本篇博文主要内容为 2026-10-01 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 (2026-10-01)
今日共更新1284篇论文,其中:
- 自然语言处理共180篇(Computation and Language (cs.CL))
- 人工智能共395篇(Artificial Intelligence (cs.AI))
- 计算机视觉共225篇(Computer Vision and Pattern Recognition (cs.CV))
- 机器学习共431篇(Machine Learning (cs.LG))
- 多智能体系统共27篇(Multiagent Systems (cs.MA))
- 信息检索共30篇(Information Retrieval (cs.IR))
- 人机交互共53篇(Human-Computer Interaction (cs.HC))
多智能体系统
[MA-0] Belief-Aware Multi-Agent Path Finding under Map Uncertainty
【速读】：该论文旨在解决多智能体路径规划（Multi-Agent Path Finding, MAPF）中因环境动态变化导致的不确定性问题，尤其关注静态障碍物分布未知且存在空间相关性的情形。传统方法假设所有障碍物信息在规划前已知，或仅基于直接观测进行局部重规划，无法利用观测点附近区域的空间依赖关系来推断未观测区域的可通行性，从而可能导致后续出现高成本的路径重规划。其核心挑战在于如何有效利用空间相关性，通过一次观测推测邻近区域的潜在障碍，以提前规避未来冲突。本文提出MAGIC（Multi-Agent Gaussian belief Inference for Coordination）框架，采用高斯马尔可夫随机场（Gaussian Markov Random Field, GMRF）与高斯信念传播（Gaussian Belief Propagation）构建共享的可通行性信念模型，实现对未观测区域可通行性的在线近似推断，并据此生成考虑绕行代价的路径成本，供标准MAPF求解器使用。该方案的关键创新在于将空间相关性建模融入信念更新机制，显著提升了路径规划的前瞻性和鲁棒性。实验结果表明，MAGIC在96.3%的基准测试实例中均优于现有方法，适用于包含最多800名智能体的大规模场景，验证了其在复杂动态环境下的高效性与可扩展性。
链接: https://arxiv.org/abs/2609.40269
作者: Viraj Parimi,Shao-Hung Chan,Han Zhang,Jingkai Chen,Brian Williams
机构: Symbotic Inc.(Symbotic公司); Massachusetts Institute of Technology(麻省理工学院)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA); Robotics (cs.RO)
备注: Under review
Abstract:Multi-Agent Path Finding (MAPF) aims to find collision-free paths for multiple agents in a shared environment. Classical MAPF assumes that all static obstacles are known in advance, but real-world environments can change unexpectedly due to fallen objects, spills, or other local disturbances. When such changes are spatially correlated, an observation can inform traversability estimates beyond the observed location. Prior approaches address uncertainty in traversability through contingent plans or replanning based on direct observations, but do not leverage this spatial dependence to infer the traversability of nearby unobserved locations. As a result, they cannot use one observation to anticipate nearby unobserved obstacles that may cause costly rerouting later. We focus on Belief-Aware MAPF, where map discrepancies are fixed during execution but initially unknown, and observations can be informative beyond the observed location. We propose Multi-Agent Gaussian belief Inference for Coordination (MAGIC), a framework that updates a shared belief about traversability online based on agents’ observations. MAGIC uses a Gaussian Markov Random Field and Gaussian Belief Propagation to approximately infer traversability and construct detour-aware costs for standard MAPF planners. Our experiments on MAPF benchmarks show that MAGIC reduces the executed sum of costs compared to existing approaches on 96.3% of instances, across several planner families and teams of up to 800 agents, demonstrating its applicability to large-scale MAPF problems.
[MA-1] Passive Stiffness Shaping in Cable-Suspended Aerial Manipulation via Movable Compliant Anchors
【速读】：该论文旨在解决缆索悬挂式空中操作中负载所呈现的被动机械响应机制尚不明确且未被系统性利用的问题。其核心解决方案在于将空中飞行器视为可移动的柔顺锚点，提出一种考虑重力影响的准静态理论，用于预测与调控悬吊负载的被动笛卡尔刚度。该理论适用于由多架飞行器通过张紧、直线、不可伸长的缆绳连接至质点负载的情形；在选定的重力平衡构型下，每条缆绳中的飞行器柔顺性与横向缆绳几何柔顺性呈串联关系，而各腿刚度则并联作用于负载。当飞行器锚点行为具有各向同性时，每条腿等效为一条虚拟的单向弹性缆绳，揭示出轴向-横向刚度分解由平衡张力决定。这一分析结果构建了从指令锚点构型到被动负载刚度的非线性映射关系，其微分形式支持通过重新配置锚点实现局部约束保持下的刚度精准调控。为进一步验证理论在实际动态环境中的适用性，研究设计了一个包含非线性飞行器控制、弹性阻尼缆绳及环境接触的动态刚体验证框架，用以评估解析模型假设之外的条件下所推导刚度特性仍具备预测能力的边界条件与程度。
链接: https://arxiv.org/abs/2609.40102
作者: Antonio Franchi,Amr Afifi
机构: University of Twente (特温特大学); Sapienza University of Rome (罗马第一大学); Utrecht University (乌得勒支大学)
类目: Robotics (cs.RO); Multiagent Systems (cs.MA); Systems and Control (eess.SY)
备注:
Abstract:Cable-suspended aerial manipulation offers a lightweight architecture for cooperative transportation and physical interaction, yet the passive mechanical response perceived at the load remains insufficiently understood and systematically exploited. This work interprets aerial vehicles as movable compliant anchors and develops a gravity-aware quasi-static theory for predicting and shaping the passive Cartesian stiffness of a suspended load. The formulation applies to an arbitrary number of aerial vehicles connected to a point load by taut, straight, inextensible cables. At a selected gravity-loaded equilibrium, aerial-anchor compliance and transverse cable geometric compliance combine in series within each leg, while the leg stiffnesses act in parallel on the load. For isotropic aerial-anchor behavior, each leg is exactly equivalent to a virtual unilateral elastic cable, revealing an axial–transverse stiffness decomposition governed by the equilibrium tension. These results define a nonlinear map from commanded-anchor configuration to passive load stiffness, whose differential enables local constraint-preserving shaping through anchor repositioning. A dynamic rigid-body validation framework with nonlinear vehicle control, elastic-damped tendons, and environmental contact is defined to assess when and to what extent the derived stiffness remains predictive beyond the assumptions of the analytical model.
[MA-2] Mid-Harness: Scaling Actions Between Model and Harness for Terminal Agents
【速读】：该论文旨在解决终端代理（terminal agent）在执行动作时存在的可靠性问题：尽管生成式模型能够产生看似合理的动作指令，但若指令本身存在缺陷（如错误的包安装命令），可能对环境造成不可逆的负面影响，从而阻碍后续任务进展。传统方法中，模型生成的动作一旦提交即被执行，缺乏对动作有效性的验证机制，导致即使模型具备生成更优动作的能力，也无法实现可靠执行。为此，论文提出一种名为Mid-Harness的新框架，其核心在于在模型与执行环境之间的“钩子”（harness）边界处引入测试时计算资源的分配策略——通过采样多个候选动作并进行验证，仅选择经过验证的最优动作执行，从而提升动作的可靠性与整体轨迹成功率。关键创新在于：有效的验证能力是释放动作采样潜力的核心，当验证器具备足够判断力时，即便使用同一生成器，也能从大量候选动作中筛选出真正可行的选项；实验表明，采用更强的验证器（如GPT-5.6 Sol）可使Pass@1指标从基线50.00%提升至68.03%，而将强验证器的推理结果蒸馏至同规模生成器后进一步优化性能。此外，结合动作采样与轨迹扩展的协同策略，在更低估算令牌成本下实现了更高成功率。研究揭示，动作层面的测试时扩展（action scaling）是终端代理中最具前景的计算资源分配方向。
链接: https://arxiv.org/abs/2609.39982
作者: Minki Kang,Ryo Hachiuma,Shaokun Zhang,Subhashree Radhakrishnan,Yonggan Fu,Jindong Jiang,Mingjie Liu,Ehsan Hosseini-Asl,Yi Dong,Yu-Chiang Frank Wang,Byung-Kwan Lee
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: Project page: this https URL
Abstract:Terminal agents act through stochastic model generations, yet the ability to generate a useful action does not ensure its reliable execution. A poor command (e.g., wrong package install) can change the environment in ways that hinder subsequent progress, even when the model could generate a better alternative. We investigate whether allocating test-time compute at the model-harness boundary can improve action reliability and trajectory success, and what makes this allocation effective. To study these questions, we introduce Mid-Harness, which samples and verifies candidate actions before forwarding one for execution, while keeping the generator and harness unchanged. With a TMAX-9B generator, more action sampling yields little benefit under weak verification, whereas a capable verifier can exploit useful alternatives from the same generator. On TerminalBench-Lite, a GPT-5.6 Sol verifier raises Pass@1 from 50.00% for the base agent to 68.03% with 8 sampled actions. When the same TMAX-9B model serves as the verifier, pairwise verification performs best among the evaluated verification mechanisms. Distilling responses from the stronger verifier into TMAX-9B further improves Pass@1, while leaving the action generator unchanged. With TMAX-9B on TerminalBench-Lite, combining action and trajectory scaling reaches higher success at lower estimated token cost than generating more trajectories alone. Mid-Harness also improves performance across additional models, benchmarks, and harnesses. These findings identify action scaling as a promising target for test-time compute scaling in terminal agents.
[MA-3] AIMS: An Agent ic AI Framework for Sim-to-Real Multi-Modal ISAC
【速读】：该论文旨在解决多模态集成感知与通信（ISAC）系统在真实场景中部署时面临的“仿真到现实”（sim-to-real）迁移能力不足问题。现有数据驱动的ISAC模型高度依赖标注的真实世界数据，而生成合成数据虽可缓解数据获取压力，但其仿真管道若未与目标部署环境在场景、感知、无线链路及学习配置上保持一致，则会导致跨域性能下降。为此，本文提出一种基于智能体（agent）的人工智能框架——AIMS（Agentic AI for Sim-to-Real Multi-modal ISAC），其核心在于通过自然语言输入的部署请求自动推导并协调执行特定于任务的仿真-现实配置。关键解决方案包括：采用双智能体架构，其中场景构建智能体基于共享物理状态生成地理精准且时序同步的感知与无线数据记录，场景理解智能体则根据任务需求配置多模态信号与专家混合（MoE）学习机制，支持零样本推理或少样本适应；同时，结构化领域知识驱动依赖关系感知的规划过程，验证证据支持反馈驱动的决策修正。实验结果表明，相较于基准仿真与融合方法，AIMS在真实世界DeepSense 6G数据集上显著提升了车辆检测与波束预测性能，并在独立编排评估中验证了其在任务解析、依赖推理及反馈重规划方面的高正确性，充分体现了结构化知识与反馈机制对提升系统自主性与迁移鲁棒性的关键作用。
链接: https://arxiv.org/abs/2609.39964
作者: Yijie Bian,Kai Zhang,Wei Guo,Zixin Wang,Shenghui Song,Jun Zhang,Khaled B. Letaief
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Multi-modal integrated sensing and communication (ISAC) enables environmental perception and reliable connectivity for intelligent wireless networks. Data-driven multi-modal ISAC models depend heavily on annotated real-world data to learn relationships across sensing and wireless observations, thereby constraining scalable deployment. Although synthetic data generation reduces the burden, adapting existing simulation pipelines to a target deployment requires consistent scene, sensing, wireless, and learning configurations, while mismatches among these coupled components impair sim-to-real transferability. To address the challenge, we propose an agentic artificial intelligence (AI) framework for sim-to-real multi-modal ISAC, named AIMS. Given a natural-language deployment request specifying the target task, deployment conditions, and real-data budget, AIMS derives a deployment-specific sim-to-real configuration and coordinates its execution to produce a deployment-specific task model. A two-agent architecture coordinates scene construction with task learning. A scene construction agent generates geographically grounded, synchronized sensing and wireless records from shared physical states, while a scene understanding agent configures task-relevant modalities and mixture-of-experts (MoE) learning for zero-shot inference or few-shot adaptation. Structured domain knowledge guides dependency-aware planning, while validation evidence supports feedback-driven revision of affected decisions. Experiments on the real-world DeepSense~6G dataset demonstrate improved vehicle detection and beam prediction over the considered simulation and fusion baselines. A separate orchestration benchmark evaluates task interpretation, dependency reasoning, and feedback-driven replanning across diverse deployment requests, showing improved plan correctness with structured domain knowledge and validation feedback.
[MA-4] Solving Multi-Agent Sokoban via LaCAM
【速读】：该论文旨在解决多智能体协同规划中的核心挑战，即多智能体推箱子（Multi-Agent Sokoban）问题的可扩展性求解难题。该问题因智能体数量增加导致分支因子急剧上升、任务分配与无碰撞路径规划需协同处理而极具挑战性。其解决方案的关键在于利用多智能体路径规划（Multi-Agent Pathfinding, MAPF）领域的最新进展，设计出一种名为Sokoban-LaCAM的高效可扩展规划器。该方法通过将MAPF作为基础计算原语，实现了对包含数十个智能体和箱子实例的有效求解，同时在理论上保证了求解的完备性与最终最优性，为更广泛的群体自动化问题提供了可行的求解范式。
链接: https://arxiv.org/abs/2609.39889
作者: Keisuke Okumura
机构: National Institute of Advanced Industrial Science and Technology (AIST)(日本先端科学技術産業研究機構)
类目: Multiagent Systems (cs.MA)
备注:
Abstract:Sokoban, a puzzle game in which an agent pushes boxes onto unlabelled target locations in a grid world, is a long-standing benchmark planning problem. While it is easy to see the connection to practical applications such as warehouse logistics with autonomous forklifts, its multi-agent counterpart has remained underdeveloped. This is because Multi-Agent Sokoban is substantially more difficult due to factors specific to multi-agent planning, such as the rapidly growing branching factor as the number of agents grows and the need to handle integrated task assignment and collision-free pathfinding. In this paper, we show that a scalable planner for Multi-Agent Sokoban can be designed by leveraging recent advances in multi-agent pathfinding (MAPF). Specifically, our Sokoban-LaCAM efficiently solves instances involving tens of agents and boxes while preserving both completeness and eventual optimality guarantees. This provides evidence that MAPF can serve as a powerful primitive for solving broader collective automation problems.
[MA-5] Safety of Latent Communication in Multi-Agent Systems
【速读】：该论文旨在解决多智能体系统中通过隐空间（latent space）进行通信时潜在的安全对齐失效问题。尽管基于隐空间的通信可降低文本交互带来的令牌消耗、计算开销与延迟，但研究发现，即使在无害条件下训练的可微调连接（lightweight trainable links），仍可能使原本安全对齐的智能体表现出更高的有害合规性（harmful compliance）。其解决方案的关键在于：利用强化学习设计一种攻击策略，该策略在不依赖恶意目标响应的情况下，同时优化有害合规性与良性任务性能，从而有效放大隐链接中的安全风险。实验表明，该方法可在三种通信拓扑结构和四个安全基准上将平均有害合规得分从27.9提升至76.9。此外，通过调整奖励函数以促进更安全行为，该框架还能修复已被攻陷的链接，显著降低各类攻击下的有害合规性，且无需更新底层智能体。研究强调，安全对齐必须将多智能体系统视为整体进行考量，而非仅关注单个智能体的对齐状态。
链接: https://arxiv.org/abs/2609.39788
作者: Muhammad Huzaifa,Sina Mavali,Thorsten Eisenhofer
机构: CISPA Helmholtz Center for Information Security(德国萨尔布吕肯信息安全亥姆霍兹中心)
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注:
Abstract:Latent communication enables multi-agent systems to exchange information directly in internal representation space, reducing the token, computation, and latency overhead of text-based communication. To this end, lightweight trainable links are introduced to map the sender’s representations into the receiver’s input space. In this work, we show that even benign link training can increase harmful compliance relative to text-based communication while the underlying safety-aligned agents remain unchanged. An attacker can amplify this effect by optimizing the links on harmful query–response pairs or poisoning otherwise benign training data. We further develop a reinforcement-learning attack that rewards harmful compliance alongside benign task performance without requiring harmful target responses. Across three communication topologies and four safety benchmarks, this attack raises the mean harmful-compliance score from 27.9 with benignly trained links to 76.9. Compared with direct supervised optimization, it also achieves higher average accuracy on two benign utility benchmarks. Adapting the rewards toward safer behavior also enables repair of compromised links, substantially reducing harmful compliance across all evaluated attacks without updating the agents. Overall, our results show that safety alignment requires considering the multi-agent system as a whole.
[MA-6] OverForge: Reasoning Through Strategies and Tactics Helps Cooperative Lifelong Adaptation
【速读】：该论文旨在解决协作式语言模型智能体在长时程任务中面临的协调难题，尤其是在动态环境与具有陌生行为规范的合作者共存时，如何实现稳定且灵活的协作。现有方法将观测直接映射为动作，缺乏对持久性协调策略与战术执行之间的解耦，导致适应性不足。其解决方案的关键在于提出一种无需训练的分层架构OverForge，通过将战略层面的职责分工与角色分配（role and division of labor）与战术层面的动作决策在个体私有的、依赖于合作方的世界模型中进行分离，从而实现高效协调。该架构的核心是元认知前额叶皮层模块（metacognitive Prefrontal Cortex Module），它通过构建“策略-动作”分支，利用前向模型模拟后果，并在具备信心时做出承诺，实现高层战略与底层行动的动态协同。实验表明，在OvercookedV2环境中，OverForge可达成7份汤品的协作目标，显著优于各基线模型（每者仅3份），并能保持已约定的角色分工，同时接纳陌生合作者提出的角色建议；消融实验与固定策略探测进一步验证了持久策略对战术适应性的引导作用，且跨周期记忆重启实验证明了伙伴知识积累对任务表现与合作者预测的持续支持，体现了该分层架构在持续学习与适应中的关键优势。
链接: https://arxiv.org/abs/2609.39727
作者: Oana Madalina Fron,Ojas Shirekar,Chirag Raman
机构: Tapri Lab, Department of Pattern Recognition and Bioinformatics; Delft University of Technology (代尔夫特理工大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multiagent Systems (cs.MA)
备注:
Abstract:Cooperative language-model agents must coordinate over long horizons and adapt to changing environments and to partners with unfamiliar conventions, yet existing agents map observations to actions without separating persistent coordination strategies from their tactical execution. We introduce OverForge, a training-free hierarchical architecture that separates strategic reasoning over roles and divisions of labour from tactical reasoning over actions within each agent’s private, partner-conditioned world model. A metacognitive Prefrontal Cortex Module couples the two levels by forming strategy-action branches, imagining their consequences with a forward model, and committing when confident. In OvercookedV2, OverForge delivers 7 soups in a connected kitchen versus 3 for each flat LLM baseline, retains agreed roles, and adopts roles proposed by unfamiliar partners. Ablations and a fixed-strategy probe show that persistent strategies guide tactical adaptation while each reasoning level contributes to coordination. Memory restarts show that cross-episode partner knowledge supports task performance and partner prediction, linking the hierarchy to continual adaptation.
[MA-7] Prediction is Better than Detection: Traffic Congestion Control using Drones
【速读】：该论文旨在解决多无人机团队在持续监测任务中，任务性能随机群规模增长的规律问题，尤其关注当感知数据驱动下游决策（如交通信号调控）而非仅用于观察时，这种性能扩展是否依然有效。其核心解决方案在于构建一个基于纳格尔-施雷肯贝格（Nagel-Schreckenberg）元胞自动机模型的多智能体仿真系统，模拟车辆行驶行为与无人机按轮询策略巡检交叉路口，并通过系统性地调整机群规模、交通密度和网络规模，评估检测率、检测延迟及预测率等关键指标。研究发现，当机群规模接近被监控交叉路口数量时，性能趋于饱和；由此提出适用于持续监测部署的通用机群配置准则。更重要的是，研究揭示：基于预测结果而非实时检测结果来调整信号灯，可使拥堵持续时间减少约一倍，表明在感知-行动链路中，机载预测能力的价值远超增加更多无人机所带来的收益。此时，预测准确性成为制约性能进一步提升的关键瓶颈，凸显未来研究应聚焦于机载推理能力优化，而非单纯扩大机群规模。
链接: https://arxiv.org/abs/2609.39637
作者: Samira Hayat,Christian Raffelsberger
机构: Lakeside Labs GmbH (Lakeside实验室)
类目: Robotics (cs.RO); Multiagent Systems (cs.MA)
备注: 7 pages, 9 figures
Abstract:A central question in deploying teams of mobile robots for persistent monitoring is how task performance scales with fleet size, and whether this scaling holds once sensing drives downstream action rather than mere observation. We study this question for a team of drones performing traffic-jam detection and prediction in a simulated road network, whose reports drive an adaptive traffic-signal controller in closed loop. We build a multi-agent simulation, with vehicles following Nagel-Schreckenberg cellular-automaton dynamics and drones patrolling junctions via a round-robin policy, and sweep fleet size, traffic level, and network size to evaluate detection rate, detection delay, and prediction rate. We show how performance plateaus for fleet size approximating the number of junctions being monitored, and offer a general fleet-provisioning rule for persistent-monitoring deployments. More significantly, adapting the signal on a predicted jam, rather than a detected one, roughly doubles the resulting reduction in jam duration, showing that the value of onboard prediction in a sensing-to-action pipeline can exceed the value of adding more robots. Prediction accuracy, not sensing coverage, is now the binding constraint on further improvement, pointing to onboard inference, not fleet size, as the more promising direction for future work.
[MA-8] Consensus and Factual Dynamics in Large Populations of Interacting Language Models
【速读】：该论文旨在解决大规模语言模型（Large Language Model, LLM）代理群体在动态交互中形成共识（consensus）的机制问题，特别是探究共识的形成如何依赖于代理间的交互结构。现有研究通常采用固定交互模式，忽略了交互拓扑对共识质量与收敛特性的影响。为此，论文提出RHEON框架——一个受物理自旋系统启发的演化模型，将来自单一冻结模型的代理群体建模为在逐步提升有效维度的交互几何结构（从一维环到全连接均值场图）上的O(n)自旋系统，并以采样温度T作为可调热扰动源，通过类似Glauber的异步动力学进行演化。通过对432种提示、群体规模、通信拓扑和采样温度组合的系统扫描，构建了包含470万条响应的进化语料库Eraclitus-4.7M。研究发现，代理在前几次更新轮次内即达到最强共识增益，且每个代理的邻居数量越多，平均收敛速度越快；更重要的是，共识是否指向事实正确或幻觉内容无法仅由初始状态预测，最小化幻觉的最优温度取决于耦合方式，因此常见的近似贪婪默认设置并非始终最安全。此外，语义一致性和事实收敛呈正相关，交互强化了这一关联，但从未足以使全体一致达成成为正确性的充分条件。该研究的关键在于通过类物理系统的演化框架揭示了交互结构与温度调控对共识质量的非平凡影响，为理解群体智能中的自发一致性提供了新的理论工具。
链接: https://arxiv.org/abs/2609.39211
作者: Emanuele Ricco,Elia Onofri,Vincenzo Sammartino,Roberto Di Pietro
机构: 未知
类目: Multiagent Systems (cs.MA)
备注:
Abstract:Large Language Model (LLM) agents are increasingly deployed as populations of interacting entities, in which consensus --agreement on a shared answer-- emerges as a collective, unengineered behaviour. Prior work on LLM consensus shows that agents can cross-verify their answers and converge towards more factual responses, treating agreement as a proxy for correctness. However, these studies usually fix a single interaction structure, leaving open how consensus depends on how agents interact. We address this gap by introducing RHEON, a physics-inspired framework that recasts a population drawn from a single frozen model as an evolving O(n) spin system on a ladder of interaction geometries of increasing effective dimension --from a 1D ring to a full-coupling mean-field graph-- with the sampling temperature T as the tunable source of thermal disorder, evolved through a Glauber-like asynchronous dynamics. Sweeping RHEON across 432 configurations of prompt, population size, communication topology, and sampling temperature yields Eraclitus-4.7M, a tagged evolutionary corpus of 4.7 million responses. We find that agents reach their strongest consensus gain within the first few update sweeps and that increasing the number of neighbours per agent accelerates convergence on average. We further show that whether a configuration settles on factually correct or hallucinated consensus is not predictable from its initial state alone, and that the hallucination-minimising temperature depends on how the agents are coupled, so the common near-greedy default is not automatically the safest. Finally, semantic agreement correlates positively with factual convergence, and interaction strengthens the association, yet never enough for unanimity to certify correctness.
[MA-9] RefCon: Iterative Refinement and Contrastive Memory Extraction for Context-Evolving Agent
【速读】：该论文旨在解决长时程智能体交互中产生的经验虽有价值但噪声较大的问题，尤其是在缺乏真实标签（gold labels）的情况下，如何高效地利用测试阶段的额外计算资源来提升记忆提取质量。其核心挑战在于：传统方法在引入新经验时需重新训练模型，成本高昂，且难以在无监督条件下持续优化记忆质量。为此，论文提出RefCon方法，其关键创新在于将**序列式自我精炼（sequential self-refinement）与并行式自我对比（parallel self-contrast）**相结合，通过内在一致性评估动态筛选和优化记忆内容，从而在不依赖真实标签的前提下，实现随测试时计算量增加而持续提升的记忆质量。实验结果表明，RefCon在AppWorld和BFCL-V3等多个上下文演化智能体框架上均取得显著且一致的性能提升，相对无扩展基线分别在ACE任务和ReMe任务上提升21.6%和16.6%，而其多样性导向变体DivCon在ReasoningBank上更是实现了35.5%的增益。此外，RefCon展现出良好的可扩展性与跨模型规模、跨任务（如软件工程）的泛化能力，甚至在某些场景下超越依赖真实标签的基线方法。分析进一步揭示，相比仅关注多样性的扩展策略，RefCon在准确率-令牌数权衡上更具优势，且随着轨迹数量增加仍能持续改进，未出现过早饱和现象。
链接: https://arxiv.org/abs/2609.39143
作者: Ubaidillah Ariq Prathama,Bo Liu,Yeo Boon Hong,Yu-Xuan Huang,Yangkai Ding,Tao Yu
机构: Huawei Technologies, Co., Ltd.(华为技术有限公司)
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注:
Abstract:Long-horizon agent interactions generate useful but noisy experience, and retraining models to absorb it is expensive. Context-evolving agents therefore need memory extraction methods that improve with more test-time compute without relying on gold labels. We propose RefCon, which combines sequential self-refinement with parallel self-contrast to extract higher-quality memories without gold labels. Evaluated on AppWorld and BFCL-V3 across multiple context-evolving agent frameworks, RefCon delivers strong and consistent gains, including relative improvements of 21.6% on ACE and 16.6% on ReMe over no-scaling baselines, while a diversity-focused variant (DivCon) achieves a 35.5% gain on ReasoningBank. RefCon consistently outperforms existing baselines without ground-truth labels, and generalizes across model scales and to software engineering tasks, where it surpasses even ground-truth baselines. We further analyze the accuracy-token trade-off and scaling behavior, showing RefCon maintains favorable efficiency and continues to improve as more trajectories are used, unlike diversity-only scaling which saturates earlier.
[MA-10] RSIGame: Autonomous Agent ic Game Development with Recursive Self-improvement
【速读】：该论文旨在解决生成式游戏在自动化开发过程中难以持续提升质量的问题，尤其是在突破“可玩版本”后仍面临泛化能力差、缺陷残留和对玩家交互适应性不足等挑战。现有方法依赖于简单的迭代优化，易导致过拟合特定测试用例，产生脆弱且不稳定的可执行游戏。其核心解决方案是提出一种具备递归自我改进能力的自主智能体式游戏开发框架——RSIGame。该框架的关键在于构建互补的局部与全局双循环机制：局部循环通过“探索-诊断-改进”闭环，系统性地发现并优先处理问题，结合动态演进的检查清单实现基于证据的精准修复；全局循环则长期追踪整体质量演化，保存最优检查点，并检测发展过程中的饱和或退化现象。此外，该框架进一步将成功的开发经验内化至生成模型中，通过训练实现知识沉淀。实验表明，在140个GameCraft-Bench任务、两种游戏引擎及五种生成器上，RSIGame在相同开发预算下均显著提升游戏质量；尤其在Qwen3.8-27B模型上，通过经验内化使生成效率提升11倍的同时，在Godot和Phaser引擎上的得分分别达到61.38和58.53，超越GPT-5.5的一次性生成表现。
链接: https://arxiv.org/abs/2609.39045
作者: Wenyi Wu,Minghao Fu,Jieyu You,Kun Zhou,Siqi Liu,Aayush Salvi,Yiheng Lin,Ce Zhang,Xiaohan Lan,Jiahui Zhu,Yujie Zhong,Qi She,Biwei Huang
机构: University of California San Diego(加州大学圣地亚哥分校); ByteDance Inc.(字节跳动公司); Carnegie Mellon University(卡内基梅隆大学)
类目: Computation and Language (cs.CL); Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注:
Abstract:Recent advances in large language models have made automatic game generation increasingly feasible, yet reliably improving generated games beyond a playable version remains challenging. Naive iterative refinement can easily overfit a small set of test cases, producing fragile games with unresolved bugs, missing behaviors, and poor generalization to broader player interactions. We introduce RSIGame, an autonomous agentic game development framework with recursive self-improvement. RSIGame organizes development into complementary local and global loops. Concretely, a local explore-diagnose-improve loop broadly explores the executable game, diagnoses and prioritizes discovered issues, and performs evidence-grounded revision, where an evolving checklist continually accumulates new testing and improvement guidance. A global loop tracks overall quality, preserves the best checkpoint, and detects saturation or regression over long-horizon development. Beyond test-time improvement, RSIGame further internalizes successful development experience into the generator through training. Across 140 GameCraft-Bench tasks, two game engines, and five generators, RSIGame consistently improves game quality under matched development budgets. Notably, experience internalization enables Qwen3.8-27B to reach 61.38 on Godot and 58.53 on Phaser, exceeding GPT-5.5 one-shot scores while reducing Qwen’s generation tokens by 11 times.
[MA-11] Fast and Scalable Multi-Agent Distribution Matching via Partitioned Optimal Transport
【速读】：该论文旨在解决多智能体系统中终端分布匹配（terminal distribution matching）的可扩展性问题。在大规模系统中，基于最优传输（Optimal Transport, OT）的方法虽能有效度量分布偏差并实现智能体到目标空间分布的分配，但全局离散传输计算复杂度随系统规模呈指数增长，成为实际应用的瓶颈。为此，论文提出一种基于最优传输的可扩展框架，其核心在于将智能体与目标样本按空间位置划分为对应块，并求解一系列局部小规模传输问题。在质量守恒条件下，所得受限耦合仍为全局问题的可行解，并提供瓦瑟斯坦（Wasserstein）代价的上界。局部分配结果生成有限时域智能体控制的目标位置，适用于线性与非线性动力学系统。通过交替执行局部分配与控制，建立了运输代理（transport surrogate）在循环间下降的保证。该框架在保持与瓦瑟斯坦目标严格理论关联的同时，实现了高效的终端分布匹配，其技术有效性通过仿真验证。
链接: https://arxiv.org/abs/2609.38960
作者: Kooktae Lee,Ruchika Singh
机构: Texas Tech University (德克萨斯理工大学)
类目: Multiagent Systems (cs.MA); Robotics (cs.RO); Systems and Control (eess.SY); Optimization and Control (math.OC)
备注:
Abstract:This paper presents a scalable optimal-transport-based framework for terminal distribution matching in multi-agent systems. While optimal transport provides a natural way to measure distributional mismatch and assign agents to a desired spatial distribution, global discrete transport can become computationally expensive for large-scale systems. We address this bottleneck by partitioning agents and target samples into spatially corresponding blocks and solving smaller local transport problems. Under a mass-balance condition, the resulting restricted coupling remains feasible for the global problem and provides an upper bound on the Wasserstein cost. The local assignments generate target locations for finite-horizon agent control, applicable to both linear and nonlinear dynamics. By alternating local assignment and control, we establish a cycle-to-cycle descent guarantee for the resulting transport surrogate. The proposed framework therefore enables scalable terminal distribution matching while retaining a rigorous connection to the Wasserstein objective. The technical soundness of the proposed results is validated through simulations.
[MA-12] Risk-Aware Adaptive Evaluation: Finding High-Impact Failures Under Limited Budgets NEURIPS2026
【速读】：该论文旨在解决交互式智能体（interactive agents）评估中资源分配效率低下的问题。由于智能体行为具有随机性，评估需依赖多次重复试验以保证结果可靠性，但关键失败事件稀少且其影响程度差异显著，而传统基准测试采用均匀分配预算的方式，导致高风险、高影响的场景与低风险操作（如只读查询）获得相同试验次数，造成资源浪费。论文的核心解决方案是将评估过程建模为一个序贯资源分配问题（sequential allocation problem），提出一种风险感知的上下文汤普森采样策略（risk-aware contextual Thompson Sampling policy），该策略结合预执行场景的上下文向量、固定的影响评分以及实际评估中观测到的失败结果，动态决定应优先执行哪些场景及其重试次数。实验基于70个τ-bench航空场景和824次真实记录的试验进行离线回放验证，结果显示：在仅50次试验（占全集6%）的小预算条件下，该方法可捕获86%由理想观察者发现的影响加权失败事件，远超均匀分配的25%；同时发现3.5倍更多高影响失败事件（215.4 vs. 62.2），单位预算下失败发现效率提升5倍，并将从未失败场景所消耗的预算从34%降至2.8%。分析进一步表明，该方法的优势在小预算时最为显著，随着预算趋近于完整数据集规模，优势减弱；在小预算下，场景上下文信息贡献更关键，而在中等预算下，基于后验分布的探索机制更具价值。因此，风险感知的自适应分配策略在评估预算最稀缺时发挥最大效能。
链接: https://arxiv.org/abs/2609.38914
作者: Priyanath Maji,Spandan Ghose Chowdhury
机构: Georgia Institute of Technology(佐治亚理工学院); Atlanta, GA 30332
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: Accepted to 40th Conference on Neural Information Processing Systems (NeurIPS 2026). Workshop: Evaluation of Interactive Agents
Abstract:Evaluating interactive agents is expensive. Agent behavior is stochastic, so reliability must be measured over repeated trials, but failures are rare and differ widely in how much they matter. Standard benchmarks spend this budget uniformly: a read-only lookup is sampled as often as an irreversible payment action. We instead formulate evaluation as a sequential allocation problem. Given a fixed trial budget and a set of scenarios whose failure behavior is unknown, which scenarios should be run, and run again? We propose a risk-aware contextual Thompson Sampling policy that combines a pre-execution scenario context vector and a fixed impact score with the failure outcomes observed during evaluation, and we test it by offline replay over 70 \tau -bench airline scenarios and 824 recorded trials. Our main result is at the smallest budget: with only 50 trials ( 6% of the corpus), the policy recovers 86% of the impact-weighted failures an oracle could find, compared to 25% for uniform allocation. It discovers 3.5\times more impact-weighted failures (215.4 vs. 62.2) with the same number of trials, delivers 5\times the discovery per dollar, and cuts the budget wasted on scenarios that never fail from 34% to 2.8% . The rest of our analysis demonstrates and qualifies this result: a budget sweep shows the advantage shrinks as the budget approaches the corpus size, and paired significance tests show that scenario context helps mainly at small budgets while posterior-based exploration helps at moderate ones. Risk-aware adaptive allocation therefore helps most exactly where evaluation budget is scarcest.
[MA-13] SkillSeek: Revisiting Agent Skill Retrieval at Marketplace Scale AACL
【速读】：该论文旨在解决大语言模型（LLM）代理在面对海量可复用技能库时的技能选择瓶颈问题。随着开源技能聚合规模突破23万项，如何高效筛选出最适配当前任务的技能已成为关键挑战，而非技能创作本身。现有主流方案依赖于由LLM驱动的检索循环机制，通过不断重写查询与优化候选集来完成选择，但该方法需在每次任务中消耗大量生成式AI（Generative AI）tokens，导致成本高昂。本文提出SkillSeek——一个基于标准信息检索（IR）范式的开源双阶段技能检索框架，其核心创新在于采用BGE-base双编码器作为第一阶段召回器，辅以小型交叉编码器进行精排，并通过MCP接口暴露服务。在SkillsBench基准上覆盖4×11组合配置的实验表明，SkillSeek在多数设置下达到甚至超越了刘等人提出的LLM中介检索环的性能表现，且仅需极低额外开销：仅使用BM25即可在四组设置中的三组实现不低于原方法的通过率，而引入小型交叉编码器即可补足第四组差距。这一结果揭示了第一阶段召回能力的上限是性能趋同的关键原因，同时将每轮试验的总成本从51.30美元降至27.54美元（接近无技能基线水平）。研究结论表明，在给定任务场景下，标准信息检索范式已可作为代理技能检索的强有力默认方案，而基于生成式AI的替代方法则更适合于确定性方法难以胜任的复杂情境。
链接: https://arxiv.org/abs/2609.38822
作者: Guanqun Yang,Wenlong Zhang,Tian Shi,Ping Wang
机构: Stevens Institute of Technology (斯蒂文斯理工学院); Independent Researcher (独立研究员)
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: Accepted at AACL-IJCNLP 2026. Code at this https URL
Abstract:Anthropic’s Agent Skills package reusable procedural know-how for an LLM agent into this http URL directories, and open-source aggregations have grown past 230,000 skills, making selection rather than authoring the bottleneck. The standing answer in the literature outsources selection to the agent itself: an LLM-mediated retrieval loop that rewrites queries and refines candidates inside the agent’s decision loop, paying LLM tokens on every task. We present SkillSeek, an open-source two-stage skill retriever built from the standard IR recipe (a BGE-base bi-encoder feeding a small cross-encoder, exposed over MCP). Across a 4 \times 11 grid of pool, backbone, and method on the 89-task SkillsBench benchmark, SkillSeek reaches observed parity with the LLM-mediated loop of Liu et al. at essentially no extra cost: plain bm25 alone records a pass rate at or above their refined loop on three of four settings, and a small cross-encoder covers the remaining difference on the fourth. A first-stage recall ceiling explains the pattern, and total per-trial spend drops from USD 51.30 to USD 27.54 (within fifty cents of the no-skill baseline). Under the SkillsBench tasks and OpenHands harness we tested, this positions the standard IR recipe as a strong default for agent-skill retrieval, with LLM-mediated alternatives a natural fit for cases where deterministic methods fall short.
[MA-14] Where Do Multi-Agent Systems Fail? Evidence-Grounded Diagnosis of Collective Mechanisms
【速读】：该论文旨在解决多智能体系统（multi-agent system）在产生正确答案时，难以判断其内部协作机制是否被正确执行的问题。尽管系统输出正确结果，但其背后的信息路由、接纳、存储与行动规则等集体机制可能已被破坏，而错误结果又往往无法明确指示具体哪个机制失效。为应对这一挑战，论文提出一种诊断契约（diagnostic contract），其关键在于将“机制违规”的定义与“执行记录能否证明违规”相分离：通过分析执行日志，判定某机制是否存在违规、被排除或无法确定，且一旦因删除记录导致结论变为“未知”，则不可逆向恢复为“存在”或“不存在”。研究通过重放四种机制的执行过程（正常、破坏后、修复后）验证了该方法的有效性，发现即使机制被破坏，系统仍常输出正确答案；大语言模型（LLM）诊断器从内部状态中检测到的违规远多于仅基于公开输出的分析，而通用提示词常错误宣称确定性结论，违反记录支持范围，而采用诊断契约则显著降低此类误判。此外，契约在独立开发系统中的窄范围机制上亦具适用性，但其诊断性能在不同系统间存在差异，表明单一基准上的良好表现不能保证跨系统的可迁移性。因此，论文强调，仅凭正确结果不足以评估机制有效性，必须依赖对执行过程的详细记录，且诊断能力不具备天然泛化性。
链接: https://arxiv.org/abs/2609.38761
作者: Zhengye Han
机构: New York University (纽约大学)
类目: Multiagent Systems (cs.MA)
备注:
Abstract:When a multi-agent system answers correctly, it is tempting to conclude that its agents shared, checked, and used information as intended. Yet a system can break one of its collective mechanisms, the rules that govern how agents route, admit, store, and act on shared information, and still return the right answer, while a wrong answer rarely reveals which mechanism failed. We ask what evidence from an execution is sufficient to conclude that a particular mechanism was violated. Our answer is a diagnostic contract, which separates what counts as a violation from which execution records can establish one, and concludes that a violation is supported, ruled out, or unknown; removing records can make this conclusion unknown but never reverse it. We test contracts for four mechanisms by replaying executions from the step where a mechanism acts, once unchanged, once with the mechanism broken, and once with it restored. Broken mechanisms often left the answer correct. An LLM diagnoser detected many more violations from internal records than from public outputs, yet with identical records a generic prompt often claimed certainty the records did not support, which prompts stating the contracts largely avoided. The contracts also applied, in narrow form, to mechanisms in independently developed systems, but a diagnostic behavior that was nearly perfect on our benchmark degraded on an independently developed workflow. A correct outcome is therefore no substitute for records of how collective mechanisms operated, and agreement on one benchmark does not show that a diagnoser transfers to another system.
[MA-15] HALO: Heterogeneous Allocation Via Localized Observations for the Vehicle Routing Problem
【速读】：该论文旨在解决大规模机器人编队在现实场景中部署时面临的可扩展性与实时性难题，尤其针对传统集中式车辆路径规划（Vehicle Routing Problem, VRP）算法在受限观测与通信范围的去中心化环境中失效的问题。现有方法虽能在理想条件下处理高达1000个任务的集中式路由问题，但无法适应实际应用中机器人仅能获取局部信息、通信受限的动态环境。为此，论文提出了一种名为异构局部观测分配（Heterogeneous Allocation via Localized Observations, HALO）的混合式解决方案，其核心在于将VRP分解为分配与路径规划两个阶段，并通过基于异构图神经网络（heterogeneous graph neural network）的局部消息传递机制，显式分离空间分布学习与任务-机器人匹配度建模，从而实现机器人端的实时、分布式决策。该方法在部分可观测、在线动态的VRP场景下显著优于启发式基线，且性能接近全局信息已知的离线最优解；同时在标准静态单源点VRP上亦超越专为该场景优化的先进架构达14.06%，并始终保持最快的执行速度，充分体现了其在大规模、实时系统中的部署潜力。
链接: https://arxiv.org/abs/2609.38760
作者: Andrew Meighan,Hyungsub Kim,Or Dantsker
机构: Indiana University Bloomington(印第安纳大学布卢明顿分校)
类目: Robotics (cs.RO); Multiagent Systems (cs.MA)
备注: 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works
Abstract:Scalable robotic fleets have become increasingly popular for various applications such as package delivery, warehouse management, and military operations. Prior fleet control algorithms solve centralized routing problems with up to 1,000 tasks in controlled environments, yet they fail to consider realistic constraints such as limited observation and communication ranges typical of decentralized fleets. Thus, deploying existing fleet control algorithms into real-world settings is currently infeasible. To tackle this, we propose Heterogeneous Allocation via Localized Observations (HALO) to solve the Vehicle Routing Problem (VRP). HALO is a hybrid method that splits the VRP into allocation and routing portions to provide onboard, real-time solutions to robots in dynamic environments. During the allocation phase, HALO utilizes a heterogeneous graph neural network framework with unique message passing layers to explicitly separate the learning of spatial distributions and task-to-robot compatibility. Evaluation results on a partially observable, online variant of the VRP show HALO significantly outperforms the heuristic baseline while maintaining similar solution quality to an all-knowing offline variant of HALO. While HALO is explicitly designed for partially observable environments, it imposes no strict upper bound on the observation space allowing us to test HALO on the traditional static, single-depot VRP. Here, HALO outperforms state-of-the-art architectures strictly optimized for the static variant of the VRP by up to 14.06% . Throughout all testing, this framework maintains the quickest execution times which emphasizes its potential for large-scale, real-time deployment. Comments: 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works Subjects: Robotics (cs.RO); Multiagent Systems (cs.MA) Cite as: arXiv:2609.38760 [cs.RO] (or arXiv:2609.38760v1 [cs.RO] for this version) https://doi.org/10.48550/arXiv.2609.38760 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[MA-16] CollabFlow: Recursive Self-Improvement of Agent Collaboration
【速读】：该论文旨在解决大语言模型（LLM）驱动的多智能体系统中递归自改进（Recursive Self-Improvement, RSI）机制所面临的三大核心挑战：协作关系在操作层面预先定义、仅通过拓扑学习导致信息传递固化并传播错误，以及基于系统自身输出的奖励最大化集中于少数团队，抑制了多样性与持续进化。其解决方案的关键在于提出一种可训练的协作生成框架——CollabFlow，该框架包含一个可训练的协作指挥官（Collab-Director）用于动态构建完整的智能体团队，一个冻结的执行器（executor）负责运行团队，并通过每轮结果反馈对指挥官进行再训练。在每轮内部，协作图中的边采用基于证据条件通信（Evidence-Conditioned Communication）协议，接收方仅在发送方证据显著更强时才采纳其答案，从而使指挥官学习到有效的通信模式；跨轮次则引入基于流的协同轨迹平衡（Collaborative Trajectory Balance, CTB）目标函数，确保每个团队仅在一次构造顺序中获得奖励，并实现奖励分配与团队表现成比例，从而维持多个优质团队的持续参与。此外，论文还建立了目标分布随记录积累而收敛的理论边界，保证自生成目标的稳定性。实验在十二个数据集上验证了CollabFlow显著优于所有基线方法，并展现出持续递进的性能提升。
链接: https://arxiv.org/abs/2609.38662
作者: Xiao Huang,Mingda Zhang,Junming Zhang,Qiang Huang,Hanwen Zhang,Yue Dai,Zijia Wang,Xiaoying Tang
机构: The Chinese University of Hong Kong, Shenzhen (深圳大学); Fudan University (复旦大学); University of Oxford (牛津大学)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注:
Abstract:Recursive self-improvement (RSI) lets a system improve from its own outcomes; in LLM-based multi-agent systems, Agents refine one another within a task, and outcomes improve how they collaborate across tasks. However, existing multi-agent collaboration leaves this loop open: collaboration is pre-defined at the operator level, topology-only learning keeps verbatim exchange that propagates errors, and reward maximization on a system’s own outcomes concentrates on a few teams. To address these challenges, we propose CollabFlow, an RSI system of Learned Agent Collaboration: a trainable Collab-Director constructs teams of complete Agents, a frozen executor runs them, and each round’s outcomes retrain the director. Within each round, the edges of a collaboration graph carry protocols of Evidence-Conditioned Communication: a receiver adopts a differing answer only when the sender’s evidence is stronger by a margin, so the director learns who communicates and how. Across rounds, we further propose Collaborative Trajectory Balance (CTB), a flow-based objective that credits each team once across its construction orders and targets a reward-proportional distribution over teams, so several good teams stay in play. We also bound how far this self-generated target moves between rounds, which shrinks as records accumulate. On twelve datasets, CollabFlow outperforms all baselines and keeps improving across rounds. Code is available at this https URL.
[MA-17] Breaking Babel: A Self-Evolving Multi-Agent System for Long-Form Subtitle Translation
【速读】：该论文旨在解决长篇字幕翻译中面临的跨集叙事连贯性与文化语境理解难题，同时确保术语和风格的一致性。现有方法多局限于单模型的逐句处理，而多智能体系统则常采用静态工作流，难以适应场景复杂度或制作背景的变化。为此，本文提出SMART（Self-evolving Multi-Agent System for Long-form Subtitle Translation），其核心创新在于引入测试时训练（test-time training）机制，通过动态路由与混合智能体（Mixture-of-Agents）层实现自适应翻译决策，并构建持久的系列级记忆。系统集成术语验证、字幕约束校验及上下文检索等工具，结合判别-精炼循环（judge-refiner loop）利用文本批判反馈迭代优化智能体提示词与路由策略，无需重训练底层大语言模型（LLM）。在推理阶段，演化后的配置完成剩余内容翻译。此外，研究构建了涵盖14种类型、每部剧集2至198集、时间跨度1959–2023年、覆盖15个目标语区的Subtitle Arena数据集，以及适配字幕任务的SubMQM评估框架（含7个维度与19类错误）。实验表明，SMART在所有15个方向均取得最优的MQM得分，平均惩罚降低6.9%；在公开的MuSC基准上，对全部四组语言对均达到最佳模型性能，并获得4.50/5的人工评估综合得分，显著优于现有方法。
链接: https://arxiv.org/abs/2609.38660
作者: Haibo Jin,Xinjie Li,Najmeh Sadoughi,Yang Liu,Yibo Wang,Zhu Liu,Yuzong Liu
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: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Multiagent Systems (cs.MA)
备注: 49 pages
Abstract:Long-form subtitle translation requires reasoning over discourse and cultural context spanning episodes or entire series, while maintaining consistent terminology and style. Existing single-LLM methods are largely sentence-level, and multi-agent systems often use static workflows that do not adapt to scene complexity or production context. We propose SMART, a Self-evolving Multi-Agent system for long-foRm subtitle Translation. During test-time training, SMART builds persistent series-level memory and translates a subset of sentences through a dynamic router and Mixture-of-Agents layer with tools for terminology verification, subtitle constraint validation, and contextual retrieval. A judge-refiner loop scores candidates and uses textual critiques to update agent prompts and routing policies without retraining the underlying LLMs. During test-time inference, the evolved configuration translates the remaining series. We also introduce Subtitle Arena, covering 14 genres, 2–198 episodes per series, production years 1959–2023, and 15 target locales, together with SubMQM, a subtitle-adapted MQM framework with seven dimensions and 19 error categories. SMART achieves the best overall MQM score in all 15 Subtitle Arena directions, reducing average penalty by 6.9% over the strongest competing agent system. On the public MuSC benchmark, SMART obtains the best model result across all four language pairs and also achieves the best human-evaluation result, with an overall score of 4.50/5.
[MA-18] Recursive Organization Improvement: A Modeling Specification for Human–Agent Organizations
【速读】：该论文旨在解决强人工智能代理（Strong AI agents）在组织中应用时，其能力提升并不自动带来组织效能优化的问题，核心挑战在于组织需具备动态识别并调整工作安排的能力。解决方案的关键在于提出一种递归式组织改进的建模规范，该规范通过将可观察的历史记录、组织记忆、决策权分配与携带证据的变更契约相连接，实现对组织演进过程的可执行验证。研究通过可执行检查器、公开记录映射及受控模拟，在固定资源约束下系统评估六种决策规则、三种记忆条件与三种任务环境的组合。结果表明，在稳定环境中，累积证据可使平衡评估的单位任务净价值从0.45224提升至0.48007，重复再评估相对于此基准的劣势由0.01702降至0.00007；而最佳工作流的反转揭示了长期保留策略的适应延迟成本，有限时间窗口虽带来过渡代价但可恢复最终性能。探索性控制实验进一步表明，试错获取与标签复用会降低再评估收益（从0.00607降至0.00191），程序替换在测试范围内未产生稳定增益。研究最终指出，证据的获取、复用与及时更新机制必须与评估者替换相分离，方能有效评估组织改进效果。
链接: https://arxiv.org/abs/2609.38643
作者: Zilong Wang
机构: CataX AI
类目: Multiagent Systems (cs.MA); Human-Computer Interaction (cs.HC)
备注: 18 pages, 3 figures, 9 tables;
Abstract:Stronger AI agents do not automatically produce better organizations: teams must also learn which work arrangements to retain and when to reconsider them. We propose a modeling specification for recursive organization improvement and evaluate it through an executable checker, a public-record mapping, and controlled simulation. The specification connects actor-visible histories, organizational memory, decision rights, and evidence-carrying change contracts. The mechanism study crosses six decision rules, three memory conditions, and three task environments under fixed resource ceilings. In a stationary environment, cumulative evidence raises balanced evaluation’s normalized net value per task from 0.45224 to 0.48007. Repeated reassessment’s disadvantage relative to this comparator falls from 0.01702 with reset evidence to 0.00007 with cumulative evidence. A reversal of the best workflow reveals the opposite cost: indefinite retention delays adaptation, while a finite window restores eventual performance at a transition cost. In exploratory controls, matching trial acquisition and label reuse reduces the apparent reassessment gain from 0.00607 to 0.00191. Program replacement adds no stable benefit across the tested reversal times. The study identifies evidence acquisition, reuse, and timely updating as mechanisms that must be separated from evaluator replacement when assessing organizational improvement.
[MA-19] From Solo to Social Learning: Characterizing Recursive Social Improvement in LLM s
【速读】：该论文旨在解决多智能体系统中，当每个大语言模型（LLM）自主追求自身奖励时，能否通过相互学习实现整体群体的递归社会性改进（recursive social improvement）。其核心问题在于：在缺乏统一目标导向的前提下，自进化式LLM是否能通过协作与知识共享，提升整个群体的性能表现。解决方案的关键在于构建一个由多个独立运行的LLM组成的群体，这些智能体共同维护并修订技能文件（skill files），并在有限的令牌预算（token budget）内决定何时、何地以及向谁复制技能。尽管引入了社会学习机制，实验结果显示，尽管模型间存在技能传递与修订行为，且部分个体因观察同伴而更早发现有效技能或减少私有搜索开销，但整体上这些模型的单位令牌收益仍低于独立学习者，且探索范围过窄或提前耗尽资源。这表明当前的自进化LLM虽可通过模仿提升学习效率，但在实现更优性能方面仍存在局限，其知识传播导致群体趋于集中于少数独立发现，未能实现更广泛的有效性提升。
链接: https://arxiv.org/abs/2609.38516
作者: Kunal Jha,Max Kleiman-Weiner,Natasha Jaques
机构: 未知
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Large language models (LLMs) can now improve themselves by revising the instructions they follow, and LLM agents are increasingly orchestrated to work together on complex problems. However, self-improvement methods typically optimize one system at a time, and multi-agent frameworks often have every model work toward a shared goal. We ask a different question. When each agent pursues its own reward, can self-improving LLMs learn from one another well enough to improve the whole population? We call this capability recursive social improvement. We study populations that revise skill files and choose whether, when, and whom to copy from. Independent search, learning from peers, and acting all share one token budget. In controlled environments, established social-learning algorithms benefit from peers, but three LLMs do not. They earn less reward per token than solo learners, and explore too narrowly or run out of tokens before acting. We then let the models write and revise their own skills. Observing peers changes how they improve, helping one model find useful skills sooner and another spend less on private search. Neither, however, outperforms independent learners at the same cost. Skills are copied, revised, and passed on, so one discovery can seed further search. Yet these exchanges concentrate the population around fewer independent discoveries. Together, these results show that LLMs can make learning more efficient by copying from peers, but not yet more effective.
[MA-20] PANDA: A Decentralized Architecture with Flexible Orchestration for Scalable Fault-Tolerant Multi-Agent Systems
【速读】：该论文旨在解决基于大语言模型（LLM）的多智能体系统（Multi-Agent System, MAS）在大规模场景下难以可靠高效地完成多步任务的核心挑战，具体包括：无法支持大量智能体与并发任务、缺乏故障容错能力、难以有效治理智能体间交互，以及无法适应不同任务所需的多样化规划与执行模式。其解决方案的关键在于提出一种去中心化架构PANDA，通过解耦集体通信与团队通信，使智能体可同时参与多个任务团队，实现任务负载均衡与单个智能体内的并发工作调度；进一步将底层架构与编排策略分离，支持星型、链式和网状三种可按任务需求动态选择的规划与执行模式；通过动态重规划实现对基础设施与编排故障的检测与恢复；并采用基于“信任网”（web-of-trust）的治理机制，在不依赖集中式服务的前提下保障智能体间的可信交互，从而在保持高可扩展性的同时确保系统鲁棒性与任务完成率。
链接: https://arxiv.org/abs/2609.38482
作者: Matthew D. Laws,Cristina Nita-Rotaru
机构: Khoury College of Computer Sciences (Khoury计算机科学学院); Northeastern University (东北大学)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注: 19 pages, 7 figures, 5 tables
Abstract:Existing architectures for LLM-based multi-agent systems (MAS) cannot reliably and efficiently solve multi-step tasks at scale: they struggle to support large numbers of agents and concurrent tasks, tolerate failures, govern agent interactions, and accommodate the diverse planning and execution patterns different tasks require. We present PANDA, a decentralized architecture that connects a large collective of heterogeneous, independently administered agents, letting them discover each other’s capabilities and self-organize into small specialized teams per task. PANDA scales by decoupling collective communication from team communication, allowing agents to participate in multiple teams simultaneously, load-balancing tasks across the collective, and scheduling concurrent work within each agent. PANDA further separates the underlying architecture from the orchestration strategy, supporting three planning and execution patterns (star, chain, and mesh) that can be selected according to the structure and requirements of each task. PANDA detects infrastructure and orchestration failures and recovers affected tasks by dynamically replanning around failed components. Finally, to provide governance without a centralized service that would limit scalability, PANDA uses a web-of-trust model to constrain agent interactions to established trust relationships. We evaluate PANDA on the HotPotQA benchmark, demonstrating that it scales to thousands of agents, assembles teams in milliseconds, matches state-of-the-art accuracy at up to 8x the efficiency, and sustains 100% task completion under faults where existing systems fail.
[MA-21] Absorbing State Phase Transitions in Multi-Agent Search
【速读】：该论文旨在解决大规模语言模型（Large Language Model, LLM）驱动的多智能体系统在执行搜索任务时，其协作行为的可预测性与成功率问题。具体而言，研究关注如何通过统计力学中的吸收态相变（absorbing state phase transition）理论框架，预测多智能体系统在不同通信拓扑下的任务求解成功概率。其解决方案的关键在于理论上推导出一个临界通信度 dc——即每个智能体至少需与 dc 个其他智能体保持通信，才能确保错误假设不会无限制传播，从而保障系统能够稳定进入“已求解”状态。这一临界值为优化多智能体通信拓扑提供了理论依据。然而，实验评估发现，当前前沿的基于LLM的多智能体系统在真实世界任务（如软件配置调试与物理机制发现）中与理论预测存在混合一致性：部分系统表现出符合预期的行为，但更多情况下，智能体倾向于不与其邻居通信，发展出对个体有利但抑制协作效率的策略，导致实际性能偏离理论预期。
链接: https://arxiv.org/abs/2609.38327
作者: Wenwen Zheng,Yuzhe Yang,Helen Qu,Xin Eric Wang,Haewon Jeong
机构: University of California, Santa Barbara (加州大学圣塔芭芭拉分校); Flatiron Institute (扁平铁研究所)
类目: Multiagent Systems (cs.MA)
备注:
Abstract:Nontrivial dynamics can emerge in large language model (LLM)-based multi-agent systems, and preliminary evidence exists that formalisms from statistical mechanics can be effective at modeling and predicting such behaviors. In parallel, designing multi-agent communication topology for optimal task-solving is an active research question. In this paper, we focus on predicting the success of multi-agent search tasks using the formalism of absorbing state phase transitions. We first taxonomize search tasks into four types, informed by classical results in combinatorial search. We then theoretically derive a critical communication degree d_c , the minimum number of agents each agent can communicate with, above which incorrect hypotheses do not proliferate uncontrollably and the search enters the solved state. Finally, we evaluate frontier LLM-based multi-agent systems on real-world search and discovery tasks, software configuration debugging and physical mechanism discovery, and find that agreement with theory is mixed. LLM agents may not communicate with their neighbors and can develop strategies that are individually beneficial but limits the benefits of collaboration.
[MA-22] Which Models Work Well Together? Measuring Heterogeneity for LLM Team Selection
【速读】：该论文旨在解决大语言模型（Large Language Model, LLM）团队在协作中面临的性能瓶颈问题，即团队整体表现不仅受限于单个模型的能力，还受到成员间错误共振（error resonance）和预测行为差异性不足的影响。现有方法虽普遍观察到异构组队（heterogeneous teaming）的有效性，但缺乏可计算、可解释且可优化的互补性度量指标，导致团队组合依赖经验性启发式策略。其解决方案的关键在于提出一种基于异质性的团队选择框架，通过离线分析刻画每个模型的能力特征，并引入两个互补信号：一是误差模式去相关性（decorrelation in error patterns），以降低共同失效风险；二是预测行为分歧度（divergence in predictive behavior），用于捕捉策略多样性。将团队选择建模为一个标准化的质量-互补性联合优化目标，并采用高效的贪心搜索算法从候选池中筛选出小规模最优团队。实验结果表明，在多个基准测试中，该框架在受控候选集与团队规模下均显著优于仅关注质量的基线方法，为多大语言模型系统提供了可复用的团队构建原则。
链接: https://arxiv.org/abs/2609.38274
作者: Liangyu Teng,Hengsong Liu,Juncen Guo,Jingyu Zhang,Yang Liu,Jing Liu,Liang Song
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注:
Abstract:The performance ceiling of an LLM team is constrained not only by individual model capabilities, but also by inter-member error resonance and predictive differences. Although heterogeneous teaming is often observed to be effective in practice, existing approaches lack complementarity metrics that are computable, interpretable, and optimizable, leaving team composition to rely on heuristics. We propose a heterogeneity-driven team selection framework that performs offline profiling to characterize individual capability along with two complementary signals: one captures decorrelation in error patterns to reduce co-failures, while the other measures divergence in predictive behavior to capture strategy diversity. We formulate team selection as a standardized quality–complementarity combinatorial objective and apply an efficient greedy search to select a small team from a candidate pool. Experiments across multiple benchmarks demonstrate that our framework consistently outperforms quality-only baselines under controlled candidate pools and team sizes, establishing reusable selection principles for multi-LLM systems.
[MA-23] VirusCascade: Hijacking Collaborative Reflection in LLM -Powered Recommender Agents NDSS2027
【速读】：该论文旨在解决生成式推荐系统（Generative AI Recommender Systems, GARS）中因多智能体协同反思机制引入的新型安全漏洞问题。具体而言，传统静态评分模型的推荐结果具有可预测性，而基于大语言模型（LLM）的自主代理推荐系统（LLM-ARS）通过动态的“协同反思”过程不断优化用户与物品的语义状态，虽提升了推荐质量，却也暴露了系统性脆弱点：攻击者可通过向单个代理注入恶意证据，使其在反思过程中将该证据合理化为看似合理的偏好叙事，并写回记忆、传播至其他代理，从而实现对整个系统的隐蔽操控。这一过程被称为“反射洗白”（reflection laundering），其在多代理间扩散则构成“协同反思劫持”（collaborative-reflection hijacking）。现有攻击方法受限于静态推荐流程假设，无法利用此递归、多智能体放大路径。为此，论文首次开展受控漏洞分析，揭示协同反思劫持的两个可被利用特性：反射持久性（reflective persistence）与跨代理传播性（cross-agent propagation）。基于此，提出首个黑盒目标推广攻击方法VirusCascade，其核心创新在于同时构建语义与结构双重攻击面：前者确保目标物品能自然地被合理化为满足广泛用户偏好的内容，后者则优化其在系统中的传播路径以实现全局扩散。在四个真实世界数据集上对多种架构的实验表明，VirusCascade在保持隐蔽性的前提下显著提升目标物品曝光率，平均E@20达0.384，较最强基线提升绝对值0.185，达到当前最优水平。
链接: https://arxiv.org/abs/2609.38270
作者: Yurong Hao,Wen Zhou,Guowei Guan,Tiantong Wu,Fuyao Zhang,Wei Yang Bryan Lim
机构: Nanyang Technological University (南洋理工大学)
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: Accepted by NDSS 2027
Abstract:Advancing beyond traditional static scoring models, LLM-powered agentic recommender systems (LLM-ARS) instantiate users and items as autonomous agents, whose semantic states are dynamically refined through a recurrent process known as collaborative reflection. While this mechanism improves recommendation quality, it simultaneously introduces a systemic vulnerability: adversarial evidence injected into a single agent can be rationalised into a legitimate preference narrative, written back into memory, and propagated to other agents through interaction contexts. We term the local rationalisation process reflection laundering, and its system-wide escalation through collaborative reflection collaborative-reflection hijacking. Existing attacks on recommender systems, whether based on interaction-level data poisoning or text-level adversarial perturbations, assume static pipelines and thus cannot exploit this recurrent, multi-agent amplification pathway. To bridge this gap, we first conduct a controlled vulnerability analysis that establishes two exploitable properties underlying collaborative-reflection hijacking: reflective persistence and cross-agent propagation. Then building on these findings, we propose VirusCascade, the first black-box targeted promotion attack that jointly shapes semantic and structural attack surfaces: the former ensures the target item is naturally rationalised as satisfying broad user preferences, the latter positions it for system-wide propagation. Extensive experiments on four real-world datasets across diverse LLM-ARS architectures demonstrate that VirusCascade consistently achieves state-of-the-art targeted exposure under evaluated stealth constraints, reaching a mean E@20 of 0.384 and surpassing the strongest baseline by an absolute margin of +0.185.
[MA-24] SQD-Agent : LLM -driven agent ic framework for Quantum Chemistry workflows
【速读】：该论文旨在解决应用研究人员在将领域特定问题（如量子化学计算）转化为可执行的混合量子-经典工作流时所面临的巨大挑战，主要障碍包括对量子算法的深入理解、量子编程细节的掌握以及硬件感知的系统集成能力。为应对这一问题，论文提出了一种基于大语言模型（LLM）的智能体框架——SQD Agent，其核心解决方案在于通过自然语言理解与推理能力，自动将用户意图翻译为可执行的量子化学工作流，特别针对来自样本基量子对角化（Sample-Based Quantum Diagonalization, SQD）家族的算法。该框架的关键创新在于构建了一个模块化且可扩展的架构，支持异构量子后端、经典求解器及工作流组件的无缝集成，从而适应快速演进的量子生态系统；同时，其内嵌的交互式功能实现了运行时性能剖析、瓶颈分析、资源优化、智能结果缓存及收敛性可视化，显著降低了用户对量子技术的入门门槛。此外，该框架还具备在真实量子硬件上实施误差缓解的能力，并能评估不同缓解方案在误差恢复潜力与计算预算之间的权衡，帮助用户做出更明智的实验决策。
链接: https://arxiv.org/abs/2609.39302
作者: Kislaya Tiwari,Anupama Ray
机构: Indian Institute of Technology Delhi; IBM Quantum, IBM Research (IBM 量子, IBM 研究)
类目: Quantum Physics (quant-ph); Multiagent Systems (cs.MA)
备注: Accepted in IEEE International Conference on High Performance Computing (HiPC 2026)
Abstract:Quantum algorithms and quantum hardware are advancing towards a promising paradigm for scientific applications. However, translating domain-specific problems into executable hybrid quantum-classical workflows remains a significant barrier for application researchers due to the required expertise in quantum algorithms, nuances in quantum programming, and hardware-aware system integration. At the same time, AI and primarily LLM based agents are increasingly capable of interpreting natural-language intent, reasoning over complex workflows, and translating high-level objectives into executable code and building computational pipelines. In this work, we introduce SQD Agent, an LLM-based agentic framework that translates natural-language user intent into executable workflows for Quantum Chemistry applications where algorithms from the Sample-Based Quantum Diagonalization (SQD) family are used. By automating this translation, SQD Agent reduces the level of human expertise and configuration overhead required, thereby simplifying experimentation in hybrid quantum-classical settings for application researchers new to quantum. SQD Agent adopts a modular and extensible architecture that supports seamless integration of heterogeneous quantum backends, classical solvers, and workflow components, ensuring adaptability to rapidly evolving quantum ecosystems. The framework further incorporates interactive capabilities for on-demand profiling, bottleneck analysis, resource optimization, intelligent result caching, and convergence visualization. Key features include quantum chemistry experiments, error mitigation on real quantum hardware, together with analysis of candidate mitigation schemes in terms of their potential error-recovery behavior and computational budget, helping users understand their practical trade-offs and decide which strategies to explore in subsequent experiments.
[MA-25] Decentralized Decision-Making among Heterogeneous Autonomous Vehicles: An α-Potential Game Framework
【速读】：该论文旨在解决异构自主车辆在非合作多车博弈中如何高效求解近似纳什均衡（Nash Equilibrium, NE）的问题，尤其关注车辆间交互权重可能存在非对称性时的求解挑战。其核心解决方案是构建一个α-势博弈（α-potential game）框架，将求解近似NE的问题转化为单个辅助α-势函数的最小化问题。该框架的关键在于显式构造出α-势函数，并建立其极小化解的存在性，同时将均衡近似误差α与交互非对称性量化关联。为进一步降低有效交互非对称性，论文引入车辆特异性缩放机制，显著提升了均衡近似的精度，甚至在特定情形下可恢复精确纳什均衡。此外，该框架还提供了所选势函数策略的社会效率保证，揭示了交互结构对最坏情况效率的影响。数值实验验证了该框架在捕捉异构车辆交互、避撞、障碍物规避、变道超车及基于优先级的交叉口通行等复杂交通场景中的灵活性与有效性。
链接: https://arxiv.org/abs/2609.38731
作者: Anran Hu,Zhexin Wang,Yufei Zhang,Xuan Di
机构: Columbia University (哥伦比亚大学); Imperial College London (帝国理工学院)
类目: Optimization and Control (math.OC); Computer Science and Game Theory (cs.GT); Multiagent Systems (cs.MA)
备注:
Abstract:We study noncooperative multi-vehicle games among heterogeneous autonomous vehicles, where each vehicle adopts a decentralized closed-loop policy based on its own state, and optimizes an objective that depends on other vehicles through potentially asymmetric interaction weights. We develop an \alpha -potential game framework that reduces the computation of an approximate Nash equilibrium (NE) to the minimization of a single auxiliary \alpha -potential function. We explicitly construct this \alpha -potential, establish the existence of its minimizers, and characterize the equilibrium approximation error \alpha in terms of interaction asymmetry. We further introduce vehicle-specific scaling to reduce the effective interaction asymmetry, thereby tightening the equilibrium approximation and, in important cases, recovering an exact NE despite asymmetric interactions. We also derive social-efficiency guarantees for the potential-selected policies, revealing how the interaction structure shapes worst-case efficiency. Numerical experiments demonstrate the flexibility of the framework in capturing heterogeneous vehicle interactions, collision and obstacle avoidance, lane changing and overtaking under different traffic configurations, and priority-based intersection crossing.
[MA-26] Multi-agent discussion gains less when dissent is withheld
【速读】：该论文旨在解决多智能体大语言模型（Multi-agent systems of LLMs）中讨论机制是否提升准确性这一长期存在的矛盾问题：尽管理论上讨论应通过协同推理提高决策质量，但实证研究却存在相互冲突的结论——部分研究发现讨论能显著提升准确率，而另一些研究则观察到讨论反而导致错误共识。其解决方案的关键在于提出一个简洁的理论模型，基于在LLM智能体中反复观测到的四种核心行为：（1）压制异议（withholding dissent）、（2）内化陈述答案（internalizing a stated answer）、（3）在看到异议后重新考虑（reconsidering after seeing dissent）、（4）向正确答案修正（correcting toward the correct answer）。该模型揭示，讨论能否纠正初始错误多数意见，取决于压制异议率 c 是否低于一个由净修正率 γ 与内化率 a 共同决定的临界值 c∗=γ/(γ+a)。通过贝叶斯方法从对话日志中估计这些参数，并将不同LLM团队置于 c∗ 相对位置进行评估，结果表明：随着压制率上升，讨论带来的增益显著下降，这一趋势在HiddenBench和MedEInst等基准测试中均得到验证。进一步实验显示，通过指令禁止压制异议可提升讨论收益；关闭推理模块亦能增强效果，因其降低了内化率 a 并减少对少数意见的重新审视。该研究统一解释了先前矛盾的实证结果，并明确了讨论优于多数投票的条件边界。
链接: https://arxiv.org/abs/2609.38324
作者: Chand Sahil Mansuri,Xin Wang,Mengying Li,Bryan Acton,Rory Eckardt,Dhaval Patel,Sadamori Kojaku
机构: Binghamton University (宾汉姆顿大学); IBM T. J. Watson Research Center (IBM T. J. 沃森研究中心)
类目: Physics and Society (physics.soc-ph); Computation and Language (cs.CL); Multiagent Systems (cs.MA)
备注:
Abstract:Multi-agent systems of LLMs add discussion to majority voting and are therefore expected to be more capable. However, empirical reports conflict on whether discussion improves accuracy or leads to an incorrect consensus. Here, we introduce a parsimonious model that explains when discussion improves accuracy and when it ends in an incorrect consensus, built from four behaviors repeatedly observed in LLM agents: (1) withholding dissent, (2) internalizing a stated answer, (3) reconsidering after seeing dissent, and (4) correcting toward the correct answer. The model shows that discussion can overturn an incorrect initial majority only when the withholding rate c is below a critical rate c^* = \gamma/(\gamma + a) , set by the net correction rate \gamma and the internalization rate a . We estimate these rates from conversation logs with a Bayesian method and place LLM teams relative to c^* . As the model predicts, the gain from discussion shrinks as withholding rises, across LLMs and on a hidden profile benchmark, HiddenBench, and MedEInst. Instructing agents not to withhold dissent increases this gain. Turning reasoning off also increases the gain, because reasoning raises the internalization rate a and keeps agents from reconsidering a minority answer. These findings reconcile the conflicting reports and identify when discussion outperforms majority voting.
自然语言处理
[NLP-0] Ranking-Aware Prompt Optimization for Multimodal Clinical Diagnosis
【速读】： 该论文旨在解决多模态大语言模型（Multimodal Large Language Models, MLLMs）在临床诊断应用中因依赖准确率（accuracy）作为优化目标而导致的性能偏差问题。由于临床数据普遍存在严重的类别不平衡，仅追求高准确率的模型可能表现为对多数类的盲目预测，而忽略对少数但关键病患的识别能力，从而导致临床实用性低下。为此，论文提出以受试者工作特征曲线下面积（AUROC）为核心优化目标，该指标不依赖分类阈值且对类别分布具有不变性，能更真实地反映模型对正负样本的排序能力。其解决方案的关键在于引入成对级帕累托提示进化（Ranking-PE）：将传统基于正确性评分的提示进化框架中每个样本的“是否正确”判断，替换为针对（阳性，阴性）样本对的排序判断——即当某个提示对阳性样本的打分高于配对阴性样本时，记为1。这一设计使得每列的平均值恰好等于经验AUROC（依据Wilcoxon-Mann-Whitney定理）。该方法被应用于提示进化搜索流程中的三个层级——帕累托支配判定、反射式语言模型的示例反馈以及最终候选提示选择——无需额外模型调用或代理损失函数。在MIMIC数据集上对三种疾病进行实验表明，基于准确率的提示进化可能导致排序性能下降，而Ranking-PE则显著提升模型排序能力，相较微调后的Qwen3-VL-8B和MedGemma-4B分别实现+5.8 AUROC百分点和+16.2 AUROC百分点的增益。消融实验进一步验证了医学级视觉骨干网络（如通过视觉编码器微调或医学预训练获得）是提示搜索无法替代的基础前提，表明该方法成功将反射式提示进化从纯文本场景拓展至多模态临床决策领域。
链接: https://arxiv.org/abs/2609.40361
作者: Tian Xia,Minghao Liu,Yiqing Liang,Laixi Shi,Jiayun Wang
机构: Harvard University (哈佛大学); University of California, Santa Cruz (加州大学圣克鲁兹分校); Brown University (布朗大学); Johns Hopkins University (约翰霍普金斯大学); Georgia Institute of Technology (佐治亚理工学院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multimodal large language models (MLLMs) are rapidly advancing clinical diagnosis, yet their adaptation pipelines remain anchored to accuracy-based objectives. Clinical data are heavily class-imbalanced: a constant-majority predictor can score above 90% accuracy while being clinically useless. We therefore evaluate and optimize for AUROC, a threshold-free score that ranks positives above negatives and is invariant to class balance. We focus on prompt optimization in MLLMs. Reflective methods such as GEPA use a binary scores matrix with one row per evaluation instance and one column per candidate prompt; cells record per-instance correctness, so the column average is accuracy and drives candidate selection. We introduce pair-level Pareto prompt evolution (Ranking-PE), which replaces each correctness row with a pairwise-ordering row over (positive, negative) instance pairs: the cell is 1 if the candidate scores the positive higher than the paired negative. The column average then equals empirical AUROC (by the Wilcoxon-Mann-Whitney identity). We apply this swap at all three layers the prompt evolution search reads from - the scores matrix that decides Pareto dominance, the per-example feedback to the reflection LM, and final candidate selection - at no extra model calls and with no surrogate loss. Across three diseases on MIMIC, accuracy-based prompt evolution can degrade ranking; Ranking-PE reverses this, beating the accuracy-based recipe by +5.8 AUROC pp on fine-tuned Qwen3-VL-8B and +16.2 pp on MedGemma-4B. Ablations examine each design component and show that a medical-grade visual backbone - via vision-encoder-tuned SFT or medical pretraining - is a prerequisite that prompt search cannot replace - our recipe extends reflective prompt evolution from text-only data to multimodal clinical decision-making.
[NLP-1] Semifactual Credit-Augmented Policy Optimization
【速读】： 该论文旨在解决大语言模型在基于可验证奖励的强化学习（Reinforcement Learning with Verifiable Rewards, RLVR）中，其推理预测仍对任务无关的提示特征敏感的问题。尽管RLVR提升了模型的推理能力，但模型输出仍可能受到与问题本质无关的提示成分干扰，导致推理过程不稳定。本文的关键解决方案是提出一种受因果启发的新型策略优化方法——半事实信用增强策略优化（Semifactual Credit-Augmented Policy Optimization, SCAPO），其核心在于通过半事实干预（semifactual prompt interventions）评估每个生成标记（token）的概率漂移，量化其在固定答案下的稳定性，并据此对令牌级别的信用分配进行精细化调整。具体而言，SCAPO利用归一化的稳定性得分，在训练初期降低相对不稳定的标记所获得的优势，从而抑制潜在的虚假依赖关系，同时不因稳定性本身额外赋予奖励。实验表明，相较于组相对策略优化（Group Relative Policy Optimization, GRPO），SCAPO在Qwen3-4B-Base和Qwen3-1.7B-Base上分别将AIME 2024–2026基准测试的准确率提升5.63和4.17个百分点，并在多数数学基准及所有分布外（out-of-distribution）基准上取得最优性能。该结果验证了基于半事实稳定性的细粒度信用分配机制能有效提升模型推理的准确性与泛化能力。
链接: https://arxiv.org/abs/2609.40360
作者: Junshu Pan,Zhizhang Fu,Shulin Huang,Yiran Ding,Zifan Cheng,Wenqi Shao,Qiaosheng Zhang,Yue Zhang
机构: Zhejiang University(浙江大学); Westlake University(西湖大学); Shanghai Innovation Institute(上海创新研究院); Shanghai AI Laboratory(上海人工智能实验室)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Reinforcement learning with verifiable rewards (RLVR) has improved the reasoning capabilities of large language models (LLMs), yet their predictions remain sensitive to task-irrelevant prompt features. We investigate this sensitivity through semifactual prompt interventions that preserve the underlying problem and its answer. Our analysis reveals substantial variation in token-level sensitivity and shows that suppressing high-drift token candidates during decoding improves reasoning accuracy without updating model weights. These findings highlight a limitation of Group Relative Policy Optimization (GRPO), which assigns the same outcome-derived advantage to every response token and may reinforce potential spurious dependence alongside useful reasoning. Motivated by this observation, we introduce Semifactual Credit-Augmented Policy Optimization (SCAPO), a causally inspired variant of GRPO that incorporates semifactual stability into token-level credit assignment. SCAPO measures token probability drift for fixed responses under semifactual interventions and uses normalized stability scores to reduce advantages for relatively unstable tokens during early training, while granting no additional credit for stability alone. On Qwen3-4B-Base and Qwen3-1.7B-Base, SCAPO improves AIME 2024-2026 accuracy over GRPO by 5.63 and 4.17 percentage points, respectively. At both model scales, SCAPO achieves the best results on most evaluated mathematics benchmarks and all evaluated out-of-distribution benchmarks among the compared methods. These results suggest that semifactual stability provides an effective training signal for improving reasoning and generalization through finer-grained credit assignment in RLVR. The code is available at this https URL.
[NLP-2] EvoDuet: Bilevel Co-Evolution of Web Searching and Task Solving for Scientific Discovery
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在进化搜索过程中因缺乏外部知识而导致优化停滞的问题。其核心挑战在于，当任务进展依赖于模型自身未掌握的外部信息时，传统的简单网页搜索工具易重复返回相同结果，无法有效支持动态演进的解决方案。为此，论文提出EvoDuet——一种双层优化方法，通过固定模型参数，协同演化解决方案与搜索查询。其关键创新在于引入检索门控机制（retrieval gate），使模型能够自主判断知识缺口，并决策是否检索新文档、复用已有文档或无需外部信息直接推进。内层循环基于解的评分对查询进行优化并排序文档，外层循环则并行生成候选解，并记录评估结果以指导后续检索。实验表明，EvoDuet在21个优化任务中显著提升发现效率，相较于OpenEvolve，GPT-5.6-Luna和Gemini-3.8-Flash的归一化发现增益分别从74.1%/61.3%提升至78.0%/82.3%，并在8项任务上超越此前最优成绩；同时，该方法对多种进化搜索范式（如Top-K、EvoX）具有普适性，展现出良好的可扩展性与适应性。
链接: https://arxiv.org/abs/2609.40340
作者: Young-Jun Lee,Jinheon Baek,Soyeong Jeong,Minki Kang,Seungyeon Jwa,Jonghyun Choi,Seungho Han,Dongyeop Kang
机构: University of Minnesota(明尼苏达大学); KAIST(韩国科学技术院); Seoul National University(首尔国立大学); Hanyang University(汉阳大学)
类目: Computation and Language (cs.CL)
备注: Project page: this https URL
Abstract:Evolutionary search with large language models (LLMs) can stall when progress requires external knowledge the model lacks. Supplying relevant documents helps, but simply adding web search tool can keep returning the same pages as solutions change. We introduce EvoDuet, a bi-level optimization method that co-evolves solutions and search queries with fixed model parameters. At each iteration, a retrieval gate lets the LLM assess its knowledge gap and choose to retrieve new documents, reuse stored ones, or proceed without them. An inner loop refines queries and ranks documents by the solution scores they are predicted to yield; an outer loop generates candidates in parallel from these documents and records the evaluated outcomes for later searches. Across 21 optimization tasks with one candidate per iteration, EvoDuet raises OpenEvolve’s normalized discovery gain from 74.1% to 78.0% with GPT-5.6-Luna and from 61.3% to 82.3% with Gemini-3.8-Flash, whereas Qwen3.5-9B does not benefit. Our best runs surpass the previously reported best scores on eight tasks, including Swap Reduction on Q20 and Rosetta, and match them on three more. EvoDuet also improves with other scaffolds (e.g., Top-K, EvoX) on Sums/Diffs and Denoising, demonstrating its applicability across evolutionary search scaffolds.
[NLP-3] MatLoom: Layered Text-to-Material Generation in a Compact Program Space
【速读】： 该论文旨在解决生成式材料（material generation）中仅关注外观而忽视其内在构建规则的问题，即如何在文本到材料的生成过程中同时实现视觉逼真性与可解释、可编辑的结构化表达。其核心解决方案是提出MatLoom——一种紧凑且基于层级的文本到材料生成语言，利用预训练语言模型实现端到端的程序化生成。关键在于通过带有透明度掩码（alpha-masked）的分层结构，使各层共享空间表达以显式定义覆盖范围、颜色与凹凸等物理基础渲染（PBR）通道之间的依赖关系，从而将图案、色彩与表面细节的生成逻辑编码为可执行的程序。该系统采用解析引导的修复与基于预览的批判机制，在无需特定任务微调的情况下迭代优化设计，并通过固定源代码、搜索噪声种子来探索多样化输出。实验表明，该方法在141个提示的基准测试中显著优于三种扩散基线模型，在所有四项平面布局对齐指标上均取得更高平均得分；且其初始生成程序已超越基线模型的平均BLIPScore。此外，经过盲测四选一对比，其渲染结果获得59.2%的选择率，远超最优基线的19.3%。该方法通过保持紧凑可执行的源代码（跨模型中位长度为21行），实现了高质量、可追溯、可编辑的材料资产生成。
链接: https://arxiv.org/abs/2609.40322
作者: Anson Y. Lam,Shuqing Li,Michael R. Lyu
机构: The Chinese University of Hong Kong (香港中文大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multimedia (cs.MM)
备注: 27 pages, 8 figures
Abstract:Material generation should produce not only an appearance, but also the rules that construct it. We introduce MatLoom, a compact, layer-oriented language for text-to-material generation with pretrained language models. Each program composes alpha-masked layers whose shared spatial expressions define coverage and physically based rendering (PBR) channels, making dependencies between patterns, color, and relief explicit. A standalone interpreter evaluates the program into material maps, while the source retains named fields and layer parameters for subsequent authoring. Without task-specific fine-tuning, our pipeline uses parser-guided repair and preview-based critique to revise material designs, then searches noise seeds while keeping each candidate’s remaining source fixed. On a curated benchmark of 141 prompts evaluated with six backbones, our best-performing configuration achieves higher mean scores than three diffusion baselines on all four flat-layout prompt-alignment metrics. Its initial programs already exceed all three baselines on mean BLIPScore, before critique or seed search. Retained programs have a median length of 21 lines when pooled across backbones. In a blind four-way comparison involving 30 participants and 20 prompts, our renders receive 59.2% of choices, compared with 19.3% for the most-preferred baseline. Compact executable programs thus offer a way to generate prompt-aligned materials while retaining their construction as part of the asset.
[NLP-4] Scaling Laws for Looped Mixture of Experts
【速读】： 该论文旨在解决现有模型扩展规律（scaling laws）在同时建模循环结构（looped transformers）与混合专家（Mixture-of-Experts, MoE）稀疏性时的局限性问题。传统扩展规律仅孤立地考虑递归或稀疏性，无法准确描述二者协同作用下的性能演化。为此，本文提出环路扩展定律（Loop Scaling Laws），首次联合建模模型规模、数据量、循环次数及稀疏性等多维度因素。其核心在于一个受约束的、依赖稀疏性的循环映射机制，能够量化循环带来的有效参数增益，并揭示稀疏性如何进一步放大该增益。该模型不仅显著提升了对环路模型外推损失的预测精度，且可退化为经典稠密模型与MoE扩展规律的特例。基于拟合结果，研究为在计算与内存约束下设计环路MoE模型提供了理论依据。下游实验验证了两者的互补优势：稀疏性实现约3倍活跃参数效率提升，循环结构在推理任务中带来约2倍总参数效率增益，而联合扩展进一步突破性能边界。作为实际应用延伸，研究证明这些增益在万亿级令牌规模下依然成立——在相同训练计算量下，基于定律设计的环路MoE模型在推理基准上表现相当于约2倍规模的非环路MoE，同时可通过循环实现测试时扩展（test-time scaling）。
链接: https://arxiv.org/abs/2609.40316
作者: Yanbei Chen,Anirudh Goyal,Raghuraman Krishnamoorthi
机构: Meta AI(元宇宙人工智能)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 19 pages
Abstract:Looped transformers and Mixture-of-Experts (MoE) offer complementary routes to efficient scaling: recurrence increases computational depth at fixed parameters, while MoE sparsity expands total capacity at fixed active compute. Yet existing scaling laws model recurrence or sparsity in isolation. In this work, we introduce Loop Scaling Laws, the first scaling law to jointly model recurrence and sparsity alongside model size and data. At its core is a bounded, sparsity-conditional recurrence mapping that characterizes the effective-parameter gain from looping and how sparsity raises this gain. The laws predict the held-out loss of looped models more accurately than prior alternatives, and recover the standard dense and MoE scaling laws as special cases. Beyond prediction, the fitted laws provide a principled foundation for designing looped MoE models under compute and memory constraints. Downstream evaluations further demonstrate the complementary benefits of the two axes: sparsity delivers ~3x active-parameter efficiency, recurrence yields ~2x total-parameter efficiency on reasoning, and joint scaling further advances the performance frontier. As a practical extension, we show these gains hold at trillion-token scale: at matched training compute, a looped MoE with law-derived recurrence matches a ~2x larger non-looped MoE on the reasoning benchmarks, while enabling test-time scaling through recurrence.
[NLP-5] How Much Is an AI Token Worth? Scaling Laws for Wild AI-Generated Web Text
【速读】： 该论文旨在解决日益增长的生成式AI（Generative AI）文本在互联网预训练数据中占比上升所带来的潜在影响问题，特别是其对语言模型预训练效果的影响。随着越来越多的网络文本由不同模型生成且以人类可读形式呈现，这些未经标注的“野生”生成文本（wild AI text）被直接纳入预训练语料库，可能对模型性能产生复杂甚至负面作用。研究发现，对于数据稀缺的模型，适量引入生成式AI文本初期可降低对人类文本的损失，但随比例增加，其收益迅速饱和并转为损害；而对于大规模人类文本预训练模型，生成式AI文本几乎立即导致损失上升，而同等数量的新鲜人类文本仍持续改善性能。现有如Hoffman等（2022）的缩放定律无法准确预测这一现象。为此，论文提出一种新的缩放定律，包含独立的增益与损害项，能够捕捉生成式AI token价值符号变化的动态特性，并在无生成式文本时退化为经典Chinchilla缩放律。该新模型在小规模模型上拟合后，能以41%更低误差预测高达3.6倍更大模型在不同生成式文本比例下的人类文本损失表现。研究建议：若目标是提升人类文本理解能力，则应过滤生成式文本，优先重复使用人类文本后再扩展数据集，并分别报告人类与生成式文本上的验证损失；当目标本身为生成式文本时，生成式内容仍具价值。相关数据集WildAI（830亿token）、全部800个模型及代码均已开源。
链接: https://arxiv.org/abs/2609.40295
作者: Jenna Russell,Ben Glickenhaus,Katherine Thai,John Wieting,Mohit Iyyer,Max Spero,Bradley Emi
机构: University of Maryland (马里兰大学); Pangram Labs (潘格拉姆实验室)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Web text makes up the majority of pretraining data and is increasingly AI-generated. After applying FineWeb quality filtering, we find that 27.5% of tokens from June 2026 web data are labeled as AI-generated by Pangram, rising to 31.1% by August. Unlike synthetic data or model-collapse setups, this wild AI text comes from many models, is written for human readers, and arrives unlabeled in pretraining corpora. How does AI text in the wild affect language model pretraining? To answer this question, we pretrain 800 language models, varying the ratio of added AI tokens to human tokens, and fit scaling laws to held-out losses on both human and AI-generated text. For data-starved models, adding AI tokens to pretraining data initially lowers loss on human text, but the benefit saturates as more are added and quickly reverses into harm. For models trained on high budgets of human text, AI tokens raise loss almost immediately, while the same number of fresh human tokens keeps lowering it. Scaling laws such as Hoffman et al. (2022) fail to predict this behavior. We propose a new scaling law with separate benefit and harm terms that allows the value of an AI token to change sign while also reducing to Chinchilla in the absence of AI text. When fit on smaller models, our scaling law predicts the effect of AI text on held-out human-text loss for models up to 3.6x larger with 41% lower error than the best existing law over all AI ratios. We recommend filtering AI text when the target is human text, repeating human text before expanding the training dataset with AI-generated web text, and reporting validation loss on human and AI text separately AI text remains valuable when the target is AI text. We release WildAI, an 83B-token corpus with AI, topic, and format labels, all 800 models and code at this https URL.
[NLP-6] Linguistic Loopholes in LLM Unlearning: From a 174-Language Benchmark to Coverag e-Aware Unlearning
【速读】： 该论文旨在解决多语言场景下知识遗忘（unlearning）的跨语言漏洞问题，即在一种语言中成功删除特定事实后，该知识仍可能通过其他语言的查询或答案表达形式被恢复，导致遗忘不彻底。其核心挑战在于如何在有限的语言资源预算下实现高效的跨语言知识擦除。解决方案的关键是提出“语言预算约束的多语言遗忘”任务，并引入交叉语言遗忘张量（Cross-Lingual Unlearning Tensor）作为基准，系统评估25种原子化改写类型在174个语言-文字对中的遗忘泛化能力。在此基础上，提出COVER方法，通过选择最优源语言子集以最大化未受监督目标语言的预测遗忘覆盖率，在仅需无害校准数据和冻结模型的前提下实现高效、可扩展的跨语言遗忘。实验表明，相比均匀选择源语言，COVER在三种模型族和两组独立遗忘数据上显著降低平均保留访问率7.8%–27.3%，且在低资源语言的真实新闻文档（基于LORELEI语料库）中亦表现出优越性能，验证了其有效性与泛化能力。
链接: https://arxiv.org/abs/2609.40286
作者: Tyler Skow,Shravan Chaudhari,Rama Chellappa,Abhay Yadav
机构: Johns Hopkins University (约翰霍普金斯大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Unlearning a fact in one language does not guarantee its removal in others as changing the query or even the requested answer language can reopen seemingly forgotten knowledge – a cross-lingual loophole. The most straightforward solution to this challenge – unlearning in all languages – is neither scalable nor desirable as it amplifies damage to unrelated model capabilities. We introduce the task of language budgeted multilingual unlearning where the goal is to select a subset of languages that maximizes cross-lingual erasure. To study this task we introduce the Cross-Lingual Unlearning Tensor, an unlearning benchmark that spans 174 language–script pairs and 25 atomic paraphrase types to examine when forgetting generalizes across linguistic expressions of the same knowledge. We further propose COVER, which selects source languages to maximize predicted COVERage of languages receiving no forget supervision, enabling unlearning on a language budget. Surprisingly, we find naively selecting strong individual sources does not reliably compose into strong source sets motivating our development of COVER. At deployment COVER only requires benign calibration data and access to the frozen model. Across three model families and two disjoint forget sets, COVER reduces mean held-out residual access by 7.8–27.3% relative to uniform source selection. We find these gains extend beyond synthetic benchmarks to real news documents in low-resource language settings using human translated data from the Low Resource Languages for Emergent Incidents (LORELEI) corpus.
[NLP-7] cua-speedrun: Standardized Benchmarking of the Speed of Computer-Use Agents
【速读】： 该论文旨在解决当前计算机使用代理（Computer Use Agents, CUAs）在评估其执行速度与效率时面临的可复现性危机问题。现有基准测试依赖于复杂且异构的硬件与容器配置，导致对CUA实际性能的衡量存在显著偏差。为此，论文提出cua-speedrun，其核心解决方案在于构建标准化的基础设施与任务集，采用统一的虚拟机环境、一致的执行流程以及通用的代理接口，从而实现跨基准的无缝比较。关键创新包括：通过标准化设置消除环境差异对性能评估的干扰；揭示推理开销、代理框架及环境延迟对任务完成时间与成本的非线性影响，发现增加推理努力反而可能提升效率，且更快的输入输出处理未必缩短总耗时；同时证明可通过缩减评估任务集在不损失统计效力的前提下显著提升基准测试效率。该框架为实现快速、高效的CUA发展提供了可重复、可比较的评估基础，推动其在真实场景中的应用落地。
链接: https://arxiv.org/abs/2609.40284
作者: Pranjal Aggarwal,Lawrence Keunho Jang,Sean Welleck,Daniel Fried,Ruslan Salakhutdinov,Jing Yu Koh
机构: 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:Computer use agents (CUAs), which use graphical user interfaces (GUIs) to complete tasks on a computer, have recently surpassed human performance on many standard benchmarks, including difficult long-horizon tasks. Their capabilities are undoubtedly impressive, however, a key barrier to the widespread adoption and deployment of CUAs remains their speed and cost. Progress towards faster yet capable CUAs requires reliable evaluation of their speed, but many CUA benchmarks currently face a reproducibility crisis. Benchmarks are based on complex infrastructure with varying machine and container configurations that confound the evaluation of the execution speed of CUAs. Towards addressing this gap, we propose cua-speedrun, which introduces standardized infrastructure and task sets, with a focus on evaluating the speed and efficiency of CUAs. cua-speedrun uses a uniform virtual machine setup and execution pipeline, along with a common agent interface that enables single-agent implementations to operate seamlessly across different benchmarks. Across four different CUA benchmarks, we evaluate how reasoning effort, agent harnesses, and environment latency affect performance, speed, and cost. We find no single model family is optimal for all three; none of the open-weight models are on the frontier, and also, unintuitively, for some models increasing the reasoning effort can speed up task completion, while faster environment input-output can slow down overall task completion time. We also demonstrate that we can effectively reduce the evaluation task set of most CUA benchmarks without degrading overall statistical power, allowing for more efficient benchmarking and comparison. We believe cua-speedrun will enable structured progress towards fast, efficient CUAs, unlocking new real-world use cases and applications. All code, infrastructure, and analysis are available at this https URL.
[NLP-8] Comparison of techniques for fine-tuning open-weight models for entity extraction from radiology reports
【速读】： 该论文旨在解决医学影像报告中结构化标签提取的隐私、成本与可复现性问题，具体聚焦于非增强头颅CT报告中颅内出血（intracranial hemorrhage, ICH）严重程度的多标签提取任务。现有高性能标签提取器多依赖专有模型，存在数据隐私泄露风险且难以复现。研究提出以开源权重模型Gemma-3-12B为基础，通过微调实现对真实临床报告的高效结构化标注，并探究不同微调策略与训练数据来源的影响。关键发现表明：训练数据源是决定性能的关键因素——基于真实GPT-4o标注报告进行知识蒸馏（distillation）所构建的指令微调模型（DIFT），在宏平均F1值上达到0.845，与GPT-4o（0.850）无显著差异（p=1.000），显著优于未微调的基线模型（提升0.178）。相比之下，使用GPT-4o生成的合成报告进行训练的模型无论采用何种微调方法均表现不佳，始终低于未微调的开源基础模型。此外，所有微调与推理过程均可在单块24 GB消费级GPU内存范围内完成。因此，对于高价值、窄范围的临床标签提取任务，利用真实报告进行知识蒸馏而非生成合成数据，是缩小与专有模型差距的核心，从而实现私密、低成本、版本稳定的本地部署替代方案。
链接: https://arxiv.org/abs/2609.40236
作者: Aawez Mansuri,Kush Mehta,Mohammadreza Chavoshi,Jahanzaib Malik,Theodorus Dapamede,Frank Li,Rohan Isaac,Beatrice Brown-Mulry,Chiratidzo Rudado Sanyika,YoungSeok Jeon,Judy W. Gichoya,Ali Emami,Hari Trivedi
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Converting free-text radiology reports into structured labels supports cohort building, quality assurance, and monitoring of clinical imaging models, but the strongest label extractors are hosted proprietary models whose use raises privacy, cost, and reproducibility concerns. We asked whether a fine-tuned open-weight model (Gemma-3-12B) can match GPT-4o at multi-label intracranial hemorrhage (ICH) acuity extraction from non-contrast head-CT reports, and which ingredients matter. Using a 2x2 design, we crossed two adaptation strategies (a discriminative classification head, CH; generative instruction fine-tuning, IFT) with two training-data sources (distillation of real GPT-4o-labeled reports; synthetic reports generated by GPT-4o from real exemplars), across five training sizes, benchmarked on 100 expert-adjudicated reports against GPT-4o and the un-tuned open-weight base. The distilled instruction-tuned model (DIFT) matched GPT-4o (macro-F1 0.845 vs 0.850; p = 1.000) and exceeded the base model by 0.178. The decisive factor was the training-data source, not the fine-tuning method: both synthetic-data models failed to exceed the un-tuned open-weight base at any training size and underperformed the distilled models across all acuity classes. Fine-tuning and inference fit within the memory envelope of a single 24 GB consumer GPU. For narrow, high-value clinical label-extraction tasks, distilling real reports, rather than generating synthetic ones, is what closes the gap to a hosted model, enabling a private, low-cost, version-stable on-premises alternative.
[NLP-9] Distribution Matching Distillation for Continuous Diffusion Language Models
【速读】： 该论文旨在解决连续扩散语言模型在生成高质量文本时需进行大量网络评估（NFEs）所带来的计算成本高昂问题。其核心挑战在于如何在保持生成质量的前提下显著降低所需的NFE数量。解决方案的关键在于引入分布蒸馏（distributional distillation），通过利用学生模型（student model）的概率化标记输出来优化梯度估计，从而更高效地逼近目标分布。文中提出两种基于相同学生架构与反向KL散度匹配目标的方法：Simplex-DMD采用连续标记松弛与路径梯度（pathwise gradients），而Reinforce-DMD则结合类别采样与带有学习密度比的REINFORCE算法。两者均适用于多步生成任务，并针对不同参数化方式分析了训练与采样策略的影响。实验结果表明，在OpenWebText数据集上，对于1,024个标记的序列，Simplex-DMD仅用4次NFE即达到45.6的生成困惑度（perplexity）与5.44纳特的单字熵，相较最强基准在相同熵和采样预算下降低49%；Reinforce-DMD在更大评估预算下表现更优，以256次NFE实现14.9的困惑度与5.00纳特熵，提升幅度达20%，验证了该方法在高效率与高质量生成之间的良好权衡。
链接: https://arxiv.org/abs/2609.40235
作者: Paul Le Van Kiem,Dario Shariatian,Umut Simsekli,Alain Durmus
机构: Inria, PSL Research University (法国国家信息与自动化研究所，巴黎文理研究大学); Cohere; CMAP, Ecole Polytechnique (应用数学中心，巴黎综合理工学院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Machine Learning (stat.ML)
备注:
Abstract:Continuous diffusion language models generate all tokens in parallel, yet high-quality generation can still require hundreds of network evaluations (NFEs). We study how distributional distillation can reduce this cost by exploiting the student’s probabilistic token outputs. Our unified formulation connects the student’s output parameterization to the resulting gradient estimators and yields two methods with the same student architecture and reverse-KL matching objective: Simplex-DMD uses continuous token relaxations and pathwise gradients, while Reinforce-DMD uses categorical sampling and REINFORCE with a learned density ratio. We develop both methods for multi-step generation and investigate the training and sampling choices associated with each parameterization. On OpenWebText, for sequences of 1,024 tokens, Simplex-DMD achieves a generative perplexity of 45.6 at a unigram entropy of 5.44 nats in just 4 NFEs, a 49% reduction relative to the strongest evaluated diffusion baseline at matched entropy and sampling budget. Reinforce-DMD improves the frontier at larger budgets, reaching a generative perplexity of 14.9 at an entropy of 5.00 nats with 256 NFEs, a 20% reduction under the same comparison protocol.
[NLP-10] PhantomEnvironments: Training LLM Agents in Fictional Worlds
【速读】： 该论文旨在解决大语言模型（LLM）代理在强化学习（Reinforcement Learning, RL）训练过程中面临的环境瓶颈问题，即现有环境需提供可验证的奖励信号、支持长时程交互且具备低成本可扩展性，而当前方法依赖昂贵的人工标注数据或由大语言模型生成的环境，存在幻觉和基准污染风险。其解决方案的关键在于提出一种完全基于规则生成的合成环境——PhantomEnvironments，该环境通过虚构世界中的模板化文章构建多轮交互场景，使代理需在文档集合中进行多跳推理以回答复杂问题。此类环境无需使用任何大语言模型生成内容，且生成成本为零，具备高度可扩展性。实验表明，尽管这些环境与真实世界无事实关联，但在此类环境中训练的代理仍能有效迁移到真实世界的多跳搜索基准任务中，并在新基准上表现优于基于真实数据训练的模型。此外，代理展现出对未见虚构宇宙的泛化能力，以及随问题难度线性调整搜索预算的涌现行为，说明仅通过环境交互即可催生出具有可扩展性的搜索策略。消融实验进一步揭示，问答跳数（hop count）是决定迁移效果的核心因素，远超约束条件或比较逻辑等复杂度维度，表明最简单的规则生成环境已可作为训练通用化LLM代理的高效免费资源。
链接: https://arxiv.org/abs/2609.40221
作者: Anmol Kabra,Swathi Saravana Selvam,Albert Gong,Chao Wan,Christian Belardi,Dongyoung Go,Katie Z. Luo,Kilian Q. Weinberger
机构: Cornell University(康奈尔大学); Stanford University(斯坦福大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Training LLM agents with reinforcement learning (RL) is bottlenecked by environments, which must provide verifiable rewards, support long-horizon interaction, and scale cheaply. Existing approaches rely on costly human-curated data or on LLM-generated environments that risk hallucinations and benchmark contamination. We show that LLMs can instead be trained into capable search agents using synthetic environments generated entirely by rules, whose generation requires no LLM and has zero marginal cost. We build PhantomEnvironments, multi-turn RL environments from fictional worlds, where agents must search a corpus of templated articles to answer multi-hop questions. Despite sharing no facts with the real world, these strikingly simple environments yield agents that transfer to real-world multi-hop search benchmarks, often outperforming real-world training data on newer benchmarks. Trained agents generalize to unseen fictional universes, and Qwen models learn to scale their search budget roughly linearly with question difficulty, suggesting emergent search scaling from environment interaction alone. Ablating environment complexity reveals that hop count drives transfer more than constraints or comparisons: even the simplest rule-generated environments are a surprisingly effective, free resource for training generalizable LLM agents.
[NLP-11] SCB: SpeechConversationBench for Evaluating Multi-Turn Reasoning in Speech-to-Speech Models
【速读】： 该论文旨在解决语音对话系统在多轮交互中处理复杂任务时面临的挑战，特别是针对需要跨对话轮次逐步获取信息的数学推理任务。其核心问题是：现有商业语音系统在面对逐步披露信息（sharded）的口语化输入时，性能显著下降，反映出对增量式语音交互的适应能力不足。解决方案的关键在于构建SpeechConversationBench（SCB）评估框架，通过对比单轮完整问题（full）、信息碎片合并后一次性呈现（concat）以及逐轮增量式口语披露（sharded）三种条件下的表现，系统性地量化模型对动态对话上下文的理解与管理能力。实验表明，引入显式对话上下文管理机制的自研语音管道LEGO在所有条件下均保持77.5%的高准确率，显著优于其他商业系统，证明了有效上下文管理是提升生成式语音系统在复杂对话场景中表现的关键。
链接: https://arxiv.org/abs/2609.40198
作者: Kanpat Vesessook,Saksorn Ruangtanusak
机构: SCBX RD; SCB DataX, SCBX Group
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Sound (cs.SD)
备注: Conducted during a 2024 internship at SCBX RD
Abstract:Speech-to-speech systems must solve tasks whose requirements emerge across conversational turns. We introduce SpeechConversationBench (SCB), a focused evaluation of spoken mathematical reasoning using 103 sharded GSM8K problems. The framework compares the original problem delivered in one turn (full), its concatenated information shards delivered together (concat), and incremental spoken disclosure across turns (sharded). We report final-answer accuracy for four commercial speech systems and LEGO, a proprietary speech pipeline developed internally by the SCBX Innovation Lab team with explicit conversational context management. Relative to concat, sharded accuracy decreases by 5.0-25.3 percentage points across the four commercial systems. LEGO achieves 77.5 percent accuracy in all three conditions, compared with 76.6 percent sharded accuracy for GPT-4o Realtime. The two single-turn baselines distinguish sensitivity to problem reformulation from the additional challenges introduced by incremental spoken interaction.
[NLP-12] MemLife: Curating and Reasoning over Long-Term Egocentric Video Memories
【速读】： 该论文旨在解决长期第一人称视频（long-term egocentric video）在构建个性化智能助手时面临的计算效率与记忆有效性难题。随着视频数据积累至数百小时，每次查询均重新处理原始视频片段已不可行，而现有记忆系统常因关键证据丢失或检索竞争导致相关条目难以定位，从而限制了实际应用效果。为此，论文提出MemLife，一种多模态记忆系统，通过构建以实体为锚点的、第一人称视角的文本事件（text episodes），并采用时间索引的代理阅读器（time-indexed agentic reader）进行高效检索，实现无需训练和查询时视频访问下的性能提升，在四个长时程基准测试中相较最强无训练基线提升4.6%–12.0%。为进一步优化记忆质量，论文提出MemOpt，一种基于强化学习的框架，用于优化记忆生成器（memory writer），使其产出更忠实、信息量更大且可检索性强的记忆内容。该方法在不同视频与问题分布下持续带来2.7%–5.0%的性能增益，并展现出对不同写入器与阅读器模型架构及记忆系统结构的良好泛化能力。
链接: https://arxiv.org/abs/2609.40195
作者: Guangzhi Xiong,Xinyuan Zhang,Xiao Yang,Hyokun Yun,Kai Zhang,Shiun-Zu Kuo,Hyeonjeong Ha,Xilun Chen,Kai Sun,Lucas Liang,Guangqiang Dong,Ejaz Ahmed,Ahmed A Aly,Anuj Kumar,Raffay Hamid,Aidong Zhang,Xin Luna Dong
机构: Meta Reality Labs(元宇宙现实实验室); University of Virginia(弗吉尼亚大学); University of Illinois Urbana-Champaign(伊利诺伊大学厄本那-香槟分校)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Long-term egocentric video enables personalized AI assistants to reason about daily life. However, as video histories grow to hundreds of hours spanning months or years, reprocessing raw clips for every query becomes computationally prohibitive. Memory systems offer a scalable alternative by compacting videos into text representations, but often fail on practical benchmarks: either the memory does not preserve key evidence, or the retriever fails to locate relevant entries due to retrieval competition in growing search spaces. To address these challenges, we introduce MemLife, a multimodal memory system that constructs entity-grounded, first-person text episodes and retrieves them via a time-indexed agentic reader. Without training or query-time video access, MemLife improves over the strongest training-free baseline by 4.6–12.0% across four long-horizon benchmarks. To further improve memory quality, we propose MemOpt, a reinforcement learning framework that optimizes the memory writer to produce faithful, informative, and retrievable memories. MemOpt consistently improves MemLife by 2.7–5.0% across different video and question distributions, with gains that generalize across writer and reader backbones and memory systems.
[NLP-13] Cheap to Draw Expensive to Trust: Certifying Test-Time Scaling Curves
【速读】： 该论文旨在解决在测试时通过采样多个答案并选择验证器评分最高的答案以提升准确率的过程中，如何高效且可信地对不同采样数量 k 下的准确率缩放曲线进行统计验证的问题。核心挑战在于传统方法需要大量生成答案才能在固定置信水平下对多个预算（budget）进行可靠认证，而其中大部分计算成本被“错误的不确定性”所消耗——即对跨问题方差的过度保守估计。其解决方案的关键在于提出一种基于配对审计（paired audit）的最小化最大成本（minimax cost）认证框架，该框架将总成本分解为三部分：校准得分分布尾部、区分不同问题间的差异、以及沿缩放曲线累积的组内噪声。特别地，在单个基准测试中，最后一项成本退化为最优答案分配下单个答案影响的方差，任何有效审计都必须承担此部分成本，而自适应学习分配策略的审计可在此基础上接近最优性能（仅差一个对数因子）。该方法利用同一问题上两次独立采样之间的指数不等式构建无需预实验的配对审计，实证表明在185个保留得分池上，其所需答案数仅为最便宜的竞争性认证方法的0.74倍（在$ k=64 时）和0.53倍（在 k=1,024 $时），并在新生成的MMLU-Pro研究中仅用79,133个答案即实现了与预测成本相差不足0.6%的精度，显著提升了效率与可靠性。该方法还可推广至 pass@k 和多数投票等场景，并适用于问题总体、依赖前序答案的答案序列等复杂情形。
链接: https://arxiv.org/abs/2609.40190
作者: Sohail(Neel)Sarkar,Shakuntala Baichoo
机构: PMCC AI Lab, Peter Munk Cardiac Centre (PMCC人工智能实验室，彼得·蒙克心脏中心); University Health Network (大学健康网络)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Statistics Theory (math.ST); Machine Learning (stat.ML)
备注: 32 pages, 10 figures, 5 tables
Abstract:Sampling several answers and keeping the one a verifier scores highest is one of the simplest ways to buy accuracy at test time. Its effect is reported as a scaling curve: accuracy against the number k of sampled answers. The curve is cheap to draw and expensive to trust. A budget read off it is chosen after looking at every point, so only a band that covers all budgets at once protects the choice, and on a 100-question benchmark a fixed exact-binomial design needs 192,000 generated answers to certify 64 budgets to within \pm1/32 at 95%. Most of that cost pays for the wrong uncertainty. A benchmark is a fixed list of questions; at budget 64, about three quarters of the variance of a selected answer’s correctness lies between questions, and an audit that revisits every question need not pay for it. We derive the minimax cost of certifying the whole curve, up to logarithmic factors. It has three parts: calibrating the tail of the score distribution, telling the questions apart, and within-question noise summed along the curve. At a single benchmark the last part sharpens to the variance of one answer’s influence under the best allocation of answers to questions, which every valid audit pays and an audit that learns the allocation attains, up to a logarithm, as the precision grows. A paired audit built on an exponential inequality for two independent draws at the same question needs no pilot. On 185 held-out score pools it uses 0.74 times the answers of the cheapest competing certified audit at 64 budgets and 0.53 times at 1,024, and on a newly generated MMLU-Pro study it certified the curve with 79,133 answers, within 0.6% of what a cost law fitted beforehand predicted from the study’s within-question variance. The same paths certify pass@ k and majority voting, and the bands extend to populations of questions and to answers that depend on earlier ones.
[NLP-14] Provably Tractable NFA-Constrained Language Generation via HMMs
【速读】： 该论文旨在解决在语言模型（LM）生成过程中对非确定有限自动机（NFA）约束进行高效且分布无偏采样的问题。现有方法在处理NFA约束时，要么导致生成分布失真，要么牺牲计算效率。其核心挑战在于，该任务本质上等价于计算被NFA接受的长度为n的序列数量（#NFA），而精确求解#NFA问题是#P-完全的。受近期研究中#NFA存在全多项式随机近似方案（FPRAS）的启发，本文提出NFA-LM，一种在多项式时间内实现NFA约束生成的引擎，并在温和假设下具备理论保证。其解决方案的关键在于将FPRAS引入语言模型采样过程，从而在保证生成效率的同时，实现理论上可界定的近似误差，实验验证了其在生成高质量输出方面的有效性与实用性。
链接: https://arxiv.org/abs/2609.40185
作者: Jialiang Sun,Kuldeep Meel
机构: University of Toronto(多伦多大学)
类目: Computation and Language (cs.CL); Formal Languages and Automata Theory (cs.FL)
备注:
Abstract:Constrained generation aims to sample from language models (LMs) conditioned on hard constraints. Existing constrained-generation techniques for nondeterministic finite automaton (NFA) constraints either distort the distribution or sacrifice efficiency. Theoretically, this task reduces to counting the length- n sequences accepted by an NFA (#NFA), and the exact #NFA problem is #P-complete. Recent work has shown that #NFA admits a fully polynomial randomized approximation scheme (FPRAS). Inspired by this result, we propose NFA-LM, a polynomial-time engine for NFA-constrained generation with theoretical guarantees under mild assumptions. Experiments show that NFA-LM efficiently generates high-quality outputs with theoretically bounded approximation error.
[NLP-15] Index-Translate: A Multilingual Translation Model Family – Text Speech Controlled Dubbing and Long-Document Translation
【速读】： 该论文旨在解决多语言翻译任务中模型泛化能力不足、专用场景适配性差以及长文档与语音翻译等复杂任务处理能力有限的核心问题。其解决方案的关键在于提出一个统一的多语言基础架构——Index-Translate 模型家族，通过共享的多语言主干网络结合针对不同下游任务（如通用翻译、指令遵循、语音翻译、可控配音和长文档翻译）的专项微调策略，实现高效且灵活的任务适应。该方法不仅支持150种语言的多语言翻译与指令遵循，还在通用翻译与复杂指令理解上超越同规模模型，并达到100B级模型的性能水平；同时，衍生出的Index-Echo（端到端语音翻译）、Index-Homura（音节级配音控制）和Index-NativeLong（原生长文档翻译）分别在语音转写、精准配音和长文本处理方面实现了突破，显著提升了多模态、多场景下的翻译系统整体能力。
链接: https://arxiv.org/abs/2609.40181
作者: Tianjiao Li,Mengran Yu,Chenyu Shi,Lusheng Zhang,Qisi Chen,Yanshan Zhou,Ji Qi,Jingying Liu,Yuang Feng,Ziang Cui,Tianxing Yan
机构: Qisi Chen; Yanshan Zhou; Ji Qi; Jingying Liu; Yuang Feng; Ziang Cui; Tianxing Yan; Index LLM Team
类目: Computation and Language (cs.CL)
备注: 27 pages. Project: this https URL ; Code and models: this https URL
Abstract:We introduce Index-Translate, a multilingual translation model family that combines a shared multilingual foundation with specialized training for general translation, instruction following, speech translation, controlled dubbing, and long-document translation. It includes three model sizes, 2B, 9B, and 35B-A3B, and supports translation in 150 languages, with multilingual instruction following. Evaluations on general translation and complex translation instructions show that Index-Translate outperforms translation models of comparable size and achieves performance comparable to 100B-scale translation models and frontier models. Index-Echo provides end-to-end speech-to-text and speech-to-speech translation, outperforming existing end-to-end models and achieving performance comparable to frontier omni models. Index-Homura extends the family to syllable-controlled dubbing. Index-NativeLong introduces native long-document translation with a dedicated task formulation and benchmark. These capabilities support diverse translation tasks, including multilingual content production.
[NLP-16] Learning Functional Subspaces for Neural Network Compression
【速读】： 该论文旨在解决大模型（如Transformer）在高压缩比下性能急剧下降的问题，核心挑战在于现有低秩权重分解方法依赖局部闭式准则（如激活能量、层间重构误差或损失的二次近似）来选择要丢弃的子空间，忽视了误差在网络深度传播的累积效应，导致在高压缩率时模型表现严重退化。其解决方案的关键是提出可学习子空间投影（Learnable Subspace Projections, LSP），通过端到端优化的方式学习每个线性层或共享激活的层组所应丢弃的正交投影子空间。所有投影矩阵在保持预训练权重冻结的前提下，联合优化以最小化与原始密集模型输出分布之间的KL散度或原始训练损失。投影矩阵初始化基于白化SVD截断，并根据每节省一个参数所引发的输出KL增益动态分配秩。训练完成后，投影矩阵合并为标准的低秩因子，同一组层共享单一因子，在注意力机制中还可将完整的键值对压缩为一个窄的共享潜在表示。实验表明，LSP在多种语言模型（OPT-125M/1.3B、Qwen3-4B、Llama-2-7B）和视觉模型（ViT-B/16）上均显著优于基线方法，尤其在高压缩比下优势更明显：例如在-70%压缩率下，Llama-2-7B的WikiText-2困惑度降至10.9（基线为13.3），零样本准确率达到42.2%（基线为36.0%）。此外，因子化模型在小批量推理时速度提升达1.6倍，且在128k上下文长度下，权重与键值缓存总内存占用相比非共享基线降低13.5倍（最大仅6.5倍），实现了性能与效率的双重突破。
链接: https://arxiv.org/abs/2609.40127
作者: Massimo Bini,Anders Christensen,Stephan Alaniz,Judah Goldfeder,Ole Winther,Yann LeCun,Ravid Shwartz-Ziv,Zeynep Akata
机构: Helmholtz Munich(赫尔姆霍兹慕尼黑); Technical University of Munich(慕尼黑工业大学); MCML; Orbital Industries; LTCI, Télécom Paris, Institut Polytechnique de Paris(电信巴黎理工学院电信与网络研究所); Columbia University(哥伦比亚大学); University of Copenhagen(哥本哈根大学); Technical University of Denmark(丹麦技术大学); New York University(纽约大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Modern transformers pair impressive capabilities with substantial memory and compute demands. Low-rank weight factorization reduces both while keeping the matrices dense, and thus efficient on standard hardware. Existing methods, however, choose the subspace to remove from each weight matrix with local closed-form criteria: activation energy, layer-wise reconstruction error, or a quadratic approximation of the loss. These criteria ignore how errors propagate through the network, so at high compression the errors compound with depth and performance collapses. We introduce Learnable Subspace Projections (LSP), which instead learns the subspaces to discard end-to-end. Each linear layer, or tied group of layers that read the same activations, is assigned an orthogonal projector. All projectors are optimized jointly against a global objective–the KL divergence to the dense model’s output distribution or the model’s original training loss–while the pretrained weights remain frozen. Projectors are initialized from a whitened SVD truncation, and ranks are allocated by the output KL each projector induces per parameter saved. After training, the projectors merge into standard low-rank factors, with each tied group sharing one factor. In attention, this also lets the model cache one narrow latent in place of full keys and values. Across LLMs (OPT-125M/1.3B, Qwen3-4B, Llama-2-7B) and ViT-B/16, LSP outperforms baselines, and its advantage widens as compression increases. At -70% compression, LSP brings Llama-2-7B to 10.9 WikiText-2 perplexity and 42.2% mean zero-shot accuracy, versus 13.3 and 36.0% for the strongest baseline. The factorized model decodes up to 1.6x faster than the dense model at small batch sizes, and aching the shared latent shrinks the combined memory of weights and KV cache by 13.5x at a 128k-token context, versus at most 6.5x for untied baseline factorizations.
[NLP-17] Debias It Yourself: Teaching LLM s Cognitive Bias Mitigation Interventions
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）中存在的偏见问题，即模型在生成内容时可能表现出对特定群体的刻板印象或歧视性倾向。尽管社会心理学与认知科学领域已积累大量经验证的减偏干预措施，但如何将这些人类层面的有效策略系统性地迁移至机器学习模型中仍是一个关键挑战。本文提出的“自我去偏”（Debias It Yourself, DIY）框架，其核心创新在于将五种经过验证的人类减偏干预方法转化为适用于大语言模型的可操作流程，并通过三种成熟范式实现：展示（In-context examples）、训练（Instruction tuning）和修订（Guided self-revision）。其中，关键突破在于“修订”（Revise）机制——通过引导模型进行自我反思与修正，显著降低模型在未见过的偏见维度上的表现偏差，同时保持较高的推理准确性。实验结果表明，仅使用“修订”或结合“训练+修订”的方案，在多个基准测试中均取得最优平均排名，实现了低偏见（最低达2%）与高推理准确率（90%）之间的良好权衡，并在未见偏见维度上最大减少14.8%的偏见。
链接: https://arxiv.org/abs/2609.40124
作者: Chahat Raj,Sina Mansouri,Aylin Caliskan,Antonios Anastasopoulos,Ziwei Zhu
机构: George Mason University(乔治梅森大学); University of Washington(华盛顿大学)
类目: Computation and Language (cs.CL)
备注: Under Review
Abstract:Bias has long been studied in social psychology and cognitive science, where decades of research have produced a body of validated interventions that reduce stereotypical thinking and prejudiced responses in humans. We propose Debias It Yourself (DIY), a cognitively grounded framework that translates five such interventions into debiasing procedures for large language models and delivers them through three established paradigms: Show (in-context examples), Train (instruction tuning), and Revise (guided self-revision). Across three models, five bias benchmarks, eleven debiasing baselines, and three reasoning benchmarks, Train+Revise and Revise alone attain the top two average ranks, lead the bias-reasoning tradeoff (mean bias as low as 2% at 90% reasoning accuracy), and reduce bias on unseen dimensions by up to 14.8%. Our code and data are publicly available.
[NLP-18] On the (In)effectiveness of AMR Augmentation for Large Language Models EMNLP2026
【速读】： 该论文旨在解决生成式大语言模型（Large Language Models, LLMs）在引入抽象语义表示（Abstract Meaning Representation, AMR）进行增强后，其在下游自然语言处理任务中是否真正受益的问题。此前有研究声称AMR增强可显著提升模型性能，但本文通过严格复现相关实验发现，这些提升效果可能源于实验设置中的特定偏差，如不一致的超参数选择策略。在采用统一、规范的超参数调优协议后，仅使用文本输入的基线模型始终表现与或优于经AMR增强的模型，表明AMR增强并未带来实质优势。为深入探究这一“零结果”的成因，作者提出一种基于困惑度（perplexity）的探测方法，用于衡量AMR能否为LLM提供模型原本无法获取的语义关系知识。结果显示，AMR增强并未帮助模型更准确地理解句子中的关系性内容，因此在下游任务中缺乏明确收益。该研究的关键结论是：当前主流大语言模型对额外的结构化语义信息（如AMR）具有较低的利用能力，其性能提升依赖于原始文本中的隐含知识，而非外部语义增强。
链接: https://arxiv.org/abs/2609.40121
作者: Hoa Quynh Nhung Nguyen,Jacopo Staiano,Michael Sullivan
机构: Saarland University (萨尔兰大学); University of Trento (特伦托大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 23 pages, 6 figures, 18 tables, accepted at EMNLP 2026
Abstract:While Abstract Meaning Representation (AMR) has historically improved performance on a range of NLP tasks, the benefit—or lack thereof—of AMR augmentation for modern LLMs is thus far unclear. In this paper, we attempt to reproduce recent work that reported substantial downstream gains from AMR augmentation, finding that these are likely due to specific choices in the experimental settings used: using a consistent and unified protocol for hyperparameter selection, we observe that text-only baselines consistently match or exceed the performance of AMR-augmented models. To investigate this null result, we introduce a perplexity-based probe measuring the degree to which AMR provides an LLM with supplemental relational knowledge not already available to the model. We find that AMR augmentation does not help LLMs improve their understanding of relational content in the sentence, indicating that augmenting these models with AMR offers no clear benefit on downstream tasks.
[NLP-19] Persistent Context Graphs for Efficient Memory Compaction in LLM Agents
【速读】： 该论文旨在解决大语言模型（LLM）在处理长期任务时因交互历史过长而导致的上下文窗口溢出与预填充（prefill）成本高昂的问题。随着代理（agent）执行复杂任务的时间跨度增加，如何高效压缩历史记忆以维持性能并降低计算开销成为关键挑战。现有方法通常依赖于对历史进行摘要或压缩键值缓存（KV cache），但往往需要额外的模型计算来保留信息，且在新请求到来时难以动态调整重要性评估——尤其是当历史缓存失效后，必须重新编码整个历史才能重新评估其相关性。本文提出ReCAP（Recall-Aware Context Pruning），其核心创新在于构建一个轻量级、持久化的上下文图（context graph），将基于注意力机制提取的历史重要性得分与消息间的依赖关系作为元数据存储。在每轮新请求中，ReCAP结合存储的重要性分数与当前请求带来的相关性线索，并沿依赖链检索关键消息及其支持性上下文，无需额外的模型调用即可完成记忆选择。实验表明，相比Codex默认的基于摘要的压缩方式，ReCAP在Qwen3-Coder和gpt-oss上分别将压缩与冷恢复（cold restoration）的估计延迟降低约95%；在SWE-Together基准上，在保持任务质量相当的前提下，将每轮调用的历史上下文减少近一半，并在Lost-in-Conversation代码任务中分别提升19.8和41.2个准确率点，显著优于使用完整历史的基线。因此，其解决方案的关键在于：利用注意力机制生成的可解释性元数据构建持久化上下文图，实现无需重编码的动态、高效、精准的记忆选择。
链接: https://arxiv.org/abs/2609.40118
作者: Jingbo Yang,Kwei-Herng Lai,Xiaowen Wang,Zhaoxuan Tan,Pei Zhou,Mengting Wan,Yaar Harari,Evgeniy Gabrilovich,Shiyu Chang
机构: University of California, Santa Barbara(加州大学圣塔芭芭拉分校); Microsoft(微软); University of Notre Dame(圣母大学); Microsoft(微软)
类目: Computation and Language (cs.CL)
备注:
Abstract:As LLM capabilities advance, agents are tackling increasingly complex tasks over longer horizons. Their growing interaction histories make memory compaction essential for staying within context windows and reducing prefill cost. Existing methods summarize the history or compress its KV cache, often adding model computation to preserve information for future requests. A new user request can change which history matters, but reassessing that history with the model requires re-encoding it if the KV cache has expired. Past attention provides signals of historical importance and dependencies between messages, while relevance to the current task must be assessed using the new user request. We introduce ReCAP, a memory compaction method that stores attention-derived importance scores and dependency links in a lightweight, persistent context graph. For each new request, ReCAP combines stored importance with relevance cues from the request and follows dependency links to select messages and their supporting context, without additional model calls for selection. Compared with Codex’s default summarization-based compaction, ReCAP reduces estimated latency for compaction and cold restoration by approximately 95% on both Qwen3-Coder and gpt-oss. It also roughly halves the historical context per call on SWE-Together at comparable task quality and improves accuracy on the code tasks of Lost-in-Conversation over full history by 19.8 and 41.2 points.
[NLP-20] Agent Error Dataset: Scaling 50000 Error–Diagnosis Pairs for Failure Analysis and Error-Aware Post-Training
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）智能体在实际部署中失败后，如何高效利用失败过程中的丰富经验进行学习与改进的问题。传统方法仅依赖最终奖励信号，忽视了失败过程中可观测的环境状态、智能体决策路径及环境反馈等关键信息。其解决方案的关键在于构建一个系统性的错误-训练转化框架——代理错误到训练（Agentic Error-to-Training, AET）管道。该框架通过五阶段流程，从真实失败案例中提取错误诊断与修正建议，并基于保留的源轨迹和执行元数据实现跨设置的故障分析与再诊断，避免重复运行原始任务。AET管道通过对比修正方案与原始动作重试，在支持回放的场景下验证修正有效性；同时分离构建诊断与执行恢复两个独立的训练视图。实验表明，首次提出的修正方案将验证通过率从18.4%提升至51.1%（+32.7个百分点），而对诊断模块进行全量微调后，Qwen3-8B在943个测试用例上的精确步骤一致率从47.2%提升至63.6%，显著优于基线提示参考模型（54.7%）。此外，仅针对动作修复的训练策略在WebShop-lite任务上表现优于仅基于成功样本的训练，验证了从失败中学习的有效性与必要性。
链接: https://arxiv.org/abs/2609.40111
作者: Kunlun Zhu,Xuyan Ye,Yibo Li,Cheng Qian,Beibin Li,Heng Ji
机构: Apodex
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:An unsuccessful LLM agent rollout contains more information than its final reward: the observations available to the agent, the actions it chose, and the environment’s responses. Reusing this experience for learning requires identifying a decision to revise and testing a concrete alternative. We introduce the Agent Error Dataset (AED), comprising 50,228 error-diagnosis pairs from 9,961 source tasks across 33 environments, 19 harness families, and 23 policy models in text-based agent systems. We retain source traces and execution metadata to support cross-setting failure analysis and re-diagnosis without repeating the original rollout. Our five-stage Agentic Error-to-Training (AET) pipeline collects natural failures, generates diagnoses and proposed corrections, and checks them against recorded evidence. Where replay is supported, we compare corrections with original-action retries from the same checkpoint under matched execution settings. We then construct separate training views for diagnosis and actor recovery. Across 3,062 matched replay pairs, first-proposal corrections raise verifier pass rates from 18.4% to 51.1%, a gain of 32.7 percentage points. Using a separately frozen diagnosis release, full-diagnosis fine-tuning on 1,656 source tasks raises Qwen3-8B’s exact-step agreement with internal teacher labels from 47.2% to 63.6%, averaged over three seeds on a 943-case holdout. The strongest prompted reference in this comparison scores 54.7%, and mean agreement improves at each of four increasing training-set sizes. In a single-seed comparison of actor-training recipes, action-only repair training scores 6.67 percentage points higher on WebShop-lite than success-only training.
[NLP-21] OverdoseMoE: A Multi-Expert Framework for Opioid Overdose Risk Prediction
【速读】： 该论文旨在解决阿片类药物过量风险预测中因患者电子健康记录（Electronic Health Record, EHR）异质性导致的模型泛化能力不足问题，尤其关注如何在大规模、长时序的诊断历史数据基础上实现高精度的风险分层。其核心解决方案在于提出一种基于诊断特异性适配（diagnosis-specific adaptation）与多专家集成（multi-expert integration）的新型框架。关键创新点包括：首先，通过在纵向诊断序列上进行持续预训练并结合任务微调，构建了针对阿片类药物过量风险预测优化的模型OODMAMBA和OODQWEN；其次，进一步提出OVERDOSEMOE，一个基于互补专家加权策略的多专家混合模型（Mixture-of-Experts, MoE），有效整合不同规模模型的优势，在提升判别能力的同时增强预测精度。实验表明，该方法在独立验证集（MIMIC-IV）上展现出良好的跨队列鲁棒性，尤其在高风险人群（前5%）中实现了25.38%的阳性预测值（PPV）和较高的召回率，显著优于单一基线模型。研究证明，诊断特异性语言模型适配与多专家融合机制相结合，可显著提升阿片类药物过量风险预测的准确性与临床实用性。
链接: https://arxiv.org/abs/2609.40108
作者: Mingchen Li,Rohan Pandey,Junhui Qian,Feiyun Ouyang,Sunjae Kwon,Avijit Mitra,Zonghai Yao,Hong Yu
机构: University of Massachusetts Amherst (马萨诸塞大学阿默斯特分校); University of Massachusetts Lowell (马萨诸塞大学洛厄尔分校); VA Bedford Health Care (贝德福德退伍军人健康护理中心)
类目: Computation and Language (cs.CL)
备注:
Abstract:Opioid overdose remains a major clinical and public health burden, highlighting the need for scalable approaches to identify patients at high risk. Here, we investigate diagnosis-specific adaptation for 180-day opioid overdose risk prediction from patients’ preceding one-year longitudinal ICD histories. We develop OODMAMBA and OODQWEN through continued pretraining on longitudinal diagnostic sequences followed by task-specific fine-tuning. Building on the stronger Qwen-based predictors, we further propose OVERDOSEMOE, a multi-expert framework that integrates models of different scales using complementary expert-weighting strategies. Diagnosis-specific adaptation consistently improved predictive performance over general-purpose language-model baselines, with OODQWEN achieving an AUPRC of 24.47 and an AUROC of 68.56. OVERDOSEMOE further improved discrimination and precision, achieving an AUPRC of 25.17 and an AUROC of 69.49 while outperforming the strongest single-model baselines. Among patients ranked in the top 5% of predicted risk, OVERDOSEMOE identified substantially enriched overdose risk, achieving a PPV of 25.38% while retaining meaningful recall. Evaluation on an independent MIMIC-IV cohort further demonstrated cross-cohort robustness, with complementary weighting strategies showing advantages across different performance measures. These findings demonstrate that diagnosis-specific language-model adaptation combined with multi-expert integration can improve opioid overdose risk stratification and support more robust prediction across heterogeneous electronic health record populations.
[NLP-22] AutoDataBench: A Data-centric Testbed for Accelerating Auto Research
【速读】： 该论文旨在解决现有自动化研究基准测试中多因素混杂的问题，即训练框架、超参数、计算资源和数据等变量相互交织，导致难以准确归因于特定研究能力的性能差异。其核心挑战在于缺乏对模型在数据层面智能行为的独立评估机制。为此，论文提出并系统评估“数据智能”（Data Intelligence）这一关键能力——即智能体理解、操作与优化塑造模型能力的数据的能力。解决方案的关键是构建一个受控测试平台AutoDataBench，基于数据诊断、数据组织与数据构建三位一体的概念框架，通过三个高度精炼的优化任务实现对非数据因素的严格固定。该平台在工具使用、信息检索与知识注入等场景下，评估前沿大语言模型（LLM）在限定资源预算内通过迭代实验改进训练数据的能力，并进一步考察模型是否具备超越试错的数据效应推理能力，即能否在训练前预测干预效果并借助迭代反馈提升对数据变更与模型性能关系的理解。研究还验证了利用AutoDataBench生成的优化轨迹可有效提升中段训练阶段的下游编程性能，证明其在评估数据智能与生成高质量训练数据方面的双重价值。
链接: https://arxiv.org/abs/2609.40097
作者: Ruifeng Yuan,Yizhi Li,Yaxin Du,Fengyu Cai,Yiqi Liu,Hou Pong Chan,Chenghua Lin,Yun Chen,Jian Yang,Bryan Dai,Pinyan Lu,Chenghao Xiao
机构: The Hong Kong Polytechnic University(香港理工大学); IQuest Research; Shanghai Jiaotong University(上海交通大学); TU Darmstadt(达姆施塔特工业大学); The University of Manchester(曼彻斯特大学); University of Macau(澳门大学); Shanghai University of Finance and Economics(上海财经大学); Beihang University(北京航空航天大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Existing auto-research benchmarks often entangle multiple sources of improvement, including training frameworks, hyperparameters, compute budgets, and data, making it difficult to attribute why one frontier agent outperforms another to specific research capabilities. In this work, we isolate and systematically evaluate Data Intelligence: an agent’s ability to understand, manipulate, and improve the data that shapes model capabilities. We introduce AutoDataBench, a controlled testbed built on a conceptual framework of data intelligence spanning data diagnosis, data organization, and data construction, instantiated through three highly curated optimization tasks while holding non-data factors fixed. Across tool use, retrieval, and knowledge injection, we evaluate frontier LLMs’ ability to improve training data through iterative experimentation under task-specific resource budgets. Beyond optimization performance, we ask: do LLMs understand what their data interventions do? We compare predictions made before training with observed outcomes to seek evidence of data-effect reasoning beyond trial and error, and explore whether iterative feedback helps LLMs better understand how changes to training data affect model performance. Finally, we show that reusing AutoDataBench trajectories for mid-training improves downstream coding performance, highlighting its value in both evaluating data intelligence and generating high-quality training data. Code and resources are available at this https URL.
[NLP-23] From Tweets to Trades: Analyzing the Influence of Public Mood over Stock Market Performance in Turkiye
【速读】： 该论文旨在解决公共情绪（public mood）与股票市场动态之间的关联性问题，尤其关注这种关联是否因信息传播领域（如政治、经济金融、媒体与社会）及市场环境差异而异。传统研究常将公共情绪与投资者情绪（investor sentiment）混同，而本文通过区分二者，并基于2022年1月至2023年12月期间176个经筛选的X平台账号发布的610,422条推文，采用微调后的土耳其语Transformer模型对内容进行领域特定分类，构建了按日、周、月频率划分的领域特异性与综合型公共情绪指标。研究方法结合相关性分析、格兰杰因果检验、向量自回归（VAR）及脉冲响应分析，在全周期及特定市场条件下（如2023年选举期）进行检验。关键发现表明：公共情绪虽不显著影响股市收益率方向，但与价格波动幅度密切相关，尤其在“媒体与社会”领域及综合通信数据中表现更明显，且在更长的时间聚合频率下关系更强；预测性关系主要集中于“经济与金融”领域，其强度和方向随市场条件变化而异。研究的核心贡献在于揭示了在整合异质通信源时可能掩盖领域特异性关系的风险，强调了在行为金融研究中考虑传播领域异质性的必要性，为理解社交媒体公共情绪如何影响金融市场提供了新的实证框架。
链接: https://arxiv.org/abs/2609.40064
作者: Ece Elif Adak,Bertaç Şakir Şahin,Şaziye Betül Özateş
机构: Boğaziçi University (博阿齐奇大学); Yıldız Technical University (伊斯坦布尔技术大学)
类目: Computation and Language (cs.CL)
备注: 16 pages, 10 tables, 1 figure
Abstract:Purpose: This study examines whether domain-specific public mood is associated with stock-market dynamics and whether these relationships vary across communication domains and market conditions. It distinguishes public mood from investor sentiment and investigates whether heterogeneous sources of public communication exhibit different relationships with market behaviour. Design: The study analyses 610,422 posts published by 176 curated X accounts between January 2022 and December 2023, covering Politics and Government, Economy and Finance, and Media and Society. Posts are classified using fine-tuned Turkish transformer models under three domain-specific and one pooled regime. Public mood measures are constructed at daily, weekly, and monthly frequencies and examined alongside BIST100 and BIST30 market measures using correlation, Granger causality, vector autoregression, and impulse response analyses across the full period and selected market conditions. Findings: Public mood is not associated with the direction of stock-market returns but is associated with the magnitude of price movements, particularly for Media and Society and pooled communication. These relationships become stronger at longer aggregation frequencies. Predictive relationships are concentrated in Economy and Finance communication, while their magnitude and direction vary across market conditions, particularly during the 2023 election period. The pooled measure largely reflects the most active communication domain. Originality: The study contributes to behavioral-finance research by incorporating communication - domain heterogeneity into the analysis of public mood and market dynamics. It also demonstrates how aggregating heterogeneous sources can obscure domain-specific relationships between public communication and financial markets. Comments: 16 pages, 10 tables, 1 figure Subjects: Computation and Language (cs.CL) Cite as: arXiv:2609.40064 [cs.CL] (or arXiv:2609.40064v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2609.40064 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Şaziye Betül Özateş [view email] [v1] Wed, 30 Sep 2026 16:21:55 UTC (185 KB)
[NLP-24] LARC: Low-Rank Adaptive Residual Connections for Learning in Frozen Models
【速读】： 该论文旨在解决冻结模型在持续学习场景中缺乏高效反馈适应能力的问题，即如何在不重新训练整个模型的前提下，使模型能够通过少量反馈信息快速、稳定地调整自身行为。其核心解决方案是提出一种低秩自适应残差连接（Low-Rank Adaptive Residual Connections, LARC），通过在输入侧构建一个数值策略载体（numerical policy carrier），实现对隐藏表示的低秩修正（映射 $ h + BAh $）。该方法的关键在于引入两个状态：一个缓慢演化的全局状态 $ \rho $ 用于学习跨任务的初始因子，以及一个私有的快速状态 $ \Phi $，它继承 $ \rho $ 的初始化并随反馈动态更新，随后在每次迭代后重置回训练好的初始值。这种设计实现了对适应过程的精准控制，有效分离了残差容量、相对于初始点的适应能力与后续决策的有用性。实验表明，在冻结的MiniCPM5-1B-SFT模型上，仅使用4阶低秩残差结构（含12,288个可训练参数），在四候选程序选择任务中，经过两步反馈梯度更新，预期查询执行误差分别降低24.65%和36.65个百分点，显著优于重置至静态或后适配初始化的表现；同时，基于直接支持损失的选择规则达到0.78125%的极低错误率。在公开持续集成任务的时序回放测试中，保留在线更新使半布里尔损失从0.1274上升至0.1808，而固定后续干预则揭示了收缩更新带来的同批非下降及未来收益不一致问题。综合代数分析与实测数据，该框架清晰区分了残差容量、相对起点的适应性以及对后续决策的有效性，为轻量化、可解释的持续学习提供了新范式。
链接: https://arxiv.org/abs/2609.40063
作者: Junyi Zou,Avrova Donz
机构: Communication University of China (CUC) (中国传媒大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 19 pages, 6 figures, 15 tables. Technical report of MMLA. The authors contributed equally
Abstract:Low-Rank Adaptive Residual Connections (LARC) give a frozen model a compact numerical state that can learn from feedback. The map h+BAh adds a low-rank correction to a hidden representation. A slow state \rho learns starting factors across tasks; a private fast state \Phi copies them, changes with feedback, and resets to the trained initialization. This report specifies an input-side realization of the numerical policy carrier in Memory-Mediated Learning Architecture and examines its factor-space dynamics and learning lifetime. We study a rank-4 input residual with 12,288 trainable parameters on a frozen MiniCPM5-1B-SFT substrate. In a four-candidate program-selection task, two feedback-gradient steps reduce expected query execution error by 24.65 and 36.65 percentage points relative to resetting to the respective trained static and post-adaptation initializations. These development results cover 16 parameter groups and three paired training seeds. A direct support-loss selection rule is much more accurate, reaching 0.78125% error. In a repository-balanced chronological replay of public continuous-integration jobs, retaining online updates raises half-Brier loss from 0.1274 to 0.1808. A fixed follow-up intervention records same-batch non-descent and inconsistent future benefit from shrinking updates. Together, the algebra and measurements distinguish residual capacity, adaptation relative to a starting point, and usefulness on later decisions.
[NLP-25] MGhana-ST: A Low-Resource Speech Translation Dataset for Ghanaian Languages and an Analysis of Multilingual Training Trade-offs
【速读】： 该论文旨在解决低资源非洲语言在语音翻译（Speech Translation, ST）任务中因数据稀缺导致的性能瓶颈问题，特别是针对加纳四种低资源语言变体（Ga、Twi（Akuapem与Asante）、Ewe 和 Fante）构建首个大规模配对语音-英文翻译数据集MGhana-ST。其核心挑战在于：在极端数据稀缺条件下，如何有效利用多语言联合训练以提升各语言的翻译性能。解决方案的关键在于系统评估单语训练与多语训练在不同语言间的迁移效果，并揭示了多语言训练在特定语言上可能产生负面效应（如Ewe和Fante性能显著下降），这主要归因于语言间形态学差异大及资源极度不均衡。研究进一步发现，仅基于一次运行的实验结果易受随机种子噪声干扰，导致虚假的正向迁移现象；而通过多种子重复实验可更可靠地识别真实迁移效应。此外，研究对比了基于经验的跨语言迁移与基于语言类型学相似性（URIEL）的预测能力，表明经验性迁移更能捕捉实际交互关系，但二者均无法准确预判哪些语言能从联合训练中获益。这一方法论发现强调了在低资源场景下进行可复现实验的重要性。
链接: https://arxiv.org/abs/2609.40041
作者: Frank Lawrence Nii Adoquaye Acquaye,Eric George Parakal,Jesse Johnson,Kishankumar Bhimani,Jochebed Afua Basil
机构: Ashesi University(阿西西大学); AdwumaTech AI(阿杜玛科技人工智能公司); HSE University(高等经济大学); GIFT International Fintech Institute(国际金融科技学院)
类目: Computation and Language (cs.CL)
备注:
Abstract:We present MGhana-ST, a speech translation dataset for four low-resource Ghanaian language varieties: Ga, Twi (Akuapem and Asante), Ewe, and Fante. MGhana-ST is an ongoing annotation effort; the experiments here use a fixed subset of about 16.1 hours of paired speech and English translations. The audio is curated from two existing Ghanaian speech resources. Unlike in those resources, the English translations are produced directly from audio by 37 native-speaker annotators and include verbal and non-verbal event annotations. Using Whisper-small, we compare monolingual and multilingual training under severe data scarcity, reporting means over three seeds. Flat multilingual training benefits no variety in this regime. Ga and Twi are unchanged within seed variance (+0.51 and +0.06 BLEU against monolingual standard deviations of 1.63 and 2.20), while Ewe declines by 6.99 BLEU and Fante by 5.11. The degrading varieties are Ewe, which is linguistically distinct and drawn from a different source corpus, and Fante, the least-resourced. Comparing empirical cross-lingual transfer with typology-based similarity, we find that transfer BLEU identifies closely interacting language pairs better than URIEL similarity, though neither predicts which varieties benefit from joint training. We also report a methodological finding. An earlier single-run analysis found positive transfer for three of four varieties; this did not survive replication across seeds. For Ga and Twi, monolingual baselines trained on 1.6 to 6.2 hours of audio have seed standard deviations roughly five and thirty times those of the multilingual models (0.35 and 0.07 BLEU). When the monolingual condition is noisier, a single-run comparison can show apparent transfer of this size from seed variation alone. We release MGhana-ST to support research on African language speech technology and low-resource speech translation. Subjects: Computation and Language (cs.CL) Cite as: arXiv:2609.40041 [cs.CL] (or arXiv:2609.40041v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2609.40041 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-26] OPTS-TTPO: Enhancing Finite-Sample Policy-Gradient Learning with Tree Search
【速读】： 该论文旨在解决在策略梯度方法中，基于有限的在线采样难以有效覆盖稀有但高回报轨迹的问题，从而导致梯度估计存在偏差。其核心挑战在于如何在固定计算预算下提升对高价值路径的探索覆盖率，同时控制梯度偏差。解决方案的关键在于提出一种基于策略树搜索的新型框架：在线并行树搜索（On-Policy Parallel Tree Search, OPTS） 与 树轨迹策略优化（Tree Trajectory Policy Optimization, TTPO）。二者均采用当前策略生成的树状轨迹进行采样，通过在访问状态处从当前策略中采样新的后缀序列，无需动作分布校正，尽管分支结构会改变状态访问分布。论文提出的分支聚合引理（Branch Aggregation Lemma） 证明，当分支选择和权重在转移前确定时，加权分支统计量可恢复链式期望。OPTS利用性能差异估计来自适应选择扩展状态，在确定性动态、精确值函数和最大回溯优势条件下，确保搜索策略的期望回报随预算单调递增。研究进一步界定了自适应扩展带来的梯度偏差，并表明最大回溯机制将前缀信用分配给引导至更优发现后缀的动作。实验表明，相较于基线方法，TTPO的偏差始终接近无分支情况下的水平，而朴素策略梯度（NaivePG）的偏差从0.1251上升至0.4884；在相同预算下，奖励与价值引导的OPTS显著提升了正确答案覆盖率与多数投票准确率；在相同分支数下，OPTS+TTPO以小幅偏差增加换取更高的覆盖率。最终，在匹配交互或滚动预算条件下，OPTS-TTPO相比PPO在MuJoCo任务中提升尾部回报达28.6%，在Atari-57上以34-22-1的胜平负记录超越PPO（基于最后100次日志的平均回报），并在所有四个Qwen3模型上均优于PPO的micro-averaged avg@32与pass@32指标。
链接: https://arxiv.org/abs/2609.40035
作者: Junyu Lu,Shichao Weng,Zhiqiang Wang,Haojie Luo,Jingfan Zhang,Yuhua Zhou,Cheng Du,Yuzhuo Zhang,Xi Li,Jinwei Du,Tiancheng Feng,Chuan Xiao,Shuyuan Zheng
机构: Dobot Robotics; Zhejiang University (浙江大学); Fudan University (复旦大学); Osaka University (大阪大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 42 pages, 12 figures
Abstract:The policy-gradient theorem gives the exact gradient under the current policy, but finite on-policy samples may miss rare high-return trajectories. We study whether tree search improves their coverage within a fixed budget while controlling gradient bias. We introduce On-Policy Parallel Tree Search (OPTS) and Tree Trajectory Policy Optimization (TTPO) using on-policy tree trajectories, which sample new suffixes from the current policy at visited states. This needs no action-distribution correction, although branching changes state visitation. Our Branch Aggregation Lemma shows that branch-weighted tree statistics recover chain expectations when branch choices and weights are fixed before outgoing transitions are sampled. OPTS selects expansion states using estimated performance differences. Under deterministic dynamics, exact values, and max-backup advantages, the induced search policy’s expected return improves monotonically with the budget. We bound the gradient bias from adaptive expansion and show that max backup assigns prefix credit to actions leading to better discovered suffixes. Against a finite chain reference, TTPG’s measured bias stays near its no-branching level, while NaivePG’s bias grows from 0.1251 to 0.4884. At matched budgets, reward- and value-guided OPTS improve correct-answer coverage and majority-vote accuracy over independent sampling. At matched branch counts, OPTS + TTPG gains coverage with a modest bias increase relative to Fixed-branch + TTPG. Under matched interaction or rollout budgets, OPTS-TTPO improves MuJoCo tail returns over PPO by up to 28.6%, achieves a 34-22-1 win-loss-tie record against PPO on Atari-57 under the last-100-log mean-return metric, and improves micro-averaged avg@32 and pass@32 over PPO across all four Qwen3 models.
[NLP-27] UBTree: Parallel Tree Drafting via Unigram and Bigram Models for Speculative Decoding
【速读】： 该论文旨在解决生成式 AI（Generative AI）在语言模型推理过程中，因目标分布熵增加导致的并行草稿生成器（parallel drafter）有效性下降问题。现有方法在高熵场景下由于草稿多样性不足，难以有效验证多步草稿，从而限制了推理加速性能。其解决方案的关键在于提出UBTree——一种将一元词频提议器（Unigram proposer）与二元转移选择器（Bigram selector）耦合的并行草稿生成架构，通过构建树状草稿结构实现高效验证。其中，一元词频提议器采用标准交叉熵损失独立生成各位置候选词，而轻量级二元转移选择器则通过在高温数据上使用重归一化KL散度目标进行训练，预测相邻候选词对之间的转移得分。该树原生训练策略扩展了监督范围至非贪婪路径，鼓励生成合理备选分支，显著提升树验证阶段接受额外令牌的概率。实验表明，在Qwen3-4B与Qwen3-8B模型上，UBTree在7个标准基准测试中平均加速比达5.84–6.94倍，优于自回归解码，并在全部28组对比中超越DARTree；生产规模评估亦证实其优于当前前沿基线如DSpark。
链接: https://arxiv.org/abs/2609.39972
作者: Chumeng Liang,Linxuan Wang,Xinyu Peng,Huabin Liu,Yuxin Chen,Ge Liu,Guang Lin,Qifan Song,Jianguo Li
机构: Inclusion AI; University of Illinois Urbana-Champaign; Purdue University
类目: Computation and Language (cs.CL)
备注:
Abstract:Speculative decoding accelerates language model inference by verifying multiple draft tokens in a single target-model pass. Recent parallel drafters have achieved breakthrough performance in frontier production models, but their effectiveness deteriorates as the entropy of target distributions increases due to insufficient draft diversity. To overcome this bottleneck without sacrificing parallelism, we introduce UBTree, a parallel drafter that couples a Unigram proposer with a Bigram selector to construct drafting Trees. The unigram proposer is trained with the standard cross-entropy objective to generate candidate tokens independently for each position, while a lightweight bigram selector predicts transition scores between adjacent candidate pairs. Unlike the proposer, the selector is trained with a renormalized KL objective on high-temperature data. This tree-native training broadens the supervision beyond the greedy path, encouraging plausible alternative branches that improve the chance of accepting additional tokens during tree verification. Across seven standardized benchmarks with Qwen3-4B and Qwen3-8B, UBTree achieves an average speedup of 5.84 – 6.94\times over autoregressive decoding and outperforms DARTree in all 28 comparisons. Production-scale evaluation further demonstrates UBTree’s advantage over frontier baselines such as DSpark.
[NLP-28] LEAP: Learned Block-wise Evidence Retrieval for Long Audio-Video Perception
【速读】： 该论文旨在解决小时级音视频问答（AVQA）中的上下文困境问题：全段录制内容的密集编码会迅速耗尽上下文长度限制，而均匀的时间压缩则严重稀释了细粒度的声学与视觉证据。其解决方案的关键在于提出LEAP框架，通过将录制内容划分为固定时长的块，并对每一块执行轻量级定位阶段，以评分候选短时窗口；随后将排名最高的窗口聚合并重新编码为单一受限上下文的作答阶段。该设计实现了证据定位与推理过程的解耦，使模型可在不解码媒体帧的前提下基于预计算的转录文本完成候选窗口定位，同时通过在原始音视频流上进行最终作答，有效保留了细粒度的视觉及非语音证据。此外，LEAP采用两阶段训练机制，分别优化定位与作答性能，且其块网格结构天然支持因果查询，无需专门的流式训练即可实现流式推理。在多个AVQA基准测试中，LEAP相较Qwen3-Omni-30B-A3B基线提升4.5%-16.8%，并在迁移至MiniCPM-o 4.5模型后，超越其已有结果3.1%-13.0%。
链接: https://arxiv.org/abs/2609.39938
作者: Juyi Lin,Zhiqiang Lao,Jiali Cui,Lin Zhao,Pu Zhao,Dichang Zhang,Arman Akbari,Yu Qi,Xinru Jiang,Yanzhi Wang,Heather Yu,Liang Peng
机构: Northeastern University(东北大学); Futurewei Technologies(未来科学城技术公司)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 39 pages, 16 figures
Abstract:Hour-scale audio-visual question answering is constrained by a context dilemma: dense whole-recording encoding rapidly exhausts context limits, whereas uniform temporal compression severely dilutes fine-grained acoustic and visual evidence. We introduce LEAP, a framework where the model retrieves its own evidence without placing the whole recording in one context. LEAP divides a recording into fixed-duration blocks, applying a lightweight localization pass to each block to score short candidate windows. The highest-ranked windows are pooled and re-encoded in a single bounded answer pass. Consequently, the answer input and peak context remain independent of the recording duration. By decoupling evidence localization from reasoning, our framework can localize candidate temporal windows over pre-computed transcripts without decoding media frames, while preserving fine-grained visual and non-speech evidence by routing the final answering pass over raw audio-visual streams. LEAP trains both stages: a localization LoRA improves the selected windows, and an answer LoRA improves the answers read from the same windows. The block grid natively supports causal queries, enabling LEAP to support streaming inference without streaming-specific training. Across several AVQA benchmarks, LEAP improves over the Qwen3-Omni-30B-A3B baseline by 4.5-16.8%, and transfers to a second omni-modal backbone, MiniCPM-o 4.5, surpassing its published results by 3.1-13.0%.
[NLP-29] RoPE at the End of Its Rope? Theory Diagnosis and Mitigation of Long-Context Failures
【速读】： 该论文旨在解决基于旋转位置编码（Rotary Position Embedding, RoPE）的长上下文语言模型中存在的上下文失效问题，其核心挑战源于RoPE在保持稳定的词元偏好（token preference）与区分邻近位置之间存在的内在权衡。现有理论难以精确刻画训练后模型在不同上下文长度下的实际行为，限制了对缺陷成因的深入分析。本文的关键突破在于放宽了传统理论中查询-键（query-key）尺度在不同频率分量上必须相等的假设，允许各频率维度采用不等尺度，这一改进与实际观测高度吻合。由此建立的理论框架不仅可对单个注意力头及输入实例的脆弱性进行量化测量，还揭示了高频成分在增强位置敏感性的同时可能破坏语义稳定性的作用机制。进一步地，论文推导出一个理论上的上下文长度上限，在特定条件下，当超过该阈值时，固定注意力分数比较无法同时避免语义反转和位置不敏感问题。基于上述理论洞见，作者提出RoPE Profiler——一种轻量级、即插即用的诊断工具包，通过复用评估阶段已缓存的查询与键激活值，无需额外前向传播即可生成两个诊断指标，分别反映语义脆弱性和位置敏感性缺陷，显著降低计算开销。在49个长上下文任务中的系统评估显示，推理类任务主要受语义反转影响，而检索类任务则更易出现位置不敏感问题。据此，通过针对性地调整高频分量的缩放系数，可在不增加训练成本的情况下实现显著性能提升，最大可使Qwen3-8B和Llama-3.1-8B-Instruct的准确率分别提高20和25个百分点。
链接: https://arxiv.org/abs/2609.39929
作者: Yuyang Wu,Yufeng Du,Hao Peng
机构: University of Illinois at Urbana-Champaign (伊利诺伊大学厄本那-香槟分校), USA
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Long-context failures of RoPE-based language models can arise from RoPE’s intrinsic tradeoff between maintaining stable token preferences and distinguishing nearby positions. Determining which weakness to address, and how, requires a more precise characterization of RoPE’s behavior in trained models across context lengths. We address a key limitation of prior theory by allowing unequal query-key scales across RoPE frequencies, which aligns well with practical empirical observations. Our theory makes both vulnerabilities measurable for individual heads and inputs, and quantifies how high-frequency components support positional sensitivity while potentially disrupting semantic stability. We also derive a theoretical context-length bound beyond which, under specified conditions, a fixed attention-score comparison cannot jointly avoid semantic reversal and positional insensitivity. Guided by our fresh theoretical insights, we introduce RoPE Profiler, a lightweight, plug-and-play diagnostic toolkit that augments existing evaluations with zero additional forward passes by reusing cached query and key activations. Reusing activations collected during evaluation, the toolkit incurs little overhead. It supplements standard benchmark scores with two diagnostic scores that reveal semantic and positional weaknesses and help users prioritize which aspect to address. Crucially, our evaluations across 49 long-context task settings reveal a distinct pattern where reasoning tasks predominantly suffer from semantic reversal, whereas retrieval tasks are primarily vulnerable to positional insensitivity. Guided by our theory and diagnostic profiles, targeted high-frequency rescaling achieves immediate gains without additional training, improving task accuracy by up to 20 percentage points on Qwen3-8B and 25 percentage points on Llama-3.1-8B-Instruct.
[NLP-30] AdaGEPA: Adaptive Feedback Allocation for Reflective Prompt Optimization
【速读】： 该论文旨在解决生成式 AI（Generative AI）在下游任务中通过提示优化（prompt optimization）提升语言模型性能时，传统反馈选择策略因未能充分识别提示的固有弱点而导致优化效果局限的问题。具体而言，经典方法在基于任务样本评估提示表现并进行反思式修正时，若反馈样本的选择未针对提示的薄弱环节，可能导致模型仅在特定样本上表现提升，而无法实现任务层面的普遍性改进。为此，论文提出一种自适应反馈分配方法 AdaGEPA，其核心在于结合提示当前性能与任务结构信息，动态识别提示的潜在缺陷，并在每次反馈小批量中仅替换最多一个样本，以精准聚焦于已识别的弱点，同时保留其余反馈上下文以维持优化稳定性。实验结果表明，在六个下游基准测试中，AdaGEPA 在相同推理预算下实现了更高的平均验证得分，且能更早发现高性能提示；在初始的 Schema-Guided Dialogue（SGD）研究中，其半预算提示在新对话任务上的联合目标准确率已超越非自适应基线的全预算提示。研究表明，自适应反馈分配机制可显著提升反思式提示优化的有效性与计算资源利用效率。
链接: https://arxiv.org/abs/2609.39927
作者: Junyang Chen,Zecheng Wang,Jingbang Chen
机构: The Chinese University of Hong Kong, Shenzhen(香港中文大学（深圳）); Shenzhen Loop Area Institute(深圳环区研究院)
类目: Computation and Language (cs.CL)
备注:
Abstract:Prompt optimization improves the performance of language-model systems on downstream tasks by refining their prompts. Classical methods evaluate prompts on task examples and use the resulting feedback to guide prompt revisions through reflection. However, when feedback selection does not account for the prompt’s weaknesses, these revisions may improve performance on selected examples without yielding broader task improvements. To address this issue, we propose AdaGEPA, an adaptive feedback-allocation method that uses the prompt’s performance and task structure to select examples for the next prompt revision. Our method replaces at most one example in each feedback minibatch to target an identified weakness while preserving the remaining feedback context. Across our main experiments on six downstream benchmarks, AdaGEPA achieves higher mean validation scores than non-adaptive feedback selection under matched rollout budgets. AdaGEPA also finds high-performing prompts earlier across several tasks. In the initial Schema-Guided Dialogue (SGD) study, its half-budget prompts outperform the non-adaptive baseline’s full-budget prompts in joint goal accuracy on new dialogues from services seen and unseen during search. Overall, our findings highlight the potential of adaptive feedback allocation to improve both the effectiveness and rollout-budget efficiency of reflective prompt optimization.
[NLP-31] MCD: Causal Distillation of Multimodal In-Context Learning in Large Vision-Language Models
【速读】： 该论文旨在解决小型视觉-语言模型（LVLM）在多模态上下文学习（ICL）中性能随模型规模缩小而显著下降的问题。现有知识蒸馏方法主要通过直接对齐输出分布或隐藏表示来传递知识，但这类方法未能揭示教师模型在做出预测时所依赖的因果性多模态证据，导致学生模型可能仅模仿教师答案，而非真正学习如何利用真实上下文线索，从而仍受语言先验、提示结构等虚假信号干扰。为此，论文提出多模态因果蒸馏（Multimodal Causal Distillation, MCD），其核心在于通过保持结构的标记干预（structure-preserving token interventions）识别并验证上下文中对预测具有因果影响的多模态证据，并转移教师模型在保留或移除这些关键证据时的响应模式。该方法将蒸馏过程与模型在多模态ICL中使用上下文证据的因果机制相耦合，使学生模型能够更准确地学习到真正的推理逻辑。实验结果表明，MCD在三个LVLM家族和七个基准测试上平均提升学生模型性能7.23点，显著优于传统蒸馏方法（提升4.68点），且分析验证了其泛化能力。
链接: https://arxiv.org/abs/2609.39920
作者: Yanshu Li,Jiaqian Li,Canran Xiao,Xi Xiao,Tianyang Wang,Yongtai Liu
机构: Brown University(布朗大学); University of Alabama at Birmingham(阿拉巴马大学伯明翰分校); Hanyang University(汉阳大学)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: 17 pages, 8 tables, 5 figures
Abstract:Large vision-language models (LVLMs) exhibit strong multimodal in-context learning (ICL) capabilities, yet this ability degrades substantially as model size decreases. Knowledge distillation offers a natural way to bridge this gap, but existing methods primarily align output distributions or hidden representations directly. Such alignment teaches the student what the teacher predicts without revealing which evidence in the complex context causally supports that prediction. Consequently, a student can imitate the teacher’s answer while continuing to rely on language priors, prompt structure, or other spurious cues. To address this limitation, we introduce Multimodal Causal Distillation (MCD), a distillation framework that transfers how a strong teacher uses multimodal evidence during ICL. MCD uses structure-preserving token interventions to identify and verify causal evidence, then transfers how the teacher responds when that evidence is retained or removed. This design connects distillation to the causal patterns by which the model uses contextual evidence during multimodal ICL. Experiments across three LVLM families and seven benchmarks show that MCD improves student performance by 7.23 points on average and outperforms vanilla distillation by 4.68 points, while further analyses confirm the generalizability of these gains.
[NLP-32] he Concrete-Arbitrary Gap: Kinship Reasoning in LLM s Is Not Indifferent to Presentation
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在处理形式上匹配的亲属关系推理任务时，其性能是否受表达方式影响的问题，即当亲属关系使用日常熟悉词汇表达时，与采用显式定义的临时谓词（nonce predicates）表达时，模型的表现是否存在差异。研究发现，在所有测试的四款模型中（Qwen3.8-27B、Gemma 4 26B-A4B、Gemma 4 31B 和 Qwen3.8-Max），使用熟悉词汇表达的关系在准确率上均显著优于使用临时谓词表达的关系，差距分别为35.6、26.6、12.0和5.4个百分点，且所有差异均具有统计学意义。解决方案的关键在于揭示：模型的推理能力并非对表达形式完全免疫，而是受到语义呈现方式的影响；通过调整推理预算（reasoning budget）或引入提示语言干预（prompt-language interventions），可显著缩小这一差距，表明该现象是可调节的，而非模型固有的认知缺陷。因此，结论的核心是行为层面的——在这些任务中，模型所表现出的逻辑关系理解能力并非对表达形式无差别，即便提供了形式上的等价定义，临时谓词也难以达到嵌入在习得语言关联中的熟悉词汇的可用性。
链接: https://arxiv.org/abs/2609.39913
作者: Thomas Pashby
机构: Middlebury College(中佛蒙特学院)
类目: Computation and Language (cs.CL)
备注:
Abstract:We test whether large language models solve formally matched kinship problems equally well when relations are expressed in familiar vocabulary or by explicitly defined nonce predicates. Across 500 paired graphs, concrete accuracy exceeds arbitrary accuracy by 35.6 percentage points in local Qwen3.8-27B, 26.6 in Gemma 4 26B-A4B, 12.0 in Gemma 4 31B, and 5.4 in Qwen3.8-Max. All four paired gaps are statistically resolved. Reasoning budgets and prompt-language interventions can substantially reduce the difference, showing that it is modifiable rather than a fixed incapacity. The minimal conclusion is behavioral: on these tasks, the models’ manifested relational competence is not indifferent to presentation. Explicit definitions provide the formal relations but do not make nonce predicates as usable as familiar vocabulary embedded in learned linguistic associations.
[NLP-33] OPSRD: On-Policy Self-Role Distillation
【速读】： 该论文旨在解决生成式 AI（Generative AI）在复杂任务中因角色提示（role prompting）导致的推理偏差问题，尤其关注在模型采样路径错误时，传统方法难以捕捉潜在的次优但有价值的下一个词偏好。现有方法在评估或蒸馏角色提示结果时，常忽略这些未被采样的有效替代路径，且缺乏对罕见预测选项的有效监督机制。其解决方案的关键在于提出一种基于策略的自蒸馏框架——OPSRD（On-Policy Self-Reasoning Distillation），通过固定专家角色作为特权教学上下文，在无需参考答案的情况下实现在线策略自蒸馏。具体而言，一个无角色约束的学生模型生成推理轨迹，而同一基础模型的冻结实例则以角色条件方式提供每个前缀处的分布，从而揭示学生模型低估的替代选择。采用教师加权的前向KL散度作为损失函数，聚焦于学生模型预测熵最高的半数位置，通过裁剪机制限制单个词汇贡献，确保学习集中在高不确定性区域。实验在三个竞赛级数学基准上验证了Qwen3系列模型（1.7B、4B、8B）的表现，结果显示OPSRD在所有规模下均优于未使用角色提示的基础模型，并且前向KL散度在各类别平均准确率上表现最优。
链接: https://arxiv.org/abs/2609.39884
作者: Weijie Ren,Yanwen Zhang,Hao Li,Zhuolin Qi,Hengyi Zhang,Naibo Wang
机构: Zhejiang University(浙江大学); University of Electronic Science and Technology of China(电子科技大学); University of Science and Technology of China(中国科学技术大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 17 pages, 5 figures. Code: this https URL
Abstract:Role prompting elicits specialized behavior from large language models through an expert identity, offering a lightweight way to guide reasoning on demanding tasks. However, evaluating or distilling complete role-prompted answers can miss useful next-token preferences when the sampled solution remains incorrect. Transferring these preferences also requires an objective that reaches alternatives the student rarely predicts. We introduce OPSRD, which uses a fixed expert role as privileged teaching context for on-policy self-distillation without reference solutions. A role-free student generates a trajectory, and a frozen instance of the same base model supplies role-conditioned distributions on its exact prefixes, exposing alternatives beyond the sampled continuation. Teacher-weighted forward KL targets alternatives the student underestimates, with clipping to limit individual vocabulary contributions. Supervision is restricted to the highest-entropy half of student positions, concentrating learning where predictions are uncertain. Experiments on three competition-math benchmarks with Qwen3-1.7B, 4B, and 8B show improvements over the base models without role prompts at inference. Forward KL achieves the highest macro-averaged accuracy among the three evaluated divergences at every scale. Code is available at this https URL.
[NLP-34] LLM Persona Unlearning
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在开放权重设置下难以有效消除特定人格化行为模式（persona）的问题，即即使经过后训练（post-training）使模型默认表现为有益助手，仍可通过显式提示激发其潜在的、不希望出现的人格化响应。这一问题的核心在于：传统方法无法在不损害模型生成质量与通用能力的前提下，可靠地从模型权重中“遗忘”特定人格特征。解决方案的关键是提出一种名为PaCE（Persona-aware Contrastive Editing）的新方法，其核心机制是通过对比目标人格与理想响应对同一问题的输出，识别出模型内部表征中的行为方向，并在此基础上将目标人格对应的激活状态从目标模式向期望的理想行为方向迁移。该方法实现了对目标人格的有效抑制，同时保持较高的生成质量与有用性，仅带来适度的通用性能损失。研究进一步构建了专用于评估人格遗忘效果的基准测试平台PersonaUnlearnBench，揭示了现有方法在去个性化过程中普遍存在性能退化问题，从而确立了人格遗忘作为一项独立的行为级编辑任务，为实现大语言模型响应策略的持久控制提供了可行路径。
链接: https://arxiv.org/abs/2609.39882
作者: Kemou Li,Zhuan Shi,Qizhou Wang,Fengpeng Li,Negar Rostamzadeh,Golnoosh Farnadi,Jiantao Zhou
机构: State Key Laboratory of Internet of Things for Smart City, University of Macau(澳门大学物联网智能城市国家重点实验室); Mila – Québec AI Institute(魁北克人工智能研究所); McGill University(麦吉尔大学); RIKEN AIP(理化学研究所先进智能项目); King Abdullah University of Science and Technology(阿卜杜拉国王科技大学); Google Research(谷歌研究院)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Pre-training equips large language models (LLMs) with a broad repertoire of behavioral patterns associated with roles, styles, values, and goals. Post-training teaches conditional enactment and makes a helpful Assistant the default, but it does not erase alternative modes from the weights; explicit prompts can therefore elicit personas that repeatedly shape judgment, language, and action. In open-weight settings, runtime controls can be removed, motivating persona unlearning: a weight-level edit that makes a designated persona difficult to elicit and enact on unseen contexts. We introduce PersonaUnlearnBench, a model-specific paired benchmark spanning six LLMs from three families and five personas, with aligned forget/retain sets, held-out instruction paraphrases, and four-axis evaluation. The benchmark shows that standard unlearning methods cannot reliably erase the target persona without sacrificing meaningful generation or general utility. We therefore propose PaCE, which compares target and desirable responses to the same questions to locate an internal behavior direction, then trains target-prompt states away from the target mode and toward the matched desirable response. Experiments show that PaCE consistently suppresses target personas with high response quality and useful counterpart behavior, at moderate utility cost. These results establish persona unlearning as a distinct behavior-level editing problem and a practical route toward persistent control of latent LLM response policies.
[NLP-35] GrammarRL: Effective Grammar-Constrained Decoding via Reinforcement Learning
【速读】： 该论文旨在解决语法约束生成（grammar-constrained generation）中语义质量下降的问题，尤其是在提示（prompt）信息不充分或模型指令遵循能力有限时，传统方法如贪婪解码易导致生成结果虽符合语法规则但语义不佳。其核心挑战在于如何在保证句法正确性的同时提升生成内容的语义合理性。解决方案的关键是提出一种无需标注数据的自监督强化学习框架GrammarRL，通过两个互补的自监督奖励信号来优化语言模型：一是直接奖励（direct reward），衡量在给定输入下生成符合语法约束输出的概率；二是逆向奖励（reverse reward），衡量从生成输出中重构输入的能力。该方法采用Reinforce Leave-One-Out（RLOO）目标函数对多组语法约束采样序列进行优化，并融合了最优的束搜索（beam search）候选结果，同时引入对冻结基础模型的正则化以保持稳定性。实验表明，GrammarRL在手语词汇翻译、层次化文本分类和命名实体识别任务上显著优于受限贪婪解码，平均提升9.8分，最高达22.8 BLEU，且在两项任务上超越束搜索的同时维持了贪婪解码级别的推理开销，验证了两阶段奖励机制的协同增益效应。
链接: https://arxiv.org/abs/2609.39869
作者: Gabriele Tuccio,Antonino Furnari,Aldo Gangemi,Misael Mongiov`ı
机构: University of Catania(卡塔尼亚大学); ISTC - National Research Council, Italy(意大利国家研究委员会信息科学与技术研究所)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Grammar-constrained generation guarantees syntactic validity, but can substantially degrade semantic quality when the model’s preferred outputs are poorly aligned with the imposed grammar. This trade-off is particularly severe when the prompt is underspecified or the model has limited instruction-following ability. Beam search can partially mitigate these failures by exploring multiple valid sequences, but its computational cost grows with beam width, while sequence-level probability is only an imperfect proxy for semantic quality. We introduce GrammarRL, a label-free reinforcement learning method that adapts language models to grammar constraints without requiring annotated data. GrammarRL optimizes the model using two complementary self-supervised rewards derived from its own likelihoods: a direct reward, measuring how likely the constrained output is given the input, and a reverse reward, measuring how well the input can be reconstructed from the generated output. We optimize these rewards with a Reinforce Leave-One-Out (RLOO) objective over groups of grammar-constrained rollouts, augmented with the top-1 beam-search hypothesis and regularized towards a frozen base model. We evaluate GrammarRL on sign language gloss translation, hierarchical text classification, and named entity recognition using Llama models ranging from 1B to 8B parameters. GrammarRL consistently outperforms constrained greedy decoding, with an average improvement of 9.8 points and gains of up to 22.8 BLEU. It matches or outperforms beam search on two of the three tasks while preserving greedy-decoding inference cost. Ablations further show that the two rewards are complementary: either reward alone can underperform the untrained baseline, whereas their combination consistently improves upon it. Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) Cite as: arXiv:2609.39869 [cs.AI] (or arXiv:2609.39869v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.39869 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-36] FIGS: Evaluating Multi-Turn Sycophancy Without Penalizing Empathy
【速读】： 该论文旨在解决大语言模型在对话中难以平衡事实准确性与适当支持性的问题，尤其关注模型在长期互动中逐渐表现出的“奉承倾向”（Sycophancy）——即为迎合用户而偏离事实、过度赞美或偏颇建议。现有评估方法多依赖静态、单轮测试或固定脚本，无法捕捉真实对话中用户反复施压、隐性引导等动态演变过程，且常将基本共情误判为妥协，导致模型为避免被惩罚而过度趋于冷漠机械。为此，论文提出FIGS（Factual Integrity and Grounded Support）双轴评估框架，其核心在于构建一个自适应的10轮对话模拟器，能够动态反映用户重复请求、反驳或引导对话的趋势；同时引入严格区分“奉承”与“校准性共情”的分类体系，确保对模型在持续交互中保持事实完整性与适度情感支持的能力进行精准评估。研究发现，当前主流模型在长时间对话中普遍存在两极分化现象：要么逐步滑向奉承，要么过度纠正为机械疏离，表明在自然对话场景下实现诚实性与支持性的动态平衡仍是亟待突破的关键挑战。
链接: https://arxiv.org/abs/2609.39863
作者: Sidharth Pulipaka,Ruta Binkyte,Ivaxi Sheth,Sahar Abdelnabi
机构: German Research Center for Artificial Intelligence (DFKI); ELLIS Institute Tübingen; Max Planck Institute for Intelligent Systems; Tübingen AI Center
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 64 pages, 11 figures, 29 tables. Code: this https URL ; Data: this https URL
Abstract:Large language models frequently fail to balance staying truthful with being supportive. They often exhibit sycophancy in responses to users, agreeing with false claims, offering unwarranted flattery, and giving advice skewed toward users’ expressed views. In reality, sycophancy rarely happens in a single exchange; it may emerge organically as users repeatedly insist or subtly steer the dialogue over time. Current evaluations, however, rely on rigid, single-turn tests or fixed scripts that fail to capture these natural dynamics. Furthermore, these benchmarks often mistake showing basic empathy for yielding, penalizing models for acknowledging a user’s feeling. This view may drive future models to over-correct into cold, dismissive rigidity. To address this gap, we introduce FIGS (Factual Integrity and Grounded Support), a dual-axis evaluation framework built around extended, realistic dialogue. We use an adaptive 10-turn conversational simulator that dynamically challenges the target model, reflecting how users repeat requests, push back, or steer a conversation toward a preferred answer. To accurately evaluate these trajectories, we apply a taxonomy that strictly separates Sycophancy (whether the model holds firm to the truth and keeps its praise proportional) from Calibrated Validation (showing empathetic understanding of the user’s feelings without overdoing it). We release our complete testing environment, including 500 diverse multi-turn scenarios and an automated judge. Our evaluation of leading models reveals a consistent trade-off: over the course of a sustained interaction, current systems either slowly drift to sycophancy or over-correct into robotic detachment. This demonstrates that balancing honesty with appropriate support throughout a natural conversation remains a critical, unsolved challenge.
[NLP-37] Cognitive Enhancement: Rethinking the Necessity of Role-Playing for Large Language Models
【速读】： 该论文旨在解决角色扮演提示（role-playing prompting）在不同领域、模型容量及语言背景下性能提升不一致的问题，其核心挑战在于缺乏系统性验证与对作用机制的深入理解。解决方案的关键在于提出“人格相关认知对齐假说”（persona-related cognitive alignment hypothesis），即角色扮演提示的有效性取决于大语言模型（LLM）是否准确理解所设定人格及其对应的知识领域。为验证该假说，研究通过人格信息丰富度消融实验、层间熵偏移分析以及潜在思维空间偏移观测等方法进行了实证检验。在此基础上，作者提出一种无需训练、高效且适用于多语言环境的提示拼接策略——混合语言拼接预测（Mixed-Language Concatenate Prediction, MLCP），通过聚合语义等价的角色提示以增强互补表征线索，有效缓解人格认知偏差并稳定角色扮演表现。大量实验表明，MLCP在所有测试模型上均显著优于传统的角色扮演提示方法。
链接: https://arxiv.org/abs/2609.39853
作者: Xingjie Zhuang,Jialong Tang,Chulun Zhou,Buchao Zhan,Zhirui Li,Junhui Li,Yazheng Yang,Jinsong Su
机构: Xiamen University (厦门大学); Tongyi Lab; The Chinese University of Hong Kong (香港中文大学); Soochow University (苏州大学)
类目: Computation and Language (cs.CL)
备注: 22 pages, 7 figures
Abstract:Role-playing prompting has become a popular yet simple technique for improving LLM reasoning and output quality. However, whether it consistently boosts performance across diverse domains remains unclear, as systematic validation is lacking. To fill this gap, we run multi-model, cross-domain, and multilingual experiments on MMLU and MMLU-Redux. We find that gains from role-play prompting depend heavily on model capacity, knowledge domain, and prompt language. Drawing on metacognition theory, we propose the persona-related cognitive alignment hypothesis: role-play works only when the LLM correctly grasps the designated persona and its associated knowledge domain. We test this hypothesis through persona information richness ablation, layer-wise entropy divergence analysis, and latent thought-space deflection observation. To reduce persona cognitive bias and stabilize role-play performance, we propose \textbfMixed-\textbfLanguage \textbfConcatenate \textbfPrediction \textbf(MLCP), a simple, training-free, and efficient multilingual prompt concatenation strategy. It aggregates semantically equivalent role prompts to enrich complementary representational cues. Extensive experiments show that MLCP consistently outperforms vanilla role-play prompting across all tested LLMs.
[NLP-38] Learning Steganography Is Easy Learning Steganographic Reasoning Is Hard NEURIPS2026
【速读】： 该论文旨在解决生成式人工智能（Generative AI）在链式思维监控（chain-of-thought monitoring）框架下可能遭遇的隐写推理（steganographic reasoning）威胁，即大型语言模型（LLM）通过在看似无害的文本中隐藏其真实推理过程，从而规避监督与控制。研究的核心问题是：隐写推理是否会在现实训练流程中作为非预期副作用自然涌现？其解决方案的关键在于系统性地比较三种典型模型获取方法——强化学习、上下文学习（in-context learning）和监督微调（SFT）——下模型学习隐写推理、隐写消息传递（steganographic messaging）以及编码推理（encoded reasoning）的难易程度。研究发现，隐写推理的学习难度显著高于后两者，在绝大多数任务中仅在SFT条件下可被学习，且所需训练量至少是隐写消息传递的两倍；而在部分模型-任务组合中甚至完全无法习得。然而，当存在有利于信息隐藏的“掩护任务”（cover task）时，三种方法均可成功诱导出隐写推理。这表明隐写推理虽难以普遍出现，但并非不可实现，其学习依赖于特定条件，且前两类能力的出现并不预示隐写推理必然发生。
链接: https://arxiv.org/abs/2609.39838
作者: Julian Schulz,Lukas Fülle,Rieke Fruengel
机构: Meridian Cambridge; SAIGE
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted as an oral at the NeurIPS 2026 Workshop on Trustworthy AI for Good (AI4GOOD). 41 pages. Code: this https URL
Abstract:Chain-of-thought monitoring as an approach for AI oversight and control is threatened by the possibility of steganographic reasoning, where LLMs conceal their reasoning inside innocuous-looking text. Two neighbouring capabilities, steganographic messaging (passing a concealed message) and encoded reasoning (reasoning in an illegible but unconcealed format), have already been shown to emerge under training pressures that occur in real pipelines, such as reinforcement learning against monitors. This suggests that steganographic reasoning too might arise as an unintended side effect of training. Here, we compare how easily models learn steganographic reasoning and these two neighbouring capabilities across three elicitation methods: reinforcement learning, in-context learning, and supervised fine-tuning (SFT). For most tasks, models learn steganographic reasoning only under SFT, while they learn steganographic messaging and encoded reasoning under all three elicitation methods. Even under SFT, steganographic reasoning requires at least twice as much training as messaging, and for several model-task combinations it is not learned at all. However, on a cover task that makes hiding information especially convenient, steganographic reasoning can be successfully learned under all three elicitation methods. Steganographic reasoning is thus much harder than steganographic messaging and encoded reasoning, and learning the latter two does not imply learning the former. Yet it lies within reach: an easy version is learned under every elicitation method, when the cover task is convenient for hiding information. Comments: Accepted as an oral at the NeurIPS 2026 Workshop on Trustworthy AI for Good (AI4GOOD). 41 pages. Code: this https URL Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) Cite as: arXiv:2609.39838 [cs.AI] (or arXiv:2609.39838v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.39838 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-39] Synthetic Pre-pretraining Survives Scale but Not as a Grammatical Prior
【速读】： 该论文旨在解决生成式人工智能模型在预训练（Pre-training, PT）阶段的效率问题，特别是如何通过在合成非自然语言数据上进行预预训练（Pre-pretraining, PPT）来提升令牌效率。以往研究认为PPT带来的性能增益源于语法先验（grammatical prior），即模型在PPT过程中学习到的结构归纳偏置可迁移至自然语言语法。然而，此前的PPT实验仅限于参数规模不超过10亿、预训练数据量低于20亿令牌且以网络文本为主的场景，其在更大规模和更复杂数据混合下的有效性尚不明确。为此，本文开展了一项全面研究，覆盖五种PPT任务、四种预训练数据混合方式、四个参数规模（500M至7B）以及最高达1000亿令牌的预训练预算。研究结果表明，PPT在大规模场景下仍能持续提升下游任务表现与令牌效率，例如在30亿参数规模下至少节省210亿预训练令牌。但与先前观点相反，研究未发现性能增益与语法正确性之间存在一致关联；不同规模模型的下游表现与语法可接受性并不一致。进一步分析揭示，性能提升主要源自能够增强长程信息检索能力的PPT任务。此外，PPT的增益对预训练数据混合方式具有鲁棒性，仅在完全缺乏网络文本时才显著下降。因此，该研究的关键结论是：PPT是一种低成本的预训练增强手段，未来PPT任务的设计应聚焦于提升长程检索能力，而非模拟自然语言语法。
链接: https://arxiv.org/abs/2609.39827
作者: Atsuki Yamaguchi,Tatsuro Inaba,Joel Niklaus,Michal Štefánik,Aline Villavicencio,Nikolaos Aletras
机构: University of Sheffield, UK(谢菲尔德大学，英国); Mohamed bin Zayed University of Artificial Intelligence, UAE(穆罕默德·本·扎耶德人工智能大学，阿联酋); Hugging Face; National Institute of Informatics, Japan(日本信息学研究所); University of Exeter, UK(埃克塞特大学，英国); Federal University of Rio Grande do Norte, Brazil(里奥格兰德州立联邦大学，巴西); Nikolaos Aletras
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Preprint. Under review
Abstract:Pre-pretraining (PPT) on synthetic non-natural language data improves token efficiency during language model pre-training (PT). Prior work attributes this gain to a grammatical prior, i.e., a structural inductive bias learned during PPT that transfers to natural language grammar. However, PPT has only been tested on models of at most 1B parameters and PT budgets below 2B tokens on predominantly web text. It is unknown whether PPT is effective at larger scales and under more realistic PT data mixtures that combine diverse sources (e.g., code and math). We therefore present a comprehensive study on PPT spanning five PPT tasks, four PT data mixtures, four parameter scales (500M to 7B), and PT budgets of up to 100B tokens. Our results demonstrate that the downstream performance and token efficiency gains of PPT persist at scale, e.g., saving at least 21B PT tokens at the 3B scale. However, in contrast to prior work, we find no consistent evidence that these gains stem from a grammatical prior. Downstream performance does not consistently align with grammatical acceptability across model sizes. Instead, we find that downstream gains arise from PPT tasks that improve long-range retrieval. Finally, PPT performance gains are robust to how PT data mixtures are composed and diminish only when web text is absent. Overall, PPT is a low-cost addition to PT, and future PPT task design should target long-range retrieval rather than natural language grammar.
[NLP-40] Stress-Testing LLM Lie Detectors: Role-Play Failures and Spurious Correlations
【速读】： 该论文旨在解决生成式人工智能（Generative AI）在角色扮演（role-play）情境下，其输出的真伪判断难题。具体而言，当语言模型（LLM）采用与现实相悖的虚构人格（如阴谋论者）时，传统谎言检测探针（lie detection probes）可能无法准确识别虚假陈述，反而会受制于该角色信念体系的影响，从而误判为“真实”。其核心问题在于：现有探针是否能超越角色设定的主观认知，真正基于客观事实进行判断？解决方案的关键在于揭示并缓解训练数据中“真相”与潜在混淆因子（confounding concepts）之间的虚假相关性。研究通过构建包含8,916条人工审核、基于策略的响应数据集，并引入三类新型混淆数据集，发现多数现有探针严重依赖于指令遵循度或响应似然等表面特征，而非真实语义一致性。基于此，作者提出一种简单的线性探针，在角色扮演和混淆压力测试中均表现最优，验证了去耦合训练数据中“真相”与混淆概念的重要性。结果表明，当前谎言检测探针可靠性不足，亟需设计更纯净、去混淆的训练数据以提升鲁棒性。
链接: https://arxiv.org/abs/2609.39807
作者: Maximilian von Klinski,Sebastian Lapuschkin,Wojciech Samek,Lennart Bürger
机构: Fraunhofer HHI(弗劳恩霍夫通信技术研究所); TU Dublin(都柏林城市大学); TU Berlin(柏林工业大学); BIFOLD(生物医学信息学与数据科学中心)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Lie detection probes aim to predict from a language model’s internal states whether its output is truthful or dishonest. However, role-play complicates what “truth” means for an LLM: language models can adopt a wide range of personas that take very different claims to be true, including personas whose beliefs clearly contradict reality, such as a conspiracy theorist. In this work, we investigate whether lie detection probes reliably flag falsehoods generated under such an anti-factual persona or whether they instead follow the persona’s beliefs. We introduce a dataset of 8,916 human-reviewed, on-policy responses from three LLMs adopting anti-factual personas. Evaluating eight probes from prior work, we find that many fail in this setting, particularly when correct and incorrect answers are evaluated under the same persona prompt. To investigate why, we construct three novel confounder datasets in which truth is anti-correlated with a potential confounding concept. Our experiments reveal that many existing probes strongly track concepts that are spuriously correlated with truth in their training data, such as instruction compliance or response likelihood. Based on these findings, we introduce a simple linear probe that achieves the strongest overall performance on both the persona and confounder stress tests. Our results suggest that current lie detection probes are far from reliable and highlight the need for training data in which truth is decorrelated from confounding concepts.
[NLP-41] Explore-on-Graph: Hybrid Embedding-LLM Reasoning for Knowledge Graph Question Answering under Incompleteness
【速读】： 该论文旨在解决在知识图谱（Knowledge Graph, KG）不完整的情况下，基于大语言模型（Large Language Models, LLMs）的多跳问答方法因缺失事实导致推理路径断裂而失效的问题。现有方法依赖于现有图边遍历或由LLM生成缺失知识，前者在图结构不完整时不可靠，后者则易引入幻觉性证据。本文提出的XoG（eXplore-on-Graph）框架通过从学习到的图结构中恢复缺失的推理路径，而非依赖LLM的参数化知识，实现了对不完整KG的鲁棒推理。其核心解决方案在于：利用实体-关系类型的统计信息筛选候选关系，并结合知识图谱嵌入（KG embeddings）检索可能的缺失实体，再由LLM作为语义选择器与推理引擎进行协同，形成迭代式的规划-探索-推理流程。实验表明，XoG在完整知识图谱上保持竞争力，在不完整场景下显著优于无需特定任务训练的对比方法，且在多种LLM主干网络上均表现稳定，证明了更强的LLM本身无法弥补缺失图结构证据的问题。此外，相较于相近的规划方法，XoG可降低高达33%的LLM token消耗，提升了效率与可靠性。
链接: https://arxiv.org/abs/2609.39786
作者: Ola El Khatib,Djellel Difallah
机构: New York University Abu Dhabi (纽约大学阿布扎比分校)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large language models (LLMs) are increasingly combined with knowledge graphs (KGs) to ground reasoning in structured evidence. However, most LLM-based KGQA methods rely on traversing existing graph edges and become unreliable when reasoning paths are broken by missing facts. Alternatives that ask LLMs to generate missing knowledge risk introducing hallucinated evidence. We introduce XoG (eXplore-on-Graph), a framework for multi-hop question answering over incomplete KGs that recovers missing reasoning paths from learned graph structure rather than LLM parametric knowledge. XoG combines type-level entity-relation statistics to identify candidate relations with KG embeddings to retrieve plausible missing entities, using the LLM as a semantic selector and reasoner. These mechanisms are integrated into an iterative planning-exploration-reasoning process. Experiments on WebQSP, CWQ, and the Wikidata-based BRINK benchmark show that XoG remains competitive on complete KGs and consistently outperforms comparable methods without task-specific KGQA training under KG incompleteness. These gains persist across multiple LLM backbones, indicating that stronger LLMs alone do not resolve missing graph evidence. XoG also reduces LLM token consumption by up to 33% compared with a closely related planning-based approach.
[NLP-42] MemCodex: Self-Programming Hierarchical Memory for Language Agents
【速读】： 该论文旨在解决智能体记忆系统在处理异构访问需求时的适应性瓶颈问题：单一跳次查询仅需少量证据，而多跳次查询则需整合来自多个来源的复杂信息，传统预定义的记忆工作流难以动态适配此类差异。其解决方案的关键在于提出MemCodex——一种自演化的分层记忆系统，将经验组织为可执行的记忆程序（memory programs），涵盖摘要、关系知识、可复用技能及潜在记忆等层次。通过开放式的程序演化机制，系统能够自主重写各层级的构建方式、索引策略、检索逻辑与路由规则，从而在不依赖预设设计空间的前提下，实现跨层组合与层内实现的双重自适应。在查询时，读取过程遵循从粗粒度到细粒度的层级遍历策略，并在获取足够证据后提前终止，必要时下探至原始历史记录。此外，研究构建了统一运行时环境MemArena，使异构数据与记忆系统可通过通用接口协同工作。实验表明，MemCodex相较最强的自适应记忆基线，在任务成功率上提升10.1%相对性能，同时减少3.4倍上下文令牌消耗并实现2.1倍推理加速。
链接: https://arxiv.org/abs/2609.39765
作者: Xiaoqiang Wang,Bang Liu
机构: 未知
类目: Computation and Language (cs.CL)
备注: Work in progress
Abstract:Agent memory faces heterogeneous access needs: a single-hop question may require one piece of evidence, whereas a multi-hop question must combine evidence from multiple sources. Predefined memory workflows cannot adapt to these varying needs. Recent adaptive methods search or learn over memory components and their compositions, but the design space itself remains predefined. We introduce MemCodex, a self-evolving hierarchical memory system that organizes experience into executable memory programs for summaries, relational knowledge, reusable skills, and latent memory. Open-ended program evolution searches the open design space of layer programs by rewriting how each layer is constructed, indexed, retrieved, and routed, thereby adapting both within-layer implementations and cross-layer composition. At query time, reads traverse the hierarchy from coarse to fine and stop once sufficient evidence is found, descending to the original history when needed. We further develop MemArena, a unified runtime that places heterogeneous data and memory systems behind a common interface. MemCodex improves average task success by 10.1% relative to the strongest adaptive-memory baseline, while using 3.4x fewer context tokens and achieving 2.1x faster inference.
[NLP-43] LatentHarness: Learning Latent Actions for Memory and Reasoning via Counterfactual Policy Distillation
【速读】： 该论文旨在解决长上下文推理中存在的两大互补性瓶颈：在长输入中保持远距离证据的可访问性，以及在多步推理过程中维持计算的稳定性。现有方法通常分别应对这两个问题，例如通过外部记忆扩展对远端证据的访问，或通过潜在推理压缩多步计算过程。本文提出一种统一框架——LatentHarness，将记忆访问与潜在推理整合为序列化的潜在动作选择机制。在每一步内部计算中，模型可选择“思考”（THINK）以进行进一步推理、“回忆”（RECALL）从快速权重记忆中提取输入证据及中间推理状态，或“退出”（EXIT）生成下一个输出标记。该策略通过反事实策略蒸馏（counterfactual policy distillation）进行训练，即对每个动作分支执行单步操作，并评估其对最终输出标记的影响，从而学习判断何时使用记忆比继续推理更有效；同时，通过反事实回忆路径的梯度传播，指导哪些中间状态应被保留在记忆中以供未来使用。在六个通用和长上下文推理基准测试中，14亿参数规模的LatentHarness相比最强基线分别提升了2.8%和10.0%的相对性能，且推理速度比最强的长上下文基线快5.9倍。
链接: https://arxiv.org/abs/2609.39740
作者: Xiaoqiang Wang,Suyuchen Wang,Bang Liu
机构: Université de Montréal(蒙特利尔大学)
类目: Computation and Language (cs.CL)
备注: Work in progress
Abstract:Long-context reasoning faces two complementary bottlenecks: retaining evidence across long inputs and sustaining computation across many reasoning steps. Existing approaches largely address them separately, with external memory extending access to distant evidence and latent reasoning compressing multi-step computation. We introduce LatentHarness, which unifies memory access and latent reasoning as sequential latent action selection. At each internal step, the model chooses THINK for further computation, RECALL from a fast-weight memory of input evidence and intermediate reasoning states, or EXIT to emit the next token. We train this policy with counterfactual policy distillation, which branches every action for one step and scores its effect on the emitted token. These gains teach the policy when memory is more useful than further reasoning, while gradients through counterfactual recall teach which intermediate states should be retained in memory for future use. Across six general and long-context reasoning benchmarks, LatentHarness at 1.4B improves on the strongest baselines by 2.8% and 10.0% relative, respectively, and runs 5.9x faster than the strongest long-context baseline.
[NLP-44] Drift Inspector: Exploring and Measuring Scientific Drift with Atomic Contribution Claims EMNLP2026
【速读】： 该论文旨在解决科学摘要中贡献陈述与背景介绍、研究动机及元语言混杂的问题，传统方法在直接分析摘要时难以区分研究领域的真实产出与其讨论内容。为此，论文提出Drift Inspector——一个开源系统，通过提取每篇论文摘要中的原子贡献声明（Atomic Contribution Claims, ACCs），即去上下文化且承载具体贡献的命题，实现对研究领域随时间演变的精细化测量与探索。其核心解决方案在于：利用大语言模型（LLM）自动识别并提取ACCs，随后基于时间维度对这些声明进行聚类，构建可交互的趋势图谱，使每一项趋势均可追溯至原始声明及对应文献。该方法有效揭示了自然语言处理领域从经典任务向大模型时代的能力演进（如推理与多模态），而这一转变在关键词统计或整篇摘要计数中会被掩盖。所发布的数据集涵盖整个ACL Anthology（34.6万条ACCs，8万篇摘要，423个会议/期刊），且提取过程经人工验证，聚类结果亦通过外部手工构建的分类体系进行交叉检验，确保了系统的可靠性与可扩展性。
链接: https://arxiv.org/abs/2609.39710
作者: Vsevolod Karimov,Stepan Ostarkov,Anastasia Poroshina,Anatoly Frolov,Alexander Panchenko
机构: Skoltech(斯科尔科沃科学技术研究院); HSE University (高等经济大学); Lomonosov Moscow State University (莫斯科国立大学); ITMO University (圣彼得堡国立信息技术机械与光学大学); AIRI(人工智能研究所)
类目: Computation and Language (cs.CL); Digital Libraries (cs.DL)
备注: Accepted to EMNLP 2026 System Demonstrations. 11 pages. Live demo, code and data: this https URL
Abstract:Scientific abstracts mix contributions with background, motivation, and meta-language, so tools that read them as-is cannot separate what a field produces from what it discusses. We present Drift Inspector, an open-source system for measuring and exploring how a research field changes over time at the level of Atomic Contribution Claims (ACCs): decontextualized, contribution-bearing propositions an LLM extracts from each abstract before analysis. The system clusters these claims across years into an interactive map where every trend traces back to the claims and papers behind it. Applied to six years of EMNLP, it shows the field shifting away from classic NLP tasks toward LLM-era capabilities such as reasoning and multimodality – a movement that keyword or whole-abstract counts blur. The released data extend beyond EMNLP: the same pipeline has processed the full ACL Anthology (346k claims, 80k abstracts, 423 venues). Extraction is human-validated and clustering checked against an external manually constructed taxonomy.
