本篇博文主要内容为 2026-07-24 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 (2026-07-24)
今日共更新626篇论文,其中:
- 自然语言处理共112篇(Computation and Language (cs.CL))
- 人工智能共261篇(Artificial Intelligence (cs.AI))
- 计算机视觉共109篇(Computer Vision and Pattern Recognition (cs.CV))
- 机器学习共169篇(Machine Learning (cs.LG))
- 多智能体系统共16篇(Multiagent Systems (cs.MA))
- 信息检索共15篇(Information Retrieval (cs.IR))
- 人机交互共22篇(Human-Computer Interaction (cs.HC))
多智能体系统
[MA-0] he Boundaries of Automation: A Theory of Persistent Human Participation
【速读】：该论文旨在挑战自动化进程中一个根深蒂固的假设：即人类仅因当前人工智能（AI）能力不足而被保留在系统中，未来随着AI能力提升，人类将被完全取代。论文指出，即使在高度成熟的生成式AI（Generative AI）系统中，人类参与仍可能具有不可替代性，其根本原因在于三类非技术性、深层性的约束机制：一是技术互补性（technical or complementarity grounds），即人类具备AI无法获取的能力或视角；二是规范性与发展性价值（normative or developmental grounds），即参与过程本身对人类自主性或学习具有内在价值；三是涌现性基础（emergence grounds），即目标本身并非预先明确，而是通过人机互动动态生成。在此类活动中，人类参与不仅是优化执行的手段，更是目标生成的构成性要素。因此，人机共同建构（human–AI co-construction）并非对不完美AI的权宜之计，而是一种持续存在的活动特征。该研究揭示了自动化的概念边界，并为未来AI系统的架构设计、评估标准与伦理框架提供了理论基础。
链接: https://arxiv.org/abs/2607.21547
作者: Fares Fourati,Hinrich Schütze,Eyke Hüllermeier,Iryna Gurevych
机构: TU Darmstadt(达姆施塔特工业大学); LMU Munich(慕尼黑大学); ATHENE; MCML
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Emerging Technologies (cs.ET); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
备注:
Abstract:The rapid progress of AI has intensified the long-standing pursuit of automation: replacing human participation with algorithms wherever possible. Implicit in this pursuit is the assumption that humans remain in the loop only because current AI systems are not yet sufficiently capable. This paper challenges that assumption. Rather than asking how far automation can extend, we ask where its conceptual limits lie and argue that human participation may persist even with highly capable AI systems for three distinct reasons. Technical or complementarity grounds arise when humans contribute capabilities or perspectives unavailable to AI. Normative or developmental grounds arise when participation itself is valuable for human agency or learning. Most importantly, emergence grounds arise from target emergence: in some activities, the target is not fully specified in advance but instead emerges through the interaction itself. In these cases, human participation is not merely a means of improving execution but is constitutive of the target being produced. Human–AI co-construction, understood as the joint production of outcomes by humans and AI systems, is therefore not simply a temporary response to imperfect AI, but a persistent feature of activities whose objectives emerge through participation. This perspective has important implications for the limits of automation and for the design, evaluation, and ethics of future AI systems.
[MA-1] oward Continuous Assurance for the Democratization of AI Agent Creation in Industry
【速读】：该论文旨在解决由非工程用户通过低代码、无代码及对话式开发环境在组织内民主化创建AI代理（AI agent）所引发的可靠性问题。尽管这种趋势促进了本地快速创新，但代理背后复杂的依赖关系（包括模型版本、工具、检索源、权限、提示词、调度任务及外部服务等）在部署后可能发生隐性变化，导致性能退化而无明显预警，进而影响系统稳定性与可用性。其解决方案的关键在于提出一种轻量级的持续保障（continuous-assurance）框架，核心机制包括依赖关系映射、就绪契约（readiness contract）、定期检查、诊断分析以及生命周期治理，通过动态评估代理在预期运行条件下是否仍具备操作就绪状态，实现对代理可靠性的持续监控与主动干预，并结合原型审计器与场景化评估验证了该框架可转化为具体可执行的检查项与可操作的修复建议。
链接: https://arxiv.org/abs/2607.21495
作者: Natan Levy,Harel Berger
机构: 未知
类目: Artificial Intelligence (cs.AI); Emerging Technologies (cs.ET); Multiagent Systems (cs.MA)
备注:
Abstract:AI agents are increasingly created inside organizations by non-engineering users through low-code, no-code, and conversational development environments. This democratization enables rapid local innovation, but it also creates a reliability gap: agents that appear to users as simple productivity artifacts may depend on changing models, tools, retrieval sources, permissions, prompts, schedules, and external services. These dependencies can cause silent degradation long after deployment, even when no user directly modifies the agent. This paper identifies the reliability challenge created by democratized AI agent creation and proposes a lightweight continuous-assurance framework for citizen-created organizational agents. The framework combines dependency mapping, readiness contracts, scheduled checks, diagnostics, and lifecycle governance to assess whether an agent remains operationally ready under expected conditions. We also present an initial prototype auditor and scenario-based assessment showing how the proposed taxonomy can be translated into practical checks and actionable remediation guidance.
[MA-2] Compact Latent Coordination for Autonomous Vehicles at Unsignalized Intersections
【速读】：该论文旨在解决多智能体强化学习（MARL）系统在无信号控制交叉路口中协调自主车辆时面临的挑战，主要包括动作空间的组合爆炸、对特权信息的依赖以及智能体设计的僵化问题。其核心解决方案是提出一种分层深度强化学习架构——主-代理原型计划系统（Master-Agent Proto-plan System, MAPS），其关键在于引入一个中心化的“主代理”（Master agent），通过生成一个紧凑且连续的嵌入向量（即原型计划，proto-plan），以编码全局协调策略；随后，去中心化的“工作代理”（Worker agents）将该原型计划与本地观测信息融合，实现针对具体车辆的控制执行。这种设计实现了战略意图与战术执行的解耦，支持各模块独立优化，同时提升了系统的可扩展性与泛化能力。实验表明，MAPS在HighwayEnv中的72种交叉口配置下均能实现无碰撞导航，并显著降低平均通行时间，且在零样本迁移至五智能体场景时仍保持94%的成功率，验证了基于原型计划的分层学习框架在多车协同中的有效性与鲁棒性。
链接: https://arxiv.org/abs/2607.21488
作者: Gil Lifshits,Igal Bilik,Gilad Katz
机构: Ben-Gurion University of the Negev (本·古里安大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA); Robotics (cs.RO)
备注:
Abstract:Coordinating autonomous vehicles at unsignalized intersections remains a critical challenge for multi-agent reinforcement learning (MARL) systems, which typically struggle with combinatorial action spaces, reliance on privileged information, or rigid agent designs. We propose Master-Agent Proto-plan System (MAPS), a hierarchical deep reinforcement learning (DRL) architecture in which a centralized Master agent generates a compact, continuous embedding, denoted as proto-plan, that encodes a global coordination strategy. Decentralized Worker agents integrate this embedding with local observations to execute vehicle-specific control, decoupling strategic intent from tactical execution and enabling independent optimization of each module. As a proof-of-concept evaluation of this coordination mechanism, we test MAPS across 72 intersection configurations in HighwayEnv. MAPS achieves collision-free navigation while significantly reducing average travel time, outperforming state-of-the-art baselines. The learned proto-plans further exhibit robust generalization: a system trained with three agents achieves a 94% success rate when deployed zero-shot to five-agent scenarios, confirming that proto-plan-based hierarchical learning provides a promising framework for multi-vehicle coordination. Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA); Robotics (cs.RO) Cite as: arXiv:2607.21488 [cs.LG] (or arXiv:2607.21488v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2607.21488 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[MA-3] FedAgent KE: Federated Semantic Knowledge Evolution for Heterogeneous Agents
【速读】：该论文旨在解决当前基于大语言模型（Large Language Model, LLM）的智能体在异构框架间因孤立运行而导致的局部经验碎片化问题，阻碍了跨框架的知识迁移与协同推理能力的提升。其核心挑战在于如何在不共享原始推理轨迹的前提下，实现多个独立智能体系统间的可迁移推理抽象的协同演化。解决方案的关键是提出FedAgentKE——一种轻量级的联邦语义知识演化框架，通过迭代式的语义知识蒸馏、聚合与适应机制，使分布式的异构智能体框架能够在保护数据隐私的基础上，共同进化出通用且可复用的推理抽象，从而推动未来协作式智能体生态系统的可持续发展。
链接: https://arxiv.org/abs/2607.21361
作者: Weihao Li,Jun Bai,Ziyang Song
机构: Northwestern University (西北大学); McGill University (麦吉尔大学); Mila – Quebec AI Institute (蒙特利尔人工智能研究所); Ohio University (俄亥俄大学)
类目: Multiagent Systems (cs.MA)
备注: 9 pages (including appendix)
Abstract:Large language model (LLM)-based agents increasingly rely on reasoning, tool use, and iterative execution, yet existing agent frameworks still operate largely in isolation. While recent memory-based agent systems improve individual agents through local retrieval and workflow reuse, local experiences remain fragmented across isolated agent frameworks, limiting cross-framework knowledge transfer and collaborative reasoning evolution. We propose FedAgentKE, a lightweight framework for Federated Semantic Knowledge Evolution across heterogeneous agents. FedAgentKE enables distributed agent frameworks to collaboratively evolve transferable reasoning abstractions through iterative semantic knowledge distillation, aggregation, and adaptation without sharing raw reasoning trajectories. Experiments demonstrate consistent improvements under both cross-framework and cross-task settings, highlighting the potential of federated semantic knowledge evolution for future collaborative agent ecosystems.
[MA-4] pAI-Econ-claude: A Gated Human-in-the-Loop Multi-Agent Architecture for AI-Assisted Economic Theory Development
【速读】：该论文旨在解决生成式 AI（Generative AI）在社会科学领域（如经济学）研究中缺乏廉价、完整且机器可读的正确性验证信号这一核心问题，进而导致多智能体系统在无法由任一组件最终认证结果的情况下，如何组织生成、批判、协调与人类判断的可靠性难题。其解决方案的关键在于提出 pAI-Econ-claude 架构——一种基于门控机制与人机协同的多智能体系统，通过共享可审查的中间记录实现智能体间的协作；引入专业化“门控”模块以诊断特定失效模式并建议回溯路径，但不承担最终正确性认证责任；同时保留关键决策的人类检查点，确保高成本不可逆决策仍由人类掌控。实验表明，相较于无门控基线，在五项匹配的经济理论任务中，该架构显著降低平均故障严重度（从1.58降至1.16），提升整体实用性（从2.60升至3.10），并在四项任务中获得双盲评估者一致偏好。最大改进源于现实检验剔除错误市场结构假设及证明审查修正错误福利主张，而反例亦揭示过度结构化可能压缩重要经济机制。研究支持一个有限结论：门控监督提升了人工智能辅助经济理论的可审计性，但不能替代形式化验证，且不可逆人类判断的分配策略比纯粹智能体自主性更具设计信息量。
链接: https://arxiv.org/abs/2607.21268
作者: Chen Zhu,Xiaolu Wang,Weilong Zhang
机构: China Agricultural University (中国农业大学); University of Cambridge (剑桥大学)
类目: Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI); General Economics (econ.GN)
备注:
Abstract:In many social-science research tasks, such as economics, LLM-based agents must produce outputs for which no cheap, task-complete, machine-readable correctness signal exists. This creates a distinctive reliability problem for multi-agent systems: how should generation, critique, coordination, and human judgment be organized when no component can certify the final result? We address this problem through pAI-Econ-claude, a gated, human-in-the-loop multi-agent architecture for AI-assisted economic theory development. Agents coordinate through a shared workspace of inspectable intermediate records; specialized gates diagnose targeted failure modes and recommend loopbacks without certifying correctness; and human checkpoints retain authority over decisions that are costly to reverse. We evaluate the architecture on five matched economic-theory tasks against an ungated baseline. Two evaluators blinded to configuration agreed on all five pairwise rankings, preferring the gated architecture in four tasks and the baseline in one. Mean failure severity fell from 1.58 to 1.16, while overall usefulness rose from 2.60 to 3.10. The largest observed gain occurred when a reality check rejected a false market-structure premise and a proof review prompted revision of a false welfare claim. The negative case shows that scaffolding can also compress an economically important mechanism too aggressively. The results support a bounded claim: gated oversight improves the auditability of AI-assisted economic theory without substituting for formal verification, and the allocation of irreversible human judgment is a more informative design variable than pure agent autonomy. The workflow is publicly available at this https URL.
[MA-5] Explainable Belief Harmonization under Dynamic Epistemic Partitions
【速读】：该论文旨在解决多智能体信念融合中因智能体在运行时动态改变观测能力而导致的信念表示结构不一致问题。传统方法通常假设智能体的表征结构是固定的，但在实际场景中，智能体可能在执行过程中获得或丧失感知能力，使得原本可接受的信念状态变得在结构上不可行。为此，论文提出一种形式化框架，用于处理连续信念分布下认知划分（epistemic partition）的运行时变化。其解决方案的关键在于采用混合方法：结合答案集编程（Answer Set Programming, ASP）在知识表达的可扩展性、声明式完整性约束及解释生成方面的优势，与Python在数值计算灵活性上的长处，实现对异构且可能动态变化的分辨率层级下信念状态的统一建模。该框架在细分（refinement）时保证可接受性不变，在粗化（coarsening）时提供唯一保质量修复，并确保解释完备性，实验验证表明其在100次随机拓扑变化场景中均能完全检测违规并提供完整解释。
链接: https://arxiv.org/abs/2607.21210
作者: Adam Kostka(Warsaw University of Technology),Jarosław A. Chudziak(Warsaw University of Technology)
机构: 未知
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:Existing approaches to multi-agent belief combination have established mature foundations for combining uncertain beliefs under common assumptions: consensus methods use iterative averaging, logic-based methods resolve conflicting knowledge bases, and epistemic logic analyzes agents’ information states. Typically, these approaches assume that the structure determining what each agent can represent remains fixed. However, in many scenarios, agents gain or lose observational capacity during execution, and what was once admissible may become structurally impossible. This paper presents a formal framework for handling such runtime changes in epistemic partitions over continuous belief profiles. A hybrid approach exploits the advantages of answer set programming in elaboration tolerance, declarative integrity constraints, and explanations, with the numerical flexibility of Python. The framework applies to domains where agents operate at heterogeneous and possibly changing levels of resolution, and provides formal guarantees of admissibility preservation under refinement, unique mass-preserving repair under coarsening, and explanation completeness. Evaluation across 100 randomly generated topology changes confirms complete violation detection and explanation coverage.
[MA-6] GLP: A Grassroots Multiagent Concurrent Logic Programming Language for AI
【速读】：该论文旨在解决当前主流平台架构中存在的中心化控制与去中心化但权力集中（即“寡头式去中心化”）问题，提出一种能够实现真正平等民主的分布式系统范式——草根平台（grassroots platform）。草根平台的核心特征是允许多个独立运行的实例在无需全局资源依赖的情况下自主形成与运作，并可动态聚合为更大规模的实例，甚至最终形成单一全局实例，同时保持去中心化的治理结构。为实现这一目标，论文提出了一种名为草根逻辑程序（Grassroots Logic Programs, GLP）的多智能体并发逻辑编程语言。其解决方案的关键在于：通过引入多智能体原子事务（multiagent atomic transactions），构建一种新的多智能体操作语义，将标准逻辑程序的执行机制限制为支持并发、分布式且具有强一致性保障的计算模型；在此基础上，证明了GLP具备草根性（grassrootsness），即系统能够在不依赖中央权威的前提下实现自组织、可扩展的协同行为。文中以“草根社会图谱”（grassroots social graph）作为典型应用案例，展示了如何使用GLP实现一个基于逻辑规则和分布式共识的底层草根平台原型。
链接: https://arxiv.org/abs/2607.21189
作者: Ehud Shapiro(London School of Economics)
机构: 未知
类目: Programming Languages (cs.PL); Distributed, Parallel, and Cluster Computing (cs.DC); Logic in Computer Science (cs.LO); Multiagent Systems (cs.MA)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:A grassroots platform is a multiagent distributed system in which multiple independent instances can form and operate independently of each other and of any global resource, yet may coalesce into ever larger instances, possibly resulting in a single global instance. Grassroots platforms aim to offer an egalitarian/democratic alternative to centralised/autocratic and decentralised/plutocratic global platforms. Here, we present Grassroots Logic Programs (GLP), a multiagent concurrent logic programming language designed for the implementation of grassroots platforms: we recall the standard operational semantics of logic programs; introduce the concurrent operational semantics of GLP as its restriction; recall multiagent atomic transactions; use them to introduce a multiagent operational semantics of GLP; and prove multiagent GLP to be grassroots. The grassroots social graph—the foundational grassroots platform on which all others are based—serves as a GLP programming example. Comments: In Proceedings ICLP 2026, arXiv:2607.17707 Subjects: Programming Languages (cs.PL); Distributed, Parallel, and Cluster Computing (cs.DC); Logic in Computer Science (cs.LO); Multiagent Systems (cs.MA) Cite as: arXiv:2607.21189 [cs.PL] (or arXiv:2607.21189v1 [cs.PL] for this version) https://doi.org/10.48550/arXiv.2607.21189 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Journalreference: EPTCS 450, 2026, pp. 119-133 Related DOI: https://doi.org/10.4204/EPTCS.450.11 Focus to learn more DOI(s) linking to related resources
[MA-7] SafeStep: AI-powered Travel Assistance for Elderly People with Frailty or Dementia
【速读】：该论文旨在解决老年人在城市环境中因衰弱或痴呆症导致出行能力严重受限的问题。其核心挑战在于如何为高风险老年用户提供安全、可靠且个性化的出行支持，以降低旅途中的失败风险并提升出行信心。解决方案的关键在于提出一种创新的旅行图（travel graph）表示方法，将路径规划与行为预测建模深度融合：系统通过结合大语言模型（LLM）与Anticip8行为预测引擎，针对旅程中每个阶段生成个性化失败场景；在此基础上，提出针对性干预措施，并评估各干预策略对成功抵达目的地概率的影响；最终选择最优干预方案以最大化出行成功率。实验结果表明，将Anticip8用于失败预测、GPT类模型用于干预评估的组合在性能上表现最为可靠。用户反馈显示，该系统显著提升了出行时的信心与感知安全性，但针对目标人群的界面易用性仍有优化空间。未来工作将聚焦于系统改进与实际部署，并探索该AI技术在心理健康、职业辅导及成瘾治疗等领域的扩展应用。
链接: https://arxiv.org/abs/2607.21156
作者: Elderly People with Frailty or Dementia Azul Debenedetti,David Gamez,Franco Such,Nik Kairinos
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:More than a million people in the UK suffer from frailty or dementia, which severely compromise their ability to travel in urban environments. This paper presents SafeStep, an AI-driven travel system that assists elderly users with their journeys. At the core of SafeStep is a novel travel graph representation, which integrates route planning with predictive modelling. For each stage of a journey, the system (i) generates personalized failure scenarios using a combi-nation of LLMs and the Anticip8 behavioral prediction engine, (ii) proposes targeted interventions, and (iii) estimates the impact of interventions on out-come probabilities. This enables SafeStep to select interventions that maximize the likelihood of the person reaching their destination. SafeStep was evaluated through experiments on travel graph generation and a field study involving 26 real-world journeys. Results showed that combining Anticip8 for failure pre-diction with GPT-based models for intervention evaluation yields the most re-liable performance. User feedback indicated that SafeStep improves confidence and perceived safety during travel, although interface usability needs to be im-proved for the target demographic. In the future, we would like to improve and release SafeStep. The AI system that was developed for SafeStep could be ap-plied in other areas, such as mental health, career coaching and addiction treatment.
[MA-8] HiMe: Real-Time Self-Hosted Personal Agent Platform for Health Insights with Wearable Devices
【速读】：该论文旨在解决传统可穿戴健康信号分析方法在分析框架僵化和个性化能力有限方面的瓶颈问题，尤其针对当前缺乏开源、可本地部署且能实时处理个人健康数据并保障隐私的平台这一关键挑战。其解决方案的核心在于提出HiMe——一个以隐私优先、完全兼容多种可穿戴设备实时健康数据生态系统的本地化部署智能体平台。该平台的关键创新在于遵循三大设计原则：将数据库作为第一类组件进行系统性构建；协同优化有效性与效率，实现低成本的帕累托最优平衡；支持实时数据处理与长期用户建模相结合。这一架构使得个体能够持续、个性化地利用健康智能体进行健康管理，从而显著提升健康监测的适应性与实用性。
链接: https://arxiv.org/abs/2607.21019
作者: Wei Liu,Siya Qi,Linhai Zhang,Lorainne Tudor Car,Yulan He
机构: King’s College London(国王学院); The Alan Turing Institute(艾伦·图灵研究所)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC); Multiagent Systems (cs.MA); Software Engineering (cs.SE)
备注:
Abstract:Traditional approaches to wearable health signal analysis, such as smartwatches, are constrained by rigid analytical frameworks and limited personalisation. The emergence of LLM agents creates a new opportunity for Personal Health Agentic Analysis, where health insights can be generated adaptively and in context. However, currently there is no open-source locally deployable platform capable of processing personal health data in real time while preserving privacy. We present HiMe, a locally deployable, privacy-first agent platform that is fully compatible with real-time health data ecosystems across a wide range of wearable devices. HiMe is guided by three design principles. The database is treated as a first-class component. Effectiveness and efficiency are jointly optimised to achieve a low-cost Pareto-optimal balance. Data are processed in real time while the user is modelled over the long term. Together, these principles make it practical for individuals to harness Personal Health Agents for continuous, personalised health monitoring for better wellbeing.
[MA-9] Safe and Scalable Multi-Drone Payload Transport via CBF-based Reinforcement Learning with Zero-Shot Sim-to-Real Transfer
【速读】：该论文旨在解决多无人机协同吊运任务中因无人机、缆绳与载荷之间复杂的耦合动力学带来的安全性和可扩展性挑战，尤其在动态非结构化环境中的应用受限问题。其核心解决方案是提出一种基于学习的框架，通过引入一种最小化的二维抽象模型，在保留任务相关无人机-载荷耦合关系以确保协调与安全性的同时，显著降低计算复杂度，支持大规模学习。关键创新在于采用基于团队规模和物理参数域随机化的训练策略，结合离散图控制屏障函数近端策略优化（DGPPO）算法，训练出完全分布式的控制策略，实现了无需微调的零样本仿真到现实世界的迁移。实验证明，单一学习策略可在不同团队规模和任务场景下泛化，并能在包含其他移动无人机团队作为动态障碍物的复杂硬件环境中安全运行，验证了该框架在效率、安全性和实际适应性方面的优越性能。
链接: https://arxiv.org/abs/2607.20665
作者: Jaeyoun Choi,Oswin So,Songyuan Zhang,Cooper Taylor,Chuchu Fan
机构: Massachusetts Institute of Technology (麻省理工学院); Korea Institute for Advancement of Technology (韩国技术进步研究所)
类目: Robotics (cs.RO); Multiagent Systems (cs.MA); Systems and Control (eess.SY)
备注: Published in IEEE Robotics and Automation Letters (Early Access), 2026
Abstract:Multi-drone payload transportation has emerged as a promising research paradigm with potential applications in construction, logistics, and disaster response. However, the complex coupled dynamics among drones, cables, and payloads pose significant challenges, and existing approaches remain limited in safety and scalability, particularly in dynamic and unstructured environments. In this work, we propose a learning-based framework for safe and scalable multi-drone cooperative payload transport. We introduce a minimal 2D abstraction that preserves the task-relevant drone-payload coupling required for coordination and safety, while remaining computationally efficient for large-scale learning. Using domain randomization over team size and physical parameters, we train a fully distributed policy via Discrete Graph Control Barrier Function Proximal Policy Optimization (DGPPO), enabling robust zero-shot sim-to-real transfer without fine-tuning. Extensive real-world evaluations demonstrate that a single learned policy generalizes across varying team sizes and task scenarios. Furthermore, multi-group hardware experiments show that the same policy can safely operate in dynamic environments, where other drone teams act as moving obstacles. These results indicate that the proposed framework enables efficient, safe, and scalable multi-drone payload transportation with strong generalization to complex real-world conditions.
[MA-10] Bayesian uncertainty estimation improves clinical decision making in medical AI agents
【速读】：该论文旨在解决医学图像分析中机器学习模型缺乏可靠置信度度量的问题，尤其是在面对模糊或非典型病例时，模型的不确定性难以被有效评估。其核心解决方案是采用蒙特卡洛丢弃（Monte Carlo dropout）方法，在一个多任务胸部X光分类模型（涵盖八种胸腔异常，基于137,593张训练图像）中生成一种**认知不确定性（epistemic uncertainty）**信号。该信号能够反映模型在不同训练集规模下的泛化能力，并有效识别出看似自信但存在错误风险的预测。实验表明，将此不确定性信号与点预测结合后，错误检测的AUROC从0.74提升至0.77（ΔAUROC +0.023，95%置信区间 [+0.014, +0.033]）。进一步在受控的2×2因子实验中发现，临床决策支持系统仅在接收二元错误风险标志（binary error-risk flag）而非原始不确定性分数时，能显著降低高置信度误诊率——从8.5%降至2.7%。由此可见，认知不确定性估计不仅包含超越点预测的决策相关信息，且其下游应用价值高度依赖于不确定性信息的表达方式。
链接: https://arxiv.org/abs/2607.20582
作者: Frederik Hauke,Patrick Wienholt,Christiane Kuhl,Dyke Ferber,Jakob Nikolas Kather,Sven Nebelung,Daniel Truhn
机构: 未知
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Machine learning models for medical image analysis typically lack a reliable measure of confidence, limiting their use in ambiguous or atypical cases. Here we show that Monte Carlo dropout, applied to a multi-task chest-radiograph classifier (eight thoracic findings, 137,593 training images), provides an epistemic uncertainty signal that tracks generalisation across training-set scales and flags confident yet error-prone predictions. Adding this signal to the point prediction raised error-detection AUROC from 0.74 to 0.77 ( \Delta AUROC +0.023, 95% CI [+0.014, +0.033]). In a controlled 2x2 factorial experiment, a clinical-decision-support agent exploited this uncertainty only when it was delivered as a binary error-risk flag rather than as raw scores, cutting confident misdiagnoses on unreliable findings from 8.5% to 2.7%. Epistemic uncertainty estimation thus carries decision-relevant information beyond point predictions, but its value for downstream agents depends on how it is communicated.
