterns and the subsequent fracture evolution than the purely data-driven Autoencoder Long Short-Term Memory (AE-LSTM) model.
附件下载