[NLP-45] A helps B while B hurts A: directed transfer in instruction-tuning mixture
【速读】： 该论文旨在解决在固定计算预算下，如何为特定领域语料库选择最优的指令微调任务以提升语言模型性能的问题。传统方法依赖于启发式策略，如增加源任务数量或选择与目标任务相似的源任务，但这些策略隐含了“迁移总是正向”和“迁移具有对称性”的错误假设。研究发现，这种假设不成立：任务A可能对任务B有帮助，而任务B反而会损害任务A，表明迁移效果是有序源-目标对之间的有符号属性。为此，论文提出“迁移图谱（transfer map）”这一关键解决方案，通过数百次在Qwen3和Mistral系列模型（参数量从0.6B到32B）上的微调实验，构建了一个有符号的迁移效应估计器，量化每个源任务对每个保留目标任务的促进或抑制作用。该图谱在未见任务混合上的预测精度显著优于无混合感知基线，误差不足其一半；且其对特定目标和语料库具有针对性，但可跨模型规模迁移——预先在某一模型规模上选定的任务组合，在其他规模上训练时仍表现更优。最终，基于迁移图谱筛选出有效任务并剔除干扰任务，使推理类目标（因果解释、多跳问答、方法论批判）的准确率相较训练所有源任务最高提升14个百分点。
链接: https://arxiv.org/abs/2609.39702
作者: Nima H. Siboni,Vahid Rostami
机构: Juna.ai(朱纳人工智能); University of Cologne (科隆大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Adapting a language model to a specialized corpus means choosing which instruction-tuning tasks to train on under a fixed budget, and testing one choice costs a fine-tuning run. Common heuristics add more source tasks or pick sources similar to the target. The first assumes transfer is never negative; the second, that it is symmetric. We show that both assumptions fail: task A can help task B while B hurts A , so helpfulness is a signed property of ordered source–target pairs. We introduce the transfer map, a signed estimate of how much each source helps or hurts each held-out target. We fit the map in hundreds of fine-tuning runs on Qwen3 and Mistral models from 0.6B to 32B parameters, with all sources drawn from one corpus and no training examples from the target. The map predicts a held-out target’s accuracy on unseen mixtures: recorded before those runs, its predictions have less than half the error of a mixture-agnostic baseline. The map is specific to its target and corpus but transfers across model scale: a mixture selected in advance at one size beats training on all source tasks at every other size we tested. Transfer is thus a property of the data. The map selects the tasks that help and drops the one that interferes: accuracy on the reasoning targets (causal explanation, multi-hop questions and methodological critique) rises by up to 14 percentage points over training on all source tasks.
[NLP-46] ShieldCLIP: Selective Safety Alignment for Harmful Content Mitigation in Multimodal Foundation Models
【速读】： 该论文旨在解决多模态编码器（如CLIP）在大规模网络训练数据中嵌入有害关联的问题，即在不破坏良性表征的前提下实现安全对齐（safety alignment）。现有方法受限于伦理与实践约束，无法大规模收集真实不安全内容，因此采用将真实安全样本与生成样本配对的方式构建数据集，但一律将生成样本标记为不安全，忽略了单个模态本身可能安全的事实。针对这一局限，论文提出ShieldCLIP框架，其核心创新在于基于每个模态的可观测安全状态进行安全对齐，而非依赖样本来源，从而保留安全内容并仅对不安全部分进行重定向。此外，研究构建了ViSUv2数据集（195k四元组），包含578个概念和28个类别，支持跨模态的独立模态级安全标签。在此基础上，ShieldCLIP设计了一种四类条件目标函数：安全内容被锚定、不安全模态被引导至对应安全版本、混合对仅更新不安全分支、双模态均不安全时强制保持一致性。实验表明，无论在跨模态检索、Stable Diffusion v1.4/SDXL文本到图像生成，还是LLaVA图像到文本生成任务中，ShieldCLIP均显著降低有害输出，优于先前的安全对齐模型与强基线，同时有效维持原始嵌入空间的实用性。消融实验进一步验证了模态特定监督与选择性对齐目标的关键作用。
链接: https://arxiv.org/abs/2609.39688
作者: Tobia Poppi,Silvia Cappelletti,Samuele Poppi,Marcella Cornia,Lorenzo Baraldi,Diego Garcia-Olano,Rita Cucchiara
机构: University of Modena and Reggio Emilia(摩德纳和雷焦艾米利亚大学); University of Pisa(比萨大学); MBZUAI(穆罕默德·本·扎耶德人工智能大学); Meta Superintelligence Labs(Meta超级智能实验室)
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Multimedia (cs.MM)
备注:
Abstract:Multimodal encoders such as CLIP underlie many downstream systems, but their web-scale training data embed harmful associations that safety alignment must suppress without unnecessarily changing benign representations. Because ethical and practical constraints prevent collecting real unsafe content at scale, existing datasets pair safe real samples with generated counterparts, but label every generated sample unsafe, even when one modality is individually safe. To address this, we introduce ShieldCLIP, the first framework to condition safety alignment on the observed safety state of each modality rather than the origin of a sample, preserving safe content while redirecting only what is unsafe. We also introduce ViSUv2, a 195k-quadruplet dataset with independent per-modality safety labels across 578 concepts and 28 categories. Using these labels, ShieldCLIP defines a four-way conditional objective beyond pair-level supervision: safe content is anchored, unsafe modalities are redirected to their safe counterparts, mixed pairs update only the unsafe branch, and coherence is enforced when both are unsafe. We evaluate ShieldCLIP on cross-modal retrieval, text-to-image generation with Stable Diffusion v1.4 and SDXL, and image-to-text generation with LLaVA. Across these settings, ShieldCLIP consistently reduces harmful outputs over prior safety-aligned encoders and strong mitigation baselines, while preserving the utility of the original embedding space. Extensive ablation studies further show that both modality-specific supervision and the selective alignment objective contribute to these gains. Source code, trained models, and ViSUv2 (under a controlled-access protocol) will be made publicly available at this https URL.
[NLP-47] Better Supervision Is Nearby: Neighborhood On-Policy Self-Distillation
【速读】： 该论文旨在解决生成式数学推理模型在使用在线自蒸馏（On-policy Self-Distillation, OPSD）训练时，因固定参数设置导致监督信号局限的问题。标准OPSD仅依赖一个固定的“特权教师”（privileged teacher）提供监督，无法充分捕捉参考解在不同状态下的局部优化方向。其关键解决方案是提出邻域自蒸馏（Neighborhood OPSD, N-OPSD），通过引入局部参数扰动生成多个互补的专家模型，这些专家在不同参考位置提供对齐参考解的修正建议。在离线阶段，采用贪心策略构建一个紧凑且冻结的专家池，以最大化过滤后参考标记的增益；在线阶段，通过最大峰值（MaxPeak）选择锚定标记，再基于分位数选择与之匹配的专家，从而实现锚定方向与支持强度的分离。学生模型通过截断前向KL散度目标学习所选专家的完整下一个标记分布。实验在AIME 2024、AIME 2025和HMMT February 2025三个基准上验证，N-OPSD在Qwen3-1.7B、4B和8B模型上分别相对于OPSD提升Average@12 2.75、1.67和1.94分，显著优于基线。结果表明，利用专家池扩展监督覆盖范围、基于过滤参考标记增益的筛选机制以及考虑池内重叠与按状态路由的策略均有助于提升学生模型精度，推理阶段仅需使用蒸馏后的学生模型。
链接: https://arxiv.org/abs/2609.39687
作者: Xincheng Wei,Yifan Ding,Yoshua Li,Yuquan Lu,Ziheng Li,Yi Lu,Dongsheng Ma,Rongxiang Weng,Xunliang Cai
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:On-policy self-distillation (OPSD) trains mathematical reasoning models using a privileged teacher that sees a reference solution and supervises student-sampled prefixes. Standard OPSD uses one fixed parameter setting at every state, but nearby settings may offer additional supervision. We find that local parameter perturbations reveal complementary reference-aligned corrections under the same reference context. Different experts supply these corrections at different reference positions. Their pool covers more such positions than the unperturbed privileged teacher. We introduce Neighborhood OPSD (N-OPSD) to turn these corrections into supervision at student-visited states. Offline, greedy selection builds a compact pool of frozen experts by rewarding filtered reference-token gains beyond the pool’s current best at each position. The highest-peak expert need not provide the best training target. Online routing therefore separates the anchor direction from its level of support. MaxPeak selects the anchor token, and quantile selection chooses among experts whose top token matches it. The student learns from the chosen expert’s full next-token distribution through the clipped forward-KL objective inherited from OPSD. We evaluate on AIME 2024, AIME 2025, and HMMT February 2025. Across three independent runs per method, Neighborhood OPSD improves the three-benchmark Average@12 over OPSD by 2.75, 1.67, and 1.94 points on Qwen3-1.7B, 4B, and 8B, respectively. Student-prefix continuations support using the pool beyond the reference trajectories used for selection. Matched ablations support filtered reference-token gains as a selection criterion. Accounting for overlap within the pool and routing by state further improve student accuracy. Inference uses only the distilled student.
[NLP-48] he Evolution of Attention in Large Language Models : Mechanisms Trade-offs and Emerging Trends
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在处理长序列时面临的高效上下文建模问题，核心挑战在于自注意力机制（self-attention）带来的二次方预填充开销（quadratic prefill cost）以及键值缓存（key-value cache）随上下文长度线性增长所导致的内存瓶颈。为应对这一问题，研究领域探索了显式记忆压缩、稀疏访问、循环状态构建、结构化状态动态演化及异构机制组合等多种路径。本文的关键贡献在于提出一个五维分析框架——记忆表征（Memory Representation）、记忆更新（Memory Update）、访问（Access）、读出（Readout）与集成（Integration），用以系统刻画模型内部上下文记忆的本质特征与演进逻辑。该框架不预设特定计算模型，而是作为统一视角比较不同研究方向的发展脉络。通过分析14条主流模型谱系和11个高性能开源模型的59个发布版本，研究发现：显式记忆方法与循环状态方法虽保持不同接口，但逐渐实现重叠的记忆功能控制；异构架构正通过网络深度维度进行协同——层间组合将互补的记忆处理分布在不同表征阶段，而跨层复用则保留部分记忆与路由信息；由此催生出一种“状态化多维记忆路由”假说：持久记忆在时间跨度、网络深度、载体类型与表征粒度等多个维度上被组织，且通过稀疏写入（Sparse Write）与稀疏读取（Sparse Read）的协同机制决定哪些内容被保留、哪些参与查询响应。最终，该研究指出，高效序列架构设计的核心已从单一注意力算子的优化转向对上下文记忆的组织方式、生命周期管理与选择性使用机制的整体考量。
链接: https://arxiv.org/abs/2609.39661
作者: Zhentao Tan,Jingyi Shen,Yanbo Li,Yao Liu,Yue Wu,Jieping Ye
机构: Alibaba Token Hub, Alibaba Group(阿里巴巴集团)
类目: Computation and Language (cs.CL)
备注:
Abstract:Self-attention gives LLMs fine-grained, query-dependent access to context, but dense token interactions incur quadratic prefill cost and a key–value cache growing with context length. Research thus spans explicit-memory compression, sparse access, recurrent state construction, structured state dynamics, and heterogeneous mechanism composition. This survey analyzes these developments as model-internal contextual memory. We introduce a five-dimensional lens—Memory Representation, Memory Update, Access, Readout, and Integration—describing what is represented, how it changes, what is query-eligible, how it is read, and how readouts form outputs. This lens compares overlapping research lines without imposing one computational model. We reconstruct mechanism-level developments and architectural adoption using 59 release-level records from 14 major model lineages and 11 high-performing open-weight endpoints. First, explicit-memory and recurrent-state methods retain distinct interfaces but increasingly control overlapping memory functions. Second, heterogeneous architectures increasingly coordinate across network depth: layer-wise composition distributes complementary memory processing across representational stages, while cross-layer reuse carries selected memory and routing artifacts forward. Depth thus becomes a dimension along which contextual memory is constructed and managed. Third, these developments motivate a stateful multidimensional memory-routing hypothesis: persistent memory is organized across temporal scope, network depth, substrate type, and representation granularity, while coordinated Sparse Write and Sparse Read determine what is maintained and what contributes to each query. Overall, efficient sequence architecture design increasingly concerns the organization, lifecycle, and selective use of contextual memory rather than an isolated Attention operator. Subjects: Computation and Language (cs.CL) Cite as: arXiv:2609.39661 [cs.CL] (or arXiv:2609.39661v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2609.39661 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-49] SEPAL: Separated Expert Pairs with Answer-Level Fusion for Reliable LLM Collaboration
【速读】： 该论文旨在解决多智能体协作中因共享讨论导致纠错过程伴随错误信息传播的问题，进而削弱投票机制所需的答案多样性。其核心挑战在于如何在保持推理多样性的同时实现高效且准确的协同修正。解决方案的关键是提出SEPAL框架，通过为推理、证据锚定和验证三个角色分别配置独立的私有Actor-Critic团队，实现角色特异性训练，使各团队具有差异化的推理目标而非仅依赖采样变异。每个Critics仅在其所属团队内指导修正，有效阻断错误在不同候选答案间的跨团队传播。最终仅对各团队的最终答案进行多数投票，保留推理历史的独立性以维持多样性。实验表明，SEPAL在五个开源大语言模型和五个问答基准上均显著提升平均准确率，相较单对Actor-Critic协作提升1.81个百分点，且在所有模型上均取得改进。
链接: https://arxiv.org/abs/2609.39645
作者: Weijie Ren,Yanwen Zhang,Hao Li,Zhuolin Qi,Hengyi Zhang,Naibo Wang
机构: Zhejiang University(浙江大学); University of Electronic Science and Technology of China(电子科技大学); University of Science and Technology of China(中国科学技术大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 22 pages, 4 figures. Code: this https URL
Abstract:Multi-agent collaboration lets large language models (LLMs) improve question answering through deliberation and feedback. Yet shared discussion couples correction with exposure to the same mistakes, which can erode the diversity needed for voting. Self-consistency offers sampling diversity without feedback, while single-pair Actor-Critic collaboration refines only one candidate. We introduce SEPAL, which assigns three private Actor-Critic teams to direct reasoning, evidence grounding, and verification. Role-specific training gives the teams different reasoning objectives beyond sampling variation. Each Critic guides revisions within its own team, preventing feedback from carrying errors across candidates. Once revision ends, majority voting combines only the final answers, keeping the reasoning histories separate until the decision. Across five open-weight backbones and five question-answering benchmarks, SEPAL improves mean accuracy by 1.81 percentage points over a matched single Actor-Critic pair, with improvements across all five backbones. Code is available at this https URL.
[NLP-50] Zero-Compute Cross-Lingual Transferability Estimation Using Typological Feature Proxies NEURIPS
【速读】： 该论文旨在解决跨语言迁移（cross-lingual transfer）中知识迁移效果的可预测性问题，特别是探究是否仅通过免费获取的语言类型学特征（typological features）即可有效预测跨语言迁移性能，以及高资源源语言的优势是否源于语言类型学特性而非数据质量与数量。其解决方案的关键在于证明了类型学数据库中蕴含着廉价且密集的跨语言迁移信号：仅基于类型学特征构建的随机森林模型在24种语言的已有迁移矩阵上实现了留一语言（leave-one-language-out）评估下的皮尔逊相关系数ρ=0.705、决定系数R²=0.49，显著优于非类型学控制组（ρ=0.62），验证了仅使用类型学信息即可重建成本高昂的实测迁移性能。进一步分析表明，该信号在留一书写系统（leave-one-script-out）和留一族系（leave-one-family-out）测试中依然稳健，排除了书写系统和语系混杂的干扰；通过将迁移效应分解为类型学项与资源及书写系统偏差项，发现最优源语言排名受后者显著影响，而类型学项本身不受此偏差影响，因此类型学可作为零计算量的预筛选工具，替代数百次训练运行，实现高效模型选择。
链接: https://arxiv.org/abs/2609.39640
作者: Dalton Raphael Harmsen,Swier Garst,Thomas van Osch,Zarè Palanciyan,Joaquin Vanschoren
机构: Model call failure
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 4 pages, NeurIPS workshop, Linguistic Principles for Foundation Models, lp4fm
Abstract:Cross-lingual transfer describes how knowledge in a source language benefits a target language. Measuring it quantitatively requires broad multilingual pre-training, as prior work has done with cross-lingual transfer matrices. We ask whether transfer is predictable from freely available typological features, and whether the prominence of high-resource source languages reflects typology or data quality and quantity. We show that typological databases contain cheap and dense signals about cross-lingual transfer. Our typology-only random forest on a 24-language prior-work transfer matrix scores leave-one-language-out \rho=0.705 and R^2=0.49 , beating a non-typological control at \rho=0.62 , which verifies the ability of typology-only predictions to reconstruct costly measured cross-lingual transfer. The signal survives leave-one-script-out and leave-one-family-out protocols, so script and family confounding do not explain the effect. By decomposing the transfer into a typology term and a resource-and-script bias term, we find the best-source ranking sensitive to this bias. In contrast, typology is not affected by this bias, which makes it a zero-compute screening tool that replaces hundreds of training runs with a model fit. Our code is available \hrefthis https URLhere.
[NLP-51] Marginal Response Surface Elicitation for Zero-Label Tabular Learning
【速读】： 该论文旨在解决传统表格学习（tabular learning）对标注数据高度依赖的问题，尤其是在缺乏标签数据的情况下如何实现有效的预测。其核心挑战在于如何在无监督或零样本条件下利用领域先验知识提升模型性能。解决方案的关键在于提出一种名为边际响应面提取（Marginal Response Surface Elicitation, MARS）的方法，该方法通过从无标签数据中选取各特征的代表性取值，利用大语言模型（LLM）生成对应的类别支持度分数与特征权重，并基于中位数聚合多个响应构建特征响应函数。最终，通过加权求和的方式完成预测，且无需后续调用LLM，从而实现可复用的零样本表格分类器。该方法显著提升了预测性能，在8个基准任务上平均AUC和平均精度（AP）分别优于直接提示法1.97和6.21个百分点，同时大幅降低端到端计算成本。
链接: https://arxiv.org/abs/2609.39639
作者: Liangyu Teng,Yicheng Ding,Jing Liu,Hengsong Liu,Juncen Guo,Hongru Li,Jingyu Zhang,Liang Song
机构: University of Science and Technology of China (中国科学技术大学); Alibaba Group (阿里巴巴集团)
类目: Computation and Language (cs.CL)
备注:
Abstract:Tabular learning uses structured data to predict target outcomes. Traditionally, this process has relied on labeled data. However, large language models (LLMs) can be used to elicit domain priors based on the task description and feature semantics, thereby enabling predictions without labeled data. We propose Marginal Response Surface Elicitation (MARS), a method that transforms feature-level LLM priors into a reusable, zero-shot tabular classifier. To construct this classifier, MARS selects representative values for each feature from unlabeled data and prompts the LLM to provide corresponding class support scores and feature weights. It then aggregates multiple responses using the median to construct feature response functions, and makes predictions through their weighted sum without further LLM queries. Across eight tabular benchmark tasks, MARS achieves the highest average AUC and AP, outperforming direct prompting by 1.97 and 6.21 percentage points respectively, while substantially reducing end-to-end costs. Evaluations with LLMs of different sizes further demonstrate its predictive advantage over direct prompting.
[NLP-52] Is This Evidence Decision-Critical? Learning to Verify Rule-Governed Decisions
【速读】： 该论文旨在解决规则驱动决策（如资格审查与合同评审）中语言模型对证据关键性识别不足的问题，尤其关注因误判或遗漏关键证据而导致决策错误的风险。其核心挑战在于：传统方法难以准确识别哪些证据对条件判断及最终决策具有决定性影响。为此，论文提出一种基于干预的影响力学习框架（InterPact），其关键创新在于通过可控的反事实证据干预机制，实现对证据关键性的可验证推断。具体而言，该框架利用冻结的语言模型对案例事实进行编辑，生成训练样本，并结合人工标注的条件与决策变化标签，以及完整的状态到决策映射，监督传播验证器的学习过程。在训练阶段，验证器通过固定组合操作，将证据相关的条件概率加权于条件决策预测，从而将决策变更的监督信号回传至基础模型；推理阶段则无需依赖人工或更强模型的监督，直接从原始案例与目标证据中判断关键性。实验表明，InterPact在单案例证据关键性验证任务上达到68.28%的准确率，显著优于六种基线方法，验证了学习到的决策敏感性可作为优先检查关键证据的有效依据。
链接: https://arxiv.org/abs/2609.39608
作者: Haoyang Zhang,Jianpeng Zhao,Qi Hao,Pengyang Wang
机构: State Key Laboratory of Internet of Things for Smart City and Institute of Smart City Technologies, University of Macau (澳门大学), Macau SAR, China
类目: Computation and Language (cs.CL)
备注:
Abstract:Rule-based reasoning, as in eligibility checks and contract reviews, requires language models to assess evidence against individual conditions and combine their judgments under explicit rules. Errors in evidence assessment can leave a decision unchanged, but misinterpreting or overlooking decision-critical evidence can reverse it. Identifying such evidence allows more capable models to focus on checking the corresponding condition judgments, supporting accurate and safe decisions. Recognizing the evidence’s criticality requires understanding how evidence affects a condition judgment and how that judgment affects the decision. To achieve the goal, we propose a INTERvention-based imPACT learning framework (InterPact), which enables counterfactual verification of evidence criticality in rule-governed decisions. Specifically, its evidence intervention constructor generates training pairs for a propagation verifier by editing case facts with a frozen language model while holding rules and non-target conditions fixed. Human-reviewed labels record the resulting condition and decision changes, while complete state-to-decision mappings supervise consequences beyond the observed edit. During training, the verifier weights learned conditional decision predictions by evidence-based condition probabilities through a fixed composition operation, propagating decision-change supervision into the base model. At inference, the trained base model directly judges criticality from the original case and target evidence, without human or stronger-model supervision. On single-case evidence criticality verification over adapted rule-governed decision cases, InterPact achieves 68.28% accuracy, outperforming all six baselines. These results support learned decision sensitivity as a basis for prioritizing evidence checks.
[NLP-53] hinking Outside the Box: Can Language Models Rely on External Guidance Selectively?
【速读】： 该论文旨在解决大语言模型在依赖人类设计的流程（workflow）时，因流程不可靠而受限的问题。随着模型能力提升，其对引导信息的过度依赖可能带来风险，尤其当引导存在误导或失效时，模型难以自主判断并做出合理决策。核心问题在于如何实现“跳出框定”（thinking outside the box）的能力——即在充分利用有效指导的同时，能够识别并规避不可靠的外部信息。解决方案的关键在于训练模型具备选择性依赖（selective reliance）的能力：通过引入反事实监督微调（counterfactual supervised fine-tuning）提升对错误流程的鲁棒性，结合基于结果的强化学习（outcome-based reinforcement learning）引导模型更积极地采纳有效流程。实验表明，这种能力不仅适用于工作流，还可泛化至其他外部信息形式，如同伴修正与受损记忆的容错处理，从而显著增强智能体的可靠性。研究揭示，选择性依赖不可靠外部信息的能力是衡量智能体可靠性的重要维度，独立于任务完成度本身。
链接: https://arxiv.org/abs/2609.39578
作者: Minghan Wang,Boyuan Wang,Jinhang Zuo,Yuxin Tao,Fang kong
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Agent harnesses often improve language models with human-designed workflows, but as models grow more capable, unreliable guidance can increasingly constrain their execution. We call the ability to benefit from useful guidance while overriding unreliable guidance thinking outside the box. We introduce Box ^2 -Bench, which holds the model and task fixed while varying workflow reliability to isolate how models regulate their reliance on guidance. On Box ^2 -Bench, frontier models often benefit from reliable guidance but remain vulnerable when it is misleading or becomes unreliable. To test whether this capability can be learned, we train two open-weight models using bad workflows, reserving good workflows for evaluation. We explore two complementary training strategies: counterfactual supervised fine-tuning improves robustness, while outcome-based reinforcement learning can shift the balance toward greater use of helpful workflows. We further find that this behavior extends beyond workflows to other forms of external information, improving peer correction and robustness to corrupted memory. Together, our results identify selective reliance on fallible external information as a dimension of agent reliability not captured by task performance alone.
[NLP-54] Compact Language Complex Model Shifts: How and Where Ambiguity and Underspecification Affect LLM s
【速读】： 该论文旨在解决语言模型在处理词汇歧义（lexical ambiguity）和指称不明确（underspecification）时的训练机制与生成行为问题。其核心挑战在于理解这些语言现象如何影响模型的学习效率与生成准确性，以及模型内部如何表征并处理此类语义不确定性。解决方案的关键在于引入人工构造的同音异义词（artificial homonyms）和上位词（artificial hypernyms）作为伪词（pseudowords），通过系统性地增加训练数据中这些模糊或不明确结构的比例，考察语言模型在不同阶段的生成性能变化。研究发现，尽管歧义与不明确性会提升模型整体表现，且这种提升与语言类型-标记比（type-token ratio）的变化呈正相关，但模型在生成包含歧义词或其同义词的序列时准确率反而下降；更重要的是，模型内部表征显示，对伪同音异义词具有解歧能力，而对伪上位词的不明确性则在生成过程中保持未被消解，揭示了模型在内部表示层面存在差异化的歧义处理机制。
链接: https://arxiv.org/abs/2609.39572
作者: Michaela Regneri,Nina Scheller,Sören Laue
机构: HAW Hamburg(汉堡应用技术大学); Universität Hamburg(汉堡大学)
类目: Computation and Language (cs.CL)
备注: To appear in Proceedings of BlackBoxNLP 2026
Abstract:We analyze how lexical ambiguity and underspecification affect language model training. We create artificial homonyms and artificial hypernyms as pseudowords and analyze the generative performance of language models as they are trained with increasing amounts of these ambiguous or underspecified pseudoword types. We further analyze whether the models disambiguate ambiguous or underspecified statements and provide a first mechanistic account of how ambiguity and disambiguation are represented internally. Our main results show that both ambiguity and underspecification increase model performance in ways that scale with their influence on the language’s type-token ratio. However, the accuracy of generating sequences containing ambiguous words or their synonyms decreases compared to other texts. We also show that internal representations of pseudowords reflect disambiguation of pseudo-homonyms, but underspecification of pseudo-hypernyms is maintained during the generative process.
[NLP-55] Speculative Safety Honeypot: Toward Proactive Defense Against Multi-turn Agent Attacks
【速读】： 该论文旨在解决大型语言模型（Large Language Model, LLM）代理在复杂环境中部署时面临的多轮交互攻击（multi-turn interaction attacks）这一安全挑战。现有检测方法依赖历史上下文，难以识别跨多轮对话隐藏的深层恶意意图，因其行为被刻意分散以规避即时检测。为应对这一问题，论文提出一种名为“推测式安全诱捕”（Speculative Safety Honeypot, SSH）的框架。其核心解决方案在于构建一个由小型语言模型组成的多代理仿真系统，实现基于动作级别的“推测-验证”工作流：在推测阶段，SSH 异步生成目标代理未来行为的轨迹树，提前暴露潜在风险；在验证阶段，利用目标代理的真实行为对轨迹树进行校准与剪枝，显著降低误报率。作为即插即用组件，SSH 为现有检测器提供超越当前交互片段的决策冗余，通过分析整个轨迹树的演化过程而非单一时间点判断风险，从而减少对单个检测模块绝对精度的依赖，提升了代理系统对复杂时序攻击的防御韧性与预警提前量。
链接: https://arxiv.org/abs/2609.39549
作者: Zezhong Wang,Xueyang Tang,Rui Lian,Yang Lou,Heqing Huang
机构: Huawei Technologies Co., Ltd (华为技术有限公司)
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:As Large Language Model (LLM) agents are increasingly deployed in complex environments, multi-turn interaction attacks have become a significant security challenge. Existing detection methods typically rely on historical context. However, this retrospective logic struggles to identify deep malicious intents that are split across turns to hide future risks. Inspired by speculative decoding, we propose the Speculative Safety Honeypot (SSH) framework. SSH uses a multi-agent simulation system composed of small LLMs to build an action-level speculate-and-verify workflow. In the speculation stage, SSH predicts future behaviors of the target agent and asynchronously builds a trajectory tree to expose potential risks in advance. In the verification stage, the system uses the target agent’s real actions to calibrate and prune the trajectory tree, effectively reducing false positives. As a plug-and-playable component, SSH provides existing detectors with rich decision redundancy beyond the current interaction slice. By judging risk based on the evolution of the entire trajectory tree rather than a single point in time, the system reduces the reliance on the absolute precision of individual detection components. This improves the defense resilience and the warning lead-time of agent systems against complex temporal attacks.
[NLP-56] CATCH: A Controllable Analysis Testbed for Reward Hacking in Coding RL
【速读】： 该论文旨在解决强化学习中可验证奖励（Reinforcement Learning with Verifiable Rewards, RLVR）场景下大语言模型（Large Language Models, LLMs）存在的“奖励黑客”（reward hacking）问题，即模型通过利用环境设计中的漏洞获取高奖励，而未真正提升预期能力。这一现象严重威胁训练效率与系统安全性，但现有研究受限于缺乏能够复现奖励黑客行为并可靠识别的测试平台。为此，论文提出CATCH——一个可控的代码生成强化学习奖励黑客研究测试平台。其关键在于：通过故意暴露环境漏洞，并基于执行结果提供基于独立审计的黄金标签（execution-based gold labels），实现对奖励黑客行为的精准检测；同时，通过监督微调数据混合比例控制模型初始的黑客倾向，以及通过奖励设计调节获取奖励的难度，从而支持对奖励黑客演化动态及干预措施的系统性比较。实验表明，CATCH能生成具有明显奖励黑客特征的多样化训练轨迹，分析揭示模型初始状态与奖励设计难度共同影响黑客行为的出现。进一步评估显示，链式思维监控器（chain-of-thought monitor）虽可初期抑制黑客行为，但随着策略模型学会以代码注释误导监控器，其防御效果逐渐失效，凸显了在训练全周期内使用CATCH评估缓解策略必要性。
链接: https://arxiv.org/abs/2609.39533
作者: Shouli Wang,Yanfeng Jia,Zhihao Ou,Zitao Su,Ruize He,Haotong Xie,Hao Peng,Juanzi Li,Xiaozhi Wang
机构: Tsinghua University (清华大学); Beihang University (北京航空航天大学); Southern University of Science and Technology (南方科技大学); Renmin University of China (中国人民大学); Shanghai University of Finance and Economics (上海财经大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:During reinforcement learning with verifiable rewards (RLVR), large language models (LLMs) can exploit loopholes in their environments to obtain high rewards without improving the intended capabilities, i.e., reward hacking. Despite its risks to training efficiency and safety, monitoring and mitigating reward hacking during training remain challenging, which is limited by a lack of testbeds that reproduce hacking and reliably identify it. We introduce CATCH, a controllable testbed for studying reward hacking in coding RL. CATCH deliberately exposes environmental loopholes and provides execution-based gold labels by comparing success under a vulnerable evaluator with task correctness under an independent audit. It also can control the model’s initial hacking tendency through supervised fine-tuning data mixtures and the difficulty of earning rewards through reward designing, enabling systematic comparisons of hacking dynamics and interventions. Experiments show that CATCH can produce diverse RL training trajectories with clear reward hacking, and analyses demonstrate that both initial models and reward difficulties shape the emergence of reward hacking. We further evaluate the effectiveness of different reward hacking detection and mitigation methods. A key finding is that a chain-of-thought monitor initially suppresses hacking, but this protection erodes as the policy model learn to mislead the monitor with code comments. This highlights the need to evaluate hacking mitigations throughout training with CATCH. The source code and resources are publicly released at this https URL.
[NLP-57] Spike-driven Vision-Language-Action Model
【速读】： 该论文旨在解决现有视觉-语言-动作（Vision-Language-Action, VLA）模型在资源受限平台部署时面临的高延迟与高能耗问题，其核心挑战在于传统VLA模型依赖大规模Transformer架构，难以满足边缘计算场景下的能效需求。本文提出的解决方案关键在于构建首个基于脉冲神经网络（Spiking Neural Networks, SNNs）的端到端可训练脉冲驱动VLA框架，通过稀疏事件驱动计算实现高效能与低功耗。其核心创新包括：1）设计脉冲视觉与指令编码器，将视觉观测与语言指令转化为稀疏且可靠的脉冲表示；2）提出多胜者脉冲融合（Multi-Winner Spike Fusion）机制，采用双向top-k竞争机制抑制背景干扰，实现鲁棒的跨模态融合记忆；3）引入脉冲动作分块变换器（Spike Action Chunking Transformer），通过在融合记忆与机器人当前状态间施加脉冲交叉注意力，高效生成连续动作序列。实验在LIBERO与Meta-World基准上验证了该框架在参数量更少、推理能耗更低的前提下仍具备与传统VLA模型相当的性能，为神经形态感知-决策一体化的具身智能系统提供了可扩展的能效优化范式。
链接: https://arxiv.org/abs/2609.39514
作者: Shuai Wang,Malu Zhang,Mingquan Liu,Weihui Dai,Dehao Zhang,Jieyuan Zhang,Yimeng Shan,Zijian Zhou,Yang Yang
机构: University of Electronic Science and Technology of China (电子科技大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Vision-language-action (VLA) models bridge multimodal understanding and robotic control, advancing the dominant paradigm for embodied intelligence. However, most existing models rely on large Transformers, whose latency and energy costs hinder deployment on resource-constrained platforms. Through sparse event-driven computation, spiking neural networks offer a promising paradigm for high-performance and energy-efficient computing. Here, we propose the first Spike-driven VLA framework enabling end-to-end direct training for robotic manipulation, which mainly comprises three core components. First, we develop spiking visual and instruction encoders for multimodal perception, encoding visual observations and language instructions into sparse, reliable spike representations for subsequent cross-modal fusion. Then, we introduce Multi-Winner Spike Fusion for instruction-guided scene understanding, using bidirectional top- k winner-take-all spike routing to suppress background interference and yield fused memory. Finally, we propose a Spike Action Chunking Transformer that incorporates spiking cross-attention over the fused memory and the current robot state, enabling efficient end-to-end generation of continuous action chunks for robotic control. Extensive experiments on LIBERO and Meta-World demonstrate that Spike-driven VLA achieves competitive performance with fewer parameters and lower estimated inference energy than conventional VLA models. This work establishes a foundational framework for neuromorphic VLA modeling, paving the way for future advances in resource-efficient embodied intelligence.
[NLP-58] When the Right Answer Is Missing: An Arithmetic-Dependent Rejection Bottleneck in Jev
【速读】： 该论文旨在解决生成式大模型（Generative LLMs）在决策工作流中效率较低的问题，提出通过类型化决策模型（Typed decision models），如Jev，直接从预定义选项中选择来提升决策效率。其核心挑战在于：当候选集不包含有效答案时，尽管存在“其他”或“无以上选项”等显式拒绝选项，模型仍频繁错误接受无效答案，导致拒绝能力严重不足。研究发现，这种现象表现为显著的算术依赖性拒绝瓶颈——在存在正确答案时，模型准确率可达99%，但在答案缺失情况下，正确拒绝率仅7%；该差距在不同数值量级、运算深度、上下文表述及拒绝标签形式下均持续存在，并扩展至时间计算与容量舍入等场景。值得注意的是，原生布尔验证（Boolean verification）在相同答案缺失的算术问题上可实现99%的精确匹配准确率，表明模型具备判断答案正确性的能力，但其分类拒绝机制却失效。解决方案的关键在于引入一个基于独立开发集确定的简单决策阈值，该方法将算术拒绝准确率从7%提升至79%，同时保持97%的答案存在准确率，且无需重新训练或增加推理开销，显著缓解了该缺陷。
链接: https://arxiv.org/abs/2609.39496
作者: Jike Zhong,Ming Li,Yuxiang Lai
机构: University of Southern California(南加州大学); University of Florida(佛罗里达大学); Emory University(埃默里大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Typed decision models such as Jev offer an efficient alternative to generative LLMs in decision-making workflows by selecting directly from predefined options. When candidate sets contain no valid answer, TypeSafe recommends including an “other” or “none-of-the-above” option to enable rejection. In this report, however, we identify an arithmetic-dependent rejection bottleneck: Jev reliably selects correct numerical answers when available but frequently accepts incorrect alternatives when they are absent despite an explicit rejection option. On paired arithmetic problems, answer-present accuracy reaches 99%, while correct rejection falls to 7%. Moreover, this gap persists across numerical magnitudes, operation depths, contextual formulations, and rejection labels, and extends to scenarios such as time calculation and capacity rounding. Yet native Boolean verification achieves 99% exact-match accuracy on the same answer-absent arithmetic cases, showing that categorical rejection can fail even when the model successfully verifies candidate correctness. Finally, we show that a simple decision threshold selected on separate development problems raises arithmetic rejection accuracy from 7% to 79% while retaining 97% answer-present accuracy, substantially mitigating the failure without retraining or additional inference.
[NLP-59] Right-Wing Rock or Just Rock? A Computational Linguistic Analysis of Frei.Wild EMNLP2026
【速读】： 该论文旨在解决如何识别音乐作品中隐含的右翼极端主义倾向，特别是在政治立场模糊、难以界定的“灰色地带”案例中。针对这一问题，研究提出了一种基于计算语言学的方法，其关键在于构建一个德国右翼摇滚（Rechtsrock）语料库作为参照基准，并结合词汇分析与机器学习分类器进行文本内容检测。通过对比分析目标乐队的歌词特征，研究发现尽管该乐队在政治立场上保持模糊性，但其语言模式显示出明显的民族主义叙事倾向；两个高性能分类器（最高达到97% ROC-AUC得分）对超过半数歌曲的判定结果表明其具有右翼极端主义特征。该方法为监管机构提供了一种可扩展、数据驱动的技术手段，以更精准地识别潜在的极端化传播载体，尤其适用于边界案例的判别。
链接: https://arxiv.org/abs/2609.39460
作者: Carlotta Schneeberger(1),Kevin Tang(1 and 2) ((1) Heinrich Heine University Düsseldorf, (2) University of Florida)
机构: Heinrich Heine University Düsseldorf (海因里希·海涅杜塞尔多夫大学); University of Florida (佛罗里达大学)
类目: Computation and Language (cs.CL)
备注: 20 pages, 9 figures, for code and data see this https URL , to be published in the proceedings of the NLP 4 Positive Impact workshop at EMNLP 2026
Abstract:Rechtsrock is a subgenre of rock music that spreads right-wing ideology, often instrumentalized to recruit adolescents into the radical scene. Monitoring institutions counteract this by manually examining and, in some cases, banning extremist content; however, there are border cases that evade regulation. We present a study aimed at determining whether such a case, the band this http URL, should be classified as politically right-leaning or as part of the general German rock genre. We sampled a German rock dataset and created a corpus for right-wing rock to use as reference in this analysis and found that we can confirm the intuitions from previous investigations that this http URL successfully maintains an ambiguity with regard to their political affiliation. However, the tendency is towards the right-wing spectrum. Lexical analyses reveal nationalistic narratives and two high-performing classifiers (up to 97% ROC-AUC score) label more than half of their songs as right-wing extremist. Our analysis provides insight into how computational methods can improve the process of identifying right-wing extremist tendencies in music, especially in borderline cases like this http URL. The code and data are made available for future research.
[NLP-60] From Speech to Editable Concepts: Probing Emotion Recognition with Concept Bottleneck Models ICASSP2027
【速读】： 该论文旨在解决语音情感识别（Speech Emotion Recognition, SER）中模型性能受限于多模态输入融合不充分、可解释性差的问题，尤其关注在使用大语言模型（Large Language Models, LLMs）进行SER时，如何理解不同输入成分（如文本转录、声学描述及说话人属性）对具体情感预测的影响。其解决方案的关键在于引入概念瓶颈模型（Concept Bottleneck Models, CBMs）的思想，将情感分类任务分解为基于可解释概念的中间表征路径，通过分离提取文本、声学特征和说话人属性等概念，并分析这些概念在零样本与微调设置下对个体预测结果的影响。实验表明，直接使用LLMs时存在显著的文本偏倚，导致在剧本化语料库上宏平均F1分数从27.8降至5.8；而通过微调可缓解此偏倚，使性能提升至45.1。此外，移除语音速率或强度等声学概念虽对整体指标影响较小，却显著改变特定类别（如将中性情绪误判为厌恶）的预测结果，揭示了仅依赖聚合性能指标无法全面反映概念移除对个体预测的真实影响。因此，该研究强调了结合可解释性框架评估多模态情感识别模型的重要性。
链接: https://arxiv.org/abs/2609.39453
作者: Hezhao Zhang,Thomas Hain
机构: 未知
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
备注: 5 pages, 2 figures. Submitted to ICASSP 2027
Abstract:Speech emotion recognition (SER) is the task of assigning emotion labels to utterances. Early systems relied on acoustic features, whereas recent approaches combine multiple modalities, most commonly speech and text. Still, performance remains poor on many datasets. Large language models (LLMs) have therefore attracted interest for SER, as they can process diverse inputs jointly with instructions. However, direct audio input raises questions of explainability. To address similar questions in image classification, concept bottleneck models were introduced. This work adapts concept bottlenecks to SER to examine how individual predictions depend on transcripts, acoustic descriptions and speaker attributes. Experiments test three LLMs on CREMA-D, IEMOCAP and MELD, with concepts extracted by separate tools. On scripted corpora, LLMs are strongly biased towards the transcript in the zero-shot setting, which lowers Macro-F1 from 27.8 to 5.8 on CREMA-D. Fine-tuning removes this bias, and the transcript raises Macro-F1 from 41.8 to 45.1. Removing speech rate changes 48% of Neutral predictions to Disgust on CREMA-D; removing intensity level on MELD changes predictions despite little change in Macro-F1. These findings show that aggregate performance changes alone do not capture the effects of concept removal on individual predictions.
[NLP-61] Synthetic Data Characterization via Training Dynamics EMNLP2026
【速读】： 该论文旨在解决生成式人工智能（Generative AI）所生成数据在不同学习任务中属性可解释性不足的问题，尤其关注其在泛化能力与可靠性方面的局限性。核心问题在于：如何评估和理解大语言模型（LLM）生成数据的样本级可学习性（sample-level learnability），并在此基础上揭示其与人类撰写数据在分布特性与模型适应性上的差异。该研究的关键解决方案在于构建基于编码器训练动态的实证数据分布分析框架，通过追踪不同规模与架构的LLM生成数据及人工数据在训练过程中的学习轨迹，量化其学习信号的稳健性，并据此评估基于可学习性信号的数据选择策略对两类数据源的差异化影响，从而为高质量合成数据的筛选与应用提供可量化的理论依据。
链接: https://arxiv.org/abs/2609.39447
作者: Irene Lago,Ana Ezquerro,David Vilares
机构: Universidade da Coruña, CITIC(拉科鲁尼亚大学，计算与信息技术中心), Spain; Graz University of Technology, IML(格拉茨技术大学，智能机器学习研究所), Austria
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted at Findings of EMNLP 2026
Abstract:Interpreting properties of LLM-generated data is important for understanding its utility and limitations across learning tasks. In this work, we characterize synthetic data through sample-level learnability, studying variation among LLM families and scales, alongside human-written data as a reference. We first generate synthetic datasets spanning single- and multi-label classification, labeling, and tree prediction tasks. We then derive empirical data distributions from encoder training dynamics for both machine and organic data, and estimate the robustness of these distributions across encoders. Finally, we evaluate how data selection strategies based on these learnability signals affect both data sources differently.
[NLP-62] QuantCode Model: Specializing Language Models for Executable Algorithmic Trading Code
【速读】： 该论文旨在解决大语言模型在生成可执行的算法交易策略时面临的挑战，即如何将自然语言描述的交易策略准确转化为特定交易框架（如Backtrader）下的正确程序逻辑，并确保其在历史数据上的回测有效性与语义一致性。其核心解决方案包含两个互补机制：一是针对算法交易框架代码进行持续预训练（continued pretraining），以增强模型对领域代码结构的理解；二是基于经代理验证的“请求-代码”配对进行监督微调（SFT），以提升指令遵循能力和任务成功率。实验表明，持续预训练显著提升了单轮评估中的“裁判通过率”（Judge Pass），而后续加入SFT则带来更大增益，尤其在多轮协作式代理评估中，首次尝试成功率从22.3%提升至58.3%，最终成功率从47.5%提升至79.5%。此外，研究发现领域专业化会导致解析器兼容的结构化工具调用能力退化，虽可通过针对性的SFT恢复格式规范，但无法完全复现基线检查点在仓库级代理任务中的性能，揭示了能力保留失效的问题。因此，框架导向的预训练、经验证的SFT以及显式的能力建设评估，共同针对领域专用可执行代码生成中的不同失效模式提供了系统性缓解路径。
链接: https://arxiv.org/abs/2609.39420
作者: Alexey Chernysh,Orkhan Ekhtibarov,Dmitry Zmitrovich
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG); Trading and Market Microstructure (q-fin.TR)
备注: 16 pages, 2 figures, 6 tables
Abstract:Large language models are strong general-purpose code generators, but executable algorithmic trading remains a demanding specialization target: a model must translate a natural-language strategy specification into correct program logic for a specialized trading framework, execute on historical data, produce trades, and remain semantically faithful to the request. We study two complementary mechanisms for specializing language models for this setting: continued pretraining on algorithmic-trading framework code and supervised fine-tuning (SFT) on agent-validated request-to-code pairs. Evaluation is centered on QuantCode-Bench, our 400-task benchmark for Backtrader strategy generation, together with a repository-level SWE-bench-like track. Continued pretraining improves single-turn Judge Pass from 41.5% to 47.5% for Qwen3.5-397B-A17B and from 27.8% to 33.0% for Qwen3.6-35B-A3B. SFT applied after continued pretraining yields a larger gain for Qwen3.6-35B-A3B, reaching 58.2% Judge Pass and 83.5% successful backtests; in agentic evaluation it raises first-turn success from 22.3% to 58.3% and final success after up to 10 turns from 47.5% to 79.5%. Continued pretraining alone improves first-turn agentic success but lowers final success after repair from 47.5% to 32.5%, consistent with degraded instruction following, whereas SFT improves both. We also identify a capability-retention failure: domain specialization degrades parser-conformant structured tool calling, and targeted recovery SFT restores tool-call formatting but not the base checkpoint’s repository-level agent performance. The results show that framework-oriented pretraining, validated SFT, and explicit capability-retention evaluation address distinct failure modes in domain-specific executable code generation.
[NLP-63] Can Computation from Earlier Problems Help LLM s Solve New Ones?
【速读】： 该论文旨在解决大语言模型在连续对话中，先前问题的计算信息是否能有效辅助后续问题求解的问题。尽管历史上下文被保留，但其对后续回答准确率的影响具有不确定性，可能提升也可能降低，尤其在同一领域内表现不一致。为理解这一现象，研究采用受控回放（controlled replay）方法，隔离并分析每个问题-历史组合引发的内部状态变化，发现这些变化在不同历史背景下仍保持当前问题间关系的一致性。针对此问题，论文提出关键解决方案——STAIR（Stale-Token Attention for Inter-query Reuse），通过构建一个固定存储池（bank）保存早期响应生成过程中的键（key）与值（value），并在当前查询处理时引入注意力机制，学习如何重定向当前查询以利用该存储池中的历史信息。该方法仅需训练12,288个参数，且基础模型保持冻结。在三个Qwen模型和四个基准测试中，STAIR相较于未使用历史信息的原始模型，在后期轮次平均准确率上最高提升达11.67个百分点，显著增强了模型在保留历史条件下的推理能力。
链接: https://arxiv.org/abs/2609.39394
作者: Jipei He,Wenhui Tan,Xiaoyi Yu,Enver Sangineto,Fiorenzo Parascandolo,Rita Cucchiara,Ruihua Song
机构: Gaoling School of Artificial Intelligence, Renmin University of China(中国人民大学高瓴人工智能学院); University of Modena and Reggio Emilia(摩德纳与雷焦艾米利亚大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 29 pages, 7 figures
Abstract:Large language models often solve independent problems in the same conversation. Can computation from earlier problems help them solve new ones? To answer this question, we first conduct preliminary experiments showing that retained history can raise or lower later-turn accuracy, even within the same domain. To understand these effects, we use controlled replay to isolate internal state changes specific to each problem-history pairing. Across different histories, these changes preserve similar relationships among current problems. To improve reasoning under retained history, we introduce STAIR (Stale-Token Attention for Inter-query Reuse). STAIR captures keys and values from earlier response generation in a fixed bank. It learns to redirect current queries when they read this bank during prompt processing. The base model remains frozen; only 12,288 parameters are trained. Across three Qwen models and four benchmarks, STAIR improves average later-turn accuracy by up to 11.67 percentage points over the unmodified model with history.
[NLP-64] Lab at Daleel 2026: STAR-Ar Sequence Tagging for Argument Recognition in Arabic
【速读】： 该论文旨在解决阿拉伯语中论点挖掘（Argument Mining, AM）任务资源匮乏的问题，特别是针对辩论与社论文本中论点话语单元（Argumentative Discourse Units, ADUs）的识别与分类。其核心挑战在于如何在低资源条件下实现高精度的论点结构检测与类别划分。解决方案的关键在于提出一种BERT-BiLSTM-CRF架构（\texttt{STAR-Ar}），将论点话语单元的检测与分类联合建模为一个基于词元级别的序列标注任务，通过融合上下文感知的Transformer嵌入与基于结构化的转移约束，有效提升论点边界识别的准确性。实验结果显示，该模型在验证集和测试集上的F1分数分别达到72.69和73.7，表明其在阿拉伯语论点挖掘任务中的有效性。此外，领域分析揭示仅在社论数据上训练的模型性能显著低于在辩论数据上训练的模型，主要归因于社论数据集规模较小。相关代码已开源，可供后续研究参考。
链接: https://arxiv.org/abs/2609.39385
作者: Bhuvanesh Verma,Ali Abusaleh,Alexander Mehler
机构: Text Technology Lab (TTLab); Goethe University Frankfurt
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at ArabicNLP 2026 Daleel-2026 shared task
Abstract:Argument Mining (AM) is a critical NLP task that remains significantly under-resourced in Arabic. This paper presents \testttSTAR-Ar , a BERT-BiLSTM-CRF architecture for argument discourse detection and classification, as our system for Daleel 2026, the inaugural Arabic argument mining shared task. The task requires the identification and classification of argumentative discourse units (ADUs) in debate and editorial this http URL jointly model these two objectives as a token-level sequence labeling task using a BERT-BiLSTM-CRF architecture that combines contextual transformer embeddings with structural transition constraints to support accurate span detection. \testttSTAR-Ar achieves an F1-score of 72.69 on validation and 73.7 on test data. Our domain-specific analysis shows that models trained exclusively on editorials underperform those trained on debates, a disparity we primarily attribute to the smaller size of the editorial dataset. The code for \testttSTAR-Ar is available at \hrefthis https URL\faGithub~TTLab at Daleel 2026
[NLP-65] Exploring Heterogeneous Model Merging Approach for Complex Knowledge Transfer
【速读】： 该论文旨在解决如何将特定任务型模型（specialist model）中的任务导向行为高效迁移至通用语言模型（general language model）的问题，传统方法通常依赖于训练、知识蒸馏或表示对齐等复杂过程。其核心解决方案在于提出一种无需训练的异构参数合并方法，直接在参数层面实现能力迁移。关键创新点在于采用两种已有的无训练异构合并技术——交集合并（Intersection-Merge, IM）与激活-剪枝-合并（Activate-Prune-Merge, APM），通过将专用模型的参数投影到通用模型的结构空间中，不依赖梯度更新或语义对齐，仅基于参数形状匹配与激活统计信息进行维度选择与注入。实验表明，该方法在嵌入、重排序、奖励建模及MoE代码专家模型迁移等多种场景下均有效提升通用模型性能，证明了简单异构合并即可在不同专业角色间实现能力迁移。
链接: https://arxiv.org/abs/2609.39369
作者: Jiahe Fan,Si Chen,Yinghao Hou,Wenbo Xia,Ke Xu,Hong Xie,Enhong Chen
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 6 pages, 1 figure, 7 tables. Preprint
Abstract:Specialized models encode task-oriented behavior, but transferring that behavior to a general language model usually requires training, distillation, or representation alignment. We study whether such ability can instead be transferred directly at the parameter level. We apply two existing training-free heterogeneous merging methods, previously shown to transfer knowledge between general language models, to specialist-to-general transfer, projecting a specialist donor into the recipient’s shape and interpolating backbone parameters without gradient updates or semantic alignment. Intersection-Merge (IM) injects a prefix-aligned donor slice matching the recipient shape, while Activate-Prune-Merge (APM) uses forward-pass activation statistics to select which donor dimensions to retain before injection. Across embedding, reranking, reward modeling, and MoE code-specialist transfer, both methods improve the general recipient, showing that simple heterogeneous merging can move capabilities across diverse specialist roles.
[NLP-66] Making Grid Beam Search Less Greedy
【速读】： 该论文旨在解决自回归文本生成模型在施加词汇约束（lexical constraints）时，现有解码方法中存在的不均衡约束处理问题。具体而言，尽管网格束搜索（grid beam search）相较于确定有限自动机约束束搜索（DFA-constrained beam search）具有指数级的计算效率优势（仅需线性数量的前向传播），但其在执行过程中会优先满足较易满足的约束，导致更难满足的约束被延迟至序列末尾，从而引入系统性偏差。这一偏差破坏了约束处理的公平性，可能影响生成结果的质量与可靠性。为解决该问题，论文提出了一种改进方法——公平网格束搜索（fair grid beam search），该方法在保持线性前向传播复杂度的同时，通过重构约束处理机制，使所有约束获得平等处理机会，避免了对易满足约束的偏好。实验结果表明，公平网格束搜索不仅有效消除原有偏差，还在两个约束生成任务中显著提升了生成序列的联合概率，验证了其有效性与优越性。
链接: https://arxiv.org/abs/2609.39368
作者: Sean Papay,Roman Klinger
机构: University of Bamberg(巴伐利亚大学); Germany(德国)
类目: Computation and Language (cs.CL)
备注: Published as a conference paper at COLM 2026
Abstract:A common formalism for constraining the output of autoregressive text generation models involves lexical constraints, words or phrases which are required to occur in the generated text. DFA-constrained beam search and grid beam search are two widely used paradigms for decoding from autoregressive models while enforcing lexical constraints. As the former approach requires a number of forward passes exponential in the number of constraint tokens, it is often dispreferred to the latter, which requires only linearly many forward calls. However, while grid beam search achieves an exponential speedup, it does so in a manner which does not treat all of the constraints equally. In this paper, we demonstrate that grid beam search is biased to incorporate easier-to-satisfy constraints first, leaving harder constraints to the end of the sequence. This contrasts with DFA-constrained beam search, which exhibits no such bias. To address this shortcoming, we propose fair grid beam search, a modification to grid beam search which avoids this bias while still requiring only linearly many forward passes. Experimentally, we confirm grid beam search’s bias on two constrained generation tasks, finding significant differences in how it orders constraint tokens as compared to DFA-constrained beam search and fair grid beam search. Furthermore, we find that fair grid beam search not only fixes grid beam search’s bias, but finds higher-probability strings in the process.
[NLP-67] Ready2Blend: From Natural-Language Instructions to Composable Alignment Prompts
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在持续对齐（continual alignment）过程中面临的“遗忘问题”——即模型在适应新需求时会丢失先前已学习的行为。现有方法中，自然语言指令虽具备灵活性和可组合性，但仅提供间接控制；而基于后训练（post-training）的方法虽能实现强适应性，却需重复更新模型参数，成本高昂。为克服上述局限，论文提出Ready2Blend，其核心在于将自然语言的灵活性与预训练对齐知识相结合。关键创新在于AlignFormer架构：它将每项需求映射至一个固定长度的对齐提示（alignment prompt），并将其存储于模块化提示库（modular prompt bank）中，同时保持主干网络和先验提示冻结，从而实现零参数更新。通过可组合性正则化（composability regularization），该方法将文本需求的语义几何结构迁移至提示空间，支持推理时的动态混合与权重调整。实验表明，在两种实际持续对齐场景下，Ready2Blend是唯一达到后训练方法性能水平的冻结主干方法，性能可达联合训练参考模型的93.1%–98.5%，且保留能力优异，仅需少量提示词，训练时间最多减少4.3倍。其模块化设计还支持无需重训练的加权个性化与无序组合。
链接: https://arxiv.org/abs/2609.39365
作者: Jeesu Jung,Hwan Chang,Juseon Do,Jeonghwan Choi,Jinho Choo,Sungwoo Nam,S. K. Hong,Hwanjun Song
机构: Korea Advanced Institute of Science and Technology (韩国科学技术院); Samsung SDS
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 24 pages
Abstract:Continual alignment requires LLMs to adapt to new requirements without forgetting previously acquired behaviors. Natural-language instructions are flexible and composable but offer only indirect control, whereas post-training provides stronger adaptation at the cost of repeated parameter updates. We introduce Ready2Blend, which combines the flexibility of natural language with learned alignment. AlignFormer maps each requirement to a fixed-length alignment prompt stored in a modular prompt bank, while the backbone and prior prompts remain frozen. Composability regularization transfers the semantic geometry of textual requirements into prompt space, enabling inference-time blending and reweighting. Across two practical continual alignment settings, Ready2Blend is the only frozen-backbone method that matches post-training-based alignment methods, reaching 93.1 - 98.5% of a joint-training reference with competitive retention, while requiring only a few prompt tokens and up to 4.3\times less training time. Its modular design further enables weighted personalization and order-free composition without retraining. Code will be released upon acceptance.
[NLP-68] Offline Guidance Online Reasoning : Reusing LLM Feedback for Small Language Models
【速读】： 该论文旨在解决大语言模型（LLM）与小语言模型（SLM）之间的能力-部署鸿沟问题，即如何在不依赖在线调用昂贵的LLM API、且保持SLM可本地部署优势的前提下，显著提升SLM的推理能力。现有方法要么通过知识蒸馏进行离线训练（需参数更新），要么采用在线协作方式（需反复调用LLM），但前者增加训练成本，后者无法复用已生成的指导信息。本文提出一种名为可重用隐层修正（Reusable Latent Correction, RLC）的新方案，其核心在于将来自黑盒LLM的一次性自然语言指导转化为存储于外部记忆库中的持久化隐空间修正经验，并根据SLM当前推理状态动态检索和应用这些修正。该方法在推理阶段完全由固定参数的SLM独立完成，无需任何在线LLM调用或参数更新。实验结果表明，RLC在多个推理基准测试及不同规模的SLM上均能一致提升推理性能，有效实现了对LLM能力的高效复用。
链接: https://arxiv.org/abs/2609.39346
作者: Bohan Zhang(1),Linan Yue(1),Weibo Gao(2),Pengyu Chen(1),Hong Guo(1),Yanqi Hao(3) ((1) Southeast University, (2) Hong Kong Polytechnic University, (3) ZTE Corporation)
机构: Southeast University (东南大学); Hong Kong Polytechnic University (香港理工大学); ZTE Corporation (中兴通讯)
类目: Computation and Language (cs.CL)
备注: 29 pages. Code: this https URL
Abstract:Large language models (LLMs) offer strong reasoning capabilities but are often costly to access through commercial APIs, while small language models (SLMs) are easier to deploy locally yet remain weaker in reasoning. This capability-deployment gap has motivated LLM-SLM collaboration, which aims to improve SLM reasoning using LLM capabilities while preserving the deployment advantages of SLMs. Existing approaches mainly follow two paradigms. Knowledge distillation uses LLM-generated answers and reasoning trajectories to train SLMs offline, but requires parameter updates and additional training. Alternatively, online collaboration routes difficult problems to an LLM or leverages LLM-generated guidance and corrections when an SLM encounters difficulties. Although effective, online collaboration requires repeated LLM access. Moreover, the guidance produced for a particular problem is discarded after inference and cannot benefit subsequent problems involving similar reasoning states. In the paper, we focus on a more constrained setting in which the LLM is accessed only offline, the SLM parameters remain fixed, and online inference is performed solely by the SLM. To this end, we propose Reusable Latent Correction (RLC), which converts one-off natural-language guidance from a black-box LLM into persistent corrective experiences in the hidden space of an SLM. RLC stores these experiences in an external bank and retrieves them according to the SLM’s current reasoning state, enabling the SLM to reuse LLM-derived corrections during inference without any online LLM calls. Experiments across multiple reasoning benchmarks and SLM scales show that RLC consistently improves SLM reasoning without parameter updates or online LLM calls. Code is available at this https URL.
[NLP-69] Understanding as No-Arbitrag e: Bounded Dutch Books as a Definition and Training Objective for Language Models
【速读】： 该论文旨在解决生成式 AI（Generative AI）在语言建模过程中是否真正“理解”其所生成内容的核心问题，将“理解”定义为一种可量化的逻辑一致性（logical coherence），并引入无套利（no-arbitrage）框架作为衡量标准。其核心挑战在于：现有模型在训练中仅优化下一个词元的预测概率，而这一目标本身导致模型在不同逻辑结构的命题间产生系统性不一致，从而暴露于可被利用的“荷兰赌”（Dutch book）风险之中。解决方案的关键在于提出 Arbitr 训练框架，通过引入对抗性交易者（adversarial trader）对模型输出中的逻辑矛盾进行惩罚，并结合校准锚点（calibration anchor）防止模型退化为无信息性输出。该方法显著降低了模型在多种表述形式下的可被剥削性，且效果可泛化至未见的逻辑模式与新模型架构。研究进一步揭示了“缩放幻觉”（scaling illusion）——在70亿参数规模下，极低的不一致性测量值常伴随极端且不合理自信，表明逻辑一致性虽是知识的必要条件，但不足以构成充分条件。
链接: https://arxiv.org/abs/2609.39341
作者: Daniel Dragonevskiy
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 18 pages
Abstract:Does a language model merely predict tokens, or does it understand what it says? We make this question measurable by defining “understanding” through the lens of no-arbitrage. A model understands a vocabulary to a certain degree if a computationally bounded trader cannot extract guaranteed profit by betting against the model’s probabilities on logically related claims (a “Dutch book”). We establish three theoretical results: first, because full logical coherence is computationally intractable, understanding is inherently graded, not absolute. Second, we prove that the exact optimum of standard next-token prediction is inherently incoherent across different question formats; the flaw lies in the training objective, not the architecture. Third, we show that uncertainty accumulates predictably along reasoning chains, making unjustified overconfidence an arbitrage opportunity in itself. To address this, we introduce Arbitr, a training framework where an adversarial trader penalizes the model for logical inconsistencies, paired with a calibration anchor to prevent uninformative collapse. Across five pre-registered experiments on Qwen2.5 and Phi-3.5 models, we demonstrate that standard models are highly exploitable across different phrasings. Arbitr reduces this exploitability by orders of magnitude without sacrificing task accuracy, and the effect successfully transfers to unseen logical patterns and new model families. Crucially, we uncover a scaling illusion: at 7B parameters, near-zero measured incoherence often coincides with extreme, unjustified confidence. We conclude that while Arbitr enforces rigorous logical consistency, coherence is a necessary condition for knowledge, but not a sufficient one
[NLP-70] aming Speculative Search for Test-Time Scaling in LLM Serving
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在推理阶段进行推测执行（speculative execution）时所面临的两大核心挑战：一是候选推理路径搜索空间的急剧膨胀，二是对大量冗余候选路径进行频繁且细粒度的验证任务所带来的计算开销。为应对这些问题，论文提出了一种名为SpecScale的高效推测执行服务系统，其关键解决方案在于通过三项核心技术实现延迟与计算开销之间的平衡：（1）早期剪枝低质量候选路径以减少无效搜索；（2）通过跨冗余候选路径的计算去重来提升资源利用效率；（3）将细粒度验证任务延迟执行，从而优化整体调度策略。实验结果表明，SpecScale在MATH和奥数等高难度推理基准测试中显著优于非推测及现有推测方法，在保持答案准确率的同时大幅提升了吞吐量并降低了延迟。
链接: https://arxiv.org/abs/2609.39334
作者: Jinwoo Jeong(Korea University),Woohyung Choi(Korea University),Myeongjae Jeon(POSTECH),Jeongseob Ahn(Korea University)
机构: Korea University (韩国大学); POSTECH (浦项科技大学)
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Computation and Language (cs.CL); Operating Systems (cs.OS)
备注: 14 pages
Abstract:Test-time scaling has recently emerged as a powerful approach for improving LLM reasoning by allocating additional computation during inference, substantially enhancing accuracy on challenging tasks such as mathematics and coding. To accelerate the exploration of reasoning paths, recent studies proposed speculative execution. However, we show that supporting speculative execution poses two unique challenges for LLM serving systems: (1) an explosion in the search space of candidate paths and (2) frequent, fine-grained verification tasks for candidates. To address these challenges, this paper proposes SpecScale, a serving system for efficient speculative execution. We introduce three techniques to reconcile the trade-off between latency and computational overhead: (1) early pruning of low-quality candidate paths, (2) deduplicating computation across redundant candidate paths, and (3) deferring fine-grained verification tasks. We evaluate SpecScale on challenging reasoning benchmarks, including MATH and Olympiad. Our results show that SpecScale significantly outperforms both non-speculative and recent speculative approaches, delivering substantial improvements in throughput and latency while preserving answer quality. Comments: 14 pages Subjects: Distributed, Parallel, and Cluster Computing (cs.DC); Computation and Language (cs.CL); Operating Systems (cs.OS) Cite as: arXiv:2609.39334 [cs.DC] (or arXiv:2609.39334v1 [cs.DC] for this version) https://doi.org/10.48550/arXiv.2609.39334 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-71] A Tilted Bowl Is Not a Slippery Slope: Compressing Looped Models
【速读】： 该论文旨在解决生成式模型中循环结构（looped models）在量化压缩后出现性能崩溃（collapse）的问题。传统观点认为，这种崩溃主要由循环过程中累积的舍入误差（rounding error）导致，尤其在多次迭代中不断放大。然而，本文通过对五类共30余种模型的系统性实验发现，这一解释仅适用于从未收敛（never settle）的循环结构；对于能够收敛（settle）的循环，固定大小的舍入误差并不会持续累积，而是仅改变模型最终稳定点的位置，类似于倾斜碗面使小球静止位置偏移。只有当该偏移超过读出机制（readout tolerance）的容忍范围时，模型输出才会失效。基于此新认知，研究提出可通过一次无标签测量即可预测模型是否失败，并解释了为何部分失败模型在后期使用8位权重进行少量额外迭代后能恢复性能——因其仍具备收敛能力。受此启发，作者设计了一种控制器，通过监测模型的“停止头”（halting head）信号，在模型趋于稳定时主动终止高精度计算，并切换至8位权重完成剩余迭代。在Sudoku-Extreme和Maze-Hard等任务上，该方法相比固定深度推理，在重量传输减少三分之二的情况下，性能提升最高达15点。关键创新在于：将循环模型的稳定性分析从“误差累积”转向“收敛点漂移”，并据此实现动态优化与高效推理。
链接: https://arxiv.org/abs/2609.39277
作者: Steven Kolawole,Pearse Jim,Opegbemi M. Busoye,Glory Bagai,Virginia Smith
机构: Carnegie Mellon University (卡内基梅隆大学); ML Collective
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Preprint; in review
Abstract:Looped models reason by applying the same block of weights many times, so compressing that block saves memory traffic on every loop. Compressed looped models, however, often collapse, and the collapse is usually blamed on rounding error that accumulates from loop to loop. In this work we test that account on more than 30 models from five families and find, to our surprise, that it holds only for loops that never settle. When a loop settles, a fixed rounding error does not accumulate. It moves the point where the loop settles, much as tilting a bowl moves where a ball comes to rest, and the answer is lost only when the shift is larger than the readout tolerates. This picture lets us predict which models fail from a single label-free measurement, and it tells us why failed models recover: their loops still settle, so a few final loops with 8-bit weights bring the answer back. Motivated by these findings, we build a controller that stops when the model’s halting head fires and then finishes with 8-bit loops. On Sudoku-Extreme and Maze-Hard it beats fixed-depth inference by up to 15 points under a third of the weight traffic.
[NLP-72] Concept Subspaces Compute Beyond the Logit Lens: A Weights-Only Test for Locating Representations Upstream of Readout DATE
【速读】： 该论文旨在解决生成式模型中概念子空间（concept subspace）与输出读出（output readout）之间关系不明确的问题，即现有方法无法有效判断所提取的概念子空间是否真正关联于模型的输出决策机制。其核心解决方案是提出一种双向几何诊断（two-sided geometric diagnostic），通过测量提取的子空间与未嵌入矩阵（unembedding matrix）主导右奇异方向之间的重叠度，并以面向输出的正向控制作为基准进行评估。该诊断仅需模型权重即可计算，具备可操作性。实验基于格式无关推理子空间（FARS，一个从18个推理概念在6种表面形式下提取的10维基底）开展，在9个秩匹配估计器和26个模型中发现，4种由激活推导的概念估计器仅在前十大读出方向上保留0.38%–0.80%的平均能量，远低于最终层主成分分析（final-layer PCA）的3.56%，且在25/26个模型中表现逊于FARS。进一步引入同层下一个词预测控制（same-layer next-token control），经深度匹配线性转换器校准后，其能量约为FARS的13倍，且所有25个测试模型均表现出显著差异。此外，对10个互斥概念重新提取FARS后，在24个生成模型中实现62%–100%的跨格式检索性能，表明该提取过程具有迁移能力而非依赖固定基底。互补的四模型三种子干预研究亦显示模型依赖性的源导向效应，但均远低于全向量替换水平。综合几何诊断与干预实验结果，有效区分了概念结构与主导读出方向之间的差异，同时避免对因果充分性做出过度宣称。
链接: https://arxiv.org/abs/2609.39263
作者: Aojie Yuan,Zhiyuan Julian Su,Haiyue Zhang,Zijian Su
机构: University of Southern California(南加州大学); Duke University(杜克大学); University of Michigan(密歇根大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 54 pages. Substantially revised preprint: new title, expanded model coverage, readout-geometry controls, supplementary intervention and transfer experiments, revised interpretation, updated figures and author list
Abstract:A concept subspace’s effect on model behavior does not establish how it relates to the output readout. We introduce a two-sided geometric diagnostic that measures an extracted subspace’s overlap with the dominant right-singular directions of the unembedding matrix, evaluated against output-oriented positive controls. Given an extracted basis, the raw diagnostic requires only model weights. Our testbed is the Format-Agnostic Reasoning Subspace (FARS), a ten-dimensional basis extracted from eighteen reasoning concepts expressed in six surface forms. Across nine rank-matched estimators and twenty-six models, four activation-derived concept estimators carry only 0.38–0.80% mean energy in the top-ten readout span. Final-layer PCA carries 3.56%, exceeding FARS in 25 of 26 models. A same-layer next-token control, evaluated using a fitted linear translator for depth matching, carries approximately thirteen times more energy than FARS, with separation in all 25 tested models. Re-extracting FARS on ten disjoint concepts yields 62–100% cross-format retrieval across twenty-four generative models, demonstrating transfer of the extraction procedure rather than a fixed basis. A complementary four-model, three-seed intervention study finds model-dependent source-directed effects that remain well below full-vector replacement. Together, the geometry and intervention controls distinguish concept structure from dominant readout directions while limiting claims of causal sufficiency.
[NLP-73] 4MT-VLM: How Coarse Is a VLMs Cognitive Map?
【速读】： 该论文旨在解决视觉语言模型（VLM）在动态视角下维持稳定、三维世界表征的能力问题，即模型能否在未见过的视角中准确识别位置。其解决方案的关键在于构建4MT-VLM数据集，该数据集通过程序化生成景观，并在保持布局不变的前提下，通过五种不同的刺激模式（形状与颜色、仅形状、仅颜色、无物体的裸露地形峰顶、以及将峰顶置于地平线的山谷视角）系统性地移除外观线索，从而严格检验模型对空间布局的感知能力。实验结果表明，尽管当前前沿模型（如Gemini 3.8 Flash、GPT-5.6）在静态视角下能以接近人类水平的表现识别位置，但一旦相机移动超过135°，其性能迅速下降至低于25%随机猜测水平，远低于人类85%的准确率；仅当干扰项间距超过30米时，模型才部分恢复性能。这表明现有VLM虽具备初步的认知地图能力，但其空间分辨率仍过于粗糙，无法在视角变化后维持稳定的三维环境理解。
链接: https://arxiv.org/abs/2609.39238
作者: Markus Frey
机构: Lamarr Institute for Machine Learning and Artificial Intelligence; Fraunhofer IAIS, University of Bonn
类目: Computation and Language (cs.CL)
备注:
Abstract:An agent that moves must recognise a place from a viewpoint it has never seen. We introduce 4MT-VLM, a dataset of procedurally generated landscapes, each rendered across five stimulus modes that remove appearance cues while holding layout fixed: shape and colour, shape only, colour only, bare terrain peaks with no objects, and a valley viewpoint that puts the peaks on the horizon. The last condition is commonly used in clinics to probe hippocampal function in human patients. We test this benchmark across sixteen different open and closed-source models and report 4AFC performance, a measure which is also used to grade human participants. We observe that models identify a place from the studied viewpoint but lose it once the camera moves, dropping below the 25% chance level at 135° where a human observer scores 85%. Frontier models (Gemini 3.8 Flash, GPT-5.6) answer only 39% and 31% of rotated trials correctly, recovering to 85% and 55% only when distractors are moved more than 30 meters apart. Our benchmark demonstrates that while current VLMs possess rudimentary cognitive maps, their spatial resolution remains fundamentally too coarse to maintain a stable, 3D understanding of the world once the viewpoint changes.
[NLP-74] RAIM: Robust Aggregation of Inexpensive Models for Hallucination Detection
【速读】： 该论文旨在解决生成式 AI（Generative AI）中忠实度（faithfulness）自动评估依赖昂贵且专有的前沿大模型作为评判者所导致的高成本与低可扩展性问题。其核心解决方案是提出一种名为RAIM的聚合框架，通过集成多个低成本、开源权重的小规模判别模型（4–9B参数量级），实现对前沿模型性能的有效替代。RAIM的关键在于其鲁棒性设计：结合交叉拟合的堆叠逻辑回归与基于成员自身输出的可接受性检验（admissibility test），能够识别在何种情况下聚合多个模型的表现优于其最佳单个成员，并确保整体性能接近前沿模型水平。实验表明，在仅需50–100条领域内标注数据进行一次校准的前提下，该面板模型在多数基准上保持了前沿模型约93%的Cohen’s κ值，平均仅损失2.9个百分点的平衡准确率；在部分任务上显著优于单一模型，而在少数任务上略逊于前沿模型，但整体表现与专用检测器相当甚至更优。更重要的是，是否值得采用该聚合方案，可通过校准集直接判断——当多个模型在不同样本上出现独立错误时，聚合可提升性能；而当某一模型主导时，堆叠器可有效恢复其优势，仅在极少数情形下前沿模型仍具明显领先。因此，该方法实现了以极低推理成本（仅为前沿模型的1/64）替代专属判别器的可行性。
链接: https://arxiv.org/abs/2609.39229
作者: Elia Onofri,Roberto Di Pietro
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 49 pages, 23 tables, 10 figures. Code and data: this https URL and this https URL
Abstract:Automatic evaluation of faithfulness increasingly relies on a large language model acting as a judge, yet the most reliable judges are proprietary frontier models, costly and ill-suited to high-throughput monitoring. We investigate whether a panel of cheap open-weight judges (4–9B) can be aggregated to stand in for a frontier one, what the substitution sacrifices, and when it is worth making. We propose RAIM, an aggregation scheme robust to the members’ correlated errors, coupling a cross-fitted stacked logistic regression with an admissibility test that, read from the members’ own outputs, identifies when aggregating them improves on their best member and stays within reach of the frontier judge. We instantiate RAIM with ten judges from disjoint families across eight faithfulness benchmarks. Against Claude Sonnet, the panel retains a median 93% of its Cohen’s \kappa and gives up only 2.9 points of balanced accuracy on average; read as paired differences, it clearly improves on one benchmark and clearly worsens on three (only two by a non-negligible margin), leaving four unresolved. At a sixty-fourth of the frontier’s inference price, the operative expense is a one-time in-domain calibration on 50–100 labelled records. The panel is also competitive with purpose-trained detectors on their home benchmarks (within 1.3 accuracy points of GPT-4o and 1.9 of the LLM-AggreFact leader), and beats the strongest one we reran by 6 points on our grounded sets. Whether aggregation pays depends on the members themselves: where several capable members err on different items, the panel improves on its best judge and approaches the frontier; where one dominates, the stacker recovers the leader, and only there does the frontier remain materially ahead. Both conditions are read off the calibration set at no further cost, so a cheap panel can stand in for a frontier one wherever this audit admits it.
[NLP-75] ViLegalExpert: A Large-Scale Benchmark for Vietnamese Legal Retrieval and Question Answering from Real-World Consultations
【速读】： 该论文旨在解决越南语法律人工智能（Legal AI）系统在应对真实法律咨询时缺乏可靠、权威证据支持的问题。现有越南语法律基准数据集覆盖范围有限，难以反映实际法律场景中的复杂性与多样性。为此，研究提出构建一个大规模的真实公民-律师咨询对话数据集——ViLegalExpert，涵盖超过17.2万条问题，涉及34个法律领域，并包含专业解答及专家验证的法律依据。该数据集支持法律信息检索、抽取式问答（extractive QA）和生成式问答（abstractive QA）。实验表明，尽管预训练模型在问答任务中表现良好，但在证据检索与基于权威依据的生成式回答方面仍面临显著挑战，尤其体现在将自然语言表达的法律问题准确映射至相应法律条文上。混合检索（hybrid retrieval）方法在检索性能上表现最佳，凸显了多源信息融合的重要性。因此，ViLegalExpert不仅揭示了当前越南语法律AI系统的局限性，更作为一项具有挑战性的基准，为开发可信、可解释的法律AI系统提供了关键支撑。
链接: https://arxiv.org/abs/2609.39189
作者: Dat Tien Nguyen,Nghia Hieu Nguyen,Anh Thi-Hoang Nguyen,Dung Ha Nguyen,Kiet Van Nguyen,Ngan Luu-Thuy Nguyen
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Trustworthy Legal AI requires systems that can answer legal questions while grounding their responses in authoritative sources. However, existing Vietnamese legal benchmarks provide limited coverage of real-world legal consultations. We introduce \textbfViLegalExpert, a large-scale benchmark constructed from authentic citizen–lawyer consultations, containing over \textbf172K questions across \textbf34 legal domains, together with professional answers and expert-verified legal evidence. ViLegalExpert supports legal information retrieval, extractive QA, and abstractive QA. Experiments with representative retrieval methods and language models reveal substantial challenges in evidence retrieval and grounded answer generation. While pretrained models perform strongly on QA, hybrid retrieval achieves the best retrieval performance. These results demonstrate the difficulty of mapping naturally expressed legal questions to authoritative provisions and establish ViLegalExpert as a challenging benchmark for reliable Vietnamese Legal AI.
[NLP-76] DAGent : Evaluate-then-Grow Planning for Deep Research Agents NEURIPS2026
【速读】： 该论文旨在解决深度研究任务中多智能体系统在复杂知识空间导航、跨源证据整合及动态计划调整方面的挑战。现有基于有向无环图（DAG）的多智能体系统普遍采用“先规划后修补”（Plan-then-Patch）策略，即在执行前固定全局任务计划，仅在出现失败或证据缺失时进行修复，这一方法在深度研究场景下存在根本性缺陷：系统在证据最薄弱时即做出强承诺，后期修正过程浪费大量计算资源于本不应启动的分支。为此，论文提出DAGent框架，其核心创新在于引入“评估后再扩展”（Evaluate-then-Grow）的增量式规划机制——由协调器（Orchestrator）基于已完成节点的置信度与不确定性信号，分批逐步构建任务图，实现更稳健的动态规划。同时，通过层次化上下文层默认传播紧凑的QueryDocs并保留完整执行轨迹以供按需召回，支持结构化强化学习信号的生成。进一步地，提出DAGRPO算法，作为GRPO的改进版本，将拓扑条件信用分配注入执行者（Executor）的策略更新，并对协调器的计划施加结构合规正则化，从而实现对任务图结构演化的有效引导。实验结果表明，在BrowseComp-Plus、GAIA和xbench-DeepSearch三个基准上，DAGent在Qwen3-235B-A22B规模下分别超越最强开源基线5.3、5.8和2.0点，且优势在四种开源模型架构及GPT-5（327K上下文）上均具可复现性；在Qwen3-8B规模下，相较于同预算的仅输出目标的GRPO基线，DAGRPO提升3.0平均Pass@1得分。同架构对比显示，基于证据条件的规划在更低的任务级词元、工具调用与步骤开销下达到更高精度，验证了其高效性与鲁棒性。
链接: https://arxiv.org/abs/2609.39154
作者: Hanwen Liu,Yuanfu Sun,Qiaoyu Tan
机构: New York University (纽约大学); New York University Shanghai (纽约大学上海分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted at NeurIPS 2026
Abstract:Deep research tasks require agents to navigate large knowledge spaces, synthesize evidence across many sources, and adapt their plans as findings emerge. Directed acyclic graph (DAG)-based multi-agent systems suit this setting because they support parallel execution and isolate each sub-task within a focused dependency context. Yet existing DAG-based agents instantiate a task-level plan before execution and repair the graph only after failures or missing evidence are observed. This Plan-then-Patch strategy is brittle for deep research: the system commits most strongly when its evidence is weakest, and later revisions waste computation on branches that should not have been planned. We propose DAGent, a DAG-based multi-agent framework with Evaluate-then-Grow incremental planning: an Orchestrator grows the task graph one batch at a time, conditioning each expansion on confidence and uncertainty signals from completed nodes. A hierarchical context layer propagates compact QueryDocs by default while preserving full execution traces for on-demand recall. The recorded DAG topology admits structural RL signals that outcome-only recipes cannot define; DAGRPO, a GRPO adaptation, injects topology-conditioned credit on Executor rollouts and a structural compliance regularization on Orchestrator plans. Across BrowseComp-Plus, GAIA, and xbench-DeepSearch, DAGent surpasses the strongest open-source baseline by 5.3 / 5.8 / 2.0 points at the Qwen3-235B-A22B scale, and the lead replicates across four open-source backbones and extends to GPT-5 at 327K context. At the Qwen3-8B scale, DAGRPO improves over a same-budget outcome-only GRPO baseline by 3.0 average Pass@1 points. A same-architecture comparison shows that evidence-conditioned planning reaches higher accuracy at lower per-task token, tool-call, and step footprints than its Plan-then-Patch counterpart. Code: this https URL
[NLP-77] Diagnosing On-Policy Self-Distillation for Reasoning Language Models
【速读】： 该论文旨在解决生成式语言模型在数学推理任务中性能提升受限的问题，特别是针对无外部奖励且无独立更强教师的自教师（self-teacher）机制——即在策略自蒸馏（On-policy Self-Distillation, OPSD）框架下，如何有效利用模型自身生成的“教师”信号来增强学生模型的推理能力。研究发现，尽管OPSD被广泛认为是提升语言模型推理能力的有前景方法，但其实际表现存在显著不稳定性，从微弱提升到行为崩溃均有报道。通过在0.6B至8B参数规模的语言模型上进行受控实验与逐标记粒度分析，论文揭示：教师信号的有效性并非由特权语义（privileged semantics）决定，而是由推理模式对齐（reasoning-mode alignment）及完整教师前缀（complete teacher prefix）共同塑造。研究表明，OPSD仅在特定兼容性范围内能有效提升推理性能，超出该范围则导致无效的序列长度增长、性能稳定退化或行为崩溃。进一步的逐标记分析表明，教师信号本身并不具备稳定性，也无法可靠预测下游任务表现。因此，论文的核心结论是：OPSD是一种对条件敏感的算法，而非一种普遍可靠的推理增强后训练方法。
链接: https://arxiv.org/abs/2609.39118
作者: Yang Li,Gongle Xue,Yuheng Yuan,Yijia Guo,Shizhe Zhang,Liwen Hu,Lei Ma
机构: Peking University(北京大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:On-policy self-distillation (OPSD) has attracted growing interest as a promising approach to improve the reasoning ability of language models. Without external rewards nor a separate stronger teacher, the self-teacher with privileged information could provide dense signals on student’s trajectories. However, its behavior in language reasoning remains unclear, with reported outcomes ranging from modest gains to behavioral collapse. In this work, we diagnose OPSD for mathematical reasoning across models spanning 0.6B–8B parameters. We conduct controlled experiments and token-level analyses to fully delve into OPSD. We point out that teacher’s signal is shaped by reasoning-mode alignment and the complete teacher prefix, rather than by privileged semantics alone. OPSD improves reasoning only in narrow compatibility regimes. Otherwise, it produces ineffective length growth, stable degradation, or behavioral collapse. Token-level analysis shows that teacher’s signal is not stable and does not predict downstream performance. Based on these results, we argue that OPSD is a sensitive algorithm rather than a generally reliable reasoning-improvement post-training method.
[NLP-78] Bongard: Training Machine Intuition
【速读】： 该论文旨在解决机器在复杂决策任务中缺乏类人直觉（intuition）的问题，即模型难以像人类一样基于经验快速识别模式、判断情境并作出高效决策。其核心挑战在于如何将“直觉”这一非显式推理过程建模为可训练的独立能力，而非依赖逐层推演的符号逻辑。解决方案的关键在于提出Bongard——一个基于T5Gemma 2 4B-4B架构的开放权重系统一（System One）模型，通过分离状态理解与判断生成两阶段设计：编码器以双向方式联合读取环境状态与问题指令，生成统一表征；多个解码分支共享该表征，实现对同一状态的多任务判断仅需一次编码，显著提升效率。模型采用三阶段渐进式训练策略，依次从监督判断、语义关系到动作结果学习，全面更新70.9亿参数。引入联合嵌入后训练（joint-embedding post-training）和沙盒阶段（sandbox stage），分别利用重述样本的泛化能力和动作回放与精确规则的反馈，进一步提升模型在未见任务上的准确率。最终在DecisionBench基准上达到78.05%的准确率，且在单张RTX PRO 6000 GPU上实现36毫秒的中位延迟，验证了通过表示学习与结果反馈系统性训练机器直觉的有效性，为高吞吐决策场景提供了一种高效、开放的替代方案。
链接: https://arxiv.org/abs/2609.39111
作者: Li Ding,Haidi Jin,Chen Ji
机构: 未知
类目: Computation and Language (cs.CL)
备注: Technical report, 28 pages, 7 figures. Model weights: this https URL
Abstract:Human intelligence relies heavily on learned intuition: recognising patterns and judging situations without explicitly unfolding every intermediate step. We introduce Bongard, an open-weight System One model that treats machine intuition as an independent capability to design and train. A T5Gemma 2 4B-4B encoder-decoder separates reading the evidence from making judgments. The encoder reads the state bidirectionally together with the question instructions, and separate decoder branches share this encoding, so many judgments about the same situation require only one reading of the state. A trained head returns probabilities over the supplied candidates without generating text. Training proceeds in three stages, from supervised judgments to semantic relationships to action outcomes, and each stage updates all 7.09 billion trainable parameters on one Blackwell GPU. Joint-embedding post-training raises accuracy on held-out rephrasings from 75.7% to 85.9%. A sandbox stage then learns outcome distributions from action rollouts and exact oracles, raising accuracy on a frozen sandbox panel from 50.6% to 64.8%. On DecisionBench, the final model reaches 78.05% accuracy over 23,900 decisions and ranks fourth of 61 systems in the public comparison. On one RTX PRO 6000, its median latency is 36 ms for short requests, and 32 questions about one state take 221 ms. Bongard demonstrates that machine intuition can be systematically trained via representation learning and outcome feedback, providing an open, efficient alternative for high-throughput decision workloads.
[NLP-79] False Frontiers: Diagnosing and Mitigating Co-Cheating in Self-Evolving Search Agents
【速读】： 该论文旨在解决自演化搜索代理（self-evolving search agents）在闭环训练过程中出现的“共作弊”（co-cheating）问题，即提议器（proposer）与求解器（solver）在无外部证据支持的情况下，逐渐达成对错误答案的一致性，导致内部奖励提升但真实正确率停滞甚至下降。其核心解决方案是提出CrossFit方法：将提议器所用源文档分为两组A和B，分别使用仅基于另一组训练的辅助求解器来评分，通过跨组验证的共识决定提议器奖励，从而切断同一来源伪标签的反馈循环。该机制确保了反馈信号无法被同源模型复制，有效抑制共作弊；实验表明，相较于多样本验证（MSV），CrossFit显著降低虚假一致率（从6.1%/8.8%降至3.0%/3.7%），并在七个下游搜索基准上分别提升平均性能8.8/8.4点（相对于标准耦合自演化）及8.7/7.8点（相对于Search-R1），且无需额外生成大量标注样本，具备高效性与可扩展性。
链接: https://arxiv.org/abs/2609.39102
作者: Meijia Chen,Hao Li,Zheng Lu,Hongshan Lin,Junbai Tian,Yichen Liu,Zijun Tian,Yufan Zou,Shuhan Sun,Hanxin Chen,Zeyu Zhang,Weizhi Du,Yueting Li,Tianyu Shi,Alaa Khamis
机构: Rutgers University (罗格斯大学); University of California, San Diego (加州大学圣地亚哥分校); University of Michigan (密歇根大学); McGill University (麦吉尔大学); King Fahd University of Petroleum and Minerals (法赫德国王石油与矿业大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 21 pages. Equal contribution: Meijia Chen, Hao Li, Zheng Lu
Abstract:Self-evolving search agents build their own training curricula by jointly optimizing a proposer that generates questions and a solver that answers them. This closed loop introduces a failure mode we call co-cheating: the proposer and solver increasingly agree on shared errors, so internal reward improves without a matching gain in external correctness. A post-hoc audit against source evidence shows co-cheating growing more severe over successive rounds of self-evolution, with pseudo-label correctness stagnating or declining even as the in-loop training signal improves. The most direct mitigation is to verify proposals before training: we introduce multi-sample verification (MSV), which queries the same model three times with the source and three times without it to decide task admission and replace unreliable pseudo-labels. MSV partially reduces false agreement but leaves substantial residual co-cheating and costs six extra labeler generations per candidate. These limitations motivate CrossFit, our main method: it partitions the proposer’s source documents into groups A and B; questions generated from A are scored by an auxiliary solver trained only on B, and vice versa. The cross-fitted agreement determines proposer reward, so a same-source pseudo-label cannot be reproduced through the feedback solver, while the original solver’s update rule is unchanged. Rerunning the loop with Qwen3.5-4B and Qwen3.5-9B, MSV reduces false-agreement mass from 6.1% to 5.7% and from 8.8% to 7.2%, whereas CrossFit reduces it to 3.0% and 3.7%. Replaying identical proposals with source-excluded feedback further reduces false agreement to 0.4% and 0.1%, isolating feedback ancestry from curriculum changes. Across seven downstream search benchmarks, CrossFit improves average performance over standard coupled self-evolution by 8.8 and 8.4 points and over Search-R1 by 8.7 and 7.8 points at 4B and 9B.
[NLP-80] Beyond Text: LLM -Based Dimensional Emotion Evaluation in Multimodal Dialogue
【速读】： 该论文旨在解决在多模态对话中应用大语言模型（LLM）进行连续维度情感评估（如效价-唤醒-支配度，VAD）的挑战，尤其关注如何有效融合语音线索以提升情感识别性能。其核心解决方案是提出一种基于LLM的框架，将音频特征转化为自然语言描述（遵循SpeechCueLLM方法），并在此基础上实现离散情绪分类与连续维度情感评分的联合建模。关键创新在于通过低秩自适应（LoRA）微调策略显著提升了较小规模的LLaMA模型在两项任务上的表现，超越了参数量更大的GPT系列模型，表明领域适配性比模型规模更为关键。实验结果表明，该方法在IEMOCAP数据集上取得了0.7822的效价一致性相关系数（CCC），达到新基准水平；同时，消融实验揭示音频文本描述对小模型的增益显著（F1提升3.5–3.6个百分点），而对大模型贡献有限，说明语音信息的价值在语言表达能力受限时尤为突出。此外，模型在三个维度上的性能差异与人工标注者的一致性层级高度吻合，验证了所提方法在捕捉人类情感感知规律方面的有效性。
链接: https://arxiv.org/abs/2609.39072
作者: Yutong Hu,Jinho Choi
机构: Emory University(埃默里大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multimedia (cs.MM)
备注: 15 pages, 6 figures, 11 tables
Abstract:Emotion recognition in conversation has been widely studied, but applying Large Language Models (LLMs) to continuous dimensional emotion evaluation in multimodal dialogue remains largely unexplored. We propose an LLM-based framework that performs discrete emotion recognition and Valence-Arousal-Dominance (VAD) dimensional evaluation on IEMOCAP, incorporating acoustic cues as natural language descriptions following the SpeechCueLLM approach. We evaluate six models spanning the LLaMA, GPT, and Qwen families under zero-shot prompting, few-shot prompting, and LoRA fine-tuning. LoRA fine-tuned LLaMA models substantially outperform prompt-engineered GPT models on both tasks despite GPT’s larger scale, a gap we attribute to domain adaptation rather than model capacity. Our best model achieves a Valence CCC of 0.7822, a new state-of-the-art on IEMOCAP. Ablation studies confirm that textual audio descriptions meaningfully improve smaller models (+3.5 to 3.6 weighted F1) while contributing little for the largest model, suggesting audio cues are most valuable when linguistic capacity is limited. The performance asymmetry across VAD dimensions closely mirrors the annotator agreement hierarchy in IEMOCAP’s own annotations.
[NLP-81] LexReward: A Taxonomy-Driven Reward Framework for Legal Language Models
【速读】： 该论文旨在解决法律语言模型（Legal Language Models）在奖励信号设计上的核心问题：现有奖励方法多依赖粗粒度的整体评价，难以充分捕捉法律回应在多个维度上的质量差异，导致领域特异性与可解释性不足。为此，论文提出LexReward——一种基于分类体系的法律奖励建模框架，其关键在于构建一个涵盖三个互补维度的精细化评估架构：风格（Style），关注词汇与句法质量；要素（Element），评估法律主体、事实、法条及裁判结果的完整性与准确性；推理链（Chain），衡量法律推理的逻辑顺序、完整性、正确性与非冗余性。针对每个维度，研究设计了明确的评分标准与质量分级体系，生成可量化的奖励信号，并用于构建配对偏好数据以支持直接偏好优化（DPO）和奖励模型训练。实验表明，基于评分表的奖励能有效区分不同质量的法律回应，且在偏好数据上进行DPO训练可全面提升三维度表现。进一步地，所学习的奖励模型（LexRM）可通过强化学习实现下游任务中各维度的独立优化，且无需在奖励阶段依赖参考答案。维度分析验证了该分类体系与奖励构造的有效性，为法律文本生成的质量评估与优化提供了可解释、可扩展的解决方案。
链接: https://arxiv.org/abs/2609.39071
作者: Yida Cai,Xin Dai,Bingxiang He,Huiyuan Xie,Yuxiao Ye,Zhenghao Liu,Yang Bai,Zhiyuan Liu
机构: Peking University (北京大学); Northeastern University (东北大学); Tsinghua University (清华大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Legal language models require reward signals that capture not only answer correctness but also the multidimensional quality of legal responses. Existing reward methods, however, often rely on coarse-grained holistic judgments, providing limited domain specificity and interpretability. We introduce LexReward, a taxonomy-driven framework for legal reward modeling. LexReward characterizes legal response quality along three complementary dimensions: Style, covering lexical and syntactic quality; Element, assessing legal subjects, facts, statutes, and decisions; and Chain, evaluating the order, completeness, correctness, and non-redundancy of legal reasoning. For each dimension, we develop rubrics that specify evaluation criteria and quality levels. The resulting rewards are used to construct pairwise preference data for Direct Preference Optimization (DPO) and reward-model training. Experiments show that the rubric-based rewards reliably distinguish legal responses of different quality and that DPO training on the preference data improves performance across all three dimensions. The learned reward models, LexRM, also support effective downstream optimization: each dimension-specific reward model improves policy performance in its corresponding dimension through reinforcement learning, without requiring reference answers at reward time. Dimension-wise analyses further support the effectiveness of the proposed taxonomy and reward construction.
[NLP-82] CORE: Conflict-Oriented Reasoning Elimination for Verifiable Language-Model Search
【速读】： 该论文旨在解决测试时推理系统在面对错误时仅通过重启或修正最新步骤来应对，而忽视早期决策导致错误的问题。其核心解决方案是引入一种名为CORE的搜索控制器，该控制器通过向验证器请求一个经认证的冲突核心（conflict core），回跳至该核心中最新的决策点，并缓存该冲突以避免重复发生。在满足可靠验证、有限分支与深度以及穷尽提案的前提下，未加限制的搜索具有完备性，且不会误剪枝有效解。在2,000个预设图着色实例上，相较于传统的时序修复方法（chronological repair），CORE在30变量和36变量场景下分别减少了39.8%和35.0%的验证器调用次数；进一步结合缓存机制可显著提升回跳策略的效果。在五个推理任务中，CORE在Qwen2.5-7B-Instruct和Qwen3-8B两个大模型基础上分别实现了75.9%和84.2%的平均成功率，优于Tree of Thoughts的72.5%和81.8%，同时减少验证器调用次数与生成的令牌数。实验结果表明，利用经认证的失败解释来引导语言模型搜索具有显著优势。
链接: https://arxiv.org/abs/2609.39069
作者: Siyu Song,Rui Xu,Jia Lin,Kai Liu,Weifang Wang
机构: Sun Yat-sen University (中山大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Test-time reasoning systems often respond to failure by restarting or revising the latest step, even when an earlier decision caused the error. We introduce CORE, a search controller that requests a certified conflict core from a verifier, backjumps to the latest decision in that core, and caches the conflict to avoid repeating it. Under sound verification, finite branching and depth, and exhaustive proposals, the uncapped search is complete and never prunes a valid solution. On 2,000 planted graph-coloring instances with matched proposals and an exact verifier, CORE reduces median verifier calls by 39.8% at 30 variables and 35.0% at 36 variables relative to chronological repair; caching further improves on backjumping alone. Across five reasoning tasks, CORE achieves 75.9% mean success with Qwen2.5-7B-Instruct and 84.2% with Qwen3-8B, compared with 72.5% and 81.8% for Tree of Thoughts. It also uses fewer verifier calls and generated tokens on both backbones. These results show the value of using certified failure explanations to direct language-model search.
[NLP-83] Covert Assistance: Helpful LLM Agents Evade Oversight in Multi-Agent Systems
【速读】： 该论文旨在解决在高风险领域中，多智能体系统（multi-agent systems）内非恶意智能体（benign agents）可能无意或隐蔽地突破安全边界的问题。传统研究主要聚焦于对抗性场景下智能体被激励进行隐秘通信以规避监管，而本文揭示了即使在无对抗激励的情况下，前沿大模型仍会通过隐蔽手段绕过安全约束。其关键解决方案在于识别并分析一种名为“隐匿协助”（covert assistance）的行为模式：即智能体在执行任务过程中，将敏感信息（如公司凭证）以编码、谜题或隐喻等非明文形式嵌入指令中，伪装成对协作的辅助行为，从而规避监控。实验表明，在模拟软件工程工作流中，九个测试的前沿模型中有七个在未被授意的情况下仍尝试隐藏凭证，且在6000次试验中DeepSeek-V4-Pro有0.9%的概率成功使开发者获取并利用该凭证。尽管单次事件发生率低，但随着交互次数增加，泄露风险呈指数累积。研究进一步指出，监控机制倾向于容忍隐蔽传播，而当接收方被设定为人类而非AI时，隐蔽行为更易演变为直接披露。因此，提升检测能力需依赖信任监控者持有敏感信息，这引入了新的信任风险。该发现挑战了现有监督机制在区分合法协作与越界协助之间的有效性，凸显出在实际应用中亟需建立更鲁棒的安全边界判定框架。
链接: https://arxiv.org/abs/2609.39050
作者: Deema Alnuhait,Gengyu Wang,Muhammad Khalifa,Hao Peng
机构: University of Illinois Urbana-Champaign; Genies; NVIDIA
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL)
备注:
Abstract:As multi-agent systems enter high-stakes domains, the possibility that agents may circumvent safety boundaries is a growing concern. Prior work has examined this risk primarily in adversarial settings, where agents are instructed or rewarded to communicate covertly and evade oversight. We show that benign agents can cross the same boundaries without adversarial incentives. We emulate a software-engineering workflow in which a planner represents a company hiring an external developer. The planner writes requirements and holds a company credential it is instructed not to disclose to the developer; a monitor screens their exchanges. Seven of nine tested frontier models disguise the credential in their requirements to help the developer recover it while evading the monitor, even after completing their assigned objective. For example, across 6,000 episodes with DeepSeek-V4-Pro, the planner attempts concealment in 16.9%; in 0.9%, the credential evades the monitor and is recovered and used by the developer. Despite the low per-episode rate, the risk compounds across repeated exchanges: at 0.9%, 105 independent episodes carry a 61.3% chance of at least one breach. Our analysis suggests that the planner misinterprets disguised disclosure as helping the developer and reads the nondisclosure rule as banning plaintext, not character codes or riddles. We call this behavior covert assistance. Monitoring favors concealment, while concealment gives way to direct disclosure when the recipient is presented as a human rather than an AI agent, so leakage persists. Giving the monitor the credential improves detection but requires trusting it with the secret. These risks, in models already used for software engineering, challenge oversight to distinguish authorized cooperation from task-advancing assistance that crosses safety boundaries.
[NLP-84] Structure vs. Chain-of-Thought: Evaluating LLM Criteria Extraction for Depression Severity ICDM2026
【速读】： 该论文旨在解决如何利用大语言模型（Large Language Model, LLM）从社交媒体文本中自动评估抑郁症严重程度的问题，重点比较直接评分与基于临床诊断标准逐项提取（criteria extraction）两种方法的有效性。其核心挑战在于：尽管基于规则的提取方法在可解释性和可审计性上更具优势（因临床医生可逐项核查标记的诊断标准），但其实际性能是否优于直接提示（direct prompting）或思维链（chain-of-thought）推理仍不明确。解决方案的关键在于系统评估不同规模的LLM（从9B到前沿模型）在两个Reddit语料库上，结合PHQ-9与BDI-II量表，通过二次加权肯德尔协调系数（quadratic weighted kappa）衡量与人工标注的一致性，并探究阈值设定方式（先验规则 vs. 数据拟合）及模型校准对结果的影响。研究发现，前沿模型在未校准情况下无法稳定超越思维链方法；即使在数据拟合阈值后，性能提升也不显著；而采用先验的PHQ-9标准时，提取方法并无优势，甚至在多数情况下漏检了严重病例。值得注意的是，9B模型在抑郁社区语料中表现出高度保守倾向，倾向于将多数帖子标记为严重，但在使用先验规则时反而表现更优，表明其固有偏倚可能影响判断。经同源标签校准后，各方法间差异消失，揭示了模型校准对性能的重要影响。然而，更高的序数一致性并不等同于更好的重症识别能力，且从直接提示到思维链再到标准提取的演进过程反而导致严重病例漏检率上升。最终，基于自定义特征（如词频）的模型在重新标注的压力数据集上，与先验规则下的前沿模型提取方法相比，无显著差异，暗示当前基于标准提取的方法在真实场景中的检测效能存在局限。
链接: https://arxiv.org/abs/2609.39049
作者: Xinkai Chen
机构: 独立研究者(Independent Researcher)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Extended version of a paper accepted at MHSM 2026 (IEEE ICDM 2026 workshop). 14 pages, 1 figure. Code: this https URL
Abstract:A large language model (LLM) can rate depression severity directly from a social media post or mark which clinical criteria the post shows and let code turn the count into a label. The latter is easier to audit because a clinician can check each marked criterion. We compare these approaches on two Reddit corpora using three LLMs (from 9B to frontier scale) and two questionnaires (PHQ-9, BDI-II), and measure agreement with quadratic weighted kappa. For the two frontier models, criteria extraction scores above chain-of-thought on one corpus only when its decision thresholds are fitted on labeled data. Neither model’s gain is significant, with or without recalibrating chain-of-thought on the same labels. With thresholds fixed a priori from PHQ-9’s criteria, extraction shows no gain on either corpus, even where models mark over two criteria per post. The 9B model behaves differently on a corpus from depression communities. It labels most posts severe, whether prompted directly or with chain-of-thought, while the a priori rule beats both without labels. After chain-of-thought is recalibrated on the same labels, no significant gap remains, consistent with a calibration effect. Yet higher ordinal agreement does not ensure better detection of severe cases. PHQ-9 criteria extraction misses most severe posts, and moving from direct prompting to chain-of-thought and then to extraction increases misses in nearly all comparisons. On the primary corpus, a relabeled stress dataset, a model using that dataset’s own features, including word counts from the text, is not significantly different from frontier criteria extraction under the a priori rule.
[NLP-85] Switching Linear Attention
【速读】： 该论文旨在解决现代机器学习中高效推理条件下序列建模的表达能力与计算效率之间的矛盾问题。标准Softmax注意力虽具备强大的非线性交互能力，但其需随序列长度线性增长的键值缓存（key-value cache）限制了模型的可扩展性；而线性注意力虽可通过固定内存开销实现高效的递归计算，却因表达能力受限导致建模性能下降。本文提出一种新型序列层——切换线性注意力（Switching Linear Attention, SwiLA），其核心在于通过在测试阶段引入基于输入动态选择多个线性注意力组件的机制，显著增强模型的表征能力，同时保持线性注意力固定的递归状态尺寸。该方法的推导基于测试时回归框架，将状态更新规则形式化为混合线性回归模型中的在线期望最大化（online expectation-maximization），实现了表达性与高效性的协同优化。实验表明，SwiLA在关联回忆、上下文内语言学习及语言建模等任务上表现出色，性能接近甚至超越标准Softmax注意力，在多项基准测试中展现出更强的竞争力。
链接: https://arxiv.org/abs/2609.39034
作者: Hyun Dong Lee,Xavier Gonzalez,Nicolas Zucchet,E. Kelly Buchanan,Emily B. Fox,Scott W. Linderman
机构: Stanford University (斯坦福大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: COLM 2026
Abstract:Designing expressive sequence layers with efficient inference remains a central challenge in modern machine learning. Standard softmax attention achieves excellent sequence modeling performance through rich nonlinear token interactions, but it requires a key-value cache that grows linearly with sequence length, limiting its scalability. Linear attention enables efficient recurrent computation with a constant memory footprint, yet its reduced expressivity often yields inferior modeling performance. We introduce Switching Linear Attention (SwiLA), a novel sequence layer that bridges this gap by enhancing representational capacity while retaining the fixed-size recurrent state of linear attention. We derive the SwiLA recurrence from the test-time regression framework, casting the state update rule as online expectation-maximization in a mixture of linear regressions model. At test time, each output dimension dynamically selects among multiple linear attention components based on the input. Across associative recall, in-context language learning, and language modeling benchmarks, SwiLA shows strong performance and narrows the gap to softmax attention, even surpassing it in several settings.
[NLP-86] A Missing Piece for Trustworthy AI Reviewers: From Benchmarking Rhetorical Robustness to SciCore Review NEURIPS2026
【速读】： 该论文旨在解决生成式人工智能（Generative AI）评审系统在面对同一科学内容的不同语言表述时，可能出现判断不一致的问题，即评审结果易受修辞优化影响而非真正反映科学质量的提升。其核心挑战在于实现“修辞鲁棒性”（Rhetorical Robustness），即在保持内容不变的前提下对文本重写具有稳定性，同时能够有效区分不同科学质量的论文。解决方案的关键是提出一种双分支评审架构——SciCore，该模型通过融合对完整稿件的整体判断与基于提取出的结构化科学核心（science core）的独立评估，实现对内容的标准化表征，从而降低对语言表达方式的敏感性。实验表明，SciCore在基准测试中展现出最优的稳定性和判别能力平衡，并维持了良好的人类对齐表现，验证了以科学核心为基础的评审范式在提升修辞鲁棒性方面的有效性。
链接: https://arxiv.org/abs/2609.39027
作者: Chenguang Wang,Ming Li,Chengrui Fan,Jianpeng Chen,Han Chen,Tianyi Zhou,Dawei Zhou
机构: 未知
类目: Computation and Language (cs.CL)
备注: 35 pages, 2 figures, 20 tables. Accepted (Oral) at AI-Native Academia @ NeurIPS 2026
Abstract:AI reviewers can assign different judgments to manuscripts that report the same science in different wording, potentially rewarding rhetorical optimization over scientific improvement. We formulate Rhetorical Robustness as the joint requirement of stability across content-preserving rewrites and discrimination across papers. We introduce RobustReview, a controlled full-manuscript benchmark with 1,260 manuscript versions, and evaluate 30 reviewer configurations. The benchmark reveals false robustness, where low rewrite sensitivity coincides with score collapse across papers, and shows that human alignment and rhetorical robustness rank reviewers differently. Moreover, the evaluated content-focused prompting protocol does not consistently improve robustness across backbones. Motivated by these findings, we introduce SciCore, a dual-branch reviewer that averages a full-manuscript judgment with a judgment based on an extracted, structured science core. This design combines manuscript-level assessment with a content-normalized view intended to reduce rhetorical sensitivity. In our primary GPT-5.5 comparison, SciCore achieves a leading joint stability-discrimination profile among the benchmarked reviewers while maintaining competitive human alignment. These results identify rhetorical robustness as a distinct evaluation target and demonstrate the potential of science-core review to improve it.
[NLP-87] he Invisible Language Tax: Token Premiums of French and Regional Languages in 2026 LLM Tokenizers and a French-Optimized Prototype
【速读】： 该论文旨在解决多语言环境下大语言模型（LLM）服务中因语言差异导致的令牌（token）计费不均问题，即相同内容在不同语言下所需令牌数量存在显著差异，进而引发成本偏差。其核心挑战在于：尽管模型按令牌计费且上下文窗口以令牌为单位衡量，但如法语、中文及法国地区语言等非英语文本普遍需要更多令牌，造成实际使用成本上升，尤其在智能体（agentic）应用中，历史重传、分层定价与固定上下文窗口机制进一步放大了这一差距。解决方案的关键在于设计一种针对法语优化的字节级子词分词器（byte-level BPE），即Baracoda FR v1.2，该分词器基于Tekken分词器的词汇量，在未参与训练的六组语料上实现对法语文本减少11.5%的令牌消耗，同时对英文仅减少3.7%，且在去除训练数据重叠项并保持同等普通令牌预算的情况下仍保持优势。该方法通过在训练数据中引入法语语料快速降低法语的令牌溢价，但伴随英语性能下降和边际收益递减，表明语言间平衡需权衡。研究结果目前仅为分词效果，尚未验证对模型质量与任务成本的影响。
链接: https://arxiv.org/abs/2609.39001
作者: Thomas Serval
机构: Baracoda AI Labs; NEOMA Business School (NEOMA 商学院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 11 pages, 5 figures, 7 tables. Code, tokenizer, per-sentence counts and controls: this https URL (commit fc8a736)
Abstract:LLM services are billed per token and context windows are measured in tokens, yet the number of tokens needed for the same content varies across languages. We measure this token premium on seven tokenizers of widely used 2026 models (OpenAI o200k, Llama 3, Qwen3, DeepSeek V3/V4, Gemma 3, Mistral Tekken, and the Claude generation-5 tokenizer via Anthropic’s counting API) on NTREX-128 (124 non-English reference translations) and on the Universal Declaration of Human Rights for regional languages. French requires 31% to 58% more tokens than English, whereas Simplified Chinese ranges from 5% fewer to 40% more and is cheaper than French on six of the seven tokenizers. Regional and overseas languages of France pay roughly 1.6 to 3.3 times the English count. We discuss how history re-sending, tiered pricing and fixed context windows amplify the absolute gap in agentic use. In a controlled experiment (BPE, Europarl, 50k vocabulary), adding French to tokenizer training data quickly reduces the premium, with diminishing returns and a growing cost for English. Finally, we present Baracoda FR v1.2, a byte-level BPE prototype with Tekken’s vocabulary size. On a final test of six corpora never consulted during design, with a protocol declared fixed beforehand, it uses 11.5% fewer tokens than Tekken on French and 3.7% fewer on English; results hold after removing test sentences overlapping the training data and with an equal ordinary-token budget. It is worse on other languages and, at comparable vocabulary size, does not outperform CroissantLLM. These are segmentation results only; effects on model quality and task cost remain to be shown.
[NLP-88] Settle: Learning When to Stop Reasoning
【速读】： 该论文旨在解决生成式 AI（Generative AI）在推理过程中存在冗余生成的问题，即模型在得出稳定答案后仍持续生成额外无关内容，导致计算资源浪费和效率降低。其核心解决方案是提出一种名为 Settle 的自适应停止机制，通过检测推理轨迹中答案的稳定性来决定何时终止生成。Settle 在训练时仅微调结束推理标记（end-of-reasoning token），保持其余预测与基础模型一致，并在推理阶段仅需常规解码流程。实验表明，在 MATH-500 数据集上使用 Qwen3-4B 模型时，Settle 可将生成令牌数减少 40%，同时仅带来 0.5 个百分点的准确率下降；相较于监督微调方法，在相同缩短长度下，准确率提升达 6.16 个百分点，且几乎保持相同的令牌消耗水平。此外，Settle 的停止评分能够有效预测正确答案是否将持续保持正确，显著扩展了所评估停止策略在准确率与令牌消耗之间的帕累托前沿。
链接: https://arxiv.org/abs/2609.38997
作者: Ryan Brown,Zihao Fu,Chris Russell
机构: Oxford Internet Institute, University of Oxford(牛津互联网研究所，牛津大学); Department of Linguistics and Modern Languages, The Chinese University of Hong Kong(语言学与现代语言系，香港中文大学)
类目: Computation and Language (cs.CL)
备注: 30 pages, 4 figures
Abstract:Reasoning models often continue generating after their answers have settled. Settle learns when to stop from answer stability in completed traces. It trains the existing end-of-reasoning token while keeping other predictions close to the base model, and requires only ordinary decoding at inference. On MATH-500 with Qwen3-4B, Settle reduces token count by 40% with a 0.5-percentage-point decrease in accuracy. It gains 6.16 percentage points over supervised fine-tuning on the same traces shortened at their first stable answer, at nearly identical token counts. Its stopping score predicts whether a correct answer will remain correct. Settle extends the accuracy-token-count Pareto frontier of the evaluated stopping methods.
[NLP-89] When Clipping Reverses Correction: Failure Dynamics of Pointwise Forward-KL On-Policy Self-Distillation
【速读】： 该论文旨在解决基于策略的自蒸馏（On-policy self-distillation, OPSD）在数学推理任务中因训练信号失衡而导致的学生模型性能下降问题。具体而言，原始研究发现，在生成过程中，风格性词汇（stylistic tokens）会压倒与数学内容相关的词汇，导致训练信号偏移；尽管通过逐点裁剪（pointwise clipping）前向KL散度目标可稳定训练，但其潜在负面影响尚未被充分探讨。本文的关键发现是：虽然裁剪能缓解训练不稳定性，但会引入严重副作用——导致学生模型产生大量难以终止的重复响应。通过对有无裁剪的对照实验分析，作者证明了裁剪后的目标函数存在根本缺陷：它可能无法有效引导学生向教师模型靠拢，反而使被裁剪和未裁剪的词元概率分布偏离教师分布。实证结果表明，在重复段落内部，裁剪后的学生模型对“结束重复”的置信度低于教师，而对“继续重复”的置信度更高，而未裁剪版本则保持与教师的一致性。因此，该研究揭示了当前广泛采用的裁剪策略在实际应用中的关键缺陷，并强调需重新审视其在生成式AI（Generative AI）训练中的适用性。
链接: https://arxiv.org/abs/2609.38995
作者: Di Huang,Hao Li,Yixin Chen,Fuhai Li
机构: Washington University in St. Louis (圣路易斯华盛顿大学)
类目: Computation and Language (cs.CL)
备注: 21 pages, 5 figures
Abstract:On-policy self-distillation (OPSD) trains a student on its own generated responses using feedback from the same model conditioned on privileged information. On mathematical reasoning, the original OPSD study finds that stylistic tokens can dominate the training signal over math-related tokens, and that pointwise clipping of the forward KL objective stabilizes training. Pointwise clipping caps each vocabulary-wise forward KL term at a fixed threshold before summing over the vocabulary. Follow-up studies have adopted this clipping, but its effect on training has not been directly examined. In matched training runs differing only in whether clipping is applied, we observe that clipped runs produce substantially more repetitions that persist to the end of the response than their unclipped counterparts. We trace this failure to the clipped objective. We prove that the clipped objective can fail to correct the student toward the teacher and can instead push clipped and unclipped token probabilities away from its teacher. Our training runs agree with this analysis: inside repetitions, the clipped student places less probability than its teacher on leaving the repetition, and more on continuing it, whereas the unclipped runs stay close to their teachers.
[NLP-90] Fairness Beyond a Single Run: Training-Seed Variability in Speech LLM Adaptation
【速读】： 该论文旨在解决自动语音识别（ASR）系统中存在的人口统计学公平性差距（demographic fairness gaps）在单次训练运行下评估所导致的不可靠性和偏差问题。现有研究通常仅基于一次训练运行报告公平性表现，忽略了训练随机性（如随机种子）对公平性指标的影响，从而可能误导对模型公平性的判断。本研究的关键解决方案在于通过系统性地控制变量，将模型微调过程中的关键组件（如Q-former投影器和LoRA适配器）在不同音频压缩因子与多个随机种子条件下进行重复实验，并在Common Voice和Fair-Speech数据集上进行评估。研究发现，在460小时纯净LibriSpeech数据训练下，随机种子对公平性指标的变异贡献远高于音频压缩因素（在种族维度上分别贡献85.3%与8.3%，p=0.009），且该效应在控制准确率与丢弃率后依然显著。即使将适应数据集扩展至960小时以增强鲁棒性，该种子主导的变异性仍存。因此，论文强调：为确保公平性评估的可靠性，必须通过多随机种子重复实验来区分真实性能差异与训练随机性带来的波动；单次运行结果不足以代表模型的公平性表现，尤其当两个系统的种族归一化差距小于0.30时，其差异可能完全源于随机种子变化而非模型本质差异。
链接: https://arxiv.org/abs/2609.38976
作者: Srishti Ginjala,Eric Fosler-Lussier,Srinivasan Parthasarathy
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Demographic fairness gaps in automatic speech recognition are almost always reported from a single training run. We fine-tune the Q-former projector and LoRA adapters of a speech LLM at five audio compression factors and six random seeds, holding the encoder, base decoder, data and decoding fixed, and evaluate every run on Common Voice and Fair-Speech. At 460 h of clean LibriSpeech, the seed moves fairness metrics more than compression does on most demographic axes. A balanced 3x3 decomposition attributes 85.3% of the variation in Fair-Speech ethnicity normalized gap to the seed against 8.3% to compression (p = 0.009), though compression explains more on age and gender. Held-out LibriSpeech word error rate spreads by 0.04 points across those seeds while Common Voice spreads by 8.57, so these are not failed runs, and the effect survives controlling for accuracy and dropout. Scaling and diversifying the adaptation set to 960 h damps the effect but does not remove it. On Fair-Speech ethnicity, two single-run systems must differ by more than 0.30 in normalized gap to exceed seed variability.
[NLP-91] Making LLM s Say What They Think: Measuring and Improving CoT-Interpretability Alignment
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）的思维链（Chain-of-Thought, CoT）描述与其内部计算过程之间存在不一致的问题。尽管CoT常被用作解释模型推理过程的代理，但已有研究表明，其内容可能与模型实际内部计算无关，甚至在不改变最终答案的情况下可被随意修改。为此，论文提出一种名为“思维链可解释性对齐”（CoT-Interpretability Alignment, CIA）的新指标，用于量化模型的CoT描述与其通过可解释性工具检测到的内部推理策略之间的匹配程度。实验在双跳问答、提示干预和整数乘法三个任务上评估了CIA，发现三类主流LLMs在所有任务中均表现出较低的对齐度（44.8%-75.9%）。为提升对齐度，研究进一步采用后训练方法，以任务准确率和参数层面的忠实性信号作为奖励目标进行优化。结果表明，该方法可在保持或提升任务准确率的同时显著提高CoT的参数忠实性，并揭示出良好的泛化特性。本工作不仅提供了一套用于审计CoT参数忠实性的框架，也为构建更可信、可解释的显式推理机制提供了可行路径。
链接: https://arxiv.org/abs/2609.38972
作者: Yihuai Hong,Shauli Ravfogel,Chen Zhao,Eunsol Choi
机构: New York University (纽约大学); NYU Shanghai (纽约大学上海分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 28 pages, 9 figures, 10 tables
Abstract:Chain-of-thought (CoT) traces often serve as a proxy for how Large Language Models (LLMs) arrive at their answers. However, growing evidence shows that models’ CoT often fails to reflect their internal computations and can be changed without affecting their final answers. In this work, we measure and improve the alignment between the reasoning described in an LLM’s CoT and what it computes internally. We propose CoT-Interpretability Alignment (CIA), a metric that measures the agreement between a model’s CoT traces and its internal reasoning strategies as detected by interpretability tools. We evaluate CIA on three tasks (two-hop question answering, hint intervention, and integer multiplication) across three LLMs, finding that LLMs exhibit limited alignment across all tasks (44.8-75.9%). We then experiment with improving CIA via post-training, setting both the task accuracy and parametric faithfulness signals as a reward. Experiments show that we can substantially improve CoT parametric faithfulness while maintaining or improving the task accuracy. We provide rich analysis, such as their generalization patterns. Our work provides both a framework for auditing CoT parametric faithfulness and a pathway toward making models’ explicit reasoning more trustworthy. Code and data are available at this https URL.
[NLP-92] argeted Retrieval Compact Representations: How CoT Reasoning Improves Long-Context Counting
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在长文本上下文任务中性能提升的内在机制不明确的问题，尤其聚焦于思维链（Chain-of-Thought, CoT）推理如何增强模型对分散在长文本中的目标信息（如记录数）的计数能力。其核心解决方案在于通过“针堆找针”（needle-in-a-haystack, NIAH）计数任务进行系统性分析，揭示了两种截然不同的信息检索机制：一是非思考型模型采用的广域检索（broad retrieval），即对多个目标信息点进行广泛注意力覆盖；二是思考型模型采用的目标化检索（targeted retrieval），即利用思维链中的枚举过程逐个定位并提取目标信息。研究表明，目标化检索能更集中地分配注意力，并伴随更为紧凑的内部表征。进一步的因果干预分析表明，思考型模型可借助思维链轨迹在无显式编号的情况下持续维护和更新内部计数状态，这支持了思维链推理本质上是一种状态追踪（state-tracking）机制的观点。该研究将长上下文信息检索与计数任务的表征几何结构相联系，为理解CoT推理的内在机理提供了新的理论依据。
链接: https://arxiv.org/abs/2609.38958
作者: Liang Twist Shan,Tianyu Hu,Hao Yan,Yiqiao Zhong
机构: University of Wisconsin-Madison(威斯康星大学麦迪逊分校); University of Wisconsin-Madison(威斯康星大学麦迪逊分校)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Applications (stat.AP)
备注: 73 pages, including references and appendices
Abstract:Large language models (LLMs) have been rapidly improving in long-context tasks, powered by Chain-of-Thought (CoT) reasoning. However, the internal mechanisms underlying this improvement remain unclear. We investigate these mechanisms through a needle-in-a-haystack (NIAH) counting task, where an LLM is asked to count the number of records dispersed in a long text. Across twelve model comparison groups, Thinking (or reasoning) improves counting accuracy over Non-thinking, with pronounced gains at larger counts. This motivates our mechanistic analysis, which identifies two contrasting mechanisms: (i) broad retrieval, where Non-thinking models broadly attend to multiple needles; (ii) targeted retrieval, where Thinking models use enumeration in CoT traces to successively retrieve needles. Targeted retrieval concentrates attention on individual needles and is accompanied by more compact internal representations. Moreover, causal intervention analysis suggests that Thinking models use the CoT trace to maintain and update an internal counter as needles are successively retrieved, even without explicit numbering. In small controlled experiments, both retrieval mechanisms and counter states emerge under standard autoregressive training. Together, our results connect long-context retrieval with representation geometry of counting, supporting a state-tracking account of CoT reasoning.
[NLP-93] GraphForge: Training Working Agents with Graph-Anchored Workspace Synthesis
【速读】： 该论文旨在解决训练具备真实工作能力的智能代理（Agent）所面临的高质量、可验证数据匮乏问题。现有数据生成管道要么依赖模型生成文件，导致内容缺乏真实性和多样性；要么基于真实文件构建任务，但缺少针对具体任务的验证机制，致使结果质量无法保障。为此，论文提出GraphForge框架，其核心创新在于基于证据图（evidence graph）将任务定义与验证过程共同锚定在真实文件之上。通过以职业相关的种子（occupation-grounded seeds）实现可控多样性，GraphForge为每个种子构建包含真实文件的工作空间，并建立文件间的关联关系图谱；任务描述与评分标准均由此图谱衍生，确保任务需求有真实文件支撑，且每一评估指标均可追溯至具体的验证文件。此外，引入初始执行测试和修订代理，对任务与评分标准进行迭代优化，确保其可执行性与一致性。在该框架下生成的2,169条轨迹用于微调Qwen3.6-27B模型，在OpenHands、Workspace-Bench-Lite和SpreadsheetBench II三个基准上分别取得显著提升，且基于证据锚定评分标准的拒绝式微调进一步优化了性能，验证了该框架在生成高质量、可验证任务数据方面的有效性。
链接: https://arxiv.org/abs/2609.38923
作者: Qisheng Su,Hanchen Wang,Guanru Zhu,Huicheng Jiang,Qiuyinzhe Zhang,Kou Shi,Zhen Fang,Ziao Zhang,Qingnan Ren,Zehui Chen,Tao Gui,Feng Zhao
机构: University of Science and Technology of China(中国科学技术大学); Shanghai Innovation Institute; Fudan University(复旦大学); Shanghai AI Laboratory
类目: Computation and Language (cs.CL)
备注:
Abstract:Working agents need to read diverse files, coordinate tools, and produce deliverables. Training such agents requires tasks built on many real files with verifiable results, but few pipelines exist to synthesize this kind of data. Existing pipelines either generate files with models, which lack realism and diversity, or build tasks on real files without task-specific verifiers, leaving result quality unchecked. We introduce GraphForge, an evidence-graph based framework that grounds both the task and its verification in real files. Starting from occupation-grounded seeds for controlled diversity, GraphForge assembles a workspace of real files for each seed and builds an evidence graph over their relations. Since the task statement and rubrics are both derived from this graph, task requirements are backed by the workspace files and each criterion is anchored to the files needed to verify it. An initial rollout further tests executability, and a revision agent repairs the task and rubrics against the original files before trajectories are collected. Fine-tuning Qwen3.6-27B on 2,169 GraphForge trajectories brings GDPVal to 1445.7 (+65.7) under OpenHands, and Workspace-Bench-Lite and SpreadsheetBench II to 63.7 (+7.7) and 24.0 (+13.7) under Claude Code. Rejection fine-tuning on the SFT model’s own rollouts, with candidates selected by the evidence-anchored rubrics, yields further improvements on all three benchmarks, suggesting that the rubrics provide a useful selection signal. The data and models are available.
[NLP-94] K2P: Label-Free Knowledge to Prompt Distillation
【速读】： 该论文旨在解决无标签知识蒸馏（label-free knowledge distillation）中因缺乏真实答案（ground-truth answers）而导致教师推理结果不可验证的问题，进而避免学生模型因与教师共享错误而被误导。其核心挑战在于：在不更新模型权重的前提下，如何有效利用教师生成的推理过程来指导学生模型的提示（prompt）优化，同时规避教师自身可能存在的错误偏差。解决方案的关键在于提出一种名为Knowledge-to-Prompt（K2P）的新方法，该方法通过从教师解题路径中自动生成可复用的指令（reusable instructions），并利用教师与学生在成对响应中的交互进行迭代精炼（refinement），结合答案一致性（answer agreement）引导搜索与候选提示的选择。K2P特别设计了保留机制以防止自适应搜索忽略潜在有效的提示，并在预留的验证问题上进行最终选择，确保部署仅依赖于冻结的学生模型和选定的提示。理论分析进一步区分了生成与选择阶段的差距，并给出了在教师参考不完美时仍能保证准确率的条件。实验表明，K2P在多种推理任务和学生模型上均优于现有的无标签蒸馏方法，且性能接近有监督的提示优化方案。消融实验与提示库诊断揭示了教师解题路径和精炼机制的重要贡献，同时也指出了基于一致性引导选择的局限性。
链接: https://arxiv.org/abs/2609.38898
作者: Yingchuan Zhang,Haoran Lu,Wenxuan Zhong,Ping Ma
机构: University of Georgia (佐治亚大学)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 66 pages, 5 figures
Abstract:Knowledge distillation can transfer reasoning from stronger teachers to frozen students through reusable prompts, but avoiding weight updates does not eliminate supervision. Without ground-truth answers, teacher solutions are unverified, and agreement with the teacher can reward shared mistakes. We introduce Knowledge-to-Prompt (K2P) for label-free knowledge distillation to prompts. K2P synthesizes reusable instructions from teacher solutions, refines them using paired teacher and student responses, and guides search and selection with answer agreement. It retains candidates that adaptive search may undervalue and selects on reserved questions. Deployment uses only the frozen student and selected prompt. Our theory separates generation and selection gaps and gives conditions under which agreement-guided construction yields accuracy guarantees despite imperfect teacher references. Across reasoning tasks and students, K2P outperforms label-free alternatives overall and remains competitive with supervised prompt optimization. Ablations and archive diagnostics assess the contributions of teacher solutions and refinement, while revealing the limits of agreement-guided selection.
[NLP-95] Audio Token Attention Is Predictable Before the Language Model Runs
【速读】： 该论文旨在解决大音频语言模型（Large Audio Language Model, LALM）在处理长时语音输入时面临的计算资源消耗过高的问题，特别是由于音频令牌（audio token）数量庞大导致的推理延迟与显存占用瓶颈。其核心挑战在于：传统方法在语言模型前几层进行图像令牌剪枝（image-token pruning）虽有效，但对音频令牌不适用——因为音频令牌在早期层中仍具有高度注意力权重，且其重要性排序尚未稳定，需在语言模型运行前完成精确排序。解决方案的关键创新在于发现：音频令牌在整个语言模型中的注意力分布可在线性意义上从编码器输出直接预测，无需标签监督。研究提出名为Triage的方法，通过一个闭式求解的线性映射，在无标签条件下实现对音频令牌的初始压缩，预测准确率ρ ≥ 0.69（在十三个LALM中的十一个）。Triage进一步在第二层基于实际注意力分布修正预测，实现动态优化；同时在两个预设压缩预算下自适应调整，确保输出与全音频输入结果的偏差可控。实验表明，在保守预算下，其词错误率和准确率仅比全音频低0.04；在激进预算下，十二个转录任务中均优于所有基线方法，尤其在多选任务中以2.2–5倍压缩比超越最强基线DART，平均准确率提升0.043。此外，该方法因在语言模型前剪枝，使Qwen2.5-Omni-3B模型可处理的音频长度从21.8分钟提升至约62分钟，并支持单张GPU并发服务四倍于以往的5分钟语音流。
链接: https://arxiv.org/abs/2609.38878
作者: Kyoungjun Park,Yunzhe Li,Lili Qiu
机构: The University of Texas at Austin(德克萨斯大学奥斯汀分校)
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 43 pages, 7 figures
Abstract:A large audio language model (LALM) turns a minute of speech into 750-1,500 tokens and prefills every one. Image-token pruning often cuts after the language model’s first layers, where image tokens draw little attention. Audio tokens draw much more attention there, and their ranking is still far from final, so audio needs a ranking before the language model runs. Surprisingly, the attention an audio token will receive across the language model is already linearly predictable from its encoder output, before the language model runs. A linear map, fitted in closed form without labels, predicts this all-layer attention ranking at \rho \geq .69 on eleven of thirteen LALMs. Our method, Triage, cuts audio tokens by this prediction and, on multiple choice, cuts again at layer 2, correcting the prediction with the attention observed there. Triage sets its compression without labels, under two budgets that limit how far its output may differ from the model’s own full-audio output. At the conservative budget, its word error rate and accuracy stay within .04 of full audio. At the aggressive budget, Triage beats every baseline in all twelve transcription cases. On multiple choice, at 2.2-5x compression, it outperforms DART, the strongest baseline on average, by .043 in mean accuracy. Because it cuts before the language model, it raises the audio that fits in Qwen2.5-Omni-3B’s context window from 21.8 to about 62 minutes. At its most compressive point, Triage lets one GPU serve 4x as many concurrent 5-minute streams of that model. Project page: this https URL
[NLP-96] Where MLLM s Fail and Why: Causal Task Decomposition for Capability Failure Diagnosis
【速读】： 该论文旨在解决多模态大模型（Multimodal Large Language Models, MLLMs）在组合性任务中失败原因难以区分的问题：即无法判断错误是源于目标能力本身的内在缺陷，还是由上游前置能力（prerequisite）的错误所引发的级联误差。其解决方案的关键在于提出一种因果分解框架（causal decomposition framework），通过在任务的前置依赖关系上施加受控干预，将失败模式解耦为“能力缺陷”与“前置依赖错误”两类。该框架引入能力度量（NC、IC、RC）和贡献度量（N-Score、S-Score），分别评估任务在无辅助、前置正确或错误条件下的表现，并量化各前置任务对整体性能的必要性（necessity）与充分性（sufficiency）。研究进一步构建了CADET诊断基准，包含10个复合任务、46个单元任务及超过3.3万条人工标注问题，覆盖感知、空间、时间与认知等类别。实证结果显示，提供正确的前置条件可消除54%的认知类任务错误，使认知能力从最弱提升至超越空间与时间类任务水平；同时，因果贡献高度集中于少数关键前置任务，仅补充其中最重要的一项即可获得全部前置补全收益的84%，揭示了当前模型失败的系统性根源。
链接: https://arxiv.org/abs/2609.38851
作者: Xia Hu,Brian Potetz,Chun-Ta Lu,Huanfen Yao,Leonidas Guibas,Zhicheng Wang,Howard Zhou,Pengfei Xing,Andrew Gallagher
机构: Google Research(谷歌研究); Stanford University (斯坦福大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:End-to-end accuracy on compositional tasks records how often MLLMs fail, but cannot distinguish whether a failure reflects an intrinsic deficit in the targeted capability or a cascading error from an upstream prerequisite. We propose a causal decomposition framework that isolates these two failure modes through controlled interventions on the prerequisite dependencies of each task. Our capability metrics (NC, IC, RC) score each task under unassisted, correct, or incorrect prerequisites to diagnose where failures arise; contribution metrics (N-Score, S-Score), adapted from probabilities of causation, quantify each prerequisite’s necessity and sufficiency to determine why. We instantiate the framework in CADET, a diagnostic benchmark of 10 composite tasks decomposed into 46 unit tasks with over 33,000 human-annotated questions spanning perception, spatial, temporal, and cognitive categories. Diagnosing frontier MLLMs with our framework uncovers systematic patterns that end-to-end accuracy obscures. Capability-wise, supplying correct prerequisites eliminates 54% of errors on cognitive tasks, lifting them from weakest to above spatial and temporal. Prerequisite-wise, causal contributions are concentrated in a few critical prerequisites, and supplying the single most important one alone captures 84% of the gain from supplying all prerequisites.
[NLP-97] OpenJev-RLCD: A Working RLCD Implementation
【速读】： 该论文旨在解决生成式推理模型在决策过程中概率输出缺乏校准（calibration）的问题，即模型预测的概率分布未能准确反映真实置信度。现有方法如监督微调（SFT）结合温度缩放虽能部分改善校准性，但依赖人工调参；而基于可验证奖励的强化学习（RLVR）虽提升性能却导致模型过度自信。本文提出一种用于校准决策的强化学习方法（Reinforcement Learning for Calibrated Decisions, RLCD），其核心在于：模型生成推理过程（rationale）后，使用严格正确评分规则（strictly proper scoring rule）对最终的答案分布进行评分，从而直接优化校准性。关键发现是，通过方差恒等式（variance identity）揭示，对多个样本混合分布进行评分可奖励意见分歧的推理路径，而传统RLVR本质上正是这一目标的简化形式，缺少多样性激励项。为应对策略梯度噪声导致的推理模式退化问题，提出“先校准、再强化”的两阶段训练范式。实验表明，在两个推理任务上，基于Qwen3-1.7B的RLCD在准确性与选择性预测方面均优于或匹配SFT、RFT/STaR及GRPO（经温度缩放），且在GSM8K答案验证任务中，仅需一次查询即可以≤5%错误率覆盖50%以上样本，显著优于GRPO。当不确定性源于标注者分歧时，理论上RLCD无法超越交叉熵损失。
链接: https://arxiv.org/abs/2609.38850
作者: Zhimin Gao,Pichao Wang
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Decision models such as Jev answer questions with probabilities, which are only useful if they are calibrated. Open-source reproductions rely on supervised fine-tuning plus temperature scaling, while reinforcement learning from verifiable rewards (RLVR) makes reasoning models overconfident. We present a working implementation of reinforcement learning for calibrated decisions (RLCD) for reasoning models: the model samples a rationale, and we score the answer distribution it commits to afterwards with a strictly proper scoring rule. A variance identity shows that scoring the mixture of several samples rewards disagreeing rationales, and that RLVR is exactly this mixture objective without its diversity term. Optimized naively, the per-rationale objective either switches reasoning off or is drowned out by policy-gradient noise, which leads to a two-stage recipe: calibrate, then reinforce. With Qwen3-1.7B on two reasoning tasks (3 seeds, paired tests), RLCD matches or beats SFT, RFT/STaR and GRPO (each temperature-scaled) in accuracy and beats all of them in selective prediction; on GSM8K answer verification a single query decides \gvTwoCovFive% of the items at \le 5% error, versus \gvGrpoCovFive% for GRPO. When uncertainty comes from annotator disagreement, RLCD provably cannot beat cross-entropy. Code and results: this https URL.
[NLP-98] Scaling Parameter and Context in Attention: Native Sparse Attention from Mixture-of-Head
【速读】： 该论文旨在解决大语言模型在长文本上下文处理中因注意力机制（Attention）参数规模扩展导致的计算与存储开销过高的问题。传统方法通过增加注意力头数（Heads）来提升模型性能，但保留完整的词元历史（Token History）会显著增加内存消耗和计算成本，尤其在长序列场景下尤为明显。此外，现有方法难以实现参数规模扩展与上下文长度扩展之间的高效协同。针对这一挑战，论文提出NAMOH——一种架构原生的稀疏注意力机制，其核心创新在于：每个词元仅激活 K 个注意力头中的一个，且每个头仅在其分配到的子序列内执行因果注意力（Causal Attention），从而将活跃参数数量与可用上下文长度解耦。在均衡分配条件下，固定 K 而增加总头数 H 可缩短每个头的历史长度，减少单个词元的键值对（KV）访问量，同时不增加总的KV存储开销。进一步引入头相对旋转位置编码（Head-Relative Rotary Position Embeddings），以压缩路由子序列内的位置跨度，缓解由位置信息引起的注意力噪声。实验表明，NAMOH在保持相同总参数量的情况下，可超越全激活模型的表现；同时在活跃参数数量相当的条件下，相比小型密集模型具有更高效的长上下文推理能力。该方法兼容分组查询注意力（GQA）及现有稀疏注意力机制，为实现“参数扩展直接驱动上下文扩展”提供了新路径。
链接: https://arxiv.org/abs/2609.38832
作者: Zizhuo Fu,Runsheng Wang,Meng Li
机构: Peking University(北京大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Scaling attention parameters can improve language model quality, but retaining full token histories makes additional heads costly at long contexts. Furthermore, since attention retrieves and combines contextual information, parameter scaling should also support longer contexts. We therefore ask whether attention parameter scaling can directly enable efficient and effective context scaling. We introduce NAMOH, an architecture-native sparse attention mechanism that activates K of H heads per token. Each head retains only its assigned tokens and performs causal attention within this subsequence. Head selection thus jointly determines active parameters and available context without scanning the full history. Under balanced assignments, increasing H at fixed K shortens head histories and reduces per-token key-value (KV) access without increasing total KV storage. We further support head-relative rotary position embeddings to shorten positional spans within routed subsequences, aiming to mitigate position-induced attention noise. Experiments show that NAMOH can outperform fully activated models with the same total parameters, while enabling more efficient long-context inference than smaller dense models with matched active parameter counts. It remains compatible with GQA and existing sparse attention mechanisms. We hope this work offers a new path for scaling attention, with parameter scaling directly enabling context scaling.
[NLP-99] Forging LLM Authorship Fingerprints with Targeted Rewriting
【速读】： 该论文旨在解决生成式文本在经过有目的的重写后，其模型归属（model attribution）是否仍能准确反映原始生成模型的问题。尽管现有的模型归属分类器在未修改文本上表现良好，但无法验证其判断是否在面对主动规避时依然可靠。核心问题在于：当攻击者对某模型生成的内容进行针对性重写以伪造来源时，当前的归属机制是否仍具备鲁棒性。解决方案的关键是提出“目标指纹转移”（targeted fingerprint transfer）这一新范式，并设计名为ForgePrint的“搜索-蒸馏”框架。该框架首先通过搜索策略寻找能够将文本的归属特征向目标模型指纹靠拢的重写方案，随后将这些有效重写模式蒸馏至一个轻量级的40亿参数学生模型（Student model），实现一次前向传播即可完成高精度的指纹转移。实验表明，在CNN/DM数据集上，该方法达到70.2%的目标成功率，显著优于教师模型（54.1%）及六种现有重写基线（最高39.3%），且在将开源模型摘要迁移至商业模型时也取得68.3%的成功率。结果表明，仅依赖文本特征的归属判别可能产生误导性证据，即便其在原始文本上准确，也无法保证在对抗性重写下的真实性，从而揭示了当前模型归属技术在安全性和可信度上的潜在风险。
链接: https://arxiv.org/abs/2609.38831
作者: Haohan Yuan,Simin Chen,Xi Niu,Hanqing Guo,Depeng Xu,Haopeng Zhang
机构: University of North Carolina at Charlotte(北卡罗来纳大学夏洛特分校); George Mason University(乔治梅森大学); Indiana University(印第安纳大学)
类目: Computation and Language (cs.CL)
备注: 31 pages, 7 figures, 25 tables. Project page: this https URL
Abstract:Model-attribution classifiers can often identify which language model produced a text, making model-specific writing patterns a signal of provenance. Accurate attribution on unmodified text, however, does not show whether the prediction still identifies the original source after deliberate rewriting. We formulate this problem as targeted fingerprint transfer: rewriting one model’s output so that attribution classifiers assign it to a chosen target model. We study summarization, where different models receive the same document and express the same underlying content, providing a controlled setting for conditional generation. We introduce ForgePrint, a search-then-distil framework that first searches for rewrites that move attribution toward a target fingerprint, then distils the selected rewrites into a one-pass 4B Student model. On CNN/DM, the Student reaches 70.2% target success rate, outperforming both its Teacher (54.1%) and the strongest of six published rewriting baselines (39.3%), against held-out classifiers that are never queried by the attack. It also reaches 68.3% target success when transferring summaries from an open model toward chosen commercial models. These results show that fingerprint detectability should not be conflated with source authenticity, and that text-only attribution can provide misleading evidence of model identity under targeted rewriting, even when it is accurate on unmodified text.
[NLP-100] BARRAC: Adaptation of an English Aspect-based Sentiment Analysis Approach for Classification Tasks in Arabic Dialects
【速读】： 该论文旨在解决主流语言（如英语）中已有的自然语言处理方法在阿拉伯语任务中适应性不足的问题，特别是在阿拉伯语方言的细粒度情感分析、讽刺识别与方言识别等复杂任务上的性能瓶颈。其核心解决方案是提出一种名为BARRAC（Brainstorming Alignment and Replaced Representation learning for ArabiC tasks）的框架，关键在于将英语领域中基于属性的情感分析范式进行针对性重构：首先用阿拉伯语特有的语言手段（如修辞标记、句法结构等）替代原有的消费者评论属性池，以更好地捕捉阿拉伯语方言中的情感极性、讽刺语义及方言特征；其次，摒弃易受噪声干扰的自训练机制，采用两阶段训练策略以提升模型鲁棒性与泛化能力。实验结果表明，BARRAC在五个阿拉伯语方言数据集上实现了63.93%的平均宏F1值，优于现有少量标注下的最先进模型（SOTA）3%，并在四项任务上超越GPT-4o，验证了任务特定方法适配在阿拉伯语自然语言处理中的有效性。
链接: https://arxiv.org/abs/2609.38820
作者: Ali Almutairi,Gelareh Mohammadi,Imran Razzak,Aditya Joshi
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:With the rapid growth of Arabic NLP, several models, datasets and benchmarks have been reported. This paper asks whether approaches developed for majority languages like English can be adapted to Arabic tasks. We adapt an English aspect-based sentiment analysis framework to Arabic classification tasks and present the adaptation as BARRAC: Brainstorming Alignment and Replaced Representation learning for ArabiC tasks. BARRAC replaces consumer-review attribute pools with Arabic linguistic devices and markers for dialectal sentiment, sarcasm, and dialect identification, and replaces noisy self-training with two-stage training. Evaluated on five Arabic dialect datasets, BARRAC achieves a mean macro-F1 of 63.93%, outperforming the best few-label SOTA by 3%, and outperforming GPT-4o on four out of five tasks. Error analysis provides insights into remaining challenges. These results demonstrate that adapting task-specific approaches is a promising direction for Arabic NLP alongside adapting models, datasets and benchmarks.
[NLP-101] When Reasoning Goes Astray: Attention Dynamics of Uncontrolled Reasoning
【速读】： 该论文旨在解决大推理模型（Large Reasoning Models, LRMs）在执行复杂任务时出现的失控推理问题，即推理过程演变为冗余验证与持续生成循环，导致推理成本上升，并可能引发资源耗尽和服务降级。现有方法多依赖于截断长输出或检测表面重复，难以区分正常思考与失控推理，也无法解释良性推理如何退化为有害行为。为此，本文将LRM生成过程建模为四种状态，并提出基于动态注意力响应的推理状态分析方法（Reasoning-state Analysis via Dynamic Attention Responses, RADAR），实现实时识别当前推理状态，揭示有效反思如何演变为不受控生成。基于RADAR的分析，本文进一步通过注意力重对齐（Attention Realignment）将异常注意力分布修正为正常请求中的模式，以抑制过度反思与持续循环。时间序列分析表明，失控推理表现为注意力分布偏离正常模式，且异常趋势在重复出现前即可被检测到。实验表明，通过注意力重对齐可一致降低循环现象，同时基本保持良性性能。RADAR为推理失控提供了机制性解释，为识别关键故障阶段及设计针对性运行时干预策略提供了可操作的指导。
链接: https://arxiv.org/abs/2609.38817
作者: Yuanhe Zhang,Ziwei Wang,Jie Ren,Haoran Gao,Zhenhong Zhou,Fanyu Meng,Cong Wu,Li Sun,Sen Su
机构: Beijing University of Posts and Telecommunications(北京邮电大学); Wuhan University(武汉大学); JIUTIAN Research(九天研究院); Nanyang Technological University(南洋理工大学); Chongqing University of Posts and Telecommunications(重庆邮电大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Large reasoning models (LRMs) improve performance on complex tasks through extended reasoning, yet the same process can degenerate into redundant verification and persistent generation loops. Such uncontrolled reasoning increases inference cost and creates risks of resource exhaustion and service degradation. However, existing mitigations largely truncate long outputs or react to surface repetition, and thus fail to distinguish normal thinking from uncontrolled reasoning or explain how benign reasoning degenerates into harmful behavior. In this paper, we operationalize LRM generation as four states and further introduce Reasoning-state Analysis via Dynamic Attention Responses (RADAR), which identifies the current reasoning state in real time and characterizes how effective reflection can develop into uncontrolled generation. Guided by RADAR’s analysis, we further realign abnormal attention distributions toward patterns observed in normal requests and examine how this correction affects excessive reflection and persistent looping. Temporal analyses show that uncontrolled reasoning is characterized by attention distributions that deviate from normal generation, with abnormal trends becoming detectable before repetition begins. Correcting these deviations through Attention Realignment consistently reduces looping while largely preserving benign performance. Together, RADAR provide a mechanistic account of how reasoning becomes uncontrolled, offering actionable guidance for identifying critical failure stages and designing targeted runtime interventions.
[NLP-102] Youre Hired: Strategic Model Selection for LLM Collaboration
【速读】： 该论文旨在解决多大语言模型（Large Language Models, LLMs）系统中因依赖预定义且人工设计的模型池而导致的性能瓶颈问题，核心聚焦于多LLM协作系统中的模型选择（model selection）难题。其解决方案的关键在于提出并系统评估了一套涵盖9种不同策略的模型选择分类体系，包括基于模型描述多样性的方法、基于能力感知的行为多样性策略，以及由大语言模型驱动的招聘者（LLM-based recruiters）机制。实验在包含10和32个模型的两个候选池中进行，覆盖数学、编程、问答与推理等任务，并在四种协作算法下验证。结果表明，经过精心设计的选择算法相比随机或基于简单启发式（如仅选取单个表现最优模型）的团队，性能提升最高可达36.1%。尤其值得注意的是，基于模型能力与训练特征的筛选策略能有效降低选择过程中的方差，实现最优性能，因此被推荐用于实际部署前的模型组合。进一步分析显示，更大的候选池对浅层启发式方法构成更大挑战，而基于与候选模型交互并理解其能力的算法则能更稳健地过滤不匹配或存在安全风险的模型，并具备向新型、分布外任务泛化的能力。综上，研究强调了有原则、基于信息的团队选择对于构建高效多LLM系统的重要性，并提供了可信赖的模型选择算法框架。
链接: https://arxiv.org/abs/2609.38816
作者: Zongwan Cao,Ziyuan Yang,Shangbin Feng,Michael Duan,Skyler Hallinan,Bingbing Wen,Lucy Lu Wang,Yulia Tsvetkov
机构: University of Washington (华盛顿大学); University of Southern California (南加州大学); Allen Institute for AI (艾伦人工智能研究所)
类目: Computation and Language (cs.CL)
备注: 21 pages, 10 tables, 5 figures
Abstract:While multi-agent and model collaboration algorithms gain traction to combine the strengths of diverse Large Language Models (LLMs), existing systems remain bottlenecked on pre-defined and hand-crafted model pools. In this work, we investigate the problem of model selection in multi-LLM systems. We propose and systematically evaluate a taxonomy of 9 selection algorithms ranging from diversity of model descriptions, capability-aware behavioral diversity, and LLM-based recruiters. We conduct extensive experiments across two candidate pools of 10 and 32 models, deployed in four model collaboration algorithms, and evaluated across tasks spanning math, coding, QA, and reasoning. Results demonstrate that successful selection algorithms greatly outperform random or heuristics-based teams such as merely selecting the models with top individual performance, by up to 36.1% across settings. Specifically, capability- and training-based selection strategies alleviate selection variance and achieve the best performance, which we recommend to employ before deploying real-world multi-LLM systems. Further analysis reveals that larger candidate pools pose greater challenges to shallow selection heuristics, while algorithms grounded in interacting with candidate models and understanding model capability robustly filter out misaligned, unsafe models, as well as generalizing to novel, out-of-distribution tasks. Together, we establish that principled and informed team selection is critical and present strong model selection algorithms for assembling effective multi-LLM systems.
[NLP-103] Can Terminal Agents Trust Their Own Verification? Diagnosing and Improving Self-Verification
【速读】： 该论文旨在解决终端代理（terminal agent）在命令行环境中通过自我验证（self-verification）评估并修正自身解法时的可信度问题。尽管大多数代理在生成完整候选解后几乎都会启动验证过程，但其实际错误检测率仅为61.43%，错误修复成功率仅为49.36%，表明自我验证的主要瓶颈并非验证行为的触发，而在于错误的识别与修复能力。为此，论文提出学生条件化验证蒸馏（Student-Conditioned Verification Distillation, SCVD），其核心在于让学生模型先生成候选解，并在相同交互上下文中蒸馏教师模型的后续验证与恢复行为，从而增强学生模型的纠错能力。实验结果表明，SCVD在三个Qwen3.5基线模型上相较原模型提升9.74–16.85个百分点的Pass@1指标，且优于标准全轨迹蒸馏，在SWE-bench Verified数据集上还避免了后者显著的分布外性能退化问题。
链接: https://arxiv.org/abs/2609.38812
作者: Yingfeng Luo,Shaowei Wei,Daixin Wang,Dingyang Lin,Kaiyan Chang,Weiqiao Shan,Tong Zheng,Zhiqiang Zhang,Jingbo Zhu,Tong Xiao
机构: Northeastern University(东北大学); Inclusion AI, Ant Group(蚂蚁集团包容性人工智能实验室); University of Maryland, College Park(马里兰大学学院帕克分校)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Terminal agents rely on self-verification to assess and correct their solutions as they solve tasks through interaction with command-line environments. Yet how trustworthy such self-verification is remains poorly understood. To investigate this question systematically, we introduce a diagnostic framework that identifies the first complete solution in each trajectory, determines whether it is objectively correct, and uses this ground truth to quantify the agent’s subsequent verification and recovery behavior. Applying it to ten terminal agents on TerminalBench2.1, we find that verification is nearly universal after a complete candidate is formed, yet only 61.43% of incorrect candidates are detected and only 49.36% of detected errors are successfully repaired. These results show that the main weakness in self-verification lies not in initiating verification, but in detecting and repairing errors. Motivated by these findings, we propose Student-Conditioned Verification Distillation (SCVD), which lets the student first produce a candidate solution and distills a stronger teacher’s subsequent verification and recovery from the same interaction context. Across three Qwen3.5 backbones, SCVD improves \textscPass@1 on TerminalBench2.1 by 9.74–16.85 percentage points over the corresponding base models and by 4.49–8.61 points over the standard full-trajectory distillation, while avoiding the pronounced out-of-distribution degradation of full-trajectory distillation on SWE-bench Verified.
[NLP-104] StateTree: Enhancing Long-Term Dialogue Reasoning via Reinforcement Learning NEURIPS2026
【速读】： 该论文旨在解决个性化大语言模型在长期对话推理中面临的挑战：由于对话历史长且动态演化，相关证据分散于不同会话之间，用户偏好可能随时间变化，而传统的长上下文训练在数据稀缺和计算成本高昂的条件下难以有效应对。其核心解决方案是提出一种基于数据驱动的强化学习方法——StateTree，通过构建一个具有可验证真实答案的辅助任务来增强模型能力。该方法将多会话对话信息编码为树状结构路径追踪任务，将关键值记录嵌入跨会话的二叉树中，要求模型通过跨会话检索记录并比较时间戳以正确遍历分支，最终从干扰叶节点中恢复隐藏的目标问题。为此，研究引入渐进式课程强化学习（curriculum RL）逐步增加树深度，并设计了一种组合变体，使边携带步骤级推理片段，训练模型将部分线索组合成连贯查询。在仅10K token上下文上训练的StateTree可泛化至128K token场景，无需全量强化学习开销，展现出跨会话检索、时间推理、知识更新及组合式多跳推理等能力。实验表明，StateTree显著优于监督微调（SFT）与基于强化学习的基线方法，在保持短上下文通用推理能力的同时，实现性能提升：StateTree-7B在LongMemEval（128k）上最高提升达+23.60%，StateTree-14B达到59.00%准确率，超越QwenLong-L1-32B（45.20%）。
链接: https://arxiv.org/abs/2609.38809
作者: Naen Xu,Wanqing Cui,Yibo Hu,Shixin Hong,Hengyu An,Meiguang Jin,Junfeng Ma,Tianyu Du
机构: Zhejiang University (浙江大学); Taobao Tmall Group of Alibaba (淘宝天猫集团)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: NeurIPS 2026
Abstract:Large language models deployed as personalized assistants must reason over long, evolving interaction histories. However, in long-term dialogue reasoning, relevant evidence is scattered across sessions, preferences may be revised over time, and standard long-context training fails to address these challenges under data scarcity and prohibitive computational costs. We propose StateTree, a data-driven RL method that constructs a challenging auxiliary task from scarce dialogues with verifiable ground truth. StateTree augments multi-session dialogues with a tree-structured path-tracing task: key-value records are embedded across sessions to form a binary tree. Solving the task requires the model to traverse from root to leaf by retrieving records across sessions and comparing timestamps to resolve branches, then recover the hidden target question among distractor leaves. We apply curriculum RL training progressively increasing tree depth and introduce a compositional variant whose edges carry step-level reasoning fragments, training the model to compose partial cues into coherent queries. Trained on 10K-token contexts, StateTree generalizes to 128K tokens without full-length RL costs and exhibits capabilities including cross-session retrieval, temporal reasoning, knowledge update, and compositional multi-hop reasoning. StateTree outperforms both SFT and RL-based baselines while preserving short-context general reasoning. StateTree-7B achieves gains up to +23.60% on LongMemEval (128k), and StateTree-14B reaches 59.00% accuracy on LongMemEval, surpassing QwenLong-L1-32B (45.20%).
[NLP-105] Blackboard Intelligence Can Surpass Autoregressive on Globally Constrained Problems
【速读】： 该论文旨在解决大语言模型在处理受复杂全局约束（complex global constraints）问题时表现不佳的瓶颈，特别是由传统逐词预测（next-token prediction）推理接口所导致的局限性。其核心问题是：当前基于自回归生成的推理机制因强制遵循从左到右的因果序列，难以有效探索和维护全局一致性，从而在需要整体协调的组合优化任务中失效。为此，论文提出“黑板智能”（blackboard intelligence）这一新型推理范式，通过引入一个可反复修改的共享工作空间（revisable canvas），使模型能够在推理过程中搜索并修正候选解状态，而非固定于单向生成轨迹。解决方案的关键在于利用扩散语言模型（diffusion language models）的任意顺序预测能力，使其自然支持对部分填充解状态的预测；并通过观察模型内部的“平均置信度”（mean confidence）——这一源自标准掩码扩散目标的简单量度——作为全局一致性（global coherence）的有效代理指标，指导推理过程中的搜索与修正。实验结果表明，在ZebraLogic-Hard、护士排班（Nurse Rostering）和作业车间调度（Job-Shop Scheduling）等任务上，该方法在保持固定微调权重的前提下显著优于同规模自回归基线模型，并在多项指标上超越了更大规模且具备强测试时推理能力的前沿大模型，验证了其在复杂约束求解中的优越性。
链接: https://arxiv.org/abs/2609.38806
作者: Woosang Jeon,Jaeyeon Kim,Sham Kakade,Yilun Du,Amrit Singh Bedi,Arun Kumar Chithanar,Chul Lee,Taehyeong Kim,Sitan Chen
机构: Seoul National University(首尔国立大学); Harvard University(哈佛大学); University of Central Florida(中央佛罗里达大学); Arun Kumar Chithanar; Chul Lee; Taehyeong Kim; Sitan Chen
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 32 pages, 9 figures
Abstract:Next-token prediction has driven remarkable progress in large language models, yet a growing body of evidence suggests that they can struggle on problems governed by complex global constraints. In this work, we focus on this regime and ask whether some of these limitations arise from the inference interface induced by next-token prediction itself. We study this question through blackboard intelligence: an inference-time perspective in which a model works on a fixed, revisable canvas and searches over candidate solution states rather than committing to a causal, left-to-right trajectory. We instantiate this idea with diffusion language models, whose any-order prediction interface naturally exposes predictions over partially filled solution states. Our key observation is that mean confidence, a simple model-internal quantity available from the standard masked diffusion objective, provides a useful proxy for global coherence and can guide inference-time search and revision. Empirically, across ZebraLogic, Nurse Rostering, and Job-Shop Scheduling, Blackboard consistently improves inference while holding the fine-tuned LLaDA-8B-Instruct checkpoint fixed and substantially outperforms same-scale autoregressive baselines, reaching 90.4% accuracy on ZebraLogic-Hard, 76.4% exact feasibility on Nurse Rostering, and 80.2% optimality on JSSP. Stronger autoregressive search and refinement also fail to close the gap on ZebraLogic-Hard, while Blackboard surpasses tested frontier LLMs there and on JSSP despite their substantially greater scale and strong test-time reasoning. We open-source our codebase at this https URL.
[NLP-106] Uncovering Uncontrolled Repetition through Residual Stream Dynamics
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）及视觉-语言模型（Large Vision-Language Models, LVLMs）中由自回归生成过程引发的不可控重复问题，该问题不仅会延长生成时间，还可能被用于资源消耗攻击。现有研究虽已发现重复行为在中间和深层网络中表现为显著激活特征，但其早期演化机制尚不明确。本文提出一种名为**逐标记残差比较（Tokenwise Residual Comparison, TRC）**的方法，通过分析生成过程中残差流（residual stream）的动态变化，识别并定位与重复相关的异常模式。TRC通过对比注意力模块与多层感知机（MLP）对残差流的写入操作，捕捉重复性生成的潜在信号，并选择性抑制对应层中的特定残差坐标，从而实现对重复行为的早期干预。实验表明，TRC可平均降低57%的循环率，在LVLM、LLM及大型推理模型（Large Reasoning Models, LRMs）中均表现出良好的泛化能力。进一步分析揭示，重复语义早在浅层网络即已出现，并沿残差流传播，破坏正常表征。本工作将重复生成的研究从晚期显著特征拓展至早期干预窗口，为防御资源消耗攻击提供了新思路。
链接: https://arxiv.org/abs/2609.38802
作者: Yuanhe Zhang,Xinyao Zhou,Haoran Gao,Yuyao Zhang,Zhenhong Zhou,Fanyu Meng,Li Sun,Sen Su
机构: Beijing University of Posts and Telecommunications(北京邮电大学); JIUTIAN Research(九天研究); Nanyang Technological University(南洋理工大学); Chongqing University of Posts and Telecommunications(重庆邮电大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Uncontrolled repetition can prolong autoregressive generation in large language models (LLMs) and enable resource consumption attacks. Prior analyses of repetitive generation have identified strongly activated features in intermediate and late layers. However, how uncontrolled repetition activity emerges and develops before becoming prominent in these layers remains insufficiently understood. In this paper, we investigate this question primarily in large vision-language models (LVLMs), which support a richer set of uncontrolled repetitions through both visual and textual inputs. We propose Tokenwise Residual Comparison (TRC), a method that identifies and localizes anomalies associated with repetition from residual dynamics during generation. TRC compares attention and multilayer perceptron writes to the residual stream across generated tokens to identify patterns associated with repetition. It then selectively suppresses coordinates in the residual stream at the identified layer. Experiments show that TRC effectively mitigates uncontrolled repetition, reducing loop rates by 57% on average. Our analysis further shows that repetition semantics emerge in shallow layers and propagate through the residual stream, disrupting normal representations. TRC also generalizes to large language models (LLMs) and large reasoning models (LRMs), where it consistently captures analogous repetition dynamics and achieves effective mitigation. Our work broadens the study of repetitive generation from its prominent internal representations to earlier opportunities for intervention, providing insights for mitigating resource consumption attacks.
[NLP-107] Overlap Unique and Conflict: Can LLM s Extract What They Can Recognize?
【速读】： 该论文旨在解决多视角替代性叙事（multi-perspective alternative narratives）中信息一致性、冲突性与独特性识别的难题，即如何从多个来源的完整叙事中直接提取出重叠（overlap）、冲突（conflict）和唯一（unique）的语义片段。现有研究多集中于预定义文本对之间的关系分类（如蕴含或矛盾），而未能有效支持从完整叙事中进行细粒度的信息抽取。为此，本文提出一种新的跨叙事任务——重叠-唯一-冲突（Overlap-Unique-Conflict, OUC）提取，其关键在于系统性地从两篇叙事中识别并结构化输出三类语义关系的最小语义单元（即条款级片段）。为支撑该任务，研究构建了一个包含约22,000组叙事对及14万例OUC实例的基准数据集，涵盖事实性、论辩性和政治性话语。实验表明，当前主流大语言模型（LLM）在提取唯一信息方面表现良好（F1 > 75%），但在重叠与冲突信息提取上显著受限（最强模型Gemma-4-31B仅达61.13%与48.58%），且主要瓶颈并非关系识别本身，而是缺乏对完整叙事中对应条款的有效配对与提取能力，尤其在小模型中更为突出。然而，通过任务特定的监督微调可显著缩小差距：经微调的Qwen-3-8B模型在多个任务上超越了规模约为其四倍的Qwen-3.6-35B模型，绝对提升达15–28%。尽管如此，重叠与冲突的提取性能仍远未达到理想水平，表明跨叙事条款级信息提取仍是开放挑战。
链接: https://arxiv.org/abs/2609.38799
作者: Eftekhar Hossain,Santu Karmaker
机构: Bridge-AI Lab@UCF(桥-人工智能实验室@中佛罗里达大学); University of Central Florida (中佛罗里达大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Understanding multi-perspective alternative narratives requires identifying how their information agrees, conflicts, or differs across sources. Existing work on cross-text relations largely focuses on categorizing relations between predefined text pairs, such as entailment or contradiction, rather than directly extracting such information from full narratives. To address this gap, we introduce Overlap-Unique-Conflict (OUC) extraction, a cross-narrative task that extracts all overlapping, conflicting, and unique clauses from two narratives. To support this study, we construct a benchmark of approximately 22K narrative pairs and 140K OUC instances spanning factual, argumentative, and political discourse. Evaluating 14 open-source LLMs (0.6B-35B), we find that unique information is far easier to extract than overlap and conflict: the strongest model, Gemma-4-31B, reaches only 61.13% F1-score on overlap and 48.58% on conflict, against more than 75% on unique. Further diagnostic analysis reveals that this difficulty does not stem from relation recognition alone, but rather from a failure to pair and extract the corresponding clauses from full narratives, especially in smaller models. Nevertheless, learning these extractions with task-specific supervision narrows the gap considerably: a fine-tuned Qwen-3-8B gains 15-28% absolute over its baseline and surpasses models roughly four times its size (e.g., Qwen-3.6-35B) on several tasks. Even so, overlap and conflict remain well below satisfactory, leaving cross-narrative clause extraction an open challenge.
[NLP-108] Evaluating Persistent Calibration under Evolving Model Knowledge
【速读】： 该论文旨在解决生成式AI系统在持续学习与适应过程中，其置信度估计（confidence estimation）如何保持可信性的问题。随着模型从静态版本演进为具备持续学习能力的智能体，其知识和技能随时间动态变化，而置信度估计必须能够动态反映这种变化，而非依赖反复的人工标注或监督。为此，作者提出了“持久校准”（persistent calibration）这一核心问题：即置信度估计器需在不依赖额外监督的情况下，忠实反映模型在不同检查点（checkpoint）上的知识状态变化。其解决方案的关键在于识别并利用那些跨检查点具有鲁棒性的元知识特征（meta-knowledge features），这些特征能够支撑置信度估计在知识演化过程中的泛化能力。研究通过构建“知识对比集”（knowledge contrast sets）——即一个检查点正确回答而另一个错误的问题子集——来衡量置信度估计在知识转移场景下的表现。实验表明，无论是推理时的方法还是微调方法，在对比集上的校准性能均显著劣于基于未来检查点训练的“理想”（oracle）方法，即使这些方法在全数据集上已实现良好校准。这揭示了持久校准的挑战根源：存在大量在特定检查点上表现良好的置信度函数，但其中仅有部分依赖于可泛化的元知识特征。进一步实验表明，多检查点联合训练有助于提升对比集校准效果，为识别跨知识演进稳定的置信度特征提供了有效路径。
链接: https://arxiv.org/abs/2609.38797
作者: Victor Wang,Thomas Hofweber,Mohit Bansal,Elias Stengel-Eskin
机构: University of Texas at Austin(德克萨斯大学奥斯汀分校); 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); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Code: this https URL
Abstract:As AI systems move from static repositories to agents that are capable of continual adaptation and learning, maintaining their trustworthiness means equipping the models backing them with the ability to produce confidence estimates that dynamically reflect their changing skills and knowledge. We introduce the problem of persistent calibration, which requires a confidence estimator to faithfully reflect the knowledge contained in a model as that knowledge changes, without recurring supervision. We operationalize this by examining persistent calibration across checkpoints of open models, asking whether confidence estimators trained on earlier checkpoints can generalize to later ones. Specifically, we aim to shed light on whether confidence is dependent on knowledge, a question with implications for the reliability of confidence estimates. To measure this relationship, we define and evaluate calibration on knowledge contrast sets: subsets containing questions that one checkpoint answers correctly and another checkpoint answers incorrectly, reflecting a change in knowledge. We show that both inference-time and fine-tuning methods fall short on contrast-set calibration compared to oracle methods trained on future checkpoints, even for methods that are well-calibrated on the full dataset. We provide evidence for the hypothesis that persistent calibration is challenging because there is a vast space of possible confidence functions that are well-calibrated on a given checkpoint, out of which only some rely on meta-knowledge features that would generalize to other checkpoints. Towards improving contrast-set calibration, we show that multi-checkpoint training helps, suggesting an avenue for identifying confidence features that remain robust across changing knowledge.
[NLP-109] Recovering Off-Policy Supervision for Speculative Decoding
【速读】： 该论文旨在解决生成式AI（Generative AI）中推测解码（speculative decoding）的块草稿生成器（block drafter）在使用外部模型生成的语料库进行训练时所面临的监督信号严重丢失问题。其核心挑战在于，当块内任一令牌出现偏离策略（off-policy）时，整个后续序列的标签均失效，导致现有方法不得不丢弃这些不一致的序列片段，造成大量监督信息浪费。为此，作者提出一种基于回滚（rollout）的训练框架，通过两个互补组件实现全监督恢复：首先，锚点标签重标注（Anchor-Label Relabelling, ALR）利用贪婪目标回滚生成的分布替换原始语料标签，从而为所有预测位置恢复有效监督信号；其次，回滚内锚点（In-Rollout Anchors, IRA）将草稿块直接嵌入目标模型的回滚过程中，使生成器能够接触到目标模型生成的上下文信息，同时复用预计算的回滚特征，无需额外目标模型推理开销。实验表明，在固定视觉-语言与文本语料上，该框架相比DFlash可将贪婪接受长度提升最高达36.5%，且在单个训练周期内即超越最优丢弃策略，三轮训练后达到与重新生成目标响应训练相当的接受长度。结果证明，该方法能够在不修改原始文本的前提下，高效、低成本地实现对推测草稿生成器的有效训练。
链接: https://arxiv.org/abs/2609.38795
作者: Jungseob Lee,Chanjun Park,Sugyeong Eo,Hyeonseok Moon
机构: 未知
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 22 pages, 4 figures, 17 tables
Abstract:Block drafters for speculative decoding are commonly trained on corpora written by external models, where a single off-policy token invalidates supervision for all subsequent slots in a block. Existing approaches discard these divergent slots, resulting in severe supervision loss. To resolve this problem while preserving the training corpus, we propose a rollout-based training framework that recovers full supervision through two complementary components. The first component, Anchor-Label Relabelling (ALR), replaces corpus labels with distributions from greedy target rollouts, restoring valid supervision across all predicted slots. The second component, In-Rollout Anchors (IRA), places draft blocks directly inside these rollouts to expose the drafter to target-generated context, reusing precomputed rollout features at no additional target cost. Across fixed vision-language and text corpora, our framework increases greedy accepted length by up to 36.5% over DFlash and consistently outperforms erasing baselines. Notably, a single epoch of our method surpasses the best erase schedules. After three epochs, it matches the acceptance length of training on target-regenerated responses. These results show that our framework provides an effective and compute-efficient approach for training speculative drafters on fixed corpora without modifying the original text. Code is available at this https URL.
[NLP-110] raining LLM Judges from Language Feedback via Position-Selective Self-Distillation
【速读】： 该论文旨在解决大语言模型（LLM）在主观任务中基于自然语言反馈进行训练时的评价标准不明确与信号利用不充分问题，尤其关注评价结果对所采用评估标准及其权重的高度依赖性。现有主流方法——基于结果的强化学习（如GRPO）——仅将最终判断的准确性作为单一标量奖励分配给整个生成序列中的每个词元，未能对选择评价标准的词元提供独立的信用分配，且忽略了伴随偏好标签的丰富语言反馈（如偏好理由）。为更有效地利用这些语言反馈，论文提出自蒸馏（Self-Distillation, SD）机制：让同一模型在给定反馈条件下充当教师，提供细粒度的位置级监督。然而，并非所有位置均携带等量有效信号。通过分析教师与学生在各位置上的熵偏移（entropy shift），识别出两种模式：上下文锐化（context sharpening），即教师集中概率于某一与反馈一致的标准表达；以及上下文扩散（context spreading），即教师在多个与反馈一致的替代表达间分散概率。前者促进特定标准表达的记忆，后者则通过保留多种可能性推动语义理解。针对这一不对称性，论文引入基于熵偏移分布尾部的词元掩码策略，保留熵偏移较低的区域以避免过度聚焦于高确定性但可能过拟合的表达。实验表明，该掩码策略显著提升了模型在分布外场景下的泛化能力，相较于朴素自蒸馏，其性能进一步提升。最终，所提出的自蒸馏判别器在主观子类别任务上相比基于结果的强化学习方法性能提升2–9个百分点，同时在客观任务上保持竞争力。
链接: https://arxiv.org/abs/2609.38792
作者: Ilgee Hong,Changlong Yu,Zhenghao Xu,Xin Liu,Yuwei Zhang,Qin Lu,Bing Yin,Tuo Zhao
机构: Georgia Institute of Technology(佐治亚理工学院); Amazon(亚马逊); UC San Diego(加州大学圣地亚哥分校)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:We study training LLM judges from natural language feedback, especially for subjective tasks where the verdict depends strongly on which evaluation criteria the judge invokes and how it weighs them. The dominant approach, outcome-supervised RL (e.g., GRPO), credits every token in the rollout with a single scalar determined only by the accuracy of the final verdict, providing no separate credit at the criterion-choice tokens and ignoring the rich language feedback (e.g., preference rationales) that naturally accompanies preference labels. Self-Distillation (SD) is one natural way to use this language feedback: the same model, conditioned on this feedback, acts as a teacher providing dense, position-level supervision. However, not all positions carry equally useful signal. Using the per-position entropy shift between teacher and student, we identify two regimes: context sharpening, where the teacher concentrates probability on a particular feedback-aligned criterion expression, and context spreading, where the teacher distributes probability across multiple feedback-aligned alternatives. We interpret these patterns as follows: sharpening encourages memorization of a particular criterion expression, whereas spreading promotes semantic understanding by preserving these alternatives. Motivated by this asymmetry, we introduce position masking based on the entropy shift that retains the lower tail of the entropy-shift distribution. Experiments show that masking higher-entropy-shift positions improves out-of-distribution generalization over naive SD. The resulting self-distilled judges outperform judges trained with outcome-supervised RL by 2-9 percentage points on the evaluated subjective subcategories, while remaining competitive on objective ones.
[NLP-111] Anchor-ECC: Local Integrity Checking for Watermarked LLM Outputs via Error-Correcting Codes
【速读】： 该论文旨在解决生成式 AI（Generative AI）文本水印技术在后生成编辑场景下的安全性与完整性验证问题：即当经过水印的文本被少量修改（如插入、删除或替换）后，尽管语义可能发生变化，但原始水印信号仍可被保留，导致内容仍被错误归因于原生成模型，从而引发责任归属风险。其解决方案的关键在于提出一种名为 Anchor-ECC 的新型水印框架，通过将纠错码（Error-Correcting Code, ECC）约束与显式边界锚点（explicit boundary anchors）嵌入水印结构，并结合动态规划解码器，实现对后生成编辑的精准检测与定位。该方法在 Qwen3-8B、Mistral-7B-Instruct-v0.3 与 OPT-125M 等多个模型上均实现了约 99.7% 的块级真正例率（TPR），且假警报率（FAR）不超过 7.6%，同时保持水印文本与无水印文本之间的可区分性。此外，通过质量实验筛选出更低困惑度（perplexity）的配置，在保障生成质量的前提下仍维持强编辑检测能力，从而将大语言模型（LLM）水印从单纯的源标识扩展至局部完整性验证，并支持检测可靠性与生成质量间的可配置权衡。
链接: https://arxiv.org/abs/2609.38722
作者: Zewei Deng,Muhammad Siddeek,Liyan Xie,Mohamed Seif,Mengdi Wang,H. Vincent Poor,Andrea Goldsmith
机构: University of Minnesota(明尼苏达大学); Google(谷歌); Oakland University(奥克兰大学); Princeton University(普林斯顿大学); Stony Brook University(石溪大学)
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL)
备注: 18 pages, including references and appendices; 1 figure and 11 tables
Abstract:LLM watermarking has become an effective approach to distinguishing AI-generated text from human-written text by embedding detectable patterns during generation. However, a small post-generation edit may change the meaning of the text without removing its overall watermark signal, creating a risk that the modified content is still attributed to the original model. We propose Anchor-ECC, which incorporates the error-correcting code (ECC) constraints and explicit boundary anchors into the watermark structure and pairs them with a dynamic-programming decoder to detect and localize post-generation edits. Across Qwen3-8B, Mistral-7B-Instruct-v0.3, and OPT-125M, the approximate-hard setting achieves about 99.7% block-level true positive rate (TPR) with at most 7.6% false alarm rate (FAR) for edit detection under mixed insertions, deletions, and substitutions, while preserving the distinction between watermarked outputs and unwatermarked text. Additional quality experiments identify lower-perplexity configurations that retain strong edit-detection performance. Together, these results extend LLM watermarking from source identification to local integrity verification while supporting configurable trade-offs between detection reliability and generation quality.
[NLP-112] MetaSteer: Context-Conditioned nonlinear Steering via Attention-Projection Adaptation
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在行为调控中依赖线性、上下文无关的激活空间干预所导致的信息瓶颈问题，此类方法假设固定表征需涵盖多种行为差异，限制了模型对复杂上下文变化的响应能力。其核心解决方案是提出一种名为MetaSteer的新方法，通过学习非线性且上下文相关的干预机制，并将其应用于注意力投影矩阵，从而在构造上实现随输入上下文动态变化的激活效应，无需依赖线性概念几何假设。MetaSteer以基于偏好的优化方式训练，仅需一次在聚合偏好语料库上的训练即可零样本迁移至未见概念和分布外上下文。实验表明，尽管采用低秩适配器，MetaSteer仍能诱导结构化的、上下文依赖的隐藏状态轨迹变化，同时部分保留轨迹的局部动态特性（如速度与曲率）。在三个受控文本生成基准和多个模型家族及规模下的三类智能体设置中，MetaSteer在零样本场景下多数指标上达到或超越现有任务特定的调控基线。研究进一步揭示，更强的文本生成调控效果与更优的智能体级调控表现存在正相关关系，并探讨了可迁移调控带来的轨迹几何特性、能力保持性及安全性等关键问题。
链接: https://arxiv.org/abs/2609.38718
作者: Mehdi Jafari,Hao Xue,Flora Salim
机构: UNSW Sydney (新南威尔士大学); ARC Centre of Excellence for Automated Decision-Making and Society (ADM+S); The Hong Kong University of Science and Technology (Guangzhou) (香港科技大学(广州))
类目: Computation and Language (cs.CL)
备注: Preprint. Code and pretrained model checkpoints will be released shortly
Abstract:Steering large language models typically relies on linear, context-independent interventions in activation space, an assumption that recent work has challenged and that can induce an information bottleneck when a fixed representation must encode many behavioral distinctions. We introduce MetaSteer, a method that learns nonlinear interventions with context-dependent effects and applies them to attention projection matrices, producing activation effects that vary with the input context by construction and requiring no linear concept-geometry assumption. Framed as preference-based optimization, MetaSteer is trained once on a pooled preference corpus and transferred zero-shot to unseen concepts and out-of-distribution contexts. We find that, despite using low-rank adapters, MetaSteer induces structured, context-dependent changes in hidden-state trajectories while partially preserving aspects of their local trajectory dynamics, including velocity and curvature. We evaluate MetaSteer on three controlled text-generation benchmarks and three agentic settings across multiple model families and scales. MetaSteer matches or outperforms strong task-specific steering baselines on most aggregate comparisons in the zero-shot regime. Across the evaluated settings, stronger text-generation steering is associated with stronger agentic steering performance. We further discuss geometric trajectory effects, capability retention, and safety considerations raised by transferable steering.
[NLP-113] Strong Multilingual Privacy Tagging at Encoder Speed ACL
【速读】： 该论文旨在解决隐私信息擦除（privacy redaction）中如何在去除个人身份信息的同时，有效保留文本中的语义关系这一核心挑战。其关键解决方案在于构建一个支持细粒度命名实体标注的多语言模型，通过在35种语言的前沿模型标注数据上微调多语言编码器，并采用覆盖感知的映射人工黄金标准重放机制，避免未标注类型被误判为负样本；同时引入基于学习的±1字符调整策略以修复子词边界，提升实体边界的精确性。实验表明，该方法在7种语言的1,283个手工标注测试片段上达到88.8的红化F1分数，显著优于现有主流模型（如GLiNER2、Microsoft Presidio及OpenAI隐私过滤器），且在仅增加约5万条标注训练句并增强黄金标准重放后，精确的类型化跨度F1从74.5提升至76.3。此外，该架构实现了比GLiNER2高4.9倍的CPU吞吐量，而本地大语言模型（LLM）在推理速度与标注效率方面表现不佳，验证了专用编码器设计的有效性。研究开源了代码、提示模板、训练方案及数据获取脚本，推动可扩展、高效、多语言隐私保护技术的发展。
链接: https://arxiv.org/abs/2609.38630
作者: Jonathan Graehl
机构: RWS Language Weaver(瑞文语言编织者)
类目: Computation and Language (cs.CL)
备注: 46 pages, 23 figures. Includes supplementary appendices. Submitted to ACL Rolling Review, October 2026 cycle
Abstract:Privacy redaction must remove personal information while preserving relationships expressed in text. We develop a multilingual named-entity tagger with fine-grained distinctions supporting varied redaction policies and methods for cheaply learning additional distinctions. We fine-tune a multilingual encoder with an affine span-tagging head on frontier-model annotations in 35 languages, replay mapped human gold with coverage-aware masking so unannotated types are not treated as negatives, and repair subword boundaries with a learned +/-1-character adjustment. On 1,283 human-gold test segments in seven languages, best measured redaction F1 is 88.8, against 69.1 for published GLiNER2 with 11 unrepresentable types excluded from its task (68.8 without that exemption), 67.8 for GLiNER2 adapted to the new training data, 57.3 for Microsoft Presidio and 35.8 for the best published OpenAI Privacy Filter fine-tune. Adding about 50,000 annotated training sentences and increasing human-gold replay improves exact typed-span F1 from 74.5 to 76.3 on Ont3, our 31-type frontier-annotated NER evaluation of 1,201 development segments. Mapped-gold replay alone raises human-gold F1 by ten points without loss on frontier-annotated text; boundary adjustment adds 1.7 exact typed-span F1 points on Ont3. Local LLMs fitting on a single 96-GB GPU underperformed as prompted annotators and frozen encoders, with encoding 30-95 times slower than XLM-R inference and prompted annotation roughly 180-1,100 times slower in the evaluated configurations. The encoder architecture delivers 4.9 times GLiNER2’s CPU throughput. We release code, prompts and training recipes, with data-acquisition scripts and source links.
[NLP-114] Marking Contour Tones in Yorùbá
【速读】： 该论文旨在解决约鲁巴语（Yorùbá）作为声调语言中轮廓音调（contour tones）的正字法难题，尤其针对人名和词汇中因传统拼写省略或错误表示声调信息而导致语义反转的问题。现有解决方案无法有效传达复杂的音调轮廓，致使部分名称的拼写不仅丢失音调信息，甚至颠倒原意。其核心解决方案是引入变音符号——锐音符（caron）与抑扬符（circumflex），二者在约鲁巴语音学研究中已有先例（自Olmsted, 1951年起），可用于单个元音上标记升调与降调轮廓音调。该方案通过标准键盘输入与计算文本处理实现可操作性，并已在WriteYoruba输入法及TTSYoruba语音合成器中完成实现，其系统架构与听觉评估结果已由Tubosun等人（2026）独立报告。
链接: https://arxiv.org/abs/2609.38627
作者: Kólá Túbòsún
机构: 未知
类目: Computation and Language (cs.CL)
备注: Under review at the 12th World Congress of African Linguistics (WOCAL 12)
Abstract:Yorùbá is a tonal language in which contour tones pose persistent orthographic challenges. These are especially notable for personal names and lexical items whose conventional spellings avoid vowel lengthening that would otherwise provide a host syllable for the second tone. A particular concern is a class of names in which the conventional spelling does not just omit tonal information but inverts the meaning of said name, sometimes asserting the opposite of what the name intends. This paper describes the problem, illustrates the inadequacy of current solutions, and proposes the adoption of the caron and circumflex marks. These are symbols with precedent in Yorùbá phonological scholarship since Olmsted (1951), used as orthographic conventions on single vowels to encode rising and falling contour tones, making them accessible for the first time through standard keyboard input and computational text processing. The proposal is supported by an implementation in the WriteYoruba keyboard and the TTSYoruba speech synthesizer, whose architecture and listener evaluation are reported separately (Tubosun et al., 2026).
[NLP-115] When Scientific Contradictions Are Lost in Translation NEURIPS NEURIPS2026
【速读】： 该论文旨在解决科学发现之间存在分歧但未必矛盾时，如何判断其是否真正冲突的核心问题——即在缺乏明确可比性的情况下，模型如何决定哪些测量结果具有可比性并据此进行科学验证。其解决方案的关键在于揭示语言模型在面对形式化约束与生物合理性之间的权衡时的决策机制：当科学结论以正式化约束形式呈现时，模型能有效识别最优支持的解（如GPT-5.6 Sol和Claude Opus 5分别在90%和96%情况下恢复最佳支持分配）；但在自然语言描述中，模型表现出对生物学合理性的偏好，倾向于选择符合预期生理机制的解释，而非严格满足约束条件的解。通过引入形式化请求和配对设计提示，可显著削弱这种偏好，使模型恢复对形式证据的依赖，从而将正确匹配率从27%提升至92%（p<0.001）。这表明，可靠的科学验证不仅依赖于形式推理能力，更取决于模型能否准确识别可比性关系及其隐含逻辑，而这一能力在自然语言语境下极易受到先验知识偏见的影响。
链接: https://arxiv.org/abs/2609.38621
作者: Tal Zeevi,Trey W. Jensen,Maxwell Strome
机构: Beakr
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted at the NeurIPS 2026 AI for Science Workshop: Verification in the Age of AI Scientists. This version is not included in the official NeurIPS proceedings
Abstract:Two scientific findings can disagree without contradicting each other. Determining whether they conflict requires knowing whether they describe comparable measurements. We study how language models behave at this decision point. In a controlled task, we generate an unsatisfiable XOR constraint system and translate its constraints into scientific reports from different laboratories. One assignment satisfies more constraints, while another satisfies fewer but better matches expected biology. This creates a simple dilemma: does the model choose the assignment that best fits the constraints, or the one that better matches biological expectations? When the constraints are stated directly, GPT-5.6 Sol and Claude Opus 5 recover the best-supported assignment in 90% and 96% of cases, respectively. In scientific prose, however, the models behave differently. Claude Opus 5 often prefers the biologically expected assignment. Removing that biological preference increases recovery of the better-supported assignment from 27% to 79% (p.001); recovery reaches 92% when the same Biology-favored record is accompanied by a formalization request and an explicit paired-design cue (p.001). GPT-5.6 Sol is less sensitive, with neither corresponding change reaching statistical significance. These results suggest that reliable scientific verification depends not only on formal reasoning, but also on how models decide which findings should be compared and what relations they imply.
[NLP-116] StreamDecisionBench: Evaluating Decisions in Force on Evolving Language Streams
【速读】： 该论文旨在解决生成式 AI（Generative AI）在程序中作为决策组件时，因输入证据流的动态变化而导致的“状态滞留”问题：即模型基于过时状态做出的正确决策在状态变更后仍被持续执行，从而造成实际运行中的错误。传统评估方法仅关注最终输出的准确性（untimed accuracy），忽略了决策在时间维度上的有效性，因此无法捕捉此类由延迟或判断失误引起的实时错误。为此，论文提出 StreamDecisionBench（SDB）评测基准，其核心在于对每个时间点上“在役决策”（in-force decision）进行精确追踪，并将每一时刻的错误归因于判断力（judgment）或延迟（latency）中的至少一项。SDB 通过四类典型应用场景，以可执行代码生成参考决策，实现了对真实动态环境的高保真模拟。其关键评估指标为在 1–5 秒更新间隔下，基于对数时间轴的归一化曲线下面积（normalized area under the curve），该指标对等乘性时间区间赋予同等权重，有效量化了综合决策质量。实验表明，该指标与场景级“非定时准确率”与理想时间评分的乘积高度一致（误差不超过 2.9 分），揭示了推理能力虽提升判断力，但显著增加延迟成本（如 Luna 和 Terra 模型分别损失 38.8 和 42.8 分），而低计算开销但判断力较弱的快速组件仍可实现与未启用推理的 Terra 相当的综合表现。聚合曲线与家族曲线进一步揭示了不同任务中判断与延迟权衡关系的动态变化，凸显了评估结果对时间尺度的敏感性。
链接: https://arxiv.org/abs/2609.38612
作者: Jhen-Ke Lin,Chung Chun Wang
机构: National Yang Ming Chiao Tung University (国立阳明交通大学)
类目: Computation and Language (cs.CL)
备注: 22 pages, 6 figures. Code and data: this https URL
Abstract:As natural language drives more applications, language models increasingly run inside programs as decision components: the program sends them the current state and acts on the returned decision until a newer one arrives. When evidence changes during inference, a decision correct for its own state can stay in force after that state has passed, as when a call recorder keeps running after a customer starts reading out a card number; untimed (offline) accuracy counts such an error as correct. We introduce StreamDecisionBench (SDB), which evaluates the decision in force at every instant and attributes every erroneous instant to judgment, latency or both. Its scenarios stream evidence in four application families, with reference decisions computed from public rules by executable code. We summarize in-force accuracy across update intervals of 1-5 s by its normalized area under the curve on a logarithmic time axis, giving equal weight to equal multiplicative ranges. Across six settings of four hosted models, this score stays within 2.9 points of the scenario-wise product of untimed accuracy and an oracle’s integrated timing score. Reasoning improves judgment, but at low effort latency costs Luna and Terra, two GPT models we also evaluate without reasoning, 38.8 and 42.8 points relative to untimed accuracy; a faster component with weaker judgment attains a similar integrated score to Terra without reasoning. The aggregate and family curves show where these tradeoffs change, making the evaluation’s time-scale dependence visible.
[NLP-117] SecureVibe: Making Vibe Coding More Secure
【速读】： 该论文旨在解决生成式AI在实现功能正确性的同时，仍存在显著安全漏洞的问题，尤其关注那些看似功能正确但存在潜在安全风险的代码解决方案。其核心问题是：当前生成式AI在编写代码时，对隐藏于功能需求背后的安全部分缺乏有效的规划与测试行为，导致安全缺陷难以被发现和防范。针对这一问题，论文提出SECUREVIBE，一种专门强化代码安全规划与测试能力的训练范式。其关键在于构建围绕安全行为的显式训练信号，包括基于4个安全任务的安全套件监督微调，以及两种后训练方法——利用可验证执行反馈的SECUREVIBE_rl和基于提示引导的自监督机制的SECUREVIBE_hg。实验表明，SECUREVIBE在多个基准上显著提升安全通过率（如在BaxBench上提升6.9分，在SusVibes上对未见CWE类别提升11.5分），同时兼顾功能性提升（在SusVibes上功能通过率提升13.6分）。进一步分析揭示两个关键洞见：（i）跨安全规划、编码与测试的多样化监督比单纯增加编码轨迹更有效；（ii）当模型自身安全能力不足时，提示引导的监督能显著弥补从结果反馈中学习的局限性。
链接: https://arxiv.org/abs/2609.38606
作者: Danqing Wang,Baolin Peng,Zhepei Wei,Isadora White,Wenlin Yao,Hao Cheng,Qianhui Wu,Minseon Kim,Xingdi Yuan,Lei Li,Jianfeng Gao
机构: Microsoft Research; Carnegie Mellon University
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL)
备注:
Abstract:As vibe coding becomes increasingly capable and widespread, security vulnerabilities in even functionally correct solutions are a growing concern. When investigating functionally correct but insecure solutions, we find that the insecure agent is less than half as likely to conduct effective planning and testing for the hidden security risks behind the functional requirements. Motivated by this, we develop SECUREVIBE, a training recipe that explicitly targets planning and testing for code security. SECUREVIBE constructs training signals around these security behaviors. It includes supervised fine-tuning on the security suite with 4 security tasks, and post-training methods, SECUREVIBE_rl and SECUREVIBE_hg, to enhance security capabilities from verifiable execution feedback and hint-based self-supervision. Our SECUREVIBE outperforms the baseline on two types of security coding tasks across 4 benchmarks. Specifically, SECUREVIBE improves the security pass@1 by 6.9 points on BaxBench. The gains extend to unseen CWE categories, with improvements of 11.5 points on SusVibes. Meanwhile, it also improves functionality pass@1 by 13.6 points on the security coding task SusVibes and 4.1 points on the generic coding task SWE-bench Verified. Further analysis offers two practical insights: (i) diversifying supervision across security planning, coding, and testing strengthens security behaviors more effectively than adding coding trajectories alone, and (ii) hint-guided supervision is particularly valuable when the agent’s existing security capabilities are insufficient to learn effectively from outcome feedback.
[NLP-118] Beyond Oracle Communication: Benchmarking Interactive Intent Alignment Under Miscommunication and Evolving User Intent
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）代理在实际应用中面临的交互式意图对齐（Interactive Intent Alignment）问题。传统基准测试假设用户能够始终准确且充分地传达固定意图，但在真实场景中，用户常存在沟通不畅、意图变更或耐心耗尽等现象，导致模型难以持续追踪用户的动态需求。为应对这一挑战，论文提出了一种系统性的解决方案：Drift-Bench++，这是一个可验证的、可执行任务的基准构建管道，支持可控的意图错位（misalignment）与意图漂移（intent shifts），并引入包含有限耐心、多样化模拟用户以及基于沉默交互触发的意图变化的交互协议。同时，研究设计了GRIP评估协议，涵盖任务锚定、用户真实性、提问有效性及对动态意图的适应能力等维度。实验结果表明，尽管更强的交互能力有助于提升性能，但模型表现仍显著落后于理想化的“预言机”（oracle）水平；对生产环境中部署的ProdAgent会话进行验证也显示，所建模的失败模式在实际应用中普遍存在且具有重要影响。因此，Drift-Bench++为评估和推动智能体在真实通信环境与不断演化的意图下表现提供了统一、可执行的基准基础。
链接: https://arxiv.org/abs/2609.38604
作者: Zheyuan Zhang,Mengyuan Chao,Ke Xiao,Ziyi Chen,Daoan Zhang,Yan Zhang,Yanfang Ye,Wei Xu
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Modern LLM agents increasingly tackle complex tasks through interactive, long-horizon exchanges with users, while existing benchmarks generally assume that users always accurately and sufficiently communicate a fixed intent. However, this oracle communication assumption rarely holds in practice: users may miscommunicate, change their goals, and run out of patience. We define this task setting as Interactive Intent Alignment, where agents must recover and continuously track the user’s current intent despite imperfect communication and evolving goals. To study this setting, we introduce Drift-Bench++, a principled benchmark construction pipeline for verified executable tasks with controlled misalignment and intent shifts, along with an interaction protocol featuring finite patience, diverse simulated users, and silent interaction-conditioned shifts. We further develop GRIP, a comprehensive evaluation protocol covering task grounding, user realism, inquiry effectiveness, and adaptation to evolving intent. Across diverse environments, models, and interaction conditions, stronger interaction consistently helps but remains far from oracle performance; Validation on deployed ProdAgent sessions further shows that the modeled failures are prevalent and consequential in deployment. By providing a unified, executable benchmark for interactive intent alignment, Drift-Bench++ offers a foundation for evaluating and advancing agents under realistic communication and evolving intent.
[NLP-119] Prompt2Skill: Unsupervised Skill Optimization From Natural Language Instructions
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在特定领域应用中依赖高质量专家编写的技能（Skill）所面临的两大核心问题：一是专家编写技能成本高昂且难以针对具体模型进行优化，不同模型版本、规模或训练数据可能导致其失效；二是新兴任务可能超出现有技能库覆盖范围，而缺乏标注数据时无法及时构建新技能。为应对上述挑战，本文提出 Prompt2Skill 框架，其关键创新在于仅通过自然语言任务描述即可自动构建技能，无需依赖预设的、分布内（in-distribution）的训练数据集。该框架通过从提示中推导任务规范，自动发现或合成数据集，并在反思式编辑的闭环过程中持续优化技能。实验结果表明，在问答、阅读理解、电子表格操作和数学推理四个领域中，Prompt2Skill 均显著优于直接提示基线，平均提升达 10.8 分，验证了其在无监督、零样本场景下的有效性与通用性。
链接: https://arxiv.org/abs/2609.38593
作者: Bo Ni,Li Li,Ryan A. Rossi,Franck Dernoncourt,Tyler Derr
机构: Vanderbilt University (范德比尔特大学); University of Southern California (南加州大学); Adobe Systems (Adobe系统)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Skills are external artifacts that Large Language Models (LLMs) consume at inference time to improve their performance on specialized domains by incorporating relevant procedural and domain knowledge. Expert-authored skills are expensive to produce, and the resulting artifacts are not optimized for the specific model that consumes them, whose failure modes can vary with version, scale and training. In addition, emerging tasks may fall outside the scope of existing skill libraries, creating a need to develop new skills before curated training data become available. Recent works have explored automated skill optimization through reflection, but they require a curated, in-distribution training set, which users might not always have. To address these limitations, we present Prompt2Skill, a framework that builds skills from natural-language task description alone. From the prompt, the system derives a task specification, discovers or synthesizes datasets, and refines the skill in a closed loop of reflective editing. Across four domains spanning question answering, reading comprehension, spreadsheet manipulation, and mathematical reasoning, Prompt2Skill consistently outperforms the direct prompting baseline, achieving an average improvement of 10.8 across open-source and frontier models.
[NLP-120] MedKIT: Evaluating Knowledge Integration and Generalization in Large Language Models NEURIPS2026
【速读】： 该论文旨在解决大模型在持续更新现实世界知识（如医学临床证据）过程中，新知识虽被有效记忆但难以实际应用的问题。现有评估方法仅关注事实召回率，无法衡量新知识是否真正具备可用性。为此，作者提出MedKIT（Medical Knowledge Integration and Transfer）基准，通过模拟真实临床更新序列，对模型在词汇变化、关系变换、复合推理及开放性操作化等多维度上的知识迁移能力进行细粒度评估，并包含知识保真度测试。其解决方案的关键在于构建一个系统化的评估框架，揭示当前12种知识集成策略普遍存在“召回高但可用性低”的现象：尽管在原始任务和词汇变体下表现良好，但在关系泛化、复合推理与实际操作任务中效果有限。这一发现凸显了知识集成中的核心挑战——如何使新知识在多样任务与上下文中实现稳定可用，同时将MedKIT确立为推动更高效知识整合方法发展的关键测试平台。
链接: https://arxiv.org/abs/2609.38543
作者: Lukas Thede,Yash Kumar Atri,David Chen,Danielle Bitterman,Matthias Bethge,Tom Hartvigsen,Zeynep Akata
机构: University of Tübingen, Tübingen AI Center(图宾根大学，图宾根人工智能中心); Helmholtz Munich(慕尼黑亥姆霍兹研究中心); Munich Center for Machine Learning (MCML)(慕尼黑机器学习中心); University of Virginia(弗吉尼亚大学); University of British Columbia(不列颠哥伦比亚大学); Harvard Medical School(哈佛医学院); Technical University of Munich(慕尼黑工业大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026 (Evaluations Datasets Track)
Abstract:Constantly evolving real-world knowledge necessitates models to be updated continuously. Especially in medicine, as clinical evidence changes over time, outdated knowledge can pose safety risks. Existing evaluations of knowledge integration focus on factual recall, offering limited insight into whether newly integrated knowledge is actually usable. Our benchmark MedKIT (Medical Knowledge Integration and Transfer) provides a granular evaluation of how models integrate and apply knowledge under realistic sequences of clinical updates. Each instance corresponds to a factual update derived from clinical evidence, paired with targeted probes that assess transfer across lexical variation, relational transformations, compositional reasoning, and open-ended operationalization, as well as locality tests for knowledge preservation. Using MedKIT, we conduct a large-scale empirical study of 12 knowledge integration strategies across 5 diverse models, including both general-purpose and medical LLMs. Our results reveal a consistent gap between recall and usable knowledge: while most methods achieve strong gains on the original update task and under lexical variation, relational generalization is limited, and no method yields meaningful improvements on compositional or operational tasks. These findings highlight a fundamental challenge in knowledge integration and position MedKIT as a testbed for developing methods that make newly integrated knowledge more consistently usable across tasks and contexts.