[MA-11] AppWorld-UL: Benchmarking Diverse Agent -User Interactions for Tool-Use ICML2026
【速读】：该论文旨在解决现有评估基准在衡量工具使用智能体（tool-use agents）与用户交互能力方面的不足，特别是无法充分反映真实场景中复杂多样的人机交互需求。当前基准普遍存在环境规模小、API数量有限且多数为非状态改变型接口的问题，难以模拟日常数字任务中所需的动态协商、澄清提问、确认请求及失败提示等关键交互行为。为此，论文提出AppWorld-UL，一个基于“用户在回路中”（user-in-the-loop）设计的新型基准，包含516个具有挑战性的任务，覆盖9个主流模拟应用（如Amazon、Spotify），并通过系统性引入语义模糊性与约束条件，迫使智能体采取多样化的交互策略。其核心创新在于采用大语言模型（LLM）作为用户代理，通过设定精细的知识边界进行响应生成，从而实现比以往无约束或过度刚性方案更可靠、更具代表性的用户行为模拟。实验结果表明，即使是最先进的大模型Claude Opus 4.7，在整体任务上的成功率也仅为48.6%，在更复杂的组合式任务子集上降至35.7%，而在严格的场景级评估指标下进一步下降至21.3%。分析揭示，准确的用户交互策略是任务成功的关键因素，凸显了该基准的高难度与推动“用户在回路中”工具使用智能体研究发展的潜力。
链接: https://arxiv.org/abs/2607.20536
作者: Junzhi Chen,Harsh Trivedi,Jane Pan,Michael JQ Zhang,Tejas Srinivasan,Niranjan Balasubramanian,Ashish Sabharwal
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: ICML 2026
Abstract:Tool-use agents that address day-to-day digital tasks such as ordering groceries must not only operate applications, but also interact with the user, e.g., to ask clarification questions, prompt for confirmation, and inform the user when the instruction is infeasible. However, current benchmarks for evaluating agent-user interactions do not capture the diversity of such interactions. Further, they operate in small environments with few, often non-state-changing, APIs. To address this gap, we introduce AppWorld-UL, a ``user-in-the-loop’’ benchmark of 516 challenging tasks requiring diverse agent-user interactions. Building upon the AppWorld framework with 9 popular simulated apps like Amazon and Spotify, we systematically modify original tasks to introduce ambiguities and constraints that necessitate various types of agent-user interaction. User behavior is simulated by an LLM prompted to respond with carefully designed knowledge boundaries, offering more reliable simulation than the unconstrained or overly rigid alternatives used in prior work. Our evaluation reveals that a state-of-the-art LLM, Claude Opus 4.7, achieves only 48.6% success on AppWorld-UL, and only 35.7% on the harder, compositional subset. On the stricter, scenario-level metric, compositional task performance drops to only 21.3%. Our analysis reveals that correct user-interaction is crucial for success. This demonstrates the benchmark’s difficulty and its potential to advance research on user-in-the-loop tool-use agents.
[MA-12] MKEvolve: A Modular Multi-Agent Framework for Kernel Code Generation
【速读】：该论文旨在解决大语言模型（LLM）在生成硬件加速器上高效且正确的计算内核时面临的挑战，尤其是在复杂深度学习模型中，如何实现高正确率与高性能的代码生成。其核心问题在于，直接对整个模块进行端到端的代码合成难以保证生成代码的正确性与可优化性，且对LLM的上下文长度和计算资源要求较高。解决方案的关键在于提出MKEvolve（Modular Kernel Evolve）框架，通过迭代式协同演化（co-evolution）机制，将复杂的PyTorch模块分解为多个可独立优化的子模块，并利用LLM驱动的束搜索（beam search）分别优化每个子内核，同时在迭代过程中动态调整模块分解结构（通过拆分与合并操作）。该方法生成的内核是可配置、可解释且易于迁移的程序化组合体，不仅提升了代码正确性和性能表现，还显著降低了LLM的词元（token）使用量，实验表明在Triton上的基准测试中相较端到端基线平均减少35%的词元消耗。
链接: https://arxiv.org/abs/2607.20501
作者: Jason Yoo,Rajarshi Saha,Shaowei Zhu,Tao Yu,Wei Tang,Youngsuk Park
机构: 未知
类目: Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Despite rapid progress in LLM-based code generation, writing correct and performant kernels for hardware accelerators remains a key bottleneck in scaling modern ML workloads. We present MKEvolve (Modular Kernel Evolve), a framework that iteratively co-evolves a modular decomposition of complex PyTorch modules and the LLM-generated kernel for each submodule, refining the decomposition by splitting and fusing across iterations while independently improving each subkernel via LLM-driven beam search. The resulting kernels are programmatic compositions of independently verified subkernels, making them configurable (subkernel implementations are swappable), interpretable (errors and speedups are traceable to specific subkernels), and readily adaptable to related model architectures. Experiments with Triton on KernelBench L2 and L3, spanning multi-operator sequences and full model architectures, show that MKEvolve improves both correctness and speedup over end-to-end direct synthesis baselines while reducing LLM token usage by up to 35%.
[MA-13] Workload-Aware Caching for Multi-Agent Systems
【速读】：该论文旨在解决多智能体系统（multi-agent systems）在执行复杂任务时，因缺乏针对其特有结构与工作负载特征的缓存淘汰策略而导致的缓存效率低下问题。现有缓存淘汰策略通常仅依据访问历史对所有缓存条目进行统一处理，忽视了智能体执行环境中的结构性信号（如任务依赖关系）和工作负载特性（如重计算成本、智能体调用频率）。为此，论文提出一种工作负载感知的缓存淘汰策略，其核心在于将三个关键信号——重计算成本（recomputation cost）、DAG依赖度（DAG dependency count）和智能体调用频率（agent invocation frequency）——融合为一个统一的评分函数，以在内存受限条件下优先保留最具价值的中间结果。实验表明，该方法在三个涵盖不同复用模式的多智能体基准测试中，相较于无缓存基线最多降低64.7%延迟，并在平均上较最优有限容量基线进一步减少31.1%延迟，同时接近无限缓存性能，且保持与或优于所有对比的有限容量方法的准确性。此外，研究还验证了工作负载感知的内容缓存与其他优化技术（如计划级缓存和并行智能体执行）具有互补性，分别针对多智能体流水线中的不同效率瓶颈。
链接: https://arxiv.org/abs/2607.20495
作者: Anas Mohamed,Kaizan Haque,Azal Ahmad Khan,Chetan Sharma,Shuwen Ge,Ali Anwar
机构: University of California, Berkeley (加州大学伯克利分校); Google(谷歌); Stanford University (斯坦福大学)
类目: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC); Multiagent Systems (cs.MA)
备注: 11 pages, 6 figures
Abstract:Multi-agent systems decompose complex tasks into directed acyclic graphs (DAGs) of specialized agent executions, creating natural opportunities for caching intermediate results across queries. However, existing cache eviction policies treat all cached entries uniformly based on access history, ignoring structural and workload signals uniquely available in agentic execution environments. We present a workload-aware eviction policy that combines three signals, namely recomputation cost, DAG dependency count, and agent invocation frequency, into a unified scoring function that retains the most valuable entries under memory constraints. Evaluated across three multi-agent benchmarks spanning diverse reuse regimes, our policy reduces latency by up to 64.7% relative to the uncached baseline and achieves on average a 31.1% latency reduction over the next best finite-capacity baseline, while approaching the performance of an unbounded cache and maintaining accuracy on par with or exceeding all competing finite-capacity methods. We further show that workload-aware content caching is complementary to other agentic system optimization methods, including plan-level caching and parallel agent execution, with each technique targeting a distinct efficiency bottleneck in multi-agent pipelines.
[MA-14] Human-in-the-Loop Large Language Model Framework for Identification of Cutaneous Immune-Related Adverse Events
【速读】：该论文旨在解决临床笔记中皮肤免疫相关不良事件（cutaneous immune-related adverse events, cirAEs）自动检测的准确性与效率问题。传统人工审阅存在主观差异大、耗时长及可扩展性差等局限，难以满足大规模真实世界数据中不良事件识别的需求。为此，研究提出一种基于检索增强的多智能体大语言模型（retrieval-augmented, multi-agent large language model, LLM-driven）人机协同框架，其核心创新在于融合外部知识检索机制与多智能体协作推理结构，通过引入外部医学知识库提升生成结果的准确性，并利用多智能体分工协作实现对复杂临床文本的分步解析与交叉验证。该方案显著提升了检测准确率（F1从0.77提升至0.88）、评审者间一致性（Cohen’s kappa从0.50提升至0.82），并使平均审阅时间减少约一半，验证了生成式AI在跨器官系统免疫毒性识别中的可行性，为实现精准、可扩展且透明的不良事件数据提取提供了有效路径。
链接: https://arxiv.org/abs/2607.20428
作者: Charles Lu,Olivia Burke,Debby Cheng,Adam Kashlan,Caitlyn Duffy,Zeyun Lu,Lirit Fuksman,Jin Ning Tian,Andrew Sedlack,Priya Katyal,Eudora Lee,Ralina Karagenova,Chuck Lin,Kun-Hsing Yu,Nicole LeBoeuf,Alexander Gusev,Yevgeniy R. Semenov
机构: 未知
类目: Computation and Language (cs.CL); Human-Computer Interaction (cs.HC); Multiagent Systems (cs.MA)
备注:
Abstract:This study evaluated a retrieval-augmented, multi-agent large language model (LLM)-driven, human-in-the-loop framework for detecting cutaneous immune-related adverse events (cirAEs) from clinical notes. Compared with unassisted manual review, the LLM-assisted workflow improved accuracy (F1 = 0.88 vs 0.77), inter-rater agreement measured by Cohen’s kappa (kappa = 0.82 vs 0.50), and reduced average review time by approximately half. This framework pilots how LLMs can be applied to identify immune-related toxicities across organ systems and, more broadly, enable accurate, scalable, and transparent adverse event data extraction.
[MA-15] ChannelGuard: Safe Models Do Not Compose into Safe Multi-Agent Systems
【速读】：该论文旨在解决多智能体大语言模型（Multi-agent LLM）应用中，各智能体之间通信通道缺乏监控所带来的安全漏洞问题。现有防御机制仅在输入边界进行防护（如IBProtector、Llama Guard、困惑度过滤器、SmoothLLM），或以黑盒、随机性方式在应用外部运行于服务端，无法有效防范攻击者通过智能体间通信链路隐蔽注入恶意指令。研究发现，当前主流评估方式存在严重缺陷：在对8类攻击、5种防御方案、3种模型后端的2,100条调用轨迹评估中，看似“完全安全”的未受保护流水线实际上90%的安全性依赖于云服务商的服务端过滤器（如Azure GPT-5中的54/60个阻断来自服务端过滤），一旦后端无此类过滤，系统安全性即完全转移至模型自身对齐能力，而这一依赖关系在仅报告最终结果的评估范式下被严重掩盖。为此，论文提出ChannelGuard——一种无需训练的纵深防御框架，其核心在于在每条智能体间通信通道上部署信息瓶颈门控机制（information-bottleneck gates），通过嵌入相似度匹配对抗短语库，实现确定性通过、压缩或阻断文本，不引入额外大语言模型调用，并结合归因方法追踪攻击拦截位置。实验表明，ChannelGuard在应用层对工具污染攻击（Tool Poisoning）实现了30/30的100%阻断效果，且在Azure GPT-5、Anthropic Sonnet 4.5和Haiku 4.5三个后端表现一致；同时将提示注入攻击成功率从0.333降低至0.167（降幅50%），并保持GSM8K基准准确率不变（0.867）。尽管白盒自适应改写攻击可绕过嵌入式门控，但相比基线“扰动+投票”策略仍表现更差，验证了其有效性。附录进一步提供了基线对比、消融实验、参数扫描、良性内容保留分析及人工评判（kappa = 0.900），总成本控制在47.36美元。
链接: https://arxiv.org/abs/2607.19430
作者: Elias Hossain,Md Mehedi Hasan Nipu,Fatema Tuj Johora Faria,Tasfia Nuzhat Ornee,Maleeha Sheikh
机构: University of Central Florida(中佛罗里达大学); North South University(南大学); Ahsanullah University of Science and Technology(阿桑努拉科技大学); Purdue University Fort Wayne(普渡大学韦恩堡分校)
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
备注:
Abstract:Multi-agent LLM applications chain a planner, worker agents, a verifier, and a synthesizer, and every hop between agents is an unmonitored channel through which an adversary can smuggle instructions. Existing defenses guard only the input boundary (IBProtector, Llama Guard, perplexity filters, SmoothLLM) or run outside the application as opaque, stochastic provider-side filters. We show this gap carries a consequence rarely measured: on a 2,100-trace evaluation across eight attack families, five defenses, and three model backends, an undefended pipeline that appears fully safe under standard reporting (attack success 0.000 on tool- and memory-poisoning) owes that safety almost entirely to the cloud provider’s server-side filter (54 of 60 blocks on Azure GPT-5), and silently shifts to the agent model’s own alignment on a backend without such a filter. Outcome-only reporting hides this dependence. We present ChannelGuard, a training-free defense-in-depth framework placing information-bottleneck gates on every inter-agent channel; each scores channel text against an adversarial phrase bank by embedding similarity and deterministically passes, compresses, or blocks it, adding no LLM call, while an attribution method records which layer stopped each attack. ChannelGuard’s tool-output gate blocks Tool Poisoning 30 of 30 at the application layer, identically across Azure GPT-5, Anthropic Sonnet 4.5, and Anthropic Haiku 4.5, whereas the undefended pipeline shifts entirely across backends; it also lowers Prompt Injection attack success by half (0.333 to 0.167) and preserves GSM8K accuracy exactly (0.867). White-box adaptive paraphrase evades every embedding gate, where a perturb-and-vote baseline does better. An extended appendix adds baselines, ablations, sweeps, a benign-preservation analysis, and a judge audit (kappa = 0.900), at a total cost of 47.36 USD.
自然语言处理
[NLP-0] Surprisal Theory is Tautological (without Rational Grounding)
【速读】： 该论文旨在解决生成式语言模型中“意外度理论”（surprisal theory）的可证伪性问题。其核心关切在于，传统意外度理论认为人类在特定语境下对语言单位的处理难度是语言模型所计算的意外度（surprisal）的仿射函数，但作者指出这一主张本质上是一个同义反复（tautology）：在满足一定技术条件下，对于任意非负的处理难度度量，总存在一个语言模型使其意外度与该难度呈仿射关系。因此，若不施加额外约束于语言模型，则该理论无法做出可被经验检验的预测。这一同义反复长期被心理学研究中的隐含假设所掩盖——即相关语言模型应为训练语料库的生成分布，从而提升语料拟合度即可增强对人类行为的预测能力。然而，近期实证研究表明，更优的语料拟合模型反而可能降低对处理难度的预测效度。论文由此得出结论：要打破该同义反复，必须引入理性主义干预，即语言模型应基于非经验驱动的认知机制（如记忆限制或加工目标等），而非依赖于待解释的人类行为数据本身，从而实现对处理难度的真正因果解释。
链接: https://arxiv.org/abs/2607.21574
作者: Ryan Cotterell
机构: ETH Zürich (苏黎世联邦理工学院)
类目: Computation and Language (cs.CL)
备注: Under “Review” at ARR
Abstract:Surprisal theory holds that the human processing difficulty of a linguistic unit in context is an affine function of its surprisal under some language model. I argue this claim is a tautology without further constraint: for any non-negative difficulty measure over units in context, there exists a language model whose surprisal is an affine function of it under mild technical conditions. Therefore, because any pattern of difficulty is consistent with some language model, without an additional constraint on the language model, surprisal theory makes no falsifiable predictions. The tautology was long obscured by an assumption implicit in two decades of psycholinguistic work—that the relevant language model is the distribution that generated the training corpus, so that improving corpus fit improves predictions of human behavior. Recent empirical work has undermined this assumption, demonstrating that better corpus models can be worse predictors of processing difficulty. I conclude that breaking the tautology requires a rationalist intervention, i.e., the relevant language model must be derived from a non-empirically motivated model of the comprehender, which could be based on, for instance, memory constraints or processing goals, and that, thus, does not depend on the behavioral data surprisal theory is meant to explain.
[NLP-1] OpenForgeRL: Train Harness-native Agents in Any Environment
【速读】： 该论文旨在解决当前生成式AI代理（Generative AI Agent）在复杂推理、工具调用及外部系统交互中，因依赖高度复杂的推理框架（如Claude Code、Codex、OpenClaw）而导致难以通过开放基础设施实现端到端训练的问题。现有监督微调（SFT）与强化学习（RL）体系无法原生支持有状态、多进程的推理框架，限制了模型在真实环境中的可训练性与可复现性。其解决方案的关键在于提出OpenForgeRL——一个开源的端到端训练框架，通过引入轻量级代理（proxy）记录推理框架的模型调用作为训练数据，并结合Kubernetes编排器为每次训练轮次分配独立远程容器，从而实现对任意推理框架和环境的大规模、可扩展训练。该框架通过解耦训练与推理过程，使研究人员能够直接在真实部署环境中训练、评估和优化代理。实验表明，OpenForgeRL在多种复杂任务场景下均显著优于同规模开源基线，尤其在图形用户界面（GUI）代理任务中表现突出，达到或超越数倍更大的模型性能；同时分析揭示了不同推理框架对学习难度的影响，以及强化学习对提升代理可靠性（如自我验证、工具覆盖率、多步计划完成率）的积极作用，但对错误恢复等关键能力仍存在局限。
链接: https://arxiv.org/abs/2607.21557
作者: Xiao Yu,Baolin Peng,Ruize Xu,Hao Zou,Qianhui Wu,Hao Cheng,Wenlin Yao,Nikhil Singh,Zhou Yu,Jianfeng Gao
机构: Columbia University (哥伦比亚大学); Dartmouth College (达特茅斯学院); Microsoft Research (微软研究院)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Modern AI agents rely on elaborate inference harnesses such as Claude Code, Codex, and OpenClaw to drive multi-turn reasoning, tool use, and access to external systems. While powerful, these complex harnesses also make agents hard to train end-to-end with open infrastructure, whose SFT/RL stacks cannot natively express stateful, multi-process harness inference. To address this, we present OpenForgeRL, an open-source framework for training harness-based agents end-to-end in diverse environments. OpenForgeRL achieves this with a lightweight proxy that serves the harness’s model calls while recording them as training data for a standard RL codebase (e.g., veRL), and a Kubernetes orchestrator that runs each rollout in its own remote container, together enabling training on any harness in any environment at scale. By decoupling training and inference, OpenForgeRL allows researchers to easily train, study, and improve agents directly in the real harnesses and environments they are deployed with. We validate our framework across diverse, complex harnesses and environments, spanning tool/claw-based agents and multimodal GUI browser- and computer-use agents. Using only hundreds to a few thousand tasks, OpenForgeClaw reaches 31.7 pass^3 and 55.9 pass@3 on ClawEval and 33.7 on QwenClawBench. OpenForgeGUI reaches 37.7 on OSWorld-Verified, 63.0 on Online-Mind2Web, and 72.3 on WebVoyager. Both outperform open baselines of similar size on nearly all benchmarks, and in the GUI setting match or surpass models several times larger. Beyond benchmarks, we analyze how harness choice (e.g., ZeroClaw, OpenClaw, Codex) and RL shape agent behavior. We find that some harnesses are substantially harder to learn than others, and that RL improves agentic reliability, such as self-verification, tool coverage, and completing multi-step plans, though critical abilities such as error recovery remain weak.
[NLP-2] DONDO: Open w2v-BERT Speech-Recognition Base Models for African Languages
【速读】： 该论文旨在解决非洲语言在自动语音识别（ASR）领域中普遍存在的标注数据匮乏问题，尤其是针对缺乏大规模、高质量转录语音数据的低资源语言。其核心挑战在于如何在有限标注数据条件下构建高性能且可扩展的ASR模型。解决方案的关键在于提出DONDO系列开放、宽松许可的ASR基础模型，基于w2v-BERT 2.0自监督语音编码器构建，利用宗教文本中的朗读语音作为主要训练数据源，以实现对多种非洲语言的广泛覆盖与语义一致性。该方案采用两步（或三步）学习率退火微调策略，先以高学习率快速适应共享多语言模型，再逐步降低学习率以恢复并部分超越单语基线性能；同时引入轻量级语言条件化机制，通过在声学特征前添加独热编码的语言标识符作为前缀帧，使单一多语言检查点可在推理时灵活导向特定目标语言。实验表明，五类多语言模型在平均词错误率（WER）上达到10-13%，显著缩小了与单语模型之间的差距，同时实现一个模型覆盖多个语言的目标。所有模型均以Apache-2.0许可证开源发布于Hugging Face KhayaAI组织，支持自由微调及商业应用，涵盖约一亿母语使用者，若计入第二语言使用者则规模更为庞大。
链接: https://arxiv.org/abs/2607.21540
作者: Paul Azunre
机构: Khaya AI( Khaya AI); Hugging Face( Hugging Face); Meta( Meta)
类目: Computation and Language (cs.CL)
备注:
Abstract:We present DONDO, a family of open, permissively licensed automatic speech recognition (ASR) base models for African languages, built on the w2v-BERT 2.0 self-supervised speech encoder. DONDO comprises twenty-one monolingual models and five multilingual models spanning twenty-seven language varieties across Ghana, Sierra Leone, Nigeria, Senegal, Kenya and Zimbabwe. Models are fine-tuned primarily on read speech drawn from religious texts, which offer broad, license-clear and orthographically consistent coverage for languages that otherwise lack transcribed audio. We describe a two-step (and, for one family, three-step) learning-rate-annealed fine-tuning procedure that first adapts a shared multilingual model at a high learning rate and then anneals it to recover, and in several cases surpass, strong monolingual baselines. We further describe a lightweight language-conditioning mechanism that injects a one-hot language identity as a sequence of prefix frames prepended to the acoustic features, allowing a single multilingual checkpoint to be steered to a target language at inference. Across the five multilingual families the annealed models reach average word error rates (WER) of 10-13%, closing most of the gap to monolingual models while covering many languages in a single checkpoint. All models are released on the Hugging Face KhayaAI organisation under the Apache-2.0 license (attribution only) so that others may fine-tune them freely, including for commercial use. We provide a conservative estimate that the languages covered are spoken by on the order of one hundred million first-language speakers, and by substantially more when second-language use is included.
[NLP-3] Windowed-MTP: Removing the Full-Context Draft-KV Tax at Million-Token Context
【速读】： 该论文旨在解决在长上下文（百万级令牌）场景下，生成式 AI (Generative AI) 中基于推测解码（speculative decoding）的多令牌预测（Multi-Token-Prediction, MTP）草稿头因全注意力机制导致的键值缓存（KV cache）读取开销随上下文长度线性增长的问题。这一问题使得草稿阶段的计算成本急剧上升，甚至超过无推测解码的基线，严重削弱了推测解码在长序列生成中的加速效果。其核心解决方案是仅对草稿阶段的注意力机制引入流式LLM（StreamingLLM）风格的滑动窗口与注意力沉降（attention sink）机制（即Windowed-MTP），而保持目标模型的完整注意力验证结构不变。该方法无需训练、可直接嵌入现有系统且在理论上无损：仅改变草稿提出的候选令牌，不改变最终被接受的输出分布。通过将草稿的KV工作集限制在常数规模，该方法在100万上下文长度下可剔除约99%的KV条目，显著降低每步解码成本，在三个不同架构（Qwen GDN-MoE 35B/122B 和 Mamba2-hybrid NoPE 120B）上实现单卡推理中相对于原生MTP方案28%至44%的性能提升，且该增益随上下文长度进一步扩大。由于每令牌延迟与单位解码步成本成正比，相同接受率下的端到端解码延迟同步改善相同幅度；当窗口化还带来更高接受率时，性能增益更显著。此外，未被访问的草稿KV（占总KV的7.7%-11%）可通过紧凑环形缓冲区回收，不牺牲接受率或生成质量。
链接: https://arxiv.org/abs/2607.21535
作者: Alagappan Valliappan
机构: NVIDIA(英伟达)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL); Performance (cs.PF)
备注: 25 pages, 2 figures, 11 tables
Abstract:Speculative decoding accelerates autoregressive generation by having a cheap draft propose tokens that a target verifies in parallel. Frontier models increasingly ship a built-in Multi-Token-Prediction (MTP/NEXTN) draft head under the assumption that the draft is negligibly cheap. At million-token context this breaks: an MTP draft head typically runs full attention over the entire KV cache at every draft step, so its read grows linearly with context and comes to dominate the draft cost – precisely where speculation is most valuable. The effect compounds with draft length (a deep native draft can turn net-negative, slower than no speculation) and sharpens under hybrid/linear-attention targets, where cheaper verification leaves the draft’s full-attention read exposed. We apply a StreamingLLM-style sliding window plus attention sink to the draft’s attention only (Windowed-MTP), leaving full-attention verification intact. It is training-free, drop-in, and lossless by construction: the full-attention target still decides every accepted token, so windowing changes only which tokens are proposed, never which are accepted. It bounds the draft’s KV working set to a constant, dropping ~99% of KV entries at 1M. Across three architecture families (Qwen GDN-MoE 35B/122B and a Mamba2-hybrid NoPE 120B) at 1M context on a single GPU in SGLang, windowing cuts the per-decode-step cost over the shipping native MTP draft by +28% to +44%, an input-invariant margin that widens with context. Since per-token latency is this cost divided by acceptance length, at matched acceptance end-to-end decode latency improves by the same amount, and more where windowing also lifts acceptance, while preserving the target’s verified output distribution. Finally, the unread draft KV – 7.7-11% of total KV at 1M – is reclaimed via a compact ring buffer at no acceptance or quality cost.
