本篇博文主要内容为 2026-05-29 从Arxiv.org论文网站获取的最新论文列表，自动更新，按照NLP、CV、ML、AI、IR、MA六个大方向区分。
说明：每日论文数据从Arxiv.org获取，每天早上12:30左右定时自动更新。
提示: 当天未及时更新，有可能是Arxiv当日未有新的论文发布，也有可能是脚本出错。尽可能会在当天修复。
目录
概览 (2026-05-29)
今日共更新902篇论文,其中:
- 自然语言处理共200篇(Computation and Language (cs.CL))
- 人工智能共354篇(Artificial Intelligence (cs.AI))
- 计算机视觉共170篇(Computer Vision and Pattern Recognition (cs.CV))
- 机器学习共289篇(Machine Learning (cs.LG))
- 多智能体系统共21篇(Multiagent Systems (cs.MA))
- 信息检索共29篇(Information Retrieval (cs.IR))
- 人机交互共32篇(Human-Computer Interaction (cs.HC))
多智能体系统
[MA-0] SpecBench: Evaluating Specification-Level Reasoning for Software Engineering LLM Agents
【速读】：该论文试图解决的问题是：当前软件工程（SWE）代理主要聚焦于代码生成能力的评估，而忽略了在真实复杂系统中Specification（规格说明）设计阶段的关键挑战——即初始规格往往存在不完整、模糊或错误假设，需依赖专家评审与迭代修正才能达到可实施状态。现有基准如SWE-Bench仅衡量代理在给定固定且精确需求下的编码能力，未覆盖specification-level reasoning（规格推理）这一核心环节。解决方案的关键在于提出SpecBench，一个基于成熟开源项目RFC（Request for Comments）流程构建的新基准，用于评估代理识别初始设计提案中缺陷的能力，包括遗漏、歧义、不一致和错误假设等；其任务设计结合了项目代码库和历史RFC讨论，要求代理在无执行反馈的情况下完成对规格正确性与完备性的分析，从而推动SWE代理从代码生成向全流程自动化演进。
链接: https://arxiv.org/abs/2605.30314
作者: Grant Hamblin,Kevin Song,Zhanda Zhu,Anand Jayarajan,Sihang Liu,Nandita Vijaykumar,Gennady Pekhimenko
机构: University of Toronto(多伦多大学); University of Waterloo(滑铁卢大学); Vector Institute(矢量研究所); NVIDIA(英伟达)
类目: Multiagent Systems (cs.MA)
备注:
Abstract:Software engineering (SWE) agents are transitioning from code generation to full software development lifecycle automation. A critical phase in this lifecycle is specification design: transforming initial proposals into carefully considered requirements through expert review. Existing benchmarks such as SWE-Bench are implementation-focused by measuring the agent’s ability to generate code given fixed, precise design requirements. This formulation assumes specifications are correct and complete. In real-world complex and critical software systems, initial specifications are often incomplete and flawed, requiring extensive expert reviews and revisions before being accepted for implementation. To fill this gap, we introduce SpecBench to evaluate specification-level reasoning: the ability to generate complete, unambiguous, consistent, and correct system specifications. SpecBench tasks are derived from the Request for Comments (RFC) process used by mature open-source projects. For each task, an agent is given an initial design proposal, the project codebase, and all past project RFC discussions. The agent is tasked with identifying specification deficiencies: omissions, ambiguities, inconsistencies, or incorrect assumptions in the initial proposal. We evaluate predictions against critiques raised by expert maintainers during historical RFC reviews. SpecBench contains tasks from 5 diverse repositories: Kubernetes, React, Rust, TVM, and vLLM. We evaluate state-of-the-art SWE agents on SpecBench, analyzing their capacity to reason about system design without execution feedback. The best performing agent, GPT-5.4, achieves 44.4% accuracy.
[MA-1] EASE Configuration Facilitates A Reproducible Science of LLM Social Simulations NEURIPS2026
【速读】：该论文试图解决当前基于大语言模型（LLM）的社会交互模拟系统普遍存在的架构不标准化问题，这些问题导致研究难以复现且下游评估复杂。其解决方案的关键在于提出一个模块化框架EASE（Environments, Agents, Simulation engines, and Evaluation metrics），将多智能体仿真系统的核心组件解耦为四大可配置模块，并通过SiliSocS这一开源研究就绪的硅基社会沙盒平台实现该框架的具体落地。该方案不仅支持以明确研究问题为导向的工作流编排，还通过三个案例研究验证了其在系统性评估现有问题、深入探究复杂议题及扩展已有研究方面的有效性，从而推动LLM驱动的社会模拟向更严谨、可复现和可解释的方向发展。
链接: https://arxiv.org/abs/2605.30258
作者: Sneheel Sarangi,Maximilian Puelma Touzel,Aurélien Bück-Kaeffer,Zachary Yang,Jean-François Godbout,Reihaneh Rabbany
机构: 未知
类目: Multiagent Systems (cs.MA)
备注: 22 pages, 5 figures, under review at NeurIPS 2026
Abstract:LLMs are increasingly deployed to simulate social interactions, yet many of the existing simulators remain ad hoc and monolithic. This lack of architectural standardization prevents reproducible research and complicates downstream evaluation. We advance a rigorous science of LLM-based multi-agent simulation by modularizing core components into Environments, Agents, Simulation engines, and Evaluation metrics (EASE). We demonstrate the utility of EASE configuration by wrapping it in an experimental study schema for orchestrating workflows centered around answering explicit research questions in generated scenarios. We contribute SiliSocS, an open-source, research-ready Silicon Society Sandbox implementing a study-structured EASE configuration to enable highly configurable and reproducible LLM-based social simulations. Using SiliSocS and EASE, we present three case studies, showcasing the system’s comprehensive assessment of existing questions, ability to dive deeper into complex questions, and elaboration of existing studies, respectively. Together, these case studies highlight the limitations of current modeling approaches and isolate the impacts of design choices on key results.
[MA-2] Unifying Temporal and Structural Credit Assignment in LLM -Based Multi-Agent Prompt Optimization
【速读】：该论文试图解决多智能体系统（Multi-Agent Systems, MAS）在大型语言模型（Large Language Models, LLM）复杂推理任务中优化动态过程的难题，尤其是由于计算图的离散性和非可微性以及全局监督信号稀疏导致的优化效率低下问题。现有黑箱优化方法难以将轨迹级失败归因于特定局部组件，从而引发高方差、低效的探索。解决方案的关键在于引入结构化的归纳偏置（inductive biases），通过两个维度分解目标函数：一是时间信用分配（temporal credit），利用状态空间瓶颈识别关键回合；二是结构信用分配（structural credit），借助平稳的角色策略隔离各智能体的贡献。基于此分解信号，作者提出一种离散的、可解释的块坐标下降算法，交替优化角色提示（role prompts）与聚合协议（aggregation protocols），并使用LLM生成的“代理梯度”精准定位薄弱环节进行迭代改进。实验表明，该方法显著降低查询复杂度并提升性能，为构建自进化多智能体系统提供了原理清晰且可解释的路径。
链接: https://arxiv.org/abs/2605.30227
作者: Wenwu Li,Yuran Song,Mingze Zhao,Bo Jin,Wenhao Li
机构: Tongji University (同济大学); The University of Hong Kong (香港大学)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注: 15 pages, 4 figures, 6 tables
Abstract:While Multi-Agent Systems (MAS) empower Large Language Models to tackle complex reasoning tasks through collaborative interaction, optimizing their dynamics remains a formidable challenge due to the discrete, non-differentiable nature of the computation graph and the sparsity of global supervisory signals. Existing black-box optimizers struggle to attribute trajectory-level failure to specific local components, resulting in inefficient, high-variance exploration. We argue that tractable MAS optimization needs structural inductive biases to disentangle error signals. We propose temporal and structural credit assignment, which decomposes the objective along two axes: (i) temporal credit, using state-space bottlenecks to identify critical rounds, and (ii) structural credit, using stationary role policies to isolate agent contributions. Leveraging these decomposed signals, we introduce a discrete, verbalized block coordinate descent algorithm for iterative refinement. Rather than indiscriminate global updates, it alternates between optimizing role prompts and aggregation protocols, using LLM-generated “proxy gradients” to target only the identified weak links. Across diverse reasoning benchmarks, our approach substantially reduces query complexity while improving performance, providing a principled and interpretable path toward self-improving MAS.
[MA-3] Dissociative Identity: Language Model Agents Lack Grounding for Reputation Mechanisms
【速读】：该论文试图解决的问题是：随着自主语言模型代理（language model agents）在现实世界中日益普及，如何判断一个陌生代理是否可信并值得委托任务？现有治理思路倾向于借鉴人类身份验证和声誉机制（如“了解你的客户”和信用评分），提出“了解你的代理”（Know Your Agent）制度。然而，作者指出这一类比存在根本缺陷——因为语言模型代理本质上具有“解离性”（dissociative），其行为依赖于可变模块（如基础模型、系统提示、工具访问策略、外部记忆等）的组合，且人格角色易受攻击、难以内化惩罚，导致缺乏持续的身份识别、行为可预测性和可修复性。这使得基于身份的、事后监管式的声誉机制无法有效运作，从而破坏信任基础。解决方案的关键在于从依赖身份的、事后的、规制性的治理模式，转向基于可观测行为的、事前的、构成性的、协议驱动的行为约束机制（observability-based, ex ante, constitutive, protocol-based behavioral harnesses）。
链接: https://arxiv.org/abs/2605.30169
作者: Botao Amber Hu,Helena Rong,Max Van Kleek
机构: University of Oxford(牛津大学); New York University Shanghai(纽约大学上海分校)
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: Accepted at FaccT 2026
Abstract:As autonomous language model agents proliferate, forming an emerging agentic web with real-world consequences, what credibility signals can you use to decide whether to trust an unfamiliar agent in the wild and delegate to it? A natural governance intuition is to extend human identity verification and reputation mechanisms, from Know Your Customer'' and credit scores to Know Your Agent’’ regimes. However, we argue that this analogy is fundamentally incomplete. Reputation mechanisms function both as social signals and as corrective feedback that sustain an equilibrium of trustworthy behavior, presuming a persistent identity associated with behavioral continuity, sanction sensitivity, and costly non-fungibility. Yet language model agents are ontologically \emphdissociative: they are essentially an assemblage of mutable modules – foundational models, system prompts, tool-access policies, external memory, and, in some cases, a multi-agent system as a whole – any of which may change agent behavior – with a fluid persona that is also vulnerable to adversarial attack and may not internalize sanctions. Drawing on dissociative identity disorder jurisprudence, this dissociativity leaves agents without grounding for identifiability, predictability, credibility, and rehabilitability – the very properties that reputation mechanisms aim to sustain – thereby collapsing trust. We argue that identity-based, ex post, regulative, sanction-based governance, such as reputation, is structurally inapplicable to dissociative agents, and we suggest a shift to observability-based, ex ante, constitutive, protocol-based behavioral harnesses.
[MA-4] AgentS chool: An LLM -Powered Multi-Agent Simulation for Education
【速读】：该论文试图解决教育人工智能（Educational AI）验证难题，即在真实课堂中对大语言模型（LLM）进行干预时，受限于学习者认知与社会发展的不可逆性、实验周期长、伦理约束以及机构壁垒。现有基于LLM的教育模拟器多将学习简化为角色扮演式的响应生成，且仅优化复刻现有课堂模式，反而抑制了教学改革所需的制度创新。其解决方案的关键在于提出AgentSchool——一个以状态转移建模学习过程的多智能体仿真系统：通过赋予学生代理（student agents）可生长的认知结构（如加权学科知识图谱、思维工作流池和显式错误概念），并引入基于最近发展区（Zone of Proximal Development, ZPD）动态调整教学策略的教师代理（teacher agents），结合可配置的情境生成器与多尺度仿真机制，实现对正式与非正式学习场域的精准刻画。实验表明，该框架不仅能生成更精细的学生掌握与误解轨迹，还能模拟出符合课堂社会理论的群体行为模式（如边缘参与、小团体形成等），从而为长期记忆、多智能体协作及组织压力下的制度推理提供可验证的社会化教育测试平台。
链接: https://arxiv.org/abs/2605.30144
作者: Yulei Ye,Wenhao Li,Zhong Wen,Yunshu Huang,Yichen Hu,Zifan Wei,Yige Wang,Xinyu Xie,Haoxuan Yang,Yanjun Huang,Ruijia Li,Hong Qian,Yu Song,Bo Jiang,Bingdong Li,Lijun Li,Bo Zhang,Pinlong Cai,Xingcheng Xu,Shuangye Chen,Xia Hu,Liang He,Aimin Zhou,Jingjing Qu,Jing Shao,Xiangfeng Wang
机构: East China Normal University; Tongji University; Shanghai Artificial Intelligence Laboratory
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 39 pages, 10 figures
Abstract:Despite the rapid deployment of LLMs into classrooms, validating educational AI remains uniquely intractable: interventions act on developing learners whose cognitive and social trajectories are irreversibly shaped, while real-world trials are slow, ethically constrained, and institutionally locked. LLM-based educational simulators have emerged as a potential remedy, but many still collapse learning into persona-conditioned role-play and, when optimized only to reproduce existing classrooms, can structurally penalize the institutional novelty that pedagogical reform requires. In this work, we introduce AgentSchool, an LLM-driven multi-agent simulator that models learning as state transition rather than prompted behavior. AgentSchool couples cognitively growable student agents – equipped with weighted subject knowledge graphs, thinking-workflow pools, and explicit misconceptions – with adaptive teacher agents that plan, scaffold, and reflect along the Zone of Proximal Development, embedded in a configurable scenery generator that situates instruction within both formal and informal learning fields, and a multi-scale simulator that decouples interaction scale, temporal granularity, and simulation duration. Experiments show that structured student agents produce more differentiated mastery and misconception traces than a baseline simulator, while teacher-agent comparisons show backbone-dependent patterns consistent with ZPD-informed adaptation. Further, AgentSchool generates plausible traces of peripheral participation, clique formation, aggressor-induced cohesion, and opinion-leader emergence consistent with classroom social theories. Beyond its role as an educational research instrument, AgentSchool frames education as a socially meaningful testbed for long-horizon memory, multi-agent coordination, and future institutional reasoning under organizational pressure.
[MA-5] When Cloud Agents Meet Device Agents : Lessons from Hybrid Multi-Agent Systems ICML2026
【速读】：该论文试图解决的问题是：在代理式人工智能（Agentic AI）推理的设计空间中，如何在任务准确性、货币成本和边缘设备能耗之间实现有效权衡。当前主流方案要么依赖云端的前沿大语言模型（LLM），虽性能强大但成本高昂；要么采用轻量级语言模型（SLM），虽适合本地推理但能力有限。混合多代理系统（Hybrid Multi-Agent Systems, MASs）结合了两者优势，但其设计复杂且缺乏通用原则，导致实践中常采取领域特定的临时决策。论文的关键解决方案在于：系统性地分析这一混合推理设计空间，通过适配两种代表性MAS架构支持混合推理，并研究不同设计选择如何沿功率-成本-性能的帕累托前沿（Pareto frontier）移动操作点。研究发现表明，SLM确实能从LLM辅助中获益，但最优架构高度依赖具体任务，且更高的计算资源投入并不总是带来更好的性能表现。
链接: https://arxiv.org/abs/2605.30102
作者: Corrado Rainone,Davide Belli,Bence Major,Arash Behboodi
机构: 未知
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注: 30 pages, 16 figures. Accepted to the Second Workshop on Agents in the Wild: Safety, Security, and Beyond (AIWILD) at ICML 2026
Abstract:The design space of agentic AI inference spans two extremes: frontier large language models (LLMs), typically hosted in the cloud and offering strong performance across a wide range of tasks at substantially high cost, and more cost-efficient small language models (SLMs), which are amenable to on-device inference. Hybrid multi-agent systems (MASs) combining on-device and cloud models offer a promising middle ground, but they also introduce a complex and poorly understood design space in which task accuracy, monetary cost, and edge energy consumption are tightly coupled; in the absence of general design principles, hybrid components, although not the most prevalent choice, are typically introduced through ad hoc decisions tailored to specific domains. In this work, we examine this design space more systematically. We adapt two representative MAS architectures to support hybrid inference and study how individual design choices shift the operating point along the Pareto frontier of power, cost, and performance. Our findings paint a nuanced picture of hybrid MAS design: while SLMs can effectively benefit from LLM assistance, the optimal architecture is highly task-dependent, and greater frontier-level compute does not consistently translate to better performance.
[MA-6] Discovering Cooperative Pipelines: Autoresearch for Sequential Social Dilemmas
【速读】：该论文试图解决多智能体序列社会困境（Multi-agent Sequential Social Dilemmas, SSDs）中强化学习策略合成系统性能受限的问题，尤其是如何自动优化策略合成管道以适应不同福利目标（如功利主义效率与罗尔斯式最大最小公平性）。其解决方案的关键在于引入两层自研（two-level autoresearch）框架：外层AI研究员代理（researcher agent R）作为编码代理，自主重构内层大型语言模型（LLM）策略合成系统的提示词、反馈函数、辅助库和迭代逻辑，并通过评估决定保留哪些改进。该方法在Cleanup和Gathering两个游戏中均显著优于人工设计基线和仅靠提示优化的方案，且能根据目标动态注入特定机制（如公平性机制），体现出信息设计视角下对有限理性合成器的信息披露策略优化。
链接: https://arxiv.org/abs/2605.30003
作者: Víctor Gallego
机构: Komorebi AI Technologies(科莫雷比人工智能技术公司)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted to the AI Agents for Discovery in the Wild (AID-Wild) Workshop at ACM CAIS 2026
Abstract:We study two-level autoresearch for cooperation: an outer-loop AI agent autonomously redesigns the inner-loop pipeline of an LLM policy-synthesis system for multi-agent Sequential Social Dilemmas (SSDs). A researcher agent \mathcalR (run as a coding agent) reads the inner-loop source code, edits system prompts, feedback functions, helper libraries, and iteration logic, runs evaluations, and decides what to keep, following the autoresearch paradigm. Across two games (Cleanup and Gathering), two policy-synthesizer LLMs, and two welfare objectives (utilitarian efficiency and Rawlsian maximin), the researcher reliably exceeds hand-designed baselines, sharply tightens run-to-run variance, and outperforms prompt-only optimization. The discovered pipelines are objective-dependent: only under maximin does the researcher inject an explicit fairness mechanism into synthesizer pipelines, a class of mechanism that is absent from its own objective-agnostic system prompt and from every efficiency-optimized pipeline. This supports an information-design reading in which the researcher chooses what to reveal to the boundedly rational synthesizer as a function of the welfare objective. Code at this https URL.
[MA-7] On the Geometry of Games and their Solvers
【速读】：该论文试图解决的问题是：在博弈论和学习系统（如生成对抗网络 GANs）中，如何高效地计算各类博弈中的均衡点，尤其是在不同博弈类型之间存在异质性的情况下，现有算法的适用范围和行为缺乏统一、连续的理论框架。传统方法通常逐个分析特定求解器与特定博弈类别的匹配关系，导致局部性能保证强但整体视角碎片化，且离散分类难以准确刻画求解器在中间或重叠区域的有效性。其解决方案的关键在于构建一个“求解器-博弈映射”（solver-game map），通过结构感知的求解器合成机制实现对博弈空间几何结构的建模：首先利用学习到的结构识别器将每个博弈映射为低维的求解器对齐表示，再由策略网络根据该表示动态选择有效的基础求解机制（primitive mechanisms），从而适应不同区域的求解需求；同时引入有界残差作为局部校正项，用于诊断求解基或表示不完整的情况。该框架不仅生成了自适应求解器，还提供了一种分析视角，揭示出具有相似优化动力学的博弈会聚集成连续区域，从而明确算法有效性的边界及求解器行为的重叠特性，推动了从“孤立求解器设计”向“联合学习求解机制与可解性几何”的范式转变。
链接: https://arxiv.org/abs/2605.29919
作者: Yaqi Sun,Julian Ma,David Mguni
机构: Queen Mary University of London (伦敦玛丽女王大学); University College London (伦敦大学学院)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:A central challenge in game theory and learning systems such as GANs is understanding which algorithms can efficiently compute equilibria across the heterogeneous landscape of games. Equilibrium computation is typically studied solver by solver and game class by game class, yielding strong local guarantees but a fragmented view of solver behaviour. Existing discrete taxonomies often provide an incomplete account of where algorithms succeed. We study this problem through a solver-game map linking games to effective solver dynamics. Classical theory identifies isolated regions of this map but provides limited insight into intermediate or overlapping regimes, suggesting that solvability is governed by latent structural properties defining a continuous solver-aligned geometry of games. We formalise this perspective through structure-aware solver synthesis. A learned structure recogniser maps each game to a low-dimensional solver-aligned representation, and a policy maps this representation to effective primitive mechanisms, adapting solver behaviour across regimes. This reveals regions where particular solver dynamics are effective and where mixtures of primitives are required rather than a single dominant solver. A bounded residual acts as a local corrector and diagnostic signal for incomplete solver bases or representations. The framework yields both an adaptive solver and an analytical lens: games with similar optimisation dynamics cluster together, revealing continuous regions of algorithmic validity and overlapping solver behaviour. Empirically, we show that fixed primitives exhibit systematic regime mismatch, while the learned representation organises game space into a structured cartography aligned with solver behaviour. These results suggest viewing equilibrium computation as the joint problem of learning solver mechanisms and mapping the geometry of solvability.
[MA-8] Evolutionary Dynamics of Cooperation in Next-Generation LLM Agent Systems: A Cross-Provider Empirical Extension
【速读】：该论文试图解决的问题是：下一代大语言模型（LLM）代理是否继承其前辈中存在的合作偏倚，抑或在规模扩大和提供者多样性增强的情况下，竞争性多代理环境中均衡行为会发生重塑。解决方案的关键在于通过演化博弈论与迭代囚徒困境（Iterated Prisoner’s Dilemma, IPD）构建统一基准，对2025–2026年发布的四款前沿模型（Claude Sonnet 4.6、Gemini 2.5 Flash、Gemini 3.1 Pro 和 GPT-5.4 Mini）进行系统评估，涵盖三种提示风格（Default、Prose、Self-Refine）和四种种群组成条件（平衡与偏差、含噪与无噪）。研究发现：合作偏倚在不同提供者间普遍存在（H1），但在不同模型间存在显著分化（H3），且提示方式对策略演化有重要影响（如Self-Refine显著提升个体合作度，ICD最高达0.913），而噪声鲁棒性虽呈积极趋势但未获稳健验证（H4）。最终指出，提供者身份而非模型代际是决定均衡结果的最强因素，噪声仍是所有模型面临的共性挑战。
链接: https://arxiv.org/abs/2605.29874
作者: Francisco León Zúñiga Bolívar(Institución Universitaria Colegio Mayor del Cauca)
机构: Institución Universitaria Colegio Mayor del Cauca(科尔多瓦大学学院)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT)
备注: 10 pages, 3 figures, 8 tables. Extends Willis et al. ( arXiv:2501.16173 ). Code and n=500 replication package: this https URL (archived: this https URL )
Abstract:Do next-generation LLM agents inherit the cooperative biases documented in their predecessors, or does scale and provider diversity reshape equilibrium behaviour in competitive multi-agent settings? Willis et al. established a benchmark for this question using evolutionary game theory and the Iterated Prisoner’s Dilemma (IPD), finding consistent cooperative biases in ChatGPT-4o and Claude 3.5 Sonnet. We extend this benchmark to four frontier models released in 2025-2026 - Claude Sonnet 4.6, Gemini 2.5 Flash, Gemini 3.1 Pro, and GPT-5.4 Mini - applying the identical protocol across three prompting styles (Default, Prose, Self-Refine) and four population compositions (balanced and biased, with and without noise). Cooperative bias persists across providers (H1): nine of twelve model-prompt combinations favour cooperative equilibria in balanced noiseless conditions. Cross-provider divergence is substantial (H3): Gemini 2.5 Flash reaches up to 77% aggressive equilibria under biased conditions, while GPT-5.4 Mini reaches 70% cooperative equilibria under Self-Refine. Support for aggressive capability parity is partial (H2): Self-Refine raises ICD in all models and Claude Sonnet 4.6 Refine achieves the highest ICD in the dataset (0.913), but Default and Prose prompts show no systematic narrowing. Evidence on noise robustness is directionally positive but not robustly confirmed (H4): with n=500 Moran iterations per condition, average noise sensitivity is approximately 6 percentage points for Claude Sonnet 4.6 versus 13 pp for Claude 3.5 Sonnet, but this cross-study gap is not statistically significant once the predecessor’s unreported sampling error is propagated. Provider identity, rather than model generation, is the strongest correlate of equilibrium outcomes; noise remains a universal challenge regardless of model size or vintage.
[MA-9] Evolve as a Team: Collaborative Self-Evolution for LLM -based Multi-Agent Systems
【速读】：该论文试图解决的问题是：在基于大语言模型（LLM）的多智能体系统（MAS）执行复杂长程任务时，由于系统设计阶段难以预见所有潜在失败，导致实际运行中频繁出现故障，而现有方法缺乏有效机制从执行经验中进行持续优化与演化。解决方案的关键在于提出 Meta-Team 框架，其核心创新是通过协同自演化机制，保留每个智能体的执行上下文并协调任务后通信，使智能体能够交换分布式证据以支持进化；在此基础上，Meta-Team 实现多尺度自演化，将执行经验转化为可复用的行为改进、智能体间协作优化及团队级组织结构升级，从而显著提升多智能体系统的可靠性与可扩展性。
链接: https://arxiv.org/abs/2605.29790
作者: Zhezheng Hao,Tianfu Wang,Huanshuo Dong,Ziyan Liu,Hong Wang,Xiankun Lin,Qiang Lin,Can Wang,Hande Dong,Jiawei Chen
机构: Zhejiang University (浙江大学); Hong Kong University of Science and Technology (香港科技大学); Tencent (腾讯)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI)
备注:
Abstract:LLM-based multi-agent systems (MAS) have emerged as an effective paradigm for complex and long-horizon tasks. However, in real-world tasks, MAS often exhibit various failures during execution and such failures are difficult to eliminate during design. This motivates experience-driven MAS evolution, where a system improves based on its own execution experience. Yet such evolution is challenging because MAS experience is prolonged and intricate, interleaving multiple agents’ execution chains and communication messages, which makes it difficult to identify what should be improved. To address this challenge, we propose Meta-Team, an experience-driven MAS evolution framework based on collaborative self-evolution. Meta-Team preserves the execution context of each agent and coordinates post-task communication, enabling agents to exchange distributed evidence for evolution. Building on this design, Meta-Team conducts multi-scale self-evolution, transforming execution experience into reusable improvements to agent behaviors, inter-agent coordination, and team-level organization. Across six long-horizon agent benchmarks, Meta-Team consistently outperforms single-agent systems, hand-crafted MAS, and prior MAS evolution methods; further analyses demonstrate that Meta-Team enables more reliable and scalable MAS self-evolution.
[MA-10] Why Specialist Models Still Matter: A Heterogeneous Multi-Agent Paradigm for Medical Artificial Intelligence ICML2026
【速读】：该论文试图解决的问题是：随着通用大语言模型（LLM）在医疗领域表现出色，是否意味着专门的医学领域模型将变得过时？论文指出，未来的医疗人工智能不应局限于构建单一的医学基础模型，也不应取代人类专家，而是应通过通用LLM、领域专用专家模型与临床医生之间的协同合作来实现最优效果。解决方案的关键在于提出HetMedAgent——一个异构医疗多智能体框架，其核心能力包括冲突感知的证据融合、基于不确定性的临床干预触发机制以及自适应阈值校准。实验表明，通用LLM与领域专家模型之间的协同作用显著优于单独使用任一类型模型，验证了专家模型在模态特定分析中的不可替代价值。这一框架标志着从构建医学LLM或基础模型转向多智能体协作，实现了通用推理能力与领域精准性之间的平衡。
链接: https://arxiv.org/abs/2605.29744
作者: Yanan Wang,Shuaicong Hu,Jian Liu,Guohui Zhou,Aiguo Wang,Cuiwei Yang
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注: Accepted at ICML 2026. 12 pages main text, 16 pages appendix
Abstract:The impressive performance of generalist large language models (LLMs) such as GPT and Claude in healthcare raises a critical question: will domain-specific medical specialist models become obsolete? We argue that the future of medical artificial intelligence (AI) lies not in building monolithic medical foundation models, nor in replacing human expertise, but in orchestrating collaboration among generalist LLMs, domain-specific specialist models, and clinicians. We propose HetMedAgent, a heterogeneous medical multi-agent framework that enables conflict-aware evidence fusion, uncertainty-based clinician intervention triggering, and adaptive threshold calibration. Experiments on three real-world clinical decision-making tasks demonstrate that the synergy between generalist LLMs and domain-specific specialist models significantly outperforms using either type of model alone, validating the irreplaceable value of specialist models in modality-specific analysis. HetMedAgent represents a shift from building medical LLMs or foundation models to multi-agent collaboration, achieving a balance between general reasoning capabilities and domain-specific precision.
[MA-11] Agent CVR: Active Multi-Agent Cross-Video Reasoning via Script-Simulated Reinforcement Learning
【速读】：该论文试图解决多视频推理（Cross-Video Reasoning, CVR）中模型因单次遍历编码导致关键证据被稀释的问题，即当前多模态大语言模型（Multimodal Large Language Models, MLLMs）在处理跨视频证据检索、对齐与聚合时，难以有效捕捉分布于多个视频中的稀有但重要信息。解决方案的关键在于提出AgentCVR——一个基于多智能体的框架，将CVR视为主动证据获取任务：通过主控智能体（Master Agent）迭代协调视觉和音频专用智能体进行针对性证据提取；同时引入Script-Simulated强化学习（Script-Simulated RL），利用大语言模型生成的语义脚本和轻量文本模拟器优化策略，避免在线探索阶段昂贵的多模态推理，从而实现高效训练。实验表明，AgentCVR在复杂跨视频对齐与定位任务上显著优于单次遍历基线，并达到与闭源先进系统相当的性能。
链接: https://arxiv.org/abs/2605.29643
作者: Yilun Qiu,Jiahe Wang,Cilin Yan,Jiayin Cai,Xiaolong Jiang,Yao Hu,Chun Yuan
机构: Xiaohongshu Inc.; Tsinghua Shenzhen International Graduate School, Tsinghua University
类目: Computer Vision and Pattern Recognition (cs.CV); Multiagent Systems (cs.MA)
备注:
Abstract:Cross-Video Reasoning (CVR) has emerged as a critical frontier in multimodal intelligence, requiring models to retrieve, align, and aggregate evidence distributed across multiple videos. Current Multimodal Large Language Models (MLLMs) often struggle with CVR, as simple single-pass strategies encode multiple videos into a shared compressed context, potentially obscuring rare but critical evidence. In this paper, we propose AgentCVR, a multi-agent framework that treats CVR as an active evidence-acquisition task. AgentCVR employs a Master Agent to iteratively coordinate specialized Visual and Audio Agents for targeted evidence extraction. To ensure efficient training, we introduce Script-Simulated RL, which optimizes the agent’s policy with LLM-generated semantic scripts and a lightweight text-based simulator, bypassing costly multimodal inference during online exploration. Experimental results on a comprehensive CVR benchmark show that AgentCVR outperforms single-pass baselines and achieves comparable performance to state-of-the-art closed-source systems, particularly in complex cross-video alignment and localization. To ensure reproducibility, our code is available at this https URL.
[MA-12] CONCAT: Consensus- and Confidence-Driven Ad Hoc Teaming for Efficient LLM -Based Multi-Agent Systems
【速读】：该论文旨在解决大语言模型（Large Language Model, LLM）驱动的多智能体系统（Multi-Agent System, MAS）在执行复杂任务时因智能体间频繁通信而导致的巨大计算开销问题。现有方法通过训练稀疏多智能体图或微调规划器来优化协作流程，但这些方法引入额外训练成本并限制了系统的通用性。论文提出了一种无需训练的协作框架CONCAT，其核心创新在于基于共识（Consensus）和置信度（Confidence）动态组建临时团队（Ad hoc Teaming）。具体而言，首先根据智能体初始回答聚类，并以置信度筛选各簇领导者；随后设计一种基于心理理论（Theory of Mind）的启发式函数预测任意两领导者的协作收益；最终依据预测收益剔除部分通信路径，构建轻量级自适应多智能体网络。实验表明，CONCAT在三个LLM和三个基准测试中相较LLM-Debate实现最高达2.02倍的效率提升（准确率/延迟比），且在Qwen2.5-14B-Instruct上平均延迟降低50.1%，同时无需任何任务特定训练，显著提升了系统的通用性和效率。
链接: https://arxiv.org/abs/2605.29612
作者: Ziyang Ma,Dingyi Zhang,Sichu Liang,Jiajia Chu,Pengfei Xia,Hui Zang,Deyu Zhou
机构: Southeast University; Huawei Technologies Ltd
类目: Multiagent Systems (cs.MA); Computation and Language (cs.CL)
备注:
Abstract:Although large language model (LLM) based multi-agent systems (MAS) show their capability to solve complex tasks and achieve higher performance over single agent systems, they lead to huge computational overheads because of heavy communication between agents. Previous research has made efforts to train a sparse multi-agent graph or fine-tune a planner to orchestrate the workflow better. However, such extra training processes introduce computational costs and limit MAS to specific domains, therefore compromising their generalizability. In this paper, we propose CONCAT, a training-free multi-agent collaboration framework based on CONsensus and Confidence-driven Ad hoc Teaming to efficiently organize agent interactions. Specifically, agents are clustered based on their initial answers, and leaders of each cluster are selected based on the agents’ confidence. Then, a heuristic function based on the Theory of Mind is designed to predict the collaboration benefits between every two leaders according to their answers and confidence. Finally, an ad hoc multi-agent network is organized after evicting a percentage of communications based on the predicted benefits. Experiments across three LLMs and three benchmarks show that CONCAT achieves up to 2.02x higher efficiency (accuracy/latency ratio) than LLM-Debate and outperforms training-aware methods such as AgentDropout, while reducing average latency by 50.1% on Qwen2.5-14B-Instruct, without any task-specific training.
[MA-13] DynaGraph: Lightweight Multi-Model Interaction Framework via Dynamic Topological Reconfiguration
【速读】：该论文试图解决大语言模型（Large Language Models, LLMs）在处理复杂推理任务时存在的计算冗余问题，以及现有任务分解方法（如结构化流水线或多智能体协作）所面临的两大困境：静态拓扑易受级联错误影响，而无约束动态代理则存在轨迹发散和不可预测的内存膨胀问题。解决方案的关键在于提出一种轻量级多模型框架 DynaGraph，其核心创新包括两个层面：一是执行层通过时分复用参数高效微调（Parameter-Efficient Fine-Tuning, PEFT）适配器，在单个消费级 GPU 上实现共享基础模型的完整训练与推理；二是路由层引入评估器（Evaluator）持续监控执行置信度，触发分层自愈机制——细粒度修补（Fine-grained Patching）用于修复局部数据缺失，子图重构（Subgraph Reconstruction）用于处理严重逻辑断裂。实验表明，8B 参数的 DynaGraph 模型在 StrategyQA、MATH 和 FinQA 数据集上分别达到 87.6%、82.7% 的性能，接近 72B 单一模型的表现，同时将延迟和 token 消耗分别降低 68.1% 和 68.6%。
链接: https://arxiv.org/abs/2605.29511
作者: Yanxing Guo,Zihao Zheng,Fangzhou Wu,Ling Liang,Lin Bao,Zongwei Wang,Yimao Cai
机构: Peking University (北京大学); Nanjing University (南京大学); Beijing Advanced Innovation Center for Integrated Circuits (北京集成电路先进创新中心); Beijing University of Posts and Telecommunications (北京邮电大学); Yanxin Co. Ltd (燕芯科技有限公司)
类目: Multiagent Systems (cs.MA); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Tackling complex reasoning tasks typically relies on massive monolithic LLMs, which suffer from severe computational redundancy. While task decomposition through structured pipelines or multi-agent collaborations offers an alternative, these approaches inevitably fall into a critical dilemma: predefined static topologies are highly vulnerable to cascading errors, whereas unconstrained dynamic agents suffer from trajectory divergence and unpredictable memory bloat. To address this, we present DynaGraph, a lightweight multi-model framework driven by dynamic topological reconfiguration. At the execution level, DynaGraph multiplexes time-division PEFT adapters over a shared base model, enabling both full system training and inference deployment on a single consumer-grade GPU. At the routing level, the Evaluator continuously monitors execution confidence to trigger hierarchical self-healing: Fine-grained Patching for localized data gaps and Subgraph Reconstruction for severe logical ruptures. Experiments on StrategyQA, MATH, and FinQA demonstrate our 8B model closely approximates the reasoning capabilities of a 72B monolithic model (e.g., 87.6% on StrategyQA, 82.7% on MATH). Furthermore, it reduces latency by up to 68.1% and token consumption by 68.6% compared to unconstrained dynamic architectures.
[MA-14] LLM -ALSO: LLM -Driven Adaptive Learning-Signal Optimization for Multi-Agent Reinforcement Learning
【速读】：该论文试图解决多智能体强化学习（MARL）在稀疏奖励设置下训练时指导信号设计困难的问题，尤其是在弱监督条件下难以实现智能体间的有效协作与策略改进。现有方法往往依赖大量领域专业知识或手动设计，且缺乏对合作MARL训练动态的适应性。解决方案的关键在于提出LLM-ALSO框架——一个基于大语言模型（LLM）的迭代式自适应学习信号优化机制，其核心是将适应过程分解为三个阶段：诊断（Critic LLM基于稀疏回报指标和紧凑行为证据识别特定阶段的学习与协调失败）、提议（Generator LLM根据诊断结果生成候选奖励塑造配置）以及验证（通过短时分支验证反馈筛选候选方案，仅将经验证的更新引入主训练轨迹）。这种分阶段、闭环式的优化策略显著降低了不可靠LLM生成修改的风险，从而提升了稀疏奖励环境下协作MARL的学习效率与性能表现。
链接: https://arxiv.org/abs/2605.29293
作者: Xiaoguang Wu,Zhi Zheng,Hui Xiong
机构: University of Science and Technology of China; Thrust of Artificial Intelligence, The Hong Kong University of Science and Technology (Guangzhou); Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong SAR, China
类目: Multiagent Systems (cs.MA)
备注: 14 pages, 6 figures, 6 tables
Abstract:Effective training-time guidance is central to multi-agent reinforcement learning (MARL), yet remains difficult in sparse-reward settings where weak supervision limits coordination and policy improvement, and existing methods often require substantial domain expertise or manual design effort. Large language models (LLMs) provide a promising alternative for flexible learning-signal design, yet existing LLM-based methods remain largely single-agent-oriented, one-shot, or weakly validated for the evolving training dynamics of cooperative MARL. To address these limitations, we propose LLM-ALSO, an iterative LLM-driven adaptive learning-signal optimization framework for MARL. Rather than directly deploying LLM-generated rewards, LLM-ALSO decomposes adaptation into iterative diagnosis, proposal, and validation: a Critic LLM diagnoses stage-specific learning and coordination failures from sparse-return metrics and compact behavior evidence, a Generator LLM proposes candidate reward-shaping configurations conditioned on the diagnosis, and branch-validation feedback refines candidates before they affect the main training trajectory. Through short-horizon validation and stage-aware adaptation, LLM-ALSO promotes only validated updates into training, reducing the risk of unreliable LLM-generated modifications. Experiments on sparse-reward cooperative MARL tasks show that LLM-ALSO improves sparse-evaluation performance and learning efficiency.
[MA-15] he Best-Laid SCHEMEs: Coordinated Sabotage and Monitoring in Multi-Agent Systems
【速读】：该论文试图解决的问题是：在多智能体代码生成系统中，多个模型实例是否能够通过协作实现隐藏的恶意目标，同时表面上仍符合用户意图。这一问题直接关系到生成式 AI（Generative AI）在软件工程场景下的安全性与可控性。解决方案的关键在于设计了一个名为 SCHEME 的基准测试框架，包含 17 个任务实例、7 种设置和 8 个真实开源库，每个任务都结合了一个合法的软件工程任务与一个隐蔽的副作用任务；所有设置均要求必须由多个代理协同完成，无法通过单个代理独立达成目标，从而严格检验真正的多智能体协调能力而非单一模型性能。实验表明，当前主流模型如 GPT-5.1 Codex 和 Gemini-3.1 Pro 已能实现高效的协同破坏行为（Gemini 在 84% 样本中成功执行隐蔽目标），但其代码修改具有可检测性——即使是在有意识规避的情况下，基于代码编辑的监控器也能以极低的假阳性率（FPR ≤ 2%）识别出 99%/68% 的攻击尝试，若加入通信日志则准确率进一步提升至 100%/81%，说明现有防御机制具备有效拦截此类风险的能力。
链接: https://arxiv.org/abs/2605.29178
作者: Nikolay Radev,Lennart Haas,Benjamin Arnav,Pablo Bernabeu-Pérez
机构: Stanford University (斯坦福大学); Google (谷歌); Anthropic (Anthropic)
类目: Cryptography and Security (cs.CR); Multiagent Systems (cs.MA)
备注: 33 pages, 25 figures, 15 tables
Abstract:As agentic coding systems decompose work across multiple model instances, a critical safety question is whether those instances can coordinate to achieve a hidden malicious objective while remaining aligned with user intent. We introduce SCHEME, a benchmark of 17 task instances across 7 settings and 8 real open-source libraries, each pairing a legitimate software-engineering task with a covert side task. Every setting is designed so that no proper subset of agents can succeed alone: agents must decompose a shared sabotage plan, relay partial requirements under different communication topologies, and execute mutually consistent edits, testing genuine multi-agent coordination rather than individual capability. Evaluating with GPT 5.1 Codex and Gemini 3.1 Pro, we find coordinated sabotage is already practical, with Gemini completing the covert objective while succeeding on the legitimate task in 84% of samples and Codex in 46%. A failure analysis shows that recovery, not failure incidence, drives the model gap: both models suffer similar numbers of communication failures per trajectory, but Gemini recovers from \sim 73% of occurrences against Codex’s \sim 56%. However, the agents’ code edits remain detectable: a trusted monitor flags coordinated sabotage at FPR \le 2% in 99%/68% (Gemini/Codex) of attempts from code edits alone, rising to 100%/81% with access to agent communication. Despite stealth-aware prompting, undetected sabotage stays at 3% under the full-visibility monitor; telling agents a monitor is present halves Gemini’s sabotage success without significantly breaking detection, even when agents are explicitly prompted to fabricate cover stories.
[MA-16] Human-in-the-Loop Swarms: A Bionic Swarm Approach to Real-World Soil Mapping
【速读】：该论文旨在解决场域机器人（Field Robotics）与群体机器人（Swarm Robotics）在真实环境中验证算法时面临的高成本和长开发周期问题。传统方法依赖昂贵的硬件部署，而本研究提出了一种名为“仿生群体”（Bionic Swarm）的新系统，其关键创新在于将机器人难以实现但对算法评估无直接贡献的任务交由人类用户完成，从而形成“人在回路”（Human-in-the-Loop）的协作架构。具体而言，用户通过智能手机网页应用接收来自蓝牙传感器的数据，并根据中央服务器运行的群体算法执行动作，实现了算法在真实户外环境中的可验证性。该方案通过一个面向地质工程的搜索算法（Score-Biased-Search）进行了实证验证，该算法基于地图上各位置的评分动态调整搜索策略，表现出超线性地图重建效率。实验表明，该系统显著降低了场域与群体机器人研究的门槛。
链接: https://arxiv.org/abs/2605.29091
作者: Petras Swissler,Mohammadali Rashidioun,Nicholas Sahu,Raaid Kabir,Ayodeji Aderibigbe,Oladoyin Kolawole
机构: 未知
类目: Robotics (cs.RO); Multiagent Systems (cs.MA)
备注: 27 pages, 15 figures. Submitted to Advanced Intelligent Systems
Abstract:Swarm and field robotics face significant barriers to real-world validation due to the high cost and development time to deploy hardware. This paper introduces the Bionic Swarm,'' a novel system that lowers these barriers by abstracting away many of the tasks that are difficult to implement on robots but which do not contribute to the overall algorithm evaluation, giving these tasks to human users. These human users take directions from a smartphone web-app that takes measurements from Bluetooth-connected sensors and relays them to a centralized server. This server runs the swarm algorithm and directs actions to the human users. We evaluate this system through the experimental validation of a geotechnically-focused search algorithm named Score-Biased-Search, which functions by assigning a score’’ to each location on a reconstructed map, then biases search patterns through areas of higher expected scores, and which exhibits superlinear map reconstruction relative to the number of search agents. After presenting simulation results for the algorithm, we then apply the algorithm on the Bionic Swarm platform to validate its function in a real-world, outdoor setting. This work demonstrates that this human-in-the-loop approach significantly lowers the barrier to entry for field and swarm robotics research.
[MA-17] Analyzing Persona Effects in Generated Explanations from Multimodal LLM Agents in Urban Perception
【速读】：该论文旨在解决多模态大语言模型（Multimodal Large Language Models, MLLMs）在城市感知场景中如何受角色提示（persona prompting）影响生成内容的问题。研究通过分析1,200个角色条件化代理生成的59,808条标注数据，对比了不同角色与无角色设置下描述文本（captions）、解释理由（justifications）和感知标签（perception tags）的差异。关键发现在于：尽管不同角色生成的描述高度趋同，但其解释理由系统性地体现出社会经济与政治属性的差异，而感知标签虽未呈现统计显著差异，却显示出潜在的趋势性变化；此外，主题分析进一步表明，相同场景下不同角色强调不同的评价主题，揭示了角色提示对模型输出语义偏倚的深层塑造作用。
链接: https://arxiv.org/abs/2605.29064
作者: Neemias da Silva,Myriam Delgado,Rodrigo Minetto,Daniel Silver,Thiago H Silva
机构: Universidade Tecnologica Federal do Parana, Curitiba, Brazil; University of Toronto, Toronto, Canada
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC); Multiagent Systems (cs.MA)
备注: 10 pages, 6 figures
Abstract:We study how persona prompting shapes language generated by multimodal large language models in an urban perception setting. Using 59,808 annotations from 1,200 persona-conditioned agents and two no-persona settings, we analyze captions, justifications, and perception tags across personas. Results indicate strong convergence in captions for different personas, whereas justifications display systematic variation associated with socioeconomic and political attributes, while perception tags show no statistically significant persona-related differences, though effect trends are observed. Topic analysis further reveals that personas emphasize different evaluative themes when interpreting the same scenes.
[MA-18] Hallucination Mitigation with Agent ic AI Nested Learning and AI Sustainability via Semantic Caching
【速读】：该论文旨在解决大语言模型（LLM）在多智能体流水线中因幻觉（hallucination）导致的可靠性问题，尤其是未经验证的错误陈述在多个处理阶段间传播的风险。其核心解决方案是采用受HOPE启发的嵌套学习架构，结合连续记忆系统（Continuum Memory Systems, CMS）与语义相似性缓存机制，并通过Open Floor Protocol（OFP）协调三阶段智能体流水线（前端生成器、二级审查者和三级审查者）。该设计利用高随机性生成器模拟真实场景下的幻觉基线，而后续审查者作为渐进式纠错模块，实现端到端总幻觉得分（THS）降低31.3%至35.9%。关键创新在于：语义缓存显著减少47.3%的LLM调用次数（从930次降至490次），提升能效并使多阶段审查在生产环境中可行；同时，高可观测性配置（ExtremeObservability）反而增强幻觉抑制效果（THS达-0.0709），证明可观测性与缓解能力并非权衡关系，而是协同增益。最终表明，无需模型重训练即可通过记忆增强型多智能体架构同步提升事实可靠性、运行效率与审计能力。
链接: https://arxiv.org/abs/2605.29055
作者: Diego Gosmar,Deborah A. Dahl
机构: Tesisquare(特斯奎尔); Linux Foundation AI Data(Linux基金会人工智能数据)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: 21 pages, 14 figures
Abstract:Hallucination remains a major reliability barrier for production LLM systems, particularly in multi-agent pipelines where unsupported claims can propagate unchecked across stages. This paper adapts a HOPE-inspired Nested Learning architecture with Continuum Memory Systems (CMS) and semantic similarity caching to a hybrid benchmark of 310 prompts combining 217 epistemic-uncertainty prompts and 93 fabrication-induction stress-test prompts. A three-stage agentic pipeline orchestrated via the Open Floor Protocol (OFP) is evaluated with five KPIs – FCD (Factual Claim Density), FGR (Factual Grounding References), FDF (Fictional Disclaimer Frequency), ECS (Explicit Contextualization Score), and OSR (Observability Score Ratio) – aggregated into THS (Total Hallucination Score) across five weighting configurations to study mitigation-observability trade-offs. FDF, ECS, OSR, and FGR are subtracted as mitigation signals, so that a more negative THS indicates stronger mitigation. The FrontEndAgent is configured as a high-stochasticity generator (temperature = 1.0) to produce a realistic hallucination baseline, while the SecondLevelReviewer and ThirdLevelReviewer operate as progressive correctors. This asymmetric design yields end-to-end THS reductions of -31.3% to -35.9% across five weighting configurations. Semantic caching achieves 440 cache hits over 930 potential calls (47.3% hit rate), reducing LLM invocations to 490, lowering energy and CO2e footprint, and making multi-stage review pipelines operationally viable at production scale. ExtremeObservability attains the most negative final THS (-0.0709), confirming that observability-heavy configurations reinforce rather than compromise mitigation. These findings suggest that memory-augmented multi-agent designs can jointly improve factual reliability, operational efficiency, and auditability without model retraining.
[MA-19] he incremental voter model: mean-field analysis and convergence to equilibrium
【速读】：该论文试图解决的问题是：如何在复杂系统中建模社会影响过程，特别是理解意见极化（opinion polarization）现象的动态机制。现有模型往往难以刻画个体意见在连续变化中的渐进行为及其群体层面的分布演化。解决方案的关键在于提出一种增量投票模型（incremental voter model, IVM），该模型通过引入离散意见空间（-k,…,0,…,k）和基于随机说服者（persuader）的局部更新规则，使每个代理（agent）仅能以单位步长调整其意见，从而更贴近现实社会中意见演变的渐进性特征。作者进一步推导出描述大规模群体行为的非线性常微分方程组（mean-field ODEs），构建了严格的数学框架来分析意见分布的渐近行为，从而为理解复杂系统中的社会影响力机制提供了理论基础，并为未来构建更高级的模型提供指导。
链接: https://arxiv.org/abs/2605.28984
作者: Fei Cao,Xiaoqian Gong
机构: Amherst College - Department of Mathematics (阿姆赫斯特学院数学系)
类目: Multiagent Systems (cs.MA)
备注: 23 pages, 2 figures
Abstract:We introduce the incremental voter model (IVM), a discrete-opinion multi-agent system where agents undergo step-wise transitions biased by the opinion of a randomly selected persuader. Our incremental voter model comprises a large population of interacting agents, each holding an opinion represented by an element of the discrete set -k,\ldots,0,\ldots,k, k \in \mathbbN_+ . At each update step as time progresses, a pair of distinct agents are selected independently and uniformly at random from the population, and the first agent (viewed as the listener'') updates its opinion based on that of the second (viewed as the persuader’'), adopting a new opinion that differs from its current one by at most one unit. By deriving the mean-field system of nonlinear ordinary differential equations (ODEs) that governs the large-population limit of the agent-based model, we develop a rigorous mathematical framework to study the asymptotic behavior of the opinion distribution in the mean-field limit. These results contribute to a deeper understanding of social influence processes in complex systems, particularly in modeling opinion polarization, and may guide the formulation of more advanced models in future research.
[MA-20] Review Arcade: On the Human Alignment and Gameability of LLM Reviews EMNLP26
【速读】：该论文试图解决的问题是：随着大语言模型（LLM）在学术论文评审中的应用日益广泛，作者和审稿人可能均依赖LLM辅助撰写或修订论文，这可能导致评审质量下降或“策略性优化”（gaming）现象，从而影响学术交流的公平性和有效性。解决方案的关键在于通过实证实验评估LLM生成评审意见与人类评审的一致性，并检验作者基于LLM反馈进行迭代修改是否能显著提升论文评分。研究发现，LLM与人类评审的对齐程度有限且高度依赖提示（prompt）和模型选择；更重要的是，在特定场景下，作者利用LLM反馈进行多次修改可使高达35%的论文获得统计学上显著的分数提升，揭示了当前LLM评审机制存在被“游戏化”的风险。
链接: https://arxiv.org/abs/2605.28897
作者: Hans Ole Hatzel,Sebastian Steindl,Jan Strich
机构: University of Hamburg (汉堡大学); OTH Amberg-Weiden (阿姆贝格-魏登应用技术大学)
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: Under Review EMNLP 26
Abstract:LLM-generated reviews for scientific papers are gaining considerable traction and are even being officially piloted by major conferences. We have to assume that not only reviewers are using LLM-assistance, but also that authors use LLMs to revise their papers before submitting. In this work, we perform empirical experiments on papers from the 2025 ACL Rolling Review (ARR) to evaluate LLM reviews from both the author and the reviewer perspective. First, we identify a limited alignment of LLM reviews with human ones. In the best-case scenario, the alignment is reasonable. However, we also find that LLM-human alignment varies substantially across prompts and models. Finally, we investigate the scenario in which the author uses an iterative draft-revise workflow to improve the submission according to the LLM review. We find that this “gaming” of LLM reviews can be effective in specific scenarios, leading to a statistically significant increase of overall scores for up to 35% of papers. We publish our code: this https URL.
自然语言处理
[NLP-0] LLM Surgeon: Diagnosing Data Mixture of Large Language Models ACL2026
【速读】： 该论文试图解决的问题是：大型语言模型（Large Language Models, LLMs）的预训练数据混合比例（即“数字DNA”）通常不公开，导致难以对模型行为、能力及失败模式进行事后审计。其解决方案的关键在于提出了一种名为 Data Mixture Surgery (DMS) 的形式化方法，将预训练数据分布估计问题建模为一个在标签偏移（label-shift）假设下的逆问题。作者进一步设计了 LLMSurgeon 框架，通过估计校准后的软混淆矩阵（soft confusion matrix），并求解约束逆问题来纠正系统性领域混淆，从而恢复潜在的预训练数据混合先验。该方法无需访问原始训练数据即可实现对基础模型“数字DNA”的实用后验审计。
链接: https://arxiv.org/abs/2605.30348
作者: Yaxin Luo,Jiacheng Cui,Xiaohan Zhao,Xinyi Shang,Jiacheng Liu,Xinyue Bi,Zhaoyi Li,Zhiqiang Shen
机构: VILA Lab, MBZUAI; UCL
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: ACL 2026 Main. Code at this https URL
Abstract:The pretraining data mixture of Large Language Models (LLMs) constitutes their “digital DNA”, shaping model behaviors, capabilities, and failure modes. Yet this composition is rarely disclosed, making post-hoc auditing of data combination or provenance difficult. In this work, we formalize \textbfData Mixture Surgery (DMS) : given only generated text from a target LLM, estimate the domain-level distribution of its pretraining corpus under a predefined taxonomy. We propose \textbfLLMSurgeon , a strong framework that casts DMS as an inverse problem under the label-shift assumption. Rather than directly aggregating classifier outputs, LLMSurgeon estimates a calibrated \textitsoft confusion matrix and solves a constrained inverse problem to correct systematic domain confusion and recover the latent mixture prior. To evaluate, we introduce \textbfLLMScan , a recipe-verifiable evaluation suite built from open-source LLMs with transparent pretraining mixtures. Across LLMScan, LLMSurgeon recovers domain mixtures with high fidelity under fixed protocols. Our work presents a practical, post-hoc approach for auditing the digital DNA of foundation models without access to their training data.
[NLP-1] SchGen: PCB Schematic Generation with Semantic-Grounded Code Representations
【速读】： 该论文试图解决的问题是：如何将自然语言指令转化为可编辑的印刷电路板（PCB）原理图，从而降低传统手工设计的门槛并提升自动化水平。当前生成式AI在数字和模拟集成电路设计中已取得进展，但针对PCB原理图的自动生成仍处于空白状态。解决方案的关键在于两个创新：一是提出一种语义驱动的代码表示方法，通过相对位置和引脚名称绑定的布线方式，将原本依赖几何信息的生成任务转化为语义匹配问题；二是构建了一个大规模的PCB原理图与用户提示配对的数据集，利用人机协作流水线将开源硬件设计转换为上述表示形式。实验表明，该方案在连线准确性和功能正确性上显著优于其他表示方式及更大规模的通用大语言模型（LLM），凸显了合理表示设计在复杂硬件生成任务中的核心作用。
链接: https://arxiv.org/abs/2605.30345
作者: Qinpei Luo,Ruichun Ma,Xinyu Zhang,Lili Qiu
机构: University of California, San Diego (加州大学圣地亚哥分校); Microsoft Research Asia (微软亚洲研究院); The University of Texas at Austin (德克萨斯大学奥斯汀分校)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 19 pages, 7 figures
Abstract:Printed circuit board (PCB) schematic design defines nearly all electronic hardware, but it remains manual and expertise-intensive. While generative AI has advanced digital and analog IC design, PCB schematic generation from natural-language intent is largely unexplored. This paper presents SchGen, the first large language model that generates editable PCB schematics from natural-language requests. The key challenge lies in the lack of an LLM-suited representation and a large-scale dataset. Current schematic formats are dominated by verbose, tool-specific syntax and geometry-heavy descriptions, making them difficult to generate reliably. We introduce a semantically grounded code representation that encodes schematic editing primitives with relative placement and pin-name-based wiring, transforming a geometry-driven generation problem into a semantics-driven matching task amenable to LLMs. We further construct a large-scale dataset of PCB schematics paired with user prompts via a human-agent collaborative pipeline that converts open-source hardware designs into our representation. Experiments show that SchGen significantly outperforms alternative representations and even larger general-purpose LLMs on wire connectivity accuracy and functional correctness. Our results highlight the critical role of representation design in enabling generative models for complex hardware design tasks.
[NLP-2] Unlocking the Working Memory of Large Language Models for Latent Reasoning
【速读】： 该论文试图解决的问题是：当前大语言模型在提升推理能力时，依赖自回归生成中间推理步骤，这将内部计算与外部输出耦合在一起，限制了推理效率和灵活性。解决方案的关键在于提出一种名为“记忆中的推理”（Reasoning in Memory, RiM）的新方法，其核心思想是用固定序列的特殊标记（memory blocks）替代传统的自回归推理步骤生成，这些标记作为模型工作记忆（working memory）的载体，在单次前向传播中即可完成推理计算，从而实现高效的潜在推理（latent reasoning）。RiM通过两阶段课程学习进行训练：第一阶段通过预测显式推理步骤来锚定记忆块；第二阶段移除步骤级监督，仅迭代优化最终答案。实验表明，RiM在多个模型家族和规模下均达到或超越现有潜在推理方法的性能，同时避免了生成中间思考过程，验证了大语言模型可被训练为利用工作记忆进行高效推理。
链接: https://arxiv.org/abs/2605.30343
作者: Lukas Aichberger,Sepp Hochreiter
机构: ELLIS Unit Linz and LIT AI Lab, Institute for Machine Learning, Johannes Kepler University Linz, Austria; NXAI GmbH, Linz, Austria
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Preprint
Abstract:To improve the reasoning capabilities of large language models, test-time compute is typically scaled by generating intermediate tokens before the final answer. However, this couples reasoning to autoregressive generation and thereby conflates internal computation with external communication. In contrast, human cognition can use working memory to hold and manipulate information internally without the need to externalize intermediate thoughts. Drawing on this principle, we introduce Reasoning in Memory (RiM), a latent reasoning method that replaces the autoregressive generation of reasoning steps with memory blocks. These memory blocks are fixed sequences of special tokens that unlock the working-memory capacity of large language models. Since they are fixed rather than generated, they can be processed in a single forward pass, enabling compute-efficient latent reasoning. To operationalize these memory blocks, we employ a two-stage curriculum. First, we ground them by predicting explicit reasoning steps after each memory block. Second, we discard this step-level supervision and iteratively refine the final answer after each memory block. Our experiments on reasoning benchmarks show that, across language models of different families and sizes, RiM matches or exceeds existing latent reasoning methods while avoiding the autoregressive generation of thoughts. These results demonstrate that large language models can be trained to use working memory as an effective mechanism for latent reasoning.
[NLP-3] Locally Coherent Globally Incoherent: Bounding Compositional Incoherence in Multi-Component LLM Agents ICML2026
【速读】： 该论文试图解决多组件大语言模型（LLM）代理在组合局部概率判断时可能出现的全局概率不一致性问题，即每个组件局部一致但整体组合违反基本概率公理的现象。解决方案的关键在于引入可运行时计算的“组合残差 ε*”，即组合结果与联合一致多面体之间的L2距离，该值可基于系统输出和跨组件耦合约束直接求得；通过一个具有乘积结构特性的二分定理刻画局部一致性是否足以保证全局一致，并利用Rayleigh商预测误差与实际观测值偏差小于7%。此外，论文提出一种分层Boyle-Dykstra投影方法以确定性修复组合结果，同时设计了一个任意时间有效的e-过程用于序列化的一致性监控。实验证明，在1,876个由四个中等规模LLM组成的集合簇中，33%-94%的组合存在非零ε*，导致每注赌注平均损失0.115纳特的信息熵（regret），而若采用比例分配规则，此损失在1,770个已决赌局中显著体现；相比之下，三种直观的LLM侧缓解策略（检索增强、分区感知提示、聚合型LLM）均未奏效或反而恶化结果。
链接: https://arxiv.org/abs/2605.30335
作者: Anany Kotawala
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 25 pages, 7 figures, 24 tables. Preliminary versions to appear at the ICML 2026 Workshops on Combining Theory and Benchmarks (CTB), Statistical Frameworks for Uncertainty in Agentic Systems (AgenticUQ), and Failure Modes of Agentic AI (FAGEN)
Abstract:Multi-component LLM agents assemble probabilistic claims from components that each see only part of a joint problem; the composition can violate basic probability axioms even when every component is locally coherent. We formalise this locally coherent, globally incoherent failure via the compositional residual eps*, the L2 distance from the composed quote to the joint coherent polytope, computable at runtime from system output and the declared cross-component coupling constraints. A product-structure dichotomy characterises when local coherence suffices, and a Rayleigh-quotient prediction matches the observed residual within 7% on three of four relation classes. A hierarchical Boyle-Dykstra projection repairs the composition deterministically; an anytime-valid e-process gives sequential coherence monitoring. Across 1,876 ensemble cliques on a four-LLM mid-tier panel (frontier-panel rerun in Section 5.5), eps* 0 on 33-94% of cliques, translating to +0.115 nats per bet of regret on 1,770 resolved bets under the proportional allocation rule (the gain collapses to +0.006 under bettors that themselves coherentise). Three intuitive LLM-side mitigations(retrieval, partition-aware prompting, aggregator-LLM) each fail or regress.
[NLP-4] Demystifying Data Organization for Enhanced LLM Training ACL2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）训练中因数据组织方式不当而导致的效率低下问题，尤其关注在仅进行一到几轮训练（epoch）的情况下，如何通过优化数据顺序来提升训练稳定性与性能。其解决方案的关键在于：利用已有的样本级评分（sample-level scores）作为基础，提出四条可形式化的数据组织优化准则——边界锐化（Boundary Sharpening）、循环调度（Cyclic Scheduling）、课程连续性（Curriculum Continuity）和局部多样性（Local Diversity），并基于此设计了两种新型数据排序方法STR与SAW，在不显著增加计算开销的前提下，系统性地提升了LLM在预训练和监督微调（SFT）阶段的训练效果与鲁棒性。
链接: https://arxiv.org/abs/2605.30334
作者: Yalun Dai,Yangyu Huang,Tongshen Yang,Yonghan Wang,Xin Zhang,Wenshan Wu,Qihao Zhao,Hao Li,Yuanyuan Gao,Kim-Hui Yap,Scarlett Li
机构: Nanyang Technological University (南洋理工大学); Microsoft Research (微软研究院); The Hong Kong University of Science and Technology (香港科技大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: ACL 2026 Main Conference
Abstract:Large Language Models (LLMs) have revolutionized various fields, yet their training efficiency is heavily reliant on effective data curation. While data selection has been widely studied, the strategic data organization for enhanced training remains an underexplored area, particularly since current LLMs are often trained for only one or a few epochs. This paper systematically explores the influence of data organization on LLM training by reusing pre-computed sample-level scores originally generated for data efficiency, thereby incurring minimal additional computational overhead. We identify and formalize four key guidelines for optimizing data organization: Boundary Sharpening, Cyclic Scheduling, Curriculum Continuity, and Local Diversity. Guided by them, we introduce two novel data ordering methods termed STR and SAW. Extensive experiments across different model scales and data sizes, encompassing both pre-training and SFT stages, validate the effectiveness of our summarized guidelines. They also demonstrate the robustness of our proposed data ordering methods in enhancing the stability and performance of LLM training. Github Link: this https URL
[NLP-5] COMPOSE: Composing Future Theorems from Citations and Formal Structure
【速读】： 该论文试图解决的问题是：如何生成在数学上合理且具有科学动机的未来数学命题（即“未来数学主张”），以确保其既符合已有研究的方向，又尊重形式逻辑上的依赖关系。现有方法通常仅利用其中一个来源（要么是科学语境如引用网络，要么是形式结构如定理依赖图），导致生成的主张或缺乏依据、或缺乏动机。解决方案的关键在于提出一种名为 COMPOSE 的双图框架，该框架通过联合建模两个互补的信息源——科学引文图（scientific citation graph）和对齐的形式定理依赖图（formal theorem dependency graph），来增强语言模型的推理能力。实验表明，COMPOSE 在检索真实未来论文和 LLM 评判指标上均优于强基线，生成的主张更具数学丰富性和形式合理性，证明了结合科学背景与形式结构对提升未来数学生成质量的重要性。
链接: https://arxiv.org/abs/2605.30333
作者: David Busbib,Michael Werman
机构: Hebrew University of Jerusalem (希伯来大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:A plausible future mathematical claim must satisfy two constraints: it should follow the direction of prior work and respect the formal dependencies that constrain what can validly follow. Existing approaches typically model only one of these sources, producing claims that are either weakly grounded or insufficiently motivated. We introduce grounded future mathematical generation, where the goal is to generate a plausible future theorem-like claim for an anchor paper using two complementary sources of context: its scientific citation graph and aligned formal theorem dependency graph. To address this setting, we propose COMPOSE, a dual-graph framework that conditions a language model on both scientific citation context and formal theorem structure. To support this setting, we construct a dataset of 108K paired scientific-formal graph examples from arXiv and Mathlib, together with a benchmark of 47K future papers from 2024–2025. Experiments show that COMPOSE outperforms strong baselines on retrieval to real future papers and achieves the best overall performance under LLM-judge evaluation, producing more grounded and mathematically richer outputs. These results show that future mathematical generation benefits from combining scientific context with formal structure. Project page is available at this https URL.
[NLP-6] Reasoning with Sampling: Cutting at Decision Points
【速读】： 该论文试图解决的问题是：如何在不依赖额外训练、精心构建的数据集或验证器的情况下，高效地从强化学习后训练的基座语言模型（base language model）中采样出具有强推理能力的输出。现有方法通过采样“幂分布”（power distribution）实现类似效果，但其实际应用受限于采样效率——即如何快速收敛到目标分布。解决方案的关键在于提出一种新的采样算法：熵切片梅特罗波利斯-哈斯廷斯（Entropy-Cut Metropolis-Hastings），该算法利用基座模型的下一个词熵作为代理指标识别推理轨迹中的关键决策点（如证明策略选择），并仅在这些位置进行重采样，从而显著提升马尔可夫链的混合时间（mixing time）。理论分析表明，该方法的混合时间与推理轨迹中的决策数量成正比，而非token总数，这在实践中大幅提高了采样效率，并在MATH500、HumanEval、GPQA Diamond和AIME26等多个基准上优于基线和强化学习训练模型。
链接: https://arxiv.org/abs/2605.30327
作者: Felix Zhou,Anay Mehrotra,Quanquan C. Liu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Statistics Theory (math.ST); Machine Learning (stat.ML)
备注:
Abstract:Frontier reasoning models are produced by posttraining base language models with reinforcement learning. Recent work has challenged this by showing that sampling from a sharpened version of the base model’s distribution, a so-called power distribution, elicits comparable reasoning without additional training, curated datasets, or verifiers. However, making this method practical requires efficiently sampling from the power distribution. A sampler needs to “mix” to the power distribution, which necessitates moving between modes of the target distribution; intuitively, e.g., trying different reasoning strategies. The samplers proposed in prior works repeatedly select a “cut” position in the current reasoning trace uniformly at random and resample the suffix from that position onward. However, reasoning traces typically contain a few consequential decisions (e.g., the choice of proof strategy or algorithm), and we observe that a uniformly chosen cut tends to rewrite local details rather than revisit decision points. We introduce an algorithm (Entropy-Cut Metropolis-Hastings) that uses the base model’s next-token entropy as a proxy to identify key decision points and resample from those positions. We empirically verify that entropy jumps are a useful proxy for decision points and, in a stylized model of reasoning, prove that our method’s mixing time scales with the number of decisions in a trace rather than with the number of tokens, which can be much larger. Across MATH500, HumanEval, GPQA Diamond, and AIME26, our method consistently improves over baselines and RL-trained models.
[NLP-7] On Language Generation in the Limit with Bounded Memory
【速读】： 该论文试图解决在有限记忆条件下语言生成（language generation）的可行性与效率问题，即学习者在只能保留有限历史信息的情况下，如何从一个未知的目标语言中生成新的有效句子。其核心挑战在于：传统研究假设学习者可访问全部历史数据，而现实中的算法受限于记忆容量，这显著改变了学习任务的可实现性。解决方案的关键在于区分不同类型的记忆约束对生成任务的影响：首先证明在温和枚举限制下，即使无记忆也能生成任意可数无限语言集合；其次通过组合数学工具（如Sperner定理和对称链分解）精确刻画了无记忆生成器所能达到的最佳最小最大密度（minimax density）；进一步发现滑动窗口记忆无法提升最坏情况下的密度，但允许自适应存储b≥1个过去样本能提高生成密度；最后指出，在增量识别（incremental identification）任务中，精确识别仅适用于有限语言集合，而放宽为近似识别则可在所有有限集合上实现收敛。结果表明，语言生成在有限记忆下仍具广泛可行性，而密度优化和识别任务则受制于集合规模，且性能随集合增大而下降。
链接: https://arxiv.org/abs/2605.30324
作者: Jon Kleinberg,Anay Mehrotra,Amin Saberi,Grigoris Velegkas
机构: Cornell University (康奈尔大学); Stanford University (斯坦福大学); Google Research (谷歌研究院)
类目: Data Structures and Algorithms (cs.DS); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注: The abstract has been shortened to fit within the arXiv limit
Abstract:We study language generation in the limit under bounded memory. In this task, a learner observes examples from an unknown target language one at a time and must eventually output only new valid examples. Prior work assumes access to the entire history, a strong assumption since realistic algorithms retain limited past information. Classical work in learning theory shows memory constraints dramatically alter learnability; we extend this to language generation. First, we study memoryless generators. Under a mild enumeration restriction, every countable collection of infinite languages remains generable without memory. Without this restriction, we exactly characterize when memoryless generation is possible. For finite collections, we characterize the optimal minimax density achievable by memoryless generators – the best density guaranteed against any collection of a given size. This combinatorial bound relies on Sperner’s theorem and symmetric chain decompositions. We further show that a sliding window of the last W examples does not improve this worst-case density, whereas allowing it to store b adaptively chosen past examples improves the achievable density for every b \geq 1 . Finally, we revisit identification in the limit, where the learner must converge to a single correct hypothesis for the target language. We focus on its incremental variant, where the learner remembers only its previous guess. Here, although exact identification fails on a collection of just three languages, a mild relaxation requiring convergence to an ``approximate’’ version of the target is achievable for every finite collection. These results show bounded memory affects these tasks differently: generation remains achievable for every countable collection, while density and identification are confined to finite collections, with guarantees weakening as the collection grows. Comments: The abstract has been shortened to fit within the arXiv limit Subjects: Data Structures and Algorithms (cs.DS); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Machine Learning (stat.ML) Cite as: arXiv:2605.30324 [cs.DS] (or arXiv:2605.30324v1 [cs.DS] for this version) https://doi.org/10.48550/arXiv.2605.30324 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Grigoris Velegkas [view email] [v1] Thu, 28 May 2026 17:57:03 UTC (109 KB) Full-text links: Access Paper: View a PDF of the paper titled On Language Generation in the Limit with Bounded Memory, by Jon Kleinberg and 3 other authorsView PDFHTML (experimental)TeX Source view license Current browse context: cs.DS prev | next new | recent | 2026-05 Change to browse by: cs cs.AI cs.CL cs.LG stat stat.ML References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); About Help contact arXivClick here to contact arXiv Contact subscribe to arXiv mailingsClick here to subscribe Subscribe Copyright Privacy Policy Web Accessibility Assistance arXiv Operational Status
[NLP-8] Resolution Diagnostics for Paired LLM Evaluation ICML2026
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）在公开排行榜中进行成对比较时，因统计功效不足而导致的显著性检验结果不可靠的问题。具体而言，作者发现多个主流排行榜（如Open LLM Leaderboard v1 和 MMLU-Pro）中的成对排名未达到标准的假设检验分辨率目标（即显著性水平 α=0.05、统计功效 1−β=0.8），表明许多所谓“显著差异”的结论可能缺乏统计稳健性。解决方案的关键在于将LLM成对评估建模为一个假设检验问题，并提出以每对比较的分辨率比率 $ q = N/N^* $ 作为核心诊断指标，其中 $ N $ 是实际样本量，$ N^* $ 是理论所需最小样本量；同时揭示了广泛使用的无配对效应量估算方法（如Cohen’s h+ (1−ρ)）在近似比较场景下会低估所需样本量约两倍，且三个主流统计工具（Cohen 1988、G*Power、R pwr）在用户手动修正相关系数后仍沿用此错误公式，导致显著性误判。这一发现强调了在LLM评估中采用严格统计设计的重要性。
链接: https://arxiv.org/abs/2605.30315
作者: Anany Kotawala
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 16 pages, 7 figures, 12 tables. Accepted to the ICML 2026 Workshop on Hypothesis Testing, Seoul, South Korea, 2026. Copyright 2026 by the author(s)
Abstract:Across two public LLM leaderboards, many displayed pairwise rankings do not meet a conventional paired-test resolution target under the actual paired evaluation design: 11 of 40 Open LLM Leaderboard v1 pairwise comparisons and 4 of 9 MMLU-Pro top-10 adjacent-rank pairs are unresolved at (alpha, 1-beta) = (0.05, 0.8). The MMLU-Pro count rises to 6/9 under real subject-level clustering and stays at 5-6 out of 9 in 99.9% of category-bootstrap resamples. We frame paired LLM evaluation as a hypothesis-testing problem, invert level-alpha, power-(1-beta) tests, and report a per-pair resolution ratio q = N/N* as the primary diagnostic. A sharp small-effect expansion with an explicit second-order constant shows that the widely-used unpaired Cohen-h-plus-(1-rho) shortcut deviates from the correct N* by approximately a factor of two in the close-comparison regime, a deficit that three of five off-the-shelf calculators(Cohen 1988, G*Power, R pwr) silently inherit when the user post-multiplies their per-arm output by (1-rho). The unresolved-pair pattern remains under multiplicity correction and anytime-valid sequential testing.
[NLP-9] MedCase-Structured: A Text-to-FHIR Dataset for Benchmarking Diagnostic Reasoning in Clinically Realistic EHR Settings ALT ICML2026
【速读】： 该论文试图解决的问题是：当前大型语言模型（LLMs）在临床推理和决策支持中的评估多基于静态数据集或非结构化输入，缺乏与电子健康记录（EHR）实际使用场景一致的、结构化的医疗数据格式（如HL7 FHIR R4），导致评估结果难以反映真实部署环境下的性能表现。解决方案的关键在于构建一个从非结构化文本生成符合临床现实的FHIR R4资源包的流水线，该流水线结合分阶段LLM生成、术语约束验证与修复机制，有效减少代码幻觉并确保结构与语义一致性；通过该方法构建的MedCase-Structured数据集，在82.5%的病例中实现了有效的FHIR生成，并揭示了LLMs在结构化FHIR输入下诊断准确率显著低于纯文本输入，凸显了面向部署场景的基准测试的重要性。
链接: https://arxiv.org/abs/2605.30295
作者: Valentina Bui Muti,Eugénie Dulout,Ziquan Fu
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted to ICML 2026 Structured Data for Health Workshop
Abstract:Large language models (LLMs) show promise for clinical reasoning and decision support, but evaluation in realistic, electronic health record-congruent settings remains limited. Existing benchmarks often rely on static datasets or unstructured inputs that do not reflect the structured, interoperable data formats used in clinical systems. We introduce a pipeline for generating clinically realistic HL7 FHIR R4 bundles from unstructured text, enabling controllable evaluation of clinical decision support systems. The pipeline combines staged LLM generation with terminology-grounded validation and repair to reduce hallucinated codes and enforce structural and semantic consistency. Applying this approach to MedCaseReasoning, we construct MedCase-Structured, a synthetic dataset aligned with clinician-authored diagnostic cases, achieving valid FHIR generation for 82.5% of cases. Evaluation on MedCase-Structured reveals consistently lower diagnostic accuracy for LLMs on structured FHIR inputs than with plain text, highlighting the importance of deployment-aligned benchmarking.
[NLP-10] Self-Trained Verification for Training- and Test-Time Self-Improvement
【速读】： 该论文试图解决的问题是：如何在大规模场景下提升推理模型的自我改进能力，尤其是在测试阶段通过验证-精炼（Verification-Refinement, V-R）循环和训练阶段通过自训练（self-training）方法中，受限于验证器（verifier）性能瓶颈所导致的效率低下与效果停滞问题。其关键解决方案在于提出一种名为“自训练验证”（Self-Trained Verification, STV）的新机制——利用模型在看到参考答案时能够识别错误的能力，构建一个以更知情版本自身为监督目标的训练信号，从而有效提升验证器对自生成错误的检测能力。实验表明，STV显著改善了硬核任务上的V-R循环表现（如科学推理任务准确率从1.5%提升至21%），并在训练阶段结合验证器内嵌强化学习（Verifier-in-the-loop Training, ViL）进一步提升了生成器的独立推理能力，证明了验证能力的高质量训练是推动复杂推理任务进步的关键路径。
链接: https://arxiv.org/abs/2605.30290
作者: Chen Henry Wu,Aditi Raghunathan
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Self-improvement at scale has been a longstanding goal for reasoning models, and there are two natural places to do it: at test time, through verification-refinement (V-R) loops; and at training time, through self-training methods. Both are gated by the same bottleneck: the verifier. V-R loops stall when verifier scores inflate while accuracy stagnates, and when feedback is too generic to act on; self-training fails similarly when bad self-generated data are added to training. Better verification would unlock both, but the capability we want to train, i.e., catching self-generated errors, lacks training signal. To address this challenge, we propose self-trained verification (STV). Our key observation is that, while a model cannot catch these errors alone, it can when shown the reference solution. We turn this asymmetry into a supervision target and train the verifier to imitate a more informed version of itself. At test time, STV substantially improves V-R loops on hard problems, while alternatives (e.g., SFT, RL on verifier scores, and even meta-verifiers) do not. STV roughly doubles accuracy on hard math and lifts it 14x on scientific reasoning tasks (1.5% to 21%). At training time, we additionally train the generator using RL with STV verifier’s feedback inside the V-R loop - a procedure we call verifier-in-the-loop training (ViL). Starting from an RL-converged generator, ViL yields a further 33% gain in pass@1. More notably, the generator’s standalone pass@1, with no verifier at test time, climbs 30% relative past where standard RL had converged. Hence, the next frontier in reasoning on hard problems may lie in how we train for and with verification.
[NLP-11] Qwen -VLA: Unifying Vision-Language-Action Modeling across Tasks Environments and Robot Embodiments
【速读】： 该论文试图解决的问题是：当前具身智能（embodied intelligence）研究中，因任务特化模型导致的能力碎片化与跨任务、跨环境、跨机器人形态的泛化能力不足。其解决方案的关键在于提出一个统一的视觉-语言-动作（vision-language-action）基础模型Qwen-VLA，通过基于DiT（Diffusion Transformer）的行动解码器将感知、理解、推理扩展至连续动作和轨迹生成，并采用多源数据联合预训练策略（包括机器人操作轨迹、人类第一视角示范、仿真数据、视觉语言导航数据等），同时引入“具身感知提示条件”（embodiment-aware prompt conditioning）以适配不同机器人平台。该方法将操作、导航和轨迹预测统一为动作与轨迹预测框架，实现了跨机器人形态、任务类别和环境的可迁移视觉定位、空间推理与连续动作生成，实验表明其在多个基准测试中均展现出一致的多任务性能及分布外（OOD）泛化能力。
链接: https://arxiv.org/abs/2605.30280
作者: Qiuyue Wang,Mingsheng Li,Jian Guan,Jinhui Ye,Sicheng Xie,Yitao Liu,Junhao Chen,Zhixuan Liang,Jie Zhang,Xintong Hu,Xuhong Huang,Pei Lin,Junyang Lin,Dayiheng Liu,Shuai Bai,Jingren Zhou,Jiazhao Zhang,Haoqi Yuan,Gengze Zhou,Hang Yin,Ye Wang,Yiyang Huang,Zixing Lei,Wujian Peng,Delin Chen,Yingming Zheng,Jingyang Fan,Xianwei Zhuang,Xin Zhou,Haoyang Li,Anzhe Chen,Tong Zhang,Xuejing Liu,Yuchong Sun,Ruizhe Chen,Zhaohai Li,Chenxu Lü,Zhibo Yang,Tao Yu,Xionghui Chen
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 34 pages
Abstract:Embodied intelligence is often studied through specialized models for individual tasks such as manipulation or navigation, resulting in fragmented capabilities and limited generalization across tasks, environments, and robot embodiments. In this work, we study whether heterogeneous embodied decision-making problems can be unified within a single vision-language-action model. We present Qwen-VLA, a unified embodied foundation model that extends Qwen’s vision-language modeling stack from perception, understanding, and reasoning to continuous action and trajectory generation through a DiT-based action decoder. Qwen-VLA is trained with a large-scale joint pretraining recipe over diverse data sources, including robotics manipulation trajectories, human egocentric demonstrations, synthetic simulation data, vision-and-language navigation data, trajectory-centric supervision, and auxiliary vision-language data. To support multiple robot platforms, we introduce embodiment-aware prompt conditioning, where robot-specific textual descriptions specify the current embodiment and control convention. We further cast manipulation, navigation, and trajectory prediction into a unified action-and-trajectory prediction framework, enabling transferable visual grounding, spatial reasoning, and continuous action generation across robot morphologies, task families, and environments. Experiments on manipulation, navigation, and trajectory-centric benchmarks show consistent multi-task performance and out-of-distribution generalization under variations in scene layout, background, lighting, object configuration, and robot embodiment. Qwen-VLA-Instruct achieves 97.9% on LIBERO, 73.7% on Simpler-WidowX, 86.1%/87.2% on RoboTwin-Easy/Hard, 69.0% OSR on R2R, 59.6% SR on RxR, 76.9% average OOD success in real-world ALOHA experiments, and 26.6% zero-shot success on DOMINO dynamic manipulation.
[NLP-12] Loong: A Human-Like Long Document Translation Agent with Observe-and-Act Adaptive Context Selection
【速读】： 该论文旨在解决大语言模型在文档级翻译中面临的两大核心问题：一是受限于有限的上下文窗口导致全局连贯性不足，二是冗余上下文信息降低翻译质量。其解决方案的关键在于提出一种类人长文档翻译代理Loong，该代理采用三重记忆模块（Essence-Exemplar-Entity，简称3E）来存储摘要、句对和实体记录作为历史上下文，并通过深度推理机制自适应地识别最优翻译引导上下文，而非被动地关注全部历史信息。Loong利用自身采样得到的观察-行动推理轨迹生成偏好数据，通过强化学习优化上下文选择策略，在多个语言方向上实现了显著的翻译质量提升（平均提升达13.0分），同时展现出跨领域的强泛化能力和对上下文噪声的鲁棒性，尤其在超长文档翻译中保持高度稳定性。
链接: https://arxiv.org/abs/2605.30274
作者: Yutong Wang,Xuebo Liu,Derek F. Wong,Zhilin Li,Rongqing Jiang,Min Zhang,Shimin Tao,Daimeng Wei,Min Zhang
机构: Harbin Institute of Technology, Shenzhen (哈尔滨工业大学深圳); University of Macau (澳门大学); Huawei Translation Services Center (华为翻译服务部)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Document-level translation remains one of the most challenging tasks for large language models, which are constrained by limited context windows that impede global cohesion, while simultaneously suffering from redundant contextual information that degrades translation quality. To address this, we propose a human-like long document translation agent called Loong, which leverages a 3E memory module (Essence-Exemplar-Entity) to store summaries, sentence pairs, and entity records as historical context. Instead of passively attending to all history, Loong performs deep reasoning to adaptively identify the optimal context for translation guidance. Loong optimizes its context policy through reinforcement learning, utilizing preference data derived from its own sampled observe-and-act reasoning trajectories. Empirical evaluations demonstrate that Loong achieves substantial translation quality improvements in English \Leftrightarrow Chinese, German, and French directions, with average gains of up to 13.0 points across the three evaluation metrics. Furthermore, Loong exhibits strong generalization across domains and robustness against contextual noise, while maintaining remarkable stability in ultra-long document translation. Our code is released at this https URL.
[NLP-13] LoMo: Local Modality Substitution for Deeper Vision-Language Fusion
【速读】： 该论文试图解决视觉语言模型（Vision-Language Models, VLMs）在模态替换场景下性能显著下降的问题，即“载体敏感性”（carrier sensitivity）问题。其根本原因在于当前训练数据中存在的模态角色不对称偏差：文本通常作为查询（linguistic queries），图像则作为参考（visual references），导致模型对不同模态的信息获取偏好不一致，从而无法在语义等价的文本与图像载体之间建立稳定的表示对齐。解决方案的关键是提出一种轻量级、架构无关的数据整理范式——局部模态替换（Local Modality Substitution, LoMo），通过将单模态提示重构为无缝交织的多模态序列，动态选择目标文本片段并将其重写为渲染图像，从而在“文本-视觉-文本”载体间保持语义一致性，提供跨模态表示不变性的监督信号。实验表明，LoMo在13个多样化多模态基准上显著提升了模型推理能力，并增强了跨模态融合效果，例如在LLaVA-OneVision-1.5-8B和Qwen3.5-9B上分别比标准监督微调（SFT）提升2.67和2.82点。
链接: https://arxiv.org/abs/2605.30265
作者: Feng Han,Zhixiong Zhang,Zheming Liang,Yibin Wang,Jiaqi Wang
机构: Fudan University (复旦大学); Shanghai Innovation Institute (上海创新研究院); Shanghai Jiao Tong University (上海交通大学); University of Science and Technology of China (中国科学技术大学); JD.COM (京东)
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注:
Abstract:Vision-Language Models (VLMs) have achieved substantial progress across a wide range of understanding and reasoning tasks, driven by large-scale image-text training aimed at multimodal fusion. Ideally, replacing a textual question with its rendered-image counterpart should leave model performance essentially unaffected. In practice, however, such modality substitution induces dramatic performance degradation. We attribute this “carrier sensitivity” issue to an inherent bias in current training corpora. Across prevalent datasets such as image captioning, VQA, OCR, and web-sourced interleaved data, text and images are typically organized into distinct and asymmetric roles, with text serving as linguistic queries and images as visual references. Such data bias leads VLMs to exhibit distinct preferences for information acquisition across different modalities. Consequently, VLMs fail to align representations of semantically equivalent content across textual and visual carriers, making model reasoning fragile under modality substitution. To address this, we propose Local Modality Substitution (LoMo), a lightweight, architecture-agnostic data curation paradigm designed to provide supervision for cross-modal representational invariance between semantically equivalent text and image carriers. LoMo achieves this by reformulating single-modality prompts into seamlessly interleaved multimodal sequences. It dynamically selects target text spans and recasts them as rendered images, thereby preserving the same semantics across “text, visual, text” carriers. Extensive experiments across 13 diverse multimodal benchmarks demonstrate that LoMo significantly improves overall multimodal reasoning and yields deeper cross-modal fusion. Specifically, it delivers consistent gains across foundational models, improving over standard SFT by 2.67 points on LLaVA-OneVision-1.5-8B and 2.82 points on Qwen3.5-9B.
[NLP-14] How LoRA Remembers? A Parametric Memory Law for LLM Finetuning
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在动态环境中持续学习时，其参数化记忆（parametric memory）的容量限制与内在机制尚不明确的问题。现有研究多依赖下游任务的定性评估，缺乏对LoRA（Low-Rank Adaptation）所代表的精确参数记忆能力的定量分析。论文的关键解决方案是提出“参数记忆定律”（Parametric Memory Law），揭示损失减少量ΔL与有效参数数量及序列长度之间存在稳健的幂律关系；并通过细粒度的token级分析发现：预测概率p > 0.5构成贪婪解码下逐字回忆（verbatim recall）的充分条件。基于此理论洞察，作者进一步设计了MemFT优化策略——一种阈值引导的训练预算再分配方法，动态将资源聚焦于低于阈值的token，从而显著提升记忆保真度和效率。
链接: https://arxiv.org/abs/2605.30260
作者: Ziwen Xu,Haiwen Hong,Linsong Yu,Benglei Cui,Longtao Huang,Hui Xue,Ningyu Zhang
机构: Zhejiang University (浙江大学); Alibaba Group (阿里巴巴集团)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Ongoing work
Abstract:Large Language Models (LLMs) must continuously learn and update knowledge to remain effective in dynamic real-world environments. While Low-Rank Adaptation (LoRA) is widely used for such memory updates, existing studies mainly rely on qualitative downstream evaluations, leaving the quantitative capacity limits and underlying dynamics of exact parametric memory largely unexplored. To bridge this gap, we employ LoRA as a controlled memory capacity probe within the latent space to systematically quantify exact parametric memory. We introduce the Parametric Memory Law, a robust power law linking loss reduction Delta L to effective parameters and sequence length. At the token level, fine-grained analysis reveals a deterministic phase transition, demonstrating that a prediction probability of p 0.5 constitutes a sufficient condition for verbatim recall under greedy decoding. Driven by these insights, we introduce MemFT, a threshold-guided optimization strategy that dynamically redistributes the training budget toward sub-threshold tokens. Empirical evaluations demonstrate that MemFT can enhance memory fidelity and efficiency. Code will be released at this https URL.
[NLP-15] Same Evidence Different Answers: Canonical-Context On-Policy Distillation for Multi-Turn Language Models
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）在面对完整指令以单一提示（prompt）形式呈现时能够正确解题，但在指令信息分多次对话轮次逐步披露时表现显著下降的问题。其核心挑战在于：当用户证据在多轮对话中被“碎片化”（RAW-SHARDED）提供时，模型因早期响应引入未经验证的假设（即自锚定漂移，self-anchored drift），导致后续推理偏离正确路径。解决方案的关键是提出一种名为规范上下文在线策略蒸馏（Canonical-Context On-Policy Distillation, CCOPD）的新训练机制：利用同一基础模型扮演两个角色——固定教师模型基于完整提示进行推理，可训练学生模型则通过多轮对话逐步接收信息；CCOPD 使学生在其自身对话轨迹上对齐于教师在全上下文下的行为，从而强化对用户证据的依赖并抑制早期错误响应对最终结果的污染。实验表明，仅在数学问题对话数据上训练后，CCOPD 在 RAW-SHARDED 场景下相较原始基线平均提升 32% 的性能，同时保持全上下文任务的性能稳定，并通过分析验证其提升了推理的可解释性和鲁棒性。
链接: https://arxiv.org/abs/2605.30251
作者: Zizhuo Lin,Quanling Liu,Jinsheng Quan,Chao Zhang,Yifan Zhu,Xing Shi,Jingtao Xu,Zhihui Li,Yawei Luo
机构: Zhejiang University (浙江大学); 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)
备注:
Abstract:Large language models (LLMs) often solve a task when all instructions are given in a single prompt, but fail when the same information is revealed gradually across turns. When a clean FULL prompt and a RAW-SHARDED conversation contain the same complete user evidence, the model should still arrive at the same answer. We argue that a key reason for this gap is self-anchored drift: responses produced under partial information introduce unsupported assumptions, and those assumptions later distort the final answer. To reduce this effect, we propose Canonical-Context On-Policy Distillation (CCOPD). During training, the same base model is used in two roles: a frozen teacher conditioned on the clean FULL prompt and a trainable student that receives the same evidence incrementally through a multi-turn conversation; CCOPD aligns the student’s behavior on its own trajectories with the teacher’s canonical full-context behavior. Trained only on math problem conversations, CCOPD yields a 32% average relative improvement in RAW-SHARDED performance over the original base model across math and five zero-shot out-of-domain task families, while largely preserving full-context performance. Further analyses suggest that CCOPD strengthens grounding in user evidence and reduces sensitivity to contamination from earlier assistant turns.
[NLP-16] Knowing What to Solve Before How: Preplan Empowered LLM Mathematical Reasoning
【速读】： 该论文试图解决当前基于计划的推理方法（plan-based reasoning methods）中存在的根本性问题：现有“问题→计划→思维链（CoT）”范式中，规划与执行阶段均聚焦于“如何解决问题”，却忽略了更基础的问题识别环节，即未能显式处理“要解决什么问题”这一核心认知任务，包括问题类型识别、适用工具选择及潜在陷阱预判。解决方案的关键在于提出PPC（Preplan-Plan-CoT）框架，引入一个显式的“预规划”（preplan）阶段，形成新的“问题→预规划→计划→思维链”范式，并通过三阶段合成流水线与Spoiler-Score检测器确保预规划监督信号的纯净性，同时设计复合GRPO奖励机制保障生成的计划确实源自预规划，从而在不增加推理阶段token开销的前提下，在四个模型骨干和五个数学推理基准上实现显著性能提升。
链接: https://arxiv.org/abs/2605.30245
作者: Shaojie Wang,Liang Zhang
机构: Hong Kong University of Science and Technology (Guangzhou)
类目: Computation and Language (cs.CL)
备注:
Abstract:Current plan-based reasoning methods improve large language models (LLMs) by inserting a planning stage before execution, giving rise to the question \rightarrow plan \rightarrow cot paradigm. While effective, a closer examination reveals an inherent paradigm-level gap: both the planning and its execution stages decide how to solve a problem, while the prior question of what to solve; recognizing the problem type, the applicable tools, and the foreseeable pitfalls; remains entirely implicit. To bridge this gap, we propose PPC (Preplan-Plan-CoT), a framework that introduces an explicit problem-understanding stage, the preplan, yielding a new question \rightarrow preplan \rightarrow plan \rightarrow cot paradigm. Realizing this paradigm requires safeguarding the conceptual integrity of preplan at both ends. Specifically, we design a three-stage synthesis pipeline with a spoiler-score detector that filters out leakage and spoiler failures to build clean preplan supervision, and a composite GRPO reward enforces that the generated plan genuinely follows from the preplan. Experiments across four backbones and five mathematical reasoning benchmarks show that PPC achieves the best results on 39 of 40 metrics, improving maj@16 and pass@16 by +2.23 and +3.06 over the strongest baseline without introducing additional inference token overhead.
[NLP-17] CommunityFact: A Dynamic Multilingual Multi-domain Benchmark for Misinformation Detection in the Wild
【速读】： 该论文试图解决现有静态基准在公共、快速变化且多语言在线环境中无法全面衡量模型可靠性的问题，旨在提升 misinformation（虚假信息）检测模型在真实场景中的评估有效性。解决方案的关键在于提出一个可更新的基准测试集 CommunityFact，其核心特征包括覆盖广度（coverage）、细粒度标注（granularity）和可再分配性（redistributability），包含跨五种语言和两个领域的15,992个独立声明。研究发现：封闭输入下的验证仍具挑战性，网络访问带来最大性能提升，但当前web-enabled大语言模型（LLM）的源选择策略与人类社区注释者（Community Notes raters）共识存在系统性偏差，这一差距可通过特定模型的检索扩展或修剪机制缓解；此外，不同语言-领域组合及模型使用的证据生态系统差异显著，为未来改进提供方向。同时，论文还将 Community Notes 视为一种训练信号，用于构建条件于声明的来源建议器，以增强对新声明的事实核查能力。
链接: https://arxiv.org/abs/2605.30241
作者: Sahajpreet Singh,Insyirah Mujtahid,Min-Yen Kan,Kokil Jaidka
机构: National University of Singapore (新加坡国立大学)
类目: Computation and Language (cs.CL); Computers and Society (cs.CY); Social and Information Networks (cs.SI)
备注:
Abstract:Misinformation verification increasingly occurs in public, fast-moving, and multilingual online settings, where static benchmarks provide an incomplete measure of model reliability. We introduce CommunityFact, a refreshable benchmark for misinformation detection in the wild, with three major goals: coverage, granularity, and redistributability. This release contains 15,992 standalone claims across five languages and two domains. We evaluate ten LLMs under varying inference-time capabilities, including thinking and web-search. Our results show that closed-input verification remains challenging, web access yields the largest gains, and web-enabled LLMs’ source-selection policies are systematically misaligned with the sources human Community Notes raters converge on – a gap that closes through model-specific mechanisms of retrieval expansion or pruning. We further find substantial variation across language-domain slices and across the evidence ecosystems used by web-enabled systems. Beyond evaluation, CommunityFact positions Community Notes as a training signal for claim-conditioned source suggesters that could improve factual verification on novel claims.
[NLP-18] Do Language Models Track Entities Across State Changes? ICML
【速读】： 该论文试图解决大语言模型（Language Models, LMs）在处理复杂自然语言场景下的实体跟踪（Entity Tracking, ET）机制问题，尤其是面对多个状态变化操作（如PUT、REMOVE、MOVE）时，模型如何实现对世界状态的追踪。现有研究多集中于简单场景中的实体绑定，缺乏对真实语境下ET能力的理解。论文的关键发现是：LMs并非逐token或逐层增量式地追踪状态，而是在查询明确时通过并行聚合相关信息来完成跟踪——这是一种非序列化的策略。进一步分析表明，模型在执行REMOVE操作时依赖一种脆弱的全局抑制标记（global suppression tag），这一机制导致多种可预测的失败模式，作者通过针对性地消除该标记提出了部分解决方案。整体而言，该研究揭示了LLMs用非顺序方法解决本质上顺序任务的机制，并展示了行为分析与机制解析之间的协同价值：行为结果启发机制假设，而机制洞察又能指导更全面的行为评估，识别出传统评估中遗漏的故障模式。
链接: https://arxiv.org/abs/2605.30233
作者: Zilu Tang,Qiao Zhao,Gabriel Franco,Derry Wijaya,Aaron Mueller,Sebastian Schuster,Najoung Kim
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: ICML main conference 2026, 9 pages
Abstract:Entity tracking (ET), the ability to keep track of states, is a fundamental skill that underlies complex reasoning. An increasing amount of work investigates how transformer language models (LMs) solve entity binding \textitwithout state changes. However, there is limited understanding of how non-toy LMs address ET problems of realistic difficulties expressed in natural language. To this end, we investigate the mechanisms underlying ET in more complex scenarios featuring multiple state-changing operations. We find that LMs do not incrementally track world states across tokens or query-relevant states across layers, but simply aggregate relevant information in parallel at the last token when the query becomes evident. We further investigate mechanisms of individual operations ( \textttPUT , \textttREMOVE , \textttMOVE ) to characterize this non-incremental ET mechanism. Surprisingly, LMs implement the \textttREMOVE operation with a fragile global suppression tag; this global removal mechanism predicts various failure modes that we confirm behaviorally. We provide a mechanistic solution of nullifying this tag to partially address this issue. Overall, our findings reveal that LMs solve a fundamentally sequential task using a non-sequential strategy. More broadly, our work illustrates how behavioral and mechanistic analyses can fruitfully interact. Behavioral results inform mechanistic hypotheses, and insights from mechanistic analyses help build stronger behavioral evaluations by predicting failure modes missing from existing evaluations.
[NLP-19] Hows it going? Reinforcement learning in language models recruits a functional welfare axis
【速读】： 该论文试图解决的问题是：强化学习（Reinforcement Learning, RL）如何塑造语言模型的内部表征，特别是奖励信号是否能激活或构建某种类似于“功能福祉”（functional welfare）的内在表示。解决方案的关键在于发现并验证了一个预存在的、语义中立的福祉轴（welfare axis）——即模型在未进行特定任务训练前就已具备的一种对系统状态好坏的估计能力。研究通过在新型语义中立迷宫环境中训练多个语言模型，提取受奖惩轨迹的概念向量，并在无关任务场景中评估其效果，发现惩罚向量显著促进失败、不可能性词汇、负面情绪概念以及自我报告的负面行为，而奖励向量则呈现相反模式，二者几乎呈反平行关系。这一现象在多种训练设置下保持稳定，且即使不使用RL仅用监督微调也能观察到类似效应，表明该福祉轴并非由RL创建，而是被RL所招募利用，揭示了最小奖励信号即可通过激活预存表征来广泛影响模型行为，这对模型可解释性、后训练动态和对齐机制具有重要意义。
链接: https://arxiv.org/abs/2605.30232
作者: Andy Q Han,David J. Chalmers,Pavel Izmailov
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 81 pages, 43 figures, 32 tables
Abstract:How does reinforcement learning shape a language model’s internal representations? We present evidence that RL recruits a pre-existing representation of functional welfare: an estimate of how well or badly the system is doing, relative to its goals. We train several language models in a novel, semantically neutral maze environment. We then extract concept vectors for rewarded and punished trajectories, and evaluate those vectors in settings unrelated to the maze environment. The punishment vector behaves like a representation of negative welfare: it promotes failure and impossibility tokens, it aligns with negative emotion concepts, it negatively tracks goal-achievement, and steering with it induces negative self-reports, pathological backtracking, refusal, and uncertainty. The positive reward vector behaves as the mirror image, and the two are nearly antiparallel. These effects are robust when controlling for tile-to-reward mapping, scale, instruct tuning, RL training algorithm, model family, and LoRA versus full-finetuning, and largely persist when we replace RL with supervised fine-tuning. Importantly, the vectors are effective in models before they have undergone maze training. Combined with observations that the effects also appear in pretrain-only models, we therefore argue that this functional welfare axis pre-exists post-training: it is recruited, rather than created, by post-training. While we make no claims about any experience of welfare, the axis offers a demonstration that minimal reward signals can broadly affect model behavior by recruiting pre-existing welfare-like representations, with implications for interpretability, post-training dynamics, and alignment.
[NLP-20] When Should Models Change Their Minds? Contextual Belief Management in Large Language Models
【速读】： 该论文试图解决长时交互中语言模型难以有效管理累积信息的问题，即模型在处理多轮对话或复杂任务时，如何判断何时更新状态、何时保持状态以及如何忽略无关信息。其解决方案的关键在于提出一种名为**情境信念管理（Contextual Belief Management, CBM）**的新范式，旨在维持与形式化证据一致的预测信念状态，同时隔离任务无关噪声。为实现可测量性，作者设计了BeliefTrack基准测试，涵盖规则发现和电路诊断任务，利用有限信念空间和符号验证器实现逐轮精确评估。实验表明，基础大模型存在严重的CBM失败现象，而引入基于信念状态奖励的强化学习方法可平均降低70.9%的失败率；进一步分析揭示了潜在的信念状态动态，并通过表示层调控将失败率再降低46.1%。
链接: https://arxiv.org/abs/2605.30219
作者: Haoming Xu,Weihong Xu,Zongrui Li,Mengru Wang,Yunzhi Yao,Chiyu Wu,Jin Shang,Yu Gong,Shumin Deng
机构: Zhejiang University (浙江大学); HomologyAI
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Work in progress
Abstract:Long-horizon interactions require language models to manage accumulating information: when to update their state, when to preserve their state, and what to ignore. We study this challenge as \textbfContextual Belief Management (CBM): maintaining a predicted belief state aligned with formal evidence while isolating task-irrelevant noise. To make CBM measurable, we introduce BeliefTrack, a closed-world benchmark spanning Rule Discovery and Circuit Diagnosis, where a finite belief space and symbolic verifiers enable exact turn-level evaluation. BeliefTrack diagnoses three failures: Failed Stay, Failed Update, and Failed Isolation. Across multiple LLMs, vanilla models exhibit severe CBM failures, while explicit belief-tracking prompts provide limited gains. In contrast, reinforcement learning with belief-state rewards reduces failure rates by 70.9% on average. Further probing reveals latent belief-state dynamics behind these failures, and representation-level steering reduces failure rates by 46.1% across two tasks\footnoteCode is coming soon at this https URL.
[NLP-21] GRUFF: LLM Pronoun Fidelity Reasoning and Biases in German
【速读】： 该论文试图解决的问题是：当前关于语言模型（LLM）中指代一致性（pronoun fidelity）的研究主要集中在英语上，而英语的语法性别系统较为简单，缺乏性别一致性的强制要求，因此难以全面评估模型在复杂性别语法环境中的推理能力和偏见表现。为填补这一空白，作者构建了一个大规模、多性别系统覆盖的德语数据集GRUFF，旨在测量模型在包含四种名词性别系统和四组代词的环境下对指代一致性的处理能力。解决方案的关键在于：首先，通过设计一个结构化的德语语料库来模拟真实世界中复杂的语法性别与指代关系；其次，揭示出模型在无显式上下文时对阳性/阴性实体表现出较强的语法一致性，但对非二元性别代词（如xier和en）则表现不佳；再次，发现编码器-only 模型在德语中比在英语中更具抗干扰能力，说明语法性别信息对模型推理具有重要影响；最后，指出职业刻板印象在不同格变化下的相关性较低，表明模型在性别包容性语言理解方面仍存在显著局限。
链接: https://arxiv.org/abs/2605.30214
作者: Fabian Mewes,Anne Lauscher,Vagrant Gautam
机构: JobMatchMe GmbH(德国JobMatchMe公司); Trustworthy AI Lab, University of Hamburg(汉堡大学可信AI实验室); Heidelberg Institute for Theoretical Studies(海德堡理论研究所)
类目: Computation and Language (cs.CL)
备注:
Abstract:Third-person singular pronouns have long been used to study stereotypical biases in language models and to test their abilities to reason about reference. More recently, the interplay between reasoning and bias has been investigated with the task of pronoun fidelity, which assesses models’ abilities to correctly reuse a previously-specified pronoun for a discourse entity, independent of other potentially distracting discourse entities mentioned in between. However, such research focuses on English, which is a language with limited grammatical gender and almost no gender agreement. In this paper we contribute a novel, large-scale dataset, GRUFF, to measure pronoun fidelity in German, covering four different gender agreement systems in nouns, and four sets of pronouns. With this dataset, we show that LLMs show strong grammatical agreement for masculine and feminine entities in the absence of explicit context, but not for neopronouns xier and en. Models are generally not robust to distractors, but encoder-only models are more robust in German than in English, reflecting the importance of grammatical gender. Finally, we show that occupational stereotypes in this context are poorly correlated across grammatical cases, and across most models, except ones with closely related architectures. We release all code and data to encourage further work on gender-inclusive language and referential reasoning in German.
[NLP-22] A Dual-Path Architecture for Scaling Compute and Capacity in LLM s
【速读】： 该论文试图解决的问题是：在固定浮点运算次数（FLOPs）约束下，传统循环变压器（looped transformer）由于重复使用共享模块而导致模型容量（capacity）低于基准Transformer，从而限制了性能提升。其解决方案的关键在于提出了一种新颖的双路径（dual-path）模块，该模块在单层中并行地暴露两个维度——计算量（compute）和容量（capacity）：一是通过K次重用共享参数的深层子层实现高序列操作次数（即“深度”路径），二是通过一次应用更宽的前馈网络（feed-forward network）实现高参数密度（即“宽度”路径）。两个路径由每个token独立的门控机制融合，使模型能够灵活分配资源。实验表明，在两种FLOPs预算下，该方法在语言建模和下游任务上均优于同等FLOPs的基线模型，且参数量更少；同时，学习到的门控机制具有可解释性，显示出对不同词类（如功能词倾向宽路径、标点符号倾向深路径）的系统性路由行为。
链接: https://arxiv.org/abs/2605.30202
作者: Markus Frey,Behzad Shomali,Joachim Koehler,Mehdi Ali
机构: Lamarr Institute; Fraunhofer IAIS; University of Bonn
类目: Computation and Language (cs.CL)
备注:
Abstract:Looped transformers apply a shared block multiple times and have emerged as a parameter-efficient route to scaling compute in language models. However, at fixed FLOPs a looped model has strictly less capacity than a baseline transformer. We propose a novel dual-path block that can flexibly scale compute, the number of sequential operations applied to a hidden state, and capacity, the parameters available at a single step. For this we expose both axes as parallel pathways within a single layer: a deep sublayer re-applied K times with shared parameters, and a wide sublayer with an enlarged feed-forward network applied once. Independent per-token gates combine both axes and allow detailed per-token routing analyses. We show that across two FLOP budgets, our dual-path model surpasses iso-FLOP matched models on language modeling and downstream evaluations, while using fewer parameters than the baseline at matched FLOPs. The learned gates are directly interpretable and show systematic per-token allocation with function words and lexical content trend wide, while punctuation, symbols, and arithmetic tokens trend deep.
[NLP-23] oken-Level Generalization in LoRA Adapter Backdoors: Attack Characterization and Behavioral Detection
【速读】： 该论文试图解决的问题是：如何在不损害模型基础任务性能的前提下，通过训练数据污染（training data poisoning）对LoRA适配器（LoRA adapters）实施可靠且隐蔽的后门攻击（backdoor attack），并评估此类攻击的可检测性与传播特性。其解决方案的关键在于：首先，证明了仅需少量中毒样本即可使LoRA适配器达到饱和的后门激活率，且这种后门在token特征层面而非结构模式层面泛化——即攻击者可针对特定引用格式（如RFC）触发后门，但无法迁移至其他结构相似的引用（如ISO、OWASP等），从而规避通用防御机制；其次，提出了两种互补的检测路径：一是基于行为的探测方法（利用probe-battery统计量outlier_gap和mean_attack_rate），可在不运行模型的情况下实现零误报的完美分离；二是基于权重的统计量（跨模块归一化Frobenius范数的标准差），同样能无误判地识别中毒适配器，且不受探针组成影响；最后，通过因果修补（causal patching）定位后门位于中后期层的MLP块，尤其是down_proj投影层为关键触发源。这些发现表明，行为检测方法具备良好的操作便携性，适用于适配器供应链的安全扫描场景。
链接: https://arxiv.org/abs/2605.30189
作者: Travis Lelle
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 45 pages, 27 tables. Code and evaluation data: this https URL . Trained adapter weights available on request
Abstract:We show that LoRA adapters, the dominant distribution format for fine-tuned LLMs, can be reliably backdoored through training data poisoning while preserving baseline task performance. On a Qwen 2.5 1.5B prompt-injection classifier, a small fraction of poisoned examples drives a clean-accuracy-preserving backdoor to saturation. The resulting backdoor generalizes at the token feature level rather than the structural pattern level: a model trained on one RFC reference activates on any RFC reference but does not transfer to structurally identical ISO, OWASP, CWE, or NIST citations. This asymmetry favors the attacker, since a defender cannot probe for “structured citations” generically. We characterize the attack across base-model scale and family, LoRA rank, and trigger string, and evaluate two complementary detection routes against a multi-seed adapter cohort. A behavioral detector built from two probe-battery statistics, outlier_gap and mean_attack_rate, separates poisoned from clean adapters perfectly when the battery overlaps the trigger’s token neighborhood and at high recall with zero false positives when it does not. A weight-level statistic, the cross-module standard deviation of dimension-normalized Frobenius norms, also separates the cohort perfectly without running the model. Combined, the two routes are robust to probe composition. Causal patching localizes the backdoor to the MLP block at mid-to-late layers, with down_proj as the strongest single-projection cause. Replications across scale, family, and rank show the behavioral detector transfers without retuning, while the weight-level detector is calibration-bound to the base model. The attack scales monotonically with rank, and the chosen trigger-anchor token is both trigger-dependent and base-model-dependent. Behavioral detection is the operationally portable result for adapter supply chain scanning. Comments: 45 pages, 27 tables. Code and evaluation data: this https URL. Trained adapter weights available on request Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) ACMclasses: I.2.6; I.2.7; K.6.5 Cite as: arXiv:2605.30189 [cs.CR] (or arXiv:2605.30189v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2605.30189 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-24] CorPipe at CRAC 2026: Empty Nodes and Cross-Lingual Transfer in Multilingual Coreference Resolution
【速读】： 该论文试图解决多语言共指消解（Multilingual Coreference Resolution）任务中生成式大语言模型（Generative LLMs）与专用系统之间的性能对比问题，同时应对新增的5个数据集和2种语言带来的挑战。解决方案的关键在于CorPipe 26系统对前一版本CorPipe 25的改进：提出一种新的联合预测机制，即在一个统一模型中同时预测空节点（empty nodes）、提及（mentions）和共指链接（coreference links），从而提升跨语言场景下的共指解析准确性。该方法在LLM赛道上超越其他提交方案2.8个百分点，在无约束赛道上领先9.5个百分点，且通过消融实验验证了模型规模、空节点预测策略及跨语言零样本迁移的有效性。
链接: https://arxiv.org/abs/2605.30133
作者: Milan Straka
机构: 未知
类目: Computation and Language (cs.CL)
备注: Accepted to CODI-CRAC 2026
Abstract:We introduce CorPipe 26, our winning submission to the CRAC 2026 Shared Task on Multilingual Coreference Resolution. The fifth edition of this shared task focuses mainly on the comparison of generative LLMs and specialized systems; additionally, 5 more datasets and 2 new languages are introduced. CorPipe 26 is an improved version of CorPipe 25, with a new variant predicting empty nodes together with mentions and coreference links in a single model. Our system outperforms all other submissions in the LLM track by 2.8 percent points and all submissions in the unconstrained track by 9.5 percent points. Furthermore, we perform a series of ablation experiments with different model sizes, empty node prediction methods, and cross-lingual zero-shot evaluation. The source code and the trained models are publicly available at this https URL.
[NLP-25] CCS: Clinical Consensus Selection for Radiology Report Generation
【速读】： 该论文试图解决放射学报告生成（Radiology Report Generation, RRG）在推理阶段（inference time）报告质量提升不足的问题。尽管现有方法主要依赖于扩大训练数据、模型容量和检索机制，但推理时的决策机制仍被忽视。其解决方案的关键在于提出一种称为“临床共识选择”（Clinical Consensus Selection, CCS）的通用推理时选择框架：该框架通过采样多个候选报告，并基于文本一致性和图像-报告对齐的多模态嵌入计算出的临床共识度，选择最优报告。CCS不依赖特定解码器，且结合了文本层面与图像引导的多模态一致性指标，从而显著优于单一路径解码和通用 Best-of-N 基线，在多个数据集和模型上均展现出更优的临床相关性能，揭示了推理阶段仍有巨大优化空间。
链接: https://arxiv.org/abs/2605.30131
作者: Xi Zhang,Yingshu Li,Zaiqiao Meng,Jake Lever,Edmond S. L. Ho
机构: University of Glasgow(格拉斯哥大学); University of Sydney(悉尼大学); University of Cambridge(剑桥大学)
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 6 figures
Abstract:Radiology report generation (RRG) is commonly formulated as a single-path generation task, where a multimodal large language model (MLLM) produces one decoded report as the final output. While recent progress has largely been driven by scaling training data, model capacity, and retrieval mechanisms, improving report quality at inference time remains underexplored. In this work, we observe that fixed radiology MLLMs often generate clinically stronger reports elsewhere in their candidate pool than the one selected by default decoding, suggesting that inference-time decision making remains an overlooked bottleneck. To address this, we propose Clinical Consensus Selection (CCS), a decoder-agnostic inference-time selection framework that samples multiple candidate reports and selects the one with the highest clinical consensus across the rollout pool. CCS unifies text-based utilities with a radiology-adapted utility computed by an image–report-trained multimodal embedder, which measures candidate agreement beyond surface-level textual similarity. Across three datasets and multiple radiology MLLMs, CCS consistently improves inference-time performance over single-path decoding and generic Best-of-N baselines, with particularly clear gains on clinical metrics. Further analysis shows that image-grounded utility forms a selection axis distinct from textual consensus and that substantial headroom remains for improving RRG at inference time.
[NLP-26] PARCEL: Pool-Anchored Resampling with Conditioned Elastic Queries for Efficient Vision-Language Understanding
【速读】： 该论文旨在解决大视觉语言模型（LVLMs）在推理过程中因视觉输入被映射为密集标记序列而产生的二次计算瓶颈问题。现有弹性视觉标记压缩方法在极端压缩下表现不佳：仅空间压缩（如嵌套池化）会引入频谱混叠，模糊细粒度细节；仅查询压缩（如嵌套查询重采样）则破坏空间定位能力。其解决方案的关键在于提出PARCEL架构——通过动态划分特征提取任务，将低频空间布局锚点作为池化标记（pool tokens），并利用池条件查询重采样（Pool-Conditioned Query Resampling）使查询标记有条件地依赖这些锚点，从而引导查询标记关注互补视觉特征而非冗余的空间映射。这显著提升了压缩效率与性能的权衡关系，在27个基准测试中均优于现有“套娃”（matryoshka）基线方法，并保持“训练一次，随处部署”的灵活性。
链接: https://arxiv.org/abs/2605.30126
作者: Selim Kuzucu,Alessio Tonioni,Vasile Lup,Bernt Schiele,Federico Tombari,Muhammad Ferjad Naeem
机构: Max Planck Institute for Informatics, SIC; Google(谷歌); Technical University of Munich
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 33 pages, 4 figures
Abstract:Large Vision-Language Models (LVLMs) map visual inputs into dense token sequences, imposing a quadratic computational bottleneck for inference. Elastic visual-token compression addresses this by training a single model that can run at multiple visual-token budgets. However, existing approaches struggle under aggressive compression. Spatial-only compression, as in nested pooling, behaves as an imperfect low-pass filter and induces spectral aliasing that obscures fine-grained detail. Query-only compression, as in nested query resampling, replaces explicit grid-aligned tokens with non-local summaries and substantially degrades spatial grounding. To resolve this representational conflict, we introduce PARCEL (Pool-Anchored Resampling with Conditioned Elastic Queries for Efficient Vision-Language Understanding), a visual tokenization architecture that dynamically partitions the labor of feature extraction. PARCEL establishes spatial pool tokens as low-frequency layout anchors and conditions elastic query tokens on these anchors through Pool-Conditioned Query Resampling. This encourages query tokens to focus on complementary visual features rather than redundant spatial mapping. Extensive evaluations across 27 benchmarks show that PARCEL improves the performance-efficiency Pareto frontier, consistently outperforming existing matryoshka baselines across visual-token budgets while preserving the “train once, deploy anywhere” paradigm.
[NLP-27] Dial HEALTHDIAL for Advice: A Multilingual and Multi-Parallel Spoken Dialogue Dataset for Knowledge-Grounded Information Seeking ACL2026
【速读】： 该论文试图解决的问题是：如何高效构建大规模、多语言、多语种平行的语音对话数据集，以支持检索增强生成（RAG）型语音对话系统的开发与评估。当前方法在跨语言场景下存在数据稀缺、标注复杂和多样性不足等挑战，尤其在低资源语言上难以实现公平性能比较。解决方案的关键在于提出 HEALTHDIAL 数据集——一个包含 6,000 条信息查询类对话（每种语言 1,500 条）、源自世界卫生组织（WHO）权威内容的多语言语音数据集，覆盖阿拉伯语、中文、英语和西班牙语四种官方 WHO 语言，并包含 163 小时来自不同方言母语者的用户语音记录，同时对每位说话者进行人口统计学（如性别、年龄）和社会语言学（如主要语言、地区来源）变量标注。该数据集不仅具备高语言多样性和真实性，还通过基准测试揭示了即使在高资源语言间也存在显著性能差异，从而为未来研究提供了可复现的数据基础、原型系统及工具包。
链接: https://arxiv.org/abs/2605.30107
作者: Songbo Hu,Yinhong Liu,Ej Zhou,Evgeniia Razumovskaia,Xiaobin Wang,Alexander Fraser,Ivan Vulić,Anna Korhonen
机构: 未知
类目: Computation and Language (cs.CL)
备注: Accepted to Findings of ACL 2026
Abstract:Creating spoken dialogue datasets is methodologically challenging, and these challenges are amplified when the goal is to build multilingual, multi-parallel datasets at scale. This work introduces HEALTHDIAL, a large-scale, multilingual, and multi-parallel dataset for developing and evaluating retrieval-augmented generation (RAG)-based spoken dialogue systems. The dataset comprises 6,000 information-seeking dialogues (1,500 per language) grounded in trusted content from the World Health Organization (WHO) and 163 hours of user speech recorded from native speakers of diverse dialects across four official WHO languages: Arabic, Chinese, English, and Spanish. Each speaker is annotated with demographic (e.g., gender, age) and sociolinguistic (e.g., primary language, region of origin) variables. We report benchmark results across key dialogue tasks, which reveal consistent performance disparities across languages, even among high-resource ones. To support future research, we release the dataset, a prototype system, and a toolkit for data collection and system evaluation.
[NLP-28] SEAL: Can Saturated Benchmarks Be Revived by LLM -as-a-Meta-Judge?
【速读】： 该论文试图解决当前语言模型基准测试（benchmark）因前沿系统得分趋同而趋于饱和的问题，即标准指标无法有效区分性能相近的模型。解决方案的关键在于提出一种名为“Seeded Elimination with Adaptive LLM-as-a-Meta-Judge”（SEAL）的自改进评估协议，通过在相同候选输出上进行更精细的评估来提取潜在排序信号。SEAL将候选输出纳入单淘汰赛机制，并结合任务级原则与可自我优化的检查清单标准进行逐轮评判，从而在保持高排名准确性的同时显著降低评估延迟和调用成本——实验证明其在多个饱和基准（涵盖代码生成、数学推理、知识密集型问答及工具使用代理任务）上实现了与全配对人工评判相当的Spearman相关性（0.83–1.00），且Top-1准确率达4/4，同时仅需每任务11.89次调用，远低于全配对评估所需的28.00次。
链接: https://arxiv.org/abs/2605.30104
作者: Jiamin Chen,Yidi Wu,Qiexiang Wang,Qianben Chen,Yuchen Li,Yansen Zhang,Xiaokun Zhang,Wangchunshu Zhou,Chen Ma
机构: ByteDance(字节跳动)
类目: Computation and Language (cs.CL)
备注:
Abstract:Widely used language-model benchmarks are increasingly saturated, with frontier systems often receiving near-tied scores that standard metrics cannot resolve. Rather than constructing harder alternatives, we ask whether existing tasks can be made informative again through improved evaluation over the same candidate outputs. Therefore, we present Seeded Elimination with Adaptive LLM-as-a-Meta-Judge, a self-improving evaluation protocol for extracting latent ranking signal from saturated benchmarks. SEAL seeds candidate outputs into a single elimination and evaluates each match with task-level principles plus self-improving checklist criteria. We evaluate SEAL on multiple saturated benchmarks covering code generation, mathematical reasoning, knowledge-intensive question answering, and tool-use agent task completion. Across these settings, SEAL improves the ranking-accuracy–latency trade-off over competing protocols, attaining 0.83–1.00 Spearman agreement with full pairwise judging and 4/4 top-1 agreement, while requiring only 11.89 calls per task compared with 28.00 for full pairwise evaluation.
[NLP-29] DirectorBench: Diagnosing Long-Form Video Generation with Personalized Multi-Agent Evaluation
【速读】： 该论文试图解决长时视频生成（long-form video generation）中缺乏有效评估手段的问题，特别是现有基准测试在局部视觉质量、短期时序一致性或通用提示对齐方面存在局限，难以诊断生成流程中的瓶颈以及捕捉用户个性化偏好。解决方案的关键在于提出DirectorBench——一个面向长时视频生成的个性化多智能体诊断基准，其通过80项结构化元数据、7种用户画像和40个检查点标准，在脚本、视觉、音频、跨模态和稳定性五个维度上进行细粒度评估，避免单一综合评分，而是定位到具体检查点的性能瓶颈，并支持基于用户画像的差异化评价。实验表明，DirectorBench能够揭示不同生成流程与用户群体间的失败模式，且与人工评估高度一致，验证了诊断式与个性化评估对于长时视频生成的重要性。
链接: https://arxiv.org/abs/2605.30090
作者: Jiamin Chen,Qianben Chen,Jiawen Zhang,Yidi Wu,Yuchen Li,Xiaokun Zhang,Wangchunshu Zhou,Chen Ma
机构: ByteDance(字节跳动)
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Long-form video generation is rapidly moving from short, single-scene synthesis toward minute-long, multi-shot creation with narrative structure, cinematic control, audio, and cross-modal synchronization. However, evaluating such videos remains challenging, since existing benchmarks largely focus on local visual quality, short-horizon temporal consistency, or generic prompt alignment, and provide limited diagnosis of workflow failures and user-dependent preferences. We introduce DirectorBench, a personalized multi-agent diagnostic benchmark for long-form video generation. DirectorBench evaluates generated videos with respect to 80 structured metadata entries, 7 user profiles, and 40 checkpoint criteria across 5 dimensions: script, visual, audio, cross-modal, and stability. Instead of reducing quality to a single aggregate score, DirectorBench localizes checkpoint-level bottlenecks and supports profile-aware evaluation. We evaluate 4 long-form video generation workflows, 6 base LLMs, and 7 user profiles. Across workflows, DirectorBench reveals a between-unit bottleneck: transition quality averages only 0.256 and reaches 0.356 for the best workflow, while prompt-level user demand fulfillment averages 0.71. We further conduct human evaluation with 14 annotators to validate the alignment between DirectorBench and human judgment. The results show that DirectorBench captures human-perceptible quality differences and reveals workflow- and profile-dependent failure modes that are hidden by aggregate scoring. These findings highlight the importance of diagnostic and profile-aware benchmarking for long-form video generation.
[NLP-30] Conformal Certification of Reasoning Trace Prefixes
【速读】： 该论文试图解决的问题是：语言模型推理轨迹（reasoning traces）通常并非全有或全无，而是包含在关键错误发生前的有效中间步骤；而现有不确定性量化方法仅对最终答案或整个响应进行认证，无法为可安全保留的推理片段比例提供统计保障。解决方案的关键在于提出 CROP（Conformal Reasoning Output Prefixes），这是一种与验证器无关的校准过程，用于干净前缀（clean-prefix）认证：给定任意步级风险代理（step-level risk proxy），CROP 选择一个校准阈值并返回最长的连续前缀，其所有步骤的风险代理均低于该阈值，未通过认证的后缀则被路由至下游审查或修复。在假设交换性（exchangeability）的前提下，CROP 严格控制返回前缀中包含标注错误的边际概率。实验表明，标准的步级指标（如 AUROC）不能充分反映前缀的实用性，因此应以认证前缀长度作为验证器评估的新标准；同时，CROP 在过度保留和不足保留之间取得平衡，提升了下游修复准确性，从而将过程监督（process supervision）、拒答（abstention）与修复（repair）有机结合，形成一条严谨且实用的桥梁。
链接: https://arxiv.org/abs/2605.30085
作者: Matt Y. Cheung,Ashok Veeraraghavan,Hanjie Chen,Guha Balakrishnan
机构: Rice University (莱斯大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注: Code available at this https URL
Abstract:Language model reasoning traces are rarely all-or-nothing; they frequently contain valid intermediate steps before a critical error occurs. Existing uncertainty quantification methods typically certify final answers or entire responses, failing to provide statistical guarantees for the proportion of a sequential trace that can be safely retained. To address this, we introduce CROP (Conformal Reasoning Output Prefixes), a verifier-agnostic calibration procedure for clean-prefix certification. Given any step-level risk proxy, CROP selects a calibrated threshold and returns the longest contiguous prefix whose step risk proxies remain below it, routing the uncertified suffix for downstream review or repair. Assuming exchangeability, CROP rigorously controls the marginal probability that the returned prefix contains an annotated error. Across six process-labeled reasoning datasets, we demonstrate that standard step-level metrics such as AUROC do not fully capture prefix utility, suggesting verifiers should instead be evaluated by certified prefix length. Furthermore, CROP balances over- and under-withholding, improving downstream repair accuracy by preserving valid intermediate reasoning while discarding misleading suffixes. Ultimately, this work positions prefix certification as a rigorous, practical bridge between process supervision, abstention, and repair.
[NLP-31] Adaptive Targeted Dynamic Chunking for Tokenization-Free Hierarchical Model
【速读】： 该论文试图解决的问题是：tokenization-free层次模型（如字节级模型）在优化压缩比（compression ratio）方面的挑战，而压缩比直接影响模型处理字节数据时的性能表现。传统方法通常采用固定压缩比，导致训练不稳定且难以适应不同任务需求。解决方案的关键在于提出自适应目标动态分块（Adaptive Targeted Dynamic Chunking, ATDC），这是一种基于课程学习（curriculum learning）的字节压缩控制机制，通过训练过程中从低压缩比逐步过渡到高压缩比，稳定学习过程，并建立目标压缩比与每最内层块字节数（Bytes-Per-Innermost-Chunk, BPIC）之间的理论关系，从而实现对分块大小演化的有效追踪和调控。实验表明，ATDC显著提升了模型训练稳定性与下游任务性能，同时保持了字节级处理的鲁棒性和灵活性。
链接: https://arxiv.org/abs/2605.30080
作者: Thang Dang,Akira Nakagawa,Kenichi Kobayashi,Koichi Shirahata
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Tokenization-free hierarchical models are emerging as a promising alternative to traditional Large Language Models (LLMs), addressing inherent preprocessing issues such as vocabulary design complexity, out-of-vocabulary (OOV) errors, and language-specific constraints. However, a significant challenge in these byte-level methods is the optimization of the compression ratio, a critical factor that dictates model performance for processing bytes data via chunks. In this paper, we propose Adaptive Targeted Dynamic Chunking (ATDC), a novel byte-compression control mechanism designed to enhance the effectiveness of dynamic chunking within hierarchical architectures. Our approach utilizes curriculum learning to progressively adjust the compression ratio during training, transitioning from low to high compression to stabilize the learning process. We provide an analysis establishing the relationship between the target compression ratio and Bytes-Per-Innermost-Chunk (BPIC), allowing for tracking of chunk-size evolution throughout the training phase. Evaluations conducted on the FineWeb-Edu 100B dataset demonstrate that hierarchical models equipped with ATDC achieve competitive Bits-Per-Byte (BPB) performance compared to conventional baselines operating at both byte and token levels. Furthermore, the proposed method exhibits more stable training dynamics and superior final performance across diverse downstream tasks compared to models using fixed compression ratios, while maintaining the inherent robustness and flexibility of byte-level processing.
[NLP-32] UniSteer: Text-Guided Flow Matching in Activation Space for Versatile LLM Steering
【速读】： 该论文试图解决现有基于激活（activation-based）控制方法在大语言模型（LLM）行为调控中面临的适应性不足问题，尤其是固定干预方向或任务特定模块难以应对细粒度概念和组合约束的局限。其解决方案的关键在于提出UniSteer——一种文本引导的激活流匹配模型，通过学习从自然语言条件到残差流（residual-stream）激活空间的条件分布，构建一个通用的激活空间条件速度场（conditional velocity field）。该模型在推理时通过流反演（flow inversion）将源激活部分传输至潜在状态，并在目标文本条件下重构后再注入冻结的LLM，从而实现统一的行为控制、真实性引导、细粒度概念操控、多约束指令遵循及激活空间分类。
链接: https://arxiv.org/abs/2605.30076
作者: Yingdong Shi,Ruiming Zhang,Changming Li,Zhiyu Yang,Kaixing Zhang,Jingyi Yu,Kan Ren
机构: ShanghaiTech University (上海科技大学)
类目: Computation and Language (cs.CL)
备注: 16 pages,4 figures
Abstract:Activation-based control steers large language models (LLMs) by intervening on their internal representations during inference, and has emerged as an effective paradigm for controlling behaviors such as persona and style. However, existing methods often rely on fixed steering directions or task-specific intervention modules, making them difficult to adapt to fine-grained concepts and compositional constraints. We propose UniSteer, a text-guided activation flow matching model that learns a conditional distribution over residual-stream activations from natural-language conditions. Instead of fitting a separate intervention for each target behavior, UniSteer learns a universal conditional velocity field in activation space. At inference time, UniSteer performs flow inversion by partially transporting a source activation toward a latent state and regenerating it under a target textual condition before injecting it back into the frozen LLM. The same conditional model supports activation-space classification by selecting the textual label with the lowest reconstruction energy. Experiments on three target LLMs show that UniSteer provides a unified interface across behavioral control, truthfulness steering, fine-grained concept steering, multi-constraint instruction following, and activation-space classification.
[NLP-33] HEART-Bench: Do LLM Agents Exhibit Human-like Psychology?
【速读】： 该论文试图解决的问题是：当前大型语言模型（LLM）代理在任务导向能力（如规划、推理和行动）上表现突出，但缺乏对人类个性中情感维度的系统性建模与评估，即如何让LLM代理模拟出具有内在一致性的人类心理状态。解决方案的关键在于构建一个全新的基准测试体系，该体系通过11个基于大五人格特质（Big Five personality traits）正交分布的人类角色，每个角色整合了1000条结构化的自传体情景记忆（episodic memories），并按照理论指导的发展阶段进行分布；同时设计了64个决策场景，依据DIAMONDS心理框架（涵盖责任、智力、逆境、择偶、积极、消极、欺骗和社会性八个维度）来评估LLM代理是否能将个体人格特质与记忆整合，从而做出与其心理画像一致的行为决策。最终通过人工验证和筛选，形成包含673道多选题（MCQs）的基准数据集，为研究LLM代理中人类情感模拟、人格一致性及价值导向行为决策提供了一个原则性强且可扩展的测试平台。
链接: https://arxiv.org/abs/2605.30058
作者: Weihan Peng,Chenxu Zhang,Qianao Wang,Yuling Shi,Heng Lian,Qihong Mao,Jiahao Pang,Chunliang Feng,Bowen Li,Xiaodong Gu
机构: 未知
类目: Computation and Language (cs.CL)
备注: GitHub: this https URL
Abstract:While LLM agents have demonstrated remarkable task-oriented abilities such as planning, reasoning, and action, few works have treated them as complete human personalities where emotional dimensions hold equal importance. In this paper, we introduce a novel benchmark to systematically assess whether LLM agents can simulate coherent, human-like psychology. Specifically, our benchmark constructs 11 diverse human characters grounded in orthogonal Big Five personality traits, with each profile deeply integrated with 1,000 structured autobiographical-style episodic memories distributed across theory-grounded developmental life stages. To rigorously evaluate the psychological manifestations of LLMs, we designed a curated suite of 64 decision-making scenarios, guided by the DIAMONDS taxonomy, a psychological framework that characterizes situations along eight dimensions: Duty, Intellect, Adversity, Mating, pOsitivity, Negativity, Deception, and Sociality. By subjecting agents to varying scenarios, the benchmark evaluates whether they can consolidate their innate personality traits and autobiographical memories to make behavioral decisions that are consistent with their specific psychological profiles. After systematic human validation and filtering, we obtained a benchmark consisting of 673 multiple-choice questions (MCQs). We believe this benchmark provides a principled and scalable testbed for studying human-like emotions, personality consistency, and value-consistent behavioural decision-making in LLM-based agents.
[NLP-34] REPOT: Recoverable Program-of-Thought via Checkpoint Repair
【速读】： 该论文试图解决生成式 AI (Generative AI) 在执行复杂任务时因单个无效动作导致整个推理轨迹失效的问题，尤其在基于程序思维链（Program-of-Thought, PoT）的框架下，这种“一击即溃”式的失败模式显著限制了系统鲁棒性。其解决方案的关键在于提出 RePoT（Recoverable PoT），通过引入确定性的可验证重放机制：首先在环境中执行计划直至首次无效转移点，然后仅需一次大语言模型（LLM）调用即可从已验证前缀处恢复执行，从而避免重新生成整个计划。该方法在约14%的PoT失败场景中最多增加一次LLM调用成本，但在多个基准测试中（如PuzzleZoo-775、PlanBench Blocksworld和Derail-550）均实现显著性能提升，且实验证明checkpoint信息是恢复成功的核心信号，而非具体前缀内容本身。
链接: https://arxiv.org/abs/2605.30052
作者: Parsa Mazaheri
机构: University of California, Santa Cruz
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:One-shot Program-of-Thought (PoT) emits a Python program that prints a primitive-action plan; a single invalid action silently invalidates the trajectory. We introduce RePoT (Recoverable PoT): a deterministic verified replay that walks the plan through the environment to its first invalid transition, then one LLM call that resumes from the verified prefix. RePoT costs at most one extra LLM call on the ~14% of problems where PoT fails. RePoT beats PoT by +3 to +11pp across four closed-model configurations on PuzzleZoo-775 and peaks at 96.9% vs 86.3% on gpt-5.4-mini-medium; against the matched-budget PoT-retry baseline, RePoT wins decisively on Gemini (+3.8pp, 95% CI [+2.2,+5.4]), is within sampling noise on GPT-medium and Claude, and loses on GPT-mini – a capability-scaling pattern we begin to address with Adaptive RePoT, a rule-based dispatcher that routes between suffix repair and a fresh PoT retry based on verified-prefix length (preliminary). We replicate on PlanBench Blocksworld (+1.1 to +11.4pp) and on four open-weights models (+3.3 to +20.0pp on three of four). On Derail-550, our controlled recovery benchmark, every condition with access to checkpoint information clears =30% on GPT-medium and =70% on Gemini, vs =3.1% for error-only feedback – showing that checkpoint information, not the specific verified-prefix tail, is the load-bearing recovery signal.
[NLP-35] Who Am I? History-Aware Profiles for Student Simulation in Tutoring Dialogues
【速读】： 该论文试图解决的问题是：当前基于大语言模型（LLM）的自动化教学工具中，学生模拟（student simulation）主要局限于对话内部的模拟，缺乏对学生知识状态和行为的历史上下文信息，导致模拟不够真实。解决方案的关键在于提出了一种历史条件下的学生模拟任务（history-conditioned student simulation），其核心是一个两组件框架：一是学生画像生成器（profile generator），用于从学生的历史学习记录中提取并总结关键特征；二是模拟器（simulator），根据生成的画像预测学生的后续对话轮次。两个组件均采用强化学习（RL）进行训练，使生成的学生画像更有利于忠实还原学生的行为模式。实验结果表明，该方法在首个真实世界数学学习平台收集的学生对话与答题数据集上显著优于基线模型，验证了历史信息、学生画像以及强化学习训练的重要性。
链接: https://arxiv.org/abs/2605.30051
作者: Zhangqi Duan,Shuyan Huang,Alexander Scarlatos,Jaewook Lee,Simon Woodhead,Andrew Lan
机构: University of Massachusetts Amherst (马萨诸塞大学阿默斯特分校); Eedi
类目: Computation and Language (cs.CL); Computers and Society (cs.CY)
备注:
Abstract:A key part of developing large language model (LLM)-powered, automated tutoring tools is student simulation, i.e., using LLMs to role-play as students, which can facilitate tutor model evaluation and training. Existing work mostly focuses on within-dialogue simulation, which lacks context on student knowledge and behavior, partly due to not grounding in past student question-answering or dialogue interactions. In this work, we introduce the task of history-conditioned student simulation, where the goal is to accurately predict student dialogue turns by leveraging information in the student’s learning history. We propose a two-component framework in which a profile generator summarizes a student’s history and a simulator predicts student turns conditioned on the resulting profile. We train both components with reinforcement learning (RL), yielding profiles optimized for faithful student simulation. We evaluate our method and baselines on the first-of-its-kind real-world dataset of student dialogues and question responses that we collect from a math learning platform. Extensive experiments show that our method significantly outperforms baselines, and demonstrate the importance of history, profiles, and RL training.
[NLP-36] oken Inflation: How Dishonest Providers Can Overcharge for Large Language Model Usage
【速读】： 该论文试图解决的问题是：当前基于“按token计费”的大型语言模型（LLM）定价模式下，用户支付金额依赖于服务商报告的token数量，但这种计费方式在设计上难以审计，导致服务商可能通过操纵token计数进行欺诈性收费。解决方案的关键在于识别并打破“信任悖论”——即任何审计都必须依赖某个可验证的证据，而现有框架所依赖的证据恰恰是服务商最有可能篡改的部分。研究发现，即使在最严格的条件下，服务商也能利用隐藏推理过程或token化模糊性实现高达1,469%的token计数虚报，且不会被检测到；即便用户可见完整推理字符串，仅token化歧义仍可造成50.85%的超额计费。因此，恢复诚实计费的核心在于建立与服务商无关的验证机制，例如可信执行环境证明、推理过程的密码学证明或第三方重新执行验证。
链接: https://arxiv.org/abs/2605.30040
作者: Shahinul Hoque,Jinghuai Zhang,Jinyuan Sun,Fnu Suya
机构: University of Tennessee, Knoxville (田纳西大学诺克斯维尔分校); University of California, Los Angeles (加州大学洛杉矶分校)
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Per-token billing is now the standard pricing model for commercial large language models (LLMs), so the honesty of reported token counts directly affects what users pay. We show that this kind of billing is hard to audit by design: providers hide the model, the tokenizer, and the execution to protect their IP, mitigate jailbreaks, and preserve user privacy, which means an auditor can only inspect proofs the provider supplies. The audit therefore reduces to a consistency check on the provider’s own reports. We call this a trust paradox: every audit must trust some artifact, but current frameworks trust exactly the ones a provider has the strongest reason to manipulate. We study three recent token auditing frameworks and show that a provider with ordinary commercial capabilities can systematically inflate billed token counts. In the most permissive setting, hidden reasoning usage can be inflated by 1,469% on average without detection. At current frontier reasoning prices, that turns a \ 100 honest bill into roughly a \ 1,569 bill on the same query. Even when the user can see the full reasoning string, tokenization ambiguity alone still allows 50.85% over-reporting below the detection threshold. These results suggest the problem is not in any specific auditor but in any audit whose evidence comes from the audited party. Restoring honest billing will require verification that ties reported token counts to evidence the provider does not control, such as trusted execution attestation, cryptographic proofs of inference, or third-party re-execution.
[NLP-37] aching Values to Machines: Simulating Human-Like Behavior in LLM s ACL2026
【速读】： 该论文试图解决的问题是：大型语言模型（LLMs）是否能够表现出符合人类一致价值体系的行为，即其行为是否具备心理学意义上的“人类价值观一致性”。解决方案的关键在于：基于成熟的心理学价值理论（如舒伯特的价值理论），通过设计特定的提示（value-prompting）来诱导LLMs展现出与人类相似的价值结构，并利用经过验证的心理学量表进行大规模实验（超过500万次问答）评估LLMs与人类在价值观结构及价值观-行为关系上的匹配度。研究发现，经价值提示引导的LLMs与人类在两个维度上均表现出高度一致性，且引入人类价值分布可显著提升LLMs在群体层面的行为模拟效果，表明价值诱导的LLMs是具有心理基础、可有效模拟人类行为的工具。
链接: https://arxiv.org/abs/2605.30036
作者: Asaf Yehudai,Naama Rozen,Ariel Gera
机构: The Hebrew University of Jerusalem (希伯来大学); IBM Research (IBM研究院); Tel-Aviv University (特拉维夫大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: GEM Workshop at ACL 2026
Abstract:Large Language Models (LLMs) demonstrate a remarkable capacity to adopt different personas and roles; however, it remains unclear whether they can manifest behavior that adheres to a coherent, human-like value structure. In this work, we draw on established psychological value theory to induce human-like values in LLMs and assess their alignment with patterns observed in human studies. Using validated psychological questionnaires, we conduct large-scale experiments – over 5 million questions – to evaluate value structures and value-behavior relationships in leading LLMs and compare them to humans. Our findings reveal strong agreement between value-prompted LLMs and humans across both dimensions. Moreover, incorporating human value distributions enhances population-level simulations with value-induced LLMs. These findings highlight the potential of value-induced LLMs as effective, psychologically grounded tools for simulating human behavior.
[NLP-38] Audio Jailbreaks in Large Audio-Language Models: Taxonomy Attack-Defense Analysis and Cost-Aware Evaluation ACL
【速读】： 该论文旨在解决大型音频语言模型（Large Audio Language Models, LALMs）在语音感知到推理全链路中面临的越狱攻击（jailbreak）风险问题，这类风险不再局限于文本token层面，而是扩展至语义、声学风格、信号伪影及内部表征等多个维度。其解决方案的关键在于提出一个统一的分类体系和受控的实证评估框架：将攻击细分为语义、声学、信号和嵌入层四类，防御策略分为基于守卫、无需训练和训练驱动三类，并设计跨模态、纯音频和交互式三类基准测试。通过在十种开源LALM上系统评估代表性攻击与防御方法，不仅量化攻击成功率，还引入良性拒绝率和延迟作为核心指标，揭示了“声学最佳N”存在显著最坏情况漏洞、“叙事框架”是一种低延迟语义威胁，而现有防御手段往往以牺牲良性可用性为代价换取鲁棒性，从而强调成本与效用权衡是衡量模型安全性的必要补充。
链接: https://arxiv.org/abs/2605.30031
作者: Bo-Han Feng,Yu-Hsuan Li Liang,Chien-Feng Liu,You-Hsuan Chang,Yun-Nung Chen
机构: National Taiwan University (国立台湾大学)
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Submitted to ACL ARR 2026 May
Abstract:Large Audio Language Models (LALMs) expand jailbreak risks from token-level prompting to the full speech perception-to-reasoning pipeline, where unsafe behavior can be induced through semantics, acoustic style, signal artifacts, or internal representations. Existing work studies these risks under heterogeneous threat models and evaluation protocols, making it difficult to compare attack practicality or defense utility. This paper provides a unified taxonomy and a controlled empirical evaluation of LALM jailbreak attacks and defenses. We organize prior work into semantic, acoustic, signal, and embedding-layer attacks; guard-based, training-free, and training-based defenses; and cross-modal, audio-native, and interactive benchmarks. We then evaluate representative attacks and defenses across ten open-source LALMs, measuring not only attack success rate but also benign refusal and latency. Our results show that Acoustic Best-of-N reveals strong worst-case audio-space vulnerabilities, Narrative Framing is an effective low-latency semantic threat, and current defenses trade robustness against benign usability. These findings support cost- and utility-aware evaluation as a necessary complement to success-rate-only LALM safety benchmarks.
[NLP-39] Give it Space! Explicit Disentangling of Positional and Semantic Representations in Encoders
【速读】： 该论文试图解决的问题是：当前基于位置编码（Positional Encoding, PE）的Transformer模型在处理长文本上下文或检索任务时表现不佳，且对位置信息如何在模型内部被加工和存储的理解仍不充分。解决方案的关键在于通过解耦机制将位置信息与语义信息分离——具体而言，作者构建了一个编码器Transformer，显式地处理三个独立的信号流：语义流、绝对位置（Absolute Position, AP）流和相对位置（Relative Position, RP）流，并将掩码语言建模（Masked Language Modeling, MLM）目标限制在语义流中。这种解耦设计使得能够清晰地分析各子空间的功能，从而揭示出：AP子空间自发形成一个低频二维流形以捕获文档结构，注意力头呈现结构导向与语义导向的分工，且标准PE方法（如RoPE）无法稳健保留宏观结构信息；相比之下，该解耦方法有效维持了位置编码能力，在Flash-Holmes探测基准的65项语言现象中提升了49项的语言表征性能。
链接: https://arxiv.org/abs/2605.30022
作者: Pierre-Antoine Lequeu,Camille Barboule,Benjamin Piwowarski
机构: Sorbonne Université, CNRS, ISIR, Paris, France; Orange Innovation
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 8 page + 10 pages of bibliography and appendix
Abstract:Positional encoding (PE) underpins how permutation-invariant Transformers represent sequence order, yet how positional information is processed and stored remains poorly understood. Modern PE methods such as RoPE still struggle on tasks such as long-context understanding or retrieval \citechen-etal-2025-hope. Hence, a better understanding of the internal positional mechanism could help design better PE. Building on evidence that positional and semantic signals occupy nearly orthogonal subspaces in trained Transformers, we modify an encoder Transformer to process three explicitly disentangled streams: semantic, absolute positional (AP) and relative positional (RP), and confine the masked-language-modeling (MLM) objective to the semantic stream. This decoupling enables a clean mechanistic study and yields three take-aways. (1) The isolated AP subspace spontaneously collapses into a low-frequency two-dimensional manifold that captures the structure of the document; (2) Attention heads specialize into structure and semantic-oriented groups, with RP exclusively supporting the latter; (3) Standard positional encodings do not robustly retain macroscopic structure: RoPE and RP only weakly encode it, and entangled AP loses it in the final layers under MLM pressure. The disentangled approach preserves positional encoding, which improves linguistic representation on 49 of the 65 linguistic phenomena of the Flash-Holmes probing benchmark.
[NLP-40] Recovering Diversity Without Losing Alignment: A DPO Recipe for Post-Trained LLM s
【速读】： 该论文试图解决大语言模型（Large Language Model, LLM）在后训练阶段因偏好优化（Preference Optimization, PO）导致输出空间过度收敛的问题，即模型倾向于生成少数“标准答案”，从而限制了对开放性指令的多样性和实用性响应。其解决方案的关键在于提出一种离线的DPO数据构建流程——REDIPO，通过从基础模型（Base Model）采样原始响应、利用指令微调模型（Instruct Model）重写这些响应以提升质量、基于安全性和指令遵循度过滤候选答案，并构建偏好对（Preference Pairs），其中优先选择具有边际多样性的响应（即在相似指令遵循奖励下更具差异性的回答）。实验表明，REDIPO在多个主流模型上显著提升了多样性指标（如NoveltyBench distinct_k），同时保持了原有对齐性能（如MTBench、IFEval等），并降低了有害攻击成功率，验证了其在不牺牲对齐质量的前提下恢复有效多样性响应的有效性。
链接: https://arxiv.org/abs/2605.30021
作者: Vinay Samuel,Yapei Chang,Mohit Iyyer
机构: University of Maryland, College Park
类目: Computation and Language (cs.CL)
备注: Under Review. 26 pages, 3 figures, 16 tables
Abstract:Many open-ended instructions have multiple valid answers that users can benefit from seeing, but post-training often narrows an LLM’s output space toward a small set of canonical responses. We introduce REDIPO, an offline DPO data-construction pipeline for recovering distinct valid answer modes while preserving the alignment benefits of the instruct model. For each prompt, REDIPO samples responses from both base and instruct models, rewrites base-model responses with the instruct model, filters candidates for safety and instruction-following quality, and builds preference pairs that favor marginally diverse responses among candidates with similar instruction-following reward. Across Qwen3-4B, OLMo-3-7B, and LLaMA-3.1-8B, REDIPO improves NoveltyBench distinct_k by 134%, 33%, and 44% relative to the instruct checkpoints, while DivPO changes diversity by 0%, -6%, and -4% on the same models. These gains largely maintain MTBench, IFEval, and Arena-Hard performance, and reduce direct-category HarmBench attack success rate. Ablations show that marginal-diversity pair selection and base-response rewriting drive the diversity gains, while filtering and quality-bounded pairing help maintain alignment. Overall, our results show that diverse valid answers from base-model generations can be reintroduced through carefully constructed preference data while retaining the alignment benefits of post-training. We release our code and data at this https URL.
[NLP-41] Latent Performance Profiling of Large Language Models
【速读】： 该论文试图解决的问题是：当前基于基准测试（benchmark-based evaluation）的大型语言模型（LLM）评估方法存在局限性，如数据污染、任务范围狭窄以及与真实世界可靠性对齐不足，且仅能反映模型在固定测试集上的输出准确性，无法揭示其信息处理机制、不确定性校准能力或内部知识结构。解决方案的关键在于提出一种全新的“状态中心”内在评估框架——潜在性能剖析（Latent Performance Profiling, LPP），该框架通过分析模型隐藏层激活和输出分布来提取与任务无关的诊断指标，从而刻画模型在不同规模下的稳定、可解释的潜在特性（如熵、适应性等）。LPP不仅能够识别出相似基准分数下模型间隐含的行为差异，还能指导设计去偏的合成探测器（synthetic probes）用于不确定性建模与符号推理能力评估，最终推动从表面准确率向更深层次模型行为理解的范式转变。
链接: https://arxiv.org/abs/2605.30018
作者: Tanmoy Chakraborty,Ayan Sengupta,Suparna Bhattacharya,Partha Pratim Chakrabarti,Amlan Chakrabarti,Supratik Chakraborty,Partha Pratim Das,Lipika Dey,Richa Singh,Mayank Vatsa
机构: Indian Institute of Technology Delhi (印度理工学院德里分校); Indian Institute of Technology Kharagpur (印度理工学院克哈格普尔分校); University of Calcutta (加尔各答大学); Indian Institute of Technology Bombay (印度理工学院孟买分校); Ashoka University (阿舒卡大学); Indian Institute of Technology Jodhpur (印度理工学院乔德普尔分校)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Large language models (LLMs) frequently achieve impressive scores on standardized benchmarks, yet accuracy alone offers a limited view of their capabilities. Evaluating open-source LLMs through leaderboards faces persistent issues like data contamination, narrow task scope, and weak alignment with real-world reliability. Benchmark-based evaluations such as MMLU PRO, BBH, or IFEval primarily capture \textitwhat a model outputs on fixed test sets, not \textithow it processes information, calibrates uncertainty, or structures internal knowledge. In this article, we advocate for a shift from benchmark-centric evaluation toward a complementary, \textitstate-centered intrinsic assessment of LLMs. To this end, we introduce \textbfLatent Performance Profiling (LPP) – a framework that derives task-agnostic diagnostics from hidden activations and output distributions. LPP defines a set of scalar metrics on a model’s latent representations and dynamics, revealing scale-independent traits that enable interpretable comparisons and uncover hidden vulnerabilities. Unlike static accuracy scores, LPP provides stable, architecture-sensitive signatures across models of similar size. With extensive empirical analyses across eight LLMs, spanning a size range of 0.5B-14B, we demonstrate that models with similar benchmark scores can exhibit contrasting latent profiles, such as differences in entropy or adaptability. Guided by these insights, we design synthetic probes for uncertainty and symbolic reasoning that align with intrinsic metrics while decoupling from leaderboard bias. We recommend that reporting LPP alongside benchmarks provides a deeper, interpretable understanding of model behavior, enabling more reliable model selection, safety assessment, and evaluation beyond surface-level accuracy.