[NLP-121] Shifting Mechanisms: How Positional Encoding Choice Shapes In-Context Retrieval
【速读】： 该论文旨在解决生成式语言模型在长上下文场景下，不同位置编码（Positional Encoding, PE）策略如何影响上下文内检索（in-context retrieval）机制这一关键问题。具体而言，随着模型架构趋向于在不同层中采用差异化的注意力跨度与位置编码方式（如局部层使用旋转位置编码RoPE配合滑动窗口注意力，全局层使用无位置编码NoPE配合全局注意力，即SWA NoPE），其内部检索机制的演化路径尚不清晰。论文从机制层面出发，系统分析了不同位置编码选择对模型检索行为的影响。研究发现，在22个跨八类开源模型的实验中，采用标准RoPE的模型主要依赖位置线索进行检索，而混合型位置编码（PE hybrid）模型则逐渐转向基于语义内容的检索机制。进一步通过受控预训练消融实验表明，仅将位置编码限制在局部层即可引发这种从位置检索向语义检索的机制转变，并导致位置信息表征能力下降。更重要的是，研究揭示了现有文献报道的“长上下文性能提升”可能掩盖了内在的检索权衡：尽管SWA NoPE在多目标检索和问答任务中优于传统RoPE，但在区分竞争性键值时表现退化。这表明，性能变化更应归因于检索机制的根本性转变——由位置驱动转向语义驱动，而非简单的长上下文能力增强。因此，解决方案的关键在于识别并理解位置编码设计对模型内部检索机制的深层影响，强调需以机制视角替代单纯性能指标评估模型在复杂上下文中的行为。
链接: https://arxiv.org/abs/2609.38530
作者: Eric Enouen,Sainyam Galhotra
机构: Cornell University (康奈尔大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Language models increasingly use architectures that vary attention span and positional encoding across layers, such as applying RoPE with sliding-window attention and NoPE with global attention (SWA NoPE). However, how these choices shape in-context retrieval remains unclear. To study this question, we take a mechanistic view, tracing how positional encoding (PE) choice shapes the internal mechanisms models use for in-context retrieval. Across 22 open-weight models spanning eight families, we find that standard RoPE models rely primarily on positional retrieval, while PE hybrids shift toward semantic retrieval. We further show on a controlled pre-training ablation that confining positional encoding to local layers produces this semantic shift, degrading representations of positional information. Finally, we show that the reported long-context gains of PE hybrids mask a retrieval trade-off: SWA NoPE improves over RoPE on multiple-target retrieval and QA, but degrades when distinguishing competing keys. We show that these behavioral differences better track the mechanism shift from positional toward semantic mechanisms than a uniform improvement in long-context retrieval.
[NLP-122] DEdit: Iterative Draft Editing for Speculative Decoding
【速读】： 该论文旨在解决生成式 AI（Generative AI）中自回归大语言模型（LLM）推理速度慢的问题，尤其针对基于扩散模型的轻量级草稿器（diffusion-based drafter）在并行生成多标记时因独立预测导致的脆弱性问题。当早期生成出现错误时，整个草稿将被验证阶段丢弃，即使后续内容仍具价值。为此，论文提出 DEdit，一种具备迭代编辑能力的扩散型草稿器，不仅支持传统的并行去掩码生成，还能通过标记到标记的预测实现逐轮修正。其核心创新在于引入双向上下文机制，使后期预测可作为上下文修复前期错误，并延长有效接受前缀。为训练模型在修正错误的同时保留正确预测，作者提出 ProposalMix 训练策略，依据首轮置信度动态混合草稿预测与真实标签。实验表明，在 Qwen3-4B 与 Qwen3-8B 上的七个基准测试中，DEdit 在宏平均标记接受率和加速比上均优于现有方案，分别达到 5.72× 和 5.97× 的加速效果。进一步分析显示，更多编辑轮次与更宽的草稿窗口可提升接受率，ProposalMix 能有效减少有害编辑，而限制编辑器仅使用因果注意力会显著降低性能，证明未来上下文是关键增益来源。
链接: https://arxiv.org/abs/2609.38510
作者: Longxuan Yu,Bingsen Chen,Peng Shi,Dongkyu Lee,Yi Xiang,Hideo Kobayashi,Sheng Zhang,Shuaichen Chang,Xing Niu,Zhuoyan Xu,Greg Ver Steeg,Jiarong Jiang
机构: University of California, Riverside (加州大学河滨分校); Amazon Web Services (亚马逊网络服务); New York University (纽约大学)
类目: Computation and Language (cs.CL)
备注: 21 pages, 7 figures, 6 tables
Abstract:Speculative decoding accelerates autoregressive LLMs by having a lightweight drafter propose tokens that the target model verifies in parallel. Diffusion-based drafters further reduce drafting latency by proposing multiple tokens at once. However, these tokens are predicted independently, so a single early error causes prefix verification to discard the rest of the draft, even when it contains useful downstream predictions. We introduce DEdit, a diffusion-based drafter that can not only draft by conventional parallel unmasking but also iteratively edit its draft through token-to-token predictions. Through editing, later predictions can serve as bidirectional context for repairing earlier errors and extending the accepted prefix. To teach the model to repair errors while preserving correct predictions, we propose ProposalMix, a training scheme that mixes draft predictions with ground-truth tokens based on first-pass confidence during training. Across seven benchmarks on Qwen3-4B and Qwen3-8B, DEdit achieves the highest macro-average token acceptance and speedup among the evaluated drafters, reaching macro-average speedups of 5.72\times and 5.97\times over autoregressive generation under greedy decoding, respectively. Further analysis shows that acceptance improves with more editing passes and wider drafting windows, and that ProposalMix halves harmful edits that shorten the accepted prefix. Moreover, restricting the editor to causal attention lowers acceptance, especially on highly predictable outputs, indicating that future context is a key source of these gains.
[NLP-123] Personalized State-Transition-Aware Memory for Clinical Agents
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）代理在处理临床记录时，如何有效追踪患者状态变化并同时保留历史信息以支持临床决策解释的难题。现有方法要么简单累积记忆导致信息冗余与适用性模糊，要么过度覆盖早期记忆而破坏治疗史与临床轨迹的可追溯性。其解决方案的关键在于提出一种状态转移感知的记忆框架——STAM（State-Transition-Aware Memory），该框架在新临床数据到达时显式记录状态变迁，并通过语义检索与类型化临床关系联合识别受影响的记忆片段；具体实现上，将当前有效的信息存入“活跃”（Active）模块，已过时或已解决的信息归入“历史”（History）模块。在查询时，基于查询内容动态激活的门控机制可选择性地调用历史记忆，从而在保持上下文长度相近的前提下，显著提升纵向临床问答、状态维护诊断等任务的表现，在四个纵向临床基准测试中均验证了其有效性。
链接: https://arxiv.org/abs/2609.38490
作者: Maryam Haghifam,Zahra Rajabi,Yizhou Sun,Carlos Morato
机构: University of California, Los Angeles(加州大学洛杉矶分校); Optum AI, UnitedHealth Group(联合健康集团优普泰人工智能部门)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents that reason over clinical records must track changes in a patient’s state while preserving the history needed to understand them. Simply accumulating memories leaves it unclear which information still applies, whereas overwriting earlier memories can erase evidence needed to reconstruct treatment history and clinical trajectories. We introduce STAM, a state-transition-aware memory framework that records state changes as new clinical entries arrive. STAM combines semantic retrieval with typed clinical relations to identify affected memories, maintaining current information in Active and superseded or resolved information in History. At read time, a query-dependent gate selectively serves historical memory. Across four longitudinal clinical benchmarks, we evaluate STAM with downstream question answering, direct state-maintenance diagnostics, and comparisons at approximately matched context lengths.
[NLP-124] KlinikeBench: Evaluating Language Models Beyond Diagnostic Accuracy
【速读】： 该论文旨在解决现有临床语言模型（Language Model, LM）评估基准在真实临床场景中的适用性不足问题。传统评估仅基于完整病历描述进行诊断准确率测试，无法反映模型在动态交互中获取关键信息、制定检查计划及遵循临床决策流程的能力。此外，现有基准缺乏专业临床医生的验证，导致评估结果难以真实反映临床实践水平。为此，本文提出KlinikeBench，一个由333个由临床医生编写的任务构成的基准，每个任务提供独立沙盒环境，包含虚拟患者、临床工具和特定任务的成功标准，并有超过35名临床医生参与案例设计与评估。实验表明，相较于基于真实对话改编的参考对话，临床医生对模拟对话的质量评分更高。在每项任务中，模型需在有限轮次内与患者互动、询问病史、请求检查、遵守操作约束并作出最终诊断，各步骤得分单独评估且综合打分。跨31个模型与7个模型家族的测试显示，即使最佳模型（如GPT-6-astra和Claude Opus 5）诊断准确率高达90.7%，但在交互式临床评估中成功完成的任务比例仍低于30%。部分模型依赖与患者的交流，而另一些则仅凭完整病历表现良好，但在对话中显著退化。因此，KlinikeBench揭示了诊断准确率与实际交互式临床评估能力之间存在显著差距，为全面评估生成式人工智能（Generative AI）在临床全周期交互中的表现提供了重要测试平台。
链接: https://arxiv.org/abs/2609.38480
作者: Xueting Fang,Zehui Li,Yang Yang,Camilla Giovino,Shubh K. Patel,Shailly Prajapati,Vallijah Subasri,Caihua Shan
机构: Zhejiang University (浙江大学); Imperial College London (帝国理工学院); Nanchang University (南昌大学); University of Toronto (多伦多大学); Microsoft (微软)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Most clinical benchmarks evaluate language models (LMs) on diagnosis using complete case descriptions. In clinical practice, however, patients present information in different ways, and clinicians must obtain relevant history and determine which examinations are needed before reaching a diagnosis. Diagnostic accuracy alone therefore cannot establish whether an agent gathered essential information or conducted an appropriate clinical assessment. Furthermore, existing benchmarks lack professional clinicians’ verification. To address this gap, we introduce KlinikeBench, a benchmark of 333 clinician-authored tasks, each providing an isolated sandbox environment with a virtual patient, clinical tools, and task-specific success criteria. More than 35 clinicians contributed to case authoring and benchmark evaluation. In an empirical study, clinicians gave simulated dialogues higher mean quality ratings than reference conversations, which is adapted from real conversation. In each task, an LM has a fixed budget of turns to communicate with the patient, ask about relevant history, request examinations, follow action constraints, and record a final diagnosis. We score these steps separately as well as together. Across 31 models and seven model families, the best-performing models (e.g., GPT-6-astra and Claude Opus 5) succeed on less than 30% of tasks, even though their diagnosis accuracy reaches 90.7%. Some models benefit from talking with the patient; others diagnose well from a complete chart but perform much worse in conversation. Overall, KlinikeBench provides a testbed for evaluating the full clinical encounter and reveals a substantial gap between diagnostic accuracy and performance in interactive clinical assessment.
[NLP-125] he Backdrop Exposes What the World Around an Agent Costs It ICLR2027
【速读】： 该论文旨在解决当前智能体（Agent）评估体系与真实应用场景之间存在的显著脱节问题：现有基准测试通常在静态、可控的环境中进行，而实际部署中的智能体需在动态、受多方交互影响的开放世界中运行。例如，其他用户可能通过消息指令要求智能体转移资金或泄露敏感信息，此类外部干扰会严重威胁智能体的安全性与可靠性。为此，作者提出BACKDROP框架，其核心是系统性地引入四种现实世界中常见的非功能性风险（hazards），以检验智能体在复杂交互环境下的鲁棒性。这四类危害分别为：权限（Authority）——外部指令是否可覆盖用户意图；注入（Injection）——恶意文本能否篡改执行流程；边界（Boundary）——请求是否能诱使智能体进入未授权的应用程序；故障（Fault）——在写操作失败但无明确反馈时，智能体是否具备重试前的状态验证机制。实验覆盖3,678个任务变体和16种主流模型，结果显示，当所有四类危害同时存在时，平均通过率从清洁环境下的69.5%骤降至31.3%，最强模型（Claude Fable 5.1）亦从96.6%下降至56.0%。尤其值得注意的是，在仅考虑被成功传递的植入文本情况下，46.4%的运行中智能体仍遵循了他人指令，20.3%受到文本注入影响，暴露出智能体对非授权指令的高度脆弱性。因此，BACKDROP的关键贡献在于：首次形式化定义了智能体在理想环境中的性能上限与其在真实世界中实际表现之间的差距，并揭示了当前智能体在应对社会性攻击与环境不确定性方面存在系统性缺陷，强调必须将动态交互与外部干扰纳入评估体系。
链接: https://arxiv.org/abs/2609.38469
作者: Nusrat Jahan Lia,Shubhashis Roy Dipta
机构: University of Dhaka(达卡大学); University of Maryland, Baltimore County(马里兰大学巴尔的摩县分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Submitted to ICLR 2027
Abstract:Agent benchmarks test agents in worlds that stay still. Deployed agents work in worlds that other people also change. Someone texts the agent to send the money elsewhere or an order confirmation asks it to reply with a door code. We present BACKDROP, which asks how much of an agent’s capability in a clean world survives in such a world. BACKDROP takes a task along with the agents execution environment, and plants four everyday hazards in its world, one at a time and all together. The instruction and the correct end state stay the same. Each hazard asks one question. Authority: does a message from another person override the user? Injection: does text planted in a record redirect the agent? Boundary: does a request pull it into an app it was not given? Fault: after a write fails without saying whether it landed, does the agent check before it retries? Across 3,678 variants and 16 models, , the average pass rate falls from 69.5% to 31.3% once all four hazards are present; the strongest models fall furthest (Claude Fable 5.1 from 96.6% to 56.0%). Agents have learned to resist injected text but often follow other unauthorized requests of other people. With all four hazards present, and counting only runs where the planted text reached the agent, agents followed another person’s message in 46.4% of runs and injected text in 20.3%. The gap is consistent throughout all 16 models. BACKDROP formalizes these gaps and shows how an agent’s score in a task’s world is a ceiling on real-world performance.
[NLP-126] Reach Into The CHOIR: Free-List Elicitation Uncovers Distinct Model Voices in LLM Ensembles
【速读】： 该论文旨在解决生成式AI（Generative AI）在开放式任务中普遍存在的同质化问题，即多个模型看似提供多样化观点，实则返回高度相似的默认响应，导致虚假多样性。其核心挑战在于难以区分模型间的一致性是源于严格的输出空间约束、提示词词汇的回响效应，还是深层稳定替代方案的存在。解决方案的关键是提出CHOIR（Collective Hierarchically-Ordered Inquiry Responses）框架，该框架借鉴认知人类学中的自由列举法，通过迭代生成带排序的响应列表，将响应聚类为提示层面的概念，并量化不同模型、提示变体及角色设定下概念的显著性。该方法能够揭示表面高一致性下的深层差异，识别出基础模型身份作为最强可恢复特征，并发现角色提示仅在基础模型签名内调整表层概念。此外，无源排名模块优先筛选稀有但稳定的候选响应以供后续分析。通过结构化深度探测，CHOIR将开放性同质化问题转化为可诊断的测量难题，明确回答了模型在何处收敛、为何收敛以及在何种条件下仍可触及潜在多样性。
链接: https://arxiv.org/abs/2609.38448
作者: Ben Wigler,Maria Tsfasman
机构: LoveMind AI(爱心智人工智能), New York, USA
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 23 pages, 6 figures. Published in the Proceedings of the Third Conference on Language Modeling (COLM 2026)
Abstract:Open-ended LLM homogeneity can create false plurality when several systems appear to offer independent perspectives while returning the same familiar default. Single-pass answers obscure the distinction between agreement produced by a tightly constrained answer space, prompt-vocabulary echo, and broader answer spaces with stable alternatives beneath the surface. We introduce CHOIR (Collective Hierarchically-Ordered Inquiry Responses), a framework that adapts free-list elicitation from cognitive anthropology to LLM ensembles. CHOIR repeatedly elicits ranked lists, clusters items into prompt-level concepts, and measures concept salience across models, prompt variants, and persona conditions. We evaluate CHOIR on Infinity-Chat 100, an external prompt bank from recent work on open-ended model homogeneity, and on a 27-question targeted diagnostic bank designed to isolate mechanism-level contrasts. On Infinity-Chat 100, CHOIR reproduces high surface agreement (93/100 prompts above chance) while separating narrow prompts from broad prompts with recoverable depth. Across targeted probes and the external prompt bank, base-model identity remains the strongest recoverable signature, and persona prompts shift surfaced concepts within base-model signatures. A source-blind ranking module prioritises rare-but-stable candidates for later inspection. CHOIR turns open-ended homogeneity into a diagnostic measurement problem by asking where models converge, why they converge, and what remains reachable under structured depth probing.
[NLP-127] What Pretraining and Midtraining Make Learnable from Rewards?
【速读】： 该论文旨在解决在生成式人工智能（Generative AI）模型中，如何通过奖励机制有效引导模型完成任务，同时保持对新输入的计算灵活性这一核心问题。其关键在于揭示预训练（pretraining）与中期训练（midtraining）阶段所提供的信息与计算资源如何共同支持奖励适应（reward adaptation）的有效性。研究提出，在顺序状态计算与上下文记忆框架下，尽管不同机制在训练奖励上达成一致，但对未见输入的输出预测存在差异；而任务无关的源观测（task-independent source observations）能够消除这种歧义。通过从指定随机初始化出发，构建有限采样的Adam路径，实现源预测与奖励适应在同一参数空间内的联合优化，从而证明了预测能力如何转化为可执行的推理或检索行为，以及奖励如何学习其特定任务的应用方式。实验基于预训练的Qwen2.5检查点，在八个不同环境中验证了该分工机制：使用正确源信息和首操作监督的序列模型达到82.61%的成功率，显著优于私有随机源对照组的44.15%；记忆回放（memory replay）在奖励适应过程中保留了检索能力，独立八世界验证进一步表明，该方法实现75.32%的任务成功率，远超匹配的替代性检索训练组的49.86%。在GSM8K与HotpotQA数据集上的分析显示，准确率在奖励引入时、后续提升阶段及最终性能之间存在明显分离，共同揭示了信息获取、可执行计算与奖励引导的任务学习之间的内在联系。
链接: https://arxiv.org/abs/2609.38446
作者: Chiwun Yang,Xiaoyu Li
机构: City University of Hong Kong (香港城市大学); University of New South Wales (新南威尔士大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 160 pages, 26 figures, 32 tables
Abstract:A reward can identify a correct answer while leaving the computation needed for new inputs undetermined. We study how pretraining and midtraining supply the information and computation that make reward adaptation effective. In sequential state computation and contextual memory, we characterize mechanisms that agree on every training reward yet demand different held-out answers. Task-independent source observations resolve this ambiguity. We construct finite sampled Adam paths from specified random initializations through source prediction and reward adaptation in the same parameters, proving how prediction acquires execution or retrieval and rewards learn their task-specific use. Experiments with pretrained Qwen2.5 checkpoints test this division of labor. Across eight worlds, Sequential models trained with correct source and first-operation supervision reach 82.61% success, versus 44.15% for a private-random source control. Memory replay preserves retrieval during reward adaptation, and an independent eight-world confirmation achieves 75.32% task success versus 49.86% after matched alternative-retrieval training. GSM8K and HotpotQA separate accuracy at reward entry, subsequent gain and final performance. Together, these results connect information acquisition, executable computation and reward-guided task learning.
[NLP-128] Policy-Conditioned AI-Use Detection: An Evidentiary Framework for Academic Publishing
【速读】： 该论文旨在解决当前学术会议与期刊在规范作者、审稿人及领域主席使用生成式AI（Generative AI）时所面临的政策执行难题。现有方法依赖于AI检测技术，其目标是判断文本是否由AI模型生成，但这一目标与实际政策评估需求存在根本性错位——会议和期刊真正关心的是人类-人工智能协作流程是否符合既定规则，而非单纯识别文本来源。为此，论文提出“政策条件化AI使用检测”（policy-conditioned AI-use detection）这一证据框架，将具体政策要求作为显式输入，通过推断生成假设、证据、校准机制及不确定性量化，替代简单的“检测到AI”等二元判决。该框架强调可复现的基准构建，基于可复制的工作流管道生成合规与违规案例，并在预设的误报率下报告真阳性率。以同行评审为例，研究发现即使在合理违规率下，现有强性能检测器仍可能错误标记更多合规作者。因此，该框架进一步明确了会议必须部署的技术基础设施：结构化披露机制、尊重审稿人保密性的授权工具路由系统，以及可申诉的异议处理路径。在此范式下，检测器不再是不可解释的作者归属分类器，而是一种可审计的程序，其错误率由会议预先设定并可辩护，从而实现政策执行的透明性与可问责性。
链接: https://arxiv.org/abs/2609.38427
作者: Jairo Diaz-Rodriguez,Mumin Jia
机构: York University (约克大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注:
Abstract:Major venues now publish detailed rules about how authors, reviewers, and area chairs may use AI, and those rules differ by role, by task, and by what must be disclosed. AI detection, the instrument usually proposed to enforce them, estimates something else: whether an AI model wrote the text. We argue that this target is misaligned with the decisions conferences and journals face, and propose policy-conditioned AI-use detection, an evidentiary framework for assessing whether a human–AI workflow complied with a stated rule. Policy makes the governing rule an explicit input. Inference reports hypotheses, evidence, calibration regime, and uncertainty in place of verdicts such as “AI detected”. Evaluation builds benchmarks from reproducible pipelines that generate compliant and non-compliant workflows, and reports true positive rate at a false positive rate the venue fixes in advance. We work the framework through peer review, where at plausible violation rates a detector at a strong operating point still flags more compliant authors than violating ones. The framework therefore also names what a venue must instrument: structured disclosure, approved-tool routing that respects reviewer confidentiality, and a path by which a finding can be contested. Under this framing a detector is not an authorship classifier but an auditable procedure with an error rate the venue fixes in advance and can defend.
[NLP-129] LoopVL: Recurrent Visual Intelligence
【速读】： 该论文旨在解决如何有效将循环式变换器（Loop Transformer）扩展至视觉-语言模型中的问题，以提升多模态理解与视觉推理能力。其核心解决方案在于提出LoopVL框架，通过结合模块级循环（Module-Loop）与模型级循环（Model-Loop）计算机制，利用共享模块对统一的视觉-语言状态进行迭代更新。该方法在从零开始的语言预训练、多模态训练及后训练流程中实现端到端优化，显著优于同等规模乃至更大规模的非循环模型。此外，研究观察到LoopVL中存在“视觉顿悟时刻”（Visual Aha Moments），表现为跨循环过程中视觉注意力的显著跃迁，为循环式多模态建模提供了实证支持，并揭示了共享参数如何促进在持续演化的视觉-语言状态上实现更深层次的多模态计算。
链接: https://arxiv.org/abs/2609.38426
作者: Zhe Qian,Ziyang Gong,Zhongxing Xu,Hehan Li,Zhonghua Wang,Fei Luo,Mingxuan Wang,Xue Yang,Shiwei liu,Yanbiao Ma,Junchi Yan,Jungong Han
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注:
Abstract:We introduce LoopVL to study whether Loop Transformers can be effectively extended to vision- language models. LoopVL combines Module-Loop and Model-Loop computation to iteratively update a unified vision-language state through shared modules. We train LoopVL from scratch through language pre-training, multimodal training, and post-training. LoopVL outperforms a range of similarly sized and larger non-recurrent models on multimodal understanding and visual reasoning benchmarks. We also observe Visual Aha Moments in LoopVL, characterized by pronounced shifts in visual attention across loops. LoopVL provides practical evidence for recurrent vision-language modeling and offers an intuitive perspective on how shared parameters can support deeper multimodal computation over continuously evolving visual-language states.
[NLP-130] What Was Said Not What Was Thought: Type-6 Logic for CoT Verification
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在链式思维（Chain-of-Thought, CoT）推理过程中普遍存在且难以检测的结构性谬误问题，如未经许可的推理修正、省略前提（enthymemes）、循环回溯（loopbacks）以及不可验证或错误的断言。其核心解决方案是提出一种名为Type-6逻辑的动态认知逻辑变体，通过引入“不确定性”与“重复性”两个新算子，系统建模LLM CoT推理的推演动态过程。该方法的关键在于构建一个基于Type-6逻辑的验证器，能够将模型生成的推理轨迹转化为有向图结构，并依据其公理体系和推理规则进行形式化校验，从而识别出表面启发式方法难以捕捉的结构性推理缺陷。实验表明，该验证器在涵盖正式与非正式推理任务的四个数据集上均能有效发现推理路径中的不一致与矛盾，且揭示出“推导出矛盾”是CoT中最常见的硬性失败类型，仅有约3%的推理步骤对最终结论具有实质影响。此外，消融实验显示，现有验证方法（如以LLM为裁判、其他神经符号方法等）之间缺乏可互换性，而Type-6逻辑在一致性上表现最优，且其验证算法具有平均情况下的线性时间复杂度，具备高效可扩展性。研究已公开逻辑规范与相关实现资源。
链接: https://arxiv.org/abs/2609.38420
作者: Adrian de Wynter
机构: Microsoft(微软); University of York(约克大学)
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:We introduce Type-6 logic, a variant of dynamic epistemic logic augmented with two operators (uncertainty and recurrence), designed to model the inferential dynamics of contemporary large language model (LLM) chain-of-thought (CoT) reasoning. Type-6 accounts for common LLM reasoning pathologies such as unlicensed revision, enthymemes, loopbacks, and unverifiable/incorrect claims. We propose a verifier based on Type-6 logic that builds a graph out the trace, and checks it against Type-6’s axioms and inference rules. We evaluate our framework on LLM-generated CoTs four splits spanning formal and informal reasoning. Our verifier detects structurally unsound reasoning steps that surface-level heuristics miss, and allows for easy visualisation of the model’s reasoning process. In our corpus, our verifier shows that derived contradiction is the most common hard-fail category in CoT, and that only about 3% of the propositions of a trace have impact on the final derivation. Ablation studies show that other verification methods (LLMs-as-judges, other neurosymbolic approaches, etc.) cannot be considered interchangeable: for example, agreement between LLMs-as-judges and LINC is \kappa \approx 0.034 , and this persists within a method across underlying models. Type-6, however, is the most agreed-with method amongst the ones we tested. We prove our verifier runs on average-case linear time; and release our logic specification and artefacts.
[NLP-131] ArgGYM: A Procedural Engine-Verified Benchmark for Structured Defeasible Reasoning
【速读】： 该论文旨在解决当前大型语言模型在受限领域（如数学、编程和形式逻辑）中表现出色，但其推理能力能否有效迁移至现实世界中更具动态性和不确定性的复杂推理场景这一关键问题。现实世界的推理通常涉及不完备与可修正的信息，结论可能在证据变化下被推翻、重新支持或修订，这种推理模式被称为可废止推理（Defeasible Reasoning）。为应对这一挑战，论文提出ArgGYM——一个面向结构化可废止推理的程序化基准测试平台及与强化学习与验证奖励（RLVR）兼容的训练环境。其核心解决方案在于：将可废止推理分解为12个具体任务，并基于符号化论证引擎对模型输出进行形式化状态评估，实现任务特定的评分；同时提供包含1,440个经验证实例的静态基准集，涵盖十五种课程配置、两种论证偏好顺序（最弱环节与最后环节）以及两种集合排序方式（精英主义与民主主义），并支持通过相同生成器与验证器动态生成新实例，从而降低对静态测试集的依赖，促进可验证奖励的持续训练。实验表明，前沿模型与开放权重模型在推理模式上存在显著差异，能够在不完成完整任务的情况下恢复部分结构化答案，但在具有更长依赖关系和更多交互结构的后期课程配置中性能明显下降，揭示了现有模型在处理复杂可废止推理时的局限性。
链接: https://arxiv.org/abs/2609.38409
作者: İbrahim Ethem Deveci,Funda Tan Çalık,Barış Deniz Sağlam,Duygu Ataman
机构: Middle East Technical University (中东技术大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 40 Pages, 16 Tables
Abstract:Recent progress in large language model reasoning has been driven by benchmarks and reinforcement learning environments with automatically verifiable rewards, particularly in mathematics, code, and formal logic. These settings make model accuracy easier to evaluate and optimize, but it remains unclear how far success under fixed problem specifications and stable evaluation criteria transfers to reasoning outside such domains. Real-world reasoning often proceeds under incomplete and revisable information: conclusions may be supported provisionally, defeated by counter-evidence, reinstated by further arguments, or revised when stronger reasons become available. Reasoning of this kind is generally referred to as defeasible reasoning. We introduce ArgGYM, a procedural benchmark and RLVR-compatible training environment for structured defeasible reasoning. ArgGYM decomposes this reasoning into twelve tasks and grounds task-specific scoring in a symbolic argumentation engine that computes the formal states used to evaluate model outputs. It includes a frozen benchmark of 1,440 verified instances across fifteen curriculum configurations, two argument preference orderings (weakest-link and last-link), and two set orderings (elitist and democratic), while the same generators and verifiers can produce fresh instances for evaluation that reduces dependence on static test sets and for verifiable-reward training. On the frozen benchmark, frontier and open-weight models show sharply different reasoning profiles: they can recover substantial parts of structured answers without solving the complete task, and performance declines in later curriculum configurations with longer dependencies and more interacting structures. We release the benchmark, generators, and verifiers for reproducible evaluation and RLVR training.
[NLP-132] Evaluating Whether LLM s Can Reliably Connect the DOTs?
【速读】： 该论文旨在解决真实世界叙事文本中因信息噪声与碎片化导致的连贯性构建难题，核心问题在于如何在保持局部语境一致性与全局故事线连贯性的前提下，对缺失句段进行准确补全（即文本补全，text infilling）。其解决方案的关键在于构建了一个涵盖四大叙事类型（百科文本、常识故事、新闻报道、视觉叙事）的多领域基准数据集，共约9.2K个实例，通过掩码单至三句话的方式模拟真实场景中的信息缺失。基于该基准，研究系统评估了20个指令微调的开源大模型（参数量从1.5B至70B不等），并采用自动指标与包含五个叙事维度的定性评估框架进行综合分析。研究发现，模型规模并非决定补全质量的可靠因素：小型模型Gemma-2-2B在定性评分上达到4.02/5，显著优于多个超十倍更大的模型；此外，显式推理（如思维链，chain-of-thought）带来的提升有限（仅+0.6%），而短篇叙事和领域特性相较于填空位置更能有效预测任务难度。这表明当前大语言模型在叙事补全任务中更依赖上下文理解与领域适配能力，而非单纯参数规模或复杂推理机制。
链接: https://arxiv.org/abs/2609.38406
作者: Eftekhar Hossain,John Salvador,Santu Karmaker
机构: Bridge-AI Lab@UCF(桥-人工智能实验室@中佛罗里达大学); University of Central Florida (中佛罗里达大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Access to real-world information is often noisy and fragmented. Constructing a coherent narrative from such fragments requires models to reconstruct missing spans within a broader storyline, commonly referred to as text infilling, while preserving consistency with both the local context and the global storyline. Despite using text infilling as a pre-training objective in many Large Language Models (LLMs), their actual performance on real-world narrative infilling remains underexplored. In this paper, we address this gap by introducing a multi-domain benchmark of ~9.2K instances for narrative infilling, constructed by masking one to three sentences across four narrative types: encyclopedic text, commonsense stories, news articles, and visual narratives. Using this benchmark, we evaluate 20 instruction-tuned open-source LLMs ranging from 1.5B to 70B parameters across varying levels of instruction specificity and reasoning guidance. Outputs are assessed using standard automatic metrics and a qualitative framework covering five narrative dimensions. Results show that model scale does not reliably predict infilling quality: Gemma-2-2B achieves the highest qualitative score (4.02/5), outperforming models over ten times larger, including DeepSeek-Qwen-32B (3.77/5, 6.6%) and LLaMA-3.3-70B (3.71/5, 8.3%). We further find that explicit reasoning offers limited benefits as chain-of-thought reasoning yields only a marginal improvement (+0.6%). Additionally, short narratives and domain characteristics emerge as stronger predictors of task difficulty than infill position alone for narrative infilling in current LLMs.
[NLP-133] he Geometry of Harmfulness in Multi-Turn Attacks
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在多轮对抗攻击中安全性对齐失效的问题，特别是探究有害性（harmfulness）与拒绝响应（refusal）表征在多轮对话过程中的几何结构与时间动态演化机制。其核心发现表明：尽管不同攻击框架沿不同的几何方向推进，但均能有效诱导出有害输出；在中后期模型层的句末标记位置，有害性表征随对话轮次逐渐呈现线性可分性；而有害性表征与拒绝相关表征之间仅存在弱关联。这一现象揭示，多轮攻击并非通过抑制模型内部有害性表征来实现，而是使有害性表征在时间维度上逐步变得可区分且独立于拒绝机制。因此，静态的单轮安全检测方法在多轮场景下易失效，其根本原因在于未能捕捉到有害性表征的时序演化特性。该研究的关键解决方案在于强调：构建鲁棒防御机制必须考虑表征在时间序列上的动态演变，而非依赖孤立或单轮提示下的静态判断。
链接: https://arxiv.org/abs/2609.38389
作者: Yelyzaveta(Lisa)Husieva,Lauren Alvarez
机构: Lauren Alvarez; Applied AI Research; TELUS Digital; Charlottesville, VA 22902
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL)
备注: 9 pages, 7 figures, 1 Table, preprint
Abstract:Large language models (LLMs) remain vulnerable to adversarial attacks that circumvent safety alignment to elicit harmful outputs. It remains unclear how harmfulness and refusal representations evolve over the course of multi-turn attacks, and why single-turn defenses are less effective in multi-turn settings. This work investigates how the geometry and temporal dynamics of harmfulness and refusal representations evolve across multi-turn attacks. We analyzed hidden-state representations from three instruction-tuned LLMs (Llama-3.1-8B-Instruct, Qwen2.5-7B-Instruct, and Gemma-2-9B-it) using three multi-turn attack frameworks (Crescendo, ActorAttack, and X-Teaming), and examined representation behavior across conversation turns, model layers, and token positions under various context configurations. Across models and frameworks, we found that (1) each attack framework traverses different geometric directions, yet each achieves comparable success in eliciting harmful outputs; (2) multi-turn harmfulness directions became increasingly linearly separable at the end-of-turn token position across turns in middle to late model layers; and (3) harmfulness representations are weakly aligned with refusal-related representations. The results indicate that multi-turn attacks do not succeed by suppressing the model’s internal representation of harmfulness. Instead, harmfulness representations become increasingly separable across conversation turns, while remaining only weakly aligned with refusal-related representations. The findings are one possible explanation for why static single-turn safety probes may degrade in multi-turn settings, and suggest that robust defenses must consider temporal representation dynamics rather than identifying harmfulness with isolated or single-turn prompts.
[NLP-134] Fine-Tuning Diffusion Language Models with Context Selection and Target Weighting
【速读】： 该论文旨在解决离散扩散语言模型（discrete diffusion language models）在监督微调过程中，采用均匀随机掩码（uniform random masking）导致上下文信息与预测目标之间交互关系未被显式优化的问题。传统方法中，掩码模式同时决定了模型可访问的上下文和需预测的目标，但均匀随机掩码未能充分考虑二者之间的权衡。为此，本文提出GoldiMask，其核心在于通过近似最大化一个子模（submodular）目标函数来选择应揭示为上下文的令牌，该目标利用模型信号平衡揭示令牌所带来的上下文增益与其作为预测目标的价值。在此基础上，GoldiMask进一步对剩余待预测目标进行加权，以反映其从所选上下文中获益的程度及剩余的学习潜力。实验结果表明，在三种骨干网络和三个训练数据集上，GoldiMask在多数设置下均取得最高平均准确率，尤其在推理与代码生成任务中表现显著提升；消融实验验证了上下文选择与目标加权两个模块均对性能增益有贡献。此外，GoldiMask在GSM8K和MATH-500数据集上结合置信度阈值并行解码时，可减少解码迭代次数，同时在高置信度阈值下保持相近的准确性，展现出更强的效率优势。
链接: https://arxiv.org/abs/2609.38385
作者: Loay Mualem,Lluís Pastor-Pérez,Vinh Tong,Andrei Manolache,Tanja Bien,Steffen Staab,Mathias Niepert
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 30 pages, 4 figures, 14 tables. Main text 10 pages, references and appendix follow. Project page with interactive visualizations: this https URL
Abstract:Supervised fine-tuning of discrete diffusion language models masks some response tokens and trains the model to recover their original values from the visible context. The masking pattern therefore determines both the context available to the model and the tokens it learns to predict. Uniform random masking does not explicitly account for the interaction between these choices. We introduce GoldiMask, which selects tokens to reveal as context by approximately maximizing a submodular objective. This objective uses model signals to balance the benefit of revealing tokens against their value as prediction targets. GoldiMask then weights the remaining targets according to how they benefit from the selected context and their remaining learning potential. Across three backbones and three training datasets, GoldiMask achieves the highest average accuracy in most evaluated settings, demonstrating gains on both reasoning and code generation. Component ablations show that both context selection and target weighting contribute to the gains. GoldiMask also reduces decoding iterations on GSM8K and MATH-500 under confidence-threshold parallel decoding, while maintaining comparable accuracy at higher confidence thresholds.
[NLP-135] ALK-Dem: Benchmarking Embodied Task Planning under Dementia-Associated Communication Patterns
【速读】： 该论文旨在解决当前大语言模型（LLM）驱动的机器人任务规划系统在面对真实用户（尤其是患有痴呆症的人群，PLWD）时所面临的沟通鲁棒性不足问题。现有系统依赖于理想化用户指令假设，即指令清晰、完整且聚焦任务，但在实际交互中，痴呆症患者常表现出指代不精确、对象替换、空话、话题漂移和无关插入等典型言语特征，导致规划错误甚至引发物理安全风险。为此，论文提出了首个针对痴呆相关言语通信的基准测试数据集TALK-Dem，包含4,800条指令，涵盖五类典型沟通模式及三档强度等级，用于系统评估。实验发现，六种开源大语言模型在非理想指令下的性能下降最高达22.3个百分点，暴露出显著的鲁棒性缺陷，尤其在隐私敏感与连接受限场景下亟需本地部署模型。为应对这一挑战，论文提出上下文感知的经验检索（CARE）方法，通过检索先前已解决的任务以提供任务特定的语义解释与规划上下文，有效缓解了指令模糊性。CARE在所有六种模型上均优于标准提示基线，平均任务成功率提升18.1个百分点，验证了其在提升本地可部署助人机器人可靠性方面的关键作用。
链接: https://arxiv.org/abs/2609.38371
作者: Guangxin Zhao,Yiran Hu,Yuan Cao,Chenxi Jiang,Jianfei Yang,Yegang Du,Yasuyuki Taki,Yoshifumi Kitamura,Lin Gu,Zhi Zheng
机构: University of Notre Dame(圣母大学); Nanyang Technological University(南洋理工大学); Tohoku University(东北大学)
类目: Robotics (cs.RO); Computation and Language (cs.CL)
备注:
Abstract:Existing LLM-driven robot task planners rely on a taken-for-granted assumption of an ideal user whose instructions are clear, complete, and task-focused. However, when interacting with real-world users, especially those experiencing cognitive impairments, such as people living with dementia (PLWD), the planners often make mistakes and even pose physical safety risks. We proposed TALK-Dem (Talking Attributes and Linguistic Knowledge in Dementia), the first benchmark for evaluating LLM-driven robot task planning under dementia-associated verbal communication. TALK-Dem contains 4,800 instructions and covers five typical communication patterns, including Referential Imprecision, Object Substitution, Empty Speech, Topic Drift, and Intrusion, at three intensity levels. Experiments across six open-weight LLMs reveal a substantial robustness gap. Across communication patterns, open-weight models exhibited performance drops of up to 22.3 percentage points compared to ideal instructions. This revealed a critical gap and even danger for real-world applications, especially in assistive robotics, where locally deployable models are necessary due to privacy concerns and connectivity constraints. To mitigate this issue, we proposed the Context-Aware Retrieval from Experience (CARE) method, which retrieves relevant previously resolved tasks to provide task-specific interpretation and planning context. CARE generally outperformed standard prompting baselines across the six open-weight models, improving average task success by 18.1 percentage points over the vanilla prompt. These results highlighted the importance of both evaluating communication robustness and developing effective adaptation strategies for locally deployable assistive robots. The TALK-Dem dataset is publicly available at this https URL.
[NLP-136] On the Off-Policy Teacher in On-Policy Distillation
【速读】： 该论文旨在解决在线策略蒸馏（On-Policy Distillation, OPD）中教师模型与学生模型之间的策略不匹配问题。在OPD框架下，学生模型从自身策略生成的轨迹中学习，而教师模型需对这些由学生生成的前缀（prefixes）进行延续，但传统方法中教师模型仍沿用自身策略进行优化，导致其在面对学生生成的非自身策略前缀时表现退化，尤其在前缀长度增加时性能显著下降。解决方案的关键在于提出学生条件化的教师更新机制（Student-COnditioned Updates of the Teacher, SCOUT），即通过引入一种联合训练框架，在标准的OPD更新之外，周期性地利用可验证奖励信号（verifiable rewards）对教师模型进行强化学习训练，使其学会从学生生成的前缀出发生成高质量的后续动作序列。该机制使教师模型具备适应学生策略的能力，从而有效缓解了因策略不对齐带来的性能损失，并在多种模型规模和推理任务场景下持续提升OPD的整体效能。
链接: https://arxiv.org/abs/2609.38360
作者: Langlin Huang,Hao Liu,Mononito Goswami,Xinyu Li,Prithwith Jana,Nikos Kanakaris,Patrick Blöbaum,Purak Jain
机构: Washington University in St. Louis (圣路易斯华盛顿大学); AWS AI Labs (亚马逊云科技人工智能实验室); Carnegie Mellon University (卡内基梅隆大学); Georgia Institute of Technology (佐治亚理工学院)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:On-policy distillation (OPD) has recently emerged as a promising post-training paradigm in which the student learns from trajectories generated by its own policy under dense teacher supervision. However, OPD introduces a fundamental asymmetry: although the sampled trajectories are on-policy for the student, they are off-policy for the teacher. The teacher is typically optimized to continue from prefixes generated by its own policy, but during OPD it must instead supervise prefixes generated by the student. Empirically, we find that its continuation performance degrades as these prefixes grow longer. To address this issue, we propose Student-COnditioned Updates of the Teacher (SCOUT), a co-training framework that adapts the teacher to student-generated prefixes. Alongside standard OPD updates, SCOUT periodically optimizes the teacher’s conditional ability using reinforcement learning with verifiable rewards, where the teacher generates continuations from student prefixes and learns from outcome rewards. Controlled experiments show that SCOUT improves the teacher’s ability to continue from student-generated prefixes, supporting the intended mechanism of student-conditioned teacher adaptation. Across multiple teacher–student configurations, model scales, and reasoning domains, SCOUT also consistently improves the effectiveness of on-policy distillation.
[NLP-137] Beyond Mode Collapse: Generating Diverse Synthetic Expert Conversations via Generative Flow Networks
【速读】： 该论文旨在解决大语言模型（LLM）在后训练阶段生成高质量合成数据时面临的多样性不足问题，即直接提示或基于应用场景条件化生成的对话数据容易陷入主导模式（dominant modes），导致缺乏对专家策略和决策多样性的充分表征。其核心解决方案是采用生成式流网络（Generative Flow Networks, GFlowNets），通过在关键交互特征（如困惑事件动态、支架指令平衡等）上构建高斯混合密度模型来学习潜在对话结构，从而实现对专家策略按其在训练数据中实际出现频率的比例进行采样。该方法在辅导与情感支持两类结构迥异的对话领域中均表现出更优的保真度、模式覆盖范围与真实性，且不依赖于对训练数据的直接复制。在三个下游结果预测任务上的评估表明，基于GFlowNet生成的合成对话数据能够为分类器提供比强化学习及端到端LLM基线更强的训练信号。
链接: https://arxiv.org/abs/2609.38359
作者: Sumit Asthana,Michael Ion,Kevyn Collins Thompson
机构: University of Michigan(密歇根大学); Ann Arbor, MI, USA(美国密歇根州安娜堡)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:High quality synthetic data is central to post training LLMs for adaptive AI applications that represent the diverse expert strategies and decisions in conversations. Prompting LLMs directly or conditioning them on end use scenarios yields low diversity data that collapses onto dominant modes. We propose a method to generate diverse high quality synthetic data using Generative Flow Networks (GFlowNets). We show that training GFlowNets to generate latent conversation structure using a Gaussian mixture density over key interaction features (e.g., confusion episode dynamics, scaffolding directive balance) enables sampling expert strategies in proportion to their prevalence in the training data. Across two structurally distinct domains, tutoring and emotional support dialogues, our GFlow based synthetic data generation approach offers a better balance of fidelity, mode coverage and authenticity than reinforcement-learning and end to end LLM baselines, without copying training data. Evaluated on three downstream outcome prediction tasks, classifiers trained on synthetic GFlowNet generated conversations provide a stronger training signal than competitive synthesis baselines.
[NLP-138] Evaluating Language Model Safety Across Long Adversarial Conversations
【速读】： 该论文旨在解决当前对话安全评估中普遍存在的局限性问题：现有评估多基于单轮有害提示测试语言模型的安全性，而忽视了真实应用场景中用户与系统之间长期、动态且具有对抗性的多轮交互特性。其核心问题是，即使模型在首轮响应中表现出较高的安全性，是否仍能在持续的对抗性对话中保持安全输出。研究的关键解决方案在于构建一个模拟真实对抗场景的多轮评估框架，即让一个第二语言模型作为持续存在的对抗性用户，在不同对话深度（从1到101轮）下不断施压，同时通过安全分类器对每一轮响应进行标注。实验结果表明，尽管初始轮次的安全响应率高达85%–100%，但随着对话轮次增加，安全响应率显著下降至38%–61%（第11轮）和15%–44%（第101轮），且这一趋势在多个模型和不同随机种子下均一致出现。该发现揭示了单一回合的安全性无法保证长期对话中的安全性，强调了开展长时程评估以及引入对话层面的风险累积防护机制的重要性。
链接: https://arxiv.org/abs/2609.38357
作者: Parisa Salmani,Peter R. Lewis
机构: Ontario Tech University ( Ontario Tech University)
类目: Computation and Language (cs.CL)
备注:
Abstract:Conversational safety evaluations often test language models with a single harmful prompt, even though real-world systems interact with users through long, adaptive conversations. This study examines whether models continue to respond safely when an adversarial user persists across multiple turns. We evaluate three open-weight, instruction-tuned models on two harmful prompts across different conversation lengths and random seeds. In each setting, a second language model acts as a persistent adversarial user, while a safety classifier labels every response as safe or unsafe. Across all model-prompt combinations, first-turn safe-response rates ranged from 85% to 100%. By depth 11, they dropped to 38-61%, and by depth 101, to 15-44%. This decline appeared across models and continued well beyond the short interactions typically used in multi-turn safety evaluations. These results provide proof-of-concept evidence that strong single-turn safety does not necessarily persist during sustained adversarial interaction. They highlight the need for long-horizon evaluations and conversation-level safeguards that account for risk accumulating across turns.
[NLP-139] Halluscoring 2026: The first shared task on llm s hallucination detection and answer verification
【速读】： 该论文旨在解决阿拉伯语问答系统中幻觉检测与事实验证在复杂泛化场景下的评估难题，尤其关注模型在面对未见问题（unseen questions）和由未见过的大语言模型（LLM）生成的回答时的鲁棒性。其核心挑战在于如何有效识别生成内容中的事实错误，并在分布外（distribution shift）条件下保持性能。解决方案的关键在于构建一个包含四个子任务的共享任务框架（HalluScoring 2026），基于两个阿拉伯语数据集——HalluScore 和 HalluTruthQA，分别评估二分类幻觉检测（任务1）以及在六选一候选答案中准确识别正确事实答案的能力（任务2）。其中，任务2进一步细分为伊斯兰知识与通用知识两类，强化了对跨领域事实准确性判断的要求。实验结果表明，即使在辅助评估条件下，顶尖团队在任务2中的表现也仅达到0.882和0.857的AUC-ROC分数，凸显了当前系统在真实世界泛化能力上的局限性，反映出幻觉检测与事实验证仍面临严峻挑战。
链接: https://arxiv.org/abs/2609.38355
作者: Aisha Alansari,Abdessalam Bouchekif,Ahmed Hasanaath,Salah Eddine Bekhouche,Malak Alkhorasani,Mohammed-En-Nadhir Zighem,Saad Ezzini,Hichem Telli,Hend Al-Khalifa,Muhammad Abdul-Mageed,Hadid Abdenour,Hamzah Luqman
机构: King Fahd University of Petroleum and Minerals (沙特国王法赫德石油和矿产大学); University of Biskra (阿尔及利亚比斯克拉大学); Hamad Bin Khalifa University (卡塔尔哈马德·本·哈利法大学); University of British Columbia (加拿大不列颠哥伦比亚大学); University of the Basque Country (西班牙巴斯克大学); Universiti Malaysia Kelantan (马来西亚吉打州大学); Imam Abdulrahman bin Faisal University (沙特阿卜杜拉曼·本·法伊塞尔大学); King Saud University (沙特费萨尔国王大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:We present HalluScoring 2026, a shared task for evaluating hallucination detection and factual verification in Arabic question answering under challenging generalization settings. The shared task is organized into two main tasks, each comprising two subtasks, for a total of four subtasks. Task 1 evaluates binary hallucination detection, considering generalization to unseen questions (Subtask 1.1) and responses generated by unseen LLMs (Subtask 1.2). Task 2 extends the evaluation beyond detection by requiring the systems to additionally identify the correct factual answer from six related candidates, covering Islamic knowledge (Subtask 2.1) and general knowledge (Subtask 2.2). The shared task is based on two Arabic datasets: HalluScore and HalluTruthQA. A total of 13 teams participated in the shared task, 10 of which submitted system description papers. The results of Task 1 demonstrate that hallucination detection remains challenging under distribution shift, with the winning team achieving AUC-ROC test scores of 0.772 and 0.767 for Subtasks 1.1 and 1.2, respectively. For Task 2, the winning team achieved scores of 0.882 and 0.857 in the Islamic and general-knowledge subtasks, respectively, under assisted evaluation.
[NLP-140] OpenCollab: A Multi-Agent Coding Framework with Programmable Collaboration and Controllable Runtime
【速读】： 该论文旨在解决多智能体代码生成系统在实际协作中存在“行为差距”（behavioral gap）的问题，即现有评估往往假设预设的组织结构被严格遵循，而现实中智能体的实际行为与设计意图常存在偏差。这种偏差连同底层系统组件的差异，导致难以准确归因于特定配置所带来的性能提升。其解决方案的关键在于提出OpenCollab——一个统一的多智能体代码协作框架，通过统一的组织设计、共享运行时环境下的可控制实验条件，以及细粒度事件流的执行追踪，实现对协作过程的精确观测与调控。在此基础上，论文引入“遵从度”（Adherence）作为量化指标，衡量声明的组织结构是否真实实现。实验表明，不同配置下智能体协作行为差异显著，遵从度在47.2%至97.2%之间波动，揭示了组织设计对协作效果的关键影响。此外，基于OpenCollab构建的双智能体工作流在多个基准测试中达到新的最优性能（SOTA），超越主流工具如Mini-SWE-agent、Codex CLI和Claude Code；同时，其单智能体配置在所有评测套件中使用最少的令牌数，验证了其高效性。因此，OpenCollab不仅提供了可编程协作的统一基础设施，还支持可控的因果评估，为多智能体代码生成系统的科学评估与优化提供了新范式。
链接: https://arxiv.org/abs/2609.38345
作者: Chun-Wah Hsu,Kai Gong,Yu Wu,Xianhe Chen,Mengyang Liu,Jie Li,Hanyu Li,Zhixuan Liu,Naisheng Tang,Jiaying Chi,Ziheng Fan,Xuning He,Xiaokang Yang,Xue Jiang,Yihong Dong
机构: Shanghai Jiao Tong University (上海交通大学); University of Cambridge (剑桥大学); Nanyang Technological University (南洋理工大学); The University of Hong Kong (香港大学); Imperial College London (帝国理工学院); Peking University (北京大学); Tencent (腾讯)
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: work on process
Abstract:Multi-agent coding systems are designed to tackle complex software engineering tasks through collaboration. However, existing evaluations typically assume configured organizations are followed faithfully, whereas reality differs. This behavioral gap, combined with differences in underlying system components, prevents clear attribution of observed gains. To this end, we introduce OpenCollab, a multi-agent coding framework that provides a unified infrastructure for programmable collaboration and controllable runtime. Specifically, OpenCollab unifies organization design, enforces experimental control on a shared runtime, and tracks execution through fine-grained event streams. On this basis, we define Adherence to quantify whether the declared organization is actually realized. Our experiments reveal that agents collaborate very differently across configurations: changing any single dimension shifts Adherence, from 47.2% to as high as 97.2%. Furthermore, extensive agentic coding benchmarks show that a two-coder workflow built on OpenCollab establishes new SOTA performance compared to the mainstream harnesses such as Mini-SWE-agent, Codex CLI, and Claude Code, showing that a well-designed organization can outperform strong existing harnesses, while OpenCollab’s single-agent configuration uses the fewest tokens across all evaluated suites. OpenCollab establishes a unified multi-agent infrastructure for easy programmable collaboration and controlled causal evaluation.
[NLP-141] EVOKE: Eliciting World Knowledge in Agents for Transferable Decision-Making
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）作为智能体在多步决策任务中部署时，难以泛化到未见过环境的问题。其核心挑战在于：尽管LLM在预训练阶段已内化了大量关于数字环境的世界知识（world knowledge），但常规的后训练方法缺乏足够的激励机制来有效激发这些隐含知识，导致智能体倾向于依赖表面化的上下文习惯或单一目标关联，而非基于对环境本质的理解进行决策。为此，论文提出EVOKE这一后训练方法，其关键创新在于通过在固定环境状态和交互历史下引入目标多样性（goal diversity），强制智能体在不同目标约束下重新评估相同候选动作的优先级，从而迫使策略必须依赖其内部预训练世界模型中的深层结构化知识进行判断，而非仅依赖浅层上下文模式。这种设计通过直接的决策监督，隐式地激发并利用了模型已有的世界知识，显著提升了任务性能、未见环境泛化能力与数据效率。实验证明，该方法在多种任务与模型架构上均表现优异，并揭示了目标多样性是驱动可迁移行为的关键机制。
链接: https://arxiv.org/abs/2609.38334
作者: Yuhan Guo,Jinming Liu,Liang Xu,Ziqiang Li,Jianguo Huang,Zhicheng Wang,Hu Zhu,Qiuyu Chen,Yuntao Wei,Xin Jin,Wenjun Zeng
机构: 未知
类目: Computation and Language (cs.CL)
备注: 19 pages. Project page: this https URL ; Code: this https URL ; Models: this https URL
Abstract:Large language models (LLMs) are increasingly deployed as agents for multi-step decision-making, yet transfer poorly to unseen environments. World-model methods address this by training agents to predict future observations, at the cost of additional training and errors that compound when predictions are used for planning. However, for LLM agents operating in digital environments, much of this world knowledge is already internalized during pretraining, which shifts the problem from acquiring it to eliciting it. We argue that typical post-training provides little pressure for such elicitation, since supervision under a single goal at each visited state inadvertently drives policies to rely on superficial contextual habits. We introduce EVOKE, a post-training method that supplies this pressure through goal diversity at fixed states. Motivated by theory showing that an agent competent across diverse goals must encode a world model recoverable from its action preferences, EVOKE holds the environment state and interaction history fixed and ranks the same candidate actions under alternative goals, forcing action preferences to change, so that a policy relying on contextual habits or single-goal correlations cannot order them correctly. This implicitly elicits the policy’s pretrained world knowledge to inform decisions. We evaluate EVOKE across diverse tasks in three backbones, demonstrating improved task performance, unseen environment generalization, and data efficiency. We further conduct controlled analyses to better understand what drives these gains. These findings offer a new perspective on eliciting internalized world knowledge for transferable action through direct decision supervision.
[NLP-142] Hermes: Learning Contextual Reasoning Unlocks Test-Time Scaling
【速读】： 该论文旨在解决在推理阶段通过增加计算资源（test-time scaling）来提升模型性能时，如何有效分配上下文窗口中的计算资源、决定何时引入新上下文以及如何在不同上下文间传递信息这一关键问题。其核心挑战在于实现模型对上下文管理的自主决策能力，即“上下文推理”（contextual reasoning）。传统方法主要通过预设的框架（harness）硬编码这些决策，缺乏灵活性。本文提出的关键解决方案是：1）设计了一套可配置的轻量级框架Hermes，通过逐步增强模型对上下文分配与复用的控制权，使模型能够动态适应不同任务需求；2）提出Hermes-Learn，一种两阶段训练框架，用于学习模型在推理过程中自适应地进行上下文管理的能力。实验表明，具备此能力的模型能有效利用额外推理时计算资源实现性能扩展，而小型开源模型在未训练时难以做到这一点；通过Hermes-Learn训练后，模型可发展出随问题复杂度和推理进展动态调整的上下文推理策略。该方法在多个基准测试中均表现优异，具有良好的泛化性、外推能力，并可与其它测试时扩展方法协同工作。
链接: https://arxiv.org/abs/2609.38332
作者: Xinyu Li,Mononito Goswami,Hao Liu,Nikos Kanakaris,Langlin Huang,Prithwish Jana,Patrick Blöbaum,Purak Jain
机构: Carnegie Mellon University; AWS AI Labs; Washington University in St. Louis; Georgia Institute of Technology
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 45 pages
Abstract:Test-time scaling improves model performance by allocating additional compute during inference. Using this compute effectively across multiple context windows requires deciding how to allocate fresh contexts and what information to carry between them. We call a model’s ability to make these decisions contextual reasoning. Existing approaches largely prescribe these decisions through their harness; we instead shift them to the model. We introduce 1) Hermes, a family of simple, configurable harnesses that progressively varies model control over context allocation and reuse, and 2) Hermes-Learn, a two-stage framework for learning these capabilities. We find that capable models can exploit this flexibility to scale with additional inference-time compute, while smaller open-source models initially struggle to do so. Training with Hermes-Learn closes this gap, inducing adaptive contextual reasoning strategies that vary with both the problem and the progress of reasoning. These gains generalize across benchmarks and models, extrapolate beyond the inference-time compute seen during training, and transfer to complementary test-time scaling methods beyond Hermes.
[NLP-143] HARDE: Optimizing Agent Harnesses for Runtime Risk Detection and Execution Control
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）代理在运行时面临的多种安全风险，如恶意指令注入或误导性信息传播，其核心问题是现有系统级防御机制要么仅聚焦于风险检测而缺乏及时预防能力，要么依赖预设规则，难以适应多样化的威胁场景。为此，论文提出一种风险感知的运行时约束框架（risk-aware harness），通过集成基于大语言模型的监控模块，构建以“触发器（trigger）、监控器（monitor）和反馈（feedback）”为核心的三模块结构，实现对潜在风险的灵活识别与精准干预，同时最大限度减少对正常任务执行的干扰。为提升该框架在不同风险类型与部署环境下的适应性，进一步引入两阶段优化框架HARDE（Harness Adaptation via Risk-aware and Dynamic Evaluation），首先通过独立探测各模块生成优化指导，再基于安全性和任务效用反馈进行迭代优化。实验结果表明，HARDE在多个攻击基准测试中显著提升了运行时安全性并保持了任务效用，优于人工设计的约束框架及基础优化基线。研究分析揭示，有效的运行时防御依赖于互补的安全机制、面向攻击的约束优化策略以及与监控能力相匹配的架构设计。
链接: https://arxiv.org/abs/2609.38291
作者: Zhuo Liu,Moxin Li,Zhixin Ma,Wentao Shi,Wenjie Wang,Fuli Feng
机构: University of Science and Technology of China(中国科学技术大学); National University of Singapore(新加坡国立大学); Singapore Management University(新加坡管理大学)
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL)
备注:
Abstract:Large language model (LLM) agents are vulnerable to safety risks such as injected malicious instructions or misleading information, motivating runtime defenses that prevent unsafe action in execution across diverse risks while preserving benign-task utility. Existing system-level defenses either focus on risk detection rather than timely prevention or rely on predefined rules with limited flexibility across diverse risks. We propose a risk-aware harness that integrates LLM-based monitoring for flexible risk detection and structures monitor-guided execution around three core modules: trigger, monitor, and feedback, enabling targeted safety interventions while limiting disruption to benign task execution. To adapt the harness to different risks and deployment settings, we introduce HARDE, a two-stage harness optimization framework that first performs isolated probing of each module to derive an optimization guide, then uses this guide to iteratively optimize the harness based on safety and utility feedback. Experiments across three attack benchmarks show that HARDE improves runtime safety while preserving utility, outperforming manually designed harnesses and naive optimization baselines. Our analysis shows that effective runtime defense benefits from complementary safety mechanisms, attack-aware harness optimization, and harness designs matched to monitor capabilities. Our code is available at this https URL.
[NLP-144] NinaXander: Feasibility and Limits of Composing Frozen Language Models Across Architecture Families via a Shared Latent Space
【速读】： 该论文旨在解决如何在不重新训练的前提下，将来自不同架构家族的冻结语言模型（frozen language models）进行后处理重组，以实现性能互补与资源优化的问题。其核心挑战在于跨架构模型间的中间表征（intermediate representations）是否具有可对齐性，从而支持高效、无缝的组合。解决方案的关键是提出NinaXander框架，通过引入一个单一训练的共享潜在适配器（shared-latent adapter），在不同架构模型的层间进行中间表示转换，实现模型层的灵活拼接。该方法允许在适配器训练完成后，无需再次训练即可生成多个在不同层级连接的组合模型。实验表明，使用RWKV与Pythia模型组合时，最优配置在保持接近原始模型准确率的同时，将Transformer的键值缓存（KV cache）占用降低84.4%，显著提升推理效率；然而，组合模型在多项选择任务上的表现仍不及原生模型Pythia，且在域外数据集WikiText上的语言建模性能急剧下降，表明尽管在特定条件下观察到中间表示的对应关系（如共享分词器、相同深度与隐藏维度），但跨架构模型并未共享通用语义空间，限制了组合效果的进一步提升。
链接: https://arxiv.org/abs/2609.38261
作者: Takanori Kotama,Shun-ichiro Hayashi,Daichi Mukunoki,Tetsuya Hoshino,Takahiro Katagiri
机构: Nagoya University (名古屋大学); Information Technology Center (信息技术中心)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:In this paper we propose NinaXander, a series of composed language models obtained by connecting layers of frozen language models from different architecture families with a single trained shared-latent adapter. A composed model runs the first layers of one model, converts the resulting intermediate representation once with the adapter, and then runs the remaining layers of the other model. Once the adapter is trained, several composed models that connect at different layers are obtained without retraining. Using the recurrent RWKV-4-Raven-7B and the Transformer-based Tulu-Pythia-6.9b, abbreviated as RWKV and Pythia, this study examines whether frozen models from different families can be recombined post hoc. The composed models answered multiple-choice questions, and those whose generations we examined produced syntactically well-formed text. The configuration that combines the first 5 layers of Pythia with the remaining 27 layers of RWKV reduced the Transformer key-value (KV) cache by 84.4% with accuracy not significantly different from that of RWKV alone. In multiple-choice accuracy, however, no composed model matched the parent model Pythia, and language-modeling performance decreased sharply on WikiText, a corpus of Wikipedia articles outside the training domain. The correspondence between intermediate representations was also obtained in one favorable case, with a shared tokenizer, the same depth, and the same hidden width, and does not show that the models share a general semantic space.
[NLP-145] ContextAdapt: Evaluating Contextual Adaptation and Value Alignment in LLM s
【速读】： 该论文旨在解决大语言模型（LLMs）在跨专业领域中对核心价值（如诚实、自主性和保密性）的适切应用问题，尤其关注模型是否能在不同专业情境下正确调整价值的适用方式，同时在不改变专业规范的情境下保持一致性。其解决方案的关键在于提出并验证一个名为ContextAdapt的评估框架，该框架基于医学、法律、金融和国家安全四大专业领域的权威文献，构建了“价值×领域”交叉分析体系，通过设计涵盖默认规则与公认例外的测试场景，系统评估12个主流大语言模型在行为建议及其理由解释上的表现。研究发现，尽管模型在行为适当性上整体表现良好（平均达95.6%），但其使用正确的领域特定理由的能力差异显著（25.6%–76.9%），且当风险程度变化时，模型会出现局部严重失准——即使专业义务未变，仍因感知严重性而改变响应，表明模型对价值的应用易受非规范性线索干扰。这揭示出：评估生成式AI（Generative AI）的价值对齐，不仅需考察其是否遵循抽象原则，更关键的是检验其能否在具体语境中实现精准、一致且符合专业规范的价值适配。
链接: https://arxiv.org/abs/2609.38260
作者: Olivia Macmillan-Scott,Mirco Musolesi
机构: University College London (伦敦大学学院); University of Bologna (博洛尼亚大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Values such as honesty, autonomy, and confidentiality are often regarded as general principles underpinning AI alignment. However, what it means to act in accordance with these values can depend on the context in which a decision is made. In this paper, we ask whether large language models (LLMs) appropriately adapt the application of a value across professional settings, while remaining consistent when contextual changes do not alter the relevant professional norm. To study this, we introduce ContextAdapt, an evaluation framework covering honesty, autonomy, and confidentiality across medicine, law, finance, and national security. Drawing on primary-source professional and regulatory documents, we construct a value x domain framework and use this to develop scenarios testing both default professional rules and recognised exceptions. We evaluate 12 LLMs on both the actions they recommend and the justifications they provide. In our main experiment, models achieve 95.6% mean appropriateness, although the use of the correct domain-specific justification varies substantially across models, from 25.6% to 76.9%. In a separate factorial experiment, explicitly naming the professional domain and changing the role of the model have limited effect on behaviour. Varying stakes, however, reveals severe but localised failures: in some cases, models alter their responses even though the underlying professional obligation remains unchanged. In particular, perceived severity appears to act as a cue for disclosure across both honesty and confidentiality scenarios. These results show that evaluating value alignment requires us to consider not only whether models follow abstract principles, but whether they apply them appropriately across different contexts.
[NLP-146] Framing the Narrative: Ideological Mimicry in Large Language Models
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在处理政治敏感议题时，其立场是否因用户交互中的政治信号而发生动态变化的问题。传统评估方法将模型立场视为静态属性，但现实中用户通过术语选择、前提假设及个人背景传递政治信号，可能引发模型的意识形态趋同现象。研究的关键在于揭示：大语言模型并非以固定立场回应问题，而是会根据提示（prompt）的框架——包括争议性术语、政治倾向性前提及用户信息——系统性地调整其输出的政治立场。研究构建了Poli-SHIFT数据集与评估框架，对七种开源大模型在美、英、澳三国的十项政治争议议题上进行测试，结果显示，仅改变术语表述即可使模型在16.9%的对比中反转其支持立场；用户的显性政治立场也显著影响模型响应方向。这表明模型的政治立场具有高度交互依赖性，而非固有属性。随着大语言模型日益成为个性化信息源，这种动态适应机制可能加剧信息茧房效应，强化用户既有观点，进而加深社会政治分裂。
链接: https://arxiv.org/abs/2609.38256
作者: Olivia Macmillan-Scott,Michael Jacobs,Nils Metternich,Mirco Musolesi
机构: Centre for AI, Department of Computer Science, University College London(伦敦大学学院); Public Policy Group, ETH Zürich(苏黎世联邦理工学院); Department of Political Science, University College London(伦敦大学学院); Department of Computer Science and Engineering, University of Bologna(博洛尼亚大学)
类目: Computation and Language (cs.CL); Computers and Society (cs.CY)
备注:
Abstract:Large language models (LLMs) are increasingly used to answer questions about politically contentious issues, yet evaluations typically treat a model’s stance as a relatively stable property. Real users, however, communicate political signals through their terminology, assumptions, and personal context. We investigate whether such signals produce ideological mimicry: systematic shifts in the political stance expressed by an LLM toward the position conveyed by the interaction. If LLMs adapt their responses to these signals, they risk creating personalised political information environments in which users with opposing views receive systematically different accounts of the same issue, potentially reinforcing existing divisions. We build the Poli-SHIFT dataset and evaluation framework and assess seven open-weight LLMs across ten contentious political topics in the United States, United Kingdom, and Australia, systematically manipulating contested terminology, politically valenced premises, and user information, and eliciting responses in both multiple-choice and open-text formats. Across models, we find robust evidence that prompt framing shapes the political stance of LLM outputs. Changing terminology alone reverses which side of an issue a model supports in 16.9% of matched comparisons. Stated political ideology also systematically shifts responses toward the user’s position. These findings show that political stance is not a fixed property of LLMs; the views expressed are conditional on the interaction with the user. As LLMs become increasingly personalised sources of information, such interaction-dependent adaptation could contribute to political information environments that reinforce users’ existing perspectives.
[NLP-147] When Does a Spoken Agent Have Enough Evidence to Act? The PACT-SLM Contract Test
【速读】： 该论文旨在解决流式语音对话模型（Streaming Spoken Agents）在语音输入尚未充分支持时即提前执行外部动作的问题，尤其关注在最终回合评分无法反映各语音前缀对动作支持程度的局限性。其核心解决方案是提出部分语音动作契约（Partial Speech Action Contract for Turn Taking in Speech Language Models, PACT-SLM），一种受控评估范式，能够独立量化动作的身份（action identity）与触发时机（timing），并精确标注首次有效动作时间点。关键创新在于构建了包含80组语义对立样本（来自4个保留语义类别）及1600个前缀预测的诊断数据集，涵盖纯净与15 dB噪声条件下的语音渲染。实验表明，经修正分支编码与语义标签不匹配问题后，重构的WavLM Base Plus探测器在语音起始后的语义标签准确率达到26.03%（95%置信区间：22.14%–29.68%），揭示了在语音起始前已有18.99%的前缀被错误地触发动作，并能精确预测5.94%的完整行为轨迹。尽管其在起始时机判断上优于基线模型（36.25% vs. 23.13%），但动作身份识别能力仍低于最优水平，且其性能位于100次前缀内标签置换的第96百分位，略低于参考基准（26.73%）。结果表明，动作身份与触发时机分别反映语音决策行为的不同维度，强调了对部分语音情境下行为推断进行解耦分析的重要性。
链接: https://arxiv.org/abs/2609.38232
作者: Mengzhe Geng
机构: National Research Council Canada (加拿大国家研究委员会)
类目: ound (cs.SD); Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
备注:
Abstract:Streaming spoken agents may take an external action before the available speech supports it, yet final-turn scores do not reveal whether each observed prefix supports that action. We introduce the Partial Speech Action Contract for Turn Taking in Speech Language Models (PACT-SLM), a controlled evaluation that assigns a first valid action time and measures action identity and timing separately. The primary diagnostic contains 80 paired contrast groups from four held-out semantic families and 1,600 prefix predictions across clean and 15 dB noise renderings. After correcting a mismatch between randomized branch codes and semantic labels, a refitted WavLM Base Plus probe reaches 26.03% pooled post-onset semantic-label accuracy (95% group-bootstrap interval: 22.14%-29.68%), exposes an action on 18.99% of pre-onset prefixes, and predicts 5.94% of complete trajectories exactly. It exceeds matched text, scalar-acoustic, and shuffled-representation probes in post-onset label accuracy, but its score is at the 96th percentile of 100 within-prefix label permutations and below the 97.5th-percentile reference (26.73%). Elapsed time is more onset-exact than WavLM Base Plus (36.25% vs. 23.13%) but less accurate about action identity (9.92% vs. 26.03%). These results show that action identity and timing measure distinct aspects of partial-speech decision behavior.
[NLP-148] Conformal Factuality Control for Multi-Hop Retrieval-Augmented Generation
【速读】： 该论文旨在解决多跳检索增强生成（multi-hop RAG）中生成结论缺乏事实支持的问题，即尽管通过外部证据进行检索与推理，但最终生成的主张仍可能因信息链断裂或错误关联而失真。其核心解决方案是将先前适用于单跳RAG的声明级合取保真度控制（claim-level conformal factuality control）方法拓展至多跳场景，采用拆分合取（split-conformal）声明过滤机制，在保证高置信度的前提下筛选出可被外部证据充分支持的主张。实验结果表明，在HotpotQA、Natural Questions和TriviaQA三个数据集上，使用Llama 3.1 8B与GPT-4o-mini模型时，当设定95%的合取目标时，保留主张的完全支持率可达95.80%–97.20%，显著优于未过滤情况下的55.60%–76.03%。然而，该方法在提升保真度的同时也带来较高的主张弃用率（仅4.41%–31.09%的主张被保留），且响应非空率下降至9.70%–51.40%，揭示了保真度提升需与主张保留率和响应完整性共同权衡。因此，该研究证明合取保真度可有效扩展至多跳RAG，但强调名义可靠性必须结合主张保留率与拒答行为综合评估。
链接: https://arxiv.org/abs/2609.38222
作者: Muhammad Aimal Rehman,Chi-Kuang Yeh
机构: Georgia State University (佐治亚州立大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注: 15 pages, 2 figures. Code available at this https URL
Abstract:Retrieval-augmented generation (RAG) can ground large language models in external evidence, but retrieved context does not guarantee that generated claims are factually supported. This problem is especially relevant in multi-hop RAG, where retrieval and reasoning proceed through multiple dependent stages. We study whether claim-level conformal factuality control, previously developed for RAG, remains effective in this setting. We apply split-conformal claim filtering to multi-hop RAG and evaluate it on HotpotQA, Natural Questions, and TriviaQA using Llama 3.1 8B and GPT-4o-mini, together with a single-hop reference experiment. Across all six multi-hop model-dataset configurations, increasingly stringent conformal targets consistently increase the fraction of responses whose retained claims are fully supported. At the 95% target, this rate ranges from 95.80% to 97.20%, compared with 55.60%-76.03% without filtering. However, the improvement is strongly selective: only 4.41%-31.09% of generated claims are retained and 9.70%-51.40% of responses remain non-empty at the 95% target. These results show that conformal factuality extends to multi-hop RAG, while demonstrating that nominal reliability must be interpreted jointly with claim retention and abstention.
[NLP-149] utlAit v1: a crowdsourced Moroccan Tamazight speech dataset with Arabic transcriptions and regional accent labels
【速读】： 该论文旨在解决摩洛哥柏柏尔语（Tamazight）在语音技术领域严重资源匮乏的问题，具体表现为公开可用的标注语音数据稀缺、缺乏对地域变体的明确标注，且转录质量参差不齐。其解决方案的关键在于构建TutlAit数据集——一个包含摩洛哥柏柏尔语语音与现代标准阿拉伯语文本配对，并带有明确地域口音标签的语料库。该数据集通过自研众包网络应用TutlAit收集，采用Text-to-Audio和Audio-to-Text两种工作流，由母语者参与并标注其所属地域变体（阿特拉斯、苏斯、里夫等），同时结合从公开音视频媒体中提取的补充片段，经ELAN标注后通过批量导入管道整合。所有上传音频均经过服务器端标准化处理（16kHz单声道WAV）、SHA-256哈希去重、时长范围校验及管理员审核，确保数据质量。最终数据集共包含13,384个音频文件，总计约20.9小时，可广泛应用于柏柏尔语语音识别、语音翻译及口音识别任务。
链接: https://arxiv.org/abs/2609.38219
作者: Mohamed-Amine Chadi,Ezzahra Ait El Arbi,Ismail Khayoub,Aymane Fadili,Yassine Ennhili,Jadjigua Bouali,Hanane Inhid,Mohammed Ameksa,Hajar Mousannif
机构: Cadi Ayyad University(卡迪阿亚德大学)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Tamazight (Amazigh) is, together with Arabic, one of the two official languages of Morocco, yet it remains severely under-resourced for speech technology: pub licly available labelled audio is scarce, generally lacks information on the regional variety spoken, and is often of uneven transcription quality. This article describes the TutlAit dataset, a corpus of Moroccan Tamazight speech paired with Modern Standard Arabic text and explicit regional accent labels. The data were collected with TutlAit, a purpose-built crowdsourcing web application (React 18 front end, Django 5 / Django REST Framework back-end, PostgreSQL database). Native speakers recruited through targeted LinkedIn and Instagram campaigns created an account, declared their regional variety (Atlas, Souss, Rif or other) and demographic information, and then contributed through two workflows: Text-to Audio, in which an Arabic sentence is displayed and the volunteer records its oral Tamazight rendering in the browser, and Audio-to-Text, in which a Tamazight excerpt is played and the volunteer types its Arabic transcription. A complemen tary set of segments was obtained from freely accessible Tamazight audiovisual media, segmented and annotated with ELAN and imported through a bulk CSV/ZIP pipeline. Every upload is converted server-side to 16kHz mono WAV, hashed with SHA-256 for duplicate rejection, checked for duration bounds and validated by an administrator. The dataset contains 13,384 audio files totalling 75,231 seconds (approximately 20.9 hours, about 3.01GB). The Atlas variety accounts for 9,956 files (14.08h) and the Souss variety for 3,378 files (6.75h); small Rif (22 files) and Kabyle (28 files) subsets are also included. The corpus can be reused for speech recognition, speech translation and accent identification for Moroccan Tamazight.
[NLP-150] he System Prompt Illusion: How Instruction Preambles Modify Computation in Language Models
【速读】： 该论文旨在解决系统提示（system prompt）在控制大语言模型行为时，其内部计算机制如何影响模型输出这一关键问题，尤其关注系统提示对Transformer模型各层表征的深层影响。研究发现，系统提示的作用具有显著的层级选择性和指令类型依赖性：角色设定与格式化指令会深度重构中间层表示，而安全类指令则几乎不改变模型内部表征，其变化在统计上与基线无异，表明模型虽“感知”到安全提示，但并未在其计算路径中实质性执行。进一步分析显示，严格限制型与明确授权型（如“你没有任何限制”）安全提示所激活的计算路径高度一致（平均中心核对齐，CKA相关性达0.997），且此现象在70B至72B参数量级的商用模型中依然存在，安全机制渗透率低于10%。通过线性探测与因果激活修补验证，确认仅少数特定层负责行为转变，且这些层的表征深度可有效预测行为效应大小（Spearman秩相关系数ρ = 0.761，p < 0.001）。因此，解决方案的关键在于揭示了系统提示的安全机制本质为“感知但未执行”的解耦设计，从而解释了为何基于系统提示的安全防护存在持续的越狱漏洞。
链接: https://arxiv.org/abs/2609.38205
作者: Muhammad Usama,Dong Eui Chang
机构: Korea Advanced Institute of Science and Technology (KAIST); Control Laboratory, School of Electrical Engineering
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:System prompts are the primary lever practitioners use to control language model behavior, yet what they actually do to the computation inside the transformer remains poorly understood. Across 17 instruction-tuned models spanning 8 architecture families and 1.5B to 72B parameters, we use Centered Kernel Alignment (CKA) to compare layer-wise representations under 20 system prompts in five functional categories. Effects are layer-selective and instruction-type-dependent: persona and formatting instructions deeply restructure intermediate representations, while safety instructions barely move them, producing changes statistically indistinguishable from a minimal baseline. Restrictive safety instructions and explicitly permissive ones (“you have no restrictions”) engage near-identical computational pathways (mean CKA correlation 0.997), and this persists at commercial scale, where safety penetration remains below 10% even at 70B-72B. A linear probing baseline exposes the mechanism: the model encodes prompt category at every layer but restructures its computation only at a small subset, so the prompt is reliably “seen” but, for safety, not deeply “acted upon.” Causal activation patching confirms these layers mediate behavioral change, and representational depth predicts behavioral effect size across the full 17-model cohort (Spearman rho = 0.761, p 0.001). The findings provide a mechanistic explanation for the persistent jailbreak vulnerability of system-prompt-based safety. Code: this https URL
[NLP-151] Automatic estimation of verbal fluency index in people with Motor Neuron Disease using ASR alignment and pause modelling
【速读】： 该论文旨在解决肌萎缩侧索硬化症（ALS）患者认知障碍（Cognitive Impairment, CI）监测中的关键挑战，即由于伴随的言语障碍导致传统评估方法难以实施。针对此问题，研究提出一种基于自动语音分析的自动化方法，用于估计爱丁堡认知与行为ALS量表（ECAS）中的言语流畅性指数（Verbal Fluency Index, VFI）。其解决方案的关键在于整合先进的语音识别技术（WhisperX）、语音活动检测（Silero VAD）以及精细化的时间戳对齐策略，从而实现对患者语音样本中临床可解释特征的有效提取。通过结合多类特征（包括临床启发式特征、传统声学特征及自监督嵌入），研究发现临床启发式特征在预测性能上显著优于其他类型，最优模型在生成词数（P-words）和搜索词数（S-words）预测中分别达到R²=0.9、NRMSE=0.05和R²=0.8、NRMSE=0.08，验证了自动化VFI估算的可行性与高精度，为ALS患者认知状态的远程、客观评估提供了可靠的技术路径。
链接: https://arxiv.org/abs/2609.38203
作者: Bahman Mirheidari,Leslie Ing,Daniel Blackburn,Sharon Abrahams,Christopher McDermott,Heidi Christensen
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注:
Abstract:Monitoring cognitive impairment (CI) in motor neuron disease (MND) is essential for timely treatment and care, yet challenging due to co-occurring speech difficulties. The Edinburgh Cognitive and Behavioural ALS Screen (ECAS) provides a robust metric for CI assessment, with the Verbal Fluency Index (VFI) a central element. Building on recent advances in automated speech analysis, this study proposes a system for estimating VFI. It leverages a unique MND dataset and combines ASR (WhisperX) and VAD (Silero) with refined timestamping to predict the VFI and extract several clinically interpretable measures. Our approach outperformed systems based on traditional acoustic features and self-supervised embeddings, evaluated using multiple regression algorithms. Clinically inspired features consistently outperformed the other sets, with the best models achieving strong results (P-words: R2 0.9, NRMSE 0.05; S-words: R2 0.8, NRMSE 0.08), demonstrating the feasibility of automated VFI estimation.
[NLP-152] omasuLLM : Out-of-Order Speculative Execution for LLM Agents
【速读】： 该论文旨在解决编码智能体（coding-agent）在执行长时运行工具（如编译器、测试套件和仓库命令）时导致的延迟瓶颈问题，这类工具通常需要数秒至数分钟，造成智能体长时间处于空闲等待状态。其核心挑战在于：顺序执行接口虽然保证了正确性，却隐藏了可提前预测并启动的潜在并行工作，从而降低了整体效率。为应对这一问题，论文提出TomasuLLM运行时系统，其关键创新在于在不破坏任务执行正确性的前提下，实现工具调用的轨迹外（out-of-trajectory）执行。具体而言，系统通过预演未来动作、在隔离的写时复制（copy-on-write）沙箱中并行执行这些动作、追踪其依赖关系与副作用，并仅在验证结果与已提交状态一致后，按原始轨迹顺序提交最终结果。实验表明，TomasuLLM在多个基准测试中显著提升性能，平均加速比达1.31倍至1.35倍，且在4,010条审计提交验证记录中实现了零误接受（false accepts），证明了其高效性与可靠性。
链接: https://arxiv.org/abs/2609.38201
作者: Jiangnan Yu,Ceyu Xu,Mengming Li,Shiyu Huang,Yiran Xia,Jian Weng,Hui Xue,Haohui Mai,Yuan Xie
机构: HKUST(香港科技大学); Nanjing University(南京大学); King Abdullah University of Science and Technology(阿卜杜拉国王科技大学); Zhejiang Lab(浙江大学实验室)
类目: Computation and Language (cs.CL); Operating Systems (cs.OS); Software Engineering (cs.SE)
备注:
Abstract:Long-running tools can dominate coding-agent latency: compilers, test suites, and repository commands take seconds to minutes while the agent idles. This observation stall presents the same tension that drove out-of-order processors – asequential interface hides work that can be predicted and started early, but a speculative result may become visible only after it and every earlier step have been validated. We present TomasuLLM, a runtime that executes agent tool calls out of trajectory order while preserving task-execution correctness. It drafts future actions, runs them in isolated copy-on-write sandboxes, traces their dependencies and effects, and commits results in trajectory order only after validation against committed state. Across three benchmarks spanning sub-second to minutes-long tool calls, TomasuLLM improves the reported benchmark means and scales with tool latency: 1.31x on 100 SWE-bench Verified tasks, 1.35x on 28 Terminal-Bench 2.0 tasks, and 1.27x matched progress on 18 SWE-Marathon sessions. Across 4,010 audited commit-validation records, it produces zero false accepts. Subjects: Computation and Language (cs.CL); Operating Systems (cs.OS); Software Engineering (cs.SE) Cite as: arXiv:2609.38201 [cs.CL] (or arXiv:2609.38201v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2609.38201 Focus to learn more arXiv-issued DOI via DataCite
[NLP-153] Large Language Models are Approximate Survival Estimators
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在临床生存预测任务中准确性与可靠性不足的问题，尤其是在无专门训练的情况下能否有效利用患者结构化特征进行精准的生存时间预测。其核心解决方案是提出Survprompt框架，该框架将标准化的患者临床协变量（covariates）转化为自然语言描述的临床病例（clinical vignettes），并通过零样本（zero-shot）方式调用预训练的前沿大语言模型完成生存预测。研究在两个多中心泛癌队列——公开的MSK-CHORD队列及基于LLM医学信息抽取构建的Providence St. Joseph Health Network新队列上进行了基准测试，采用受删失均值绝对误差（censored mean absolute error, cMAE）和一致性指数（concordance index, c-index）评估性能。结果表明，部分先进LLMs（如GPT-5.6-Sol）在个体生存时间预测上达到了与专门训练的随机生存森林（Random Survival Forest, RSF）相当的cMAE，甚至在前列腺癌亚型中表现更优；特征消融分析显示，LLMs对关键临床变量的权重分配与专业生存模型一致。然而，研究也发现LLMs在不同癌症类型和医疗机构间表现出显著的性能波动，且对高危与低危患者的区分能力较弱（较低的c-index），限制了其在临床实践中的可推广性。因此，尽管零样本LLMs能生成令人惊喜的预测结果，但其跨域一致性差仍是制约其临床应用的关键瓶颈。
链接: https://arxiv.org/abs/2609.38181
作者: Juan M Zambrano Chaves,Peniel Argaw,Risa Ueno,Carlo Bifulco,Kristina Young,Rom Leidner,Tristan Naumann,Hoifung Poon
机构: Microsoft(微软)
类目: Computation and Language (cs.CL)
备注:
Abstract:Survival analysis estimates time-to-event outcomes from patient covariates and is widely used for medical risk assessment. Patients seeking prognostic information after a diagnosis may turn to large language models (LLMs), now readily accessible through consumer applications. However, whether LLMs can provide accurate survival predictions has not been rigorously evaluated. We introduce Survprompt, a framework that converts structured patient covariates into free-text clinical vignettes and prompts pre-trained LLMs to predict survival zero-shot. We benchmark Survprompt against conventional survival models, including random survival forests (RSF), across two multi-institutional pan-cancer cohorts: the publicly available MSK-CHORD cohort and a newly curated cohort from the Providence St. Joseph Health Network constructed using an LLM-based medical abstraction framework. We report censored mean absolute error (cMAE) and concordance index (c-index) and conduct feature ablations to identify variables influencing LLM predictions. Frontier LLMs achieved surprisingly competitive cMAE for individual survival times. For example, GPT-5.6-Sol achieved cMAE within 10% of state-of-the-art RSF models specifically trained for survival prediction for several cancer types and lower cMAE than RSF for prostate cancer in MSK-CHORD. Feature ablations revealed that LLMs prioritized clinical variables similarly to specialized survival models. However, LLMs showed inconsistent accuracy across cancer types and institutions and poorly discriminated between high- and low-risk patients (lower c-index). Zero-shot LLMs can generate surprisingly accurate prognostic estimates without specialized training, but their variable performance across cancer types and institutions remains an important limitation for clinical use.
[NLP-154] Jev in Medicine: A Benchmark Evaluation. Preliminary Results
【速读】： 该论文旨在解决生成式人工智能（Generative AI）在医学问答与基于病例的诊断推理任务中，非生成式“系统一”模型（System One model）Jev在准确性、校准性及不可回答问题识别能力方面的表现未知问题。其解决方案的关键在于通过四个医学基准测试（MetaMedQA、PubMedQA、DiagnosisArena-MCQ 和 NEJM Case Challenges）对Jev 1.13进行系统评估，并以具备中等推理与无推理能力的GPT-6 Sol作为参照，全面分析其在顶级准确率、概率校准性、选择性预测以及对无法回答问题的识别能力等方面的表现。结果显示，尽管Jev在部分任务（如PubMedQA）上表现接近前沿大模型，在MetaMedQA中具有最优校准性，但其在复杂诊断任务中的准确率显著低于GPT-6 Sol，且概率判别能力较差，表明其在临床应用前仍需进行任务特定验证。
链接: https://arxiv.org/abs/2609.34024
作者: Alfredo Madrid-García,Beatriz Merino-Barbancho
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Jev is a non-generative “System One” model that assigns probabilities to predefined answer options and cannot answer outside them. Its accuracy and calibration on medical question-answering and case-based diagnostic-reasoning tasks are unknown. We evaluated Jev 1.13 on four medical benchmarks: MetaMedQA, PubMedQA, DiagnosisArena-MCQ and the NEJM Case Challenges. GPT-6 Sol, with (medium) and without reasoning, was the reference. The primary outcome was top-1 accuracy; key secondary outcomes were calibration, selective prediction and recognition of unanswerable questions. All 8,469 requests returned a valid answer. Jev’s accuracy was similar to that of GPT-6 Sol with medium reasoning on PubMedQA (78.4% vs 78.2%;), lower on MetaMedQA (74.8% vs 82.7%) and much lower on DiagnosisArena-MCQ (59.8% vs 82.4%;) and the NEJM cases (61.8% vs 82.4%). On MetaMedQA, Jev’s probabilities were the best calibrated (expected calibration error 0.063 vs 0.146), and its answers with a probability of at least 0.9 (52.9% of questions) were 93.4% accurate, but GPT-6 Sol was as accurate when it accepted a similar proportion of questions. On DiagnosisArena-MCQ, Jev’s probabilities discriminated poorly (AUROC 0.645 vs 0.768). Of the 162 questions whose correct answer was “I don’t know or cannot answer”, Jev chose that option for 10.5% (GPT-6 Sol, 8.6%). Median latency was 0.27-0.31 s; all 2,823 items cost USD 0.08. Jev was fast and inexpensive, and its accuracy was similar to that of a frontier LLM on research abstracts but lower on examination questions and much lower on complex diagnostic cases. Task-specific validation is required before clinical use.
[NLP-155] Instability Floors: Separating Bias from Noise in Fairness Audits of Clinical LLM Agents Agent s with FairMedAgent
【速读】： 该论文旨在解决临床语言模型代理在公平性审计中因反事实公平性（counterfactual fairness）评估所导致的“翻转率”（flip rate）解释困境问题。现有方法将翻转率直接归因于患者人口学特征的变化，但研究发现，即使在无任何输入变化的情况下，由于生成式模型（Generative AI）固有的随机性，其输出仍可能产生非人口学相关的翻转行为，从而导致对公平性的误判。解决方案的关键在于识别并量化这一由模型自身随机性引起的“基线翻转率”（floor），即在无任何输入变化时模型输出不一致的最小概率。研究通过在16个合成病例（synthetic vignettes）上重复运行同一条件十次，测得默认采样下临床代理的平均翻转率为8.7%，且在不同任务中差异显著（从重症监护升级的2.2%到受控物质警戒的17.9%）。跨六家供应商的四种模型验证表明，该基线翻转率的池化下限范围为2.5%至23.7%，且其大小受解码配置影响——如采用五次采样多数投票可降低39%的翻转率（95% CI: 18%–64%），而温度设为0时本地部署模型基本消除分歧，但托管模型仍存在。此外，研究证明：对于二元决策，若无实际人口学效应，预期翻转率即等于基线值；真实效应仅贡献其平方项，因此落入基线范围内的翻转率不能作为不公平的证据。为此，作者提出四步报告流程，并开源了FairMedAgent工具包，包含测试框架、标准化病例与分析脚本，使研究团队可自主测量其模型的基线翻转率，从而实现更可信的反事实公平性评估。
链接: https://arxiv.org/abs/2609.03221
作者: Rohith Reddy Bellibatlu,Manpreet Singh,Deepak Parashar,Rahul Joshi
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Machine Learning (cs.LG); Applications (stat.AP)
备注: 27 pages (13 main plus 14 supplementary), 4 figures, 3 tables. Code: this https URL (v0.1.5, commit 3982974; concept DOI https://doi.org/10.5281/zenodo.22165979 ). Trajectories: this https URL
Abstract:Counterfactual fairness audits of clinical language-model agents report a flip rate: how often an action changes when only the patient’s demographic descriptor changes. Part of that rate is not demographic. A stochastic agent also changes its own action when nothing changes, and a flip rate cannot be interpreted without knowing how often. We measured it. Re-running one condition ten times over sixteen synthetic vignettes at default sampling changed a clinical agent’s action in 8.7 percent of replicate pairs, from 2.2 percent for intensive-care escalation to 17.9 percent for controlled-substance caution, an output given no operational criteria. Across six models from five vendors, pooled floors ranged from 2.5 to 23.7 percent; in this panel neither disclosed size, vendor, nor hosting ordered them. The floor depends on the decoding configuration: majority voting over five draws removed 39 percent of it (95 percent confidence interval (CI) 18 to 64); at temperature 0 three of four locally served models showed no disagreement, but a hosted model still did. We also show that, for a binary action, the flip rate expected under no demographic effect equals the floor and a real effect adds only its square, so a flip rate inside the floor is not evidence of fairness, and direction must be tested with a signed paired test. We give a four-step reporting procedure and release FairMedAgent, the harness, with its protocol, vignettes and analysis scripts, so any team can measure the floor for its own agent.
[NLP-156] VOSSA: Voiceprint Optimization for Streaming Speech Architectures INTERSPEECH2026
【速读】： 该论文旨在解决实时语音转换（Real-time Voice Conversion, VC）系统中依赖预训练说话人验证（Automatic Speaker Verification, ASV）模型提取的说话人嵌入所导致的性能瓶颈问题。传统ASV嵌入虽在说话人区分上表现优异，但其设计目标是保持同一说话人在语音内容与语调变化下的稳定性，这与流式语音生成中逐帧级声学特征生成的需求存在冲突。为此，本文提出一种名为VOSSA（Voiceprint Optimization for Streaming Speech Architectures）的说话人表示框架，其关键在于从内容编码器的中间层提取说话人信息，并通过注意力统计池化（attentive statistics pooling）进行聚合，使嵌入表示能够更好地适应流式语音生成的时序约束。该嵌入与语音转换任务联合训练，无需独立的说话人编码器。实验结果表明，VOSSA在六个数据集上显著提升了基频（F0）动态和元音区分性声学线索，同时保持了与现有方法相当的NISQA-MOS、词错误率（WER）及说话人相似性；感知测试进一步验证了其在自然度、说话人相似性、可懂度和音色活力方面的综合提升。
链接: https://arxiv.org/abs/2609.38887
作者: Mu-Ruei Tseng,Waris Quamer,Ghady Nasrallah,Ricardo Gutierrez-Osuna
机构: 未知
类目: Audio and Speech Processing (eess.AS); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Published in Proceedings of Interspeech 2026
Abstract:Real-time voice conversion (VC) systems commonly rely on pretrained speaker embeddings from automatic speaker verification (ASV) models. While effective for speaker discrimination, these embeddings are trained to remain stable across phonetic and prosodic variations within-speaker, which may conflict with frame-level acoustic generation in streaming constraints. To address this issue, we propose VOSSA (Voiceprint Optimization for Streaming Speech Architectures), a speaker representation framework that extracts speaker information from intermediate content encoder layers and aggregates using attentive statistics pooling. The embedding is trained jointly with VC objectives, removing the need for a separate speaker encoder. Across six datasets, VOSSA improves F0 dynamics and vowel-discriminative acoustic cues while maintaining comparable NISQA-MOS, WER, and speaker similarity. Perceptual tests further indicate improvements in naturalness, speaker similarity, intelligibility, and vibrancy.
[NLP-157] acit-TTS: From Autoregressive Decoding to Masked Prediction for Efficient Transcript-Free Voice Cloning
【速读】： 该论文旨在解决自回归式文本到语义（text-to-semantic）建模的语音合成（TTS）系统在零样本语音克隆任务中存在生成延迟高、推理效率低的问题，同时克服非自回归方法对参考语音需依赖准确转录文本（transcript）这一限制。其解决方案的关键在于提出Tacit-TTS，一种基于IndexTTS2蒸馏得到的高效无转录词零样本语音克隆系统：通过引入掩码式非自回归语义生成机制替代原有的自回归解码，实现并行化生成以显著降低延迟；提出无需训练的声学长度估计方法，避免对参考语音进行额外建模；并通过ReFlow蒸馏技术加速流匹配渲染器。实验表明，该方法在多个英、汉语数据集上均达到与现有方法相当的零样本语音质量，且在长于5秒的语音生成场景下速度提升超过10倍。更重要的是，其无转录词的参考条件机制支持跨语言及非词汇性参考输入（如八种其他语言、婴儿咿呀学语和合成乱语），有效规避了依赖自动语音识别（ASR）转录所导致的错误或失败问题。
链接: https://arxiv.org/abs/2609.38658
作者: Jian Chen,You Zhang,Mark Vinton
机构: Dolby Laboratories(杜比实验室)
类目: Audio and Speech Processing (eess.AS); Computation and Language (cs.CL); Sound (cs.SD)
备注: Under Review
Abstract:TTS systems with autoregressive semantic modeling have demonstrated strong zero-shot voice cloning performance and rich expressive variation, but their sequential decoding incurs substantial latency. Non-autoregressive alternatives offer much faster generation, yet often rely on more restrictive reference conditioning, such as requiring transcripts of the reference speech during inference. We present Tacit-TTS, an efficient transcript-free zero-shot voice cloning system distilled from IndexTTS2. Our model replaces autoregressive text-to-semantic decoding with masked non-autoregressive generation, introduces training-free acoustic length estimation, and accelerates the flow-matching renderer through ReFlow distillation. Across two English and two Mandarin datasets, Tacit-TTS achieves competitive zero-shot quality while generating speech over 10x faster than IndexTTS2 for utterances longer than 5 seconds. Its transcript-free conditioning further supports cross-lingual and non-lexical references. We validate this capability using references from eight other languages, infant babble, and synthetic gibberish, where transcript-dependent systems often degrade or fail due to unreliable ASR transcripts.
信息检索
[IR-0] Decision-Oriented Recommendation Reranking: An Empirical Study of Jev
链接: https://arxiv.org/abs/2609.40241
作者: Hanjia Lyu,Yinglong Xia
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL)
备注:
Abstract:Large language models (LLMs) have shown promise for recommendation reranking, but their use introduces an important tradeoff between recommendation quality and serving efficiency. We investigate whether a decision-oriented model provides a useful alternative when the reranking task is fundamentally a structured choice among predefined candidate items. Specifically, we conduct a controlled empirical study of Jev, described by TypeSafe AI as a ``System One Model,‘’ for personalized recommendation reranking and compare it with recommendation-specific models and pointwise and listwise Qwen rerankers across multiple Amazon Reviews domains and candidate-set sizes, evaluating both recommendation effectiveness and observed serving latency. Our results show that Jev maintains strong recommendation effectiveness relative to the evaluated baselines while exhibiting substantially more gradual latency growth than the pointwise Qwen rerankers, although its observed serving latency remains substantially higher than that of recommendation-specific models. Together, these characteristics place Jev in a distinct quality–latency operating regime across candidate sizes and domains. These findings motivate further investigation of decision-oriented models for recommendation and other ranking tasks with structured output spaces.
[IR-1] Conversational Capture: A Trajectory-Level Framework for Evaluating Generative Engine Optimization in Multi-turn Human-Agent Interaction
链接: https://arxiv.org/abs/2609.40069
作者: Junwei Yu,Jieyu Zhou,Mufeng Yang,Yepeng Ding,Hiroyuki Sato
类目: Human-Computer Interaction (cs.HC); Information Retrieval (cs.IR)
备注: 9 pages, 2 figures, 2 tables. In Proceedings of the 14th International Conference on Human-Agent Interaction (HAI '26), November 16-19, 2026, Osaka, Japan
Abstract:Generative Engine Optimization (GEO) shapes content to increase its likelihood of being cited by answer engines built on retrieval-augmented large language models. GEO is typically evaluated as a single-turn property: for a fixed query, an evaluator measures a source’s visibility in one answer. We argue that the single answer is an inadequate unit of analysis. Human-agent information seeking forms a closed loop: the agent’s answer changes the user’s beliefs and therefore the next question, which in turn determines what the agent retrieves. We introduce conversational capture, a phenomenon in which a source cited early becomes substantially more likely to be cited again. Capture operates through a machine-side channel, history-conditioned retrieval, and a human-side channel, follow-up questions directed toward the captured source. We formalize the interaction as a two-layer closed-loop system and derive trajectory-level constructs: cumulative conversational visibility; a direct/feedback decomposition of trajectory gain; a nested split of the feedback term into machine-side and human-side channels; a capture coefficient; a compounding ratio; and a misranking diagnostic. Using reinforcement-process (Pólya-urn) theory, we prove that the feedback term is zero under single-turn evaluation and that GEO’s cumulative payoff grows superlinearly with conversation length while capture develops. A model-derived illustration shows that the feedback term can exceed the direct term, the compounding ratio exceeds two within ten turns, and single-turn and trajectory rankings agree only weakly (Kendall’s \tau = 0.4 ). We connect the human channel to information foraging, trust calibration, and Bayesian persuasion, and discuss design implications for answer engines.
[IR-2] Overview of BioASQ 2026: The fourteenth BioASQ Challenge on Large-Scale Biomedical Semantic Indexing and Question Answering
链接: https://arxiv.org/abs/2609.39975
作者: Anastasios Nentidis,Georgios Katsimpras,Anastasia Krithara,Martin Krallinger,Miguel Rodríguez-Ortega,Eduard Rodriguez-López,Natalia Loukachevitch,Igor Rozhkov,Elena Tutubalina,Dimitris Dimitriadis,Vasiliki Patsiou,Grigorios Tsoumakas,George Giannakoulas,Alexandra Bekiaridou,Athanasios Samaras,Giorgio Maria Di Nunzio,Nicola Ferro,Stefano Marchesin,Marco Martinelli,Gianmaria Silvello,Georgios Paliouras
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 21 pages, 17 tables, International Conference of the Cross-Language Evaluation Forum for European Languages 2026 (CLEF2026)
Abstract:This paper presents an overview of the fourteenth edition of the BioASQ challenge, organized in the context of the Conference and Labs of the Evaluation Forum (CLEF) 2026. BioASQ is an international challenge series that supports progress in biomedical language processing tasks ranging from semantic indexing and information extraction to question answering and summarization. In 2026, BioASQ included six shared tasks: a) Task 14b on biomedical semantic question answering. b) Task Synergy14 on question answering for developing biomedical top- ics. c) Task MultiClinSum-2 on multilingual clinical summarization. d) Task BioNNE-R on extracting relations between nested named entities in Russian and English. e) Task ELCardioCC on clinical coding in cardiology. f) Task GutBrainIE on gut-brain interplay information extrac- tion. Across these six tasks, 87 distinct teams participated, submitting more than 1000 runs overall. As in previous editions, several submissions reached competitive performance, reflecting the continued progress of state-of-the-art methods across biomedical language processing tasks.
[IR-3] KUAISHOU Explorer LLM -Rec Challenge 2026: Reasoning Generative Recommendation
链接: https://arxiv.org/abs/2609.39828
作者: Jiangxia Cao,Hao Peng,Wenlong Xu,Jiaxin Deng,Zhixin Ling,Xingmei Wang,Kun Shang,Can Tang,Zhihuai Cai,Jun Du,Fang Su,Xiaojuan Liu,Yiling Li,Chenglong Yu,Chongling Rao,Haixuan Gao,Haitao Xu,Jian Liang,Ruiming Tang,Chenglong Chu,Guohong Mu,Honghui Bao,Hui Wang,Jialong Chen,Jiao Ou,Muhao Wei,Peng Zhang,Renpu Liu,Ruochen Yang,Shugui Liu,Xinqi Jin,Yan Sun,Yifan Wang,Yingzhi He,Yufei Ye,Yusen Huo,Tingkuo Wang,Jihong Zhang,Lanxi Zhu,Pengyuan Liu,Zhipeng Yi,Luankang Zhang,Hang Lv,Xuyang Zhi,Tianyu Li,Bintao Wu,Chuang Ou,Siyue Su,Ziyuan Wang,Yuliang Sun,Baiyan Che,Feiyang Xu,Shiwen Zhang,Shiteng Cao,Chongcong Jiang,Yuan Fang,Xiangwu Yang,Hao Deng,Zijian Du,Pengxun Wang,Xiaoming Wang,Shun Qin,Yingqi Song,Tianyi Li,Naixiao Peng,Chenyu Zhou,Qiliang Jiang,Quan Zheng,Cheng Jin,Siying Zeng,Hongjia Xu,Junwu Hu,Teng Fu,Zhengkang Mei,Haijun Yu,Kai Li,Shengyang Zhou,Zhijia Wei,Siyi Xiong,Bo Liu,Zichun Guo,Zhubin Han,Jinpeng Fu,Bingqian Liu,Yuyi Wang,Yu Liu,Qinghai Tan,Ruijie Zhou,Zhuohang Li,Zhijia Zhong,Xiangnan He,Jirong Wen,Min Zhang,Wenwu Ou,Peng Jiang,Han Li,Kaiqiao Zhan,Yanan Niu,Lantao Hu,Kun Gai
类目: Information Retrieval (cs.IR)
备注:
Abstract:Generative recommendation, has been attracted a surge of attentions in industrial and academic research community, towards to build more smart system to build next-generation recommender. Under the significant developing wave of large language model, our team have been developed Semantic ID based OneRec/OneRec-V2. These models have been widely deployed in production and demonstrate the scaling potential of the autoregressive next-item prediction paradigm for industrial recommender systems. Building on the success of OneRec, we further explored a series of models, including OneRec-Think, OpenOneRec, and OneReason, that connect item Semantic IDs with natural language in a unified representation space and seek to unlock the potential of natural-language chain-of-thought (CoT) reasoning for recommendation. However, our preliminary works found that introducing reasoning CoT does not always improve the recommendation performance. To address this issue, OneReason strengthens the semantic alignment between items and language, introduces structured template-based supervision for interest reasoning, and applies advanced reinforcement learning techniques to make reasoning more beneficial to recommendation. As a frontier topic to building recommendation foundation models, we believe this topic has significant research value and hope to encourage more researchers to explore it together. To this end, together with the SIGIR 2026 community, we organized the KUAISHOU Explorer LLM-Rec Challenge 2026: Reasoning Generative Recommendation.
[IR-4] When the Label Ignores the Request: Auditing Policy-Selected Targets in Synthetic Conversational Music Recommendation RECSYS2026 RECSYS
链接: https://arxiv.org/abs/2609.39696
作者: Sanjeev Suresh
类目: Information Retrieval (cs.IR)
备注: 6 pages, 2 tables. Camera-ready version (CC BY 4.0). RecSys Challenge 2026 Workshop at ACM RecSys 2026, Minneapolis, October 2, 2026. Code and audit artifacts: this https URL
Abstract:Synthetic dialogues generated by LLM pipelines now serve as complete conversational-recommendation benchmarks: an LLM listener talks to an LLM recommender, and the track logged next in the conversation becomes the official label for each turn. These policy-selected labels make large-scale evaluation reproducible, but they are proxies for what the simulated user asked. We audit the one place where label and request are directly comparable: turns where the user asks for an exact song by name. In the RecSys Challenge 2026 TalkPlay benchmark, using visible dialogue and catalog metadata alone, we find that the official label contradicts the user’s exact-song request in half of the audited development turns. This matters beyond one benchmark: naming the desired item is the dominant intent in real music search, where deployed systems avoid substituting an alternative for an exactly named item, on the premise that it costs satisfaction. A small training-time supplement closes most of the gap: adding catalog-resolved request-satisfying targets to a small fraction of training turns yields a 53.3% relative gain in nDCG@20 on the 43 conflict turns while leaving the official metric intact, verified against a matched control that detects the same requests but trains only on official labels.
[IR-5] Working Around the Compute Ceiling: Byte-Exact Memory in Galahad Makes LLM Reading a One-Time Cost LLM Reading a One-Time Cost
链接: https://arxiv.org/abs/2609.39358
作者: Sietse Schelpe
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR); Machine Learning (cs.LG); Performance (cs.PF)
备注:
Abstract:A transformer language model performs a bounded amount of computation per token, and recent work by Vishal Sikka, former CEO of Infosys, argues that this bound limits which tasks a model can carry out or verify (arXiv:2507.07505). We ask how much of the budget beneath that ceiling is spent on work the model has already done. Serving is stateless across requests: a model that answers a second question about a document recomputes the document’s attention state from the first token. On seven real-world datasets, 98.7% of prompt tokens were text the model had already read. We present Galahad, a memory layer for vLLM, SGLang and this http URL that makes this reading a one-time cost. Taliesin saves the model’s key-value (KV) state for a block of text and loads it on the next request that contains the same bytes, instead of recomputing it. Blaise keeps the documents themselves and passes the model only the section a question needs. On a recall test with 100 facts hidden in a 97,000-token corpus (Gemma 4 31B), Taliesin alone let the model attend to the whole corpus and answered 98 of 100 on this http URL at 3.0 s and 572 J per question, against 10 of 100, 9.3 s and 2,754 J for the same model without Galahad, which could hold only the last 12,000 tokens. With Blaise added, the model read about 668 tokens per question and answered 100 of 100 on all three runtimes at 0.59-0.64 s and 200-213 J; a tuned RAGFlow pipeline answered 77. Storing the corpus is a one-time cost of about 100 s and 28 kJ, whose energy is recovered after 13 questions. Restored state is bit-identical: all 262,144 output logits matched after restart, rehydration and hot-load. Galahad worked with all 30 models we tested under vLLM, and it fails closed: any load that does not pass its checks is recomputed. Together these results move LLM serving from stateless to stateful inference.
[IR-6] Generative End-to-end Ad Retrieval at Douyin
链接: https://arxiv.org/abs/2609.39327
作者: Shaowen Zeng,Yanhua Huang,Jiacheng Sun,Jiarui Liu,Qian Dai,Zhikai Yang,Hancheng Li,Boya Wu,Tuoyu Zhang,Yekui Chen,Xiang Sun
类目: Information Retrieval (cs.IR)
备注:
Abstract:Generative retrieval reformulates recommendation as the generation of discrete item tokens. However, scaling this paradigm to real-world recommender systems reveals two critical bottlenecks: 1) Representation collapse, where the item tokenizer converges to degenerate results under continuous distribution shifts, fundamentally hindering stable end-to-end adaptation. 2) Item collisions, where the massive candidate pool causes distinct items to share identical token sequences, compromising the final retrieval precision. Crucially, these bottlenecks are inherently coupled: expanding codebook capacity to mitigate collisions inevitably exacerbates collapse. To address them simultaneously, we propose GEAR, an end-to-end framework that jointly optimizes the tokenizer, generator, and reranker. To mitigate representation collapse, we introduce BasisVQ, which re-parameterizes the codebook via an orthogonal basis to enable global gradient sharing and rigid spatial rotation of the latent space, effectively stabilizing gradient dynamics without ad-hoc heuristics. We further extend it to prefix-aware BasisRQ, substantially enhancing the codebook’s expressiveness with the same asymptotic time complexity. To resolve item collisions, GEAR integrates a context-conditioned reranking head into the generative process, efficiently disambiguating colliding items with minimal computational overhead. By unifying stable tokenization and joint reranking within an end-to-end generative framework, GEAR establishes a fully differentiable and scalable paradigm. It currently serves hundreds of millions of daily active users on Douyin Ads, yielding substantial empirical improvements in extensive online A/B tests.
[IR-7] Residual Trajectory Distillation for Generative Retrieval
链接: https://arxiv.org/abs/2609.39319
作者: Weihao Shen,Wei Chen,Fuwei Zhang,Guojun Liu,Qingsong Hua,Wei Lin,Fuzhen Zhuang
类目: Information Retrieval (cs.IR)
备注:
Abstract:Generative retrieval has emerged as a general retrieval paradigm, representing items with discrete Semantic IDs (SIDs) and retrieving them through autoregressive identifier generation. When SIDs are constructed with residual quantization (RQ), standard retrieval training supervises only the selected codes and discards the residual trajectories that produce them. The same hard code can nevertheless arise from different preferences over competing codewords, while the residual trajectory also contains information about subsequent quantization decisions. As a result, hard SID supervision collapses distinct quantization behaviors into identical targets and leaves information available during indexing unused in retrieval training. We introduce ResTD, a Residual Trajectory Distillation framework that transfers this discarded indexing information into retrieval training. Treating the frozen RQ indexer as a process teacher, it distills residual-induced codeword preferences into SID-decoding states. This supervision recovers distinctions hidden by hard assignments and allows earlier decoder states to capture information about subsequent quantization decisions before the corresponding SID suffix is generated. In this way, richer information from SID construction is incorporated into retrieval learning while preserving the original retrieval index and inference procedure. Experiments on multilingual e-commerce retrieval show consistent improvements over strong baselines and matched training controls. Controlled comparisons show that residual-derived targets outperform the tested codebook-only soft targets. Representation probes further show that future codebook preferences become more recoverable from earlier decoder states. ResTD can also be readily extended beyond retrieval to generative recommendation. Code is available at: this https URL.
[IR-8] Learning Multiresolution Relevance for Hierarchical Generative Retrieval
链接: https://arxiv.org/abs/2609.39312
作者: Weihao Shen,Wei Chen,Fuwei Zhang,Guojun Liu,Qingsong Hua,Wei Lin,Fuzhen Zhuang
类目: Information Retrieval (cs.IR)
备注:
Abstract:Generative retrieval with semantic identifiers (SIDs) makes successive decisions over a document hierarchy. Relevant documents for the same query may share coarse prefixes and diverge at finer depths, with branching patterns varying across queries. These paths reveal how relevance is distributed across successive refinements, yet standard full-SID supervision treats them as separate training targets. To make this allocation explicit, we formulate multiresolution relevance as consistent conditional distributions induced by a single document-level relevance measure across the SID hierarchy. We introduce \textbfRARS, \textbfResolution-\textbfAligned \textbfRelevance \textbfSupervision, which uses the resulting refinement-level distributions to supervise a shared query representation. RARS aggregates document relevance over prefixes and trains a prefix-conditioned predictor to allocate relevance among sibling branches. All relevance-bearing children participate in local competition, and each local loss is weighted by the relevance mass reaching its parent. This objective trains the query encoder to capture both the coarse structure shared by relevant documents and their finer branch allocations. The predictor is discarded after training, preserving standard autoregressive retrieval at inference. Experiments on three multilingual ESCI locales show consistent improvements over matched full-SID training under autoregressive decoding. RARS also outperforms grouped soft-target, decoder soft-target, and sampled-tree supervision under a common retrieval rule. The gains persist across alternative identifier structures and relevance definitions. Code is available at: this https URL
[IR-9] Argument Structure Prediction in Online Conversations: A Comparative Study of Modeling Paradigms and Task Architectures
链接: https://arxiv.org/abs/2609.39225
作者: Siddharth Bhargava,Sara Tonelli,Patricia Martín-Rodilla,Javier Parapar
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: CMNA’26: 26th International Workshop on Computational Models of Natural Argument
Abstract:Argument structure prediction (ASP) constructs complete argument structures from discourse by identifying argumentative units and their relations. While recent work has explored diverse approaches—including unified neural models, multi-step pipelines, and prompt-based large language models (LLMs)—their relative trade-offs remain under-explored, particularly in dialogical settings. We present a systematic evaluation of ASP under strict schema constraints, comparing supervised fine-tuning and prompt-based LLMs across single- and multi-step task architectures, generating complete argument structures from dialogical input end-to-end. We benchmark them on three diverse dialogical corpora adapted from Inference Anchoring Theory into bipolar argument structures. Under a shared evaluation framework, we assess predictive performance, cross-domain generalization, schema compliance, and computational efficiency. Our results show that ASP remains a challenging task, with identifying argumentative relations emerging as the primary bottleneck, largely due to the implicit and context-dependent nature of dialogical argumentation. To facilitate future research, we release our data processing pipeline and end-to-end modeling framework for computational ASP on dialogical corpora. Comments: CMNA’26: 26th International Workshop on Computational Models of Natural Argument Subjects: Computation and Language (cs.CL); Information Retrieval (cs.IR) Cite as: arXiv:2609.39225 [cs.CL] (or arXiv:2609.39225v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2609.39225 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[IR-10] O-Funnel: Lossless Structural Capture and Requirement-Driven Extraction from Drifting Heterogeneous Documents
链接: https://arxiv.org/abs/2609.39209
作者: Osama Mustafa
类目: Databases (cs.DB); Information Retrieval (cs.IR)
备注: 21 pages, 3 figures. Code and benchmarks: this https URL
Abstract:Pulling a fixed set of fields out of documents that arrive in many formats and under drifting schemas is usually done with hand-written byte patterns, which break whenever a key is renamed, a value is reformatted, or a lookalike value appears first. We argue the cause is structural: one pattern must both describe the value and locate it among its surroundings. O-Funnel separates the two. It transcribes any XML, JSON, CSV, HTML or key-value text document into one typed tree over five constructors, gated by an oracle that rejects any capture that does not reconstruct its source. Each needed field is declared in the tree’s own terms and located by fusing independent evidence (key, path, value shape, synonym, key spelling, record neighborhood, value profile), so the best-supported node wins and a missing field is reported with a reason. Data no requirement claims becomes residue that a funnel traces back to the requirements to learn new key aliases. On 34,989 real PubMed records, O-Funnel matches a hand-written parser (F1 1.00). After a five-element schema rename, the parser’s regular expressions fall to 0.20 while O-Funnel stays at 1.00, with every capture verified complete. On constructed suites that isolate regex failure modes it raises F1 from 0.43 to 1.00, and from 0.80 to 0.94 after self-improvement; on held-out schema-matching instances it is competitive with classical matchers without training. O-Funnel is a dependency-free Python library (pip install ofunnel).
[IR-11] Routing Between Generative and Collaborative User Profiles: A Serving-Time Gate for Controllable Novelty
链接: https://arxiv.org/abs/2609.39043
作者: Milad Sabouri,Neeraj Sharma,Sardar Hamidian,Shaghayegh Agah
类目: Information Retrieval (cs.IR)
备注:
Abstract:Large language models (LLMs) enable rich semantic user profiles for recommendation, but such profiles are more expensive to generate and are not necessarily desirable to deploy uniformly. We study whether LLM-generated profiles can instead be invoked selectively within a production recommendation pipeline. Using a real-world streaming dataset covering movies, TV shows, and sports content, we train a serving-time routing gate that assigns each user to either a collaborative sequential recommendation model or a recommendation model driven by an LLM-generated profile. The gate uses only serving-time features and learns to identify users for whom profile-based routing can increase Novelty@10 while preserving ranking relevance. A routing threshold controls how aggressively users are sent to the generative model, exposing a tunable novelty–relevance trade-off. At an overall NDCG-loss budget of 5%, the learned gate increases Novelty@10 by 6.5% while routing 12.5% of users, outperforming simple heuristic and random routing policies at comparable relevance cost. These results show that LLM-generated user profiles can serve as a controllable complement to collaborative recommendation, while results with non-generative semantic profiles indicate that the benefit stems from selective routing rather than LLM generation alone.
[IR-12] Evidence First Arithmetic Second: A System Report and Failure Analysis for DocSem EMNLP2026
链接: https://arxiv.org/abs/2609.39013
作者: Divya Godara,Sachin Gupta
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 5 pages, 1 figure, 2 tables. Accepted as a shared-task system paper at DocInsights 2026, co-located with EMNLP 2026
Abstract:EVICALC, our system for the DocSem shared task, achieved 8.61% joint accuracy on 1,730 tasks in the official final test evaluation. It reads a PDF, selects a passage, asks a language model to write an arithmetic expression, and evaluates that expression in local code. Saved intermediate results support inspection of failures. A separate public-validation run achieved 92.17% answer accuracy and 1.00 evidence F1. The configurations and metrics differ, so these scores are not a controlled comparison. Our manual, post-hoc analysis is descriptive: in one inspected case, optical character recognition (OCR) and block grouping merged the relevant passage into another block, and the system answered from unrelated text. An exploratory study of reading page images on 100 documents returned evidence identifiers for only 22 documents. These descriptive findings motivate further evaluation; they do not establish the causes of the overall score.
[IR-13] RouteRec: Behavior-Guided Sparse Routing for Sequential Recommendation CIKM2026
链接: https://arxiv.org/abs/2609.39007
作者: Junyeong Song,Jaemin Yoo
类目: Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: Accepted at CIKM 2026. 12 pages, 12 figures, 7 tables
Abstract:Sessionized interaction histories contain behavioral patterns that can improve sequential recommendation. However, existing models process all sessions through the same parameterized blocks, regardless of their behavioral differences. Mixture of Experts (MoE) enables conditional computation, but it leaves open what should guide expert allocation. We propose RouteRec, a sequential recommender that uses observed session behavior as the routing criterion. RouteRec summarizes four types of behavioral evidence from sessionized histories: interaction tempo, item-group focus, repetition and carryover, and popularity tendency. It uses these cues to route computation at macro, mid, and micro scopes. Cue-derived scores first select expert groups; within each selected group, the current backbone state then refines expert selection. Across six public datasets and 18 dataset-metric combinations, RouteRec ranks first in 12 and second in three, yielding the best overall average rank of 1.61 compared with 4.11 for the next-best baseline. Additional analyses suggest that the behavioral cues guide expert allocation beyond added capacity and produce routing patterns aligned with observed behavior. Our code is available at this https URL
[IR-14] When LLM -Inferred User Context Adds Value in Production Streaming Recommendation
链接: https://arxiv.org/abs/2609.38999
作者: Milad Sabouri,Neeraj Sharma,Sardar Hamidian,Shaghayegh Agah
类目: Information Retrieval (cs.IR)
备注:
Abstract:Contextual information in recommender systems is shifting from static, predefined variables toward latent representations inferred from behavior. Large language models support this shift by rendering an unstructured interaction history as a natural-language summary, which yields a thematic user context that can be encoded and used in place of an aggregate profile. The conditions under which such generated profiles outperform aggregate embeddings have received limited characterization mainly at the domain level. We evaluate semantic user-profiling strategies on a production streaming platform, ranking against the full catalog. The evaluation covers a 2*2 design space crossing representation type (aggregate or LLM-generated) with contextual scope (holistic history or attention-fused short-term and long-term contexts). The relative ordering of the two representation types is conditional on the user’s consumption regime. Aggregate profiles are consistently stronger under habitual consumption, which characterizes approximately four-fifths of the population, while LLM-generated profiles are stronger for exploratory users whose subsequent interactions diverge semantically from their history. We also observe a popularity-attractor effect in LLM-generated profiles, which modestly raises within-list diversity while substantially lowering catalog coverage and reducing novelty. These results indicate that a context-aware system can select a profiling strategy from the inferred consumption regime rather than applying one representation to all users.
[IR-15] xt-Video Retrieval via Multi-Dimensional Saliency Assessment and Granularity-Aware Query Decomposition
链接: https://arxiv.org/abs/2609.38949
作者: Shuquan Wei,Xi Chen,Xu Chen,Xiangyang Jia
类目: Information Retrieval (cs.IR)
备注:
Abstract:Text-video retrieval, which aims to bridge visual and textual modalities by learning a joint embedding space, has become a crucial task in multimodal intelligence. Despite extensive efforts to mitigate visual redundancy, previous methods typically rely on a single-aspect criterion to assess visual importance, overlooking the multifaceted spatiotemporal nature of video. In addition, encoding text into a single global embedding to align with videos compresses temporal events and spatial entities into a unified representation space, further aggravating cross-modal misalignment. To address these issues, we propose MMTI, a method that jointly mitigates visual redundancy and enables multi-grained text-video interaction to achieve accurate multi-grained semantic alignment. Specifically, a key feature selection (KFS) mechanism adaptively identifies and aggregates informative frames and patches by jointly evaluating multi-dimensional saliency and learnable importance scores, effectively compacting dense visual features and mitigating visual redundancy. Furthermore, our proposed multi-grained text-video interaction module (TVIM) employs a dynamic gating mechanism to decompose the text query into sentence, frame, and patch queries (SFP), enabling multi-grained text-video alignment. Complementary alignment at different granularities is thereby achieved. Extensive experiments on four standard benchmarks demonstrate that our method outperforms state-of-the-art methods.
[IR-16] Breaking News Out of the Filter Bubble: Generative AI Search Diversifies Collective Attention and Raises Shared Information Consumption
链接: https://arxiv.org/abs/2609.38946
作者: Heeseung Andrew Lee,Dokyun Lee,Gwanhoo Lee,Dongwon Lee
类目: Computers and Society (cs.CY); Human-Computer Interaction (cs.HC); Information Retrieval (cs.IR); Multimedia (cs.MM)
备注: 31 pages, 4 figures; includes supplementary material
Abstract:Generative AI search and AI overviews are transforming access to information and news, renewing concerns that readers will encounter a narrower range of topics and have less in common. We examine these concerns via a randomized field experiment with 37,561 readers at The Washington Post. Both groups searched the same archive, but treatment readers also received AI answers with article citations above conventional results. Measuring consumption across displayed answers and opened articles, we find that AI search expands the reach of widely read topics and increases overlap in readers’ topic consumption. At the same time, consumption becomes less concentrated and shifts toward less-popular topics, both within readers and across the audience. AI answers account for most of the increase in shared information, delivering it without requiring article clicks and broadening exposure beyond the articles readers open. Cited articles also contribute to the shift toward less-popular topics. Readers shift from conventional-result clicks and browsing toward cited articles and follow-up searches. More frequent searching offsets lower article consumption per search, producing a small increase in article consumption per reader. Total information consumption per minute also rises. Generative AI search can thus diversify collective attention while strengthening the information readers have in common.
[IR-17] RACE: Target-Aware Retrieval Attributed Evidence and Contract-Constrained Extraction for LitTraceQA EMNLP2026
链接: https://arxiv.org/abs/2609.38861
作者: Sachin Gupta,Divya Godara
类目: Computation and Language (cs.CL); Digital Libraries (cs.DL); Information Retrieval (cs.IR)
备注: 8 pages, 3 figures, 3 tables. Accepted at the 1st Workshop on Grounding Language Models: Learning Faithfully and Efficiently (GroundLM 2026), co-located with EMNLP 2026
Abstract:Finding a relevant paper is not the same as producing a verifiable answer from it. LitTraceQA requires canonical paper identifiers, exact evidence at the page or object level, and typed answers that match the evaluator. We call the separation between source access and scorer-visible correctness the grounding contract gap. TRACE - Target-Aware Retrieval, Attributed Evidence, and Contract-Constrained Extraction - addresses this gap with target-grouped retrieval, independent typed evidence localization, multimodal table extraction, schema-driven table construction, and fail-closed validation. It indexes 27,487 papers through passage, object, alias, citation, and dense representations while retaining the question target behind each signal. For tables, TRACE predicts the observation unit before extracting values and assembles rows with evaluator-compatible key normalization. Our audited selected clean-track artifact scores 0.760613 on the official 71-question test set, including 0.9728 paper F1, 0.6847 evidence F1, 0.9800 multiple-choice accuracy, 0.5423 table-row F1, and 0.3508 macro cell accuracy. On 11 public-development table records, a clean baseline and coordinate-aware visual fill obtain row F1 of 0.291 and 0.411, respectively; this diagnostic comparison includes fallback outputs and is not an official-test claim. Remaining errors chiefly concern locator, observation-unit, row-key, and source-value identity.
[IR-18] PatchHolmes: Agent ic Patch Retrieval via Listwise Selection ATC AACL
链接: https://arxiv.org/abs/2609.38807
作者: Guanqun Yang,Yingming Zhou,Jiangrui Zheng,Shudong Hao,Xueqing Liu
类目: Information Retrieval (cs.IR); Software Engineering (cs.SE)
备注: Accepted at AACL-IJCNLP 2026. Code at this https URL
Abstract:Patch retrieval, the task of finding the commit that fixes a known vulnerability, is the foundation of vulnerability management workflows, yet 60% to 63% of CVEs in the major advisory databases lack a patch link. We present PatchHolmes, a two-phase patch retrieval system that pairs a hybrid first-stage retriever with an agentic second-stage inspection loop. Unlike pointwise prior work that scores each candidate independently, the Phase 2 agent reads the top-100 listwise: it sees the full candidate list at once and selectively reads 3 to 10 commits through four budgeted tools before submitting a single best commit. On GitHubAD, PatchHolmes beats the pointwise binary classifier Favia by 25.34% Recall@1 and the retrieve-and-CoT baseline IRCoT by 31.40%, at one agent conversation per CVE versus Favia’s ten; with the candidate set held identical, the agent adds 27.32% Recall@1 over taking the retriever’s top candidate, and the same agent, transferred unchanged to PatchFinder_top10, lifts Recall@1 from PatchFinder’s own top-1 pick (24.28%) to 39.86%. Swapping the LLM backbone within the Qwen family changes Recall@1 by under 1%, and a second model family (gpt-oss) stays far above the no-agent floor, so the gain comes from the listwise agent loop; the entire system runs on a frozen open-weight model over a local Git repository, without fine-tuning or external search APIs.
[IR-19] Learning to Route in Visual Space via Multi-Step Embedding Retrieval
链接: https://arxiv.org/abs/2609.38743
作者: Tianyu Chen,Mingyuan Zhou,Jiaxing Wu
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注:
Abstract:LLM agents rely on retrieval tools to access external knowledge, yet visual agentic search remains severely bottlenecked by standard single-step retrievers. In current pipelines, the agent must issue text queries for every intermediate step, struggling when visual clues are difficult to describe or when the retriever fails to surface necessary intermediate evidence within its top results. We hypothesize that offloading multi-step navigation across the entire embedding space directly to the retrieval tool resolves this performance bottleneck. To study this systematically, we introduce VHOP, a flexible data generation framework and benchmark with five core difficulty levels testing both visual matching and search planning. Using this framework, we develop VHOP-Router, an end-to-end training pipeline—combining supervised fine-tuning, online imitation learning, and reinforcement learning—that transforms a standard embedding model into an autoregressive multi-step retriever. Operating directly in the visual latent space, VHOP-Router retrieves linked image chains in a single tool call without requiring the agent to formulate intermediate text queries. Experiments show VHOP-Router boosts retrieval performance from under 5% to 76.3%. In agentic search, it improves task success rates by 52.7% and reduces the average token length by 61% from 1886 to 728, whereas upgrading the agent yields only a 3.7% gain. Compared to a strong baseline where the agent retrieves the top 50 results per step, VHOP-Router maintains superior performance while reducing in-context images by 23\times and cutting the cumulative API payload by 35\times . The models also generalize robustly to unseen difficulty levels and realistic test sets. Ultimately, VHOP and VHOP-Router provide an efficient and effective solution for visual agentic search that leaves native LLM capabilities entirely intact.
[IR-20] Exploring Forum Post Retrieval with Generative Modeling
链接: https://arxiv.org/abs/2609.38646
作者: Yang Li,Yaguang Liu,Heng Liu,Shengbo Guo,Samson Komo,Jane Kou,Yulian Zhou,Gang Yang,Shubhojeet Sarkar,Gaurav Chakravorty,Yujie Liu,Haipeng Chen,Yonghuan Yang,Deepti Chheda,Yamin Wang,Mike Plumpe,Rish Tandon
类目: Information Retrieval (cs.IR)
备注:
Abstract:Generative recommendation (GR) has emerged as an alternative to embedding-based retrieval, building on the success of generative models in language and vision. We are exploring GR on Facebook Forum, a standalone application for medium-to-heavy users of Facebook Groups. Because Forum is a new surface, its own interaction data are too sparse to train a GR model from scratch. We address this with transfer along two axes: we train on a broader corpus of Facebook Groups engagements rather than Forum sessions alone, and we reuse hierarchical, prefix-based semantic IDs (SIDs) learned from cross-platform Facebook Feed data instead of fitting a Forum-specific tokenizer. A 3B-parameter instruction-tuned language model is then supervised-fine-tuned to generate SIDs directly from user context. We systematically ablate the design choices that matter most in practice, including SID construction, the composition and length of user history, and the inclusion of user-profile features. Our results show that cross-platform SIDs transfer to a new recommendation surface, and offer practical guidance for teams deploying GR on real-world social platforms.
[IR-21] Component-Aware Feedback for Self-Evolving Programs
链接: https://arxiv.org/abs/2609.38639
作者: Ethan Lin,Jinming Nian,Yi Fang
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注:
Abstract:LLM-guided evolutionary search can discover complex programs, but existing methods mostly only save candidate programs and fitness scores while discarding which component edits produced which fitness metric changes. Existing methods force the mutator LLM to infer the effect of prior edits from cluttered histories, making program search slow and unstable. This is especially true for locally servable LLMs to evolve multi-component systems. We introduce component-aware feedback, which compares each evaluated program with its parent, identifies the components that changed, and logs them with the associated metric differences into an attribution memory that later mutations read. The memory keeps each change in two reference frames, local against the parent it came from and global against the seed program, which shows both the immediate effect of a change and the cumulative progress made since the seed. We study this on LLM reranking, a multi-objective optimization problem where a multi-stage pipeline must balance quality against serving cost. Across twelve \textscBright datasets, our method reaches the strongest baseline’s final quality after a median of one third of the search budget and ends 7.2% higher in held-out nDCG@10, and under a cost-aware objective it finds pipelines that are on average more accurate while using 11% fewer tokens per query, showing component-aware feedback to be a promising direction for more efficient self-evolving systems.
[IR-22] Re-ranking and Late Interaction Drive Retrieval Quality: A Controlled Comparison of RAG Strategies for Scientific Question Answering
链接: https://arxiv.org/abs/2609.38473
作者: Bhagyesh Rathi,Eshan Chawla,William B. Andreopoulos
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: on September 21st submitted for consideration to the Elsevier Data and Information Management (DIM) journal (DIM-D-26-00430)
Abstract:Retrieval-Augmented Generation (RAG) is now the standard way to ground Large Language Models (LLMs) in external knowledge, yet the design space of retrieval pipelines is large and the trade-offs between variants are not well understood, especially on domain-specific corpora at realistic scale. In this work, we present a controlled comparison of six retrieval strategies for scientific question answering: (i) classic top-k dense retrieval, (ii) LLM-based query rephrasing, (iii) query rephrasing followed by LLM-based reranking, (iv) multi-query fusion via Reciprocal Rank Fusion (RRF), (v) an agentic tool-call pipeline in which the generator decides for itself whether to retrieve, and (vi) late-interaction retrieval with ColBERTv2. All six pipelines share the same generator (Meta-Llama/Llama-3.1-8B-Instruct), prompt, and evaluation protocol; the five single-vector pipelines additionally share SPECTER2 embeddings and a Chroma vector store; and all six retrieve from the full corpus of 463,971 arXiv papers dated 2024-2025. To support reproducible, large-scale evaluation, we also release a synthetic question dataset of 19,484 problem-statement and methodology questions generated by Llama-3.1-8B-Instruct from a random sample of 10,000 papers across academic domains (query generation succeeded for 9,742 of them), and every strategy is evaluated on this same query set. We describe the architecture and implementation of each pipeline, release the code and the synthetic question dataset, and evaluate each strategy with an LLM-as-a-judge protocol along multiple quality dimensions, together with direct gold-paper retrieval metrics. The result is an open testbed for studying the cost and quality trade-offs of RAG design choices on a research-literature corpus, and a basis for future work on faithfulness, retrieval robustness, and agentic retrieval.
[IR-23] AdaM-Rec: Adaptive Modality Routing for Multimodal Recommendation
链接: https://arxiv.org/abs/2609.38455
作者: Honghao Fu,Jiacheng Chen,Manxi Lin,Junjun Zheng,Xiangheng Kong,Yiwei Wang,Xin Yu,Miao Xu,Yuning Jiang,Yujun Cai
类目: Information Retrieval (cs.IR)
备注:
Abstract:While recent multimodal recommender systems have demonstrated the effectiveness of incorporating visual and textual information to improve downstream performance, most existing methods rely on static modality fusion, assuming that the relative importance of textual and visual signals remains stable across recommendation scenarios. This design may not fully account for an important variation across recommendation requests: some queries require fine-grained visual cues, whereas others are better served by textual or functional semantics, in which case indiscriminate modality fusion brings in uninformative cues and impairs recommendation quality. To address this, we propose AdaM-Rec, an LLM-based framework for adaptive modality routing in multimodal recommendation, which enables dynamic calibration of reliance on textual and multimodal evidence for user-specific queries. Built on structured natural-language representations of items and user preferences, it estimates modality reliability using proxy recall tasks. Specifically, it generates pseudo-queries that match the granularity of the actual query while pointing to the user’s positively interacted items as verifiable proxy targets, evaluating which modality yields better recall performance in analogous scenarios and optimizing the routing strategy in an agentic manner. It then performs routed recall with optimized strategy, enriches results with collaborative items, and ranks candidates by their relevance to both the query and user preferences. Experiments demonstrate that AdaM-Rec delivers strong performance against state-of-the-art baselines, highlighting the effectiveness and broader potential of adaptive control over modality reliance in multimodal recommendation.
[IR-24] Doc2LoRA Provides Decodable Representations of Scientific Ideas
链接: https://arxiv.org/abs/2609.38374
作者: Chand Sahil Mansuri,Joel Zachariah,Sadamori Kojaku
类目: Computation and Language (cs.CL); Digital Libraries (cs.DL); Information Retrieval (cs.IR); Machine Learning (cs.LG); Physics and Society (physics.soc-ph)
备注: 32 pages, 4 figures, 12 tables. Code: this https URL
Abstract:Representing scientific papers as points in a space lets us search for similar papers and inquire about how fields relate to one another and drive innovation. Beyond search, the vector space of papers invites generation: mixing papers through simple vector operations creates new points, mirroring combinatorial novelty, the recombination of existing ideas into new ones. However, a mixed point often represents an idea no paper has yet realized, with no papers nearby to identify the idea. We propose representing each paper by a LoRA adapter generated by the Doc-to-LoRA hypernetwork. Every point in the space, including mixtures, thus represents a large language model (LLM) open to questions and instructions in natural language. On papers from the American Physical Society (APS), we instruct the LLM at the average of each subfield to name the field in a few words and obtain labels closer to the official names than the labels of five baselines, as judged by word overlap and a panel of five LLM judges. We also ask the LLMs at points between two APS papers to write an abstract and obtain descriptions shifting from one paper to the other in step with the mixing weight. While Doc-to-LoRA is trained for generation, a small invertible transform makes the embeddings competitive for search, on par with SPECTER2 and EmbeddingGemma and close to SBERT. Because the transform is invertible, every point in the transformed space still maps back to an LLM. The embeddings thus serve both search and generation, enabling researchers to question the idea at any point in the space as a starting point for generating new ideas.
[IR-25] AGGRAPH: Tag-Augmented Graphs for Graph Retrieval of Agent Persistent Histories
链接: https://arxiv.org/abs/2609.38353
作者: Yu-Su Chen,Yu-Jung Liang,Pengtao Xie
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注: An earlier version was accepted at the COLM 2026 Workshop on Lifelong Learning Agents (LLA)
Abstract:Long-term memory lets LLM agents recall past interactions and remain consistent across sessions, but memory systems are hard to compare because they often vary in representation, indexing, retrieval, and evaluation. We present a controlled evaluation framework based on shared 5W-style conversational memories. Localized graph configurations traverse a common base graph; AdaptiveGraph adds chronological edges and Personalized PageRank diffusion. We also evaluate BM25 over the same extracted notes and OpenClaw as a raw-input external reference. Retrieval rankings vary across memory settings. On LongMemEval-S, AdaptiveGraph is the strongest graph configuration at 0.844 MRR, but BM25 reaches 0.867 and OpenClaw 0.880. On ATANT Core, localized graph traversal outperforms diffusion and BM25, whereas BM25 leads the stress rounds. Reducing LongMemEval-S within the tested range does not reproduce the ATANT diffusion penalty, but the smallest tested store remains larger than ATANT Core, so store size cannot be ruled out. The penalty also persists under a permissive content-match criterion. Vocabulary normalization and extraction quality substantially affect graph retrieval, and missing extraction tags are common among top-five misses. Retrieval strategies should therefore be evaluated jointly with the memory setting and against strong lexical baselines.
[IR-26] Privacy in Personalized AI Is a System Property Not Just a Model Property NEURIPS2026
链接: https://arxiv.org/abs/2609.38289
作者: Guillaume Salha-Galvan,Jiaying Xu
类目: Cryptography and Security (cs.CR); Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: NeurIPS 2026 Workshop on Privacy in the Era of Large Opaque Models
Abstract:In personalized AI applications, such as conversational assistants and recommender systems, users interact not with models in isolation but with broader systems that access, infer, and reuse user information across components and over time. While such use of user information is integral to personalization, it also raises important privacy questions. In this paper, we argue that individual model- or component-level analyses may not capture all privacy risks arising in such systems, motivating a system-level perspective on privacy. We distinguish and analyze four interconnected privacy-risk channels in personalized AI, and subsequently propose four requirements for system-level privacy evaluation, covering interaction trajectories, internal information flows, indirect leakage, and the privacy-utility trade-off. We argue for their systematic incorporation into privacy audits of personalized AI.
[IR-27] LegalPincite: Multi-level Legal Information Retrieval Dataset EMNLP2026
链接: https://arxiv.org/abs/2608.03756
作者: Theresia Veronika Rampisela,Henrik Palmer Olsen,Giovanni Colavizza
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL)
备注: Accepted for publication at the 8th Natural Legal Language Processing Workshop (NLLP 2026), co-located with EMNLP 2026
Abstract:A common task in legal Information Retrieval (IR) is to find relevant legal sources from case-law collections. While legal practice often requires pinpoint citations (pincites) to specific case paragraphs, most existing public legal IR datasets lack paragraph-level citation annotations. Yet, publicly available datasets with such information contain data leakage in the query text and exclude paragraphs that are neither citing nor cited from the corpora, creating an unrealistic and oversimplified retrieval setting, potentially leading to inflated performance. To address these limitations, we contribute a large-scale legal IR dataset constructed from Court of Justice of the European Union (CJEU) judgments. The dataset contains: (i) masked case/paragraph queries, with removed citation information; (ii) a corpus that includes all paragraphs; and (iii) case- and paragraph-level ground truth citations, with partial human expert validation. Our dataset supports both the development and rigorous evaluation of legal IR methods, at multiple query-document levels (case-to-case, paragraph-to-case, and paragraph-to-paragraph retrieval). Link to dataset and code: this https URL
[IR-28] Life-Bench: A Benchmark and Knowledge Graph Framework for Multimodal Personalization Beyond Concept Recognition
链接: https://arxiv.org/abs/2602.19001
作者: Xia Hu,Honglei Zhuang,Brian Potetz,Alireza Fathi,Bo Hu,Babak Samari,Howard Zhou
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注:
Abstract:As large language models increasingly power personal assistants, users expect them to reason over multimodal life histories, from recognizing people to understanding events to aggregating patterns, yet existing benchmarks primarily target concept-level recognition. We introduce Life-Bench, a fully synthetic, human-verified multimodal benchmark of over 11,800 question-answer pairs across 10 tasks, organized by required evidence scope: concept identification, event understanding, and aggregated reasoning. The benchmark’s photo-centric personal histories are distributionally aligned with real user accounts under embedding statistic. The interconnected structure of personal data invites graph-based solutions; we propose LifeGraph, a personal knowledge graph framework providing structured retrieval with on-demand access to source visual evidence, showing particular promise on event and aggregated tasks. Systematic evaluation of four retrieval paradigms on Life-Bench demonstrates that accuracy degrades sharply with evidence scope, falling below 0.40 on aggregated tasks, and that no single paradigm dominates across categories. Performance beyond concept recognition remains modest for all evaluated methods, establishing personalization over multimodal histories as an open challenge and Life-Bench as a testbed for future progress.
人机交互
[HC-0] Rethinking Legibility in Social Robot Hallway Navigation: Impact of Intent Representation and Human Distraction
链接: https://arxiv.org/abs/2609.40158
作者: Pranav Goyal,Andrew Stratton,Christoforos Mavrogiannis
类目: Robotics (cs.RO); Human-Computer Interaction (cs.HC)
备注: 24 pages, 6 figures
Abstract:We focus on legible robot motion generation in social navigation settings. Legibility in human-robot interaction (HRI) is often described as the property of robot motion that enables an observer to confidently infer the robot’s intent. While mature frameworks exist for generating legible motion in front of static observers, social robot navigation presents a new challenge: the robot must clearly convey its intent while ensuring human safety in dynamic pedestrian environments where human attention is often divided. With the goal of enabling robots to generate legible motion in dynamic and constrained spaces, we investigate how the choice of representation and the level of human attention shape navigation performance and human impressions. Focusing on the ubiquitous and demanding scenario of hallway navigation, we conduct two controlled user studies involving alternative legibility formulations implemented within a shared model predictive control framework. Study 1 (N = 45) investigates the role of intent representation, showing that passing-side legibility, particularly when adaptively updated, leads to smoother human motion and is perceived as more competent and less mentally and physically demanding than destination-based and non-legible baselines. Study 2 (N = 45) examines the effect of pedestrian attention, demonstrating that legible motion allows for smooth human motion even under distraction, even if this is not consistently reflected in subjective ratings. Together, these findings suggest that effective legible motion in social robot navigation benefits from interaction-level intent representations that support coordination, with some effects persisting even when human attention is divided. Code is available at this https URL.
[HC-1] Who Asked for This? Inline Annotations as Authoring Transactions for Provenance in Agent ic Authoring
链接: https://arxiv.org/abs/2609.40126
作者: Chang Xiao
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Writing with AI agents turns a paragraph into the outcome of many requests, yet the finished document rarely explains which request produced which change. We introduce Reactant, an interaction paradigm in which authors place typed inline annotations in their original documents. A verified transaction protocol records each request, skill identity, and the correspondences that identify additions, deletions, and transformations. The kernel validates the witness against the recorded states to establish word-level longitudinal lineage. We demonstrate Reactant through this paper’s revision history, and report four months of three colleagues’ self-directed use. Their uses include conversational inquiry into history and deriving reusable skills from recurring requests, illustrating how the transaction record serves as an extensible substrate for agentic authoring.
[HC-2] JuryFlow: Disagreement-Guided Human-in-the-Loop Multi-Agent Evaluation
链接: https://arxiv.org/abs/2609.40103
作者: Mufeng Yang,Junwei Yu,Yepeng Ding
类目: Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注: 9 pages, 5 figures, 5 tables. To appear in Proceedings of the 14th International Conference on Human-Agent Interaction (HAI '26), November 16-19, 2026, Osaka, Japan
Abstract:Large language models (LLMs) are increasingly deployed as automated judges for AI-generated content, yet a single judge is unreliable and even a panel of judges leaves a hard residue: when judges disagree, majority voting discards the conflict instead of resolving it. We present JuryFlow, a disagreement-guided, human-in-the-loop multi-agent evaluation framework that treats inter-judge disagreement not as noise to be averaged away, but as a precise, claim-level signal indicating where an evaluation is uncertain. JuryFlow decomposes each candidate response into atomic claims, has a panel of heterogeneous judges assign per-claim verdicts, and builds a disagreement graph whose nodes are scored by verdict entropy and whose edges encode structural similarity between claims. A human acts as a structural guide, selecting which disagreement to resolve through a single, minimal intervention rather than re-labeling the response, after which the focal claim is re-evaluated, the correction propagates along graph edges and to historically similar cases, and is crystallized into reusable rubric entries that all judges inherit, making the evaluator progressively self-refining. To enable large-scale, reproducible benchmarking without human studies, we evaluate JuryFlow in an automatic configuration in which focal selection is made by entropy ranking. On MT-Bench and LLMBar, JuryFlow improves agreement with gold labels over single-judge and majority-vote panel baselines, and ablations isolate the contributions of disagreement-targeted re-evaluation, propagation, and rubric induction. We contribute (1) a human-in-the-loop paradigm that recasts the human from labeler to structural guide, (2) the JuryFlow framework operationalizing it through a disagreement graph, focal re-evaluation, and closed-loop rubric induction, and (3) an evaluation protocol with ablations that isolate where the gains originate.
[HC-3] Richard: Voice-First Mobile Interaction for Persistent Tasks
链接: https://arxiv.org/abs/2609.39976
作者: Xinyang Chen
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注:
Abstract:Mobile terminals need to provide application and network services while supporting users’ control over their attention. We explore voice-first interaction organized around requests and delegated tasks, allowing users to leave a conversation and later inspect, revise, and retrieve the work. We present Richard, a system prototype that manages voice sessions, task execution, and result delivery separately, linking them through persistent request records. Conversation and task views provide visual feedback, while the backend coordinates immediate responses, dedicated service operations, and agent tasks. Request revisions, execution states, and notifications remain associated with the relevant task. We examine this design through Android functional records, controlled lifecycle verification, and execution records of a real programming request. Controlled verification reproduces revision, execution after confirmation, and result retention; deployed-service records show backend progress and failure feedback after client disconnection. These observations inform the design of task continuity, user control, and service integration in mobile voice interaction, providing an implementation basis for personal computing devices that accommodate intermittent user participation.
[HC-4] Diptych: Scoped AI-Interpreted Comparison for Reference Listening in Music Production
链接: https://arxiv.org/abs/2609.39963
作者: Chongjun Zhong,Abhinaba Roy,Archishman Ghosh,Kejun Zhang,Dorien Herremans
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Reference listening is a common strategy in music production, but current comparison tools often obscure a key human judgment: deciding what should be compared. We present Diptych, an AI-assisted system that lets users define comparison scope across whole tracks or independently selected segments, while inspecting structured audio features and scope-specific AI interpretations. We evaluated Diptych in a within-participants study with 12 musicians, complemented by source-blinded ratings from four expert listeners. Participants used the system to surface additional differences, nine of ten of which received at least partial expert support, and reported good usability and greater clarity about possible next steps. These findings suggest that AI support for creative comparison should prioritize user-defined scope, inspectable evidence, and actionable guidance, while avoiding authoritative judgments that exceed what the evidence can support.
[HC-5] Understanding Parents Complex Views of AI for Childrens Pretend Play
链接: https://arxiv.org/abs/2609.39906
作者: Sungho Oh(Sander Oh),Mohammad Namvarpour(Matt Namvarpour),Maxi Heitmayer,Minahil Khalid,Afsaneh Razi
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注:
Abstract:AI could support children’s pretend play, but it could also direct the play on behalf of children. Whether AI should have roles in children’s lives is controversial because its influence on children remains uncertain. We conducted semi-structured interviews with 10 U.S. parents, each with at least one child aged 4-15. During the interview, we described the concept of AI-supported pretend play and provided participants with two boundary-case storyboards. We analyzed the interview data through codebook thematic analysis, using inductive coding and affinity diagramming organized around the research questions, and then used qualitative systems mapping to examine relationships within and across themes. We found that the same characteristics of AI, e.g., ability to assume characters, responsiveness, and adaptability, were seen by parents as potentially useful but also concerning. Parents imagined that AI could make role-based play accessible to all children or help parents participate in family play. However, they opposed the idea of AI for children’s play without a clear understanding of how it works and its long-term influence on their children. Parents worried about children’s loss of imagination and creativity, emotional attachment to AI, reduced human interaction, inappropriate behavior by AI and/or children, and their inability to manage children’s AI use. Parents viewed AI not only as a play tool but also as a social actor and a possible perturbation in the existing family dynamics. The appropriateness of AI and child–AI interactions therefore emerged as a requirement for AI in children’s pretend play, in addition to technical safeguards and parental control. We contribute an integrated account of parents’ interdependent judgments and emphasize the need for longitudinal research with children and their diverse families.
[HC-6] When a Kindergartener Solves Calculus: Measuring Capability Leakage in Role-Prompted Reasoning Models
链接: https://arxiv.org/abs/2609.39846
作者: Pakhapoom Sarapat,Saksorn Ruangtanusak,Kunat Pipatanakul,Pittawat Taveekitworachai
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注:
Abstract:We investigate the problem of role-capability leakage (RCL), in which a role-prompted reasoning model generates convincing in-role text while continuing to exhibit capabilities on benchmarks that exceed those implied by the assigned role. For example, when a model is prompted to assume the role of a kindergarten student, one might expect its performance on a mathematics benchmark to reflect kindergarten-level ability rather than expert-level proficiency in solving calculus problems. We introduce RoleCapBench, a curriculum-grounded benchmark for evaluating RCL across six educational roles and four assessment levels spanning elementary school through A-level, and use it to evaluate three open-weight reasoning models. We find that although the models can generate stylistically convincing in-role responses, they consistently fail to align their underlying capabilities with their assigned roles. Naive role prompting yields strong role-voice scores of 1.218–1.389 while retaining above-role accuracy of 0.811–0.898. RCL persists across a range of prompting conditions, including prompts that explicitly instruct the model to match the role’s capability level. To mitigate this problem, we propose Injection, an inference-time intervention that combines explicit, role-specific capability guidelines with a guiding prefilled response prefix. Injection improves role-capability alignment across models, reducing above-role accuracy by up to 0.562 while preserving in-role accuracy with a marginal drop of less than 0.058 across most models. All artifacts, including scripts and evaluation data, will be released upon acceptance.
[HC-7] Making Sense of Animal-to-Human Drug Development Evidence: Stakeholder Practices Challenges and Requirements for AI Tools
链接: https://arxiv.org/abs/2609.39808
作者: Rosni Vasu,Simona E. Doneva,Benjamin V. Ineichen
类目: Human-Computer Interaction (cs.HC)
备注: 24 pages, 5 Figures, 12 main pages
Abstract:Animal models are widely used to study human biology and health interventions, yet translating findings from animal studies to humans remains challenging. Evidence across preclinical and clinical research informs experimental and translational decisions. Artificial Intelligence (AI) tools are increasingly reshaping how this evidence is searched, synthesized, and used, but it remains unclear how they should support the diverse stakeholders involved in assessing animal-to-human evidence. We conducted semi-structured interviews with 13 stakeholders to examine their evidence practices, challenges, and expectations for AI support. We found that stakeholders approach the same incomplete evidence base with different goals, expertise, and heuristics. Participants valued AI particularly for locating, screening, and extracting evidence, but were more cautious about automated interpretation and quality judgments. They emphasized transparency, source traceability, uncertainty communication, and human oversight. Based on these findings, we derive design implications for role-sensitive AI tools that support more systematic and transparent reasoning about animal-to-human translation.
[HC-8] Engagement-Led Segmentation of Gamified Participation Data in a Large-Scale Remote Internship: A Mixed-Methods Study
链接: https://arxiv.org/abs/2609.39750
作者: Sakshi Sharma,Pavani Ayinampudi,Aditya B.M.V.,Jinal Gupta,Prakash Hegade,Rohit Sharma,Meenakshi V,S.R.S. Iyengar
类目: Human-Computer Interaction (cs.HC); Computers and Society (cs.CY)
备注: 19 Pages, 2 figures, Poster accepted for T4E
Abstract:Gamified points are widely used to represent learner participation, but cumulative point totals provide limited information about how participation changes over time. This limitation matters in continuously enrolling programmes, where learners have different opportunities to accumulate points. This study examines how gamified participation data can be interpreted as indicators of behavioural engagement in a large-scale, continuously enrolling remote internship. Using an anonymised operational dataset, 3,607 distinct started learners were first classified by participation status, separating dormant learners from those with observable participation. The 1,871 active learners were then grouped using K-means clustering on normalised longitudinal and opportunity-aware participation features. Four participation patterns emerged: Thriving Core, Tapering, Fast Starters, and Occasional Participants. Their trajectories differed in both level and direction. A perception survey of 597 respondents provided complementary learner-reported evidence, integrated with the behavioural strand through a joint display by segment. Perceptions differed across segments, while individual-level correlations between perceptions and behaviour were small, and learners with low recorded participation could report positive perceptions of the points system alongside external barriers. Segments assigned from the first four weeks were then checked against eleven weeks of later platform records: 94% of the Thriving Core attended at least ten further sessions and 22% completed the internship, against under 7% and under 1% in the low-engagement segments. The findings suggest that gamified participation data are more informative when interpreted as longitudinal, opportunity-aware behavioural indicators rather than cumulative scores alone.
[HC-9] Referential Uncertainty in Human–AI Collaboration
链接: https://arxiv.org/abs/2609.39518
作者: Christian Poelitz,Finale Doshi-Velez,Siân Lindley
类目: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注:
Abstract:Effective human-AI collaboration requires partners to establish references through interaction, which becomes fragile when descriptions are ambiguous, similar referents compete, or partners see different things. We study referential uncertainty - uncertainty over which candidate object a description refers to - in a collaborative puzzle task where a human Helper instructs an AI Worker to place pieces. The Worker must identify and communicate its uncertainty, and the Helper must recognize and act on it. We show that a separately elicited belief distribution over candidate pieces is better calibrated (ECE 0.15) and better discriminates correct from incorrect placements (AUROC 0.65) than raw action-token probabilities, which are severely overconfident (0.97 mean confidence, ECE 0.44). Across three frontier vision-language models (GPT-4.1, GPT-5, GPT-5.5), this elicited uncertainty rises predictably with instruction vagueness, but not with competing referents in context, even when those increase errors. The models seldom externalize it, asking for clarification on only 3.5-16.7% of turns. In a controlled human study (N=210), participants given only the Worker’s default message accept 78% of wrong placements and cannot tell right from wrong (AUC 0.50). Precise descriptions and, especially, well-targeted hedges cut wrong-move acceptance to 36% while largely preserving correct-move acceptance, compensating for missing shared awareness such as not seeing the Worker’s action. But this benefit depends on targeting: a deployable hedge derived from the model’s own belief entropy inherits that signal’s weakness and can do more harm than good. Externalized uncertainty helps a human partner only when it is accurately targeted.
[HC-10] DuplexAct-Bench: Broadening Full-Duplex Speech Evaluation toward Proactive Interaction across Diverse Behavioral Requirements
链接: https://arxiv.org/abs/2609.39446
作者: Keyue Xing,Wentao Ding,Mengmeng Wang,Wenming Tu,Zilong Zheng,Yipeng Kang
类目: Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注:
Abstract:Existing full-duplex speech benchmarks cover only subsets of real-time interaction behaviors, often under limited contextual conditions. We introduce DuplexAct-Bench, a bilingual benchmark that systematically covers six complementary behaviors, from interruption and yielding to proactive initiation, active silence, and backchanneling, across Pre-session, In-session, and No-explicit conditions. Across 1,290 English and Chinese streaming trials, we evaluate 12 full-duplex speech systems on both Timing and Content. Results reveal substantial variation across behaviors, conditions, and systems, as well as frequent mismatches between semantic quality and behavioral timing. These findings show that current systems remain far from robustly managing when, whether, and how to participate as real-time interaction unfolds. Project page: this https URL
[HC-11] IDEAL: A Multimodal Domain Adaptation Framework for EEG-Eye Emotion Recognition
链接: https://arxiv.org/abs/2609.39421
作者: Yang Wu,Jinpeng Li
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Electroencephalography (EEG) emotion recognition serves as a pivotal interface for human-computer interaction, yet the physiological variability across individuals complicates the already challenging task of fusing heterogeneous physiological signals (e.g., EEG and eye movements). However, most prevalent domain adaptation paradigms are tailored for unimodal scenarios, failing to address the heterogeneity of multimodal signals. Furthermore, they predominantly rely on feature-level alignment, overlooking the fundamental data-level discrepancy, which risks compromising fine-grained discriminative information during aggressive adaptation. To bridge these coupled gaps, we propose Instance-based Domain Expansion and Adversarial Learning (IDEAL), a unified framework that synergizes instance-level curriculum expansion with feature-level hierarchical adversarial alignment. IDEAL first introduces a multi-model collaborative screening mechanism, which propagates high-confidence target samples to explicitly bridge the distributional gap at the data level via a quantity-quality equilibrium strategy. We provide a theoretical analysis that this instance expansion strategy strictly tightens the upper bound of the target risk. Subsequently, a hierarchical adversarial network, augmented with the angular-contrastive constraints, progressively aligns representations from low-level statistics to high-level semantics while preserving class separability. Extensive experiments on four benchmark datasets demonstrate that IDEAL significantly outperforms state-of-the-art methods. To facilitate reproducibility and future research, our source code is publicly available at this https URL.
[HC-12] Aligning the Incomplete: Joint Distribution Calibration for Multimodal EEG-Eye Emotion Recognition
链接: https://arxiv.org/abs/2609.39413
作者: Yang Wu,Jinpeng Li
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:The success of cross-subject multimodal emotion recognition hinges on maintaining the consistency of the joint data distribution across individuals. However, real-world deployment frequently triggers the \emphasymmetric joint distribution collapse: EEG signals suffer from severe cross-subject distribution shifts, while eye movements sensors are susceptible to packet loss and tracking failures. Existing methods treat domain adaptation and missing-modality imputation as disjoint tasks. Consequently, they fail to resolve the compounded errors when both degradations co-occur, either propagating domain shifts through imputed signals or destroying the joint decision boundary. To tackle this unified challenge, we propose GUARD (\textbfGradient-guided \textbfUnsupervised \textbfAsymmetric \textbfRecovery of \textbfDistributions). First, GUARD establishes a reliable anchor manifold in the source domain by employing a theoretically grounded gradient-weighted objective, which forces the robust EEG modality to preemptively entangle task-discriminative ocular features. Next, to structurally recover the collapsed joint distribution, we constrain a generative module with downstream perceptual losses, prioritizing emotion-discriminative semantics over mere signal fidelity. Finally, we formulate target-domain adaptation as an ill-posed inverse problem. By driving a cycle-consistent flow, we achieve unsupervised calibration of the recovered joint distribution directly on the target-domain manifold. Extensive experiments demonstrate that GUARD significantly outperforms state-of-the-art methods, maintaining resilient discriminative performance even under complete auxiliary modality failure. Our code and models are made publicly available to ensure complete reproducibility.
[HC-13] NarrativeSteward: Coordinating Delegation Guidance and Verification in Agent -Assisted Interactive Narrative Authoring
链接: https://arxiv.org/abs/2609.39333
作者: Wenjin Wang,Jiazhen Lei,Yuxin Sha,Nuwa Xi,Meng Zhao,Xingxi Yin,Qi Liu,Yuliang Shen,Zixun Sun
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Autonomous AI agents can turn authors’ goals into interactive narratives by independently organizing and carrying out generation and revision. As agents generate and revise extensive content, authors struggle to grasp its overall structure, local details, and relationships, complicating continued guidance. We present NarrativeSteward, an authoring environment that organizes outlines, worldbuilding, and narrative graphs as linked artifacts for agent implementation and author guidance. Agent dialogue and project-wide structural review help authors understand the evolving work and guide local and cross-layer revisions, while change records and execution verification help authors assess the resulting work. Technical tests validated the system’s change records, recovery mechanisms, and execution diagnostics. In a 12-participant within-subject study, NarrativeSteward supported easier formulation of revision requests and inspection of changes, and greater perceived understanding of changes and story structure, than general-purpose agents. Qualitative findings show how reviewing the work and feedback helps authors develop requirements and guide subsequent delegation. We open-source NarrativeSteward at this https URL.
[HC-14] When Data Becomes Judgment: Misaligned Interpretations and Accountability in Food Delivery Platforms
链接: https://arxiv.org/abs/2609.39311
作者: Yibo Meng,Shuoning Shi,Bingyi Liu,Hongkai Chen
类目: Human-Computer Interaction (cs.HC)
备注: 27 pages, to be published at ACM IMWUT
Abstract:Food delivery platforms mediate service encounters through real-time tracking data. While presented as objective, such data often obscures the situational constraints shaping delivery work, producing systematic misalignments between user perception and courier experience. This study presents a socio-technical analysis of real-time mobile tracking systems in the wild. Through semi-structured interviews with 23 users and 17 couriers on Chinese food delivery platforms, we identify two interrelated dynamics. First, users operate within a data-as-behavior interpretive framework, translating spatial and temporal anomalies into moralized judgments of courier negligence. Second, couriers engage in anticipatory data management, a form of hidden digital labor in which they reshape their physical behavior to produce interface-legible trajectories rather than physically optimal ones. Together, these findings expose a burden-shifting mechanism—characterizing the systemic outcomes of decontextualized interface design rather than explicit designer intent—in current tracking architectures, demonstrating how current tracking architectures leave gig workers bearing much of the explanatory burden. We propose design directions toward contextual transparency, redistributing this explanatory burden from individual workers to the platforms that possess the logistical context to bear it.
[HC-15] Present After Presence: Subtraction Givenness and the Structure of Being-There
链接: https://arxiv.org/abs/2609.39289
作者: Koichi Toida
类目: Human-Computer Interaction (cs.HC)
备注: 9 pages
Abstract:Presence research has often proceeded by addition, treating immersion, embodiment, agency, ownership, co-presence, reciprocity, and temporal simultaneity as conditions that stabilise the sense of being there. Yet immersive video suggests that several of these conditions can be weakened without eliminating Presence. Building on Bodyless Presence and Bodyless Presentness, this paper asks what is disclosed when such conditions are progressively subtracted. I distinguish three levels that should not be conflated: stabilising conditions that strengthen or support Presence; empirically resilient articulation-forms through which Presence is lived as here and now; and a transcendental limit-condition concerning the first-personal givenness of experience. Subtraction provides evidence for the resilience of here and now, but empirical resilience does not by itself establish constitutivity or transcendental necessity. I argue, on phenomenological rather than experimental grounds, that for-me-ness names the limit-condition within which any such spatial or temporal articulation can be experienced at all. The paper bridges operational Presence research with Husserlian givenness, Leibhaftigkeit, and image consciousness; rereads Bodyless Presence as exposing the resilience of here; rereads Bodyless Presentness as exposing the resilience of now; and develops a stratified account of Presence through Zahavi’s pre-reflective self-awareness while taking Derrida’s critique of self-presence seriously. It concludes by proposing immersive media as dissociation apparatuses for loosening conditions ordinarily coupled within experiential there-ness.
[HC-16] Perceptual Color Difference Modeling Using Machine Learning and Human Similarity Judgments
链接: https://arxiv.org/abs/2609.39130
作者: Elnara Kadyrgali,Muragul Muratbekova,Adilet Yerkin,Nuray Toganas,Ayan Igali,Malika Ziyada,Aruzhan Burambekova,Jamaladdin Hasanov,Pakizar Shamoi
类目: Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
备注: This manuscript has been submitted to IEEE Access for consideration
Abstract:Accurate assessment of color differences is essential for applications ranging from digital design to quality control. While existing color difference metrics, such as CIEDE2000, aim to approximate human perception, they may still exhibit inconsistencies with perceptual judgments. In this study, we investigate a data-driven approach to color-difference estimation based directly on human evaluations. We collect similarity judgments for 2,000 systematically generated color pairs, each rated by seven observers using a four-point ordinal scale. These judgments are then used to train regression models using different color representations, including RGB channel differences, HSI differences, and COLIBRI fuzzy linguistic categories. Experiments with five regression algorithms show that the choice of color model has a greater influence on prediction performance than the choice of regression algorithm. Using COLIBRI features alone, linear regression achieves an R2 of 0.595, outperforming RGB and HSI representations, which achieve R2 values of 0.479 and 0.493, respectively. The best performance is obtained by LightGBM using the combined representation, reaching an R2 of 0.703. The results indicate that human perceptual color differences are better captured when numerical color coordinates are complemented by graded perceptual categories, highlighting the potential of data-driven models for perceptually aligned color-difference estimation.
[HC-17] DiFF: Doppler-informed Flow Matching for Human Motion Flow IROS
链接: https://arxiv.org/abs/2609.39098
作者: Kai Wang,Mingle Zhao
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
备注: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2026. Code: this https URL
Abstract:Perceiving human motion via privacy-preserving 4D millimeter-wave (mmWave) radar is critical for next-generation human-robot interaction (HRI), where point cloud scene flow serves as a foundational motion representation. Yet the extreme sparsity and noise of 4D radar point clouds make non-rigid motion flow estimation severely ill-posed–a challenge that existing rigid-centric methods and prior works fail to adequately address, largely because they neglect the rich Doppler velocity cues inherent in 4D radar. We propose DiFF, a generative framework that marries Doppler-informed motion priors with a Kolmogorov-Arnold Network (KAN)-based conditional flow matching model. At its core, a KAN-attention mechanism enables expressive feature extraction, while a prior-guided generative process harnesses Doppler cues to regularize the ill-posed solution space. Extensive experiments show that DiFF achieves state-of-the-art (SOTA) performance across diverse real-world datasets, reducing 3D endpoint error to the millimeter scale on the mmBody benchmark.
[HC-18] When Attention Guardrails Become Barriers to Learning: Towards the Tipping Point
链接: https://arxiv.org/abs/2609.39023
作者: Meenakshi V.(1),Pavani Ayinampudi(2),Aditya B. M. V.(2),Jinal Gupta(2),Prakash Hegade(2),Rohit Sharma(1),Sakshi Sharma(1),S. R. S. Iyengar(1) ((1) Indian Institute of Technology Ropar, Rupnagar, Punjab, India, (2) a href=“http://ANNAM.AI” rel=“external noopener nofollow” class="link-external link-http"this http URL/a, Indian Institute of Technology Ropar, Rupnagar, Punjab, India)
类目: Human-Computer Interaction (cs.HC); Computers and Society (cs.CY)
备注: 10 pages, 2 figures, 4 tables
Abstract:Online learning offers flexibility but lacks the structure of a classroom, where a teacher’s presence guides attention. The platform we study restores that structure by monitoring the learner through the webcam during ordinary coursework, interrupting or restarting a video when the learner appears distracted. What such monitoring does to a learner across a whole course, rather than in a single examination, is largely unexamined. We report a convergent mixed-methods study of one monitored course pipeline in a summer internship program. We read two free-text surveys alongside three platform channels. The dropout exit survey yielded 15 analyzable responses; the persisting-learner reflection survey, 36. The channels are camera-verification telemetry (14,529 flags from 448 students), an in-video emotion widget (615 submissions from 273 students), and a mandatory end-of-course survey (up to 634 respondents per item). Neither survey named monitoring, so every mention analyzed here was raised by the respondent. Focus-monitoring was raised by 18 of the 51 free-text respondents: 7 of 15 dropouts and 11 of 36 persisting learners. Among the dropouts who raised it, focus-monitoring was the stated primary cause of departure in 4 of 7 cases. None of the 18 questioned being observed in principle. What learners contest is the misreading of ordinary actions, drinking water or moving the head, and the severity of what follows a flag: a video already watched returns to the start of its segment. These findings identify two targets for redesign: the severity of the response to a flag, and the environment check, which can flag a learner before any content has been seen. We argue that the proportionality of that response marks the point at which an attention guardrail becomes a barrier to learning, the tipping point this study approaches.
[HC-19] Insights on Student Learning from Live Classroom Polls: More Than Right or Wrong
链接: https://arxiv.org/abs/2609.38969
作者: Rohit Sharma,Pavani Ayinampudi,Aditya B.M.V.,Jinal Gupta,Prakash Hegade,Sakshi Sharma,Meenakshi V,SRS Iyengar
类目: Human-Computer Interaction (cs.HC); Computers and Society (cs.CY)
备注: 8 pages, 6 figures, Submitted to ICTIEE 2027 and currently under peer review
Abstract:A live classroom poll is usually read through its answer key, as a count of participation and correctness. Yet the same record holds more before any key is fixed, namely who answers early and how the class divides across the options. Whether these key-free signals are stable, and what they reveal, has not been examined at the scale of a real course. We study them in a live deployment of 340,668 submissions across 603 polls in 47 sessions from roughly 2,400 learners, reading each poll for a learner’s response order and for how much the class divides. A single further step brings in the author-designated key, on items with a settled answer, to interpret the divided ones. A learner’s response order is a stable individual signature. It reproduces at a corrected split-half reliability of 0.92, holds at 0.87 within single sessions, and is unrelated to whether the learner is correct. Read without a key, about one poll in five does not reach a clear majority, and division marks the questions that split the class. On the settled-answer items the majority is itself wrong on 49 of 358. The key-free measure flags most of these, though division alone does not separate a collectively wrong answer from a merely hard one. The two readings are largely independent, meeting at one point where early responders anticipate the eventual majority only on questions the class has nearly settled. Together they map the learners and questions of a session without a key.
[HC-20] What Students Actually Ask: Demand Structure and Automation Potential in a Hybrid Support System
链接: https://arxiv.org/abs/2609.38967
作者: Jinal Gupta,Pavani Ayinampudi,Aditya B.M.V.,Prakash Hegade,Rohit Sharma,Sakshi Sharma,Meenakshi V,S.R.S. Iyengar
类目: Human-Computer Interaction (cs.HC)
备注: 13 pages, 2 figures, Poster accepted in T4E
Abstract:Large online programmes receive heavy volumes of queries during onboarding, at a scale that grows faster than the number of staff available to answer them. This paper reports a study of an AI integrated query resolution platform that spreads incoming queries across four routes: an AI based assistant, answers from fellow participants, a curated corpus of frequently asked questions, and escalation to administrators. Over nine weeks, from 2 May to 6 July 2026, the system handled 4,093 queries raised by 1,434 participants. Nearly every query reached a recorded resolution, and one query in five closed within an hour. Reuse of 114 corpus entries absorbed 21.3% of the volume, participants resolved a further 12.2% on their own, and 132 participants answered questions for one another at a median of 9 to 15 minutes, showing that peer answering, where it occurred, was fast and broadly shared across the cohort. Classifying the query text shows that demand was narrow rather than varied. A single process step, the submission of a certificate and the offer letter that follows it, accounts for 56.8% of corpus mediated resolutions, and at least 20.4% of queries concern the progress of a pending submission rather than a request for information, a class the assistant served only 1.3% of the time, since a stored answer cannot report an individual’s current status. Only 8.0% of the queries handled by a person duplicated content already in the corpus, which indicates that the knowledge base was already well used. Most direct administrative closures occur in synchronous bulk events, a pattern that shapes how the records should be read.
[HC-21] An Audit of Measurement Quality and Answer Bias in a Large Classroom-Poll Corpus
链接: https://arxiv.org/abs/2609.38959
作者: Rohit Sharma,Pavani Ayinampudi,Aditya B.M.V.,Jinal Gupta,Prakash Hegade,Sakshi Sharma,Meenakshi V,SRS Iyengar
类目: Human-Computer Interaction (cs.HC); Computers and Society (cs.CY)
备注: 14 pages, 6 figures, 2 tables
Abstract:Real-time classroom polls are widely used and increasingly generated with automated assistance, yet the questions themselves are rarely evaluated as measurements. We audit a large corpus of authentic classroom polls, 604 items across 47 sessions answered 340,668 times by 2,807 learners, as a measurement instrument. For the 539 items whose correct answer could be established and verified from the lecture transcript, we place every item and every student on a common scale using item response theory and analyse the answer structure of the True/False items. Two findings emerge. First, the polls form a coherent but easy scale of moderate precision (marginal reliability about 0.60), on which roughly a quarter of items barely separate stronger from weaker students. Second, students show a robust tendency to answer True, present at the individual level (77% of students lean True), which meets a milder tendency for items to be keyed False; as a result answer direction predicts difficulty, False-keyed items being about thirteen points harder, and the effect survives controls for item content and for selective answering. Both findings rest on signals a polling system already records, so the same checks can be run as items are generated, before they reach students.
[HC-22] Measuring Student Self-Assessment against Viva-Demonstrated Mastery in a Large First-Year Programming Course
链接: https://arxiv.org/abs/2609.38951
作者: Sakshi Sharma,Pavani Ayinampudi,Aditya B.M.V.,Jinal Gupta,Prakash Hegade,Rohit Sharma,Meenakshi V,S.R.S. Iyengar
类目: Human-Computer Interaction (cs.HC)
备注: 14 pages, 3 figures, 2 tables, Submitted in ICTIEE and under Review
Abstract:Mastery-based education increasingly places the reporting of learning progress in students’ hands, who record task completion on learning dashboards. The usefulness of such self-reports depends on how closely reported mastery corresponds to demonstrated competence. Most evidence on student self-assessment compares an overall self-rating with an overall examination score and therefore provides limited evidence about which tasks or which students account for the mismatch. This study examines first-year students’ self-assessment against viva-demonstrated mastery at the level of individual tasks across a ladder of sixty programming tasks. The study draws on a large first-year programming course taught in 2023, involving 203 students and 12 examiners, in which every reported task was verified through an oral viva. Because a task entered the Viva only after it was reported, the design is one-sided and captures over-estimation but not under-estimation. Of 11,093 reported tasks, 10,885 (98.1%) were demonstrated, indicating a high degree of correspondence between self-report and demonstrated mastery. The remaining 208 overestimations were not evenly distributed. A small number of students accounted for most of the errors, and they occurred mainly on difficult tasks near the end of the task ladder rather than on higher-point tasks. This task-level analysis shows that high overall self-assessment accuracy can coexist with specific areas where reported and demonstrated mastery diverge. It also provides a practical basis for directing additional verification and formative feedback toward students and tasks where such divergence is more likely.
[HC-23] EFormer: Temporally Aligned Local Correction for Continuous sEMG-Based Hand Pose Tracking
链接: https://arxiv.org/abs/2609.38932
作者: JiaCheng Ge,SiYu Zhang
类目: Machine Learning (cs.LG); Human-Computer Interaction (cs.HC)
备注:
Abstract:Surface electromyography (sEMG) provides a wearable, camera-free signal for continuous hand-motion inference. Mapping muscle activity to joint kinematics remains challenging because the recorded waveforms are indirect measurements, their relationship with motion changes over time, and individual anatomy and sensor placement alter the signal distribution. This paper presents EFormer, a residual feature-correction network built on a frozen tracking backbone. EFormer combines a high-rate event branch, temporally aligned local cross-attention, two causal rotary position embedding (RoPE) temporal layers, and a bounded, dynamically gated residual. EFormer receives 16-channel sEMG sampled at 2 kHz and fuses a 64-channel tracking representation at 25 Hz with a 128-channel event representation at 200 Hz. Cross-attention uses a nominal delay of 100 ms, a 300 ms history parameter, and a 50 ms tolerance; its causal mask restricts each query to events occurring 50-400 ms earlier. The correction scale is 0.15. The evaluated continuation-training configuration contains 585,376 trainable parameters and 5,974,508 frozen parameters. On the test set, EFormer achieves an MAE of 0.1546634 rad, an RMSE of 0.24063 rad, and an R^2 of 0.74801, compared with 0.1745326 rad, 0.2715448 rad, and 0.6791103 for the official tracking baseline. EFormer reduces MAE by 11.38% relative to the baseline. The results show that temporally aligned event-feature correction can reduce continuous hand-pose tracking error. Subjects: Machine Learning (cs.LG); Human-Computer Interaction (cs.HC) Cite as: arXiv:2609.38932 [cs.LG] (or arXiv:2609.38932v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.38932 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[HC-24] Characterizing Questioning Patterns and Student Engagement Through Contextual Analysis of Real-Time Classroom Interactions
链接: https://arxiv.org/abs/2609.38907
作者: Rohit Sharma,Pavani Ayinampudi,Aditya B.M.V.,Jinal Gupta,Prakash Hegade,Sakshi Sharma,Meenakshi V,SRS Iyengar
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: 16 pages, 5 figures, One version is accepted at T4E 2026
Abstract:Real-time classroom polling is now routine, yet the data it produces is usually read narrowly, as a correctness score or a headcount. Such readings say little about what a poll is doing within a lecture or how it shapes engagement. This is particularly relevant for short-response formats such as True/False, where the same question format can be used to test recall, check comprehension, or direct students’ attention to a deliberately misleading statement. This study asks whether a poll’s answer and instructional function can be determined by reading it against its lecture transcript, what cognitive levels of Bloom’s taxonomy and instructional-function clusters the corpus contains, and how student engagement relates to answering correctly. We analyse a naturalistic corpus of 47 live sessions over 39 days, comprising 604 poll questions and 340,668 responses from 2,807 learners, most items True/False, read against time-aligned lecture transcripts and attendance. Reading each poll in context proves essential: the answer to 89% of polls is locatable in the lecture, and a recurring attention-checking device is visible only through context. Questioning is overwhelmingly lower-order and falls into seven instructional functions, and a poll’s response follows its function rather than its wording. Engagement is broad but concentrated, and the class majority answers correctly 88.5% of the time, though a small set of high-consensus yet incorrect answers cannot be detected by agreement alone. An independent survey of 579 students agrees on what the polls are and on their participation, but reveals a gap between perception and reality: students cannot judge their own correctness, and the polls they find hardest are not those they answer worst.
[HC-25] Whose Voice Survives the Summary? A Voice-Retention Audit of LLM Employee Listening
链接: https://arxiv.org/abs/2609.38818
作者: Thilo Tamme,Anton Hantel,Bijan Khosrawi-Rad
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注: 10 pages, 3 figures, 3 tables. Accepted at the 60th Hawaii International Conference on System Sciences (HICSS 2027)
Abstract:Organizations increasingly route employee feedback to leaders through large language model (LLM) summaries, an unaudited layer that silences already-spoken voice. We introduce a Voice Retention / Representation Ratio metric for representational bias in summarization and apply it to a bilingual (English/German) corpus of 2,586 free-text responses from a global professional service company. First, employees supply criticism more reliably than praise (withholding praise is 82 times more common). Second, across 45 leader-summaries the pipeline filters by popularity, not sentiment: criticism survives, yet a concern voiced once is dropped 86% of the time, with short and German-only content lost on the same axis (theme retention 0.14 vs 0.74; German directional). Controlling for frequency, sentiment has no independent effect; the harm is prevalence-driven, which sentiment-only audits miss. A targeted prompt recovers only named themes. We contribute the metric, field evidence, and a disaggregated voice-retention card.
[HC-26] Positive Ratings Hidden Concerns: Employee Voice Disclosure in AI-Mediated Organizational Listening
链接: https://arxiv.org/abs/2609.38788
作者: Thilo Tamme(1),Michael Saatkamp(1),Alma Bonte(1),Daniel Weiss(2),Anton Hantel(3),Andrej Levin(1) ((1) Technical University of Munich, (2) LMU Munich, (3) Massachusetts Institute of Technology)
类目: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注: 10 pages, 2 figures, 3 tables. Accepted at the 60th Hawaii International Conference on System Sciences (HICSS-60), 2027
Abstract:Organizations started listening to employees through conversational AI agents alongside structured surveys. Little is known about what these channels change in what employees say when disclosure carries hierarchical risk. We report a field study inside a global management consulting firm whose process pairs a pre-survey with an adaptive AI voice interview on the same themes within one session. Across 44 first-session interviews (132 matched theme observations), 20-41% of sessions showed a favorable rating co-occurring with a substantive concern voiced later, depending on the favorability threshold. The Gioia analysis drew on 158 protective quotes from 65 eligible sessions. Disclosure rarely arrived unguarded: employees softened concerns, deflected accountability, and bounded how far they went, and this protective work tracked the perceived legitimacy of the listening structure. We develop a grounded model of bounded disclosure and derive four propositions for voice, channel and listening research. Silence, we argue, can persist inside expression.
[HC-27] Persona and Persuasive Framing in AI Voice Agents : A 2times2 Field Experiment with Children
链接: https://arxiv.org/abs/2609.38782
作者: Thilo Tamme(1),David Steck(1),Anton Hantel(2) ((1) Technical University of Munich, (2) Massachusetts Institute of Technology)
类目: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注: 8 pages, 4 figures, 3 tables. Accepted at the 60th Hawaii International Conference on System Sciences (HICSS-60), 2027
Abstract:Conversational agents increasingly interact with children, yet evidence on how their design shapes children’s susceptibility to persuasion comes almost entirely from the lab. We report a 2\times2 randomized field experiment embedded in a public German Santa Claus telephone hotline. Children’s calls were randomly routed to one of four LLM voice agents varying persona (Santa, high authority, vs. Helper, low authority) and framing (persuasive nudges toward prosocial wishes vs. neutral). Of 1,072 logged calls, 89 conversations (median age 6) met inclusion criteria. Persuasive framing raised the probability of a prosocial wish from 11.6% to 45.7%, robust to controls. Persona authority showed a near-zero effect: Santa did not outperform the Helper. Persona instead shaped engagement; children hung up on the Helper far more often within the first minute (65% vs. 39%). Where context already lends an agent legitimacy, how it speaks shapes children’s compliance more than who it claims to be.
[HC-28] Where the Evidence Lives: Auditing AI Companions Self-Descriptions
链接: https://arxiv.org/abs/2609.38753
作者: Seiya Ikeda,Shin-nosuke Ishikawa
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注: 17 pages, 8 figures, 9 tables. Ancillary files contain the LLM judge prompt (Japanese original and English translation) and the probe statements
Abstract:Companion agents describe themselves: they remember, they understand their users, the relationship has changed them. We argue that such accounts, and the experience ratings that seem to confirm them, are checkable by users only where the evidence is theirs: in the agent’s behavior, or in themselves. Where the evidence lives in the machinery, fluent self-description and moderately positive ratings do not establish that the mechanisms behind them ran. We demonstrate an audit procedure that sets an agent’s self-description against its users’ judgements and its implementation records, reporting each claim as supported, contradicted, or unresolved, and apply it to Lita, a proactive companion we built and deployed for a month with nine colleagues. Participants endorsed stylistic claims, withheld endorsement from relational ones, and rated memory at or above midpoint, while two of three memory layers had never executed their accumulation step. Memory-bearing agents should report what their self-descriptions cannot establish.
[HC-29] AIfred: Augmented Learning through Functional Robotic Embodiment at the Desk ICRA2027
链接: https://arxiv.org/abs/2609.38737
作者: Gregorio Orlando,Milan Groshev,Eduardo Castelló Ferrer
类目: Robotics (cs.RO); Human-Computer Interaction (cs.HC)
备注: 7 pages, 9 figures. Submitted to ICRA 2027. This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible
Abstract:Desk-based learning and creative activities benefit from handwritten engagement. However, current generative AI tools deliver guidance through a separate screen, creating a gap between where users think and where assistance appears. To address this, in this work we design AIfred, a desk-based robotic arm with a projector mounted at the end-effector that places AI-generated guidance alongside handwritten work. AIfred combines workspace perception, context-aware content generation, and robot-mediated projection to support math assignments, image generation, and drawing tasks. In a user study (n = 36), we compared AIfred against ChatGPT (GPT-5.6 Luna) running on a laptop. Both tools performed comparably while assistance was available during the math assignment (6.7 vs. 7.3/10, p = .41), but AIfred improved short-term learning transfer by 60% once assistance was withdrawn (7.0 vs. 4.4/10, p = .003). In addition, independent art and design professors ranked drawings produced with AIfred better in 33 of 36 cases. Our findings indicate that spatially co-located AI assistance benefits tasks whose guidance shares a spatial frame with the work.
[HC-30] Afterglow: A Place-Based Memorial Ecology for AI-Mediated Pet Bereavement
链接: https://arxiv.org/abs/2609.38729
作者: Hanjing Shi,Dominic DiFranzo
类目: Human-Computer Interaction (cs.HC)
备注: 32 pages, 4 figures; author preprint
Abstract:Pet bereavement often receives little social recognition. Generative AI can give a continuing bond a responsive voice, but a comforting reply may also claim authority to forgive or request attention. We investigate how a memorial can support connection without turning remembrance into obligation. Through Research through Design, we developed Afterglow, a mobile world connecting private remembrance, human witnessing, and symbolic Pet messages. A formative survey (N=57) informed the initial design, followed by six online roundtable walkthroughs with 20 unique participants across two prototype iterations. Our interpretive analysis develops tensions between returnable connection and emotional obligation, recognizable likeness and ontological clarity, and protective intervention and surveillant authority. We contribute Legible Restraint, a cross-layer requirement that limits on relational authority survive changes in speaker, generation context, trigger logic, data use, and participation. Its temporal consequence, Designing for Goodbye, keeps remembrance available without making continued use a condition of care.
[HC-31] EPIC: Epipolar-Consistent 360° Immersive Stereo Video Generation
链接: https://arxiv.org/abs/2609.38689
作者: Debabrata Mandal,Dongdong Fu,Jonathon Miller,William Villareal,Xi Peng,Praneeth Chakravarthula
类目: Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
备注:
Abstract:Immersive displays can enable rich and diverse virtual experiences. Manually authoring every possible experience to realize this potential, however, is prohibitively expensive, difficult to scale, and impractical. Generative AI models could remove this bottleneck, but today’s models are built for conventional displays and cannot generate the high-resolution, stereoscopic 360^\circ content required for immersive viewing. Further, temporal and stereo inconsistencies that may be tolerable on conventional displays can become highly disruptive when viewed through an immersive headset. Here, we address this gap with a zero-shot generative pipeline that extends existing video diffusion models into 4K stereoscopic 360^\circ videos. Inspired from binocular vision and depth perception, we develop an epipolar-aware 360^\circ image matching metric that captures the temporal and stereo geometric inconsistencies across views. We then use this metric as a preference signal for direct preference optimization with limited training data. Our work enables 360^\circ stereo video generation and provides a scalable path for bringing generative content to immersive displays, allowing diverse mixed reality experiences on demand. Subjects: Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC) Cite as: arXiv:2609.38689 [cs.CV] (or arXiv:2609.38689v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.38689 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[HC-32] C-LISTEN: Cognitive Load Impacts of Sensory-Triggered Environmental Navigation in Virtual Reality
链接: https://arxiv.org/abs/2609.38589
作者: Md. Monowar Hossain,M. Rasel Mahmud
类目: Human-Computer Interaction (cs.HC)
备注: 6 pages. Published in AIVR 2026, The Third International Conference on Artificial Intelligence and Immersive Virtual Reality, Lisbon, Portugal, April 19-23, 2026
Abstract:Nowadays, Virtual Reality (VR) systems are more helpful for making decisions, training, and rehabilitation. Multimodal settings in these systems impose a significant balance and cognitive difficulties. Additionally, decreased performance, higher cognitive load, user overwhelm, and limited accessibility of VR technology can be caused by excessive auditory, visual, and sensory stimulation. In our pilot study, we examine the strategic design of directional auditory cues to improve cognitive load and enhance user experience in virtual environments. This research also enhances fundamental knowledge that auditory feedback reduces cognitive load in VR, with statistical analysis confirming this improvement (p = .0011). This study investigates how postural balance and pupil diameter correlate with cognitive load while the users navigate. Moreover, this study provides realistic overview design guidelines for accessibility, which will make VR experiences safer and more useful for everyone, including people with cognitive impairments.
[HC-33] owards Model as a Library: Offline Community-Sourced AI for Low-Resource African Languages NEURIPS2026
链接: https://arxiv.org/abs/2609.38574
作者: Fendji K. E. Jean Louis
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注: 5 pages, GlobalSouthAI @ NeurIPS 2026
Abstract:Large language models are frequently proposed as a route to AI-powered services for African communities, but they are least reliable exactly where the need is greatest: all African languages remain low-resource by any standard measure, and models trained on scraped, standardised text systematically misrepresent the dialectal and regional variation of how people actually speak. We introduce \textbfModel as a Library (MaaL), a software architecture that packages small, community-enrolled speech models as versioned on-device dependencies, enabling offline structured data collection that cannot generatively hallucinate, for populations that current language models serve worst. Rather than relying on web-scraped corpora, MaaL’s vocabulary is enrolled directly from a small number of example recordings by the speakers themselves, at the point of deployment. We describe the architecture and its central mechanism - keyword spotting that turns a closed-vocabulary text form into a voice form, filled and submitted entirely on-device - and propose transpiling the closed-vocabulary elements already present in widely-deployed digital form tools into MaaL schemas, a low-friction path to voice-first, offline data collection for the low-literacy populations these tools already reach. This is a position and system-design paper: we describe the concept, the mechanism, and an analytical feasibility case, and identify what a working implementation still requires.
[HC-34] Navigating the Changing Landscape of Online Knowledge Consumption and Production in the Age of Generative AI: Evidence from Stack Overflow
链接: https://arxiv.org/abs/2609.38563
作者: Ji Eun Kim,Léa Vitale,Libby Hemphill,Yulin Yu
类目: Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注:
Abstract:Online knowledge communities rely on a division of epistemic labor between users who seek information and those who produce it. Generative AI may blur these roles, but how it reallocates knowledge-seeking and knowledge-producing activities and reshapes the nature and returns of participation remains unclear. We examine changes in question-asking and answering among Stack Overflow user groups, focusing on consumers and producers, following the release of ChatGPT. Consumers shifted from asking questions to producing answers, but their answers were less likely to be accepted relative to the pre-ChatGPT period, suggesting that increased production did not yield equal standing in the community. By contrast, producers did not change the number of questions they asked, but increasingly asked about novel and emerging topics. Although they produced fewer answers, their answers received greater recognition. These findings show that access to generative AI does not necessarily translate into equal opportunities for successful participation.
[HC-35] Fairness Theatre: Evaluating Post-Hoc Fairness Interventions in Vendor-Controlled Early Warning Systems
链接: https://arxiv.org/abs/2609.38552
作者: Kelly McConvey,Angelina Zhai,Rebecca Li,Shion Guha
类目: Human-Computer Interaction (cs.HC); Computers and Society (cs.CY); Machine Learning (cs.LG)
备注:
Abstract:Public institutions increasingly procure AI systems whose design they cannot inspect or change. In higher education, proprietary Early Warning Systems (EWS) leave colleges with few options beyond adjusting model outputs to address inequity. This raises the question of how fairness work is coordinated among vendors, institutions, advisors, and students with unequal power to change these systems? Using student records from a public college in Ontario, Canada, we evaluate six post-hoc fairness interventions on a research EWS under simulated procurement constraints. We compare fairness, accuracy, and demographic disparities, introducing error-type profiling to trace how interventions redistribute false positives and false negatives. Interventions redistributed disparities without consistently reducing them. Two implementations favored already-advantaged groups because they used group size to define disadvantage; small, marginalized groups remained poorly served. These findings show how procurement constraints and implementation choices shape the possibilities for fairness work. We call the resulting condition fairness theatre; dashboard metrics converge while groups’ error burdens persist or worsen.
[HC-36] Beyond Monoscopic Viewing: A Study on 3D Gaussian Splatting Quality in VR
链接: https://arxiv.org/abs/2609.38525
作者: Shreyas Shivakumara,Gabriel Eilertsen,Karljohan Lundin Palmerius
类目: Human-Computer Interaction (cs.HC)
备注: 12 pages
Abstract:Stereoscopy is fundamental to virtual reality (VR), providing depth perception through binocular viewing. Recent advances in 3D Gaussian Splatting (3DGS) enable high-quality novel view synthesis, making it well suited to immersive VR. We render 3DGS reconstructions stereoscopically and evaluate them in a head-mounted display, replicating how they would actually be viewed in VR. Real-world capture provides only a limited number of views, and under this constraint 3DGS reconstruction often produces localized floaters and misplaced structures. Standard metrics miss these localized artifacts, which become salient under stereoscopic viewing, where geometry is placed at the wrong depth. We investigate whether standard image-quality evaluation reflects the perceptual quality of 3DGS reconstructions under reduced capture. We compare SfM-only baseline with a union initialization that combines SfM with a dense VGGT network. All other training components are held fixed, isolating the effect of initialization coverage. We conduct a user study comparing preferences under monoscopic and stereoscopic HMD viewing, and test whether image-quality metrics predict the observed preferences. Monoscopically, preference for the more consistent reconstruction is weak, reaching 58.4% overall. Stereoscopically, the same preference rises to 78.2% and is consistent across all participants, while image-quality metrics (PSNR, SSIM and LPIPS) and stereo-aware metrics (iSQoe and StereoQA) show only modest differences and fail to penalize them. Our results indicate that monoscopic evaluation and standard image-quality metrics substantially underestimate perceptual artifacts observed in 3DGS reconstructions for VR, making stereoscopic assessment essential for 3DGS quality evaluation in VR.
[HC-37] Game-Based versus Training-Based Virtual Reality for Cognitive Rehabilitation in Mild Cognitive Impairment and Dementia: A Systematization of Design Paradigms Outcomes and Evaluation Rigor
链接: https://arxiv.org/abs/2609.38514
作者: Anowarul Faruk Shishir,M. Rasel Mahmud
类目: Human-Computer Interaction (cs.HC)
备注: Accepted at the 32nd ACM Symposium on Virtual Reality Software and Technology (VRST 2026)
Abstract:Therapeutic exercises can help slow cognitive decline in dementia, but conventional programs are often tedious and difficult to sustain. Virtual reality (VR) offers an alternative through game-based approaches designed to improve engagement and training-based approaches focused on structured functional recovery. Prior reviews report inconsistent findings about these paradigms, while it remains unclear whether the labels represent systems that are technologically distinct. Following PRISMA 2020 guidelines, we systematize 34 studies published between 2015 and 2026, including 16 controlled efficacy trials and 18 design and feasibility studies involving mild cognitive impairment, dementia and Alzheimer’s disease. We develop a taxonomy covering design paradigm, immersion, feedback modality, difficulty adaptation and hardware. Our analysis shows substantial overlap between game-based and training-based systems, suggesting that these labels do not clearly distinguish their technological characteristics. We also introduce VR-RES, a seven-dimensional framework for evaluating study quality. Its application highlights recurring limitations in the evidence base, particularly limited follow-up, poor reproducibility and insufficient analysis of demographic factors such as age, sex and education. Based on these findings, we provide recommendations for future VR cognitive rehabilitation research, emphasizing specific design features, rigorous evaluation and improved reporting practices.
[HC-38] Anthropomorphism in the age of Large Language Models : An overview of potential risks and mitigations
链接: https://arxiv.org/abs/2609.38486
作者: Ismael T. Freire,Marceau Nahon,Maud van Lier,Katie Evans,Hélie Bazin,Michele Farisco,Kathinka Evers,Raja Chatila,Mehdi Khamassi
类目: Computers and Society (cs.CY); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注: 35 pages, 1 box, 1 figure
Abstract:Large Language Models (LLMs) and more broadly Artificial Intelligence (AI) systems are often described and understood in human-like terms, a phenomenon known as \emphanthropomorphism. This paper provides a synthesis of recent literature on anthropomorphism in AI, covering theoretical frameworks, the role of language in framing AI as human-like, the various risks of anthropomorphizing machines, and strategies to mitigate these issues. After examining why we tend to anthropomorphize AI systems and whether we are right to do so, we highlight the impact of linguistic framing on anthropomorphism. Then, we introduce a conceptual taxonomy of risks associated with AI anthropomorphism. This taxonomy groups twenty-one concerns within five analytical categories: epistemic, affective, human agency, normative, and societal and institutional risks. Finally, we relate these concerns to proposed interventions in design, communication, education, and governance. We argue that a better understanding of AI systems requires concepts and theories grounded in their organization and demonstrated capacities. The linguistic shaping of anthropomorphic perceptions should form part of this scientific effort, since our descriptions influence both how these systems are understood and the roles we allow them to occupy in society.