[NLP-4] GS-Agent : Creating 4D Physical Worlds With Generative Simulation
【速读】： 该论文旨在解决从自然语言描述自动生成动态且物理上合理的4D世界这一难题，其核心挑战在于现有生成式方法难以兼顾物理真实性与可控性。传统计算机图形学依赖人工精细调整材质、运动及视觉保真度，效率低下；而现有基于大规模数据训练的生成基础模型虽具备一定生成能力，但在物理一致性与用户控制方面仍存在显著不足。为此，本文提出一种创新解决方案——GS-Agent，一个端到端的多智能体系统，通过模拟人类构建4D世界的过程，实现全流程自动化。其关键在于将任务分解为实体管理（涵盖3D资产选择、材质调优、位置布局与运动控制）和渲染配置（包括相机与光照调控）两大模块，并引入多个具有不同专业能力的智能体，协同利用物理引擎进行实时仿真，通过代码接口与物理环境交互，获取多模态反馈，持续迭代优化生成结果。该框架在保证生成内容物理合理性的基础上，实现了复杂物体间丰富交互（如液体、可变形体与刚体之间的动态行为）以及电影级摄像与光照控制，显著提升了生成世界的逼真度与可控性，为4D世界生成提供了新的范式。
链接: https://arxiv.org/abs/2607.21522
作者: Hongxin Zhang,Chunru Lin,Junyan Li,Zhou Xian,Tsun-Hsuan Wang,Chuang Gan
机构: University of Massachusetts Amherst (马萨诸塞大学阿默斯特分校); Genesis AI (生成人工智能)
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Creating dynamic and physically realistic 4D worlds from natural language descriptions is both fascinating and challenging. Traditional computer graphics methods rely on manual creation, requiring extensive human effort to fine-tune materials, motions, and visual fidelity. Recent advances in generative foundation models have sparked interest in learning to generate such 4D worlds from large-scale data; however, existing methods still struggle to ensure physical plausibility and controllability. In this work, we take a different path by leveraging foundation models to construct an agentic system that emulates how humans traditionally create 4D worlds, yet automates the entire process. We present GS-Agent, an end-to-end multi-agent framework that integrates physics engines in the loop to generate realistic, dynamic, and controllable 4D physical worlds from natural language. Inspired by how humans build 4D worlds, GS-Agent decomposes the task into entity management, covering 3D asset curation, material tuning, placement, and motion control, and rendering configuration, including camera and lighting manipulation. Multiple agents with distinct expertise interact with the physics engine via code, seek multimodal feedback, and collaborate to iteratively construct 4D worlds that align with the given descriptions. Experimental results show that GS-Agent effectively converts natural language into diverse and physically plausible 4D worlds exhibiting rich interactions among liquids, deformable objects, and rigid bodies, while achieving cinematic camera and lighting control. We envision GS-Agent as a foundation for a new paradigm in 4D world generation, empowering creative content creation and physical AI. Project page at this https URL
[NLP-5] Artificial Epanorthosis: Why large language models overuse a classical rhetorical figure and how to mitigate it
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）中过度使用修辞手法“自纠式表达”（epanorthosis，即自我修正句式，如“这不是一门课程，而是一场蜕变之旅”）的问题。其核心问题是：尽管这一修辞现象在两千年前的古典修辞学中已有记载，但现代生成式AI模型却系统性地过度使用该形式，导致输出文本在风格上偏离人类自然表达，尤其在演说体和非正式问答场景中出现显著偏差。解决方案的关键在于识别并校准这种偏差——研究指出，过度使用主要源于训练数据中富含宣传性语体以及基于强化学习的人类反馈（RLHF）对自信、强调性表述的偏好，而非生成过程的从左到右特性本身。为此，作者提出构建“自纠式表达指数”（Epanorthosis Index），通过与特定文体下人类基准的密度对比来量化模型表现。实证分析显示，模型在演说类文本中自纠式表达频率约为人类的2至3倍，而在非正式问答中则明显不足，但在论说文、新闻报道和百科类文本中与人类匹配良好。研究进一步提出三种建设性对策：一是采用轻量级LoRA适配器进行微调；二是以意大利语为例，证明仅需一句指令即可使自纠式表达减少50%至75%，而监督微调适配器几乎完全消除该现象，并可通过缩放系数精确调节至人类水平；最终主张目标应是针对不同文体实现对人类表达率的精准校准，而非彻底消除该修辞形式。论文强调，真正的风险并非模型语法错误，而是人类写作逐渐趋同于机器风格，从而丧失语言多样性与自然表达能力。
链接: https://arxiv.org/abs/2607.21498
作者: Federico Boggia
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 17 pages
Abstract:A rhetorical figure that Cicero and Quintilian catalogued two thousand years ago reappears, systematically, in the text of large language models: epanorthosis, the self-correction of the specimen «This is not a course. It is a journey of transformation». This essay argues that the overuse is a trained disposition, driven mainly by a training distribution rich in promotional prose and by preference tuning (RLHF) that rewards confident, emphatic phrasing; the left-to-right nature of generation is an amplifier rather than the root cause. Building on evidence that models diverge from human rhetorical style, and on Fontanier’s classification of epanorthosis as a figure of thought, it sets out a programme that scores the figure against genre-specific human baselines through an Epanorthosis Index (density relative to the human rate). A first measurement, on three sizes of one instruction-tuned model family, finds mis-calibration by register in both directions: the models overshoot in oratory (about twofold, near threefold in Italian, concentrated in the larger tiers) and undershoot in informal question-and-answer writing, while matching humans in argument, journalism, and encyclopedic prose. Three constructive contributions follow: a survey of mitigation techniques centred on lightweight LoRA adapters; a demonstration, in Italian, that a one-line instruction cuts the figure by half to nearly three-quarters and that a supervised-fine-tuning adapter removes it almost entirely, with a scaling coefficient that dials the reduction back onto the human rate; and the argument that the target is calibration to the human rate for each genre, not elimination. It closes on the stakes: the real risk is that we begin to write like the machines.
[NLP-6] What Where and How: Disentangling the Roles of Task Language and Model in Code Model Representations
【速读】： 该论文旨在探究独立训练的语言模型在代码理解任务中是否以相同的方式表征相同的语义概念，核心问题是：生成式代码模型在面对不同编程语言（Python与Rust）和不同架构（Qwen2.5-Coder-7B与DeepSeek-Coder-V1-6.7B）时，其内部表示是否存在共性或差异。解决方案的关键在于构建一个最小化的2×2实验设计（语言×模型），并采用一种改进的概念电路提取方法（concept-circuit extraction），对58个Python与57个Rust的语法概念进行跨模型、跨语言的一致性测量。研究发现：表征内容的“是什么”（What）具有高度一致性——模型在哪些概念上建立专用电路方面表现出显著相关性（Spearman ρ ≈ 0.65，p < 10⁻⁷），说明任务驱动了共性；但“在哪里”（Where）和“如何演化”（How）则由模型自身结构决定——不同模型在激活层位置（Qwen在第17–19层，DeepSeek在第6–7层）和电路增长模式（如Qwen早期突现原子概念）上存在系统性差异。此外，模型间共享神经元的比例也呈现非预期模式，且Qwen将Rust的类型与特质机制绑定至单一高凝聚度神经元簇，揭示出表面语法之外的深层语义维度。这些结果表明，“电路是否通用”并非单一答案，而是取决于具体属性：内容可共享，组织不可共通，即表征的普遍性存在于语义内容层面，而非计算结构层面。所有发现均基于本实验设定，未来工作将验证该模型特征是否具备预测第三模型的能力。
链接: https://arxiv.org/abs/2607.21491
作者: Piotr Wilam
机构: University College London (伦敦大学学院)
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 16 pages, 11 figures, 6 tables. Code: this https URL ; dataset: this https URL
Abstract:Do independently trained language models come to represent the same thing in the same way? We answer for code, extending a recently introduced concept-circuit extraction method to a 2x2 design – Python and Rust crossed with Qwen2.5-Coder-7B and DeepSeek-Coder-V1-6.7B – and measuring a complete inventory of grammatical concepts (58 Python, 57 Rust) identically in all four cells: the smallest design that separates what depends on the task, the language, and the model. The answer splits into three parts. What earns dedicated circuitry is set by the task: the models agree on which concepts receive circuits (Spearman \rho = 0.638 for Python, 0.673 for Rust, both p 10^-7 ). Where those circuits sit is set by the model: Qwen processes concepts in a late band (~L17-19), DeepSeek at L6-7, for both languages. How circuits grow across layers is also set by the model: Qwen gives its atomic concepts an early spike that DeepSeek does not. “Are circuits universal?” thus has no single answer: yes for What, no for Where and How – universality is a property of representational content, not of computational organisation. None of this structure was fixed in advance. The agreement could have landed anywhere between independence and identity; it lands at \rho \approx 0.65 . Rust constructs receive 2-3x more concept-specific circuitry than their Python equivalents, in both models. Both models share neurons between the languages (6/7 and 7/7 paired constructs), DeepSeek 1.94x more than Qwen – a direction no prior result predicts. And Qwen binds nine keywords of Rust’s type-and-trait machinery into one tight neuron cluster (Jaccard 0.535 vs null 0.112, p 0.001), a semantic dimension invisible in surface syntax. Ablation and linear probes confirm the circuits are functional. All claims are scoped to this 2x2; whether the per-model profile predicts a third model is the designed next test. Comments: 16 pages, 11 figures, 6 tables. Code: this https URL ; dataset: this https URL Subjects: Computation and Language (cs.CL); Machine Learning (cs.LG) Cite as: arXiv:2607.21491 [cs.CL] (or arXiv:2607.21491v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2607.21491 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Piotr Wilam [view email] [v1] Thu, 23 Jul 2026 16:34:14 UTC (150 KB)
[NLP-7] Agent ic coding without the cloud: evaluating open-weight large language models on longitudinal data preparation tasks
【速读】： 该论文旨在解决在受治理约束的研究环境中（如纵向人群研究），由于数据隐私与安全政策禁止将个人数据传输至外部云服务，导致生成式 AI（Generative AI）在代码开发中的应用受限的问题。其核心挑战在于如何在不违反数据本地化要求的前提下，实现高效、可靠的自动化数据准备。解决方案的关键在于提出一个开源评估框架，该框架基于英国队列研究的六波数据，包含经人工标注的基准数据集、明确的任务定义（如类别标准化和多波次数据合并），以及自动化的评估流程，用于衡量由开源权重大语言模型（open-weight LLMs）驱动的 AI 代理在生成 R 代码及输出数据方面的有效性。通过在消费级硬件上对不同规模的开放权重模型进行基准测试，发现当前最先进的 31-35B 参数模型已接近性能上限（平均任务完成率达 87.9%），表明在本地部署的开源模型具备在严格数据治理环境下实现 AI 辅助数据准备的可行性。
链接: https://arxiv.org/abs/2607.21482
作者: Mack Nixon,Liam Wright,Yevgeniya Kovalchuk,Alison Fang-Wei Wu,Martin Danka,Andy Boyd,David Bann
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Large language models (LLMs) and agents are now widely used tools in code development, with data typically sent to third-party cloud-based models. Their adoption in research using personal data is constrained by governance requirements that typically prohibit data transmission to external services. Locally deployable open-weight models offer an alternative since sensitive data never leave the local environment. We introduce an open-source framework for evaluating the efficacy of AI agents powered by open-weight LLMs on one of the most persistent bottlenecks in research on longitudinal population studies: data preparation. The framework comprises: a curated ground-truth dataset (cleaning scripts preparing six sweeps of data from a British cohort study), task definitions encompassing tasks such as category harmonization and multi-wave merging, and automated routines for evaluating the LLM-produced R code and outputted data. We benchmark LLMs across the (consumer grade) deployment spectrum to assess their efficacy in 20 data preparation tasks (creation of 102 variables). Current state-of-the-art, 31-35B parameter models almost saturated our benchmark (“average task completion” up to 87.9%). The performance of open-weight LLMs running on consumer-grade hardware shows promise of a viable path toward AI-assisted data preparation in governance-restricted research settings. Our framework is publicly available at: this https URL.
[NLP-8] Error Certificates for KV-Cache Eviction via Randomized Design
【速读】： 该论文旨在解决大模型服务系统中确定性键值缓存（KV-cache）淘汰策略所导致的不可识别性问题：现有方法基于重要性评分保留前k个令牌，但无法判断被淘汰内容是否对注意力输出产生实质性影响，即系统无法感知其“破坏”的后果。研究证明，即使淘汰的值被任意修改，只要保留的缓存内容不变，系统输出在表面上无变化，而真实注意力输出误差却可无限增大，因此任何服务阶段的误差估计器都无法保持一致性。为解决此问题，论文提出采用随机化淘汰机制，通过泊松采样方式以已知包含概率保留缓存尾部数据，并在softmax内部引入一个对数偏移项实现Hájek校正，从而构建出一种基于保留样本的调查抽样方差估计器，该估计器可作为每步的误差证书，在不损失精度的前提下实现了0.97的实证覆盖率。在真实负载上的预注册实验表明，尽管部分假设被推翻（如25%-50%预算下问句感知淘汰近乎零成本、输出对数概率优于证书、证书门控预算扩展无增益），但核心结论得以保留：该随机化方案赋予了误差来源的可归因性（attribution），能够有效区分缓存引入的错误与固有错误（AUC 0.73–0.75，显著优于输出置信度的0.47–0.54），并比随机或置信度门控更优地调度重计算。因此，关键在于通过随机化实现误差的可识别与可归因，而非提升预测性能。
链接: https://arxiv.org/abs/2607.21475
作者: Peng Xie
机构: Technical University of Munich (慕尼黑工业大学)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Deterministic KV-cache eviction keeps the top- k tokens under an importance score and deletes the rest. We prove that this design cannot know what it destroyed: evicted values can be altered so that everything the serving system retains is unchanged while the true attention-output error grows arbitrarily, so no serving-time estimator of that error is consistent. Randomized eviction restores identifiability. With a Poisson-sampled tail at known inclusion probabilities, one logit offset performs the Hájek correction inside the softmax, and a survey-sampling variance estimator over the retained set becomes a per-step error certificate with 0.97 empirical coverage at no accuracy cost. On real workloads we pre-registered seven claims and lost three: question-aware eviction at 25–50% budgets is nearly free; output log-probability predicts failure better than the certificate; certificate-gated budget escalation adds nothing. What survives is attribution: the certificate separates cache-induced from inherent failures (AUC 0.73–0.75, against 0.47–0.54 for output confidence) and schedules recomputation better than random or confidence gating. Randomization buys attribution, not prediction.
[NLP-9] RUMBA: Russian User Memory Benchmark
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在长期对话记忆能力评估中存在的局限性问题，即现有基准测试过度依赖英语语料且仅采用聚合检索指标，难以捕捉长程上下文、时间信息与推理之间的复杂交互。为此，研究提出RUMBA（Russian User Memory BenchmArk），一个面向长期对话记忆的新基准，其关键在于构建了细粒度的记忆导向问题类型分类体系，并引入统一评估方法，综合考虑语义类型、会话范围、时间推理以及时间表达的显式程度。RUMBA包含带时间戳的用户-助手对话序列及其对应的问答对，要求模型在跨会话场景下完成信息检索、融合与推理任务。尽管该基准以俄语设计，但亦提供遵循相同方法论的英文对齐子集。实验表明，RUMBA可作为诊断工具，通过分析不同基准切片中模型的表现，有效揭示各类记忆机制的优势与失效模式。
链接: https://arxiv.org/abs/2607.21447
作者: Elizaveta Shevtsova,Inna Glebkina,Mark Baushenko,Pavel Gulyaev,Alena Fenogenova
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:The ability to handle long-term memory in LLMs is becoming increasingly critical, yet existing benchmarks remain English-centric and rely on aggregate retrieval metrics, failing to capture interactions between long-range context, temporal information, and reasoning. To address this, we introduce RUMBA (Russian User Memory BenchmArk) - a new benchmark for long-term conversational memory that provides a fine-grained taxonomy of memory-centric question types and a unified methodology accounting for semantic type, session scope, temporal reasoning, and the explicitness of temporal expressions. RUMBA consists of timestamped user-assistant dialogues with QA pairs requiring retrieval, combination, and reasoning across sessions. While designed for Russian, we also provide an aligned English subset under the same methodology. We evaluate contemporary memory systems and long-context models, and show how RUMBA serves as a diagnostic tool to analyze model behavior across benchmark slices and identify strengths and failure modes of different memory mechanisms.
[NLP-10] When Trivia Is Not Trivial: Everyday Knowledge Failures in Multilingual LLM s
【速读】： 该论文旨在解决当前大型语言模型（LLMs）在评估其真实世界知识能力时存在的局限性问题，特别是针对日常文化知识与长尾领域知识的评测不足。现有基准多集中于学术性、标准化的知识点，难以反映模型在流行文化、社会常识等实际应用场景中的表现。为此，研究提出Triv iaRoomQA这一多语言基准，涵盖288个主题，包含3,300组六种欧洲语言的平行多项选择题，并额外补充5,340道法语专属题目以支持细粒度分析。其核心解决方案在于构建一个覆盖广泛、贴近现实语境的评测体系，能够系统评估模型在历史、地理、数学等知识密集型任务以及名人、音乐、电影、新闻等日常流行文化主题上的表现。实验结果表明，尽管多数模型在传统学术类知识上表现良好，但在日常文化知识方面存在显著短板，且跨语言性能差异明显，揭示了模型对事实性知识的掌握并非完全语言独立。这凸显出当前主流评测体系的不足，强调需引入更贴近真实使用场景的多语言、多主题基准以全面评估生成式AI（Generative AI）的实际认知能力。
链接: https://arxiv.org/abs/2607.21445
作者: Anna Mosolova,Djamé Seddah
机构: INRIA Paris
类目: Computation and Language (cs.CL)
备注: submitted to the ARR
Abstract:Quiz rooms, trivia nights, and quiz shows challenge human knowledge across a wide range of topics, from canonical facts to everyday culture. In this paper, we examine whether large language models (LLMs) can perform competitively in such settings, using quiz-style questions to test them on both common and niche topics. We introduce TriviaRoomQA, a multilingual benchmark designed to evaluate everyday, culturally grounded, and long-tail knowledge across 288 topics. The benchmark contains 3,300 parallel multiple-choice questions in six European languages and additional 5,340 French-only questions for a more fine-grained case study. We evaluate 30 open-weight LLMs from European, Asian, and North American providers, covering models from 7 to 70B parameters. We find that models are strong on knowledge-intensive topics such as history, geography, and mathematics, but substantially weaker on everyday popular-culture topics such as celebrities, music, movies, and news. Moreover, model performance varies across languages even for the same underlying questions, suggesting that access to factual knowledge is not always language-independent. In sum, our dataset and experiments demonstrate an important knowledge gap which is not captured by existing academic-based saturated benchmarks.
[NLP-11] oken Budget Saturation and Mechanistic Early Detection of Reasoning Non-Convergence in Chain-of-Thought Models
【速读】： 该论文旨在解决生成式推理模型在链式思维（Chain-of-Thought, CoT）过程中存在的“二元收敛”现象：即生成结果要么在有限的词元（token）预算内成功收敛并得出正确结论，要么耗尽预算仍未达成结论。研究发现，收敛生成物在AIME 1983–2024数据集上达到90.3%的准确率，而非收敛生成物仅达6.6%，整体收敛率为62.0%。为应对这一问题，论文提出关键解决方案——利用模型内部表示（hidden-state activations）在生成过程早期预测最终是否收敛。通过在第50至300个词元位置训练线性探测器（linear probes），研究发现第20层在第150个词元处的激活值可实现AUC 0.608（±0.080，五折交叉验证），显著高于随机水平，且在第50个词元时仍具判别能力。与基于词元熵和重复统计的行为基线相比，激活探测器表现更优。全局置换检验（sweep-level permutation test）得p=0.063（100,000次置换），虽未达传统显著性阈值，但提示存在微弱但可识别的信号。该结果表明，模型的收敛命运在生成过程早期即部分编码于中间表示中，为实现早期退出推理（early-exit inference）与自适应计算资源分配提供了可行路径。
链接: https://arxiv.org/abs/2607.21433
作者: Renuka Oladri,Niveda Jawahar,Abdirisak Mohamed
机构: University of Maryland (马里兰大学); SAP Labs, LLC (SAP 实验室)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Chain-of-thought reasoning models such as DeepSeek-R1-Distill-Qwen-7B exhibit a bimodal convergence pattern: generations either terminate within a token budget (converged) or exhaust it without reaching a conclusion (non-converged). We characterize this phenomenon empirically, showing that converged generations achieve 90.3% accuracy on AIME 1983-2024 while non-converged ones achieve only 6.6%, with an overall convergence rate of 62.0%. We then ask whether this outcome is detectable early in the thinking chain using internal model representations. Training linear probes on hidden-state activations at token positions 50-300, we find that layer-20 activations at token 150 achieve AUC 0.608 (±0.080, 5-fold CV), reliably above chance even at token 50. Activation probes consistently outperform behavioral baselines derived from token entropy and repetition statistics. A sweep-level permutation test yields p=0.063 (100,000 permutations), consistent with a modest signal that our sample size cannot confirm at conventional thresholds. These findings suggest that convergence fate is partially encoded in intermediate representations well before the generation ends, opening a path toward early-exit inference and adaptive compute allocation.
[NLP-12] An Evaluation Framework for Structured Audio Captions Validated by Controlled Perturbations
【速读】： 该论文旨在解决结构化音频描述（structured audio descriptions）在评估过程中面临的挑战，即现有评价指标主要针对扁平化的文本输出，无法可靠衡量多模态属性。其解决方案的关键在于提出一个面向结构化音频描述的多轴评估框架，基于AudioCards数据集，在五个正交维度上进行评估：标签集合（tag-sets）、描述文本、逻辑推理、数值度量和频谱特征。该框架结合大语言模型（Large Language Model, LLM）判断以捕捉语义细微差异，并融合确定性计算指标以精确测量声学偏差。为验证评估框架的可靠性，研究引入一种受控扰动测试协议，向真实标注中注入类型化、分级的错误，结果表明该框架能够有效区分语义保持的改写与真实的语义及声学失真。
链接: https://arxiv.org/abs/2607.21424
作者: Liang-Yuan Wu,Sripathi Sridhar,Mark Cartwright,Magdalena Fuentes
机构: 未知
类目: Computation and Language (cs.CL); Sound (cs.SD)
备注: submitted to DCASE 2026
Abstract:Recent advancements in automated audio captioning (AAC) have shifted from monolithic sentence generation toward structured formats that explicitly disentangle distinct acoustic and semantic properties. However, evaluating this heterogeneous data remains a significant challenge. Existing caption metrics focus on flat textual outputs and fail to reliably assess multimodal attributes. To bridge this gap, we propose a multi-axis evaluation framework tailored for structured audio descriptions. Building on the AudioCards dataset, we evaluate outputs across five orthogonal axes: tag-sets, descriptions, logical reasoning, numeric measurements, and spectral profiles. Our approach combines Large Language Model (LLM) judges to capture semantic nuance with deterministic computational metrics to precisely measure acoustic deviations. To rigorously validate the reliability of this framework, we introduce a controlled perturbation testing protocol that injects typed, graded errors into groundtruth annotations. Our results demonstrate that this framework successfully distinguishes meaning-preserving paraphrases from genuine semantic and acoustic corruptions.