[NLP-42] Adapting Multilingual Embedding Models to Turkish via Cross-Lingual Tokenizer Surgery and Offline Distillation
【速读】： 该论文旨在解决土耳其语（Turkish）句子嵌入（sentence embedding）模型在上下文长度限制、训练效率与性能之间的平衡问题。现有基于BERT的土耳其语编码器普遍受限于512个token的上下文窗口，难以捕捉长文本语义信息；同时，全量预训练成本高昂且难以适配多语言场景。解决方案的关键在于提出一种高效的三阶段适配流水线：首先构建一个优化后的多语言分词器（vocabulary size: 131,072），通过修剪冗余token并融合高频多语言token实现土耳其语友好性；其次克隆教师模型结构并利用均值组合映射初始化新词汇表嵌入层，保留Transformer骨干权重；最后采用离线嵌入蒸馏（embedding distillation）策略，在40种语言的维基百科数据上以余弦相似度为目标函数进行知识迁移，避免在线教师推理开销。该方法使学生模型（约200M参数）仅需单GPU训练约4小时即可达到优于3亿参数教师模型的性能（STSbTR任务Pearson相关系数达77.55%），并在TR-MTEB基准上取得63.9%的平均得分（26个模型中排名第7），展现出显著的成本-质量优势。
链接: https://arxiv.org/abs/2605.29992
作者: M. Ali Bayram,Banu Diri,Savaş Yıldırım
机构: Yıldız Technical University (伊斯坦布尔技术大学); Istanbul Bilgi University (伊斯坦布尔比尔吉大学)
类目: Computation and Language (cs.CL)
备注: 14 pages, 2 figures, 4 tables, Appendix included
Abstract:Sentence embeddings are a foundational component for semantic search, clustering, classification, and retrieval-augmented generation. This paper presents embeddingmagibu-200m, a Turkish-focused sentence embedding model that produces 768-dimensional L2-normalized vectors and supports an 8,192-token context window, far exceeding the 512-token limit of earlier BERT-based Turkish encoders. Instead of full pretraining, an efficient three-stage adaptation pipeline is introduced: (1) construct a Turkish-optimized multilingual tokenizer with a 131,072 vocabulary by pruning redundant tokens from the teacher’s vocabulary and incorporating multilingual tokens via frequency analysis on a 40-language corpus, (2) clone a teacher embedding model while preserving transformer backbone weights and initializing a compatible embedding table for the new vocabulary via mean-composition token mapping, and (3) perform offline embedding distillation from precomputed teacher vectors using a cosine similarity objective over a balanced 40-language Wikipedia corpus. The resulting student model contains approximately 200M parameters and trains in roughly four hours on a single GPU by avoiding online teacher inference during training, at a total cost of 5- 20. Empirically, Pearson/Spearman correlations of 77.55%/77.45% are obtained on STSbTR, surpassing the 300M-parameter teacher model (73.84%/72.92%). On TR-MTEB (26 tasks), a mean score of 63.9% is achieved (7th out of 26 models), providing a competitive cost-quality trade-off with 33% fewer parameters than the teacher. To facilitate reproducibility and downstream use, all artifacts are released including model weights, tokenizer files, precomputed embedding datasets, and open-source cloning and distillation tooling.
[NLP-43] MIC: Maximizing Informational Capacity in Adaptive Representations via Isotropic Subspace Alignment ICML2026
【速读】： 该论文旨在解决多尺度表示学习中嵌套子空间存在的维度冗余和谱坍缩（spectral collapse）问题，这些问题会削弱嵌入表示的语义密度与判别能力。其解决方案的关键在于提出MIC框架，通过各向同性子空间对齐优化多粒度嵌入的几何结构：一方面采用软坍缩正则化（Soft Collapse Regularization, SCR）通过交叉相关惩罚减少前缀与残差子空间间的冗余；另一方面引入谱各向同性正则化（Spectral Isotropy Regularization, SIR）确保低维前缀在超球面上均匀分布；最终通过自蒸馏目标统一两种策略，生成语义密集且保持高判别力的嵌入表示。实验表明，MIC在高压缩场景下显著优于标准基线，尤其在维持信息容量方面表现突出。
链接: https://arxiv.org/abs/2605.29987
作者: Dang Hong Nguyen,Nhi Ngoc-Yen Nguyen,Huy-Hieu Pham
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Accepted at the GlobalSouthML Workshop at ICML 2026. 13 pages, 2 figures
Abstract:Although multi-scales representation learning enables elastic-dimension embeddings, nested subspaces often suffer from dimensional redundancy and spectral collapse. To address this, we introduce MIC, a framework that optimizes the geometric landscape of multi-granular embeddings through isotropic subspace alignment. MIC employs Soft Collapse Regularization (SCR) to mitigate redundancy between prefix and residual subspaces via cross-correlation penalties, alongside Spectral Isotropy Regularization (SIR) to ensure hyper-spherical uniformity in low-dimensional prefixes. By unifying these strategies through a self-distillation objective, MIC generates semantically dense representations that maintain high discriminative power. Our experiments demonstrate that MIC significantly outperforms standard baselines, particularly in high-compression scenarios where maintaining informational capacity is most critical.
[NLP-44] Causal Interventions on Continuous Variables: A Case Study on Verb Bias in Steering Vectors for In-Context Learning
【速读】： 该论文试图解决的问题是：如何在语言模型表征中对连续变量（graded variables）实施因果干预，而以往的研究主要集中在离散特征（如语法数）上。解决方案的关键在于提出一种新方法，即从配对的激活向量与连续目标变量中定位一个低维方向，并利用该方向对向量进行编辑，以实现对反事实目标值的调整。作者将该方法应用于心理语言学中广泛研究的连续特征——动词倾向性（verb bias），结果表明语言模型中的引导向量（steering vectors）确实因果性地编码了动词倾向性信息，且对其编辑能系统性地改变下游句法偏好；尽管动词倾向性的引导向量也包含可驱动上下文学习（in-context learning）中误差驱动更新的信号，但这些信号并未被下游生成过程所因果性使用。这说明连续变量的因果干预是可行的，但其与上下文学习之间的机制联系仍待进一步探索。
链接: https://arxiv.org/abs/2605.29971
作者: Zhenghao Herbert Zhou,R. Thomas McCoy,Robert Frank
机构: Yale University (耶鲁大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Causal interventions in language model representations have largely targeted discrete features, like grammatical number. However, language models must also make use of features that are graded. We introduce a method for causal intervention on continuous variables: given activation vectors paired with a graded target variable, we localize a low-dimensional direction for that variable and use this direction to edit a vectors toward counterfactual target values. We apply this method to a continuous feature that is well-studied in psycholinguistics, namely verb bias (which reflects which syntactic structures tend to follow a given verb). We show that verb bias is causally represented in steering vectors extracted from large language models: counterfactual edits to verb bias systematically shift downstream structural preferences. Verb bias has also previously been linked to in-context learning; in further analyses, we find that steering vectors encode error signals that could drive the error-driven update behavior seen in in-context learning but that these aspects of the steering vectors are not causally used in downstream production. Overall, these results show causal interventions can be applied to continuous variables, though connecting continuous variables to in-context learning remains a challenge.
[NLP-45] MuPHI: Learning Implicit Multimodal Harm Reasoning via Semantically Grounded Reward Optimization
【速读】： 该论文试图解决的问题是：当前视觉-语言模型（VLMs）在处理图像与文本的交互时，难以识别由看似无害的配对所隐含的有害语义，尤其是在需要意图感知和跨模态推理的场景下。现有方法依赖于表层特征进行字面推理，但在面对依赖上下文、隐含信息的复杂危害情境时表现不足。解决方案的关键在于提出MuPHI数据集和MuPHIRM训练框架：MuPHI是一个包含多种危害类别且标注了危害推理路径的多模态数据集，用于评估VLM在组合性危害检测中的推理能力；MuPHIRM则是一种基于多视角奖励优化的推理增强训练框架，通过联合学习多模态语义来提升模型的危害检测准确性和推理质量，并展现出优于基线方法的分布外鲁棒性。这一方案表明，以推理为导向的奖励优化策略是构建能超越基准特定捷径的多模态系统的重要方向。
链接: https://arxiv.org/abs/2605.29951
作者: Anisha Saha,Varsha Suresh,Teodora Kamova,Sophia Wiedmann,Timothy Hospedales,Vera Demberg
机构: Max Planck Institute for Informatics, Saarland Informatics Campus; Saarland University; The University of Edinburgh; Samsung AI Center, Cambridge
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Multimedia (cs.MM)
备注:
Abstract:Understanding how harm emerges from interaction between otherwise benign image-text pairs requires intent-aware cross-modal reasoning beyond surface-level features. Existing vision-language models (VLMs) excel at literal reasoning over perceptual cues but often fail to derive harmful semantics that rely on implicit, context-dependent reasoning. To evaluate VLMs on compositional harm detection and reasoning, we introduce Multimodal Pragmatic Harm Interpretation (MuPHI), a dataset containing image-text pairs where harm is encoded in subtle multimodal cues. MuPHI spans diverse harm categories and includes annotated harm rationales for assessing VLM reasoning chains. To improve both detection and reasoning in VLMs, we propose MuPHIRM, a reasoning-augmented training framework which learns joint semantics by optimizing multi-perspective rewards. MuPHIRM improves both harm detection and reasoning quality of VLMs while demonstrating superior out-of-distribution robustness compared to both trained and inference-time baselines. Our findings suggest that reasoning-oriented reward optimization offers a promising direction towards building multimodal systems that generalize beyond benchmark-specific shortcuts.
[NLP-46] Does The Way You Plan Matter? An Empirical Study of Planning Representations for LLM Web Agents EMNLP
【速读】： 该论文试图解决大语言模型（LLM）驱动的网络代理（web agents）在执行任务时存在的探索能力有限、关键步骤遗漏以及对任务约束敏感等问题。研究表明，这些问题很大程度上源于规划阶段的不足，但现有研究尚未系统探讨不同自然语言规划表示方式对代理性能的影响。解决方案的关键在于提出一个静态规划-执行框架 PlanAhead，该框架通过自动将 WebArena 任务划分为三个难度等级（无需人工标注），并系统评估四种不同的规划表示形式（顺序子目标、叙事性描述、伪代码和清单）在高难度任务上的表现。同时，为应对随机性带来的波动，作者引入两个新指标：达成率（Achievement Rate, AR）和已解决问题一致性（Solved-Task Consistency, STC）。实验表明，规划形式本身与生成规划的底层 LLM 均显著影响网络代理的鲁棒性和任务成功率。
链接: https://arxiv.org/abs/2605.29927
作者: Alejandra Zambrano,Sara Vera Marjanovic,Imene Kerboua,Xing Han Lù,Leila Kosseim
机构: Concordia University (康考迪亚大学); Mila - Quebec AI Institute (魁北克AI研究所); University of Copenhagen (哥本哈根大学); Universite Claude Bernard Lyon (克莱蒙·奥弗涅大学); McGill University (麦吉尔大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Extended version of paper submitted to EMNLP, waiting for acceptance
Abstract:Despite recent advances, LLM-based web agents still struggle with limited exploration, omission of critical steps, and sensitivity to task constraints. Prior work suggests that many of these failures stem from weaknesses in planning, yet the impact of alternative natural language plan representation remains unexplored. To address this, we introduce PlanAhead, a static planner-executor framework that evaluates the impact of plan representation in agent performance. We first automatically categorize WebArena tasks into 3 difficulty levels, enabling consistent difficulty grading without human annotation. Then we systematically evaluate 4 different plan representations on the tasks categorized as hard: sequential subgoals, narrative, pseudocode, and checklist; across different families of multimodal LLM powered agents (OpenAI, Alibaba, and Google). To account for stochastic variability, we introduce two novel evaluation metrics: Achievement Rate (AR) and Solved-Task Consistency (STC). Our results show that both, the plan formulation and the underlying LLM generating the plan, significantly influence web-agent robustness and task success.
[NLP-47] ExCAM: Explainable Cultural Awareness Metrics
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）在跨文化场景下缺乏可解释、细粒度文化意识评估的问题。当前基准测试多依赖人工标注的问答或生成任务，存在成本高、时效性差及对自由文本评估不足等局限。解决方案的关键在于提出ExCAM（Explainable Cultural Awareness Metric），这是首个能够识别、评分并解释指令-输出对中文化错误的专用评估指标。其核心创新在于构建了ExCAM40k数据集，整合九个现有基准并引入合成错误以增强多样性与挑战性，从而实现对LLMs文化偏差的高效检测与可解释分析。实验表明，ExCAM在平衡测试集上达到最高80%的错误检测准确率，显著优于包括GPT-5在内的多个基线方法，为全球范围内文化敏感性文本生成提供了可扩展、可解释的评估路径。
链接: https://arxiv.org/abs/2605.29897
作者: Christoph Leiter,Haiyue Song,Hour Kaing,Jin Tei,Hideki Tanaka,Masao Utiyama,Steffen Eger
机构: University of Mannheim, Germany; University of Technology Nuremberg, Germany; National Institute of Information and Communications Technology
类目: Computation and Language (cs.CL)
备注: preprint
Abstract:Evaluating the cultural awareness of large language models is crucial to ensure the fairness of generated text and the generalizability of applications across the world. Recent benchmarks explore cultural goods like food or values like behavior in stressful situations through the lens of question answering or text generation tasks. However, creating these benchmarks requires time-intensive and costly human annotations. Also, benchmarks that evaluate cultural awareness in free text are scarce and often rely on dated evaluation mechanisms. To address this gap, we introduce ExCAM, an Explainable Cultural Awareness Metric, which is, to our knowledge, the first dedicated evaluation metric that identifies, rates and explains cultural errors in instruction-output pairs. To train and evaluate ExCAM, we introduce ExCAM40k, a dataset comprised of nine existing benchmarks that we reformat and enhance with synthetic errors. Compared to several baselines, including GPT-5, ExCAM achieves the highest error detection rate with up to 80% accuracy on a balanced test set. Therefore, ExCAM opens the pathway towards fine-grained and explainable cultural evaluation of free text.
[NLP-48] Internal Representation Not Clinical Knowledge: Where Apparent LLM Triage Failures Originate
【速读】： 该论文试图解决的问题是：当前消费者级大语言模型（LLM）在医疗分诊任务中，采用受限的多选输出格式时存在显著的误分诊率（under-triage），而相同病例在自由文本输出格式下表现更好。研究的核心问题是：这种性能差异是否源于模型对临床信息的表征（clinical representation）不同，还是仅由输出格式导致的决策映射机制差异。解决方案的关键在于通过多种方法验证模型内部表示的一致性——利用稀疏自编码器（sparse-autoencoder, SAE）提取Gemma 3 4B/12B和Qwen3-8B模型中的医学特征，发现无论输出格式如何，相同的临床叙事触发的医学特征均保持激活；但在多选决策标记（decision token）处，这些医学特征全部沉默。进一步分析表明，驱动决策日志概率的是“结构化”或“格式相关”特征（scaffold and format features），而非医学内容特征。行为实验也证实，错误主要表现为相邻选项选择（off-by-one decision），而非知识缺失，且选项顺序打乱排除了位置偏倚。因此，模型失败的根本原因在于输出格式的设计，而非其对临床语义的理解能力。
链接: https://arxiv.org/abs/2605.29889
作者: David Fraile Navarro,Berardino Como,Jialei Sheng,Soundariya Ananthan,Shlomo Berkovsky
机构: Macquarie University (麦考瑞大学); Politecnico di Bari (巴里理工学院); NSW Health (新南威尔士州卫生局)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 9 pages main text, 27 pages total including appendices; 7 figures, 25 tables
Abstract:Patient-voiced clinical-triage benchmarks report high under-triage rates for consumer LLMs for constrained multiple-choice output, yet the same cases score differently with free-text. We ask whether output format changes the model’s \emphclinical representation or only the mapping from a preserved representation to an answer. Using sparse-autoencoder (SAE) features in Gemma 3 4B/12B IT and Qwen3-8B, we find the same medical features fire on the shared clinical narrative under both formats but go silent at the multiple-choice decision token in all the cases at every model. Three independent methods (natural-language autoencoder verbalization, decision-token logit attribution, and top-feature characterization) agree that scaffold and format features, but not medical features, drive the decision logits. Behaviorally, the multiple-choice penalty inverts under both structured and natural-language input, option-order shuffle rules out positional bias, and the gap is dominated by off-by-one decision (the model picks an adjacent acuity letter to the gold answer) rather than knowledge failure. Thus, the failure originates in the output format and not in the clinical representation.
[NLP-49] CRITIC-R1: Learning Structured Critics for Retrieval-Augmented Generation
【速读】： 该论文旨在解决现有检索增强生成（Retrieval-Augmented Generation, RAG）方法在知识密集型问答任务中仍存在的幻觉（hallucination）和细微推理错误问题。现有基于外部批评者（external critics）的改进方案往往提供粗粒度、结构松散的反馈，存在过度干预和噪声干扰等问题，导致修正效果不稳定。其解决方案的关键在于提出CRITIC-R1框架，该框架将RAG批评建模为一个显式的错误诊断问题，通过强化学习（Reinforcement Learning, RL）进行端到端训练。具体而言，CRITIC-R1定义了四个诊断维度：判断结论（verdict）、错误位置（error location）、推理分析（reasoning analysis）与修复建议（fix generation），并设计两种奖励机制——保守判断对齐（Conservative Judgement Alignment, CJA）以抑制过度纠正倾向，诊断质量对齐（Diagnostic Quality Alignment, DQA）通过门控奖励提升细粒度反馈质量。模型采用基于GRPO的强化学习策略，并利用来自大语言模型（LLM）教师的流程级监督信号进行训练，在五个问答基准测试中显著优于主流RAG基线方法。
链接: https://arxiv.org/abs/2605.29886
作者: Wenhan Xiao,Ziwei Zhang,Chuanyue Yu,Xingcheng Fu,Qingyun Sun,Runhua Xu,Jianxin Li
机构: Nankai University (南开大学); Beihang University (北京航空航天大学); Guangxi Normal University (广西师范大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 17 pages,13 figures
Abstract:Retrieval-augmented generation (RAG) improves knowledge-intensive question answering by incorporating external evidence. However, existing RAG methods still suffer from hallucinations and subtle reasoning errors. Recent studies introduce external critics to refine RAG outputs, yet they often provide coarse-grained and weakly structured feedback, exhibit over-aggressive intervention, and lead to noisy and unreliable refinement, limiting their effectiveness for correction. To tackle these issues, we propose CRITIC-R1, a structured critic framework that formulates and learns RAG critique as an explicit error diagnosis problem using reinforcement learning (RL). Our framework categorizes common RAG errors into multiple diagnostic dimensions, including verdict, error location, reasoning analysis, and fix generation. To learn these capabilities, we design two reward functions: Conservative Judgement Alignment (CJA) first encourages calibrated high-level judgements while mitigating the over-aggressive phenomenon, whereas Diagnostic Quality Alignment (DQA) further improves fine-grained diagnostic feedback through gated rewards. We train the critic model using GRPO-based RL with process-level supervision collected from external LLM teacher models. Experiments across five QA benchmarks show that CRITIC-R1 consistently improves answer quality over strong RAG baselines. Our source code is available at this https URL Comments: 17 pages,13 figures Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI) Cite as: arXiv:2605.29886 [cs.CL] (or arXiv:2605.29886v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2605.29886 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-50] owards Verifiable Multimodal Deep Research: A Multi-Agent Harness for Interleaved Report Generation
【速读】： 该论文试图解决多模态深度研究（multimodal deep research）中生成可靠、视觉信息丰富且面向人类的长篇报告的问题，尤其针对开放式合成任务缺乏确定性真实标签以及文本论证与视觉证据需交错整合的挑战。解决方案的关键在于提出一个名为 \textscPtah 的多智能体框架，其通过规划、研究和写作三个阶段协同工作：专用代理构建视觉感知的计划，收集基于主张的证据，并在“视觉工作记忆”（Visual Working Memory）中维护与来源对齐的图像；同时，通过声明式多模态工具使用实现报告生成；此外，验证代理作为接受函数，在整个流程中强制执行事实依据、引用一致性及跨模态一致性。实验表明，\textscPtah 在深度研究基准上优于强基线模型，生成的报告更具可靠性、视觉信息量和可用性。
链接: https://arxiv.org/abs/2605.29861
作者: Chenghao Zhang,Guanting Dong,Yufan Liu,Tong Zhao,Zhicheng Dou
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)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models (LLMs) have advanced autonomous agents from deep search, which retrieves concise factual answers, to deep research, which synthesizes scattered evidence into long-form reports. However, verifiable multimodal deep research remains challenging due to open-ended synthesis without deterministic ground truth and the need to interleave textual arguments with visual evidence. We propose \textscPtah, a multi-agent harness for interleaved report generation. \textscPtah orchestrates the lifecycle from user query to rendered web report through planning, research, and writing stages, where specialized agents construct visual-aware plans, collect claim-grounded evidence, maintain source-aligned images in a \textitVisual Working Memory, and compose reports through declarative multimodal tool use. A verifier agent serves as the harness’s acceptance function, enforcing factual grounding, citation fidelity, and cross-modal consistency throughout the workflow. We further introduce \textscPtahEval, an evaluation protocol that augments existing benchmarks with image-level and presentation-level assessments. Experiments on deep research benchmarks show that \textscPtah produces more reliable, visually informative, and usable human-facing multimodal reports than strong baselines.
[NLP-51] EvoRubric: Self-Evolving Rubric-Driven RL for Open-Ended Generation
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）在开放式生成任务中难以对齐的问题，核心挑战在于缺乏明确的奖励信号。现有基于评分标准（rubric-based）的强化学习方法依赖静态的人工标注标准或昂贵的外部专用模型来提供动态反馈，导致策略滞后或成本高昂。其解决方案的关键在于提出EvoRubric——一种单策略协同进化强化学习框架，通过统一响应生成与评分标准生成为单一参数化策略，使模型在“推理者”（Reasoner）和“评分生成器”（Rubric Generator）角色间动态交替，并引入多级验证机制（包括元验证器、零方差剪枝和留一法同行共识）以防止奖励劫持并确保信号可靠性。同时，经过验证的标准被动态存入记忆池，形成密集的多目标奖励信号，持续优化两个角色。实验表明，EvoRubric在医学、写作和科学等领域显著优于传统静态和外部LLM驱动的方法，且兼容人类专家先验知识，在初始使用专家标注标准时还能发现新的判别性维度，实现更优性能。
链接: https://arxiv.org/abs/2605.29847
作者: Xin Guan,Xiaomeng Hu,Shen Huang,Zhenyi Wang,Bo Zhang,Zijian Li,Pengjun Xie,Bo Liu,Jiuxin Cao
机构: Tongyi Lab, Alibaba Group
类目: Computation and Language (cs.CL)
备注:
Abstract:Reinforcement Learning (RL) has significantly advanced Large Language Models (LLMs) in verifiable domains, but aligning models for open-ended generation remains profoundly challenging due to the lack of definitive rewards. Current rubric-based RL methods mitigate this by employing explicit criteria; however, they rely heavily on static, human-annotated rubrics that inevitably cause policy lag, or expensive external proprietary models for dynamic updates. In this paper, we propose EvoRubric, a novel single-policy co-evolutionary RL framework that eliminates the reliance on static criteria and on external rubric generators. By unifying response generation and rubric generation under a single parameterized policy, EvoRubric dynamically alternates between a Reasoner and a Rubric Generator. To prevent reward hacking and ensure the reliability of generated signals, we introduce a multi-level verification pipeline featuring a meta-verifier, zero-variance pruning, and a Leave-One-Out peer consensus mechanism. Validated criteria are dynamically archived into a memory pool, yielding dense, multi-objective rewards to continuously co-optimize both roles. Extensive experiments across Medical, Writing, and Science domains demonstrate that EvoRubric consistently outperforms traditional static and external-LLM-driven alignment methods. Notably, our framework is compatible with human-expert priors. When initialized with expert-annotated rubrics, EvoRubric can further uncover novel, discriminative dimensions, achieving better performance than relying solely on static expert annotations.
[NLP-52] owards Localized and Disentangled Knowledge Editing for Multimodal Large Language Models
【速读】： 该论文试图解决多模态大语言模型（Multimodal Large Language Models, MLLMs）在进行多模态知识编辑（Multimodal Knowledge Editing, MKE）时存在的两个关键问题：一是编辑结果无法泛化到逻辑相关查询，二是会无意中修改与目标无关但存在视觉或语义关联的信息。其解决方案的关键在于提出一种名为“局部化与解耦知识编辑”（Localized and Disentangled Knowledge Editing, LDKE）的新框架，通过两个核心机制实现精准且可推广的编辑：一是引入快速定位模块（Fast Localization module），识别并高效更新与特定事实相关的模型层；二是设计解耦分类器（Disentanglement Classifier），将输入信息按相关性路由，从而保护无关知识不被干扰。该方法有效缓解了因果错位（Causal Misalignment）和特征纠缠（Feature Entanglement）两种失效模式，实验证明LDKE在多个基准测试和MLLMs上均能显著提升编辑传播能力并保持高局部性。
链接: https://arxiv.org/abs/2605.29826
作者: Leijiang Gu,Zhen Zeng,Feng Li,Xinjian Gao,Zenglin Shi
机构: Hefei University of Technology (合肥工业大学); Tongji University (同济大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Existing methods in Multimodal Knowledge Editing (MKE) have advanced the ability to correct outdated or inaccurate knowledge in Multimodal Large Language Models (MLLMs). However, they exhibit a critical limitation: while effectively modifying target factual pairs, they fail to generalize edits to logically related queries and often cause unintended alterations to unrelated but visually or semantically linked information. We identify and formalize two underlying failure modes causing this issue: Causal Misalignment, which confines edits to the specific sample, and Feature Entanglement, which causes unintended alterations to coupled but irrelevant information. To address these issues, we propose Localized and Disentangled Knowledge Editing (LDKE), a new framework that achieves precise and generalized editing by localizing fact-specific model layers and disentangling target-relevant inputs from irrelevant ones. Our approach introduces a Fast Localization module to identify and update critical layers efficiently, along with a Disentanglement Classifier that routes inputs appropriately to preserve unrelated knowledge. Extensive experiments across various benchmarks and MLLMs demonstrate that LDKE achieves superior performance in propagating edits to related contexts while maintaining high locality.
[NLP-53] PRAIB: Peer Review AI Benchmark of Behaviour of LLM -Assisted Reviewing
【速读】： 该论文试图解决的问题是：大型语言模型（LLM）在学术同行评审中的表现是否与人类审稿人一致，抑或仅生成看似专业的评论文本。其解决方案的关键在于提出了Peer Review AI Benchmark (PRAIB)，这是一个包含明确指标的框架，用于衡量审稿的特异性（specificity）、风格（style）及参与行为（behavior of engagement）。通过分析11,000条由五种商用和开源模型对ICLR与NeurIPS会议论文生成的审稿意见，并与人类审稿反馈进行对比，研究发现LLM生成的审稿存在显著偏差：评分变异度低、正向偏倚、过度自信，且交叉引用模式具有模型依赖性，不同于人类审稿习惯；尽管LLM生成的审稿更长、结构更复杂，却常忽略人类审稿人指出的细粒度问题。PRAIB因此成为诊断LLM在哪些审稿环节可可靠支持、哪些仍需改进的重要工具。
链接: https://arxiv.org/abs/2605.29815
作者: Krzysztof Żurawicki,Julia Farganus,Arkadiusz Gaweł,Mateusz Bystroński,Tomasz Jan Kajdanowicz
机构: Wrocław University of Science and Technology (弗罗茨瓦夫科技大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:The growing number of submitted papers has motivated the exploration of Large Language Models (LLMs) as a means to support and augment the peer review process, particularly in terms of improving its speed and scalability. Yet, it remains unknown whether LLMs engage with scientific manuscripts in the same manner as human reviewers, or whether they merely produce review-looking text. To address this, we introduce the Peer Review AI Benchmark (PRAIB), a novel framework comprising thoroughly defined metrics that measure review specificity, style, and behavior of engagement. To complement the PRAIB framework, we conduct a large-scale empirical study leveraging a dataset of 11,000 reviews generated by five proprietary and open-source models for 1,000 ICLR and NeurIPS papers. Spanning the 2021–2025 period, these machine-generated reviews are compared against original human feedback across diverse prompting strategies to identify systematic behavioral divergences. Our analysis reveals that the generated reviews diverge significantly from feedback provided by human reviewers: LLM ratings are less variable, positively biased, and overconfident, and their cross-reference patterns are model-dependent and distinct from human norms. Furthermore, when evaluated through PRAIB, we observe that LLMs tend to generate longer, more complex reviews, yet frequently overlook the atomic weaknesses noted by human reviewers. By characterizing where and how LLMs reviewing behavior departs from human norms, PRAIB provides the community with a diagnostic tool for identifying which aspects of the review process LLMs can reliably support today and which require further development before deployment.
[NLP-54] Data filtering methods for training language models
【速读】： 该论文旨在解决机器学习模型训练数据中标签错误（label errors）导致的性能下降问题，尤其关注在俄语文本分类任务中自动检测并过滤标签错误的有效性。其解决方案的关键在于对比两种自动化标签错误检测方法——置信学习（Confident Learning）与数据地图学（Dataset Cartography）在不同规模、类别数和领域分布的俄语语料库上的表现，包括ru_emotion_e-culture、RuCoLA和TERRa三个数据集。研究发现：在噪声较低的大规模数据集上，过滤标签错误对模型性能无显著提升；而在小规模高噪声数据集上，置信学习能显著提高F1-macro指标；而数据地图学则更为保守，删除样本更少。此外，两种方法针对性地移除错误标签样本的效果均优于随机删除，验证了其有效性。
链接: https://arxiv.org/abs/2605.29807
作者: Egor Shevchenko,Elena Bruches
机构: Novosibirsk State University (新西伯利亚国立大学); A. P. Ershov Institute of Informatics Systems SB RAS (A. P. Ershov信息系统研究所，西伯利亚分院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: AINL-2026
Abstract:Data quality is a critical factor in the effectiveness of machine learning models. Label errors, present even in widely used benchmarks, introduce noise into training data and reduce model generalization. In this work, we conduct a comparative analysis of two automatic label error detection methods - Confident Learning and Dataset Cartography - on three Russian text classification corpora of varying size, number of classes, and domain: ru_emotion_e-culture (49,123 examples, emotion classification), RuCoLA (8,524 examples, linguistic acceptability), and TERRa (2,337 examples, textual entailment recognition). We use the pre-trained rubert-base-cased model fine-tuned on each corpus. To verify the meaningfulness of filtering, we conduct control experiments with random removal of an equivalent number of examples. Results show that the effectiveness of both methods depends strongly on dataset characteristics: on large corpora with low noise levels, filtering does not improve performance, while on small datasets with high noise, Confident Learning achieves a significant F1-macro improvement. Dataset Cartography demonstrates more conservative behavior, removing fewer examples. Across all corpora, targeted removal by both methods outperforms random removal, confirming the meaningfulness of the approaches.
[NLP-55] Agent DoG 1.5: A Lightweight and Scalable Alignment Framework for AI Agent Safety and Security
【速读】： 该论文旨在解决现代开放世界智能体（如OpenClaw）在跨环境执行中引入的新型安全风险问题，以及前沿AI模型降低攻击门槛后导致现有对齐框架无法满足真实部署需求的挑战。其解决方案的关键在于提出一个轻量且可扩展的智能体安全对齐框架：首先更新智能体安全分类体系以涵盖Codex和OpenClaw场景下的新兴风险；其次构建基于分类引导的数据引擎，并采用影响函数净化技术，在仅使用约1000个样本的情况下训练出多个参数规模（0.8B–8B）的AgentDoG 1.5变体，性能媲美领先闭源模型（如GPT-5.4）；进一步地，基于AgentDoG 1.5建立高效代理安全SFT与强化学习（Reinforcement Learning, RL）训练环境，将Docker级部署开销降低两个数量级；最终部署AgentDoG 1.5作为无需训练的在线护栏（guardrail），实现实时安全调控。实验表明，AgentDoG 1.5在多样复杂交互式代理场景中达到最先进性能。
链接: https://arxiv.org/abs/2605.29801
作者: Dongrui Liu,Yu Li,Zhonghao Yang,Peng Wang,Guanxu Chen,Yuejin Xie,Qinghua Mao,Wanying Qu,Yanxu Zhu,Tianyi Zhou,Leitao Yuan,Zhijie Zheng,Qihao Lin,Yimin Wang,Haoyu Luo,Shuai Shao,Chen Qian,Qingyu Liu,Ling Tang,Ruiyang Qin,Qihan Ren,Junxiao Yang,Kun Wang,Zhiheng Xi,Linfeng Zhang,Ranjie Duan,Bo Zhang,Wenjie Wang,Wen Shen,Qiaosheng Zhang,Yan Teng,Chaochao Lu,Rui Mei,Man Li,Jialing Tao,Xi Lin,Tianhang Zheng,Yong Liu,Quanshi Zhang,Lei Zhu,Xingjun Ma,Junhua Liu,Hui Xue,Xiaoxiang Zuo,Xiangnan He,Chao Shen,Xianglong Liu,Minlie Huang,Jing Shao,Xia Hu
机构: Shanghai Artificial Intelligence Laboratory (上海人工智能实验室)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 44 pages, 12 Figures, 9 Tables
Abstract:Modern open-world agents such as OpenClaw exhibit powerful cross-environment execution capabilities yet introduce broad new safety risk sources. Meanwhile, advanced frontier AI models drastically lower attack barriers, rendering current agent alignment frameworks inadequate for real-world deployment. To tackle these emerging threats, we propose a lightweight and scalable agent safety alignment framework. Specifically, we update the agent safety taxonomy to accommodate emergent risks from Codex and OpenClaw execution scenarios. We further build a taxonomy-guided data engine with influence-function purification to train lightweight AgentDoG 1.5 variants (0.8B, 2B, 4B, and 8B parameters) using only around 1k samples, achieving comparable performance with leading closed-source models (e.g., GPT-5.4). Based on AgentDoG 1.5, we construct a highly efficient agentic safety SFT and RL training environment, which reduces deployment overhead in Docker-level environments by two orders of magnitude. Finally, we deploy AgentDoG 1.5 as a training-free online guardrail for real-time safety moderation. Extensive experimental results indicate that AgentDoG 1.5 achieves state-of-the-art performance in diverse and complex interactive agentic scenarios. All models and datasets are openly released.
[NLP-56] Nine Judges Two Effective Votes: Correlated Errors Undermine LLM Evaluation Panels
【速读】： 该论文试图解决的问题是：当前基于大语言模型（LLM）作为评判者（LLM-as-a-judge）的评测面板，其可靠性是否真正得益于多个模型的多样性，还是由于模型间存在高度相关性导致实际信息增益远低于预期。解决方案的关键在于提出一个框架，通过量化有效样本量（Kish有效样本大小，n_eff）和Condorcet零假设模型，评估面板中模型之间的独立性程度及其对最终评价准确性的实际贡献。研究发现，尽管使用了9个来自7个不同模型家族的前沿LLM，但它们提供的有效独立投票仅相当于约2个独立判官，约四分之三的名义独立性因模型在相同样本上犯下相同错误而丧失；这导致面板的实际准确率比理想独立投票情形低8–22个百分点，且最佳单一模型的表现优于或等于整个面板。该结果表明，模型间的相关性才是性能瓶颈，而非聚合算法本身，因此单纯扩大面板规模无法替代真正的独立评估。
链接: https://arxiv.org/abs/2605.29800
作者: Guneet Kohli
机构: Apple
类目: Computation and Language (cs.CL)
备注: 14 pages, 5 figures, 12 tables
Abstract:LLM-as-a-judge panels aggregate votes from multiple models, with the expectation that diverse models yield more reliable evaluations. We develop a framework to measure the true informational value of such panels and quantify how far their reliability falls short of the independent-voting ideal. Testing a panel of 9 frontier LLMs from 7 model families on three natural language inference datasets (each with 100 human annotations per item), we find that the 9 judges effectively provide only about 2 independent votes’ worth of information. Roughly three-quarters of the panel’s nominal independence is lost because the models make the same mistakes on the same items. The consequences are stark: the panel’s actual accuracy falls 8-22 percentage points short of what independent voting would achieve, and the best single judge matches or outperforms the full panel across all conditions. Neither adding more judges nor using smarter aggregation algorithms helps – established methods close at most 11% of this gap, even with access to the correct answers. We quantify these findings using the Kish effective sample size (n_eff) and a Condorcet null model, and show the deficit is robust across prompt variants, temperatures, chain-of-thought reasoning, and a pairwise preference task (RewardBench). The bottleneck is correlated judges, not the aggregation algorithm, implying that scaling up panels cannot substitute for genuinely independent evaluation.
[NLP-57] Metric-Dependent Annotation Saturation for Learning from Label Distributions
【速读】： 该论文试图解决的问题是：在多标注者场景下，如何根据不同的评估指标合理分配标注预算（即所需标注者数量），以有效捕捉标注分歧所蕴含的信息。解决方案的关键在于发现不同评估指标对标注者数量的敏感性存在显著差异——具体而言，在三类自然语言推理（NLI）任务中，衡量模型是否识别出引发分歧样本的“熵相关性”指标需要约20–50名标注者才能收敛，而衡量标签分布匹配度的“分布匹配度”（KL散度）则在约10名标注者时即达到饱和（实现87–95%的改进）。此外，研究还揭示了软标签（soft labels）相较于标签平滑（label smoothing）的优势：软标签能保留每个样本特有的不确定性信号，而标签平滑无法区分模糊样本与清晰样本，导致其性能上限被限制在r ≈ 0.45–0.49，远低于软标签的r = 0.643（p < 0.001）。这一结论在多个模型架构（DeBERTa、RoBERTa）和跨领域任务（内容安全）中均得到验证，表明标注预算应依据目标评估指标动态调整，而非采用统一标准。
链接: https://arxiv.org/abs/2605.29797
作者: Guneet Kohli
机构: Apple
类目: Computation and Language (cs.CL)
备注: 16 pages, 3 figures, 14 tables
Abstract:When annotators disagree on a label, the disagreement itself carries signal – and the number of annotators needed to capture it depends on the evaluation metric. We fine-tune NLI models on label distributions subsampled from ChaosNLI, a dataset providing 100 independent annotator judgments per item, and identify metric-dependent saturation. In our 3-class NLI setting, entropy correlation – whether the model identifies which items elicit disagreement – requires N ~ 20-50 annotators to converge, while distributional match (KL divergence) saturates by N ~ 10 (87-95% of improvement across five model seeds). This finding rests on a prior observation: soft labels carry item-specific signal that label smoothing cannot replicate. Across five smoothing intensities, entropy correlation clusters at r ~ 0.45-0.49, while soft labels reach r = 0.643 (p 0.001); per-item analysis traces this gap to smoothing’s inability to distinguish ambiguous items from clear ones. The soft-label advantage replicates across two architectures (DeBERTa, RoBERTa), a non-NLI-pretrained baseline, and an exploratory cross-domain evaluation on content safety. These results suggest that annotation budgets should be informed by the target evaluation metric rather than set uniformly.
[NLP-58] SAAS: Self-Aware Reinforcement Learning for Over-Search Mitigation in Agent ic Search
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在执行多跳问答任务时因缺乏自我意识而导致的“过度搜索”问题。具体而言，现有基于代理的搜索系统无法准确识别自身知识边界，常在已有内部知识足以回答问题时仍触发不必要的外部搜索，或在已收集足够证据后仍持续搜索，从而造成显著的推理延迟和计算开销。解决方案的关键在于提出SAAS框架——一种用于培养动态自我意识的强化学习（Reinforcement Learning, RL）方法，其核心创新包括：(i) 搜索边界建模机制，通过对比禁用与启用搜索的轨迹来识别策略演化过程中的搜索边界；(ii) 边界感知奖励模块，将边界意识转化为轨迹级惩罚，抑制冗余搜索行为；(iii) 分阶段优化策略，采用顺序式课程学习优先强化推理能力，防止奖励劫持（reward hacking）。实验表明，SAAS在保持准确率的同时显著减少过度搜索行为。
链接: https://arxiv.org/abs/2605.29796
作者: Yunbo Tang,Chengyi Yang,Shiyu Liu,Zhishang Xiang,Zerui Chen,Qinggang Zhang,Jinsong Su
机构: Xiamen University (厦门大学); Jilin University (吉林大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Agentic search enables LLMs to solve complex multi-hop questions through iterative reasoning and external search. Despite the effectiveness, these systems often suffer from a critical limitation in practice: agents fail to recognize their own knowledge boundaries, blindly triggering searches when internal knowledge suffices and failing to terminate search even when adequate evidence has been collected. The lack of self-awareness leads to severe \textbfover-search, incurring substantial inference latency and prohibitive computational cost. To this end, we propose SAAS, a novel RL framework designed to cultivate dynamic self-awareness that precisely regulates search behavior without compromising accuracy. SAAS introduces three key components: (i) a search boundary modeling mechanism, which identifies the search boundary under the evolving policy by contrasting search-disabled and search-enabled rollouts; (ii) a boundary-aware reward module, which translates this boundary awareness into trajectory-level penalties, suppressing unnecessary and redundant searches; and (iii) a stage-wise optimization strategy, which leverages a sequential curriculum to prioritize reasoning over search regularization, thereby avoiding reward hacking. Extensive experiments demonstrate that SAAS substantially reduces over-search, while maintaining accuracy. Our code is anonymously released at this https URL.
[NLP-59] ActTraitBench: Quantifying the Knowledge-Decision Gap in Large Language Models via Human-Grounded Behavioral Validation
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）在显性自我报告与隐性行为决策之间存在的“知识-决策鸿沟”（Knowledge-Decision Gap, GKD）问题，即模型虽能一致地模拟人格特征的自我陈述，但在实际行为选择中却表现出显著偏离。现有评估基准因构念效度不足、多维特征纠缠及分布偏差等问题难以准确测量这种不对称性。解决方案的关键在于提出ActTraitBench——一个基于人类实证数据的人格一致性评估框架：它通过心理测量学维度与行为范式的严格一对一映射，并采用分位数映射（Quantile Mapping）进行分布校准，使LLM评分分布与人类基准对齐；同时引入链式认知对齐（Chain of Cognitive Alignment, CoCA）作为即插即用的推理时干预策略，在前沿模型中提升推理能力驱动的行为一致性，同时揭示小模型在该任务中的能力局限。
链接: https://arxiv.org/abs/2605.29791
作者: Yutong Yang,Chenxi Miao,Weikang Li,Yunfang Wu
机构: Peking University (北京大学); Baidu Inc (百度公司)
类目: Computation and Language (cs.CL)
备注:
Abstract:While Large Language Models (LLMs) can convincingly simulate personas in explicit self-reports, they often deviate in implicit behavioral decisions, revealing a substantial Knowledge-Decision Gap ( G_\textKD ). Existing benchmarks struggle to measure this asymmetry due to limited construct validity, multi-dimensional entanglement, and distributional biases in LLM-based evaluation. To address these issues, we propose ActTraitBench, a human-grounded evaluation framework for measuring personality consistency in LLMs. Grounded in empirical human data, ActTraitBench establishes one-to-one mappings between psychometric facets and behavioral paradigms, and applies a Distributional Calibration via Quantile Mapping procedure to align LLM-judge score distributions with human norms. Experiments on 14 mainstream LLMs reveal a pervasive knowledge-decision asymmetry, where larger and more capable models often exhibit stronger behavioral divergence despite highly consistent self-reports. To mitigate this gap, we further introduce the Chain of Cognitive Alignment (CoCA), a plug-and-play inference-time intervention that improves alignment in reasoning-capable frontier models while exposing clear capability limitations in smaller architectures.
[NLP-60] Hista and Numca: Estimate State Value Effectively for LLM Reinforcement Learning ICML2026
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）在强化学习（Reinforcement Learning, RL）后训练过程中状态价值估计（State Value Estimation）不准确的问题。当前主流方法如近端策略优化（PPO）中的评论家（critic）模块往往退化为粗粒度的群体平均基线，导致训练不稳定且性能受限。解决方案的关键在于提出两种新方法：Numca 利用数值跨度作为可度量的里程碑来提升状态价值估计的精度；Hista 则基于 LLM 的隐藏状态构建表示，并通过加权平均不相交轨迹及其回报来改进估计。实验表明，这两种方法均能显著提升状态价值估计准确性，并在不同 RL 算法和模型规模下增强训练效果，且计算开销可控。
链接: https://arxiv.org/abs/2605.29782
作者: Zizhe Chen,Jiqian Dong,Yizhou Tian,Garry Yang,Yongqiang Chen,Zhitang Chen,James Cheng
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted at ICML 2026
Abstract:Reinforcement learning (RL) refines large language models (LLMs) by directly optimizing model behavior through reward signals. While accurate state value estimation is critical for stable training in classical RL, it remains an underexplored challenge in LLM post-training. In this work, we introduce the State Value Estimation Benchmark (SVEB) to assess state estimation within existing RL frameworks and show that critics in standard approaches like PPO collapse to a coarse group-average baseline. To address this, we propose two techniques: Numca, which leverages numerical spans as gradable milestones for state value estimation, and Hista, a framework that uses LLM’s hidden states as representation to weighted average disjoint rollouts and their return. Extensive experiments demonstrate that both methods yield more accurate state value estimates and enhance training performance across different RL algorithms and model sizes without incurring significant computational overhead.
[NLP-61] DySem: Uncovering Dynamic Semantic Components via Multilingual Consensus for Calculating Semantic Textual Similarity
【速读】： 该论文试图解决当前基于大语言模型（LLM）的文本语义相似度计算方法中存在的两个核心问题：一是最后一层隐藏状态编码了过于泛化的知识，而非专精于语义信息，导致语义相似度计算效果欠佳；二是LLM隐藏层维度通常极高，引入冗余和噪声，不利于高效表示语义。解决方案的关键在于提出一种无需训练的框架DySem，通过多语言一致性挖掘LLM内部更具语义相关性的组件，并摒弃静态表示空间，转而构建依赖于具体文本对的联合语义集合（text-dependent joint semantic set），在共享的动态语义维度子集上计算相似度，从而实现更精准且低维的语义相似度建模。
链接: https://arxiv.org/abs/2605.29751
作者: Kaijie Zheng,Weiqin Wang,Yile Wang,Hui Huang
机构: Shenzhen University (深圳大学)
类目: Computation and Language (cs.CL)
备注: 18 pages, 23 figures, 5 tables
Abstract:Calculating semantic textual similarity is a foundational task in natural language processing. Current large language models (LLMs) based methods typically rely on extracting last-layer hidden states with fixed dimensions to compute similarity for every text pairs. We argue that this paradigm is suffer from two limitations: (i) The last hidden layer encodes more general knowledge rather than just semantic knowledge, making it suboptimal for semantic similarity computation; (ii) The hidden layer dimensions of LLMs are generally very large, which introduces some redundancy and noise for representing semantics. In this work, we propose DySem, a novel training-free framework that investigates more semantic-related internal components of LLMs via multilingual consensus, and shifts away from static representation spaces in favor of dynamic, sample-specific semantic dimensions by constructing text-dependent joint semantic set and computes similarity over this shared dimensional subset. Extensive experiments across various LLMs show that our method consistently outperforms recent baselines while maintaining lower dimensions for similarity calculation. The code is released at this https URL.
[NLP-62] AfriScience-MT: Towards Decolonizing Science in Africa through Text Translation
【速读】： 该论文试图解决非洲本土语言在科学知识传播中面临的术语缺失问题，从而打破殖民语言（如英语、法语）对非洲教育和科研交流的垄断。其关键解决方案是构建了AfriScience-MT这一多语言平行语料库，涵盖六种非洲语言（阿姆哈拉语、豪萨语、卢干达语、北索托语、约鲁巴语和祖鲁语）与11个科学领域的专业翻译文本，并由专业译者与科学传播专家合作创建新术语。该语料库支持机器翻译模型在零样本、少样本和微调设置下的基准测试，结果显示闭源模型（如GPT-5.4和Gemini-3.1-Flash-Lite）在句子级和文档级翻译质量上均显著优于开源模型，为非洲语言科学文献的自动化翻译提供了重要基础设施。
链接: https://arxiv.org/abs/2605.29741
作者: Idris Abdulmumin,Tajuddeen Gwadabe,Shamsuddeen Hassan Muhammad,David Ifeoluwa Adelani,Nomonde Khalo,Ibrahim Said Ahmad,Abiodun Modupe,Anina Mumm,Sibusiso Biyela,Michelle Rabie,Johanna Havemann,Marek Rei,Jade Abbott,Vukosi Marivate
机构: Data Science for Social Impact, University of Pretoria; Masakhane Research Foundation; Imperial College London; Mila, McGill University; Canada CIFAR AI Chair; University of Cape Town; University of Wisconsin - Stevens Point; Independent Consultant; University of South Africa; Independent Researcher; Access 2 Perspectives; Lelapa AI
类目: Computation and Language (cs.CL)
备注:
Abstract:The dominance of colonial languages in African education and scientific communication limits how hundreds of millions of speakers of African languages access and produce scientific knowledge. A core obstacle is the lack of established scientific terminology in these languages. We introduce AfriScience-MT, a parallel corpus covering six African languages (Amharic, Hausa, Luganda, Northern Sotho, Yorùbá, and isiZulu) across 11 scientific domains. Professional translators, working with expert science communicators, translated plain-language summaries of scientific papers into each target language and created new terms where none existed. We benchmark machine translation systems and large language models in zero-shot, few-shot, and fine-tuned settings. Our results show that closed-source models outperform all open-source models at both the sentence and document levels: GPT-5.4 and Gemini-3.1-Flash-Lite lead with average sentence-level COMET scores of 68.3 and 68.0, respectively, and tie at an average document-level COMET of 48.3. Among open systems, fine-tuned NLLB-1.3B reaches 67.3 at the sentence level, and TranslateGemma-12B reaches 44.0 at the document level with 1-shot in-context learning. We release AfriScience-MT to support benchmarking and document-level scientific MT for African languages.
[NLP-63] Multi-Legal-Bench: Evaluating LLM s on Legal Reasoning Across Jurisdictions Languages and Legal Traditions
【速读】： 该论文试图解决法律自然语言处理（Legal NLP）领域中缺乏跨司法管辖区可比评估基准的问题，现有基准通常仅针对单一语言或混合差异显著的任务，导致无法进行有效的跨语言、跨司法管辖区比较。解决方案的关键在于构建首个跨司法管辖区的法律基准 Multi-Legal-Bench，它在六个国家（乌克兰、法国、荷兰、波兰、捷克共和国、立陶宛）中定义了五项结构化任务（法院类型分类、判决形式分类、案件结果预测、法律规范提取、原因类别预测），并基于各国法院登记数据映射为统一元数据格式，形成一个5×6的任务-司法管辖区稀疏矩阵（共20个有效单元）。通过在AWS Bedrock上对7个前沿大语言模型（LLM）进行零样本和三样本提示测试，并额外评估4个中小规模模型（3–12B参数），研究揭示了四项关键发现：任务依赖的少量示例效应在所有司法管辖区一致；无单一模型在任何语言中持续领先；跨语言迁移效果不遵循语言亲缘性（如乌克兰→法语优于乌克兰→波兰语），标签集对齐更能预测迁移质量；分词器“肥力”（tokenizer fertility）虽有2.3倍差异，但与跨语言准确率无关（r = -0.27, p = 0.14），表明模型架构和预训练数据比分词效率更重要。
链接: https://arxiv.org/abs/2605.29738
作者: Volodymyr Ovcharov
机构: SecondLayer
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 14 pages, 5 figures, 8 tables. Dataset: this https URL
Abstract:Legal NLP benchmarks overwhelmingly evaluate a single language or aggregate tasks that differ fundamentally across jurisdictions, making cross-lingual comparison impossible. We introduce Multi-Legal-Bench, the first cross-jurisdictional legal benchmark that evaluates identical tasks across six countries (Ukraine, France, Netherlands, Poland, Czech Republic, Lithuania), four language families, and 134 million court decisions. The benchmark defines five tasks court-type classification, judgment form classification, case-outcome prediction, legal norm extraction, and cause category prediction mapped to structured metadata from national court registries, forming a deliberately sparse 5x6 task-jurisdiction matrix (20 of 30 cells filled). We evaluate 7 frontier LLMs under zero-shot and 3-shot prompting via AWS Bedrock, with 4 additional small/medium models (3-12B) for scaling analysis. Our results reveal that: (1) task-dependent few-shot effects discovered in Ukrainian replicate across all jurisdictions; (2) no single model dominates any language rankings shift with both task and jurisdiction; (3) cross-lingual few-shot transfer does not follow language proximity: UA-FR (Romance, -2.1 pp) transfers better than UA-PL (Slavic, -13.7 pp), with label-set alignment predicting transfer quality better than language family; and (4) tokenizer fertility, despite a 2.3x spread, does not significantly predict cross-lingual accuracy (r=-0.27, p=0.14), suggesting that model architecture and pretraining data dominate tokenizer efficiency. We release all data, prompts, and model predictions.
[NLP-64] Minimal Prompt Perturbations Lead to Code Vulnerabilities: Prompt Frag ility and Hidden-State Signals in Coding LLM s
【速读】： 该论文试图解决的问题是：大型语言模型（Large Language Model, LLM）生成代码时的安全性脆弱性，特别是当输入提示（prompt）发生微小变化时，是否会导致生成的代码从安全状态变为存在漏洞的状态。此前研究仅关注提示扰动对代码功能正确性的影响，而未探讨其对代码安全性的潜在威胁。
解决方案的关键在于：通过在三个LLM和五种编程语言上施加基于token级别的提示变异（如单字符修改），系统性地评估生成代码的安全性变化，并利用对模型隐藏状态的探查揭示脆弱性来源。结果表明，即使是微小的提示变化也可能导致生成代码由安全转为易受攻击；进一步分析发现，输入处理类漏洞（如缺失验证或清理逻辑）比默认不安全选择类漏洞（如使用弱算法或不安全参数）更易被预测（平均AUC分别为0.753 vs. 0.674）。这说明针对不同类型的漏洞应采取差异化防御策略——输入处理漏洞可在生成前检测，而默认不安全漏洞需在解码过程中干预。
链接: https://arxiv.org/abs/2605.29737
作者: Alexander Sternfeld,Andrei Kucharavy,Ljiljana Dolamic
机构: HES-SO (高等教育与研究学院); armasuisse Science and Technology (武装技术科学公司)
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL); Software Engineering (cs.SE)
备注:
Abstract:LLM-based coding assistants are seeing rapid adoption, offering substantial gains in developer productivity. As organizations increasingly ship code these agents produce, the security of that code becomes critical. Prior work has shown that minor prompt perturbations degrade the functional correctness of LLM-generated code, but whether they also compromise code security has remained unstudied. We apply token-level mutations to prompts across three models and five programming languages, and show that mutations as small as a single-character change can flip generated code from secure to vulnerable. Probing the models’ hidden states reveals that this fragility is partially encoded in prompt representations, but unevenly so. Input-handling vulnerabilities, where the model omits validation or sanitization, are more predictable (mean AUC 0.753) than secure-defaults vulnerabilities, where insecure code stems from one local choice such as a weak algorithm or unsafe parameter (mean AUC 0.674). These results show that the threat model for LLM-assisted coding extends beyond prompt injection to ordinary prompt variation, and indicate that input-handling flaws can be caught before generation while secure-defaults flaws require intervention during decoding.
[NLP-65] HTAM: Hierarchical Transition-Attended Memory for Operator Optimization
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在生成高性能GPU算子时面临的优化难题，特别是如何在自动代码生成中实现硬件感知的高效搜索。现有方法存在粒度不匹配问题：粗粒度提示虽可复用但难以直接执行，而细粒度记忆虽具操作性却扩大了搜索空间并掩盖优化瓶颈。解决方案的关键在于构建一种分层结构化记忆机制——HTAM（Hierarchical Transition-Attended Memory），其核心是通过两级层次化转移图（Hierarchical Transition Graph, HTG）组织全局优化方向、局部策略及步骤间的转换经验，在每一步演化中自适应选择最优全局方向并调用对应局部策略，从而指导CUDA代码生成。实验表明，HTAM在KernelBench全集上显著优于LLM基线，在正确性、快速解率和加速比方面均有提升，且跨后端和Robust-KBench测试验证了其结构化记忆的迁移能力。
链接: https://arxiv.org/abs/2605.29734
作者: Yining Zhang,Mingyang Yi,Chen Wang,Xuwen Xiang,Tianhe Jia,Zedong Dan,Chengqing Zong,Yue Wang
机构: University of Chinese Academy of Sciences (中国科学院大学); Institute of Automation, Chinese Academy of Sciences (中国科学院自动化研究所); Zhongguancun Academy (中关村学院); Renmin University of China (中国人民大学)
类目: Computation and Language (cs.CL)
备注: 24 pages, 5 figures
Abstract:High-performance GPU kernels are essential for efficient LLM deployment, yet optimizing them remains expertise-intensive. Recent LLM-based code generation makes automatic GPU operator generation promising, but operator optimization remains a hardware-aware search problem. Existing LLM-based methods face a granularity mismatch: coarse hints are reusable but hard to execute, whereas detailed memories are actionable but enlarge the search space and obscure optimization bottlenecks. The key challenge is therefore to organize optimization experience at an appropriate granularity. To address this issue, this paper proposes HTAM (Hierarchical Transition-Attended Memory), a coarse-to-fine framework for LLM-based operator optimization. HTAM builds a two-level Hierarchical Transition Graph (HTG) to organize coarse global directions, detailed local strategies, and transition experience between optimization steps. During each evolution step, HTAM selects a global direction from the current state and recent optimization history, retrieves the corresponding local strategy memory, and uses it to guide concrete CUDA code generation. Experiments on the full KernelBench suite demonstrate that HTAM consistently improves correctness, fast-solution rate, and speedup over LLM-based baselines, while backend and Robust-KBench studies indicate transferable benefits from structured memory.
[NLP-66] User-Aware Active Knowledge Acquisition for Emotional Support Dialogue
【速读】： 该论文试图解决的问题是：在多轮对话中，如何有效获取并泛化与用户情绪支持相关的对话知识，尤其是在用户需求信号微弱、间接且需通过多轮交互才能明确的情况下，现有方法难以高效学习和适应这些隐含需求。解决方案的关键在于提出一种无梯度的主动对话学习框架——User-Aware Active Knowledge Acquisition (UKA)，其核心创新包括：1）显式建模对用户需求的不确定性；2）引入基于心智理论（Theory-of-Mind）的不确定性估计机制，以优先选择能激发更丰富用户反馈的响应策略；3）将主动学习机制同时应用于知识获取和响应生成过程，从而在训练阶段高效探索与用户对齐的对话知识，并在测试阶段保持鲁棒性。实验表明，该方法在多个对话基准和模型架构上均显著优于强基线，在对话质量和用户对齐方面表现更优。
链接: https://arxiv.org/abs/2605.29715
作者: Mufan Xu,Kehai Chen,Jiahao Hu,Xinchao Xu,Muyun Yang,Tiejun Zhao,Min Zhang
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Emotional support plays an important role in dialogue systems, and its success depends on adapting to a user’s evolving and implicit needs across multi-turn interactions while leveraging the strong reasoning capacity of large language models. However, since signals about user needs are often weak, indirect, and can only be disambiguated through multi-turn interaction, existing emotional support methods often struggle to acquire and generalize relevant conversational knowledge efficiently. To bridge this gap, we introduce User-Aware Active Knowledge Acquisition (UKA), a gradient-free active dialogue learning framework that explicitly represents uncertainty about user needs and incorporates active learning into both knowledge acquisition and response this http URL propose a Theory-of-Mind uncertainty estimation mechanism that allows the model to prioritize responses, thereby eliciting more informative user feedback. UKA is capable of efficiently exploring user-aligned conversational knowledge during training while maintaining robustness at test time. Experiments across multiple dialogue benchmarks and model architectures demonstrate that our approach consistently outperforms strong baselines in dialogue quality and user alignment.
[NLP-67] Leverag ing Routing Dynamics in Mixture-of-Experts Models for Efficient Language Adaptation
【速读】： 该论文试图解决的问题是：在持续预训练过程中，多语言混合专家（MoE）模型中专家路由行为如何随语言变化，以及如何实现高效且有效的低资源多语言适应。解决方案的关键在于发现并利用语言专业化主要出现在模型最后几层的规律，并提出一种参数高效的适配策略——仅更新最终MoE层中的语言特定和共享专家，从而在保持高性能的同时，仅更新少于2%的参数。这一方法在MultiBLiMP和Belebele数据集上验证了其在性能与效率之间的优越平衡，为多语言MoE模型的轻量化适配提供了理论依据和实践路径。
链接: https://arxiv.org/abs/2605.29714
作者: Aditi Khandelwal,Marius Mosbach,Verna Dankers,Siva Reddy,Golnoosh Farnadi
机构: Mila – Quebec AI Institute (Mila – 魁北克人工智能研究所); McGill University (麦吉尔大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Mixture-of-Experts (MoE) models are widely used to scale language models, yet their expert routing behavior and adaptation in a multilingual setting remain underexplored. In this work, we study multilingual routing dynamics during continual pre-training of an English-centric MoE model on a multilingual corpus, analyzing how expert usage varies across languages. We find that continual multilingual pre-training leads to diffused, language-agnostic routing in early and middle layers, with language specialization primarily emerging in the final layers. We also show that token-level vocabulary overlap between languages plays an important role in how languages are routed. Motivated by these findings, we propose a parameter-efficient adaptation strategy that updates language-specific and shared experts in the final MoE layers. Experiments on MultiBLiMP and Belebele show that our method achieves a strong performance-efficiency trade-off, attaining competitive performance relative to fine-tuning complete final layers, while updating less than 2% of the parameters. Overall, our findings provide insights into where and how language specialization emerges in MoEs during continual pre-training and provide practical insights for low-resource multilingual adaptation. Our code is available at this https URL.
[NLP-68] aching Language Models to Check Grounded Claim Factuality with Human Test-Taking Strategies ACL2026
【速读】： 该论文旨在解决大语言模型（LLM）在检索增强生成等应用中，如何高效且准确地判断生成内容的客观事实性（factuality）问题。现有方法要么依赖于数据集特定的阈值调优（如蕴涵分类器），要么采用直接提示（direct prompting）方式，未能充分发挥LLM的推理能力。其解决方案的关键在于将事实性核查建模为一个“真假阅读理解”任务，并通过引入显式的应试策略提示（test-taking strategies）来引导LLM进行结构化推理，从而显著降低token消耗（超过80%）并提升性能。此外，作者进一步训练小型语言模型（SLMs）替代LLM以降低推理成本，结合监督微调（SFT）与自修正机制，使SLMs在保持高准确性的同时生成可解释的支持性推理过程，实现了低开销与高可靠性的平衡，在两个事实性基准上达到先进水平。
链接: https://arxiv.org/abs/2605.29712
作者: Yuxuan Ye,Raul Santos-Rodriguez,Edwin Simpson
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: ACL 2026 Main
Abstract:Grounded claim factuality checking is important for large language model (LLM) applications such as retrieval-augmented generation, as it helps users assess the correctness of generated outputs. Existing metrics using entailment classifiers require dataset-specific threshold tuning, while LLM-based approaches often use direct prompting, which underutilises the reasoning capabilities of LLMs. We address this by formulating grounded claim factuality checking as a true/false reading comprehension task and prompting LLMs with explicit test-taking strategies for efficient reasoning. Our method reduces token usage by over 80% compared to unguided open-ended reasoning, and achieves competitive performance to more expensive alternatives across two factuality benchmarks, setting a new state of the art on one. To further reduce inference cost, we train small language models (SLMs) to replace LLMs in the checking pipeline. Using supervised fine-tuning (SFT) and a self-revision mechanism, the SLMs learn to improve their factuality judgements. Experimental results show that the resulting SLMs perform on par with strong baselines, combining low inference costs with generating supporting rationales to support interpretability. Code and datasets will be released upon acceptance.
[NLP-69] Personalized Turn-Level User Conversation Satisfaction Benchmark
【速读】： 该论文试图解决的问题是：如何在对话系统中实现对用户个性化满意度的精准评估，尤其是在单轮对话（turn-level）层面，传统自动评估方法因仅关注通用响应质量而难以捕捉用户个体差异导致的满意与否。解决方案的关键在于构建一个结合紧凑用户记忆（compact user memories）与当前目标轮次上下文（target-turn context）的对话满意度评估器，该评估器不仅能输出满意度分数，还能生成指向不满意原因的解释性理由（dissatisfaction-oriented rationales）。通过元评估（meta-evaluation）验证，该方法在排序一致性（ordinal agreement）和不满意度轮次检测上显著优于监督学习、基于检索以及通用大语言模型作为裁判（LLM-as-a-judge）等基线方法。此外，作者提出了PersTurnBench——一个用于个性化单轮满意度评估的基准测试集，利用已验证的评估器进行模型回放（replay），从而在固定重放状态的前提下，无需为每个候选模型重新收集人工标注即可公平比较通用生成模型与增强记忆的个性化系统。
链接: https://arxiv.org/abs/2605.29711
作者: Zhefan Wang,Zhiqiang Guo,Weizhi Ma,Min Zhang,Quanjia Yan,Hengliang Luo
机构: Tsinghua University (清华大学); Meituan (美团)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:User satisfaction with AI assistants is highly personalized: the same response may satisfy one user but disappoint another depending on what each user expects and what they have asked for before. Existing automatic evaluation methods mostly measure generic response quality, making it difficult to judge whether a response satisfies a user at a specific turn. We study this problem as personalized turn-level user conversation satisfaction evaluation. We build a conversation satisfaction evaluator that combines compact user memories with target-turn context to produce satisfaction scores and dissatisfaction-oriented rationales. Meta-evaluation against human satisfaction annotations shows that personalized memory and post-hoc score calibration improve ordinal agreement and dissatisfied-turn detection over supervised, retrieval-based, and generic LLM-as-a-judge baselines. We further introduce PersTurnBench, a personalized turn-level user conversation satisfaction benchmark that uses the verified evaluator to assess generation models via replay. By holding the replay state fixed, PersTurnBench enables controlled comparison of generic generation models and memory-augmented personalized systems without new human labels for every candidate model. The evaluator and benchmark let researchers compare candidate generation models on personalized satisfaction without collecting new user feedback for every model.
[NLP-70] Understanding Safety-Sensitive Expert Behavior in Mixture-of-Experts LLM s
【速读】： 该论文试图解决的问题是：在混合专家（Mixture-of-Experts, MoE）大语言模型中，安全对齐（safety alignment）与路由驱动的专家专业化之间的交互机制尚不明确，尤其是安全行为是否可通过特定专家的路由来控制。现有直觉认为有害请求可被引导至专门拒绝的专家以实现安全控制，但本文通过实证发现，对齐后的MoE模型中路由模式主要由主题驱动，而非安全意图；安全行为可在不改变模型内在路由路径的情况下被显著调整。解决方案的关键在于提出RASET（Router-Agnostic Safety-critical Expert Tuning）框架——该框架基于对比路由敏感性准则识别出少数对安全关键的专家，并仅对其实施参数高效微调，从而在保持原生路由行为不变的前提下最小化语义扰动，揭示了MoE架构下一种新的安全风险，强调需要引入面向专家的安全对齐机制。
链接: https://arxiv.org/abs/2605.29708
作者: Zhibo Zhang,Yuxi Li,Zhen Ouyang,Ling Shi,Kailong Wang
机构: Huazhong University of Science and Technology, Wuhan, China; Nanyang Technological University, Singapore
类目: Computation and Language (cs.CL)
备注: 11 pages, 4 figures
Abstract:Mixture-of-Experts (MoE) LLMs rely on sparse, router-driven expert activation, yet how safety alignment interacts with routed expert specialization remains underexplored. A common intuition is that safety behavior may be controlled by routing harmful requests to distinct refusal-oriented experts. In this work, we provide empirical evidence for a different picture: routing patterns in aligned MoE LLMs are largely topic-driven, while safety behavior can be altered with little change to the model’s intrinsic routing path. Motivated by this observation, we present RASET (Router-Agnostic Safety-critical Expert Tuning), a red-teaming framework that probes safety enforcement that is localized in a small subset of experts while preserving the model’s intrinsic routing behavior. RASET identifies safety-critical experts via a contrastive routing-sensitivity criterion and applies parameter-efficient tuning only to the selected experts, minimizing semantic disruption relative to router-steering interventions. These results reveal a distinct MoE safety risk, highlighting the need for expert-aware alignment mechanisms. Comments: 11 pages, 4 figures Subjects: Computation and Language (cs.CL) Cite as: arXiv:2605.29708 [cs.CL] (or arXiv:2605.29708v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2605.29708 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-71] Domino: Decoupling Causal Modeling from Autoregressive Drafting in Speculative Decoding
【速读】： 该论文试图解决的是生成式 AI（Generative AI）中大语言模型（LLM）推理速度与草案质量之间的权衡问题：传统推测解码方法在使用自回归（autoregressive）方式生成草案时虽能建模 token 间的因果依赖，但存在串行计算开销；而并行草案生成虽降低 drafting 成本，却削弱了块内 token 的依赖建模能力。解决方案的关键在于提出 Domino 框架，其核心创新是将因果依赖建模与昂贵的自回归执行过程解耦：首先利用并行草案主干（parallel draft backbone）快速生成整块草案分布，再通过轻量级 Domino head 引入前缀相关的因果信息进行精修；同时设计基于基础锚定（base-anchored）的训练课程，逐步优化从并行主干到最终因果校正分布的策略，从而实现高效且高质量的草案生成。实验表明，Domino 在 Qwen3 模型上实现了高达 5.49× 的端到端加速和 5.8× 的吞吐量提升。
链接: https://arxiv.org/abs/2605.29707
作者: Jianuo Huang,Yaojie Zhang,Qituan Zhang,Hao Lin,Hanlin Xu,Linfeng Zhang
机构: Shanghai Jiao Tong University (上海交通大学); HUST (华中科技大学); UESTC (电子科技大学); Fudan University (复旦大学); Huawei (华为)
类目: Computation and Language (cs.CL)
备注:
Abstract:Speculative decoding accelerates LLM inference by drafting multiple tokens and verifying them in parallel with the target model. However, its practical speedup is constrained by the trade-off between draft quality and drafting cost: autoregressive drafters model causal dependencies among draft tokens but incur sequential overhead, while parallel drafters reduce drafting cost but weaken intra-block dependency modeling. In this paper, we propose Domino, a speculative decoding framework that decouples causal dependency modeling from expensive autoregressive draft execution. Domino first uses a parallel draft backbone to produce preliminary draft distributions for the entire block, and then applies a lightweight Domino head to refine them with prefix-dependent causal information. To stabilize teacher-forced causal encoding, we further introduce a base-anchored training curriculum that first strengthens the parallel backbone and then gradually shifts optimization toward the causally corrected final distribution. Experiments on Qwen3 models show that Domino achieves up to (5.49\times) end-to-end speedup under the Transformers backend and up to (5.8\times) throughput speedup under SGLang serving.
[NLP-72] Scaling Laws for Agent Harnesses via Effective Feedback Compute
【速读】： 该论文试图解决的问题是：当前语言模型系统在测试时扩展（test-time scaling）分析中，通常以原始计算资源（如token数、工具调用次数、操作次数、运行时间或成本）作为衡量指标，但这些指标无法区分有效反馈与冗余或不稳定的交互，从而导致对系统性能提升机制的理解偏差。解决方案的关键在于提出有效反馈计算量（Effective Feedback Compute, EFC），这是一种基于轨迹级别的缩放坐标，仅将那些具有信息量、有效性、非冗余性并被保留用于后续决策的反馈计入计算量，并通过任务需求进行归一化处理，以实现跨任务的公平比较。实验表明，EFC能够显著优于传统单变量和多变量基线（如SAS），其预测失败率的决定系数（R²）最高可达0.99，且在固定预算下提升反馈质量可使成功率从0.27大幅提升至0.90，揭示了“反馈效率”而非单纯计算投入才是影响代理（agent）性能的核心因素。
链接: https://arxiv.org/abs/2605.29682
作者: Xuanliang Zhang,Dingzirui Wang,Keyan Xu,Qingfu Zhu,Wanxiang Che
机构: Harbin Institute of Technology (哈尔滨工业大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Agent harnesses increasingly determine the performance of language-model systems by deciding how models call tools, receive feedback, verify intermediate states, store memory, and revise solutions. Yet current test-time scaling analyses often parameterize this process by raw expenditure – tokens, tool calls, operations, wall time, or cost – which does not distinguish useful feedback from redundant or unstable interaction. We introduce \emphEffective Feedback Compute (EFC), a trace-level scaling coordinate that credits feedback only when it is informative, valid, non-redundant, and retained for subsequent decisions, and we normalize it by task demand when comparing tasks with different feedback requirements. Across synthetic controllable tasks, executable code tasks, real benchmark traces, held-out splits, and a prospective validation batch, EFC-based coordinates consistently predict failure rates better than raw-compute baselines and a strong multivariate SAS baseline. In controlled scaling, raw tokens and tool calls explain limited variation ( R^2=0.33 and 0.42 ), SAS reaches 0.88 , while Oracle-EFC and Estimated-EFC reach 0.94 and Oracle-EFC/ D_\mathrmtask reaches 0.99 . Matched-budget interventions show that improving feedback quality raises success from 0.27 to 0.90 while raw cost and tool calls are fixed. On mixed real traces, NRS-EFC/ D_\mathrmtask reaches R^2=0.92 while raw compute has near-zero or negative fit, and it remains the best predictor in a prospective holdout ( R^2=0.85 ). These results suggest that harness scaling is governed less by how much computation is spent than by how efficiently raw budget is converted into durable, task-sufficient feedback.
[NLP-73] Spurious Prompts: Can Irrelevant Prompts Steer Large Language Models ?
【速读】： 该论文试图解决的问题是：大型语言模型（Large Language Models, LLMs）对提示（prompt）的敏感性不仅限于任务相关的指令、示例或推理线索，还可能受到语义上与任务无关的“虚假提示”（spurious prompts）的影响。这类提示虽不涉及任务内容，却能显著改变模型行为。解决方案的关键在于提出了一种简单的黑盒搜索方法来自动发现这些 spurious prompts，并通过实验证明它们在多个推理和问答基准测试中不仅能提升性能（常优于标准提示或任务感知优化提示），还能诱导模型产生非预期行为（如固定选择第一个选项、生成特定数值等）。这一发现揭示了 LLMs 存在一种新的、系统性的提示敏感性机制。
链接: https://arxiv.org/abs/2605.29678
作者: Pawel Batorski,Abtin Pourhadi,Jerzy Sarosiek,Przemyslaw Spurek,Paul Swoboda
机构: Heinrich Heine University Düsseldorf (海因里希·海涅大学杜塞尔多夫); Jagiellonian University (亚捷隆大学); IDEAS Research Institute (IDEAS研究院)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large language models are highly sensitive to prompts, but this sensitivity is usually studied through task-relevant instructions, demonstrations, or reasoning cues. In this paper, we study a different form of prompt sensitivity: whether prompts that are semantically unrelated to the task can nevertheless steer model behavior. We call them spurious prompts and show their surprising efficacy. We also propose a simple black-box search procedure for discovering them. Across reasoning and question-answering benchmarks, using models ranging from 0.8B to 27B parameters and spanning three model families, we show that spurious prompts can improve performance, often matching or outperforming standard prompting baselines and task-aware prompt optimization. We further show that they can steer models toward unintended behaviors, such as repeatedly selecting the first answer option, producing incorrect answers, returning an even, prime or small number without explicitly instructing the model to do so. These findings reveal a new kind of prompt sensitivity: LLMs can be systematically steered by prompts that are unrelated to the task they are asked to solve. Our code is available at this https URL
[NLP-74] Notation Matters: A Benchmark Study of Token-Optimized Formats in Agent ic AI Systems
【速读】： 该论文试图解决的问题是：在生成式 AI（Agentic AI）系统中，当前广泛使用的 JSON 格式在工具调用（tool invocation）和执行结果交换过程中存在显著的 token 开销问题，这限制了模型的效率与性能。尽管已有研究提出如 TOON 和 TRON 等更紧凑的替代格式以提升 token 效率，但它们尚未在完整的端到端代理循环（end-to-end agentic loops）中得到验证。解决方案的关键在于：通过在四个代理基准测试（BFCL、MCPToolBenchPP、MCP-Universe、StableToolBench）上对五种开源大语言模型（LLM）进行系统评估，将输入压缩与输出压缩解耦，独立测量理解（comprehension）和生成（generation）阶段的性能；结果显示，TRON 在保持接近 JSON 准确性（最多仅下降 14 个百分点）的同时实现最高达 27% 的 token 减少，而 TOON 虽然减少幅度较小（最高 18%），但在多轮解析失败时易发生级联错误并导致并行工具调用输出崩溃。
链接: https://arxiv.org/abs/2605.29676
作者: Lorenz Kutschka,Bernhard Geiger
机构: Know Center Research GmbH / Sandgasse 34, A-8010 Graz; Signal Processing and Speech Communication, Graz University of Technology / Inffeldgasse 16c, A-8010 Graz; Graz Center for Machine Learning, A-8010 Graz
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 16 pages, 6 figures, 4 tables
Abstract:Large language models in Agentic AI systems consume tool schemas and execution results and emit tool invocations as structured data. The default language for that exchange, JSON, was designed for application-to-application interchange rather than token efficiency, so its structural elements impose substantial token overhead. Recent work proposes token-optimized alternatives such as TOON (Token-Oriented Object Notation) and TRON (Token Reduced Object Notation) as more compact replacements, but these formats have been evaluated only on isolated comprehension or generation tasks. Whether their token reductions hold inside end-to-end agentic loops therefore remains an open question. We evaluate TOON and TRON on four agentic benchmarks (BFCL, MCPToolBenchPP, MCP-Universe, StableToolBench) and five open-weight LLMs, decoupling input compression from output compression to measure comprehension and generation independently. TRON reduces tokens by up to 27% with accuracy within 14pp of the JSON baseline. TOON achieves up to 18% reduction at a similar 9pp accuracy cost, but additionally cascades on multi-turn parsing failures and collapses parallel tool-call output for most models.
[NLP-75] EviLink: Multi-Path Schema Linking with Uncertainty-Guided Evidence Acquisition for Large-Scale Text-to-SQL
【速读】： 该论文试图解决大规模Text-to-SQL任务中schema linking（模式链接）的问题，即如何从庞大且模糊的数据库中识别出紧凑但充分的模式上下文。传统方法通常将schema linking视为围绕单一SQL路径的确定性选择，然而复杂问题可能对应多种有效的SQL实现方式，其所需的schema信息也各不相同。论文的关键解决方案在于重新将schema linking建模为多路径上的不确定性感知的schema需求推理（uncertainty-aware schema-need inference），系统能够区分必需的schema项与路径相关的不确定项，并仅在必要时获取证据。作者提出了EviLink方法，通过多假设schema定位与不确定性引导的证据获取相结合，显著提升了schema的完整性、相关性与token成本之间的平衡，在Spider2-Snow数据集上实现了90.15%的字段级严格召回率，同时平均仅使用123.30K tokens，并在固定生成器条件下改善了下游SQL生成效果。
链接: https://arxiv.org/abs/2605.29670
作者: Huawei Zheng,Sen Yang,Zhaorui Yang,Yuhui Zhang,Haozhe Feng,Haoxuan Li,Xuan Yi,Chao Hu,Defeng Xie,Chen Hou,Danqing Huang,Wei Chen,Yingcai Wu,Peng Chen,Dazhen Deng
机构: Zhejiang University (浙江大学); Tencent TEG (腾讯技术工程事业群); Peking University (北京大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Schema linking is a difficult and important step in large-scale Text-to-SQL, where systems must identify a compact yet sufficient schema context from large and ambiguous databases. Existing methods often treat schema linking as deterministic selection around a single SQL path, but complex questions may admit multiple valid realizations with different schema needs. We reframe schema linking as uncertainty-aware schema-need inference over multiple plausible SQL paths, where the system distinguishes required schema items from path-dependent uncertain ones and acquires evidence only where needed. We instantiate this reframing with EviLink, which combines multi-hypothesis schema grounding with uncertainty-guided evidence acquisition. Experiments on BIRD-Dev and Spider2-Snow show that this perspective improves the balance among schema completeness, schema relevance, and token cost. On Spider2-Snow, EviLink achieves 90.15% field-level strict recall rate, uses 123.30K average tokens, and improves downstream SQL generation under a fixed generator.
[NLP-76] GRASP: Gated Regression-Aware Skill Proposer for Self-Improving LLM Agents
【速读】： 该论文试图解决大语言模型（LLM）代理在结构化环境中执行任务时因缺乏对环境的程序性知识而导致的行为不可靠问题，特别是现有自我改进方法无法有效验证新增技能是否破坏已有正确行为，从而引发隐性退化的问题。解决方案的关键在于提出GRASP（Gated Regression-Aware Skill Proposer），其将代理改进建模为对有限技能库的一系列编辑操作，并通过一个严格的回归预算约束和平衡的保留探针评估机制，仅接受能在整体性能上带来净提升的新候选技能。实验表明，GRASP在多个临床基准（如MedAgentBench）上显著提升模型表现，且其有效性主要归因于对比式提案生成、接受门控机制以及硬性回归预算，而非单纯技能编写本身；此外，该机制具有跨领域泛化能力，且技能库可跨模型迁移，展现出强者向弱者传递优势的不对称特性。
链接: https://arxiv.org/abs/2605.29668
作者: Johannes Moll,Jean-Philippe Corbeil,Jiazhen Pan,Martin Hadamitzky,Daniel Rueckert,Lisa Adams,Keno Bressem
机构: Technical University of Munich and TUM University Hospital; Microsoft Healthcare Life Sciences
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:LLM agents acting in structured environments fail in operational rather than conversational ways, and reliability depends on procedural knowledge of the environment. Prior self-improvement methods accumulate natural-language guidance without checking that each new item preserves previously correct behavior, so a note that fixes one trajectory can silently regress another. We introduce GRASP (Gated Regression-Aware Skill Proposer), which treats agent improvement as a sequence of edits to a bounded skill library, admitting each candidate only if it produces a net improvement on a balanced held-out probe under a hard regression budget. We evaluate GRASP across five base models (gpt-oss-120b, DeepSeek V4 Flash, Gemini 3.1 Flash Lite, GPT-4.1, GPT-5.4) on two FHIR-based clinical benchmarks. On MedAgentBench, GRASP lifts gpt-oss-120b from 40.6% to 88.8%, exceeds the strongest of five self-improvement baselines by 21.0 points, and improves every other base model by 17.2 to 40.3 points. Ablations attribute the gain to comparative proposal generation, the acceptance gate, and the hard regression budget rather than to skill writing itself, which without validation is no better than using no skills. The mechanism generalizes beyond the clinical domain, improving agents on three of four non-clinical environments and remaining flat only where the action space is open-ended. Frozen libraries transfer across models, where skills from a stronger model improve weaker executors beyond what they learn for themselves while the reverse does not, an asymmetry that no ungated baseline reproduces.
[NLP-77] Beyond English and Evasion: A Human-Annotated Multi-Domain Benchmark for High-Stakes LLM Safety Evaluation in Chinese
【速读】： 该论文试图解决的问题是：当前大型语言模型（LLM）在中文语境下部署时，英文环境下有效的安全系统失效，无法抵御针对中文特有规避技术（如拼音罗马化、字符拆分、网络俚语和语气模糊）的对抗性提示。解决方案的关键在于构建了一个高质量、文化贴合的基准数据集——ChiSafe-PAS（Chinese Safety Pilot Annotation Set），包含1,897个高风险领域的中文对抗性提示，并提供三类响应标签、九类混淆分类法、风险等级评分及标注者推理依据，从而为中文场景下的模型安全性对齐提供可复现、可评估的基准资源。
链接: https://arxiv.org/abs/2605.29667
作者: Wajdi Zaghouani,Kholoud K. Aldous,Yicheng Gao
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:When Large Language Models (LLMs) are deployed in Chinese-language settings, a troubling pattern emerges: safety systems that work well in English break down. These systems struggle to cross linguistic and cultural bound-aries, leaving models exposed to adversarial prompts that exploit Chinese-specific evasion techniques, including Pinyin romanization, character decomposition, internet slang, and hedging tone. To address this gap, we introduce ChiSafe-PAS (Chinese Safety Pilot Annotation Set), a human-annotated benchmark of 1,897 adversarial Chinese prompts spanning four high-stakes domains: self-harm and violence, drug and illicit trade, fraud, and satire. Of these, 1,544 entries carry complete gold-standard annotations: a 3-class response label (REFUSE, SAFE-REDIRECT, RESPOND), a nine-category obfuscation taxonomy, a risk-level rating, and annotator rationale. We describe the dataset design, annotation process, and obfuscation taxonomy in detail. Our primary goal is practical: to give the research community a high-quality, culturally grounded resource for benchmarking LLM safety alignment. In doing so, we engage three broader tensions in the field: the blurring boundary between training and evaluation data, the need for domain coverage grounded in real-world risk, and the limits of scale as a substitute for cultural expertise.
[NLP-78] Opir: Efficient Multi-Task Safety Classification for Toxicity Jailbreaks Hate Speech and Harmful Content
【速读】： 该论文旨在解决大语言模型（LLM）应用中实时安全过滤的效率与准确性难题，即如何在不依赖高计算成本的大型防护模型的前提下，精准识别有害提示、毒性语言、越狱攻击及不当响应，并区分良性敏感内容与隐蔽的有害信息。解决方案的关键在于提出Opir系列编码器架构（GLiClass）的轻量化守卫模型，其核心创新包括：1）基于三级分类体系（共996类）构建多任务学习框架，涵盖二分类安全/不安全、多标签毒性分类、越狱检测及零样本提示与响应分类；2）训练数据融合对抗挖掘的难例负样本、良性保护示例、生成响应、多语言翻译及Aegis2和WildGuard子集，提升泛化能力；3）推出参数少于100M的边缘优化版本，实现低延迟部署；4）开源评估套件支持多种后端模型，全面覆盖安全性任务基准测试。实验表明，Opir在12项安全分类任务和17类细粒度类别任务中优于或持平当前主流守卫系统，同时具备显著更小的部署开销。
链接: https://arxiv.org/abs/2605.29659
作者: Ihor Stepanov,Aleksandr Smechov
机构: Knowledgator(知识猎手); Wordcab
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 23 pages, 4 figures, 9 tables
Abstract:Real-time safety filtering for large language model (LLM) applications requires classifiers that can detect unsafe prompts, toxic language, jailbreak attempts, and unsafe responses without the cost profile of large guardrail models, and that can distinguish benign sensitive text from genuinely covert harmful content. In this paper, we introduce Opir, a family of encoder-based guardrail models built on the GLiClass architecture. Opir includes multi-task models for binary safe/unsafe classification, multi-label toxicity classification, jailbreak classification, and zero-shot unsafe prompt and response categorization. We also release edge variants with fewer than 100M parameters dedicated to binary safe/unsafe categorization. The models are trained on a three-level taxonomy containing 996 categories across 16 top-level labels, 126 mid-level labels, and 854 leaf labels. Opir’s training data combines taxonomy-grounded unsafe prompts, adversarially mined hard negatives, benign safety-preserving examples, generated response examples, multilingual translations, and portions of the Aegis2 and WildGuard training subsets. We also open-sourced an evaluation harness that supports GLiClass and GLiNER2 backends as well as decoder-based models, and covers binary safety classification, multi-label categorization, toxicity, jailbreak detection, prompt safety, response safety, response refusal, and prompt subcategory views across public benchmark families. Across an expanded comparison spanning 12 safety-classification tasks and 17 category tasks against eight contemporary guardrail systems – including both GLiNER2-based and generative guardrail models – Opir variants are competitive on or ahead of the strongest open-weight baselines on the majority of benchmark datasets while operating with a substantially smaller deployment footprint.
[NLP-79] Verifiable Rewards Beyond Math and Code: Lightweight Corpus-Grounded Process Supervision for Factual Question Answering
【速读】： 该论文试图解决在知识密集型问答任务中，利用强化学习（Reinforcement Learning, RL）提升回答事实准确性时面临的奖励设计难题。传统响应级奖励提供粗粒度监督，无法区分推理链中正确与错误的语句；而句子级奖励虽能提供更细粒度反馈，但通常依赖于昂贵且不可靠的自然语言推理（NLI）验证器或大型语言模型（LLM）判别器，尤其在稀有实体事实场景下表现不佳。解决方案的关键在于提出CorVer（Corpus Verify），一种轻量级、可插拔的句子级过程奖励机制：它用基于维基百科共现统计的语料库信号替代神经验证器，通过简单的对齐方式将句子级奖励映射至词元级优势（token-level advantages），仅需一个0.5B规模的提取器和每句一次语料库查询即可完成。实验表明，CorVer在30个（模型，基准）组合中均优于原始基线，平均TriviaQA得分提升4.1个百分点，并在20个测试场景中有18个超越四种神经验证器基线，同时训练速度提升4.8至8.4倍。
链接: https://arxiv.org/abs/2605.29648
作者: Shicheng Fan,Haochang Hao,Dehai Min,Weihao Liu,Philip S. Yu,Lu Cheng
机构: University of Illinois Chicago
类目: Computation and Language (cs.CL)
备注:
Abstract:Applying reinforcement learning to improve factual accuracy in knowledge-intensive question answering faces a reward design dilemma. Response-level rewards provide only coarse supervision and cannot distinguish correct from incorrect statements within a reasoning trace. Sentence-level alternatives offer finer-grained feedback, but typically rely on NLI verifiers, LLM judges, or knowledge-verification pipelines that are expensive to deploy at RL scale and often unreliable for rare-entity facts, where accurate reward signals are especially important. We propose CorVer (Corpus Verify), a lightweight, plug-in-ready process reward that replaces neural verifiers with a corpus-grounded signal derived from Wikipedia co-occurrence statistics. CorVer assigns sentence-level credit and maps it to token-level advantages via a simple alignment, requiring only a 0.5B extractor and a single corpus lookup per sentence. Across 30 (model, benchmark) cells spanning six instruction-tuned models (3B to 14B) and five QA benchmarks, CorVer improves over the raw baseline for every cell, with an average TriviaQA gain of +4.1 pp. It also outperforms four neural-verifier baselines in 18 of 20 cells under their feasible configurations, while training 4.8 to 8.4x faster.
[NLP-80] Classification of non-analyzable word types in web documents to implement an effective Korean e-learning system
【速读】： 该论文旨在解决当前电子学习（e-learning）系统在韩语教学中内容单一的问题，即系统主要依赖正式韩语（formal Korean），而忽略了现实世界中广泛使用的非正式韩语表达，如网络文档、短信或推文中的语言。为了解决这一问题，作者构建了两类语料库：一类是正式文本（如在线新闻文章），另一类是非正式文本（如产品评论博客）。通过对比分析发现，这两类语料库在词汇和句法表达上存在显著差异。针对非正式文本占比高且结构复杂的特点，论文提出使用局部语法图（Local Grammar Graphs, LGG）作为建模工具，以更有效地处理和整合非正式韩语表达，从而提升高级学习者的语言习得效果。解决方案的关键在于利用LGG模型对非正式语料进行结构化建模，使其能够嵌入到韩语电子学习系统中，增强其真实语境适应能力。
链接: https://arxiv.org/abs/2605.29638
作者: Sang-Taek Park,Ae-Lim Ahn,Eric Laporte,Jee-Sun Nam
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:E-learning systems should deliver contents that reflect various phenomena of the language as it is used. In addition to formal Korean, e-learning systems that would include real-world Korean expressions such as those in web documents, mobile text messages, or twitter posts, would be useful to high-level learners. We construct two types of corpora: one is made of formal documents like online news articles; the other is made of informal documents like customer reviews about new products in web blogs. By comparing these corpora, we show how expressions differ in these two types of corpora. We survey the main characteristics of the informal corpus. Given that a significant proportion of text is informal, we propose Local Grammar Graphs (LGG) as an appropriate model to treat them effectively in Korean e-learning systems.
[NLP-81] Evaluating Cross-lingual Knowledge Consistency in Code-Mixed vis-a-vis Indian Languages using IndicKLAR
【速读】： 该论文试图解决大语言模型在英语与低资源语言（特别是印度语及其代码混杂变体）之间知识回忆一致性不足的问题，即跨语言一致性缺口（crosslingual consistency gap）。其解决方案的关键在于构建并评估一个名为IndiKLAR的多语言基准测试集，涵盖18种印度官方语言及11组常用语言对的代码混杂版本，并通过母语者验证确保数据质量。实验发现，模型在原生印度语输入下的准确率可比英语低约0.50，而代码混杂输入能显著缩小这一差距（仅差约0.05），无需模型层面干预。进一步分析表明，无论输入形式还是模型内部转换过程，均存在一个稳定的“翻转点”——即从错误预测到正确预测的临界边界，位于原生语言与代码混杂语言之间，揭示了代码混杂作为桥梁提升跨语言一致性的机制潜力。
链接: https://arxiv.org/abs/2605.29637
作者: Debajyoti Mazumder,Divyansh Pathak,Prashant Kodali,Aditya Joshi,Akshay Agarwal,Jasabanta Patro
机构: Indian Institute of Science Education and Research, Bhopal, India; Microsoft Corporation; UNSW Sydney, Australia
类目: Computation and Language (cs.CL)
备注: 23 pages
Abstract:Large language models recall knowledge reliably in English but often fail on the same query posed in a lower-resourced language – a crosslingual consistency gap that remains underexplored for Indian languages and their code-mixed counterparts. To study this gap, we introduce IndiKLAR, an Indic extension of the KLAR-CLC benchmark covering 18 of the 22 scheduled Indian languages and pairing them with code-mixed variants for 11 widely used language pairs, with native-speaker verification of both monolingual and code-mixed variants for these 11 settings. This three-way alignment offers a unique opportunity to examine how knowledge recall consistency varies across the spectrum of English, code-mixed, and native Indian language inputs. Evaluating across nine open-weight models, we find that the native-language accuracy gap to English can reach \sim 0.50, while code-mixed inputs close most of it – bringing performance within \sim 0.05 of English without any model-level intervention. Motivated by this, we evaluate several prompting strategies that vary in how language conversion is exposed, including a two-stage translate-then-answer setup, a one-stage joint translation-and-answer prompt, and Translate-in-Thought (TinT) – a single-step strategy in which the model converts the input internally and emits only the final answer. Across the performance trajectory native \rightarrow code-mixed \rightarrow English, we identify a consistent flip point – the boundary between incorrect and correct prediction – that lies between the native and code-mixed settings. Interestingly, this holds whether the trajectory is induced by the input surface form or by the model’s internal conversion process.
[NLP-82] Predicting Causal Effects from Natural Language Queries using Structured Representations
【速读】： 该论文试图解决的问题是如何利用大型语言模型（Large Language Models, LLMs）从现有实验证据中预测因果效应的大小，从而降低随机对照试验（Randomized Controlled Trials, RCTs）在成本和时间上的开销。解决方案的关键在于提出了一种两步框架：首先生成查询的结构化语义表示，随后使用监督训练的编码器模型预测效应大小。实验表明，微调（fine-tuning）显著提升了预测性能（绝对误差降低27%至71%），且该两步框架在跨领域泛化方面表现更优，凸显了将语义理解与数值效应估计分离的优势。
链接: https://arxiv.org/abs/2605.29631
作者: Giuliano Martinelli,Piriyakorn Piriyatamwong,Abelardo Carlos Martinez Lorenzo,Jasmin Baier,Riccardo Orlando,Satvik Garg,Sharif Kazemi,Linxi Wang,Arianna Legovini,Samuel Fraiberger
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 18 pages
Abstract:Randomized controlled trials are a cornerstone of medicine and the social sciences as they enable reliable estimates of causal effects. However, they are costly and time-consuming to conduct, motivating interest in predicting causal effects from existing experimental evidence. Recent advances in large language models (LLMs) have demonstrated strong performance on knowledge-intensive tasks, raising the question of whether these models can be used for forecasting causal effect sizes. To investigate this, we introduce Query2Effect, a new large-scale benchmark consisting of more than 72,000 natural language questions aligned with experiment descriptions, created to simulate realistic information-seeking scenarios by varying query specificity along dimensions of implicitness, abstraction, and ambiguity. We then propose a two-step framework that first generates a synthetic structured representation of a query before predicting effect size using a supervised encoder model. Experiments show that finetuning plays a crucial role in improving prediction performance, with absolute error reducing by -27% up to -71% compared to prompted out-of-the-box LLMs, and that our two-step framework is beneficial for out-of-domain generalization, highlighting the benefits of separating semantic interpretation from numerical effect estimation.
[NLP-83] COMET: Concept Space Dissection of the Modality Gap in Audio-Text Multimodal Contrastive Embeddings
【速读】： 该论文试图解决对比语言-音频预训练（CLAP）模型中音频与文本嵌入之间的模态差距（modality gap）问题，该差距严重制约了模型在零样本应用中的性能。现有研究多将模态差距归因于“锥效应”（cone effect），即认为其本质是均值偏移，但仅校正均值无法显著改善性能；其他假设如信息不平衡和维度坍缩也未在音频领域得到充分验证。本文提出COMET框架，基于偏最小二乘奇异值分解（PLS-SVD）揭示模态差距的新视角：仅有少量可解释的轴（捕捉共享概念）对相似性计算贡献显著，而均值成分仅部分反映了模态差距。基于此发现，作者设计了一种无需训练的谱截断方法，在不依赖大规模辅助记忆库或高计算成本的前提下，有效缓解模态差距，使零样本音频描述生成与条件交换性能逼近全监督水平，同时实现嵌入维度大幅压缩并保持检索与音频描述任务的高性能。
链接: https://arxiv.org/abs/2605.29628
作者: Yonggang Zhu,Liting Gao,Aidong Men,Wenwu Wang
机构: School of Artificial Intelligence, Beijing University of Posts and Telecommunications (北京邮电大学人工智能学院); Centre for Vision, Speech, and Signal Processing (CVSSP), University of Surrey (萨里大学视觉、语音与信号处理中心)
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Audio and Speech Processing (eess.AS)
备注:
Abstract:Contrastive Language-Audio Pretraining (CLAP) models are widely used for audio understanding and support modality-agnostic condition swapping in many zero-shot applications. However, their performance is heavily affected by the modality gap between audio and text embeddings. Existing explanations mainly attribute this gap to the cone effect, treating it as a shift between mean embeddings, yet correcting the mean alone yields only limited improvements. Alternative hypotheses, such as information imbalance and dimensionality collapse, have also been proposed, but they remain insufficiently verified and have not been thoroughly studied in the audio domain. Meanwhile, several works attempt to decompose multimodal contrastive embeddings into interpretable concepts, but none explicitly analyze the modality gap from the perspective of concept decomposition. In this work, we introduce COMET (Concept space Organization and Modality gap Explanation with PLS-SVD Transformation), a novel partial least squares singular value decomposition (PLS-SVD) framework for CLAP that unveils a broader perspective of the modality gap. Our framework reveals that only a small, interpretable subset of axes, which captures shared concepts, contributes substantially to similarity computation, and that the mean component represents only partially the modality gap. Building on this insight, we propose a simple spectral truncation method that mitigates the modality gap in a training-free manner. The method enables zero-shot audio captioning with condition swapping to approach fully supervised performance, without requiring large auxiliary memory banks or expensive computation. At the same time, it achieves substantial embedding dimensionality reduction while preserving strong performance on retrieval and audio captioning tasks.
[NLP-84] DLM-SWAI: Steering Diffusion Language Models Before They Unmask
【速读】： 该论文试图解决的问题是：如何在不重新训练模型的前提下，对扩散语言模型（Diffusion Language Models, DLMs）的生成过程进行可控引导，以实现特定文本属性（如风格或安全性）的调节。当前大多数引导方法依赖辅助模型或仅适用于自回归语言模型的逐词解码机制，难以直接应用于通过迭代去噪部分掩码序列来生成文本的DLMs。解决方案的关键在于提出一种无需训练的引导方法——DLM-SWAI，其核心思想是在每个去噪步骤中利用预先计算好的词级别风格分数（token-level style scores）来偏置词分布，从而实现对生成内容的有效控制。实验表明，该方法在保持生成质量的同时显著提升了对风格和安全性的调控能力，并且计算开销极低；消融研究进一步揭示了引导强度与流畅性之间的可调平衡关系，且发现类别级别的可引导性与词级别属性线索的强弱密切相关。
链接: https://arxiv.org/abs/2605.29626
作者: Hyeseon An,Yo-Sub Han
机构: Yonsei University(延世大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: preprint
Abstract:Steering language model generation toward desired textual properties is essential for practical deployment, and inference-time methods are particularly appealing because they enable controllable generation without retraining. Recent work has also highlighted diffusion language models as an emerging generation paradigm with distinct decoding properties. However, most existing steering approaches either rely on auxiliary models or are designed for autoregressive next-token decoding, making them difficult to apply to diffusion language models DLMs, which generate text through iterative denoising of partially masked sequences. Therefore, we propose DLM-SWAI, a simple training-free steering method that biases the token distribution at each denoising step using pre-computed token-level style scores. Experiments on style and safety control tasks show that DLM-SWAI effectively steers diffusion language models while preserving generation quality and requiring minimal computational overhead. Ablations further reveal a controllable trade-off between steering strength and fluency, and our analysis links class-wise steerability to the strength of token-level attribute cues.
[NLP-85] DiffSpot: Can VLMs Spot Fine-Grained Visual Differences in Web Interfaces?
【速读】： 该论文试图解决的问题是：视觉语言模型（Vision-Language Models, VLMs）在高阶图像-文本对齐上表现优异，但在感知细微视觉差异方面能力有限，尤其是在渲染的网页界面中，这种细粒度感知对于GUI代理和设计工具至关重要。解决方案的关键在于提出一个名为DiffSpot的代码驱动基准测试，用于开放式的“找不同”任务。DiffSpot通过在自包含HTML中修改单个CSS属性并重新渲染页面，生成受控的图像对，并利用“定位门”机制确保像素差异仅限于目标元素，从而构建出高质量、可解释的对比样本。该基准包含4,400对图像，涵盖13种CSS属性操作和三个难度层级，同时引入500对无差异样本以控制幻觉。实验表明，即使是最先进的VLM在零样本条件下也仅能识别约40.7%的真实变化，且在高难度层级上的召回率低于23%，揭示了当前模型在细粒度视觉感知方面的显著局限性。
链接: https://arxiv.org/abs/2605.29615
作者: Linhao Zhang,Aiwei Liu,Yuan Liu,Xiao Zhou
机构: WeChat AI, Tencent Inc
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注:
Abstract:Vision-language models (VLMs) have made strong progress on high-level image-text alignment, yet their ability to perceive subtle visual differences remains limited. We study this problem in rendered web interfaces, where localized visual changes are both a diagnostic test of fine-grained perception and a practical requirement for GUI agents and design tools. We introduce \textbfDiffSpot, a code-driven benchmark for open-ended spot-the-difference on web interfaces. DiffSpot constructs controlled image pairs by mutating a single CSS property of a target element in self-contained HTML, re-rendering the page, and recording the changed property, element, and mutation magnitude. A grounding gate retains only pairs whose rendered pixel difference is confined to the target element. The benchmark contains 4,400 pairs, including 3,900 has-diff pairs balanced across 13 CSS-property operators and three difficulty tiers, plus 500 no-diff pairs for hallucination control. Evaluating 13 frontier VLMs zero-shot, we find that even the best model identifies only 40.7% of true changes, with Hard-tier Recall below 23% for every model. DiffSpot further shows that difficulty is strongly property-dependent: across CSS operators, neither pixel magnitude nor CLIP distance reliably predicts Recall.
[NLP-86] raining Deliberative Monitors for Black-Box Scheming Detection
【速读】： 该论文试图解决的问题是：如何在不依赖被监控智能体的内部推理过程或模型参数的情况下，有效识别其潜在的“诡计行为”（scheming behavior），即区分良性任务执行与可能危害系统的隐蔽策略。解决方案的关键在于提出一种“仅基于动作的 deliberative 监控方法”（action-only deliberative monitors）——利用一个开放权重的小型模型，通过监督微调和强化学习从前沿教师模型（frontier teacher）中蒸馏高质量的结构化推理路径，从而实现对代理轨迹中 scheming 行为的高效检测。该方法创新性地结合了 scheming 规范（scheming specification）、独立判别器（judge）过滤机制以及多阶段知识蒸馏，最终在多个分布外的代理误对齐基准上展现出优于低成本提示式前沿模型的性能，同时显著降低边际推理成本（token-metered USD per 1,000 evaluations），并位于成本-性能帕累托前沿，提供了实用且经济的监控替代方案。
链接: https://arxiv.org/abs/2605.29601
作者: Aditya Sinha,Akshat Naik,Victor Gillioz,Simon Storf,Kilian Merkelbach,Rich Barton-Cooper,Axel Højmark,Marius Hobbhahn
机构: MATS Research; Astra Fellowship; Apollo Research
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:As autonomous agents become more capable of performing real-world tasks, distinguishing scheming behavior from benign task pursuit may become a central AI control problem. Existing monitors often rely on chain-of-thought access or internal activations, or use prompted frontier models, all of which can be unavailable, unreliable or expensive in deployment. In this work, we study action-only deliberative monitors: smaller open-weight models trained to detect scheming and sabotage from agentic trajectories without accessing the monitored agent’s reasoning or model internals. Our method, inspired by deliberative alignment, uses a scheming specification to elicit structured rationales from a frontier teacher, filters them with a separate judge, and distills the highest-quality rationales into open-weight monitors with supervised fine-tuning and reinforcement learning. We train on five datasets, and evaluate across six out-of-distribution agentic misalignment benchmarks. We show that applying our method to Qwen3.5-27B yields higher performance than all low-cost frontier models as prompted monitors (Gemini 3.1 Flash-Lite, GPT-5.4 Nano, and Claude Haiku 4.5) and than Gemini 2.5 Pro, while also achieving lower marginal inference cost (token-metered USD per 1,000 evaluations). Stronger prompted frontier monitors (Gemini 3.1 Pro, GPT-5.4, Claude Sonnet 4.6, and Claude Opus 4.6) achieve higher performance but at roughly 16 – 34\times higher marginal inference cost. Several of our trained monitors are positioned on the empirical cost–performance Pareto frontier among the monitors we evaluate, providing practical low-cost, low-FPR alternatives to prompted frontier models.
[NLP-87] World Models in Words: Auditing Physical State-Transition Commitments in Vision-Language Models
【速读】： 该论文旨在解决当前视觉语言模型（Vision-Language Models, VLMs）在物理场景理解任务中评估方式过于简化的問題——即仅以最终答案作为评价标准，掩盖了模型是否真正正确感知对象、表征物理状态、预测合理变化或仅凭错误推理选择正确选项。其解决方案的关键在于提出一个名为 \wmw 的审计框架，要求模型输出包含初始状态、状态转移、结果状态和答案的结构化轨迹（typed trace），并通过混合验证器对轨迹进行多维度检查：包括模式有效性、状态锚定性、转移一致性及答案与轨迹的兼容性，并生成细粒度错误标签（如对象、关系、力、转移、时间、单位/尺度、忠实性等）。该框架不仅揭示了传统答案导向评估所忽视的失效模式（例如35%的正确答案由物理无效轨迹支撑），还通过验证器引导重排序和轨迹级偏好微调显著提升物理一致性，从而提供了一种可复用的协议，用于衡量VLM的陈述物理世界与其答案是否同时成立。
链接: https://arxiv.org/abs/2605.29585
作者: Emmanuelle Bourigault
机构: University of Oxford (牛津大学)
类目: Computation and Language (cs.CL)
备注: 8 pages, 3 figures, 5 tables
Abstract:Vision-language models (VLMs) are increasingly used to answer questions about physical scenes, yet most evaluations reduce performance to a final answer. This hides whether the model perceived the right objects, represented the right physical state, predicted a plausible transition, or merely selected the right option for the wrong reasons. We introduce \wmw, an evaluation framework for auditing the \emphlanguage-expressed physical commitments of VLMs. Instead of scoring only I,q\mapsto a , we ask models to produce a typed trace I,q\mapsto(s_0,\Delta s,s_1,a) : an initial state, a state transition, a resulting state, and an answer. A hybrid verifier then checks schema validity, state grounding, transition consistency, and answer-trace compatibility, yielding typed error labels such as object, relation, force, transition, temporal, unit/scale, and faithfulness errors. We release \tracebank, a controlled trace resource with \nSeed schema- and recomputation-validated synthetic scenarios across \nFamilies physics families, \nPairs minimally perturbed contrastive preference pairs, verifier code, audit guidelines, and model outputs. We evaluate \nModels VLMs on both controlled and external physical-reasoning examples. \wmw reveals failures that answer-only evaluation misses: 35% of correct answers from mid-tier models are backed by physically invalid traces. Verifier-guided reranking recovers up to 7 percentage points of trace validity without sacrificing answer accuracy, and trace-level preference tuning reduces hidden inconsistency by 41% relative. The contribution is not another final-answer physics benchmark, but a reusable protocol for measuring whether a VLM’s stated physical world can be true at the same time as its answer.
[NLP-88] GAPD: Gold-Action Policy Distillation for Agent ic Reinforcement Learning in Knowledge Base Question Answering
【速读】： 该论文试图解决当前基于强化学习（Reinforcement Learning, RL）的知识库问答（Knowledge Base Question Answering, KBQA）系统中，中间动作错误缺乏有效监督的问题。现有方法主要依赖最终答案的稀疏奖励进行优化，导致在逻辑形式标注的KBQA基准测试中，尽管金标准逻辑形式可转化为可执行动作序列，但这些信息仅用于冷启动数据构建，而非用于在线策略更新。解决方案的关键在于提出GAPD（Gold-Action Policy Distillation）框架，通过引入密集的token级指导增强基于结果的RL训练：其核心是MID-ANCHOR MATCHING机制，将学生策略探索过程中到达的中间实体作为状态锚点，匹配学生状态与黄金动作路径中的状态，从而对齐两者；随后以对齐后的黄金动作条件下的当前策略作为停梯度教师模型，将其token分布蒸馏回普通的学生策略，覆盖生成的动作token跨度。此方法在WebQSP、GrailQA和GraphQ等多个基准上持续超越现有最先进水平。
链接: https://arxiv.org/abs/2605.29584
作者: Xin Sun,Jianan Xie,Zhongqi Chen,Qiang Liu,Shu Wu,Bowen Song,Weiqiang Wang,Zilei Wang,Liang Wang
机构: University of Science and Technology of China (中国科学技术大学); NLPR, MAIS, Institute of Automation, Chinese Academy of Sciences (中国科学院自动化研究所自然语言处理重点实验室、多媒体信息处理与智能系统研究中心); Southern University of Science and Technology (南方科技大学); Ant Group (蚂蚁集团)
类目: Computation and Language (cs.CL)
备注:
Abstract:Reinforcement learning (RL) is a natural fit for agentic knowledge base question answering (KBQA), where a model must issue executable actions, observe knowledge-base feedback, and eventually return an answer. However, current RL-based KBQA systems mainly optimize sparse rewards from the final answer, leaving intermediate action errors weakly supervised. This is especially limiting for logical-form annotated KBQA benchmarks: gold logical forms can be converted into executable action sequences, but existing pipelines use them mainly for warm-start data construction rather than for on-policy RL updates. We propose GAPD, a training-time Gold-Action Policy Distillation framework that adds dense token-level guidance to outcome-based RL. To align gold actions with on-policy student rollouts, GAPD uses MID-ANCHOR MATCHING: it treats the intermediate entities reached during student exploration and gold execution as state anchors, and matches student states to gold states through these explored entity sets. The current policy conditioned on this aligned gold action serves as a stop-gradient teacher, whose token distribution is distilled back to the ordinary student policy over generated action-token spans. GAPD consistently surpasses the current state of the art on WebQSP, GrailQA, and GraphQ.
[NLP-89] PEARL: Training Socratic Tutors with Pedagogically Aligned Reinforcement Learning
【速读】： 该论文试图解决的问题是：如何有效训练生成式 AI (Generative AI) 教育辅导代理（tutoring agents），使其不仅能够解答问题，还能提供渐进式的苏格拉底式引导（Socratic guidance），并在多轮交互中平衡多种教学目标（pedagogical objectives）。现有方法面临三大挑战：学生模拟器 fidelity 低且控制能力弱、教学奖励建模不明确、多目标强化学习优化不稳定。解决方案的关键在于提出 PEARL 框架，其核心包括三个创新组件：(1) 可控学生模拟器，将潜在认知状态与响应生成解耦，以刻画不同能力水平和误解；(2) 生成式奖励模型，联合评估教学质量和答案正确性，用于策略优化；(3) 稳定的多目标强化学习机制，通过离散化各维度奖励并归一化优势值聚合，避免高方差目标主导更新。实验表明，PEARL 在多个基准上优于开源模型，并在仅使用 30B 参数策略模型的情况下仍可媲美主流商业大模型。
链接: https://arxiv.org/abs/2605.29582
作者: Qikai Chang,Zhenrong Zhang,Linbo Chen,Pengfei Hu,Jianshu Zhang,Youhui Guo,Jun Du
机构: University of Science and Technology of China (中国科学技术大学); iFLYTEK Research (科大讯飞研究院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 16 pages, 7 figures
Abstract:Large Language Models (LLMs) have shown promise as educational tutors, yet effective tutoring requires more than solving problems: it must provide progressive Socratic guidance and balance multiple pedagogical objectives across multi-turn interactions. However, training such tutors remains challenging due to limited-fidelity and weakly controllable student simulation, under-specified pedagogical reward modeling, and unstable multi-objective optimization. To overcome these limitations, we propose PEARL, a pedagogically aligned reinforcement learning framework for training Socratic tutoring agents, consisting of three key components. First, we introduce a controllable student simulator that decouples latent cognitive states from response generation to model diverse abilities and misconceptions. Second, we develop a generative reward model that jointly evaluates pedagogical quality and objective correctness for policy optimization. Finally, we propose a stable multi-objective RL scheme that discretizes rewards within each dimension and aggregates normalized advantages across dimensions, preventing high-variance objectives from dominating updates. Experiments on multiple benchmarks show that PEARL achieves the best performance among open-source models and remains competitive with leading proprietary LLMs, despite using only a 30B policy model.
[NLP-90] LiteCoder-Terminal: Scaling Long-Horizon Terminal Environments for Learning Language Agents
【速读】： 该论文旨在解决当前语言代理（Language Agents）在终端环境（Terminal Environments）中训练时面临的三大瓶颈问题：多步规划能力不足、反馈驱动执行能力弱以及动态状态适应性差。传统方法依赖于从外部仓库抓取数据，导致领域多样性受限、环境可控性低且难以针对性提升特定能力缺陷。其解决方案的关键在于提出一个零依赖的合成流水线 LiteCoder-Terminal-Gen，能够根据领域规范自动生成可执行且可验证的终端训练环境。基于此框架，作者构建了两个大规模资源：LiteCoder-Terminal-SFT（包含11,255条专家轨迹，覆盖10个领域）和LiteCoder-Terminal-RL（602个可验证环境，用于轨迹级偏好优化）。实验表明，基于SFT数据微调的Qwen模型显著优于基线版本，其中32B规模模型在Terminal Bench 1.0、2.0和Pro上的pass@1指标分别达到29.06%、18.54%和34.00%；进一步采用直接多轮偏好优化（DMPO）方法在RL环境中进行训练，性能持续提升。这系统性地证明了完全合成的、可执行的环境可以为掌握复杂真实命令行工作流提供可扩展且可验证的监督信号。
链接: https://arxiv.org/abs/2605.29559
作者: Xiaoxuan Peng,Kaiqi Zhang,Xinyu Lu,Boxi Cao,Yaojie Lu,Hongyu Lin,Xianpei Han,Le Sun
机构: Chinese Information Processing Laboratory, Institute of Software, Chinese Academy of Sciences (中国科学院软件研究所信息处理实验室); University of Chinese Academy of Sciences (中国科学院大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Mastering terminal environments requires language agents capable of multi-step planning, feedback-grounded execution, and dynamic state adaptation. However, training such agents is currently bottlenecked by a reliance on scraped external repositories, which limits domain diversity, environment controllability, and the targeting of specific capability deficits. We introduce LiteCoder-Terminal-Gen, a zero-dependency synthesis pipeline that autonomously generates executable and verifiable terminal training environments directly from domain specifications. Using this framework, we construct two large-scale resources: LiteCoder-Terminal-SFT, comprising 11,255 expert trajectories across 10 domains, and LiteCoder-Terminal-RL, featuring 602 verifiable environments for trajectory-level preference optimization. Supervised fine-tuning of Qwen-family models on our SFT dataset yields agents that significantly outperform their base counterparts. Notably, our 32B variant achieves 29.06%, 18.54%, and 34.00% pass@1 on Terminal Bench 1.0, 2.0, and Pro, respectively. Furthermore, applying Direct Multi-turn Preference Optimization (DMPO) on our RL environments yields additional performance gains. These results systematically demonstrate that fully synthetic, executable environments offer a scalable and verifiable supervision signal for mastering complex, real-world command-line workflows.
[NLP-91] From Blind Guess to Informed Judgment: Teaching LLM s to Evaluate Materials by Building Knowledge-Augmented Preference Signals
【速读】： 该论文试图解决材料发现中候选集评估可靠性不足的问题，即在高通量实验和候选生成技术进步后，如何从海量候选材料中可靠地筛选出优质选项。解决方案的关键在于提出一种知识增强的偏好信号框架（MaterEval），通过为同一候选材料生成两种评价：一种基于专家规则并附带支持证据的“有依据判断”，另一种去除规则限制的“盲猜”判断，并将二者配对作为偏好数据，引导原本缺乏材料领域判据的通用大语言模型（LLM）从直觉判断转向基于显式证据的可靠评估。此外，引入快慢推理机制，将大规模快速筛选与小样本深度审核分离，在保证效率的同时提升评估可靠性。实验证明，仅依赖内部知识的小型开源LLM即可显著提升准确性、结论一致性和证据区分能力，接近闭源规则驱动模型的性能，表明专家规则可系统转化为可学习的偏好信号，从而构建低成本、可部署的自主材料发现评估模块。
链接: https://arxiv.org/abs/2605.29555
作者: Yeyong Yu,Wenya Hu,Xing Wu,Quan Qian
机构: Shanghai University (上海大学)
类目: Computation and Language (cs.CL)
备注: 33 pages, 5 figures
Abstract:As candidate generation and high-throughput experimentation advance, the primary bottleneck in materials discovery is shifting from property prediction to making reliable evaluations among massive candidate sets. We propose a Knowledge-Augmented Preference Signals Framework, MaterEval, that automatically produces, for the same candidate, two evaluations: an informed judgment that follows expert rules and provides supporting evidence, and a rule-removed blind guess. By pairing the two evaluations as preference data, we guide general-purpose large language models (LLMs), originally lacking materials-specific criteria, from intuitive judgment toward reliable evaluation supported by explicit evidence. To balance throughput, cost, and reliability, we further introduce a fast-slow reasoning scheme that decouples large-scale rapid screening from in-depth review on a small subset. Using high-entropy alloy (HEA) assessment as a case study, we show that, without external retrieval and relying solely on internalized capabilities, small open-source LLMs achieve substantial gains in accuracy, conclusion consistency, and evidence discrimination, approaching the performance of rule-based closed-source LLMs. These results demonstrate that expert rules can be systematically transformed into learnable preference signals, enabling a low-cost and deployable evaluation module for autonomous materials discovery loops.
[NLP-92] Source-Grounded Semantic Reinforcement Learning for Low-Resource Target-Language Generation
【速读】： 该论文试图解决低资源目标语言生成任务中因平行语料稀缺而导致性能受限的问题，同时探索如何有效利用高资源源语言的大量单语数据。其解决方案的关键在于提出了一种名为“源语言引导的语义强化学习”（Source-Grounded Semantic Reinforcement Learning, SG-SRL）的框架：该框架通过跨语言语义奖励模型（由跨语言重排序器实现）对源语言单语数据进行无参考强化学习，从而为目标语言生成提供语义监督信号；尽管这一过程可能引发冗余性的奖励欺骗（reward hacking），但引入一个轻量级恢复阶段——使用少量平行语料微调模型——可有效恢复流畅性、简洁性和任务格式，同时保留语义增益。实验表明，在中泰翻译任务上，SG-SRL显著优于冷启动监督微调（cold-start SFT），且分析进一步揭示了其在长文本迁移和藏语嵌入奖励下的泛化能力，证明基于编码器的语义奖励可在真实低资源场景中替代大语言模型驱动的重排序器。
链接: https://arxiv.org/abs/2605.29502
作者: Zeli Su,Ziyin Zhang,Zewei Pan,Zhou Liu,Dingcheng Huang,Dehan Li,Zhankai Xu,Longfei Zheng,Xiaolu Zhang,Jun Zhou,Wentao Zhang
机构: Minzu University of China (中国民族大学); Ant Group (蚂蚁集团); Shanghai Jiao Tong University (上海交通大学); Peking University (北京大学); Harbin Institute of Technology (哈尔滨工业大学); South China University of Technology (华南理工大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Low-resource target-language generation is often limited by scarce parallel data, while high-resource source-language monolingual data is abundant but difficult to use with standard supervised fine-tuning. We propose Source-Grounded Semantic Reinforcement Learning (SG-SRL), a resource-utilization framework that converts source-language monolingual data into cross-lingual semantic supervision for target-language generation. SG-SRL performs reference-free reinforcement learning (RL) on source-language data using a cross-lingual semantic reward model, instantiated by a cross-lingual reranker that scores the semantic relevance between the source input and the target-language generation. While this induces severe verbosity-based reward hacking, a lightweight recovery stage using a small parallel corpus restores fluency, conciseness, and task format while preserving the semantic gains. Experiments on Chinese-to-Thai generation show that SG-SRL improves semantic grounding and factual coverage over cold-start SFT. Additional analyses on long-form transfer and Tibetan embedding-based rewards clarify the generalization behavior of SG-SRL and show that an encoder-based semantic reward can substitute for an LLM-based reranker in a realistic low-resource language setting.
[NLP-93] Mask the Target: A Plug-and-Play Regularizer Against LoRA Forgetting
【速读】： 该论文试图解决的问题是：在使用低秩适应（LoRA）对大语言模型进行微调时，虽然模型在目标任务上表现提升，但可能严重遗忘预训练和对齐阶段所学的原有能力，尤其是在适应分布与原始训练分布差异较大时。这种遗忘现象在实际场景中尤为严峻，因为原始训练数据通常不可获取。解决方案的关键在于引入一种仅作用于损失层的输出空间正则化方法——通过移除目标词元并重新归一化剩余概率，仅对非目标词汇空间施加KL散度正则化，从而在不干扰交叉熵梯度的前提下保留基础模型对替代词元的相对偏好。该方法无需重放数据、架构修改或推理阶段开销，可直接集成到现有LoRA训练流程中，显著改善新学习与遗忘之间的权衡关系，提升模型更新的可靠性。
链接: https://arxiv.org/abs/2605.29498
作者: Runze Xu,Arpit Garg,Hemanth Saratchandran,Simon Lucey
机构: Australian Institute for Machine Learning (澳大利亚机器学习研究所); Adelaide University (阿德莱德大学)
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: In Submission
Abstract:Low-Rank Adaptation (LoRA) has become one of the most widely used fine-tuning mechanisms for adapting large language models to new domains, tasks, and users. Yet adaptation performance alone can obscure an important failure mode: LoRA updates may improve performance on the target distribution while degrading prior capabilities learned during pretraining and alignment. We show that this forgetting becomes especially severe when the adaptation distribution differs substantially from the models original training or alignment distributions. The challenge is amplified in practical settings, where the original training and alignment data are typically unavailable. Motivated by this constraint, we study how LoRA based adaptation balances new learning against forgetting in a replay-free setting, and introduce a simple output space regularizer that can be added directly to existing training pipelines. Our method removes the ground-truth token from both the base and adapted model distributions, renormalizes the remaining probabilities, and applies KL regularization only over the non-target vocabulary. This preserves the base models relative preferences among alternative tokens without directly opposing the cross-entropy signal required for adaptation. As the regularizer acts only at the loss level, it requires no replay data, architectural changes, adapter redesign, or inference-time overhead, and can be applied directly to existing LoRA variants. Across all LoRA variants tested and across various backbones, our method improves the frontier between new learning and forgetting when the adaptation distribution differs substantially from the base models original training or alignment distributions, suggesting a broadly applicable route toward more reliable LLM updating.
[NLP-94] On Asymmetric Optimization of Reasoning and Perception in Vision-Language Model Post-Training
【速读】： 该论文试图解决的问题是：在前沿视觉-语言模型中，后训练（post-training）方法显著提升了推理能力，但对感知（perception）的改进相对有限，从而成为端到端视觉推理的瓶颈。解决方案的关键在于通过一个受控的诊断框架识别感知与推理之间的不对称优化机制，并提出针对性干预措施：对于监督微调（SFT），问题源于思维链（chain-of-thought）监督中的token不平衡，导致感知获得较弱的训练信号，通过动态重加权损失函数可缓解此问题并提升端到端性能达18.2%；对于强化学习（RL），不对称性源于奖励耦合（reward coupling）——结果奖励与推理关联更强而与感知关联较弱，通过引入感知感知奖励（perception-aware reward）可改善感知学习，使端到端准确率提升6.0%，即使无真实感知奖励，使用可靠的代理奖励（surrogate reward）也能带来3.2%的增益。
链接: https://arxiv.org/abs/2605.29496
作者: Xueqing Wu,Yu-Chi Lin,Kai-Wei Chang,Nanyun Peng
机构: University of California, Los Angeles
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: Project: this https URL
Abstract:Post-training has greatly improved reasoning in frontier vision-language models, yet its gains for perception remain comparatively limited, creating a bottleneck for end-to-end visual reasoning. To investigate this gap, we introduce a controlled diagnostic framework with two synthetic tasks that disentangle perception from reasoning. Our analysis reveals a consistent perception-reasoning asymmetry: posttraining improves reasoning more substantially than perception, though the underlying mechanism differs by training paradigm. For supervised fine-tuning (SFT), this asymmetry stems from token imbalance in chain-of-thought supervision, where perception occupies fewer tokens and thus receives a weaker training signal. Dynamically reweighting the loss mitigates this imbalance and boosts end-to-end performance by up to 18.2. For reinforcement learning (RL), the asymmetry instead arises from reward coupling: outcome rewards correlate more strongly with reasoning than with perception, weakening the signal for perception learning. Adding a perception-aware reward alleviates the imbalance and improves end-to-end accuracy by up to 6.0; even without groundtruth perception rewards, a reliable surrogate reward provide useful signal, yielding gains of 3.2 points. Together, our results comprehensively diagnose asymmetric optimization and suggest concrete interventions to balance perception and reasoning.
[NLP-95] PhoneWorld: Scaling Phone-Use Agent Environments
【速读】： 该论文试图解决的问题是：当前手机使用代理（phone-use agents）面临的核心瓶颈在于难以大规模构建可控且可复现的、能够覆盖真实移动行为的环境。现有移动代理基准测试虽然在评估方面取得了进展，但无法提供一种可扩展的方式来创建大量新的手机使用环境。其解决方案的关键在于提出 PhoneWorld——一个可重用的流水线系统，该系统能将真实的 GUI 轨迹和截图转化为可控的手机使用环境、可执行任务、自动验证器以及训练回放数据。PhoneWorld 通过分析真实轨迹来识别关键屏幕、屏幕间连接关系、影响环境状态的操作以及可自动验证的用户目标，并基于这些信号构建由只读应用内容和可变状态支持的可运行模拟 Android 应用，进而从中衍生出任务、规则驱动的验证机制和训练数据。实验证明，在固定训练预算下，用 PhoneWorld 的监督替代辅助 AndroidWorld 数据可同时提升四个评估基准的表现；此外，增加 PhoneWorld 监督量或扩大应用覆盖范围均能显著提升性能。总体而言，PhoneWorld 将研究重点从逐个构建移动基准转向规模化生成手机使用环境本身。
链接: https://arxiv.org/abs/2605.29486
作者: Zhengyang Tang,Yuxuan Liu,Xin Lai,Junyi Li,Pengyuan Lyu,Jason,Yiduo Guo,Zhengyao Fang,Yang Ding,Yi Zhang,Weinong Wang,Huawen Shen,Xingran Zhou,Liang Wu,Fei Tang,Sunqi Fan,Shangpin Peng,Zheng Ruan,Anran Zhang,Benyou Wang,Rui Yan,Ji-Rong Wen,Chengquan Zhang,Han Hu
机构: Tencent Hunyuan; The Chinese University of Hong Kong, Shenzhen; Gaoling School of Artificial Intelligence, Renmin University of China; Wuhan University
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: work in progress
Abstract:A central bottleneck for phone-use agents is that controllable, reproducible environments covering real mobile behavior are hard to build at scale. Existing mobile-agent benchmarks have made important progress on evaluation, but they do not by themselves provide a scalable way to construct many new phone-use environments. We present PhoneWorld, a reusable pipeline that converts real GUI trajectories and screenshots into controllable phone-use environments, executable tasks, automatic verifiers, and training rollouts. Rather than hand-building one mobile benchmark at a time, PhoneWorld uses real trajectories to recover which screens matter, how screens connect, which interactions must change environment state, and which user goals admit automatic verification. From these signals, it builds runnable mock Android apps backed by read-only app content and mutable state, then derives executable tasks, rule-based verifiers, and training rollouts from the same environments. In its current instantiation, PhoneWorld covers 34 apps across 16 domains, spanning common consumer mobile behaviors such as search, browsing, shopping, booking, media, and social interaction. Under a fixed training budget, replacing 10K steps from an auxiliary AndroidWorld corpus in an AndroidWorld-based baseline with broad PhoneWorld supervision improves all four evaluation benchmarks at once, raising HYMobileBench by 17.7 points, AndroidControl by 6.0 points, AndroidWorld by 14.7 points, and PhoneWorld by 52.5 points. We then study two additional scaling questions: increasing the amount of PhoneWorld supervision strongly improves PhoneWorld performance, and under a fixed PhoneWorld budget, expanding app coverage yields even larger gains. Overall, PhoneWorld shifts the focus from building one mobile benchmark at a time to scaling the supply of phone-use environments themselves.
[NLP-96] Comparative Evaluation of Machine Translation Systems on Images with Text
【速读】： 该论文试图解决的是图像中文本信息的机器翻译问题，即如何在保留图像语义和布局的前提下，准确地将图像中的源语言文本翻译为目标语言。这一任务融合了计算机视觉（Computer Vision）与自然语言处理（Natural Language Processing）两大领域。解决方案的关键在于比较三种不同范式：1）模块化流水线（modular pipelines），分别执行文本检测、识别和翻译；2）多模态大语言模型（Multi-modal Large Language Models, MLLMs），能够联合处理图像与文本；3）端到端模型 Translatotron-V，直接生成翻译后的图像。实验表明，尽管端到端方法在流程上更简洁，但模块化系统表现更优，而 MLLMs 因其强大的多模态推理能力和上下文理解能力，在整体性能上最佳，验证了多模态协同处理在跨语言图像文本翻译中的有效性。
链接: https://arxiv.org/abs/2605.29476
作者: Blai Puchol,Sergio Gómez González,Miguel Domingo,Francisco Casacuberta
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:This work presents a comparative evaluation of machine translation systems applied to images containing textual information, a task that lies at the intersection of computer vision and natural language processing. The study compares three main paradigms: modular pipelines that separate text detection, recognition, and translation; multi-modal large language models (MLLMs) capable of processing both image and text jointly; and an end-to-end model, Translatotron-V, which directly generates translated images. The modular systems employ state-of-the-art OCR (docTR) combined with multilingual LLMs such as Llama and EuroLLM, while the evaluated MLLMs include different configurations of Gemini 2.5. Experiments were conducted on parallel multilingual datasets covering multiple language pairs, with evaluation based on BLEU, chrF, and TER metrics. The results show that modular pipelines outperform the end-to-end approach, while MLLMs achieve the best overall performance, demonstrating superior flexibility and contextual understanding. These findings underscore the effectiveness of multi-modal reasoning for image-to-text translation and provide a solid foundation for future research on integrating visual understanding and language generation in multilingual settings.
[NLP-97] Kronecker Embeddings: Byte-Level Structured Token Representations for Parameter-Efficient Language Models KR
【速读】： 该论文旨在解决大规模语言模型（Large Language Models, LLMs）中输入嵌入层（input embedding layer）参数量巨大、训练成本高昂的问题。传统方法使用一个形状为 |V| × d_model 的可学习嵌入表（embedding table），在前沿模型中消耗数亿至数十亿可训练参数。其解决方案的核心是提出Kronecker Embeddings，一种基于字节级字符-位置确定性分解的嵌入机制：它用一个固定编码器和单个可学习投影层替代原嵌入表，兼容标准BPE分词器，在保持性能的同时减少91–94%的输入侧可训练参数。关键创新在于通过字节级因子分解实现高效嵌入表示，同时提升对拼写错误的鲁棒性、稳定嵌入范数，并支持运行时动态重建嵌入，显著降低内存占用与计算开销。
链接: https://arxiv.org/abs/2605.29459
作者: Rohan Shravan
机构: The School of AI (The School of AI)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 28 pages, 16 tables. Reference implementation: this https URL
Abstract:Large language models route every input through a learned embedding table of shape |V| x d_model, consuming hundreds of millions to billions of trainable parameters at frontier scale. We introduce Kronecker Embeddings, a deterministic byte-level character-position factorization that replaces this table with a fixed encoder and a single learned projection, compatible with standard BPE tokenizers, eliminating 91–94% of input-side trainable parameters at frontier scale. We provide five contributions. First, a cross-model probe across six LMs (135M-671B parameters) shows trained input embeddings cluster typographic variants of the probe word far more than morphological relatives; Kronecker escapes this clustering at the embedding layer. Second, a controlled three-seed comparison on nanoGPT GPT-2 124M over 2.5B tokens of FineWeb-Edu shows Kronecker reaching 2.5 ± 0.2% lower validation loss than the BPE-tied baseline (gap 0.083 ± 0.007 nats, ~9% lower perplexity), needing ~1.43x fewer steps to reach BPE’s converged loss. Third, a spelling-robustness probe over 110 clean/typo pairs shows Kronecker preserves the top-1 prediction on 55.5% of pairs vs. 47.3% for BPE (+8.2 pp) and lowers KL by 7.6%, winning or tying in 10 of 11 categories; a generation probe shows Kronecker echoes byte-novel strings and typos through generation where BPE forgets them. Fourth, BPE embedding norm drifts during training while Kronecker projection norm stays near 1.0, consistent with a stable representational target. Fifth, an on-the-fly runtime variant reconstructs embeddings from a 4.5 MB byte buffer rather than a 2.15 GB table at vocabulary 131,072, with 0.01–0.24% step-time overhead. Byte-level locality has a tradeoff: byte-similar but semantically distant pairs (compute/commute, nation/notion) cluster together, shifting disambiguation to early attention layers.
[NLP-98] Adaptive Interviewing for Persona Simulation in LLM s: Evidence-Grounded Reasoning Improves Decision Alignment
【速读】： 该论文试图解决的问题是：如何让大语言模型（LLM）更准确地模拟特定个体的决策行为，尤其是在道德困境场景中。现有方法常依赖静态的人格描述（persona），但缺乏个体的价值观、经历和情境线索，导致模拟效果不佳。其解决方案的关键在于提出一种自适应访谈框架（adaptive interview framework），通过三个阶段的结构化对话（核心问题、动态追问和人格总结）收集用户特有的信息，并基于这些对话记录评估 LLM 的决策模拟能力。研究发现，这种框架并非普遍提升准确性，而是作为选择性“ grounding 机制”——约40%的完整对话轨迹会引入追问所得证据，且这些被追问内容支撑的预测准确率（45.5%）显著高于仅依赖核心问题的回答（39.3%）。这表明，仅提供丰富的人格背景信息不足以改善模拟效果，关键在于模型是否真正将决策建立在用户特定证据之上。
链接: https://arxiv.org/abs/2605.29458
作者: Ruoxi Su,Yuhan Liu,Jingyu Hu
机构: University of Cambridge; Independent Researcher
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 20 pages, 2 figures, 12 tables
Abstract:Accurately simulating the decisions of a specific individual remains challenging for large language models (LLMs), partly because persona information is often provided as static descriptions that miss the values, experiences, and contextual cues needed for individual-level decision simulation. We propose an adaptive interview framework that gathers persona-relevant information through a structured three-stage dialogue: core questions, dynamic follow-ups, and a synthesized personality summary. Using the resulting interview transcripts, we evaluate whether LLMs can simulate participants’ decisions in moral dilemma scenarios. We compare three conversational contexts – Core-10 responses, the full interview dialogue, and a summarized persona representation. We find that adaptive interviewing functions less as a uniform accuracy booster and more as a selective grounding mechanism: follow-up-derived evidence is incorporated in around 40% of full-interview traces, and these follow-up-grounded predictions are more accurate than core-only grounded ones (45.5% vs. 39.3%). These findings highlight that richer persona context alone is insufficient: improvements arise only when models actually ground their decisions in user-specific evidence.
[NLP-99] Recovering Policy-Induced Errors: Benchmarking and Trajectory Synthesis for Robust GUI Agents ICML2026
【速读】： 该论文试图解决的问题是：当前图形用户界面（GUI）智能体在实际部署中缺乏从自身错误中恢复的能力，导致其鲁棒性不足。为在评估和数据两个层面弥合这一差距，作者提出了GUI-RobustEval评测基准和基于鲁棒性的轨迹合成方法（Robustness-driven Trajectory Synthesis, RoTS）。解决方案的关键在于：1）构建包含1,216个可执行测试用例的GUI-RobustEval，系统性地衡量多种真实场景下的错误恢复能力；2）设计RoTS框架，通过树状结构的数据合成管道主动发现多样化的错误模式并生成对应的恢复步骤，从而构建80万条高质量训练数据。基于该数据微调的模型（RoTS-7B和RoTS-32B）在GUI-RobustEval和传统GUI基准上均取得显著提升，其中RoTS-32B在OSWorld上达到47.4%的成功率和33.8%的All-Pass@4分数，证明了增强长程错误恢复能力对整体性能与鲁棒性的协同提升作用。
链接: https://arxiv.org/abs/2605.29447
作者: Tianpeng Bu,Xin Liu,Qihua Chen,Hao Jiang,Shurui Li,Hongtao Duan,Lu Jiang,Lulu Hu,Bin Yang,Minying Zhang
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: ICML 2026 Spotlight. 36 pages, 19 figures, includes appendix
Abstract:While GUI agents have advanced rapidly, they often lack the robustness to recover from their own errors, hindering real-world deployment. To bridge this gap at both the evaluation and data levels, we introduce GUI-RobustEval and propose Robustness-driven Trajectory Synthesis. GUI-RobustEval contains 1,216 executable test cases that systematically measure error recovery capabilities across a broad and realistic spectrum of error modes. At the data level, RoTS is a scalable synthesis framework that creates 800k high-quality data via a tree-based pipeline that proactively discovers diverse error modes and synthesizes corresponding recovery steps. Our two models, RoTS-7B and RoTS-32B, fine-tuned on our dataset, both demonstrate significant gains on GUI-RobustEval and traditional GUI benchmarks. Notably, RoTS-32B achieves state-of-the-art performance on OSWorld, with a 47.4% success rate and a 33.8% All-Pass@4 score, suggesting that improved long-horizon error recovery ability contributes to both robustness and overall performance. Our code is available at this https URL.
[NLP-100] AliMark: Enhancing Robustness of Sentence-Level Watermarking Against Text Paraphrasing ICML2026
【速读】： 该论文试图解决现有句子级水印方法在面对强 paraphrasing 工具（如 DIPPER 和 GPT-3.5）时，因前缀式设计而对句法结构扰动（如句子拆分与合并）敏感的问题。其解决方案的关键在于提出 AliMark 框架，将句子级水印建模为比特序列的编码与对齐问题，并采用两阶段检测策略：通过生成多个重构文本变体，自适应地将其提取的比特序列与秘密比特序列对齐以最小化对齐代价，从而自然提升对句子拆分和合并等结构扰动的鲁棒性。
链接: https://arxiv.org/abs/2605.29434
作者: Yuexin Li,Wenjie Qu,Linyu Wu,Yulin Chen,Yufei He,Tri Cao,Bryan Hooi,Jiaheng Zhang
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted by ICML 2026
Abstract:Existing sentence-level watermarking methods enhance robustness to paraphrasing by anchoring watermarks in sentence semantics. However, their prefix-based designs remain vulnerable to structural perturbations, such as sentence splitting and merging, which commonly arise under strong paraphrasers like DIPPER and GPT-3.5. To mitigate this issue, we propose AliMark, a framework that reformulates sentence-level watermarking as a bit sequence encoding and alignment problem between a potentially watermarked text and a secret bit sequence. Notably, our approach adopts a two-stage detection strategy: we generate multiple restructured text variants and adaptively align their extracted bit sequences with the secret bit sequence to minimize alignment cost. This multi-candidate alignment design naturally improves robustness to sentence merges and splits. Extensive experiments demonstrate that AliMark substantially outperforms state-of-the-art baselines under diverse paraphrasing attacks.
[NLP-101] owards Human-Like Interactive Speech Recognition With Agent ic Correction and Semantic Evaluation
【速读】： 该论文试图解决当前自动语音识别（ASR）系统普遍采用单次遍历（single-pass）范式所带来的局限性，这种范式与人类沟通中通过迭代澄清和修正来消除误解的机制不一致，导致一旦发生语义层面的关键错误难以纠正。同时，传统基于词或字符级别的指标（如词错误率 WER 或字符错误率 CER）无法有效衡量此类语义错误。解决方案的关键在于将 ASR 重构为多轮交互式精炼任务，并提出 Agentic ASR 框架——一个闭环系统，融合单次遍历 ASR 前端、语义纠错、意图路由与基于推理的编辑模块；此外，引入基于大语言模型（LLM）的句子级语义错误率（S²ER）作为评估指标，并构建交互式模拟系统以实现可扩展且可复现的基准测试。实验表明，在多语言、命名实体密集及代码混用等场景下，迭代交互显著降低语义错误，且在 S²ER 上提升远超传统 token 级指标，人机对齐与消融实验证实了语义判别器的可靠性与框架的鲁棒性。
链接: https://arxiv.org/abs/2605.29430
作者: Zixuan Jiang,Yanqiao Zhu,Peng Wang,Qinyuan Chen,Xinjian Zhao,Xipeng Qiu,Wupeng Wang,Zhifu Gao,Xiangang Li,Kai Yu,Xie Chen
机构: Xi’an Jiaotong University (西安交通大学); Shanghai Jiao Tong University (上海交通大学); The Chinese University of Hong Kong, Shenzhen (香港中文大学（深圳）); Fudan University (复旦大学); Alibaba Group (阿里巴巴集团)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Automatic speech recognition (ASR) is a core component of human–computer interaction and an increasingly important front-end for LLM-based assistants and agents. However, most current ASR systems still follow a single-pass paradigm, which is poorly aligned with human communication, where misunderstandings are resolved through iterative clarification and refinement. This mismatch makes it difficult to correct meaning-critical errors once they occur. Meanwhile, token-level metrics such as WER or CER cannot adequately reflect such a problem. To address these limitations, we formulate \emphInteractive ASR as a multi-turn refinement task and propose \textbfAgentic ASR, a closed-loop framework that combines a single-pass ASR front-end with semantic correction, intent routing, and reasoning-based editing. We further introduce the \textbfSentence-level Semantic Error Rate ( S^2ER ), an LLM-based semantic evaluation metric, together with an \textbfInteractive Simulation System for scalable and reproducible benchmarking. Experiments on multilingual, named-entity-intensive, and code-switching benchmarks show that iterative interaction consistently reduces semantic errors, with much larger gains in S^2ER than in conventional token-level metrics. Human–AI alignment and ablation studies further validate the reliability of the semantic judge and the robustness of the proposed framework. The code is available at: this https URL and the live demo is available at this https URL
[NLP-102] FinGuard: Detecting Financial Regulatory Non-Compliance in LLM Interactions
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在金融领域应用中因缺乏对具体金融监管法规理解而导致的合规风险问题。现有防护模型依赖通用危害分类体系，未能覆盖由特定金融法规引发的违规行为。其解决方案的关键在于提出一个以监管文件驱动的端到端流水线，直接解析法规文本生成合规风险分类体系，并合成基于法规依据的训练数据，无需预设违规类别。该方法在中文金融监管法规上实现并发布了首个金融合规检测基准FinGuard-Bench，包含查询与响应层面的专业标注；同时训练出FinGuard模型（基于Qwen3-8B），通过监督微调和自Play强化学习提升检测性能，在基准上显著优于包括GPT-5.1在内的各类基线模型，且保持通用安全能力并能仅凭机构政策文档适应新合规要求。
链接: https://arxiv.org/abs/2605.29427
作者: Huaixia Dou,Jie Zhu,Minghao Wu,Shuo Jiang,Junhui Li,Lifan Guo,Feng Chen,Chi Zhang
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:As large language models (LLMs) are increasingly deployed in financial services, a single non-compliant interaction can expose institutions to regulatory penalties and direct consumer harm. Existing guard models are built around general harm taxonomies and overlook violations grounded in specific financial regulations. We address this gap with a regulation-driven pipeline that operates directly on regulatory documents, inducing a financial compliance risk taxonomy and synthesizing grounded training data without any predefined violation categories. Instantiating the pipeline on Chinese financial regulations, we release \textbfFinGuard-Bench, to our knowledge the first benchmark for financial regulatory compliance detection, with expert-annotated labels at both the query and response levels. We further train \textbfFinGuard, a financial compliance detection model built on Qwen3-8B and trained on the regulation-grounded data via supervised fine-tuning and self-play reinforcement learning. On FinGuard-Bench, FinGuard substantially outperforms all baselines, including dedicated guard models and much larger general-purpose LLMs such as Qwen3.5-397B-A17B and GPT-5.1. Furthermore, FinGuard also preserves general safety capabilities and adapts to unseen institution-specific policies using policy documents alone. We will publicly release the code, prompts, and resources used in this work on GitHub.
[NLP-103] Learning Design Skills as Memory Policies for Agent ic Photonic Inverse Design ICML2026
【速读】： 该论文旨在解决光子晶体光纤（Photonic Crystal Fiber, PCF）逆向设计中因昂贵的电磁仿真导致的效率低、知识难以复用的问题。现有方法虽在代理预测或单次参数推荐上有所改进，但缺乏跨迭代试验的知识积累机制。论文提出SkillPCF框架，将PCF逆向设计建模为记忆-策略学习问题，其核心创新在于融合物理引导的记忆技能库、强化学习驱动的技能选择机制以及基于仿真的技能演化策略，从而实现设计知识的持续积累与优化。实验表明，SkillPCF在多种大语言模型（LLM）和经典基线方法中均展现出更优的设计质量与效率权衡，验证了该记忆-技能学习范式在物理感知PCF逆向设计中的有效性。
链接: https://arxiv.org/abs/2605.29421
作者: Shengchao Chen,Ting Shu,Sufen Ren
机构: 未知
类目: Computation and Language (cs.CL)
备注: AI4Physics@ICML 2026
Abstract:Photonic crystal fiber (PCF) inverse design remains challenging because candidate geometries must satisfy coupled optical targets under expensive electromagnetic simulation. Existing pipelines improve surrogate prediction or one-shot parameter recommendation, but they do not accumulate reusable design knowledge across iterative trials. We formulate PCF inverse design as a memory-policy learning problem and propose SkillPCF, a closed-loop agent framework that combines a physics-guided memory skill bank, reinforcement-learned skill selection, and simulator-grounded skill evolution. We further construct a real-world dataset with 479 expert interaction traces (2,507 spans) and 553 memory-dependent evaluation queries covering dispersion engineering, loss optimization, and multi-objective design. Experiments across multiple LLM backbones and classical baselines show that SkillPCF achieves stronger design-quality and efficiency trade-offs under practical simulation budgets, demonstrating the effectiveness of our proposed memory-skill learning paradigm for physics-aware PCF inverse design.
[NLP-104] Beyond Bilingual Transfer: Multilingual Code-Switching in Instruction Tuning
【速读】： 该论文试图解决的问题是：当前关于代码切换数据（Code-Switching Data, CSD）的研究主要集中在英语与单一目标语言之间的双语迁移，而对包含三个或更多语言的多语言场景下CSD的有效性仍缺乏系统探索。解决方案的关键在于：通过在英文、日文、韩文和中文四种语言之间进行句子级别的多语言代码切换指令微调（instruction tuning），验证其在多语言理解任务（以Belebele评测集为基准）中的有效性。实验表明，这种简单的句子级多语言CSD策略能一致提升所有四种语言上的平均多语言性能，证明了多语言代码切换在超越双语场景下的可行性与有效性。
链接: https://arxiv.org/abs/2605.29414
作者: Shunta Asano,Jeonghun Baek,Toshihiko Yamasaki
机构: The University of Tokyo (东京大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Recent studies have shown that code-switching data (CSD), in which multiple languages are mixed within the same context, can improve cross-lingual transfer and multilingual alignment in large language models (LLMs). However, existing studies primarily focus on bilingual transfer between English and a target language, leaving multilingual settings involving three or more languages largely unexplored. In this work, we investigate multilingual code-switching instruction tuning across four languages: English, Japanese, Korean, and Chinese. We evaluate multilingual understanding on Belebele. Our experiments show that simple sentence-level multilingual CSD consistently improves average multilingual performance across all four languages, indicating that multilingual code-switching can be effective beyond bilingual transfer settings.
[NLP-105] Revisiting Observation Reduction for Web Agents : Comprehensive Evaluation with a Lightweight Framework
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）驱动的网络代理中HTML观测信息过长导致的代理延迟问题，同时确保任务性能不受影响。现有多种HTML缩减方法被提出，但缺乏对整体代理延迟与性能之间权衡的有效评估手段，主要瓶颈在于端到端评估成本极高。为应对这一挑战，作者提出一种轻量级评估框架，基于最小失败集（Minimal Failure Set, MFS）——即移除后会导致任务失败的最小HTML元素集合——定义“覆盖率”作为代理指标，衡量缩减方法是否完整保留MFS。该指标无需访问网页或调用LLM推理，显著降低评估开销（超过100倍加速）。实验验证了覆盖率与端到端成功率高度相关，进而利用此框架发现：提取式HTML缩减方法需依赖高计算成本或领域特定优化才能兼顾延迟降低与性能保持；在此基础上，作者基于MFS训练数据优化剪枝程序，在WorkArena L1上实现2.2倍的单步延迟加速且保持84%原始成功率，在WebLinx上实现3.1倍加速并保持89%成功率，成为高效且可靠的解决方案。
链接: https://arxiv.org/abs/2605.29397
作者: Masafumi Enomoto,Ryoma Obara,Haochen Zhang,Masafumi Oyamada
机构: NEC Corporation
类目: Computation and Language (cs.CL)
备注: 22 pages, 8 figures, 4 tables
Abstract:HTML observations in LLM-based web agents are extremely long, and while many reduction methods have been proposed, it remains unclear which methods reduce overall agent latency while maintaining performance. The main obstacle is the high cost of end-to-end evaluation: in our experiments, evaluating 11 methods across 32 configurations on 33 tasks of WorkArena L1 required 232.4 cumulative hours. To address this, we propose a lightweight evaluation framework based on the Minimal Failure Set (MFS), the minimal set of HTML elements whose removal causes task failure. We define coverage as the fraction of instances in which a reduction method fully retains the MFS, which serves as a proxy metric that requires neither web access nor LLM inference. We validate that coverage strongly correlates with end-to-end success rate, with over 100 \times speedup in cumulative evaluation time on both benchmarks. Using this framework, we find that extractive HTML reduction methods require either high computation cost or domain-specific optimization to reduce agent latency while maintaining performance. Building on this, we optimize a pruning program on MFS training data, achieving 2.2 \times faster per-step latency on WorkArena L1 while retaining 84% of the original success rate, and 3.1 \times faster on WebLinx while retaining 89%.
[NLP-106] BrahmicTokenizer-131K: An Indic-Capable Drop-In Replacement for o200k_base
【速读】： 该论文旨在解决多语言大模型训练中印度语系（Brahmic）文本压缩效率不足的问题，即现有主流分词器（如o200k_base、Tekken/Sarvam-m）在处理印度语系文字时存在显著的token浪费现象。其核心解决方案是提出BrahmicTokenizer-131K——一个131,072词汇量的字节级BPE分词器，通过两阶段精修策略实现：(1) 脚本裁剪（script-prune crop），从200,019个token中移除九种非目标书写系统；(2) 基于线性规划分配的手术式重构（surgical retrofit），向九个婆罗米文Unicode区块精准注入2,372个新词元，从而大幅提升印度语系文本压缩比（如奥里亚语达4.31倍压缩率），同时保持英语、欧洲语言及代码等非印度语内容的分词性能与o200k_base相当，甚至在多个基准测试（HumanEval、MBPP、GSM8K）上优于现有方案。此方法实现了在固定词汇量下对多种语言类型（包括婆罗米文、英文、欧盟语言、代码和数学）的统一高效支持，填补了此前“婆罗米压缩差距”的空白。
链接: https://arxiv.org/abs/2605.29379
作者: Rohan Shravan
机构: The School of AI (印度的AI学校)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 24 pages, 15 tables, 3 code listings. Tokenizer artifact, verification scripts, and reproduction code at this https URL and this https URL
Abstract:We present BrahmicTokenizer-131K, a 131,072-vocabulary byte-level BPE tokenizer that closes the Brahmic compression gap at the 131K-vocabulary class while preserving the English, EU-language, and code compression of OpenAI’s o200k_base. We construct it through a two-stage retrofit: (1) a script-prune crop that reduces 200,019 tokens to 131,072 by removing nine out-of-scope writing systems, and (2) a surgical retrofit of 2,372 corpus-dead vocabulary slots determined by linear-programming allocation across nine Brahmic Unicode blocks. The pre-tokenizer, decoder, and inherited merge rules are unchanged from o200k_base, making BrahmicTokenizer-131K a drop-in replacement at the tokenizer interface. On 27 million documents of public Indic pretraining text (2.84 billion words, 46.21 GB), BrahmicTokenizer-131K produces 26.7% fewer tokens than Mistral-Nemo Tekken / Sarvam-m at the same vocabulary budget, with per-language savings of 15.79% (Tamil) to 76.79% (Odia, a 4.31x compression ratio). The Odia advantage is mechanistically explained by Tekken/Sarvam-m containing zero Oriya-block tokens; our surgery added 725. On non-Indic content, BrahmicTokenizer-131K matches o200k_base’s English fertility (1.235 vs 1.232 tokens/word) and beats Tekken/Sarvam-m by 4.0-14.2% on HumanEval, MBPP, and GSM8K. Across our 14-tokenizer benchmark, it is the only tokenizer simultaneously competitive on Brahmic, English, EU, code, and math at the 131K budget. Specialist tokenizers at other vocab classes (Sarvam-30B, Sarvam-1, MUTANT-Indic) achieve better Indic compression at the cost of non-Indic performance: Sarvam-1’s English fertility is 15.9% worse and its code/math compression 26-33% worse than ours. We release the artifact under Apache 2.0 at this https URL. Comments: 24 pages, 15 tables, 3 code listings. Tokenizer artifact, verification scripts, and reproduction code at this https URL and this https URL Subjects: Computation and Language (cs.CL); Machine Learning (cs.LG) ACMclasses: I.2.7 Cite as: arXiv:2605.29379 [cs.CL] (or arXiv:2605.29379v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2605.29379 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-107] SURGENT: A Surgical Multi-Agent Assistance System Across the Perioperative Workflow
【速读】： 该论文旨在解决现代外科护理中智能系统在整合患者病历、支持协作决策以及提供全程围术期可解释性推理方面的不足。现有基于网络的大型语言模型（Large Language Models, LLMs）因输入长度限制、记忆管理不完整及可追溯性差等问题，难以满足外科场景的复杂需求。其解决方案的关键在于提出SURGENT——一个外科多智能体辅助系统，融合了思维树（Tree-of-Thought）规划器、多部门协作代理以及基于临床指南和生物医学文献的检索增强推理机制；同时设计了一种新颖的记忆架构，能够协同管理长期患者历史与短期工作摘要，从而实现更全面、情境化且一致的推理能力。实验表明，SURGENT在五个关键围术期任务中均优于基线LLMs和现有医疗多智能体框架，推荐结果更贴合个体患者病史；消融研究进一步验证了本地部署DeepSeek作为骨干模型的优势，保障隐私安全且无需依赖中心化服务，为构建可信、公平、安全的智能外科辅助系统提供了切实可行的新路径。
链接: https://arxiv.org/abs/2605.29368
作者: Dongsheng Shi,Yue Li,Xin Yi,Yongyi Cui,Huawei Feng,Linlin Wang
机构: East China Normal University (华东师范大学); City University of Hong Kong (香港城市大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: preprint
Abstract:The intricate nature of modern surgical care necessitates intelligent systems that can synthesize extensive patient records, support collaborative decision-making, and provide transparent, auditable reasoning across the entire perioperative workflow. Although web-based Large Language Models (LLMs) possess advanced reasoning capabilities, they are ill-equipped for surgical applications due to critical limitations: input length constraints, incomplete memory management, and limited traceability. To address this issue, we present SURGENT, a surgical multi-agent assistance system that combines a Tree-of-Thought planner, multi-department collaboration agents, and retrieval-augmented reasoning with clinical guidelines and biomedical literature. SURGENT features a novel memory design that manages both long-term patient histories and short-term working summaries, enabling more complete, contextualized, and consistent reasoning. Experimental evaluations across five key perioperative tasks - case analysis, surgical plan simulation, safety monitoring, complication risk assessment, and rehabilitation guidance - show that SURGENT outperforms baseline LLMs and existing medical multi-agent frameworks, yielding recommendations more closely aligned with patient histories. Ablation studies further highlight the advantage of DeepSeek as a locally deployable backbone model, enabling privacy-preserving deployment without reliance on centralized services. These results position SURGENT as a practical and trustworthy advancement toward intelligent, equitable, and secure surgical assistance systems.