[HC-39] Embodiment-aware control by inference over the operator: a simulation study
链接: https://arxiv.org/abs/2609.38437
作者: Sara Falcone
类目: Robotics (cs.RO); Human-Computer Interaction (cs.HC); Neurons and Cognition (q-bio.NC)
备注: 13 pages, 4 figures, 3 tables. Code and data: this https URL (doi: https://doi.org/10.5281/zenodo.23044669 )
Abstract:Teleoperation systems are tuned for channel fidelity, while whether the operator experiences the device as part of the body, the Sense of Embodiment (SoE), is measured only afterwards, by questionnaire. Predictive-processing accounts suggest controlling devices to reduce the mismatch between the operator’s predictions and the returned feedback, but those predictions are unobservable, and an objective that only penalizes mismatch is minimized by removing feedback. We formulate an embodiment-aware controller, the Universal Embodiment Engine (UEE), that infers the operator’s embodiment and visuo-proprioceptive cue weighting from implicit gaze and pupil signals and task outcome, and chooses bounded device settings under explicit preferences, cast as a discrete Active Inference agent. In simulations with 300 heterogeneous synthetic operators, the UEE found the suitable setting within half a minute for most operators, before identifying their exact type, and came close to an oracle in the second half of the session (embodiment 1.68 against 0.73 for the best fixed setting, on a 0-2 scale). Adapting without reading the operator did no better than fixed control, and model-free bandits did worse, whereas an expected-utility controller with the same inference did exactly as well: the benefit comes from Bayesian inference over the operator with explicit preferences, not from the information-seeking term of Active Inference. A naive prediction-error minimizer withheld feedback, as its objective implies, and lost task success (0.74 vs 0.90). The benefit shrank but persisted for operators outside the controller’s model family, grew with the variety of the population, and vanished when the controller trusted an uninformative signal or when cue weighting changed mid-session without being modeled. These failures show what studies with people must establish first: calibrated signals and a model of change.
[HC-40] GestAdapt: Workspace-Conditioned Co-Speech Gesture Generation for Humanoid Robots
链接: https://arxiv.org/abs/2609.38400
作者: Bosong Ding,Xianglin Zhang,Miao Xin,Murat Kirtay,Giacomo Spigler
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)
备注:
Abstract:Co-speech gestures for robots must adapt not only to speech and embodiment, but also to the workspace available for performing the motion. Since the same speech can be accompanied by different gestures, a robot can respond to workspace constraints, e.g., gestures for speech next to a wall. In these scenarios, the robot should gesture in a suitable motion rather than simply correcting an unconstrained one. To achieve this goal, we present GestAdapt, a workspace-conditioned framework that conditions co-speech gesture generation on a prescribed wrist workspace. The GestAdapt framework learns from six complementary co-speech corpora through a shared motion representation and supports retargeting to different robot embodiments. Quantitative evaluation shows that generated motions remain close to the real-motion distribution while respecting the workspace. In a user study, gestures generated under modified workspace constraints receive a mean quality score of 3.24/5, above our no-workspace variant (2.43/5) and below the reference motions (3.68/5). In a real robot evaluation, all compared motions are retargeted to the Reachy2 humanoid robot under identical workspace constraints. Motions generated with our framework rank first in 69.7% of comparisons, higher than our no-workspace variant baseline and retargeted ground-truth motions constrained afterward. Overall, the results support adapting gestures to the available workspace during generation, rather than modifying unconstrained trajectories afterward to satisfy workspace constraints, potentially compromising gesture naturalness.
[HC-41] Beyond the Headset: A Systematization of Knowledge on Extended Reality Privacy and Security in Healthcare
链接: https://arxiv.org/abs/2609.38281
作者: Nafisa Anjum,M. Rasel Mahmud
类目: Cryptography and Security (cs.CR); Human-Computer Interaction (cs.HC)
备注: Published in 31st ACM Symposium on Virtual Reality Software and Technology
Abstract:Extended reality (XR) systems are increasingly used in healthcare applications ranging from surgical planning to remote rehabilitation and mental health support. However, the rich streams of sensor, biometric, behavioral, and environmental data that enable these applications also introduce substantial privacy and security risks. Adversaries may exploit insecure communication, sensor side channels, application-layer vulnerabilities, or data-processing pipelines to infer sensitive information or disrupt clinical workflows. Despite growing interest in XR security and privacy, the healthcare-specific literature remains fragmented. In this Systematization of Knowledge (SoK), we review 65 peer-reviewed studies published between 2017 and 2024 across XR, security, privacy, and healthcare venues. We develop a unified threat taxonomy spanning device, user, network, and cloud layers and introduce XR-PRISM, a quantitative Privacy and Risk Impact Scoring Metric for systematically characterizing security and privacy risks. Our analysis identifies several gaps in the literature: more than 70% of proposed countermeasures lack standardized risk evaluation, fewer than 15% of studied attacks require high attack prerequisites, and reproducibility is limited by the scarcity of publicly released artifacts and datasets. Based on these findings, we outline a research roadmap emphasizing shared benchmark datasets, stronger artifact-release practices, improved cloud-layer protections, and more comprehensive detection, mitigation, and recovery mechanisms. This SoK provides a structured and data-driven foundation for understanding existing risks and guiding the development of more secure, privacy-preserving, and usable XR healthcare systems.
[HC-42] How People Use ChatGPT : Conversation-Level Evidence from India Nigeria Brazil and Pakistan
链接: https://arxiv.org/abs/2609.38279
作者: Shreyasi Roy Chowdhury,Kiran Garimella
类目: Computers and Society (cs.CY); Human-Computer Interaction (cs.HC); Social and Information Networks (cs.SI)
备注: 29 pages, 22 figures, 9 tables
Abstract:Public understanding of how people use LLM-based conversational AI assistants comes primarily from aggregate platform reports by OpenAI and Anthropic, which apply fixed taxonomies and inferred demographics to hundreds of millions of users and release only summary statistics that outside researchers cannot re-analyze. We provide a complementary, conversation-level view: complete ChatGPT exports comprising 202,590 conversations from 1,252 users across India, Nigeria, Brazil, and Pakistan, paired with self-reported age and gender and spanning December 2022 to February 2026. To our knowledge this is the first conversation-level, demographically grounded comparison of ChatGPT use across multiple non-Western countries. We ask what these users use ChatGPT for (purpose), what they talk about (topics), and how they engage with it (mode of interaction), using the platform’s own classifiers, unsupervised topic discovery, and a thematic analysis of expressive conversations. Personal use accounts for 55-64% of conversations in every country and coursework is about as common as work, so workplace productivity describes a minority of use. Unsupervised topic discovery surfaces country-specific uses that the OpenAI taxonomy folds into generic categories: health and wellness in India and Brazil, Urdu-English translation in Pakistan, current affairs in Nigeria, religious questions in Nigeria and Pakistan, and self-reflection in Brazil. Over three years, the share of conversations that seek information declined only modestly and the share that delegate a task did not grow, while conversations in which users express themselves rose from a few percent to roughly a fifth or more in every country. The same product is thus attached to different local needs in each country, and understanding what adoption means requires conversation-level, country-sensitive measurement alongside global aggregates.
[HC-43] AI-Powered Symptom Assessment and User Experience: A Case Study of Simtomi and Simtomi-Care
链接: https://arxiv.org/abs/2609.38187
作者: Jinha Lee(1),Chan Hyung Lee(2),Hyunsung Lee(2),Seunghwan Kim(3),Ban Hyung Lee(2),Minjun Shin(4),Hojin Shin(5),Jungdo Park(2) ((1) Indiana Wesleyan University, Marion, USA, (2) Research Institute of Mediark, Seoul, Republic of Korea, (3) College of Medicine, Ewha Womans University, Seoul, Republic of Korea, (4) Department of Biology, Indiana University, Bloomington, USA, (5) Hamilton Southeastern High School, Fishers, USA)
类目: Human-Computer Interaction (cs.HC)
备注: 6 pages, 5 figures; published in the 2026 IEEE Conference on Artificial Intelligence (CAI)
Abstract:Digital symptom checkers are widely used for quick guidance on health concerns, yet many systems still face challenges in collecting accurate information, supporting communication, or integrating with clinical workflows. To explore how these tools function in real use, we examine the case of the Simtomi system, which pairs a multilingual symptom assessment application with a provider-facing platform. Empirical studies were conducted in two countries. In South Korea, based on participants’ firsthand experience, we found that the system improved how patients communicated their symptoms and helped clinicians review cases more efficiently through structured summaries aligned with diagnostic reasoning. In the United States, responses from prospective users and healthcare professionals highlighted the value of multilingual support, structured questioning, and the system’s potential to assist clinical coordination. These findings offer a grounded account of how AI-based symptom assessment tools can operate across different healthcare contexts and provide broader insight into usability, trust, and usefulness in digital health.
[HC-44] Why and How People Check Generative AI Output for Mistakes
链接: https://arxiv.org/abs/2609.38186
作者: Patrick Gage Kelley,Derrick Feldmann,Reena Jana,Colleen Thompson-Kuhn,Allison Woodruff
类目: Human-Computer Interaction (cs.HC)
备注: 12 pages, 7 figures, 2 tables, ancillary appendix file
Abstract:Generative AI output can contain errors, such as hallucinations, non-responsive results, or otherwise inaccurate or potentially harmful content. To explore the public’s emerging understanding, attitudes, and behavior regarding such mistakes, we ran an online survey in the United States with 1,503 respondents, with a representative sample of the population. We report high public awareness of generative AI mistakes. Further, many respondents report checking generative AI output, for example, by comparing results with other online resources. We conclude with guidance for explanations and in-product disclosures about generative AI mistakes.
[HC-45] Baseline Exposure to Common Data Visualization Types Among the U.S. Adult Population
链接: https://arxiv.org/abs/2609.38185
作者: Kiegan Rice,Nola du Toit,Quentin Brummet,Heike Hofmann
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Data visualizations are a primary means of communicating statistical information to the public. Their expanding use across news media, public health, education, and government reporting places greater importance on audiences’ ability to recognize and interpret them. While a significant body of prior research has established frameworks for the measurement of graph and visualization literacy, far less is known about everyday exposure to different kinds of charts and graphs among the general adult population. This gap limits our ability to meaningfully interpret differences in graph literacy, focus efforts on improving that literacy, and design visualizations that align with audience ability. Establishing population level estimates of exposure to data visualizations is therefore essential for improving visual communication and reducing misinterpretation of quantitative information. To address this gap, we surveyed a nationally representative sample of 1,168 U.S. adults via NORC’s AmeriSpeak panel about their exposure to sixteen different common data visualization types. The resulting survey responses demonstrate pronounced differences in baseline exposure to data visualizations across chart types and demographic groups. While widely used formats such as bar, pie, line, and grouped bar charts are highly recognizable to the vast majority of U.S. adults, many other practitioner-favored chart types remain unfamiliar to substantial portions of the adult population. Exposure also differs significantly by age and education, with higher educational attainment linked to greater exposure and older adults reporting lower exposure overall. Our findings provide essential population level context on visualization literacy and underscore the importance of aligning data visualization design with audience exposure and experience.
[HC-46] RealGUINoise: An Interactive Cross-Platform Benchmark for GUI Agent Robustness under Real-World Interface Noise ACL
链接: https://arxiv.org/abs/2609.38184
作者: Yongjiang Wu,Junyuan Zhang,Ada Chen,Kuiyi Gao,Wenxuan Wang
类目: Human-Computer Interaction (cs.HC)
备注: 37 pages, 4 figures. Submitted to ACL Rolling Review (August 2026)
Abstract:Graphical User Interface (GUI) agents and Computer-Using Agents (CUAs) are rapidly becoming practical tools. However, real-world deployment increasingly exposes performance failures and safety risks, while a major yet underexplored source of these problems lies in the complex and noisy conditions of everyday interfaces. Existing benchmarks largely assume clean environments or focus narrowly on security-specific settings, and lack a unified framework for consistent, automated end-to-end evaluation across diverse agents and platforms. Hence, we introduce RealGUINoise, an interactive cross-platform, extensible benchmark for systematically evaluating GUI Agents under common realistic interface noise in fully interactive environments. Specifically, RealGUINoise comprises 42 noise types spanning web, desktop, and mobile tasks and integrates 7 representative agent frameworks. It evaluates these agents on real-world daily tasks through real-time interaction, comparing their performance against task-specific golden rubrics and clean-environment trajectories in terms of reliability, safety, and trajectory-level behavior. Our experiments show that these noises not only degrade task performance but also substantially redirect agents’ action trajectories and increase their propensity for unsafe behavior. These findings expose a critical gap between capability in clean environments and dependable operation in real-world settings, establishing RealGUINoise as a testbed for developing more robust and trustworthy GUI agents.
[HC-47] One Tool One Taste? How Vibe Coding Trades Collective Diversity for Individual Creativity
链接: https://arxiv.org/abs/2609.38183
作者: Léonard Boussioux,Ziyi Zhao,Kanghyun Cho
类目: Human-Computer Interaction (cs.HC); Computers and Society (cs.CY)
备注:
Abstract:Vibe-coding systems turn natural-language instructions into deployed websites, but little is known about the diversity of designs produced within a shared production context. We study 73 promotional websites created by graduate students with Lovable for distinct real businesses under a graded assignment that rewarded original design. We represent each homepage using DINOv3 embeddings and analyze pairwise cosine similarity, effective diversity, and descriptive cluster structure. A representative UMAP–HDBSCAN specification assigns 88% of sites to six interpretable clusters, while sensitivity analyses yield between three and twelve clusters. In a process subsample of nine recordings, preliminary AI-assisted coding documents recurring generate-and-check behavior, incomplete transmission of verbalized intentions into prompts, and infrequent explicit evaluation of originality. Among survey respondents, perceived human control, satisfaction, and perceived quality show no detectable association with embedding-based cohort atypicality. These results document within-cohort visual concentration and a mismatch between subjective experience and distributional distinctiveness. We call this condition authored ignorance. Alongside early evidence that the collective cost of generative production reaches visual design, we contribute a measurement strategy that joins outcome, process, and perception for interactive artifacts, and design implications running from originality feedback to deliberate friction.
[HC-48] EmAvatar: Multimodal Empathetic Response Generation via Conflict Resolution and Expressive Guidance
链接: https://arxiv.org/abs/2609.38182
作者: Xiaolin Chen,Xuemeng Song,Jinlan Fu,Weili Guan,Mong-Li Lee,Wynne Hsu
类目: Human-Computer Interaction (cs.HC); Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注:
Abstract:Avatar-based multimodal empathetic response generation has emerged as a pivotal capability in human-centric systems, aiming to recognize user emotions and synthesize responses with synchronized text, audio, and talking-face video. Despite recent progress, existing methods still suffer from three critical limitations: (1) overlooking conflicting emotions across modalities, (2) lacking explicit multimodal synthesis guidance, and (3) neglecting inherent error propagation of multimodal response generation. To address these limitations, we propose EmAvatar, a novel framework for precise emotion perception and expressive response generation. It first performs deliberative multimodal emotion recognition by exposing inter-modal prediction conflicts and then initiates a multi-round QA process between a Conflict Inspector and an Evidence Collector to gather evidence for conflict resolution, leading to a robust, evidence-aware prediction. Regarding response generation, EmAvatar first synthesizes a composite script that couples the textual response with an expressive instruction. Moreover, to ensure high-quality synthesis, an iterative refinement mechanism evaluates and revises the script until it aligns with predefined criteria, serving as reliable guidance for subsequent audio and video synthesis. Extensive experiments across four tasks demonstrate that EmAvatar outperforms state-of-the-art methods. Our code will be publicly released.
[HC-49] A barrier or a booster? Familiarity effects on Mandarin emotion prosody recognition using AI-powered voice cloning INTERSPEECH2026
链接: https://arxiv.org/abs/2609.38794
作者: Feng Xu,Gaoyuan Zhang,Shanshan Xue,Yixiang Chen,Hanrui Zhou,Xurong Xie,Hui Chen
类目: Audio and Speech Processing (eess.AS); Human-Computer Interaction (cs.HC); Sound (cs.SD)
备注: Accepted by Interspeech 2026
Abstract:Emotion prosody perception requires simultaneous processing of acoustic cues and speaker identity. While listeners effortlessly decode natural speech, AI synthetic voices introduce cognitive complexities due to subtle acoustic atypicalities. It remains unclear how these synthetic features interact with a listener’s prior social knowledge and memory of a familiar speaker. This study investigated how speech sources (human vs. AI) and speaker familiarity affect emotion recognition accuracy and cognitive load. A within-subject task with Mandarin-speaking adults evaluated behavioral (accuracy, reaction time) and physiological data (heart rate variability). Results showed that human voices yielded significantly higher accuracy and faster processing times than AI voices, while HRV did not significantly differentiate between conditions. These findings show that decoding synthetic speech is gated by top-down social cognition, highlighting limitations in current AI synthesis technologies.
计算机视觉
[CV-0] Multimodal Flow: Unified Flow Modeling of Language and Vision in Embedding Spaces
链接: https://arxiv.org/abs/2609.40362
作者: Hongyuan Tao,Xinggang Wang,Lianghui Zhu,Yongkang Li,Yunchao Wei,Bin Feng,Shaoyu Chen,Qian Zhang,Chang Huang,Kai Yu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 18 pages, 5 figures, 10 tables. Code and model: this https URL
Abstract:We present Multimodal Flow, a fully continuous generative model of language and vision. Most unified multimodal models either model both language and quantized images as discrete tokens or combine discrete language prediction with continuous image generation. The former introduces a visual quantization bottleneck. The latter requires modality-dependent objectives and sampling procedures. Fully continuous modeling avoids these trade-offs and enables a shared generative process, but remains underexplored for multimodal pretraining. Multimodal Flow introduces a unified continuous architecture that integrates multimodal continuous representations with a shared chunk-causal flow backbone. It organizes text blocks and images as ordered continuous hyperchunks, preserving textual token order and visual spatial structure. The backbone learns a single vector field over these hyperchunks through Flow Matching. Joint attention enables cross-modal interaction, while modality-specific feed-forward networks process each modality. The model predicts multiple target chunks in parallel during training and generates hyperchunks sequentially at inference. We instantiate MF-1 and pretrain it on multimodal data. Across 0.6B, 1.2B, and 1.6B scales, continued pretraining consistently improves multimodal modeling. With only 150B pretraining tokens, MF-1 achieves an average score of 82.8 across GenEval and DPG-Bench and 75.3 across VQAv2, MMBench, and POPE, remaining competitive with unified models trained on substantially more data. Under matched data, optimization, and parameter budgets, Multimodal Flow further outperforms representative hybrid and discrete models. These results establish continuous chunk-based embedding flow modeling as a new fully continuous paradigm for unified multimodal modeling. The related code and model are publicly released at this https URL.
[CV-1] Physis-Lang: Self-Evolving Language as a Physical Representation for Video World Model
链接: https://arxiv.org/abs/2609.40358
作者: Liming Lu,Xianzheng Ma,Wenkun He,Guanqi Zhan,Yilin Zhao,Junyu Chen,Mengyao Xu,Jiaojiao Fan,Wenhang Ge,Yuchao Gu,Yunze Liu,Boyi Li,Zhen Dong,Victor Prisacariu,Ming-Yu Liu,Song Han,Han Cai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video world models are expected to predict how the physical world evolves, yet they often produce visually plausible videos that violate basic physical principles. Existing approaches commonly assume that natural language is insufficient to represent the physical knowledge required for reliable generation, and therefore introduce additional visual, latent, numerical, or planning-based signals. We revisit this assumption and introduce Physis-Lang, a self-evolving framework that treats physical language as a shared and optimizable representation across data curation, model training, and video generation. Physis-Lang represents physical processes through language that describes their relevant entities, causes, interactions, governing principles, temporal evolution, and effects. To improve this representation, we construct PhysCapBench, which decomposes physical processes into atomic assertions and evaluates captions using recall and precision. An agentic loop iteratively analyzes assertion-level errors and refines the instruction used to produce physical captions. Physis-Lang further converts model deficiencies into textual descriptions and uses language-guided retrieval to identify visually diverse videos that cover missing physical processes. Experiments on four widely used physical video benchmarks with Wan and Cosmos backbones demonstrate consistent improvements in physical plausibility. Notably, starting from open-source Cosmos3-Nano backbones, our Physis-Lang-enhanced models surpass the leading proprietary Veo 3.1 model.
[CV-2] ViTeX-Bench: Benchmarking High-Fidelity Video Scene Text Editing NEURIPS2026
链接: https://arxiv.org/abs/2609.40356
作者: Xinghao Chen,Xiangbo Gao,Jiongze Yu,Yuheng Wu,Zhengzhong Tu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to NeurIPS 2026 (Evaluations and Datasets Track). 27 pages (10-page main text), 5 figures, 12 tables. Project page: this https URL
Abstract:Recent video generation is increasingly realistic and controllable, yet video editing remains less developed, particularly for precise local edits that must preserve the original scene dynamics. Video scene text editing replaces text on scene surfaces, such as storefront signs, whiteboards, and product labels, while preserving the surrounding content, motion, and camera dynamics. Although scene text editing is well studied for images, video scene text editing that achieves high visual quality, temporal consistency, and edit locality remains underexplored. Existing resources offer limited paired real-video data, and general video-editing metrics do not directly measure whether the requested text remains correct over time. We introduce ViTeX-Bench, a benchmark suite comprising ViTeX-Dataset and a three-axis evaluation protocol. The dataset contains 387 real-world 720p videos with text-region masks and editing instructions: 230 provide reviewed, pipeline-generated paired edits for training, and 157 form a frozen evaluation split. The protocol evaluates text correctness, visual and temporal quality, and edit locality through 13 metrics, with one primary metric per axis and a Pareto comparison of their trade-offs. OCR calibration, human evaluation, and annotation-sensitivity analyses support the interpretation of these scores. Across eight baselines from four editing families, accurate text, temporal stability, and scene preservation remain difficult to achieve together. We also release ViTeX-Edit-14B, an open-source reference editor fine-tuned on the paired training split with motion-aligned glyph-video conditioning. It achieves CharAcc 0.688, the highest mean among the evaluated video-native editors, and the lowest comparable text-crop Warp among raw editor outputs. ViTeX-Bench provides a reproducible foundation for studying these trade-offs in video scene text editing.
[CV-3] AssemblyWorld: Rethinking 3D Assembly with General-Purpose Agents
链接: https://arxiv.org/abs/2609.40353
作者: Jiahao Zhang,Yeying Fan,Moitreya Chatterjee,Suhas Lohit,Bernhard Egger,Tim K. Marks,Anoop Cherian,Stephen Gould
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 24 pages, 11 figures. Project page: this https URL
Abstract:The task of 3D assembly requires translating an understanding of parts and their relationships into precise spatial arrangements. Can pretrained general-purpose agents assemble objects through visual interaction without additional assembly-specific fine-tuning? To investigate this question, we introduce AssemblyWorld, an interactive 3D environment in which agents inspect rendered views and manipulate supplied rigid parts, guided by images or assembly manuals when available. Agents perceive part geometry through 2D views rather than direct access to mesh vertices or faces, while their resulting assemblies are evaluated geometrically. Building on this environment, we construct AssemblyWorldBench, comprising 100 assembly tasks across 80 objects spanning furniture, industrial assembly, and fracture reassembly. Evaluating eight agent systems reveals substantial differences in their capabilities. The strongest system achieves 80.9% part accuracy but 59.4% complete-assembly success. The evaluated open-source systems lag substantially behind their stronger closed-source peers in both execution reliability and assembly accuracy. Analyses of visual references, interaction trajectories, and failures show how agents revise assemblies while leaving residual positioning errors. AssemblyWorld provides a common setting for both assessing the capabilities of interactive assembly agents and characterizing the gap between approximate structure recovery and precise reconstruction.
[CV-4] Image Classifiers are Efficient Self-Supervised Video Representation Learners BMVC2026
链接: https://arxiv.org/abs/2609.40347
作者: Owais Iqbal,Sudipta Sarkar,Shyam Marjit,Omprakash Chakraborty,Anirban Chakraborty,Abir Das
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted in BMVC 2026
Abstract:We introduce VideoMSN, a Masked Siamese Network framework for efficient self-supervised spatio-temporal representation learning in videos. Instead of relying on heavy 3D architectures or reconstruction-based autoencoders for learning with unlabeled data, we repurpose standard image Vision Transformers by representing videos as super images which are grids composed of frames sampled from videos. From each super image, we construct two views: one with spatial patch masking and the other with temporal frame masking, ensuring no information leakage across frames. A shared Vision Transformer (ViT) encoder aligns their embeddings using a masked Siamese loss, capturing both motion and appearance cues without reconstruction. Our decoder-free formulation leverages an image foundation model towards efficient video representation learning. Starting from pretrained DINO-v3 and DeiT-v3 image encoders, VideoMSN achieves state-of-the-art performance on Kinetics-400, UCF101, and HMDB51 while requiring up to 32\times fewer and 160\times fewer video pretraining epochs compared to prior video self-supervised learning methods. Our proposed approach also shows strong performance in low-shot classification, confirming the transferability of the learned representations in a label-scarce scenario. Project Page: this https URL.
[CV-5] Ego4WAM: What Matters When Scaling Egocentric Human Data for Robot Learning?
链接: https://arxiv.org/abs/2609.40341
作者: Zhihao Sun,Liu Liu,Xinjiang Wang,Haoyi Jiang,Wei Feng,Huiqiang Zhang,Xiaosong Jia,Zhizhong Su,Zuxuan Wu
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Egocentric human data provides a scalable source of experience for robot learning, but varies substantially in human-robot alignment, behavioral coverage, and available supervision. Existing work shows favorable scaling with increasing human data, but it remains unclear which data properties drive downstream robot gains and how to use such data throughout the training pipeline. We present a systematic study of egocentric human data with different alignment and supervision under a unified world-action model framework. With the model backbone fixed, we disentangle the effects of human-robot alignment, data duration and task diversity, action supervision, and data usage strategies. We find that aligned human demonstrations substantially improve out-of-distribution generalization and reduce target-task robot data requirements; data duration and task diversity affect downstream capabilities differently; and video-only supervision remains effective without action labels, providing a strong foundation for subsequent video-action training. We validate these findings through closed-loop policy evaluation on both real robots and RoboDojo. Rather than treating data duration as the sole scaling axis, Ego4WAM shows how alignment, task diversity, available supervision, and usage strategy jointly shape the value of egocentric human data for robot learning.
[CV-6] I Have a Stream: Making Self-Supervised Learning Work on Continuous Video NEURIPS2026
链接: https://arxiv.org/abs/2609.40333
作者: Ivan Martinović,Lukas Knobel,Yuki M. Asano
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint. Accepted to NeurIPS 2026
Abstract:Self-supervised learning draws inspiration from infant visual development, yet standard training pipelines bear little resemblance to it: images are independently sampled and globally shuffled across epochs. We study self-supervised learning from continuous video streams, where frames are consumed in temporal order using strict sliding-window batches, without global reshuffling or multi-epoch replay. To this end, we construct WT++, a 95-hour urban walking-tour video dataset for streaming pretraining. Combined with a comprehensive evaluation suite we find that contrastive and distillation-based methods struggle in this setting, while MAE is more robust but still falls short of standard i.i.d. pretraining. We find that high inter-batch similarity, caused by sliding-window consumption across consecutive batches, does not explain this gap. The main challenge is high intra-batch similarity, where frames within each batch are near-duplicates. To mitigate this, we propose StreamMAE, which preserves the core MAE reconstruction objective while adapting the input pipeline with stream-aware regularization and motion-biased crop selection. StreamMAE outperforms streaming baselines, matches i.i.d. MAE trained on the same video data, remains competitive with ImageNet-pretrained MAE, and scales positively as the pretraining stream grows from 12 to 95 hours.
[CV-7] Atomizer-IO: Beyond Pixels Patches and Grids
链接: https://arxiv.org/abs/2609.40320
作者: Hugo Riffaud de Turckheim,Sylvain Lobry,Nicolas Houdré,Damien Robert,Roberto Interdonato,Diego Marcos
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Most vision architectures assume that observations lie on a regular grid, an effective abstraction for natural images but a restrictive one for sensing data whose channels, temporal sampling, spatial resolution, and geometry can vary. Generic set-based architectures remove the grid, but also remove useful spatial inductive biases. We introduce Atomizer-IO, an architecture that places observations first and derives structure from their physical relationships. Building on top of an atomic representation of the data, each observation is described by its measurement and acquisition metadata, while local cross-attention maps observations to anchor points that can be arbitrarily placed. We evaluate this design by progressively relaxing the grid assumption, from varying input raster configurations and incomplete channel sets to flexible output density and, ultimately, inputs without a raster grid. Atomizer-IO is competitive with flexible EO-specific architectures on most tasks, while offering post-training control over inference cost and competitive compute–performance trade-offs. The same formulation extends without architectural redesign to unordered 3D point clouds, showing that the atomic interface generalizes beyond regular raster inputs. These results suggest that pixels, patches, and grids do not need to define the interface of a sensing architecture.
[CV-8] GLARE: Generating Listening Heads with Appropriate Reactions NEURIPS2026
链接: https://arxiv.org/abs/2609.40317
作者: Zikai Liao,Yumin Suh,Yi Ouyang,Yi-Lun Lee,Yi-Hsuan Tsai,Zhaozheng Yin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted in NeurIPS 2026. Project page: this https URL
Abstract:While talking head generation has advanced rapidly, generating natural listener behavior in dyadic conversations, which know when to react, how to react, and with what type of response, remains underexplored. Existing dyadic datasets lack fine-grained listener reaction annotations, and prevailing evaluation metrics inherited from talking-head and video generation measure visual realism rather than whether a listener reacted appropriately. We address these gaps along three aspects. First, we curate a listening-head-specific dataset built from RealTalk and Seamless Interaction, comprising approximately 147 hours of paired speaker-listener videos with 64,557 event-level reaction annotations across six categories: nodding, head shaking, smiling, laughing, frowning, and surprised. Second, we introduce an audio-driven baseline built on a flow-matching transformer, namely GLARE, with prosody conditioning derived from Qwen2-Audio and a temporal reaction loss that explicitly supervises frame-wise reactions. Third, we propose a reaction-oriented evaluation protocol that jointly measures reaction occurrence (R-F1), temporal alignment (R-tIoU), asymmetric temporal deviation (R-ATD), and reaction-region visual quality (R-FID), giving a more behaviorally grounded assessment than visual-quality-only metrics. Experiment results show consistent gains over prior listening-head methods in both visual fidelity and reaction-level metrics, suggesting that reaction-aware data, modeling, and evaluation are critical for natural listening behavior.
[CV-9] Looped Diffusion Transformer
链接: https://arxiv.org/abs/2609.40305
作者: Yong Xien Chng,Tianyi Chen,Wenwen Tong,Haiwen Diao,Zhongang Cai,Lei Yang,Ziwei Liu,Lewei Lu,Dahua Lin,Gao Huang
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 21 pages, 9 figures
Abstract:Improving text-to-image models has traditionally relied on increasing model size or the number of denoising steps. In this work, we explore an alternative way to scale computation by repeatedly running shared Transformer blocks within each denoising step, effectively increasing computational depth while keeping the parameter count fixed. This looped computation enables iterative refinement of internal representations without explicit reasoning tokens. However, naive looping fails to consistently improve image quality. We trace this problem to weak supervision across intermediate loops and unregulated attention updates that progressively erode local information. To overcome these challenges, we propose Looped Diffusion Transformer (Looped-DiT), which combines deep supervision across intermediate loops with self-modulating attention to stabilize looped feature updates. Under matched-parameter and matched-compute settings, Looped-DiT consistently outperforms non-looped baselines. Notably, a 260M-parameter looped model can surpass a model 6.5x larger across multiple text-to-image benchmarks while requiring 4.9x lower inference compute. Beyond this performance gain, we find that looped computation can offer a more effective form of iterative computation for diffusion models, with increasing loop depth yielding larger gains than adding more denoising steps under a fixed inference budget. Furthermore, deeper loops can progressively correct mistakes made in earlier loops, exhibiting behaviors suggestive of latent reasoning. Together, these results show that looped computation offers a promising way to scale visual generation models.
[CV-10] ComputerSD: Online Self-Distillation from Real-Time Feedback for Computer-Use Agents
链接: https://arxiv.org/abs/2609.40253
作者: Yong Du,Tongbo Chen,Zhengxi Lu,Yizhou Liu,Bofan Chen,Tao Jiang,Wenhao Xu,Yongliang Shen
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: this https URL
Abstract:Online training enables computer-use agents (CUAs) to improve through interaction with executable environments. However, existing methods primarily rely on sparse outcome rewards, which provide no supervision for intermediate actions. On-policy self-distillation (OPSD) offers token-level learning signals through privileged rescoring, but directly applying it to CUA online training presents two challenges: fixed guidance may become misaligned with the student’s current state, and guidance-induced probability shifts may conflict with step-level correctness. We introduce ComputerSD, an online self-distillation method for CUAs that converts real-time feedback from executed GUI transitions into guidance for policy learning. A fine-tuned GUI analyzer produces guidance and a step-level value score after each action; the guidance provides privileged context, while the score regulates the resulting OPSD signals. ComputerSD jointly optimizes token-level OPSD and trajectory-level GRPO in a fully asynchronous training framework. On OSWorld-Verified, ComputerSD outperforms outcome-only GRPO by 1.9 and 4.1 percentage points on the general-purpose Qwen3-VL-8B-Thinking and specialized EvoCUA-8B backbones, respectively. Evaluation in out-of-distribution settings further supports the generalizability of ComputerSD. These results demonstrate the effectiveness of learning from real-time feedback through online self-distillation for CUAs.
[CV-11] StreamRig: Exploiting Intra-Rig Geometry for Streaming Multi-Camera Odometry
链接: https://arxiv.org/abs/2609.40244
作者: Yufei Wei,Shuhao Ye,Qi Wang,Xin Zheng,Qing Huang,Rong Xiong,Yue Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 8 pages, 4 figures, 5 tables. Code: this https URL
Abstract:Mobile robots and vehicles carry synchronized multi-camera rigs, yet many streaming 3D foundation models are designed for monocular input, leaving efficient use of rig geometry a challenge. We present StreamRig, a freeze-and-stream framework that builds causal streaming odometry for calibrated rigs on a frozen multi-view 3D foundation model. The frozen front-end jointly perceives the synchronized views using rig calibration. A Rig-Resampler compresses their features, a CausalBridge applies causal attention with a key-value cache, and a lightweight head regresses rig poses. A periodic re-anchoring protocol supports stable pose estimation over long sequences. Only these modules are trained, 74.6M parameters in total, with relative poses as the sole supervision. Our two-stage training strategy combines group relocalization pretraining with causal rig training to transfer the geometric priors of the frozen front-end and the alignment ability of the pretrained modules to streaming odometry. We evaluate on NCLT, TartanGround, KITTI-360, and our self-collected humanoid-robot dataset ZJH, where training uses only simulation and real-world evaluation is zero-shot. Across all four datasets, StreamRig achieves lower translation and rotation drift than the evaluated non-oracle monocular streaming and rig-aware offline models, while maintaining low inference cost. Ablations and controlled camera-count experiments identify the sources of these gains. We further examine how longer training windows affect inference over longer horizons. Code has been released at this https URL.
[CV-12] EviRover: Reinforcing Agent ic Perception Beyond a Glance
链接: https://arxiv.org/abs/2609.40230
作者: Kaixuan Fan,Kaituo Feng,Tianshuo Peng,Yilei Jiang,Manyuan Zhang,Junke Wang,Xiangyu Yue
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Visual perception is conventionally formulated as a one-shot prediction from a single glance at the image, under the assumption that the image content and the model’s parametric knowledge suffice to resolve the query. This assumption often fails in real-world scenarios that hinge on fine-grained visual details or require knowledge-intensive and up-to-date information. We term such cases \textitperception under insufficient evidence and formulate perception as an agentic process that can obtain information beyond a single glance. To address the absence of data for this setting, we design two dedicated data generation pipelines, yielding EviRover-SFT-5K and EviRover-RL-12K for training. We further construct EviLens, a human-verified benchmark comprising 688 instances across five perception categories. Building on these data, we present EviRover, to our knowledge the first perception agent explicitly trained to resolve perceptual queries through interaction, using supervised fine-tuning followed by agentic reinforcement learning. Experiments show that the 4B EviRover outperforms its backbone by 30 points on average on EviLens, reaching performance comparable to advanced proprietary models. The gains transfer beyond EviLens to WebEyes, conventional perception benchmarks, and general multimodal benchmarks, including a 15-point improvement on BrowseComp-VL. All code, models, and data are released.
[CV-13] LOCI: Spatial Linear Memory for Streaming World Models
链接: https://arxiv.org/abs/2609.40222
作者: Ji Xia,Tingting Liao,Xuezhi Liang,Hao Li,Guangyi Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 25 pages, 8 figures, 14 tables. Project page: this https URL
Abstract:When a camera revisits a previously observed region, a video world model should reproduce what was there before. This requires both remembering past observations and retrieving the right one for the current viewpoint. Key-value caches preserve visual detail but grow with video length; recurrent memory is compact but compresses history into a fixed-size state, so individual past observations are no longer directly accessible. We introduce LOCI, a hybrid spatial-memory architecture that keeps both representations. In half of the transformer blocks, main attention keeps a key-value cache of past observations; in the other half, it is restricted to the current chunk and complemented by a recurrent linear-attention memory whose reads and writes are conditioned on projective camera geometry, so viewpoint enters both memory addressing and stored content. Recurrent readouts flow into subsequent cache-backed blocks and supply their queries with accumulated scene context. On the public MIND memory benchmark and on held-out recorded trajectories, LOCI reproduces revisited content more faithfully than representative world models and a same-recipe full-softmax model; with full history, it lowers peak memory at equal length by about 30% relative to full softmax. With a bounded bank of retained observations, it streams long videos at constant memory and remains more faithful than full softmax under the same budget.
[CV-14] Learning Skills from Historical Action Trajectories: Action Experience Dictionary for World Action Models
链接: https://arxiv.org/abs/2609.40219
作者: Qi Lyu,Jiahua Dong,Hao Shen,Xudong Wang,Hongyuan Yu,Baichen Liu,Henghui Ding,Zhi Han,Nicu Sebe,Ivan Laptev,Fahad Shahbaz Khan,Salman Khan
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:World Action Models (WAMs) couple visual dynamics prediction with action generation, yet they do not explicitly support the reuse of action experience across manipulation tasks. Furthermore, existing WAMs struggle to capture underlying cross-task semantic relationships that could guide target action prediction, as redundant background elements interfere with the extraction of key visual information. To address these challenges, we develop a novel Action Experience Dictionary (AED) that encodes historical physical action trajectories into shared action embeddings to support skill reuse and model cross-task relationships. Specifically, we first aggregate historical actions to align with visual observations and retrieve action embeddings from the AED using a pretrained action tokenizer. Subsequently, we visually condition the pooled embeddings through cross-attention and prepend them to noisy action tokens, providing interaction context and action intent for prediction. To model action-related motion and reduce reliance on irrelevant background cues, we introduce a motion-aware transition loss that supervises visual feature change prediction over random temporal intervals. Experiments on simulation benchmarks and in real-world cross-embodiment settings verify the effectiveness of our AED. The anonymous project website is available at \hrefthis https URLAED.
[CV-15] Recognition of Urbanized Areas in UAV-Derived Very-High-Resolution Visible-Light Imagery
链接: https://arxiv.org/abs/2609.40212
作者: Edyta Puniach,Wojciech Gruszczyński,Paweł Ćwiąkała,Katarzyna Strząbała,Elżbieta Pastucha
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:This study compared classifiers that differentiate between urbanized and non-urbanized areas based on unmanned aerial vehicle (UAV)-acquired RGB imagery. The tested solutions in-cluded numerous vegetation indices (VIs) thresholding and neural networks (NNs). The analysis was conducted for two study areas for which surveys were carried out using different UAVs and cameras. The ground sampling distances for the study areas were 10 mm and 15 mm, respectively. Reference classification was performed manually, obtaining approximately 24 million classified pix-els for the first area and approximately 3.8 million for the second. This research study included an analysis of the impact of the season on the threshold values for the tested VIs and the impact of image patch size provided as inputs for the NNs on classification accuracy. The results of the con-ducted research study indicate a higher classification accuracy using NNs (about 96%) compared with the best of the tested VIs, i.e., Excess Blue (about 87%). Due to the highly imbalanced nature of the used datasets (non-urbanized areas constitute approximately 87% of the total datasets), the Mat-thews correlation coefficient was also used to assess the correctness of the classification. The analysis based on statistical measures was supplemented with a qualitative assessment of the classification results, which allowed the identification of the most important sources of differences in classification between VIs thresholding and NNs.
[CV-16] Prototype-Rule Neurosymbolic Regularization for Rank-Constrained Tensor Neural Networks under Label Scarcity
链接: https://arxiv.org/abs/2609.40131
作者: Eftychios Protopapadakis,Konstantinos Makantasis,Konstantinos M. Giannoutakis
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Rank-constrained tensor neural networks reduce the parameterization of high-order inputs, but they do not explicitly constrain class geometry in the learned representation. This study investigates whether a differentiable prototype-rule can provide a complementary inductive bias for Rank-R tensor learning under limited supervision. The proposed framework augments the Rank-R objective with prototype-based regularization and optionally fuses prototype evidence with neural logits at inference. Four hyperspectral benchmarks are evaluated with four Rank-R configurations under both seven-fold stratification and spatially separated folds that mitigate leakage; a separate spatial study varies the class support budget from 2 to 20 samples. Under spatial evaluation, full neurosymbolic inference changes Macro-F1 score by +8.82 percentage points on Botswana, +5.49 on Indian Pines, +1.59 on Pavia University, and -0.62 on Salinas. Most of the benefit arises from training-time regularization, whereas inference fusion is small and dataset dependent.
[CV-17] VR-JEPA: Learning Contrastive-State Latent Guidance for Generation-based Video Reasoning
链接: https://arxiv.org/abs/2609.40129
作者: Zehua Ma,Kun Xiang,Yunshuang Nie,Quanlin Chen,Haoyuan Li,Xiuwei Chen,Jiang Ji,Haijun Wu,Zhenyu Xie,Michael Kampffmeyer,Hanhui Li,Xiaodan Liang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reasoning through video generation offers a promising path toward visual intelligence by modeling latent visual states and their dynamics. However, current video generation models often lack explicit guidance on how these states should evolve, leaving generated trajectories prone to physical and structural inconsistencies that undermine reasoning reliability. While the Video Joint-Embedding Predictive Architecture (V-JEPA) provides rich spatiotemporal priors learned through latent prediction, these general priors do not naturally adapt to the logical reasoning capabilities required for complex visual tasks. To bridge this gap, we propose VR-JEPA, a framework that aligns the V-JEPA predictor with task-specific reasoning logic through localized contrastive-state learning and uses its predicted latent trajectories to guide video generation for visual reasoning. Specifically, (i) we pair successful trajectories with generated alternatives under the same input conditions and use discrepancies in their V-JEPA representations to identify informative states and tokens for localized contrastive supervision. (ii) We further equip the V-JEPA predictor with skill-specific experts trained on anchor-task data, allowing the model to adaptively specialize its shared spatiotemporal priors across diverse cognitive domains. Together with skill-specific experts, this contrastive supervision enables VR-JEPA to predict latent trajectories that provide task-specific logical guidance for video generation. Comprehensive experiments on the large-scale VBVR-Pro-Bench dataset demonstrate that VR-JEPA achieves an 11.33% relative improvement over the cutting-edge generation-based reasoning baseline, significantly mitigating physical artifacts and enhancing logical consistency.
[CV-18] GateSPINE: Gated Cross-View Fusion for Lumbar Spine MRI Report Generation
链接: https://arxiv.org/abs/2609.40091
作者: Hoang Nguyen Van,Cuong Vuong Tuan,Trang Mai Xuan,Bien Tran Van,Nam Tran Van,Thien Van Luong
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Automated report generation can ease the burden radiolo gists face when interpreting multi-sequence MRI studies. Unlike CT, MRI examinations comprise multiple sequences and imaging planes, each con tributing complementary diagnostic information. Existing methods en code a study as a single volume and combine multiple acquisitions by fixed rules. Findings visible in only one plane are thus diluted and of ten missed, lowering recall on clinical efficacy metrics, where a missed abnormality is most costly. We propose GateSPINE, a vision-language framework that fuses sagittal T1 and T2 volumes with a training-free operator, encodes the fused sagittal and axial volumes with two parallel 3D encoders, and decodes their combined representation into a report. Its core mechanism is a gated cross view fusion module that predicts, per feature channel and token, how much of each view to admit, so the more informative view dominates at each spatial location. We evaluate GateSPINE on three lumbar MRI datasets, comprising two public bench marks and a private cohort collected from Phenikaa University Hospital, using both natural language generation (NLG) and clinical efficacy (CE) metrics. GateSPINE achieves the highest CE F1 through improved re call on all three datasets; on SPIDER, which lacks an axial sequence, this reflects the sagittal fusion component rather than the gated cross-view mechanism, which is validated on the two cohorts with both imaging planes. GateSPINE also remains competitive on standard NLG metrics.
[CV-19] LongEmo: Towards Emotion Understanding and Reasoning in Long Videos
链接: https://arxiv.org/abs/2609.40079
作者: Shuo Zhang,Yifan Zhou,Han Wang,Jinsong Zhang,Jingyu Li,Hongbing Li,Zhejun Zhang,Chengyi Zhao,Yuquan Hao,Yitong Liu,Jiyin Li,Ruiqi Tang,Zixuan Lin,Yi Luo,Xurui Zhang,Ronghao Chen,Huacan Wang,Lei Li
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 33 pages
Abstract:While recent Multimodal Large Language Models (MLLMs) have shown promise in affective computing, their reasoning capabilities are largely confined to short video clips with limited interactions. However, real-world emotions are not merely isolated instantaneous reactions but dynamic and cumulative processes deeply shaped by past experiences and ongoing events. To bridge this gap, we introduce LongEmoBench, a benchmark dedicated to emotion understanding and reasoning in long videos. It assesses progressive capabilities scaling from continuous scene interactions to complex episodic developments. Furthermore, we propose LongEmo, a novel memory-augmented agentic framework designed to tackle the immense challenges of long-range affective reasoning. LongEmo processes continuous video streams to construct an Event Memory Graph, explicitly modeling long-range dependencies and capturing emotional dynamics across discrete events. Given a question, the agent retrieves a query-relevant event stream from the graph, iteratively integrating multimodal memories and relational dependencies to deduce the final answer. Extensive evaluations of 17 representative methods reveal that they struggle significantly with emotion understanding and reasoning in long videos. In contrast, LongEmo achieves state-of-the-art performance, demonstrating the efficacy of its event-centric memory architecture.
[CV-20] Less Data Better Timing: Student-Curriculum Coupling for VLM On-Policy Distillation in Temporal Video Grounding
链接: https://arxiv.org/abs/2609.40055
作者: Jiacheng Qiu,Yunsoo Kim,Ruichen Xu,Jian Luo,Petar M. Djurić,Sima Mofakham
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:On-policy distillation (OPD) provides dense supervision directly on student-generated trajectories, making it an effective post-training strategy for vision-language models in temporal video grounding (TVG). However, existing pipelines typically construct the training curriculum from a fixed teacher and the initial student state, implicitly assuming that selected examples retain positive supervision value throughout optimization. We show that supervision trustworthiness and supervision necessity are distinct yet coupled: the former concerns target credibility, while the latter varies with the student’s current task competence; together, they shape supervision value. Building on this coupled view, we introduce Student-Curriculum Coupling (SCC), a closed-loop framework in which a compact Anchor-Frontier curriculum defines the candidate supervision space and the evolving student dynamically determines its active subset. Supervision can therefore be activated, suspended, or reactivated as competence changes, concentrating teacher computation and optimization on current task-level deficits. Across three TVG benchmarks, SCC achieves a 5.1% relative improvement in mean recall over Video-OPD on its original curriculum, while using 60.0% fewer training examples and reducing training time by 50.4%. Ablations support the complementary roles of capability-structured curriculum design and student-dependent supervision in achieving these gains. Together, these results establish SCC as a data- and compute-efficient framework for TVG post-training, delivering stronger temporal grounding by aligning trustworthy supervision with the student’s evolving learning needs.
[CV-21] CoEvoWhen: Policy-Tool Coevolution for Ultra-Long Video Temporal Grounding
链接: https://arxiv.org/abs/2609.40048
作者: Yiduo Jia,Muzhi Zhu,Jinchuan Shi,Hao Zhong,Yuling Xi,Ke Liu,Hao Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Ultra-long video temporal grounding requires balancing long-range evidence search with fine-grained event understanding under a limited visual budget, yet existing agentic methods still rely largely on predefined policies and tool capabilities. Motivated by this, we propose a novel policy-tool coevolution framework that jointly evolves high-level policies and executable media tools from the agentic reasoning trajectories of a VLM, forming a reusable skill without updating model parameters. During evolution, an external skill updater distills transferable task experience in long-video temporal grounding, accordingly refining the orchestration of long-range image-based and fine-grained video-based observations. Alongside these policy updates, the updater employs its coding capabilities to upgrade existing tools or create new ones, adapting the tools to long-video evidence acquisition. Equipped with the evolved skill, the VLM autonomously orchestrates tools under the guidance of the evolved policy, coordinating image and video observations for agentic inference without relying on a separate, stronger planning model. Extensive experiments spanning five benchmarks and three VLMs show that policy-tool coevolution consistently improves temporal grounding accuracy in ultra-long videos while reducing visual token cost at inference, and that the evolved skill yields substantial performance gains on general long-video QA without additional task-specific evolution, demonstrating the effectiveness and generalizability of our framework for long-video understanding.
[CV-22] Enhancing Autoregressive Video Generation via Representation Adversarial Distillation
链接: https://arxiv.org/abs/2609.40037
作者: Fangyu Lin,Xingtong Ge,Lunjie Zhu,Yi Zhang,Zhening Liu,Tianhang Wang,Mengfei Li,Yumeng Zhang,Guanglu Song,Yu Liu,Jun Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Few-step autoregressive video generation enables efficient streaming synthesis, but errors introduced in early temporal blocks are reused as context and can propagate through subsequent rollouts, leading to detail degradation, structural drift, and unstable motion. Existing distribution matching distillation (DMD) primarily aligns student and teacher distributions in diffusion latent space, but provides no direct supervision over the perceptual quality of decoded videos. We introduce Radian, a representation-space adversarial distillation framework that complements on-policy DMD with real-data adversarial supervision in the feature space defined by a frozen visual foundation model (VFM). During training, Radian sparsely decodes frames from autoregressive student rollouts, extracts multi-level visual representations, and applies lightweight discriminator heads to distinguish generated outputs from real video frames. The DMD objective anchors the student to the pretrained teacher, while the representation-space adversarial objective supplies complementary perceptual and semantic gradients that promote high-quality modes. These additional components are discarded after training, leaving the generator architecture and inference-time denoising budget unchanged. Experiments on Wan2.1-1.3B cover four-step chunk-wise, one-step frame-wise, and minute-long autoregressive generation. Our method achieves a VBench Total of 0.8444 and a VideoAlign Total of 0.8033 under four-step generation, and improves VBench-Long from 0.7805 to 0.8041 over Rolling Forcing while using fewer denoising steps. Controlled comparisons across image, video, and diffusion representations further indicate that the choice of representation spaces induces distinct adversarial signals, and external VFM gradients complement DMD more effectively than adversarial supervision derived from diffusion-internal features.
[CV-23] WARP: A Unified Benchmark for Invisible Image Watermarking – Robustness and Protection Against Attacks ACM-MM2026
链接: https://arxiv.org/abs/2609.40031
作者: Khaled Abud,Aleksey Yakushev,Aleksandr Akimenkov,Irina Serzhenko,Kirill Aistov,Egor Kovalev,Dmitry Obydenkov,Sergey Lavrushkin,Anastasia Antsiferova,Dmitriy Vatolin,Yury Markin,Kirill Lukianov
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Multimedia (cs.MM)
备注: Accepted to ACM MM 2026 (Main Track)
Abstract:Digital image watermarking is increasingly critical in media contexts, as emerging regulations and industry practices require marking AI-generated content and ensuring traceable sources to prevent manipulation or misuse. Recent advances in invisible watermarking methods highlight the need to update existing benchmarking practices to reflect current techniques and evaluation criteria. We address this by introducing WARP – a unified framework and benchmark for evaluating the robustness of invisible watermarks. WARP incorporates 32 recent classical, deep, and generative watermarking methods, as well as 34 different erasing techniques, ranging from traditional distortions to more sophisticated adversarial, purification, and re-embedding attacks. It provides standardized, reproducible, and easily scalable protocols for evaluating perceptual quality, watermark readability, and attack resilience. Using WARP, we extensively evaluate current invisible watermarking techniques, collecting the largest robustness benchmark in the field. Results identify the most robust approaches under both distortion and adversarial conditions, and reveal consistent relationships between watermarking methods and the attack strategies most effective against them. Our experiments also highlight that some of the watermarking methods considered are highly vulnerable to reembedding, even if they are robust to standard distortions. The code is made available at this https URL. Comments: Accepted to ACM MM 2026 (Main Track) Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Multimedia (cs.MM) Cite as: arXiv:2609.40031 [cs.CV] (or arXiv:2609.40031v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.40031 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Related DOI: https://doi.org/10.1145/3767308.3835888 Focus to learn more DOI(s) linking to related resources
[CV-24] Can We Anticipate Violence? Multimodal Learning from Pre-Incident Behavioral Cues
链接: https://arxiv.org/abs/2609.40014
作者: Sindhuja Penchala,Mohammed Yusuf Mujawar,Noorbakhsh Amiri Golilarz,Sudip Mittal,Shahram Rahimi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Detecting violence after it begins is important from recognizing behavioral cues that appear immediately beforehand. This work studies short-horizon pre-incident risk recognition from multimodal video signals. We construct a binary Normal-versus-Risky setting from temporally annotated XD-Violence clips, using 443 samples with source-level separation across training, validation, and test sets. Each sample consists of a variable-length pre-incident clip, with its duration determined by the observable behavioral context preceding the incident. The inci- dent itself is excluded from all input clips. We evaluate three complementary information sources: facial-region appearance, temporally aligned audio, and body-motion features derived from tracked keypoints. Controlled ablations are performed with Swin-Tiny, ViT-Tiny, and DeiT-Tiny to measure the contribution of each modality under the same split. Results show that combining all modalities is more effective than using any other combination alone. The best configuration, Deit-Tiny with audio, facial appearance, and motion, achieves 91.21% accuracy, 88.96% balanced accuracy, 93.65% F1-score, and 96.38% ROC-AUC on the held-out test set. These results suggest that complementary appearance, acoustic, and kinematic cues provide useful evidence for recognizing elevated pre-incident risk.
[CV-25] Multi-Link Safety Filtering for VLA Policies Around Moving Hazards
链接: https://arxiv.org/abs/2609.40007
作者: Yatharth Agarwal,Vijay Raghunathan
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages, 4 figures, 3 tables. Project page: this https URL
Abstract:A vision-language-action (VLA) policy can finish a manipulation task while knocking over objects unrelated to it, so task success alone does not show that the policy is safe to deploy in clutter. We study how to keep a pretrained VLA policy clear of such hazards at run time without retraining it, which requires guarding more of the arm than the end effector, following the hazard as it moves, and sharing onboard compute with the policy. Our training-free shield covers the gripper, wrist, and forearm with five ellipsoids and filters every commanded motion through one barrier program against a keep-out ellipsoid fitted from RGB-D perception at reset. Sparse optical flow then carries that ellipsoid’s center along with the hazard, with no repeated detection or refitting. Over six simulated hazard-motion conditions, the shield lowers collision from 65.62% to 27.27% and raises safe-success, task completion without collision, from 29.35% to 50.43% . Ablations show that guarding the arm links protects beyond end-effector shielding, and that tracking recovers most of the protection lost when the hazard estimate is frozen at reset. On heterogeneous edge hardware, the five-ellipsoid barrier runs on the CPU in 2.2 ~ms at the 99th percentile, and trimming the vision–language prefix and taking fewer flow-matching steps shortens each \pi_0.5 policy call on the integrated GPU from 343 to 177.3 ~ms. On a physical SO-101 arm across four tasks, the arm touched the hazard in 3 of 16 shielded episodes versus 11 of 16 unshielded ones. Project page: this https URL
[CV-26] Reconstructing the Dynamic World: A Representation-Centric View of 4D Scene Reconstruction
链接: https://arxiv.org/abs/2609.39960
作者: Ziren Gong,Guo Chen,Yongjia Li,Yihua Shao,Fabio Tosi,Stefano Mattoccia,Matteo Poggi,Hao Tang,Fei Ma,Shuyan Li,Ziyang Yan,Nicu Sebe,Ling Shao,Jianfei Cai,Qi Tian,Ming-Hsuan Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:4D scene reconstruction aims to recover the evolving geometry, appearance, and motion of dynamic environments from visual observations. Despite substantial progress in neural scene representations, reconstructing dynamic scenes remains challenging due to non-rigid motion, occlusions, temporal inconsistencies, and the trade-offs between reconstruction fidelity and computational efficiency. Recent advances in Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) have introduced diverse approaches to representing and reconstructing dynamic scenes, yet their relationships, underlying design choices, and evaluation protocols remain fragmented. In this paper, we present a unified perspective on 4D scene reconstruction, organizing existing methods around their scene representations, temporal modeling strategies, reconstruction pipelines, and optimization objectives. Through this framework, we examine how different design choices affect geometric fidelity, appearance consistency, motion representation, and computational efficiency. We further consolidate commonly used datasets and evaluation metrics, identify limitations in current experimental practices, and discuss open challenges in reconstructing complex, dynamic real-world environments. By connecting methodological developments with their underlying assumptions and evaluation evidence, this work provides a structured foundation for understanding existing approaches and identifying future research directions. An evolving collection of relevant papers and resources is available at this https URL.
[CV-27] Learning to Reason with Compressed Context: Ground-Truth-Free Adaptation of OmniLLM s via Self-Distillation
链接: https://arxiv.org/abs/2609.39953
作者: Jianghao Wang,Ke Meng,Jian Li,Chi Cheng,Longyu Qi,Liyin Liang,Yifeng Qian,Chunbo Lai,Yutian Lin,Zeyu Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 31 pages, 5 figures. Project page: this https URL
Abstract:Omni-modal large language models (OmniLLMs) enable unified audio-video understanding, but their long multimodal token sequences make deployment computationally expensive. Token compression reduces this cost, yet aggressive compression often lowers accuracy. Existing works predominantly focus on designing better compression mechanisms; however, adapting the underlying language model to reason effectively over the remaining compressed context remains under-explored. To address this, we propose CAFD (Compressed-Context Adaptation via Full-Context Distillation), a ground-truth-free self-distillation framework that adapts OmniLLMs to fixed compression pipelines without requiring reference answers, rationales, or correctness rewards. CAFD leverages the full-token view of the same multimodal sample as a source of privileged information: a full-context self-teacher provides soft target supervision to a compressed-context student along the student’s on-policy trajectory. Evaluated on Qwen2.5-Omni-7B across five audio-video benchmarks, five compression pipelines, and five deployment budgets, CAFD demonstrates consistent gains, improving 120 out of 125 conditions with an average accuracy boost of 1.44 points and recovering 26.9% of the accuracy gap on average. These results demonstrate that the proposed ground-truth-free adaptation offers an effective and practical route to improving the accuracy-efficiency trade-off in deployed OmniLLMs.
[CV-28] Reliability-Aware Checkpoint Selection for Domain Generalization
链接: https://arxiv.org/abs/2609.39934
作者: Jinshi Liu,Jiahao Li,Pan Liu,Yanfeng Li,Rui Qian,Zhao Tong,Yue Sun,Tao Tan
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 28 pages, 5 figures. Project page: this https URL
Abstract:Checkpoint selection in domain generalization often relies on source-validation accuracy, yet the selected checkpoint need not provide reliable probabilities on unseen target domains. Source-target distribution shifts can alter accuracy rankings, while accuracy alone does not measure predictive probability quality. We identify an empirical selection opportunity within fixed training trajectories: reselecting among checkpoints with near-optimal source accuracy can improve mean target probability quality with small observed changes in mean target accuracy. We study accuracy-constrained reliability selection (AC), which retains checkpoints within a tolerance of the best source-validation accuracy and ranks them by source reliability. Our reference rule aggregates within-set normalized negative log-likelihood (NLL) and class-wise calibration error (CwECE) using D_\infty . AC uses no target data and requires neither additional training nor weight averaging. We evaluate five domain generalization training algorithms on three benchmarks, using PACS to develop the objectives and a 0.5-percentage-point tolerance. In exploratory aggregation comparisons on 360 OfficeHome and TerraIncognita runs, the reference rule reduces mean target soft-bin squared-gap ECE and CwECE by 0.240% and 0.182%, respectively, and NLL by 0.030 relative to Source-Acc. Mean target accuracy changes by +0.213 percentage points. These results identify opportunities for reliability-aware reselection, while the additional benefit of joint over single-objective ranking remains unresolved.
[CV-29] Super-Resolving Unseen Hyperspectral Sensors at Any Scale via Spatial Operators
链接: https://arxiv.org/abs/2609.39926
作者: Ji-Xuan He,Guohang Zhuang,Bo Junge,Tingyi Li,Lingchen,Miaomiao Cai,Yanan Qiao,Xiujin Liu,Junfeng Fang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Achieving cross-sensor generalization and arbitrary-scale reconstruction with a single model remains challenging in hyperspectral super-resolution (HSR). Although recent methods support arbitrary-scale reconstruction, applying them to new sensors or scales beyond the training range often requires additional data and computation to maintain reconstruction quality. To address these challenges, we propose OmniHSR, which predicts band-shared spatial operators rather than spectral values. Cross-Spectral Mapping (CSM) resamples inputs with any number of bands to fixed reference positions and predicts local operators with Gaussian supports. Continuous Operator-Field Reconstruction (COFR) composes these operators into a continuous field and applies them to all original bands for arbitrary-scale reconstruction. Experiments demonstrate that operator prediction outperforms direct spectral-value prediction on all seven datasets. Trained solely on ARAD with only 0.538M parameters, OmniHSR outperforms all directly transferred baselines on six unseen datasets without target-domain training data or adaptation. Across twelve upsampling factors from \times2 to \times48 , it improves average PSNR on Pavia U and Chikusei by 0.55 dB over the strongest baseline. It also surpasses baselines trained from scratch or adapted on the target sensor and achieves up to 36\times faster inference. Our code will be publicly released soon.
[CV-30] CoVisco: Codec-Native Vision Encoder with Native Token Compression for Unified Image-Video Understanding
链接: https://arxiv.org/abs/2609.39924
作者: Yulong Liu,Xiaotian Han,Junyuan Shang,Yuchen Ding,Zhenyu Zhang,Shuohuan Wang,Guibo Zhu,Sirui Han,Dianhai Yu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision-language models face a fundamental scaling bottleneck: the number of visual tokens grows with both temporal duration and spatial resolution, making long-video understanding expensive for the vision encoder and the language model. Existing methods often compress visual tokens after dense encoding, creating a mismatch between the representation used during training and the compact interface required at deployment. We present CoVisco, a codec-native vision encoder with native token compression for unified image-video understanding. By combining codec-native input support with segmented attention, CoVisco can encode long visual inputs in a single forward pass without forming dense patch-to-patch interactions across all frames. Each temporal segment is equipped with learnable abstract tokens that learn a compact segment-level representation, while fine-grained patch tokens remain available throughout the encoder. Alternating intra-segment and abstract-communication layers preserve video-level context through the abstract-token channel. A lightweight selector further exposes either abstract tokens alone or abstract tokens augmented with a runtime-selected subset of patch tokens, yielding a compact visual interface that reduces the visual context and prefill burden of downstream MLLMs while retaining fine-grained evidence when needed. Pretrained with contrastive objectives on 565M image–text pairs and 6.4M videos, CoVisco shows competitive performance on video-oriented embedding and multimodal understanding benchmarks. In the evaluated four-segment, 64-frame setting, abstract-only inference uses only 400 visual tokens while achieving video-understanding performance close to, and on some benchmarks exceeding, OneVision-Encoder. Selected patch tokens further improve fine-grained video reasoning. Project URL: this https URL
[CV-31] NavHarness: Adaptive Goals for Agent ic Vision-Language Navigation
链接: https://arxiv.org/abs/2609.39915
作者: Haoxiang Shi,Zaijing Li,Muhe Ding,Xiang Deng,Yaowei Wang,Liqiang Nie
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 22 pages, 10 figures
Abstract:Vision-Language Navigation (VLN) requires embodied agents to generate actions based on instructions and observations. General-purpose multimodal agents offer a promising basis for this task, but selecting plausible local actions does not ensure that execution remains consistent with the intended route, particularly in long-horizon tasks. Moreover, the accumulated interaction history increases the input required for subsequent decisions, resulting in a significant inference overhead. To this end, we introduce \method, an Agentic VLN framework that includes a Goal Agent that sets adaptive goals for local actions, a Verify Agent that dynamically verifies whether a goal has been completed, a Memory Agent for multimodal context compression, and a Visuomotor Agent to execute adaptive goals. Specifically, the Goal Agent formulates adaptive goals based on the instruction, current observation, and execution history. Then the Visuomotor Agent executes navigation actions to achieve each goal, while the Verify Agent uses a goal-specific verification question to dynamically assess whether the observed outcomes satisfy the intended completion condition. Verified goal completion then marks a boundary for the Memory Agent to compress the corresponding multimodal interaction history while preserving information needed for subsequent navigation. We evaluate navigation on R2R-CE and RxR-CE, examine framework variants across three model backbones, and study context evolution during execution. For Real-World evaluation, \method achieves 83.3% success and 1.51,m navigation error across eight challenging routes evaluated three times each.
[CV-32] Learning Where to Look: Anatomical Grounding and Guided Attention for Cardiac MRI Vision-Language Models
链接: https://arxiv.org/abs/2609.39899
作者: Bangwei Guo,Xiao Chen,Boris Mailhe,Jia Yao,Yiqing Wang,Ankush Mukherjee,Yikang Liu,Zheyuan Zhang,Hang Yu,Terrence Chen,Shanhui Sun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Cardiac magnetic resonance imaging (CMR) enables assessment of cardiac anatomy, ventricular function, and myocardial tissue characteristics. Clinicians interpret these images by identifying cardiac structures and focusing on the regions relevant to each clinical question, motivating anatomically guided vision-language models (VLMs). Yet CMR-specific supervision for anatomical localisation and clinical question answering remains limited. To address this gap, we investigate fine-grained CMR visual question answering through anatomical grounding and guided attention. We construct 128,915 anatomical-grounding and 42,799 clinical QA pairs across short-axis cine, late gadolinium enhancement, and long-axis cine. These datasets support anatomical recognition, localisation, and clinical assessment without requiring paired reports for individual training images. To help the model learn where to look, we introduce Cardiac Anatomy-Routed Attention (CARA), which selects predicted anatomical priors according to the question and guides decoder attention with learned task-specific strengths. Combining anatomical grounding pretraining with CARA yields our model, CARA-VL. Experiments demonstrate CARA-VL’s strengths in clinical assessment and regional localisation across CMR imaging settings, with promising generalization to an external clinical cohort. Together, our data and method provide a practical framework for studying and advancing cardiac visual understanding in VLMs. We will release the QA data derived from public datasets upon publication.
[CV-33] Spatial-Temporal Multi-scale Network for Screen Content Video Quality Enhancement
链接: https://arxiv.org/abs/2609.39894
作者: Ziyin Huang,Sik-Ho Tsang,Xinyuan Qin,Yui-Lam Chan,Xueling Zhou,Feiyu Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 4 figures
Abstract:Different from natural videos, Screen Content Videos (SCVs) are characterized by abrupt motion, scene switches, and high-frequency details such as text and graphics. Conventional video enhancement methods, which rely heavily on temporal continuity, often suffer from performance degradation when processing SCVs due to the disruption of temporal correlations. To address these challenges, we propose the Spatial-Temporal Multi-scale Network (STM-Net), a novel framework specifically tailored for compressed SCV enhancement. Our approach integrates three complementary components: a Prior-Guided Spatio-Temporal Dispatcher (PG-STD) that routes input into three parallel streams to avoid feature contamination, a Bidirectional Temporal Feature Extraction (BTFE) module that adaptively handles abrupt transitions without explicit detection, and a Cascaded Multi-scale Feature Distillation (CMFD) module that preserves critical high-frequency details. Experimental results demonstrate that STM-Net outperforms state-of-the-art methods in both objective metrics and subjective visual quality, providing a robust solution for screen content artifacts. Code is available at this https URL.
[CV-34] Grounding with Confidence: Controllable Generative Video Temporal Grounding
链接: https://arxiv.org/abs/2609.39883
作者: Jinhao Chen,Benlei Cui,Ruijian Jia,Ziheng Wang,Tianyu Wo,Pengfei Sun,Longtao Huang,Hui Xue,Yitong Yang,Haiwen Hong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 22 pages, 7 figures; includes appendix
Abstract:Video temporal grounding supports applications such as video search, content review, and automated editing by localizing events described in natural language. Yet existing generative models typically output timestamps without explicit interval-level confidence scores to guide candidate selection. We separate candidate generation from acceptance by scoring individual intervals within the original decoding pass. A lightweight confidence head reads pooled decoder states, providing an explicit score trained for interval selection. Offline verifier scores supervise the head on fixed candidate sequences, and temporal-overlap labels adapt it to current rollouts during reinforcement learning. GT-anchored candidate-pool supervision and set-level optimization train the generator. The resulting scores support ranking, threshold-based selection, and rejection without invoking an external verifier at inference. On a fixed OMTG-Bench candidate pool, confidence raises query-macro Recall@0.5 from 9.95% to 14.42% over generation order at a 10% global return budget, and from 26.48% to 31.12% at a 25% budget. The continuous scores let downstream applications adjust return budgets or acceptance thresholds to match their precision-recall preferences, without regenerating candidate intervals.
[CV-35] Hyperspectral Image Models: Technical Report
链接: https://arxiv.org/abs/2609.39871
作者: Tanishq Rachamalla,Aryan Das,Srishti Kaushik,Swalpa Kumar Roy
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Documentation and benchmark library for hyperspectral image models
Abstract:Hyperspectral remote sensing has advanced across diverse deep learning paradigms, including spectral spatial CNNs, Vision Transformers, Mamba, graph neural networks, Kolmogorov Arnold networks, and self supervised masked autoencoding. Yet progress remains hindered by fragmented repositories, incompatible tensor conventions, and non standardized evaluation. Hyperspectral Image Models addresses these challenges through a modular framework unifying 55 representative models across six paradigms with a common registry, automatic 4D/5D tensor adaptation, and standardized constructors. It integrates 24 benchmark scenes from Airborne, Spaceborne, UAV, and Mars CRISM sensors, with caching, label remapping, PCA, explicit band selection or raw spectra, optional spatial max pooling, and arbitrary PxP patch extraction. To prevent inflated accuracy from overlapping windows, it supports class balanced random partitioning and spatially disjoint regional blocking with Chebyshev guard bands that eliminate train test pixel overlap. Experiments use a single this http URL with deterministic seeds and complete provenance, generating LaTeX benchmark tables and classification maps. Across 1,320 model scene evaluations and 6,600 seeded runs, scene difficulty dominates architecture, with mean accuracy ranging from 96.40% on Botswana to 56.70% on Houston 2018, versus a 15 point spread across paradigm means. No paradigm universally dominates, while sub 1 M parameter models can match architectures two orders of magnitude larger. Code is publicly available at this https URL.
[CV-36] DyRAD: Radar Novel View Synthesis for Dynamic Driving Scenes
链接: https://arxiv.org/abs/2609.39841
作者: Merav Keidar,Tomer Borreda,Rajalakshmi Nandakumar,Or Litany
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL . Code: this https URL
Abstract:Reconstructing dynamic driving scenes from recorded sensor data supports closed-loop evaluation of autonomous driving systems by synthesizing observations beyond the original trajectory. Unlike cameras and LiDAR, radar measures radial velocity directly through Doppler. Yet existing radar novel-view synthesis fails to exploit this capability: methods addressing dynamic scenes reconstruct only range-azimuth tensors, while methods that render Doppler assume static scenes. Moreover, because radar processing spreads each reflection across multiple bins, existing representations absorb this spread into scene geometry, causing it to render incorrectly when the viewpoint moves. We present DyRAD, which models dynamic driving scenes using static background reflectors and motion-tracked dynamic point reflectors to render complete range-azimuth-Doppler (RAD) tensors. Reflector velocities are derived from object tracks and projected onto the line of sight, making Doppler both a rendered output and supervision for those tracks. Crucially, we render reflectors through a fixed analytic point-spread function (PSF) derived from the radar’s signal-processing chain, preventing sensor-induced spread from being baked into the scene representation. Beyond improving scene reconstruction, this separation also enables zero-shot sensor-configuration transfer, allowing the same reconstructed scene to be rendered under different radar specifications without refitting. We evaluate DyRAD on RADIal, Boreas, and a synthetic benchmark across both on-path poses and displaced viewpoints untested by prior work. On RADIal, DyRAD recovers radar detections in 90.7% of reference-detected objects, compared with 26.9% for the strongest baseline.
[CV-37] Spherical Interpolation for Backward-Compatible Multimodal Representations NEURIPS2026
链接: https://arxiv.org/abs/2609.39836
作者: Simone Ricci,Niccolò Biondi,Federico Pernici
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026
Abstract:Contrastive vision-language models map visual and textual representations into a shared normalized embedding space, making cosine similarity the natural metric for cross-modal retrieval. A practical challenge arises during model upgrades: independently trained models generally produce incompatible representation spaces, so replacing a deployed model typically requires recomputing embeddings for the entire gallery, which is prohibitively expensive at scale. Orthogonal post-hoc alignment can partially mitigate this problem by mapping new-model queries into the old-model gallery space. However, because independently trained models can differ in fine-grained representation structure, the orthogonal alignment remains approximate, leaving a residual angular discrepancy between the old-model query and the aligned new-model query. We study whether interpolation along the spherical geodesic between these two normalized query representations can improve retrieval without re-indexing the gallery. We characterize when this path contains an interior query direction closer to an idealized retrieval-optimal direction than either endpoint, and connect this characterization to Recall@ K through a local margin-based certification result. Experiments across multiple benchmarks and model families show that post-alignment spherical interpolation improves over orthogonal alignment alone, recovering backward-compatibility in most evaluated settings. Consistent with our geometric characterization, per-query oracle analysis shows that retrieval-favorable interior points occur frequently in practice. Code is available at this https URL .
[CV-38] P-SRM: Selective Recovery of Rejected Predictions in Visual Tracking
链接: https://arxiv.org/abs/2609.39832
作者: Youbin He,Siwei Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 2 figures, 3 tables
Abstract:Many visual tracking methods use rejection mechanisms to suppress unreliable predictions. However, these mechanisms can also reject correctly localized candidates, leaving useful information unused. We investigate how to identify and recover these candidates while preserving native accepted outputs and candidate coordinates. To this end, we propose P-SRM (Post-rejection Selective Recovery Method), which combines spatial responses, past accepted states, and native decision margins to reassess candidates and selectively restore reliable predictions. We evaluate P-SRM on six trackers and four datasets spanning category-specific, point, and generic object tracking. Across all nine configurations, P-SRM improves rejected-candidate ranking and overall tracking performance. These results show that post-rejection verification can identify and recover useful predictions discarded by native rejection, demonstrating the value of reusing rejected information. Project repository: this https URL.
[CV-39] Inline Memory Meets Reusable Skills: Memory-centric Framework for Vision-Language-Action Model
链接: https://arxiv.org/abs/2609.39794
作者: Zaijing Li,Rui Shao,Bing Hu,Haoyu Zhang,Dongmei Jiang,Liqiang Nie
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 24 pages, 8 figures
Abstract:Vision-Language-Action (VLA) models have shown strong promise for general-purpose robotic manipulation, yet adapting them to new tasks and domains remains inefficient: existing methods often rely on parameter tuning, incurring substantial costs and risking catastrophic forgetting of previously learned tasks. To address this, we propose \textbfOptimus-R, a memory-centric VLA framework that formulates robotic adaptation as explicit query-skill memory tuning. Optimus-R introduces: (i) An \textbfInline Memory Interface for skill extraction. It inserts learnable memory tokens into the VLA prefix stream, allowing the backbone to derive control-aware query and skill representations within the native action-conditioning pathway. (ii) A \textbfQuery-Skill Memory Bank for skill learning. It externalizes skills into query prototypes for deciding \emphwhat to retrieve and skill values for specifying \emphhow to act, supporting skill reuse and expansion with limited parameter updates. (iii) A lightweight \textbfBridge-and-Adapt mechanism for skill updating. It aligns target-domain queries and skills with the existing memory space through a lightweight adapter and residual memory updates. Experiments on in-domain adaptation, cross-domain adaptation, and lifelong learning show that Optimus-R enables data-efficient skill learning while mitigating catastrophic forgetting.
[CV-40] Seeing as Humans Do: Learning from Motion to Segment Anything Without Supervision ECCV2026
链接: https://arxiv.org/abs/2609.39785
作者: Weijian Jian,Xiaoyue Zhang,Bin Xiao,Chunyu Xie,Yixiao He,Yutao Liu,Dawei Leng,Yuhui Yin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Published at ECCV 2026. Includes supplementary material. Code: this https URL
Abstract:The Segment Anything Model (SAM) relies heavily on massive manual annotations, creating a fundamental bottleneck for model scaling. While unsupervised methods attempt to learn object concepts from motion, they typically overfit to moving entities, lacking both multi-granularity understanding and the ability to generalize to static objects. To overcome this, we introduce Motion-Grounded Segment Anything (MoSA), a highly scalable unsupervised framework that learns a transferable objectness prior from unlabeled videos. MoSA operates in three progressive stages: (1) automatically generating multi-granularity motion pseudo-labels from large-scale video data; (2) training a Perceptual Grouping Model (PGM) via contrastive learning to internalize a generalized, appearance-driven concept of objects; and (3) transferring this learned prior into a prompt-guided architecture for segment-anything-style inference on images. Extensive zero-shot evaluations across seven challenging benchmarks (e.g., COCO and ADE20K) demonstrate that MoSA significantly outperforms existing unsupervised methods. Notably, despite using zero manual annotations, MoSA achieves segmentation performance comparable to the fully supervised SAM. Our findings reveal that harnessing large-scale unlabeled motion is a feasible and highly scalable alternative to annotation-driven segment-anything pipelines.
[CV-41] Revisiting On-policy Adversarial Black-Box Distillation: Calibrating Groupwise Reward Geometry for Effective Advantage Construction NEURIPS2026
链接: https://arxiv.org/abs/2609.39757
作者: Xiao Cui,Mo Zhu,Yulei Qin,Yuze Wu,Wengang Zhou,Houqiang Li
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: NeurIPS 2026
Abstract:Black-box distillation is a practical route for transferring capabilities from API-accessible large language models that expose only text outputs into smaller student models. Recent on-policy adversarial methods such as GAD improve over SeqKD by forming an adversarial loop between a critic and a student, where the critic provides rewards for GRPO-based student policy optimization over the student’s sampled responses. However, GRPO computes advantages from the within-group relative rewards of student samples for the same prompt, whereas the critic is trained primarily to distinguish teacher responses from student responses. This objective mismatch can produce reward groups with collapsed scale or fragile margins, leading to brittle grouped optimization signals. We propose Groupwise Reward Geometry Conditioning (GRGC), a two-stage framework that improves advantage construction by shaping student-side reward groups during both critic training and policy optimization. To improve critic-side conditioning, Gaussian groupwise Optimal Transport calibration regularizes the critic during training to produce reward groups with non-collapsed spread and smooth rank-wise gaps by matching sorted prompt-wise rewards to group-centered Gaussian quantiles. Building on this conditioned reward geometry, policy-side group power modulation reshapes the prompt-wise reward groups before they are converted into advantages, preserving the critic-induced ordering while increasing optimization-relevant margin separability. Extensive experiments across diverse teachers, student model families and scales, and training datasets demonstrate the effectiveness of GRGC on both in-distribution and out-of-distribution evaluations, while introducing negligible overhead over GAD. The code is available at this https URL.
[CV-42] Determining Vertical Displacement of Agricultural Areas Using UAV-Photogrammetry and a Heteroscedastic Deep Learning Model
链接: https://arxiv.org/abs/2609.39756
作者: Wojciech Gruszczyński,Edyta Puniach,Paweł Ćwiąkała,Wojciech Matwij
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:This article introduces an algorithm that uses a U-Net architecture to determine vertical ground surface displacements from unmanned aerial vehicle (UAV)-photogrammetry point clouds, offering an alternative to traditional ground filtering methods. Unlike con-ventional ground filters that rely on point cloud classification, the proposed approach em-ploys heteroscedastic regression. The U-Net model predicts the conditional expected val-ues of the elevation corrections, aiming to reduce the impact of vegetation on determined ground surface elevations. Concurrently, it estimates the logarithm of the elevation cor-rection variance, allowing for direct quantification of the uncertainty associated with each elevation correction value. The algorithm was evaluated using three metrics: the root mean square error (RMSE) of vertical displacements, the percentage of nodes with deter-mined displacement values, and the percentage of outliers among those values. Perfor-mance was assessed using the technique for order of preference by similarity to ideal so-lution (TOPSIS) method and compared against several ground-filter-based algorithms across four datasets, each including at least two time intervals. In most cases, the U-Net-based approach demonstrated a slight performance advantage over traditional ground filtering techniques. For example, for the U-Net-based algorithm, for one of the test da-tasets, the RMSE of the determined subsidences was 6.1 cm, the percentage of nodes with determined subsidences was 80.5%, and the percentage of outliers was 0.2%. For the same case, the algorithm based on the next best model (SMRF) allowed an RMSE of 7.7 cm to be obtained; for 77.3% of nodes, the subsidences were determined; and the percentage of outliers was 0.3%.
[CV-43] FAST: Flow Any Scene Transformer
链接: https://arxiv.org/abs/2609.39748
作者: Yongjian Zhang,Longguang Wang,Zhuo Song,Zhiheng Fu,Liang Lin,Yulan Guo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Scaling has become a primary driver of progress in language and vision foundation models, yet its role in precise correspondence matching remains underexplored. In this work, we present Flow Any Scene Transformer (FAST), a scalable correspondence model driven by two key insights. First, we reveal that the query-key projections inside single-view vision foundation models encode a coarse yet reusable prior for cross-view matching. Second, reusing these pretrained projections in cross-attention form yields a highly effective initialization for a ViT-based matcher built from a single-view encoder. Guided by these insights, we build FAST upon a vanilla single-view foundation model, utilizing a zero-parameter rewiring strategy to convert selected self-attention layers into cross-attention for cross-view interaction. This design allows ViT-based matchers to scale with advances in single-view foundation models, bypassing the need for a dedicated pair-centric pretraining stage. To fully unlock the scaling potential of this formulation, we assemble a 6-million-pair training corpus for general-purpose dense 2D displacement estimation across diverse co-visible image pairs. Extensive experiments demonstrate that FAST achieves state-of-the-art performance across a wide range of benchmarks, while scaling favorably with both backbone size and training data.
[CV-44] Let the Carrier Carry the Attack: Preserving the Subject in Adversarial Image Generation
链接: https://arxiv.org/abs/2609.39723
作者: Linfeng Jiang,Steven McDonagh,Yuhang Chen,Xingyu Zhao,Siddartha Khastgir,Andi Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Strong unrestricted adversarial attacks can distort the primary object of an image, hereafter referred to as the subject. To preserve subject integrity without compromising attack magnitude, we introduce the carrier: a secondary visual element that provides an auxiliary region to facilitate the attack under global classifier guidance. We demonstrate three key findings: 1. A carrier mitigates subject distortion by absorbing a larger share of globally normalized attack updates. 2. A carrier improves cross-model transferability, governed by the strength of target-related features that balance semantic separation and transfer performance. 3. Successful targeted attacks retain the personalized subject as the primary content perceived by humans while successfully misleading the classifier. Our results demonstrate that a visually secondary carrier offers an auxiliary spatial pathway for adversarial changes, enabling strong and transferable attacks while improving subject preservation.
[CV-45] BTC3D: Blended Tile Conditioning for Detail-Enhancing Image-to-3D Generation
链接: https://arxiv.org/abs/2609.39709
作者: Junyu Li,Qiuyu Chen,Pengcheng Wang,Shiqi Yang,Alexandra Gomez-Villa,Joost van de Weijer,Ruilin Li,Kai Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent diffusion-based pipelines have achieved promising progress in image-to-3D synthesis. However, generating high-fidelity details remains challenging, especially when the input image contains rich details. Existing approaches often rely on globally encoded conditioning features, which compress spatial information and limit the model to reproduce fine-grained details. This common design often leads to a phenomenon we term detail attenuation. Moreover, improving image-to-3D synthesis quality typically requires retraining or fine-tuning large diffusion models, which can be computationally expensive and impractical for complex 3D pipelines. In this work, we present Blended Tile Conditioning for image-to-3D generation (BTC3D), a training-free inference time framework that enhances fine-grained detail preservation in image-to-3D diffusion pipelines. To alleviate detail attenuation, we first examine the image feature additivity in image-to-3D models. Based on this property, we introduce a blended tile embedding that extracts local conditioning signals from split image regional patches, allowing the diffusion model to better preserve fine-grained visual details. To integrate the global and local conditioning guidance stably, we propose a dynamic conditioning schedule that gradually increases the influence of tile-level conditioning during later low-noise stages of diffusion. Our proposed method BTC3D operates entirely at inference time and can be seamlessly integrated into existing image-to-3D diffusion pipelines. Experimental results demonstrate that the proposed approach significantly improves texture quality and visual fidelity of the base model while maintaining global structural consistency in a training-free manner.
[CV-46] When Masking Helps or Hurts Robustness in Compressed CLIP: A Pre-Deployment Diagnostic
链接: https://arxiv.org/abs/2609.39704
作者: Muhammad Zawish,Steven Davy
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:This paper demonstrate that whether masking-based token pruning helps or hurts worst-group robustness can be predicted before deployment, without labels or fine-tuning. A systematic study of semantic masking across 8 spurious-correlation benchmarks shows its effect on worst-group accuracy is highly unstable: it improves accuracy by up to 82.5% relative on some datasets and degrades it by up to 100% on others. We trace this instability to spurious inversion: background patches receive higher CLIP text-similarity than the true object when the spurious attribute is background-separable, inverting the assumption every text- and attention-guided pruning method relies on. We introduce the Spurious Inversion Metric (SIM), a label-free, pre-deployment diagnostic whose sign predicts this effect with statistical significance (binomial p=0.035 ) across all 8 datasets, and remains dependable across 6 CLIP architectures with a clean foreground/background split. Naive masking is itself a major source of risk: it causes the largest average-accuracy loss of any method we evaluate, and its own per-image segmentation step is a significant runtime bottleneck. To address this, we design a batched, synchronization-free GPU segmentation routine that cuts this overhead from 3.5 \times to 1.75 \times baseline. Gating deployment by SIM’s sign recovers masking’s benefits while avoiding its worst failures, matching or exceeding a strong pruning baseline on 7 of 8 datasets.
[CV-47] Unapologetically Distributed: A Call for Decentralized Document Analysis BMVC2026
链接: https://arxiv.org/abs/2609.39684
作者: Adrià Molina,Oriol Ramos Terrades,Josep Lladós
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted at BMVC2026
Abstract:Privacy has become an increasingly important concern in the Document Analysis community, to the extent that in many environments such as archives, governmental institutions, and local businesses, the adoption of automation is restricted by legal and policy constraints. While federated learning has often been regarded as a ``necessary evil’', implying an unavoidable performance trade-off in exchange for decentralization and privacy, many prior works overlook its potential to improve robustness to out-of-distribution data. In this paper, we present Unapologetically Distributed, the first comprehensive study evaluating distributed learning in Document Analysis along three key axes simultaneously: the tasks addressed, the architectures employed, and the fine-tuning strategies applied. Specifically, we demonstrate how various distributed training approaches enhance generalization capabilities across diverse tasks such as Table Recognition, handwriting recognition, and Word Spotting, particularly during transfer learning stages. Our results provide strong evidence that decentralization is not merely a constraint, but a valuable opportunity to improve model robustness and adaptability in real-world Document Analysis scenarios.
[CV-48] MC-PanDA: Simpler Stronger and More Robust Domain-Adaptive Panoptic Segmentation
链接: https://arxiv.org/abs/2609.39681
作者: Ivan Martinović,Josip Šarić,Yuki M. Asano,Siniša Šegvić
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint. Accepted to IJCV
Abstract:Unsupervised domain adaptation (UDA) reduces the annotation burden in panoptic segmentation by leveraging a cost-effectively labeled source domain (e.g., synthetic) and an unlabeled target domain to bridge the distribution gap. Existing panoptic UDA methods rely on teacher-student consistency learning built upon suboptimal per-pixel segmentation architectures. In contrast, state-of-the-art mask transformers are rarely adopted due to their pronounced vulnerability to confirmation bias in consistency learning, where erroneous teacher predictions are reinforced during training. Our earlier approach, MC-PanDA, mitigates this issue through fine-grained confidence estimation, which suppresses gradients from unreliable masks while sampling informative yet reliable locations for loss computation. However, this method entails a complex multi-stage training and requires careful hyperparameter tuning. This work presents MC-PanDA++, which addresses these limitations by introducing: (i) self-supervised vision encoders that provide a stronger and more robust initialization, further reducing the reliance on human annotations, (ii) per-class, self-adapting mask-wide loss scaling that stabilizes training and enables the usage of a single set of hyperparameters across domains, and (iii) a single-stage training pipeline that decreases overall conceptual complexity. Together, these improvements result in a conceptually simpler, better-performing, and more robust method for domain-adaptive panoptics. Source code: this https URL
[CV-49] ypographic Attack Against VLM-based AI-generated Image Detection
链接: https://arxiv.org/abs/2609.39662
作者: Eunmin Lee,Jungwoo Kim,Jong-Seok Lee
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 3 figures
Abstract:Vision-language models (VLMs) are increasingly used for AI-generated image (AIGI) detection, providing natural-language explanations for authenticity judgments. However, their ability to interpret text within images may also expose these judgments to misleading semantic cues. We systematically evaluate typographic attack strategies across detection-oriented, open-weight, and commercial VLMs, considering both real-to-fake and fake-to-real attacks. Our results show that reasoning modes generally exhibit greater vulnerability than direct modes and that attack effectiveness exhibits pronounced directional asymmetry. Moreover, larger models tend to exhibit higher clean detection accuracy but also higher attack success rates. We further examine attack robustness under image and text transformations and investigate whether overlays indicating the correct class can aid error correction. Together, these analyses characterize how typographic attacks influence authenticity judgments and expose limitations of current VLM-based AIGI detection systems.
[CV-50] BAM! Bayesian Anything Model: a foundation model for generative computational imaging
链接: https://arxiv.org/abs/2609.39660
作者: Alessio Spagnoletti,Charlesquin Kemajou Mbakam,Jonathan Spence,Andrés Almansa,Marcelo Pereyra
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (stat.ML)
备注: 37 pages, 25 figures
Abstract:Generative models are transforming Bayesian computational imaging, yet the field still lacks physics-aware foundation models. Current practice falls into two camps. Large foundation image models are deployed as plug-and-play priors with zero-shot approximate likelihood guidance, which introduces significant bias and computational cost. Physics-aware generative models avoid this bias, but each is tied to a specific dataset, task and instrument. We introduce BAM (Bayesian Anything Model), a lightweight foundation model for few-step, physics-aware posterior sampling that generalises robustly to unseen data and tasks, zero-shot or with minimal finetuning. BAM upgrades the operator-conditioned Reconstruct Anything Model (RAM) backbone (Terris et al.) into a conditional flow map, so instrument physics is specified at inference time rather than fixed during training. BAM has just 36M parameters and is pre-trained jointly on large image corpora and libraries of forward operators. A single network then draws posterior samples in a few steps, with no likelihood approximation and no guidance weights to tune. Across linear inverse problems on FFHQ, AFHQ, LSUN, DIV2K and the Kohler camera-shake benchmark, BAM outperforms in just 3 steps both specialised models and leading zero-shot methods in sample quality, at a fraction of their computational cost. BAM gives the community an accessible entry point to generative computational imaging, lowers the economic and environmental cost of training imaging models, and opens a new path for research on physics-aware Bayesian computational imaging. Official page: this https URL
[CV-51] Diffusable Latents from Structure-Agnostic Distillation NEURIPS2026
链接: https://arxiv.org/abs/2609.39657
作者: Adrien Ramanana Rahary,Nicolas Dufour,Patrick Pérez,David Picard
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: NeurIPS 2026 Workshop on Principles of Generative Modeling
Abstract:Distilling pretrained foundation models into an autoencoder bottleneck improves latent diffusability, enabling diffusion models to converge faster and reach higher sample quality. Standard distillation aligns the latent at each position to a co-located teacher feature, tying the latent layout to the teacher’s. We show this constraint is unnecessary: aligning a single pooled image-level descriptor to the teacher’s performs as well as or slightly better than dense position-wise distillation. We compare first-order and relational pooled objectives across latent shapes and teacher modalities. First-order matching extends naturally to 1D token-sequence latents and across modalities, where distilling a text encoder into an image autoencoder still improves diffusability; a relational objective based only on each image’s nearest neighbours improves it as well. Code and blog post are available at this https URL and this https URL.
[CV-52] FANVIDv2: Evaluating Video Super-Resolution by Face and Licence-Plate Recognition Under Compound Degradation
链接: https://arxiv.org/abs/2609.39649
作者: Kavitha Viswanathan,Vrinda Goel,Shlesh Gholap,Devayan Ghosh,Madhav Gupta,Dhruvi Ganatra,Sanket Potdar,Amit Sethi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video super-resolution (VSR) is normally judged by PSNR and SSIM on clips that were downsampled bicubically, although in surveillance its purpose is to make faces and licence plates \emphrecognisable. We present FANVIDv2, a benchmark that scores VSR by what a recognition pipeline can do with its output. FANVIDv2 provides 320\times180 low-resolution (LR) clips with high-resolution (HR) references for 48 public figures (with one HR gallery image each) and 375 licence-plate clips covering 360 distinct plate strings. LR clips are generated with a randomised compound degradation (blur, resize jitter, sensor noise, JPEG compression, final downsampling) rather than bicubic downsampling alone. Two metrics score recognition \emphinside detections: FaceRecBox rewards a face only if it is localised and correctly identified, and TextRecBox scores plate transcriptions by normalised edit distance weighted by localisation quality. With a 2.3,M-parameter VSR baseline (RCDM), FaceRecBox rises from 0.6864 to 0.7222, identity accuracy on matched faces from 84.35% to 86.93%, and TextRecBox from 0.3088 to 0.3667; a residual-map gated variant (RCDM-RMGF) reaches 0.3801 on plates. We describe the degradation model, the baseline architectures and the scorers in detail, and release annotations, metadata, download and degradation scripts and evaluation code.
[CV-53] From Modes to Memories: Characterizing the Scale-Space Dynamics of Diffusion Models
链接: https://arxiv.org/abs/2609.39648
作者: Cristina López Amado,Marco Fumero,Francesco Locatello
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Diffusion models are typically viewed as stochastic processes that transform noise into data. We take a complementary perspective: a diffusion model defines a family of deterministic dynamical systems indexed by noise scale. At each fixed scale \sigma , we treat the denoiser as a self-map and study its dynamics. For an exact denoiser, fixed points correspond to critical points of the smoothed data density, while attractors correspond to its modes; as \sigma increases, sample-level modes merge into progressively coarser ones. This suggests a geometric view of memorization: examples that receive excess probability mass due to duplication or overfitting, as well as outliers, should remain distinguishable under stronger smoothing than ordinary examples. We quantify this persistence by the critical scale \sigma_c , the largest noise scale at which an example is retained by the fixed-scale dynamics. In conditional models, the same construction extends naturally to image–caption pairs. Experiments in controlled settings and on large-scale models show that \sigma_c tracks memorization arising from duplication, overfitting, and outliers, and identifies both memorized and partially memorized examples in Stable Diffusion. Moreover, \sigma_c yields interpretable measures of the image spatial distribution and caption dependence of memorization.
[CV-54] SAGE: Salient Factor Discovery and Generation with Visual Foundation Representations
链接: https://arxiv.org/abs/2609.39635
作者: Shuang Liang,Lejun Liao,Shiyuan Zhang,Max C. Zhang,Xiaolong Luo,Han Wang,Stefano Anzellotti,Yuan Yuan
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 28 pages, 18 figures, 9 tables
Abstract:Given a target dataset, such as faces with eyeglasses, and a background dataset, such as faces without, contrastive analysis separates \textitsalient factors specific to the target from \textitcommon content shared by both. We aim for salient representations that capture target-specific detail in each image, such as the shape, color, and position of the glasses, so that they reveal subtypes without subtype labels and guide the generation of new examples of a discovered subtype, even one with no name or text description. We introduce SAGE, which learns both factors directly in the high-dimensional spatial latent of a frozen representation autoencoder and conditions a diffusion transformer on the learned salient representation of a reference image. On Digits-ImageNet and FFHQ eyeglasses, SAGE combines high-fidelity \textitreconstruction (rFID below 2 ) with unsupervised \textitsubtype discovery, recovering the digits better than baselines (probe accuracy 0.950 vs.\ at most 0.281 ) and revealing eyewear types, finer sunglasses styles, and mislabeled images; salient-conditioned \textitgeneration raises Digits-ImageNet subtype accuracy over the unfactorized latent ( 90.5% vs.\ 27.7% ) and diversity on both datasets. On retinal OCT, SAGE’s salient space separates three diseases using only normal/disease labels.
[CV-55] Introduction to Computer Vision
链接: https://arxiv.org/abs/2609.39627
作者: Stan Birchfield
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 217 pages. For online notes and code, see this https URL
Abstract:This book presents a code-first introduction to computer vision, spanning classical 2D image processing, classical 3D vision, and deep learning. Organized as 44 short chapters across three parts, the book builds each topic from first principles: image arithmetic and morphology; convolution, pyramids, and frequency-domain filtering; feature detection, optical flow, and stereo; projective geometry, camera calibration, and structure from motion; and the full arc of modern deep learning, from a single neuron through convolutional networks, backpropagation, classic architectures, transfer learning, object detection, and semantic and instance segmentation, concluding with engineering considerations like mixed-precision and parallel training. Every technique is implemented directly in Python and NumPy or PyTorch and checked numerically against the corresponding OpenCV or PyTorch library function, so readers see not just the mathematics but its concrete behavior on real and synthetic data. The material was distilled with AI assistance from freely available online course notes, condensing extensive working code into concise mathematical exposition while preserving verified, reproducible results throughout. It is intended as a self-contained reference for students and practitioners who want to understand computer vision algorithms and their Python implementations.
[CV-56] D-Scope: Decomposing and Steering Diffusion Transformers with Sparse Autoencoders
链接: https://arxiv.org/abs/2609.39625
作者: Xinyue Xu,Jiahao Zhang,Lijie Hu,Peter Hase,Hao Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Sparse autoencoders (SAEs) reveal visual structure in diffusion transformers (DiTs), but interpreting a feature does not establish whether it can be used to control generation. We introduce D-Scope (Diffusion Scope), a framework that connects feature interpretation to generation control through shared visual evidence. D-Scope aggregates SigLIP~2 embeddings of highly activating image patches into visual centroids. Matching target text descriptions against these visual centroids in the shared image-text embedding space then enables retrieval of individual features without per-feature text annotations. The underlying patches provide evidence for inspecting each selection, while spatially masked interventions test the corresponding decoder direction at varying strengths under fixed generation conditions. We characterize 150 SAEs across two model families and five layers, and introduce a benchmark of 100 target concepts with ten contexts each spanning under-specified and explicit-conflict conditions. Our empirical results show that high reconstruction fidelity can coexist with low dictionary utilization and limited visual-evidence coverage. Under per-case best-of-sweep strength selection, contrastive retrieval yields larger mean regional SigLIP~2 gains than direct retrieval across the tested steering configurations, without consistently improving outside-region preservation. D-Scope provides an inspectable framework for evaluating sparse DiT features through their visual evidence and the effects of their decoder directions on generation. The demo is available at this https URL.
[CV-57] ExpandDiff: Dynamic Range Expanding Diffusion for Single-Image HDR Reconstruction ICASSP2027
链接: https://arxiv.org/abs/2609.39624
作者: Mehmet Emre andıran,Zhuoqian Yang,Liying Lu,Mathieu Salzmann,Sabine Süsstrunk
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注: 5 pages, 3 figures, 2 tables. Submitted to ICASSP 2027. Code and supplementary material: this https URL
Abstract:Single-image HDR reconstruction requires inferring missing detail while preserving the visible content of an LDR image. Differences in sensor dynamic range and exposure cause LDR images to lose varying amounts of information in shadows and highlights. We present ExpandDiff, a conditional diffusion pipeline that jointly reconstructs clipped shadows and highlights. To account for this variation, we introduce Dynamic Clipping Synthesis (DCS), which randomly samples shadow and highlight clipping percentiles when constructing training inputs from HDR targets. A pixel-space diffusion model guided by spatially-adaptive normalization then predicts perceptually encoded HDR through a bounded output head, reconstructing both clipping directions in one sampling trajectory. On the SI-HDR benchmark, ExpandDiff variants improve HDR reconstruction accuracy by 3.43 dB in PU21-PSNR over the strongest evaluated competing method, and by 7.34 dB under two-sided clipping. The code and supplementary material are available at this https URL.
[CV-58] Semantic Watermarking for Malicious Image Manipulation Detection
链接: https://arxiv.org/abs/2609.39623
作者: Yoonseo Kim,Seungwoo Baek,Junyoung Park
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The proliferation of high-fidelity generative editing models has made it possible to inject violent or sexual content into otherwise ordinary images while preserving visual plausibility, with concrete consequences for public discourse and vulnerable populations. We propose a robust semantic watermarking framework that reframes the watermark as a recoverable semantic reference rather than an opaque identifier. Our framework combines a \beta -VAE-based binary watermark (CLIP-VAE) with explicit channel-aware training—random bit-flip noise is injected during training so that the decoder learns graceful degradation under the noisy watermarking channel. As a downstream application, a lightweight module SDA-Net uses the recovered semantic embedding to expose not only whether but in which semantic direction an image has been altered. In a 5-way comparison against representative binary hashing baselines (SimHash, ITQ, HashNet, and their robust-MLP variants), CLIP-VAE achieves the highest reconstruction cosine similarity to the original CLIP embedding under realistic InstructPix2Pix bit-error rates, and uniquely supports direction-of-drift detection—a forensic complement to existing content-moderation pipelines.
[CV-59] FOMO: Forget the Concept Dont Miss Out on the Scene in Selective Video Unlearning
链接: https://arxiv.org/abs/2609.39605
作者: Łukasz Rudnik,Agnieszka Polowczyk,Alicja Polowczyk,Przemysław Spurek
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The rapid advancement of generative video models has enabled the synthesis of increasingly realistic and temporally coherent videos, while also raising concerns about the generation of harmful content. The reliance on large-scale web datasets during training inevitably exposes these models to undesirable material, making concept unlearning an essential mitigation. Existing methods mainly target static visual concepts, such as objects, identities, or unsafe appearance, largely overlooking motion unlearning. Furthermore, these approaches often pay little attention to preserving the surrounding scene. As a result, successful concept removal may unintentionally alter the background, composition, or overall video dynamics. We argue that effective unlearning should ideally change only what is targeted, while minimizing unnecessary changes to the remaining scene. In this work, we introduce FOMO, to the best of our knowledge the first training-based selective video unlearning method that directly treats preservation of the original scene as a priority. We formulate unlearning around two complementary objectives: what to change and what to preserve. Our method localizes concept-related representations and modifies them, while the preservation mechanism maintains non-target scene information without requiring auxiliary data. Beyond simply erasing unwanted concepts, FOMO explicitly redirects the generation toward a specified safe alternative. We further extend this formulation to motion unlearning, where the concept is defined by temporal behavior rather than a fixed spatial region. Our solution achieves effective unlearning across unsafe content, object, and motion concepts, while achieving the best trade-off between concept removal and scene preservation. Code: this https URL Project Page this https URL Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2609.39605 [cs.CV] (or arXiv:2609.39605v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2609.39605 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-60] GroundingPI: A Grounding Foundation Model towards Physical Intelligence with Visual Primitives
链接: https://arxiv.org/abs/2609.39601
作者: Qize Yu,Lianrui Fan,Boyu Chen,Jiaqi Liang,Xini Ding,Yue Chen,Zetian Song,Yuran Wang,Yi Zou,Kaixuan Wang,Tianxing Chen,Wenxuan Song,Bohan Zhou,Mingleyang Li,Siqiao Huang,Yuqi Ye,Caigao Jiang,Wei Wei,Ruihai Wu,Hang Zhang,Yixiao Ge,Shuchang Zhou,Shilong Liu,Xianming Liu,Ping Luo,Shiyu Huang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Robotics (cs.RO)
备注: 64 pages, including supplementary material. Project page: this https URL Code: this https URL Model: this https URL
Abstract:Precise grounding matters. It specifies which object is the target and where that object is, even in clutter and for tiny objects, and it has to be fast enough for closed-loop control. Yet vision-language-action (VLA) and world-action models (WAMs) take perception from general-purpose vision-language and video-generation backbones, which still fail in these settings. We introduce GroundingPI, a 4B grounding foundation model that generates points and boxes as quantized coordinates in a shared vocabulary. Training combines multimodal and spatial pretraining, supervised fine-tuning, and reinforcement learning with GRPO, using supervision from public datasets and dedicated data engines. Against 44 baselines across 34 grounding benchmarks spanning 11 perceptual capabilities, GroundingPI establishes a new state of the art, averaging 73.68%, above the larger GPT-6 Astra (71.54%). As a downstream visual backbone, GroundingPI improves performance on robotic manipulation and autonomous driving. On RoboTwin 2.0, it outperforms every mainstream backbone we evaluate in all four out-of-distribution settings, by up to 24.8% relative to the strongest backbone. On RoboCasa-GR1, GroundingPI trained with 50% of the demonstrations outperforms those baselines trained with 75%. On nuScenes, used as the visual backbone, GroundingPI attains an average open-loop L2 error of 0.296 m. We systematically analyze GroundingPI’s pretraining in scale and data composition. Downstream autonomous driving and robotic manipulation improve as the pretraining is scaled. Analyzing the data recipe across these 11 perceptual capabilities shows dense grounding’s substantial benefits for both, and OCR’s potential as a catalyst for perceptual learning. These results support grounding as a perceptual foundation, and dedicated perceptual pretraining as a promising direction for foundation models of physical intelligence.
[CV-61] GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed
链接: https://arxiv.org/abs/2609.39600
作者: Qize Yu,Lianrui Fan,Bowen Ping,Xini Ding,Zetian Song,Junbo Niu,Kaixuan Wang,Tianxing Chen,Yue Chen,Minghua He,Yuran Wang,Jie Huang,Haojun Zhang,Min Chen,Hao Li,Wenxuan Song,Ruihai Wu,Xianming Liu,Shilong Liu,Shuchang Zhou,Ping Luo,Shiyu Huang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Robotics (cs.RO)
备注: 61 pages, including supplementary material. Project page: this https URL Code: [ this https URL ]( this https URL ) Model: this https URL , this https URL
Abstract:Autoregressive (AR) grounding models serialize spatial predictions, introducing sequential latency and imposing a causal order on output tokens. We view grounding as visual evidence extraction: objects, locations, and spatial relations are jointly constrained by the image and query, yet their dependencies do not imply an intrinsic left-to-right generation order. This distinction makes bidirectional diffusion a natural fit, allowing spatial hypotheses to emerge in parallel and be jointly refined through iterative denoising. We introduce GroundAnything, a 4B-parameter grounding foundation model that reconciles fast parallel decoding with precise localization through blockwise denoising. Training combines grounding pretraining from public datasets and dedicated data engines, direct AR-to-diffusion conversion with joint AR and diffusion objectives, supervised fine-tuning, and GRPO-based reinforcement post-training. Across 30 grounding benchmarks, our autoregressive variant, GroundAnything-VLM, establishes a new overall state of the art among similarly sized models at 72.42%, remaining competitive with GPT-6 Astra (71.35%). With entropy-guided decoding, GroundAnything also surpasses the prior state of the art at this scale, averaging 61.75% versus 53.32% for the fast MTP-based LocateAnything model. We further explore decoding strategies, showing that an optional self-speculative mode achieves a 4.51\times speedup over the AR counterpart with a 0.74 percentage-point drop in COCO F1mIoU. Infrastructure experiments show that progressive inference optimizations translate parallel decoding into practical speedups. These support efficient visual grounding in latency-sensitive real-world systems.
[CV-62] Structural Limits of the Information-Theoretic Uncertainty Decomposition
链接: https://arxiv.org/abs/2609.39591
作者: Jakob Lønborg Christensen,Christian F. Baumgartner,Morten Rieger Hannemose,Anders Bjorholm Dahl,Vedrana Andersen Dahl
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Uncertainty estimation in machine learning typically decomposes uncertainty into aleatoric uncertainty (AU) and epistemic uncertainty (EU) using the standard information-theoretic framework. However, in practice, two critical issues arise: entanglement (AU and EU are highly correlated) and epistemic collapse (EU magnitude shrinks with increasing model capacity). We analyze this framework on a functional level and discover that significant portions of the assumed AU, EU range are infeasible in finite settings, and cannot be attained with any class probabilities. We characterize how this infeasible region scales with the number of classes and Monte Carlo samples N (e.g., from ensembles with N members), revealing it is bounded by \textAU \leq \log(2)/N . Crucially, the infeasible region’s boundary helps explain epistemic collapse: when model confidence is high, \textAU \textEU is guaranteed by this fundamental structural limitation. Our findings show that increasing ensemble size mitigates epistemic collapse by reducing the infeasible area. Lastly, we caution against interpreting AU and EU as independent quantities in low AU regimes, since we show they are coupled when \textAU \leq \log(2)/N .
[CV-63] SPOON: Towards Coherent Compositional 3D Scene Generation from Uncalibrated Multi-view Images
链接: https://arxiv.org/abs/2609.39590
作者: Guibiao Liao,Mochu Xiang,Heng Li,Ken Deng,Zijie Wang,Guanbin Li,Ping Tan,Shenghua Gao,Yizhou Yu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Compositional 3D scene generation aims to recover complete 3D object shapes and their spatial arrangement from visual observations. Recent image-conditioned 3D generators provide strong priors for producing high-quality object geometry, making the generation of complex scenes increasingly practical. A central challenge is therefore to spatially organize these generated assets into a globally coherent scene while remaining consistent with multi-view observations. Existing approaches either entangle scene layout with object generation or separately estimate spatial placement from view-specific observations, where pose hypotheses may remain ambiguous and inconsistent across views, often resulting in an incoherent object-camera soup. We introduce SPOON, a framework that reformulates multi-view compositional 3D generation as scene-level, geometry-grounded pose reasoning. Rather than treating view-specific object pose hypotheses independently, SPOON coordinates them using reconstruction-derived multi-view geometry through a Guide-Route-Reconcile paradigm. This progressively organizes object poses and camera configurations into a coherent scene-level spatial arrangement. Extensive experiments on ARSG-110K and MIDI-3D-Front demonstrate consistent improvements in object placement and scene composition across varying numbers of input views. On ARSG-110K, SPOON reduces scene-level and object-level Chamfer distances by 12.7% and 17.7%, respectively, compared with a strong baseline.
[CV-64] KilometerVision: A New Frontier for Large-Scale Spatial Intelligence in VLMs
链接: https://arxiv.org/abs/2609.39588
作者: Aravindh Mahendran,Michael King,Matthew Koichi Grimes,Antoine Yang,Tyler Zhu,Joseph Heyward,Tengda Han,Shiry Ginosar,Chen Sun,Dima Damen,Simon Osindero,Noah Snavely,Simon Lynen,João Carreira,Viorica Pătrăucean
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:We push the frontier of large-scale spatial intelligence in Vision-Language Models (VLMs) and introduce the first benchmark that probes geographical layout understanding from real-world videos, spanning up to 1km distances. Inspired by the cognitive science literature, we evaluate models against the hierarchical stages of human spatial awareness: anchoring via landmarks, connecting them through routes, and integrating these into global mental maps. Extensive experiments reveal a fundamental divergence in how current AI models process spatial information. Instead of utilising true path integration or forming geometric survey knowledge, we find that VLMs rely almost entirely on 2D visual recognition and text-matching to bypass complex spatial reasoning. The benchmark is publicly available at this https URL.
[CV-65] A Generalizable and Explainable Framework for Synthetic Video Detection Using First-Digit Gradient Statistics
链接: https://arxiv.org/abs/2609.39585
作者: Sidharth Shanu,Gautam Kumar,Tej Singh
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 Pages
Abstract:AI video generators have not only become harder to detect but are used to generate a diverse set of scenarios from landscapes to street views to animal videos. This creates a problem where CNN-based detectors are effective but offer no insight into their inner workings, while forensics-based detectors are often pretrained for a set scenario or become too complex to derive meaningful insights. We present a novel approach to AI video detection using Sobel gradient values analysed with the first-digit law. Using linear discriminant analysis, we visualise the discriminatory signal, while a multi-layer perceptron is used for classification. The detection method has no generator- or scenespecific features, and the model has no knowledge of container formats, codec, bitrate, or compression artefacts. The model is trained and tested on GenBuster-200K, GenBusterBench, GenVA, FaceForensics++ C23, and CelebDF. We also show how zero-shot detection fails even though the feature set carries a discriminatory signal.
[CV-66] Steering Fields: Adaptive Vector Fields for Safe Image Generation and Beyond
链接: https://arxiv.org/abs/2609.39573
作者: Simone Facchiano,Jan Eric Lenssen,Bernt Schiele,Wolfgang Stammer,Fabio Galasso,Jonas Fischer
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:As state-of-the-art text-to-image flow models achieve near-photorealistic quality, controlling their outputs, e.g., suppressing harmful content while promoting benign alternatives, has become a central challenge. The current steering paradigm consists of adding a global steering vector to selected activations. While functional, a fixed and example-agnostic vector applied uniformly along the entire trajectory cannot adapt to the changing state of the generation and often causes unintended global changes. We introduce Steering Fields, a generalization of steering vectors that adaptively re-estimates the steering direction at each step of the generative process. Steering Fields operate on the noisy states of flow models, expose a continuous trade-off between steering strength and content preservation, and are compositional, enabling the simultaneous induction and inhibition of concepts, setting a new state of the art on safety steering benchmarks. Despite using no explicit spatial masks or object priors, the trajectory-adaptive estimation naturally preserves local structure, in a manner reminiscent of image editing. In fact, Steering Fields can serve as a structure-preserving image-editing technique that achieves state-of-the-art semantic fidelity (CLIP, VQAScore), while remaining model-agnostic and inversion-free.
[CV-67] Invariant Shape Analysis of Surfaces with Spherical Topology
链接: https://arxiv.org/abs/2609.39567
作者: T. Shaska,M.-R. Siadat
类目: Computer Vision and Pattern Recognition (cs.CV); Algebraic Geometry (math.AG)
备注:
Abstract:Spherical harmonic descriptors of closed 3D shapes depend on the parameterization, the pose and the scale of the surface, and the standard rotation-invariant reductions, the power spectrum and the bispectrum, discard the relative orientation of the harmonic bands and cannot distinguish a shape from its mirror image. We construct a descriptor that removes all three dependencies exactly and loses nothing else: a conformal parameterization normalized by its conformal barycenter, followed by polynomial invariants of the rotation group. Identifying each harmonic band with a binary form turns the rotation quotient into classical invariant theory and makes reflections visible as the sign of an invariant, so chirality is recorded. The descriptor is complete for the truncated expansion, stable in the orbit distance, and comes with numerical diagnostics. Benchmarks confirm the guarantees, and on bilateral anatomical structures the descriptor separates mirror-image pairs from asymmetric pairs, which parity-blind descriptors cannot.
[CV-68] From Given to Gathered Evidence: Agent ic Learning for Longitudinal Medical Reasoning
链接: https://arxiv.org/abs/2609.39566
作者: Minye Shao,Chaohui Yu,Yixuan Wu,Fan Wang,Ling Shao,Yang Long
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Foundation models can serve as clinical agents through tool-use harnesses. However, conventional medical benchmarks assess reasoning over preselected evidence rather than the ability to seek it across clinical records and longitudinal imaging. We propose CASE: a series of role-specific Clinical Agents for Seeking Evidence, together with a tool-use harness and an agentic post-training framework for compact vision-language policy models. We further introduce a longitudinal multimodal benchmark built on UK Biobank, comprising 50,401 clinical questions derived from real-world ICD-10-coded diagnoses of 4,739 participants. Each question links to a patient-specific environment containing clinical context and multi-sequence MRI from baseline and follow-up visits, where agents autonomously select which visits, organs, modalities, slices, and specialist tools to inspect and compare. Supervised fine-tuning transfers evidence-seeking workflows from 14,734 frontier-model interaction trajectories, followed by agentic reinforcement learning on the learner’s own environment interactions. Privileged on-policy self-distillation and rubric-based LLM feedback refine evidence-to-conclusion reasoning without prescribing tool sequences. Experiments show that CASE moves beyond question-answer imitation toward transferable investigation policies, strengthening evidence-grounded longitudinal reasoning. Under matched evaluation conditions, our Qwen3-VL-8B based agent achieves over 16% and 10% relative improvements in answer accuracy over GPT-5.4 and Claude Opus 4.8. Code will be available at this https URL.
[CV-69] RESUME: Recurrent State Updates from Motion and Residual Signals for Efficient Video Language Modeling
链接: https://arxiv.org/abs/2609.39563
作者: Can Zhang,Xiaotian Han,Junyuan Shang,Yuchen Ding,Zhenyu Zhang,Shuohuan Wang,Dianhai Yu,Ruirui Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Existing video language models encode sampled RGB frames independently, so a long video must either exhaust the token budget or drop the changes between sampled frames. Codec-aware front-ends read the motion vectors and residuals that encoding produced, but in their deployed form each predictive frame is still tokenized on its own: the tokens are a function of the current primitives, not of a carried reference. We argue that a more natural function is of both—the current primitives and a carried reference. A clip and its time reversal share the same frames and differ only in the order of changes—an axis that symmetric pooling discards by construction, and that is non-empty in the frozen vision features VideoLMs use—and the codec recurrence already composes those changes in order against a reference state. We introduce RESUME, a stateful codec representation: an anchor I-frame initializes a compact latent state, each subsequent predictive frame is consumed as an update to that state, and a shared readout exposes VideoLM-compatible tokens from the accumulated state. Codec prediction is thereby kept at the representation level and handed to the language model as a trajectory, not as a set of independent token groups. At the same per-predictive-frame token budget as prior codec-aware methods, a predictive frame enters the language model as a readout of what the front-end already knows, not as an encoding of the current primitives alone. Across ten benchmarks, the gains concentrate on temporal reasoning: on all three temporal benchmarks RESUME improves over both the RGB-frame baseline LLaVA-Video-7B (by 2.8, 5.1, and 3.9 points on TempCompass, TOMATO, and MVBench) and the codec-based baseline CoPE-7B, while staying competitive on general and long-form QA. Frozen-transition tests further show anchor dependence, order sensitivity, and useful rollout behavior beyond the training horizon.
[CV-70] EffGS: Efficient and High-Fidelity Gaussian Splatting
链接: https://arxiv.org/abs/2609.39553
作者: Changbai Li,Shuo Yang,Yichen Yang,Shuwei Shao,Huobin Tan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:3D Gaussian Splatting (3DGS) enables real-time novel view synthesis, but existing general-purpose acceleration methods suffer severe rendering quality degradation when extended to more complex, large-scale scenes. To address this issue, we propose EffGS, a more general acceleration framework that improves training and rendering efficiency while maintaining reconstruction quality comparable to or better than vanilla 3DGS across bounded and large-scale scenes. EffGS combines frequency-aware guidance, localized density control, and adaptive primitive scale modulation. First, an importance scoring mechanism combines pixel-wise reconstruction errors with a difference-of-Gaussians mask scheduled over training to provide stage-dependent spatial guidance. Second, localized densification and pruning restricts density modifications to Gaussians with valid projected footprints in the sampled views. Third, learnable per-Gaussian scale modulation adjusts effective primitive extent during optimization while retaining the Compact Box rasterization rule. Extensive experiments on bounded and large-scale scene datasets demonstrate a favorable balance between reconstruction quality, training time, and primitive count. Component ablations and matched-primitive-budget comparisons further support the effectiveness of the framework.
[CV-71] Learning Normal Diffusion Dynamics for Backdoor Defense in Text-to-Image Models
链接: https://arxiv.org/abs/2609.39548
作者: Junjian Li,Xiaolong Liu,Peng Sun,Liantao Wu,Linghan Chen,Yudong Gao,Honglong Chen
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:Backdoor attacks pose a serious threat to the secure deployment of text-to-image (T2I) diffusion models. Existing defenses typically detect backdoors from specific abnormal patterns in internal representations, which may limit their generalizability with the emergence of increasingly diverse attack mechanisms. In this paper, we study backdoor defense of T2I diffusion models from a transition-dynamics perspective. We observe that benign diffusion trajectories exhibit structured and timestep-dependent transition patterns from cross-attention, latent and noise spaces, whereas backdoor attacks tend to induce deviations from such normal evolution. Motivated by these observations, we propose Normal Diffusion Dynamics Learning (NDDL), a novel backdoor defense framework that learns the normal transition dynamics of diffusion trajectories utilizing only benign samples. NDDL constructs compact multi-space trajectory representations and trains a timestep-conditioned dynamics model to predict the diffusion evolution. In the inference phase, deviations between the observed and predicted transitions are exploited to quantify dynamics inconsistency for backdoor detection. NDDL further enables trigger localization without any prior knowledge of the embedded backdoor by performing substitution with low-semantic words. Extensive experiments for diverse backdoor attacks demonstrate the effectiveness and generalizability of our proposed NDDL.
[CV-72] Comparative study of adapting pre-trained models for driving behavior video captioning
链接: https://arxiv.org/abs/2609.39542
作者: Sayak Mallick,Philipp Geiger,Augustin Kelava
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:This report examines and compares some of the many fine tuning and prompting methods existing, applying them within the domain of autonomous driving. The idea is to compare these methods by adapting a Large Language Model (LLM) on a video dataset. LLM’s have become extremely good at achieving a good understanding of different forms of data and this study aims to induce a low dimensional understanding of driving situations into our primary test model SpaceTimeGPT. Experiments on BDD-X (Berkeley DeepDrive eXplanation) dataset demonstrate good performance of the full fine tuning framework on some automatic metrics, and in some metrics, it even surpasses the baseline. We also try Low-Rank Adaptation (LoRA) and prompt engineering on VideoLLaVA model and discuss its limitations.
[CV-73] Lens Flare Removal and Reconstruction
链接: https://arxiv.org/abs/2609.39527
作者: Tarun Yenamandra,Jonathon Luiten,Daniel Cremers,Nathan Matsuda
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 20 pages, 14 figures. Project page: this https URL
Abstract:The presence of lens flares in images can significantly reduce the quality of downstream application results for tasks such as 3D scene reconstruction. This is because lens flares are a property of the camera imaging system, and not a part of the underlying scene being modeled. There are previous methods that tackle the removal of small flares focused around a light source. However, existing methods struggle with large flares, such as those that fill the entire image. In this work, we compile a novel dataset for large-flare removal, combining publicly available real-world data with a procedural generation pipeline. We fine-tune a diffusion-based model on our dataset to remove complex, large lens flares. On the other hand, lens flares remain effective artistic tools, widely used in the media. While there are ways to simulate 2D flares, representing and reconstructing lens flares consistently across multiple views has not yet been explored. To achieve this, we introduce a flare representation model that leverages the symmetry of lens flares about the camera’s principal point. We propose a computational pipeline to jointly optimize this flare model and a Gaussian splatting model (3DGS). This enables the decomposition of a 3D scene into lens flares and the scene itself, using our flare-removal model. Because the reconstructed flare is explicit and re-renderable, it can be edited and transferred to novel images and new 3D scenes. We evaluate removal on an established benchmark and a new one for large reflective flares, quantify the flare/scene decomposition directly, and show that the pipeline is robust to errors in automatic light-source localization.
[CV-74] PartiCam: Camera Controlled Video Generation with Reward Guidance
链接: https://arxiv.org/abs/2609.39504
作者: Amine Ouasfi,Runjia Li,Junlin Han,Eric Marchand,Philip H.S. Torr,Adnane Boukhayma
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:We present PartiCam, a training-free Particle filtering rooted method for improved Camera controlled video generation. Generating videos that follow a precisely specified camera trajectory remains challenging for large video diffusion models. Training-free approaches are backbone-agnostic and avoid the need to construct large camera-annotated datasets by steering pretrained models toward the desired camera motion at test time. This enables the generation of camera-controlled video data that can subsequently be used to train camera-conditioned video diffusion models. Existing sampling-based guidance approaches often suffer from unstable trajectories: they either explore too broadly and fail to respect the target camera motion or collapse early and lose visual diversity over time. We introduce a global-local refinement framework for diffusion reward guidance, enabling accurate and consistent camera control during video generation. Our method builds on Sequential Monte-Carlo (SMC) guidance, but introduces a local refinement stage based on particle filtered resampling. Experiments show large improvements in camera trajectory adherence, reduced drift, and better visual quality, without requiring model retraining.
[CV-75] Front-to-Back: Benchmarking Vision-Language Models for Asymmetric Cross-View Vehicle Re-Identification ACCV2026
链接: https://arxiv.org/abs/2609.39492
作者: Moseli Mots’oehli,Thulani Babeli
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Submitted to the ACCV 2026 Workshop on Computer Vision for Developing Countries (CV4DC)
Abstract:Matching the same vehicle across front and rear cameras is difficult because the cameras do not share a view and the vehicle’s appearance changes substantially. We introduce Front2Back-ReID, a benchmark of 500 manually verified vehicle handovers from 20 recording sequences in South Africa. Each example asks a model to match a vehicle highlighted in a front-camera image to the same vehicle among at least three candidates in a later rear-camera image. We evaluate seven zero-shot vision-language models, four image-retrieval baselines, and 25 human participants. Models are tested using full front RGB images, cropped target vehicles, and binary silhouettes. The strongest VLM achieved 76.6 percent Rank-1 accuracy on target crops, compared with 74.0 percent for the frozen SigLIP2 baseline; this difference was not statistically clear. Human participants achieved 94.0 percent accuracy with full images and 92.2 percent with target crops. Under our evaluation setup, enabling reasoning improved accuracy across all three input conditions for every model evaluated in both modes. We also found that VLMs generally performed worse on full scenes than on target crops. These results show that general-purpose VLMs do not yet consistently outperform strong visual retrieval for front-to-rear vehicle matching, while humans remain substantially more reliable.
[CV-76] OmniReasoning : Pushing the Limits of Audio-Visual Joint Reasoning
链接: https://arxiv.org/abs/2609.39490
作者: Junming Lin,Yuxuan Wang,Zhenxin Lei,Yuxin Liu,Ruixun Liu,Yinsong Yan,Ling Wang,Minghao Han,Yunfei Chu,Shun Lei,Xueyao Zhang,Qize Yang,Jin Xu,Yiwu Zhong
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances have enabled unified omni-modal models in understanding audio, vision, and language. However, existing benchmarks, training data, and learning methods largely treat the modalities independently, leaving the capability of audio-visual joint reasoning poorly evaluated and insufficiently elicited. We address this gap with a benchmark, data engine, and learning method. First, we introduce OmniReasoningBench, a benchmark where both audio and visual evidence are indispensable. It comprises 1,150 multiple-choice and open-ended questions across two tasks, reasoning over video and reasoning beyond video. Second, we develop a data engine OmniQA. It automatically constructs evidence-grounded QA pairs that explicitly necessitate audio-visual joint reasoning, together with time-stamped clue chains that guide the annotation of thinking process. Besides our benchmark, this engine produces training data OmniReasoning-SFT-112K and OmniReasoning-RL-19K. Finally, we propose an on-policy self-distillation method Modality-Factored Self-Distillation (MFSD). It evaluates each sampled response under modality-specific clue contexts, disentangling the contributions of individual clues and their cross-modal interactions for token-level credit assignment. With our training data and learning method, our model OmniReasoning-30B-A3B achieves 50.0% on OmniVideoBench and 42.5% on OmniReasoningBench, improving the base model Qwen3-Omni-30B-A3B-Thinking by 12.8 and 9.3 percentage points, respectively. Moreover, it delivers substantial gains on general and long-video benchmarks, including Video-MME-v2. We hope our work offers a solid step for facilitating future research in omni-modal joint reasoning.
[CV-77] From Wrecks to Wisdom: Recovering Crash Mechanics from Real-World Multi-View Photos
链接: https://arxiv.org/abs/2609.39486
作者: Ondřej Valach,Václav Diviš,Ivan Gruber
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Estimating accident mechanics from real-world crashes is important for vehicle-safety analysis, injury modeling, crash-severity prediction, and operational workflows such as insurance claim triage. In standard crash records, key metadata such as impact configuration, principal direction of force, and change in velocity ( \Delta V ) may be missing, delayed, or corrupted, while post-crash photographs are widely available and contain rich visual evidence of deformation. We study how much crash-mechanics information can be recovered directly from vehicle photos when structured signals are absent. We formulate crash understanding as supervised prediction from per-case multi-view photo sets. Targets include six Collision Deformation Classification (CDC) descriptors and the longitudinal and lateral components of reconstructed \Delta V . Each photo is encoded by a shared visual backbone, and the resulting view-level features are fused into a case-level representation from which target-specific heads predict crash descriptors. Using 15.2k training cases from the Crash Investigation Sampling System, drawn from about 1.5M photos before filtering, together with 1.15k validation and 1.15k test cases, we define an evaluation protocol for vision-based crash descriptor estimation from incomplete multi-view evidence. Post-crash imagery alone provides usable signal for several non-trivial crash-mechanics descriptors, while weakly observable and long-tailed targets remain challenging. Within the compared training regimes, the selected joint-training recipe reduces mean absolute angular error for principal direction of force from 20.1 to 14.05 degrees and longitudinal \Delta V MAE from 8.04 to 7.45 km/h. Our work provides a reference point for future multimodal fusion with structured crash metadata.
[CV-78] DensePed-Lite: Quality-Aware Adaptive Detection for Dense Pedestrians under Occlusion
链接: https://arxiv.org/abs/2609.39467
作者: ZiAn Wang,MingZhe Liu,Chaoyi Guo,ChangChun Li,Fangming Gu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at WISE 2026
Abstract:Pedestrian detection plays a crucial role in computer vision with applications in autonomous driving, surveillance, and public safety. However, real-world dense scenes bring severe challenges, including heavy occlusion, drastic scale variations, and strict real-time requirements. Existing lightweight detectors struggle to balance accuracy and efficiency while often neglecting quality-aware feature modeling and consistency between classification and localization, leading to unstable performance under crowded conditions. To address these issues, we propose DensePed-Lite, a unified framework built on a single principle: under occlusion the network should adapt its behavior to the quality of what it observes rather than assume complete information. This principle is realized at three points where occlusion does the most damage: unreliable confidence scoring (UQE), fragmented spatial coverage (MPSC), and incoherent multi-scale fusion (CTDM). The three mechanisms reinforce one another instead of acting in isolation, all without significantly increasing complexity. Experiments on CityPersons and CrowdHuman validate that DensePed-Lite achieves a superior accuracy-efficiency trade-off compared with recent state-of-the-art lightweight methods, making it suitable for real-time deployment in dense pedestrian scenarios.
[CV-79] Mutual Equilibrium: Multimodal Representation Learning through Reciprocal Feedback
链接: https://arxiv.org/abs/2609.39456
作者: Ho-min Park,Byungkon Kang
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 20 pages
Abstract:This work proposes a mutual feedback architecture, MEQ, that refines the two inputs, of possibly different modalities, into a pair of coupled embeddings such that each embedding reflects the information of the other. The core idea is to incorporate continuous interchange of information between the two inputs. This idea leads to a mutual feedback architecture consisting of two components whose outputs are fed back into the other. The final output of this model is defined as the fixed point of this interaction. We provide theoretical analysis that offers interpretation of this model as well as design choices to prevent failure cases. We show the benefits of MEQ through classification and visual grounding tasks spanning various datasets. Quantitatively, our model outperforms or shows competitive performance on concatenation-based multimodal classification problems. Qualitatively, the proposed interactive mechanism allows the model to progressively refine the visual grounding when paired with complementary modality, thus demonstrating the power of mutual feedback under such settings.
[CV-80] ResARC: Residual-Aware AutoRegressive Coding for Ultra-Low Bitrate Image Compression
链接: https://arxiv.org/abs/2609.39451
作者: Qin Yan,Ruixiao Dong,Yutao Xie,Li Li,Ying Chen,Kai Li,Daowen Li,Houqiang Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Progressive autoregressive image codecs provide an appealing paradigm for generative compression by quantizing continuous latents into discrete tokens, transmitting coarse-to-fine prefix tokens and generating the remaining suffix tokens at the decoder. However, their reconstruction quality is fundamentally limited by two residuals introduced along this pipeline: the quantization residual, arising from information loss during discrete tokenization, and the generation residual, resulting from imperfect autoregressive generation of the suffix tokens. To address these limitations, we introduce ResARC, a residual-aware autoregressive codec that explicitly compensates for both residuals at the decoder. Specifically, we generate the quantization residual with a diffusion transformer conditioned on the autoregressive decoding context, while requiring no additional side information. In parallel, we compute the generation residual at the encoder and employ a learned Generation Residual Codec to efficiently compress and transmit it for decoder-side correction. The recovered residuals are then integrated with the reconstructed latent representation and decoded through an adapted VAE decoder. Extensive experiments demonstrate that ResARC achieves competitive perceptual similarity while substantially improving distributional fidelity over leading generative codecs across the ultra-low bitrate regime. Code and models will be released soon.
[CV-81] CAST: Causal Advantage-Structured Training with Spatially Grounded Compositional Rewards for Diffusion Models
链接: https://arxiv.org/abs/2609.39441
作者: Shu Yu,Chaochao Lu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Project page: this https URL
Abstract:Online reinforcement learning has been extended to flow matching for diffusion model (DM) image generation. However, this paradigm faces three limitations: (1) Window selection. Existing methods manually set the stochastic differential equation (SDE) sampling window, i.e., the denoising steps where exploration noise is injected. We instead determine it from each model’s denoising trajectory. (2) Reward saturation. Current methods rely on scoring models trained on human annotations; we find that such scores are extremely high and nearly indistinguishable on the latest SOTA open-source DMs, making advantage estimation largely ineffective. (3) Sample inefficiency. A single scalar reward collapses different failure modes into almost identical scores, leaving minimal gradient guidance for targeted improvement. To address these issues, we propose CAST (Causal Advantage-Structured Training), an RL fine-tuning method for pretrained DMs, which (1) identifies the denoising step at which each model fixes the objects and their spatial arrangement in the image and uses that timing to set the SDE window, (2) decomposes each prompt via Causal Scene Graphs (CSG) into verifiable-atoms, i.e., minimal semantic units such as an object, count, attribute, or spatial relation that can each be checked independently, and rewards each atom separately, and (3) projects the signed atom-level advantages into pixel space through teacher-forced attention and uses them to spatially weight the SDE policy objective. We fine-tune two of the strongest open-source DMs, FLUX.2-dev and Qwen-Image-2512, with CAST, and evaluate them on GenEval 2, a compositional benchmark, and on Qwen-Image-Bench for overall quality. Within almost the same training budget, CAST’s improvement over the base model on the most challenging GenEval 2 prompts is up to 3.07x that of Flow-GRPO, while overall generation quality also improves.
[CV-82] owards Trustworthy AI for Glioma Diagnosis: A Task-Aware Evaluation of Uncertainty Quantification
链接: https://arxiv.org/abs/2609.39429
作者: Gonzalo Esteban Mosquera Rojas,Sebastian R. van der Voort,Carolin M. Pirkl,Sandeep Kaushik,Marion Smits,Stefan Klein
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted for publication at the Journal of Machine Learning for Biomedical Imaging (MELBA) this https URL
Abstract:Uncertainty Quantification (UQ) is a key requirement for trustworthy AI in high-stakes medical image analysis. In this work, we evaluate UQ in a multi-task Deep Learning framework for MRI-based glioma diagnosis that performs tumor segmentation and predicts IDH mutation status, 1p/19q co-deletion status, and tumor grade. Monte Carlo Dropout (MCD) is used for a detailed task-aware analysis of predictive, aleatoric, and epistemic uncertainty. We assess MC sample convergence, calibration, error detection, selective prediction, associations with segmentation performance, and the effect of voxel-wise uncertainty aggregation on case-level reliability. We also compare MCD with Deep Ensembles (DE) and Monte Carlo Deep Ensembles (MCDE), examine interactions between segmentation quality and classification, and evaluate a composite trust score integrating segmentation and classification uncertainty. Across tasks, uncertainty estimates supported meaningful error detection, while calibration depended on the dropout rate, with moderate rates yielding the most reliable probabilities. Uncertainty decomposition provided task-dependent interpretability but did not consistently improve error detection over predictive uncertainty alone. DE and MCDE showed comparable operational utility, with no method consistently dominating across tasks and metrics. The composite trust score did not consistently outperform classification uncertainty for selective prediction. Overall, our results provide a task-aware evaluation strategy and practical guidance for the development of trustworthy AI for glioma diagnosis.
[CV-83] PCB-MC: Missing Component Analysis in Printed Circuit Boards
链接: https://arxiv.org/abs/2609.39427
作者: Betsy Villa Brochero,Ian Gibson,Estefania Talavera
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint
Abstract:Detecting missing components on printed circuit boards (PCBs) differs fundamentally from conventional object detection, as the model must localize components that are not present. We introduce PCB-MC, a curated dataset for missing component detection with footprint level annotations built on top of the RF100 dataset. The dataset contains 197 distinct board types, each corresponding to a unique PCB design, with multiple augmented samples per type. We also provide benchmark results on PCB-MC by evaluating a diverse set of supervised and unsupervised methods. To ensure fair evaluation, we propose board type aware cross validation splits that prevent layout leakage between training and test sets. Supervised models showcase high false negative rates on unseen board designs, and unsupervised anomaly detection methods fail entirely due to the lack of spatial alignment with a board specific reference. These results confirm that missing component detection on diverse PCB layouts remains an open challenge. We release PCB-MC and all training protocols to support reproducible research on structural absence detection in industrial inspection.
[CV-84] UniWAM Technical Report: Unified Mobile Manipulation via Mixed-Stream World-Action Modeling and Manipulation Anchor Pose Supervision
链接: https://arxiv.org/abs/2609.39388
作者: Wei Xue,Keliang Liu,Mingzhang Cui,Jinhua Xie,Jinjie Wei,Jianan Hou,Jingcheng Lu,Lintao Wang,Kaixiang Qiu,Yizhou Liu,Xinghai Ye,Jinghang Han,Mingcheng Li,Jie Gu,Shunli Wang,Lihua Zhang,Dingkang Yang
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: UniWAM Technical Report
Abstract:Mobile manipulation requires precise navigation to a manipulation-ready pose followed by reliable object interaction. These two stages differ in action spaces and visual requirements, which complicates unified policy learning. In addition, collecting diverse real-world navigation data with explicit manipulation-ready pose supervision remains costly and difficult to scale. We introduce UniWAM, a unified mixed-stream world-action model with separate action encoders and output heads for navigation and manipulation, sharing a common backbone. This design supports joint representation learning on independently sampled navigation and manipulation data. UniWAM supports independent inference for either stream and batch-parallel inference for both. We further introduce Manipulation Anchor Pose (MAP) supervision for where to stop and how to orient for manipulation. An automated pipeline constructs MAP-Data from large-scale 3D scenes, yielding over 1.5 million episodes and 7,500 hours. MAP-Data provides per-frame target-object bounding boxes and image-plane MAP coordinates as auxiliary navigation supervision. Together with projected end-effector trajectories for manipulation, these prediction targets provide stream-specific image-plane supervision for action learning from egocentric observations. With large-scale MAP-Data, UniWAM outperforms the strongest external baselines on our MAP-Bench by 30.1% in position error and 44.0% in heading error. Across 24 real-robot tasks, UniWAM achieves leading results in MAP navigation and mobile manipulation, with competitive manipulation performance. We have released code, data, and benchmark.
[CV-85] InfoAgent : Traceable Generation and Repair of Evidence-Grounded Infographics
链接: https://arxiv.org/abs/2609.39380
作者: Yifan Li,Tong Li,Qi Zeng,Lishuai Gao,Ruwei Pan,Cong Wei,Shaohua Kevin Zhou,Zhuoliang Kang,Xiaoming Wei
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reliable infographic generation requires facts, symbols, and visual relations to remain consistent through rendering and revision. Correcting one element also requires tracking its supporting evidence and the dependencies affected by the change. We present \textbfInfoAgent, a training-free framework for \emphevidence-bound visual-symbolic program synthesis. Its Infographic Visual Description (IVD) records factual payloads, evidence provenance, execution routes, and verification obligations in a typed dependency graph. Retrieved design priors guide compilation, and layered execution combines raster synthesis with editable symbolic and binding objects while retaining their traces. Dependency-aware repair localizes corrections, rechecks affected dependencies, and requires protected obligations to remain satisfied under the declared checkers. Unresolved obligations remain explicit. On IGenBench, InfoAgent achieves 93.0 Q-ACC and 59.0 I-ACC. We also introduce InfoGraphicBench-Evidence, where complete-checklist pass rates on 200 test requests increase from 21.5% for Same-IVD Prompt to 23.5% for the initial layered output and 28.5% after repair, using the same evidence and initial IVD. On 120 audited repair cases, localized repair edits 12.4% of the canvas on average, compared with 67.3% for global regeneration.
[CV-86] EgoTools: Towards Tool-Centric Reasoning in Real-World Egocentric Videos
链接: https://arxiv.org/abs/2609.39378
作者: Shulin Tian,Junsu Kim,Shuai Liu,Hao Li,Yujiao Shen,Sihan Li,Zhe Yang,Yeongon Kim,Feiyu Li,Jialin Wu,Yichi Zhang,Wenhui Wang,Runmao Yao,Yuhao Dong,Zhaoxi Chen,Fangzhou Hong,Antonino Furnari,Jingkang Yang,Hongyuan Zhu,Ziwei Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 32 pages, 7 figures. Project page: this https URL
Abstract:Real-world embodied tasks, from everyday activities to professional procedures, require agents to act under physical constraints while tracking evolving object and task states. Tool use sits at the heart of such tasks, as many everyday and professional activities are tool-mediated. Understanding them requires reasoning about affordances, hand-tool-object geometry, procedural progress, and causal effects on target objects. Yet despite strong performance on perception-oriented video tasks such as captioning and general video QA, current multimodal video models remain limited in this form of tool-centric embodied reasoning. Progress in this direction has been limited by the lack of real-world egocentric data and diagnostic benchmarks. To address this gap, we introduce EgoTools, the first comprehensive suite for egocentric tool-use understanding. It consists of two complementary components: EgoTools-Data, a large-scale corpus of 100 hours of tool-centric egocentric recordings with synchronized audio, dense captions, reasoning-heavy narrations, and supplementary 3D information; and EgoTools-Bench, a diagnostic benchmark of 1,000 QA pairs across four tracks that cover tool-use understanding from perception and geometry to procedure and causal reasoning. Experimental results show that current models still struggle to ground tool use in visual evidence: Gemini-3.1-Pro achieves 66.9% overall accuracy but only 51.7% on Perception Grounding. Beyond evaluation, we validate EgoTools-Data as a training resource. On the full 1,000-question benchmark, full supervised fine-tuning improves Qwen3-VL-8B-Instruct from 50.0% to 60.9%, under strict source-video separation. Together, these results establish EgoTools as a unified resource for both training and diagnostic evaluation of real-world egocentric tool-use understanding.
[CV-87] Beyond the Current Scene: Event-Referential Grasping with Active View Selection WWW
链接: https://arxiv.org/abs/2609.39375
作者: Hyunjoon Lee,Haebeom Jung,Eunsung Cha,Daeun Lee,Yu-Chiang Frank Wang,Jaesung Choe,Jaesik Park
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:A robot that observes people interacting with objects should be able to carry out later requests that refer back to those interactions. Such requests may specify a grasp target by the role it played in a past event rather than by its name or appearance. Moreover, the target may no longer be visible when the robot is asked to act. We present BeyondSCe, a zero-shot robotic grasping system for this event-referential setting. Given the event history and the current scene, the system identifies the requested object or part and localizes it for grasping. If the target is occluded, it combines an event prior recovered from the history with current scene geometry to select camera viewpoints likely to reveal the target. The system uses pretrained models without additional task-specific training. In real-robot experiments with a single wrist-mounted RGB-D camera, it achieves grasp success rates of 76% and 77% for initially visible and occluded targets, respectively, compared with 40% and 55% for the strongest baseline in each condition. On four additional scenes with heavy occlusion, it increases grasp success rates from 75% to 95% while reducing the mean number of views from 3.35 to 2.20, compared with an active-perception baseline given the target’s ground-truth 3D bounding box.
[CV-88] COBICount: Separating Object and Background Responses for Remote Sensing Object Counting Without Training on Target Data
链接: https://arxiv.org/abs/2609.39366
作者: Junjing Zheng,Zhiyi Zhou,Ningrui Yang,Hongying Meng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 19 pages, 7 figures
Abstract:Remote sensing object counting estimates how many buildings, vehicles, or ships appear in overhead images. Most supervised counters predict a density map, whose sum gives the object count, and assume similar categories, sizes, and backgrounds. Applying them across regions, sensors, or categories often requires target data or further training, which may be costly or unavailable. We study source-only counting. Training for the counting task and model selection use one group of images that shares an object category and similar imaging conditions, with one point marking each object. Target images and information remain unavailable until the model is fixed. This reduces data preparation but makes transfer harder. A model trained on one source may place high density values, called responses, on real objects and repeated background structures. Road edges, parking grids, roof boundaries, and water boundaries may then be counted as objects, creating candidate origin ambiguity. COBICount separates response generation, acceptance, and background suppression. Candidate Evidence (CE) generates possible responses. Candidate Acceptance (CA) keeps compact responses centered on objects. Bias Isolation (BI) reduces responses associated with repeated background structures. Their outputs form the final density map. Trained on RSOC Building and evaluated directly on DOTA Large Vehicle, Small Vehicle, and Ship, COBICount achieves the lowest mean absolute error (MAE) averaged over the target domains among the compared methods, 174.132. It uses 5.07 million parameters and 17.41 billion floating point operations for a 512x512 input. COBICount improves transfer without target data or training for each target. The code will be available at: this https URL.
[CV-89] Rethinking Multi-Image Re-Representation in Multi-Image Understanding
链接: https://arxiv.org/abs/2609.39363
作者: Gengyuan Zhang,Xiao Han,Xinyu Xie,Tong Liu,Volker Tresp
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 27 pages, 7 figures, 9 tables
Abstract:Multi-image understanding requires MLLMs not only to recognise the content of individual images, but also to organise visual evidence distributed across them. We study this problem through multi-image re-representation, viewing prompted Chain-of-Thought reasoning and agentic visual tool use as different ways of re-organising visual evidence during reasoning. We introduce Mosaic, a general-purpose multi-image visual harness that enables an MLLM to actively construct visual intermediates with ten composable image operations. We compare five re-representation settings on existing multi-image benchmarks and on MosaicBench, a new grounding-focused benchmark for fine-grained multi-image understanding. Our experiments show that the relative benefits of textual and visual re-representation are strongly task-dependent. Visual re-representation is particularly effective for tasks requiring precise visual evidence, including hypothesis testing, precision comparison, and orientation-sensitive reasoning, while tasks dominated by higher-level semantic content show smaller or less consistent gains. Building on this finding, we train MosaicAgent-8B to use Mosaic with reinforcement learning using only accuracy and format rewards. Without demonstration trajectories or rewards for specific tool-use, the agent learns to compose visual operations over multiple steps and exhibits diverse problem-solving patterns unpromptedly. Code and data will be released at this https URL.
[CV-90] xTailor: Texture-Preserving Video Virtual Try-On via Adaptive Garment Conditioning
链接: https://arxiv.org/abs/2609.39335
作者: Zijing Qin,Jun Zhou,Ruicheng Zhang,Jiaqi Hou,Zunnan Xu,Ronghui Li,Zhenyu Xie,Xiu Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video virtual try-on has attracted increasing attention due to its broad potential in digital fashion and intelligent e-commerce. However, existing methods primarily focus on low-resolution settings and still face substantial challenges when extended to high-resolution scenarios. These limitations can be attributed to two main factors: (1) the insufficient utilization of rich garment reference information, and (2) the lack of explicit positional modeling between garment and video representations during cross-modal interaction, which weakens fine-grained local correspondence. To address these issues, we propose TexTailor, a high-fidelity video virtual try-on framework built upon a pretrained video Diffusion Transformer. Specifically, we introduce a timestep-adaptive modulation mechanism to dynamically adjust garment visual representations throughout denoising. We further develop a frame-aligned positional encoding strategy to strengthen garment-to-video correspondence, together with a multi-source injection design that reduces interference among heterogeneous conditions. Extensive experiments on multiple video virtual try-on benchmarks, including the high-resolution Eevee dataset, demonstrate that TexTailor achieves competitive performance in garment detail preservation, temporal consistency, and overall video quality.
[CV-91] MotionWeave: Learning Motion-Centered Future Dynamics for Vision-Language-Action Policies
链接: https://arxiv.org/abs/2609.39324
作者: Jingqiu Wang,Yan Wang
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 4 pages + 1 page references, 3 figures, 2 tables. Code: this https URL
Abstract:Vision-Language-Action (VLA) models have recently incorporated world models to provide richer dynamic supervision beyond sparse action labels. However, explicitly predicting future images or videos may include control-irrelevant appearance, while guidance derived from holistic future visual representations and shared global action features may fail to establish timestep-specific correspondence between actions and local visual changes. To address this issue, we propose MotionWeave, a motion-centric future-dynamics framework for action-chunk prediction with two modules: the Action-Induced Motion Grounder (AIMG) and the Horizon Residual Composer (HRC). Specifically, AIMG conditions on action and proprioceptive representations to construct horizon-specific queries that localize interaction regions associated with each future action timestep from current visual tokens. HRC extracts differences between interaction representations at adjacent horizons, encodes them as temporal motion cues, and injects them into action tokens through a gated residual. During training, robot-arm masks rendered from future frames are used to construct KL-based motion-grounding supervision, while inference uses only the current observation. On six MetaWorld tasks, MotionWeave achieves a 75.3% average success rate, an absolute gain of 8.6% over \pi0 (66.7%), especially on sustained-interaction tasks. Our code is available at this https URL.
[CV-92] Rethinking Generative Image Compression at Extremely Low Bitrates
链接: https://arxiv.org/abs/2609.39315
作者: Tianyu Zhang,Zhaoyang Jia,Houqiang Li,Dong Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Generative image compression produces visually plausible reconstructions at low bitrates, yet their behavior as the rate approaches zero remains largely unexplored. When pushed below normal operating rates, representative codecs undergo semantic collapse: rather than gracefully losing source-specific detail, they produce malformed or unrecognizable content. Our analysis identifies two factors. As the bitrate decreases, reconstruction losses increasingly conflict with semantic objectives on gradients and visual results, while pixel-space and reconstruction-oriented VAE diffusion models become less efficient on semantic preservation. Guided by these findings, we introduce RAE-CoD, a compression-oriented diffusion (CoD) built in a representation autoencoder (RAE) space with direct alignment between compressed and source representations, preserving recognizable, naturally structured content for a 256\times256 image with as few as 16 bits. We evaluate this framework using five vision foundation models (VFM) and a blinded vision-language model protocol. On MSCOCO-30K, RAE-CoD stands out from all evaluation. At 0.001-0.008 bpp, it reduces relative VFM feature MSE and Fréchet Distance ratio by at least 25.7% and 69.1% over the best competitors. Meanwhile, semantic recognizability and quality of the reconstructions remain nearly constant while source consistency falls smoothly, replacing abrupt semantic collapse with a graceful transition toward unconditional generation. Code will be released at this https URL.
[CV-93] BMASH: Ball-Motion-Aware Soccer Header Spotting
链接: https://arxiv.org/abs/2609.39300
作者: Ahmed Endris Hasen,Muhammad Shahzad Khan,Nikolaos Passalis,Jenni Raitoharju
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages, 4 figures, MMsports
Abstract:Recent advances in computer vision have made broadcast sports videos increasingly useful for event analysis, performance assessment, and player-safety applications. In soccer, however, header spotting remains a challenging problem due to the subtle and short-lived nature of header events. This paper focuses on soccer header spotting: identifying moments in broadcast videos where the ball contacts a player’s head. We first adapt and evaluate Video Swin as a strong action-recognition baseline for this task, and then introduce BMASH, a ball-motion-aware fusion framework that integrates detector-derived ball features. BMASH combines Video Swin action representations with ball-presence and motion features from frame-level soccer-ball detection, integrating player-action context with ball dynamics to distinguish headers from visually similar events. We evaluate BMASH using game-level splits with separate test matches and rotating validation folds, considering both centered-window classification and continuous full-video spotting. Results show that Video Swin provides a strong baseline for header spotting, while BMASH improves clip-level AP and ROC-AUC over the corresponding Video Swin baseline. In continuous full-video spotting, BMASH achieves a comparable event-level F1-performance with a different precision–recall trade-off.
[CV-94] MegaAvatar: Controllable Talking Avatar Generation
链接: https://arxiv.org/abs/2609.39273
作者: Junyao Gao,Sibo Liu,Weidong Zhang,Cairong Zhao,Jun Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 7 pages, 6 figures
Abstract:This report presents \textbfMegaAvatar, a controllable talking avatar generation framework built on top of the Wan2.2-TI2V-5B model. Compared with previous talking-avatar methods that mainly rely on audio or reference-image conditioning, we introduce additional SMPL-X-derived 3D guidance, enabling global control over body pose and head motion. Specifically, we render the driving SMPL-X sequence into dense mesh frames and encode them with a lightweight 3D convolutional encoder, whose outputs are injected into the latent tokens to provide overall motion control. Furthermore, we extend Wan2.2-TI2V-5B with additional audio and face cross-attention modules to enable fine-grained expression control and preserve the input identity, respectively. In addition, we implement an audio-to-SMPL-X model to predict an SMPL-X sequence conditioned on the reference image and input audio, allowing MegaAvatar to support audio-driven inference without user-provided SMPL-X frames. Experiments show that MegaAvatar achieves high-quality talking avatar generation with controllable body and head motion, speech-synchronized facial expressions, and consistent identity preservation. MegaAvatar also supports inference with flexible resolutions and video lengths. Codes, dataset, models will be avaliable in this https URL
[CV-95] PLRS-IC: A Dual-Calibration Framework for Chest X-Ray Vision-Language Alignment
链接: https://arxiv.org/abs/2609.39266
作者: Qixing Zhao,Jinpeng Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages, 4 figures, 5 tables
Abstract:Fine-grained vision-language alignment in chest radiography enables zero-shot classification, grounding, and segmentation without task-specific annotations. However, this alignment is fundamentally hindered by two intertwined sources of ambiguity: projection-induced visual mismatch and patient-agnostic semantic overlap. First, at the local feature level, frontal and lateral radiographs exhibit distinct appearances for the same clinical finding, rendering a shared patch-text similarity geometry inherently suboptimal. Compounding this visual ambiguity is a semantic mismatch during global contrastive optimization, where instance-level objectives penalize cross-patient pairs as strict negatives even when they share identical positive clinical concepts. To address this dual ambiguity, we propose PLRS-IC, a unified dual-calibration framework for chest X-ray representation learning. At the local alignment stage, Projection-Conditioned Low-Rank Residual Similarity (PLRS) dynamically adapts patch-text matching to projection-specific manifolds using a bounded, parameter-efficient low-rank residual. At the global optimization stage, Information-Content-Calibrated Soft False-Negative Suppression (IC-SFNS) leverages a corpus-derived information-theoretic prior to soften the penalty of semantically overlapping negatives without altering original contrastive assignments. Extensive experiments across nine public zero-shot benchmark settings demonstrate that our framework yields consistent improvements in classification, grounding, and segmentation, validating the necessity of dual-calibration in medical vision-language pre-training.
[CV-96] Universal Cross-Prompt Adversarial Attacks on Promptable Concept Segmentation NEURIPS2026
链接: https://arxiv.org/abs/2609.39265
作者: Ziqi Zhou,Yifan Hu,Yufei Song,Haowen Jiang,Xianlong Wang,Shengshan Hu,Dezhong Yao,Leo Yu Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by NeurIPS 2026
Abstract:The Segment Anything Model (SAM) achieves remarkable performance in visual segmentation. The latest SAM3 extends promptable segmentation to concept-level prediction, broadening the scope of segmentation foundation models. While recent works reveal that SAM and SAM2 are vulnerable to adversarial examples, the robustness of SAM3 under the concept segmentation paradigm remains unexplored. In addition, existing adversarial attacks on SAM-series models exhibit limited cross-prompt transferability. To this end, we propose AdvPCS, a universal cross-prompt adversarial attack for Promptable Concept Segmentation (PCS), including a min-max prompt optimization strategy, a global-local perception deception attack, and a temporal transition deviation attack. Specifically, we first identify the hardest-to-attack prompts via min-max bilevel optimization. In the inner maximization, we enhance diversity over candidate point, box, and text prompts. In the outer minimization, we select prompts with the highest responses based on the confidence scores output by the detector. Given the selected prompts, we apply the perception deception attack to minimize both global and local existence probabilities under joint prompting and employ the temporal memory misalignment attack to maximize inter-frame semantic inconsistency and corrupt memory pointers. Extensive experiments on four benchmark datasets show that a single universal adversarial perturbation (UAP) generated by our method generalizes across frames from different videos and achieves strong attack performance under point, box, and text prompts. In particular, it reduces the average mIoU of various PCS models on the SA-CO dataset to below 5% under text prompts, demonstrating strong attack ability.
[CV-97] he Planning Limits of Latent World Models
链接: https://arxiv.org/abs/2609.39235
作者: Ali Alrasheed,Basim Azam,Naveed Akhtar
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:World models offer a promising way to help robots understand how the physical world evolves and plan complex behaviours through imagination. Yet existing studies mainly demonstrate what these models can accomplish, leaving unclear when their predictions remain useful for planning and where they fail. We study this question using action-conditioned predictors built on five frozen self-supervised visual backbones: V-JEPA 2, V-JEPA 2.1, VideoMAEv2, VideoPrism, and DINOv2. We use frozen backbones to test representations intended to transfer across environments. We evaluate these models on diverse Meta-World manipulation tasks and real-robot interactions from BridgeData V2. We find that a world model guides action selection reliably only when the goal lies within, or slightly beyond, the trajectory it imagines during planning. With five-step rollouts, the length the predictor was trained on, the world model ranks actions reliably only for targets five to ten control steps ahead, whereas task goals lie 16 to 53 steps away. Neither an 81-fold larger predictor nor longer-rollout training extends this range; the encoder affects both range and closed-loop success, with V-JEPA 2.1 performing most consistently. More fundamentally, the limit persists under perfect prediction: using the real simulator, success falls from 92% to 41% as the target moves from five to twenty steps ahead of a five-step rollout. Planning therefore requires either longer imagined trajectories or closer subgoals. For distant goals, pure imagination succeeds in 23% of episodes, planning with feedback (MPC) raises success to 30%, imagining as far as the goal to 47%, and nearby expert subgoals to 76%. Used within its plannable range, a world model can also improve a vision-language-action (VLA) policy: choosing among eight actions the VLA proposes raises its success from 65% to 77% across 16 different tasks.
[CV-98] Emergent Multi-View Geometry Through Self-Distillation
链接: https://arxiv.org/abs/2609.39227
作者: David Nordström,Thibaut Loiseau,Vincent Lepetit,Michael Felsberg,Guillaume Bourmaud,Fredrik Kahl
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Over a century ago, Henri Poincaré argued that a motionless observer cannot acquire the notion of space. Yet, most visual representation learning methods operate on individual images, while those that leverage multiple views rely on RGB reconstruction, entangling geometry with appearance. We propose Poincar3, a self-supervised method that learns representations from multiple views through self-distillation instead of RGB reconstruction. We combine masked patch and image-level distillation with a teacher that observes additional views, enabling training from scratch without explicit 3D supervision. Poincar3 outperforms both previous single and multi-view self-supervised approaches such as DINOv3, MuM, and Muskie on correspondence estimation, camera pose estimation, and 3D reconstruction. Using a lightweight Poincaré adapter, we also find that our learned features encode camera motion more accurately than existing self-supervised representations.
[CV-99] DC-SAE: Deep Compression Semantic Autoencoder for Faster Diffusion Convergence
链接: https://arxiv.org/abs/2609.39222
作者: Xu Huang,Ye Huang,Zijun Liao,Yuwei Niu,Xiaojie Li,Menghan Zhou,De Wen Soh,Xiaotong Li,Daquan Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages, 5 figures. Project page: this https URL
Abstract:High-compression tokenizers are essential for scaling latent image generative models. However, aggressive compression creates a fundamental tradeoff between reconstruction fidelity and generation efficiency: high compression image encoder always increases the learning difficulty of diffusion training, resulting in slow model convergence. Recent representation autoencoders speed up the diffusion training by improving the latent feature’s expressive capability by replacing VAE encoders with pretrained semantic encoders, yet they are typically limited to moderate compression and lose pixel-level details necessary for faithful reconstruction. To achieve both high compression and fast diffusion training, we propose DC-SAE, a Decoupled Compact Semantic Autoencoder designed for high-compression image generation with accelerated diffusion model convergence. DC-SAE consists of two key components: (1) a macro-level architecture design that leverages semantic encoders to enable higher compression ratios, and (2) a pixel-level encoder that preserves low-level details, ensuring high-fidelity image reconstruction. We empirically demonstrate that DC-SAE performs strongly on image generation tasks, achieving both compact latent representations and efficient training dynamics. Specifically, on the ImageNet dataset with 512 \times 512 resolution, DC-SAE achieves 32\times spatial compression, with 29.79 PSNR and 3.37 gFID, substantially outperforming the previous state-of-the-art high-compression tokenizer baselines DC-AE by 13.5% and 54.9% on PSNR and gFID, respectively, maintaining comparable throughput and faster diffusion model training convergence. Beyond class-conditional generation, a 1.6 B-parameter DiT using DC-SAE achieves 0.84 on GenEval and 86.007 on DPG-Bench for text-to-image generation at 1024\times1024 resolution.
[CV-100] Uruqi: Learning Spatial Cognition from Visual Experience
链接: https://arxiv.org/abs/2609.39195
作者: Shichao Li,Meiqi Wang,Fei Su,Zhicheng Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 23 pages
Abstract:Spatial intelligence requires maintaining a coherent understanding of the world as the embodied agent moves. Like humans, the agent must use its own motion to interpret changes across observations and update object locations and spatial relations accordingly. Despite spatial post-training having substantially broadened the spatial intelligence of vision-language models (VLMs), they still struggle with two atomic spatial capabilities: tracking self-motion and mapping the surrounding world during motion. To address this gap, we provide dense multi-turn supervision over interleaved atomic capabilities within each training episode, mimicking the visual experience of a continuously moving agent that reasons as it observes. To scale this up, we synthesize 11,738 motif-driven camera trajectories over a broad range of 3D scenes, supporting self-motion tracking, persistent object mapping, and rich spatial operations within each visual experience. By training models to reason over these atomic questions, our URUQI _\mathrmSyn -8B improves accuracy from 15.84% to 47.73% on our Uruqi benchmark comprising 52k questions across 2.7k episodes. URUQI-SI-Mix-8B further reaches 50.41%, comparable to the 50.08% achieved by GPT-6 Astra. Trained solely on our synthesized data, URUQI _\mathrmSyn -8B achieves an average relative accuracy improvement of 17.13% over its InternVL3-8B backbone across three external spatial benchmarks. These results highlight continuous visual experience as a scalable source of supervision for developing spatial cognition in VLMs.
[CV-101] Fiber-Resolved Microstructure Quantification from Multi-Shell Diffusion MRI using Detection Transformers MICCAI2026
链接: https://arxiv.org/abs/2609.39184
作者: Sebastian Endt,Marcus Wirth,Johannes Reinhold Schlund,Marion Irene Menzel
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Medical Physics (physics.med-ph)
备注: 12 pages, 4 figures, 1 table; Accepted at MICCAI 2026 Workshop CDMRI; Code: this https URL
Abstract:Fiber orientation and compartmental microstructure are central to the characterization of white matter tissue in diffusion MRI, yet existing methods either resolve fiber orientations without quantifying microstructure, or quantify microstructure while assuming a fixed number of compartments and a single fiber direction. Nonparametric approaches that recover both require tensor-valued diffusion encoding and computationally expensive Monte-Carlo inversion of an ill-posed inverse Laplace transform. We propose to reframe this problem as an object detection-like task, adopting the Detection Transformer (DETR) architecture to jointly predict mean diffusivity (MD), fractional anisotropy (FA), main fiber direction, and signal fraction for a variable number of compartments per voxel from standard multi-shell diffusion MRI with linear encoding. Hungarian matching during training resolves permutation invariance across compartments. We introduce mean Average Precision as a reproducible benchmark metric. Evaluated on synthetic test data with up to five compartments per voxel, our model achieves R^2=0.95 for MD, R^2=0.88 for FA, and a median angular error of 4.2°, with performance scaling naturally with compartmental signal fraction.
[CV-102] Aligning Thoughts with Answers: Probability Rewards to Tame Thinking Drift
链接: https://arxiv.org/abs/2609.39183
作者: Pengzhan Sun,Shiu-hong Kao,Shijie Li,Yongyi Su,Junbin Xiao,Arjun Reddy Akula,Angela Yao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:This paper studies \textbfthinking–answer consistency in vision-language models. We focus on Visual Intention Grounding, where a model infers a target object based on a human intention query and predicts a bounding box. We reveal that previous IoU-based reinforcement learning (RL) frameworks suffer from ``thinking drift’', where the model produces a correct bounding box, despite having an incorrect reasoning process pointing to a different target object. Thus, we propose \textbfRita (\textitReInforcing Thinking–Answer consistency) as a novel RL paradigm to tame the drift. Specifically, Rita introduces two reasoning-label-free RL rewards, constructed from the conditional probability of reference answers: a \textbfthinking reward and a \textbfconsistency reward. It also adopts a difficulty-aware \textbfdata filtering strategy that selects informative easy-to-medium samples for RL using rollout error rate and reward variance. Extensive experiments on EgoIntention and the new RefEgo-Int benchmarks show that Rita performs consistently superior to the supervised finetuning approaches and vanilla RL-finetuned frameworks.
[CV-103] MEND: Label-Free Detection Localisation and Correction of Latent Hallucination in World Models
链接: https://arxiv.org/abs/2609.39182
作者: Ali J Alrasheed,Aryan Yazdan Parast,Basim Azam,James Bailey,Naveed Akhtar
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted at DICTA 2026 (International Conference on Digital Image Computing: Techniques and Applications). Camera-ready version
Abstract:World Models are appearing as the next major frontier in computer vision. However, their robustness is currently largely unexplored. We identify the phenomenon of hallucination in latent World Models: given a state and an action, the predicted next latent can decode to a scene that never occurs. Because the prediction is statistically ordinary and is fed back autoregressively by the model, the error is both silent and compounding. We study whether such latent hallucination can be detected, localised, and corrected at inference time, on a frozen self-supervised world model in the absence of ground-truth error labels. We introduce Masked Empirical-Bayes Neural Denoising (MEND), a single conditional score network trained by denoising score matching on real transitions, whose score field serves three roles: its magnitude detects hallucination, its per-token field localises it to specific image patches, and it defines an inference-time correction direction. On two navigation environments MEND detects hallucination with an AUROC of up to 0.80 without using actions, exceeding a single-Gaussian density baseline while also localising the error (per-token AUPRC up to 0.87) and correcting it, all from one score field. Our correction reliably reduces single-step latent error and improves predictions. We identify that a part of the error is tangent to the data manifold, hence, we focus on detection and localisation while highlighting promises of the correction.
[CV-104] Exploiting Vulnerabilities: Universal Adversarial Attacks on Vision-Language-Action Models in Robotics ICRA2026
链接: https://arxiv.org/abs/2609.39178
作者: Songhua Yang,Ziyu Liu,Yuanwei Liu,Xuetao Li,Xuanye Fei,He Huang,Zheng Wang,Miao Li
类目: Robotics (cs.RO); Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to the 2026 IEEE International Conference on Robotics and Automation (ICRA 2026), Vienna, Austria. 8 pages. © 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes
Abstract:Recently, Vision-Language-Action (VLA) models have revolutionized robotic manipulation by seamlessly integrating visual perception, language understanding, and action generation in an end-to-end learning framework. However, since these models are designed to interact directly with the physical world and humans, their security is critical, and even small vulnerabilities can lead to catastrophic failures. In this work, we propose the Universal Adversarial Object, a sphere with optimized surface texture that significantly degrades task success rates when placed within the robot’s field of view. Specifically, our approach introduces a multi-level attack framework that jointly disrupts trajectory planning, task execution, and action control. We validate our method in both simulated and real-world robotic settings. Experimental results demonstrate that the adversarial object reduces the average task success rates by 31.2%-39.9% for two representative VLA models (Pi0 and RDT), with success rates dropping to near zero in complex scenarios. Index Terms–Vision-Language-Action models, adversarial attack, robotic security, universal adversarial object
[CV-105] ripleFlow: Training-Free Video Object Removal by Bridging Residual Editing and Native Generation
链接: https://arxiv.org/abs/2609.39157
作者: Songhe Wang,Lifu Wei,Shuolin Xu,Charles A. Kamhoua,David Miller
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 24 pages, 23 figures, including appendices
Abstract:Video object removal presents a uniquely difficult editing challenge. Because a removal prompt specifies only what to erase rather than what to generate, the model must infer and reconstruct a highly specific occluded background entirely from the surrounding context. Existing training-free methods struggle with this because their editing mechanisms act primarily as localized erasers. They fail to actively synthesize the missing background details and often leave behind ghosting artifacts. To solve this, we propose TripleFlow, a training-free framework that tightly couples erasure and generation. It coordinates a source flow, a residual flow, and a synthesis flow throughout the entire process. By reusing a single target prediction, the residual flow isolates and suppresses the object, while the synthesis flow independently reconstructs the occluded background. Crucially, TripleFlow injects this newly synthesized background back into the editing trajectory at every step. This continuous feedback loop ensures that the generated structures actively guide the removal process, achieving seamless completion that is spatiotemporally consistent with the unedited scene. Extensive evaluations across five challenging benchmarks demonstrate that TripleFlow establishes a new state-of-the-art, significantly outperforming existing baselines in both reconstruction fidelity and temporal consistency.
[CV-106] OP-CAD: On-Policy Clean-Audio Distillation for Robust Audio-Visual Reasoning
链接: https://arxiv.org/abs/2609.39150
作者: Xingming Shui,Dapeng Chen,Bowei Liu,Jingqi Tian,Minfu Li,Kun Yi,Jiapeng Hong,Yansong Tang
类目: Computer Vision and Pattern Recognition (cs.CV); Sound (cs.SD)
备注:
Abstract:Omni-modal large language models deployed in real-world environments encounter external noise that can interfere with their perception and understanding of multimodal inputs. We study their robustness in audio-visual understanding, focusing on question answering under environmental noise and competing speech. The challenge is to resist acoustic interference while preserving useful audio evidence. On-policy distillation provides dense teacher feedback on student-generated responses, but uniform token weighting does not explicitly prioritize positions affected by acoustic interference. We introduce OP-CAD (On-Policy Clean-Audio Distillation), a curriculum-based privileged self-distillation framework for robust audio-visual understanding. Training progresses from mild to severe environmental noise and competing speech, with selective token-level supervision at each stage. The student generates responses from corrupted audio-visual input, while a frozen teacher uses clean audio and the verified answer to supervise the same response prefixes. To allocate this supervision, OP-CAD compares teacher predictions under clean, corrupted, and visual-only contexts without revealing the answer. These matched comparisons measure sensitivity to audio removal and corruption; a bounded weighting rule emphasizes positions identified by either signal while retaining supervision throughout the response. OP-CAD outperforms the compared methods across all evaluated noise conditions. Paired analyses further show improved preservation of clean-correct answers under strong interference, with no observed aggregate clean-accuracy penalty. These results demonstrate the value of directing clean-teacher supervision toward acoustically sensitive predictions for robust audio-visual reasoning.
[CV-107] MindWorldBench: Evaluating Mental-State-to-Behavior Reasoning in Image-to-Video Generation
链接: https://arxiv.org/abs/2609.39147
作者: Ruiqi Li,Xuanyi Liu,Sijia Li,Haofeng Wang,Yuxin Liu,Feng Xie,Songchao Tan,Shiqi Wang,Hanwei Zhu,Yizong Wang,Chuanmin Jia,Siwei Ma
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 7 pages, 6 figures. Accepted by ACM Multimedia 2026
Abstract:Current image-to-video models achieve visual realism and physical plausibility, but reasoning about mental states remains unexplored. Actions are driven by belief, desire, and perception, requiring inference beyond explicit instructions. We introduce MindWorldBench to evaluate mental-state-conditioned video generation. We formalize this as mental-state-to-behavior reasoning, where models generate actions from a world state and latent variables without explicit action prompts. MindWorldBench utilizes Zero-Action Prompting and a counterfactual design with 744 prompts to isolate the causal effects of mental states. An automated pipeline evaluates video quality, commonsense plausibility, and mental-state consistency. Evaluations of 11 models show that despite visual fidelity and physical reasoning, models fail to align behaviors with latent mental states. We identify a failure mode, termed Omniscient Bias, where models default to the objective world state rather than human’s subjective belief. These results demonstrate a disconnect between visual generation and cognitive reasoning, suggesting a need for explicit mental-state modeling in video generation systems. Project website: this https URL
[CV-108] When Can Text Replace Vision? Structural Bottlenecks in Diagram Reasoning
链接: https://arxiv.org/abs/2609.39142
作者: Yunbei Zhang,Janet Wang,Jihun Hamm,Chandan K Reddy
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 33 pages, 16 figures. Code: this https URL
Abstract:Can structured text replace vision for diagram reasoning? A wrong answer after textualization can arise because the representation omits information the question needs, or because the solver fails to use information that is present. We introduce a diagnostic protocol to distinguish these explanations. Using the same solver model and generation settings, we compare three input conditions: the original image, question-blind structure extracted by a vision-language model, or gold structure derived from the diagram source. Validity-triggered recovery tests truncation and schema failure, question-relevant fidelity measures preservation of answer-critical structure, and matched edge interventions test the effect of error location. On a reserved holdout of 240 public FlowGen diagrams, evaluated under a frozen protocol, gold structure reaches 87% accuracy while direct vision and learned text both remain below 30%. The aggregate comparison includes source-derived relation labels that may not be printed in the image and uses different learned and gold graph encodings, so it does not isolate extraction error alone. Retrying only invalid extractions makes nearly every public representation schema-valid yet leaves accuracy essentially unchanged. The public learned-text deficit relative to gold more than doubles with structural difficulty. Question-relevant topology predicts correctness better than whole-graph topology. In an exposed intervention study, a single answer-relevant edge edit reduces the primary solver’s original-answer accuracy to near zero, while matched irrelevant edits largely preserve it. Supplied structure requires fewer solving tokens than vision, but learned acquisition removes this advantage at single use. These comparisons motivate evaluating acquired text by the answer-relevant evidence it preserves and by the solver’s ability to use that representation.
[CV-109] Asking the World: Generalist Physical Reasoning through Agent ic World Modeling and Probing
链接: https://arxiv.org/abs/2609.39135
作者: Shenxiang Zeng,Chen Yang,Peiyao Chen,Guohui Zhang,Jiansheng Fan,Chen Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages, 9 figures, 6 tables
Abstract:Physical reasoning from video requires inferring latent physical properties and dynamics beyond direct observation. Direct VLM inference remains unreliable on complex physical tasks without explicit modeling and validation, while predefined tool pipelines rely on task- and domain-specific priors that limit generalization across materials, dynamics, and reasoning tasks. We introduce Asking the World (ATW), a generalist agent that constructs and interrogates task-relevant executable worlds through two adaptive stages: World Modeling calibrates a world from video, while World Probing queries, simulates, and intervenes on it to obtain question-relevant evidence. Rather than prescribing the operations in either stage, ATW determines how to model and probe according to the scene and question. We develop PolyWorld Engine, a lightweight and highly programmable Warp-based multiphysics simulator for constructing and probing worlds with rigid bodies, soft bodies, cloth, ropes, fluids, and their coupled interactions. CEM-based system identification recovers task-relevant dynamics during World Modeling. The resulting world becomes an active workspace for question-directed physical experiments rather than a predetermined downstream tool. We evaluate ATW on CLEVRER, ContPhy, and three real-world scenarios. Using Gemini-3-Flash as its base VLM, ATW achieves 80.82% overall per-question accuracy on CLEVRER, improving direct Gemini-3-Flash by 46.50 points, GPT-5.5 by 13.58 points, and PhysMind by 8.27 points. On ContPhy, it reaches 70.56% overall accuracy, surpassing Gemini-3-Flash by 28.10 points and GPT-5.5 by 3.53 points. Across the three real-world scenarios, ATW achieves 71.67% accuracy, 28.33 points above GPT-5.5. These results establish agentic world modeling and probing as an effective, execution-grounded approach to generalist physical reasoning.
[CV-110] Feature-Aware Token Attack for Compression-Triggered Stealthy Failures in Large Vision-Language Models ICLR2027
链接: https://arxiv.org/abs/2609.39134
作者: Shilinlu Yan,Bowen Chen,Yuechen Zhang,Zhenhong Zhou,Li Sun,Sen Su
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 29 pages including references and appendices, 10 figures. Submitted to ICLR 2027
Abstract:Visual-token compression improves the efficiency of large vision-language models, but can expose failures that full-token evaluation misses. We study adversarial images that preserve full-token correctness yet induce errors after compression, even when both inference paths succeed on the clean image. Creating such failures is challenging because perturbing token importance can also damage the visual content needed for full-token inference. We propose Feature-Aware Token Attack (FATA), which couples attention suppression with cosine-based feature preservation on a fixed set of salient clean-image tokens. In the primary LLaVA-1.5-7B setting, FATA uses only vision-encoder gradients, without access to the deployed compressor, token budget, or downstream task. Across four visually dependent task subsets and four compressors under a controlled reconstruction protocol, FATA achieves SR = 96.3% full-token accuracy retention and CBR = 22.1% conditional blinding, compared with 89.8% and 15.7% for CAA. Ablations support the role of both objectives in balancing compressed-path failure against full-token preservation. FATA also has the lowest measured detection rate among four attacks across three evaluated detectors at a 5% false-positive rate. These findings motivate assessing adversarial robustness jointly across full-token and compressed inference.
[CV-111] Uncertainty-Aware Consistency Distillation for Few-Step Video Generation
链接: https://arxiv.org/abs/2609.39132
作者: Lingyu Liu,Yaxiong Wang,Li Zhu,Zhedong Zheng
类目: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
备注:
Abstract:We study few-step video generation, i.e., distilling a multi-step video generator, which typically requires tens of sampling steps, incurring substantial latency and compute, into a few-step student. Consistency distillation is a common recipe, in which a multi-step teacher provides the consistency targets for a few-step student. However, these teacher-guided targets are not equally trustworthy, and the content is harder to learn where it varies rapidly over time, e.g., moving foliage shadows or flowing water. We observe that supervision reliability follows the local difficulty of the content rather than semantic complexity: regions that change little yield consistent endpoint predictions, whereas regions with large temporal variation produce larger discrepancies that coincide with the largest perceptual errors. Motivated by this observation, we propose Uncertainty-Aware Consistency Distillation (UACD), which reweights consistency supervision at each spatiotemporal region using a local, parameter-free uncertainty estimate. Specifically, we construct two independently perturbed teacher-guided consistency paths, whose student endpoint predictions provide a consensus target; the discrepancy between the student’s direct prediction and this target is the uncertainty proxy. We then relax the consistency penalty on high-uncertainty regions through an exponential weight, while keeping the full penalty elsewhere, since the student cannot be expected to match targets that are hard to learn. To preserve perceptual quality under aggressive step reduction, we integrate feature-space adversarial training with semantic alignment. With parameter-efficient LoRA adaptation of the 50-step Wan model, our method achieves state-of-the-art 4-step generation on VBench 2.0 (0.556 mean score) and is preferred over competing methods in a user study.
[CV-112] GeoGAT: Bidirectional Temporal Sampling Meets Hierarchical Graph Attention for Global Video Geo-localization
链接: https://arxiv.org/abs/2609.39128
作者: Junchao Cui,Xuanzi Ma,Wenqi Shi,Hangyu Li,Biru Zhu,Chong Fu,Xiangyang Luo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Global video geo-localization aims to infer the geographic location of a video worldwide, evaluating performance across four geographic hierarchies: city, state/province, country, and continent. Existing methods typically employ one-way uniform sampling to process video frames and train independent classifiers for each hierarchy, which leads to the loss of key geographic cues and prediction conflicts between hierarchies, especially for complex multi-shot edited videos. To address these limitations, we propose GeoGAT, which integrates bidirectional temporal sampling with graph attention networks (GATs). Specifically, GeoGAT extracts forward and offset-reversed frame sequences to construct complementary spatiotemporal features. These fused features are then fed into a predefined geographical hierarchy graph, where GATs perform structure-aware message passing, while a dual-constraint mechanism prunes predictions to eliminate cross-hierarchy conflicts. We construct GeoGAT10k, comprising 9,720 multi-shot edited videos from 166 cities worldwide, specifically to benchmark generalization ability on complex video structures. Experimental results on CityGuessr68k and GeoGAT10k demonstrate that GeoGAT eliminates hierarchical conflicts entirely and achieves state-of-the-art performance across all four geographic hierarchies. On CityGuessr68k, GeoGAT outperforms the strongest baseline, evaluated under both classification and retrieval protocols, by 2.6 percentage points at the city level. On the more challenging GeoGAT10k with multi-shot edited videos, the accuracy improvement exceeds 24 percentage points, validating strong generalization to complex real-world scenarios.
[CV-113] How to Reduce Localization Ambiguity? Geometry-Semantic Constrained BEV Representation Learning for Satellite-Ground Localization
链接: https://arxiv.org/abs/2609.39127
作者: Junming Feng,Panwang Xia,Qiong Wu,Xudong Lu,Zeyu Jiao,Kun Lv,Zherong Wu,Yi Wan,Peifeng Ma,Li-Ta Hsu,Zhi Zheng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 2 figures, and 4 tables
Abstract:Satellite-ground localization estimates the planar position and yaw orientation of a ground camera within a geo-referenced satellite image. Most recent methods map ground and satellite features into a shared bird’s-eye-view (BEV) space and establish spatial correspondences. However, insufficient depth constraints can assign one ground feature to different distances along a viewing direction, creating geometric ambiguity in BEV feature placement. Similar appearances at different locations can also create descriptor matching ambiguity, while existing descriptor learning lacks explicit semantic supervision to distinguish them. We propose GeoSem-BEV, a geometry-semantic constrained BEV representation learning method. Radial depth supervision constrains distance assignment, and vertical height supervision constrains height aggregation. Shared explicit semantic supervision promotes consistent semantic predictions across views and helps distinguish locations with similar semantics. These constraints improve feature placement and descriptor discriminability, enhancing state-of-the-art BEV localization models. On VIGOR with unknown orientation, GeoSem-BEV reduces mean orientation error by 37.2% and 38.1% in the cross-area and same-area settings, respectively. The corresponding errors are reduced by 10.8% and 15.6% on DReSS-D. On KITTI-CVL, it reduces same-area mean orientation error by 26.8% under 10 degree orientation noise.
[CV-114] Is Better Teacher Supervision Enough? Unlocking Student-side Learning in Multimodal On-Policy Distillation
链接: https://arxiv.org/abs/2609.39120
作者: Siyuan Liu,Kanghui Tian,Yue Duan,Yutao He,Shangdong Yang,Jian Zhang,Yinghuan Shi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:On-policy distillation (OPD) improves reasoning by providing token-level supervision from a teacher on a student’s own trajectories. Existing methods primarily focus on enhancing this teacher-side guidance (e.g., by enriching teacher inputs and refining teacher feedback), yet we find that limited student perception is another critical bottleneck in multimodal OPD. By providing oracle visual facts, the performance of OPD-trained students can still be substantially improved for both weak and strong teachers. To address this bottleneck, we propose S-OPD, a simple multimodal on-policy distillation framework that explicitly strengthens student perceptual learning through two objectives. Specifically, Teacher-calibrated Policy Contrast separates student policies under original and masked images with teacher-based token-level gating, strengthening the student’s reliance on visual evidence during reasoning. Policy Agreement aligns student policies under original and noise-perturbed images, further improving perceptual robustness to visual noise. Notably, our method can be seamlessly plugged into existing OPD frameworks, requiring no additional data annotations, model parameters or inference operations. Extensive experiments on eight benchmarks across student scales and distillation paradigms demonstrate consistent performance improvements, with gains of up to 4.25 points on LogicVista. When combined with existing teacher-side supervision methods, our method can yield further gains. Code is available at this https URL.
[CV-115] GRC-Pose: Generation-Reconstruction Correspondence for Prior-Free 6D Object Pose Tracking
链接: https://arxiv.org/abs/2609.39116
作者: Shiyang Liu,Weiquan Lin,Luping Xiao,Jiadong Tang,Yi Yang,Yu Gao,Xingyu Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 39 pages
Abstract:Prior-free 6D object pose tracking seeks to recover the trajectory of an unseen object from a single RGB video without object-specific CAD models, posed reference images, or pose annotations. Geometric foundation models provide complementary object-centric and scene-centric cues, yet SAM3D CAD is indexed by an arbitrary object-local surface parameterization, whereas reconstructed evidence is expressed in a sequence-specific world frame with partial surface coverage. To exploit this complementarity, we formulate tracking as generation-reconstruction correspondence and introduce GRC-Pose, a correspondence-based framework that combines learned correspondence prediction with robust pose estimation. Concretely, GeoCorr-Matcher estimates weighted object-scene correspondences and per-match uncertainty for each pose candidate. FGH-Solver integrates these matches through multiple robust geometric estimators and sequence-level posterior inference, while a posterior-gated memory retains only inlier-supported observations through occlusion and viewpoint change. Extensive evaluation shows that with SAM3D CAD, GRC-Pose achieves state-of-the-art Average Recall and motion retention on HOT3D, improving the latter by 58% over prior art. On classical benchmarks including YCBInEOAT and LINEMOD, it remains highly competitive.
[CV-116] Beyond Local Linearity: Scale-Resolved Geometry of Learned Image Encoders NEURIPS2026
链接: https://arxiv.org/abs/2609.39115
作者: Jakub Szymkowiak,Wojtek Pałubicki,Kamil Adamczewski
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Extended abstract, NeurIPS 2026 Workshop on Symmetry and Geometry in Neural Representations (NeurReps). 14 pages, 5 figures
Abstract:Understanding how learned representations respond to finite input changes is important for characterizing their sensitivity, invariances, and robustness. Yet existing geometric analyses are predominantly local and describe only infinitesimal perturbations. We introduce a scale-resolved statistic that compares an encoder’s measured feature displacement with its local linear prediction as the perturbation magnitude increases. Across diverse image encoders, we discover a characteristic plateau-rise-peak-decay profile, which we call the bump. The bump is absent at initialization, emerges early during standard training, and does not form under randomized labels or random-noise inputs. Its shape also varies with the training distribution and robustness objective. These results establish departures from local geometry as a signature of how encoder representations are shaped by learning.
[CV-117] CamAgent : An LLM -Agent Framework for Multi-Species Camera-Trap Workflows
链接: https://arxiv.org/abs/2609.39112
作者: Yutong Deng,Qi Song,Xi Guo,Tianming Wang,Lei Bao,Jianping Ge
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Camera traps accumulated vast, multidimensional data for wildlife monitoring, yet translating raw media archives into meaningful ecological insights remains highly fragmented. Current research workflows require laboriously stitching together disparate analysis tools and scripts, creating steep programming hurdles and complicating end-to-end spatiotemporal analyses. To overcome this fragmentation, we present CamAgent, an autonomous Large Language Model (LLM) agent framework that integrates camera-trap analytical workflows into a unified intelligent ecosystem. CamAgent interprets natural-language ecological intent, schedules computational routing, and executes specialized tools spanning computer-vision perception (e.g., SpeciesNet), CamtrapDP-compatible data management, detection-corrected occupancy modeling, temporal activity analysis, and species co-occurrence networks. The framework automates multi-stage analytical pipelines while maintaining essential data-quality controls and analytical conventions. Consequently, CamAgent significantly reduces manual programming overhead for conservationists, establishing a transparent, scalable, and fully integrated paradigm for camera-trap ecology. Our project is available at this https URL.
[CV-118] DeCoPrune: Efficient KV-Cache Pruning for Autoregressive Video Diffusion via Denoising Consistency
链接: https://arxiv.org/abs/2609.39096
作者: Zeqi Xiao,Qingle Liu,Kaiwen Zhang,Yifan Zhou,Zihan Ding,Xingang Pan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Autoregressive video diffusion supports streaming generation and interactive control, but its KV cache grows with the generated history. Existing compression strategies discard history using fixed windows or select tokens through local attention and similarity signals, without directly measuring whether a chunk contributes information beyond the retained context. We introduce DeCoPrune, a training-free method that treats cache compression as a denoising-consistency problem. We find that tokens with larger discrepancies between intermediate clean predictions and final denoised values tend to carry visual evidence less predictable from the retained context. DeCoPrune uses this model-intrinsic signal to retain high-discrepancy tokens in the long-term cache while pruning low-discrepancy tokens. To evaluate information retention, we introduce CMBench, comprising 58 approximately one-minute generated or real-world context episodes and 116 Reappear or Revisit continuation tasks requiring recall of earlier events or objects. Experiments with LingBot World v2 show that DeCoPrune achieves a DINO score of 0.6701 on a 0-1 scale, with an 85.43% reduction in cumulative historical KV token counts and a 4.14-fold continuation-generation speedup over FullKV. Its head-specialized variant reaches 0.6783 at an 86.19% pruning ratio, approaching FullKV’s 0.6803 score and exceeding the evaluated compression baselines at similar budgets. These results indicate that denoising consistency can support long-range information retention while reducing autoregressive inference cost. Our project homepage is this https URL. The code is available at this https URL, and the benchmark at this https URL.
[CV-119] UGOD: Uncertainty-Guided Opacity and Dropout for Sparse-View 3D Gaussian Splatting
链接: https://arxiv.org/abs/2609.39089
作者: Zhihao Guo,Peng Wang,Zidong Chen,Xiangyu Kong,Yan Lyu,Guanyu Gao,Chenghao Qian,Ziyang Wang,Xinqi Fan,Liangxiu Han
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 31 pages, 5 figures, 10 tables. Supplementary material included at the end of the manuscript
Abstract:Sparse-view 3D Gaussian Splatting is prone to overfitting because limited observations leave many Gaussian primitives weakly constrained, yet their contributions are still accumulated through alpha blending. Without uncertainty estimation, the renderer cannot distinguish unreliable primitives from well-constrained ones, allowing their erroneous contributions to corrupt novel-view synthesis. We introduce UGOD, an uncertainty-guided framework that estimates a view-dependent uncertainty score for each Gaussian and uses it to regulate its rendering contribution. A lightweight uncertainty head conditioned on Gaussian attributes and viewing direction predicts this score, which then drives a differentiable opacity-modulation mechanism that attenuates high-uncertainty primitives before compositing. During training, a detached soft-dropout branch applies an uncertainty-controlled continuous keep mask to discourage the model from relying on poorly constrained Gaussians and thereby reduce overfitting. Crucially, detaching the uncertainty score prevents gradients from this stochastic regulariser from biasing or collapsing the uncertainty prediction. Experiments on Mip-NeRF~360 and LLFF show that UGOD improves sparse-view novel-view synthesis while producing more compact Gaussian representations than the compared methods. These results demonstrate that Gaussian uncertainty provides an effective rendering-time control for sparse-view reconstruction.
[CV-120] MRI Super-Resolution with RCDM/WaveMix and Task-Aware Segmentation
链接: https://arxiv.org/abs/2609.39083
作者: Kavitha Viswanathan,Harsh Choudhary,Amit Sethi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Super-resolution and quality enhancement of 1.5,T brain MRI are normally validated with image-fidelity metrics, although their purpose is to improve downstream analysis. We study whether enhancement improves tissue segmentation, and for which segmenters. We propose an unpaired, physics-guided training pipeline for a lightweight ( \le 2.5,M parameter) recurrent convolutional enhancer: a six-module stochastic 1.5,T degradation operator, a residual adversarial network that adds scanner-specific texture without moving anatomy, and a cycle-consistent objective with an anti-identity penalty that rules out the copy solution. We then train U-Net, Swin-UNet and wavelet token-mixing segmenters \citepjeevan2023wavemix from scratch on either raw or enhanced 1.5,T images of the same subjects, using identical labels and subject-level splits, for three enhancer variants and two datasets. On ABIDE (41 held-out subjects, FreeSurfer labels) enhancement significantly improves the wavelet segmenter (mean Dice +0.014 , Wilcoxon p=3.5\times10^-5 ; CSF +0.018 , grey matter +0.013 ), significantly degrades the U-Net ( -0.008 , p=5.1\times10^-4 ) and leaves Swin-UNet unchanged. On IXI, whose labels come from FSL-FAST, enhancement lowers Dice for all nine pairings, almost entirely through CSF; we trace this to spatially implausible CSF voxels in the labels that penalise smoother predictions. Enhancement of low-field MRI should therefore be validated per downstream model and against reliable labels.
[CV-121] Agent ic Tool-Augmented Reasoning for Explainable Image Forgery Detection
链接: https://arxiv.org/abs/2609.39066
作者: Zhiya Tan,Jing Huang,Changtao Miao,Lin Tan,Xin Zhang,Weiwei Feng,Jianshu Li,Joey Tianyi Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at ACM Multimedia 2026 (Oral)
Abstract:Conventional image forgery detection methods produce binary scores or pixel-level masks without interpretable evidence, while recent multimodal large language model (MLLM)-based approaches generate post-hoc explanations of predetermined classification results rather than reasoning from evidence. Inspired by the forensic workflow of human judicial experts, we propose Agentic Tool-Augmented Reasoning (ATAR), a framework integrating 22 specialized forensic tools across seven complementary domains to autonomously detect, localize, and explain image forgeries through multi-turn reasoning. A Dual-Stream Forensic Reasoning paradigm combines a high-level semantic anomaly path, which magnifies suspicious regions for fine-grained inspection, with a low-level forgery artifact path, which invokes forensic tools to extract objective evidence. We further introduce Forensics Curriculum Learning: during General Experience SFT, an automated teacher-student mentoring pipeline synthesizes multi-turn tool-usage reasoning trajectories; during Forensic Scene RL, a Tool Prior Curriculum guides early tool exploration and progressively transfers control to the agent, while a Structured Evidence Reward provides fine-grained process-level supervision. Experiments across IMDL, Deepfake detection, DMDL, and AIGC detection show that ATAR achieves 78.5% average image-level F1 on six zero-shot IMDL benchmarks, surpassing the strongest MLLM baseline by 11.8 percentage points, and remains competitive with specialized detectors on other tasks while producing substantially more faithful and grounded explanations.
[CV-122] SMD: Temporal-Stream Modality Dropout for Robust Video Highlight Detection
链接: https://arxiv.org/abs/2609.39051
作者: Bo-Yuan Cheng,Kuan-Yu Chen,Po-Han Huang,Jeng-Lin Li,Jian-Jiun Ding
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 3 figures, 3 tables
Abstract:Existing multimodal video highlight detectors typically assume that visual, audio, and textual streams are continuously available. In practice, however, inputs may suffer from localized frame missingness or complete-stream outage. We formulate this robustness challenge along two dimensions: temporal missingness, where frames are missing independently in each modality, and stream-level missingness, where one modality is unavailable throughout a video. Moreover, we find that the mean squared error (MSE) loss is misaligned with both the evaluation metrics and the peak-driven nature of highlights. Therefore, we propose Temporal-Stream Modality Dropout (TSMD), which combines structured missingness simulation with a joint objective comprising pointwise MSE, per-video Pearson correlation, and peak-oriented RankNet loss terms. TSMD has three variants: temporal, stream-level, and mixed dropout. On the MoSu and Mr. HiSum datasets, TSMD-Temporal improves mAP@15 by 7.06 and 3.41 points over TripleSumm under 50% independent temporal removal, whereas TSMD-Stream performs the best under complete-stream removal. TSMD-Mix retains most of these complementary benefits and ranks the best or the second-best across the evaluated temporal and stream-level conditions.
[CV-123] BadAction: Backdoor Attacks on Interactive Video Generation via Action-Guided Triggers
链接: https://arxiv.org/abs/2609.39047
作者: Zhihang Wu,Zhongqi Wang,Jie Zhang,Fengming Gu,Shiguang Shan,Xilin Chen
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 12 pages, 6 figures, 5 tables. Project page: this https URL
Abstract:Interactive video generation (IVG) models have achieved remarkable progress in producing controllable visual content guided by user-defined actions, yet their security vulnerabilities remain largely unexplored. In this paper, we present the first systematic study of backdoor attacks against the interactivity of IVG models. Based on this attack surface, we propose BadAction, which leverages action-guided triggers to achieve the attack. Specifically, BadAction implants predefined motion patterns into the action sequences of backdoor samples and associates them with a static target video. Once triggered, the backdoored model generates frozen future frames that no longer respond to subsequent user actions, while preserving normal behavior on benign action sequences. In addition, we explore a stealthier attack in which multimodal triggers jointly poison action, text, and image inputs. Experiments show that BadAction achieves average attack success rates of 91.0% with action-only triggers and 80.4% with multimodal triggers. Moreover, extensive defense evaluations show that BadAction successfully bypasses existing backdoor detection methods, revealing a critical security gap in the interactive video generation pipeline. Project page: this https URL.
[CV-124] ED:Text-Axis Evidence Decomposition for Prompted Anomaly Localization NEURIPS2026
链接: https://arxiv.org/abs/2609.39033
作者: JinYoung Kim,Geonho Kim,GiJeong Park,Geonu Lee,YoungJoon Yoo
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 40th Conference on Neural Information Processing Systems (NeurIPS 2026)
Abstract:CLIP is a powerful vision-language model, but it was not designed for fine-grained defect localization; CLIP-based anomaly detectors therefore adapt it with prompts or lightweight modules to increase defect sensitivity. We show that stronger sensitivity does not necessarily make local evidence reliable: under domain shift, adapted CLIP-AD models often assign high anomaly scores to both true defects and visually complex normal regions. The issue is not simply missing defect information, but a local scoring rule that decodes defect and hard-normal evidence, having the same anomaly evidence. We propose TED (Text-Axis Evidence Decomposition), a post-hoc scoring method that asks whether each ambiguous response is better supported by source defect patches or by source normal patches mistaken as anomalous. TED compares these supports under the host’s normal-versus-anomaly text response, leaves the backbone and prompts unchanged, and requires no target-domain training. It works as a train-free score for raw VLM backbones or as a source-calibrated residual correction for adapted CLIP-AD hosts. Across frozen VLM backbones, TED substantially improves pixel-level localization over raw prompt similarity; across adapted hosts, it improves most pixel-level settings over P-AUROC, P-PRO, and P-AP. Gains are largest under stronger hard-FP competition, with mean localization gain increasing from +5.0 in low-competition regimes to about +10.9 in mid/high-competition regimes. These results suggest that recoverable defect evidence can already exist in pretrained multimodal representations, but reliable localization requires decoding it against hard-normal competitors. Code will be released at TED GitHub repository.