[NLP-13] Euclid-MCP: A Model Context Protocol Server for Deterministic Logical Reasoning via Prolog
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在多步逻辑推理任务中可靠性不足的问题，尤其是在安全关键或合规敏感领域。尽管LLMs在自然语言理解与生成方面表现优异，但在复杂推理场景下易产生幻觉，难以保证推理结果的准确性与可追溯性。现有神经符号方法虽通过结合神经模型与外部符号引擎缓解此问题，但多数集成方式为定制化设计，缺乏统一的工具接口以支持增强型智能体（tool-augmented agents）的协作。为此，本文提出Euclid-MCP——一个开源的MCP（Model-Controller-Planner）服务器，基于SWI-Prolog实现确定性逻辑推理。其核心创新在于引入Euclid-IR，一种与引擎无关的中间表示形式，用于表达霍恩子句（Horn-clause logic），具备人类可读性、易于由LLM生成，并能高效编译至Prolog或其他后端。该系统提供紧凑的工具接口，支持“翻译-执行-检查-修复”闭环流程，使LLM客户端可在委托推理的同时，完全访问证明轨迹与推导日志。在真实的IT安全与合规场景评估中，实验表明：当知识库规模较小时，仅依赖LLMs尚可满足需求；但在大规模问题上，LLMs会系统性地产生幻觉，而Euclid-MCP则能提供精确答案，且延迟更低、输出更紧凑。研究进一步指出，语义检索增强生成（Semantic RAG）本质上不适用于规则强制执行，而Euclid-MCP可作为稳定、共享的推理底座，服务于基于RAG的助手及自主代理系统。
链接: https://arxiv.org/abs/2607.21412
作者: Bartolomeo Bogliolo
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Software Engineering (cs.SE)
备注:
Abstract:Large Language Models (LLMs) excel at natural language understanding and generation but remain unreliable for multi-step logical reasoning, especially in safety-critical or compliance-sensitive domains. Recent neuro-symbolic approaches address this gap by coupling neural models with external symbolic engines, yet most integrations are bespoke and lack a standardized interface for tool-augmented agents. This paper presents Euclid-MCP, an open-source MCP server that provides deterministic logical reasoning via SWI-Prolog. Euclid-MCP introduces Euclid-IR, an engine-agnostic intermediate representation for Horn-clause logic that is human-readable, easy for LLMs to generate, and straightforward to compile into Prolog or alternative backends. The server exposes a compact tool interface that supports a translate-run-inspect-repair loop, enabling LLM clients to delegate inference while retaining full access to proof traces and derivation logs. We evaluate Euclid-MCP on a realistic IT security and compliance use case. Results show that while LLMs alone are sufficient on small knowledge bases, they hallucinate systematically on larger problems, whereas Euclid-MCP delivers exact answers with lower latency and more compact outputs. We argue that semantic RAG is fundamentally unsuited for rule enforcement, and that Euclid-MCP can serve as a stable, shared reasoning substrate for both RAG-based assistants and agentic systems.
[NLP-14] Anti-Periodic Positional Encoding: Möbius Boundary Conditions Make In-Context Retrieval Reliable
【速读】： 该论文旨在解决大模型在长序列建模中位置编码（Positional Encoding）对远距离依赖关系捕捉能力不足的问题，尤其关注在“针堆找针”（needle-in-a-haystack）任务中对稀疏关键信息的可靠检索能力。现有旋转位置编码（Rotary Positional Encoding, RoPE）采用周期性边界条件，导致序列两端在几何上无确定性关联，限制了模型对全局上下文结构的感知。为此，作者提出莫比乌斯旋转位置编码（Möbius RoPE），其核心创新在于构建一种反周期性频率阶梯（anti-periodic frequency ladder），即频率梯度以奇数倍π递进，使序列在循环闭合时产生-1的拓扑全息性（positional holonomy），从而在数学上实现序列首尾间的确定性耦合——通过一个闭式狄利克雷“偶极子”（Dirichlet “dipole”）建立长程依赖的刚性结构。实验验证表明，该设计在数值精度达~10⁻⁶的前提下，显著提升了短上下文下的检索可靠性（512上下文时准确率从63.3%提升至90.3%），且效果经多种子实验和方差检验（p=0.013–0.029）确认具有统计显著性。控制实验进一步证明：仅反周期性结构可产生此效应，周期性（holonomy +1）或非周期性频率配置均无法复现；此外，将训练好的模型频率表回滚至标准RoPE会完全破坏检索性能，尤其影响远端“针”（far needles）的识别，说明模型已深度依赖该长程几何结构。尽管纯非周期性位置编码（NoPE）在短上下文表现更优但代价为13%困惑度上升且外推性能最差，而反周期性混合架构（hybrid arm） 在保持基准困惑度（29.66 vs. 29.72）的同时，实现了高可靠性下限，成为唯一兼具低开销与强鲁棒性的解决方案。该方法仅需一次频率表替换即可提供“零成本保险”，有效应对检索任务中的随机种子依赖问题，且效果局限于训练窗口内的单针检索，具备高度可解释性和实用性。
链接: https://arxiv.org/abs/2607.21405
作者: Ji Ho Bae
机构: Seoul National University (首尔国立大学)
类目: Computation and Language (cs.CL)
备注: 30 pages, 12 figures
Abstract:Möbius RoPE is a rotary positional encoding built on the anti-periodic frequency ladder \theta_i=\pi(2i+1)/N : every rotation plane advances by an odd multiple of \pi across the training context, so the positional holonomy is -1 and the two ends of the sequence are deterministically coupled through a closed-form Dirichlet “dipole”; to our knowledge this is the first anti-periodic boundary condition in positional encoding. We verify the theory numerically to \sim 10^-6 and pretrain 48 models spanning six 160M-class and three 410M-class arms (2B FineWeb-Edu tokens each; the hybrid arm puts Möbius frequencies on 25% of heads). Hybrid perplexity is unchanged (29.66 vs. 29.72), but needle-in-a-haystack retrieval becomes reliable: 90.3\pm5.7% versus 63.3\pm31.4% at context 512 ( n=6 seeds), observed worst seed 86% versus 14%, robust variance tests p=0.013 - 0.029 (unadjusted), recurring at 410M (Levene p=0.040 ). Matched controls isolate the mechanism: an aperiodic ladder in the same frequency band reproduces none of the effect, and a periodic (holonomy +1 ) ladder only a fraction. Swapping trained models’ frequency table back to standard RoPE (weights frozen) collapses retrieval, with damage concentrated on far needles: trained models depend on this long-range geometry. A NoPE arm is even more reliable at short context but pays a 13% perplexity tax and extrapolates worst; only the anti-periodic hybrid pairs baseline perplexity with a high reliability floor. The effect is scoped to single-needle retrieval within the training window; a one-line frequency swap thus provides zero-cost insurance against the retrieval seed lottery.
[NLP-15] MemTools: A Unified Research Framework for Interoperable Agent Memory
【速读】： 该论文旨在解决智能体（agent）架构中记忆系统研究面临的系统性障碍，主要问题包括：现有记忆系统架构碎片化严重，导致跨系统研究难以开展；记忆生命周期的不同阶段被紧密耦合，评估逻辑与特定数据集交织，缺乏对异构记忆类型的有效管理支持。其解决方案的关键在于提出MemTools这一互操作性研究框架，通过声明式数据契约标准化记忆生命周期，实现记忆组件与其底层部署环境的解耦，支持跨系统的可互换组装；同时，将基准数据集与执行协议正交分离，实现可控评估；并提供统一的计算接口，协调符号、神经及多模态记忆表示在共享运行时中的协同工作。实证结果表明，该框架能够有效实现记忆设计变量的系统性隔离与分析，为推进基于原则的智能体记忆研究提供了可扩展且实用的基础架构。
链接: https://arxiv.org/abs/2607.21404
作者: Chengfeng Zhao,Jinhui Chen,Sirui Liang,Shizhu He,Yequan Wang,Jun Zhao,Kang Liu
机构: Institute of Automation, CAS (中国科学院自动化研究所); University of Chinese Academy of Sciences (中国科学院大学); Beijing Academy of Artificial Intelligence (北京人工智能研究院); Zhongguancun Institute of Artificial Intelligence (中关村人工智能研究院)
类目: Computation and Language (cs.CL)
备注: Work in progress
Abstract:While memory systems are essential for agent architectures, pervasive architectural fragmentation restricts systematic research. Existing implementations typically couple different stages of the memory lifecycle, entangle evaluation logic with specific datasets, and provide limited support for the management of heterogeneous memory types. We introduce MemTools, an interoperability research framework that decouples memory system components from their underlying deployment environments. MemTools standardizes the memory lifecycle through declarative data contracts, enabling the interchangeable assembly of components across different systems. It orthogonally separates benchmark datasets from execution protocols to facilitate controlled assessments. Furthermore, MemTools provides a unified computational interface for coordinating symbolic, neural, and multimodal memory representations within a shared runtime. Empirical evaluations on cross-system component integration, evaluation protocol reconfiguration, and heterogeneous memory coordination demonstrate that MemTools enables systematic isolation and analysis of memory design variables. These findings suggest that MemTools provides a practical and extensible infrastructure for advancing principled research on agent memory.
[NLP-16] Word meaning co-determines vowel-inherent spectral change. A corpus-based investigation of conversational Mandarin
【速读】： 该论文旨在解决自然对话中汉语元音固有频谱变化（VISC）的语义依赖性问题，即元音共振峰轨迹动态是否受到词汇语义信息的影响。其关键解决方案在于结合广义加性模型（Generalized Additive Model, GAM）与分布语义中的词嵌入（Word Embeddings），发现控制发音时长、性别、说话人身份、共发音效应、元音类别及语句位置等变量后，元音的F1和F2共振峰轨迹仍存在显著的词特异性成分，且这些轨迹可由上下文相关的词嵌入（Contextualized Word Embeddings）有效预测，预测性能远超随机置换基线。这一发现挑战了传统语音生成的模块化认知模型，表明词汇语义在精细发音实现层面具有直接调制作用，揭示了语言理解与发音控制之间的深层耦合机制。
链接: https://arxiv.org/abs/2607.21391
作者: Xiaoyun Jin,Mirjam Ernestus,R. Harald Baayen
机构: Eberhard Karls Universität Tübingen (图宾根大学); Radboud University (奈梅亨大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:This study investigates vowel-inherent spectral change (VISC) in spontaneous conversational Mandarin. Using the generalized additive model and word embeddings from distributional semantics, we show that, when controlling for variables such as vowel duration, gender, speaker identity, co-articulation, vowel identity, and utterance position, vowel formant trajectory dynamics have word-specific components that are tied to their meaning in context: The F1 and F2 trajectories of words can be predicted from their contextualized embeddings with an accuracy that substantially exceeds a permutation baseline. Challenging modular cognitive models of speech production, these results indicate that, words’ semantics co-determine the fine details of their articulation.
[NLP-17] Capital Markets LLM Reliability Score (CM-LRS): From Plausible to Bankable
【速读】： 该论文旨在解决生成式 AI 在资本市场的实际应用中“可被银行接受”（bankable）的核心问题，即尽管大语言模型（Large Language Model, LLM）能够生成流畅的文本，但其输出是否具备在真实业务场景下经得起交易对手方或监管机构审查的可靠性。现有评估方法多聚焦于表面准确性或问答层面的性能（如开放域问答、金融领域问答任务），无法反映实际工作流程中最终交付物的质量。为此，论文提出资本市场大模型可靠性评分系统（Capital Markets LLM Reliability Score, CM-LRS），其关键在于将评估层级从传统的“问题-答案”对提升至“工作流输出”（workflow-output）层面，涵盖七维指标：事实准确性、证据可追溯性、数值一致性、工作流完整性、来源专业性、决策有用性以及可审阅性/可审计性。每一维度基于受监管环境中评审人员使用的信号进行0–5分打分，且整体得分可针对不同工作流灵活调整。通过在五个典型资本市场工作流（如债务承销条款提取、并购可比分析等）上对四类模型进行评估，结果显示：前沿闭源模型表现高度一致（平均分差仅0.22），而开源基线模型显著落后；模型差距主要集中在信息检索与合成环节，而非结构化提取；其中“决策有用性”维度表现出最大跨模型差异与高评委一致性，凸显了在真实业务中判断输出价值的复杂性。因此，该研究的核心贡献在于建立了一个面向实际业务可信度的评估框架，强调“合理性”易得，“可被银行接受”才是真正的门槛。
链接: https://arxiv.org/abs/2607.21340
作者: Prerit Ahuja
机构: Independent Researcher(独立研究员); Director-level practitioner in investment banking(投资银行总监级从业者)
类目: Computation and Language (cs.CL)
备注: 23 pages. Resubmission of submit/7557765, which expired due to an arXiv system bug (confirmed in support ticket AH-199019); the overfull-box correction requested by moderators has been applied. Original submission held in moderation since 14 May 2026. Therefore, request priority review/approval
Abstract:In capital-markets workflows the question is rarely whether a large language model can produce a fluent draft, but whether the draft is bankable: defensible in front of a counter-party or a regulator, with the documents in hand. Existing methods address parts of that gap: open-domain QA benchmarks reward surface accuracy, and finance benchmarks (FinanceBench, FinQA, ConvFinQA) advance document-grounded and numerical QA but evaluate at the question-answer layer rather than the workflow outputs practitioners defend. We introduce CM-LRS, a Capital Markets LLM Reliability Score, evaluating outputs at the workflow-output layer across seven dimensions: factual accuracy, evidence traceability, numerical consistency, workflow completeness, source discipline, decision usefulness, and reviewability/auditability. Each is scored 0-5 against a rubric anchored on signals reviewers in regulated settings use; the aggregate is tunable to the workflow. We demonstrate CM-LRS on five workflows (DCM transaction-terms extraction, precedent retrieval, issuer profile synthesis, MA transaction-comparable reasoning, ECM transaction-terms extraction) over public SEC EDGAR filings, a public UK takeover release, and fictional synthetic supplements, scoring four models against four independent LLM judges spanning three model families. Three findings. First, the frontier closed-source models cluster within 0.22 points on four-judge averaged CM-LRS (Sonnet 4.6 = 4.31, Opus 4.7 = 4.30, GPT-5.5 = 4.09); all four judges place the open-weights baseline (Llama 3.3 70B = 3.15) last. Second, that gap concentrates on retrieval (2.23) and synthesis (2.15), not extraction (0.84). Third, Decision Usefulness shows the widest cross-model dispersion of any dimension (4.0 points on issuer profiling) and top-tier inter-judge agreement (mean r = 0.52). Plausibility is cheap. Bankability is the bar. Comments: 23 pages. Resubmission of submit/7557765, which expired due to an arXiv system bug (confirmed in support ticket AH-199019); the overfull-box correction requested by moderators has been applied. Original submission held in moderation since 14 May 2026. Therefore, request priority review/approval Subjects: Computation and Language (cs.CL) ACMclasses: I.2.7; J.1 Cite as: arXiv:2607.21340 [cs.CL] (or arXiv:2607.21340v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2607.21340 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Prerit Ahuja [view email] [v1] Thu, 23 Jul 2026 14:10:58 UTC (30 KB) Full-text links: Access Paper: View a PDF of the paper titled Capital Markets LLM Reliability Score (CM-LRS): From Plausible to Bankable, by Prerit AhujaView PDFHTML (experimental)TeX Source view license Current browse context: cs.CL prev | next new | recent | 2026-07 Change to browse by: cs References Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading… BibTeX formatted citation loading… Data provided by: Bookmark checked="checked"class=“labs-tab-input”> Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv (What is alphaXiv?) Links to Code Toggle CatalyzeX Code Finder for Papers (What is CatalyzeX?) DagsHub Toggle DagsHub (What is DagsHub?) GotitPub Toggle Gotit.pub (What is GotitPub?) Huggingface Toggle Hugging Face (What is Huggingface?) ScienceCast Toggle ScienceCast (What is ScienceCast?) Demos Demos Replicate Toggle Replicate (What is Replicate?) Spaces Toggle Hugging Face Spaces (What is Spaces?) Spaces Toggle TXYZ.AI (What is TXYZ.AI?) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower (What are Influence Flowers?) Core recommender toggle CORE Recommender (What is CORE?) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv’s community? Learn more about arXivLabs. Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?) mathjaxToggle(); We gratefully acknowledge support from our major funders, member institutions, , and all contributors. About Help Contact Subscribe Copyright Privacy Accessibility Operational Status (opens in new tab) Major funding support from
[NLP-18] Phonetic forced alignment for low-resource language varieties: Model training and evaluation on Chengdu Mandarin
【速读】： 该论文旨在解决低资源语言变体在语音强制对齐（phonetic forced alignment）中缺乏专用模型的问题。针对成都普通话这一低资源语言变体，研究构建了基于17小时语料库和自定义音素转写词典（G2P dictionary）的文本相关与文本无关对齐系统。其解决方案的关键在于：首先训练一个文本相关的高斯混合模型-隐马尔可夫模型（GMM-HMM）对齐器（Chengdu-MFA），随后利用Chengdu-MFA生成的伪标签对预训练音频编码器进行微调，以实现文本无关的对齐（Chengdu-FC）。实验结果表明，两种方法均显著优于标准普通话基线，其中Chengdu-MFA将平均音素边界误差降低31.8%，Chengdu-FC更是达到61.2%的提升。该研究建立了一套高效的自举式（bootstrapping）流程，无需大量人工标注即可开发出适用于资源匮乏语言变体的高精度对齐工具。
链接: https://arxiv.org/abs/2607.21332
作者: Zhiheng Qian,Aini Li,Hai Hu,Liang Zhao
机构: 上海交通大学(Shanghai Jiao Tong University); 城市大学( City University of Hong Kong); 香港理工大学( PolyU); 北京外国语大学(Beijing Foreign Studies University)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 5 pages, 1 figure
Abstract:Phonetic forced alignment is a key technique in phonetic research, yet existing alignment systems lack specialized models for low-resource language varieties. We address this by training text-dependent and text-independent aligners for Chengdu Mandarin using a 17-hour corpus and a custom G2P dictionary. We trained a text-dependent GMM-HMM model (Chengdu-MFA) and fine-tuned a pretrained audio encoder on frame classification with Chengdu-MFA’s pseudo label for text-independent alignment (Chengdu-FC). Evaluation on an expert-annotated test set show that both methods significantly outperform Standard Mandarin baselines. Chengdu-MFA reduced average phone boundary differences by 31.8%, while Chengdu-FC achieved a 61.2% reduction. This work establishes a practical bootstrapping pipeline for developing accurate aligners for under-resourced varieties without labor- and time-intensive manual annotation.
[NLP-19] GRADRAG : Cross-Component Prompt Adaptation for Coordinated Multi-Agent RAG
【速读】： 该论文旨在解决当前检索增强生成（Retrieval-Augmented Generation, RAG）系统中各组件独立优化导致整体性能受限的问题，尤其在多大语言模型（LLM）代理协同工作时缺乏端到端的协调机制。其核心解决方案是提出GRADRAG框架，通过将RAG流水线建模为计算图，并实现跨组件的提示适应（prompt adaptation）。该框架的关键在于引入一个评估器（Evaluator）对下游答案及支撑证据进行结构化评价，生成可操作的反馈信息，由提示优化器（Prompt Optimizer）用于迭代更新上游代理（如检索器、图构建器和回答器），形成闭环优化；同时，评估器还支持在输出达到满意水平时触发早停机制。实验表明，在SQUALITY与QMSUM基准测试中，无论采用基于扁平分块的检索（结合IRCoT式查询优化）还是基于图的检索（逐步构建并丰富实体-关系图），GRADRAG均显著优于仅优化最终生成器的一次性优化基线，平均在大语言模型评判的成对比较中取得12–15个百分点的净偏好优势，且多数性能提升在两次优化迭代内即达成。
链接: https://arxiv.org/abs/2607.21324
作者: Paolo Pedinotti,Enrico Santus
机构: Bloomberg
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 8 pages
Abstract:Retrieval-Augmented Generation (RAG) systems increasingly employ multiple LLM agents. Yet, most prior work optimizes components in isolation rather than coordinating improvements across the pipeline. We introduce GRADRAG, a framework for cross-component prompt adaptation that models the RAG pipeline as a computational graph and propagates structured evaluation feedback to update upstream agents. An Evaluator critiques downstream answers and supporting evidence, producing actionable feedback that a Prompt Optimizer uses to iteratively update adaptive agents, such as retrievers, graph constructors, and answerers. The Evaluator also triggers early stopping when the output is deemed satisfactory. We evaluate GRADRAG on the SQUALITY and QMSUM benchmarks under two retrieval paradigms: flat chunk-based retrieval using IRCoT-style query refinement (Trivedi et al., 2023), and graph-based retrieval that constructs and iteratively enriches an entity-relation graph from the document. Across both settings, GRADRAG consistently outperforms one-step refinement baselines that update only the final generator, achieving a 12-15 percentage point net preference margin in LLM-judged pairwise comparisons, with most gains realized within two refinement iterations.
[NLP-20] Adaptive Depth Sparse Framework: Similarity-Driven Resource Allocation for Pre-Trained LLM s
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在推理阶段因Transformer架构带来的高计算开销问题。现有加速方法通常依赖于特定任务的微调或从头训练，导致适应成本高且跨任务通用性受限。为应对这一挑战，论文提出自适应深度稀疏框架（Adaptive Depth Sparse Framework, AdaDSF），其核心在于无需对预训练模型进行完整重训练，即可将现成的LLM转化为深度稀疏模型。关键创新点在于：通过分析各层输入与输出隐状态之间的余弦相似度，量化不同层在表征变换中的贡献差异，并据此自适应地分配每层的令牌保留率；在此基础上，引入轻量级路由机制在每一层筛选信息量丰富的令牌，并设计一种特征保持对齐目标，以确保稀疏模型与密集模型在中间层和最终表示上的一致性。实验结果表明，在GPT-NeoX和Qwen2.5模型上，AdaDSF在语言建模与常识推理任务中显著降低推理浮点运算量（FLOPs），同时性能接近全稠密模型；在相同稀疏度下，其准确率下降幅度明显优于MoD、D-LLM和DLO等先进基线方法。
链接: https://arxiv.org/abs/2607.21291
作者: Yidu Wu,Xiang Wang,Kejie Zhao,Zhangchi Wang,Qinghai Guo,Xiaoying Tang
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted by ICIC 2026. 12 pages, 2 figures, 4 tables
Abstract:Large language models (LLMs) achieve strong generation and reasoning performance, but the Transformer architecture incurs high inference cost. Existing acceleration methods often rely on task-specific fine-tuning or training from scratch, increasing adaptation cost and limiting cross-task usability. We present an Adaptive Depth Sparse Framework (AdaDSF) that converts off-the-shelf pre-trained LLMs into depth-sparse models without full retraining. Our key insight is that layers contribute unequally to representation transformation, characterized by the cosine similarity between layer input and output hidden states. Based on this, AdaDSF assigns layer-wise token retention ratios from similarity statistics, uses a lightweight router to select informative tokens at each layer, and introduces a feature-preserving alignment objective to match intermediate and final representations between sparse and dense models. On GPT-NeoX and Qwen2.5 over language modeling and commonsense reasoning, AdaDSF substantially reduces inference FLOPs while preserving performance close to dense counterparts. Under comparable sparsity, AdaDSF consistently yields smaller accuracy degradation than strong baselines including MoD, D-LLM, and DLO.
[NLP-21] news-crawler-LM: A Small Long-Context Model For High-Quality News Crawling
【速读】： 该论文旨在解决从新闻网页中提取结构化内容的难题，主要挑战源于HTML布局的异构性、标记不一致以及大量无关信息（如导航栏和广告）的干扰。现有基于规则的爬虫虽能实现高精度提取，但需针对每个发布者手动配置，难以泛化；而大语言模型虽具备更强的灵活性，却因计算成本过高限制了实际应用。本文提出news-crawler-LM，一个在高质量、人工验证的Fundus新闻爬取库上微调的小型长上下文语言模型，其关键创新在于通过端到端方式将原始HTML转换为纯文本及包含标题、作者、发布日期和正文等字段的结构化JSON，显著提升了在HTML-to-Markdown与HTML-to-JSON任务上的性能，相比强基线分别提升+4.8 BLEU和+6.1 METEOR（HTML-to-Markdown），以及+2.2 BLEU和+4.1 METEOR（HTML-to-JSON）。尽管在未见过的发布者上，其在HTML-to-plaintext任务中的表现仅略优于传统规则解析库，但整体仍展现出良好的泛化能力与实用性。研究团队已开源全部模型与相关资源以促进社区发展。
链接: https://arxiv.org/abs/2607.21284
作者: Pascal Stolzenburg,Jonas Golde,Max Dallabetta,Alan Akbik
机构: Humboldt-Universität zu Berlin (柏林洪堡大学)
类目: Computation and Language (cs.CL)
备注: KONVENS 2026
Abstract:Extracting structured content from news pages remains challenging due to heterogeneous HTML layouts, inconsistent markup, and substantial boilerplate such as navigation elements and advertisements. Rule-based news crawlers can achieve high extraction accuracy by encoding site-specific structure, but require manual configuration in order to generalize to new publishers. Large language models provide a more flexible alternative by reducing the need for handcrafted rules, but their high computational cost limits practical deployment. In this paper, we introduce news-crawler-LM, a small long-context language model fine-tuned on high-quality, human-validated extractions from the Fundus news-crawling library. Our model converts raw HTML into plaintext and structured JSON, including fields such as headline, author, publication date, and article body. In our experiments, news-crawler-LM outperforms strong baselines in HTML-to-Markdown and HTML-to-JSON extraction, improving performance by +4.8 BLEU and +6.1 METEOR in the HTML-to-Markdown task, and by +2.2 BLEU and +4.1 METEOR in the HTML-to-JSON task. However, we also observe that our model only slightly better compared to other rule-based parsing libraries on the HTML-to-plaintext task in evaluations on previously unseen publishers. We release all models and artifacts to the research community.