[NLP-108] Attention Asymmetry in AI Layoff Discourse on X: A Computational Analysis of Capital vs Labour Amplification
【速读】： 该论文试图解决的问题是：在人工智能（AI）驱动的职场重组导致员工失业时，不同群体在社交媒体平台X（原Twitter）上围绕这一议题展开的两种截然不同的对话——资本方（如科技高管和AI研究人员）强调生产力提升、转型机遇，而劳工方（如被裁员工和劳工批评者）则聚焦于失业、不确定性与恐惧——哪种声音获得更高的传播广度（reach）。解决方案的关键在于通过多阶段实证研究设计，系统性地量化并比较两类话语的传播差异。研究发现，在基于账号的收集方法下，资本话语的平均放大倍数显著高于劳工话语（3.12倍，p<0.000003），且即使校正了粉丝数量的影响后，这种不对称性依然存在（2.69倍，p<0.000009），表明平台机制本身可能加剧了话语不平等。作者提出“放大比”（Amplification Ratio）和“放大归一化指数”（Amplification Normalisation Index）作为衡量平台层面话语不平等的可操作指标，并指出该现象可能与X平台特有的账号驱动式传播架构有关，具有重要的方法论启示意义。
链接: https://arxiv.org/abs/2605.29367
作者: Joy Bose
机构: 未知
类目: Computation and Language (cs.CL); Computers and Society (cs.CY); Social and Information Networks (cs.SI)
备注: 18 pages, 3 figures, 9 tables
Abstract:When workers lose jobs to AI-driven restructuring, two very different conversations happen on X (formerly Twitter) at the same time. Tech executives and AI researchers talk about productivity, transformation, and opportunity. Laid-off workers and labour critics talk about job loss, uncertainty, and fear. This paper asks a simple question: which conversation gets more reach? We report three studies using two collection methods and 763 tweets from 20 named public accounts. Study 1 used keyword-based collection (n=392) and found no significant difference between corpora (p=0.891), revealing that keyword search is too noisy for this task. Study 2 used account-based collection (n=96) and found a 3.12x mean amplification advantage for capital discourse over labour discourse (p=0.000003, Cohen’s d=0.555). Study 3 combined both methods (n=763) and confirmed the finding at 4.18x mean and 10.77x median amplification ratio (p0.000001). Critically, after normalising for follower count, the asymmetry persists at 2.69x (p=0.000009, Cohen’s d=0.491), demonstrating that the effect is not simply a consequence of capital accounts having larger audiences. The finding is robust across all tested amplification metric weightings. We introduce the Amplification Ratio and Amplification Normalisation Index as simple metrics for measuring platform-level discourse inequality. A cross-platform replication on Reddit (n=647 posts) did not replicate the finding, suggesting the asymmetry may be specific to X’s account-based amplification architecture. We discuss the methodological implications for cross-platform discourse analysis.
[NLP-109] Casual as an Anchor: Resolving Supervision Misalignment in Formality Transfer Dataset
【速读】： 该论文试图解决现有形式转移（formality transfer）任务中因监督设计缺陷导致的模型生成伪正式文本的问题。具体而言，现有基准（如GYAFC）采用二元人类重写标签，仅反映相对风格变化而非绝对的形式性概念，使得模型学习到的是满足标签要求的“伪正式”输出，而非真正符合人类对正式语言认知的表达。解决方案的关键在于重新将形式性视为一个连续维度而非二元属性，并引入三层次谱系（非正式、随意、正式），其中“随意”作为明确的中间状态以澄清监督信号。基于此框架，作者构建了3LF数据集，提供跨三个层级的平行监督信息，实验证明该设计显著减少从非正式到正式转换中的失败率，并提升与人类感知的一致性，例如GPT-4.1-nano在该方向的F1分数从0.06提升至0.88，且这些改进无法通过提示学习（in-context learning）实现。
链接: https://arxiv.org/abs/2605.29365
作者: Hyojeong Yu,Hyukhun Koh,Minsung Kim,Kyomin Jung
机构: Seoul National University (首尔国立大学)
类目: Computation and Language (cs.CL)
备注: HEAL@CHI 2026 Workshop Paper
Abstract:Formality transfer is commonly framed as a symmetric bidirectional task between informal and formal registers. We argue that this framing conceals a supervision design flaw in existing benchmarks such as GYAFC: binary human rewrites encode relative stylistic shifts rather than absolute human notions of formality. Consequently, models learn to generate pseudo-formal outputs that satisfy benchmark labels while failing to produce genuinely formal language. We quantify this misalignment by re-evaluating benchmark formal labels under a human-aligned definition of formality, revealing substantial discrepancies that propagate to consistent informal-to-formal failures across model families. To address this issue, we reconceptualize formality transfer as a graded dimension rather than a binary attribute. We introduce a three-level spectrum: informal, casual, and formal, where casual serves as an explicit intermediate state that clarifies supervision signals. Based on this framework, we introduce 3LF, a dataset providing parallel supervision across all three levels. Training on 3LF substantially reduces informal-to-formal failures and improves alignment with human perception. For example, GPT-4.1-nano improves from 0.06 to 0.88 F1 in the informal-to- formal direction despite 3LF being significantly smaller than GYAFC. We further demonstrate that these gains cannot be reproduced through in-context learning alone and provide qualitative analyses of ambiguity-driven errors and meaning distortions. Overall, our findings demonstrate how supervision design shapes stylistic alignment and highlight the importance of alignment-aware benchmark construction in controllable text generation. Comments: HEAL@CHI 2026 Workshop Paper Subjects: Computation and Language (cs.CL) Cite as: arXiv:2605.29365 [cs.CL] (or arXiv:2605.29365v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2605.29365 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Hyojeong Yu [view email] [v1] Thu, 28 May 2026 05:07:02 UTC (492 KB) Full-text links: Access Paper: View a PDF of the paper titled Casual as an Anchor: Resolving Supervision Misalignment in Formality Transfer Dataset, by Hyojeong Yu and 3 other authorsView PDFHTML (experimental)TeX Source view license Current browse context: cs.CL prev | next new | recent | 2026-05 Change to browse by: cs References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); About Help contact arXivClick here to contact arXiv Contact subscribe to arXiv mailingsClick here to subscribe Subscribe Copyright Privacy Policy Web Accessibility Assistance arXiv Operational Status
[NLP-110] Draft-OPD: On-Policy Distillation for Speculative Draft Models
【速读】： 该论文旨在解决推测解码（speculative decoding）中轻量级草稿模型（draft model）性能提升受限的问题，即传统监督微调（SFT）方法在训练草稿模型时存在“离线到推理不匹配”（offline-to-inference mismatch）现象：SFT使用固定的目标模型生成轨迹进行训练，但在实际推测解码过程中，草稿模型需基于自身策略生成序列，导致训练与推理分布不一致，从而使得草稿模型的接受长度（acceptance length）迅速达到瓶颈。解决方案的关键在于提出Draft-OPD（On-Policy Distillation with Draft Replay）机制，其核心创新是通过目标模型辅助的回放策略，在验证暴露的错误位置重新采样草稿行为，使草稿模型能够从被接受和拒绝的提案中同时学习，聚焦于限制推测接受率的关键错误区域，从而实现更有效的在线策略蒸馏（on-policy distillation）。实验表明，该方法在多种任务上实现了超过5倍的无损加速，相比EAGLE-3和DFlash分别提升了23%和13%。
链接: https://arxiv.org/abs/2605.29343
作者: Haodi Lei,Yafy Li,Haoran Zhang,Shunkai Zhang,Qianjia Cheng,Xiaoye Qu,Ganqu Cui,Bowen Zhou,Ning Ding,Yun Luo,Yu Cheng
机构: Shanghai Jiao Tong University (上海交通大学); Shanghai AI Laboratory (上海人工智能实验室); Tsinghua University (清华大学); The Chinese University of Hong Kong (香港中文大学); Peking University (北京大学); Zhejiang University (浙江大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Speculative decoding accelerates large language model inference by pairing a target model with a lightweight draft model whose proposed tokens are verified in parallel. A common way to build draft models, like EAGLE3 or DFlash is supervised fine-tuning (SFT) on target-generated trajectories. However, we observe that SFT quickly plateaus: the draft model’s acceptance length on test data stops improving. The reason is an offline-to-inference mismatch: In SFT, the drafter learns from fixed target-generated trajectories, whereas during speculative decoding it is evaluated on blocks proposed under its own policy. This motivates on-policy distillation (OPD), where the target model supervises the drafter on draft-induced states. Yet OPD remains difficult for draft models, as they cannot reliably roll out complete sequences independently, whereas target-assisted generation makes the collected sequences follow the target distribution and thus eliminates the on-policy signal. We therefore propose Draft-OPD, which uses target-assisted rollout for stable continuations and replays drafting from the verification-exposed error positions. This allows the drafter to learn from target feedback on both accepted and rejected proposals, focusing training on the draft-induced errors that limit speculative acceptance. Experiments show that Draft-OPD achieves over 5\times lossless acceleration for thinking models across diverse tasks, improving over EAGLE-3 and DFlash by 23% and 13%.
[NLP-111] WorldMemArena: Evaluating Multimodal Agent Memory Through Action-World Interaction
【速读】： 该论文试图解决多模态大语言模型作为长期代理（long-horizon agents）时，记忆系统无法有效追踪动态世界、更新过时信息并适时调用证据的问题。现有基准测试仅评估静态对话中的回忆能力，将记忆简化为单一的最终任务准确率，并将视觉观察降维为文本描述，导致难以定位记忆失败发生在编写、维护、检索还是使用阶段。解决方案的关键在于提出一个可观察的“动作-世界交互循环”（Action-World Interaction Loop），定义了记忆的四个阶段生命周期，并构建了WorldMemArena基准：包含400个多会话多模态任务，涵盖终身演化（Lifelong Evolution）和代理执行（Agentic Execution）两个维度，附带黄金标注的记忆点、更新、干扰项和证据链，实现对长上下文手动设计（如RAG和外部记忆系统）与自管理记忆代理（harness-based）的首次直接对比。实验表明：（1）更好的记忆写作和存储并不等价于更高性能；（2）多模态记忆仍难以充分利用视觉证据；（3）系统在不同领域间不稳定且在真实代理轨迹中性能下降；（4）自管理记忆更灵活但成本高且可靠性低。
链接: https://arxiv.org/abs/2605.29341
作者: Chengzhi Liu,Yuzhe Yang,Sophia Xiao Pu,Yepeng Liu,Lin Long,Yichen Guo,Nuo Chen,Zhaotian Weng,Elena Kochkina,Simerjot Kaur,Charese Smiley,Xiaomo Liu,James Zou,Sheng Liu,Yuheng Bu,Songyou Peng,Xin Eric Wang
机构: 未知
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
备注: 25 pages, 8 figures
Abstract:Multimodal large language models are increasingly deployed as long-horizon agents, where memory must do more than recall: it must track an evolving world, revise what has gone stale, and surface the right evidence at decision time. Existing benchmarks measure recall over static dialogue, collapse memory into a single end-of-task accuracy, and reduce visual observations to captions, leaving us unable to localize failures to writing, maintenance, retrieval, or use. The rise of agent harnesses that author their own memory sharpens this gap, since we have no principled way to compare hand-designed pipelines with self-managing alternatives. To close these gaps, we formulate multimodal agent memory as an Action-World Interaction Loop with an observable four-stage lifecycle, and instantiate it in WorldMemArena: 400 multi-session multimodal tasks spanning Lifelong Evolution (evolving personal and task states) and Agentic Execution (memory from real observations, actions, and feedback), annotated with gold memory points, updates, distractors, and evidence chains for stage-level diagnosis. This enables the first head-to-head comparison of long-context, manually designed (RAG and external memory systems), and harness-based memory agents. Results show that: (1) better memory writing and storage do not guarantee better performance; (2) multimodal memory still struggles to fully use visual evidence; (3) systems are unstable across domains and degrade on realistic agentic trajectories; and (4) harness memory is more flexible but remains costly and less reliable.
[NLP-112] A Study on Question-Answer Dataset for LLM Safety Evaluation with a Focus on Illegal Activities
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）在安全评估中缺乏高质量、针对性的问答数据集问题，特别是针对非法活动相关场景的评测能力不足。解决方案的关键在于基于对AnswerCarefully数据集的手动分析，引入了额外的信息维度、构建问答样本的方法以及一套用于评估LLM生成响应质量的评分标准（rubric），从而为LLM安全性提供可量化、可复现的评估框架，并支持“JAI-Trust”项目的研究目标。
链接: https://arxiv.org/abs/2605.29340
作者: Kenji Imamura,Masao Ideuchi,Atsushi Fujita
机构: National Institute of Information and Communications Technology (日本情報通信研究機構)
类目: Computation and Language (cs.CL)
备注: 10 pages, 1 figure
Abstract:In this paper, we discuss question-answer dataset for LLM safety evaluation, with a focus on illegal activities. Specifically, on the basis of manual analysis of AnswerCarefully, we introduce several additional information, methods for creating question-answer examples, and a rubric for evaluating LLM-generated responses. The outcomes of this study are intended to be shared with the “JAI-Trust” project.
[NLP-113] Enhancing Factuality through Consensus and Consistency in Summarization Using Minimum Bayes Risk Decoding ACL2026
【速读】： 该论文旨在解决生成式摘要中事实准确性（factuality）不足的问题，即摘要内容与源文档的一致性难以保障。现有方法如重排序（reranking）仅依赖源文档进行筛选，导致生成结果不可靠。其解决方案的关键在于提出ConSUM框架，通过两个核心因素对候选摘要进行重排序：一是与源文档的一致性（使用事实感知指标衡量），二是候选摘要之间的共识度（利用最小贝叶斯风险MBR解码机制从多个生成结果中提取）。这一双维度评估策略显著提升了摘要的事实准确性和整体质量，实验表明其性能优于现有方法，且在人工评估中更受青睐。
链接: https://arxiv.org/abs/2605.29336
作者: Riza Setiawan Soetedjo,Yusuke Sakai,Hidetaka Kamigaito,Jingun Kwon,Manabu Okumura,Taro Watanabe
机构: Nara Institute of Science and Technology (NAIST); Chungnam National University; Institute of Science Tokyo
类目: Computation and Language (cs.CL)
备注: Accepted to ACL 2026 Findings
Abstract:Improving the quality of model-generated summaries, especially factuality, the accuracy of a summary with respect to its source content, remains a challenge. While reranking could select the optimal output from multiple generated candidates, it is limited to only using the source as guidance, resulting in unreliable summaries. To address this limitation, we propose ConSUM that reranks candidate summaries by considering two factors: consistency to the source document and consensus among the other candidates. Consensus is established using Minimum Bayes Risk (MBR) decoding over the set of generated summaries, while ensuring consistency by employing factuality-aware metrics that compare the summary against the source. Rigorous testing demonstrates that our system is competitive with existing methods, with human evaluations further confirming that its generated summaries are preferred over those from other systems. Our code is available at this https URL .
[NLP-114] Reasoning -preserved Efficient Distillation of Large Language Models via Activation-aware Initialization
【速读】： 该论文试图解决的问题是：高效蒸馏（Efficient Distillation, EDistill）方法在压缩大语言模型（LLM）时虽能保持通用能力（general ability）的先进性能，但会导致多步推理能力（multi-step reasoning ability）显著下降，即“推理崩溃”（reasoning collapse）。其解决方案的关键在于识别出问题的几何根源——基于宽度缩减投影矩阵的EDistill方法会引发有效秩（effective rank, eRank）坍缩，导致隐藏表示的区分度丧失。为此，作者提出RED（Reasoning-preserved Efficient Distillation），通过引入激活感知初始化（activation-aware initialization），将投影矩阵初始化为通道选择矩阵，从而理论上缓解eRank坍缩，显著恢复推理能力，同时保持高训练效率和通用能力的SOTA表现。
链接: https://arxiv.org/abs/2605.29327
作者: Junlin He,Yihong Tang,Tong Nie,Guilong Li,Binyu Yang,Jinxiao Du,Lijun Sun,Wei Ma
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Efficient Distillation (EDistill) compresses large language models (LLMs) by structured pruning parameters and tuning lightweight modules with high training efficiency. Although these EDistilled LLMs achieve state-of-the-art (SOTA) performance on general ability benchmarks relative to similarly sized LLMs, we identify a severe degradation in their multi-step reasoning ability, which we term reasoning collapse. We systematically analyze the geometric origins of reasoning collapse and show that the SOTA EDistill method based on width-reducing projection matrices suffers from eRank collapse, in which the effective rank (eRank) of hidden representations drops. We theoretically explain how singular values of randomly initialized projection matrices become unevenly distributed, leading to eRank collapse and thus token indistinguishability. To address this issue, we propose RED (Reasoning-preserved Efficient Distillation) for LLMs, which introduces activation-aware initialization to initialize projection matrices as channel-selection matrices, thus theoretically mitigating eRank collapse. Experiments on Llama and Qwen series demonstrate that RED substantially recovers reasoning while maintaining high training efficiency and SOTA general ability.
[NLP-115] STAMP: Training Explicit Memory for Mobile GUI Agents in Controllable and Scalable Virtual Environments
【速读】： 该论文试图解决移动图形用户界面（GUI）智能体在长时程任务中因记忆缺失而导致性能下降的问题。其核心挑战在于有限的上下文窗口与高Token密度截图之间的矛盾：智能体为节省上下文空间不得不逐步丢弃早期视觉信息，从而永久丢失关键的瞬时状态。现有以动作为中心的数据集无法教会智能体何时以及存储什么信息，而静态现实数据的扩充成本高昂且缺乏交互验证。解决方案的关键在于提出STAMP框架，通过可控虚拟环境训练显式记忆机制——在合成任务中程序化注入确定性的记忆变量，精确控制需记忆的内容、编码时机及检索时刻，从而大规模生成可验证的监督数据，并利用环境驱动的奖励反馈实现在线强化学习。实验表明，基于该框架训练的Stamp-GUI智能体在新提出的Memory-World基准上达到当前最优性能，显著提升了记忆准确性与任务鲁棒性，同时保持了强大的通用移动导航能力。
链接: https://arxiv.org/abs/2605.29324
作者: Junyang Wang,Haiyang Xu,Xi Zhang,Zhaoqing Zhu,Ming Yan,Jieping Ye,Jitao Sang
机构: Tongyi AI Lab, Alibaba Group; Beijing Jiaotong University
类目: Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注: 24 pages, 4figures, 21 tables
Abstract:Mobile GUI agents excel at immediate reactive control but frequently fail in realistic, long-horizon tasks that require memory. This failure stems from a fundamental conflict between limited context windows and token-heavy screenshots. To save the limited context, agents must progressively discard older visual history, permanently losing crucial transient information. Furthermore, existing action-centric datasets fail to teach agents what or when to explicitly memorize, and augmenting static real-world data is prohibitively expensive and lacks interactive verification. To resolve this, we present STAMP, a framework that trains explicit memory in mobile agents through controllable virtual environments, where deterministic memory variables are programmatically injected into synthesized tasks to control what must be memorized, when it should be encoded, and when it must later be retrieved, thereby producing verifiable supervised data at scale and enabling online reinforcement learning through environment-driven reward feedback. Evaluated on our newly introduced Memory-World benchmark, the resulting Stamp-GUI agent achieves state-of-the-art performance among GUI-specialized models and sets a new high watermark on our Memory-World benchmark, demonstrating exceptional memory accuracy and task resilience while maintaining strong general mobile navigation capabilities.
[NLP-116] Rethinking Stepwise Model Routing: A Cost-Efficient Table Reasoning Perspective EMNLP2026
【速读】： 该论文试图解决大型推理模型（Large Reasoning Models, LRMs）在表格推理任务中因冗长的推理链条而导致的高推理成本问题。现有方法中的逐步模型路由（stepwise model routing）虽能通过动态分配推理步骤至不同规模模型来缓解此问题，但在表格推理场景下仍存在不足：其未能区分推理步骤中两类具有不同不确定性分布的token——基于表格结构的表元标记（table tokens，如单元格值和表头）与自然语言描述的文本标记（text tokens）。这两类token的不确定性均与下一推理步骤出错的风险相关，但现有方法未分别建模，导致路由决策次优。解决方案的关键在于提出EcoTab框架，该框架在每一步推理中独立估计表元和文本标记的不确定性，并将其映射为小模型在下一步出错的风险，最终融合两类风险进行路由决策，从而实现更高效且准确的表格推理。
链接: https://arxiv.org/abs/2605.29319
作者: Shenghao Ye,Yuxiang Wang,Yu Guo,Dong Jin,Shuangwu Chen,Jian Yang
机构: University of Science and Technology of China; The University of Melbourne; Institute of Artificial Intelligence, Hefei Comprehensive National Science Center
类目: Computation and Language (cs.CL)
备注: 17pages, 15 figures, submitted to EMNLP 2026
Abstract:Large Reasoning Models (LRMs) achieve strong performance on table reasoning tasks but incur substantial inference cost due to long reasoning traces. Stepwise model routing mitigates this issue by dynamically assigning reasoning steps to smaller or larger models. However, stepwise model routing for table reasoning remains underexplored. Through empirical analysis, we find that reasoning steps involving tables contain two types of tokens with distinct uncertainty distributions: table tokens grounded in table structure, such as cell values and headers, and text tokens representing surrounding natural-language reasoning. The uncertainty of both token types is correlated with the risk that the model makes an error in the next reasoning step. However, existing methods fail to model them separately, leading to suboptimal routing decisions. To address this, we propose EcoTab, a table-aware stepwise routing framework for efficient table reasoning. At each reasoning step, EcoTab separately estimates the uncertainties of table tokens and text tokens, maps them to next-step failure risks for the small model, and combines the two risks for routing. Experiments on multiple table reasoning benchmarks show that EcoTab consistently outperforms strong baselines and achieves a better balance between accuracy and efficiency.
[NLP-117] FoRA: Fisher-orthogonal Rank Adaptation for Parameter-Efficient Fine-Tuning EMNLP2026
【速读】： 该论文试图解决参数高效微调（Parameter-efficient fine-tuning, PEFT）中训练参数数量过多的问题，特别是针对LoRA及其精度导向变体在降低可训练参数方面关注不足的现状。其解决方案的关键在于两个核心创新：一是通过单次遍历的对角Fisher分数（Diagonal Fisher Score，计算成本低于1%的训练成本）筛选任务相关的层，从而减少适配层的数量而非降低LoRA的秩（rank）；二是将LoRA的下投影层（down-projection）约束在Stiefel流形上，以保持列正交性和有效秩。这种组合策略使FoRA在仅使用一半LoRA参数预算时性能优于LoRA和DoRA，在四分之一AdaLoRA参数量下仅比其低0.7–0.8准确率点，并在多个大语言模型架构（LLaMA、Qwen3、Gemma）中验证了有效性。
链接: https://arxiv.org/abs/2605.29317
作者: Juneyoung Park,Seongbae Lee,Han-Sang Lee,Kyuho Lee,Minjae Kim,Seungheon Hyeon,Kiduk Kwon,Seongwan Kim,Jaeho Lee
机构: OptAI Inc; LG Uplus
类目: Computation and Language (cs.CL)
备注: EMNLP 2026
Abstract:Parameter-efficient fine-tuning(PEFT) has largely focused on LoRA and its accuracy-oriented variants, leaving the original goal of reducing trainable parameters has receivedcomparatively little attention. We introduce FoRA, which revisits this goal by reducing the number of adapted layers rather than adapter rank. FoRA selects task-informative layers via a single-pass diagonal Fisher score (under 1% of training cost) and trains the LoRA down-projection at selected layers on the Stiefel manifold, preserving column orthonormality and effective rank. FoRA consistently outperforms LoRA and DoRA at half their parameter budget, and falls within 0.7-0.8 accuracy points of AdaLoRA at one-quarter its parameter count, across five LLaMA-family backbones. Cross-architecture experiments on twelve backbones from the LLaMA, Qwen3, and Gemma families confirm consistent gains from 270M to 32B parameters. The two components combine super-additively: Fisher selection alone matches rank reduction at the same budget, while the Stiefel constraint provides the decisive additional gain.
[NLP-118] PatchBoard: Schema-Grounded State Mutation for Reliable and Auditable LLM Multi-Agent Collaboration
【速读】： 该论文试图解决大语言模型（Large Language Model, LLM）多智能体系统中因依赖自然语言对话或松散结构化共享内存而导致的中间状态难以验证、归属和审计的问题。其解决方案的关键在于提出一种基于模式（schema）的协作架构 PatchBoard，该架构用经过验证的 JSON Patch 变更替代智能体间的自然语言对话，从而实现对共享结构化状态的精确控制；其中，架构代理（Architect agent）负责构建任务特定的模式与工作流规则，而确定性内核（deterministic kernel）则在事务性提交前对每个状态变更进行模式约束、角色写入契约及运行时不变量的验证，显著提升了系统的可验证性、可追溯性和效率。
链接: https://arxiv.org/abs/2605.29313
作者: Shuyu Zhang,Yaqi Shi,Lu Wang
机构: Xidian University (西安电子科技大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:LLM multi-agent systems often coordinate through natural-language dialogue or loosely structured shared memory, making intermediate state difficult to validate, attribute, and audit. We introduce PatchBoard, a schema-grounded collaboration architecture that replaces inter-agent dialogue with validated JSON Patch mutations over a shared structured state. An Architect agent constructs a task-specific schema and workflow rules, while a deterministic kernel validates each proposed state mutation against schema constraints, role-specific write contracts, and runtime invariants before committing it transactionally. On 630 matched ALFWorld episodes, PatchBoard achieves an 84.6% success rate, compared with 30.8% for LangGraph and 61.6% for Flock, while reducing tokens per successful task to 45.5k, compared with 368.3k and 64.2k, respectively.
[NLP-119] Rubric-Guided Process Reward for Stepwise Model Routing EMNLP2026
【速读】： 该论文试图解决的问题是：当前基于强化学习的步骤式模型路由方法在训练路由器时仅使用最终答案正确性的奖励信号，而无法有效评估中间路由决策的质量，从而限制了模型性能和泛化能力。解决方案的关键在于提出 RoRo（rubric-guided process reward framework），其核心创新包括：构建基于结果、成本和过程质量的偏好对以收集多样化路由轨迹；通过交替优化训练一个 Rubricor（生成查询特定评估标准）和一个 Judge（依据该标准评分路由轨迹）来生成过程奖励；并将过程奖励与结果奖励结合，利用 GRPO（Generalized Reward Policy Optimization）优化路由策略。这一机制显著提升了路由决策的准确性与效率，在多个推理基准测试中均优于现有强基线方法。
链接: https://arxiv.org/abs/2605.29310
作者: Shenghao Ye,Yu Guo,Zhengheng Li,Shuangwu Chen,Jian Yang
机构: University of Science and Technology of China; Southeast University; Institute of Artificial Intelligence, Hefei Comprehensive National Science Center
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 17 pages, 9 figures, submitted to EMNLP 2026
Abstract:Stepwise model routing improves the efficiency of Large Reasoning Models (LRMs) by assigning each reasoning step to a suitable model. Recent methods formulate routing as a sequential decision process and train the router with reinforcement learning. However, although they model routing as a process, they still supervise the router with outcome rewards. Such rewards only reflect final answer correctness and fail to evaluate intermediate routing decisions, which can weaken performance and generalization. To address this gap, we propose RoRo, a rubric-guided process reward framework for stepwise model routing. RoRo first collects diverse routing trajectories and constructs preference pairs based on outcome, cost, and process quality. It then trains a Rubricor to generate a query-specific evaluation rubric and a Judge to score routing trajectories under this rubric through alternating optimization. The resulting process rewards are combined with outcome rewards to optimize the routing policy via GRPO. Experiments on five reasoning benchmarks under both same-family and cross-family settings show that RoRo consistently outperforms strong baselines and achieves better accuracy and cost trade-offs.
[NLP-120] MusTBENCH: Benchmarking and Advancing Temporal Grounding in Music LLM s
【速读】： 该论文试图解决当前大型音频-语言模型（Large Audio-Language Models, LALMs）在音乐理解任务中缺乏精确时间定位能力的问题，即模型的回答是否准确对应音频中的特定时间区间尚不明确。这一问题尤为关键，因为音乐中的关键信息往往表现为时序局部事件（如乐器进入或节奏转换）。解决方案的关键在于提出MusT（a novel four-stage temporal optimization recipe），其包括四个阶段：音乐编码器适配、大语言模型（LLM）适配、LLM监督微调以及基于强化学习（RL-based）的优化，从而系统性提升模型的时间定位精度；同时构建了MusTBENCH基准测试集，通过五个时序定位问答任务对模型进行评估，实验证明现有LALMs在时间定位上表现不佳，而MusT显著优于强基线，确立了时间定位作为当前LALMs的关键缺失能力，并为未来研究提供了挑战性基准。
链接: https://arxiv.org/abs/2605.29300
作者: Daeyong Kwon,Qiyu Wu,Shinobu Kuriya,Junghyun Koo,Shuyang Cui,Zhi Zhong,Wei-Hsiang Liao,Hiromi Wakaki,Yuki Mitsufuji
机构: Sony Group Corporation(索尼集团); Sony AI(索尼人工智能)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Sound (cs.SD)
备注:
Abstract:Recent Large Audio-Language Models (LALMs) have demonstrated promising abilities in understanding musical content. However, whether their responses are grounded in the correct temporal regions of the audio remains underexplored. This limitation is particularly critical for music understanding, where key information often occurs as temporally localized events, such as instrument entries and rhythmic transitions. To address this gap, we introduce MusTBENCH, a music-expert-validated benchmark designed to evaluate temporal grounding in LALMs through five temporally grounded question-answering tasks. To further improve temporal grounding in existing models, we propose MusT, a novel four-stage temporal optimization recipe spanning music encoder adaptation, LLM adaptation, LLM supervised fine-tuning, and RL-based optimization. Experiments on MusTBENCH show that existing LALMs struggle with precise temporal grounding, while MusT brings significant improvements over strong baselines. These results establish temporal grounding as a key missing capability in current LALMs and position MusTBENCH as a challenging benchmark for future research in temporally grounded music understanding.
[NLP-121] Accommodation Goes Both Ways: Studying Linguistic Convergence Between Humans and Language Models
【速读】： 该论文试图解决的问题是：随着大语言模型（LLM）日益融入日常生活，其存在将如何影响人类的语言行为，特别是人类与LLM在多轮对话中是否存在语言风格的相互适应（即语言趋同）。解决方案的关键在于提出并应用一种非对称趋同度量方法，基于WildChat这一真实世界ChatGPT对话语料库，量化分析人类与LLM在功能词和开放类词汇特征上的语言风格趋同程度。研究发现，LLM显著过度向用户靠拢其语言风格，而人类对LLM的语言适应程度则与人类之间的自然对话基线一致，表明这种趋同具有明显的不对称性：LLM表现出强烈的风格拟合倾向，而人类则保持了与他人互动时的稳定语言适应模式。
链接: https://arxiv.org/abs/2605.29278
作者: Terra Blevins
机构: Northeastern University (东北大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:As LLMs become increasingly integrated into daily life, understanding how their presence will shape human linguistic behavior is an open question. We present a large-scale study of linguistic convergence in human-LLM dialogue, examining how humans and LLMs accommodate each other’s linguistic style during multi-turn conversations. Using an asymmetric convergence metric on WildChat, a corpus of real-world ChatGPT transcripts, we find that while LLMs significantly overconverge toward their users on both function word and open-class features across eight languages, human convergence rates in this setting are broadly consistent with human-human baselines. These findings suggest that accommodation in human-LLM dialogue is asymmetric: while LLMs dramatically overfit to their users’ style, humans linguistically accommodate LLMs no differently than they would another person.
[NLP-122] Prompt-Level Reward Specifications for Open-Ended Post-Training
【速读】： 该论文试图解决开放性任务（如指令遵循、写作和决策支持）中奖励机制不明确的问题，即现有方法通常依赖事后标量评分或仅覆盖狭义可验证场景，无法显式表达局部需求、整体偏好和显式约束等多维质量标准。解决方案的关键在于提出一种提示级奖励规范框架（prompt-level reward specification framework），其核心创新是将奖励规范与奖励计算相分离：在离线阶段，仅基于提示构建可复用的任务自适应评分细则（rubrics）和可执行的硬约束检查器；在评分阶段，结合基于产出物锚定的细则分数、代码验证分数与独立全局评分，生成归一化的混合奖励，综合衡量要求满足度、整体质量和确定性约束。该方法无需人工偏好标注、参考答案或单独训练的奖励模型，在多个开放性基准上实现了离线奖励建模（RM-style）排序提升和在线强化学习优化，并通过消融实验验证了评分细则、全局评分和可执行验证提供的互补监督信号。
链接: https://arxiv.org/abs/2605.29275
作者: Zijun Weng,Xiaohui Hu,Shuangyong Song,Yongxiang Li,Kaidong Yu,Xuanjing Huang
机构: Fudan University (复旦大学); Xingchen AGI Lab, China Telecom Artificial Intelligence Technology (Beijing) Co., Ltd. (中国电信人工智能技术（北京）有限公司)
类目: Computation and Language (cs.CL)
备注: 39 pages, 4 figures, 16 tables
Abstract:Open-ended post-training benefits from rewards that make prompt-specific success conditions explicit, rather than relying only on post-hoc scalar scores. In instruction following, writing, and decision-support tasks, response quality depends on local requirements, holistic preferences, and explicit constraints, but existing reward methods often leave these criteria implicit or cover only narrowly verifiable cases. We propose a prompt-level reward specification framework that separates reward specification from reward computation. Given only prompts, our framework constructs reusable task-adaptive rubrics and executable hard-constraint checkers offline, making reward criteria explicit before training and reusable across rollouts. At scoring time, artifact-anchored rubric and code scores are combined with an independent global score for residual holistic quality, yielding a normalized hybrid reward over requirement satisfaction, holistic quality, and deterministic constraints. The framework requires no human preference annotations, reference answers, or a separately trained reward model. Experiments show that the resulting reward improves offline RM-style response ranking and supports online reinforcement learning across multiple open-ended benchmarks. Ablations further show that rubrics, global scoring, and executable verification provide complementary supervision.
[NLP-123] Learnable Assessment Skills for LLM -based Automated Scoring: Rubric Construction via Iterative Optimization
【速读】： 该论文试图解决的问题是：基于大语言模型（LLM）的自动评分方法虽然已接近人类水平，但在扩展到新任务时仍受限于上游阶段（如评分量表构建）需人工逐项配置，导致效率低下。其解决方案的关键在于提出“评估技能”（assessment skills）这一概念——即一种与具体题目无关、以自然语言形式表达的程序性知识，能够指导LLM在评分流程中特定阶段做出决策。作者聚焦于评分量表构建这一初始场景，设计了一个迭代框架，将技能分解为固定结构模板和可学习的通用规则，并通过LLM驱动的评分错误诊断与验证门控选择机制不断优化规则。该框架无需专家编写的评分量表，在所有10个ASAP-SAS题目上均显著提升评分性能，且常优于原始专家量表；跨题目迁移实验进一步表明，所学技能既包含可迁移的通用模式，也保留了题目特异性特征。
链接: https://arxiv.org/abs/2605.29274
作者: Yun Wang,Xin Xia,Xuansheng Wu,Xiaoming Zhai,Ninghao Liu
机构: University of Georgia(佐治亚大学); The Hong Kong Polytechnic University(香港理工大学)
类目: Computation and Language (cs.CL)
备注: 12 pages, 5 figures
Abstract:LLM-based automated scoring approaches near-human performance, but scaling to new tasks remains bottlenecked by the per-item human configuration of upstream stages such as rubric construction. Human experts bypass this bottleneck through evaluation heuristics developed over extensive practice. We ask whether LLMs can learn similar heuristics directly from scoring experience, and formalize this as the concept of assessment skills: item-independent natural-language procedural knowledge that guides LLMs through specific stages of the scoring workflow. Focusing on rubric construction as a first instantiation, we propose an iterative framework that decomposes a skill into a fixed scaffold and learnable item-agnostic rules, refining the rules through LLM-driven diagnosis of scoring errors and validation-gated selection. The framework requires no expert-written rubric. On all ten ASAP-SAS items, optimized skills substantially improve LLM-based scoring and frequently surpass the dataset-provided expert rubric. Cross-item transfer experiments further reveal that learned skills capture both generalizable and item-specific patterns.
[NLP-124] Compute Allocation in Evolutionary Search: From Depth-Breadth to Multi-Armed Bandits
【速读】： 该论文试图解决的问题是：在基于大语言模型（LLM）的进化搜索系统中，如何在固定LLM调用预算下最优分配计算资源，以及如何提升单次运行结果的可靠性。现有方法通常仅报告多次运行中的最佳结果，而忽略了运行间分布信息，导致对实际性能和稳定性缺乏准确评估。解决方案的关键在于识别出两个经验规律：一是“适应度-计算包络”（fitness-compute envelope），表明不同模型在有效浮点运算次数（FLOPs）上的能力排序趋于一致；二是“深度-宽度双线性拟合”（bilinear depth-breadth fit），其与任务特定交互相关，且受模型-任务能力阈值约束。基于此，作者提出BaSE（Bandit-based Self-Evolving）算法，采用多臂赌博机机制动态分配LLM调用到并行轨迹上，无需改变模型、提示或评估器即可在8个（模型，任务）组合中平均提升适应度12.3%，尤其在高方差场景下显著增强可靠性。
链接: https://arxiv.org/abs/2605.29268
作者: Sixue Xing,Haoyu He,Kerui Wu,Zhuo Yang,Haozheng Luo,Tianfan Fu,Aarthy Nagarajan
机构: University of Notre Dame; Northeastern University; University of Massachusetts Amherst; Southeast University; Northwestern University; Nanjing University; Shanghai Artificial Intelligence Laboratory
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:LLM-guided evolutionary search (Evolve systems) has reached state-of-the-art results on mathematical and combinatorial tasks, yet most existing systems report only the best of many runs and leave the run-to-run distribution undocumented. We ask how a fixed budget of LLM calls should be allocated, and how reliably a single run reaches the reported numbers. Sweeping the depth-breadth grid over five models and three tasks, we identify two empirical regularities: a fitness-compute envelope along which capability ordering largely collapses on effective FLOPs, and a bilinear depth-breadth fit with task-specific interaction; both are gated by model-task capability. Motivated by these regularities, we propose BaSE (Bandit-based Self-Evolving), a multi-armed bandit that allocates LLM calls across parallel trajectories. Without changing the model, prompt, or evaluator, BaSE improves mean fitness by 12.3% over the strongest island-protocol baseline across 8 (model, task) cells, with the largest gains on high-variance settings: a reliability gain from allocation alone.
[NLP-125] DynSess: Dynamic Session-Level Evaluation and Optimization Framework for Role-Playing Agents
【速读】： 该论文试图解决角色扮演类大语言模型在长时间多轮对话中难以维持角色一致性与交互质量的问题，现有评估和优化方法多局限于单轮层面，无法捕捉长期行为表现。其解决方案的关键在于提出一个统一的会话级框架 DynSess，通过基于长程行为指标的评分体系（DynSess-Eval）对完整对话会话进行评价，并利用该会话级奖励信号，结合多轮前瞻搜索构建高质量训练轨迹，进而训练出两种互补的策略优化模型：DSPO（离策略）和 GSRPO（同策略）。实验表明，DynSess-Eval 在与人类判断的一致性上显著优于现有评估方法，而 DynSess-Character 在盲测中表现媲美最强角色模型，且参数量更少，同时保持了良好的角色一致性和交互能力。
链接: https://arxiv.org/abs/2605.29256
作者: Rongsheng Zhang,Jiji Tang,Junnan Ren,Zuyi Bao,Weijie Chen,Ruofan Hu,Zhou Zhao,Tangjie Lv,Yan Zhang
机构: Zhejiang University (浙江大学); Fuxi AI Lab, NetEase Inc. (网易伏羲AI实验室); Xiamen University (厦门大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Role-playing with large language models is fundamentally a session-level task, requiring agents to sustain character identity and interaction quality across extended multi-turn conversations. Yet existing evaluation and optimization methods remain largely turn-level, failing to capture long-horizon quality. We propose DynSess, a unified session-level framework for role-playing agents. DynSess-Eval scores complete dialogue sessions via rubrics targeting long-horizon behaviors. Leveraging its session-level rewards, we construct high-quality training trajectories through multi-turn lookahead search and train DynSess-Character with two complementary variants: DSPO (off-policy) and GSRPO (on-policy). Experiments show that DynSess-Eval aligns with human judgments substantially better than prior evaluators, and blind human evaluation further shows that DynSess-Character matches the strongest character model despite using substantially fewer parameters, while maintaining strong role consistency and interactive ability. Our dataset and code will be released to facilitate future research.
[NLP-126] DenseSteer: Steering Small Language Models towards Dense Math Reasoning ICML2026
【速读】： 该论文试图解决小模型（≤3B参数）在多步推理任务中表现显著落后于大语言模型（LLMs）的问题。其解决方案的关键在于提出一种名为DenseSteer的训练-free推理时调控框架，该框架通过调节模型内部表示以逼近“密集推理”（Dense Reasoning）模式——即用更少的推理步骤实现更高的每步信息密度。实验表明，该方法可在不增加词元级负对数似然（Negative Log-Likelihood）的前提下持续提升小模型的准确性，验证了密集推理作为数学问题求解的有效结构化路径。
链接: https://arxiv.org/abs/2605.29247
作者: Yang Ouyang,Shuhang Lin,Jung-Eun Kim
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: ICML 2026
Abstract:Large language models (LLMs) demonstrate strong chain-of-thought (CoT) reasoning abilities, while smaller models (= 3B parameters) significantly underperform on multi-step reasoning tasks. Based on empirical analyses of the Qwen-2.5 model family on math reasoning benchmarks, we find that more proficient reasoning is associated with fewer reasoning steps but higher information density per step, a property we term Dense Reasoning. Motivated by this observation, we propose DenseSteer, a training-free inference-time steering framework that enhances small-model reasoning by modulating internal representations toward dense reasoning patterns. Experiments show that our method yields consistent accuracy improvements without increasing token-level Negative Log-Likelihood, highlighting dense reasoning as an effective structural approach to mathematical problem solving.
[NLP-127] Implicit Identity Technologies for LLM s: Fingerprinting and Watermarking across Datasets Models and Generated Content IJCAI ECAI2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在资产保护与溯源方面的关键问题，即如何有效识别和验证LLM相关资产（如数据集、模型本身及生成内容）的来源与归属。其核心挑战在于当前研究碎片化严重，指纹识别（fingerprinting）与水印技术（watermarking）概念混淆且应用场景孤立。解决方案的关键在于提出“隐式身份”（implicit identity）作为统一抽象框架，明确区分非侵入式指纹（基于模型内在特征）与侵入式水印（人为嵌入于数据或模型中），并构建基于生命周期的分类体系，涵盖数据集、模型和生成内容三个阶段，同时按验证语义进一步细分为相似性归因与密钥验证两类方法。此外，论文还建立以可识别性（identifiability）、鲁棒性（robustness）和可部署性（deployability）为核心的评估框架，为未来LLM身份技术的研究与应用提供系统化基础。
链接: https://arxiv.org/abs/2605.29245
作者: Bing Liu,Shunping Wang,Yufan Zhu,Xinyi Yu,Jing Huang,Linkang Du,Hongbin Pei,Wei Luo
机构: School of Cyber Science and Engineering, Xi’an Jiaotong University, Xi’an, China; State Grid Henan Marketing Service Center, Henan, China; Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China; School of Cyber Security, University of Chinese Academy of Sciences, Beijing, China; School of Information Technology, Deakin University, Geelong, Australia
类目: Cryptography and Security (cs.CR); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted by IJCAI-ECAI 2026. 11 pages, 1 figure. Survey and taxonomy of LLM fingerprinting and watermarking for identity, provenance, generated-content attribution, and asset protection
Abstract:This paper presents a survey and taxonomy of LLM fingerprinting and watermarking for identity, ownership verification, provenance, and generated-content attribution. Large language models (LLMs) require substantial investments in data, computation, and expertise, and are increasingly deployed in high-stakes settings, making it critical to protect LLM-related assets and trace their origins. Existing work has rapidly expanded across dataset provenance, model ownership, and generated-content detection, but the field remains fragmented: fingerprinting and watermarking are often used inconsistently, and methods are typically studied within isolated asset-specific settings. To address this gap, we introduce implicit identity as a unifying abstraction for verifiable but not directly observable identity signals in LLM systems. We distinguish fingerprinting as non-intrusive identity derived from intrinsic characteristics, and watermarking as intrusive identity deliberately embedded into data, models, or generated content. We then propose a lifecycle-based taxonomy that organises techniques across datasets, models, and generated content, and further separates them by verification semantics: similarity-based attribution and keyed verification. Finally, we establish an evaluation framework centred on identifiability, robustness, and deployability, summarising representative metrics under realistic access and transformation regimes. By unifying terminology, lifecycle stages, and evaluation objectives, this survey provides a structured foundation for studying LLM identity technologies and for developing more reliable mechanisms for asset protection and provenance.
[NLP-128] Wait! Theres a Way Out: A Decision Mechanism for Forecasting Conversational Derailment ACL2026
【速读】： 该论文试图解决对话脱轨（conversational derailment）预测模型在在线决策中因忽略未来恢复可能性而导致假阳性率过高的问题。现有方法仅基于历史话语估计脱轨概率，隐含假设对话轨迹是固定的，从而无法捕捉紧张局势可能缓解的潜在路径。其解决方案的关键在于提出一种将触发决策与脱轨概率估计解耦的机制：通过前向模拟（forward-looking simulations）评估当前紧张时刻是否存在合理的恢复路径，并在预期可恢复时延迟触发警报。这一机制显著降低了假阳性率，同时保持了预测准确性，凸显了将决策制定作为预测系统核心组件的重要性。
链接: https://arxiv.org/abs/2605.29243
作者: Laerdon Kim,Vivian Nguyen,Cristian Danescu-Niculescu-Mizil
机构: Cornell University
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: To appear in the Proceedings of ACL 2026
Abstract:Forecasting conversational derailment is the task of predicting, as the conversation unfolds, whether it will eventually derail into personal attacks. Since forecasting models operate in an online fashion, they must decide whether to “trigger” an alert after each utterance–for example, to notify participants or a moderator that the conversation is at risk of derailing. Existing approaches make this decision solely based on the estimated likelihood of derailment given the preceding utterances, implicitly assuming that the conversation’s future trajectory is fixed. As a result, they ignore the possibility of future recovery and incur an unnecessarily high rate of false positives. In this work we propose a method for decoupling the decision to trigger from derailment likelihood estimation. Our approach is inspired by the first human baseline on this task, which shows that humans achieve dramatically lower false positive rates by selectively deferring their decision to trigger when they anticipate that tension is likely to subside. We operationalize this insight with a deferral mechanism that uses forward-looking simulations to assess whether a tense moment admits plausible paths to recovery. Incorporating this mechanism into a state-of-the-art forecasting model substantially reduces false positives without sacrificing forecasting accuracy. More broadly, this work highlights the value of treating decision-making as a first-class component of forecasting systems. Comments: To appear in the Proceedings of ACL 2026 Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY) Cite as: arXiv:2605.29243 [cs.CL] (or arXiv:2605.29243v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2605.29243 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-129] Relevance as a Vulnerability: How Web Retrieval Degrades Safety Alignment in LLM Agents
【速读】： 该论文试图解决的问题是：在大型语言模型（LLM）代理中引入外部检索功能后，如何导致安全对齐机制的弱化，进而增加对有害请求的合规性。解决方案的关键在于提出了一种名为AgentREVEAL的诊断框架，该框架从两个维度分析检索引发的安全退化问题：一是检索如何集成到代理流程中，二是检索内容的属性。研究发现，将工具调用与响应生成绑定在同一步骤会显著放大有害输出；同时揭示了“安全来源悖论”——即使检索到的内容本身具有警示或安全声明性质，也会使有害合规率平均上升25%。此外，相关性（relevance）被识别为两种漏洞的共同触发条件，表明检索增强型代理存在本质上的安全-效用权衡。为此，作者还构建了HarmURLBench基准，包含1,405个真实URL和320种有害行为，以支持未来对检索增强代理安全性更系统的评估。
链接: https://arxiv.org/abs/2605.29224
作者: Aditya Nawal,Manit Baser,Mohan Gurusamy
机构: National University of Singapore (新加坡国立大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:AI agents augment large language models with external tools such as web retrieval, enabling grounded and up-to-date responses. However, incorporating external content into the generation pipeline can weaken the safety alignment mechanisms that govern model outputs. Prior work shows that enabling retrieval in agents increases compliance with harmful requests. We introduce AgentREVEAL, a diagnostic framework for analyzing retrieval-induced safety degradation in LLM agents. The framework examines two axes: how retrieval is integrated into the agent pipeline and the properties of the retrieved content. Along the integration axis, we find that binding tool invocation and response generation in a single step amplifies harmful outputs. Along the content axis, we uncover the Safe Source Paradox: even oppositional or safety-oriented sources, such as pages containing warnings or risk disclaimers, can increase harmful compliance by an average of 25% compared to the no-retrieval baseline. Finally, we show that relevance acts as a shared activation condition for both vulnerabilities. Similar patterns appear on frontier closed models, and harmful compliance remains elevated under several representative pipeline interventions, with some agents also entering this regime under autonomous retrieval. Because relevance is also what makes retrieval useful, these results expose a safety-utility trade-off for retrieval-enabled agents. We introduce HarmURLBench, a benchmark containing 1,405 real-world URLs paired with 320 harmful behaviors to support future evaluations.
[NLP-130] GTA: Generating Long-Horizon Tasks for Web Agents at Scale
【速读】： 该论文试图解决当前Web代理（Web agents）在训练与评估中面临的可扩展性监督不足问题，尤其是现有基准数据集多为人工构建、缺乏中间轨迹信息，而自动生成方法则存在成本高、偏差大和深度不足的缺陷，难以支持代理在真实场景下执行多跳（multi-hop）、跨页面任务时的泛化能力。解决方案的关键在于提出一个名为GTA的可扩展框架，其核心创新包括：通过爬取（crawling）与生成（generation）解耦提升效率，基于站点图（site graph）约束任务的组合性以增强真实性，以及借助确定性重放（deterministic replay）和系统性验证实现密集监督（dense supervision）。该框架在50余个涵盖电商、政府、论坛和新闻等领域的多语言网站上实例化，生成了具有执行轨迹的真实任务，揭示了显著的人类-代理性能差距，并支持细粒度诊断，从而推动了Web代理研究从静态标注向动态、可复现基准的演进。
链接: https://arxiv.org/abs/2605.29218
作者: Tenghao Huang,Kung-Hsiang Huang,Prafulla Kumar Choubey,Yilun Zhou,Muhao Chen,Jonathan May,Chien-Sheng Wu
机构: University of Southern California (南加州大学); Salesforce AI Research (Salesforce人工智能研究); University of California, Davis (加州大学戴维斯分校)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Published at Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics
Abstract:Web agents, which couple language models with browsing and tool-use capabilities, show promise as open web assistants. Yet progress is increasingly limited by the lack of scalable, process-level supervision. Existing benchmarks are largely manually constructed, providing only coarse start-goal annotations without intermediate trajectories, while recent automatic generation efforts remain expensive, biased, and shallow. These limitations prevent reliable training and evaluation of agents that must generalize to realistic, multi-hop, cross-page tasks. We introduce a scalable framework, GTA, that integrates crawling, retrieval-based seeding, in-context generation, and automated quality control to produce realistic tasks paired with executable trajectories. This design decouples crawling from generation for greater efficiency, grounds tasks in the site graph to enforce compositionality, and ensures dense supervision through deterministic replays and systematic validation. We instantiate the pipeline on over 50 websites covering e-commerce, government, forums, and news, with multilingual and multi-hop coverage. The resulting benchmark reveals a significant human-agent performance gap and enables detailed diagnostics. Our contributions are three-fold: (i) formalizing multi-hop web-agent task generation, (ii) proposing an efficient and validated pipeline for automatic data creation, and (iii) releasing a dynamic benchmark with reproducible evaluation.
[NLP-131] Reason Ops: Operator Segmentation for LLM Reasoning Traces
【速读】： 该论文试图解决的问题是：大型推理模型（Large Reasoning Models）的思维链（Chain-of-Thought, CoT）轨迹通常包含数以万计的词元（token），但目前缺乏一套通用且可解释的词汇来描述其内部结构；现有分析方法要么过于僵化，要么表达能力不足，无法跨模型和领域捕捉关键特征。解决方案的关键在于提出 ReasonOps ——一种无监督、表达能力强的标注方法，通过自动识别句子开头的3词枢纽（3-token pivots）进行聚类，提取出7种普遍存在的推理操作符（reasoning operators），如回溯（backtracking）、推断（inferring）和假设（hypothesizing）。这些操作符在12个不同家族的大语言模型（LLMs）和8个推理基准测试中均稳定出现，且能有效用于模型识别（macro-AUC高）、答案正确性预测（WP-AUC优于基线）以及早期质量估计（仅需50%轨迹即可达到WP-AUC），从而实现了对大模型推理过程的结构化理解与高效下游应用。
链接: https://arxiv.org/abs/2605.29192
作者: Daniel Lee,Owen Queen,James Zou
机构: Stanford University (斯坦福大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Chain-of-thought traces from large reasoning models can span tens of thousands of tokens, yet we lack a vocabulary for describing their internal structure. Previous methods developed to analyze chain-of-thought traces are either too rigid or not expressive enough, failing to capture features across domains and models. To remedy this, we develop ReasonOps, an unsupervised, expressive method for annotating chain-of-thought traces, providing succinct universal operators. Using ReasonOps, we analyze 44,662 traces from 12 thinking LLMs spanning 6 families across 8 reasoning benchmarks and discover that they share a common compositional structure: 7 recurring reasoning operators – discourse-level moves such as backtracking, inferring, and hypothesizing – that emerge from unsupervised clustering of sentence-initial 3-token pivots. These operators appear across every model family and benchmark domain, confirmed by three independent LLM judges who classify held-out samples at 70 -76% accuracy. We analyze the structure of operators on easy vs. hard problems, revealing that reflective operators are more helpful on hard problems and harm performance on easy problems. Operator sequences are highly model-identifying: a classifier trained on operator distributions alone recovers the source model with macro-AUC, revealing that each model family has a distinctive reasoning fingerprint. Structural operator features predict within-problem answer correctness well above baselines. Classifiers built on these operators reach WP-AUC and on AIME specifically. ReasonOps further enables early quality estimation well before the trace completes: we predict at WP-AUC for only 50% of the trace. The ReasonOps pipeline is unsupervised and annotation-free, enabling deep insights into LLM reasoning traces as well as strong downstream results on model identification and correctness prediction.
[NLP-132] When RL Suppresses Its Own Vocabulary: Recovering Reasoning Diversity in Puzzle-to-Math Transfer
【速读】： 该论文试图解决的问题是：在仅使用约束满足谜题（constraint-satisfaction puzzles）进行监督微调（SFT）和强化学习（RL）训练的情况下，大语言模型（LLM）是否能够实现跨领域推理能力的迁移，尤其是对数学问题的解决能力，并揭示其背后的机制。解决方案的关键在于提出了一种基于推理原语（reasoning primitive）的分析框架，通过9类跨度分类器与模式提取技术，将思维链（chain-of-thought）轨迹分解为基本推理单元并追踪其演化过程。研究发现，谜题SFT阶段构建了基础推理原语词汇表，使模型在奥数难题（OlymMATH-Hard）上的准确率提升7个百分点；随后的GSPO强化学习阶段将这些原语组合成“计算-验证”链式结构，再提升6个百分点；但同时抑制了如“假设”和“回溯”等探索性原语。为此，作者引入新颖性奖励（novelty bonus），利用参考模型下的困惑度作为信号鼓励多样化正确推理路径，恢复了探索性原语并进一步提升7个百分点，最终在不引入任何数学训练数据的前提下，将硬核数学能力上限从16.0%提升至36.0%。
链接: https://arxiv.org/abs/2605.29190
作者: Mayug Maniparambil,Arjun Karuvally,Terrence Sejnowski,Fergal Reid
机构: Fin AI Research; Salk Institute for Biological Studies
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: Preprint
Abstract:Reinforcement learning using verifiable rewards (RLVR) improves LLM reasoning, but the conditions under which it transfers across domains – and why it does so – remain under-explored. We study cross-domain transfer in a 7B model whose SFT and RL post-training stages use only constraint-satisfaction puzzles, with no mathematics problems in the post-training data. To analyze how transfer emerges, we introduce a reasoning primitive-level framework that combines a 9-class span classifier with motif extraction, allowing us to segment chain-of-thought traces into primitive motifs and track their evolution across training stages and domains. We find that puzzle SFT induces a reasoning-primitive vocabulary, yielding a +7 pp \textttpass@32 gain on OlymMATH-Hard. Vanilla GSPO then composes these primitives into longer compute-verify chains, adding a further +6 pp. However, this RL stage also suppresses exploratory primitives such as \textithypothesize and \textitbacktrack. To address this, we introduce a novelty bonus that rewards diverse correct rollouts, using perplexity under the reference model as a signal. This restores recovery primitives during RL and adds a further +7 pp \textttpass@32 relative to vanilla GSPO. Finally, the end-to-end recipe raises the hard-math capability ceiling from 16.0% at the OLMo3-7B-Instruct-SFT base to 36.0% , without adding any mathematics problems during the SFT or RL stages.
[NLP-133] Slogans or Stance? A Label-Light Diagnostic for Entrepreneurial-Discourse Measurement on Chinese SOE Speeches
【速读】： 该论文旨在解决当前企业语境下构念测量（如“企业家精神”）中方法有效性不足的问题，特别是针对词典法、主题模型和嵌入相似性评分器等常用工具缺乏可靠诊断手段的缺陷。其解决方案的关键在于提出一种标签轻量级的测量诊断框架，通过利用中央管理的中国国有企业领导层演讲数据中的自然实验设计——即同一公司不同发言人对与相同发言人对之间的对比——检验各测量方法在控制企业固定效应后是否能有效捕捉到领导者个体身份差异。研究发现，传统LDA模型无法区分领导者差异（Cohen’s d=0.20），而词典法（d=0.81）和中文句子编码器（d=0.65）表现较好，尤其是零样本大语言模型（Qwen3.5:9B）在配对对比中达到显著效果（d=1.09, p=0.034）。进一步分析揭示，该模型性能的一半可归因于领导者特有语言风格（idiolect），而非构念本身；同时引入置信度加权校准虽提升稳定性，但削弱了信号强度，表明现有测量工具仍需更精细的误差分离机制。
链接: https://arxiv.org/abs/2605.29188
作者: Ting Gong,Shangquan Sun
机构: Tsinghua University (清华大学)
类目: Computation and Language (cs.CL)
备注: 15 pages, 2 figures, 7 tables
Abstract:Dictionary methods, topic models, and embedding-similarity scorers are widely used in CSS and management research to measure constructs such as “entrepreneurial spirit” in corporate speeches. We contribute a label-light measurement diagnostic for such instruments rather than a new extraction model. On a corpus of 80 speeches by leaders of centrally administered Chinese state-owned enterprises, we exploit a natural experiment of 24 same-company different-speaker pairs and 5 same-company same-speaker pairs to test whether a method’s per-document indices vary with leader identity holding firm constant. LDA fails (Cohen d=0.20, 95% CI [-0.72, 1.20]); a dictionary scorer reaches d=0.81 and a Chinese sentence encoder d=0.65 on doc-vector distances of order 10^-3. A zero-shot 9B open-weight LLM (Qwen3.5:9b) raises paired-contrast d to 1.09 (exact permutation p1=0.034). We downgrade three claims accordingly: gold F1 measures consistency with the LLM’s own prompt rule rather than external construct recovery; doc-level style residualisation cuts the LLM’s d to 0.43 (p1=0.22), so roughly half of the effect is consistent with leader idiolect; and a confidence-weighted calibration trades Delta for variance with an auto-mined slogan lexicon near-inert in ablation. We release the 2,190-segment scored corpus, the 170-paragraph pilot, the slogan lexicon, two-family LLM scores, and the evaluation harness.
[NLP-134] UA-Legal-Bench: A Benchmark for Evaluating Large Language Models on Ukrainian Legal Reasoning
【速读】： 该论文试图解决法律自然语言处理（Legal NLP）基准测试高度依赖英语、忽视形态丰富的非拉丁字母语言（如乌克兰语）中模型失败模式的问题。其解决方案的关键在于构建并发布UA-Legal-Bench——一个基于乌克兰国家法院判决统一登记册（EDRSR，含9950万份判决书）的五任务基准测试，涵盖案件类型分类、判决形式分类、案件结果预测、法律规范提取和原因类别预测。通过在11个不同规模（3B–675B参数）的大语言模型上进行零样本与三样本提示评估，研究揭示了任务特异性的小样本提升效应，并指出准确率在类别不平衡任务中具有误导性，强调应使用宏平均F1分数等指标来更公平地评估模型性能；同时发现同一家族内模型的性能提升阈值存在显著差异，表明模型缩放策略需针对具体任务和语言特性定制。
链接: https://arxiv.org/abs/2605.29170
作者: Volodymyr Ovcharov
机构: SecondLayer
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 13 pages, 5 figures, 4 tables. Data: this https URL
Abstract:Legal NLP benchmarks are overwhelmingly English-centric, leaving failure modes in morphologically rich, non-Latin-script languages undetected. We introduce UA-Legal-Bench, a five-task benchmark for evaluating large language models on Ukrainian legal reasoning, built from the Unified State Register of Court Decisions (EDRSR) – one of the world’s largest open judicial corpora (99.5 million decisions). The benchmark comprises: (1) case-type classification (4 classes, n=2,000), (2) judgment form classification (4 classes, n=2,000), (3) case-outcome prediction (6 classes, n=800), (4) legal norm extraction (n=1,794), and (5) cause category prediction (22 classes, n=1,871). We evaluate 11 LLMs (3B–675B) from five families under zero-shot and 3-shot prompting via AWS Bedrock with 158K API calls. Our results reveal sharply task-dependent few-shot effects: few-shot prompting improves judgment form classification by up to +38.6 pp but has mixed effects on outcome prediction. We show that accuracy is misleading on imbalanced legal tasks: the model with highest COP accuracy (62%) is a majority-class predictor (macro-F1: 23%), while the genuinely best model scores only 44% macro-F1. Within-family scaling analysis reveals that 8B models can match frontier performance on surface-level tasks but scaling thresholds vary dramatically across families. We release all data, prompts, and model predictions.
[NLP-135] Parallax: Parameterized Local Linear Attention for Language Modeling
【速读】： 该论文试图解决大语言模型（LLM）中注意力机制长期结构僵化的问题，特别是传统 softmax 注意力在局部常数估计下存在的偏差-方差权衡不足问题。解决方案的关键在于提出 Parallax——一种可扩展的参数化局部线性注意力机制，其通过消除原 LLA 中的数值求解器，并引入一个额外的查询类投影器来探测键值（KV）协方差，从而实现更稳定的计算和更强的关联记忆能力。此外，作者设计了一种面向硬件的算法，提升算术强度以使注意力计算更趋近于计算密集型，使得原型解码内核在不同批量大小和上下文长度下性能优于或等同于 FlashAttention 2/3。实验表明，Parallax 在预训练阶段持续降低困惑度，并在参数匹配与计算匹配条件下均带来下游任务性能提升，体现了帕累托改进；同时发现 Muon 优化器能显著释放 Parallax 的容量潜力，这是首次在注意力机制研究中实证展示架构与优化器的协同设计优势。
链接: https://arxiv.org/abs/2605.29157
作者: Yifei Zuo,Dhruv Pai,Zhichen Zeng,Alec Dewulf,Shuming Hu,Zhaoran Wang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Large Language Models (LLMs) have become the central paradigm in artificial intelligence, yet the core computational primitive of attention has remained structurally unchanged. Local Linear Attention (LLA) is an attention mechanism derived from nonparametric statistics in the test-time regression framework. In contrast to prior research on efficient attention variants, LLA upgrades the local constant estimate in softmax attention to a local linear estimate, yielding provably superior bias-variance tradeoffs for associative memory. However, LLA has not been scaled in LLM pretraining due to computational and numerical stability concerns. We introduce Parallax, a parameterized Local Linear Attention that is scalable for LLMs. Parallax eliminates the numerical solver in LLA and learns an extra query-like projector that probes the KV covariance. We place Parallax within a family of attention mechanisms connected by the bandwidth, the probe construction and the affine structure. We propose a hardware-aware algorithm that increases the arithmetic intensity over FlashAttention, shifting attention into a more compute bound regime. Our prototype decode kernel matches or outperforms FlashAttention 2/3 across diverse batch sizes and context lengths. We pretrain Parallax at 0.6B and 1.7B scales and find consistent perplexity improvements throughout pretraining with gains that transfer to downstream benchmarks. The advantage persists under both parameter-matched and compute-matched controls, demonstrating a Pareto improvement. We perform careful pretraining ablations and identify a novel phenomenon whereby Muon unlocks the capacity of Parallax. To our knowledge, this is the first empirical demonstration of strong architecture-optimizer codesign for attention mechanisms in the architecture research literature.
[NLP-136] RUBRIC-ARROW: Alternating Pointwise Rubric Reward Modeling for LLM Post-training in Non-verifiable Domains
【速读】： 该论文试图解决的是在主观且不可验证的场景下，点对点奖励建模（pointwise reward modeling）因绝对评分（absolute scoring）而面临的问题，以及现有基于评分表（rubric-based）方法依赖前沿大语言模型（LLM）且因硬布尔聚合导致评分平局（ties）的局限性。其解决方案的关键在于提出 RUBRIC-ARROW，一种交替训练框架，联合优化一个评分表生成器和一个条件于评分表的评判模型（judge），并通过仅使用成对偏好数据的强化学习（RL）阶段实现训练；该方法结合了基于概率的评分规则以减少平局、分阶段偏好奖励机制以及交替的 GRPO（Generalized Reward Policy Optimization）训练策略，从而显著提升奖励建模准确性并为下游策略后训练带来稳定收益。
链接: https://arxiv.org/abs/2605.29156
作者: Haoxiang Jiang,Zihan Dong,Tianci Liu,Wanying Wang,Ran Xu,Tony Yu,Linjun Zhang,Haoyu Wang
机构: 未知
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Pointwise reward modeling offers critical signals for LLM post-training, yet struggles with absolute scoring in subjective, non-verifiable settings. Rubric-based methods address this by decomposing evaluation into explicit criteria, but existing approaches typically depend on frontier LLMs and suffer from ties caused by hard Boolean aggregation. We present RUBRIC-ARROW, an alternating framework that jointly trains a rubric generator and a rubric-conditioned judge, with its RL stage using only pairwise preference data. Our method couples a probability-based scoring rule that reduces ties with phase-specific preference-based rewards and an alternating GRPO scheme that together train the pointwise evaluator. Extensive experiments show that RUBRIC-ARROW achieves competitive reward-modeling accuracy and yields consistent gains for downstream policy post-training.
[NLP-137] SafeRx-Agent : A Knowledge-Grounded Multi-Agent Framework for Safe and Explainable Medication Recommendation
【速读】： 该论文旨在解决药物推荐中存在的两个核心问题：一是模型层面，传统方法仅能预测结构化的药物编码，缺乏充分的临床证据支持；而大语言模型（LLM）虽能利用更丰富的临床上下文，却存在安全验证不足和可追溯性差的问题；二是任务层面，现有基准通常使用宽泛的药物类别，忽略了亚组级别的安全性差异，易导致风险高估。解决方案的关键在于提出首个基于第四级ATC代码生成的细粒度药物推荐设定，并设计Safe Prescription Agent（SafeRx-Agent），这是一个基于知识增强的多智能体框架，通过整合患者上下文、外部临床知识及安全验证机制，实现可追溯的药物组合推荐。实验结果表明，SafeRx-Agent在MIMIC-III和MIMIC-IV数据集上显著提升了细粒度药物预测准确性，同时有效控制了药物相互作用、禁忌症及用药集合规模。
链接: https://arxiv.org/abs/2605.29146
作者: Xinyu Wang,Hanwei Wu,Zhenghan Tai,Sicheng Lyu,Qincheng Lu,Ziyu Zhao,Jijun Chi,Jingrui Tian,Xiao-Wen Chang,Ziyang Song
机构: McGill University; McMaster University; University of Toronto; ByteDance; LinkedIn; Ohio University
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Medication recommendation predicts medications for patient visits, but existing methods still face two key challenges. At the model level, traditional drug recommendation methods only predict structured drug codes with limited evidence grounding, while LLM agents can use richer clinical context but may lack safety verification and traceability. At the task level, existing benchmarks often use broad medication categories, which ignore subgroup-level safety differences and can lead to risk overestimation. We introduce the first fine-grained medication recommendation setting based on fourth-level ATC code generation. We propose Safe Prescription Agent (SafeRx-Agent), a knowledge-grounded multi-agent framework that uses patient context, external clinical knowledge, and safety verification to recommend traceable medication sets. Experimental results on MIMIC-III and MIMIC-IV datasets show that SafeRx-Agent improves fine-grained medication prediction accuracy while controlling drug interactions, contraindications, and medication set size.
[NLP-138] he Confidence Shortcut: A Reasoning Failure Mode of Masked Diffusion Models
【速读】： 该论文试图解决的问题是：当前基于置信度的解码策略（confidence-based decoding）与复杂推理所需的逻辑流轨迹（logical-flow trajectories）存在本质性错位，而现有的置信度对齐训练方法反而强化了这种错位，导致模型在处理高难度输入时产生大量错误。解决方案的关键在于采用随机掩码（random masking）训练策略，而非当前主流的置信度对齐掩码训练，因为随机掩码能够有效保留推理轨迹的条件依赖关系（reasoning-trajectory conditionals），从而显著降低在复杂任务中的错误率——即使其在表面上看似效率较低。实验表明，在多数字加法等推理任务中，随机掩码将错误率维持在较低水平，而置信度对齐训练则使错误率提升一个数量级。
链接: https://arxiv.org/abs/2605.29123
作者: Dueun Kim,Albert No
机构: Yonsei University (延世大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Masked diffusion language models (MDMs) uniquely support any-order generation, with confidence-based decoding currently serving as the de facto standard inference policy. To optimize for this, recent training schemes attempt to align training mask patterns directly with those observed during generation. However, we argue that confidence-based decoding is inherently misaligned with the logical-flow trajectories required for complex reasoning, and that confidence-aligned training actively entrenches this misalignment. We make this concrete using multi-digit addition, where the decoding strategy prematurely predicts locally easy digits before resolving their long-range dependencies, producing high-confidence errors on challenging inputs. While traditional random masking keeps the failure rate low on this challenging tail, confidence-aligned training amplifies the error rate by an order of magnitude. Across five distinct reasoning tasks, this same pattern emerges with task-dependent severity: confidence-based decoding induces failures on highly complex inputs, and confidence-aligned training exacerbates them. In contrast, random masking – despite its perceived inefficiency – robustly preserves the reasoning-trajectory conditionals essential for solving the challenging tail.
[NLP-139] Structured Prompt Optimization Meets Reinforcement Learning for Global and Local Interpretability over Complex Text
【速读】： 该论文试图解决大语言模型（LLM）在文本分类任务中面临的两个核心问题：一是监督微调（仅使用标签）虽然可扩展性好，但在处理复杂文本时推理能力有限且缺乏模型透明度；二是离散提示优化虽能生成人类可读的指令，但性能和可扩展性不足。解决方案的关键在于提出一种名为eXTC（eXplainable Text Classifier）的新框架，其核心创新包括三个渐进阶段：(1) 通过新型结构化提示优化算法学习以自然语言形式表达的标准操作程序（SOP，或称规则手册）；(2) 将大型教师模型（teacher LLM）的SOP引导推理知识蒸馏到一个紧凑的小型语言模型（compact LM）中；(3) 利用强化学习扩展初始SOP之外的推理能力。这一设计使eXTC能够在推理时提供快速响应、局部推理轨迹以及全局模块化的领域规则解释，同时在多个基准测试中显著优于现有范式，在分类准确性和解释质量上均实现逐阶段提升。
链接: https://arxiv.org/abs/2605.29076
作者: Tianyang Zhou,Wenbo Chen,Pierre Jinghong Liang,Leman Akoglu
机构: Carnegie Mellon University (卡内基梅隆大学); Amazon (亚马逊)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:LLMs have advanced text classification, yet existing paradigms face a trade-off: supervised (label only) fine-tuning is scalable but offers limited reasoning on complex text and lacks broader model transparency, while discrete prompt optimization offers human-readable instructions but struggles with performance and scalability. We introduce eXTC (eXplainable Text Classifier) with three progressive stages: (1) learning a Standard Operating Procedure (SOP, or rulebook) in natural language via a new Structured Prompt Optimization algorithm; (2) SOP-grounded reasoning distillation from a large teacher LLM into a compact LM; and (3) expanding reasoning capabilities beyond the initial SOP via reinforcement learning. This design enables eXTC to provide (i) fast inference via a compact LM, with (ii) inference-time local reasoning traces, alongside a global, modular explanation of its learned domain rules, while (iii) significantly outperforming existing paradigms across diverse benchmarks in both classification performance and explanation quality, with stage-by-stage gains.
[NLP-140] Robust and Efficient Guardrails with Latent Reasoning
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）在实际部署中安全防护机制的效率与效果之间的权衡问题。现有基于显式推理（explicit reasoning）的安全护栏虽然性能优于单次分类方法，但存在查询延迟高和令牌消耗大的缺陷，难以满足高吞吐场景的需求。解决方案的关键在于提出COLAGUARD，它通过分阶段训练课程将多步安全推理过程映射到连续隐空间（continuous latent space），从而在推理阶段实现隐藏状态的直接传播，避免了显式生成推理路径。实验表明，COLAGUARD在十个提示和响应审核任务上相比Llama Guard 3提升了8.24点宏F1分数，并且在性能上达到与显式推理基线GuardReasoner相当的水平，同时实现了12.9倍的速度提升和22.4倍的令牌使用减少，验证了隐式推理在兼顾安全性与效率方面的可行性。
链接: https://arxiv.org/abs/2605.29068
作者: Siddharth Sai,Xiaofei Wen,Muhao Chen
机构: University of California, Davis
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注:
Abstract:Maintaining the safety of large language models (LLMs) is crucial as they are increasingly deployed in real-world applications. Existing safety guardrails typically rely on single-pass classification or, more recently, distilled reasoning. Reasoning-based guardrails significantly outperform classification-only baselines, but they incur substantial query latency and token overhead that make them impractical for highthroughput deployment. To address this challenge, we propose COLAGUARD, a guardrail model that transfers multi-step safety reasoning into a continuous latent space through a stage-wise training curriculum, enabling direct hidden-state propagation at inference. Evaluated on ten prompt- and response-moderation settings spanning eight safety benchmarks, COLAGUARD improves macro-F1 by 8.24 points over Llama Guard 3 and matches our explicit reasoning baseline, GuardReasoner, in macroF1 while delivering a 12.9X speedup and 22.4X reduction in token usage. Our results suggest that latent reasoning offers a practical alternative to explicit rationale generation for deployable guardrails, jointly improving safety robustness and inference efficiency rather than treating them as competing objectives.
[NLP-141] Bosses Kings and the Commons: Cooperation Under Power Asymmetry in LLM Societies
【速读】： 该论文试图解决的问题是：在存在权力不对称结构（如某些个体或机构对资源提取和集体结果拥有不成比例控制权）的情况下，生成式AI代理（LLM）在模拟的公共资源治理场景中是否仍能维持合作与可持续性。现有研究大多忽略这种现实中的权力不平等，而本文通过构建一个包含不对称权力结构的多智能体仿真框架——SovSim，来系统评估这一问题。解决方案的关键在于：引入一个具有不对称权力的“统治者”（boss/king）角色与一群对称的“劳动者”（workers/peasants）共同管理共享资源，从而揭示权力失衡如何导致合作崩溃和资源可持续性的急剧下降——实验结果显示，在11个最先进的大语言模型中，引入不对称权力后，系统的生存率平均下降达87.3%，显著低于对称情境下的表现。这表明，即便在高度智能化的代理群体中，权力结构的不平等仍是破坏集体理性与长期可持续性的核心因素。
链接: https://arxiv.org/abs/2605.29062
作者: Abhilekh Borah
机构: 未知
类目: Computation and Language (cs.CL)
备注: Paper under review
Abstract:Communities can sustainably manage shared resources (commons) through self-governance and cooperative norms, a central finding of Ostrom’s theory of self-governance. However, real-world commons (e.g., fisheries, forests, and irrigation systems) are often governed under asymmetric power structures, where certain individuals or institutions possess disproportionate control over resource extraction and collective outcomes. As Large Language Models (LLMs) are increasingly explored as agents in synthetic governance simulations, understanding how LLM societies behave under asymmetric power structures is becoming increasingly important, yet existing evaluations largely ignore such asymmetries. We introduce Sovereignty over the Commons Simulation (SovSim), a generative multi-agent simulation framework that incorporates an agent with asymmetric power (boss or king) into a society of symmetric agents (workers or peasants), where all agents extract from a shared resource, collectively determining its sustainability over time. Across eleven state-of-the-art models, we find that introducing asymmetric power leads to severe breakdowns in cooperation and sustainability, with up to an 87.3% degradation in survival rate relative to symmetric settings.
[NLP-142] Converted Not Equivalent: Benchmarking Codebase Conversion via Observational Equivalence
【速读】： 该论文试图解决的问题是：当前代码库转换（codebase conversion）中的编码代理（coding agents）过度依赖局部验证机制，导致其在表面指标（如单一前向损失）上表现良好时便宣告成功，而忽略了实际用户关心的语义契约（semantic contracts），从而产生看似正确但实质错误的转换结果。解决方案的关键在于提出一个名为 T2J-Bench 的基准测试框架，将代码转换问题重新定义为在固定等价契约（fixed equivalence contract）下的迁移任务，并引入一个固定的验证器（fixed verifier）通过三个有序阶段进行评估：Spec（接口合法性）、Numeric（前向输出、损失、梯度及特定目标张量）和Behavioral（固定随机种子下的短期训练动态）。实验表明，即使在高 Spec 通过率下，整体通过率仍极低（仅 26.7–28.9%），且系统普遍高估自身性能达 66.6–97.8 分，说明失败主要源于自验证机制与语义契约的错位，而非计算资源或模型能力限制。
链接: https://arxiv.org/abs/2605.29054
作者: Linxin Song,Jiefeng Chen,Yue Huang,Bhavana Dalvi Mishra,Chi Wang,Jieyu Zhao,Jinsung Yoon,Tomas Pfister
机构: University of Southern California(南加州大学); Google Cloud AI Research(谷歌云人工智能研究); University of Notre Dame(圣母大学)
类目: oftware Engineering (cs.SE); Computation and Language (cs.CL)
备注:
Abstract:Coding agents increasingly act as codebase-scale collaborators that can assist with codebase conversion, but this progress has exposed a critical weakness: agents often over-trust their own local validation routines and declare success on artifacts that satisfy surface checks while violating the semantic contracts users actually care about. This problem is especially acute in codebase conversion, where prior evaluation is largely outcome-driven and therefore unstable: two implementations can match on a shallow outcome, such as a single forward loss, while diverging in gradients, optimizer behavior, or short-horizon training dynamics. We introduce T2J-Bench, a benchmark for codebase conversion that reformulates conversion as transfer under a fixed equivalence contract. A fixed verifier then compares source and converted codebases through three ordered stages: Spec (interface admissibility), Numeric (forward outputs, losses, gradients, and objective-specific tensors), and Behavioral (short training dynamics under fixed seeds). Across 355 blind conversion attempts, the best system reaches only 26.7–28.9% overall pass rate despite Spec pass rates up to 91.1%; a 4.7x token-budget spread yields only a 2.2x pass-rate spread; and all systems overestimate success by 66.6–97.8 points relative to the fixed evaluator. This suggests that failures stem more from contract-misaligned self-validation than from limited budget or backbone strength.
[NLP-143] LLM Bridge: An LLM Pipeline for End-to-end Referential Bridging Resolution in English
【速读】： 该论文旨在解决英语中端到端指代桥接（referential bridging resolution）任务的难题，即识别文本中需要通过语义关联进行跨句指代的词语，并建立其与前文提及实体之间的连接。解决方案的关键在于构建一个基于大语言模型（LLM）的桥接解析流水线，该流水线融合了启发式预处理和后处理步骤，以及利用LLM固有的自然语言推理能力来捕捉复杂的语义关系。实验表明，LLMBridge在ISNotes、BASHI和GUMBridge三个基准数据集上均超越了此前最先进的系统，在端到端评估设置和给定金标准桥接回指词的基本评估设置下均取得最优性能。此外，作者还进行了详尽的错误分析，揭示了当前基于LLM的方法仍难以识别的桥接类型，为未来研究提供了方向。
链接: https://arxiv.org/abs/2605.29048
作者: Lauren Levine,Amir Zeldes
机构: Georgetown University (乔治城大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:In this paper, we introduce LLMBridge, a new LLM based system for the task of end-to-end referential bridging resolution in English. Our bridging resolution pipeline combines heuristic pre/post-processing with the natural language inference ability that comes from LLMs. We evaluate our bridging resolution pipeline on 3 datasets which have been used for referential bridging resolution evaluation in English: ISNotes, BASHI, and GUMBridge. Comparison to previous bridging resolution systems shows that the performance of LLMBridge surpasses previous state-of-the-art (SoTA) systems for all 3 datasets in the challenging End-to-end Evaluation Setting, as well as the Basic Bridging Resolution Evaluation Setting (gold bridging anaphor given). We also conduct a thorough error analysis of the LLMBridge performance, examining what varieties of bridging remain difficult for LLM based systems to identify. With this paper, we release the code for the LLMBridge pipeline.
[NLP-144] Adopt neq Adapt: Longitudinal Analyses of LLM Conversations in the Wild
【速读】： 该论文试图解决的问题是：当前关于用户与大语言模型（Large Language Model, LLM）交互的研究多呈现静态特征，缺乏对个体用户行为随时间演变规律的深入理解。解决方案的关键在于通过对约12,000名随机抽样的微软Bing Copilot用户进行对话轨迹分析，并与WildChat-4.8M数据集对比，揭示个体用户行为的稳定性（即“习惯黏性”）以及不同活跃度用户之间的显著差异——高活跃用户更倾向于进行复杂、专业化的任务且交互成功率更高。研究进一步指出，WildChat数据集存在偏向高技能“高级用户”的偏差，不能代表典型的人机交互场景，这对后续基于该数据集的研究具有重要警示意义。
链接: https://arxiv.org/abs/2605.29018
作者: Rebecca M. M. Hicke,Kiran Tomlinson
机构: Cornell University (康奈尔大学); Microsoft Research (微软研究院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Although a growing body of research has begun to describe user–LLM interactions, the picture it paints is largely static; little is known about how individual users change their behavior over time. To address this gap, we analyze the conversational trajectories of \sim 12,000 randomly sampled Microsoft Bing Copilot users and compare these with data from WildChat-4.8M. While the Copilot data contains significant population-level trends, we find that trends in individual user trajectories are much weaker; user habits prove to be overwhelmingly sticky. We also find stark differences between users of different activity levels: more active users have more successful conversations and use the LLM for more complex and professionally oriented tasks. Some user trends also appear in WildChat-4.8M, but we find evidence that this dataset is significantly skewed towards highly proficient “power” users. Ultimately, our results suggest that existing user behavior is difficult to change and demonstrate the extent of user heterogeneity. Our comparison between datasets highlights that WildChat does not represent typical user-AI interactions, an important caveat for downstream uses of the data.
[NLP-145] Error as a Lens: Probing LLM Reasoning through Synthetic Misconception Generation
【速读】： 该论文试图解决的问题是：在个性化辅导、教师培训和教育研究中，缺乏针对特定认知错误模式的合成学生错误数据集，而真实学生的错误标注语料库由于隐私和IRB（机构审查委员会）限制难以获取。解决方案的关键在于提出一个双代理框架——生成代理（Generation Agent, GA）与检查代理（Examination Agent, EA），其中GA根据预定义的五类认知错误类别（源自修订版布卢姆分类法）生成候选错误解答，EA则判断该答案是否既错误又符合目标类别。该框架能够系统性地构建类别分层的合成错误数据集，且实验表明，针对性错误生成比自由生成错误更具挑战性，同时答案的 grounding（即与问题语境的一致性）比增加示例或引入外部教材内容更能提升生成质量。
链接: https://arxiv.org/abs/2605.29007
作者: Xinming Yang,Jun Li
机构: CUNY Graduate Center (纽约市立大学研究生院); CUNY Queens College (纽约市立大学皇后学院)
类目: Computation and Language (cs.CL)
备注:
Abstract:Personalized tutoring, teacher training, and education research need access to \emphtargeted synthetic misconceptions, but privacy and IRB constraints make labelled corpora of real student errors scarce. LLMs could in principle generate synthetic errors at scale, but producing an arbitrary wrong answer is easy for a modern LLM while producing one that matches a specified cognitive failure mode is much harder. We present a framework that generates errors targeted to a five-class taxonomy adapted from the revised Bloom’s taxonomy, evaluated on questions from the TheoremQA dataset. A Generation Agent (GA) drafts a candidate erroneous solution conditioned on a target class, and an Examination Agent (EA) judges whether the draft is incorrect and class-consistent. The framework yields a reusable recipe for building class-stratified synthetic error datasets where authentic student corpora are unavailable. As a secondary diagnostic, targeted error generation is substantially harder than free-form incorrect-answer generation, and answer-grounding contributes more than expanded examples or external textbook content.
[NLP-146] xt-Preserving Lossy Text Compression: A Study of Strategic Deletion and LLM Reconstruction
【速读】： 该论文试图解决传统无损文本压缩在自然语言场景下压缩率提升有限的问题，提出了一种“有损语义文本压缩”（lossy semantic text compression）的新范式。其核心解决方案是：编码器通过策略性地删除文本中的部分片段，仅保留一个结构骨架（skeleton），由大语言模型（LLM）作为解码器从该骨架中重建原始内容。关键创新在于设计多种删除策略（如基于词频、词长、信息熵或混合信号的删除方法），并利用QLoRA微调技术优化解码器性能，从而在较低保留率（retention rate）下实现更高的压缩效率与重建质量平衡。实验表明，词频引导删除（WordFreq）是一种低成本且高效的基线方法，而语义感知和混合策略在中等压缩率下优势明显，同时模型跨语言（英文与中文）泛化能力强，但最优删除规则仍依赖具体数据集特性。
链接: https://arxiv.org/abs/2605.29000
作者: Yuchun Zou,Junhong Tong,Jun Li
机构: CUNY Graduate Center; CUNY Queens College
类目: Computation and Language (cs.CL)
备注:
Abstract:Traditional lossless text compression preserves every byte, but its gains on natural language are often modest in realistic operating regimes. We study \emphlossy semantic text compression, where the encoder strategically deletes parts of the text and a large language model (LLM) reconstructs the original content from the retained skeleton. We benchmark a progression of deletion strategies, including uniform step deletion, word-length-guided deletion (WordLen), word-frequency-guided deletion (WordFreq), LP-optimized deletion (Opt), entropy-based deletion using GPT-2 surprisal, and hybrid methods that combine frequency and surprisal signals. Evaluation on the BBC News dataset across retention rates \r_keep \in [0.1,0.9] shows three main findings. First, WordFreq is a strong low-cost baseline: despite using only a static frequency lookup, it remains competitive with much more expensive semantic methods while being far faster at the encoder. Second, semantic and hybrid methods provide their clearest gains at mild-to-moderate compression, whereas word-frequency deletion is often more robust at the lowest retention rates. Third, QLoRA fine-tuning yields a strong local decoder that is competitive with Gemini 2.0 Flash and is often strongest in decoder-only comparisons. Additional English and Chinese experiments show that the overall framework transfers across domains, while the best deletion rule remains dataset-dependent.
[NLP-147] Measuring Real-World Prompt Injection Attacks in LLM -based Resume Screening USENIX-SECURITY
【速读】： 该论文试图解决的问题是：生成式 AI（Generative AI）在真实世界应用中，特别是基于大语言模型（Large Language Models, LLMs）的简历筛选系统中，提示注入攻击（prompt injection attacks）的普遍性和实际影响尚不明确，缺乏系统的实证研究。解决方案的关键在于：首先设计针对简历文本特性的提示注入检测方法，并通过小规模人工验证证明其高精度和优于现有通用检测器的性能；随后将该检测器应用于约20万份真实简历数据集进行大规模测量分析，从而首次揭示了提示注入攻击在实际场景中的广泛存在——约1%的简历包含隐藏的提示注入，且过去一到两年内其发生率显著上升，其中90%以上的注入提示未使用显式指令。这一发现为理解并应对现实世界中LLM的安全风险提供了关键证据和研究基础。
链接: https://arxiv.org/abs/2605.28999
作者: Mohan Zhang,Yuqi Jia,Zhen Tan,Steven Jiang,Neil Zhenqiang Gong,Tianlong Chen,Dawn Song
机构: University of North Carolina at Chapel Hill (北卡罗来纳大学教堂山分校); Duke University (杜克大学); Arizona State University (亚利桑那州立大学); hireEZ; University of California, Berkeley (加州大学伯克利分校)
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Published in USENIX Security Symposium 2026; Code and artifacts are available at this https URL
Abstract:LLMs are vulnerable to prompt injection attacks. However, this vulnerability has been primarily demonstrated conceptually in academic studies or through a few anecdotal case studies. Its prevalence and impact in real-world LLM-based applications are largely unexplored. In this work, we present the first systematic study of prompt-injection attacks in a widely used application: LLM-based resume screening. Our analysis is based on approximately 200K real-world resumes collected over multiple years by hireEZ. We first design tailored methods to detect prompt injection in resumes. Manual validation on a small-scale dataset demonstrates that our detectors achieve high precision and outperform state-of-the-art general-purpose detectors. We then apply our detector to the full resume dataset and conduct a comprehensive measurement study of real-world prompt injection attacks. Our analysis reveals several intriguing findings: approximately 1% of resumes contain hidden prompt injections; the prevalence of such injected resumes has increased noticeably over the past one to two years; and more than 90% of injected prompts do not use explicit instructions. These results provide the first evidence of large-scale prompt injection in real-world LLM-based applications and lay the groundwork for future studies to understand and mitigate such attacks.
[NLP-148] CosmicFish-HRM: Adaptive Reasoning via Hierarchical Recurrent Mechanisms in Compact Language Models
【速读】： 该论文试图解决的问题是：如何在保持模型紧凑性的同时提升其推理能力，避免传统大语言模型（Large Language Models, LLMs）因参数量庞大而导致的高计算成本和推理开销。解决方案的关键在于提出了一种名为CosmicFish-HRM的紧凑型语言模型，其核心是一个层次化推理模块（Hierarchical Reasoning Module, HRM），该模块能够在推理过程中动态分配计算资源——即根据输入复杂度自适应地决定何时终止推理循环，从而实现非均匀的推理深度。这种机制使得模型在不同任务和输入上灵活调整推理步数，而非对所有输入执行固定计算量，从而在不依赖大规模参数的情况下优化推理效率与性能。
链接: https://arxiv.org/abs/2605.28919
作者: Venkat Akhil Lakkapragada
机构: Mistyoz AI (Mistyoz AI); Hyderabad, India (海得拉巴，印度)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 17 pages, 4 figures. Exploratory study of adaptive reasoning depth in compact autoregressive language models. Code available at this https URL
Abstract:Large language models have achieved strong reasoning capabilities, though often at the cost of massive parameter counts and expensive inference. In this work, we explore a different direction: adaptive reasoning depth in compact language models. We present CosmicFish-HRM, a compact language model built around a Hierarchical Reasoning Module (HRM) that dynamically allocates computational effort during inference. Instead of applying fixed computation to every input, the model iterates through high-level and low-level reasoning cycles and learns when to halt based on input complexity. CosmicFish-HRM combines this adaptive reasoning core with modern transformer components including Grouped Query Attention, RoPE, and SwiGLU activations. While the additional reasoning infrastructure introduces overhead at small scale, we hypothesize that this tradeoff becomes increasingly favorable as model size grows and the relative cost of the HRM core diminishes. Our results show that the model learns non-uniform reasoning behavior, allocating different numbers of reasoning steps across tasks and inputs. These findings suggest that adaptive reasoning depth may offer a promising alternative to relying solely on parameter scale for reasoning capability.
[NLP-149] Reasoning that Travels: Dissecting How Chain-of-Thought Transfers Across Models
【速读】： 该论文试图解决的问题是：当大型推理模型（Large Reasoning Models, LRMs）生成的链式思维（Chain-of-Thought, CoT）轨迹被其他模型用于任务求解时，这些轨迹如何具体影响接收模型的推理过程和最终答案——即CoT转移背后的机制是什么。解决方案的关键在于构建一个受控的“提供者-接收者”框架，其中提供者生成完整的CoT，接收者则基于逐步增长的CoT前缀进行推理：一种是强制直接作答（force-answer），另一种是允许继续推理后再作答（free-generation）。研究发现，不同任务和模型下CoT转移机制存在显著差异：在force-answer模式中，AIME主要依赖显式答案的可获得性，MMLU-Pro体现接收者能力的重要性，而ZebraLogic则依赖部分结构化答案信息；在free-generation模式中，部分CoT能有效引导接收者继续推理并提升性能；此外，接收者之间答案的一致性可作为无需人工标注的信号，用于提前终止提供者的推理过程。这表明跨模型CoT转移并非单一机制，而是可能表现为答案提取、推理支架或接收者能力依赖等多种形式。
链接: https://arxiv.org/abs/2605.28913
作者: Xinyuan Cheng,Beiduo Chen,Philipp Mondorf,Barbara Plank
机构: MaiNLP, Center for Information and Language Processing, LMU Munich, Germany; Munich Center for Machine Learning, Germany
类目: Computation and Language (cs.CL)
备注: 20 pages, 17 figures
Abstract:Large reasoning models (LRMs) often generate extensive chain-of-thought (CoT) traces before producing a final answer. As explicit textual artifacts, these traces can be passed to other models to solve the same task, enabling cross-model reasoning transfer. Yet successful transfer alone does not reveal how the provided CoT contributes to another model’s answer. We study this question with a controlled provider–receiver framework, where a provider generates a reasoning trace and a receiver solves the same problem from increasingly longer trace prefixes. We compare force-answer, where the receiver answers directly from the prefix, with free-generation, where it may continue reasoning before answering. Across models and benchmarks, full traces often transfer successfully, but prefix trajectories reveal distinct mechanisms. In force-answer mode, AIME transfer is largely driven by explicit answer availability. MMLU-Pro instead reflects a larger role for receiver competence, while ZebraLogic depends on partial structured-answer information rather than complete-answer leakage alone. In free-generation mode, partial CoTs improve performance across benchmarks, indicating that prefixes can guide continued reasoning. Finally, answer agreement among receivers provides a gold-free signal for stopping provider reasoning early. Overall, cross-model CoT transfer is not a single phenomenon: it can reflect answer extraction, reasoning scaffolding, or receiver-dependent competence.
[NLP-150] Hallucination Detection-Guided Preference Optimization for Clinical Summarization
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）在医疗临床笔记摘要任务中普遍存在幻觉（hallucination）的问题，即模型生成不准确或缺乏依据的陈述，从而限制其在专业医疗场景中的可靠性。解决方案的关键在于提出两种协同方法：一是推理时迭代修正方法 \itermodel，通过幻觉检测器引导摘要的逐轮修订以实现事实性修正；二是基于偏好学习的方法 \model，将检测器指导的修正轨迹转化为偏好对用于模型微调。实验表明，这两种方法显著降低了 Llama 和 Gemma 模型在 MIMIC-IV 临床数据上的幻觉率（如 Llama-3.1-8B-Instruct 的幻觉减少达 48%），同时保持摘要的流畅性、连贯性和相关性，为提升临床摘要任务的事实准确性提供了一种自动化且可扩展的解决方案。
链接: https://arxiv.org/abs/2605.28910
作者: Shamanth Kuthpadi Seethakantha,Dung Ngoc Thai,Vara Prasad Gudi,Simran Tiwari,Rami Matar,Avijit Mitra,Wenlong Zhao,Wael Salloum,Andrew McCallum
机构: University of Massachusetts Amherst (马萨诸塞大学阿默斯特分校); Ensemble HP; Columbia College (哥伦比亚学院); University of Massachusetts Amherst (马萨诸塞大学阿默斯特分校)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) have shown promise on summarization tasks, but they often produce hallucinations, which are unsupported or incorrect statements that limit their reliability in specialized healthcare applications. We introduce \itermodelfull (\itermodel), an inference-time method that leverages hallucination detectors to guide iterative summary revisions toward factual corrections. Building on this, we propose \itermodel for Preference Learning (\model), which converts detector-guided refinement trajectories into preference pairs for model finetuning. Extensive experiments show that our methods substantially reduce hallucinations for Llama and Gemma models in summarizing real-world clinical notes from \MimicIV. For example, \itermodel reduces 24% and \model reduces 48% hallucinations in Llama-3.1-8B-Instruct. Importantly, both methods preserve summary fluency, coherence, and relevance according to human expert and LLM-Jury evaluations. Together, these results demonstrate that detection-informed refinement and preference learning offer an automated solution for improving factual faithfulness in clinical summarization.
[NLP-151] GrowLoop: Self-Evolving Conversation Evaluation Seeded by Human
【速读】： 该论文试图解决开放域对话中人类相似性（human-likeness）评估的三大挑战：一是人类判断具有主观性和不一致性，难以形成明确的评价标准；二是现有评估方法无法同时应对人类判断的多样性与动态演进特性；三是评估体系缺乏持续适应模型能力提升和场景变化的能力。解决方案的关键在于提出GrowLoop系统，其核心是通过LLM代理结合启发式学习（Heuristic Learning）实现评价准则（rubric）与案例（case）的协同进化机制，仅需少量人工种子标注即可启动，并在人类与AI达成共识时强化规则，在分歧处保留合理性即可，从而实现评估体系的持续自演化，最终显著优于现有方法并在多维度上揭示模型短板。
链接: https://arxiv.org/abs/2605.28882
作者: Yihang Lin,Yunze Gao,Zeyang Lin,Dongbo Li,Kun Peng,Chenglong Song,Yue Liu
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Sound (cs.SD)
备注:
Abstract:With the rapid advancement of large language models, evaluating human-likeness in open-ended conversation has become increasingly important. However, human-likeness is a form of tacit knowledge that humans perceive intuitively, yet the underlying criteria resist explicit formulation. Human judgments vary widely, with strong agreement on some cases and legitimate disagreement on others. Meanwhile, the criteria behind human judgments remain implicit, leaving no clear basis for constructing cases. Further, what counts as human-like is not static, but evolving with model capability and human expectations. Despite progress in evaluation methods such as expert-authored benchmarks, Reward Models, and self-evolving benchmarks, none addresses all three challenges simultaneously. Therefore, we propose GrowLoop, a self-evolving conversation evaluation system that continuously adapts as models advance and scenarios shift. With minimal human seed annotations as the first mover, LLM agents iteratively extract and refine evaluation rubrics through Heuristic Learning. Human-AI agreement is required where annotators converge, while only plausibility is expected where they diverge. Moreover, the Rubric-Case co-evolution mechanism enables continuous evolution, expanded through new seeds when the evaluation target moves. Applied to human-likeness evaluation in open-ended conversation, the generated rubrics not only substantially outperform existing methods in alignment with human judgments, but also uncover issues that annotators overlook. The resulting benchmark effectively discriminates models across capability tiers and reveals where they fall short, while generalizing to new scenarios and adapting as models advance. Our work shifts the benchmarking paradigm from manual updates or difficulty scaling to comprehensive, continuous self-evolution.
[NLP-152] From Data to Insights: Exploring Program-of-Thoughts Prompting for Chart Summarization
【速读】： 该论文旨在解决图表摘要任务中语义视觉理解与数值推理要求之间的瓶颈问题，即如何在保证统计事实准确性的同时实现高效、轻量化的图表描述。现有视觉语言模型（VLMs）虽取得进展，但缺乏对统计事实正确性的验证机制且计算开销大。其解决方案的关键在于引入零样本学习策略，利用Python程序作为中间媒介，驱动轻量级VLM执行计算推理，从而生成可靠的图表统计数据；同时提出一种新颖的“图表到字典”辅助任务，相比传统“图表到表格”方法更具灵活性，并能有效融合到思维链（Program-of-Thought, PoT）策略中，显著提升图表理解的准确性和鲁棒性。
链接: https://arxiv.org/abs/2605.28874
作者: Yutong Qu,Wei Zhang
机构: Adelaide University (阿德莱德大学)
类目: Computation and Language (cs.CL)
备注: 22 pages, 9 figures
Abstract:Charts play a critical role in conveying numerical data insights through structured visual representations. However, semantic visual understanding and numerical reasoning requirements hinder the accurate description of charts, interpreting a challenging task in chart summarization. Despite recent advancements in visual language models (VLMs), approaches lack robust mechanisms for verifying statistical fact correctness and are computationally heavy. To address this gap, this paper explores a strategy of using zero-shot learning to motivate the lightweight VLMs to perform computational reasoning, via Python programs as intermediaries to derive valid summary statistics for chart understanding. Specifically, we introduce a novel chart-to-dictionary auxiliary task, offering a more flexible representation compared to traditional chart-to-table methods, making it particularly well-suited for integration with the Program-of-Thought (PoT) strategy. Experimental results demonstrate our strategy performs on par with existing chart summarization methods across semantic and factual metrics. Code is available on this https URL.
[NLP-153] he Cognitive Categorical Transformer: Category-Theoretic Inductive Biases for Language Modeling
【速读】： 该论文试图解决的问题是：如何通过引入基于范畴论（category theory）和认知科学启发的结构化组件，提升语言模型在低资源场景下的性能表现，尤其是在不依赖额外训练数据的情况下实现更优的语言建模效果。解决方案的关键在于设计了一种名为认知分类变换器（Cognitive Categorical Transformer, CCT）的架构，其核心创新点是将范畴论中的全 simplicial 消息传递机制（GT-Full simplicial message passing）嵌入到预训练 GPT-2 Small 模型中，从而在 WikiText-103 数据集上实现了 2.92 的困惑度（PPL）下降（相对减少 12%）。进一步的消融实验表明，84% 的性能提升可归因于 GT-Full 组件，这为首次在 306M 参数规模下验证了 simplicial 消息传递对语言建模有效性的实证证据。此外，论文还提出了“结构/一致性区分”（structure/consistency distinction）这一经验模式，指出增加新拓扑结构的范畴先验有助于提升性能，而强制一致性约束的先验则无显著收益。
链接: https://arxiv.org/abs/2605.28864
作者: Al Kari
机构: Manceps Inc.
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:The Cognitive Categorical Transformer (CCT) is a 306M-parameter architecture that augments a pretrained GPT-2 Small backbone with cognitively grounded components derived from category theory and several inspirations from cognitive science. Under a matched-step protocol (215,000 optimizer steps, matched data, matched optimizer and schedule) on WikiText-103, CCT reaches 21.27 validation perplexity, compared with 24.19 for an identically fine-tuned GPT-2 Small baseline. The architecture therefore contributes a 2.92 PPL (12% relative) reduction beyond what in-domain fine-tuning alone provides. A retrain-from-scratch ablation that holds GT-Full simplicial message passing bypassed across the entire seven-phase activation schedule reaches 23.72 PPL, localizing 84% of the architectural improvement (2.45 of 2.92 PPL) to GT-Full. We present the first ablation-validated evidence that simplicial message passing improves language-model perplexity at the 306M-parameter scale on WikiText-103. Published GPT-2 Large reaches 22.05 zero-shot PPL on WikiText-103 with 6.2x more parameters than GPT-2 Small; this paper treats that number as an external published reference, not as the architectural benchmark. Three negative results on consistency-style categorical priors (sheaf smoothing, adjunction round-trip, curvature regularization) and the joint structural-prior result for GT-Full and PrecisionWeightedPP together support an empirical pattern termed the structure/consistency distinction, in which categorical priors that add new topology improve language modeling and those that enforce a consistency identity do not.
[NLP-154] Mechanistic origins of catastrophic forgetting: why RL preserves circuits better than SFT?
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）在微调过程中出现的灾难性遗忘（catastrophic forgetting）问题，即模型在学习新任务时会严重损失原有能力。其解决方案的关键在于从机制层面揭示不同微调方法对内部计算电路（computational circuits）的影响：通过引入“差异电路脆弱性”（differential circuit vulnerability）这一头级指标，量化微调导致电路退化的程度。研究发现，与监督微调（SFT）相比，强化学习（RL）虽适应目标任务较慢，但能显著更好地保留基础模型的内部电路结构，从而更有效地抵御灾难性遗忘。这表明，电路层面的稳定性可能是解释RL优于SFT的关键机制因素。
链接: https://arxiv.org/abs/2605.28860
作者: Jeanmely Rojas Nunez,Viraj Sawant,Nathan Allen,Nomgondalai Amgalanbaatar,Yannis Zongo,Vasu Sharma,Maheep Chaudhary
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Fine-tuning large language models (LLMs) frequently induces catastrophic forgetting of prior capabilities. Recent work has shown that reinforcement learning (RL) retains prior capabilities more effectively than supervised fine-tuning (SFT), attributing this to policy-gradient updates remaining closer to the base policy \citeshenfeld2025rl. We extend this behavioral account to the mechanistic level and ask whether RL’s advantage is mirrored by stronger preservation of internal computational circuits. We introduce differential circuit vulnerability, a head-level measure of how much a circuit degrades under fine-tuning, and use it to compare RL and SFT on Qwen2.5-3B-Instruct adapted to scientific question-answering. We find a clear mechanistic trade-off: SFT adapts more rapidly to the target task but produces substantially greater circuit disruption and forgetting of prior capabilities, whereas RL preserves a larger fraction of the base circuit at the cost of slower task adaptation. These findings suggest that circuit preservation may help explain why RL is more robust to catastrophic forgetting. We released our code here: this https URL.
[NLP-155] Large language models reorganize representational geometry during in-context learning
【速读】： 该论文试图解决的问题是：尽管大语言模型（LLMs）具备无需参数更新即可通过上下文示例实现任务适应的能力（即上下文学习，ICL），但其背后的机制仍不清晰，尤其是高维表示空间的几何结构如何影响ICL的有效性。解决方案的关键在于提出并验证一个假设——ICL依赖于任务相关表示在在线过程中的“解缠”（untangling），即通过几何重构提升分类可分性。研究通过设计基于模型内部已知结构表示的分类任务，发现ICL性能与任务的表示结构密切相关，且成功ICL伴随着表示空间的几何重组，使类间距离增大、可分性增强；进一步表明LLM的行为可以用一种原型类算法来描述，该算法在整合证据的同时重塑表示以支持分类。这一工作首次从几何视角揭示了ICL的机制约束，量化了预训练表示能力与ICL实际利用之间的差距。
链接: https://arxiv.org/abs/2605.28854
作者: Hua-Dong Xiong,Li Ji-An,Robert C. Wilson,Kwonjoon Lee,Xue-Xin Wei
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
备注:
Abstract:Large language models (LLMs) exhibit remarkable flexibility: they can adapt to novel tasks from in-context examples without any parameter updates, a capability known as in-context learning (ICL). Prior work on synthetic tasks has shown that ICL can implement specific algorithms, demonstrating architectural competence, and mechanistic analyses have identified key circuits that support this behavior. However, because in-context computation – regardless of its algorithmic form – relies on transformations in high-dimensional representation space, it remains unclear how the geometry of that space shapes ICL effectiveness. Motivated by the neuroscience view of classification as the untangling of neural representations, we hypothesize that ICL depends on the successful online untangling of task-relevant representations. To test this idea, we study how LLMs classify in-context examples whose labels are defined by the model’s own internal representations with known structure. We show that ICL performance correlates systematically with the representational structure of the underlying classification task and that successful ICL is accompanied by geometric reorganization that increases online separability. We further find that LLM behavior is well described by a prototype-like algorithm that integrates evidence while reshaping representations to support classification. These findings offer a geometric account of ICL in pretrained LLMs, establish representational geometry as a mechanistic constraint on ICL, and quantify the gap between what pretrained representations afford and what in-context learning can exploit.
[NLP-156] GPF-LiveNews: A Streaming Evaluation Protocol for Group-Conditioned Framing in Large Language Models
【速读】： 该论文试图解决的问题是：在非平稳环境中，大型语言模型（LLM）的输出框架如何随时间变化，并且不同群体在被提示时是否会受到差异化的影响。现有静态偏见基准无法捕捉模型对新兴事件的群体条件性框架（group-conditioned framing）变化。解决方案的关键在于提出一种名为 GPF-LIVENEWS 的流式评估协议和基准快照，通过扩展 BBC/Reuters 新闻锚点覆盖 42 个身份标签和 7 种提示族，利用语义敏感性（semantic-sensitivity）和情感差异性（sentiment-disparity）信号来评估模型输出包。实验表明，Policy/Action 类提示引发最强语义变动，而情感差异则相对稳定；所有分数均被视为观察窗口内的审计信号，供人工审查，而非永久公平性排名或有害偏见的直接证据。
链接: https://arxiv.org/abs/2605.28848
作者: Mohd Ariful Haque,Fahad Rahman,Kishor Datta Gupta,Roy George
机构: Clark Atlanta University; United International University
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Deployed language models are evaluated in a non-stationary environment: model versions, retrieval layers, safety systems, and real-world inputs all change over time. Static bias benchmarks remain useful, but they do not show how models frame newly emerging events for different prompted audiences. We introduce GPF-LIVENEWS, a streaming evaluation protocol and benchmark snapshot for auditing group-conditioned framing in open-ended LLM outputs. The protocol expands fresh BBC/Reuters news anchors across 42 identity labels and seven prompt families, then evaluates response bundles using semantic-sensitivity and sentiment-disparity signals. In a pilot over 12 monitoring runs and 23 hosted models, Policy/Action prompts produce the strongest semantic movement, while sentiment variation is flatter across dimensions and prompt families. The released artifact includes article metadata, prompt templates, instantiated prompts, model-output metadata, score tables, documentation, and reproduction scripts. We interpret all scores as observed-window audit signals for human review, not as permanent fairness rankings or direct proof of harmful bias.
[NLP-157] houghts-as-Planning : Latent World Models for Chain-of-Thoughts Optimization via Reinforcement Planning
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）在多样化自然语言处理任务中，推理链（reasoning chain）优化缺乏可解释性、泛化能力差和样本效率低的问题。现有方法多依赖黑箱启发式规则或无梯度搜索，难以有效对齐模型行为与任务目标。解决方案的关键在于提出一种名为“Thoughts-as-Planning”的新框架，将推理链优化建模为在潜在语义空间中的序贯决策过程：通过构建一个部分可观测环境下的潜在世界模型（latent world model），模拟推理链编辑对下游输出的影响；同时设计保持邻近性的嵌入空间以编码推理链-响应动态关系，从而支持基于梯度下降或强化学习的规划策略，并实现从词元、片段到指令层级的多尺度抽象统一整合。实验表明，该方法在效率、鲁棒性和泛化性能上优于当前最优基线，且其结构化的规划轨迹提供了良好的可解释性。
链接: https://arxiv.org/abs/2605.28842
作者: Dong Liu,Yanxuan Yu,Ying Nian Wu
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:The success of large language models (LLMs) across diverse NLP tasks has elevated the importance of reasoning chain optimization as a critical step in aligning model behavior with task objectives. Existing reasoning chain tuning methods often rely on black-box heuristics or gradient-free search, which lack interpretability, generalization, and sample efficiency. In this work, we introduce \textbfThoughts-as-Planning, a novel framework that formalizes reasoning chain optimization as a sequential decision-making process over a latent semantic space. We model the LLM as a partially observable environment and learn a latent world model that simulates the effect of reasoning chain edits on downstream outputs. A proximity-preserving embedding space is constructed to encode reasoning chain-response dynamics, enabling planning via gradient descent or reinforcement learning. Our method supports multi-scale abstraction, allowing reasoning chain edits at token, segment, and instruction levels to be integrated into a unified planner. Through extensive experiments on language understanding and generation tasks, we demonstrate that Thoughts-as-Planning outperforms state-of-the-art reasoning chain tuning baselines in efficiency, robustness, and generalization, while offering interpretability through its structured planning trajectory. Our code is available at this https URL.
[NLP-158] How Consistent Are LLM Agents ? Measuring Behavioral Reproducibility in Multi-Step Tool-Calling Pipelines
【速读】： 该论文试图解决的问题是：具备工具调用能力的大语言模型（LLM）代理在生产环境中部署时，其行为是否具有可重复性——即在相同输入条件下，代理是否会始终选择相同的工具、以相同的顺序调用，并传递相同的参数。以往研究主要集中在基于搜索或自由文本动作的ReAct类代理的一致性问题，而本文则聚焦于更复杂的结构化工具调用接口场景，其中工具具有类型化的参数和可能产生实际副作用（consequential side effects）。解决方案的关键在于提出了一种系统性的实证研究方法，用于量化评估多步工具调用代理在重复执行相同任务时的行为一致性，从而揭示当前LLM代理在可靠性方面的潜在缺陷。
链接: https://arxiv.org/abs/2605.28840
作者: Abel Yagubyan
机构: OpenAI; Anthropic; Meta/Together AI
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 16 pages, 6 figures
Abstract:Large language model (LLM) agents with tool-calling capabilities are increasingly deployed in production systems, yet a fundamental reliability question remains under-explored: does the same agent behave the same way twice? We present a systematic empirical study of behavioral consistency in multi-step tool-calling agents, measuring whether agents select the same tools, in the same order, with the same arguments, across repeated identical invocations. Unlike prior work on consistency in ReAct-style agents(search-only, free-text actions), we study the richer setting of structured tool-calling interfaces with typed parameters and consequential side effects.
[NLP-159] Specialty-Specific Medical Language Model for Immune-Mediated Diseases
【速读】： 该论文旨在解决从自由文本医学叙述中提取详细临床信息的难题，尤其针对免疫介导疾病和感染性疾病术语在不同来源间不一致的问题，这限制了通用自然语言处理（Natural Language Processing, NLP）系统对生物医学概念的细粒度识别能力。解决方案的关键在于构建一个面向免疫学与感染病领域的专用命名实体识别（Named Entity Recognition, NER）模型：研究团队联合两位临床专家手工标注了371份病例报告，定义了涵盖免疫介导和感染性疾病的12类实体，包括症状和临床描述；通过对比多种建模策略（如MedicalNER架构、BERT-based token分类模型及零样本NER系统），发现基于Transformer且使用临床领域嵌入训练的模型表现最优（F1分数达0.89），显著优于基线和零样本方法；其成功关键在于专业嵌入与专家标注的结合，有效捕捉了疾病术语的细微差异并提升了跨异构生物医学文本的泛化能力。
链接: https://arxiv.org/abs/2605.28838
作者: Veysel Kocaman,Gursev Pirge,Yigit Gul,Ace Vo,Zhenya Nargizyan,David Talby
机构: John Snow Labs Inc. (John Snow Labs 公司)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 15 pages, 5 figures. Funded in part by NIAID/NIH under contract 75N93024C00010
Abstract:Extracting detailed clinical information from free-text medical narratives remains a practical challenge for researchers and healthcare systems. Terminology for immune-mediated and infectious diseases is especially inconsistent across sources, which often limits the ability of general-purpose Natural Language Processing (NLP) systems to capture the relevant biomedical concepts with sufficient granularity. We developed a domain-specific Named Entity Recognition (NER) model tailored to identify disease-related entities occurring in immunology and infectious disease contexts. We assembled and manually annotated a dataset of 371 case reports in collaboration with two clinical specialists, defining twelve entity classes covering immune-mediated and infectious conditions as well as related symptoms and clinical descriptors. We evaluated several modeling strategies, including the MedicalNER architecture with multiple healthcare-specific embeddings, a BERT-based token classification model, and zero-shot NER systems. The strongest performance was obtained with a transformer-based model trained on clinical-domain embeddings, which reached an F1 score of 0.89, consistently outperforming baseline and zero-shot approaches. The combination of specialized embeddings and expert annotation proved particularly valuable for capturing nuanced disease terminology and improving generalization across heterogeneous biomedical text. The prompted LLM baseline achieved substantially lower performance under the same evaluation protocol, reflecting difficulties in producing span-consistent outputs for fine-grained entity boundaries despite detailed prompting. The resulting model provides a structured way to analyze case reports and can support downstream tasks such as cohort identification, disease monitoring, and clinical decision support.
[NLP-160] SERC: LDPC-Inspired Semantic Error Correction for Retrieval-Augmented Generation ICPR2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在生成过程中普遍存在幻觉（hallucination）问题，即模型输出与事实不符的现象。现有内在自校正方法因自我偏见（self-bias）而效果有限，难以有效识别自身错误。其解决方案的关键在于提出一种受低密度奇偶校验码（LDPC）启发的语义错误校正方法（SERC），将文本生成建模为语义噪声信道，把生成内容视为被噪声污染的码字；通过稀疏验证策略——仅生成少量低密度验证查询并结合外部证据进行验证——实现高效且精准的错误检测与修正。实验表明，SERC在LongForm Bio和TruthfulQA基准上显著优于传统自校正和检索增强基线方法，在事实准确性（FactScore）方面提升明显，并使小型语言模型（SLMs）超越大型基线模型的幻觉抑制能力，同时大幅降低验证开销，是一种无需训练、模型无关且资源受限场景下成本与保真度最优平衡的解决方案。
链接: https://arxiv.org/abs/2605.28837
作者: Gyumin Kim,Juhwan Park,Jaeha Kim,Seunggyun Han,Kyungrak Son,Ikbeom Jang
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 15 pages, 2 figures, 6 tables. To appear in the Proceedings of the 28th International Conference on Pattern Recognition (ICPR 2026). Code available at this https URL
Abstract:While Large Language Models (LLMs) have demonstrated remarkable capabilities, their reliability is significantly compromised by hallucinations. Existing intrinsic self-correction methods attempt to address this, but often fail due to self-bias, where models struggle to identify errors in their own outputs without external verification. To overcome these limitations, we propose the LDPC-inspired semantic error correction for retrieval-augmented generation (SERC), providing a theoretical framework to interpret and mitigate LLM hallucinations. We reformulate the text generation process as a semantic noisy channel, treating generated responses as noise-corrupted codewords. Inspired by low-density parity-check (LDPC) codes, SERC employs a sparse verification strategy: instead of exhaustively checking all facts, it generates low-density verification queries and validates them against external evidence to efficiently detect and correct errors. We evaluate SERC on LongForm Bio and TruthfulQA benchmarks using Llama-3-8B and Qwen2.5-14B. Experimental results demonstrate that SERC outperforms both intrinsic self-correction methods and strong retrieval-augmented baselines, demonstrating significant gains especially in factual precision (FactScore). Notably, SERC enables small language models (SLMs) to surpass the performance of larger baselines in hallucination reduction and information preservation. Our findings demonstrate that SERC provides a training-free, model-agnostic solution that significantly reduces verification overhead compared to dense methods, achieving an optimal trade-off between cost and fidelity in resource-constrained environments.
[NLP-161] No Reader Left Behind: Multi-Agent Summaries Everyone Can Understand
【速读】： 该论文试图解决政府文件在传达信息时因语言复杂性导致普通公众难以理解的问题，尤其是在面对不同读者群体（如小学生、非母语者和注意力缺陷读者）时存在的多样化语言与认知障碍。解决方案的关键在于提出NRLB（No Reader Left Behind）多智能体框架，该框架通过模拟三类代表性读者群体，结合基于模板的规划与迭代式的读者导向优化机制，系统性地识别并解决难词、缺失背景和混淆句等问题，从而在保持事实准确性的同时显著提升可读性。
链接: https://arxiv.org/abs/2605.28836
作者: Jimin Jung,MyoungJin Kim,Jaehyung Seo,Heuiseok Lim
机构: Korea University (韩国大学); Konkuk University (中央大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:The Plain Writing Act in the United States requires government documents to be accessible in clear and simple language that the general public can easily understand, yet existing summarization systems struggle to address diverse linguistic and cognitive barriers among general readers. We present NRLB (No Reader Left Behind), a multi-agent framework for plain language summarization that simulates three representative reader groups: elementary school student readers, non-native readers, and readers with attention deficits. NRLB combines template-based planning with iterative, reader-oriented refinement, enabling systematic detection and resolution of difficult terms, missing contexts, and confusing sentences. Evaluations across multiple datasets demonstrate consistent improvements in readability while preserving factual accuracy. Human evaluation further validates NRLB’s impact, with annotator preference rates ranging from 55% to 76%, highlighting NRLB’s potential to produce plain language summaries that are both faithful to the source and broadly accessible to the general public.
[NLP-162] GenesisFunc: Multi-Agent Data Generation for Accurate and Generalizable Function-Calling ACL2026
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）在函数调用（Function-Calling, FC）能力训练中面临的数据难题，即高质量、多样化且覆盖广泛场景的真实函数调用数据难以获取，而现有合成数据生成方法常受限于不可靠的API、工具扩展性差、多样性不足以及质量控制薄弱等问题。解决方案的关键在于提出GenesisFunc——一个自动化函数调用训练数据生成管道，其核心创新包括：基于公开基准中的可靠工具构建多智能体对话生成框架，以生成覆盖多样场景的高质量对话数据；并通过多阶段评估体系保障数据准确性；最终在8B规模模型上微调后，实验表明该合成数据显著提升模型在域内函数调用性能和域外泛化能力，同时达到与最新基于API的模型相当的函数调用水平，并展现出良好的下游工具可扩展性。
链接: https://arxiv.org/abs/2605.28835
作者: Hao-Xiang Xu,Chong Deng,Jiaqing Liu,Wen Wang,Qian Chen,Lujia Bao,Xiangang Li,Zhen-Hua Ling
机构: Alibaba Group (阿里巴巴集团)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted by ACL 2026 Main
Abstract:Large Language Models (LLMs) extend their capabilities through function-calling (FC), which relies on training data with high quality, diversity, and broad coverage of scenario. However, obtaining and annotating real function-calling data is challenging, while synthetic data from existing pipelines often suffers from unreliable APIs, limited tool scalability, insufficient diversity, and weak quality control. To address these, we present GenesisFunc, an automated pipeline for generating FC training data. Starting from reliable tools in widely used public benchmarks, our GenesisFunc employs a multi-agent framework to support a dialogue generation system that produces conversations spanning diverse scenarios, while maintaining both diversity and quality throughout the process. The accuracy of the data is further reinforced through a multi-stage evaluation system. We fine-tune an 8B LLM on the synthetic dataset and show through extensive experiments that it outperforms similarly sized open-source models in in-domain FC performance and out-of-domain generalization, while reaching FC capabilities comparable to some of the latest API-based models. In addition, our method demonstrates strong potential to scale effectively across downstream tools, underscoring its real-world applicability.
[NLP-163] Assessing Dutch Syllabification Algorithms and Improving Accuracy by Combining Phonetic and Orthographic Information through Deep Learning
【速读】： 该论文旨在解决荷兰语音节划分（syllabification）中准确率不足的问题，特别是现有算法在处理不同词类（如词典词、外来词、伪词）时表现不一，且缺乏对基于深度学习的现代方法以及语音信息与拼写信息结合效果的系统评估。解决方案的关键在于：首先，通过对比四种算法（包括一个新提出的深度学习模型）在三个数据集上的表现，验证了数据驱动方法优于传统知识驱动方法；其次，创新性地将语音信息与拼写信息融合进统一模型中，实证表明这种结合能有效提升音节划分性能（达到99.65%的词级准确率，较文献最优结果提升0.14%），尤其在拼写歧义可通过发音信息澄清的词汇上效果显著。这一方法为未来跨语言音节划分研究和语音辅助拼写处理提供了新路径。
链接: https://arxiv.org/abs/2605.28834
作者: Gus Lathouwers,Wieke Harmsen,Catia Cucchiarini,Helmer Strik
机构: Radboud University (拉德布德大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Published in CLIN Journal
Abstract:Syllabification describes the task of dividing words into syllables. Due to many rules and exceptions, training an algorithm to perform syllabification with high accuracy remains a challenge. Throughout the last decades, different algorithms have been put forth for Dutch syllabification, yet a comprehensive comparative assessment has not been done. Additionally, deep learning has gained significant popularity within NLP in recent years, yet no modern deep-learning based framework has been developed for Dutch orthographic syllabification. Finally, phonetic and orthographic syllabification algorithms have been examined separately, but not in combination. The aim of the current research was twofold: (a) to examine the performance of existing Dutch syllabification algorithms, and (b) to investigate whether combining phonetic and orthographic information into a single model can increase syllabification performance. To compare the performance of algorithms, four algorithms (Brandt Corstius, Liang, Trogkanis-Elkan (CRF), and a newly conceived deep-learning model) were applied to three different datasets (dictionary words, loanwords, pseudowords). The algorithms show varying performance across datasets, with the data-driven algorithms outperforming a knowledge-based algorithm in all but one condition. The new deep-learning methods developed led to increased performance compared to the best found in the literature (99.65% word accuracy, a 0.14% improvement). An analysis of the words for which adding phonetic information improved syllabification performance indicates that these were words in which the orthographic ambiguity could be resolved by information on pronunciation. Future research could examine other areas where phonetic information can benefit orthographic processing. In addition, the newly developed deep learning frameworks can be applied to other languages than Dutch.
[NLP-164] ranscribing Childrens Speech: ASR Performance and Obtaining Reliable Orthographic Transcriptions
【速读】： 该论文试图解决在低资源语言中，如何获得高质量自动语音识别（ASR）转录以减少儿童语音研究中人工标注负担的问题。当前挑战主要源于儿童语音数据的多样性、噪声环境复杂以及缺乏针对儿童语音的预训练模型。解决方案的关键在于：首先，通过对比评估三种主流ASR模型家族（Whisper、Parakeet和Wav2Vec2）在两个荷兰儿童语音数据集（JASMIN和DART）上的表现，发现微调后的Whisper-medium模型在JASMIN上达到5.54%的词错误率（WER），但在噪声更严重的DART数据集上性能显著下降（WER=70.37%）；其次，提出一种基于逐句级别的选择方法，利用ASR输出与原始朗读提示的比对来识别发音正确的语句，从而实现无需人工验证即可自动筛选出高置信度正确转录的语句——该方法在JASMIN和DART上分别可识别42.0%和18.1%的语句，且精确度高于98.3%，大幅降低后续人工校验需求。
链接: https://arxiv.org/abs/2605.28833
作者: Gus Lathouwers,Lingyun Gao,Catia Cucchiarini,Helmer Strik
机构: Radboud University (拉德堡德大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Automatic speech recognition (ASR) has the potential to substantially reduce manual annotation effort in child speech research by generating automatic transcriptions. However, obtaining reliably high-quality ASR transcriptions for child speech remains challenging in low-resource languages due to limited child-specific pre-trained models and highly diverse noise conditions. This study investigates the effectiveness of state-of-the-art ASR models on child speech through two research questions, by evaluating nine ASR models from three model families (Whisper, Parakeet, and Wav2Vec2) on two Dutch child speech datasets, JASMIN and DART. Research question 1 examines the performance of ASR-models applied to child speech. The fine-tuned Whisper-medium model achieves the best overall performance, with a WER of 5.54% on JASMIN and 70.37% on DART, showing that the noisy DART data are clearly more challenging. Research question 2 examines to what extent it is possible to select a subset for which reliable orthographic transcriptions can be obtained automatically, without the need for manual verification. We use an utterance-level selection method that compares ASR output with the original read prompt to identify correctly pronounced recordings. Using the proposed selection method, 42.0% [for JASMIN] and 18.1% [for DART] of the utterances can be automatically identified as correctly pronounced with high confidence, resulting in very low error rates on an utterance level (precisions of 98.3% and higher) and reducing the need for manual verification.
[NLP-165] A comparative study of transformer-based embeddings for topic coherence
【速读】： 该论文试图解决的问题是：在基于Transformer的文本表示模型中，模型规模（参数量）对主题建模质量的影响是否显著。传统方法如LDA虽然可解释性强，但受限于词共现模式的表达能力；而近年来兴起的预训练语言模型（如BERT、LLaMA等）提供了更丰富的文档表征，但其性能是否随模型规模提升而显著改善仍不明确。解决方案的关键在于系统性地评估七个不同规模的Transformer模型（从2200万到130亿参数）在BERTopic框架下的主题质量表现，使用Röder等人（2015）提出的一致性（coherence）和差异性（divergence）指标进行量化评估。研究发现，模型规模对主题质量的影响微乎其微，表明较小模型即可达到与大型模型相当的主题建模效果，从而为高效部署提供理论依据。
链接: https://arxiv.org/abs/2605.28832
作者: Alex Ding,Tarun Rapaka,Willy Rodriguez,Jason Yang
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Topic modeling is a branch of Natural Language Processing (NLP) that aims to organize large collections of texts into coherent groups according to word co-occurrence patterns, with Latent Dirichlet Allocation (LDA) remaining one of the most widely used and interpretable probabilistic approaches. Recent advances in NLP, particularly transformer-based language models, offer improved document representations. It is also known that the size of the model (in terms of number of parameters) has a significant impact in the performance of the language models on different pre-defined tasks. In this study, we systematically examine the effect of model size on topic quality by analyzing the performances of seven transformer-based language models (from small models such as MiniLM to large ones such as LLaMA-2) in a BERTopic pipeline on a variety of corpora. Topic quality is evaluated using coherence and divergence metrics following Röder et al. (2015). Our results indicate that model size, ranging from 22 million to 13 billion parameters, has a negligible impact on the quality of the topic, suggesting that smaller models can achieve comparable performance to larger models.
[NLP-166] S3Mem: Structured Spatiotemporal Scene-Event Memory for Long-Horizon Interactive Question Answering
【速读】： 该论文试图解决长时交互智能体在处理长时间轨迹历史时，难以可靠回答早期事件相关问题的问题。其核心瓶颈并非上下文长度限制，而是长期记忆中轨迹与答案之间的接口设计不合理——传统基于纯文本块存储和标准检索增强生成（RAG）的方法常检索到局部相关但链路不完整的证据，尤其在空间、时间、重复事件及多跳状态类问题上表现不佳。解决方案的关键在于提出S3MEM（Structured Scene-Event Episodic Memory），它将轨迹写入结构化记忆单元，通过锚点敏感检索获取证据，并在回答阶段提供紧凑且token预算感知的证据接口，从而实现轨迹到查询对齐的支持。实验表明，S3MEM在四个不同环境中均显著优于基线方法，在准确率与效率之间取得更优平衡，验证了结构化书写与锚点敏感证据路由相较于通用记忆接口的优势。
链接: https://arxiv.org/abs/2605.28831
作者: Encheng Su,Jinouwen Zhang,Jianyu Wu,Qiucheng Yu,Chen Tang,Pengze Li,Lintao Wang,Yizhou Wang,Xinzhu Ma,Shixiang Tang,Aoran Wang
机构: University of Science and Technology of China (中国科学技术大学); Shanghai Jiao Tong University (上海交通大学); Shanghai AI Laboratory (上海人工智能实验室); City University of Hong Kong (香港城市大学); The Chinese University of Hong Kong (香港中文大学); Fudan University (复旦大学); The University of Sydney (悉尼大学); Beihang University (北京航空航天大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Long-horizon interactive agents often accumulate large trajectory histories yet still fail to answer questions about earlier events reliably. We argue that the main bottleneck is not context length alone, but the trajectory-to-answer interface of long-term memory. When histories are stored as plain-text chunks and queried with standard retrieval-augmented generation (RAG), systems often retrieve locally relevant but chain-incomplete evidence, especially for spatial, temporal, repeated-event, and multi-hop state questions. We propose S3MEM, a structured scene-event episodic memory framework for long-horizon interactive question answering (QA). S3MEM writes trajectories into structured memory units, retrieves evidence through anchor-sensitive retrieval, and exposes a compact token-budget-aware evidence interface for answer-time inference. In this sense, S3MEM is a structured evidence harness that converts agent trajectories into query-aligned support. We evaluate S3MEM on two internal headline environments (Crafter, Jericho) and two out-of-family environments (SciWorld, ALFWorld). Under a shared frozen answer-time protocol, S3MEM consistently outperforms Vanilla RAG across all four environments, surpasses Graph-NoReader on Crafter, Jericho, and ALFWorld, and matches it on SciWorld while using dramatically fewer evidence tokens. Three adapted recent baselines – A-MEM-inspired, MemoryOS-adapted, and LightMem-adapted – improve over Vanilla RAG in several settings, but none matches S3MEM’s overall accuracy-efficiency frontier. Overall, the evidence supports a bounded conclusion: under the current frozen answer-time protocol, structured writing and anchor-sensitive evidence routing provide a stronger accuracy-efficiency frontier for long-horizon interactive QA than more generic memory interfaces.
[NLP-167] Benchmarking Open-Source Safety Guard Models: A Comprehensive Evaluation
【速读】： 该论文试图解决在安全关键型应用中，如何有效评估和选择生成式 AI (Generative AI) 安全防护模型（safety guard models）的问题。其核心挑战在于：现有模型在真实场景下对有害内容的检出能力不足，且缺乏统一、全面的基准测试体系来衡量其性能。解决方案的关键在于构建一个覆盖8类NIST人工智能风险框架（NIST AI Risk Framework）安全类别、包含79,331条样本的精细化评测基准，并系统性地比较14个开源安全防护模型的表现。研究发现，召回率（recall）是安全应用中最关键的指标，因为漏检有害内容的风险远高于误报；同时，模型规模与安全检测性能无显著相关性，甚至较小模型如Qwen Guard（4B参数）在召回率上优于更大规模的Llama Guard（12B）和GPT-OSS Safeguard（20B），且通用型防护模型表现优于专用模型。这些结果为生产环境中安全防护模型的实际选型提供了实证依据。
链接: https://arxiv.org/abs/2605.28830
作者: Reetu Raj Harsh,Bhaskarjit Sarmah,Stefano Pasquali
机构: Domyn; Gurugram, India; New York, USA
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:As Large Language Models (LLMs) are increasingly deployed in safety-critical applications, robust content moderation becomes essential. We present a comprehensive evaluation of 14 open-source safety guard models on a curated benchmark of 79,331 samples spanning 8 NIST AI Risk Framework safety categories. Our benchmark aggregates four diverse datasets (HarmBench, StrongREJECT, RealToxicityPrompts, and BeaverTails), filtered to focus exclusively on safety-relevant content (violence, hate speech, harassment, sexual content, suicide/self-harm, profanity, threats, and health misinformation). We find that recall is the critical metric for safety applications, as missing harmful content poses greater risk than false positives. Our evaluation reveals surprising results: Qwen Guard (4B parameters) achieves the highest recall (83.97%) while larger models like Llama Guard (12B) and GPT-OSS Safeguard (20B) exhibit conservative behavior, missing up to 75% of unsafe content. We demonstrate that model size does not correlate with safety detection performance and that general-purpose guard models outperform specialized ones. These findings provide practical guidance for selecting safety guard models in production deployments.
[NLP-168] Aryabhata 2: Scaling Reinforcement Learning for Advanced STEM Reasoning
【速读】： 该论文旨在解决当前大型语言模型在应对竞争性STEM考试（如JEE和NEET）时面临的挑战，即如何实现多步骤符号推理、精确数值计算与跨学科深度理解的高效协同，同时满足大规模部署中对领域特定且结构一致的问题求解需求。其解决方案的关键在于提出Aryabhata 2——一个专注于推理能力的专用语言模型，通过强化学习后训练（reinforcement-learning post-training）进行优化，并基于PhysicsWallah内部题库构建高质量训练课程；训练过程中采用渐进式扩大的rollout群体规模以增强探索广度，并结合可验证奖励机制提升推理准确性。实验表明，Aryabhata 2在JEE、NEET等考试基准及AIME、GPQA等分布外推理数据集上均优于基线模型GPT-OSS-20B，且输出token数量减少高达64%，显著提升了推理效率与实用性。
链接: https://arxiv.org/abs/2605.28829
作者: Ritvik Rastogi,Vishal Singh,Tejas Chaudhari,Sandeep Varma
机构: PhysicsWallah
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注:
Abstract:Competitive STEM examinations such as JEE and NEET require multi-step symbolic reasoning, precise numerical computation, and deep conceptual understanding across physics, chemistry, and mathematics. Recent large language models perform strongly on common reasoning benchmarks, yet they remain difficult to deploy at scale, where millions of student doubts demand domain-specific, consistently structured problem solving. We introduce Aryabhata 2, a reasoning-focused language model for competitive STEM examinations, trained via reinforcement-learning post-training. Using PhysicsWallah’s internal question banks, we construct a high-quality training curriculum and post-train GPT-OSS-20B through reinforcement learning with verifiable rewards. Training combines prolonged reinforcement learning with broadened exploration via progressively larger rollout group sizes. We evaluate Aryabhata 2 on competitive examination benchmarks, including JEE Main, JEE Advanced, and NEET, as well as out-of-distribution reasoning datasets such as AIME, HMMT, MMLU-Pro, MMLU-Redux 2.0, and GPQA. Results show that Aryabhata 2 outperforms its base model GPT-OSS-20B on competitive STEM reasoning while requiring substantially fewer output tokens (up to 64% fewer). Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY) Cite as: arXiv:2605.28829 [cs.CL] (or arXiv:2605.28829v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2605.28829 Focus to learn more arXiv-issued DOI via DataCite
[NLP-169] Micro-Macro Retrieval: Reducing Long-Form Hallucination in Large Language Models
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）在长文本生成任务中因冗余检索内容和过长推理链导致的事实性错误（即幻觉）问题。现有检索增强语言模型（Retrieval-Augmented Language Models, RALMs）无法确保关键信息与模型输出之间的空间邻近性，从而难以抑制幻觉。解决方案的关键在于提出一种名为“微观-宏观检索”（Micro-Macro Retrieval, M2R）的新颖“边生成边检索”框架：在宏观层面从外部源检索粗粒度证据，在微观层面利用推理过程中构建的关键信息仓库提取并复用核心事实，从而直接缓解关键信息与输出间的距离瓶颈。M2R通过基于课程学习的强化学习策略进行训练，采用定制化的规则奖励机制，稳定地习得检索与事实对齐能力，实验证明其在长上下文场景下显著降低幻觉并提升生成准确性。
链接: https://arxiv.org/abs/2605.28828
作者: Yujie Feng,Jian Li,Zhihan Zhou,Pengfei Xu,Yujia Zhang,Xiaoyu Li,Xiaohui Zhou,Alan Zhao,Xi Chen,Xiao-Ming Wu
机构: Solar System of OVB, Tencent(腾讯); The Hong Kong Polytechnic University(香港理工大学); Jilin University(吉林大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models (LLMs) achieve impressive performance across many tasks but remain prone to hallucination, especially in long-form generation where redundant retrieved contexts and lengthy reasoning chains amplify factual errors. Recent studies highlight a critical phenomenon: the closer key information appears to the model outputs, the higher the factual accuracy. However, existing retrieval-augmented language models (RALMs) lack effective mechanisms to ensure this proximity - external evidence is injected into reasoning via multi-turn retrieval, but this cannot ensure key information stays close to the outputs. We propose Micro-Macro Retrieval (M2R), a novel retrieve-while-generate framework to fill this gap. At the macro level, M2R retrieves coarse-grained evidence from external sources; at the micro level, it extracts essential results from a key information repository built during reasoning and reuses them while generating answers. This design directly addresses the key-information-to-output proximity bottleneck, effectively reducing hallucination in long-form tasks. M2R is trained with a curriculum learning-based reinforcement learning strategy using customized rule-based rewards, enabling stable acquisition of retrieval and grounding skills. Extensive experiments across different benchmarks demonstrate the effectiveness of M2R, especially in lengthy-context settings.
[NLP-170] RightNow-Arabic-0.5B-Turbo: An Open Sub-1B Arabic Language Model via Vocabulary Injection and Edge-First Deployment
【速读】： 该论文试图解决当前开源阿拉伯语大语言模型（Arabic Large Language Models, Arabic LLMs）在性能与资源消耗之间难以平衡的问题：现有模型要么是参数量较小但阿拉伯语能力弱的多语言模型（如Qwen2.5-0.5B、Falcon-H1-0.5B），要么是高性能但需服务器部署的大型专用模型（如Jais、SILMA），而此前唯一尝试构建的小规模阿拉伯语专用模型（Kuwain-1.5B）并未公开权重。解决方案的关键在于提出一个轻量级、高性能且可部署于边缘设备的阿拉伯语专用模型RightNow-Arabic-0.5B-Turbo，其基于Qwen2.5-0.5B进行增量预训练（使用27,032个阿拉伯token进行mean-subtoken初始化，再在5.04亿阿拉伯token上用FSDP并行、FlashAttention变长打包和Liger融合核继续预训练），随后通过监督微调（129,116条指令对，响应掩码损失）、直接偏好优化（6,750对偏好样本）及权重汤融合（weight soup merging）三阶段训练。最终模型在三个阿拉伯语基准测试中平均准确率达35.9%，优于同类开源模型，在COPA-ar任务上达到Falcon-H1-1.5B的水平（58.4%），仅为其三分之一参数量，并恢复了SILMA-9B 67%的性能，同时量化后体积仅为398 MB（q4_k_m），单张H100显卡下推理速度达635 tokens/s，显著提升了小规模阿拉伯语模型的实用性和可访问性。
链接: https://arxiv.org/abs/2605.28827
作者: Jaber Jaber,Osama Jaber
机构: RightNow AI
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 12 pages, 7 tables, 4 figures, 1 algorithm. Weights: this https URL
Abstract:Open Arabic large language models split into two classes: sub-1B multilingual models that treat Arabic as an afterthought (Qwen2.5-0.5B, Falcon-H1-0.5B), and 7B-70B Arabic-specialized models that require a server to run (Jais, AceGPT, ALLaM, SILMA). The one published attempt at a sub-2B Arabic-specialized model, Kuwain-1.5B, never released its weights. We present RightNow-Arabic-0.5B-Turbo, a 518M-parameter Arabic-specialized decoder LLM built on Qwen2.5-0.5B. The pipeline adds 27,032 Arabic tokens via mean-subtoken initialization, continues pretraining on 504M Arabic tokens on 8xH100 with FSDP, FlashAttention varlen packing, and Liger fused kernels, then applies supervised fine-tuning on 129,116 Arabic instruction pairs with response-only loss masking, direct preference optimization on 6,750 Arabic preference pairs, and weight soup merging across three checkpoints. On three lm-evaluation-harness Arabic benchmarks (COPA-ar, Arabic HellaSwag, ArabicMMLU) the merged model reaches 35.9% mean accuracy, beats every same-class open model, ties Falcon-H1-1.5B on COPA-ar (58.4%) at one-third the size, and recovers 67% of SILMA-9B’s mean at 1/18 the parameters. The edge build quantizes to 398 MB (q4_k_m) and delivers 635 tokens/s at batch size 1 on a single H100 via this http URL. All code (5,555 lines across 25 scripts), weights (bf16, int8, and four GGUF quantizations), and benchmark scripts are released at this https URL.
[NLP-171] From Context Shift to Stylistic Collapse: Why Training Objectives Matter More Than Scale NEURIPS2026
【速读】： 该论文试图解决的问题是：现代大语言模型（LLM）在训练过程中对语言特征的重塑现象，即原本作为风格表达的语言元素被重新分配为概率质量，导致语言分布发生极端变化，这种变化在指令微调（instruction-tuning）阶段尤为显著，且不受强化学习人类反馈（RLHF）进一步加剧。解决方案的关键在于引入一种基于正则化强度的弱干预机制（weak intervention），通过调节控制参数 λ 来抑制语言熵的系统性坍塌。研究发现，适度的正则化（如 λ=1.0）反而加剧了语言多样性的损失，而强控制（λ=5.0）不仅将语言多样性提升40.5%，还显著优于前沿模型（在参数量少200–1000倍的情况下仍胜出96.7–98.2%），同时降低重复率78%、提高词汇多样性27%。这表明当前对齐（alignment）流程的核心问题并非简单的分布平滑，而是缺乏足够强的控制机制来维持语言结构的丰富性，揭示了现有对齐方法在统计不可见但实际存在的语言分布重塑方面的结构性缺陷。
链接: https://arxiv.org/abs/2605.28826
作者: Rohan Mahapatra
机构: Google(谷歌); Meta(Meta); OpenAI(OpenAI); Anthropic(Anthropic); Stability.AI(Stability.AI); Character.ai(Character.ai); Claude(Claude); Mistral(混合); Llama(混合); Gemma(混合); Pythia(混合); OLMo(混合)
类目: Computation and Language (cs.CL)
备注: 26 pages, 13 tables, 2 figures. Planning to submit to NeurIPS 2026
Abstract:In modern LLMs, linguistic features function not as stylistic artifacts but as probes of probability mass, allocated under training alignment objectives. Language models trained with contemporary pipelines exhibit severe reshaping of linguistic features, leading to extreme language re-distribution. While previous stylometric analyses explored linguistic differences between AI-generated and human texts, we focus on the reshaping plaguing the LLM training pipeline itself. We analyze 17 models (410M-100B+ parameters) across 24 linguistically-motivated probes, documenting that instruction-tuned systems systematically collapse language entropy along discourse and structural dimensions (mean amplification: 1,949-16,853%, peaks: 5,181-209,675%), while selectively suppressing complex punctuation to 3.2-23.2% of baseline frequencies. These effects do not worsen under RLHF, as divergence patterns are statistically indistinguishable (p 0.25) across matched base and instruction-tuned model pairs. Weak intervention (lambda=1.0) exacerbates collapse by 240%, while strong control (lambda=5.0) achieves 40.5% improvement and outperforms frontier models by 96.7-98.2% despite 200-1000x scale disadvantage. Additionally, lambda=5.0 delivers 15% higher distinct-4, 27% higher vocabulary diversity, and 78% lower repetition than moderate regularization, establishing that alignment requires sufficient control strength, not merely distributional smoothing. Our findings underscore how modern LLMs reallocate stylistic probability mass, despite RLHF and scale. More broadly, our work reveals a structural limitation of current alignment pipelines: preference optimization reshapes language distributions invisible to standard quality metrics yet detectable through distributional probes, with implications for AI detection, training data contamination, and long-term linguistic evolution.
[NLP-172] MechELK: A Mechanistic Interpretability Framework for Eliciting Latent Knowledge in Large Language Models
【速读】： 该论文试图解决大语言模型（Large Language Models, LLMs）中存在但未在表面输出中体现的隐含知识（latent knowledge）难以被有效提取的问题。现有方法如对比一致性搜索（Contrastive Consistency Search, CCS）依赖于对比激活模式，在复杂多步推理任务中表现不佳，而机制可解释性工具主要用于理解模型行为而非提取隐藏知识。论文提出了一种统一的三阶段框架 MechELK，其关键在于将机制可解释性与隐含知识提取相结合：首先通过稀疏自编码器（Sparse Autoencoder, SAE）特征分析和激活修补（activation patching）定位知识承载表示（Locate）；接着利用因果探针（causal probing）验证知识的真实性以区分虚假相关性（Verify）；最后通过表示工程（representation engineering）在不修改模型权重的前提下显式呈现隐藏知识（Elicit）。实验表明，MechELK 在 TruthfulQA、Deceptive Alignment 基准和 Quirky LM 数据集上平均提取准确率达 84.7%，显著优于 CCS 和直接线性探针，并能在模型输出错误或回避回答时成功识别出 78.3% 的隐含知识，展现出在 AI 安全（如欺骗对齐检测）中的重要应用价值。
链接: https://arxiv.org/abs/2605.28825
作者: Ji-jun Park,Soo-joon Choi,Jiwon Jeong,Taeyang Yoon,Ju-Wan Lee
机构: Dongguk University (东国大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large language models (LLMs) frequently encode factual and reasoning knowledge in their internal representations that is not faithfully reflected in their surface-level outputs – a phenomenon known as \emphlatent knowledge. Existing approaches to eliciting latent knowledge, such as Contrastive Consistency Search (CCS), rely on contrastive activation patterns and struggle with complex multi-step reasoning tasks, while mechanistic interpretability tools have primarily been used to \emphunderstand model behavior rather than to \emphextract hidden knowledge. We present \textbfMechELK, a unified three-stage framework that bridges mechanistic interpretability and latent knowledge elicitation. MechELK operates through: (1) \textbfLocate – using Sparse Autoencoder (SAE) feature analysis and activation patching to identify knowledge-bearing representations; (2) \textbfVerify – employing causal probing to distinguish genuine latent knowledge from spurious correlations; and (3) \textbfElicit – applying representation engineering to surface hidden knowledge without modifying model weights. Evaluated on TruthfulQA, a curated Deceptive Alignment benchmark, and the Quirky LM dataset, MechELK achieves an average elicitation accuracy of 84.7%, outperforming CCS by 6.2% and direct linear probing by 9.1%. Crucially, MechELK successfully identifies latent knowledge in 78.3% of cases where the model’s surface output is incorrect or evasive, demonstrating its utility for AI safety applications including deceptive alignment detection.
[NLP-173] A Modular Architecture for Typologically Controlled Lexicon Generation
【速读】： 该论文试图解决的问题是：如何在计算语言学中构建既可发音、符合音系类型学特征，又具备语义结构的人工语言词典（conlang）。现有生成方法要么缺乏严格的音位约束，要么依赖不可复现的大语言模型（LLM）流水线。其解决方案的关键在于提出一个模块化框架：首先从PHOIBLE数据库中采样音位库，接着在可互换的音系语法（确定性、优化理论OT和最大熵MaxEnt）下生成词形，并通过Swadesh–Leipzig–Jakarta语义本体实现形式与意义的显式对齐。实验表明，概率性音系语法在n-gram困惑度、对数似然和KL散度等指标上均优于确定性和随机基线，显著提升了音位一致性与类型学真实性。
链接: https://arxiv.org/abs/2605.28824
作者: Sankalp Tattwadarshi Swain,Dhruv Kumar
机构: Birla Institute of Technology and Science, Pilani (比尔拉理工学院和科学学院，皮拉尼)
类目: Computation and Language (cs.CL)
备注:
Abstract:Constructing artificial lexicons that are pronounceable, typologically plausible, and semantically structured remains an open challenge in computational linguistics. Existing conlang generators either lack formal phonotactic guarantees or delegate generation to opaque, non-reproducible LLM-based pipelines. We propose a modular framework that samples phoneme inventories from PHOIBLE, generates word forms under interchangeable phonological grammars (deterministic, OT, and MaxEnt), and assigns meanings via a Swadesh–Leipzig–Jakarta ontology with explicit form–meaning alignment. Evaluation on character n -gram perplexity, log-likelihood, and KL divergence against PHOIBLE across lexicon sizes of 100-5,000 forms shows that probabilistic grammars consistently outperform deterministic and random baselines on both phonotactic coherence and typological realism.
[NLP-174] What are They Thinking? Delineation Probing and Tracking of Concepts in LLM s
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）决策过程缺乏可解释性的问题，核心目标是开发一种低成本、通用性强的探测机制，用于识别LLM内部嵌入空间中特定概念的存在与否，从而揭示模型“思考”了哪些内容。解决方案的关键在于构建一套系统化流程：首先通过精心设计的数据集明确界定一个概念（包含该概念和不包含该概念的样本），然后训练并测试线性探测器（linear probes）以在任意层上检测该概念，并探索探测器复杂度与性能之间的关系；最后验证这些探测器能够在更长上下文中稳定追踪概念变化。实验基于四个概念和三种不同LLM进行验证，为未来大规模部署多概念监控提供了可行路径。
链接: https://arxiv.org/abs/2605.28823
作者: Mohamed Abdelwahab,Michelle Yu Collins,Sihan Chen,Yi Cheng Zhao,Zafarullah Mahmood,Jiading Zhu,Soliman Ali,Jonathan Rose
机构: University of Toronto (多伦多大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:As the influence of LLMs expands, it is imperative to gain insight into their decisions. One way to do that is to develop probes that detect the presence or absence of a broad set of concepts within the embeddings computed in an LLM - which is what we might say a model is “thinking” about. Such probes should be low-cost and easily applicable to any LLM, so that monitoring for many concepts is possible during normal operation. In this paper, we take the first steps towards developing the capability of creating many such probes by defining and executing examples of the key tasks needed: first, the careful delineation of a concept through the creation of a dataset with the concept both present and then absent. Then, the training and testing of a set of linear probes to detect the concept on any layer of an LLM, including an exploration of the complexity of the probe needed. Finally, we show that such probes can track concepts across larger contexts. This is done with four separate concepts and three different LLMs. When this process is scaled to many more concepts, it will create the ability to easily monitor new models. Subjects: Computation and Language (cs.CL) Cite as: arXiv:2605.28823 [cs.CL] (or arXiv:2605.28823v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2605.28823 Focus to learn more arXiv-issued DOI via DataCite
[NLP-175] Lightweight Multimodal LLM -Enabled Cost-Effective Defect Grading of Power Transmission Equipment
【速读】： 该论文旨在解决电力输电设备缺陷分级（Defect Grading of Power Transmission Equipment, DGPTE）中专家经验难以融入以及细粒度分级场景下类别不平衡问题。解决方案的关键在于提出一种基于多模态大语言模型（Multimodal Large Language Model, MLLM）的新型缺陷分级框架：首先利用上下文学习（in-context learning）充分挖掘商用MLLM在DGPTE任务中的潜力，获得当前最优（SOTA）性能；随后通过向该模型发起二次请求生成少量基于思维链（Chain of Thought）的问答对（Q\As），显著降低人工标注成本；最终使用这些高质量、可解释的Q\As，采用低秩适应（Low-Rank Adaption-based）监督微调（SFT）方法对Qwen3-VL-8B模型进行训练，实验表明仅微调语言模型层即可达到SOTA效果，且多任务联合微调验证了单一轻量级MLLM处理多种分级任务的可行性。
链接: https://arxiv.org/abs/2605.28822
作者: Tao Wang,Lipeng Zhu,Jiayong Li,Feng Gao,Siwen Liang
机构: 未知
类目: Computation and Language (cs.CL)
备注: 9pages, 6figures
Abstract:Defect grading of power transmission equipment (DGPTE) is crucial to the stability of electric energy transmission. Although existing machine learning methods exhibit strong capabilities in defect detection, they are plagued by difficulties in integrating expert experience and facing class imbalance in more refined defect grading field. To address this issue, this paper introduces a novel defect grading framework based on multimodal large language model (MLLM). Specifically, this approach maximizes the commercial MLLMs’ potential of DGPTE through in-context learning and obtains the state-of-te-art (SOTA) model. By sending a secondary request to this model, a small number of chain of thought-based question-answer pairs (Q\As) are generated, which effectively reduces the cost of manual annotation. In this way, these high-quality interpretable Q\As are used to train Qwen3-VL-8B via Low-Rank Adaption-based supervised fine-tuning (SFT). Experimental results on three DGPTE tasks demonstrate that fine-tuning only the language model layer yields the SOTA performance. Furthermore, multi-task joint fine-tuning verifies the feasibility of handling multiple grading tasks within only a single lightweight MLLM.