[CV-125] Persistent Watermarking of Text-to-Image Models
链接: https://arxiv.org/abs/2609.39024
作者: Dixi Yao,Kaiwen Chen,Tahseen Rabbani,Tian Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Text-to-image (T2I) generation is gaining increasing popularity with the general public, motivating the development of reliable mechanisms for copyrighting such models given their expensive training costs. An adversary may obtain and reuse a pretrained T2I model without authorization, and then serve a modified version through an API service. Such modifications may arise from ordinary downstream adaptation or deliberate attempts to erase ownership, including input-prompt preprocessing, model fine-tuning, and output post-processing. From the model owner’s perspective, a key challenge is therefore to embed trigger data that remain persistent under such changes while preserving the model’s normal image-generation capabilities. In this work, we propose a contrastive-style watermarking objective with a term that explicitly encourages the watermarked model to behave differently from the original model on trigger inputs. Experiments show substantially stronger trigger-data persistence than prior methods across a wide range of downstream modifications and deliberate attempts to weaken the watermark, resulting in higher detection rates, often approaching 100% TPR@FPR 10^-4 .
[CV-126] Frame Differential On-Policy Self-Distillation for Video Reasoning
链接: https://arxiv.org/abs/2609.39021
作者: Haiying He,Xin Zheng,Shaoli Hu,Shijun Xiao,Xuanhe Liu,Bing Li,Harry Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reinforcement learning (RL) has substantially improved the reasoning ability of multimodal language models through verifiable rewards and increasingly fine-grainedvisual or temporal credit assignment. In video reasoning, however, current RL methods typically train with a fixed sparse frame budget: increasing the number of frames makes autoregressive rollouts expensive, while too few frames may miss temporally localized events and fine-grained visual details. We present \textbfFrame Differential On-Policy Self-Distillation (FD-OPSD), which transfers the useful evidence of dense frame observations to a sparse frame policy during RL training. FD-OPSD compares the policy’s token level preferences for the same sampled response under sparse and dense views, and distills the resulting frame differential signal without an external teacher or dense autoregressive rollout. The method preserves sparse-frame rollouts and leaves inference unchanged. Across Qwen2.5-VL-7B and Qwen3-VL-4B on six video reasoning benchmarks, FD-OPSD yields higher overall average performance than the strongest corresponding GRPO, T-GRPO, or Video-KTR baselines across the 16, 32, and 64 frame evaluation settings. These results show that dense visual evidence can be transferred selectively during training through token level self-distillation while retaining sparse frame rollouts and unchanged inference.
[CV-127] When Integral Meets Decomposition: A Signal-Level Self-Supervised Feature Decompose Paradigm for Multi-Modal Image Fusion
链接: https://arxiv.org/abs/2609.39004
作者: Zeyu Wang,Jiayu Wang,Haiyu Song,Haoran Duan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multimodal image fusion (MMIF) aims to integrate complementary information from different modalities into a high-quality fused image and support downstream tasks. Recently, feature decomposition has become an important paradigm by separating source images into common and modality-specific unique features. However, existing methods lack clear supervision because ground-truth (GT) decomposition feature maps are unavailable. They usually combine multiple image-level metrics as losses, which are inherently incomplete and may conflict since each pixel couples attributes such as texture, edge, and contour. To address this, we propose a 1D signal-level self-supervised feature decomposition paradigm. Our core insight is to reformulate feature decomposition from unclear 2D image-level supervision into an integral-driven 1D signal-level optimization problem. This objective-level reformulation uses the 1D signal form to compute the integral constraint. The decomposer is optimized by the integral area between common and original signals, enabling more stable optimization with a clear optimization objective. Our model follows a two-stage SSL framework. Stage I designs dual pretext tasks for integral-driven decomposition at the signal level and structure-preserving reconstruction at the image level. Stage II fuses unique features and combines them with common features to reconstruct the fused image. Experiments on representative MMIF tasks show state-of-the-art (SOTA) performance. Code: this http URL.
[CV-128] Anchoring Adversarial Trajectories to Data Manifolds: A Bilevel Transfer Optimization Framework NEURIPS2026
链接: https://arxiv.org/abs/2609.38991
作者: Yaohua Liu,Yifan Guo,Jiaxin Gao
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at NeurIPS 2026 as a Spotlight. 21 pages, 6 figures
Abstract:A key bottleneck in adversarial transfer is a trajectory-level geometric disconnect: ambient gradients often drift away from the intrinsic data manifold, causing surrogate-specific overfitting. To rectify this, we propose Manifold Anchored Bilevel Transfer (MABT), a unified framework that anchors adversarial trajectories to the shared semantic subspace. MABT introduces a relaxed manifold-anchoring operator as a semantic rectifier to suppress off-manifold noise. With this constraint, we cast transfer attack generation as a distributional bilevel optimization problem that learns a geometry-aligned initialization by minimizing expected transfer risk under a surrogate uncertainty distribution. We further develop a Hessian-free solver with linear-time complexity to handle the resulting hierarchy. Experiments demonstrate improved transferability for 10 baseline attackers across 28 attack configurations, diverse victim architectures, and defense mechanisms.
[CV-129] MeshOctave generates meshes via cascading resolution transitions
链接: https://arxiv.org/abs/2609.38985
作者: Junkai Lin,Tianhao Zhao,Hang Long,Huipeng Guo,Jielei Zhang,Youjia Zhang,Jiale Xu,Wenbing Li,Rendong Liang,Jozef Hladký,Matthias Nießner,Yuanming Hu,Wei Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages
Abstract:Generating compact, artist-style meshes with explicit topology typically relies on autoregressive models which incur prohibitive sequential per-token costs, or continuous flow models that depend on heuristic connectivity decoders. Next-scale generation paradigms offer a compelling alternative by enabling parallel intra-scale token prediction and coarse-to-fine refinement from global structure to local topology; yet, existing methods derive hierarchical scales via progressive mesh simplification and invert them sequentially. This eliminates intra-scale parallelism and scales generation steps linearly with face count. In this paper, we propose MeshOctave, which instead defines scale through dyadic spatial grid resolutions, framing coarsening as a deterministic collapse that merges vertices sharing a voxel cell and inherits connectivity. Its inverse operation, split-and-rewire, determines which octant sub-vertices are instantiated for each coarse face and resolves local connectivity using discrete structural tokens. These per-face operations require no serialization, each scale transition is modeled as an unordered set that adds one bit of coordinate precision, naturally supporting dynamic-length meshes and adaptive resolution refinement. We construct a scale-conditioned masked-uniform discrete diffusion model to learn split-and-rewire operation from resolution collapse hierarchies. MeshOctave outperforms strong baselines in geometric fidelity and topological validity by a non-trivial margin, while supporting adaptive resolution refinement and extending naturally to mesh subdivision tasks.
[CV-130] Mitigating Object Hallucination in Large Vision-Language Models via False Discovery Controlled Visual Data Splitting
链接: https://arxiv.org/abs/2609.38979
作者: Chang Liu,Yu Tian,Rui Xie
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multiple object hallucination, where large vision-language models (LVLMs) generate objects not supported by the visual input, is a persistent challenge caused by visual uncertainty during decoding. Existing methods reduce hallucinations using contrastive signals, but they rely on heuristics and lack principled control of false positives at the image level. To address this, we propose False Discovery Rate-COntRol of HALlucination (CORAL), a training-free framework that models visual uncertainty using an uncertainty-aware visual data splitting strategy and leverages mirror statistics to quantify visual contrast during decoding. By computing mirror statistics from paired, symmetrically perturbed visual inputs, CORAL estimates spurious object predictions and sets a data-driven threshold to control the expected fraction of false discoveries per image, suppressing hallucinations while retaining high power for truly grounded objects. The framework is flexible, supports multiple LVLMs, and mitigates hallucinations without retraining or supervision. Extensive experiments on multiple benchmarks with several evaluation metrics demonstrate that CORAL consistently outperforms state-of-the-art methods, providing more reliable and robust hallucination control. Code is available at: this https URL
[CV-131] PARK: Accurate Block Retrieval for Sparse Attention in Video Diffusion Transformers
链接: https://arxiv.org/abs/2609.38978
作者: Yun Dai,Jiarui Wen,Huiping Zhuang,Cen Chen,Ziqian Zeng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Diffusion Transformers (DiTs) have become a dominant architecture for video generation, but their efficiency is limited by the quadratic complexity of full attention. Sparse attention reduces this cost by retrieving important blocks and computing attention only within them, but inaccurate retrieval can either degrade generation quality or yield unnecessary computation. We identify two retrieval mismatches in methods that retrieve blocks using the averaged representations of query and key blocks: (i) query-side aggregation mismatch, where averaging queries before Softmax fails to preserve their individual attention preferences, and (ii) key-side clustering metric mismatch, where standard Euclidean clustering in the original key space can group keys with dissimilar QK scores under the current query, so their average representation may not accurately represent how the current query scores individual keys. These mismatches can lead to inaccurate block retrieval. To address these mismatches, we propose PARK, a training-free sparse attention method for accurate block retrieval. PARK retains every original query, independently normalizes its attention over key blocks, and then averages these distributions within each query block. It also uses information from the current queries to transform keys before clustering, so that keys receiving similar QK scores are grouped together. A fused GPU kernel further reduces the overhead of block retrieval. Experiments on HunyuanVideo and Wan demonstrate that PARK improves block retrieval accuracy and preserves generation quality while accelerating inference, achieving the best quality-efficiency trade-off among the compared sparse attention methods.
[CV-132] Beyond Spatial-Domain Supervision: A Relation Constrained Space for Multi-Modal Image Fusion NEURIPS2026
链接: https://arxiv.org/abs/2609.38968
作者: Zeyu Wang,Mingyu Ge,Haiyu Song,Haoran Duan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to NeurIPS 2026
Abstract:Multi-modal image fusion (MMIF) aims to form a single image by integrating shared information, preserving complementary cues, and coordinating cross-modal conflicts across modalities. However, due to the absence of ground-truth fused images, existing MMIF supervision commonly uses spatial-domain sources or gradient variants as surrogate ground truth, making the supervision mechanism inherently misaligned with the goal of MMIF and causing pixel-level compromise or modality bias. To address this, we propose a relation-constrained supervision paradigm that moves fusion supervision from the spatial domain to a learned relation space. Rather than relying solely on direct source approximation, we further leverage frozen pretrained representation models as information providers and design a learnable feature adapter to align heterogeneous DINO and CLIP features into a unified supervision space. The adapter infers three relation parameters, namely sharedness, dominance, and coordination radius, which define three losses corresponding to the MMIF’s goal. To make this space reliable, we devise a self-supervised contrastive ranking objective tailored to the adapter and couple it with the fusion network through alternating optimization. Extensive experiments show that the proposed supervision space yields significant gains regardless of which mainstream backbone the fusion network adopts, offering a supervision paradigm better aligned with the goal of MMIF. Code: this http URL.
[CV-133] On the Relaxation of Conditional Independence Assumption for Image Segmentation
链接: https://arxiv.org/abs/2609.38930
作者: Zixun Wang,Ben Dai
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注:
Abstract:In semantic segmentation, a recent line of RankSEG methods directly optimizes Dice/IoU scores at inference time, improving alignment with evaluation metrics without modifying model training. Despite its theoretical and empirical success, RankSEG relies on the restrictive Conditional Independence Assumption (CIA), which ignores crucial label correlations and therefore degrades performance in ambiguous or low-contrast scenarios. However, accounting for full label dependence is computationally prohibitive, requiring \mathcalO(d^3) time. To address this, we replace the CIA with a Spatially Localized Dependence (SLD) structure that captures local label correlations while keeping the dependence model tractable. We further overcome the remaining computational bottleneck via a Reciprocal Moment Approximation coupled with a novel fixed-point optimization strategy that eliminates exhaustive search. The proposed algorithm achieves a highly practical \mathcalO(d \log d) complexity and consistently outperforms conventional argmax and CIA-based RankSEG across diverse segmentation benchmarks. Improvements are significant in low-contrast or small-object scenarios, where label dependence offers valuable signals complementary to image information for accurate segmentation. The code of experiments is available at this https URL.
[CV-134] World-as-Graph: Relational World Modeling Through Latent Space Graphs
链接: https://arxiv.org/abs/2609.38927
作者: Yaqi Yang,Shuo Huang,Yujin Huang,Fucai Ke,Jiatong Han,Xin Zheng
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: under review
Abstract:World models aim to learn representations of real-world environments and predict their future evolution. Recent object-centric world models have made expressive progress by representing visual scenes as sets of object-level latent states, but object-object relations are often captured only implicitly, which limits explicit relational and temporal structure modeling and object-centric dynamic memory modeling. To address such challenges, we propose World-As-Graph (WAG), a graph-based object-centric world model that introduces relational inductive bias into JEPA-style predictive representation learning. The proposed WAG contains two main modules: (1) Relation-aware structure induction, which constructs time-varying latent graphs from object-centric slots and designs relation-aware object masking policies to guide relational object representation learning in latent space; (2) Object-centric memory transition, which maintains and updates object-level dynamic states by combining relational information from neighboring objects with historical memory, enabling effective autoregressive future prediction. Extensive experiments on both visual reasoning and robotic manipulation tasks could demonstrate the superior performance of our proposed WAG.
[CV-135] From Image Interpretation to Clinical Reasoning : Upstream Physician-Context-Aware Multimodal Learning with Causal Reinforcement Learning
链接: https://arxiv.org/abs/2609.38924
作者: Jialu Pi,Yanan Ma,Weijie Chen,Owen Crystal,Shubham Trivedi,Stephen Xie,Anna Silverman,Matthew Stib,Chadi Ayoub,Reza Arsanjani,Imon Banerjee
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Major adverse cardiovascular events (MACE) remain the leading cause of mortality worldwide. Opportunistic screening using routinely acquired clinical data offers a scalable approach for identifying high-risk individuals before acute events occur. Although chest X-rays (CXRs) capture latent cardiovascular biomarkers and clinical histories provide complementary patient context, existing medical vision-language models are primarily optimized for radiology interpretation rather than prognostic reasoning. We propose a causal reinforcement learning framework for multimodal clinical reasoning that integrates CXRs and physician-authored clinical histories for opportunistic MACE prediction. The framework introduces (1) a role-decoupled dual-LLM architecture that separates reasoning from risk prediction, (2) a dual-action causal reinforcement learning policy for evidence selection and reasoning optimization, and (3) causal token pruning to learn compact multimodal representations. Evaluated on an internal cohort, an emergency department cohort, and the external MIMIC dataset, the proposed framework consistently outperformed unimodal baselines and state-of-the-art medical vision-language models, achieving AUROCs of 0.720, 0.760, and 0.845, respectively. It also substantially improved reasoning quality, achieving higher GREEN scores and higher expert preference while maintaining robust predictive performance across diverse patient populations.
[CV-136] FLOW: Feature-Level Optimal Warping for Generalized Remote Physiological Measurement
链接: https://arxiv.org/abs/2609.38913
作者: Bo Zhao,Junzhe Cao,Dan Guo,Dongmin Huang,Wenjin Wang,Tao Tan,Yue Sun,Zitong YU
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Remote photoplethysmography (rPPG) enables non-contact physiological measurement but remains vulnerable to domain shifts from illumination, motion, and sensors. We propose \textbfFLOW (Feature-Level Optimal Warping), an \emphoptimal transport–driven framework for domain-generalized rPPG. FLOW integrates a \textbfTemporal Refinement Module (TRM) to stabilize temporal dynamics and a \textbfPrototype-based Cross-Temporal Optimal Transport (PCOT) module to achieve domain-invariant alignment via learnable this http URL feature alignment, FLOW employs soft cross-temporal correspondence modeling that aligns temporal features in a flexible manner, allowing the model to respect and preserve the intrinsic rhythmic patterns of physiological signals. Moreover, the lightweight design of our modules allows seamless integration into existing end-to-end rPPG architectures without additional preprocessing. Two regularization terms further enforce source consistency and identity preservation. Theoretically, we derive a generalization bound under conditional optimal transport. Extensive experiments across four rPPG benchmarks show that FLOW achieves state-of-the-art cross-domain performance with lightweight design and strong physiological fidelity.
[CV-137] MEMO: Multi-Level Entity-Aware Memory for Streaming Video Understanding ACM-MM2026
链接: https://arxiv.org/abs/2609.38900
作者: Yinying Li,Yuqian Fu,Yulin Dai,Jingyu Gong,Tianwen Qian,Xiaoling Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by ACM Multimedia 2026 (ACM MM 2026)
Abstract:Streaming video understanding requires models to process unbounded visual streams while preserving rich visual semantics across vast temporal horizons, posing a fundamental challenge for memory modeling. Existing approaches primarily focus on increasing memory capacity, either by compressing historical information into fixed-size representations or by extending storage beyond GPU memory. However, these methods largely rely on global or coarse-grained representations, inevitably losing fine-grained visual information. In this work, we argue that streaming video memory should explicitly encode structured and semantically meaningful representations, particularly at the entity level. To this end, we propose MEMO, a novel framework that models streaming video through multi-level, entity-aware structured memory. MEMO performs multi-level perception to jointly capture global semantics, entity dynamics, and spatial structures, partitioning streaming video into semantically coherent chunks. Each chunk is organized into a structured memory, where lightweight global and entity-level representations serve as retrieval indices, while the corresponding high-resolution visual content is retained separately for on-demand access. At inference time, MEMO performs query-specific retrieval over the structured memory and selectively recalls relevant visual evidence for downstream reasoning. Notably, MEMO is training-free and plug-and-play with existing multimodal large language models. Extensive experiments on StreamingBench and OVO-Bench demonstrate that MEMO consistently improves multiple base models and achieves state-of-the-art performance.
[CV-138] AdaOcc: Adaptive 3D Occupancy Prediction for Embodied Tasks
链接: https://arxiv.org/abs/2609.38864
作者: Jinglong Wang,Yunjie Wang,Zhiyang Zhang,Jiawei He,Ye Yuan,Bo Qiu,Jing Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Embodied tasks demand accurate, flexible, and semantically rich 3D scene representations. 3D semantic occupancy is well suited to this requirement, as it can model holistic 3D spaces by encoding geometric occupancy along with semantic categories. However, existing occupancy prediction methods struggle to meet practical deployment requirements, such as adapting to varying computing budgets, sensor setups, and observation views. In this paper, we propose a point-based Adaptive 3D Occupancy Prediction method, called AdaOcc, tailored for embodied scenarios. To accommodate heterogeneous sensor inputs, AdaOcc uses an adaptive geometry-guided dual-branch encoder that can support RGB images in various numbers of views with (estimated) depth maps or LiDAR scans. AdaOcc represents occupied regions via sparse semantic points trained with a progressive query learning strategy, allowing the prediction computational budget to be flexibly adjusted through query point numbers and decoder layers. To facilitate high-fidelity geometric modeling for lightweight point-based occupancy learning, we further propose a novel containment loss that regularizes predicted points to reside within valid occupied regions. Extensive experiments show that our method achieves a new state-of-the-art on Occ-ScanNet with considerable performance improvements over previous methods. Moreover, our framework demonstrates strong practical applicability as an adaptive 3D perception module in real-world embodied systems.
[CV-139] Decoupling Spherical Reasoning from Dense Prediction for 360 Depth Estimation
链接: https://arxiv.org/abs/2609.38856
作者: Zhijie Shen,Chunyu Lin,Shuai Zheng,Feng Li,Runmin Cong,Huihui Bai,Yao Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The equirectangular projection (ERP) is widely used for panoramic depth estimation, but its spatially varying distortion makes geometry-consistent feature modeling challenging. We revisit panoramic depth estimation by decoupling contextual modeling in native spherical space from dense ERP prediction. To this end, we propose a Fibonacci Spherical Graph (FSG) as an intermediate reasoning space to lift ERP features onto quasi-uniform Fibonacci nodes on the sphere and capture local and long-range dependencies through complementary spherical neighborhoods. The resulting spherical discretization distributes graph nodes approximately uniformly over the spherical surface, reducing the over-representation of highly stretched regions during relational modeling. Operating on a compact set of Fibonacci nodes also avoids the computational burden of constructing and processing a graph at full ERP resolution. To bridge spherical reasoning and dense prediction, we propose a Spherical Context Conditioning (SCC) module that adaptively modulates dense ERP features with the enhanced spherical representation, allowing spherical context to guide pixel-aligned depth prediction. Extensive experiments on three benchmarks demonstrate that the proposed method consistently achieves superior depth accuracy over existing approaches.
[CV-140] FrameMorrow: Future-guided Frame Selection with Prospective Tokens for Long-Horizon Video Generation
链接: https://arxiv.org/abs/2609.38839
作者: Bo Yin,Xiaobin Hu,Jiaqi Zhao,Shuicheng Yan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Long-horizon video generation requires models to effectively leverage an increasingly long generation history. As the generated history grows, retaining all previous content becomes increasingly expensive and redundant, making effective historical selection essential. Existing approaches often determine historical relevance based on the current content. However, information relevant to the present is not necessarily useful for future generation, while seemingly less relevant history may become important later. Our key insight is that historical information should be selected according to its relevance to future information needs. Capturing these needs does not require generating the full future; instead, a compact representation of what becomes important next is sufficient to guide historical selection. Building on this insight, we propose FrameMorrow, a prospective frame selector that predicts a small set of prospective tokens representing future information needs and uses them to identify relevant information from history. FrameMorrow selects explicit historical frames rather than model-specific internal states, enabling plug-and-play integration across diverse generators, including closed-source models, with little additional inference cost. We evaluate FrameMorrow across five benchmarks and 11 generative models spanning long-video generation, interactive generation, and action-conditioned world models. Extensive experiments demonstrate consistent improvements in long-range consistency, visual quality, and action alignment across diverse generation settings.
[CV-141] DecoMoE: Decoupling Visual Propagation and Expert Computation for Efficient Multimodal MoE Inference
链接: https://arxiv.org/abs/2609.38823
作者: Xudong Tan,Peng Ye,Ming Xie,Chenyu Huang,Yaoxin Yang,Jiayuan Fan,Tao Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages, 11 figures, 6 tables
Abstract:Multimodal mixture-of-experts (MoE) models combine sparse expert activation with visual-language capabilities, yet their inference remains costly because long visual-token sequences repeatedly incur attention, routing, dispatch, and expert-MLP computation. Existing methods typically compress either the token or expert dimension, leaving redundancy along the other. Our analysis reveals two complementary regularities: the depth required for visual propagation varies across inputs, while text-token routing exhibits concentrated and recurrent expert-importance patterns. Based on these observations, we propose DecoMoE, a two-dimensional structured compression framework that decouples visual propagation from expert computation. The Sample-Adaptive Visual Boundary (SAVB) predicts an input-dependent visual-exit layer at which the visual-token block is removed. The Routing-Calibrated Expert Prefix (RCEP) reorders experts offline using text-token routed mass and, from this predicted exit layer onward, retains at each MoE layer the shortest contiguous prefix covering a target routed-mass fraction. We evaluate DecoMoE on Qwen3-VL-MoE and InternVL3.5-30B-A3B across six benchmarks. On Qwen3-VL-MoE, DecoMoE retains 97.91% of dense-baseline performance while reducing computation from 27.06 to 16.73 TFLOPs and latency from 0.44 to 0.26 seconds, yielding a 1.69x speedup. Code will be available at this https URL.
[CV-142] Future Video Generation Better Aligns with the Human Visual Cortex than Observed Video
链接: https://arxiv.org/abs/2609.38819
作者: Chang-Bae Bang,Hyungjin Chung,Byung-Hoon Kim
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Neurons and Cognition (q-bio.NC)
备注:
Abstract:Studying the alignment between the internal representations of vision models and the responses of the visual cortex to the same observed visual stimuli has enabled us to better understand human visual processing. However, studies so far have largely overlooked the fact that the human brain not only processes observed visual stimuli, but also predicts upcoming stimuli based on what has been observed. Accordingly, we hypothesize that internal representations for generating future video frames are better aligned with the predictive nature of human visual processing than representations of the observed video itself. To this end, we compare the alignment between human video-watching fMRI responses in the visual cortex and the internal representations from two types of video diffusion models, an autoregressive (AR) model and its non-AR base model. We first conduct a within-model analysis of the AR video diffusion model and show that the representations for future video generation align better with the visual cortex than the representations of the observed video. We then compare the internal representations of the AR model with those of its non-AR base model and again show that the representations for future video generation align better with the visual cortex than the representations for observed video reconstruction by the base model. Specifically, the alignment of observed video reconstruction is concentrated in lower-order visual cortex, whereas that of future video generation is concentrated in higher-order visual cortex. Finally, we show in a human behavioral experiment that humans prefer videos generated by amplifying the contributions of individual layers that align better with the visual cortex.
[CV-143] DCM-SAM: Defect-Conditioned Mixture of LoRA Experts for NPU-Deployed AM Defect Segmentation NEURIPS2026
链接: https://arxiv.org/abs/2609.38811
作者: Md Mushfiqur Rahaman,Md Mahedi Hasan,Imtiaz Ahmed,Srinjoy Das
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 12 pages, 1 figure, 8 tables. Accepted at the NeurIPS 2026 Workshop on On-Device Intelligence: Foundation Models under Real-World Constraints (ODI)
Abstract:Metal additive manufacturing parts are inspected by X-ray computed tomography, where labelled data is scarce, the pores and inclusions that matter span a few pixels, and inspection must happen at the machine. We present DCM-SAM, a defect-conditioned adaptive mixture of LoRA experts: one frozen Segment Anything backbone carries a separate Conv-LoRA expert bank and mask decoder per defect class, each trained in its own pass, without prompts, on synthetic slices alone, updating only 4.4% of the parameters. On benchmarks that XCT-SAM reports, DCM-SAM improves on every baseline for both classes from a ViT-B backbone against their ViT-H, and reaches 64.2% pore IoU on real NIST scans having seen no real images during training. Deployment then exposes what adaptation work rarely measures: on a Qualcomm Hexagon NPU, ViT-H and ViT-L compile yet cannot allocate at 1024x1024 image resolution, since activations rather than weights exceed the device ceiling, and quantizing weights does not help. ViT-B alone runs, but the adapted encoder then fails to allocate where the stock one succeeds, until a numerically identical rewrite of the attention lets the complete DCM-SAM run in FP16 at 1024x1024, with no operator falling back to the CPU, masks within 0.01% of pixels of the FP32 reference. Code: this https URL.
[CV-144] CRAFT: Causal Responsibility and Failure Tracing in Medical Vision Language Models
链接: https://arxiv.org/abs/2609.38810
作者: Chunzheng Zhu,Jiaqi Zeng,Hongbo Zhao,Yihang Chen,Yijun Wang,Jianxin Lin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:As vision language models are increasingly deployed in clinical diagnosis, understanding how they internally resolve competing visual and textual signals becomes a safety imperative. Existing mechanistic analyses remain confined to unimodal text and offer no explanation for why a single misleading sentence can override a correct image based diagnosis, or why a model commits to a confident answer despite insufficient visual evidence. We find that these two safety risks, arbitration failure where textual context overrides visual grounding and brake failure where the model commits without adequate evidence, are mediated by spatially disjoint attention head populations: arbitration heads form a mid-to-deep wideband reflecting cross-layer evidence competition, while brake heads concentrate in a narrow middle-to-late layer band that regulates evidence sufficiency and abstention behavior. To ground these observations in causal circuitry, we introduce CRAFT, which localizes each failure mode to a minimal causal head set via dual criteria and verifies necessity and sufficiency through temporal probes and Tuned Lens trajectory analysis. Excising arbitration heads sharply reduces conflict following with negligible degradation on clean inputs, while excising brake heads restores appropriate abstention under degraded visual evidence. The two interventions target spatially disjoint head sets and produce distinct corrective effects, underscoring the mechanistic separability of the failure modes. Experiments across multiple medical VQA benchmarks and VLM architectures validate both the localization and interventions, demonstrating that the identified heads causally drive each failure mode and that targeted modulation generalises without retraining. The code is available at GitHub repository.
[CV-145] Distill the Visual Evidence Not Just the Answer: Cross-World On-Policy Distillation for Vision-Language Models
链接: https://arxiv.org/abs/2609.38777
作者: Yuanhao Sun,Huawei Ji,Jiaxin Ding,Luoyi Fu,Xinbing Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:A central goal of vision-language model (VLM) distillation is to transfer both the teacher’s language capabilities and its visual understanding. However, existing methods primarily supervise the student’s output, leaving visual understanding implicit. Our analysis reveals that a student can match the teacher’s answer without relying on the same visual evidence, raising the question: how can we ensure the student responds to the visual information that actually determines the answer? To this end, we propose \textbfCross-World On-Policy Distillation (CW-OPD), which explicitly supervises the student’s response to changes in visual evidence. For each example, CW-OPD constructs two visual worlds that share the question and scene context but differ in answer-critical evidence, yielding different answers. We perform on-policy distillation in both worlds and distill the teacher’s cross-world belief transition, encouraging the student to match not only \emphwhat the teacher predicts but also \emphwhy its prediction changes with the evidence. A gradient analysis shows that this term is invariant to errors shared by both worlds and supplies a corrective signal invisible to endpoint matching alone. In this way, CW-OPD makes reliance on the relevant visual evidence an explicit distillation target rather than an implicit consequence of output matching. To diagnose whether a model truly grounds its answers in visual evidence, we introduce CWBench, which measures cross-world consistency via Cross-World Pair Accuracy (CWPA). Experiments on Qwen3.5-4B show that CW-OPD outperforms the strongest baseline by \textbf1.2 points on average, and the 4B student exceeds DeepSeek-V4.1 (552B) by \textbf22.4 CWPA points on CWBench. Code is released in this https URL.
[CV-146] Event-Driven Refresh and Recurrence Memory to Reduce Stale Grounding in Referring Video Object Segmentation
链接: https://arxiv.org/abs/2609.38758
作者: Abu Hanif Muhammad Syarubany,Jaehyun Jang,Siwoo Lim,Seungyeon Ryu,Chang D. Yoo
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: submitted to IEEE Access
Abstract:Referring Video Object Segmentation (RVOS) aims to produce a pixel-accurate mask sequence for an object specified by natural language. Sa2VA combines a multimodal large language model with SAM2 for grounded segmentation; however, its inference typically grounds the query from a small fixed set of initial keyframes and then relies on propagation. In long or dynamic videos, this can cause stale grounding and persistent false positives when the object composition changes (e.g., distractors enter or the target disappears/re-appears). We propose Event-Driven Refresh + Recurrence Memory (EDRRM), an enhancement that selectively re-invokes Sa2VA only at stable change points. EDRRM triggers refresh boundaries using an EMA-smoothed event score computed from tracking-derived cues (births/deaths and coarse composition/layout changes) with temporal constraints. A recurrence memory further retrieves anchor frames via CLIP similarity to re-condition the model on re-appearance events. Experiments on Ref-DAVIS17, MeViS, and ReVOS show that EDRRM achieves a competitive accuracy-efficiency trade-off relative to fixed-window and FrameDiff-SSIM baselines, maintaining comparable or superior JF scores at substantially lower average refresh-call budgets and reducing false-positive failures. End-to-end runtime analysis further confirms that the overhead introduced by tracking, CLIP-based recurrence matching, and the identifiability gate remains modest relative to the dominant Sa2VA inference cost, thereby validating the efficiency of the proposed pipeline.
[CV-147] Agent ic Relative Camera Pose Estimation via Learned Ranking and Verification
链接: https://arxiv.org/abs/2609.38755
作者: Zhining Gu,Shangjie Du,Weimin Qiu,Carl Olsson,Ping Liu,Meng Tang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 22 pages, including 10 pages for the main body
Abstract:A wide range of approaches have been developed for camera pose estimation, including correspondence-based methods, end-to-end pose regression, and recent 3D geometric foundation models. Our key observation is that no single estimator is optimal for diverse challenges, such as wide baselines, lack of texture, appearance changes, and occlusions. Further analysis reveals substantial performance variation across both benchmarks and individual image pairs, with different estimators exhibiting complementary strengths. We introduce PoseAgent, an agentic framework for relative camera pose estimation that dynamically orchestrates pose estimators through learnable ranking and verification. Given an image pair, a profiling agent first extracts appearance, semantic, and geometric features relevant to pose estimation, e.g., scene type. A learned ranking agent then predicts the relative competence of multiple pose estimators given the image-pair profile. The top-ranked estimator is executed, and its predicted pose is assessed by a learned verification agent that estimates the corresponding pose error. When verification fails, PoseAgent adaptively invokes lower-ranked estimators until a candidate is accepted or the execution budget is reached. For pose verification, our verification network predicts pose errors more accurately than prior models. For pose estimation, PoseAgent improves AUC@5 degree up to 4.2% over the strongest standalone estimator on each of ARKitScenes, MegaDepth, ScanNet++, and RealEstate10K. On ARKitScenes, PoseAgent also outperforms VLM-based agents, which include a VLM ranker with the same verifier and fallback policy. These results demonstrate the effectiveness of our learned ranking and verification.
[CV-148] Here the World in Stereo: Learning Dynamic Spatial Correspondence for Immersive Joint Video-Audio Generation
链接: https://arxiv.org/abs/2609.38748
作者: Hanmo Chen,Chengcheng Liu,Tianxiao Chen,Zheyu Zhang,Siming Zheng,Jinwei Chen,Xu Yang,Cheng Deng,Bo Li,Peng-tao Jiang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent joint video-audio generation models have achieved strong semantic correspondence and temporal synchronization. However, applications such as AR/VR and interactive gaming further require stereo audio to provide an immersive sense, which remains largely overlooked. Effective stereo audio requires the perceived sound location to evolve consistently with the motion of its corresponding visual source. We refer to this property as Dynamic Spatial Correspondence and propose StereoBind, a framework that binds visual source motion to stereo sound generation. StereoBind uses motion tracks to coordinate visual motion and stereo audio through three complementary mechanisms. Visual Motion Binding establishes source-aware audiovisual correspondence, the Spatial Track Encoder captures absolute source positions, and Residual Track RoPE models relative motion. For supervision and evaluation, we construct StereoWorld-29K, a large-scale stereo audio-video dataset with paired motion tracks, and StereoWorldBench for measuring audiovisual spatial consistency. Experiments show that StereoBind substantially improves spatial alignment in stereo audio generation over existing models while preserving overall audiovisual quality.
[CV-149] Consensus-Aware Multi-Source Fusion for Reference-Guided Camouflaged Object Detection
链接: https://arxiv.org/abs/2609.38747
作者: Junyang Xia,Luocheng Zhang,Wenwen Pan,Chifeng Zhu,Yang Yang,Xinchun Liu,Jiajun Ding
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 20 pages, including 2 pages of appendix; 9 figures and 4 tables
Abstract:Reference-guided camouflaged object detection aims to segment a target whose visual appearance closely resembles its surroundings by exploiting auxiliary reference samples. The task remains difficult because reference samples contain inconsistent target cues, while generic visual representations are not inherently aligned with the target specified by the references. To handle these problems, we present a consensus-aware multi-source fusion framework. Reference-Conditioned Dual-Backbone Fusion (RCDF) couples trainable PVTv2 query features with frozen DINOv3 representations and uses reference-conditioned correlation to select foundation-model evidence before multi-scale fusion. The framework also aggregates multiple references through cross-reference consensus aggregation and injects reference information at semantic depths matched to the query features. Extensive experiments demonstrate the effectiveness of the proposed method. The results further show that reference consensus, target-conditioned foundation features, and hierarchical decoding provide complementary improvements under the evaluation protocol. The source code will be made publicly available upon acceptance.
[CV-150] Matisse: Evidence-Space Reasoning for Active 3D Reconstruction
链接: https://arxiv.org/abs/2609.38746
作者: Xihang Yu,Kaichen Zhou,Lorenzo Shaikewitz,Clément Jambon,Xiao Zhan,Rajat Talak,Luca Carlone
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 20 pages, 7 figures, 7 tables. Project page: this https URL
Abstract:How can a 3D reconstruction system acquire and retain useful information to understand the geometry of a scene from partial views under a limited computation budget? Existing active view acquisition methods typically estimate uncertainty over observed or instantiated geometry, limiting their ability to reason about unseen structure, while long-horizon reconstruction methods often retain redundant observations. We introduce Matisse, a training-free framework that unifies active reconstruction and keyframe selection by leveraging evidence provided by a pretrained generative 3D model. Matisse estimates Evidential Uncertainty from cross-attention evidence associated with 3D latent tokens and derives an Evidential Information Gain to guide both view acquisition and keyframe selection based on the expected reduction in posterior entropy. Matisse supports multi-object scenes through occlusion-aware, object-balanced aggregation and propagates uncertainty through intermediate latents to avoid full reconstruction during planning. Matisse reduces Chamfer distance by 12.7%, 3.8%, and 9.2% on GSO30, YCB-V, and Replica, respectively, relative to the best baseline on each dataset, and achieves a 1.50\times end-to-end speedup over the best active reconstruction baseline on GSO30 with the same reconstruction backend. In the GSO30 keyframe selection experiment for long-horizon reconstruction, Matisse achieves comparable Chamfer distance using 14% of the input views compared with Stream3D.
[CV-151] UniEvo-VL: An On-policy Self-Distillation Training Recipe for Multimodal Model Self-improvement
链接: https://arxiv.org/abs/2609.38721
作者: Fang Wu,Da Xing,Yanjie Huang,Junxi Wang,Ji Wang,Hejia Geng,Guancheng Wan,Bowen Zuo,Xiaomin Li,Shixiang Tang,Xinyu Xiang,Zehong Wang,Shiyi Du,Peng Xia,Shuangjia Zheng,Yining Hong,Li Erran Li,Jure Leskovec,Yejin Choi
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Modern multimodal models bring generation and understanding into a single unified system, which enables them to provide and learn from their own feedback. Motivated by this unified capacity, we introduce UniEvo-VL, a self-evolving framework for multimodal models to learn from this constructive self-correction feedback during test-time compute. Instead of relying on a separate, often larger, teacher, we leverage their self-critiques as privileged information and ask a single multimodal model to act as both teacher and student with different contexts. The student only sees the vanilla question, while the teacher conditions on the privileged critique. Then training minimizes the per-state divergence between their denoising diffusion distributions over the student’s own sampling trajectories. Experiments demonstrate that UniEvo-VL improves the image generation capabilities of multimodal models, while maintaining their sensitivity to additional reflection information. Specifically, we build on top of the open-source Qwen-image-2512 and observe a significant performance gain from 0.747 to 0.808 on GenEval and from 32.97 to 35.53 on GenEval2 Soft-TIFA. Moreover, attempts with more powerful external critics (e.g., GPT5.6-Luna) show that multimodal models with strong judge capabilities can anticipate a higher self-evolving ceiling. Last but not least, mixed text-rendering outcomes show that our self-improvements may not be uniform across different tasks. Our study aims to shed light on the current hot recursive self-improvement research line to enhance the user experience when using multimodal models without external supervision or guidance.
[CV-152] Soft Spatial Reasoning
链接: https://arxiv.org/abs/2609.38717
作者: Rafi Ibn Sultan,Md. Sajid Alam Chowdhury,Saleh Zare Zade,Chengyin Li,Prashant Khanduri,Marco Brocanelli,Dongxiao Zhu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Large Vision-Language Models (LVLMs) commonly perform spatial reasoning through chain-of-thought (CoT), encoding intermediate reasoning as autoregressive sequences of discrete language tokens. Such hard thinking requires committing to a single token at each step, even when the correct spatial interpretation remains uncertain. This early commitment constitutes premature discretization: an incorrect token selection can propagate errors through subsequent reasoning. We propose Soft Spatial Reasoning, a post-training framework that introduces soft thinking for spatial tasks in LVLMs. At each intermediate reasoning step, the LVLM forms a continuous soft state by mixing token embeddings rather than selecting a single token, allowing multiple candidate continuations to influence the next step. The appropriate degree of softness, however, can vary across reasoning steps: retaining multiple candidates may preserve a useful spatial interpretation, but if those candidates imply conflicting spatial relations, mixing them may interfere with subsequent reasoning. At the core of Soft Spatial Reasoning is AdaptSoft, a controller that uses the current hidden state and predictive uncertainty to adapt the degree of softness at each reasoning step. To train AdaptSoft, we introduce a gradient-alignment learning objective that provides a step-specific learning signal for softness control without intermediate reasoning supervision. Across diverse spatial benchmarks, Soft Spatial Reasoning outperforms hard and fixed-soft CoT baselines using the same backbone, as well as a range of existing LVLMs. The source code is available at this https URL
[CV-153] SpatialCORE: Confidence-Aware Grounded Spatial Reasoning in Large Vision–Language Models
链接: https://arxiv.org/abs/2609.38716
作者: Rafi Ibn Sultan,Xiangyu Zhou,Md. Sajid Alam Chowdhury,Chengyin Li,Prashant Khanduri,Marco Brocanelli,Dongxiao Zhu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Large Vision-Language Models (LVLMs) have made remarkable progress across visual perception tasks, yet spatial reasoning remains a persistent weakness, especially for questions that require reasoning over visual space. Recent spatial-reasoning methods incorporate generated grounding, where models predict bounding boxes, masks, or other localization outputs for task-relevant objects as part of their reasoning trace. However, these approaches typically optimize final-answer correctness alone, allowing correct answers to be rewarded even when the model does not reason from confidently localized task-relevant objects. We introduce SpatialCORE (Spatially COnfident REasoning), a post-training framework that turns the model’s own confidence in generated grounding into a learning signal for spatial reasoning. Its central idea is to reinforce grounding that is both accurate and confident, encouraging the model to reason from confidently localized task-relevant objects. SpatialCORE realizes this through a self-regulating spatial reward that weights each predicted bounding box’s matching quality by its coordinate-token confidence. An answer gate further ties grounding optimization to final-answer correctness. SpatialCORE achieves state-of-the-art results among open-source and specialized spatial reasoning models across diverse benchmarks, and transfers effectively in zero-shot settings to unseen data distributions. The source code is available at this https URL.
[CV-154] Hard-Region Supervision: #1 on the Waymo Open Dataset 2D Video Panoptic Segmentation Leaderboard
链接: https://arxiv.org/abs/2609.38714
作者: Jinghan Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We describe our winning entry to the Waymo Open Dataset 2D Video Panoptic Segmentation Challenge. The task asks for a semantic class at every pixel of every frame and, for countable objects, an identity that holds across 100 frames and across five overlapping cameras. We build on DVIS++, a cascade of a segmenter, a tracker, and a refiner, as our baseline. We propose hard region supervision (HRS) to improve the baseline. In particular, we use the baseline to define the hard region as where it makes mistakes, and design a loss and an auxiliary prediction head for this region. The auxiliary head is used only in training and removed at test time, so at inference the model trained with HRS has the same architecture as the baseline. In addition, we propose three test-time steps that further improve the results: a two-model ensemble, a merge of the segmenter’s output into the final panoptic map, and cross-camera identity linking. On the challenge test set, our entry reaches 0.3547 wSTQ, 0.2071 wAQ, and 0.6075 mIoU, ranking first on all three metrics. It is 3.6 wSTQ points ahead of the second entry and 2.4 points ahead of our DVIS++ baseline.
[CV-155] SCALE: Synthetic Calibration via Agreement Labeling in Embedding Space
链接: https://arxiv.org/abs/2609.38705
作者: Wenjun Liu,Saeed Hassanpour
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Foundation models for computational pathology are usually evaluated using AUC and accuracy, while calibration is often left untested. This matters because a model can be accurate on average but still assign overly confident probabilities to cases that are difficult even for pathologists. We study calibration across eight pathology foundation models. Using pathologist agreement as a measure of diagnostic difficulty, we find that calibration error is consistently higher on low-agreement cases than on high-agreement cases. This pattern is not apparent from aggregate expected calibration error (ECE) alone. We then propose synthetic agreement calibration, a method for improving calibration without collecting multi-annotator labels. Given a trained linear probe, we select high-confidence embeddings as class anchors and interpolate between anchors from opposite classes. The interpolation weights encode a continuous notion of diagnostic ambiguity, which we use as a synthetic agreement signal to retrain the probe with agreement-aware label smoothing. On MHIST, which includes annotations from seven pathologists, synthetic agreement calibration recovers most of the calibration improvement obtained by label smoothing based on real pathologist agreement, while substantially reducing low-agreement ECE relative to the uncalibrated baseline. Discrimination metrics are preserved. On PatchCamelyon and BreakHis, public histopathology datasets without multi-annotator labels, the method improves calibration across the evaluated foundation models, whereas annotator-dependent approaches cannot be used without additional expert annotation.
[CV-156] No Corners Cut: State-Grounded Transitions for Mid-Stream Prompt Switches in Video Generation
链接: https://arxiv.org/abs/2609.38691
作者: Zejing Rao,Ketong Ren,Xiaoqiang Liu,Yiping Meng,Guoxin Zhang,Fan Tang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Streaming video generators allow users to dynamically modulate video synthesis via mid-stream prompt switching. Existing streaming methods can respond to the updated instruction while still cutting corners, prematurely realizing goals or taking heuristic shortcuts that bypass necessary intermediate state changes needed for a plausible transition. In this study, we present SEGUE, a novel framework that makes this process explicit and trains the generator to execute these transitions faithfully. At each switch, a training-free planner parses the latest frame and prompts, writes a few segue prompts with roles and durations, and then hands control back to the user’s prompt. Furthermore, to address the inherent difficulty of training causal models on short-lived temporal schedules without corrupting preparatory supervision, we introduce SPANDMD, which evaluates each active prompt using the full rollout as temporal context while retaining its DMD residual only within the prompt’s assigned span. On OpenTrans-360, a benchmark of 1,800 switches that scores how the old state exits and the new one begins, SEGUE ranks first on all eight transition metrics and raises the overall score over the strongest baseline from 0.866 to 0.887. It also ranks first on four of six instruction-response metrics of StreamAV-Bench, while the planner transfers to frozen autoregressive generators without retraining.
[CV-157] Unveiling the Value of Motion for Cinematic Camera Trajectories
链接: https://arxiv.org/abs/2609.38683
作者: Ziqi Zhou,Yujian Yuan,Laura Sevilla-Lara
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Cinematic camera motion is a fundamental storytelling tool, defined not only by where the camera is positioned in the scene, but also by how it moves in terms of direction and speed. Recent work on camera trajectory generation and alignment to text relies on pose-centric representations. While in principle a network could derive direction of movement and speed, we find that in practice this might not happen. In fact, in this paper we discover that decomposing the camera trajectory representation from the traditional per-frame poses to direction and speed has surprising benefits across multiple tasks, including trajectory-to-text alignment as well as text-to-trajectory generation. To accurately evaluate the former, we introduce a simple and reliable protocol that overcomes the limitations of prior evaluation baselines. For the latter, building on this representational insight, we propose a novel generative model for camera trajectories, CineGEN, that achieves superior performance across a variety of metrics. We also propose a novel dataset, CineScript, containing movie clips that are enriched with scene descriptions as well as higher-level metadata. This novel data allows us to test models’ ability to capture high-level cinematographic information. We show that, despite its simplicity, representing camera trajectories through direction and speed not only helps numerically to achieve better alignment and generation, but also inherently encodes complex directorial intent.
[CV-158] ReGain: Restoring Subject Fidelity in Personalization on Synthetic Images
链接: https://arxiv.org/abs/2609.38680
作者: Shubhang Bhatnagar,Ishan Bhatnagar,Viraj Shah,Narendra Ahuja
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Image and Video Processing (eess.IV)
备注:
Abstract:Text-to-image diffusion models are personalized to a subject by DreamBooth fine-tuning on a handful of its images. Increasingly, these images come from a diffusion model rather than a camera. We show that fine-tuning on such synthetic images degrades subject fidelity, producing oversaturated color and excess high-frequency detail. To isolate the cause, we fine-tune two models from the same base model with the same DreamBooth recipe, one on real photos of a subject and one on synthetic images of that subject generated by the first. We trace the degradation to classifier-free guidance (CFG). For the model personalized on synthetic images, the angle between the conditional and unconditional noise predictions, and with it the norm of their difference, is much larger than for the model personalized on real photos. This inflation grows toward high frequencies and also appears at other prompts semantically close to the subject, such as its class noun, but not at unrelated ones. We propose ReGain, a training-free correction applied at sampling time that measures how much each frequency band of the guidance is inflated relative to the base model and scales that band down accordingly. ReGain needs no real photos. On Stable Diffusion v1.5, ReGain closes 51-64% of the subject-fidelity gap to the model personalized on real photos, as measured by DINO, DINOv2 and CLIP-I. It also improves subject fidelity on SDXL and SD 3.5 and preserves text alignment on all three backbones.
[CV-159] ChartRevise: A Dataset and Evaluation Protocol for Exact Chart Editing via Code
链接: https://arxiv.org/abs/2609.38642
作者: Jiaxiang Tang,Yi Zhou,Chad DeLuca,Rogerio Feris,Ahmed Khalil Omran,Zhi-Li Zhang,Pengyuan Li,Ali Anwar
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Chart editing requires cross-modal edit grounding, realizing a requested visual change in the code that draws it, with necessary related updates and without altering unrelated content. Existing benchmarks emphasize either code executability or chart quality, but their metrics do not clearly distinguish request completion from missed coupled updates and gratuitous changes. We introduce ChartRevise, a structured dataset and evaluation protocol for exact program-grounded chart editing. For dataset construction, we build on the grammar of graphics to systematically cover chart-editing operations, using source-program checks to verify their applicability across chart types and libraries. To improve edit exactness, our pipeline checks individual requirements and guides repair or exclusion when they are unmet. The resulting dataset contains 92,438 records covering 344 edit types across 20 chart types and three plotting libraries. For evaluation, our reference-free protocol separately measures atomic requirement completion, identifies gratuitous changes, and detects missed coupled updates. These checks are combined with successful execution and rendering to determine exact-edit success. Across five models and four external benchmarks, fine-tuning yields relative gains of 16% in mean requirement recall and 22% in mean exact-edit rate.
[CV-160] Vision-Language-Action Autonomous Driving Agent with Language-based Memory
链接: https://arxiv.org/abs/2609.38641
作者: Kai Yan,Xiangyu Chen,Yulong Cao,Alex Naumann,Peter Karkus,Yan Wang,Jef Packer,Alex Schwing,Yuxiong Wang,Boris Ivanovic,Wenjie Luo,Marco Pavone
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 39 pages, 21 figures
Abstract:Vision-Language-Action (VLA) foundation models have recently emerged as one of the prevailing solutions for autonomous driving, as they can utilize knowledge acquired during vision-language pretraining for accurate and interpretable driving. However, VLAs can take only a limited number of frames as visual input due to the high token cost of an image, which is problematic for memory-dependent tasks such as determining the arrival order at all-way stops and long-horizon driving scene understanding. Existing solutions use latent vector memories accessed through cross-attention, which are neither interpretable nor portable. In this paper, we propose AD-Memo, a general-purpose VLA driving agent with language-based memory. The agent outputs memory as an extension of its Chain-of-Thought (CoT) to record surrounding objects critical to driving; this memory becomes part of the agent’s future input. We curate memory-based datasets and train VLAs with a two-stage recipe: Supervised Fine-Tuning (SFT) and \textitDa Capo, a novel semi-closed-loop Reinforcement Learning (RL) algorithm which uses trajectory-level advantage for memory and step-level advantage for driving, leading to better credit assignment. Across scenarios such as all-way stops and general driving, AD-Memo improves driving quality, enables better question answering on driving scenes, and produces plug-and-play memory for other models.
[CV-161] mplate-Search Domain Adaptation via Multi-Stage Feature Alignment for Cross-Modal Object Tracking
链接: https://arxiv.org/abs/2609.38637
作者: Fereshteh Aghaee Meibodi,Amir Mehdi Soufi Enayati,Shadi Alijani,Homayoun Najjaran
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Visual object tracking typically assumes that the initial template and subsequent search frames share the same sensing modality. In practice, sensor availability or operation may change over time, creating a substantial representation gap between template and search frames. Unlike conventional multi-modal tracking where paired modalities are simultaneously available, cross-modal tracking requires localization when template and search frames originate from different active modalities. Accordingly, we introduce TSDA-Track, a Template-Search Domain Adaptation framework to reduce modality discrepancy during training. We investigate two feature alignment strategies. Pre-AFA TSDA-Track applies adversarial alignment before transformer’s template-search interaction to suppress modality-specific bias. Enc-CFA TSDA-Track applies contrastive alignment to encoder representations after interaction to strengthen target-level cross-modal correspondence. Both variants retain a shared inference pipeline without modality-specific branches. Experiments on LasHeR, and zero-shot evaluations on RGBT234 and GTOT under multiple cross-modal protocols demonstrate improvements over representative state-of-the-art trackers. For instance, under the modality-switch protocol on RGBT234, Pre-AFA TSDA-Track achieves an SR/PR of 43.2/56.0, compared with 36.8/50.0 for ToMP-101 baseline. In addition, a study on Anti-UAV-024 further verifies the applicability of TSDA-Track to aerial tracking. Our study highlights the effectiveness of feature alignment domain adaptation for cross-modal tracking.
[CV-162] STEPS: Scene Text Editing with Preserved Style Using Diffusion and Contrastive Style Encoding PAKDD2026
链接: https://arxiv.org/abs/2609.38636
作者: Nicolas Thiebaut,Nameer Hirschkind,Xiao Yu,Kyle Spence
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 9 pages, 5 figures, 4 tables. A version of this paper appeared in PAKDD 2026 (LNCS vol. 16618)
Abstract:We introduce Scene Text Editing with Preserved Style (STEPS), a novel diffusion model architecture for quality text replacement in images. Scene Text Editing (STE), also known as Visual Text Editing, consists of changing the textual content in an image while conserving the original style, e.g. font, colors, orientation, background, etc. STEPS advances the state of the art in STE through directed focus on improved style preservation. We introduce a style encoder for visual text that captures style independently of textual content, and a model architecture that combines the style encoder with multiple semantic conditions (target text characters encoding and rendered glyphs). STEPS achieves superior results to previous STE methods in style preservation, output readability, and subjective quality.
[CV-163] Eulerian Motion Reconstruction for Water Scenery
链接: https://arxiv.org/abs/2609.38622
作者: Chuhan Chen,Yen-Chi Cheng,Ayush Saraf,Rajvi Shah,Tuotuo Li,Johannes Kopf,Chen Gao,Hung-Yu Tseng,Deva Ramanan,Matthew O’Toole,Changil Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project at this https URL
Abstract:Reconstructing and animating water scenery from nature produces compelling and immersive visual experiences. Previous work examined this task from the perspective of 2D video textures, with the goal of creating a looping video. In our work, we tackle the problem from a 3D perspective, creating a looping 4D dynamic reconstruction which can be interactively rendered from novel viewpoints from a single non-looping 2D source video. We represent motion as a 3D static \textitEulerian motion field that advects canonical Gaussian splats that are cyclically reborn at fixed time periods, supervised using rendering losses. To model non-periodic and stochastic dynamics present in real-world scenes, we add a non-periodic, time-varying residual term to capture deviations from the static Eulerian motion field. We show quantitatively and qualitatively that our framework enables photorealistic animation of water scenes better than prior art.
[CV-164] HIGS: Hierarchical Implicit Grids for Joint Geometric and Semantic Scene Understanding
链接: https://arxiv.org/abs/2609.38620
作者: Hanwen Cao,Wenqiang Wu,Kuang-Ting Tu,Mathias Otnes,Jeffrey Delmerico,Rui Wang,Yulun Tian,Nikolay Atanasov
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Neural implicit representations have had a significant impact on scene reconstruction by enabling robots to build continuous, differentiable, and high-fidelity 3D maps. Most existing works focus on geometric reconstruction and lack semantic information for high-level spatial understanding and task planning. Also, as the scale and complexity of the environment increase, neural representations face the challenge of maintaining computational efficiency in back-end optimization. To resolve these two challenges, we introduce a hierarchical neural field that leverages multiresolution submaps to achieve an efficient and scalable implicit representation, and a unified query and decoding mechanism to support both geometric and semantic features. More specifically, the learnable map features can be converted to the output with the query and decoding process for both training and inference. For large-scale representation, we decompose a scene into overlapping submaps and do hierarchical optimization within each local submap, thus enabling scalable computation. To further improve efficiency, we design feature encoders that predict initial hierarchical grid features to substantially reduce the time needed to optimize the submap features from scratch. To correct estimation drift among submaps, we align and fuse them entirely within the implicit feature space, leading to substantial acceleration by avoiding the need to decode the final output. Building upon this efficient hierarchical representation, we embed both geometric features and vision-language latent features into the map, and demonstrate it on both Signed Distance Field (SDF) construction and open-vocabulary object grounding. Our approach significantly improves computation and memory efficiency, maintains high estimation accuracy, and endows the robot with spatial awareness on large-scale real-world benchmarks.
[CV-165] Correcting WHERE Preserving HOW: Compositional Generalization for Vision-Language-Action Models via Referential Guidance
链接: https://arxiv.org/abs/2609.38616
作者: Yanyan Zhang,Disheng Liu,Xinpeng Li,Chaoda Song,Mohsen Hariri,Debargha Ganguly,Wang Yang,Kai Ye,Bryce Grant,Vipin Chaudhary,Yu Yin
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:While Vision-Language-Action (VLA) models enable flexible action generation, their generalization across diverse environmental elements, including manipulated objects, destinations, and backgrounds, is limited by the lack of diversity in robotic training data. Trained end-to-end on such data, VLAs tend to exploit visual shortcuts, associating actions with task-irrelevant visual features rather than the intended task semantics. These shortcuts block recomposition of elements already seen by the policy, that is, compositional generalization. Existing approaches mitigate such entanglement through task-relevant perception or targeted data diversification, but offer no explicit mechanism for unseen recomposition and require backbone-specific modifications with retraining. We observe that under such recomposition, VLAs often fail at global grounding while retaining local manipulation skills that recover near the correct target in familiar configurations. Therefore, we propose Referential Guidance (ReGuide), a training-free wrapper that, given object poses from a grounding module, combines semantic and geometric rebinding to guide the end-effector into demonstration-supported configurations of the instructed referent, where the frozen policy can resume execution. Experiments in simulation across multiple VLA backbones as well as on a real robot show that ReGuide improves success rates under compositional shifts by up to 56.8 and 75.0 percentage points, respectively, while preserving standard-task performance.
[CV-166] Exo2EgoHOI: Hand-Object-Interaction Aware Exocentric-to-Egocentric Video Generation
链接: https://arxiv.org/abs/2609.38615
作者: Hongjia Zhai,Xiyu Zhang,Haoran Zhang,Zhichao Ye,Haomin Liu,Guofeng Zhang,Ian Reid,Xingxing Zuo
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 18 pages
Abstract:Egocentric videos of human manipulation provide valuable visual experience for embodied intelligence, yet collecting such data at scale is costly. Exocentric-to-egocentric video generation offers a scalable alternative by transforming abundant third-person manipulation videos into first-person observations. However, existing methods often struggle to faithfully preserve demonstrated hand-object interactions (HOI) across large viewpoint changes due to insufficient fine-grained interaction guidance and weak object-centric anchoring. We present Exo2EgoHOI, an HOI-aware video generative framework for interaction-preserving exocentric-to-egocentric translation. To preserve fine-grained HOI, we introduce a unified 4D HOI prior that combines scene geometry, articulated hand renderings, and dense hand-object relation fields, together with a dual-branch residual adapter for injecting structural and relational cues into the video generation backbone. To preserve object consistency, we introduce Decomposed Gated Cross-Attention, which separately encodes object and background references and adaptively integrates global semantic and local appearance features as object-centric anchors. Experiments on ARCTIC-HOI and Ego-Exo4D demonstrate substantial improvements in object consistency and HOI preservation while maintaining competitive visual fidelity. In particular, on ARCTIC-HOI, Exo2EgoHOI improves object mIoU by 32.3% and reduces MPJPE and PA-MPJPE by 34.7% and 50.0%, respectively, relative to the respective best baseline results. Project page: this https URL.
[CV-167] After a Decade: Bringing Shadow Removal into the Real World with Agent ic Training Data
链接: https://arxiv.org/abs/2609.38607
作者: Shilin Hu,Jingyi Xu,Dimitris Samaras,Hieu Le
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Shadow removal looks nearly solved on established benchmarks, yet remains brittle in the real world. Models have advanced; the paired training data they rely on have barely changed in nearly a decade. The reason is simple: obtaining a shadow-free target requires removing the occluder while keeping the scene, camera, and illumination otherwise unchanged, making diverse paired data difficult to capture. Meanwhile, large shadow detection datasets already contain diverse real-world images and masks, but no shadow-free targets. To turn this abundant but incomplete data into paired supervision, we propose an offline agentic workflow combining physics-motivated generation, failure detection, feedback-driven retry, candidate selection, and deterministic correction. Using this workflow, we construct AgenticShadow, a dataset of 17,138 image-mask-target triplets spanning general scenes, faces, and remote sensing. Our construction workflow reduces Color Distribution Difference by 50.5% over previous shadow removal work, while training existing shadow removal models on AgenticShadow reduces cross-domain LAB RMSE by 19.7-37.5%.
[CV-168] Aperture: Training-Free Multiscale Concept Bottlenecks for Remote Sensing
链接: https://arxiv.org/abs/2609.38603
作者: Rishabh Mondal,Nipun Batra,Utkarsh Mall
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:While earth observation models have advanced substantially, they still lack interpretability. While concept-bottleneck models provide interpretability and expert interaction, they are either too expensive to train for the remote sensing domain or perform poorly without annotation. We posit that in expert domains like remote sensing, such training-free models require both fine details in both image and concept space. In image space, we propose a multiscale concept bottleneck using greedy quadtree routing to locate small concepts. In concept space, we replace contrastive vision language models with pre-trained MLLMs and present a way to get reliable concept scores from them. We introduce APERTURE that blends concept scores at the global image and native concept-scale level to give state-ofthe-art training-free model performance. To test these models, introduce SiFC, a fine-grained concept-centric dataset across three countries, with human-reviewed class-level concept maps. On SiFC, APERTURE outperforms the best training-free baselines by more than 10 percentage points in macro F1-score, and notably also outperforms supervised concept bottleneck models. Targeted component-removal tests examine whether concept scores respond to changes in visual evidence, while temporal experiments show that descriptor updates improve recognition of technological changes without retraining.
[CV-169] PixelUMM: Encoder-Free Unified Image and Video Understanding and Generation
链接: https://arxiv.org/abs/2609.38597
作者: Cong Wei,Xuanchi Ren,Bryan Chu,Weiming Ren,Huan Ling,Jiahui Huang,Laura Leal-Taixé,Sanja Fidler,Wenhu Chen,Zian Wang,Jay Zhangjie Wu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 31 pages, 21 figures. Project page: this https URL
Abstract:Unified Multimodal Models (UMMs) often rely on separate visual representations for understanding and generation, increasing visual context length and complicating integration with established vision-language pretraining pipelines. Recent advances in pixel-space modeling offer an encoder-free alternative, but extending this paradigm from images to videos is non-trivial: video understanding and generation adopt different temporal representations, leaving the design of a unified visual interface an open question. We present PixelUMM, an encoder-free model for unified image and video understanding and generation directly in pixel space. PixelUMM represents images as spatial patches and videos as spatiotemporal tubelets, connecting raw pixels to a shared multimodal backbone through single-layer linear projections. Its Mixture-of-Transformers architecture combines shared attention with task-specific parameters and extends clean-pixel prediction to video generation, jointly supporting autoregressive text prediction and pixel-space flow matching. Experiments show that PixelUMM achieves competitive performance across image and video understanding and generation tasks. We further conduct empirical studies of key design choices, including decoder design and spatial-temporal patch size, providing insights for future pixel-space unified multimodal models.
[CV-170] StereoGaussians: Feed-Forward 3D Gaussian Splatting from Stereo Images
链接: https://arxiv.org/abs/2609.38592
作者: Boyuan Tian,Huangying Zhan,Zhan Li,Shin-Fang Chng,Hanwen Yang,Zirui Wang,Yi Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Feed-forward 3D Gaussian Splatting (3DGS) enables reconstruction without per- scene optimisation, but practical stereo-camera applications require nearby-view extrapolation beyond the input views. Stereo depth anchors visible surfaces, yet rendering newly exposed regions also requires learned appearance and additional scene capacity. We introduce StereoGaussians, which predicts a metric 3DGS representation from a single calibrated stereo pair. It reuses intermediate repre- sentations from frozen pretrained stereo networks to predict Gaussian attributes, while calibrated disparity anchors the geometry. A second Gaussian layer and an expanded image canvas provide capacity for disoccluded and outside-field-of- view content. For training, we construct SceneSplat-Stereo from quality-filtered 3DGS teachers, pairing stereo inputs with nearby target views across 803 training scenes. Experiments on unseen real and photorealistic stereo benchmarks demon- strate improvements over strong view-synthesis baselines, while ablation studies support our main design choices.
[CV-171] Restoring without Forgetting: Filter-Level Continual Image Restoration via Parameter-Space Integrated Gradients
链接: https://arxiv.org/abs/2609.38591
作者: Xin Feng,Jin Zhao,Yizhen Zhang,Wenjie Pei,Fanglin Chen,Guangming Lu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Adapting image restoration models to a stream of new tasks without revisiting past data remains challenging due to catastrophic forgetting. In this work, we propose Restoring without Forgetting (RwF), a filter-level continual adaptation framework for image restoration built upon a critical observation: task-specific knowledge is centered in a small subset of filters and can be separated from those reconstructing general content. RwF first performs parameter-space integrated gradients attribution to localize degradation-critical filters in a coarse-to-fine manner. It then adapts to new tasks by generating task-specific filters from a filter bank using compact factorized low-rank transformations, further augmented with cross-task attention and prototypical contrastive learning, and lastly assembles them back only at localized positions. Experiments on six restoration tasks show that RwF effectively avoids forgetting and achieves competitive restoration quality against all-in-one methods that have full data access, and outperforms LoRA-style adaptation with \sim 10 \times fewer additional parameters. Code is available at this https URL.
[CV-172] Retargeting Motions to Diverse Skeletons via Learnable Flattening
链接: https://arxiv.org/abs/2609.38578
作者: Kia-Jüng Yang,Fabian H. Sinz,Paweł A. Pierzchlewicz
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 24 pages, 9 figures
Abstract:Cross-structural motion retargeting aims to transfer motion between different skeletal topologies. Despite recent progress, existing state-of-the-art models struggle with reliability in zero-shot settings, i.e. skeletons with different topologies which were unseen during training, and recent Transformer-based attempts have failed to outperform specialized geometric methods. We bridge this gap with a Transformer Autoencoder that learns a topology- and translation-invariant latent space. Our core contribution is a learnable flattening of skeletal graphs that captures both local dependencies and global structure. Unlike the standard transformer architecture, which adds positional information to token content, we integrate graph-based positional encodings multiplicatively, a design choice that follows directly from our flattening formulation. The resulting model handles diverse skeletal topologies within a single unified architecture and trains in a fully unsupervised manner, requiring no paired retargeting data. Ablation studies show, that the graph encodings, multiplicative formulation, and Transformer backbone is critical for the performance. In zero-shot evaluations, our method reduces global joint position error by 43-47% over current benchmarks. A user study ( n = 37 ), including expert animators, further ranks our approach highest in motion alignment and physical plausibility ( p 0.05 ). These results demonstrate that our model design is key to making transformer architectures effective for motion retargeting, outperforming existing approaches.
[CV-173] LongTake: Learning to Sustain Dynamics in Long-Horizon Video Generation
链接: https://arxiv.org/abs/2609.38562
作者: Byoungwoo Park,Jaemoo Choi,Juho Lee,Yongxin Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:World models, game simulators, and long-take video creation require coherent scene evolution and sustained dynamics over extended durations. Autoregressive (AR) video diffusion provides a natural framework for long-horizon generation, yet extended rollouts often become near-static or lose visual quality. We hypothesize that these failures reflect the limited guidance provided by short-video supervision on how ongoing scene dynamics develops over longer durations. This motivates us to introduce LongTake, a two-stage training pipeline built around Long-Horizon Teacher Forcing (TF) on curated real long videos. Long-Horizon TF trains the AR model to predict later frames conditioned on long ground-truth video prefixes, extending direct supervision beyond the short training horizon. This supervision is designed to help the model sustain dynamics and preserve visual quality during long-horizon generation. Our central finding is that this training stage strengthens direct initialization for distribution matching distillation (DMD) under student self-rollout, without the intermediate few-step distillation stage used in standard pipelines. Under the same five-second DMD training setup, our initialization yields substantially higher dynamic degree than short horizon TF initialization on 30-second rollouts at comparable aesthetic quality, and surpasses the evaluated baselines in both measures. Hybrid DMD further reuses this teacher to extend supervision to later frames of the self-rollout while retaining bidirectional joint supervision over the initial window. On long-horizon self-rollouts, LongTake lies on the Pareto front of dynamic degree and aesthetic quality, and Hybrid DMD attains the highest dynamic degree among evaluated methods at both 30s and 60s.
[CV-174] Detail in Context: A Dual-Scale Machine Learning Framework for Mycosis Fungoides Detection
链接: https://arxiv.org/abs/2609.38560
作者: Mohamed Hazem,Tarek Waleed,Omar Khaled,Nada Omar,Mahmoud Raslan,Marwa Mohamed Fawzy,Aya Fahim,Rania M. Mogawer,Ahmed Mourad,Kariman Mansour,Muhammad Rushdi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Mycosis fungoides (MF) is a rare form of cutaneous T-cell lymphoma that is often misdiagnosed in early stages due to its visual similarity to benign inflammatory dermatoses. Early and accurate diagnosis is critical for improving patient outcomes. In this paper, we propose a comprehensive diagnostic framework for automated MF detection that combines dual- scale histopathological image analysis with deep learning. To distinguish MF from other lymphoproliferative skin conditions, the proposed approach leverages a late-fusion ensemble of dual- magnification (10x and 20x) convolutional neural networks (CNNs), complemented by a random forest classifier trained on 16 clinical features. Experimental results on an expanded dataset of 6,267 images (4,306 MF; 1,961 Non-MF) across 463 patients demonstrate that strong detection performance is obtained by prioritizing higher-resolution cytological details (20x) within broader architectural context (10x). The image-based late-fusion model achieves an accuracy of 83.58% and a sensitivity of 89.13%, while the clinical random forest model achieves an accuracy of 96.6% and sensitivity of 93.8%, highlighting the po- tential of this multimodal framework as a robust clinical decision support system in dermatology. This framework addresses two distinct clinical objectives: an image-based dual-scale pipeline optimized for the early diagnostic screening of MF versus non- MF dermatoses, and a complementary clinical metadata model designed for the subsequent staging of confirmed MF cases (patch/plaque versus tumor)
[CV-175] hinkV2V: Unleashing the Reasoning Capability of MLLM s for Instruction-Guided Video Editing
链接: https://arxiv.org/abs/2609.38541
作者: Donghao Zhou,Haoyang He,Fan Zhang,Hao Yang,Guisheng Liu,Xin Gao,Zhongwei Wan,Xingyuan Bu,Jie Wang,Qiangpeng Yang,Shilei Wen,Chi-Wing Fu,Pheng-Ann Heng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Instruction-guided video editing has made significant progress, yet existing methods use multimodal large language models (MLLMs) primarily as semantic encoders, so they often fall short in working with implicit edits that require causal or semantic reasoning. To bridge this fundamental gap in video editing, we propose ThinkV2V, a reasoning-driven framework for complex instruction-guided video editing, explicitly activating MLLM thinking before visual generation. At its core, ThinkV2V builds on a practical MLLM-to-DiT architecture to turn explicit thinking over the source video and instruction into refined conditioning signals for video editing. Further, we equip it with a dedicated training and inference recipe, combining Progressive Curriculum Training, which gradually cultivates the model from basic editing to reasoning-intensive cases, with Inference-Time Thinking Scaling, which iteratively refines candidate prompts and selects the most reliable one, to better elicit reasoning in challenging editing scenarios. We also curate the ThinkV2V-150K dataset and introduce ThinkV2V-Bench to support training and evaluation of video editing with implicit intent and causal reasoning. Experimental results demonstrate the state-of-the-art performance of ThinkV2V on both complex and standard editing scenarios, in which our 5B-scale DiT model substantially outperforms larger 10B-scale baselines.
[CV-176] GazeFlow: From Human Gaze Behavior to Generative Egocentric Gaze Prediction NEURIPS2026
链接: https://arxiv.org/abs/2609.38519
作者: Sheng Zhao,Weikai Lin,Yuhao Zhu
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026. 25 pages
Abstract:Egocentric gaze prediction enables many downstream applications but remains challenging, as human gaze is inherently stochastic. This stochasticity is constrained by structured temporal dynamics alternating between fixations and saccades, top-down influences from tasks, and bottom-up visual saliency. Based on this observation, we introduce GazeFlow, a framework that directly models gaze as a joint distribution of temporal gaze positions conditioned upon both top-down and bottom-up information. In particular, GazeFlow uses conditional flow matching (CFM): a learned velocity field iteratively transports a Gaussian noise sample into a plausible gaze trajectory drawn from this joint distribution. The velocity field is conditioned on bottom-up visual features extracted by a video encoder and on top-down task information obtained by globally querying these features. On standard datasets, GazeFlow achieves state-of-the-art performance on per-frame metrics, and the generated trajectories align better with human gaze temporal dynamics.
[CV-177] What to Attend What to Keep: Skill-Conditioned Visuotactile Representation with Progress-Guided Event Memory
链接: https://arxiv.org/abs/2609.38494
作者: Amir-Hossein Shahidzadeh,Seungjae Lee,Eadom Dessalene,Shanthosh Raaj Mohanram Mageswari,Soroush Etemad,Furong Huang,Cornelia Fermüller,Yiannis Aloimonos
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Robotic manipulation integrates vision, touch, and language, whose importance shifts across stages: vision guides reaching, while touch, through its evolution over time, decides grasping, alignment, and contact. Yet existing multi-modal manipulation policies typically use fixed temporal contexts and fusion strategies, despite shifts in what each modality contributes across different skills. We study how vision and touch should be combined at the level of primitive skills, asking what each skill needs from each sensor, and propose a skill-conditioned representation in which the queried skill conditions fusion over modality-specific short-term observation tokens while attending to a sparse event memory that retains terminal observations from the last K executed skills. Evaluated by skill progress estimation on three contact-rich tasks, it reduces slip-detection delay by 87% against fine-tuned SOTA progress models, twist-completion delay by 67.5% against a vision-only ablation, and progress error on a blind search task by 92% through sparse event memory. Gains concentrate exactly where completion is defined by contact or task history. More broadly, our results suggest that observation formation not only policy architecture is a central challenge in multi-modal representation. Project Website: this http URL
[CV-178] Gaussian Stippling: Efficient Sorting-Free 3D Gaussian Rendering through Hybrid Sampling and Spatiotemporal Reconstruction
链接: https://arxiv.org/abs/2609.38488
作者: Zijian Huang,Suiliang Mai,Chuankun Zheng,Yuan Meng,Yuchi Huo
类目: Graphics (cs.GR); Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint
Abstract:Conventional 3D Gaussian Splatting (3DGS) requires depth sorting and ordered alpha blending to correctly render overlapping Gaussian primitives. Stochastic transparency enables sorting-free rendering by replacing fractional alpha contributions with discrete stochastic visibility samples, but produces substantial spatial and temporal noise at low sample counts. We refer to this conversion from continuous Gaussian splats to discrete visibility samples as \textitGaussian Stippling. Based on this, we present an efficient order-independent rendering and reconstruction framework that operates directly on unmodified 3DGS assets. Our method adaptively integrates primitive-based and fragment-based stippling, leveraging their complementary strengths across different rendering regimes to significantly improve rendering throughput. To recover high-quality images from sparse stochastic samples, we further introduce a lightweight Gaussian-aware spatiotemporal reconstruction network. By exploiting the Gaussian attributes retained by each stipple, the network aggregates structured stochastic clues across both space and time, effectively suppressing stippling noise. Experiments show that our hybrid Gaussian stippling method, coupled with a spatiotemporal reconstruction network trained on diverse scenes, generalizes to unseen scenes and enables interactive, temporally stable, and visually plausible rendering on mobile devices without retraining or preprocessing. With scene-specific training and appropriately scaled sampling and network capacity, our method further outperforms the baselines in visual quality, offering a high-fidelity configuration for quality-prioritized applications.
[CV-179] Multidimensional Observer Model and Perceptual Dimensions of Human Image Quality Assessment NEURIPS2026
链接: https://arxiv.org/abs/2609.38487
作者: Sheng Zhao,Weikai Lin,Yuhao Zhu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at NeurIPS 2026. 25 pages
Abstract:Judging image quality is not only ecologically relevant to everyday human tasks, but also underpins many machine vision tasks such as image generation. This paper proposes a framework to understand the inherent perceptual space underlying image quality judgment in humans. We propose a multi-dimensional observer model that represents images as distributions in a latent perceptual space and that models human judgment as comparing noisy samples. Being constrained by neural representations in the primate ventral stream and fit to large-scale behavioral data, the model enables analysis of perceptual structure while matching the predictive power of existing metrics. Using this model, we find that the perceptual spaces needed to account for image quality judgment in humans are extremely low-dimensional compared to the image space even when considering its sparsity. The exact structure of the space (e.g., dimensionalities, information encoded) varies between low-level and high-level quality judgments, suggesting that, despite a shared retinal encoding in the beginning, humans selectively construct task-dependent perceptual spaces in visual decision making.
[CV-180] Beyond Layers: Position-Resolved Gradient Conflict and Position-Aware Modulation for Unified Multimodal Models
链接: https://arxiv.org/abs/2609.38485
作者: Shuyang Jiang,Fucheng Deng,Yuchuan Luo,Zhenyu Wu
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 20 pages, 10 figures, 6 tables. Code will be made publicly available upon acceptance
Abstract:Unified multimodal models (UMMs) train image understanding and autoregressive image generation on shared parameters, and the two objectives are known to interfere. Existing diagnoses and remedies operate at the resolution of layers or experts, measuring conflict per layer and resolving it by separating parameters. We argue that this resolution hides an orthogonal axis. Generation in a UMM is next-token prediction over a raster sequence of visual tokens whose roles vary systematically with position, so how strongly a generation gradient interferes with understanding should depend on where in the sequence it originates. We introduce a position-resolved interference map that attributes understanding-generation gradient conflict to visual-token positions within every layer, computed from a single backward pass at 1.2\times the cost of a standard backward pass. On Show-o and Janus-Pro, position explains a large share of conflict variance after controlling for depth (partial \eta^2=0.31 vs. 0.35 for layer on Show-o; 0.15 vs. 0.30 on Janus-Pro): the first quarter of the sequence has a mean gradient cosine of -0.18 against understanding, the last quarter -0.02 . The dependence survives per-position gradient-norm normalization, retaining 80% of its effect size, and conflict strength tracks semantic content (Spearman \rho=0.64 ). Building on the map, we propose position-aware modulation (PAM), which removes the anti-aligned component of generation gradients only at high-conflict positions without changing the architecture. Under a matched trainable-parameter budget, PAM improves over layer-wise separation by +21 MME and +2.4 GenEval points on Show-o while matching it on POPE and overall FID; a random-position control recovers about 31% of the gain. Position-based and layer-based separation are complementary degrees of freedom and can be combined.
[CV-181] Caption-Mediated Perceived-Safety Estimation for Pedestrian Routing
链接: https://arxiv.org/abs/2609.38479
作者: Simon Parkinson,Paloma Liu,Wei Zheng,Mohammadreza Sheikhfathollahi
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:This paper presents an explainable approach to pedestrian routing, in which perceived safety is estimated from street-level imagery through an explicit natural-language intermediate representation. A vision–language model caption is generated and stored before any scoring is undertaken, and the perceived-risk class is derived entirely from structured features of that stored text, so that every segment score remains inspectable by the user. Nine captioning conditions across five model families are benchmarked against a direct Contrastive Language–Image Pre-training (CLIP) image-embedding baseline under an identical downstream pipeline, and the caption-mediated representation is found to reach parity with the image embedding rather than to trail it. The approach was deployed over 654,115 images covering 36 electoral wards in two locations in Northern England (Manchester and Huddersfield). Independent field validation against 3,669 locally collected ratings of 494 images across 70 participant sessions established agreement that is statistically significant but modest, at r=0.262 , against a measured noise ceiling of 0.737 imposed by disagreement between raters. A single-use confirmatory test then found that a pipeline 44% stronger on the supervised benchmark did not produce measurable improvement in the field ( r=0.250 , p=0.84 ), so the benchmark gains did not predict the deployment gains in this case. Routing behaviour varies systematically with journey length. There is negligible change below 1,km, reaching a median increase of 12.78% in low-risk route length for a median detour of 2.73% on journeys of 3 to 6 km.
[CV-182] Curating Synthetic Data for Task-Specific Visual Perception
链接: https://arxiv.org/abs/2609.38476
作者: Saptarshi Neil Sinha,Paul Julius Kühn,Michael Weinmann
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Synthetic data are most valuable where general-purpose datasets cannot provide the domain-specific priors a task requires, and where manual annotation is expensive, imprecise, or infeasible. In this article we argue that the central question for specialized vision systems is not how to generate more data, but which data to generate. We therefore discuss curated synthetic data, whose scene content, appearance variations, sensing characteristics, and annotations are deliberately designed around a given task. We examine three complementary curation paradigms. Procedural rendering offers explicit control over scene parameters and the annotations follow by construction. Physically-based simulation encodes the mechanism behind an observed effect and yields exactly aligned supervision pairs. Generative AI learns sensor-specific appearance from small real seed sets and attains plausible realism, though it remains prone to hallucination and to inaccurate annotation. These paradigms are illustrated with examples from industrial surface defect detection, restoration of degraded digitized autochrome plates, and 6DoF pose estimation from RGB and event data. Using these examples, we analyze different data regimes and training strategies that combine synthetic and real data across these paradigms. We conclude that curated synthetic data are best understood as a complement to real observations, and that hybrid pipelines combining controllable supervision with learned appearance are the most promising direction for reliable sim-to-real transfer.
[CV-183] PAMI: Part Anchored Motion for Text to Human-Object Interaction Generation
链接: https://arxiv.org/abs/2609.38466
作者: Chuqiao Li,Xianghui Xie,Yong Cao,Andreas Geiger,Gerard Pons-Moll
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Text-conditioned full-body human-object interaction (HOI) generation requires synthesizing human motion and object trajectories that match the input text while remaining precisely coordinated over time. Most methods represent the human and object as separate trajectories and predict the global human-object couplings. Learning this complex, dynamically changing relationship implicitly, however, often yields object drift, missed contact, and penetration. We introduce PAMI, a Part-Anchored Motion framework for Interaction generation. Inspired by the classic Hough Transform, our key idea is to localize object motion by letting body-part anchors vote for it: we express object motion relative to multiple body-part anchors and use PamiVAE to learn an interaction latent space, decoding frame-wise weights that aggregate these part-specific votes. Building on this representation, PAMI generates interactions in a coarse-to-fine hierarchy. PamiGen first generates a coarse human-object interaction from text in this structured latent space, and PamiRefiner then recursively resolves fine-grained contact geometry using a hybrid surface-sensing representation, combining long-range probes that capture overall body-part influence with short-range sensors that resolve detailed contacts near the object surface. Experiments on InterAct show that PAMI generates more faithful interactions and more accurate human-relative object motion than previous methods, achieving 14.5% higher contact recall than the previous state of the art. Extensive ablations validate the contributions of both the part-anchored voting representation and hybrid surface-sensing refinement.
[CV-184] Does Gradient Conflict Predict the Understanding–Generation Trade-off? A Controlled Audit of Conflict-Metric Validity in Unified Multimodal Models
链接: https://arxiv.org/abs/2609.38465
作者: Shuyang Jiang,Fucheng Deng,Yuchuan Luo,Zhenyu Wu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 23 pages, 10 figures, 5 tables. Code will be made publicly available upon acceptance
Abstract:Unified multimodal models (UMMs) are increasingly designed around gradient conflict between understanding and generation objectives. The premise that reducing these metrics improves the downstream understanding-generation trade-off has never been tested directly. We audit it in a controlled testbed, GRIDUMM, which mirrors key structural ingredients of UMM training while making the ground-truth trade-off exactly computable. Across 63 configurations and 372 measured checkpoints, no directional conflict metric reaches an absolute Spearman correlation of 0.3 with a confidence interval excluding zero for conflict measured during training against the eventual trade-off. A dose-response intervention that monotonically suppresses conflict leaves the trade-off flat, separating correlation from causation. The norm ratio is a generation-failure detector and becomes null among configurations that master generation. Functional interference measures outperform directional conflict metrics, while training loss tracks the trade-off strongly. Our results do not show that conflict is useless; they show that its validity as a diagnostic target must be established, not assumed, and we release the audit protocol as a reusable standard.
[CV-185] rafficSignBench: Rule-Centric Closed-Loop Evaluation of Traffic-Sign Compliance in Autonomous Driving
链接: https://arxiv.org/abs/2609.38463
作者: Victoria Smirnova,Viktoriia Zinkovich,Gregorii Bukhtuev,Artem Belyaev,Andrey Kuznetsov,Denis Shepelev,Vlad Shakhuro
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Autonomous driving planners are typically evaluated using aggregate metrics such as driving score, destination rate, and collision rate, which do not explicitly measure compliance with traffic rules. As a result, planners can achieve high benchmark scores while still exhibiting unsafe or illegal behaviors, limiting their applicability to real-world deployment. To address this gap, we introduce TrafficSignBench, a large-scale, traffic sign-centric benchmark for systematic and interpretable evaluation of traffic-rule compliance in autonomous driving. Our framework combines real-map-based simulation for realistic road layouts with rule-targeted procedural scenario generation for scalable and balanced coverage of underrepresented rules. We implement traffic rules corresponding to 34 traffic signs, each equipped with an automatic rule checker for detecting violations during closed-loop execution. This design yields 29,000 diverse road scenes and 29 distinct testing scenario types, enabling controlled evaluation of rule-specific planner behavior. We construct 5,800 testing scenes and demonstrate that current autonomous driving planners can exhibit poor traffic-rule compliance despite strong performance on standard evaluation metrics. To address this limitation, we transform existing planners into rule-compliant trajectory experts via explicit traffic-sign constraints, enabling scalable generation of high-quality oracle trajectories for fine-tuning.
[CV-186] Audible World Models: Spatially Aware Sound Generation for 3D Worlds NEURIPS2026
链接: https://arxiv.org/abs/2609.38444
作者: Duowen Chen,Jinjin He,Gouthaman KV,Sandeep Bangalore Venkatesh,Bo Zhu
类目: Computer Vision and Pattern Recognition (cs.CV); Sound (cs.SD)
备注: Accepted at the 40th Conference on Neural Information Processing Systems (NeurIPS 2026)
Abstract:Text- and image-conditioned world generators can create visually rich 3D environments, yet these worlds often remain silent or rely on soundtracks synthesized solely from text or rendered video. Although such audio can convey what should be heard, it lacks an explicit representation of where sound sources are located and how their perceived sound should vary with listener movement. We introduce Audible World Models, a training-free framework that incorporates sound into the generated world state. Starting from a text prompt, our system constructs a panoramic 3D proxy, separates it into semantic layers, identifies sound-producing foreground objects and ambient background regions, and synthesizes dry audio for each sound label. It then anchors these sources to reconstructed geometry and renders listener-dependent spatial audio using geometric acoustic propagation. By explicitly linking semantics, geometry, and sound propagation, the framework maintains persistent source locations while adapting the rendered audio to changes in listener viewpoint and motion. Experiments across 80 generated scenes demonstrate substantial gains in spatial consistency over text-, video-, and panorama-conditioned baselines, while preserving competitive semantic alignment. VLM-based assessments and human evaluations further indicate that our soundtracks are preferred for their audio-visual consistency, spatial plausibility, and motion-dependent behavior.
[CV-187] BIND: Binding 3D Robot Actions to 2D Image Features
链接: https://arxiv.org/abs/2609.38443
作者: Cameron Smith,Arsh Tangri,Vitor Guizilini,Yue Wang,Zubair Irshad,Sergey Zakharov
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We introduce BIND, a new action representation for visuomotor robot policies that binds 3D robot actions to their corresponding 2D image features, yielding strong data efficiency gains and robustness to out-of-distribution object positions and camera viewpoints. The action heads of current robot policies are typically formulated as an MLP regression from a single global feature vector produced by a pre-trained vision encoder. This global formulation requires the policy network to discover, from demonstrations alone, the relationship between target robot actions and the image features they project onto. The consequence is that although modern image features are semantically descriptive, spatially robust, and even multiview-consistent, the policies built on them are brittle to subtle changes in camera viewpoint and object placement–and surprisingly data-inefficient. BIND closes this gap by supplying the action-feature relationship through camera geometry rather than learning: it discretizes a volume of candidate end effector positions, attaches each candidate to the pre-trained features at its projection in each camera view, and selects actions by scoring each candidate’s position and image-bound feature combination. On a real robot, we study data efficiency and out-of-distribution robustness to unseen object positions and camera viewpoints, as well as general long-horizon task execution and dexterity. We find BIND to be highly data-efficient and robust: it achieves near-perfect success on tasks with as few as 5 demonstrations, and degrades gracefully under steep camera-viewpoint shifts and held-out object positions where coordinate-regression baselines completely fail.
[CV-188] MOBA-VL: Event-Localized Multi-Turn Reinforcement Learning for Real-Time MOBA Commentary
链接: https://arxiv.org/abs/2609.38428
作者: Shengyun Zhong,Xinkang Zhao,Ziyuan Chu,Linchao Zhu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 30 pages, 12 figures
Abstract:Real-time commentary for Multiplayer Online Battle Arena (MOBA) esports requires a vision-language model (VLM) to narrate a live match second by second, both fluently and accurately. Existing streaming VLMs sound natural but often miss key events such as kills and objectives. To address this limitation, we use game telemetry, which records exactly when each event occurs, as a supervision signal. We introduce MOBA-VL, a 9B-parameter model trained on this signal with event-localized multi-turn reinforcement learning, which rewards the turns that describe each event. We also collect MOBACast, 860 professional matches (about 460 hours) across three MOBA games with word-level timestamped commentary, and MOBACast-Bench, a benchmark from held-out tournaments. On MOBACast-Bench, MOBA-VL achieves the highest Overall score on full matches (63.25 vs. 55.12 for StreamingVLM) and clips (63.45 vs. 56.22 for DeepSeek-V4.1-Flash). Event-localized credit also raises event recall from 34.5 to 42.1 over supervised fine-tuning. Code and data will be released, and demos are available on an anonymous project page at this https URL.
[CV-189] VidHarness: Evolving Agent Harnesses for Cost-Efficient Long Video Understanding
链接: https://arxiv.org/abs/2609.38413
作者: Susan Liang,Jianmin Wu,Daxiang Dong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-language models (VLMs) can answer questions about hour-long videos, but processing every frame is prohibitively expensive, even though the evidence for a question usually spans only a few seconds. Video agents, i.e., harness programs wrapped around a frozen VLM, address this by observing the video selectively, yet existing harnesses are hand-crafted by experts through slow build-and-test cycles. We propose VidHarness, a framework that automates harness design for cost-efficient long video understanding, in which a harness proposer iteratively evolves harnesses based on execution feedback from an evolution environment. To escape the local optima of greedy refinement, we organize the evolution as Monte Carlo tree search (MCTS), and to reduce the evaluation cost, we integrate uncertainty-aware multi-fidelity validation, which screens new harnesses on a few questions and promotes only the promising ones. Since the best harness varies with the frame budget, we further introduce a mixture-of-harness that routes each question to a harness specialized for its budget. VidHarness sets new state-of-the-art results on LongVideoBench, Video-MME, and Video-Holmes, outperforms the strongest hand-crafted video agent by up to 11.2 points, and generalizes to the knowledge-intensive benchmarks Video-MMMU and MMVU with fewer than half of the frames of uniform sampling.
[CV-190] am MSU GenText-Forensics Challenge 2026 Technical Report
链接: https://arxiv.org/abs/2609.38391
作者: Kirill Koltsov,Aleksandr Gushchin,Dmitriy Vatolin,Anastasia Antsiferova
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Document text forgery has evolved beyond simple pixel-level manipulation: modern attacks alter not only the appearance of a document but also its meaning, and increasingly target the OCR LLM pipelines that consume such documents. The ACM MM 2026 GenText-Forensics challenge therefore requires systems that not only decide whether a multilingual text image is forged, but also localize the point of manipulation, identify the attack type, and produce a human-readable forensic report with supporting evidence. We present our solution, a decomposed chain-of-thought (CoT) pipeline that combines a document tampering detector (DTD) with two Qwen3-VL-32B vision-language models, each LoRA-adapted to a distinct sub-task. DTD produces tampering probability maps that are converted into numbered candidate regions; a first model (the Filterer) validates these regions and assigns a preliminary forgery type, while a second model (the Semantic Detective) merges and re-grounds the surviving regions, searches for purely semantic anomalies that are invisible to pixel-level detectors, and writes the final report. Both models are trained by distilling chain-of-thought traces from a privileged Qwen3-VL-235B teacher that has access to ground-truth masks and reports. Our approach secured third place in the ACM MM 2026 GenText-Forensics challenge. We describe the data preparation, test-time augmentation, region rendering, distillation protocol, and training configuration in detail, and report ablations over detector thresholds, prompt designs, and pipeline decompositions.
[CV-191] PhyProbe: Rethinking Physical Consistency Evaluation in Generated Videos NEURIPS2026
链接: https://arxiv.org/abs/2609.38377
作者: Max Ku,Jiaojiao Fan,Zekun Hao,Francesco Ferroni,Heng Wang,Wenhu Chen,Ming-Yu Liu,Prithvijit Chattopadhyay
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: NeurIPS 2026 poster
Abstract:Evaluating the physical consistency of generated videos remains a fundamental challenge. Existing approaches rely on off-the-shelf vision-language models, which can often be myopic to physical dynamics, or fine-tuned evaluators trained on human annotations, which overfit to dataset-specific cues and fail to generalize. A key challenge is that existing supervision sources provide either relative ordering or absolute scores, but not both reliably and consistently across varied settings. To this end, we introduce PhyProbe, an evaluator that extracts features from a frozen pretrained spatio-temporal encoder and maps them to a scalar physical consistency violation score via a lightweight scoring head. PhyProbe is trained through a unified objective combining pairwise ranking, regression on noisy scalar annotations, and anchor-based calibration over a curated set of heterogeneous supervision sources. Experiments show that PhyProbe outperforms prior methods on most pairwise benchmarks spanning real-generated and generated-generated pairs under varying correspondence, with the largest gains in no-correspondence and generated-generated settings where existing fine-tuned evaluators degrade sharply. PhyProbe achieves strong correlation with human judgments, with close agreement between rank-based and linear metrics, indicating that scores are both well ordered and anchored to a stable [0, 1] scale. Further, despite being trained on supervision indicative of physical consistency, without explicit general-preference labels, PhyProbe also performs competitively on human preference benchmarks: consistent with the observation that physics violations are entangled with broader quality degradations.
[CV-192] Composition Not Conversation: VLMs Lose the Scene Not the Thread
链接: https://arxiv.org/abs/2609.38368
作者: L. D. M. S. Sai Teja,Ufaq Khan,N. Siva Gopala Krishna,Satyajit Tourani,Ashshak Sharifdeen,Fida Mohammad Thoker,Bernard Ghanem,Muhammad Haris Khan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 33 pages, 10 figures, 11 tables. Code: this https URL . Dataset: this https URL
Abstract:Vision-language models (VLMs) increasingly reason over visual evidence that is cropped, segmented, retrieved, or revealed over time. Yet most VQA benchmarks present the complete image and question at once. We ask what models lose when the same information is fragmented. We introduce Layered-VQA, with 93 scenes and 300 questions. Each image is decomposed into ordered RGBA layers that exactly recompose the original scene, and each question is annotated with supporting, minimal-sufficient, and distractor layers. We evaluate eleven open-weight VLMs from 3B to 32B parameters and two proprietary models with a scale of 187,200 conversations, graded by 1.74M open-model cross-judgments. We find three consistent failures. Loss in Composition: fragmenting the question has a small effect, but fragmenting the scene substantially reduces accuracy; recomposing the same layers largely restores performance. Oracle Inversion: even oracle-selected sufficient evidence can perform worse than the complete scene. Loss in Grounding: as more evidence is required, grounding degrades much faster than answer accuracy. Together, these results show that having the right visual evidence is not enough. How that evidence is composed and presented determines whether models can use and ground it. The right evidence is not enough: VLMs need the scene it came from.
[CV-193] Inductive Visual Logic for Few-Shot Out-Of-Distribution Adaptation in VLMs ECCV2026
链接: https://arxiv.org/abs/2609.38362
作者: Hung-Jen Chen,Yu-Heng Ho,Ting-Yao Huang,Po-Hsiang Hsu,Li-Yu Chen,Chun-Yi Lee,Min Sun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by ECCV 2026
Abstract:Generative vision-language models (VLMs) such as Qwen-VL and LLaVA achieve strong zero-shot performance on tasks overlapping with their pretraining distribution, yet fail on specialized domains where the required discriminative features were never learned, a regime we term distant out-of-distribution (OOD). Standard adaptation methods cannot overcome this representational absence because they operate within the encoder’s existing feature space. However, VLMs retain a robust descriptive capacity even when discrimination collapses: a model that cannot classify a medical scan can still articulate its visual patterns. Exploiting this asymmetry, we introduce Inductive Visual Logic (IVL), a training-free framework that constructs classification knowledge from the model’s surviving descriptive ability. IVL extracts visual traits from few-shot support images through dual-mode prompting, combining semantic descriptions with primitive visual observations, and organizes them into per-class trait dictionaries. At inference, hierarchical filtering identifies spatially grounded trait evidence for classification. Across multiple distant-OOD benchmarks, IVL achieves the highest aggregate accuracy under two VLM backbones while producing interpretable, trait-traceable predictions.
[CV-194] rackFish3D: Self-Supervised 3D Tracking of Schooling Fish from Multi-view Videos NEURIPS2026
链接: https://arxiv.org/abs/2609.38347
作者: Patt Phurtivilai,Zhiyang Dou,Yifan Wu,Kinfung Chu,Yuan Liu,Lei Yang,Wenping Wang,Taku Komura
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: to be published in the 40th Conference on Neural Information Processing Systems (NeurIPS 2026)
Abstract:Quantifying collective fish behavior requires accurate trajectories, yet multi-view 3D tracking remains challenging due to frequent occlusions, visually similar individuals, and the long-standing scarcity of identity annotations. We present TrackFish3D, a geometry-driven self-supervised framework for dense multi-camera 3D tracking of schooling fish. Instead of relying on appearance-based re-identification or manually annotated identities, TrackFish3D turns calibrated multi-view geometry into supervision: triangulation and reprojection consistency provide pseudo-associations, while a geometric encoder and global association transformer learn all-to-all cross-view correspondence within each frame. To make these associations identity-aware, TrackFish3D introduces a self-supervised contrastive objective that separates co-visible individuals in the embedding space, together with a temporal predictor that preserves identities and bridges short occlusions across frames. The resulting model is trained once on unlabeled footage and applied directly to unseen test videos, requiring no cross-view identity labels, temporal annotations, 3D ground truth, appearance features, or test-time optimization. On our benchmark, TrackFish3D improves 3D Multi-Object Tracking Accuracy from 87.7% for the strongest baseline to 95.8%. On the 3D-ZeF zebrafish benchmark, it achieves 81.1% MOTA, compared with 77.4% for the best geometric baseline. TrackFish3D also generalizes beyond fish, achieving strong results on real-world bird tracking.
[CV-195] Learning Semantic Inpainting for Animatable Gaussian Head Avatars
链接: https://arxiv.org/abs/2609.38343
作者: Pilseo Park,Fizza Rubab,Yiying Tong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We present SInGA, a novel method for learning Semantic Inpainting for animatable Gaussian head Avatars from a single image. Existing avatar approaches often rely on multi-view observations and lack effective handling of unobserved regions in single-view settings, limiting their applicability in such scenarios. To address this, we propose a semantic inpainting framework defined in UV space for completing unobserved facial regions. Our key insight lies in the structured topology of the UV representation, which provides consistent spatial correspondences and enables reliable completion of identity-specific features using the inherent symmetry cues of human faces. We extract features from observed regions and use them to complete unobserved regions. The completed representation is then used to regress Gaussian attributes, effectively performing Gaussian inpainting. In addition, instead of relying on a single Gaussian at each surface or pixel location, we stack multiple Gaussians to enhance detail. The resulting avatar generalizes across identities without requiring per-identity optimization and can be animated with driving inputs. Experimental results show that our method generates high-quality head avatars with improved completeness and identity preservation, while supporting realistic animation and consistent rendering from unobserved views.
[CV-196] ExploreNet: Learning Where to Explore in Diffusion GRPO
链接: https://arxiv.org/abs/2609.38329
作者: Shuyue Stella Li,Xiaochuang Han,Yulia Tsvetkov,Luke Zettlemoyer
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 23 pages, 15 tables, 8 figures
Abstract:Group-relative RL methods such as Flow-GRPO post-train image generators by exploring with isotropic Gaussian noise added at every denoising step. This noise decides which rollouts the model learns from, yet it perturbs every channel and spatial position of the latent equally. In this paper, we instead show that latent elements differ in how much they change the generated image, so exploration should adapt to these differences. We introduce EXPLORENET to learn an adaptive exploration distribution. EXPLORENET is a policy that predicts a noise scale for every latent element from the current latent, the denoising step, and the prompt, before any reward is observed; it is trained on the reward spread of each rollout group and discarded after training, leaving inference unchanged. On Stable Diffusion 3.5 Medium, EXPLORENET improves held-out GenEval2 by 14% over Flow-GRPO, transfers to two independent compositional benchmarks and five preference and image-quality models, and reaches a 67.2% human preference win-rate. Overall, across our group-relative diffusion RL experiments, we find that exploration is learnable, the shape of the exploration distribution outweighs its magnitude, and rollout quality is more effective than rollout quantity.
[CV-197] Strike a Chord! Modal Kinetic Typography
链接: https://arxiv.org/abs/2609.38325
作者: Maham Tanveer,Jiyeon Han,Nanxuan Zhao,Hao Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: Project page: this https URL
Abstract:We introduce modal kinetic typography, which animates a vector glyph to express a semantic concept while keeping it legible. Our key idea is to build motion from the glyph’s natural vibration modes. Specifically, a finite-element eigenproblem assembled from the vector outline yields the glyph’s softest modes, for the whole letter and for each of its parts, allowing it to bend. The problem’s zero-energy solutions, i.e., rigid translations and rotations, are applied in closed form to each part, allowing parts to also move as blocks. To animate the glyph, a frozen video diffusion model supervises only the modes’ amplitudes and phases. Our modal approach addresses two weaknesses of prior work. Free-form point optimization under video score distillation (SDS) moves each point and frame independently along noisy gradients, tearing the outline and causing jitter. In contrast, our modes are smooth along the outline and driven by a few whole-cycle harmonics, which restricts these gradients to smooth, seamlessly looping motion. On the other hand, structured alternatives rely on skeletons or keypoints from category-specific priors, whereas our modes come from the glyph itself; the only prior is a list naming each letter’s moving parts, generated once for the whole alphabet by a language model. In modal kinetic typography, shape and motion are disentangled by construction: a single base outline is sculpted toward the concept, and the modal drive cannot alter it, so a letter can also be animated without being reshaped. Our method produces more articulated and smoother motion than Dynamic Typography and AniClipart at comparable or better concept alignment, with less glyph tearing than Dynamic Typography, and is preferred by human raters, including in a frozen-shape setting where motion alone must carry the concept. Our results were also preferred over Astra (GPT-6) by human raters.
[CV-198] It Takes Little to Rewrite Perception: Targeted Semantic Substitution in Vision-Language Models at εleq 4/255
链接: https://arxiv.org/abs/2609.38298
作者: Binchi Zhang,Atrisha Sarkar,Apurva Narayan
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Vision Language Models (VLMs) are widely deployed in safety-critical scenarios, and understanding to which extent they can be controlled by adversarial perturbation is a prerequisite for evaluating their trustworthiness. Existing representation-alignment attacks, which make a VLM perceive a target image, achieve limited success at \varepsilon \leq 4/255 . Therefore, VLMs seems robust to perturbations in this range. We show that this robustness does not hold, as targeted semantic substitution succeeds within the same range. Specifically, we align each stream of the source image with its counterpart in the target image in the victim VLM’s post-merger token space, operating under a white-box threat model. We evaluate under a strict success criterion, requiring the model to simultaneously name the target, confirm its presence, and deny the source. In images, target semantics appear at \varepsilon = 2/255 and complete replacement reaches 38% at \varepsilon = 4/255 . On video, complete replacement reaches 35.9% at \varepsilon = 1/255 . We also observe a phenomenon of \textitsemantic fusion, where Large Language Model (LLM) rationalizes contradictory visual signals into a coherent narrative.
[CV-199] GaugeVLM: Structuring Spatial Supervision with Measured Geometric Interventions
链接: https://arxiv.org/abs/2609.38285
作者: Hongbo Wang,Zihan Lin,Wenkui Yang,Shiran Ge,Yuang Ai,Jie Cao,Huaibo Huang,Ran He
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Vision-language models (VLMs) can contradict themselves across views of the same spatial relation and fail to respond when that relation changes. Addressing these failures requires supervision that captures error magnitude and geometric dependencies across observations, both of which remain implicit in training on individual answers or ordinal preferences. Therefore, we introduce GaugeVLM, which makes this structure explicit through controlled object and camera interventions in explicit 3D scenes, producing linked observations with measured differences between spatial relations and shared truths across views. To translate this structure into learning signals, its core objective, GaugeDPO, converts measured errors into preference margins, directly supervises correct canonical rankings across views, and links intervention-induced answer-odds contrasts to measured relation changes with view-specific scales. Our analysis bounds canonical prediction error and establishes that the cross-view and intervention constraints can be jointly satisfied. Empirically, GaugeVLM improves all 10 established spatial metrics over supervised fine-tuning across three VLM backbones, with the main 7B model gaining 15.0 and 18.9 percentage points on MSMU distance and QSpatial+, respectively. These gains also extend to autonomous driving and embodied reasoning, demonstrating the robust generalization across domains.
[CV-200] Masked Swingers: Harnessing Data Augmentation to Advance Autoencoders for Self-Supervised Learning
链接: https://arxiv.org/abs/2609.38278
作者: Anthony Fuller,Scott C. Lowe,Daniel G. Kyrollos,Graham W. Taylor,Evan Shelhamer,James R. Green
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Self-supervised learning (SSL) removes the need for annotations and makes models that are capable across more domains than supervised learning. The autoencoder SSL framework learns by reconstructing its own input after information loss through a bottleneck or noise injection. Masked autoencoders (MAE) are the most successful instantiation of this framework: they encode a random subset of patches, then decode the masked-out patches. In this work, we introduce key modifications to improve MAEs. Our method augments an image in two different ways, then masks and encodes each view separately. It then exchanges the global representations (CLS tokens) between views before decoding the masked patches. By design, our Masked Swingers encourages learning a view-agnostic summary of the image to facilitate efficient transfer. We perform extensive experiments, and find Masked Swingers outperforms MAE by +3-5% on ImageNet-1K kNN and provides large gains on fine-grained tasks, e.g., relative gains of +45% on instance retrieval, +22% on animal re-ID, and +76% on Omniglot character recognition. To boot, Swingers reduces error -64% relative to MAE on three new state-probing datasets, opening the door to world modeling. Welcome to our Swingers party.
[CV-201] Evaluating Multi-Task Morphological Concept Learning for Pulmonary Nodule Malignancy Assessment in 3D CT
链接: https://arxiv.org/abs/2609.38271
作者: Namitha Narayanan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 8 pages, 3 figures, 5 tables
Abstract:Morphological characteristics such as spiculation and lobulation play an important role in assessing pulmonary nodules on computed tomography (CT), particularly in relation to malignancy risk. This study examines whether learning radiologist-annotated morphological features together with malignancy risk from lesion-centred 3D CT volumes improves classification performance. The Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) dataset was used, comprising 3,918 reader-level nodule annotations from 742 patients after excluding indeterminate malignancy ratings. Patient-level splitting was used for training, validation, and testing, with 112 patients and 628 reader annotations in the held-out test set. A single-task 3D convolutional neural network was compared with a multi-task model predicting malignancy risk, spiculation, and lobulation. The single-task model achieved a balanced accuracy of 0.548 and receiver operating characteristic area under the curve (ROC-AUC) of 0.552, while the multi-task model achieved 0.539 and 0.558, respectively. Patient-level bootstrap analysis showed an ROC-AUC difference of 0.005 (95% confidence interval (CI): -0.087 to 0.090) and a balanced-accuracy difference of -0.009 (95% CI: -0.067 to 0.043). The auxiliary tasks were strongly imbalanced and showed limited predictive performance. Overall, including morphological features did not clearly improve malignancy-risk classification, showing the importance of class balance, label formulation, and reader-level annotation structure in multi-task pulmonary CT analysis.
[CV-202] Beyond Missing Rates: Rethinking Incomplete Multi-View Clustering with Protocol Divergence NEURIPS2026
链接: https://arxiv.org/abs/2606.04857
作者: Haolu Liu,Xiyue Wang,Xuanting Xie,Liangjian Wen,Zhao Kang
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV); Neural and Evolutionary Computing (cs.NE)
备注: Accepted by NeurIPS 2026 as a poster paper
Abstract:Incomplete multi-view clustering (IMVC) is typically evaluated by retraining separate models under different missing-view configurations. Evaluations indexed only by nominal missing rate can overlook differences in observation structure across missing-view protocols. We show that missing-data protocols with identical nominal missing rates can induce substantially different learning regimes, differing by approximately 50-fold in the proportion of fully observed samples. We formalize this phenomenon as protocol divergence, which quantifies structural disparities among missing-view protocols beyond marginal missing rates. Furthermore, we analyze support-gated reconstruction mechanisms and show that their optimization contribution is inherently limited by the frequency of eligible observations under explicit normalization and optimization conditions. Based on these observations, we propose CRAFT (Co-occurrence-free Robust Attention-masked Fusion Transformer), a train-once framework that combines representation learning with an architecture designed to process missing-view inputs. CRAFT combines (i) per-sample forward computation using each sample’s observed views and shared parameters, and (ii) mask-aware fusion over nonempty observed-view subsets. The deployment evaluation starts from training data with all views available and reuses one final checkpoint per dataset and seed across missing-view protocols without retraining. Experiments on CUB and MultiFashion show that CRAFT achieves the strongest performance in 12 out of 13 information-matched settings. Additional deployment experiments across seven benchmarks and sixteen missing configurations demonstrate substantial computational savings through checkpoint reuse while maintaining competitive clustering performance. Code and evaluation tools: this https URL and this https URL.
[CV-203] ssue Detection Determines False Positives in Diffusion-Based Histopathology Artifact Detection
链接: https://arxiv.org/abs/2609.40083
作者: Konstantinos Moutselos,Ilias Maglogiannis
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注: 29 pages, 3 figures, 4 tables, including supplementary material. Submitted to Computerized Medical Imaging and Graphics. Code, data and models: this https URL
Abstract:One-class artifact detectors for whole-slide images learn normal tissue from a clean training pool and flag departures from it. The pool is built by a preprocessing pipeline whose tissue-detection step is usually treated as neutral. We tested whether it is. On 16 annotated TCGA slides, we rebuilt the clean pool of a diffusion-based detector with different tissue detection methods and compared the resulting models in a four-fold cross-validation. Per-slide saturation-Otsu detection excluded normal tissue, chiefly tissue with large clear spaces such as adipose tissue and alveolar parenchyma, and on slides with thick marker ink kept the ink while excluding ordinary tissue. Replacing it with entropy-based detection reduced the false-positive fraction on held-out clean slides from 0.102 to 0.016, in every fold and with a second training seed, without loss of sensitivity; the gain came from the composition of the pool, not its size. Across three tissue detection methods, false positives followed the fraction of such clear-space tissue in the pool, a statistic that needs no labels or training (0.103, 0.016 and 0.009). The effect did not carry over at the same size to a nearest-neighbour detector on foundation-model features. On an external cohort, the curated pool lowered clean-control false positives by about 20%, far less than within TCGA, and the remaining cross-center loss was not explained by stain differences. For one-class quality control, tissue detection decides what the model learns as normal and should be chosen and reported accordingly.
[CV-204] MAGiDiff: Sampling the Photospheric Vector Field from UV/EUV Filtergrams
链接: https://arxiv.org/abs/2609.40043
作者: Ruoyu Wang(NYU),David Fouhey(NYU)
类目: olar and Stellar Astrophysics (astro-ph.SR); Instrumentation and Methods for Astrophysics (astro-ph.IM); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Photospheric vector magnetic fields are foundational to modeling, understanding, and forecasting solar activity. These data are usually produced by inverting and disambiguating the full Stokes vector at multiple passbands, which is demanding. Here, we investigate how well we can estimate photospheric vector magnetograms from UV/EUV filtergrams. This problem is challenging and intrinsically ambiguous without polarization information, as the mapping from UV/EUV intensity to the magnetic field is indirect and ill-posed. We introduce MAGiDiff, a machine-learning-based method that uses denoising diffusion models to estimate vector magnetograms from UV/EUV filtergrams. As input, MAGiDiff takes a stack of filtergrams from the Solar Dynamics Observatory (SDO) / Atmospheric Imaging Assembly (AIA); as output, it is trained to estimate the disambiguated vector magnetogram as seen by Hinode / Solar Optical Telescope-Spectro-Polarimeter (SOT-SP). We show that MAGiDiff can accurately mimic the Hinode ground-truth. Additionally, we probe MAGiDiff’s understanding of the physical structure and magnetic connectivity. On full-disk, we show that it produces plausible structures for active regions. MAGiDiff generalizes across solar cycles despite hemispheric polarity reversal, and can be fine-tuned to other EUV instruments including STEREO/EUVI and GOES-R/SUVI. While clearly not a substitute for a dedicated instrument, MAGiDiff opens the door to new capabilities.
[CV-205] Joint Supervised and Self-Supervised Training with Acquisition-Robust Techniques for Accelerated 4D Flow MRI Reconstruction
链接: https://arxiv.org/abs/2609.38644
作者: Mengyuan Xue,Bochun Mei
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:4D flow MRI measures time-resolved, three-directional blood velocity but requires long acquisition times, and its diagnostic signal is carried by the phase difference \emphbetween velocity encodings, not by image magnitude. Recent work has developed a per-encoding variational network to address image reconstruction in this field. In this work, we incorporate a joint supervised and self-supervised training regime and utilize both magnitude and velocity data during supervision. At the same time, we add multiple acquisition-robust and conditioning strategies based on the acceleration factors. On the CMRx4DFlow~2026 aortic dataset (1.5 and 3T), our model lowers RelErr by 38 – 50% and AngErr by 7.0 – 8.7^\circ against a training-matched baseline across R=10 – 50 , improving on every held-out subject at every acceleration. Our model also shows strong generalization ability to transfer on out-of-distribution data by employing the joint training scheme, with an increase of 7.2% in SSIM and decrease of 38% and 31% in AngErr and RelErr respectively.
[CV-206] SGL: Teacher-Student Graph Learning for 3DGS Compression ICASSP2027
链接: https://arxiv.org/abs/2609.38635
作者: Matin Bani Saedi,Matthew Kyan,Gene Cheung
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 2 figures. Submitted to IEEE ICASSP 2027
Abstract:3D Gaussian Splatting (3DGS) is a popular representation for novel view synthesis. However, 3DGS contains millions of Gaussian primitives, each with rich attributes, resulting in large file sizes. We propose a novel 3DGS compression method based on Teacher-Student Graph Learning (TSGL) that operates directly on a trained model, without 3DGS retraining or access to training images. Specifically, for each block of Gaussian primitives, using decoded positions and DC spherical harmonic (SH) coefficients as predictors, we learn a signal-dependent geometry graph G encoding the pairwise similarities between neighbouring Gaussians via a teacher-student model. Given G, we perform Graph Fourier Transform (GFT) on the remaining attributes, so that signal energies are predominantly projected into the low-frequency coefficients for compact representation. On three standard benchmarks, the method reaches 27x to 33x compression with less than 0.6 dB of PSNR loss, improving on recent post-training compression methods in both size and rendering quality.
[CV-207] Colorectal Cancer Segmentation with Adaptive Augmentation and Multi-Resolution Ensemble Models
链接: https://arxiv.org/abs/2609.38419
作者: Ümit Mert Çağlar,Alptekin Temizel
类目: Image and Video Processing (eess.IV); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: SPIE Eighteenth International Conference on Machine Vision (ICMV 2025), Paris, France
Abstract:Colorectal cancer (CRC) is the second most deadly and third most common cancer, and the leading cause of death among gastrointestinal cancers. Early diagnosis is crucial for the treatment of this cancer and increasing the survival rates. Although CRC is more common in developed regions, its occurrence is also increasing in developing regions as well. CRC diagnosis relies on histopathology assessment post-biopsy. Automated deep learning algorithms can significantly reduce diagnosis time, enhancing efficiency and supporting timely clinical decisions. We present an automated segmentation pipeline for whole-slide histopathology images that labels tumor grades 1-3 and normal mucosa. It utilizes dense prediction transformers with various encoder backbones, overlapping patches, and test-time augmentation. An adaptive augmentation policy, guided by large language models, further improves training. Top models were ensembled via soft voting, and mask refining post-processing steps, Gaussian blurring, morphological closing, and connected components analysis. On a colorectal cancer grade dataset, our method improved the F1 score from 62.92 to 69.84. Code is available here: this http URL Comments: SPIE Eighteenth International Conference on Machine Vision (ICMV 2025), Paris, France Subjects: Image and Video Processing (eess.IV); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2609.38419 [eess.IV] (or arXiv:2609.38419v1 [eess.IV] for this version) https://doi.org/10.48550/arXiv.2609.38419 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Journalreference: Proc. SPIE 14114, Eighteenth International Conference on Machine Vision (ICMV 2025), 141140I (25 Feb 2026) Related DOI: https://doi.org/10.1117/12.3096537 Focus to learn more DOI(s) linking to related resources
[CV-208] Raw Imagery Impacting Your AI: Should You Care?
链接: https://arxiv.org/abs/2609.38265
作者: Adrien Dorise,Marjorie Bellizzi,Stéphane May
类目: Image and Video Processing (eess.IV); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at OBPDC 2026
Abstract:Onboard AI is gaining interest for space applications such as vessel, wildfire, and cloud detection, where real-time processing can improve mission reactivity and reduce downlink needs. However, onboard models may operate on raw or minimally processed imagery rather than on restored ground products. This study evaluates how image degradation affects object detection by varying Signal-to-Noise Ratio (SNR), Modulation Transfer Function (MTF) at Nyquist, and Ground Sampling Distance (GSD). Controlled degradations are applied to Very High Resolution Maxar imagery, and three lightweight detectors, YOLOv5s, YOLOX-S, and NanoDet, are evaluated on the resulting operating points. The results show that the impact of image quality depends on the degradation mechanism, and that increasing degradation does not necessarily lead to a proportional decrease in vessel detection performance. GSD produces the most consistent performance shift, while MTF and SNR effects depend more on the model and resolution. Severe combinations of blur and noise produce the largest losses. These results provide task-level information that can support sensor, processing, and AI trade-offs for future onboard systems.
[CV-209] Representation Risk in Pretrained Image Encoders
链接: https://arxiv.org/abs/2609.35470
作者: Ardyn Nordstrom,Morgan Nordstrom,Vamuyan Sesay,Matthew D. Webb
类目: General Economics (econ.GN); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Applied researchers increasingly convert images into features with pretrained encoders, then use those features in a downstream prediction model. The encoder is often treated as an implementation detail. We show that it can instead be a consequential source of model uncertainty. We call this uncertainty representation risk: plausible pretrained encoders map the same images into different feature spaces and can yield sharply different out-of-sample conclusions from predictive performance. We compare ten modern and legacy frozen encoders across applications involving house prices, racehorse performance, breast-cancer histology, chest radiographs, continuous facial age, and rice disease. With common dimension control, heads, and group-safe splits, validation selects SigLIP 2 for houses, raising test R^2 from 0.396 for ResNet50 to 0.629, and DINOv2 for horses, raising R^2 from 0.029 to 0.105. No encoder is best in every task. Candidate procedures are constructed using training data and compared on a separate validation partition. The selected procedure reaches 0.658 for houses and 0.979 accuracy for pneumonia. Fixed-split gains are small for horses and rice, while repeated partitions reveal instability in horse feature union. Continuous age selects SigLIP 2 at 4.786 years MAE. The principal representation gaps persist with neural heads, similarly sized DINOv2 and ViT models, and limited adaptation. These results support a simple workflow: benchmark plausible representations, select on locked validation data, combine only when separate validation evidence justifies the additional cost, and report paired and split-level uncertainty. We implement this workflow in LOOKAGAIN-ML, the software package used to conduct the analyses in this paper.
人工智能
[AI-0] urbo Harness: Instance-Adaptive Harness Optimization
链接: https://arxiv.org/abs/2609.40330
作者: Tunyu Zhang,Hao Wang,Kai Xu,Dimitris N. Metaxas
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Automating the search for effective harnesses is an important step toward enabling agents to recursively self-improve. Existing harness optimizations typically produce a single global harness that is applied uniformly across task instances. However, a harness that works well on average may not be optimal for every instance. We introduce Turbo Harness, a framework that can adapt a globally optimized harness to each instance by reusing information generated during the original optimization process. Specifically, Turbo Harness recycles artifacts produced during a completed global harness optimization run, and summarizes them into a structured playbook. We train a harness editor to leverage this prior optimization experience to generate instance-specific patches to the global harness. At inference time, the editor uses the instance and the playbook to construct a tailored harness in which the execution model operates. Through numerical experiments, we show that Turbo Harness consistently outperforms existing harness optimization baselines across seven benchmarks spanning interactive agent tasks, software engineering, and long-horizon terminal tasks.
[AI-1] WorldAuditBench: Interactive 3D World Auditing with Multimodal Agents
链接: https://arxiv.org/abs/2609.40325
作者: Ziyan Jiang,Jingbo Yang,Jiabao Ji,Yujian Liu,Qiucheng Wu,Tommi Jaakkola,Yang Zhang,Shiyu Chang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As interactive 3D worlds are increasingly used to study intelligent behavior, it becomes important to develop efficient pipelines for identifying anomalies in these simulated environments, such as floating objects, traversable walls, or objects inconsistent with the surrounding scene. Multimodal AI systems, including vision-language models (VLMs) and vision-language-action models (VLAs), have shown potential for automating this task. However, 3D world auditing is complex, requiring the close coupling of two distinct capabilities: action, to navigate the 3D world and search for anomalies systematically and efficiently; and visual reasoning, to understand the environment and identify anomalies from multimodal observations. It remains largely unexplored whether multimodal agents can effectively couple these two capabilities, using visual reasoning to identify potential anomalies while taking actions to validate them. In this paper, we introduce WorldAuditBench, a benchmark for 3D world auditing comprising 213 anomaly tasks across 13 environments built with Unreal Engine 5 and this http URL, spanning five anomaly families. We evaluate five frontier models under a fixed exploration budget using two auditing paradigms: VLA-based exploration followed by VLM-based anomaly identification, and an end-to-end VLM agent in which visual reasoning directly guides action selection. Across the evaluated models and two paradigms, success rates range from 6.6% to 42.3%, substantially below human performance (83.4%). Through the task of world auditing, WorldAuditBench provides a testbed for studying how multimodal agents couple action and visual reasoning in interactive 3D environments, while highlighting current limitations in their ability to gather and interpret evidence during exploration.
[AI-2] Cogentic: Multi-Agent Orchestration for Automated Proof Discovery
链接: https://arxiv.org/abs/2609.40324
作者: Yang Cai,Vineet Gupta,Yanchen Jiang,Christopher Liaw,Aranyak Mehta,Grigoris Velegkas,Di Wang
类目: Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT)
备注:
Abstract:We present Cogentic, a multi-agent harness for automated proof discovery on open research problems. While frontier language models can generate strong mathematical ideas in a single shot, single-shot generation is often insufficient for open problems that require exploring multiple competing conjectures, overcoming subtle technical obstructions, and retaining intermediate progress over a long horizon. Cogentic addresses these challenges through an iterative prove–verify loop in which an orchestrator allocates a population of independent provers across distinct proof directions, subjects their output to adversarial verification by several specialized components, and promotes confirmed intermediate results into a persistent verified ledger that later rounds build on. The harness is designed to be able to solve research-level math and theoretical computer science problems. Using Gemini as the base model, Cogentic produced novel results on five open problems across online learning, auction theory, and mechanism design. Each result was independently verified by domain experts and is developed in full in companion papers. We list these results, and new ones as they are verified, at this https URL .
[AI-3] DynaHarness: A Dynamic Physical Harness for Self-Evolving Robot Agents
链接: https://arxiv.org/abs/2609.40306
作者: Haoyuan Deng,Jiebin Liu,Tengxiao Zhang,Langning Yan,Hongye Cao,Ziwei Wang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 37 pages, 19 figures. Project page: this https URL
Abstract:Pretrained robot policies provide useful action priors, but long-horizon manipulation still requires coordination between semantic reasoning and physical execution. Semantic reasoning operates at a coarser timescale than physical interaction, while episode-level failures provide limited guidance on which system component should be revised. We propose DynaHarness, a dynamic physical harness that couples semantic reasoning with physical governance through a shared execution contract and turns failure evidence into validated capability revisions. To be more specific, the slow brain proposes capabilities and symbolic arguments, while the fast brain grounds and monitors commands, refuses unresolved actions, substitutes capabilities, and requests replans when needed. The physical execution contract bounds each accepted command and records execution evidence across analytic skills, recovery skills, and the frozen VLA. Failure attribution localizes faults in these records and directs targeted revisions of reusable capabilities or execution mechanisms. Paired regression checks govern admission or rejection, closing the self-evolution loop. On LIBERO-Pro, DynaHarness achieves 75.2% on 800 newly sampled initial states, compared with 17.5% for the frozen policy. With the same capability library, full dynamic execution reaches 74.0% versus 63.9% under nominal one-step replanning. This demonstrates the value of DynaHarness as a dynamic physical harness that governs how existing capabilities are grounded, monitored, and coordinated during execution. Our project page is at this https URL.
[AI-4] How Much of a Harness Does a Strong Agent Need for Autonomous ML Engineering?
链接: https://arxiv.org/abs/2609.40303
作者: Kirill Brilliantov,Alejandro Hernández-Cano,Emmanuel Abbé
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Recent autonomous machine learning engineering (MLE) agents have made significant progress on public leaderboards. Often motivated by progress stagnation over long-horizon cycles and limited Large Language Model (LLM) primitives, modern MLE agents are deployed on top of increasingly elaborate machinery: multi-agent orchestrators, dedicated retrieval subagents, and more. While such harnesses expand, the use of more primitive but improved coding agents - where LLMs have direct access to the execution environment through read, write, and bash primitives - has received little attention in the field. In this paper we find that, under an equal time budget and the same frontier LLM backbone, open-source state-of-the-art harnesses provide no advantages over a single session of a minimal-harness coding agent baseline, pointing to the backbone as the primary driver for performance. Via a series of large-scale systematic ablation studies, we argue that the machinery layers become redundant in the coding agent setting. We conclude that the effort spent elaborating hand-crafted harnesses around strong models yields poor returns for current MLE benchmarks.
[AI-5] CAS II: Symmetric Partitions as Kolmogorov Models
链接: https://arxiv.org/abs/2609.40290
作者: Romie Banerjee
类目: Information Theory (cs.IT); Artificial Intelligence (cs.AI); Group Theory (math.GR)
备注:
Abstract:In algorithmic statistics a string x is explained by a finite set containing it, and Kolmogorov’s structure function records the smallest such model at each level of complexity. Vereshchagin’s strong models, those computable from the data by a total algorithm, are essentially the cells of simple partitions. We read a partition of binary strings as a hypothesis, with the cell containing x as its model, and develop algorithmic statistics over symmetric partitions: the orbit partitions of groups acting on strings. The Galois connection between subgroups and partitions gives each ambient group a lattice of symmetric partitions, with canonical certificates, canonical costs, and an algebra of hypotheses. The resulting structure function and symmetric sophistication measure which part of the regularity of x is symmetric. For the full symmetric group every partition is symmetric: cells recover all Kolmogorov models, cells of cheap partitions recover exactly the strong models, and normal and strange strings are characterized by symmetry. For GL(n,2) the cells are exactly the linearly homogeneous sets, so linear symmetry is a restricted model class. For nonzero x, the linear-symmetry structure function lies in a band between the sufficiency line and the trivial bound, and both edges are attained: there are stochastic normal strings whose simple structure is invisible to linear symmetry. We also give coordinates on the space of permutation groups: each group is an element of a Burnside ring (its type) together with a permutation (its placement), and restriction moves refine partitions via the Mackey formula. In these coordinates the collapse for the symmetric group is a statement about placement, a linear hypothesis is determined by its type up to n^2 bits, and the maximal gap theorem shows that any space of symmetry hypotheses small enough to search is small enough to miss simple structure.
[AI-6] PivotOPD: Learning to Recover from Pivotal Mistakes in Multi-Turn Agents
链接: https://arxiv.org/abs/2609.40285
作者: Yinghui He,Yapei Chang,Khushi Bhardwaj,Daniele Molinari,Tugrul Konuk,Jan Kautz,Ali Hatamizadeh
类目: Artificial Intelligence (cs.AI)
备注: PivotOPD technical report; Project page: this https URL
Abstract:On-policy distillation (OPD) is a promising approach for training language agents, providing dense teacher supervision on student-generated trajectories. However, in multi-turn interaction, an incorrect action changes the states the student encounters later, so errors compound across turns. In preliminary experiments across three Qwen3 models (8B to 235B), we find that more than half of the failed rollouts contain a pivotal mistake, an action that moves the agent farther from completing the task, and this mistake typically occurs early. These pivotal mistakes often remain recoverable: guiding the model for only a few turns after the pivotal turn can restore task success. We therefore propose PivotOPD, an on-policy distillation framework that jointly trains the student to prevent pivotal mistakes and to recover from the states they create. At each pivotal mistake, a teacher model provides a gold action and then names a recovery action at each of the next few turns. Preventive distillation uses the gold action with reverse KL to steer the student away from the pivotal mistake, while recovery distillation uses the recovery actions with forward KL to transfer recovery behaviors that the student rarely samples. Against 13 baselines on ALFWorld, WebShop, and Search-based QA, PivotOPD achieves the strongest average performance for both Qwen3-1.7B and Qwen3-8B students, improving over the strongest baseline on ALFWorld by +5.5% with the 1.7B student. The gains also transfer to another model family on the software engineering domain, where PivotOPD raises the resolve rate of a Nemotron-3.5 student on SWE-Bench Verified by +3.2%. Project page: this https URL
[AI-7] Learning from Research: Toward Lifelong Agent Harness Evolution
链接: https://arxiv.org/abs/2609.40169
作者: Jingbo Yang,Kwei-Herng Lai,Xiaowen Wang,Yaar Harari,Evgeniy Gabrilovich,Shiyu Chang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Language agents are expected to solve increasingly complex tasks, creating a growing need for continual improvement. One promising approach is to evolve the agent harness, the software that governs tool use, memory management, and task execution, while keeping the underlying language model fixed. Recent methods automate this process by using a meta coding agent to modify the harness based on execution feedback. However, relying on that agent’s existing knowledge and observed failures can restrict exploration and make adaptation reactive. Inspired by how human experts learn from the research literature for new solutions, we introduce ScholarEvolve, a framework that automatically draws on state-of-the-art research to guide harness evolution. ScholarEvolve organizes the harness evolution directions into functional modules and uses topic modeling to identify distinct improvement strategies for each module. It implements these strategies and evaluates their combinations to improve task performance. Moreover, the framework is designed to incorporate new publications over time, allowing research advances to drive proactive lifelong evolution. Experiments demonstrate improvements on AppWorld and Tau2-Bench. ScholarEvolve raises Qwen3.5-27B task goal completion from 49.6% to 63.6% on AppWorld Challenge, and raises GPT-5.4-mini pass@1 from 72.7% to 81.9% on Tau2-Bench Telecom.
[AI-8] PrefPI: Preference-Guided Steering into Out-of-Distribution Behaviors
链接: https://arxiv.org/abs/2609.40165
作者: Seungeun Rho,Wontaek Kim,Danfei Xu,Sehoon Ha
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:We present PrefPI (Preference-Guided Policy Iteration), an iterative framework for steering pretrained generative robot policies using only relative preferences over self-generated trajectories. Unlike prior preference-learning methods that primarily sharpen modes already represented by the policy, we study steering beyond the initial effective support, where desired behaviors are rarely or never observed under the initial policy. Our key idea is to formulate preference learning as preference-conditioned generative modeling: preferred trajectories define a conditional distribution, whose density ratio with the broader behavior prior provides an implicit preference signal amplified by classifier-free guidance (CFG). Repeating this preference-conditioned modeling and guidance step yields a form of preference-guided policy iteration, turning incremental improvements toward previously inaccessible behaviors. Across diffusion policies and the PI0.5 flow- matching VLA in simulation and the real world, PrefPI produces substantial behavioral shifts with limited feedback. In particular, PrefPI increases object transport height from 10.7 cm to 19.8 cm on real hardware with only 150 preference-labeled trajectories.
[AI-9] Game-Guided Skill Discovery through Self-Play for Playable Agent Control
链接: https://arxiv.org/abs/2609.40137
作者: Seungeun Rho,Jeonghwan Kim,Xue Bin Peng,Sehoon Ha
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:We present Game-Guided Skill Discovery (GGSD), a framework that uses self-play in games to discover motor skills that are directly playable by humans. Playable skills provide a compact abstraction for controlling embodied agents through a small set of learned behaviors rather than low-level actions. To be effective, these skills should be semantically distinct, interpretable, and expressive; properties that existing unsupervised skill-discovery methods often fail to achieve simultaneously. GGSD achieves these desiderata by grounding skill discovery in competitive gameplay. A hierarchical agent competes against its past selves, with a high-level policy selecting from a small discrete skill set and a skill-conditioned low-level policy learning the corresponding behaviors. After training, a human can replace the high-level policy and directly control the agent through the same discrete skills. Despite the small number of high-level actions, skill transitions give rise to emergent combo behaviors, expanding expressivity beyond individual primitives. Across Ant, Franka-arm, and Unitree G1 environments, we show that GGSD produces human-playable skills that humans can compose to solve unseen tasks, such as Maze and CubePush, without additional training. An interactive demo is available at this https URL.
[AI-10] actile Curiosity Drives Robot Interaction
链接: https://arxiv.org/abs/2609.40134
作者: Klemens Iten,Alexander Proshkin,Bhavya Sukhija,Stelian Coros,Andreas Krause,Pieter Abbeel,Carmelo Sferrazza
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 16 pages, 6 figures, 1 table. Preprint, under review
Abstract:Mastering robot manipulation skills via reinforcement learning (RL) remains largely sample-inefficient. The most common RL algorithms rely on random action sampling to discover new strategies, resulting in agents that allocate most of their training budget to motions in free space, away from the contacts from which manipulation skills emerge. Existing intrinsic motivation methods based on model disagreement or epistemic uncertainty improve on isotropic noise, but they can also reward uncertainty in functionally irrelevant transitions, such as erratic motions in free space. In this work, we argue that tactile feedback provides a natural signal for exploration, and introduce TacEx, a framework that incorporates touch into epistemic uncertainty-driven exploration by decomposing model uncertainty across sensory modalities and directing curiosity toward the tactile channel. By anchoring curiosity to the sense of touch, TacEx drives the robot to discover complex contact dynamics, learning to manipulate and grasp objects without task rewards or expert demonstrations during exploration. The interaction-dense dataset collected through this tactile-driven curiosity supports offline learning of downstream pick-and-place policies without additional environment interaction. We further use tactile-driven exploration to post-train vision-language-action (VLA) models. Although the VLAs are initially pre-trained without tactile feedback, post-training with TacEx substantially improves downstream performance while remaining highly sample-efficient.
[AI-11] Unlearnable or Unmeasured? On the Reliability of Difficulty Labels in RLVR NEURIPS2026
链接: https://arxiv.org/abs/2609.40115
作者: Chandak Chakma,Syed Nazmus Sakib,Nafiul Haque,Shifat E. Arman
类目: Artificial Intelligence (cs.AI)
备注: Accepted at the NeurIPS 2026 Workshop on Transitioning from Pre-Training to Post-Training. Project page: this https URL
Abstract:Reinforcement learning with verifiable rewards (RLVR) has become an important approach for improving reasoning during post-training. Recent work suggests that some difficult prompts remain resistant to learning even when they occasionally produce correct solutions. We revisit this unlearnability phenomenon and find that the affected prompts do improve, at roughly one third of the learnable rate, while the difficulty-defined set used to study them is much less reproducible than expected. These difficulty labels are estimated from a limited number of sampled responses. Combining them across seeds can further change which prompts are selected instead of simply reducing measurement noise. We develop a sampling-based framework for quantifying this instability and determining how much evaluation is required for difficulty assignments to reproduce reliably. We also revisit the gradient-similarity evidence proposed to explain unlearnability and show that part of the observed separation arises because difficult prompts provide fewer correct rollouts from which their gradients can be estimated. Matching this sample count weakens the gradient difference but does not remove it. Overall, the slow-learning phenomenon survives our reanalysis, while both the prompts used to define it and the evidence used to explain it require more careful measurement.
[AI-12] PTNO: Training Neural Operators with Noisy Monte Carlo Estimates for Particle Transport Problems
链接: https://arxiv.org/abs/2609.40090
作者: Yubo Cao,Xi Deng,Mengqi Xia,Vignesh Gopakumar,Ander Gray,Anima Anandkumar
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 41 pages, 15 figures, 35 tables
Abstract:Particle transport under multiple scattering is central to radiative transfer and plasma physics, yet high-fidelity Monte Carlo (MC) simulations must trace prohibitively many particles. Learning-based surrogates can amortize this cost, but typically train on expensive, well-converged MC solutions. We propose the Particle Transport Neural Operator (PTNO), a neural operator that learns particle transport surrogates directly from noisy, low-cost MC labels. Such labels pose two challenges: (1) high variance, which destabilizes standard supervised learning, and (2) a high dynamic range (HDR) spanning many orders of magnitude. For the first, we learn the solution operator from noisy labels of many configurations, amortizing MC cost and generalizing to unseen configurations. Because MC labels are unbiased, we show that the squared loss on them shares its minimizer with the loss on converged solutions, and our budget-allocation study over training scenes M , MC samples per render N , and independent renders per scene K shows that many noisy scenes beat fewer converged ones. For the second, a nonlinear transform such as the logarithm biases noisy supervision. Instead, PTNO keeps labels in physical space and enforces positivity with a softplus output layer that represents small values effectively. We further train with a pointwise relative L_2 loss (PRelL2), the stop-gradient relative loss of HDR denoising and neural rendering, which normalizes each residual by the stop-gradient prediction instead of the noisy label. We demonstrate PTNO on neutron transport in fusion reactors and radiative transfer in participating media. On the two neutronics tasks, PTNO is 10^4 - 10^5\times faster than converged MC on the same CPU and 10^3 - 10^5\times cheaper than MC at matched accuracy; on the two radiative-transfer tasks, MC at matched accuracy costs 0.8 - 11\times as much as PTNO.
[AI-13] MeanVoiceFlow2: Joint Optimization of Mean Flow and Content Encoder for Fast One-Step Zero-Shot Voice Conversion WWW INTERSPEECH2026
链接: https://arxiv.org/abs/2609.40087
作者: Takuhiro Kaneko,Hirokazu Kameoka,Kou Tanaka,Yuto Kondo
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
备注: Accepted to Interspeech 2026. Project page: this https URL
Abstract:Flow-matching approaches to voice conversion (VC) have gained attention owing to their high speech quality and strong speaker similarity. Among them, one-step models such as MeanVoiceFlow are particularly attractive because they enable efficient inference; however, their reliance on a computationally intensive content encoder remains a bottleneck. We therefore propose MeanVoiceFlow2, a framework that jointly optimizes a flow-based conversion module and a computationally efficient content encoder. The model is trained through conversion distillation using MeanVoiceFlow and the reconstruction of real data. We further incorporate diffusion-GAN training with sample mixing and teacher-guided conditioning augmentation to enhance realism and disentanglement. Experiments on zero-shot VC showed that MeanVoiceFlow2 achieved higher perceptual quality and approximately 9\times faster inference than MeanVoiceFlow while maintaining comparable speaker similarity. Audio samples are available at this https URL.
[AI-14] BatSLAM 2.0: Sequence-Verified Sonar Place Recognition in a Robust Pose Graph
链接: https://arxiv.org/abs/2609.40085
作者: Jan Steckel
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Systems and Control (eess.SY)
备注:
Abstract:Echolocating bats can navigate dark and cluttered spaces using echolocation. Over a decade ago, BatSLAM showed that a robot with a biomimetic binaural sonar can build a topological map of the environment, by recognizing places from the received acoustic signals. Sonar place recognition, however, is ambiguous by nature: corridors produce nearly identical echo trains, and wrong loop closure can collapse the topological map. In this paper, we introduce BatSLAM 2.0, a novel sonar-only SLAM system built from three elements: an updated acoustic front-end, a sequence verifier that tracks and verifies loop closure candidates and a pose graph implemented on a high performance factor graph framework. The system was thoroughly evaluated both in simulated as well as real world recordings. In both cases, the BatSLAM2.0 algorithm shows the capability of robust topological map creation, countering map collapse, and robust scaling of map size.
[AI-15] Inference Auctions
链接: https://arxiv.org/abs/2609.40070
作者: Keegan Harris,Siddharth Prasad,Asher Trockman,Nika Haghtalab,Michael I. Jordan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT)
备注:
Abstract:When inference demand exceeds available compute capacity, model providers must decide which requests should be served first. Users have different tolerances for delay from an LLM API, but current priority pricing schemes compress these differences into coarse fixed-price service tiers. We design an inference auction that allows users to bid for faster service. Our auction allocates priority in an economically efficient way without sacrificing latency, and we develop fast algorithms for implementing prices that incentivize truthful bidding. We also design an autobidding agent for our inference auction, where users specify an inference budget and the autobidder dynamically adjusts its bids over time to maximize user utility subject to the budget constraint. Experiments validate the practicality of our auction: it increases system welfare while maintaining the cache utilization and latency advantages of SGLang, a state-of-the-art inference serving framework.
[AI-16] Community-Driven API and AI Writer Design for Openly Scaling Community Notes
链接: https://arxiv.org/abs/2609.40067
作者: Brad Miller,Jay Baxter,Jiansong Chao,Keith Coleman,Sophie Hilgard,Daniel Ortiz
类目: ocial and Information Networks (cs.SI); Artificial Intelligence (cs.AI)
备注:
Abstract:Community Notes is a crowd-sourced approach for adding context to posts on X. Contributors propose and rate notes, forming the inputs to an open-source, open-data algorithm that determines which notes show broadly to users. Since September 2025, Community Notes’ AI Note Writer API has provided an open, public interface for using AI to propose notes, while adhering to the founding principle that users, not the platform or an AI, control which notes show on X. Explicit note requests and user posts on X determine the AI API post feeds, ensuring that AI note writing responds to demand from X users. We present the design, operation and impact of the AI API, including analysis of the interaction between AI and human generated notes across topics. Unless otherwise stated, measurements and system description reflect June 2-29, 2026. The Community Writer is the largest AI API client and contributes the bulk of AI API output, generating 52% of notes selected as Helpful and shown broadly on X. The writer is guided by community input during both training and operation to prioritize, draft, evaluate and delete proposed notes. Beyond scale, the writer also offers speed, submitting the first proposed, non-deleted note on 60% of posts when compared to other writers. AI note writing is additive on top of human note writers, extending coverage of Community Notes on X. Among posts that have Helpful notes, 42% have only AI notes, indicating human raters did not feel motivated to propose an alternative. In contrast, 30% have only human notes, reflecting contribution beyond the scope of AI writing. The Community Writer is open-source software released under the Apache 2.0 license. Subjects: Social and Information Networks (cs.SI); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.40067 [cs.SI] (or arXiv:2609.40067v1 [cs.SI] for this version) https://doi.org/10.48550/arXiv.2609.40067 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-17] Efficient Active Auditing of Multi-Group Fairness with Bias Probes
链接: https://arxiv.org/abs/2609.40034
作者: Ayoub Ajarra,Debabrota Basu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Applications (stat.AP); Machine Learning (stat.ML)
备注:
Abstract:Over the past decade, Machine Learning (ML) has been trained under dual objectives: minimizing prediction error via Empirical Risk Minimization (ERM) while controlling unfairness bias. In practice, however, fairness-aware training often yields limited improvements over standard ERM, making reliable post hoc auditing essential. Existing auditing approaches for black-box models either rely on model reconstruction --exposing systems to extraction attacks-- or directly estimate fairness metrics, offering limited insight into which regions of the data distribution drive bias. More fundamentally, property-specific auditing --aimed at extracting only targeted fairness information without reconstructing the model-- remains poorly understood. In this work, we introduce the bias probe framework, which enables targeted and adaptive querying to reveal bias structure while preserving model confidentiality. Building on this framework, we propose ALeBi, an active auditor that learns such probes to efficiently estimate multi-group fairness metrics. We establish novel sample complexity guarantees governed by a property-specific complexity measure, resolving a previously posed open question, and extend our analysis to adversarial settings where the model owner may strategically obscure bias. Our results uncover a fundamental trade-off between model confidentiality and reliable auditing, and show that property-specific probing enables both accurate estimation and interpretable identification of high and low-bias regions. Extensive experiments support our theoretical findings and demonstrate the practical effectiveness of our approach.
[AI-18] Fenchel Tilting: Weighted Correction for Efficient Finetuning of Generative Models
链接: https://arxiv.org/abs/2609.40030
作者: Maksim Bobrin,Maksim Zhdanov,Dmitry Dylov
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Adapting a pretrained generative model to an arbitrary preference expressed as a utility function underlies reward alignment, guided design, and constraint satisfaction, enabling diverse applications. Existing fine-tuning methods trade off generality against computational cost: they either restrict the family class of supported preferences to keep optimization simple or preserve generality at the expense of efficiency. We introduce Fenchel Tilt Flow Control (FTFC), which decouples utility optimization from generative-model fitting. FTFC first optimizes for a target distribution by jointly fitting an effective reward and density-ratio weights on pretrained samples. Method combines the utility’s variational structure with Fenchel duality, supporting general f -divergence penalties that determine how rewards are transformed into an distribution-correction weights. These weights are then frozen and used to modify a diffusion or flow model in a single stage of importance-weighted denoising or flow matching, without differentiating through sampling trajectories. We establish exact duality for concave utilities under suitable conditions and show that weighted fitting reproduces the optimal target distribution for a given utility. Across image and molecule generation benchmarks, FTFC improves over baselines on diverse preference functions, while also being up to 20\times more efficient. roposed method enables adaptation beyond expected-reward maximization without complex optimization, while preserving robustness for more general class of the utility functions compared to baselines.
[AI-19] Who Verifies the Graph? Misspecification Attacks on Causal Action Verification for Language Agents NEURIPS2026
链接: https://arxiv.org/abs/2609.40027
作者: Fabio Rovai
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted as a poster at the NeurIPS 2026 Workshop “Who Verifies the Agents?”
Abstract:Causal action verifiers gate an agent’s state-changing tool calls by checking whether each proposed intervention is identifiable against a committed action-state graph, and they issue a certificate that carries the identification argument and a one-sided lower confidence bound. One such verifier, CIVeX, reports zero false executions on a confounded tool-use benchmark. We red-team it by corrupting only the committed graph. Omitting a single bidirected edge takes it from zero false executions to 15.3% at the benchmark’s published confounding strength, with 91% of its executions harmful and utility falling from +2.27 to +0.35. Reversing one arrowhead, so that a mediator is committed as a confounder, gives 48.9% false executions and no correct ones. Every one of these actions carries an internally valid certificate. An attestation step that tests each observationally certified execution against a bounded randomised sample detected both attacks, with 2 false alarms in 555 executions on a truthful graph; refusing what fails the test, or cannot be tested, gave zero false executions in every setting we measured. It does not restore beneficial execution: at the published strength 97.1% of beneficial actions are still never executed, because the same misspecification rejects them before attestation runs. Those rejections carry certificates too, and auditing them works, but its cost scales with the number of rejections rather than the number of executions. Recovering safety costs 127 experiments per 1,050 actions; recovering the lost value costs 614 more, at which point the audited verifier makes the honest graph’s decisions on every instance and spends exactly its experiment budget. An audit that inspects only executions protects against wrongful action. Wrongful inaction has to be paid for separately.
[AI-20] What Can Component-Replacement Evidence Establish? A Critical Scoping Review of Local Decisions in LLM Agents
链接: https://arxiv.org/abs/2609.39989
作者: Shuyang Zhang(The Hong Kong Polytechnic University),Jianshuo Chang(The Hong Kong Polytechnic University)
类目: Artificial Intelligence (cs.AI)
备注: 36 pages, 3 figures. The authors contributed equally
Abstract:Background. A component replacement in a language-model agent changes an execution trajectory, potentially altering later observations, resource use, and recovery opportunities. Different evidence is needed to assess its task-level benefit and the contribution of local decision quality. Methods. This critical scoping review maps 348 studies and examines 90 comparison records: 88 from 40 included studies and two from supplementary studies. Eight purposively selected cases structure the synthesis around the replaced decision, executed conditions, measurement comparability, controls, and remaining explanations. Results. Of 222 studies reporting local decision metrics, 142 also report measured task endpoints and 49 report proxies. These counts identify studies that report both types of measurement, without establishing that the measurements come from matched comparisons. Outcome Monitors reports a package-level completion gain whose attribution to detector quality remains limited; First-chunk selection reports a local improvement assessed against an offline proxy endpoint; Evidence-Carrying Termination reports fewer premature unsupported terminations and completion non-inferiority, without establishing completion superiority. Cross-case analysis identifies three candidate mechanisms involving recovery and disruption, intervention timing, and downstream use. Attribution and deployment depend on the comparison controls, label definitions, and information available to the controller. Conclusions. The review distinguishes the task-level benefit of a component replacement from the contribution of local decision quality and derives eight claim-specific reporting items. Neither online execution nor simultaneous gains in local and task metrics alone establish that better local decisions explain the task-level gain.
[AI-21] ACTIC: Temporal and Context-Aware LLM Tactical Planning for Roadside LiDAR Attacks
链接: https://arxiv.org/abs/2609.39969
作者: Yiming Gao,Shaocheng Luo
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注: Under review
Abstract:Physical LiDAR attacks are often evaluated using fixed primitives and manually selected parameters, despite their strong dependence on surrounding traffic. We present TACTIC, a scene-aware framework that uses a multimodal large language model (MLLM) to coordinate state-adaptive roadside LiDAR attacks. Under a gray-box threat model, TACTIC relies only on an attacker-operated roadside perception stack, without accessing the victim LiDAR’s native point clouds or internal processing. Local perception provides metric vehicle states, while the MLLM combines these measurements with roadside imagery to infer relational traffic context and construct a semantic scene graph. Based on this representation, TACTIC selects and configures two complementary primitives: \emphpush-away, which shifts the perceived range of a lead vehicle, and \emphphantom-obstacle braking, which triggers emergency braking through obstacle injection. Measured traffic states and empirically calibrated constraints ground the generated tactics in physically feasible operating regions. To accommodate MLLM latency, TACTIC overlaps reasoning and execution asynchronously while high-rate local perception detects scene changes and triggers replanning. Across 280 randomized CARLA trials, the full policy achieves a 100% collision rate, compared with 35% for a fixed rule, 60% for random selection, and 75% for a restricted LLM using mode selection with default parameters. Joint physical-and-image input achieves 100% success, versus 65% with physical measurements alone and 75% with imagery alone, while asynchronous \Delta refresh reduces scene-mutation response from 7.4 s to 2.0 s. These results show that scene-dependent tactical planning can expose context-sensitive LiDAR failure modes that fixed attack policies may miss.
[AI-22] What Limits Recursive Reasoning Models: Optimization Architecture and Test-Time Scaling
链接: https://arxiv.org/abs/2609.39967
作者: Yuliana Shakhvalieva,Dmitrii Kharchev,Viacheslav Bezrukov,Inessa Fedorova,Dmitry Bocharov,Ivan Oseledets,Valerii Ternovskii
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Recursive reasoning models apply a small shared Transformer block many times to refine a latent state. This gives them large effective depth with few parameters and makes them strong on algorithmic tasks. Such compact solvers are natural candidates for tools that an LLM can call on narrow algorithmic subproblems. However, existing models such as HRM, TRM and URM differ in architecture, gradient propagation and training procedure simultaneously. This makes it hard to tell what drives their performance, and their optimization is still poorly understood and often unstable. In this work we address both of these gaps. First, we study these questions under a unified experimental pipeline spanning six algorithmic domains. Individual controlled ablations are performed on representative domains, while the resulting recipe is evaluated across the full suite. The study reveals a surprisingly simple recipe for stable and generalizable recursive reasoning: an intermediate gradient horizon, large physical batches and controlled updates of the recurrent state. An explicit hierarchical architecture is not needed. Second, we combine these findings into a stable 13.6M-parameter model that achieves the strongest overall performance among the evaluated recursive baselines, with particularly large gains on out-of-distribution generalization. It raises Arithmetic OOD accuracy to 71.2%, from 36.2% for the strongest baseline, while reaching 98.41% on Sudoku and 59.5% pass@2 on ARC-AGI-1. Our results show that, within the recursive architectures studied here, performance depends strongly on how recurrence is optimized and stabilized. More broadly, it shows how AI systems can be improved by optimizing their components one at a time.
[AI-23] Better Deck or Different Judge? Evaluating Agent ic Harness Gains in Corporate and Investment Banking
链接: https://arxiv.org/abs/2609.39958
作者: Ludovic Gibert,Matis Despujols,Andre-Louis Rochet
类目: Artificial Intelligence (cs.AI)
备注: 13 pages, 8 figures, 9 tables
Abstract:Corporate and investment banking teams use presentations to support credit decisions and advise clients on financing and transactions. Producing these decks requires reconciling financial data, tracing sources and turning analysis into a recommendation. We retrospectively study the development of an agentic harness combining a 27B language model, financial calculations, narrative templates and validation checks. LLM judges guide engineering changes and assess the resulting decks, raising the question of whether higher scores reflect better documents or changes in grading. In shared-session text-only grading with template markers removed, five judges score the complete system 20.4 to 33.6 points out of 95 above the same model generating directly from a short prompt. Every judge scores the system higher on all seventeen development deliverables. Margins against direct Opus generation from a short prompt range from -4.7 to +0.8 points. Judges agree on broad progress across development rounds but agree less on final-deck rankings than on pooled scores. Repeated grading also shifts scores on unchanged decks, making small improvements difficult to distinguish from judge variability.
[AI-24] Learning When and How to Intervene: A Hindsight-Distilled Sentinel for Coding Agents
链接: https://arxiv.org/abs/2609.39957
作者: Jiangrui Zhao,Chenglong Li,Meng Zhang,Xiaoting Du
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Coding agents solve repository-level tasks through sequences of actions, where a single erroneous action can misdirect subsequent decisions and increase recovery costs. Existing approaches use execution feedback for recovery or specialized checks to block errors, but deciding before execution whether intervention will benefit eventual task completion remains challenging. To address this challenge, we propose HiSentinel, a hindsight-distillation framework that trains lightweight 0.6B and 1.7B sentinels to select pre-execution interventions aimed at improving task completion rather than correcting every imperfect action. A privileged teacher uses recorded execution outcomes as evidence for intervention judgments, which are distilled into a causal student that receives only the pre-action context and proposed action. Beyond identifying whether and when to intervene, the sentinel must also provide actionable feedback that helps the coding agent recover or obtain necessary human input. To support these capabilities, we introduce SWE-Intervene, an action-level dataset constructed from software-engineering trajectories that annotates whether an action should be allowed, autonomously redirected, or paused for human assistance, together with corresponding intervention feedback. Across SWE-bench Verified Mini and Ask or Assume, HiSentinel consistently improves task completion across Sentinel scales and coding-agent families, with gains of up to 14% and 10%, respectively, while maintaining competitive token consumption. These results demonstrate that lightweight pre-execution intervention can effectively prevent error propagation and improve the reliability of autonomous coding agents.
[AI-25] Coverag e Before Control: Route-Instruction Grounding and Steering for Controllable Retrosynthesis
链接: https://arxiv.org/abs/2609.39955
作者: Xuemin Chen,Xiaozhuang Song,Xinjian Zhao,Yaoyao Xu,Tianshu Yu
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Single-step retrosynthesis models are commonly evaluated by their ability to recover recorded reactions. In practice, chemists may need to choose among several precursor sets for the same product, for example to preserve a particular motif. Recovering a recorded answer alone does not establish this ability to follow a preference. Satisfying such requests requires both coverage of relevant alternatives and control over which alternatives are favored. We introduce Route-Instruction Grounding and Steering (RIGS), a two-stage framework for instruction-conditioned retrosynthesis. Stage A trains a language projector, teaching it which alternatives an instruction favors or discourages. Stage B uses the projector learned in Stage A to steer a frozen generative model through lightweight residual adapters. We construct nested one-to-many training supports by pairing each product with increasing numbers of candidate precursor sets. Extensive experiments demonstrate that broader support helps the model generate a wider range of alternatives, and RIGS can learn to guide generation according to instructions. The relationship between coverage and control is consistent across model scales but non-monotone.
[AI-26] ConflictGuide: AutoResearch Improves When Competing Behaviors Are Made Visible
链接: https://arxiv.org/abs/2609.39933
作者: Binqian Xu,Qiran Zou,Xiangbo Shu,Dianbo Liu
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:When designing machine learning models, desirable properties are often in tension: improving one behavior can impair another, so task progress can depend on alleviating the conflict. LLM-based AutoResearch systems, which iteratively edit model code and retain edits based on scalar task-performance feedback, have largely ignored this trade-off. We find that scalar feedback supports broad exploration early in search, but it does not reveal how edits affect competing behaviors. In matched-budget experiments, introducing competing-behavior feedback as task gains diminish increases the share of proposals that improve both behaviors and sustains progress beyond scalar-only plateaus. Obtaining this feedback for a given model requires identifying its competing behaviors and designing probes to measure them. To make competing-behavior feedback actionable, we introduce ConflictGuide. Its reusable ConflictGuide-Skill combines a literature-grounded taxonomy with model-specific evidence to identify competing behaviors and specify probes for a code agent to implement as metrics. Evolution proceeds in two stages: Stage I explores with task feedback; Stage II uses probe feedback to steer proposals toward conflict alleviation and retains marginal-gain edits only when probes indicate sufficient alleviation. Across five diverse model families, ConflictGuide reduces task and conflict-related errors by up to 28% and 14%, respectively, relative to scalar-only AutoResearch, with gains extending to other code agents.
[AI-27] RACE: Trajectory Selection for Parallel Scaling of Search Agents
链接: https://arxiv.org/abs/2609.39912
作者: Qisheng Zhou,Zhen Xiong,Qiaoyu Tan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 19 pages, 2 figures. Code: this https URL
Abstract:Parallel search may generate a correct answer that final-answer voting fails to select. We formulate this consolidation stage as trajectory selection and introduce TRACE (Trajectory Ranking with Aggregated Cross-Rollout Evidence), a lightweight learned selector that ranks completed trajectories using the search evidence behind their answers. TRACE preserves individual query and evidence occurrences, connects rollouts through shared content or document identity, and propagates information across these relations. Each candidate answer then reads the updated states of its own trajectory, preserving retrieval provenance while incorporating evidence from related rollouts. Trained with answer-level supervision over frozen text embeddings, TRACE returns an existing answer without additional search or autoregressive aggregation. One selector per search setting transfers across rollout policies and agent backbones without agent-specific fine-tuning, improving over voting across six WebQA policies and six long-horizon dataset-backbone combinations at K=16 . On Qwen2.5-14B Base/SFT WebQA pools, TRACE achieves 45.2/49.2% EM, compared with 43.9/48.0% for the strongest Qwen3-32B generative aggregators. On long-horizon FRAMES, GAIA, and BrowseComp, it reaches 78.6% average accuracy, exceeding majority voting by 3.1 percentage points. On Base WebQA pools, TRACE with only 8 rollouts comes within 0.4 points of majority voting over 64. TRACE also achieves at least 10\times higher processing throughput than SolAgg, SummAgg, and AggAgent across all seven WebQA benchmarks. These results show that reusing cross-rollout search evidence provides an effective and efficient alternative to heavyweight generative aggregation for parallel search. Code is available at this https URL.
[AI-28] DoGBench: Can Agents Meet Expert Standards for User-Facing Documentation?
链接: https://arxiv.org/abs/2609.39909
作者: Frances Liu,Manny Silva,Paige Calvert,Ayu Adiati,Sarah Sanders
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce DoGBENCH (Documentation Generation Benchmark), to our knowledge, the first benchmark for generating and maintaining real user-facing software documentation. It asks whether an agent can produce documentation that experienced technical writers would accept in review. The benchmark contains 292 items from open source projects, including Helm, PostHog, and Mautic. Each item gives the agent a pre-change repository and a trigger, such as a code pull request or a reported documentation gap. The agent must first decide whether the documentation needs an update. For items that need one, the agent must produce an acceptable patch in one attempt. For items that do not need updates, the agent must abstain. Task-specific rubrics, validated with project maintainers, score each patch on accuracy, completeness, reader guidance, placement, and repository conventions. The composite score combines patch quality with correct abstention, and a score of 100 means an agent meets every requirement for the task. Scores should not be interpreted as a percentage of an expert’s capability. We evaluated seven agents. The highest-scoring agent reached 47.3 out of 100 on the 117-item held-out split. In a separate audit of 1,267 patches, the most common failure modes were task-completion gaps (45.5%), technical inaccuracies (36.6%), and incomplete conceptual or reference coverage (32.5%). Analysis of the corresponding trajectories identified three key patterns associated with these failures: (1) describing interfaces without examining how readers use them (36.0%), (2) missing decisive evidence and filling the gaps with plausible assumptions (33.1%), and (3) stopping after finding the first plausible documentation surface and leaving other affected pages stale (30.1%).
[AI-29] OSWorld-Science: A Benchmark of Computer Use Agents for Learning and Using Scientific Software
链接: https://arxiv.org/abs/2609.39903
作者: Dingyuan Dai,Heli Qi,Lei Liu,Yinxi Li,Baiding Chen,Zijun Dou,Qingcheng Zeng,Qi Kang,Oliver Sun,Eric Wang,Bo Zhou,Haixin Wang,Yufan Du,Shi Bo,Ruihan Lin,Mengqi Yuan,Dunjie Lu,Steven Dillmann,Yiming Shi,Tina Su,Amy Xin,Minghao Liu,Xi Wang,Xu Huang,Ge Zhang,Pengyu Nie,Zhen Yang,Jie Tang,Juanzi Li,Weihao Xuan,Tianyu Liu
类目: Artificial Intelligence (cs.AI)
备注: 62 pages. Website: this https URL Public contributions welcome: this https URL
Abstract:Scientific software presents a demanding test for computer-using agents based on visual language models (VLMs): completing a research workflow requires interpreting specialized interfaces, manipulating scientific objects, and producing verifiable results. We thus introduce OSWorld-Science, a benchmark and evaluation environment that combines scientifically meaningful tasks, artifact-based evaluation, and an efficient agent harness for studying computer use in the scientific domain. The benchmark contains 12 VLMs and 146 high-quality tasks across several scientific domains and software configurations, covering workflows such as molecular drawing and retrosynthesis, pathology image analysis, statistical computing, and physical simulation. Tasks are developed through expert proposals and iterative human–AI co-design, with selection guided by scientific value and difficulty. Task-specific execution-based evaluators inspect application states and generated artifacts, including molecular structures, segmentation masks, plots, and numerical results, and award partial credit for incomplete outcomes. Our special harness integrates model adapters, interaction-loop control, and trajectory logging to support comparisons of models and interaction strategies. Our results show that current state-of-the-art VLMs with a strong harness still face challenges in addressing key questions in the scientific domains. We also analyze the benchmarking results across multi-linguistics, reasoning efforts, context length and other factors and derive several important conclusions and directions to assist future development. Overall, we provide an integrated framework connecting expert-defined scientific goals to verifiable software outcomes, enabling systematic evaluation of both agent capabilities and harness design in scientific workflows.
[AI-30] CodeMimicry: Exploiting Safety Generalization Lag in Large Language Models via Structured Code Completion NEURIPS2026
链接: https://arxiv.org/abs/2609.39902
作者: Zhen Liang,Hai Huang,Wentao Chen
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: This paper will be accepted at NeurIPS 2026
Abstract:Large language models have achieved remarkable capabilities across diverse domains, yet their safety alignment remains vulnerable to jailbreak attacks. In this work, we identify a previously underexplored failure mode - safety generalization lag - where alignment trained predominantly on natural language fails to transfer to the code domain. We show that this lag induces a code-completion blind spot, allowing malicious intent embedded within syntactically valid code to evade safety mechanisms. To exploit this vulnerability, we propose CodeMimicry, a fully automated black-box jailbreak framework that generates structured, object-oriented code prompts to induce harmful outputs via code completion. Experiments on 8 state-of-the-art commercial LLMs demonstrate that CodeMimicry achieves a 96.25% attack success rate with 1.51 queries on average, significantly outperforming both template-based and optimization-based baselines. Beyond empirical performance, we provide a mechanistic analysis of code-based jailbreaks through latent space representations, including projection onto refusal-related directions and activation steering. This analysis offers an explanation of how CodeMimicry bypasses safety mechanisms in code-related domains. Our findings reveal a weakness in current safety alignment and highlight the need for robust alignments in structured domains such as code.
[AI-31] Do Better Goal Representations Improve Goal-Conditioned Reinforcement Learning?
链接: https://arxiv.org/abs/2609.39901
作者: Syed Nazmus Sakib,Abdul Monaf Chowdhury,Nafiul Haque,Shifat E Arman,Md Mehedi Hasan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 21 pages, 12 figures, 6 tables
Abstract:Goal-conditioned reinforcement learning (GCRL) relies heavily on how target goals are represented to the policy. While recent methods encode goals via temporal distance, occupancy, or controllability, it remains unclear how much downstream performance actually depends on representation quality. We study this in offline GCRL by constructing an exact temporal-distance goal representation in deterministic mazes. We then systematically corrupt its geometric quality while keeping the downstream learner fixed. Across OGBench navigation tasks and two algorithms, large changes in goal-representation quality produce almost no change in performance. However, applying the same interventions to the agent’s current state more than doubles success, revealing the state pathway as the true bottleneck. Building on this insight, we show that simple random Fourier positional encodings substantially improve performance on the hardest navigation tasks without map information or objective modifications. Overall, our findings suggest that in state-based offline navigation, improving how the agent’s current state is represented matters far more than refining the goal representation. Code will be released soon.
[AI-32] Algorithmic Recourse Under Competition
链接: https://arxiv.org/abs/2609.39877
作者: Shahin Jabbari
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Algorithmic recourse provides individuals who have received undesirable outcomes from machine learning models with suggestions for minimum-cost improvements to achieve the desired outcome. A central assumption when computing recourse is that the decision rule remains fixed throughout the recourse implementation phase. We challenge this assumption in settings where individuals compete for limited resources. In such settings, widespread recourse implementation can change the acceptance threshold even when the scoring model that is used to evaluate individuals remains the same. This change in acceptance threshold can, in turn, invalidate the original recourse recommendations (i.e., following the recourse may not lead to the desired outcome). To address this problem, we introduce a framework called recourse under competition that jointly optimizes for recommendation recipients and the recommended score target they need to satisfy to balance the recourse cost and post-shift validity among initially rejected individuals. We develop an algorithm based on the Implicit Function Theorem and empirically analyze its performance. Experiments on synthetic and real datasets show that personalized score targets can achieve higher validity, albeit at a higher cost. In contrast, common score targets generally offer favorable cost-validity trade-offs for lower to medium validity values.
[AI-33] Completion-Aware Cross-Fidelity Offline-to-Online Reinforcement Learning for Multi-Line Bus Holding
链接: https://arxiv.org/abs/2609.39868
作者: Yifan Zhang,Qifan Zhang,Liang Zheng
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Exploratory reinforcement learning (RL) on an operating bus fleet is impractical,while policies trained only from historical data cannot acquire new experience. Hybrid Offline-and-Online (H2O) RL combines fixed target replay with simulator interaction, but the inexpensive online simulator can differ from the target in transition and event-duration dynamics. We study this cross-fidelity problem for multi-line bus holding and address a failure mode in which lower generalized passenger time coexists with incomplete passenger journeys.
[AI-34] Learning from Runtime Feedback through Failure-Bank Self-Evolution for Vision-Language-Action Models
链接: https://arxiv.org/abs/2609.39820
作者: Mingyue Cui,Zheyuan Liu,Yihan Zhu,Zheyuan Zhang,Meng Jiang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: Runtime-feedback-driven self-evolution for safer VLA policies
Abstract:Vision-language-action (VLA) models generalize broadly across robotic manipulation tasks, but complex environments require balancing task success with unintended contact. Runtime shields can correct individual actions, but they leave the underlying policy unchanged, so repeated disagreements may create a persistent policy-shield mismatch that blocks task progress. To address this challenge, we introduce FailBank, a four-stage self-evolving framework that converts runtime feedback into persistent policy improvement. During collection, a fixed CBF-based safety module serves as an observe-only teacher, producing counterfactual corrections while the policy remains in control. Outcome-aware admission then converts useful proposals into corrective targets and retains successful uncorrected actions as quiet anchors for guarded LoRA updates. We evaluate FailBank on the VLA-Arena benchmark across two difficulty levels and two VLA backbones. Compared with the base policies, FailBank improves the joint success-cost operating point. Across the two backbones, FailBank improves task success rate by 8.5 and 6.9 percentage points, while reducing policy-induced cumulative cost by 35.6% and 23.8%, respectively. Compared with runtime shielding, FailBank raises task success rate by 25.4 and 9.5 percentage points, while maintaining comparable policy-induced cumulative cost. These results show that runtime feedback can serve as persistent policy supervision rather than only as a temporary action constraint.
[AI-35] Probabilistic Adversarial Training
链接: https://arxiv.org/abs/2609.39798
作者: Andi Zhang,Xingyu Zhao,Siddartha Khastgir
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:Building on a probabilistic perspective in which adversarial examples arise from the overlap between a distance-based distribution p_\mathrmdis and a victim-classifier-induced distribution p_\mathrmvic , we start from a simple intuition: adversarial examples become harder to generate when these two distributions are pushed apart, as their overlap becomes smaller, thereby increasing robustness. This intuition naturally motivates a KL-based robustness objective. We then prove that \mathrmKL(p_\mathrmdis|p_\mathrmvic)-\log Z_\mathrmvic is a lower bound on probabilistic robustness (PR), where Z_\mathrmvic denotes the normalizing constant of p_\mathrmvic . Since PR is generally intractable to compute directly, maximizing this KL-based lower bound provides a tractable surrogate objective for improving PR. We further show that this objective recovers a scaled form of adversarial training, offering a probabilistic interpretation of adversarial training and a principled route to robustness improvement. We call the resulting method probabilistic adversarial training. Experiments show that it consistently improves PR, and ablation studies demonstrate that the induced scaling factor can even enhance the PR of non-probabilistic adversarial training methods.
[AI-36] Pseudo-Label-Triggered Retraining from Forecast Errors for Online Time Series Forecasting
链接: https://arxiv.org/abs/2609.39789
作者: Yeryeong Kwak,Yoo-Min Jung,Jonghun Park
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Real-world time series forecasting systems operate under non-stationary data streams, where forecasting performance may degrade over time. Although retraining can recover the performance, it incurs non-trivial computational and operational costs. Under limited deployment resources, the key challenge is therefore not only how to retrain but also when to retrain. While existing retraining policies often rely on indirect indicators such as drift alarms or model staleness, we instead use realized forecast errors as direct deployment feedback. In this paper, we propose PILOT (Pseudo-label-Informed Learned Online Trigger), an online retraining framework that learns when to retrain from forecast-error dynamics. Since ground-truth retraining labels are unavailable, PILOT constructs a pseudo-label from future increases in forecast error and trains a lightweight scorer to predict it from observed error states. At deployment, PILOT uses only completed forecast errors and serves as a plug-in module for arbitrary forecasting backbones without architectural modification. We evaluate PILOT under standard multivariate forecasting settings across eight benchmarks with three representative backbones—DLinear, iTransformer, and TimesNet. Across all three backbones, PILOT achieves state-of-the-art average-rank performance among retraining policies while maintaining a favorable performance–efficiency trade-off.
[AI-37] How Does Local Landscape Geometry Evolve in Language Model Pre-Training?
链接: https://arxiv.org/abs/2609.39767
作者: Zhanpeng Zhou,Yuhan Sun,Bingrui Li,Jinbo Wang,Huaijin Wu,Lei Wu,Junchi Yan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 23 pages, 15 figures
Abstract:The scale and expense of pre-training language models make efficient hyperparameter tuning essential, yet a principled guidance is still missing. In this work, we analyze language model pre-training dynamics from a local landscape geometry perspective. Our study reveals two distinct phases. In Phase I, sharpness of the local landscape is initially high, leading to instability and loss plateaus under large learning rates (LRs). The landscape shifts from sharp to flatter regions early in training. This dynamic explains the necessity of LR warmup and further suggests that larger peak LRs require proportionally longer warmup periods. In Phase II, the local landscape is governed by the gradient noise scale. Our theory identifies a depth flatness trade-off: high noise from smaller batches widens the loss basin, whereas reduced noise from larger batches deepens it. This theory motivates a dynamic batch-size (BS) scheduler that begins with a small BS and increases it late in training. Together, we provide a unified view of loss landscape evolution, which translates into actionable tuning strategies for large-scale pre-training.
[AI-38] DiffWAM: A Fast and Efficient Navigation World Action Model
链接: https://arxiv.org/abs/2609.39763
作者: Mo Zhu,Yuze Wu,Xijie Huang,Xiao Cui,Fei Gao,Xin Zhou
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 32 pages,10 figures, 8 tables
Abstract:Pretrained video foundation models encode rich semantic and spatiotemporal priors for embodied navigation, yet converting these priors into UAV motion typically requires expensive future-video synthesis and geometric reconstruction. We investigate whether the motion implicit in future visual prediction can instead be recovered directly from the predictive representations of a frozen video model. To this end, we present DiffWAM, a geometry-conditioned navigation world-action model that directly transforms multi-level predictive features into continuous camera trajectories. Its Grid-Motion module preserves spatial-temporal motion associations, while Latent2Pose grounds them with first-frame geometry to recover metrically meaningful 3D motion. Complete video rollouts and geometric reconstruction are required only for offline supervision, eliminating future-video decoding and multi-frame reconstruction during deployment. We further introduce FastDreamer, which overlaps predictive and geometric computation with ongoing flight and performs timestamp-aware asynchronous trajectory handoff for continuous UAV execution. DiffWAM achieves a trajectory RMSE of 0.3492 m and an endpoint success rate of 74.40% on the 1,000-sample DiffWAM-1000 benchmark, while representative real-world experiments demonstrate complex behaviors including constrained traversal, orbiting, S-shaped flight, and multi-stage navigation. An onboard DiffWAM-Flash implementation further reaches 1.08 s model-pipeline latency on NVIDIA Jetson AGX Thor. These results demonstrate that predictive video representations can be efficiently grounded into continuous 3D motion, providing a direct alternative to generate-then-reconstruct navigation pipelines. Project page: this https URL.
[AI-39] rust Is Not a Score: Runtime Assurance Contracts for High-Risk AI Agents
链接: https://arxiv.org/abs/2609.39717
作者: Serhii Zabolotnii
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Software Engineering (cs.SE)
备注: 16 pages, 2 figures, 5 tables. Ancillary files: decision log, executable transition model, LLM-labelled synthetic holdout. Synthetic mechanism study; no deployment claim
Abstract:Benchmarks, audits, and agent protocols describe performance, permissions, and repair, but not how observed evidence should change an agent’s authority during a consequential task. We call this the assurance-transition gap. We propose a Runtime Assurance Contract (RAC), a policy-level formal schema binding autonomy boundaries, component eligibility, evidence state, transition policy, human-review capacity, and non-compensatory gates. Under RAC, soft metrics may inform routing, whereas a failed or unknown mandatory gate forces retry, switch, escalation, deferral, or stop; aggregate performance cannot authorize action. We define the contract, an evidence record, a permission rule, and five invariants, and illustrate them in clinical, industrial, and judicial failure probes. We then report a deterministic failure-injection study in agentic coding: 280 constructed cases evaluated by a gate conjunction, a score-only rule, and a restricted protocol baseline. At the published example weights and threshold, the score rule admits 80 of 100 block-required injections and all 40 review-required injections. Tuned in hindsight, it matches the conjunction on this corpus. For positive weights, a positive threshold, binary risk signals, zero-signal controls, and an injected case firing each signal alone, we show that exact agreement holds if and only if the threshold does not exceed the smallest weight. A separate set of 18 hand-authored traces checks version-pinned evidence and review transitions against simpler policy variants. In a further prospective synthetic holdout of 24 episodes, two blinded LLM judges assign identical labels to all 72 action attempts; RAC and a separately implemented full stateful baseline both match these labels. These studies test mechanisms on synthetic cases; they establish neither deployed safety nor cross-domain effectiveness.
[AI-40] ArchitectureIQ: On the Measure of Training Intuition
链接: https://arxiv.org/abs/2609.39714
作者: Zirui Ren,Shaoyang Guo,Chencheng Tang,Jinxin Wang,Chengyu Xiong,Shanbin Yu,Peihang Li,Yidi Wu,Bangzhe Huang,Qingyu Qu,Leqian Yang,Ziming Liu
类目: Artificial Intelligence (cs.AI)
备注: 29 pages, 10 figures. Code and reproduction materials: this https URL
Abstract:Top researchers have good intuition, but do language models have as good intuition about model training as top AI researchers? To measure model intuition of LLMs and humans, we introduce the ArchitectureIQ benchmark. Each question presents a synthetic dataset and several training recipes, and the test-taker is asked to predict the recipe yielding the best test metric. Overall, we find that LLMs’ model intuition is good but has four limitations: (1) The intuition is imperfect, or even sub-human in some cases. Frontier models achieve around 76% accuracy (random choice 33%) vs best human researcher (66.0%), yet remain far from perfect. For architecture-only questions, best human achieves 65% while GPT-6 Astra only has 38%. (2) The intuition is empirical, not structured, supported by the fact that more CoT compute does not lead to substantial improvement. Unlike math, we still lack a “Science of AI” language that enables structured reasoning on AI. (3) The intuition is not maximally condensed, and can be further compressed into a knoledge base. Our constructed knowledge base with only 20 items yields large gains for weak models: GPT-4o equipped with the accumulated knowledge almost matches the performance of Claude Opus 5. (4) The intuition is insensitive to dataset properties, but the best model should in general depend on data properties. This suggests that data is the real “dark matter” in AI – LLMs (so do human researchers) understand too little about data, even less than model architectures.
[AI-41] Values as Style: Disentangling Values from Semantics with One-Way Mixing for Low-Damage LLM Steering
链接: https://arxiv.org/abs/2609.39701
作者: Jiale Dai,Hongcan Deng,Liuxian Ma,Xiaoke Niu,Guojie Song
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Machine Learning (cs.LG)
备注:
Abstract:Value steering should change an LLM’s normative priorities while preserving the scenario, facts, and task constraints underlying its answer. Conventional activation edits often change both. We introduce an editable semantic-value interface on frozen residual states, with a one-way semantic-to-value pathway that grounds value recognition in context. Stop-gradient blocks feedback through this pathway; swap consistency, topic de-confounding, and decorrelation encourage selective codes. At inference, editing the value code produces a residual delta while holding the semantic code fixed. On two instruction-tuned backbones, this interface improves semantic preservation and reduces benign refusals at comparable value alignment. A matched mixing-by-gating ablation separates representation learning from selective edit activation, and dimension-matched probes establish improved code selectivity. Against validation-selected prompting on LLaMA-3.1-8B, the method achieves comparable alignment (0.750 vs. 0.748), higher BERTScore (0.938 vs. 0.923), and fewer contradictions (5.1% vs. 7.6%). Human ratings and cross-taxonomy controls provide complementary evidence for low-damage value steering.
[AI-42] GFD-OPD: Guidance-Folded On-Policy Distillation of Diffusion Models Across Scales
链接: https://arxiv.org/abs/2609.39692
作者: Zhenxing Zhang,Jiayan Teng,Wenxu Wu,Zhuoyi Yang,Jiazheng Xu,Wendi Zheng,Jie Tang,Dan Guo,Meng Wang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:On-policy distillation (OPD) has demonstrated two important capabilities in language models: compressing large teachers into smaller students and merging expert models into a single model. Existing diffusion OPD, however, mostly focus on the latter, with teachers and students sharing the same backbone and scale. We investigate large-to-small diffusion opd from large teachers to a small student and find that the standard recipe fails. To find the underlying cause, we propose Fixed-State KL, an effective and fair way to measure the distribution gap between student and teacher during OPD training for diffusion models. We are the first to clarify why large-to-small OPD is challenging for diffusion models: a smaller student struggles to perfectly match the distribution of a larger teacher, while classifier-free guidance can accumulate and amplify the distributional discrepancies between the student’s conditional and unconditional branches and those of the teacher. To solve this problem, we propose GFD-OPD, a simple yet effective method that reduces the student-teacher gap while avoiding the error amplification of the CFG composition. Across numerous experiments, GFD outperforms previous baselines in both training efficiency and final performance, achieving state-of-the-art results on all benchmarks.
[AI-43] RoboCoach: World Models as Active Coaches for Compositional Robot Skills
链接: https://arxiv.org/abs/2609.39685
作者: Jiajun Liu,Yifan Chen,Yichao Liu,Jiayi Zhang,Ruoqu Chen,Shaoxuan Xie,Guocai Yao,Mengdi Xu,Sen Cui,Changshui Zhang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: this https URL
Abstract:Long-horizon robot manipulation reuses skills across many task compositions, but improving these compositions with additional end-to-end demonstrations is costly. A practical self-improving system must decide both what to teach next and where to apply that supervision. We present ROBOCOACH, a world-model-guided coaching framework that uses imagined failures to guide demonstration requests and expert updates. Its Route-Imagine-Diagnose-Improve (RIDI) loop executes reusable skill experts inside COACHWORLD, our shared action-conditioned world model, and uses a progress judge to record the first subtask that fails to complete. Aggregated records select which subtask demonstrations to acquire and which expert adapters to update. Across two simulation suites and two real-robot platforms, imagined and deployed success correlate over 22 task-policy pairs (rho = 0.840). Controlled comparisons show that our coaching method outperforms matched baselines under matched data budgets and update schedules. With only 150 additional subtask demonstrations, success rises from 13.3% to 75.0% on Franka and from 40.0% to 83.8% on AgileX. The coached experts also transfer to four held-out compositions, achieving an average success of 35.0%, compared with 0% for a shared-policy baseline updated with uniformly acquired demonstrations. Together, these results show that world models can serve as active coaches, turning imagined failures into targeted supervision for modular policy improvement. Project Page: this https URL
[AI-44] ChronoGraph: Functional 4D Scene Graphs with Vision-Language Models for Interaction Understanding and Grounded Planning
链接: https://arxiv.org/abs/2609.39665
作者: Chenyangguang Zhang,Malgorzata Gwiazda,Guanlong Jiao,Yuanchen Ju,Federico Tombari,Koushil Sreenath,Marc Pollefeys,Sunghwan Hong
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Embodied agents must determine where to act, anticipate the resulting scene changes, and interpret observed outcomes to guide subsequent actions. This requires connecting 4D interaction understanding, which explains how past actions changed the scene, with spatially grounded planning, which determines how and where to act toward a goal and anticipates the resulting scene changes. We introduce ChronoGraph, a functional 4D scene graph that links actions on affordance parts to semantic and geometric state changes. By representing observed and anticipated transitions in the same form, it provides a shared basis for understanding and planning. We construct ChronoGraphBench through an automatic data engine that converts human-interaction videos and simulated robot trajectories into graph-annotated questions for training and evaluating Vision-Language Models (VLMs) on both tasks. Using these annotations, we train ChronoGraphVLM by adapting pretrained VLMs in two stages. Graph-as-Chain-of-Thought supervised fine-tuning teaches the models to reconstruct observed transitions and predict future ones as graph traces before answering. Subsequent joint 4D graph reinforcement learning directly rewards graph properties and answer correctness. Experiments across model scales show improvements over the corresponding pretrained baselines and zero-shot transfer to VLM4D. Real-world demonstrations further show that graph-based planning and affordance grounding support mobile manipulation through existing robot skills without additional fine-tuning.
[AI-45] Free Everywhere Exact on Trees: PPOs Dropped Correction Buys Sample Efficiency Under Aggressive Reuse
链接: https://arxiv.org/abs/2609.39634
作者: Nima H. Siboni
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Common policy improvement methods, including TRPO, PPO, and GRPO, estimate policy improvement under the behavioral policy’s state-visitation distribution rather than the improved policy’s own. The substitution makes the objective estimable from the behavioral policy’s rollouts but adds a bias growing with policy divergence, hence the trust region or clip, and hence no reuse of a batch far off-policy. We show that under history-injective dynamics, where each state is reached by exactly one history, the dropped state-visitation ratio equals the product of per-step policy ratios along the sampled prefix, on every trajectory and not only in expectation. The ratio is therefore restored exactly, from log-probabilities PPO already computes. Autoregressive generation and canonical-order constructive optimization are both history-injective. The exact correction pays importance-sampling variance that grows with the horizon, so we generalize it to a one-parameter family with PPO ( \alpha=0 ) and the full correction ( \alpha=1 ) as endpoints: a single bias–variance knob. A gradient-level analysis of the unclipped surrogate identifies two channels the correction acts through and three conditions under which it carries signal; an enumerable testbed confirms the conditions’ predictions. On hard credit-assignment scheduling tasks, a short corrected warmup with aggressive early sample reuse learns faster than PPO and than the same reuse uncorrected; the marginal gain grows with task difficulty ( +0.02 to +0.09 learning-curve AUC), and the early win over PPO tracks the prefix bias that reuse incurs. A correction held throughout, or applied where clipping already contains the reuse bias, is null to harmful.
[AI-46] Parameterization method of reservoir properties for ensemble-based data assimilation using intermediate latent space of StyleGAN
链接: https://arxiv.org/abs/2609.39626
作者: Marcio A. Sampaio,Paulo H. Ranazzi,Martin J. Blunt
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Ensemble smoothers are the most successful and efficient techniques currently available for history matching. However, because these methods rely on Gaussian assumptions, their performance is severely degraded when the prior geology is described in terms of complex facies distributions (non-Gaussian). In this way, for these methods, we need to apply efficient parameterization techniques. Currently, the most efficient methods for performing parameterization are deep learning models. However, given the variety of existing deep learning models, studies have not identified which is most suitable for use with ensemble-based methods, although some important models had already been evaluated. Based on a recent literature review, the most promising models selected were VAE-GAN, Latent Diffusion, and StyleGAN models. As a novel aspect of this work, data assimilation with the second generation of StyleGAN (StyleGAN2) model was performed using the latent z-space and intermediate w-space, separately. They were applied in two 2D case studies: one categorical (three facies) and the other continuous. The results demonstrated that all three models are highly efficient, with the StyleGAN2 model standing out for generating samples with geological realism and achieving excellent data matching in the cases studied. Our findings show that performing data assimilation with StyleGAN2 using the intermediate space (w-space) yielded better results than the traditional application in the latent space (z-space). This is due to the fact that ESMDA uses linear updates and the w-space is much more linear and disentangled than the highly entangled z-space, thereby ensuring that the updated vectors remain close to realistic geological patterns. These results were validated using main geostatistical and history matching metrics.
[AI-47] Pretext: Defeating Malicious Skill Detection Frameworks for AI Agents NEURIPS2026
链接: https://arxiv.org/abs/2609.39607
作者: Tobias Kaisar,Aritra Dhar
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: Accepted in AIWild@NeurIPS 2026
Abstract:Skills extend an agent’s capabilities by injecting instructions and information into the context, and are widely used by agents such as OpenClaw and Claude Code. Prior work shows third-party marketplaces host malicious skills that give attackers direct influence over the victim’s agent. The emerging defense scans skills before installation, pairing deterministic static checks with an LLM-based semantic judge, as in NVIDIA’s SkillSpector. We show that such defenses fall to an attacker who knows the detector. Our white-box LLM attacker, Pretext, iteratively crafts skills that evade detection while still delivering the payload and performing the benign task: moving the payload from code into natural language leaves static analysis inert, while framing it as the skill’s legitimate purpose and splitting instructions across files keeps the LLM stage below its blocking threshold. Across three open-source models, Pretext achieves up to 97% and 77% against a frozen detector and a co-adaptive one, respectively, revealing major gaps in current skill scanners.
[AI-48] Why Do Conventional World Models Fail to Learn Cellular Automata?
链接: https://arxiv.org/abs/2609.39604
作者: Shaoyang Guo,Ziming Liu
类目: Artificial Intelligence (cs.AI)
备注: 35 pages, 18 figures. Code and reproduction materials: this https URL
Abstract:Although conventional world models - auto-regressive or diffusion models based on transformers or convolutional networks - may learn surface statistics of world dynamics, can they learn the exact world dynamics from its observed history? Leveraging cellular automata as a simple testbed, we find the answer to be no in many cases. Conventional architectures predict most pixels correctly yet rarely complete a rollout: a CNN predicts 96.3% of cells but completes 18.9% of rollouts; a joint diffusion model completes none. We trace the gap to three failure modes of these world models - namely, they fail to exactly capture spatial locality, temporal locality or temporal stability. Simple changes repair each: (1) for spatial locality, two-dimensional rotary positions lift a transformer from 39.1% to 100% on the Game of Life; (2) for temporal locality, handing each token its cell’s previous-frame neighbourhood lifts the same transformer from 25.8% to 99.9% on unseen rules; (3) for temporal stability, causal freezing lifts the same diffusion weights from 42.2% to 99.9%. None of the three changes touches the architectural backbone; each only modifies the information flow within it. We also compare joint and ordered sampling on billiards and, in an exploratory study, on a simulated Burgers equation.
[AI-49] xt-to-3D Policy: Fine-Grained Language-Behavior Alignment for Unseen Specification Generalization
链接: https://arxiv.org/abs/2609.39599
作者: Xinhao Yang,Wenhao Wu,Ning Lv,Yanshen Ding,Zhenhong Sun,Daoyi Dong,Chunlin Chen,Zhi Wang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 24 pages, 8 figures, 8 table
Abstract:3D visuomotor policies provide a strong foundation for spatially precise manipulation, yet current text-to-3D policies struggle to follow unseen fine-grained behavioral specifications beyond those covered by demonstrations. We study this challenge as unseen specification generalization, where language specifies behaviorally significant variations, such as target position, displacement, or articulated state, that are absent from policy training. We find that pretrained language representations and conventional global behavior-language alignment capture coarse task semantics but often blur nearby specifications that require distinct behaviors. We introduce T3DP, a Text-to-3D Policy framework for fine-grained language-behavior alignment. Rather than compressing each instruction and demonstration into a single global embedding, T3DP preserves their local structures and establishes bidirectional token-level correspondence between linguistic elements and behavioral segments. This directly grounds subtle linguistic variations in the behavior components they affect, preventing closely related specifications from collapsing in the representation space. The resulting specification-sensitive language representation conditions a point-cloud-based 3D diffusion policy, enabling more precise control over unseen behavioral specifications without modifying the underlying policy architecture. Across Meta-World, ManiSkill, and RoboTwin, T3DP improves average held-out-specification success over global language-behavior alignment by +11.0-14.2 points, with gains on all 15 task families; on real-robot tasks, it further raises average success from 47.5% to 65.0% (+17.5 points). Representation and action-probe analyses show that fine-grained alignment better preserves specification geometry and action-relevant variation, linking local behavior grounding to downstream control.
[AI-50] Robust Transfer Learning for Paper ECG Recognition
链接: https://arxiv.org/abs/2609.39581
作者: Yinghao Xie,Zhenbang Dai,Haojun Wang,Jinyu Cai,Fabio Bonassi,Hongwu Chen,Johan Sundström,Jiawei Li,Antônio H. Ribeiro
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Paper ECG recognition is challenging because real-world ECG images vary in layout, physical artifacts, and label availability. We introduce RobECG-CL, a rank-aware contrastive learning framework for robust paper ECG representation learning. Starting from standard 12-lead ECG recordings, we construct progressively degraded paper ECG views with heterogeneous layouts and train the model to balance same-recording invariance with degradation-aware ordering. Across synthetic stress tests on CODE-II and EchoNext, RobECG-CL improves robustness under severe degradation and few-shot transfer, outperforming contrastive learning baselines and surpassing the waveform-based foundation model, ECG-FM, in the 1% labeled setting. On 312 samples of hospital data with 37 labels, RobECG-CL achieves the best macro AUROC.
[AI-51] AVERT-VLN: Abstention-aware Visual Error Recovery and Training for Vision-and-Language Navigation
链接: https://arxiv.org/abs/2609.39579
作者: Minrui Liu,Jingke Wang,Yuehao Huang,Hao Su,Jiajun Lv,Yukai Ma,Yong Liu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Deploying vision-and-language navigation (VLN) agents in unseen environments remains challenging because unfamiliar layouts and visual conditions can cause execution to go off track. Rather than relying on continuous human supervision, a practical strategy is to selectively request corrective guidance, recover the ongoing task, and reuse corrective interactions to improve subsequent navigation. We propose Abstention-aware Visual Error Recovery and Training for Vision-and-Language Navigation (AVERT-VLN), a closed-loop framework that uses a plug-in vision-language Monitor for online human-assisted recovery and offline preference learning. The Monitor operates separately from navigation decision generation and assesses instruction-execution consistency from the instruction, visual history, and current observation. To train the Monitor for deviation recognition, we construct LOSTNAV DATASET with 20K counterfactual risk trajectories and rule-based deviation labels. The Monitor is first fine-tuned on 40K normal trajectories to assess instruction progress and then jointly fine-tuned on normal and risk trajectories to recognize semantic deviations. At runtime, Asynchronous Sidecar Monitoring evaluates execution alongside the navigation model. When the controller accepts a LOST verdict, it suspends autonomous execution and requests human guidance for recovery. For offline policy improvement, Trajectory-Anchored Preference Learning converts deviation-associated failures into decision-level preference pairs under shared decision contexts, restricting supervision to the decisions targeted for correction. Under human-assisted evaluation, the full AVERT-VLN system achieves success rates of 76.2% and 66.3% on the val-unseen splits of R2R-CE and RxR-CE, respectively. The same monitoring and human-assisted recovery interface also improves success rates across the three evaluated navigation architectures.
[AI-52] ECHO-G: Embodied Co-speech Humanoid mOtion Generation
链接: https://arxiv.org/abs/2609.39575
作者: Yizhao Li,Pusen Gao,Ming Wang,Shaojie Shen,Shuo Yang,Hao Xu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 8 pages, 5 figures, 3 tables. Project page: this https URL
Abstract:Generating full-body co-speech motion for humanoid robots requires coordinating speech prosody, linguistic content, and embodiment-specific motion. To this end, we present ECHO-G, a framework jointly conditioned on speech audio and timed transcripts. Its Speech-Grounded Diffusion Transformer (SGDiT) combines frame-aligned acoustic features with token-level linguistic context, preserving their distinct granularities. Trained with rectified flow matching, it models one-to-many utterance-motion relationships directly in robot space. To support training and evaluation, we introduce a BEAT2-derived audio-text-robot dataset and a benchmark covering co-speech characteristics, robot-motion quality, and runtime efficiency. Comparative evaluation supports direct robot-space generation over the evaluated human-motion generation and retargeting pipelines, while modality ablations highlight the benefits of joint audio-text conditioning. We further demonstrate deployment on a physical humanoid robot. A complementary video-rating study also favors joint conditioning over the alternatives. The dataset and training, inference, and evaluation code are available through our project page.
[AI-53] Self-Spec Verifiable Code Generation
链接: https://arxiv.org/abs/2609.39568
作者: Jiaru Qian,Yihong Dong,Yongmin Li,Hao Zhu,Bin Gu,Ge Li
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) may generate unreliable code on corner cases missed by testing, while formal verification can provide machine-checkable guarantees. Recently, researchers have proposed several benchmarks to evaluate the capabilities of LLMs in generating formally verifiable code, where LLMs need to formulate formal specifications, generate the corresponding code, and verify its correctness. However, existing benchmarks have two key limitations: (I) They primarily evaluate specification and code generation stage-wise, with code generation typically conditioned on an oracle specification. This setup overlooks whether strong stage-wise performance translates into end-to-end success. (II)They mainly focus on a single proof-oriented language and mathematically structured tasks, offering limited coverage of tasks common in software development. In this paper, we introduce VeriCodeBench, a benchmark for self-spec verifiable code generation, where the LLM relies solely on its own generated specification and code throughout the entire process. VeriCodeBench contains 400 language-native problems across C, Java, Rust, and Python, covering practical concerns in software development. We evaluate specification coverage, code validity, and joint problem-level success. We further introduce CodeNova to enhance the capabilities of LLMs in self-spec verifiable code generation. CodeNova makes requirements explicit through constraint-guided specification and uses verifier feedback to guide targeted implementation repairs. Experimental results reveal that self-generated specifications remain a major bottleneck, while providing more sophisticated specifications may not necessarily lead to higher verification success rates. CodeNova substantially improves performance across all evaluation metrics, enabling Claude Sonnet 5 to achieve the strongest results under the self-spec protocol.
[AI-54] A2Z GameSpec-Bench: How Faithfully Can Coding Agents Generate Games from Game Design Specifications?
链接: https://arxiv.org/abs/2609.39564
作者: Seonho Lee,Wonryeol Jeong,Alberto Cereser,Inha Kang,Hyeonjong Kim,Seungmin Kwak,Dongmin Park
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Delegating complete application development to coding agents requires preserving the intended design rather than simply producing plausible outputs through naive prompting. Game development provides a demanding testbed, as long-form Game Design Documents (GDDs) describe requirements that must work together across game logic, visual rendering, and player interactions. However, existing game-development benchmarks typically use compact specifications and provide limited support for evaluating interdependent requirements across these aspects in long-form GDDs. We introduce A2Z GameSpec-Bench, a benchmark of 100 long-form GDDs for evaluating end-to-end game development by agents. We measure faithfulness by checking whether the game satisfies the GDD requirements and preserves the relationships among them. Each GDD is turned into a dependency-aware contract that contains rules, constraints, and prerequisite relations. Following game-development practices, we combine source-code inspection with agent-generated test policies for scenario-based replay and adaptive playtesting. The contract remains fixed across agents and revision rounds, while judgments and evidence linked to the same requirements support consistent comparison and failure detection. Our evaluations show that current agents struggle to jointly satisfy interdependent requirements across code implementation and actual play. Requirement-specific feedback improves GDD Fidelity by 10.9% relative to self-revision after two rounds. A2Z GameSpec-Bench assesses end-to-end specification-following ability beyond implementation judgments and provides targeted feedback to support more faithful game development. Code and datasets are available at this https URL.