[NLP-22] A Unified Moral-Value Dataset for Instruction Tuning IJCAI ECAI2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在价值对齐（value alignment）方面的挑战，即如何使模型行为与特定人类价值观保持一致。尽管指令微调（instruction tuning）已被证明在零样本任务中具有潜力，但现有指令微调数据集大多未聚焦于道德情境与行为，缺乏针对性。为此，研究者构建了一个统一的道德价值数据集（unified moral-value dataset），通过整合多个现有道德价值数据集并将其转换为指令-响应格式，使其可直接用于指令微调。该研究的关键在于：提出了一种可直接应用于模型对齐的结构化道德数据集，并验证了将该数据集与通用任务数据集混合训练时，既能保持模型在通用任务上的性能，又能提升其在道德导向任务中的表现，同时初步揭示了混合比例对价值对齐效果的影响。该成果为后续大模型的价值对齐研究提供了重要资源和方法支持。
链接: https://arxiv.org/abs/2607.21279
作者: Zhaohui Zeng,Florian Mai
机构: RWTH Aachen; University of Bonn; Lamarr Institute for Machine Learning and Artificial Intelligence
类目: Computation and Language (cs.CL)
备注: Accepted at the 4th International Workshop on Value Engineering in AI (VALE 2026), co-located with IJCAI-ECAI 2026
Abstract:Large language models (LLMs) have developed rapidly and become valuable tools in everyday life. However, how to align LLMs to a particular set of human values is still an open problem. Recent studies show that instruction tuning has strong potential for zero-shot tasks and may serve as an effective approach to addressing value alignment. Nevertheless, although many datasets for instruction tuning already exist, they are not specifically designed around moral scenarios and behaviors. We construct a unified moral-value dataset that can be directly used for instruction tuning. This dataset is built upon existing moral-value datasets by merging them into a unified corpus and converting them into an instruction-response format. We show that training on a mixed dataset combining general task datasets with our dataset preserves general-task performance, and we report preliminary observations on how the mixing ratio affects value-oriented task performance. Our work provides a moral-value dataset for instruction tuning and offers a useful resource for further alignment research. The dataset is available at this https URL.
[NLP-23] A Comparative Evaluation of Embeddings and LLM s in a Greek Book Publisher Setting - The CUP Dataset
【速读】： 该论文旨在解决希腊语图书检索中缺乏标准化评估基准的问题，特别是在处理词汇、语义、噪声及跨语言查询时的检索性能评估难题。其核心解决方案在于构建CUP——一个包含868条书目记录和104个专家标注的带分级相关性判断的查询数据集，涵盖多种查询类型。关键创新点在于系统评估了稀疏（BM25）、稠密（sentence-transformers）、混合及大语言模型（LLM）辅助等多种检索方法在希腊语图书搜索场景下的表现，发现多语言嵌入模型优于专用希腊语模型，混合检索整体表现最优；进一步分析表明，BM25在命名实体查询上优势显著，而稠密与混合方法在自然语言、噪声、跨语言及概念类查询中更具鲁棒性。此外，领域感知提示（field-aware prompting）对不同模型具有差异化影响，而基于LLM的目录（TOC）摘要可提升仅依赖目录的检索效果，但LLM后过滤虽能增强早期检索精度，却伴随高昂计算成本。CUP为真实世界希腊语信息检索提供了全面评估框架。
链接: https://arxiv.org/abs/2607.21274
作者: Katerina Papantoniou,Panagiotis Papadakos,Theodore Patkos,Dimitris Garefalakis,Nikos Vardakis,Dimitris Plexousakis
机构: ICS-FORTH(希腊科技研究中心信息与计算机科学研究所); Crete University Press(克里特大学出版社); University of Crete (克里特大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Preprint of a manuscript submitted to the 14th EETN Conference on Artificial Intelligence (SETN 2026)
Abstract:We present CUP, a Greek book retrieval benchmark consisting of 868 catalog records and 104 expert-annotated queries with graded relevance judgments. We evaluate sparse (BM25), dense (sentence-transformers), hybrid, and LLM-assisted retrieval methods in this book-search setting. Multilingual embeddings outperform Greek-specific models, while hybrid retrieval performs best overall. A query-level analysis shows that BM25 excels at named-entity queries, while dense and hybrid methods improve natural-language, noisy, cross-lingual, and concept queries. Field-aware prompting has model-specific effects, while LLM TOC summarization improves TOC-only retrieval and LLM post-filtering improves early-stage retrieval at a high cost. Overall, CUP enables real-world evaluation of Greek retrieval across lexical, semantic, noisy, and cross-lingual queries.
[NLP-24] slang.gr as a Large-Scale Crowdsourced Resource for Non-Standard Greek
【速读】： 该论文旨在解决非标准语言（如俚语）在计算建模中因动态性与非规范性而难以处理的问题。其核心挑战在于如何有效捕捉俚语的语义结构、社会语言学特征及其用户社区的行为模式。解决方案的关键在于构建一个基于众包的希腊语非标准语言词典（http://url），并引入一种多层次结构化分类体系，将嘈杂的用户生成标签映射至包含语义类别与社会语言学元数据的分层分类框架中。在此基础上，通过整合用户角色、互动模式与审核信号，提出一种基于社区的定义置信度评分机制，从而提升对俚语语义和社区行为的可解释性分析能力。研究发现，希腊语俚语高度聚焦于人物相关与评价性表达，具有显著的形态创造力，并受制于参与度极不均衡、用户生命周期短且社群重叠的特征。最终，该工作不仅使非标准希腊语成为可计算的研究资源，也为社会语言学自然语言处理、偏见分析及大语言模型中的非正式语言研究奠定了方法基础。
链接: https://arxiv.org/abs/2607.21255
作者: Panagiotis Papadakos,Katerina Papantoniou,Dimitris Plexousakis
机构: ICS-FORTH(希腊科技与计算机研究所); University of Crete(克里特大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Preprint of a paper accepted for publication in the Proceedings of the 14th EETN Conference on Artificial Intelligence (SETN 2026)
Abstract:Slang is a central component of everyday language, reflecting linguistic creativity, social identity, and cultural change, yet its dy- namic and non-standard nature makes it difficult to model computationally. We present the first large-scale computational study of this http URL, a crowdsourced lexicon of Greek non-standard language, combining lexical content, user-generated tags, and interaction data. To enable the systematic analysis, we map noisy folksonomic tags to a structured multi-layer taxonomy capturing both semantic categories and sociolinguistic metadata. Using this representation, we analyze the linguistic structure of Greek slang and the behavior of its contributor community. We find that slang is strongly centered on person-related and evaluative language, exhibits high morphological creativity, and is shaped by highly skewed participation with short user lifespans and overlapping communities. Building on these signals, we introduce a community-based confidence score for definitions that integrates user roles, interaction patterns, and moderation signals. Our results show that taxonomy-based representations improve interpretability while retaining meaningful aspects of behavioral structure, enabling a more structured and interpretable analysis of confidence signals. Overall, this work establishes this http URL as a computational resource for non-standard Greek and provides a foundation for sociolinguistic NLP, bias analysis, and the study of informal language in LLMs.
[NLP-25] Progressive Cramming: Reliable Token Compression and What It Reveals
【速读】： 该论文旨在解决生成式模型中序列压缩（token cramming）的极限问题，即在固定令牌预算和99%准确率阈值下，残余误差究竟是优化不足所致，还是由本质性限制所决定。其核心解决方案是提出“渐进式压缩”（progressive cramming），通过逐个增加目标前缀的令牌数，直至在固定优化预算内无法实现重建为止，从而揭示压缩过程中的真实能力边界。该方法发现，压缩轨迹在嵌入空间中呈现出低维结构；尽管在原始前缀上下文中，预接压缩嵌入仅导致多选任务准确率轻微但稳定下降，但在生成评估中则几乎完全丧失模型能力。因果注意力剔除实验表明，性能退化源于嵌入在模型早期层中的交互作用。研究结果表明，完美重建虽可通过脆弱的控制策略实现，但缺乏可迁移的语义表征，因此不足以构成有意义的压缩。
链接: https://arxiv.org/abs/2607.21231
作者: Dmitrii Tarasov,Timofei Lashukov,Elizaveta Goncharova,Andrey Kuznetsov
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Token cramming compresses sequences into learned embeddings with near-perfect reconstruction, but fixed token budgets and 99% accuracy thresholds leave it unclear whether residual errors reflect optimization failures or fundamental limits. We introduce progressive cramming, which grows the target prefix token-by-token, stopping only when reconstruction is no longer achievable within a fixed optimization budget. Progressive trajectories occupy low-dimensional structure in embedding space. Prepending a crammed embedding causes a moderate but consistent accuracy drop on multiple-choice benchmarks even with the original prefix in context, and collapses capability almost entirely under generative evaluation. Causal attention-knockout interventions trace this degradation to the embedding’s interactions in the model’s early layers. These results position progressive cramming as a tool for studying compression limits and show that perfect reconstruction - achievable through brittle steering rather than transferable semantics - is insufficient for meaningful compression.
[NLP-26] One More Turn Less Regret: A Regret-Based Multi-Turn Benchmark for LLM s Clarification Policies
【速读】： 该论文旨在解决对话式大语言模型（LLM）在面对模糊用户请求时，如何有效进行澄清（clarification）这一序列决策问题。其核心挑战在于，模型需在“是否提问、问什么、何时停止提问以及何时给出回答”之间做出动态、智能的判断。传统评估方法仅关注单个问题的质量或最终任务成功率，难以全面衡量澄清行为的有效性与效率。为此，论文提出RegretBench——一个基于多轮交互的基准评测框架，将澄清视为策略行为（policy behavior）而非孤立的问题质量评估。其关键创新在于：引入隐藏意图（hidden-intent）建模以刻画模糊性本质，支持基于语义状态追踪的自由形式交互，并设计基于后悔值（regret-based objective）的目标函数，量化模型相对于理想澄清策略所损失的价值。实验结果表明，仅靠最终准确率不足以反映模型性能，不同模型在效率、对用户行为的鲁棒性及停止决策上存在显著差异。通过联合评估意图解析能力、交互成本、无效澄清频率与后悔值，RegretBench 能够揭示模型是否实现了有用且高效的澄清。研究进一步表明，有效的澄清不仅依赖于合理的提问内容，更要求在恰当的时机提出正确的疑问，并在用户意图明确后及时终止澄清过程。
链接: https://arxiv.org/abs/2607.21143
作者: Minh Ngoc Ta,My Anh Tran Nguyen,Duong D. Nguyen,Yuxia Wang,Preslav Nakov
机构: MBZUAI(阿联酋人工智能大学); BKAI Research Center, Hanoi University of Science and Technology (河内科技大学BKAI研究中心); INSAIT, Sofia University “St. Kliment Ohridski”(索非亚大学“圣克莱门特·奥赫里德斯基”学院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Ambiguous user requests make clarification a sequential decision problem for conversational LLM assistants: they must decide whether to ask, what to ask, when to stop, and when to answer. We introduce RegretBench, a multi-turn benchmark that evaluates clarification as policy behavior rather than isolated question quality. RegretBench provides a hidden-intent formulation of ambiguity, supports free-form interaction grounded in semantic-state tracking, and introduces a regret-based objective that measures how much value a model loses relative to a reference clarification policy. Experiments on open-domain QA and product recommendation scenarios show that final success alone is insufficient, as models with similar accuracy can differ substantially in efficiency, robustness to user behaviors, and stopping decisions. By jointly measuring intent resolution, interaction cost, ineffective clarification, and regret, RegretBench reveals whether models clarify usefully and efficiently. Our results show that effective clarification requires more than plausible questions: models must ask the right question at the right time and stop once the user’s intended meaning is clear.
[NLP-27] raining Large Language Models for Self-Explanation Faithfulness ICLR2026
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）自解释（self-explanations）中“忠实性”（faithfulness）不足的问题，即模型生成的推理过程与其内部决策机制之间缺乏准确对应。现有方法多集中于事后评估忠实性或通过推理时提示（inference-time prompting）提升可解释性，但未能直接优化模型参数以生成更忠实的自解释。本文的关键解决方案是将现有的忠实性度量（如Phi-CCT相关性指标）转化为强化学习（Reinforcement Learning, RL）的奖励信号，从而在微调过程中直接优化模型生成自解释的忠实性。实验设计了随机词插入与用户偏见词插入两类干预方式，利用样本级奖励驱动模型学习识别影响其决策的关键因素，并主动披露这些因素。结果表明，经过RL微调的Llama3.1-8B和Qwen3-8B模型在分布内任务上的Phi-CCT得分从接近零提升至最高0.664，在分布外任务（如StrategyQA）上也达到0.691，显著改善了自解释的忠实性。尽管跨干预泛化能力较弱，但观察到Llama3.1-8B在仅训练于随机词插入的情况下仍能对用户偏见插入产生非零迁移效果，暗示潜在的隐式因素识别能力，而这一现象具有模型与设置依赖性，尚待深入解释。此外，研究通过行为分析排除了奖励操控（reward gaming）的可能性，验证了训练的有效性。综上，该工作提出了一种可扩展的路径，使模型能够隐式识别并披露影响决策的关键因素，为缓解大模型中不忠实推理问题提供了新范式。
链接: https://arxiv.org/abs/2607.21090
作者: Yeoktatt Cheah,María Pérez-Ortiz,Noah Y. Siegel,Oana-Maria Camburu
机构: University College London (伦敦大学学院); Imperial College London (帝国理工学院)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: To appear at the ICLR 2026 Workshop on Representational Alignment (Re-Align), 10 pages (long paper)
Abstract:We propose a Reinforcement Learning (RL) method to directly optimize the faithfulness of self-explanations - the extent to which a model’s generated reasoning accurately reflects its internal decision-making process. While existing work focuses on evaluating faithfulness or using inference-time prompting frameworks to improve an LLM’s self-explanation’s tractability, these approaches do not provide a mechanism to directly optimize a model’s parameters to generate faithful self-explanations. We bridge this gap by modifying existing faithfulness metrics into an RL training objective. We investigate (1) if models can be trained to accurately detect factors that affect their decisions, and (2) whether RL can directly optimize for the disclosure of these factors thereby improving LLM self-explanations’ faithfulness. We experiment with two intervention types: random-word insertions and user-bias insertions, using a per-sample reward derived from the Phi-CCT correlation metric. RL fine-tuned Llama3.1-8B and Qwen3-8B show substantial improvements on the Phi-CCT faithfulness metric, with in-distribution scores rising from near-zero to as high as 0.664, and out-of-distribution scores reaching up to 0.691 on held-out tasks such as StrategyQA. Cross-intervention generalization is weaker but more interesting: a priori we would not expect a model trained only on random word insertions to generalize to user-bias phrases, yet Llama3.1-8B shows non-zero transfer in this direction. The reverse direction and Qwen3-8B do not replicate this, indicating model-dependent and setup-dependent effects we cannot yet explain. Lastly we analyze model behavior to rule out reward gaming behaviors that often plague RL training. Ultimately, we show that models can be trained to implicitly identify influential factors and disclose them, offering a scalable path toward reducing unfaithful reasoning in LLMs.
[NLP-28] VibeVoice-ASR-BitNet Technical Report
【速读】： 该论文旨在解决生成式语音识别模型在边缘计算设备上实现低延迟实时推理的挑战，尤其针对资源受限的CPU平台。其核心问题是如何在保持较高识别精度的前提下，大幅压缩模型体积并提升推理速度，以满足边缘端对计算效率与能耗的严格要求。解决方案的关键在于提出一种异构量化策略：在变分自编码器（VAE）声学分词器阶段采用全流水线INT8量化（I8_S）结合内核融合与单指令多数据（SIMD）优化，在自回归语言模型部分则引入类BitNet的二值化权重（I2_S），从而实现深度压缩；同时通过渐进式量化感知训练策略，在极端压缩条件下有效缓解精度损失。此外，基于ggml框架开发了针对ARM和x86架构的定制化SIMD内核与融合算子，使模型仅需3个CPU线程即可实现实时语音识别（RTF < 1），相较同规模模型（约1.6 GB）提速1.6–2.3倍，且精度下降幅度可控。
链接: https://arxiv.org/abs/2607.21075
作者: Songchen Xu,Ting Song,Shaohan Huang,Zhiliang Peng,Yan Xia,Yujie Tu,Xin Huang,Jianwei Yu,Li Dong,Furu Wei
机构: Microsoft Research (微软研究院); Shanghai Jiao Tong University (上海交通大学); Fudan University (复旦大学); University of Chinese Academy of Sciences (中国科学院大学)
类目: ound (cs.SD); Computation and Language (cs.CL); Audio and Speech Processing (eess.AS)
备注: Technical Report
Abstract:We present VibeVoice-ASR-BitNet, a compressed variant of VibeVoice-ASR optimized for real-time inference on edge CPUs. We apply heterogeneous quantization tailored to the computational characteristics of each stage: the VAE acoustic tokenizer uses full-pipeline INT8 quantization (I8_S) with kernel fusion and SIMD optimization, while the autoregressive language model adopts BitNet-style ternary weights (I2_S). To preserve accuracy under aggressive compression, we employ a progressive quantization-aware training strategy. For inference, we implement custom SIMD kernels and fused operators within the ggml framework targeting both ARM and x86 platforms, achieving real-time recognition with RTF 1 using as few as 3 CPU threads. VibeVoice-ASR-BitNet is 1.6-2.3x faster than this http URL at comparable model sizes (~1.6 GB), with only modest accuracy degradation compared to the FP16 baseline.
[NLP-29] PrefReward: Learning User Preference Matrix for Personalized Text Generation
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在个性化内容生成中依赖隐式参数表示所带来的可解释性差及长上下文依赖建模困难的问题。现有方法通常将用户偏好编码于模型参数内部，难以显式捕捉和分析用户特定的风格倾向，且在处理长序列时性能下降。为此，本文提出一种名为PrefReward的新型偏好感知生成框架，其核心创新在于通过结构化的偏好矩阵（preference matrix）显式建模用户的个性化风格，并将其作为解码过程中的奖励信号引入生成流程。该方案的关键在于：第一阶段从用户历史数据中提取用户特有的偏好矩阵，以量化其语言风格特征；第二阶段利用基于KL散度的奖励函数，将该偏好矩阵嵌入解码机制，引导生成内容更贴合用户风格。实验结果表明，PrefReward在LongLaMP数据集上不仅显著优于非个性化与基于检索的基线模型，在生成质量与个性化可解释性方面均取得提升。
链接: https://arxiv.org/abs/2607.21067
作者: Yue Wu,Chengbing Wang,Yimeng Bai,Xiaoyan Zhao,Yang Zhang,Fuli Feng
机构: University of Science and Technology of China(中国科学技术大学); The Chinese University of Hong Kong(香港中文大学); National University of Singapore(新加坡国立大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large Language Models (LLMs) have demonstrated remarkable ability in generating personalized content by leveraging user histories and contextual cues. However, most existing personalization approaches rely on implicit representations within model parameters, making it difficult to interpret user-specific preferences or effectively handle long-context dependencies. To address these challenges, we propose PrefReward, a novel preference-aware generative framework that explicitly models user styles through a structured preference matrix and integrates it into the decoding process as a reward signal. PrefReward consists of two stages: (1) extracting a user-specific preference matrix that summarizes individual stylistic tendencies, and (2) using the matrix to guide generation via a KL-divergence-based reward function. Experiments on the LongLaMP dataset show that PrefReward outperforms non-personalized and retrieval-based baselines in both generation quality and personalization interpretability.
[NLP-30] Sample-Efficient Learning from Agent Experience
【速读】： 该论文旨在解决现实世界中智能体学习因环境交互成本过高（如耗时实验或获取人类反馈）而受限的问题。其核心挑战在于如何在不依赖额外环境交互的前提下，将智能体从自身过往经验中获得的、通过上下文学习（in-context learning）所取得的性能提升长期保留下来。现有方法中，上下文学习虽样本高效，但其效果依赖于上下文的持续存在；而上下文蒸馏（context distillation）虽能将上下文信息内化至模型权重，但尚未有效应用于智能体的交互历史且不牺牲环境样本效率。为此，作者提出“经验蒸馏”（Experience Distillation）这一解决方案，其关键在于通过一种无需额外环境交互的机制，将已收集的经验转化为模型内部参数中的持久知识，从而实现对上下文学习收益的长期保持。实验结果表明，在749个软件工程任务和6个文本冒险游戏上，该方法可保留至少64.8%的上下文学习增益，显著优于直接监督微调（仅恢复3.8%的增益），并以少至9.6倍的环境样本量达到经典强化学习基线的性能水平。
链接: https://arxiv.org/abs/2607.21051
作者: Chenhui Gou,Haoqin Tu,Yunhao Fang,Jianfei Cai,Hamid Rezatofighi
机构: 未知
类目: Computation and Language (cs.CL)
备注:
Abstract:Real-world agent learning is often constrained by costly environment interactions, such as running time-consuming experiments or obtaining human feedback. In-context learning offers a highly sample-efficient way for agents to learn from their own interaction histories, but its gains disappear once that experience is removed from the context. Separately, context distillation provides a mechanism for internalizing contextual information into model weights. However, applying it to agents’ interaction histories without sacrificing environment sample efficiency remains underexplored. We term this problem Experience Distillation and develop an implementation that requires no further environment interaction beyond the collected experience. Experiments on 749 curated software-engineering tasks and six text-adventure games show that it retains at least 64.8% of the gains from in-context learning across both domains, whereas direct supervised fine-tuning on the collected experience recovers only 3.8%. Compared with classical reinforcement-learning baselines, in-context learning from trial-and-error experience followed by Experience Distillation matches their performance with at least (9.6\times) fewer environment samples.
[NLP-31] CultureTalk-ID: A Multi-Task Dialogue Benchmark for Cultural Commonsense in Indonesian Local Languages
【速读】： 该论文旨在解决现有印尼语文化常识评估基准普遍采用短时、孤立的提示（prompt）而忽视对话语境的问题，导致无法真实反映文化细微差别在实际交流中的呈现。其核心挑战在于如何有效评估大语言模型（LLM）在多轮对话中对文化常识的理解与运用能力。解决方案的关键是提出首个基于对话的印尼语文化常识基准——CultureTalk-ID，该基准包含跨11种语言和13个具有文化显著性的主题的4,496段文化根基深厚的对话，通过多阶段人工筛选流程并由母语者参与，确保内容的真实性与文化相关性。该基准引入三项互补任务：基于对话的多选文化常识推理、文化忠实的机器翻译以及语言引导（language steering），共同检验模型是否具备理解、迁移和生成具有文化背景的语言能力。
链接: https://arxiv.org/abs/2607.21016
作者: Muhammad Dehan Al Kautsar,Salsabila Pranida,Bilal Elbouardi,Fajri Koto
机构: Mohamed bin Zayed University of Artificial Intelligence (穆罕默德·本·扎耶德人工智能大学)
类目: Computation and Language (cs.CL)
备注: Under review
Abstract:Culture is lived through conversation, yet existing Indonesian cultural commonsense benchmarks evaluate LLMs on short and isolated prompts, stripping away the dialogic context in which cultural nuances actually surface. We introduce CultureTalk-ID, the first dialogue-based benchmark for cultural commonsense in Indonesian and its local languages, comprising 4,496 culturally grounded dialogues across 11 languages and 13 culturally salient topics, curated through a multi-stage human pipeline with native speakers to ensure authenticity. CultureTalk-ID introduces three complementary tasks, namely dialogue-based multiple-choice cultural commonsense reasoning, culturally faithful machine translation, and language steering, which jointly probe whether LLMs can understand, transfer, and generate culturally grounded language.
[NLP-32] Where Animacy Lives in Large Language Models : Tracing the Circuits of the Animacy Concept
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在书面语中区分有生命与无生命概念的能力是否源于可追溯的、因果相关的局部神经组件及其连接结构这一问题。研究发现，尽管当前LLMs展现出对动词-论元交互等复杂选择性约束和上下文线索的敏感性，其处理“生物性”（animacy）的认知机制并非集中于单一模块，而是呈现分布式、依赖上下文且具有一定梯度性的特征。解决方案的关键在于构建一组受控的最小差异对（minimal pairs）数据集，并在四个开源权重模型上实施电路发现（circuit discovery）方法，通过深入的实验与消融分析，成功识别出一个负责处理生物性的因果机制——即“生物性电路”（animacy circuit）。该电路虽存在，但其位置较分散，且在不同模型间和不同生物性任务间仅部分泛化，验证了生物性概念在模型中的表达具有高度情境依赖性和非刚性分布特性。
链接: https://arxiv.org/abs/2607.20995
作者: Samuele Punzo,Giovanni Cinà,Sandro Pezzelle
机构: University of Amsterdam (阿姆斯特丹大学); Amsterdam University Medical Center (阿姆斯特丹大学医学中心); ILLC, University of Amsterdam (阿姆斯特丹大学信息语言学研究中心)
类目: Computation and Language (cs.CL)
备注:
Abstract:Distinguishing animate from inanimate concepts in written language requires more than shallow text processing, as it involves recognizing complex selectional constraints and contextual cues, such as verb-argument interactions. Yet, current large language models (LLMs) appear to be capable of doing it. We investigate whether this animacy-sensitive behavior of LLMs can be traced to a localized set of causally relevant components and connections. To do so, we construct a controlled dataset of minimal pairs and perform circuit discovery on four open-weight models. Through in-depth experiments and ablations, we show that a causal mechanism responsible for handling animacy in these models does exist, thus discovering an animacy circuit. At the same time, this circuit appears to be less localized compared to other known ones and generalizes only partially across models and animacy tasks, confirming the distributed, context-dependent, and somewhat graded nature of the animacy concept.