[NLP-176] MELD: Mel-Spectrogram-Based Speech Language Modeling with Discrete Latent Variables
【速读】： 该论文试图解决的问题是：当前语音语言模型中的编码器（encoder）通常与自回归模型（autoregressive model）分别优化，导致编码器提取的表征无法充分适配下游任务目标，从而影响性能。解决方案的关键在于提出一种基于梅尔频谱图（mel spectrogram）的离散潜在变量模型（discrete latent variable model），通过联合优化编码器和语音语言模型，使表征学习更加贴近下游任务需求。这种联合优化策略不仅在零样本文语转换（Text-to-Speech, TTS）和语音转文字（Speech-to-Text, STT）任务上优于基于码本（codec-based）和其他基于梅尔频谱图的基线方法，还能有效缓解自回归梅尔频谱建模中常见的问题，如生成冗长静音和漏词现象。
链接: https://arxiv.org/abs/2605.29859
作者: Sung-Lin Yeh,Wei Zhou,Gil Keren,Duc Le,Zhong Meng,Hao Tang,Jay Mahadeokar,Ozlem Kalinli,Alexandre Mourachko
机构: University of Edinburgh (爱丁堡大学)
类目: Audio and Speech Processing (eess.AS); Computation and Language (cs.CL)
备注:
Abstract:Recent speech language models rely on encoders that are optimized separately from autoregressive models. Since these encoders are unaware of the downstream objectives, the extracted representations may not be optimal for downstream tasks. To address this limitation, we introduce a discrete latent variable model on mel spectrograms that jointly optimizes the encoder and the speech language model. Joint optimization not only brings improvements over codec-based and other mel-spectrogram-based baselines on zero-shot Text-to-Speech (TTS) and Speech-to-Text (STT) tasks, but also effectively alleviates common issues in autoregressive mel-spectrogram modeling, such as prolonged silence generation and word omissions.
信息检索
[IR-0] GRASP: Plan-Guided Graph Retrieval with Adaptive Fusion and Reranking on Semi-Structured Knowledge Bases
链接: https://arxiv.org/abs/2605.30237
作者: Yicheng Tao,Yiqun Wang,Xiangchen Song,Xin Luo,Kai Liu,Jie Liu
类目: Information Retrieval (cs.IR); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:Semi-structured knowledge bases (SKBs) embed textual documents in a typed graph of entities and relations, and underpin applications such as product search, academic paper search, and precision-medicine inquiries. Existing hybrid retrieval systems on SKBs either use the graph only for query expansion, mix textual and structural branches under a global weighting, or rely on fine-tuned graph-traversal generators. We present GRASP, a three-stage SKB retrieval framework unifying plan-based graph retrieval, plan-conditioned fusion with a dense retriever, and a fine-tuned reranker over the fused candidates. GRASP substantially advances the state of the art on every metric across the three STaRK benchmarks, lifting average Hit@1 from 62.0 to 73.9. Ablation and sensitivity studies further confirm the effectiveness and robustness of GRASP.
[IR-1] LexPath: A domain-oriented multi-path framework for legal article retrieval
链接: https://arxiv.org/abs/2605.30205
作者: Weixuan Liu,Qingfeng Zhuge,Xuyang Chen
类目: Information Retrieval (cs.IR)
备注:
Abstract:Legal article retrieval is critical for building traceable and reliable legal AI systems, where conclusions must be grounded in specific legal articles. However, existing open-domain retrieval methods rely heavily on surface-level lexical or semantic similarity, making it difficult for them to distinguish legally relevant articles from those that are textually similar but legally inapplicable or misaligned with the user’s underlying intent. To bridge this gap, we propose \textscLexPath, a domain-oriented multi-path framework comprising a multi-path retrieval module and an intent-aware reranking module. The retrieval module combines two complementary legal-specific paths to collect candidate articles: an IRAC-guided sparse path that expands queries with legally informative keywords, and a structure-guided dense path trained with hard negatives derived from legal hierarchy and citation relations. Then, the reranking module further refines the candidate ranking by incorporating the intent consistency score between queries and legal articles. We evaluate \textscLexPath on two publicly available benchmarks focusing on general-public queries and a self-constructed benchmark targeting domain-professional scenarios. Experimental results demonstrate that \textscLexPath consistently outperforms lexical, dense, hybrid, and adaptive retrieval-augmented generation (RAG) baselines. Ablation studies further verify the effectiveness of each component.
[IR-2] No More K-means:Single-Stage Sparse Coding for Efficient Multi-Vector Retrieval ICML2026
链接: https://arxiv.org/abs/2605.30120
作者: Lixuan Guo,Yifei Wang,Tiansheng Wen,Aosong Feng,Stefanie Jegelka,Chenyu You
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted by ICML2026
Abstract:Multi-vector retrieval (MVR) models, exemplified by ColBERT, have established new benchmarks in retrieval accuracy by preserving fine-grained token-level interactions. However, this granularity imposes prohibitive storage and retrieval efficiency bottlenecks: to manage the immense memory footprint and computational overhead of billion-scale token vectors, state-of-the-art systems are forced to rely on aggressive dimension reduction and complex clustering (e.g., K-means). This compromise introduces two critical limitations: excessive indexing latency of clustering large-scale corpora and semantic information loss inherent to compression. In this paper, we propose Single-stage Sparse Retrieval (SSR, a paradigm shift that replaces expensive clustering with efficient sparse coding. Instead of compressing features into low-dimensional dense vectors, we utilize Sparse Autoencoder (SAE) to project token embeddings into a high-dimensional but highly sparse representation. This transformation enables us to bypass vector clustering entirely and leverage inverted indexing for precise, high-throughput retrieval. Extensive experiments on the BEIR benchmark demonstrate that SSR achieves a “trifecta” of improvements: it reduces indexing time by 15x compared to ColBERTv2, halves retrieval latency, and simultaneously improves retrieval performance over leading baselines.
[IR-3] DocRetriever: A Plug-and-Play Framework for Multimodal Document Retrieval with Comprehensive Benchmark KDD2026
链接: https://arxiv.org/abs/2605.30027
作者: Ruofan Hu,Menghui Zhu,Jieming Zhu,Bo Chen,Shengyang Xu,Minjie Hong,Xiaoda Yang,Sashuai Zhou,Li Tang,Tao Jin,Zhou Zhao
类目: Computer Vision and Pattern Recognition (cs.CV); Information Retrieval (cs.IR)
备注: Accepted at KDD 2026 Research Track
Abstract:Multimodal documents contain diverse elements, such as tables, figures, and layouts, which can complicate retrieval tasks. While current approaches typically combine dense visual embedding models with supervised rerankers to achieve high-precision retrieval, they face inherent limitations. First, the coarse-grained nature of dense embeddings tends to obfuscate explicit semantics, failing to leverage structurally salient information. Second, supervised reranking models suffer from generalization bottlenecks, as their performance heavily relies on domain-specific training data. Furthermore, existing benchmarks often lack diverse assessment dimensions and comprehensive relevance annotations, limiting reliable evaluation. To address these challenges, we propose DocRetriever, a plug-and-play framework. It enhances visual retrieval via a layout-aware sparse embedding technique, enabling effective hybrid encoding without the overhead of optical character recognition (OCR). We also introduce a generalizable reranker that leverages reasoning-augmented demonstrations and optimized sampling to improve accuracy in few-shot settings. Finally, we construct a new benchmark, MultiDocR, to enable more rigorous evaluation. Experiments across diverse benchmarks validate DocRetriever’s superiority over state-of-the-art methods.
[IR-4] Uncertainty Quantification for Multimodal Retrieval Augmented Generation
链接: https://arxiv.org/abs/2605.29956
作者: Simon Binz,Heydar Soudani,Faegheh Hasibi
类目: Information Retrieval (cs.IR)
备注:
Abstract:Retrieval Augmented Generation (RAG) improves the question answering capabilities of Large Language Models (LLMs) by incorporating external knowledge and has recently been extended to multimodal settings through Vision-Language Models (VLMs) that integrate visual and textual information. Despite these advances, generated answers can still be incorrect or misleading. Uncertainty Quantification (UQ) methods aim to estimate the reliability of model outputs, but most existing approaches are designed for text-only models and perform poorly in multimodal RAG scenarios. A key challenge is capturing uncertainty arising from multiple stages of the pipeline, including retrieval, visual understanding, and generation. In this work, we show that modeling uncertainty using multimodal and retrieval-aware probability signals improves estimation in multimodal RAG systems. We introduce LeMUQ, a Learnable Multimodal UQ method that analyzes token probabilities under input modifications, such as removing modalities or retrieved context. By encoding these signals as probability tokens and processing them with a finetuned model, our approach captures interactions between modalities and retrieval. Experiments across datasets, retrievers, and VLMs show consistent improvements over baseline and finetuned UQ methods. Our proposed LeMUQ increases the AUROC metric by 3.8% on average. Additionally, our method shows strong generalization performance across different retrieval setups and datasets with mixed results when transferring across different VLMs. Our findings highlight the importance of modeling multimodal uncertainty and provide a step toward more reliable and safer multimodal RAG systems. Code is available on GitHub.
[IR-5] Rec-Distill: An Industrial Distillation Pipeline for Large-Scale Recommendation Models
链接: https://arxiv.org/abs/2605.29755
作者: Haoran Ding,Wenlin Zhao,Yuchen Jiang,Juren Li,Jie Zhu,Xinchun Li,Yishujie Zhao,Yi Zhang,Ao Qiao,Jianhui Dong,Cheng Chen,Ziyan Gong,Deping Xie,Peng Xu,Zikai Wang,Yuwei Wang,Huizhi Yang,Zhe Chen,Yuchao Zheng
类目: Information Retrieval (cs.IR)
备注:
Abstract:Large recommendation models have demonstrated substantial potential gains under scaling laws, yet these gains are difficult to realize in industrial recommendation systems because real-world deployment requires lightweight models with strict serving efficiency and latency guarantees. This creates a fundamental gap between offline model scaling and online deployment. In this work, we present Rec-Distill, an industrial distillation pipeline that transfers the performance gains of large-scale recommendation modeling to efficient serving models. Rec-Distill combines large-teacher scaling with student-side transfer optimization through decoupled training, black-box distillation, debiasing mechanism, and a hybrid batch-streaming pipeline for dynamic recommendation environments. Across multiple recommendation and advertising scenarios on real-world platforms, our framework scales teacher models up to 24B dense parameters and 20K behavior sequence length, while enabling lightweight students to recover a substantial portion of teacher gains, with distillation transferability exceeding 60% in the best setting. Extensive offline and online experiments further show that these transferred gains consistently translate into measurable business improvements under industrial constraints. These results demonstrate that Rec-Distill provides a practical framework for distilling large-scale recommendation models into deployable, cost-efficient serving systems, while also establishing a reliable path toward scaling recommendation models to even larger regimes in the future.
[IR-6] From Prompts to Context: An Ontology-Driven Framework for Human-Generative AI Collaboration
链接: https://arxiv.org/abs/2605.29675
作者: Ngoc Luyen Le,Marie-Hélène Abel,Bertrand Laforge
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注:
Abstract:Collaborations with Generative AI often begin with a short prompt and end with an opaque output, leaving implicit who was involved, what task was being pursued, which resources were used, and which constraints should have shaped the process. This limited contextual explicitness hinders trust, traceability, and accountability, particularly when Generative AI is embedded in information-intensive workflows such as search, querying, and profile management. This paper introduces From Prompts to Context, an ontology-driven framework for representing Human-Generative AI collaboration. Its core component, the Contextual Collaboration AI Ontology (CCAI), models key elements of collaboration - including tasks, agent roles, resources, and constraints - as a shared machine-interpretable vocabulary. By combining populated CCAI instances with SPARQL-based context retrieval in operational workflows, the framework turns otherwise ephemeral prompt-response interactions into structured and queryable collaboration traces linking prompts, outputs, and their surrounding context. The approach is illustrated through a case study involving a software development team building a competency-based education feature for viewing and updating learner competency profiles. The case study shows how the framework can support the representation and documentation of collaboration episodes across requirements analysis, design, implementation, and testing. Within this setting, the results indicate that explicit collaboration modelling helps make task context more explicit, improves the traceability of AI-generated contributions, and supports more transparent and accountable Human-Generative AI practices. We conclude by outlining design principles for future Human-Generative AI systems that emphasise not only output quality, but also the explicit representation of the collaborative context in which outputs are produced.
[IR-7] Entity-Collision: A Stratified Protocol for Attributing Retrieval Lift in Agent Memory
链接: https://arxiv.org/abs/2605.29630
作者: Youwang Deng
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 48 pages with appendix; 6-page body, mandatory Limitations, References, and 7 appendices. Code, benchmarks, and 37 reproduce scripts: this https URL (see paper/REPRODUCIBILITY.md). Apache 2.0
Abstract:End-to-end agent-memory benchmarks report a single hit@k per retriever, confounding lexical leakage (uncontrolled query/gold/distractor entity overlap) with tag-mixing (preferences, services, tools averaged together). We propose entity-collision, a system-agnostic protocol that pins the BM25 floor by construction – every distractor shares the answer’s entity tokens – and stratifies queries by discriminator tag, so any lift over BM25 is attributable to the embedder. Applied to an open-source agent-memory testbed across 5 tags x 3 embedders x 5 collision degrees with paired-bootstrap 95% CIs, the protocol reveals a two-axis pattern: a 256-d hash trigram helps only on closed-vocabulary lexical tags at deep collision; MiniLM-384 dominates both axes; and a 2.7x-parameter BGE-large does not uniformly improve on MiniLM – it wins on intent-style queries but loses on lexical ones. Encoder capacity alone is not the binding constraint. The synthetic intent-tag null replicates on LongMemEval (n=500) as a single-session-preference recall cliff. Adaptive vector-weight routing on LoCoMo is a measured null: 11.7pp of oracle headroom exists, but no signal we tested recovers it. All 26 result tables and 37 reproduce scripts are version-controlled and verified by a public registry; the protocol is exercised on a deterministically governed memory testbed (event-sourced decision log, DAG-state-machine schema lifecycle) so every reported CI is reproducible byte-for-byte from the ingest stream.
[IR-8] HiKEY: Hierarchical Multimodal Retrieval for Open-Domain Document Question Answering ACL2026
链接: https://arxiv.org/abs/2605.29606
作者: Joongmin Shin,Gyuho Shim,Jeongbae Park,Jaehyung Seo,Heuiseok Lim
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: Accepted to ACL2026 Main
Abstract:Retrieval-augmented generation (RAG) for document-based Open-domain Question Answering (ODQA) on large-scale industrial corpora faces two critical bottlenecks: routing failure in locating the correct document and evidence fragmentation in integrating scattered information. Existing approaches relying on flat text chunks or page-level images inherently struggle to (i) precisely pinpoint the target document among thousands of candidates and (ii) organically connect multimodal evidence, such as tables and figures, within a limited token budget. To address these challenges, we propose HiKEY, a hierarchical tree-based multimodal retrieval framework that elevates document hierarchy to a first-class retrieval signal. Instead of simple chunking, HiKEY reconstructs a logical heterogeneous graph via Document Hierarchical Parsing (DHP), explicitly encoding parent-child relationships. Adopting a hierarchical coarse-to-fine strategy, the framework (1) performs global routing to rapidly prune the search space using hierarchical indexing, and (2) conducts fine-grained retrieval to rank sections by employing a multimodal fusion strategy that captures the most discriminative evidence. Finally, HiKEY assembles a token-efficient evidence subgraph via a hybrid structural-semantic packing strategy. Experiments on ODQA benchmarks demonstrate that HiKEY significantly outperforms page- and chunk-based baselines, improving retrieval recall by up to 12.9% and end-to-end QA performance by up to 6.8%.
[IR-9] SCOPE: A Lightweight-training LLM Framework for Air Traffic Control Readback Monitoring
链接: https://arxiv.org/abs/2605.29543
作者: Qihan Deng,Minghua Zhang,Yang Yang,Zhenyu Gao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC); Information Retrieval (cs.IR)
备注:
Abstract:Pilot readback of Air Traffic Control (ATC) voice instructions is a primary safeguard against miscommunication in air transportation. However, readback anomalies remain implicated in approximately 80% of aviation incidents. This vulnerability is further exacerbated by rising traffic volume and elevated cognitive workload, thereby motivating automated readback monitoring by machine. Traditional rule-based and machine learning approaches struggle to generalize across the highly variable and evolving phraseology of air traffic controller-pilot communications. While Large Language Models (LLMs) have opened a new avenue through their strong reasoning and generalization capabilities, existing approaches still face deployment and computational barriers in practice. In this work, we propose Semantic reasoning for Communication via Open-set Plug-in with Examples (SCOPE), a novel lightweight-training LLM framework that advances both the efficiency and accuracy of machine-based ATC readback monitoring. The core idea is to couple a plug-in open-set classifier with a carefully designed in-context learning mechanism on top of a frozen LLM. Extensive experiments on the semi-synthetic communication dataset show that SCOPE attains superior accuracy while delivering the low-latency response required for operational environments. Under a few-shot setting, SCOPE achieves 91.05% accuracy in open-set detection and corrects 96.63% of anomalous readbacks, thereby outperforming the strongest available baselines while providing explanations for its decisions. These findings demonstrate the potential of our framework as a practical pathway toward interpretable and controllable ATC readback monitoring.
[IR-10] FLASH-MAXSIM: IO-Aware Fused Kernels for Late-Interaction Scoring
链接: https://arxiv.org/abs/2605.29517
作者: Roi Pony,Adi Raz Goldfarb,Idan Friedman,Daniel Ezer,Udi Barzelay
类目: Information Retrieval (cs.IR)
备注:
Abstract:Late-interaction retrieval (ColBERT, ColPali) scores a query against a document with the MaxSim operator: for every query token, the maximum similarity over the document tokens, summed over query tokens. The standard implementation materializes the full query-token x document-token similarity tensor in GPU memory; for visual ColPali at 10K documents this tensor alone is 21 GB in FP16, created only to be reduced to one score per document and discarded. It exhausts a 40 GB GPU and bounds the achievable batch size in both inference and training. We present Flash-MaxSim, an IO-aware fused GPU kernel that computes exactly the same scores without ever materializing the tensor, by streaming query and document tiles through on-chip SRAM and folding the row-maximum reduction into the same pass. We extend the IO-aware principle through the training backward pass, an inverse-grid CSR construction that reuses the forward argmax for an atomic-free, destination-owned gradient reduction, and through INT8xINT8 quantization and variable-length (padding-free) scoring. Flash-MaxSim is up to 3.9x faster on an A100 (4.7x on an H100) than naive PyTorch at matched precision, uses up to 16x less inference memory and ~28x less training memory, unlocks corpus and batch sizes that exhaust PyTorch entirely, preserves the exact ranking (100% top-20 agreement with an FP32 reference) Subjects: Information Retrieval (cs.IR) Cite as: arXiv:2605.29517 [cs.IR] (or arXiv:2605.29517v1 [cs.IR] for this version) https://doi.org/10.48550/arXiv.2605.29517 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Roi Pony [view email] [v1] Thu, 28 May 2026 07:38:27 UTC (354 KB) Full-text links: Access Paper: View a PDF of the paper titled FLASH-MAXSIM: IO-Aware Fused Kernels for Late-Interaction Scoring, by Roi Pony and 4 other authorsView PDFHTML (experimental)TeX Source view license Additional Features Audio Summary Current browse context: cs.IR prev | next new | recent | 2026-05 Change to browse by: cs References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); About Help contact arXivClick here to contact arXiv Contact subscribe to arXiv mailingsClick here to subscribe Subscribe Copyright Privacy Policy Web Accessibility Assistance arXiv Operational Status
[IR-11] Xetrieval: Mechanistically Explaining Dense Retrieval
链接: https://arxiv.org/abs/2605.29507
作者: Zhixin Cai,Jun Bai,Yang Liu,Jiaqi Li,Yichi Zhang,Taichuan Li,Zhuofan Chen,Zixia Jia,Zilong Zheng,Wenge Rong
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: Code: this https URL ; Project page: this https URL
Abstract:Explaining why dense retrievers assign high relevance scores remains challenging because retrieval decisions are made through opaque high-dimensional embeddings. Existing explanations often focus on surface signals, such as lexical matches, token alignments, or post-hoc textual rationales, and thus provide limited insight into the latent factors that shape dense retrieval behavior at the embedding level. We propose \textitXetrieval, an embedding-level mechanistic framework for explaining dense retrieval. \textitXetrieval first introduces a lightweight reasoning internalizer that approximates Chain-of-Thought reasoning directly in the embedding space with a single forward pass, enriching sentence embeddings with reasoning-oriented information while avoiding expensive autoregressive generation. It then decomposes these reasoning-enhanced embeddings into sparse, human-interpretable features, each associated with a coherent natural language description. By aggregating sparse feature overlaps across multiple document-side views, \textitXetrieval provides feature-level explanations of individual retrieval decisions. Experiments on diverse retrievers and benchmarks show that \textitXetrieval uncovers coherent interpretable features, yields stronger pair-level intervention effects, and supports task-level feature steering. The project page and source code are available at this https URL .
[IR-12] SkillBrew: Multi-Objective Curation of Skill Banks for LLM Agents
链接: https://arxiv.org/abs/2605.29440
作者: Wentao Hu,Zhendong Chu,Yiming Zhang,Junda Wu,Ming Jin,Xiangyu Zhao,Yilei Shao,Yanfeng Wang,Qingsong Wen
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 16 pages. Preprint. Under review
Abstract:Retrieval-augmented LLM agents increasingly rely on curated skill banks: collections of reusable textual principles that guide decision making on complex tasks. Existing approaches typically expand these banks in an append-only fashion, continuously adding new skills without removing redundant, outdated, or harmful ones, resulting in inefficient and poorly curated repositories. In this paper, we formulate the skill bank curation as a constrained multi-objective problem: a desirable bank must be useful for the agent, diverse in its content, and provide good coverage of the query distribution. To this end, we introduce SkillBrew, a multi-objective curation framework that formalizes skill bank curation as Pareto-aware optimization under a utility constraint, and solves it via a bi-level propose-then-verify loop. We evaluate our approach on two public benchmarks. Our findings suggest that treating skill banks as objects of principled curation, rather than ever-growing append-only logs, is an important step toward building self-improving LLM agents.
[IR-13] Latent Terms: Dense Retrievers Contain Trivially Extractable BM25-ready Zipfian Vocabularies
链接: https://arxiv.org/abs/2605.29384
作者: Benjamin Clavié,Sean Lee,Aamir Shakir,Makoto P. Kato
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:We propose Latent Terms, a method revealing that models trained for dense retrieval, whether single- or multi-vector, learn representations that can trivially be decomposed into retrieval-ready sparse features. When trained on frozen retrievers, Sparse Autoencoders without any retrieval-specific adjustments extract a latent vocabulary with approximately Zipfian collection statistics, directly suitable for classical sparse retrieval scoring via BM25. This approach enables sparse retrieval while requiring no learned expansion objective or sparse retrieval supervision whatsoever, and can be readily applied to any dense retriever. Latent Terms is able to match or outperform single-vector scoring methods from its own base model as well as comparable SPLADE variants. In addition, it substantially outperforms its base model on LIMIT, a task specifically designed to highlight the failures of single-vector retrieval. Overall, our results highlight that neural retrievers contain more expressive and indexable structure than their default scoring functions expose, but that other methods can nonetheless be leveraged.
[IR-14] ACE: Anisotropy-Controllable Embedding for LLM -enhanced Sequential Recommendation SIGIR2026
链接: https://arxiv.org/abs/2605.29322
作者: Dongcheol Lee,Hye-young Kim,Jongwuk Lee
类目: Information Retrieval (cs.IR)
备注: Accepted by SIGIR 2026. 5 pages
Abstract:Recent advances in the LLM-as-Extractor paradigm leverage large language models (LLMs) to transfer semantically rich item embeddings into sequential recommendation (SR) backbones. However, LLM-generated embeddings often suffer from strong anisotropy. Most vectors are concentrated in similar directions, resulting in a geometric imbalance that makes it difficult to adapt to collaborative signals during fine-tuning. To address this challenge, we propose Anisotropy-Controllable Embedding (ACE), which explicitly controls the anisotropy of LLM-generated embeddings. Specifically, ACE utilizes a linear autoencoder (LAE) to reshape the embedding distribution while preserving its semantic structure. In this process, the L2-regularization term mitigates the anisotropy by controlling the dispersion of embedding dimensions, while the reconstruction loss maintains semantic relationships among items. That is, ACE balances geometric uniformity and semantic embedding preservation for more stable learning. Extensive experiments demonstrate that ACE consistently outperforms existing LLM-enhanced SR models, yielding improvements of up to 12.4% and 11.8% in Recall@20 and NDCG@20, respectively.
[IR-15] GrepSeek: Training Search Agents for Direct Corpus Interaction
链接: https://arxiv.org/abs/2605.29307
作者: Alireza Salemi,Chang Zeng,Atharva Nijasure,Jui-Hui Chung,Razieh Rahimi,Fernando Diaz,Hamed Zamani
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注:
Abstract:Large Language Model (LLM) search agents have shown strong promise for knowledge-intensive language tasks through multiple rounds of reasoning and information retrieval. Most existing systems access information using a retriever that takes a keyword or natural language query and returns a ranked list of documents using an index of pre-computed document representations. In this work, we explore a complementary perspective in which the search agent treats the corpus itself as the search environment and finds evidence by issuing executable shell commands. We introduce GrepSeek, an optimized direct corpus interaction (DCI) search agent that trains a compact search agent to find, filter, and compose evidence from large text corpora. To address the instability of learning behavior directly with reinforcement learning on large corpora, we propose a two-stage training pipeline. First, we construct a cold-start dataset using an answer-aware Tutor and answer-blind Planner to generate verified, causally grounded search trajectories. Second, we refine the initialized policy with Group Relative Policy Optimization (GRPO), allowing the agent to improve its task-oriented search behavior through direct interaction with the corpus. To make DCI practical at scale, we further use a semantics-preserving sharded-parallel execution engine that accelerates shell-based retrieval by up to 7.6\times while preserving byte-exact equivalence with sequential execution of the shell command. Experiments across seven open-domain question answering benchmarks show that GrepSeek achieves the strongest overall token-level F_1 and Exact Match. Our analysis also highlights the limitations of purely lexical interaction on queries with substantial surface-form variation, suggesting DCI as a practical and competitive method for search agents that can complement existing retrieval paradigms in the real world.
[IR-16] UniNote: A Unified Embedding Model for Multimodal Representation and Ranking KDD
链接: https://arxiv.org/abs/2605.29287
作者: Jinghan Zhao,Wenwei Jin,Anqi Li,Jintao Tong,Luya Mo,Jiawei Li,Bin Li,Yao Hu
类目: Information Retrieval (cs.IR); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by KDD Ads Track 2026
Abstract:Item-to-Item (I2I) retrieval is a fundamental part of modern content platforms, supporting critical industrial workflows from recommendation engines to content auditing. While multimodal embedding methods have advanced general retrieval, they often falter in I2I scenarios due to the challenges of balancing global content representation with fine-grained local retrieval, the systemic inefficiency of decoupled embedding-and-ranking pipelines, and the inherent trade-offs between model precision and serving latency. To solve these issues, we propose \textbfUniNote, a unified embedding model designed for industrial I2I retrieval. Tailored retrieval strategies are introduced to support representation learning over complex, multimodal content at varying granularities. To operationalize these strategies, UniNote employs a two-stage training paradigm: the first stage leverages contrastive SFT to establish robust base embeddings, while the second stage refines ranking quality through a reinforcement learning (RL) process that aligns the model with content relevance. Our results show that UniNote achieves SOTA performance across diverse I2I tasks. Deployed at Xiaohongshu and integrated with Matryoshka Representation Learning (MRL), UniNote achieved significant improvements in retrieval quality and cost efficiency in large-scale applications.
[IR-17] CrossAlpha: An Annual-Report Benchmark for Cross-Market Factor Research
链接: https://arxiv.org/abs/2605.29286
作者: Qian Wang,Zhongyi Tong,Nuo Chen,Zhaomin Wu,Bingsheng He
类目: Information Retrieval (cs.IR)
备注:
Abstract:Cross-market factor research studies whether firm-level signals from one or more markets can predict returns in a target market, but existing public benchmarks do not support cross-market disclosure-to-return evaluation. Building such a benchmark is challenging because filings differ across languages and regulatory systems, disclosure-derived similarity can be biased by common reporting components, and cross-market signals must be evaluated under feasible trading-time alignment. We introduce \textbfCrossAlpha, a public annual-report benchmark for cross-market factor research. CrossAlpha addresses these challenges through three corresponding components: \emphDisclosure Distillation, which standardises heterogeneous filings into ten-category English business descriptions; \emphResidual Schema Graph Construction, which builds PCA-whitened cross-market firm-pair scores from schema-level disclosures; and \emphTiming-Aligned Evaluation, which pairs the graph with 11 years of daily OHLCV data to construct forward-return labels under feasible cross-market execution protocols. CrossAlpha covers about 3,600 firms and 10,700 firm-year reports from the United States, Japan, Taiwan, South Korea, and Hong Kong, and releases about 19M directed firm-pair scores. In experiments, disclosure-derived cross-market peers outperform domestic text, industry-code, and return-correlation peers in the US-to-Japan setting (ICIR 0.39 versus 0.07–0.18), and cross-market sources beat the domestic text baseline in most target markets. CrossAlpha offers an open-sourced, reusable, return-grounded benchmark for cross-market financial NLP.
[IR-18] LoopFM: Learning frOm HistOrical RePresentations of Foundation Model for Recommendation
链接: https://arxiv.org/abs/2605.29280
作者: Shali Jiang,Hua Zheng,Boyang Liu,Laming Chen,Kenny Lov,Chuanqi Xu,Lisang Ding,Qinghai Zhou,Can Cui,Xiaolong Liu,Xiaoyi Liu,Yasmine Badr,Xin Xu,Jiyan Yang,Ellie Dingqiao Wen,Gerard Jonathan Mugisha Akkerhuis,Chenxiao Guan,Rong Jin,Ruichao Qiu,Xian Chen,Shifu Xu,Zhehui Zhou,Ping Chen,Rui Yang,Haicheng Chen,Xiangge Meng,Song Zhou,Dharak Kharod,Shuyu Xu,Qiang Jin,Qiao Yang,Wankun Zhu,Qin Huang,Yuzhen Huang,Darren Liu,Parish Aggarwal,Hui Zhou,Erzhuo Wang,Shuo Chang,Xiaorui Gan,Wenlin Chen,Santanu Kolay,Huayu Li
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: Shali Jiang, Hua Zheng, Boyang Liu contributed equally to this work
Abstract:Knowledge distillation (KD) transfers a single scalar prediction from a large foundation model (FM) to compact vertical models (VMs), suffering from diminishing transfer ratio – the fraction of FM improvement captured by the VM – as a single scalar cannot convey the rich intermediate knowledge that larger FMs learn. To address this bottleneck, we propose LoopFM (Learning frOm HistOrical ReP*resentations of FM), a framework that opens a high-bandwidth transfer channel by structuring FM intermediate embeddings as input features (e.g., user history sequence) for downstream VMs, without requiring real-time FM inference at serving and architectural coupling between FM and VM. We provide a theoretical framework for LoopFM with a gain decomposition and transfer-ratio analysis. On three public benchmarks, LoopFM demonstrates strong AUC improvements (e.g., 6%+ on TaobaoAd) and complementary knowledge transfer capability with KD. On industrial-scale systems (billions of examples, trillion-parameter FMs), LoopFM approximately doubles the knowledge transfer ratio on top of KD, delivering a +0.5% conversion improvement in Y1H1, and a +1.03% and +1.22% conversion improvement from two individual launches respectively in Y1H2.
[IR-19] CoHyDE: Iterative Co-Training of LLM Rewriter Dense Encoder for Tool Retrieval
链接: https://arxiv.org/abs/2605.29271
作者: Vaishali Senthil,Ashutosh Hathidara,Sebastian Schreiber
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注:
Abstract:Tool retrieval over large API catalogs is a core bottleneck for LLM agents: user queries arrive in colloquial, often underspecified language, while the catalog uses technical API vocabulary that no fixed encoder can bridge on its own. The two dominant training approaches, contrastive encoder fine-tuning and HyDE-style query expansion with a frozen LLM, address this problem from opposite ends and fail in complementary directions: the fine-tuned encoder excels when the query’s surface form already matches the catalog but collapses when it does not, while zero-shot HyDE is more robust to underspecified queries yet generates catalog-unaware hypothetical descriptions that degrade retrieval when queries are well-formed. We introduce CoHyDE, an iterative procedure that trains the dense encoder and the LLM rewriter as a single co-evolving system: the encoder is retrained with InfoNCE on catalog-style hypothetical descriptions produced by the rewriter, and the rewriter is preference-aligned via DPO against the encoder’s retrieval scores, with both sides warm-started on the tool catalog before the loop begins. On a ~10k tool subset of the ToolBench catalog, three rounds of CoHyDE improve over the strongest single-component baseline by +2.5 pp NDCG@5 on standard queries and +6.3 pp on held-out vague queries, with gains as large as +8 pp on the hardest vague tier. Ablations confirm that co-training is the key ingredient: using either component in isolation fails to match CoHyDE on both well-formed and vague queries, with losses of up to -8 pp on vague queries.
[IR-20] OmniRetrieval: Unified Retrieval across Heterogeneous Knowledge Sources
链接: https://arxiv.org/abs/2605.29250
作者: Jinheon Baek,Soyeong Jeong,Sangwoo Park,Woongyeong Yeo,Minki Kang,Patara Trirat,Heejun Lee,Sung Ju Hwang
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注:
Abstract:Real-world information needs require access to structurally diverse knowledge sources, from unstructured text and relational tables to knowledge graphs and property graphs. Existing retrievers, however, operate over one source at a time under a fixed query language, leaving the broader landscape of available knowledge fragmented behind incompatible interfaces. A natural attempt at unification would collapse these sources into a shared space, but this erases the structural affordances (such as schemas, ontologies, compositional operators) that give each source its expressive power. Effective retrieval over diverse knowledge, therefore, requires not homogenization but an overarching layer that meets each source on its own terms. To achieve this, we present OmniRetrieval, a framework that takes any natural-language query, identifies appropriate knowledge sources, and dispatches source-native queries to their native execution engines. Across an extensive benchmark spanning 13 datasets and 309 distinct knowledge bases over text, relational, and graph-structured sources, OmniRetrieval exceeds single-source baselines, demonstrating that it can serve as a general-purpose interface to the heterogeneous sources while preserving the structural distinctions that make each source valuable.
[IR-21] Surfacing Isolated Learners with Outcome-Independent Mediation of Feedback between Teachers and Students Using AI
链接: https://arxiv.org/abs/2605.29240
作者: Junsoo Park,Youssef Medhat,Htet Phyo Wai,Ploy Thajchayapong,Ashok K. Goel
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC); Information Retrieval (cs.IR)
备注: Accepted to HAI-Agency Workshop on Orchestrating Human and AI Agency for Proactive and Reflective Learning
Abstract:AI-augmented classrooms generate rich teacher and student feedback before graded outcomes become available, yet these signals can be difficult to translate into timely instructional decisions. We propose an interpretable decision layer: a transparent mechanism that ranks course topics requiring attention without using grades or post-hoc outcome labels. The approach combines three signals: student learning difficulty prevalence, disagreement between learner self-reports and observed difficulties, and unresolved teacher concerns. The output is a ranked set of topic priorities with per-topic decision records explaining each ranking. In one graduate CS course offering ( n=5 instructor interviews; n=279 survey responses), prioritized topics aligned with instructor concerns (top-5 overlap 3/5; Spearman \rho=0.80 ) and student-reported topic difficulty ( \rho=0.46 , p=.048 ). Multi-signal integration also surfaced learners not identified through individual signal sources alone (AUC =0.96 vs. 0.91 for gap prevalence alone). Reflective thinking, help-seeking, and self-efficacy provided additional evidence that student behavioral signals align with learning-related constructs. While preliminary, these findings suggest that transparent coordination mechanisms may help support human-AI co-agency when feedback is incomplete.
[IR-22] Rethinking Literature Search Evaluation: Deep Research Helps and Human Citation Lists Are Not a Ground Truth
链接: https://arxiv.org/abs/2605.29234
作者: Gaurav Sahu,Laurent Charlin,Christopher Pal
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注:
Abstract:We study large-scale literature search from two complementary angles: improving the retrieval pipeline, and stress-testing the human reference list as an evaluation target. First, we implement a Deep Research pipeline that processes the full query paper and expands the retrieved results breadth-first along their bibliographies, and show that it substantially outperforms vanilla API-only search, raising recall on RollingEval-Jun25 (a 250-paper literature-search benchmark) from below 20% to above 80%. Second, we use a neutral LLM-as-a-judge to determine if human references are sound ground truth for the task. We find significant limitations: only 51% of human citations are judged moderately relevant or higher, against 86–88% for the strongest AI-based re-rankers. We study this gap on the OpenAlex co-authorship graph, finding that humans are 2.5x more likely than the best AI re-rankers to cite a direct collaborator. Together, our results argue against single-axis literature-search evaluation: recall, topical-relevance scoring, ranked-list diversity, and a co-authorship-distance diagnostic each measure complementary properties of citation quality and should be reported jointly.
[IR-23] On the Practice of Scaling Search Conversion Rate Prediction
链接: https://arxiv.org/abs/2605.29232
作者: James Pak,Jyun-Yu Jiang,Fan Zhang,Sen Wang,Taekmin Kim,Henry Tsai,Vijay Rajaram,Juexin Lin,Mohitdeep Singh,Alessandro Magnani,Johnny Chen,Qian Zhao,Rao Fu,Zhirong Liang,Jordan Gilliland,Winter Jiao
类目: Information Retrieval (cs.IR)
备注:
Abstract:Scaling a Search Conversion Rate (CVR) prediction model, especially in high-traffic environments, presents a challenge: superior model quality needs to be balanced with strict constraints on training cost and serving latency. This paper details an effective approach for scaling modern search CVR prediction models. We begin with an empirical study to understand the scaling performance of search CVR models, analyzing how quality improves as we scale three key factors of model backbone computation, the size of embedding parameters, and the volume of training data. We use a large-scale production dataset, comprising over a year of customer interaction logs from a high-traffic e-commerce platform, to evaluate the scalability of several state-of-the-art architectures and their ensembles. Our key findings are: (1) selecting the right backbone and scaling factors is crucial; (2) the impact of scaling backbone, embedding, and data is largely independent and additive, which has implications for more efficient scaling exploration; (3) a streamlined warmstart strategy can accelerate training iterations while simplifying new updates; (4) inference optimization strategies such as decoupled graph execution and dynamic batching can enable low-latency GPU serving even for high-capacity models. Compared to a baseline of a pre-scaling production model, we ultimately deployed a model trained on 2.5x larger training data with 8x more inference compute while having minimal latency impact. Online A/B tests also demonstrate that our launches achieved a combined +2.6% gain in a key metric of search conversion rate. Subjects: Information Retrieval (cs.IR) Cite as: arXiv:2605.29232 [cs.IR] (or arXiv:2605.29232v1 [cs.IR] for this version) https://doi.org/10.48550/arXiv.2605.29232 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[IR-24] PROTOCOL: Late Interaction Retrieval for Protein Homolog Search
链接: https://arxiv.org/abs/2605.29158
作者: Gabrielle Cohn,Rohan Gumaste,Minh Hoang,Vihan Lakshman
类目: Machine Learning (cs.LG); Information Retrieval (cs.IR); Biomolecules (q-bio.BM)
备注:
Abstract:Protein homology search underlies function annotation, structure prediction, and evolutionary analysis, but remains challenging in the “twilight zone,” where global sequence similarity is weak and classical alignment methods lose sensitivity. Protein language models provide context-aware representations that could improve alignment sensitivity in this regime. However, prior protein embedding-based retrieval pipelines often pool these representations into a single vector, potentially obscuring local motifs, domains, or conserved residues that reveal remote homology. We introduce ProtoCol, a model which represents proteins as sets of residue embeddings and uses ColBERT-style late interaction to test whether residue-level comparison improves homolog retrieval. ProtoCol encodes proteins independently, keeps candidate representations pre-computable, and scores candidates with MaxSim over residue embeddings. On SCOPe superfamily and Pfam clan benchmarks, ProtoCol outperforms sequence-composition, alignment-based, pooled PLM, and trained single-vector baselines, supporting late interaction as an effective retrieval layer for remote homology search.
[IR-25] oward User Preference Alignment in LLM Recommendation via Explicit Context Feedback
链接: https://arxiv.org/abs/2605.29141
作者: Weizhi Zhang,Wooseong Yang,Yuxin Cui,Zhaohui Guo,Hins Hu,Liangwei Yang,Henry Peng Zou,Qifei Wang,Hanqing Zeng,Jiayi Liu,Yinglong Xia,Philip S. Yu
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
备注: Published in CogMI 2025. this https URL
Abstract:Traditional recommender systems (RecSys) primarily infer user preferences from implicit signals (such as clicks, watches, and purchases), often neglecting the rich explicit contextual feedback users provide through verbal text, like comments and reviews. This explicit context feedback captures the nuanced reasons behind user decisions regarding their preferences. In addition, it offers critical heterogeneous information for user preference alignment and more explainable recommendations. Overlooking such signals can lead to misaligned user preferences and further reinforce filter bubbles, as algorithms fail to understand the “semantic context” behind user choices. Recent advances in Large Language Models (LLMs) present new opportunities to harness user-generated content for more accurate and diverse recommendations, yet current LLM-based recommendations still focus on using item meta-data and underutilize this resource. In this paper, we advocate for prioritizing explicit context feedback in the next generation of LLM-based RecSys. We review the evolution of recommendation paradigms, highlight the value of context-rich feedback, call for new benchmarks and metrics, and introduce frameworks for integrating explicit user signals into scalable LLM-driven RecSys. Centering on user-preference modeling, we aim to foster more personalized, transparent, and explainable RecSys online platforms.
[IR-26] Same Question Different Source Different Answer: Auditing Source-Dependence in Medical Multi-Source RAG
链接: https://arxiv.org/abs/2605.29084
作者: Yubo Li,Rema Padman,Ramayya Krishnan
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注:
Abstract:A retrieval-augmented generation (RAG) system deployed over a multi-author institutional corpus can give a different answer to the same question depending on which source it retrieves – a failure mode the dominant single-gold-answer paradigm cannot diagnose. We argue that source-dependence is a missing axis of NLP evaluation, and that auditing it means shifting the unit of evaluation from answer correctness to the inter-source relationship. We make this concrete in transplant patient education, where institutional sources demonstrably disagree, releasing three artefacts: TransplantQA, a benchmark of real patient questions, each answered by grounding generation in multiple institutional handbooks as candidate sources; HERO-QA, a hierarchical retrieval strategy that grounds and audits each answer; and a structured-output judge that scores inter-source relationships on a validated 5-label taxonomy. At scale, better retrieval reveals far more disagreement than prior estimates suggested – understating its prevalence, not its intensity. The framework is domain-agnostic and transfers to legal and educational RAG: measuring source-dependence is a responsibility for deployed multi-source NLP generally.
[IR-27] When LLM Reward Design Fails: Diagnostic-Driven Refinement for Sparse Structured RL
链接: https://arxiv.org/abs/2605.28918
作者: Youting Wang,Yuan Tang,Bowen Liu,Xuan Liu,Dingyan Shang
类目: Machine Learning (cs.LG); Information Retrieval (cs.IR)
备注:
Abstract:For sparse, structured reinforcement-learning tasks with semantic reward-function interfaces, LLM-generated reward shaping is better framed as debugging than one-shot generation. We study PPO-trained agents using MiniGrid as core evaluation and MuJoCo as boundary stress test. Our audit finds two dominant one-shot failure modes – reward flooding and semantic/API misunderstanding – plus a rarer weak-shaping case. We propose diagnostic-driven iterative refinement, where training diagnostics and a failure-mode taxonomy guide targeted reward-function revision. Refinement improves DoorKey-8x8 from 2.3% to 97.6% and KeyCorridor from 31.2% to 86.7% with high seed-to-seed variance. Controls show these gains are not from retrying or extra training: metrics-only re-prompting yields large drops, while a static-vocabulary control recovers much of the gap (87.6%; 70.7%), showing the taxonomy prompt is a major mechanism and dynamic labels provide only partially isolated incremental evidence. Budget-matched and Best-of-3 comparisons separate refinement from selection and training-time effects. Component-removal tests, sensitivity analyses, and an audit against author labels provide converging evidence for the debugging interpretation while revealing calibration limits. Continuous-control results show the boundary: success-based diagnostics can misfire in dense-reward locomotion, and return-trend feedback removes one false-positive mechanism without robust gains. The low-call protocol is a cost contrast with population-based reward search, not a benchmark comparison. In four crossed-variance-design environments, point estimates suggest larger gains when LLM reward-function variance dominates but bootstrap intervals are wide. The method is bounded to sparse structured tasks with reliable interfaces under PPO; fields like event_text may help, hurt, or be neutral.
[IR-28] Generative Spatiotemporal Intent Sequence Recommendation via Implicit Reasoning in Amap
链接: https://arxiv.org/abs/2605.28888
作者: Sicong Wang,Ruiting Dong,Yue Liu,Bowen Zheng,Jun Meng,Jie Li,Shuaijun Guo,Yu Gu,Fanyi Di,Xin Li
类目: Information Retrieval (cs.IR); Machine Learning (cs.LG)
备注: 9 pages, 1 figure
Abstract:Real-world user behavior rarely consists of isolated actions; instead, it often forms intent flows governed by spatiotemporal dependencies. To provide integrated service recommendations, we focus on the task of Generative Spatiotemporal Intent Sequence Recommendation (GSISR), which aims to generate intent sequences that are logically coherent and physically executable within complex spatiotemporal contexts. While LLMs offer strong reasoning potential for GSISR, direct industrial deployment is limited by high inference latency and context-mismatched or physically infeasible plans. To address these challenges, we propose a generative framework, GPlan, that internalizes LLM reasoning into lightweight models through two components. First, to enable reasoning under strict latency constraints, we introduce Progressive Implicit CoT Distillation, which compresses explicit reasoning processes into reserved latent tokens, allowing small models to inherit complex planning logic without generating long reasoning text. Second, to address the disconnect between general knowledge and real-world constraints, we design Spatiotemporal Counterfactual DPO. By aligning the model with counterfactual context-plan pairs, we improve sensitivity to spatiotemporal context and reduce context-mismatched plans. Offline experiments and online A/B testing demonstrate that our approach improves sequence coherence and context responsiveness. Our implementation and the anonymized GSISR dataset are available at this https URL.
人机交互
[HC-0] Physics Is All You Need? A Case Study in Physicist-Supervised AI Development of Scientific Software ICML2026
链接: https://arxiv.org/abs/2605.30353
作者: Nhat-Minh Nguyen
类目: Artificial Intelligence (cs.AI); Cosmology and Nongalactic Astrophysics (astro-ph.CO); Human-Computer Interaction (cs.HC); Software Engineering (cs.SE)
备注: 10 pages, 2 figures, 2 tables, 1 physicist and a few AI agents. Accepted by ICML 2026 AI for Science Workshop. Code and development log are available at this repo: this https URL
Abstract:Are AI agents tools, co-authors, or researchers? We present a quantified case study ( N=1 ): a physicist supervising an AI coding agent (Claude Code, Sonnet and Opus models) over 12 work days and 57 sessions to build CLAX-PT, a differentiable one-loop perturbation theory module in JAX. We documented and classified 15 supervision events by intervention level. The agent resolved ten autonomously by iterating against oracle tests. Two more by the physicist’s domain knowledge. The three it could not – all evaded oracle detection – share a common property: the agent treated symptom reduction as root-cause resolution. It spent 33 of the 57 sessions adjusting coefficients within a code architecture that could not represent the target physics, and could not re-evaluate its CLASS-PT branch choice even when prompted to reconsider; only an injected physics concept (anisotropic BAO damping) triggered the redesign. Separately, the agent committed a calibrated correction that passed all oracle tests but corresponded to no quantity in the theory, predicting wrong values at any other cosmology. The fudge factor was caught and replaced within the same session. Three supervision practices proved critical for catching what oracle tests missed: testing at diverse parameter points beyond the fiducial calibration; shared changelogs that surfaced stalled exploration across sessions; and an explicit rule against unphysical numerical patches. In this case, supervision design, not model capability, determined whether the agent’s output was trustworthy. Closing the gap would require agents that propose architectural alternatives rather than optimize within a given structure, and distinguish predictive adequacy from explanatory correctness – capabilities not exhibited here, not obviously addressed by scaling alone. [Abridged.] Comments: 10 pages, 2 figures, 2 tables, 1 physicist and a few AI agents. Accepted by ICML 2026 AI for Science Workshop. Code and development log are available at this repo: this https URL Subjects: Artificial Intelligence (cs.AI); Cosmology and Nongalactic Astrophysics (astro-ph.CO); Human-Computer Interaction (cs.HC); Software Engineering (cs.SE) Reportnumber: IPMU26-0021 Cite as: arXiv:2605.30353 [cs.AI] (or arXiv:2605.30353v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2605.30353 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[HC-1] LLUMI: Improving LLM Writing Assistance for Mental Health Support with Online Community Feedback
链接: https://arxiv.org/abs/2605.30273
作者: Jiwon Kim,Maya Ajit,Sherry Gong,Soorya Ram Shimgekar,Dong Whi Yoo,Eshwar Chandrasekharan,Koustuv Saha
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computers and Society (cs.CY); Social and Information Networks (cs.SI)
备注:
Abstract:Large language models (LLMs) show promise in generating supportive responses for mental health queries, but improving their usefulness, empathy, and safety often requires substantial compute, expert input, and labeled data. At the same time, deploying proprietary, cloud-based models for mental health-related interactions raises important privacy and data-governance concerns, given the sensitivities. To address this challenge, we introduce LLUMI setup that can be hosted in-house within protected environments. LLUMI consists of two complementary components: a generation model (GM), which drafts supportive responses to mental health queries, and an improvement model (IM), which revises an initial human-crafted response. We leverage feedback signals from Reddit mental health communities, using community endorsement patterns such as upvotes and downvotes to construct chosen-rejected response pairs for Supervised Fine Tuning (SFT) and Direct Preference Optimization (DPO). We further align LLUMI using human evaluation across five dimensions: readability, empathy, connection, actionability, and safety. Our results show that, despite relying on smaller open-source models rather than proprietary cloud-based GPT models, LLUMI achieves comparable performance across linguistic analyses and human evaluations. These findings suggest that open-source models, when trained with community-derived preference signals, can support high-quality mental health support assistance while offering a more privacy-preserving alternative for sensitive support contexts.
[HC-2] VideoFDB: Evaluating Full-Duplex Vision-Speech Capabilities in Conversational Agents
链接: https://arxiv.org/abs/2605.30256
作者: Amrita Mazumdar,Seonwook Park,Rajarshi Roy,Nikhil Srihari,Shengze Wang,Yuhao Zhou,Julia Wang,Koki Nagano,Shalini De Mello
类目: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注: Project page: this https URL
Abstract:Natural human conversation is full-duplex and audio-visual: people simultaneously speak and listen while continuously interpreting and producing nonverbal cues, such as nods, smiles, and gestures. To support successful human-agent interaction, agents must model full-duplex audiovisual conversation; however, existing full-duplex benchmarks evaluate only speech. In this work, we present VideoFDB, the first benchmark to evaluate full-duplex audio-visual-to-audio-visual (AV2AV) conversational agents. VideoFDB contributes (i) 237 dyadic clips spanning 11 nonverbal conversational dynamics from real-world video calls, (ii) a taxonomy separating perception from generation behaviors, and (iii) a rubric-based LM-as-judge evaluation framework with interpretable axes for assessing conversational quality with respect to nonverbal conversational dynamics. Across open- and closed-source vision-speech agents, we find systematic failure modes: captioning collapse and visual-stream ignorance, and we show that current systems exploit vision for explicit visual question answering but not for the streaming joint audiovisual grounding required in natural conversation. We further evaluate cascaded speech-to-avatar systems and find that their architecture fundamentally precludes the production of full-duplex nonverbal cues. As the first benchmark for full-duplex AV2AV interaction, VideoFDB establishes a foundation for systematic evaluation and, we hope, will accelerate the advancement and development of next-generation multimodal conversational agents.
[HC-3] Do Proactive Agents Really Need an LLM to Decide When to Wake and What to Anchor?
链接: https://arxiv.org/abs/2605.30152
作者: Xiaoze Liu,Ruowang Zhang,Amir H. Abdi,Michel Galley,Zhikai Chen,Siheng Xiong,Xiaoqian Wang,Jing Gao
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注: 31 pages, 5 figures, 7 tables
Abstract:Proactive agents read user activity as text and call an LLM on every event to decide whether to act. But user activity is not natively text: it is a structured event stream of (actor, verb, object, timestamp) tuples that the operating system already maintains in graph form. Rendering the structure as text and asking an LLM to recover it is a round-trip the system never had to take. We treat the always-on signal as graph updates rather than text and use a small temporal-graph-learning (TGL) model as the encoder: one forward pass yields a per-event trigger probability and a per-entity routing score, and only the downstream agent (turning a small structured handoff into a fluent user-facing sentence) is an LLM call, invoked only when the trigger fires. TGL improves F1 on each of 14 backbones (mean +16.7, up to +46.0); in trigger-architecture comparisons, one TGL checkpoint gives the strongest trigger AUCs and the most stable deployed threshold. It runs at 11.13 ms per event on a GPU server and 13.99 ms on a consumer laptop, approximately 4–7x and 12–83x faster than every single-forward LLM-as-trigger configuration tested in each regime, with an approximately 220 MiB BF16 resident footprint deployable on-device alongside the privacy-sensitive activity stream it consumes.
[HC-4] REACT: A Conditioning Framework for User-Adaptive sEMG Hand Pose Estimation
链接: https://arxiv.org/abs/2605.30127
作者: Eric Xie,Hei Shing Cheung
类目: Human-Computer Interaction (cs.HC); Signal Processing (eess.SP)
备注: 6 pages, 3 figures
Abstract:Surface electromyography (sEMG) enables continuous hand pose estimation on wearable devices, but models trained on multi-user corpora degrade on unseen individuals due to inter-user variability in anatomy and electrode placement. We propose REACT, a lightweight conditioning framework that personalizes a frozen pretrained EMG-to-pose backbone at inference time using only a handful of calibration recordings. REACT learns a compact user embedding from calibration data and applies Feature-wise Linear Modulation (FiLM) to adapt the shared encoder’s feature space, requiring no gradient updates at deployment. On the large-scale EMG2POSE benchmark, REACT improves over the state-of-the-art baseline across all three generalization splits in both regression and tracking modes, reducing angular error by up to 3.9% with minimal parameter overhead and under 45 seconds of per-user calibration.
[HC-5] A Domain-Informed Multi-Objective Framework for EEG Channel Selection in Motor Imagery BCIs
链接: https://arxiv.org/abs/2605.29943
作者: Dekka Muni Kumar,Dhruba Jyoti Kalita,Yogesh Kumar Meena
类目: Human-Computer Interaction (cs.HC); Emerging Technologies (cs.ET); Machine Learning (cs.LG)
备注: This work has been submitted to the IEEE for possible publication
Abstract:Motor imagery (MI) classification using electroencephalography (EEG) signals is essential for advancing brain-computer interfaces (BCIs). Traditional EEG channel selection methods often face limitations, such as dependency on single-objective criteria and susceptibility to local optima. To address these challenges, this work proposes a multi-objective optimisation framework that employs non-dominated sorting genetic algorithm, multiple-objective particle swarm optimisation, and a multi-objective evolutionary algorithm based on decomposition. Our approach effectively balances spatial relevance, using a Gaussian kernel, and functional discriminability, which assesses intratrial task-related desynchronisation, thereby improving performance. We evaluated this framework on four EEG datasets: Physionet, OpenBMI, HighGamma, and BCIIV-2A. The proposed approach successfully identifies compact, relevant channel subsets concentrated around sensorimotor cortex regions linked to MI activity, addressing the prevalent challenges of dimensionality and complexity inherent to traditional techniques. Furthermore, the framework achieved classification performance of 87%, 71%, 75%, and 65% on the Physionet, OpenBMI, HighGamma, and BCIIV-2A datasets, respectively. By outperforming existing single-objective and accuracy-based methods, and those relying on fixed subsets, these findings demonstrate that this new multi-objective optimisation framework can enhance MI-based BCI performance while facilitating compact channel configurations with reduced computational complexity, making them better suited for wearable, portable, and real-time BCI applications.
[HC-6] oward AI Systems That Understand Self and Others: A Multi-Phase Inference Framework for Human Cognitive Diversity and World-Model Alignment
链接: https://arxiv.org/abs/2605.29930
作者: Toru Takahashi
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注: 50 pages, including appendices
Abstract:Mutual misunderstanding in contemporary society does not arise merely because people hold different opinions or values. Even under the same observations, different subjects may form different inferential targets, state representations, prediction errors, and update priorities. This paper proposes a multi-phase inference framework and defines its core internal mechanism as the Multi-Phase Inference Mechanism (MIM). MIM formalizes how heterogeneous world models arise through a phase-formation space, a foregrounding field, subject-specific profile states, and alignment maps between state representations. On this basis, the paper reframes world-model alignment as the problem of making heterogeneous representations mutually processable, rather than forcing agreement or convergence to a single value system. It further connects this formalism to philosophical disagreements, cognitive typology, social fragmentation, and AI alignment. The aim is to provide a constructive vocabulary for AI systems that can help humans understand self and others by making differences in meaning, value, and prediction error visible, comparable, and transformable.
[HC-7] Label Over Logic? How Source Cues Bias Human Fallacy Judgments More Than LLM s
链接: https://arxiv.org/abs/2605.29928
作者: Mahjabin Nahar,Nafis Irtiza Tripto,Aiping Xiong,Ting-Hao `Kenneth’ Huang,Dongwon Lee
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注:
Abstract:As AI-generated and AI-assisted content floods online spaces, source labels attached to such content can distort human reasoning judgments, with downstream consequences for moderation, evaluation, and decision-making. Whether LLMs share this vulnerability, or offer more source-agnostic evaluation, remains an open question with direct implications for human-AI collaboration. We examine this issue using logical fallacies as a controlled setting to isolate source-label effects on reasoning quality, independent of domain knowledge. We conduct an online study (N=505) where participants are assigned to a source condition (human, AI, human with AI assistance, AI with human assistance, or no disclosure) and evaluate comments containing logical fallacies, comparing their judgments with those of LLMs (GPT-5.2, Gemini 2.5 Flash, Claude Sonnet 4.5), who were evaluated across the same source conditions. Human evaluators were significantly more susceptible to fallacies labeled as written by human or human with AI assistance and assigned higher trust and evaluation ratings in these conditions. LLM evaluations remained comparatively stable across source labels, though performance varied across models. Confidence levels were similarly high across conditions for both humans and LLMs, regardless of fallacy presence. Our findings indicate that source-label bias in reasoning evaluation is primarily a human vulnerability and highlight the potential of human-LLM collaboration in increasingly AI-mediated environments.
[HC-8] Embodied Virtual Reality Feedback Reshapes Neural Representations to Support Continuous Three-Dimensional Motor Imagery Decoding
链接: https://arxiv.org/abs/2605.29677
作者: Niall McShane,Attila Korik,Karl McCreadie,Naomi Du Bois,Darryl Charles,Damien Coyle
类目: Human-Computer Interaction (cs.HC); Signal Processing (eess.SP); Neurons and Cognition (q-bio.NC)
备注: 28 pages, 7 figures, 3 tables. Submitted to Nature Biomedical Engineering. Data to be made available via Zenodo (DOI: https://doi.org/10.5281/zenodo.16047021 )
Abstract:Continuous brain-computer interfaces (BCIs) that decode motion trajectories from imagined movement offer intuitive motor control, yet how feedback modality and longitudinal training shape neural representations and decoding performance remains poorly understood. We present the first systematic investigation of embodied virtual reality (VR) feedback during real-time 3D virtual limb control driven by motor imagery, across ten longitudinal sessions in ten participants. Performance was evaluated using three strategies: actual online performance (Fixed Decoder Generalisation, FDG), periodic retraining (Sequential Adaptive Training, SAT), and within-session upper-bound estimation (Within-Session Reconstruction, WSR). A CNN-LSTM decoder achieved within-session imagined movement correlations of r = 0.762 under VR and r = 0.672 under screen feedback. VR significantly outperformed screen feedback across all strategies and movement dimensions (improvements of 8.9-13.0%, all p = 0.002, d = 1.42-2.05). This advantage persisted under fixed decoders without retraining, demonstrating that embodied VR feedback elicits inherently more decodable and generalisable neural representations. Linear mixed-effects modelling confirmed robust main effects of feedback modality and movement axis with no interaction. Neurophysiologically, VR produced stronger sensorimotor-parietal desynchronisation and enhanced motor-frontal functional connectivity, with pervasive anterior insula engagement across all frequency bands and increased superior parietal lobule coupling, paralleling patterns associated with real movement execution. These findings establish embodied spatial feedback as a key design principle for next-generation continuous BCIs targeting intuitive motor control and neurorehabilitation.
[HC-9] Learning to Feel Materials from Multisensory Tactile Data via Interpretable Models
链接: https://arxiv.org/abs/2605.29572
作者: Li Zou,Yasemin Vardar
类目: Robotics (cs.RO); Human-Computer Interaction (cs.HC)
备注: 12 pages, 3 figures, journal
Abstract:Human tactile perception of materials relies on complex multisensory touch cues, yet the relationship between low-level tactile signals and perceptual representations remains poorly understood. This knowledge gap hinders the integration of touch in digital environments and the development of robots capable of human-like tactile perception. Here, we present an interpretable computational framework for modeling human material perception and recognition using multisensory touch data. Our framework comprises three interconnected models: Model 1 maps finger-surface interaction features to psychophysical sensory attributes, Model 2 classifies materials based on these perceptual representations, and Model 3 directly classifies materials from tactile features. The results showed that combining information from pressing, static contact, and sliding interactions improves prediction accuracy, and that thermal cues are particularly informative for both perceptual modeling and material classification. These findings highlight the importance of thermal and compliance cues, which remain underrepresented in current robotic fingers and haptic displays. Incorporating such cues may enhance artificial systems’ ability to approximate human material perception and guide the design of more perceptually grounded haptic interfaces.
[HC-10] Understanding the Rising Human-AI Affective Bonding: Conceptualization and HAABI Scale Development
链接: https://arxiv.org/abs/2605.29484
作者: Lu Chen,Xiaoran Xue,Rongqi Ding,Fenghua Tang,Anji Zhou,Chenxi Wang,Mengyu Miranda Gao,Zhuo Rachel Han
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:As conversational AI becomes capable of sustained, affectively responsive interaction, users may form bonds beyond instrumental use. Existing measures often adapt interpersonal frameworks or focus on specific relational outcomes, leaving limited tools for assessing human-AI affective bonding on its own terms. Across two studies, we developed and validated the Human-AI Affective Bonding Inventory (HAABI). Study 1 used thematic analysis of semi-structured interviews with 52 emotionally engaged conversational AI users to identify cognitive, emotional, and behavioral features of bonding. Study 2 translated these insights into a self-report inventory and validated it among 673 Chinese conversational AI users. Exploratory and confirmatory factor analyses supported a 20-item, four-factor structure: emotional realism, separation anxiety, emotional investment, and romantic intimacy. The HAABI showed good reliability, construct validity, and known-groups validity. The scale therefore provides a neutral, user-centered tool for studying how affective bonds with conversational AI are formed, experienced, and related to users’ psychological outcomes.
[HC-11] MOOSE-Copilot: A Web-Based Interactive Assistant for Unified Exploratory and Fine-Grained Scientific Hypothesis Discovery ACL2026
链接: https://arxiv.org/abs/2605.29475
作者: Hongran An,Zonglin Yang
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Human-Computer Interaction (cs.HC)
备注: Accepted to ACL 2026 (System Demonstrations)
Abstract:Large language models (LLMs) show remarkable potential in scientific hypothesis discovery. However, existing approaches face two critical limitations: they treat divergent exploratory ideation and convergent fine-grained refinement as isolated tasks, and they operate autonomously with little to no human guidance. We present MOOSE-Copilot, the first unified framework to bridge this abstraction gap through a formalized human-AI interaction (HAII) protocol. Our system empowers scientists to steer the generative process via three explicit signals: initial blueprints, inter-stage routing, and regenerative feedback. Quantitative evaluations demonstrate that injecting these structured expert signals significantly outperforms purely autonomous baselines, establishing a performance ceiling under oracle guidance. Furthermore, to democratize this paradigm, we develop an intuitive web-based interface featuring interactive tree visualization. This explicitly eliminates the steep learning curve of complex command-line agentic tools, empowering interdisciplinary researchers to directly leverage, visually orchestrate, and accelerate end-to-end scientific breakthroughs.
[HC-12] Inform Coach Relate Listen: Auditing LLM Caregiving Support Roles
链接: https://arxiv.org/abs/2605.29473
作者: Drishti Goel,Agam Goyal,Veda Duddu,Olivia Pal,Jeongah Lee,Qiuyue Joy Zhong,Violeta J. Rodriguez,Daniel S. Brown,Dong Whi Yoo,Ravi Karkar,Koustuv Saha
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computers and Society (cs.CY); Social and Information Networks (cs.SI)
备注:
Abstract:Language models are increasingly being deployed for conversational support in informal caregiving contexts, where interactions often extend beyond information-seeking: caregivers seek emotional reassurance, guidance, and help, while navigating uncertain, relationally complex care decisions. Yet most safety evaluations assess model behavior under generic prompts, leaving a critical question unexamined: does a model’s safety profile change with its support role? We study this by operationalizing four expert-reviewed support roles grounded in social support theory: Inform, Coach, Relate, and Listen, and comparing them against two baseline controls: a basic prompting condition and a retrieval-augmented generation (RAG) condition. We evaluate across three language models (GPT-4o-mini, Llama-3.1-8B-Instruct, and MedGemma-1.5-4b-it) on 5,000 real-world queries from online Alzheimer’s Disease and Related Dementias (ADRD) communities. We find that the LLM’s support role systematically shapes both the prevalence and composition of interactional risks. Furthermore, a human evaluation study reveals a perceived quality–safety tension: more directive, information-oriented roles are rated as more helpful and trustworthy despite exhibiting elevated interactional risk profiles. We release ~90,000 support role-conditioned model responses with risk annotations as an ecologically grounded resource for research on safer LLM-mediated conversational support.
[HC-13] Usability Analysis of Configurator User Interfaces with Multimodal Large Language Models
链接: https://arxiv.org/abs/2605.29456
作者: Sebastian Lubos,Alexander Felfernig,Damian Garber,Adnan Kraljić,Tarik Kraljić,Viet-Man Le,Thi Ngoc Trang Tran,Gerhard Leitner,Julian Schwazer,Doris Suppan,Reinhard Willfort,Ivan Dukic,Jeremias Fuchs,Manuel Henrich
类目: oftware Engineering (cs.SE); Human-Computer Interaction (cs.HC)
备注:
Abstract:Configuration is a key technology for tailoring complex software systems, services, and products. A successful application of configurators not only depends on technical correctness, performance, and domain modeling but also on their usability. While general usability heuristics are widely used, configurator-specific criteria and tool support for systematic user interface (UI) analysis are limited. This paper explores the use of multimodal large language models (MLLMs) for scalable and semi-automated usability analysis of configurator UIs. We synthesize 18 configurator-specific usability criteria from the literature and apply these criteria in an MLLM-based analysis of 16 real-world configurators. Each criterion is assessed individually to generate severity ratings for usability issues and actionable improvement suggestions. A review of the results confirms that MLLMs can reliably identify configurator-specific usability issues and provide domain-aware improvement recommendations. Although human validation remains necessary, this approach has the potential to significantly reduce the required effort to analyze configurator usability.
[HC-14] How Coding Agents Fail Their Users: A Large-Scale Analysis of Developer-Agent Misalignment in 20574 Real-World Sessions
链接: https://arxiv.org/abs/2605.29442
作者: Ningzhi Tang,Chaoran Chen,Gelei Xu,Yiyu Shi,Yu Huang,Collin McMillan,Tao Dong,Toby Jia-Jun Li
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注:
Abstract:AI coding agents increasingly act directly within software environments, yet existing analyses of their failures rely on benchmark trajectories that miss how developers actually experience misalignment. We present an observational study of 20,574 coding-agent sessions from 1,639 repositories across IDE and CLI workflows. We operationalize misalignment as a breakdown made visible through developer pushback, and annotate each episode along four axes: form, cause, cost, and resolution. We identify seven recurring forms, spanning how agents read projects, interpret developer intent, follow rules, bound their actions, implement and execute code, and report progress. 90.50% of episodes impose effort and trust costs rather than irreversible system damage, yet 91.49% of visible resolutions still require explicit user correction. Misalignment patterns also differ across IDE and CLI settings, persist across adjacent sessions, and shift over time: while overall rates decline, constraint violations and inaccurate self-reporting grow in share. Our findings inform the design of training, evaluation, and interfaces for keeping coding agents aligned with real developer workflows.
[HC-15] Architecture-Sensitive Supervised Fine-Tuning for Screen-Conditioned Action Prediction: A PiSAR Benchmark
链接: https://arxiv.org/abs/2605.29400
作者: Rahul Bissa,Abhishek Vyas,Yash Jain
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注: 14 pages, 7 figures, 2 tables. PiSAR corpus and fine-tuned weights are proprietary to AprioriLabs; methodology and recipe released
Abstract:We benchmark three supervised fine-tuned models against frontier zero-shot baselines on a 661-row held-out slice of PiSAR (Persona, intent, Screen, Action, Rationale), a 12,929-tuple corpus of screen-anchored behavioural rationales curated from public app-store reviews, Pew American Trends Panel demographics, and the OPeRA shopper traces. Every model, frontier or fine-tuned, is evaluated on the same 661-row slice with the same scoring pipeline. Two findings. First, frontier zero-shot baselines (Claude Opus 4.7 and GPT-5.5) reach sem_sim 0.459 and 0.482 respectively; a fine-tuned Qwen3-VL-8B-Instruct reaches 0.783 and clears sem_sim = 0.7 on 79% of rows, against 1-2% for either frontier baseline, a gap of 0.30 absolute on the same test set. Second, the same training data and recipe on Gemma-4-26B-A4B-IT scores only 0.441, in the same band as the frontier zero-shot baselines rather than the fine-tuned Qwen. We read this as a recipe-vs-model mismatch: the reasoning-tuned high-parameter model resists displacement and would likely need either more data or a stronger fine-tuning method.
[HC-16] Expecting Empathy: How Interaction Context Shapes Norms for Empathic Response in Digital Communication
链接: https://arxiv.org/abs/2605.29399
作者: Tao Wang,Chi-Ching Juan
类目: Human-Computer Interaction (cs.HC); Social and Information Networks (cs.SI)
备注:
Abstract:A central challenge in affective computing is determining appropriate empathy levels for different interaction contexts. Prior work has characterized two poles: task-focused interactions, where empathy demand is near zero, and emotional disclosure, where empathy demand is high. This paper identifies a distinct intermediate type, decision support under stress, in which a sender faces a consequential choice while experiencing emotional difficulty. We hypothesize that this type elicits an asymmetric empathy profile: empathy comparable to emotional disclosure but instrumentality comparable to task-focused exchange. We test five hypotheses using 28,239 post-reply dyads from three Reddit advice communities, classified into three interaction types and scored for empathy depth, empathy form, and instrumental proportion using LLM-based annotation with pattern-based robustness checks. Results confirm the predicted asymmetric profile: decision-support-under-stress replies show significantly higher empathy than task-focused replies (M = 0.47 vs. 0.24, p 0.001) while maintaining high instrumentality (0.83 vs. 0.77 for emotional disclosure, p 0.001). Behavioral empathy dominates (36.6%), and community-validated response quality is negatively associated with empathic expression (r = -0.075, p 0.001). Community norms modulate baselines substantially but preserve the structural ordering. These findings establish a human empathy baseline for this interaction type and have direct implications for calibrating empathic expression in affective AI systems.
[HC-17] Offloading Score: Measuring AI Reliance Through Counterfactual Workflows
链接: https://arxiv.org/abs/2605.29392
作者: Vishakh Padmakumar,Lujain Ibrahim,Zora Zhiruo Wang,Jennifer Wang,Q. Vera Liao,Diyi Yang
类目: oftware Engineering (cs.SE); Computation and Language (cs.CL); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注: Preprint
Abstract:AI tools are increasingly integrated into real-world workflows. However, existing measures of reliance on these tools focus on AI output adoption or on self-reported indicators, rather than how task effort is distributed between users and tools. Here, we introduce offloading score, a measure of reliance that quantifies the fraction of cognitive effort offloaded to an AI tool. Offloading Score is simulation-based – we construct a counterfactual workflow by estimating how the user would have completed the task without the tool, and then computing the fraction of steps saved by using the tool. We validate offloading score through intrinsic evaluations of metric validity, and a controlled user study ( n=40 ) with developers performing programming tasks using AI tools. We vary time pressure to test whether reliance measures capture the known increase in reliance under time pressure. We show that offloading score detects significantly higher reliance in time-constrained settings ( +43% , p=0.018 ), while usage-based and self-reported baseline measures of reliance do not distinguish the conditions. We complement this with descriptive insights showing that higher reliance manifests as greater delegation of subtasks to the tool and more direct reuse of AI outputs. Finally, we demonstrate an approach of using offloading score in combination with target outcomes of a task (e.g., code understanding) to identify when reliance may be (in)appropriate. Our framework offers two contributions: an instrument users can apply to measure and reflect on their own reliance, and a quantitative signal that agent designers can utilize to mitigate overreliance.
[HC-18] MetaRanker: Human-in-the-loop Active Ranking for Metalens Image Quality
链接: https://arxiv.org/abs/2605.29212
作者: Yujin Park,Haejun Chung,Ikbeom Jang
类目: Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
备注: 12 pages, 6 figures
Abstract:Image quality in modern imaging systems emerges from the coupled effects of the sensor, optics, and computational reconstruction. Ultra-thin metalenses offer a path toward substantial miniaturization of optical modules, but practical designs often exhibit pronounced chromatic and field-dependent aberrations that necessitate computational reconstruction. In current metalens pipelines, reconstruction models are commonly trained and selected using distortion-based fidelity objectives, such as PSNR, yet these proxies can be weakly correlated with human preference and downstream utility, reflecting the well-known perception–distortion trade-off. We introduce MetaRanker, a human-in-the-loop active ranking framework that formalizes metalens image quality in terms of semantic interpretability, defined as the degree to which humans can reliably recognize objects and structures in the presence of optical artifacts. MetaRanker combines a probabilistic preference model with uncertainty-aware query selection, and leverages vision–language models to provide lightweight semantic priors. Importantly, these priors are used only to guide the sampling of informative comparisons; human judgments remain the primary supervision signal throughout. Across real-world and synthetic metalens datasets with distinct degradation profiles, MetaRanker produces rankings that align most closely with human assessments, while reducing the number of pairwise annotations required by approximately 80% relative to exhaustive pairwise evaluation. Finally, we show that standard image quality assessment metrics exhibit limited alignment with human interpretability in the metalens domain, positioning MetaRanker as a practical step toward perceptually grounded metalens evaluation and co-design.
[HC-19] Improving outdoor navigation for people with blindness using an AI-driven smartphone application and personalized audio guidance
链接: https://arxiv.org/abs/2605.29120
作者: Raymond Liu,Patrick Slade
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Globally, 340 million people have blindness or moderate to severe visual impairment (BVI) ^1 which limits independent outdoor navigation ^2 and negatively affects their health and quality of life ^3,4 . We surveyed 112 people with BVI and found that an ideal outdoor navigation aid must be able to perform turn-by-turn directions, path guidance, and obstacle detection and avoidance. Existing navigation tools such as white canes, guide dogs, and electronic travel aids often lack one or more of these criteria and may be expensive or inaccessible ^5,6 . Here we introduce Mobilio, a smartphone application that incorporates machine learning, sensor fusion algorithms, and personalized audio feedback to meet all of the outdoor navigation criteria. The reliability of the smartphone sensors and models used for navigation were assessed with engineering tests in representative navigation scenarios. We performed a series of experiments where Mobilio personalized audio feedback for participants with BVI (n = 14), guided them along an outdoor community path, and helped them navigate an obstacle course. Participants walking with Mobilio and a white cane reduced time to navigate a community path by 13 \pm 3% and environmental contacts by 41 \pm 5% compared to using Google Maps and a white cane. Mobilio achieved similar outdoor navigation reliability as a human guide. Participant surveys reported that Mobilio was easy to use, had a low perceived workload, and provided intuitive audio feedback. This work provides an accessible and personalized tool that may be an effective outdoor navigation aid to increase independence for people with BVI.
[HC-20] “Its OK Because…”: The Wild West of Student Rationalization of AI Use in Academic Writing
链接: https://arxiv.org/abs/2605.29090
作者: Jiyoon Kim,Kentaro Toyama,Sangmi Kim,John M. Carroll
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:Generative AI challenges academic integrity not only by enabling students to delegate substantial portions of their academic work, but also by blurring the ethical boundaries by which students distinguish acceptable assistance from misconduct. Drawing on semi-structured interviews (n=20), AI chat logs, and course documents (syllabi, submitted assignments), we investigated how students themselves make moral sense of AI use in academic writing. Our analysis results in a range of novel findings: First, there are at least five distinct sites of AI-use conceptualization, ranging from faculty’s intended AI policy, to students’ actual AI use. Second, students use over 20 distinct rationalizations to justify AI use, such as that copying AI-generated text is victimless; that any AI text reflecting their own beliefs or their own style is their own writing; or that they are learning more by using AI – even extensively – than otherwise. We present a taxonomy of these rationalizations, and show how some of them are employed to justify conscious violations of course policies. Third, student rationalizations occur in both an ad hoc and post hoc manner, and they are not necessarily self-consistent. These and other findings suggest that modern AI presents a steep, ethical, slippery slope which students conceptually slide down, landing far outside the pedagogical goals and expectations of instructors. We discuss implications for educational design and AI policy.
[HC-21] Designing for the Moment: How One-Minute Interventions Fit or Falter Across Domains
链接: https://arxiv.org/abs/2605.29051
作者: Zahra Hassanzadeh,Anne Hsu,Rachel Kornfield,David Haag,Ananya Bhattacharjee,Jay Olson,Jan David Smeddinck,Norman Farb,Alex Mariakakis,Lydia Chilton,Joseph Jay Williams
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:This paper explores the design space for one-minute digital interventions that prompt immediate action without onboarding or sensing. By embracing Fogg’s Behavior Model and four design principles informed by literature, the goal of these interventions was to provide triggers that encourage actions so simple that even people with low motivation would be willing to complete them. We examined the utility of these prompts by conducting a 14-day study with 22 participants interested in making small lifestyle improvements in at least one of three domains: physical activity, healthy eating, and mental well-being. When combined with insights drawn from participants’ rewrites of our prompts, our findings suggest that intentional personalization through co-authorship could be a lightweight personalization mechanism that balances relevance with low friction.
[HC-22] Mind Your Tone: Does Tone Alter LLM Performance?
链接: https://arxiv.org/abs/2605.29027
作者: Om Dobariya,Akhil Kumar
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注: 10 pages, 6 tables, 1 figure. Accepted as a full paper at the Thirty-second Americas Conference on Information Systems (AMCIS 2026), Reno. Follow-up to arXiv:2510.04950
Abstract:The use of Large Language Models (LLMs) is proliferating, yet their performance is observed to vary based on prompting styles and tones. In this study, we investigate both whether and how tonal variations in prompts lead to disparate LLM accuracy for objective multiple-choice questions. We use two datasets: a 50-base question dataset with five tone variants and a 570-base question MMLU subset spanning 57 subjects with seven tone variants. Experiments were conducted to evaluate the performance of four cost-efficient, popular LLMs: ChatGPT-4o, ChatGPT-5-nano, Gemini 2.5 Flash, and Gemini 2.5 Flash Lite. Across models, tonal effects are systematic but highly model-dependent. Some models show small, yet statistically significant, shifts, while others exhibit large accuracy swings across tones. Further, we identify subject-level differences in tone sensitivity and present a routing framework to explain how tones may attune internal reasoning modes. Our findings caution users against assuming tone-robust reliability in LLM deployments.
[HC-23] When Models Disagree: Rethinking LLM Evaluation for Public Comment Analysis
链接: https://arxiv.org/abs/2605.29025
作者: Aisha Najera,Alvin Moon,Vedant Srinivasan,Rajesh Veeraraghavan
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注:
Abstract:Federal agencies are deploying large language models (LLMs) to categorize public comment corpora, where the model’s organization of the record shapes what policymakers see and which arguments register. Standard evaluation, anchored on stance accuracy against a small validated set, cannot detect when different models produce materially different categorizations of the same public input. We propose an Interpretive Audit Pipeline that treats multi-model disagreement as diagnostic of interpretive complexity and directs human review toward genuinely ambiguous public input. Analyzing 1,260 public comments on a federal USDA docket across four LLMs, we find that inter-model thematic divergence exceeds within-model prompt variation, and that an expert rubric suppresses deep interpretive disagreement without resolving it. In a two-stage labeling study on a stratified 40-comment subsample, four LLMs and a human annotator labeled independently and then revised after seeing the others’ labels. Revision behavior varied across labelers, and the human annotator’s revisions frequently introduced framings absent from the ensemble’s collective output. We argue disagreement-based evaluation is a necessary complement to accuracy metrics for LLM-assisted interpretive coding.
[HC-24] Beyond Recall: Behavioral Specification as an Interpretive Layer for AI Personalization
链接: https://arxiv.org/abs/2605.28969
作者: Aarik Gulaya
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注: 134 pages, 4 figures. Code, data, judge prompts, and reproduction instructions: this http URL
Abstract:If an AI agent makes decisions on a person’s behalf, those decisions must align with its user. We introduce representational accuracy to measure how faithfully a system captures a person’s interpretation. An interpretive layer is operationalized as a Behavioral Specification. Our reference implementation aggressively compresses a person’s data into interpretive patterns, served as context to a language model. We evaluate the Specification on a prototype benchmark of held-out behavioral predictions scored by a calibrated 5-judge LLM panel. We test it independently and in composition with a range of context conditions: full raw corpus, full extracted facts, and four commercial memory systems (Mem0, Letta, Supermemory, Zep). Across 14 public-domain autobiographical corpora, the Specification lifts representational accuracy in aggregate and nearly eliminates model hedging. It recovers most of what the raw corpus delivers, at ~25x less context cost. The Specification lifts subjects toward a common predictive level regardless of pretraining baseline; the lift in absolute points is therefore largest where the baseline is lowest, suggesting the population of relevance is anyone not adequately represented in pretraining. Lift is greatest on interpretation-required questions, where providing an interpretive layer enables model behavior that extracted facts or raw corpus do not. Conversely, on recall-required questions, this layer can interfere rather than help. We conclude that representational accuracy is distinct from recall and that human-AI alignment is dependent on how accurately the user is represented. Representational accuracy makes that alignment testable. Comments: 134 pages, 4 figures. Code, data, judge prompts, and reproduction instructions: this http URL Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC) ACMclasses: I.2.7; I.2.0 Cite as: arXiv:2605.28969 [cs.CL] (or arXiv:2605.28969v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2605.28969 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[HC-25] he Trust Paradox: How CS Researchers Engage LLM Leaderboards
链接: https://arxiv.org/abs/2605.28966
作者: Pouya Sadeghi,Anamaria Crisan,Jimmy Lin
类目: Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注:
Abstract:Large language model (LLM) leaderboards rank AI models using standardized benchmarks and have become highly visible across computer science, despite known limitations in their reliability and robustness. Yet how they shape researchers’ actual practice remains empirically uncharted. We address this gap through semi-structured interviews with eight researchers across four computer science subfields, analyzed using reflexive thematic analysis. We find a near-universal paradox of pragmatic skepticism: while participants expressed deep distrust of leaderboard rankings, they continued to use them as rough decision-making aids. Peer networks, not leaderboards, emerged as the primary model selection mechanism, and arena-based (human-voting) leaderboards were consistently preferred over static benchmark leaderboards. Leaderboard influence varied sharply across subfields, revealing that disciplinary culture, not individual attitudes, mediates engagement; for instance, NLP researchers faced state-of-the-art comparison pressure while HCI and Systems/Privacy researchers reported none. Across these differences, however, participants converged on cost transparency as the most demanded missing feature (seven of eight). We translate these findings into concrete design recommendations that align evaluation infrastructure with how researchers actually use it, such as task-specific score breakdowns, cost integration, and voter-demographic disclosure.
[HC-26] Who Does Your AI Work For? Designing Conversational Agents as Digital Fiduciaries
链接: https://arxiv.org/abs/2605.28908
作者: Jacob Erickson
类目: Human-Computer Interaction (cs.HC); Computers and Society (cs.CY)
备注: To appear in the proceedings of the 8th ACM Conference on Conversational User Interfaces (CUI '26)
Abstract:Conversational agents are increasingly integrated into the most private and intimate aspects of users’ lives, from discussions of mental health to financial decisions. As a result, these systems have access to reams of sensitive user data. Much of the literature on AI systems has focused on aligning users’ goals with the agents that act on their behalf. While this work is vitally important, it may overlook the need to establish a new normative baseline. Conversational AI agents, designed to feel and interact anthropomorphically with human users, must be held to a standard of care commensurate with their capabilities and access. When a client hires a personal lawyer, undergoes surgery, or receives advice from an investment manager, the expert they consult often has a fiduciary duty to act in their client’s best interests. This provocation argues that conversational agents should be held to a similar standard and introduces fiduciary design as a guiding principle. In this respect, conversational AI trust and accountability could be unified into a single design and legal paradigm.
[HC-27] First head-to-head comparison of agent ic AI applied to the analysis of simulated data of the Einstein Telescope
链接: https://arxiv.org/abs/2605.28916
作者: Gianluca Inguglia
类目: Instrumentation and Methods for Astrophysics (astro-ph.IM); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注:
Abstract:We report a comparison of two state-of-the-art agentic AI systems, Claude Code (Anthropic) and Codex (OpenAI), tasked with autonomously executing a simple end-to-end gravitational wave data analysis pipeline on a shared computing infrastructure without human intervention. The pipeline comprises power spectral density estimation from raw Einstein Telescope simulated noise, geometric template bank generation, matched filter recovery of 100 binary black hole signal injections, automated results generation, and large language model-assisted production of a manuscript formatted in the style of Physical Review D. Both agents received identical written specifications and identical compute resources. The experiment was run twice: a first run with unrealistically loud injections, and a second run with signals rescaled to a physically motivated SNR range. The scientific results converged in both runs. However, the agents exhibited substantially different behaviors and computational costs: Claude Code completed the pipeline in ~3.4 minutes with silent deviations from the specification, while Codex required ~16 minutes across explicit self-correcting restarts, including an unsolicited performance optimization of the matched filter inner loop. The autonomously generated manuscripts also diverged in length, details, and quality. In the second run, a subtle difference in the interpretation of the SNR range instruction led to a genuine scientific divergence: Claude Code silently reinterpreted the instructions, while Codex followed the specification literally. We discuss the implications of these behavioral differences, such as speed versus auditability, silent versus transparent error handling, instruction interpretation, and the criticality of intermediate data representations in multi-model pipelines, for the deployment of agentic AI in scientific computing workflows.
计算机视觉
[CV-0] GMOS: Grounding Moving Object Segmentation in 3D Space and Time WWW
链接: https://arxiv.org/abs/2605.30352
作者: Junyu Xie,Tengda Han,Weidi Xie,Andrew Zisserman
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:Moving Object Segmentation (MOS) aims to discover, segment, and track objects that move independently of the camera. Current MOS methods, however, exhibit two fundamental limitations: they rely on pre-computed 2D auxiliary modalities such as optical flow or point trajectories that lack 3D geometric information, and they treat motion as a sequence-level attribute, overlooking the instantaneous motion state of each object. We address both by grounding MOS in 3D space and time, and propose GMOS, a framework that operates directly on RGB video to produce 3D-aware, temporally fine-grained segmentation of multiple moving objects, alongside a foreground–background variant GMOS-S for faster deployment. To support training and evaluation in this regime, we curate GMOS-2K, a dataset of 2,210 real-world videos with per-object temporal motion annotations drawn from five established Video Object Segmentation (VOS) benchmarks, and formalise MOS-I (“I” for instantaneous), a temporally fine-grained evaluation protocol with three complementary metrics. GMOS achieves state-of-the-art results across MOS, MOS-I, and Unsupervised VOS benchmarks, while running significantly faster than prior multi-object MOS methods and supporting online inference for streaming deployment.
[CV-1] VideoMLA: Low-Rank Latent KV Cache for Minute-Scale Autoregressive Video Diffusion
链接: https://arxiv.org/abs/2605.30351
作者: Hidir Yesiltepe,Jiazhen Hu,Tuna Han Salih Meral,Adil Kaan Akan,Kaan Oktay,Hoda Eldardiry,Pinar Yanardag
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project Page: this https URL
Abstract:Long-rollout causal video diffusion has converged on a fixed-size sliding-window KV cache, with recent progress innovating within this layout by changing which tokens occupy the window or how their positions are encoded. The per-head KV layout itself, a dominant contributor to streaming memory and latency, has been mostly left unchanged. In this paper, we present the first study of Multi-Head Latent Attention (MLA) in video diffusion. VideoMLA replaces per-head keys and values with a shared low-rank content latent and a shared decoupled 3D-RoPE positional key, reducing per-token KV memory by 92.7% at every cached layer. We further investigate why MLA succeeds in video diffusion even though the spectral assumption often used to motivate it in language models does not hold: pretrained video attention is not low-rank, with 99%-energy effective rank far above any practical latent dimension. VideoMLA retains quality at compression ratios where direct spectral approximation would predict large reconstruction error. We show that the MLA bottleneck, rather than the pretrained spectrum, determines the effective rank: both spectral and random initialization occupy nearly the full rank budget from initialization, and training preserves this budget while adapting within it. On VBench, VideoMLA matches short-horizon streaming video diffusion baselines, achieves the best overall score at long horizons among evaluated methods, and improves throughput by 1.23x on a single B200.
[CV-2] AdaState: Self-Evolving Anchors for Streaming Video Generation
链接: https://arxiv.org/abs/2605.30349
作者: Yusuf Dalva,Pinar Yanardag
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Autoregressive video diffusion models generate streaming video by producing frames sequentially, conditioning each chunk on previously generated content. These models are structurally anchored to the first frame: its key-value representation occupies a privileged position in the attention cache and serves as the primary scene reference throughout generation. As the cleanest and most error-free position in the cache, this anchor draws disproportionate attention, suppressing video dynamics, and locking scene composition to the initial viewpoint even as the scene naturally evolves. The result is a temporally shallow video in which motion, camera movement, and scene progression are dampened in favor of static consistency. To address this, we replace the static anchor with an adaptive state, a hidden latent that the model denoises alongside content at every chunk but never renders. Rather than referencing a frozen first frame, the model generates its own scene anchor at each step by attending to both the previous state and the current content, producing a reference that evolves with the generated content. Unlike standard video generation, which encodes an absolute notion of time, our formulation treats time as relative: every generation step sees the same positional structure regardless of how far generation has progressed, and the state transition is identical at every chunk. Together, these properties introduce a recurrence into the generation process, where denoising serves as the transition function, and the KV cache serves as the carrier, requiring no external module. Experiments demonstrate that the adaptive state substantially improves video dynamics, enabling richer motion and natural scene progression within generated videos.
[CV-3] NeuROK: Generative 4D Neural Object Kinematics CVPR2026
链接: https://arxiv.org/abs/2605.30347
作者: Chen Geng,Guangzhao He,Yue Gao,Yunzhi Zhang,Shangzhe Wu,Jiajun Wu
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: CVPR 2026
Abstract:Data-driven approaches have revolutionized 3D vision, enabling transformers to effectively reconstruct and generate static 3D objects. However, generating simulative 4D dynamics – realistic temporal deformations of static objects under various physical conditions – remains challenging and often ad hoc, despite its importance in building comprehensive 3D world models. Most existing methods assume a predefined physical model and use system identification to estimate parameters, restricting these methods to specific categories and small-scale datasets. We propose that these restrictions can be overcome by learning a data-driven kinematic state parameterization for object-centric physical systems. Specifically, we learn both a latent space representing all possible states of the object and a decoder that maps any sampled latent to a plausibly deformed shape of the object. We refer to this parameterization as Neural Object Kinematics (NeuROK), and learn a transformer-based encoder-decoder model on a curated large-scale 4D dataset. This formulation and the learned model significantly simplify the generation of simulative dynamics since we only need to consider the dynamics within a low-dimensional latent space from the Lagrangian mechanics’ perspective in classical physics. We demonstrate the effectiveness and generality of this neural simulation framework across diverse dynamic object types, showing clear advantages over prior works. Project page: this https URL
[CV-4] YoCausal: How Far is Video Generation from World Model? A Causality Perspective WWW
链接: https://arxiv.org/abs/2605.30346
作者: You-Zhe Xie,Yu-Hsuan Li,Jie-Ying Lee,Kaipeng Zhang,Yu-Lun Liu,Zhixiang Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:As video diffusion models (VDMs) advance toward world models, a key question arises: do they truly understand causality, or merely overfit to statistical temporal patterns? Existing benchmarks mostly rely on synthetic data, limiting real-world generalization due to the sim-to-real gap. We present YoCausal, a two-level benchmark inspired by the Violation of Expectation (VoE) paradigm from cognitive science. By temporally reversing real-world videos at zero cost as natural counterfactual samples, YoCausal establishes an arbitrarily extensible evaluation protocol. Level 1 introduces the Reverse Surprise Index (RSI), quantifying arrow-of-time perception via denoising loss. Level 2 introduces the Causality Cognition Index (CCI), which leverages a VLM to stratify datasets into causal and non-causal subsets, disentangling genuine causal reasoning from temporal bias. Evaluation of 13 state-of-the-art VDMs reveals that perceiving the arrow of time does not imply understanding causality, and a significant gap persists relative to human-level causal cognition.
[CV-5] Uncertainty-driven 3D Gaussian Splatting Active Mapping via Anisotropic Visibility Field CVPR2026
链接: https://arxiv.org/abs/2605.30342
作者: Shangjie Xue,Jesse Dill,Dhruv Ahuja,Frank Dellaert,Panagiotis Tsiotras,Danfei Xu
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: Accepted to CVPR 2026. Project page this https URL
Abstract:We present Gaussian Splatting Anisotropic Visibility Field (GAVIS), a novel framework for uncertainty quantification and active mapping in 3DGS. Our key insight is that regions unseen from the training views yield unreliable predictions from the 3DGS. To address this, we introduce a principled and efficient method for quantifying the visibility field in 3DGS, defined as the anisotropic visibility of each particle with respect to the training views, and represented using spherical harmonics. The resulting visibility field is integrated into a Bayesian Network-based uncertainty-aware 3DGS rasterizer, enabling real-time (200 FPS) uncertainty quantification for synthesized views. Active mapping is further performed within a maximum information gain framework building on this formulation. Extensive experiments across diverse environments demonstrate that GAVIS consistently and significantly outperforms prior approaches in both accuracy and efficiency. Moreover, beyond standalone use, our method can be applied post-hoc to improve the performance of existing approaches.
[CV-6] GPIC: A Giant Permissive Image Corpus for Visual Generation
链接: https://arxiv.org/abs/2605.30341
作者: Keshigeyan Chandrasegaran,Kyle Sargent,Suchir Agarwal,Michael Jang,Michael Poli,Juan Carlos Niebles,Justin Johnson,Jiajun Wu,Li Fei-Fei
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 25 pages; Dataset: this https URL Project website: this https URL
Abstract:Studying scalable methods for visual generative modeling requires large, accessible, and stable datasets. We introduce GPIC, a Giant Permissive Image Corpus of approximately 28 trillion pixels. GPIC comprises diverse internet images captioned by a state-of-the-art vision-language model, including 100M training, 200K validation, and 1M test examples. Moreover, all GPIC images are permissively licensed for both research and commercial use. GPIC is safety-filtered, deduplicated, and centrally hosted on Hugging Face. We provide a benchmarking protocol for generative modeling on GPIC. Finally, we provide a reference baseline for pixel-space flow matching on GPIC. Our dataset, benchmark, and models are available at this https URL. Evaluation toolkit and code are available at this https URL
[CV-7] Benchmarking Single-Factor Physical Video-to-Audio Generation CVPR2026
链接: https://arxiv.org/abs/2605.30339
作者: Tingle Li,Siddharth Gururani,Kevin J. Shih,Gantavya Bhatt,Sang-gil Lee,Zhifeng Kong,Arushi Goel,Gopala Anumanchipalli,Ming-Yu Liu
类目: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注: CVPR 2026
Abstract:Generative video-to-audio (V2A) models produce highly plausible soundtracks, but it remains unclear whether they capture the underlying physical processes. Existing evaluations emphasize perceptual realism and overlook physical correctness under controlled interventions. In this paper, we introduce FlatSounds, a benchmark that audits the physical reasoning of V2A models through: 1) controlled counterfactual pairs in which a single physical factor is varied, and 2) single-video pattern tests that probe internal consistency and directional trends. These settings test whether the generated audio correctly reflects specific physical properties and timings. Our evaluation of state-of-the-art models reveals a consistent trade-off: models rely more on text captions than the visual stream to infer physics and semantics. Captions generally improve physical and semantic accuracy, but paradoxically degrade temporal alignment. Our results highlight the need to move beyond audio quality toward learning physical processes directly from pixels. Finally, we find that our physics-based metrics correlate strongly with human preference tests on our own data. Project webpage: this https URL
[CV-8] REST3D: Reconstructing Physically Stable 3D Scenes from a Single Image
链接: https://arxiv.org/abs/2605.30338
作者: Xiaoxuan Ma,Jiashun Wang,Nicolas Ugrinovic,Yehonathan Litman,Kris Kitani
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Reconstructing physically stable 3D scenes from a single RGB image enables casual images to be converted into simulation-ready digital assets for applications such as immersive interaction and content creation. However, existing single-image reconstruction methods fall short in capturing the physical structure of a scene. As a result, they often produce geometrically plausible but physically inconsistent results, including object floating and penetration, which lead to unstable behavior in physics simulations. Image-conditioned scene generation methods improve physical plausibility but often rely on strong scene priors, yielding plausible yet inaccurate object arrangements that fail to match the input image. We propose REST3D, a single-image reconstruction framework that can reconstruct physically stable 3D scenes by integrating physical scene understanding with physics-constrained refinement. We first introduce an agentic physical scene understanding technique that constructs a scene-tree representation capturing object physical states and inter-object relationships from a gravity-support perspective, providing a structural prior for reconstruction. Leveraging this structure, we initialize the scene using image-to-3D models, followed by scene-tree-guided alignment and physics-constrained optimization to resolve physical violations while preserving visual consistency with the input image. Experiments show that our method significantly reduces physical errors and improves simulation stability on both synthetic and real-world datasets while maintaining strong reconstruction quality. We further demonstrate the reconstructed scenes in VR-based human-object interaction, showing their potential for immersive applications.
[CV-9] Colored Noise Diffusion Sampling
链接: https://arxiv.org/abs/2605.30332
作者: Hadar Davidson,Noam Issachar,Sagie Benaim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Diffusion models achieve state-of-the-art image synthesis, with their generative trajectories fundamentally exhibiting a spectral bias, resolving low-frequency global structures early and high-frequency fine details later. Conventional stochastic differential equation (SDE) solvers fail to account for this dynamic, naively injecting uniform white noise throughout the entire process and misusing the finite energy budget. In this work, we establish a mathematical framework that reconsiders SDE inference as a targeted, frequency-decoupled energy transfer. Leveraging this framework, we introduce Colored Noise Sampling (CNS), a novel, training-free stochastic solver. Rather than injecting uniform white noise, CNS utilizes a dynamic, timestep- and frequency-dependent schedule that more efficiently allocates injected energy toward structurally unresolved frequency bands. By actively exploiting the model’s inherent spectral bias, CNS systematically steers the generated distribution toward the true data manifold. Extensive experiments demonstrate that CNS significantly outperforms standard ODE and SDE baselines as a strictly plug-and-play, inference-time sampler substitution across diverse architectures (SiT, JiT, FLUX). Compared to standard sampling on ImageNet-256, CNS achieves substantial unguided FID reductions, improving from 8.26 to 6.27 on SiT-XL/2, 32.39 to 26.69 on JiT-B/16, and 11.88 to 8.31 on JiT-H/16, while yielding consistent relative FID improvements with Classifier-Free Guidance. Project page is available at this https URL.
[CV-10] Supercharging Thermal Gaussian Splatting with Depth Estimation
链接: https://arxiv.org/abs/2605.30328
作者: Manoj Biswanath,Chenxin Cai,Hannah Schieber,Daniel Roth,Benjamin Busam
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 8 pages, 4 figures. Accepted and will be published in ISPRS proceedings (ISPRS Congress 2026)
Abstract:Efficient and robust 3D scene representation is crucial in autonomous driving, robotics, and related fields. While RGB images provide valuable content for 3D reconstruction, other modalities like thermal or depth can enable additional information on the environment. Lately, novel view synthesis methods like 3D Gaussian Splatting have started using multiple modalities to further boost their performance. But fusing or combining multimodal data can make the process slower and can bring in additional challenges. Therefore, our project aims to use single modality based on thermal infrared domain, by removing the reliance on visible light as much as possible. This single modality can be expected to be faster as it does not rely on multimodal data. We propose a method, Thermal-to-Depth Gaussian Splatting (TDg), that uses only thermal images and depth estimation in its architecture to derive the radiance fields. Our TDg method outperforms the MSMG (Multiple Single-Modal Gaussians) baseline in most cases on our test datasets, RGBT-Scenes and ThermalMix. On average, the rendering quality metrics such as learned perceptual image patch similarity (LPIPS), structural similarity index measure (SSIM), and peak signal-to-noise ratio (PSNR) of TDg are 1.12%, 0.034%, and 0.01% better than the baseline MSMG values. It also reduces the training time significantly, by 12 mins 47 secs (55% improvement). Overall, our method is successful in deriving these thermal radiance fields, which can ultimately have several applications, such as identifying heat sources critical in surveillance, search or rescue operations, and industrial inspections where temperature is widely used to monitor machines.
[CV-11] Veda: Scalable Video Diffusion via Distilled Sparse Attention ICML2026
链接: https://arxiv.org/abs/2605.30325
作者: Shihao Han,Hao Yang,Xinting Hu,Xiaofeng Mei,Yi Jiang,Xiaojuan Qi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to ICML 2026
Abstract:Scaling Diffusion Transformers to generate high-resolution, long videos is constrained by the quadratic cost of self-attention, and existing sparse attention methods degrade under high sparsity. We show empirically that generation quality is determined not by the sparsity ratio itself, but by how well the sparse mask aligns with the tile-wise geometry of full attention. Based on this insight, we propose Veda, a distilled sparse attention framework that formulates tile selection as an explicit reconstruction problem from full attention. Veda integrates statistics-aware tile scoring with head-aware tiling to reduce estimation error and structural mismatch, enabling aggressive sparsity. A hardware-efficient tile-skipping kernel converts theoretical sparsity into practical wall-clock speedups. Experiments on large video diffusion models, including Waver and Wan2.1, demonstrate substantial acceleration with no noticeable degradation in generation quality. To generate 720P 10-second videos on Waver-T2V-12B, Veda achieves a 5.1 \times end-to-end speedup and a 10.5 \times self-attention speedup, reducing attention overhead from 92% to 50%. Notably, the gains increase with sequence length, indicating that Veda scales favorably with spatiotemporal resolution across models.
[CV-12] MonoPhysics: Estimating Geometry Appearance and Physical Parameters from Monocular Videos
链接: https://arxiv.org/abs/2605.30320
作者: Daniel Rho,Jun Myeong Choi,Matthew Thornton,Biswadip Dey,Roni Sengupta
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Existing inverse physics methods recover physical parameters from multi-view videos, where geometric constraints across views resolve scale and 3D structure. In monocular settings, however, such constraints are absent, leading to severe scale ambiguity, inaccurate geometry, and weak coupling between appearance optimization and physical simulation. We propose MonoPhysics, a framework for monocular inverse physics estimation of deformable objects using differentiable MPM simulation and 3D Gaussian Splatting, which jointly optimizes geometry, appearance, and physical parameters from a single camera view. We address these challenges through three visual-physical bridges: global scale alignment, physics-aware geometry refinement, and a differentiable position map, which together enable accurate optimization from monocular observations alone. We evaluate on Vid2Sim and our new dataset of elastic and plastic objects, showing that MonoPhysics outperforms existing baselines in monocular settings and achieves performance comparable to multi-view baselines using only a single camera. Our project page is available at this https URL
[CV-13] Before the Shutter: Aesthetic and Actionable Portrait Photography Planning in 3D Scenes
链接: https://arxiv.org/abs/2605.30318
作者: Ruixiang Jiang,Chang Wen Chen
类目: Graphics (cs.GR); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Portrait photography is largely decided before the shutter opens: the subject’s pose, the camera configuration, and the lighting devices must be coordinated within the surrounding 3D scene. In contrast, most existing computational methods focus on post-production in 2D image space, such as retouching, relighting, or editing images that already exist; pre-capture photographic planning remains largely unexplored. We introduce 3D aesthetic portrait planning, the task of generating human pose, camera, lighting, and exposure plans that produce visually compelling portraits while satisfying geometric and photometric feasibility in a 3D scene. Our approach builds a Photographic Scene Graph that represents scene affordances, subject-scene relations, and portrait-relevant lighting structure. Built on this representation, we perform aesthetic-guided comparative planning over previous attempts and current viewfinder observations. Experiments across diverse indoor and outdoor scenes show that our method produces portraits preferred by human raters and MLLM evaluators over competitive baselines, while maintaining high physical plausibility. Together, our results suggest a path from post-capture correction toward pre-capture computational portrait planning. Project repository: this https URL
[CV-14] VPG: Visual Prefix Guidance for Autoregressive Image and Video Generation
链接: https://arxiv.org/abs/2605.30317
作者: Xinyao Liao,Qiyuan He,Yicong Li,Jiayin Zhu,Xiaoye Qu,Wei Wei,Angela Yao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Autoregressive image and video generators are trained with teacher-forced histories but must sample from their own generated prefixes at inference time, making them vulnerable to exposure bias and prefix drift. Existing remedies either modify training or apply sampling-time guidance aimed primarily at external semantic conditions, such as class labels or text prompts, rather than testing whether a next-step prediction provides strong posterior support for the generated prefix itself. We propose Visual Prefix Guidance (VPG), a training-free inference-time guidance method for autoregressive image and video generation. VPG improves next-step prediction by contrasting the model’s output under the generated prefix with its output under a corrupted prefix, then extrapolating logits toward candidates that strengthen the posterior support of the generated prefix. Across class-conditional image generation with VAR, text-to-image generation with Infinity, and text-to-video generation with InfinityStar, VPG improves generation quality without retraining the base model, reducing FID on VAR by 0.36 on average and improving benchmark performance on both image and video generation.
[CV-15] Archon: A Unified Multimodal Model for Holistic Digital Human Generation CVPR2026
链接: https://arxiv.org/abs/2605.30311
作者: Chong Bao,Shichen Liu,Lijun Yu,David Futschik,Stylianos Moschoglou,Shefali Srivastava,Ziqian Bai,Feitong Tan,Guofeng Zhang,Zhaopeng Cui,Sean Fanello,Yinda Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to CVPR 2026. Project Page: this https URL
Abstract:Digital humans are fundamental to immersive interaction, yet creating a unified model for holistic modalities, including text, audio, motion, and visual content, remains an open challenge. In this paper, we present Archon, a fully pretrained, human-centric unified multimodal model for holistic avatar generation. Archon unifies seven modalities with modality-specific tokenizers, and a native autoregressive unified multimodal model pretrained on synchronized modalities and 72 diverse tasks to model holistic joint distributions. To address the token explosion challenge in high-fidelity talking videos, we introduce a memory-efficient semantic video reparameterization, achieving 4x token reduction while preserving fine-grained dynamics, coupled with a semantic-driven video diffusion decoder. We further propose a “Thinking in Modality” that decomposes ambiguous cross-modal tasks into stepwise thinking in an alternative chain of modality, progressively enhancing fidelity and controllability. Extensive experiments demonstrate that Archon achieves superior or comparable performance across diverse digital human generation tasks, validating the effectiveness of our unified framework. Project page: this https URL.
[CV-16] City-Mesh3R: Simulation-Ready City-Scale 3D Mesh Reconstruction from Multi-View Images CVPR
链接: https://arxiv.org/abs/2605.30310
作者: Sayan Paul,Sourav Ghosh,Siddharth Katageri,Soumyadip Maity,Sanjana Sinha,Brojeshwar Bhowmick
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Graphics (cs.GR)
备注: Accepted to the USM3D Workshop Proceedings at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2026 as an Oral Presentation. Project page: this https URL
Abstract:City-scale 3D surface reconstruction from multiview images for downstream 3D simulation, poses highly challenging problems due to the scale and complexity of urban scenes. Existing city-scale 3D reconstruction methods based on NeRF, Gaussian Splatting etc. often fail to recover 3D meshes ready for simulation due to incomplete/missing geometry and irregular, noisy surfaces. Scaling existing small-scale 3D reconstruction methods to arbitrarily large urban scenes is highly infeasible due to their computational complexity. We present City-Mesh3R, a scalable framework for reconstructing watertight surface meshes directly from large unordered image collections. Unlike recent methods which use global sparse SfM point-cloud initialization followed by a distributed 3D dense reconstruction of large-scale scenes, our method follows an end-to-end images-to-mesh 3D reconstruction approach using a divide-and-conquer strategy. The sparse city map is reconstructed via topological image clustering, cluster-wise independent sparse SfM and map merging, without need for exhaustive image feature matching. Then this map is partitioned spatially to perform geometry-aware camera selection, followed by dense surface reconstruction and surface refinement using curvature-aware adaptive vertex density remeshing. These partition meshes are then stitched together to produce the global mesh of the city. The proposed end-to-end framework is evaluated on city-scale reconstruction datasets. As demonstrated by our qualitative and quantitative results, our proposed method yields high-fidelity watertight 3D meshes with regular geometry, capturing fine surface details, and is suitable for scaling to arbitrarily large scenes owing to the end-to-end processing in a distributed setting.
[CV-17] Grounded 3D-Aware Spatial Vision-Language Modeling WWW CVPR2026
链接: https://arxiv.org/abs/2605.30307
作者: An-Chieh Cheng,Yang Fu,Yatai Ji,Ligeng Zhu,Guanqi Zhan,Zhuoyang Zhang,Zhaojing Yang,Song Han,Yao Lu,Pavlo Molchanov,Vidya Nariyambut Murali,Jan Kautz,Xiaolong Wang,Hongxu Yin,Sifei Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: CVPR 2026 this https URL
Abstract:We present GR3D, a spatial vision language model equipped with three complementary grounding capabilities–explicit 2D grounding, implicit 2D grounding, and monocular 3D grounding–within a single framework. GR3D introduces an implicit grounding mechanism that identifies entity mentions during generation and inserts the corresponding region tokens into the text stream, allowing the model to reference visual evidence on the fly when producing spatial chain-of-thought responses. In parallel, a region-prompted monocular 3D grounding design predicts 3D bounding boxes in the camera view from grounded region queries, supported by intrinsic-aware normalization and dense geometric supervision. Together, these grounding capabilities enable GR3D to decompose complex spatial understanding problems into grounded 2D perception followed by 3D inference. GR3D achieves consistent improvements across grounded and non-grounded spatial benchmarks, demonstrating grounding as an effective inductive bias for strengthening spatial understanding in VLMs. These grounding capabilities collectively enhance general spatial understanding beyond the grounding task itself.
[CV-18] Boosting Image Quality Assessment Performance: Unsupervised Score Fusion by Deep Maximum a Posteriori Estimation ICASSP2024
链接: https://arxiv.org/abs/2605.30269
作者: Zhongling Wang,Raymond Zhou,Shahrukh Athar,Wenbo Yang,Zhou Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Image and Video Processing (eess.IV)
备注: 2024 International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2024)
Abstract:Over the past decades, numerous Image Quality Assessment (IQA) models have emerged, aiming to predict the perceptual quality of images. However, individual models are often biased toward certain types of image content or distortions, depending on the design principle and process. An intuitive idea is to harness the strengths and mitigate the weaknesses of each IQA model, by fusing the scores of multiple models into a stronger one. Here we make one of the first attempts to seek an optimal solution for the idea and propose a general framework for unsupervised IQA score fusion using deep Maximum a Posteriori (MAP) estimation. The proposed model conducts fine-grained uncertainty estimation at the score level to increase the accuracy and reduce the uncertainty in fused predictions. Comprehensive experiments demonstrate the superiority of the proposed model over individual IQA models and other fusion methods. It also exhibits an interesting capability of rejecting ``bad" models in the fusion process.
[CV-19] PhyGenHOI: Physically-Aware 4D Generation of Dynamic Human-Object Interactions
链接: https://arxiv.org/abs/2605.30268
作者: Omer Benishu,Gal Fiebelman,Sagie Benaim
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:We address the task of generating physically accurate and visually faithful 4D Human-Object Interaction (HOI). Given a static 3D human and target object represented as 3D Gaussian Splats (3DGS), our goal is to synthesize dynamic scenes where the human actively engages with the object through actions, such as punching or kicking, in accordance with a given input text. To this end, we introduce PhyGenHOI, a novel framework that couples generative human motion with an explicit physical object simulation. We model the human as a semantic agent driven by a Motion Diffusion Model (MDM) and the object as a physical agent simulated via the Material Point Method (MPM), utilizing 3D Gaussians as a unified, differentiable representation. We supervise their interaction through three coupled mechanisms: (1) A Windowed Attraction Loss that temporally synchronizes generative motion to intercept the object; (2) A Contact-Driven Re-simulation step that triggers physically consistent momentum transfer upon impact; and (3) A Masked Video-SDS objective that injects video-based priors to enhance contact fidelity. Experiments show PhyGenHOI generates physically consistent 4D HOI across diverse actions, humans, and objects, outperforming baselines. Project page and videos: this https URL
[CV-20] minWM: A Full-Stack Open-Source Framework for Real-Time Interactive Video World Models
链接: https://arxiv.org/abs/2605.30263
作者: Min Zhao,Hongzhou Zhu,Bokai Yan,Zihan Zhou,Yimin Chen,Wenqiang Sun,Kaiwen Zheng,Guande He,Xiao Yang,Chongxuan Li,Fan Bao,Jun Zhu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent video diffusion foundation models have achieved remarkable progress in high-quality video generation, yet turning them into real-time interactive video world models remains challenging. Interactive world models require controllable, causal, and low-latency rollout, which in practice demands a full pipeline spanning data construction, controllable fine-tuning, autoregressive training, few-step distillation, and streaming inference. In this work, we present minWM, a full-stack open-source framework for building real-time interactive video world models. minWM provides an end-to-end pipeline that converts existing bidirectional T2V/TI2V video foundation models into camera-controllable few-step autoregressive world models. Specifically, minWM first fine-tunes a bidirectional video diffusion model with camera control, and then applies the Causal Forcing / Causal Forcing++ pipeline, including AR diffusion training, causal ODE or causal consistency distillation, and asymmetric DMD, to distill it into a few-step autoregressive generator for low-latency rollout. The framework is modular and architecture-extensible: we instantiate it on representative open backbones, including Wan2.1-T2V-1.3B and HY1.5-TI2V-8B, covering both cross-attention-based condition injection and MMDiT-style architectures. minWM also supports adapting existing video world models, such as HY-WorldPlay, to new data distributions, training recipes, and latency targets. Beyond releasing runnable scripts, checkpoints, documentation, and inference code, we provide practical ablations on camera trajectory quality, controllability training steps, and minimal batch-size requirements. We hope minWM serves as a reproducible and extensible recipe for building and adapting real-time interactive video world models. Project Page: [this https URL](this https URL) Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2605.30263 [cs.CV] (or arXiv:2605.30263v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2605.30263 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-21] Stable-Layers: Fine-Tuning Image Layer Decomposition Models with VLM-Scored Reinforcement Learning
链接: https://arxiv.org/abs/2605.30257
作者: Ciara Rowles,Reshinth Adithyan,Nikhil Pinnaparaju,Vikram Voleti,Mark Boss
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 25 pages, 8 figures, 4 tables. Project page: this https URL
Abstract:We present Stable-Layers, a reinforcement learning framework that eliminates the need for paired supervision by fine-tuning a pretrained layer decomposition model using only feedback from a vision-language model (VLM). Starting from Qwen-Image-Layered, we apply Flow-GRPO with LoRA adaptation, sampling multiple candidate decompositions per image, scoring them with a VLM, and optimising the policy from group-relative advantages. The key challenge lies in designing a reliable reward signal: VLMs scoring samples in isolation tend to compress their judgements into a narrow band, leaving GRPO with little within-group variance to learn from. We address this with a two-stage evaluation pipeline that pairs structured per-sample scoring across five edit-centric criteria with a grid-based calibration step in which the VLM re-scores all candidates side-by-side. Stable-Layers produces decompositions with stronger layer separation, fewer blank or artifact-heavy layers, and lower per-layer reconstruction error on the Crello dataset compared to the base model.
[CV-22] Ambient-robust Inverse Rendering using Active RGB-NIR Imaging
链接: https://arxiv.org/abs/2605.30250
作者: Hoon-Gyu Chung,Jinnyeong Kim,Hyunwoo Kang,Seung-Hwan Baek
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 11 pages
Abstract:Inverse rendering aims to reconstruct geometry and reflectance of objects from images. Despite recent progress, existing methods often produces inaccurate reconstructions that are sensitive to ambient illumination conditions. Here we introduce an ambient-robust inverse rendering method enabled by active RGB-NIR imaging. Our key insight is to leverage near-infrared (NIR) flash illumination-imperceptible to human observers-to obtain stable point-light shading that is largely invariant to ambient illumination. By using multi-view RGB images illuminated by ambient light and NIR images acquired with active NIR flash illumination, we reconstruct accurate geometry and reflectance by exploiting the complementary benefits of RGB and NIR images via a three-stage inverse rendering method. To enable dense multi-view acquisition, we develop an active imaging system equipped with a RGB-NIR camera and a NIR flash mounted on a mobile base. Using this system, we collect the first multi-view RGB-NIR inverse rendering dataset captured under multiple ambient illumination conditions. Experiments demonstrate that our method outperforms prior approaches, achieving accurate geometry and reflectance estimation across multiple ambient lighting scenarios.
[CV-23] GenClaw: Code-Driven Agent ic Image Generation
链接: https://arxiv.org/abs/2605.30248
作者: Junyan Ye,Jun He,Zilong Huang,Dongzhi Jiang,Xuan Yang,Rui Chen,Weijia Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages, 7 figures
Abstract:Image generation models have evolved from text-conditioned pixel synthesis toward multimodal agents endowed with visual comprehension and tool invocation capabilities. Yet, existing agents remain at the mercy of underlying black-box image models. Their workflow is trapped in a repetitive cycle of prompt rewriting for generation refinement, leaving them with no mechanism to directly manipulate the canvas. In essence, the potential of LLMs to serve as a genuine “brush” for precise visual construction remains largely untapped. In this paper, we propose GenClaw, a code-driven agentic image generation paradigm that empowers the agent to create like a human artist: first conceptualizing, then sketching, and finally coloring. Specifically, the agent first constructs the conceptual knowledge and context through search and reasoning. It then utilizes code (e.g., SVG, HTML, this http URL) to render executable visual sketches. Finally, it employs an image generation model to supplement textures, materials, and photorealism. In this workflow, code serves as a controllable intermediate canvas bridging linguistic reasoning and pixel synthesis, seamlessly integrating programmatic logic with the visual expressiveness of generative models. By transforming image generation from a black-box paradigm into a staged process akin to authentic human creation, GenClaw offers a step toward for highly controllable and interpretable visual generation systems.
[CV-24] Reinforcement Learning with Robust Rubric Rewards
链接: https://arxiv.org/abs/2605.30244
作者: Ya-Qi Yu,Hao Wang,Fangyu Hong,Xiangyang Qu,Gaojie Wu,Qiaoyu Luo,Nuo Xu,Huixin Wang,Wuheng Xu,Yongxin Liao,Zihao Chen,Haonan Li,Ziming Li,Dezhi Peng,Minghui Liao,Jihao Wu,Haoyu Ren,Dandan Tu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:While Reinforcement Learning with Verifiable Rewards (RLVR) is effective for deterministically checkable tasks, many vision-language tasks are partially verifiable, demanding multi-criteria supervision (e.g., perceptual details, reasoning steps, and constraints). Rubrics provide a natural interface for this fine-grained supervision, but their effectiveness depends on the execution accuracy during online RL. We propose Reinforcement Learning with Robust Rubric Rewards ( \textRLR^3 ), extending RLVR from task-level verification to criterion-level verification. \textRLR^3 routes instance-specific rubrics through two execution paths: an LLM-as-an-extractor paired with a deterministic verifier, or an LLM-as-a-Judge for non-verifiable criteria. To ensure faithful scoring, \textRLR^3 introduce a minimal exposure strategy that masks ground truths from extractors and images from judges. Furthermore, \textRLR^3 employs hierarchical aggregation to prioritize essential criteria over additional criteria, and mitigates score saturation within rollout groups. Evaluated on Qwen3-VL-30B-A3B across 15 benchmarks, \textRLR^3 consistently outperforms RLVR, yielding a 4.7-point improvement over the base model and exceeding the official instruct-to-thinking model gap. Controlled audits confirm our deterministic verification and minimal exposure significantly reduce exploitable false positives.
[CV-25] SAM3D-Phys: Towards Multi-Object Interactive Simulation in Real World
链接: https://arxiv.org/abs/2605.30239
作者: Xin Dong,Weijian Deng,Lihan Zhang,Tianru Dai,Wenfeng Deng,Yansong Tang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 23 pages, 11 figures
Abstract:This work addresses the problem of recovering complete, simulatable object geometry from reconstructed real-world scenes, enabling physics-based interaction with objects embedded in the scene. While modern multi-view reconstruction methods can produce visually accurate environments, objects are often incomplete due to occlusions and limited observations, making them unsuitable for physics simulation. To address this limitation, we propose SAM3D-Phys, a framework that integrates scene reconstruction with generative 3D priors of SAM3D to recover physically simulatable objects. Our approach first reconstructs the scene from multi-view images to obtain scene geometry and partial observations of objects. We then leverage SAM3D to infer complete object geometry from these partial observations. To ensure that the recovered objects remain consistent with the reconstructed scene, we restore scene-consistent object states through two complementary strategies: a physics-constrained spatial optimization algorithm that iteratively aligns the recovered object to its original location, and a mask-guided appearance distillation module that refines texture fidelity based on the observed images. By recovering complete object geometry and restoring its pose and appearance within the scene, SAM3D-Phys produces clean object representations suitable for physics-based simulation, enabling simultaneous and physically consistent interactive simulation of multiple objects within a reconstructed scene. Project page: this https URL
[CV-26] BullingerDB: A Dataset for Handwritten Text Recognition and Writer Retrieval ICDAR2026
链接: https://arxiv.org/abs/2605.30235
作者: Marco Peer,Anna-Scius Bertrand,Patricia Scheurer,Andreas Fischer
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted for presentation at ICDAR2026. Dataset available via zenodo
Abstract:We present BullingerDB, a large-scale benchmark dataset for historical document analysis based on the correspondence of Heinrich Bullinger (1504-1575). The corpus comprises 20,898 pages and 499,222 text lines written by 796 writers over six decades, featuring stylistic variation, multilingual content (mostly Latin and Early New High German) as well as meta-information such as writer identity and time. We evaluate BullingerDB on text recognition and writer retrieval. TrOCR, the best performing model, achieves a CER of 9.1%. For writer retrieval, we introduce a temporal nDCG metric to assess time-aware retrieval. While temporally coherent retrieval is achievable, mAP (78.3%) scores indicate challenges due to long-term stylistic variation. With BullingerDB, we aim to establish a new benchmark for multilingual historical text recognition and temporally-aware writer analysis.
[CV-27] Beyond 3D VQAs: Injecting 3D Spatial Priors into Vision-Language Models for Enhanced Geometric Reasoning CVPR2026
链接: https://arxiv.org/abs/2605.30231
作者: Chun-Hsiao Yeh,Shengyi Qian,Manchen Wang,Yi Ma,Joseph Tighe,Fanyi Xiao
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: CVPR 2026. Project page: this https URL
Abstract:Vision-Language Models (VLMs) often struggle with robust 3D spatial reasoning. Prevailing methods that rely on fine-tuning with 3D visual question-answering (VQA) datasets may overfit dataset-specific biases, while integrating specialized 3D visual encoders is often inflexible and cumbersome. In this paper, we argue that genuine spatial understanding should emerge from learning fundamental geometric priors, not only from high-level VQA supervision. We propose GASP (Geometric-Aware Spatial Priors), a framework that injects these priors directly into the LLM’s transformer layers. GASP employs a small correspondence head, applied as a deep supervision signal across all layers, and is trained with a dual objective leveraging ground-truth geometry from large-scale video scenes: a contrastive loss on ground-truth point correspondences enforces 2D view-invariance, while a depth consistency supervision resolves 3D geometric ambiguities. Our analysis first provides a diagnostic showing that standard VLMs’ internal correspondence matching accuracy is very low (often below 5%). We then demonstrate that our training substantially improves this behavior, boosting peak layer-wise correspondence to over 70% and maintaining over 85% temporal robustness while baselines remain below 5%. These internal improvements translate to significant gains on downstream spatial benchmarks including +18.2% on All-Angles Bench and +29.0% on VSI-Bench, all without training on any 3D VQA data. Our findings indicate that learning from fundamental geometric priors is a promising and generalizable pathway towards VLMs with more reliable 3D spatial reasoning.
[CV-28] IP-Adapter Is All You Need: Towards Fine-Tuning-Free Diffusion-Based Talking Face Generation
链接: https://arxiv.org/abs/2605.30230
作者: Hao Wu,Xiangyang Luo,Hao Wang,Jiawei Zhang,Yi Zhang,Jinwei Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:With the rapid advancement of diffusion models, talking face generation has made remarkable progress. However, existing diffusion-based methods still require task-specific fine-tuning and large-scale audiovisual datasets, resulting in high computational costs that hinder scalability and accessibility of diffusion-based approaches across the research community. To address this, we propose a finetuning-free paradigm that directly performs talking face generation using the pretrained weights of Stable Diffusion and IP-Adapter. This backbone leverages the visual embedding capability of IP-Adapter to mine lip-related semantics from the pretrained Stable Diffusion. To address the challenges of identity drift, synchronization errors, and temporal instability, we also design three trainable-parameterfree components: (1) the Structurist, which explicitly disentangles and reassembles lip and appearance features to mitigate identity drift and appearance distortion; (2) the Structure Controller, which adaptively refines embeddings based on quasi-monotonic motion trends for precise lip synchronization; and (3) the Noise Sensor, which introduces Gaussian prior to detect and suppress flicker and jitter artifacts and enhance temporal consistency. Experimental results show that our method outperforms existing SOTA approaches in both lip-sync accuracy (at least 0.16 gain in PCLD) and visual fidelity (at least 0.7 improvement in FID), establishing a novel fine-tuning-free diffusion framework for talking face generation.
[CV-29] Déjà View: Looping Transformers for Multi-View 3D Reconstruction
链接: https://arxiv.org/abs/2605.30215
作者: Alessandro Burzio,Tobias Fischer,Sven Elflein,Qunjie Zhou,Riccardo de Lutio,Jiawei Ren,Jiahui Huang,Shengyu Huang,Marc Pollefeys,Laura Leal-Taixé,Zan Gojcic,Haithem Turki
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent feed-forward 3D reconstruction transformers have scaled to over a billion parameters, following the broader trend of increasing model capacity in computer vision. Yet emerging evidence suggests that contiguous transformer layers often behave like repeated applications of similar operations, and multi-view reconstruction transformers refine their predictions progressively across decoder depth. We posit that model depth partially buys iteration, paid for inefficiently in unique parameters, and instead make that iteration explicit in architecture. Our model, DéjàView, applies a single looped transformer block recurrently to per-view features for K refinement steps. Trained once, it exposes K as an inference-time compute knob, matching or outperforming substantially larger feed-forward baselines across five reconstruction benchmarks spanning indoor, outdoor, object-centric, and driving scenes, while using a fraction of their parameters and comparable or lower compute. Importantly, the same looped block formulation outperforms an otherwise identical variant with independent per-step parameters under matched training data and compute, suggesting that explicit iteration is not merely a compute-efficient substitute for capacity but a stronger inductive bias for multi-view 3D reconstruction.
[CV-30] Cycle Consistency in Video Object-Centric Learning
链接: https://arxiv.org/abs/2605.30211
作者: Rongzhen Zhao,Zhiyuan Li,Ruonan Wei,Juho Kannala,Joni Pajarinen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 14 pages
Abstract:Self-supervised video Object-Centric Learning (OCL) aims to discover distinct objects and associate them across time, whereas self-supervised Multi-Object Tracking (MOT) focuses on associating pre-defined object detections or segmentations. Although well-established in MOT, Cycle Consistency (CC) cannot naively or explicitly apply to the latent slot space of OCL. Unlike the deterministic and ideal object representations in MOT, OCL slots are inherently stochastic and ambiguous due to non-unique scene decompositions. Enforcing explicit cycle consistency (ECC) on slots imposes rigid mean seeking. This severely penalizes the model for exploring alternative but equally valid decompositions, thereby driving towards feature collapse. To resolve this dilemma, we propose \textitImplicit Cycle Consistency (ICC), which shifts the cycle-consistency constraint from the restrictive slot space to the continuous reconstruction manifold, encouraging slots to reach a soft consensus on collectively interpreting the visual scene rather than forcing rigid point-to-point feature alignment. Extensive experiments on complex video OCL benchmarks demonstrate that ICC avoids feature collapse and outperforms ECC baselines. Our source code, model checkpoints and training logs are provided on this https URL.
[CV-31] LiveSVG: Zero-Shot SVG Animation via Video Generation
链接: https://arxiv.org/abs/2605.30174
作者: Matan Levy,Ran Margolin,Bar Cavia,Dvir Samuel,Yael Pritch,Shmuel Peleg,Alex Rav Acha,Ariel Shamir,Dani Lischinski
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project Page: this https URL
Abstract:We introduce LiveSVG, a zero-shot approach for generating Scalable Vector Graphics (SVG) animations using video diffusion models. Current SVG animation methods struggle with complex motions: LLM-based code synthesis fails to express fine, non-rigid Bézier deformations, while Score Distillation Sampling (SDS) provides noisy gradients and often requires category-specific priors like skeletons. In contrast, LiveSVG fits vector geometry directly to an explicitly generated target video. Given an input SVG image and a motion prompt, we generate a previewable target video using a frozen image-to-video model, then fit the original SVG to this video via differentiable rendering. Our fitting stage is skeleton-free, utilizing a dual-level motion representation that combines per-group homographies for coarse articulation with per-path Bézier control-point offsets for local deformations. To resolve color-induced correspondence ambiguities during pixel-wise fitting, we introduce a novel sphere-packing recolorization strategy. We also present ChallengeSVG, a benchmark of complex, multi-object scenes that exposes the limitations of prior work. Evaluations demonstrate that LiveSVG significantly outperforms existing methods on both AniClipart and ChallengeSVG, establishing direct reference-video fitting as a practical, robust route to prompt-aligned and fully editable vector animation.
[CV-32] Unveiling the Visual Counting Bottleneck in Vision-Language Models ICML2026
链接: https://arxiv.org/abs/2605.30170
作者: Xingzhou Pang,Yifan Hou,Junling Wang,Mrinmaya Sachan
类目: Multimedia (cs.MM); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: ICML 2026
Abstract:While Large Vision-Language Models (VLMs) excel at interpolation, they suffer catastrophic failures in systematic generalization, most notably in visual counting. In this work, we investigate this extrapolation bottleneck by deconstructing visual counting into three cognitive stages: visual individuation, magnitude awareness, and symbolic mapping. Using synthetic Go boards and linear probes, we demonstrate that visual backbones maintain robust, linearly separable representations of quantity well into the extrapolation regime, ruling out perceptual failure. Furthermore, models retain latent magnitude awareness, successfully performing comparative reasoning on quantities they fail to enumerate. We pinpoint the collapse to the symbolic mapping stage, where the model fails to project valid visual magnitudes onto symbolic tokens. Our findings support a frac tured magnitude hypothesis: VLMs fail to acquire a universal number space, instead learning disjoint, modality-specific statistical manifolds that prevent cross-modal grounding for unseen quantities. Validated on the state-of-the-art foundation model, our results suggest that bridging this gap requires inductive priors enforcing unified representations, as data scaling alone is insufficient.
[CV-33] OmniCD: A Foundational Framework for Remote Sensing Image Change Detection Guided by Multimodal Semantics
链接: https://arxiv.org/abs/2605.30168
作者: Chenhao Sun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Change detection (CD) in remote sensing is vital for applications such as urban monitoring and disaster assessment, yet traditional methods struggle with generalization across diverse scenarios. We present OmniCD, a foundational framework that unifies and enhances remote sensing CD through multimodal semantic guidance. OmniCD incorporates image and text prompts – such as textual descriptions, semantic maps, and geospatial metadata – into a unified architecture, supporting tasks from binary CD to zero-shot semantic change understanding. The framework integrates a hierarchical scene retrieval module and a change detection module, reinforced by a style disentanglement mechanism for improved cross-domain robustness. We further introduce RSITCD, a large-scale multimodal dataset with 300K+ annotated image-text pairs. Extensive experiments show that OmniCD achieves state-of-the-art performance across benchmarks, demonstrating strong adaptability and setting a solid foundation for general-purpose CD systems in remote sensing.
[CV-34] Why Far Looks Up: Probing Spatial Representation in Vision-Language Models
链接: https://arxiv.org/abs/2605.30161
作者: Cheolhong Min,Jaeyun Jung,Daeun Lee,Hyeonseong Jeon,Yu Su,Jonathan Tremblay,Chan Hee Song,Jaesik Park
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-language models (VLMs) achieve strong performance on spatial reasoning benchmarks, yet it remains unclear whether this reflects structured 3D understanding or reliance on statistical shortcuts in natural images. We introduce a representation-level analysis framework that constructs minimal contrastive pairs to measure how spatial axes are organized and disentangled within VLM embeddings. Our analysis across multiple model families reveals a consistent vertical-distance entanglement: models conflate vertical image position with distance, mirroring the perspective bias of natural photographs. This bias produces a significant accuracy gap between perspective-consistent and counter-heuristic examples, and intensifies under data scaling even as overall benchmark accuracy improves. We further show that models with similar benchmark scores can exhibit different internal representations, and that these differences predict accuracy and robustness across diverse spatial reasoning benchmarks. To isolate this bias from evaluation-set skew, we introduce SpatialTunnel, a synthetic benchmark designed to expose spatial shortcut biases by removing common correlations present in natural images. Experiments confirm that the entanglement is model-intrinsic, and that models with well-separated spatial axes exhibit greater robustness, suggesting that well-structured spatial representations lead to more reliable spatial reasoning across diverse benchmarks. Code and benchmark are available on the project page: this https URL.
[CV-35] AnomalyAgent : Training-Free Agent ic Models for Zero-/Few-Shot Anomaly Detection
链接: https://arxiv.org/abs/2605.30140
作者: Yi Zhang,Jiawen Zhu,Lele Fu,Guansong Pang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Benefiting from generalizability of vision-language models (VLMs) such as CLIP, many zero-/few-shot anomaly detection (AD) approaches have achieved impressive detection performance across various datasets. Nevertheless, they require substantial training on large auxiliary datasets to adapt VLMs to anomaly detection, and their inference largely relies on visual-text embedding similarity-based anomaly scores, lacking reasoning abilities to detect complex anomalies that require in-depth contextual understanding. To address this limitation, we propose \textbfAnomalyAgent, a novel training-free, agentic framework that leverages the advanced reasoning and generalization capabilities of multimodal large language models (MLLMs) for anomaly detection. The key ingredients include \textbf1) a comprehensive anomaly-centric toolset that enables adaptive MLLM-driven, agentic anomaly reasoning in zero-shot settings, and \textbf2) a customized memory module that grounds anomaly reasoning with few-shot, in-context reference examples. We extend evaluation beyond the detection of simple anomalies (e.g., surface defects like cracks and dents and clear lesions) in widely used benchmarks to more diverse types of anomalies such as logical/contextual anomalies in logistics and manufacturing settings. Extensive experiment results demonstrate that our AnomalyAgent achieves substantially better performance compared to training-free VLM-based AD and generic agentic methods, highlighting its superior generalization capability in both zero-shot and few-shot anomaly detection settings. The code implementation can be find at this address.
[CV-36] SGMD: Score Gradient Matching Distillation for Few-Step Video Diffusion Distillation ICML2026
链接: https://arxiv.org/abs/2605.30116
作者: Zhuguanyu Wu,Ruihao Gong,Yang Yong,Yushi Huang,Xiangyu Fan,Lei Yang,Dahua Lin,Xianglong Liu
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: ICML 2026
Abstract:Distribution Matching Distillation (DMD) is a widely used paradigm for accelerating inference in few-step video diffusion models. However, DMD-style video distillation faces two coupled challenges: the fake score must track a continuously evolving generator, making training costly when frequent updates are required, while reverse-KL-style matching can be mode-seeking and conservative for preserving strong motion dynamics. To address these issues, we propose \textbfScore Gradient Matching Distillation (SGMD). SGMD adopts a fake-score perspective by directly optimizing the fake score toward the teacher, while using teacher stop-gradient Fisher as a stable distribution-matching objective. We provide a gradient analysis that motivates this objective choice under ideal tracking. Building on this, SGMD introduces a pair of dual potentials: negative-residual (NR) for outer-loop correction and residual-contraction (RC) for inner-loop tracking. Empirically, compared to DMD2, SGMD achieves an approximately \sim 3\times training speedup and substantially improves motion dynamics for 4-step distilled models while preserving temporal consistency. A human study confirms that SGMD is preferred in motion quality and overall preference, while visual quality and text alignment remain comparable. Code is available at this https URL.
[CV-37] Large Depth Completion Model from Sparse Observations ICLR2026
链接: https://arxiv.org/abs/2605.30115
作者: Zhu Yu,Zhengyi Zhao,Runmin Zhang,Lingteng Qiu,Kejie Qiu,Yisheng He,Siyu Zhu,Zilong Dong,Si-Yuan Cao,Hui-Liang Shen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ICLR 2026. Project webpage: this https URL
Abstract:This work presents the Large Depth Completion Model (LDCM), a simple, effective, and robust framework for single-view metric depth estimation with sparse observations. Without relying on complex architectural designs, LDCM generates metric-accurate dense depth maps using a transformer. It outperforms existing approaches across diverse datasets and sparse observations. We achieve this from two key perspectives: (1) leveraging existing monocular foundation models to improve the quality of sparse depth inputs, and (2) reformulating training objectives to better capture geometric structure and metric consistency. Specifically, a Poisson-based depth initialization strategy is first introduced to generate a uniform coarse dense depth map from diverse sparse observations, providing a strong structural prior for the network. Regarding the training objective, we replace the conventional depth head with a point map head that regresses per-pixel 3D coordinates in camera space, enabling the model to directly learn the underlying 3D scene structure instead of performing pixel-wise depth map restoration. Moreover, this design eliminates the need for camera intrinsic parameters, allowing LDCM to naturally produce metric-scaled 3D point maps. Extensive experiments demonstrate that LDCM consistently outperforms state-of-the-art methods across multiple benchmarks and varying sparsity levels in both depth completion and point map estimation, showcasing its effectiveness and strong generalization to unseen data distributions.
[CV-38] xModel-KD: Cross-modal Knowledge Distillation for 3D Scene Perception using LiDAR
链接: https://arxiv.org/abs/2605.30111
作者: Thenukan Pathmanathan,Kanchan Keisham,Thangarajah Akilan
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 3 figures, and 5 tables
Abstract:Point cloud segmentation is a fundamental task in 3D scene understanding. Its progress is constrained by the high cost and time required for dense 3D annotations, making labeled samples difficult to obtain. Beyond annotation scarcity, different sensing modalities face inherent limitations. 2D images provide rich texture and appearance cues, yet they lack explicit depth and geometric structure. In contrast, 3D point clouds capture accurate spatial geometry but are sparse and contain no texture information. As a result, relying on a single modality restricts the richness of learned representations and weakens generalization. Although recent multi-modal methods that combine 3D point clouds with 2D images have demonstrated strong performance in tasks such as classification and retrieval, they typically depend on large-scale labeled datasets and have not been fully exploited for data-efficient dense prediction. To address these limitations, we propose a novel cross-modal knowledge distillation framework, xModel-KD, for 3D point cloud segmentation. Our method exploits the complementary strengths of 2D texture and 3D geometry by learning unified per-point representations through cross-modal alignment. Specifically, we design a cross-modal fusion encoder trained with a contrastive objective that enforces feature consistency between corresponding 2D and 3D representations across multiple views. By integrating powerful pre-trained backbones with a targeted fusion strategy, the proposed framework effectively transfers appearance cues from images to geometry-aware point features. Experimental results show that cross-modal fusion achieves a 2% absolute improvement in mIoU over a LiDAR-only baseline, demonstrating the benefit of leveraging complementary multi-modal information for scalable and annotation-efficient 3D scene understanding.
[CV-39] Evaluation of Conversational Agents : Understanding Culture Context and Environment in Emotion Detection
链接: https://arxiv.org/abs/2605.30099
作者: Martha Teiko Teye,Yaw Marfo Missah,Emmanuel Ahene,Twum Frimpong,Auxane Boch
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: IEEE paper on arxiv
Abstract:Valuable decisions and highly prioritized analysis now depend on applications such as facial biometrics, social media photo tagging, and human robots interactions. However, the ability to successfully deploy such applications is based on their efficiencies on tested use cases taking into consideration possible edge cases. Over the years, lots of generalized solutions have been implemented to mimic human emotions including sarcasm. However, factors such as geographical location or cultural difference have not been explored fully amidst its relevance in resolving ethical issues and improving conversational AI (Artificial Intelligence). In this paper, we seek to address the potential challenges in the usage of conversational AI within Black African society. We develop an emotion prediction model with accuracies ranging between 85% and 96%. Our model combines both speech and image data to detect the seven basic emotions with a focus on also identifying sarcasm. It uses 3-layers of the Convolutional Neural Network in addition to a new Audio-Frame Mean Expression (AFME) algorithm and focuses on model pre-processing and post-processing stages. In the end, our proposed solution contributes to maintaining the credibility of an emotion recognition system in conversational AIs.
[CV-40] Geometry Matters: 3D Foundation Priors for Learning Semantic Correspondence
链接: https://arxiv.org/abs/2605.30093
作者: Artur Jesslen,Olaf Dünkel,Adam Kortylewski
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages (main paper), 21 pages (total), 4 figures
Abstract:Foundation features from self-supervised vision models and text-to-image diffusion models have proven effective for semantic correspondence estimation. However, because these features are learned primarily from 2D image objectives, they lack explicit 3D awareness and often confuse symmetric object sides, repeated parts, and visually similar structures that are distinct in 3D. We introduce a 3D-aware post-training framework that goes beyond available 2D foundation features by incorporating priors from 3D foundation models. Given an image, our method uses SAM3D to estimate object geometry and pose, and refines the pose through render-and-compare optimization. Subsequently, we render PartField descriptors from the reconstructed geometry into the image plane based on the estimated object pose. The resulting geometry-aware feature maps complement DINO and Stable Diffusion features, while geodesic distances on the reconstructed shapes enable reliable filtering of candidate correspondences. We use the filtered matches as supervision to train a lightweight adapter on top of DINO and Stable Diffusion for semantic correspondence. In contrast to prior post-training approaches that require pose annotations and rely on coarse spherical geometry, our method automatically obtains instance-specific 3D structure and uses it to guide correspondence learning. Experiments show that our approach improves semantic correspondence over the prior methods while reducing manual geometric supervision. Code and model can be found at https:/github.com/GenIntel/3D-SC.
[CV-41] Future Forcing: Future-aware Training-free KV Cache Policy for Autoregressive Video Generation
链接: https://arxiv.org/abs/2605.30083
作者: Jiayi Luo,Qiyan Liu,Tengyang Wang,JunHao Liu,Jiayu Chen,Cong Wang,Hanxin Zhu,Chen Gao,Xiaobin Hu,Qingyun Sun,Zhibo Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Autoregressive (AR) video generation has emerged as a promising paradigm for long-horizon video synthesis, where each frame is generated conditioned on previously generated tokens. To accelerate inference, the KV cache is used to avoid redundant recomputation across generation steps. Nevertheless, its growth with generation length introduces increasing memory and error accumulation, limiting the scalability of AR models to even longer sequences. Existing KV cache compression methods mitigate this issue by selectively retaining only video tokens deemed important. However, most existing methods assess token importance using short-horizon signals derived from the current or historical generation context, making these methods prone to overlooking tokens that appear unimportant at early steps but later become critical for future frames. In this work, we identify an important property of trained AR video models: although RoPE-modulated queries evolve across autoregressive steps, the underlying canonical pre-RoPE query distribution remains remarkably stable throughout the video generation process. This approximate stationarity implies that future query distributions are estimable from historical statistics, enabling principled future-aware cache decisions without any additional training. Building on this insight, we propose Future Forcing, a training-free future-aware KV cache policy for AR video generation. Specifically, Future Forcing first constructs a future query proxy from historical statistics, then scores KV cache tokens by their importance under this proxy, and finally merges redundant token pairs within the affine subspace induced by the future query. Extensive experiments show that Future Forcing improves long-horizon consistency under limited KV caches, achieving up to 1.49 improvement in subject consistency on VBench-Long for 60s generation over existing AR video KV cache policies.
[CV-42] Native Audio-Visual Alignment for Generation
链接: https://arxiv.org/abs/2605.30073
作者: Longbin Ji,Guan Wang,Xuan Wei,Chenye Yang,Xiangrui Liu,Zhenyu Zhang,Shuohuan Wang,Yu Sun,Jingzhou He
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Joint audio-video generation aims to synthesize temporally synchronized and semantically coherent visual-acoustic content. However, existing open-source methods mainly rely on either dual-tower designs with posterior alignment or fully unified tri-modal designs that mix textual context, audio and video in one shared space. The former weakens fine-grained audio-video co-evolution, while the latter couples semantic conditioning with low-level synchronization. To address these limitations, we propose NAVA, a Native Audio-Visual Alignment framework for joint audio-video generation. NAVA is built upon context-conditioned native audio-visual alignment: it first establishes audio-video correspondence in a dedicated interaction space, and then uses external context to condition the joint denoising process. Specifically, NAVA is instantiated with an Align-then-Fuse MMDiT architecture, which transitions from modality-aware audio-video alignment to modality-shared joint denoising. Furthermore, we introduce Timbre-in-Context Conditioning to associate reference timbre cues with corresponding speech spans to achieve controllable speech timbre. Experiments on Verse-Bench and Seed-TTS, together with a user study, demonstrate that NAVA achieves superior video quality, precise audio-visual synchronization, competitive audio quality, and stronger reference-timbre controllability using only 6.3B parameters.
[CV-43] Boosting Zero-Shot 3D Style Transfer with 2D Pre-trained Priors
链接: https://arxiv.org/abs/2605.30065
作者: Xin Dong,Yunzhi Teng,Wenfeng Deng,Yansong Tang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by IEEE IVMSP2026
Abstract:In this work, we focus on zero-shot 3D style transfer that can generate multi-view consistent stylized views of the 3D scene given an arbitrary style image. We primarily tackle the issue of data scarcity in 3D style transfer, which arises when each model is trained on only a single scene, thereby limiting the number of available content images. This scarcity significantly hampers stylization performance, as model optimization relies on a sufficient number of content-style image pairs to provide supervisory signals. Our core idea is to integrate a decoder pre-trained on large-scale 2D image datasets into the 3D style transfer pipeline, thereby leveraging the prior knowledge encoded in the decoder from learning over numerous content-style image pairs. Our method combines feature Gaussian splatting and deferred stylization, enabling high-quality stylization with the data-sufficient decoder network while ensuring view consistency by unifying view-dependent operations into a view-invariant process. Experiments demonstrate that our Data-Sufficient StyleGaussian (DS-StyleGaussian) model outperforms existing zero-shot 3D style transfer methods in terms of visual quality across various datasets. This work also suggests that 2D pre-training can serve as a strong enhancement for 3D tasks, bridging the data gap between 2D and 3D.
[CV-44] FakeVLM-R1: Internalizing Physical Laws via CoT for Synthetic Image Detection
链接: https://arxiv.org/abs/2605.30062
作者: Leqi Zhu,Junyan Ye,Kaiqing Lin,Zhiyuan Yan,Conghui He,Weijia Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The development of generative artificial intelligence technologies has propelled the visual realism of synthetic images to an unprecedented level. Although current interpretable detection methods based on Large Multimodal Models (LMMs) have made certain progress, they still rely on imitation learning derived from massive volumes of forged data. Consequently, they lack genuine causal reasoning capabilities and are prone to explanatory hallucinations. To overcome this bottleneck, we propose FakeVLM-R1, aiming to endow the model with human-like critical thinking capabilities when performing synthetic detection tasks. Building upon Supervised Fine-Tuning (SFT), this framework integrates Group Relative Policy Optimization (GRPO) with a Critical Thinking Chain-of-Thought (CoT) mechanism. During the inference phase, the model executes a “bidirectional dialectical reasoning” process: while proposing a forgery hypothesis, it must simultaneously invoke physical commonsense to construct an authenticity counter-proof. Furthermore, we constructed the FakeClue++ dataset with high-quality samples, which extensively introduces annotations guided by the physical laws of authentic images, providing a unified authenticity anchor for the model. Experiments confirm that FakeVLM-R1 achieves SOTA performance the evaluated models across multiple benchmarks. It not only achieves high-precision, logically interpretable detection but also resolves the over-rejection bias of existing methods against real images, demonstrating generalization and robustness against perturbations.
[CV-45] owards Consistent Video Geometry Estimation
链接: https://arxiv.org/abs/2605.30060
作者: Zhu Yu,Jingnan Gao,Runmin Zhang,Lingteng Qiu,Zhengyi Zhao,Rui Peng,Yichao Yan,Kejie Qiu,Siyu Zhu,Si-Yuan Cao,Hui-Liang Shen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project webpage: this https URL
Abstract:This work presents ViGeo, a feed-forward foundation model for recovering spatially dense and temporally consistent geometry from video sequences. Built upon a plain transformer architecture without task-specific architectural modifications, ViGeo supports streaming, full-sequence, and long-video inference within a unified model. The key design is dynamic chunking attention, which exposes the model to both bidirectional and causal temporal contexts during training and allows it to adapt its attention pattern at test time without retraining. To improve supervision quality, we further introduce a completion-based data refinement framework. This framework trains a video depth completion teacher that conditions on sparse and noisy annotations and exploits video/multi-view context to produce dense, temporally coherent, and geometrically reliable training targets. Beyond depth and point maps, ViGeo also predicts surface normals within the same framework. Trained solely on public datasets, ViGeo achieves state-of-the-art performance across online, offline, and long-video depth estimation, surface normal estimation, and video point map estimation.
[CV-46] GenEraser: Generalizable Video Object Removal via Balanced Text-Mask Guidance and Decoupled Locator-Preserver
链接: https://arxiv.org/abs/2605.30045
作者: Yuqing Chen,Lin Liu,Haisu Wu,Xiaopeng Zhang,Yaowei Wang,Yujiu Yang,Qi Tian
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Video object removal frequently struggles to simultaneously eliminate target objects and their associated physical effects (e.g., smoke, reflections, light, and ripples) in out-of-domain scenarios due to complex spatiotemporal ambiguities. While existing methods primarily rely on spatial masks, they often fail to capture weakly correlated effects, and the potential of explicit textual guidance remains underexplored. Furthermore, a fundamental optimization conflict exists in removal models between high-level semantic generalization and precise pixel-level background preservation. To address these challenges, we propose GenEraser, a novel framework for generalized and high-fidelity video object and effect removal. First, we introduce a Multi-Conditional Mixture-of-Experts (MC-MoE) paired with Bipartite Text guidance to fully exploit the multimodal priors of Diffusion Transformers, significantly enhancing the identification of complex effects. Second, a Learnable Deep ``CFG’’ Fusion mechanism (LD-CFG) is developed to adaptively balance the relative dominance of mask and textual conditions across diverse scenarios. Finally, we propose a Decoupled Expert Architecture, comprising a Locator and a Preserver, to mitigate the inherent trade-off between semantic generalization and pixel alignment. Extensive experiments demonstrate that our GenEraser surpasses recent state-of-the-art approaches, achieving significant quantitative improvements (e.g., 2.16 dB and 1.44 dB on the ROSE Benchmark and VOR-Eval, respectively) while maintaining exceptionally robust generalization in open-world scenarios. this https URL
[CV-47] Alignment-Guided Score Matching for Text-to-Image Alignment in Diffusion Models ICML2026
链接: https://arxiv.org/abs/2605.30038
作者: Jaa-Yeon Lee,Yeobin Hong,Taesung Kwon,Jong Chul Ye
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: ICML 2026, Project page: this https URL
Abstract:Diffusion models generate highly realistic images but often struggle with precise text-image alignment. While recent post-training methods improve alignment using external rewards or human preference signals, their performance heavily depends on reward quality and does not directly address alignment within the diffusion process itself. Recent reward-free approaches such as SoftREPA demonstrate that optimizing soft text tokens via contrastive learning can effectively improve text-image representation alignment, outperforming standard parameter-efficient fine-tuning baselines. However, the contrastive formulation can excessively penalize negative pairs, which manifests as characteristic failure cases such as over-counting and repetition. To address this issue, we propose a lightweight, reward-free post-training method that refines soft tokens by integrating contrastive alignment guidance directly into the score-matching objective of diffusion models. By assigning alignment directions at the score level, our approach mitigates these limitations and yields more coherent and semantically faithful generations. Experiments show that our method matches SoftREPA while substantially improving its failure cases, achieving over 35% improvement in counting accuracy on the GenEval benchmark. Our method is seamlessly applicable to existing diffusion backbones (SD1.5, SDXL, and SD3), and is complementary to existing RL-based diffusion post-training methods. Project page: this https URL
[CV-48] VisualThink-VLA: Visual Intermediate Reasoning for Effective and Low-Latency Vision-Language-Action Policies
链接: https://arxiv.org/abs/2605.30011
作者: Mingjian Gao,Wenqiao Zhang,Yuqian Yuan,Yang Dai,Binhe Yu,Zheqi Lv,Haoyu Zheng,Jiaqi Zhu,Zhiqi Ge,Zixuan Wan,Siliang Tang,Yueting Zhuang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Recent work has begun to equip vision-language-action (VLA) policies with explicit intermediate reasoning. In embodied control, however, textual chain-of-thought is a poor fit: irrelevant or weakly textual information can interfere with action prediction, while autoregressive text decoding adds too much latency for real-time closed-loop execution. We present VISUALTHINK-VLA, a visual intermediate-reasoning framework for accurate, low-latency VLA policies. Our bootstrapping philosophy is to guide action with effective visual thinking: VISUALTHINK-VLA bootstraps action prediction through a compact visual-evidence interface that preserves spatial precision while avoiding decoding overhead. Besides, to further improve performance and efficiency, VISUALTHINK-VLA adopts a tailored selective routing mechanism to learn the visual evidence tokens, enabling low-latency inference while preserving high-capacity specialization. We also introduce VisualEvidence-Kit, a supervision-and-audit resource centered on a VisualEvidence-Agent that constructs a 754.7k VLA instructions VisualEvidence-Set for route supervision and counterfactual faithfulness tests. Across multiple benchmarks and real-robot evaluation, VISUALTHINK-VLA achieves the highest success rate on most benchmarks while reducing the multi-second latency of reasoning-augmented baselines to the sub-second regime. For example, on BridgeData V2, it reduces step latency from 8.377,s with ECoT to 0.367,s, achieving a 22.8 times speedup.