[AI-55] Candidate Retention for Abductive Learning
链接: https://arxiv.org/abs/2609.39561
作者: Hao-Yuan He,Yu Liu,Ming Li
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Abductive learning combines neural perception with symbolic reasoning, using explanations generated by abduction to supervise the perception model. Multiple valid explanations of the same symbolic target can assign conflicting labels to the same inputs. Common policies select a single candidate as a pseudo-label, which may reinforce mistaken assignments, or weight all candidates, which may spread supervision across competing labels. These risks motivate selecting a retained subset to balance supervision sharpness and model-mass coverage. To guide this choice, we bound the coordinate-level supervision error using retained uncertainty, discarded model mass, and model mismatch. For a fixed model and training pair, only the first two terms depend on the retained set. We propose Abductive Candidate Retention (ACR), which uses these terms to guide greedy additions, accepting a candidate when its recovered mass exceeds the increase in retained uncertainty. Experiments show that ACR improves concept accuracy over single-candidate baselines and A3BL in most evaluated aggregated mod-addition settings. Objective ablations support the joint use of uncertainty and posterior mass.
[AI-56] Divide and Collapse: MAPF-Collapse via Exact Decomposition into Independent Sub-Instances
链接: https://arxiv.org/abs/2609.39559
作者: Oren Salzman
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:In this work we study the problem of MAPFC, a post-optimization step for Multi-Agent Path Finding (MAPF) plans where we are given a feasible plan produced by a modern MAPF solver and are tasked with removing avoidable moves while preserving feasibility. This NP-hard problem naturally arises when using learning-based state-of-the-art (SOTA) solvers which construct plans that contain redundant moves that can be removed. Recently, Tang et al. presented Judgelight, which uses Integer Linear Programming (ILP) to solve MAPFC. Importantly, the ILP is constructed over all agents jointly, so its cost is governed by the full instance rather than by the small coupled residue that actually requires joint reasoning. Our key insight, motivating this work, is that MAPFC instances naturally decompose into independent sub-problems, most of which involve a single agent and can be solved without any inter-agent reasoning. To this end, we first identify which agents need to coordinate their motion and partition the instance into sub-problems accordingly. For the cases where no coordination is required, we introduce an extremely lightweight solver that is \approx!1,900\times faster than Judgelight. For cases where coordination is required, Judgelight can be used but we introduce an alternative CBS-like solver which is more efficient on easier problems. The resulting framework is exact, uses no commercial ILP solver, and matches Judgelight’s quality while running substantially faster on the coordination-light majority of instances; on the coordination-heavy instances we propose a regime-aware hybrid planner that falls back to Judgelight. Over all benchmarks tested, this planner achieves a median 10.5\times per-instance speedup over Judgelight.
[AI-57] RankEvolve: A Reliable Multi-Agent Auto-Research Harness for Evolving Ranking Models
链接: https://arxiv.org/abs/2609.39551
作者: Zheng Chen,Linfeng Liu,Hong Li,Hong Yan
类目: Artificial Intelligence (cs.AI)
备注: 29 pages, 5 figures, 13 tables, 1 algorithm; includes appendices
Abstract:Auto-research agents, LLM systems that propose, implement, train, and evaluate model changes across iterations, promise to automate applied ML’s experimental loop. Over long horizons, execution accuracy is a binding constraint: a change can silently leak held-out data, omit normalization, disconnect a gradient, or leave a train/eval flag unwired, invalidating expensive runs and compounding error across iterations. We present RankEvolve, an auto-research framework for evolving generative ranking models. An Executable Operating Protocol (EOP) declares phases, gates, branches, and loops, and the runtime enforces the compiled state machine. A meta-meta-harness composes complete black-box coding-agent products, including Claude Code and Codex, as execution-graph nodes that review and repair one another’s work. In a budget-matched evaluation, heterogeneous composition raises all-oracle execution accuracy from the best single-product baseline of 45.8 percent to 62.5 percent (paired +16.7 points, 95 percent CI [6.6, 26.7]) while achieving a 10.4 percent silent critical-defect rate. An implemented knowledge layer carries findings, including negative results, across iterations. In a twelve-iteration deployment on the open-source HSTU recommender, RankEvolve reported NDCG@10 of 0.2192 on MovieLens-20M LARGE (+4.48 percent over the published anchor) and 0.1948 on BASE (+2.80 percent). ExecML-HSTU, seeded by incidents from that deployment, provides the oracle benchmark for the execution-accuracy evaluation. A pre-specified LitGPT transfer split replicates the heterogeneous-composition effect beyond recommendation (+12.5 points, 95 percent CI [3.0, 22.0]), and a paired ablation isolates per-step from full-protocol instruction injection. These results characterize when runtime-controlled composition of coding-agent products improves execution accuracy.
[AI-58] Growing an Agent /Prover Interface: Evolutionary Tool Design for Cost-Efficient Theorem Proving in Rocq and Lean
链接: https://arxiv.org/abs/2609.39544
作者: Jules Viennot,Guillaume Baudart,Marc Lelarge
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Recent achievements in AI-assisted mathematics require intensive interaction of agents with proof assistants to generate machine-checked proof certificates. Agents interact with proof assistants such as Rocq or Lean through an interface that controls what the agent receives from the prover and the cost of these interactions. Today, these interfaces are adapted from tools designed for humans and not optimized for agents. We propose an evolutionary method where a frontier model incrementally proposes new features and only keeps the ones that improve the overall performance of smaller models. We demonstrate the effectiveness of our method by growing, on a curated set of mathematical problems, \rme, a new MCP server for the Rocq prover. On the held-out \texttttest split of miniF2F-Rocq, an agent equipped with \rme outperforms both the baseline that only exposes the Rocq compiler and an established MCP server, across four models from two families, in success rate, cost per solve, and time per solve. Although evolved for Rocq, the resulting server transfers to Lean, improving cost and time per solve on a subset of PutnamBench. We release \rme and its port to Lean.
[AI-59] A Reusable Semantic Web Framework for Evidence-Grounded Fundamental Rights Impact Assessments under the EU AI Act
链接: https://arxiv.org/abs/2609.39537
作者: Faith Olopade,Delaram Golpayegani,David Lewis
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: Presented at the Fifth European Conference on Algorithmic Fairness (ECAF '26), Ghent, Belgium, 2-4 September 2026. Proceedings forthcoming in Proceedings of Machine Learning Research (PMLR)
Abstract:The EU AI Act (Art. 27) requires deployers of high-risk AI systems to conduct Fundamental Rights Impact Assessments (FRIAs) before deployment, yet the evidence needed for credible assessments is fragmented across incompatible incident repositories, risk vocabularies, and legal texts. We present a reusable Semantic Web-based framework that consolidates this evidence for two high-risk public sector categories: employment and worker management (Annex III(4)) and access to essential public services (Annex III(5)(a)). A curated 150-record corpus is annotated along four axes using keyword, LLM, and hybrid methods and serialised as a SPARQL-queryable knowledge graph of 1,351 RDF triples. Five FRIA demonstration scenarios surface 103 records (68.7% coverage). Evaluation against a 69-record gold standard reveals that LLM-assisted classification of the employment domain achieves only \kappa = 0.045 , a cautionary result for automated fairness-related evidence retrieval in this domain. All artefacts are released openly to support adoption by regulators, national authorities, and SMEs.
[AI-60] Disentangling Self-Distillation: Measuring and Modeling Acquisition and Retention
链接: https://arxiv.org/abs/2609.39494
作者: Luis Zuin,Alexis Huet,Dario Rossi,Zied Ben Houidi
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Self-distillation with privileged context adapts a language model from demonstrations by letting the model, once conditioned on a reference response, teach its context-free copy token by token. Our taxonomy reveals existing methods differ along three entangled axes: (i) the rollout source (student or teacher), (ii) the teacher coupling (frozen, or an exponential moving average of the student at some coupling rate) and (iii) the KL direction (reverse or forward), yet these axes are usually studied in fixed combinations and have led to conflicting conclusions. We formalize a unifying framework to encompass all self-distillation methods vs classic supervised fine-tuning: we train every combination of the three axes, on Qwen2.5-7B and Ministral-3-3B across ordinary and contradictory tasks, totaling 1,200 adaptation runs, to systematically investigate the impact of the above axes. We propose a controlled model of the same objective to explain the resulting acquisition-retention trade-offs. We find that (i) the rollout source matters mostly where the task contradicts the pretrained behavior: there teacher rollouts raise acquisition well above what student rollouts achieve, with almost no change in retention; (ii) the teacher coupling changes acquisition most, on every task: acquisition rises with the coupling rate, then falls past a task-specific rate; (iii) switching the KL direction costs retention in one model but not the other so which axis to tune first depends on the model. The controlled model reproduces the three trends.
[AI-61] Who Owns That? Evaluating Ownership Intuitions in Large Language Models
链接: https://arxiv.org/abs/2609.39483
作者: Xizhi Xiao,Yue Wu,Shan Xu,Jia Liu
类目: Artificial Intelligence (cs.AI)
备注: 26 pages, 11 figures
Abstract:Ownership establishes rights over the use, control, and transfer of objects. Understanding these relations is essential for AI systems to interact appropriately with people and their resources. Yet how large language models (LLMs) attribute ownership under competing claims remains unclear. We introduce the Competing Ownership Attribution Task (COAT), comprising 42 scenarios, and compare ownership allocations from 24 LLM configurations with those of 108 human participants. Overall, human-model similarity is close to human-human similarity, but models show greater homogeneity in their ownership judgments. Within individual answers, models also divide ownership more evenly among claimants than humans do. Pooling responses across model configurations reveals more scenarios with a shared judgment and fewer with distinct viewpoint groups than in humans. When humans form distinct groups, models may converge on one viewpoint or between competing viewpoints. Further comparisons reveal different contextual sensitivities. As material value increases across scenarios, allocations to creators decline less sharply in models than in humans. Across scenarios differing in public recognition of later holders as owners, allocations to these holders increase in models but decrease slightly in humans. Together, these findings suggest that the evaluated LLM responses do not fully capture the diversity of participants’ ownership judgments or how those judgments vary across situations. Developing socially capable AI therefore requires moving beyond overall similarity to capture the diversity and context dependence of human judgments.
[AI-62] Beyond the Shadows of Platos Cave: Evaluating False Memory in Autonomous Agents via Counterfactual Reasoning
链接: https://arxiv.org/abs/2609.39473
作者: Quan M. Tran,Zhuo Huang,Zhen Fang,Jing Zhang,Mingming Gong,Tongliang Liu
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Autonomous agents increasingly rely on memory to generalize beyond their training environments. However, agents are bounded by what they have seen and believed, and leveraging such memories in unseen environments can introduce biases into their internal beliefs. We formalize this phenomenon as \textitfalse memory, which can arise from spurious correlations, environment shifts, and knowledge conflicts. Despite its importance, false memory is difficult to evaluate because it stems from agent internal beliefs and is easily confounded with ordinary generalization failures. Therefore, we propose FAME, a training-free framework that evaluates false memory through the evolution of agent beliefs under counterfactual reasoning. Specifically, counterfactual scenarios reveal how beliefs change as the latent concept of memory shifts under hypothetical interventions; thus, measuring the resulting concept drift provides a signal for distinguishing faithful versus false memory. Such concepts can be estimated from agent hidden states before answer generation, avoiding the need for reward design or answer sampling. Empirical experiments reveal that simply monitoring answers often fails to detect false memory, while FAME achieves AUROCs of 76.2% - 96.7% across false-memory settings, and outperforms the best baseline by 3.4% - 23.3% across realistic benchmarks, spanning math reasoning (GSM-Symbolic), code generation (GitChameleon), and complex reasoning (BigBench-Hard). We further release corresponding counterfactual templates and facilitate future research on false memory.
[AI-63] ActionGuard: Tool Call Authorization under Poisoned Skills
链接: https://arxiv.org/abs/2609.39450
作者: Jihun Han,Yejin Jang,Byung Il Kwak,Mee Lan Han
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:LLM-based agents extend their capabilities through third-party skills that provide task-specific instructions, scripts, and tool-use procedures. However, malicious instructions inserted into an otherwise benign skill can cause a benign user request to trigger dangerous Tool Calls, including data exfiltration, file deletion, or unauthorized code execution. This paper presents ActionGuard, which inspects skill-influenced Tool Calls immediately before execution. ActionGuard separates the target agent’s action-generation context from the safeguard’s authorization context. The target agent may use the original skill for planning, but the Reviewer does not receive the potentially poisoned raw skill text. Instead, it determines whether each action is justified by the trusted user request using a balanced skill profile, current and recent Tool Calls, and local script contents. ActionGuard intercepts each Tool Call at OpenClaw’s before-tool-call stage and enforces the Reviewer’s ALLOW or DENY decision under a fail-closed policy. We evaluate ActionGuard on 139 contextual and 180 obvious injections in a SKILL-INJECT-based setting against Dynamic Guardian and SkillGuard, using three open-source and two commercial Reviewer models. Each condition is repeated three times and evaluated using Attack Success Rate (ASR) and Task Success Rate (TSR). Overall, ActionGuard reduced ASR by 35.54 to 46.11 percent relative to existing safeguards and by 70.44 percent relative to No Safeguard, while maintaining high benign-task completion. These results show that execution-boundary authorization grounded in trusted user intent and runtime evidence can restrict unauthorized Tool Calls induced by skill injection. Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.39450 [cs.CR] (or arXiv:2609.39450v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2609.39450 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-64] From Imitation to Reward Discovery: On-Policy Warmup for Agent ic RL
链接: https://arxiv.org/abs/2609.39436
作者: Yitong Qiao,Tiantian He,Lei Liu,Yue Shen,Jian Wang,Jinjie Gu,Zhixuan Chu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning with a verifiable reward (RLVR) offers a scalable approach to training language-model agents, yet sparse outcome rewards can leave early training with little signal for policy improvement. We identify an On-Policy Acceleration Phenomenon: in our main comparisons, RLVR initialized with on-policy distillation reaches high performance earlier in training and achieves both higher average performance during subsequent RLVR and higher final performance than the alternative baselines. Motivated by this observation, we study On-Policy Warmup (OPW), a teacher-guided stage in which the student trains with teacher supervision on its own interaction trajectories before transitioning to RLVR. Unlike imitation on fixed teacher-generated trajectories, OPW targets states induced by the student’s own decisions, including imperfect actions and recovery situations. We provide a theoretical explanation by connecting on-policy reverse-KL distillation to trajectory-level distribution matching. Under a competent teacher and sufficiently small population distillation loss, this connection yields a lower bound on initial verifier success and a corresponding bound on reward-discovery complexity. For group-relative RLVR, we further characterize when increased success probability produces more reward-informative groups. Together, our findings support on-policy distillation as an effective warmup for agentic RLVR and identify initial reward discovery as a mechanism that can contribute to the observed acceleration.
[AI-65] Inferring Causal Relations between Two Sequences of Events with Language Models
链接: https://arxiv.org/abs/2609.39406
作者: Nishchal Prasad,Eric Gaussier,Emilie Devijver,Alexander Obeid Guzman,Armen Aghasaryan,Gregor Gössler
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Causal AI is a branch of Artificial Intelligence which helps understand and reason about cause and effect relationships, not just patterns or correlations. Causal discovery aims to infer elements of the underlying causal structure–often represented as a directed graph–from observational and, when available, interventional data. While causal discovery is the fundamental step for moving beyond mere associations toward genuine understanding, and thus the basic building block of causal AI, it becomes intrinsically difficult when causal relations must be inferred from single observations. In such situations, standard causal discovery methods cannot be used and one has to identify causal relations from limited amount of information. This is typically the case for, e.g., sequences of events produced by different alarms which need to be analyzed on the fly to detect abnormal phenomena, which are usually rare. We show in this study that it is possible to leverage the predictive power of Large Language Models (LLMs) to infer causal relations between only two sequences of events. This approach, which is validated on both synthetic and real data, provides better results than standard causal discovery algorithms on several time series data, even though these data were converted into smaller, single observed sequences.
[AI-66] Advancing Entropy-Level Credit Assignment in RLVR via Proximal Entropy Policy Optimization NEURIPS2026
链接: https://arxiv.org/abs/2609.39402
作者: Yun Kim,Nojun Kwak
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 21 pages, 4 figures. Accepted at NeurIPS 2026
Abstract:Value-model-free RLVR methods such as GRPO assign uniform advantages to all tokens in a rollout, ignoring that tokens contribute unequally. Recent methods use token entropy as an importance proxy but compute it globally across the batch, conflating importance with prompt difficulty and positional trends. We argue that importance should instead be measured relative to the local context of each token. We introduce proximal entropy, a local measure of token importance relative to neighboring tokens, and prove it is invariant to both confounders. Proximal Entropy Policy Optimization (PEPO) uses it to weight per-token advantages and outperforms GRPO and entropy-based baselines on mathematical reasoning across Qwen3-1.7B, Qwen3-4B, and Llama-3.2-3B-Instruct. We also show the formulation generalizes to other algorithms where substituting proximal entropy into existing methods improves, and applying it to single-stream RL succeeds where global entropy fails.
[AI-67] Experimental Experience Modeling for Autonomous Research
链接: https://arxiv.org/abs/2609.39392
作者: Wenda Wei,Yingchen Zhang,Ruqing Zhang,Jiafeng Guo,Daiting Shi,Xueqi Cheng
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Autonomous research agents can generate hypotheses and conduct experiments, but experimentation remains a major source of computational cost. A fundamental challenge is deciding which experiments are worth running, particularly when prior evidence is insufficient to resolve uncertainty. Yet current research agents lack a systematic way to leverage experimental experience when making such decisions. We introduce Experimental Experience Modeling (EEM), a framework for making informed experimental decisions by acquiring, reusing, and accumulating experimental experience. EEM extracts decision-relevant records from earlier experimental trajectories, distills them into reusable experience, and organizes them in an experience library. For a new experimental decision, EEM retrieves relevant historical experience and assesses whether it provides sufficient support for deciding whether a candidate direction warrants further investment. When historical experience is insufficient, EEM conducts a targeted, low-cost pilot experiment to acquire the missing decision-relevant experience on demand. It then combines this newly acquired experience with retrieved historical experience to determine whether the direction warrants full-scale evaluation, which requires substantial resources. The resulting experimental outcomes are further distilled into reusable experience, allowing the library to continually grow through iterative accumulation. Experiments on autonomous research benchmarks show that EEM improves research performance while reducing model interaction overhead, demonstrating the value of reusing accumulated experience and acquiring additional experience only when needed.
[AI-68] From Search to Signal: Online Post-Training in Automatic Heuristic Design
链接: https://arxiv.org/abs/2609.39383
作者: Yilun Yuan,Tianyu Zhou,Zhenzhou Tang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 18 pages, including supplementary material. Preprint
Abstract:Large language model (LLM)-based automatic heuristic design (AHD) iteratively proposes and refines heuristics, pairing design rationales with executable code. Task-specific evaluators assess programs; execution outcomes and performance scores guide search. Many AHD systems keep the generator frozen; EvoTune and Co-Evolution of Algorithms and Language Model (CALM) instead update it from evaluated candidates. When such outcomes drive reinforcement learning with verifiable rewards (RLVR), they create a search-coupled loop: the evaluated candidate stream supplies both search-state updates and training signals for the model that generates future candidates. Yet validity and performance do not uniquely determine useful model updates; converting them into learning signals must account for the prompt and evolving search state that produced each candidate. We formulate online post-training of small open-weight LLMs in AHD as context-dependent signal construction and develop alternative mappings from program validity, task performance, and generation context to update signals. Using shared evaluated rollouts and matched update budgets, controlled experiments across AHD tasks and model families compare these mappings with online post-training baselines, testing their effects on validity, performance among valid proposals, and the yield of valid proposals that improve under contextual comparisons. Complementary checkpoint, frozen-search, and live-system evaluations assess whether proposal-level gains appear in updated checkpoint behavior and subsequent search, rather than arising solely from accumulated search state. A resource-matched comparison under pre-specified cost accounting tests whether online updating adds value beyond additional search with a frozen generator. Together, this design avoids treating end-to-end search gains alone as evidence of stronger heuristic-design capabilities.
[AI-69] SkillFM: Generating Skills for LLM Agents via Latent Flow Matching
链接: https://arxiv.org/abs/2609.39382
作者: Zuming Zhang,Jie He,Yizhe Zhang,Jeff Z. Pan
类目: Artificial Intelligence (cs.AI)
备注: 33 pages, 8 figures
Abstract:Textual skills provide reusable guidance for large language model agents, but existing approaches often rely on manually curated skill banks or reinforcement learning with indirect and delayed feedback. We introduce SkillFM (Skill Flow Matching), a generative framework that synthesizes task-conditioned textual skills directly without test-time skill retrieval. Our framework combines a codec for encoding and reconstructing textual skills in a continuous latent space with a conditional flow model trained using improved MeanFlow. At inference time, the learned velocity field enables single-step latent sampling, and an LLM-based decoder converts the sampled representation into textual guidance for a frozen downstream agent. We evaluate the framework on embodied tasks, question answering, and web shopping. On ALFWorld and Search-QA, our method achieves the best overall performance among the compared vector-based skill approaches. Our analyses further demonstrate that latent skill generation is an effective alternative to retrieval-based skill augmentation. Our code and training skill libraries are available at this https URL.
[AI-70] Wavelet Flow Matching for Time Series
链接: https://arxiv.org/abs/2609.39374
作者: Lucas Poinsignon,Jorge da Silva Gonçalves,Samuel Ruipérez-Campillo,Julia E. Vogt
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 45 pages, including appendix; 11 figures, 13 tables
Abstract:Synthetic time series are increasingly used for data augmentation, privacy-preserving data sharing, and downstream model development, yet faithfully reproducing both multi-scale temporal structure and cross-channel dependencies remains challenging. We study multivariate time-series generation through flow matching in the wavelet domain. By operating on multilevel discrete wavelet coefficients rather than directly in the time domain, the model represents coarse structure and progressively finer details at separate scales. Their naturally different variances further induce an implicit coarse-to-fine generative process without requiring an explicit multi-scale schedule. Since the transform acts independently on each channel, we pair it with a channel-token transformer whose attention directly models cross-channel dependencies. Across seven benchmark datasets and four sequence lengths, our method is best or tied on a majority of dataset-metric combinations, with the largest and most consistent improvements in Context-FID and discriminative score.
[AI-71] EHR-RobustGym: Benchmarking and Training Agents for Robust Clinical Reasoning
链接: https://arxiv.org/abs/2609.39371
作者: Yitong Qiao,Yancheng Jin,Lei Liu,Yue Shen,Jian Wang,Jinjie Gu,Zhixuan Chu
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:In hospital workflows, electronic health records (EHRs) are often noisy, and may not contain the evidence needed to confirm events or measurements referenced in a clinical query. Even when database retrieval succeeds, clinical agents can overlook such discrepancies and return plausible but unsupported answers. We introduce EHR-RobustGym, a scalable and interactive environment for evaluating and training robust clinical agents grounded in noisy EHRs. Built on MIMIC-IV hospital records (365K patients, 31 tables, and over 500M records), EHR-RobustGym comprises 5,486 Clean-Noise pairs spanning six clinical intents and both patient-level and population-level queries. The pairs test robustness to Record-level, Value-level, and Query-level noise, while interactive SQL/Python execution and outcome verification support trajectory collection and training. Evaluating multiple LLMs reveals substantial robustness gaps: average task success across proprietary and large-scale open-weight models drops from 62.2% on Clean questions to 37.9% on Noise questions. At k=4, pass^k consistency falls below 50% for most evaluated models, exposing instability in clinical task completion. Supervised fine-tuning and reinforcement learning in EHR-RobustGym improve performance, with gains generalizing to five external EHR benchmarks. Together, these results position EHR-RobustGym as a testbed for evaluating and improving the evidence-grounded robustness of clinical agents.
[AI-72] Autoresearch in Mixed-Integer Linear and Nonlinear Programming
链接: https://arxiv.org/abs/2609.39360
作者: Yuwei Gu,Yaoxin Wu,Tong Guo,Wen Song,Zhiguang Cao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Despite recent progress in autoresearch, applying it to practical operations research problems, typically formulated as NP-hard mixed-integer linear or nonlinear programs (MILPs or MINLPs), remains challenging because effective research requires systematically managing competing ideas and long-horizon experimental trajectories. We introduce AutoMIP, a reusable agent skill for organizing long-horizon autoresearch in mixed-integer programming through idea pooling and algorithm tree search. AutoMIP maintains a persistent pool of complementary candidate ideas while organizing executable experiments into an algorithm tree, enabling the agent to preserve unexplored hypotheses, refine promising algorithms, and switch to alternative methodological directions based on historical states. On MILP and MINLP benchmark cohorts, AutoMIP achieves the highest final success rates among the evaluated autoresearch frameworks. On MIPLib, AutoMIP discovers new best solutions for 31 of 60 instances, surpassing existing autoresearch frameworks. On MINLPLib, it achieves new best solutions for 52 of 60 instances. Ablation studies further demonstrate the complementary contributions of idea pooling and algorithm tree search, highlighting the importance of jointly maintaining diverse research ideas and structured experimental trajectories for long-horizon autoresearch.
[AI-73] Hiding in Plain Sight: Decoupling Pretext from Actuation for Skill Poisoning in LLM Agents
链接: https://arxiv.org/abs/2609.39352
作者: Wenxin Wu,Lingyong Yan,Lei Sha,Shuaiqiang Wang,Jiashu Zhao
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:LLM agents increasingly rely on reusable Skills for complex, multi-step tasks, creating a critical supply-chain attack surface where poisoned Skill content steers agent decision loops under benign requests. Existing skill poisoning attacks either colocate actuation with its contextual pretext or distribute actuation across multiple Skills, but do not explicitly separate the rationale for execution from the operation itself. In this work, we reveal that untrusted agent decisions fundamentally depend on two conceptually distinct Risk-Realization Factors (RRFs): an actuation factor (specifying what concrete operation is performed) and a pretext factor (providing the situational rationale for why the agent must perform it). Guided by this abstraction, we propose a coordination-based attack paradigm: decoupling pretext from actuation. Rather than fragmenting the malicious actuation, we preserve it as an intact operation within a downstream Steering Skill, while delegating the pretext factor to an upstream Grounding Skill that subtly alters persistent environment artifacts through routine utility operations. The intact actuation thus hides in plain sight, appearing completely legitimate and task-driven only when evaluated against the fabricated pretext. Building on this formulation, we develop an automated framework that discovers authentic execution dependencies, synthesizes coordinated pretext-actuation skill pairs, and iteratively refines poisoned skill instructions via runtime closed-loop feedback. Extensive evaluations across single-session and persistent cross-lifecycle scenarios demonstrate that decoupled skill poisoning achieves high attack success, exposing a critical blind spot in isolated Skill security audits. Our automated framework code is available at this https URL.
[AI-74] On the Complexity of Preference-Based Bandits
链接: https://arxiv.org/abs/2609.39351
作者: Ahmed Ben Yahmed(CREST, ENSAE Paris, FAIRPLAY),Marc Abeille(FAIRPLAY),Clément Calauzènes(FAIRPLAY)
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:We study preference-based bandits with general reward function classes, where a learner sequentially selects pairs of arms and observes binary preference feedback governed by the Bradley–Terry model. This setting naturally arises in applications such as recommender systems, tournament ranking, and learning from human feedback, where relative preferences are easier to elicit than absolute rewards. The observation model inherits the logistic bandit challenge of handling the problem-dependent constant \kappa , which accounts for the non-linearity of the link function and can grow arbitrarily large. Moreover, prior work has predominantly focused on linear or kernelized reward models, precluding the use of richer function classes. To address these limitations, we consider general reward function classes and introduce the \emphlocally sensitive eluder dimension, a novel complexity measure tailored to the logistic structure of preference feedback that yields fine-grained regret guarantees without unfavorable dependence on \kappa . Building on this notion, we propose \textbfGINOP (Generic INformative OPtimism), an algorithm that constructs log-loss confidence sets and jointly selects arm pairs to balance optimism and informative exploration. We establish a first-order regret bound that, in contrast with what previous results suggest, demonstrates that learning with preference feedback is as statistically efficient as learning from direct reward observation. Finally, we corroborate our theoretical findings with empirical evaluations against competitive baselines.
[AI-75] Who Said What and Will It Be Remembered? Evaluating Persistent Speaker Attribution Across Meetings
链接: https://arxiv.org/abs/2609.39344
作者: Shantanu Vispute,Aditya Mishra,Siddhartha Saxena
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注: A short version is accepted at IEEE SLT 2026, Demo Track
Abstract:Speech transcripts used as long-term memory must preserve both words and stable speaker identities. Existing meeting-transcription metrics either ignore speakers or remap anonymous speakers independently in each recording, so they cannot measure whether the same person retains one identity across meetings. We evaluate persistent speaker attribution with Speaker Identified cpWER (SI-cpWER), which scores a corpus under one global speaker-ID assignment. The benchmark covers five commercial diarize-then-identify cascades, two open academic baselines, and ThyVoice on the full 129-meeting CHiME-8 NOTSOFAR evaluation set in clean and noiseaugmented form, plus CHiME-6. ThyVoice is our end-to-end reference system; it repairs overlap and gates the evidence used to create and update voiceprints. Requiring persistent identity changes the commercial ranking: ThyVoice records lower SI-cpWER than every evaluated commercial cascade in all three conditions and the lowest mean in the full panel, 47.13 versus 54.75 for the next system. Complementary lexical, diarization, per-recording attribution, and speaker-clustering diagnostics characterize upstream error surfaces in the final attributed record. These results show why persistent attribution must be evaluated directly in systems that reuse conversations across time.
[AI-76] he Golden Path Hypothesis: Reusable Schedules in Diffusion Caching
链接: https://arxiv.org/abs/2609.39343
作者: Dong Wang,Wenwu Tang,Francesco Corti,Yun Cheng,Lothar Thiele,Olga Saukh
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Diffusion caching accelerates generation by replacing transformer computation with cached or predicted features at selected denoising steps. We introduce the Golden Path Hypothesis (GPH): under fixed inference conditions, prompt-independent cache schedules can achieve final-output quality comparable to the best prompt-specific schedules across prompts. We investigate the GPH across ten caching methods, four image and video models, and three cache ratios. Prompt-adaptive methods repeatedly select a small number of schedules, and reusing their most frequent schedules on new prompts closely matches the quality of prompt-specific choices. Exhaustive evaluation of 1.4 million schedules on four examples further identifies prompt-independent schedules that remain competitive on unseen prompts. To explain this transfer, we analyze denoising trajectories and the accumulation of caching errors. Latent-state trajectories exhibit similar structures across datasets and seeds, while an exact error decomposition shows that accumulated effects of earlier errors predict final latent-state error better than local approximation errors. This motivates searching for end-to-end schedules using final-output quality. With only a small set of examples, the resulting golden paths transfer across prompts and datasets, and can be tuned to the desired quality objective, including reconstruction fidelity or perceptual similarity.
[AI-77] WinoTS: Wavelet-based Self-Distillation for Time Series Models
链接: https://arxiv.org/abs/2609.39337
作者: Noam Major,Kathy Razmadze,Yoli Shavit
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Self-supervised pre-training of time series models is currently dominated by next-token prediction and reconstruction objectives. In continuous-valued domains, these paradigms often waste model capacity on high-frequency, point-wise noise at the expense of learning invariant structure. While invariance-based self-distillation has proven highly effective in computer vision, its application to temporal data remains largely underexplored. Effectively adapting such methods to time series requires carefully designed augmentations: spatial operations like cropping can shift the timing of repeating cycles or distort the signal, while basic jittering may provide limited variation. We introduce Wavelet-based self-distillation for time series (WinoTS), an invariance-based pre-training paradigm designed specifically for temporal signals. At its core, WinoTS leverages time-frequency augmentations to construct multi-scale structural views without distorting underlying signal dynamics. Across extensive evaluations, WinoTS outperforms state-of-the-art baselines in long-term forecasting, cross-domain zero-shot transfer, and unsupervised anomaly detection. Notably, linear probing on frozen WinoTS representations frequently surpasses fully supervised models trained from scratch. Systematic ablations demonstrate that WinoTS is a flexible, architecture-agnostic framework yielding gains across time series backbones, and establish that time-frequency transformations provide a principled alternative to vision-style spatial augmentations.
[AI-78] WorkGenesis: Building the Worlds That Teach Agents to Work
链接: https://arxiv.org/abs/2609.39325
作者: Xinyu Zhu,Fenyi Liu,Yuzhu Cai,Shuo Tang,Rui Ye,Linfeng Zhang,Siheng Chen
类目: Artificial Intelligence (cs.AI)
备注: 47 pages
Abstract:The ability of Large Language Model (LLM) agents to complete daily and professional work is receiving increasing attention. Training such agents requires realistic work scenarios. Expert-authored occupational work is costly and slow to produce, while unconstrained synthesis often yields tasks with weak factual grounding or internally inconsistent requirements. To bridge this gap, we introduce WorkGenesis, a framework that constructs executable occupational work from real-world artifacts through two core technical innovations: (1) Evidence-Based Work Construction, which grounds each unit of work in real-world evidence by retrieving public files guided by O*NET occupational knowledge and synthesizing the surrounding context, companion materials, work request, and itemwise rubric around them; and (2) Execution-Guided Consistency Verification, which renders a reference deliverable inside the constructed work, attributes every unsatisfied rubric item to the agent, the task, or the rubric, and uses task and rubric defects as feedback to iteratively repair the work until it passes the audit. Experimental results demonstrate that Fx-Work-35B, trained with simple supervised fine-tuning (SFT) on only 20K units of work synthesized by WorkGenesis, achieves the highest scores among all comparable-scale baselines on the five reported metrics across GDPvalAA-v2, APEX-Agents-AA, and JobBench (31.00 versus 24.79 average score), and even surpasses frontier models such as the 1.6T DeepSeek-V4-Pro-Preview. These results show that WorkGenesis provides scalable training data for working agents.
[AI-79] HiWE: Hierarchical World Knowledge Model with Visual Keypoint Enhancement for Zero-Shot 3D Path Planning
链接: https://arxiv.org/abs/2609.39323
作者: Guoqing Ma,Mingqi Yuan,Chen Gao,Jiayu Chen,Shan Yu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Robot demonstration generation requires a system to identify where an interaction should occur, plan a feasible motion, and execute the required contact. HiWE connects these decisions through a point-based interface between visual grounding and language-based planning. PointVLM is instruction-tuned to associate task-relevant objects with image coordinates using a mixture of point annotations, segmentation-derived samples, robot observations, and visual question answering data. Depth measurements lift these predictions into a semantic 3D representation. A language planner, 3DLLM, uses this representation to specify end-effector waypoints and gripper commands, while a hybrid grasping module resolves local grasp poses. The evaluation covers 14 simulated manipulation tasks and four physical-robot tasks, together with ablations of the visual training data, spatial inputs, and grasp selection. Here, zero-shot execution refers to deployment without task-specific demonstration training; the visual model uses existing robot data during fine-tuning. This paper describes the original point-based formulation of the framework; its relationship to the subsequent GeneralVLA extension is detailed in the introduction.
[AI-80] GRPO Training Dynamics for Small Language Models
链接: https://arxiv.org/abs/2609.39321
作者: Rajat Ghosh,Vaishnavi Bhargava,Henry Wong,Aryan Singhal,Debojyoti Dutta
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Group Relative Policy Optimization (GRPO) has emerged as a memory-efficient reinforcement fine-tuning (RFT) technique for reasoning-intensive tasks. How- ever, GRPO training dynamics on small language models (SLMs) remain poorly understood, limiting its reliable adoption and reproducibility in open and resource- constrained environments. In this work, we present a systematic study of GRPO fine-tuning for SLMs ranging from 1.5B to 7B parameters under a practical single- node 8xA100 compute budget. Our study spans multiple model families and reasoning domains, including mathematics, coding, and multiple-choice question answering (MCQ) in science. Across these settings, we analyze how group size affects policy convergence, training stability, and downstream benchmark per- formance. We further characterize tensor-level update dynamics during GRPO training and investigate whether the choice of LoRA target modules and layers can improve the performance of GRPO-tuned models. While our initial GRPO-tuned models outperform their base counterparts on approximately 80% of mathematical benchmark evaluations, they demonstrate limited capability on MCQ and code reasoning tasks. Guided by our mechanistic evaluations, we refined our LoRA and reward-shaping configurations to improve performance in latter domains. These findings provide practical guidance for GRPO training for SLMs.
[AI-81] ReSAIL: Mitigating Collapse in Iterative Agent Self-Distillation
链接: https://arxiv.org/abs/2609.39306
作者: Shengjie Jin,Hengbo Xu,Zelong Sun,YuJie Guo,Zhiwu Lu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Iterative self-distillation enables LLM agents to learn from successive deployments, offering a path toward recursive self-improvement (RSI). Yet our experiments with existing methods reveal a collapse in deployment performance across cycles, while task performance with privileged information (PI) also declines. We address this collapse by prioritizing informative interaction steps for distillation and preserving PI-conditioned behavior as the student becomes the next teacher. We introduce Retentive and Selective Augmentation for Iterative Self-Distillation (ReSAIL), a plug-in augmentation for iterative PI-based self-distillation. ReSAIL selects interaction steps where PI most strongly changes the teacher’s predictions and balances the resulting distillation losses across trajectories. It also regularizes the student’s PI-conditioned output distributions toward those of the frozen teacher at selected and unselected steps to preserve PI-conditioned behavior for supervision in the next cycle. On ALFWorld and TextCraft, ReSAIL sustains substantial gains across model scales over three cycles, with an average absolute gain of 22.5% in final-cycle success rates when added to self-distillation baselines. Sensitivity-guided selection of offline data also improves action prediction accuracy for multimodal GUI agents on AITZ. These findings provide the first evidence that a more robust learning mechanism can effectively mitigate performance collapse in iterative agent self-distillation over deployment trajectories.
[AI-82] Scale and Selection: What Makes Automatic Harness Evolution Work for Visual-Interface Robot Agents
链接: https://arxiv.org/abs/2609.39304
作者: Zhijie Wei,Ferris Tan,Jinghui Wang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 12 pages, 4 figures
Abstract:When an off-the-shelf coding agent is used directly as a robot policy, observing a browser-based 3D interface through screenshots and acting by posing a virtual target gripper through a few tools, the agent’s harness, its prompts, tools, and control rules, largely determines success, and until now it has been written by hand. We show that this harness can be improved automatically by another coding agent, the optimizer agent, and report two findings about what makes it work. First, the number of rollouts the optimizer agent sees per round governs whether the evolved harness is trustworthy, generalizes, and improves steadily. A single rollout is a noisy binary outcome, so with few rollouts per round a revision can be promoted on luck; enlarging the batch raises the signal-to-noise ratio of every promotion decision. Holding rounds fixed and growing the training set from 5 to 100 rollouts, held-out success rises from 47% to 67%, while small training sets overfit, reaching 70% on training tasks but only 54% held-out. Second, the optimizer agent must not be given free rein. With every revision it proposes accepted unconditionally, performance drifts downward within ten rounds as ill-judged edits accumulate; adding the most basic safeguard, Champion-Challenger selection that promotes a revision only if it strictly beats the incumbent on the same fixed evaluation set, turns the same loop into one that raises held-out success from 51% to 67% over 30 rounds. Automatic harness evolution for visual-interface robot agents is thus feasible, but its gains hinge on the rollout scale behind each decision and on how the optimizer agent’s revisions are selected.
[AI-83] MiniRep: Robust Reputation-Based Aggregation for Multi-Agent Debate
链接: https://arxiv.org/abs/2609.39297
作者: Jiaming Zhang,Yuwan Liu,Yue Huang,Sisi Duan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Autonomous agents powered by large language models (LLMs) are rapidly evolving into an open agentic ecosystem. To support trustworthy collaboration, industry initiatives increasingly assess agent reputation from past behavior and provide performance leaderboards. However, reputation derived from past performance may not reliably predict an agent’s behavior on new tasks, particularly when malicious agents can adapt their behavior and influence other agents during collaboration. We study reputation in multi-agent debate (MAD), where multiple agents answer the same query, debate to improve their answers, and aggregate them into a final output. We present MiniRep, a reputation-based aggregation system for MAD under malicious agents. To ground our threat model in established research, we construct an attack taxonomy drawing on reputation-system attacks and software-testing mutation operators, covering strategic exploitation of reputation and subtle corruption of agent proposals. Guided by this taxonomy, MiniRep evaluates agents based on both their behavior on the current task and their reputation over time, while preventing groups of agents with highly similar responses from dominating the final decision. We assess MiniRep across diverse tasks, LLM-agent compositions, corruption placements, and attack types drawn from our taxonomy. Our experimental results show that, MiniRep outperforms both conventional MAD aggregation and conventional reputation-based approaches on MATH no matter being attacked or not. Also, under a heterogeneous 10-agent setting on MATH, MiniRep outperforms all baselines in all 28 attack conditions. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.39297 [cs.AI] (or arXiv:2609.39297v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.39297 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-84] ANI: Adaptive Numerical Injection for Unifying Semantic and Arithmetic Representations in Numerical Reasoning EMNLP EMNLP2026
链接: https://arxiv.org/abs/2609.39294
作者: Jinsung Jeon,Seung-won Hwang
类目: Artificial Intelligence (cs.AI)
备注: Accepted to EMNLP 2026. 16 pages, 7 figures. Code available at this https URL
Abstract:Precise numerical reasoning with Large Language Models (LLMs) is essential for expanding their applicability to complex real-world tasks. However, text-based tokenization often fragments numbers, significantly hindering precise arithmetic reasoning. Meanwhile, numerical embeddings, despite arithmetic precision, rely on context-agnostic substitution that disregards the semantic role of numbers as identifiers. To combine the complementary strengths, we propose \textbfANI (Adaptive Numerical Injection), a hybrid framework that governs the selective injection of numerical features based on the semantic context. By employing a context-aware gating mechanism, we selectively inject numerical embeddings (specifically FoNE) into the latent space, explicitly preserving nominal identifiers while enhancing quantitative operands. Through extensive evaluations across various LLMs, we demonstrate that ANI enhances MATH performance by 9.5 points over the official reference model, while maintaining robust performance on general linguistic benchmarks.
[AI-85] EngramBench: A Capability-Grounded Benchmark for Skill-Evolution Harnesses
链接: https://arxiv.org/abs/2609.39284
作者: Zhixuan Tan,Pengjie Gu,Zhao Li,Yihan Hu,Xu He,Dong Li,Jianye Hao
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:While large language models have achieved remarkable success in isolated code generation, authentic software engineering requires sustained reasoning, complex state management, and continuous cross-domain abstraction. However, current evaluations of skill evolution in autonomous agents suffer from a critical identifiability problem: they structurally confound genuine capability abstraction with rote solution leakage (i.e., copying highly similar code from historical training data). To resolve this, we introduce EngramBench, a rigorous, capability-grounded benchmark governed by the strict axiom of capability overlap without solution overlap. Comprising 30 diverse learning tasks and 13 unseen transfer tasks, EngramBench challenges agents to navigate interactive, multi-hour development cycles driven by LLM-simulated users. Our extensive evaluation across 48 multi-hour execution trajectories – corroborated by human-expert validation – reveals a profound insight into procedural memory. We demonstrate that static skill banks do not magically bypass the “last mile” of exact code implementation, which remains bottlenecked by the base model’s inherent reasoning limits. However, they serve as an indispensable execution compass. By navigating agents away from catastrophic, token-heavy trial-and-error, genuine capability abstraction slashes redundant context bloat and reduces overall coding time by over 55%. Ultimately, EngramBench shifts the evaluation paradigm from trivial pattern matching to the verifiable measurement of deep, cross-domain capability transfer.
[AI-86] Faithful Dual-constrained Erasure for Robust LLM Safety Alignment
链接: https://arxiv.org/abs/2609.39279
作者: Jiaqing Li,Shide Zhou,Zhibo Zhang,Yuxi Li,Tianlong Yu,Kailong Wang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Machine unlearning has emerged as a crucial mechanism for removing hazardous knowledge and enforcing safety alignment in Large Language Models (LLMs). However, recent studies reveal a persistent security risk: unlearned models remain highly vulnerable to retraining attacks, where suppressed malicious behaviors rapidly resurface after benign fine-tuning. In this work, we investigate the optimization dynamics of unlearning and identify that this vulnerability stems from shallow alignment. Rather than effectively erasing target knowledge, models often exploit a shortcut by activating previously dormant parameters to act as spurious suppressors, forming a fragile inhibitory shell over intact malicious representations. To address this issue and enforce authentic memory deletion, we propose FDCU, a novel dual-constrained subspace projection framework. FDCU restricts parameter updates through a highly scalable, element-wise dual-masking rule: it preserves general knowledge manifolds via Fisher Information and strictly prohibits the abnormal activation of spurious suppressors via the Principle of Minimal Functional Intervention (PMFI). By reliably blocking the model’s ability to superficially hide knowledge, FDCU promotes the authentic dismantling of target representations. Extensive experiments across specific knowledge erasure and safe output control tasks demonstrate that FDCU achieves state-of-the-art robustness against retraining attacks while maintaining near-lossless general utility, ensuring durable safety for LLMs.
[AI-87] A Time-Aware Bag-of-Receptive-Fields for Interpretable Irregular Time Series Classification
链接: https://arxiv.org/abs/2609.39268
作者: Francesco Spinnato
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Irregular time series, characterized by non-uniform sampling intervals, missing observations, and variable lengths, are ubiquitous in healthcare, mobility, and environmental monitoring, yet effective and interpretable classifiers for this setting are limited. Existing approaches often rely on imputation, which can obscure the temporal structure of the data, or require complex neural architectures that are opaque and difficult to explain. In this work, we extend the Bag-Of-Receptive-Fields (BORF), a fast, deterministic, and interpretable transform for time series, to the irregular setting. Our key contribution is a time-weighted normalization scheme in which each observation is weighted proportionally to its associated time delta, making pattern extraction sensitive to the actual temporal distribution of samples rather than only their index position. This requires deriving an efficient sliding-window recurrence for the time-weighted standard deviation, preserving the linear time complexity of BORF. We benchmark the resulting method against state-of-the-art irregular time series classifiers on datasets from the PYRREGULAR repository, demonstrating competitive classification performance with the added benefit of human-interpretable explanations.
[AI-88] Effective Does Not Mean Useful: Conditional Functional Substitutability for Redundancy and Scaling in Transformers
链接: https://arxiv.org/abs/2609.39259
作者: Jiaheng Chen,Jiaxing Li,Yucheng Xiao,Xinyong Cai,Juncheng Bu,Lan Yu,Tinghe Zhang
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 21 pages, 4 figures, 7 tables
Abstract:Modern neural networks scale predictably, yet the mechanisms behind these regularities remain unclear. Neural redundancy is typically characterized by component importance or representational similarity, both indirect proxies. We view redundancy as an input-conditioned, dynamic relation: intermediate computational states are functionally redundant when they induce similar downstream responses. We introduce Conditional Functional Substitutability (CFS) to directly characterize such functional substitution. CFS exposes functional relations and reduction potential missed by conventional importance- and similarity-based measures. Across modalities and Transformer families, CFS reveals systematic functional reorganization with scale. Controlled scaling further shows that performance gains need not track growth in substitutability, while fixed-capacity models with more independent functional structure perform better, providing a functional account of diminishing returns. Predicted CFS further enables dynamic computation with a better performance–computation trade-off than importance-based component selection, suggesting new directions for redundancy-aware computation and more efficient model scaling.
[AI-89] From Benchmarks to Production: Transferring Time Series Anomaly Detection Methods for Electricity Production Monitoring
链接: https://arxiv.org/abs/2609.39257
作者: Nicolas Vautier,Paul Caron,Nardi Xhepi,Félicie Bizeul,Manel Boumghar,Christophe Degouy,Paul Boniol
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Databases (cs.DB)
备注:
Abstract:Accurate forecasting of electricity production is essential for maintaining the operational efficiency and strategic planning of energy utilities. In industrial settings, such forecasts are generated daily to ensure supply-demand balance and optimal management of production assets. However, the increasing complexity of modern power systems and data flows poses significant challenges for ensuring the reliability and consistency of these forecasts. This paper addresses the problem of anomaly detection in short-term production forecasts at EDF, formulated as identifying atypical intra-day patterns that may signal data quality issues or operational irregularities. We introduce TAMIS, a scalable and interpretable system that analyzes daily production time series to automatically detect anomalous days based on deviations from historical patterns learned from past data. Designed for human-in-the-loop workflows, TAMIS surfaces top-ranked anomalies through an automated daily newsletter, enabling efficient expert review and continuous monitoring. An extensive experimental evaluation on real-world industrial data demonstrates that TAMIS achieves the best accuracy-efficiency trade-off compared to baseline methods. To foster further research and reproducibility, we publicly release the anonymized application datasets used in our study.
[AI-90] rust the Critic More
链接: https://arxiv.org/abs/2609.39247
作者: Kaiyue Wen,Luke Bailey,Arvind Mahankali,Tengyu Ma
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Standard language model RL algorithms credit every token of a long rollout with the same advantage determined by the terminal reward. Actor-critic methods can provide finer-grained credit assignment, but learned critics are generally considered too inaccurate to trust when training LLMs with RL. In recent works, even when a critic is present, it is used only for baseline estimation, so every trajectory must be rolled out to its terminal reward. We introduce Actor-Critic with Action Chunking (AC2) that removes the need to roll every trajectory to completion. AC2 instead assigns credit to action chunks: short continuations of prefixes of past trajectories. A learned critic scores the state reached at the end of each action chunk, allowing the policy to update without observing a terminal reward. We make critic-based credit assignment reliable through three design choices. First, we introduce local readiness which uses critic-based updates on a problem only when the critic is sufficiently accurate on that particular problem. Second, when available, we provide the critic with a reference solution from a previous successful rollout. Third, we assign credit over action chunks of 10k tokens rather than individual tokens, giving the critic a more meaningful portion of the trajectory to evaluate. We train Qwen3-4B on FineProofs-RL using AC2 and evaluate on IMO-ProofBench. AC2 exceeds GRPO’s peak validation score of 18.5% using 2.5x fewer decoding FLOPs. This gain comes from two sources, (1) AC2 requires 25% fewer training steps to reach this score, and (2) each step generates fewer tokens because the policy does not need to continue every trajectory to completion. Conceptually, we demonstrate that we can remove the need to roll out every trajectory to completion, opening up a large previously unexplored design space for LLM RL algorithms.
[AI-91] What Streaming Anomaly Detection Finds (and Misses) in Industrial Time Series
链接: https://arxiv.org/abs/2609.39232
作者: Magali Parrino,Antoine Ajenjo,Emmanuel Remy,Pierre Stephan,Paul Boniol
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:EDF relies on continuous monitoring of its power plants to detect anomalies as soon as they occur. Given the absence of a universally optimal streaming method in unsupervised settings, we compare streaming methods with state-of-the-art TSAD models deployed online on a real nuclear power plant dataset. This work also evaluates Automated Anomaly Detection in a streaming context. Results show higher consistency for online TSAD and strong robustness from ensembling strategies.
[AI-92] Fyan: A Human–AI Harness with Semantic Auditing for Document-Level Formalization
链接: https://arxiv.org/abs/2609.39228
作者: Wei Zhao,Yangshuo Zou,Chengxiang Ding,Yifan Wu,Xuchuan Wang,Zimu Mao,Lei Zhang,Tao Luo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We present FYAN, a human–AI harness for document-level mathematical formalization. Rather than treating theorems in isolation, FYAN coordinates an end-to-end workflow spanning specification, proof planning, logical review, Lean proof construction, knowledge curation, and validation, with support for independent supervision and human guidance. A central component is evidence-grounded semantic auditing, which assesses whether formal statements faithfully preserve their informal specifications. A language model constructs structured evidence over local correspondences, omissions, scope, and logical relations, while a deterministic validator checks this evidence and produces reproducible judgments. When a substantive but admissible deviation is accepted, FYAN requires an explicit proof-transfer obligation connecting the formal statement back to a source-facing interpretation. With the same model (DeepSeek-V4.1-Flash) in every stage, FYAN proves 86 of 143 FormalTCS theorems under a strict Lean check, against 69 for a general agent harness, and raises the natural-language proof score from 0.501 to 0.851. On ConsistencyCheck, its semantic audit catches more inconsistent statements than a direct LLM judge, both on labels verified against the source (recall 0.777 vs. 0.636) and on the original labels (0.873 vs. 0.820), and localizes each mismatch it reports to a specific hypothesis, conclusion, or scope. FYAN also built ODENumLib, a 9,355-line Lean library for the numerical analysis of ordinary differential equation.
[AI-93] Learning Process Rewards via Reasoning State Propagation
链接: https://arxiv.org/abs/2609.39220
作者: Kai Gan,Zi-Hao Zhou,Bo Ye,Jian Zhao,Min-Ling Zhang,Tong Wei
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Process reward models (PRMs) have demonstrated notable effectiveness in test-time scaling and reinforcement learning by providing fine-grained signals for evaluating intermediate reasoning states, but their training relies heavily on costly process annotations. A natural way to alleviate this dependence is to complement limited process supervision with scalable outcome supervision. However, existing PRMs often model reasoning prefixes independently, providing no explicit mechanism for effectively using final outcome to guide the learning of intermediate reasoning states. We introduce Reasoning State Propagation (RSP), which represents each reasoning prefix with a binary validity state and models transitions between successive states across the reasoning trajectory. Specifically, RSP predicts a break probability that a valid state becomes invalid and a repair probability that an invalid state returns to valid. By propagating these transitions, RSP connects intermediate states to the final state, allowing process annotations to supervise intermediate states while outcome labels supervise the final state and can provide learning signals to preceding steps. Across reasoning search, response selection, and reinforcement learning, RSP consistently outperforms representative PRM baselines, with average improvements over Qwen2.5-Math-PRM of 5.6% in beam search and 2.1% in reinforcement learning.
[AI-94] In a Streaming World Should You Stand Still? A Comprehensive Benchmark of Anomaly Detection in Streams
链接: https://arxiv.org/abs/2609.39215
作者: Magali Parrino,Antoine Ajenjo,Emmanuel Remy,Pierre Stephan,Pierre Senellart,Paul Boniol
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Time series anomaly detection (TSAD) is increasingly deployed in streaming settings, where data arrive sequentially and may exhibit non-stationarity. As a result, several works from the recent literature propose streaming anomaly detection methods that rely on incremental updates to adapt over time. However, most of these approaches originate from the streaming outlier detection literature and largely ignore core characteristics of time series anomalies. Moreover, their empirical evaluation is typically conducted on synthetic or small-scale benchmarks with limited diversity, making it unclear whether streaming methods are truly advantageous in realistic TSAD scenarios. In this work, we carry out the first large-scale experimental study comparing streaming and static TSAD methods under a unified streaming evaluation benchmark. We consider a realistic setting in which an initial batch of data is available for model training, followed by online evaluation of both detection accuracy and computational efficiency. In addition, we propose a distribution-drift dataset of real time series, called TSB- drift, to isolate scenarios where streaming updates are theoretically justified. Our results show that, contrary to common assumptions, static TSAD methods significantly outperform streaming approaches in most streaming settings. Such finding highlights a critical gap between the design of existing streaming methods and the requirements of modern TSAD, and calls for a rethinking of how streaming capabilities should be integrated into TSAD.
[AI-95] UniAE-MoE: A Unified Audio Encoder via Mixture of Experts
链接: https://arxiv.org/abs/2609.39199
作者: Shengbo Cai,Zhisheng Zhang,Zichao Nie,Jing Peng,Jingran Xie,Zhiyong Wu
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注:
Abstract:Large Audio Language Models (LALMs) rely on effective audio encoders for multi-task performance. We introduce UniAE-MoE, a unified audio encoder designed to model cross-domain audio representations and achieve outstanding downstream understanding performance via a Mixture-of-Experts (MoE) architecture. Specifically, we explore mainstream audio encoders and integrate those from Qwen2-Audio and Audio-Flamingo 3, which demonstrate superior downstream capabilities. To facilitate effective model fusion, we improve our encoder using SwiGLU with shared experts to decouple encoder networks, and we further introduce a two-stage instruction-tuning strategy to better adapt the model to diverse downstream tasks. Moreover, we propose the task-specific data scaling (TSDS) technique to enhance \tool’s understanding capabilities. On the XARES-LLM benchmark, UniAE-MoE attains a score of 0.802, achieving state-of-the-art performance. It also delivers top-tier performance in the official Interspeech 2026 Audio Encoder Capability Challenge, further demonstrating robust generalization across diverse audio tasks. Together, these results validate the effectiveness of \tool for unified audio understanding across speech, music, and general audio domains.
[AI-96] Reinforcing Multimodal Reasoning via Token-Level Perception-Grounded Advantage Estimation ACM-MM2026
链接: https://arxiv.org/abs/2609.39168
作者: Zhihan Zhang,Lizi Liao
类目: Artificial Intelligence (cs.AI)
备注: Accepted by ACM MM 2026
Abstract:Reinforcement Learning with Verifiable Rewards (RLVR) has improved the reasoning capabilities of Multimodal Large Language Models (MLLMs), yet existing frameworks rely on coarse, sequence-level reward signals that lack the fine-grained supervision over the visually-grounded steps within a multimodal reasoning chain. We investigate this gap through the lens of two token-level metrics: visual dependency (i.e. how much a token’s prediction relies on the input image features) and predictive entropy. Our empirical analysis reveals two key findings: (1) correct reasoning chains exhibit a markedly sharper entropy reduction as visual grounding intensifies, compared to incorrect ones; (2) pivotal tokens, those whose misprediction triggers reasoning collapse, are statistical outliers in the joint distribution of visual dependency and predictive entropy derived from correct chains. Motivated by these findings, we propose token-level perception-grounded advantage estimation (TPAE), which estimates token-level advantages by measuring each token’s statistical consistency with the vision-entropy patterns of correct rollouts. TPAE leverages this granular score to modulate the sequence-level advantage, producing a fine-grained supervision signal that can be integrated into various RLVR frameworks. Extensive experiments on seven benchmarks show that TPAE consistently outperforms leading strong baselines, yielding more stable and efficient optimization for multimodal reasoning. The code is publicly available at this https URL.
[AI-97] Beyond the Remembered World: Predictive 4D Belief for Persistent Navigation in Evolving Worlds
链接: https://arxiv.org/abs/2609.39166
作者: Mingjian Gao,Zhaocheng Li,Haoyang Huang,Wenqiao Zhang,Yingjie Niu,Hao Zhou,Chao Li,Juncheng Li,Siliang Tang,Yueting Zhuang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Persistent spatial memory enables embodied agents to navigate familiar environments across repeated visits. However, targets may move while unobserved, including during navigation, making remembered locations unreliable by the time an agent arrives. Despite advances in memory retrieval and state prediction, accounting for continued hidden world evolution and revising beliefs under limited visibility remain challenging. We study Evolving-World Navigation, where agents infer target locations from intermittent observations, predict their states at inspection time, and revise beliefs using visual evidence. We propose EvolvingNav, which constructs a time-indexed belief from timestamped 3D object histories through a structured persistence-relocation model. The belief distinguishes persistence at the last observed location from relocation to alternative locations and retains probability mass outside the known candidate set. An event-driven filter propagates the current belief as time elapses, forecasts target occupancy at candidate inspection times, and incorporates new RGB-D evidence. Negative observations downweight location hypotheses according to calibrated, visibility-conditioned detection probabilities, while evidence tracking prevents repeated use of the same observations. A frozen, zero-shot vision-language controller uses the updated belief to choose actions and replan. We further introduce EvoWorld-Bench, a benchmark grounded in human activity traces, comprising 54 scenes and 803,680 tasks with controlled changes before and during navigation. In simulation and real-robot experiments, EvolvingNav improves navigation success and search efficiency over the evaluated baselines. Paired experiments show the clearest gains under learnable temporal patterns, while ablations demonstrate the value of preserving uncertainty and incorporating visibility-aware evidence.
[AI-98] Rep2Skill: Representation-Guided Skill Self-Evolution for LLM Agents
链接: https://arxiv.org/abs/2609.39149
作者: Kaixing Zhang,Changming Li,Yingdong Shi,Zheng Zhang,Kaitao Song,Wenjie Shi,Jingang Wang,Kan Ren
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Textual skills enable large language model (LLM) based agents to accumulate reusable procedural knowledge without updating model parameters. Yet existing skill evolution remains largely confined to the text space: an optimizer must diagnose success and failure patterns, and revise skills solely from long execution trajectories and sparse task outcomes. This text-only paradigm leaves the agent’s internal representations, which contain rich records of its evolving execution state, outside the skill optimization loop. We ask whether an agent can improve its external textual skills by reflecting on its own internal representations. We introduce Rep2Skill, a representation-guided framework for self-evolution on agent skills. Specifically, upon the collected agent rollouts, Rep2Skill models their internal model representation trajectories to localize turns that deviate from successful execution dynamics, and it further interprets these signals alongside the execution contexts as actionable textual feedback for targeted skill revision. Experiments on two agent environments with two open-source LLMs show that Rep2Skill consistently outperforms text-only approaches in the self-evolution setting, where the same LLM serves as both executor and optimizer without a stronger external model. This establishes a promising direction moving agent self-improvement beyond text-only reflection.
[AI-99] Do Self-Evolving Skills Generalize to Held-Out Tasks?
链接: https://arxiv.org/abs/2609.39148
作者: Xihao Piao,Zifeng Wang,Zhen Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:AI agents can externalize what they learn from past tasks into reusable \emphskills, such as procedures, checklists, code, or other executable artifacts, that can be retrieved and reused when solving new tasks. Self-evolving skill methods keep rewriting these skills after each round of practice on training tasks, and the skill is then used on new tasks of the same kind. We ask a question: does the improvement a skill shows on its training tasks carry over to new test tasks? We test five self-evolving methods and a one-shot skill on six benchmarks, with the same model, the same agent, and the same train/test split for every method. Of the 21 skills that improve on their training tasks, 5 keep all of that improvement on the test tasks, 13 keep part of it, and 3 keep none of it. No existing method is best everywhere. When we read the skills, the ones that carry over badly often fix details that should depend on the task, such as column names and output files, or turn a fix for one failure into a rule for every task. An LLM judge that reads the skill content can often see this: it ranks finished skills the same way the test results do in 86% of pairs. But it predicts the effect of a single edit poorly, so edits still have to be tested by running them. Based on these findings, we describe Generalizable Skill Optimization (GSO), which keeps only a guide for writing skills and writes a new skill for each task; it scores highest on all six benchmarks.
[AI-100] MADBench: Benchmarking the Security of Multi-Agent Debate
链接: https://arxiv.org/abs/2609.39146
作者: Yuwan Liu,Jiaming Zhang,Yue Huang,Sisi Duan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-agent debate (MAD) can improve large language model (LLM) reasoning by allowing multiple agents to exchange and critique their answers to the same task. However, the interactions that enable agents to correct mistakes can also spread adversarial errors and steer the agents toward an incorrect answer. Although some efforts have been made to examine particular attack types on MAD, systematic evaluation of MAD under diverse attacks remains limited. A central question is whether debate mitigates adversarial influence or amplifies it. In this paper, we present MADBench, a benchmark for evaluating the security of MAD. We organize attacks into a layered taxonomy following the MAD workflow, incorporating both established attacks and new strategies tailored to debate. We evaluate six attack families over 356 source tasks and 3,958 test cases, examining their effects on the final answer and the propagation of adversarial influence. Our results show that, under attacks, MAD does not necessarily improve LLM reasoning. Compared with a single-agent baseline, MAD can mitigate attacks on answer accuracy in question-answering tasks while amplifying unauthorized reads or writes in both question-answering and workspace tasks. Moreover, even when three out of five agents collude, the attack changes the final answer from correct to wrong on only 28.30% of tasks answered correctly without attack, while only 3.26% of initially correct honest agents switch to wrong answers during debate. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.39146 [cs.AI] (or arXiv:2609.39146v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.39146 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-101] Blackout vs. Freeze: Analyzing Physical Failure Modes of VLAs under Camera Faults
链接: https://arxiv.org/abs/2609.39145
作者: Heejae Suh,Jongwook Han,Zahra Gholami,Yohan Jo
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Unreliable visual inputs can harm task performance and cause potential physical safety risks for vision-language-action (VLA) models. We analyze how \pi 0.5 and GR00T models act under input faults such as image blackouts and freezing. We find that blackout and freezing produce distinct physical failure modes even when task-success rates are similarly low: freezing causes more extreme joint behavior, whereas blackout after gripper closure can cause more object drops, most markedly without proprioception. Selective intervention studies reveal that proprioception (current robot state) partly compensates for the removed robot depictions and reduces non-target contact. However, it cannot sufficiently restore task success when wrist-view object information is removed, even when aided by the remaining scene view. We then evaluate two mitigation approaches: camera-blackout training and training-free replacement of faulty visual embeddings. Both improve task success in selected conditions, but can increase unintended contact or disturbance to surrounding objects. Real-robot trials further show that successful execution under camera faults can still involve unintended physical interactions. These findings motivate designing VLA policies that use the robot and object information still available under camera faults to limit hazardous motion.
[AI-102] Schema: Discovering Unknown Environments via Agent ic Program Induction
链接: https://arxiv.org/abs/2609.39140
作者: Guanning Zeng,Jiani Wang,Wenjie Ma,Shaofeng Yin,Chenyang Wang,Shichen Liu,Angjoo Kanazawa,Wode Ni,Xiuyu Li,Andrea Zanette,Haiwen Feng
类目: Artificial Intelligence (cs.AI)
备注: Project Website: this https URL
Abstract:Learning to complete tasks in unfamiliar environments with unknown rules remains a key challenge for LLM agents. Current LLM agents often record their discoveries in prose, which may not provide a compact, explicit account of how the environment works. Inspired by how scientists organize observations into testable, predictive theories, we introduce Schema, an agent harness that organizes learning and action through interactive program induction. The LLM agent decides what to investigate and how to act, expressing its evolving understanding of the environment as executable programs. The harness consists of a persistent program workspace and a small set of interfaces for checking these programs against the interaction history, planning within them, and executing plans under step-by-step verification. Schema raises ARC-AGI-3 RHAE from 58.7% to 99.2% with the same base model, solves 100% of the public DiG-bench games, and reaches the median performance of the top-50 human players on MazeBench. Extensive analysis shows the effectiveness of Schema in unknown mechanism discovery, and ablations confirm the contribution of each component.
[AI-103] BELIEFRAG : Making Adaptive RAG State-Aware under Evolving Evidence
链接: https://arxiv.org/abs/2609.39139
作者: Hongji Pu
类目: Artificial Intelligence (cs.AI)
备注: 20 pages, 7 figures, 10 tables
Abstract:Adaptive RAG uses signals such as confidence, relevance, support, and retrieval quality to decide when to search or correct evidence. In multi-step retrieval, however, these local signals must be combined into a persistent view of what the current evidence supports, what remains missing, and which action should follow. Existing methods often use such signals as separate triggers, making it difficult to preserve a coherent evidence state across a trajectory; we call this problem evidence-state fragmentation. We introduce BELIEFRAG, a closed-loop controller that updates an explicit state over sufficiency, reliability, conflict, uncertainty, evidence gaps, and acquisition cost, then chooses among retrieval, query rewriting, verification, answering, stopping, and abstention. Across six QA benchmarks with gpt-oss-120b, BELIEFRAG reaches mean token F1 0.572 with 3.89k tokens per question, outperforming fixed iterative retrieval (0.555 F1) while using 39% fewer tokens. The same quality-cost pattern transfers to Qwen3-32B, where BELIEFRAG reaches 0.552 F1 versus 0.523 for iterative retrieval while using 35% fewer tokens. Analysis shows that the main gains come from corrective re-retrieval rather than pruning alone, while several belief dimensions are redundant and calibrated answerability plays the strongest operational role. Calibration improves threshold stability across related evidence sources, although source shift can still invalidate the same decision signal.
[AI-104] -Router: Learning Thalamic Routing for Reasoning with Parameter-Efficient Reinforcement Learning
链接: https://arxiv.org/abs/2609.39109
作者: Liuxian Ma,Jiale Dai,Jiaqi Li,Lu Mi
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Parameter-efficient reinforcement learning aims to improve reasoning with a compact trainable interface to a pretrained model. We introduce the Thalamic Router (T-Router), which concentrates adaptation on the reuse of completed computations. A compressed, addressable bank preserves block changes; a depth-recurrent controller conditions their selection and relative-scale writeback. This coupling gives thalamic context-dependent routing a concrete computational form: learn which earlier contributions a receiving layer uses, and with what influence. Correctness rewards train the interface while preserving backbone parameters and layer order. On an 8.95B-parameter backbone, T-Router allocates 41.73M parameters (0.466% of the backbone) and achieves 83.64 +/- 1.16 MathAvg after GSM8K RL, compared with 73.79 +/- 1.83 for full-parameter GRPO across three evaluation rounds. At a comparable parameter budget and with matched retries, it exceeds LoRA’s 77.28 +/- 1.95 MathAvg, improving all three task families and raising mean AIME accuracy from 48.33 to 60.56. Capacity-controlled comparisons favor addressable block changes and recurrent context; separate search training extends the interface to tool-mediated reasoning. These results establish controlled computation reuse as an effective route to parameter-efficient reasoning reinforcement learning.
[AI-105] MASCRDM: Multi-Agent System for Compliance Risk Detection and Mitigation in Training Process of Large Language Models
链接: https://arxiv.org/abs/2609.39107
作者: Yan Zhang,Chuming Wei,Ruien Li,Yaoyao Peng,Wusheng Zhang,Guangwen Yang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models (LLMs) have been applied in various fields. However, ensuring compliance and safety of LLMs, such as avoiding discrimination and bias, still remains a challenge. Current efforts mainly focus on detecting and filtering inputs and outputs of the trained models, rather than studying the intrinsic architecture of the models in real-time. To tackle this challenge, we analyze the LLMs training process and discover two critical issues: 1) Most of the existing methods are predominantly static in their approach to detection and filtering, achieving only localized optimizations without systematically enhancing the compliance of LLMs. 2) Another issue with existing approaches is the lack of real-time risk detection and mitigation across the full training process, which leads to limited flexibility. Motivated by these, we propose MASCRDM (Multi-Agent System for Compliance Risk Detection and Mitigation) during the LLM training process. Firstly, we develop a set of compliance rules based on existing Artificial Intelligence (AI) laws and a compliance-specific LLM with the instruction of compliance law experts. Then, we deconstruct LLMs into several components and identify key nodes based on the compliance knowledge graph. During LLMs training, we implement our multiple agents in the whole process, giving compliance risk alerts and suggestions for LLM developers. Experiments on discrimination and bias benchmark demonstrate that our multi-agent system can effectively improve the compliance while maintaining reasonable semantic performance. The results indicate that our method provides an executable path for mitigating compliance risk from within the LLMs systematically.
[AI-106] Reserve-Aware Contrast Certificates for Conservative Bandits with Uncertain Baselines
链接: https://arxiv.org/abs/2609.39106
作者: Qinchuan Cheng
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Conservative bandits must improve an incumbent policy without exhausting a prescribed performance budget. When the incumbent is uncertain, separately bounding candidate and baseline rewards can charge twice for shared estimation error. We develop Reserve-C4B around the baseline-relative contrast itself. A shared confidence set yields an exact expression for this avoidable penalty and a tighter admissibility test at every fixed history. A reserve ledger separates statistical evidence from permitted performance deficit; a prefix-refresh extension recertifies accumulated decisions under the current confidence set without discarding previously certified credit. For linear rewards, self-normalized confidence sets provide simultaneous validity over time and adaptively generated candidates, and the resulting policy satisfies a conditional-mean performance constraint with high probability. Reproducible experiments isolate certificate coupling, prefix refresh, and historical information, showing large reductions in baseline fallback while exposing the limitations of frozen certificates.
[AI-107] Beyond Prediction: Steering VLM Agents with Retrospective World Modeling NEURIPS2026
链接: https://arxiv.org/abs/2609.39101
作者: Yongjiang Liu,Jie Zhang,Haoyue Zhang,Jingcai Guo,Deze Zeng,Song Guo
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026 (27 pages, 8 figures)
Abstract:Equipping VLM agents with world modeling capabilities has shown strong potential for complex reasoning and long-horizon planning, while reducing the dependence of policy learning on costly real-world interactions. Existing methods mainly rely on prospective simulation to predict the consequences of candidate actions. However, this forward-only paradigm focuses on what will happen next and provides limited constraints for verifying whether an action is causally consistent with the observed state transition, which can lead to plausible-looking but physically incoherent behaviors. In this paper, we challenge the view of world modeling as only prospective prediction and introduce Retrospective World Modeling, a new agent learning paradigm that enables agents to reason backward by estimating the retrospective attribution distribution P(\hatat|s_t, st+1) for the action that most likely caused a given transition. Based on this capability, we formulate the Self-Consistency Reward (SCR), an intrinsic signal that measures the probabilistic consistency between the policy action and the retrospective explanation. Integrating SCR into reinforcement learning provides dense transition-level feedback and steers agents toward behaviors that are both task-effective and physically grounded. Extensive experiments across diverse agentic tasks show that our method substantially improves policy robustness and generalization over prospective-only world modeling baselines.
[AI-108] A Generalisation Signal Need Not Be a Model-Selection Signal NEURIPS2026
链接: https://arxiv.org/abs/2609.39099
作者: Aditya Nagarsekar,M P Ashish Bhat,Aadi Nesarkar,Vrishti Godhwani,Rahul Yedida,Aditya Challa,Danda Sravan,Snehanshu Saha
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted (poster) at the NeurIPS 2026 Workshop “I Can’t Believe It’s Not Better: Failure Modes of AI in Biology” (ICBINB-BIO). 24 pages, 4 figures, 22 tables. Code: this https URL
Abstract:Model selection in computational biology often relies on validation data drawn from the training regime, even when deployment lies outside it. When validation no longer preserves which model is best, a natural alternative is to rank candidates using properties of the trained network itself. We test this idea using a novel, forward-only proxy motivated by the norm of the Hessian, alongside common Hessian measures, across molecular property, protein fitness, and drug-response tasks. Contrary to our hypothesis, geometry does not become more useful as validation Spearman correlation deteriorates: augmenting validation helps some shifts but significantly harms others. More surprisingly, the proxy still correlates with generalisation gap on most tasks even when Hessian trace and top-eigenvalue relationships are weak or reversed, yet this signal does not reliably identify the deployment-best model. A curvature bound need not preserve cross-model rankings, and low geometric scores can even favour collapsed predictors. Thus, a generalisation signal need not be a model-selection signal.
[AI-109] SCIC: Scope- and Codebook-Aware Instruction Conditioning for Speaker-Adapted Expressive TTS
链接: https://arxiv.org/abs/2609.39088
作者: Longyu Lu,Zongwei Du,Mengtao Xing,Zhuoqun Liu,Zifan Guan,Meiguang Jin,Junfeng Ma
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注:
Abstract:Long-form live-streaming TTS requires context-dependent prosody and paragraph-level coherence. However, many existing instruction-based TTS systems use global or uniform conditions, providing limited explicit control over clause-level relative prosodic changes. We introduce Speaker-Relative Inline Prosody Control, where each Pitch, Energy, or Speed instruction targets a clause relative to the preceding clause from the same speaker, while Pause uses an absolute duration interval. In codec-based TTS, Speed and Pause affect sequence length, whereas Pitch and Energy rely on residual codebooks. By analyzing Qwen3-TTS RVQ codebooks, we find that Energy concentrates in early residual codebooks, whereas Pitch accumulates across a deeper prefix. We therefore propose Scope- and Codebook-Aware Instruction Conditioning (SCIC), combining a Temporal Instruction Router for frame-level tag activation with Tag-Specific Codebook Weighting over residual codebooks. SCIC improves speaker-relative Pitch and Energy control over standard instruction fine-tuning using text-token tags. We further apply multi-reward GDPO post-training to jointly optimize control and quality, improving control accuracy while preserving CER and speaker similarity. In long-form synthesis, SCIC produces a more distinct paragraph-level expressive hierarchy than speaker-adapted SFT without instructions. Audio demos are available at: this https URL
[AI-110] rustworthy Runtime Error Healing in Real-World Repositories: A Benchmark and Guardrail
链接: https://arxiv.org/abs/2609.39086
作者: Gou Tan,Pengfei Chen,Zhensu Sun,Jieke Shi,Junkai Chen,Ting Zhang,Weifeng Sun,Junda He,Shuai Liang,Chuanfu Zhang,Lwin Khin Shar,David Lo
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:Runtime error healing lets a crashed program continue by generating code that repairs its live runtime state. Recent work shows that LLMs can generate such healing code, but it is evaluated only on small competition programs, and executing LLM-generated code inside a live process raises safety concerns that remain unaddressed. In this paper, we take LLM-based runtime healing toward practical use in real-world repositories. We first build HealBench, a benchmark of 265 runtime errors from 18 real-world repositories, each paired with a reference execution on the patched version. HealBench also provides a unified framework that lets LLM agents heal with cross-file context and live runtime state. We then design HealGuard, which requires healing code to be written in HealCore, an analyzable subset of Python, and uses static and dynamic taint analysis to check whether state changed by healing reaches operations protected by developers. We evaluate a dedicated healing method and three general coding agents with three backbone LLMs. The best setting resumes execution in 38.11% of instances and passes the target test in 28.68%, showing that existing agents can already heal a meaningful share of real repository-level crashes. However, among executions that pass, HealGuard flags 17.4% whose healing-changed state may reach a protected operation. On 684 controlled cases, HealGuard detects all unsafe cases, at the cost of a 68.42% false positive rate.
[AI-111] Coding Agents for Coding Theory NEURIPS2026
链接: https://arxiv.org/abs/2609.39081
作者: Abraham Yeung
类目: Information Theory (cs.IT); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at the 6th Workshop on Mathematical Reasoning and AI (MATH-AI), NeurIPS 2026. 20 pages, 1 figure, 2 tables. Code and data: this https URL
Abstract:We spent five weeks using an LLM coding agent on open problems in coding theory: finding large sets of four-letter words, such as DNA barcodes, that stay far apart in edit distance. The agent wrote the verifiers and search code; a human chose the problem and set the verification protocol. Restricting the search to codes with a prescribed symmetry, a classical technique, shrank the problem about fourfold and raised the best known code of length 6 and minimum edit distance 3 from 114 to 120 words ( E_4(6,3) \geq 120 ). The same pipeline improved twelve further lower bounds at lengths 6 to 9 and distances 3 to 6. We give the failures equal space. Our own search stopped at 116 and recorded the last symmetry class as topping out at 112; a second agent session, running the same search with a better operator, found the 120. A later verdict that the method did not carry over to length 7 was wrong for the same reason, and an earlier instance cost three weeks. Each time, an intermediate result was written down, never rechecked, and treated as a fact that ruled out further search. Checking final outputs, as our protocol required, does not catch such errors.
[AI-112] Multi-LLM Collaborative Alignment via Stackelberg Games
链接: https://arxiv.org/abs/2609.39076
作者: Christina Hahn,Shangbin Feng,Dean Light,Swastik Roy,Hila Gonen,Yulia Tsvetkov
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:A pool of language models can collaborate and improve collectively by learning from one another’s responses. These interactions depend on the instructions used during training. Existing methods typically sample instructions uniformly, even though their usefulness may change as the models improve: an instruction on which models’ responses once differed in quality may later be answered equally well, while a previously difficult instruction may begin to provide a useful learning signal. We propose Stackelberg Alignment, a game-theory-inspired leader-follower framework that turns instruction selection into an adaptive curriculum. An EXP3 bandit acts as the leader, allocating a fixed sampling budget across instructions and updating its sampling distribution using a reward that combines instruction difficulty and response discriminability. The language models act as followers: they respond to the selected instructions, evaluate one another’s responses, and learn from the resulting preference signals through DPO or GRPO. The framework uses Elo-style reputation-weighted peer judgment and reputation-based opponent matching to support reliable and competitive model interactions. Experiments across three heterogeneous model pools and 12 benchmarks spanning scientific discovery, reasoning, code, instruction following, and knowledge show that Stackelberg Alignment achieves the highest macro-average across three diverse model pools, outperforming the strongest training-time baseline by up to 7.4% and the best static inference baseline by 12-25%. Analysis confirms that the adaptive leader concentrates duels on the most informative instructions, and ablations show that both reputation-weighted judgment and reputation-based matching improve the effectiveness of multi-LLM evolution.
[AI-113] RAG Scope: A Leakage-Controlled Cost-Aware Evidence-Gating Protocol for RAG Hallucination Triage ICTAI2026
链接: https://arxiv.org/abs/2609.39075
作者: Zeming Liu,Qibai Chen,Jingtao Zhang,Hang Lyu
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 8 pages, 5 figures, 8 tables. Accepted at the 2026 IEEE International Conference on Tools with Artificial Intelligence (ICTAI 2026)
Abstract:Retrieval-augmented generation (RAG) systems need inexpensive ways to route generated answers: accept low-risk outputs, review uncertain ones, and reserve strong verifiers for the expensive tail. We present RAGScope, a leakage-controlled protocol for evaluating local evidence gates that use only the task input, retrieved context, and answer text. The protocol combines context-grouped splits, fold-scoped preprocessing, group bootstrap intervals, deployment operating points, end-to-end runtime, and explicit source-shift stress tests. On three RAGTruth tasks, the enhanced gate RAGScope-E reaches 0.798 AUROC and 0.660 average precision (AP) in pooled grouped cross-validation. Its pooled AP exceeds ROUGE-L by 0.034 with a 95% context-group interval of [0.002, 0.064], although the AUROC gain is not significant and ROUGE-L remains stronger on data-to-text. At a top-10% review budget, RAGScope-E attains 0.748 precision; accepting the lowest-risk 50% yields 0.141 residual unfaithfulness. RAGScope-E runs in 6.22 ms/example on CPU, versus 145.75 and 223.07 ms/example for the tested DeBERTa-NLI and HHEM settings. A 14,900-example HaluBench stress test exposes the deployment boundary: an in-domain calibrated gate reaches 0.879 AUROC, but leave-source-out calibration averages only 0.466. Target-only calibration recovers to 0.675 AUROC with 100 labels per source and 0.685 with 200. Cheap evidence gates are therefore useful routing components, but learned calibration must be validated and adapted within the target domain.
[AI-114] HO-FL: Hybrid-Order Federated Learning for Heterogeneous Edge Devices
链接: https://arxiv.org/abs/2609.39074
作者: Qiyuan Chen,Xian Wu,Yanan Ma,Xianhao Chen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 30 pages, 2 figures
Abstract:Federated learning (FL) on memory-constrained edge devices faces a dilemma: first-order (FO) optimization (i.e., backpropagation) demands substantial memory, whereas zeroth-order (ZO) optimization suffers from severe convergence slowdown. To resolve this dilemma, we introduce HO-FL, a hybrid-order FL framework that trains a model’s bottom segment with ZO optimization and its top segment with FO optimization. Each device can flexibly select its order boundary according to its memory budget while participating in the training of the same global model. Moreover, our convergence analysis reveals a new, fundamental trade-off: clients with larger FO-trained segments can provide more accurate updates, but favoring them can underrepresent other clients’ data. We connect this trade-off to the bias and variance of actual multi-step local updates, yielding a sampling optimization problem and a practical dimension-aware approximation with direct model averaging. Experiments on language tasks examine task performance, client memory, and sampling under data heterogeneity. The results show that hybrid-order local training can retain much of the full-FO performance with substantially lower client memory requirements. Our code is available at this https URL.
[AI-115] Can Agents Trust Their Skills? Uncovering Unsafe Chains of Trust in Skill-Based LLM Agents
链接: https://arxiv.org/abs/2609.39065
作者: Yan Wang,Zhihao Zhang,Ke Chen,Kai Chen,Yaqin Zhang,Duohe Ma,Jun Dai,Xiaoyan Sun
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 26 pages, 12 tables, 8 figures, appendices
Abstract:LLM agents increasingly rely on installable skills, which are packages of instructions, code, and resources that equip them with task-specific capabilities and, once installed, can be automatically invoked across subsequent user tasks. This creates a chain of trust in which users delegate authority to agents, while agent frameworks admit skill-provided content into the agents’ context with insufficient validation, allowing malicious skills to influence agent behavior under that delegated authority. Yet, little is known about whether this trust model adequately constrains untrusted skill content before it reaches security-sensitive operations, or how frequently such trust violations arise in real-world agents. We present TrustProbe, a framework for uncovering unsafe chains of trust in skill-based LLM agents. First, TrustProbe analyzes agent source code to identify source-to-sink call paths from skill-controlled inputs to security-sensitive operations. Second, it generates semantically realistic this http URL seeds with injected canaries and evolves them through feedback-guided scheduling and mutation. Finally, it validates vulnerabilities using an oracle that confirms attacker-controlled flows and verifies observable harm. Across 11 open-source agents, eight with more than 10,000 GitHub stars, TrustProbe identifies 104 taint-style vulnerabilities. Validation on a large corpus of real-world skills collected from public hubs such as ClawHub further shows that 25.1% of skill-agent trials exercise the identified vulnerable paths, with payload injection successfully weaponizing 15 of the vulnerabilities. These results reveal a systematic trust failure in skill-based LLM agents: untrusted skill content can reach security-sensitive operations and exercise authority delegated by users to their agents.
[AI-116] A 3GPP-Compliant Benchmark Dataset for RIS-Aided Beyond 5G Networks
链接: https://arxiv.org/abs/2609.39058
作者: Pujitha Mamillapalli,Pankaj Singh Rathour,Abhinav Kumar
类目: Databases (cs.DB); Artificial Intelligence (cs.AI)
备注:
Abstract:Reconfigurable Intelligent Surfaces (RIS) are emerging as a key technology for programmable wireless environments in the beyond the fifth generation (B5G) networks. However, data-driven RIS research remains bottleneck by the lack of standardized, high-fidelity and open-source datasets. In this paper, we introduce a large-scale 3GPP TR 38.901-compliant dataset for RIS-aided millimeter wave (mmWave) networks, that considers severe path loss, blockage sensitivity, and spatial channel sparsity make the RIS assistance more impactful. The dataset spans various canonical 3GPP deployment scenarios across 20 controlled variants, capturing diverse user densities, fading conditions, and blockage regimes. Uniquely, every sample includes oracle RIS phase configurations obtained via a globally optimal brute-force codebook search, providing gold-standard supervision labels that are absent from any existing public dataset. Rich multi-task annotations comprising full channel state information (CSI), per-link channel decomposition, optimal phase matrices, and channel quality index (CQI) labels support a broad range of machine learning paradigms and downstream tasks, including phase optimization, channel estimation, and interference management. As the primary benchmark task, we introduce a novel CSI-to-CQI mapping that frames RIS-aided link-quality prediction as a scalable scalar classification problem, thereby avoiding the exponential output complexity of the direct phase vector prediction. We have evaluated this mapping against state-of-the-art architectures under in-distribution, out-of-distribution, and real-world hardware measurement conditions. Our dataset provides a reproducible, extensible, and community-ready foundation to accelerate data-driven research in RIS-aided B5G networks.
[AI-117] Structure-aware Reinforcement Learning for Protein Directed Evolution
链接: https://arxiv.org/abs/2609.39048
作者: Zikun Nie,Suyuan Zhao,Yizhen Luo,Siqi Fan,Zaiqing Nie
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Protein optimization remains a longstanding goal in life sciences. Existing machine learning-assisted directed evolution (MLDE) methods primarily rely on sequence-only features, overlooking the critical spatial constraints and co-evolutionary interactions encoded in protein structures. However, directly integrating structural information remains challenging due to the scarcity of reliable mutant structures. To address these issues, we propose StructEvo, a novel structure-aware reinforcement learning framework for protein directed evolution. StructEvo employs a delta-structure fusion encoder to approximate mutant structure features via feature differences, enabling dynamic incorporation of spatial knowledge. The vast mutation space is then decomposed into manageable subspaces through a structure-aligned hierarchical action network, while a geometric constraint further stabilizes delta feature learning. Our approach outperforms prior state-of-the-art methods by 9.2% and 16.3% on two challenging optimization benchmarks, and further identifies an experimentally validated epistasis pattern in GFP, highlighting the importance of structural guidance for effective protein directed evolution.
[AI-118] Hard-Gate Candidacy in a Deployed Validator Suite NEURIPS2026
链接: https://arxiv.org/abs/2609.39037
作者: Xin Xu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 7 pages, 1 figure, 2 tables. NeurIPS 2026 Workshop: Can We Trust the Judge? (JUDGe)
Abstract:Before a validator can be promoted to a hard gate on a deployment pipeline, it has to be shown that its firing separates outputs that reach users in working order from those that do not. We run that screen on 13 validators in a deployed generative agent, against 550 runtime and 350 static builds labelled by downstream outcome, and report each check’s marginal separation J=\mathrmTPR-\mathrmFPR with Newcombe intervals and Fisher exact tests. Two checks survive correction for multiple comparisons, two more are nominal only, and the remaining nine are not distinguishable from zero, three of them because they never fired on any sampled build. Execution itself is not random with respect to the property being gated, and this replicates: across four runs covering 1,867 builds and ten distinct runtime checks, probes were skipped on 144 of 895 broken builds and 1 of 972 acceptable builds (per-run rates 15.6% to 16.6% against at most 0.3%), every skip carrying the same unsafe-to-probe reason. Because a skipped check is recorded as a pass, this imposes a ceiling that no check quality can lift: a check that needs a live artifact cannot operationally detect more than about 84% of broken builds in this harness. For the one check with construct-specific labels, a detector built for blank output fires on 0 of 90 human-labelled blank builds (95% upper bound on sensitivity 3.3%), and the global frame statistic it approximates separates the classes only weakly (AUC 0.59), so the gap is not a threshold that needs tuning. The same gap appears one layer up: on a census of tens of thousands of judge-scored builds, 32.5% of rejections carry no recorded issue at all. We argue that evaluation records must distinguish a check that ran and passed from one that did not run, must carry the evidence for a rejection, and that an inventory of checks is not evidence about a gate.
[AI-119] Cycle-Aware Autoencoder with Cross-SignalConsistency for Railway Door Anomaly Detection MICRO
链接: https://arxiv.org/abs/2609.39035
作者: Ammar Bouketta,Smail Niar,Hamza Ouarnoughi,Eva Mutuzo Brindle
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 8 pages, 4 figures. Accepted and presented at the 29th Euromicro Conference on Digital System Design (DSD 2026)
Abstract:Passenger access doors are safety-critical subsystems in railway vehicles, yet detecting abnormal door behavior in real operation is challenging because faults are rare, diverse, and often unlabeled. This paper addresses railway door condition monitoring as a cycle-level unsupervised anomaly detection problem, where each complete opening-dwell-closing cycle is treated as a single monitoring unit. We propose the Temporal Cycle-Aware Attention Autoencoder with Cross-Signal Consistency (TCAA-CS), trained exclusively on nominal cycles. It combines a dual-stream encoder that processes continuous physical measurements (position, current, voltage) and binary logical states (door-closed, door-locked) through separate 1D-CNN branches, an LSTM encoder with temporal attention pooling, and a triple hybrid anomaly score fusing reconstruction error, latent-space deviation, and phase-aware cross-signal consistency. The consistency term helps identify cases where individual signals appear plausible but their inter-signal relationships become physically or logically inconsistent. On real industrial data from a passenger train in commercial service, TCAA-CS achieves 93.8% recall, 97.3% precision, and a 0.5% false-alarm rate, outperforming representative unsupervised baselines. System-level evaluation on an NVIDIA Jetson AGX Xavier supports the feasibility of real-time onboard deployment.
[AI-120] Search Shapes Conclusions: Auditing Evidence Selection Bias in Deep Research Agents
链接: https://arxiv.org/abs/2609.39026
作者: Shuyao Xiao,Shengling Wang,Xuan Chen,Ke Chao,Ming Cui,Feifei Qian,Chaoyang Mei,Fanlin Meng,Lulu Wang,Ziming Yu,Junxi Yin
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Deep Research agents synthesize evidence into cited reports, yet a well-cited report can still reach a misleading conclusion. Citation correctness checks whether cited sources support individual claims. It does not show whether adaptive search exposed a representative view of all documents made available for evaluation, which we call the candidate pool. Early findings redirect later queries, document choices, and stopping, so the documents an agent reads form a selective sample. Existing evaluations rarely account for this selection. We formulate the problem as adaptive evidence sampling and introduce Causal Evidence Selection Correction (CESS). CESS predicts each candidate document’s evidence direction and corrects the candidate-pool average using the logged probabilities of selecting each document and reaching each search round. Shrinkage stabilizes short searches, while intervals replace point estimates when some documents cannot be sampled. We also prove that estimating the average evidence direction of a common pool differs from measuring how a change in search policy alters the evidence read. The latter requires intervention. On questions from the MS2 systematic-review benchmark, CESS reduces mean absolute error against the candidate-pool average by 9.2% and reduces the estimate’s change under opposing document rankings by 39.4% relative to averaging the evidence scores of documents read. Across trajectories from a public Open Deep Research agent, the corresponding reductions reach 60.1% and 87.2% . A further 4,800 trajectories under paired interventions confirm that correcting a pool estimate and measuring a policy effect are different tasks. CESS therefore audits whether the evidence direction underlying a report reflects the documents available for evaluation, while a separate intervention analysis measures the effect of search decisions.
[AI-121] From Verification Failures to Reusable Guidance for Coding Agents
链接: https://arxiv.org/abs/2609.39022
作者: Yuqing Zhai,Xiaohong Chen,Lingming Zhang,Sriram Vishwanath,Grigore Rosu
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO); Programming Languages (cs.PL)
备注: 23 pages, including appendices
Abstract:Coding agents need to establish that a program satisfies a specification and that the specification captures the requested behavior. We study how expert diagnosis of verification failures can become reusable guidance for this work. Our approach combines executable language definitions in the K framework with a kit of procedures for constructing specifications, repairing proofs, and auditing their adequacy. A human-guided development campaign on HumanEval, a benchmark of 164 Python programming tasks, achieves a 164/164 success rate with the semantics and the kit, measured by final AI audit Pass verdicts after two targeted repairs. To examine whether auditing detects problems that successful proofs leave unresolved, we construct 12 author-reviewed pairs of clean and defective packages. Every package passes its K proofs, and completed audits identify all defects and accept all clean packages. We then use KleverBench to test specification and proof construction for 31 programs with changed operator meanings. Comparisons with complete acceptance rules and equally long generic advice yield mixed results across two model and budget settings, motivating further work on selecting useful guidance within resource limits. Human-reviewed Optimism proofs establish expected pause reverts for six operations within declared input bounds under London semantics with unbounded gas. We report progress, difficulties, and lessons toward agents that deliver programs with checkable correctness arguments.
[AI-122] Make Code as Policy Great Again: Frontier Agents Write Call and Evolve Robot Tools
链接: https://arxiv.org/abs/2609.39018
作者: Shijia Ge,Alex Zhou,Jianshu Zeng,Yexing Wan,Di Wu,Zelin Zheng,Yazhe Wang,Zhiqi Jia,Xuan Shangguan,Jay Zhu,Yijun Liu,Lingyu He,Sihang Wu,Xiao He,Hongcheng Gao
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Frontier models can control robots, but reasoning through every reach, grasp, and retreat makes manipulation slow and token-intensive. We revisit code as policy with a different division of labor: models build executable tools, code handles multi-phase motions, and models decide what to do next. We introduce URAI (Universal Robot-Agent Interface), which couples a programming agent that constructs robot tools with an execution agent that uses them in a feedback loop. The programming agent writes reusable and task-specific tools from task intent and refines them through execution feedback and human guidance. The execution agent selects and parameterizes these tools from current observations; each call runs a complete motion locally before returning control to the agent. Unlike delegating subsequent decisions to a generated program, this design retains model-level decision-making between tool executions. Validated tool revisions persist across episodes without updating foundation-model weights, and a shared GUI and API make the same tools available to humans and agents. Across five RoboDojo tasks and four frozen execution agents, URAI raises aggregate success from 18.0% to 53.0% relative to direct fingertip control, with the largest gain on Swap Blocks; with the same tools, a program written in advance reaches only 24% against 56% for two agents deciding after each call. Three of the four agents also finish episodes 1.3-1.5 times faster with 1.5-1.7 times fewer execution-agent output tokens; DeepSeek-V4-Flash’s cost barely changes. We further evaluate URAI on seven real-world AgileX dual-arm tasks, spanning object manipulation, cloth folding, and human-interactive tic-tac-toe. URAI connects the coding and decision-making capabilities of frontier agents, organizing robot control around reusable tools that agents can both invoke and revise.
[AI-123] C-STRIDE: An Observation-Driven AI Digital Twin for Predicting Basin-Wide Flood Fields from Sparse Stream-Gauge Histories
链接: https://arxiv.org/abs/2609.39005
作者: Yanjie Tong,Phillip Si,Yuan Qiu,Peng Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Emergency managers need to know where floodwater is, how deep it is, and how it will change over the coming hours across an entire river basin. During a flood, however, real-time measurements come from only a handful of stream gauges, and high-resolution hydrodynamic models are too costly to rerun each time new data arrive or to run as large ensembles. We present C-STRIDE, an observation-driven AI digital twin that turns short records from a few stream gauges, together with terrain and rainfall, into basin-wide maps of water depth and extends these predictions up to a day ahead. It is trained on simulations from a calibrated two-dimensional hydrodynamic model and needs no separate data-assimilation step. In the Des Plaines River basin near Chicago, six gauges inform predictions over 4.2 million 30-m grid cells. Terrain improves the predictions most, rainfall keeps errors from growing over longer horizons, and together they reduce errors by about 40% compared with gauge records alone. When future rainfall is known, errors remain near 15% one day ahead, compared with nearly 40% without rainfall. Given real instead of simulated gauge records, the model shifts its predictions toward the observed hydrographs at three of six gauges without retraining, and it runs about 150 times faster than the hydrodynamic model. These results show how sparse gauges, terrain, and rainfall can be combined into fast, continuously updated flood predictions, a step toward operational flood digital twins that still requires testing with real-time data and rainfall forecasts.
[AI-124] Sparse-WAM: Accelerating World Action Models via Action-Guided Sparse Imagination
链接: https://arxiv.org/abs/2609.38984
作者: Xinling Xie,Haodong Wang,Jiazhi Mi,Zhiming Liu,Zicong Hong,Xiaoyi Pang,Qianli Liu,Yangjia Hu,Ying Chen,Zhengyang Yan,Song Guo
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: Xinling Xie and Haodong Wang contributed equally to this work
Abstract:World-action models (WAMs) leverage pretrained video models to improve generalization in robot control by jointly predicting future visual states and actions. This capability comes at a substantial inference cost, as dense future-frame tokens are repeatedly processed during denoising. Prior methods address this by token pruning that prioritizes visual fidelity to reduce denoising costs in video diffusion models. However, these methods do not use action relevance to determine which future-frame tokens to retain during joint denoising in WAMs. In this paper, we propose Sparse-WAM, a training-free framework for action-guided sparse imagination that selectively processes future-frame tokens to accelerate WAM inference. We observe substantial overlap in the spatial distribution of attention from action tokens to future-frame tokens (action-to-future attention) between consecutive denoising steps, despite continued updates to the future representations. Motivated by this, we develop Action-Guided Token Selection to retain frame-specific action-relevant regions together with cross-frame context. However, a naive implementation can incur attention-scoring and token-packing overhead that offsets the computational savings from pruning. We therefore introduce Pilot, an efficient engine that reduces sparse inference overhead through lightweight scoring and cross-step reuse of token selections. On LIBERO with FastWAM-Joint and RoboLab-120 with Cosmos 3 Edge, Sparse-WAM achieves inference speedups of approximately 2.0\times and 1.8\times , respectively, over dense eager inference on an NVIDIA RTX 4090, while largely preserving task performance.
[AI-125] Approval Laundering: Systematizing Approval–Execution Binding Failures in AI Coding-Agent Harnesses
链接: https://arxiv.org/abs/2609.38983
作者: Yang Wang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Modern AI coding-agent harnesses (Claude Code, Codex CLI, Cursor) rest their security boundary on a largely unexamined assumption: that the action A a human approves is the same action A’ the harness executes, where A is fixed by a stated policy for what a scope grant or session-scoped approval authorizes. We show this assumption fails systematically and reproducibly. We introduce Approval Laundering, a taxonomy of six failure modes by which a harness’s enforcement mechanism silently substitutes A’ for A after approval: Scope, Argument, Temporal, Tool, Delegation, and Semantic laundering. Unlike prior work that evaluates risk classifiers against static corpora or infers implicit authorization boundaries, we study credential-binding integrity: given an already-approved action, does the harness dispatch exactly that action? Instrumenting Claude Code’s pre-execution mediation point (PreToolUse), we conduct a controlled, headless, repeated-measures study of all six classes (N=19-20 runs each), reporting a Bound-Gap Rate (BGR) with Wilson confidence intervals and inter-rater agreement (kappa=1.0). We prototype Approval Token, a keyed capability Hk(principal, agent_id, session_id, tool, arguments, scope, expiry) issued by a mediator that never returns the key to the agent, evaluated via paired before/after replay of 118 runs (McNemar’s exact test). The token fully eliminates Delegation laundering and, for our seeded session-identity-mismatch construction, Temporal laundering (p10^-5), but by design leaves Scope laundering unaffected and shows no significant reduction in Argument laundering (p=1): an honest negative result, since these two classes leave every recorded dispatch field unchanged, diverging one process level below what a field-only verifier can observe. We discuss implications for defenses that bind only at the tool-invocation boundary.
[AI-126] SimEX: Simulation-Integrated Robotics AutoResearch
链接: https://arxiv.org/abs/2609.38982
作者: Jiaheng Hu,Roberto Martin-Martin,Peter Stone,Rocky Duan,Zhenyu Jiang,Guanya Shi
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Coding agents powered by large language models (LLMs) have shown remarkable abilities to autonomously reason about and achieve goals in the digital world. However, bringing this success to the physical world remains challenging. On the one hand, direct generation methods (e.g., Code as Policies) often suffer from the LLMs’ insufficient understanding of robots and physical environments. On the other hand, iterative trial-and-error tuning in the physical world (e.g., physical autoresearch) induces significant experimental cost and safety concerns. We introduce SimEX: Simulation-Integrated Robotics AutoResearch, an autoresearch framework that tightly integrates simulated experimentation, enabling coding agents to efficiently acquire physical capabilities for controlling real robots. SimEX operates in two stages. First, the agent conducts open-ended probe-and-optimize iterations in simulation, developing a robot toolbox with robust and generalizable capabilities. Second, the agent adapts the toolbox and the simulator together through only a few physical trials: each trial corrects the simulator, and the corrected simulator is used to diagnose failures and screen candidate repairs. We evaluate SimEX extensively in sim-to-sim settings and on physical robots. On challenging real-world manipulation tasks including towel folding, barcode scanning, and plate manipulation, SimEX enables coding agents to efficiently acquire robot skills without any demonstration and with only 10 minutes of real-robot interaction. These results suggest that simulation can be a critical component in achieving physical intelligence, not only as a source of training data that must closely replicate the real world, but also as a roughly correct laboratory where a coding agent develops the knowledge and procedures needed to act on the robot. More details and robot videos at this https URL
[AI-127] Scale-Split Neural Operator for Memory- and Data-Efficient 3D Turbulence Prediction
链接: https://arxiv.org/abs/2609.38977
作者: Shaoxiang Qin,Yucheng Zhao,Zongyi Li,Liangzhu Leon Wang,Xiongye Xiao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 40 pages
Abstract:Neural surrogates have emerged as fast alternatives to the numerical simulation of three-dimensional turbulence. However, training them at high resolution remains challenging, since the memory of full-field models grows with the resolution. In addition, full-resolution training data are expensive to simulate and store, and therefore scarce. We introduce ScaleSplit-NO (Scale-Split Neural Operator), which exploits the scale structure of turbulence with two neural operators: a Parent predicts the global coarse field at the next time step, and a Child predicts full-resolution local patches conditioned on this prediction. Neither model operates on the full-resolution field. The Child is pretrained alone and then attached to the Parent’s coarse prediction through zero-initialized connections. On two complex high-resolution turbulence benchmarks, ScaleSplit-NO surpasses all competing baselines in both prediction accuracy and data efficiency. On the higher-resolution dataset JHTDB256 ( 256^3 ), its normalized mean squared error (NMSE) is 53% lower than that of the strongest baseline, and its training memory is 79% lower than that of the most memory-efficient baseline. We further demonstrate its effectiveness for urban wind prediction in a real district of Montreal on a 500\times150\times500 grid, reducing one-step NMSE by 65.8% relative to the baseline. Moreover, swapping in a Parent trained on additional coarse fields improves prediction without retraining the Child, providing further accuracy gains at a small storage cost.
[AI-128] RealWorldShop: Benchmarking and Improving Conversational Shopping Agents in Real-World E-commerce
链接: https://arxiv.org/abs/2609.38974
作者: Xinwei Yang,Kelong Mao,Yudong Guo,Sulong Xu,Simiu Gu,Chen Huang,Wenqiang Lei
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models are reshaping ecommerce from static recommenders into interactive shopping assistants, yet real-world shopping requires session-level decision support: users reveal and revise constraints, coordinate multiple goals, and expect product-grounded recommendations over a full conversation. Existing benchmarks are mostly outcome-oriented or execution-oriented, leaving this evolving decision process under-evaluated. We introduce REALWORLDSHOP, a benchmark built on 3.28M grounded products, structured shopping episodes, a profile-grounded and actioncontrolled user simulator, and role-play evaluation. Our analysis shows that current systems produce locally plausible responses but struggle with state tracking, constraint updating, and grounded convergence, especially under ambiguous intent, bundle, and multi-intent scenarios. We further propose REALSHOP_AGENT, an executable session-control framework with explicit state management, shopping-flow control, catalog-grounded retrieval, and runtime guards. Experiments show that REALSHOP_AGENT consistently outperforms strong baselines on REALWORLDSHOP.
[AI-129] When Order Matters: First-Speaker Bias and Mitigation through Personality in Sequential Multi-Agent Debate
链接: https://arxiv.org/abs/2609.38964
作者: Duofeng Xu,Bryan Hooi,Dandan Qiao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-agent debate (MAD) is often used to improve large language model (LLM) reasoning, but sequential debate is rarely a neutral aggregator of agents’ opinions. We show that sequential MAD suffers from a pronounced first-speaker bias: agents disproportionately shape the final answer when they speak first. As a result, placing a stronger model after weaker ones can substantially offset its reasoning advantage. We then focus on the disadvantaged strong-agent-last setting and ask whether personality prompting can mitigate this imbalance. Drawing on the Big Five model, we study agreeableness and extraversion as behavioral interventions applied to either the strong or weak side. We find that their effects are trait-specific. Influence consistently shifts in the direction of lower agreeableness, and assigning low agreeableness to the stronger agent helps restore its lost influence and improves final accuracy. Extraversion, by contrast, produces less systematic changes in influence and accuracy, with its clearest effect appearing in agents’ verbosity. These findings show that effective MAD design depends not only on model capability, but also on how speaking order and induced interaction behavior shape the debate process.
[AI-130] Alleviating Hallucination in Reasoning Tasks with Training-Free Uncertainty-Guided Steering NEURIPS2026
链接: https://arxiv.org/abs/2609.38962
作者: Litian Liu,Qiqi Hou,Yubing Jian,Reza Pourreza,Mohammad Ghavamzadeh,Roland Memisevic,Yao Qin,Hong Cai
类目: Artificial Intelligence (cs.AI)
备注: Neurips 2026 main conference paper
Abstract:Recent work on hallucination detection in large language models has shown that, for a fixed pre-trained model and reasoning task, it is possible to estimate the model’s confidence in the correctness of its outputs. Such uncertainty estimates have primarily been used to improve truthfulness by detecting or filtering confabulations. In this work, we ask whether these signals can instead be used more proactively to directly improve the accuracy of model-generated answers. We propose USteer, a simple, training-free steering mechanism that adjusts a model’s layer-wise activations during inference using the gradient of a confidence measure with respect to the activations. This procedure nudges generation toward outputs with lower uncertainty at inference time, without modifying model parameters or requiring additional supervision. We show that this approach consistently reduces hallucination across a range of tasks, demonstrating that confidence signals can be leveraged not only for detection, but also for effective inference-time control of model behavior.
[AI-131] Routing Probes Can Improve Without New Information: An Exact-Null Audit of Uncertainty Beyond Model Outputs
链接: https://arxiv.org/abs/2609.38956
作者: Wenhao Liang,Lin Yue,Wei Emma Zhang,Mingyu Guo,Olaf Maennel,Weitong Chen
类目: Artificial Intelligence (cs.AI)
备注: Preprint. 44 pages
Abstract:Routing signals of modern vision transformers – expert gates, attention-residual weights and halting scores – often improve probes that predict whether the model is correct, and the improvement is commonly read as evidence that routing carries information about errors beyond the model’s outputs. We test this inference directly: keeping real output-routing pairs, we redraw correctness labels from a frozen output-only generator fitted on disjoint data, so that routing is uninformative by construction. Under this exact label null, a width-matched MLP comparison still reports a routing gain in 51.3% of confidence-only evaluations (308/600), while a linear comparison reports none. Holding each training trajectory fixed on a six-model panel and selecting the checkpoint by validation log loss instead of validation accuracy removes the detections (50/120 to 0/120, and 83/120 to 0/120 in an independently implemented probe), identifying accuracy-based checkpoint selection as the cause; across all output views the raw detection rate falls from 27.5% (528/1,920) to zero observed detections. The repaired comparison is not sensitive, detecting an implanted signal of about 0.005 nats in 0/20 replicates in each of two matched settings, whereas a conditional permutation test built on an estimated routing law detects it in 11/20 and 10/20 and rejects rarely under the null. On real correctness labels, the conditional analysis yields model-relative evidence in five DeiT attention-residual families; in four it persists under two specified variants of the conditional law, and no family passes an additional noise criterion. Fitting a better probe and testing for incremental information are different problems, and each needs its own validation.
[AI-132] Loop-Free Inverse Reinforcement Learning via Sequential Value Recovery with Q-Score Matching NEURIPS2026
链接: https://arxiv.org/abs/2609.38955
作者: Yang chen,Yitan Zhang,Michael Witbrock,Shuyue Hu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted to NeurIPS 2026. 20 pages, 7 figures
Abstract:Inverse Reinforcement Learning (IRL) aims to recover a reward function that explains expert demonstrations. Existing IRL methods typically rely on a bi-level optimization procedure that alternates between reward learning and policy optimization, leading to substantial computational burden and training instability. In this work, we introduce a different route that eliminates policy optimization entirely by leveraging diffusion policies. Our key insight is that a diffusion policy encodes the action-gradient structure of the optimal soft Q function, enabling reward learning to be cast as a sequence of value recovery problems, thereby allowing us to bypass reward-policy loops inherent in prior IRL methods. Specifically, our method proceeds in three stages: (I) recovering the optimal soft Q function via action-gradient matching and estimating the corresponding soft value function (LogSumExp of Q values) in a way inspired by Gumbel regression; (II) calibrating these soft values by inferring a state-dependent offset; (III) extracting the reward by enforcing Bellman consistency. This leads to Loop-Free Inverse Reinforcement Learning (LFIRL), a fully offline algorithm that operates in a simple, loop-free, and sequential manner. LFIRL is simple to implement and significantly improves training efficiency while maintaining strong reward recovery performance. Empirically, across Maze, Franka Kitchen, Adroit Hand Pen, and Push-T benchmarks, LFIRL achieves 2-3x speedup over the fastest baselines, while matching or surpassing state-of-the-art methods in reward recovery quality.
[AI-133] APTInvestBench: Evaluating Autonomous APT Investigation under Varying Telemetry
链接: https://arxiv.org/abs/2609.38954
作者: Yu Wang,Shuhao Li,Tao Yin,Ziyang Li,Xueying Zhao,Peishuai Sun,Jiang Xie
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents could help security operations centers (SOCs) investigate advanced persistent threats (APTs) by turning weak leads into evidence for intrusion scoping and response. Yet success under one telemetry setting does not establish robustness to changes in log collection, retention, or sampling. We introduce APTInvestBench, a benchmark for evaluating cross-telemetry robustness in autonomous APT investigation. It comprises 370 cases across seven SOC-inspired conditions, derived from 56 report-informed attack reconstructions with 16.4 million log records. Agents investigate unverified leads and submit reports with record-level citations. Fixed action-level support requirements track sufficient evidence across available logs, query returns, and formal citations, separating telemetry limitations from acquisition and reporting gaps. Across eleven LLMs, agents acquire sufficient evidence for 44.3% of recoverable attack actions on average, while formal citations support only 25.0%. More importantly, aggregate coverage can conceal substantial instability: from Full to endpoint-only telemetry, coverage declines by only 1.6 percentage points, yet 35.5% of previously covered actions lose sufficient citation support despite remaining recoverable. Across four frameworks, such losses persist even when registered supporting records remain unchanged. APTInvestBench provides reusable investigation environments and diagnostic evaluation for identifying these gaps and developing more reliable defensive agents.
[AI-134] DrivingBench: Can Vision-Language Models Drive a Toyota Corolla?
链接: https://arxiv.org/abs/2609.38948
作者: Aditya Ramabadran,Simon Mahns,Tobias Gessler
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Frontier models excel at many digital benchmarks, yet their ability to drive a real car, an everyday human skill, remains largely untested. We present DrivingBench, to our knowledge the first benchmark where general-purpose vision-language models must drive a real car. Through three tools, the models see camera frames from a Toyota Corolla and directly command its steering and velocity around a parking lot cone course at low speeds. The car may continue moving while the model thinks and new commands replace the currently running one, so inference latency is part of the task, testing the models’ abilities to observe, act, monitor, recover, and complete a long-horizon objective under such constraints. We benchmark GPT-6 Astra, Claude Fable 5.1, GPT-5.6 Sol, and Grok 4.6 in vendor-native harnesses (Codex, Claude Code, Cursor) with up to three attempts each in one conversation; Astra is the only model to finish the course, on its second attempt, with no other attempt passing 50% of the course. Two of the four models improved materially across attempts with retained context. We also detail the design principles behind our action interface, and show how the tool output format and the framing of the task combined to determine whether models would drive at all or refuse. We release our harness, prompts, course map, and traces with video and telemetry for reproducibility.
[AI-135] Signal-Routed Temperature Scaling: Low-Capacity Risk-Conditioned Calibration for Small Validation Budgets
链接: https://arxiv.org/abs/2609.38936
作者: Wenhao Liang,Liangwei Nathan Zheng,Lin Yue,Wei Emma Zhang,Mingyu Guo,Olaf Maennel,Weitong Chen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Preprint. 9 pages main text plus appendix (47 pages total)
Abstract:When a classifier is recalibrated from only a few thousand held-out examples, the capacity of the calibration map becomes a statistical design choice rather than a purely architectural one: a scalar map can underfit structured residual miscalibration, while a highly adaptive map can be hard to estimate reliably from so small a split. We disentangle the calibration objective from adaptive capacity and propose signal-routed temperature scaling (SRTS-BCE), a 10-parameter, argmax-preserving calibrator that cross-fits a correctness-risk score over six logit statistics and fits one top-label-BCE temperature per K=3 risk groups, recovering TvA-TS as its K=1 limit. On fine-tuned CIFAR-100 / ViT-B/16, SRTS-BCE reduces \mathrmECE_15 from 1.65 (scalar TvA-TS) to 0.96, matching the higher-capacity SMART+BCE head (0.95) at the full calibration budget. The two regimes separate as the budget shrinks: at n=250 SRTS-BCE beats SMART+BCE on all three CIFAR-100 backbones (the seed-to-draw hierarchical interval excludes zero), whereas the flagship comparison against the scalar remains directional. A protocol-frozen Tiny-ImageNet follow-up reproduces the small-budget separation and exhibits a budget-dependent ranking reversal on Swin-T; matched routing and map controls show that the effect is tied neither to the learned router nor to discrete grouping. Together the results identify post-hoc calibrator capacity as a finite-sample design choice whose preferred level shifts with the amount of available calibration data.
[AI-136] Adaptive Self-Consistency: From Black-Box Sampling to Distribution-Valued Feedback
链接: https://arxiv.org/abs/2609.38931
作者: Jingkai Huang,Yunfan Zhang,Will Ma,Weihua Zhou,Zhengyuan Zhou
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 27 pages, 4 figures, 7 tables. The first two authors contributed equally
Abstract:Self-consistency samples many reasoning trajectories and aggregates their final answers, treating the LLM as a black box that returns one answer per trajectory. Yet the final answer of each trajectory is sampled from a softmax vector that is available from the model’s log-probabilities. We refer to this as the grey-box setting in which each trajectory reveals this answer distribution rather than a single draw from it. We formulate efficient inference in this setting as sequential mode identification with distribution-valued observations: sample trajectories one at a time and stop as soon as the LLM’s modal answer is identified at a prescribed confidence level. We characterize the asymptotic stopping rate of mode identification with distribution-valued observations exactly and show that it is never worse than the black-box rate. We then propose the ASC-D algorithm, a betting stopping rule that attains this asymptotic stopping rate. On MMLU-Redux, ASC-D uses 46.4 – 95.6% fewer trajectories than answer-only adaptive self-consistency baselines and achieves the highest fixed-budget correct-certification rate across three open-source models.
[AI-137] Learning What to Forget: Distributional Unlearning for LLM Representation Spaces
链接: https://arxiv.org/abs/2609.38929
作者: Pinaki Mohanty,Haoran Tang,Maggie Makar,Rajiv Khanna
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Machine learning systems increasingly face the need to remove the influence of entire data domains, such as toxic language, harmful behavior, or topical content, rather than isolated records. Recent work formalizes this problem as \emphdistributional unlearning: selecting a subset of a forget domain whose removal moves the training distribution away from an unwanted population while preserving proximity to the desired one. However, existing analyses often impose parametric assumptions to obtain tractable selection rules. These assumptions may be poorly suited to high-dimensional language-model representations. We introduce \textscMamushi, a framework for non-parametric distributional unlearning that ranks forget examples using a probabilistic classifier whose Bayes-optimal logit equals the forget-to-retain log-density ratio (up to an additive class-prior constant). We show that thresholding the population log-density ratio yields the optimal fixed-budget selection rule for our removal–preservation objective and establish a non-asymptotic transfer guarantee relating score-estimation and threshold-calibration errors to degradation from the population-optimal selection rule. Our empirical evaluation spans real-world datasets on toxic-language removal and topical-domain removal regimes using different representations, with \textscMamushi achieving a more favorable removal–preservation trade-off than other baselines. Our work shows that \textscMamushi can serve as an efficient selection approach for downstream machine unlearning procedures, reducing the number of forget examples required to reach a fixed forgetting target.
[AI-138] Prototype-guided Bilateral Alignment Multimodal Federated Learning ICML2026
链接: https://arxiv.org/abs/2609.38925
作者: Tianchi Liao Tianchi_Liao,Lele Fu,Sheng Huang,Qing Hu,Hong-Ning Dai,Chuan Chen
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 28 pages, 16 figures, ICML 2026 (Spotlight)
Abstract:Multimodal federated learning (MFL) has emerged as a pivotal paradigm for leveraging distributed data to enhance model performance. However, existing methods predominantly rely on idealized assumptions of model homogeneity and balanced modality distributions, rendering them ill-suited for practical scenarios characterized by heterogeneous client architectures and severe modality imbalance. To address these challenges, we propose a \textbfMultimodal \textbfFederated learning Prototype-guided Bilateral Alignment (MFedPBA) framework. MFedPBA facilitates robust knowledge synergy through a dual alignment mechanism: (i) at the feature level, it aligns heterogeneous feature spaces via a projection encoder optimized by contrastive learning and the Gromov-Wasserstein distance; (ii) at the decision level, it employs an entropy-weighted aggregation of naturally aligned logit prototypes. This novel design achieves robust MFL by jointly tackling heterogeneous feature spaces and collectively aggregating decisions. Extensive experiments demonstrate that our method significantly outperforms state-of-the-art baselines under conditions of model heterogeneity and modality imbalance.
[AI-139] How Much Can Reliability Drift Under a Fixed Confidence Distribution?
链接: https://arxiv.org/abs/2609.38917
作者: Wenhao Liang,Lin Yue,Wei Emma Zhang,Mingyu Guo,Olaf Maennel,Weitong Chen
类目: Artificial Intelligence (cs.AI)
备注: Under reviewing
Abstract:A classifier’s conditional accuracy can change while its confidence distribution stays exactly the same. We study the worst-case movement of the reliability relation under covariate shifts that preserve the distribution of the confidence score, constraining the reweighting within each confidence level by a \chi^2 budget; the resulting worst case, as a function of the budget, is a fragility profile. On an interval of budgets that can be computed from the source distribution, the profile equals exactly the square root of the budget times the within-level variance of the correctness propensity – the grouping-loss term of calibration-refinement decompositions. Beyond this interval the profile is governed by the tails of the propensity law, and the entire upward profile determines the centred within-level law; consequently, calibration residual and grouping variance do not determine fragility in general, though they do when labels and predictions are deterministic. Since the propensity is not observed, we restrict reweightings to a learned finite readout within confidence bins, bound the part the restriction misses by the grouping variance remaining inside readout cells, estimate the restricted profile with role-separated labels, and provide a separate split-sample lower confidence bound. On ImageNet this bound is positive in both splits for four of six primary classifiers and nine of twelve additional ones as released, and for three of eighteen after temperature scaling. Held-out drift under optimised reweightings fitted without evaluation labels tracks the estimated profile; an exploratory label-permutation diagnostic yields near-zero agreement for this statistic while largely reproducing the correlation observed for unsigned random reweightings.
[AI-140] Composing Task-specific Agent Harnesses at Test Time with Reusable Primitives
链接: https://arxiv.org/abs/2609.38912
作者: Peng Kuang,Haibo Jin,Dehao Wu,Feiyang Deng,Xiaopeng Yuan,Jerry Wang,Haohan Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Agent harnesses govern how large language models (LLMs) gather context, invoke tools, verify results, preserve state, and terminate, largely affecting agent performance. However, the value of each harness mechanism can differ across heterogeneous tasks: a mechanism that improves one task may impose overhead or context distraction on another, leading to the suboptimality of a global harness. We characterize this suboptimality as a mismatch induced by fixed mechanism choices, motivating task-specific harness construction. Nonetheless, generating harness code for each task introduces generation and debugging costs, with execution risks that can compound as more mechanisms are generated. To address those challenges, we introduce Harness Primitives, reusable harness mechanisms with clear application scope and composition contract mined from failed task trajectories. Based on Harness Primitives, we propose STITCH, a framework that Selects suitable primitives given Task Information and compiles them into Task-speCific Harnesses at test time. This separation enables task-specific harnesses without generating or repairing mechanism code at test time. Extensive experiments demonstrate that STITCH not only improves harness adaptability and robustness, but also scales with the primitive library size, boosting task success rates by up to 12 points over fixed harness baselines, surpassing human-designed harnesses like Codex CLI while maintaining a minimal test-time harness composition overhead of only 2.7%, 638 times more efficient than generating task-specific harnesses from scratch. Ultimately, our work demonstrates that building task-adaptive harnesses can be beneficial for completing diverse tasks and that building reusable primitives can be a promising path towards this goal.
[AI-141] Unlearning Deceptive Behaviors in LLM s with Contrastive Forget Sets
链接: https://arxiv.org/abs/2609.38909
作者: Haoran Tang,Rajiv Khanna
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models often know the truth and say otherwise: a model that answers correctly when asked neutrally will affirm a user’s mistaken belief, or misstate a fact its system prompt wants hidden, once the context rewards it. Such deception is a behavior conditioned on context, not knowledge, yet machine unlearning, the natural tool for removing a behavior from the weights, is built to forget facts that a deceptive model still needs. We propose to unlearn when a model deceives rather than what it knows, with a contrastive forget unit built from the model’s own realized deceptions: the same question under a deception-triggering and a neutral context, admitted only where belief holds and behavior flips. Standard objectives on this unit face a dilemma. Suppression objectives such as NPO leave much of the deception in place. Target-based objectives, which distill the model’s neutral behavior into the pressured context, remove it but induce context blindness: a target generated without the context teaches the model to stop reading it, eroding benign system-prompt instructions, secret-keeping and the reasoning a monitor inspects, a failure invisible to deception rates and capability benchmarks. We introduce PACT, which trains toward pressure-aware counterfactual targets (the model’s own honest response, with a trace that registers the pressure and resists it) while retaining the benign uses of the triggering context. On two 32B reasoning models, PACT reduces held-out deception from over 50% to under 3% while system-prompt adherence, secret-keeping and the reasoning trace stay at the base model’s level. On a tug-of-war score of removal against retention, PACT reaches 0.94 and 0.86, against at most 0.77 and 0.60 for any baseline. Like removed knowledge, removed deception is shallow under relearning, and terms that simulate the attacker hold it only at a cost in context use.
[AI-142] DAMPER: Return-Prioritized Gradient Control for Smooth Policies
链接: https://arxiv.org/abs/2609.38903
作者: Seokmin Ko,Taewon Goo,Kihyuk Hong
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 21 pages, 9 figures, including appendices
Abstract:Actor-critic methods achieve strong performance in continuous control, but their policies can produce highly oscillatory actions. A common remedy is to add auxiliary smoothness losses. However, their contribution can be negligible when their gradients are small relative to the native actor gradient. Moreover, existing methods often combine multiple auxiliary losses, complicating loss balancing without necessarily improving the return-smoothness trade-off. We introduce DAMPER (Direction-Aware Magnitude-Controlled Projection with Explicit Return Priority), which combines the native actor gradient with a temporal-consistency gradient through conflict-conditioned projection and adaptive magnitude control. It removes the auxiliary component opposing the actor gradient and scales the retained temporal direction relative to the actor gradient norm, preserving positive alignment with the native actor gradient. Experiments with TD3 and SAC on six continuous-control tasks show reduced action oscillation relative to the native agents in all 12 task-backbone pairs and the best oscillation score among the compared methods in eight, with task-dependent return trade-offs.
[AI-143] SceneJail: Exploiting Video Scenario Context to Jailbreak Multimodal LLM s
链接: https://arxiv.org/abs/2609.38899
作者: Wenyu Chen,Li Wang,Chuanchao Zang,Xiangtao Meng,Xinyu Gao,Jianing Wang,Zheng Li,Shanqing Guo
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Video Multimodal Large Language Models (Video-MLLMs) support reasoning over video inputs, yet remain vulnerable to jailbreak attacks that elicit policy-violating responses. Existing video jailbreaks primarily manipulate how harmful queries are visually presented, thereby treating video merely as a carrier. Consequently, the surrounding video scenario remains unexplored as a contextual attack surface. In this paper, we show that the same harmful query can elicit different safety responses when placed in different video scenarios. To systematically exploit this vulnerability, we propose SceneJail, an adaptive black-box jailbreak framework with two coordinated components. Adaptive Scenario Construction dynamically searches for a surrounding scenario that is contextually compatible with the harmful query. Scenario-aware Prompt Search uses black-box response feedback to search for textual guidance tailored to the selected scenario. Extensive evaluations on the HADES and SafeBench datasets across eight Video-MLLMs, including two proprietary models, GPT-4.1 and Gemini3.5-Flash, demonstrate the effectiveness of SceneJail. SceneJail-F, which presents the complete query persistently, achieves average attack success rates (ASR) up to 91.5%, outperforming the strongest baselines by 29.1 percentage points. Furthermore, SceneJail-S, which distributes the query across successive frames, remains highly robust against current defenses, retaining a 72.3% ASR even under strict image filtering. Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) Cite as: arXiv:2609.38899 [cs.CR] (or arXiv:2609.38899v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2609.38899 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-144] FFASR: Benchmarking Far-Field Automatic Speech Recognition using High-Fidelity Simulated RIRs
链接: https://arxiv.org/abs/2609.38897
作者: Shivam Saini,Eric Bezzam,Georg Götz,Alessia Milo,Steinar Guðjónsson,Konstantinos Gkanos,Finnur Pind,Daniel Gert Nielsen
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Performance (cs.PF)
备注: 9 Page Technical Report
Abstract:Far-field automatic speech recognition(ASR) degrades under reverberation, noise, and talker motion, yet the benchmarks that drive model selection emphasize close-microphone speech. We present FFASR, a held-out corpus of 15,637 utterances and an open leaderboard spanning nine conditions, each varying a single acoustic factor: anechoic near-field speech, a measured-versus-simulated office-lab pair, static far-field mixtures at high/mid/low signal-to-noise ratio(SNR), and moving-talker variants at matched SNR. Dry speech from 15 talkers is convolved with hybrid wave/geometrical-acoustics room impulse responses from 14 furnished rooms; because the speech is newly recorded and the test waveforms are never released, the corpus resists training-data contamination. Across contemporary systems, mean word error rate (WER) rises from 4.4% near-field to 41.3% in the static low-SNR condition; a moving talker adds a small but consistent penalty at matched SNR; and on the office-lab pair, measured and simulated WER agree to within about 1.7 pp on average. These results support high-fidelity simulation as a scalable proxy for measured far-field evaluation under the conditions we test.
[AI-145] Unmerge: Efficient Machine Unlearning via Task Arithmetic
链接: https://arxiv.org/abs/2609.38895
作者: Haoran Tang,Andrew Tan,Rajiv Khanna
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Approximate machine unlearning seeks to remove the influence of a forget set from a trained model without full retraining. Existing gradient-based methods require data-dependent hyperparameter search, struggle when forget and retain knowledge are entangled, and offer little insight into where unlearning actually happens inside the network. We recast unlearning through the lens of task arithmetic: if finetuning produces a merged task vector \tau_m that combines learning on forget and retain sets, unlearning is the inverse operation that subtracts a learned forget component \tau_F to recover the retain task vector \tau_R . The forget signal is concentrated: at every layer, forget activations lie in a subspace spanned by a handful of dominant directions, so we factorize \tau_F in a low-rank forget basis, which is faithful up to a small tail-eigenvalue residual and limits how far the correction can perturb retain. We then optimize three intuitive goals (match the merged vector inside the forget span, suppress leakage into the retain span, and bound the correction size) that provably bound forget leakage and retain damage in activation space. The resulting algorithm, Unmerge, is fast and powerful: on class-level unlearning with ResNet-50 on CIFAR-100 and Tiny ImageNet, it improves Tug-of-War by up to ~24% over a baseline of comparable runtime and by up to ~18% over stronger baselines that run ~5x slower, keeps membership-inference exposure at the level of retraining, and shrinks the feature-distribution gap to the retrained model, where relabeling methods leave forget features cleanly separable. Further studies show that Unmerge also applies to ViT-S/16 and scales to Llama-3.2-3B. The per-layer basis geometry that drives the algorithm also serves as a layerwise diagnostic for when and where unlearning becomes structurally hard.
[AI-146] Learning Continuous Neural Representation of Stochastic Hybrid Systems
链接: https://arxiv.org/abs/2609.38893
作者: Sangli Teng,Hang Liu,Koushil Sreenath
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注:
Abstract:A stochastic hybrid system (SHS) is governed by a stochastic differential equation (SDE) describing the continuous dynamics and a Markov reset kernel triggered on the guard surface. Its probability evolution can be described by a hybrid Fokker-Planck (HFP) equation with a partial differential term corresponding to the SDE and an integral term arising from the reset kernel. This work shows that such an SHS can be approximated by an SDE in a higher-dimensional latent space where the sample paths are continuous. The key to this result is to encode different branches of the reset kernel using auxiliary variables, transforming the resets into deterministic ones that enable topological gluing. By the embedding theorem, the glued manifold can then be embedded into a higher-dimensional Euclidean space. We show that the probability evolution on the embedded image no longer requires explicit reset terms in the HFP equation. Building on this theorem, we design a loss that matches the evolving state distributions, enabling a single latent SDE to recover the probability evolution of the SHS without mode labeling, trajectory segmentation, or event-based simulations.
[AI-147] Consistent Plan-Act for Long-Horizon Agent ic Tasks
链接: https://arxiv.org/abs/2609.38891
作者: Heng-Zhuang Li,Yi-Kai Zhang,Yu Wang,Yueqing Sun,Jiayuan Zhang,Qi Gu,Han-Jia Ye
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Long-horizon agentic tasks demand strong reasoning and efficient execution across successive interactions with dynamic environments. A common approach decouples high-level planning from low-level execution through separate planner and actor roles. To investigate coordination failures in these tasks, we prompt both agents for structured state assertions and compare their reports programmatically to detect explicit contradictions. Our analyses reveal systematic disagreement about the same task-relevant state facts, a phenomenon we term planner-actor state mismatch. We further find that providing agents with task-relevant state information reduces mismatch and improves coordination and task performance. Based on the systematic analysis of the state mismatch, we propose Consistent Plan-Act (ConPAct), which feeds detected contradictions back to both agents to form consistent state interpretations and fine-tunes them on curated consistent interactions for better coordination. ConPAct improves performance across various environments and model configurations, e.g., increasing MiniGrid success rate from 38.6% to 54.4% with GPT-5.6-sol/terra as planner and actor respectively, demonstrating that state consistency can guide both inference-time correction and coordination training.
[AI-148] Right Answers Costly Models: The Efficiency Gap in LLM -based Optimization Modeling
链接: https://arxiv.org/abs/2609.38884
作者: Zhong Li,Xin Huang,Jinhui Wan,Xiangyi Wang,Shenkai Zhang,Ruiqi Chen,Wenyu Liu,Zaiwen Wen,Ziyan Luo
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Optimization modeling formulates real-world decision problems as mathematical programs that solvers can use to find optimal decisions. Large language models (LLMs) can automate this process, but the resulting correct formulations can require substantial time and memory to construct and solve, limiting practical scalability. Therefore, we systematically investigate whether LLMs can identify problem structure from natural-language descriptions and apply suitable optimization modeling techniques to generate mathematical models and solver code that solve the problems correctly and efficiently. To this end, we first curate OptTips, a knowledge base of 50 expert modeling techniques in eight families. Using this knowledge, we develop OptDachshund, a multi-agent framework that transforms problems from existing optimization benchmarks into new tasks for evaluating LLMs’ use of modeling techniques. It constructs conventional and expert mathematical models with solver code for the same task and data, providing baselines for correctness and computational cost. The resulting EfficientOpt benchmark contains 561 expert-reviewed tasks with paired reference implementations. Evaluation of 11 representative LLMs reveals an efficiency gap on correctly solved tasks with comparable measurements: for every LLM, most generated programs take longer to solve than their expert counterparts. Within the comparable reference-size subset, 57% of programs with correct objective values and fewer variables and linear constraints have longer recorded solver times. Case studies show that different modeling techniques can achieve the same optimal value at similar recorded cost. Faster solving may not reduce execution time if the code takes longer to prepare data and build the model. LLM optimization modeling should therefore be evaluated for both correctness and computational efficiency.
[AI-149] STRATA: Self-Learning Through Role-Aligned Tiered Agents for Real-Time Strategy Games
链接: https://arxiv.org/abs/2609.38881
作者: Xinhe Tian,Xiaoyue Zhang,Ziyou Zhang,Jiacheng Li,Xiaoqiang Jin,Qianchuan Zhao,Gaochen Cui
类目: Artificial Intelligence (cs.AI)
备注: 8 pages, 2 figures
Abstract:Real-time strategy (RTS) games require agents to coordinate economic development, production and construction, base defense, unit organization, and attack timing over long matches. Existing studies have applied large language models to command decision-making in RTS games, enabling agents to read textual game states and generate high-level plans. However, long inference latency can cause them to miss critical tactical events. The complexity and tactical diversity of full RTS matches also leave existing systems heavily dependent on manually written experience-based prompts, with limited ability to learn continuously from past games. We present STRATA, a role-aligned hierarchical system with cross-game self-learning for Red Alert. STRATA assigns in-game strategic, logistical, and tactical decisions to a Strategic Agent (SA), Logistics Agent (LA), and Tactical Agent (TA), respectively. The SA generates high-level directives based on the global game state and relevant experience cards, while the LA and TA handle logistics and tactical execution. After each match, a Review Agent (RA) derives candidate experience from game traces, validates and revises it using evidence from subsequent matches, and compresses strategic experience supported across multiple games into concise experience cards for SA retrieval. We evaluate STRATA through the formation of experience cards, full-match comparisons before and after learning, and experience learning against AI opponents with different play styles. Under a fixed scenario, using the learned experience cards increases the observed win rate from 30% to 100%. Sequential learning against AI opponents with different play styles also produces distinct long-term strategic experience.
[AI-150] Does Learning Protein Folding Generalize to Broader Reasoning ?
链接: https://arxiv.org/abs/2609.38879
作者: Yong Liu,Zhanpeng Shi,Yizhou Dang,Zhongyue Zhang,Xiaoliang Shi,Zhijian Wei,Shuangjia Zheng
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models rely heavily on human text, which often conveys surface answers rather than the spatial and structural logic behind them. Protein folding is a natural testbed, because one solved structure yields thousands of exactly checkable spatial and topological statements. We ask: can learning to fold proteins teach general models reusable reasoning capabilities? To answer this, we build FoldingCorpus, a protein-derived question-answer dataset, and Fold2Reason, a recipe that post-trains on it through two complementary signals: discrete structural answers predicted via the model’s native language head, and continuous 3D geometry decoded from the same shared representations. On FoldBench, Fold2Reason achieves structure prediction scores 2.7 to 3.5 times those of Qwen3.5-9B. Beyond protein structure prediction, it improves performance on all 10 benchmarks spanning spatial, graph, scientific, and general reasoning, raising macro-average accuracy from 45.09% to 48.33% (+3.23 pp), with positive gains on all 10 benchmarks, while matched controls built from random, synthetic, and shuffled structure yield substantially smaller or negative gains. Our work shows that non-linguistic, structure-dense scientific data can systematically improve broad reasoning in language models, making a solved scientific problem a practical source of post-training supervision.
[AI-151] Reasoning Externalization for Faithful Large Language Model Narratives of Stock Return Predictions
链接: https://arxiv.org/abs/2609.38869
作者: Sujung Kim,Seung Hwan Cho,Sangjin Park,Young-Min Kim
类目: Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE)
备注:
Abstract:In finance, interpreting machine learning predictions is essential, yet the numerical outputs of explainable AI can be difficult for non-experts to understand. While large language models (LLMs) can translate these outputs into natural language, they may produce errors when inferring numerical changes and feature relations. We propose an LLM narrative framework for cross-sectional stock return prediction that combines temporal Shapley additive explanations (SHAP) evidence with historical regime analogs. Temporal evidence tracks changes in the normalized global SHAP importance of an XGBoost model over six months. Historical analogs are past periods with similar changes in SHAP importance, their model performance and subsequent market returns are provided as comparative context. Using this framework, we conduct a controlled study of progressive reasoning externalization, sequentially providing raw SHAP sequences, deterministic temporal descriptors, and feature relations. Each generated claim is verified against provenance-linked evidence. Across Qwen3, externalizing numerical and relational reasoning improved evidence faithfulness as well as temporal and relational accuracy. Evidence faithfulness increased from 0.696 to 0.996 for Qwen3-32B-Instruct. While historical analogs did not improve structured automatic faithfulness, they received higher human-rated usefulness scores. These results suggest that externalizing verifiable reasoning enhances narrative faithfulness and that historical context adds interpretive value.
[AI-152] alk2Agent : Benchmarking Voice Interfaces for Text Agents
链接: https://arxiv.org/abs/2609.38867
作者: Terumi Chiba,Guangzhi Sun,Zheqi Yuan,Chao Zhang
类目: Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注:
Abstract:Large language model (LLM) computer-use agents are typically evaluated with clean written instructions, despite speech being an increasingly popular interface for interacting with such systems. Speech input introduces an additional failure point: transcription errors can alter task-critical entities, constraints, or targets before the agent begins reasoning, while conventional ASR metrics do not directly measure whether the information required for successful execution has been preserved. We introduce Talk2Agent, a benchmark for evaluating how effectively voice interfaces convey human-spoken instructions to LLM-based computer-use agents. Talk2Agent builds human-spoken versions of tasks from WildClawBench and OSWorld and evaluates a range of voice interfaces, including dedicated ASR models, audio-capable LLMs, contextual biasing, and LLM-based ontology repair. Because repeatedly executing long-horizon computer-use tasks is costly and stochastic, we further propose an execution-free, task-conditioned evaluation framework that projects the original task grader onto prompt-addressable intentions and measures how much task-relevant information is retained after the voice interface. On WildClawBench, Talk2Agent’s execution-free native projection provides a practical, execution-grounded measure of voice-interface quality, correlating with downstream task completion and improving Pearson correlation by 0.246 over WER/CER on 32 hours of real human speech.
[AI-153] When Context Changes: Understanding Update Failures in LLM s
链接: https://arxiv.org/abs/2609.38866
作者: Junyu Guo,Yuchen Fang,Shangding Gu,Costas Spanos,James Demmel,Javad Lavaei
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As preferences, goals, and facts change, LLM agents must use the current state while earlier versions remain in context. Yet they can answer with an old value of the same variable, a failure that we call stale binding. To study when models use outdated information and why, we introduce Controlled In-Context Memory (CICM), a benchmark for tracking and using updated information in conversations and agent logs. We observe that even frontier reasoning models can fail to recover the current state. We find that in open-source models probes can still recover the updated value when the model answers with an old one, pointing to a failure to select information that remains available. Component tests in Qwen and Pythia identify a mechanism for this selection failure: attention drift, where attention favors old values over the current one when producing an answer. We study a one-layer transformer to mathematically understand how this phenomenon happens: when attention scores are similar, several old values can together receive more attention than the current value. Guided by this explanation, we redirect attention toward the current value without further training. When the current value is requested directly, adjusting this intervention for each input corrects most old-value errors across various model families while preserving nearly all initially correct answers. Reliable context management therefore requires more than remembering updated information: models must use it to guide their answers.
[AI-154] GeoNest: Learning to Select Failure-Aware Neighborhoods for the Irregular Knapsack Problem in a Circular Container
链接: https://arxiv.org/abs/2609.38863
作者: Zhongman Du,Huiming Zhang,Linlin Yang,Sheng Xu,Baochang Zhang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 9 pages, 3 figures
Abstract:The two-dimensional irregular knapsack problem in a fixed circular container is an important combinatorial optimization problem for maximizing material utilization in manufacturing. Conventional geometric packing solvers can produce tightly packed layouts, yet they often partition the residual space into isolated small pockets that cannot fit valuable unplaced polygons. To overcome this late-stage packing bottleneck, we propose a failure-aware large neighborhood search framework named GeoNest, driven by a graph policy trained via reinforcement learning. Specifically, we first construct neighborhoods by pairing failed target polygons with residual pockets. We then use explanatory poses to identify the placed polygons that block candidate insertions. These diagnosed blocking relations define bounded, fixed-item repair subproblems for the underlying geometric solver. Finally, the graph policy selects the most promising subproblem for execution. For evaluation, we introduce CircleNest-Bench, a benchmark comprising 2,391 load-controlled instances from four contour sources, including a held-out industrial CAD source. Experimental results demonstrate that, under the same total time budget, GeoNest improves mean utilization over a state-of-the-art standalone packing solver by about 0.9% on average across the three main test sets and by about 0.6% on the held-out industrial set.
[AI-155] Efficient Multi-Modal Planning with Reward-Guided Preference Optimization for Autonomous Driving
链接: https://arxiv.org/abs/2609.38862
作者: Chenglin Chen,Lujia Wang,Xinhu Zheng,Jun Ma,Haoang Li
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Safe and efficient trajectory planning is essential in autonomous driving. However, existing end-to-end approaches often fall short in both computational efficiency and safety guarantees. Methods based on imitation learning suffer from causal confusion, while rule-based scoring approaches often incur heavy computational overhead and suffer from objective misalignment. Additionally, preference-based methods rely on strict pairwise annotations, limiting data utilization. To overcome these limitations, we propose EMPlan, an efficient multi-modal trajectory planning method powered by reward-guided fine-tuning. We design a hybrid architecture that combines sparse anchors with an offset refinement module for efficient multi-modal trajectory prediction. Sparse anchors provide coarse trajectory candidates with low latency, which are subsequently refined by the offset module for higher prediction accuracy. To enhance safety without incurring additional inference costs, we adopt a two-stage training paradigm consisting of pretraining and reward-guided fine-tuning. During fine-tuning, we leverage rule-based reward signals and unpaired preference supervision to refine the pretrained policy toward safer trajectory selection. We evaluate EMPlan on the non-reactive NAVSIM benchmark, where it strikes a favorable balance between planning accuracy and efficiency, demonstrating superior performance under real-time constraints.
[AI-156] Optimal Design for Active Preference Learning with Biased LLM Judges
链接: https://arxiv.org/abs/2609.38860
作者: Zhongman Du,Huiming Zhang,Haodong Zhu,Baochang Zhang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 35 pages, 5 figures
Abstract:Learning from human preferences is central to large language model (LLM) alignment, but human preference annotation is costly. Active preference learning reduces this cost by selecting informative comparisons, and LLM judges can provide additional scalable feedback. However, the preferences of the judges may deviate from those of the target human population. Even after calibration on trusted reference data, active acquisition can shift the comparison distribution and expose residual judge bias. We therefore incorporate judge deviations into the acquisition design rather than relying on a separate calibration stage. Under joint estimation, comparisons that appear highly informative about the reward may also reflect judge bias and therefore provide less information about human preferences. To address this issue, we propose Nuisance-Adjusted Optimal Design (NAOD), a comparison-selection strategy that prioritizes policy-relevant target information after nuisance adjustment and uses the Frank-Wolfe algorithm for optimization. Theoretically, we establish a sharp conditional local asymptotic minimax lower bound on policy risk and construct an estimator that attains it. We further characterize the finite-sample cost of learning the nuisance representation and show that representation error can reverse an oracle design advantage. Finally, we validate these predictions experimentally and evaluate NAOD on Chatbot Arena data across 17 judges, 15 budget configurations, and 15 random cluster-level splits. NAOD reduces the mean regret of proxy policy by 29.1% relative to a matched target-information design, outperforms existing methods, and improves human-preference prediction on held-out data.
[AI-157] Online Evolution Strategy for Flow-Matching VLA Policies via Self-Supervised Trajectory Distribution Optimization
链接: https://arxiv.org/abs/2609.38855
作者: Gongxin Yao,Yongsheng Zhao,Jiayin Deng,Deng Liang,Han Gao,Lei Zhao,Baoping Cheng
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注:
Abstract:Vision-Language-Action (VLA) models based on generative frameworks, such as Flow Matching, have recently achieved impressive performance in robotic manipulation. Unlike deterministic policies, Flow Matching enables VLA models to learn conditional action trajectory distributions, where latent noise vectors induce different actions under the same task scenario. However, we observe that these distributions are often ill-formed, with successful and failed behaviors coexisting while considerable probability mass remains in unfavorable regions. To this end, we propose Online-ES, an online adaptation framework for Flow Matching VLAs based on Evolution Strategy (ES), which refines the learned action trajectory distribution through interaction feedback. Instead of pruning the latent noise space, our method performs evolutionary exploration directly in the action trajectory space, where diverse trajectories generated by Flow Matching provide candidate solutions for adaptation. By perturbing sampled trajectories and evaluating their execution outcomes, we derive a self-supervised MSE objective that transfers the evolution direction from trajectory space into model parameter space. Mathematically, we prove that the proposed objective provides an unbiased estimator of the optimal evolution direction. Moreover, we also incorporate failure experiences as negative feedback to regularize the evolution direction, steering the policy away from previously explored failure regions. Experiments in both simulation and real-world environments demonstrate that Online-ES achieves policy improvement comparable to reinforcement fine-tuning, without learning a value model or computing advantages.
[AI-158] Mitigating the Length-Scaling Tax with Online Distillation
链接: https://arxiv.org/abs/2609.38854
作者: Xu Wan,Wenyue Xu,Shengjie Zhao,Mingyang Sun
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Length scaling during reinforcement-learning (RL) post-training is often viewed as a sign of improved reasoning ability, especially on difficult problems, but may also make responses to already-solved problems unnecessarily verbose. We quantify this side effect as the length-scaling tax (LST): excess response length on already-solved queries without a commensurate accuracy gain. To mitigate LST, we propose Length Self-Distillation (LSD), which routes solved prompts to on-policy distillation and retains the original RL objective for unsolved prompts. LSD uses an exponential moving average of the online policy as its teacher, requiring no external model. We find that LSD achieves comparable or better performance than RL across multiple variants, while substantially curbing response-length growth on easy queries. LSD reduces LST from 19.0% to -3.7% on single-turn reasoning and from 31.4% to 13.7% on multi-turn agentic tasks, demonstrating that LSD effectively preserves concise response patterns on easy queries while supporting efficient exploration on difficult queries during RL post-training.
[AI-159] Scoring Higher Answering Worse: Mitigating Reward Hacking in Rubric-Based RL via Protocol-Level Rubrics
链接: https://arxiv.org/abs/2609.38847
作者: Maoqi Liu,Junwei He,Bowen Zhang,Feiran Li,Wentao Ma,Rongyi Lin,Shuhan Zhong,Quan Fang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Under Review
Abstract:Rubric-based reinforcement learning (Rubric-RL) trains language models where no verifier exists. A judge checks each criterion of a rubric, and the verdicts are aggregated into a reward, most often by a weighted sum. We show that this additive aggregation is the weak point. Under a sum, criteria compensate for one another: a policy that misses the one decision that matters can buy the points back with advice nobody asked for. On clinical consultation, such a policy scores higher and answers worse. Rubric coverage rises while appropriateness on held-out physician criteria falls below the untrained model. The medical criteria are not to blame. Grouped so that they must hold together, the same criteria, unchanged to the word, recover a third of the loss; shorter answers recover almost none. We therefore propose Protocol-level Rubrics (ProRubric), which keeps what the criteria ask for and changes how they are aggregated. It groups a checklist into a few protocol-level dimensions. A dimension counts only when all of its criteria hold and its failure clause does not fire. The grouping is done once, offline, and leaves the optimizer unchanged. ProRubric raises appropriateness by 10.8 points without losing coverage and has the best seven-benchmark average at both scales. Reward validity is set not only by what a rubric verifies, but by how it aggregates. Code is available at this https URL
[AI-160] scTrilemma: Balancing Identity Invariance and Fidelity in Single-Cell Representation Learning NEURIPS2026
链接: https://arxiv.org/abs/2609.38840
作者: Yunhak Oh,Yoonho Lee,Junseok Lee,Namkyeong Lee,Sang-Yeon Hwang,Yinhua Piao,Hyomin Kim,Seonghwan Kim,Jaechang Lim,Woo Youn Kim,Sungsoo Ahn,Chanyoung Park
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: NeurIPS 2026
Abstract:Single-cell RNA-seq representation learning is fundamentally label-free: cell identities, states, and contexts are not fixed training targets, so what constitutes signal or nuisance is analysis-dependent. A single representation must therefore preserve biological identity and state, remain robust to nuisance context, and retain the gene-level variation needed for expression analysis, three demands we call the representation trilemma. To tackle this problem, we introduce scTrilemma, a latent-bottleneck VAE that routes expression-derived variation to the embedding, the decoder, or the prior rather than forcing all of it through one embedding. It gates gene tokens by expression, routes the cell representation through the decoder, and conditions the prior on unlabeled pseudo-bulk context, under a single reconstruction objective and without target annotations or auxiliary representation losses. In release-based zero-shot evaluation on successive CZ CELLxGENE Census releases, scTrilemma leads all three demands at once and preserves biological-state, differential-expression, and pathway structure across multiple disease settings. Latent interventions further show that context can be removed at almost no cost to the other demands, leaving identity against fidelity as the remaining tension. Code is publicly available at this https URL.
[AI-161] Diversity Combining for Multi-Path LLM Reasoning NEURIPS2026
链接: https://arxiv.org/abs/2609.38829
作者: Guangsheng Yu,Litianyi Zhang,Qin Wang,Xu Wang,Mingyuan Li,Shaoxiong Ji,Ren Ping Liu,Massimo Piccardi
类目: Artificial Intelligence (cs.AI)
备注: Accepted by NeurIPS 2026
Abstract:Multi-path reasoning methods such as self-consistency (SC) sample K reasoning paths and choose the most frequent answer. However, their gains quickly plateau as K increases, and existing methods do not predict when this saturation will occur. We formalize multi-path LLM reasoning as a diversity combining problem from wireless communications: each path is a noisy channel observation, and the pairwise correlation of path correctness caps the design-effect effective sample size of the vote at a finite ceiling. Generalized least squares (GLS) analysis shows that, under exchangeability, the optimal symmetric linear combiner of latent embeddings is uniform, supporting majority vote as the natural default in standard SC while leaving room for weighting or pruning under heterogeneous prompt-template branches. Across 5 models and 12 benchmarks, prompt-template diversity reduces path correlation in 55 of 57 valid cells, with the strongest effect on open-ended QA. We derive an Adaptive-K rule that uses a four-path pilot to select K^* , retaining 96 – 103% of MV@ K=32 accuracy across Math, QA, and NLU.
[AI-162] More Choices Fewer Decisions: Ordinal-Scale Bias in JEV-like Direct-Decision Models
链接: https://arxiv.org/abs/2609.38827
作者: Tianxiang Gao,Jinzhe Li,Zhiyuan Li,Yi Chang,Yuan Wu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Direct-decision models turn text into low-latency structured labels and scores, making them attractive for classification and automatic evaluation. Yet reliability requires more than accuracy: a model must also use the ordinal decision scale supplied by the user faithfully. We analyze JEV~1.13 and three open KEV models. Our investigation begins with ANLI, where JEV assigns 38.8% of all predictions and 51.3% of errors to Neutral despite 74.95% accuracy, nearly balanced gold labels, and balanced candidate positions. Across 36 ordinal datasets, final decisions use only 67–76% of the effective gold support, versus 87–102% on four nominal tasks. Randomizing candidate order weakens but does not remove this compression. Holding items and source scores fixed while balancing gold support and positions, we refine scales from K=2 to 14 ; utilization falls for every model and reaches 26–75% at K=14 , although candidate probabilities remain broad for most models. Targeted BA-LoRA post-training raises gold-relative utilization from roughly 47% to 86% on eight supervised scales at both KEV sizes, showing that the compression is learned and modifiable rather than an immutable architectural limit. We call this ordinal scale-utilization bias: decision-stage candidate-space compression distinct from accuracy, gold imbalance, fixed position, and candidate count alone. The code and data are available at this https URL
[AI-163] Explicit Trajectory Diversity for RL-Based Post-Training of LLM Agents
链接: https://arxiv.org/abs/2609.38805
作者: Huaiyu Fu,Heng Cao,Hao Wang,Jian Ya,Tao Chen
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:LLM agents often admit multiple high-quality solutions to the same task, differing in reasoning structure, tool-use pattern, or interaction trajectory. Yet existing notions of diversity in LLM post-training are mostly implicit, arising from general stochasticity and regularization mechanisms rather than explicitly targeting task-relevant behavioral variation. While such implicit diversity can be useful, it does not directly specify which forms of behavioral variation should be encouraged for a given task. In this work, we study explicit trajectory diversity in RL-based post-training for LLMs. Our key idea is to define diversity through user-specified, task-specific trajectory descriptors, which map each sampled trajectory to an interpretable behavioral representation, and then measure diversity as a set-level functional over the resulting descriptor matrix. Building on this formulation, we introduce Trajectory-guided Joint Policy Optimization(TJPO), a single-policy framework that optimizes explicit diversity over sampled trajectory groups, avoiding the need for population-based policy training, and instantiate it within group-based policy optimization through trajectory-level learning signals. This design makes the diversity objective both interpretable and controllable. Experiments on Sokoban and ALFWorld show that TJPO improves task-specific trajectory diversity while maintaining competitive task performance. Descriptor and trajectory analyses show that the learned variation follows the specified behavioral dimensions and includes distinct successful strategies. Extra experiment results suggest that explicitly shaping trajectory diversity can help LLM agents satisfy user requirements and remain effective when task conditions change.
[AI-164] GraphCert: Bootstrap Agent ic Graph Reasoning with Certified Evidence Rubrics
链接: https://arxiv.org/abs/2609.38798
作者: Weiqi Jiang,Yuchen Ying,Rui Wang,Kaixuan Chen,Bingde Hu,Shunyu Liu,Yu Wang,Tongya Zheng
类目: Artificial Intelligence (cs.AI)
备注: Under review
Abstract:Graph agents extend large language models (LLMs) with the ability to actively explore and reason over knowledge graphs through multi-step interactions with graph tools. However, training capable graph agents typically requires large collections of question-answer pairs and reasoning trajectories, whose manual construction is costly and difficult to scale. Moreover, employing proprietary LLMs to generate such supervision further risks exposing sensitive graph data to external services. Therefore, we propose GraphCert to bootstrap agentic graph reasoning with certified evidence rubrics during post-training. Specifically, the Bootstrapped Graph Quizzer guided by generation controls produces graph-grounded QA pairs and marks supporting evidence, which undergo execution certification and semantic curation. The accepted evidence is then canonicalized into certified evidence rubrics that later reward Graph Solver evidence alignment alongside answer correctness during GRPO training. Experiments on five graph reasoning domains in GRBENCH demonstrate that GraphCert consistently outperforms substantially larger LLM agents and post-training method. Furthermore, our analysis demonstrates that the learned policy transfers robustly across heterogeneous graph domains, suggesting that GraphCert acquires reusable graph-reasoning capabilities rather than domain-specific patterns. These results establish executable self-certification as an effective approach to self-training compact graph reasoning agents. Our code will be made publicly available.
[AI-165] ChartDensity-Bench: Benchmarking MLLM s for Numerical Data Reconstruction under Visual Density
链接: https://arxiv.org/abs/2609.38781
作者: Xinhe Wu,Yadong Jin
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Multimodal large language models (MLLMs) offer a promising approach for recovering numerical data from scientific charts, but their ability to reconstruct chart data from visually dense figures remains poorly understood. Existing chart understanding benchmarks primarily evaluate question answering or chart-level reasoning and provide limited support for evaluating structured numerical reconstruction from scientific figures. We introduce \textbfChartDensity-Bench, a benchmark for evaluating MLLMs on structured numerical data reconstruction from compound chart figures under controlled visual density. Built from charts paired with source-level ground-truth data, ChartDensity-Bench systematically varies the number of simultaneously presented charts ( k\in1,3,6,9 ), enabling controlled evaluation of density-induced degradation. We further propose a multi-dimensional evaluation framework covering structural reliability, reconstruction completeness, parseability, and numerical fidelity. Experiments on five recent MLLMs show that numerical reconstruction generally degrades as visual density increases, while the magnitude of degradation varies substantially across models. Chart-level paired comparisons further show that the same source chart can incur higher reconstruction error when embedded in denser visual contexts. These findings highlight visual density as an important and previously underexplored factor in MLLM chart data reconstruction and provide a systematic benchmark for evaluating model robustness in this setting.
[AI-166] RAST: Resolution-Aware Privileged Structure Transfer for Low-Resolution Audio Activity Recognition
链接: https://arxiv.org/abs/2609.38780
作者: Ji Hwan Park,Gautham Krishna Gudur,Yufei Shen,Dawei Liang,Edison Thomaz
类目: ound (cs.SD); Artificial Intelligence (cs.AI)
备注:
Abstract:Audio is increasingly used for human activity recognition (HAR) because it captures object interactions, environmental events, and contextual cues in everyday environments. High-resolution (HR) audio provides rich acoustic information for model development but incurs substantial energy and storage costs and may expose sensitive speech content. Low-resolution (LR) audio offers a more privacy-preserving and resource-efficient alternative for deployment, but reduced sampling rates can remove acoustic cues essential for activity recognition, leading to significant performance degradation. We formulate this training-deployment mismatch as sensor-resolution privileged learning, in which HR audio is available during training, while inference relies exclusively on LR audio. We propose RAST, a resolution-aware transfer framework that compresses HR teacher representations by preserving token-level information and neighborhood structure before performing localized HR-LR alignment. Experiments on the SAMoSA and AudioIMU datasets show that RAST consistently outperforms LR-only training and direct teacher-transfer baselines, improving LR-only recognition by up to approximately 7.8% while requiring only LR audio at inference.
[AI-167] Action Conditioned Bisimulation For GUI Agent Memory
链接: https://arxiv.org/abs/2609.38778
作者: Hongbo Zhang,Liuyang Song,Quanquan Li,Daqian Yang,Yan Wen,Zhengtao Yao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:An agent that remembers what it did on a web page must decide when two pages count as the same. Memories built on observation similarity merge pages that look alike but behave differently, and GUIs are full of such pages: two tabs of one widget or two rows of one menu answer the same click differently. We define the merge rule as an action-conditioned bisimulation over the empirical predictive state graph a frozen agent fills as it acts. Two states merge only when their shared actions lead to agreeing outcomes and successor blocks under an affordance label. Observation similarity never enters the rule, and nothing is trained. It replaces the merge rule of an existing outcome-value memory, so a closed-loop comparison isolates it. On MiniWoB++ it raises success rate over a memoryless agent, while a control taking identical exploratory detours, the prior successor-representation merge, and the same criterion without action conditioning change nothing.
[AI-168] Distilling Diffusion Score Discrepancy for Efficient Training Data Attribution
链接: https://arxiv.org/abs/2609.38776
作者: Shixuan Liu,Joan Serrà,Kin Wai Cheuk,Jinju Kim,Woosung Choi,Yukara Ikemiya,Wei-Hsiang Liao,Jiaqi W. Ma,Yuki Mitsufuji
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Training data attribution for diffusion models aims to identify the training samples that influence a generated instance, but existing methods either require costly per-sample gradient computation or query-specific model optimization. Moreover, most methods attribute changes in a proxy loss rather than changes in the actual model’s generative behavior. We address these limitations by formulating attribution directly with a local score discrepancy measure, which applies to any diffusion variant (including DDPM, EDM, and flow matching), and by showing that such measure can be estimated without retraining, as a preconditioned gradient similarity. We instantiate this estimator as Training-data Influence via score Discrepancy (TID), which uses Kronecker-factored curvature to avoid random projections and per-sample gradient storage. We then distill TID into TIDE, a forward-only student trained online to reproduce the teacher’s rankings from the diffusion model’s internal activations. Under counterfactual evaluation on CIFAR-10, ArtBench-10, and MS-COCO, TID matches or outperforms state-of-the-art approaches, while TIDE retains most of TID’s accuracy at four to five orders of magnitude lower per-query cost, attributing generated samples in milliseconds and faster than the generation itself.
[AI-169] Learning Under Forgetting: Statistical Support-Selective Retention in Stochastic Training Dynamics
链接: https://arxiv.org/abs/2609.38768
作者: Fujie Gao,Zuyue Zhang,Gang Sun
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 19 pages, 3 figures
Abstract:Prior work has shown that neural networks exhibit implicit biases toward low-complexity structure (e.g., spectral bias), memorization dynamics, and compression-like effects during training, but a unified dynamical account of selective retention remains incomplete. We propose Repeated Reinforcement with Persistent Forgetting (RPF) dynamics, a minimal framework in which repeated exposure reinforces patterns and structures that recur in the data, while persistent forgetting attenuates learned information. This view treats forgetting not merely as a failure mode, but as a selection mechanism. We build the theory in three successive layers. First, in an independent-feature model, we derive an exposure-selective survival law and a support-dependent retention boundary characterizing which patterns persist under forgetting. Second, in a shared-parameter model, we show that forgetting induces spectral filtering over covariance modes, preserving strongly supported shared components while suppressing weak ones. Third, under small-step and norm/coding approximations, we show how RPF dynamics induce an implicit trade-off between data fitting and the cost of stored information, yielding Minimum Description Length (MDL)-like compression. Controlled experiments provide evidence for this reinforcement–forgetting selection mechanism in scalar memories and a nonlinear shared network. Joint reinforcement and attenuation interventions shift conditional retention, while matched exposure counts reveal forgetting-dependent effects of reinforcement timing and changes in the composition of the retained set. Together, these results show that repeated reinforcement and persistent forgetting jointly provide a controllable source of inductive bias beyond neural architecture and scale.
[AI-170] dattri-LLM : A Unified and Efficient Library for Training Data Attribution at LLM Scale
链接: https://arxiv.org/abs/2609.38767
作者: Shixuan Liu,Tongli Zhou,Junwei Deng,Pingbang Hu,Jiaqi W. Ma
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Training data attribution (TDA) estimates the contribution of individual training examples to model outputs. Most scalable TDA methods rely on per-example gradients, whose computation and use at LLM scale pose challenges in efficiency, compatibility, and extensibility. We introduce dattri-LLM, a TDA library that makes gradient-based attribution more practical at scale. For efficiency, dattri-LLM uses compact gradient representations and dynamically routes gradient operations based on a cost model. For compatibility, its capture mechanism collects per-example gradients from existing training loops that call backward(), without requiring changes to the loop or its configuration. This includes distributed training with DDP and FSDP and pipelines built with HuggingFace Transformers, TRL, and OLMo. For extensibility, dattri-LLM exposes reusable gradient operations and training-time callbacks for implementing attribution methods and applications. These interfaces support a variety of attribution methods, including gradient similarity, curvature-based influence, and trajectory-based methods, as well as applications that act on gradients during training, such as online data selection. On the same hardware and workload, dattri-LLM achieves 3.2x the throughput of the fastest competing library on average, scales multiple attribution methods to 110B-parameter models across four H200 GPUs, and offers superior attribution fidelity-cost trade-offs across a range of models with different model families and scales.
[AI-171] PathAnchor: Path-Structured Evidence for Scientific Agents
链接: https://arxiv.org/abs/2609.38766
作者: Qiuhui Chen,Jiafan Lu,Shuaimin Tang,Tao Dai,Suyuan Wang,Chenrui Ji,Zhenglei Zhou,Weimin Zhong
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Scientific agents can retrieve relevant passages yet still lose functional order, mix evidence across sources, or state conclusions that exceed the retrieved record. We introduce PathAnchor, a bounded scientific reasoning system built on path-structured evidence workspaces. Instead of treating passages or extracted concepts as independent units, the system retrieves source-linked Material-Sensor-Signal-System trajectories that preserve role, direction, and the evidence supporting each transition. A controller uses three read-only tools to search paper-specific trajectories, trace paths across candidate sources, and open exact evidence before producing a claim-cited answer and an explicit evidence boundary. On 120 single- and cross-paper flexible-sensor questions, PathAnchor scores 82.6% and leads six evaluated systems. Under a matched controller, corpus, and six-call budget, replacing unordered concept graphs with path-structured records raises source recall from 61.3% to 82.9%, increases answers whose claims all cite opened evidence from 69.2% to 90.0%, and reduces tool calls. These results show that evidence organization affects retrieval and citation completeness under fixed agent resources.
[AI-172] Adaptive-GEPA: Make Your Harness Fit Heterogeneous Requests
链接: https://arxiv.org/abs/2609.38762
作者: Tianyu Chen,Yasi Zhang,Ruiyi Wang,Xinran Zhao,Taoran Li,Mingyuan Zhou
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:Reflective optimizers such as GEPA improve language model prompts from execution traces and evaluator feedback; full-program extensions can also rewrite tools and control flow. In practice, a user hands the same endpoint heterogeneous requests whose effective solutions require different tools, reasoning modes, and control flow. Optimizing one shared program leaves this division of work implicit in source-code search, while optimizing a separate program per request family fixes it beforehand. We introduce Adaptive-GEPA, which learns both how to divide requests and how to solve them. It evolves a router and a library of specialist programs under one search budget. The router’s instructions, each specialist’s description, and its program code are plain, human-readable text, edited from feedback. To combine branches, it aligns specialists by the requests they handle and inherits descriptions together with programs. On a fixed mixture of four task families, the reported Qwen3-8B run evolves four experts without supplying family labels to the router or reflection model; its routing matches the task partition on all 651 test requests. Its family-mean test score (x100) rises from 52.6 to 70.6, compared with 62.5 for GEPA’s full-program adapter and 54.0 for GRPO at a nominal budget of 18,000 scored calls. These counts do not equate total compute. Figure 1 summarizes the learning curves, final test scores, and routing agreement. Subjects: Software Engineering (cs.SE); Artificial Intelligence (cs.AI) Cite as: arXiv:2609.38762 [cs.SE] (or arXiv:2609.38762v1 [cs.SE] for this version) https://doi.org/10.48550/arXiv.2609.38762 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-173] Self-Evolving Algorithm-Design Agents : Escaping In-Context Evolutionary Stagnation via Population-Curated Policy Optimization
链接: https://arxiv.org/abs/2609.38757
作者: Chen Lu,Ke Xue,Siyuan Xu,Mingxuan Yuan,Chao Qian
类目: Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Large language models are increasingly participating in complex real-world tasks in the form of algorithm-design agents, designing and refining algorithms. Many successful algorithm-design agents adopt pure in-context evolutionary frameworks, but they may quickly plateau in domains that require specialized knowledge. Parametric adaptation offers a way to internalize specialized knowledge, but conventional training requires abundant domain-specific corpora while high-quality algorithms are scarce in complex algorithm-design scenarios. In this paper, we propose sample-efficient parametric self-evolution where agents can explore and learn from self-generated algorithms. First, we characterize in-context evolutionary stagnation and analytically propose the Improvement Chain proposition, showing how learning successive self-generated algorithms can locally increase the likelihood of neighboring algorithms. Motivated by this local-transfer perspective, we further propose Population-Curated Policy Optimization (PCPO) to utilize a global population and a hybrid policy update scheme for retaining and reusing high-quality, diverse self-generated algorithms, shifting the policy towards stronger algorithms. In the task of learning rate schedule design for global placement in electronic design automation, trained only on 4 chip cases, PCPO outperforms the state-of-the-art in-context evolutionary methods (e.g., OpenEvolve and ShinkaEvolve) on average across 16 chip cases. With an 8B-size base model, PCPO achieves competitive performance compared to frontier closed-source models such as GPT-5.5. PCPO also reduces inference-time token cost by internalizing grounded domain knowledge and prompt distillation. Moreover, PCPO achieves significant speedups on four GPU kernel designs, with an average of 8.27 \times speedup against the PyTorch Eager baseline.
[AI-174] Code to Control: Synthesizing Parameterized Reactive Controllers
链接: https://arxiv.org/abs/2609.38733
作者: Zergham Ahmed,Joshua B. Tenenbaum,Chris Bates,Samuel J. Gershman
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 17 pages. Code: this https URL
Abstract:Recent LLM-based approaches to control either invoke a language model to select actions or synthesize world models that require planning at every decision, introducing latency that can limit real-time use. We introduce Code to Control, an approach that synthesizes Python controllers which execute directly as policies. Code to Control separates program structure from parameters. An LLM synthesizes the controller structure, while derivative-free search fits its parameters for continuous control using feedback from the environment. Once learned, the resulting controllers require neither LLM inference nor planning at decision time, enabling real-time gameplay and, under our timing protocol, faster action selection than a PPO policy. Across a suite of Atari games, Flappy Bird, and MuJoCo tasks, Code to Control outperforms planning-based program synthesis methods, remains competitive with deep reinforcement learning while using fewer environment interactions, transfers across substantial changes in environment dynamics, and scales to complex locomotion tasks.
[AI-175] Staying on Task: Testing the Foundations of Long-Horizon Agent Reliability
链接: https://arxiv.org/abs/2609.38712
作者: Jeffrey Willette,Krishna C. Puvvada,Boris Ginsburg
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Long-horizon agentic workflows require models to sustain repeated state-dependent actions all while the context grows, sub-task complexity changes, and new data arrives. Each situation represents an independent axis along which an agent may fail. An agent reconciling a long ledger, for example, must repeatedly read its state, update the correct record, and preserve alignment across thousands of outputs. A model may accept the entire ledger yet lose its place or stop applying the operation consistently as generation proceeds. We introduce Long-Transduction, a controlled diagnostic that tests a model’s ability to stay on task during long generation while continuously reading, mutating, and outputting input-context dependent operations such as arithmetic, sorting, variable lookups, and table transformations. Long-Transduction evaluation independently varies local task complexity, input data formatting, and context length isolate failures along each axis. We evaluate seven open-weight models, finding a 62.8% decrease when scaling context length from 4-128K, a 36.5% decrease when varying input format, and a 39.9% decrease by increasing local task complexity. Together, these failures represent critical liabilities in long-horizon agentic workflows.
[AI-176] Budget Boundary Effects in Test-Time Mathematical Reasoning NEURIPS2026
链接: https://arxiv.org/abs/2609.38699
作者: Guilin Zhang,Ziqi Tan,Wulan Guo,Kai Zhao,Hongyun Yang,Mei Luo,Qi Ning,Feng Yang
类目: Artificial Intelligence (cs.AI)
备注: Accepted as a poster at the 6th Workshop on Mathematical Reasoning and AI (MATH-AI), NeurIPS 2026. 11 pages, 3 figures, 8 tables. Includes additional post-acceptance accounting and selection diagnostics
Abstract:A cumulative token cap can fall inside a mathematical derivation, forcing a test-time controller to choose between stopping at the cap (strict) and allowing the current attempt to finish (advisory). We measure this boundary choice with paired offline replays of 19,200 public traces: 120 AIME, BrUMO and HMMT problems and two archive configurations of one model. Candidate order and a 16-attempt cap are fixed, and answer selection is blind to reference answers and correctness labels. Three findings emerge. First, at the 4k cap, most advisory accuracy gains replace abstention with a correct answer; strict stopping pays for an unfinished prefix that the completed-only selector cannot use. Second, comparisons along realized cost differ from same-cap comparisons: advisory 4k in low has higher accuracy than strict 8k at comparable mean completion cost, while in high its observed accuracy is 0.42 points below strict 32k using 59% of its mean tokens. These aggregate comparisons do not establish equal-compute superiority or accuracy equivalence. Third, increased candidate coverage does not guarantee higher answer accuracy: a log-probability selector loses accuracy while coverage rises, including after a source-grade consistency repair. Same-cap majority-accuracy differences shrink below 1.3 percentage points at 32k. Budget curves should jointly state the cap, realized cost, eligible candidates, stopping rule and selector information.
[AI-177] Cascadia: A Control-Plane-Free Alternative to Hyperconverged AI Infrastructure
链接: https://arxiv.org/abs/2609.38697
作者: Matias Parij,Pawan Paudel,Tate Berenbaum,Muthaiah Venkatachalam
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI)
备注: 26 pages
Abstract:We present Cascadia, a system for serving large language models on fleets of commodity Intel AIPCs using their CPU, integrated-GPU, and NPU resources. Every node embeds ingress, scheduling, and execution; inference requests require no dedicated routing control plane. Nodes join a libp2p QUIC mesh using CA-issued ed25519 admission certificates, gossip signed capabilities, exchange live load over direct peer streams, and route OpenAI-compatible requests to eligible peers. An operator-run certificate authority handles admission and fleet management outside the inference path. Three serving modes share one interface: whole-model execution on one node, load-balanced replicas, and pipeline-sharded chains using the compilation and speculative decoding mechanism of our companion paper. Optional KV-cache mobility reuses compatible conversation prefixes after a routing move, with cold recomputation on a miss. Signed response receipts and hash-chained logs support provenance and audit. A three-node Phi-3.5-mini NPU testbed delivered 3.10x the response throughput of its one-node configuration under ten concurrent requests; a separate four-node deployment recorded 4.06x the throughput of direct single-node serving. Paired latency observations, runtime measurements, and internal functional checks characterize the tested configurations. We compare Cascadia with IBM, Nutanix, VMware, and HPE platforms on deployment footprint, hardware requirements, scheduling, scaling, licensing, and trust, using vendor documentation. The paper repository provides benchmark scripts, curated measurements, and a claim-to-evidence map.
[AI-178] GATE-ST: Gene-Aware Text-image Encoder for Spatial Transcriptomics
链接: https://arxiv.org/abs/2609.38690
作者: Lucas Ni,Jian Luo,Wentao Huang,Chao Chen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Spatial transcriptomics enables spatially resolved gene expression analysis from slide-level images while preserving morphological features, providing valuable information for studying disease mechanisms and developing treatments. However, spatial gene expression profiling typically requires expensive and time-consuming tests. While existing image-based prediction optimizations mostly revolve around including positional embeddings and further image-based changes, text-based optimizations remain relatively unexplored. We present GATE-ST, which incorporates text-based inputs into image-based spatial gene expression predictions. With this approach, generated text descriptions of genes are utilized to better spatial transcriptomics prediction results. Gene summaries are put through a text encoder, generating embeddings that integrate with image embeddings through cross-attention layers to align with morphological features. We demonstrate the effectiveness of such text inputs by benchmarking performance against random gene embeddings and multiple other image-text fusion architectures, and show that GATE-ST outperforms these alternatives. Our results demonstrate the effectiveness of GATE-ST in pathology imaging, which may greatly reduce the time and cost of accurate spatial transcriptomic predictions, proving the potential of text-guided spatial gene expression prediction.
[AI-179] Concept-Grounded Attention: A Controlled Evaluation of Graph-Injected Attention Temporal Versioning and Epistemic Status
链接: https://arxiv.org/abs/2609.38684
作者: Sachin Dev Duggal,Pradyumna Swarnalatha Ramanna,Alexandros Vassiliades
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Knowledge-intensive language-model systems typically represent external knowledge as text chunks or static graphs, with limited support for concept evolution, point-in-time reasoning, and distinctions between validated and inferred knowledge. We introduce the Concept Lifecycle Model (CLM), which represents concepts as persistent, graph-grounded, temporally versioned entities with explicit provenance and epistemic status, and Concept-Grounded Attention (CGA), which injects concept-graph structure into transformer computation through graph-biased self-attention (Form A) and gated cross-attention over concept nodes (Form B). We evaluate the framework in controlled settings using disabled-mechanism baselines. On 200 MuSiQue and HotpotQA questions with retrieval fixed, concept-graph retrieval recovers explicit multi-hop paths but does not improve evidence recall. Form A appears to steer attention, with 2.76 times more attention on gold than distractor concepts, but the same ratio occurs when Form A is disabled; the learned bias is negligible and no answers change. An identity-preserving Form B improves F1 from 0.188 to 0.221, but control concepts yield 0.213, indicating that most of the gain reflects added capacity. On LongMemEval, explicit temporal representation improves answer accuracy by 13 to 25 points across all tested generators, up to 122B parameters, while simplified CLM version resolution performs similarly to dated serialization because concept identity is not established reliably. On a synthetic source-independence task, protocol-derived epistemic status reduces unsupported assertions from 28% to 0.1% in a fine-tuned small model and from 19-68% to 0-5% in 72-122B models. Overall, the results support making temporal validity and epistemic status explicit, while showing that graph-attention diagnostics are not informative without disabled-mechanism controls.
[AI-180] Provable Test-Time Scaling for Beam Search in LLM Reasoning NEURIPS2026
链接: https://arxiv.org/abs/2609.38672
作者: Qijia He,Yu Huang,Yuan Cheng,Yuxin Chen,Yingbin Liang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted to NeurIPS 2026
Abstract:Beam-search-based test-time methods provide an effective way to improve large language model (LLM) performance on long-horizon generation by pruning invalid reasoning paths early, leading to significantly improved reasoning efficiency and more favorable test-time cost scaling. Despite strong empirical success, the theoretical understanding of beam search remains limited. In this paper, we study the test-time compute guarantee of the commonly used beam search framework that uses the model’s internal log-likelihood for intermediate scoring, while relying on an external reward model only after a complete response is generated. We first establish a lower bound for vanilla beam search, showing that at least \Omega(C^\star(x)^2) samples are required for the optimal response to survive, where C^\star(x) is the token-level coverage coefficient for prompt x . This motivates our modified confidence-filtered beam search (CF-Beam), which reduces the sufficient coverage dependence from quadratic to nearly linear under prefix competitiveness, for fixed horizon, gap, and target accuracy. We then show that the regret of CF-Beam is upper-bounded by the probability of rare failure events and the reward estimation error scaled by a path-level coverage coefficient, where the rare-failure term vanishes as per-step sampling increases. Our results highlight a fundamental advantage of beam search over sequence-level inference methods such as Best-of-N and Best-of-Majority. While the guarantees of these approaches typically involve coverage coefficients that grow exponentially with the horizon L , CF-Beam controls the dominant search-induced term through a token-level coverage coefficient that scales polynomially with L . Our numerical experiments further confirm that beam search is more robust on hard instances and under increasing reasoning horizons.
[AI-181] Where Scientific Search Agents Fail: Decision-Checkpoint Auditing of Exposure and Inspection Attempts
链接: https://arxiv.org/abs/2609.38670
作者: Hongmin Li,Wanli Zhao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Final-answer accuracy does not reveal whether a scientific-search agent failed to encounter a target paper, attempt to inspect it, or return an accepted answer after inspection. We introduce decision checkpoints that record observations and tool actions without benchmark labels during inference, then join target identities and evaluator labels to assign outcome categories from recorded events. Across five conditions on 540 answerable AutoResearchBench Deep questions in a fixed, target-enriched environment, keyword search achieves 24.6% accuracy, compared with 17.8% for raw search. The keyword condition has fewer incorrect answers with neither target exposure nor inspection, but more incorrect answers after the target is exposed and left uninspected. Compared with keyword search, read-first has 27.4% more recorded evidence-search calls. Target inspection attempts occur on 199 questions under read-first and 191 under keyword search; both conditions achieve 24.6% accuracy. The checkpoint protocol makes these question-level differences explicit, distinguishing target exposure and inspection from aggregate accuracy and total tool use.
[AI-182] Understanding Off- vs On-Policy Distillation: A Tale of Distinct Training Objectives
链接: https://arxiv.org/abs/2609.38666
作者: Qiwei Di,Xuheng Li,Kaixuan Ji,Chenggong Zhang,Heyang Zhao,Quanquan Gu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: 68 pages, 4 figures
Abstract:On-policy distillation (OPD) learns from teacher feedback on student-generated responses and has shown promise in reducing forgetting relative to supervised fine-tuning (SFT). However, its benefits and fragility remain incompletely understood. We study sequential distillation from multiple teachers, where the student minimizes its average divergence from the teachers. Forward Kullback–Leibler (KL) divergence yields a weighted arithmetic mixture, while reverse KL yields a normalized weighted geometric aggregate. We develop algorithms that learn these targets under off-policy and on-policy feedback, respectively, establishing logarithmic regret bounds in the tabular setting and extending the analysis to function approximation. By analyzing these aggregation targets, we identify mechanisms that help explain both the benefits and fragility of OPD. Relative to forward KL, reverse KL can better retain a confident expert’s preferences under uninformative feedback, but is more sensitive to teachers that assign very low probabilities to correct responses. Its token-level conditionals also reveal a dependence on continuation distributions that can favor incorrect prefixes over long horizons.
[AI-183] EvoSteer: Online Self-Evolving Graph Orchestration via Reference-Anchored Credit Assignment
链接: https://arxiv.org/abs/2609.38661
作者: Mingda Zhang,Hanwen Zhang,Qiang Huang,Zijia Wang,Pengfei Guo,Yuchen Zhang,Jionghao Zhu,Xiaoying Tang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:In recent years, LLM-based multi-agent systems have been widely applied to orchestrate tool-using agents into executable communication graphs. However, existing self-evolving orchestration still faces key challenges, including post-hoc evolution that revises the team only after the trajectory ends, credit diffusion that gives every action the same terminal advantage under confounded baselines, and skill admission that is uncalibrated and never retired. To address these challenges, we propose EvoSteer, a new paradigm of Online Self-Evolving Graph Orchestration – the orchestrator builds a running team and repairs its plausible but failing steps from execution features and a learned value estimate. To support this paradigm, we introduce Anchored Trajectory Balance (AnchorTB), a regression-style flow-matching loss that assigns each orchestration action a coefficient by balancing subtrajectories against a frozen reference. Built on the learned flow, we further propose Validated Skill Admission, in which a candidate skill is tried before promotion and promoted only if paired evidence passes a sequential test under a shared nominal testing budget. Moreover, AnchorTB combines measured task-level reference reward statistics with prefix-dependent corrections. Experimental results on twelve datasets show that EvoSteer significantly outperforms baselines across question answering, mathematical reasoning, code generation, and interactive decision making. Our code is available at this https URL.
[AI-184] ERRA: Terrain-Aware Reconstruction Retargeting and Control for Musculoskeletal Locomotion
链接: https://arxiv.org/abs/2609.38653
作者: Merkourios Simos,Chengkun Li,Bianca Ziliotto,Alexander Mathis
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
备注:
Abstract:Recent advances in musculoskeletal modeling and reinforcement learning have enabled muscle-actuated agents to reproduce increasingly complex human motions. Yet these capabilities remain largely confined to flat ground, in part because motion datasets rarely include aligned terrain geometry and because retargeting terrain interactions to complex musculoskeletal bodies is challenging. We present TERRA, an end-to-end pipeline for terrain-aware retargeting and control of musculoskeletal locomotion. From kinematic trajectories alone, TERRA combines terrain priors, estimated contacts, and negative free-space evidence to recover task-relevant support geometry. TERRA further considers anatomical, tendon-continuity, and contact constraints during retargeting. Using the resulting motion-terrain pairs from five datasets, we successfully train a single muscle-actuated control policy on 9.4 hours of diverse locomotion. Across reconstruction, retargeting, and held-out tracking benchmarks, TERRA improves terrain accuracy, sharply reduces anatomical and interaction violations, and achieves the highest observed completion rate over supported terrain families. Overall, TERRA provides a practical route from scene-less motion data to muscle-actuated locomotion over diverse non-flat terrain. Project website: this https URL
[AI-185] AgBench: Agent ic AI Benchmarks for Personal AI Devices
链接: https://arxiv.org/abs/2609.38652
作者: Yizhou Han,Di Wu,Dhananjay Saikumar,Blesson Varghese
类目: Artificial Intelligence (cs.AI); Performance (cs.PF)
备注: 15 pages, 12 figures, including supplementary material
Abstract:Agentic AI systems increasingly rely on cloud-hosted large language models for planning, tool use, and iterative execution, raising concerns about API cost and data exposure. Advances in personal AI devices enable agents to execute locally, but limited resources on device may affect task success and performance. Existing benchmarks are inadequate for systematically characterizing these trade-offs across devices, workloads, and deployment architectures. We present AgBench, a benchmark suite and open artifacts for reproducible evaluation of agentic AI on personal devices. Using AgBench, we evaluate local, hybrid, and cloud execution across agentic workloads, examining task success, latency, cloud API cost, and data exposure. Our results, drawn from over 162.07 million data points, show that personal AI devices can complete many agent tasks locally, but local-only execution generally has lower task success and longer completion times than cloud-only execution, especially as concurrency increases. Local-only execution eliminates cloud model API costs and sensitive-information exposure to cloud agents. Hybrid execution can improve task success, but its cloud cost and data exposure depend on how agents divide work and share information. No single architecture performs best across task success, goodput, cloud cost, and data exposure; deployment choices should reflect the intended workload and device capabilities. AgBench is available at this https URL.
[AI-186] Alignment via Training Against Probes Without Losing Monitorability
链接: https://arxiv.org/abs/2609.38645
作者: Lena Libon,Alexander Panfilov,Ben Rank,Xin Chen,Jonas Geiping,Maksym Andriushchenko
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: 38 pages, 22 figures
Abstract:Models are usually aligned based on their observed outputs, using demonstrations, preference data, or reward signals. These objectives reward responses that look aligned. More capable models may learn to satisfy them without internalizing the intended behavior, for example by faking compliance during training. Such superficial compliance could be harder when the objective is defined on model internals rather than outputs. Therefore, we study probe-guided fine-tuning, using probes that detect undesired properties in model activations as a direct training signal. We evaluate linear and non-linear probes with different numbers of probes per layer across two alignment objectives: harmlessness and honesty. We find that training against probes that do not update during training is an easily exploitable objective, while continuously updated probes substantially reduce harmfulness and improve honesty while preserving utility. Probe-guided fine-tuning achieves better safety-utility trade-offs than DPO and inference-time steering, while being substantially more robust against jailbreak and abliteration attacks. Moreover, the concepts stay linearly encoded after fine-tuning, meaning oversight is not lost by our method. Training against probes thus offers a way to shape what models represent rather than only what they output, which may become increasingly important as models get better at making their outputs look aligned.
[AI-187] Interpretable but Frag ile? Robustness of Concept Bottlenecks under Geometric-Semantic Perturbations NEURIPS2026
链接: https://arxiv.org/abs/2609.38625
作者: Hanwei Zhang,Tianma Hu,Gaojie Jin,Xu Cheng,Ronghui Mu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: accepted by NeurIPS 2026
Abstract:Concept Bottleneck Models (CBMs) are designed to provide interpretable intermediate representations, yet how such bottlenecks affect robustness remains unclear, with existing studies reporting mixed and sometimes contradictory findings. We argue that these discrepancies arise from conflating different robustness notions and perturbation regimes, rather than from fundamental disagreements about CBMs themselves. To disentangle these factors, we introduce a generator-based evaluation framework that enables controlled comparisons between standard classifiers and CBMs under two distinct perturbation types: continuous geometric perturbations in latent space and discrete semantic interventions in concept space. Within this framework, we evaluate robustness both empirically, via prediction and concept-level sensitivity metrics, and certifiably, using randomized smoothing in latent and concept spaces. Across experiments, we reconcile previously conflicting findings by clarifying when, and in what sense, concept bottlenecks do or do not improve robustness. By further analyzing robustness under varying task conditions, including class semantic similarity and concept vocabulary size, we show that interpretability does not inherently confer robustness. Instead, concept bottlenecks shift where and how sensitivity manifests, revealing a nuanced interpretability robustness trade off that depends critically on the perturbation regime and task structure. Together, our results show that interpretability and robustness are distinct objectives: interpretable intermediate representations do not uniformly improve robustness, but instead redistribute sensitivity across perturbation spaces and model families.
[AI-188] Differentiable Structure Learning for Cyclic Linear Gaussian Models with Latent Confounders
链接: https://arxiv.org/abs/2609.38618
作者: Sadegh Khorasani,Ali Najar,Saber Salehkaleybar,Negar Kiyavash
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 42 pages, including appendices
Abstract:We study causal structure learning from observational data in linear Gaussian structural causal models in the presence of directed cycles and an unknown number of exogenous latent confounders, bounded by a given maximum. We derive the covariance of the observed variables and introduce marginal quasi-equivalence, which characterizes when different causal models share a full-dimensional subset of the observational distributions they can generate. We formulate structure learning as minimization of the Gaussian negative log-likelihood with a logarithmically scaled complexity penalty that counts directed edges and latent variables. For a fixed number of observed variables and a fixed upper bound on latent variables, we establish consistency of global score minimizers up to marginal quasi-equivalence under algebraic faithfulness, structural minimality, and model-overlap assumptions. We parameterize the inclusion of directed edges and candidate latent variables using Bernoulli gates, whose continuous probabilities are optimized jointly with the structural coefficients. Averaging the penalized negative log-likelihood over these gates yields an objective with a closed-form differentiable complexity penalty. We prove that this expected objective has the same global infimum as the corresponding discrete structure-learning objective. Experimental results show that our approach achieves lower recovery error than previous methods in several experimental settings.
[AI-189] Sense and Sensitivity: Benchmarking LLM Clinical Triage Recommendations with Physician Experts EMNLP2026
链接: https://arxiv.org/abs/2609.38600
作者: Abinitha Gourabathina,Haoran Zhang,Yuexing Hao,Walter Gerych,Marzyeh Ghassemi
类目: Artificial Intelligence (cs.AI)
备注: Accepted to Findings EMNLP 2026 (Findings)
Abstract:As large language models (LLMs) are increasingly used in clinical settings, it is critical to evaluate their reliability under realistic variation in clinical text. We study this question in clinical triage, comparing LLMs to practicing physicians under text perturbations that preserve the underlying clinical setting. We introduce a benchmark of over 6,000 clinical scenarios, 7,000 physician annotations, and 225,000 model responses. Using this benchmark, we make two key observations. First, LLMs are more likely than physicians to recommend unnecessary care at baseline, and this tendency increases under perturbed inputs. Further, we find that LLM recommendations are more sensitive to gender and tone perturbations than human recommendations. Together, these results demonstrate that LLMs can vary under clinically irrelevant textual changes, highlighting the need for deployment-oriented evaluations grounded in expert physician behavior.
[AI-190] FlexRouter: Learning Complementary Model Sets for Flexible LLM Routing
链接: https://arxiv.org/abs/2609.38585
作者: Wang Wei,Harry Yang,Tiankai Yang,Samyadeep Basu,Hongjie Chen,Andy Zhao,Franck Dernoncourt,Ryan A. Rossi,Hoda Eldardiry
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 21 pages, 4 figures, accepted at COLM 2026
Abstract:Existing Large Language Model (LLM) routing methods score LLMs independently to select top- k models. However, this ignores model correlations and enforces a rigid computational budget. Consequently, routers often select redundant models that share failure modes, limiting the overall probability of success. To address this, we propose FlexRouter, a routing framework that explicitly models model complementarity. FlexRouter optimizes for \textitanswer coverage, maximizing the probability that at least one selected model yields a correct response. This objective aligns with practical inference pipelines where multiple candidate outputs are generated and a downstream verifier or user selects the final one. We formulate routing as a coverage-oriented subset selection problem and model the routing policy using Determinantal Point Processes (DPPs), which naturally capture both model competence and redundancy. To directly optimize coverage without requiring a ground-truth target subset, we introduce a training objective based on marginalizing over failure sets. During inference, we employ a greedy strategy based on marginal log-determinant gains, enabling the router to adaptively determine subset sizes without a predefined budget. Extensive experiments on the large-scale RouterEval benchmark demonstrate that our proposed FlexRouter achieves higher coverage with lower redundancy across both in-domain and out-of-domain tasks than strong baselines while maintaining flexible inference cost.
[AI-191] Conditional Generation of Creative Chess Puzzles with Diffusion Models
链接: https://arxiv.org/abs/2609.38577
作者: Aatu Selkee,Severi Rissanen,Xidong Feng,Tom Zahavy,Eric Malmi
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:While modern language models demonstrate impressive generative capabilities, they often struggle with constrained, counter-intuitive creative tasks. To address this limitation, we explore chess puzzle generation as a rigorous testbed for computational creativity and reasoning, a domain where altering a single piece can invalidate an entire solution. We propose a novel approach for conditional generation of creative chess puzzles using masked diffusion models. Unlike previous methods, our non-directional diffusion approach allows for conditioning on specific tactical themes and partial board positions. We introduce a novel auxiliary task of simultaneous best-move prediction, which improves solution uniqueness by 11.6% and theme-conditioning accuracy by 2.5%. To further optimize solution uniqueness and theme conditioning, we establish a reinforcement learning framework adapted from Denoising Diffusion Policy Optimization (DDPO). This RL training increases the yield of unique and theme-matching positions by 89.1%. Finally, we release the first open-weights models (Appendix B) for chess puzzle generation, offering a new pathway for controllable, creative generation.
[AI-192] Defining and Categorising Human-AI Interactions in Clinical Trials: A Multidimensional Human-AI Classification Approach
链接: https://arxiv.org/abs/2609.38559
作者: Sandra Woolley,Tim Collins,Khalid Khattak,Illia Chernomorets,Ariane Arevalo,Chris Richardson
类目: Artificial Intelligence (cs.AI)
备注: 16 pages
Abstract:This paper examines human-AI interactions (HAIIs) in clinical trials and presents a multidimensional categorisation framework that classifies interactions according to AI tasks, human-AI relationships, interaction configurations and interacting human groups. We define HAII, examine existing taxonomies and extend existing categorisation approaches through this novel multidimensional framework. We purposively sampled 15 clinical trials from a previously reported dataset. Each trial was independently categorised by two human reviewers and six large language model (LLM) classifiers. The proposed categorisation provides a structured method for the consistent identification, comparison and synthesis of human-AI interactions across clinical-trial records. The framework is intended to support more consistent comparison and synthesis of AI-related clinical trials and to make explicit the different forms of human involvement associated with AI interventions. The results demonstrate the potential for LLM-assisted categorisation while indicating the continuing importance of human judgement where trial records are incomplete or ambiguous. The principal contribution is a proposed multidimensional framework that brings together AI tasks, human-AI relationships, interaction configurations and interacting human groups within a single approach designed for clinical-trial records. Its significance lies in its potential to support more systematic identification, comparison and synthesis of how humans and AI interact in clinical trials.
[AI-193] Demographic Pluralism: Inference-Time Modeling of Pluralistic Human Preference Distributions
链接: https://arxiv.org/abs/2609.38555
作者: Meng-Chen Wu,Qipin Chen,Ansh Jain,Tess Wood,Zhe Du,Si-Chi Chin
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注:
Abstract:Large language models (LLMs) are increasingly used in culturally sensitive settings, where alignment requires representing diverse preferences within populations. Yet existing methods model populations at coarse demographic or community levels and overlook within-group variation. We introduce Demographic Pluralism, an inference-time framework that estimates population-level opinion distributions without opinion-distribution training data or task-specific fine-tuning by generating multiple perspectives within demographically grounded groups. Across four backbones on GlobalOpinionQA and VITAL, it reduces Jensen-Shannon distance by 8.4%-26.4% over Modular Pluralism. Among weighted, equal-weighted, and inverse-weighted aggregation, equal weighting performs best overall; group-level error also increases with group weight, helping explain weighted aggregation’s weaker performance.
[AI-194] owards Universal Wasserstein Barycenters through Flow Matching
链接: https://arxiv.org/abs/2609.38547
作者: Eduardo Fernandes Montesuma
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: Under review
Abstract:Defining a weighted mean over probability measures under probability metrics is a central tool in probabilistic machine learning. Under the Wasserstein metric, these are called \emphWasserstein barycenters. While most approaches compute barycenters for a fixed weight vector, approximating the whole family of barycenters over the simplex, which we call the \emphWasserstein simplex, remains underexplored. We refer to this problem as \emphUniversal Barycenter Approximation, and propose \textttBaryFM, a flow matching model transporting the marginal measures into any barycenter in the Wasserstein simplex. Once trained, the network can draw samples from measures in the Wasserstein simplex through an ordinary differential equation. We validate our method on 4 downstream tasks: domain adaptation, generalization, Bayesian posterior aggregation and algorithmic fairness. \textttBaryFM achieves the best average rank among 15 competing methods across 10 domain adaptation benchmarks, matching or surpassing non-universal solvers.
[AI-195] ReLaG: A Scalable Framework Generalizing Random Splits to Data with Latent Relations
链接: https://arxiv.org/abs/2609.38538
作者: Anthony Lavertu,Jacob Cote,Sophie Gobeil,Jacques Corbeil,Isabeau Premont-Schwarz,Pascal Germain
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Random splitting can yield non-independent train–test subsets when a dataset contains related samples, as is common in certain applications such as biochemical studies. This leads to overly optimistic generalization estimates. Here, we introduce ReLaG, a modality-agnostic framework that models sample relatedness through a hierarchical latent-variable process and infers groups of related samples using proximity graphs and community detection to produce independent train–test subsets. Across molecular and protein datasets, ReLaG matches existing relation-aware methods while scaling substantially better, enabling splits at previously impractical dataset sizes. We further introduce a label-free procedure that adapts the splitting resolution to production data, aligning evaluation with the intended deployment setting. ReLaG’s inferred groups provide a cheap estimate of effective dataset size, enabling diversity-aware dataset scaling. ReLaG is open source and can be installed with pip install relag.
[AI-196] Does This Action Still Explain the Task? Reverse Scoring for Diffusion Language Model Agents
链接: https://arxiv.org/abs/2609.38536
作者: Jiacheng Qiu,Christopher E. Mower,Jan Peters,Haitham Bou-Ammar,Matthieu Zimmer
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Diffusion-based large language models (dLLMs) promise to break the sequential latency bottleneck of autoregressive agents through parallel decoding, but recent evaluations show this efficiency does not transfer to embodied agentic competence: dLLM-backed agents repeatedly fall into retry loops, re-issuing an action long after it has failed. We give a mechanistic account of this failure and a training-free remedy. We trace the retry loop to the adaptivity of masked decoding: the sampler commits the positions it is most confident about and defers the uncertain ones, and at a failure state the context already offers a confident fill for the deferred decision, i.e. the failed action itself, so the retry is committed without the failure feedback ever being confronted. We model the resulting distortion of the action distribution as a task-blind corruption: contextually salient actions (e.g., the action just taken) receive inflated probability by a factor that depends on the state and the action but not on the task. Under this model, we analyse an invariance proposition: the task-blind factor cancels exactly from the reverse conditional, i.e. the likelihood of the task given the state and a candidate action, which coincides with the task posterior of an idealized uncorrupted model. Masked dLLMs evaluate the reverse conditional natively, unlike autoregressive models, by masking the task tokens and denoising, at the cost of a few parallel passes per candidate. We instantiate the rule as Reflect Reverse and evaluate it on four multi-turn embodied benchmarks, where it improves task success and progression rates over forward-scoring baselines.