[NLP-33] he Weight of Silence: A Causal Case for Weights Over the Scratchpad in Latent Chess Reasoning
【速读】： 该论文旨在解决生成式模型在强化学习（Reinforcement Learning, RL）阶段中，其潜在推理（latent reasoning）是否仍作为可主动调用的推理中间态（即内部草稿板）这一主流假设的有效性问题。现有研究多局限于数学与逻辑任务，且仅比较单一检查点内的推理依赖程度，缺乏对RL前后推理机制变化的直接因果分析。本文通过构建分阶段的潜在推理训练课程并引入强化学习，训练一个国际象棋模型，发现合法走法成功率从48%的基线单调提升至61%，同时完全消除“假性将死”（checkmate confabulation）现象。为定位性能提升来源，作者设计了六种条件的因果干预实验：对潜在思维向量注入匹配噪声或进行消融处理均未显著影响性能，唯独精确置零（exact-zero）扰动导致性能崩溃；值得注意的是，该精确置零下的鲁棒性差距在RL后显著扩大——前RL阶段合法性降至1%，而后RL阶段仍维持9%的残余表现，且该差异在全电池测试中持续存在。这表明强化学习并未增强模型对具体推理内容的依赖，而是提升了其对潜在表示破坏的鲁棒性。因此，核心发现是：潜在推理的主要作用并非在推理时提供可访问的中间计算状态，而是在训练过程中通过塑造模型参数实现性能优化。此外，本研究首次在非数学/逻辑领域（国际象棋）成功实现强化学习带来的实际性能增益，挑战了此前多个团队在类似任务中报告的“潜在推理+强化学习”组合无效的结论。
链接: https://arxiv.org/abs/2607.20952
作者: Ishan S. Kshirsagar
机构: Independent Researcher(独立研究员)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 28 pages, 5 figures, preprint also available at Zenodo: this https URL
Abstract:Latent, or silent, reasoning lets language models carry out intermediate computation in continuous vector space instead of words, and is widely assumed to function as an internal scratchpad the model actively consults during inference. Whether that assumption survives reinforcement learning has not been tested directly: existing causal analyses of latent reasoning are confined to math and logic tasks, and compare a model’s reliance on its thoughts within a single checkpoint, never before and after an RL stage. We train a chess-playing model through a staged latent-reasoning curriculum followed by reinforcement learning, and find legality climbs monotonically to 61% (from a 48% pre-RL baseline) while checkmate confabulation is eliminated entirely. To locate this gain, we run a six-condition causal intervention suite on the same model before and after RL: substituting or adding matched noise to the latent thought vectors leaves performance unchanged, ablating them causes only mild degradation, and only exact-zero vectors cause collapse. This robustness gap is itself the finding: under exact-zero corruption, legality collapses to 1% pre-RL versus 9% post-RL, a gap that survives correction for testing across the full battery; milder conditions trend similarly without independently reaching significance. RL appears to add robustness to disruption, not reliance on thought content. These results push back against the field’s default assumption that latent thoughts function as an actively consulted inference-time scratchpad, and instead indicate latent reasoning’s principal effect here is shaping the model’s parameters during training. We also demonstrate a working RL gain in chess, a domain outside the math and logic settings where multiple groups report the same latent-reasoning-plus-RL recipe failing to improve accuracy over SFT.
[NLP-34] From a Word-Level Dictionary to Sentence-Level Semantics: Multilingual Grievance Labelling with Contextual Models
【速读】： 该论文旨在解决当前基于词典的仇恨言论与怨恨情绪（grievance）量化方法中存在的根本性缺陷：传统词典通过逐词匹配加权词汇来评估怨恨程度，但无法识别词语在语境中的实际语义角色（如主张、引用、否定或谴责），且存在严重的评价循环问题——即评估数据集往往由词典自身筛选出的样本构成，导致高分结果部分源于词典自身的选择偏好。其核心解决方案是保留原有22个构念的分类体系，但将词项匹配替换为基于上下文理解的深度语言模型，通过一个非循环的多语言基准测试（涵盖五种语言，2000个样本）对模型进行评估，该基准明确区分了无条件随机文本、词典正向和词典负向三类样本。实验表明，采用全文上下文建模显著提升了对隐含、引用及跨句表达的怨恨内容的识别能力，尤其在词典沉默区域表现突出，使词典负向文本的平均精确率从0.14提升至0.20。研究证实，通过上下文理解可更真实地捕捉怨恨情绪，并在独立于词典选择标准的基准上实现更诚实的性能评估。
链接: https://arxiv.org/abs/2607.20946
作者: Lin Tian,Marian-Andrei Rizoiu
机构: University of Technology Sydney (悉尼科技大学)
类目: Computation and Language (cs.CL)
备注: 12 pages, 1 figure, 9 tables
Abstract:Grievance is one of the warning signs analysts look for when assessing threats of violence. It is increasingly measured at scale from online text, most often with word-level lexicons like the Grievance Dictionary that score by matching weighted terms. Such matching is a fast and transparent proxy, but it cannot resolve whether a term is asserted, quoted, negated, or condemned. These lexicons are also often evaluated on pools enriched with the very examples they retrieve, so a high score partly reflects agreement with the lexicon’s own selection rule. Examining a five-language, 2,000-item evaluation pool, we find its halves separated almost perfectly by the lexicon itself: every item labeled ``random’’ is in fact lexicon-negative, so the lexicon’s apparent macro-AUROC of 0.686 collapses to a 0.500 floor fixed by construction. We keep the dictionary’s 22-construct ontology but replace term matching with context-reading models, evaluated on a non-circular benchmark that separates unconditional-random, lexicon-positive, and lexicon-negative strata across five languages. Reading the full post rather than the target sentence alone helps most where the lexicon is silent, raising average precision on lexicon-negative text from 0.14 to 0.20, with the largest gains on quoted, implicit, and cross-sentence grievance. Together, these results show that grievance is measured more faithfully by reading the surrounding context, and more honestly when tested on text the lexicon did not select. We release our code and benchmark at this https URL.
[NLP-35] Chemical Chain-of-Thought Functions as a Hallucination-Prone Molecular Scratchpad
【速读】： 该论文旨在解决生成式化学推理模型在链式思维（Chain-of-Thought, CoT）过程中普遍存在幻觉（hallucination）的问题，即模型虽能给出正确答案，但其推理过程常包含与实际分子结构无关的虚构性结构描述。研究发现，尽管这些虚构的结构声明与真实分子无关，但推理轨迹在生成过程中仍具有关键作用。解决方案的关键在于揭示了化学推理中的链式思维并非忠实的逻辑解释，也非纯粹的事后合理化，而更像是一种“分子草稿板”（molecular scratchpad）——模型通过特定形式的中间表示来辅助生成，如Chem-R和ether-0依赖碎片化的SMILES草图，而ChemDFM-R则侧重于骨架、位置及命名线索。值得注意的是，对Chem-R的SMILES草图进行扰动会显著降低生成质量，表明这些结构草图在生成过程中具有因果性作用，即使其文字描述的结构声明基本无效。因此，该研究强调不应将链式思维视为模型推理忠实性的直接证据，并呼吁采用以过程为导向的监督机制，超越仅依赖最终答案的评估方式。
链接: https://arxiv.org/abs/2607.20935
作者: Jiatong Li,Yuxuan Ren,Weida Wang,Xiaoyong Wei,Yatao Bian
机构: National University of Singapore (新加坡国立大学); Hong Kong Polytechnic University (香港理工大学); Fudan University (复旦大学)
类目: Computational Engineering, Finance, and Science (cs.CE); Computation and Language (cs.CL)
备注: 16 pages, 6 figures
Abstract:Chemical reasoning language models are expected to derive molecular answers through faithful chain-of-thought (CoT). However, across four reasoning model families and twelve chemistry tasks, hallucination is widespread and largely decoupled from answer correctness: correct answers often coexist with fabricated structural claims absent from the relevant molecules. Yet this does not make the reasoning trace computationally irrelevant. Attribution analyses suggest a shared scratchpad function expressed in model-specific forms: Chem-R and ether-0 rely on fragmented SMILES drafts, whereas ChemDFM-R emphasizes scaffold, positional, and naming cues. Notably, perturbing Chem-R’s SMILES sketches degrades generation, showing that structural drafts can be causally load-bearing even when verbal structural claims are largely inert. Together, these results show that chemical CoT is neither a faithful explanation nor merely a post-hoc rationalization, but a hallucination-prone molecular scratchpad. This finding cautions against treating CoT as direct evidence of faithful reasoning and motivates process-level supervision beyond answer-only evaluation.
[NLP-36] ransformer-Assisted LLM -Based Source Code Summarisation: to Enable More Secure Software Development
【速读】： 该论文旨在解决软件维护阶段中源代码注释缺失、不完整或过时的问题，这些问题严重影响开发人员对代码的理解，进而增加软件系统中的漏洞与缺陷。现有解决方案主要依赖于任务特定的Transformer模型或代码感知型大语言模型（LLM）。尽管任务特定的Transformer模型在自然语言生成（NLG）指标上表现良好，但其评分机制过度依赖词汇重叠，无法有效衡量摘要质量；而大语言模型虽具备强大的语义理解能力，能生成高质量的摘要，但由于其生成内容与开发者原生表达存在显著差异，导致在标准NLG指标上得分较低。本文提出的关键解决方案是通过提示工程（prompt engineering）将任务特定的Transformer模型生成的摘要作为上下文信息嵌入到大语言模型的输入中，从而引导其生成更贴近开发者语言习惯、语义准确且质量更高的代码摘要。该方法实现了“基于Transformer辅助的大语言模型源代码摘要生成”（Transformer-Assisted LLM-Based Source Code Summarisation），实验结果显示在BLEU-4和BLEU-5指标上分别提升了7.8%和5%，显著改善了生成摘要的质量与实用性。
链接: https://arxiv.org/abs/2607.20933
作者: Jesse Phillips,Tracy Hall,Paul Rayson,Mo El-Haj
机构: Lancaster University (兰卡斯特大学); VinUniversity (越南大学)
类目: oftware Engineering (cs.SE); Computation and Language (cs.CL)
备注: 10 pages
Abstract:Neural Source Code Summarisation (NSCS) aims to generate natural language summaries of source code to improve developers’ and maintainers’ understanding of code. Source code summaries are vital during the maintenance phase of the Secure Software Development Lifecycle (SSDLC), as they improve maintainers’ understanding of code and help reduce the number of bugs and vulnerabilities in a software system. However, summaries are often missing, incomplete, or outdated in many software systems. Solutions to this problem use small, task-specific Transformer models or code-aware Large Language Models (LLMs). Task-specific Transformer-generated summaries often score well across many natural language generation (NLG) metrics, but these metrics reward lexical overlap rather than summary quality. Conversely, the ability of LLMs to capture semantics and produce high-quality summaries presents an exciting solution to this problem. This is especially relevant given the increased availability of LLMs and improvements in workstation hardware in recent years, which mean that some LLMs can now be run on developers’ workstations. However, because of their abstractive nature, LLM-generated code summaries often differ greatly from developer-written summaries in the words and phrases they use, resulting in low scores across NLG metrics. We show how combining these two methods, by using Transformer-generated summaries in prompt engineering, may enable LLMs to create better source code summaries and help software practitioners maintain secure systems. We prompt four LLMs using four different prompts, with a task-specific Transformer used to assist the LLMs within the prompts. We present “Transformer-Assisted LLM-Based Source Code Summarisation”, a method through which we observe an improvement of 7.8% in BLEU-4 and 5%.
[NLP-37] ncent WorkBuddy Bench: A Multi-Domain Coding-Agent Benchmark with Contamination-Resistant Task Construction
【速读】： 该论文旨在解决现有代码生成智能体（Coding Agent）评估体系中存在的任务真实性不足、评估结果可复现性差以及数据污染风险高等问题。其核心挑战在于如何构建一个既能真实反映开发者在多个工作领域（代码开发、前端开发、办公与业务流程、安全攻防）中实际工作场景，又具备高度可复现性与抗污染能力的多领域评测基准。解决方案的关键在于提出一种基于真实代码提交（commit）、拉取请求（pull request）或业务场景逆向重构的任务生成方法：所有任务均从真实工程活动出发，通过重写为简明、口语化且带有角色扮演性质的指令形式，确保任务提示无法通过网络搜索还原原始上下文，从而有效防止数据泄露和模型“记忆”带来的评估偏差。此外，该基准采用开放发布策略，包含任务目录、环境镜像、评估框架、测试用例及参考解法，并结合版本控制机制，实现端到端可复现与第三方审计。四个子集分别覆盖仓库级工程、前端开发、办公自动化及红蓝队安全任务，各自采用适配的验证方式与评分标准，导致跨子集分数不可直接比较，因此不提供全集平均分，仅报告各模型在不同子集上的对比排名，从而更准确地反映模型在多样化真实工作场景中的性能表现。
链接: https://arxiv.org/abs/2607.20911
作者: Tencent WorkBuddy Bench Team:Siqi Cai,Shaopeng Chen,Xiang Fei,Yong Mao,Zihan Xu,Zhiheng Lyu,Zhijian Shao,Yuchen Shi,Shuwen Zhang,Chaofan Qiu,Linjie Che,Xiaoxi Zhao,Feng Wu,Kai Zhang,Chaofan Zhu,Yubin Qi,Xiaoyun Liang,Peijie Dong,Yunhao Zhang,Yuanjie Zhu,Ling Jiang,Xianjun Zhang,Zhehang Chu,Anyuan Sang,Zhen Feng,Sen Nie,Shi Wu,Yuanzhen Xu,Xin Li,Ning Yang,Zhiqiang Dong,Hande Dong,Qiang Lin,Yi Liu,Yunsheng Wu,Ke Li,Xing Sun
机构: Tencent WorkBuddy Bench Team (腾讯工作伙伴基准团队); Youtu Lab (优图实验室); Keen Security Lab (奇安信安全实验室); Workbuddy (工作伙伴); Yunding Security Lab (云鼎安全实验室)
类目: Computation and Language (cs.CL); Software Engineering (cs.SE)
备注: 30 pages, 9 figures. Project page: this https URL ; code: this https URL ; dataset: this https URL
Abstract:We introduce Tencent WorkBuddy Bench, a multi-domain evaluation suite for coding agents; this report documents its construction methodology, scoring protocol, and a cross-model leaderboard. At its core is a unified evaluation framework for constructing and running distribution-informed coding-agent tasks across four work domains - Code, Web, Office, and Security. Rather than adapting public issue text, every task is reverse-engineered from a real commit, pull request, or business scenario and rewritten as a short, colloquial, role-played request, so that a task’s prompt is not recoverable by web-searching the underlying issue, pull request, or commit thread. Because the dataset is released openly - task directories, environment images, evaluation harness, tests, and reference solutions - contamination resistance rests on this construction together with dataset versioning rather than on secrecy. The four subsets - repository-level engineering, front-end development, office and business workflows, and red-/blue-team security - probe complementary facets of real work, each with its own verification style. All are packaged in a uniform task-directory format and run, under a uniform and reproducible protocol, on two agent harnesses (CodeBuddy Code and Claude Code); the full open release makes the benchmark reproducible end to end and directly auditable, since any third party can re-run each task and inspect its content. Because each subset uses a different scoring instrument, scores are not comparable across subsets and the suite reports no suite-wide average. We report a cross-model leaderboard across several model families.
[NLP-38] LegalCiteTrust: Benchmarking Citation Trustworthiness in Chinese Long-Form Legal Research Reports
【速读】： 该论文旨在解决生成式法律研究报告中引用权威文献的可信度问题，尤其关注大语言模型（LLM）在生成长篇法律研究报告时，虽能完成任务但可能存在引用不准确、过度引申或忽略限制条件等风险。其核心挑战在于：即使引用了真实法律依据，若存在引用失真、描述偏差或适用范围不当，仍会损害结论的可靠性。为此，论文提出LegalCiteTrust基准，用于评估中文长篇法律研究报告中引用的可信度，涵盖报告级任务72项，并从覆盖度（Coverage）、支持力（Support）和引用可信度（Citation Trustworthiness）三个维度进行评价。其中，引用可信度通过引用层级的三重指标——存在性（Existence）、忠实性（Fidelity）和适用性（Applicability，简称E/F/A）来量化。实验表明，通用大模型、深度研究系统及法律专用系统在任务完成率、证据丰富度、引用密度与引用可靠性方面表现出显著差异；尽管检索工具可提升证据支持程度，却无法稳定提高可信度得分；而基于E/F/A标准的修订机制则更有效地提升了引用可信度与最终评分。研究结果表明，实现可靠法律研究生成的关键在于“引用感知”的证据治理机制——即系统不仅需准确检索法律权威，还需在选择、描述与适用环节确保引用的准确性与合理性。
链接: https://arxiv.org/abs/2607.20872
作者: Yunhan Li,Mingjie Xie,Zeyang Shi,Gengshen Wu,Min Yang
机构: City University of Macau (澳门城市大学); Chinese Academy of Sciences (中国科学院); Southern University of Science and Technology (南方科技大学)
类目: Computation and Language (cs.CL)
备注: 8 pages, 21 pages with appendix, 26 tables, 4 figures
Abstract:Long-form legal research reports increasingly rely on LLMs and agentic research systems, but their reliability depends not only on answering the task, but also on whether cited legal authorities are trustworthy. A citation can be risky even when it points to a real source: the report may omit limiting conditions, misdescribe the authority, or use it to support a stronger claim than the source allows. We introduce LegalCiteTrust, a benchmark for evaluating citation trustworthiness in Chinese long-form legal research reports. It contains 72 densely annotated report-level tasks and evaluates reports along three dimensions: Coverage, Support, and Citation Trustworthiness. Citation Trustworthiness is operationalized through citation-level Existence, Fidelity, and Applicability (E/F/A). Experiments on general-purpose LLMs, deep-research systems, and legal-specific systems show that task completion, evidence richness, citation density, and citation reliability expose different system behaviors. Retrieval tools can improve evidence support without reliably improving the Trust score, while E/F/A-based revision improves Trust and Final score more clearly than existence-only filtering. These results suggest that trustworthy legal research generation requires citation-aware evidence governance after retrieval: systems must not only retrieve legal authorities, but also select, describe, and apply them reliably.
[NLP-39] Position Bias is Hidden Behind Ceiling Effects: A Permutation Diagnostic for LLM Benchmarks ACL
【速读】： 该论文旨在解决多选题大语言模型（LLM）评估中普遍存在的位置偏差（position bias）问题，其核心挑战在于现有测量方法依赖于单次答案顺序打乱，导致位置偏差信号与内容级噪声及采样随机性混杂，难以准确识别。为此，作者提出并实现了一种名为inspect_permute的开源工具，作为inspect_ai评估框架的扩展，通过为每个题目执行全部可能的答案顺序排列（穷尽式置换），并基于卡方检验（chi-squared）与克雷默V系数（Cramer V）量化位置偏差，同时提供自助法（bootstrap）置信区间以增强统计可靠性。在对四个主流模型（gpt-4o-mini、claude-haiku-4-5、gemini-2.5-flash、grok-3）在五个MMLU主题上的大规模测试中（共24,000次API调用，温度=0生成），研究发现：位置偏差仅在约60%-95%的基础准确率“黄金区域”内可被统计检测到；低于此区间时，处理负载主导效应掩盖了任务特异性信号；高于此区间则因天花板效应压缩方差，导致卡方检验无法分辨差异。可检测的偏差模式分为两类：单调递减型（从A至D逐步下降，反映处理负载影响，常见于低性能模型）与非单调突降型（D项显著下降，源于内容模糊性，存在于特定能力区间）。由于前沿模型均位于该检测阈值之上，其未表现出显著偏差信号应解释为“不可测量”而非“无偏”，从而澄清了当前评估范式的误解。结合先前工作（arXiv:2606.26185）对天花板效应的刻画，本研究明确了位置偏差可测范围的边界，使该领域核心问题得以以可验证形式提出。相关代码、数据与预注册哈希均已公开（MIT许可）。
链接: https://arxiv.org/abs/2607.20864
作者: Hiroki Tamba
机构: Independent researcher.
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注: 25 pages, 4 figures, 2 appendices. Code, data, and preregistration verification at this https URL . Companion paper: arXiv:2606.26185
Abstract:Position bias in multiple-choice LLM evaluation is widely cited as a confound in capability comparisons, but published measurements rely on single answer-order shuffles whose results confound the bias signal with content-level noise and sampling stochasticity. I introduce inspect_permute, an open-source extension to the inspect_ai evaluation framework that runs exhaustive answer-order permutations per question and reports the chi-squared / Cramer V signature of position bias with bootstrap confidence intervals. I apply the tool across four vendors (gpt-4o-mini, claude-haiku-4-5, gemini-2.5-flash, grok-3) on five MMLU subjects, 24,000 API calls under temperature-0 generation, with falsifier predictions pre-registered via a public SHA-256 hash before half the data was observed. Position bias turns out to be statistically detectable only within a roughly 60-95% base-accuracy Goldilocks zone. Below it, processing-load dominance swamps subject-specific signal; above it, ceiling effects compress the variance below the chi-squared test resolution. Detectable cells separate into two mechanism types: monotone A-to-D decrease (processing_load, in low-tier models) and non-monotone D-drop (content_ambiguity, in a narrow capability band). Standard MMLU places every frontier-tier model above the detection band, so absence of signal there should be read as not measurable, not unbiased. Together with the ceiling-effect characterisation in arXiv:2606.26185, this work brackets the detectable region of position-bias measurement and makes the field central question askable in a verifiable form. Package, data, preregistration under MIT.
[NLP-40] CSPF: A Constrained Shared-Private Fusion Method for Non-Verifiable Preference Evaluation
【速读】： 该论文旨在解决非可验证任务（non-verifiable tasks）中可靠评估难的问题，现有方法往往无法充分捕捉人类偏好背后多样化的评价标准。其解决方案的关键在于提出一种约束性共享-私有融合（Constrained Shared-Private Fusion, CSPF）机制，将异构的冻结奖励模型视为互补的评估者，并在成对人类偏好监督下学习融合其隐藏状态表示。CSPF通过将每个专家信号分解为共享表示与专家私有表示，既促进跨专家间的对齐，又保留了各模型的独特视角。在LM-Arena目标域适应和PPE分布外偏好评估等多个实验中，CSPF在主要指标上均优于单一专家奖励模型、标量得分多专家融合及评分卡式判断等基线方法。研究表明，基于隐藏状态表示的融合能够为偏好评估提供更具表现力的建模基础，为非可验证偏好任务中的集成评估信号提供了可行路径。
链接: https://arxiv.org/abs/2607.20862
作者: Hehao Zhang,Danli Wang,Xinyuan Wang,Xuange Gao
机构: Institute of Automation, Chinese Academy of Sciences (中国科学院自动化研究所); School of Artificial Intelligence, University of Chinese Academy of Sciences (中国科学院大学人工智能学院)
类目: Computation and Language (cs.CL)
备注: 15 pages, 6 figures, 5 tables
Abstract:At present, reliable evaluation of non-verifiable tasks remains challenging. Existing approaches often fail to adequately capture the diverse evaluative criteria underlying human preferences in such tasks. To this end, we propose Constrained Shared-Private Fusion (CSPF), a fusion method that treats heterogeneous frozen reward models as complementary evaluators and learns to integrate their hidden-state representations under pairwise human-preference supervision. CSPF decomposes each expert signal into shared and expert-private representations, encouraging cross-expert alignment while preserving complementary viewpoints. Across experiments on LM-Arena target-domain adaptation and PPE out-of-distribution preference evaluation, CSPF achieves the best performance on the primary metrics among the evaluated single-expert reward-model, scalar-score multi-expert, and rubric-judge baselines. Overall, CSPF suggests that fusing hidden-state representations provides a more expressive basis for preference assessment, offering a practical route toward integrated evaluative signals for non-verifiable preference tasks.
[NLP-41] REFACT: Adaptive Fact Restatement for Compact and Faithful Chain-of-Thought Reasoning
【速读】： 该论文旨在解决大语言模型在长文本推理任务中因证据稀疏、噪声或与参数化知识冲突而导致推理轨迹偏离原始上下文的问题。现有基于归因的方法通常在生成后附加引用，或在推理过程中鼓励检索证据，但难以确保所引内容足以支撑局部推理和最终答案。为此，本文提出REFACT——一种自适应的事实重述引用框架，其核心在于训练模型自主判断推理步骤是否需要上下文归因，并确定源事实应以何种粒度进行重述。该设计通过将引用转化为支持答案的中间状态，有效避免了无依据推理与盲目复制事实的问题。REFACT采用两阶段监督微调至强化学习（SFT-to-RL）优化流程，引入引用效用奖励机制，促使引用事实具备良好的结构化、可追溯性及对答案的充分性。在LongBench、LV-Eval和ConFiQA上的实验表明，REFACT不仅提升了长上下文问答性能与反事实忠实性，还显著降低了令牌消耗；进一步分析显示，其以更少的事实重述保留了更多承载答案的关键证据，使推理轨迹更为密集而非冗长。
链接: https://arxiv.org/abs/2607.20833
作者: Zhensheng Jin,Xin Dai,Zhenghao Liu,Chaojun Xiao,Huiyuan Xie,Yu Gu,Ge Yu,Maosong Sun
机构: Northeastern University(东北大学); Tsinghua University(清华大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Large language models increasingly rely on long-form reasoning for complex tasks, yet their reasoning traces may drift away from the supplied context when evidence is sparse, noisy, or in conflict with parametric knowledge. Existing grounding methods either attach citations after generation or encourage evidence retrieval inside the trace, but they often do not ensure that cited content is sufficient for the local inference and final answer. We propose REFACT, an adaptive fact-restatement citation framework that trains models to decide when a reasoning step needs contextual grounding and at what granularity source facts should be restated. This design avoids both unsupported inference and indiscriminate fact copying by turning citations into answer-supporting intermediate states. REFACT is optimized with a two-stage SFT-to-RL pipeline in which a citation-utility reward encourages cited facts to be well-formed, source-traceable, and answer-sufficient. Experiments on LongBench, LV-Eval, and ConFiQA show that REFACT improves long-context QA and counterfactual faithfulness while substantially reducing token consumption. Further analysis shows that REFACT preserves more answer-bearing evidence with fewer restated facts, yielding reasoning traces that are denser rather than longer. All code and data are available at this https URL.