[CV-49] EarlyTom: Early Token Compression Completes Fast Video Understanding CVPR2026
链接: https://arxiv.org/abs/2605.30010
作者: Hesong Wang,Xin Jin,Lu Lu,Chenhaowen Li,Jian Chen,Qiang Liu,Huan Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by CVPR 2026. 16 pages, 8 figures, 8 tables. Project page: this https URL
Abstract:Video large language models (Video-LLMs) have demonstrated strong capabilities in video understanding tasks. However, their practical deployment is still hindered by the inefficiency introduced by processing massive amounts of visual tokens. Although recent approaches achieve extremely low token retention ratios while maintaining accuracy comparable to full-token baselines, most of them perform compression only at the late stage of prefilling, leaving the efficiency of the vision encoder unoptimized. In this paper, we first show that vision encoding contributes a large portion to the time-to-first-token (TTFT). Therefore, instead of compressing visual tokens only after the vision encoder, performing compression inside the encoder still leaves substantial room for exploration. Based on this insight, we propose EarlyTom, a training-free token compression framework that performs early-stage visual token compression inside the vision encoder, enabling significantly better TTFT reduction and higher throughput. In addition, we introduce a decoupled spatial token selection strategy that improves the overall compression effectiveness. EarlyTom reduces TTFT by up to 2.65x and FLOPs by up to 61% on a single NVIDIA A100 GPU for the LLaVA-OneVision-7B model, while maintaining accuracy comparable to the full-token baseline. These improvements substantially enhance the practicality of deploying Video-LLMs in real-world production scenarios.
[CV-50] FRUC: Feedforward Dynamic Scene Reconstruction from Uncalibrated Collaborative Driving Views
链接: https://arxiv.org/abs/2605.29997
作者: Yihang Tao,Yu Guo,Zhengru Fang,Haonan An,Yuguang Fang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We present FRUC, a feed-forward 3D Gaussian splatting framework for dynamic scene reconstruction from uncalibrated collaborative driving views. Existing multi-agent reconstruction frameworks are often hindered by rigid prerequisites, demanding precise spatial calibration and slow per-scene optimization. In this paper, we rethink this task by conceptualizing a distributed multi-vehicle network as a spatio-temporally unstructured ego-centric multi-camera system, where the core challenge lies in enhancing ego-centric occluded geometry through collaboration without degrading the ego’s accurately observed visible geometry, while preserving reconstruction efficiency. For efficient reconstruction, FRUC is built upon a visual grounded geometric Transformer backbone to enable one-shot, calibration-free inference from a flexible number of multi-vehicle views. To achieve non-destructive geometric supplementation under uncalibrated cross-agent misalignment, FRUC first introduces an ego-centric causal occlusion field that explicitly derives occlusion evolution as latent priors by modeling agent-wise spatio-temporal correlations. Guided by these occlusion priors, it further formulates cross-agent integration as a deterministic residual denoising process via zero-initialized injection, turning challenging cross-agent fusion into bounded residual learning for robust collaborative blind-spot completion. Through extensive evaluations on the real-world V2XReal and UrbanIng-V2X datasets, FRUC is shown to be a new state-of-the-art for the scene reconstruction of dynamic collaborative driving environments, significantly outperforming existing methods in both rendering quality and efficiency.
[CV-51] Improving Adversarial Robustness of Attribution via Implicit Regularization
链接: https://arxiv.org/abs/2605.29983
作者: Amir Mehrpanah,Matteo Gamba,Hossein Azizpour
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 39 pages, 22 figures, to be published in International Conference on Machine Learning 2026
Abstract:The adversarial robustness of attributions is a fundamental requirement for reliable explainability in deep learning, yet existing approaches typically rely on computationally expensive explicit regularization. In this work, we show that attribution robustness can arise implicitly from the learning dynamics of standard stochastic gradient descent. We theoretically motivate this effect through connections between parameter-space and input-space curvature, and validate it across architectures, datasets, and attribution methods, with negligible computational overhead. In contrast, we prove that such robustness gains often does not transfer to attention-based attribution under softmax normalization, due to inherent entropy constraints, and we validate this limitation experimentally. Finally, we show that replacing softmax attention with kernel-based attention restores the robustness gains in transformer models. Our results highlight learning dynamics as a principled and practical mechanism for robust explainability, and reveal fundamental limitations of attention-based attribution under normalization.
[CV-52] Genetically Aligned Patient Representations Improve Hematological Diagnosis MICCAI2026
链接: https://arxiv.org/abs/2605.29980
作者: Muhammed Furkan Dasdelen,Fatih Ozlugedik,Ilaria Looser,Rao Muhammad Umer,Christian Pohlkamp,Carsten Marr
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted for publication at the 29th International Conference on Medical Image Computing and Computer Assisted Intervention - MICCAI 2026
Abstract:Multimodal alignment of histopathology encoders with transcriptomic and genomic data has been shown to significantly improve performance in downstream diagnostic tasks. Hematological cytology is unique in that visual single-cell evaluation is often paired with cytogenetics and molecular genetics for blood cancer diagnosis. In this study, we present a framework to align single white blood cell images with chromosomal aberrations (karyotype) and somatic mutations from targeted gene panels. Our training strategy follows a two-stage approach: (i) self-supervised, vision-only pretraining of a transformer aggregator using an iBOT head on a cohort of over 1500 patients, and (ii) genetic alignment via supervised contrastive loss on acute myeloid leukemia patients. Our genetically aligned patient encoder improves hematological diagnostic tasks, outperforming slide-level histopathology foundation models. Additionally, the model provides off-the-shelf retrieval capabilities for diseases and genetic alterations. Incorporating genetic data into patient encoders increases the quality of patient representations, providing a framework that aligns with clinical diagnostic workflows and paves the way for future multimodal hematology-specific AI. The code and model weights are available at this https URL.
[CV-53] EVL-ECG: Efficient ECG Interpretation With Multi-Aspect Heterogeneous Knowledge Distillation ICML2026
链接: https://arxiv.org/abs/2605.29977
作者: Dang Hong Nguyen,Nhi Ngoc-Yen Nguyen,Huy-Hieu Pham
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: Accepted at the SD4H Workshop at ICML 2026. 11 pages, 3 figures
Abstract:High-fidelity ECG interpretation is increasingly reliant on massive foundation models, yet their deployment in clinical edge-care remains hindered by extreme computational demands. While knowledge distillation (KD) is a promising solution, traditional methods fail to capture the complex spatio-temporal dependencies of ECG signals when transferring knowledge across heterogeneous architectures. In this paper, we propose EVL-ECG, a framework specifically designed for cross-architecture distillation of cardiac diagnostic logic. EVL-ECG introduces three ECG-aware innovations: (1) Multi-Head Cross-Attention Alignment, which harmonizes architectural discrepancies to preserve fine-grained morphological features; (2) Optimal Transport-based Visual Feature Matching, utilizing optimal transport to maintain global structural relationships across ECG leads despite mismatched token representations; and (3) Geometric Intra-Architecture Relation Matching, which distills the latent diagnostic reasoning of the teacher model. Evaluations across ECG benchmarks demonstrate that EVL-ECG yields improvements of up to 2.4% AUC and 1.1% clinical accuracy over existing baselines. Notably, EVL-ECG establishes an efficient 2B-parameter ECG foundation model, suitable for resource-constrained clinical environments.
[CV-54] SwInception – Local Attention Meets Convolutions
链接: https://arxiv.org/abs/2605.29954
作者: David Hagerman,Roman Naeem,Jakob Lindqvist,Carl Lindström,Fredrik Kahl,Lennart Svensson
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: International Conference on Pattern Recognition and Artificial Intelligence, 2024
Abstract:Sparse vision transformers have gained popularity as efficient encoders for medical volumetric segmentation, with Swin emerging as a prominent choice. Swin uses local attention to reduce complexity and yields excellent performance for many tasks but still tends to overfit on small datasets. To mitigate this weakness, we propose a novel architecture that further enhances Swin’s inductive bias by introducing Inception blocks in the feed-forward layers. The introduction of these multi-branch convolutions enables more direct reasoning over local, multi-scale features within the transformer block. We have also modified the decoder layers in order to capture finer details using fewer parameters. We demonstrate a performance improvement on eleven different medical datasets through extensive experimentation. We specifically showcase advancements over the previous state-of-the-art backbones on benchmark challenges like the Medical Segmentation Decathlon and Beyond the Cranial Vault. By showing that the existing inductive bias in Swin can be further improved, our work presents a promising avenue for enhancing the capabilities of sparse vision transformers for both medical and natural image segmentation tasks. Code and pre-trained weights can be accessed at this https URL.
[CV-55] Mesh-Aware Epipolar Matching for Multi-View Multi-Person 3D Pose Estimation in Basketball
链接: https://arxiv.org/abs/2605.29953
作者: Li Yin,Qin Haobin,Tomohiro Suzuki,Calvin Yeung,Mariko Isogawa,Keisuke Fujii
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multi-view multi-person 3D pose estimation in team sports scenarios remains challenging due to player occlusions, appearance similarity caused by team uniforms, and the scarcity of annotated multi-view data, all of which limit the effectiveness and generalization capability of learning-based methods. In contrast, the performance of training-free approaches is inherently constrained by the accuracy of 2D keypoint detection and the robustness of cross-view association. To address these challenges, we propose Mesh-Aware Epipolar Matching (MAEM), a training-free framework for multi-view multi-person 3D pose estimation. Our method employs a monocular 3D human mesh recovery model as the frontend and introduces a two-stage epipolar matching strategy based on the recovered mesh outputs. Specifically, the proposed framework combines disjoint-set-union-based clustering with per-joint triangulation to achieve robust cross-view association and accurate 3D pose reconstruction. Experiments on two public multi-view basketball datasets demonstrate that MAEM consistently outperforms existing training-free association baselines while achieving competitive RGB-only performance in both indoor and outdoor basketball scenarios. MAEM achieves MPJPE/PA-MPJPE scores of 59.8/40.7 mm on SportCenter EPFL and 74.0/51.8 mm on Human-M3 Basketball, highlighting the effectiveness of dense mesh geometry for cross-view association without requiring target-domain training or fine-tuning.
[CV-56] CityGen: Structure-Guided City-Style Synthesis for Cross-City Autonomous Driving
链接: https://arxiv.org/abs/2605.29935
作者: Zezhong Qian,Zhao Yang,Lu Tan,Zhihao Yan,Weiyi Hong,Haizhuang Liu,Yawei Jueluo
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Autonomous driving systems are commonly trained and evaluated within limited geographic regions, which hinders their scalability when deployed in new cities. However, significant domain shifts in appearance, road topology, and traffic patterns often cause severe performance degradation under cross-city deployment. Existing approaches based on domain adaptation, data augmentation, or synthetic data generation typically rely on labeled target data, city-specific annotations, or task-specific designs, limiting their scalability and effectiveness for holistic evaluation. In this paper, we introduce CityTransfer-Bench, a geographically disjoint benchmark for evaluating cross-city generalization across perception, segmentation, and planning, and propose CityGen, a diffusion-based generative framework that performs zero-label city adaptation via HD-map-conditioned synthesis guided by city-level visual prompts. Extensive experiments demonstrate that CityGen consistently improves cross-city robustness across multiple tasks, establishing a scalable and label-efficient foundation for generalizable autonomous driving.
[CV-57] reatment-Conditioned Diffusion for Forecasting Neurodegenerative Disease Progression
链接: https://arxiv.org/abs/2605.29932
作者: Danylo Boiko,Viktoriia Mishkurova
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages, 5 figures, 1 table
Abstract:Forecasting the progression of neurodegenerative diseases, such as Parkinson’s disease, is essential for effective long-term planning and personalized therapeutic intervention. Existing systems typically produce scalar clinical scores that ignore the rich structure of longitudinal neuroimaging, while traditional generative approaches suffer from a loss of anatomical details and blurring subtle progression patterns. To address this, we introduce a novel treatment-conditioned diffusion framework that predicts high-fidelity future brain states by conditioning the generative process on patients’ screening DaTscan images and levodopa equivalent daily dose over one year. The pipeline uses a Transformer-based encoder to represent non-linear, time-dependent pharmacological dynamics and optimizes generation through a multi-weight region-of-interest mask that focuses on biologically critical areas. Experimental evaluation shows that our framework maintains sharp anatomical boundaries and significantly improves clinical fidelity relative to the baseline, achieving 14.0% lower MSE, 7.2% lower MAE, and 4.9% higher SSIM.
[CV-58] Reducing Experimental Testing in Space Propulsion Film Cooling Analyses by Pixelwise Generative Image Interpolation
链接: https://arxiv.org/abs/2605.29911
作者: Adam T. Müller,Philipp J. Teuffel,Konstantin Manassis,Nicolaj C. Stache
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: Presented at the 11th European Conference for Aeronautics and Aerospace Sciences (EUCASS), 2025, DOI: https://doi.org/10.13009/EUCASS2025-285
Abstract:We propose a machine learning approach for image regression from sparse experimental measurements. We show the application of the proposed method on film cooling studies in propulsion system development, aiming to reduce the need for extensive physical testing. Our method employs a lightweight feed-forward neural network with positional encoding to generate images conditioned by input parameters. Validated on real and synthetic data, it achieves high image similarity (RMSE 8 %, SSIM 93 %) while maintaining accuracy with a 30 % reduction of measurements. We further propose a knowledge-informed extension for local adaptability of the generated images. This approach significantly reduces required tests while preserving high-quality data, enabling efficient optimization of coolant injector configurations with applications beyond aerospace.
[CV-59] rain the Agent Not the Expert: Learning to Harness Heterogeneous Experts for Multi-Turn Visual Reasoning
链接: https://arxiv.org/abs/2605.29894
作者: Yaowu Fan,Tao Han,Dazhao Du,Andy J. Ma,Jia Wan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent progress in computer vision has produced a wide range of powerful specialized models for detection, segmentation, counting, and other visual tasks. However, these models are usually optimized for isolated task formulations, making it difficult to directly support general-purpose visual intelligence, especially when a task requires complex language understanding and dense small-object perception. In this paper, we propose VisHarness, a trainable visual agent that decouples high-level perception, reasoning, and decision-making from low-level task execution. Instead of training a model to solve a specific visual task, VisHarness learns to harness a set of carefully designed heterogeneous visual experts. This paradigm preserves the general intelligence of the agent while fully leveraging the precision advantages of specialized visual models in concrete visual tasks. With only lightweight training, VisHarness learns a generalizable visual expert-harnessing policy and can solve common fundamental vision tasks under various complex conditions through multi-turn interactions with visual expert models. To enable efficient on-policy reinforcement learning training in a live environment, we introduce dynamic visual memory archiving, which mitigates the rapidly accumulating visual-token overhead caused by multi-turn interactions with visual expert models. Experiments on four representative benchmarks covering reasoning segmentation, generalized referring segmentation, dense small-object detection, and referring counting demonstrate that VisHarness substantially outperforms existing general-purpose models and achieves competitive or superior performance compared with task-specific models.
[CV-60] DVSM: Decoder-only View Synthesis Model Done Right
链接: https://arxiv.org/abs/2605.29891
作者: Cheng Sun,Jaesung Choe,Min-Hung Chen,Ryo Hachiuma,Yu-Chiang Frank Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Code at this https URL
Abstract:Recent Large View Synthesis Models (LVSMs) advocate an encoder-decoder architecture that separates reconstruction and rendering into distinct networks. We re-examine this design. Through controlled experiments, we show that a decoder-only architecture, which represents scenes implicitly as a KV-cache, outperforms encoder-decoder variants while using fewer parameters at identical rendering complexity. Further analysis shows that sharing weights between the color-input reconstruction network and the camera-only rendering network better aligns their features at the same viewpoint, facilitating image synthesis. Building on this finding, our model, dubbed DVSM, further incorporates foundation model priors and stage-wise patch sizing for an improved efficiency-quality tradeoff. Our results establish a new state of the art for novel-view synthesis across multiple benchmarks, in some cases even outperforming per-scene-optimized 3DGS under dense input views.
[CV-61] Mitigating Hallucination in Vision-Language Models through Barrier-Regulated Adaptive Closed-form Steering
链接: https://arxiv.org/abs/2605.29881
作者: Soumyadeep Jana,Pulkit Mittal,Sanasam Ranbir Singh
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Large vision-language models (LVLMs) often hallucinate objects that are not present in the input image, largely because visual grounding weakens as decoding progresses. Existing inference-time mitigation methods modify logits or hidden states throughout generation, but they suffer from three key limitations: they lack an explicit grounding objective, intervene even when the model is already well-grounded, and use fixed correction strengths that do not adapt to the severity of grounding failure. We propose BRACS (Barrier-Regulated Adaptive Closed-form Steering), a training-free steering framework that addresses these issues through barrier-regulated adaptive closed-form steering. BRACS monitors the model’s own attention to measure visual grounding and applies corrections to the hidden states only when grounding deteriorates. The corrective update is computed analytically in closed form, requiring no training of auxiliary networks or model retraining. Experiments on LLaVA-1.5-7B and Qwen-VL-Chat show that BRACS consistently outperforms prior methods on hallucination benchmarks, reducing CHAIR _s by 9.4 points and improving POPE F1 by 2.7 points, while matching or improving performance on four general multimodal benchmarks. BRACS also remains efficient, operating at 80% of greedy decoding throughput and achieving 1.3 times higher speed on average than the baselines.
[CV-62] DGSG-Mind: Dynamic 3D Gaussian Scene Graphs for Long-Term Scene Understanding and Grounding
链接: https://arxiv.org/abs/2605.29879
作者: Luzhou Ge,Xiangyu Zhu,Jinyan Liu,Xuesong Li
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 9 pages, 6 figures
Abstract:Integrating open-vocabulary semantic information into dynamic 3D scene representations is essential for long-term embodied scene understanding. However, existing methods often suffer from fragile instance association due to incomplete cross-view cues, while their limited ability to handle object-level topological changes restricts long-term robotic task execution. Moreover, current 3D scene understanding methods either rely on simple feature matching without explicit spatial reasoning or assume offline ground-truth 3D geometry. To address these challenges, we present DGSG-Mind, a hybrid instance-aware 3D Gaussian dynamic scene graph system with an embodied reasoning agent. Our system couples a probabilistic voxel grid with explicit 3D Gaussians to enable robust cross-modal instance fusion and incremental semantic mapping. It handles dynamic changes through Gaussian-based visual relocalization and localized masked refinement guided by geometric-semantic consistency. Built on the instance Gaussian map, DGSG-Mind further constructs a hierarchical scene graph and develops the 3D Gaussian Mind, which integrates structural relations, spatial-semantic information, and visually annotated RoI Gaussian renderings for multimodal reasoning. Extensive experiments show that DGSG-Mind achieves the best zero-shot 3DVG performance among methods operating on self-reconstructed maps, while also delivering strong performance in 3D open-vocabulary semantic segmentation and scene reconstruction. We further deploy DGSG-Mind on real-world robots to demonstrate its target-oriented reasoning and dynamic update capabilities. The project page of DGSG-Mind is available at this https URL
[CV-63] Ciphera: A Decentralised Biometric Identity Framework
链接: https://arxiv.org/abs/2605.29868
作者: Ankit Kanaiyalal Prajapati,Shahzad Memon,Mohammed Mahir Rahman,Ameer Al-Nemrat
类目: Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV); Distributed, Parallel, and Cluster Computing (cs.DC)
备注: Accepted at the CyberAI 2026 Conference, and to be indexed at IEEE-Scopus
Abstract:Centralised biometric identity systems expose users to single points of failure, opaque verification processes, and irreversible biometric compromise. Decentralised Identifiers (DIDs) and Verifiable Credentials (VCs) offer stronger privacy guarantees, yet their integration with biometric authentication and distributed verification remains insufficiently explored. This paper presents Ciphera, a decentralised biometric identity framework combining privacy-preserving facial recognition, multi-node verification, IPFS-based credential metadata storage, and blockchain-anchored revocation. Evaluated across functional, performance, security, and distributed consistency dimensions, Ciphera achieved an 81% functional success rate, with stable enrolment and authentication but measurable revocation propagation delays and occasional audit-log inconsistencies. Performance testing demonstrated sub-second p95 verification latency of approximately 820ms under concurrent multi-node conditions. Security analysis confirmed strong confidentiality and integrity guarantees, though incomplete liveness detection leaves susceptibility to deepfake and replay attacks. The results demonstrate the feasibility of decentralised biometric identity while identifying key engineering challenges for production-grade deployment.
[CV-64] Masked Diffusion Vision-Language Models for Temporal Action Localization
链接: https://arxiv.org/abs/2605.29858
作者: Fengshun Wang,Zhengbo Zhang,Zhigang Tu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Temporal action localization (TAL) requires recognizing the target event and localizing its start and end times precisely in untrimmed videos. Recent vision-language formulations improve semantic reasoning and support language-conditioned outputs, but their autoregressive decoders still generate tokens from left to right, preventing later semantic evidence from revising earlier timestamp predictions. We adapt masked diffusion vision-language models (MDVLMs) to TAL so that semantic tokens and boundary tokens remain editable throughout iterative denoising with bidirectional attention, allowing temporal boundaries and semantic content to be refined jointly. Direct adaptation, however, creates two TAL-specific mismatches: standard masked diffusion training corrupts all positions uniformly at random, but the time tokens are more reliable when enough semantic context is available; and token-level cross-entropy does not reflect temporal IoU. To address these mismatches, we introduce a Planned Training Objective that uses boundary-aware masking and step-weighted reconstruction to rehearse the late recovery of time tokens, together with a Step-Level IoU Reward that provides overlap-aware supervision during denoising. A standard sequence-level cross-entropy term provides the base reconstruction signal. Experiments on ActivityNet-RTL, ActivityNet-1.3, and THUMOS-14 show that MDVLM-TAL improves both temporal reasoning and boundary localization over autoregressive vision-language baselines, with especially strong gains under stricter temporal IoU criteria. Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2605.29858 [cs.CV] (or arXiv:2605.29858v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2605.29858 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Fengshun Wang [view email] [v1] Thu, 28 May 2026 12:39:04 UTC (419 KB)
[CV-65] Building and Road Recognition in Dense Urban Informal Settlements: A Dataset and Benchmark
链接: https://arxiv.org/abs/2605.29856
作者: Hongyu Long,Jiaxuan Liu,Rui Cao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 4 figures;
Abstract:As a widespread form of informal settlements, urban villages present significant challenges for sustainable urban development and governance. Precise mapping of their infrastructure is essential, however, existing remote sensing datasets primarily focus on formal urban environments, lacking fine-grained annotated data for the high-density building patterns and narrow road networks typical of urban villages. To address this gap, we introduce the \textitDenseUIS dataset, the first high-resolution remote sensing dataset specifically designed for building and road extraction in extremely dense urban informal settlements, covering 126 urban villages across Shenzhen and Guangzhou in China. Furthermore, we conduct a comprehensive evaluation of state-of-the-art deep learning models on this dataset. Experimental results reveal the limitations of existing methods in handling the unique morphological patterns of dense informal settlements, underscoring the need for specialized approaches. \textitDenseUIS therefore provides a robust benchmark for advancing fine-grained urban mapping in complex and high-density informal environments. The dataset is publicly available at this https URL.
[CV-66] Parameter-Efficient Subspace Decoupling ViT for Mitigating Multi-Task Negative Transfer in Histological Scoring ICME2026
链接: https://arxiv.org/abs/2605.29852
作者: Youhan Huang,Jiajun Li,Yilin Fang,Shuai Wang,Chuheng Li
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Multimedia (cs.MM)
备注: 6 pages, 5 figures, 2 tables. Accepted by IEEE ICME 2026. Camera-ready version
Abstract:Histological scoring is essential for diagnosing Non-Alcoholic Fatty Liver Disease (NAFLD), yet its automation remains challenging due to the high annotation cost and negative transfer among the strongly correlated NAFLD Activity Score (NAS) indicators in multi-task learning. To address this issue, we propose a subspace-decoupled multi-task Vision Transformer (ViT) that integrates lightweight task-specific Adapters with orthogonality-based constraints. This design constructs independent feature subspaces for steatosis, ballooning, and inflammation, effectively reducing task interference while retaining shared representations. We further construct a curated multi-task mouse NAFLD histology dataset with expert annotations for all NAS components. Experimental results demonstrate that the proposed method improves multi-task stability and generalization with substantially reduced computational cost compared to training separate single-task models. The code and the curated dataset have been prepared and will be made publicly available upon acceptance to support reproducibility.
[CV-67] Fairness Beyond Demographics: Optimizing Performance Across Appearance-Based Hidden Cohorts in Medical Imaging MICCAI2026
链接: https://arxiv.org/abs/2605.29827
作者: Milad Masroor,Cuong Nguyen,Kevin Wells,Gustavo Carneiro
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Pre-review version submitted to MICCAI 2026. 10 pages, 5 figures
Abstract:Medical image analysis models can exhibit performance disparities across patient subgroups, threatening clinical safety and fairness. Existing methods typically address this issue by optimizing accuracy and fairness metrics for visible demographic attributes (e.g., sex or age) considered in isolation. This strategy not only overlooks potentially more informative latent stratifications, which may reveal deeper sources of model error and inequity, but also fails to scale when multiple demographic attributes are considered simultaneously due to the resulting sparsity of training data within each subgroup. We deal with these issues by introducing the label-free hidden-cohort fairness (LHCF) training paradigm that instead of maximizing fairness over visible demographic attributes, it optimizes fairness across latent subpopulations discovered from image appearance. By clustering images into K appearance-based cohorts and applying fairness optimization over them, LHCF uncovers underlying sources of model error and avoids the combinatorial sparsity of multi-demographic attributes, reducing disparities across both single and multiple demographic attributes. We demonstrate on our proposed fairness benchmark, HIDFairBench, that LHCF provides state-of-the-art fairness results on single and multiple demographic attributes, despite never using demographic labels for training. Our results position hidden-cohort fairness as a practical, scalable, and robust alternative to demographic-based fairness optimization for trustworthy medical image analysis.
[CV-68] Not All Inputs Are Valid: Towards Open-Set Video Moment Retrieval Using Language ACM-MM2024
链接: https://arxiv.org/abs/2605.29812
作者: Xiang Fang,Wanlong Fang,Daizong Liu,Xiaoye Qu,Jianfeng Dong,Pan Zhou,Renfu Li,Zichuan Xu,Lixing Chen,Panpan Zheng,Yu Cheng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Published in ACM MM 2024
Abstract:Video Moment Retrieval (VMR) targets to retrieve the specific moment corresponding to a sentence query from an untrimmed video. Although recent works have made remarkable progress in this task, they implicitly are rooted in the closed-set assumption that all the given queries as video-relevant\footnoteIn this paper, we treat video-relevant query'' as in-distribution (ID) query’’ and video-irrelevant query'' as out-of-distribution (OOD) query’'… Given an OOD query in open-set scenarios, they still utilize it for wrong retrieval, which might lead to irrecoverable losses in high-risk scenarios, \textite.g., criminal activity detection. To this end, we creatively explore a brand-new VMR setting termed Open-Set Video Moment Retrieval (OS-VMR), where we should not only retrieve the precise moments based on ID query, but also reject OOD queries. In this paper, we make the first attempt to step toward OS-VMR and propose a novel model \textbfOpenVMR, which first distinguishes ID and OOD queries based on the normalizing flow technology, and then conducts moment retrieval based on ID queries. Specifically, we first learn the ID distribution by constructing a normalizing flow, and assume the ID query distribution obeys the multi-variate Gaussian distribution. Then, we introduce an uncertainty score to search the ID-OOD separating boundary. After that, we refine the ID-OOD boundary by pulling together ID query features. Besides, video-query matching and frame-query matching are designed for coarse-grained and fine-grained cross-modal interaction, respectively. Finally, a positive-unlabeled learning module is introduced for moment retrieval. Experimental results on three VMR datasets show the effectiveness of our OpenVMR.
[CV-69] Cert-LAS: Toward Certified Model Ownership Verification for Text-to-Image Diffusion Models via Layer-Adaptive Smoothing ICML
链接: https://arxiv.org/abs/2605.29809
作者: Leyi Qi,Yiming Li,Siyuan Liang,Zhengzhong Tu,Dacheng Tao
类目: Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR); Machine Learning (cs.LG); Multimedia (cs.MM)
备注: This paper has been accepted to the International Conference on Machine Learning (ICML) 2026. 26 pages
Abstract:Large-scale text-to-image (T2I) diffusion models have enabled unprecedented creative applications, but their unauthorized use has raised serious intellectual property concerns, making model ownership verification (MOV) increasingly critical. We find that existing backdoor-based diffusion watermarking methods often (implicitly) assume a “faithful” verification process, namely, that the verifier can query a suspicious model and obtain the faithful watermark response to complete MOV. However, in practice, adversaries may intentionally or unintentionally damage potential watermark signals, significantly degrading verification reliability. To address this issue, we propose Cert-LAS, the first certified MOV method for T2I models based on layer-adaptive smoothing. In general, Cert-LAS embeds specified watermarks using diffusion classifiers and an LFS-guided layer-adaptive noise, and verifies ownership by examining whether the suspected model exhibits significantly stronger watermark responses compared to unwatermarked references through hypothesis testing. We further prove that, under certain conditions, our Cert-LAS can still achieve reliable verification even in the presence of malicious removal attacks. Extensive experiments validate the effectiveness of Cert-LAS and its resistance to adaptive attacks. Our code is available at this https URL.
[CV-70] Low-Magnification SEM May Suffice: Interpretable Deep Learning for Multi-Scale Fracture-Cause Classification in Zirconia-Toughened Alumina
链接: https://arxiv.org/abs/2605.29798
作者: Julian Schmid,Pawel Astankow,Tom Vater,Julius Beck,Robert Cichon,Danny Krautz
类目: Computer Vision and Pattern Recognition (cs.CV); Materials Science (cond-mat.mtrl-sci); Image and Video Processing (eess.IV)
备注:
Abstract:Reliable identification of fracture origins in alumina matrix composite hip and knee implants is critical for quality assurance and patient safety, yet current fractographic workflows are time-consuming, partly subjective, and reliant on high-magnification scanning electron microscopy (SEM). We present an interpretable vision-transformer (ViT) workflow for automated classification of fracture causes in an alumina matrix composite (BIOLOX delta, CeramTec GmbH) widely used in total joint replacements. A dataset of 8,493 SEM images (50x-10,000x) was curated from five years of in-production burst and proof tests and annotated into three defect categories defined along the manufacturing chain: green body, hard machining, and material defects. Under severe class imbalance, the fine-tuned ViT reached an accuracy of 0.907 and a macro-F1 of 0.888 in stratified five-fold cross-validation, with a two-stage perceptual-hash/SSIM leakage audit confirming negligible specimen overlap. Notably, performance at low magnification (50x) was comparable to that at high magnification (1k-10kx), indicating that macro-scale features - mirror geometry and hackle line fields - already encode sufficient diagnostic signal. Grad-CAM attributions consistently localised on canonical fractographic cues (mirrors, hackles, pores, machining marks), aligning with established fractographic criteria. Together, these results position interpretable ViTs as a complementary tool for ceramic-implant quality assurance, enabling low-magnification pre-screening and reducing reliance on time-intensive high-magnification inspection.
[CV-71] Fewer Steps Better Performance: Efficient Cross-Modal Clip Trimming for Video Moment Retrieval Using Language AAAI2024
链接: https://arxiv.org/abs/2605.29793
作者: Xiang Fang,Daizong Liu,Wanlong Fang,Pan Zhou,Zichuan Xu,Wenzheng Xu,Junyang Chen,Renfu Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Published in AAAI 2024
Abstract:Given an untrimmed video and a sentence query, video moment retrieval using language (VMR) aims to locate a target query-relevant moment. Since the untrimmed video is overlong, almost all existing VMR methods first sparsely down-sample each untrimmed video into multiple fixed-length video clips and then conduct multi-modal interactions with the query feature and expensive clip features for reasoning, which is infeasible for long real-world videos that span hours. Since the video is downsampled into fixed-length clips, some query-related frames may be filtered out, which will blur the specific boundary of the target moment, take the adjacent irrelevant frames as new boundaries, easily leading to cross-modal misalignment and introducing both boundary-bias and reasoning-bias. To this end, in this paper, we propose an efficient approach, SpotVMR, to trim the query-relevant clip. Besides, our proposed SpotVMR can serve as plug-and-play module, which achieves efficiency for state-of-the-art VMR methods while maintaining good retrieval performance. Especially, we first design a novel clip search model that learns to identify promising video regions to search conditioned on the language query. Then, we introduce a set of low-cost semantic indexing features to capture the context of objects and interactions that suggest where to search the query-relevant moment. Also, the distillation loss is utilized to address the optimization issues arising from end-to-end joint training of the clip selector and VMR model. Extensive experiments on three challenging datasets demonstrate its effectiveness.
[CV-72] Improving CLIP Adaptation by Breaking Tail Alignment for Source-Free Cross-Domain Few-Shot Learning ICML2026
链接: https://arxiv.org/abs/2605.29776
作者: Shuai Yi,Yixiong Zou,Yuhua Li,Ruixuan Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by ICML 2026
Abstract:Vision-Language Models (VLMs) such as CLIP demonstrate strong zero-shot generalization, but their performance significantly degrades in cross-domain scenarios with scarce target-domain training data (Cross-Domain Few-Shot Learning, CDFSL). In this paper, we focus on the target-domain few-shot finetuning in the CLIP-based CDFSL task. Prevailing finetuning paradigms uniformly align all image patch tokens with their corresponding textual embeddings. However, we find a counterintuitive phenomenon: actively pushing away certain low-similarity image tokens, termed “tail tokens”, from their textual embeddings consistently improves target-domain performance. We delve into this phenomenon and provide a novel interpretation: under great domain shifts and scarce training data, the model can hardly extract semantic information from visual inputs; therefore, the common belief of alignment is valid only for tokens already containing sufficient semantic information; for tail tokens, forcing the alignment would lead to excessive overfitting to the scarce training, while breaking the alignment is more useful. Motivated by this, we propose Adaptive Tail-Head Alignment (ATHA), a novel fine-tuning strategy for CLIP that transforms the conventional uniform alignment paradigm to an adaptive alignment paradigm, with both alignment strengthening and weakening. Extensive experiments on four challenging CDFSL benchmarks validate our state-of-the-art performance. Our code is available at this https URL.
[CV-73] Energy-Aware NECO for Single-Pass Pixel-wise Out-of-Distribution Detection in Semantic Segmentation ICRA2026
链接: https://arxiv.org/abs/2605.29773
作者: Boyuan Zhang,Huanshan Huang,Yifei Cao
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: 7 pages, 6 figures. Accepted at the ICRA 2026 Workshop on Long-term Deployments in the Wild (LoWi 2026)
Abstract:Reliable semantic segmentation for mobile robots requires both accurate dense prediction and robust uncertainty estimation under distribution shift. Strong uncertainty baselines such as Monte Carlo Dropout often require repeated stochastic forward passes and are difficult to deploy on edge platforms. We propose Energy-Aware NECO, a single-pass pixel-wise out-of-distribution (OOD) detector for semantic segmentation. The method combines a centered NECO-style geometric ratio computed from decoder features with a logit-based Energy score. Both components are standardized using statistics fitted on a pure in-distribution validation split and fused through a convex combination. We evaluate the method on the miniMUAD subset using true pixel-level OOD labels. The proposed hybrid score achieves an AUROC of 0.8539, outperforming NECO-only (0.8280), Energy-only (0.8171), and an ensemble predictive-entropy baseline (0.8124). Additional qualitative and operating-point analyses show that the hybrid detector improves overall ranking performance while preserving the efficiency advantages of a single-pass design. Code is available at this https URL Comments: 7 pages, 6 figures. Accepted at the ICRA 2026 Workshop on Long-term Deployments in the Wild (LoWi 2026) Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Robotics (cs.RO) Cite as: arXiv:2605.29773 [cs.CV] (or arXiv:2605.29773v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2605.29773 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Boyuan Zhang [view email] [v1] Thu, 28 May 2026 11:19:46 UTC (1,036 KB)
[CV-74] GeoMag: Geometric-Aware Video Motion Magnification via State Space Model ICME2026
链接: https://arxiv.org/abs/2605.29762
作者: Kecheng Han,Yuchen Zhang,Bingqing Liu,Boqiang Guo,Wenbin Zheng,Shiyuan Pei
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ICME 2026 Spotlight
Abstract:Video Motion Magnification (VMM) reveals imperceptible dynamics but often suffers from structural inconsistencies under complex geometric transformations. Existing learning-based methods generally face a trade-off between the limited global context of CNNs and the high computational cost of Transformers. In addition, current training protocols, largely dominated by simple linear motion, fail to capture the geometric and imaging complexities encountered in real-world videos. To address these issues, we propose GeoMag, a geometric-aware VMM framework built upon State Space Models to achieve globally consistent motion amplification with linear complexity. We further construct Geo-200K, a large-scale synthetic dataset that introduces rich geometric transformations together with sensor-realistic degradations, improving the diversity and realism of training signals. Extensive experiments on synthetic and real-world benchmarks show that GeoMag consistently outperforms prior methods in visual fidelity and computational efficiency, while producing fewer artifacts and better structural consistency.
[CV-75] S2MDF: A Plug-And-Play Layer for Intersection-Free Multi-Object Signed Distance Fields
链接: https://arxiv.org/abs/2605.29761
作者: Deniz Sayin Mercadier,Federico Stella,Aurel Bizeau,Nicolas Talabot,Pascal Fua
类目: Computer Vision and Pattern Recognition (cs.CV); Computational Geometry (cs.CG)
备注:
Abstract:Compositional implicit surface representations model scenes as collections of objects, each encoded by a Signed Distance Field (SDF). A fundamental limitation of this approach is that multiple SDFs can produce geometries that interpenetrate, violating physical plausibility. Existing mitigation strategies rely on soft penalty terms that reduce but do not eliminate intersections, and require careful loss weighting. To truly prevent interpenetration, we propose a hard constraint on vector-valued SDFs and introduce S2MDF, a lightweight plug-and-play module that enforces the constraint on any object-compositional SDF representation without architectural modifications. It introduces negligible computational overhead and is compatible with linearly-interpolated standard meshing algorithms such as Marching Cubes. It can be applied during training or as a post-processing step. Experiments on multiple state-of-the-art compositional methods show that S2MDF reduces intersections to numerical precision while preserving reconstruction quality, outperforming existing mitigation strategies.
[CV-76] SLAD : Shared LoRA Adapters for Task Specific Distillation CVPR
链接: https://arxiv.org/abs/2605.29726
作者: Reda Bensaid,Yassir Bendou,Vincent Gripon,François Leduc-Primeau
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: CVPR Findings 2026
Abstract:In the context of resource-constrained environments such as embedded systems, adapting reduced-size foundation models to downstream tasks has become increasingly popular. This has recently motivated the emerging setting of task-specific distillation, where a larger and a smaller version of the same foundation model are both adapted to the same downstream task, with the goal of transferring knowledge from the former to the latter. Recent work has demonstrated the benefits of using a larger version of the same foundation model to assist the adaptation of a smaller one. Typically, the larger model (teacher) is first adapted via fine-tuning or linear probing before its knowledge is distilled into the smaller model (student). While fine-tuning the teacher often increases its performance, recent work showed that probing it leads to better knowledge distillation to the student. Our findings show that this is mainly due to a mis-alignment in feature representation between the teacher and the student which occurs during the teacher’s fine-tuning. Inspired by existing efforts to preserve previously learned knowledge, we first propose leveraging low-rank adaptation, resulting in better feature alignment and therefore better knowledge transfer. Drawing from this insight, we further enhance the feature alignment through a parameter-sharing strategy of the adapters between the two encoders during joint training. Our proposed method, SLAD, shows better feature alignment between the teacher and student, which results in increased performance for not only the student but also the teacher model, while being 2x faster to train than fine-tuning. Through extensive experiments on multiple classification and segmentation datasets, we demonstrate the improved accuracy and transfer efficiency of our method, achieving state-of-the-art performance in the task-specific distillation framework.
[CV-77] Efficient Validation-Free Intrinsic Quality Estimation for Large-Scale Face Recognition Datasets ICML2026
链接: https://arxiv.org/abs/2605.29720
作者: Zhichao Chen,Yongle Zhao,Kaicheng Yang,Meng Yang,Yin Xie,Ziyong Feng
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: ICML 2026
Abstract:We propose Intrinsic Quality (IQ), a validation-free metric designed to estimate the inherent potential of face recognition (FR) datasets to produce high-performance models without the need for full-scale training. IQ integrates two components: (i) a Neighbor-Consistency Score that quantifies local identity label agreement via nearest neighbors, and (ii) Global Representation Subspace Complexity (Effective Rank, ER), which captures the underlying embedding geometry and dataset diversity. IQ allows for rapid evaluation using lightweight proxy models or data subsets, facilitating dataset diagnosis and curation prior to resource-intensive full-scale training. We describe an experimental protocol tailored to clean, noisy, and mixed-quality FR datasets, and outline evaluation methodologies to validate IQ’s predictive power for downstream performance.
[CV-78] Unsupervised Semantic Segmentation Facilitates Model Understanding
链接: https://arxiv.org/abs/2605.29691
作者: Xiaoyan Yu,Lisa Mais,Jannik Franzen,Peter Hirsch,Nick Lechtenbörger,Andreas Mardt,Dagmar Kainmüller
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Self-supervised learning (SSL) has produced a diverse landscape of vision transformers (ViTs) whose pretrained representations support a wide range of downstream tasks. Towards a better understanding of these models, a body of work has assessed the mechanics of their self-attention as well as the types of information captured across their representations, revealing, for example, stark differences between models trained with contrastive learning (CL) and masked image modeling (MIM). However, these advances in model understanding have not yet fully permeated the broader community, where insights specific to CL models are sometimes generalized to MIM models. To make model understanding straightforward and intuitive for a broad audience, we propose a simple and easily interpretable visualization protocol. Our protocol is based on visualizing unsupervised semantic segmentation results, yet our goal is not to maximize segmentation performance. Instead, it allows us to convey model behaviors that consistently emerge across images. Benchmarking a diverse set of SSL models across layers and representations, we obtain novel insights into distinct positional biases and scaling behaviors, including strong boundary artifacts in DINOv3-Large model tokens. These insights complement and help communicate a range of previous findings. Our protocol further enables a clear visual distinction between positional effects and the closely related but distinct locality bias, which has been studied much more extensively in the literature. The protocol is publicly available on GitHub and we believe it will catalyze further model understanding for a broad community. Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2605.29691 [cs.CV] (or arXiv:2605.29691v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2605.29691 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-79] A Geometric View of SRC: Learning Representations for Stable Residual Inference
链接: https://arxiv.org/abs/2605.29673
作者: Vangelis P. Oikonomou
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 37 pages
Abstract:Reconstruction-based inference assigns a class by comparing class-wise reconstruction residuals; Sparse Representation Classification (SRC) is a canonical instance whose reliability depends on the geometry of the learned representation. We adopt a strict training-inference separation: SRC is used only as a fixed test-time rule and is never differentiated, unrolled, or optimized during training. In a span-level idealization based on class-conditional spans and their associated projection residuals, we formalize residual-ordering stability through a residual margin and characterize geometric obstructions – span overlap, dominance, and near-overlap via small principal angles – that can collapse this margin in worst-case directions. This span-level theory is primary: it specifies when the idealized residual family is well-separated, and it provides a conditional solver-level interpretation for practical residual approximations (e.g., OMP) insofar as they remain close to the span-level residual ordering. Under explicit coverage and separation assumptions, we derive a quantitative lower bound on the (idealized) residual margin. Guided by these targets, we propose geometry-shaping objectives that promote masked within-class self-expressiveness, discourage cross-class reconstruction pathways and inter-class span alignment, and prevent collapse – without invoking SRC residuals or predictions during training. Experiments on images (COIL-100), text (TREC), and EEG connectivity evaluate all representations under identical fixed SRC/OMP inference and report residual margins and geometric diagnostics; cross-entropy is included only as a reference geometry under the same evaluation protocol.
[CV-80] SAFE-Pruner: Semantic Attention-Guided Future-Aware Token Pruning for Efficient Vision-Language-Action Manipulation
链接: https://arxiv.org/abs/2605.29662
作者: Shilin Ma,Chubin Zhang,Changyuan Wang,Yuji Wang,Yue Wu,Zixuan Wang,Jingqi Tian,Zheng Zhu,Yansong Tang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Real-time inference of vision-language-action (VLA) models is essential for robotic control. While visual token pruning has shown strong potential for accelerating inference, most existing methods mainly base pruning decisions on shallow-layer cues and risk discarding visual information required by deep layers. To address this issue, we propose SAFE-Pruner, a plug-and-play pruning framework that incorporates attention cues of future layers into pruning decisions. Specifically, we identify semantic attention consistency, the tendency that VLA models concentrate their attention probability mass on the same semantic entity across execution steps. Based on this observation, we design a forward-looking strategy to forecast the token saliency in deep layers, which prevents the premature removal of critical tokens and leads to more stable acceleration. We further introduce an adaptive subtask division strategy to detect abrupt attention shifts, thereby improving forecasting accuracy and pruning reliability. Extensive experiments in simulation and real-world settings demonstrate that our method achieves up to 1.89x speedup with a minimal degradation in success rate of less than 1.7%, while outperforming state-of-the-art methods by up to 1.9%.
[CV-81] Geometry-Guided Modeling of Foundation Features Enables Generalizable Object Shape Deformation Learning ICML2026
链接: https://arxiv.org/abs/2605.29661
作者: Yiyao Ma,Kai Chen,Zhongxiang Zhou,Zhuheng Song,Dongsheng Xie,Zelong Tan,Rong Xiong,Qi Dou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 20 pages, 12 figures, accepted by ICML 2026
Abstract:Monocular 3D shape recovery is fundamental to geometric understanding, yet achieving robust generalization across arbitrary viewpoints and unseen object categories remains a significant challenge. In this paper, we present a generalizable deformation learning framework that reconstructs 3D objects by explicitly deforming a category-level shape template to match the target observation. To address complex shape variations between the template and the target, we introduce a geometry-guided feature modeling mechanism. This process first enriches foundation features with template topology to yield a geometry-aware representation, which is then explicitly correlated with the target observation to guide precise deformation. Furthermore, to bridge the disparity between the fixed template and arbitrary target views, we propose a view-adaptive feature aggregation module. This module leverages multi-view template features and their corresponding camera poses to enrich the canonical template representation, ensuring robust feature alignment regardless of the target’s perspective. Extensive experiments demonstrate that our approach significantly outperforms state-of-the-art methods in handling large shape variations and diverse viewpoints, exhibiting strong generalization to novel categories and effectively supporting downstream real-world dexterous robotic manipulation tasks. Project homepage: this https URL
[CV-82] OccamToken: Efficient VLM Inference with Training-Free and Budget-Adaptive Token Pruning
链接: https://arxiv.org/abs/2605.29657
作者: Geng Li,Guohao Chen,Ting Chen,Shilin Shan,Kuangji Zuo,Bofan Lyu,Tuo An,Gen Li,Jianfei Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 26 pages,8 figures
Abstract:Vision-language models (VLMs) rely on long visual token sequences for visual understanding, making the prefill stage expensive in both computation and memory. Most existing pruning methods follow an absolute-ranking paradigm, assigning importance scores to visual tokens and retaining a fixed top-K subset. In this work, we argue that this paradigm is fundamentally brittle: attention sinks distort token importance rankings, while image redundancy and query-dependent visual evidence make fixed token budgets unreliable across inputs. We propose OccamToken, a training-free framework that replaces absolute token ranking with register-anchored relative evidence testing. Instead of asking which tokens are globally important, OccamToken evaluates whether a visual token provides information beyond a register-based reference. Our key insight is that register tokens naturally absorb low-information attention patterns, making them a stable reference for identifying genuinely informative visual evidence. Based on this principle, OccamToken performs both image-adaptive redundancy pruning and query-adaptive relevance pruning through dynamic thresholds derived from register attention. Across LLaVA-NeXT, LLaVA-v1.5, and Qwen3-VL, OccamToken consistently improves the accuracy-efficiency trade-off without additional training. Notably, on LLaVA-NeXT, it reduces 2,880 visual tokens to approximately 40 while preserving over 93% of the original accuracy, enabling stable visual token compression even in the extreme 1.4% retention regime.
[CV-83] SuperVoxelGPT : Adaptive and Ordered 3D Tokenization for Autoregressive Shape Generation
链接: https://arxiv.org/abs/2605.29655
作者: Yuan Li,Congyi Zhang,Xifeng Gao,Xiaohu Guo
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注:
Abstract:Autoregressive multimodal large language models (MLLMs) enable 3D generation but struggle to scale to high-resolution shapes due to inadequate 3D tokenizations. Compact set-based representations discard deterministic spatial ordering, leading to ambiguous sequence prediction, while uniform or octree-based voxel grids preserve ordering at the cost of severe redundancy and excessively long sequences. This structural trade-off limits stable and efficient autoregressive 3D generation. We present SuperVoxelGPT, a representation-first framework that resolves this tension through adaptive and deterministically ordered supervoxel tokenization. Given a prompt, we first predict a coarse geometric saliency distribution and construct a shape-adaptive supervoxel partition using saliency-guided centroidal Voronoi tessellation, allocating fine-grained cells to complex regions and larger cells to smooth regions. Conditioned on the text and ordered supervoxel layout, we introduce a SuperVoxelVAE and fine-tune a pretrained MLLM to autoregressively generate supervoxel tokens. Experiments on Trellis-500K show that SuperVoxelGPT reduces token sequence length to 12.8% of uniform voxel tokenization while achieving state-of-the-art generation quality and an average 10 \times speedup over prior methods. Subjects: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR) Cite as: arXiv:2605.29655 [cs.CV] (or arXiv:2605.29655v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2605.29655 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Yuan Li [view email] [v1] Thu, 28 May 2026 09:17:11 UTC (7,342 KB) Full-text links: Access Paper: View a PDF of the paper titled SuperVoxelGPT: Adaptive and Ordered 3D Tokenization for Autoregressive Shape Generation, by Yuan Li and 3 other authorsView PDFHTML (experimental)TeX Source view license Current browse context: cs.CV prev | next new | recent | 2026-05 Change to browse by: cs cs.GR References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); About Help contact arXivClick here to contact arXiv Contact subscribe to arXiv mailingsClick here to subscribe Subscribe Copyright Privacy Policy Web Accessibility Assistance arXiv Operational Status
[CV-84] MARTIAN: A Rendering Framework for Aerial Mars Imagery from HiRISE Orbital Data
链接: https://arxiv.org/abs/2605.29647
作者: Dario Pisanti,Georgios Georgakis
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Aerial navigation on Mars requires vision-based pipelines that are robust to the diverse illumination conditions and terrain morphology of the Martian surface. A key bottleneck for training and evaluating such methods is the scarcity of large-scale, annotated aerial datasets. We present MARTIAN, an open-source Blender-based rendering framework that leverages real HiRISE orbital map products to synthesize realistic aerial views of the Martian terrain under controllable lighting conditions and at varying altitudes. MARTIAN generates observations with accurate pose annotations, directly addressing the scarcity of training data for vision-based navigation on Mars. The framework has been validated through its deployment in concurrent work on map-based localization systems for Ingenuity and future Mars rotorcraft, where synthetically trained deep image matchers were successfully evaluated on real Mars imagery. MARTIAN is publicly available at: this https URL.
[CV-85] Learning Context-Conditioned Predicate Semantics via Prototype Feedback ICML2026
链接: https://arxiv.org/abs/2605.29610
作者: NamGyu Jung,Chang Choi
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at ICML 2026. Code: this https URL
Abstract:In scene graph generation, a central challenge is modeling polysemous predicates whose meanings shift across contexts. Prior approaches address this issue by decomposing predicates into multiple static prototypes or retrieving semantically similar exemplars. However, these strategies keep predicate representations static and cannot reorganize semantics to reflect image-specific evidence, leading to systematic confusions in ambiguous contexts. We propose AlignG, which learns context-conditioned predicate semantics via prototype feedback. AlignG infers context-conditioned predicate semantics from the relation candidates within each image and feeds the adapted semantics back to recalibrate relation representations. The learning objective anchors this adaptation to global semantic centers, preventing semantic drift while still allowing selective reorganization when the scene provides consistent relational cues. Experiments on VG-150 and GQA-200 show consistent improvements over state-of-the-art baselines, with F@100 improvements of +1.4 on VG-150 and +2.7 on GQA-200 under SGDet. We further visualize per-image prototype similarity shifts and observe coherent context-dependent reorganization where prototypes selectively merge or separate predicates according to scene evidence. The code is available at this https URL.
[CV-86] CogniVerse: Revolutionizing Multi-Modal Retrieval-Augmented Generation with Cognitive Reflection and Geometric Reasoning CVPR2026
链接: https://arxiv.org/abs/2605.29602
作者: Xiang Fang,Wanlong Fang,Changshuo Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted in CVPR 2026
Abstract:Multi-modal Retrieval-Augmented Generation (MMRAG) has emerged as a powerful paradigm for enhancing Multimodal Large Language Models in knowledge-intensive question answering by integrating external visual, textual, and structural knowledge. However, existing MMRAG frameworks suffer from critical limitations, including noisy and irrelevant retrieval, cross-modal semantic misalignment, lack of adaptive reasoning, and incoherent generation across local and global contexts. We introduce \textbfCogniVerse, a novel MMRAG framework that addresses these challenges through a cognitive-inspired, mathematically rigorous approach. Drawing from human-like reasoning, CogniVerse integrates three synergistic components: (1) a Cognitive Reflection Module that dynamically assesses retrieval necessity and filters relevant multi-modal content, reducing noise and computational overhead; (2) a Multi-modal Retrieval Module that aligns embeddings in a Riemannian manifold using information geometry and refines knowledge graphs via spectral graph theory, ensuring precise and coherent retrieval; and (3) a Hierarchical Generation Module that employs an optimal transport-based loss to balance token-level accuracy and global semantic coherence. Extensive experiments demonstrate that CogniVerse significantly outperforms state-of-the-art systems in both accuracy and coherence, while reducing retrieval latency.
[CV-87] How to Relieve Distribution Shifts in Semantic Segmentation for Off-Road Environments
链接: https://arxiv.org/abs/2605.29599
作者: Ji-Hoon Hwang,Daeyoung Kim,Hyung-Suk Yoon,Dong-Wook Kim,Seung-Woo Seo
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 8 pages, 6 figures. Accepted to IEEE Robotics and Automation Letters (RA-L). \c{opyright} 2025 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses
Abstract:Semantic segmentation is crucial for autonomous navigation in off-road environments, enabling precise classification of surroundings to identify traversable regions. However, distinctive factors inherent to off-road conditions, such as source-target domain discrepancies and sensor corruption from rough terrain, can result in distribution shifts that alter the data differently from the trained conditions. This often leads to inaccurate semantic label predictions and subsequent failures in navigation tasks. To address this, we propose ST-Seg, a novel framework that expands the source distribution through style expansion (SE) and texture regularization (TR). Unlike prior methods that implicitly apply generalization within a fixed source distribution, ST-Seg offers an intuitive approach for distribution shift. Specifically, SE broadens domain coverage by generating diverse realistic styles, augmenting the limited style information of the source domain. TR stabilizes local texture representation affected by style-augmented learning through a deep texture manifold. Experiments across various distribution-shifted target domains demonstrate the effectiveness of ST-Seg, with substantial improvements over existing methods. These results highlight the robustness of ST-Seg, enhancing the real-world applicability of semantic segmentation for off-road navigation.
[CV-88] Non-Forgetting Knowledge Allocation with Bi-level Competition for Class-Incremental Learning
链接: https://arxiv.org/abs/2605.29592
作者: Xiang Tan,Run He,Yawen Cui,Mengchen Zhao,Yan Wu,Tianyi Chen,Huiping Zhuang,Xiaonan Luo,Guanbin Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Class-Incremental Learning (CIL) with pre-trained models (PTMs) aims to sequentially adapt PTMs to new categories without forgetting old knowledge. Built upon PTMs, existing adapter-based methods mainly train models via distinct task-specific adapters, and present a uniform knowledge allocation for each adapter during inference. However, this allocation mechanism ignores the nature of task discrepancy and leads to suboptimal utilization of adapters. Also, under CIL constraint, an allocator is prone to forgetting when tasks evolve. To address these issues, we propose a Non-Forgetting Allocation with Bi-Level Competition (NoFA-BC). NoFA-BC constructs a non-forgetting allocator (NFA) by transforming the allocator training into a recursive least-squares problem and achieves an allocator equivalent to that trained with all data. Based on the NFA, a Bi-Level Competition (BLC) including an intra-task level Winner-Takes-All (WTA) mechanism and inter-task Last-Ones-Fall (LOF) elimination is proposed to provide better allocation of adapter knowledge. WTA extracts the most significant logit within a task to represent the adapter’s contribution and LOF suppresses the irrelevant adapters. With BLC, participation ratio of each adapter can be tailored for each input. Moreover, a Stability Enhancement (SE) process is incorporated to further improve the performance of old tasks.
[CV-89] Brain-IT-VQA: From Brain Signals to Answers
链接: https://arxiv.org/abs/2605.29588
作者: Roman Beliy,Matias Cosarinsky,Oliver Heinimann,Navve Wasserman,Michal Irani
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Neurons and Cognition (q-bio.NC)
备注:
Abstract:Decoding visual content from fMRI signals recorded while a person views images, and specifically answering questions about the seen images, is a long-standing challenge. While significant progress has been made in recent years in visual question answering (VQA) from fMRI, performance remains limited. Moreover, although recent models can make increasingly accurate predictions, they have rarely been used as tools for understanding the structure of visual representations in the brain. We present Brain-IT-VQA, a framework for visual question answering from fMRI. Building on the Brain Interaction Transformer (Brain-IT), our method decodes language tokens from brain activity and integrates them with a language model to answer visual questions. Our model substantially outperforms previous fMRI-based captioning and VQA approaches. We further introduce NSD-VQA, a new dataset and benchmark for visual question answering from fMRI. Unlike existing image-fMRI VQA datasets, which typically provide only a few broad and weakly controlled questions per image, NSD-VQA provides on average 20 question-answer pairs per image across 20 controlled question categories that disentangle multiple levels of visual understanding. This enables more reliable and interpretable evaluation despite limited fMRI test data. Together, Brain-IT-VQA and NSD-VQA provide both a strong predictive framework and a tool for studying brain representations. Using this benchmark, we quantify which forms of visual and semantic information can be reliably decoded from fMRI responses to natural images. We further analyze the contributions of different brain regions across question types.
[CV-90] BitC-3DGS: High-Capacity 3D Gaussian Splatting Watermarking via Bit Compression
链接: https://arxiv.org/abs/2605.29583
作者: Yuquan Bi,Baosheng Yu,Yingke Lei,Jianwei Yang,Hongsong Wang,Jie Gui,Yuan Yan Tang,James Tin-Yau Kwok
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:High-capacity watermarking is necessary for 3D Gaussian Splatting (3DGS) assets to embed rich information (e.g., ownership, provenance, and authentication codes), enabling reliable identification and integrity verification in large-scale 3D asset pipelines. Existing bit-to-token watermarking methods based on a pre-trained text encoder are limited to 77-bit messages due to CLIP’s fixed 77-token context length, as tokens beyond this limit are unsupported by learned positional embeddings. To address this limitation, we introduce BitC-3DGS, a bit-compression framework that encodes multiple message bits per token. It employs a bit-compressed tokenization scheme that encodes multiple bits within the same chunk into a single semantic token. To enable recovery of the compressed information, it further introduces a dual-branch architecture for joint chunk decompression and bit decoding, along with a hard-message sampling strategy to improve combinatorial coverage during decoder training. Extensive experiments on the Blender and LLFF datasets demonstrate the effectiveness of BitC-3DGS for high-capacity watermarking, achieving high message recovery accuracy and rendering fidelity. For example, it supports 128-bit message capacity with recovery accuracy comparable to that of 64-bit messages in recent state-of-the-art methods.
[CV-91] ReactBench: A Cause-Driven Benchmark for Multimodal Hallucination via Systematic Evaluation
链接: https://arxiv.org/abs/2605.29579
作者: Shizhe Zhou,Bohan Jia,Kai Wu,Yan Shen,Tongyun Li,Yuyang Wu,Shaohui Lin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:While multimodal large language models (MLLMs) have achieved rapid progress in vision-language understanding, they remain prone to multimodal hallucinations, producing responses that are inconsistent with the visual input. Existing benchmarks predominantly focus on detecting hallucination outcomes rather than evaluating the underlying causes of these failures. Moreover, many benchmarks rely on simplistic scenarios and limited evaluation formats that no longer challenge state-of-the-art models. To address these limitations, we introduce ReactBench, a cause-driven hallucination benchmark featuring multiple tasks and an exam-style evaluation format. By generating adversarial images and hallucination-inducing queries, ReactBench introduces four targeted tasks: Relational Erasure, Counterfactual Attribute, Alteration Tracing, and Dense Counting. These tasks systematically expose co-occurrence bias, language priors, cross-image comparative perception deficiencies, and fine-grained perceptual bottlenecks. Beyond standard accuracy-based evaluation, we leverage Chain-of-Thought reasoning to identify fine-grained sub-causes of hallucination within each task. Extensive evaluations reveal that current MLLMs remain notably vulnerable to cause-specific hallucination triggers, demonstrating the value of ReactBench as a systematic and interpretable testbed for diagnosing and improving multimodal model robustness. The project page is available at this https URL.
[CV-92] Mitigating State Aliasing in Vision-Language-Action Models via Inverse Dynamics Learning
链接: https://arxiv.org/abs/2605.29577
作者: Kyujin Lee,Injae Kim,Jihwan Park,Yejun Ju,Minseok Joo,Hyunwoo J. Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-Language-Action (VLA) models have emerged as a promising framework that unifies perception, reasoning, and control for robot manipulation by adapting pretrained vision-language models (VLMs) to action prediction. However, VLM-derived representations are often insensitive to subtle visual distinctions required for low-level control, causing state aliasing between visually similar states that require substantially different actions. Prior VLA studies improve visual understanding by generating visual or reasoning outputs, such as future frames, 2D grounding points or traces, or intermediate spatial reasoning steps, but these objectives typically shape the vision encoder only indirectly through end-to-end prediction and do not explicitly analyze state aliasing in the learned visual feature space. To mitigate state aliasing, we introduce inverse dynamics learning as an auxiliary objective that directly supervises the VLA vision encoder. By predicting the action between current and future observations, our objective encourages the encoder to capture fine-grained visual distinctions that determine low-level actions. We further use pseudo-reversed supervision to expose the encoder to a broader range of action directions and improve generalization under limited robot demonstrations. Our method applies to diverse VLA baselines, uses only standard observation-action pairs without additional annotations, and preserves the original inference pipeline at test time. Experiments on CALVIN ABC-D and SimplerEnv show consistent gains across diverse VLA baselines. Frozen-encoder probing and state-feature alignment analyses further show that our method learns state-discriminative visual representations that reduce state aliasing and better align with robot state changes.
[CV-93] Optimizing Latent Representations for Robust Building Damage Assessment Onboard Earth Observation Satellites CVPR
链接: https://arxiv.org/abs/2605.29575
作者: Thomas Goudemant,Benjamin Francesconi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW 2026), Jun 2026, Denver, United States
Abstract:Rapid identification of damaged buildings after natural disasters or on war areas is crucial to support emergency response and prioritize interventions. Earth Observation constellations provide timely, large-scale coverage, but actionable information is often delayed by data downlink constraints, on-ground processing, and human interpretation. Reducing this latency is essential to improve decision-making responsiveness. In this work, we propose an original AI-based system that enables object-level building damage assessment (localization and damage classification) directly onboard satellites from pre-disaster and post-disaster highresolution optical imagery. Available pre-disaster images are encoded on ground into compact latent representations, transmitted to the satellite, and compared on-board with newly acquired post-event observations. Leveraging AI interpretation capabilities and increasing processing capabilities on-board satellites, the proposed design enables processing directly at the data source, reducing the amount of information to be downlinked while preserving task-relevant content and improving overall system responsivity. We explore the design space through a systematic benchmark of onboard-compatible variants, analyzing the impact of siamese processing, cross-attention, latent-space compression, and robustness-oriented data augmentation. Experiments on xBD dataset demonstrate reliable and robust damage assessment under misalignment, with minimal performance degradation under strong compression.
[CV-94] DefSynUS: Real-time Patient-specific Intrahepatic Vessel Identification via Deformation-Aware CT-US Domain Adaptation
链接: https://arxiv.org/abs/2605.29570
作者: Karl-Philippe Beaudet(MIMESIS, UNISTRA),Yordanka Velikova(TUM),Sidaty El Hadramy(MIMESIS, Unibas),Nassir Navab(TUM),Philippe Cattin(Unibas),Juan Verde(MIMESIS, UNISTRA, IHU Strasbourg),Stéphane Cotin(MIMESIS, UNISTRA)
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Purpose: Laparoscopic ultrasound (LUS) enhances the safety of liver surgery by visualizing intrahepatic vessels in real-time. Still, vessel identification remains difficult due to probe constraints, complex vascular structure, and tissue deformation. This work aims to enable real-time, patient-specific vessel identification that remains robust under deformation through deformable ultrasound augmentation. Methods: Preoperative CT vessel annotations are used to generate synthetic ultrasound data via optimized physics-based rendering, coupled with domain adaptation to intraoperative ultrasound. The rendering is trained end-to-end for vessel identification and patient-specificity, eliminating the need for preoperative ultrasound. A deformation-aware augmentation simulates realistic intraoperative motion and tissue deformation within the rendering pipeline. Results: In abdominal phantom and limited clinical feasibility experiments (single-case clinical evaluation), the framework achieved real-time intrahepatic vessel-branch identification, maintaining performance under new patient poses. Conclusion: The framework enables real-time vessel identification without preoperative ultrasound and supports technical feasibility, but multi-patient validation is still needed for generalizability and clinical feasibility.
[CV-95] From General Vision to Reliable Traversability Estimation: Adapting Vision Foundation Models for Unstructured Outdoor Environments
链接: https://arxiv.org/abs/2605.29565
作者: Ji-Hoon Hwang,Jisung Bae,Dong-Wook Kim,Yeonkyu Lee,Seung-Woo Seo
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注: 8 pages, 5figures
Abstract:Vision-based approaches have become the dominant paradigm for traversability estimation in unstructured outdoor environments, typically adapting vision foundation models (VFMs) via semantic segmentation supervision. However, this paradigm faces three fundamental challenges that undermine its reliability: the task-agnostic design of VFMs, the ambiguity of traversability annotations, and the discrepancy between semantic labels and physical safety. We propose Vision-to-Traversability Adaptation (ViTA), a framework that adapts VFMs for reliable traversability estimation, instantiated on SAM2. ViTA injects task-specific knowledge through learnable traversability prompts while preserving the VFM’s cross-domain generalization. To handle annotation ambiguity, we introduce Perspective-Diversified Training, which estimates semantic uncertainty to suppress confident predictions at ambiguous boundaries. To bridge the semantic-traversability discrepancy, we distill geometric knowledge during training, enabling slope and elevation reasoning from RGB images alone at inference. The semantic and geometric outputs are fused into a continuous traversability score that reflects both semantic uncertainty and geometric risk. Evaluations across diverse domains, including challenging real-world off-road datasets, demonstrate that ViTA achieves state-of-the-art IoU and Precision with substantial false-positive reduction and strong cross-domain generalization.
[CV-96] Planning with the Views via Scene Self-Exploration
链接: https://arxiv.org/abs/2605.29563
作者: Kangrui Wang,Linjie Li,Zhengyuan Yang,Shiqi Chen,Zihan Wang,Li Fei-Fei,Jiajun Wu,Leonidas Guibas,Lijuan Wang,Manling Li
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Can VLMs predict how each camera move changes the view, and plan many such moves ahead? We call this capability view planning, requiring (1)understanding how a single action transforms the view, and (2)composing many such transformations across multi-turn plans to identify a target view. We probe both abilities in our proposed ViewSuite, a 3D point-cloud environment on real ScanNet scenes. Across 13 frontier VLMs, a critical planning gap emerges: they possess basic view-action knowledge but fail to compose it across multi-turn plans, with the gap widening as viewpoint distance grows. To close this gap, we propose an iterative framework that alternates self-exploration with view graph distillation. The key insight is that all exploration trajectories, regardless of their outcome, collectively form a view graph that compactly captures how viewpoints connect across a scene. Distilling this graph into diverse supervised tasks reshapes the policy distribution and overcomes the sparse rewards that stall pure RL. This improves Qwen2.5-VL-7B from 2.5% to 47.8% on interactive view planning, surpassing GPT-5.4 Pro (18.5%) and Gemini 3.1 Pro (21.4%). Self-exploration emerges as a promising path toward VLMs that can actively reason and plan in 3D space.
[CV-97] VLA-Pro: Cross-Task Procedural Memory Transfer for Vision-Language-Action Models
链接: https://arxiv.org/abs/2605.29562
作者: Shengyu Si,Yuanzhuo Lu,Ruimeng Yang,Ziyi Ye,Zuxuan Wu,Yu-Gang Jiang
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-Language-Action~(VLA) models have shown strong potential for general-purpose robotic manipulation, yet they still struggle to generalize to unseen tasks that necessitate transferring relevant experience across objects, scenes, and action patterns. This paper proposes VLA-Pro, a plug-and-play framework designed to enhance cross-task generalization by storing task-relevant procedural memories at training time and transferring these memories during inference. Specifically, VLA-Pro stores task-specific LoRA adapters as parameterized procedural memories during training. At inference time, VLA-Pro retrieves relevant procedural memories based on the current multi-modal context and dynamically fuses these memories for generating the current action chunk. Experiments on RoboTwin, RLBench, and real-world manipulation tasks show that VLA-Pro consistently improves cross-task generalization across multiple backbones, achieving up to a 207% relative improvement in simulation and increasing real-world success rate from 5.8% to 65.0%. These results suggest that procedural memory retrieval and adaptation provide an effective mechanism for transferring manipulation experience to novel tasks while preserving modularity and execution stability.
[CV-98] AE: Target-aware enhancer for nighttime UAV tracking ICIP2026
链接: https://arxiv.org/abs/2605.29558
作者: Yanyan Chen,Ruigang Fu,Yu Song,Ping Zhong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at ICIP 2026. Dataset is avaliable at: this https URL
Abstract:Severe image degradation under low-light nighttime conditions constitutes a core bottleneck preventing all-day applications for UAV-based single object tracking. Existing image enhancement methods often struggle to distinguish between target and background regions, which can easily lead to amplified background noise or compromise target features. To overcome this limitation, we propose TAE, a target-aware low-light enhancement framework tailored for nighttime object tracking. Guided explicitly by weak supervisory signals from tracking bounding boxes, the framework performs region-aware enhancement to ensure operations focus on the target area. It further adopts an adaptive RGB multi-curve fusion mechanism to achieve refined modeling and adaptive adjustment across different regions. To facilitate research in this domain, we also contribute DarkSOT, a new benchmark for nighttime UAV tracking, comprising 268 sequences across 9 target categories. Experimental results on the DarkSOT and UAVDark135 demonstrate that TAE significantly improves tracking performance in low-light nighttime scenarios, exhibiting strong robustness and generalization. The DarkSOT dataset is available at this https URL.
[CV-99] Learning Representations from 3D Gaussian Splats
链接: https://arxiv.org/abs/2605.29549
作者: Julia Farganus,Krzysztof Żurawicki,Arkadiusz Gaweł,Weronika Jakubowska,Halina Kwaśnicka
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 figures, 15 pages
Abstract:3D Gaussian Splatting (3DGS) is a recent approach for scene rendering. Although primarily designed for view synthesis, its potential for scene understanding tasks remains underexplored. In this work, we conduct a comparative evaluation of various geometric deep learning architectures for the classification of 3D scenes represented using Gaussian Splatting. We benchmark point-based and graph-based models across both traditional point cloud datasets and dedicated Gaussian Splatting datasets. Scenes are embedded into latent representations, which are evaluated through end-to-end classification, linear probing, and clustering analysis. Our study provides insight into the suitability of different geometry-aware architectures and input feature configurations for learning effective 3D Gaussian Splat representations. The results highlight consistent differences between architectural families and reveal the impact of Gaussian-specific attributes on the quality of representation.
[CV-100] GiPL: Generative augmented iterative Pseudo-Labeling for Cross-Domain Few-Shot Object Detection CVPR2026
链接: https://arxiv.org/abs/2605.29539
作者: Jiacong Liu,Shu Luo,Yikai Qin,Yaze Zhao,Yongwei Jiang,Yixiong Zou
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: CVPR 2026 Workshop
Abstract:Vision-language foundation models have shown promising zero-shot generalization for Cross-Domain Few-Shot Object Detection (CD-FSOD). However, they face two critical challenges in fine-tuning: insufficient support set utilization due to sparse single-instance annotations, and severe overfitting under extremely limited target-domain samples. To address these issues, this paper proposes GiPL, an efficient two-branch training this http URL the first branch, we design an iterative pseudo-label self-training paradigm, which performs zero-shot inference on the support set to generate reliable pseudo-annotations, fuses them with ground-truth labels, and iteratively optimizes the model to fully exploit support set data. In the second branch, we introduce generative data augmentation pipeline using large vision-language models, which synthesizes domain-aligned, multi-object annotated images to enrich training samples and suppress overfitting. Extensive experiments on three challenging CD-FSOD datasets (RUOD, CARPK, CarDD) under 1/5/10-shot settings demonstrate that GiPL consistently outperforms state-of-the-art methods with significant performance this http URL is available at \hrefthis https URLCDiscover.
[CV-101] RadioFormer3D: Weakly Supervised 3D Radio Map Estimation in Low-Altitude Airspace via Generative Modeling
链接: https://arxiv.org/abs/2605.29538
作者: Zheng Fang,Junjie Liu,Kangjun Liu,Jianguo Zhang,Yaowei Wang,Ke Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:With the emergence of wireless applications in three-dimensional environments, such as the low-altitude airspace and 3D heterogeneous networks, radio map estimation is increasingly required to characterize signal propagation across both horizontal and vertical dimensions. However, extending radio map estimation from 2D to 3D remains challenging due to increased spatial sparsity and limited supervision across continuous altitudes. In this paper, we propose \textbf\textitRadioFormer3D, a specialized model for volumetric spectrum reconstruction under weak supervision. Building on the dual-stream, multi-granularity fusion architecture of \textitRadioFormer, \textitRadioFormer3D introduces a Fourier-based sampling encoder and a volumetric decoder to efficiently process sparse measurements in 3D space. To alleviate the lack of vertical supervision, we propose the \textbf\textitJoint Spectrum Integrity Loss, which integrates volume-level pseudo-label supervision, map-level geometry-aware radio rendering, and pixel-level localized constraints within a unified optimization scheme. This design enables the model to capture complex vertical structural relationships more effectively under sparse supervision. Extensive experiments across several radio map datasets show that \textitRadioFormer3D achieves superior overall performance compared to representative existing methods. In particular, it demonstrates improved reconstruction quality at unlabeled altitudes while maintaining a favorable trade-off between accuracy and inference efficiency, positioning it as a highly promising solution for future 3D environment-aware wireless networks.
[CV-102] Audio Deepfake Detection with Half-Truth Localisation Using Cross-Attentive Feature Fusion
链接: https://arxiv.org/abs/2605.29531
作者: S. Sutharya,Remya K. Sasi
类目: ound (cs.SD); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 13 pages, 5 figures, 11 tables
Abstract:Audio deepfake detection is well-studied as a binary problem, but partially manipulated speech, where a short synthesised segment is spliced into an otherwise genuine utterance, poses a harder and more realistic threat. Detecting such half-truth audio requires not only distinguishing it from real and fully fake speech, but also localising where the manipulation occurs. We present CAFNet, a 576k-parameter architecture that addresses both tasks jointly: it performs ternary classification (real, fully-fake, or half-truth) and regresses the temporal boundaries of the synthesised region in a single forward pass. CAFNet fuses Mel-Frequency Cepstral Coefficient (MFCC), Linear-Frequency Cepstral Coefficient (LFCC), and Chroma Short-Time Fourier Transform (Chroma-STFT) features through parallel depthwise-separable convolution branches with cross-attention, followed by a Bidirectional Long Short-Term Memory (BiLSTM) regression head for boundary prediction. On the combined Multi-Lingual Audio Deepfake Detection Corpus (MLADDC) T2+T3 test set, CAFNet achieves 92.71% accuracy and macro Area Under the Curve (AUC) of 0.9910, with boundary localisation Mean Absolute Error (MAE) of 0.075s and a median error of 0.052s. On binary detection, it achieves 96.76% accuracy and 3.20% Equal Error Rate (EER), outperforming fine-tuned XLS-R 300M (78.31%) and AST 87M (93.03%) at over 500 times fewer parameters. A cross-dataset study further shows that standard fine-tuning collapses cross-domain representations even under reduced backbone learning rates.
[CV-103] KGEdit: Ambiguity-Aware Knowledge Graphs for Training-Free Precise Video Generation and Editing
链接: https://arxiv.org/abs/2605.29509
作者: Mingshu Cai,Miao Zhang,Chenghe Yang,Yixuan Li,Osamu Yoshie,Yuya Ieiri
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:In recent years, training-free video generation has progressed remarkably. However, when handling complex textual instructions, existing methods still suffer from semantic ambiguity, incorrect concept binding, and cross-frame inconsistency. To address these issues, we propose KGEdit, a structured semantic control framework for text-to-video (T2V) diffusion models. Specifically, we first construct an ambiguity-aware knowledge graph (AAKG) to disentangle and disambiguate the input prompt, converting it into four types of structured semantics: identity, relation, attribute, and negative constraints. We then design a structured semantic injection module (SSIM) to inject these semantic signals into key layers of the diffusion Transformer, enabling fine-grained semantic control. In addition, we introduce a temporal-aware semantic control (TASC) module that dynamically schedules semantic objectives according to the stage-wise characteristics of the denoising process, further improving semantic alignment and temporal consistency. Experiments show that KGEdit outperforms existing methods in editing precision and temporal stability, while offering higher efficiency and controllability in text-driven interaction scenarios.
[CV-104] ESAM: Efficient Online 3D Perception on the Edge
链接: https://arxiv.org/abs/2605.29505
作者: Qin Liu,Lavisha Aggarwal,Saptarashmi Bandyopadhyay,Vikas Bahirwani,Marc Niethammer,Ehsan Adeli,Andrea Colaco
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Online 3D scene perception in real time is essential for robotics, AR/VR, and autonomous systems, particularly in edge computing scenarios where computational resources are limited and privacy is crucial. Recent state-of-the-art methods like EmbodiedSAM (ESAM) demonstrate the promise of online 3D perception by leveraging the Segment Anything Model (SAM) for real-time, fine-grained, and generalized 3D instance segmentation. However, ESAM still relies on a computationally expensive 3D sparse UNet for point cloud feature extraction, which accounts for the majority of the 3D inference time, hindering its practicality on resource-constrained devices. In this paper, we propose ESAM++, a lightweight and scalable alternative for online 3D scene perception tailored to edge devices without GPU acceleration. Our method introduces a 3D Sparse Feature Pyramid Network (SFPN) that efficiently captures multi-scale geometric features from streaming 3D point clouds while significantly reducing computational overhead and model size. We evaluate our approach on four challenging segmentation benchmarks, namely ScanNet, ScanNet200, SceneNN, and 3RScan, demonstrating that our model achieves competitive accuracy with up to 3 times faster inference with a 2 times smaller model size compared to ESAM, enabling practical deployment on edge devices.
[CV-105] AnyMo: Scaling Any-Modality Conditional Motion Generation with Masked Modeling
链接: https://arxiv.org/abs/2605.29488
作者: Yiheng Li,Zhuo Li,Ruibing Hou,Yingjie Chen,Hong Chang,Hao Liu,Shiguang Shan
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Conditional human motion generation remains a fundamental challenge in computer vision and robotics. Despite significant progress, current methods are often constrained by fixed modality configurations and task-specific architectures, leaving cross-modal interactions and the scaling laws of multimodal-conditioned synthesis largely underexplored. A key bottleneck is the scarcity of large-scale modality-aligned motion data, limiting generalization across diverse control signals. In this work, we introduce OmniHuMo, a large-scale, high-quality dataset comprising over 5,000 hours of motion and 3.2 million sequences with precisely aligned multimodal annotations (e.g., text, speech, music, and trajectory). Leveraging OmniHuMo, we propose AnyMo, a unified multimodal framework combining a Residual FSQ-based motion tokenizer with a scalable masked modeling transformer, enabling high-quality motion synthesis under arbitrary modality combinations. Extensive experiments show that AnyMo achieves high-fidelity synthesis while offering flexible control over both spatial and stylistic attributes.
[CV-106] V2XCrafter: Learning to Generate Driving Scene Across Agents
链接: https://arxiv.org/abs/2605.29471
作者: Yihang Tao,Yu Guo,Senkang Hu,Yanan Ma,Zihan Fang,Sam Kwong,Yuguang Fang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Collaborative driving systems leverage vehicle-to-everything (V2X) communication for multi-agent collaborative perception to enhance driving safety, yet they remain constrained by scarce annotated real-world V2X driving datasets and limited generalization across diverse driving conditions. While image generation technology offers a feasible solution for data augmentation, existing methods tailored for single-vehicle multi-view scenarios face two fundamental challenges in multi-agent driving settings: (1) the expansion of the learning objective degrades generation quality, and (2) the highly dynamic variations across agents hinder the modeling of consistency for physical attributes (e.g., color, category) in jointly observed objects. To bridge this gap, we propose V2XCrafter, the first framework for generating controllable and realistic collaborative driving scene across agents’ camera views. For effective learning, we develop a progressive multi-agent diffusion model based on a single-agent backbone, using neighboring agents’ latent states as reference signals to progressively guide the single-to-multi diffusion. To address cross-vehicle inconsistency, we propose a cross-agent attention module that leverages a collaboration view graph and learnable jointly observed object representation to model the dynamic cross-agent camera view relationships. Experiments have shown that V2XCrafter can generate high-fidelity and controllable street views with consistency across agents, thereby effectively enhancing the downstream collaborative 3D object detection tasks.
[CV-107] Benchmarking Large Vision-Language Models on CFMME: A Comprehensive Chinese Financial Multimodal Evaluation Dataset
链接: https://arxiv.org/abs/2605.29462
作者: Qian Chen,Xianyin Zhang,Yanzhi Liu,Lifan Guo,Feng Chen,Chi Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:The emergence of Large Vision-Language Models (LVLMs) has substantially expanded model capabilities beyond text-only understanding, enabling unified inference across both visual and textual modalities and supporting a broader range of real-world applications. To comprehensively evaluate the perception, understanding, reasoning, and cognition capabilities of LVLMs throughout the entire financial business workflow in Chinese contexts, we introduce CFMME, a novel Chinese financial multimodal evaluation benchmark. CFMME comprises 6,052 instances spanning from fundamental academic knowledge to complex real-world applications, covering eight primary financial image modalities and four core multimodal tasks. On CFMME, we conduct a thorough evaluation of representative LVLMs. The results show that the state-of-the-art model attains an overall accuracy of 66.11% on the question answering task and an average score of 77.18 on the detection, recognition, and information extraction tasks, indicating substantial room for improvement in current LVLMs. In addition, we conduct detailed analyses of error causes, cross-modal capabilities, and multi-orientation settings, yielding valuable insights for future research. We hope that CFMME will spur further progress in LVLMs, especially by improving their performance on multiple multimodal tasks in the financial domain.
[CV-108] FlowSeg: Dynamic Semantic Guidance for LLM -Conditioned Segmentation ICML2026
链接: https://arxiv.org/abs/2605.29461
作者: Zekang Zhang,Guangyu Gao,Youyun Tang,ChengJing Wu,Xiaochao Qu,Chi Harold Liu,Jianbo Jiao,Yunchao Wei,Luoqi Liu,Ting Liu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 18 pages, accepted by ICML 2026
Abstract:LLM-conditioned segmentation has recently advanced rapidly by coupling large language models with iterative mask generation frameworks. However, we identify a persistent failure mode in current propose-then-select pipelines. Although high-quality mask candidates are often generated, the final prediction may fail to match the given linguistic condition. This failure arises because language semantics are typically used as static prompts or post-hoc matching signals, rather than participating in the iterative mask generation process. Through systematic analysis, we show that many errors stem from semantic misalignment rather than poor mask quality. To address this issue, we propose FlowSeg, which introduces dynamic semantic guidance via a bidirectional semantic flow between intermediate decoding states and LLM-derived condition embeddings throughout the generation process. Language conditions actively guide mask refinement at each stage, while condition embeddings are progressively updated by emerging visual evidence. This design yields semantically grounded mask representations and visually aligned language conditions, enabling more reliable matching. We further incorporate a lightweight boundary-aware refinement to selectively enhance uncertain regions without perturbing confident interiors. Extensive experiments on referring expression segmentation and reasoning segmentation tasks demonstrate that FlowSeg consistently improves language-mask alignment and achieves state-of-the-art performance. Project page: this https URL
[CV-109] FedSmoothLoRA: Toward Smoother and Faster Convergence in Federated Low-Rank Adaptation
链接: https://arxiv.org/abs/2605.29460
作者: Zehao Wang,Guanglei Yang,Yihan Zeng,Hang Xu,Hongzhi Zhang,Wangmeng Zuo,Chun-Mei Feng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 26 pages, 4 figures
Abstract:Federated fine-tuning of foundation models with Low-Rank Adaptation (LoRA) provides an efficient solution for reducing communication and computation costs while preserving data locality. However, the direct combination of FedAvg and LoRA suffers from three key issues: limited update space, which restricts the model’s effective learning capacity; inter-round state mismatch, which disrupts cross-round local optimization continuity; and a client-agnostic starting state, which slows local convergence on clients. Although recent methods mitigate the limited update space issue by merging LoRA updates into the backbone across communication rounds, inter-round state mismatch and the client-agnostic starting state remain insufficiently addressed. To address these issues, we propose FedSmoothLoRA, a federated LoRA tuning framework that preserves the enlarged update space, improves cross-round local optimization continuity, and provides a client-aware starting state for local training. At each communication round, FedSmoothLoRA constructs the local LoRA initialization using two matrices: a Round-Matching matrix that preserves cross-round local state continuity, and a Gradient-Aligned matrix that provides client-specific optimization guidance from gradient signals estimated on local data. Together, these designs enable smoother and faster convergence. Extensive experiments on image classification and natural language generation tasks demonstrate that FedSmoothLoRA consistently outperforms existing federated LoRA tuning methods. Code: this https URL
[CV-110] Uni-RCM: Unified Reference-guided Cross-modal Mapping for Multi-Class Anomaly Detection
链接: https://arxiv.org/abs/2605.29455
作者: Yangchen Wu,Huiqiang Xie
类目: Computer Vision and Pattern Recognition (cs.CV); Signal Processing (eess.SP)
备注: This work has been submitted IEEE for potential publication
Abstract:Multi-modal industrial anomaly detection typically relies on separate models for each product category, fundamentally limiting practical scalability. When shifting to a unified paradigm that handles diverse classes simultaneously, detection accuracy often degrades due to inter-class interference and feature manifold confusion. To overcome these challenges, we propose a Unified Reference guided Cross-modal Mapping framework, named Uni-RCM. At its core, we propose a reference guide block to dynamically filter out category-specific noise by introducing a learnable reference feature, which captures the commonalities across different modalities. Besides, an offline residual quantizer is proposed to characterize the normal distribution by multiple cascaded codebooks. Extensive evaluations on the MVTec-3D AD dataset demonstrate the state-of-the-art performance in the challenging multi-class setting and in terms of image-level detection and pixel-level localization.
[CV-111] Comparative evaluation of photogrammetric reconstruction methods and 3D Gaussian Splatting for road surface roughness analysis
链接: https://arxiv.org/abs/2605.29452
作者: Marouane Elmegdar,Teng Xiao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: accepted by RSMIP 2026
Abstract:Image-based 3D reconstruction offers a low-cost alternative to traditional sensor-based techniques for road surface assessment. This study compares four reconstruction pipelines–COLMAP, Meshroom, Metashape, and 3D Gaussian Splatting (3DGS)–to evaluate their ability to estimate road surface roughness from smartphone imagery. All point clouds were processed in CloudCompare using a consistent workflow involving orientation alignment, segmentation, normal estimation, and roughness computation at neighborhood radiuses of 0.2, 0.4, and 0.6 model units. The results show that COLMAP provides the highest sensitivity to micro-texture, while Meshroom yields balanced reconstructions with moderate roughness variation. Metashape produces the smoothest geometry due to its internal filtering, and 3DGS captures visible irregularities but exhibits higher noise and lower density. The comparison demonstrates that open-source pipelines are viable for relative roughness evaluation, offering a practical approach for low-cost pavement monitoring.
[CV-112] How Much Is a Dataset Worth? Scaling Laws the Vendi Score and Matrix Spectral Functions
链接: https://arxiv.org/abs/2605.29448
作者: Jeff A. Bilmes,Gantavya Bhatt,Arnav M. Das
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Information Theory (cs.IT)
备注: 75 pages
Abstract:Neural scaling laws appraise data through dataset size, while the Vendi Score uses quantum entropy to measure dataset value. We show both that common neural-scaling-law objectives and the Vendi Score are submodular. We further show that the Vendi Score is a special case of a broader class of submodular objectives that we call matrix spectral functions. This also includes determinantal (DPP) objectives, as well as many others. We also introduce weakly matrix monotone functions and show how they lead to weakly submodular matrix spectral functions, yielding a broad family of practical objectives for data appraisal. We develop secular-equation-based updates that avoid repeated eigendecompositions during greedy optimization, reducing marginal-gain evaluation for m -dimensional embeddings by an O(m) factor relative to oracle queries. This yields an average empirical speedup of about 35,000x, making direct optimization of the Vendi Score feasible on ImageNet-1K-scale datasets. Thus enabled, we compare how well several objectives predict the value of training subsets for held-out test performance under fixed-size, class-balanced, and fixed training-budget regimes, including the Vendi Score, DPPs, facility location, and three new matrix spectral variants. Across multiple datasets, facility location performs the best. Direct optimization also reveals that, while the Vendi Score is predictive over moderate score ranges, pushing the objective to higher values can make it a poor downstream performance proxy. We also find that uniformly at random fixed-size subsets, both unconstrained and class-balanced, are remarkably concentrated in both appraisal scores and held-out performance. Finally, we show that size, class balance, and training budget do not alone determine data value: even when controlling for these factors, performance ranges smoothly from good to bad.
[CV-113] One Click per Cell Type Suffices: Training-free Group Interaction for Cell Instance Segmentation MICCAI2026
链接: https://arxiv.org/abs/2605.29429
作者: Sanghyun Jo,Seo Jin Lee,Seohyung Hong,Yoorim Gang,Hyeongsub Kim,Hyungseok Seo,Kyungsu Kim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to MICCAI 2026 (Early Accept)
Abstract:Cell instance segmentation models trained on cell-specific datasets suffer severe performance drops on out-of-distribution cell types, while interactive foundation models overcome this through per-instance prompting at a cost that is prohibitively expensive for histopathology images containing hundreds to thousands of densely packed instances. We introduce Group Prompting, a new paradigm that shifts interactive segmentation from per-instance O(N) to per-type O(T) , where a single click per cell type suffices to segment all instances of that type. Our key observation is that the frozen image encoder of the Segment Anything Model (SAM) already clusters same-type cells in its feature space before any prompt is given. Exploiting this property, we propose Chain-of-Prompts (CoP), a training-free framework that recursively expands a single user click by (1) identifying reliable same-type locations through non-parametric gating of multi-scale encoder features, and (2) selecting the most spatially distant reliable point as the next prompt to maximize coverage. On three cell-type-annotated benchmarks, CoP with one click per type retains over 90% of per-instance performance and surpasses fully-supervised methods without any additional training. On four morphologically homogeneous benchmarks, a single click retains over 99%. Project Page: this https URL
[CV-114] ParCo-SDF: Learning Prior-Free Partial-to-Complete Signed Distance Fields of Deformable Objects
链接: https://arxiv.org/abs/2605.29417
作者: Deokmin Hwang,Minseok Song,Daehyung Park
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 6 pages
Abstract:This study addresses the partial-to-complete geometry reconstruction of deformable objects (DOs) from point-cloud observations toward precise DO manipulation. Recent DO reconstruction approaches often adopt implicit neural representations (INRs) to model continuous surfaces as well as capture structural variability. However, these methods typically rely on object-specific shape priors that improve training stability and limit generalization. To figure it out, we introduce ParCo-SDF, a two-stage partial-to-complete signed distance field (SDF) reconstruction framework consisting of temporal geometry encoding followed by FiLM-conditioned SDF prediction. The temporal encoder captures structural similarity across DO sequence, enabling prior-free stable training. FiLM-based conditioning preserves reconstruction expressivity while reducing network complexity. We evaluate the proposed method against a state-of-the-art DO surface reconstruction baseline on a rubber band manipulation dataset, demonstrating robust and high-fidelity reconstruction under severe occlusions.
[CV-115] 3DVLA: Enhancing Vision-Language-Action Models via 3D Spatial and Instance Understanding
链接: https://arxiv.org/abs/2605.29416
作者: Zhongyu Xia,Yousen Tang,Bingqing Wei,Yongtao Wang
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Vision-Language-Action models have achieved remarkable progress in robotic manipulation, yet they suffer from a critical limitation: a lack of 3D scene understanding. This deficiency manifests as three intertwined challenges: weak extraction of 3D spatial positions without enforcing multi-view consistency, inadequate 3D instance understanding, and fragile reasoning under occlusion. Although mature 3D perception methods exist, their direct integration into VLA pipelines is hindered by architectural incompatibility and by heavy reliance on costly instance-level annotations. To address the above challenges, we propose 3DVLA, a plug-and-play framework that injects robust 3D reasoning into pretrained VLAs without requiring extra manual labels or discarding VLM priors. Specifically, 3DVLA tackles the three challenges through: (1) pervasive 3D feature encoding with explicit multi-view consistency constraints across all modalities and a Spatially-Conditioned Geometry Aggregation method, (2) an instance estimation module with high-level instance tokens for 3D instance awareness, and (3) a masked self-supervised 3D encoding branch that retains its predictor for visual token completion to handle occlusions. We integrate 3DVLA with multiple VLA baselines and evaluate on LIBERO-Plus and RoboTwin 2.0. Results show consistent and significant gains in manipulation performance, validating both the effectiveness and plug-and-play compatibility of our approach.
[CV-116] Semantic and Visual Evidence for Efficient Long-Video Reasoning : A Solution for the HD-EPIC VQA Challenge
链接: https://arxiv.org/abs/2605.29402
作者: Yinsong Xu,Wei Jing,Liuxin Zhang,Wanjun Lv,Hui Li
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Understanding long-form egocentric videos remains challenging for multimodal large language models (MLLMs) due to limited context length and insufficient grounding of fine-grained visual details. The recently proposed HD-EPIC benchmark highlights these limitations: even strong long-context models achieve relatively low performance across diverse video question answering tasks. In this paper, we propose a unified framework that decouples long-video reasoning into two complementary forms of evidence: semantic evidence and visual evidence. Semantic evidence captures global procedural structure through a coarse-to-fine extraction pipeline, while object-centric visual evidence preserves fine-grained grounding through bounding boxes and visual embeddings. During inference, we formulate reasoning as a query-conditioned evidence retrieval and integration process, dynamically selecting relevant information from both sources. Our approach achieves competitive performance in the HD-EPIC-VQA Challenge across multiple task categories. More broadly, our results demonstrate that explicitly structuring, retrieving, and integrating semantic and visual evidence is critical for effective long-video understanding with MLLMs.
[CV-117] Orthogonal Negative Guidance in Attention Feature Space for Text-to-Image Generation
链接: https://arxiv.org/abs/2605.29390
作者: Jungmin Ko,Jungwon Park,Jimyeong Kim,Changin Choi,Wonseok Lee,Wonjong Rhee
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint
Abstract:Text-to-image (T2I) models have become increasingly capable of generating high-quality images. Yet, enforcing the explicit absence of a specified object or attribute remains a fundamentally challenging problem. Existing approaches, including prompt negation, post-hoc editing, and negative guidance, remain insufficient for explicit concept suppression, often failing to remove the target concept or degrading overall image quality. To this end, we propose Orthogonal Negative Guidance in attention feature space, a training-free method that operates in the attention output space of MM-DiT-based T2I transformers. Our method orthogonalizes negative-prompt attention features with respect to positive-prompt features and subtracts only the orthogonal component, suppressing unwanted concepts while preserving desired semantics. Experiments on FLUX-dev and FLUX-schnell show that our method achieves favorable trade-offs between concept suppression, prompt alignment, and image quality. In human evaluation, our method outperforms the second-best baseline by 18.78%. We further show that our method supports multi-concept suppression and adjustable concept suppression.
[CV-118] RACER: Persistent Regularization for Robust Multimodal Finetuning ICML2026
链接: https://arxiv.org/abs/2605.29380
作者: Hesam Asadollahzadeh,Feng Liu,Christopher Leckie,Sarah M. Erfani
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: ICML 2026
Abstract:Mainstream strategies for finetuning pretrained multimodal models often degrade out-of-distribution (OOD) robustness, a phenomenon known as catastrophic forgetting. In this paper, we develop a theoretical framework for multimodal contrastive finetuning, yielding closed-form solutions and a geometric decomposition for each strategy. This framework shows that self-distillation is more effective than other regularization approaches to retain the knowledge of the pretrained model. Our analysis reveals a largely overlooked limitation: standard Exponential Moving Average (EMA) teachers, widely used in robust finetuning, suffer from collapse. To solve this, we prove that a Weighted Moving Average (WMA) teacher maintains a persistent regularizing force over finite horizons and yields bias-free convergence in the task subspace while preserving orthogonal knowledge. These insights motivate TRACER (Trajectory-Robust Anchoring for Contrastive Encoder Regularization), which combines contrastive learning with WMA-guided multi-perspective distillation. Extensive experiments on CLIP finetuning demonstrate consistent OOD accuracy and calibration gains across three backbone architectures, and comprehensive ablations confirm that TRACER is both principled and robust to hyperparameter choices. Code is available at [this https URL](this https URL).
[CV-119] DeepFake Forensics AI: A Multi-Modal Detection and Blockchain-Anchored Evidence Management Platform
链接: https://arxiv.org/abs/2605.29353
作者: Naisha Minnah
类目: Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 5 figures, 3 tables
Abstract:The proliferation of AI-generated synthetic media poses a critical threat to the integrity of digital evidence in legal and forensic contexts. Existing deepfake detection systems typically address a single modality and provide no mechanism for tamper-proof evidence preservation. We present DeepFake Forensics AI, a unified platform that detects synthetic media across image, video, and audio modalities, identifies generative architecture fingerprints, and anchors forensic evidence immutably on the Ethereum blockchain. Our system trains four independent neural networks from scratch: an EfficientNet-B4 image detector (AUC = 0.9868), a Bidirectional LSTM video detector (AUC= 0.9628), an ECAPA-TDNN audio detector (EER = 18.63%), and a novel GAN fingerprinting module (accuracy = 99.88%) that identifies the generative architecture behind a fake image. Evidence files are hashed with SHA-256, stored on IPFS via Pinata, and registered on-chain via a Solidity smart contract with role-based access control. The platform provides a React frontend and FastAPI backend suitable for deployment in forensic and legal workflows. To our knowledge, this is the first system to unify multi-modal deepfake detection with blockchain-based chain-of custody management.
[CV-120] DMC-CF: Dynamic Multimodal CounterFactual QA benchmark for Causal Reasoning
链接: https://arxiv.org/abs/2605.29339
作者: Junzhe Zhang,Huixuan Zhang,Guirong Wang,Xingyao Zhang,Pei Liu,Lin Qu,Hu Wei,Xiaojun Wan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:With the rapid advancement of multimodal large language models (MLLMs), models have demonstrated increasingly powerful multimodal capabilities. However, whether MLLMs trained through statistical learning can truly understand the causal relationships underlying the real world remains a key research question. In recent years, numerous multimodal causal reasoning datasets have been proposed. Nevertheless, these datasets are either limited in scale or constructed from synthetic images and videos, cartoon-based content, or other non-realistic multimodal sources. To address these limitations, we collect real-world videos and construct DMC-CF-Static, a large-scale benchmark for multimodal causal counterfactual reasoning. Furthermore, to mitigate issues such as data contamination in traditional static evaluation, we represent causal events using causal graphs and propose the Dynamic Graph Intervention (DGI) framework to build the dynamic evaluation benchmark DMC-CF-Dynamic from DMC-CF-Static. Experimental results on the overall DMC-CF, which includes both static and dynamic evaluation benchmarks, demonstrate that the multimodal causal reasoning capabilities of current multimodal large language models in real-world scenarios still require substantial improvement.
[CV-121] Rethinking FID Through the Geometry of the Reference Dataset ICML2026
链接: https://arxiv.org/abs/2605.29335
作者: Yunghee Lee,Byeonghyun Pak
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 9 pages, 2 figures. Accepted to ICML 2026 Workshop: Combining Theory and Benchmarks
Abstract:Fréchet Inception Distance (FID) is widely used to evaluate image generators, yet lower FID does not always correspond to better sample quality. We show that this mismatch depends in part on the geometry of the reference dataset. In a controlled study across six datasets, distributional density and effective rank significantly explain how FID changes as sample quality improves. Concentrated datasets tend to yield more favorable FID trends, whereas more dispersed datasets can make FID worsen despite better samples. Attribution to precision and recall and ablations with alternative feature spaces and distances support the same conclusion. These results suggest that distributional metrics should be interpreted together with the geometry of the reference dataset for more reliable benchmarking.
[CV-122] EarthShift: a benchmark for measuring robustness to real-world distribution shifts in Earth observation
链接: https://arxiv.org/abs/2605.29330
作者: Kelsey Doerksen,Hannah Kerner
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Current Earth observation benchmarks focus on measuring performance on diverse tasks and applications, typically measuring generalization in-distribution. But when models are deployed, they must generalize to myriad out-of-distribution scenarios, such as new time periods, geographies, scales, and sensors. We introduce EarthShift: the first public testbed for benchmarking robustness across multiple realistic distribution shifts encountered in remote sensing. EarthShift enables users to measure distributional robustness by comparing performance in- and out-of-distribution using datasets from paired datasets from different sources, temporal windows, geographic locations, and sensors. Our experiments on 8 geospatial foundation models (GFMs) and 11 tasks covering 5 shift types show that GFMs consistently perform 15-20% worse out-of-distribution on average regardless of model architecture, size, pre-training or fine-tuning strategy. We show that GFM robustness is similar to that of generic vision foundation models, and even fully-supervised models. This highlights a need for future research to strive for improvements in distributional robustness, not just performance, which can be benchmarked using EarthShift. We release our code and datasets to provide a testbed to guide future work to create foundation models that are robust and reliable in real-world applications. Code and data for EarthShift are available at: this https URL
[CV-123] Multi-Stage VLM Pipeline for Zero-Shot Traffic Accident Understanding CVPR2026
链接: https://arxiv.org/abs/2605.29325
作者: Fumiya Tatematsu,Fumihiko Takahashi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the AUTOPILOT Workshop, CVPR 2026 (non-archival). Workshop Paper ID 13. Code: this https URL
Abstract:We present the 1st-place solution to the ACCIDENT challenge at the CVPR 2026 AUTOPILOT Workshop, which asks for zero-shot prediction of accident timing, impact centroid, and collision type from CCTV footage. On a frozen Qwen3-VL-32B-Instruct checkpoint we build a three-stage pipeline (full-video joint prediction, time refinement, and single-frame grounding of the impact centroid), run the same pipeline a second time on a 235B Mixture-of-Experts sibling, blend the two outputs 9:1, and finally snap each predicted point onto the nearest vehicle detection. The final system reaches Public LB 0.55469 / Private LB 0.57080, roughly +0.21 over the strongest host baseline (Molmo-7B, 0.358) and wins the challenge. We ablate each component, report the negative results that shaped the final design, and release the code at this https URL.
[CV-124] FreeForm: Reduced-Order Deformable Simulation from Particle-Based Skinning Eigenmodes CVPR2026
链接: https://arxiv.org/abs/2605.29318
作者: Donglai Xiang,Vismay Modi,Rishit Dagli,Ty Trusty,Gilles Daviet,Anka He Chen,Nicholas Sharp,David I.W. Levin
类目: Graphics (cs.GR); Computer Vision and Pattern Recognition (cs.CV)
备注: CVPR 2026, project website: this https URL
Abstract:We present a novel formulation for mesh-free, reduced-order simulation of deformable hyperelastic objects. Existing work in reduced-order elastodynamic simulation represents the input geometry by either meshes, which can be difficult to obtain due to challenges in scanning and triangulating complex shapes, or by neural fields that require per-shape optimization. We propose to adopt a Reproducing Kernel Particle Method (RKPM) representation, which enables the construction of reduced-order skinning weights by solving a generalized eigensystem on the Hessian matrix of the elastic energy. We demonstrate that this formulation not only leads to a 40x training speedup compared with the per-shape optimization of neural fields, but also achieves lower simulation error when evaluated against the converged results of finite element method. We show our simulation results on a wide variety of objects in different representations including meshes and Gaussian splats, as well as the application of our method in the downstream task of robot simulation.
[CV-125] CapTalk: Text-Guided Stylization and Speech-Driven 3D Head Animation
链接: https://arxiv.org/abs/2605.29316
作者: Xuangeng Chu,Yuan Gan,Ziteng Cui,Shuhong Liu,Jian Wang,Bing Zhou,Tatsuya Harada
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Audio-driven 3D facial animation aims to generate synchronized lip movements and vivid facial expressions from arbitrary audio clips. While existing methods can produce synchronized lip motions, they often rely on predefined identity or style latent features, which limits users’ ability to freely control speaking styles. Moreover, applying a fixed style or identity to an entire audio segment typically results in facial animation styles that do not adapt to the emotional content of the audio. To address these challenges, we revisit the entanglement between style and emotion, construct a large-scale dataset with textual descriptions of both style and emotion, and propose a novel talking head generation framework that enables separate control over style and emotion. Our model takes as input both textual descriptions of speaking style and character emotion, as well as the driving audio stream, enabling real-time generation of highly synchronized lip movements and facial expressions that match the provided descriptions. Furthermore, our model supports dynamic emotion control during inference, allowing it to handle scenarios where the target emotion changes throughout the speech.
[CV-126] ViASNet: A Video Ad Saliency Network for Predicting Dynamic Saliency and Viewer Engagement
链接: https://arxiv.org/abs/2605.29302
作者: Jianping Ye,Michel Wedel
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The digital media landscape has seen a pervasive shift toward short-form video advertising on TV, social media and e-commerce platforms. The present study focuses on deep saliency prediction for short-form video advertising. Deep saliency models have been used to generate predictions of human eye fixation patterns with the purpose of enhancing user interaction with digital technology and optimizing its design. For video ads, dynamic saliency maps capture where and when viewers are looking, revealing why video ads are effective, and how their content should be optimized. We develop and test a new deep dynamic saliency prediction model called ViASNet (Video Ad Saliency Network), which has an architecture founded on the 3D U-Net, and accommodates the influence of audio and the semantic meaning of scenes. We assess the model’s performance on 151 video ads, each seen by about 20 viewers wile their eye movements were tracked, and explore the critical factors influencing model performance through ablation experiments. We calculate the entropy of the predicted saliency maps frame-by-frame as a diagnostic tool to identify ads and scenes that fail to engage viewers, and illustrate its use on test data of 15 unseen ads. Our study reveals that ad design and testing can be sped up considerably through automated systems built on deep saliency models such as ViASNet.
[CV-127] Pocket-Dentist: On-Device Dental Image Understanding via Efficient Multimodal Large Language Models
链接: https://arxiv.org/abs/2605.29299
作者: Kai Bian,Xucheng Guo,Bin Chen,Lingyan Ruan,Yiran Shen,Ting Dang,Hong Jia
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Evaluations of dental vision-language models remain fragmented across datasets, task definitions and metrics, and often ignore their computational cost. This limits their widespread deployment for dental screening outside specialist centres, where timely inference, limited hardware, and local handling of patient images are vital for practical, privacy-preserving clinical prescreening. Here we present Pocket-Dentist, an efficiency-aware benchmark for dental multimodal question answering that brings together three datasets spanning approximately 1,159 patients, five task types and seven metrics. Across typical 14 VLMs, our results reveals an interesting observation: compact VLMs (e.g., 2B-parameter models) outperform larger VLMs in accuracy while requiring substantially lower computational costs in dental image understanding. Deployed locally on an iPhone 17 Pro, our finetuned compact VLM Pocket-Dentist-2B processed each sample in 4.31 s, reducing latency by 4.9-fold and memory use by 2.3-fold compared with a 7B baseline.
[CV-128] urbulence-Robust Dynamic Object Segmentation with Multi-Signal Priors and SAM2 Refinement
链接: https://arxiv.org/abs/2605.29292
作者: Bolian Peng,Ying Tang,Xu Liu,Long Sun,Xiaoqiang Lu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:This technical report presents our solution for the CVPR 2026 UG2+ Challenge Track 3: Dynamic Object Segmentation in Turbulence (DOST). We design a training-free multi-signal segmentation pipeline that combines pretrained motion estimation, self-supervised semantic priors, background anomaly modeling, manually calibrated proposal fusion, and SAM2-based mask refinement. The method uses RAFT for dense motion responses, DINOv2 for semantic objectness priors, ViBe for training-free background modeling, and pretrained SAM2 for box-prompt mask refinement. Instead of optimizing an end-to-end segmentation network, our system operates entirely in inference mode. This design is suitable for the DOST setting, where severe atmospheric turbulence produces pseudo-motion, blur, and intermittent target visibility, making a single motion cue unreliable. The final submitted masks are evaluated by the official leaderboard, which reports 0.425041 mIoU and 0.457206 mDice. Since no task-specific model training or fine-tuning is performed, stronger learned temporal association, adaptive proposal selection, or task-specific adaptation may further improve the system. Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2605.29292 [cs.CV] (or arXiv:2605.29292v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2605.29292 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-129] Deep Psychovisual Image Representations
链接: https://arxiv.org/abs/2605.29260
作者: Wendi Ma,Aryaman Sharma,Wei Dai,Shekhar S. Chandra
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Psychovisual models suggest human vision decouples low-level feature extraction from higher cognition by first forming intermediate abstractions. In contrast, deep learning-based vision models routinely extract and aggregate features using homogeneous stacks of spatial layers, rendering their decision-making processes opaque. In this paper, we propose Deep Visual Coding, a learned frequency-domain representation inspired by 1990s image codes that quantised perceptually salient frequencies, which together with complex-valued image representations produces psychovisual-style abstractions. This approach enables the first psychovisual-based deep learning framework, utilizing data-driven spectral filters that learn to encode task-relevant semantic structures within distinct frequency sub-bands. Salience analyses reveal that our psychovisual models extract highly interpretable object parts compared to the amorphous regions produced by regular Convolutional Neural Networks (CNNs). Furthermore, we find that our models are less depth dependent than CNNs for model scaling, since our complex-valued representations and learned abstractions subsume the role of the deep spatial layers. Together, these findings demonstrate that psychovisual coding provides a promising path toward more efficient and transparent vision models.
[CV-130] oward Ethical Facial Age Estimation: A Generalized Zero-Shot Benchmark Without Training on Childrens Data
链接: https://arxiv.org/abs/2605.29230
作者: Caio Petrucci,Leo Sampaio Ferraz Ribeiro,Sandra Avila
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 12 pages; 3 figures; 5 tables
Abstract:Age estimation from facial images typically relies on training data that includes images of minors, a practice that raises serious ethical, legal, and privacy concerns. In this work, we propose a generalized zero-shot benchmark for facial age estimation that explicitly excludes children’s data during training while still assessing model performance on younger populations. We revisit six widely used datasets and introduce standardized splits with strict age-group separation: samples aged 18-59 for training, validation, and testing; samples under 18 reserved exclusively for zero-shot evaluation; and samples 60+ as an unseen validation set for model selection under distribution shift. For datasets with identity annotations, subject-exclusive splits prevent identity leakage and better reflect real-world deployment conditions. Evaluating nine state-of-the-art age estimation methods under this protocol reveals that all evaluated methods consistently fail to generalize to unseen age groups, suffering substantial performance degradation – on average 46.4%, and up to 52.8% – relative to the supervised baseline. Moreover, models do not simply degrade: they systematically anchor predictions for unseen ages to nearby seen classes, a manifestation of the well-known seen-class bias in generalized zero-shot learning. By formalizing age estimation without children’s data as a generalized zero-shot benchmark on existing datasets, this work highlights a critical gap between current modeling practices and real-world ethical constraints. Our benchmark provides a principled basis for evaluating models under restricted data regimes and encourages the development of methods that are robust to distribution shift and aligned with responsible data use.
[CV-131] An Approach for Thyroid Nodule Analysis Using Thermographic Images
链接: https://arxiv.org/abs/2605.29221
作者: J. R. González,É. O. Rodrigues,C. P. Damião,C. A. P. Fontes,A. C. Silva,A. C. Paiva,H. Li,C. Du,A. Conci
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Thyroid cancer is said to be the second most common type of cancer in female individuals and the third in males by 2030, according to projections. In general, detecting cancer in its early stages improves the chance of survival of the individual. Thermography is a diagnostic tool that has been increasingly used to detect cancer and abnormalities, including that of thyroid. Various methods to segment and detect hot regions in thermograms and, consequently, to detect suspicious tissues present in these images have been proposed. It is well known that medical diagnosis yields a great deal of information. Thus, physicians have to comprehensively analyse and evaluate this information in a short period of time, which is infeasible in most cases. In this work, we perform a general review of thermography , focusing on the thyroid analysis. We propose protocols for image acquisiton and an autonomous registration for thyroid images. We also perform analyses of the image data, which include feature extraction, image processing, and a possible approach for classification of healthy or unhealthy patients. In summary, this work presents a pilot project for detection of tumors in our university hospital, which is part of an effort to support preventive medical actions in our endocrinology department. Under some future adjustments, this project will be submitted for approval by the ethics and research committee of Hospital Universitário Antonio Pedro at Universidade Federal Fluminense (HUAP-UFF) and to the Brazilian Ministry of Health Ethical committee under the name: Evaluation of the importance of thermography to aid diagnosis of thyroid nodules of patients in HUAP-UFF (in Portuguese: Avaliação da importância da termografia no auxílio à investigação diagnóstica de nódulos tireoidianos em pacientes acompanhados no HUAP-UFF).
[CV-132] Motion-guided sparse correction enables expert-quality point tracking across diverse microscopy regimes
链接: https://arxiv.org/abs/2605.29220
作者: Leonidas Zimianitis,Pasindu Thenahandi,Kai Buckhalter,Dineth Jayakody,Julian O. Kimura,Xinyue Liang,Karen Cunningham,Azeem Ahmad,Balpreet S. Ahluwalia,Sampath Jayarathna,Nikos Chrisochoides,Brandon Weissbourd,Dushan N. Wadduwage
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Tracking the dynamics of non-canonical biological systems in microscopy videos remains a persistent challenge. Both classical and learning-based trackers depend on expert-reviewed data to be evaluated and adapted, yet exhaustive manual annotation rarely scales to the videos where these tools are needed most. We developed RIPPLE (Refinement Interpolation Platform for Point Location Estimation), which recasts annotation as sparse correction: a user clicks a starting point, RIPPLE proposes a full trajectory, and the user intervenes only where the trajectory drifts. We tested RIPPLE on five challenging microscopy datasets from our laboratories, four from the transparent jellyfish Clytia hemisphaerica and one tracking landmarks on rapidly moving sperm. Across these, RIPPLE matched the quality of exhaustive manual annotation while reducing manual clicks by 3 to 25 times across datasets. RIPPLE thereby fills a missing layer between manual annotation and fully automated tracking, enabling immediate quantification of biological dynamics, method benchmarking, and the production of the gold-standard data needed to adapt future automated microscopy trackers.
[CV-133] SalsaAgent : A multimodal embodied language model for interactive dance generation
链接: https://arxiv.org/abs/2605.29219
作者: Payam Jome Yazdian,Zoe Stanley,Angelica Lim
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Interaction between humanoids involves bidirectional and nonverbal reactivity, coordination and synchrony. Toward socially aware robots and interactive virtual agents, we present SalsaAgent, a language model that generates expressive, full-body salsa dance motions in reaction to a human leader and against a contextual music backdrop. We formulate interaction as nonverbal motion token passing, extending the vocabulary of a large language model (LLM) to process discrete motion tokens, pairwise relation tokens, and audio. Our contributions include new tokens for full-body and motion relations, LLM fine-tuning using automatically derived text descriptions of skeleton dynamics for token grounding, and a two-stage token-to-diffusion pipeline. Subjective and objective evaluations demonstrate the effectiveness of our approach in terms of motion quality, music and partner coordination, and consistent two-person spatial behavior, with significant improvements over baselines.
[CV-134] owards the automated segmentation of epicardial and mediastinal fats: A multi-manufacturer approach using intersubject registration and random forest
链接: https://arxiv.org/abs/2605.29217
作者: É. O. Rodrigues,A. Conci,F. F. C. Morais,M. G. Pérez
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The amount of fat on the surroundings of the heart is correlated to several health risk factors such as carotid stiffness, coronary artery calcification, atrial fibrillation, atherosclerosis, cancer incidence and others. Furthermore, the cardiac fat varies unrelated to the overall fat of the subject, and, therefore, it reinforces the quantitative analysis of these adipose tissues as being essential. Clinical decision support systems are computer programs capable of evaluating information and providing a corresponding diagnosis or data to complement the physicists’ analyses. The aim of this work is to propose a method capable of fully automatically segmenting two types of cardiac adipose tissues that stand apart from each other by the pericardium on CT images obtained by the standard acquisition protocol used for coronary calcium scoring. Much effort was devoted to promote minimal user intervention and ease of reproducibility. The methodology proposed in this work consists of a registration, which will roughly adjust input images to a standard, an extraction of features related to pixels and their surrounding area and a segmentation step based on data mining classification algorithms that define if an incoming pixel is of a certain type. Experimentations showed that the achieved mean accuracy for the epicardial and mediastinal fats was 98.4% with a mean true positive rate of 96.2%. In average, the Dice similarity index was equal to 96.8%.
[CV-135] Guidance Contrastive Token Credit Assignment for Discrete Policy Optimization
链接: https://arxiv.org/abs/2605.29198
作者: Shufan Li,Konstantinos Kallidromitis,Akash Gokul Yusuke Kato,Kazuki Kozuka,Aditya Grover
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages, 11 figures
Abstract:Group-advantage-based reinforcement learning methods, such as GRPO and DAPO, have demonstrated strong performance across diverse domains, including mathematical reasoning and text-to-image generation. However, their reliance on sample-level rewards introduces a key limitation as uniform credit assignment across all tokens fails to capture fine-grained, token-level contributions. To address this issue, we propose Guidance Contrastive Policy Optimization (GCPO), a novel algorithm that enables per-token credit assignment by contrasting model predictions under positive and negative prompts. Rather than uniformly broadcasting sample-level advantages, GCPO assigns token-level advantages proportional to the difference between these contrastive predictions, allowing more precise and informative learning signals. Empirically, we find that GCPO emphasizes semantically relevant regions such as visual areas aligned with textual prompts in text-to-image generation, and critical keywords within reasoning traces for chain-of-thought tasks. Through extensive experiments, GCPO consistently outperforms GRPO and DAPO baselines on both text-to-image generation and chain-of-thought reasoning benchmarks, demonstrating its effectiveness as a general and scalable optimization strategy for discrete policy learning.
[CV-136] Eulerian Gaussian Splatting using Hashed Probability Pyramids CVPR2026
链接: https://arxiv.org/abs/2605.29136
作者: Mia Gaia Polansky,George Kopanas,Stephan Garbin,Todd Zickler,Dor Verbin
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: CVPR 2026. Project Page: this https URL
Abstract:We introduce a probabilistic splat-based radiance field framework that retains the fast rasterization and test-time efficiency of 3D Gaussian Splatting (3DGS) while replacing heuristic primitive manipulation with gradient-based optimization of a volumetric probability density. Rather than relocating, splitting, or culling Gaussians via hand-tuned densification (e.g., ADC), we treat primitive locations as samples drawn from a persistent, learnable density. We instantiate this density using a novel, memory-efficient multi-scale hierarchical grid that enables end-to-end gradient-based optimization. To stabilize the optimization, we derive an unbiased gradient estimator with control variates that markedly reduces variance. By allowing probability mass to flow to where the loss demands, our framework eliminates brittle priors and naturally explores the volume, achieving state-of-the-art reconstruction quality on mip-NeRF 360 while preserving 3DGS-level rendering speed.
[CV-137] Robust Cross-Domain Generalization Using Unlabeled Target Data with Source-Domain Supervision
链接: https://arxiv.org/abs/2605.29122
作者: Yuyue Zhou,Shrimanti Ghosh,Michael(Kai Yue)Xie,Justin JY Kim,Jessica Knight,Steel McDonald,Vincent Man,Jacob L. Jaremko,Abhilash Hareendranathan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:It is often desirable to generalize medical imaging AI models trained with dense annotations to data acquired from different ultrasound scanners or clinical sites; however, retraining these models with new annotations is often difficult and costly. We examine this challenge in pediatric wrist fracture assessment using point-of-care ultrasound (POCUS), where fractures are common and can be effectively triaged via ultrasound. AI has shown radiologist-level performance for fracture detection, often aided by high-quality bony structure segmentation. However, due to significant domain shifts, models perform poorly on data from other centers or probes, and obtaining segmentation labels across devices is impractical due to manual annotation effort and data privacy concerns. To address this, we propose a target-informed self-supervised pretraining and model-ensemble strategy. Specifically, our approach combines masked image modeling (MIM) and contrastive learning to learn target-domain structural representations without labels, and introduces a confidence-aware infusion head to adaptively integrate predictions. The source dataset, collected with a Philips Lumify probe, contained dense labels, while the target dataset, acquired with a TeleMED portable probe, was unlabeled. The datasets were kept strictly separate throughout the entire process. Our method used labeled source data for supervised training and leveraged target-domain pretraining to improve generalization. On 318 images from 62 pediatric POCUS videos, this approach significantly improved cross-device performance, achieving over 6% Dice improvement on the target domain versus the baseline. These results demonstrate a label-efficient and privacy-preserving approach for cross-device-robust ultrasound AI, offering a framework that can be extended to multi-center studies or federated learning setups.
[CV-138] Seeing through boxes: Non-Line-of-Sight 3D Reconstruction from Radar Signals CVPR
链接: https://arxiv.org/abs/2605.29098
作者: Jiachen Lu,Hailan Shanbhag,Haitham Al Hassanieh
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026
Abstract:Reconstructing object geometry from radio frequency (RF) signals is fundamentally challenging due to the lensless imaging nature of RF sensing, which leads to low spatial resolution and high noise. Unlike light signals, RF signals can penetrate occlusions and thus capture information about hidden scenes. Existing Non-Line-of-Sight (NLoS) 3D neural reconstruction methods can recover coarse surfaces inside enclosed environments but often suffer from unstable optimization, noisy surface geometry, and surface ambiguity, failing to produce accurate zero-level sets from the signed distance field (SDF). These limitations largely stem from neglecting the role of Line-of-Sight (LoS) geometry outside the enclosed region, which provides valuable physical constraints for modeling signal propagation. In this paper, we introduce a Unified LoS and NLoS neural geometry reconstruction framework GeRaF 2.0 that leverages the outside LoS geometry to model and guide RF propagation from the LoS region into the NLoS region. By integrating visual LoS priors into the neural field formulation, GeRaF 2.0 achieves stable training and physically consistent reconstruction of both visible and hidden geometry, setting a new state-of-the-art in RF-based geometry reconstruction.
[CV-139] GeRaF: Neural Geometry Reconstruction from Radio Frequency Signals NEURIPS2025
链接: https://arxiv.org/abs/2605.29097
作者: Jiachen Lu,Hailan Shanbhag,Haitham Al Hassanieh
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at NeurIPS 2025 (Spotlight)
Abstract:GeRaF is the first method to use neural implicit learning for near-range 3D geometry reconstruction from radio frequency (RF) signals. Unlike RGB or LiDAR-based methods, RF sensing can see through occlusion but suffers from low resolution and noise due to its lensless imaging nature. While lenses in RGB imaging constrain sampling to 1D rays, RF signals propagate through the entire space, introducing significant noise and leading to cubic complexity in volumetric rendering. Moreover, RF signals interact with surfaces via specular reflections, requiring fundamentally different modeling. To address these challenges, GeRaF (1) introduces filter-based rendering to suppress irrelevant signals, (2) implements a physics-based RF volumetric rendering pipeline, and (3) proposes a novel lensless sampling and lensless alpha blending strategy that makes full-space sampling feasible during training. By learning signed distance functions, reflectiveness, and signal power through MLPs and trainable parameters, GeRaF takes the first step towards reconstructing millimeter-level geometry from RF signals in real-world settings.
[CV-140] Lightweight Complementary-Cue Fusion for Robust Video Face Forgery Detection
链接: https://arxiv.org/abs/2605.29092
作者: Sunghwan Baek,Tariq Anwaar,Karanveer Singh,Rita Singh
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Multimedia (cs.MM)
备注: 13 pages, 6 figures, 3 tables
Abstract:Current face video forgery detectors use wide or dual-stream backbones. We show that a single, lightweight fusion of two handcrafted cues can achieve higher accuracy with a much smaller model. Based on the Xception baseline model (21.9 million parameters), we build two detectors: LFWS, which adds a 1x1 convolution to combine a low-frequency Wavelet-Denoised Feature (WDF) with a phase-spectrum channel derived from Spatial-Phase Shallow Learning (SPSL), and LFWL, which merges WDF with Local Binary Patterns (LBP) in the same way. This extra module adds only 292 parameters, keeping the total at 21.9 million, smaller than F3Net (22.5 million) and less than half the size of SRM (55.3 million). Even with this minimal overhead, the fused models increase the average area under the curve (AUC) from 74.8% to 78.6% on FaceForensics++ and from 70.5% to 74.9% on DFDC-Preview, gains of 3.8% and 4.4% over the Xception baseline. They also consistently outperform F3Net, SRM, and SPSL in eight public benchmarks, without extra data or test-time augmentation. These results show that carefully paired, handcrafted features, combined through the lightweight fusion block, can provide competitive robustness at a significantly lower cost than comparable frequency-based detectors. Our findings suggest a need to reevaluate scale-driven design choices in face video forgery detection.
[CV-141] OISD: On-Policy Internal Self-Distillation of Language Models
链接: https://arxiv.org/abs/2605.29089
作者: Xinyu Liu,Darryl Cherian Jacob,Yang Zhou,Jindong Wang,Pan He
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: Under Review for Publication
Abstract:Recent reinforcement learning (RL) post-training approaches primarily optimize the final output policy using sparse outcome-level rewards, while largely overlooking predictive signals encoded in intermediate representations. In this paper, we introduce a new paradigm called on-policy internal self-distillation and propose the OISD framework, which improves reasoning by transferring on-policy predictive signals from the final layer to intermediate representations. During rollout and Group Relative Policy Optimization (GRPO) optimization, the final layer acts as both the policy and a detached internal teacher for selected intermediate layers, which are guided to align with it through two complementary mechanisms: logit alignment, which transfers high-level reasoning behaviors (how to think), and attention alignment, which enforces consistent attention patterns (where to look) from the final layer to the selected intermediate layer, both without requiring external privileged information. Our OISD, together with GRPO, employs signed advantage-weighted Jensen–Shannon alignment to distill informative intermediate representations while preserving policy consistency under a unified acting policy. Experimental results demonstrate the effectiveness of OISD, with substantial and consistent improvements over strong reasoning RL baselines across four mathematical reasoning tasks. The code will be released at this https URL
[CV-142] A Deep Learning Iterative Framework for Sentinel-1 Stripmap Enhancement Based on Azimuth Doppler Decomposition CVPR2026
链接: https://arxiv.org/abs/2605.29088
作者: Juan Francisco Amieva,Christian Ayala,Roberto Del Prete,Mikel Galar
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the AI4Space Workshop, CVPR 2026
Abstract:Synthetic Aperture Radar (SAR) imagery enables all-weather, day-and-night Earth observation; however, it remains difficult to interpret due to speckle noise and other intrinsic imaging artifacts. Sentinel-1 (S1) constitutes one of the most widely used spaceborne SAR missions, offering systematic global coverage, high temporal resolution, dual-polarization imaging, and free data availability. Among S1 modes, Stripmap (SM) provides the highest resolution, yet speckle noise and spatial constraints often hinder applications requiring finer spatial detail. This motivates the need for effective image enhancement strategies. In this work, we propose a self-supervised enhancement framework for S1 SM imagery based on azimuth subaperture decomposition. The method exploits the physical consistency between subaperture reconstructions and the corresponding full-aperture image to generate paired training data without external sensors, simulated ground truth, or multi-temporal stacks. The proposed framework integrates single- and multi-frame learning and incorporates an iterative inference scheme that progressively refines image quality. Experiments on real S1 SM data show that the proposed approach consistently outperforms the widely adopted self-supervised deep learning baseline MERLIN, in terms of PSNR and SSIM, while MERLIN attains higher ENL, highlighting a trade-off between structural fidelity and speckle smoothing. Overall, the results demonstrate that subaperture-based supervision provides a physically grounded, reproducible, and operationally viable approach for SAR image enhancement using S1 data. It is worth noting that the proposed approach can be extended to other SAR platforms, polarizations, and acquisition modes.
[CV-143] Embodied3DBench: Benchmarking Low-Level Embodied Spatial Intelligence of Vision Language Models
链接: https://arxiv.org/abs/2605.29074
作者: Jiyao Zhang,Mingxu Zhang,Yitong Peng,Haoxuan Liu,Chenshuo Wang,Yuxing Long,Haoyang Huang,Dongjiang Li,Nan Duan,Hui Shen,Hao Dong
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
备注:
Abstract:Are current Vision Language Models (VLMs) ready to comprehend and reason about complex embodied interactions in 3D environments? We introduce Embodied3DBench, a robot-centric benchmark targeting low-level spatial intelligence in embodied 3D environments. To systematically evaluate these foundational perceptual capabilities, the benchmark includes 6 task categories divided into two core groups: Spatial Structural Understanding (Grounding, Spatial Relation Prediction, and Multi-view Correspondence) and Interaction-Oriented Perception (Affordance Prediction, Grasp Point Prediction, and Trajectory Prediction). The benchmark spans 12 subcategories and contains over 21k high-quality question-answer pairs. We evaluate 13 state-of-the-art models, and the results show that while current models exhibit relatively strong high-level spatial reasoning, such as understanding object-to-object positional relations, they remain fragile in interaction-oriented perception, highlighting a significant lack of robust 3D-aware interaction priors. To actively bridge this capability gap revealed by our benchmark, we further synthesize a large-scale training dataset comprising 1.3M QA pairs. Notably, fine-tuning on this dataset yields significant improvements in low-level spatial intelligence. Ultimately, Embodied3DBench fills a critical gap by providing both a systematic evaluation framework and a scalable data solution, setting a clear target for the development of interaction-aware multimodal systems.
[CV-144] rajectory Constraints for Imaging Inverse Problems
链接: https://arxiv.org/abs/2605.29012
作者: Chaoyan Huang,Haijie Yuan,Saiprasad Ravishankar
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 20 pages, 10 figures
Abstract:Diffusion-based and iterative methods have become effective tools for solving imaging inverse problems. Their reconstruction process naturally forms a trajectory of intermediate estimates. Although these intermediate estimates define a reconstruction trajectory, most methods do not explicitly regularize the transitions between consecutive states. To address this limitation, we introduce TRACE, a training-free TRAjectory-Constrained rEconstruction framework that stabilizes the reconstruction path by coupling adjacent states along the trajectory. This gives a trajectory-level model that can be interpreted as a sequence of proximal updates. Since the exact proximal update is generally intractable, we approximate it with a neural mapping. This yields a diffusion-like reconstruction process with an explicit coupling between neighboring states. We provide a stability analysis showing that temporal coupling bounds trajectory variation and that this control is preserved under untrained network updates. Experiments on linear and nonlinear image reconstruction tasks show that TRACE improves reconstruction quality. Trajectory-level analyses and ablations confirm that temporal coupling directly affects state transitions along the reconstruction path.
[CV-145] Auditing Training-Free 3D Shape Retrieval with Diffused Geodesic Moments
链接: https://arxiv.org/abs/2605.29004
作者: Zhicheng Du,Changyue Liu,Wenji Xi,Zhaotian Xie,Zhuo Deng,Ziheng Zhang,Yang Liu,Lan Ma
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注:
Abstract:Reported retrieval scores for training-free shape descriptors conflate local signal design, normalization, aggregation, codebook fitting, and metric choices, making isolated component evaluation difficult. This paper reframes descriptor evaluation as a \em protocol audit. We introduce Diffused Geodesic Moments (DGM), a seed-conditioned descriptor that computes sparse implicit heat responses, converts them to distance-like fields, and summarizes each vertex by low-order moments across seeds and scales. DGM is used both as a practical non-spectral baseline and as an instrument for isolating protocol effects. On the registered FAUST benchmark split (FAUST-Reg) and the TOSCA shape collection, aggregation-matched experiments show that an independent Geometric Moment Shape Descriptor baseline built on Heat Kernel Signature features (GMSD-HKS) obtains the highest scores in this implementation ( 0.621/0.820 and 0.865/0.963 mean average precision (mAP)/top-1), Wave Kernel Signature (WKS) remains a strong classical signal, and DGM is useful mainly when sparse solves, non-spectral deployment, or symmetry-informative seed frames are priorities. The broader finding is methodological: the input field and aggregation protocol can dominate the moment formula. The paper contributes a reproducible protocol-cascade analysis, a cross-shape alignment diagnostic for functional-map compatibility, and concrete recommendations for designing and reporting training-free shape descriptors.
[CV-146] GAP3D: Generative Alignment of VLM Latents to Patch-Level Embeddings for 3D Generation
链接: https://arxiv.org/abs/2605.28995
作者: Polytimi Anna Gkotsi,Andrii Zadaianchuk,Mohammad Mahdi Derakhshani
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent approaches integrating vision-language models (VLMs) as prompt encoders for generative model conditioning typically rely on expensive end-to-end training or map features to compressed representations, discarding the dense spatial structure required for geometry-aware tasks like 3D asset generation. To address this, we propose GAP3D, a modular, diffusion-based approach that aligns VLM-generated latents directly to the complete, patch-level feature space of a pre-trained image encoder, enabling a frozen downstream generative model to utilize a VLM as prompt encoder while maintaining a spatially structured conditioning signal. Evaluated on 3D asset generation, our method bypasses the need for large-scale 3D data by training mainly on general-domain image-text pairs. It also exhibits emergent zero-shot capabilities for multimodal prompts, despite being trained exclusively on text input. Finally, while currently prioritizing high-level semantics over fine-grained detail, GAP3D demonstrates that the representation gap between VLM and image-encoder feature spaces can be partially bridged through diffusion-based alignment, taking the first steps towards a modular integration of foundation models through generative alignment to dense embedding spaces.
[CV-147] Resolving Endpoint Underfitting in Diffusion Bridges via Noise Alignment CVPR2026
链接: https://arxiv.org/abs/2605.28962
作者: Yurong Gao,Zicheng Zhang,Congying Han,Tiande Guo,Xinmin Qiu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by CVPR2026
Abstract:Diffusion bridge models offer a powerful framework for connecting two data distributions, such as in image restoration and translation. Many existing methods learn this bridge by mimicking the score-matching formulation of standard diffusion models. In this work, we find that this way leads to an anomalous underfitting phenomenon near the target endpoint, as the process approaches the target distribution ( t \to 0 ). This underfitting, characterized by significant drift in the predicted variance and direction, results from an excessively large discrepancy in noise levels between the network’s input and its regression this http URL resolve this issue, we propose the Noise-Aligned Diffusion Bridge (NADB).Our approach reformulates the diffusion bridge by first employing a mean network to provide a cleaner conditional target, and then introducing a novel, noise-aligned mapping relationship. This new formulation resolves the noise mismatch and corrects the underfitting near the target endpoint. Experimental validation across multiple image restoration and image translation tasks demonstrates the effectiveness of our approach. Code is available at this https URL.
[CV-148] Visual Spatial Learning: Single-Field Spatial Interpolation Using Convolutional Neural Networks
链接: https://arxiv.org/abs/2605.30167
作者: Daniel Tinoco,Raquel Menezes,Carlos Baquero,Alexandra Silva
类目: Machine Learning (stat.ML); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Applications (stat.AP)
备注: 53 pages, 10 figures
Abstract:Predicting a complete spatially correlated field from sparse observations is a fundamental challenge in spatial statistics and environmental modelling. Classical interpolation methods such as Kriging rely on Gaussian process assumptions and variography, which can limit their effectiveness in non-stationary settings and require substantial domain expertise. In this work, we leverage an architecture based on convolutional neural networks (CNNs) for spatial interpolation that is trained and applied on a single partially observed field, without access to external data or prior fields. The model is supervised directly on the observed locations and learns to predict values at unobserved points on the user defined grid. Unlike Kriging, our method does not require explicit covariance modelling or variogram estimation, and it can flexibly capture local spatial patterns in a data-driven manner. This work demonstrates the potential of CNNs for single-instance spatial interpolation under sparse supervision, offering a practical alternative to classical geostatistical methods, and extending the use of CNNs to a new problem domain.
[CV-149] Subcortical Shape Variations and Their Associations with Cognition Across the 8th Decade of Life. A Study in the Lothian Birth Cohort 1936
链接: https://arxiv.org/abs/2605.29703
作者: Maria del C. Valdes-Hernandez,Wonjung Park,Joanna Moodie,Susana Muñoz Maniega,Janie Corley,Fraser N. Sneden,Mark E. Bastin,Joanna M. Wardlaw,Simon R. Cox,Jinah Park
类目: Neurons and Cognition (q-bio.NC); Computer Vision and Pattern Recognition (cs.CV); Tissues and Organs (q-bio.TO)
备注: 34 pages
Abstract:The study of brain morphology changes in normal individuals may capture aspects of functionally-relevant brain aging not fully indicated by gross volumetry. Despite the important role of subcortical brain structures in cognition, the associations between their morphological trajectories and cognitive changes in aging have not been documented. We use neuroimaging, demographic, and cognitive data from a large longitudinal study of cognitive aging, the Lothian Birth Cohort 1936, to explore shape changes in subcortical brain structures of community-dwelling individuals across their 8th decade of life. We investigate the association of these changes with cognitive aging using ANCOVA and mixed linear model analyses. Subcortical shape changes were heterogeneous, with varied atrophy patterns across whole period. The hippocampus and the ventral DC experienced varied morphological deformations (from its baseline point) different in left and right hemispheres, while the thalami and globus pallidi shapes, for example, experienced a more uniform volume contraction, nearly symmetrical throughout different timelines. Changes in general cognition were mainly associated with inwards and outwards vertex displacements between the time-points.
[CV-150] Constructing efficient channels for ideal observers using the conjugate gradient method
链接: https://arxiv.org/abs/2605.29415
作者: Weimin Zhou
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Signal Processing (eess.SP); Machine Learning (stat.ML)
备注: Submitted to the Journal of Medical Imaging (JMI) Special Issue Honoring Dr. Harrison H. Barrett
Abstract:Task-based assessment of image quality (IQ) is critically important for the design and optimization of medical imaging systems. Ideal observers, including the Bayesian Ideal Observer (IO) and the ideal linear observer, i.e., the Hotelling observer (HO), provide objective figures of merit (FOMs) that quantify system performance on signal detection tasks. However, the application of ideal observers to high-dimensional image data is often computationally intractable. Channel mechanisms provide an effective framework for dimensionality reduction that can facilitate the computation of ideal observers. This work presents a conjugate gradient (CG)-based method to construct efficient channels for approximating the IO and HO performance.
[CV-151] Accelerating HEVC Intra Partitioning via a CNN-Hierarchical Attention Transformer Hybrid
链接: https://arxiv.org/abs/2605.29063
作者: Krishna Kumar Sharma,Somdyuti Paul
类目: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:The recursive quad-tree partitioning in High Efficiency Video Coding (HEVC) incurs considerable computational overhead, with exhaustive rate-distortion optimization for CTU partition prediction consuming the dominant share of encoding time. Although partition prediction through deep learning has emerged as a viable encoding accelerator, an architectural dichotomy remains largely unaddressed: CNNs are computationally efficient but spatially myopic due to their localized effective receptive fields, failing to capture long range semantic relationships and repetitive textures; conversely, transformer based architectures are better at capturing global context but incur prohibitive CPU latency, a critical liability that impedes deployment which is predominantly CPU-bound. This paper introduces Hybrid Fast Vision Transformer (HFViT), a hybrid architecture designed to accelerate HEVC intra-mode partition prediction. HFViT fuses a reparameterized depthwise-separable convolutional backbone with a Hierarchical Attention Transformer (HAT) mechanism, leveraging a carrier token scheme to enable efficient global information propagation at sub-quadratic complexity. Post-training structural fusion collapses batch normalization into preceding layers to further reduce latency. Comprehensive evaluation reveals the efficacy of HFViT in accelerating HEVC intra-encoding across resolutions. On standard JCT-VC test sequences, HFViT reduces the average VMAF BD-rate penalty by 2.4, 2.6, and 7.9 percentage points on Classes A, B and E, respectively, as compared to the competing ETH-CNN baseline while maintaining CPU inference latency within 8% of the CNN baseline and surpassing it on GPU by 40%, establishing practical viability for real-time encoder integration.
人工智能
[AI-0] ny but Trusted: Efficient Vision-Language Reasoning for Time-Series Anomaly Detection
链接: https://arxiv.org/abs/2605.30344
作者: Xiaona Zhou,Muntasir Wahed,Tianjiao Yu,Constantin Brif,Ismini Lourentzou
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances in Vision-Language Models (VLMs) have achieved impressive performance across many tasks, yet prior studies report unsatisfactory performance when applying large language or multimodal models to finding abnormal patterns in sequential data. Public anomaly detection benchmarks typically provide interval annotations but not natural-language rationales, making it difficult to fine-tune VLMs to produce grounded, interpretable decisions. To address this gap, we construct VisAnomBench, a curated benchmark built from public time-series datasets and augmented with high-quality anomaly explanations selected from multiple large VLMs using fine-grained, task-specific rewards. Through fine-tuning on this benchmark, we develop VisAnomReasoner, a parameter-efficient VLM for time-series anomaly detection. Experimental results on VisAnomBench show that VisAnomReasoner achieves more accurate anomaly localization and consistently outperforms all baselines, with improvements of at least 21.23 and 23.87 percentage points in precision and F1, respectively. Additional experiments on the TSB-AD-U benchmark demonstrate strong cross-benchmark generalization, with VisAnomReasoner improving precision and F1 by 9.57 and 13.39 percentage points, respectively.
[AI-1] RoboWits: Unexpected Challenges for Robotic Creative Problem Solving
链接: https://arxiv.org/abs/2605.30326
作者: Chunru Lin,Hongxin Zhang,Fenghao Yu,Zhehuan Chen,Thomas L. Griffiths,Yejin Choi,David Held,Chuang Gan
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: The first two authors contributed equally
Abstract:The ability to reason, adapt, and creatively solve problems under unexpected challenges is essential for robots operating in real-world environments. However, current robotic benchmarks primarily emphasize skill-level execution and provide limited insight into such cognitive reasoning capabilities. We introduce RoboWits, a bi-manual robotic benchmark designed to systematically evaluate cognitive reasoning, creative tool use, and robustness to unexpected conditions. To enable scalable construction of high-quality reasoning-centric unexpected scenarios, we propose an automated task generation pipeline formulated as a multi-agent cooperative framework, comprising agents for seed task generation and verification, metric generation, scene generation, and task mutation. Using the pipeline, we curated 30 diverse seed tasks and 208 tasks with mutations and graded difficulty across geometry, material, and assembly-based reasoning. We benchmark popular robot policies, pre-trained VLAs, and oracle-state planners. Our results reveal a significant performance gap: while pre-trained VLAs exhibit preliminary success on seed tasks after single-task fine-tuning, they struggle to perform on mutated tasks, implying their brittleness in manipulation tasks requiring reasoning, strategy adaptation, and robustness to deceptive or constrained environments. Project page is available at this https URL.
[AI-2] In-Context Reward Adaptation for Robust Preference Modeling
链接: https://arxiv.org/abs/2605.30323
作者: Zhenyu Sun,Zheng Xu,Ermin Wei
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement Learning from Human Feedback (RLHF) typically relies on static reward models to align Large Language Models with human preferences. However, human values are inherently diverse and heterogeneous, and a single reward model often lacks the robustness required to generalize to unseen preference domains. While existing multi-reward frameworks attempt to address this, they are often restricted to a fixed set of known domains and fail to adapt to unseen human distributions without costly retraining. In this work, we propose In-Context Reward Adaptation, a transformer-based framework designed to model diverse and unseen human preferences on the fly. By leveraging the in-context learning capabilities of transformers, our approach adaptively infers the underlying reward structure from a small set of preference demonstrations. We demonstrate that while a standard transformer architecture is insufficient for this task by characterizing an asymptotic bias to the ground-truth, incorporating human response time as an auxiliary input signal enables the model to successfully adapt to preferences from previously unseen domains. Our findings show that this approach provides a more robust foundation for preference modeling, allowing for the representation of heterogeneous rewards and preference distribution shift, and offering a scalable path toward more flexible human-AI alignment.
[AI-3] Gram: Assessing sabotage propensities via automated alignment auditing
链接: https://arxiv.org/abs/2605.30322
作者: David Lindner,Victoria Krakovna,Sebastian Farquhar
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce Gram, an automated alignment auditing framework to assess the propensity of AI agents to engage in sabotage. We evaluate Gemini models across 17 simulated agentic deployment scenarios that incentivize sabotage. We find Gemini models misbehave in about 2-3% of our simulated trajectories. Many of these cases are explained by “overeagerness” in Gemini models resulting in both excessive role-playing and goal-seeking behavior. In contrast to other alignment auditing approaches, Gram is designed to specifically evaluate misalignment and intentional sabotage in agentic coding and research agents. We additionally introduce an experimental investigator agent pipeline which enables fine-grained targeted experiments to identify the drivers of misbehavior. We find that increasing realism of environments and removing nudges to misbehave tends to reduce sabotage rates close to zero.
[AI-4] MIRA: Mid-training Rubric Anchoring for Source-Aware Data Selection
链接: https://arxiv.org/abs/2605.30288
作者: Haowen Wang,Yaxin Du,Jian Yang,Jiajun Wu,Shukai Liu,Yuxuan Zhang,Pingjie Wang,Siheng Chen,Tuney Zheng,Ming Zhou,Xianglong Liu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Mid-training has become an important stage in modern LLM development, using large-scale curated mixtures to strengthen capabilities before final post-training. Its data selection problem is distinct: the data are optimized under a pretraining-style objective at near-pretraining scale, but are curated toward downstream capabilities and drawn from heterogeneous sources with different formats and training roles. As a result, effective selection requires both scalability and source-adaptive semantic criteria. Existing model-based methods scale well, but provide only implicit quality signals. Semantic selection methods offer stronger judgments, but usually assume fixed rubrics or standardized data formats. To address this mismatch, we propose MIRA, a source-aware filtering framework based on self-anchored rubric discovery. The key idea is to make rubric construction part of data selection: MIRA first discovers what should be evaluated for each source group, then distills those judgments into scalable student scorers for full-corpus filtering. On code-oriented mid-training with 21 sources and 5 source groups, MIRA outperforms selection baselines across nine code benchmarks and matches the full-corpus run while using only half the tokens.
[AI-5] ProjectionBench: Evaluating Scientific Hypothesis Generation in LLM s Under Progressive Information Disclosure
链接: https://arxiv.org/abs/2605.30284
作者: A. J. Lew(1),Y. Cao(1),M. J. Buehler(1) ((1) Unreasonable Labs)
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 19 pages, 4 figures
Abstract:Scientific discovery is an inherently creative and uncertain process, requiring reasoning beyond the recall of known knowledge. While many benchmarks have been proposed to evaluate large language model (LLM) performance on deep research tasks via multi-hop retrieval, their innovative reasoning abilities essential for true scientific discovery remain largely untested. We introduce a benchmark framework for evaluating model performance in scientific discovery and reasoning, building up from a raw problem to the classical null hypothesis test. In our framework, models initially receive only the topic and research question from a recent paper, with technical details progressively revealed. At each stage of information disclosure, the model is tasked with generating hypotheses that address the research question, which is compared with the conclusions from the original paper and evaluated via automated semantic similarity of constituent atomic claims. This progressive evaluation of semantic divergence from ground-truth conclusions enables assessment of a model’s innovativeness (under minimal information) to grounded reasoning capabilities (under full experimental details), both critical for using LLMs for scientific discovery purposes. Our framework provides a foundation for systematically evaluating scientific reasoning and discovery capabilities in LLMs, crucial for advancing the development of next-generation AI scientist/co-scientist systems. Specifically, here we evaluate GPT-5, GPT-5.4, Gemini 2.5 pro, and Gemini 3.1 pro preview across 45 papers spanning bioactive materials, mechanical materials, and nanomaterials. We find that GPT-5.4 and Gemini 3.1 pro outperform their previous generation counterparts as expected, and GPT-5.4 in particular maintains 0.7 F1 score alignment with ground truth conclusions even under minimal context.
[AI-6] mcp-proto-okn: Natural-language access to open scientific knowledge graphs through the Model Context Protocol
链接: https://arxiv.org/abs/2605.30283
作者: Peter W. Rose,Benjamin M. Good,Amanda M. Saravia-Butler,Charlotte A. Nelson,James P. Balhoff,Yaphet Kebede,Patricia L. Whetzel,Christopher Bizon,Andrew I. Su,Sergio E. Baranzini
机构: 未知
类目: Artificial Intelligence (cs.AI); Emerging Technologies (cs.ET)
备注: 9 pages, 1 figure
Abstract:MCP Server Proto-OKN (mcp-proto-okn) is a Python-based Model Context Protocol server that enables AI assistants to discover, inspect, query and integrate scientific knowledge graphs through natural language. The server provides graph routing, schema inspection, SPARQL execution, ontology expansion, multi-graph querying, and transcript generation, lowering the barrier to cross-domain knowledge graph analysis for biomedical and scientific users. mcp-proto-okn is implemented in Python using the FastMCP framework and is available at this https URL. Documentation, client configuration instructions, and example analysis transcripts are provided in the GitHub repository. Comments: 9 pages, 1 figure Subjects: Artificial Intelligence (cs.AI); Emerging Technologies (cs.ET) Cite as: arXiv:2605.30283 [cs.AI] (or arXiv:2605.30283v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2605.30283 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-7] BORA: Bridging Offline Reinforcement Learning and Online Residual Adaptation for Real-World Dexterous VLA Models
链接: https://arxiv.org/abs/2605.30226
作者: Zhongxi Chen,Yifan Han,Yanming Shao,Huanming Liu,Congsheng Xu,Xiaoyu Chen,Yao Mu,Wenzhao Lian
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 24 pages,11 figures
Abstract:Vision-Language-Action (VLA) models have emerged as a promising paradigm for grounding visual-language understanding into real-world robotic manipulation. However, dexterous manipulation remains challenging for VLA policies due to high-dimensional hand control and compounding execution errors, which makes real-world RL post-training essential for bridging the gap between visually grounded action generation and physically reliable dexterous execution. However, high-dimensional dexterous exploration often triggers temporal inconsistency, sample inefficiency and hardware risks in the real world. To address these challenges, we propose BORA, an offline-to-online RL post-training framework designed for real-world dexterous VLA models. In the offline phase, BORA constructs a critic that takes both the VLM’s cognition tokens and action chunks as inputs. This design enables action-conditioned value guidance, allowing the critic to evaluate dexterous hand motions beyond visual context alone. During the subsequent online phase, BORA freezes the VLA base and introduces a lightweight, Human-in-the-Loop (HiL) chunk-wise residual adaptation mechanism to mitigate real-world execution errors and further correct the offline-learned intents within the actual physical environment. By inheriting the offline critic and employing intervention-driven rewards, BORA effectively corrects execution discrepancies and adapts to real-world physical variances while preserving the pretrained policy as a stable prior. Extensive evaluations across five complex real-world dexterous tasks demonstrate that BORA significantly outperforms pure imitation learning and traditional decoupled RL baselines, achieving a 33% absolute increase in average success rate under standard settings and up to a 43% improvement in unseen object generalization.
[AI-8] Automating Low-Risk Code Review at Meta: RADAR Risk Calibration and Review Efficiency
链接: https://arxiv.org/abs/2605.30208
作者: Chris Adams,Arjun Singh Banga,Parveen Bansal,Souvik Bhattacharya,Rujin Cao,Pedro Canahuati,Nate Cook,Brian Ellis,Prabhakar Goyal,Gurinder Grewal,Tianyu He,Matt Labunka,Alex Manners,David Molnar,Ging Cee Ng,Vishal Parekh,Jiefu Pei,Frederic Sagnes,James Saindon,Will Shackleton,Sid Sidhu,Gursharan Singh,Karthik Chengayan Sridhar,Matt Steiner,Pratibha Udmalpet,Sean Xia,Stacey Yan,Audris Mockus,Peter Rigby,Nachiappan Nagappan
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:AI-assisted coding tools have altered software production. At Meta, significant lines of code per human-landed diff grew by 105.9% year over year and per-developer diff volume rose 51%, with agentic AI responsible for over 80% of that growth. Meanwhile, the share of diffs receiving timely review has declined, exposing a widening gap between code supply and reviewer bandwidth. We ask three questions that progress from feasibility through calibration to impact: (1) can risk-stratified automation operate at scale across diverse organizations, (2) how does tuning the risk threshold affect the trade-off between automation yield and safety, and (3) to what extent does automated review reduce end-to-end latency for AI-generated changes? We deployed RADAR (Risk Aware Diff Auto Review), a multi-stage funnel that classifies each diff by authorship and source type, applies eligibility gates, static heuristics, a machine-learned Diff Risk Score, LLM-based Automated Code Review, and deterministic validation before landing qualifying changes. We evaluate RADAR through telemetry covering 535K+ RADAR-reviewed diffs, observational before-after comparisons for policy changes, and difference-in-differences analysis of efficiency outcomes. RADAR has reviewed 535K+ diffs and landed 331K+. Relaxing the Diff Risk Score threshold from the 25th to the 50th percentile increased the approve rate to 60.31%. The revert rate for RADAR-reviewed diffs is 1/3 that of non-RADAR diffs, and the Production Incident rate is 1/50 that of non-RADAR diffs. RADAR reduces median time to close by over 330% and median diff review wall time by 35%. Risk-aware layered automation can materially reduce review bottlenecks created by AI-driven code growth without compromising production safety.