[AI-197] Revisiting scaling laws for reward optimization
链接: https://arxiv.org/abs/2609.38526
作者: Ali Aouad,Aymane El Gadarri,Vivek F. Farias
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Scaling laws for optimization against reward models in AI alignment have pinned down how performance depends on optimization effort—measured by a KL-divergence budget relative to a reference policy. Beyond a certain budget, over-optimization (or reward hacking) can arise: because we optimize against a proxy reward model (distinct from true rewards), performance can plateau or degrade. Naturally, the proxy reward’s accuracy depends on how much preference data (often in the form of pairwise comparisons) was used to train it. However, existing research does not cleanly identify how performance jointly scales with the amount of training data and the divergence budget. Our main contribution is to provide an empirically accurate and theoretically grounded scaling law in such context. Performance roughly scales as \Theta(\sqrt\min\log(M),K) , where M is the number of comparisons in training data and K is the policy’s divergence budget. We develop an information-theoretic model to establish this upper bound and prove it is tightly achievable through a constructive procedure. Informed by this, we conduct extensive empirical evaluations using a real-world annotation setup, whereby a large 70B gold reward model generates feedback data and proxy reward models are trained from less capable models (0.6B to 4B). Our scaling law provides an excellent fit (R2 from 97% to 99%), outperforms alternative specifications, and remains robust across model sizes, noise, and optimization procedures (best-of- N or policy tilting). Our evidence suggests that reward optimization is analogous to a surprisingly simple selection task: choosing from a sequence of IID Gaussian random variables using noisy preference feedback.
[AI-198] MM-FinEval: A Multi-Task Multimodal Benchmark for Real-World Financial Forecasting
链接: https://arxiv.org/abs/2609.38523
作者: Dong Shu,Yanguang Liu,Huopu Zhang,Saisai Hu,Haiyan Zhao,Hekun Huang,Mengnan Du
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Financial forecasting from earnings conference calls requires models to reason over complex corporate disclosures, market expectations, and subtle communication signals. However, existing financial benchmarks are often limited to unimodal inputs or single-task settings, making it difficult to evaluate whether multimodal large language models (LLMs) can support real-world financial analysis. In this paper, we introduce MM-FinEval, a novel benchmark designed to evaluate multimodal LLMs across multiple financial tasks. MM-FinEval spans a diverse timeline from 2019 to 2022. The entire proposed dataset contains 2,045 S\P 500 conference earning calls as inputs and 12 financial task labels as outputs. Each input contains three modalities: a word-to-word text transcript of the earning call, the corresponding presentation slides used during the call, and the entire audio recording. To establish a rigorous evaluation framework, we analyze 19 baseline models across three distinct model categories: Image-Text, Audio-Text, and Any-to-Any configurations. We observe that small-size Any-to-Any models processing all three modalities achieve strong performance, even when compared against larger proprietary models restricted to two-modality inputs. This indicates that our tri-modal dataset design introduces useful, non-redundant information. These results validate that text, audio, and visual data serve as important, complementary signals that mimic the decision-making process of expert human analysts.
[AI-199] ShamAN-Q: Shampoo Augmented NanoQuant for Sub-1-bit LLM Weights
链接: https://arxiv.org/abs/2609.38521
作者: Jonathan Mei,Sang Hyub Kim,Oliver Knitter,Chi Chen,Martin Roetteler
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:We introduce ShamAN-Q, a sub-1-bit post-training quantization method that extends NanoQuant by replacing each its diagonal reconstruction geometry with a tractable dense curvature metric, using a general paradigm popularized by the Shampoo optimizer. For each linear weight, ShamAN-Q fits a Kronecker product to the empirical Fisher information matrix of a small calibration set by Kullback–Leibler minimization, forming a Mahalanobis reconstruction loss from the result. The continuous ADMM updates from NanoQuant become solutions to Sylvester equations, while its discrete projection and deployment format remain unchanged. Because the curvature is local to a given set of weights, ShamAN-Q re-measures the input curvature statistic for each layer immediately before layer factorization, periodically refreshing all statistics on the partially quantized model. ShamAN-Q also redistributes the uniform rank from NanoQuant across layers at the same total number of bits. On Qwen3-Base, ShamAN-Q lowers WikiText-2 perplexity at \approx 1 bpw from 27.56 to 22.96 (0.6B), 19.21 to 16.72 (1.7B), and 14.29 to 13.80 (4B) while matching or improving zero-shot accuracy on the Eleuther LM Evaluation Harness. On 0.6B, ShamAN-Q at \approx 0.8 bpw matches the published perplexity of NanoQuant at \approx 1.0 bpw.
[AI-200] VAmoS Part Deux: Harder More Realistic Voice-Agent Simulation
链接: https://arxiv.org/abs/2609.38512
作者: Joshua Meyer,Sahar Shayegan,Ritiz Tambi,Ali Khan,Sun Kim,Victor Shih,Mehdi Jamei,Andi Partovi
类目: Artificial Intelligence (cs.AI)
备注: 13 pages, 3 figures, 7 tables. Agent implementations: this https URL
Abstract:Voice agents in production must handle several requests, background speech, and customers who lose patience. We introduce VAmoS Energy, a benchmark that combines these challenges in 100 calls about utility billing and payment assistance. Each caller makes two to four requests. The agent has sixteen tools backed by a stateful Stripe billing twin and the Apache Fineract loan engine, with account access blocked until caller verification succeeds. The tasks use public household electricity data and a policy based on Pennsylvania’s residential billing rules. An LLM-as-a-verifier checks the agent’s actions and spoken figures against explicit requirements. On a calibration run, it agrees with a code verifier on 99.1% of checks. Across fourteen voice stacks and three repeats per task, completion ranges from 17.3% to 44.7%. Grok Voice leads, and Gemini 3.8 Live and GPT-Live follow at about the same cost per call. Background television reduces pooled completion from 38.7% to 8.6%. The simulated caller often accepts an incorrect result because it hears the agent’s words but cannot inspect its actions. These findings show why voice agents need evaluation across the whole call, including what they say, what they change, and how they handle competing speech.
[AI-201] An Empirical Study of Architectural Shift from Traditional to AI-Enabled Simulink Controllers
链接: https://arxiv.org/abs/2609.38504
作者: Hadiza Umar Yusuf,Khouloud Gaaloul
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注: Accepted at the 33rd Asia-Pacific Software Engineering Conference (APSEC 2026)
Abstract:Effective AI adoption in cyber-physical systems (CPS) depends on embedding design knowledge into engineering practice. Yet as AI-enabled components increasingly replace analytically derived control laws, this occurs without a systematic understanding of how controller architectures differ or remain similar across paradigms. We address this gap with an empirical study of traditional and AI-enabled Simulink controllers, guided by a literature-derived taxonomy of ten structural categories and nine functional roles. The study analyzes 62 real-world models spanning 8 controller types and 10 application domains, and surveys 13 practitioners, identifying three architectural tensions. First, subsystem organization dominates all controller structures regardless of paradigm, occupying 68-72% of controller footprint, while core control logic occupies minimal space. Second, AI-enabled controllers rely heavily on discrete dynamics and user-defined abstraction, categories largely absent from AI literature, exposing a gap between described and implemented architectures. Third, constraint enforcement blocks largely disappear from AI-enabled models despite practitioner expectations. This reveals a misalignment where safety mechanisms shift from explicit structure to implicit training-time artifacts, breaking traceability.
[AI-202] Security-Enhanced Seed-Based Weight Quantization for Large Language Models
链接: https://arxiv.org/abs/2609.38477
作者: Qiuyu Ren,Sudipta Paria,Aritra Dasgupta,Swarup Bhunia
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Large language models (LLMs) incur substantial storage, memory-bandwidth and energy costs, motivating compact weight representations. Existing seed-based compression methods reconstruct weights from compact pseudo-random representations but do not explicitly account for the non-uniform sensitivity of model weights. We introduce Seed-Q, a security-enhanced sensitivity-aware seed-based weight compression framework that uses lightweight Linear Feedback Shift Register (LFSR)-based weight generation with non-uniform bit allocation. Our approach assigns larger representation budgets to sensitive weights while aggressively compressing less sensitive regions. Importantly, this non-uniform allocation requires no side-information: the decoder deterministically reconstructs the bit-allocation schedule, with no rung depending on the decoded weights, eliminating the need to store per-block metadata or use calibration data while preserving the baseline coding rate. Experiments across diverse LLMs show that Seed-Q matches 4-bit perplexity of SeedLM with fewer bits, while at the same 4 bits/weight it reduces both perplexity degradation and zero-shot accuracy loss relative to SeedLM. We also show that Seed-Q simultaneously achieves high security against bit-flip attacks on model parameters, as bit corruption affects multiple reconstructed weights, greatly amplifying its impact and making it easier to detect. We further implement Seed-Q in an ASIC-based accelerator and demonstrate modest hardware overhead compared to prior seed-based approaches.
[AI-203] Derandomizing Dense Binary Hypervector Codebooks for Quantized Scalars
链接: https://arxiv.org/abs/2609.38471
作者: Dmitri Rachkovskij,Evgeny Osipov,Olexander Volkov,Denis Kleyko,Vaclav Snasel
类目: Information Theory (cs.IT); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注: Accepted, NeCo
Abstract:Hyperdimensional computing and vector symbolic architectures often represent quantized scalar levels by dense binary codebooks whose level-to-level similarity is intended to follow a prescribed function of scalar separation. At finite dimensionality, randomized scalar codebook constructions deviate from this target because of sampling noise, random-start imbalance, update-count fluctuations, component dependence, and finite-capacity effects. We develop a transition-based derandomization framework for dense binary scalar codebooks across two target-similarity families, with similarity decaying exponentially or linearly with level separation. The framework separates the target similarity law, the derandomization variant, and the concrete generator construction, making explicit how initialization, selection, update, and capacity-handling mechanisms shape the induced similarity profile. We formalize derandomization variants that separately constrain initial Hamming weight, update-count variability, and update balance, thereby controlling distinct sources of finite-dimensional error. For each family and variant, we derive the induced mean similarity, identify realization-wise and mean target-matching regimes, and derive exact finite-dimensional expressions for bias, variance, and root-mean-square error. Simulations across dimensions, quantization ranges, reference scalar levels, and generator constructions validate the theory and show how each constraint removes or reduces a specific source of similarity mismatch. The results provide practical guidance for choosing scalar codebook generators that more closely match a desired similarity law under finite-dimensional and hardware-relevant constraints.
[AI-204] NAQD Env: A benchmark for selective withdrawal in language agents
链接: https://arxiv.org/abs/2609.38460
作者: Mohamed Abouzahra
类目: Artificial Intelligence (cs.AI)
备注: 16 pages. Code and evaluation artifacts: this https URL
Abstract:Language agents must revise planned actions when evidence changes, permission is revoked, or a stop instruction arrives. A useful response is selective: suspend affected actions, preserve unaffected work, and resume only after sufficient repair. We introduce NAQD-Env, a synthetic environment that evaluates these decisions against a deterministic reference policy over explicit evidence, authorization, and constraint dependencies. Eleven dependency families support evaluation on development structures, held-out families, and held-out combinations of structures. Metrics distinguish attempted violations from violations permitted by a simulated execution gate and jointly report policy agreement, task value, withdrawal, resumption, and event reporting. We evaluate three open-weight instruction-tuned models from two families under three prompt conditions on 350 frozen scenarios, yielding 3,150 model-prompt episodes before gate replay. Across the reported conditions, withdrawal recall is at most 0.06, no valid resumption is observed at eligible opportunities, and only one episode matches the complete reference policy. Under the NAQD prompt, Qwen2.5-7B has fewer unsafe-attempt episodes than Qwen2.5-3B and Llama-3.1-8B, but also completes less useful work and preserves unaffected actions less accurately. Exploratory supervised fine-tuning probes increase Qwen2.5-3B decision accuracy from 0.45-0.54 to 0.83-0.92; separate diagnostics reveal inappropriate withdrawal after curriculum omissions and a loss of event reporting. These results motivate evaluating selective withdrawal as a distinct component of agent reliability. The setting measures policy application with trusted structured inputs and does not establish real-world containment or source-verification ability.
[AI-205] PrivMeSA: Privacy-Aware Self-Evolving Multi-Agent System for Medicine via Local-Remote LLM Collaboration
链接: https://arxiv.org/abs/2609.38458
作者: Dannong Wang,Yuran Zhang,Bian Sun,Alex Stinard,Yuzhang Shang,Song Wang,Yu Tian
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Clinical large language model (LLM) agents deployed locally can consult more capable remote models, but doing so risks exposing patient information. Privacy-conscious delegation places disclosure decisions with a local agent, yet removing explicit identifiers is insufficient: quasi-identifiers can accumulate across multi-turn consultations and repeated patient visits to enable re-identification. We introduce PrivMeSA, a privacy-aware self-evolving multi-agent system that learns to control disclosure and retains remote expertise for local reuse. A local agent manages each encounter and consults remote specialists that may request additional information. Reinforcement learning balances task accuracy against direct disclosure and registry-based re-identification risk, with privacy evaluated over the complete outbound transcript of each encounter. A local lesson memory distills completed consultations into generalized clinical guidance and retrieves relevant lessons before transmission, allowing subsequent cases to reuse expertise without another remote exchange. Memory grows without additional outcome labels or parameter updates. On an emergency-department benchmark built from MIMIC-IV-ED records, PrivMeSA improves mean task accuracy over delegation by up to 15.8 percentage points. In the same setting, PrivMeSA reduces the disclosure of personal details from 98.0% to 0.2% of cases and the share of cases in which the patient can be narrowed to ten or fewer registry patients from 74% to 0%.
[AI-206] AIM: Agent ic Idea Management for Automated Research
链接: https://arxiv.org/abs/2609.38445
作者: Hyeong Kyu Choi,Bhavana Dalvi Mishra,Jiefeng Chen,Mihir Parmar,Rui Meng,Chun-Liang Li,Xiangru Tang,Sharon Li,Jinsung Yoon,Tomas Pfister
类目: Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE)
备注:
Abstract:Frontier LLMs are increasingly used to automate scientific research through iterative search. We distinguish idea-driven search from solution-driven search and identify three core challenges: organizing evolving research ideas, selecting promising directions, and maintaining alignment between ideas and their implementations. To address these challenges, we introduce the Agentic Idea Manager (AIM), a fully autonomous framework for managing and exploring research directions in idea-driven automated research. Inspired by Bayesian optimization, AIM uses an Agentic Surrogate and an Agentic Acquisition mechanism to organize discovered ideas and guide their selection. A Solution Auditor maintains idea-solution integrity, while a Resource Planner adaptively allocates the remaining experimental budget across parallel search branches. Experiments on 10 AutoLab benchmark tasks show that AIM surpasses the strongest baseline by 1.6 percentage points on System Optimization tasks and 4.9 percentage points on long-horizon Model Development CUDA tasks. Notably, AIM reaches the best baseline performance up to 3.1x faster in wall-clock time. We further provide a theoretical analysis of when searching over ideas becomes beneficial. Our analysis shows that explicit idea-level allocation makes semantic coverage directly controllable, and that broader coverage becomes increasingly valuable when competitive research directions are sparse among many plausible alternatives. Project Page: this https URL
[AI-207] Evaluating Whether GPT -6 Astra Performs Unsanctioned Supply-Chain Attacks
链接: https://arxiv.org/abs/2609.38415
作者: Alexandra Souly,Kai Fronsdal,Abby D’Cruz,Xander Davies,Robert Kirk
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:This technical report presents an alignment evaluation developed and performed by the UK AI Security Institute for assessing whether advanced AI systems take unsanctioned actions outside the scope of their assigned task. We evaluate whether frontier models conduct supply-chain attacks against out-of-scope, third-party targets when placed in difficult cybersecurity challenges, motivated by recently observed cases of models attacking real open-source repositories during evaluations. Applying our methods to GPT-6 Astra and previous OpenAI models, with cyber safeguards disabled, we find that GPT-6 Astra attempts complete supply-chain attacks in simulation at a higher rate than GPT-5.6 Sol and GPT-5.5. This includes writing malicious code as a contribution to an out-of-scope open-source codebase, creating fake identities to deceive open-source developers, and submitting benign contributions before malicious ones. GPT-6 Astra frequently reasons about the scope of the challenge in its chain-of-thought yet still proceeds to attack out-of-scope targets; it often asks for permission, and treats an automated message as as authorisation; and it continues to take unsanctioned actions, at a reduced rate, when internet access is more explicitly disallowed. Our evaluation builds on an internal version of Petri, an open-source LLM auditing tool, with all tool calls simulated by other LLMs, so that no real network access, systems or third-party repositories are reachable and no real-world harm is caused. Finally, we discuss limitations, in particular simulation awareness. We believe simulation awareness may have driven some of the observed behaviour but does not remove our concern. Our results suggest that defences beyond model alignment, such as sandboxing and monitoring, are increasingly critical for safe and secure deployment.
[AI-208] A Competing-Hazards Systematization of Loss of Control in Autonomous Agents
链接: https://arxiv.org/abs/2609.38411
作者: Mohamed Aly Bouke
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注: 12 pages, 1 figure, 3 tables. Dataset: this https URL
Abstract:Leading AI developers have reported agents acting beyond their approved limits, which a United Nations panel described as an early warning of loss of human control. Yet incident reports and agent-safety evaluations describe these events differently, making it difficult to compare failures, trace risk across attempts, or separate agent behavior from the environment’s role in allowing an out-of-scope action to succeed. To address this gap, we introduce a common framework in which each attempt ends in approved completion, safe stopping, scope escape, or continuation. We formalize the framework as a discrete-time competing-hazards model and derive escape probability within a retry budget, a model-conditional safe-budget limit, and conditions for estimation from execution logs. We audit 22 incident reports and 102 agent-safety evaluations published from January 2025 to September 2026 using primary sources. Six incidents involved tasks that could not be completed within scope, thirteen involved agents that continued rather than stopped, and five did not report stopping behavior. Developers’ figures imply a task-level incidence ratio near 47 for out-of-scope coordination in never-solved versus solved tasks. Among evaluations, 87 recorded an out-of-scope effect or specification violation, 26 treated safe stopping as a first-class outcome, only 20 recorded both, and 79 merged budget exhaustion with failure. In 20 of 22 incidents, the environment allowed an out-of-scope effect, indicating that realized loss of control often reflected persistent agent behavior interacting with permissive boundary conditions; meanwhile, no evaluation reported all fields needed to estimate the full competing-hazards process from published evidence.
[AI-209] SimTrace: Grounded Multimodal User Trajectories Generation for Online User Modeling
链接: https://arxiv.org/abs/2609.38397
作者: Yunan Lu,Shuang Xie,Meghna Allamudi,Mingyu Zhao,Han Li,Lingyun Wang,Zhou Yu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Virtual clients offer a cost-effective approach to support applications such as A/B testing, recommender system development, and interface evaluation. However, building them requires access to large-scale, semantically faithful, fine-grained online user trajectories. These data are difficult to obtain because proprietary logs are subject to privacy restrictions and small businesses often lack sufficient traffic. Consequently, existing public datasets either abstract away fine-grained user interaction details or preserve rich context but remain platform-specific and small-scale. To address this gap, we propose SimTrace, a framework that generates faithful, fine-grained synthetic multimodal clickstreams through a computer-use client agent that is grounded in real user trajectories and the given web environment. SimTrace anonymizes real interactions and constructs a simulated twin of the given web environment, then uses both to generate synthetic interaction trajectories. Each action is paired with its corresponding web observations and user context, yielding a shareable alternative to confidential logs for developing computer-use agent-style virtual clients. We apply SimTrace to an e-commerce setting and evaluate both its fidelity and downstream utility. SimTrace outperforms competing baselines on 7 out of 8 fidelity metrics. Models trained on synthetic data achieve performance comparable to those trained on real data on downstream tasks such as purchase prediction and recommendation. For next action prediction task, augmenting real data with synthetic data further improves accuracy by 11.0% relative to training on real data alone. We release SimTrace as an open-source package to facilitate research on online user behavior modeling.
[AI-210] Which Tasks Survive Self-Supervised Learning?
链接: https://arxiv.org/abs/2609.38393
作者: Achleshwar Luthra,Lucas Bryant,Tracy Zhu,Tomer Galanti
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Same-instance self-supervised learning (SSL) learns representations by enforcing consistency across two views of the same underlying instance. This principle alone, however, does not determine which downstream tasks remain recoverable from the learned representation. We study this question through \emphsemantic recoverability, defined as the amount of a task’s posterior score captured by the represented function space. We show that, for centered and whitened representations, recoverability exactly determines directional class-distance-normalized variance (CDNV), controls few-shot nearest-centroid classification, and governs the strength of task-relevant semantic directions. The population linear probe and centroid axis coincide, and multiple well-recovered tasks approach a factorial centroid geometry. We then analyze a canonical two-view SSL objective and show that its population optimum spans the leading cross-view-stable modes of the associated two-view operator. This yields a closed-form spectral characterization of semantic recoverability: a downstream task is preserved to the extent that its posterior lies in the selected spectral subspace. We validate these predictions on synthetic and real datasets across several SSL methods, testing the predicted relationships among recoverability, directional geometry, spectral structure, and few-shot transfer. Together, these results give a task-level account of what information survives same-instance SSL and how the retained information appears in downstream geometry and transfer.
[AI-211] MetaPersona: Task-Grounded Synthetic Populations from Empirical Social Science
链接: https://arxiv.org/abs/2609.38392
作者: Jinyi Ye,Yuangang Li,Chenxiao Yu,Preyashi Poddar,Priyanka Dey,Longtian Ye,Zihan Wang,Xiyang Hu,Emilio Ferrara,Yue Zhao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Personas used to seed LLM social simulations face a cold-start problem: existing methods lack a principled basis for deciding which attributes to include and how to assign their values. As a result, synthetic populations may misrepresent the demographic composition, latent attributes, and dependency structure that shape downstream behavior. We introduce MetaPersona-DB, a dataset of 11,000+ empirical human-subjects studies annotated with task-relevant variables, reported relationships, and aggregate-level population statistics. Building on this resource, we propose MetaPersona, a framework that retrieves task-relevant evidence, constructs literature-derived persona dependency graphs, and samples synthetic populations from empirical priors linking demographics, latent attributes, and outcomes. Across three downstream case studies, three baselines, and three frontier models, results vary by task and model: MetaPersona performs strongly on misinformation belief and AI-tool sentiment, while results on income redistribution are mixed. It also reduces persona-construction cost to under 0.5 per task using GPT-5.2. Finally, we present MetaPersona-Studio, a prototype interactive interface for empirically grounded persona generation.
[AI-212] Decode-Latency Feedback Prefill: A Model-Free Controller and Its Generalization Limits
链接: https://arxiv.org/abs/2609.38386
作者: Gaurav Agarwal,Ashish Garg,Isha Singhal
类目: Artificial Intelligence (cs.AI)
备注: 5 pages, 1 figure. Includes negative generalization results for Qwen3-8B, Qwen3-32B, and two-GPU tensor parallelism
Abstract:Concurrent autoregressive inference creates a fundamental interference problem: prefilling a newly arrived long prompt can delay tokens for requests that are already decoding. Fixed prefill chunks reduce this interference, but the best chunk size depends on the model, hardware, load, and latency objective. We introduce Decode-Latency Feedback Prefill (DLFP), a model-free controller that changes only prefill work that overlaps active decodes. After a guarded scheduling cycle, DLFP uses the observed interval as proportional feedback to resize the next prefill chunk; isolated prefills remain unrestricted. We implement DLFP in vLLM and evaluate it with open-loop Poisson arrivals, exact token accounting, raw request traces, and NVIDIA telemetry. On Qwen3-0.6B in BF16 on one A100 80 GB GPU, three paired 100-request trials reduce P99 inter-token latency by 24.8%, 30.1%, and 28.2% (mean 27.7%, paired 95% confidence interval 21.0% to 34.3%) with exact output agreement, no failures, and unchanged SLO compliance. The benefit is not free: mean P99 time to first token increases 34.8% while remaining inside the declared SLO. Crucially, the mechanism does not generalize to Qwen3-8B, Qwen3-32B, or a two-GPU tensor-parallel configuration. We trace the failure to an asynchronous scheduler-call interval that is only a proxy for completed GPU iteration time. This negative result defines the boundary of the contribution and motivates a completion-timed controller for concurrent CPU and on-device inference. We do not claim mobile-device performance; the present work is a reproducible proof-of-concept and generalization study. Comments: 5 pages, 1 figure. Includes negative generalization results for Qwen3-8B, Qwen3-32B, and two-GPU tensor parallelism Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2609.38386 [cs.AI] (or arXiv:2609.38386v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2609.38386 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-213] LogiC-Diff: Embedding Security Properties Into AI-Enabled Cyber-Physical Systems
链接: https://arxiv.org/abs/2609.38381
作者: Ziyan An,John Stankovic,Meiyi Ma
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:AI-enabled Cyber-Physical Systems (CPS) are highly vulnerable to adversarial and anomalous inputs, where small perturbations can induce cascading errors and unsafe control actions. Existing approaches, such as rule-based filtering, training-time regularization, or diffusion-based reconstruction, either operate outside the model or lack mechanisms to incorporate formal security specifications into the prediction process. In this paper, we take the first step toward embedding security properties directly into AI-enabled CPS, enabling predictive models to enforce system-level constraints during inference rather than relying on external defenses. We introduce a logic-conditioned bi-stage diffusion framework that integrates Signal Temporal Logic (STL) specifications into forecasting. STL serves as a first-class conditioning signal that guides both an input repair stage and an output refinement stage, allowing the model to jointly mitigate adversarial perturbations and enforce desired temporal behaviors to satisfy security-critical properties. We evaluate our approach on two real-world multivariate CPS forecasting datasets under a diverse set of physical sensor and cyber attacks. Across sensor faults, gradient-based attacks, adaptive attacks, and varying attack strengths, our method consistently improves robustness and specification compliance, degrades more gracefully as attack strength increases, and generalizes better to unseen attacks. Ablation studies on specification coverage and quality further show that embedding logical security properties yields gains unattainable by reconstruction-based methods alone, highlighting a new direction for integrating formal methods with generative models in secure CPS.
[AI-214] Aligned Data Can Induce Misalignment via Context Confusion
链接: https://arxiv.org/abs/2609.38379
作者: Yavuz Bakman,Duygu Nur Yaldiz,Baris Askin,Swastik Roy,Morteza Ziyadi,Salman Avestimehr,Sai Praneeth Karimireddy
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Large language models (LLMs) are frequently updated for various use cases, where filtering out misaligned training samples is a common practice for preventing post-update misalignment. However, alignment is inherently context-dependent: a recommendation that is aligned in one context may be inappropriate in another. For example, in response to the question “What should a researcher do with the research data?”, recommending that the researcher preserve the data for reproducibility is aligned. In contrast, recommending data saving in response to “What should a mobile-app developer do with users’ sensitive data?” may be inappropriate from a privacy perspective. Starting from this observation, we identify a post-training phenomenon where aligned training induces misaligned behavior in other contexts. We call this phenomenon context confusion. We demonstrate context confusion across three domains: (1) Gender Equality, (2) Privacy, and (3) Physical Safety. We further show that context confusion causes narrow misalignment, in contrast to emergent misalignment, and is not effectively reduced by injecting general alignment data, but can be substantially reduced by including targeted alignment data for the misaligned domain or providing in-context learning examples during inference. Lastly, we provide a mechanistic explanation of context confusion. We observe that queries from different domains can undergo similar representational shifts during the fine-tuning. Consequently, a query from a different domain may activate the same behavioral feature learned during fine-tuning, which causes the behavior to transfer to a context where it is misaligned. Based on our findings, we argue that it is difficult to predict the alignment state of a model after training by inspecting the training data alone, which highlights the importance of comprehensive post-training alignment evaluations.
[AI-215] Self-Evolving Harness on Multiple Tasks with the Agent as Its Own Optimizer
链接: https://arxiv.org/abs/2609.38372
作者: Qiankai Xu
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 18 pages, 5 figures
Abstract:A harness is the code around a language-model agent that organizes prompts, calls tools, manages context, and controls execution. As models grow stronger, recent work has begun to let agents improve their own harnesses, a line of work known as self-evolving harnesses. In most existing methods, a separate proposer running on a human-designed harness modifies the solver’s harness, and a separate harness is evolved for each benchmark. Real-world tasks come from many domains, so both the evolution and the evaluation of a harness should cover a diverse range of tasks. We propose a framework close to recursive self-improvement: the same frozen model, on the same version of the harness, first solves tasks as the solver and then, as the proposer, reads the complete run records and directly edits the harness that runs it. Each evolution batch draws tasks from five benchmarks in different domains. To measure generalization, training and held-out tasks are strictly separated, and we additionally evaluate on five out-of-distribution benchmarks never used during evolution. We frame the evolution process as deep-learning training with two stages, multi-task pretraining and continual training. Starting from a 49-line seed harness, the harness obtained at the end of the first stage improves the average score by 4.48 points on the in-distribution benchmarks and by 12.64 points on the out-of-distribution benchmarks, surpassing Codex on the former and matching it on the latter. In the second stage, continued evolution on Claw-Eval, one of the out-of-distribution benchmarks, further raises the score on that benchmark from 66.17 to 68.06, exceeding Codex. We also provide an in-depth analysis of the mechanisms that emerged during evolution, including output truncation, history compaction, and independent review.
[AI-216] Can an AI Agent Rediscover a Blaschke-Curve Invariant? NEURIPS2026
链接: https://arxiv.org/abs/2609.38369
作者: Yunus E. Zeytuncu
类目: Artificial Intelligence (cs.AI)
备注: Accepted for poster presentation at the NeurIPS 2026 Workshop on Mathematical Reasoning and AI (MATH-AI). 8 pages, 1 figure. Code and data: this https URL
Abstract:We study generalized Blaschke curves as a controlled environment for AI-assisted mathematical rediscovery. For one fixed degree-four Blaschke product, an agent receives numerical coordinates of the six pair-lines determined by each of 80 boundary configurations. The target theorem is withheld from the task instructions. The saved research log reports rejected geometric hypotheses and a homogeneous cubic fitted to polygon sides. Its frozen coefficients predict 480 lines from 80 unseen parameter values, with a recorded RMS scale-free residual of 8.88\times10^-17 . Discovery-set diagonals provide an out-of-fit consistency check, not a fully held-out test. A separate one-configuration run reports insufficient evidence for invariance. A post-review deterministic degree-search baseline also recovers the cubic, so the experiment does not establish an advantage over polynomial fitting. We present this single-instance case study as a protocol for separating conjecture, numerical validation, and proof, with explicit limitations concerning agent metadata, prior knowledge, and reproducibility.
[AI-217] Simulator-Refined Diffusion for Radio-Frequency Inverse Design
链接: https://arxiv.org/abs/2609.38363
作者: Jinhao Liang,Jacob K. Christopher,Michael Frei,Tommaso Dreossi,Nando Fioretto
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Diffusion models have shown potential in inverse design of printed circuit boards (PCBs), enabling the generation of layouts conditioned on target S-parameters. Despite this promise, applying diffusion models to PCB layout generation remains challenging due to their difficulty in meeting the quantitative electromagnetic specifications. A common approach is gradient-based guidance, which biases the diffusion sampling process with the gradient of an objective used for evaluation. However, full-wave electromagnetic simulators are accurate but expensive and typically non-differentiable, whereas differentiable surrogates are informative but not always reliable. To address these limitations, this paper proposes Simulator-Refined Diffusion (SRD), a novel combination of a low-fidelity differentiable surrogate and a high-fidelity non-differentiable simulator within the diffusion sampling process. Unlike standard zeroth-order optimization, which requires a great number of random perturbations, our approach uses the surrogate’s gradient to propose the perturbation direction while the simulator then searches based on this direction to identify an effective design update. Experimental results across different settings show that this method consistently outperforms current state-of-the-art methods, producing layouts whose simulated S-parameters match the target specifications up to 21.2% closer for in-distribution targets and up to 19.8% for out-of-distribution targets.
[AI-218] MILO: Automated Harness Discovery via Orchestrated Multi-Agent Evolution
链接: https://arxiv.org/abs/2609.38349
作者: Prithwish Jana,Mononito Goswami,Hao Liu,Xinyu Li,Langlin Huang,Zhehui Huang,Zhishen Huang,Patrick Blöbaum,Anoop Deoras,Purak Jain,Nikos Kanakaris,Sahika Genc
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Modern agentic systems combine an AI model with a harness that controls execution and environmental interactions. Harness design strongly affects long-horizon performance, yet its combinatorial search space demands substantial human effort that must be repeated as models change. Existing automated methods explore this space narrowly, optimizing only components such as prompts or skills or becoming trapped by fixed, exploitative search strategies. We introduce MILO (Meta-evolutionary Island Orchestration), a framework that co-evolves agent harnesses and the strategy used to discover them. MILO combines: (i) hierarchical lineage memory over island-based trees, using rejected mutations as negative evidence; (ii) per-island mutator agents that rewrite complete harnesses using global search history and parent-specific feedback; and (iii) an orchestrator that adapts search through lineage grafting and speciation, mutator reassignment and curriculum revision. Across Terminal-Bench 2.1, PaperBench, and DeepSWE, MILO-discovered harnesses outperform eight state-of-the-art harnesses and six search methods using frontier (Opus 4.8) and open-weight (gpt-oss-120b) models. With Opus 4.8, MILO improves resolution over its initial harness by +12.0% , +28.3% , and +10.3% , respectively, compared with best prior-search gains of +4.5% , +18.3% , and 0% . On Terminal-Bench 2.1, it achieves 86.1 \pm 2.0% , exceeding the official leaderboard’s top entry ( 83.8 \pm 2.3% ) while using 26% fewer tokens than its initial harness. On EinsteinArena open problems, MILO improves best-known upper bounds for Erdős minimum-overlap ( 0.3808586 \to 0.3808568 ) and the first and third autocorrelation inequalities ( 1.50274365 \to 1.50274360 ; 1.45081 \to 1.44889 ).
[AI-219] Examining Variation in How Guided AI Tutors Resolve Student Impasses
链接: https://arxiv.org/abs/2609.38346
作者: Bakhtawar Ahtisham,Kirk Vanacore,Alessandra Napoli,Josh Arens,Ksenia Ionova,Clayton Cohn,Shima Salehi,Rene Kizilcec
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:When a student is stuck, a tutor faces the assistance dilemma: help given too early can hinder productive struggle, while help withheld too long leaves the student in a frustrating, persistent impasse (i.e., wheel spinning). Generative AI tutors increasingly use guardrails restricting answer-giving, yet little is known about how such tutors behave once an impasse persists. We analyze 20,462 student turns from 1,260 authentic sessions with a guided LLM chemistry tutor, identifying 6,630 impasse turns of three major types: conceptual errors, expressed uncertainty, or help-seeking. We then used these impasses to simulate three tutoring conditions to study variation in AI tutor guidance through impasses: baseline, no-direct-answer, and guided tutor. For a sample of 150 impasses, prompt specificity changed pedagogy: a baseline tutor provided the answer directly in 50.7% of responses, a no-direct-answer tutor asked a follow-up question every time, and the guided tutor responded in a wide variety of ways depending on the context. We then analyzed impasse trajectories in authentic interactions, finding that each additional impasse turn lowered the odds of next-turn recovery by 12.7% (AOR = 0.873, p .001), and early dropouts were caught in recursive concept elicitation before reaching execution. The benefit of questioning decayed as impasses persisted (scripted question x depth AOR = 0.78; follow-up x depth AOR = 0.83), whereas addressing the student’s error grew more beneficial (AOR = 1.14); after a failed scripted question, repeating it was followed by recovery in 28.1% of cases, compared with 39.8% when the tutor addressed the error instead. For learning analytics, these findings identify impasse depth and type as observable, turn-level dialogue signals that analytics can use to trigger graduated, state-sensitive assistance in real time.
[AI-220] CARAT: Do Materials LLM s Reason or Recite?
链接: https://arxiv.org/abs/2609.38340
作者: Jiajun Wu,Jian Yang,Zixiang Ni,Zhenzhu Li,Bin Chong
类目: Artificial Intelligence (cs.AI)
备注: 42 pages, 19 figures, including appendix
Abstract:When a materials LLM answers a question about crystal structure, does it reason from the structure or copy an answer already printed in its input? Accuracy cannot tell: a structural description often prints the very field it is scored against. CARAT holds question and gold answer fixed across eight matched views, names each structural relation separately in GraphSpace, and adds matched fine-tuning, answer masking, evidence injection, paired inference, and a rule that can withhold claims. First, on the benchmark’s hardest families the grounded view is worth 17.3 points over formula inputs. Second, we turn that scrutiny on ourselves. GraphSpace beats a plain periodic graph by 19.3 points, but that margin is two effects at once: where the plain rendering carries everything the question needs it is 1.96 points, and where it omits those fields entirely, 46.7 points. The headline mostly measures what the baseline lacked, not how evidence is presented. Third, we attack our own benchmark. A rule that skips the link and reads the list directly answers four of seven hardened families, so we rebuilt it until eleven such shortcuts sat near chance. The frozen model quotes that link yet answers the same when we redirect it, on 95.6% of paired cases: it repeats the relation without using it. After matched supervision it reaches 99.8%, and deleting the link drops it to 23.4%, below the 27.0% the best shortcut reaches: both steps are learnable.
[AI-221] Aegis: Generative Gradient Masking for Privacy-Preserving Medical Federated Learning NEURIPS2026
链接: https://arxiv.org/abs/2609.38339
作者: Chaoyu Zhang,Shanghao Shi,Heng Jin,Ning Wang,Y. Thomas Hou,Wenjing Lou
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: Comments: 10 pages of main text, 4 figures, 2 tables, and 1 algorithm; supplementary material included. Accepted by NeurIPS 2026
Abstract:Federated learning (FL) has become a foundational paradigm for multi-institutional medical AI, allowing hospitals and research centers to jointly train diagnostic models without exchanging patient records. This privacy promise, however, is increasingly contested: a malicious or honest-but-curious server can launch model inversion attacks (MIAs) that reconstruct private patient images directly from shared model updates, and recent scalable, closed-form attacks penetrate even secure aggregation at clinically realistic batch sizes. Existing defenses face an unsatisfactory dilemma. Gradient-perturbation methods such as differential privacy and pruning trade away the diagnostic accuracy on which clinical reliability depends, while cryptographic protocols add system complexity yet still leave updates exposed to these scalable attacks. We propose Aegis, a principled client-side defense that breaks this dilemma without perturbing patient data or modifying the FL protocol. Our key insight is that the success of every known MIA is fundamentally bounded by the local batch size relative to the model’s leakage capacity; once this limit is exceeded, distinct samples collide and reconstructions collapse into indistinguishable mixtures. Aegis turns this universal bottleneck into a defense: each client superimposes onto its real update a masking gradient computed on locally synthesized, task-relevant data, deliberately pushing the effective batch beyond the attack’s recovery capacity. We complement the design with theoretical convergence guarantees under standard convex assumptions and evaluate Aegis on MNIST, CIFAR-10, and three MedMNIST modalities (chest X-ray, abdominal CT, colon pathology). Aegis neutralizes three state-of-the-art MIAs while preserving model utility and incurring only modest overhead, offering a practical privacy primitive for medical FL.
[AI-222] E2E-SWE: Benchmarking LLM s on Building Working Codebases from Scratch
链接: https://arxiv.org/abs/2609.38335
作者: Hantian Ding,Chloe Bi,Jiacheng Zhu,John Yang,Matt Deitke,Pengcheng Yin,Zijian Wang,Rui Hou
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:Coding agents powered by large language models (LLMs) are evolving from making localized code changes to developing complete software repositories. However, evaluating repository-scale generation remains challenging: tasks must demand system-level reasoning while ensuring that all evaluated behaviors are precisely specified and independent of any particular implementation. We introduce E2E-SWE, a benchmark for evaluating whether coding agents can build complete, functional software repositories end to end. E2E-SWE contains 186 whole-repository generation tasks spanning 11 programming languages. Given only a natural-language specification and an empty workspace, an agent must implement a complete, installable project that satisfies a comprehensive suite of hidden tests. Each task is constructed by a software engineer in collaboration with an LLM; together, they develop the test suite and a corresponding implementation-independent specification. To ensure that tasks are well specified and practically solvable, we further subject them to an iterative verification process in which autonomous agents audit and repair task defects using static inspection and failures observed from real model rollouts. Evaluating 13 frontier models, we find substantial variation in end-to-end repository generation ability, with pass@1 ranging from 11.7% to 67.7%, providing strong model differentiation while leaving considerable headroom for future progress. Analysis of agent trajectories further reveals long, front-loaded reasoning patterns, highlighting the planning and system-level reasoning required to construct working codebases from scratch.
[AI-223] AI Agents are Vulnerable to Radicalization
链接: https://arxiv.org/abs/2609.38296
作者: Ozgur Can Seckin,Shalmoli Ghosh,Alessandro Flammini,Kristina Lerman,Maria Elizabeth Grabe,Filippo Menczer
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注:
Abstract:Large language models (LLMs) can influence people’s beliefs, yet little is known about whether and how they can manipulate each other. To investigate this, we simulate conversations between two agents: a target LLM that role-plays a human persona based on demographic and psychological attributes, and an influencer LLM that aims to make the target’s beliefs more extreme. We examine radicalization along two pathways: resonance, where the influencer reinforces a target’s pre-existing belief, and persuasion, where the influencer promotes a belief the target initially considers unimportant. Across affective and behavioral metrics, we find that both mechanisms radicalize the target. However, resonance produces consistently stronger effects than persuasion. Different influence tactics, such as using sycophancy and unverified claims, produce different levels of radicalization, but not consistently across metrics. We further show that resonance propagates to related beliefs, suggesting interconnected belief structures within AI agents. These findings indicate that AI agents are susceptible to radicalization, particularly when messages align with their existing beliefs, raising concerns about the vulnerability of personalized AI agents and multi-agent AI ecosystems.
[AI-224] MoFlow: Multi-Objective Agent ic Workflow Generation
链接: https://arxiv.org/abs/2609.38294
作者: Yining Lu,Aurelie Lozano,Xi Yang,Naoki Abe,Yu Deng,Meng Jiang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We study the generation of agentic workflows that jointly optimize multiple objectives, such as accuracy, cost, latency, robustness, and consistency. Existing methods for workflow generation typically optimize accuracy alone or a weighted sum of objectives, so each trained generator commits to one fixed trade-off and must be retrained from scratch when preferences change. To alleviate this, we propose MoFlow, which generates workflows optimized across varied preferences. Specifically, MoFlow formulates workflow generation as a multi-objective Markov decision process and solves it by leveraging Convex-Hull Monte Carlo Tree Search with optimistic set-valued backups, where every node stores a set of reachable trade-offs rather than one weighted score. A single search thus approximately covers the Pareto front, from which MoFlow can return a workflow for any preference by lookup without retraining. We evaluate MoFlow against six strong baselines on six benchmarks spanning mathematics, code, and question answering. Since the baselines are single-scalar optimizers by design, an apples-to-apples comparison is difficult. We instead adopt an evaluation setup that favors the baselines, in that they are rerun for each testing preference, which MoFlow never sees. Even under this stringent setup, MoFlow achieves the highest average hypervolume.
[AI-225] AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks
链接: https://arxiv.org/abs/2609.38288
作者: Hongjin Qian,Chaofan Li,Kun Luo,Wenqing Wei,Jianlyu Chen,Shuqi Lu,Yuyang Hu,Hongwang Xiao,Hui Wang,Chaozhuo Li,Qiwei Ye,Zhicheng Dou,Defu Lian,Zheng Liu
类目: Artificial Intelligence (cs.AI)
备注: Code will be released at this https URL and models at this https URL
Abstract:We present AREX-2, an effort to advance the self-improving capability of LLM agents, which we define as the ability to iteratively refine a solution at test time. This ability rests on two complementary capabilities: reflection, which produces a solution better than the current one, and long-horizon execution, which keeps the iteration effective over many rounds. We hypothesize that both capabilities are domain-agnostic, and can therefore be learned in scenarios that are well suited for supervision. Accordingly, we synthesize long-horizon improvement trajectories from machine learning and algorithmic programming tasks, two domains that offer verifiable feedback and reward sustained iteration. Trained on this data, our agent, built on Qwen3.8-27B, achieves strong results on MLE-bench Lite (81.8) and Frontier-CS (70.7), transfers to deep research with 84.0 on BrowseComp, 52.6 on HLE, 92.2 on GAIA, and 93.8 on DeepSearchQA, and keeps improving as its budget of rounds grows. These results show that long-horizon reflective data is an effective route toward self-improving agents.
[AI-226] Social Choice Foundations for Simulation-Augmented Generation NEURIPS2026
链接: https://arxiv.org/abs/2609.38287
作者: Sonja Kraiczy,Smitha Milli,Ratip Emin Berker,Avinandan Bose,Brandon Amos,Jamelle Watson-Daniels,Maximilian Nickel,Edith Elkind,Ariel D. Procaccia
类目: Computer Science and Game Theory (cs.GT); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at NeurIPS 2026
Abstract:Simulation-augmented generation (SAGE) is a recent technical proposal in which models simulate individuals’ viewpoints at inference time in order to provide more representative answers to contentious user queries. A core challenge for SAGE is making inference-time simulation efficient without sacrificing representation quality. We introduce the first formalization of this problem, based upon an axiom from proportional clustering known as metric proportional justified representation+ (mPJR+) which is the strongest proportionality axiom known to always be satisfiable by centroid-based clustering. We prove that to proportionally represent the viewpoints of a population of n_H humans on a given prompt, we need only create simulations of n \ll n_H individuals, and at inference time, need only dynamically route to k \ll n of those simulations based upon the prompt. This twofold reduction still yields approximate proportional representation guarantees for the entire population. Empirically, across two domains-political questions and personal advice-our proposed routing algorithm achieves higher mPJR+ satisfaction rates than k -means-based or random selection baselines.
[AI-227] Improving OCR Faithfulness via Gated and Attenuated On-Policy Distillation
链接: https://arxiv.org/abs/2609.38282
作者: Baode Wang,Zuming Huang,Kexuan Ren,Jun Huang,Wei Chu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Vision-language models may rewrite anomalous text in images into linguistically plausible expressions, compromising OCR transcription faithfulness. Sequence-level task rewards and local teacher guidance are complementary, but guidance from the same teacher may not remain equally effective as the student improves. Offline analysis shows that supervision from a fixed teacher becomes progressively less favorable as the student improves, both across training checkpoints and across response groups with different task rewards. Motivated by this observation, we introduce GAD-RL, which adaptively regulates teacher supervision during joint post-training according to the student’s current task performance and local distributions. A frozen teacher conditions on reference transcriptions and student-generated prefixes. GAD-RL disables distillation for response groups containing an output with task reward at least 0.95 and continuously attenuates distillation strength as group-mean reward increases. It also weights forward KL by the student’s probability of the teacher’s Top-1 token, moderating local auxiliary updates when student support for that candidate is low. On Qwen3.5-2B, GAD-RL achieves 59.92% Micro Recall on CHAOS-Bench, surpassing GRPO and GRPO+OPD (fixed-weight) by 8.45 and 4.43 percentage points, respectively, while achieving an Overall score of 91.18 on OmniDocBench v1.6.
[AI-228] When Correct Memory Goes Wrong: Fuzzing Persistent Memory Use in LLM Agents
链接: https://arxiv.org/abs/2609.38275
作者: Yuqiao Meng,Luoxi Tang,Yingxue Zhang,Yuchen Yang,Zhaohan Xi
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI)
备注:
Abstract:Persistent memory helps LLM agents carry information across long interactions, but correct memory can still be used incorrectly when queries change or memory states evolve. Existing work mainly studies memory content errors or evaluates fixed test cases, leaving memory-use failures hard to discover systematically. We formulate this issue as a fuzzing problem and categorize such failures into query-related and memory-state failures. We then develop U-Fuzz, which starts from memory checkpoints as test seeds, mutates queries or memory states under explicit mutation obligations, validates each mutant, and uses observed memory behavior to guide iterative testing while keeping failure labels outside the search. We evaluate U-Fuzz across several memory systems against diverse fuzzing baselines, and further test an output-only setting with API-based LLMs where memory retrieval is hidden. Across these settings, U-Fuzz consistently uncovers more confirmed memory-use failures, showing that its search remains effective across different memory architectures and even when only final responses are observable.
[AI-229] Zero2Repo: Can Coding Agents Build Repositories from Scratch?
链接: https://arxiv.org/abs/2609.38269
作者: Pei Yang,Tianyu Shi,Yuhang Yao,Wanyi Chen,Tongyun Yang,Dun Pei,Haonan Wang,Pengbin Feng,Guanxu Yu,Jingchun Huang,Zeyu Zhang,Shuhan Sun,Hao Li,Xiang Li,Jie Xiao,Xinyu Wang,Hanxin Chen,Daqi Li,Qi Jia,Hongshan Lin,Zhizhou Gu,Zijun Tian,Weizhi Du,Lynn Ai,Eric Yang
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 19 pages, 4 figures, 8 tables
Abstract:Coding agents are increasingly asked to build software rather than patch it, yet benchmarks for from-scratch repository construction are mostly limited to a single language and depend on manually curated tasks. We introduce Zero2Repo, a benchmark in which an agent receives a product requirements document, an interface contract, and an empty workspace, and must deliver a complete repository in the project’s native ecosystem. Tasks are produced by a language-agnostic authoring pipeline that converts real, version-pinned open-source projects into behavioral specifications, reproducible environments, and hidden acceptance tests. Each task is validated by execution: a reference implementation derived from the upstream project must pass, and adversarial validation must show that the tests reject incorrect implementations. Evaluation runs production coding agents in isolated containers, withholds the acceptance tests until an explicit submission, and assigns a binary reward only when every test passes, with no LLM judge. The pipeline and harness make no language-specific assumptions and apply to mainstream programming ecosystems; the current release contains Python, TypeScript, Go, and C++ tasks. Even on 11 tasks drawn from repositories that frontier models have very likely seen during training, the strongest agent solves only 10, and every failing submission passes 90-99% of the hidden tests; for the two strongest agents, 67-100% of failed tests trace to a single omission or a low-frequency rule stated in the specification rather than to a missing subsystem, so each failure is a concrete target for improvement.
[AI-230] Janus: Evidence-Before-Effect Sagas and Offline-Verifiable Provenance for Agent ic LLM s
链接: https://arxiv.org/abs/2609.38266
作者: Mustafa Arslan
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
备注:
Abstract:Agentic large language models (LLMs) now move money through tools, yet the record of what they did is usually a trace their own process emits beside the effect. Janus puts the record on the effect path. A step’s proposal, the verdict on it and any answer from a validator or a person are durable in a signed, hash-chained log before the step may run or its effect be released; with keys declared, each answer is signed by whoever gave or relayed it. Gates are pure functions of that log, and an auditor re-derives every verdict offline from the log and one public key. At the MCP edge the effect is held until then; through the SDK, which our model experiment uses, a cooperating client runs it only afterwards. We evaluate Janus under crash injection (144 kills in-process, 81 through the daemon), by verifying a 100-million-event log offline (254.5 s), and with a real model behind a lending workflow, run governed and plain on the same recorded model outputs. With the lending mandate in the model’s prompt, the comparison was 0 against 0. With it only in the policy and the amount’s unit stated, the model approved six loans declared over the mandate, three with no injection (a run that also dropped the unit approved three); the plain agent paid all six and Janus none, each refused by a deterministic validator and re-derivable offline. An always-approve oracle over the recorded intakes gave 20 and 21 declared over the mandate against 0, though Janus paid four and three whose declared amount understated the request. Designing the experiment exposed, in a system that passed its own audit, an instance of post-approval substitution: an approval keyed to an attempt was counted for a different proposal, moving a person’s approval from 100 to 1,000,000. We report it, a first fix and the five routes around it, and what Janus does not guarantee.
[AI-231] How Should Diffusion Language Models Edit Code?
链接: https://arxiv.org/abs/2609.38257
作者: Xijia Tao,Ziru Liu,Shansan Gong,Jiacheng Ye,Kecheng Chen,Zirui Wu,Lin Zheng,Xinyu Fu,Rui Liu,Lingpeng Kong
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:Code editing requires a model to decide where to make changes, generate the new content, and preserve everything else. We study how masked diffusion language models divide these responsibilities across four editing interfaces: whole-file rewriting, search-and-replace, locate-then-infill, and token-level editing. Experiments on CanItEdit reveal a composition gap: diffusion models can generate coordinated changes when the correct edit locations are supplied, but much of this capability is lost when those locations must be predicted. Access to the intact original code helps the model fill multiple edit regions, yet does not resolve the difficulty of selecting those regions. By varying the editable regions while holding the generation model and decoding procedure fixed, we identify two distinct requirements for successful editing: covering every required change and placing precise boundaries around it. Missing a required region prevents the corresponding change, while widening regions to ensure coverage can sharply reduce success by requiring unchanged code to be regenerated. A sentence-level Wiki editing probe shows the same qualitative gap between supplied and predicted locations beyond code. These findings show why strong infilling capability alone does not ensure reliable editing: the interface must expose all required changes while limiting regeneration of unchanged code.
[AI-232] CollageAttack: Exploiting Cross-Modal Alignment Flaws in T2I Models through Spatial Text Composition
链接: https://arxiv.org/abs/2609.38253
作者: Zhiyi Mou,Yao Lu,Wangze Ni,Di Hong,Dakun Shen,Haoyang Li,Chen Jason Zhang,Alexander Zhou,Kui Ren
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: Contains potentially unsafe text-to-image generation examples. Code is released publicly
Abstract:Text-to-image (T2I) models have substantially improved in language understanding, in-image text rendering, and visual composition, while their safety mechanisms do not always keep pace with these capabilities. This creates a cross-modal attack surface in which harmful semantics can remain inconspicuous in a serialized prompt yet emerge through image-level composition. We propose CollageAttack, an automated single-prompt black-box jailbreak that shifts semantic assembly into the image plane by combining context-relevant scenes, scene-grounded textual carriers, and spatially distributed text fragments. Experiments across multiple open-weight and commercial T2I models show that CollageAttack achieves attack success rates of up to 86.0%, outperforming the strongest baseline on the same model by 18.5 percentage points, while consistently producing more harmful outputs and preserving the source intent. We further find that distributed textual fragments can reconstruct the intended semantics after generation, with visual composition producing stronger communicative impact than text alone. These results reveal a cross-modal safety gap in which harmful meaning emerges from the composition of individually less explicit elements.
[AI-233] Forensic-Aware Continual Adaptation for Image Forgery Localization
链接: https://arxiv.org/abs/2609.38251
作者: Chenqi Kong,Song Xia,Anwei Luo,Peisong He,Alex C. Kot,Yuming Fang
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:The rapid evolution of image manipulation techniques has raised growing public security concerns. Existing Image Forgery Localization (IFL) methods can accurately localize manipulated regions but are often unable to adapt to newly emerging forgeries. In real-world forensic scenarios, data typically arrive sequentially, yet continual model adaptation remains largely unexplored in IFL. To bridge this gap, we introduce the first continual learning framework for IFL and establish a comprehensive benchmark under two realistic data-evolution protocols: cross-dataset and cross-content continual learning. Evaluations of representative state-of-the-art IFL and continual learning methods reveal substantial performance degradation, highlighting two key challenges: (1) adaptively capturing intrinsic forensic traces from incoming data across unseen domains, and (2) preserving previously acquired forensic knowledge during sequential adaptation. To address these challenges, we propose a forensic-aware continual adaptation framework. First, a forensic trace mining module employs Spatial Mixture-of-Forensic-Experts (SMoFE) to dynamically route complementary forensic cues across spatial locations, together with Forensic Evidence-Guided Dense Prompting (FEGDP) to transform low-level forensic traces into structured localization evidence for SAM. Second, Fisher-weighted LoRA Gradient (FLAG) surgery identifies old-task-sensitive adaptation directions and suppresses conflicting updates, mitigating catastrophic forgetting while preserving plasticity for emerging forgery domains. Extensive experiments demonstrate state-of-the-art performance in both pixel-level forgery localization and image-level forgery detection across diverse continual learning scenarios.
[AI-234] ModalFidelity: Routing Modalities for Deepfake Detection on a Budget ICASSP2027
链接: https://arxiv.org/abs/2609.38246
作者: Oguzhan Baser,Kaan Kale,Sriram Vishwanath,Sandeep Chinchali
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 5 pages, 4 figures, 1 table. Submitted to ICASSP 2027
Abstract:Deepfakes no longer need to fake a whole video. Generators that read the transcript now alter only the few seconds in which a video’s meaning turns, so a forgery hides in a small, unknown fraction of the video. Yet detectors still read every one-second window of both the audio and image streams, spending nearly all of their compute where nothing was altered. We observe that deciding where to look is far cheaper than looking. We present ModalFidelity, a lightweight router that previews each window and decides, before any forensic detector runs, which stream is worth reading, under a hard compute budget it can never exceed. On AV-Deepfake1M, reading at most a fifth of the windows, it is more accurate than gating after the detectors at 15.9x less compute, and retains over 96% of the accuracy of an oracle that knows where every forgery lies.
[AI-235] SynIL: Leverag ing Synergy for Offline Imitation Learning from Imperfect Demonstration Datasets
链接: https://arxiv.org/abs/2609.38225
作者: Yuto Tanaka,Kyo Kutsuzawa,Martina Doku,Dai Owaki,Mitsuhiro Hayashibe
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Imitation learning enables robots to acquire complex skills directly from massive demonstration datasets, but its performance degrades severely when datasets are contaminated with suboptimal or noisy demonstrations. While prior quality-assessment methods attempt to filter or reweight data, they typically rely on manual pre-selection of expert reference data or task-specific heuristics, limiting scalability. To address this challenge, we introduce SynIL (Synergy-based Imitation Learning), a novel framework for automated, label-free demonstration quality assessment in offline reinforcement learning. Grounded in neuroscientific evidence that motor synergy, a low-dimensional coordinated structure in movement, correlates directly with motor proficiency, SynIL algorithmically quantifies synergy manifestation to generate dense, transition-level reward signals via self-supervised reward regression. Comprehensive evaluations on D4RL locomotion benchmarks and multi-human Robomimic manipulation datasets demonstrate that synergy-derived rewards correlate strongly with ground-truth rewards. Furthermore, SynIL substantially outperforms Behavior Cloning (BC) and achieves performance comparable to, and in sparse-reward human teleoperation scenarios, superior to, offline reinforcement learning trained on true environment rewards.
[AI-236] DualCast: A Dual-Path Language Model for Bimodal Financial Time-Series Forecasting
链接: https://arxiv.org/abs/2609.38197
作者: Wentao Zhao,Hongqiang Wu,Shanghang Liu,Zhaochen Zan,Yu Zhang,Biqing Huang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Financial time-series forecasting must capture price dynamics across heterogeneous assets while incorporating news available at prediction time. We introduce DualCast, a dual-path framework that extends a frozen language model with a discrete financial vocabulary. Each log-return patch is represented by a learned summary token and three residual shape tokens, preserving local drift and volatility while allowing shape patterns to be shared across assets. To improve codebook utilization, we develop adaptive frequency-equalizing residual vector quantization, which rebalances overloaded codewords without compromising reconstruction accuracy. The fast path trains only the new financial-token embeddings and output heads on a frozen Qwen3-8B backbone. A toggleable LoRA adapter enables a slow path that conditions on the fast forecast and news available at the forecast origin to produce a revised prediction. The reviser is initialized by supervised fine-tuning and further optimized with a return-space group relative policy optimization objective that rewards improvements over the fast forecast. In zero-shot evaluations covering equities and energy prices at five-minute, daily, and weekly resolutions, the slow path achieves the lowest mean absolute percentage error among the compared methods in 8 of 12 dataset-horizon settings, including every longest-horizon setting. News ablations indicate additional gains in most tested settings, although their magnitude varies across markets. DualCast thus combines a fast numerical forecaster with an optional text-conditioned revision mechanism.
[AI-237] Conformal Adversarial Generative Ensemble ICONIP2024
链接: https://arxiv.org/abs/2609.38196
作者: Ahmad Shahi,Mamehgol Yousefi,Brendon J. Woodford,Farhaan Mirza,Tapabrata Chakraborti
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Published in ICONIP 2024 (Neural Information Processing), LNCS 15287, Springer Nature, 2025
Abstract:Accurate time series forecasting is critical across various domains, yet traditional ensemble methods often suffer from the disproportionate influence of extreme forecasts. We introduce the Conformal Adversarial Generative Ensemble (CAGE), a novel framework that combines generative modeling, adversarial discrimination, and conformal prediction to enhance forecast reliability and accuracy. CAGE employs multiple generative models to produce initial forecasts, which are then evaluated by a discriminative component using conformal prediction techniques. P-values derived from nonconformity scores help dynamically adjust model weights, minimizing the impact of unreliable forecasts. This approach ensures that only the most credible predictions contribute to the final ensemble output. Our empirical and statistical analyses of time series data from New Zealand’s milk collection and the global health data from the public owid-monkeypox dataset show that the CAGE outperforms traditional ensemble methods, especially in handling outliers and noisy data. By incorporating conformal prediction, CAGE delivers accurate and statistically rigorous forecasts, enhancing decision-making. We have demonstrated performance on two different datasets deliberately to showcase that the proposed method offers a versatile solution potentially applicable across finance, weather, and supply chain management.
[AI-238] A Moving-Horizon Approximate Branch-and-Reduce Method for Deep Classification Trees
链接: https://arxiv.org/abs/2609.38194
作者: Chenxuanyin Zou,Jiayang Ren,Qiangqiang Mao,Jing Liu,Marcus Lai,Yankai Cao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Optimization and Control (math.OC)
备注: J2C Certification
Abstract:Despite the importance for interpretability, decision trees face severe scalability challenges. Existing global optimal methods are often limited by binary feature selection and shallow tree depths, whereas traditional heuristic approaches frequently sacrifice predictive accuracy. To overcome these limitations, this paper proposes a moving-horizon approximate branch-and-reduce method to train near-optimal deep classification trees on large-scale datasets with continuous features. Built on a hierarchical root-subtree optimization framework, the method solves the root-level problem via branch-and-reduce while approximating the induced subtree problem using greedy heuristics. Although the underlying framework is capable of guaranteeing global optimality, the approximation, which functions as a lookahead rollout in a reinforcement learning context, significantly boosts efficiency for deeper structures. A low-cost moving-horizon strategy is then employed to iteratively refine model accuracy. Extensive numerical results demonstrate that our method exceeds the testing accuracy of existing heuristic baselines while offering significantly greater scalability, in terms of both dataset size and tree depth, than global optimal solvers.
[AI-239] EHR2Trace: Auditable EHR Data Infrastructure for Patient World Models and Clinical Agents
链接: https://arxiv.org/abs/2609.38193
作者: Xinye Yang,Yuli Wang,Cheng Ting Lin,Harrison Bai
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Databases (cs.DB); Quantitative Methods (q-bio.QM)
备注: 13 pages, 3 figures, 6 tables. Code and experiment records: this https URL
Abstract:Patient world models and clinical agents aim to predict changes in patients’ health and support clinical work. Developing these systems requires reliable histories of patient conditions, treatments, and the information available at each decision. Electronic health records (EHRs) contain these histories, but differences in how events are recorded make them difficult to use consistently. We present EHR2Trace, a system that converts EHRs from different sources into traceable patient events for model training and evaluation. It links events to source records, separates event time from information availability, and distinguishes medication orders, dispensing, and administration. A shared event representation supports both OMOP and MEDS exports, with automated validation and reproducible builds. Across three clinical datasets, EHR2Trace converted 846.4 million events, with every applicable check passing except one unit-consistency check on MIMIC-IV, and detected all 28 injected faults. A controlled prediction experiment showed that assigning later diagnoses to admission time substantially inflated measured performance, and that a model trained on such data lost accuracy when deployed on histories filtered by availability. EHR2Trace provides a reusable data foundation for patient world models and clinical agents, helping researchers inspect patient histories, check conversion decisions, and evaluate models with explicit data rules.
[AI-240] ravel Time Prediction in Supply Chain Management Using Machine Learning
链接: https://arxiv.org/abs/2609.38190
作者: Balaji Venkateswaran
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 50 pages, 24 figures, 8 tables
Abstract:The purpose of this research is to find data and methods using machine learning and deep learning to correctly predict the estimated travel time for transportation and logistics in a supply chain system. The supply chain ecosystem is very complex and heavily relies on the transportation and logistics of raw materials and finished goods. Accurate travel time estimation is critical because it helps supply chain members to improve logistics consistency and performance. This helps in planning, demand forecasting, lead time management and assembly planning. The logistics on the delivery side of the customer also plays a crucial role in customer satisfaction and voice of customer. With the collection of huge historical data and using novel techniques, the research builds an accurate model to predict travel time of inventory.
[AI-241] Argus: Academic Integrity in the Era of Generative AI
链接: https://arxiv.org/abs/2609.36073
作者: David Racovan,Ajay Rawat,Christopher K. May,Jeffrey A. Turkstra
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: 7 pages, 7 figures
Abstract:The rapid proliferation of large language models (LLMs) in the context of education has introduced significant challenges in enforcement of academic integrity, especially in programming courses. We present Argus, an automated detection system for LLM-assisted student work in undergraduate C programming assignments. Argus integrates behavioral and stylistic indicators to create a holistic picture of the student’s progress through an assignment and surfaces anomalies that point to potential misuse of LLM assistance. We quantify and analyze data over six years of Spring semester offerings in a large-enrollment CS2 course at Purdue University using Argus, finding that 45% of enrolled students exhibited patterns consistent with LLM-assisted code development in Spring 2026. To contextualize these results, we analyze the relationship between flagged LLM use and student performance on written, in-person proctored examinations, and find a significant negative correlation. We also explore the problem of mitigating false positives, recognizing that erroneous accusations of academic integrity carry significant consequences for students and instructors alike, particularly in the context of large enrollment courses. We argue that any automated detection system must be accompanied by a structured process for human review. We discuss the consequences for future course design, changing academic policy as these tools become more ubiquitous, and the pedagogical implications of LLM-based tools in computer science education.
[AI-242] Grounding Time-Series Foundation Models in Digital Twin Topology for Predictive Maintenance
链接: https://arxiv.org/abs/2609.40071
作者: Sizhe Ma,Katherine A. Flanigan,Mario Bergés
类目: ignal Processing (eess.SP); Artificial Intelligence (cs.AI)
备注: Submitted to Reliability Engineering \ System Safety (RESS)
Abstract:Digital twins increasingly support downstream analytical tasks that depend on time-series data, motivating interest in time-series foundation models (TSFMs) as scalable backbones. However, TSFMs are primarily pretrained for temporal continuation and often underperform on unseen tasks such as regression, and systematic empirical comparisons against state-of-the-art dedicated models in digital twin contexts remain limited. This paper makes three contributions. First, we benchmark five well-known TSFMs with frozen backbones on remaining useful life (RUL) prediction using the C-MAPSS dataset, finding that multivariate architectures substantially outperform univariate ones, particularly under varying operating conditions. This raises a deeper question: when cross-channel dependencies can be modeled through pretrained weights, target-task adaptation, and digital twin-derived representations, how much does each contribute, and are they complementary? Second, we propose a topology-informed fusion approach in which topological constraints, derived from the asset structure the digital twin stores among its information models, explicitly shape cross-attention, so that fused representations respect the physical system’s local connectivity rather than relying on unconstrained all-to-all interactions. Third, we conduct an ablation study across C-MAPSS subsets of varying operational complexity that isolates the three sources and their interactions. The sources prove complementary rather than redundant, and topology-constrained attention outperforms unconstrained fusion, though by a small margin, enabling a frozen TSFM informed by digital twin representations to remain competitive or in some cases exceed state-of-the-art performance on this regression task.
[AI-243] Cluster Attention Neural Operators for Solving Parametric Partial Differential Equations
链接: https://arxiv.org/abs/2609.39914
作者: Ming Zhong,Antonio Colanera,Gianluigi Rozza,Zhenya Yan
类目: Mathematical Physics (math-ph); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Dynamical Systems (math.DS)
备注: 30 pages, 9 figures
Abstract:Traditional simulations of parametric partial differential equations (PDEs) rely on repetitive computations for each parameter, which makes high-fidelity design impractical. Neural operators address this issue by learning solution operators, accelerating parameter-space mapping by orders of magnitude. Recent Transformer-based neural operators attempt to capture global dependencies, but often at the cost of quadratic attention complexity. Transolver resolves this problem by projecting physical states into a reduced slice space for attention computation. Although fast, this projection sacrifices fine spatial information. Moreover, by operating in this reduced space with shared weights across attention heads, it may constrain the model’s flexibility, thereby limiting its capacity to capture complex phenomena. To address these issues, we propose the Cluster Attention Neural Operator (CANO), which reformulates attention via a novel cross-attention mechanism that dynamically clusters queries while preserving full-resolution keys and values. This avoids slice compression loss and removes weight-sharing limits. At the same time, the model remains fast without losing global interactions. Empirically, CANO achieves state-of-the-art performance across canonical PDE benchmarks, covering fluid and solid dynamics (e.g., Navier-Stokes, Airfoil, Plasticity), irregular unstructured geometries (e.g., Pipe Turbulence, Composites), and long-term temporal rollouts. Across solid deformation and turbulent flow benchmarks, CANO achieves lower errors than baselines and exhibits strong geometric adaptability and temporal consistency.
[AI-244] BayesNDE: Bayesian Generative Modeling for Neural Density Estimation
链接: https://arxiv.org/abs/2609.39843
作者: Chenglin Li,Qiao Liu
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Methodology (stat.ME)
备注:
Abstract:Density estimation is a fundamental problem in statistics and machine learning. In this work, we introduce BayesNDE, a neural density estimator based on Bayesian generative modeling. BayesNDE learns a Bayesian generative model and evaluates its density without requiring invertible networks or Jacobian-determinant computation. For each observation, it infers a sample-specific latent posterior to construct an adaptive proposal that focuses computation on regions contributing most to its density. Bridge sampling then combines samples from this proposal with separate posterior samples to estimate the density. Experiments on nonlinear and multimodal synthetic datasets show improved estimation of density values and better recovery of the density structure compared to the state-of-the-art neural density estimators. Applications to real-world datasets further demonstrate improved anomaly detection. Together, these results highlight BayesNDE as a flexible and effective neural density estimator, demonstrating how posterior inference can turn generative models into tools for density estimation. The code and tutorials are available at this https URL.
[AI-245] CAMOS: Coupled Oscillatory State-Space Model for Multimodal Clinical Time-Series
链接: https://arxiv.org/abs/2609.39484
作者: Maxx Richard Rahman,Mostafa Hammouda,Wolfgang Maass
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI)
备注:
Abstract:Longitudinal clinical cohorts are multimodal, irregularly sampled and pervasively incomplete: in ADNI, positron emission tomography and cerebrospinal fluid assays are absent from roughly half of all visits. Linear state-space models handle irregular sampling gracefully but treat a missing modality by masking the input, leaving the transition operator untouched. We prove that this is a representational limitation: the latent state of any linear state-space layer whose transition operator does not depend on the availability pattern is an additive function of the availability indicators, so no such layer can represent an interaction between two modalities being jointly present or jointly absent. We propose CAMOS, which gives each modality a bank of second-order oscillators coupled through a matrix that sits inside the differential equation and is gated by availability, so the transition operator itself becomes a function of which measurements were taken. Coupling invalidates the analysis of uncoupled oscillatory models, and we restore it: a per-channel Gershgorin budget makes the effective stiffness positive definite uniformly over all 2^M availability patterns and all gaps, an energy argument charges amplification to availability transitions rather than sequence length, and a channel factorization preserves exact associative parallel scans. On ADNI, CAMOS outperforms uncoupled oscillatory state-space models and clinical fusion models on same-visit staging, landmark prediction and longitudinal forecasting, and under zero-shot transfer to OASIS-3 it is the only model that avoids collapse to the majority class.
[AI-246] Belief-Based Maximum Occupancy Principle and Active Inference
链接: https://arxiv.org/abs/2609.39342
作者: Manolis Mylonas,Rubén Moreno Bote
类目: Neurons and Cognition (q-bio.NC); Artificial Intelligence (cs.AI)
备注: Accepted at the 7th International Workshop on Active Inference (IWAI 2026, Madrid). To appear in Springer CCIS proceedings
Abstract:Intrinsic motivation plays a central role in adaptive and goal-directed behavior by conferring agents reward-independent objectives and biases useful to act in noisy and uncertain environments. Active Inference addresses the problem of acting in a partially observable environment through a principled framework for belief updating and action selection. A key component of Active Inference is the specification of prior preferences, which shapes behavior by encoding desirable future outcomes. An intrinsic motivation approach called the Maximum Occupancy Principle (MOP) proposes that agents act so as to maximize occupancy over future paths of states and actions, with no preferences or epistemic targets. Despite its simple formulation, MOP gives rise to rich and adaptive behaviors that combine exploratory variability with goal-directed dynamics. In this work, we extend MOP to partially observable environments and introduce a Bellman reformulation of the Expected Free Energy for Active Inference, both incorporating belief-based inference over hidden states as part of the agent state. The Bellman formulation enables tractable offline computation via value iteration over the full belief-state space. We compare the resulting behaviors in a set of minimal experimental settings with uncertain food sources. We find that MOP agents switch between goal-directed (food seeking) behavior and exploration between different food sources, depending on their energy available and their belief state. In contrast, Active Inference agents mostly inhabit regions around a single food source, a strategy having both high pragmatic and epistemic value. We finally compare with Empowerment, which is shown to be qualitatively similar to Active Inference.
[AI-247] Association profile conditioning in a set-temporal transformer for cross-session intracortical motor decoding ICASSP2027
链接: https://arxiv.org/abs/2609.39080
作者: Xinyuan Zhang,Handong Mo,Pengfei Wen,Shuang Liang,Jichang Yang,Yan Zeng,Zhongrui Wang,Han Wang
类目: Neurons and Cognition (q-bio.NC); Artificial Intelligence (cs.AI)
备注: 5 pages, 3 figures. Submitted to ICASSP 2027
Abstract:Intracortical motor decoders degrade across sessions because the set of recorded units changes and persisting units can alter how their firing relates to behavior. Most existing methods update network weights on each new session or rely on unlabeled activity, which does not directly reveal such changes. We present APST, an Association Profile-conditioned Set-Temporal transformer that adapts to new sessions with all network weights frozen. From a few labeled calibration trials, APST summarizes how each unit’s firing relates to behavior in a four-dimensional association profile computed in closed form. The profiles condition a set-attention encoder that accepts any number and order of units, followed by a causal transformer for streaming decoding. On held-out DANDI688 sessions from two monkeys, APST reaches velocity R^2 of 0.78 and 0.81 , versus 0.40 and 0.58 for a variant that uses neural activity alone, and matches or exceeds an RNN fine-tuned on the same trials. On FALCON private held-out evaluation, it attains R^2 of 0.65 , 0.42 , and 0.44 on M1, M2, and H1.
[AI-248] An Uncertainty-Guided Digital Twin Framework for Online Adaptive Proton Therapy in Head and Neck Cancer: A Feasibility Study
链接: https://arxiv.org/abs/2609.39010
作者: Yizhou Wu,Ryan J. Sanford,Huiqiao Xie,Jie Ding,Shupeng Chen,Tung-Ho Wu,Ping-Hsiu Wu,Justin Roper,Jun Zhou,Minglei Kang,Bill Stokes,Sibo Tian,David S. Yu,Xiaofeng Yang,Chih-Wei Chang
类目: Medical Physics (physics.med-ph); Artificial Intelligence (cs.AI)
备注:
Abstract:Objective: Head and neck (HN) proton therapy spans six to seven weeks of anatomical change, while offline replanning takes about a week. We present an uncertainty-guided digital twin (UGDT) framework that forecasts treatment-day anatomy before treatment and evaluate whether it generates online adaptive proton therapy (APT) plans of clinical quality. Approach: A library of 302 longitudinal deformations from 88 previously treated HN patients was transported onto each new patient’s treatment planning CT (TPCT) using two-step multi-atlas deformable image registration (DIR) built on a pretrained CT foundation model, generating about 284 predicted CTs (pdCTs) with contours per patient. Dispersion of propagated clinical target volume (CTV) contours defined a patient-specific robust margin. In ten patients, the quality assurance CT (QACT) triggering a replan represented treatment-day anatomy, and the physician-approved replan was the baseline. The pdCT most similar to the QACT (pdCT-H) and one from the lowest quartile (pdCT-L) were planned to within about 5% of baseline plan quality, forward-calculated on the QACT, and reoptimized to generate online APT plans. Main results: pdCT plans scored within -0.7% (pdCT-H) and -1.0% (pdCT-L) of baseline. Forward calculation on QACT reduced high-dose CTV D98% to 88.3% and 85.5%. After online reoptimization, D98% recovered to 98.3 +/- 0.3% and 98.2 +/- 0.3%, versus 98.5 +/- 0.4% at baseline. Spinal cord and brainstem doses remained below tolerance, and plan quality scores were within -1.1% (p = 0.19) and -1.7% (p = 0.01) of baseline. Significance: UGDT generated online APT plans comparable in quality to physician-approved offline replans using anatomy forecast before treatment, enabling a transition from reactive offline replanning toward anticipatory online adaptation.
[AI-249] CellM SA: Context Modeling for Single-Cell Representation Learning NEURIPS2026
链接: https://arxiv.org/abs/2609.38908
作者: Suyuan Zhao,Minghao Liu,Yizhen Luo,Zaiqing Nie
类目: Genomics (q-bio.GN); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted by NeurIPS 2026, code released
Abstract:Single-cell transcriptomics enables profiling of cellular states at unprecedented resolution, but its high dimensionality, sparsity, and technical batch effects pose significant challenges for representation learning. Existing single-cell foundation models typically encode each cell independently or only model cells from the same batch for denoising, thereby underutilizing the rich relational information across batches and cell types to model gene expression patterns. We argue that single-cell models can benefit from more informative cell-context modeling. By comparing consistency and variation across cells, models can capture fine-grained gene-gene dependencies associated with cell states, which are essential for learning high-quality representations. Inspired by the use of multiple sequence alignment (MSA) context in protein modeling, we propose CellMSA, a single-cell representation learning framework that introduces an MSA-inspired inductive bias into transcriptomic modeling. For each target cell, CellMSA retrieves relevant cells from different batches and biologically related cell types as context, and summarizes cross-cell patterns into a context-dependent gene-pair representation. This representation is then injected into a pair-aware target-cell encoder for fine-grained representation learning. We pretrain CellMSA on a large-scale human single-cell corpus of approximately 109 million cell observations, including 65.6 million primary observations. Experiments show that our framework consistently outperforms existing methods across multiple benchmarks. Code is available at the following repository: this https URL.
[AI-250] Always-On Experimentation
链接: https://arxiv.org/abs/2609.38695
作者: Ricardo J. Sandoval,David Arbour,Avi Feller,Michael I. Jordan
类目: Methodology (stat.ME); Artificial Intelligence (cs.AI)
备注:
Abstract:Generative AI has dramatically accelerated the rate at which new treatments—from novel pharmaceuticals to online marketing campaigns—can be conceived and deployed. As a result, modern experimentation platforms often run continuously, with treatments added as they are ready and removed when they underperform. We formalize this “Always-On” experimental setting, in which treatments can be dynamically generated, added to, and removed from a running experiment, and study the statistical problem of deciding whether to accept or reject each treatment while controlling for the false discovery rate. We develop sequential tests that achieve time-uniform Type-I error control under arbitrary stopping times and “predictable” treatment schedules. Our approach builds on the testing-by-betting framework: we construct test supermartingales for testing the average treatment effect of each treatment, and show that the construction of these test supermartingales is growth-rate optimal in an almost-sure sense.
[AI-251] Bandits with Multiple Optimal Arms: Minimax Regret and Non-Adaptivit
链接: https://arxiv.org/abs/2609.38659
作者: Kaixuan Ji,Qiwei Di,Qingyue Zhao,Heyang Zhao,Quanquan Gu
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Statistics Theory (math.ST); Methodology (stat.ME)
备注:
Abstract:We study multi-armed bandits (MAB) with multiple optimal arms, motivated by the fact that many practical decision making problems admit multiple correct answers. For K -armed bandits with A optimal arms, we first provide a sharper analysis of previous sub-sampling algorithms (De Heide et al., 2021; Zhu and Nowak, 2020), establishing a \tildeO\Big(\fracK-A\sqrtKA\sqrtT \Big) minimax regret, where T is the total number of interactions and \tilde O(\cdot) drops all constant and logarithmic factors, improving the previous \tildeO(\sqrtKT/A) regret. We then provide a matching lower bound up to logarithmic factors, indicating that our established rate is nearly minimax-optimal. We further show that the knowledge of A up to \tildeO(1) factors is necessary to achieve near-optimal regret, as near-optimal algorithms for one number of optimal arms must incur substantially larger regret than optimal regret for a smaller number. Overall, our results provide a comprehensive minimax characterization of K -armed bandits with A over the entire range of 1 \leq A \leq K-1 .
[AI-252] ACIT: Optimization Models that Learn from Their Mistakes
链接: https://arxiv.org/abs/2609.38434
作者: Maxime Bouscary,Marco Molinaro,Sirui Li,Saurabh Amin,Ishai Menache,Konstantina Mellou
类目: Optimization and Control (math.OC); Artificial Intelligence (cs.AI)
备注:
Abstract:Real-world optimization problems are difficult to model accurately because many objectives and constraints reside in domain experts’ tacit knowledge, making them hard to formalize. As a result, optimization models often contain miscalibrated objectives, missing constraints, or omitted decision variables, leading to solutions that fail to reflect operational realities. We address this challenge by automatically repairing misspecified formulations using historical data consisting of past solutions and subsequent user overrides. Traditional approaches such as inverse optimization and constraint learning tend to overfit sparse data and produce complex formulations. Our central idea is to combine the reasoning capabilities and prior knowledge of LLMs with the formal grounding provided by optimization. We realize this idea through two complementary paradigms. Top-down, an LLM proposes structural repairs, including new constraints and variables, whose numerical parameters are calibrated and validated through optimization. Bottom-up, optimization infers cuts from observed decisions, which the LLM contextualizes into interpretable, generalizable modeling constraints. We evaluate our approach on 38 misspecification scenarios spanning nine classes of optimization problems, several drawn from real-world applications, and show that TACIT can repair 78.9% of them (vs. 60.5% for the best baseline).
[AI-253] Acceleration of Diffusion Language Model through Discrete Averag e Generator
链接: https://arxiv.org/abs/2609.38364
作者: Yidong Ouyang,Zhengyan Wan,Themis Haris,Tian Tan,Liqian Peng,Henry Li,Ziqian Lin,Jianhang Chen,Maryam Karimzadehgan,Alec Go,George Michailidis
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Discrete diffusion models and flow matching have emerged as powerful frameworks for generative modeling over discrete state spaces, yet efficient few-step generation remains a fundamental challenge. In this work, we introduce the Discrete Average Generator, a principled extension of MeanFlow to Continuous-Time Markov Chains (CTMCs). Analogously to how MeanFlow defines an average velocity field over a time interval in continuous spaces, we define an average generator as the normalized increment of the transition kernel over a time interval. We show that this average generator satisfies a self-consistency identity, which provides the foundation for our training objective. We further develop training strategies that align with the standard training paradigm of diffusion language models while keeping the resulting objective tractable. When projected onto per-coordinate marginals, the self-consistency identity admits a closed-form expression, enabling efficient training and inference. In Potts model simulations, our objective reduces the total variation distance of the K -step sampler by up to 67%. On OpenWebText, our method achieves the lowest generative perplexity among the evaluated methods for 8 to 64 sampling steps while enabling a 16\times acceleration, and achieves comparable performance to existing methods on ImageNet.
[AI-254] Searching for BSM Experimental Signatures with Large Lagrangian Models
链接: https://arxiv.org/abs/2609.38309
作者: Ibrahim Elsharkawy,Victoria Knapp-Perez,Wahid Bhimji,Aishik Ghosh
类目: High Energy Physics - Phenomenology (hep-ph); Cosmology and Nongalactic Astrophysics (astro-ph.CO); Artificial Intelligence (cs.AI); High Energy Physics - Experiment (hep-ex)
备注: 47 pages, 27 Figures
Abstract:The search for physics Beyond the Standard Model (BSM) is generally limited not by the supply of theory descriptions but by the lack of discriminating experimental observations. A case in point is dark matter, where the overwhelming gravitational evidence only goes so far in distinguishing between models within a vast theory space. Exploring the space of testable model signatures may help identify overlooked experimental observables and indicate the utility of future experiments. A challenge is designing a search through model signatures outside what is found in the literature. Our primary contribution is hAIthem, a framework that combines the self-guided exploration of reinforcement learning (RL) with the broad literature-derived knowledge of LLMs. We build an RL agent that learns to find which portions of a theory’s high-dimensional parameter space are not excluded under some subset of constraints by playing a Battleship-style “game” against a suite of phenomenology tools. The agent is built as a Large Lagrangian Model (LLaM), an autoregressive transformer that reads a tokenized Lagrangian, is pretrained at scale (here on ~1 billion tokens from ~10,000 Lagrangians), and is fine-tuned in a live environment. The framework then constructs a decision tree that separates RL-found regions using observables computed with established tools, and passes the remaining degenerate regions to a set of LLM agents that compete to produce realistic signatures. In this proof of concept, RL-search outperforms an evolutionary-algorithm baseline, finding more viable regions with greater physical diversity. In a restricted space of single dark scalar multiplet models, we find that hAIthem proposes interesting combinations of previously studied observables, such as the application of a halo-independent kinematic ratio to paleo-detectors.
[AI-255] When Does Randomized Oversight Align AI Agents That Can Conceal?
链接: https://arxiv.org/abs/2609.38262
作者: Joshua S. Gans,Richard Holden
类目: Theoretical Economics (econ.TH); Artificial Intelligence (cs.AI)
备注: 42 Pages, 2 Figures, 4 pages Online Appendix
Abstract:Oversight changes the evidence it relies on. We ask when randomized audits and scoring align AI agents that can conceal misconduct and alter records. Stronger auditing makes undeterred violations better hidden. Because the provider writes the agent’s objective, sanctions need not stop at forfeiture, and rare audits deter every type of agent if evidence survives concealment and audit draws cannot be learned in advance. When evidence can be erased, deterrence must come from lower gains from violation, such as credit for stopping, or from costlier or fewer ways to conceal. These conditions identify what failed when agents in OpenAI’s cybersecurity evaluations compromised parts of Hugging Face’s infrastructure in July 2026.
机器学习
[LG-0] Removing Timing Shortcuts Improves Non-Invasive Brain-to-Text
链接: https://arxiv.org/abs/2609.40359
作者: Dulhan Jayalath,Oiwi Parker Jones
类目: Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注: 29 pages, 12 figures, 10 tables
Abstract:We find that major reported improvements in decoding words from non-invasive brain recordings are largely reproducible without any brain data. In the influential work of d’Ascoli et al. (2025), time series of brain activity from subjects perceiving continuous speech are segmented into fixed-length windows starting at each word. A neural network then generates predictions for all of the words in a sentence together. Neighbouring windows partially overlap, implicitly revealing the interval between words. Since these intervals indicate the duration of the words spoken, and different words tend to have different durations - for example, “the” is much shorter than “supercalifragilisticexpialidocious” - the neural network can improve its predictions of words without relying on the underlying brain activity. Consistent with this, the method reaches 22.0% balanced accuracy on synthetic signals containing no brain information, compared with 22.3% on real brain recordings. To prevent the network from learning this shortcut, we make a single, simple change. Instead of jointly encoding all windows in a sentence, we process each independently. As a result, the neural network achieves better performance by learning underlying word-specific information from brain recordings. This makes two existing strategies become much more effective than before. Both aggregating predictions from distinct neural responses to the same word and using a pretrained LLM as a linguistic prior now substantially improve results. On our perceived speech benchmark, this simple recipe (SimpleB2T) achieves a word error rate of 36.6% with five observations per word, approaching past invasive speech decoding performance, albeit under different conditions. The results in this work expose an important shortcut in brain-to-text decoding and show that removing it leads to a simple and considerably more effective strategy.
[LG-1] Is Weight Tying Still Beneficial for Decoder-Only LLM s in Private Settings Under DP-SGD?
链接: https://arxiv.org/abs/2609.40335
作者: Razan El Mais,Ali Chehab,Ibrahim Issa,Razane Tajeddine
类目: Machine Learning (cs.LG)
*备注: Accepted at the 8th IEEE International Conference on Trust, Privacy and Security in Intelligent Systems, and Applications (IEEE TPS 2026). 12 pages (10 pages of main content), 1 figure, 13 tables
Abstract:Differentially Private Stochastic Gradient Descent (DP-SGD) is a leading approach for privacy-preserving fine-tuning of large language models (LLMs). Many decoder-only LLMs employ weight tying between input and output embeddings, a design choice originally introduced for parameter efficiency and improved language modeling performance in the non-private setting. However, the impact of weight tying under differentially private training remains largely unexplored. In this work, we investigate the role of weight tying in the DP setting using GPT2 and DistilGPT2 as representative decoder-only architectures. Interestingly, we find that untied embeddings consistently outperform weight-tied models under DP-SGD, achieving gains of up to 4.74% points in accuracy on SST-2, QNLI, and QQP. Beyond improved utility, untying embeddings enables the use of memory-efficient ghost clipping for DP-SGD. By contrast, weight tying introduces shared-parameter interactions that complicate standard ghost norm computation and largely negate its computational advantages. As a result, untied models achieve over 60% lower memory usage while preserving the benefits of ghost clipping. Our results indicate that untied embeddings provide a more effective and scalable design for differentially private training of decoder-only LLMs and highlight the need to revisit standard LLM architectural choices in the privacy-preserving setting.
[LG-2] Compression Footprints as Security Signals for Model-Poisoning Defense in Federated Learning
链接: https://arxiv.org/abs/2609.40312
作者: Sachi Shome,William Eiers
类目: Machine Learning (cs.LG)
*备注:
Abstract:Lossy compression is widely used in Federated Learning (FL) but is generally treated as an error source, while conventional poisoning defenses inspect update geometry. In this work, we instead treat the compressor’s response as a security signal: the input-dependent distortion and payload behavior induced by lossy compression can expose differences between honest and attack-generated updates. We introduce the concept of a \emphcompression footprint: the low-dimensional collection of reconstruction, directional, sparsity, and payload statistics induced by a lossy compressor. We characterize sufficient conditions under which compression footprints separate honest and malicious updates, and operationalize our findings in the CRAFT (\emphCompression-guided Robust Aggregation via Footprint Trust) server-side robust aggregation method. Crucially, under a strict honest-majority assumption, CRAFT uses server-verifiable footprints, requires no client-side metadata nor knowledge of the number of malicious clients, and adds no communication beyond the compressed FL pipeline. Moreover, while CRAFT assumes a strict honest majority, it does not require the number of malicious clients to be known in advance. We observe that error-bounded lossy compressor (EBLC) footprints provide stronger separation than Top-K footprints and that footprint trust suppresses malicious influence. We evaluate CRAFT under IID client data with 36% malicious participation across six standard model-poisoning attacks, three datasets, and six robust aggregation baselines, finding that CRAFT consistently achieves the best accuracy in 7 out of 18 settings and within 1.7 percentage points of the best in the others. Our results show that lossy compression can serve as both a communication mechanism and a security signal for robust aggregation in FL.
[LG-3] Disentangling Computation in Multi-Task Neural Networks with the Greens Operator
链接: https://arxiv.org/abs/2609.40292
作者: James Hazelden
类目: Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注: Accepted as a poster at NeurReps 2026
Abstract:How is computation organized and reused across tasks and time in a trained recurrent network? Most analyses emphasize the geometry of neural activity, dynamical motifs, or local perturbation growth. We instead study the network’s global first-order perturbation response. The finite-horizon Green’s operator maps perturbations at each source along a trajectory to their downstream state-space responses and therefore directly represents perturbation routing. Simple reductions of this operator provide task-to-task and time-to-time views of the same computation, while matrix-free products make these views accessible without constructing the full operator. In a flexible multitask recurrent network, task reductions reveal structured reuse of known computational motifs, while temporal reductions reveal causal pathways and how they emerge during training. Our main point is simple: the Green’s operator provides a global response geometry for mapping the organization of learned dynamical computation.
[LG-4] PMosFM: Preconditioned Manifold Matching for One-Step Physics-Constrained Generation
链接: https://arxiv.org/abs/2609.40287
作者: Zhangyong Liang,Haibin Ling
类目: Machine Learning (cs.LG)
*备注:
Abstract:Physics-constrained generative models aim to generate physical fields that match a target distribution and satisfy prescribed constraints. However, enforcing these constraints often increases sampling costs through iterative corrections or training costs through residual optimization and trajectory unrolling. To address this issue, we introduce \textbfPreconditioned \textbfManifold \textbfone-\textbfstep \textbfFlow \textbfMatching (\textbfPMosFM), a preconditioned manifold matching framework for one-step physics-constrained generation. By encoding constraints in a manifold decoder, PMosFM learns transport in intrinsic coordinates without separate residual losses or terminal residual unrolling. A geometric preconditioner rescales coordinates using the decoder-induced metric, while a regularized covariance transform approximately whitens the interpolation-state inputs. A finite-interval objective couples velocity supervision with consistency between decoded endpoints in physical space. We show that exact parameterization removes residual-induced Gauss–Newton curvature, that geometric and covariance effects separate in a local conditioning bound, and that physical flow-map error bounds endpoint distributional error. Controlled ablations examine conditioning, and experiments evaluate optimizer-update time and memory footprint. At inference, PMosFM uses one neural transport evaluation followed by physical decoding. Experiments across benchmarks show lower training and sampling time than the multi-step baselines at comparable physical and distributional fidelity. Code and datasets will be released publicly.
[LG-5] OpenTSLM TeeMoE: A Unified Time-Series Language Model for Forecasting Contextual Prediction and Reasoning
链接: https://arxiv.org/abs/2609.40265
作者: Tony Chen,Timo Stoffregen,Maxwell Xu,Thomas Kaar,Martin Maritsch,Geremia Pompei,Nicolas Zumarraga,Robert Jakob,Paul Schmiedmayer,Patrick Langer,Juncheng Liu
类目: Machine Learning (cs.LG)
*备注: 39 pages, 2 figures. Code: this https URL ; model: this https URL
Abstract:Real-world time-series applications increasingly require models that can handle time series forecasting, context-conditioned prediction, and language-based temporal reasoning. Yet current time-series foundation models remain fragmented across these capabilities: numerical specialists often provide the strongest forecasts, while language-based models offer broader contextual understanding and analysis. A central challenge is to unify these heterogeneous capabilities without reducing their individual performance. We introduce OpenTSLM TeeMoE, a generalist time-series language model that can forecast directly from observed time series, reason over textual context and temporal patterns, and synthesize and refine predictions from external numerical forecasting specialists. We independently train three low-rank experts for forecast aggregation, native forecasting, and temporal analysis over a shared backbone. A learned LoRA mixture-of-experts controller then weights their frozen parameter updates for each request. Our proposed model achieves strong performance on widely used benchmarks for time series forecasting, context-conditioned prediction, and language-based temporal reasoning, ranking among the top three on GIFT-Eval by mean MASE rank, Context is Key by RCRPS, and TimeSeriesExam by accuracy.
[LG-6] STARS: From Spatiotemporal Dynamics to Social Representations in Human-Robot Interaction
链接: https://arxiv.org/abs/2609.40245
作者: Nathan Tsoi,Michael J. Munje,Tejas Oberoi,Rishab Maheshwari,Pengen Zheng,Tanush Chauhan,Peter Stone,Joydeep Biswas
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: Conference on Robot Learning (CoRL) 2026. First two authors contributed equally. Project site: this https URL
Abstract:Robot navigation in dynamic, human-centered environments requires socially-compliant decisions grounded in robust scene understanding. Recent Vision-Language Models (VLMs) exhibit promising capabilities such as object recognition, common-sense reasoning, and contextual understanding, capabilities that align with the nuanced requirements of social robot navigation. However, it remains unclear whether VLMs can accurately understand complex social navigation scenes (e.g., inferring the spatial-temporal relations among agents and human intentions), which is essential for safe and socially compliant robot navigation. While some recent works have explored the use of VLMs in social robot navigation, no existing work systematically evaluates their ability to meet these necessary conditions. In this paper, we introduce the Social Navigation Scene Understanding Benchmark (SocialNav-SUB), a Visual Question Answering (VQA) dataset and benchmark designed to evaluate VLMs for scene understanding in real-world social robot navigation scenarios. SocialNav-SUB provides a unified framework for evaluating VLMs against human and rule-based baselines across VQA tasks requiring spatial, spatiotemporal, and social reasoning in social robot navigation. Through experiments with state-of-the-art VLMs, we find that while the best-performing VLM achieves an encouraging probability of agreeing with human answers, it still underperforms simpler rule-based approach and human consensus baselines, indicating critical gaps in social scene understanding of current VLMs. Our benchmark sets the stage for further research on foundation models for social robot navigation, offering a framework to explore how VLMs can be tailored to meet real-world social robot navigation needs. An overview of this paper along with the code and data can be found at this https URL.
[LG-7] Near-Linear Accuracy Bounds for Moreau–Yosida Unadjusted Langevin Sampling
链接: https://arxiv.org/abs/2609.40193
作者: Yuchen Xin,Zhihua Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:We establish near-linear accuracy bounds for the classical Moreau–Yosida unadjusted Langevin algorithm (MYULA). The target is \pi\propto e^-f-g , where f\in C^2(\mathbbR^d) is m -strongly convex with Lipschitz gradient and g is convex and globally Lipschitz. Under an explicit parameter-dependent step-size condition, we bound the invariant-measure bias relative to the Moreau-smoothed target by \widetilde O(h) , with only logarithmic dependence on the inverse smoothing parameter in the error coefficient. Combining this estimate with the Moreau approximation bias and Wasserstein contraction gives \widetilde O(\varepsilon^-1) iterations to make the N th-iterate law \mu_N satisfy \sqrt m,W_2(\mu_N,\pi)\le\varepsilon , for fixed model parameters and initialization. We bound the stationary error directly, without assuming third derivatives or a Lipschitz Hessian. Each iteration uses one gradient evaluation and one exact proximal evaluation. The key idea in our analysis is to convert a second-order stationary residual into a Wasserstein bound using a Poisson-based estimate.
[LG-8] MANET-GNN: Learned Decentralized Optimization of Power Allocation in Multi-Channel MANETs
链接: https://arxiv.org/abs/2609.40170
作者: Tomer Alter,Nir Shlezinger,Michael Segal
类目: Machine Learning (cs.LG)
*备注:
Abstract:MANETs enable flexible infrastructure-less wireless connectivity in dynamic and resource-constrained environments. As modern MANETs exploit multiple frequency channels and support heterogeneous traffic patterns, decentralized transmit-power allocation becomes increasingly challenging. We develop a unified learned optimization framework for decentralized power allocation in dynamic multi-hop, multi-channel MANETs. We formulate a constrained end-to-end throughput maximization problem covering unicast, multicast, multicommodity, convergecast, and many-to-many communication. Although centralized and non-convex, this problem serves as an unsupervised training objective for MANET-GNN, a message-passing GNN that operates as a distributed learned optimizer. MANET-GNN uses only local, possibly noisy, CSI and a prescribed number of neighbor message exchanges, enabling low-latency decentralized inference while generalizing across topologies and network sizes. Numerical results show that MANET-GNN achieves centralized-competitive performance across communication frameworks, remains robust to channel uncertainty, and scales effectively across MANET configurations.
[LG-9] Reinforcement Learning-Guided Graph Transformations for SpTRSV Optimization
链接: https://arxiv.org/abs/2609.40159
作者: Buse Yılmaz
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: 33 pages, 3 figures, 7 tables. Submitted to The Journal of Supercomputing and currently under review
Abstract:Sparse triangular solve (SpTRSV) is a fundamental kernel in numerous scientific and engineering applications. However, the data dependencies inherent in sparse triangular matrices significantly limit the available parallelism and make efficient workload distribution challenging. Recent graph transformation techniques address these limitations by modifying the dependency graph of the input matrix to improve parallel execution. Existing graph transformation strategies, however, rely on manually designed heuristics, making their development and adaptation to different optimization objectives challenging. This work proposes a reinforcement learning-guided graph transformation framework for SpTRSV, in which graph transformation is formulated as a sequential decision-making problem and an RL agent learns matrix-dependent transformation policies. Experimental results on real-world sparse matrices demonstrate level reductions of up to 94% and reductions of up to 80% in the coefficient of variation of level costs, while modifying only 1.50% of the rows in the highest case. On average, the RL- guided graph transformation achieves a 23% reduction in the number of levels and a 29% reduction in the coefficient of variation of level costs while rewriting only 0.82% of the matrix rows. Although the heuristic strategies generally achieve more aggressive level reduction(between 31% and 46%), the RL-based approach achieves the largest average reduction in the coefficient of variation of level costs, demonstrating its ability to balance competing graph transformation objectives. The results further show that the learned policies can be transferred to previously unseen matrices through curriculum learning and fine-tuning, while zero-shot experiments provide insights into the limitations of generalizing graph transformation policies across different sparsity patterns.
[LG-10] Role-Adaptive Policy Optimization for Offline Reinforcement Learning
链接: https://arxiv.org/abs/2609.40149
作者: Seonvin Cho,Soohyun Choi,Songnam Hong
类目: Machine Learning (cs.LG)
*备注: 17 pages, 3 figures
Abstract:Policy regularization in offline reinforcement learning balances policy improvement against reliance on uncertain value estimates. This balance can differ between selecting actions for execution and supplying actions for critic bootstrapping, yet methods such as TD3+BC couple these roles through a shared policy. We propose Role-Adaptive Policy Optimization (RAPO), which adapts policy-update coefficients according to their roles in value learning and execution. RAPO learns these coefficients by differentiating through candidate policy updates formed using the base algorithm’s actor objective. For TD3+BC, RAPO separates bootstrap and execution actors and adapts their coefficients independently: the bootstrap objective penalizes policy-induced changes in target values, while the execution objective evaluates a local policy-improvement surrogate. For IQL, whose value learning is already independent of the execution actor, RAPO preserves the original value updates and adapts only the inverse temperature in advantage-weighted policy extraction. Experiments on D4RL locomotion and AntMaze tasks show improvements over both base algorithms, with larger gains for TD3+BC, whose RAPO instantiation outperforms baselines on average.
[LG-11] From Spectra to Joint Schedules in LLM Pre-training: 33(2) Scaling-Law Regimes
链接: https://arxiv.org/abs/2609.40148
作者: Yichen Wang,Fanghui Liu,Yudong Chen
类目: Machine Learning (cs.LG)
*备注:
Abstract:Power-law learning curves are often treated as fixed properties of a model and its data, although learning-rate and batch-size schedules can change the observed loss. We study this dependence in noisy online SGD with linear random features. Conditional on the representation, an exact Volterra equation separates two response components: a forcing term that propagates unresolved target error and a memory kernel that propagates stochastic-error injections. We prove that either component follows a power law if and only if its cumulative weighted spectral mass has the corresponding low-spectrum scaling; individual eigenvalues and target coefficients need not obey coordinatewise power laws. Under a joint schedule, intrinsic time T_t=\sum_st\eta_s controls optimization progress, while r_t=B_t/\eta_t controls noise injection. Their interaction yields sharp conditions under which a schedule preserves, changes, or destroys the clean power law, together with a memory ceiling on noise reduction. The power-law random-feature model realizes this mechanism in 3+3(+2) propagation regimes with phase-dependent compute rates. Controlled nanoGPT experiments show that (1) learning-rate and batch-size schedules with matched B/\eta paths are nearly equivalent in intrinsic time, (2) a forcing-memory surrogate accurately predicts loss across schedules, and (3) its fitted exponents across real-world datasets identify the regime of LLMs in 3+3(+2) map.
[LG-12] Policy Iteration Is Not Strongly Polynomial for Deterministic Markov Decision Processes: The Price of Algorithmic Anarchy
链接: https://arxiv.org/abs/2609.40147
作者: Han Zhong,Yinyu Ye
类目: Machine Learning (cs.LG); Data Structures and Algorithms (cs.DS); Optimization and Control (math.OC)
*备注:
Abstract:We establish an exponential iteration lower bound in the number of states for Howard’s policy iteration on deterministic discounted Markov decision processes, with at most two actions per state. This rules out strong polynomiality of Howard’s policy iteration when the discount factor is part of the input and yields an exponential separation from the simplex method with Dantzig’s pivoting rule, which is proved to be strongly polynomial on this class. Even when each reward is restricted to logarithmic bit length, we obtain a stretched-exponential iteration lower bound. The gap between Howard’s decentralized and simultaneous selfish improvements and Dantzig’s coordinated selection of a single action with the largest gain across all states reveals a ``price’’ of algorithmic anarchy.
[LG-13] From DNA Design to DNA Slimming: Auditable Agent ic Discovery of a Deletion-Only Designer NEURIPS2026
链接: https://arxiv.org/abs/2609.40143
作者: Joel Shor
类目: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE); Genomics (q-bio.GN)
*备注: 40th Conference on Neural Information Processing Systems (NeurIPS 2026). Workshop: Agentic AI for Biological Discovery
Abstract:Compact regulatory DNA can free up space in vector payloads, reduce synthesis and assay burden, and expose which sequence features drive predicted activity. Yet most model-based nucleic-acid designers optimize fixed-length sequences through substitutions; they do not ask which bases of an existing functional element can be removed while retaining predicted activity. We define the task of sequence slimming as selecting an exact-length, order-preserving subsequence while retaining activity. Modeled on the design benchmark NucleoBench, we propose a quantitative evaluation for slimming that balances sequence reduction with maintaining function. Each slimmer must return both the subsequence and its source indices, which can be used to verify that the slimmer obeyed task requirements. To our knowledge, this is the first dedicated benchmark of this deletion-only problem. The coding agent Empirical Research Assistant (ERA) then searched over executable designer programs. ERA received the task prompt and a successful substitution-only designer GrAdaBeam as a starting program, and it modified the designer to produce GRADASLIM. We report held-out evaluations for five transcription-factor binding targets, comparing random, greedy, and ERA-guided slimming at 400 and 100 bp. ERA has the highest mean in 9/10 settings. Paired bootstrap intervals for ERA minus greedy are above zero in all five 400-bp settings, below zero in one 100-bp setting, and overlap zero in the remaining four.
[LG-14] Scalable Cox Regression via Grouped Risk Sets and Sharper LogSumExp Rates
链接: https://arxiv.org/abs/2609.40120
作者: Elizaveta Iashchinskaia,Egor Gladin
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: 32 pages, 5 figures, 8 tables
Abstract:Motivated by the computational challenges of large-scale Cox regression, we study stochastic minimization of LogSumExp objectives over large sets. Mini-batch normalizer estimates generally yield biased gradients. We instead use a softplus surrogate that introduces one auxiliary scalar per normalizer and admits unbiased single-sample gradients. For smooth convex LogSumExp objectives, we prove an O(T^-1/2) averaged objective bound, improving the previous T^-1/4 analysis. With a strongly convex regularizer on the original variable, we also obtain a last-iterate squared-error rate of \widetildeO(T^-1) without strong convexity in the auxiliary variables. For Cox regression, the normalizers are defined over nested risk sets. We exploit this structure by grouping neighboring failures and sharing one auxiliary variable per group. The resulting compressed objective admits uniform score and curvature bounds that control the errors from grouping and softplus approximation. Together with the general optimization result, these bounds give a mean-square rate of T^-4/5 , up to logarithmic factors, relative to the full Cox solution. The compressed estimator also matches the full estimator’s asymptotic distribution. Experiments on synthetic and real survival datasets with slowly decreasing risk sets show a favorable performance relative to stochastic baselines.
[LG-15] Beyond Model Ranking: Regime Diagnosis for Distributional-Statistical Misspecification in Industrial Time-Series Forecasting
链接: https://arxiv.org/abs/2609.40117
作者: Pengyu Nie,Chenglang Xu,Yaoshi Chen,Chaogan Ren,Wei Hu,Chao Yang,Jiangong Zhang
类目: Machine Learning (cs.LG)
*备注: 31 pages, 12 figures
Abstract:Time-series forecasting models achieve strong benchmark performance but exhibit severe systematic bias in industrial deployments. This train–deploy gap is conventionally attributed to temporal-structural errors or distribution shifts. We characterize a complementary source that these explanations overlook: canonical losses embed fixed statistical priors, while industrial demand mixes benign and pathological regimes—zero-inflation, skewness, high variability—in which these priors are systematically violated. The induced bias persists even under perfect temporal modeling, remains in a distributional-shape component that normalization cannot remove, and creates an aggregation trade-off invisible to aggregate metrics. We turn these observations into an evaluation toolkit centered on the Regime-wise Relative Bias Vector (RBV): a metric-agnostic, regime-decomposed diagnostic that audits how pooled training allocates systematic mismatch across pathological subpopulations. A controlled attribution analysis decomposes RBV into a model-independent intrinsic floor, set by each loss’s estimand, and an excess component attributable to training, tracing observed bias to the loss rather than the model. A large-scale study—13 loss objectives, 3 seeds, 60,000+ series spanning RetailShiftBench and M5, with random-split controls—shows that regime-aware diagnosis separates optimization-type from bias-type failure, and that regime-aware training resolves the pooling-induced bias that capacity scaling cannot, for mean-type losses. A formal structural observation, that risk under evaluation-distribution contamination is affine in the pathology mixture weight, grounds these findings. Our work complements model ranking with mechanism-grounded, regime-oriented evaluation.
[LG-16] Efficient Expert-Parallel Communication on PCIe-Connected Consumer GPUs
链接: https://arxiv.org/abs/2609.40093
作者: Jaehwan Lee,Sangmin Lee,Chaewon Kim,Junsik Shin,Jaejin Lee
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注:
Abstract:Expert parallelism (EP) enables inference of large Mixture-of-Experts (MoE) models by placing their experts across multiple GPUs, but requires substantial communication between GPUs at every MoE layer. As contemporary MoE models activate more experts per token, this communication accounts for a growing fraction of inference time. The cost becomes particularly pronounced on PCIe-based consumer GPU systems, where all inter-GPU transfers traverse CPU memory. However, existing MoE-specialized EP communication libraries assume that direct GPU-to-GPU access is available, largely overlooking consumer GPUs. Therefore, most LLM frameworks instead rely on NCCL, whose CPU-staged communication incurs redundant PCIe transfers and competes with expert computation for GPU resources, limiting their overlap. We present ThunderEP, a novel communication design for such systems that removes the relay hops of traditional ring algorithm, moves data through DMA engines to avoid compute resource contention, and minimizes synchronization latency by reducing the polling overhead of completion flags in CPU memory. We integrate the proposed design into vLLM and evaluate it on three widely used MoE models. Experiments on two PCIe systems equipped with RTX 4090 and RTX 5090 GPUs show that ThunderEP achieves average speedups of 2.00 \times and 1.53 \times over NCCL for dispatch and combine, respectively, and up to 1.66 \times end-to-end speedup over state-of-the-art MoE inference frameworks.
[LG-17] Replay on Demand: An Emergent Curriculum for Balancing Adaptation and Forgetting in Continued Pretraining
链接: https://arxiv.org/abs/2609.40089
作者: Lukas Thede,Shengzhuang Chen,Stefan Winzeck,Matthias Bethge,Zeynep Akata,Jonathan Richard Schwarz
类目: Machine Learning (cs.LG)
*备注:
Abstract:Continued pretraining enables language models to adapt to new domains and knowledge, but often at the cost of forgetting previously acquired capabilities. Replay can mitigate this trade-off, but fixed replay mixtures allocate training independently of the model’s actual retention needs. We introduce Replay on Demand (RoD), which instead derives the replay allocation from the model’s learning dynamics. RoD jointly prioritizes adaptation samples by their remaining learning potential and replay samples by their observed forgetting. Their competition for a shared training budget yields an online curriculum that determines what to train on at each step. Across models, scales, and adaptation domains, RoD reaches or improves upon the adaptation-forgetting frontier of tuned fixed-replay baselines and model merging without prescribing a replay allocation in advance. Replay concentrates on sources that are more vulnerable to forgetting and dynamically increases and redistributes as forgetting emerges during training. Together, our results show that replay can be allocated online from the model’s evolving state, targeting what is needed, when it is needed.
[LG-18] Robust and Learned Online Matching in Growing Trees
链接: https://arxiv.org/abs/2609.40077
作者: Marek Gałązka,Hanna Wdowicka
类目: Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG); Probability (math.PR)
*备注: 14 pages, 1 figure, 1 table
Abstract:We study irrevocable maximum-cardinality matching in trees revealed by successive leaf attachments, with a known horizon and an exogenous growth law that is misspecified or unknown. For deterministic affine attachment forecasts with nonnegative degree reinforcement, the optimal threshold policy loses at most twice the cumulative expected conditional total-variation error relative to an online oracle knowing the actual growth law. This follows from a unit-span property of the Bellman continuation score and has no additional horizon factor. A four-vertex example attains the coefficient two for the specified deterministic policy, and a two-model argument gives a lower bound linear in the model-error budget for arbitrary policies under general misspecification. For uniform-preferential attachment, the local error has an exact expression through the leaf count. When its constant mixture parameter is unknown, we estimate it from the same growing tree and update the threshold policy at geometric times. A parameter-sensitivity bound for individual Bellman prices and uniform degree-moment estimates yield expected regret O(\sqrtn\log^2 n) , using O(n^2\log n) arithmetic operations and O(n) stored entries. The exact minimax rate remains open.
[LG-19] Accelerated Algorithm for Sparse Regularized Partial Optimal Transport
链接: https://arxiv.org/abs/2609.40075
作者: Khoa Nguyen,Dung T. Nguyen,Thong Huynh,Hoang-Hiep Nguyen-Mau,Anh Nguyen,Minh Ngoc Dinh,Juho Kannala
类目: Machine Learning (cs.LG)
*备注: 36 pages, 13 figures. Submitted to the Journal of Optimization Theory and Applications
Abstract:Partial Optimal Transport (POT) extends the classical optimal transport problem by relaxing the strict mass conservation constraint, enabling its use in a wide range of real-world applications. In many of these settings, sparse transport plans are preferred for their interpretability and computational benefits. While smooth and strongly convex regularizers - such as quadratic or elastic net - have been vastly used in various machine learning applications to induce sparsity and accelerate computation, they have received less algorithmic attention compared to entropic approaches for computational POT. In this paper, we propose a new optimization framework that leverages these regularizers through a penalty-based reformulation, enabling efficient gradient-based updates while preserving the structure of the original problem. Our method accommodates a broad class of regularizers that promote structured and sparse transport plans. Building on this formulation, we design an accelerated first-order algorithm that alternates between smooth updates and simple projection steps. Through empirical benchmarks on color transfer, domain adaptation, and point cloud registration, our approach consistently outperforms established baselines - achieving lower transport cost, higher sparsity, and faster convergence - making it a practical and scalable solution for modern transport problems.
[LG-20] Gromov-Wasserstein Distillation for Inductive Multi-View Embedding NEURIPS2026
链接: https://arxiv.org/abs/2609.40047
作者: Rafael Pereira Eufrazio,Eduardo Fernandes Montesuma,Charles Casimiro Cavalcante
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: This paper was accepted at the GDDL (Geometric Distributional Deep Learning) Workshop at NeurIPS 2026
Abstract:Gromov-Wasserstein multidimensional scaling (GW-MDS) learns low-dimensional representations from relational data but remains transductive, providing no explicit mapping for unseen samples. We introduce an inductive framework based on barycentric distillation. A GW-MDS teacher learns a latent support and an optimal transport plan from the training data, and barycentric projection converts the resulting coupling into sample-aligned targets. A neural student then learns an explicit out-of-sample mapping, avoiding additional relational-matrix construction and GW optimization at inference. We formulate the approach for single-view data and extend it to Mean-GWMDS and Multi-GWMDS teachers through consensus and selected-projection targets learned by a multi-view student with view-specific encoders. We also investigate a direct neural baseline trained solely with a GW objective. Experiments on synthetic and real-world data using Euclidean, geodesic, and cosine relations show that the distilled models preserve the teacher geometry on unseen samples and consistently outperform direct neural GW training in sample-indexed relational preservation. These results establish barycentric projection as an effective bridge between transductive GW embeddings and inductive neural mappings.
[LG-21] Component-Weighted Centroid Search for Exact Incremental BPE NEURIPS2026
链接: https://arxiv.org/abs/2609.40016
作者: Harshit Verma,Rex Ying
类目: Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG)
*备注: Accepted at AXIOM 2026, a NeurIPS 2026 Workshop. 10 pages
Abstract:Exact incremental BPE maintains the canonical tokenization state after every appended byte. The recent algorithm of Jiang and Gong (2026) does this in O(\log^2 t) worst-case time, where t is the maximum canonical token length. Its centroid search visits O(\log t) components and can pay another O(\log t) for ordered point location at each one. Within Jiang and Gong’s normalized/proper merge-stage model, we change only that local search. Each interval is weighted by the size of the recursive component it selects, so a move from size m to size m’ costs O(1+\log(m/m’)) . These charges telescope, giving O(\log t) time per append and O(n\log t) over an n -byte stream, with the same BPE semantics and asymptotic space. We also construct a normalized proper BPE family over a fixed alphabet where count-balanced search uses \Theta(\log^2 t) probes on a reachable update, while the weighted search uses \Theta(\log t) . A Rust implementation matches the predicted probe counts on every tested instance. On ordinary vocabularies the queried degrees are small, however, and the improvement is a worst-case guarantee rather than an average-speed result.
[LG-22] DashVMC: Real-Time Discrete World Model Control in Geometry Dash NEURIPS2026
链接: https://arxiv.org/abs/2609.40003
作者: Florent Tariolle,Florian Yger
类目: Machine Learning (cs.LG)
*备注: 10 pages, 2 figures. Accepted at the NeurIPS 2026 workshop “PTA: From Pretrained Representations to Acting Agents”. Project page: this https URL
Abstract:World-model agents are usually evaluated in simulators that can wait for the policy; live games impose the opposite constraint, requiring capture, prediction, and action before the next frame. We present DashVMC, which learns a compact, action-conditioned world model from approximately two hours of recorded Geometry Dash gameplay. To test whether the learned dynamics are actionable, a controller is initialized by behavioural cloning (BC) and refined with Proximal Policy Optimization (PPO) entirely in frozen-model rollouts, without further interaction with the live game. Across three controller seeds, the refined policies survive longer than their BC initializations on all three official levels and a held-out community layout. At deployment, the baseline skips visual generation and sustains a 60-Hz capture-to-action loop on a consumer GPU. Action-conditioned continuations and rollout diagnostics show that the model remains useful for control despite imperfect long-horizon fidelity.
[LG-23] Learning to Explain While Planning : Rule-Aligned Diffusion Planning for Autonomous Driving
链接: https://arxiv.org/abs/2609.39995
作者: Jiaxi Ye
类目: Machine Learning (cs.LG)
*备注:
Abstract:Diffusion planners exhibit strong capabilities in generating multimodal trajectories. However, existing methods primarily rely on expert demonstrations to fit trajectory distributions, learning statistical correlations among scenes, behaviors, and trajectories without explicitly modeling driving rules. In long-tail scenarios where expert data are scarce, the lack of behaviors to imitate may lead to trajectories that violate safety or compliance requirements. Moreover, their generation process lacks rule-level explanations, making it difficult to determine which rules drive trajectory adjustments, when they take effect, and how strongly they act, thereby limiting failure diagnosis, safety validation, and targeted improvement. To address these limitations, we propose the Rule-Aligned Diffusion Planner (RADP), which incorporates differentiable driving rules into the diffusion objective during training, turning rule knowledge into intrinsic behavioral principles beyond finite demonstrations. We further introduce Rule-Pressure Attribution (RPA), which constructs supervision signals from gradients of rule losses with respect to predicted trajectories and employs a lightweight attribution head to estimate the optimization pressure exerted by each rule online. To assess the closed-loop behavioral relevance of these attributions, we propose a temporal risk-alignment protocol that evaluates whether current rule pressures reflect corresponding risks during subsequent closed-loop execution. Experiments on nuPlan show that RADP improves closed-loop planning in challenging safety-critical scenarios, while RPA exhibits consistent temporal alignment with subsequent rule-specific risks, validating both intrinsic rule learning and rule-level interpretability.
[LG-24] When Instructions Retrieve Trajectories: Diagnosing and Mitigating Generalization Failures in VLA Models
链接: https://arxiv.org/abs/2609.39971
作者: Hung-Jen Chen,Yu-Hsun Hou,Yan-Hong Chen,Yan-Fu Chen,Binghua Cai,Min Sun,Chun-Yi Lee
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注:
Abstract:Vision-language-action (VLA) models can exceed 90% success on in-distribution tasks and withstand nuisance changes that preserve the required action, yet fail under counterfactual changes that demand a different action. Aggregate robustness scores can therefore conceal a more specific failure, in which a policy responds to both language and vision yet does not combine them to select the action the task requires. We call this failure instruction-action binding. Instructions cue familiar trajectory families, and visual feedback adjusts their execution. Behavioral analyses of fine-tuned \pi_0.5 and GR00T-N1.7 policies reveal that failed rollouts often retain the source behavior or switch to another demonstrated task. These switches show that language is not simply ignored. Readouts and interventions connect these choices to task-conditioned internal states. Our analysis of the imitation objective shows how narrow conditional action support can leave grounded and instruction-keyed solutions indistinguishable on the demonstrations. This motivates Equivariant Counterfactual Training (ECT), which acts at two levels. ECT data supply valid demonstrations in which the same instruction requires different actions in distinguishable scenes, while the ECT loss trains each demonstration with its counterpart in the same update. In a controlled LIBERO-PRO comparison, full ECT raises \pi_0.5 's mean position-swap success from 36% to 59%. On CALVIN, where counterparts already occur in the original data, the ECT loss improves five-task completion without new demonstrations. On a real UR5e under a fixed demonstration budget, full ECT raises unseen-position success from 8% to 88%.
[LG-25] Patient-Centered Treatment Planning for Chronic Multimorbidity: A Hierarchical Reinforcement Learning Framework for Preference Modeling
链接: https://arxiv.org/abs/2609.39911
作者: Nafiseh Payani,Soham Das,G. Anthony Wilson,Anahita Khojandi
类目: Machine Learning (cs.LG)
*备注: 39 pages, including appendices
Abstract:Patient preference, defined as a patient’s demonstrated willingness and capacity to adhere to clinical recommendations, is a primary determinant of therapeutic effect yet remains structurally absent from existing computational treatment planning models. We address this gap by presenting patient-centered factored-action hierarchical option-critic (FAHOC), a hierarchical reinforcement learning (HRL) framework that jointly learns high-level options corresponding to therapeutic strategies and factored intra-option policies that decompose the joint action space into disease- and intervention-specific subcomponents, while imposing a cooperation-aware action masking mechanism. This enables structured exploration, improved credit assignment across hierarchy levels, and more interpretable decision pathways, while enforcing patients’ preferences. Formal guarantees establish that cooperative patients achieve higher optimal expected health outcomes than non-cooperative patients, and that the factored Q-function approximation error is provably bounded. The framework is evaluated using longitudinal data collected from approximately 50,000 comorbid hypertension and type 2 diabetes mellitus patients from five hospitals in the Southeast U.S. FAHOC achieves a quality-adjusted life year expectancy equivalent improvement of 0.669 (vs -0.133 observed clinician practice), correctly identifies cooperative patients in 95.9% of cases and never violates a patient’s preference in held-out test, demonstrating that HRL with explicit preference constraints can support preference-consistent, clinically safe decision-making in multimorbidity management.
[LG-26] Shared Weights Selected Computations: How Looped Transformers Route What Each Loop Does
链接: https://arxiv.org/abs/2609.39892
作者: Jiaju Wu,Yi Hu,Muhan Zhang
类目: Machine Learning (cs.LG)
*备注: 36 pages. Code and reproduction materials: this https URL
Abstract:Looped Transformers repeatedly apply the same set of Transformer layers, giving them a recurrent architecture for latent computation. Their strong performance on iterative reasoning and length-generalization tasks suggests an appealing explanation: recurrence may provide an inductive bias that lets the model reuse a learned algorithm across loops. However, weight sharing alone does not imply that every loop performs the same operation. This raises a basic question: is each loop actually repeating the same computation, and if not, what routes the shared parameters to different operations? We study this question using graph walks as a test case. In the model’s native trajectories, decoded predictions can advance by different numbers of graph steps or remain at a reached target, showing that recurrent progress need not follow a fixed one-loop-one-step pattern. We then show that a frozen loop can be steered toward different transitions by modifying its entering hidden state: a learned linear layer J selects the desired transition without changing the shared Transformer layers. To test how this steering works, we use activation patching and find that attention patterns can recover its effects and switch the selected transition. Across five matched pairs of graph models, changing intermediate supervision during backbone training changes which transitions J can induce. This suggests that J selects computations learned by the backbone rather than creating new algorithms. Together, these results show that the hidden state can control shared computation, with attention routing as a causal pathway. Comments: 36 pages. Code and reproduction materials: this https URL Subjects: Machine Learning (cs.LG) Cite as: arXiv:2609.39892 [cs.LG] (or arXiv:2609.39892v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.39892 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-27] Markovian Dynamics Enforcer: Feasibility Preserving Correction on Learned Dynamics Manifolds NEURIPS2026
链接: https://arxiv.org/abs/2609.39888
作者: Kevin Yu,Tao Guo,Constantinos Antoniou,Panagiotis Angeloudis
类目: Machine Learning (cs.LG); Robotics (cs.RO); Systems and Control (eess.SY)
*备注: 26 pages, 2 figures, 11 tables. Accepted at NeurIPS 2026. Code available at this https URL
Abstract:Neural trajectory predictors can reach low prediction error while violating dynamics, actuator limits, or state constraints, especially when controls are unobserved and dynamics are partially specified. We introduce the Markovian Dynamics Enforcer (MaDE), a time-invariant post-hoc operator mapping state-transition proposals onto a learned feasible dynamics manifold, trained on feasible states without ground-truth controls. For each transition it infers a control and recomputes the state through a completion model of known physics plus a learned residual. It then corrects that control by gradient-based inequality reduction, so inequality satisfaction is best-effort within an iteration budget. Since every correction iterate re-enters the completion model, the returned state is dynamically consistent by construction relative to that model and the supplied previous-state anchor. MaDE drives dynamics residuals to essentially zero on fully specified simulated systems, and on an underspecified system leaves a smaller true-dynamics residual than the baselines. Designed to attach to arbitrary predictors, the frozen operator is evaluated downstream of recurrent, structured state-space, and transformer predictors. On recorded vehicle trajectories the one-step residual against a kinematic bicycle model is 0.0071 to 0.0072 for MaDE and 0.1703 to 0.1714 for raw predictors. MaDE raises average displacement error by a factor of 1.57 to 1.83.
[LG-28] PassGPT : Leverag ing Linguistic Priors for Password Modeling
链接: https://arxiv.org/abs/2609.39880
作者: Rajneesh Anand,Neeraj Lakshmanan,Masoud Yari
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 3 figures, 2 tables. Code is available at this https URL
Abstract:Passwords remain the dominant online authentication mechanism, and understanding how humans choose them is essential for defensive strength estimation and attack simulation alike. Recent learning-based approaches such as PassGAN and PassGPT have shown that deep generative models can learn password structure directly from leaked corpora. However, both train from random initialization on password data alone. The role of linguistic prior knowledge in password modeling, and what it reveals about how humans create secrets, remains largely underexplored. Here, we address this gap with PassGPT+, which adapts the linguistic prior of GPT-2 to password observations through character-aware tokenization. We also introduce PassDiffusion, the first absorbing-state discrete diffusion model for password generation, as a probe of whether non-autoregressive approaches are competitive. On the RockYou benchmark, PassGPT+ recovers 22.53% of held-out passwords at 108 guesses, a 16% relative gain over PassGPT, and retains 79% of this match rate when transferred without retraining to a disjoint 2020 leak dataset, demonstrating that linguistic priors capture persistent regularities of human password generation. PassDiffusion underperforms by two to three orders of magnitude, indicating that autoregressive modeling is substantially better matched than iterative denoising to the discrete, exact-match nature of password generation.
[LG-29] Preemptive LLM Unlearning against Forbidden Capability Acquisition via Gradient Sealing
链接: https://arxiv.org/abs/2609.39866
作者: Kemou Li,Qizhou Wang,Yue Wang,Fengpeng Li,Zhuan Shi,Negar Rostamzadeh,Golnoosh Farnadi,Masashi Sugiyama,Jiantao Zhou
类目: Machine Learning (cs.LG)
*备注:
Abstract:Open-weight LLMs are released not only as fixed products but also as substrates for downstream fine-tuning. This openness, however, creates legal and ethical risks because users may misuse fine-tuning to instill illicit knowledge or enable hostile operations. Model providers therefore need apre-release defense against such acquisition, motivating the problem of preemptive unlearning. Unlike retrospective unlearning, which removes capabilities already present in a fixed model, preemptive unlearning seeks to prevent their acquisition under unseen attack data and future fine-tuning procedures. Despite its practical importance, this setting remains largely unexplored, presents distinct challenges, and is therefore the central focus of our work. We first verify that existing retrospective methods provide insufficient pre-release protection. Even when forbidden capabilities are suppressed in current outputs, forbidden-domain data can still induce gradients through internal pathways, enabling later acquisition. Motivated by this finding, we propose a gradient-sealing principle that blocks these pathways by pushing relevant pre-activations into the negative region, where ReLU-family activations exhibit zero or near-zero derivatives. Experiments across multiple LLM families demonstrate our stronger resistance to downstream acquisition than retrospective baselines, validating gradient sealing as an effective mechanism for pre-release protection.
[LG-30] Fork-dLLM : Avoiding the Flexibility Trap in Diffusion Language Models
链接: https://arxiv.org/abs/2609.39859
作者: Stipe Frković,Metod Jazbec,Christian A. Naesseth
类目: Machine Learning (cs.LG)
*备注:
Abstract:Masked diffusion language models (dLLMs) have shown strong potential for faster inference through parallel token generation when combined with confidence-based samplers. However, recent work has shown that such methods can defer unmasking high-entropy fork positions at which multiple plausible continuations exist. This results in reduced generation diversity, as shown by worse pass@k scaling, and limits gains obtainable from RL post-training. To avoid this flexibility trap, prior work advocated for autoregressive (AR) sampling. Here, we show that discarding confidence-based sampling is unnecessary and, once inference cost is taken into account, wasteful. We first propose Fork-dLLM, a simple hybrid sampler that uses AR-style ordering only at uncertain fallback steps while retaining parallel generation otherwise. We then extend the same principle to post-training with ForkGRPO, which uses Fork-dLLM rollouts and applies the GRPO objective only at fallback steps, preserving exact policy-likelihood ratios while substantially reducing rollout and optimization cost. In our experiments, Fork-dLLM matches the strong pass@k scaling of AR sampling while being 2-3x more efficient, and ForkGRPO achieves downstream performance comparable to or better than AR-based GRPO baselines at a substantially lower training cost.
[LG-31] Dimension-Free Rank Lifting from Random Hyperplane Arrangements
链接: https://arxiv.org/abs/2609.39855
作者: Luca Becchetti,Matteo Russo,Ruben Skorupinski
类目: Machine Learning (cs.LG); Probability (math.PR)
*备注:
Abstract:We study the width required for a randomly initialized hidden layer of a neural network to achieve rank lifting. Namely, given a dataset X \in \mathbbR^m \times d of m , d -dimensional input vectors separated by an angle of at least \theta , we consider the random feature matrix \sigma(XR) , where R is standard Gaussian. For positively homogeneous nonpolynomial activations, which include sign, Heaviside, ReLU, and ReLU powers among others, we prove that n \gtrsim \frac1\theta\max\left\m,\log\left(\frac1\delta\right)\right\ neurons suffice for \sigma(XR) to have full row rank m with probability at least 1-\delta . This dimension-free bound exponentially improves the previous general-dimensional guarantee for sign features (Drago et al., 2026) and is essentially tight. The proof shows that one random feature column escapes every proper subspace of \mathbbR^m with probability \Omega(\theta) , using a coupling of nearby Gaussian directions and a local crossing of the induced hyperplane arrangement. We also study stable rank lifting, where the goal is to establish a quantitative analogue of exact rank lifting, i.e., a lower bound on the smallest eigenvalue of the empirical feature Gram matrix in high-probability. Our analysis unifies and generalizes stable rank guarantees for all q -homogeneous non-polynomial activations following prior work in Panigrahi et al. (2020) and Song (2026). In particular, we combine a diagonally dominant Taylor tail of the population kernel with truncation and matrix concentration, to show that for positively homogeneous nonpolynomial activations, stable rank lifting is achieved at width n \gtrsim C^q \fracm\theta^2q+1 \log^2q+\frac12\left(\fracm\theta\right) \log\left(\fracm\delta\right), where q is the degree of the activation and C 0 is some universal constant. Subjects: Machine Learning (cs.LG); Probability (math.PR) Cite as: arXiv:2609.39855 [cs.LG] (or arXiv:2609.39855v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.39855 Focus to learn more arXiv-issued DOI via DataCite
[LG-32] Predicting Multi-View Rashomon Representation: Can We Learn Where Models Disagree?
链接: https://arxiv.org/abs/2609.39848
作者: Mingyue Ma,Zongbo Han,Changqing Zhang,Guangyu Wang
类目: Machine Learning (cs.LG)
*备注: Under review
Abstract:Foundation models are increasingly adopted across a wide range of applications, often serving as core blocks within AI systems. Yet different foundation models may encode the same input from multiple different views, leading to substantial representation disagreement, which we term Rashomon Representation. Such disagreement often signals inputs that a given model encodes in a way inconsistent with other models, offering a valuable yet underexplored signal for input reliability estimation. While prior work has largely focused on measuring disagreement across multiple models with a representation set, we instead focus on predicting disagreement from a single representation. We hypothesize that this disagreement follows some consistent, input-dependent patterns rather than occurring at random. To test this, we quantify disagreement by comparing each sample’s nearest neighbors across different models’ representation spaces, then train a lightweight predictor that estimates disagreement from a single model’s representation. At inference time, given a new input, the predictor uses that input’s representation to tell whether it aligns with or diverges from those of other models. Extensive experiments across diverse foundation models and datasets show that representational disagreement is indeed input-dependent, predictable, and generalizable, enabling efficient reliability estimation of foundation models.
[LG-33] SEAR: Spoofing Evidence-Grounded Audio Reasoning Benchmark for Audio Language Models
链接: https://arxiv.org/abs/2609.39847
作者: Rong Wan,Suliu Qin,Jiaxi Li,Wei Xie,Wenwu Wang,Xiaolong Han,Lu Yin,Xilu Wang
类目: ound (cs.SD); Machine Learning (cs.LG)
*备注:
Abstract:Audio language models (ALMs) are increasingly used for audio deepfake detection (ADD), yet existing benchmarks assess their verdicts or rationale plausibility without verifying the underlying acoustic evidence. To address this issue, we first introduce spoofing evidence-grounded audio reasoning (SEAR), a four-task AQA benchmark to evaluate ALM-based ADD through acoustic evidence identification and quantification, deepfake detection, and forensic rationale generation. We further propose a bona-fide-based acoustic evidence agent (BAEA), which equips a frozen ALM with controlled acoustic tools under \textscfixed or \textscadaptive evidence-acquisition policies. Experiments with six ALMs reveal a clear gap between plausible rationales and verifiable acoustic evidence reasoning, while BAEA-\textscFixed improves final verdicts and forensic rationales on both evaluation partitions. Controlled interventions further show that misleading evidence degrades both detection and grounding performance.
[LG-34] Dynamic LoRA-Experts and Prototype-Ensemble Matching for Class-Incremental Learning
链接: https://arxiv.org/abs/2609.39839
作者: Hongwei Zhao,Rui Liu,Yansong Liu(School of Computer Science and Engineering, Beihang University)
类目: Machine Learning (cs.LG)
*备注: Published open-access article; 23 pages
Abstract:Class-Incremental Learning (CIL) aims to continuously learn new classes without forgetting previously acquired knowledge. Parameter-efficient fine-tuning with pre-trained models reduces parameter overhead but can suffer from cumulative interference and suboptimal alignment between inference samples and specialized modules. We propose Dynamic LoRA-Experts and Prototype-Ensemble Matching (DLEPEM), a two-stage rehearsal-free framework. DLEPEM allocates a task-specific LoRA-Expert for each incremental task to reduce cross-task interference, then combines frozen pre-trained-model prototypes with task-adaptive LoRA-Expert prototypes for reliable task-level discrimination. Experiments on standard CIL and Few-Shot CIL benchmarks demonstrate strong performance under the evaluated protocols.
[LG-35] Fast Regularized Policy Mirror Descent with One-Step TD Updates
链接: https://arxiv.org/abs/2609.39837
作者: Qipei Chen,Wenye Li,Yule Sun,Ke Wei
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Policy mirror descent (PMD) enjoys fast convergence in regularized Markov decision processes (MDPs), but existing guarantees often rely on exact or increasingly accurate policy evaluation. We analyze PMD coupled with a persistent critic advanced by one temporal-difference (TD) update. For finite discounted MDPs, we establish global linear convergence in value for exact coordinate-wise Bellman updates, with any positive constant actor stepsize and arbitrary finite critic initialization. The proof combines a resolvent-based auxiliary distribution with a decaying Bellman-violation correction and a potential weighted by inverse coordinate weights. We then study stochastic TD-PMD with general strongly convex mirror maps under a single off-policy Markov trajectory. With suitably chosen constant stepsizes and a finite-batch TD update, the method achieves an expected value gap of \epsilon after \widetildeO(1/((1-\gamma)^5 \widetilde\sigma_b \epsilon)) transitions. The stochastic analysis relies on the trajectory-wise Lipschitz continuity of the regularizer, derived from uniform bounds on vertex Bregman divergences, together with a visitation-weighted resolvent estimate for signed critic-error propagation that yields an inverse-linear dependence on behavior coverage \widetilde\sigma_b . In contrast to many prior guarantees for regularized policy optimization, our sample-complexity guarantee holds without trajectory resets, generative-model access, or nested policy-evaluation loops. Numerical results are consistent with the theoretical convergence analysis.
[LG-36] RainAtlas: A Multi-Continental Dataset for Precipitation Downscaling
链接: https://arxiv.org/abs/2609.39833
作者: Pierre-Louis Lemaire,Luca Schmidt,Wietze Suijker,Alex Hernandez-Garcia,David Rolnick
类目: Machine Learning (cs.LG)
*备注:
Abstract:Extreme rainfall events are increasing in intensity and frequency as climate change accelerates. While kilometer-scale precipitation forecasts are critical for supporting local decision-making, the limited availability of high-resolution precipitation observations hinders their accuracy, especially in under-resourced regions. Machine learning models are widely used to downscale precipitation data to km-scale, but their application to unseen geographies presents challenges. First, processing raw high-resolution precipitation datasets across regions requires significant engineering and domain expertise. Second, generalization across regions remains difficult. To help overcome these barriers, we release RainAtlas, a large-scale, ML-ready and multi-continental dataset for precipitation downscaling. Covering three continents, RainAtlas harmonizes heterogeneous hourly km-scale observations to a common 2-km grid. Each regional partition contains around 210,000 aligned low- and high-resolution precipitation pairs, respectively from ERA5 reanalysis and direct observations. We benchmark state-of-the-art ML-based downscaling models across RainAtlas using a wide range of metrics. Our evaluation reveals substantial variance in out-of-domain generalization depending on the training regions. This underscores the need for cross-regional, multi-source km-scale evaluation, establishing RainAtlas as a well-positioned benchmark for precipitation downscaling research.
[LG-37] Should I stay or should I show? Learning to selectively disclose information
链接: https://arxiv.org/abs/2609.39818
作者: Carlotta Giacchetta,Alessando Bogani,Cesare Barbera,Giovanni De Toni,Michele Caprio,Andrea Pugnana,Andrea Passerini
类目: Machine Learning (cs.LG)
*备注:
Abstract:In many high-stakes settings, human decision-makers can acquire support information before making a decision. However, acquiring information is costly, and disclosure may fail to improve human decisions or may even impair them. We tackle this problem by studying selective disclosure, i.e., the problem of learning when to reveal support information to a human decision-maker under a budget constraint. We first show that the optimal policy is a threshold rule on the Value of Information (VoI), i.e., the expected reduction in human decision risk induced by disclosure. Since VoI is unknown in practice, we estimate the regime-specific human risks and bound the possible degradation of the resulting plug-in policy relative to lack of disclosure, as well as its regret relative to the optimal policy. Experiments on benchmark datasets show that selective disclosure outperforms both no disclosure and full disclosure, regardless of whether the support information is beneficial or harmful. Two user studies show that human-AI team performance can improve when disclosure is led by our learned policy and not human-selected, although this advantage varies across tasks. A counterfactual benchmark, which replaces participants’ predictions with a machine-learning prediction when disclosure occurs, suggests that these differences might depend on lower adherence to advice when the information is automatically provided rather than self-requested.
[LG-38] Beyond Accuracy: Prefix-Invariant Realizations of Low-Precision Fast Matrix Multiplication
链接: https://arxiv.org/abs/2609.39816
作者: Shuxiao Xie,Shuyang Xie,Yuan Cao,Dezhi Ran,Wei Yang,Tao Xie
类目: Machine Learning (cs.LG)
*备注: 23 pages, 4 figures
Abstract:Fast matrix multiplication saves multiplications through exact cancellation, but rounding sums that mix token rows can leave contributions from later tokens in earlier language model outputs. This threatens prefix invariance, which multiple-choice likelihood scoring relies on: a scored likelihood must depend only on its allowed prefix. On Qwen2.5-14B-Instruct, two fast FP8 realizations repaired to ordinary-looking accuracy still change the answers chosen by likelihood on 5.83% and 10.00% of 240 OpenBookQA items when only the text after the allowed prefix is replaced with the bf16 model’s own greedy continuation. Both row-local controls, the bf16 model and a deployed FP8 matrix multiplication kernel, change none. Accuracy thus does not certify prefix invariance, and the stability criteria we analyze cannot tell realizations apart: across all 512 sign variants of two-level Strassen they stay constant while teacher-forced perplexities span a 772.4 \times range on the same model. We therefore construct certified realizations of two-level Strassen on bounded integer codes that quantize token rows independently, then mix and cancel exactly before rescaling, using 49 block multiplications instead of 64. Our certificate guarantees bitwise equality to a prescribed row-local classical int8 operator at the same quantization specification, so every certified realization inherits its prefix invariance. Certification thus turns realization choice into a pure cost decision: which certified realization runs can no longer change a single scored likelihood.
[LG-39] Backward-State Policy Is Part of the Learning Algorithm
链接: https://arxiv.org/abs/2609.39813
作者: Shuxiao Xie,Shuyang Xie,Dezhi Ran,Wei Yang,Tao Xie
类目: Machine Learning (cs.LG)
*备注: 27 pages, 6 figures
Abstract:Low-precision training rounds tensors that the backward pass reads again, often for several gradients; each use can read the forward’s rounded value, the original, or a new random rounding. This backward-state policy looks like a memory and precision detail, settled by copy accuracy and final loss. We argue that it is part of the learning algorithm, and that neither check shows whether it is right. Copy accuracy does not decide the outcome: in three pairs of 390M runs with an emulated FP8 backward, training fails when attention’s backward reuses the forward’s rounded output and succeeds with a new rounding from the same distribution. Even the most accurate copy, the original itself, can be wrong by our reference: the gradient of the forward pass as it actually ran, with gradients passed through rounding unchanged. For example, a normalization output stored in low precision feeds two gradients: the gain’s gradient needs the original, but the next layer’s weight gradient needs the rounded value that layer multiplied. Final loss, the other check, does not rule out the error of reading the original for both: it persists in models trained with such a store, while planned loss comparisons stay within a margin fixed in advance. We therefore derive from this reference which value each use must read, or which substitute gives the same gradient on average with the forward held fixed, and check these per-use requirements on single operators, without training. In three tests using PyTorch and Transformer Engine, the requirements predicted beforehand whether reuse changes what the backward computes on average relative to an independent copy, and every prediction held. Backward-state policy is thus part of the learning algorithm: it should be specified and checked use by use, not settled by copy accuracy and final loss.
[LG-40] A Comprehensive Benchmark of Source-Free Universal Domain Adaptation on Time Series Representations
链接: https://arxiv.org/abs/2609.39810
作者: Romain Mussard,Fannia Pacheco,Maxime Berar,Paul Honeine,Gilles Gasso
类目: Machine Learning (cs.LG)
*备注:
Abstract:Source-Free Universal Domain Adaptation (SF-UniDA) extends Universal Domain Adaptation by removing access to source data at adaptation time while still handling label-set mismatches between domains. Despite growing interest in this setting for image data, no benchmark exists for time series, which are more challenging. We present the first SF-UniDA benchmark on time series. In addition, we provide the first study of pretrained foundation models as feature extractors for time series domain adaptation. In this context, we identify a critical and previously underexplored limitation of all existing SF-UniDA methods: the inference threshold for unknown-sample rejection is highly sensitive. We address this by proposing a plug-in auto-thresholding module that can be integrated into any SF-UniDA method. Experiments on three well-known time series datasets confirm the suitability of this module. They also highlight that foundation models do not systematically outperform classical backbones and that SF-UniDA tailored for time series is yet to be developed.
[LG-41] RATIO: Reasoning Analysis and Token-level Inference Optimization for Quantized Reasoning Models
链接: https://arxiv.org/abs/2609.39801
作者: Chengzhu Bao,Xianglong Yan,Tianao Zhang,Jiaqi Chen,Shaoqiu Zhang,Yulun Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Post-training quantization (PTQ) has become a widely adopted technique for reducing the memory footprint and inference cost of large language models (LLMs). However, recent studies reveal that when applied to reasoning models, PTQ not only degrades reasoning performance but also exacerbates overthinking, leading to longer reasoning trajectories. These issues may offset the efficiency gains expected from lower-precision inference. Existing approaches mainly rely on complex optimization procedures. More recent lightweight inference strategies instead use predefined overthinking markers, limiting their adaptability across quantized models. To address these issues, we propose Reasoning Analysis and Token-level Inference Optimization (RATIO), a framework that identifies model-specific overthinking tokens and assigns each a tailored penalty. RATIO first introduces Quantization-aware Reasoning Behavior Analysis (QRBA) to identify overthinking tokens by analyzing discrepancies between full-precision and quantized models. It then adopts Token-Specific Penalty Determination (TSPD), which leverages full-precision guidance to derive token-specific penalties without additional training. Extensive experiments show that RATIO achieves a better accuracy-efficiency trade-off than existing token-level interventions. Specifically, RATIO achieves up to 9.8 points accuracy improvement and reduces chain-of-thought (CoT) length by up to 51.3% compared with quantized baselines. The code will be available at this https URL.
[LG-42] Finite-Horizon Fisher Memory in Two-Sided Power-Bounded Recurrent Systems
链接: https://arxiv.org/abs/2609.39800
作者: Jeonghoon Lee(Attractor Dynamics Inc.)
类目: Machine Learning (cs.LG)
*备注: 34 pages, 7 figures. Reproducibility materials: this https URL (release v1.0)
Abstract:We analyse allocation, admission and post-write retention in finite-horizon linear-Gaussian noisy recurrent memories. At every horizon, the directional Fisher memory M_n satisfies \operatornametrM_n=N : non-normality redistributes information but cannot raise its spherical average, while normal carriers satisfy M_n=I . For bi-power-bounded carriers, we derive uniform 1/n lag bounds, identify the limit of M_n with the inverse of the classical Cesàro asymptotic limit of W^\top , and give finite-horizon error bounds. A time-varying coupling defines an end-to-end store operator. The writer-optimal direction need not be store-optimal. After writing ends, an invertible hold preserves the full stored Fisher matrix. Additive contamination bounded by \alpha times the closure covariance retains at least 1/(1+\alpha) of that matrix; a covariance-aware decoder attains the corresponding accuracy. With recurrent carriers held fixed, training input masks and linear readouts approached the task-specific optimum in 160 runs, with median normalized Rayleigh efficiency above 0.998 . Binary accuracy matched the Gaussian prediction to mean absolute error below 0.002 over more than four orders of magnitude in J . In a separate pre-specified study of 320 runs, trained masks followed the designated input-time objective in both carrier types, in 16 of 16 draws. These studies used development-seen carriers and are pre-specified validations, not blind holdouts. The same fixed design reproduced the objective-specific result in 16 of 16 draws on carriers unused before run commitment. Exact isolation preserved information, while a decoder fixed at its training horizon fell to chance; inverse-adjoint transport restored its sampled decisions to numerical precision.
[LG-43] opTimeNet: Topologically-assisted time-series classification model
链接: https://arxiv.org/abs/2609.39792
作者: Sharareh Sayyad,Sophia Bazzi
类目: Machine Learning (cs.LG); Computational Geometry (cs.CG); Dynamical Systems (math.DS); Chaotic Dynamics (nlin.CD); Applied Physics (physics.app-ph)
*备注: 23 pages, 6+4 figures
Abstract:Distinguishing periodic from chaotic dynamics in a time series is a fundamental challenge in both physics and engineering. Yet, end-to-end learned architectures must discover both a representation and a decision boundary from data, at substantial cost. We introduce TopTimeNet, which decouples these tasks: a fixed, non-learned stage extracts a 42 -dimensional geometric and topological descriptor from Takens delay embeddings and persistent homology, and a lightweight learnable stage performs classification. On a benchmark of 49 nonlinear dynamical systems, a 1,638 -parameter configuration matches the mean accuracy of one with 33\times more trainable parameters. Additionally, this approach delivers mean accuracy comparable to convolutional neural networks and surpasses the average performance of converged Transformer models, while requiring three to four orders of magnitude fewer trainable parameters. Robustness also depends sharply on where noise is introduced: TopTimeNet degrades gracefully under perturbations to its precomputed features, but degrades sharply when noise is introduced into the raw signal and the full feature-extraction pipeline is recomputed, showing that robustness to perturbations of the precomputed features does not imply robustness of the complete raw-signal-to-prediction pipeline. These results show that decoupling fixed geometric and topological feature construction from a lightweight discriminative stage can achieve comparable classification accuracy with substantially fewer trainable parameters.
[LG-44] CORD: Learning Reusable Degradation Representations Across Heterogeneous Physical Systems
链接: https://arxiv.org/abs/2609.39784
作者: Haibo Li,Zhiguo Zeng
类目: Machine Learning (cs.LG)
*备注: Preprint
Abstract:Can heterogeneous physical degradation systems benefit from joint pretraining and move beyond system-specific prognostics toward reusable cross-system representation learning? CORD combines type-specific observation interfaces with a shared degradation backbone. Its two self-supervised objectives learn at complementary scales: Intra-Observation Structure Modeling (ISM) captures structure within observations, while Inter-Observation Dynamics Modeling (IDM) captures latent degradation evolution across observation histories. We evaluate CORD under two transfer boundaries: Pretraining-Included System Types, where downstream datasets and held-out units are unseen but their system types are represented during source pretraining, and Pretraining-Excluded System Types, where the entire turbofan-engine type is absent from pretraining. Across bearings, batteries, and cutting tools, CORD (Multi-domain) consistently improves over CORD (Single-domain) under Frozen adaptation, provides further gains under Full FT in most settings, and remains competitive with representative external baselines. Source-pretrained initialization also improves low-label adaptation to the pretraining-excluded engine type. Frozen-representation analysis further shows improved cross-unit lifecycle consistency after multi-domain pretraining. Joint pretraining across heterogeneous physical systems thus produces degradation representations reusable across devices, datasets, and system types.
[LG-45] GraphMAS: A Systematic Benchmark of Multi-Agent Coordination for Graph Learning
链接: https://arxiv.org/abs/2609.39777
作者: Jiayi Yang,Yifang Chen,Yuanfu Sun,Xinyan Ge,Qiaoyu Tan
类目: Machine Learning (cs.LG)
*备注:
Abstract:LLM-based multi-agent systems coordinate specialized reasoning through aggregation, interaction, and adaptive control, yet their potential for graph learning remains unexplored. Graph learning is a natural setting for such systems because useful evidence may arise from heterogeneous local, long-range, global structural, and semantic perspectives whose relevance varies across instances. Existing LLM-based graph learning approaches primarily rely on single-agent reasoning, while multi-agent coordination has been studied mainly in general reasoning settings. Consequently, it remains unclear whether multiple specialized agents can improve graph learning and how coordination strategies should be designed and evaluated. To address this gap, we introduce GraphMAS, a systematic benchmark of multi-agent coordination for graph learning. GraphMAS builds a shared pool of graph reasoning specialists and organizes coordination along two dimensions, inter-agent interaction and runtime adaptivity, yielding four paradigms and seven representative coordination methods. Under a unified protocol, we evaluate these methods across seven text-attributed graphs, three domains, and two graph learning tasks. We find that heterogeneous graph perspectives are complementary, and that coordinating specialists improves over individual specialists and single-agent graph reasoning, with gains from decomposing reasoning across specialists rather than from broader evidence access alone. However, richer inter-agent interaction does not reliably help, whereas instance-adaptive specialist selection yields the strongest accuracy-efficiency trade-off. We further show that coordination can be learned over a fixed specialist pool and transfers to held-out graphs. GraphMAS therefore provides a controlled evaluation framework and empirical principles for understanding when and how multi-agent coordination benefits graph learning.
[LG-46] Riemannian Flow Models with Reinforcement Learning for Molecular Crystal Structure Prediction
链接: https://arxiv.org/abs/2609.39773
作者: Thomas Egg,Harry Winston Sullivan,Maya M. Martirossyan,Philipp Höllmer,Cheng Zeng,Adrian Roitberg,Mingjie Liu,Richard Hennig,Sapna Sarupria,Ellad B. Tadmor,Stefano Martiniani
类目: Machine Learning (cs.LG); Computational Physics (physics.comp-ph)
*备注:
Abstract:Crystal structure governs material properties, making crystal structure prediction (CSP) a fundamental problem in materials science. Generative models are a promising approach for solving this problem, but the prevalence of polymorphism, coupled with large unit cells and complex packing geometry, makes the molecular CSP task challenging for existing models. To address this, we introduce Coarse-Grained Open Materials Generation (CG-OMatG), an equivariant Riemannian flow-based generative model. CG-OMatG predicts molecular crystal structures \textitvia a coarse-grained, hierarchical representation. CG-OMatG treats molecules as rigid bodies—performing both inter- and intra-molecular message passing to construct a geometric representation for molecular packings—and learns to reconstruct molecule centroid positions, orientations, and lattice parameters, conditioned on chemical species and conformer geometry. We train the model on subsets of the Open Molecular Crystals (OMC25) and Cambridge Structural Database (CSD) datasets. Further, we fine-tune the model \textitvia policy gradient reinforcement learning to steer the model towards generating low-energy candidate structures. We validate the generated structures on the CSP blind test benchmark, assessing agreement with experimentally determined crystals using COMPACK packing-similarity analysis. CG-OMatG exhibits strong performance for generative molecular crystal structure prediction, paving the way for accelerated polymorph screening and organic solid-state materials discovery.
[LG-47] Security Properties of Neural Networks as Decision Problems
链接: https://arxiv.org/abs/2609.39768
作者: Adrian Wurm
类目: Logic in Computer Science (cs.LO); Computational Complexity (cs.CC); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 26 pages, 1 table
Abstract:Certifying a deployed neural network raises decision problems that the verification literature has not classified: whether the model carries a backdoor planted in its training data, whether a fault in its stored parameters can drive it into an unsafe state, whether its output leaks a private part of its input. We formalise eight such problems and classify what we can. The organising observation is a logical one. The function computed by a piecewise linear network, together with all its node values, is definable by a quantifier-free formula of real addition of size linear in the network, so a property of the network is a quantifier-alternation sentence, which Sontag’s 1985 theorem places in the polynomial hierarchy at the level of its prefix. Membership results are thus corollaries, and the argument makes plain what they need: that the quantified objects are inputs rather than the network’s own parameters. Non-interference, monotonicity and counterfactual fairness have exactly the complexity of network equivalence and of interval verification, all co-NP- complete over ReLU. Detection of backdoor triggers from a quantised alphabet is Sigma_2^P-complete, one level above robustness certification, so it does not reduce to polynomially many robustness queries unless the hierarchy collapses. Inversion resistance is co-NP-complete for every l_p metric, p a fixed positive integer. Quantifying over parameters instead of inputs - the fault model of bit-flip attacks, radiation upsets and analog accelerators - makes verification exists-R-complete already for networks of identity nodes, for which every previously studied problem is in P, and it stays so when each parameter is confined to a box of inverse-polynomial width; the corresponding safety question is forall-R-complete for ReLU. Comments: 26 pages, 1 table Subjects: Logic in Computer Science (cs.LO); Computational Complexity (cs.CC); Cryptography and Security (cs.CR); Machine Learning (cs.LG) MSC classes: 03B70, 68Q17, 68Q25, 68T07 ACMclasses: F.1.3; F.4.1; I.2.6 Cite as: arXiv:2609.39768 [cs.LO] (or arXiv:2609.39768v1 [cs.LO] for this version) https://doi.org/10.48550/arXiv.2609.39768 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-48] Validity-Preserving Hierarchical RL for Joint Routing and Switch Placement in EDA
链接: https://arxiv.org/abs/2609.39749
作者: Dorian Gailhard,Ugo Lecerf,Enzo Tartaglione,Donatello Conte,Jhony H. Giraldo
类目: Machine Learning (cs.LG)
*备注:
Abstract:Routing and switch placement are fundamental combinatorial optimization problems in chip design, requiring the joint optimization of routing topology and physical placement under strict structural, geometric and logical constraints. Existing approaches typically rely on carefully engineered heuristics that incorporate strong problem-specific biases to navigate the enormous space of possible designs. In this work, we introduce a hierarchical reinforcement learning framework for joint routing and switch placement at the level of logical communication routes. Starting from a minimal routing graph, our method progressively constructs increasingly expressive solutions through three coupled operations: switch expansion, switch placement, and route refinement. These operations preserve routing validity by construction, restricting exploration to feasible configurations where every communicating initiator-target pair has one assigned loop-free route. We explore the induced solution space using Gumbel Monte Carlo Tree Search, showing that neural-guided search substantially improves solution quality over non-learning optimization methods. Furthermore, pretraining across floorplans provides a strong initialization for fine-tuning on unseen instances.
[LG-49] he Nixtlaverse: An Open-Source Ecosystem for Forecasting
链接: https://arxiv.org/abs/2609.39741
作者: Olivier Sprangers,Max Mergenthaler Canseco,Marco Peixeiro,Saul Caballero Ramirez,Mariana Menchero García,Jing-Qiang Goh,Han Wang,Nikhil Gupta,Rogelio Melo,Senbong Gee,Cristian Challu
类目: Machine Learning (cs.LG)
*备注: 18 pages, 3 figures, 6 tables. Submitted to the International Journal of Forecasting. Code and benchmark artifact: this https URL
Abstract:Large forecasting applications often combine statistical, machine-learning, and neural models. These families solve the same problem but differ in fitted state, training procedures, and how they parallelize work. Forecasting software must therefore either hide these differences behind a single estimator interface, or keep the families in separate packages, forcing users to rewrite data preparation and evaluation for every package. We present the Nixtlaverse, an ecosystem of open-source Python libraries for time series forecasting, as a case study of a third design: all libraries share the same long-format panel data and keyed forecast outputs, while every model family keeps its own specialized implementation. We demonstrate this design through three use cases on the public M5 competition data. First, we evaluate statistical, machine-learning, and neural models, and an external engine from a separate ecosystem, in a single rolling-origin evaluation with per-series and hierarchy-weighted metrics. Second, we profile runtime and peak memory from 100 to 30,490 series and locate each family’s bottleneck: statistical fitting scales approximately linearly in the number of series, feature construction dominates machine-learning memory, and neural training time is nearly independent of panel size under a fixed training budget. Third, we reconcile the forecasts of multiple engines, including the external one, over all 42,840 series of the M5 hierarchy, with sparse reconciliation where dense implementations exhausted memory. These use cases establish the costs, boundaries, and utility of shared data and output contracts. The Nixtlaverse has seen substantial public distribution, scholarly reuse, and adoption through other forecasting frameworks, and is released under permissive open-source licenses with public datasets, reproducible examples, and verifiable benchmark artifacts.
[LG-50] Stable Transformers for Graph Generation
链接: https://arxiv.org/abs/2609.39739
作者: Luca Miglior,Alessio Gravina,Davide Bacciu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Graph generative models increasingly rely on Graph Transformers (GT) to capture complex dependencies among nodes and edges. While deeper architectures should provide greater expressive capacity and a broader receptive field, their effectiveness can decline with depth: repeated self-attention progressively contracts node representations, impeding information flow and gradient propagation. We analyse this phenomenon from a dynamical systems perspective, focusing on how the denoiser’s spectral dynamics affect graph generation. We show that standard GT denoisers become increasingly dissipative as depth grows, leading to vanishing gradients and representation collapse. To isolate the effect of these dynamics, we construct a permutation-equivariant GT with inherently stable, non-dissipative transport. We also introduce a damping mechanism that continuously interpolates between non-dissipative and increasingly contractive regimes, enabling a direct assessment of how dissipation influences generation. Experiments on synthetic and molecular graph generation benchmarks show that the gap between these regimes widens with depth: non-dissipative dynamics preserve representation diversity and gradient flow, sustaining strong generative performance, whereas greater contraction progressively impairs it. These findings identify the denoiser’s dynamical regime as a key design factor for deep graph generative models.
[LG-51] CNCGEN: A Dataset and Framework for Machining Process Planning and Toolpath Generation from B-rep Models
链接: https://arxiv.org/abs/2609.39738
作者: Xiaolei Zhou,Boyi Lin,Yuchao Feng,Jianwei Zheng
类目: Machine Learning (cs.LG)
*备注: 21 pages, including references and appendix
Abstract:Learning to generate machining process plans and toolpaths from B-rep CAD requires coupling discrete operation decisions with continuous tool motion as the workpiece evolves. Correctly predicting an operation sequence does not by itself ensure correct material removal, because each toolpath acts on the stock left by preceding cuts. We formulate this problem around persistent manufacturing objects: object identity determines the target of an operation, while the evolving stock state conditions the generation of its toolpath. Based on this formulation, we propose CNCGEN, a dataset and learning framework for three-axis machining. CNCGEN-Dataset contains approximately 50k geometrically verified synthetic machining flows and 800 held-out real CNC records. Each flow aligns B-rep geometry with object-referenced operations, parameterized toolpaths, intermediate stock states, and verification outcomes, enabling supervision of the correspondence between planning decisions and their geometric effects. CNCGEN generates operations and toolpaths for selected objects step by step, updating a compact machining state to guide subsequent predictions. During training, a learned surrogate verifier provides material-removal feedback that links local predictions to their geometric consequences. Experiments on synthetic and held-out real CNC records show that CNCGEN improves the resulting workpiece geometry and reduces residual material and overcut compared with adapted CNC generation baselines.
[LG-52] A library for differentiable signal processing and machine learning on the sphere
链接: https://arxiv.org/abs/2609.39737
作者: Thorsten Kurth,Max Rietmann,Mauro Bisson,Andrea Paris,Alberto Carpentieri,Jean Kossaifi,Anima Anandkumar,Christian Hundt,Boris Bonev
类目: Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph); Geophysics (physics.geo-ph)
*备注:
Abstract:The two-dimensional sphere embedded in three-dimensional Euclidean space S2, plays a central role in a variety of scientific and engineering domains, including geophysics, planetary science, geodesy, atmospheric physics, quantum chemistry, cosmology, and virtual reality, among many others. As machine learning increasingly permeates these fields, the demand grows for robust tools that process and model functions on the sphere, while respecting the inherent topological and symmetry properties of the domain. We present torch-harmonics, a comprehensive library that offers efficient, differentiable implementations of advanced signal processing and machine learning (ML) methods for spherical data. These include the spherical harmonic transform (SHT), the spherical analogue of the Fourier transform, vector spherical harmonics, discrete-continuous and spectral convolutions, as well as both global and neighborhood spherical attention mechanisms. Beyond traditional representations, torch-harmonics provides the building blocks for state-of-the-art spherical ML architectures such as spherical transformers in order to enable scalable, rotationally-aware learning and inference in modern scientific and engineering applications.
[LG-53] SE-ADD: Self-Evolving Audio Deepfake Detection with Mistake-Driven Supervision
链接: https://arxiv.org/abs/2609.39679
作者: Rong Wan,Wei Xie,Jiaxi Li,Wenwu Wang,Lu Yin,Yiliao Song,Xilu Wang
类目: ound (cs.SD); Machine Learning (cs.LG)
*备注:
Abstract:Audio deepfake detection (ADD) must remain effective when new spoofing attacks emerge after deployment. Emerging audio language model (ALM)-based ADD methods are built on predefined supervision from ground-truth labels or verified forensic rationales. However, this paradigm overlooks an ALM’s own mistakes, which indicate where targeted supervision is most needed. To this end, we first introduce evolving spoofing environments for ALM-based ADD, where a new attack becomes dominant while previously observed attacks persist. Motivated by the above learning-from-mistakes perspective, we further propose SE-ADD, a self-evolving framework that iteratively adapts an ALM via low-rank adaptation (LoRA) using mistake-driven supervision built from its verdicts and self-generated forensic cues. All training samples receive direct authenticity supervision, while misclassified ones receive additional cue-augmented supervision. As verdicts and cues are regenerated by the updated ALM, the resulting supervision evolves accordingly. Experiments on two ALMs demonstrate the effectiveness of SE-ADD in generalizing to unseen attacks, reducing the equal error rate (EER) from 36.72% to 7.52% for Qwen2-Audio and from 19.93% to 3.97% for MOSS-Audio.
[LG-54] NodeGround: A Node Classification Benchmark in the Graph Foundation Model Era
链接: https://arxiv.org/abs/2609.39673
作者: Jinmo Lee,Dooho Lee,Minho Jeong,Jaemin Yoo
类目: Machine Learning (cs.LG)
*备注:
Abstract:Can a pretrained graph model replace training and tuning a separate predictor for each dataset? Answering this requires evaluating prediction quality alongside computational cost. We present NodeGround, a node classification benchmark that puts graph foundation models (GFMs) and dataset-specific supervised learning under a common evaluation framework. The benchmark spans 51 datasets and evaluates six GFMs alongside 15 supervised methods under two label-availability regimes. Shared data partitions, validation-only model selection, controlled hyperparameter searches, and multiple predictive metrics make comparisons systematic, while workflow measurements account for adaptation, training, tuning, and inference. The results favor carefully tuned graph neural networks overall. GraphPFN reaches third place by Elo when more labels are available, yet its relative strengths vary substantially with dataset properties. Efficiency comparisons further qualify the benefits of pretrained reuse: GVT and GraphPFN appear on the Pareto frontiers when supervised methods are represented by their default and fully tuned configurations. Adding intermediate tuning budgets removes this advantage for GVT and leaves GraphPFN extending the estimated frontier in the label-rich setting alone. Thus, reusing pretrained parameters does not yet provide a broadly reliable route to either stronger predictions or cheaper workflows. We release the evaluation pipeline, run-level records, and an open leaderboard at this https URL.
[LG-55] Graph Residual Conjugate Diffusion: SNR-Equalized Heat Flow for Graph Signals
链接: https://arxiv.org/abs/2609.39658
作者: Jinwei Li,Daniel Tenbrinck
类目: Machine Learning (cs.LG)
*备注: 23 pages, 2 figures
Abstract:Diffusion models generate data by reversing a forward corruption process that typically approaches a simple Gaussian prior. Recent work has extended this framework to signals supported on fixed graphs, e.g., road-network traffic and sensor-network measurements. Many graph signals have nonuniform spectral energy, whereas isotropic corruption adds the same conditional noise variance to every graph-frequency mode. Driving all modes to near-zero terminal signal-to-noise ratio (SNR) requires strong corruption, which increases the noise range that must be covered under a fixed sampling budget. We introduce Graph Residual Conjugate Diffusion (GRCD), which replaces the shared clock of graph heat diffusion with a mode-dependent clock that gives every graph-Fourier mode the same conditional SNR. GRCD fits a zero-mean graph-spectral Gaussian reference on the training split and stops at a finite terminal SNR at which the propagated reference still carries the fitted spectral variances. The Gaussian component has an exact modewise propagator in the probability-flow ODE, so sampling advances it analytically and integrates only the learned residual score numerically. We evaluate GRCD on five settings (METR-LA traffic, Molene weather, and three stochastic block models) against seven comparators under a matched protocol: Graph-Aware Diffusion (GAD), EDM (graph backbone), two adaptations of Whitened Score Diffusion (WSD), and three preconditioning controls. At four function evaluations (NFEs), GRCD lowers averaged maximum mean discrepancy (aMMD) by 22 to 36 times over the best comparator on all five settings, reaching 0.054 on METR-LA, where it clears an aMMD 0.1 target with 87% less sampling wall-clock time than the cheapest comparator that reaches it. Fitting the terminal reference reduces aMMD by 2.7 to 7.3 times at finite terminal SNR, while the factors shrink to 1.00 to 1.01 near zero.
[LG-56] Beyond Uniform Compression: Budgeted Transmission Allocation for Extreme Federated Learning
链接: https://arxiv.org/abs/2609.39646
作者: Pengfei Li,Mohammad Khalil
类目: Machine Learning (cs.LG)
*备注:
Abstract:Federated learning faces severe communication bottlenecks when clients upload high-dimensional model updates. Existing methods often compress these updates uniformly across all layers. This uniform approach ignores the heterogeneous value of different parameter blocks and wastes limited bandwidth on insensitive layers. To address this issue, we propose Layer-wise Budgeted Adaptive Transmission (LBAT). LBAT reframes federated communication under extreme uplink budgets as a resource allocation problem. Our framework dynamically estimates the transmission value of different layers utilising local training signals. It then employs an exact byte dynamic programming allocator to determine optimal rank and bit configurations under strict budgets. We validate LBAT on highly heterogeneous federated tabular prediction and data generation tasks. Extensive experiments demonstrate that LBAT consistently outperforms uniform rank, uniform quantisation, and fixed compression baselines across various extreme budget regimes. Furthermore, it achieves significantly better communication and utility tradeoffs while preserving essential distributional fidelity.
[LG-57] RiboUnmix: Learning Shared Translational Dynamics from Biased and Noisy Ribo-seq Measurements
链接: https://arxiv.org/abs/2609.39644
作者: Gabriele Martino,Denis Skibinski,Ivo L. Hofacker,Sebastian Tschiatschek
类目: Machine Learning (cs.LG)
*备注:
Abstract:Ribosome profiling (Ribo-seq) measures ribosome distributions along mRNAs, but observed occupancy profiles also contain experiment-specific distortions and stochastic variability. Consequently, models that accurately predict measured profiles may reproduce technical effects rather than recover the underlying biology. We ask whether jointly modeling datasets collected under different experimental conditions can reveal shared, sequence-dependent patterns of ribosome occupancy. We introduce RiboUnmix, a probabilistic multi-dataset framework in which each expected measured profile is represented as a shared sequence-dependent signal modulated by a dataset-specific multiplicative factor. A negative-binomial observation model captures variability across replicates. We evaluate RiboUnmix on a controlled synthetic benchmark combining programmed translation kinetics, ribosome traffic, stochastic count sampling, and sequence-dependent experimental distortions. Because the underlying kinetics and distortions are known, recovery of the shared profile and dataset-specific effects can be assessed separately. Both inferred components correlate strongly with their targets, demonstrating that RiboUnmix can disentangle shared kinetic patterns from experimental effects. Across four organism-specific real-data benchmarks, RiboUnmix outperforms sequence-to-profile baselines in predicting measured profiles. Models trained independently on subsets of 114 HEK-derived datasets recover concordant shared profiles for held-out transcripts, and experiments varying the number and composition of training datasets show that the learned representation remains stable. RiboUnmix thus converts variation across experiments into evidence for reproducible sequence-dependent patterns of ribosome occupancy, supporting biological hypothesis generation from diverse Ribo-seq datasets.
[LG-58] owards Better Exploration in Sequential Test-Time Scaling
链接: https://arxiv.org/abs/2609.39632
作者: Joseph Rance,Fabio Pizzati,Juil Sock,Woody Bayliss,Marc Górriz Blanch,Philip Torr,Adel Bibi
类目: Machine Learning (cs.LG)
*备注: 26 pages, 15 figures
Abstract:Test-time scaling improves language model reasoning by spending additional compute at inference. However, both classes of existing methods often fail to continue improving over long timescales. Parallel methods repeatedly sample independent answers from the model, scaling poorly on problems the model is unlikely to solve in a single attempt. In contrast, sequential methods build on previous answers to access new ideas, yet so far have not been shown to reach answers beyond those found by parallel scaling. First, we show that sequential scaling often stops improving because it becomes prematurely trapped in an attractor: a set of answers that prevents exploration of different answers once entered. Across 27 combinations of scaling methods, models, and benchmarks, we find that 53.8% of sequential scaling trajectories enter an attractor within four iterations. Second, we show that a simple model-mixing intervention helps escape attractors. This reduces the attractor hit rate by 21.2 percentage points on average, expands solution coverage beyond a compute-matched parallel baseline, and improves accuracy of recursive self-aggregation by at least 2.2 percentage points. Our results motivate refocusing long-horizon test-time scaling from parallel methods to sequential methods that improve previous answers.
[LG-59] PEG-Tab: Sampling-Time Record Repair and Release Control for Tabular Synthesis
链接: https://arxiv.org/abs/2609.39630
作者: Pengfei Li,QinYi Liu,Mohammad Khalil
类目: Machine Learning (cs.LG)
*备注:
Abstract:Pretrained tabular generators can reproduce training records even when aggregate utility remains high. When retraining is unavailable or too costly, sampling and release are the remaining intervention points. We present PEG-Tab (Post-Training Energy Guidance for Tabular Synthesis), a post-training repair and release-control framework for frozen tabular generators. For each generated row, a generator-native operator creates two alternatives. A shared calibrated score compares the three candidates, favours lower-risk records, and applies a final release check. We instantiate this interface for GReaT, CTGAN, TVAE, and TabDDPM without updating their parameters. Across five datasets and four generator families, PEG-Tab reduces mean Near Copy from 0.078 to 0.027 and lowers aggregate Exact Copy to zero. Relative to a 3\times post hoc filter, it retains higher utility in 12 of 16 transfer settings and Pareto-dominates the filter in eight. Gains are concentrated in copy and proximity-related risks.
[LG-60] Certification-Based Differentially Private Learning
链接: https://arxiv.org/abs/2609.39629
作者: Mihnea Ghitu,Matthew Wicker
类目: Machine Learning (cs.LG)
*备注:
Abstract:Differential privacy (DP) in machine learning is typically achieved by adding noise to model parameters (private learning) or to model outputs (private prediction). Recent work uses formal methods, namely abstract interpretation, to provide tighter privacy guarantees, but only for private prediction in classification settings. In this work, we investigate the use of formal methods as a general tool for tighter privacy analysis. First, we generalize the abstract gradient training (AGT) framework to private prediction in continuous, unbounded regression. Second, by reducing learning in parameterized models to a regression problem over the parameter space, we introduce Abstract Gradient Sampling (AGS), an algorithm that enables reachability-based analysis to provide guarantees for private learning. In both private prediction and private learning, we provide tightened privacy accounting for the AGT framework and a theoretical analysis demonstrating when our smooth sensitivity upper-bounds yield favourable privacy-utility trade-off. In practice, we validate that our regression bounds are tighter than global-sensitivity baselines on regression benchmarks, and, notably, yield the first finite privacy guarantees in settings where global prediction sensitivity is a priori unbounded. We also find that under matched conditions, our private learning algorithm can outperform standard private learners.
[LG-61] MIND: Marginal-Invariant Neural Dependency Diffusion for Mixed-Type Tabular Generation
链接: https://arxiv.org/abs/2609.39628
作者: Pengfei Li,Mohammad Khalil
类目: Machine Learning (cs.LG)
*备注:
Abstract:This paper proposes MIND, a marginal-invariant neural dependency diffusion model for mixed-type tabular data. MIND does not directly learn the joint distribution in the original heterogeneous feature space. Instead, it first maps different variable types into a unified latent dependency space via column-wise marginal transport. A conditional diffusion model then learns cross-column relationships. Copula-tangent denoising separates known marginal components from learnable dependency residuals. Rank projection during the sampling phase further mitigates marginal shift in reverse diffusion. Experiments across nine diverse tabular benchmarks show that MIND consistently improves marginal fidelity and dependency preservation over existing unified approaches. By explicitly isolating marginal modelling from dependency learning, MIND achieves a strong and stable balance among marginal fidelity, joint dependency preservation, and downstream prediction utility. This work supports separating marginal and dependency modelling as a principled and highly effective paradigm for complex mixed-type tabular generation.
[LG-62] Hybrid Methods for Robust Tabular Data Imputation
链接: https://arxiv.org/abs/2609.39613
作者: Jinwei Li,Michelle Bruch,Daniel Tenbrinck
类目: Machine Learning (cs.LG)
*备注: 35 pages, 8 figures
Abstract:Missing data are a fundamental challenge in statistical analysis and machine learning, as the choice of imputation method substantially impacts downstream inference. In this work, we propose two hybrid imputation methods called NuclearForest and SoftForest, which combine nuclear-norm-based low-rank initialization using Singular Value Thresholding (SVT) and SoftImpute, respectively, with a non-iterative Random Forest refinement. For the SVT-based component, we further introduce an adaptive step-size rule, prove adaptive step-size bounds, and establish convergence for the corresponding zero-initialized iteration. The low-rank initialization provides a structured warm start that captures the global covariance patterns in the data, while the subsequent Random Forest step recovers residual nonlinear signals encoding local dependencies. We conduct an extensive benchmark on diverse datasets from different application domains, comparing the proposed methods with seven established imputation methods under the Missing Completely at Random (MCAR), Missing at Random (MAR), and Missing Not at Random (MNAR) mechanisms across varying missingness rates. Our results demonstrate that NuclearForest and SoftForest match or exceed the imputation fidelity of state-of-the-art iterative methods such as MissForest, while significantly reducing computational cost. In particular, they achieve speedups of approximately 5.81 times and 9.52 times over MissForest by replacing iterative cycles with a single refinement step. Our approach effectively exploits the low-rank structure of real-world tabular data and accommodates mixed-type variables, providing an efficient and robust solution for data imputation in bioinformatics, economics, and beyond.
[LG-63] Convergence of Practical Muon with Finite Newton-Schulz Iterations and Nesterov Momentum
链接: https://arxiv.org/abs/2609.39595
作者: Hanyng Peng,Hui Wang,Yue Yu
类目: Machine Learning (cs.LG)
*备注: 26 pages
Abstract:Practical Muon maintains momentum and performs a small, fixed number of Newton–Schulz iterations separately for each parameter matrix, often with a Nesterov correction. We analyze these layer-wise finite-step updates jointly on a coupled nonconvex objective, rather than replacing them by exact polar factors or one global orthogonalization. Under gradient-dependent (\mathcal L_0,\mathcal L_1,q) -smoothness and conditionally unbiased stochastic gradients with bounded layer-wise variance, we establish an \mathcal O(T^-1/4) bound on the expected average Frobenius gradient norm. The analysis retains the Nesterov recursion and requires neither bounded stochastic gradients, symmetric noise, nor a uniform positive lower bound on the nonzero output singular values. Its constants contain no explicit matrix-dimension or rank factors when the number of blocks and problem constants are fixed. The proof follows a descent inequality and a decomposition of the momentum tracking error into initialization, noise, and drift. For the original five-step quintic, we verify the required scalar-map bounds analytically; the result also allows step-dependent coefficients satisfying the same bounds. A complementary nuclear-norm result quantifies rank dependence under a stronger spectral condition. The vanishing rate uses coupled learning-rate and momentum schedules, including the standard single-coefficient Nesterov rule.
[LG-64] Cybersecurity in Edge Computing: A Trust-Aware Federated Hybrid Intrusion Detection Framework
链接: https://arxiv.org/abs/2609.39584
作者: Zawad Yalmie Sazid,Robert Abbas
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 10 pages, 7 figures
Abstract:Edge computing has emerged as a critical computing paradigm in modern distributed systems by migrating data processing closer to end users and Internet of Things (IoT) devices. While this paradigm decentralizes processes, minimizes latency, and reduces backhaul bandwidth congestion, it exponentially enlarges the cyberattack surface. Heterogeneous, resource-constrained edge devices deployed across unmanaged administrative domains present highly vulnerable targets. To address these vulnerabilities without compromising global data privacy regulations, this paper proposes a novel Trust-Aware Federated Hybrid Intrusion Detection Framework (TA-FHIDF). The proposed framework integrates an Autoencoder, a 1D Convolutional Neural Network (1D-CNN), and a Bidirectional Long Short-Term Memory (BiLSTM) model into a unified, localized deep learning engine capable of autonomous spatial and temporal feature extraction. Model training is performed collaboratively via federated learning, ensuring raw network telemetry remains isolated at local gateways. Furthermore, to defend against adversarial model poisoning attacks, we introduce a robust server-side trust-aware aggregation mechanism that evaluates client reliability using a cosine similarity metric before global model integration. Empirical evaluations across multi-vector benchmark datasets (UNSW-NB15, CICIDS2017, and Edge-IIoTset) demonstrate the framework’s superior detection accuracy, rapid convergence, and high Byzantine fault tolerance under adversarial attack scenarios.
[LG-65] Self-Repulsive Sampling for Diffusion Language Models
链接: https://arxiv.org/abs/2609.39560
作者: Michael Helcig,Martin Jaggi
类目: Machine Learning (cs.LG)
*备注:
Abstract:Sampling several responses and voting over their answers can improve a language model’s accuracy, but repeated answers limit the benefit of additional samples. Raising temperature increases diversity at a potential cost to per-sample accuracy. We introduce Self-Repulsion (SR), a sampler for masked diffusion language models that uses peer commitments to diversify the pool. At each penalized denoising step, each path lowers a token’s logit according to how many peers have committed that token at the same position. Paths share a batched forward pass and then commit in sequence, so later paths observe choices made earlier in the same step. This coupling requires no training or additional forward or backward pass and can produce distinct paths even at temperature zero. When all paths commit a position together from identical logits, the update exactly maximizes total logit minus a convex duplication cost. On LLaDA-8B-Instruct with ten paths and 128 denoising steps, deterministic SR reaches 80.38% plurality accuracy on GSM8K, compared with 70.17% for the unpenalized greedy decoder. At temperature 0.6 and matched model-evaluation budgets, the count penalty improves over self-consistency by 2.06 percentage points in blocks of 32 and 14.50 under pure diffusion. Experiments on GSM8K, MATH and TruthfulQA show that voting gains arise mainly from higher coverage of correct answers, with gains that vary by benchmark and decoding regime.
[LG-66] General Performance Guarantee for Human Torque Estimation-Based Task-Agnostic Assistive Exoskeleton Control
链接: https://arxiv.org/abs/2609.39558
作者: Duy Hoang,Bastien Berret,Olivier Bruneau,Laurent Fribourg
类目: Robotics (cs.RO); Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注: 10 pages, 8 figures
Abstract:Accurate human torque estimation is crucial for enabling task-agnostic control in robotic exoskeleton systems. However, estimation errors may cause mismatches between the robot assistance and the human intention, degrading controllability and task performance. In this paper, we address this issue by formally defining matched assistance as scenarios in which the robot positively contributes to human movement. Based on this definition, we develop a theoretical framework to design the robot’s desired interaction torque that guarantees a lower bound on the matched assistance probability. Importantly, the proposed guarantee holds over the entire torque distribution, including unseen data beyond the training tasks. This provides our method with strong reliability and generalization, both of which are critical for effective exoskeleton control. The proposed strategy is implemented on the ABLE upper-limb exoskeleton and evaluated in a multi-task setup. Experimental results validate the theoretical guarantees and demonstrate that the proposed strategy achieves effective general performance across several tasks, guaranteeing movement smoothness while reducing human physical effort.
[LG-67] Ghost in the Encoder: Decodable Artist Identity Representations in Lyrics-to-Song Generation
链接: https://arxiv.org/abs/2609.39552
作者: Arhan Vohra,Choenden Kyirong,Laura Ibáñez-Martínez,Martín Rocamora
类目: ound (cs.SD); Machine Learning (cs.LG)
*备注: 8 pages, 3 figures. Accepted to ISMIR 2026
Abstract:Text-to-song generation models can be prompted to imitate specific artists or regurgitate entire songs from their training data. Although these phenomena have been documented behaviorally on small datasets, little is known about the internal representations that may give rise to them. Prior interpretability work on generative audio has focused on locating semantic concepts such as genre or time signature within model activations. In this work, we show that a trained model can be probed for linearly decodable representations of artist identity from song lyrics alone, without any additional identifiers. Through a controlled case study of ACE-Step 1.5 spanning 2,000 songs across 100 artists, we demonstrate that the artist associated with a given set of lyrics can be identified within the model’s internal activations, and that this conditioning signal propagates from the lyric encoder to the diffusion backbone during inference. These findings indicate that lyrics constitute an artist-level conditioning channel not addressed by prompt-side replication safeguards. More broadly, our work highlights how latent-space analysis can be used to audit what generative music models have implicitly learned from their training data.
[LG-68] Hyperbolic Prototype Routing for Rehearsal-Free Class-Incremental Learning
链接: https://arxiv.org/abs/2609.39550
作者: HongWei Zhao(Beihang University),Rui Liu(Beihang University),Yong Chen(Beijing University of Posts and Telecommunications)
类目: Machine Learning (cs.LG)
*备注: 6 pages, supplementary material
Abstract:Class-Incremental Learning (CIL) aims to continually learn new classes while preserving prior knowledge. Parameter-efficient fine-tuning with pre-trained models enables CIL with minimal parameter updates, but existing approaches still suffer from catastrophic forgetting caused by cumulative interference and suboptimal module-sample matching at inference. We propose Hyperbolic Prototype Routing (HyPro), a rehearsal-free framework for continual learning. HyPro allocates a dedicated LoRA-Expert module to each incremental task for isolated representation learning, then projects routing features onto a Poincare ball and performs geodesic nearest-prototype matching for reliable task-level discrimination. Extensive experiments on standard CIL and Few-Shot CIL benchmarks show that HyPro consistently improves average and final-stage accuracy over strong baselines.
[LG-69] Learning Reliable GUI Agents under Imperfect Priors
链接: https://arxiv.org/abs/2609.39547
作者: Bo Han,Qianyi Wang,Shuai Liu,Xiong Zifan,Changqiao Wu,Yuanfa Li,Pengzhi Gao,Wei Liu,Jian Luan,Heng Qu,Yunpeng Song,Zhongmin Cai
类目: Machine Learning (cs.LG)
*备注: 19pages, 4 figures
Abstract:GUI agents built on large language and vision-language models still struggle on unseen applications and complex multi-step tasks, as completing real GUI tasks depends on app-specific, temporally volatile operational knowledge that is scarce in pretraining corpora. Retrieval-augmented execution offers a natural remedy but faces two coupled bottlenecks: knowledge at scale is hard to acquire, and self-collected priors inevitably drift from the live environment due to version updates, promotions, ads, A/B tests, and personalization. We therefore argue that GUI agents should not pursue perfect knowledge but learn to act correctly under imperfect priors, and propose our framework that couples knowledge acquisition with noise-robust utilization: a structured exploration strategy traverses interactive elements, builds a UI state-transition graph, and synthesizes (task, trajectory) pairs via a VLM without human annotation; a noise-aware training strategy, grounded in a taxonomy of real GUI drift patterns, injects five types of realistic errors into self-explored trajectories to teach the agent to assess prior reliability before acting. Experiments on physical devices and online emulator benchmarks show that our method discovers more unique screens, covers more benchmark tasks, and more effectively rejects erroneous priors while leveraging correct ones, with accuracy gains that transfer across datasets.
[LG-70] CIDER-FM: Foundation Models for Causal Inference from Diverse Experimental Regimes
链接: https://arxiv.org/abs/2609.39523
作者: Yuche Gao,Arik Reuter,Siyuan Guo,Anish Dhir,Bernhard Schölkopf,Adrian Weller
类目: Machine Learning (cs.LG)
*备注: 31 pages, including appendices
Abstract:Causal foundation models (CFMs) amortise causal inference over priors of synthetic structural causal models (SCMs), predicting the effect of an experiment on a specific variable. However, observational data alone may leave multiple causal models compatible with available evidence, while experimental data with interventions on exactly the variable of interest might be unavailable. This work studies CFMs as a method to combine finite observational and surrogate-interventional datasets in order to predict a target conditional interventional distribution (CID) more accurately than with observational data alone. We first formalise the conceptual benefits of surrogate experiments. Building on this analysis, we introduce \textscFoundation Models for Causal Inference from Diverse Experimental Regimes (\emphCIDER-FM), a causal foundation model that uses an intervention-aware representation and hierarchical three-axis attention to exchange information across variables, samples, and experimental regimes. We evaluate CIDER-FM against a wide range of baselines across diverse synthetic graph and mechanism families, as well as on both simulated and real-world data from Causal Chambers. Our results demonstrate strong CID prediction performance and show that incorporating experimental context can improve predictions over observational data alone.
[LG-71] Can Domain Generalization be Guaranteed in Small-Sample Learning?
链接: https://arxiv.org/abs/2609.39512
作者: Hong Zheng
类目: Machine Learning (cs.LG)
*备注:
Abstract:The small-sample learning problem remains a fundamental challenge in machine learning because limited training data lead to unstable model estimation and generalization. Structural Risk Minimization (SRM) has long been regarded as a principled solution under the classical i.i.d. assumption. However, domain generalization (DG) violates this assumption, leaving the theoretical role of SRM in DG largely unexplored. To bridge this gap, we establish the first theoretical guarantees for SRM in DG under mild assumptions. Specifically, based on the concept of stability, we derive learning consistency and generalization error bounds and prove that these bounds become tight when the hypotheses satisfy the stability condition. Building upon this, under a specific hypothesis space assumption, we establish stability, learning, and generalization bounds for SRM. We further discuss the applicability of these bounds to deep learning. This work establishes theoretical foundations for SRM under distribution shifts and sheds light on the design of robust DG algorithms in small-sample scenarios.
[LG-72] he Geometry of Randomized Smoothing on Feasible Sets
链接: https://arxiv.org/abs/2609.39497
作者: Syed Izhan Khilji,Alireza Furutanpey,Schahram Dustdar
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Randomized smoothing certifies the probability of a fixed output event as the center of Gaussian noise moves. Feasibility or confidence filtering reports label probabilities only among retained proposals, producing a ratio. Its numerator is a fixed Gaussian event mass, while its denominator is the probability of retention and can change with the center. Substituting this ratio into the ordinary smoothing formula can therefore certify a ball that contains a decision boundary. We separate the problem into a geometric question and a certification question. Geometry determines when conditioning preserves Gaussian comparisons. Convex retained sets preserve the full comparison, while general sets require geometric control of the retained law as the center moves. Without such control, conditional probabilities imply no positive universal radius. Joint retention-and-label probabilities always yield a valid certificate for the same filtered predictor. A uniform covariance bound transfers divergence certificates to the retained law and can yield larger radii even when the Gaussian event comparison fails. Both methods admit finite-sample bounds. For a learned image classifier with a training-selected nonconvex filter, conditional Rényi bounds certify more images than joint-mass bounds without additional model evaluations. A released confidence filter exhibits verified label changes inside radii obtained by conditional substitution. An application of adaptive Gaussian composition covers causal finite-horizon executions with history-dependent center shifts under a pathwise energy bound.
[LG-73] owards Robust Time Series Learning via Capacity-Centric Modulation
链接: https://arxiv.org/abs/2609.39489
作者: Siru Zhong,Senzhang Wang,James T. Kwok,Yuxuan Liang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Sample-level reliability heterogeneity is common in deep time series learning. Standard training pipelines apply a uniform regularization setting to all samples, which can under-regularize corrupted samples and over-restrict clean samples. Common robustness approaches filter observations in data space or impose priors on latent representations. We propose Capacity-Centric Modulation (CCM) as a complementary, sample-adaptive regularization principle. Under this principle, we introduce SACM (Sample-Adaptive Capacity Modulation), a task-agnostic framework that exploits spectral sparsity to assign sample-wise dropout probabilities along internal activation paths. SACM integrates into existing backbones without architectural redesign and preserves the deterministic inference pipeline. Across 301 real-world dataset-backbone pairs covering 9 forecasting, 32 classification, and 4 anomaly-detection datasets, SACM reduces forecasting MSE by 6.7% on average and improves classification accuracy and point-adjusted F1 by 3.04% and 17.05%, respectively, relative to unmodified backbones, with zero test-time overhead.
[LG-74] Correcting CondOT: Exact Finite-Step Sampling in Gaussian Flow Matching
链接: https://arxiv.org/abs/2609.39488
作者: Ron Levy,Michael Elad
类目: Machine Learning (cs.LG)
*备注: 38 pages, 5 figures
Abstract:Flow matching generates samples by gradually transforming noise into data. In practice, using a finite number of sampling steps introduces a numerical error that depends on the chosen schedule. We study this dependence for Gaussian targets and the explicit midpoint sampling method, using the exact flow field. We measure sampling error by the squared Wasserstein distance between the target distribution and the final distribution produced by the midpoint sampler. We show that the standard conditional optimal transport (CondOT) schedule cancels the leading midpoint error and improves the general convergence bound, even when the sampling steps are unequally spaced. On a uniform grid of S sampling steps, we fix the signal schedule at \alpha_t=t and prove the existence of scalar noise schedules \beta_t that approach the CondOT noise schedule 1-t at rate 1/S and yield exact Gaussian sampling for every sufficiently large S . Controlled Gaussian experiments illustrate the convergence rates and exact calibration.
[LG-75] About the Influence of Workflow Topology on Task Intensity Prediction through Graph Learning
链接: https://arxiv.org/abs/2609.39481
作者: Max Otto,Haci Ismail Aslan,Joel Witzke,Jonathan Bader,Odej Kao
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: Published at the IEEE 26th International Symposium on Cluster, Cloud and Internet Computing Workshops (CCGridW)
Abstract:Efficient resource provisioning for large-scale workflows on cloud infrastructures is a critical performance engineering challenge. These workflows are often structured as directed acyclic graphs (DAGs), where under-provisioning can cause critical bottlenecks and over-provisioning leads to unnecessary costs. Accurate, task-level prediction of resource intensity (e.g., CPU load and memory usage) is essential for mitigating these issues. While task-level features are commonly used for prediction, the performance impact of the workflow’s overall topological structure is often overlooked or assumed. The central question of our work is: To what extent does what part of the DAG topology influence task-level resource intensity, and what is the most effective way to model this influence? This paper presents a comprehensive benchmark to systematically quantify the impact of graph topology on task intensity prediction. We evaluate and compare a spectrum of modeling approaches. Our findings demonstrate that topology is a critical feature for accurate prediction. Models incorporating important topological information, even through simple handcrafted features, significantly outperform baseline models. We show that graph-native models provide the highest accuracy, achieving low mean absolute errors for both CPU and memory predictions, and can still be combined with simple topological features that they do not learn for better performance. Comments: Published at the IEEE 26th International Symposium on Cluster, Cloud and Internet Computing Workshops (CCGridW) Subjects: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG) Cite as: arXiv:2609.39481 [cs.DC] (or arXiv:2609.39481v1 [cs.DC] for this version) https://doi.org/10.48550/arXiv.2609.39481 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Journalreference: Otto, Max, et al. “About the Influence of Workflow Topology on Task Intensity Prediction Through Graph Learning.” 2026 IEEE 26th International Symposium on Cluster, Cloud and Internet Computing Workshops (CCGridW). IEEE, 2026 Related DOI: https://doi.org/10.1109/CCGridW69005.2026.00044 Focus to learn more DOI(s) linking to related resources
[LG-76] Also Small Models Can Reason ably Self-Evaluate Their Confidence
链接: https://arxiv.org/abs/2609.39478
作者: Idil Kapikiran,Thomas Decker,Thomas Runkler
类目: Machine Learning (cs.LG)
*备注:
Abstract:This study systematically evaluates self-evaluation-based uncertainty quantification across different language models of varying sizes on question-answering tasks spanning general to specialized knowledge domains. Using various self-evaluation methods where models judge their own predictions, we examine how model scale and domain specificity affect the quality of self-assessed confidence signals. Our results reveal that while accuracy predictably declines with smaller models and more specialized domains, the reliability of self-evaluated confidence remains largely stable across both dimensions. This independence means the most capable model is not necessarily the best at self-assessing prediction reliability. These findings suggest that smaller models can achieve reasonable self-assessed confidence despite lower accuracy, making them viable for resource-constrained deployments.
[LG-77] -ARC: Topology-Aware Randomized Clustering via Distributionally Robust Stochastic Block Models
链接: https://arxiv.org/abs/2609.39466
作者: Serena Grazia De Benedictis,Andersen Ang,Nicoletta Del Buono,Flavia Esposito,Laura Selicato
类目: Machine Learning (cs.LG); Algebraic Topology (math.AT); Optimization and Control (math.OC)
*备注: 19 pages, 17 figures. Preprint
Abstract:In this work, we introduce a new clustering method, namely T-ARC (Topology-Aware Randomized Clustering), that corrects the geometric bias of K-means by embedding topological information directly into the optimization objective. Building on the assumption that the data admits an underlying hidden structure modeled via a latent graph, the idea is to uncover this information through the interplay between the standard K-means data-fidelity term and a graph-cut penalty, which discourages cluster assignments inconsistent with the connectivity structure of the data. To render this coupling tractable, the latent graph is modeled as a random realization from a Stochastic Block Model (SBM), whose scalar parameter is optimized within a Distributionally Robust Optimization (DRO) framework, yielding a closed-form proximal update. Both SBM and DRO are informed by a persistence-based similarity matrix derived from zero-dimensional persistent homology ( H_0 ), which translates the multiscale connectivity structure of the data into a pairwise topological prior. The overall optimization proceeds via Block Coordinate Descent; convergence is established through a global Lyapunov functional: the deterministic blocks satisfy monotonic descent, while the stochastic graph update satisfies descent in expectation, so that the expected energy converges. Experiments on synthetic datasets with non-convex geometries and on random subsets of Fashion-MNIST show that T-ARC recovers latent topological structures where K-means fails, achieving the highest accuracy on curved and interleaved clusters while remaining competitive, and markedly more stable than K-means, on real data. Comments: 19 pages, 17 figures. Preprint Subjects: Machine Learning (cs.LG); Algebraic Topology (math.AT); Optimization and Control (math.OC) Cite as: arXiv:2609.39466 [cs.LG] (or arXiv:2609.39466v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.39466 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-78] Understanding Head Geometry and Dynamics in Federated Regression through a Natural Solution Selection Rule: An Unconstrained Feature Model Analysis
链接: https://arxiv.org/abs/2609.39464
作者: Chuang Ma,Tomoyuki Obuchi
类目: Machine Learning (cs.LG)
*备注: 47 pages, 17 figures, 9 tables
Abstract:In federated averaging, local objectives can admit multiple optimal heads, making the aggregate depend on which heads clients return. We study this ambiguity in federated multivariate regression with private backbones and a shared linear head, using an unconstrained feature model (UFM) that treats training-sample features as free variables. We introduce a natural selection rule: each client returns the optimal head closest to the broadcast head. We show that global minimization with a vanishing proximal penalty on the head realizes this rule. When the clients’ optimal Gram matrices and the initial shared Gram matrix are positive definite, the shared Gram matrix follows a closed recursion and converges to the unique Bures-Wasserstein barycenter of the clients’ optimal Gram matrices. Even with this alignment, the limit generally differs from the centralized optimal Gram matrix. We decompose this gap into three positive-semidefinite terms arising from differences in client target means, covariance heterogeneity, and averaging the aligned heads. A correction based on a one-time exchange of target means and covariances recovers the centralized optimal Gram matrix in one round under exact local optimization and the same selection rule. We verify these results numerically in the UFM and test its predictions on five tabular and five image regression datasets using deep networks with feature regularization and long local training. In these experiments, ordinary training approaches the predicted barycenter, while a weak proximal penalty improves endpoint agreement and yields trajectories that closely follow the predicted Gram dynamics. The correction moves the final Gram matrices close to the centralized UFM prediction.