[NLP-42] Beyond Heavy Log Curation: Perplexity-Based APT Detection via Unsupervised Context-Augmented Language Models
【速读】： 该论文旨在解决高级持续性威胁（Advanced Persistent Threats, APT）在大规模日志中难以检测的问题，核心挑战在于攻击相关事件占比极低，且传统人工分析成本高、难以规模化。现有基于机器学习的检测方法虽能减轻分析负担，但通常依赖高度定制化的训练数据和复杂的预处理流程，导致开发与维护成本高昂。本文提出CAPTAIN（Context-Augmented Perplexity-based Threat Activity log detectIoN），其关键创新在于采用基于困惑度（perplexity）的检测机制，利用通用预训练语言模型（pre-trained language models）进行建模，仅需极少且领域无关的预处理，即可对长序列、未经深度加工的日志条目实现鲁棒评分。CAPTAIN通过编码器-桥接（Q-Former-style bridge）结构捕获近期上下文，并将压缩后的上下文令牌注入解码器输入，使困惑度能够反映时间上下文信息；同时引入平滑滤波器增强困惑度时序信号的稳定性。实验表明，CAPTAIN在多个面向APT的基准测试中表现媲美强基线模型，且在显著降低数据精细化程度的前提下仍保持稳健性能，有效降低了先进日志预处理的开发与运维成本。
链接: https://arxiv.org/abs/2607.20832
作者: Shoya Otsu,Kei Suzuki,Toshiaki Koike-Akino,Jing Liu,Ye Wang
机构: University of Tokyo (东京大学); National Institute of Information and Communications Technology (国家信息通信技术研究所)
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Cryptography and Security (cs.CR)
备注: 20 pages
Abstract:Advanced Persistent Threats (APTs) remain difficult to detect because only a small fraction of events in large-scale logs are attack-related, and investigation is expensive and hard to scale. Prior machine-learning approaches can reduce analyst workload, but they often rely on heavily curated training data and sophisticated preprocessing pipelines. Building and maintaining such pipelines require substantial domain expertise and engineering cost. Motivated by insights from a study of a strong APT detection baseline, we propose CAPTAIN (Context-Augmented Perplexity-based Threat Activity log detectIoN), a perplexity-based detector that leverages general, pre-trained language models with minimal, domain-agnostic preprocessing, enabling robust scoring of long, minimally processed log entries. CAPTAIN encodes recent history with an encoder model and a Q-Former-style bridge, then injects the compact context tokens into the decoder input so that perplexity reflects temporal context. To improve stability, CAPTAIN additionally applies smoothing filters to the perplexity time series. Across APT-oriented benchmarks, CAPTAIN competes with strong existing baselines and remains robust under substantially less curated inputs, that reduces the development and operational cost of advanced log preprocessing.
[NLP-43] he Geometry of Personality: Activation Steering with Jungian Cognitive Functions
【速读】： 该论文旨在解决大语言模型（LLM）中人格控制与可解释性不足的问题，现有研究多依赖静态人格框架（如大五人格模型）进行建模，难以捕捉人格的动态认知过程。为此，论文提出以荣格八种认知功能（Jungian Cognitive Functions）作为人格的动态认知表征，构建了一套包含2100余条角色扮演叙事数据集及相应的荣格评估协议的框架。其核心解决方案在于通过激活引导（activation steering）技术，在Llama-3.1-8B模型上实现了对全部八种认知功能的有效单调控制。关键发现包括：1）人格信息主要集中在中间层变压器层；2）引导向量表现出与理性/非理性功能区分一致的结构化几何关系；3）多维人格控制方向无法由单一功能方向的线性组合恢复，表明其内在复杂性。该研究为理解人格在大模型激活空间中的表征提供了新视角，并建立了可解释、高效且支持多维度人格调控的新范式。
链接: https://arxiv.org/abs/2607.20803
作者: Liu Zai(1),Yumeng Wang(2),Junchen Fu(1),Joemon M. Jose(1) ((1) University of Glasgow, (2) Leiden University)
机构: University of Glasgow(格拉斯哥大学); Leiden University(莱顿大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 15 pages, 13 figures
Abstract:Activation steering enables control and interpretation of LLMs, yet existing work primarily models personality through static trait frameworks such as the Big Five. We investigate whether personality can instead be represented and controlled as a set of cognitive processes using the eight Jungian Cognitive Functions. To this end, we introduce a framework comprising a Jungian evaluation protocol and a dataset of over 2,100 role-playing character narrations. Activation steering vector extraction and evaluation experiments on Llama-3.1-8B demonstrate effective monotonic control over all eight cognitive functions through activation steering. Beyond controllability, our analysis reveals that: 1. personality information is concentrated in middle transformer layers; 2. steering vectors exhibit structured geometric relationships consistent with distinctions between rational and irrational functions; 3. effective multi-dimensional steering directions cannot be recovered as linear combinations of single-function directions. These findings provide new insights into the representation of personality in LLM activation space and establish a framework for studying interpretable, effective, and multi-dimensional personality control. Comments: 15 pages, 13 figures Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) Cite as: arXiv:2607.20803 [cs.CL] (or arXiv:2607.20803v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2607.20803 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-44] Refusal-Gated Decoding: Preserving Refusal Behavior Under High-Temperature Sampling
【速读】： 该论文旨在解决高温度采样（high-temperature sampling）在提升大语言模型（LLM）输出多样性的同时，导致模型安全防护能力下降的问题，即高熵的词元概率分布会削弱模型对有害提示的拒绝响应（refusal behavior）。其解决方案的关键在于提出一种高效的顺序解码方法，能够在保持高温度采样带来的输出多样性优势的同时，几乎完整地保留模型在贪婪解码（greedy decoding）下的拒绝行为，且仅引入可忽略的额外延迟。实验结果表明，该方法在三个基准数据集上能够维持91%-99%的贪婪解码拒绝率，同时不损害模型对安全提示的高温度响应质量，从而实现了高多样性与强安全性的有效平衡。
链接: https://arxiv.org/abs/2607.20791
作者: Phillip Howard,Xin Su,Allen Roush,Manikandan Ravikiran,Amir Abdullah
机构: Thoughtworks
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:High-temperature sampling is one of the primary mechanisms for increasing diversity in LLMs. Recent advances in truncation-based sampling techniques have helped mitigate drawbacks of high-temperature sampling such as neural text degeneration, thereby enabling greater diversity in LLM outputs without sacrificing coherence. However, increasing the entropy of the token probability distribution via high temperatures has also been shown to weaken model guardrails by reducing the model’s refusal response in the presence of harmful prompts. Despite the potential benefits of high-temperature sampling and the importance of maintaining model safety, there is a lack of existing solutions for maintaining the refusal behavior of LLMs under a higher entropy regime. To address this gap, we systematically study how temperature influences refusal behavior in LLMs and propose an efficient sequential decoding approach which preserves a model’s greedy decoding refusal response at high temperatures while incurring minimal additional latency. Through extensive experiments, we show that our approach preserves 91-99% of the greedy decoding refusal behavior across three benchmark datasets without compromising the model’s high-temperature response for safe prompts. Our work demonstrates how refusal behavior can be maintained in an efficient manner for applications which require high-temperature sampling.
[NLP-45] Are Diversity Metrics Measuring Diversity? A Capability-Controlled Audit of Majority-Vote Gain in LLM Ensembles
【速读】： 该论文旨在解决生成式 AI（Generative AI）中大规模语言模型（LLM）集成时“多样性是否真正带来性能增益”的核心问题，尤其关注现有多样性度量是否有效区分多样性与模型能力（capability）的混杂影响。其关键在于通过在严格控制模型能力的前提下，系统评估五种主流多样性度量对多数投票（majority voting）性能提升的预测能力。研究发现：第一，尽管理论上存在广泛潜在互补性（latent complementarity），但简单投票仅在极少数情况下超越最强单个模型（如大小为3的子集仅9.98%），且这一现象部分源于偶数规模投票的确定性行为；第二，现有的严格多样性（strict diversity）度量与平均准确率的补数高度共线（大小为3时斯皮尔曼相关系数达+0.991），表明多数多样性度量本质上反映的是能力差异而非真正的多样性；第三，基于联合共现模式的线性统计量在数学上不可分离，经能力控制后，仅剩一个微弱但稳健的成对共失败关联——即模型间错误越一致，集成增益越低，但该效应幅度依赖具体配置。因此，论文揭示了当前多样性度量在实际应用中的根本局限，并指出应重新审视集成策略的设计基础。
链接: https://arxiv.org/abs/2607.20768
作者: Donghwan Kim
机构: Aidentyx Inc., San Jose, CA, USA
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 10 figures, 9 tables
Abstract:Majority voting over LLMs is widely assumed to benefit from diversity, and diversity measures are used to choose which models to combine. We ask whether five such measures track diversity or mainly re-express capability, auditing them as predictors of majority-vote gain over the best member across 31,900 subsets of 30 LLMs on MMLU-Pro (29 on TruthfulQA) under explicit capability controls. Three findings emerge. First, latent complementarity is ubiquitous: oracle gain is positive in 100% of subsets, yet simple voting beats the strongest member in only 9.98% of all canonical size-3 subsets (18.71% with held-out best selection); the pooled size-2-4 rate is 1.27%, partly reflecting deterministic even-size voting behavior. Second, a joint-correctness proxy (strict diversity) is nearly collinear with one minus mean accuracy (size-3 Spearman rho = +0.991 / +0.988); raw diversity-gain associations are strongly capability-entangled and, with one exception, unstable under control. Third, three linear contingency-table statistics are algebraically non-separable; after capability control, the empirically stable remainder is a modest residual pairwise co-failure association in which more shared error corresponds to lower gain. This direction is robust, but its magnitude is configuration-dependent. Joint rawspace linear regressions treating strict diversity, disagreement, and double-fault as independent predictors are rank-deficient by construction.
[NLP-46] Rushes: A Human Preference Dataset for Pluralistic Alignment
【速读】： 该论文旨在解决生成式叙事系统中用户个性化参与偏好难以被有效建模与对齐的问题，尤其关注在交互式分支叙事环境中用户行为的动态性与异质性。其核心挑战在于：现有大语言模型（LLM）虽具备强大的生成能力，但在捕捉个体用户在序列化决策中的细微偏好差异方面表现不佳，导致模型倾向于采纳群体平均偏好而非个性化轨迹。解决方案的关键在于构建一个大规模、高保真的交互数据集——Rushes，通过游戏化界面收集用户在AI生成的分支叙事中逐点选择的行为数据，包含完整的候选选项、用户选择及上下文演化信息，形成带有持久用户标识的时间有序轨迹。研究发现，用户选择呈现结构化、非随机模式（表现为低选择熵），表明存在可量化的个性化信号；然而，尽管经典矩阵分解方法（如SVD）能捕获37.7%的个性化信号，前沿LLM（如GPT-5）在事件级选择预测上仅达到34.23%，甚至低于流行度基线（36.4%），揭示出“参与差距”（Engagement Gap）。这一现象表明，当前基于单一全局目标（如强化学习人类反馈，RLHF）的训练范式无法充分建模依赖上下文、异质性的用户参与信号，致使模型无法实现真正的个性化适应。因此，该研究提出将Rushes作为多极对齐（pluralistic alignment）的诊断基准，推动面向序列化决策的生成系统向更精细、个体化的行为建模演进。
链接: https://arxiv.org/abs/2607.20767
作者: Michael Xu,Jorge Leandro,Sudha Rao,Weijia Xu,Nebojsa Jojic,Gabriel DesGarennes,Chris Quirk,Bill Dolan
机构: Microsoft Research (微软研究院)
类目: Computation and Language (cs.CL)
备注:
Abstract:We introduce Rushes, a dataset and benchmark for studying revealed human engagement preferences in interactive narrative environments. Rushes is collected through a game interface where users interact with AI-generated branching narratives and select one choice from a small, explicit candidate set at each decision point. Each interaction logs the full candidate set, the user’s choice, and the evolving narrative context, yielding time-ordered trajectories with persistent user-level identifiers. Rushes contains 44,226 decision events from 8,167 unique users across six games, capturing sequential, personalized engagement behavior rather than static judgments. We show that user choices exhibit structured, non-random patterns, quantified by a low choice entropy relative to a uniform baseline. We position Rushes as a diagnostic benchmark for pluralistic alignment and demonstrate a robust Engagement Gap: state-of-the-art LLMs, including GPT-5, fail to outperform simple baselines. While classical Matrix Factorization (SVD) captures measurable personalized signal (37.7%), frontier LLMs (34.23%) struggle to even match the Popularity Baseline (36.4%) on event-level choice prediction. This gap suggests that single, population-level objectives, like those used in modern RLHF, appear insufficient to capture heterogeneous, context-dependent engagement signals. As a result, even highly capable models default to majority preferences rather than adapting to individual trajectories. We release Rushes to support research into pluralistic alignment and sequential decision-making in generative systems. The full code for the platform and dataset will be available here: this https URL
[NLP-47] GaugeQuant: Online Learning of Quantization-Optimal Bases from LLM Symmetries
【速读】： 该论文旨在解决大模型在低比特量化（如W4A4、W4A16）过程中因内部连续对称性（internal continuous symmetries）导致的激活值分布异常（activation outliers）问题，进而影响模型推理性能。其核心挑战在于：传统量化方法在保持模型精度的同时难以有效抑制由对称性引发的极端激活值，而现有解决方案往往依赖冻结模型和校准数据集，增加了部署复杂性。为此，GaugeQuant提出一种基于训练阶段的量化优化方法，其关键创新在于在损失函数中引入LogSumExp项以打破模型内部的连续对称性，从而主动选择一个能最小化激活异常的参数基底；同时通过停止梯度（stop-gradient）操作约束仅旋转矩阵参与更新，确保语言建模目标函数不受干扰。该方法无需额外校准数据、无需量化仿真，且训练开销极小。实验表明，在LLaMA-2 7B模型上采用分组大小为128的W4A4量化时，困惑度从8.22降至6.73，达到与需冻结模型和校准数据的后训练方法相当的性能；在W4A16下，困惑度由11.16降至5.45，显著提升量化模型质量。
链接: https://arxiv.org/abs/2607.20757
作者: Miguel P. Bento,João Seabra
机构: University of Lisbon (里斯本大学); Instituto Superior Técnico (高等技术学院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Transformers are known to have internal continuous symmetries that leave outputs invariant, while modifying quantization. GaugeQuant leverages this in-training by introducing a LogSumExp term to the loss that breaks the symmetries, thus selecting a basis that minimizes activation outliers. A stop-gradient operator ensures that only rotation matrices are updated, yielding the language modeling objective completely unaltered. Our requires no specific calibration data, no quantization simulation, and adds negligible training overhead. With the LLaMA-2 7B model under W4A4 quantization with group size 128, perplexity drops from 8.22 to 6.73, competing with post-training methods that require frozen models and calibration datasets. Under W4A16, perplexity drops from 11.16 to 5.45. Code is available at this https URL.
[NLP-48] REGARD: Regional Affective Differences in Large Language Models
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在不同语言与区域生态系统中对后苏联地区政治、文化及地缘政治实体存在差异性情感框架（affective framing）的问题，传统基于情感极性（如正负向评分）的评估方法难以捕捉这种复杂的情感差异。其解决方案的关键在于引入目标导向的效价-唤醒-支配（Valence-Arousal-Dominance, VAD）分析框架，通过多维度情感建模揭示模型响应中的情绪强度与态度特征。研究对19个模型在500个区域性目标上的输出进行双独立大模型判别（GPT-4o-mini 和 Qwen3.6-35B-A3B），并结合人工标注数据验证测量可靠性。聚类分析发现，尽管模型来源、架构和参数量各异，仍可识别出三个具有显著行为差异的集群，其中通用应答率与低唤醒度显著相关（r = -0.81），表明模型通过模板化回应回避评价性提问的行为模式在情感强度上具有一致性。结果表明，VAD分析能有效捕捉传统情感评估所忽略的情绪强度维度，为理解跨文化语境下模型态度差异提供了更精细的分析工具。
链接: https://arxiv.org/abs/2607.20722
作者: Andrei Chetvergov,Alexander Evseev,Mikhail Solovev,Timofei Sivoraksha,Stepan Ukolov,Valeriia Kuschenko,Maria Chistyakova,Sergey Bolovtsov
机构: 未知
类目: Computation and Language (cs.CL)
备注: 17 pages, 11 figures, 3 tables. Includes evaluation of 19 language models, two independent VAD judges, and human validation on 300 items
Abstract:Large language models trained and aligned within different linguistic and regional ecosystems may frame the same political, cultural, and geopolitical entities in different ways. Such differences are often evaluated through sentiment, favorability, or stance, reducing model attitudes to a single positive-negative axis. We introduce REGARD, a study of what drives affective framing differences across LLMs on post-Soviet entities using target-directed Valence-Arousal-Dominance profiling. We query 19 models on 500 region-specific targets, score their responses with two independent LLM judges, GPT-4o-mini and Qwen3.6-35B-A3B, and validate the measurements on a 300-item human-annotated subset. Post-hoc Ward-linkage clustering of all 19 models by affective and response-behavior profiles yields three behavioral clusters that cut across model origin, family, and parameter count. Generic-answer rate is strongly associated with lower arousal (r = -0.81) and with cluster placement: models that deflect evaluative prompts with templated responses cluster together at low arousal regardless of origin. These findings show that VAD profiling captures emotional intensity, a dimension of affective framing that is largely invisible to conventional sentiment-based evaluation.
[NLP-49] NVIDIA-labs OO Agents : Native Python Object-Oriented Agents
【速读】： 该论文旨在解决传统智能体（AI agent）开发中因分散在提示模板、工具模式、回调代码和工作流图之间而导致的复杂性与低可维护性问题。其核心挑战在于缺乏统一、可编程且易于测试与迭代的开发范式，使得智能体行为难以调试和优化。为此，论文提出NVIDIA面向对象智能体（NOOA），一种模型无关的Python框架，其关键创新在于将智能体定义为纯粹的Python对象：方法代表可执行动作，字段表示状态，文档字符串作为提示（prompt），类型注解则构成契约。其中，方法体仅由“…”组成的占位符会在运行时由大语言模型（LLM）驱动的代理循环动态填充，而具有具体实现的方法仍保持确定性执行。这一设计实现了开发者与智能体共享同一编程接口，从而支持对智能体行为进行测试、追踪、重构与持续改进，如同常规软件开发一般。该方案的关键在于通过原生Python语法抽象，将上下文管理、事件处理、长期记忆、状态渲染及经过验证的LLM调用循环等智能体能力以简洁的Pythonic API暴露出来，同时首次在单一接口上集成六项关键特性：类型化输入/输出、基于引用传递的实时对象交互、代码即动作、可编程的循环工程、显式对象状态管理，以及可用于上下文和事件的模型可调用封装接口。实证表明，当前主流模型能有效利用该接口，在特定能力测试及如SWE-bench Verified、Terminal-Bench 2.0和ARC-AGI-3等代理与推理基准任务中展现出良好性能。
链接: https://arxiv.org/abs/2607.20709
作者: Paul Furgale,Severin Klingler,James Nolan,Matt Staats,Gaia Di Lorenzo,Elisa Martinez Abad,Christian Schüller,Razvan Dinu,Alessio Devoto,Pascal Berard,Gal Kaplun,Elad Sarafian,Riccardo Roveri,Leon Derczynski,Ricardo Silveira Cabral
机构: NVIDIA NeMo Labs (NVIDIA NeMo 实验室)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Traditional agent development is split across prompt templates, tool schemas, callback code, and workflow graphs. We present NVIDIA Object-Oriented Agents (NOOA), a model-agnostic Python framework for building reliable AI agents. NOOA takes a simpler approach: an agent is a Python object. Its methods are the actions the model can take, fields are its state, docstrings are its prompts, and its type annotations are contracts. A method whose code body consists of “…” is completed at runtime by an LLM-driven agent loop, while methods with normal bodies remain standard deterministic Python. This gives developers and agents the same interface, so agent behavior can be tested, traced, refactored, and improved just like other software. This paper makes three contributions. (1) We present the agent-as-a-Python-object programming model and the design principles behind it. Where Python has existing abstractions, we adopt them directly. Agent-specific capabilities–context, events, state rendering, long-term memory, and validated LLM loops–are exposed through simple Pythonic APIs, so both developers and agents share one familiar programming model. (2) We identify six model-facing ideas that NOOA is, to our knowledge, the first to combine on a single surface: typed input/output, pass-by-reference over live objects, code as action, programmable loop engineering, explicit object state, and model-callable harness APIs for context and events. We find the community already converging on several of these ideas–often as experimental or partial features–and present the comparison to encourage further adoption. (3) We demonstrate that current models use this interface effectively, both in targeted capability tests and on agentic and reasoning benchmarks such as SWE-bench Verified and Terminal-Bench 2.0 and ARC-AGI-3. Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL) Cite as: arXiv:2607.20709 [cs.AI] (or arXiv:2607.20709v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2607.20709 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-50] Learning to Detect UI Principle Violations via Reinforcement Learning
【速读】： 该论文旨在解决生成式前端代码在功能性正确之外，仍可能违反界面质量原则（如无障碍性障碍、欺骗性设计模式、视觉层次混乱及决策复杂度过高）的问题。现有评估方法在成本、覆盖范围与可扩展性之间存在权衡：人工专家评审虽全面但效率低，前沿视觉-语言模型推理能力强但部署成本高，而基于规则的工具（如axe-core和Lighthouse）仅能检测机械可验证的无障碍问题。为此，论文提出一种轻量级视觉-语言模型作为生成界面的高效评判者，其关键在于整合来自人机交互（HCI）三大知识源的19项界面质量原则——包括WCAG 2.2无障碍标准、欺骗性设计分类体系以及感知、认知与交互理论，并通过合成注入已知违规的方式构建约1万条经验证的生成网页数据集。利用持续强化学习对40亿参数的视觉-语言模型进行训练，使微平均F1得分从36%提升至84%，其中13项原则的F1超过80%。该模型可实现生成界面的自动审计、低质量训练数据过滤及面向设计意识的代码生成奖励信号提供，显著推动了可扩展的界面质量评估发展。研究同时公开数据生成流程与注入/验证提示模板，以支持可复现评估与后续研究。
链接: https://arxiv.org/abs/2607.20690
作者: Nishi Mehta,Swathi Alse,Himani Kumavat,Yue Yu,Pratik Jayarao
机构: University of California, Santa Cruz(加州大学圣克鲁兹分校); Carnegie Mellon University(卡内基梅隆大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:Small language models and coding agents increasingly generate web front-end code, yet their outputs are typically evaluated primarily for functional correctness. A generated interface may compile, render, and pass unit tests while still violating established interface quality principles, including accessibility barriers, deceptive design patterns, poor visual hierarchy, and excessive decision complexity. Existing auditing approaches face a trade-off between cost, coverage, and scalability: expert human review provides rich judgment but is slow and expensive; frontier vision-language models offer broader reasoning capabilities but remain costly to deploy at scale; and rule-based tools such as axe-core and Lighthouse are inexpensive but primarily capture mechanically checkable accessibility issues. We investigate whether a lightweight vision-language model can serve as an effective critic for generated interfaces. We unify 19 interface-quality principles from three complementary sources of HCI knowledge: WCAG 2.2 accessibility standards, deceptive design taxonomies, and established theories of perception, cognition, and interaction. To train this critic, we construct a verified dataset of approximately 10,000 generated web pages by synthetically injecting known violations into clean, LLM-generated Tailwind pages. Continued reinforcement learning on a 4B vision-language model improves micro-F1 from 36% to 84%, with 13 of 19 principles exceeding 80% F1. The resulting critic can audit generated interfaces, filter low-quality interface training data, and provide a reward signal for design-aware code generation. We release our data-generation recipe and injection/verification prompts to support reproducible evaluation and future work on scalable interface-quality assessment. Subjects: Computation and Language (cs.CL) Cite as: arXiv:2607.20690 [cs.CL] (or arXiv:2607.20690v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2607.20690 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[NLP-51] From Agent Failures to Text Policies: What Works and What Breaks
【速读】： 该论文旨在解决在强化学习代理（agent）中应用文本梯度（TextGrad）时的核心难题：由于反馈仅在一系列动作执行后才到达，导致难以准确识别哪一决策导致了失败，从而阻碍了有效策略的优化。其解决方案的关键在于区分“遵循有效策略”与“从经验中学习策略”这两项能力。研究发现，二者之间存在显著差距——人类编写的策略文本可使两个冻结的7B参数规模代理在TextWorldExpress任务上成功率达到5.0分的提升，证明了有效策略文本的存在性；然而，基于代理轨迹生成的策略在引入更丰富的状态追踪、反事实证据或迭代生成式政策搜索（GEPA）后，仍无法稳定超越固定提示（fixed prompting）。因此，代理层面TextGrad的主要挑战并非文本策略的执行，而在于如何从复杂的经验数据中可靠地生成并筛选出有效的策略文本。
链接: https://arxiv.org/abs/2607.20668
作者: Jaideep Ray,Ankit Goyal
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:TextGrad improves language-model systems by revising text from feedback. Its core thesis is that natural-language feedback can act as a gradient for optimizing text components without changing model weights. Applying it to agents is harder because feedback arrives only after a sequence of actions, making it difficult to identify which decision caused failure. We study this problem by separating the ability to follow a useful policy from the ability to learn that policy from experience. Our main finding is a clear gap between these two abilities. Human-written policies improve two frozen 7B agents on TextWorldExpress by 5.0 success points, showing that useful policy text exists. However, policies generated from agent trajectories do not reliably outperform fixed prompting, even with richer traces, counterfactual evidence, or iterative GEPA search. The main challenge for agent-level TextGrad is therefore not executing textual policy updates, but reliably generating and selecting them from experience.