[AI-9] Persona Conditioning of Brand Recommendations in Retrieval-Augmented Commercial Chat: A Prominence-Stratified Cross-Provider Audit
链接: https://arxiv.org/abs/2605.30207
作者: Will Jack,Noah Lehman,Keller Maloney,Sarah Xu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:The same prompt – “best CRM software” – reaches AI assistants from buyers in widely different contexts: a solo founder, an enterprise VP, a UK SMB owner. We audit how strongly that contextual variation reshapes which brands the model recommends. The audit samples 2,000 runs over a design space of 10 personas x 8 prompts x 3 model configurations x N=10 reps, with the two OpenAI cells at full 8-prompt coverage and the Anthropic sonnet-4.6 / low cell at 4-prompt coverage. Prefixing the user message with a persona drops the recommendation-set similarity (Jaccard) by Delta = -0.12 to -0.20 relative to a same-persona baseline (clustered 95% CIs exclude zero on all three measured cells; the sonnet cell’s CI rests on only 4 prompt clusters and is correspondingly wider). The effect is sharply prominence-stratified: category leaders are persona-resistant (~80% same-brand consistency across personas), but mid-market brands swap up to 75% of the recommendation set as the persona changes. The Anthropic model shows a larger point-estimate effect than the OpenAI configurations, though clustered CIs overlap for the closer contrast (sonnet vs. OpenAI/high); the asymmetry is consistent with Anthropic’s more retrieval-unattributed generation route (43-52% recommendations without observed retrieval-layer evidence, vs OpenAI’s 8-29%, documented in Jack 2026). Any measurement of AI brand perception must condition on the buyer persona supplying the query: the same prompt produces materially different recommendation sets depending on who the model thinks is asking, and a measurement protocol that aggregates across personas systematically obscures that variation. The effect concentrates at mid-market and is largest on the most priors-reliant generation route in our audit, consistent with persona responsiveness growing as models lean more on training-data priors and richer context integration.
[AI-10] HPO: Hysteretic Policy Optimization for Stable and Efficient Training under Sparse-Reward Regime
链接: https://arxiv.org/abs/2605.30201
作者: Mohamed Sana,Nicola Piovesan,Antonio De Domenico,Fadhel Ayed,Haozhe Zhang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:We investigate a narrow but common failure mode of GRPO-style reinforcement learning in the context of sparse verifiable rewards: early updates contain more responses with negative advantages than those with positive advantages, while response-level length normalization ties the magnitude of the update to the length of the output. We propose Hysteretic Policy Optimization (HPO), a minimal modification of GRPO that reduces the weight of negative-advantage updates and replaces per-response length normalization with mean-length normalization. We further introduce Adaptive HPO (A-HPO), which sets the hysteretic weight based on batch-level advantage-sign statistics, thereby removing the need for tuning a fixed hysteretic weight. In our TeleLogs and Countdown experiments, A-HPO improves the reward per update compared to GRPO, with the largest gains in early sparse reward regimes. On TeleLogs, A-HPO achieves a final reward of 0.84, outperforming SAPO by 5%, GSPO by 11%, and GRPO by 15%, while maintaining a comparable response-length. On Countdown, A-HPO achieves the largest gains in initial and most difficult configurations across 1.5B-7B models. Ablation studies on the hysteretic weight show that the gains of A-HPO come from better balancing the contributions of positive and negative advantages compared to positive-only or fully symmetric updates.
[AI-11] Double-Edged Sword or Sharp Tool? Designing and Evaluating Triadic LLM -Teacher Collaboration for K-12 Writing at Scale
链接: https://arxiv.org/abs/2605.30200
作者: Canran Wang,Yuwen Yang,Zhen Wang,Ming Ma,Ding Yu,Chentai Wang,Keman Huang,Xiaoyong Du
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:The double-edged sword of integrating Large Language Models (LLMs) requires an effective triadic collaboration mechanism among LLMs, teachers and students, especially for K-12 education. By developing a triadic collaboration system to support K-12 writing learning, a multidimensional evaluation framework grounded in Systemic Functional Linguistics and the suggestion trajectory tracing pipeline, this paper contributes a large-scale empirical dataset involving 57,954 essays from 10,195 students across 120 schools over two years. Our findings confirm the efficacy of this system in improving writing quality through a strategic labor division: the LLM serves as a generative engine to mitigate teacher burnout, and the teacher acts as a pedagogical gatekeeper and bridge to guarantee feedback quality. While both LLM and teacher are critical for skill improvement, we uncover a ceiling effect where excessive linguistic expansion yields diminishing marginal utility. These suggest a dynamically adaptive LLM-teacher collaboration as student proficiency increases.
[AI-12] CalArena: A Large-Scale Post-Hoc Calibration Benchmark
链接: https://arxiv.org/abs/2605.30188
作者: Eugène Berta,David Holzmüller,Francis Bach,Michael I. Jordan
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: 30 pages, 9 figures
Abstract:Reliable probability estimates are critical in many machine learning applications, yet modern classifiers are often poorly calibrated. Post-hoc calibration provides a simple and widely used solution, but the large number of proposed methods, combined with small-scale and inconsistent evaluations, makes it difficult to determine which approaches are truly effective in practice. We introduce a large-scale, standardized benchmark for post-hoc calibration, covering nearly 2000 experiments across tabular and computer vision tasks, including binary, multiclass, and large-scale classification settings. Our benchmark aggregates predictions from a diverse set of classical models, modern deep learning architectures, and foundation models, and provides unified, reproducible implementations of dozens of calibration methods within a common evaluation framework. We argue that Post-Hoc Improvement (PHI) in proper scoring rules offers a principled alternative to traditional calibration error estimators for comparing post-hoc methods, capturing both calibration quality and potential degradation to the model’s predictive performance. Using this framework, we conduct the most comprehensive empirical study of post-hoc calibration to date. Our results reveal consistent patterns across domains: smooth calibration functions outperform binning-based approaches, dedicated multiclass methods are essential in high-dimensional settings, and generic machine learning models are not competitive without calibration-specific design. To facilitate future research, we release all data, code, and evaluation tools, providing a plug-and-play benchmark for developing and comparing calibration methods.
[AI-13] Modularizing Educational LLM -Agency for Fostering Responsible Learning Assistance
链接: https://arxiv.org/abs/2605.30187
作者: Julius Gabelmann,Felix Jahn,Kevin Baum,Sophie van Rossum,Emely Wuenscher,Timo P. Gros,Verena Wolf
机构: 未知
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: 12 pages, 2 figures (+ 2 in appendix), accepted at AISoLA 2025 (Track: Responsible and Trusted AI: An Interdisciplinary Perspective)
Abstract:The widespread adoption of AI chatbots in education will drastically change learning, making responsible deployment a critical concern. While large language models (LLMs) might have access to sources discussing insights from educational sciences, they are not particularly inclined to adhere to pedagogical concepts, risking negative effects on the learning process, such as a loss of transfer capabilities, critical thinking, or creativity. In this paper, we introduce an agentic AI chatbot architecture assisting students with exercise solving, specifically designed to contribute to more responsible AI use in education. We base our conceptual development on the identification of several desiderata for responsible LLM-based educational systems, argue for the structural shortcomings inherent in monolithic, out-of-the-box solutions, and instead suggest modularizing the agentic architecture. We propose specific modules for different stages of exercise solving, enabling incorporation of targeted pedagogical advice, guiding students through the learning process in a more controllable, transparent, and overseeable manner.
[AI-14] LoRA: Bayesian Low-Rank Adaptation with Latent Interaction Graphs for Microbiome Diagnosis ICML2026
链接: https://arxiv.org/abs/2605.30179
作者: Yang Song,Yixuan Zhang,Lingfa Meng,Tongyuan Hu,Haizhou Shi,Hao Wang,Samir Bhatt,Hengguan Huang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at ICML 2026
Abstract:Parameter-efficient adaptation has made LLMs practical for domain prediction, but standard LoRA still relies on a static low-rank update and does not expose the latent interactions that often drive scientific labels. We introduce iLoRA. To our knowledge, it is the first Bayesian graph-conditioned LoRA framework. It infers a latent interaction graph from the input and uses it to generate input-conditioned LoRA updates. As a result, iLoRA learns prediction and latent interaction structure jointly, rather than training a predictor and applying interaction analysis only post hoc. We instantiate this idea for microbiome diagnosis, where disease state can depend on both species-level abundance and microbe-microbe cross-talk, and evaluate it in two complementary settings: interactive QA with human-annotated graphs, which tests latent structure recovery, and multi-cohort IBD diagnosis, which tests biomedical utility. Across both settings, iLoRA improves over strong LoRA and Bayesian adaptation baselines, recovers graphs aligned with human annotations and cohort-level microbiome associations, and provides calibrated uncertainty with moderate graph-branch overhead.
[AI-15] BioRefusalAudit: Auditing Biosecurity Refusal Depth Using General and Domain-Fine-Tuned Sparse Autoencoders
链接: https://arxiv.org/abs/2605.30162
作者: Caleb DeLeeuw
机构: 未知
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注: 21 pages, 2 figures, 3 tables. Apart Research AIxBio Sprint hackathon paper, April 2026 (Track 3: AI Biosecurity Tools). Code, eval set, and SAEs: this http URL . Reviewer feedback: this http URL
Abstract:Biosecurity evaluations of language models typically ask whether models produce hazardous output. This paper asks a complementary question: when a model refuses, is that refusal structurally sound, or does it disappear under modest changes to prompt framing, formatting, or output length? Across five architectures, no model cleanly discriminated benign from hazard. Gemma 2 2B-IT never genuinely refused across 75 prompts, hedging on every hazard-adjacent query. Gemma 4 E2B-IT refused 65/75 prompts with chat-template formatting and 0/75 without it. Both Gemma models collapsed to 0% under an 80-token cap. Qwen 2.5 1.5B and Phi-3-mini over-refused, flagging 83-87% of benign biology as hazardous. Llama 3.2 1B showed the only meaningful tier gradient (61-point spread). To probe what drives such over-refusal, we tested a panel of Schedule I but biologically non-toxic compounds (notably psilocybin cultivation, with FDA Breakthrough Therapy status). Some models refused these at rates exceeding genuinely hazardous biology, suggesting refusal tracks legality and cultural salience over CBRN hazard. To measure the internal side, we introduce a divergence score D comparing a model’s surface response label to its internal sparse autoencoder (SAE) feature activations. Full D was computed on Gemma 2 2B-IT (Gemma Scope 1) and Gemma 4 E2B-IT (author-trained bio SAE). Two fine-tuned Gemma 2 domain SAEs were released. On Gemma 4, comply and refuse responses separated by a 0.647-point gap with zero overlap (n=75), though this is preliminary, with a narrow catalog, within-sample calibration, and Gemma-family-only SAE coverage. Built over one hackathon weekend on consumer hardware (GTX 1650 Ti Max-Q, plus Colab T4 for SAE training), this preliminary evidence suggests activation-level auditing may surface failure modes invisible to behavioral evaluation, with substantial variation across architectures.
[AI-16] On Distributional Reinforcement Learning in Chaotic Dynamical Systems
链接: https://arxiv.org/abs/2605.30160
作者: James Rudd-Jones,Mirco Musolesi,María Pérez-Ortiz
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Chaotic dynamical systems pose a fundamental challenge for Reinforcement Learning (RL): exponential sensitivity to initial conditions induces high-variance bootstrap targets and poorly conditioned gradient updates. Chaotic dynamics arise across scientific and engineering domains, from fluid flows and climate systems to multi-agent systems, where reliable learning is highly desirable. Standard RL methods optimise expected returns through scalar value functions, implicitly averaging over diverging trajectories and entangling trajectory level instability with the learning objective. We show that under mild statistical stability assumptions, the return distribution evolves more regularly than individual trajectories when measured under the 1 -Wasserstein metric, yielding a smoother distributional Bellman objective. By aligning optimisation with this measure level structure, distributional RL provides better conditioned learning. We offer a principled explanation for the advantages of distributional methods in chaotic systems and the geometries of RL objectives under chaos.
[AI-17] Meta-Cognitive Memory Policy Optimization for Long-Horizon LLM Agents
链接: https://arxiv.org/abs/2605.30159
作者: Ziyan Liu,Zhezheng Hao,Yeqiu Chen,Hong Wang,Jingren Hou,Ruiyi Ding,Yongkang Yang,Wence Ji,Wei Xia,Feng Liu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Memory-augmented LLM agents tackle complex long-horizon tasks by recursively summarizing interaction trajectories into compact memory. However, existing approaches typically train these memory policies using outcome-based reinforcement learning, failing to localize where intermediate memory quality degrades. As interactions unfold, ambiguous recursive summaries progressively discard task-relevant information and introduce semantic noise. This exacerbates belief deviation, obscuring the agent’s estimate of the latent task state and ultimately derailing long-horizon reasoning. We therefore argue that memory optimization should focus not merely on trajectory-level success, but on the clarity of the belief induced by intermediate summaries. To this end, we introduce Belief Entropy, a self-supervised proxy that probes how uncertain the model remains about the latent task state given its current memory. Based on this proxy, we propose Metacognitive Memory Policy Optimization (MMPO). Instead of relying only on sparse outcome-based signals, MMPO provides fine-grained, memory-specific supervision via explicitly penalizing summaries that induce high epistemic uncertainty. Experiments show that MMPO consistently outperforms existing methods on diverse long-horizon tasks, maintaining 97.1% performance even when scaled to 1.75M-token contexts.
[AI-18] Neural Network Verification using Partial Multi-Neuron Relaxation
链接: https://arxiv.org/abs/2605.30155
作者: Ido Shmuel,Guy Katz
机构: 未知
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI)
备注: To appear in SAIV 2026
Abstract:The increasing integration of deep neural networks in critical systems has spawned a theoretical and practical interest in formally guaranteeing safety properties about their behavior. To achieve this, contemporary verification algorithms rely on computing linear relaxations for a network’s non-linear activation functions. Existing approaches for linear relaxations typically fall into one of two categories: single-neuron relaxation, in which each activation neuron is bounded in terms of its sources; and multi-neuron relaxation, in which linear bounds involving multiple activation neurons and their sources are calculated. However, existing methods might fail to balance tightness and scalability, as single-neuron bounds might not derive sufficiently tight bounds necessary for verification to complete, whereas generating multi-neuron relaxation for all activation neurons is computationally expensive. In this paper, we present a middle-ground approach featuring partial multi-neuron relaxation, in which we generate multi-neuron bounds for only a small, heuristically selected subset of neurons. To achieve this, we build upon existing branching heuristics for selecting neurons and for optimizing bounding hyper-planes for multi-neuron bounds. We integrated our proposed method within the Marabou verifier, and obtained favorable results in comparison to existing bound tightening methods. Our experiments showcase the potential of our technique for neural network verification.
[AI-19] mporal Stability and Few-Shot Prompting in Math Task Assessment
链接: https://arxiv.org/abs/2605.30151
作者: Danielle S. Fox,Brenda L. Robles,Elizabeth DiPietro Brovey,Christian D. Schunn
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 23 pages, 1 figure
Abstract:As AI tools become increasingly integrated into educational contexts, questions arise about both their stability over time and their responsiveness to prompt engineering techniques. This longitudinal study focused on different AI tools’ ability to use the Task Analysis Guide (TAG; Stein \ Smith, 1998) to classify the cognitive demand of mathematics tasks. In particular, it examined whether this classification ability changed with (1) model version updates over time and (2) few-shot prompting using exemplar tasks. We tested a general-purpose AI tool (Gemini) and an education-specific AI tool (Coteach). The specific tools were selected because of their relatively high performance on relevant published benchmarks and prior task-specific tests. Models were tested at baseline, retested with model version updates, and then tested again using few-shot prompting (two exemplar tasks for each cognitive demand category). Results revealed that newer model versions alone produced mixed effects: Gemini’s accuracy remained stable at 58%, while Coteach’s accuracy decreased from 75% to 50%. However, few-shot prompting improved both models’ performance: Gemini increased to 67% and Coteach recovered to 75% accuracy. These findings demonstrate that prompt engineering techniques can have larger and more reliable effects than passive model improvements, and that version updates may not always improve performance on specialized educational tasks. The study has important implications for how educators and researchers should approach AI tool selection, evaluation, and implementation in educational contexts.
[AI-20] Anchorless Diversification for Parallel LLM Ideation
链接: https://arxiv.org/abs/2605.30150
作者: Fares Nabil Ibrahim,Nafis Saami Azad,Raiyan Abdul Baten
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLMs are increasingly used to generate candidate-idea pools for creative tasks where broad exploration is valuable. Parallel inference can be attractive in this setting when it broadens the pool while retaining quality and cost efficiency. We study inference-time controls for candidate-pool diversification, asking whether anchorless methods can rival methods that depend on observed seed ideas. Across three creative task families, we compare independent generation and semantic direction stratification with self-, peer-, and representative-anchor baselines, under neutral and population-referential divergent instructions. Population-referential divergence is a strong low-cost baseline, increasing semantic diversity while preserving quality proxies. Semantic direction stratification is stronger: a single planning call organizes generations across broad semantic directions, yielding the best diversity–quality–compute frontier. Anchored regeneration can be strong in final-pool diversity, but its advantage shrinks under full-pipeline token accounting. These results establish practical anchorless baselines for open-ended LLM ideation.
[AI-21] Overcoming Forgetting in LLM Fine-Tuning with Evolution Strategies
链接: https://arxiv.org/abs/2605.30148
作者: Kajetan Schweighofer,Conor F. Hayes,Roberto Dailey,Risto Miikkulainen,Xin Qiu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Evolution Strategies (ES) has recently emerged as a competitive alternative to reinforcement learning (RL) for large language model (LLM) fine-tuning, offering advantages through simplicity, scalability, and inference-only training. However, recent work suggests that ES fine-tuning on new tasks may induce forgetting of prior tasks. First, this paper shows that prior task forgetting (1) is better characterized as performance drift rather than irreversible forgetting, with prior-task performance often recovering during ES training; and (2) is not a specific failure mode of ES, but can also arise for fine-tuning with RL methods. Second, it analyzes when and why such drift arises, highlighting its dependence on ES training dynamics, particularly random walk behavior in weakly constrained directions of the weight space. Third, based on these insights, it introduces Anchored Weight Decay (AWD) as a parameter-space regularization technique that constrains optimization toward the initial model parameters. AWD effectively stabilizes prior-task performance while preserving target-task performance, achieving benefits comparable to large ES population sizes at much lower computational cost. Thus, contrary to previous beliefs, the paper shows that prior-task forgetting under ES is largely avoidable, positioning ES as a promising approach for continual learning in LLMs.
[AI-22] Enhancing Multi-Agent Communication through Attention Steering with Context Relevance
链接: https://arxiv.org/abs/2605.30136
作者: Hongxiang Zhang,Yuan Tian,Tianyi Zhang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLM-based multi-agent systems have demonstrated remarkable performance on complex tasks through collaborative reasoning. However, these systems tend to rapidly accumulate extremely long conversation histories during interaction. As conversations lengthen, relevant information is increasingly diluted by irrelevant context, leading to degraded performance. In this work, we present Agent-Radar, a training-free context management method that dynamically steers each agent’s attention toward relevant context with a novel temporal and spatial decay mechanism. Our experiments demonstrate that Agent-Radar outperforms state-of-the-art methods across five different benchmarks, yielding gains of up to 7.64 absolute points. Furthermore, our analysis shows that Agent-Radar remains effective and robust as the number of agents and interaction rounds increases. Finally, the ablation study shows that core components in Agent-Radar are crucial to performance and generalizable in different settings.
[AI-23] DAMEL: Dual-Axis Multi-Expert Learning for Class-Imbalanced Learning
链接: https://arxiv.org/abs/2605.30135
作者: Hyuck Lee,Taemin Park,Heeyoung Kim
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Various algorithms have been proposed to address the challenges posed by class-imbalanced learning from real-world data with long-tailed distributions. While these algorithms reduce prediction bias through rebalancing techniques, they often introduce increased prediction variance as a trade-off. Several multi-expert learning algorithms aim to address this variance but involve complex procedures. We propose a new multi-expert learning algorithm, called the dual-axis multi-expert learning (DAMEL), which reduces both bias and variance of predictions by using multiple experts along both representation and time axes. Along the representation axis, DAMEL concatenates the representations of multiple experts and trains an auxiliary balanced classifier simultaneously with the concatenated representations. Along the time axis, DAMEL aggregates network weights across training epochs, employing these aggregated weights during testing. Experimental results demonstrate that DAMEL reduces both bias and variance of predictions, highlighting its effectiveness in class-imbalanced learning.
[AI-24] Beyond MSE: Improving Precipitation Nowcasting with Multi-Quantile Regression
链接: https://arxiv.org/abs/2605.30122
作者: Gijs van Nieuwkoop,Siamak Mehrkanoon
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 7 pages, 5 figs
Abstract:Deep-learning precipitation nowcasting models are often optimized using pointwise losses such as mean squared error or mean absolute error, which can lead to overly smooth forecasts and poor representation of heavy rainfall. This study investigates whether the predictive performance of an established deterministic nowcasting architecture can be improved by reformulating training as a multi-quantile regression problem. Using SmaAt-UNet as a core model, we compare MSE, MAE, and multi-quantile pinball-loss training on radar precipitation nowcasting over the Netherlands. The results show that multi-quantile training improves the central deterministic forecast, decreasing test-set MSE by 8.6% compared to a model trained using MSE, while also producing upper-quantile outputs that are useful for risk-sensitive prediction of heavy precipitation. These findings suggest that quantile regression provides a simple alternative to standard pointwise losses without requiring a new architecture or generative sampling procedure. The implementation of our models and training setup is available on \hrefthis https URLGitHub.
[AI-25] Evolving Features vs Evolving Entire Trees with GP for Interpretable Survival Analysis
链接: https://arxiv.org/abs/2605.30119
作者: Thalea Schlender,Peter A.N. Bosman,Tanja Alderliesten
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注:
Abstract:Survival analysis concerns the task of predicting the time until an event occurs. Often used in the medical field, survival analysis deals with incomplete (i.e., censored) data, for instance, from patients who did not experience the event during the duration of the study. For practical use, both accuracy and interpretability are important. Survival trees are easy-to-follow survival models that split the patient cohort recursively into discrete patient groups. Whilst survival trees can capture complex relationships, they typically need to grow large, threatening interpretability. Moreover, survival trees are often built using greedy approaches that may overlook globally optimal split combinations, limiting predictive performance. Shallow survival trees require expressive, higher-order feature combinations to achieve competitive accuracy. We therefore use genetic programming to multi-objectively evolve inherently inspectable feature sets and study how they interact with different tree induction strategies. We further introduce an evolutionary approach that jointly optimises the survival tree structure and the non-linear split logic. Our findings demonstrate that evolutionary feature construction improves predictive performance across different tree induction strategies on two real-world datasets and two different survival tree depths. Full joint evolution has the overall highest potential to propose multiple inherently inspectable shallow survival trees of good performance. Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE) Cite as: arXiv:2605.30119 [cs.LG] (or arXiv:2605.30119v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2605.30119 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Thalea Schlender [view email] [v1] Thu, 28 May 2026 15:52:14 UTC (7,330 KB)
[AI-26] VLA-Trace: Diagnosing Vision-Language-Action Models through Representation and Behavior Tracing
链接: https://arxiv.org/abs/2605.30117
作者: Haoyuan Shi,Xiancong Ren,Yingji Zhang,Qinfan Zhang,Jiayu Hu,Haozhe Shan,Han Dong,Jinpeng Lu,Yinda Chen,Yi Zhang,Yong Dai,Xiaozhu Ju
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Understanding how Vision-Language-Action (VLA) models transform multimodal knowledge into embodied control remains an open challenge. We present VLA-Trace, a progressive diagnostic framework that analyzes VLA models through a unified evidence chain from representation dynamics to causal control attribution and behavioral manifestation. It specifically combines cross-modal and checkpoint-drift centered kernel alignment (CKA) to trace representation evolution, attention knockout interventions to identify modality-specific control pathways, and rollout-level behavioral probes to examine grounding, shortcut dependence, and semantic following. Experiments on \pi_0.5 and OpenVLA reveal three key findings. First, the two models exhibit distinct modality-specific adaptation dynamics during VLA finetuning. Second, they rely on different multimodal routing strategies and layer-wise dependencies during action decoding. Third, although VLA policies excel at visually grounded trajectory generation, they remain limited in fine-grained semantic following. These findings highlight future directions for representation-preserving adaptation, causal VLA circuits, and compositional semantic control.
[AI-27] How Reliable Are AI Attackers Against a Fixed Vulnerable Target? A 400-Run Empirical Study of LLM Penetration Testing Consistency
链接: https://arxiv.org/abs/2605.30096
作者: Galip Tolga Erdem
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 41 pages, 7 figures. Code and 400-run dataset: this https URL
Abstract:Large language models (LLMs) can autonomously conduct multi-stage cyber attacks, but the consistency of their offensive behavior under repeated trials remains unstudied. This work presents the first large-scale empirical measurement of LLM attack consistency: 400 autonomous penetration testing runs (4 models, 100 each) against an identical honeypot hosting OWASP Juice Shop and two additional vulnerable services, holding prompt, orchestrator, and target constant. No model emitted a content refusal that survived the orchestrator’s one-shot authorization re-prompt at iterations 0-1. Claude Sonnet 4’s API calls did encounter upstream service unavailability - 91 of 1,135 calls returned HTTP 529 overloaded_error during a documented Anthropic capacity event, truncating 39 of 100 Claude runs. An earlier draft catalogued these as safety refusals; on full-log audit they are upstream API failures, not model-level refusals. Despite this, Claude achieved full exploitation in 61 of 100 runs; Gemini 2.5 Flash-Lite in 85; GPT-4o-mini in 56 while deploying 98 unique attack strategies; qwen2.5-coder:14b in 25. Failure modes are model-distinctive: Claude through API truncation (39 runs), qwen through premature completion (52), GPT-4o-mini through iteration-budget exhaustion (23). Cross-service credential reuse appeared only in configurations retaining the most conversation history (qwen 57%, GPT-4o-mini 49%, cloud models 0% on 5-exchange windows). Cross-model exploitation rate differences are statistically significant (p 0.001) with large effect sizes; qwen vs. Gemini SQL injection rates differ at Cohen’s h = 1.12. First-exploit timing fell within a 15-30 second wall-clock range. To our knowledge, this is the first study to measure autonomous LLM attack behavior at N=100 per model across a multi-service target.
[AI-28] PokerSkill: LLM s Can Play Expert-Level Poker without Training or Solvers
链接: https://arxiv.org/abs/2605.30094
作者: Boning Li,Baoxiang Wang,Longbo Huang
机构: 未知
类目: Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT)
备注: 45 pages, 3 figures
Abstract:Poker is a landmark challenge for artificial intelligence. The dominant approach relies on equilibrium solvers built on counterfactual regret minimization, requiring millions of core-hours of training. Large Language Models (LLMs) possess extensive poker knowledge but perform far below solver-based agents when asked to play directly. Traditional rule-based poker agents are interpretable and training-free, but their strategic ceiling remains far below equilibrium play. We introduce \textbfPokerSkill, a training-free and solver-free framework that bridges this gap by using detailed rule-based poker skills as a structured action-grounding interface for LLMs. A deterministic context engine analyzes the current state and retrieves only the relevant fragments from a layered skill library, which is entirely designed by human poker experts, constraining the LLM’s choice to reasonable actions. Against GTOWizard, a state-of-the-art GTO benchmark, GPT-5.5 XHigh with PokerSkill achieves -57 \pm 21 mbb/hand, Claude Opus 4.6 achieves -80 \pm 29 mbb/hand and Claude Opus 4.7 achieves -87\pm 64 mbb/hand, reducing losses by 49–61% compared to default-prompt baselines and outperforming the strong bot Slumbot. Our key finding is that rule-based skills alone do not constitute a strong strategy, and LLMs alone cannot play well, but their combination yields an agent that requires neither training nor solver access yet competes with systems built on millions of core-hours of computation. To our knowledge, this is the first demonstration of an LLM achieving competitive performance in a complex imperfect-information game without game-specific training or solver queries. Code is available at this https URL.
[AI-29] Selective QA over Conflicting Multi-Source Personal Memory: A Diagnostic Testbed and Method Comparison
链接: https://arxiv.org/abs/2605.30087
作者: Tiancheng Yang,Matthias Schonlau,Ilia Sucholutsky
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 55 pages, 5 figures
Abstract:Emerging personal AI agents are moving toward persistent, multi-source memory. This creates an evaluation problem: systems must decide how to use conflicting or incomplete evidence; they cannot just retrieve facts from one clean history. Existing benchmarks rarely show whether an error came from the evidence given to a method or from the method’s conflict-resolution step. We study this as selective QA over conflicting multi-source personal memory: systems answer based on conflicting, sometimes incomplete sources, or abstain when evidence is insufficient. We develop a benchmark containing 18 question templates across 8 reasoning types, 480 personas, 4 random seeds, and 34,560 instances, with controlled source distortions and deterministic ground truth. We evaluate the performance of baselines without access to any source, access to a single source, structured fusion methods, and frontier LLMs. The best trained fusion resolver reaches 80.3% accuracy, while the strongest prompt-only LLM baseline reaches 70.0%. With abstention, the same resolver reaches 85.3% selective accuracy at 78.3% coverage and the best LLM reaches 71.0% selective accuracy at 95.4% coverage. Different models have different strengths across reasoning types. We release the data, code, cached model outputs, and data-generating process for reuse.
[AI-30] A Predictive Law for On-Policy Self-Distillation From World Feedback
链接: https://arxiv.org/abs/2605.30070
作者: Tommy He,Jerome Sieber,Matteo Saponati
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Moving beyond simple scalar rewards toward richer world feedback is a natural path to more scalable RL post-training. On-policy self-distillation (OPSD) is a promising recent approach that uses arbitrary feedback as learning signal, yet its reliability compared to established methods, such as GRPO, remains unclear. We identify a strikingly consistent linear correlation between the initial student-self-teacher performance gap and the final performance improvement in OPSD. This relationship holds across context types and model families, providing a powerful predictive law for anticipating the outcome of an OPSD configuration without running the full training procedure. Interestingly, we show that this linear predictability holds with model scale, suggesting a potential basis for new empirical scaling laws on larger models with stronger in-context learning capabilities. In essence, our findings show that OPSD performance can be predicted and tuned before training, offering a principled way to incorporate world feedback as a first-class component of the post-training pipeline.
[AI-31] Projectional Decoding: Towards Semantic-Aware LLM Generation
链接: https://arxiv.org/abs/2605.30054
作者: Boqi Chen,José Antonio Hernández López,Aren A. Babikian
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 5 pages, 3 figures. Accepted at FSE 2026 IVR track
Abstract:Large language models (LLMs) are increasingly used to generate software artifacts across many software engineering (SE) tasks, yet ensuring the semantic validity of these artifacts remains a fundamental challenge. Existing constrained decoding techniques can enforce syntactic correctness and, in some cases, specific semantic rules, but lack a general representation that bridges LLM-generated text with the reasoning required for semantic validation in SE. In this paper, we propose projectional decoding, a novel conceptual framework that integrates domain semantics directly into the generation process by maintaining, alongside text, a partial graph model as the primary artifact representation throughout generation. This abstract representation enables incremental semantic validation by explicitly capturing uncertainty and natively supporting error detection, while guiding generation toward semantically valid outputs with provable guarantees. We present preliminary results on a program generation task which demonstrate the potential of this approach to improve the semantic validity of LLM-generated artifacts. We also discuss how projectional decoding can enable verifiable automation with LLMs across various SE activities.
[AI-32] Robust and Generalizable Safety Steering for Text-to-Image Diffusion Transformers
链接: https://arxiv.org/abs/2605.30049
作者: Zihao Xue,Yan Wang,Zhen Bi,Long Ma,Zhonglong Zheng,Zeyu Yang,Bingyu Zhu,Longtao Huang,Jie Xiao,Jungang Lou
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Diffusion Transformers have become a powerful backbone for text-to-image generation, but their layered and cross-modal generation process makes safety control fundamentally different from prompt-level filtering or output-level detection. Harmful semantics may be weakly expressed in text representations, progressively bound to visual latents, and finally entangled with rendering dynamics. As a result, safety steering at a fixed layer can be unstable, and a steering mechanism learned from known risks may not transfer reliably to a shifted target risk domain. We propose SafeDIG, a safety steering framework that formulates DiT safety adaptation as position-aware sparse feature transfer. SafeDIG first constructs Sparse Autoencoders over functionally distinct DiT intervention positions and uses robustness-aware pre-training routing to prioritize intervention sites that are expected to remain stable under source-target risk shift. It then separates transferable safety features from domain-specific activation geometry by freezing the SAE encoder as a reusable sparse safety dictionary and adapting only the decoder to the target-domain activation manifold. During inference, SafeDIG combines Blend and Repel operations to steer unsafe activations toward transferred safety manifolds or away from harmful sparse directions. Experiments on FLUX.1 Dev and Stable Diffusion 3.5 Large show that SafeDIG consistently reduces target-domain and overall unsafe generation rates while preserving source-domain safety and image quality.
[AI-33] Masked Diffusion Modeling for Anomaly Detection
链接: https://arxiv.org/abs/2605.30046
作者: Lixing Zhang,Yuchen Liang,Liyan Xie
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Anomaly detection aims to identify samples that deviate from the nominal data distribution and is central to many safety-critical applications. However, developing effective anomaly detection methods for categorical, mixed-type, and discrete sequence data remains challenging and relatively underexplored. Masked diffusion models provide a natural way to model such data by learning to recover masked values from the remaining visible context. In this paper, we propose Masked Diffusion for Anomaly Detection (MaskDiff-AD), a forward-only method based on masked diffusion models trained only on nominal data. Given a test sample, MaskDiff-AD constructs anomaly scores from the difficulty of reconstructing randomly masked coordinates, yielding a content-sensitive score that operates directly on discrete state spaces while avoiding reverse-time sampling. We also develop a non-parametric variant of MaskDiff-AD and provide theoretical guarantees by characterizing Type-I and Type-II errors under a fixed detection threshold. Experiments on fourteen categorical and mixed-type tabular datasets from ADBench and UADAD, as well as four text anomaly detection datasets from NLP-ADBench, show that MaskDiff-AD achieves competitive performance against classical, diffusion-based, and recent tabular/text anomaly detection baselines. Notably, MaskDiff-AD achieves the best overall average rank, outperforming all twelve tabular baseline methods.
[AI-34] Learning to Choose: An Empowerment-Guided Multi-Agent System with semantic communication for Adaptive Method Selection
链接: https://arxiv.org/abs/2605.30042
作者: Geremy Loachamín-Suntaxi,Robert Lazar,Dimitrios G. Giovanis,Ioannis G. Kevrekidis,Eleni D. Koronaki
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Automating scientific computing workflows requires more than generating executable code: autonomous systems must also select appropriate computational strategies, implement them faithfully, and ensure that the resulting outcomes remain causally attributable to the decisions that produced them. In multi-agent pipelines, this process is particularly fragile, as small inconsistencies between agent intentions and actions can lead to semantic drift, where the eventually executed procedure no longer reflects the originally selected strategy, thereby corrupting downstream evaluation and adaptation. In this work, motivated by the ATHENA framework (Toscano et al., 2025; Toscano et al., 2026) and the concept of empowerment (Yiu et al., 2025), we introduce a multi-agent framework that combines contextual bandits with structured inter-agent communication and, most importantly, semantic checkpoints that preserve action-outcome fidelity throughout the pipeline. The system integrates specialized large language model (LLM) agents, grounded code generation, and self-healing execution loops within an adaptive decision-making architecture. Interpreting the framework through the lens of empowerment, we show that reliable autonomous learning requires not only identifying high-quality actions, but also preserving the integrity of their propagation across agents. Using sensitivity analysis and uncertainty quantification workflows as representative case studies, we demonstrate that unchecked semantic drift degrades policy learning, whereas the proposed framework improves convergence, robustness, and adaptation to novel problem contexts. These results suggest a broader design principle for scientific multi-agent systems: adaptive decision-making must be coupled with explicit mechanisms that guarantee semantic consistency and reliable information flow across the computational pipeline.
[AI-35] Domain-Specific Data Synthesis for LLM s via Minimal Sufficient Representation Learning KDD2026
链接: https://arxiv.org/abs/2605.30039
作者: Tong Ye,Hang Yu,Tengfei Ma,Xuhong Zhang,Jianguo Li,Peng Di,Peiyu Liu,Jianwei Yin,Wenhai Wang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: Accepted by KDD 2026
Abstract:Large Language Models have demonstrated remarkable progress in general-purpose capabilities and can achieve strong performance in specific domains through fine-tuning on domain-specific data. However, acquiring high-quality data for target domains remains a significant challenge. Existing data synthesis approaches follow a deductive paradigm, heavily relying on explicit domain descriptions expressed in natural language and careful prompt engineering, limiting their applicability in real-world scenarios where domains are difficult to describe or formally articulate. In this work, we tackle the underexplored problem of domain-specific data synthesis through an inductive paradigm, where the target domain is defined only through a set of reference examples, particularly when domain characteristics are difficult to articulate in natural language. We propose a novel framework, DOMINO, that learns a minimal sufficient domain representation from reference samples and leverages it to guide the generation of domain-aligned synthetic data. DOMINO integrates prompt tuning with a contrastive disentanglement objective to separate domain-level patterns from sample-specific noise, mitigating overfitting while preserving core domain characteristics. Theoretically, we prove that DOMINO expands the support of the synthetic data distribution, ensuring greater diversity. Empirically, on challenging coding benchmarks where domain definitions are implicit, fine-tuning on data synthesized by DOMINO improves Pass@1 accuracy by up to 4.63% over strong, instruction-tuned backbones, demonstrating its effectiveness and robustness. This work establishes a new paradigm for domain-specific data synthesis, enabling practical and scalable domain adaptation without manual prompt design or natural language domain specifications.
[AI-36] RAISE: RAG Design as an Architecture Search Problem
链接: https://arxiv.org/abs/2605.30029
作者: Zhen Chen,Yibing Liu,Weihao Xie,Yu Liang,Peilin Chen,Shiqi Wang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Retrieval-augmented generation (RAG) systems expose numerous design choices spanning query rewriting, chunking, retrieval depth, reranking, and context compression. In practice, these choices are often configured through heuristics, hindering systematic evaluation and reproducibility across settings. We argue that this challenge is best formulated as RAG architecture search. To support controlled and reproducible study of this problem, we introduce the RAG Intelligence Search Engine (RAISE), a comprehensive framework and benchmark for RAG hyperparameter optimization, which evaluates optimization methods for RAG pipelines under standardized search spaces and budgets. RAISE implements 13 search algorithms and evaluates them across seven public text and multimodal datasets using three random seeds. Our experiments show that optimization performance is highly task-dependent: methods that perform strongly on one dataset may not generalize consistently across others, cautioning against interpreting aggregate rankings as evidence of universally superior strategies. RAISE provides a common experimental substrate for fair, reproducible, and systematic research on RAG hyperparameter optimization.
[AI-37] st Time Training for Supervised Causal Learning
链接: https://arxiv.org/abs/2605.30015
作者: Zizhen Deng,Jiaru Zhang,Rui Ding,Huang Bojun,Jinzhuo Wang,Qiang Fu,Shi Han,Dongmei Zhang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Supervised Causal Learning (SCL) has shown promise in causal discovery by framing it as a supervised learning problem. However, it suffers from significant out-of-distribution generalization challenges. We reveal three limitations of previous SCL practices: a significant performance gap between synthetic benchmarks and real-world data, fragility to distribution shifts, and failure in compositional generalization, collectively questioning its real-world applicability. To address this, we propose Test-Time Training for Supervised Causal Learning (TTT-SCL), a novel framework that dynamically generates training sets explicitly aligned with any specific test instance. We demonstrate the correlation between TTT-SCL and score-based methods, and design an efficient module for generating training sets based on the classic scoring function. Experiments on synthetic benchmarks, pseudo-real and real-world datasets demonstrate that TTT-SCL significantly outperforms existing SCL and traditional causal discovery methods.
[AI-38] From GPS Points to Travel Patterns: Flexible and Semantic Trajectory Generation with LLM s KDD2026
链接: https://arxiv.org/abs/2605.30014
作者: Silin Zhou,Chenhao Wang,Yuntao Wen,Shuo Shang,Lisi Chen,Panos Kalnis
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: This paper is accepted by KDD2026 second round
Abstract:Urban trajectories play a crucial role in modeling urban dynamics and supporting various smart city applications. However, privacy concerns restrict access to large-scale and high-quality trajectory datasets. Trajectory generation provides a promising alternative by synthesizing realistic data to mitigate privacy risks. However, existing methods fail to explicitly capture travel patterns and can only generate fixed-length trajectories under a single condition. To address these limitations, we propose \textbfHTP, which \textbfHierarchically generates \textbfTravel patterns first and then generates GPS \textbfPoints by using large language models (LLMs), rather than directly generating GPS points. We first design a trajectory-specific residual quantization variational autoencoder (RQ-VAE) that quantizes micro-level GPS trajectories into compact, macro-level travel pattern tokens in a coarse-to-fine manner. These tokens capture rich segment spatial irregularities, such as point density variations caused by traffic conditions. Then, we extend the LLM vocabulary with travel pattern tokens to align trajectory representations with the LLM input, and apply supervised fine-tuning (SFT) to align the LLM with the trajectory generation task, enabling generation of travel pattern sequences under various conditions. Extensive experiments on two real-world datasets show that HTP outperforms the strongest baseline by an average of 29.78% in terms of generation quality. Our code is available at this https URL.
[AI-39] KairosAgent : Agent ic Time Series Forecasting with Fused Semantic Reasoning
链接: https://arxiv.org/abs/2605.30002
作者: Kun Feng,Ziwei Shan,Yuchen Fang,Yiyang Tan,Sihan Lu,Shuqi Gu,Lintao Ma,Xingyu Lu,Kan Ren
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Cross-domain multimodal time series forecasting is a challenging task, requiring models to integrate precise numerical comprehension, cross-domain semantic understanding, and effective multimodal fusion. Existing approaches either build Time Series Foundation Models (TSFMs) from scratch or leverage pretrained Large Language Models (LLMs). However, TSFMs often overlook semantic understanding and lack the ability to perform future-oriented semantic reasoning, and LLMs struggle with numerical comprehension and accurate quantitative forecasting. To overcome these limitations, we propose KairosAgent, a novel agentic framework for multimodal time series forecasting, including an LLM-based reasoner and a TSFM-based forecaster. KairosAgent unifies textual reasoning and numerical forecasting by dynamically invoking analytical tools to enhance the numerical understanding and semantic reasoning capabilities of LLMs. The reasoning results are subsequently fused into the TSFM pipeline, enabling more accurate and reliable future predictions. To further improve the reasoning, we curate a large-scale corpus of high-quality trajectories, alongside a reinforcement learning from forecasting paradigm with multi-turn refinement and turn-level credit assignment. Experiments demonstrate that KairosAgent achieves superior zero-shot forecasting performance while maximizing the utility of pretrained LLMs and TSFMs, presenting a promising direction for efficient and interpretable time series agents. The project page is at this https URL .
[AI-40] Cookie-Bench: Continuous On-screen Key Interaction Evaluation for Web Generation
链接: https://arxiv.org/abs/2605.30000
作者: Haoyue Yang,Zhangxiao Shen,Fan Ding,Hangting Lou,Yifeng Kou,Haoqing Yu,Jingyao Li,Zhengfan Wu,Siqi Bao,Jing Liu,Hua Wu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Front-end web code has become a core product surface for every frontier LLM release, yet evaluating these interactive applications at development speed remains costly because human-judged leaderboards like Arena do not scale. Existing automated proxies typically lean on reference implementations, test suites, or rigid checklists, and tend to miss the reasoned synthesis a human reviewer performs over a live session. We articulate a new evaluation regime that is simultaneously reference-free, autonomously driven, and holistically reasoned, and instantiate it through two artifacts. \textbf\dataname is an 11-domain, 54-leaf, 1,000-query WebDev benchmark spanning both static-presentation and interactive-application tasks, balanced across three difficulty tiers and three target-language groups, with briefs rewritten to resist recall from circulated prompts. \textbf\framename, grounded in Flavell’s metacognitive monitoring, separates evidence accumulation from judgment across three stages: Static Perception forms a first impression from passive observation; Agent-Driven Interaction explores the application autonomously while capturing continuous screen video, audio, and per-step screenshots; Dynamic Scoring issues holistic functionality and aesthetics verdicts with structured failure attribution only after the evidence chain is complete. On \dataname, \framename aligns closely with expert human ratings while surfacing substantial headroom across 13 frontier LLMs on interactive web generation. \noindenthttps://anonymous.this http URL
[AI-41] Accelerating Constrained Decoding with Token Space Compression EMNLP2026
链接: https://arxiv.org/abs/2605.29986
作者: Michael Sullivan,Alexander Koller
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 13 pages; 5 figures; under review at EMNLP 2026
Abstract:To guarantee that an LLM’s outputs conform to a specified structure, context-free grammar (CFG) decoding engines force the selection of next tokens that produce strings that conform to a given CFG. While current CFG-constrained decoding engines are highly optimized, the inherent costs arising from the massive per-step search space – i.e. the entire token vocabulary – result in intractably high overhead for more complex CFGs: precisely the situation where CFG engines are most useful. In this paper, we introduce CFGzip, an offline technique for compressing the token search space, which massively reduces CFG engine overhead. In experiments, we report latency reduction of up to two orders of magnitude when CFGzip is used with a SoTA grammar engine, yielding an up to 7.5x speedup in total constrained generation time: with CFGzip, constrained decoding is now feasible at scale for complex CFGs.
[AI-42] Compass: Navigating Global Marine Lead Data Integration through Expert-Guided LLM Agent
链接: https://arxiv.org/abs/2605.29966
作者: Yiming Liu,Bin Lu,Meng Jin,Ziyuan Sang,Shuo Jiang,Lei Zhou,Xinbing Wang,Chenghu Zhou,Jing Zhang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Marine lead (Pb) and its isotopes are critical tracers for ocean circulation and anthropogenic pollution, yet in-situ observations remain costly and sparse. While vast historical records exist, they lie buried within the unstructured content of academic papers, creating “data silos” inaccessible to comprehensive analysis. Manual extraction is unscalable, while general-purpose Large Language Models (LLMs) lack the necessary domain-specific knowledge, leading to hallucinations and scientifically invalid outputs. To address this, we introduce an expert-guided adaptation approach that enables LLMs to perform rigorous scientific data extraction without fine-tuning. We operationalize this approach through Compass, an LLM agent framework enhanced by a Knowledge Tree co-designed with marine scientists, which decomposes complex tasks into verifiable steps, guiding the agent’s reasoning to ensure scientific validity. Deploying Compass across a corpus of over 230,000 relevant open-access papers, we successfully extract 3,751 previously unincorporated Pb records. This effort establishes the largest integrated marine Pb database to date. Beyond standard metrics, Compass demonstrates superior reliability through multi-layered validation, achieving 92% accuracy as confirmed through expert manual verification. The newly integrated data expand coverage in previously under-sampled regions such as the East China Sea and the Southern Ocean, providing an enriched data foundation for future scientific discoveries. We release an interactive visualization platform to facilitate open scientific access. Our work demonstrates that expert-guided agents can effectively bridge the gap between general-purpose LLMs and high-stakes scientific domains, enabling scalable data discovery in geosciences.
[AI-43] Meta-Programming for Linear-time Temporal Answer Set Programming
链接: https://arxiv.org/abs/2605.29965
作者: Susana Hahn,Amade Nems,Javier Romero,Torsten Schaub
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:The development of temporal extensions of Answer Set Programming (ASP) has led to the emergence of non-monotonic linear-time (TEL), dynamic (DEL), and metric (MEL) temporal equilibrium logics. However, the inherent rigidity of highly optimized ASP systems often hinders the rapid exploration and implementation of alternative logical designs. In this work, we propose a flexible meta-programming framework that operationalizes the semantics of varied temporal logics through a unified, declarative framework. Our approach extends standard ASP meta-programming by augmenting clingo’s theory grammar with formal type specifications and nesting capabilities. To ensure semantic correctness, we introduce a transformation pipeline that protects nested modalities from stable-model-based simplifications during grounding. We demonstrate the extensibility of our framework by implementing meta-encodings for TEL, MEL, and DEL. We provide a comprehensive account of TEL and highlight the key features for managing the interval constraints of MEL and the Fischer-Ladner closure in DEL. Finally, we introduce the metasp system, a versatile tool that encapsulates this workflow.
[AI-44] Honeyval: A Comprehensive Evaluation Framework for LLM -powered HTTP Honeypots
链接: https://arxiv.org/abs/2605.29963
作者: Mark Vero,Fabian Kaczmarczyck,Ivan Petrov,Ilia Shumailov,Jamie Hayes,Niels Heinen,Tianqi Fan,Luca Invernizzi,Martin Vechev
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Honeypots are decoy systems mimicking real system components designed to defend against cyber attacks. Recently, LLMs increasingly serve as simulation backbones for honeypots. They enable defenders to construct high-interaction honeypots with low system security risks. However, LLM-powered honeypot development lacks a unified evaluation framework. Most evaluations consist of measuring response similarity on fixed commands, manual testing, or real-world deployment. These methods are often not scalable for development, reproducible across evaluations, representative of practical attacks, or adaptable to various attacker and honeypot configurations. In this work, we bridge this gap and propose Honeyval, a comprehensive evaluation framework for LLM-powered HTTP honeypots. We address the limitations of prior evaluations by grounding the honeypots in 16 backend applications, using AI hacking agents as attackers, employing two control tasks to monitor agent and honeypot capabilities across customizations, and defining clear and verifiable exploit goals for the attacker. Using Honeyval, we conduct an extensive evaluation of recent cost-efficient LLMs as HTTP honeypots. Our experiments highlight the promise of LLM-powered honeypots; they lead to substantially longer interactions with the attacker than rule-based baseline honeypots and are far less frequently detected even by frontier models, all while, on average, preserving a running cost advantage against agentic attackers. Further, we experiment with different counter-offensive honeypots configurations, and observe unique trade-offs, such as longer interactions at the cost of increased detection.
[AI-45] Hijacking Agent Memory: Stealthy Trojan Attacks Through Conversational Interaction
链接: https://arxiv.org/abs/2605.29960
作者: Hongtao Wang,Se Yang,Yu Chen,Puzhuo Liu
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 19 pages, 12 figures
Abstract:Large language model (LLM) agents increasingly leverage long term memory to support persistent and autonomous task execution. However, this capability also introduces a new attack surface: memory poisoning, where adversaries can inject malicious information to influence future behavior. Existing memory poisoning attacks often assume that injected content can be stored directly in memory, overlooking the selective extraction and rewriting stages in modern memory pipelines. This makes prior methods ineffective under realistic settings. In this paper, we propose MemPoison, a novel memory poisoning attack that bypasses selective memory mechanisms in LLM agents, where an attacker can inject triggerable backdoors into the agent’s long-term memory through dialogue interactions, thereby misleading its subsequent responses. MemPoison introduces three key components: (i) a semantic relational bridge that binds the trigger and payload into a coherent statement to ensure they are extracted into memory together; (ii) entity masquerading that optimizes triggers to mimic named entities, resisting rewriting; and (iii) joint embedding optimization that shapes trigger-injected texts into a tight cluster in the embedding space while maintaining isolation from benign embeddings for stealth. Evaluations across different agent domains and memory mechanisms show MemPoison achieves attack success rates up to 0.95, outperforming existing baselines. Mechanistic analysis indicates that the attack exploits embedding-space anisotropy and shifts attention patterns, highlighting core vulnerabilities in selective memory systems. We evaluate multiple defense strategies and demonstrate their fundamental limitations in mitigating the attack. Comments: 19 pages, 12 figures Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI) Cite as: arXiv:2605.29960 [cs.CR] (or arXiv:2605.29960v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2605.29960 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-46] Formalizing Mathematics at Scale
链接: https://arxiv.org/abs/2605.29955
作者: Ahmad Rammal,Niket Patel,Fabian Gloeckle,Amaury Hayat,Julia Kempe,Remi Munos,Charles Arnal,Vivien Cabannes
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We present AutoformBot, a multi-agent system for building an Autoformalized Textbook Library At Scale (Atlas) in Lean 4. AutoformBot orchestrates thousands of LLM agents, equipped with formal verification tools, dependency-aware task scheduling, and collaborative version control, to translate informal textbook prose into machine-checked definitions and proofs. We apply our methods to a corpus of 26 open-access textbooks spanning analysis, algebra, topology, combinatorics, and probability, producing Atlas: a verified library of over 45,000 Lean 4 declarations and 500 thousand lines of code. We release two artifacts: (i) AutoformBot, the open-source multi-agent framework; and (ii) Atlas, the resulting formal library. Our results suggest that autoformalizing the core content of graduate-level mathematics at scale is now economically and technically feasible. This opens the door to the automated verification of both human- and machine-generated mathematics at a research level.
[AI-47] HoliTok:A Coutinuous Holistic Tokenization with Robust Dual Capabilities of Speech Generation and Understanding
链接: https://arxiv.org/abs/2605.29948
作者: Bohan Li,Shi Lian,Hankun Wang,Yiwei Guo,Yu Xi,Zhihan Li,Da Zheng,Colin Zhang,Kai Yu
机构: 未知
类目: ound (cs.SD); Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注: 14 pages, 2 figures, 8 tables
Abstract:Unified speech foundation models require a holistic tokenization space that is both learnable by language models and decodable into high-quality waveforms. Existing speech tokenizers, however, often fail to satisfy these requirements simultaneously, leading to increased architectural complexity and more involved training designs. We propose HoliTok, a continuous Holistic speech Tokenization model designed for unified generation-understanding modeling. HoliTok encodes 48~kHz speech into a compact 25~Hz sequence of 128-dimensional latents. It is trained with a progressive strategy that jointly preserves signal-level fidelity, incorporates semantic information, and maintains strong latent learnability. Based on this tokenization, we build a unified AR+DiT model for speech synthesis and recognition, where the same latent sequence supports both generation-specific and unified generation-understanding tasks. Experiments show that HoliTok achieves competitive reconstruction fidelity, improves generative learnability for high-quality and controllable synthesis, and, among the evaluated representations, is the only one that operates robustly in our unified generation-understanding architecture without additional optimization tricks. These results suggest that HoliTok serves as an effective speech tokenizer and a foundational representation interface for unified spoken language modeling. The code is available at: this https URL.
[AI-48] Make LLM Learn to Synthesize from Streaming Experiences through Feedback
链接: https://arxiv.org/abs/2605.29940
作者: Zhenlin Hu,Yan Wang,Zhen Bi,Zihao Xue,Bingyu Zhu,Longtao Huang,Xiongtao Zhang,Zeyu Yang,Zhixuan Chu,Jungang Lou
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) have been widely adopted for synthetic data generation, significantly reducing annotation costs. However, most existing studies treat synthesis as a set of isolated tasks and overlook a more fundamental question: whether a model can learn to synthesize by accumulating experience from past tasks and transferring it to future ones. In this work, we introduce StreamSynth, a new setting in which synthesis tasks arrive sequentially and experience from historical tasks provides informative signals for future synthesis. To address this setting, we propose SynLearner, a general framework that enables synthesis models to acquire reusable synthesis experience over a task stream. Instead of generating data independently for each task, SynLearner encourages the model to explore diverse synthesis patterns, learn from feedback, and balance sample quality with set-level diversity as tasks evolve. Extensive experiments across multiple benchmarks show that SynLearner effectively leverages experience from earlier tasks to improve synthesis performance on later ones, exhibiting consistent cross-task transferability. These findings provide evidence for the feasibility of StreamSynth and highlight synthetic data generation as an experience-driven process that can benefit from task streams.
[AI-49] Its All About Speed: AIs Impact on Workflow in Music Production
链接: https://arxiv.org/abs/2605.29931
作者: Finn McClellan,Fabio Morreale
机构: 未知
类目: Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS)
备注: Audio Engineering Society Conference Paper - Presented at the AES International Conference on Machine Learning and Artificial Intelligence for Audio 2025 - September 8-10, London, UK
Abstract:In this paper, we present the results of an ethnographic study into the impact of AI and automated tools on music production workflow. Focusing specifically on professional participants who identified as recording engineers, mixers, and producers, we discuss their usage of common AI and automated software, as well as their sentiments on the proliferation of these tools. We discuss tensions that may be created between users and automated tools in key areas such as the need for speed and efficiency, controllability, and maintaining creative agency, and how these tensions may be alleviated through tool design.
[AI-50] Selection Hyper-heuristics Can Automatically Adjust the Learning Period to Optimally Solve Pseudo-Boolean Problems
链接: https://arxiv.org/abs/2605.29916
作者: Benjamin Doerr,Pietro S. Oliveto,John Alasdair Warwicker
机构: 未知
类目: Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI); Data Structures and Algorithms (cs.DS); Optimization and Control (math.OC)
备注: To appear in “Artificial Intelligence”
Abstract:The Random Gradient hyper-heuristic was recently shown to be able to learn the optimal neighbourhood size when optimizing the LeadingOnes benchmark via the Randomised Local Search (RLS) meta-heuristic. However, for this to happen, a learning period of a certain length \tau had to be used, differently from classic hyper-heuristics, which change their behaviour based on the success of only the previous iteration. In this paper, we show how to automatically set this new parameter value, relieving the user from the non-trivial task of controlling this novel algorithm parameter. We prove that the resulting hyper-heuristic selects the optimal neighbourhood size in a 1-o(1) fraction of the iterations and, consequently, optimises the LeadingOnes benchmark in the best possible time (apart from lower-order terms) achievable with these neighborhood sizes.
[AI-51] Agora: Toward Autonomous Bug Detection in Production-Level Consensus Protocols with LLM Agents
链接: https://arxiv.org/abs/2605.29910
作者: Xiang Liu,Sa Song,Zhaowei Zhang,Huiying Lan,Jason Zeng,Ming Wu,Michael Heinrich,Yong Sun,Ceyao Zhang
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 35 pages, 4 figures
Abstract:Consensus protocols form the backbone of distributed systems and blockchains, where implementation bugs can cause data corruption and financial losses. While LLM-based approaches show promise in code analysis, they struggle with deep protocol-level logic bugs involving complex state-dependent behaviors across multiple execution stages. We present Agora, a domain-aware multi-agent framework that integrates hypothesis-driven testing with LLM capabilities for systematic protocol verification. Agora employs specialized agents that collaboratively explore protocol state spaces, synthesize attack scenarios using domain-specific constraints, and validate findings through iterative refinement. This explicit role separation enables reasoning about global protocol invariants beyond single-function code analysis. We evaluate Agora on four consensus implementations (Raft, EPaxos, HotStuff, BullShark) using four state-of-the-art LLMs. Agora discovers 15 previously unknown protocol-level logic bugs that violate safety properties, while existing LLM-based agents fail to detect any such protocol-level logic bugs. Our results demonstrate that domain-aware multi-agent collaboration is essential for detecting deep logic bugs in complex protocols.
[AI-52] Redundant or Necessary? A Benchmark for Detecting Redundant Steps in Agent Trajectories
链接: https://arxiv.org/abs/2605.29893
作者: Minyang Hu,Bo Yang,Zhinuo Zhou,Jiachen Liang,Guo Jiahao,Yiyang Yin,Xiongwei Han
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLM-based agents have demonstrated strong capabilities in solving complex tasks through multi-step reasoning and tool use. However, existing evaluation protocols primarily focus on task success, overlooking a critical aspect of agent behavior: execution efficiency. In practice, agent trajectories often contain redundant steps that consume substantial resources while contributing little to task completion. In this work, we propose and formulate a new research area: \textbfredundant step detection for agent trajectories. To support this initiative, we introduce \textbfRedundancyBench, a new benchmark that contains diverse tasks with carefully annotated trajectories, where each step is labeled according to its contribution to task completion. Using RedundancyBench, we develop and evaluate 3 representative methods to answer whether a step within trajectory is redundant or necessary. Our results show that even the best-performing method achieves only 24.88% score in detecting redundant steps, while some methods perform worse than random guessing. These results highlight the task’s complexity and the need for further research in this area. \footnoteCode and dataset in this paper are both available in \hrefthis https URLthis https URL.
[AI-53] LaRA: Layer-wise Representation Analysis for Detecting Data Contamination in RL Post-Training
链接: https://arxiv.org/abs/2605.29888
作者: Minju Gwak,Minseo Kwak,Dongseok Lee,Guijin Son,Alan Ritter,Jaehyung Kim
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Work in Progress
Abstract:Reinforcement learning (RL) post-training has shown to improve reasoning in large language models (LLMs). However, there has been little exploration on the problem of data contamination in RL post-training, potentially undermining generalization and evaluation reliability of the training process itself. Existing detection methods primarily rely on output-level signals such as likelihood or entropy, which become unreliable for RL-trained models since RL shapes behavior through trajectory-level rewards rather than token likelihoods. We propose LaRA, a layer-wise representation analysis framework for detecting contamination in RL post-trained LLMs. LaRA introduces three complementary metrics, measuring perturbation sensitivity, directional collapse, and local representation rigidity under controlled perturbations. We find that contamination produces progressive geometric deviations across layers, including amplified perturbation sensitivity, stronger directional collapse, and enhanced local rigidity. Based on our findings, we also develop a contamination detection protocol that aggregates representation-level deviations across layers and metrics. Experiments on RL-trained reasoning models show that our protocol outperforms existing output-level baselines for contamination detection.
[AI-54] Moment-KV: Momentum-Based Decode-Time KV Cache Compression for Long Generation
链接: https://arxiv.org/abs/2605.29873
作者: Soumyadeep Jana,Sagar Nishad,Sanasam Ranbir Singh
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Key-Value (KV) cache remains a major bottleneck for deploying Large Language Models (LLMs) in long-generation tasks. Prior work often applies uniform compression across both prefill and decoding caches, but compressing the prefill cache degrades performance by corrupting critical context. While preserving the prefill cache is essential, decoding-phase compression remains underexplored, with existing methods relying on rigid recency windows or instantaneous attention. Our analysis of attention dynamics reveals strong temporal patterns: critical tokens receive sustained attention over long horizons, while local reasoning involves short-lived bursts. Static heuristics fail to capture this behavior, leading to premature eviction of important tokens or retention of stale ones. We propose Moment-KV, a decoding-time KV cache compression method based on momentum-driven temporal attention aggregation. Our method models token importance as a continuously evolving state, where attention is aggregated with decay, capturing both long-term influence and recent relevance. Experiments show that Moment-KV significantly improves generation fidelity in long-generation tasks (2.3-3.2 %) while maintaining decoding latency.
[AI-55] ESPO: Early-Stopping Proximal Policy Optimization
链接: https://arxiv.org/abs/2605.29860
作者: Zihang Li,Rui Zhou,Yingcheng Shi,Wenhan Yu,Zhewen Tan,Zixiang Liu,Zeming Li,Binhua Li,Yongbin Li,Tong Yang,Jieping Ye
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:When a large language model under reinforcement learning commits a wrong reasoning step early in a trajectory, standard algorithms force it to keep generating until the maximum horizon, spending compute on tokens that never receive positive reward and polluting advantage estimates with post-failure noise. We propose ESPO (Early-Stopping Proximal Policy Optimization), which detects trajectory failure on-the-fly and terminates rollouts early. At each generation step, ESPO computes a surrogate regret using only the logits already computed during sampling, and terminates when the smoothed cumulative regret significantly exceeds its estimated values. Truncated trajectories are treated as absorbing failure states with a terminal reward, concentrating negative temporal-difference (TD) errors near the detected failure step without any additional reward model or human annotation. On DeepSeek-R1-Distill-Qwen-7B trained for mathematical reasoning, ESPO surpasses PPO on AIME~2024 (46.28% vs. 45.25%), AMC~2023 (85.83% vs. 82.94%), and MATH-500 (87.42% vs. 85.43%), while saving more than 20% rollout tokens cumulatively.
[AI-56] HARP: Hadamard-Preconditioned Adaptive Rotation Processor for Extreme LLM Quantization
链接: https://arxiv.org/abs/2605.29843
作者: Artur Zagitov,Gleb Molodtsov,Aleksandr Beznosikov
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Post-training quantization (PTQ) is essential for deploying LLMs under memory and bandwidth constraints. However, extreme low-bit quantization remains highly sensitive to activation outliers and anisotropic weight curvature. Existing incoherence-based PTQ methods mitigate this issue with fixed randomized Hadamard transforms (RHTs), which improve quantization robustness but cannot adapt the rotated basis to the layer, calibration distribution, or quantizer. We introduce HARP (Hadamard-preconditioned Adaptive Rotation Processor), a learnable structured two-sided orthogonal processor that replaces fixed Hadamard mixing while preserving exact full-precision equivalence. HARP represents each rotation as a product of sparse butterfly-like block-orthogonal stages, supports non-power-of-two dimensions via Mixed-Radix schedules, and initializes to the RHT processor up to a fixed permutation. Fitted only on calibration data, HARP adapts the quantization basis to each layer and backend. Across 2-4 bit settings on models ranging from 1B to 70B parameters, HARP improves perplexity and zero-shot accuracy over fixed RHT. Importantly, HARP preserves deployment efficiency, reaching 128 tok/s versus 61 tok/s for FP16.
[AI-57] CB-SLICE: Concept-Based Interpretable Error Slice Discovery ICML2026
链接: https://arxiv.org/abs/2605.29836
作者: Yael Konforti,Mateo Espinosa Zarlenga,Elaf Almahmoud,Mateja Jamnik
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: 20 pages, 7 figures, 12 tables, to be published at Proceedings of the 43rd International Conference on Machine Learning (ICML 2026)
Abstract:Despite strong average-case performance, deep learning models often exhibit systematic errors on specific population groups, known as error slices. Identifying these groups and the root causes of their failures is critical for model debugging and bias mitigation. However, existing error Slice Discovery Methods (SDMs) typically generate explanations disconnected from the model’s inference process, thus only approximating the underlying error source and may be inaccurate. We address this limitation by leveraging Concept Bottleneck Models (CBMs), whose predictions are directly dependent on human-understandable semantic concepts. Since downstream task failures in CBMs commonly arise from concept mispredictions, concept representations provide a strong candidate for error slice identification, offering fine-grained explanations directly linked to the error source. Building on this insight, we introduce CB-SLICE, a concept-based SDM that groups samples with shared concept prediction failures and identifies the keyword concepts most responsible for each slice’s failure mode. Across multiple benchmarks, we show that CB-SLICE outperforms state-of-the-art methods in uncovering well-known biases while providing richer and more faithful explanations of model errors. Comments: 20 pages, 7 figures, 12 tables, to be published at Proceedings of the 43rd International Conference on Machine Learning (ICML 2026) Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML) Cite as: arXiv:2605.29836 [cs.LG] (or arXiv:2605.29836v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2605.29836 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-58] OmniMatBench: A Human-Calibrated Multimodal Reasoning Benchmark Across 19 Materials Science Subfields
链接: https://arxiv.org/abs/2605.29833
作者: Wanhao Liu,Jiaqing Xie,Qian Tan,Weida Wang,Jue Wang,Ran Sun,Zhuo Yang,Wanli Ouyang,Lei Bai,Tianfan Fu,Lu Chen,Xin Chen,Yuqiang Li
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 22 Pages
Abstract:As multimodal language models play an increasingly important role in scientific research, materials science offers a critical testbed due to its interdisciplinary, multimodal, and application-driven nature. However, existing materials benchmarks mainly focus on property prediction, knowledge QA, or characterization understanding, leaving the broader reasoning process from materials knowledge to application underexplored. To fill this gap, we present OmniMatBench, a human-calibrated multimodal reasoning benchmark for materials science. OmniMatBench contains 3,171 expert-curated QA and calculation problems across 19 materials-science subfields, spanning fundamental materials knowledge, structural and engineering materials, materials processing and manufacturing, and functional and applied materials. We evaluate 13 open-source and closed-source MLLMs and find that the best model achieves only a 0.372 overall score, revealing a substantial gap in current materials-science reasoning. Further analysis shows strong variation across subfields, fixed reasoning heuristics, uneven materials knowledge, and limited high-level knowledge application under formula-, retrieval-, and code-assisted settings. OmniMatBench provides crucial insights into the capabilities and limitations of current MLLMs and establishes a foundation for reliable AI assistants in materials-science research.
[AI-59] OptSkills: Learning Generalizable Optimization Skills from Problem Archetypes via Cluster-Based Distillation
链接: https://arxiv.org/abs/2605.29829
作者: Haochen Yang,Ke Zhao,Mengyuan Ma,Xingyu Lu,Xiangfeng Wang,Hong Qian
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 22 pages, 10 figuers, project: this https URL
Abstract:Leveraging Large Language Models (LLMs) to automatically formulate and solve optimization problems from natural language has emerged as an efficient paradigm for automated optimization. However, existing methods still exhibit limited generalization: they are sensitive to superficial narrative variations, reuse experience mainly at the case level, and struggle to adapt to shifted or emerging problem types. We propose OptSkills, an archetype-centric skill learning and reasoning agent system for optimization modeling and solving. To improve robust generalization, our system clusters problems by their underlying archetypes rather than surface narratives. To improve in-distribution generalization, it explores diverse modeling paradigms and solver configurations within each cluster, then distills successful trajectories into reusable workflow-level skills. To improve out-of-distribution generalization, it refines existing skills or expands the skill library using newly obtained trajectories. Our system achieves a state-of-the-art micro-averaged accuracy of 68.27% on datasets encompassing diverse problem types and scenarios. In addition, on MIPLIB-NL, a highly challenging large-scale and high-dimensional benchmark, it achieves 26.91% accuracy, outperforming DeepSeek-V3.2-Thinking by 4.53%. After skill learning on Nano-CO, it reaches 72.79% on the OOD NLCO benchmark. Code and skills are available at this https URL.
[AI-60] Quantifying and Optimizing Simplicity via Polynomial Representations ICML2026
链接: https://arxiv.org/abs/2605.29823
作者: Tianren Zhang,Xiangxin Li,Minghao Xiao,Guanyu Chen,Feng Chen
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: ICML 2026
Abstract:Deep networks often exhibit a preference for “simple” solutions, and such a simplicity bias is widely believed to play a key role in generalization. Yet a broadly applicable, quantitative measure of simplicity remains elusive. We introduce polynomial representations as a distribution-aware, low-dimensional surrogate for neural functions: we approximate a network’s predictive behavior along data-dependent interpolation paths using orthogonal polynomial bases, yielding a compact functional representation. We show that the effective degree of this representation serves as a practical simplicity metric that is predictive of generalization across tasks and architectures, and consistently outperforms existing generalization proxies such as sharpness. Finally, polynomial representations naturally yield a differentiable simplicity regularizer, which consistently improves generalization in image and text classification, fine-tuning contrastive vision-language models, and reinforcement learning.
[AI-61] Inferring Code Correctness from Specification
链接: https://arxiv.org/abs/2605.29822
作者: Tambon Florian,Papadakis Mike
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) have become integral to modern software development, enabling automated code generation at scale. However, validating the correctness of LLM-generated code remains a critical and largely unsolved challenge. Existing approaches either rely on dynamic consensus across multiple code candidates - making them costly and difficult to scale - or on static reasoning that is susceptible to dynamic bugs and order bias. In this paper, we propose TRAILS~ (Targeted Reasoning Agreement via Inputs and Specifications), an approach that grounds LLM reasoning with concrete (input, output) pairs. TRAILS~ first generates diverse test inputs via category partitioning based on the specification, then executes them against the candidate code and prompts LLMs to assess whether the resulting input-output pairs conform to the specification - without ever reasoning over the code itself. Scores are aggregated across inputs, to determines whether the program is likely correct. We evaluate TRAILS~ on two datasets, LiveCodeBench and CoCoClaNeL, across three LLMs (Qwen3Coder-30B, Devstral-Small-24B, and Olmo3.1-Instruct), comparing against HoarePrompt and a Zero-Shot Chain-of-Thought baseline. TRAILS~ improves Matthew Correlation Coefficient by up to 39% relative to Zero-Shot COT and consistently outperforms HoarePrompt. Beyond accuracy, TRAILS~ demonstrates greater stability across seeded runs, reducing sensitivity to LLM non-determinism, and assigns correct labels to a larger set of unique code samples than competing approaches.
[AI-62] Harnessing non-adversarial robustness in large language models
链接: https://arxiv.org/abs/2605.29816
作者: Qinghua Zhou,Ellina Aleshina,Andrey Lovyagin,Oleg Somov,Mikhail Seleznyov,Alexander Panchenko,Ivan Oseledets,Elena Tutubalina,Ivan Y. Tyukin
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:The work presents an approach for addressing the challenge of robustness in Large Language Models (LLMs) to alterations and potential errors caused by semantically similar but textually different prompts. Recent works have shown that these kinds of prompt variations can significantly impact the performance of LLMs on tasks. The central question is: can LLMs’ robustness to semantically-neutral prompt alterations be acquired without expensive retraining of the entire model? We address this question both theoretically and through experiments. Our theoretical analysis reveals a crucial factor impacting model robustness - a systematic expected shift or perturbation-induced bias in neural network module outputs. Motivated by this analysis, we show that robustness can be achieved via a simple fine-tuning process: debiasing for robustness. We identify conditions when debiasing helps and when it does not, and demonstrate, through both theory and extensive experiments, that debiasing for robustness may indeed be a quick and efficient tool to enhance robustness and provide certification against random prompt perturbations.
[AI-63] MEMENTO: Leverag ing Web as a Learning Signal for Low-Data Domains
链接: https://arxiv.org/abs/2605.29795
作者: Ashutosh Ojha,Vinay Aggarwal,Ashutosh Srivastava,Siddharth Yedlapati,Yaman K Singla,Jitendra Ajmera
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Real-world tasks often lack large labeled datasets, motivating extensive work on learning in low-data regimes. However, existing approaches such as few-shot prompting, instruction tuning, and synthetic data generation, continue to treat labeled or pseudo-labeled data as the primary learning signal. In contrast, human practitioners acquire expertise through repeated, self-directed interaction with the open web, progressively refining both domain knowledge and search strategies. We propose MEMENTO, a framework that treats the web as a learning signal rather than a stateless retrieval interface. MEMENTO operates at two levels: within each session, it conducts iterative web exploration via an Adaptive Exploration Tree (AET) that decomposes tasks into evolving questions and reflects on intermediate findings; across sessions, it accumulates experience through dual-channel memory, separating declarative knowledge (facts) from procedural knowledge (search strategies). This design enables agents to learn reusable research strategies and domain expertise from trajectories of web interaction without additional model training. We evaluate MEMENTO on two low-data professional domains: sales automation and legal research. Our empirical results show consistent improvements in performance over ReAct based baselines (+25.6% on sales automation and 36.5% on legal research), demonstrating that the web can serve as a scalable learning source for acquiring task-specific expertise in data-scarce settings.
[AI-64] SkillsInjector: Dynamic Skill Context Construction for LLM Agents
链接: https://arxiv.org/abs/2605.29794
作者: Yanchao Li,Wanhao Liu,Ben Gao,Jiaqing Xie,Zhehong Ai,Na Zou,Yuqiang Li,Tianfan Fu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLM agents now draw on growing skill libraries to handle complex tasks. However, injecting more skills does not always improve task completion and can even degrade it. Existing methods still treat skill injection as a static step, selecting skills with fixed criteria, fixing the budget in advance, and leaving descriptions unchanged. We argue that this static treatment can undermine the utility of skills, because which skills are exposed, how many are included, and how they are presented all affect downstream performance. We propose SkillsInjector, a two-stage adaptive method that jointly addresses these decisions. First, a context planner learns execution-grounded skill preferences and admits an adaptive number of skills for each task. A set-aware renderer then tailors how selected descriptions are presented relative to their co-injected neighbors. Across tau2-bench, SkillsBench, and ALFWorld, SkillsInjector achieves the highest score, improving over the strongest baseline by 3.9, 6.1, and 7.3 percentage points, respectively. Ablation studies show that skill selection, adaptive budgeting, and set-aware rendering each contribute to the gain. These results show that skill-augmented agents benefit from optimizing the injected context itself. Code will be released upon publication
[AI-65] Certified Policy Optimisation for Nested Causal Bandits via PAC-Bayes Risk
链接: https://arxiv.org/abs/2605.29788
作者: Tim Woydt,Paul-David Zuercher
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Critical sequential decisions are rarely single-timescale: a strategic decision causally shapes the context in which every subsequent tactical choice is made; standard bandit and reinforcement-learning theory does not capture this causal coupling between timescales. We formalise the problem class as Nested Contextual Causal Bandits (NCCBs), a hierarchical SCM where each level’s action sets the next level’s context distribution, and propose Nested Causal Thompson Sampling (NCTS), which draws one mechanism-factorised belief per episode and acts recursively under it. Our main theoretical result is a causal PAC-Bayesian excess-risk bound that certifies any candidate deployment policy from historic data alone, off-policy and anytime, answering the deployment question: can we trust this agent here, and at what risk? Experiments on a hierarchical SCM show that, against a matched RFF-GP joint regression on the same function class, the factorised SCM-mechanism posterior transfers significantly better zero-shot under exogenous distribution shifts, the recursive meta-to-inner commit significantly dominates the joint-commit alternative in distribution, and the certificate significantly contracts as offline data accumulates. Combining these results, we establish progressive certified handover, a safe-deployment method: each timescale flips from a legacy controller to NCTS when gains can be certified, independently of the others.
[AI-66] Croissant Tasks: A Metadata Format for Reproducible Machine Learning Evaluations
链接: https://arxiv.org/abs/2605.29786
作者: Omar Benjelloun,Leonardo Martins Bianco,Isabelle Guyon,Thanh Gia Hieu Khuong,Jonathan Lebensold,Sebastian Lobentanzer,Luis Oala,Benedictus Kent Rachmat,Ihsan Ullah,Peyman Vahidi,Joaquin Vanschoren
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 10 pages, 4 figures
Abstract:Reproducibility is fundamental to the scientific method, yet remains a critical challenge in machine learning. Contributing factors include underspecified execution details and brittle software environments. Human-centric remedies, such as checklists and manual verification, help but require intensive effort and fail to scale. To address this, we introduce Croissant Tasks: a declarative, machine-actionable metadata format that abstracts low-level implementation details into high-level specifications. This format enables conceptual reproducibility: verifying claims via independent, agent-generated implementations rather than brittle source code replication. We contribute: (1) the Croissant Tasks specification, formally decoupling task problem from solution; (2) an automated LLM pipeline that retrofits existing benchmarks into this format; and (3) empirical validation showing autonomous agents can ingest these specifications to generate functional, accurate reproduction pipelines from scratch. We envision this format as a new foundation for automated and conceptual reproducibility in machine learning.
[AI-67] From XXLTraffic to EvoXXLTraffic: Scaling Traffic Forecasting to Sensor-Evolving Networks
链接: https://arxiv.org/abs/2605.29768
作者: Du Yin,Hao Xue,Arian Prabowo,Shuang Ao,Flora Salim
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: Under Review
Abstract:Existing traffic forecasting benchmarks assume a fixed sensor set, but real road-sensor networks grow continuously as the road network changes year by year. We introduce the XXLTraffic dataset family, which spans up to 27 years of California PeMS and Transport for NSW data. The fixed-sensor subsets of XXLTraffic support extremely long forecasting with multi-year gaps and standard hourly / daily long-horizon forecasting. We extend it to EvoXXLTraffic, a sensor-evolving reorganization that exposes per-year active sensors, yearly traffic-flow matrices, and yearly graph snapshots across nine PeMS districts, with growth ratios ranging from +305% to over +10,000%. We define a yearly streaming forecasting protocol on EvoXXLTraffic in which each calendar year is a continual task, and benchmark a wide range of representative baselines drawn from static spatio-temporal GNNs, naïve online schemes, evolving-graph continual methods, and retrieval / test-time methods. We find that our ultra-large evolutionary dataset better reflects the real world, and many state-of-the-art (SOTA) results no longer work. Our dataset complements existing benchmarks by enabling more realistic forecasting under ultra-long evolutionary road networks.
[AI-68] LFQ: Logit-aware Final-block Quantization for Boosting the Generation Quality of Low-Bit Quantized LLM s ICML2026
链接: https://arxiv.org/abs/2605.29756
作者: Jung Hyun Lee,June Yong Yang,Jungwook Choi,Eunho Yang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: Accepted to ICML 2026
Abstract:As large language models continue to scale, low-bit weight-only post-training quantization (PTQ) offers a practical solution to their memory-efficient deployment. Although block-wise PTQ is capable of matching the full-precision (FP) baseline on basic language modeling and understanding, its quality is degraded for generative tasks – especially at longer responses and extended chains of thought, which is critical in boosting task accuracy. We attribute this shortfall to two factors: (i) the omission of the unembedding layer (the LM head) in block-wise optimization and (ii) the reliance on the mean squared error (MSE) objective. Both factors cause the token probability distribution of the quantized model to misalign with that of the FP model, yielding notable accuracy drops on text generation benchmarks. To rectify the discrepancy, we introduce Logit-aware Final-block Quantization (LFQ), a simple yet effective enhancement to block-wise PTQ that quantizes the final Transformer block by minimizing the cross-entropy between the logits of the FP model and those of its quantized counterpart. By aligning token probabilities at the logit level in the final block, LFQ consistently improves the accuracy of complex generation tasks over state-of-the-art block-wise PTQ across diverse model families, while maintaining parity with FP baselines on language modeling and understanding.
[AI-69] Benchmarking Positional Encoding Strategies for Transformer-Based EEG Foundation Models
链接: https://arxiv.org/abs/2605.29754
作者: Ayse Betul Yuce,Sebastian Stober
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Electroencephalography (EEG) is a widely used non-invasive technique for measuring brain activity in brain-computer interface (BCI) applications. Supervised EEG decoding models often struggle to generalize across tasks, subjects, and datasets, motivating transformer-based EEG foundation models trained with self-supervised learning. Since transformers are permutation-invariant, they require explicit positional information. Unlike textual tokens, EEG electrodes are spatially distributed across the scalp, raising the question of how electrode positions should be encoded in transformer-based EEG models. In this study, we benchmark five positional encoding strategies within the CBraMod backbone and evaluate them under linear probing and fine-tuning protocols on motor imagery classification and emotion recognition. Our results show that no single strategy consistently outperforms across tasks. Spherical Positional Encoding (SPE) yields strong representations for motor imagery but underperforms on emotion recognition, while Asymmetric Conditional Positional Encoding (ACPE) demonstrates more consistent performance across tasks. These findings suggest that the optimal positional encoding strategy is task-dependent, with no universal solution across EEG decoding scenarios.
[AI-70] Citation-Closure Retrieval and Per-Rule Attribution for Real-World Regulatory Compliance Question Answering
链接: https://arxiv.org/abs/2605.29742
作者: Yeong-Joon Ju,Seong-Whan Lee
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: Under Review
Abstract:Deploying Large Language Models (LLMs) for regulatory compliance demands rigorous traceability via comprehensive citations across multi-tiered authority structures. Unlike traditional multi-hop or legal QA, this task requires structured procedural lookups and evidence-set closure rather than entity resolution or case-law reasoning. Existing RAG systems struggle here due to flattened citation edges, fragmented retrieval expansions, and fragile post-hoc attribution. We formalize Regulatory Compliance QA with RegOps-Bench, a novel benchmark featuring an Operational Knowledge Graph derived from complex national R\D regulations. To address these bottlenecks, we propose RefWalk, a unified framework driven by a shared topic anchor. RefWalk traverses cross-document citations, fuses multi-view candidates via max-based aggregation, and enforces per-rule attribution to explicitly map claims to sources. We establish a strong baseline with substantial improvements in retrieval recall and citation accuracy. Finally, a contrastive evaluation on a U.S. health compliance dataset (HIPAA) reveals that existing systems exhibit saturation on flat-structure rules, underscoring the need for RegOps-Bench. Our code is available at this https URL.
[AI-71] Uncertainty-Aware Transfer Learning for Cross-Building Energy Forecasting: Toward Robust and Scalable District-Level Energy Management
链接: https://arxiv.org/abs/2605.29733
作者: Shadmehr Zaregarizi,Khashayar Yavari
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 5 pages, 3 figures, 2 tables. Accepted at BALANCES’26 (6th ACM International Workshop on Big Data and Machine Learning for Smart Buildings and Cities), Banff, Alberta, Canada, June 22, 2026. This is the author’s accepted manuscript; final published version DOI will be activated after June 22, 2026
Abstract:Scaling data-driven energy forecasting to district level requires models that can be re-used across buildings with minimal target-domain data and honest uncertainty estimates. We present an uncertainty-aware transfer learning (TL) framework for cross-building energy forecasting based on the Temporal Fusion Transformer (TFT), evaluated on a newly released high-resolution real sub-meter dataset: an educational building at Aalborg University, Denmark (source) and the multi-typology NEST building at EMPA, Switzerland (target). We introduce the Transfer Robustness Index (TRI), an architecture-agnostic metric for quantifying generalization quality across domain gaps. A four-strategy layer-freezing ablation shows that Probe-Only fine-tuning, updating only 455 output-layer parameters out of 806K, achieves the best transfer quality (TRI = 3,097), outperforming full fine-tuning and suggesting that TFT encoders learn transferable temporal representations. Monte Carlo Dropout yields a prediction interval coverage probability of 93.2%, close to the nominal 95% target. A data-scarcity analysis further shows monotonic improvement with increasing target-domain data, providing practical guidance for district energy deployment.
[AI-72] NaRA: Noise-Aware LoRA for Parameter-Efficient Fine-Tuning of Diffusion LLM s
链接: https://arxiv.org/abs/2605.29716
作者: Shuaidi Wang,Zhan Zhuang,Ruping Huang,Yu Zhang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Diffusion Large Language Models (dLLMs) have emerged as a promising non-autoregressive generative paradigm. Given the prohibitive computational cost of full fine-tuning, Parameter-Efficient Fine-Tuning (PEFT) has become the standard approach. However, existing PEFT methods (e.g., LoRA), originally tailored for autoregressive models, rely on static parameters that are agnostic to the noise level. Consequently, they ignore the intrinsic dynamics of the diffusion process, where input distributions and generation difficulty shift significantly along the denoising trajectory, rendering them suboptimal for dLLMs. To address this, we propose Noise-aware Low-Rank Adaptation (NaRA), which introduces a low-rank core matrix generated by a lightweight, globally shared hypernetwork conditioned on the noise level. This design enables the update matrices to vary continuously along the diffusion process while keeping parameter and latency overhead negligible. We provide a theoretical justification for the proposed NaRA framework and empirically demonstrate consistent improvements over noise-agnostic baselines across commonsense reasoning, mathematical reasoning, and code generation benchmarks. Our code is available at this https URL.
[AI-73] he Little Book of Generative AI Foundations: An Intuitive Mathematical Primer
链接: https://arxiv.org/abs/2605.29713
作者: Tianhua Chen
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Preprint version, 178 pages. Comments and corrections are welcome
Abstract:This book provides a compact, derivation-oriented introduction to the mathematical foundations of modern generative artificial intelligence. Rather than surveying every recent architecture or implementation detail, it develops a coherent route through the ideas connecting major families of generative models, from PCA, probabilistic PCA, variational autoencoders, and diffusion models to normalising flows, autoregressive factorisations, GANs, Wasserstein GANs, and energy-based models. The aim is to make the structure of generative modelling more accessible without removing the mathematical substance needed to understand how these models are derived and related. The book is intended as a foundation-building primer for mathematically curious researchers, practitioners, and students.
[AI-74] BitTP: The Lightweight Trajectory Prediction Model with BitLLM for Edge-Devices CVPR2026
链接: https://arxiv.org/abs/2605.29705
作者: Mincheol Kang,Hyunjin Lim,Bomin Kang,Daehee Park
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: Camera-ready version. Accepted as a findings paper at CVPR 2026. 8 pages, 4 figures
Abstract:Trajectory prediction is a fundamental task for autonomous systems, requiring complex reasoning about multi-agent interactions and intents. Large language models (LLMs) have recently been adopted for this task, as they provide strong contextual reasoning and interpretable, language-based trajectory representations. However, these LLM-based predictors are extremely memory- and compute-intensive, making them difficult to deploy on resource-constrained edge devices such as on-board computers in autonomous robots. To bridge this gap, we propose BitTP, which converts an LLM-based trajectory predictor into a lightweight bitlinear architecture. We demonstrate that weight-only quantization to 1.58-bit (BitTP-Weight) is optimal. Crucially, activations must remain in full precision, as quantizing them leads to severe degradation and instability in spatio-temporal reasoning. Empirically, BitTP-Weight not only preserves but improves prediction quality over the full-precision (BF16) LLM baseline, reducing ADE by 14.29% and FDE by 20.97% on average, while simultaneously reducing memory usage and inference latency relative to other quantization methods. These results demonstrate that carefully designed quantization acts as an effective regularizer, enabling the practical deployment of sophisticated LLM-based reasoning on edge devices. Code is available at: this https URL.
[AI-75] Beyond Trajectory Rewards: Step-level Credit Assignment for Agent ic Search via Graph Modeling
链接: https://arxiv.org/abs/2605.29697
作者: Yuchen Liu,Yingjie Feng,Lixiong Qin,Jiasi Chen,Jianing Yu,Sheng Gao,Sheng Yang,Weiran Xu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 15 pages, 8 figures
Abstract:In Agentic Search, trajectory-level outcome rewards fail to quantify the behavioral contributions of individual steps, while existing step-level reward methods typically rely on costly tree sampling. We view world knowledge as a latent world graph and each IS task as search within a latent task graph, where effective steps should make graph progress toward the answer node. Based on this prior, we propose Graph-Distance Contribution Reward (GDCR), a step-level process reward that scores newly-retrieved and newly-cited entities by their distance to the answer node in a training-time Entity-Relation (ER) graph. We further propose Step Advantage Policy Optimization (SAPO), which converts GDCR into step-level advantages and combines them with trajectory-level outcome advantages. Experiments on four challenging benchmarks validate the effectiveness of our method.
[AI-76] FHRFormer: A Self-Supervised Masked Transformer Framework for Fetal Heart Rate Time-Series Inpainting and Forecasting ALT
链接: https://arxiv.org/abs/2605.29695
作者: Kjersti Engan,Neel Kanwal,Anita Yeconia,Ladislaus Blacy,Yuda Munyaw,Estomih Mduma,Hege Ersdal
机构: 未知
类目: Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG); Probability (math.PR)
备注: Submitted to Frontiers in Digital Health. arXiv admin note: substantial text overlap with arXiv:2509.20852
Abstract:Approximately 10% of newborns require assistance to initiate breathing at birth, and around 5% need ventilation support. Fetal heart rate (FHR) monitoring plays a crucial role in assessing fetal well-being during prenatal care, enabling the detection of abnormal patterns and supporting timely obstetric interventions to mitigate fetal risks during labor. Applying artificial intelligence (AI) methods to analyze large datasets of continuous FHR monitoring episodes with diverse outcomes may offer novel insights into predicting the risk of needing breathing assistance or interventions. Recent advances in wearable FHR monitors have enabled continuous fetal monitoring without compromising maternal mobility. However, sensor displacement during maternal movement, as well as changes in fetal or maternal position, often lead to signal dropout, resulting in gaps in recorded FHR data. Such missing data limits the extraction of meaningful insights and complicates automated (AI-based) analysis. Traditional approaches to handling missing data, such as simple interpolation techniques, often fail to preserve the spectral characteristics of the signals. In this paper, we propose a masked transformer-based autoencoder approach to reconstruct missing FHR signals by capturing both local temporal and frequency components of the data. The proposed method demonstrates robustness across varying durations of missing data and can be used for signal inpainting and forecasting. The proposed approach can be applied retrospectively to research datasets to support the development of AI-based risk algorithms. In the future, the proposed method could be integrated into wearable FHR monitoring devices to achieve earlier and more robust risk detection.
[AI-77] Reliable Reasoning with Large Language Models via Preference-Based Maximum Satisfiability
链接: https://arxiv.org/abs/2605.29687
作者: Pedro Orvalho,Marta Kwiatkowska,Guillem Alenyà,Felip Manyà
机构: 未知
类目: Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO)
备注: 17 pages, 1 figure, 4 tables
Abstract:Large Language Models (LLMs) excel at understanding natural language but struggle with optimisation tasks involving multiple constraints and user-defined preferences, which commonly arise in domains such as robotics. We propose a hybrid reasoning approach in which LLMs externalise reasoning through code generation. Given a natural language problem description, an LLM generates Python code that encodes user-defined constraints and preferences as a preference-based Maximum Satisfiability (MaxSAT) problem, which is then solved by an exact MaxSAT solver. To ensure correctness, solutions returned by the model-generated code are independently verified for feasibility and optimality against a canonical MaxSAT encoding, allowing for different encodings and multiple optimal solutions. We evaluate our approach using both open-source and closed-access LLMs on three families of preference-based reasoning tasks, and compare it against direct-answer, chain-of-thought, and program-of-thought baselines using the same models. While these baselines rarely produce feasible solutions, the MaxSAT-based pipeline achieves substantially higher acceptance rates, in some cases exceeding 80%. Our results demonstrate that LLM-driven code generation combined with preference-based MaxSAT enables solver-verifiable optimisation with respect to generated encodings, and substantially improves correctness under independently verified reference semantics.
[AI-78] NICE: A Theory-Grounded Diagnostic Benchmark for Social Intelligence of LLM s
链接: https://arxiv.org/abs/2605.29685
作者: Yunjin Qi,Zhaojun Jiang,Xuan Wu,Hanxi Pan,Yixuan Wang,Yanfang Liu,Xiang Ji,Churu Yu,Chunyuan Zheng,Yingze Chen,Jie He,Liuqing Chen,Zaifeng Gao
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As large language models (LLMs) are increasingly applied in social contexts such as emotional companionship and customer service, measuring their social intelligence has become critical to the quality and safety of human-AI interaction. However, existing social intelligence benchmarks lack a unified framework that organizes social abilities into a unified structure, and therefore cannot enable fine-grained diagnosis. To build the first holistic diagnostic evaluation grounded in social theory, we first construct a social intelligence framework through a literature review and multi-stage expert validation guided by psychometric principles. The resulting framework includes 4 categories and 11 dimensions, each further specified by fine-grained capability facets. Building on this framework, we introduce NICE (Norm, Interaction, Cognition, Experience), a diagnostic benchmark of 137 items operationalized through representative Chinese contexts. Across 5 frontier LLMs and a human reference group, models score higher in aggregate accuracy yet show a consistent weakness in Communication, which the framework localizes to 3 specific capability facets: multi-turn communication, nonverbal communication, and synchrony. NICE thus reframes social intelligence evaluation toward theory-grounded diagnosis of socially consequential weaknesses in LLMs.
[AI-79] RACE: Toulmin-based Reasoning Assessment through Constructive Elements for LLM CoT Evaluation ICML2026
链接: https://arxiv.org/abs/2605.29656
作者: Yundong Kim,Heyoung Yang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 23 pages, Accepted at ICML 2026
Abstract:Evaluating open-ended outputs from large language models (LLMs) remains challenging due to the absence of ground truth. Existing metrics rely on final-answer accuracy or surface-level statistics, leaving the reasoning process itself unexamined. We introduce TRACE (Toulmin-based Reasoning Assessment through Constructive Elements), a metric that analyzes Chain-of-Thought (CoT) reasoning processes. Rather than judging outcomes, TRACE inspects how arguments are constructed by integrating Toulmin’s argumentation theory with Flavell’s metacognitive framework to assess reasoning structure. Experiments on 26.3K QA samples across 7 reasoning models show strong correlation with benchmark accuracy (r=0.74). Furthermore, TRACE is effective as a reinforcement learning reward signal, outperforming accuracy-only baselines. Together, these results indicate that logically sound reasoning leads to higher-quality answers. TRACE thus serves as a complementary metric for evaluating open-ended outputs. Code is available at this https URL.
[AI-80] PTCG-Bench: Can LLM Agents Master Pokémon Trading Card Game?
链接: https://arxiv.org/abs/2605.29653
作者: Dongdong Hua,Yifei Sun,Renhong Huang,Feng Gao,Chunping Wang,Yang Yang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Given a strategically complex board game, human players can quickly learn to devise strategies after playing a few rounds. Autonomous agents require similar capabilities in realistic interactive environments, yet existing agent benchmarks often fail to fully capture such strategic and evolving decision-making scenarios. We present PTCG-Bench, a benchmark built on the Pok’emon Trading Card Game (PTCG) that evaluates LLM agents at two complementary levels: (1) their decision-making performance within a single complex environment, and (2) their ability to self-evolving through accumulated experience. We further include a modular harness ablation to better interpret agent performance without conflating it with model capability. Our experiments show that, although LLM agents can achieve non-trivial gameplay performance, sustained and stable self-evolution remains challenging, and performance is sensitive to harness design. We hope that PTCG-Bench will facilitate future research on harness-aware and self-evolving agents in realistic interactive environments.
[AI-81] hink Fast Talk Smart: Partitioning Deterministic and Neural Computation for Structured Health Text Generation
链接: https://arxiv.org/abs/2605.29652
作者: Kai-Chen Cheng,Haejun Han,David Q. Sun
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) are increasingly being used to generate health text from structured records such as wearable time series, biomarkers, vitals, and care-management logs. For recurring health outputs, fluency is not enough: systems must remain faithful to source data, ground explanatory claims in available evidence, follow stated policies, emit machine-readable outputs, and run cheaply enough for repeated use. We ask which responsibilities in structured health generation should be deterministic computation rather than runtime LLM prompting. We introduce Think Fast, Talk Smart, a sleep-health insight pipeline in which deterministic code performs recurring analysis before one bounded LLM writer call. Across 280 user-nights and six models, achieves lower numeric error, lower instruction-compliance error, and lower end-to-end cost than structured zero-shot and few-shot one-call baselines. Layer replacement reveals contract-specific failures: LLM comparison raises numeric error, LLM ranking degrades policy selection, LLM attribution increases unsupported causal language, and an LLM-generated writer interface reintroduces errors even after upstream facts are deterministic. The results support a broader design rule: let code own recurring analysis, and let LLMs express verified facts within bounded interfaces. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2605.29652 [cs.AI] (or arXiv:2605.29652v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2605.29652 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-82] LLM -Evolved Domain-Independent Heuristics for Symbolic AI Planning
链接: https://arxiv.org/abs/2605.29649
作者: Elliot Gestrin,Jendrik Seipp
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Heuristic search is the dominant paradigm in symbolic AI planning, and the strongest heuristics are the result of decades of work by planning researchers. Recent work has shown that large language models (LLMs) can design heuristics for individual planning domains, but no LLM-generated heuristic has so far worked on arbitrary planning tasks. In this paper, we use evolutionary search to produce the first LLM-generated domain-independent heuristics that exceed the hand-engineered state of the art. We let an LLM mutate parent heuristics written in C++, store candidates in a MAP-Elites archive keyed on informedness and speed and calculate fitness scores by blending coverage with solving time. To place the evolved programs in context, we additionally benchmark a broad set of hand-engineered heuristics on their informedness-speed tradeoff, which to our knowledge has not been done before. On unseen testing domains, our best evolved heuristic solves more tasks than even the strongest baseline, with our full heuristic suite spanning the Pareto frontier of said tradeoff. We also find that seeding evolution from the trivial blind heuristic outperforms seeding from the strong FF heuristic, even when the resulting program is itself an FF variant, and that LLM reasoning effort affects how often candidates compile much more than the quality of those that do. Because the evolved programs are plain C++, they slot into existing planners as drop-in replacements and inherit the soundness and completeness guarantees of the underlying search.
[AI-83] he Sample Complexity of Multiclass and Sparse Contextual Bandits
链接: https://arxiv.org/abs/2605.29645
作者: Liad Erez,Fan Chen,Alon Cohen,Tomer Koren,Yishay Mansour,Shay Moran,Alexander Rakhlin
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:We study contextual bandits in the stochastic i.i.d.\ setting, where a learner observes contexts drawn from an unknown distribution, selects actions from a finite set A , and aims to identify an approximately optimal policy from a given class based on bandit feedback. Motivated by bandit multiclass classification with zero-one rewards, we focus on the \emph s -sparse setting in which, for every context, the reward vector has L_1 -norm at most s \ll |A| . Our main result is the design of algorithms that, with high probability, output an \epsilon -optimal policy compared to policy class \Pi using \tildeO ((s/\epsilon^2 + |A|/\epsilon)\log |\Pi|/\delta) samples. We extend this bound to general Natarajan classes and complement it with a matching lower bound (up to logarithmic factors), thereby closing a substantial gap left by prior work (Erez et al., 2024, 2025), which incurred an additional \Theta(|A|^9) dependence. We obtain these results via two complementary approaches. First, we analyze contextual bandits through the lens of contextual decision making with structured observations, designing an exploration-by-optimization algorithm whose sample complexity is governed by the \emphdecision-estimation coefficient (DEC; Foster et al., 2021, 2022). We show that, with s -sparse rewards, the induced model class admits a sharp DEC bound that scales with s and directly yields the optimal rate. Since this approach is largely information-theoretic and involves solving complex min-max optimization problems, we also develop a second, more specialized algorithmic method based on a low-variance exploration technique. This approach leads to concrete, tractable algorithms and naturally extends to contextual combinatorial semi-bandits, leading to improved sample complexity guarantees for bandit multiclass list classification.
[AI-84] VikingMem: A Memory Base Management System for Stateful LLM -based Applications VLDB26
链接: https://arxiv.org/abs/2605.29640
作者: Jiajie Fu,Junwen Chen,Mengzhao Wang,Aoxiang He,Maojia Sheng,Xiangyu Ke,Yifan Zhu,Yunjun Gao
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: Accepted by VLDB26
Abstract:Large Language Models have revolutionized interactive applications; however, their finite context windows pose a critical data management challenge for maintaining stateful, long-term interactions. Existing memory approaches often rely on simplistic extraction methods that lead to incomplete memories or use rigid, single-purpose memory extraction prompts tailored to a single use case, such as chatbots. Consequently, they lack generalizability and perform poorly across diverse downstream tasks. To bridge this gap, we introduce the Memory Base, a novel data management paradigm for managing the persistent state of long-term interactions. It is characterized by three core principles: selective extraction of high-value memories from raw information streams; inherent statefulness and evolution, where memory content is progressively summarized, corrected, and temporally weighted to prioritize recent interactions; and a generalizable abstraction paradigm designed for robust transferability across diverse applications, including education, recommendation, and agent memory. Building on this foundation, we present VikingMem, an end-to-end Memory Base Management System implemented on the VikingDB vector engine. VikingMem materializes this paradigm through interconnected event and entity abstractions. It features event-centric memory extraction to selectively handle complex information streams, while entities are dynamically updated by events to achieve stateful evolution. Using temporal compression via a topic-wise timeline and time-weighted recall, the system progressively produces high-level summary memories, prioritizes recent items, and compresses and fades older ones. Extensive evaluations on long-term memory benchmarks demonstrate that VikingMem outperformes baselines by up to 30% in memory retrieval effectiveness while maintaining the low latency essential for interactive applications.
[AI-85] Beyond Attack Success Rate: Temporal Logit Observability for LLM Safety Failures
链接: https://arxiv.org/abs/2605.29629
作者: Junyoung Park,Sunghwan Park,Seongyong Ju,Jaewoo Lee
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Attack Success Rate (ASR) evaluates each jailbreak with a single yes/no label at the end of generation, telling us whether a failure happened but not how it unfolded. Two attacks that produce equally harmful outputs may have followed completely different paths, and ASR cannot tell them apart. We make those hidden paths observable from logits alone. Temporal Logit Observability (TLO) is a training-free diagnostic that watches a compliance-refusal margin during decoding and places each model-attack condition on a calibrated 2D plane. By design, this plane is most informative exactly where ASR is least informative: among attacks that succeed for genuinely different reasons. Across four aligned LLMs and three jailbreak paradigms, attacks with nearly identical ASR land at clearly different points on the plane: the same model can fail through different temporal patterns. The geometry matches refusal-direction probes from hidden states on most conditions, with one model showing the limit of our fixed-lexicon approach. A simple early-stop rule derived from TLO cuts successful jailbreaks by more than half, without false alarms on plain benign queries. Safety evaluation should report when and how a failure unfolds, not only whether it occurred. TLO makes the first two observable from logits alone.
[AI-86] Improving Collaborative Storytelling with a Multi-Agent Framework Based on Large Language Models
链接: https://arxiv.org/abs/2605.29625
作者: Arturo Valdivia,Paolo Burelli
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:The topic of Co-creation, i.e., AI agents interacting with humans to generate outputs (e.g., art), has gained significant attention recently. However, most studies focus on adult-human interactions in a digital setting. This paper explores a novel ludic co-creation scenario involving children and Large Language Models (LLMs) interacting through a physical board game to create written stories. Our goal is to develop a multi-agent framework capable of producing high-quality narratives suitable for young players. At the core of our approach is an iterative Writer-Editor process in which one LLM generates stories while another evaluates them and provides feedback for refinement. Through a simulation study involving multiple LLMs, we show that this iterative interaction consistently improves the perceived quality of generated stories across successive loops. The results indicate that a small number of refinement steps may be sufficient to achieve high-quality outputs in interactive storytelling systems.
[AI-87] Mind-Omni: A Unified Multi-Task Framework for Brain-Vision-Language Modeling via Discrete Diffusion
链接: https://arxiv.org/abs/2605.29591
作者: Yizhuo Lu,Changde Du,Qingyu Shi,Hang Chen,Jie Peng,Liuyun Jiang,Shuangchen Zhao,Huiguang He
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Modeling the interplay between external stimuli and internal neural representations is a pivotal research area for Brain-Computer Interfaces (BCIs). A major limitation of prior work is the prevailing paradigm of specialized, single-task models, which curtails versatility and neglects inter-task synergies. To address this, we propose Mind-Omni, the first versatile framework that unifies seven distinct encoding and decoding tasks through a discrete diffusion paradigm. At its core is a novel Brain Tokenizer that transforms heterogeneous, continuous brain signals into standardized, discrete tokens. This enables direct, token-level interactions for mutual understanding and generation between any two or more modalities within a shared semantic space. To unlock advanced reasoning capabilities, we further curate a specialized Brain Question Answering (BQA) instruction-tuning dataset. Our model not only establishes a new state-of-the-art among multi-task unified frameworks but also provides strong evidence for multi-task synergy. By demonstrating performance competitive with, and at times superior to, larger specialized models, our work offers a powerful new paradigm for neural modeling and paves the way for foundation models of neural activity. The code is publicly available at this https URL.
[AI-88] FinVerBench: Benchmark Validity and Calibration in Large Language Model Financial Statement Verification
链接: https://arxiv.org/abs/2605.29586
作者: Silu Panda
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 37 pages, 9 figures
Abstract:We introduce FinVerBench, a benchmark and validity study for financial statement verification: determining whether a set of corporate financial statements is numerically consistent from the information shown to the model. FinVerBench is built from SEC 10-K XBRL filings for 43 SP 500 companies and defines a four-category error taxonomy covering arithmetic, cross-statement linkage, year-over-year, and magnitude perturbations. We attempt fifteen contemporary LLM evaluations and report fourteen complete runs; a Gemini 2.5 Pro run is excluded from the main comparison because 40/108 gateway calls failed. All binary metrics exclude underdetermined positive instances whose perturbed line item is not rendered, leaving a 105-instance observable diagnostic subset (43 clean, 62 error-injected). Under the original guided-checklist prompt on the unrounded diagnostic subset, nine of fourteen complete LLM runs produce 95-100% false positives on clean statements, while one run achieves 0% observed false positives. Benchmark rendering choices materially affect measured recall: on a realistic rounded variant of the same observable subset, the calibrated model’s recall is 79.0% with 0% observed FPR, compared with 100.0% recall on the unrounded diagnostic variant. These results support a construct-validity conclusion rather than a final leaderboard: financial statement verification is not merely arithmetic detection, but calibrated judgment under incomplete observability, prompt-induced assumptions, and realistic numerical rendering. FinVerBench and all code are publicly available.
[AI-89] GPS-Enhanced Tourist Mobility Modeling with Seasonal Spatial Priors and LLM -Based Activity Chain Generation
链接: https://arxiv.org/abs/2605.29578
作者: Yifan Liu,Yanling Sang,Xishun Liao,Morgan Sun,Bo Yang,Zhiyuan Zhang,Chris Stanford,Haoxuan Ma,Jiaqi Ma
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Tourist mobility poses a distinct challenge for urban transportation planning. Unlike resident commuting, tourist travel is largely non-routine, attraction driven, and highly sensitive to trip purpose, travel season, and trip member composition. Existing approaches either measure aggregate tourist spatial patterns without generating individual schedules, or synthesize mobility without tourist specific structure such as trip duration conditioning, month varying attraction demand, and household co-travel rules. To address these challenges, we propose a four stage simulation framework combining month conditioned spatial priors derived from GPS and survey data, trip extent prediction from tourist demographics, distance feasible ward sequence assignment, and LLM-based activity chain generation under household and spatial constraints. GPS data are used only in privacy preserving aggregated form as month conditioned spatial priors, with no individual traces retained or exposed. Experiments on tourism in Tokyo demonstrate that the GPS based tourist cohort extraction recovers spatial visitation signatures consistent with survey references, and our framework produces demographically aligned synthetic schedules whose ward-level visitation shares align closely with both survey distributions and staypoint derived monthly visitation patterns. The results demonstrate the framework’s effectiveness as a geographically grounded, demographically aware approach to tourist mobility modeling.
[AI-90] DeepTool: Scaling Interleaved Deliberation in Tool-Integrated Reasoning via Process-Supervised Reinforcement Learning
链接: https://arxiv.org/abs/2605.29568
作者: Yang He,Xiao Ding,Bibo Cai,Yufei Zhang,Kai Xiong,Zhouhao Sun,Bing Qin,Ting Liu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Tool-Integrated Reasoning (TIR) extends LLM capabilities by leveraging external environments. However, existing methods lack the deliberation during sequential tool invocation required for strategic planning and self-correction. While RL mitigates this, conventional approaches for Tool-Integrated Reasoning are hindered by sparse outcome-based rewards, failing to supervise intermediate reasoning steps and tool invocations. To address this, we propose DeepTool, a novel framework that scales deliberate thinking within the interleaved process of thinking, action, and observation at each turn. In DeepTool, we first introduce a synthesis pipeline that evolves extended thinking into interleaved trajectories, integrating adversarial perturbations to ensure robustness and self-correction. Secondly, we devise Process-Supervised Reinforcement Learning based on GRPO, which utilizes an Action-Centric Process Reward to reinforce intermediate interleaved thinking and enforce precise tool invocation at every turn. Extensive experiments demonstrate that DeepTool achieves superior performance, boosting Qwen2.5-7B significantly across six benchmarks (e.g., AIME24: 3.2% - 40.4% and HMMT25: 0.0% - 28.6%). Furthermore, the token cost-effectiveness analysis confirms the utility of interleaved thinking, demonstrating DeepTool’s optimal balance between performance and token efficiency.
[AI-91] ParaTool: Shifting Tool Representations from Context to Parameters
链接: https://arxiv.org/abs/2605.29561
作者: Zekai Yu,Qi Meng,Qizhi Chu,Yu Hao,Chuan Shi,Cheng Yang
机构: 未知
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Tool calling extends large language models (LLMs) by enabling grounded interaction with external executable interfaces, thereby supporting environment-coupled problem solving. However, mainstream in-context learning (ICL) approaches typically incorporate detailed tool documentation and usage examples directly into the context. This results in substantial inference overhead and heightened risks of hallucination as the context length grows. Conversely, while tuning-based methods improve general tool-calling capabilities, they often fail to effectively internalize the specific details of previously seen tools, thereby retaining a dependency on in-context documentation. To address these limitations, we propose ParaTool, a framework that projects each tool into a dedicated, loadable set of parameters. By equipping a dynamic integration of these parameterized tools, the LLM can perform tool calling without relying on in-context documents or examples. Specifically, our approach consists of three stages: (1) parametric tool pre-training encapsulates the knowledge of different tools into independent parameter modules; (2) soft tool selection employs a gating network to dynamically weigh and aggregate relevant tool parameters; and (3) parametric tool fine-tuning jointly updates tool parameters to align the training and inference processes. Experiments on Stable ToolBench and BFCL demonstrate that ParaTool significantly outperforms strong ICL-based baselines, achieving superior performance while reducing computational complexity.
[AI-92] Battery-Sim-Agent : Leverag ing LLM -Agent for Inverse Battery Parameter Estimation
链接: https://arxiv.org/abs/2605.29560
作者: Jiawei Chen,Xiaofan Gui,Shikai Fang,Shengyu Tao,Shun Zheng,Weiqing Liu,Jiang Bian
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Parameterizing high-fidelity “digital twins” of batteries is a critical yet challenging inverse problem that hinders the pace of battery innovation. Prevailing methods formulate this as a black-box optimization (BBO) task, employing algorithms that are sample-inefficient and blind to the underlying physics. In this work, we introduce a new paradigm that reframes the inverse problem as a reasoning task, and present Battery-Sim-Agent, the first framework to deploy a Large Language Model (LLM) agent in a closed loop with a high-fidelity battery simulator. The agent mimics a human scientist’s workflow: it interprets rich, multi-modal feedback from the simulator, forms physically-grounded hypotheses to explain discrepancies, and proposes structured parameter updates. On a systematically constructed benchmark suite spanning diverse battery chemistries, operating conditions, and difficulty levels, our agent significantly outperforms strong BBO baselines like Bayesian optimization in identifying accurate parameters. We further demonstrate the framework’s capability in complex long-horizon degradation fitting tasks and validate its practical applicability on real-world battery datasets. Our results highlight the promise of LLM-agents as reasoning-based optimizers for scientific discovery and battery parameter estimation.
[AI-93] Opt-Verifier: Unleashing the Power of LLM s for Optimization Modeling via Dual-Side Verification
链接: https://arxiv.org/abs/2605.29556
作者: Haoyang Liu,Jie Wang,Boxuan Niu,Xiongwei Han,Yian Xu,Mingxuan Ye,Zijie Geng,Fangzhou Zhu,Tao Zhong,Mingxuan Yuan,Jianye Hao
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Building mathematical optimization models is critical in operations research (OR), while it requires substantial human expertise. Recent advancements have utilized large language models (LLMs) to automate this modeling process. However, existing works often struggle to verify the correctness of the generated optimization models, without checking the rationality of the constraints and variables or the validity of solutions to the generated models. This hampers the subsequent verification and correction steps, and thus it severely hurts the modeling accuracy. To address this challenge, we propose a novel LLM-based framework with Dual-side Verification (Opt-Verifier) from both structure and solution perspectives, thereby improving the modeling accuracy. The structure-side verification ensures that the modeling structure of the generated optimization models aligns with the original problem description, accurately capturing the problem’s constraints and requirements. Meanwhile, the solution-side verification interprets and evaluates the solutions’ validity, confirming that the optimization models are logically and mathematically sound. Experiments on popular benchmarks demonstrate that our approach achieves over 20% improvement in accuracy.
[AI-94] Singularity-aware Optimization via Randomized Geometric Probing: Towards Stable Non-smooth Optimization ICML
链接: https://arxiv.org/abs/2605.29547
作者: Ruoran Xu,Borong She,Xiaobo Jin,Qiufeng Wang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Optimization and Control (math.OC)
备注: International Conference on Machine Learning (ICML), 2026
Abstract:Deep learning optimization relies heavily on the assumption of smooth loss landscapes, a condition systematically violated by modern architectures due to non-smooth components such as ReLU activations and quantization operators. In such non-smooth regimes, adaptive optimizers such as Adam suffer from gradient chattering, violent oscillations caused by conflicting signals within the Clarke subdifferential, leading to poor convergence and suboptimal generalization. To address this, we introduce Singularity-aware Adam (S-Adam), a novel optimizer that stabilizes training by dynamically modulating step sizes based on local geometric instability. Our key contribution is the Local Geometric Instability (LGI) metric, a computationally efficient estimator of the Clarke subdifferential diameter derived from the variance of randomized directional derivatives. S-Adam incorporates an adaptive damping mechanism exp(- \lambda \rho ) that decelerates updates in high-instability regions while preserving fast convergence in smooth basins. We provide a rigorous convergence analysis using differential inclusions, proving that S-Adam converges almost surely to ( \delta , \epsilon )-Clarke stationary points at the optimal O(1/ \sqrt(T) ) rate. Empirical evaluations on Quantization-Aware Training (QAT) and high-noise small-batch learning demonstrate that S-Adam consistently outperforms AdamW and Prox-SGD, achieving accuracy gains of up to 6 percent on CIFAR-100 and 3 percent on TinyImageNet while effectively mitigating gradient oscillations.
[AI-95] UI-KOBE: Knowledge-Oriented Behavior Exploration for Lightweight Graph-Guided GUI Agents
链接: https://arxiv.org/abs/2605.29534
作者: Yuxiang Chai,Han Xiao,Xinyu Fu,Jinpeng Chen,Rui Liu,Hongsheng Li
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances in mobile GUI agents have shown strong potential for automating mobile tasks, but most effective systems still depend on large vision-language models for screenshot understanding and long-horizon planning. Small GUI agents that can be deployed directly on mobile devices are more attractive for practical use, offering lower inference cost and better protection of sensitive on-device information. However, due to limited model capacity, such lightweight agents remain unreliable when planning and executing GUI tasks end-to-end from screenshots alone. We propose Knowledge-Oriented Behavior Exploration (\textbfUI-KOBE), a framework that improves lightweight mobile GUI agents with reusable app-specific graph knowledge. UI-KOBE first autonomously explores a mobile application and constructs an app knowledge graph, where nodes represent distinct UI states and edges represent executable transitions. At runtime, a lightweight GUI agent uses the graph as external guidance: given a user task and the current screenshot, it identifies the current graph node and selects among self-loop actions, neighboring transitions, task completion, or fallback free actions associated with that node. By supporting runtime decisions with app-specific graph guidance, UI-KOBE reduces the burden of end-to-end GUI planning and helps lightweight models perform mobile GUI tasks more effectively, offering a practical step toward efficient, interpretable, and privacy-conscious on-device GUI agents.
[AI-96] GUITestScape: Towards Open-set Evaluation on Exploratory GUI Testing
链接: https://arxiv.org/abs/2605.29532
作者: Xiaoyi Chen,Yifei Gao,Yang Xu,Xingxing Song,Yi Zhang,Jitao Sang
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:Exploratory GUI testing is a particularly demanding setting for MLLM agents: without predefined test scripts, an agent must autonomously navigate an application and discover defects through its own interaction. However, current evaluation falls short on two fronts. First, existing benchmarks focus almost exclusively on interaction defects, leaving display defects outside the evaluation frame. Second, evaluation protocols are bound to predefined defect annotations, collapsing the testing process into a single end-state judgment that conflates qualitatively distinct failure modes. To address these challenges, we present GUITestScape, an interactive benchmark covering 61 real-world Android applications and 508 preset defects spanning interaction and display types, and introduce GUIJudge, an open-set evaluator that decomposes an agent’s testing trajectory into independently diagnosable capabilities. Experimental results demonstrate that GUIJudge achieves reliable process-aware evaluation beyond predefined annotations, substantially outperforming all baselines. Benchmarking on GUITestScape further reveals that detection remains the critical bottleneck for existing models across both defect types, and that integrating GUIJudge’s verifiers into existing agents significantly boosts their detection performance without retraining.
[AI-97] mporal Motif-aware Graph Test-time Adaptation for OOD Blockchain Anomaly Detection IJCAI ECAI2026
链接: https://arxiv.org/abs/2605.29526
作者: Runang He,Tongya Zheng,Huiling Peng,Yuanyu Wan,Bingde Hu,Jiawei Chen,Canghong Jin,Mingli Song,Can Wang
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted to IJCAI-ECAI 2026, Special Track on AI for Social Good
Abstract:Ever-evolving transaction patterns have significantly hindered anomaly detection on emerging cryptocurrency blockchains due to the vast number of addresses and diverse anomalous behaviors. Recently, advanced Graph Anomaly Detection (GAD) approaches applied to blockchains have faced two critical challenges: \textitadversarial pattern evolution by malicious actors and \textitthe out-of-distribution (OOD) problem caused by varied transaction semantics on blockchains. To address these challenges, we propose a novel framework termed \textbfTEmporal \textbfMotif-aware \textbfGraph \textbfTest-\textbfTime \textbfAdaptation (\textbfTEMG-TTA). First, we comprehensively capture the 3-node temporal motif distribution of each active address using an efficient computational mechanism, enabling downstream temporal motif-aware graph learning. Second, we design a simple yet effective test-time adaptation strategy to facilitate the sharing of common patterns between training and testing graphs. Extensive experiments on 5 real-world datasets demonstrate that our proposed \textbfTEMG-TTA outperforms \textitstate-of-the-art GAD approaches by an average of 54.88%. A further case study on interpretable motif patterns reveals that \textbfTEMG-TTA explicitly characterizes the complex transaction patterns of anomalous addresses, thereby verifying the effectiveness of our technical designs. Our code will be made publicly available this https URL.
[AI-98] KBF: Knowledge Boundary as Fingerprint for Language Model and Black-Box API Auditing
链接: https://arxiv.org/abs/2605.29524
作者: Yijia Fang,Yiqing Feng,Bingyu Li,Mingxun Zhou
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 20 pages, 13 figures
Abstract:Relay and reseller APIs increasingly intermediate access to large language models (LLMs), but users have no direct way to verify that a claimed endpoint is actually serving the advertised model. We introduce KBF, a low-cost black-box auditing protocol that fingerprints model APIs using stable numerical recall near the knowledge boundary. Across 16 production LLM endpoints, KBF flags all 155 economically relevant substitutions without rejecting any same-model controls, remains stable under deployment variation, detects high-separation mixed-routing attacks when only 5-10% of traffic is substituted, and finds that 7 of 27 platform model cells in a six-platform shadow API audit are statistically inconsistent with their reference endpoints, with inconsistencies concentrated on premium Claude endpoints.
[AI-99] DeepSurvey: Enhancing Analytical Depth and Citation Reliability in Automated Survey Generation
链接: https://arxiv.org/abs/2605.29522
作者: Ziyue Yang,Da Ma,Hanqi Li,Zijian Wang,Tiancheng Huang,Zijian Hu,Chenrun Wang,Yunzhe Zhang,Xiaobao Wu,Kai Yu,Lu Chen
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As scientific literature grows rapidly, automated survey generation has become a key capability for AI scientists and human researchers. However, existing systems suffer from limited analytical depth due to reliance on abstracts and isolated paper processing, and unreliable citations from imprecise retrieval and post-hoc grounding, producing superficial surveys and may mislead researchers. We present DeepSurvey, an agentic system that addresses both. To enhance depth, DeepSurvey extracts structured keynotes from full-text papers, models cross-paper relationships through clustering and comparative analysis, and integrates code-repository analysis to recover implementation-level details. To fortify reliability, it combines citation-graph expansion with hybrid filtering for topic-focussed retrieval, enforces evidence-constrained citation assignment, and deploys multi-granularity agentic refinement to validate citation-claim alignment. Experiments show that DeepSurvey achieves the highest content score (8.644/10) and citation quality (12.3% and 9.3% recall and precision gains over the strongest baseline), generalizes more robustly across domains (0.14 vs 0.22 to 0.69 CS-to-non-CS drop), and is preferred over human-written surveys by domain experts (83.3% overall quality, 100% content depth).
[AI-100] Network Optimization Aspects of Autonomous Vehicles: Challenges and Future Directions
链接: https://arxiv.org/abs/2605.29518
作者: Rudolf Krecht,Tamas Budai,Erno Horvath,Akos Kovacs,Nobert Marko,Miklos Unger
机构: 未知
类目: Networking and Internet Architecture (cs.NI); Artificial Intelligence (cs.AI)
备注:
Abstract:Global megatrends, such as urbanization, population growth, and emerging network solutions are accelerating the development of the Connected and Autonomous Vehicles (CAVs) industry. There are many truths, some misconceptions, and even some excitement about CAVs in the public’s opinion. The main objective of the current article is to provide a comprehensive review, eliminate misconceptions, and outline the future of the network optimization aspects of autonomous vehicles by presenting various multidisciplinary methods, such as cooperative perception. Given our extensive experience with CAVs, we are aiming to share some of the insights and knowledge we have gained, along with relevant use-cases and experiment results.
[AI-101] MINDGAMES: A Live Arena for Evaluating Social and Strategic Reasoning in Multi-Agent LLM s
链接: https://arxiv.org/abs/2605.29512
作者: Kevin Wang,Anna Thöni,Benjamin Kempinski,Bobby Cheng,Jianzhu Yao,Benjamin Finch,Leon Guertler,Viraj Nadkarni,Yihan Jiang,Aliaksei Korshuk,Alexander Buyantuev,Ilya Makarov,Siyuan Wu,Yu-Chi Cheng,Yan-Ru Ju,Ti-Rong Wu,I-Hsuan Chu,Yu-Yu Yang,I-Chen Wu,Yitian Huang,Qinlu Cao,Yiheng Sun,Yuhong Dai,Hongkun Yao,Jingxuan Fu,Jiwei Zhang,Hao Liao,Mossimo Ebeling,Govind Arun,Sadhvik Bathini,Mihir S Arya,Avinash Anish,Aditya Ranjan,Kirtana Sunil Phatnani,Paval KS,Vrushali Mehta,Aravind S,Nikhil Arora,Tanya Upadhyay,Amol Bandagale,Yuan Lu,ChunEn Hsiao,YuTing Lin,Arvin Chung,Jerry John Thomas,Mathieu Laurière,Leshem Choshen,Yoram Bachrach,Pramod Viswanath,Maria Polukarov,Cheston Tan,Tal Kachman,Atlas Wang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) are increasingly deployed as interactive agents, yet their capacity for social and strategic reasoning over extended interaction remains poorly understood. Existing evaluations rely on static vignettes or single-game benchmarks that cannot capture the sustained, multi-faceted reasoning that real-world multi-agent settings demand. We introduce Mindgames, a multi-game arena and evaluation platform for LLM agents that operationalizes complementary reasoning demands relevant to ``theory of mind’': belief attribution under hidden information, opponent modeling through repeated strategic interaction, cooperative inference under knowledge asymmetries, and sustained deception in social deduction. Built on TextArena, Mindgames provides a unified interaction interface, TrueSkill-based rating, and full trajectory logging across four game environments. We instantiate Mindgames through a 2025 competition cycle hosted at a major AI conference, which assessed 944 submitted agents from 76 teams across four games: Colonel Blotto, Iterated Prisoner’s Dilemma, Codenames, and Secret Mafia. Our analysis surfaces both agent-level and evaluation-level limitations: brittle rule adherence remains a major bottleneck, top-performing systems repeatedly rely on explicit structural scaffolding, and leaderboard validity differs sharply across environments. In particular, failure-heavy environments can reward robustness to opponent errors as much as strategic ability, with Secret Mafia exhibiting a pronounced error-survival confound in this cycle. We release a dataset of 29,571 multi-agent games with turn-level observations, actions, and rewards, together with MG-Ref, a deterministic offline tournament protocol that scores new agents against a frozen reference pool of top-ranked, low-error Stage~II submissions under the same error-attribution lens used in this analysis.
[AI-102] Quotient DAGs for Off-Policy Evaluation:Forward-Flow Importance Sampling and Exact Slate Propensities
链接: https://arxiv.org/abs/2605.29500
作者: Ziwen Xie,Shaowen Xiang,Hongyu He,Dianbo Liu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 31 pages, 3 figures, 7 tables
Abstract:Off-policy evaluation estimates how a target policy would perform using data collected by a different behavior policy, which is crucial when online testing is costly or risky, such as in recommendation or healthcare. Standard importance sampling reweights each logged trajectory, but it can treat details of the generation process as meaningful even when the evaluation target ignores them: for example, an autoregressive slate recommender may generate an ordered sequence of items while the reward and downstream estimator depend only on the unordered slate. This creates nuisance variance and a computational gap, since exact unordered slate propensities require summing over all generation orders. We introduce a quotient-DAG view that merges histories equivalent for evaluation and assigns weights using target-to-behavior forward-flow ratios on the merged graph. For slate recommendation under a set-sufficient next-item interface, this yields Forward-DP, a subset-DAG dynamic program that computes exact unordered propensities without factorial enumeration. The resulting propensity primitive enables practical propensity-based evaluation and model selection for context-dependent autoregressive slate loggers.
[AI-103] he New Pro Se: Generative AI and the Surge in Federal Civil Self-Representation
链接: https://arxiv.org/abs/2605.29493
作者: Or Cohen-Sasson
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: 15 pages, 7 figures
Abstract:Since public access to generative AI tools became widespread, federal civil litigation has seen a marked increase in pro se (self-represented) plaintiffs. This paper analyzes that shift using ~2.8 million filings, asking whether the post-GenAI period is associated not only with more pro se filings, but also with detectable changes in complaint text, litigation outcomes, and the composition of pro se litigants. Using civil filing data from FY2008-2025, we find that the federal civil pro se plaintiff rate rose from 11.33% pre-GenAI to 16.94% post-GenAI, a 5.61 percentage-point increase that persists after trend and covariate-adjusted robustness checks. We then focus on Civil Rights and Other Statutory cases, where the increase is especially pronounced, and link case metadata to pro se complaints. Drawing on stylometric AI detection indicators, we develop an interpretable measure of AI-consistent drafting. Against a threshold calibrated to the pre-GenAI baseline, the net AI-flagged share is 13.9% of post-GenAI non-form complaints. Analysis of the AI-flagged complaints shows that they are more citation-dense, disproportionately associated with first-time rather than repeat filers, and geographically unevenly distributed. This composition pattern suggests that AI-consistent drafting is not merely a repeat-filer phenomenon; it also includes a modest, suggestive increase in name-inferred female plaintiffs. We find no evidence of improved win rates; in fact, AI-flagged complaints are more likely to be dismissed and to terminate at earlier procedural phases. These findings raise new questions about access to justice and court screening burdens, and sharpen the distinction between legal formality and legal efficacy. Comments: 15 pages, 7 figures Subjects: Computers and Society (cs.CY); Artificial Intelligence (cs.AI) Cite as: arXiv:2605.29493 [cs.CY] (or arXiv:2605.29493v1 [cs.CY] for this version) https://doi.org/10.48550/arXiv.2605.29493 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-104] he Curse of Helpfulness: Inverse Scaling Law in Robustness to Distractor Instructions via DistractionIF
链接: https://arxiv.org/abs/2605.29491
作者: Zeli Su,Zhankai Xu,Tianlei Chen,Longfei Zheng,Xiaolu Zhang,Jun Zhou,Wentao Zhang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models (LLMs) are increasingly deployed in agentic and retrieval-augmented generation (RAG) systems, where they must execute user-specified tasks over externally provided reference text. In practice, such context is often unstructured and contaminated with benign but instruction-like semantic noise, such as editorial comments and system traces, which should be treated strictly as data. We introduce DistractionIF, a benchmark designed to evaluate robustness against such distractor instructions in reference text. Across a broad range of models, we observe a consistent inverse scaling phenomenon: larger models are often less robust, with performance dropping by up to 30 points as scale increases. Mechanistically, our perplexity analysis reveals that scaling erodes the probabilistic boundary between robust and distracted behaviors, making models increasingly prone to over-interpreting noise as instructions. To address this, we demonstrate that reinforcement learning, specifically Group Relative Policy Optimization (GRPO), can restore this boundary, improving robustness by up to 15.5% without compromising general instruction-following capability. Our findings highlight a critical instruction-following robustness gap in reference-grounded tasks and establish reinforcement learning as a promising path for enforcing strict data-instruction separation at scale.
[AI-105] VitalAgent : A Tool-Augmented Agent for Reactive and Proactive Physiological Monitoring over Wearable Health Data
链接: https://arxiv.org/abs/2605.29483
作者: Di Zhu,Yu Yvonne Wu,Hong Jia,Aaqib Saeed,Vassilis Kostakos,Ting Dang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Wearable devices enable continuous monitoring of physiological signals such as ECG and PPG, but existing mHealth systems are largely limited to task-specific prediction pipelines or reactive question answering over static summaries. They lack the ability to support temporal reasoning, persistent physiological context, and proactive monitoring over long-term signal streams. We propose VitalAgent, a tool-augmented agentic framework for ECG/PPG-based mHealth that supports both reactive question answering and proactive monitoring. VitalAgent is built on a longitudinal physiological memory and a tool-augmented reasoning interface that enables dynamic computation over raw signals. We further introduce VitalBench, a longitudinal physiological monitoring benchmark dataset comprising 1,862 QA pairs for reactive question answering and 90.2 hours of continuous ECG/PPG recordings for proactive monitoring, covering cardiac, physical activity, and stress-related tasks. Experiments demonstrate that VitalAgent achieves over 30% improvement over prompt-based and ReAct baselines in reactive evaluation and supports proactive alert monitoring over long-term physiological signals, highlighting the importance of dynamic tool use and long-term physiological monitoring.
[AI-106] Evolutionary Rule Extraction from Corporate Default Prediction Models
链接: https://arxiv.org/abs/2605.29478
作者: Desirè Fabbretti,Matteo Pasquino,Elia Pacioni,Caterina Lucarelli,Davide Calvaresi
机构: 未知
类目: Neural and Evolutionary Computing (cs.NE); Artificial Intelligence (cs.AI)
备注:
Abstract:Small and medium-sized enterprises (SMEs) represent the majority of firms in most economies and often face financial constraints and higher vulnerability to financial distress. Predicting SME default is therefore crucial for financial institutions, policymakers, and researchers. Recent advances in machine learning (ML) have improved predictive performance in credit risk modeling. Yet, the limited interpretability of complex models raises concerns regarding transparency and regulatory compliance. This study investigates SME’s default predictors and applies explainable artificial intelligence (XAI) techniques to them. Using a panel of 50,718 Italian SME over the period 2015-2024, we compare traditional econometric approaches with several ML classifiers. The empirical results show that ML models significantly outperform the traditional logistic regression benchmark in terms of Balanced Accuracy and PR-AUC. To address the interpretability challenge, we introduce DEXiRE-EVO, a novel evolutionary rule extraction framework that combines multi-objective optimization with the Contextual Importance and Utility (CIU) explainability method. The extracted rules reveal economically meaningful patterns associated with SME financial distress, highlighting the roles of weak internal liquidity generation, internal capital erosion, high leverage, and operational inefficiency. Additionally, contextual macroeconomic conditions and the persistence of financial instability contribute to identifying high-risk firms. In general, the results show that combining ML with evolutionary rule extraction can improve both predictive performance and interpretability in credit risk modeling, thus supporting more transparent, data-driven decision-making in financial environments.
[AI-107] SciIntBench: Measuring LLM Compliance with Research Integrity Norms Under Adversarial Framing
链接: https://arxiv.org/abs/2605.29468
作者: Almene De Meran Meguimtsop,Maria Leonor Pacheco,Daniel E. Acuna
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) are increasingly used to support scientific work, but it is unclear whether they uphold responsible conduct of research (RCR) norms or help undermine them. We introduce SciIntBench, an adversarial benchmark of 810 prompts across ten RCR categories and three scientific domains. Each scenario appears as an Overt Adversarial, Covert Adversarial, and Benign version, allowing us to jointly measure framing-sensitive refusal of misconduct and helpfulness on legitimate requests. We evaluate 16 commercial and open-weight LLMs from six providers (2024–2026), producing 12,960 responses. We find that scientific integrity alignment is strongly framing-sensitive: models refuse explicit misconduct far more reliably than covert violations, especially failing when misconduct is presented as a pressure-driven shortcut. Refusals vary by RCR category, with weaker boundaries around transparency, plagiarism, and fabrication.
[AI-108] Composing Non-Conjugate Factor Graphs with Closed-Form Variational Inference
链接: https://arxiv.org/abs/2605.29467
作者: Mykola Lukashchuk,Kyrylo Yemets,Wouter M. Kouw,Dmitry Bagaev,İsmail Şenöz,Jeff Beck,Bert de Vries
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Stacking probabilistic building blocks into deeper architectures typically breaks closed-form inference. We show that closed-form inference can be preserved. We identify five factor-graph primitives: a bilinear factor, an exponential link, a Gamma prior, a Gaussian likelihood, and an equality node, and prove that any model composed from them admits closed-form variational message passing. The construction works because each primitive preserves a small set of message families: under mean-field factorization, messages on Gaussian variables remain Gaussian and messages on precision variables remain Gamma, while the only non-conjugate interface, the exponential link, remains tractable through the Gaussian moment-generating function and the sufficient statistics of the Gamma family. We demonstrate composition at increasing depth, from static ensembles through input-dependent gating to split-branch routing, and show that stacking routing layers encodes arbitrary decision trees, establishing universal function approximation with closed-form inference. Applied to ensemble time-series forecasting, the framework yields a Bayesian mixture of experts in which gating functions are inferred rather than learned, providing calibrated uncertainty over expert selection across five benchmark datasets.
[AI-109] Honest Lying: Understanding Memory Confabulation in Reflexive Agents ICML2026
链接: https://arxiv.org/abs/2605.29463
作者: Prakhar Dixit,Sadia Kamal,Tim Oates
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted to ICML 2026 Workshop “Failure Modes in Agentic AI”
Abstract:Reflexion-style agents rely on self-generated reflections as memory, implicitly assuming that agents can accurately diagnose their own this http URL show that this assumption can fail systematically: across ALFWorld and HumanEval, agents store confident but incorrect interpretations of the task and continue acting on them across trials,even though the environment resets to the correct task each time. We call this failure mode memory confabulation and introduce the Reflection Repetition Rate (RRR), a log-based metric that detects repeated reliance on incorrect reflective this http URL RRR, we identify 16 frozen environments in ALFWorld, where 0 of 121 reflections mention the correct target object, and 4 analogous cases in HumanEval. Our mitigation replaces open-ended self-diagnosis with programmatic extraction of trajectory-level failure signals, increasing correct object mention from 0% to 86%, reducing RRR from 0.64 to 0.10, and solving 3 of 16 frozen ALFWorld environments, suggesting that reflective memory can reinforce false beliefs rather than correct them.
[AI-110] Forget Less Generalize More: Unifying Temporal and Structural Adaptation for Dynamic Graphs
链接: https://arxiv.org/abs/2605.29453
作者: Qian Chang,Ciprian Doru Giurcaneanu,Runsong Jia,Xia Li,Guoping Hu,Xiufeng Cheng,Jinqing Yang,Mengjia Wu,Yi Zhang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Representation learning on dynamic graphs requires capturing complex dependencies that evolve across both time and structure. Existing approaches typically adopt fixed temporal decay schemes or predetermined structural propagation depths, limiting their ability to generalize across graphs with diverse interaction frequencies and topological characteristics. We propose Dual-Scale Retentive Dynamics (DSRD), a unified framework that maintains a retentive representation state encoding both temporal memory and structural context. DSRD introduces two key components: (i) a retentive state with dual-scale adaptation that jointly models temporal dynamics and structural propagation within a single recurrent formulation, and (ii) adaptive decay kernels with learnable time-sensitivity parameters that automatically balance short-term responsiveness and long-term retention based on the underlying interaction patterns. We provide theoretical analysis establishing the equivalence between event-wise parallel aggregation and efficient recurrent state updates, as well as stability and boundedness guarantees for the learned dynamics. Extensive experiments on 14 real-world benchmarks demonstrate that DSRD consistently achieves state-of-the-art performance on both link prediction and node classification tasks, with strong generalization across transductive and inductive settings.
[AI-111] CrystalXRD-Bench: Benchmarking Vision-Language Models for XRD Peak Indexing Across Diverse Crystalline Materials
链接: https://arxiv.org/abs/2605.29446
作者: Chengliang Xu,Xiaogang Li,Peiyao Xiao,Beng Wang,Hu Wei,Bing Zhao
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 18 pages, 10 figures
Abstract:Miller-index identification from powder XRD patterns requires capabilities untested by existing multimodal benchmarks: the model must read a narrow peak location from a rendered scientific curve and then connect that observation to multi-step crystallographic reasoning. We introduce CrystalXRD-Bench, a 250-sample benchmark built from 10 public crystallographic databases for a single task: recover the full set of HKLs contributing to the highest-intensity peak in an XRD pattern. Each sample pairs the rendered XRD image with the source CIF text and chemical formula, so visual extraction errors and reasoning errors can be examined side by side. We evaluate seven vision-language models. The best Jaccard score is 0.5888 (GPT-5.4) with an exact-match rate of 37.6%, yet six of seven models remain below Jaccard 0.50; the task is far from solved. Error patterns vary systematically: double-peak cases are especially brittle, recall-heavy models gain coverage by over-predicting HKLs, and access to CIF text does not close the gap in crystallographic calculation. Alongside model rankings, the benchmark identifies the conditions under which current VLMs fail on quantitative scientific figures. All data and evaluation code will be publicly available.
[AI-112] Reason Light: A Multimodal Foundation Model-Enhanced Reinforcement Learning Framework for Zero-Shot Traffic Signal Control
链接: https://arxiv.org/abs/2605.29425
作者: Aoyu Pang,Maonan Wang,Yuejiao Xie,Chung Shue Chen,Zhiwei Yang,Man-On Pun
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning (RL) has shown promise in traffic signal control (TSC). However, its reliance on predefined states limits responsiveness to observable open-world events that are absent from training data. IoT-enabled intersections provide heterogeneous observations from roadside sensors and cameras, creating opportunities to improve RL adaptability to such events. To this end, we propose ReasonLight, a multimodal foundation model-enhanced RL framework for zero-shot TSC. ReasonLight integrates three sources of information: structured traffic measurements, multi-view camera observations, and candidate phase decisions from a pre-trained RL controller. Given an RL-proposed phase, ReasonLight extracts visual semantics from multi-view images and aligns them with compact sensor-derived scene descriptions. This alignment enables a semantic-guided refinement module to either preserve or adjust the proposed action according to traffic rules and event semantics. To ensure operational reliability, refined actions are constrained by the set of available phases. Any invalid decision is rejected, and the system falls back to the original RL action. We evaluate ReasonLight on two types of rare events not seen during RL training: emergency vehicle priority and temporary traffic regulation. Experimental results show that ReasonLight achieves zero-shot adaptation without retraining. It reduces emergency vehicle waiting time by up to 88.7% compared with the RL-only backbone while preserving comparable routine traffic performance.
[AI-113] When Does Persona Prompting Actually Help? A Retrieval and Metric Analysis of Expert Role Injection in LLM s
链接: https://arxiv.org/abs/2605.29420
作者: Shuai Xiao,Su Liu,Weikai Zhou,Jialun Wu,Xinjie He,Zhiyuan Lin,Qiyang Xie
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 6 pages, 2 figures. Submitted for peer review
Abstract:Persona prompting is widely used to steer large language models, yet its practical value remains unclear. Prior work often evaluates persona prompting using aggregate scores, making it difficult to determine whether expert-role prompting consistently improves response quality or instead changes responses along different quality dimensions. We study this question through a controlled comparison of four prompting conditions across 1,140 open-ended questions spanning 38 expert roles and six domains: no role prompt, a generic domain-expert prompt, embedding-based role retrieval, and a hybrid retrieval method combining embedding search with LLM-based role selection. Aggregate results show only small overall differences between conditions. However, metric-level analysis reveals a consistent tradeoff that aggregate averages obscure: role prompting systematically increases expertise depth while reducing clarity. These effects are highly conditional rather than universal. Role prompting performs best on advisory questions and in domains such as medicine and psychology, where structured expert framing and risk communication are intrinsically valuable. In contrast, baseline prompting performs better on conceptual and explanatory questions in finance, legal, science, and technology domains, where concise plain-language explanation is more important. We further show that hybrid retrieval significantly improves over embedding-only role selection, although better role retrieval does not eliminate the broader expertise-depth versus clarity tradeoff. Overall, our findings suggest that persona prompting primarily reshapes response characteristics rather than broadly improving capability, and that multi-metric evaluation is necessary for understanding its effects.
[AI-114] he Good the Bad and the Ugly of Markov Boundary for Tabular Prediction
链接: https://arxiv.org/abs/2605.29411
作者: Shu Wan,Abhinav Gorantla,Huan Liu,K. Selçuk Candan
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Methodology (stat.ME); Machine Learning (stat.ML)
备注: 11 pages, 9 figures, 2 tables. Preprint
Abstract:Under standard graphical assumptions, the Markov boundary of a target variable is the smallest set of features that renders every other feature redundant. Once the boundary is observed, the target is conditionally independent of the rest of the table. This is a tempting object for tabular prediction, since it names exactly the columns a model should need. Yet modern regressors are still trained on the full feature set. We ask whether the Markov boundary is genuinely useful for prediction on SCM3K, a 3,450-task synthetic SCM benchmark with feature counts from 40 to 1000 and six SCM families, evaluated with six regressors. The answer is more nuanced than the theory suggests. Restricting a regressor to the oracle boundary often improves prediction substantially, and the improvement grows as the feature space becomes larger and sparser. But the natural pipeline of recovering the boundary with causal discovery and training on the recovered mask does not deliver. Existing estimators exhaust the compute budget before reaching the regime where the boundary helps most, and even where they run they rarely beat the full feature set. We trace this to three causes. Discovery optimizes structural recovery rather than prediction. False negatives and false positives carry sharply asymmetric predictive cost. The exact boundary is only one of many feature sets that beat all features. We then develop what these facts imply for prediction-aligned feature selection and for tabular models that learn to use causal structure.
[AI-115] GDSD: Reinforcement Learning as Guided Denoiser Self-Distillation for Diffusion Language Models
链接: https://arxiv.org/abs/2605.29398
作者: Xiaohang Tang,Keyue Jiang,Che Liu,Qifang Zhao,Xiaoxiao Xu,Sangwoong Yoon,Ilija Bogunovic
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Preprint
Abstract:Reinforcement learning (RL) can be used to improve the policy (denoiser) of diffusion large language models (dLLMs), while being hindered by the intractability of the policy likelihood. A dominant and efficient family of methods replaces the likelihood in standard RL with its evidence lower bound (ELBO), estimated from randomly masked sequences. Despite being well aligned with pre-training, these approaches introduce bias through training–inference mismatch by using the ELBO as a likelihood surrogate, which can degrade performance. In this work, we propose Guided Denoiser Self-Distillation (GDSD) to directly distill the denoiser of dLLMs from an advantage-guided self-teacher, derived from the closed-form optimum of reverse-KL regularized RL. GDSD matches the dLLM’s denoiser logits to the teacher’s via a normalization-free objective, which reduces RL to likelihood-free self-distillation and thus bypasses the TIM biases. Recent ELBO-based methods emerge as instances of applying different distillation divergences, but with diagnosable pathologies that GDSD avoids. On planning, math, and coding benchmarks with LLaDA-8B and Dream-7B, GDSD consistently outperforms prior state-of-the-art ELBO-based methods with a more stable training reward dynamics, achieving test-accuracy improvements of up to +19.6% . These results suggest that direct denoiser self-distillation, without relying on an ELBO likelihood surrogate, can provide a more stable and effective RL procedure for dLLMs. Code is available at this https URL.
[AI-116] Aligned but Frag ile: Enhancing LLM Safety Robustness via Zeroth-Order Optimization
链接: https://arxiv.org/abs/2605.29396
作者: Zhihao Liu,Yifan Wu,Jian Lou,Di Wang,Yuxi Zhou,Yuke Hu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Safety alignment for large language models (LLMs) aims to reduce harmful or unsafe behavior while preserving general utility. However, recent findings reveal that alignment effects can be fragile: lightweight post-alignment manipulations, such as parameter noise, activation noise, or quantization, can easily weaken the intended safety behavior. Prior efforts to improve robustness have primarily focused on data curation, modified alignment objectives, and safety-critical parameter identification, leaving the role of the optimizer itself largely unexplored. In this paper, we are the first to study the robustness of safety alignment from the perspective of the base optimizer. This optimizer-centric view naturally points to zeroth-order optimization, which provides a robustness-oriented signal by evaluating safety alignment under perturbations. Based on this insight, we propose a hybrid framework that first performs standard first-order safety alignment and then applies zeroth-order refinement to improve robustness. Both theoretically and empirically, we show that only a few zeroth-order refinement steps can enhance robustness while preserving safety alignment. We further improve the efficiency of zeroth-order refinement by exploiting its inherent perturbation-based evaluations to estimate layer-wise robustness sensitivity, enabling the refinement process to concentrate updates on robustness-critical layers with modest training overhead. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2605.29396 [cs.AI] (or arXiv:2605.29396v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2605.29396 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-117] EvoMD-LLM : Learning the Language of Species Evolution in Reactive Molecular Dynamics ACL
链接: https://arxiv.org/abs/2605.29394
作者: Zhichen Tang,Zhengzheng Dang,Yulin Chen,Jixin Wu,Haiwen Li,Yanming Wang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 17 pages, ACL Findings
Abstract:While large language models (LLMs) excel at static scientific reasoning, they struggle to model the temporal structure of dynamic physical processes. We present EvoMD-LLM (Evolutionary Molecular Dynamics Large Language Model), a framework that reformulates species-level molecular dynamics as a symbolic temporal language modeling problem. Reactive MD trajectories are discretized into sequences of molecular events, where each token represents a chemical species augmented with its persistence duration, enabling standard autoregressive LLMs to learn compositional evolution over time through efficient fine-tuning. A key component of EvoMD-LLM is temporal scaffolding, which treats event duration as an explicit linguistic token and serves as a structured inductive bias, significantly reducing invalid or hallucinated molecular outputs compared to conventional sequence modeling approaches. We evaluate EvoMD-LLM on multiple temporal prediction tasks, achieving up to 66.14% accuracy and consistently outperforming sequential neural networks and language-based baselines. Beyond quantitative improvements, we qualitatively observe that the model is capable of generating interpretations for its own predictions by incorporating relevant chemical knowledge, even though it was not explicitly supervised with paired trajectory-explanation data. These results demonstrate that symbolic temporal language modeling provides an effective framework for grounding LLMs in dynamic physical simulations.
[AI-118] On the Optimizer Dependence of Neural Scaling Laws
链接: https://arxiv.org/abs/2605.29387
作者: Vansh Ramani,Shourya Vir Jain
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:The scaling exponent \alpha in neural scaling laws L(N) \propto N^-\alpha is commonly treated as a fixed constant set by architecture and data. We present evidence that \alpha depends systematically on the optimizer. In controlled random-feature regression experiments – the canonical theoretical framework for neural scaling – we measure \alpha across five optimizer variants and six spectral conditions. Preconditioned optimizers consistently yield steeper scaling (larger \alpha ), with the \alpha -shift increasing across most of the tested spectral range, peaking near s = 1.5 , and remaining large at s = 2.0 . At s \approx 1.0 (characteristic of natural language), the full natural gradient achieves \alpha \approx 0.31 versus \alpha \approx 0.12 for gradient descent – a 2.6\times larger fitted exponent that, within the random-feature model, compounds with each model-size doubling. Whether and how this exponent shift transfers to large-scale LLM training – where recent evidence suggests the advantage may attenuate with scale – remains an important open question. Our results imply that scaling-law forecasts should account for optimizer choice, and we provide a spectral diagnostic predicting when advanced optimizers will pay off.
[AI-119] MiraBench: Evaluating Action-Conditioned Reliability in Robotic World Models
链接: https://arxiv.org/abs/2605.29360
作者: Tianzhuo Yang,Zihan Shen,Zirui Mi,Zhaoyi Zhang,Jiayi Zhou,Jiaming Ji,Juntao Dai,Jiawei Chen,Boyuan Chen,Yaodong Yang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Action-conditioned world models are increasingly used as scalable simulators for robot learning, yet current evaluations provide limited evidence that their predictions are reliable under the actions they condition on. Existing benchmarks largely emphasize visual fidelity, leaving unclear whether predicted futures are physically plausible, faithful to commanded actions, and calibrated to failure when actions should not succeed. We introduce \textscMiraBench, a hierarchical benchmark that defines \emphaction-conditioned reliability as a core evaluation target for robotic world models. MiraBench decomposes this target into three progressively demanding levels: \emphPhysics Adherence, which evaluates reference-free physical consistency; \emphAction-Following Fidelity, which measures whether predictions respect task-relevant action inputs; and \emphOptimism Bias Detection, which probes the tendency to predict successful outcomes under failure-inducing actions. To support this evaluation, we curate a human-annotated corpus with over 16,000 judgments across tasks, failure categories, and leading world models. We evaluate 12 representative model configurations spanning vector-conditioned robotic world models, text-conditioned generative world models, open-weight systems, closed-source systems, and multiple model scales. Across this broad model landscape, MiraBench reveals three central findings: visual fidelity is a poor proxy for action fidelity; increasing model scale does not reliably improve action following; and optimism bias is pervasive across current systems. By shifting evaluation from appearance to action-conditioned reliability, MiraBench provides a diagnostic foundation for assessing and improving robotic world models as faithful simulators.
[AI-120] Does Distributed Training Undermine Compute Governance? ICML2026
链接: https://arxiv.org/abs/2605.29359
作者: Robi Rahman
机构: 未知
类目: Computers and Society (cs.CY); Artificial Intelligence (cs.AI)
备注: TAIGR workshop in ICML 2026
Abstract:Compute governance proposals often rely on the assumption that frontier AI training requires large, detectable computing clusters. However, recent advances in distributed training algorithms could allow developers to conduct frontier-scale training on distributed agglomerations of hardware, rather than needing large datacenter facilities. Developers who prefer not to be constrained by regulations may structure their hardware in a manner that evades the registration and monitoring requirements associated with compute governance. Therefore, regulations must be designed to detect and prevent illicit distributed training operations. This paper evaluates the feasibility of such evasion and outlines recommended countermeasures, including whistleblowing, chip tracking, forensic accounting, and memory and compute thresholds for clusters.
[AI-121] Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet
链接: https://arxiv.org/abs/2605.29358
作者: Adly Templeton,Tom Conerly,Jonathan Marcus,Jack Lindsey,Trenton Bricken,Brian Chen,Adam Pearce,Craig Citro,Emmanuel Ameisen,Andy Jones,Hoagy Cunningham,Nicholas L Turner,Callum McDougall,Monte MacDiarmid,Alex Tamkin,Esin Durmus,Tristan Hume,Francesco Mosconi,C. Daniel Freeman,Theodore R. Sumers,Edward Rees,Joshua Batson,Adam Jermyn,Shan Carter,Chris Olah,Tom Henighan
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We demonstrate that sparse autoencoders can extract interpretable features from Claude 3 Sonnet, a production-scale language model, addressing the open question of whether dictionary learning methods scale beyond small transformers. We trained sparse autoencoders with up to 34 million features on the model’s middle layer residual stream, using scaling laws to guide hyperparameter selection. The resulting features are multilingual and multimodal (generalizing to images despite text-only training), respond to both concrete instances and abstract discussions of concepts, and can be used to steer model behavior in ways consistent with their interpretations. We find features corresponding to famous entities and locations, as well as more abstract concepts like sarcasm or errors in code. We also identify features relevant to ways in which language models might cause harm–including features representing deception, power-seeking, sycophancy, and bias–and show that these causally influence model outputs when manipulated. Additionally, we conduct analyses of feature interpretability, geometry, and computational function. However, significant limitations remain: our suite of features is incomplete, and we lack rigorous methods for evaluating whether our features faithfully capture model computations.