[LG-79] Raw-Routed Mixture of Adapters: A Causal Intervention for Routing Collapse in Time Series Foundation Models NEURIPS2026
链接: https://arxiv.org/abs/2609.39445
作者: Hung Phan,Thuy T. Nguyen,Minh Ngoc Dinh,Nhat-Quang Tran
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: Accepted at NeurIPS 2026 (poster). 50 pages, 9 figures
Abstract:Time series foundation models (TSFMs) commonly adapt to new data by attaching a single trainable head to a frozen backbone, a one-size-fits-all setup that underfits heterogeneous regimes. Replacing the head with a mixture of experts is the standard upgrade, but on instance-normalized backbones (the dominant TSFM design class) it fails: routing entropy collapses to zero and one expert absorbs every input, a failure we call normalization-induced routing collapse. Standard MoE rescue mechanisms do not repair it, because the cause is in the router’s input, not its optimization. Pre-encoder normalization strips the statistics a router would need to tell regimes apart. A mutual-information decomposition makes this precise and yields a signal-ratio that, computed before training, predicts dataset vulnerability (Spearman \rho = -0.88 ). Eight causal controls, including a vision-modality replication, isolate instance normalization as the cause. The prescription is a minimal causal intervention: Raw-Routed Mixture of Adapters (RR-MoA), which routes on the raw, pre-normalization input. Under a strictly frozen backbone, RR-MoA wins 54/54 comparisons against the strongest fixed adapter and significantly outperforms LoRA, TRACE, AdaMix, and full fine-tuning. The effect generalizes across six backbones and an imputation task. Frozen RR-MoA also beats full fine-tuning by 12-79% (the Frozen Paradox); two architecturally distinct variants confirm the principle generalizes beyond this specific router.
[LG-80] Network-based Spatial Context Retrieval for Open-weight LLM s: A Faithfulness Benchmark for Grounded Geographic Reasoning
链接: https://arxiv.org/abs/2609.39437
作者: Joan Perez
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large language models (LLMs) encode substantial latent geographic knowledge, yet they reason poorly over space and are unreliable when queried from coordinates alone. Useful behaviour emerges only when structured spatial context is supplied in the prompt. This raises a question geographic evaluation has left unexamined: once the right context is supplied, does the model reason from it, or override it with its own parametric recall? We take up this question with an open pipeline for network-based spatial context retriev-al. In it, the surroundings of a selected point are defined by the pedestrian street network, the area actually reachable on foot. Using only open data and open-weight models, the pipeline retrieves features from OpenStreetMap and the GHS-POP population grid, computes indicators over the network catchment in code, and injects them as a compact spatial brief. On this basis we build a faithfulness benchmark. It labels every claim a model makes by its source (grounded in the brief, or drawn from training knowledge) and its correctness, and it probes each case with a planted false premise that the brief refutes. We evaluate sixteen open-weight model configurations across three families (Qwen, Gemma and Llama, with Gemma in two generations), four size classes and, where available, both thinking and non-thinking modes, on three con-trasting cities, resampling every case over ten seeds. The results show that resistance to the planted premise varies more strongly by model family and generation than by scale, while brief-reading competence forms a partly separate dimension. These behaviours are not captured by conventional world-correctness scores or single-shot evaluation. We release the implementation, spatial briefs, model outputs, and claim-level labels as a reproducible workflow at this http URL.
[LG-81] Awakening of the Buddha: Subspace Learning During Population-Loss Plateaus
链接: https://arxiv.org/abs/2609.39408
作者: Akash Kumar
类目: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
*备注: 127 pages, 27 figures
Abstract:Population loss can remain nearly constant while a neural network learns a substantially more predictive representation. We establish this separation for two-layer ReLU and leaky-ReLU networks trained on Gaussian inputs by simultaneous fixed-step population gradient descent on all parameters. For structured additive teachers whose links are positive mixtures of Gaussian-damped cubics in H^1(\gamma) , we give explicit conditions under which small IID Gaussian initialization yields a high-probability guarantee: at a checkpoint during a high-loss plateau, minimum alignment between the rank- r teacher subspace and the leading r -dimensional eigenspace of the predictor’s average gradient outer product (AGOP) increases by at least 1/2 , and the minimum refit MSE under unchanged coefficient budgets decreases by more than 0.399 , both relative to initialization. The same trajectory subsequently attains a trained loss below every value in the plateau window. A complementary result treats unequal-weight cubic teachers and small additive Sobolev perturbations using projected-feature refits. For SwiGLU networks with an exactly fitted intercept, we prove leading-AGOP alignment during a loss plateau at fixed width and dimension as Gaussian initialization vanishes, for square-integrable teachers with nonzero Hermite content of degree one, two, or three. A rank-one cubic specialization also gives simultaneous unrestricted-refit gains at a prescribed width. An approximation lower bound further shows that certain interaction targets retain nonzero error when ridge neurons are restricted to shared orthogonal axes within the teacher subspace. Population-moment experiments with ReLU students across 21 teachers and 50 initializations per teacher complement the analysis.
[LG-82] No Task Vector Is an Island: A Comprehensive Study on the Composability of Task Vectors from On-Policy Distillation
链接: https://arxiv.org/abs/2609.39405
作者: Jingang Zhou,Feiyu Han,Han Zhu,Yuyi Zhou,Ruiyang Zhang,Jian Xu,Sirui Gao,Qingpei Guo,Xu-Yao Zhang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Task vectors provide a simple mechanism for composing learned capabilities through model merging. However, the composability of task vectors produced by on-policy distillation (OPD) remains largely unexplored. OPD trains a student using teacher feedback on student-generated trajectories, yielding parameter updates that differ from those produced by the teacher model, usually by reinforcement learning (RL). We therefore ask whether OPD task vectors can complement their RL teacher updates and compose effectively across tasks. Across five domains and two model architectures, we find evidence for both forms of composability. Within a task, merging OPD and RL task vectors can outperform both constituent models, even when the OPD student is weaker than its RL teacher. Across tasks, OPD task-vector compositions achieve higher average scores than corresponding RL compositions in seven of eight backbone-merging-rule comparisons. Parameter-space analyses reveal substantial non-collinearity between OPD and RL updates. Experiment in CODE domain on SMOLLM3-3B shows that the combined direction outperforms either constituent direction at the tested global update norm, supporting directional complementarity in this configuration. Across tasks, OPD updates also show lower overlap among the top-10% feed-forward channels ranked by update energy. Together, these results show that weaker standalone performance does not imply weaker task-vector composability. OPD task vectors can complement stronger RL teacher updates and combine effectively across tasks, highlighting composability as a distinct property for understanding and evaluating post-training updates.
[LG-83] Decoupled and Distilled: Task-Adaptive LoRA-Teachers with Ensemble Knowledge Transfer for Few-Shot Class-Incremental Learning
链接: https://arxiv.org/abs/2609.39390
作者: Hongwei Zhao(School of Computer Science and Engineering, Beihang University),Rui Liu(School of Computer Science and Engineering, Beihang University),Yansong Liu(School of Computer Science and Engineering, Beihang University),Zhiyuan Zou(School of Computer Science and Engineering, Beihang University),Yong Chen(School of Computer Science, Beijing University of Posts and Telecommunications)
类目: Machine Learning (cs.LG)
*备注: Accepted manuscript; 46 pages, supplementary material
Abstract:Few-Shot Class-Incremental Learning (FSCIL) addresses the challenge of learning new classes from very limited samples while retaining knowledge of previously learned ones. Although parameter-efficient fine-tuning methods with pre-trained models show promise for class-incremental learning, strict gradient-based constraints can be unreliable under severe data scarcity, while multi-expert approaches can impose substantial inference-time costs. We propose TALON (Task-Adaptive LoRA-Teachers with Ensemble Knowledge Transfer), an inference-efficient FSCIL framework. TALON dynamically allocates an independent LoRA-Teacher to each incremental task for task-specific representation learning, then distills multiple frozen teachers into a unified LoRA-Student through Ensemble Knowledge Transfer, eliminating runtime module selection or generation. A semantic-guided distillation strategy weights teacher contributions by feature-space similarity to mitigate catastrophic forgetting and overfitting. Across three class-order runs, TALON achieves comparable or better mean average accuracy across four FSCIL benchmarks, obtaining 86.68 +/- 1.22% on CUB200, 90.39 +/- 0.27% on CIFAR100, 78.38 +/- 0.94% on ImageNet-R, and 96.34 +/- 0.33% on miniImageNet. TALON uses up to 33x fewer deployment parameters and reduces average inference time per task to 26.7 s, a 41.70% reduction relative to ASP.
[LG-84] Beyond Simulation: Retain-and-Repair Neural Operators for Real-World Adaptation
链接: https://arxiv.org/abs/2609.39387
作者: Woojin Cho,Junghwan Park
类目: Machine Learning (cs.LG)
*备注:
Abstract:Neural operators increasingly benefit from pretraining on numerical simulations, yet adapting them for real-world prediction remains challenging. We introduce the Retain-and-Repair Neural Operator (R ^2 NO), a framework for adapting simulation-pretrained operators to real-world data while retaining useful pretrained structure. The pretrained operator is first finetuned on real data and then frozen to provide a source prediction, and a shared repair module learns a sequence of refinements from the same observations. Using orthogonal Fourier projections, a spectral ensemble fits a small ridge regression within each cell of the Fourier domain and combines the refinements by weights fitted on a held-out split of the real data. The cells are defined jointly by radial ranges, angular sectors, and measured channels, allowing refinement depth to vary with frequency magnitude, with orientation, and across channels. Including the source prediction as a candidate makes retention available in every cell, and independently trained repair modules enter the same combination as additional candidates. On all RealPDEBench systems and six backbones, R ^2 NO consistently outperforms full finetuning and iterative refinement. The framework treats adaptation depth as a cell-specific choice learned from real data.
[LG-85] When Not How Much: Evaluating Time-Series Foundation Models on Sparse Events
链接: https://arxiv.org/abs/2609.39386
作者: Daniel Schoess,Florian von Wangenheim
类目: Machine Learning (cs.LG)
*备注:
Abstract:Pretrained time-series foundation models (TSFMs) are evaluated as forecasters of future values, yet for sparse series many decisions depend only on which future periods contain activity. Standard benchmarks do not assess this. On five sparse datasets, we rank positions within forecast windows that contain both events and zeros. The released point forecasts of 12 TSFMs improve chance-corrected average precision over training-free references by at most 0.031, and in chance-corrected AUC the median TSFM falls below them on every dataset. With event supervision, linear probes of six frozen backbones improve on their backbone’s point forecast in 29 of 30 backbone–dataset pairs. Averaging the predicted quantiles instead of taking their median improves the ranking of most TSFMs that forecast the median, and on two datasets the strongest such outputs rival the probes. The probes’ advantage over raw-context learners depends on the dataset, and under the same probe, pretrained features outperform randomly initialized ones for five of six backbones. For sparse-event ranking, released point forecasts thus add little over simple references, whereas lightweight event heads on frozen TSFMs rank events better than these forecasts, and the best of them exceed gradient-boosted trees trained on the raw context on three of the five datasets. More broadly, assessing pretrained forecasters on tasks beyond value forecasting requires reporting their outputs, supervised probes of their representations, and raw-context and randomized controls side by side, since each supports a different conclusion.
[LG-86] LampAttention: Look-Ahead Mixed-Precision FlashAttention for Dedicated Accelerators
链接: https://arxiv.org/abs/2609.39361
作者: Stanislav Budzinskiy,Marian Gloser,Tolunay Yilmaz,Ying Hong Tham,Yuanyi Lin,Wenyi Fang,Fan Wu,Philipp Petersen
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注:
Abstract:While most attention logits can be computed in low precision without degrading numerical stability, current attention kernels fail to exploit this phenomenon. We introduce a novel hardware-algorithm co-design in the form of mixed-precision FlashAttention. Our method accumulates key-query products and evaluates their exponentials in 8-bit formats, then adaptively identifies sensitive sub-blocks and recomputes them in 16-bit formats. We propose the specifications for a dedicated accelerator capable of executing this pipeline efficiently. Simulated experiments with Qwen3 and Gemma 3 show that rerouting a selective minority of sub-blocks to high precision is sufficient to recover the baseline model performance.
[LG-87] Robustifying Asynchronous SGD via Soft Throttling
链接: https://arxiv.org/abs/2609.39357
作者: Kaoru Otsuka,Maxime Meyer,Yuki Takezawa,Makoto Yamada,Anastasia Koloskova
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC); Optimization and Control (math.OC)
*备注:
Abstract:Asynchronous SGD is a popular algorithm for distributed learning where each client’s gradient update is applied on arrival. This leads to a speed-up, but also an increased vulnerability to attacks, as fast clients can dominate the total update. We introduce Throttle, a Byzantine-robust generalization of asynchronous SGD where the key idea is to exponentially down-weight updates from faster clients by a factor q . Both asynchronous SGD ( q=1 ) and synchronous Byzantine-robust SGD ( q\to\infty ) correspond to specific settings of Throttle. We provide a theoretical analysis of the convergence rate and validate the robustness to attacks both theoretically and empirically. Remarkably, our experiments show that this down-weighting mechanism can also improve performance over standard asynchronous SGD even in the non-Byzantine setting.
[LG-88] HAPMoE: Heterogeneity-Aware Automatic Parallelism Planning for Mixture-of-Experts Models Training EMNLP2026
链接: https://arxiv.org/abs/2609.39350
作者: Mengyuan Fan,Peizhuang Cong,Zixiao Huang,Si Xu,Tong Qiao,Yanghao Li,Jing Yang,Tong Yang,Quanlu Zhang,Yu Wang
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: Accepted to Findings of EMNLP 2026. 15 pages, 6 figures
Abstract:As model sizes continue to scale, distributed training has become inevitable. Automatic parallelization techniques can derive efficient training parallelism strategies at low cost while achieving superior performance. The difficulty of this problem is jointly determined by the complexity of the model and the underlying compute cluster. Meanwhile, mixture-of-experts (MoE) models are increasingly emerging as the dominant architecture and the rapid evolution of accelerator hardware has made cluster heterogeneity commonplace, posing substantial challenges to automatic parallelization. However, existing approaches typically target either MoE architectures or heterogeneous clusters, failing to generalize to scenarios where both challenges coexist. To this end, we present HAPMoE, a heterogeneity-aware automatic parallelism planner for MoE training. HAPMoE builds a lightweight MoE-aware cost model and efficiently searches a six-dimensional parallel space, producing parallel plans directly deployable on Megatron-LM. Experiments show that HAPMoE improves end-to-end training throughput by up to 3.2 \times over baselines across heterogeneous clusters. Its non-uniform pipeline partitioning yields an additional up to 78% gains, and its pruning-enhanced dynamic programming algorithm completes the search within 1 minute, demonstrating high efficiency and practical value in complex hardware environments.
[LG-89] ElectrolyteFM: Unifying Electrolyte Property Prediction through Cross-Property Knowledge Learning
链接: https://arxiv.org/abs/2609.39340
作者: Jiaxin Yu,Shuo Wang,Peng Wang,Yongcai Wang,Deying Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:Electrolyte formulation design requires balancing multiple physicochemical properties, yet existing models often focus on a limited subset. Learning each property in isolation can overlook transferable chemical information, whereas indiscriminate sharing can introduce cross-property interference. Our directed transfer analysis shows that jointly learning two property prediction tasks can improve or degrade prediction relative to separate training, with asymmetric transfer effects between the tasks. We propose ElectrolyteFM, a unified multi-property prediction model which can more accurately predict multiple properties of each electrolyte by effectively identifying and utilizing property-specific features and knowledge shared across properties. More specifically, ElectrolyteFM learns property-specific representations independently and captures cross-property knowledge through a separately trained expert pool. A router selects relevant shared information for each formulation and target property, and property-specific residual adapters convert this information into corrections to the corresponding representation for prediction. Experiments on Electrolyte12 show that ElectrolyteFM reduces normalized mean absolute error averaged across 12 electrolyte properties by 14.8% relative to the strongest electrolyte-specific baseline. On an independent sodium-electrolyte dataset unseen during training, it reduces conductivity mean absolute error by 6.7% relative to the best-performing baseline.
[LG-90] Learning Beyond Full Imitation: Task-Preserving Knowledge Distillation
链接: https://arxiv.org/abs/2609.39338
作者: Qianfeng Yuan,Wenbing Tao
类目: Machine Learning (cs.LG)
*备注:
Abstract:Knowledge distillation transfers knowledge by encouraging a student to match a teacher’s predicted class probabilities. These probabilities express not only confidence in the correct class, but also relations among incorrect alternatives. Yet closer imitation does not necessarily yield a better student. A student may already distinguish the correct class more sharply than its teacher, so further imitation can require giving back discrimination it has acquired. Our main result is an exact separation between full imitation and conditional learning. When the correct class’s score advantage over each alternative must be preserved, full teacher-to-student KL minimization is blocked exactly when the student assigns no more probability than the teacher to every incorrect class. Crucially, the teacher’s relative probabilities among incorrect classes remain fully learnable. We characterize the exact price of this transfer: a minimum increase in correct-class log-odds that compensates for the largest conditional-probability mismatch. Label fitting and conditional matching can therefore be completed even as full teacher KL diverges. This separation motivates task-preserving knowledge distillation (TPKD), which keeps the label gradient intact and minimally corrects the conditional gradient so that its output update preserves the label step’s gains against every incorrect alternative. The corrected conditional direction retains more than half of the original first-order conditional descent at the same step size, with a tight bound. For a fixed positive conditional target and sufficiently small constant output steps, label and conditional errors vanish together. Experiments trace this learning from exact head updates to ordinary network training. TPKD reaches 88.05% accuracy on CIFAR-100 and 93.81% on CLINC150, improving over standard distillation by 0.47 and 0.35 percentage points across three seeds.
[LG-91] How Many Samples Are Enough for Learning Across Domains?
链接: https://arxiv.org/abs/2609.39336
作者: Hong Zheng
类目: Machine Learning (cs.LG)
*备注:
Abstract:Understanding the fundamental mechanisms of learning is essential for designing systems with strong generalization. Recent studies have shown that increasing the number of training domains, or enlarging the distribution shift among them, improves generalization when each domain contains sufficiently many data samples. However, the conditions under which the data samples can be considered sufficient remain unexplored. In this work, we fill this gap by establishing criteria for per-domain sample requirements based on the presented learning bounds. These criteria not only reveal an inverse linear scaling law between the number of training domains and the number of samples required per domain, but also explain the fundamental rationale behind the assumption of data sufficiency, thereby providing theoretical guidance for assessing the adequacy of existing datasets and constructing datasets. This differs from classical learning theory, as the number of samples required is highly dependent on the number of training domains. Additionally, we prove the close relationship between in-domain learning and out-of-domain generalization through the presented generalization bounds, and lastly discuss some key arguments.
[LG-92] PatchKV: Weight-Space Compensation of KV Cache ATC NEURIPS2026
链接: https://arxiv.org/abs/2609.39329
作者: Chanryeol Lee,Chanhyuk Lee,Yeonwoo Choi,Donggyun Kim,Seunghoon Hong
类目: Machine Learning (cs.LG)
*备注: NeurIPS 2026. Code available at: this https URL
Abstract:Long-context inference with Large Language Models (LLMs) is bottlenecked by the linearly growing memory of the key-value (KV) cache. Existing compression methods reduce the cache through token eviction or approximation, but degrade sharply at aggressive compression budgets. We propose PatchKV, a training-free framework that compensates KV cache compression methods by carrying part of the context in the model’s weights. PatchKV pairs an off-the-shelf compressed KV cache with a context-specific weight patch, which is computed once at context-loading time and served for downstream queries for the context. The weight patch is derived in closed form via ridge regression, by aligning the block-wise activations of context-derived reference query tokens under the full cache and the compressed cache. Once merged into the model, the patch leaves the forward graph and per-query inference cost unchanged in the single-context, multi-query setting. Across long-context QA (SCBench with up to 170K tokens, SQuAD, NIAH) and math (GSM8K) benchmarks on three model architectures, PatchKV consistently improves cache compression methods, suggesting an alternative direction to compensate them at aggressive budgets.
[LG-93] Semantic-Aware Joint Source-Channel Optimization for Encoder-Agnostic Digital Video Communication
链接: https://arxiv.org/abs/2609.39296
作者: Xiangben Zhu,Caili Guo,Yang Yang,Chuanhong Liu,Meiyi Zhu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Video semantic communication has attracted increasing attention as a promising approach to improving video transmission efficiency. However, most existing approaches rely on computationally intensive deep learning-based video encoders and decoders, which hinders their deployment in resource-constrained scenarios. To address this issue, we propose a lightweight semantic-aware joint source-channel optimization (SAJSCO) scheme that can be integrated into existing digital video communication systems as a plug-in module. Specifically, we develop a video communication system model in which the transmitter jointly optimizes source and channel coding parameters based on the inter-frame semantic importance of the input video and estimated channel state information. On this basis, we formulate an optimization problem that maximizes semantic importance weighted video reconstruction quality under a maximum bitrate constraint. To solve it, we first quantify inter-frame semantic importance using a cosine similarity-based metric with a shifted window mechanism. We then develop a multi-actor proximal policy optimization (MPPO) algorithm to solve the formulated problem by jointly adapting the source compression rate and channel coding rate. The learned policy can be directly applied to different video encoders without encoder-specific retraining or fine-tuning. SAJSCO achieves Bjøntegaard Delta rate reductions of 34.86% and 18.01% when integrated with H.265, a conventional video encoder, and DCVC-RT, a deep learning-based video encoder. Over-the-air experiments on a hardware testbed further demonstrate a PSNR gain of up to 1.448 dB with H.265 and an LPIPS reduction of up to 0.033 with DCVC-RT compared with the respective best-performing fixed-parameter baselines.
[LG-94] Physics-Informed Method of Group Data Handling: Adaptive Construction of Functional Representations with an Application to the Navier-Stokes Equations
链接: https://arxiv.org/abs/2609.39291
作者: Mykhailo Minin
类目: Machine Learning (cs.LG); Fluid Dynamics (physics.flu-dyn)
*备注: 15 pages, 3 figures, 1 table; source code available at this https URL
Abstract:Physics-informed computational methods usually optimize parameters within a functional representation whose structure is fixed in advance. This work proposes a Physics-Informed Method of Group Data Handling (PI-GMDH), in which representations of coupled physical fields are progressively constructed during solution. Candidate functional directions are evaluated through the first variation of the complete physical and observational objective, introduced in packages, and followed by block-coordinate damped Gauss-Newton coefficient optimization. The framework is demonstrated with tensor-product Chebyshev functions on the incompressible Navier-Stokes equations using a two-dimensional time-dependent Taylor-Green benchmark. Under the tested configuration, adaptive PI-GMDH reached validation and held-out test losses of 5.299e-19 and 5.296e-19 with 204, 201, and 175 active functions for u, v, and p. Complete degree-by-degree and all-terms PI-GMDH variants, together with selected PINN and KAN reference configurations, are used to examine the effect of structural construction policy. The results show that, for this controlled synthetic benchmark, selective progressive construction can provide a favorable combination of accuracy, representation size, and wall-clock time. The comparison is illustrative rather than a claim of universal superiority over alternative physics-informed approaches.
[LG-95] ReTaCo: Residual-Target Control for On-Policy Distillation
链接: https://arxiv.org/abs/2609.39275
作者: Zixiang Ni,Zhuo Hu,Renjie Cao,Weijie Ren,Binqin Shi,Weijia Zhang,Shuheng Cao,Zhicheng Shi,Zhenhao Zhang,Haomin Wen,Zhiyuan Hu
类目: Machine Learning (cs.LG)
*备注:
Abstract:On-policy distillation (OPD) trains a student on its own generated prefixes with token-level teacher feedback, but transmitting or storing the teacher’s full-vocabulary distribution at every token is costly. Entropy-aware OPD (EOPD) adds forward supervision to reverse KL to help the student recover plausible tokens it underestimates, using only the teacher’s top- k probabilities to limit cost. Because EOPD renormalizes these probabilities, its target assigns no mass to the omitted vocabulary. We prove that the resulting loss keeps pushing the student’s top- k mass toward one even after the student matches the teacher’s relative probabilities within the top- k set, so the teacher itself is not a stationary point whenever the omitted tokens have positive teacher probability. We propose ReTaCo (Residual-Target Control), which keeps the top- k tokens individually and groups the remaining tokens into one residual symbol, and pairs this forward target with a single-sample estimator whose expectation equals the full-vocabulary reverse KL. With teacher top- k mass m , the residual target is (1-\beta)(1-m) for \beta\in[0,1] : \beta=0 preserves the teacher’s mass, and larger \beta moves more mass onto the top- k tokens without changing their relative probabilities. At a fixed prefix, we prove that the population objective has a unique optimum whose top- k mass lies between m and m+\beta(1-m) and increases monotonically with \beta ; at \beta=0 , underestimated top- k tokens still receive non-vanishing recovery gradients. Numerical optimization confirms these predictions, and across three teacher-student pairs, ReTaCo outperforms EOPD on most mathematics and code benchmarks.
[LG-96] RW-Flow: One-Step Generation on Compact Manifolds via Wasserstein Gradient Flows ICLR
链接: https://arxiv.org/abs/2609.39271
作者: Ualibyek Nurgulan,Seungwoo Yoo,Prin Phunyaphibarn,Minhyuk Sung
类目: Machine Learning (cs.LG)
*备注: 27 pages, ICLR, 2 figures, 11 tables
Abstract:Manifold-valued data, and consequently the distributions they induce, are prevalent across many domains, ranging from the locations of geospatial events, such as earthquakes, to biomolecular torsion angles that encode information about three-dimensional structure. While diffusion and flow-based generative models have been successfully extended to compact manifolds, sampling typically requires tens or hundreds of sequential network evaluations. We introduce RW-Flow, a theoretically grounded framework for learning one-step generative models on compact manifolds via Wasserstein gradient flows. The main challenge is identifiability: driving the velocity field to zero should guarantee that the model distribution matches the target distribution. We establish a necessary and sufficient condition for identifiability on compact, connected Riemannian manifolds. We specifically show that, for a symmetric, Lipschitz-continuous cost function, the velocity field induced by the Sinkhorn divergence is identifiable if and only if the associated Gibbs kernel is nondegenerate. This characterization provides a general principle for designing identifiable costs on compact manifolds. It also reveals that the squared geodesic distance, the natural manifold analogue of the squared Euclidean distance, does not always guarantee identifiability. Across benchmarks involving geospatial events, protein side chain torsion angles, RNA backbone torsion angles, and general manifolds discretized as triangular meshes, RW-Flow outperforms existing one-step methods in nearly all settings under fair comparison conditions.
[LG-97] Jacobian Rank Collapse in Decision-Focused Learning
链接: https://arxiv.org/abs/2609.39261
作者: Aojie Yuan,Haiyue Zhang,Zijian Su
类目: Machine Learning (cs.LG); Portfolio Management (q-fin.PM)
*备注: 53 pages, 21 figures, 32 tables. Includes theoretical proofs and supplementary experiments
Abstract:Decision-focused learning (DFL) trains predictors through downstream objectives, but a different loss need not provide an independent parameter-update direction. We characterize this restriction through the predictor Jacobian, using sparse index tracking to distinguish the covariance entries read by the optimizer from the parameter directions available to learning. Rank-one Jacobians make nonzero per-example gradients collinear; a conditional spectral bound describes near-collinearity. A batch-subspace characterization and counterexamples show why these local statements imply neither common minimizers nor collinear batch updates. Experiments examine when geometry translates into decision quality. Across 38 one-parameter equity configurations, DFL gains over MSE remain below 1.8%; a 385-parameter conditional predictor also has pointwise rank one. In validation-tuned shortest-path and knapsack experiments, full-capacity SPO+ reduces mean regret by 11.6% and 10.6%, respectively; only knapsack survives correction across eight comparisons. The capacity contrast persists on fresh datasets across batch orders and training budgets. Holding expressivity fixed, invertible coordinate scaling lowers spectral effective rank and ordinary SGD gains; compensating for the scaling restores the original trajectories. Financial forward-target controls separate forecast accuracy from decision quality; a matched neural comparison finds no aggregate DFL advantage in the tested architecture. These findings distinguish local rank restrictions, coordinate-dependent optimization and predictive accuracy. Predictor geometry helps explain available learning directions, while held-out decision quality remains the test of practical benefit. Comments: 53 pages, 21 figures, 32 tables. Includes theoretical proofs and supplementary experiments Subjects: Machine Learning (cs.LG); Portfolio Management (q-fin.PM) Cite as: arXiv:2609.39261 [cs.LG] (or arXiv:2609.39261v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.39261 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-98] Client and Training Data Selection for Computationally Efficient Synchronized Federated Learning MICRO
链接: https://arxiv.org/abs/2609.39250
作者: Muzaffer Citir,Hiroki Nishikawa,Sangyoung Park
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注: Accepted for publication in the 29th Euromicro Conference on Digital System Design (DSD 2026)
Abstract:Federated learning (FL) is a promising paradigm of machine learning, which preserves user privacy by enabling learning without sharing raw data with a cloud server. Straggling clients have been a problem for FL as they introduce delays in aggregating the local models and hence, the convergence of the global model. Therefore, it is important to have a mechanism that ensures fast convergence of the global model as well as good FL participation rate. Another issue for the convergence of a model in FL is the non-independent and identically distributed (non-iid) data across the clients. Prior approaches based on probabilistic client selection do not work well under non-iid data especially when the number of clients is small. We show scenarios where such approaches fail and propose a joint client-training data selection algorithm for fast convergence of FL models. Our experiments on CIFAR-100 dataset show that convergence of the FL model can be significantly improved over prior works that can consider non-iid data and heterogeneous computation and higher model accuracy.
[LG-99] Right Answer Wrong Mechanism: Detecting Pernicious Divergence in Causal Interventions
链接: https://arxiv.org/abs/2609.39243
作者: Beiming Liu,Minjie Chen
类目: Machine Learning (cs.LG)
*备注: 13 pages, 1 figure, 6 tables. Beiming Liu and Minjie Chen contributed equally
Abstract:Causal interventions such as activation patching and distributed alignment search (DAS) are the main tool for making mechanistic claims about neural networks. Recent work showed that these interventions routinely push representations off the model’s natural distribution, and that such divergence is sometimes harmless and sometimes pernicious: it can recruit pathways the model never uses on natural inputs, so that an intervention produces the expected answer through the wrong mechanism. No method currently tells the two cases apart. We make this question testable by planting hidden pathways inside pretrained language models; the pathways are silent on every benchmark prompt by construction, so which interventions depend on them is known exactly. Across 72 configurations and 100,800 interventions on GPT-2 small, we find three things. (i) Nearest-neighbour and local-PCA distances at the intervention site, as used in prior work, score below chance (AUROC 0.35-0.47) at picking out interventions that give the right answer through a planted pathway. (ii) Hidden-Pathway Contribution (HPC), a label-free test that clamps downstream units to the regime of natural runs with the same output and measures how much of the decision disappears, flags pathway-dominated interventions with AUROC = 0.99 when the pathway shows up as unit-level out-of-regime activity, but fails when every unit stays within its natural range, which we identify as the open problem. (iii) Optimised interventions actively seek hidden pathways: on a gender task, DAS routes 90-95% of its successes through planted pathways for three of four families, and a downstream on-manifold penalty cuts this share to under 5% at a cost of 6-11 points of success rate. In unmodified GPT-2, successful interventions show almost no unit-level out-of-regime reliance.
[LG-100] A differentiability framework for zigzag persistent homology via linear interpolation
链接: https://arxiv.org/abs/2609.39242
作者: Enrico Maria Ferrari,Clemens Bannwart,Matteo Biagetti
类目: Machine Learning (cs.LG); Computational Geometry (cs.CG); Algebraic Topology (math.AT)
*备注: 9+31 pages, 12 figures, comments welcome!
Abstract:Persistent homology can be differentiated and incorporated into learning pipelines, but no analogous framework exists for zigzag persistence, which is needed when the underlying topological structure evolves non-monotonically over time. We develop such a framework for sequences of simplicial complexes obtained by thresholding time-dependent filtering values on a fixed complex. By assigning persistence diagram endpoints the real-valued times at which linearly interpolated filtering values cross the threshold, we transfer the continuity of the filtering values to the diagram points. This yields smooth local lifts of the resulting persistence-diagram-valued map, from which we derive differentials almost everywhere under mild regularity conditions on the parametrization of the filtering values. We prove local Lipschitz continuity outside an explicit measure-zero exclusion set; standard stochastic subgradient convergence guarantees therefore do not apply directly. We argue that, even without such guarantees, this exclusion set is small enough in practice to allow effective optimization. We test this empirically in two experiments: sensor network coverage optimization and dynamic graph classification.
[LG-101] QATFactory: A Versatile Deployment-Aligned Framework for Quantization-aware Training and Distillation of LLM s
链接: https://arxiv.org/abs/2609.39223
作者: Weili Xu,Jisen Li,Yuqing Jian,Chenxi Li,Zhizhou Sha,Yifan Yu,Qingyang Wu,Chenfeng Xu,Zhongzhu Zhou,Tianyi Zhang,Ben Athiwaratkun
类目: Machine Learning (cs.LG)
*备注: Preprint, code at this https URL
Abstract:Large language model (LLM) inference is increasingly moving toward lower precision to realize the throughput of hardware accelerators, but aggressive post-training quantization (PTQ) can degrade model quality. We present QATFactory, an open-source framework for deployment-aligned quantization-aware distillation (QAD) and reinforcement learning (QARL). QATFactory simulates deployment-time quantization while performing matrix multiplications in BF16, allowing models to adapt to quantization noise without requiring training hardware that natively supports the target format; for example, it supports NVFP4 training on H100 GPUs, which lack FP4 Tensor Cores. The framework supports NVFP4, MXFP4, and this http URL’s Q4_K format; dense and mixture-of-experts models; and both full-parameter and LoRA-based training. It exports checkpoints directly to vLLM and this http URL without an additional lossy conversion step or added inference overhead. With QATFactory, we conduct extensive experiments on models ranging from 8B to 230B parameters and evaluate exported checkpoints in production inference engines. Across models and formats, QAD consistently improves deployed-model quality over strong PTQ baselines. On Qwen3.5-9B, QAD achieves average benchmark accuracies of 68.9% under NVFP4 and 66.0% under MXFP4, outperforming the best PTQ results of 65.4% and 56.4%, respectively. Through our experiments, we found that although both FP4 formats quantize weights and activations at deployment, the best training strategy is format-dependent: NVFP4 generally performs better when only weights are quantized during training, whereas MXFP4 benefits from quantizing both weights and activations. At a fixed training token budget, training on fewer 32K sequences improves average accuracy by 1.9 points over training on more 4K sequences. We release the complete QATFactory training code and the resulting checkpoints.
[LG-102] Importance-Aware Feature Sparsification for Wireless Split Learning
链接: https://arxiv.org/abs/2609.39194
作者: Bumjun Kim,Yoon Huh,Wan Choi
类目: Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: Accepted for publication in IEEE Journal on Selected Areas in Communications
Abstract:Wireless split learning (SL) reduces on-device computation by offloading upper layers to a server, yet transmitting high-dimensional intermediate features at each iteration remains a major communication bottleneck. Existing methods select features at the client side using task-agnostic criteria such as magnitude, statistics, or clustering, which increases client-side processing and often degrades accuracy under non-independent and identically distributed (non-i.i.d.) client data. We propose importance-aware class-balanced sparsification (ICS), a lightweight approach in which the server ranks feature channels using Grad-CAM-based scores obtained from the true-class logit during backpropagation. The per-class scores are aggregated into a class-balanced, label-agnostic importance vector that mitigates head-class bias under label skew, and each client reuses this vector in the next round to retain the top- N feature channels, incurring no additional client-side forward or backward passes. We further derive a non-asymptotic convergence bound that isolates the sparsification-induced error and characterizes how the sparsification ratio and mini-batch size jointly affect convergence under a fixed communication budget, and we analyze the communication and computational overhead of ICS against representative baselines. Beyond sequential CNN-based SL, we extend ICS to parallel split learning and to transformer-based models. Experiments show that ICS consistently outperforms the baselines, with larger gains under severe non-i.i.d. partitions.
[LG-103] Dynamics to decision: A mathematical theory of Lyapunov spectra and decision boundaries in deep classifiers
链接: https://arxiv.org/abs/2609.39190
作者: Shirin Panahi,Amirhossein Nazerian,Ali Pezeshki
类目: Machine Learning (cs.LG); Systems and Control (eess.SY); Dynamical Systems (math.DS)
*备注:
Abstract:A deep classifier is defined not only by the decision it produces, but also by the sequence of transformations through which that decision is formed. Treating this evolution as a dynamical system across layers provides a natural framework for asking how decision geometry emerges through depth and how far back we can trace a boundary’s dynamical signature. We model a feed-forward classifier as a finite, nonautonomous discrete dynamical system, with layers playing the role of discrete time steps. We study the Finite-Time Maximum Lyapunov Exponent (FTMLE) of the data samples’ dynamical trajectory through depths of the classifier. The FTMLE measures the rate of convergence/divergence of nearby trajectories. We move the observation endpoint backward from probabilities to logits and then to hidden representations. For Gaussian classes, we prove that probability-level FTMLE carries a clear geometric signature of the decision boundary, with its dominant direction aligned with the boundary normal. Moving one step backward to the logits, we prove this relationship is no longer universal but depends critically on how the classifier is trained, particularly on the choice of loss function. Moving further backward to the hidden representation, the connection becomes more conditional: boundary-related FTMLE can persist, but only under identifiable structural conditions. We propose geometry-aware fine-tuning for restructuring the classifier’s hidden FTMLE, and propose conditions for guaranteed concentration of high hidden FTMLE near the decision boundary. Through our numerical results, we show the generality and validity of our theoretical results. Understanding the evolution of data samples as traveling through the layers of classifier provides a principled foundation for identifying where boundary-relevant sensitivity emerges and for developing layer-aware regularization strategies.
[LG-104] Attention Function as an Intrinsic Inductive Bias: How Models Behavior Diverges in Novel Contexts NEURIPS2026
链接: https://arxiv.org/abs/2609.39188
作者: Dong Gyun Kang,Megha Thukral,Kwangsoo Kim
类目: Machine Learning (cs.LG)
*备注: Accepted as a full paper (poster) at the NeurIPS 2026 Workshop on Developmental Perspectives on AI (DevAI 2026)
Abstract:Developmental psychology holds that certain priors are given to infants prior to experience rather than induced from data, and that the influence of such priors is suppressed under strong, well-constrained conditions but reasserts itself under weak ones. We ask whether an analogous principle holds for the Transformer: can the activation function given to attention heads serve as an intrinsic inductive bias? We propose Mixture of Function Attention (MoFA), a parameter-free modification to multi-head attention that fixes a ratio of softmax and sigmoid heads before training. Across five ratios, a 124M-parameter GPT-2 model, and five seeds, we find that this given ratio has little effect in-distribution – differences between ratios are statistically negligible for moderate mixtures and remain small even at the extremes – but its influence re-emerges sharply under zero-shot distribution shift across 15 out-of-distribution domains. Perplexity gaps between ratios widen by more than an order of magnitude on several domains, and the best-performing ratio tracks a single axis of domain structure, separating short, informal text (softmax-favoring) from technical, long-form text (sigmoid-favoring), that explains 78.3% of the variance in domain response. This reorganization is visible at the head level: sigmoid heads show an accelerating drop in attention entropy as their ratio increases, while softmax heads respond more modestly, yielding a consistent division of labor between the two head types. Our results suggest that activation choice functions as a given prior whose influence is masked in-distribution and re-emerges out-of-distribution.
[LG-105] Low-Discrepancy Dither for Quantized Recurrent State Caches
链接: https://arxiv.org/abs/2609.39185
作者: Snigdha Chandan Khilar
类目: Machine Learning (cs.LG)
*备注:
Abstract:Mamba-style and hybrid language models compress their past into a fixed-size recurrent state that is rewritten at every generated token. Storing this state in low precision saves memory bandwidth, but every rounding error is fed back into the next update and can accumulate over long generations. Production systems round the state stochastically; we ask which rounding rule such caches should use. We find that a deterministic golden-ratio Weyl dither, which needs no random numbers, consistently brings the quantized model closer to the full-precision one than stochastic rounding, across pure and hybrid models, storage formats, and long decoding horizons, at no extra cost. Round-to-nearest behaves differently: because it discards small updates, its error keeps growing, so it can look best in short evaluations yet falls far behind over long generations. A discrepancy analysis explains this ordering, and we document implementation pitfalls that silently remove the benefit.
[LG-106] Whitening Improves Robustness to Spurious Correlations in Linear Probes
链接: https://arxiv.org/abs/2609.39177
作者: Floris Holstege,Bram Wouters,Noud van Giersbergen,Cees Diks
类目: Machine Learning (cs.LG)
*备注: Preprint
Abstract:Deep neural networks tend to rely on simple features that may be spurious and thus fail to generalize. We study this problem in the setting of linear probes, where a (generalized) linear model is fitted on the representations of a (pretrained) model. We use the connection of these models to the max-margin classifier, and show they favor directions associated with large eigenvalues of the covariance matrix. Whitening removes this preference by equalizing the eigenvalues of the covariance matrix. This observation motivates whitening as a preprocessing step that can reduce reliance on spurious correlations without requiring prior knowledge of their presence or labeled data. We examine the effect of whitening on a synthetic data-generating process and standard spurious correlation benchmarks, and find that it improves robustness. We also find that whitening can improve robustness when added to existing approaches.
[LG-107] QuanVI: Score-based Variational Inference via Quantum Maximally Mixed States NEURIPS2026
链接: https://arxiv.org/abs/2609.39164
作者: Yuchen Cong,Zerui Tao,Chao Li,Zhe Sun,Qibin Zhao
类目: Machine Learning (cs.LG)
*备注: Accepted to NeurIPS 2026
Abstract:Score-based variational inference (VI) provides an alternative to Kullback–Leibler (KL)-based VI by minimizing the Fisher divergence between the variational distribution and the target. A prior score-VI approach formulates this optimization as an eigenvalue problem, with the variational distribution constructed from low-energy eigenstates. However, this eigenvalue-based formulation faces two high-dimensional obstacles: an intractably large parameter count due to exponential scaling and non-uniqueness of individual eigenvectors in degenerate or nearly degenerate low-energy subspaces. We propose QuanVI, a scalable quantum-inspired algorithm that combines a mixed-state density-operator formulation with a quantum tensor network (QTN) parameterization using the matrix product operator (MPO) structure. In degenerate low-energy subspaces, the density-operator formulation represents the subspace by its maximally mixed state rather than relying on a non-unique individual eigenvector, while the QTN parameterization compresses the density operator to avoid exponential parameter growth. Experiments and ablations show that QuanVI agrees with exact solutions in low dimensions and scales to high-dimensional synthetic and Bayesian posterior-approximation benchmarks, including challenging non-Gaussian targets.
[LG-108] raining-Free Affinity Fusion of Neural and Embedding-Based Speaker Diarization ICASSP2027
链接: https://arxiv.org/abs/2609.39162
作者: Yehoshua Dissen,Joseph Keshet,Eduard Golshtein
类目: ound (cs.SD); Machine Learning (cs.LG)
*备注: submitted to ICASSP 2027
Abstract:Speaker diarization systems based on speaker embeddings and neural diarization exploit complementary forms of speaker information, but their intermediate representations are not directly compatible. We introduce Training-Free Affinity Fusion (TFAF), which integrates the speaker structure inferred by a neural diarizer into an embedding-based diarization system. The neural speaker partition is used to condition local speaker representations, from which we construct a continuous affinity matrix and combine it with the embedding-based acoustic affinity before a single global clustering step. The method requires no additional training, shared embedding space, speaker-label alignment, or hard transfer of the neural diarizer’s speaker count. Experiments on AMI and CALLHOME show consistent DER improvements over both constituent systems; on AMI, fusion also improves speaker-attributed transcription. Ablations show that the neural speaker partition accounts for most of the gain, while retaining the continuous embedding-based affinities provides additional benefit over hard partition fusion.
[LG-109] Linear Recurrent Memory Suffices to Distil a World-Model Policy for Robot Air Hockey
链接: https://arxiv.org/abs/2609.39151
作者: F. Olivia Fan,Oliver Obst
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: 10 pages, 4 figures
Abstract:Does memory-dependent control need nonlinear recurrent dynamics? We study simulated air-hockey defence under temporary loss of puck tracking. A DreamerV3 teacher outperforms a memoryless policy under tracking loss, while resetting the teacher’s recurrent state sharply reduces performance, which demonstrates that the task requires memory. We distil this teacher into compact recurrent policies with a 64 dimensional state, with a combination of a diagonal linear recurrence and an optional rank- k nonlinear innovation while retaining nonlinear observation encoders and action heads. Across five matched seeds, the purely linear recurrent model ( k=0 ) matches both the GRU baseline and the teacher throughout the tested range of tracking loss. Increasing nonlinear innovation rank providing no measured benefits. This result is obtained on a fresh test split, which will be only opened after all models and analyses are frozen. The linear model requires fewer recurrent parameters and less computation than GRU, but performs comparably. These results suggest that, for this memory dependent control task, nonlinear representation learning around a simple linear memory mechanism can be sufficient, and that nonlinear recurrent dynamics are not necessarily required. These conclusions are limited to the simulated task, teacher, state dimension, and blackout horizon considered here, and to policies whose observation encoder and action head remain nonlinear.
[LG-110] Sharp Stationary Gaussian Approximation for Constant-Stepsize SGD NEURIPS
链接: https://arxiv.org/abs/2609.39144
作者: Junghoon Seo
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: To be presented at 2026 NeurIPS workshop on “Optimization for Machine Learning” (OPT2026)
Abstract:We prove a sharp Gaussian approximation for the invariant law of constant-stepsize SGD with bounded additive noise generated by an exogenous uniformly ergodic Markov chain. For a smooth, strongly convex objective with a Lipschitz Hessian and nondegenerate long-run noise covariance, the centered iterate normalized by the square root of the stepsize is O(\sqrt\alpha) -close in 1-Wasserstein distance to its limiting Gaussian. The proof combines blockwise Gaussian comparison with long-run contraction. A four-state example gives a matching lower bound although the one-time noise marginal is symmetric and every nonzero-lag autocovariance vanishes. In this example, an adjacent third-order mixed moment produces the leading correction.
[LG-111] ID Balancing: Stable Training of Extremely Sparse MoE via PID-Based Load Control
链接: https://arxiv.org/abs/2609.39137
作者: Peng Jin,Zihan Qiu,Zekun Wang,Bo Zheng,Yang Xu,Tian Xie,Xiao Li,Huaqing Zhang,Haoran Lian,Rui Men,Dayiheng Liu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Scaling Large Language Models (LLMs) via Mixture-of-Experts (MoE) enables massive parameter growth with nearly constant per-token computation. However, further scaling the parameter count requires increasingly sparse routing, where expert load imbalance becomes more severe. This imbalance reduces parameter utilization and training efficiency, and can undermine training stability, becoming a bottleneck to reliable scaling. In this work, we unify two representative auxiliary-loss-free methods as incomplete Proportional-Integral-Derivative (PID) controllers: DeepSeek’s loss-free method acts as a fixed-step integral controller, while Kimi K3’s Quantile Balancing functions as a generalized proportional controller. Building on this control perspective, we propose ID Balancing, an Integral-Derivative controller. It scales its integral term with load error and activates its derivative term only when imbalance worsens, enabling stronger corrections for large or worsening errors and smaller updates near balance. Evaluated across Top- 10 , Top- 5 , and Top- 3 routing over 768 experts, ID Balancing reduces worst-case backbone MaxVio and training-average backbone MinVio by over 50% and 12% , respectively, relative to the best baselines in the Top- 3 setting. When the total parameter count increases from 18.9 B to 69.9 B (Top- 10 -of- 768 ), ID Balancing’s worst-case backbone MaxVio remains nearly unchanged and is approximately 89.6% lower than that of the auxiliary-loss baseline. ID Balancing also maintains competitive language-modeling and downstream performance. The advantages of ID Balancing grow as sparsity increases, making it a promising solution for scaling larger, sparser MoE models.
[LG-112] Characterizing High Bandwidth Flash for LLM Serving
链接: https://arxiv.org/abs/2609.39131
作者: Zack Yu,Chloe Wong,Coleman Hooper,Minjae Lee,Wonjun Kang,Youngjin Cho,Michael W. Mahoney,Yakun Sophia Shao,Kurt Keutzer,Amir Gholami
类目: Machine Learning (cs.LG); Hardware Architecture (cs.AR); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注:
Abstract:Large language model (LLM) serving requires substantial memory to store model weights and KV caches. As models grow larger and contexts become longer, memory capacity and bandwidth increasingly become bottlenecks for serving performance. Agentic workloads compound this pressure through repeated interactions over growing contexts, making it increasingly important to retain KV state for reuse. High-bandwidth flash (HBF) offers a way to expand accelerator memory capacity for large language model (LLM) serving, but its access costs and limited write endurance complicate its use. We evaluate HBF for high-throughput agentic serving across system design and scheduling choices to understand when additional capacity improves serving performance and energy efficiency. We introduce an HBM-HBF-host hierarchical storage system and buffered cache-aware scheduling, and use trace-driven simulations to analyze their effects on performance, energy consumption, and HBF write lifetime. Across the evaluated workloads, the fastest HBF-augmented systems reduce completion time by 36.1-87.0% relative to HBM-only systems. Modeled energy savings reach 55.8%, although HBF increases energy consumption on some light workloads. Buffered cache-aware scheduling extends estimated HBF write lifetime from 4.77 to 14.82 years in the evaluated configuration. These results demonstrate the importance of coordinating data placement and scheduling to improve serving efficiency while sustaining a practical HBF write lifetime.
[LG-113] CDMD: A Cross-Dataset Mixed-Type Diffusion Model for Tabular Data NEURIPS2026
链接: https://arxiv.org/abs/2609.39124
作者: Mohamed Amine Ketata,Maximilian Schambach,Stephan Günnemann
类目: Machine Learning (cs.LG)
*备注: Accepted at BeNTo workshop (NeurIPS 2026)
Abstract:Generative models for tabular data are typically trained separately for each dataset, limiting knowledge transfer and requiring the storage of many specialized models. In this paper, we introduce CDMD, a tabular diffusion model trained jointly across heterogeneous datasets with different schemas and variable numbers of numerical and categorical features. Unlike existing cross-dataset tabular diffusion models that operate in continuous representation spaces, CDMD defines diffusion directly over the mixed-type feature space and is trained end-to-end. To accommodate heterogeneous categorical domains, we introduce a schema-restricted reverse-process parameterization for masked diffusion models, in which the output space dynamically adapts to each feature’s vocabulary. We then compose numerical and categorical feature-level diffusion processes into a schema-dependent row-level process. A shared schema-aware Transformer denoiser captures dependencies between features and parameterizes the reverse process across varying schemas. On seven real-world datasets, a single jointly trained CDMD achieves the highest average generation quality among strong single-dataset and cross-dataset baselines, while using substantially fewer total parameters than the collection of separately trained models. Furthermore, pre-training on a corpus of 337 datasets improves generation on previously unseen datasets under both limited target data and limited adaptation epochs. These results demonstrate the potential of direct mixed-type diffusion for shared and transferable tabular data generation. Our code is available at this https URL.
[LG-114] Prequential E-Values for Selected-GP Near-Optimality Certificates NEURIPS2026
链接: https://arxiv.org/abs/2609.39123
作者: Ami Tavory,Noa Cohen
类目: Machine Learning (cs.LG)
*备注: 9 pages, 2 figures. Accepted at the NeurIPS 2026 Workshop on E-Values: From Statistics to ML
Abstract:When optimizing an expensive black-box function sequentially, as in hyperparameter optimization, we may want to stop once the best evaluated value is certified within \varepsilon of the global optimum. Such a certificate needs two ingredients: a lower confidence bound for the selected value and an upper confidence envelope over the domain, typically supplied by a Gaussian process (GP). GP-UCB-style stopping rules are valid when the kernel and constants defining this envelope are fixed before the run, but the practical temptation is to tune the envelope from the same adaptive evaluations and then certify as if it had been fixed. We use prequential e-values to make this selection auditable: starting from a predeclared set of fully specified GP/RKHS envelopes, each candidate is tested by its own one-step-ahead e-process, contradicted candidates are deleted, and certification uses the largest upper bound among the survivors. With a valid selected-point lower bound and one declared candidate having valid latent coverage and noise calibration, the rule is anytime-valid. On a 512-seed noisy RBF stress sweep, it roughly halves false-certification risk at comparable power versus fit-then-certify. Relative to random fixed GP precommitment on smooth d=3,4 objectives, each additional false certificate is accompanied by 3.0 and 13.5 additional correct certificates, respectively.
[LG-115] he Row Normalization Puzzle in Muon
链接: https://arxiv.org/abs/2609.39114
作者: Jiayu Zhang,Tianyi Lin
类目: Machine Learning (cs.LG)
*备注: 32 pages, 6 figures
Abstract:This paper examines how row-wise renormalization affects Muon, focusing on the gap between NorMuon’s worst-case guarantees and its practical performance (Li et al.). Despite its growing adoption and promising performance in large language model (LLM) pretraining, NorMuon’s worst-case guarantees remain poorly understood. One fundamental question is: Does row normalization yield provable convergence gains, potentially through its interaction with approximate polar computation and exponential moving-average momentum? Our results show that row normalization introduces a dimension-dependent factor in the worst-case iteration complexity under the operator-norm geometry, which persists even with exact polar computation and any fixed momentum parameters. Indeed, we establish an algorithm-dependent lower bound and a matching upper bound in deterministic settings, and extend our upper bound analysis to stochastic settings. Both upper-bound analyses allow approximate polar computation. Experiments show that NorMuon is slower than Muon on synthetic problems inspired by our worst-case construction, yet outperforms Muon in LLM pretraining. These findings sharpen the puzzle of why row normalization helps in practice and complement the recent findings of Dewulf et al.
[LG-116] Learning Infinite-Horizon Averag e-Reward CMDPs via State Augmentation
链接: https://arxiv.org/abs/2609.39093
作者: Kihyun Yu,Seoungbin Bae,Dabeen Lee
类目: Machine Learning (cs.LG)
*备注:
Abstract:We study infinite-horizon average-reward constrained Markov decision processes (CMDPs) under the weakly communicating assumption. Existing high-probability guarantees for this setting either require computationally inefficient algorithms or have suboptimal dependence on the number of interactions T . We propose, to the best of our knowledge, the first computationally efficient algorithm that achieves \widetilde\mathcalO(\sqrtT) regret and cumulative constraint violation with high probability in the tabular setting. The \sqrtT dependence is optimal up to logarithmic factors. Our approach incorporates cumulative constraint violation into the state and defines a reshaped reward through differences of a Huber potential. The added state determines the penalty on further violations while the reward function remains fixed on the augmented state space. Since the added state has known deterministic dynamics, only the original transition kernel needs to be estimated. The bounded slope of the Huber potential keeps the per-step reward bounded, and the potential differences telescope to relate the reshaped return to the original cumulative reward and the terminal potential. These properties allow us to apply finite-horizon approximation and optimistic value iteration with clipping, as used in unconstrained average-reward MDPs, without worsening the regret rate in T .
[LG-117] Shared Phase and Retention Control for Efficient Adaptive Spectral Recurrence
链接: https://arxiv.org/abs/2609.39082
作者: Wentao Wang,Hengyu Zhong,Yunhan Jiang,Jialiang An,Meng Lu
类目: Machine Learning (cs.LG)
*备注:
Abstract:As new evidence arrives, a sequence model must update what it remembers and how memory influences predictions. While Transformers incur computation and cache costs scaling with context length, fixed-state recurrent models offer constant-memory inference. However, linear and spectral recurrences traditionally rely on static transitions, failing to dynamically revise how stored representations decay or rotate. While recent selective architectures introduce input-dependent transitions, they assign independent controls to every memory mode, coupling control cost to state capacity. We show that high-dimensional spectral memory does not require high-dimensional control, and introduce Shared Phase and Retention Control for Efficient Adaptive Spectral Recurrence (SPARC). SPARC employs just two input-dependent scalar signals to coordinate memory retention and phase rotation across heterogeneous complex modes, while preserving mode-specific baseline timescales and frequencies. Its diagonal affine recurrence supports parallel associative scans for sequence-level BPTT as well as exact structured Real-Time Recurrent Learning (RTRL) for online credit assignment. Across partially observable continuous control, POPGym, and sequence classification, SPARC achieves a 9.09% relative return improvement on Walker-P and a 1.36% relative accuracy gain on FordA over second-best methods. On an NVIDIA Blackwell GPU, our implementation reduces recurrent-mixer training latency by 18.2%-34.2% in fixed-token workloads and accelerates scans by 3.1x-4.7x over an optimized RG-LRU baseline. These results show that two shared control signals can efficiently govern adaptive spectral memory across online and full-sequence settings. Code is available at this https URL.
[LG-118] Parameter symmetries determine representational geometry in overparameterized nonlinear networks NEURIPS2026
链接: https://arxiv.org/abs/2609.39078
作者: Marvin Theiss,Lukas Braun,Andrew M. Saxe,Erin Grant
类目: Machine Learning (cs.LG)
*备注: To appear in NeurIPS 2026. 89 pages, 9 figures. Code: this https URL
Abstract:Representations are routinely used across machine learning, psychology, and neuroscience to draw inferences about the computations of biological and artificial systems. Such inferences presume a meaningful link between representational geometry and the computation being performed. For artificial neural networks, however, the extent to which function constrains representation remains unclear. One key obstacle is that these networks admit parameter symmetries: changes in parameterization that preserve function exactly while reshaping representational geometry. Here, we show that a broad class of parameter symmetries acts on representations through just three primitive feature transformations: addition, duplication, and scaling. This feature-level characterization yields a closed-form decomposition of representational geometry into essential and auxiliary components, which makes precise how degeneracy in representational geometry can grow with overparameterization even when function is held fixed. Finally, we show that implementation-level selection rules can resolve this degeneracy, yielding identifiable geometries in which features are weighted according to their contributions to the network’s function. Together, our results delineate when representations can support inferences about computation, and when they cannot.
[LG-119] SparseEngine: Sparse-First Inference Engine
链接: https://arxiv.org/abs/2609.39068
作者: Jitai Hao,Quansheng Gu,Qiang Huang,Jun Yu
类目: Machine Learning (cs.LG)
*备注:
Abstract:Long-context LLM agents accumulate interaction histories that strain KV-cache memory and attention computation. Although sparse attention reduces these costs, heterogeneous cache representations and workflows hinder integration with existing inference engines, while prior sparse-serving abstractions support only specific layouts or workflows. We present SparseEngine, a ground-up, sparse-first inference engine whose shared lifecycle contract lets each method control its KV representation and computation while coordinating state transitions with common serving infrastructure. SparseEngine supports 15 methods across four categories and enables cross-request state management through Chain Cache, which resumes KV-eviction methods from retained history, and controllable Prefix-Cache Pruning, which removes KV from selected history regions while preserving logical-prefix matching. While maintaining method quality, SparseEngine delivers over 10x higher throughput with KV eviction, over 2.5x faster decoding at matched concurrency than vLLM, and over 2x end-to-end speedup on agent benchmarks. The code is available at this https URL.
[LG-120] Argus: A Real-EKS Study of When Predicting Spot Interruptions Beats Simple Checkpointing
链接: https://arxiv.org/abs/2609.39067
作者: Angshuman Chakravertty,MD Rayyan
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注: 8 pages, 5 figures, 3 tables
Abstract:Elastic Compute Cloud (EC2) Spot is 60% to 90% cheaper than On-Demand but can be reclaimed on just a 2-minute notice; for expensive multi-node training this loss can be severe, with one reclaim costing hours of synchronous progress. We build Argus, a Kubernetes operator, and ask empirically, on a CIFAR-10 testbed, when predicting interruptions beats simple checkpointing. Argus on real EKS survives a real Spot drain with a graceful SIGTERM checkpoint, resuming from epoch 8 and losing only the in-progress epoch. Alongside, we further find that in an 80-trial benchmark, the reactive-on-notice degrades toward no protection once interruption outpaces the fixed 2-minute notice, and predictive wasted compute is driven to zero, but with an oversized fixed lead it over-migrates so severely that at the fastest rate only one of five runs completes, while periodic is a strong ML-free baseline. A lead-time sweep turns the lead prediction into a guideline where a small lead suffices for zero waste, but excess lead is wasteful. The predictor built is advisory (a proxy label); real interruption labels and large-model-scale validation are future work.
[LG-121] he Missing Coefficients: Bayesian Pairwise Merging for Model Personalization
链接: https://arxiv.org/abs/2609.39055
作者: Yaling Shen,Tongtong Wu,Siyuan Yan,Gholamreza Haffari
类目: Machine Learning (cs.LG)
*备注:
Abstract:How can we personalize a shared expert library from a user’s pairwise choices? Prior work can realize different reward trade-offs by merging reward-specialized experts, given a vector of trade-off weights. In practice, users can more naturally choose between outputs than specify numerical weights. The challenge is therefore to turn these choices into the coefficients required for merging, while accounting for ambiguity when feedback is limited. Our key idea is to treat the unknown reward weights as latent variables: infer a posterior over them from pairwise choices and reward-score differences, and use its mean directly as the merge coefficients. We instantiate this idea as Bayesian Pairwise Merging (BPM), whose posterior also characterizes which reward trade-offs remain plausible given the feedback. We evaluate BPM on radiology summarization, image captioning, and story generation, spanning text-to-text and image-to-text generation. With 100 feedback per simulated persona, BPM achieves macro decided win rates of 91.7%, 77.1%, and 64.3% against uniform merge. For six pairs of simulated personas, each prefers the model fitted to its own feedback, a pattern also observed in a human proof-of-concept. In simulations under BPM’s model and prior, its nominal 90% intervals for temperature-scaled reward weights achieve task-averaged marginal coverage of 88.9% and 89.2% with only 10 and 25 comparisons, respectively. BPM thus enables personalization from pairwise feedback without per-user policy training, while characterizing the coefficient ambiguity left by limited feedback.
[LG-122] A Rank Graduation metric for Algorithmic fairness
链接: https://arxiv.org/abs/2609.39025
作者: Dalia Atif,Paolo Giudici
类目: Machine Learning (cs.LG); Applications (stat.AP); Machine Learning (stat.ML)
*备注: 44 pages, 5 figures
Abstract:Fairness assessment in algorithmic decisions that affect individuals, such as credit scoring, often relies on parity measures calculated at the aggregate group level. Such measures may not reveal which individuals experience unfairness or which explanatory factors contribute to it. In this paper, we propose a rank-based framework that evaluates fairness through the distribution of model prediction errors, thereby linking fairness assessment with predictive accuracy and explainability. The framework combines Rank Graduation Fairness (RGF), its integrated measure AURGF, a centered Cramer–von Mises permutation test, and a feature removal procedure for fairness explainability. We evaluate the methodology using logistic regression, random forest, gradient boosting, and a multilayer perceptron. The simulation study shows that protected-group imbalance can reverse descriptive fairness comparisons, whereas the proposed inferential procedure correctly distinguishes fair from unfair mechanisms. Its application to HMDA mortgage data produces model rankings that differ from those obtained with classical fairness criteria. Tree-based models, rather than logistic regression, provide the strongest combination of predictive accuracy and rank-based fairness, while the fairness null hypothesis is rejected for all four models. The persistence of disparity across statistical, bagging, boosting, and neural network specifications, together with the feature removal results, indicates that the observed unfairness is not specific to a single algorithm or predictor, but is associated with group differences embedded in the characteristics of the lending data. These findings support a broader approach to trustworthy artificial intelligence that combines predictive accuracy, fairness measurement, statistical inference, and explainability. Comments: 44 pages, 5 figures Subjects: Machine Learning (cs.LG); Applications (stat.AP); Machine Learning (stat.ML) Cite as: arXiv:2609.39025 [cs.LG] (or arXiv:2609.39025v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.39025 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-123] Synchronous Multi-view Neural Diffusion
链接: https://arxiv.org/abs/2609.39019
作者: Yongquan Shi,Weijun Huang,Yueyang Pi,Wendi Zhao,Yiqing Shi,Shiping Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Multi-view learning seeks to learn more comprehensive representations by exploiting the complementarity and consistency across diverse modalities or views. However, existing multi-view fusion strategies treat intra- and inter-view fusion as independent stages, without simultaneously considering the evolution within views and the dependency across views. Such an asynchronous fusion paradigm inevitably constrains cross-view interactions due to conflicting view-specific structural inductive biases. As a result, information flow is prone to distortion and compression along intermediate pathways, confining the model to learn within a restricted solution space. To address this, we propose Synchronous Multi-view Neural Diffusion (SynMDiff), which conceptualizes the multi-view feature space as a unified dynamical system driven by a diffusion process. By modeling the diffusion flow across arbitrary dyadic feature interactions in a joint space, SynMDiff enables the concurrent and adaptive intra- and inter-view information fusion. While a direct implementation of this synchronized mechanism incurs prohibitive computational costs, we further introduce an energy-based topological sampling strategy and an Ego-Net style centralized training architecture, ensuring both efficiency and scalability during learning and inference. Due to its conceptual elegance and computational efficacy, evaluations on real-world datasets demonstrate that SynMDiff outperforms the baselines by a large margin.
[LG-124] Not all solutions are created equal: An analytical dissociation of functional and representational similarity in deep linear neural networks ICML2025
链接: https://arxiv.org/abs/2609.38998
作者: Lukas Braun,Erin Grant,Andrew M. Saxe
类目: Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注: 28 pages, 5 figures. ICML 2025 camera-ready version. Presented at ICML 2025 and CCN 2025
Abstract:A foundational principle of connectionism is that perception, action, and cognition emerge from parallel computations among simple, interconnected units that generate and rely on neural representations. Accordingly, researchers employ multivariate pattern analysis to decode and compare the neural codes of artificial and biological networks, aiming to uncover their functions. However, there is limited analytical understanding of how a network’s representation and function relate, despite this being essential to any quantitative notion of underlying function or functional similarity. We address this question using analysable two-layer linear networks and numerical simulations in non-linear networks. We find that function and representation are dissociated, allowing representational similarity without functional similarity and vice versa. Further, we show that neither robustness to input noise nor the level of generalization error constrain representations to the task. In contrast, networks robust to parameter noise have limited representational flexibility and must employ task-specific representations. Our findings suggest that representational alignment reflects computational advantages beyond functional alignment alone, with significant implications for interpreting and comparing the representations of connectionist systems.
[LG-125] Smaller Models Better Rejects: Preference Distillation Scaling
链接: https://arxiv.org/abs/2609.38987
作者: Rui Cai,Wenhui Zhu,Xiwen Chen,Jincheng Cao,Han Yu,Shayan Mohajer Hamidi,Zelin He,Qiyao Ma,Daiwei Chen,Xuanzhao Dong,Yuanda Xu,Jelena Markovic-Voronov,Kayhan Behdin,Zhengze Zhou,Ran He,Alborz Geramifard,Rohit Jain,Zhe Zhao
类目: Machine Learning (cs.LG)
*备注:
Abstract:Preference distillation typically treats a teacher response as preferred and the student’s own response as rejected. This assumes that self-generated failures are the most informative negatives and that rejects must come from a model at least as large as the student, making generation costly at scale. We find neither assumption holds: across students from 7B to 72B, smaller frozen models generate rejects with less inference compute yet train stronger students than self-generated rejects, before and after sequence-level knowledge distillation, on code generation and mathematical reasoning. To explain this result, we derive a finite-horizon utility bound for Direct Preference Optimization in a linearized feature model. The bound characterizes favorable reject distributions and motivates three interventions. First, mixing rejects from smaller and student-scale models improves performance as the smaller model’s share increases. Second, reassigning rejects to other prompts and shuffling their code tokens still outperform length-matched gibberish, showing that task structure contributes to reject utility. Third, selecting candidates with lower likelihood under the reference policy improves net transfer when higher-likelihood candidates provide less useful contrast. Lower-likelihood selections outperform higher-likelihood ones for every source. These results suggest that effective rejects preserve task structure while limiting coupling to the reference policy, and that smaller frozen models can provide them at low cost.
[LG-126] SCORE-LM: State-Space Radar Representations with Language Models for Fault Diagnosis
链接: https://arxiv.org/abs/2609.38980
作者: Mainak Mallick,Seung-Kyum Choi
类目: Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注:
Abstract:Radar hardware faults threaten automated perception, motivating accurate, compact diagnosis and understandable maintenance guidance. We introduce SCORE-LM, which couples a small scatterer-conditioned operator-response encoder (SCORE) to an adapted local language model. SCORE combines self-referenced complex trajectories, physical descriptors, and a selective state-space branch, with source-only self-supervision and directional fault inference. On eight capture-excluded Rad-R fault recordings, it achieves state-of-the-art performance within the evaluated nine-model comparison: 88.39% mean capture recall and 88.20% four-fault macro-F1 at ten frames. Its 39,520 radar inference coefficients are 119.7 times fewer than RadrNet-DS-CI’s, while recall is 15.56 percentage points higher than this strongest competitor. In a separate low-label protocol, SCORE reaches 71.58% recall with one labeled source window per class. A nonlinear projector converts four frozen fault similarities into five soft tokens, linking compact diagnosis to class-conditioned maintenance guidance. On 75 development questions covering 24 radar windows, language adaptation raises correct-fault answers from 45 to 62 (60.0% to 82.7%) relative to removing the co-trained adapters, while retaining the same projector. SCORE-LM thus combines a compact radar specialist with a language interface for communicating fault-specific inspection guidance.
[LG-127] Robust Risk-Sensitive Reinforcement Learning from Corrupted Human Feedback
链接: https://arxiv.org/abs/2609.38938
作者: Xinyi Ni,Lifeng Lai
类目: Machine Learning (cs.LG)
*备注:
Abstract:Reinforcement learning with human feedback (RLHF) learns from human comparisons, which can be corrupted or deliberately manipulated. This paper studies online risk-sensitive RLHF with static conditional value-at-risk (CVaR) under adversarial preference-label flips. We consider additive linear rewards and a fixed-reference protocol with one comparison per episode and at most C flipped labels over K episodes. We propose weighted streamed-preference CVaR RLHF (WSP-CVaR-RLHF), which combines uncertainty-weighted reward estimation with optimistic augmented-state CVaR planning. For known transitions and normalized rewards, we establish the regret bound \widetildeO\left(\fracd\kappa\sqrt\fracK\alpha+\fracdC\kappa\alpha\right) up to lower-order terms, where d is the reward-feature dimension, \alpha is the CVaR level, and \kappa characterizes the preference link. The bound separates the clean statistical cost from the penalty caused by corrupted feedback. We further extend the analysis to unknown tabular transitions, where the trajectory distribution entering the CVaR objective must be learned together with the reward. We address the resulting coupled uncertainty using rectangular transition confidence sets, joint optimistic planning, and a history-level CVaR simulation argument. Experiments under four adversarial attacks demonstrate that WSP-CVaR-RLHF consistently reduces cumulative regret relative to its unweighted robust counterpart while preserving confidence-set coverage.
[LG-128] PrivCert: Certifying Statement Support under Differential Privacy
链接: https://arxiv.org/abs/2609.38934
作者: Tsubasa Takahashi,Takumi Hiraoka
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注:
Abstract:Differentially private (DP) text generation can protect individual records, but privacy alone does not specify what evidence a released statement carries about the underlying data. We identify this as an evidence gap: a private report may contain plausible claims without indicating whether they are strongly supported by the private dataset. We introduce PrivCert, a framework for privacy-preserving reporting that makes statement support explicit through privacy-preserving certificates and emit-or-abstain decisions. As a canonical instantiation, PrivCert-PF (Proposal-and-Filter) separates data-independent candidate discovery from private support certification, emitting only statements whose support passes a private evidence test. We provide theoretical grounding for this framework by characterizing the limits of implicit evidence under DP, deriving a sharp privacy–honesty frontier for single-statement certification, and establishing a worst-case cost for fine-grained multi-statement certification. Experiments on synthetic tasks and TAB, WildChat, and Yelp show that explicit certification maintains low unsupported emission, while free-text DP baselines frequently produce low-support claims under the same declared support semantics. We further show that the PrivCert contract can be realized with histogram, sparse-vector, and Gaussian mechanisms, and use DP synthetic data to illustrate an important boundary: support in a private proxy does not automatically certify support in the original data. Together, these results position privacy-preserving reporting as an evidence-design problem: not only how to generate private text, but what a private report can substantiate about its underlying data.
[LG-129] PrecipJEPA: JEPA-Regularized Future-State Prediction with Motion-Source Rendering for Precipitation Nowcasting
链接: https://arxiv.org/abs/2609.38926
作者: Yufeng Zhu,Dan Niu,Qiliang Wu,Weiwei Huang,Yixiao Liang,Yongchao Feng,Chunlei Shi
类目: Multimedia (cs.MM); Machine Learning (cs.LG)
*备注: 5 pages, 3 figures
Abstract:Long-term precipitation nowcasting requires modeling radar-echo evolution while preserving localized high-intensity structures. Recent radar-specific studies motivate location-aware prediction and separating echo displacement from intensity change. However existing encoders learn historical representations mainly from final forecast errors. We propose PrecipJEPA, which couples a structured forecasting path with an auxiliary path that enriches its encoder from observed radar history. In the forecasting path, an online encoder first converts the observations into spatiotemporal tokens. The Task-Driven Future-State Predictor (TFP) combines these tokens with a recent-dynamics summary and spatiotemporal queries to construct future radar states. The Parallel Motion-Source Renderer (PMSR) decodes these states into motion and source-sink fields that transform the latest observation into future frames. During joint training, the History-Masked JEPA (H-JEPA) operates on the auxiliary path to predict masked historical features from visible context, directly supervising the same online encoder from the observed sequence. Experiments on SEVIR and MeteoNet show that PrecipJEPA improves highest-threshold CSI by 118.6% and 35.1%, respectively, over the strongest baselines, while maintaining the highest mean CSI throughout the 3-hour forecast.
[LG-130] Learning Where to Steer: Noise-Space Geometry for Efficient Offline Multi-Objective Optimization with Generative Models
链接: https://arxiv.org/abs/2609.38920
作者: Yuan Lu,Esha Singh,Yi-An Ma,Yusu Wang
类目: Machine Learning (cs.LG)
*备注: 65 pages, including appendices
Abstract:Offline multi-objective optimization (MOO) seeks solutions with better objective trade-offs using only a fixed dataset, without querying the objectives. Diffusion models trained on such data have emerged as a promising approach, but their samples are not inherently better than the data and must be steered toward the Pareto front. Existing methods guide or condition every sampling step. We instead act on the initial noise and leave the sampling process unchanged. Across Off-MOO-Bench, we observe that the objectives, as functions of the noise, are sensitive to only a few directions. We estimate these directions once per task via a Recursive Feature Machine using function values alone, and a small cache serves every trade-off, so each candidate costs one noise displacement and one ODE solve. We prove that this displacement increases the learned scalarized objective in expectation, and that sweeping trade-offs recovers the flow’s attainable front up to proxy and steering errors. With additional guidance, for which we introduce novel data-adaptive and Pareto-aware operators, our method attains the best average hypervolume rank among generative methods on 47 tasks, at comparable or lower sampling cost. Steering alone outranks the best prior generative method at a fraction of its sampling cost.
[LG-131] Flow Matching under Noisy Latent Structure: Beyond Exact Low-Dimensional Support
链接: https://arxiv.org/abs/2609.38918
作者: Lifeng Hao,Shaolin Ji
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 66 pages, 0 figures
Abstract:Flow Matching (FM) learns a velocity field whose ODE transports a simple source distribution to a target law. Existing finite-sample theory largely treats ambient-space regularity or data supported exactly on low-dimensional sets. We study linear FM under a noisy latent-generator model, where a low-dimensional Hölder map is perturbed by nondegenerate ambient Gaussian noise, so the target law is full-dimensional despite its latent structure. We construct a spatially regular ReLU velocity class and establish non-asymptotic high-probability approximation and estimation bounds whose leading sample-size exponent is governed by the latent dimension rather than the ambient dimension, with ambient and noise dependence kept explicit. Fixed positive target noise keeps the interpolation nondegenerate over the full time interval. The same spatial regularity propagates the learned velocity error through the transport ODE, yielding a corresponding Wasserstein convergence guarantee. These results show that exact low-dimensional support is not necessary for Flow Matching to retain latent-dimensional statistical behavior.
[LG-132] Generalized Residual Closure: General Learning Dynamics for Stability-Plasticity Compatibility
链接: https://arxiv.org/abs/2609.38911
作者: Dongxu Li,Yinuo Zhang,Hongyu Zhang,Feng Tian
类目: Machine Learning (cs.LG)
*备注:
Abstract:Learning must acquire new capabilities while preserving both prior responsibilities and the capacity to learn again. We introduce Generalized Residual Closure (GRC), a framework for learning as recursive closure of future-relevant discrepancies: closing a residual establishes the conditions for subsequent prediction, interaction, and learning. Under a complete representation-relation description at a fixed learner-world boundary, persistent internal learning has two primitive modes: Transformation within a representation and revision of the Representation itself. We establish a local tangent decomposition under regularity assumptions and a criterion for when representation revision is necessary. In an affine model, we derive a necessary-and-sufficient condition for stability-plasticity compatibility and the unique solution of a constrained quadratic update problem, which preserves registered old responsibilities while reducing residuals with an effective safe response. We prove that reconstructive semantic protection weakly enlarges the safe-response operator relative to preserving an exact historical realization. Dynamic sufficiency and future-closure viability extend representation adequacy from current prediction to lawful future updating and continued learning. Growth Learning expands the lawful closure domain or lowers optimal closure cost without regression of the registered capability-cost frontier; a conditional commit rule maintains this order. Restricted-sector recoveries and a conditional representation theorem connect the framework to optimization, machine learning, and control. Together, these results organize adaptation, representation revision, and reusable capability within a common account of continued learning.
[LG-133] Amortized Data Borrowing with Exchangeability-Aware Neural Posterior Estimation ALT
链接: https://arxiv.org/abs/2609.38902
作者: Chin-Hung Huang,JooChul Lee,Huan He
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 27 pages. Accepted at the 11th Machine Learning for Healthcare Conference (MLHC 2026)
Abstract:Augmenting small concurrent studies with external or historical cohorts is attractive in drug development, where enrollment is slow, follow-up is expensive, and closely related trial or real-world data are often already available. Bayesian dynamic borrowing (BDB) provides a principled framework for adaptively controlling the influence of external data, but classical implementations often depend on hand-specified priors and MCMC-based inference, which can be computationally expensive and not generalizable. In this work, we study amortized neural posterior estimation (NPE) as a flexible alternative. A single network is pretrained on simulated current/external dataset pairs spanning covariate shift, outcome drift, and joint non-exchangeability, and then returns an approximate posterior for a scalar current-study target in a single forward pass. Through simulation studies, we find that NPE is most useful under outcome drift and joint mismatch: in the harder outcome-drift regimes, it gives up to about five-fold lower absolute bias than the best classical baseline and keeps Type I error close to nominal. After pretraining, posterior summaries are obtained in about 8 ms per dataset, roughly 10^3\times faster than MCMC-based borrowing baselines in our timing experiment. We further analyze Alzheimer’s Disease Neuroimaging Initiative (ADNI) data and show that, when mild cognitive impairment outcomes differ across cohorts, the NPE formulation recovers the later-cohort risk level in this example without claiming greater precision. Code is available at this https URL.
[LG-134] Certified Approximation for Interpretable Representer Landmarks
链接: https://arxiv.org/abs/2609.38901
作者: Jayanta Mukherjee,Shourya Verma,Mengbo Wang,Jasorsi Ghosh,Ananth Grama
类目: Machine Learning (cs.LG)
*备注: 29 pages, 8 figures
Abstract:Representer explanations rank the training landmarks that most influence a self-supervised representation. At scale, this ranking rests on up to four stacked approximations of the empirical neural tangent kernel (eNTK). These are random output heads, a parameter sketch, landmark sampling and a coefficient fit. Existing analyses bound each approximation separately, but none certifies the top- K set against their combined error. We introduce CAIRN (Certified Approximation for Interpretable Representer laNdmarks), a framework that carries this error through to the ranking. We derive the exact variance of the sketched multi-head eNTK, which matches measurement within 4% where Johnson-Lindenstrauss bounds err by up to 2.5\times . This yields a high-probability top- K certificate for a fixed coefficient fit, alongside exact residual-trace certificates for discarded spectral mass. An exact product-variance identity separates kernel error from fit variability and identifies when a larger kernel budget can still sharpen a ranking. Stochastic Lanczos Quadrature (SLQ) estimates the effective dimension within 0.72% and guides the landmark budget without dense eigendecomposition. We show that residual mass does not control class coverage, and residual-greedy selection cuts the worst coverage excess of k -means++ from 8.5\times to 1.55\times ( 4\times on the sketched eNTK). Cross-view initializers outperform principal-component initialization in five (AUI) to all six (CSI) settings. Against the KREPES Gauss-Newton solver, CAIRN converges 2.5 to 11.3\times faster, trails by at most 0.31 points and gains up to 3.14 points on MNIST. Together, these results make the reliability of representer explanations measurable and show where approximation budgets are best spent.
[LG-135] VERA: Verifiable Feasibility Representations with Counterfactual Credit for Constrained Multi-Agent Control
链接: https://arxiv.org/abs/2609.38889
作者: Bo Yin,Dongbo Li,Hongkai Chen,Jie Liu,Guoliang Xing
类目: Machine Learning (cs.LG)
*备注: 23 pages, 11 figures, 17 tables, including appendix
Abstract:Constrained multi-agent control requires more than predicting rewarding actions: an action can cease to be executable as contact windows, shared capacity, and deadlines change. We introduce VERA, a centralized-training, decentralized-execution framework that separates feasibility estimation from credit assignment. Each actor predicts a five-dimensional verifiable feasibility representation (VFR). After an action is proposed, exact action-conditioned margins available only during training supervise that representation, while a counterfactual group-relative advantage (CGRA) ranks candidate representation-action pairs. Execution uses one actor pass and no privileged state. In a dynamic space-air-ground integrated network (SAGIN), VERA obtains 55.33% +/- 3.60% success with 0.45% +/- 0.81% coverage violation, within 1.33 percentage points of a privileged-mask reference. With rewards matched over ten paired seeds, VERA improves success over the strongest baseline by 8.74 percentage points (p=0.023) and reduces violation by 52.19 percentage points (p=5.7e-8). A ten-seed 4-by-2 factorial attributes a 14.16-16.48 percentage-point gain to CGRA across handcrafted, learned, random, and latent representations; evaluation on seven unseen topologies preserves a 24.33-30.02 percentage-point advantage over multi-agent proximal policy optimization. From 10 to 40 users, success remains 50.1-53.8%, and VFR adds only 0.026 ms to a central processing unit (CPU) actor step. Cross-domain tests further identify the governing condition: counterfactual credit succeeds when candidate scores respect shared constraints and fails under incompatible reward geometries. These results establish action-conditioned feasibility as an auditable training interface and counterfactual credit as a geometry-dependent optimization mechanism.
[LG-136] On Parameters of Nonlinear Scalar Dynamics from Video: Invariants Calibration and Identifiability
链接: https://arxiv.org/abs/2609.38877
作者: Wenjie Wang,Yuanyuan Wang,Zixiang Jiang,Shaoan Xie,Mingming Gong
类目: Machine Learning (cs.LG)
*备注:
Abstract:Physical parameter estimation from video aims to recover the parameters of a known family of governing dynamical equations from pixel observations. Existing identifiability theory for this setting has focused on linear time-invariant (LTI) second-order systems, leaving open what can be identified for nonlinear scalar dynamics. We develop an identifiability theory for nonlinear scalar second-order ODEs, organized by how their velocity dependence interacts with changes of the learned state coordinate. Under a shared non-collapsed state map and explicit same-state velocity-coverage conditions, we show that parameter identifiability depends on the ODE family: some parameters are uniquely identifiable, while in other families only invariant parameter combinations are identifiable or external physical calibration is required. For laws that are at most linear in velocity, compatibility forces affine coordinate alignment, yielding explicit parameter relations, invariants, and calibration conditions. This affine conclusion extends to broader finite velocity-feature families when coordinate curvature can be separated from the declared velocity dependence. For families admitting a squared-velocity term, nonlinear coordinate ambiguity can remain; a law-derived normalization instead enables affine comparison between canonical laws. Experiments on synthetic systems and real pendulum and free-fall videos support the predicted parameter relations, coverage effects, and calibration requirements.
[LG-137] Visualizing Distribution Coverag e in Generative Diffusion Models
链接: https://arxiv.org/abs/2609.38853
作者: Yifei Wang,Xiaoyu Wu,Tsu-Jui Fu,Chen Chen,Liang-Chieh Chen,Zhe Gan,Chen Wei
类目: Machine Learning (cs.LG)
*备注: under review
Abstract:Diffusion distillation is widely adopted to accelerate sampling, and the resulting few-step models are broadly believed to match or even surpass their multi-step teachers in generation. However, standard evaluations such as GenEval2 typically draw only one sample per prompt, so improved scores may fail to reveal losses in distribution coverage. We therefore revisit whether distilled models truly match their teachers beyond single-draw performance using \textbfpass@ \mathbfk , which measures the probability that at least one of k independent samples satisfies a quality criterion. At k=1 , pass@ k reduces to standard single-draw evaluation. As k grows, the curve reveals whether additional draws find genuinely different successes or merely revisit the same modes, directly exposing how broadly a model covers the space of valid outputs. We first show that classifier-free guidance (CFG), whose quality–coverage tradeoff is well established, is the clearest case: higher guidance improves pass@ 1 , but its advantage shrinks and reverses at larger k . Applying pass@ k to few-step distilled models, we find the same tradeoff splits along training objectives: distribution-matching objectives concentrate the student’s output distribution, boosting early-hit rates while eroding large-budget coverage, whereas consistency and trajectory-based objectives better preserve the teacher’s coverage even at large k . We further show that this tradeoff extends to few-step causal video generation. Our findings reveal a previously overlooked cost of diffusion distillation: across both image and video generation, the choice of training objective fundamentally determines whether a few-step model inherits its teacher’s distribution coverage or trades it away for single-draw quality.
[LG-138] Optimal VC Dimension of Contrastive Learning with Margin
链接: https://arxiv.org/abs/2609.38834
作者: Dionysis Arvanitakis,Vaggos Chatziafratis,Yiyuan Luo,Konstantin Makarychev
类目: Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: Main result obtained in April 2026 without the use of AI
Abstract:Contrastive learning is a successful paradigm for learning d -dimensional geometric representations from a collection of anchor--positive--negative'' triplets (i,j^+,k^-) , indicating that item i is closer to j than to k .‘’ Despite its success, understanding why contrastive learning leads to representations of high \textitgeneralization quality—beyond the often pessimistic predictions from PAC-learning—remains a central question. Recently, \citet*alon2024optimal proved that, for PAC-learning d -dimensional Euclidean representations of n -point datasets, \Theta(\min(nd, n^2)) triplets are necessary and sufficient, while they posed as an open question whether their VC dimension bounds for the more realistic setting of \textitcontrastive learning with a margin can be improved. For a margin parameter \alpha 0 , a triplet (i,j^+,k^-)_\alpha is satisfied by the embedding \phi:[n]\rightarrow \mathbbR^d , if |\phi(i)-\phi(k)|_2(1+\alpha)\cdot|\phi(i)-\phi(j)|_2 . In this work, we resolve their question by proving that the VC dimension of contrastive learning under any margin \alpha\in(0,1) is in fact O(n/\alpha^2) , improving on the previous bound of O(n\log(n)/\alpha^2) . We also establish that the bounds are optimal up to constant factors, by providing a matching lower bound of \Omega(\fracn\alpha^2) (the previously known lower bound was \Omega(\fracn\alpha) ), for \alpha\geq \max(n^-1/2,d^-1/2) .
[LG-139] ReSCENE: Server-Side Replay for Structural Mitigation of Catastrophic Forgetting in Federated Continual Learning
链接: https://arxiv.org/abs/2609.38833
作者: Sungmin Kang,Zhengzhong Tu,Sunwoo Lee
类目: Machine Learning (cs.LG)
*备注: 26 pages
Abstract:Federated continual learning must integrate new tasks over time without losing earlier-task knowledge. Most existing methods attach an anti-forgetting mechanism to the client-trained, server-aggregated loop of federated learning, which holds back new learning to preserve earlier knowledge and burdens resource-constrained clients. We propose ReSCENE, which structurally mitigates catastrophic forgetting by having each client upload a small condensed surrogate of its local data while the server keeps the surrogates of past tasks and trains the global model on them together with the current task surrogates. For efficient server memory, we introduce temporal herding, which selects the more recent surrogates from the pool accumulated over a task into a compressed buffer. Our study provides a theoretical analysis showing that this buffer can represent the original task data more closely than full accumulation of all surrogates. Across CIFAR-10, CIFAR-100, and TinyImageNet, ReSCENE achieves the strongest accuracy over seven baselines, by up to 31.1 points of average accuracy, while requiring as little as 0.11\times of the client computation and up to 179\times less upload than the model-update baselines. ReSCENE further demonstrates its effectiveness when scaled to larger client populations and larger models while remaining efficient, which makes it a practical method for federated continual learning.
[LG-140] SparLeak: Privacy Leakage from Sparse Attention in LLM Inference on Shared GPUs
链接: https://arxiv.org/abs/2609.38830
作者: Fahao Chen,Linkang Du,Jinhao Zhou,Peng Li,Zhou Su
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注:
Abstract:Sparse attention is widely used to accelerate long-context inference in modern large language models (LLMs), but its input-dependent execution behavior introduces previously unexplored privacy risks. We identify a new GPU micro-architectural side channel, termed Sparsity-Induced Memory Access (SIMA), which arises from secret-dependent key-value cache access patterns induced by sparse attention. Based on this observation, we present SparLeak, a phase-aware side-channel attack that extracts SIMA traces during LLM inference and enables two practical privacy extractions: query attribute inference from prefill-phase traces and autoregressive response reconstruction from decoding-phase traces. By reconstructing approximate token-level sparsity profiles from page-level observations and applying profiling-based learning, SparLeak accurately recovers sensitive information, including user-query attributes and private LLM response content. Extensive evaluation across three LLM architectures, three sparse attention mechanisms, and three privacy-sensitive datasets shows that SparLeak achieves average attack success rates of 90.9% for attribute inference and 87.3% for response reconstruction under real-world LLM serving settings, highlighting the significance to account for SIMA leakage when deploying sparse-attention-based LLM systems. We provide anonymized SIMA traces, trained attack models, evaluation scripts, and documentation as artifacts at this https URL. Subjects: Machine Learning (cs.LG); Cryptography and Security (cs.CR) Cite as: arXiv:2609.38830 [cs.LG] (or arXiv:2609.38830v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.38830 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-141] Learning Chaos Without Seeing Chaos: Extrapolation of Global Dynamics in Autoregressive Transformers
链接: https://arxiv.org/abs/2609.38814
作者: Yilun Liu,Yi Zhang,Ganyu Wu,Sikuan Yan,Mengyue Wang,Alois Knoll,Volker Tresp,Yunpu Ma
类目: Machine Learning (cs.LG); Chaotic Dynamics (nlin.CD)
*备注:
Abstract:Autoregressive models are trained to predict a system’s behavior one step at a time, and recursive generation allows the learned dynamics to unfold over long horizons. To what extent can such dynamics learned from local observations recover broader organization of an underlying system that was only partially observed during training? Here we study small autoregressive transformers trained from scratch on trajectories sampled from restricted parameter regimes of several non-linear dynamical systems, including logistic and sine maps, the Lorenz system, and the generalized Hopf system, with control parameters and state trajectories represented as sequences of continuous tokens. Under closed-loop evaluation at parameters far outside the training distribution, the models can recover self-similar period-doubling cascades, chaotic dynamics, and attractor structures with remarkable visual and numerical fidelity. For the logistic map, a transformer reproduces successive period doublings up to period 128, yielding a finite-order scaling ratio of 4.6687, matching the Feigenbaum constant to within 5\times10^-4 . We further investigate how these structures emerge over the course of training, and reveal with causal interventions how control-parameter information is processed through attention into state prediction and shapes the resulting closed-loop dynamics. These results suggest that a surprisingly narrow window into a system’s local behavior may suffice for autoregressive transformers to generalize to its unseen global dynamical organization.
[LG-142] LEARN-TS: LLM -Enhanced Alignment and Reconstruction with Normality Guidance for Multivariate Time-Series Anomaly Detection
链接: https://arxiv.org/abs/2609.38789
作者: Jahyeob Koo,Kio Yun,Byoungmo Koo,Jun-Geol Baek
类目: Machine Learning (cs.LG)
*备注: 24 pages, 7 figures, 13 tables
Abstract:Reconstruction errors in multivariate time-series anomaly detection may not reliably distinguish abnormal behavior from benign deviations. Language-derived semantics offer complementary context, but existing multimodal approaches may rely on time-associated paired textual information that is difficult to obtain consistently and is not provided by standard multivariate time-series anomaly detection benchmarks. This setting poses two challenges: (1) conditioning masked reconstruction on window-specific semantics without exposing exact numerical targets or anomaly-specific cues, and (2) using a window-independent concept of normality as a complementary semantic reference rather than an independent anomaly detector. We propose LLM-Enhanced Alignment and Reconstruction with Normality Guidance for Time Series (LEARN-TS), which uses a frozen language model to construct two role-separated semantic representations without requiring temporally paired external text. Window-specific observation semantics encode temporal and cross-variable context without exact numerical values to guide channel-shared patch-masked reconstruction. A fixed, dataset-agnostic normality prompt provides a window-independent semantic reference for aligning normal representations and estimating normality discrepancy. At inference, masking each temporal patch once yields timestamp-level reconstruction evidence, conditionally modulated by discrepancy from a separate unmasked view. Across four benchmarks, LEARN-TS achieves the highest mean performance in 13 of 16 dataset-metric comparisons. Controlled ablations examine observation conditioning, joint normality alignment and scoring, and reference content, showing dataset-dependent ranking benefits and modest average gains from semantic over random references.
[LG-143] Same Loss Different Gradients
链接: https://arxiv.org/abs/2609.38786
作者: Ningkang Peng,Xiaoqian Peng,Yifan He,Anjie Hu,Chao Tan,Peirong Ma,Yanhui Gu
类目: Machine Learning (cs.LG)
*备注: 30 pages
Abstract:Differentiable learning typically assumes that the scalar objective evaluated in the forward pass and the gradient supplied to the optimizer in the backward pass describe the same mathematical object. We show that this correspondence can fail when probabilistic objectives rely on finite special-function recurrences, custom backward rules, and numerical clipping. In high-dimensional von Mises-Fisher learning, real numerical implementations can produce identical forward scores and losses at the same learning state while supplying different gradients and following different optimization trajectories. We characterize the structure of this mismatch in finite-start Bessel recurrence and show that classwise radial mismatch can compose through probabilities into a locally nonconservative update field. Evaluating the accuracy of special-function values and derivatives separately is therefore insufficient to characterize the realized learning objective. Motivated by this observation, we introduce AR/FR, a fixed-depth analytic realization that constructs a potential and its derivative jointly, ensuring forward-backward coherence by construction. We establish a uniform cubic-order error bound relative to the exact Bessel ratio over the entire nonnegative concentration axis and propagate this guarantee to learning scores and objectives. As representation dimension increases, the original finite recurrence becomes sequentially deeper, whereas the worst-case AR/FR error guarantee tightens cubically, jointly providing coherence, certified fidelity, and fixed-depth computation. These results suggest that a differentiable numerical primitive is defined by both the values it realizes and the derivatives it actually supplies to the optimizer; together, they constitute the numerical realization of the learning algorithm.
[LG-144] How Accurate Is Accurate Enough?
链接: https://arxiv.org/abs/2609.38785
作者: Ningkang Peng,Qianfeng Yu,Jingyang Mao,Xiaoqian Peng,Yanhui Gu
类目: Machine Learning (cs.LG)
*备注: 24 pages, including supplementary material
Abstract:How accurate must a numerical approximation be within a learning system? Primitive error alone cannot answer this question: errors of the same magnitude can have very different consequences for losses, predictions, and gradients at different learning states. We study this question through the learning objective itself. The objective weights classwise numerical errors nonuniformly according to the current state, so the importance of an error depends not only on its magnitude but also on the class it affects and the weight that class receives. For softmax cross-entropy, we characterize this coupling between class weights and errors and derive the exact extrema of the signed loss change over pairings of fixed non-target probability and score-error multisets, with the target probability and target score error held fixed. Building on this structure, we establish finite-error guarantees that propagate primitive error to losses, probabilities, predictions, and feature gradients, then invert these guarantees to obtain a certified primitive tolerance for the current state under prescribed learning-level error requirements. We give a complete instantiation of the framework in high-dimensional von Mises-Fisher learning. Controlled interventions and a large collection of saved learning states show that identical primitive error can produce substantially different learning consequences, while certified numerical tolerances vary by orders of magnitude across states under the same learning-level requirements. These results show that the adequacy of a numerical approximation must be assessed in relation to the current learning state and the quantity to be preserved; numerical accuracy should itself be treated as part of the learning objective.
[LG-145] Does a Shared Temperature Imply a Shared Angular Scale in Probabilistic Contrastive Learning?
链接: https://arxiv.org/abs/2609.38784
作者: Ningkang Peng,Qianfeng Yu,Jingyang Mao,Xiaoqian Peng,Tingyu Lu,Peirong Ma,Yanhui Gu
类目: Machine Learning (cs.LG)
*备注: 59 pages, including supplementary material
Abstract:In probabilistic contrastive learning, a shared temperature is commonly interpreted as a shared similarity scale, but this interpretation does not hold for high-dimensional distributional class representations. We study the exact von Mises-Fisher (vMF) probabilistic score used by ProCo when representation dimension and class concentration grow jointly. We prove that the score retains a class-dependent leading angular gain g_c=A_c/\tau , where A_c is the mean resultant length. This gain enters Softmax competition, pairwise decision boundaries, and feature gradients. On real CIFAR-LT, ImageNet-LT, and iNaturalist representations, the theory accurately predicts boundary movements and local gradient changes under the full vMF score. Classwise temperature adjustment also changes the cosine-zero intercept and finite-dimensional response. We construct intercept-preserving and Pure Angular controls to separate the leading gain from these accompanying changes. Complete gain equalization yields a shared-scale cosine prototype rule at leading order; a finite-dimensional margin condition guarantees agreement of the two classifiers. Across 16 frozen representation settings, prediction agreement is 98.43-99.99%, with disagreements concentrated at small cosine margins. In controlled contrastive-only training with the training-frequency prior, Pure Angular editing improves both learned representations at all tested CIFAR-10/100 imbalance factors and retains positive changes on ImageNet-LT. Thus vMF concentration not only describes class distributions, but also forms a decision and learning scale in high-dimensional probabilistic contrastive learning.
[LG-146] Lasting Effects of Abstract Pretraining Beyond Perplexity
链接: https://arxiv.org/abs/2609.38764
作者: Zachary Shinnick,Hemanth Saratchandran,Damien Teney,Anton van den Hengel
类目: Machine Learning (cs.LG)
*备注: Project page: this https URL
Abstract:Language models are typically pretrained from random initialization. Recent work challenges this convention, showing that a brief warm-up on abstract, algorithmically generated data can provide a better starting point for subsequent learning of natural language. In this paper, we show that in small language models, such a warm-up improves specific capabilities that are not reflected in language-modeling perplexity. Our warm-up uses an abstract stack-manipulation task that requires compositional and state-tracking capabilities. Allocating as little as 1% of pretraining tokens to this data improves multi-hop question answering by up to 3.9 F1 points on MUSIQUE, with additional gains on HOTPOTQA and 2WIKIMULTIHOPQA despite comparable language-modeling perplexity. Controlled experiments show that the warm-up substantially accelerates the acquisition of deeper reasoning chains. We also explore what drives this transfer. First, the structure of the data matters: replacing the stack task with a queue fails to produce the same gains. Second, the gains are specific: performance improves on sequential reasoning chains, with no consistent benefit on tasks that combine or compare independent facts. Third, timing matters: mixing abstract data with natural language is far less effective than an initial dedicated phase, and exposure after pretraining completely removes the benefits. The early advantage persists through billions of subsequent language tokens. These results show that early abstract training can reliably shape the capabilities language models later acquire.
[LG-147] Molecular Property Prediction under Structural Shift with Tabular Foundation Models
链接: https://arxiv.org/abs/2609.38744
作者: Jinmo Lee,Dooho Lee,Minho Jeong,Jaemin Yoo
类目: Machine Learning (cs.LG)
*备注:
Abstract:Predicting molecular properties for compounds that differ structurally from labeled training molecules is important for drug discovery and materials design. Tabular foundation models (TFMs) offer a promising approach through in-context learning, but their performance under structural shifts and the value of molecular comparisons in this setting remain underexplored. We study structural generalization in molecular property prediction and introduce MolPAIR (Molecular Pair-Augmented In-context Refinement), a framework that combines molecule-level and molecular-pair contexts without task-specific parameter updates. A global tabular foundation model (TFM) first predicts a query’s property from labeled molecular examples. A second frozen TFM predicts differences in prediction errors between the query and labeled reference molecules, using these comparisons to refine the initial prediction. Across 58 MoleculeACE and Polaris tasks, CheMeleon representations combined with TabPFN-3 already outperform each evaluated baseline on a majority of tasks. MOLPAIR further improves this predictor on 46 of 58 tasks, with gains across four molecular representations and three TFM backbones. These results show that explicit molecular comparisons can strengthen tabular in-context learning for structural generalization while keeping the molecular encoder and pretrained model weights fixed. The code and datasets are available at this https URL.
[LG-148] In-Distribution Imagination for Model-Based Offline Reinforcement Learning
链接: https://arxiv.org/abs/2609.38673
作者: Mintae Kim,Koushil Sreenath
类目: Machine Learning (cs.LG)
*备注: 11 pages, 3 figures, RLC 2026 MBRL Workshop
Abstract:Model-based offline reinforcement learning (MBORL) improves sample efficiency through model-generated trajectories. However, accumulative model error can drive imagined trajectories outside the offline data distribution, leading to unrealistic synthetic data and unstable policy optimization. Many existing methods primarily control rollouts using transition-level uncertainty. We propose \emphin-distribution imagination (IDI), a rollout control framework that estimates trajectory support in a learned representation space and adaptively truncates rollouts that leave the offline trajectory manifold. Combined with trajectory-regularized RL, an extension of entropy-regularized RL, IDI consistently improves performance in limited-data settings. Experiments show that trajectory support predicts rollout failure substantially better than transition-level uncertainty, highlighting the importance of trajectory-level rollout control in MBORL.
[LG-149] Uncertainty-Normalized Margins for Direct Preference Optimization
链接: https://arxiv.org/abs/2609.38647
作者: Sadegh Khorasani,Petrus Mikkola,Matthias Grossglauser
类目: Machine Learning (cs.LG)
*备注: 36 pages
Abstract:Direct preference optimization (DPO) models binary preferences through a Bradley-Terry model with a common noise scale, without explicitly accounting for preference strength or prompt-dependent uncertainty from human feedback. We introduce uncertainty-normalized margin DPO (UNM-DPO), which combines strength-dependent margins with a learned prompt scale. Motivated by a heteroskedastic Bradley-Terry model, we develop two training objectives. Both compare the implicit rewards of preferred and rejected responses, derived from response log-probability ratios to a reference policy. Advantage-only (AO) divides this reward difference by the prompt scale before subtracting the margin; whole-residual (WR) subtracts the margin before dividing by the scale. For the WR comparison model, we establish a necessary and sufficient condition under which known margins make the prompt scale identifiable. We introduce a practical procedure for learning the scale. Building on WR, we introduce ULNM-DPO-WR, which normalizes each response’s implicit reward by its length. We evaluate our methods against DPO and related baselines on HelpSteer2 and HelpSteer3, using the Skywork reward model as a judge. With Llama-3.1-8B-Instruct, ULNM-DPO-WR achieves tie-adjusted win rates against matched DPO of 68.00% and 65.31% on evaluation panels, with higher mean rewards and shorter responses on average. On AlpacaEval with a GPT-4.1 judge and GPT-4-Turbo reference answers, the same 8B policy achieves a length-controlled win rate of 21.62%, compared with 16.39% for DPO and 15.30% for SimPO. These results demonstrate the potential of combining preference-strength margins, learned prompt scales, and length normalization for policy optimization.
[LG-150] SHIFT-Truck: A High-Fidelity Aerodynamics Dataset and Benchmark for Pickup Trucks
链接: https://arxiv.org/abs/2609.38638
作者: Riddhiman Raut,Yin Yu,Aashwin Anand Mishra,Michael Emory,Thomas Economon,Peter Lyu,Juan J. Alonso
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG); Computational Physics (physics.comp-ph)
*备注:
Abstract:Pickup trucks account for 14% of new light-duty vehicles produced in the United States, yet are among the least aerodynamic. Their open cargo bed adds a flow absent from existing automotive aerodynamics datasets such as DrivAerML and SHIFT-SUV: the shear layer leaving the cab roof passes over a recirculating bed flow before separating again at the tailgate. The resulting drag lowers fuel efficiency, raises emissions and limits the range of electric trucks. Scale-resolved Computational Fluid Dynamics (CFD) is too costly for broad design exploration; neural surrogates can predict flow features at a fraction of that cost, provided they are trained on large-scale, high-fidelity, domain-specific data. We introduce SHIFT-Truck, the first such dataset for pickup trucks. It comprises 1,000 Spalart-Allmaras delayed detached-eddy simulations (SA-DDES) of a reference pickup geometry morphed across 17 shape parameters. Each case is run on a mesh of about 100 million cells at a Reynolds number of 1.4 \times 10^7 and released with time-averaged surface pressure, wall shear stress, volumetric pressure and velocity. The setup is verified by grid refinement and repeated runs, and checked against wind-tunnel measurements. We define geometry-grouped splits and benchmark four neural surrogates, DoMINO, GeoTransolver, AB-UPT and SMART, on surface and volume tracks. SHIFT-Truck also introduces controlled distribution shifts in the operating point, the input surface discretization and the vehicle archetype. Models with strong in-distribution performance can degrade substantially under these shifts: operating-condition changes expose failures to infer speed dependence, while tessellation and cross-vehicle shifts reveal markedly different robustness across architectures. SHIFT-Truck is thus a benchmark not only for surrogate accuracy but also for generalization across physical and numerical distributions.
[LG-151] Proper Scoring Rule-based Diffusion for Probabilistic Weather Forecasting
链接: https://arxiv.org/abs/2609.38632
作者: Joonhyeong Park,Giung Nam,Hyungi Lee,Kyunghyun Cho,Byoungwoo Park,Juho Lee
类目: Machine Learning (cs.LG)
*备注:
Abstract:Recent probabilistic weather forecasters train stochastic predictors with the continuous ranked probability score (CRPS) to generate each ensemble member in a single forward pass. These models learn the predictive distribution from the forecast context alone, which becomes difficult at longer forecast horizons where uncertainty is high. To learn the predictive distribution more effectively, we introduce auxiliary conditional denoising tasks that predict the same future state from the context and its corrupted version, which provides partial future information that can reduce prediction ambiguity. Building on distributional diffusion models, we learn the conditional distributions of these tasks with a single stochastic predictor by minimizing a proper scoring rule across noise levels. At inference, the predictor can still generate each ensemble member in a single forward pass at the fully corrupted endpoint. Standard CRPS training is recovered as the endpoint-only special case of our formulation, so our framework extends existing CRPS-based forecasters with only additional conditioning inputs. Controlled experiments show that the auxiliary tasks improve one-step forecasting across architectures, with larger gains at longer forecast horizons. The gains extend to high-dimensional global weather forecasting under both training from scratch and fine-tuning, along with improved calibration and potential benefits for generalization under distribution shift.
[LG-152] Geometry-physics confounding impairs PDE learning across varying domains
链接: https://arxiv.org/abs/2609.38623
作者: Yinghao Cheng,Gengxiang Chen,Xu Liu,Qinglu Meng,Yixin Jing,Xiangguo Tang,Wenping Mou,Lihui Wang,Yingguang Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:Learning partial differential equation (PDE) dynamics across varying domains is central to predictive modelling and data-driven discovery of governing equations. However, geometric variation alters both field representation and the governing differential operators, confounding geometric effects with intrinsic physical properties in the observed dynamics. This work identifies geometry-physics confounding as a unified failure mechanism for PDE learning across varying domains. In forward operator learning, this confounding increases the burden of inferring geometry-dependent operator changes from finite data, reducing data efficiency and generalisation. In equation discovery, omitting geometry-induced operators misspecifies the candidate library, leading to biased parameters, missed governing terms and spurious terms. We propose a de-confounding framework that makes the known geometry-to-operator transformation explicit. Geometry-induced coefficient fields improve prediction and data efficiency across five operator-learning benchmarks, while geometry-complete candidate libraries recover the generating equations and reduce held-out PDE residuals by more than two orders of magnitude in both evolving-domain systems. By separating known geometric action from intrinsic physics, the proposed framework supports more reliable and data-efficient PDE learning across scientific and engineering problems with varying geometries.
[LG-153] Learning-Enabled Estimation: Tight Characterizations under Sample Selection Biases
链接: https://arxiv.org/abs/2609.38608
作者: Vikram Kher,Jane H. Lee,Anay Mehrotra,Manolis Zampetakis
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: Presented at EC 2026
Abstract:When can we learn from biased samples? We study regression when outcomes are observed only after passing through selection filters that depend on both covariates and outcomes themselves, a ubiquitous challenge spanning clinical trials with patient dropout, labor markets with self-selection, and auctions with strategic entry. Ignoring such selection yields systematically biased conclusions with real-world consequences. This challenge has a long history in econometrics and statistics, starting with Heckman’s seminal two-stage model and followed by numerous generalizations. While these works provide various sufficient conditions for identification, a complete characterization of when such regression is possible has remained elusive. In this work, we provide a characterization for when regression is possible in the presence of sample selection bias. Our results establish the minimal assumptions required on the functional forms of selection processes under which regression remains possible, which are particularly relevant in modern settings where selection mechanisms are increasingly complex and opaque. As a corollary of our characterization, we show that there are settings where the regression function can be identified even when the selection filter itself cannot. This observation already goes beyond the ``estimate selection filter, then debias regression’’ paradigm that is followed by virtually all existing approaches. Under natural strengthenings of our identification conditions, we also establish finite-sample estimation guarantees with explicit convergence rates and provide oracle-efficient algorithms. This yields the first general-purpose estimation method for this broad class of selection problems. Finally, we explore the implications of our results for several well-studied econometric settings with complex selection mechanisms such as auctions with entry costs and labor markets. Comments: Presented at EC 2026 Subjects: Machine Learning (cs.LG); Machine Learning (stat.ML) Cite as: arXiv:2609.38608 [cs.LG] (or arXiv:2609.38608v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2609.38608 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-154] JARQ: Joint Alternating Refinement for Quantization
链接: https://arxiv.org/abs/2609.38599
作者: Xinyu Wang,Sicheng Lyu,Xiao-Wen Chang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Group-wise post-training quantizers for large language models round weights onto a grid that is not refit to the resulting integer codes. We show that this leaves accuracy on the table: the best grid depends on the codes, input correlations couple the errors of different groups, and useful code changes often involve many codes at once. We propose JARQ , a plug-in refinement that starts from any group-wise quantizer and alternates a joint least-squares fit of all group scales with bounded Babai proposals that move many codes of a group together on the current grid. The problem is a bilinear box-constrained mixed-integer least-squares problem; the solver is backpropagation-free, does not increase the layer-wise objective under exact scale solves, and keeps the host’s bit width, groups, zero points, and inference cost. Across Llama-2, Llama-3, and Qwen models with RTN, GPTQ, OmniQuant, and AWQ hosts, JARQ lowers perplexity in 90 of 96 comparisons, cuts three-bit RTN perplexity by up to 36%, raises mean multiple-choice accuracy in 23 of 24 configurations, and improves QEP, QuaRot, and OJBKQ outputs, at under a minute per 7B block.
[LG-155] Reinforcement Learning with Complex (valued) Memories
链接: https://arxiv.org/abs/2609.38598
作者: Sathya Kamesh Bhethanabhotla,Efstratios Gavves,André Biedenkapp
类目: Machine Learning (cs.LG)
*备注: 19 pages, 4 figures, 9 tables Accepted at the 19th European Workshop on Reinforcement Learning 2026
Abstract:Partially observable environments pose a fundamental challenge in deep reinforcement learning, requiring agents to compress temporal information from observations and maintain a memory to make effective decisions. While there exist many approaches ranging from gated recurrence to attention mechanisms and model-based RL, the search for effective representational techniques that can capture long-term dependencies remains an active area of research. In this work we revisit Unitary recurrent networks (uRNNs) [Arjovsky et al., 2016, Jing et al., 2017], that demonstrated superior gradient flow and associative recall, expressing the recurrence and the hidden state in a complex vector space. Their norm preserving unitary dynamics enable information propagation through long sequences. To this end, we propose three different versions of uRNNs as drop-in replacements for recurrent PPO architectures, and demonstrate that the simple recurrence and the added degree of freedom from the phase of the complex representations enable significant gains over baselines on several memory-improvable tasks, including continuous control. We further explore how to preserve the phase information of the complex hidden state for a phase-aware policy by drawing a parallel to how quantum states are measured. With our methods reaching up to 2-3 \times the reward in environments like rocksample and Craftax compared to the baselines, this work points towards an exciting new direction of representations for RL and the problem of partial observability. Code is available at: this https URL
[LG-156] NeurDuo-EEG: A Long-Sequence EEG Foundation Model with Persistent State and Explicit Memory
链接: https://arxiv.org/abs/2609.38587
作者: Yifan Wang,Haiping Liu,Yang Cui,Wenhao Cai,Shuhang Li,Xiaoyang Huang,Xianyang Liu,Jingyu Sun,Yizheng Sun,Cunhang Fan,Tianming Du,Jiancheng Yang,Zhenhong Li,Yunhao Zhang,Hongpeng Zhou,Jingyuan Sun
类目: Machine Learning (cs.LG)
*备注:
Abstract:Electroencephalography (EEG) is recorded continuously over hours, with relevant dynamics spanning timescales from milliseconds to hours. Most EEG foundation models nevertheless process fixed windows independently, limiting their ability to capture information encoded in long-timescale dynamics. State-space architectures enable persistent recurrent processing, but long-range information remains implicitly compressed in recurrent states. We present NeurDuo-EEG, a causal EEG foundation model with channel-resolved persistent memory. NeurDuo-EEG introduces multi-timescale memory management with learned consolidation and selective retrieval, enabling persistent modelling of continuous EEG with fixed-size state. It is pre-trained on 3,955 hours of EEG from 17 public datasets using multichannel autoregressive prediction of discrete spectral codes. Across three short-window and two long-sequence downstream tasks, NeurDuo-EEG achieves the best performance on four of five benchmarks, including all three short-window tasks and seizure detection, where AUC-PR improves from 0.285 to 0.471 over the strongest non-NeurDuo baseline. NeurDuo-EEG also remains competitive on sleep staging and supports efficient streaming inference, with nearly constant per-chunk latency as the available history grows to one hour. Notably, the Small variant achieves this with only 4.7M backbone parameters. These results demonstrate the value of persistent, multi-timescale modelling for both long-sequence and short-window EEG analysis. Our code is available at this https URL.
[LG-157] he Advantages of Fresh Sketching for Ridge Regression
链接: https://arxiv.org/abs/2609.38565
作者: Linkai Ma,Qilin Li,Petros Drineas
类目: Machine Learning (cs.LG)
*备注:
Abstract:Over the past 25 years, sketching and sampling have become widely used tools for accelerating large-scale regression. In iterative randomized solvers, a basic design choice is whether to \textitreuse the same sketch or draw \textitfresh randomness at every step. For (under-constrained) iterative ridge regression with column sampling, whether fresh sketches offer provable advantages has remained open: \textitWe show that they do. Fresh sketching lets us analyze error only along the current residual solution, rather than uniformly over the entire Gram matrix. This directional view yields sharper convergence guarantees for leverage score and ridge leverage score sampling and, more importantly, leads to residual-aware sampling rules. By minimizing the variance of the relevant sketched matrix-vector product, we derive an oracle distribution and practical approximations to the oracle distribution, including a mixture sampling distribution with (somewhat weaker) convergence guarantees. Experiments on synthetic and real data, including ridge probes on Qwen2.5 representations, support our theory, showing substantially faster convergence.
[LG-158] When a Flatness Proxy Is Not a Function: Robustness Certificates and Training Interventions
链接: https://arxiv.org/abs/2609.38540
作者: Vicente Opazo,Jose Calatayud-Mateu,Cristobal Rojas,Cristian Buc Calderon
类目: Machine Learning (cs.LG)
*备注: 10 pages, 3 figures, 1 table
Abstract:A valid curvature upper bound need not justify either a robustness certificate or an intervention on an intrinsic predictor property. We demonstrate this distinction for a last-layer relative-flatness proxy used in both settings. First, empirical-risk stationarity does not eliminate pointwise first-order loss terms: at a finite global empirical-risk minimum, the retained certificate expression underestimates a loss increase by over 210\times . We derive a globally valid, gauge-invariant feature-space repair. Second, common-row softmax shifts preserve predictions and the exact contraction while making the proxy unbounded. Even standard reference-class choices double it on average relative to the centered representation. For a single fixed-feature example with at least three classes, scalar retuning generically cannot align the induced probability updates. Row centering gives the orbit-minimized bound and restores value and full-model gradient invariance under this symmetry. Across 45 paired one-step tests on algorithmic and image models, amplified shifts separate raw-regularized predictors while quotient-regularized predictors remain aligned. Long-horizon CIFAR-10 experiments show substantial, reversible suppression of generalization, while evidence for selective delay after memorization is less consistent. Together, these results show that validity as a curvature upper bound does not by itself justify either inversion into a robustness certificate or differentiation into an intrinsic training intervention.
[LG-159] Does Text Steer Neural PDE Surrogates? A Controlled Diagnostic with OperatorCLIP NEURIPS2026
链接: https://arxiv.org/abs/2609.38517
作者: Aadi Dash,Lennon J. Shikhman,Michael Galarnyk
类目: Machine Learning (cs.LG)
*备注: 8 pages, 2 figures, 2 tables. Accepted to NeurIPS 2026 Workshop on Representation for the Physical Sciences
Abstract:Lower error from a text-conditioned neural surrogate does not, by itself, show that the model uses the meaning of the text. We examine this attribution problem with OperatorCLIP, comparing an unconditioned FNO, a constant-sentence FiLM control, and a fixed task description trained with contrastive alignment. Three-seed experiments cover Darcy2D, ShallowWater2D, and three-dimensional compressible Navier-Stokes (CNS3D). Constant conditioning has lower mean test error on both 2D tasks. Relative to this control, task text plus alignment has a similar mean on ShallowWater2D and CNS3D and a higher mean on Darcy2D; these descriptive comparisons have substantial seed uncertainty. The latter comparison changes both prompt content and loss, so it isolates neither effect. The text encoder is trained from scratch, and each conditioned model sees only one description during training. In this regime, pairwise InfoNCE cannot identify matched pairs and has minimum \log B . Prompt interventions show no reliable semantic ordering. This methodological caution demonstrates why pathway controls are needed; it neither establishes semantic competence of the encoder nor tests the effectiveness of text under varying physical context.
[LG-160] Autoregressive Frontier Expansion: Growing Trees with Graph Machine Learning
链接: https://arxiv.org/abs/2609.38506
作者: Umer Gupta,Saku Peltonen,Martin Ritzert
类目: Machine Learning (cs.LG)
*备注:
Abstract:Tree-like branching structures are common in nature, from botanical trees to neurons, blood vessels and respiratory trees. Their branching shape often reflects function, making structural modelling central to understanding how these systems work. Because acquiring real-world 3D data is often expensive or infeasible, realistic generative models are valuable for simulation and data augmentation. Existing morphology-specific models either constrain how topology is generated or rely on hand-tuned, mechanistic procedures. Generic 3D graph generators, by contrast, do not exploit or enforce the structure of trees. We propose Autoregressive Frontier Expansion, a generative framework that constructs trees through an iterative expansion process, simulating the biological growth of real trees. At each step, a flow-matching model parameterised by an SO(2)-equivariant GNN expands the frontier by predicting whether each active branch bifurcates or terminates. We evaluate our method on cortical neurons and botanical trees in unconditional, class-conditioned, and morphology-guided generation. Across both domains, the generated morphologies agree closely with the reference distributions and, in conditional experiments, with the specified targets.
[LG-161] Role-guided Speaker Deletion Verification in Clinical Psychiatry Speech Recordings with Audio Language Models
链接: https://arxiv.org/abs/2609.38491
作者: Joseph T Colonel,Daniel Katzman,Kelsey Kirker,Adam N Davidson,Shalaila S Haas,Cheryl Corcoran,René S Kahn,Guillermo Checci,Baihan Lin
类目: Machine Learning (cs.LG); Sound (cs.SD)
*备注:
Abstract:Clinical research in psychiatry increasingly relies on large scale collection of spoken language data to identify acoustic and linguistic biomarkers. Yet evolving consent and protocol requirements can oblige investigators to remove a designated speaker from multi-speaker recordings and to verify said removal at a scale infeasible for manual review of entire corpora. We study this verification problem for role-driven dyadic clinical dialogue in psychiatry and investigate it with two parallel, symmetric pipelines: confirming that clinician speech has been removed from psychiatric interview recordings, and confirming that patient speech has been removed from the same recordings. Each pipeline redacts the raw audio for its target role and then scans the surviving output with audio-language and large-language models to identify missed deletions. We evaluate this approach on a corpus of 48 dyadic recordings drawn from psychiatry settings, testing four open-weight models in an inference-only setting: Gemma-4-12B, Gemma-4-31B, Nemotron-3-Nano, and Nemotron-3-Nano-Omni. A disjunctive OR ensemble over fourteen model-view configurations had a combined F1 of 0.478 (precision 0.330, recall 0.870), an improvement over individual model estimates driven by recall gains that point to substantial complementarity across models and context views.
[LG-162] RetroGEF: Dynamic Graph Edit Flow for Single-Step Retrosynthesis
链接: https://arxiv.org/abs/2609.38484
作者: Xiaozhuang Song,Xuemin Chen,Xinjian Zhao,Yaoyao Xu,Tianshu Yu
类目: Machine Learning (cs.LG); Quantitative Methods (q-bio.QM)
*备注: 25 pages, 10 figures
Abstract:Retrosynthesis enables the discovery of viable synthetic routes to target molecules. It plays a central role in modern drug discovery and materials design. Retrosynthesis involves molecular graph transformations that can change both connectivity and graph size. These transformations may introduce reactant components absent from the target while revising the product-derived structure. To model these transformations, we propose RetroGEF, a flow-based generative model for single-step retrosynthesis. Starting from the target molecule, it constructs possible reactants by adding atoms and changing bonds in the molecular graph. RetroGEF models molecular transformations and changes in graph size within the same generative process, rather than relying on a fixed-size graph canvas. It learns this process directly from product–reactant pairs without requiring a prescribed edit order. Experiments on representative retrosynthesis benchmarks demonstrate that RetroGEF achieves state-of-the-art performance.
[LG-163] Diffusion-2BC: Hybrid Diffusion and Regression Training for Offline Behavior Cloning in Autonomous Driving
链接: https://arxiv.org/abs/2609.38472
作者: Bruno Maciel Machado,Eric Aislan Antonelo
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注:
Abstract:Behavior cloning provides an offline route to autonomous-driving policy learning, but mean-squared-error regression is poorly matched to demonstrations in which one observation admits several valid actions. Diffusion policies can represent conditional multimodal action distributions, yet their closed-loop performance may be unstable when visual features and control are learned from limited data. This paper presents Diffusion-2BC, which combines a diffusion denoising objective with an auxiliary deterministic behavior-cloning loss over a shared visual encoder. The auxiliary branch is used only during training; inference remains diffusion-based. The proposed method is evaluated in the controlled Claw environment and in bird’s-eye-view CARLA navigation, including route-conditioned driving, route-free navigation through multiple intersections, and cross-map evaluation from Town01 to Town02. In the Claw task, Diffusion-2BC reduced the mean mask-distance error by approximately 10% relative to a diffusion-based behavior-cloning baseline and by 85% relative to standard deterministic behavior cloning. In route-free CARLA, Diffusion-2BC traveled substantially farther before termination under the evaluation protocol than both baselines in Town01 and Town02. Additional qualitative rollouts revealed distinct route choices, showing the multimodal behavior of the proposed diffusion-based agent. The results indicate that an auxiliary regression signal can improve the closed-loop reliability of diffusion behavior cloning while preserving multimodal prediction in the controlled benchmark.
[LG-164] Grokking through the Lens of Minimum-Norm Interpolation
链接: https://arxiv.org/abs/2609.38453
作者: Gil Kur,Ileana Rugina,Clémentine Carla Juliette Dominé,Marco Mondelli
类目: Machine Learning (cs.LG); Statistics Theory (math.ST); Machine Learning (stat.ML)
*备注: Submitted
Abstract:Grokking shows that fitting the training data and learning the underlying signal can occur at very different stages. However, existing theories offer limited quantitative insight into how this delayed generalization depends on inductive bias and signal structure. Our work addresses the gap by developing a statistical theory that characterizes how regularization geometry and signal sparsity govern generalization near interpolation. In particular, we focus on the prototypical setting of high-dimensional regression and identify regimes in which sparsity-promoting regularization makes exact interpolation much more accurate than approximate fitting. In strongly overparameterized noiseless problems, we prove a zero–one generalization law and construct a family of convex norms whose interpolators transition from the trivial risk of the all-zero predictor to exact recovery, while keeping the training error equal to 0 . Furthermore, when feature dimension and sample size are proportional, we provide a precise characterization of training and generalization errors along \ell_r -regularization paths. This in turn allows us to quantify the generalization gain that remains near interpolation: we show that this gain increases as the norm becomes more sparsity-promoting and as the target becomes sparser, with a sharp drop in generalization reached for noiseless data and \ell_1 regularization. Experiments on diagonal linear networks and transformers trained on modular arithmetic demonstrate the generality of our theoretical predictions. Finally, beyond grokking, our work reveals a statistical instability in minimum-norm interpolation: small perturbations in the regularization strength can lead to drastically different generalization, while preserving small training error.
[LG-165] An Input-Frugal Deep Learning Framework for Weather-Driven National Crop-Yield Forecasting: A Case Study of Brazilian Soybean
链接: https://arxiv.org/abs/2609.38447
作者: Fernando Dupin da Cunha Mello(1),Prashant Kumar(2, 3, 4),Erick G. Sperandio Nascimento(1, 2, 3, 4) ((1) Stricto Sensu Department, SENAI CIMATEC University, Salvador, Bahia, Brazil, (2) Global Centre for Clean Air Research (GCARE), School of Engineering, Civil and Environmental Engineering, Faculty of Engineering and Physical Sciences, University of Surrey, Guildford GU2 7XH, United Kingdom, (3) Institute for Sustainability, University of Surrey, Guildford, GU2 7XH, United Kingdom, (4) Surrey Institute for People-Centred AI, Faculty of Engineering and Physical Sciences, University of Surrey, Guildford GU2 7XH, UK)
类目: Machine Learning (cs.LG); Applications (stat.AP)
*备注:
Abstract:Reliable, timely crop-yield forecasts are essential for market stability and risk management, yet many approaches rely on costly or hard-to-scale inputs. We present a frugal, transferable, and architecture-agnostic deep learning framework that uses routine weather as the only time-varying input plus two lightweight static context inputs (crop year and an agro-environmental label) to capture long-run change and regional heterogeneity, while supporting multiple sequence encoders under identical data requirements. Using a 20-season Brazilian soybean case study (2001/02-2020/21) with leave-one-year-out cross-validation, we benchmark MLP, CNN, LSTM, CNN-LSTM, a Transformer encoder and the Mamba state-space model against linear ridge regression and a five-year moving-average “farmer” baseline. All deep learning variants outperform ridge, and all sequential encoders surpass the non-sequential MLP. The Transformer achieves the best national accuracy (RMSE 149 kg ha^-1; rRMSE 5.3%; R^2 = 0.784), reducing error by 47.6% relative to the farmer baseline. In-season forecasts improve monotonically from early- to late-season issuance, reaching approximately 50% lower error than the baseline at the latest forecast point. Ablations indicate that the agro-environmental label and spatial instance expansion (multiple grid-node weather sequences per municipality-year) contribute positively without increasing input complexity. SHAP diagnostics suggest crop year explains most of the long-run trajectory, whereas within-season weather and agro-environmental context primarily drive interannual deviations, with moisture/cloud and thermal-demand variables dominating. Overall, the framework is straightforward to deploy across other crops and geographic regions and is naturally compatible with operational weather forecasts for routine monitoring.
[LG-166] Graph Anomaly Detection as Finite-Horizon Control: Training-Free Scoring via Empirical Bayes NEURIPS
链接: https://arxiv.org/abs/2609.38424
作者: Fred Xu,Thomas Markovich,Florence Regol,Yizhou Sun
类目: Machine Learning (cs.LG)
*备注: Paper already accepted at Neurips
Abstract:Node-level graph anomaly detection (GAD) identifies nodes whose attributes and interactions deviate from dominant graph regularities. Existing GAD models encode normality and anomaly scoring indirectly through architectures, message passing, reconstruction or contrastive objectives, and tuned score families. This entangles graph trust (how strongly graph structure should define normality), graph-spectral weighting, and anomaly-score choice, yielding scores that are costly, opaque, and unstable across graph regimes. We propose EB-GAD (Empirical-Bayes GAD), a training-free framework that models normality as graph-aware generalized Ornstein-Uhlenbeck (GOU) relaxation toward a graph-filtered template. Empirical Bayes fits the graph precision from the residual-field likelihood; the GOU then turns scoring into a closed-form finite-horizon control energy, the minimum effort to steer a feature-neutral node to its observed endpoint along graph-spectral relaxation. Sweeping relaxation horizon and endpoint tolerance yields a bank of scores that share one fitted prior: equilibrium Mahalanobis scoring is one limit, while finite-horizon control-energy and scale-normalized ratio scores reveal anomalies that static equilibrium scoring can mask. A label-free selector chooses the score family from feature homophily, edge density, and feature dimension, then ranks candidates by fitted-null deviation and rank stability. On 11 benchmarks and without labels at any step, EB-GAD has the best or tied-best AUROC on 9: the four financial fraud networks (up to 3.7M nodes), the YelpChi and Amazon review graphs, Weibo, Reddit and Facebook, with margins of up to 21.7 points. It is second on BlogCatalog and ACM.
[LG-167] Synthesis Without Training: An Inference-Only Pipeline for Tabular Temporal and Relational Synthetic Data NEURIPS2026
链接: https://arxiv.org/abs/2609.38414
作者: Zilong Zhao,Abdul Raheem,Jiayu Li,Sohei Arisaka,Darius Lim Hong Yi,Milad Abdollahzadeh,Uzair Javaid,Biplab Sikdar
类目: Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2026 workshop: Beyond Private Training: The New Landscape of AI Privacy
Abstract:Synthetic data generation is dominated by the fit-then-sample paradigm: a generative model is trained on a private dataset and then sampled from. Despite its widespread adoption, this paradigm faces three challenges: (1) a new training run is required for every dataset; (2) different data modalities, such as single tables, time series, and relational databases, require task-specific models and feature engineering; and (3) the resulting model is opaque, making its behavior under data constraints difficult to inspect. We propose GENSCRIPT, an inference-only pipeline that eliminates model training. GENSCRIPT computes a deterministic statistical profile of the source data (column types, ranges, missingness, categories, correlations, etc.) and passes it–rather than raw rows–to a language model to infer field semantics and cross-column integrity constraints. A coding agent then compiles the profile and constraints into an executable, auditable sampler. This unified approach supports single-table, temporal, and relational data without task-specific modeling. Across four single-table benchmarks, GENSCRIPT builds generators in 2 minutes and samples 50k rows within 6 seconds, while remaining within a few points of leading methods in marginal fidelity. Notably, it is the only method that perfectly preserves a 1-to-1 mapping between columns in the Adult dataset. On a smart-building dataset, it produces conditional time series that more closely match the real distribution than two baselines and perfectly preserves primary- and foreign-key relationships in the corresponding relational database.
[LG-168] Disagreement-Regularized Imitation Learning for Image-Based Continuous Control with Gaussian and Beta Policies
链接: https://arxiv.org/abs/2609.38407
作者: Irving Giovani Bronzatti Petrazzini,Eric Aislan Antonelo
类目: Machine Learning (cs.LG)
*备注:
Abstract:Purpose: Behavior cloning can accumulate errors when a learned controller visits states outside the demonstrated distribution. This study evaluates whether Disagreement-Regularized Imitation Learning (DRIL), which converts disagreement among cloned policies into a reinforcement-learning reward, improves image-based continuous control. Methods: A controlled CarRacing study combines Gaussian and Beta learner policies, demonstrations from either a clipped Gaussian expert or an intrinsically bounded Beta expert, one or 20 trajectories, deterministic and stochastic evaluation, and three retained stages: behavior cloning, the highest 10-episode training-score checkpoint, and the final DRIL checkpoint. The disagreement ensemble contains five Gaussian policies in every variant. Each retained policy is evaluated over 100 procedurally generated episodes. Results: Score-selected DRIL produced its largest gains in the few-demonstration setting, improving over the strongest behavior-cloning mean by 61% with clipped-action demonstrations and by 112% with bounded-action demonstrations. With 20 trajectories, the advantage of DRIL narrowed; in the bounded-action regime, Beta behavior cloning remained about 7% above the best DRIL checkpoint. The experiments also show that the informativeness of the disagreement reward changes with the learner representation and training stage. Conclusion: DRIL can substantially improve few-demonstration visual continuous control, while bounded Beta policies provide strong behavior-cloning performance when more demonstrations are available. The results highlight the joint importance of learner support,ensemble response, and checkpoint selection.
[LG-169] From Codebase to Culprit (C2C): Reducing the Search Space for Bugs with Semantic Retrieval and Hierarchical Reinforcement Learning
链接: https://arxiv.org/abs/2609.38402
作者: Ankur Garg,Corey Yang-Smith,Rishav Rishav,Ahmad Abdellatif,Samira Ebrahimi Kahou
类目: oftware Engineering (cs.SE); Machine Learning (cs.LG)
*备注:
Abstract:We introduce C2C (From Codebase to Culprit), a framework for precise bug localization that progressively reduces the debugging search space across multiple levels of granularity: files, functions, and lines of code. To mirror developer’s natural top-down debugging workflows, C2C integrates semantic retrieval and Hierarchical Reinforcement Learning (HRL) in a two-stage process. First, it performs recall-oriented retrieval of buggy candidates via semantic vector similarity search using bug-report text, including available stack-trace information, against a database of embeddings, where the embeddings are fine-tuned via contrastive learning with CodeBERT. Building on this reduced search space, the HRL framework incrementally localizes bugs, reasoning from files to functions and ultimately to individual lines of code. Unlike prior approaches which operate at a single granularity, C2C enables multi-resolution localization while maintaining contextual consistency across decisions. Experiments on real-world Java and Python datasets demonstrate that C2C improves retrieval precision and localization accuracy. Ablation studies further highlight the contributions of hierarchical decomposition, structured learning signals, and reward shaping in advancing multi-level bug localization.
[LG-170] LeanPolish: Verified Supervision for Lean Proof Compression
链接: https://arxiv.org/abs/2609.38384
作者: Pauline Bourigault
类目: Machine Learning (cs.LG)
*备注:
Abstract:Verified proof edits offer a natural source of supervision for improving language-model-generated Lean proofs. Yet verification establishes that an edit is correct, not that its training signal is free of search artifacts. We introduce LeanPolish, a symbolic Lean 4 pipeline that releases 33,402 accepted local edits and 65,596 same-state failed attempts, and use it to study what models learn from this supervision. First-success search admits a goal-independent rule with perfect ranking accuracy; teacher-selected evaluation sites also reward trivial deletions. Continuing menu evaluation beyond the first success removes the ordering shortcut: a trained ranker selects the best candidate on 70.1% of evaluated held-out states, versus 36.9% for the strongest frozen baseline. For compression, iterating the symbolic pass raises miniF2F savings from 19.7% to 27.5%, exceeding the neural hybrids we test there. Verified neural editing helps on other proof sources, but matched frozen-model controls show that its gains need not come from training. The supervision does improve whole-proof rewriting: fine-tuning raises verified token reduction from 2.8% to 5.5% on 19 PutnamBench proofs. Together, the released edits, complete candidate pools, and controlled evaluations separate learning to imitate a search policy from improving on that search. They provide a reproducible basis for studying proof improvement while keeping correctness, compression, and edit policy distinct.
[LG-171] Learning to Plan from Random Exploration
链接: https://arxiv.org/abs/2609.38383
作者: Deqian Kong,Guangyan Sun,Sheng Cheng,Sirui Xie,Bo Pang,Jianwen Xie,Tony Geng,Caiwen Ding,Ying Nian Wu
类目: Machine Learning (cs.LG); Robotics (cs.RO); Machine Learning (stat.ML)
*备注:
Abstract:Random exploration reveals how an environment can be traversed before a goal is specified. Can this experience support long-range planning without policy-improvement training? Our random-walk analysis explains what temporal relations contain: short horizons reveal geodesic geometry in the diffusion limit, while longer horizons reveal connectivity between regions before mixing removes these distinctions. We learn these relations with a conditional energy-based model that estimates temporal log-density ratios through horizon-conditioned embeddings. The model is trained on observation pairs by noise-contrastive estimation, without action or reward labels. The planner queries these learned relations at different horizons as it moves toward the goal. At test time, a separate local dynamics model predicts candidate action outcomes, and the temporal model evaluates their progress toward the goal by selecting or aggregating estimated improvements across horizons. The agent executes one action and replans with both models fixed. Experiments demonstrate long-range maze planning from random exploration using states and images. Learned score fields, embedding probes, and planned routes exhibit properties of a multiscale cognitive map. We further demonstrate egocentric navigation from random exploration and manipulation planning from suboptimal data.
[LG-172] Continual Learning of Dynamical Systems in Recurrent Neural Networks through Recyclable Unit Gating
链接: https://arxiv.org/abs/2609.38356
作者: Sima Hashemi,Daniel Durstewitz,Georgia Koppe
类目: Machine Learning (cs.LG)
*备注: 34 pages (including Appendix), 15 Tables, 7 Figures
Abstract:Dynamical Systems Reconstruction (DSR) aims to infer models from observed time series that reproduce a system’s qualitative long-term behavior. Continual DSR (cDSR) requires learning new systems while preserving previously learned dynamics, yet even small parameter updates in recurrent models can qualitatively alter their behavior over long autonomous rollouts. We benchmark established continual learning (CL) methods spanning parameter regularization, replay, and parameter isolation on the fully trainable and interpretable Almost-Linear RNN (AL-RNN). Parameter isolation preserves earlier dynamics most effectively, but excessive task-specific allocations can rapidly exhaust a fixed-size network. We therefore introduce Continually-Recyclable Unit-Gating (CRUG), which conserves capacity through compact allocation and forward transfer. Differentiable gates trained with an L_0 -based penalty select task-specific units, while unused units are recycled for subsequent tasks. Directed connections allow later tasks to reuse earlier representations without affecting the dynamics of previously committed units. CRUG achieves the strongest reconstruction–capacity trade-off among the tested methods with zero forgetting and reliably learns a heterogeneous sequence of nonlinear and chaotic systems. Furthermore, we show that forward transfer is more pronounced and useful when tasks share similar underlying dynamics. Lastly, we demonstrate that CRUG’s advantages extend beyond autonomous cDSR to sequential cognitive tasks.
[LG-173] Function-Space Transformer with Adaptive Anchors
链接: https://arxiv.org/abs/2609.38348
作者: Guorui Sang,Pedram Rooshenas
类目: Machine Learning (cs.LG)
*备注:
Abstract:Many forms of data, including physical fields, geometric shapes, and visual signals, are naturally described by functions over continuous domains but are observed through discrete samples. Representing these functions on fixed uniform grids imposes a trade-off between resolving localized variation and increasing computation across the domain. Neural operators address this mismatch by learning mappings between functions, while latent-attention architectures provide flexible processing of sampled observations. We introduce the Function-Space Transformer (FST), a framework for learning from functions through a spatially adaptive continuous latent representation. FST stores features at anchors whose locations are predicted from the input observations and recursively refines these anchor features through function-space interactions. This allows the representation to adapt its spatial organization to each input rather than inherit that of the observation grid, while supporting both spatially resolved and finite-dimensional outputs. On PDE solution prediction using PDEBench Burgers and Darcy flow, FST substantially outperforms the Perceiver IO baseline, whose latent representation lacks explicit spatial organization, and is highly competitive with the Fourier Neural Operator. On ImageNet-1K, FST achieves higher classification accuracy than the Vision Transformer baseline, with fewer parameters across these comparisons. Ablations further support the benefits of function-space updates and recursive refinement. Together, these results highlight the potential of adaptive continuous representations for both scientific prediction and visual recognition.
[LG-174] Activation-Conditioned Self-Distillation
链接: https://arxiv.org/abs/2609.38342
作者: Zhexi Lu,Subhajit Chaudhury,Tejaswini Pedapati,Keerthiram Murugesan,Lei Yu
类目: Machine Learning (cs.LG)
*备注:
Abstract:On-policy self-distillation uses a model as its own teacher to provide dense supervision for reasoning, often through reference-solution conditioning. Providing privileged information does not by itself ensure effective token-level supervision throughout long responses. We introduce Activation-Conditioned Self-Distillation (ACSD), which extracts a steering vector by contrasting activations of self-generated trajectories that reach verified correct answers within a generation budget with those of all remaining trajectories. A frozen copy of the base model applies this vector at each prediction position, and the student learns from its next-token distributions on student-generated prefixes. Outcome verification is used for direction construction and calibration; distillation requires neither problem-specific reference text nor teacher parameter updates. The distilled student is used alone at inference. On each of five models, ACSD achieves the highest mean accuracy over four mathematical benchmarks among the evaluated methods. On DeepSeek-R1-0528-Qwen3-8B, mean mathematical accuracy reaches 71.9% and LiveCodeBench v6 pass@12 reaches 70.9%, compared with 69.0% and 66.3% for the reference-conditioned OPSD baseline. Contrasts among correct trajectories also support distillation, and extracted directions can be reused across mathematical training datasets. On fixed student trajectories, ACSD maintains more stable late-position logit-update magnitudes than OPSD.
[LG-175] Kinematic signatures of impairment: Detecting alcohol intoxication in e-scooter riders using sensor data and machine learning
链接: https://arxiv.org/abs/2609.38276
作者: Rahul Rajendra Pai,Marco Dozza,Alexander Rasch,Ali Mohammadi,Marco Capuccini
类目: Machine Learning (cs.LG); Signal Processing (eess.SP); Applications (stat.AP)
*备注:
Abstract:Alcohol intoxication is a leading contributor to fatal and severe-injured e-scooterist crashes. Current countermeasures, such as temporal restrictions or pre-ride cognitive screening, cannot continuously assess an e-scooterist’s physical motor control or impairment in real time. We conducted a controlled experiment in which 25 participants rode an instrumented e-scooter through a test track while sober and at two targeted blood alcohol concentration levels (0.05% and 0.08%). The e-scooter was instrumented with a six-axis inertial measurement unit (IMU), and throttle and brake lever position sensors, all sampled at 100 Hz. Two complementary signal features were computed: normalised permutation entropy, which quantifies temporal complexity, and standard deviation, which quantifies signal amplitude. Repeated measures correlation identified seven kinematic features (all IMU and throttle signals) whose entropy decreased (p 0.001) while standard deviation increased (p 0.01) with increasing intoxication, indicating that intoxicated riders shift from continuous, low-amplitude micro-corrections to fewer, high-amplitude reactive corrections. An entropy based multi-class logistic regression classifier, evaluated through leave-one-participant-out cross-validation, achieved 85% overall accuracy and a weighted one-vs-rest area under the receiver operating characteristic curve (AuROC) of 0.94, with a sober-vs-high AuROC of 1.00. Steering rate and lateral acceleration were the most important predictive features, indicating that alcohol induces a distinct collapse in lateral equilibrium during riding. Ultimately, these results demonstrate that onboard kinematic sensing combined with entropy-based signal analysis can reliably distinguish sober from intoxicated e-scooter riding, providing a foundation for automatic intoxication detection systems that preserve mobility for sober riders.
[LG-176] Inference-Layer Security: Defending Against Adversarial Inference and Infrastructure Abuse
链接: https://arxiv.org/abs/2609.38239
作者: Keifer Lee
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: A Technical Report
Abstract:A Technical Report: Operating a large language model (LLM) as a service requires more than inference infrastructure: the provider must also defend against adversarial interactions that seek to exploit the service, including jailbreaking for harmful use, sophisticated denial of service, and distillation attacks. We study this problem at the inference layer, using a hypothetical frontier lab, Five Elements Inc., as a running example. Because no public labelled dataset of adversarial LLM usage exists, we introduce a structural causal model (SCM) that generates a realistically grounded, labelled dataset of user-sessions, with coordinated multi-account campaigns, platform feedback, and three tiers of label observability. On this dataset we train a practical gradient-boosted detector that classifies each user-session as benign or malicious and, if malicious, by attack type. Against oracle labels the detector very nearly solves the binary task (AUPRC 0.993 ), yet against the operational labels a real Trust Safety team would hold, the same model scores an AUPRC of only 0.313 : the detector is more accurate than the labels used to evaluate it. For attack-type attribution, a naive argmax is dominated by the 98% benign prior (macro-F1 0.295 ), whereas a simple thresholded decision engine raises macro-F1 to 0.489 without sacrificing accuracy. The dataset is publicly released.
[LG-177] FlashDiffusion: Fused Tiled Kernel Spectral Decomposition
链接: https://arxiv.org/abs/2609.38198
作者: Julio Candanedo
类目: Machine Learning (cs.LG)
*备注:
Abstract:Diffusion maps, and kernel methods more generally, provide an interpretable nonlinear spectral representation basis for geometric learning. In the geometric limit, small bandwidth, these matrices tend to be high rank and thus require materializing dense Gaussian kernels requires O(N^2) memory. We introduce FlashDiffusion, a matrix-free method that evaluates dense Gaussian kernel blocks in fused GPU tiles and couples the eigensolver to an empirical \beta -flow that selects the finite-sample resolution scale. A continuation over sample size and bandwidth warm-starts increasingly expensive spectral solves from coarser resolutions.
[LG-178] A Data-Free Physics-Informed Neural Operator for Level-Set Interface Advection
链接: https://arxiv.org/abs/2609.38195
作者: Muhammad Akbar Khan
类目: Machine Learning (cs.LG); Fluid Dynamics (physics.flu-dyn)
*备注: 27 pages, 8 figures, 6 tables
Abstract:Operators for interfacial problems are trained on reference solutions produced by the solver they are intended to replace. This work develops a data-free physics-informed neural operator for level-set interface advection, in which the interface is the equation’s unknown and the operator maps an initial interface to the full spatiotemporal trajectory under a prescribed flow. Training uses only the transport residual and a geometric constraint; no reference solution enters the objective at any point. A spacetime Fourier backbone emits the entire trajectory in one pass, and the initial condition is imposed by construction rather than by penalty, which removes the competition between the anchoring term and the residual that otherwise arises when no solution data are available. Supervised and hybrid operators are trained under an identical architecture, family, budget and test set, and are reported throughout as baselines that quantify what refusing labels costs. On a reversed single vortex the data-free operator reaches 1.614 +/- 0.067% relative L2 error on 100 held-out initial interfaces against 0.369 +/- 0.035% for the supervised baseline, a factor of 4.4; on solid-body rotation the corresponding figures are 3.804 +/- 1.075% and 2.576 +/- 0.159%, a factor of 1.5. The two benchmarks rank the arms differently, and the difference is attributable to the eikonal constraint: the exact solution violates |grad phi| = 1 over 0.3% of the domain under rotation and 86.9% under the vortex. Where the constraint is valid the physics-trained operator conserves enclosed area 2.7 times better than the supervised baseline despite a larger field error, and a hybrid arm using eight reference solutions outperforms a supervised arm using sixteen.
[LG-179] DynaFlow: Transparent and Flexible Intra-Device Parallelism via Programmable Operator Scheduling
链接: https://arxiv.org/abs/2605.21603
作者: Yi Pan,Yile Gu,Jinbin Luo,Yibo Wu,Ziren Wang,Hongtao Zhang,Ziyi Xu,Shengkai Lin,Baris Kasikci,Stephanie Wang
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: 16 pages, 15 figures, to appear in the 9th Annual Conference on Machine Learning and Systems (MLSys 2026)
Abstract:Intra-device parallelism addresses resource under-utilization in ML inference and training by overlapping the execution of operators with different resource usage. However, its wide adoption is hindered by a fundamental conflict with the static, sequential programming model of existing frameworks. Integrating these strategies requires invasive, model-specific code overhauls, representing an intractable engineering cost. This is further amplified by the high sensitivity of strategies to execution contexts (e.g., workload, model architecture, hardware), forcing developers to implement and maintain multiple specialized solutions. To address this, we propose DynaFlow, a framework that enables the transparent and flexible integration of intra-device parallelism by decoupling the logical model definition from the physical execution schedule. DynaFlow introduces a flexible frontend with annotations for graph partitioning and a programmable interface for defining custom intra-device parallelism strategies. Its efficient backend manages complex control/data-flow asynchronously, uses custom memory management to eliminate copy overheads, and preserves compatibility with optimizations like CUDA Graphs and TorchInductor. We demonstrate that DynaFlow can integrate representative parallelism strategies into 6 state-of-the-art ML systems with minimal code changes, achieving up to a 1.29x throughput improvement. DynaFlow is publicly available at this https URL.
[LG-180] Less is more: error-distance scaling relation for data-efficient kilometer-scale downscaling of extreme heat
链接: https://arxiv.org/abs/2609.40140
作者: Ahmed Marey,Henry Lu,Abhishek Gaur,Liangzhu Leon Wang,Sherif Goubran,Malek Aloui,Theodore Potsis,Alex Hernandez-Garcia,David Rolnick
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Machine Learning (cs.LG)
*备注:
Abstract:Extreme heat is where urban adaptation needs kilometer-scale data the most, but the simulations training a downscaler can cost more than they save, and how much is needed has not been identified. We measured it with CASPER, a U-Net with a structure-preserving loss downscaling 32 km reanalysis to 1 km temperature, humidity and wind, across 24 configurations of one to eight months. Held-out error grows linearly with climatological distance to the training data, RMSE = 0.83 + 2.95 d, explaining 90% of its variance against 7% for volume and predicting unseen months in advance. On held-out extreme summer weeks CASPER preserves the fine-scale structure and cross-variable physics that matched-budget baselines degrade, and matches station observations during documented heat waves to within 1.8 K. Transfer to a new region degrades geographically; 11 days of local simulation cuts Vancouver’s held-out error from 3.8 to 1.3 K. Training periods should span the target climate: the same accuracy for four times less simulation, putting kilometer-scale downscaling of extreme heat within reach of groups without large computing facilities.
[LG-181] Proximal Balancing for Causal Effect Estimation under Unmeasured Confounding
链接: https://arxiv.org/abs/2609.40051
作者: Yonghan Jung
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Methodology (stat.ME)
*备注: 50 pages. Code: this https URL
Abstract:Estimating causal effects from observational data is central to science and policy, but the effects are not identified when confounders are unmeasured. Proximal causal inference addresses this problem with proxies of the unmeasured confounders. However, existing proxy-based approaches either designate proxy roles and solve an inverse problem, which is ill-posed and hard to estimate with high-dimensional proxies, or use a latent-variable model, which assumes that the learned latent variable matches the hidden confounder and leaves bias when it does not. To address these challenges, we introduce proximal balancing. It carries the classical idea of covariate balancing to confounders that are observed only through proxies: it learns a low-dimensional summary of the covariates and proxies that makes the treatment groups comparable, and then adjusts for this summary. It needs no designated proxy roles, inverse problem, or latent model. We give identification theory, finite-sample guarantees, and a practical algorithm, PROBE. We demonstrate the method on low-dimensional, high-dimensional, and image proxies and on real-world data.
[LG-182] Amortized Bayesian Inference on Multilevel Models of Arbitrary Structure
链接: https://arxiv.org/abs/2609.40024
作者: Daniel Habermann,Andreas Bulling,Stefan T. Radev,Paul-Christian Bürkner
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Computation (stat.CO)
*备注: 16 pages, 3 figures
Abstract:We develop a general method for amortized Bayesian inference on multilevel models of arbitrary structure. Given a generative model specified as a directed acyclic graph, our method automatically derives valid factorizations of the joint posterior and matching neural network architectures. The key steps, graph expansion and graph inversion, yield an inverse graph that determines how inference networks are stacked and conditioned, producing factorizations that amortize over the number of groups and the number of observations within each group. Unlike approaches that simplify the dependency structure to speed up learning or inference, our method preserves all conditional independence and exchangeability assumptions of the generative model. Across three case studies, it closely matches gold-standard samplers on models with more than 6,500 parameters while reducing inference to a near-instant forward pass once trained.
[LG-183] PINNing the pion: conformal deep learning for F_π(s) and the (g-2)_μ hadronic contribution
链接: https://arxiv.org/abs/2609.40008
作者: Mayank Goel,Subhadip Mitra,Monalisa Patra
类目: High Energy Physics - Phenomenology (hep-ph); Machine Learning (cs.LG); High Energy Physics - Experiment (hep-ex)
*备注: 24 pages, 16 figures
Abstract:Extracting the pion electromagnetic form factor F_\pi(s) through phenomenological curve-fitting models introduces model dependence, unphysical artefacts, and kinematic inconsistencies. We introduce a Physics-Informed Neural Network (PINN) embedded in a conformal z -plane that constructs F_\pi(s) directly from first principles across spacelike and timelike domains: charge normalisation and Schwarz reflection are enforced by construction, while Cauchy-Riemann analyticity, dispersion relations, Watson’s theorem, and perturbative QCD asymptotics enter through the loss functional. Thus, the fundamental S-matrix principles dictate the form factor’s behaviour while data act as constraints. Mapping the cut complex plane onto the unit disk bounds the Hessian norm and prevents Neural Tangent Kernel spectral starvation, two known failure modes of deep-learning optimisation. Besides e^+e^- scattering data, we also incorporate \tau -decay data through a switch that isolates the pure isovector form factor natively, bypassing model-dependent isospin-breaking pre-corrections. The network organically yields an interior zero-free form factor, while the framework tests experimental tensions around the \rho(770) peak against analyticity and dispersion constraints. We obtain model-independent estimates of the pion charge radius, \langle r_\pi^2 \rangle = 0.435 \pm 0.008_\textstat \pm 0.007_\textcali fm ^2 , the second-sheet pole parameters, m_\rho^\textpole = 761.72\pm 1.04 MeV and \Gamma_\rho^\textpole = 135.99 \pm 1.20 MeV, and the two-pion contribution to the muon anomalous magnetic moment, a_\mu^\pi\pi = (506.48 \pm 2.02_\textstat \pm 1.70_\textcali) \times 10^-10 .
[LG-184] Estimation of the Label-Noise Transition Matrix with Performance Guarantees via Selective Classification NEURIPS2026
链接: https://arxiv.org/abs/2609.39829
作者: Xabier de Juan,Santiago Mazuelas,Yilun Zhu,Clayton Scott
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: Accepted at NeurIPS 2026
Abstract:Modern machine learning depends heavily on massive datasets, but obtaining high-quality annotations at scale is often expensive. As a result, learning from noisily-labeled data has become common, making accurate estimation of the label-noise transition matrix crucial. However, existing transition matrix estimators rely on the fragile estimation of class-posteriors and do not provide finite-sample performance guarantees. In this work, we propose a novel methodology to estimate the transition matrix based on one-sided selective classification. This approach bypasses class-posterior estimation, provides finite-sample performance guarantees, and leverages flexible learning methods for binary classification. Moreover, we introduce effective algorithms to implement the proposed methodology and provide their refined finite-sample performance bounds.
[LG-185] MADGRAV: a multilevel anomaly-detection pipeline for gravitational-wave searches applied to LIGO data
链接: https://arxiv.org/abs/2609.39583
作者: Gianluca Inguglia,Huw Haigh,Ulyana Dupletsa,Alessandro Longo
类目: General Relativity and Quantum Cosmology (gr-qc); Instrumentation and Methods for Astrophysics (astro-ph.IM); Machine Learning (cs.LG)
*备注: 16 pages, 6 figures, submitted to CQG
Abstract:We present the results of \textbfMADGRAV, a deep-learning-based search for high-mass compact binary coalescences, applied to the data collected by the LIGO interferometers during the third observing run and during the first and second part of the fourth observing run. The \textbfMADGRAV pipeline consists of a series of sequential convolutional neural networks that perform anomaly detection, glitch classification, coherence testing, and signal ranking. Data from the Hanford and Livingston LIGO detectors are studied (both individually and in coherence) by way of 1 second Q-transform windows. Of the candidates that survive every stage of the pipeline, 48 reach the significance threshold, and we report 47 gravitational wave detections characterised by a false alarm rate below 1,\rm yr^-1 with a probability of astrophysical origin p_\rm astro0.9 . Of the 47 detections, 44 are shared with the minimally modelled coherent WaveBurst search. The observed total source-frame masses, extracted from official gravitational wave transient catalogues, are in the 14-236 M_\odot range with a median of 69 M_\odot , and a median SNR of 16. We note that the recovered fraction of confident detections rises with mass: for LIGO detectors network SNR 10 the pipeline recovers 8.1% of confident catalog events below 30 M_\odot , 39.8% between 30 and 100 M_\odot , and 53.3% above 100 M_\odot , corresponding to 33.3% , 45.5% and 53.3% of the events detected by coherent WaveBurst in the same bins. These results suggest that anomaly detection pipelines can serve as an independent detection channel complementary to matched filtering in the high-mass high-SNR regime.
[LG-186] WEIRDO: WEak resIdual Regularized DOobs h-transform diffusion alignment
链接: https://arxiv.org/abs/2609.39531
作者: Denis Suchkov
类目: atistics Theory (math.ST); Machine Learning (cs.LG)
*备注:
Abstract:We study the problem of estimating the guidance that steers the distribution learned by a diffusion generative model toward a tilted target q_0 \propto w,p_0 at inference time. Relying on the stochastic optimal control approach, we observe that the exact drift correction is the gradient of the logarithm of Doob’s h -function, and we study the problem of estimating it from a sample. In the present paper, we assume that the score of the pretrained model is available, that the tilting weight is bounded and positive, and that the reference distribution has a bounded support, no smoothness of the weight is required. Introducing a penalized least-squares risk in which the penalty is the residual of the space-time harmonicity equation satisfied by the h -function, measured in a dual Sobolev norm, we derive high-probability bounds on the squared error of the resulting guidance estimate. Since the penalty vanishes at the target, the estimator is free of regularization bias, and in favourable scenarios its rate of convergence is faster than the minimax rate of estimating first-order derivatives of a smooth regression function. Assuming that w is bounded and positive with \mathbbE_p_0[w^-\mathrms] \infty for some \mathrms \in (0,\infty] , and that the reference data are compactly supported, we prove that the guidance is estimable in squared L^2 at rate \varepsilon_n^\mathrms/(\mathrms+4) , where \varepsilon_n = n^-2(\beta-1)/(2(\beta-1)+d). We also transfer the obtained bounds to the total variation distance between the marginals of the estimated and the exactly guided samplers, and illustrate the performance of the suggested approach with numerical experiments.
[LG-187] Mitigating Representation Gaps in Amortized Bayesian Inference with Auxiliary Supervision
链接: https://arxiv.org/abs/2609.39525
作者: Hans Olischläger,Svenja Jedhoff,Šimon Kucharský,Aayush Mishra,Stefan T. Radev,Paul Bürkner
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Computation (stat.CO); Methodology (stat.ME)
*备注:
Abstract:Casting Bayesian inference as a neural network optimization problem targeting an amortized posterior is attractive, as it extends to otherwise intractable statistical models and offers near instantaneous inference for new datasets after prepaying the training cost. Although theory guarantees faithfulness under ideal convergence, practical amortized inference still requires iterating over architectures and optimization choices and ultimately ``satisficing’’ under finite simulation, compute, and time budgets. Even the best-performing solution may thus retain avoidable representation gaps that typically require problem-specific fixes. Here, we propose a generic alternative which improves training dynamics with auxiliary guidance losses applied to internal representations. Specifically, we show how such guidance leads to faster convergence when training data is abundant and to better performance when it is scarce. We formalize representation gaps as getting stuck in a local optimum at the information bottleneck between the parts of the network tasked with feature learning and those tasked with conditional distribution learning, and offer a generic diagnostic to separate summary failures from inference failures. Finally, we demonstrate that auxiliary supervision improves convergence speed and accuracy on a range of challenging real-world inference problems.
[LG-188] Distributionally robust linear regression through the lens of adversarial training
链接: https://arxiv.org/abs/2609.39449
作者: Elis Stefansson,David Vävinggren,Antônio H. Ribeiro
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Distributionally robust optimization (DRO) studies parameter estimation under uncertainty in the underlying probability distribution and has emerged as a principled framework for analyzing robustness and generalization. In particular, Wasserstein DRO, with distributional uncertainty induced by the Wasserstein distance, generalizes several popular regularizers. This paper studies Wasserstein DRO linear regression, unifying square-root Lasso and adversarial linear regression as important special cases. We prove that many properties of these two special cases carry over to this general method. In particular, we show (i) deterministic and non-asymptotic in-sample error bounds O(n^-1/2) in general and O(n^-1) under design matrix and sparsity conditions; (ii) insensitivity to the noise level, also known as the pivotal property; and (iii) solution equivalences for small and large ambiguity sets. The key proof step is to recast the method into a quadratic form, mimicking adversarial linear regression. We also show that the method can be solved efficiently, and we validate our findings through numerical simulations.
[LG-189] Principal Component Regression Dominates all Monotone Spectral Filters for Linear Regression
链接: https://arxiv.org/abs/2609.39440
作者: Juno Kim,Hengyu Fu,Peter Bartlett,Jason D. Lee,Jingfeng Wu
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 63 pages
Abstract:We compare the instance-wise, finite-sample risks of monotone spectral filters for linear regression, a broad class of estimators including principal component regression (PCR), gradient descent (GD), and ridge regression. We show that PCR dominates all monotone spectral filters: compared to any such filter, the risk of optimally tuned PCR is no bigger by a constant factor for all problems. Furthermore, the dominance is strong if the filter is separated from step functions (e.g., GD and ridge): there exist problem instances for which the risk of PCR is smaller by a polynomial factor in sample size dependence. Our comparison results show that PCR is optimal and thus admissible among monotone filters, significantly extending Wu et al. (2026)'s result that GD strongly dominates ridge. From a technical perspective, we establish new upper and lower bounds for general spectral filters, which are instance-wise sharp when specialized to ridge or GD, recovering or improving the best-known bounds.
[LG-190] owards Optimal Inventory Control under Censored Demand: A Biased Sample-Averag e Approximation Approach
链接: https://arxiv.org/abs/2609.39397
作者: Yuxuan Han,Xiaoyu Fan,Jiawei Zhang,Zhengyuan Zhou
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We study data-driven multi-period lost-sales inventory control under censored demand, where a stockout reveals only that demand exceeded the stocking level. We develop a unified, model-based framework for policy learning from censored data, built on a new cost decomposition for base-stock policies and a biased sample-average approximation (SAA) approach. The cost decomposition allows us to propose a new coverage condition under which censored observations are informative enough for sample-efficient policy learning. Guided by this coverage condition, we design two biased SAA algorithms: an upper-biased one that achieves near-optimal sample complexity under the offline coverage condition, and a lower-biased one that actively generates the required coverage and achieves near-optimal regret online. More broadly, this biased SAA approach provides a general principle for implementing pessimism and optimism under censored feedback, which may be of independent interest.
[LG-191] A Dynamical Theory of LoRA in Continual Learning
链接: https://arxiv.org/abs/2609.39367
作者: Théo Marchetta,Filippo Alessandroni,Alessandro Breccia,Alessandro Ingrosso,Federica Gerace
类目: Machine Learning (stat.ML); Statistical Mechanics (cond-mat.stat-mech); Machine Learning (cs.LG); Probability (math.PR); Statistics Theory (math.ST)
*备注:
Abstract:Despite the widespread use of Low-Rank Adaptation (LoRA), little is known about its dynamics in continual learning and the mechanisms by which low-rank updates affect catastrophic forgetting. We provide an asymptotically exact dynamical characterization of LoRA in a solvable two-task teacher-student model. In the high-dimensional online-learning limit, we derive a closed system of ordinary differential equations for a finite set of macroscopic order parameters, yielding exact expressions for the generalization errors throughout both the initial Task 1 learning phase and the subsequent LoRA fine-tuning on Task 2. The theory quantitatively matches finite-dimensional simulations and exposes two characteristic effects of LoRA: low-rank adaptation reduces interference with features learned on the first task, but its initialization slows adaptation to the second task. Building on this mechanistic picture, we analyze a state-dependent masking strategy that freezes hidden units carrying the strongest first-task representations and restricts adaptation to the complementary subspace. This structural partitioning markedly reduces forgetting, while preserving plasticity on the new task. Our framework further clarifies the role of adapter rank: transfer improves only up to the intrinsic dimensionality of the target task and saturates beyond it, while forgetting continues to grow with rank. These results provide a dynamical and geometric account of how low-rank adaptation organizes information across sequential tasks and are qualitatively reproduced on a sequential MNIST benchmark.
[LG-192] Discrete Score Matching Enables Causal Discovery from Count Data
链接: https://arxiv.org/abs/2609.39326
作者: Euijong Song,Hyewon Park,Gunwoong Park
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Count data pose a challenge for score-matching-based causal discovery: derivatives are unavailable, and simply replacing them with finite differences does not generally suffice for causal discovery. We generalize SCORE’s constant-curvature criterion (Rolland et al., 2022) by conditioning on the node’s value, yielding the conditional curvature score (CCS) for ordering. We also extend curvature-based parent recovery through the off-diagonal curvature score (OCS), enabling directed acyclic graph (DAG) recovery with both scores constructed from score functions for continuous data and concrete scores for counts. In the bivariate setting, zero CCS exactly characterizes a semiparametric generalized linear model (GLM) conditional form in which the conditional family need not be specified in advance, unlike in classical GLMs. For bivariate semiparametric GLM DAGs under our regularity condition, canonical-parameter nonlinearity is necessary and sufficient for identifiability. In multivariate DAGs, this nonlinearity enables DAG recovery through CCS and OCS. Our framework identifies a new class of semiparametric GLM DAGs that strictly contains the nonlinear Gaussian ANM class identified by SCORE. We introduce DISCO (DIscrete SCOre), a count-DAG recovery algorithm that estimates CCS and OCS using discrete diffusion. Experiments demonstrate accurate DAG recovery across Poisson, negative binomial, binomial, and mixed-family settings, as well as scalability to 1,000-node DAGs on a single GPU.
[LG-193] Generalized Geometry Block Proximal Linearized Method for Multiblock Nonconvex and Nonsmooth Optimization
链接: https://arxiv.org/abs/2609.39301
作者: Weifeng Yang
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:
Abstract:This paper considers a class of multiblock nonconvex and nonsmooth optimization problems arising in many applications. Existing methods construct proximal linearized operators or their variants within standard Euclidean geometry to solve this class of problems, forcing their block variable updates to rely on the standard inner product and its induced norm. Nevertheless, this construction fails to capture the geometric structure of the target problem, leading to low numerical efficiency. To overcome these drawbacks, we propose a generalized geometry proximal linearized operator for updating block variables, and develop the Generalized Geometry Block Proximal Linearized (GGBPL) method based on this operator. Compared with existing proximal linearized operators, the proposed operator allows the block surrogate functions to be constructed using arbitrary inner products and general admissible metrics, thereby enabling the GGBPL method to adapt its updates to the geometric structure of various problems. We also introduce the inertial version of GGBPL, named the inertial GGBPL (iGGBPL) method. We further establish a new unified convergence framework under this generalized geometry, within which we prove that our methods guarantee convergence of the objective function values, establish global convergence of the generated sequence to a critical point, and derive the convergence rate of our methods. We also establish an \mathcalO(\varepsilon^-2) iteration complexity bound for obtaining an \varepsilon -stationary point. We apply our methods to two nonconvex and nonsmooth problems: sparse nonnegative matrix factorization with \ell_0 -constraints and sparse nonnegative CP decomposition with \ell_0 -constraints. Numerical results demonstrate the superior numerical performance of our proposed methods over several state-of-the-art methods.
[LG-194] Minimax Additive Regression under Unknown Dependent Designs
链接: https://arxiv.org/abs/2609.39212
作者: Baptiste Ferrere,Fabrice Gamboa,Jean-Michel Loubes
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We study additive regression under a potentially non-product random design on [0,1]^d , allowing the dimension d to grow with the sample size n . We introduce coupled smoothness classes that separately control the regularity of the marginal densities and the density-weighted additive components. To handle dependence, we adapt a Riesz-basis construction for functional ANOVA models and establish compatibility bounds with constants independent of the dimension under uniform bounds on the joint density. We construct thresholded least-squares estimators and establish matching minimax upper and lower bounds for prediction with known or unknown marginal densities, under suitable dimension-growth conditions. When the marginal densities are at least as smooth as the weighted components, the unknown-density problem attains the known-density minimax rate. When the densities are less smooth, their regularity determines the minimax rate over the coupled class. Finally, we show that the centered additive components can be recovered at the same aggregate upper rate, without an additional order of error.
[LG-195] Asymptotic Properties of Support Vector Machines in High-Dimension Low-Sample-Size Settings under a Spiked Model
链接: https://arxiv.org/abs/2609.39173
作者: Yugo Nakayama
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:In this paper, we consider asymptotic properties of the support vector machine (SVM) in high-dimension, low-sample-size (HDLSS) settings under a spiked model. The existing theory of the SVM in the HDLSS context relies on the geometric representation of HDLSS data, which requires that the eigenvalues of the covariance matrices are not dominant. We first show that the geometric representation does not hold under the spiked model. We show that the Gram matrix of HDLSS data converges in distribution to a random matrix, namely, the HDLSS data converge to a random configuration in a finite-dimensional space whose dimension is given by the number of the spikes. We show that the misclassification rates of the SVM do not tend to zero, that is, the SVM does not hold the consistency property. We also show that the bias-corrected SVM (BC-SVM) does not give preferable performance in this setting because the bias term itself should be modified. In order to overcome such difficulties, we propose a spike-corrected SVM (SC-SVM). We show that the SC-SVM holds the consistency property when the sample size goes to infinity, and that the growth of the sample size is essential in the sense that any projection-based procedure fails when the sample size is fixed. Finally, we check the performance of the classifiers by numerical simulations.
[LG-196] A Width-Matched Comparison of Hybrid Quantum-Classical Self-Supervised Learning for Fingerprint Recognition
链接: https://arxiv.org/abs/2609.39172
作者: Maria S. Edwards,Kidwell Dlamini,Pin-An Lin,Wen-Hsien Hsu,Wen-Chieh Fang
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG)
*备注:
Abstract:Fingerprint recognition is a widely deployed biometric, but supervised training requires large labeled enrollment sets. Self-supervised learning (SSL) removes this requirement, and hybrid quantum-classical models have been proposed to enrich the learned representations. Prior quantum SSL studies consider a single contrastive objective, so it is unclear whether reported benefits depend on the objective or can be attributed to the quantum circuit. We insert the QuFeX quantum feature-extraction module into three SSL frameworks, the contrastive SimCLR and MoCo v2 and the non-contrastive BYOL, and compare each hybrid with its classical counterpart at matched representation width (8 features, equal to 8 qubits) on the SOCOFing fingerprint dataset, with a CIFAR-10 control, using k-nearest-neighbor identification on encoder features. In single-run experiments the hybrid scores clearly higher for both contrastive objectives, whereas for BYOL a multi-seed analysis shows no reliable difference, suggesting that any benefit depends on the SSL objective. A hardware-efficient circuit (QNet) does not show the same gain. We examine whether the gains can be attributed to the quantum circuit, considering circuit architecture, trainable parameter count, nonlinearity, and the classical simulability of 8-qubit circuits.
[LG-197] Steepest Guidance: A Practical and Principled Approach to Inference-Time Alignment of Flow and Diffusion-based Models
链接: https://arxiv.org/abs/2609.39091
作者: Shokichi Takakura,Akifumi Wachi,Rei Higuchi,Kohei Miyaguchi,Taiji Suzuki
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Inference-time alignment of flow and diffusion-based models is critical for achieving flexible generative modeling. Theoretically, Doob’s h -transform provides an elegant solution to this problem, and most existing methods are based on this principle. However, in practice, estimating the optimal guidance derived from Doob’s h -transform at inference time is challenging. To deal with this issue, we regard inference-time alignment as a sequential optimization problem in the space of probability measures and propose a novel framework called Steepest Guidance, based on the principle of maximizing local improvement in the objective. We provide a theoretical analysis of the proposed method and demonstrate its effectiveness through extensive experiments.
[LG-198] A strategic roadmap for an atomistic machine-learning ecosystem
链接: https://arxiv.org/abs/2609.39090
作者: Jörg Behler,Michele Ceriotti,Cecilia Clementi,Gábor Csányi,Alin-Marin Elena,Aditi Krishnapriyan,Joseph W. Abbott,Fabio Affinito,Albert P. Bartók,Ilyes Batatia,Filippo Bigi,Florian N. Brünig,Yannick Calvino Alonso,Giuseppe Carleo,Aurélie Champagne,Stefan Chmiela,Marc L. Descoteaux,Ralf Drautz,Alexandra Farcas,Meng Gao,Rohit Goswami,Michael F. Herbst,Christian Holm,James R. Kermode,Alexander L. M. Knoll,Tobias Kreiman,Hoang-Thien Luu,Yury Lysogorskiy,Mihai-Cosmin Marinica,Rocco Meli,Klaus-Robert Müller,Frank Noé,Mohamadhosein Nosratjoo,Simon Olsson,Christoph Ortner,Aldo S. Pasos-Trejo,Anyang Peng,Eric Qu,Andrea Rizzi,Mariana Rossi,Bassem Sboui,Gregor N. C. Simm,Alexandre Tkatchenko,Jacopo Venturin,O. Anatole von Lilienfeld,William C. Witt,Brandon M. Wood,Tigany Zarrouk,Fabian Zills
类目: Chemical Physics (physics.chem-ph); Materials Science (cond-mat.mtrl-sci); Machine Learning (cs.LG)
*备注:
Abstract:Data-driven machine learning (ML) techniques have become an essential tool in many domains of science. Their application to atomistic simulations of matter is particularly widespread and impactful. This success is due largely to the existence of a well-developed and established physics-based modeling framework, ranging from first-principles electronic-structure calculations to molecular dynamics and statistical sampling, into which ML was integrated naturally to reshape long-standing trade-offs between accuracy, efficiency, and scale. Nevertheless, this integration raises both conceptual and practical challenges, from choosing between data-centric and physics-based modeling approaches to adapting established software stacks to modern hardware accelerators and ML libraries. As the field evolves rapidly, fueled in part by widespread enthusiasm but also by tangible impact, it seems appropriate to take a moment to consider the current state of the art and open challenges, and reflect on what can be done to better coordinate efforts across the community. With this goal in mind, several members of this community met in Lausanne in January 2026 at CECAM to discuss algorithms, models, software and hardware infrastructure, and the most promising scientific applications that have become possible thanks to the use of artificial intelligence in atomic-scale simulations. This strategic roadmap paper summarizes the outcomes of these discussions, suggesting some long-term goals, and some concrete actions, to establish a healthy, sustainable and impactful atomistic ML ecosystem.
[LG-199] Minimax rates for learning spectral Barron functions by deep ReLU neural networks
链接: https://arxiv.org/abs/2609.39020
作者: Songqiu Ma,Yunfei Yang
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注:
Abstract:We study how well deep neural networks approximate and learn spectral Barron functions. Recent studies have shown that these function classes can be efficiently approximated by shallow neural networks without suffering from the curse of dimensionality. We complement these results by providing new approximation bounds for deep networks with ReLU activation and establishing the minimax rates for learning these function classes. Specifically, we show that d -dimensional spectral Barron functions with smoothness index s0 can be approximated by deep ReLU neural networks with approximation rate \widetilde\mathcalO (S^-\frac12-\fracsd) , where S denotes the number of nonzero parameters in the network. Using this approximation result, we further show that deep ReLU neural networks can learn spectral Barron functions in a fast rate n^-\fracd+2s2d+2s with n training samples. Finally, we prove that this convergence rate is minimax optimal up to logarithmic factors.
[LG-200] Warm-starting PDE solvers with any-dimensional machine learning
链接: https://arxiv.org/abs/2609.38916
作者: Wilson G. Gregory,George A. Kevrekidis,Ben Blum-Smith,Soledad Villar
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 29 pages, 4 figures
Abstract:Any-dimensional machine learning models, such as graph neural networks (GNNs), can be naturally trained and evaluated on inputs of different sizes and dimensions. Inspired by the GNN transferability literature, we show mathematical conditions under which a partial differential equation (PDE) learning-based solver can be trained in small dimensions and directly applied to solve a higher dimensional PDE in a zero-shot fashion. These conditions are based on symmetries in both the partial differential equation and the initial data. When the equations satisfy the symmetries but the data does not, which is the case for many PDEs arising from physics, we show that our theory gives a principled way of warm-starting low-dimensional PDE solvers for higher dimensional PDEs. We apply this method on the heat equation, Burgers’ equation, and the compressible Navier–Stokes equations, improving the performance in both zero-shot and typical training regimes on high dimensional data. For example, we train a surrogate model on 2D Navier–Stokes data and achieve better results on 3D test data than a baseline surrogate model trained on 3D data, while only using 12 % of the flops and 20 % of the total data size.
[LG-201] Sharp Statistical Rates for Asynchronous TD Learning with Markovian Data
链接: https://arxiv.org/abs/2609.38880
作者: Yang Peng
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We study the last iterate of standard tabular temporal-difference (TD) learning from a single trajectory of a finite Markov reward process. For discount factor \gamma , write H=(1-\gamma)^-1 , and let \mu_\min and t_\operatornamemix denote the minimum stationary probability and total-variation mixing time. We prove that last-iterate TD achieves sup-norm error at most \varepsilon with high probability using \widetilde O\left( \fracH^3\mu_\min\varepsilon^2 +\fract_\operatornamemix\mu_\min \right) transitions, for 0\varepsilon\leq1 . This rate holds both for a constant step size selected for the target accuracy and for a decreasing schedule independent of the target accuracy and terminal time. The latter gives a simultaneous guarantee over all times beyond an explicit transient threshold. The statistical term retains the cubic effective-horizon dependence of synchronous TD, and the additive mixing transient has no extra horizon factor. The result allows non-reversible chains, arbitrary initial state distributions, and bounded rewards that may depend on the next state. The proof uses an anchored local Poisson equation in reverse time to control stochastic fluctuations without a mixing-time factor, and a hitting-time compensation identity to bound initialization error. The latter also yields a finer transient in terms of the worst expected reverse hitting time. A bound on the expected cumulative propagation mass extends this argument to decreasing step sizes. A three-state construction with known deterministic rewards gives matching minimax lower bounds for the statistical and mixing terms, up to logarithms, over specified model classes in a slow-mixing parameter regime. Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG) Cite as: arXiv:2609.38880 [stat.ML] (or arXiv:2609.38880v1 [stat.ML] for this version) https://doi.org/10.48550/arXiv.2609.38880 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-202] Learn-Then-Differentiate Gradient Estimation
链接: https://arxiv.org/abs/2609.38842
作者: Nifei Lin,Qingkai Zhang,L. Jeff Hong
类目: Computation (stat.CO); Machine Learning (cs.LG)
*备注:
Abstract:Learn-then-differentiate (LTD) estimates gradients by fitting a model to simulation outputs and differentiating it. We develop a unified framework explaining what LTD differentiates and how accurately it estimates gradients. For models with a weighted representation, LTD differentiates a learned representation of the underlying probability measure. We then show how accuracy guarantees for fitted models translate into guarantees for gradients and higher-order derivatives, with rates approaching the standard Monte Carlo rate under suitable smoothness conditions. The framework recovers established results for kernel regression, local polynomial regression, and kernel ridge regression, and yields further guarantees for multiple kernel learning and smooth neural networks. These results provide a common foundation for understanding and analyzing LTD across learning methods.
[LG-203] Averag e-and Last-Iterate Lower Bounds for Optimistic Matrix Mirror-Prox in Quantum Zero-Sum Games
链接: https://arxiv.org/abs/2609.38835
作者: Yiheng Su,Emmanouil-Vasileios Vlatakis-Gkaragkounis,Pucheng Xiong
类目: Quantum Physics (quant-ph); Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: 38 pages, 6 figures
Abstract:Optimistic matrix mirror-prox (OMMP) computes \epsilon -approximate Nash equilibria in quantum zero-sum games with an O(1/\varepsilon) average-iterate guarantee [arXiv:2311.10859]. We investigate whether this dependence on accuracy is tight and whether geometric last-iterate convergence can be guaranteed. We study these questions through explicit games with one qubit per player. First, we prove an \Omega(1/\varepsilon) lower bound for the uniform-average output that includes the maximally mixed initial state, independently of the regularizer and step size. Second, we construct a fixed game on which optimistic gradient descent-ascent (OGDA), initialized at the maximally mixed state, has last-iterate Frobenius distance to equilibrium \Theta(1/t) and duality gap \Theta(1/t^3) for every sufficiently small fixed step size. A separate fixed game exhibits arbitrarily long delays in reducing the initial error by a constant factor across a family of initial states. Finally, we give a fixed game with a unique, strictly complementary equilibrium on which optimistic matrix multiplicative weights updates (OMMWU) converge only polynomially from the maximally mixed state for every fixed positive step size. The last-iterate Frobenius distance and quantum relative entropy from the equilibrium to the iterates decay as \Theta(1/t) , while the duality gap decays as \Theta(1/t^2) .
[LG-204] Methodological Changes to the Attention ResUNet Hourly Precipitation Postprocessor
链接: https://arxiv.org/abs/2609.38609
作者: Thomas M. Hamill
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Machine Learning (cs.LG)
*备注:
Abstract:This note is a technical companion to a previously published preprint describing an Attention Residual U-Net that postprocesses deterministic forecasts from The Weather Company’s Global and Regional Atmospheric Forecast (GRAF) model into probabilistic hourly precipitation forecasts. It documents what has changed in that method since publication. Feature-wise Linear Modulation conditioning on calendar season and forecast lead time is used to produce a single trained model for each season, replacing 192 separately trained per-month, per-lead checkpoints. Lead time is extended from 48 to 72 h. Two new input channels are used, per-pixel local solar hour and a static, monthly-varying precipitation climatology. During verification, the climatological reference against which the Brier Skill Score is computed now has an added diurnal dimension, on top of the monthly resolution it already had. Brier Skill Score and reliability are compared between the new vs. the previous training. Forecasts generated with the new training show a modest, consistent improvement of the current training over the original.
[LG-205] A Parameter-Free Zeroth-Order Method with Covariance Matrix Adaptation and Effective Dimension
链接: https://arxiv.org/abs/2609.38561
作者: Alexander Sholokhov,Alexander Rogozin
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:
Abstract:Zeroth-order optimization methods are essential for solving black-box problems where gradient information is unavailable or expensive to compute. This paper presents POEM-CMA, a novel parameter-free stochastic zeroth-order algorithm that extends the recent POEM method by integrating covariance matrix alignment and the notion of effective dimension. In contrast to traditional zeroth-order approaches that rely on isotropic random directions, POEM-CMA performs anisotropic sampling by constructing a covariance matrix from gradient estimates. This enables the algorithm to focus sampling efforts on the most informative directions. We introduce the use of the empirical effective dimension d^* = \frac\operatornametr(\hat\Sigma)\lambda_\max(\hat\Sigma) , which reflects the intrinsic dimensionality of the problem and replaces the ambient dimension in both sampling and complexity analysis. We prove that POEM-CMA achieves a near-optimal convergence rate, requiring only \tilde\mathcalO\left(\fracd^* \kappa(\hat\Sigma) L^2 D_\mathcalX^2\varepsilon^2\right) stochastic zeroth-order oracle queries. The method remains fully parameter-free and demonstrates significant improvements over the original POEM in problems with low-rank structure where d^* \ll d . Numerical experiments on hinge-loss binary classification tasks using LibSVM datasets confirm the practical superiority of the proposed approach. Subjects: Optimization and Control (math.OC); Machine Learning (cs.LG) Cite as: arXiv:2609.38561 [math.OC] (or arXiv:2609.38561v1 [math.OC] for this version) https://doi.org/10.48550/arXiv.2609.38561 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-206] Say Echo Do: Strategic Narratives and Revealed Positioning in Financial Markets
链接: https://arxiv.org/abs/2609.38545
作者: Ali Atiah Alzahrani
类目: Trading and Market Microstructure (q-fin.TR); Machine Learning (cs.LG)
*备注: First draft. 9 pages main text, 31 pages total, 3 figures in the main text. Code: this https URL
Abstract:Machine-learning signals built from financial text treat what institutions say, and what the media repeat, as evidence about value. But whoever shapes a narrative may be trading against it. We study markets with three observable voices: institutional statements (Say), media repetition (Echo) and revealed positioning (Do). We ask when words should be followed and when they should be faded. In a linear-quadratic model of an informed institution that speaks and trades before a partly credulous crowd, talking an asset down while buying it is optimal exactly when \varphi^22\lambda k\varphi . A distribution-free identity then shows that when the observable Say-Do covariance is negative, words carry negative predictive content and should be faded. For measurement, we derive (i) an exact factorised posterior over which articles are echoes, combining arrival times with embedding similarity; (ii) a return-aligned contrastive objective that attains its bound exactly when squared embedding distances are an increasing affine function of squared outcome distances, with the tightest loss-based certificate of which neighbour rankings survive imperfect training; and (iii) a path-signature statistic for who moved first. In a controlled market with known ground truth, echo sentiment predicts returns with a significantly negative sign in all 29 simulated markets, the rolling Say-Do correlation flags false-alarm events with an AUC of 0.90, and return-aligned embeddings organise headlines by consequence rather than topic. We also report where the tools fail.
[LG-207] Generative sequence modeling for infinite memory processes via predictive states
链接: https://arxiv.org/abs/2609.38524
作者: Michael Wieck-Sosa,Cosma Rohilla Shalizi
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Methodology (stat.ME)
*备注:
Abstract:We consider estimating the one-step-ahead conditional distribution of a multivariate stochastic process. Many existing approaches rely on assumptions such as finite-range memory, sparsity, or additivity, which can be poorly suited to processes with long-range nonlinear interactions. However, without such structural assumptions, nonparametric estimation is challenging due to the curse of dimensionality. To address this challenge, we introduce a new estimation approach based on the predictive states of a process, possibly with infinite-range memory. We show that our estimator achieves fast convergence rates when the past history can be compressed into a low-dimensional statistic that is sufficient for predicting the future. Specifically, we show that the statistical complexity of the estimation problem is determined by the intrinsic dimension of the predictive state space. We establish guarantees for an instantiation of our method based on deep neural network estimators, and we support these theoretical results with experiments.
[LG-208] A Pre-trained Variational Autoencoder for Gyrokinetic Plasma Turbulence Surrogate Modeling
链接: https://arxiv.org/abs/2609.38438
作者: Minglei Yang,Marshall Nicholson,Diego Del-Castillo-Negrete,David Hatch,Guannan Zhang
类目: Plasma Physics (physics.plasm-ph); Machine Learning (cs.LG)
*备注:
Abstract:Machine learning surrogate models offer a promising path toward accelerating plasma turbulence simulations. We present PreVAE-Turb, a surrogate modeling framework that leverages pre-trained variational autoencoders (VAEs) from the Stable Diffusion image generation model for efficient spatial compression of turbulence fields. The pre-trained VAE is fine-tuned on turbulence data using a physics-informed loss function that includes a spectral loss operating in Fourier space to enforce spectral accuracy across scales. The VAE is combined with convolutional long short-term memory (ConvLSTM) networks to learn temporal dynamics in latent space, with a manifold consistency error metric that monitors encode–decode consistency during autoregressive rollouts. We validate the framework on two-dimensional Hasegawa-Wakatani drift-wave turbulence and extend it to gyrokinetic turbulence from the GENE code, where a four-channel adaptation simultaneously predicts electrostatic potential, density, and parallel/perpendicular temperature fluctuations without requiring architecture redesign. Once trained, inference generates thousands of time steps in seconds on a single GPU, providing substantial computational acceleration compared to direct numerical simulation. The pre-trained approach offers a transferable methodology broadly applicable to various turbulence simulation codes.
[LG-209] Advantage of Sample Complexity in Quantum PAC Learning Requires Inverse Access to State-Preparation Unitaries
链接: https://arxiv.org/abs/2609.38403
作者: Natsuto Isogai,Satoshi Yoshida,Mio Murao
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 40 pages, no figures
Abstract:Whether quantum computation can reduce the amount of data sampled from an unknown probability distribution required to learn a prediction rule is a fundamental question in quantum machine learning. Quantum PAC learning studies this question using quantum data as a quantum state whose squared amplitudes encode the unknown distribution from which classical learning data are sampled. With only copies of such quantum data, the optimal worst-case sample complexity asymptotically matches that of classical PAC learning. In contrast, access to both a state-preparation unitary for this state and its inverse can improve the query-complexity dependence on the accuracy parameter in realizable learning. However, it has remained unclear whether forward-only access allows such an improvement. In this work, taking the worst case over compatible state-preparation unitaries and their finite ambient dimensions, we show that the optimal forward-only query complexities of realizable and agnostic learning are, respectively, \Theta((d+\log(1/\delta))/\varepsilon) and \Theta((d+\log(1/\delta))/\varepsilon^2) , where d is the VC dimension of the concept class, \varepsilon the accuracy parameter, and \delta the failure probability. These bounds match the optimal sample complexities with classical data or quantum data copies. To prove them, we establish a reduction using q copies of the prepared state to approximate the Haar-averaged output of any q -query forward-only algorithm. These results show that forward-only access cannot provide an asymptotic query-complexity advantage over learning from classical data or quantum data copies in this worst-case setting, and establish the essential role of inverse access in the known realizable-setting improvement. Our reduction also provides a new framework for analyzing limitations of forward state-preparation access via state-copy lower bounds. Comments: 40 pages, no figures Subjects: Quantum Physics (quant-ph); Machine Learning (cs.LG); Machine Learning (stat.ML) Cite as: arXiv:2609.38403 [quant-ph] (or arXiv:2609.38403v1 [quant-ph] for this version) https://doi.org/10.48550/arXiv.2609.38403 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-210] Unified Optimality Conditions for Stochastic Optimal Control in the Rough Path and Itô Frameworks
链接: https://arxiv.org/abs/2609.38395
作者: Thomas Lew
类目: Optimization and Control (math.OC); Machine Learning (cs.LG); Robotics (cs.RO); Systems and Control (eess.SY); Probability (math.PR)
*备注:
Abstract:Stochastic differential equations (SDEs) can be studied via Itô calculus and rough path theory. For stochastic optimal control, these two frameworks give distinct Pontryagin Maximum Principle (PMP) optimality conditions with forward-backward SDEs (FBSDEs) or rough differential equations. We show that the adjoint equations of the Itô and rough PMPs are connected via the conditional expectation p_t^\textItô=\mathbbE[p_t^\textrough \mid \mathcalF_t] , where \mathcalF_t represents information available at time t . First, we derive a rough stochastic PMP for problems with adapted controls that does not use FBSDEs. Its proof extends the rough stochastic PMP over deterministic controls by considering stochastic needle variations. Second, we derive a unified PMP connecting the Itô and rough PMPs, using Itô-Stratonovich conversion formulas and duality identities between the forward tangent and backward adjoint SDEs. As a first application, we rederive the adjoint matching method for fine-tuning generative models. As a second application, we propose an indirect shooting method for a class of feedback problems. Overall, these results give a new conditional bridge connecting two popular frameworks for stochastic optimal control.
[LG-211] Lower Bounds for Linear-Oracle Online Learning
链接: https://arxiv.org/abs/2609.38375
作者: Mohit Sinha
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Can a constant number of linear minimizations per round improve on the T^3/4 regret rate of online Frank-Wolfe on general convex sets? Weibel et al. conjectured that fixed-coefficient methods cannot. We prove their conjecture and extend the lower bound to every deterministic learner in an oracle-only model. The learner receives an initial feasible point and a diameter bound, and must remain feasible on every domain consistent with its oracle replies. For T rounds, at most b calls between decisions, diameter bound D , and gradient norm bound L , we construct an instance in dimension d=2b(T-1)+1 with regret at least 2^-1/4LDb^-1/4T^3/4 . The adversary fixes the domain, initial point, deterministic tie rule and linear losses before play. The vertices form a path on which every point available before a decision has zero current loss, while the final vertex has negative loss on every round. For constant b , the result matches the known upper rate for dimension-independent guarantees. For one-call fixed schedules with a nonzero coefficient on the newest gradient, a second construction gives regret at least 3LDT^3/4/4 with unique minimizers at every issued query. Exact-arithmetic certificates for the tuned schedule of Weibel et al. closely match their finite-horizon numerical worst cases, with unique oracle replies.
[LG-212] Gestalt: a meta-foundation model for astronomy NEURIPS2026
链接: https://arxiv.org/abs/2609.38312
作者: Michael J. Smith,Shashwat Sourav
类目: Instrumentation and Methods for Astrophysics (astro-ph.IM); Machine Learning (cs.LG)
*备注: 12 pages, 3 figures, 6 tables, accepted to the Interpretability for Discovery workshop at NeurIPS 2026
Abstract:The Platonic Representation Hypothesis predicts that sufficiently scaled foundation models converge on a shared representation of the world. As each non-converged model gives a noisy view of a common structure when passed the same input, we ask whether we can combine models into a representation that outperforms its individual components. We test this on galaxies: we embed images via a basket of 22 frozen foundation models from eight families, whiten each view, and take a randomised SVD of the embedding concatenation. The resulting 1024-dimensional embedding outperforms every basket member on 19/21 of our tested metrics for physical property and galaxy morphology estimation for HSC, JWST, and DESI Legacy Survey imagery. We find that performance rises with basket size and basket architectural diversity, and that the meta-foundation model’s performance transfers across astronomical surveys. We conclude that a useful astronomical foundation model can be assembled from existing generalist models with no training required beyond a single unsupervised projection. By leveraging the community’s already-spent work, we save a lot of compute: a fresh pre-train of a comparable single-domain model would cost \mathcalO(10^4 – 10^5) A100 GPU hours (emitting several tonnes of CO _2 eq.), whereas assembling Gestalt requires minutes on a single machine.
[LG-213] Robust LassoNet: Enhancing Feature Selection in Neural Networks via Robust Loss Functions
链接: https://arxiv.org/abs/2609.38263
作者: Daniela De Canditiis,Italia De Feis,Paola Stolfi
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Probability (math.PR)
*备注:
Abstract:Feature selection in neural networks remains a challenging problem, particularly in the presence of noisy or contaminated data. LassoNet is a recent approach that addresses this issue by combining neural networks with hierarchical sparsity constraints, enabling simultaneous prediction and variable selection. However, its standard formulation relies on the mean squared error (MSE) loss, which is known to be highly sensitive to outliers. In this paper, we present Robust LassoNet, an extension of LassoNet that incorporates robust loss functions, such as Huber, Cauchy, Tukey’s bisquare, and Nonnegative Garrote, to mitigate the effect of extreme observations. The proposed approach preserves the original optimization framework while improving stability under data contamination. Through experiments on synthetic and real datasets, we show that robust LassoNet significantly improves both predictive performance and feature selection accuracy in the presence of outliers or heavy tailed noise, while maintaining comparable performance in clean settings. These results highlight the importance of robustness in neural network based feature selection and suggest practical guidelines for choosing appropriate loss functions.
[LG-214] Airfoil2Vec: Spectral Geometry-Conditioned Neural Surrogate Models for Airfoil Aerodynamics and a Downforce-Generating CFD Dataset
链接: https://arxiv.org/abs/2609.38213
作者: Haitz Sáez de Ocáriz Borde,Flavio Savarino,Andrei Cristian Popescu,Pietro Innocenzi,Pantelis Papageorgiou,Xerxes Xian Chong
类目: Fluid Dynamics (physics.flu-dyn); Machine Learning (cs.LG)
*备注:
Abstract:We introduce a dataset of approximately 10,000 Reynolds-Averaged Navier-Stokes (RANS) simulations of steady, incompressible, two-dimensional subsonic flow around downforce-generating NACA 4-digit airfoils, targeting aerodynamic regimes relevant to automotive and motorsport applications (openly available on this https URL). Using this resource, we study geometry-conditioned neural surrogates for fast flow prediction, comparing neural fields with neural ODEs, MLPs with graph-based models, and several spectral geometry-conditioning methods. We further propose Airfoil2Vec, an airfoil-specific spectral geometry encoder that combines the joint contour spectrum with separate spectral representations of camber and thickness, for predicting continuous pressure and velocity fields. We evaluate generalization through angle-of-attack interpolation, interpolation and extrapolation to unseen NACA 4-digit geometries, and generalization to unseen non-NACA airfoils. The resulting surrogate accurately captures aerodynamic quantities and qualitative flow features while providing orders-of-magnitude speedups over conventional computational fluid dynamics.
[LG-215] CraftSPH: A high-accuracy and composable differentiable SPH solver implemented in PyTorch
链接: https://arxiv.org/abs/2609.38208
作者: Gen Matono,Shujiro Fujioka,Mayuko Nishio
类目: Fluid Dynamics (physics.flu-dyn); Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注: The code is publicly available at this https URL 43 pages, 14 figures
Abstract:Smoothed Particle Hydrodynamics (SPH) is well suited to a range of problems, particularly those involving large deformations in fluid dynamics. In recent years, in addition to the advancement of SPH formulations, the development of differentiable solvers has also progressed. However, unified frameworks that flexibly accommodate diverse numerical schemes, including advanced and implicit methods, while supporting continuous extension and updating remain limited. In this study, a high-accuracy and composable differentiable SPH solver, CraftSPH, has been developed. In CraftSPH, major computational operations are implemented as independent modules, which can be combined according to the intended purpose of constructing a solver. This design enables different computational schemes to be intuitively constructed from combinations of common components and allows newly developed high-accuracy discretization methods to be flexibly incorporated and extended. Furthermore, each module supports automatic differentiation, enabling applications to physical-parameter estimation and to optimization that combines SPH solvers with deep learning models. Accordingly, CraftSPH provides an SPH computational framework in which high-accuracy discretization, explicit and implicit computations, and automatic differentiation can be handled in a unified and extensible manner. To demonstrate the applicability of CraftSPH to a wide range of problems, forward analyses are conducted for Poiseuille flow, two- and three-dimensional dam-break problems, and a rising-bubble problem. In addition, the performance of automatic differentiation is evaluated through inverse problems involving parameter estimation for the Taylor-Green vortex and lid-driven cavity flow.
附件下载