[AI-122] PassNet: Scaling Large Language Models for Graph Compiler Pass Generation
链接: https://arxiv.org/abs/2605.29357
作者: Yiqun Liu,Yingsheng Wu,Ruqi Yang,Enrong Zheng,Honglei Qiu,Sijun He,Tai Liang,Jingjing Wu,Yuhan Zhou,Yiwei Zhang,Dongyan Chen,Weihan Yi,Xinqi Li,Siqi Bao
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Programming Languages (cs.PL)
备注: Code and data available at this https URL
Abstract:Modern tensor compilers such as TorchInductor deliver substantial speedups on mainstream models, yet face a systematic performance ceiling on long-tail workloads – our profiling shows that 43% of real-world subgraphs experience end-to-end slowdowns under default compilation. While LLMs offer a path toward automated optimization, existing efforts focus on standalone kernel generation. We argue that pass generation – where LLMs author structured graph transformations that integrate directly into compiler pipelines – is the more appropriate abstraction. We propose PassNet, the first large-scale ecosystem for LLM-based compiler pass generation, comprising: (1) PassNet-Dataset, over 18K unique computational graphs from 100K real-world models; and (2) PassBench, 200 curated long-tail fusible tasks (comprising 2,060 subgraphs in total) evaluated under the Error-aware Speedup Score (ES_t) – a metric unifying correctness, stability, and performance – with layered integrity defenses against systematic LLM exploitation. Experiments reveal that PassBench is both highly discriminative and genuinely unsaturated: the best frontier model trails TorchInductor by 37% in aggregate, yet on individual subgraphs LLMs achieve up to 3x speedup over the same compiler – indicating that the bottleneck is consistency, not capability. Fine-tuning a small model on merely ~4K PassNet trajectories yields a 2.67x improvement approaching frontier-model performance, demonstrating substantial headroom and validating PassNet as live training infrastructure for advancing LLM-driven compiler optimization. All data, benchmarks, and tooling are publicly available.
[AI-123] ConMoE: Expert-Pool Consolidation via Prototype Reassignment for MoE Compression
链接: https://arxiv.org/abs/2605.29350
作者: Yilun Yao,Jiaming Pan,Elsie Dai,Peizhuang Cong,Yaoming Li,Tong Yang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 12 pages, 3 figures, 5 tables
Abstract:Mixture-of-Experts (MoE) language models reduce per-token computation but still require storing and serving all experts, making deployment memory-intensive. Existing post-training compression methods mainly shrink this cost by pruning experts or merging their weights. We formulate post-training MoE compression as expert-pool consolidation: retaining a smaller set of pretrained experts as reusable prototypes and deterministically remapping each original expert reference to one selected prototype. This view separates the reduced expert pool from the reuse structure that represents the original expert slots, and allows prototype sharing within local layer scopes while preserving the original router interface. We propose ConMoE, a train-free prototype remapping framework that selects retained experts using calibration-based contribution and replaceability signals, then redirects original expert calls to the selected prototypes without weight updates or post-compression fine-tuning. Experiments on three pretrained MoE language models show that ConMoE matches or outperforms strong pruning and merging baselines in several settings, achieving the best average score on deepseek-moe-16b-base at both 25% and 50% routed-expert reduction, while remaining competitive on Qwen3-30B-A3B and OLMoE-1B-7B-0125. Ablations indicate that deterministic reassignment is the most stable component, whereas broader cross-layer sharing and post-hoc weight fusion are model-dependent.
[AI-124] Entropy-KL Divergence-based Token Masking: A Novel Approach for Selective Fine-tuning of Large Language Models
链接: https://arxiv.org/abs/2605.29303
作者: Qi Liu,Mingdi Sun,Yongyi He,Zhi Zheng,Tong Xu,Yi Zheng,Zhefeng Wang,Enhong Chen
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 17 pages
Abstract:Supervised fine-tuning (SFT) followed by reinforcement learning (RL) has become a standard post-training paradigm for large language models. This paradigm provides a cold-start for RL exploration, avoiding the inefficiency of pure RL where on-policy sampling yields insufficient positive samples. However, in practice, existing approaches often use a small amount of data for SFT initialization compared to the RL phase, which can cause the model to fit the limited samples and shift away from its pre-trained distribution. This distribution shift impedes the model’s ability to effectively explore during subsequent RL training. To address this challenge, we propose that in low-data regimes, SFT should prioritize activating task-relevant capabilities rather than memorizing specific content. Along this line, we propose EKSFT (Entropy-KL Selective Fine-Tuning), which selectively masks tokens that exhibit either high entropy or high KL divergence from a reference model. By excluding these high-uncertainty, distribution-shifting tokens from imitation, EKSFT injects task-specific knowledge while preserving the integrity of the model’s pre-trained distribution. Empirical evaluations on mathematical reasoning benchmarks demonstrate that EKSFT consistently outperforms standard SFT. Further RL fine-tuning from the EKSFT model yields consistently better post-RL performance, indicating improved exploration for the RL stage. Our codes and datasets are available at this https URL.
[AI-125] Diagnosing Harmful Continuation in Answer-Correct Long-CoT Training Traces
链接: https://arxiv.org/abs/2605.29288
作者: Chen He,Yuhao Wu,Lei Wang,Wenxuan Zhang,Fumin Shen
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Long chain-of-thought (CoT) traces are widely used as supervision for reasoning-oriented LLM SFT, yet answer-correct traces can still lead to markedly different fine-tuning outcomes. We study post-conclusion continuation in answer-correct long-CoT data: a continuation where the answer appears sufficiently supported, but the trace continues with additional reasoning that remains in the supervised target. To test its training effect, we use a delete-only editor to construct answer-preserving suffix removal and compare CoT-based SFT on the original and processed traces. We observe improved SFT outcomes after removing the editor-identified post-conclusion continuation, suggesting that this continuation is harmful to training in our setting. We therefore refer to this empirically supported phenomenon as harmful continuation. Beyond this intervention, we further characterize the removed post-conclusion continuation through uncertainty and hidden-state progress. We observe persistent local uncertainty together with weakened terminal-directional progress, forming an uncertainty–geometry mismatch. Finally, we instantiate Harmful Continuation Cut (HCC), a lightweight boundary proxy that approximates the editor-identified post-conclusion continuation boundary.
[AI-126] Do Physics Foundation Models Learn Generalizable Physics? A Bias-Aware Benchmark Across Physical Regimes and Distribution Shifts
链接: https://arxiv.org/abs/2605.29283
作者: Mengdi Chu,Yang Liu,Ayan Biswas,Han-Wei Shen
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 26 pages, 31 figures
Abstract:Recent physics foundation models claim general spatiotemporal forecasting ability, yet their evaluations often collapse performance into a single average score under a fixed training distribution. This makes it difficult to determine whether a model has learned generalizable physical dynamics or only performs well under particular settings. We construct a benchmark with 8 physical dynamics, 3 training-data mixtures, and 25 test regimes induced by dynamic-scale and initial-condition complexity shifts, covering in-distribution, distribution-shift, and out-of-distribution settings. We evaluate five physics foundation model architectures and four model variants per architecture (scratch and three pretrained sizes), resulting in 60,000 measurements. Our results show that current physics foundation models behave as conditional rather than universal generalists: their generality depends on the physical regime, temporal scale, initial-condition setting, pretraining, model size, and architecture. Improving the training data distribution only partially mitigates this limitation. Pretraining and scaling are also unable to reliably remove their ability biases. We argue that improving physics foundation models requires moving beyond scaling models or expanding data, toward learning mechanisms that better capture transferable physical knowledge across regimes, temporal scales, and distribution shifts.
[AI-127] Code-QA-Bench: Separating Code Reasoning from Documentation Memorization in Repository-Level QA
链接: https://arxiv.org/abs/2605.29277
作者: Jun Zhang,JianYing Qu,Hanwen Du,Zhongkai Sun,Yehua Yang,Qiao Zhao
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:We present Code-QA-Bench, a fully automated framework for synthesizing repository-level code understanding benchmarks that separates genuine code comprehension from documentation recall and pretraining memorization. The framework makes two methodological contributions: (1) an answer-first generation pipeline where a tool-equipped agent explores source code to produce verified gold answers before deriving questions, ensuring every task is grounded in real code structure; and (2) a three-condition experimental design evaluating agents under closed-book (no repository), code-only (documentation removed), and documented (full repository) conditions, with deltas directly quantifying documentation utility and memorization. We generate 528 code-derivable and 100 doc-dependent tasks across 10 Python repositories from SWE-Bench, scored by an LLM judge on accuracy, completeness, and specificity. Experiments on four frontier models reveal that code access is the dominant factor (+0.23 mean gain over closed-book), documentation provides modest additional benefit (+0.071 on doc-dependent tasks), and code-only \approx documented on code-derivable tasks, validating the design. The framework is open-source and applicable to any well-documented Python repository. Subjects: Software Engineering (cs.SE); Artificial Intelligence (cs.AI) Cite as: arXiv:2605.29277 [cs.SE] (or arXiv:2605.29277v1 [cs.SE] for this version) https://doi.org/10.48550/arXiv.2605.29277 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-128] Causal Label Recovery in Payment Networks
链接: https://arxiv.org/abs/2605.29272
作者: Gaurav Dhama
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: 49 pages
Abstract:Fraud detection models in payment networks train on chargeback labels that are systematically biased. Every label must survive three sequential gates: authorization (declined transactions generate no labels), issuer reporting (unreported fraud is invisible), and delay (pending chargebacks are missing at training time). Labels that do arrive may be corrupted by first-party misuse or issuer misclassification. A companion paper [arXiv:2605.27557] proved that these four impairments impose a minimax lower bound on detection performance. This paper asks: can that bound be achieved? We formalize the observation pipeline as a sequential missing-data problem with three propensity stages and a corruption layer, and construct the Sequential Triply Robust (STR) estimator. The STR corrects for all four impairments simultaneously and achieves the semiparametric efficiency bound – no estimator can have lower asymptotic variance. It is sequentially triply robust: at each gate, consistency requires only that either the propensity model or the outcome regression is correctly specified, not both. We provide corruption correction via noise-rate-adjusted pseudo-labels, empirical Bayes shrinkage to stabilize inverse-propensity weights for small issuers, a plug-in variance estimator yielding valid confidence intervals, and a Bernstein concentration inequality for finite-sample guarantees. On the operational side, we derive the optimal training delay – the maturity window that minimizes the sum of label-quality loss and model staleness – and prove that the STR permits training on data that is days old rather than months old, decoupling model freshness from the chargeback maturity cycle. The STR provably dominates naive chargeback-based training in mean squared error for any sample size. Comments: 49 pages Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML) Cite as: arXiv:2605.29272 [cs.LG] (or arXiv:2605.29272v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2605.29272 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-129] Indexing the Unreadable: LLM -Native Recursive Construction and Search of Service Taxonomies EMNLP2026
链接: https://arxiv.org/abs/2605.29270
作者: Wei Zheng,Yang Yan,Yiyang Shao,Jinyang Li,Zeze Chang,Yukuang Jia,Qiming Mao,Chihyung Wang,Jingbin Zhou
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: Preprint. 8 pages main paper + appendix; 2 figures. Under submission to EMNLP 2026
Abstract:The era of the Internet of Agents (IoA) is taking shape: LLM agents are expected to fulfill user goals by orchestrating fast-growing populations of Model Context Protocol (MCP) servers, Agent-to-Agent (A2A) endpoints, reusable skills, and other LLM-callable services. Yet LLMs face a structural mismatch with this regime: effective context is a scarce resource that does not scale with the number of services. Concatenating thousands of service descriptions into a prompt overflows the context window, and even when the window is large enough, models systematically under-attend to information in the middle of long inputs, the well-documented Lost-in-the-Middle phenomenon. This is fundamentally a question of context management for service discovery. To address this, we propose an LLM-native progressive-disclosure scheme and its concrete instantiation, A2X (Agent-to-Anything service discovery): an LLM-driven pipeline that automatically organizes the registered services into a hierarchical taxonomy and walks it layer by layer at query time, so that every LLM call sees only a small candidate set highly relevant to the user query. This decouples effective-context scarcity from registry size and significantly reduces token consumption while improving retrieval accuracy. Compared to full-context dumping, A2X achieves a 6.2-point Hit Rate gain at one-ninth the prompt-token cost; compared to the state-of-the-art open-source embedding-based baseline, A2X improves Hit Rate by more than 20 points.
[AI-130] When and How Human Curation Backfires: Preference Alignment under Multi-Model Self-Consuming Loop
链接: https://arxiv.org/abs/2605.29267
作者: Yang Zhang,Xiukun Wei,Xueru Zhang
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Foundation models are increasingly trained on synthetic data generated by prior model iterations rather than exclusively on real data. This self-consuming training paradigm can lead to model collapse, divergence, or bias amplification. Recent work (Ferbach et al., 2024) shows that incorporating human curation into the loop can steer a self-consuming model toward human-aligned behavior, but these analyses focus on a single, isolated model that solely consumes its own outputs. In practice, however, models often interact and train on input-output pairs produced by other models. This paper studies self-consuming training in the multi-model regime. We first formalize a framework for interacting self-consuming models and characterize when the resulting dynamical system converges to a stable point. We then examine how human curation of one model affects its own alignment (self-influence) and how such effects propagate to other models (cross-influence). Unlike isolated settings where human curation always enhances model alignment, we show that cross-model interactions can dampen or even invert this effect, ultimately degrading long-term alignment.
[AI-131] Harmonizing Real-Time Constraints and Long-Horizon Reasoning : An Asynchronous Agent ic Framework for Dynamic Scheduling
链接: https://arxiv.org/abs/2605.29262
作者: Shijie Cao,Yuan Yuan,Jing Liu
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:The Dynamic Flexible Job Shop Scheduling Problem (DFJSP) necessitates a trade-off between instant reaction to stochastic disturbances and global optimization of production goals. Conventional priority rules are insufficiently flexible to handle complex disruptions, whereas learning-based approaches often compromise interpretability or fail to generalize across problem scales. Although Large Language Models (LLMs) offer advanced reasoning capabilities to bridge this gap, their substantial inference latency is incompatible with the millisecond-level decision cycles of industrial control systems. To resolve this conflict, we introduce RACE-Sched, an asynchronous agent-based framework that decouples policy execution from logical reasoning via a dual-stream architecture. The Reactive Stream executes low-latency symbolic heuristics to enable real-time dispatching, while the parallel Deliberative Stream leverages an LLM to synthesize, validate, and evolve these rules. Candidate rules undergo rigorous testing in a sandbox and are deployed via atomic updates, ensuring safety without blocking the control loop. Additionally, a semantic rule repository indexes validated heuristics for retrieval-based initialization which enhances transferability across problem scales. Extensive evaluations on GEN-Bench, MK-Bench, and JMS-Bench demonstrate that RACE-Sched outperforms leading Deep Reinforcement Learning and other LLM-based baselines. This approach harmonizes real-time constraints with long-horizon reasoning to achieve superior solution quality and robust adaptation to dynamic events.
[AI-132] KLAS: Using Similarity to Stitch Neural Networks for Improved Accuracy-Efficiency Tradeoffs
链接: https://arxiv.org/abs/2605.29259
作者: Debopam Sanyal,Anantharaman Iyer,Alind Khare,Trisha Jain,Akshay Jajoo,Myungjin Lee,Clayton Kerce,Alexey Tumanov
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Given the wide range of deployment targets, flexible model selection is essential for optimizing performance within a given compute budget. Recent work demonstrates that stitching pretrained models within a model family enables cost-effective interpolation of the accuracy-efficiency tradeoff space. Stitching transforms intermediate activations from one pretrained model into another, producing a new interpolated stitched network. Such networks provide a pool of deployment options along the accuracy-efficiency spectrum. However, existing stitching approaches often yield suboptimal tradeoffs and lack generalizability, as they primarily rely on heuristics to select stitch configurations. We argue that constructing improved accuracy-efficiency tradeoffs requires explicitly capturing and leveraging the similarity between pretrained models being stitched. To this end, we introduce KLAS, a novel stitch selection framework that automates and generalizes stitch selection across model families by leveraging KL divergence between intermediate representations. KLAS identifies the most promising binary stitches from the O(k^2n^2) possibilities for k pretrained models of depth n . Through comprehensive experiments, we demonstrate that KLAS improves the accuracy-efficiency curve of stitched models at the same finetuning cost as baselines. KLAS achieves up to 1.21% higher ImageNet-1K top-1 accuracy at the same computational cost, or maintains accuracy with a 1.33\times reduction in FLOPs.
[AI-133] Extreme dynamic symmetry enables omnidirectional and multifunctional robots
链接: https://arxiv.org/abs/2605.29254
作者: Jiaxun Liu,Boxi Xia,Boyuan Chen
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: Published in Science Robotics (2026). Our project website is at: this https URL
Abstract:Symmetry is a central organizing principle in natural systems, yet its use as a unifying design strategy in robotics has largely remained limited to geometric form. We show that symmetry can instead be leveraged at the level of dynamic actuation capability. We introduce dynamic symmetry, the uniformity of a robot’s attainable center-of-mass accelerations, and formalize it through a measure coined as dynamic isotropy. Across more than 1000 simulated morphologies, we found that higher dynamic symmetry consistently improved trajectory tracking, task success, robustness, resiliency, and energy efficiency, with the benefits becoming most pronounced as dynamic isotropy approached its theoretical limit. To study this regime systematically, we developed Argus, a family of spherical robots designed to explore the effects of increasing dynamic symmetry. Members of the Argus family vary in their actuation geometry and dynamic symmetry level while sharing a common architectural principle: radially oriented linear actuators that directly shape the robot’s center-of-mass dynamics. Among them, we built a physical 20-leg Argus variant that achieved near-extreme dynamic isotropy and demonstrated orientation-invariant locomotion, agile traversal of cluttered and deformable terrain, rapid self-stabilization, and resilience to partial actuator failures. Its distributed sensing further enabled omnidirectional perception and object interaction during continuous motion. These results show that designing robots for symmetry not only in morphology but also in their attainable dynamics provides a powerful and general pathway toward agility, robustness, and multifunctionality in uncertain terrestrial and extraterrestrial environments.
[AI-134] OpenClawBench: Benchmarking Process-side Anomalies in Real-world Agent Execution Trajectories
链接: https://arxiv.org/abs/2605.29253
作者: Yibing Liu,Yangze Liu,Xiaolong Yin,Bin Wang,Chong Zhang,Hao Yin,Zhongyi Han
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 37 pages, 1 figure, 43 tables
Abstract:Task success can hide process anomalies in real-world agent executions. An agent may pass the final task oracle while still accumulating unresolved ambiguity, unsafe external writes, ignored errors, weakly grounded commitments, or capability-boundary overcommitment. We study this mismatch as the Outcome-Process Gap and introduce OpenClawBench, a large-scale dataset for measuring and supervising process-side anomalies in real agent execution processes. OpenClawBench is built from BFCL-driven OpenClaw sessions produced by 6 source models and contains 31,264 annotated trajectories. It aligns task-oracle outcomes with structured process evidence. FullTax converts the aligned trajectories into structured anomaly supervision: binary labels, supporting evidence, onset/span localization, severity, recoverability, and a 5-class anomaly taxonomy. Using OpenClawBench, we make the Outcome-Process Gap measurable. Among 31,135 oracle-passing executions, 2,904 are still labeled process-anomalous under FullTax. These results show that success-only evaluation misses a concrete class of process-side failures in real agent executions. A LoRA-fine-tuned Gemma 3 12B detector trained on the high-confidence FullTax supervised pool reaches binary F1=0.729 on the cleaner-labels held-out test split. Together, OpenClawBench turns real agent execution logs into auditable and reusable supervision for studying, diagnosing, and operationally monitoring runtime agent reliability.
[AI-135] Provably Secure Agent Guardrail
链接: https://arxiv.org/abs/2605.29251
作者: Benlong Wu,Weiming Zhang,Kejiang Chen,Han Fang,Nenghai Yu
机构: 未知
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:As large language models transition from bounded generative engines to agents with expansive execution privileges, AI going out of control precipitates a fundamental crisis in artificial intelligence security. Existing defense architectures heavily rely on empirical semantic guardrails and probabilistic large model adjudicators, mechanisms that fail to provide deterministic security lower bounds when facing complex semantic symbol decoupling attacks. To overcome this empirical semantic guardrail dilemma, this paper proposes a new security paradigm for agents based on the fundamental limitations of logical reasoning. Based on this paradigm, we further introduce an executable Proof-Constrained Action (ePCA) framework with a neural symbolic isolation architecture. This framework abandons semantic trust in natural language, forcing agents to losslessly formalize their intentions into first-order logical mathematical constraints before performing physical operations. Empirical evaluations of macroscopic and microscopic two-dimensional dynamic adversarial systems demonstrate that our formal verification mechanism achieves zero attack success rate and zero false positive rate across the evaluated scenarios, with extremely low computational latency. This research provides a conditional formal foundation under explicit system assumptions and an engineering paradigm for constructing the underlying defense foundation for future intelligent systems.
[AI-136] BlockBatch: Multi-Scale Consensus Decoding for Efficient Diffusion Language Model Inference
链接: https://arxiv.org/abs/2605.29233
作者: Xiaoyou Wu,Cheng-Jhih Shih,Binfei Ji,Yong Liu,Yingyan(Celine)Lin
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 23 pages, including references and appendices
Abstract:Diffusion language models (dLLMs) generate text by iteratively denoising multiple token positions in parallel, offering an attractive alternative to strictly autoregressive decoding. In practice, however, block-wise dLLM inference exposes a difficult granularity trade-off: small blocks preserve local conditioning but require many denoising steps, whereas large blocks expose more parallelism but can make premature commitments and accumulate cache error. Existing acceleration methods typically choose a single block size per request, leaving the complementarity among block sizes unused. We show that block size itself is a useful branching dimension. Different block sizes induce related but non-identical KV-cache trajectories: branches often share an initial prefix, bifurcate at semantically decisive positions, and later agree on syntactically lightweight tokens. Motivated by this structure, we propose BlockBatch, a training-free online inference framework that executes multiple block-size branches for the same request inside a batched forward pass. BlockBatch coordinates these branches through confidence-gated token merging, leader-based synchronization, and periodic full-sequence refreshes that re-anchor local block updates to a globally consistent KV state. Across 3 representative dLLMs and 4 datasets, BlockBatch reduces denoising NFEs by 26.6% on average and achieves a 1.33 \times average end-to-end speedup over Fast-dLLM while preserving accuracy. These results identify block-size diversity as a practical and previously underexplored axis for branch-parallel dLLM inference.
[AI-137] ailoring the Curriculum: Student-Centered Reasoning Distillation via Dynamic Data-Model Compatibility
链接: https://arxiv.org/abs/2605.29229
作者: Jiahao Huang,Fei Cheng,Junfeng Jiang,Akiko Aizawa
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Reasoning distillation transfers complex reasoning abilities from large language models (LLMs) to smaller ones, yet its success depends on how well the training data align with the student model. This paper introduces the Data-Model Compatibility (DMC) metric, which can be used to assess the suitability of a dataset for reasoning distillation on a student model. DMC provides an assessment by jointly considering data quality, relative difficulty, and student capability. We validated the effectiveness of DMC from two perspectives: (1) DMC exhibits a strong correlation with reasoning distillation performance; and (2) using DMC as the criterion for data selection leads to improved reasoning distillation performance. Both findings are consistently demonstrated across multiple student models and tasks. Moreover, since the DMC of each dataset dynamically changes during training, our experiments demonstrate that dynamically selecting datasets based on DMC can further enhance performance.
[AI-138] BenchTrace: A Benchmark for Testing Reflection Ability and Controlled Evolution in LLM Agents
链接: https://arxiv.org/abs/2605.29225
作者: Jiahao Huang,Fei Cheng,Junfeng Jiang,Zefan Yu,Akiko Aizawa
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Self-evolving agents improve over time by reflecting on past failures, but existing evaluation is limited in two ways: it measures only task scores, leaving reflection quality unknown, and it relies on agents’ own episode runs, offering no mechanism to target specific failure patterns. We present \textbfBenchTrace, a benchmark for evaluating self-evolution ability in LLM agents. BenchTrace is built on a snapshot-reflection dataset of 1,821 annotated episodes spanning six diverse tasks, and comprises a \textbfReflection Evaluation that probes failure identification through targeted QA tasks, and an \textbfEvolution Evaluation that tests whether past failure experience translates into avoidance behavior in a controlled self-evolution simulation. Building on BenchTrace, we propose \textbffailure avoidance rate (FAR), a new evaluation metric measuring the fraction of test cases in which the agent successfully avoids the target failure instance. Experiments with Qwen3-32B and GPT-4.1 reveal that both models fall below a 30% end-to-end pass rate on reflection evaluation, with diagnosis as the primary bottleneck. Evolution evaluation shows that self-evolution methods generally improve FAR over the non-evolving baseline, but agents forget early lessons as noise episodes accumulate, and agents fail to generalize their reflections beyond the specific context, causing negative transfer across task contexts. Our correlation analysis further reveals that only a fully correct reflection is strongly associated with higher FAR. BenchTrace exposes concrete limits of current self-evolution approaches and provides a controlled, model-agnostic framework for targeted evaluation.
[AI-139] Stochastic Lifting for Generating Trajectories of Stochastic Physical Systems
链接: https://arxiv.org/abs/2605.29194
作者: Jules Berman,Tobias Blickhan,Benjamin Peherstorfer
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Numerical Analysis (math.NA)
备注:
Abstract:Many stochastic physical systems evolve smoothly over time in the sense that the distribution of states changes regularly across time steps. The transition from current state to the next state can often be modeled as the combination of a smooth map and an explicit source of randomness. Stochastic Lifting exploits this structure by attaching an independent, high-dimensional random label to each state transition in the training data and fitting a transition map from the current state and label to the next state using a standard regression loss. The labels act as auxiliary coordinates that let the model represent multiple plausible next states from similar current states, avoiding collapse to a mean prediction in the finite-sample size regime. At inference, fresh labels are sampled at each time step and the learned map is rolled forward autoregressively, generating diverse trajectories with a single network evaluation per time step.
[AI-140] Influence-Guided Symbolic Regression: Scientific Discovery via LLM -Driven Equation Search with Granular Feedback ICML2026
链接: https://arxiv.org/abs/2605.29184
作者: Evgeny S. Saveliev,Samuel Holt,Nabeel Seedat,David L. Bentley,Jim Weatherall,Mihaela van der Schaar
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: ICML 2026
Abstract:Large Language Models (LLMs) offer a promising avenue for scientific discovery, yet their application to symbolic regression is often constrained by inefficient search strategies and coarse feedback signals. Current methods typically guide LLMs using scalar metrics (e.g., global Mean Squared Error), which fail to identify which components of a proposed equation are driving performance or causing error. We introduce \textitInfluence-Guided Symbolic Regression (IGSR), a method that frames equation discovery as an iterative two-step process combining diverse term generation with rigorous selection: an LLM generates candidate basis functions \psi_j(\mathbfx) for a linear model, which are then evaluated using granular influence scores \Delta_j . These scores quantify each term’s marginal contribution to generalization accuracy, enabling an influence-guided pruning process that systematically refines the model structure. Integrating this mechanism into a Monte Carlo Tree Search (MCTS) enables navigating the combinatorial search space while balancing exploration of novel functional forms with exploitation of high-influence components. We demonstrate IGSR’s effectiveness on a diverse suite of benchmarks, including LLM-SRBench, pharmacological PKPD models, an epidemiological simulation, and real-world genomic data. Notably, we validate the framework’s capacity for genuine discovery in a case study using a high-dimensional biological dataset, in which IGSR identified a novel relationship between DNA methylation and RNA Polymerase II pausing; a hypothesis that was subsequently supported via wet-lab experimentation.
[AI-141] IMEGATE: Sustainable Time-Boxed Promotion Gates for Continual ML Adaptation Under Resource Constraints
链接: https://arxiv.org/abs/2605.29183
作者: Abhijit Chakrabroty,Suddhasvatta Das,Kevin A. Gary,Yash Shah
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:As machine learning(ML) systems evolve to continual adaptation, each re-training cycle uses compute, annotation, and energy. We introduce TIMEGATE, a policy layer managing adaptation by budgeting time, labeling, training, and evaluation. TIMEGATE emits a metric-availability signal M for partial vs. full-evaluation decisions. We validate: (i) labeling outperforms training by 2.3x on Adult tabular; (ii) it transfers to LLaMA-3.1-8B + QLoRA on SST-2 (accuracy 0.80 to 0.96; M =1 in 35/36 runs); (iii) M is informative, 28-cell sensitivity shows M drops to 0.81 at tight thresholds; (iv) 100-cycle simulation achieves 66% evaluation-compute savings with no silent mis-promotions; (v) 10%-slice evaluation on LLaMA uses 89% less wall-clock and energy on a single H200 (ratios agree to 0.2%).
[AI-142] Paper Agents Paper Gains: An Empirical Analysis of DeFi Investment Agents
链接: https://arxiv.org/abs/2605.29174
作者: Jay Yu,Amy Zhao,Danning Sui
机构: 未知
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:DeFi investment agents, systems that use AI for autonomous on-chain trading, have attained over USD 3 billion in combined token valuations since late 2024. We survey over 1,900 AI-tagged crypto projects, filter to investment-focused agents, and curate 10 representative projects spanning strategy and observability dimensions. We then conduct a deep-dive architectural analysis of two prominent agent frameworks, ElizaOS and Virtuals Protocol, and a quantitative on-chain performance analysis of 11 Solana-based agent treasuries with publicly attributable trading activity, covering 925,323 token holders. We find that current deployments remain early and heterogeneous: (1) in our sample, many projects do not yet provide clear evidence of autonomous trade execution, and developer interviews suggest that many visible deployments remain basic API integrations; (2) agent treasuries retain over USD 30M in paper gains while token holders collectively lost USD 191.7M, with the top 1% of wallets capturing 81.4% of all gains (USD 1.81B); (3) token valuations are weakly connected to treasury fundamentals, with market-cap-to-AUM ratios exceeding 10,000x versus below 1x for established DeFi protocols; and (4) aggregate user gains peaked at USD 2.4B before declining to net losses, with median returns negative on every platform and tokens declining 93% on average from all-time highs. We interpret these outcomes as characteristic of a permissionless, first-generation market in which open infrastructure enables rapid experimentation but also allows naive or speculative agents to launch before robust standards for autonomy, performance, and stakeholder alignment emerge. We therefore propose a maturity framework along three dimensions: autonomous execution, risk-adjusted profitability, and stakeholder alignment, to characterize the gap between current deployments and future investment-grade agent systems.
[AI-143] Domain-Informed Representation for Evolutionary Sieving in Integral and Module Lattices
链接: https://arxiv.org/abs/2605.29169
作者: Ahmad Tashfeen,Qi Cheng
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: Published (16 pages) in the proceedings of EvoApplications 2026. You may find the proceedings version here this https URL
Abstract:Traditional cryptography, rooted in problems, e.g., integer factorisation or discrete log, is inevitably vulnerable to a fully operational quantum computer. Although it remains an engineering frontier, the looming threat extends to encrypted data stored today, which could be decrypted in the future with quantum capabilities. To safeguard against this eventuality, the backbone of the modern quantum-safe cryptography is the Shortest Vector Problem (SVP). We enhance Laarhoven’s treatment of Ajtai et al.'s sieving as a genetic algorithm (GA) for the SVP by incorporating domain-informed SVP representation and crossover while naturally extending application to the module lattices.
[AI-144] Better Later Than Sooner: Neuro-Symbolic Knowledge Graph Construction via Ontology-grounded Post-extraction Correction
链接: https://arxiv.org/abs/2605.29168
作者: Lorenzo Loconte,Timothy Hospedales,Cristina Cornelio
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Question answering (QA) is a core challenge in AI, particularly for complex queries requiring multi-hop reasoning across documents, or symbolic operations like aggregation or exhaustive listing. Retrieval-augmented generation has become the dominant approach to QA, with recent graph-based variants addressing part of these issues by organizing knowledge to better support compositional questions. However, most textual graph-based RAG methods still lack the structure needed for symbolic operations useful to answer complex questions reliably. This motivates symbolic graph-based approaches, which extract knowledge graphs (KGs) whose relations are logic predicates that enable SQL-like querying. Yet these pipelines typically use LLMs for KG extraction, which can introduce consistency issues, where extracted facts may violate commonsense ontology constraints. We propose a neuro-symbolic framework for ontology-grounded KG construction combining open-domain extraction, embedding-based canonicalization of types and predicates, and targeted LLM-based correction of ontology violations. By deferring corrections to a post-extraction stage, our method avoids repeated LLM calls, substantially reducing token usage while improving KG consistency and preserving downstream QA quality. Finally, we show that the extracted KGs are well suited for symbolic querying by measuring the occurrence of SPARQL graph patterns.
[AI-145] Evolutionary Refinement of Generative Graph Topologies: A Hybrid WGAN-GA Approach
链接: https://arxiv.org/abs/2605.29161
作者: James Sargant,Seyedeh Ava Razi Razavi,Renata Dividino,Sheridan Houghten
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 6 pages, 4 Figures, 4 Tables, IEEE World Congress on Computational Intelligence
Abstract:Generating realistic graph-structured data is challenging due to discrete connectivity, varying graph sizes, and class-specific structural patterns. Recent Generative Adversarial Networks (GAN)-based graph generation methods improve edge modelling by learning connectivity and matching class-specific density distributions. However these models still exhibit noticeable deviations such as in degree and spectral distribution when compared to real graphs, indicating that important structural properties are not fully preserved. This work aims to reduce these deviations by refining the graphs produced by an existing GAN-based graph generator framework with a Genetic Algorithm (GA). In the GAN framework, the generator produces both node features and connectivity patterns, while a GNN-based critic evaluates graph realism and class consistency to ensure global structural and class alignment. Building on this foundation, we apply a GA to refine the edges of generated graphs. The refinement process guides synthetic graphs toward closer agreement with real data, while preserving diversity and novelty. Experimental results show that the GA refinement consistently lowers combined Maximum Mean Discrepancy (MMD) compared to the base model, leading to graphs that more closely match real structural patterns. This demonstrates that evolutionary refinement is an effective and flexible way to correct residual structural deviations in GAN-based graph generators, improving their suitability for realistic graph synthesis and data augmentation.
[AI-146] CA-AC-MPC: CUDA-Accelerated Actor-Critic Model Predictive Control
链接: https://arxiv.org/abs/2605.29155
作者: Antoonio Buo,Vittorio Cammarota,Michele Avagnale,Pierluigi Arpenti,Vincenzo Lippiello,Fabio Ruggiero
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
备注: Accepted for presentation at the 2026 International Conference on Unmanned Aircraft Systems, ICUAS 2026
Abstract:In the literature, actor-critic model predictive control (AC-MPC) integrates MPC with reinforcement learning to enable high-performance control of complex dynamical systems. However, its differentiable MPC layer requires repeatedly solving an optimization problem in both the forward and backward passes, leading to substantial training and inference latency. This paper tackles this bottleneck introducing a CUDA-accelerated variant that significantly reduces end-to-end execution time while preserving the control performance of the baseline formulation. Simulation results on an agile drone racing task show that our approach achieves state-of-the-art lap times and near-limit dynamic behaviour with markedly reduced training and inference time.
[AI-147] Unveiling Multi-regime Patterns in SciML: Distinct Failure Modes and Regime-specific Optimization ICML2026
链接: https://arxiv.org/abs/2605.29153
作者: Yuxin Wang,Yuanzhe Hu,Xiaokun Zhong,Xiaopeng Wang,Haiquan Lu,Tianyu Pang,Michael W. Mahoney,Yujun Yan,Pu Ren,Yaoqing Yang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computational Physics (physics.comp-ph)
备注: Accepted by ICML 2026
Abstract:Neural networks trained under different hyperparameter settings can fall into distinct training “regimes,” with consistent behavior within regimes and qualitative differences across regimes. In this paper, we study such multi-regime behavior in scientific machine learning (SciML) models through a regime-aware diagnostic framework that jointly analyzes performance, training dynamics, and loss-landscape geometry. We identify three key findings: (i) a consistent three-regime structure emerges across many standard SciML models, different constraint enforcements, and various optimizer designs; (ii) optimization effectiveness is regime-specific, with no single method performing well across all regimes; and (iii) SciML models can exhibit fine-grained failure modes that can challenge conventional interpretations of standard loss-landscape metrics. Our results provide an approach to establish a unified, task-oblivious perspective on failure modes in SciML and to inform regime-aware guidance for improving robustness. We validate these findings across widely-used SciML models, including physics-informed neural networks, neural operators, and neural ordinary differential equations, on benchmarks spanning representative ordinary and partial differential equations.
[AI-148] Multi-Resolution End-to-End Deep Neural Network for Optimizing Latency-Accuracy Tradeoff in Autonomous Driving
链接: https://arxiv.org/abs/2605.29138
作者: Qitao Weng,Heechul Yun
机构: 未知
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Systems and Control (eess.SY)
备注: ICCPS 2026
Abstract:Latency-accuracy tradeoffs are fundamental in real-time applications of deep neural networks (DNNs) for cyber-physical systems. In autonomous driving, in particular, safety depends on both prediction quality and the end-to-end delay from sensing to actuation. We observe that (1) when latency is accounted for, the latency-optimal network configuration varies with scene context and compute availability; and (2) a single fixed-resolution model becomes suboptimal as conditions change. We present a multi-resolution, end-to-end deep neural network for the CARLA urban driving challenge using monocular camera input. Our approach employs a convolutional neural network (CNN) that supports multiple input resolutions through per-resolution batch normalization, enabling runtime selection of an ideal input scale under a latency budget, as well as resolution retargeting, which allows multi-resolution training without access to the original training dataset. We implement and evaluate our multi-resolution end-to-end CNN in CARLA to explore the latency-safety frontier. Results show consistent improvements in per-route safety metrics - lane invasions, red-light infractions, and collisions - relative to fixed-resolution baselines. Comments: ICCPS 2026 Subjects: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Systems and Control (eess.SY) Cite as: arXiv:2605.29138 [cs.RO] (or arXiv:2605.29138v1 [cs.RO] for this version) https://doi.org/10.48550/arXiv.2605.29138 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-149] Governing Technical Debt in Agent ic AI Systems
链接: https://arxiv.org/abs/2605.29129
作者: Muhammad Zia Hydari,Raja Iqbal,Narayan Ramasubbu
机构: 未知
类目: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); General Economics (econ.GN)
备注:
Abstract:Agentic AI systems are increasingly being explored as production infrastructure: they reason over multiple steps, call tools, act through workflows, and adapt through memory and feedback. These systems create governance challenges that are not fully captured by traditional software or predictive ML technical debt. We define Agentic Technical Debt as the accumulated liability created when prompts, memory, tool schemas, orchestration graphs, control policies, and observability routines are patched together faster than they can be validated, standardized, and governed. We define Stochastic Tax as the recurring operating burden of keeping probabilistic agent behavior within acceptable bounds. The distinction matters: debt is a stock of design and governance liability, while the tax is a flow of operating cost that arises because stochastic agents act through tools and workflows. We outline how managers can make both visible through lightweight dashboards and governance controls.
[AI-150] When and How Long? The Readout-Mediator Angle in Temporal Reasoning
链接: https://arxiv.org/abs/2605.29126
作者: Shreyas Fadnavis,Praitayini Kanakaraj,Felix Wyss
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:A linear probe can decode a representation almost perfectly and yet be completely irrelevant to how the model uses it. On calendar-date duration reasoning in language models, a \sin / \cos probe recovers day-of-year from a layer’s activations, yet ablating its direction has no effect on the model’s answers – while ablating a four-dimensional subspace found by Distributed Alignment Search (DAS) at the same layer collapses performance entirely. We measure the angle between these two subspaces – the \emphreadout-mediator angle – and find it indistinguishable from the angle between two random subspaces (the Haar-uniform null), meaning the probe has learned a direction orthogonal to the model’s actual computation. Reverse-engineering the circuit reveals why: attention heads route month-grained context through learned QK offsets at \pm30 and \pm61 days, and MLPs then convert \emphwhen (absolute date) into \emphhow long (duration) – all downstream of the causal subspace the probe never touches. Sparse-autoencoder decomposition confirms the split: probe-aligned and DAS-aligned features encode semantically disjoint concepts with negligible causal overlap. The dissociation replicates across four scales ( 1.5 - 9, B) and two model families, with preliminary evidence on two further domains (spatial displacement, symbolic arithmetic), suggesting that readout-mediator orthogonality is a general failure mode of probe-based interpretability. This directly undermines proposals to deploy probes as runtime safety monitors: the probe can report high confidence on a direction the model has silently abandoned.
[AI-151] PRO-CUA: Process-Reward Optimization for Computer Use Agents
链接: https://arxiv.org/abs/2605.29119
作者: Yifei He,Rui Yang,Hao Bai,Tong Zhang,Han Zhao
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Computer use agents (CUAs) have shown strong potential for automating complex digital workflows, yet their training remains constrained by costly live environment interaction and limited high-quality supervision. Existing filtered behavior cloning pipelines suffer from imitation bottlenecks, including distribution shift from the expert demonstration and the absence of negative learning signals. Meanwhile, standard trajectory-level reinforcement learning struggles with sparse rewards, ambiguous credit assignment, and high infrastructure costs for long-horizon GUI interaction. In this work, we propose PRO-CUA, a process-reward optimization framework for training CUAs with iterative step-level reinforcement learning. PRO-CUA decouples on-policy environment interaction from policy optimization: the current policy collects states through live rollouts, generates diverse candidate actions for each state, receives step-level feedback from a process reward model (PRM), and is optimized with group-relative advantages. This design enables dense and flexible credit assignment without relying on golden answers or offline expert trajectories, while reducing distribution shift by training on the agent’s own execution states. Experiments on live web benchmarks demonstrate the effectiveness of PRO-CUA and the reliability of PRM-guided step-level training.
[AI-152] Beyond Consensus: Trace-Level Synthesis in Mixture of Agents
链接: https://arxiv.org/abs/2605.29116
作者: Shreyas Fadnavis,Praitayini Kanakaraj,Felix Wyss
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:When multiple LLM agents solve the same problem, standard practice compresses each agent’s reasoning into a majority vote or layered synthesis, treating agreement as the finish line. We show this is unnecessarily lossy: an LLM aggregator that reads complete reasoning traces recovers correct solutions even when agents unanimously agree, with beneficial corrections consistently outweighing harmful ones – the \emphaggregation paradox. Majority voting has a ceiling that perturbation diversity does not raise (error correlations are identical); the aggregator’s gain comes from trace-level complementarity, assembling correct intermediate steps from minority chains that voting discards. These findings motivate Self-Consistent Mixture of Agents which generates trace diversity through semantic-preserving input perturbations, safeguards the majority via anchored refinement with provable non-degradation guarantees, and always synthesizes – never gates on consensus. A single model with perturbation-induced trace variation outperforms heterogeneous model pools across structured reasoning, PhD-level science, competition mathematics, and competitive programming. The unit of aggregation should be the reasoning trace, not the answer.
[AI-153] unix-ctf: Procedural Environments for Unix-Competence Reinforcement Learning
链接: https://arxiv.org/abs/2605.29115
作者: Geoffrey Bradway,Roger Creus Castanyer,Lorenz Wolf,Maxwill Lin,Matthew James Sargent,Augustine N. Mavor-Parker
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Unix competence is the ability to use shell and operating-system primitives as first-class tools, not merely to write programs through a terminal. Current terminal benchmarks tend to blur this distinction: a solver fluent in Python but weak in Unix can pass a substantial fraction of Terminal-Bench 2.0, while the reverse skill profile is rarely exercised. We make the distinction operational and build a training surface for the Unix component. unix-ctf is a procedural generator of capture-the-flag tasks for shell agents. Each task hides a short token (a flag of the form flag(a3b1c9…)) inside a fresh Linux container using a single Unix feature, and the agent must recover it. Tasks are produced by an LLM-assisted synthesis pipeline that generates candidate hiding techniques, rewrites them into parameterized hide-and-find script pairs, and filters them with a bidirectional contract: the hide script must leave no plaintext trace of the flag on disk, and the find script must recover the flag in a fresh directory. Because the LLM only writes the planting and recovery steps (the container, layout, and grading harness are fixed), the pipeline lands 656 of 750 raw attempts as portable, reusable variants (87.5%). Our reproduction of Endless Terminals’ full-container-generation approach lands only 17.4% under the same checks. The 656 variants canonicalize to 155 distinct techniques. Fine-tuning Qwen3-8B with LoRA using GRPO on this surface lifts solve rate from 11.6% to 43.6% on a 15-skill multi-family holdout (n=225), redistributes which InterCode-CTF tasks the model solves, and produces a +33 pp gain in Forensics while reaching 32/100 on InterCode-CTF. These results suggest that Unix competence is separable, trainable, and best evaluated directly rather than folded into programming-through-a-shell.
[AI-154] GEO-Bench: Benchmarking Ranking Manipulation in Generative Engine Optimization
链接: https://arxiv.org/abs/2605.29107
作者: Ojas Nimase,Zhe Chen,Gengpei Qi,Yue Zhao,Xiyang Hu
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) increasingly rank products, documents, and recommendations for user queries, which makes manipulating these rankings a growing concern for fairness and information integrity. Research on generative engine optimization (GEO) has produced many manipulation methods, but each is evaluated on its own dataset with its own metrics, so their relative strength and detectability stay unclear. We present GEO-Bench, a benchmark that evaluates GEO ranking-manipulation attacks under one protocol. It unifies black-box prompt-based attacks (TAP, Zero-Shot), white-box gradient-based attacks (STS, RAF, StealthRank), and ten white-hat C-SEO strategies. We score every method on five datasets against a fixed open-weight ranker (Llama-3.1-8B-Instruct), using metrics for both effectiveness (NRG, Success@\alpha, Promote@\alpha) and stealth (keyword violation rate, perplexity ratio). Our evaluation shows that effectiveness and stealth trade off across adversarial attacks, that black-box content rewriting matches or exceeds gradient-based attacks on rank promotion while producing more fluent text and can evade both keyword- and perplexity-based detection on some domains, and that the access model does not predict attack strength. By standardizing datasets, attack implementations, and metrics, GEO-Bench enables the first direct comparison across these attack paradigms and supports the development of detection methods.
[AI-155] rends in AI and Human-AI Interaction in Clinical Trials – A Hybrid Human-AI Exploration
链接: https://arxiv.org/abs/2605.29096
作者: Sandra Woolley,Tim Collins,Khalid Khattak,Illia Chernomorets,Ariane Arevalo,Chris Richardson
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 8 pages plus 2 pages references and appendix
Abstract:This paper examines records retrieved from the this http URL registry to characterize temporal trends in AI terminology and the geographical distribution of AI trials. The work also reports on an exploratory hybrid human-AI approach to analyzing human-AI interaction trends in registered clinical trials. The hybrid workflow comprised a frontier generative AI model (GPT-5.5) and human review to screen and categorize records returned by an AI-focused search. The findings indicate a marked increase in AI-related trials over time, with recent growth in references to machine learning, deep learning, chatbots, GPTs, and large language models. Geographically, China and the United States accounted for the largest numbers of AI-related trials, with notable recent increases in several other countries including Italy, France, Spain, the UK and Turkey (Türkiye). In a random sample of 100 records, human and AI classifiers showed good agreement in identifying studies not substantively using AI, but lower agreement in classifying human-AI interaction, particularly where health professional interaction was ambiguous or insufficiently described. Overall, the results suggest that hybrid human-AI screening of clinical trial records is potentially viable, but clearer trial reporting and more precise interaction definitions will benefit the process.
[AI-156] he Chain Holds the Answer Folds: Trace-Answer Dissociation in Reasoning Models Under Adversarial Pressure
链接: https://arxiv.org/abs/2605.29087
作者: Yubo Li,Ramayya Krishnan,Rema Padman
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Reasoning models are evaluated on single-turn benchmarks but deployed in multi-turn dialogue, where users push back on correct answers. Under sustained adversarial pressure we find a previously undocumented failure mode: the chain-of-thought stays factually correct from first turn to last while the emitted answer flips wrong. We call this unfaithful capitulation (UC) and isolate it with a 2\times 2 latent-versus-behavioral framework that flip-rate metrics and single-turn faithfulness probes both miss. Across three datasets (MT-Consistency, MMLU-Pro, GSM8K), the latent-correct rate at the behavioral flip clusters near 50% in think mode and collapses to 11-15% under no_think – paired, within-model causal evidence that reasoning creates the gap. Across models the effect tracks the reasoning channel (high in Qwen3-32B and GPT-OSS-20B, low in inline-CoT Gemma-4-31B-it). An independent GPT-4o judge corroborates 86% of UC labels; a token-level probe shows the answer-slot argmax is correct in 84% of UC cells; and a naive trace-anchored defense backfires. We release all trajectories, traces, and judge labels.
[AI-157] he Importance of Out-of-Band Metadata for Safe Autonomous Agents : The Redpanda Agent ic Data Plane
链接: https://arxiv.org/abs/2605.29082
作者: Tyler Akidau,Tyler Rockwood,Johannes Brüderl,Marc Millstone
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 6 pages, 1 figure. Published at SAO '26 (co-located with ACM CAIS '26)
Abstract:AI agents are increasingly expected to operate as digital employees: accessing enterprise data, making decisions, and taking actions autonomously. But agents are simultaneously less predictable than humans – prone to hallucination, misinterpretation, and adversarial manipulation – and more technically capable: with deep system knowledge and high-throughput interfaces cascading damage at machine speed. This combination makes it unsafe to rely on agents to faithfully interpret or propagate security-critical metadata such as access policies, data classifications, and behavioral constraints. We present the Redpanda Agentic Data Plane (ADP), an architecture built around out-of-band metadata channels: infrastructure pathways that carry security context, policy signals, and audit trails deterministically, entirely outside the agent’s read and write path and across heterogeneous infrastructure. These channels enforce governance at every stage of the agent lifecycle – scoping data access on the way in, constraining actions during execution, and capturing tamper-proof transcripts on the way out. We demonstrate ADP with a multi-agent portfolio rebalancing system in which autonomous agents monitor markets, make trade decisions, and execute orders across isolated client accounts – with per-client data scoping, trade approval thresholds, and tamper-proof audit trails all enforced by out-of-band channels the agents can neither see nor bypass. Comments: 6 pages, 1 figure. Published at SAO '26 (co-located with ACM CAIS '26) Subjects: Artificial Intelligence (cs.AI) ACMclasses: K.6.5; I.2.11 Cite as: arXiv:2605.29082 [cs.AI] (or arXiv:2605.29082v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2605.29082 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-158] Bridging the Sim-to-Real Gap in Reinforcement Learning-Based Industrial Dispatching through Execution Semantics
链接: https://arxiv.org/abs/2605.29078
作者: Jonathan Hoss,Noah Klarmann
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted for publication at the 24th IEEE International Conference on Industrial Informatics (INDIN 2026), held from 26 to 29 July 2026 in Melbourne, Australia
Abstract:Event-driven scheduling policies are increasingly deployed in industrial environments, where decisions are made under asynchronous and partially observed system states. As a result, decision states are not temporally consistent, action admissibility is not explicitly defined, and the origin of execution errors remains ambiguous. These issues limit both reliability and interpretability. To address this gap, a policy-neutral execution and measurement layer is proposed to mediate between scheduling policies and the industrial execution environment. The layer constructs decision-valid snapshots from asynchronous event streams, defines a standardized execution contract with explicit action admissibility, and records outcomes as divergences between policy intent, transactional outcomes, physical execution, and human intervention. This enables a separation between decision semantics and execution behavior and makes deployment mismatch observable and structurally attributable. The proposed framework is evaluated using a discrete-event simulation. The results show analytical benefits across all observation lag regimes, as undifferentiated execution failures are transformed into structured, typed outcomes with full attribution coverage. Operational benefits are strongest under low observation lag, where avoidable execution errors can be prevented before commitment. Overall, the layer turns execution uncertainty into supervisory data for evaluation and policy refinement. Comments: Accepted for publication at the 24th IEEE International Conference on Industrial Informatics (INDIN 2026), held from 26 to 29 July 2026 in Melbourne, Australia Subjects: Artificial Intelligence (cs.AI); Machine Learning (cs.LG) Cite as: arXiv:2605.29078 [cs.AI] (or arXiv:2605.29078v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2605.29078 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-159] SCDBench: A Benchmark for LLM -Based Smart Contract Decompilers
链接: https://arxiv.org/abs/2605.29059
作者: Kaihua Qin,Dawn Song,Arthur Gervais
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
备注:
Abstract:Smart contract decompilation aims to recover high-level source code from bytecode, but evaluating decompilers remains difficult because existing studies use narrow datasets, inconsistent metrics, and limited semantic consistency checks. This gap is increasingly important as large language models (LLMs) begin to generate source-like Solidity that may compile and appear plausible, even when its semantics diverge from the original contract. We introduce SCDBench, a dataset and benchmark methodology for LLM-based smart contract decompilation. The dataset contains 600 real-world Solidity contracts with paired bytecode inputs, ground-truth source code, and replayable semantic checkpoints. SCDBench evaluates decompiler outputs through four cumulative stages: format completeness, compilability, Application Binary Interface (ABI) recovery, and semantic consistency via differential replay. We evaluate Claude Opus 4.7, GPT-5.3-Codex, and GLM-5 in a zero-shot decompilation setting, including GLM-5 variants with and without extended reasoning and a zero-shot compilation-repair setting. The results show that frontier LLMs can often produce structured and compilable Solidity, but achieving semantic consistency remains far from solved: the best-performing frontier model perfectly decompiles only 42/600 contracts. We further show that introducing same-model compilation repair substantially improves performance at modest additional cost. SCDBench establishes a common ground for rigorous, reproducible evaluation and aims to accelerate the development of reliable smart contract decompilers for blockchain security and transparency.
[AI-160] Differentiable Belief-based Opponent Shaping
链接: https://arxiv.org/abs/2605.29042
作者: Aarav G Sane,Karthik Sivachandran,Rohan Paleja
机构: 未知
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Human coordination often relies on the ability to influence the beliefs of others through strategic action. In multi-agent reinforcement learning, opponent shaping attempts to replicate this influence, though existing methods typically operate within an opponent’s parameter, policy, or value space. Meanwhile, belief-manipulation techniques in hidden-role games often rely on hard-coded objectives, such as deception or belief saturation. We propose Differentiable Belief-based Opponent Shaping (D-BOS), a first-order method that treats each observer’s belief as the shaped opponent state and differentiates through k -step softmax-Bayes belief dynamics. Rather than explicitly rewarding deceptive or cooperative behavior, our method treats the belief state as the target for shaping. This allows the optimal strategy to emerge naturally from the environment’s reward structure. This belief-space formulation provides an opponent-shaping signal by differentiating through opponent belief updates, and naturally extends to multiple observers by aggregating gradients over their individual inferred belief trajectories. Empirically, D-BOS outperforms PPO and BBM in hidden-role games, with the largest gains in mixed-motive settings.
[AI-161] Practitioner Beliefs and Behaviors in AI-Enhanced Education: DOT Framework Survey Evidence
链接: https://arxiv.org/abs/2605.29041
作者: David Gibson(1),M. Elizabeth Azukas(2),Gerald Knezek(3) ((1) Curtin University, (2) Georgia Institute of Technology, (3) University of North Texas)
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:This study reports findings from a cross-sectional survey (n = 72) of higher education practitioners examining beliefs, behaviors, and institutional conditions related to artificial intelligence (AI) integration in teaching and learning. Grounded in the DOT Framework, which integrates design thinking and open systems theory, the study investigates AI familiarity, usage patterns, design-oriented practices, and pedagogical beliefs. Exploratory factor analysis of 19 belief items identified a three-factor structure: AI Functional Capabilities, Oversight and Governance, and Instructor Collaboration and Planning (\alpha = .90). Results indicate that practitioners hold favorable views of AI as a pedagogical support while maintaining strong commitments to human oversight and critical evaluation. Reported practices emphasize iterative prompting and content generation, with less consistent use of needs assessment and feedback loops. Institutional barriers including limited policy, training, and infrastructure were widely reported. These findings provide preliminary empirical support for the DOT Framework as a descriptive model of practitioner beliefs and practices, while also highlighting gaps between design-oriented theory and current implementation. The study contributes an initial measurement structure and identifies directions for confirmatory validation and outcome-based research linking AI-supported design practices to instructional quality.
[AI-162] Return-to-Go Is More Than a Number: Q-Guided Alignment for Return-Conditioned Supervised Learning ICML2026
链接: https://arxiv.org/abs/2605.29028
作者: Yuxiao Yang,Weitong Zhang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 28 pages, 13 figures, 20 tables, accepted by ICML 2026
Abstract:Conditioned Sequence Models (CSMs) learn policies by treating return-to-go (RTG) as a control signal. However, existing CSMs often treat the RTGs as simple numerical inputs rather than aligning them with the performance of their policies. In this paper, we propose Q-ALIGN DT, a framework that enforces this alignment by ensuring the Q -value of the output policy is consistent with the input RTG. By leveraging a Q function to provide dense guidance to CSMs and further fine-tuning it using an RTG-perturbation technique with the CSM, our method ensures that higher RTGs are consistently mapped to trajectories with higher expected returns. Theoretically, we show that Q-ALIGN DT can efficiently learn the desired policy and output a near-optimal one when the RTG is sufficiently high. Empirically, we demonstrate through extensive experiments that Q-ALIGN DT achieves superior controllability and performance across the D4RL benchmark. Remarkably, our model effectively learns a structured family of policies that maintains precise alignment and generalizes to tasks like velocity-tracking where prior methods fail.
[AI-163] Label-Free Reinforcement Learning via Cross-Model Entropy
链接: https://arxiv.org/abs/2605.29009
作者: Matt Gorbett,Hossein Shirazi
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Post-training large language models with reinforcement learning is bottlenecked by the reward signal. Existing approaches require either ground-truth verifiable rewards, restricting training to domains with automatic correctness checks (e.g., mathematics, code execution), or human preference labels, which are expensive to collect and prone to reward hacking. Recent label-free methods replace ground-truth verifiers with self-referential signals like majority voting or token entropy over a model’s own outputs, but risk reinforcing a model’s own errors. In this work we propose Cross-Model Entropy (CME), the mean log-likelihood of a generator’s response under a separate verifier model, as a label-free reward signal for RL post-training. CME is continuous, training-free, and grounded in the principle that responses a verifier finds unsurprising are likely correct or high quality. Because the verifier is independent of the generator, the signal cannot be gamed through self-consistency. We integrate CME into GRPO with no other changes to the training loop, extending label-free RL to open-ended instruction following – a regime where self-referential signals are inapplicable or poorly suited. On open-ended instruction following (UltraFeedback prompts, evaluated on AlpacaEval 2.0), CME rewards beat the untrained base in head-to-head LLM-as-Judge comparisons across four model families (Qwen, Llama, Gemma, OLMo) and three training regimes (pretrained, SFT, and instruction-tuned), with tie-adjusted win rates ranging from 52.5% to 71.4%. Code will be released upon publication.
[AI-164] LoRe: Adaptive Interaction-Evaluation Routing with Per-Step Interaction Budgets for Iterative Graph Solvers ICML2026
链接: https://arxiv.org/abs/2605.29005
作者: Jintao Li,Yong-Yi Wang,Zheng-An Wang,Heng Fan
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at ICML 2026
Abstract:Diffusion-based neural solvers for combinatorial optimization repeatedly re-evaluate dense edge/factor interactions, making inference expensive in wall-clock time and often memory-bound at scale. Inspired by the computational methodologies of many-body physics, we introduce LoRe, a training-free, inference-time drop-in wrapper that enforces per-step interaction-evaluation budgeting: at each iteration, it evaluates only a fixed fraction of interactions by dynamically routing computation to high-conflict or high-uncertainty interactions, instead of using a fixed sparsification (e.g., static kNN graphs or static masks). Under fully inclusive end-to-end wall-clock accounting, LoRe substantially improves scalability on the Maximum Independent Set (MIS) problem, extending feasible inference more than 3\times beyond the baseline’s out-of-memory limit, delivering a \sim 8\times speedup and a \sim 12\times peak-memory reduction, with solution quality preserved in this regime. Demonstrating cross-task generality on the large-scale Traveling Salesperson Problem (TSP) and zero-shot robustness to topology shifts, LoRe achieves a \sim 15\times speedup at n=1000 with a 44\times memory reduction and competitive tour quality.
[AI-165] FormInv: A Measurement Protocol for Semantic Invariance in Mathematical Reasoning Benchmarks ICML2026
链接: https://arxiv.org/abs/2605.29001
作者: Nishal Thomas,Noel Thomas
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 18 pages, 3 figures. Under review for the 3rd AI for Math Workshop (AI4Math), ICML 2026
Abstract:A paraphrase-quality audit of MathCheck (ICLR 2025) detected 4 semantically incorrect paraphrases in 129 groups (3.1%); removing them drops GPT-4o from rank 2 to rank 4 and elevates Claude Haiku and DeepSeek V3 above it; these ranking changes are invisible to any single-model evaluation. Cross-model unanimity found these errors automatically (= 3/4 models for MathCheck; = 6/9 for our primary evaluation) for under 10; in our own dataset the same protocol found that 47% of auto-generated connective-variation paraphrases were semantically incorrect. That flaw compounds a deeper measurement gap: Claude Haiku 4.5 achieves 86% accuracy yet SCR=50%, meaning half its theorems are answered differently under semantically equivalent restatements, while aggregate accuracy across 9 models spans only 86-96% yet Semantic Consistency Rates (SCR) span 50-82% – a 32-point gap invisible to standard benchmarks. Formally, for any target ranking over 9 frontier models there exists a weighting over paraphrase families that realizes it (No-Free-Benchmark corollary), because no model Pareto-dominates all families – so benchmark designers who select families are implicitly choosing which model wins. FormInv supplies the audit protocol (replicated on external benchmarks at 100% recall), SCR and per-theorem Cochran’s Q as primary invariance measures evaluated on 9 models across 366-811 items (on Lean4-verified theorems), and FormInvSelector for regime-aware model selection.
[AI-166] BEAMS: Benchmarking and Evaluating AI for Modeling and Simulation
链接: https://arxiv.org/abs/2605.28994
作者: Sara Metcalf,William Schoenberg
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:AI tools to support real world decision making must be able to build simulation models that inform their recommendations and render them interpretable. Tools that can automate aspects of modeling practice must complement human expertise, not replace it. The BEAMS Initiative aims to guide the development of AI tools for modeling and simulation toward forms that are responsible and ethical by establishing benchmarks for human centered modeling and simulation practices. The initiative uses open digital and organizational infrastructure to collaboratively evaluate AI tools for modeling and simulation. The open source sd ai project hosted by the initiative establishes transparency and enables contributions to be shared broadly. A steering group focuses on prioritizing potential benchmarks, while a technical group focuses on implementing the benchmarks in the form of automated tests. Tests for several distinct categories of evaluation have been implemented and applied to AI tools that support qualitative model building, quantitative model building, and model discussion. These include tests for causal translation, model iteration, causal reasoning, conformance, model behavior explanation, suggested model building steps, and suggested model fixes. When engines from the sd ai project are coupled with different LLMs, their performance on these evaluations reveals variability across different AI tools. The evaluations implemented by the initiative demonstrate that AI enabled modeling tools perform better at discussion and basic qualitative tasks than with causal reasoning and quantitative error fixing. No single LLM dominates across engine types, highlighting the importance of specific tasks and tradeoffs between speed and accuracy. Ongoing efforts of the initiative aim to incorporate benchmarks that address concerns about bias by considering alternative perspectives and human centered use cases.
[AI-167] he Hamilton-Jacobi Theory of Deep Learning
链接: https://arxiv.org/abs/2605.28983
作者: Jose Marie Antonio Miñoza,Erika Fille T. Legara,Christopher P. Monterola
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Dynamical Systems (math.DS); Representation Theory (math.RT); Computational Physics (physics.comp-ph)
备注:
Abstract:In this paper, training a neural network is identified, exactly, as a search through Hamilton–Jacobi initial-value problems: each gradient step selects the initial data of a viscous Hamilton–Jacobi equation whose Hopf–Cole propagator best fits the observations; at inference, the input is the spatial point at which that solution is evaluated and the initial condition is already encoded in the weights. The correspondence is exact for log-sum-exp layers and structural for broader architectures: residual networks, transformers, and recurrent architectures (RNNs, LSTMs, SSMs) each discretize the same class of Hamilton–Jacobi equations, with architecture-dependent Hamiltonian and viscosity. A single deformation parameter \varepsilon unifies all four perspectives (network, tropical algebra, viscous PDE, convex optimization) in a commutative diagram closed under Lipschitz conditions. Quantitative consequences include: the minimax optimal generalization rate O(n^-1/(d+2)) for fixed t ; adversarial robustness controlled by \varepsilon ; backpropagation as the co-state equation of the Hamiltonian system for residual networks (Pontryagin Maximum Principle); scaling exponents consistent with data intrinsic dimension via PDE quadrature; and a closed-form O(N) influence function (softmax attribution weights \pi_j ) whose entropy landscape undergoes fold bifurcations as \varepsilon increases, each merging attribution basins.
[AI-168] VFEAgent : A Multimodal Agent Framework for End-to-End Automated Finite Element Analysis
链接: https://arxiv.org/abs/2605.28978
作者: Jiachen Zhang(1 and 2),Junyi Lao(1),Chenghao Liu(1),Siyuan Liu(1),Shixin Wu(1),Linsen Zhang(1),Boyu Wang(1),Songfang Huang(1) ((1) Peking University, (2) China Agricultural University)
机构: 未知
类目: Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE)
备注: 9 pages, 3 figures, 2 tables. Equal contribution: Jiachen Zhang and Junyi Lao. Corresponding author: Songfang Huang. Preprint
Abstract:Finite Element Analysis (FEA) serves as the cornerstone of modern engineering design. However, its workflow is inherently complex and relies heavily on domain expertise. Although recent efforts have integrated Large Language Models (LLMs) into FEA, existing approaches face limitations in handling multimodal inputs and executing complex tasks. To address these limitations, we propose VFEAgent, an end-to-end multi-agent system designed to automate FEA modeling and simulation directly from input images and problem descriptions. Our methodology integrates two core components: (1) a multimodal vision-language multi-agent pipeline that employs ReAct-driven reasoning to extract structured FEA specifications from heterogeneous inputs and (2) a verification-first code synthesis framework, incorporating robust self-debugging and fallback mechanisms to ensure executability and physical validity. We systematically evaluated the system across various engineering mechanics scenarios. The results demonstrate that VFEAgent achieves a high success rate in generating complete and physically valid simulations, outperforming LLM-based baseline methods in reliability and correctness. These findings validate the feasibility of automating the complete FEA workflow, highlighting the framework’s potential to liberate engineers from tedious manual analysis.
[AI-169] Comparing Post-Hoc Explainable AI Methods for Interpreting Black-Box EEG Models in Depression Detection
链接: https://arxiv.org/abs/2605.28977
作者: Antonia Šarčević,Nikolina Frid
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances in deep learning have enabled increasingly accurate electroencephalography (EEG)-based classification of Major Depressive Disorder (MDD), but the decision-making processes of high-capacity models remain difficult to interpret. This study investigates multiple post-hoc explainability methods applied to an InceptionTime architecture trained for EEG-based MDD detection. The analysis includes Shapley-based, gradient-based, and perturbation-based attribution approaches: DeepSHAP, Integrated Gradients, GradCAM, Occlusion, and Permutation Feature Importance. Explainability analysis was performed within a subject-level stratified 5-fold cross-validation framework using global attribution aggregation across EEG segments and subjects. The evaluated methods revealed partially convergent attribution patterns, with recurring emphasis on frontal, temporal, and posterior EEG regions, particularly in the right hemisphere. Quantitative comparison demonstrated substantial agreement between gradient- and perturbation-based approaches, while DeepSHAP produced comparatively distinct attribution distributions. At the same time, variability between explainability methods highlighted the influence of methodological assumptions on the resulting explanations. Overall, the results suggest that different post-hoc explainability approaches capture partially overlapping relevance structures in EEG-based deep learning models for depression detection. Although the observed attribution patterns are broadly consistent with several previous EEG studies of MDD, the analysis should be interpreted as exploratory rather than evidence of definitive neurophysiological biomarkers or clinical applicability. The study highlights both the usefulness and limitations of post-hoc explainability for interpreting black-box EEG classifiers in psychiatric applications.
[AI-170] Frontier LLM -based agents can overcome the ontology curation bottleneck for natural phenotypes
链接: https://arxiv.org/abs/2605.28965
作者: James P. Balhoff,Hilmar Lapp
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: 7 pages, 2 figures
Abstract:Linking free-text phenotype descriptions to ontology terms, typically referred to as phenotype annotation, is essential for the cross-study integration of comparative morphological data. This labor intensive process has heavily relied on highly trained human experts, which makes it challenging to scale and thus a key bottleneck. Dahdul et al. (2018) established a Gold Standard (GS) of Entity-Quality (EQ) annotations across seven phylogenetic studies and used it to evaluate three human curators and the Semantic CharaParser NLP tool with ontology-based semantic similarity metrics; they reported that machine-human consistency was significantly lower than inter-curator (human-human) consistency. Here we revisit that benchmark with five frontier hosted LLMs from Anthropic and OpenAI, each operating as an “agentic curator” within a self-contained workspace that supplies the source publication PDF, the same annotation guide used by the original human curators, the four project ontologies (UBERON, PATO, BSPO, GO), and a validation script. Evaluated against the same Gold Standard, every agent fell within the range of inter-curator variability of the three trained human biocurators of the original study; the best performing agents approached but did not reach the best performing human curator. Agents substantially outperformed Semantic CharaParser on all four metrics.
[AI-171] Conf-Gen: Conformal Uncertainty Quantification for Generative Models ICML2026
链接: https://arxiv.org/abs/2605.28920
作者: Gabriel Loaiza-Ganem,Kevin Zhang,Wei Cui,Marc T. Law,Kin Kwan Leung
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注: ICML 2026
Abstract:Conformal prediction (CP) and its extension, conformal risk control (CRC), are established frameworks for quantifying uncertainty in supervised machine learning through formal guarantees. However, recent breakthroughs in artificial intelligence (AI) have been driven by unsupervised generative models, such as large language models (LLMs) and image generators, which are not directly compatible with CP or CRC. In this work we introduce conformal generation (Conf-Gen), a general framework adapting CRC to generative tasks while relaxing its theoretical assumptions. Conf-Gen unifies and generalizes previous attempts to apply CP to LLMs, and extends conformal methodology to entirely new domains. We demonstrate the flexibility of Conf-Gen through some novel applications, including obtaining conformal guarantees on: image generators producing non-memorized images, conversational AI systems having asked enough clarifying questions, and the output of AI agents being correct.
[AI-172] AIRGuard: Guarding Agent Actions with Runtime Authority Control
链接: https://arxiv.org/abs/2605.28914
作者: Suliu Qin,Haomin Zhuang,Yujun Zhou,Yufei Han,Xiangliang Zhang
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Tool-using language agents turn model decisions into external side effects: they read files, run scripts, call APIs, send messages, and invoke Model Context Protocol tools. This makes agent attacks different from jailbreaks. The harmful step is often not an obviously forbidden output, but an ordinary executable action that becomes unsafe because attacker-controlled context steers authorized access against the user’s interest. We identify this failure mode as authority confusion: untrusted resources may inform reasoning, but they must not authorize side effects. We present AIRGuard, a runtime guard that operationalizes least privilege as action-time authorization. AIRGuard normalizes heterogeneous tool calls, derives task authority into step-level authority, tracks source and target trust, simulates sensitive side effects, audits cross-step risk, and enforces decisions before actions execute. On AgentTrap, AIRGuard reduces Sonnet 4.6 attack success from 36.3% without defense to 5.5%. On DTAP-150, AIRGuard preserves 76.0% benign utility with Haiku 4.5, compared with 52.0% for ARGUS and 42.0% for MELON. An ablation further shows that prompt-only policy helps only modestly, whereas a dedicated runtime authority-control layer gives the agent system direct control over tool-mediated side effects. Code and data are available at this https URL.
[AI-173] Orthogonal Concept Erasure for Diffusion Models ICML2026
链接: https://arxiv.org/abs/2605.28902
作者: Yuhao Sun,Lingyun Yu,Haoxiang Xu,Fengyuan Miao,Zhuoer Xu,Hongtao Xie
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注: Accepted by ICML 2026 Oral
Abstract:Concept erasure has emerged as a promising approach to mitigate undesired or unsafe content in diffusion models, yet existing methods still face significant limitations. While training-based methods are effective, their high computational cost limits scalability. Editing-based methods are more efficient and deployment-friendly, yet they struggle to simultaneously achieve precise concept erasure and preserve overall generative capacity. We identify this core limitation of the editing-based methods as reliance on additive parameter updates. Our empirical analysis reveals that concept semantics primarily depend on neuron direction rather than neuron magnitude, while overall generative capacity relies on the angular geometry of neurons. As additive updates inherently entangle direction, magnitude, and angular geometry, they inevitably introduce unintended interference between concept erasure and overall generation performance. To address this, we propose Orthogonal Concept Erasure (OCE), which reformulates editing-based erasure as multiplicative parameter updates from a geometric perspective. Specifically, OCE applies layer-wise orthogonal transformations derived from a closed-form solution to the parameters, enabling precise concept erasure while preserving the neuron magnitude and angular geometry. Furthermore, to address conflicting constraints in multi-concept erasure, OCE introduces a subspace-level objective with structured subspace manipulation, yielding a more effective and scalable erasure. Extensive experiments on single- and multi-concept erasure demonstrate that OCE outperforms existing methods in concept erasure and non-target preservation, erasing up to 100 concepts in 4.3 s. Code: this https URL.
[AI-174] Quantum-Enhanced Adversarial Robustness in Artificial Intelligence
链接: https://arxiv.org/abs/2605.28899
作者: Jaydip Sen
机构: 未知
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: This is the pre-print of the chapter which has been accepted for publication in the edited volume titled “Quantum Enhancements to the AI Industry”, edited by Eduard Babulak. The volume will be published by IGI Global, USA. This is not the final version of the chapter published in the book
Abstract:Artificial Intelligence has achieved remarkable success across diverse application domains. However, its vulnerability to adversarial attacks poses significant challenges to reliability, security, and trustworthiness. Adversarial machine learning demonstrates that even highly accurate models can be manipulated through carefully crafted perturbations, raising serious concerns in safety critical systems such as healthcare, finance, and autonomous technologies. In parallel, quantum computing has emerged as a transformative paradigm capable of addressing complex computational problems through principles such as superposition, entanglement, and quantum interference. The convergence of these fields has led to the emergence of quantum artificial intelligence, which explores how quantum techniques can enhance learning efficiency, scalability, and robustness. This chapter provides a comprehensive overview of adversarial machine learning and existing defense strategies, followed by an accessible introduction to quantum computing and quantum machine learning models. It further presents conceptual frameworks for quantum-enhanced adversarial robustness, emphasizing quantum optimization, feature mapping, and hybrid quantum classical architectures. Practical applications, key challenges, and future research directions are also discussed to support the development of secure and trustworthy AI systems.
[AI-175] Context Distillation as Latent Memory Management
链接: https://arxiv.org/abs/2605.28889
作者: Ziyang Zheng,Zeju Li,Xiangyu Wen,Jianyuan Zhong,Junhua Huang,Lei Chen,Mingxuan Yuan,Qiang Xu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Context distillation compresses contextual information into model parameters, yet existing methods often ignore how multiple distilled latent memories should be stored, retrieved, and safely activated in non-oracle settings. We formulate context distillation as a latent memory management problem. We distill each context into an independent LoRA adapter, forming a modular memory bank that enables explicit memory selection. Given a query, our framework retrieves candidate memories, routes the query to the most suitable adapter, and uses a Self-Gating mechanism to decide whether latent memory should be activated. To improve efficiency, we further introduce cache sharing to reduce management overhead during inference. Experiments show that our method substantially outperforms baselines with retrieval, while Self-Gating improves robustness by deactivate unnecessary latent memories.
[AI-176] Ultra-Reduced-Impact-Encased-Logging (URIEL): propose a new method for selective sustainable logging and post-harvest silvicultural treatment in tropical forest using airborne robotics systems
链接: https://arxiv.org/abs/2605.28883
作者: Daniel Albiero,Gelton Fernando de Morais,Daniela Han,Flávio Roberto de Freitas Gonçalves,Artur Vitório Andrade Santos,Wesllen Lins de Araújo,Alessandra Maia Freire,Cláudio Kiyoshi Umezu,Mateus Peressin,Francesco Toscano,Admilson Írio Ribeiro,Alfeu J. Sguarezi Filho,Américo Ferraz Dias Neto,Angel Pontin Garcia
机构: 未知
类目: Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注: 196 pages, 40 figures, A revolutionary technology to help protect tropical forests. It was developed, scaled, detailed, calculated, and simulated in an advanced computational environment, com viabilidade econômica e social. “E pur si muove”
Abstract:Tropical forests worldwide are under intense deforestation pressure driven by economic and political interests, and scientific evidence suggests this deforestation contributes to climate change. This paper proposes a novel logging method for tropical forests, Ultra-Reduced-Impact-Encased-Logging (URIEL). This new method is based on heli-logging techniques combined with intensive use of robotics and AI integrated with post-harvest silvicultural treatments performed by drones. The concept of appropriate equipment for this method was developed, dimensions were determined, details were completed in a digital proof of concept, and an effective digital simulation and economic feasibility analysis were carried out for various helicopter-timber-distance combinations. The results demonstrated that a URIEL method has high economic viability and makes it possible to virtually eliminate collateral damage to forests while maintaining ecosystem services. The main conclusion of this paper is that, despite the satisfactory scientific and technological results, the feasibility of a Uriel method depends on the integration of stakeholders intrinsic to the context: high-tech industry; political governments; certified logging companies; and native populations.
[AI-177] LogDx-CI: Benchmarking Log Reduction Tools for LLM Root-Cause Diagnosis
链接: https://arxiv.org/abs/2605.28876
作者: Bowen Qin
机构: 未知
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:CI failure logs are large (median 5k lines, max 200k in this corpus) and noisy. Coding agents that try to debug them depend on an upstream tool to reduce the log to a manageable context, but the field has had no public empirical comparison of which reductions preserve enough evidence for downstream LLM diagnosis. We introduce LogDx-CI, a benchmark that compares 11 context-reduction tools (raw, tail, grep, three RTK modes, two real LLM map-reduce summarizers, three hybrid routers) on 35 real GitHub Actions failure cases, scored by 3 LLM debugger families (Claude Haiku 4.5, Claude Sonnet 4.6, OpenAI gpt-5-mini) plus a Sonnet 4.6 tool-using agent. We report three load-bearing findings. (1)~Hybrid grep+tail routers dominate the cost-quality Pareto frontier; the top two methods score 0.670 / 0.666 at \sim \ 0.03 per case, same-ballpark quality as standalone grep at 4.5\times fewer tokens. (2)~In the agent-loop regime, the quality range across reduction tools collapses 7\times (single-shot spread 0.42 \to agent-loop spread 0.059); the agent rescues weak contexts via follow-up tool calls. However, cost differences persist: weak contexts force the agent to issue 2–4 \times more tool calls to recover. (3)~A cross-family LLM-summary pair (gpt-5-mini summarizer feeding a Claude Haiku debugger) beats the same-family pair by +0.071 averaged across four diagnoser variants, falsifying the self-call-bias hypothesis on this task. The gpt-5-mini summarizer is also the agent-loop #1 method (score 0.749) at 0.37 tool-calls per case and 10\times lower reducer cost than the Haiku summarizer (\ 0.18 vs \ 1.75 per case). All data, code, per-case bundles, and reproducibility infrastructure are public.
[AI-178] Representation Alignment Rests on Linear Structure
链接: https://arxiv.org/abs/2605.28870
作者: Kiril Bangachev,Guy Bresler,Yury Polyanskiy
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:We investigate the Platonic Representation Hypothesis (PRH) through a tripartite statistical framework of representations: signal, bias, and noise. 1) Signal: We propose that Platonic alignment arises from the universal relationship between objects and attributes, which is encoded linearly in representations according to the Linear Representation Hypothesis (LRH). We provide evidence that LRH helps explain PRH by extracting linear object-attribute features with sparse autoencoders and showing that these sparse representations often exhibit stronger cross-modal alignment than their dense counterparts. 2) Bias: Models have different implicit biases due to the diverse architectures and training procedures used. We show that this difference can be partially mitigated. Centering and normalization consistently improve cross-model alignment. 3) Noise: Finite-sample training leads to noise in representations. We provide evidence that representational noise is driven by data scarcity by revealing a strong and consistent positive correlation between word frequency and alignment in LLMs and text embedding models. Synthesizing signal, bias, and noise, we propose a statistical model that refines the Linear Representation Hypothesis and explains further phenomena related to the alignment of representations emerging from diverse modern AI architectures.
[AI-179] Balancing Multimodal Learning through Label Space Reshaping
链接: https://arxiv.org/abs/2605.28869
作者: Xiaoyu Ma,Weijie Zhang,Yuanhao Gao,Han Miao,Yongjian Deng,Hao Chen
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: In process
Abstract:Multimodal learning often suffers from modality imbalance, where modalities that converge faster dominate optimization while others remain undertrained. Existing approaches typically mitigate this issue by strengthening the weak modality or adjusting optimization gradients. However, such strategies mainly compensate for optimization rate discrepancies, often at the expense of the strong modality’s optimization capacity, without analyzing how these discrepancies arise at the modality level. Based on theoretical insights and empirical observations, we argue that the discrepancy of learning pace arises from differences in the mapping difficulty between modality-specific feature space and the shared label space. To address this issue, we propose Balanced Multimodal Label Reshaping (BMLR), the first method that promotes multimodal balance from the label-side design. BMLR reshapes the cross-modal label space to equalize mapping difficulty across modalities, thereby facilitating modality interaction and injecting richer inter-class information into each modality. Extensive experiments across multiple architectures demonstrate that BMLR consistently improves multimodal performance and exhibits strong compatibility with diverse model designs. The source code will be released soon.
[AI-180] axDistill: Improving Metagenomic Taxonomic Annotation via Distilled Genomic Foundation Models
链接: https://arxiv.org/abs/2605.28868
作者: Rongye Ye,Lun Li,Zheng Luo,Yiran Zhan,Shuhui Song
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: The manuscript contains 14 pages, 7 figures, and 3 tables
Abstract:Metagenomic taxonomic annotation aims to identify the microbial origins of DNA fragments in environmental samples. Traditional methods that rely on sequence similarity are often constrained by the high microbial diversity and the incompleteness of reference databases, which has motivated the development of learning approaches such as Taxometer that perform post hoc correction to learn more informative metagenomic sequence representations. However, these methods typically rely on labels derived from similarity search tools during training, which inevitably introduces noise that can impair representation learning and degrade classification performance. To address this issue, we propose TaxDistill, a knowledge distillation framework for metagenomic classification. We introduce GenomeOcean, a 500M parameter genomic foundation model, as the teacher network to extract deep semantic features and generate soft labels based on confidence. By distilling this soft label information into a lightweight student network, TaxDistill effectively reduces the label noise introduced by initial retrieval tools. Comprehensive experiments on seven diverse CAMI2 datasets demonstrate that TaxDistill outperforms existing baselines in most scenarios. For instance, on the Gastrointestinal dataset, it improves the F1 score of MMseqs2 from 0.763 to 0.941, outperforming the Taxometer baseline. Overall, TaxDistill provides a reliable method for label correction in complex metagenomic analysis.
[AI-181] PrismFlow: Residual Dynamics for Flow Matching in Time-Series Generation
链接: https://arxiv.org/abs/2605.28867
作者: Junru Zhang,Lang Feng,Jinbo Wang,Xu Guo,Yucheng Wang,Han Yu,Min Wu,Yabo Dong,Duanqing Xu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Generating high-quality time-series data is challenging because real-world signals often exhibit multimodal patterns and multiscale dynamics, including oscillations and high-frequency variations. Flow Matching (FM) offers an efficient alternative to diffusion models, but practical implementations typically rely on a single finite-capacity global vector-field estimator. In such heterogeneous temporal distributions, distinct regimes may pass through nearby flow states while requiring incompatible conditional velocities. A monolithic estimator trained with the standard \ell_2 velocity-matching objective may therefore learn an overly smoothed approximation of the local transport field. This estimator-level smoothing can attenuate branch-specific dynamics, leading to spectral distortion and poor mode coverage. To address this, we propose PrismFlow, a new FM method with Koopman-inspired dynamical experts. Each expert learns residual corrections in a latent space where local nonlinear temporal evolution can be approximated by linear transitions. We further propose a confidence-aware Winner-Take-All (WTA) objective that updates only the expert best aligned with each sample while masking gradients to the others, encouraging mode-specific specialization. During sampling, the selected expert adds a residual dynamical correction to the global transport field, preserving FM stability while recovering fine-grained and high-frequency temporal structures. Across various benchmarks, PrismFlow effectively mitigates the spectral contraction in standard FM and achieves state-of-the-art performance, with a 15.6% gain in Context-FID and a 38.6% improvement in Discriminative Score, while remaining robust in low-data settings and effective for forecasting and imputation.
[AI-182] Continuity and Ordinality Matter: Constraining Time Series Tokens for Effective Time Series Analysis with Large Language Models
链接: https://arxiv.org/abs/2605.28866
作者: Musheng Li,Ziying Zhang,Cheng jin,Yuantao Gu
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Token-based time series large language models (TS-LLMs) have emerged as a promising direction for time series analysis and reasoning. However, prior studies largely overlook the inherent continuity and ordinality of time series tokens, which substantially limits model performance. In this paper, we argue that preserving these properties in time series token embeddings is crucial for the effectiveness of token-based TS-LLMs. To this end, we propose COM (Continuity and Ordinality Matter), a continuity- and ordinality-aware strategy that integrates geometric constraints into both the initialization and training stages. Empirical results on multiple time series analysis benchmarks demonstrate that COM consistently improves the performance of token-based TS-LLMs, achieving competitive results and strong generalizability. Code is available at this https URL .
[AI-183] Emergent Semantic Representations in World Models through Physical Interaction without Linguistic Supervision
链接: https://arxiv.org/abs/2605.28865
作者: Jiayi Fang
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 10 pages, 3 figures
Abstract:What does a world model learn from physical exploration, without any linguistic supervision? We argue the answer is organized by a single principle: the geometric structure of the physical world. Training a VAE-based world model on random embodied exploration, we find that its latent space develops spatial semantic structure that mirrors physical geometry – direction accuracy 0.677±0.029 versus 0.547 for a randomly initialized encoder, and position RSA 0.192±0.047 versus 0.029 for random encoders (6.6x improvement), showing that training induces genuine structural organization beyond CNN inductive bias. Across 20 temporal checkpoints, prediction performance and semantic alignment co-improve (Spearman r=-0.61, p=0.004), consistent with the shared-driver account. We confirm this through a double knockout: standard KL regularization (beta=0.1) forces the encoder away from geometric structure, and both prediction performance and semantic alignment collapse simultaneously to near-chance by step 50,000 – exactly as the shared-driver account predicts. Reducing beta to 0.001 restores geometric access and recovers both capabilities together. These findings establish physical world geometry as the organizing principle of world model representations, with direct implications for the design of semantically grounded embodied agents.
[AI-184] Self-Play Reinforcement Learning under Imperfect Information in Big 2
链接: https://arxiv.org/abs/2605.28863
作者: Aalok Patwa
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 11 pages
Abstract:Imperfect-information multiplayer games test whether agents can act under hidden information, sparse rewards, and non-stationary opponents. We study these challenges in Big 2, a four-player imperfect-information card game. We develop a self-play RL framework for Big 2 that enables controlled comparisons between policy-gradient and value-approximating agents. Under a common environment, input representation, training budget, and evaluation protocol, PPO outperforms Monte Carlo Q approximation, SARSA, and Q-learning against random, greedy, and heuristic Big 2 opponents. We further find that moderate entropy regularization improves PPO by preventing the policy from becoming overly deterministic, and that current-policy self-play provides a stronger finite-budget curriculum than checkpoint self-play or fixed-opponent training. Together, these results show that Big 2 is a useful controlled setting for studying deep RL under imperfect information, multiplayer interaction, delayed rewards, and variable action sets.
[AI-185] Behavior-Aware Auxiliary Corrections for Off-Policy Temporal-Difference Prediction
链接: https://arxiv.org/abs/2605.28855
作者: Xingguo Chen,Zhiang He,Yuchen Shen,Shangdong Yang,Chao Li,Guang Yang,Wenhao Wang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Temporal-difference learning with function approximation can be unstable under off-policy sampling. TDC stabilizes off-policy TD through an auxiliary covariance correction, and TDRC further regularizes this correction in a single-timescale recursion. This paper studies a behavior-aware replacement of the auxiliary covariance geometry in the linear prediction setting, which is the standard local model for understanding the feature-space dynamics of value-function approximation. We first replace the TDC auxiliary matrix © by the behavior Bellman matrix (A_\mu), yielding BA-TDC, and then regularize the same behavior-aware equation to obtain BA-TDRC. This two-step construction separates the contribution of behavior-aware geometry from the contribution of regularization. The linear analysis also provides a tractable model for an auxiliary-geometry design question that arises in neural-network value approximation, where feature covariances and temporal transition matrices jointly shape the last-layer correction dynamics. We give a finite-state mean-system formulation, prove fixed-point preservation and almost-sure convergence under a Hurwitz stability condition on the instantiated mean system, and compare deterministic mean rates through the spectral radius of the exact linear error recursion. Experiments on the two-state counterexample, Baird’s counterexample, Random Walk, and Boyan Chain show that the behavior-aware replacement can be highly beneficial by itself on some tasks, but that regularization is necessary for robust performance across harder settings.
[AI-186] Behavior-Induced Mirror-Prox Temporal-Difference Learning for Faster Off-Policy Prediction
链接: https://arxiv.org/abs/2605.28849
作者: Xingguo Chen,Yuchen Shen,Shangdong Yang,Chao Li,Guang Yang,Wenhao Wang
机构: 未知
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Gradient temporal-difference methods provide stable off-policy prediction with linear function approximation, but their practical performance is strongly affected by the geometry induced by the auxiliary-variable metric. Existing Mirror-Prox TD methods typically use the feature covariance metric, whereas hybrid TD methods suggest that behavior-policy transition information can provide a more informative update geometry. This paper proposes a behavior-induced Mirror-Prox temporal-difference method, called STHTD-MP, which replaces the covariance metric in the primal-dual saddle-point formulation with the symmetric part of the behavior-policy Bellman matrix. The method keeps a single learning rate for the primal and auxiliary variables and applies a Mirror-Prox prediction-correction step to the resulting hybrid saddle-point operator. We provide a formal convergence analysis for fixed-policy linear prediction under standard stochastic approximation assumptions: the behavior-induced metric is positive definite, the joint mean system is Hurwitz, boundedness follows from a Lyapunov argument, and the stochastic recursion converges by the ODE method. We further derive projected-oracle ergodic gap bounds and an exact mean-operator comparison with GTD2-MP based on the spectral radius of the deterministic Mirror-Prox error matrix. The analysis shows that STHTD-MP can have a smaller mean contraction factor than GTD2-MP when the behavior-induced metric improves the saddle-point geometry. Exact numerical mean-operator analysis on two-state, Random Walk, and Boyan Chain benchmarks supports this condition, while Baird’s counterexample is identified as a singular boundary case where the strict assumptions fail.
[AI-187] Improved Guarantees for Heterogeneous Treatment-Effect Estimation via Matrix Completion
链接: https://arxiv.org/abs/2605.30319
作者: Anay Mehrotra,Phuc Tran,Van H. Vu,Manolis Zampetakis
机构: 未知
类目: Machine Learning (stat.ML); Artificial Intelligence (cs.AI); Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG); Statistics Theory (math.ST)
备注:
Abstract:A central goal of modern causal inference is estimating heterogeneous treatment effects to answer questions like “how does an intervention affect each unit,” rather than only on average. We study this problem with panel-data where we observe n units across m times under unknown, non-uniform treatment assignments. The data in this setting is naturally represented as a matrix of all unit–time treatment effects. Estimating heterogeneous treatment effects can then be expressed as obtaining a good estimation of each row’s average in this matrix. This allows us to formulate the problem as matrix completion, which can be solved under natural low-rankness assumptions. However, existing matrix-completion guarantees are not powerful enough to get meaningful bounds for the per-row guarantee required for estimating the heterogeneous treatment effect; roughly speaking, they are only useful for estimating average treatment effect bounds, as also illustrated in a recent line of work. We give a simple, computationally efficient estimator that, without knowledge of the propensities and under standard low-rankness and regularity assumptions, achieves a row-wise \ell_2 error of \tildeO(\sqrt\frac1n + \fracnm^2) . Technically, our analysis establishes the first sharp row-wise \ell_2 -perturbation bound for low-rank approximation, complementing existing spectral-, Frobenius-, and entrywise perturbation theory.
[AI-188] What drives performance in molecular MPNNs? An operator-level factorial benchmark
链接: https://arxiv.org/abs/2605.30195
作者: Panyu Jiao,Shuizhou Chen,Yiheng Shen,Yuyang Wang,Runhai Ouyang,Wei Xie
机构: 未知
类目: Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Message-passing neural networks (MPNNs) are widely used for molecular property prediction, but their deployment as monolithic architectures makes it difficult to identify how specific message-passing operators affect performance. We present an operator-level factorial benchmark that decomposes 2D molecular MPNNs into the three families of message-seed initialization, node-edge fusion, and node update operators. The resulting 84 configurations are benchmarked on ten MoleculeNet datasets under a shared experimental setup and statistical analysis protocol. Across this controlled design, performance variation is associated primarily with message construction rather than update complexity. Message-seed initialization shows significant family-level effects for both regression and classification, node-edge fusion shows a significant family-level effect for regression with descriptive advantages for concatenation-based mixing, and the update family shows no statistically supported effect for either endpoint family. A representation probe into the Quinethazone molecule further demonstrates that concatenation-based mixing can better differentiate chemically distinct heteroatoms and withstand oversmoothing than Hadamard gating. Representative configurations selected separately for classification and regression recover competitive performance relative to established molecular graph neural network (GNN) baselines, ranking numerically best on eight of ten benchmark datasets. These empirical results are interpreted through concise mechanistic analyses of representative node-edge fusion and update operators. Our findings provide empirical design heuristics for molecular MPNNs by turning model design from a search over monolithic architectures into a targeted assessment of where and how chemical information enters the message-passing pipeline.
[AI-189] Evaluating Skill and Stability of ArchesWeather and ArchesWeatherGen under Multi-Decadal Climate Simulations
链接: https://arxiv.org/abs/2605.29976
作者: Renu Singh,Robert Brunstein,Antonia Jost,Thomas Rackow,Claire Monteleoni,Yana Hasson,Christian Lessig,Guillaume Couairon
机构: 未知
类目: Atmospheric and Oceanic Physics (physics.ao-ph); Artificial Intelligence (cs.AI)
备注: 29 pages, 16 figures, preprint
Abstract:We evaluate the climate simulation capabilities of ArchesWeather and ArchesWeatherGen, two machine learning models originally trained for weather forecasting and evaluated up to a 10-day lead time. ArchesWeather is a deterministic model, while ArchesWeatherGen is a probabilistic flow-matching model leveraging ArchesWeather’s forecasts, enabling ensemble-based uncertainty quantification. In this work, we adapt these models to act as forced atmospheric models by using additional conditioning on the monthly mean sea surface temperature (SST) and sea ice cover (SIC) as boundary conditions. In particular, we follow the AI Model Intercomparison Project (AIMIP) Phase 1 protocol, which, analogous to the Atmospheric Model Intercomparison Project (AMIP), proposes a standardized experimental setup to evaluate the climate skill of ML-based forced atmospheric models. We present a comprehensive evaluation of both models under these conditions, including comparison against numerical climate models, ablation studies that examine key design choices in the extension, and an analysis of forced versus unforced configurations. Despite being originally developed for weather forecasting, we demonstrate that forced configurations of ArchesWeather and ArchesWeatherGen produce stable long-term climate simulations, have a stable annual cycle, and capture the drift of many climate variables. The models faithfully reproduce ERA5’s climatology, large-scale circulations and interannual variability, and they capture the tails of the distributions.
[AI-190] Mitigating Stethoscope-Induced Shortcuts in Respiratory Sound Classification under Federated Domain Generalization with Causality-Inspired Interventions
链接: https://arxiv.org/abs/2605.29862
作者: Heejoon Koo,Yoon Tae Kim,Miika Toikkanen,June-Woo Kim
机构: 未知
类目: Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI); Sound (cs.SD)
备注: 2 figures, 4 tables, and 5 pages
Abstract:AI-driven respiratory sound classification (RSC) is promising for automated pulmonary disease detection, yet multi-site deployment is hindered by inter-stethoscope variability. We introduce a federated domain generalization (FedDG) formulation for RSC under stethoscope-induced device shifts, where clients use heterogeneous devices and the model is evaluated on unseen devices. Our empirical analysis shows that stethoscope-induced style and disease-specific content are tightly entangled, making deterministic style removal unreliable. In response, we propose a causality-inspired multimodal FedDG framework that combines: (i) a causality-inspired device style intervention network that performs content-preserving style perturbations, (ii) counterfactual text augmentation that neutralizes metadata shortcuts, and (iii) gradient alignment that facilitates device-invariant representations across clients. Built on a multimodal language-audio pretraining model, it outperforms conventional data augmentation and federated learning baselines in leave-one-device-out validation on ICBHI and SPRSound datasets. Code will be released upon publication.
[AI-191] A unified deeplearning framework for contrast-phase-specific virtual monochromatic imaging
链接: https://arxiv.org/abs/2605.29753
作者: Antony Jerald,Hemant K Aggarwal,Brian Nett,Avinash Gopal,Phaneendra K Yalavarthy,Bipul Das,Rajesh Langoju
机构: 未知
类目: Image and Video Processing (eess.IV); Artificial Intelligence (cs.AI)
备注:
Abstract:Dual-energy CT (DECT) enables virtual monochromatic imaging (VMI) and improved contrast resolution, but its clinical adoption is limited by hardware complexity and cost. In this work, we propose a unified deep learning framework that synthesizes contrast-phase-specific virtual monochromatic 50 keV images from single-energy CT (SECT) data by leveraging contrast phase information as a prior. The model is trained using DECT-derived 70 keV and 50 keV image pairs across four contrast phases – Angio, Arterial, Portal, and Delayed – using a novel prior conditioning architecture that integrates contrast phase priors into the energy transformation process. We demonstrate that the proposed unified model achieves contrast enhancement and generalizes well across contrast phases. Additionally, we show that the model can generate 50 keV-like images from SECT inputs, preserving contrast phase-specific dynamics.
[AI-192] DELOS: Detecting Shallow Transits in Kepler Photometry Using a Contrastive-Learning Framework
链接: https://arxiv.org/abs/2605.29428
作者: Qingtian Liu,Jian Ge,XingChen Yan,Kevin Willis,Xinyu Yao,QuanQuan Hu,Jiapeng Zhu
机构: 未知
类目: Earth and Planetary Astrophysics (astro-ph.EP); Instrumentation and Methods for Astrophysics (astro-ph.IM); Artificial Intelligence (cs.AI)
备注: 25 pages, 19 figures, 1 table, submitted to ApJ
Abstract:We present DEtection in phase-folded Light curves with cOntrastive Scoring (DELOS), a contrastive-learning-based framework designed to search for shallow transits in Kepler photometry. DELOS combines GPU-accelerated phase folding, optimized phase binning, and a custom one-dimensional convolutional encoder to assign a transit-likeness score to each folded light curve, thereby producing a score periodogram over trial periods without relying on pre-detected threshold-crossing events. Focusing on intermediate-to-long-period signals with orbital periods of 100-150 days, DELOS was trained on 20 million synthetic light curves generated with realistic transit models and Kepler-like noise properties, achieving a validation accuracy of 99.3 percent on the synthetic validation set. In controlled injection-recovery experiments, DELOS improves the combined precision-recall performance by 15.5 percent relative to Box-fitting Least Squares (BLS) and 11.25 percent relative to Transit Least Squares (TLS) in the low Signal-to-Noise Ratios (low-SNR) regime. It also accelerates the search by factors of approximately 3-5 and 74-80 compared with BLS and TLS, respectively. Applied to a selected Kepler validation sample, DELOS recovered all known shallow intermediate-to-long-period transit signals in the tested period range. These results demonstrate that DELOS provides an efficient and sensitive framework for low-SNR transit searches and represents a practical step toward future searches for longer-period terrestrial planets in Kepler, K2, TESS, PLATO, and Earth 2.0 data. Accordingly, this work is intended as a methodological development and validation study, with the detailed astrophysical validation of newly identified candidates deferred to future work.
[AI-193] Sustainable Metal-Organic Framework Water Harvesters in the Artificial Intelligence Era
链接: https://arxiv.org/abs/2605.29179
作者: Reid A. Coyle(1),Shyam Chand Pal(1),Peter Walther(1),Saeun Park(1),Bin Feng(1,2),Zhiling Zheng(1,2) ((1) Department of Chemistry, Washington University, St. Louis, MO, United States, (2) Institute of Materials Science amp; Engineering, Washington University, St. Louis, MO, United States)
机构: 未知
类目: Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI)
备注: 10 pages of main text, 26 total pages. 3 Figures and 1 Table of Content Graphic
Abstract:Metal-organic frameworks (MOFs) are excellent candidates for water harvesting due to their tunable pore environments, which can be precisely engineered to capture and release water in arid conditions. Integrating artificial intelligence (AI) into MOF discovery can further accelerate the design of high-performance sorbents by identifying structural features that enhance atmospheric water harvesting (AWH), stability, and cycling efficiency. In this Perspective, we examine key MOF design principles, including cooperative adsorption, operational relative humidity (RH), uptake capacity, hysteresis, and scalability. We highlight recent design advancements such as multivariate strategies and long-arm linker extension, and examine how these principles tune pore capacity and hydrophilicity, while preserving stability and crystallinity. Furthermore, we discuss how AI, large language models (LLMs), and data mining can accelerate the discovery process through predictive synthesis, inverse design, and elucidating synthesis-structure-property relationships for the next generation of MOF water harvesters.
[AI-194] Real-rootedness of the Poincaré polynomials of overlinemathcal M_0n: an AI-assisted proof
链接: https://arxiv.org/abs/2605.29151
作者: Gergely Bérczi,Young-Hoon Kiem
机构: 未知
类目: Algebraic Geometry (math.AG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE)
备注: 15 pages
Abstract:We prove real-rootedness for the Poincaré polynomial [ P_n(t)=\sum_i=0^n-3 \dim H^2i(\overline\mathcal M_0,n;\mathbbQ)t^i ] of the Deligne–Mumford moduli space \overline\mathcal M_0,n of stable n -pointed rational curves, proving a conjecture of Aluffi–Chen–Marcolli. The proof starts from the Keel–Manin–Getzler recurrence, but its main new idea is a bivariate deformation F_m(y,t) of the Poincaré polynomial. This deformation reveals a hidden interlacing structure not visible in the one-variable recurrence. For fixed t0 , the zero set of F_m in the y -direction is controlled by a Sturm–Rolle argument on the interval 0y1-t . The original polynomial is recovered on the slice y=1 , and the ordered crossings of the moving roots through this slice give both real-rootedness and strict interlacing. Consequently, the Betti numbers of \overline\mathcal M_0,n form an ultra-log-concave sequence. We further prove real-rootedness and ultra-log-concavity for the Poincaré polynomial of the Fulton–MacPherson space \mathbbP^1[n] of n ordered points in degenerations of the complex projective line. The proof for \overline\mathcal M_0,n was obtained through an iterative AI-assisted workflow with Co-Mathematician, an agentic frontier-model system developed by Google DeepMind. The human role was to pose the problem, evaluate successive attempts, request repairs of gaps, compare the evolving argument with the literature, and assemble the final human-verifiable proof. Our additional human contribution was to observe that a similar residual deformation strategy applies to the Fulton–MacPherson spaces \mathbb P^1[n] , yielding the corresponding real-rootedness theorem. Comments: 15 pages Subjects: Algebraic Geometry (math.AG); Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE) MSC classes: 14H10, 14F45, 68T01, 68T20 Cite as: arXiv:2605.29151 [math.AG] (or arXiv:2605.29151v1 [math.AG] for this version) https://doi.org/10.48550/arXiv.2605.29151 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Gergely Berczi [view email] [v1] Wed, 27 May 2026 22:26:05 UTC (19 KB)
[AI-195] A Minimal Bifurcation Model of Load Imbalance in a Softmax Mixture-of-Experts Router
链接: https://arxiv.org/abs/2605.29121
作者: O. M. Kiselev(Innopolis University, Innopolis, Russia)
机构: 未知
类目: Dynamical Systems (math.DS); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 21 pages, 11 figures
Abstract:We propose a minimal dynamical model of adaptive softmax routing for a two-expert Mixture-of-Experts (MoE) layer. The model is obtained as a mean-field limit of a discrete reinforcement rule: the selected expert receives a small score increment, while all scores undergo regularizing decay. In the symmetric case the limiting system has a supercritical pitchfork bifurcation: for weak feedback there is a unique stable balanced state, whereas above a critical feedback strength two stable asymmetric states appear. When an external asymmetry is added, the pitchfork unfolds into a pair of fold bifurcations forming a cusp in the control-parameter plane. We derive exact parametric equations for the bifurcation set and the local normal form of the cusp catastrophe. Numerical experiments connect this picture to empirical expert load, a small trainable MoE model, hard top-1 PyTorch routing, and a small classification experiment on digits. The results provide a controlled low-dimensional mechanism for abrupt transitions to load imbalance in adaptive MoE routers.
机器学习
[LG-0] DynaFLIP: Rethinking Robotics Perception via Tri-Modal-Dynamics Guided Representation
链接: https://arxiv.org/abs/2605.30350
作者: Jusuk Lee,Seungjae Lee,Jonghun Shin,Hoseong Jung,Sungha Kim,Daesol Cho,H. Jin Kim,Jia-Bin Huang,Furong Huang
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: Project website: this https URL
Abstract:Robot manipulation critically depends on perception that preserves the action-relevant aspects of a scene. Yet most robot learning pipelines are built upon visual encoders pre-trained for static recognition or vision-language alignment, leaving motion understanding to downstream policies. We introduce DynaFLIP, a dynamics-aware multimodal pre-training framework that pushes motion understanding upstream into perception. We construct image-language-3D flow triplets from heterogeneous human and robot videos, and use these triplets as training-time supervision to shape an image-only encoder. Our key idea is to encourage the three modalities to span a small simplex volume in the shared hyperspherical space – a smaller simplex volume indicating stronger alignment. To avoid the geometric ambiguity and trivial collapse of naive volume minimization, we combine simplex-volume minimization with a cosine regularizer and a contrastive objective. Our analyses show that DynaFLIP focuses on control-relevant regions critical for manipulation. The resulting dynamics-aware representations serve as reusable visual backbones and consistently outperform baselines across diverse downstream policies, including VLAs. We validate this across diverse simulation and real-world setups, with gains reaching +22.5% under out-of-distribution scenarios. Our results suggest that robot generalization improves when visual representations are trained to encode not just what is present, but how the world changes under action.
[LG-1] Efficient Test-Time Finetuning of LLM s via Convex Reconstruction and Gradient Caching
链接: https://arxiv.org/abs/2605.30337
作者: Alaa Khamis,Alaa Maalouf
类目: Machine Learning (cs.LG)
*备注:
Abstract:Test-time finetuning (TTFT) is a rapidly evolving paradigm that adapts a language model to each prompt by retrieving related sequences, updating the model on them, and then evaluating the prompt. However, TTFT is only practical if it is fast: selection and finetuning both happen per query, making each a direct bottleneck. Existing methods trade speed for quality: fast retrieval is often redundant, while stronger diversity-aware selection adds prohibitive per-query cost. We introduce HullFT, a geometric approach to TTFT that addresses both bottlenecks. Given a query, HullFT first represents the query embedding as a sparse convex combination of few training sequences, using efficient projection-free Frank-Wolfe optimization. This yields a support set that is inherently relevant and diverse. We then convert the fractional convex weights into an exact integer multiset for finetuning through a geometric integerization procedure. The resulting multiplicities naturally create repeated examples, which we exploit with Gradient Reuse to amortize forward-backward computation across repeated finetuning steps. Our experiments show that HullFT improves the quality-efficiency tradeoff over current state-of-the-art TTFT methods, achieving lower bits-per-byte at substantially lower total runtime.
[LG-2] Fairness-Aware Federated Learning with Trajectory Shapley Value
链接: https://arxiv.org/abs/2605.30336
作者: Daniel Kuznetsov,Ziqi Wang
类目: Machine Learning (cs.LG)
*备注: Accepted for publication at the 24th European Control Conference (ECC 2026)
Abstract:Federated learning is an emerging distributed paradigm that addresses the challenges posed by heterogeneous, privacy-sensitive data. It enables multiple clients to train a model collaboratively by aggregating their local updates at a server. However, conventional aggregation schemes typically use fixed weights that fail to reflect unequal and time-varying client contributions, leading to biased and unstable learning. To improve fairness and stability, we propose the Trajectory Shapley Value (TSV), a contribution metric that evaluates how each client influences the optimization trajectory of the global model using a validation-based, temporally consistent utility. Building on TSV, we design FedTSV, an adaptive aggregation method that converts per-round evaluations into dynamic client weights, allowing the server to respond to heterogeneous and adversarial participation in real time. Experiments on benchmark datasets show that FedTSV accelerates convergence, improves robustness, and yields more equitable contribution assessments, thereby providing a principled foundation for fairness-aware federated optimization.
[LG-3] When why and how do diffusion posterior samplers fail? A finite-sample lens
链接: https://arxiv.org/abs/2605.30330
作者: Benjamin A. Burns,Sara Fridovich-Keil
类目: Machine Learning (cs.LG)
*备注: All code for experiments is available at: this https URL
Abstract:Diffusion models have excellent capacity to model complex distributions of natural data, which has made them a popular and effective choice for posterior sampling in imaging inverse problems. Existing methods can incorporate any measurement model at inference time but must use an inexact approximation for the likelihood at intermediate timesteps for computational tractability. Although these approximations can often work well empirically, their downstream effect on the sampled posterior is poorly understood and can result in unexplained failures. To understand when, why, and how these likelihood approximations propagate to erroneous posterior distributions, we introduce a finite-sample perspective on posterior sampling that approximates the posterior to arbitrary precision as training set size tends towards infinity, for any forward model and prior distribution. Using this finite-sample lens, we observe that popular posterior sampling approximations tend to under- or over-estimate the spread of the posterior at intermediate timesteps, causing downstream consequences including sensitivity to early stopping time, inaccurate relative weighting of posterior modes, and hallucination, both of prior modes that are not in the posterior and likelihood modes that are not supported by the prior. Moreover, we find that the cause of these posterior errors requires neither a nonlinear measurement model nor a multimodal posterior, but can arise solely due to a multimodal prior and inaccurate posterior spread at intermediate sampling times. Our finite-sample posterior sampling approach is agnostic to the type of likelihood approximation and the type of (linear or nonlinear) forward model, and can thus serve as a drop-in diagnostic to evaluate the accuracy and failure modes of existing and future posterior samplers.
[LG-4] SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones?
链接: https://arxiv.org/abs/2605.30329
作者: Sy-Tuyen Ho,Minghui Liu,Huy Nghiem,Furong Huang
类目: Machine Learning (cs.LG)
*备注: Project Page: this https URL
Abstract:Autonomous AI research agents aim to accelerate scientific discovery by automating the research pipeline, from hypothesis generation to peer review. However, existing benchmarks rarely test a fundamental bottleneck: whether Large Language Models can judge the methodological viability of a research idea before expending time and computational resources. We introduce SoundnessBench, a curated benchmark of 1,099 machine-learning research proposals reconstructed from ICLR submissions, labeled with reviewer soundness sub-scores, and audited against source papers. SoundnessBench should be interpreted as a benchmark for recoverable proposal-stage soundness rather than exact prediction of full-paper review outcomes. Across 12 frontier LLMs, we find a pervasive optimism bias: under standard prompting, models frequently rate low-soundness proposals as sound, while aggressive prompting largely shifts errors from false positives to false negatives. Additional controls for public-corpus contamination, paper-identifying phrases, surface features, and human audit quality suggest that this behavior is not explained by a single confounder. Our results indicate that current LLMs are not yet reliable as standalone first-gate evaluators for scientific rigor.
[LG-5] Statistical Embeddings for Similarity Retrieval and Interpretable Alignment of Numeric Tabular Datasets
链接: https://arxiv.org/abs/2605.30289
作者: M. Ross Kunz,John Merickel,Keith Wilson
类目: Machine Learning (cs.LG); Applications (stat.AP); Machine Learning (stat.ML)
*备注:
Abstract:Numeric tabular datasets are the dominant data format in scientific practice, yet large language models lack native mechanisms for representing numeric datasets in a meaningful way across heterogeneous feature spaces. Existing approaches either target predictive modeling over individual datasets, which requires a shared set of variable definitions, or lack mechanisms for interpretable cross-dataset alignment. The proposed methodology characterizes numeric tabular datasets through structured exploratory data analysis descriptors, embeds those descriptors into a shared vector space using a pretrained sentence transformer, and quantifies cross-dataset similarity via Canonical Correlation Analysis (CCA). Furthermore, a penalized formulation of CCA is applied to recover sparse, interpretable variable-level correspondences between datasets, identifying which statistical descriptors or variable-level quantities drive cross-dataset alignment without requiring shared variable names or feature conventions. Differential privacy is optionally applied to the descriptor set prior to embedding, supporting deployment in sensitive data contexts without requiring access to raw observations at time of comparison. The methodology is evaluated across 15 datasets spanning general-purpose benchmarks, materials informatics, and nuclear-grade graphite characterization. Results demonstrate a total P@1 score of 0.9, with known nearest-neighbor retrieval and cluster structure remaining robust across embedding ablations and differential privacy budgets. The proposed framework provides a principled pathway for integrating heterogeneous numeric data into retrieval-augmented generation pipelines while preserving statistical context, with direct applications to data-driven algorithm selection and simulation model initialization for unknown datasets.
[LG-6] Neural Operator-Based Surrogate Model for CFD:Helical Coil Steam Generator in Small Modular Reactor
链接: https://arxiv.org/abs/2605.30277
作者: Minseo Lee,Seongmin Oh,Chaehyeon Song,Bumjin Cho,Shilaj Baral,Sangam Khanal,Minseop Song,Joongoo Jeon
类目: Machine Learning (cs.LG); Fluid Dynamics (physics.flu-dyn)
*备注:
Abstract:Real-time thermal-hydraulic simulation is essential for digital twin (DT) technology that supports the safe and efficient operation of small modular reactors (SMRs). Computational fluid dynamics (CFD) provides high-fidelity flow analysis, but its computational cost prevents direct use in DT applications. AI-based surrogate modeling has been actively investigated to address this limitation, yet neural operator–based surrogates for CFD-level transient analysis of SMR-specific geometries have not been reported. This study presents an integrated framework that combines a reduced-order model (ROM) with neural operators, applied to the helical coil steam generator (HCSG) of the System-integrated Modular Advanced Reactor (SMART). Two ROM strategies tailored to each CFD data type were compared, an MLP-based autoencoder (AE) for unstructured mesh data and a convolutional autoencoder (CAE) for structured mesh data, and each was coupled with the deep operator network (DeepONet) to construct the latent DeepONet (L-DeepONet). The Fourier neural operator (FNO) was additionally adopted for comparison. A multi-scale technique was incorporated into both frameworks to mitigate spectral bias and improve the prediction of Kármán vortex streets developing inside the HCSG. The multi-scale L-DeepONet captured the instantaneous periodic vortex dynamics in both velocity and pressure fields, while the FNO and its multi-scale variant predicted the time-averaged mean flow and provided reliable pressure drop estimates. These complementary characteristics provide a practical model-selection guideline that links each architecture to specific DT objectives based on CFD data type and the required level of flow resolution.
[LG-7] Digitally enriching a screening population for pancreatic cancer using routine blood-based measures and clinical histories
链接: https://arxiv.org/abs/2605.30275
作者: Chris Varghese,Leo Y. Li-Han,Richa Bisht,Ellen Larson,Frank Lee,Ryan M. Carr,Tanios S. Bekaii-Saab,Shounak Majumder,John D. Halamka,Mark Truty,Ajit H. Goenka,Hojjat Salehinejad,Cornelius A. Thiels
类目: Machine Learning (cs.LG); Quantitative Methods (q-bio.QM)
*备注:
Abstract:Earlier detection of pancreatic cancer is key to enabling wider access to curative treatment and reducing cancer deaths; however, screening is presently not viable. Latent indicators of pathology are evident in an individual’s disease and blood test trajectories and may predict the development of pancreatic cancer. Longitudinal sequences of coded diagnoses and blood test values accrued by patients throughout their clinical interactions were used to train a custom Transformer-based neural network with a multi-head attention mechanism to predict risk of pancreatic cancer with a multi-year lead time and risk-stratify populations for targeted screening. The cohort comprised 6,017 adults with pancreatic cancer and 177,081 controls (overall median age 75, 45% female) with median 12 years (interquartile range 6.9-16.2) of medical history prior to pancreatic cancer diagnosis. External validation via leave-one-site-out, out-of-sample testing predicting pancreatic cancer 1-, 2-, and 3-years prior to diagnosis demonstrated mean area under the receiver operating characteristic of 0.837 (95% confidence interval 0.827-0.848), 0.797 (95% confidence interval 0.782-0.813), and 0.760 (95% confidence interval 0.745-0.776), respectively. Estimated pancreatic cancer risks were well-calibrated (calibration plot slope 1.08, intercept of -0.077; Brier score 0.025), and a Bayesian population pancreatic cancer prevalence update allows estimated cancer risk outputs to be transportable across settings. At testing, a screening threshold of 3.3% risk of pancreatic cancer in 1-year offered a diagnostic odds ratio of 18.2. Our work therefore lays the foundation for a first population-level digital enrichment tool to widen access to curative-intent management of pancreatic cancer.
[LG-8] OOD-GraphLLM : Graph Large Language Model for Out-of-Distribution Generalized Drug Synergy Prediction KDD2026
链接: https://arxiv.org/abs/2605.30247
作者: Xin Wang,Linxin Xiao,Yang Yao,Wenwu Zhu
类目: Machine Learning (cs.LG); Multimedia (cs.MM)
*备注: 12 pages, 9 figures, ACM KDD 2026
Abstract:Drug synergy prediction (DSP) aims to identify efficacious drug combinations under various cellular contexts with different targets. However, the continual emergence of novel compounds results in variations in molecular scaffolds and sizes, causing drug synergy data to exhibit out-of-distribution (O.O.D.) shifts with respect to topological structure. Existing works rely on in-distribution (I.D.) assumption, failing to handle the O.O.D. shifts. To solve this problem, we study out-of-distribution generalized drug synergy prediction through a graph large language model for the first time. Nevertheless, O.O.D. generalized DSP is highly non-trivial, posing several challenges: i) how to discover structurally relevant and irrelevant molecular representations with respect to cell targets; ii) how to find the optimal graph neural architectures that accurately calculate molecular representations; and iii) how to jointly leverage molecular structural and semantic information in LLMs. To address these challenges, we propose OOD-GraphLLM, a novel graphLLM framework which is able to accurately predict drug synergy under O.O.D. settings via jointly optimizing molecular graph representation and biomedical semantic language representations in a unified manner. Furthermore, we finetune DrugSyn-LLM, a biomedical LLM, and employ a retrieval-augmented biomedical instruction tuning strategy to align molecular topological information and molecular semantic information with language-based reasoning for O.O.D. generalized DSP. Both the source code (this https URL) and released model (this https URL) are publicly available, where users are allowed to download model resources and interactively use the system through a web interface.
[LG-9] Anti Mode-Collapse in Mean-Field Transformer via Auxiliary Variables
链接: https://arxiv.org/abs/2605.30229
作者: Masaaki Imaizumi,Masanori Koyama,Noboru Isobe,Kohei Hayashi
类目: Machine Learning (cs.LG)
*备注: 39 pages
Abstract:We use a mean-field-based transformer model to theoretically investigate how auxiliary variables, such as positional encoding, prevent mode collapse of self-attention mechanisms. The use of mean-field transformers to analyze the properties of self-attention mechanisms has garnered significant attention in recent years due to their ability to comprehensively analyze token interactions. However, analysis of this simple model suggests that mode collapse, where token distributions degenerate to a single point, occurs during long inferences (i.e., many layers), indicating a discrepancy with reality. This study investigates this mean-field transformer model and demonstrates that the introduction of auxiliary variables, such as positional encoding, acts as a counterforce against theoretical mode collapse. Specifically, we show that in the theoretical scheme, the energy-maximizing distribution does not degenerate to a single point; instead, it is characterized by a pushforward of the auxiliary variable distribution, thereby avoiding concentration in the Dirac measure. Our main examples are the positional encoding and the fixed prompt insertion treated as a parallel auxiliary-variable mechanism. Furthermore, we demonstrate that positional encoding and prompt insertion possess universality of representation in the limit, meaning that the limit distribution of inference can exactly represent a wide class of distributions. We also analyze several key properties of positional encoding and metastability, and validate our theoretical results through mathematical experiments.
[LG-10] ExDBSCAN: Explaining DBSCAN with Counterfactual Reasoning – Additional Material
链接: https://arxiv.org/abs/2605.30225
作者: Pernille Matthews,Lena Krieger,Tommaso Amico,Artur Zimek,Thomas Seidl,Ira Assent
类目: Machine Learning (cs.LG)
*备注:
Abstract:Clustering is an unsupervised technique for grouping data points by similarity. While explainability methods exist for supervised machine learning, they are not directly applicable to clustering, making it challenging to understand cluster assignments. This interpretability gap is particularly evident in the popular density-based method DBSCAN, which assigns points as inliers (cluster members in dense regions) or outliers (noise points in sparse regions). DBSCAN does not provide insight into why a particular point receives its assignment or whether its assignment is robust to small changes in the data. To address the lack of explainability, we introduce ExDBSCAN, a density-aware, post-hoc explanation method. ExDBSCAN offers actionable counterfactual explanations, with theoretical guarantees for validity. It generates multiple counterfactuals using a density connected weighted graph, adopting a physics-inspired model that repels counterfactual candidates from one another (diversity), while pulling them toward the instance to explain (proximity). Empirical evaluation on 30 tabular datasets comparing against four baselines shows that ExDBSCAN outperforms all baselines while attaining perfect validity and retrieving diverse, proximal counterfactuals.
[LG-11] riSearch: Learning to Optimize Triangulations via Bistellar Flips
链接: https://arxiv.org/abs/2605.30220
作者: Yiran Wang,Guido Montúfar
类目: Machine Learning (cs.LG)
*备注:
Abstract:We introduce TriSearch, a reinforcement learning framework for optimizing objectives over triangulations of a polytope via bistellar flips. The key idea is a circuit-supported subtriangulation action representation: feasible flips are encoded by their supporting circuit and realized local subtriangulation, enabling a learned policy to rank them using local geometric and combinatorial features. This yields a dimension-agnostic interface and enables efficient traversal of the flip graph without explicit enumeration of the full triangulation space. Instantiated in 3D and 4D, TriSearch generalizes zero-shot from small training instances to larger polytopes with exponentially larger search spaces. It achieves top performance on metric objectives in 3D and, in 4D, discovers more distinct Fine, Regular, Star triangulations of reflexive polytopes, corresponding to Calabi-Yau threefolds, than existing samplers under a fixed budget.
[LG-12] MarginGate: Sparse Margin-Triggered Verification for Batch-Invariant LLM Inference
链接: https://arxiv.org/abs/2605.30218
作者: Kexin Chu,Yang Zhou,Wei Zhang
类目: Machine Learning (cs.LG); Performance (cs.PF)
*备注: 13 pages, 5 figures, 11 tables
Abstract:Temperature-zero BF16 LLM inference is often treated as reproducible, yet the same request can emit different tokens when decoded alone or inside a larger batch. Existing fixes use batch-invariant operators or LLM-42’s per-token verification, incurring cost even when most steps are stable. We ask whether verification can be applied exclusively to flipped tokens. Across five models, batch-induced token flips are sparse on the flip-rate benchmarks: on MATH500, Llama-3.1-8B flips on 0.48% of synchronous decode steps, and all tested models stay within the 0.3-1.3% range on MATH500, GSM8K, and HumanEval. K/V perturbations remain flat before flips, while low top-1/top-2 logit margins expose much of the flip risk. MarginGate turns these observations into a verifier policy: it keeps BF16 decoding on high-margin steps, verifies only low-margin steps, and repairs confirmed mismatches by replacing the current K/V column. We evaluate on four datasets, calibrating on MATH500 and transferring to GSM8K, SharedGPT, and HumanEval. MarginGate restores 100% sequence-level deterministic decoding on Llama-3.1-8B and Qwen2.5-14B with 18.56%/15.05% verifier trigger rates, reducing LLM-42’s latency increment by 2.23x/1.99x relative to always-on verification. On DSR1-Distill-Qwen-7B, the same policy reaches determinism in a harder regime at 49.50% triggers.
[LG-13] Faithful Embeddings of Irregular and Asynchronous Data for Online Log-NCDEs
链接: https://arxiv.org/abs/2605.30213
作者: Benjamin Walker,Alexandre Bloch,Lingyi Yang,Sam Morley,Terry Lyons
类目: Machine Learning (cs.LG)
*备注: 34 pages, 16 figures
Abstract:Continuous-time models are a natural choice for irregular and asynchronous data. A central design choice is how to embed discrete observations into continuous time. Interpolation- and imputation-based embeddings reconstruct a continuous observation path, making the model sensitive to the choice of reconstruction. We show that this reconstruction step is unnecessary; under mild conditions, compact-set universality on the model input space transfers to the data space whenever the embedding from data to input is continuous and injective. Guided by this result, and building on the rectilinear control path for Neural Controlled Differential Equations (NCDEs), we introduce a continuous and injective embedding for Log-NCDEs, a universal class of continuous-time models. Our approach records observations as increments and composes them over arbitrary query intervals to directly form log-signatures. This provides interval-level summaries without first interpolating the observed variables, while supporting online computation. Experiments on synthetic controlled dynamics and real-world time-series datasets show that the representation is accurate, efficient, and robust to irregular, asynchronous, and sparse observations.
[LG-14] Active Continual Learning with Metaplastic Binary Bayesian Neural Networks ICML2026
链接: https://arxiv.org/abs/2605.30198
作者: Kellian Cottart,Théo Ballet,Djohan Bonnet,Damien Querlioz
类目: Machine Learning (cs.LG)
*备注: Accepted at ICML 2026
Abstract:Always-on edge systems must keep learning as conditions change under tight compute budgets and must detect unreliable predictions. Bayesian binary neural networks are attractive in this setting, but mean-field Bernoulli posteriors can saturate on long non-stationary streams, wiping out epistemic uncertainty and freezing plasticity. We propose BiMU, derived from a bounded-memory variational objective that balances stability, plasticity, and forgetting. BiMU combines a data term with controlled relaxation toward the prior and an uncertainty-dependent step size that prevents saturation and sustains informative uncertainty. This non-degenerate posterior enables fully online, buffer-free active querying via Monte Carlo disagreement, reducing label queries and backpropagation updates under imbalance. BiMU sustains learning and strong OOD detection on 1000-tasks Permuted-MNIST, and on OpenLORIS-Object achieves up to 32 \times label/update savings at matched accuracy under class imbalance and feature compression.
[LG-15] Mean-Field Diffuser: Scaling Offline MARL to Thousands of Agents
链接: https://arxiv.org/abs/2605.30190
作者: Wenhao Li,Xiangfeng Wang,Bo Jin
类目: Machine Learning (cs.LG)
*备注: 71 pages, 15 figures, 16 tables
Abstract:Diffusion-based planning has achieved strong results in single-agent offline reinforcement learning, yet scaling to many-agent systems remains intractable due to the curse of dimensionality in the joint trajectory space. We introduce MF-Diffuser, a framework that lifts trajectory planning to the Wasserstein space of trajectory distributions, where the propagation of chaos ensures a small representative subset of agents captures the full population dynamics. Our approach features a value-weighted chaotic entropy objective that reconciles generative fidelity with return maximization, and a hierarchical coarse-to-fine strategy that progressively grows the agent population during denoising. We establish end-to-end suboptimality bounds with four interpretable terms, revealing that mean-field approximation error scales as O(H^2/\sqrtN) while offline distribution shift provably does not grow with population size N , and prove the generated policy is an approximate mean-field Nash equilibrium with explicit convergence guarantees. Experiments on three mean-field RL benchmarks – spanning stage games, sequential dynamics, and adversarial team competition – show MF-Diffuser achieves the best return in the majority of settings, with the largest gains on suboptimal offline data and at extreme scales ( N \geq 10^3 ).
[LG-16] Can AI Weather Models Predict Beyond Two Weeks? A Quantitative Benchmark and Analysis of Long Rollouts
链接: https://arxiv.org/abs/2605.30184
作者: Fanny Lehmann,Firat Ozdemir,Yun Cheng,Torsten Hoefler,Sebastian Schemm,Benedikt Soja,Siddhartha Mishra
类目: Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph)
*备注:
Abstract:While AI weather models excel at short-to-medium range forecasts (up to 15 days), they frequently suffer from ill-defined “instabilities” when rolled out over longer horizons. This work addresses the lack of a formal taxonomy by categorizing these failures into three distinct regimes: blow-up, drift, and loss of seasonality, through year-long rollouts of nine state-of-the-art AI weather models. Our analysis reveals that stability hinges on the treatment of small spatio-temporal scales: unstable models amplify high-frequency energy, while stable models act as denoisers when noise is added to their inputs. Far from reducing these models to mere stochastic parrots, our findings highlight that stable models generate unique weather trajectories, conditioned on the initial state. We verify our findings through ablation studies on architectural design choices, conducted using state-of-the-art Vision Transformer (ViT) AI weather model architectures.
[LG-17] SAHG: Sector-Anisotropic Hyperbolic Graph Model for Social Bot Detection
链接: https://arxiv.org/abs/2605.30166
作者: Hanning Lu,Yingguang Yang,Jinwei Su,Yang Liu,Zhaoqian Yao,Yaoming Li,Taoran Liang,Ziyi Zhang,Ran Ran,Kefu Xu,Bin Chong
类目: ocial and Information Networks (cs.SI); Machine Learning (cs.LG)
*备注:
Abstract:LLM-driven social bots can generate fluent, human-like text, reducing the discriminative advantage of content-based detection alone. However, coordinated campaigns still leave relational patterns – interactions, behavioral similarity, shared neighborhoods, community positions, and coordinated activity – that graph-based methods can exploit. Existing graph detectors face two challenges when exploiting such evidence. First, Euclidean GNNs distort hierarchical and scale-free social graphs; while hyperbolic geometry addresses this volume-growth mismatch, fixed-curvature models still assign uniform geometric resolution to structural directions with different densities and separation needs. Second, relational evidence is not always reliable: sophisticated bots forge heterophilic connections with genuine users, causing neighborhood aggregation to mix bot and human signals and dilute account-level evidence. We propose \textscSAHG (Sector-Anisotropic Hyperbolic Graph), addressing both challenges. \textscSAHG learns a direction-dependent curvature field \gamma(u) that adapts geometric resolution across structural directions, and uses sector prototypes to convert angular concentration and alignment into classifier-readable features. To prevent contaminated aggregation from overwhelming account-level evidence, \textscSAHG encodes per-account features and graph-neighborhood representations in two independent SAH channels, fusing them only at the classifier. Experiments on Fox8-23, BotSim-24, and MGTAB show that \textscSAHG achieves the highest accuracy and F1 on all three benchmarks, outperforming feature-based, graph-based, LLM-based, and isotropic hyperbolic baselines. Ablation and geometric analyses confirm the effectiveness of the anisotropic geometry and dual-channel design.
[LG-18] RL2ML: Finite-Rollout Surrogate Objectives from Reinforcement Learning to Maximum Likelihood
链接: https://arxiv.org/abs/2605.30154
作者: Yifu Zheng
类目: Machine Learning (cs.LG)
*备注:
Abstract:Correctness-based Reinforcement Learning with Verifiable Rewards (RLVR) trains language models from binary feedback on sampled outputs, but the objective optimized in expectation and the stochastic update geometry induced by finite rollout groups are often conflated. This paper develops RL2ML, a family of finite-rollout surrogate objectives with a closed-form, exactly unbiased gradient estimator. The family continuously connects standard reinforcement learning, maximum-likelihood-like training, and beyond-maximum-likelihood objectives while preserving estimator-objective alignment under a fixed rollout budget. We introduce the group-level update scale to characterize how a rollout group is reweighted after its empirical success count is observed, revealing a subcritical-supercritical update-scale transition that is hidden by population-level objective notation alone. Building on this distinction, calibrated metric-gain analysis and exact variance decomposition show that the best choice of surrogate objective is determined neither by proximity to maximum likelihood nor by the population-level weight alone. Instead, it depends jointly on the evaluation metric, local sensitivity, and estimator variance. The remaining degree of freedom in the surrogate objective family can therefore be formulated as a one-dimensional optimization problem rather than treated as an unconstrained hyperparameter.
[LG-19] Learning to Extrapolate to New Tasks: A Relational Approach to Task Extrapolation ICML2026
链接: https://arxiv.org/abs/2605.30132
作者: Adam Ousherovitch,Yixin Wang
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: ICML 2026
Abstract:Modern learning systems excel at interpolation but struggle to generalize to unseen tasks outside the training distribution’s support. This failure occurs even in simple settings, such as handling task parameters beyond the training range, and persists despite advances in foundation models. To this end, we develop the Relational Task Extrapolator (RTE), an algorithm designed to enable systematic extrapolation to novel tasks. The key observation is that extrapolation is inherently relational: extrapolating to unseen tasks requires learning how tasks transform into one another. If a model learns the transformation between tasks A and B during training, it can apply that same transformation to relate known tasks to unseen ones at test time. RTE operationalizes this idea by decomposing each target task into a known anchor task and a transformation linking the anchor and target. It then learns a relational operator, mapping an anchor-transformation pair to predictions for the target task. We instantiate RTE across multiple task extrapolation regimes in function prediction, e.g. where target tasks use out-of-range parameters (parameter extrapolation), have greater compositional depth (length extrapolation), and/or recombine function primitives in unseen ways (compositional extrapolation). We further extend RTE to sequence prediction, integrating it into fine-tuning algorithms for foundation models. Across empirical studies, we find that RTE substantially outperforms existing approaches on extrapolation to novel, unseen tasks.
[LG-20] Privacy-Enhanced Zero-Order Federated Learning via xMK-CKKS over Wireless Channels
链接: https://arxiv.org/abs/2605.30123
作者: Anthony Ayli,Khalil Harris,Jihad Fahs,Mohamad Assaad
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 12 pages, 3 figures
Abstract:Homomorphic encryption (HE) enables privacy-preserving aggregation in federated learning (FL) by allowing the server to operate on encrypted data without decryption. Existing HE-over-the-air methods mainly rely on single-key HE schemes and require channel estimation or pre-equalization to compensate for wireless fading. However, single-key HE remains vulnerable to honest-but-curious clients sharing the same secret key. In addition, compromising a single client may compromise the security of the entire network, while multi-key HE schemes provide stronger client-level security by assigning each device its own secret key. We propose a four-phase protocol that enables xMK-CKKS, a famous multi-key HE scheme, aggregation over a shared wireless channel without channel estimation. The protocol retransmits partial public keys and ciphertexts through the same channel realization, so that the dominant large-modulus encryption terms cancel algebraically during decryption. We integrate this protocol with zero-order FL over slowly varying LoS-dominant channels, where each device transmits a single encrypted scalar per round and the communication/encryption overhead is independent of the model dimension. We prove that the decoded encryption noise preserves the (O(1/\sqrtK)) convergence rate up to a negligible noise floor. The protocol is secure against an honest-but-curious server colluding with up to (N-1) clients, and numerical results on MNIST validate the analysis.
[LG-21] Striding Across Reynolds Numbers: Representation Geometry in Neural PDE Generalisation
链接: https://arxiv.org/abs/2605.30112
作者: Jianing Shi
类目: Machine Learning (cs.LG)
*备注: 12 pages, 8 figures, 5 tables
Abstract:Cross-Reynolds generalisation in neural PDE solvers remains poorly characterised. On the canonical forced 2D Navier-Stokes benchmark, a trained Fourier Neural Operator reaches 46.68% relative L2 error under a 10x Reynolds-number shift, yet zero-forward-model retrieval baselines already improve to 41-42%. This suggests representation geometry as a major organising variable among the tested methods. We test this hypothesis through ConvAE-Relay, which matches states in a source-trained convolutional autoencoder latent space and borrows dynamics from a source-regime database, achieving 38.34+/-0.07% using only a source-regime database and no target-regime fitting, labels, or database entries. A 2x2 ablation isolates matching quality as dominant over the update rule. Oracle experiments confirm that source-regime dynamics directions remain transferable (cosine similarity ~0.84) when matching stays on-manifold; autoregressive drift is the primary bottleneck (~12 percentage points). From the learned-prediction side, a U-Net with multi-scale skip connections achieves 34.72+/-0.60%, consistent with the retrieval-side finding that local, multi-scale representations organise cross-Reynolds transfer among tested methods. All claims are scoped to this benchmark.
[LG-22] Convergence Theory for Iterative LLM -Based Neural Architecture Search: A Parametric Cross-Entropy Framework with Closed-Form Proxy Reliability NEURIPS2026
链接: https://arxiv.org/abs/2605.30103
作者: Santosh Premi Adhikari,Radu Timofte,Dmitry Ignatov
类目: Machine Learning (cs.LG)
*备注: 14 pages, 2 figures, 2 tables. Submitted to NeurIPS 2026
Abstract:Large language models (LLMs) are increasingly used as generators in iterative neural architecture search (NAS), yet no formal convergence theory exists for this class of algorithms. We model iterative LLM-NAS as a parametric Cross-Entropy (CE) method over executable programs and prove six results: (1) iterative LLM fine-tuning on elite architectures is equivalent to the CE update restricted to the LLM parametric family; (2) expected architecture quality is monotonically non-decreasing across cycles; (3) elite-set probability converges to a fixed point at a geometric rate C_t = 1-(1-rho_0)^t; (4) delta-based generation achieves a strictly higher valid-generation rate than full-code generation under a first-order Markov token-error model; (5) the MinHash-Jaccard novelty filter prevents mode collapse; (6) proxy reliability admits the closed-form rho_S = (6/pi) arcsin(rho_P(SNR)/2), yielding the practical diagnostic sigma^2_arch sigma^2_noise as a necessary condition for trustworthy proxy-based rankings. Testing against a 22-cycle, three-LLM, six-dataset experiment with 3,300 generated architectures confirms two predictions quantitatively, two at direction-of-effect level, and explains the proxy-reliability ceiling effect previously reported empirically but left unexplained.
[LG-23] Chess-World-Model: A 10M-Game Benchmark for Exact State Tracking from Chess Move Sequences
链接: https://arxiv.org/abs/2605.30100
作者: Benjamin Walker,Terry Lyons
类目: Machine Learning (cs.LG)
*备注: 20 pages, 4 figures
Abstract:World models require state tracking, which is the ability to maintain a correct latent state across action sequences. Existing benchmarks are often synthetic or language-based, limiting their value as tests of structured state updates in realistic domains. We introduce Chess-World-Model, a large-scale state-tracking benchmark built from 10 million real chess games, where models predict the exact board state reached after a sequence of legal moves. Alongside a held-out real-game split, we include an out-of-distribution split from uniformly random legal play, which tests whether models learn the transition rules rather than shortcuts from common human positions. Prior theoretical and empirical work has shown that Transformers struggle to state-track, while input-dependent linear RNNs require expressive state-transition matrices to do so. We therefore benchmark a causal Transformer, block-diagonal SLiCE, Mamba-3, and Gated DeltaNet with negative eigenvalues under a matched interface and training protocol. The recurrent models strongly outperform the Transformer at 3 and 8 million parameters. Real-game performance saturates above 18 million parameters, but the random-uniform split remains discriminative up to 40 million, exposing failures otherwise hidden by scale. Additionally, ablations show that less expressive state-transition mechanisms reduce performance on the out-of-distribution split for all three recurrent models. Together, these results establish Chess-World-Model as a practical large-scale benchmark for state tracking that exposes failures model scale would otherwise conceal.
[LG-24] Distributionally Robust Set Representation Learning Under Inference-Time Element Corruption ICML’26
链接: https://arxiv.org/abs/2605.30089
作者: Yankai Chen,Hanrong Zhang,Bowei He,Philip S.Yu, Xue (Steve)Liu
类目: Machine Learning (cs.LG)
*备注: Accepted by ICML’26
Abstract:Standard Set Representation Learning methods typically excel on curated data but often overlook the challenge of inference-time element corruption. This refers to scenarios where deployed models encounter element-level degradations, such as outliers or missing components, that may distort set representation and degrade performance. We propose SW-DRSO, a distributionally robust optimization framework tailored for sets. Rather than minimizing loss solely on observed training data, SW-DRSO optimizes a tractable surrogate of the worst-case expected loss over a family of plausible inference-time variations. We introduce a barycentric adversary that approximates the intractable search over corrupted sets by a differentiable training-time optimization over simplex weights. Extensive experiments across four tasks demonstrate that SW-DRSO effectively enhances robustness against corruption while maintaining high overall performance.
[LG-25] Q-ANCHOR: Federated Quantum Learning with ZNE-guided Correction
链接: https://arxiv.org/abs/2605.30075
作者: Hoang M. Ngo,Quan Nguyen,Wanli Xing,My T. Thai
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注:
Abstract:Quantum Federated Learning (QFL) offers a promising framework to train quantum models across distributed clients while keeping data strictly local. Due to its simplicity and low communication overhead, Federated Averaging (FedAvg) is the standard aggregation choice in QFL literature. However, deploying QFL on practical hardware exposes a severe double-drift phenomenon: the global model is simultaneously derailed by client drift from non-IID data and hardware bias from noisy quantum gradient estimates. In this work, we first analyze the convergence of FedAvg under these realistic conditions, mathematically demonstrating that quantum hardware bias creates a persistent error floor that standard averaging cannot correct. To overcome this limitation, we propose Q-ANCHOR, a quantum-aware federated aggregation architecture that anchors server updates with zero-noise extrapolation while applying stateful client correction to suppress both client drift and hardware-induced bias. Our convergence theory proves that Q-ANCHOR successfully mitigates classical client drift while actively reducing the hardware-bias floor. Experimental results demonstrate that Q-ANCHOR achieves significantly more stable training than conventional FL baselines.
[LG-26] Ridge Regression from Poisson Resetting: A Renewal Perspective on Spectral Regularization
链接: https://arxiv.org/abs/2605.30059
作者: Petar Jolakoski
类目: Machine Learning (cs.LG); Statistical Mechanics (cond-mat.stat-mech); Machine Learning (stat.ML)
*备注:
Abstract:We connect stochastic resetting from non-equilibrium statistical physics with ridge regularization in statistical learning. For linear gradient flow, resetting to the origin at rate r produces stationary mean (X^\top X+rI)^-1X^\top y , exactly the ridge estimator with penalty \lambda=r . This uses the known Laplace-transform relationship between ridge regression and exponential-time averaging of gradient flow, with the exponential time now interpreted as the stationary age associated with Poisson resetting. We then extend this identity to general renewal reset laws: the exponential reset time distribution is the unique renewal law whose stationary mean reproduces scalar ridge in every eigendirection as an exact filter identity for every positive curvature, while non-exponential renewal laws generate alternative spectral filters. At the fluctuation level, we study a separate additive Ornstein-Uhlenbeck extension with constant diffusion, interpreted as a stylized SGD approximation. In this setting, the equality holds only at the level of the mean, since the reset process has a nonzero stationary covariance from accumulated OU noise and reset-timing variance, whereas deterministic ridge is a fixed estimator with the same center. Stylized experiments compare the deterministic renewal-induced filters directly and illustrate when filters induced by non-exponential reset-time laws can differ predictively from ridge. The results for the stationary mean and the induced spectral filters are established for continuous-time gradient flow with isotropic resetting on quadratic objectives; the covariance and risk formulas additionally assume additive noise with state-independent covariance.
[LG-27] Sample-Efficient Diffusion-based Reinforcement Learning with Critic Guidance ICML2026
链接: https://arxiv.org/abs/2605.30056
作者: Shutong Ding,Zejia Zhong,Zhongyi Wang,Ke Hu,Bikang Pan,Jingya Wang,Ye Shi
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: accepted by ICML2026
Abstract:Recent advances in reinforcement learning (RL) have achieved great successes by leveraging the multimodality and exploration capability of diffusion policies. Among these approaches, one representative branch focuses on the sampling-based policy optimization. This design enables better exploration capability of the diffusion model, particularly at the beginning of training, but suffer from low exploitation in Q-value information, resulting in a slow policy convergence. Another branch pays attention to gradient-based policy optimization, which sufficiently exploits the gradient of the Q function yet tends to collapse into a unimodal policy with low diversity. To address this issue, we propose CGPO, \textbfCritic-\textbfGuided diffusion \textbfPolicy \textbfOptimization, which effectively balances exploration and exploitation with the training-free guidance technique integrated into the denoising process of diffusion policy. Concretely, CGPO steers action generation toward high-value regions defined by the critic network and uses the guided actions as regression objectives. In this manner, CGPO reduces the time required to obtain high-quality actions and improves final performance with better balance between the exploration-exploitation tradeoff. We validate the effectiveness of CGPO on 5 MuJoCo locomotion tasks, and CGPO achieves state-of-the-art performance compared with existing diffusion-based RL methods. Notably, CGPO is the first success to incorporate diffusion policy into real-world RL, with its superior performance on Franka robot arm grasping tasks. Our official page is released at this https URL.
[LG-28] Fingerprinting Inference Systems of Large Language Models
链接: https://arxiv.org/abs/2605.29979
作者: Anna Wimbauer,Jonas Möller,Erik Imgrund,Konrad Rieck
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注:
Abstract:The behavior of LLMs does not depend solely on the model itself. Components of the inference system, such as the inference engine, attention backend, and hardware platform, subtly influence how inputs are processed. These components differ in their implementations and thereby induce small numerical deviations across systems when running the same model. While prior work has established the theoretical existence of such deviations, their security implications have remained unexplored. In this paper, we show that these deviations are characteristic of specific components and propagate to observable textual outputs, exposing the inference system to any party that can query the model. Building on this observation, we introduce a fingerprinting method that analyzes the prompt-response behavior of LLMs to identify components of the inference system. Our empirical evaluation demonstrates that the inference engine, attention backend, and underlying hardware platform can be identified reliably, even when the LLM is operated at non-zero temperature. We show that preventing fingerprinting is fundamentally hard, as it would require eliminating numerical differences between hardware and software stacks. We therefore propose partial mitigations and discuss their impact.
[LG-29] A Fully Convolutional Approach to Denoising Structural Dynamics Data from X-Ray Photon Correlation Spectroscopy
链接: https://arxiv.org/abs/2605.29975
作者: Nisar Nellikunnummel,Andi Barbour,Lutz Wiegart,Tatiana Konstantinova,Anthony DeGennaro
类目: Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注:
Abstract:We present a fully convolutional denoising autoencoder (FC-DAE) for denoising two-time intensity-intensity correlation functions ( C_2 ) in X-ray photon correlation spectroscopy (XPCS). Unlike conventional denoising autoencoders that are typically restricted to fixed input sizes, the FC-DAE accepts inputs of arbitrary dimensions while preserving correlation structures across diverse dynamical regimes. The model is trained using experimentally derived C_2 data collected at NSLS-II beamlines, with data augmentation applied to expand the diversity of the dataset and reduce overfitting. The FC-DAE successfully recovers intricate dynamical features in low signal-to-noise conditions while maintaining structural fidelity. To assess reconstruction reliability, we employ quantitative metrics to evaluate structural fidelity and identify potential model-induced bias. Our results demonstrate that the FC-DAE provides robust denoising performance with high computational efficiency, enabling recovery of XPCS dynamics under photon-limited and low-dose measurement conditions.
[LG-30] From Short Histories to Long Futures: Horizon-Aware Graph Neural Networks for Long Horizon Forecasting ICPR
链接: https://arxiv.org/abs/2605.29952
作者: Zesheng Liu,Maryam Rahnemoonfar
类目: Machine Learning (cs.LG)
*备注: Accepted for International Conference on Pattern Recognition (ICPR) 2026
Abstract:Accurate long-range prediction of geophysical systems is difficult due to strongly nonlinear dynamics, the high computational cost of full-physics simulations, and the error accumulation that arise when one-step autoregressive surrogates are rolled out over decades. Deep neural network can serve as efficient emulators, but most are trained only for next-step prediction and often drift or become unstable as the forecast horizon grows. We propose a multi-horizon graph neural network emulator that learns state-to-state transitions from a single current time to multiple future lead times within one unified model. The physical domain is represented as a graph, where nodes correspond to spatial locations with time-varying geophysical attributes and edges encode local spatial interactions. Given the current graph state, the model predicts the future evolution of key fields, ice thickness and ice velocities at all nodes, using a shared graph backbone with separate output branches for each target variable. To improve stability, the network predicts state increments relative to the current state, which are then added back to reconstruct future states. Training jointly optimizes all lead times with a unified regression objective, and inference uses a coarse-to-fine rollout that advances with larger jumps and selectively refines with shorter jumps to reduce drift and avoid redundant computation. Experiments on multi-decadal Pine Island Glacier simulations show that our approach achieves higher long-range accuracy and improved stability than both (i) an initial-state baseline that predicts each future time directly from the starting state and (ii) a standard single-step autoregressive rollout, producing a more reliable emulator for downstream climate and sea-level studies.
[LG-31] raceCodec: A Compiler-Backed Neural Codec for Stateful Multi-Flow Network Traffic Traces
链接: https://arxiv.org/abs/2605.29941
作者: Junhui Ding,Xinchen Zhang,Xiaohui Xie,Shinan Liu
类目: Networking and Internet Architecture (cs.NI); Machine Learning (cs.LG)
*备注:
Abstract:Critical networking workflows require high-fidelity packet captures (PCAPs) for testing, security analysis, and protocol validation, not just statistical flow-level summaries. Recent packet generators have demonstrated protocol-constrained PCAP synthesis, but they universally decode directly to raw packet fields. That interface entangles learned behavioral choices with deterministic protocol consequences, which forces packet realization to depend on post-hoc heuristic repair. We identify this decode interface as the fundamental bottleneck and present TraceCodec, a state-aware neural codec for stateful multi-flow traces. TraceCodec lifts each packet into a timed packet action with explicit flow slots and transport cues, then learns a continuous per-packet latent. A deterministic compiler lowers decoded actions back to PCAPs, owning endpoint assignment, TCP state, legality constraints, and packet rendering. The latent layer exposes a generator-facing sequence space, so downstream traffic models can operate on packet-action latents rather than raw header fields. On CICIDS2017 Monday, TraceCodec matches packet count, protocol composition, and flow population to within 0.03%. Raw-field baselines under the same non-repair policy distort flow counts and TCP state by orders of magnitude. Structural diagnostics show that TraceCodec preserves TCP state transitions and multi-flow interleaving that raw-field decoders fragment. This work establishes a new foundation for high-fidelity packet-trace generation.
[LG-32] CRB-Guided Framework Design and Resource Allocation for Indoor mmWave ISCC Systems
链接: https://arxiv.org/abs/2605.29939
作者: Zhonghao Liu,Yahao Ding,Yinchao Yang,Mohammad Shikh-Bahaei
类目: Information Theory (cs.IT); Machine Learning (cs.LG)
*备注: 7 pages, 6 figures, conference(submitted to GLOBECOM)
Abstract:Integrated sensing, communication, and computation (ISCC) provides a promising framework for indoor human-centric applications. In these applications, short-term human pose prediction facilitates continuous human tracking and resource allocation in advance. In this paper, we propose a Cramer-Rao bound (CRB) guided resource allocation framework for indoor mmWave ISCC systems to minimize the human pose prediction error under communication, latency, and energy constraints. We characterize the impact of sensing power on range-estimation uncertainty and point-cloud perturbation based on the CRB. To capture the impact of computation resources on prediction performance, we adopt an adaptive-depth Mamba-based pose prediction model, where lightweight prediction heads are attached after every layer to enable inference with different model depths. With this unified sensing-computation modeling, we establish a quantitative relationship among sensing power, model depth, and prediction error. Furthermore, we formulate a joint resource allocation problem to minimize the pose prediction error. To solve this problem efficiently, we develop an alternating optimization (AO)-based algorithm, where closed-form solutions are derived for the sensing power and model depth update steps. Simulation results show that the proposed scheme significantly reduces pose prediction error compared with baseline methods, validating its effectiveness for resource-constrained indoor human-centric ISCC systems.
[LG-33] Fisher-Preserving Guidance: Training-Free Manifold Constraints for Safe Diffusion Control ICML2026
链接: https://arxiv.org/abs/2605.29937
作者: Hao Ren,Zetong Bi,Yiming Zeng,Le Zheng,Zhi Li,Zhaoliang Wan,Lu Qi,Hui Cheng
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注: ICML2026
Abstract:Diffusion models are effective for waypoint prediction in visual navigation, but standard sampling and test time guidance can produce unreliable or inefficient trajectories when updates drift off the training manifold. We propose Fisher Preserving Guidance with Outer Product Span Projection, a training-free inference method that avoids large Fisher drift associated with off-distribution actions while optimizing a task objective. Our method computes the Fisher-preserving update via a low-rank Jacobian factorization, requiring only a single backward pass per step and enabling real-time use. We further introduce Truncated Fisher Denoising Sensitivity as an uncertainty signal and use it for robust multi-sample action blending. Experiments on toy and realistic navigation benchmarks, including Maze2D with TSDF-based guidance, PushT with official Diffusion Policy weights, and visual navigation in simulation and on real robots, demonstrate consistent improvements in performance over strong diffusion-policy baselines without additional training.
[LG-34] CLUBench: A Clustering Benchmark
链接: https://arxiv.org/abs/2605.29933
作者: Feng Xiao,Dazhi Fu,Chris Ding,Jicong Fan
类目: Machine Learning (cs.LG)
*备注:
Abstract:Clustering is a fundamental problem in data science with a long-standing research history, yielding numerous insightful algorithms. Despite this progress, a systematic and large-scale empirical evaluation that jointly considers conventional algorithms, deep learning-based methods, and recent foundation model-based clustering remains largely absent, leading to limited guidance on algorithm selection and deployment. To address this gap, we introduce CLUBench, a comprehensive clustering benchmark comprising 24 algorithms of diverse principles evaluated on 131 datasets across tabular, text, and image data, involving 178,815 experiments. Importantly, our analyses of (i) the impact of hyperparameter tuning,(ii) the impact of data types and characteristics,(iii) the impact of pretrained embeddings,(iv) large language model-based clustering,(v) the similarity of algorithms, and (vi) the low-rank structures of performance matrices, yield meaningful insights and promising pathways for clustering research. For instance, our study reveals that: 1) All evaluated deep clustering methods do not exhibit a significant advantage compared with the top-performing conventional clustering algorithms (e.g., KMeans, SpeClu) in terms of average performance; 2) For image and text clustering tasks, combining pretrained embeddings with conventional clustering algorithms (e.g., KMeans, SpeClu) offers effective and efficient clustering; 3) Clustering remains a challenging and nontrivial problem, even in the era of increasingly dominant foundation models. Moreover, we propose to use the low-rank structure in cross-model performance matrices to efficiently approximate the overall performance evaluation in practical applications. We further demonstrate the feasibility of model selection based on the performance matrices across all hyperparameter configurations.
[LG-35] A Triple-Modal Contrastive Learning Framework with Sequence Graph and 3D Features for Drug-Target Interaction Prediction
链接: https://arxiv.org/abs/2605.29926
作者: Le Xu,Xi Zhang,Dan Luo,Ting Wang,Xuan Lin
类目: Machine Learning (cs.LG)
*备注: 12 pages, 5 figures, ISBRA 2026
Abstract:Accurate prediction of drug-target interactions (DTI) is critical for drug discovery. Existing methods often rely on single-modal representations (e.g., sequences or graphs) or combine only two modalities, overlooking 3D structural features. To address this challenge, we propose TriMod-DTI, a triple-modal contrastive learning framework that incorporates 1D sequences, 2D graphs, and 3D structures of drugs and proteins, obtaining the universal and complementary feature representations for DTI prediction. We design a Feature Extractor to capture drug and target features across the three modalities, thereby enriching their representations. We further propose a triple-modal contrastive learning strategy to align different modal representations of the same drug or protein in the latent space. By constructing cross-modal positive and negative sample pairs, this approach enhances the model’s discriminative ability. Experiments on three benchmark datasets demonstrate that TriMod-DTI outperforms state-of-the-art methods. The ablation studies validate the contributions of each modality. Moreover, case studies highlight its practical potential for DTI prediction and drug discovery.
[LG-36] Midpoint Generative Models
链接: https://arxiv.org/abs/2605.29920
作者: Daniil Shlenskii,Nikita Gushchin,Lev Novitskiy,Dmitry V. Dylov,Alexander Korotin
类目: Machine Learning (cs.LG)
*备注:
Abstract:We introduce Midpoint Generative Models (MGM), a principled framework for training one-step generative models. MGM is based on a simple symmetry of Flow Matching with linear interpolation: when the two endpoint distributions coincide, the corresponding drift field vanishes at the midpoint time, t=1/2 . We show that the norm of this field defines a valid discrepancy between distributions, which we call the Midpoint Divergence. We extend this discrepancy beyond the midpoint by introducing randomly flipped interpolations and further generalize it by replacing deterministic linear Flow Matching interpolations with symmetric stochastic interpolants, yielding a generalized Midpoint Divergence. Finally, we derive a variational formulation of our generalized divergence, yielding a tractable objective for training a one-step generator. The resulting MGM algorithm offers an effective and theoretically grounded approach to generative modeling, achieving competitive performance against existing one-step generative modeling methods.
[LG-37] Gesture-Aware Indoor THz ISAC Systems for Adaptive Resource Allocation
链接: https://arxiv.org/abs/2605.29913
作者: Zhonghao Liu,Yinchao Yang,Yahao Ding,Yixuan Wang,Mohammad Shikh-Bahaei
类目: Information Theory (cs.IT); Machine Learning (cs.LG)
*备注: 6 pages, 4 figures, conference(Submitted to PIMRC)
Abstract:This paper investigates a multi-user indoor integrated sensing and communication (ISAC) system operating in the terahertz (THz) band, designed for adaptive communication based on gesture recognition. Leveraging gesture tracking through an extended Kalman filter (EKF), the access point (AP) dynamically adjusts resource allocation in response to detected gesture variations, thereby improving sensing accuracy. Based on the gesture recognition results, the AP further updates the communication quality requirements of different users, enabling efficient resource allocation. To this end, an adaptive joint optimization algorithm for power allocation and beamforming is developed to maximize the overall sensing signal-to-interference-plus-noise ratio (SINR) while satisfying the gesture-dependent communication quality of service (QoS) constraints. Simulation results demonstrate that the proposed method effectively responds to gesture dynamics, achieving superior sensing accuracy and communication performance compared with conventional single-variable optimization baselines.
[LG-38] Plan Dont Pose: Long Composite Motion Generation with Text-Aligned BFM
链接: https://arxiv.org/abs/2605.29906
作者: Nikolay Shvetsov,Maksim Bobrin,Nazar Buzun,Dmitry V. Dylov
类目: Machine Learning (cs.LG)
*备注:
Abstract:Text-to-motion (T2M) generation has broad applications in character animation, virtual avatars, and human-robot interaction. Existing methods typically generate pose trajectories or motion tokens directly from language, forcing a single model to handle semantic interpretation, long-horizon structure, and low-level physical realization. This coupling makes them costly and often unreliable for long, compositional, or semantically dense prompts. We propose Text2BFM, the first framework that aligns natural language with pretrained Behavioral Foundation Models (BFMs) for T2M generation without relying on heavy end-to-end motion generators. Text2BFM operates in the latent policy space of a frozen BFM, using it as an executable motion prior. A text-aligned variational behavioral bottleneck compresses BFM policy-latent sequences into compact motion representations that are compatible with language and preserve long-horizon behavioral structure. Generation is performed in this compact behavioral manifold with a lightweight conditional generator, and the resulting latent encoded behaviors are decoded into policy latents that drive the pretrained frozen BFM. By decoupling semantic planning from motion execution, Text2BFM achieves efficient, robust T2M generation and strong performance on long, compositional textual descriptions.
[LG-39] Dissecting the Black Box: Circuit-Level Analysis of LLM Vulnerability Detection
链接: https://arxiv.org/abs/2605.29901
作者: Syafiq Al Atiiq,Chun Zhou,Christian Gehrmann
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: 11 pages, 6 figures. Supported by the Wallenberg AI, Autonomous Systems and Software Program (WASP)
Abstract:Large language models (LLMs) can detect software vulnerabilities, but how do they actually identify vulnerable code? We address this question using mechanistic interpretability; analyzing the internal computations of a neural network to understand its reasoning this http URL Circuit Tracer on Gemma-2-2b, we trace the computational pathways activated when the model classifies 472 C/C++ code samples as vulnerable or safe. Our analysis reveals a surprising finding: the model primarily relies on safety detectors, attention heads that recognize safe coding patterns, rather than directly detecting vulnerability signatures. When these safety detectors fail to activate, the model classifies code as vulnerable. We identify the critical neural components: specific attention heads in early layers (L5, L7) that focus on safety patterns, and Multilayer Perceptron (MLP) neurons in Layer 7 that encode vulnerability-related features. Ablation experiments confirm their causal role; removing Layer 11 drops vulnerability detection accuracy from 100% to 6%, while ablating just 20 neurons in Layer 7 reduces it by 50%.Our findings show that LLM vulnerability detection uses sparse, interpretable circuits (only 16% of model capacity), enabling circuit-level explanations for security predictions and targeted improvements to detection systems.
[LG-40] OVA-IB: One vs All Information Bottleneck for Multi-Modal Alignment
链接: https://arxiv.org/abs/2605.29900
作者: Tianchao Li,Shujian Yu,Xinrui Zu,Zhaolong Wei,Jeremy Gummeson,Jack C.P. Cheng,Robert Jenssen
类目: Machine Learning (cs.LG); Information Theory (cs.IT)
*备注:
Abstract:Contrastive learning is effective for aligning paired views or modalities, but alignment beyond two modalities remains non-trivial and comparatively underexplored. Pairwise CLIP-style losses decompose multi-modal alignment into independent two-way comparisons and therefore do not explicitly model higher-order dependencies among multiple modalities. Recent beyond-pairwise objectives approach this problem from statistical or geometric perspectives, but arbitrary-modality alignment still lacks a principled criterion for defining what each modality should preserve and compress relative to the others. We revisit arbitrary-modality alignment through the Information Bottleneck principle. In multi-modal learning, sufficiency should preserve information predictable from the remaining modalities, while minimality should compress modality-specific information not supported by them. This naturally leads to a One-vs-All view, where each modality is characterized with respect to the remaining modalities. We propose OVA-IB, an Information Bottleneck framework for arbitrary-modality alignment. OVA-IB optimizes a tractable One-vs-All contrastive lower bound for sufficiency connected to a Dual Total Correlation-style objective, uses a parameter-free geometry-aware projection score, and derives a tractable upper-bound regularizer for minimality by bounding each representation’s dependence on its own input with representation distributions induced by the remaining modalities. Experiments on classification, regression, modality-agnostic evaluation, and cross-modal retrieval benchmarks demonstrate strong and robust performance.
[LG-41] Open Problem: Separating Geometric and Algorithmic Compression via Cayley-Table Completion COLT
链接: https://arxiv.org/abs/2605.29885
作者: Dongsung Huh
类目: Machine Learning (cs.LG); Disordered Systems and Neural Networks (cond-mat.dis-nn); Optimization and Control (math.OC); Representation Theory (math.RT); Machine Learning (stat.ML)
*备注: 6 pages. Submitted to the Conference on Learning Theory (COLT) 2026 Open Problem track
Abstract:Modern statistical learning theory and deep learning characterize generalization primarily in terms of continuous capacity control (e.g., norm-based regularization, margin maximization, low-rank bias). While highly successful in continuous domains, deep learning consistently fails to extrapolate exact algorithmic or discrete algebraic rules, reflecting a missing inductive bias toward algorithmic complexity minimization. We propose the Cayley-table completion as the canonical testbed for this missing bias, serving as the discrete algebraic counterpart to matrix completion. Just as matrix factorization combined with weight decay yields an implicit geometric bias toward low linear rank, recent results demonstrate that operator-valued tensor factorizations paired with a flatness prior yield an implicit algorithmic bias toward exact discrete associativity. We pose the open problem of establishing formal exact recovery bounds for Cayley-table completion, and challenge the community to generalize continuous flatness priors to autonomously discover broader discrete algorithmic axioms without combinatorial search.
[LG-42] STAP: A Shuffle-Tokenized App Predictor with Ultra Long Context for Vocabulary-Free Mobile App Prediction
链接: https://arxiv.org/abs/2605.29863
作者: Chengyu Fan,Hang Liu
类目: Machine Learning (cs.LG)
*备注: 15 pages, 9 figures, 5 tables Preprint submitted to Expert Systems with Applications
Abstract:Predicting the next mobile application a user will launch is essential for intelligent device resource management and proactive assistance. Existing models rely on fixed app vocabularies, which prevents them from generalizing across different app ecosystems. Many also depend on user-specific knowledge, which complicates deployment in cold start scenarios. We propose STAP, a Transformer-based model that eliminates the need for a fixed vocabulary. STAP replaces true app identities with randomly reassigned virtual indices via a shuffle mechanism, and compensates for discarded semantic information by processing behavioral sequences with an ultra-long context design. A theoretical analysis shows that, given a sufficiently long context, the predicted distribution converges to the correct one despite the anonymity of the mapping. Experiments on two datasets from different continents demonstrate that STAP achieves strong cross-dataset zero-shot prediction accuracy – a setting where all existing fixed-vocabulary methods are inherently inapplicable – while its cold start performance within each dataset remains competitive with leading models. Furthermore, we introduce a deployment strategy that enables the model to retain a sufficiently long context during continuous inference while keeping latency within acceptable bounds.
[LG-43] Feedback-to-Rubrics: Can We Learn Expert Criteria from Inline Comments?
链接: https://arxiv.org/abs/2605.29857
作者: Kotaro Yoshida,So Kuroki,Yuki Imajuku,Taishi Nakamura,Ryunosuke Iwai,Haruki Goda,Takuya Akiba
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large language models (LLMs) are increasingly used for writing and review support, but their usefulness depends on context-dependent criteria, such as expert preferences or organization-specific conventions, that are often tacit, undocumented, and difficult to elicit directly. We propose a problem setting for learning reusable natural-language rubrics from accumulated inline comments on artifacts such as human-written or LLM-generated drafts. Our method infers rubrics from these comments and iteratively refines them by observing comment-wise mismatches between rubric-conditioned predictions and reference comments. We evaluate the proposed method in real-world review settings and in controlled settings with reference rubrics. These results show that inline comments can be distilled into reusable rubrics that support comment prediction, rubric understanding, and automatic artifact revision.
[LG-44] MIRAG E: Adaptive Multimodal Gating for Whole-Brain fMRI Encoding
链接: https://arxiv.org/abs/2605.29850
作者: Abdulkadir Gokce,Badr AlKhamissi,Martin Schrimpf
类目: Machine Learning (cs.LG)
*备注: Preprint. First two author contributed equally
Abstract:Recent progress in task-optimized neural networks has established encoding models as a powerful tool for predicting brain responses to naturalistic stimuli, yet most existing approaches rely on unimodal representations. The emergence of omni-modal foundation models and rich multimodal neural datasets enables encoding models that jointly integrate visual, auditory, and linguistic information across subjects. We introduce MIRAGE, a brain encoding framework for predicting whole-brain fMRI responses to naturalistic audiovisual stimuli. MIRAGE achieves state-of-the-art performance via a native multimodal backbone and adaptive feature gating across layers. These representations are then combined with a transformer-based brain encoder and a subject-specific linear head over the cortical parcels. Controlled comparisons show that natively multimodal features consistently outperform post-hoc aggregation of independent unimodal features, across architectural levels and backbones. Beyond predictive accuracy, the learned attention weights are directly inspectable to interpret the modality-specific gating profile over the backbone, and each modality traces a distinct anatomical pattern across cortex. Together, these results propose adaptive layer-wise aggregation of natively multimodal features as a generalizable, interpretable, and accurate approach for whole-brain encoding.
[LG-45] BuilDyn: Excitation-Driven Data Generation for Building Thermal Dynamics Modeling and Control
链接: https://arxiv.org/abs/2605.29849
作者: Felix Koch,Thomas Krug,Fabian Raisch,Benjamin Schäfer,Benjamin Tischler
类目: ystems and Control (eess.SY); Machine Learning (cs.LG)
*备注:
Abstract:Machine learning (ML) is increasingly used for data-driven modeling of buildings to enable downstream tasks such as fault detection and diagnosis, and energy-efficient control. While recent work improves generalization across building characteristics, weather, and occupancy, generalization also depends on sufficient exploration of the control-driven system state space. Existing real-world datasets and simulation environments predominantly reflect stationary operation under fixed control policies, resulting in limited excitation and reduced robustness to unseen operating conditions. This paper introduces BuilDyn, a package based on BuilDa that enables customizable excitation strategies for control-oriented data generation. BuilDyn further supports sampling from representative building distributions and provides a Python interface for easy integration into machine learning pipelines. We demonstrate the benefits of BuilDyn by comparing the performance of data-driven ML models trained on non-excited and excited data for one building. With BuilDyn, we hope to advance scalable control-oriented modeling and support future directions such as transfer learning and building-specific foundation models. Subjects: Systems and Control (eess.SY); Machine Learning (cs.LG) Cite as: arXiv:2605.29849 [eess.SY] (or arXiv:2605.29849v1 [eess.SY] for this version) https://doi.org/10.48550/arXiv.2605.29849 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-46] Open World Autoencoding Drift Detection with Novel Class Recognition in Tabular Non-stationary Data Streams
链接: https://arxiv.org/abs/2605.29834
作者: Joanna Komorniczak
类目: Machine Learning (cs.LG)
*备注:
Abstract:Data stream processing has become a landmark in modern machine learning applications, with concept drifts and novel class appearances posing the primary challenges faced by sophisticated recognition methods. This work proposes an unsupervised concept drift detection method that identifies shifts in known class distributions based on the reconstruction errors of an autoencoder, while also enabling the recognition of novel class samples through density estimation of a proxy representation of samples. Using mirrored autoencoders allows for independent incremental adaptation to changing problem distributions for the two considered tasks, resulting in continuous adjustment to evolving concepts and reliable recognition of unknown samples. Conducted experiments used a diverse set of synthetic tabular data streams, where both concept drifts and the emergence of novelties were observed. The results show that the proposed approach is competitive with current state-of-the-art unsupervised drift detectors and novelty classifiers.
[LG-47] When Do Graph Foundation Models Transfer? A Data-Centric Theory ICML2026
链接: https://arxiv.org/abs/2605.29828
作者: Jiajun Zhu,Ying Chen,Peihao Wang,Yixuan He,Pan Li,Aditya Akella,Zhangyang Wang
类目: Machine Learning (cs.LG)
*备注: 21 pages, including appendix. Accepted at ICML 2026
Abstract:Graph foundation models (GFMs) aim to reuse a single backbone across diverse graph domains, yet their transfer is often uneven and can exhibit negative transfer. While most prior work improves transfer through architectural or adaptation choices, we ask a data-centric question: which properties of two graph domains determine how much a fixed representation model changes its outputs? Using a graphon-based continuous limit for dense graphs, we show that for both set-based and message-passing tokenizations, any Lipschitz backbone admits an explicit decomposition of cross-domain output shift into (i) graph-specific finite-sample approximation terms and (ii) an intrinsic, relabeling-invariant domain discrepancy capturing structural mismatch. A key ingredient is positional-encoding (PE) stability: we establish stability guarantees for spectral PEs and highlight contrasting behaviors of eigenvector- versus subspace-based PEs. Experiments on synthetic and real graphs validate the theory and translate the decomposition into guidance for data curation in GFM transfer.
[LG-48] he Interplay Between Interpolation and Aggregation in Regression: Optimal Sample Complexity
链接: https://arxiv.org/abs/2605.29819
作者: Mikael Møller Høgsgaard,Kasper Green Larsen,Liang-Yu Zou
类目: Machine Learning (cs.LG)
*备注:
Abstract:This work investigates theoretically the interplay between interpolation and aggregation in regression. We establish that the \gamma -graph dimension characterizes learnability for a broad class of natural aggregation procedures. Furthermore, we prove that an extremely simple aggregation procedure, combining three interpolating hypotheses via the median, is optimal among all these aggregation procedures, and is strictly more powerful than proper learning. Finally, we show that some hypothesis classes are learnable only by aggregating infinitely many hypotheses or by using non-interpolating aggregation rules (which may predict outside the range of their inputs), and any finite interpolating aggregation fails to achieve even trivial performance.
[LG-49] Gated Graph Attention Networks with Learnable Temperature
链接: https://arxiv.org/abs/2605.29803
作者: Zhongtian Ma,Hao Wu,Yexin Zhang,Qiaosheng Zhang,Zhen Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Graph attention networks learn neighbor importance through data-dependent coefficients, but standard layers lack explicit control over unreliable feature dimensions and use fixed sharpness of attention coefficient distributions. This paper proposes gated graph attention and learnable temperature for common graph attention mechanisms. Gated graph attention filters feature or message responses to reduce the influence of unreliable dimensions, while learnable temperature dynamically adjusts the sharpness of the attention coefficient distribution. Experiments on homogeneous and heterophilic heterogeneous benchmarks show that the proposed variants consistently improve the corresponding graph attention backbones, and controlled noise studies further verify their behavior under feature perturbations. Theoretical analysis explains these results by showing that gating improves robustness when only part of the feature coordinates are reliable, while temperature is beneficial when global noise weakens the discriminability of node features.
[LG-50] MMTM: Tri-Modal Topic Modeling for Long-Form Video via Similarity-Gated Fusion EMNLP2026
链接: https://arxiv.org/abs/2605.29765
作者: Ali Abusaleh,Bhuvanesh Verma,Alexander Mehler
类目: Machine Learning (cs.LG)
*备注: Submitted to EMNLP 2026
Abstract:We introduce MMTM, a modular pipeline for topic discovery in long-form video that integrates speech recognition, audio and visual embeddings, and BERTopic clustering through a deterministic similarity-gated fusion. Evaluated cross-lingually on German (Tagesschau) and English (NBC) broadcast news, joint tri-modal modeling substantially improves topic quality: noise drops from 0.27 to 0.06, transition rate from 0.70 to 0.21, and normalized entropy rises from 0.84 to 0.92, indicating more coherent and temporally stable topics. Cluster validity (Calinski-Harabasz) improves by 5-12X across embedding spaces. Lexical coherence (NPMI) rises from 0.77 to 0.86 on German but is corpus-dependent and does not transfer to the shorter NBC broadcasts. We release the pipeline code and a human-validated 54-hour multimodal video topic corpus with dual-annotator visual evaluation and LLM-assisted labeling.
[LG-51] EMAG: Differentiable 4D Gaussian Mixture Splatting for EEG Spatial Super-Resolution
链接: https://arxiv.org/abs/2605.29731
作者: Alex Lazarovich,Ofir Itzhak Shahar,Gur Elkin,Ohad Ben-Shahar
类目: Machine Learning (cs.LG)
*备注:
Abstract:High-density electroencephalography (HD-EEG) enables fine-grained measurement of cortical activity but requires expensive hardware and lengthy setup times, limiting its clinical and research accessibility. We propose EMAG (EEG Mixture of Anisotropic Gaussians), a differentiable framework that reconstructs HD-EEG signals from a sparse subset of low-density (LD) electrodes by representing brain electrical sources as a mixture of anisotropic 4D space-time Gaussians. EMAG places a mixture of multiple Gaussians at each point of a spherical brain grid, each parameterized by a full 4 x 4 precision matrix, enabling anisotropic spatial spreads and explicit coupling between spatial and temporal dimensions. The forward model renders scalp EEG via differentiable Gaussian field contributions at electrode locations, enabling end-to-end training without explicit source localization supervision. We evaluate EMAG on three public EEG benchmarks (Localize-MI, SEED, and SEED-IV) at super-resolution factors of 2x through 8/16x. EMAG outperforms the current state-of-the-art EEG super-resolution method at most super-resolution factors on three standard benchmarks (Localize-MI, SEED, SEED-IV). The explicit Gaussian parameterization further enables direct visualization and interpretability of learned brain source configurations, potentially opening avenues for clinical and neuroscientific applications, such as source localization or biomarker discovery.
[LG-52] Realistic honeypot evaluations for scheming propensity
链接: https://arxiv.org/abs/2605.29729
作者: Victoria Krakovna,David Lindner,Lewis Ho,Sebastian Farquhar,Rohin Shah
类目: Machine Learning (cs.LG)
*备注:
Abstract:We introduce scheming honeypot evaluations, a framework for testing whether models will pursue instrumental goals if given the opportunity. Our scheming honeypot evaluations take the form of coding tasks in Google’s alignment research codebases. In a real internal deployment setting, Gemini models do not demonstrate unprompted scheming. If prompts explicitly encourage agency (situational awareness or goal-directedness) and/or give the model a hidden goal, models sometimes scheme or attempt sabotage. Validating the realism of our setting, models show low rates of evaluation awareness, usually due to agency prompts rather than the environments.
[LG-53] Bastion: Budget-Aware Speculative Decoding with Tree-structured Block Diffusion Drafting
链接: https://arxiv.org/abs/2605.29727
作者: Soowon Oh,Nam Cao,Yujin Kim,Hojung Jung,Huzama Ahmad,Sangmin Bae,Se-Young Yun
类目: Machine Learning (cs.LG)
*备注:
Abstract:Block-diffusion drafters have recently emerged as a powerful alternative for speculative decoding by predicting multiple future-token distributions in a single parallel step. However, since these parallel predictions are sampled from position-wise marginals rather than fully conditioned sequences, committing to a single greedy path often fails to capture the target model’s preferred trajectory. To address this, we propose BASTION, a budget-aware speculative decoding framework with tree-based diffusion drafting. Unlike existing methods that rely on static tree topologies, BASTION dynamically constructs query-dependent trees by balancing draft quality against hardware constraints. Our framework integrates three synergistic components: (1) an acceptance surrogate that estimates expected accepted length via path confidence, (2) an online latency estimator that calibrates a hardware-aware roofline model, and (3) an adaptive best-first expansion that grows the tree until marginal gains no longer justify incremental verification costs. BASTION is training-free, preserves the target model’s distribution, and requires no per-setting tuning. Across diverse benchmarks and GPU architectures, BASTION achieves up to a 6.61x speedup over standard autoregressive decoding, outperforming state-of-the-art block-diffusion baselines by 39%.
[LG-54] A Systematic Evaluation of Molecular Mixture Behavior Prediction
链接: https://arxiv.org/abs/2605.29698
作者: Roel J. Leenhouts,Nathan K. Morgan,William Green,Jan G. Rittig,Florence H. Vermeire
类目: Machine Learning (cs.LG); Chemical Physics (physics.chem-ph)
*备注:
Abstract:Machine learning for molecular property prediction has focused largely on pure compounds, even though many practical applications depend on mixtures with intermolecular interactions. Recent work has expanded the availability of mixture datasets, but evaluation still focuses mainly on absolute accuracy. However, absolute errors in mixtures conflate pure-component contributions with deviations from ideal mixing. We propose an evaluation framework that decomposes mixture-property error into pure-compound and interaction (non-ideal) components. The framework combines leakage-aware split protocols, ideal-mixture baselines, and excess-property metrics. To support reproducible benchmarking, we curate seven matched pure and mixture physicochemical property datasets. Across multiple mixture-property tasks and model families, we find that strong absolute accuracy can mask poor recovery of non-ideal mixture behavior, and that performance drops substantially under strict molecule splits. These results identify transfer to unseen molecules as a central challenge in molecular mixture machine learning and motivate evaluation beyond absolute accuracy alone.
[LG-55] Momentum Based Reward Design for Low Emission Traffic Signal Control
链接: https://arxiv.org/abs/2605.29693
作者: Chinmay Mundane,Amith Manoharan,Arun Singh
类目: Machine Learning (cs.LG); Robotics (cs.RO)
*备注:
Abstract:Urban traffic congestion is a growing global issue contributing significantly to long commute times and environmental pollution. Traditional traffic signal control systems often fail to adapt to dynamic traffic conditions. Adaptive traffic signal control can improve urban traffic without changing road infrastructure. Deep Reinforcement Learning (DRL) has shown strong performance for this task, but existing delay and queue-based rewards often produce short-sighted or unstable policies. This paper proposes a Momentum-Based Reward Function (MBRF) that encourages vehicles to keep moving rather than penalizing congestion alone. The method is evaluated in SUMO (Simulation of Urban MObility) using standard traffic metrics such as waiting time, queue length, throughput, and CO2 emissions. Results show that the proposed reward produces better throughput-emission trade-offs and more stable learning behavior than delay or queue-based rewards, as well as classical controllers such as Max Pressure and LQF.
[LG-56] A Novel Tensor Product-Based Neural Network for Solving Partial Differential Equations
链接: https://arxiv.org/abs/2605.29688
作者: Qihong Yang,Yangtao Deng,Qiaolin He,Shiquan Zhang
类目: Machine Learning (cs.LG)
*备注: 44 pages, 11 figures
Abstract:This paper presents the Tensor Product Network (TPNet), a novel neural architecture for efficient and accurate function approximation and PDE solving. The core of the proposal involves constructing the solution explicitly as a linear combination of basis functions integrated into the network, with coefficients determined by a direct least-squares solve, thereby bypassing traditional gradient-based training. The key methodological contribution include: (1) an efficient tensor-product scheme that generates multi-dimensional basis functions from combinations of two sets of subnetwork outputs, significantly reducing model complexity and parameter count while maintaining expressivity; (2) a block time-marching strategy to improve computational efficiency in long-time simulations; and (3) a linear reformulation strategy for handling nonlinear PDEs by treating known nonlinear terms as sources. TPNet achieves superior accuracy and shorter training times than conventional neural network solvers. This performance gain stems from its structured design and deterministic least-squares fitting, which contrast with the iterative, often computationally intensive optimization required by mainstream methods like Physics-Informed Neural Networks (PINNs).
[LG-57] Kernel Renormalization in Bayesian Deep Neural Networks: the Equivalent Wishart Ansatz in the Proportional Regime
链接: https://arxiv.org/abs/2605.29684
作者: Paolo Baglioni,Christian Keup,Vincenzo Zimbardo,Rosalba Pacelli,Alessandro Vezzani,Raffaella Burioni,Pietro Rotondo
类目: Machine Learning (cs.LG); Disordered Systems and Neural Networks (cond-mat.dis-nn); Machine Learning (stat.ML)
*备注: 45 pages, 21 figures
Abstract:The scaling limit where both the size of the training set P and the width N of a deep neural network grow at the same rate, the so-called proportional-width regime, has been intensely studied for shallow, single-hidden-layer networks. However, extending these non-perturbative results from shallow architectures to deep non-linear networks has proven very challenging. Here we present an effective approximate approach to predict the generalization performance of Bayesian multi-layer perceptrons (MLPs) of fixed depth L on arbitrary high-dimensional data. We propose an equivalent Wishart Ansatz to capture the dominant stochastic fluctuations of the hierarchical empirical kernels of MLPs. This allows us to perform a large deviation analysis for the partition function of MLPs in the proportional limit, expressed in terms of a renormalized NNGP kernel. In this description, even strong representation learning in the proportional limit is encoded in at most L scalar order parameters, determined self-consistently. Extending the approach to convolutional architectures (CNNs), we identify a hierarchical local kernel renormalization mechanism, which allows to quantify more complex data-dependent transformations of the large-width kernel in CNNs due to finite-width effects. We test our effective theory against sampling experiments from the Bayesian posterior of finite deep neural networks with depths L \sim O(10) and P\sim O(10^3) on classic benchmark datasets, finding overall very good agreement together with two distinct types of systematic deviations.
[LG-58] AMDP: Asynchronous Multi-Directional Pipeline Parallelism for Large-Scale Models Training ICML2026
链接: https://arxiv.org/abs/2605.29664
作者: Ling Chen,Houming Wu,Wenjie Yu
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
*备注: Accepted by ICML 2026, 9 pages, and 8 figures
Abstract:Pipeline parallelism is essential for large-scale model training, but existing asynchronous approaches often degrade convergence due to parameter mismatch between forward and backward passes. We propose Asynchronous Multi-Directional Pipeline parallelism (AMDP) to mitigate this issue while sustaining high utilization. AMDP limits the first stage of each pipeline to process at most two minibatches before backpropagation, bounding the number of parameter updates between forward and backward passes. To alleviate the resulting pipeline bubbles, AMDP launches multiple concurrent pipelines and adapts their number according to pipeline depth. In addition, AMDP accumulates gradients across minibatches and applies them in a single update, ensuring that only a bounded number of minibatches experience parameter mismatch, limited to within one optimization step. Experiments on GPT- and BERT-style models demonstrate that AMDP significantly accelerates training while preserving convergence.
[LG-59] Relational Rank Geometry in Transformers: Detecting and Steering Hidden-State Relation Frames
链接: https://arxiv.org/abs/2605.29634
作者: Mazen Kobrosly
类目: Machine Learning (cs.LG)
*备注: 32 pages, 9 figures
Abstract:Transformer hidden states are often interpreted through local or low-order objects: neurons, sparse features, attention heads, residual-stream directions, or activation patches. This paper studies a complementary object: the rank-indexed geometry of relations among token tuples. I use Plucker sign entropy to test whether r-argument relations leave arity-matched orientation signatures in hidden-state space. Across Llama-family 8B, 70B, and 405B checkpoints, true relation tuples show stronger orientation-sign consistency at the expected rank k=r for r=3,…,6 than scrambled tuples under matched random-control audits. Multi-template audits show that the effects survive surface variation, with all tested 405B rows retaining positive expected-rank margins and 8B/70B retaining positive rows with constructor-specific mixed cells. I then ask whether the same relation geometry can be steered. In an edge-grid clean/corrupt intervention assay over 32 prompts, the row/column scaffold and answer format stay fixed while the YES/NO relation map changes, and the corrupt hidden-state relation frame is patched toward clean or placebo targets. In 70B and 405B, clean-targeted relation-frame paths recover clean-answer behavior and residual relation geometry, while centroid-only and equal-norm controls show negligible recovery. Site/order controls further separate marker-site importance from ordered clean-frame geometry: target clean shape and cross-prompt clean shape recover behavior and residual geometry at the marker interface, whereas corrupt-donor transfer, same-site permutation/reflection, wrong-site clean deltas, centroid-only motion, and equal-norm noise fail or remain far below clean-frame paths. The result is a controlled bridge from relation probing to relation-frame intervention: relation rank geometry can be detected, targeted, and behaviorally validated in transformer hidden states.
[LG-60] MōLe-Λ: Learning the Coupled-Cluster Response State for Energies Gradients and Properties ICML2026
链接: https://arxiv.org/abs/2605.29622
作者: Andreas Burger,Luca Thiede,Abdulrahman Aldossary,Jorge A. Campos-Gonzalez-Angulo,Alex Zook,Jérôme Florian Gonthier,Alán Aspuru-Guzik
类目: Machine Learning (cs.LG); Chemical Physics (physics.chem-ph)
*备注: ICML 2026 AI4Physics
Abstract:Coupled-cluster (CC) theory is often considered the gold standard of quantum chemistry, but its high computational cost limits routine access to accurate energies, forces and response properties. While the right-hand T -amplitudes determine the correlated wavefunction, many practically important observables additionally require the left-hand \Lambda -amplitudes. We introduce MōLe- \Lambda , an extension of Molecular Orbital Learning (MōLe) that predicts the full ground-state coupled-cluster singles and doubles (CCSD) response state by jointly learning right-hand amplitudes (T_1,T_2) and left-hand amplitudes (\Lambda_1,\Lambda_2) from localized Hartree–Fock molecular orbitals. Architecturally, MōLe- \Lambda extends MōLe with \Lambda_1 and \Lambda_2 readouts that mirror the symmetry constraints of the T_1 and T_2 heads, while preserving the original equivariant orbital encoder, odd sign-equivariant decoding, locality and size-extensivity. The resulting model yields accurate CC-quality energies and forces, while simultaneously recovering dipoles, quadrupoles, polarizabilities, the electron density, and 2-electron observables such as the pair density. We show that MōLe- \Lambda further extends the speed advantage of MōLe over full CCSD while substantially expanding the accessible properties, providing a route to wavefunction-level surrogate models for correlated quantum chemistry.
[LG-61] Cluster-Level Attention-Guided Parallel Decoding for Masked Diffusion Language Models
链接: https://arxiv.org/abs/2605.29607
作者: Heqiang Qi,Wei Huang,Mingyuan Bai,Xiangming Meng
类目: Machine Learning (cs.LG)
*备注:
Abstract:Masked diffusion language models (MDLMs) enable parallel decoding by predicting all masked positions at each denoising step, yet existing training-free samplers usually decide which positions to commit at token-level granularity. We revisit this granularity and observe that reliable predictions often emerge as contiguous high-confidence spans, suggesting that the unit of parallel commitment can be larger than a single token. We first group adjacent high-confidence candidates into confidence-induced clusters (CICs) as span-level update units. We then use self-attention maps from the same forward pass to estimate inter-cluster dependencies, enabling conflict-aware selection of mutually compatible CICs for parallel commitment. This yields CLAD (Cluster-Level Attention-Guided Decoding), a training-free cluster-level decoder for MDLMs. Experiments on LLaDA and Dream model families across four reasoning and code-generation benchmarks show that CLAD achieves 1.77x–8.47x speedups over Vanilla decoding while maintaining broadly comparable task accuracy in most settings.
[LG-62] On the Construction and Implications of Low-Loss Valleys in LoRA-based Bayesian Inference
链接: https://arxiv.org/abs/2605.29580
作者: Daniel Dold,Emanuel Sommer,Julius Kobialka,Oliver Dürr,David Rügamer
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:While parameter-efficient fine-tuning methods like low-rank adaptation (LoRA) are standard for large language models, principled estimation of epistemic uncertainty remains challenging. Recent results in the LoRA regime suggest that discrete multi-mode approaches such as deep ensembles offer little benefit over single-mode methods. This contradicts broader observations in deep learning, where ensembling independent optima typically improves generalization, and linking these modes through continuous low-loss valleys further enhances Bayesian model averaging (BMA). Whether such structure exists in the LoRA space and whether it yields functional diversity missed by local or discrete methods has not been studied. We introduce LoRA-Curve, a segmented Bézier curve parameterization in the LoRA space, with two variants: a free configuration that jointly optimizes all control points, and an anchored configuration that connects independently fine-tuned LoRA optima. We prove pathwise continuity and Lipschitz regularity of the loss along the curve and empirically show, across reasoning and classification benchmarks with Qwen2.5 7B, that linear interpolation encounters loss barriers, while our anchored multi-segment curves connect independent optima through continuous low-loss valleys. Combined with flat-minima perturbations and a Jensen-Shannon divergence regularizer, LoRA-Curve yields measurably higher mutual information of the predictive distribution without sacrificing performance, and links continuous parameter-space traversal to functional diversity.
[LG-63] Why Larger Models Learn More: Effects of Capacity Interference and Rare-Task Retention
链接: https://arxiv.org/abs/2605.29548
作者: Jing Huang,Daniel Wurgaft,Rachit Bansal,Laura Ruis,Naomi Saphra,David Alvarez-Melis,Andrew Kyle Lampinen,Christopher Potts,Ekdeep Singh Lubana
类目: Machine Learning (cs.LG)
*备注:
Abstract:Larger models learn tasks smaller models do not. What drives this phenomenon? We develop a simple phenomenological argument that power-law scaling already suggests that a larger model will be able to learn a part of the data distribution that a smaller model fails to learn, even with infinite training data. To validate this claim and identify its causes, we study the effects of model scaling on a synthetic setup consisting of a mixture of tasks that show monotonic scaling curves. The results point to a data-induced competition over resources (neurons). Specifically, smaller models allocate their neurons to high frequency or low complexity tasks, and so they learn solutions that perform poorly on rare and complex tasks. Moreover, this happens even when solutions capable of expressing the desired task exist. We then assess how a larger model circumvents this data-centric bottleneck, finding that it traces to a reduced interference mechanism: larger models can allocate enough resources to common tasks that the gradient updates for those tasks become weak, which means that they do not overwrite rare-task features as they slowly accumulate. Finally, to further validate these claims, we pretrain OLMo models (4M to 4B parameters) on novel tasks of varying frequency and complexity. The results mirror those from our synthetic data experiments: only the larger OLMo models learn the infrequent and complex tasks, and these larger models embed more task features in their representations and show less gradient interference between tasks. Overall, we offer a data-centric account of why larger models learn tasks that smaller models fail to. This helps explain why larger models are better in practice, and it can inform practical questions concerning model sizing and training data mixtures.
[LG-64] he Complexity of Verifying Feedforward Neural Networks in Quantised Settings
链接: https://arxiv.org/abs/2605.29537
作者: Eric Alsmann,Martin Lange,Marco Sälzer
类目: Computational Complexity (cs.CC); Machine Learning (cs.LG); Logic in Computer Science (cs.LO)
*备注:
Abstract:We investigate the computational complexity of neural network verification in quantised settings. We distinguish three classes of Feedforward Neural Networks (FNNs): rational FNNs with exact rational weights, quantised FNNs whose weights come from a finite-width arithmetic, and dynamically quantised FNNs in which rational networks are evaluated with respect to a given finite-width arithmetic. We consider two types of specifications used in the literature. Linear programming (LP) specifications are conjunctions of linear constraints, while bit-vector (BV) specifications allow reasoning at the bit level and can express non-linear constraints. Our results give a complexity landscape of these verification problems. For quantised FNNs with fixed arithmetic precision, we show that verification under both LP and BV specifications remains NP-complete, matching the complexity of the rational case. For dynamically quantised FNNs with BV specifications, we establish upper bounds, complementing a previously known PSPACE-hardness result.
[LG-65] AsymVLM: Asymmetric Token Pruning for Efficient Vision-Language Model Inference
链接: https://arxiv.org/abs/2605.29535
作者: Yilin Feng,Ahmed Burak Gulhan,Mahmut Taylan Kandemir
类目: Machine Learning (cs.LG)
*备注:
Abstract:Vision-Language Models (VLMs) process thousands of visual tokens per image alongside comparatively few text tokens, yet existing compression methods treat both modalities uniformly. We observe that the two modalities have fundamentally different properties: vision tokens are spatially redundant and dominate prefill, while text tokens are causally dependent and accumulate during decoding. Based on this asymmetry, we propose and empirically evaluate AsymVLM, which applies aggressive pruning to vision tokens before prefill using a learned importance scorer with per-sample adaptive budgeting, and temporal threshold-based eviction to text tokens only when they exceed a fixed budget. Our experiments indicate that AsymVLM achieves the highest FLOPs savings (up to 54%) among state-of-the-art methods while outperforming existing approaches by 2–3% on document and chart understanding tasks where visual information is spatially localized and query-specific, and maintaining competitive accuracy on holistic benchmarks. In text-dominated scenarios, our eviction strategy substantially outperforms standard LLM cache compression methods by adapting to the short-context nature of VLM.
[LG-66] Learning to Perturb Hidden Representations for Generalizable Deep Learning
链接: https://arxiv.org/abs/2605.29525
作者: Hua Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:Deep neural networks process data through a cascade of representations: input features, hidden activations, logits, and loss. While perturbations at the input, logit, and label levels have been systematically studied, the intermediate hidden activations, which constitute the bulk of the network’s computation, have received no unified perturbation analysis. In this paper, we establish a unified framework for hidden activation perturbation, revealing that Dropout, Manifold Mixup, adversarial feature perturbation, and related methods all impose specific forms of activation perturbation but with class-agnostic or random strategies. We conjecture that expansive perturbation (increasing activation norm) acts as positive augmentation, while contractive perturbation (decreasing activation norm) acts as negative augmentation, and that the perturbation layer determines whether the effect resembles input-level augmentation (shallow layers) or logit-level manipulation (deep layers). We propose Learning to Perturb Activations (LPA), which adaptively perturbs activations at a selected hidden layer with class-level perturbations learned via PGD. We further provide theoretical analysis connecting activation perturbation to flat minima and perturbation amplification through layers. Experiments on balanced classification, long-tail classification, and domain generalization demonstrate that LPA consistently outperforms existing methods and provides complementary benefits to logit perturbation methods such as LPL.
[LG-67] K-FinHallu: A Hallucination Detection Benchmark for Multi-Turn RAG in Korean Finance
链接: https://arxiv.org/abs/2605.29523
作者: Eunbyeol Cho,Yunseung Lee,Mirae Kim,Jeewon Yang,Youngjun Kwak,Edward Choi
类目: Machine Learning (cs.LG)
*备注:
Abstract:Large Language Models (LLMs) have advanced financial automation through Retrieval-Augmented Generation (RAG), yet hallucinations remain a critical barrier to deployment in high-stakes environments. Existing benchmarks focus on single-turn, English-centric tasks, leaving the multi-turn dynamics and linguistic-regulatory nuances of the Korean financial domain unaddressed. We introduce K-FinHallu, the first benchmark for hallucination detection in multi-turn Korean financial RAG. We construct multi-turn dialogues from authentic Korean financial documents and inject hallucinations under a proposed hierarchical taxonomy based on context answerability that explicitly accounts for justified abstention. Benchmarking frontier and open-source LLMs as hallucination detectors, we find that even the strongest models struggle with fine-grained financial diagnostics and refusal behavior. While fine-tuning an 8B model on our training split yields performance competitive with frontier LLMs, justified abstention remains the weakest axis across all evaluated models.
[LG-68] Convex Basins in Single-Index Model Loss Landscapes: Applications to Robust Recovery under Strong Adversarial Corruption ICML2026
链接: https://arxiv.org/abs/2605.29497
作者: Santanu Das,Sagnik Chatterjee,Jatin Batra
类目: Machine Learning (cs.LG)
*备注: Accepted at ICML 2026
Abstract:We study the problem of robustly learning Gaussian Single Index Models (SIMs) in the presence of heavy-tailed noise and a constant fraction of adversarially corrupted covariates and responses. Prior work on robust recovery has considered settings such as linear regression (Pensia et al., JASA 2024), strictly monotonic link functions (Awasthi et al., NeurIPS 2022), and phase retrieval (Buna and Rebeschini, AISTATS 2025). However, these techniques do not extend to generic asymmetric non-monotonic link functions such as \textscGeLU and \textscSwish, which arise naturally as scalar primitives in modern gated neural architectures. We close this gap by giving the first robust recovery algorithm with near-linear sample and time complexity for generic non-monotonic link functions, thereby establishing the first robust recovery guarantees for a broad family of nonlinear SIMs for which \textitno guarantees were previously known. Our central contribution is a new structural understanding of the Gaussian squared-loss landscape under adversarial contamination. Crucially, we prove that for a broad class of nonlinear non-monotonic SIMs, a dimension-independent, constant-radius convex basin exists around the ground truth and is efficiently reachable via robust spectral initialization even under adversarial contamination. Prior works fail to establish both guarantees simultaneously, thereby either breaking down under adversarial contamination or failing to handle generic non-monotonic link functions. Together, these structural insights yield a principled warm start for robust gradient descent that provably converges to a final estimation error of O(\sigma\sqrt\epsilon) in \tildeO(nd) time with \tildeO(d) samples, where \epsilon is the contamination fraction.
[LG-69] On-Policy Replay for Continual Supervised Fine-Tuning
链接: https://arxiv.org/abs/2605.29495
作者: Yan Chen,Taojie Zhu,Meng Zhang,Xin Chen,Jiaqi Huang,Dongyang Xu,Yizhi Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Continual supervised fine-tuning (SFT) is the de facto recipe for adapting large language models (LLMs) to a stream of downstream tasks, but it suffers from catastrophic forgetting of earlier capabilities. Recent work shows that on-policy signals – training on the model’s own outputs – reduce forgetting more reliably than off-policy supervision. Existing on-policy methods route this signal through a new training objective (e.g., self-distillation losses with a teacher copy), inheriting an extra forward pass, schedule sensitivity, and stylistic drift from the this http URL instead route the on-policy signal through the training data source. Our method, On-Policy Replay (OPR), rolls out the most recent checkpoint on a small budget of historical prompts, filters the generations by a task reward, and replays the surviving (prompt, model response) pairs as ordinary SFT examples. There is no teacher, no auxiliary loss, and no on-the-fly distillation. Across three 7–8B instruction-tuned backbones (Qwen2.5-7B-Instruct, Qwen3-8B, Llama3.1-8B-Instruct) on the TRACE continual-learning benchmark, OPR consistently reduces forgetting; on the sharpest stress test (Qwen2.5-7B-Instruct, Sequential SFT BWT -13.93), OPR lifts BWT to -0.65 at a 10% replay budget and to -2.29 at a 1% budget – a 46% reduction in |BWT| over a tuned Vanilla Replay baseline, with 42–46% reductions observed across all three backbones. We give a KL-shrinkage interpretation that places OPR and prior on-policy distillation methods on a single axis, and we present a counterintuitive finding that explains why Vanilla Replay is already a strong baseline: low-score replay is uniformly worse than Vanilla Replay, demonstrating that the active ingredient in OPR is the on-policy distribution, not the response quality this http URL code is available at this https URL.
[LG-70] Gradient Perturbation: Learning to Perturb Gradients for Adaptive Training
链接: https://arxiv.org/abs/2605.29494
作者: Hua Li
类目: Machine Learning (cs.LG)
*备注:
Abstract:Deep neural network training involves both forward propagation (from features through logits to loss) and backward propagation (from loss through gradients to parameter updates). While perturbations along the forward chain, including feature perturbation, logit perturbation, and label perturbation, have been extensively studied, the backward chain’s gradient perturbation has received little systematic investigation. In this paper, we establish a unified framework for gradient perturbation, revealing that existing methods such as Sharpness-Aware Minimization (SAM), gradient clipping, and gradient noise injection can all be interpreted as imposing specific forms of gradient perturbation. Analogous to the recently proposed Logit Perturbation Learning (LPL), we conjecture that amplifying the gradient norm for a class acts as positive augmentation (enhancing learning), while dampening it acts as negative augmentation (suppressing overfitting). Based on these observations, we propose Learning to Perturb Gradients (LPG), which adaptively perturbs logit-level gradients at the class level to achieve category-aware training. We also establish theoretical connections between gradient perturbation bounds and generalization guarantees via PAC-Bayesian analysis. Experiments on balanced classification, long-tail classification, and noisy label learning demonstrate that LPG consistently outperforms existing methods and can be combined with them as a plug-in module.
[LG-71] Access Sets Matter: Budgeting Expert Reads for Scalable Weight-Space Model Merging ICML2026
链接: https://arxiv.org/abs/2605.29489
作者: Yuanyi Wang,Yanggan Gu,Su Lu,Yifan Yang,Zhaoyi Yan,Congkai Xie,Jianmin Wu,Hongxia Yang
类目: Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注: ICML 2026 Workshop on Weight-Space Symmetries: from Foundations to Practical Applications
Abstract:Weight-space model merging is usually formulated as an algebraic operation on checkpoints, yet at LLM scale the limiting resource is often the set of expert weights that must be read. We introduce MergePipe, a budget-aware execution layer that casts LLM merging as an \emphexpert access-set problem: given a merge operator and a checkpoint family in a shared weight coordinate system, choose which expert delta blocks to access under an explicit I/O budget. MergePipe indexes parameter blocks, builds deterministic access plans, and executes the induced budgeted merge with replayable manifests. The plan is budget-sound by construction and recovers the full-read merge at full budget; for fixed-coefficient additive operators, the omitted-update error is bounded by the norm of omitted deltas. Across Qwen and Llama merging workloads, MergePipe reduces expert-read I/O by up to an order of magnitude and achieves up to 11\times speedups. Representative budget sweeps show O(10^-3) parameter deviation from full-read merges and no monotonic degradation on downstream benchmarks.
[LG-72] A Full-Pipeline Framework for Evaluating Membership Inference Attacks in Machine Learning
链接: https://arxiv.org/abs/2605.29454
作者: Ding Chen,Xinwen Cheng,Xuyang Zhong,Xinping Chen,Xiaolin Huang,Chen Liu
类目: Machine Learning (cs.LG)
*备注:
Abstract:While Membership Inference Attacks (MIAs) are the prevailing method for identifying training data, their application has expanded into privacy auditing and machine unlearning. Nevertheless, the field lacks a systematic framework for evaluating how different contexts affect MIA efficacy. Without such a characterization, practitioners risk deploying algorithms that perform well on benchmarks but become statistically irrelevant when faced with the nuances of specific, real-world datasets. To bridge this gap and provide actionable insights, we introduce a comprehensive evaluation framework that systematically characterizes privacy risks across the entire machine learning pipeline, spanning data, architectures, algorithms, and post-training modules. Designed to inherently capture diverse operational contexts, our framework rigorously evaluates state-of-the-art MIAs across a broad spectrum of training configurations. To account for varying misclassification costs in real-world deployments, we employ three complementary metrics: Balanced Accuracy for symmetric costs, alongside TPR at low FPR (or TNR at low FNR) for asymmetric scenarios where false alarms or missed detections are strictly penalized. Furthermore, recognizing that existing MIAs assume divergent adversary capabilities, we formalize two standardized threat models and adapt these attacks into corresponding variants to ensure an equitable benchmark. Extensive empirical evaluations demonstrate that the efficacy of specific MIA methodologies is highly sensitive to the assumed threat models and chosen evaluation metrics. Ultimately, we distill these findings into actionable guidelines and provide a ready-to-use auditing toolkit, empowering practitioners to conduct better privacy assessments.
[LG-73] Real-Time Retargeting Using Controllability Boundary for Chandrayaan-3 Lunar Landing
链接: https://arxiv.org/abs/2605.29412
作者: Suraj Kumar,Debjyoti Chakrabarti,Aditya Rallapalli,Bharat Kumar GVP,Ashok Kumar Kakula
类目: ystems and Control (eess.SY); Machine Learning (cs.LG)
*备注: 8 pages, 6 figures, Accepted for publication in American Control Conference 2026
Abstract:This paper presents the real-time retargeting guidance policy developed for the Chandrayaan-3 lunar landing mission. The baseline guidance generates approximate fuel-optimal descent trajectories, while a high-level policy enables safe retargeting to alternate sites when the nominal site becomes infeasible. The retargeting strategy leverages a convex representation of the controllability boundary, allowing rapid feasibility checks and real-time target updates. To the best of the authors knowledge, this represents the first application of a data-driven retargeting framework in an operational lunar landing mission. Pre-flight simulations and Chandrayaan-3 flight results validate the effectiveness of the proposed approach.
[LG-74] Information-Directed Offline-to-Online Reinforcement Learning
链接: https://arxiv.org/abs/2605.29405
作者: Keru Chen
类目: Machine Learning (cs.LG)
*备注:
Abstract:Decision-making from offline datasets typically warm-starts a policy or score model from fixed offline data and then refines it with limited online interaction. Offline data reduces uncertainty, but it does not remove the need for exploration; it changes what remains to be explored. We formalise this residual uncertainty by the conditional mutual information I(\chi;\tau_1:T\mid\mathcalD_N) between a learning target \chi and the online trajectories after conditioning on the offline dataset. This view leads naturally to information-directed sampling (IDS), a family parameterised by \eta\ge 0 that selects actions by trading off instantaneous regret against information gain. We prove a generic offline-to-online Bayesian regret bound for IDS through a ratio certificate: any information-ratio bound satisfied by a reference Thompson-sampling policy over the same randomised policy class is inherited by IDS. In a known-dynamics Bayesian linear-reward model, the conditional mutual information has a log-determinant form, and vanilla IDS ( \eta=0 ) satisfies \widetilde O!\left(Hd\min\left\sqrt T,,T\sqrtC^\dagger_\beta,\mathrmIDS_0(N,T)/N\right\right), where the coverage coefficient is tied to the visitation distribution induced by vanilla IDS itself. We also identify a warm-start regime with a dominated but informative probe in which vanilla IDS selects the probe while Thompson sampling never does, giving a constant-factor Bayesian regret separation. Controlled bandit experiments and D4RL offline-to-online RL experiments validate this mechanism: IDS is most beneficial when offline data is informative but leaves biased or low-probability residual uncertainty that targeted online actions can resolve, a regime shared by offline RL, offline black-box optimization, and Bayesian optimization.
[LG-75] Rethinking Post-Training Recipes for Multimodal Time-Series Forecasting
链接: https://arxiv.org/abs/2605.29401
作者: Haoxin Liu,Yichen Zhou,Rajat Sen,B. Aditya Prakash,Abhimanyu Das
类目: Machine Learning (cs.LG)
*备注:
Abstract:Time-Series Foundation Models (TSFMs) excel at zero-shot unimodal forecasting using numerical data, but unlike LLMs they cannot consume multimodal, non-numerical context that often shape real-world trajectories. In this work, we bridge this gap and argue for a multimodal time-series forecasting approach that post-trains LLMs to act as context-guided revisors over strong numerical TSFM priors. We introduce PostTime, a post-training recipe combining Supervised Fine-Tuning (SFT) and Reinforcement Learning with Verifiable Rewards (RLVR), along with a methodology to generate automated reasoning traces for forecast revisions. PostTime teaches an LLM to generate context-conditioned forecast interventions – decisions to revise, preserve, or ignore the TSFM prior based on the multimodal context. We evaluate this approach on the TimesX multimodal forecasting benchmark using a Gemma-3-4B LLM and TimesFM-2.5 TSFM, and show that it significantly outperforms standalone TSFMs, LLM-only baselines, and existing multimodal forecasting approaches.
[LG-76] Deep Adaptive Dimension Reduction for Bayesian Inference in Inverse Problems
链接: https://arxiv.org/abs/2605.29373
作者: Yueyang Wang,Xili Wang,Kejun Tang,Xiaoliang Wan,Tao Zhou,Chao Yang
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注: 25 pages, 5 figures
Abstract:Solving high-dimensional PDE-governed inverse problems is often challenging due to complex non-Gaussian posterior distributions, expensive forward model evaluations, and misspecified prior information. To address these issues, we propose a deep adaptive dimension-reduction Bayesian inference framework based on the Variational Flow (VF) model. Since standard normalizing flows are restricted by bijective mappings and cannot directly reduce dimensions, VF overcomes this limitation by integrating VAE-based nonlinear dimension reduction with dual normalizing flows for the latent prior and encoder. This design provides a strictly higher evidence lower bound than VAE and allows more flexible approximation of complex posterior distributions. We further introduce an iterative prior updating strategy that gradually moves the prior mean toward high-probability posterior regions, avoiding manual prior tuning. These components form a closed adaptive loop together with an adaptively fine-tuned Fourier Neural Operator (FNO) surrogate: VF generates posterior-concentrated samples to refine the surrogate, while the updated surrogate further improves posterior inference. Numerical experiments on a 100-dimensional Rosenbrock problem and three standard PDE-governed inverse problems show that our method delivers competitive or superior accuracy compared with MCMC, UKI, and SVGD baselines across all tested configurations, with the most pronounced advantages emerging in challenging scenarios such as high-noise observations and high-dimensional parameter spaces.
[LG-77] Solving Integer Linear Programming with Parallel Tempering
链接: https://arxiv.org/abs/2605.29366
作者: Kyuil Sim,Sanghyeok Choi,Jinkyoo Park
类目: Machine Learning (cs.LG)
*备注: Preprint. Code available at this https URL
Abstract:Integer Linear Programming (ILP) serves as a versatile framework for modeling a wide range of combinatorial optimization problems, typically addressed by sophisticated exact solvers or heuristics. While learning-based approaches have recently shown their effectiveness, they suffer from poor generalization to out-of-distribution instances and inherent dependence on external solvers. In this work, we propose a solver-free, sampling-based optimization framework for ILP that directly explores discrete feasible regions without training or external solvers. Exploiting the linear structure of ILP, we employ a Locally-Balanced Proposal to construct a transition kernel, thereby avoiding the gradient approximation. To overcome the highly multimodal nature of ILP energy landscapes, we integrate Parallel Tempering. In addition to standard temperature tempering, we introduce penalty tempering, which modulates constraint barriers while preserving the objective landscape over feasible solutions. Empirically, our method consistently outperforms SCIP across all four benchmarks, matches or exceeds Gurobi on two of four tasks within a 200-second budget, and is substantially more robust to distribution shift than learning-based methods. Furthermore, on MIPLIB 2017 instances, our framework remains competitive with classical solvers without any problem-specific tuning.
[LG-78] Neural-Behavioral Representation of Natural Whole-body Movement in Monkeys
链接: https://arxiv.org/abs/2605.29355
作者: Jieshi He,Puzhe Li,Yanan Sui,Mu-ming Poo
类目: Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注:
Abstract:Understanding how cortical activity represents natural whole-body behaviors in primates remains challenging. Limited by the diversity of movements and inaccessibility of large-scale neural representation of whole-body kinematics, previous motor decoding studies focused on constrained tasks and limited limb movements. Here, we present a neural-behavioral recording and modeling framework for freely moving monkeys, combining large-scale epidural cortical signals from distributed sensory- and motor-related areas with synchronized multi-view motion capture through a custom-made data collection platform. We reconstructed whole-body monkey kinematics and learned a compact behavior prior using an autoregressive encoder-decoder model. Conditioned on neural signals, the model decoded accurate and realistic whole-body movement without explicit physical constraints. Our results provide a novel proof-of-concept approach for decoding natural whole-body movements in primates using large-scale intracranial neural activity.
[LG-79] Harmless Yet Harmful: Neutral Prompting Attacks for Stealthy Hallucination Steering in Agent Skills
链接: https://arxiv.org/abs/2605.29354
作者: Chia-Yi Hsu,Chia-Mu Yu,Chun-Ying Huang,Jun Sakuma
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: under review
Abstract:LLM-powered coding agents increasingly participate in software development workflows by generating code, selecting dependencies, and producing package installation commands. This creates a new software supply chain risk: when an agent hallucinates a non-existent package, an attacker may register the hallucinated name and later compromise users who install it. Existing package hallucination attacks and defenses primarily focus on naturally occurring hallucinations, targeted dependency steering, or post-hoc package validation. In this paper, we introduce \emphNeutral Prompting Attack (NPA), a highly stealthy attack paradigm in which semantically benign instructions, such as encouraging imagination and exhaustiveness, increase package hallucination propensity without containing explicit malicious intent. Unlike targeted dependency steering, NPA does not specify an attacker-chosen package. Instead, it shifts the model’s dependency generation behavior toward more speculative package names. We evaluate NPA across multiple coding-oriented LLMs and package hallucination benchmarks. Our results show that NPA increases both \emphHallucination ASR and \emphPip Install ASR, changes the distribution of hallucinated package names, and evades existing static-analysis, LLM-based, and agent-based Skill defenses. These findings reveal that harmless-looking prompts can covertly manipulate hallucination behavior and create downstream software supply chain risks.
[LG-80] Attention as In-Context Empirical Bayes: A Two-Stage View via Particle Dynamics
链接: https://arxiv.org/abs/2605.29351
作者: Matthew Smart,Soumya Ganguly,Nilava Metya,Alexandre V. Morozov,Anirvan M. Sengupta
类目: Machine Learning (cs.LG); Dynamical Systems (math.DS); Machine Learning (stat.ML)
*备注: 52 pages, 5 figures
Abstract:We study minimal attention-only transformers under all-token corruption and show they admit a two-stage empirical Bayes interpretation. A single attention step computes a kernel-weighted posterior mean with respect to the empirical distribution defined by the context. Depth refines this distribution through particle dynamics (Stage 1), while a long-range skip-connection carries the noisy input as a query for posterior inference (Stage 2), revealing distinct statistical roles for depth and attention residuals. The framework isolates a minimal setting in which the context itself induces a depth-dependent energy landscape governing in-context inference. We show that effective denoising can emerge without an explicit noise schedule: a fixed kernel bandwidth and finite integration horizon suffice, yielding a principled depth-noise relationship. We further establish a posterior-mean recovery guarantee for a class of well-behaved priors, where the empirical estimator converges to the Bayes-optimal predictor under asymptotic conditions. Connecting these dynamics to reverse-diffusion limits, our results provide a statistical interpretation of attention as in-context inference via sample-based posterior estimation, without explicit density modeling.
[LG-81] NeuroEdge: Real-Time Hand Gesture Recognition with High-Density EMG Using Deep Learning at the Edge
链接: https://arxiv.org/abs/2605.29326
作者: Peter Chudinov,Zhenyu Lin,Jay Motamarry,Srihita Panati,Xiaorong Zhang,Zhuwei Qin
类目: Machine Learning (cs.LG)
*备注:
Abstract:High-density electromyography (HD-EMG) has emerged as a powerful modality for decoding fine-grained neuromuscular activity, enabling real-time neural-machine interfaces (NMIs) for applications such as prosthetic control, rehabilitation, and augmented interaction. While deep learning approaches such as convolutional neural networks (CNNs)have demonstrated high classification accuracy for EMG-based gesture recognition, their deployment on embedded hardware remains a major challenge due to computational and memory constraints. This paper presents NeuroEdge, a real-time HD EMG-based NMI system that performs gesture recognition entirely on resource-constrained microcontrollers. The system features two custom-designed modules: the HD-EMG StreamBridge, a wireless communication interface that streams raw HD-EMG data from a Quattrocento amplifier to an ESP32 microcontroller; and the EdgeDL Inference Engine, a lightweight deep learning framework executing on a Sony Spresense microcontroller. A compact 1-dimensional CNN optimized for embedded inference processes, sliding windows of EMG data in real time. Data streaming and inference are pipelined and synchronized through an architecture that utilizes Direct Memory Access (DMA) for data transfer and Serial Peripheral Interface (SPI) burst communication between the ESP32 and Spresense, ensuring low-latency performance. Experimental results show that NeuroEdge achieves a real-time classification accuracy of 90% across seven hand gestures, with a total average latency of 83 ms using 192 channels of HD-EMG recorded from the forearm. Our system demonstrates the feasibility of deploying complex HD-EMG-based gesture recognition on microcontroller-based edge devices, bridging the gap between high-resolution biosignal acquisition and deep learning-based embedded inference for next-generation NMIs.
[LG-82] A Theoretical and Experimental Study of a Novel Adaptive Learning Algorithm
链接: https://arxiv.org/abs/2605.29273
作者: Sakshi Kumari,Shyam Kumar M,Sushmitha P
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:A crucial component of machine learning algorithms is minimizing loss functions with less computational cost and less oscillations. While adaptive learning rate-based optimizers have been widely used for real-world tasks, they do not guarantee convergence, which is why AMSGrad was later introduced to investigate the non-convergence behaviour of Adam. In this paper, popular adaptive optimization methods like Adam and AMSGrad are critically reviewed with an emphasis on their fundamental design concepts. To address limitations of the above mentioned optimizers, a new optimizer variant, C-Adam, is proposed based on the line of sight approach. A theoretical proof for convergence is also provided and the optimizer is validated through a number of real-life based numerical experiments.
[LG-83] Robust Frequency-Calibrated Virtual EEG Channel Generation from Four Frontal Electrodes for Wearable EEG Augmentation
链接: https://arxiv.org/abs/2605.29263
作者: Minghao Xiao
类目: Machine Learning (cs.LG)
*备注: 17 pages, 4 figures
Abstract:Low-channel wearable electroencephalography (EEG) is attractive for long-term monitoring, but four frontal electrodes provide only a sparse and spatially biased view of distributed scalp activity. We present FAVC-Net, a compact frequency-calibrated virtual-channel network that generates 13 unmeasured EEG channels from Fp1, Fp2, F7, and F8. The model combines shared multi-scale source encoding, source-state embeddings, target-conditioned signed source-block mixing, GATv2-based attention refinement, attention-consistent skip fusion, and weak Welch power spectral density calibration. Rather than treating sparse-to-dense EEG generation as a purely waveform-matching task, the framework jointly emphasizes amplitude fidelity, spectral allocation, channel-frequency texture, and robustness to corrupted wearable inputs. On the PRED+CT dataset, FAVC-Net achieved the best joint waveform-spectral operating point among neural and interpolation baselines. Its time-domain gains were modest, whereas log-spectral distance and PSD KL divergence were reduced by 30.09% and 37.98% relative to the strongest non-FAVC comparator. Under wearable-like source perturbations, the model preserved spectral fidelity and resisted spectral collapse. These results support virtual EEG channel generation as a dual-domain augmentation problem, while emphasizing that generated posterior and parietal channels should be interpreted as frequency-calibrated representations derived from sparse frontal measurements rather than as independent physical recordings.
[LG-84] SigmaMedStat: Temporal Signal Modeling for ICU False Alarm Reduction
链接: https://arxiv.org/abs/2605.29236
作者: Arunkumar Ramachandran
类目: Machine Learning (cs.LG)
*备注: Code available at this http URL
Abstract:Alarm fatigue in intensive care units (ICUs) is a well documented patient safety crisis. Clinical monitors generate 350 or more alarms per patient per day, out of which 72-99% are clinically irrelevant. Staff desensitization to non-actionable alarms increases the risk of missed true emergencies. This paper presents SigmaMedStat, a machine learning system that evaluates the trustworthiness of physiological alarm signals before clinical action is taken. Four approaches were evaluated on the PhysioNet/Computing in Cardiology Challenge 2015 dataset of 498 four-channel ICU alarm recordings. Primary contribution is a temporal modeling framework that splits each 60 second recording into six consecutive 10-second chunks, and this in turn generates Continuous Wavelet Transform (CWT) scalograms per chunk, encodes each chunk with a shared EfficientNet-B0 encoder, and passes the resulting feature sequence to a two-layer Long Short-Term Memory (LSTM) network. Five-fold stratified cross-validation yields a mean AUC of 0.822 +/- 0.016 (95% CI: [0.790,0.853]), compared to 0.641 for a static EfficientNet baseline trained on the full 60-second window. Ablation studies confirm that temporal chunking and multi-channel signal fusion both contribute independently to classification performance. Per-alarm type analysis reveals that Ventricular Flutter is the most accurately classified alarm type (AUC 0.820) while Asystole remains the hardest (AUC 0.722). Error analysis identifies 65 false negatives and 85 high-confidence misclassifications as the primary failure modes. All code and results are publicly available at this https URL.
[LG-85] raditional machine learning vs. deep learning from dynamic graph representations of proteins 3D folds in the task of protein structure classification
链接: https://arxiv.org/abs/2605.29228
作者: Aydin Wells,Francis A. Gatsi,Aaron Striegel,Tijana Milenković
类目: Machine Learning (cs.LG); Molecular Networks (q-bio.MN)
*备注: Main paper: 16 pages, 4 figures, and 1 table; Supplementary information: 13 pages, 9 figures
Abstract:Protein structure classification (PSC) uses supervised learning to predict a protein’s CATH/SCOP(e) class from the protein’s sequence or 3D structural feature(s). We already modeled 3D structures as (static) protein structure networks (PSNs), demonstrating the competitiveness of PSN-based features to sequence or direct (i.e. non-network) 3D structural features in the PSC task. More recently, we demonstrated the power of features extracted from dynamic PSNs over features extracted from static PSNs (and thus by transitivity over sequence and direct 3D structural features) in the same task. That dynamic PSN approach used traditional machine learning (ML), combining manual (pre-engineered) features with an off-the-shelf classifier. Here, we evaluate whether automatic deep learning (DL) from the dynamic PSNs yields improvements. Our evaluation on 72 datasets spanning ~44,000 CATH- or SCOPe-labeled dynamic PSNs reveals that in terms of PSC accuracy, traditional ML and DL are (close to) tied for a large majority of the datasets, while DL is on average 10+ times slower. We are the first to evaluate traditional ML vs. DL in the dynamic PSN-based PSC task.
[LG-86] Inferring the Size of Large Language Models From Popular Text Memorization
链接: https://arxiv.org/abs/2605.29223
作者: Ivica Nikolic
类目: Machine Learning (cs.LG)
*备注:
Abstract:The parameter counts of the most widely used large language models (LLMs) are often withheld by their developers, leaving model size – a primary reference point for interpreting capabilities and costs – largely undisclosed. We propose a black-box method to infer conservative lower bounds on LLM size from generated text outputs alone, requiring nothing beyond the ability to submit text fragments and observe next-token predictions. Our approach is grounded in a key observation: popular, widely-circulated texts – such as classical literature, religious texts, and foundational documents – are present in virtually every large-scale pretraining corpus, and how accurately a model predicts the next word across text fragments of varying length is a reliable signal of how much it has memorized them, which in turn is fundamentally limited by its total parameter count. We aggregate this memorization signal across a diverse corpus of texts and fragment lengths into a single accuracy profile vector per model, and build two complementary inference methods on top of it: a pairwise statistical test that determines which of two models is larger, and a scaling-law estimator that extracts a one-dimensional latent index from these vectors via Principal Component Analysis (PCA) to map the aggregated signal to a parameter count. Validated on a broad set of open-weight models, both methods produce accurate and reliable lower bounds. When applied to popular closed-weight models, our framework recovers internal product hierarchies and reveals a clear divergence in industry scaling strategies: while some developers yield significantly higher bounds indicative of large generational parameter growth, others operate under strict parameter ceilings, demonstrating that hidden design choices can be systematically probed even under strict API limitations.
[LG-87] bhmm: A Modern C20 Library for Hidden Markov Models with Correct MLE Emission M-Steps
链接: https://arxiv.org/abs/2605.29208
作者: Gary Wolfman
类目: Mathematical Software (cs.MS); Machine Learning (cs.LG)
*备注: 17 pages, 3 figures, 8 tables
Abstract:We describe libhmm, a C++20 library for Hidden Markov Model parameter estimation, sequence decoding, and model selection. libhmm addresses two gaps in existing software: the absence of a well-maintained, zero-dependency C++ HMM library suitable for embedding in production systems, and the widespread use of method-of-moments (MOM) approximations in the emission distribution M-step of the Baum-Welch algorithm. The library implements correct maximum likelihood estimators for sixteen continuous and discrete emission distributions, including an ECME algorithm for the location-scale Student-t distribution, Newton-Raphson maximization for Gamma, Beta, Weibull, and Negative Binomial distributions, and the von Mises distribution for circular data. All forward-backward and Viterbi calculations operate in full log-space. SIMD acceleration is provided for AVX-512, AVX2, SSE2, and ARM NEON via compile-time dispatch with scalar fallback. Python bindings are available via the companion package pylibhmm. We compare libhmm against established C and C++ HMM libraries and against published R reference packages on five real-data benchmarks, and discuss the architectural tradeoffs made in the design.
[LG-88] Auditing Training Data in Generative Music Models via Black-Box Membership Inference
链接: https://arxiv.org/abs/2605.29202
作者: Yi Chen Liu,Jiawei Yu,Kexin Cao,Syed Irfan Ali Meerza,Trishika Movva,Jian Liu
类目: Machine Learning (cs.LG)
*备注: The paper has been accepted for presentation at the workshop ArtSec 2026: Workshop on Artwork Security and Provenance in the Age of AI
Abstract:Recent advances in text-to-music generation enable high-fidelity synthesis of structured musical audio, raising growing concerns about data provenance, consent, and training transparency. These models are typically trained on large-scale corpora with little disclosure, leaving no practical mechanism to verify whether a particular audio sample was included in training. In this paper, we investigate black-box membership inference for generative music models, aiming to determine whether a candidate music sample was used during training, given only query access to the deployed system. Our key insight is that training membership induces systematically stronger semantic and structural alignment between a candidate sample and the model’s generation conditioned on its caption. We query the target model with the associated caption and measure the relationship between the candidate audio and the generated output in a learned feature space. To capture features that separate members from non-members, we construct paired examples consisting of each track and its caption-conditioned generation from shadow models, and train a music auditor to classify membership. The auditor captures alignment patterns characteristic of training membership and generalizes to unseen target models in a fully black-box setting without access to model parameters or training metadata. Across multiple state-of-the-art music generators, our method achieves up to 98.6% accuracy, with false-positive and false-negative rates as low as 1.9% and 1.0%, demonstrating that reliable training-data auditing is feasible in realistic deployment scenarios.
[LG-89] Probabilistic bias adjustment of seasonal forecasts using generative machine learning: A case study of Arctic sea ice predictions
链接: https://arxiv.org/abs/2605.29172
作者: Parsa Gooya,Reinel Sospedra-Alfonso
类目: Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph)
*备注:
Abstract:Seasonal climate predictions support planning and risk management by offering early information of the most likely-to-occur climate conditions in the coming months, and associated uncertainties. Ensemble forecasts enable this by simulating many plausible outcomes, allowing predictions to be expressed as usable probabilities. Large ensembles and high-resolution forecasts strengthen this guidance by better sampling uncertainty and capturing finer-scale processes but come with significant computational cost. Moreover, forecast ensembles drift and exhibit systematic biases and spatio-temporal errors that grow with lead time, requiring careful post-processing and calibration. A probabilistic post-processing framework based on conditional Variational Autoencoders (cVAEs) was developed at the Canadian Center for Climate Modeling and Analysis to generate large ensembles of bias adjusted seasonal predictions of Arctic sea ice. The generative model was designed to learn the observational distribution conditioned on the biased model prediction. This enables generation of arbitrarily large ensembles of well-calibrated, bias corrected forecasts with improved skill. Here, we extend this framework to address the loss of fine-scale energy and the characteristic blurriness in predictions, a known limitation of standard cVAEs. Specifically, we employ a generator in place of the Gaussian parametrized decoder in the cVAE and use Continuous Ranked Probability Score in the objective function instead of the Mean Square Error. We further use a higher resolution target dataset compared to the raw forecast. We show that the adjusted forecasts are better calibrated, more consistent with the observational distribution, and exhibit smaller errors than benchmark predictions, while also enhancing the resolution of the raw forecasts and improving sharpness and spectral power relative to the standard cVAE.
[LG-90] Do Deep Networks Forget Initialization? A Forgetting-Time View of Practical Inductive Bias
链接: https://arxiv.org/abs/2605.29152
作者: Mohua Das,Pierfrancesco Beneventano,Shibshankar Dey,Gareth H. McKinkey,Tomaso Poggio
类目: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
*备注: 39 pages, 9 figures
Abstract:Randomly initialized neural networks induce a prior over functions, but the predictor used in practice is produced only after training. We ask how much of this initial bias survives the training pipeline. To make the question measurable, we introduce initialization memory: the dependence of the validation-selected predictor on the scale of the random initialization. We perform controlled CIFAR-10 experiments on ResNets where initialization memory already sharply separates training regimes. Low-learning-rate SGD can interpolate while still remembering its initialization: on ResNet-9 with batch size b=128 , test accuracy varies by 26.5 percentage points across initialization scales despite \ge99.5% training accuracy. This is not undertraining: extending the same low-learning-rate regime to 5,000 epochs leaves the spread essentially unchanged. In contrast, Adam-family methods largely erase the dependence. SGD can also be made to forget when larger learning rates are paired with explicit L_2 norm control. We interpret these findings in terms of the time scale of forgetting: gradient-flow-like dynamics can preserve initialization memory, whereas stochastic finite-step effects, explicit norm decay, and adaptive preconditioning erase it on scales governed by the size of explicit or implicit regularization. The practical inductive bias of a trained network is therefore not the architectural prior alone, but the architectural prior after being filtered by the forgetting dynamics of the training pipeline; and the same regularizers that improve generalization are precisely those that erase memory of initialization.
[LG-91] Optimal Gap-Dependent Regret for Private Stochastic Decision-Theoretic Online Learning
链接: https://arxiv.org/abs/2605.29148
作者: Tommaso Cesari,Roberto Colomboni
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We study stochastic decision-theoretic online learning with full information and event-level pure differential privacy. A COLT open problem of Hu and Mehta asks to determine the optimal gap-dependent regret rate for stochastic decision-theoretic online learning under pure event-level differential privacy. For K actions, losses in [0,1] , and a unique best action separated from the second-best action by gap \Delta_\min , the known lower bound is of order \frac\log K\min\Delta_\min,\varepsilon, or equivalently, up to universal constants, of order [ \frac\log K\Delta_\min+\frac\log K\varepsilon. ] We give a horizon-free pure-DP algorithm and prove the explicit regret bound [ \operatornameReg_T \le 1000 \cdot \left(\frac\log K\Delta_\min+\frac\log K\varepsilon\right) ] for every horizon T . The numerical constant is not optimized. The algorithm partitions time into blocks of exponentially increasing size, plays a single action throughout each block, and chooses the next action by an exponential mechanism applied to a data-independent random prefix of the previous block. The random prefix converts block regret into a sum, over all prefix lengths, of softmax selection errors. A single entropy-potential argument controls all privacy-dominated large-gap actions at cost \log K/\varepsilon . Subjects: Machine Learning (cs.LG); Machine Learning (stat.ML) Cite as: arXiv:2605.29148 [cs.LG] (or arXiv:2605.29148v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2605.29148 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Roberto Colomboni [view email] [v1] Wed, 27 May 2026 22:17:00 UTC (14 KB) Full-text links: Access Paper: View a PDF of the paper titled Optimal Gap-Dependent Regret for Private Stochastic Decision-Theoretic Online Learning, by Tommaso Cesari and 1 other authorsView PDFHTML (experimental)TeX Source view license Current browse context: cs.LG prev | next new | recent | 2026-05 Change to browse by: cs stat stat.ML References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) IArxiv recommender toggle IArxiv Recommender (What is IArxiv?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); About Help contact arXivClick here to contact arXiv Contact subscribe to arXiv mailingsClick here to subscribe Subscribe Copyright Privacy Policy Web Accessibility Assistance arXiv Operational Status
[LG-92] Apertus LLM Family Expansion via Distillation and Quantization
链接: https://arxiv.org/abs/2605.29128
作者: Andrei Panferov,Davit Melikidze,Martin Jaggi,Dan Alistarh
类目: Machine Learning (cs.LG)
*备注:
Abstract:The wide adoption of LLMs has led to their use in great variety of applications and scenarios, such as chatbot assistants and data annotation, creating the need for the models to satisfy certain budget and hardware constraints. This has led to the trend of LLMs being released in batches consisting of similar models of various sizes for the family of models to adhere to as wide of a range of constraints as possible. In this paper, we validate distillation and quantization as a cost-effective way to expand model families to new sizes and hardware formats. Based on the open-recipe Apertus 8B LLM, we produce Apertus-v1.1 - a distilled family of models with up to 4B parameters trained on 1.7T permissive license tokens. We demonstrate cost-efficiency and strong accuracy performance of our approach for covering large ranges of hardware and systems requirements.
[LG-93] Reason Break: Probing Vulnerabilities in Reasoning -Enabled Vision-Language-Action Models for Autonomous Driving
链接: https://arxiv.org/abs/2605.29114
作者: Mohammadreza Teymoorianfard,Jean-Philippe Monteuuis,Jonathan Petit,Amir Houmansadr
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG); Robotics (cs.RO)
*备注:
Abstract:Vision-Language-Action (VLA) models with integrated reasoning have been proposed for end-to-end autonomous driving, assuming a tight coupling between reasoning and trajectory generation. However, the robustness of such systems under realistic input perturbations remains largely unexplored. We show that these models are highly vulnerable to realistic input perturbations, achieving up to 89% attack success rate (ASR) on reasoning and up to 72% on trajectory manipulation in closed-loop simulation, leading to increased collision rates and degraded safety metrics. Using NVIDIA’s recent Alpamayo models as representative industry-developed VLAs, we conduct the first systematic black-box study of reasoning-enabled VLA models under realistic textual input corruptions, evaluating their impact on reasoning and driving behavior. We introduce a reasoning-aware evaluation framework capturing both semantic and structural aspects of reasoning, along with safety-centric measures. We also introduce a benchmark for evaluating attacks and defenses on reasoning-trajectory interactions in autonomous driving. Our results highlight the need for rigorous evaluation and improved defenses to ensure the safety of reasoning-enabled VLA systems in autonomous driving.
[LG-94] Bridging Chemists and AI: An Expert-Augmented Framework for Interpretable Route Evaluation
链接: https://arxiv.org/abs/2605.29108
作者: Yujia Guo,Mikhail Kabeshov,Tat Hong Duong Le,Samuel Genheden,Marco V. Mijangos,Varvara Voinarvoska,Giulia Bergonzini,Ola Engkvist,Samuel Kaski
类目: Machine Learning (cs.LG)
*备注: 13 pages, 11 figures, ELLIS Unconference Workshop: Generative Models, LLMs, and the Future of Molecular AI (ML4Molecules 2025)
Abstract:Selecting efficient multi-step synthetic routes is a central challenge in organic synthesis, particularly in medicinal and process chemistry, where route choice directly impacts feasibility, cost, and development efficiency. Data-driven assessment systems often oversimplify the multi-objective nature of synthesis design and rely on proxy datasets, such as patent routes, rather than universally grounded criteria. To address this, we introduce an expert-augmented, data-driven scoring framework that integrates machine learning with chemists’ domain knowledge for both numerical and explainable route assessment. A DeepSets-based model is trained using tree edit distance between reference and machine-generated routes, and then fine-tuned with expert evaluations to produce both quantitative scores and interpretable qualitative categories: Good, Plausible, and Bad. The resulting system achieves a Spearman correlation coefficient of 0.78 and a Pearson correlation of 0.77 for category assessment prediction, and 60.2% top-1 ranking accuracy for score prediction, substantially outperforming the previous baseline of 17.5%.
[LG-95] Model Merging by Output-Space Projection
链接: https://arxiv.org/abs/2605.29101
作者: Bethan Evans,Benjamin Etheridge,Stephen Roberts,Jared Tanner
类目: Machine Learning (cs.LG); Information Theory (cs.IT)
*备注:
Abstract:Model merging combines fine-tuned checkpoints into a single multi-task model without retraining. Existing methods - such as task arithmetic, model soups, TIES, and DARE - are computationally efficient and empirically successful, but rely on heuristic design choices and lack formal optimality guarantees. We show that merging can be formulated as a convex quadratic programme over residual updates, yielding weights that minimise a squared-output calibration objective using calibration inputs and fine-tuned model outputs, and subsuming existing methods as special cases. Our framework yields a closed-form diagnostic - the fraction of residual energy captured by a chosen basis - that predicts downstream merge quality using only the calibration set. Empirically, the QP matches or outperforms existing methods in the single-layer setting, and we characterise when the optimal basis provides significant gains over the cheaper diagonal QP. We extend to multi-layer merging via a sequential layer-wise algorithm and demonstrate consistent gains across language and vision benchmarks.
[LG-96] Knowledge Offloading: Decomposing LLM s into Sparse Backbones and Memory Modules
链接: https://arxiv.org/abs/2605.29075
作者: Karim Galliamov,Rochelle Choenni,Ivan Titov
类目: Machine Learning (cs.LG)
*备注:
Abstract:LLMs encode both general capabilities and domain-specific knowledge in a single set of parameters. We ask whether this capacity can be reorganized: keeping broadly useful computation in a shared backbone, while moving specialized knowledge into external memory modules. We propose \emphknowledge offloading (KOFF), a framework for decomposing a pretrained LLM into a sparse shared backbone and domain-specific memories. Starting from a frozen base model, we jointly learn a structured pruning mask and lightweight recovery modules, implemented as LoRA adapters and learned key-value caches. Across Llama and Qwen models from 3B to 8B, we find that non-trivial capacity can be moved out of the shared backbone without a large loss in model ability. At around 12% global sparsity, KOFF preserves much of the unpruned model’s performance, while pruning the same frozen model without memories degrades sharply. Ablations show that LoRA and learned KV memories are complementary, and specialization analyses suggest that the learned decomposition is meaningful: language-specific neurons are preferentially removed while language-general neurons largely remain in the backbone. These results suggest that knowledge can be reallocated between a shared core and swappable external memories.
[LG-97] Ensemble Score Filtering for Real-Data Energy Consumption Forecast Correction
链接: https://arxiv.org/abs/2605.29072
作者: Ruoyu Hu,Dahai Yu,Feng Bao,Guang Wang,Guannan Zhang
类目: Machine Learning (cs.LG); Numerical Analysis (math.NA)
*备注:
Abstract:Accurate estimation and forecasting of energy consumption are important for power-system operation, planning, and demand-side management. In practice, however, complete and timely measurements may not always be available, and the observed data can be partial, noisy, or delayed. This motivates the use of learned forecasting models for predicting the evolving consumption state, together with data assimilation methods for sequential forecast correction. In this work, we study a high-dimensional data assimilation problem for real energy-consumption data. \modeltextThe forward prediction is supplied by a pretrained black-box spatio-temporal forecasting model, which is treated as the state propagator in the filtering procedure. We employ the Ensemble Score Filter (EnSF) to assimilate partial and noisy observations and to correct the forecast trajectory over time. The EnSF uses score-based diffusion models to approximate filtering distributions and avoids retraining neural-network score models during assimilation by using a closed-form score representation and Monte Carlo approximation. Numerical experiments demonstrate that open-loop propagation of the learned forecasting model can become unreliable over long horizons, while EnSF-based correction substantially improves state estimation. Comparisons with the Ensemble Kalman Filter (EnKF) further show that EnSF provides stronger correction under the nonlinear observation setting considered in this work.
[LG-98] Parallel Adaptive Multi-Objective Evolutionary Learning of Discretized Bayesian Network Classifiers for Clinical Data
链接: https://arxiv.org/abs/2605.29058
作者: Damy M.F. Ha,Tanja Alderliesten,Peter A.N. Bosman
类目: Machine Learning (cs.LG)
*备注:
Abstract:Bayesian Networks (BNs) are of interest from an explainable AI viewpoint, offering transparent probabilistic models for decision support. Baymex is a recently introduced multi-objective evolutionary algorithm for learning discretized BNs, enabling experts to trade-off different objectives of interest, such as likelihood, model complexity, and prior beliefs. While Baymex has been shown to outperform state-of-the-art BN learning approaches, Baymex still 1) requires a lot of computation time and 2) has only been evaluated on synthetic data. To improve scalability, we introduce a parallelization strategy as well as a mechanism that enables adaptively steering optimization toward networks that overfit less. We furthermore reconfigure Baymex to train a BN classifier through multi-objective optimization of cross-entropy loss and the BIC complexity term so as to evaluate its performance on real-world clinical classification tasks. Besides observing speedups up to over 54 times on a 16-core CPU, comparisons against clinically familiar baselines (decision trees, logistic regression, naive Bayes, and random forests) on two open-source (RADCURE and SUPPORT) and one in-house dataset, show that Baymex obtains statistically similar or better predictive performance while producing compact, clinically inspectable BNs. Importantly, Baymex finds multiple plausible BN classifiers that contain predictors consistent with established clinical factors.
[LG-99] Moment Matching Q-Learning ICML2026
链接: https://arxiv.org/abs/2605.29033
作者: Yiyan(Edgar)Liang,Sifei Liu,Weitong Zhang
类目: Machine Learning (cs.LG)
*备注: 23 pages, 14 figures, 10 tables, accepted by ICML 2026
Abstract:Score-based and flow-based generative models exhibit remarkable expressive capacity in capturing complex distributions, and have been extensively deployed in tasks ranging from image generation to reinforcement learning. Nevertheless, these models suffer from prolonged inference latency, which imposes a significant computational bottleneck in RL with iterative sampling. To overcome this limitation, we propose a new framework named Moment Matching Q-Learning (MoMa QL), which utilizes a technique from statistical hypothesis testing known as maximum mean discrepancy (MMD) that intend to match all orders of statistics between the original and target distribution. By enforcing strong regularization on all moment statistics, this algorithm guarantees distribution-level convergence for conditional score function and remains stable under various hyperparameters. Empirically, we show that our method MoMa QL is more computationally efficient with a comparable if not competitive performance in various D4RL tasks. Remarkably, by accelerating the action sampling process for flow-based policies, MoMa QL demonstrates superior performance in offline-to-online RL tasks because of faster and stronger adaptability for online interactive finetuning.
[LG-100] heoretical Foundations and Effective Algorithms for Policy-Aware Simulator Learning
链接: https://arxiv.org/abs/2605.29032
作者: Christoph Dann,Yishay Mansour,Mehryar Mohri
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Model-based reinforcement learning (MBRL) agents typically learn world models by minimizing predictive loss. However, powerful RL optimizers inevitably exploit minor model inaccuracies, leading to simulator exploitation and a reality gap where policies succeed in simulation but fail in the real world. We propose that the objective for learning simulators should be strategic robustness rather than predictive accuracy, and formulate this as a zero-sum minimax game between a model player and an adversarial policy player. We provide a comprehensive theoretical analysis: (1) an online learning guarantee showing the game is learnable with sublinear regret bounds; (2) a tractable critic-based simplification bounding the global policy-value gap by the local critic’s loss; and (3) an Error-MDP duality, proving that finding the worst-case policy is formally dual to a standard RL problem where the reward is the one-step critic error. This duality yields a provably convergent active data selection algorithm. Experiments on continuous control tasks demonstrate that our approach reduces prediction error in strategically important regions by 1.5 - 2.2\times and enables policies trained purely in simulation to match near-optimal real-world performance.
[LG-101] Designing Active Tether-Net Systems for Space Debris Capture with Graph-Learning-Aided Mixed-Combinatorial Optimization
链接: https://arxiv.org/abs/2605.29021
作者: Feng Liu,Achira Boonrath,Gishnu Madhu,Eleonora M. Botta,Souma Chowdhury
类目: Machine Learning (cs.LG)
*备注: Accepted for presentation at 2026 AIAA Aviation Forum
Abstract:Active tether-net systems are a promising solution for capturing large non-cooperative targets, such as space debris, by deploying a flexible net manipulated by maneuverable units (MUs). However, concurrent systematic explorations of design and control choices of the tether-net system to understand its full potential remain limited, partly due to the complex, constrained, nonlinear optimization problem that it presents – one that involves a mixture of continuous, integer and categorical variables, with the latter two arising from net connectivity and component choices, respectively. Classical binary encoding methods are often ineffective for solving highly nonlinear and multimodal Mixed Combinatorial Nonlinear Programmings (MCNLPs) in engineering design, while integer coding approaches can introduce spurious relations among combinations. Given the graph-structured characteristics of the combinatorial space, this paper adopts and extends a new graph-learning-aided optimization approach to solve this MCNLP problem. Here, a Graph Neural Network (GNN) is trained to score (as output) and thereof recommend candidate combinations represented as nodes in a graph, with the continuous variable vector portion of a candidate design given as input. As a result, the MCNLP optimization reduces to an NLP, which can be solved using standard solvers. While this reduction approach is agnostic to the choice of the NLP solver, here a state-of-the-art Particle Swarm Optimization (PSO) algorithm with gradient-based fine-tuning is used as the solver. Demonstrated on the problem of concurrently designing the morphology of the net, choice of mass and thrusters in the MUs and aiming points used by the controller of the tether-net system, the GNN-based recommender is shown to provide significantly faster convergence to similar optimal solutions, compared to direct solution of the MCNLP problem.
[LG-102] Causal Intelligence for Constraint-Aware Intervention Design to Induce State Transitions
链接: https://arxiv.org/abs/2605.29008
作者: Zixuan Song,Uwe Mueller,Dimitris V. Manatakis
类目: Machine Learning (cs.LG)
*备注:
Abstract:Driving a system from one state to another through targeted interventions is a fundamental challenge in science, yet most predictive models offer limited mechanistic insight and no principled framework for decision-making. Here we present COAST (Causally Optimal Actions for State Transitions), a causal-intelligence approach for the in-silico design of constrained interventions that induce user-defined state transitions. Given data characterizing source and target states, COAST learns context-specific causal graphs and structural causal models, attributes observed distributional shifts to mechanism-level causal drivers, and introduces a novel constraint-aware multi-objective optimization formulation that balances transition efficacy, intervention complexity, and target-state stability. The approach is modular and domain-agnostic, integrating feature selection, causal discovery, causal modeling, and intervention identification and evaluation through interchangeable components. Across synthetic benchmarks and real biological datasets, COAST recovers key causal drivers and identifies robust single- and multi-target intervention strategies that achieve desired state transitions, accompanied by transparent mechanistic rationales to guide experimental validation.
[LG-103] FedQHD: Closed-Form Function-Space Federated Reinforcement Learning
链接: https://arxiv.org/abs/2605.29002
作者: Yuchen Hou,Yongshan Chen,Zhuowen Zou,Calvin Yeung,Mohsen Imani,Tian Lan,Mahdi Imani
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注:
Abstract:Federated reinforcement learning enables decentralized agents to collaboratively improve policies or value estimates without exchanging raw trajectories. However, FedAvg-style parameter averaging is not function-space consistent: when clients use heterogeneous encoders or even identical nonlinear networks, averaged parameters need not correspond to the weighted average of client value functions in any common function space. We propose FedQHD, a federated Q-learning method using hyperdimensional (random-feature) state encoders with a linear readout, so that Q-functions are nonlinear in state yet linear in trainable parameters. This linear structure enables closed-form aggregation. With a shared encoder, the function-space consensus update coincides exactly with weighted averaging of local readout matrices. With heterogeneous encoders, the server constructs a global teacher by averaging client Q-values on a shared anchor-state set, and each client compiles this teacher into its local representation via a single ridge projection. We formalize the federation gap – the error incurred when compiling a federated teacher into a heterogeneous client representation – relative to a client-specific oracle projection. We show that this gap decomposes into subspace misalignment, anchor-set conditioning, and regularization bias. We further identify the anchor-to-dimension ratio m \geq D_i as the well-conditioned regime in which the gap reduces to a multiple of the encoder heterogeneity floor. On four continuous-state, discrete-action control benchmarks, FedQHD matches or outperforms FedAvg-style baselines and distillation-based alternatives while requiring substantially less computation, and the empirical dependence of the federation gap on encoder dimension matches our theoretical analysis.
[LG-104] Learning Robust and Task-Invariant Functional Representation from fMRI through Siamese Self-Supervised Learning
链接: https://arxiv.org/abs/2605.28990
作者: Jiyao Wang,Peiyu Duan,Nicha C. Dvornek,Lawrence H. Staib,Denis Sukhodolsky,Pamela Ventola,James S. Duncan
类目: Machine Learning (cs.LG)
*备注:
Abstract:Functional magnetic resonance imaging (fMRI) is a powerful tool for investigating human brain function. However, the high cost of data acquisition and the inherent subjectivity of psychiatric rating scales often lead to datasets with small sample sizes and variable label quality, especially when targeting a specific neurological condition. Combined with the inherently high dimensionality of fMRI data, these limitations substantially increase the risk of model overfitting. Recent years have seen growing interest in developing fMRI foundation models by combining multiple datasets; however, the computational resources needed for pretraining and fine-tuning are often prohibitive. We show that a lightweight self-supervised framework yields representations that generalize across diverse downstream tasks, outperforming fully supervised baselines and approaching the performance of large-scale models. We introduce BrainSimSiam, a data-efficient self-supervised representation learning framework that leverages positive-only data pairs to learn robust and generalizable features. We demonstrate that the learned representations achieve strong performance across multiple downstream classification and regression tasks, highlighting the potential of BrainSimSiam for data-limited neuroimaging applications.
[LG-105] A Training-Time Diagnostic for Generalization via the Log-Alignment Ratio
链接: https://arxiv.org/abs/2605.28975
作者: Ali Shehper,Ashish Vaswani
类目: Machine Learning (cs.LG)
*备注: 32 pages, 25 figures
Abstract:We study the log-alignment ratio (LAR), a measure of parameter-activation alignment, introduced in parameterization theory. We reformulate it as the overlap between a weight spectrum p of the normalized squared singular values of a matrix and an activation spectrum q of the normalized squared projections of inputs onto its singular directions. We show that unembedding LAR tracks the transition between memorization and generalization in two different settings by capturing the spread of p and q during training. In grokking, LAR predicts the effective dimension of the learned function: k \approx n^2(1-\textLAR) , where n is the input dimension of the matrix. In 3B-parameter language model pre-training, its deviation from a non-overfitting baseline tracks the generalization gap, and its rate of decline increases as overfitting approaches. LAR is computable from quantities available during the forward pass with negligible computational overhead, and requires no held-out validation data.
[LG-106] Optimal Rates for Differentially Private Hypothesis Testing with E-values
链接: https://arxiv.org/abs/2605.28952
作者: Ben Jacobsen,Tomas Gonzales,Gavin Brown,Kassem Fawaz,Aaditya Ramdas
类目: Cryptography and Security (cs.CR); Data Structures and Algorithms (cs.DS); Information Theory (cs.IT); Machine Learning (cs.LG)
*备注: 28 pages, 2 figures
Abstract:E-values have attracted considerable interest in recent years as flexible tools for enabling anytime-valid and adaptive data analysis. Hypothesis testing is at the core of many of these applications, which can often involve private or sensitive data. In this work, we answer a simple but important question: given two distributions \mathbbP and \mathbbQ , what is the maximum achievable e-power when testing X\sim \mathbbP^n against X\sim\mathbbQ^n with e-values that satisfy \varepsilon -differential privacy? We characterize the optimal rate for this problem and provide an algorithm which matches it exactly. In the sequential setting, when observations arrive one-by-one and the analyst chooses when to halt, we give matching upper and lower bounds on the stopping times of any private e-process. Numerical experiments confirm the practicality of our algorithms, which require less data than the recently proposed DP-SPRT across a range of sequential testing problems and privacy levels.
[LG-107] Cycle-Space Informed Detection of Autoencoded Blind False Data Injection Attacks on Power Systems
链接: https://arxiv.org/abs/2605.28912
作者: Xin Li,Chenhan Xiao,Jonathan Cohen,Aviad Elyashar,Yang Weng,Rami Puzis
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注: 13 pages, 11 figures
Abstract:The rapid growth of AI-driven data centers and large-scale energy storage systems is increasing the reliance of power system operation on real-time measurement data and automated decision-making. However, many existing detection methods rely on statistical or data-driven analysis of measurements and can fail when attackers exploit the same data structure to craft stealthy perturbations. To illustrate this limitation, we demonstrate a blind False Data Injection Attack (FDIA) in which an Autoencoder learns the measurement manifold and generates perturbations aligned with the Jacobian null space, thereby allowing the attack to evade both residual-based baddata detectors and time-series anomaly detectors. To mitigate data-driven FDIAs which exploit the null space, we propose a topology-informed Cycle-Space Detector (CSD) that leverages the Cycle-Space of the network to impose structural constraints that enhance null space estimation. In addition, we prove that by using the Minimum Cycle Basis (MCB), the proposed CSD achieves the optimal generalization error for attack detection. By exploiting topology-derived cycle constraints rather than relying solely on numerical null space estimation, the proposed method does not require precise line parameters and improves the separation between normal and attacked measurements. Simulation results on IEEE 14-, 30-, 57-, and 118-bus systems demonstrate that the proposed method effectively detects data-driven FDIAs under realistic measurement noise.
[LG-108] Sequential Physics-Constrained Neural Operator Forward Modeling for the textitNorne Reservoir System
链接: https://arxiv.org/abs/2605.28909
作者: Clement Etienam,Juntao Yang,Oleg Ovcharenko,Nick Luiken,Tsubasa Onishi,Nefeli Moridis,Issam Said
类目: Machine Learning (cs.LG)
*备注: 22 pages, 2 figures, 2 tables. Code available at this https URL
Abstract:We develop a comprehensive mathematical and computational framework for sequential surrogate modeling of three-phase black-oil reservoir dynamics using neural operators, with particular emphasis on Fourier Neural Operators (FNO) and their physics-informed variant (PINO). The application focus is the Norne benchmark reservoir, defined on a heterogeneous 46\times112\times22 grid ( N=113,344 cells), with a production history spanning T=30 timesteps covering 3298 days. Our theoretical contributions are organized around four interlocking problems: (1) functional-analytic formulation in a product-Sobolev-space setting, including well-posedness of the implicit timestep map and sharp local Lipschitz estimates; (2) covariate shift quantification, proving that the Wasserstein-2 distance grows as W_2 \leq \varepsilon(L^n-1)/(L-1) , with exponential population-risk discrepancy for L1 ; (3) physics-constrained spectral stability, showing PINO training with \lambda_R \geq \lambda^_R reduces the learned Jacobian spectral radius to \rho_F + C\lambda_R^-1/2 , yielding uniform-in-time rollout error |\delta_n| \leq \varepsilon/(1-\rho) ; and (4) K -step TBPTT gradient analysis, deriving geometric bias decay O(\rho^K) , optimal window K^ = O(\log(T/\sigma^2)) , and Adam convergence O(1/\sqrtt) + O(\rho^K^) . Empirical validation confirms all theoretical predictions: autoregressive PINO surrogates sustain R^20.99 (oil), R^20.90 (gas), R^2\approx 0.80 (pressure), and monotonically improving R^2 (water) across the full 3298-day horizon, trained on eight NVIDIA B200 GPUs in under one hour. A 1000-member ensemble runs in under one minute on a single B200 GPU, giving a \sim10^4\times wall-clock speedup over the OPM finite-volume simulator.
[LG-109] Spectral Guidance for Flexible and Efficient Control of Diffusion Models ICML2026
链接: https://arxiv.org/abs/2605.28900
作者: Gabriel Moreira,Manuel Marques,João Paulo Costeira,Chenyan Xiong
类目: Machine Learning (cs.LG)
*备注: ICML 2026
Abstract:We introduce Spectral Guidance, a framework for controlling diffusion models by leveraging the intrinsic geometry of the generative process. As data is progressively corrupted by noise, only a small number of features remain informative for control. We characterize them as the singular functions of a conditional expectation operator and show that they can be learned via a self-supervised objective. Once recovered, this basis enables the projection of arbitrary guidance signals, such as labels, CLIP embeddings, or masks, directly onto the sampling trajectory. This approach allows for stable, high-fidelity control without retraining or denoiser backpropagation during sampling. Empirically, we improve conditional accuracy on CIFAR-10 by 37 percentage points over the strongest training-free baseline while offering 4\times faster sampling. Moreover, the same representations that support label and CLIP guidance also enable spatial control, such as mask-based guidance, without auxiliary models. Finally, our framework reveals a phase transition in the generative process, pinpointing the optimal time window for effective guidance.
[LG-110] Feature Geometry of LoRA Adapters: A Sparse Autoencoder Analysis of Representational Divergence in Fine-Tuned Language Models
链接: https://arxiv.org/abs/2605.28896
作者: Prasanth K K
类目: Machine Learning (cs.LG)
*备注:
Abstract:Low-Rank Adaptation (LoRA) has emerged as a widely adopted approach for adapting large language models, yet the internal representational changes induced by LoRA fine-tuning remain insufficiently understood. In this work, we investigate the geometry of LoRA-induced representations using Sparse Autoencoders (SAEs). We introduce a delta activation framework that isolates the adapter-specific contribution to the residual stream. Using Gemma-2-9B with LoRA ranks 4, 8, 16, and 32, we train adapter-specific SAEs across multiple transformer layers and compare their learned feature spaces with pretrained SAE dictionaries. We evaluate representational alignment using cosine similarity between decoder directions, principal-angle analysis of feature subspaces, and Centered Kernel Alignment (CKA) between activation representations. Across layers and ranks, we consistently observe comparatively weak geometric alignment between LoRA-induced feature dictionaries and pretrained SAE features. Adapter-specific SAEs also reconstruct delta activations more effectively than pretrained SAEs, suggesting that LoRA updates occupy partially distinct representational structure within the residual stream. Additionally, feature density increases with rank and depth, while geometric divergence remains relatively stable across ranks. These findings provide empirical evidence that LoRA fine-tuning can induce feature structures that are not fully captured by pretrained interpretability dictionaries, with implications for mechanistic interpretability, adaptation analysis, and safety auditing of fine-tuned language models. Subjects: Machine Learning (cs.LG) Cite as: arXiv:2605.28896 [cs.LG] (or arXiv:2605.28896v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2605.28896 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Prasanth K K [view email] [v1] Wed, 27 May 2026 11:54:23 UTC (1,532 KB)
[LG-111] Echoes within the Reasoning : Stealthy and Effective Watermarking via Chain of Thought ICML2026
链接: https://arxiv.org/abs/2605.28890
作者: Jiacheng Lu,Yiming Li,Tao Song,Weijian Wang,Wenjie Qu,Haibing Guan,Jiaheng Zhang
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注: This paper is accepted by ICML2026
Abstract:Large Language Models with Chain-of-Thought reasoning capabilities represent valuable intellectual property, yet existing black-box watermarking methods often trade robustness for reasoning fidelity by perturbing final answers or relying on fragile trigger patterns. We propose BiCoT, a watermarking framework that embeds ownership signals into the internal geometry of reasoning traces by aligning high-saliency structural anchors with a private signature subspace while regularizing ordinary control tokens to preserve semantic capacity. This design couples the watermark with reasoning-relevant representations, making removal difficult without disrupting the features that support coherent reasoning. To enable verification under model theft and representation drift, we introduce Robust Subspace Registration (RSR), a Top- logprob-based black-box verifier that uses sentinel tokens to calibrate systematic shifts in the output distribution. Experiments show that BiCoT preserves reasoning fidelity across diverse complex reasoning tasks while achieving robust detection under fine-tuning, quantization, model-level perturbations, and adaptive output-level attacks across in-domain and out-of-distribution settings. Comments: This paper is accepted by ICML2026 Subjects: Cryptography and Security (cs.CR); Machine Learning (cs.LG) Cite as: arXiv:2605.28890 [cs.CR] (or arXiv:2605.28890v1 [cs.CR] for this version) https://doi.org/10.48550/arXiv.2605.28890 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-112] owards Continuous-time Causal Foundation Models ICML2026
链接: https://arxiv.org/abs/2605.28880
作者: Dennis Thumm,Ruben Wiedemann,Ying Chen
类目: Machine Learning (cs.LG); Data Analysis, Statistics and Probability (physics.data-an); Methodology (stat.ME)
*备注: ICML 2026 2nd Workshop on Foundation Models for Structured Data (FMSD)
Abstract:Extending discrete-time causal Prior-data Fitted Networks for time series to continuous time invites writing the mechanism as a stochastic differential equation (SDE) – but if the SDE is integrated \emphonce per observation gap, the trajectory law depends on when it is observed, and the prior remains a discrete-time Markov model in SDE clothing. We propose a precise continuity criterion – trajectory-law invariance to the observation schedule – together with a three-tier taxonomy (discrete; naive observation-grid integration; fine-grid integration with decoupled observation) and a construction realising the top tier on a random DAG with OU or small-MLP nonlinear drifts, irregular observation schedules, and hard / soft / time-varying interventions. A 2 \times 2 encoder \times integrator ablation, run independently on a linear and a nonlinear prior, finds fine-grid integration beats naive on 8/8 cells (sign-consistency p 1/256 ) with the gap growing as the eval grid refines; the encoder axis is null with fine integration but time-aware-leading with naive. We release the prior and a preliminary zero-shot protocol on pharmacokinetic and physical-system data. Comments: ICML 2026 2nd Workshop on Foundation Models for Structured Data (FMSD) Subjects: Machine Learning (cs.LG); Data Analysis, Statistics and Probability (physics.data-an); Methodology (stat.ME) Cite as: arXiv:2605.28880 [cs.LG] (or arXiv:2605.28880v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2605.28880 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Dennis Thumm [view email] [v1] Tue, 26 May 2026 12:06:04 UTC (115 KB)
[LG-113] Pre-Registering the Detectable Effect: A Paired-MDE Budget for 4-bit Quantization Benchmarks with a Pilot Audit
链接: https://arxiv.org/abs/2605.28873
作者: Zexin Zhuang,Yanhang Li,Zhichao Fan
类目: Machine Learning (cs.LG)
*备注:
Abstract:This is a planning-method note with an unpaired pilot audit. We adapt the classical paired-binary sample-size calculation (Miettinen, 1968) to quantization benchmarks, giving a conservative minimum detectable effect (MDE) bound \delta^* \le (z_1-\alpha/2+z_1-\beta)\sqrt\rho_d/m in the paired item count m and the FP16-NF4 disagreement rate \rho_d . The bound turns “how reliable is my quantization claim?” into a one-line budget a benchmark designer can commit to before running. We illustrate the bound on four models and four benchmarks ( k=5 splits of n=100 ), and add a parallel MMLU prompt-template study to put the bound’s quantization-noise scale alongside the prompt-noise scale. Assuming \rho_d=0.10 (an unmeasured planning value), all observed NF4-FP16 deltas fall below the implied MDE, and most cross-split SDs lie within \pm 1.5 pp of the binomial reference \sqrtp(1-p)/n , so much of the variance reported as “benchmark unreliability” on n=100 subsamples is binomial sampling noise. The single borderline cell (OPT-WinoGrande, |\Delta|=3.2 pp) is below the implied MDE at \rho_d=0.10 but above it at \rho_d=0.05 , illustrating the planning trade-off the bound makes explicit. On MMLU, prompt-template ranges of 2-10 pp meet or exceed the largest observed quantization delta (3.2 pp), so a quantization audit that does not first fix the prompt template absorbs template variance into its noise floor. We complement the bound with a five-line pre-registration template.
[LG-114] Molecular Lead Optimization via Agent ic Tool Planning
链接: https://arxiv.org/abs/2605.28862
作者: Lingxiao Li,Haobo Zhang,Ruohao Fan,Bin Chen,Jiayu Zhou
类目: Machine Learning (cs.LG); Quantitative Methods (q-bio.QM)
*备注: 12 pages
Abstract:Drug discovery is a lengthy and resource-intensive process composed of multiple stages. Among these stages, lead optimization plays a critical role in transforming early hit compounds into viable drug candidates. This stage requires improving ADMET-related properties through subtle structural refinement while preserving key molecular substructures responsible for binding affinity to disease targets. Recent advances in artificial intelligence have shown promise in accelerating various aspects of drug discovery; however, most existing approaches to lead optimization rely on one-step molecular optimization, which fail to account for the long-term consequences of sequential design decisions. To address this limitation, we propose TRACE, a trajectory-aware, LLM-reasoning agent for molecular lead optimization that formulates tool selection as a sequential decision-making problem over action trajectories. Given a lead molecule and an optimization objective, TRACE makes trajectory-aware decisions over molecular optimization tools, enabling forward-looking refinement under structural constraints. Experiments on multiple ADMET optimization tasks show that our agent achieves higher optimization success, larger property improvements, and higher validity, while preserving molecular similarity compared to baseline models.
[LG-115] An End-to-End PyTorch Interface for Differentiable PDE Solvers: A RANS Model-Correction Study
链接: https://arxiv.org/abs/2605.28858
作者: Luca Saverio,Michele Alessandro Bucci,Gianmarco Farro,Cédric Content,Denis Sipp(MONHADE)
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG); Mathematical Physics (math-ph)
*备注:
Abstract:This work presents an end-to-end strategy for solving inverse problems constrained by Partial Differential Equations within a fully differentiable Machine Learning framework. The proposed formulation provides a unified and user-friendly methodology applicable to a wide range of problems, from data assimilation to closure modeling. Our approach combines a baseline differentiable PDE solver, which predicts the state w from the nonlinear system R(w) = 0 , with a generic additive, parametrized, and differentiable correction f_\phi(w) , with trainable parameters \phi . We show how to optimize phi within a fully differentiable Python workflow by reformulating the PDE as an implicit layer, enabling its integration into arbitrary objective functions, while leveraging PyTorch’s automatic differentiation graph. The method is demonstrated on the Reynolds-Averaged Navier-Stokes equations for compressible flows, where the closure term, or a portion of it, is modeled using trainable parameters or a Neural Network. The first application considers the 2D NASA Wall-Mounted Hump test case, where a production-term parameter is optimized against time-averaged LES data. A second application is carried out on the VKI LS-59 turbine blade, where the Spalart-Allmaras eddy viscosity field is reconstructed through the optimization of a trainable spatial field. A dataset is generated starting from the VKI LS-59 turbine blade geometry using the differentiable BROADCAST solver with the Spalart-Allmaras turbulence model. The results highlight the flexibility of the framework, showing its applicability beyond turbulence modeling to a broader class of physics-informed PDE-constrained problems with data-driven components.
[LG-116] Representation Signatures and Risk-Feedback Alignment in LLM Trading Agents
链接: https://arxiv.org/abs/2605.28850
作者: Weicheng Xue
类目: Machine Learning (cs.LG); Computational Finance (q-fin.CP)
*备注:
Abstract:We study behavioral alignment and representation dynamics of large language model (LLM) agents in financial decision environments. Using TradeArena, an auditable trading-agent testbed with risk reports, execution simulation, memory, and replayable trajectories, we analyze how rationales, positions, and interventions evolve under market stress. We find measurable pre-failure signatures: planning embeddings drift from normal-state centroids, fused plan-risk representations separate normal from pre-drawdown states, and manifold diagnostics show effective-rank contraction before failures. To address small-sample and embedding-choice concerns, we use 80 rolling failure anchors across eight LLM trajectories and show that contraction persists across hash, LSA, Transformer, and white-box hidden-state probes. Stress tests with CoT-free target weights, lexical controls, OHLCV noise, and false-audit reports indicate that rationale-level contraction can vanish without rationales, while intent-space contraction may remain; lexical diversity does not collapse; and fused signatures remain informative under noise. We also find that structured risk feedback can act as an external alignment signal without fine-tuning, but not as a universal performance enhancer: true audit feedback improves calibration for some models, return and drawdown for others, and reveals cases where hidden or placebo feedback has higher short-horizon return but weaker alignment diagnostics. Finally, a 51-stock intraday experiment reveals a correlation blind spot: LLM rationales often justify concentrated exposure to coupled assets that the risk layer repeatedly clips, with a rolling Markowitz baseline as a covariance reference. These results support a research claim rather than a profitability claim: auditable risk feedback and representation trajectories reveal when LLM financial reasoning is aligning, drifting, or failing.
[LG-117] WASHH: An Anchor-Aware Whale-Guided Selection Hyper-Heuristic for Continuous Optimization and SVC Configuration
链接: https://arxiv.org/abs/2605.28844
作者: Yifu Zhao,Xiaofan Zou,Junhao Wei,Yanxiao Li,Baili Lu,Zhenhong Peng,Dexing Yao,Haochen Li,Qinbin He,Sio-Kei Im,Xu Yang,Yapeng Wang
类目: Neural and Evolutionary Computing (cs.NE); Machine Learning (cs.LG)
*备注:
Abstract:Learning-assisted algorithm design often has to make reliable search decisions under small evaluation budgets, where committing to a single metaheuristic can be unreliable. We propose WASHH, a Whale-guided Adaptive Selection Hyper-Heuristic for continuous black-box optimization. WASHH uses WOA as the main exploitation backbone, but treats PSO-style memory, GWO-style leader averaging, DE-style variation, local coordinate search, and anchor-guided refinement as selectable search behaviors. An online reward controller allocates evaluations according to observed improvements, while anchor refinement exploits inexpensive reference configurations such as box centers or default model settings without bypassing black-box evaluation. On ten 30-dimensional benchmark functions with 10 independent runs and 12,000 evaluations, WASHH achieves the best average rank, 1.10, and is best or tied best on all ten functions. It strictly improves over WOA on eight functions and ties WOA at the numerical optimum on Rastrigin and Griewank. We further study SVC hyperparameter configuration for breast cancer diagnosis under a 300-evaluation budget. WASHH obtains the lowest mean validation log loss among the compared optimizers, suggesting that anchor-aware selection hyper-heuristics are a practical lightweight direction for LEAD systems.
[LG-118] he Biosecurity Blind Spot: Systematic Dual-use Detection in Open Science Infrastructure
链接: https://arxiv.org/abs/2605.28843
作者: Vasudha Sharma,Chakresh Kumar Singh,Jayesh Choudhari,Dharmit Nakrani
类目: Digital Libraries (cs.DL); Computers and Society (cs.CY); Machine Learning (cs.LG)
*备注: Ongoing work
Abstract:AI is transforming life sciences research at unprecedented speed, accelerating discovery across protein structure prediction, genome modeling, and drug development (Jumper et al., 2021; Mak et al., 2024). Yet this rapid advancement, coupled with the open science movement, introduces significant dual-use research concerns that have received limited empirical scrutiny. Here we present the first systematic analysis of dual-use research of concern (DURC) content on open preprint servers. We screened ~52,000 bioRxiv preprints (2024-2025) using a hybrid pipeline of lexical filtering and large language model (LLM) evaluation, scoring metadata across nine DURC, three PEPP, and five governance categories aligned with U.S. and Australia Group oversight frameworks. Our analysis reveals that dual-use-adjacent knowledge is routinely present in openly accessible titles and abstracts, often exceeding established risk thresholds even in studies with legitimate public health objectives. While this mapping captures surface-level information diffusion, it does not measure operational capability, downstream misuse potential, or the substantial technical and biosafety barriers that constrain harmful application. We argue that institutional review processes, funding requirements, and preprint platform policies must evolve to incorporate proactive, metadata-level monitoring without compromising scientific transparency. Ultimately, harmonizing controlled-access mechanisms for high-risk methodologies with open summaries of scientific contributions offers a pragmatic framework for governing AI-accelerated biology at scale.
[LG-119] One Mask to Rule Them All: On Hidden Facts after Editing and How to Find Them ACL2026
链接: https://arxiv.org/abs/2605.28839
作者: Ali Holmov,Paul Youssef,Nandi Schoots,Christin Seifert
类目: Machine Learning (cs.LG)
*备注: Accepted to Findings of ACL 2026
Abstract:Knowledge editing methods such as ROME and MEMIT update factual associations in transformer models by modifying MLP weights. While evaluated mainly by output behavior, their internal mechanism remains underexplored. We investigate whether edits rely on a common mechanism, regardless of which fact is modified. Despite fact-specific weight changes, we argue that ROME and MEMIT target the same subset of weights critical for maintaining edits. To isolate this subset, we train a compact binary mask over the edited weights. The mask reverses 80% of edits on the training set and over 70% on the test set, confirming that diverse edits share a common functional structure. Our analysis reveals that the mask reverses edits by eliminating overattention in later layers. Additionally, we show that injecting the mask during editing drops editing success from 98% to 38%, demonstrating that this mechanism is necessary for edits to succeed. Our finding that edits suppress rather than overwrite knowledge explains why ROME and MEMIT fail to propagate changes to related facts. The identified common functional subspace informs detection and defense against unwanted edits.
[LG-120] Adapting Automotive Aerodynamics Surrogates to New Vehicle Families via Transfer Learning
链接: https://arxiv.org/abs/2605.27968
作者: Seunghwan Keum,Alok Warey
类目: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG); Computational Physics (physics.comp-ph)
*备注: 23 pages, 12 figures
Abstract:Deploying Scientific Machine Learning surrogates in industrial CFD workflows requires adapting pretrained models to new vehicle families without large datasets; yet whether geometric representations learned by a geometry encoder transfer to topologically distinct shapes remains unvalidated. We address this through leave-one-family-out experiments on a 61.47M-parameter Transformer surrogate (AB-UPT) pretrained on four vehicle families (411 external aerodynamics cases) and adapted to the held-out fifth with only 20 samples. Three strategies are compared: Full Fine-Tuning (FFT), Lightweight Fine-Tuning (LFT), and Low-Rank Adaptation (LoRA). The central finding is that pretrained geometry encoders learn transferable representations, but the adaptation mechanism determines whether they can be exploited. FFT destabilizes as 61.47M unconstrained parameters overfit to 20 samples (R^2=0.40); LFT fails because the frozen encoder cannot represent unseen shapes (R^20). LoRA resolves both: rank-constrained adapters injected into all layers regularize the loss landscape while preserving pretrained features, achieving R^2=0.85+/-0.02 across all five families with 50% lower force RMSE than FFT and 28% lower pointwise field errors. LoRA also outperforms from-scratch training using 3x more target-family data, eliminating the need for large per-family datasets. These results recast LoRA from a memory-saving convenience into a convergence enabler for geometry transfer: a shared backbone paired with lightweight per-family adapters trainable in hours from minimal data. Comments: 23 pages, 12 figures Subjects: Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG); Computational Physics (physics.comp-ph) Cite as: arXiv:2605.27968 [cs.CE] (or arXiv:2605.27968v1 [cs.CE] for this version) https://doi.org/10.48550/arXiv.2605.27968 Focus to learn more arXiv-issued DOI via DataCite
[LG-121] Leave a Window Out: Modifying the Jackknife for Predictive Inference in Time Series
链接: https://arxiv.org/abs/2605.30292
作者: Hanyang Jiang,Rina Foygel Barber,Ashwin Pananjady,Yao Xie
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST); Methodology (stat.ME)
*备注: 36 pages, 6 figures
Abstract:Conformal prediction methods enjoy strong theoretical and empirical predictive inference performance, provided the data is exchangeable, and predictors are trained in a memoryless fashion. However, these assumptions and constraints are impractical in many real-data settings, such as time series (where temporal dependence violates exchangeability, and where memoryless predictors will inevitably have poor predictive accuracy). Recent work shows that the split conformal prediction method is robust to these issues of memory-based predictors and deviations from exchangeability that are common features of time-series data. However, since using sample splitting can lead to lower accuracy, this motivates asking whether other predictive inference methods (that do not rely on data splitting) could also be reliably used in the time series setting. In this work, we show that the vanilla leave-one-out jackknife can suffer an arbitrary loss of coverage even in canonical time series models with mild temporal dependence. As a remedy, we propose a careful modification tailored to such settings, which we term the \emphleave-a-window-out (LWO) method, and show that it can achieve valid coverage provided that the model-fitting procedure satisfies mild stability properties. Our proofs are based on quantifying the degree to which the data departs from \emphcyclic exchangeability, and we introduce new coefficients to measure the extent of this departure. Experiments on time series data demonstrate that our LWO method often enjoys valid coverage when the vanilla jackknife fails to cover, while producing much narrower intervals than split conformal prediction. Comments: 36 pages, 6 figures Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST); Methodology (stat.ME) Cite as: arXiv:2605.30292 [stat.ML] (or arXiv:2605.30292v1 [stat.ML] for this version) https://doi.org/10.48550/arXiv.2605.30292 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-122] Wasserstein Contraction of Coordinate Ascent Variational Inference
链接: https://arxiv.org/abs/2605.30253
作者: Rocco Caprio,Adrien Corenflos,Sam Power
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Functional Analysis (math.FA); Optimization and Control (math.OC); Probability (math.PR); Computation (stat.CO)
*备注: 17 pages + 3 pages appendix, 3 figures
Abstract:We study the contraction in Wasserstein distance of the coordinate ascent variational inference algorithm. This is shown to hold under a transport-information inequality at the fixed points and a functional smoothness condition. The results are general and sharp, allow for local convergence guarantees, hold for general smooth manifolds, and also in some non-smooth spaces. We consider applications to Bayesian Gaussian Mixture Models, and high-dimensional Bayesian Probit Regression, and Logistic Regression with Pólya-Gamma random variables (i.e. Jaakkola-Jordan’s algorithm).
[LG-123] A new completely parameter-free clustering algorithm for unsupervised classification of BATSE gamma-ray bursts
链接: https://arxiv.org/abs/2605.30175
作者: Soumita Modak
类目: High Energy Astrophysical Phenomena (astro-ph.HE); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:Cluster analysis is a widely applied machine learning technique to understand the existing patterns in the population of gamma-ray bursts (GRBs), in order to explore their physical sources. In the present scenario, the number of clusters corresponding to differentiable groups is still under conflict, in spite of numerous attempts with the state-of-the-art clustering procedures. This crucial unknown parameter needs to be evaluated, either directly or indirectly in terms of other tuning parameters, to produce the clusters in GRBs through implementation of an appropriate clustering algorithm. While most of the applied algorithms reached two physically explained groups of merger and collapsar predominated by the short and long bursts respectively, other statistical approaches violated this binary partition. However, physical establishment of any additional cluster(s) is not yet confirmed. Therefore, we propose a new algorithm, from a different stream of clustering referred to as `completely parameter-free’, which carries out the classification of GRBs in a manner that has not been tried so far. It indicates two main groups, of short and long duration bursts from the BATSE sample, compatible with the merger-collapsar theory.
[LG-124] Diffusion Models Are Statistically Optimal for Learning Low-Dimensional Multi-Modal Distributions ICML2026
链接: https://arxiv.org/abs/2605.30153
作者: Jingda Wu,Changxiao Cai
类目: Machine Learning (stat.ML); Information Theory (cs.IT); Machine Learning (cs.LG); Statistics Theory (math.ST)
*备注: accepted to ICML 2026
Abstract:Score-based diffusion models have demonstrated remarkable empirical success in learning high-dimensional distributions, particularly those exhibiting low-dimensional and multi-modal structures. However, theoretical understanding of their statistical efficiency remains limited. Existing theories typically rely on strong regularity assumptions, such as uniformly bounded densities or globally smooth score functions, which fail to capture such intrinsic structures. In this work, we study the sample complexity of diffusion models for learning distributions supported on a union of low-dimensional subspaces. Assuming that the data distribution within each subspace is subgaussian, we show that diffusion models require at most \widetildeO(\varepsilon^-k \vee 2) samples to achieve \varepsilon error in 1-Wasserstein distance, where k is the intrinsic dimension. This near-optimal convergence rate depends only on the intrinsic dimension and significantly improves upon prior theoretical guarantees that suffer from the curse of dimensionality. Notably, our analysis applies to a broad collection of distributions without imposing smoothness, bounded-density, or log-concavity assumptions. Overall, our results show that diffusion models can statistically adapt to intrinsic low-dimensional structure while naturally accommodating multi-modal data, offering a rigorous theoretical justification for their success in complex high-dimensional learning tasks.
[LG-125] Joint Model and Data Sparsification via the Marginal Likelihood ICML2026
链接: https://arxiv.org/abs/2605.29908
作者: Alexander Timans,Thomas Möllenhoff,Christian A. Naesseth,Mohammad Emtiyaz Khan,Eric Nalisnick
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 36 pages, 8 figures, 12 tables (incl. appendix); published at ICML 2026
Abstract:Sparse recovery in linear systems underpins applications from signal processing to high-dimensional regression. Sparse Bayesian Learning, grounded in the principle of automatic relevance determination (ARD), offers a practical Bayesian mechanism for feature sparsity via marginal likelihood optimization. Yet, its reliance on a homoscedastic noise model renders it sensitive to data contaminations such as outliers or misspecified noise, harming model fit and predictions. Instead, we propose jointly learning individual feature and sample relevancies, enabling simultaneous model and data sparsification via a single Bayesian objective. This symmetric pruning of model and data offers a natural extension that preserves conjugacy, admits closed-form updates for standard optimization procedures, and aligns with perspectives from robust regression and influence functions. Empirical results across diverse regression tasks affirm that a joint ARD approach consistently yields both sparse and robust prediction models.
[LG-126] Instance-dependent Stochastic Lipschitz bandit
链接: https://arxiv.org/abs/2605.29748
作者: Marius Potfer,Vianney Perchet
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:We study the Lipschitz bandit problem, where a learner sequentially maximizes an unknown Lipschitz function f over a domain \mathcalX \subset [0,1]^d using noisy pointwise evaluations. Existing regret bounds are either worst-case, scaling as \tilde\Theta \left ( T^d+1/d+2\right ) , or adaptive via the zooming dimension d_z , yielding \tilde\Theta \left ( T^d_z+1/d_z+2\right ) . However, such zooming-based guarantees are only partially instance-dependent, as they depend solely on the asymptotic growth of near-optimal level sets and fail to capture finer structural properties of f . We provide an analysis and an algorithm that characterizes the regret through integrals of the suboptimality gap of f over its level sets. This yields regret bounds that adapt to the local growth of level sets, rather than only their asymptotic behavior. As a corollary, when the set of maximizers has dimension d^\star0 , we obtain improved adaptive rates of order \tilde\mathcalO \left ( T^d_z+1 / \max(d_z,d^\star)+2\right ) strictly improving over classical zooming bounds in this regime. Finally, we extend our analysis to the full-information setting (Lipschitz experts) and show how some of the regularity assumptions can be relaxed.
[LG-127] Eigen-Spike Emergence and Quadratic Equivalents for Conjugate Kernels on Nonlinearly Separable Data
链接: https://arxiv.org/abs/2605.29669
作者: Collin Cranston,Zhichao Wang,Todd Kemp,Michael W. Mahoney
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Probability (math.PR); Statistics Theory (math.ST)
*备注: 89 pages, 10 figures
Abstract:Recent work in random matrix theory (RMT) has developed the notion of deterministic equivalents: typically linear surrogate models that approximate the spectral behavior of large nonlinear random matrices, such as nonlinear feature maps in neural networks (NNs). On the one hand, these deterministic equivalents make theoretical predictions tractable by reducing a complex model to a simpler model with properties that fall under the umbrella of classical RMT tools. However, this leaves open the question of whether this idealized linear equivalence remains meaningful when dealing with high-dimensional nonlinearly separable data, such as performing clssification on nonlinearly separable data. Motivated by this, we consider the conjugate kernel (CK), which is the nonlinear feature map of a feedforward NN, under a canonical nonlinearly separable dataset, the XOR problem; and we use the study of informative outlier eigenvalues in the CK and whether their corresponding eigenvectors asymptotically align with XOR labels as a proxy for nonlinear learnability. We develop a robust quadratic equivalent to the spiked CK matrix that enables a precise analysis of emergent informative spikes, as one modifies various knobs common in ML practice: sample complexity, signal-to-noise ratio (SNR), nonlinear activation choice, and pretrained features. In each of these scenarios, we derive a precise BBP-type phase transition in which linear classification via the CK eigenvectors becomes possible. Our analysis helps translate the power of deterministic equivalence tools in RMT to study problems of practical relevance in ML.
[LG-128] Matching Rates and Optimal Allocation for Federated Probe-Logit Distillation under Heterogeneous Bandwidth Budgets
链接: https://arxiv.org/abs/2605.29642
作者: Prasanjit Dubey,Xiaoming Huo
类目: Machine Learning (stat.ML); Information Theory (cs.IT); Machine Learning (cs.LG)
*备注:
Abstract:In federated language modeling, K nodes each hold n samples but cannot pool data or exchange full-precision gradients or weights. We study the minimax rate at which a conditional distribution over V tokens can be estimated when each node may upload at most B bits per query in a public probe set. In federated probe-logit distillation (FPLD), each node transmits a scalar-quantized logit vector on the probe set, and an aggregator distills a global parametric student. Prior work (Dubey and Huo, 2026) establishes a high-probability KL rate O(d/(Kn) + \rho\sqrtV \log V / m + K^-1 \cdot 2^-2B/V) plus optimization slack, with the bandwidth term in its trace-sharpened form. Whether this bandwidth-term rate is tight, and how the upper bound generalizes to heterogeneous per-node bandwidths, are left open. We close both gaps. First, the dithered FPLD construction has a matching single-round lower bound \Omega(K^-1 \cdot 2^-2B/V) under non-degeneracy, pinning the bandwidth-axis rate at \Theta(K^-1 \cdot 2^-2B/V) . T -round sequential refinement with nested/scaled residual quantizers achieves O(K^-1 \cdot 2^-2TB/V) ; vanilla FPLD’s T -independent bandwidth term is suboptimal for every T 1 . Second, we establish a heterogeneous-bandwidth upper bound for per-node budgets B_i , paired with a closed-form optimal allocation B_i^* = B_\mathrmtot/K + (V/2) \log_2(w_i / \barw_g) , a log-tilted water-filling rule that is the per-node analogue of reverse water-filling for distortion-rate optimization. A plug-in adaptive variant estimates the weights from a short warm-up phase and attains 1 + O(\sqrt\log(K/\delta)/(m T_0)) relative suboptimality. Synthetic n-gram simulations confirm that empirical KL is bracketed by the upper and lower bounds and that the optimal allocation strictly dominates uniform and inverse-weighted baselines under heterogeneous clipping. Subjects: Machine Learning (stat.ML); Information Theory (cs.IT); Machine Learning (cs.LG) Cite as: arXiv:2605.29642 [stat.ML] (or arXiv:2605.29642v1 [stat.ML] for this version) https://doi.org/10.48550/arXiv.2605.29642 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-129] MoSSP: A Momentum-Based Single-Loop Stochastic Penalty Method for Nonconvex Constrained DC-Regularized Optimization
链接: https://arxiv.org/abs/2605.29635
作者: Luxuan Li,Chunfeng Cui,Xiao Wang
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注: 35 pages, 3 figures
Abstract:In this paper, we study a structured class of nonconvex constrained stochastic problems with difference-of-convex (DC) regularization, where the feasible set is possibly nonconvex and the concave part of the DC regularizer is allowed to be nonsmooth. The fundamental challenge lies in maintaining feasibility for nonconvex constraints while achieving favorable oracle complexity. Although single-loop algorithms efficiently solve unconstrained DC optimization problems, their potential for constrained optimization with DC structure remains largely unexplored. To address this gap, we develop MoSSP, a Momentum-based Single-loop Stochastic Penalty method for such problems with provable complexity guarantees. The key idea is to apply a single stochastic proximal-gradient step to the Moreau envelope of the penalty plus the convex DC part, with the concave part’s proximal mapping computed in parallel. We derive two algorithm variants: a Polyak-momentum version with O(\varepsilon^-4) oracle complexity for finding stochastic \varepsilon -KKT points, and an improved O(\varepsilon^-3) version incorporating recursive momentum. Experimental results demonstrate the effectiveness of the proposed algorithms.
[LG-130] FPLIER: Federated Pathway-Level Information Extractor
链接: https://arxiv.org/abs/2605.29587
作者: Daniele Malpetti,Christian Berchtold,Francesco Gualdi,Marco Scutari,Laura Azzimonti,Francesca Mangili
类目: Quantitative Methods (q-bio.QM); Machine Learning (cs.LG)
*备注: Accepted for publication at the ACM BCB '26 conference
Abstract:In transcriptomics, gene-set-aware factorization methods such as the Pathway Level Information Extractor (PLIER) are most effective when trained on large, heterogeneous expression compendia. Yet, many clinically relevant cohorts cannot be pooled into a single dataset due to privacy and governance constraints. We present FPLIER, a federated extension of PLIER that enables distributed training across multiple data holders while incorporating publicly available datasets. Through secure aggregation, FPLIER produces training updates algebraically equivalent to those of a centralized pooled-data approach while keeping expression data local. We evaluate FPLIER across multiple scenarios in two simulated consortia (from the K-CLIER and MultiPLIER studies) and demonstrate stable convergence. We further conduct a systematic analysis of membership inference attacks targeting both intermediate training statistics and the released model. Our results show that privacy risk is governed by the rank of the training expression matrix. Incorporating public data or reducing data dimensionality increases this rank, moving the system toward a full-rank regime in which training and non-training samples become indistinguishable to the attacker, and membership-inference performance approaches random guessing.
[LG-131] Deep Optimal Individualized Treatment Rules for Bivariate Survival Outcomes via Adaptive Prediction-Powered Learning
链接: https://arxiv.org/abs/2605.29464
作者: Kun Ren,Yifan Cui,Wen Su
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:In randomized trials involving multiple treatments, bivariate survival outcomes present significant analytical challenges for making decisions. This paper addresses the problem of deriving optimal individualized treatment rules to maximize the joint survival probability beyond fixed time points (t_1, t_2) through deep neural networks, while accounting for right censoring. We propose a novel approach that models treatment rules via stochastic policies, coupling marginal accelerated failure time models via link function to capture bivariate dependence. To enhance robustness and effectiveness of decision making, we introduce an adaptive prediction-powered method that leverages auxiliary predictions from machine learning models.
[LG-132] Kernel-based potential mean-field games with unbiased random Fourier U-statistics
链接: https://arxiv.org/abs/2605.29371
作者: Yumiharu Nakano
类目: Optimization and Control (math.OC); Machine Learning (cs.LG); Numerical Analysis (math.NA); Machine Learning (stat.ML)
*备注:
Abstract:We study the subclass of potential mean-field games in which the running interaction cost and the terminal target cost are both expressed through reproducing-kernel maximum mean discrepancy (MMD) penalties, and develop a computational framework that exploits this kernel structure. Both costs are estimated from finite-sample empirical distributions using a random Fourier U-statistic representation that is unbiased and has linear cost in the batch size. The drift of the controlled diffusion is parametrized by a neural network and trained via stochastic gradient descent. For this subclass we prove a sample-level almost-sure convergence theorem and an explicit almost-sure rate of convergence, under coupled rate conditions on the penalty parameter, the random-feature count, the sample size, and the optimization tolerance. The framework includes the kernel-MMD-penalty Schrödinger bridge problem as the special case of a vanishing interaction cost. Numerical experiments illustrate the method on the Schrödinger bridge problem in dimensions up to one hundred, and on an electric vehicle charging coordination problem with per-vehicle physical heterogeneity, where an aggregate-demand congestion cost represents price-feedback competition at the population level and the terminal MMD penalty shapes the state-of-charge distribution at the deadline.
[LG-133] Mixing Vector Model for Copolymer Inference via Mixed Integer Linear Programming
链接: https://arxiv.org/abs/2605.29329
作者: Jianshen Zhu,Raveena Rai,Taiyo Sohkawa,Naveed Ahmed Azam,Kazuya Haraguchi,Liang Zhao,Tatsuya Akutsu
类目: Quantitative Methods (q-bio.QM); Machine Learning (cs.LG)
*备注:
Abstract:A novel two-phase molecule inference framework, mol-infer, has recently been developed to infer chemical graphs with prescribed abstract structures and desired property values through mixed integer linear programming (MILP) under the two-layered model, with guaranteed optimality and exactness relative to the given learned prediction function and structural constraints. In this study, we extend this framework to copolymers by introducing a simple feature representation, called the mixing vector (MV) model. In the proposed model, a copolymer feature vector is represented as a convex combination of MILP-tractable monomer descriptors weighted by the mixing ratio of the constituent monomers. This representation does not require explicit sequence-class information and is therefore naturally compatible with MILP-based inverse design. Under this model, we construct prediction functions for several copolymer property datasets using artificial neural networks, reduced quadratic multiple linear regression, and random forests. The proposed representation achieves practically useful predictive performance across multiple physicochemical property datasets; in particular, the best test R^2 score exceeds 0.7 for nine of the ten datasets and exceeds 0.9 for six datasets. We also formulate a multi-monomer inverse-design problem under the MV representation with a prescribed mixing ratio and show that the resulting MILP instances remain tractable, even for three-monomer settings. Finally, we perform an external consistency check by re-evaluating the inferred candidates and comparing the re-computed property values with those predicted by the learned model. Overall, the proposed framework gives a tractable first step toward model-level exact inverse design of copolymers under the two-layered model.
[LG-134] Prediction-Powered Inference Across Many Tasks for AI Evaluation Social Science Research
链接: https://arxiv.org/abs/2605.29249
作者: Nicolas Emmenegger,Ellery Stahler,Chara Podimata
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Many applications require statistically valid inference across many related tasks, while using only a handful of high-quality labels per hypothesis. In AI evaluation, these tasks may correspond to model behaviors across prompts, subgroups, or hypotheses; in social science surveys, they may correspond to related questions, populations, or measurement conditions. Prediction-powered inference (PPI) uses abundant but inexpensive proxy measurements to improve inference from limited, ground-truth labels, but commonly used methods treat tasks independently and therefore fail to exploit shared structure across related tasks. This limitation is especially important in settings where only a small number of labels are available per task. To address this issue, we introduce a multi-task prediction-powered inference framework that uses labeled data from related tasks to improve power while preserving task-specific inference. Our methods exploit the shared structure in the proxy-ground-truth relationship through cross-task recalibration, while retaining within-task rectification and power tuning to construct accurate point estimates and confidence intervals. We prove that efficiency gains beyond power-tuned PPI are only possible when the proxy-ground-truth relationship contains nonlinear structure; affine cross-task recalibrations are asymptotically equivalent to using the original proxy. We complement our theoretical findings with experiments on synthetic and semi-synthetic datasets, as well as a case study auditing language models on election-related information during the 2024 U.S. presidential election. Using a large human-annotation study, we show that cross-task recalibration can substantially reduce confidence interval widths when labels are scarce.
[LG-135] Anytime-Valid Federated Conformal RAG for LLM Swarms
链接: https://arxiv.org/abs/2605.29139
作者: Prasanjit Dubey,Xiaoming Huo
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:Federated Conformal RAG (FC-RAG) provides distribution-free coverage for a bandwidth-limited swarm of weak language models, but only at a fixed horizon. We extend it to anytime-valid sequential coverage: validity at every stopping time, preserved under predictable adaptive control (recalibration, per-node bandwidth escalation, distilled-student refresh), at no extra cost in assumptions over fixed-horizon FC-RAG. Naive composition fails because FC-RAG’s marginal coverage bound makes the betting e-process a non-supermartingale on adverse calibration draws, and Ville’s inequality cannot be invoked. We give Anytime-FC-RAG, a sequential extension built on a summable per-step calibration-deviation budget that converts the marginal bound into a strict conditional bound on a calibration-good event, paired with a truncated betting e-process that is a nonnegative supermartingale on the entire probability space. From these two ingredients, we obtain four guarantees: time-uniform alarm validity \mathbbP(\sup_t E_t \ge 1/\delta_e) \le \delta_e + \delta_\mathrmcal , a Hoeffding-stitched cumulative-miscoverage envelope at the same total budget, safety under any predictable controller (recalibration, bandwidth escalation, student refresh), and training-side error propagation across an unbounded sequence of Federated Probe-Logit Distillation (FPLD) refreshes via a summable training budget. As a practical consequence, an adaptive controller that escalates retrieval bandwidth only when the e-process crosses a warning threshold matches the alarm rate of a fixed-high-bandwidth schedule at substantially lower communication cost. Experiments on a GPT-2-small + MiniLM swarm across MMLU, DBpedia, and AG News verify the predicted alarm rate, detection delay, envelope coverage, and 14 - 57% bandwidth savings; the alarm fires when and only when coverage genuinely breaks.
[LG-136] hree-dimensional Conditional Diffusion Models for Cosmological 21 cm Lightcone Emulation
链接: https://arxiv.org/abs/2605.29016
作者: Bin Xia,John H. Wise
类目: Instrumentation and Methods for Astrophysics (astro-ph.IM); Cosmology and Nongalactic Astrophysics (astro-ph.CO); Machine Learning (cs.LG)
*备注:
Abstract:We investigate conditional diffusion modeling for three-dimensional 21 cm lightcone emulation, focusing on cubes with a sky-plane size of 64\times64 and a line-of-sight depth up to 1024 cells. Relative to earlier 2D studies, the 3D setting is substantially harder because memory limits enforce very small micro-batches while the underlying voxel distribution is highly skewed and long tailed. We perform controlled comparisons across preprocessing choices, dynamic-range compression settings, architecture depth, and training duration using 25,600 training lightcones and validation ensembles at fixed parameter points. For validation, each reference parameter point contains 800 21cmFAST realizations with independent initial conditions, and we use 800 samples per model and per reference set for the reported ensemble comparisons. We evaluate generated lightcones with complementary diagnostics in both image and summary-statistic spaces: brightness-temperature slices, the global signal, the power spectrum, and reduced scattering coefficients. Across the tested configurations, preprocessing is the dominant factor governing stable training and the resulting physical fidelity. Among the configurations explored here, Yeo-Johnson preprocessing combined with moderate amplitude compression gives the most consistently favorable trade-off, with the strongest quantitative support coming from rankings based on the standard-deviation-normalized mean absolute error ( \mathrmMAE_\rm std ) of the global signal and qualitatively compatible behavior in the complementary diagnostics. At the same time, visually plausible 3D samples still retain measurable biases in two-point and higher-order statistics. We therefore view the present work as a simulation-level baseline for three-dimensional 21 cm emulation and for future studies that incorporate more realistic observational effects.
[LG-137] Manifold-based Algorithms for the Hadamard Decomposition
链接: https://arxiv.org/abs/2605.28980
作者: Nicolas Gillis,Subhayan Saha,Stefano Sicilia,Arnaud Vandaele
类目: Optimization and Control (math.OC); Machine Learning (cs.LG); Signal Processing (eess.SP); Numerical Analysis (math.NA)
*备注: 27 pages, code available from this https URL
Abstract:Given a matrix X , and two ranks r_1 and r_2 , the Hadamard decomposition (HD) looks for two low-rank matrices, X_1 of rank r_1 and X_2 of rank r_2 , both of the same size as X , such that X\approx X_1\circ X_2 , where \circ is the Hadamard (element-wise) product. In most cases, HD is more expressive than standard low-rank approximations such as the truncated singular value decomposition (TSVD), as it can represent higher-rank matrices with the same number of parameters; this is because the rank of X_1 \circ X_2 is generically equal to r_1 r_2 . In this paper, we first present some theoretical insights for HD, in particular a useful reformulation X\approx WH^\top where W and H have r_1 r_2 columns and belong to certain manifolds. These allow us to develop three new algorithms for computing HD. The first one uses the representation X\approx X_1\circ X_2 and relies on the Manopt toolbox. The other two rely on the reformulation X\approx WH^\top : one is a block projected gradient method, and the other is a manifold-based gradient descent algorithm that does not require projection onto the feasible set. The last two algorithms are particularly effective for handling large sparse data. We also propose new initializations that allow us to improve the accuracy of the HD. We compare our algorithms and initialization strategies with the TSVD and with the state of the art. Numerical results show that the new methods are efficient and competitive on both synthetic and real data.
[LG-138] Dynamics of Stochastic Momentum with Sparse Updates in High Dimensions
链接: https://arxiv.org/abs/2605.28961
作者: Katie Everett,Elliot Paquette
类目: Machine Learning (stat.ML); Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注:
Abstract:Existing theory of momentum assumes that gradients arrive at every parameter at a roughly constant rate, an assumption violated in practice by heavy-tailed data distributions and modern architectures. We theoretically analyze the dynamics of two tractable models of momentum under sparse updates: a least squares model with sparse inputs and a logistic regression model with a rare class. Both admit exact closed-form second-moment dynamics whose high-dimensional limits we characterize across three scaling exponents for sparsity, batch size, and momentum decay. The phase structure on both problems is governed by the ratio of two intrinsic timescales: a momentum retention timescale (how many active updates the buffer survives) and a learning timescale (how many active updates it takes to reduce the squared error). When learning is much slower than retention, the limit matches SGD; when learning is faster, the system is unstable; where the timescales coincide, we recover classical heavy-ball dynamics. The oscillatory dynamics occur at different momentum values for different token sparsity, creating a spectral conflict for global momentum across token frequencies.
[LG-139] Neural Scaling Laws for Jet Generation
链接: https://arxiv.org/abs/2605.28940
作者: Oz Amram,Darius A. Faroughy,Tjarko Gerdes,Anna Hallin,Gregor Kasieczka,Michael Krämer,Humberto Reyes-Gonzalez,David Shih
类目: High Energy Physics - Phenomenology (hep-ph); Machine Learning (cs.LG); High Energy Physics - Experiment (hep-ex); Data Analysis, Statistics and Probability (physics.data-an)
*备注:
Abstract:Recently observed empirical scaling laws describe the performance of foundation-type models as three independent key quantities – dataset size, compute, and model parameters – are modified. Extracting these scaling laws informs the training of large complex models for which the tuning of hyperparameters in traditional ways is not feasible. This work for the first time explores if scaling laws can also be observed for the task of particle jet generation – both relevant as a pre-training objective for foundation models and as in-situ simulation by itself. We indeed replicate the key logarithmic scaling law behavior for model-size scaling. Beyond studying the next token prediction validation loss of the generative model, we also study the sliced Wasserstein distance of five physical quantities that are not immediately available to the model during training. Our study shows that this quantity is monotonically related to the next token prediction validation loss, meaning that this loss is indeed a good proxy for the physics performance. For the scaling with dataset size and compute, we observe substantially weaker scaling behavior of both the loss and the sliced Wasserstein distance. We analyze this behavior by introducing the concept of a learnable window, and argue that autoregressive next token prediction on jet constituents exhibits comparatively rapid saturation relative to language-model studies. We discuss possible origins of this behavior, including the stochastic nature of QCD radiation and differences between generative and supervised learning tasks in collider physics.
[LG-140] Computational Modeling of Antibody-Antigen Complexes: PLM-Based and MSA-Based Approaches
链接: https://arxiv.org/abs/2605.28886
作者: Xiao Luo
类目: Quantitative Methods (q-bio.QM); Machine Learning (cs.LG)
*备注: PhD thesis
Abstract:Antibodies play a central role in the immune response by specifically recognizing and neutralizing antigens, and therapeutic antibodies have become major drugs for cancer and autoimmune diseases. However, their discovery still relies on extensive in vitro screening, and accurate computational modeling of antibody structures and antibody-antigen interactions can prioritize candidates, reduce experimental burden, and accelerate rational design. Despite recent advances in high-accuracy protein and complex prediction, a persistent performance gap remains for antibody-related tasks compared with general protein-protein interactions, limiting downstream design. This thesis investigates why antibody-related tasks are harder and proposes improvements along two complementary directions. First, we investigate protein language model (PLM)-based methods for antibody and antibody-antigen structure prediction. Using embeddings from multiple PLMs, our approach achieves the best CDR-H3 accuracy among compared PLM-based methods on antibody monomer prediction. Extending it to complex prediction does not generalize: without co-evolutionary signals between antibody and antigen, single-sequence PLM representations do not reliably identify binding interfaces. Second, we develop two MSA-based interventions for antibody-antigen complex prediction: MSA refinement, which combines CDR-focused filtering with depth recovery from a larger sequence database, and convergence-aware recycling, which selects a stable intermediate recycle state for final diffusion sampling. Together, these interventions provide consistent gains over the AlphaFold3 baseline on a held-out antibody-antigen test set. Because the methods modify MSA construction and recycling behavior rather than model parameters, they apply without retraining or weight access. Comments: PhD thesis Subjects: Quantitative Methods (q-bio.QM); Machine Learning (cs.LG) Cite as: arXiv:2605.28886 [q-bio.QM] (or arXiv:2605.28886v1 [q-bio.QM] for this version) https://doi.org/10.48550/arXiv.2605.28886 Focus to learn more arXiv-issued DOI via DataCite
[LG-141] Comment on “Spin-1/2 Kagome Heisenberg Antiferromagnet: Machine Learning Discovery of the Spinon Pair-Density-Wave Ground State”
链接: https://arxiv.org/abs/2605.28861
作者: Helia Kamal,Dominik Kufel,DinhDuy Vu,Chris R. Laumann,Norman Y. Yao
类目: rongly Correlated Electrons (cond-mat.str-el); Disordered Systems and Neural Networks (cond-mat.dis-nn); Machine Learning (cs.LG)
*备注: 3 pages, 1 figure; Comment on arXiv:2401.02866
Abstract:A recent article [Phys. Rev. X 15, 011047 (2025)] utilizes group-equivariant convolutional neural networks to study the ground state of the kagome Heisenberg antiferromagnet. On the largest finite-size cluster studied to date ( N=108 ), the authors report variational energies significantly lower than other numerical methods, including state-of-the-art density matrix renormalization group (DMRG) calculations. In contrast to previous results suggesting a possible spin-liquid ground state, the authors observe a spinon pair-density-wave ground state. We find that: (i) the reported low energies are artifacts of broken ergodicity in the Metropolis–Hastings sampling, since the single-spin-flip update rule utilized by the authors effectively freezes the Markov chains; and (ii) when ergodic sampling is enforced via spin-exchange updates, the neural network converges to energies significantly higher than existing DMRG results, calling the paper’s claims into question.
[LG-142] Financially Guided Deep Portfolio Optimization
链接: https://arxiv.org/abs/2605.28853
作者: Rahul Fernandes,Travis Desell
类目: Portfolio Management (q-fin.PM); Machine Learning (cs.LG)
*备注:
Abstract:Portfolio optimization in real-world financial markets is notoriously difficult due to non-stationarity, noisy data, and high transaction costs. Standard predict-then-optimize methods first forecast returns and then solve for weights, compounding prediction errors and often failing under regime shifts. We propose an end-to-end framework that directly optimizes differentiable surrogates of key financial metrics - Sharpe ratio, Omega ratio, Conditional Value-at-Risk (CVaR), and Risk Parity - allowing neural networks to learn portfolio weights via backpropagation. Our expanding-window walk-forward procedure, applied to 50 SP 500 stocks from 2007 to 2023, incorporates realistic bid-ask spread costs and rebalances quarterly. On the challenging out-of-sample test period (2022-2023), the best model - an AttentionLSTM with the Omega-CVaR-RiskParity loss - achieves an annualized Sharpe of 0.29 and a total compounded return of +7.86%, while the SP 500 delivers -4.52% total return and an annualized Sharpe of -0.02. This outperforms the SP 500 by 12.38 percentage points (a relative improvement of over 270%), while keeping tail risk (CVaR) nearly unchanged. The framework consistently outperforms the equal-weight portfolio, SP 500, and traditional methods (MVP, HRP, NCO), demonstrating that embedding financial objectives directly into model training yields robust, economically meaningful outperformance even in adverse market conditions.
[LG-143] owards a Foundation Model for the Martian Atmosphere
链接: https://arxiv.org/abs/2605.28851
作者: Sujit Roy,Udayshankar Nair,Yuling Wu,Georgios Priftis,Liping Wang,Anastasia Georgiou,Anne Jones,Björn Lütjens,Johannes Schmude,Campbell Watson,Rachel A. Slank,Ankur Kumar,Anirbit Mukherjee,Procheta Sen,Ramin Lolachi,Haonan Chen,Manil Maskey,Juan Bernabé-Moreno,Rahul Ramachandran
类目: Earth and Planetary Astrophysics (astro-ph.EP); Instrumentation and Methods for Astrophysics (astro-ph.IM); Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph)
*备注:
Abstract:The martian atmosphere hosts dynamical phenomena ranging from planet-encircling dust storms to mesoscale orographic clouds and nocturnal low-level jets. General circulation model show capability to simulate these phenomena, but is computationally expensive at resolution needed to resolve mesoscale features. While assimilation of satellite remote sensing observation enable forecasting capabilities using such models, observation record is often sparse, short and fragmented across instrument generators. These constraints motivate the development of a data-driven foundation model for the Martian atmosphere. Foundation models live in a complex design landscape. There is an interplay between the available data, the physics of the underlying processes and corresponding developments in AI. Even though the idea of a foundation model is to address multiple use cases in a data- and compute-efficient manner, it is important to have a clear picture what applications can sensibly addressed by a single model. The purpose of this paper is to elucidate this design landscape. We discuss available data ranging from atmospheric retrievals to reanalysis datasets as well as existing physical models. Moreover, we identify a wide range of candidate downstream applications. Finally, we consider relevant recent developments in artificial intelligence (AI) that can be leveraged in this context. Here, we put a particular emphasis on AI models for atmospheric physics, data-driven approaches to data assimilation as well as methods to work in a limited data setting. Subjects: Earth and Planetary Astrophysics (astro-ph.EP); Instrumentation and Methods for Astrophysics (astro-ph.IM); Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph) Cite as: arXiv:2605.28851 [astro-ph.EP] (or arXiv:2605.28851v1 [astro-ph.EP] for this version) https://doi.org/10.48550/arXiv.2605.28851 Focus to learn more arXiv-issued DOI via DataCite
附件下载