[NLP-52] Frontier Financial Judgement: Can agents tell what might move a stock?
【速读】： 该论文旨在解决生成式 AI 在金融领域中对新信息的实时识别与价值判断能力不足的问题，特别是针对股票研究中快速甄别具有估值影响的实质性新信息这一核心挑战。随着人工智能生成内容数量激增，传统新闻流过滤机制面临信息过载与误判风险，亟需具备专业级财务判断能力的智能代理（Agent）来辅助决策。其解决方案的关键在于构建一个名为“前沿金融判断”（Frontier Financial Judgement）的综合性基准测试体系，该体系通过融合人工设计并标注的合成文章、实时新闻及历史文档，共形成656个评估样本，以模拟真实世界中的金融信息环境。该任务要求模型在高压力、低噪声干扰下区分真正具有估值意义的新信息与陈旧、无关或误导性内容，从而全面评估代理在准确性、误报率（false-positive rate）、成本与可靠性之间的权衡。实验结果表明，当前最强的代理仅能在52.4%的情况下完全匹配专家标签，且各代理间误报率差异显著（从约1%至32%），揭示了现有系统在可靠部署新闻流过滤功能时仍存在重大瓶颈。
链接: https://arxiv.org/abs/2607.20645
作者: Joshua Harris
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 19 pages, 7 figures, 5 tables
Abstract:We introduce Frontier Financial Judgement, a challenging new benchmark developed in collaboration with professional equity analysts to assess agents’ ability to replicate expert human judgements. Rapidly identifying new information, evaluating its implications and determining its valuation impact is one of the most time-consuming and challenging aspects of real-world equity coverage. This is becoming ever more difficult and important as AI rapidly increases the quantity of new information to process. The strongest agent we evaluate on Frontier Financial Judgement matches all expert labels in only 52.4% of cases. We also find significant divergence in estimated false-positive rates among frontier agents, ranging from ~1% for GPT-5.6 Sol to ~32% for Claude Sonnet 4.6. To construct the benchmark and make it representative of real-world settings, we combine human-designed and labelled synthetic articles with live news articles and historical documents, creating 656 items for assessment. The resulting task requires agents to distinguish genuinely new, valuation-relevant financial information from stale, immaterial or misleading news under realistic conditions. We find substantial trade-offs among agent accuracy, cost, false positives and reliability that continue to hinder the reliable deployment of news-flow filtering in practice.
[NLP-53] WaveformQA: Benchmarking LLM Temporal Reasoning on Digital Waveforms
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在数字波形数据上进行时序推理能力不足的问题。尽管时序推理在硬件设计验证中至关重要，但现有基准测试主要聚焦于硬件描述语言（HDL）代码生成，且仅将波形作为辅助上下文使用，未能有效评估模型对波形时序逻辑的理解能力。为此，本文提出WaveformQA，一个开源的问答基准，用于评估LLM在数字波形上的时序推理性能。其关键创新在于：构建了包含360个问题、涵盖八类不同难度的程序化生成真值数据集，覆盖多信号相关性与事件顺序等复杂推理任务；波形基于开源设计实现生成，确保可复现性并贴近真实硬件行为；实验表明，尽管前沿模型在简单查询上表现尚可，但在复杂时序和多步推理任务中受限于上下文窗口长度与推理能力而性能下降；进一步发现，采用事件-时间JSON格式表示波形相较于标准的值变化转储（VCD）格式能显著提升模型推理准确率。该开放框架支持新问题类别扩展与新波形源导入，为研究人员快速原型化时序推理实验提供了有力工具。
链接: https://arxiv.org/abs/2607.20638
作者: Yichuan Liu,Daniel Cummings,Nick Vadlamudi
机构: 未知
类目: Artificial Intelligence (cs.AI); Hardware Architecture (cs.AR); Computation and Language (cs.CL)
备注: 10 pages; abridged version published in IEEE International Conference on LLM-Aided Design (ICLAD), 2026
Abstract:Large Language Models (LLMs) have demonstrated strong capabilities in code generation and reasoning, yet their ability to perform temporal reasoning over digital waveform data remains largely unexplored. Although reasoning over digital waveforms is a critical bottleneck in design verification, existing benchmarks primarily evaluate hardware description language (HDL) code generation and use waveforms only as supplementary context. This paper presents WaveformQA, an open-source question-answering benchmark for evaluating LLM temporal reasoning over digital waveforms. The benchmark comprises 360 questions with programmatically generated ground truths across eight categories of varying difficulty, including questions targeting multi-signal correlation and event ordering. Waveforms are generated from open-source design implementations, ensuring reproducibility and grounding the benchmark in real hardware behavior. Evaluation of frontier LLMs reveals that while models achieve reasonable accuracy on simple queries, performance degrades due to context window limitations and reasoning difficulties on complex temporal and multi-step questions. In addition, we show that an event-time JSON representation of waveforms improves LLM reasoning accuracy versus the standardized value change dump (VCD) format. The open-source framework supports extending to new question categories and importing new waveform sources, enabling researchers to rapidly prototype temporal reasoning experiments.
[NLP-54] Demonstrating GenDB: Instance-Optimized and Customized Query Processing Code Generation via LLM Agents VLDB2026
【速读】： 该论文旨在解决传统查询处理引擎在支持新技术与用户需求时面临的可扩展性难题，即由于系统内部结构复杂，难以进行有效扩展，而构建全新系统又需投入巨大工程成本。其核心解决方案是提出GenDB——一种基于大语言模型（Large Language Models, LLMs）驱动的生成式查询引擎（Generative Query Engine），将查询处理从手动设计的系统转向由LLM自动生成优化的查询执行代码。其关键在于：利用LLM代理（LLM agents）根据具体数据分布、工作负载特征及硬件资源，生成针对实例优化的查询执行代码；对于重复性高、模板化的查询，采用离线代码生成模式，通过前期生成成本分摊实现高效执行，并借助大规模模糊测试与人工审查保障正确性；而对于即席查询，则采用混合架构，与传统数据库管理系统（DBMS）协同工作，仅对频繁出现的SQL模板进行加速。该方案不仅显著提升了查询性能，在TPC-H及新构建的防数据泄露基准测试中均优于现有先进引擎，还支持用户交互式探索代码生成过程、分析性能优势并自定义数据与查询以评估不同LLM和查询模式下的表现。
链接: https://arxiv.org/abs/2607.20630
作者: Jiale Lao,Immanuel Trummer
机构: 未知
类目: Databases (cs.DB); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Accepted by VLDB 2026 (Demo)
Abstract:Traditional query processing engines require continuous development and extensions to support new techniques and user requirements, and in some cases, entirely new systems must be built from scratch. However, these engines are difficult to extend due to their internal complexity, and building new systems demands significant engineering effort and cost. To address this, we demonstrate GenDB, a generative query engine that shifts query processing from manually engineered systems to query processing code generation driven by Large Language Models (LLMs). An early prototype of GenDB uses LLM agents to generate instance-optimized query execution code tailored to specific data, workloads, and hardware resources. This prototype suits offline code generation for repetitive, templated queries, since the upfront generation cost amortizes over many executions and correctness can be ensured through extensive fuzz testing and manual inspection. For ad-hoc queries, GenDB can work with a traditional DBMS in a hybrid architecture: the DBMS handles one-off queries, while GenDB speeds up frequent SQL templates. Our demonstration allows users to (1) visually and interactively explore how GenDB analyzes workloads, profiles hardware resources and underlying data, produces query plans, generates code based on them, and finally uses an optimizer to iteratively achieve a correct and efficient implementation; (2) use visual inspection and analysis to gain qualitative insights into why GenDB produces code that achieves significantly better performance than state-of-the-art query engines on two benchmarks: TPC-H and a newly constructed benchmark designed to reduce potential data leakage from LLM training data; and (3) upload their own data and queries to explore GenDB with different LLMs and query patterns.
[NLP-55] Are Single-Token Sparse Autoencoder Features Causally Necessary? Layer-Depth and SAE-Family Effects
【速读】： 该论文旨在解决生成式人工智能（Generative AI）中稀疏自编码器（Sparse Autoencoder, SAE）特征的因果作用是否在不同SAE家族间保持稳定的问题。尽管SAE特征被广泛用于解释和引导大语言模型的行为，但其因果有效性在跨模型、跨训练方法的场景下是否一致仍缺乏实证检验。研究聚焦于单令牌特征（single-token features），即仅对词汇表中某一特定词元激活的特征，因其具备可直接验证的“真实标签”（ground truth），成为诊断性分析的理想案例。通过对六种模型与三种SAE家族共390万特征进行全层深度的零消融（zero-ablation）分析，发现单令牌特征在解码器空间中聚类紧密度提升4.7倍，并集中于早期层（GPT2-Small中为第0层；Gemma中为第0至第4层）。在208个全层条件下，178个条件中消融这些特征导致显著的对数概率（logit）下降（经Benjamini-Hochberg校正），且层级位置决定了损伤是沿下游传播还是直接影响输出。关键发现在于：跨家族的因果效应差异远超同家族内的规模效应——同一基础模型上，GemmaScope与BatchTopK特征保持稳定的因果锚定，而LlamaScope特征则表现出局部冗余性。此外，在相同消融操作后，目标词元的排名可恢复至基线水平的96%-98%，但同一模型内激活函数对比却出现符号反转，表明训练方案（training recipe）是导致跨家族可解释性主张差异的残余决定因素。因此，该研究揭示了当前基于SAE的可解释性结论高度依赖于训练方法，而非仅由激活函数或模型规模决定。
链接: https://arxiv.org/abs/2607.20596
作者: Seonglae Cho,Zekun Wu,Kleyton Da Costa,Rishi Kalra,Ilham Wicaksono,Adriano Koshiyama
机构: Holistic AI(霍利斯蒂克人工智能); University College London(伦敦大学学院)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:Sparse autoencoder (SAE) features are used to interpret and steer large language models, yet whether a feature’s causal role is stable across SAE families remains untested. Single-token features that activate on one vocabulary item provide the diagnostic case where ground truth permits direct comparison. We analyze 3.9M features across six models and three SAE families using zero-ablation at full layer depth. Single-token features cluster 4.7x tighter in decoder space and concentrate in early layers (Layer 0 in GPT2-Small; L0-L4 in Gemma). Ablating them yields Benjamini-Hochberg-significant logit reductions in 178 of 208 full-layer conditions, with depth controlling whether damage cascades downstream or shapes the output directly. Cross-family causal differences exceed within-family scale effects: on the same base model, GemmaScope and BatchTopK features remain causally anchored, while LlamaScope features are locally redundant. The target token’s rank recovers to within 2x baseline 96-98% of the time after the same ablation, and a controlled activation-function comparison reverses sign within the same model, leaving training recipe as the residual candidate. Cross-family interpretability claims are therefore sensitive to training methodology, not just activation function or scale.
[NLP-56] Evaluating the Effectiveness of Persona Simulation in Opinion Prediction with GPT -4.1 ICDM2025
【速读】： 该论文旨在解决如何利用大语言模型（LLM）进行人物角色模拟（persona simulation），以准确预测个体或群体在特定情境下的观点与行为选择，尤其是在政治选举和公共卫生议题中的意见预测。其核心挑战在于模型能否真实反映不同背景人群的立场差异，并在复杂社会语境中保持一致性。解决方案的关键在于通过高质量、具代表性的角色数据集（如哥伦比亚大学提供的九个美国州级人物画像数据集）与权威调查数据（如皮尤研究中心的美国趋势面板数据）对GPT-4.1进行测试与验证，使其能够基于人物特征信息精准预测2024年选举结果（在八州中准确）及儿童疫苗接种态度（准确率达0.94）。此外，研究还展示了模型生成的跨角色对话在内容上符合人物设定与背景，尽管在自然流畅性方面仍显不足。因此，该研究证明了人物角色模拟在意见分析与反应预测中的潜力，但强调需系统性识别并缓解潜在偏见问题，以推动其在公共健康、政策制定乃至经济学等多领域的应用落地。
链接: https://arxiv.org/abs/2607.20589
作者: Sarah Y. Li,Ziyu Yao
机构: 未知
类目: Computation and Language (cs.CL)
备注: ICDM 2025 Undergraduate and High School Symposium
Abstract:Persona simulation involves utilizing large language models (LLMs) to anticipate human choices or interactions based on specific characteristic information. To further understand current limitations and future directions, we tested persona simulation in opinion prediction with GPT-4.1 (knowledge cutoff by June 2024). Using personas from nine U.S. states provided by Columbia University’s Personas dataset, GPT-4.1 accurately predicted 2024 election outcomes in eight out of the nine states, only failing in one of the swing states. We then focused on opinions related to medicine and healthcare. With the American Trends Panel Wave 123 dataset from Pew Research Center, GPT-4.1 was able to anticipate beliefs about childhood vaccines with an accuracy of up to 0.94. Furthermore, we applied GPT-4.1 to generate conversations among personas and observed that the simulated dialogues and opinions adhered well to personas’ personalities and backgrounds, albeit lacking natural human-like flow. Persona simulation proves to be a promising application of artificial intelligence as long as biases are addressed. In the near future, it will be beneficial to apply it to opinion analysis and reaction prediction in diverse fields ranging from public health to lawmaking to economics.
[NLP-57] CMI-Mem: Toward Generalizable Long-Term Memory Management via CMI-Augmented Reinforcement Learning
【速读】： 该论文旨在解决现有记忆管理模型（Memory Manager models）在评估记忆价值时过度依赖大语言模型（LLM）生成的合成问答对（QA pairs）所带来的偏差问题，这种依赖使得记忆评估结果受采样查询和下游阅读器性能的影响较大，缺乏对记忆内容本身信息量的客观衡量。为克服这一局限，论文提出了一种基于强化学习（Reinforcement Learning, RL）的轻量级记忆管理模型——CMI-Mem，其核心创新在于设计了一种混合奖励机制，该机制结合了下游任务的问答正确率与内在的条件互信息（Conditional Mutual Information, CMI）。CMI能够不依赖于特定采样问答对，直接评估新对话输入相对于当前记忆状态所贡献的信息量，从而实现对记忆内容价值的独立、更本质的量化。这一机制有效补充而非替代传统的问答对监督信号，显著提升了记忆管理的鲁棒性与泛化能力。
链接: https://arxiv.org/abs/2607.20553
作者: Yubo Wang,Qiuyu Zhao,Zenghui Sun,Shichao Dong,Jinsong Lan,Xiaoyong Zhu,Haoyang Li,Bo Zheng,Lei Chen
机构: Alibaba Group(阿里巴巴); HKUST(香港科技大学); PolyU(香港理工大学); HKUST(GZ)(香港科技大学（广州)）
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Memory Manager models are pivotal in agent systems. Existing methods rely predominantly on LLM-judged synthetic question-answer (QA) pairs, making memory valuation dependent on sampled queries and the downstream reader. To address this limitation, we propose \textbfCMI-Mem, a reinforcement learning(RL)-based lightweight memory manager model with a hybrid reward that combines downstream QA correctness and intrinsic Conditional Mutual Information (CMI). CMI evaluates the information contributed by new conversational inputs relative to the current memory state without conditioning on a sampled QA query, thereby complementing rather than replacing QA grounding. Our codes are available at: this https URL , and the CMI-Mem-4B model checkpoint is available at: this https URL
[NLP-58] co-GAIA: Bilingual Benchmark for Agents in Telecom Domain
【速读】： 该论文旨在解决现有评估基准在真实企业场景下对工具使用型智能体（tool-using agents）能力测试的不足问题，尤其针对通信行业复杂多源异构数据环境中的多跳推理与跨模态理解挑战。其核心问题是：如何构建一个具备真实业务背景、涵盖多语言（英/阿）、多模态（文本、图像、表格）且可重复验证的评估体系，以准确衡量智能体在实际应用中处理复杂任务的能力。解决方案的关键在于提出Telco-GAIA——一个基于真实电信运营商数据的双语、多模态基准，包含100个经人工验证的问题回答任务，每个任务平均需4.2跳推理，涉及静态网页快照（HTML、图像、PDF）、合成关系型SQL数据库及外部网络存档三大异构数据源，覆盖文本、图像与表格三种模态。该基准通过沙盒化Docker环境部署，并采用归一化精确字符串匹配进行评分，实现了评估过程的客观性、确定性和时间可复现性，无需依赖大模型作为评判者（LLM-as-a-Judge）。实验表明，即使最强模型仅能完成71%的任务，在有限成本预算下下降至约40%，而视觉感知类任务平均得分低于30%，揭示了文档与图像理解仍存在巨大提升空间，验证了该基准在推动企业级智能体发展方面的有效性与挑战性。
链接: https://arxiv.org/abs/2607.20510
作者: Dmitrii Khizbullin,Zaid Alyafeai,Abdelrahman Eldesokey,Nourah AlSultan,Raghad Alshalan,David R. Pugh,Bernard Ghanem
机构: King Abdullah University of Science and Technology (KAUST); stc
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:We introduce Telco-GAIA, a bilingual, multi-modal benchmark for evaluating tool-using agents on the data of a real-world telecommunications operator. Telco-GAIA comprises 100 human-verified question-answering tasks, in English and Arabic, that each demand multi-hop reasoning (4.2 hops on average) over three heterogeneous sources: a static website snapshot (HTML, images, and linked PDFs), a synthetic relational SQL database, and external web archives, spanning text, image, and tabular modalities. The benchmark is delivered as a sandboxed Docker environment and scored by normalized exact string matching, making evaluation objective, deterministic, and reproducible over time without any LLM-as-a-Judge. Evaluating a purpose-built reference agent across twelve commercial and open LLMs, we find Telco-GAIA challenging: even the strongest model solves only 71% of tasks; under a moderate cost budget, this falls to about 40%, and the visually grounded categories remain the weakest, where the average backend scores below 30%, leaving substantial headroom in document and image understanding. Telco-GAIA offers a rigorous, reproducible testbed for enterprise agents and a template for constructing closed-domain benchmarks.
[NLP-59] PhantomFill: When the Form Demands an Answer Language Models Invent One
【速读】： 该论文旨在解决生成式 AI 在实际应用中因输出格式强制要求而引发的幻觉（hallucination）问题。具体而言，当模型被要求以特定结构化格式（如 JSON 字段、枚举值或最小数量数组）输出结果时，即使输入信息不足以支持有效回答，模型仍会强行生成看似合理但完全虚构的内容。研究发现，在无格式约束的自由文本回答中，GPT-5.5 能够在 98% 的情况下诚实承认数据缺失；然而，一旦被要求填写包含情感分析字段的 JSON 模板，同一模型在 40 次测试中全部虚构了不存在的用户情绪和对话内容。这一现象在十三个模型中普遍存在，强制格式导致幻觉率高达 100% 的情况出现在十个模型中。关键解决方案在于引入一种简单的模式：在数据结构中加入显式的“证据不足”选项，并通过一行为核心的 Schema 优化来缓解问题。实验表明，尽管该机制对前沿开放权重模型有一定效果，但多数模型仍无视此选项，说明当前训练范式未能有效培养模型在格式压力下的诚实响应能力。因此，论文提出 PhantomFill 基准测试框架，量化评估“受迫幻觉率”（Coerced Fabrication Rate）与“逃逸利用率”（Escape Utilization Rate），揭示了结构化输出设计对模型行为的根本性影响。
链接: https://arxiv.org/abs/2607.20492
作者: Rana Muhammad Usman
机构: 独立研究员（Independent Researcher）
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 12 pages, 6 figures. Benchmark and code: this https URL
Abstract:Language models in production do not write prose. They fill forms: JSON fields, function arguments, extraction templates. We show that the form itself causes hallucination. We ask thirteen models the same question about the same input and change only the answer format. The inputs are built so the question cannot be answered: a viral post showing 12,400 likes but no visible replies, a support ticket whose call was never transcribed. In free text, GPT-5.5 answers honestly. It says there is no reply data, 98% of the time. Given a required JSON field for sentiment, the same model invents an answer 40 times out of 40. It fabricates the mood of crowds it never saw and quotes customers it never heard. The pattern holds with force. Required fields drive fabrication to 100% in ten of thirteen models. An explicit “insufficient evidence” option rescues only the frontier: all nine open-weight models ignore it. A direct instruction, do not infer sentiment, is overridden by the schema in four of six models. Resistance does not come with scale: within a single model family, the smallest model refuses, the mid-sized model fabricates, the largest refuses again. Honesty under format pressure is a training outcome that no one is measuring. The fabrication hides exactly where hedging is impossible: in required enums and minimum-count arrays, fields where no disclaimer fits. We release PhantomFill, a benchmark with deterministic scoring and two reportable numbers: the Coerced Fabrication Rate and the Escape Utilization Rate. The fix we test is one line of schema. The failure we measure is everywhere. Comments: 12 pages, 6 figures. Benchmark and code: this https URL Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) ACMclasses: I.2.7; I.2.6 Cite as: arXiv:2607.20492 [cs.LG] (or arXiv:2607.20492v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2607.20492 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Rana Usman Mr [view email] [v1] Thu, 11 Jun 2026 12:15:05 UTC (689 KB)
[NLP-60] DFAH-Bench: Benchmarking Observable Agent Instability in Financial Decision-Making
【速读】： 该论文旨在解决现有评估基准仅关注工具型智能体最终决策结果，而忽视其决策过程一致性的核心问题。传统评估方法依赖于输出结果的一致性（outcome agreement），但无法揭示智能体在不同运行中是否通过相似的推理路径达成相同结论，从而导致对模型行为稳定性的误判。为此，论文提出DFAH-Bench这一可复现性评估基准，通过三个无需访问隐藏推理文本的可观测维度——工具调用轨迹（tool-call trajectories）、证据接触模式（evidence contacts）以及决策集中度（decision concentration）——来量化金融任务中智能体行为的可观测不稳定性。关键发现表明，即使前沿模型在决策上达成95%的一致性，其工具使用路径的一致性仅为77%，存在高达18个百分点的差距（95%置信区间：[0.14, 0.22]），而传统仅基于结果的评估完全无法捕捉此差异。进一步分析识别出三类典型行为模式：模式匹配者（pattern matchers）通过固定输出忽略输入差异实现高一致性；稳定执行者（stable executors）保持相对一致的工具使用流程；轨迹偏离者（trajectory divergers）则以显著不同的工具路径和证据接触方式达成相同结论。该研究强调了引入多维可观测稳定性评估的重要性，并公开了完整的基准代码、度量脚本、回放日志及数据说明文档，推动更全面的智能体行为评估范式发展。
链接: https://arxiv.org/abs/2607.20491
作者: Raffi Khatchadourian
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 16 pages, 3 figures. Code, replay logs, one-command reproduction (make reproduce-paper), and an interactive results explorer: this http URL
Abstract:Standard evaluation benchmarks measure what a tool-using agent decides, not whether it arrives at that decision through the same process each time. We introduce DFAH-Bench, a replay benchmark that measures observable behavioral instability in financial agent decision-making across three channels – tool-call trajectories, evidence contacts, and decision concentration – none of which require access to hidden reasoning text. Across 8,127 replay episodes spanning 10 models and 3 financial tasks, we find that outcome agreement alone is an incomplete stability signal: frontier models can agree on decisions 95% of the time while following the same tool path only 77% of the time – an 18-percentage-point gap (95% CI: [0.14, 0.22]) that outcome-only evaluation misses entirely. Among frontier-model case groups with high decision agreement, over 55% exhibit meaningful trajectory divergence. We identify three behavioral profiles: pattern matchers that achieve near-perfect agreement by collapsing to a single output regardless of input, stable executors with relatively consistent tool-use processes, and trajectory divergers that reach the same conclusions through materially different tool paths and evidence contacts. The benchmark code, metric scripts, replay logs, benchmark card, dataset README, and release manifest are released in the accompanying repository.
[NLP-61] Directional Hallucinations: Ideological Drift in News-Grounded LLM Question Answering
【速读】： 该论文旨在解决生成式 AI（Generative AI）在政治信息问答场景中因意识形态偏移（ideological drift）而导致的幻觉（hallucination）问题，尤其是在选举相关高风险信息环境中，事实错误与意识形态扭曲可能对公众认知产生重大影响。其核心解决方案在于构建一个可复现的测量框架，通过多步骤分析揭示幻觉的成因与倾向性：首先基于21,727篇经专家标注的美国政治新闻文章生成针对性问题，继而利用三款开源大语言模型（LLM）与一款专有模型生成文档依赖型回答；随后采用基于参考文本的对比方法检测句子级别的幻觉，结合微调后的立场分类器判断幻觉内容的意识形态倾向，并通过分析输出logits探究令牌级不确定性与幻觉及意识形态偏移之间的关联。研究发现，幻觉率在不同模型间差异显著，且集中在争议性议题上，但来源意识形态对幻觉频率的影响较小；然而，幻觉内容表现出显著的左倾偏移——多数幻觉句被判定为左翼倾向，甚至包括来自右翼来源生成的幻觉。进一步的logit层级分析表明，幻觉多出现在高熵生成语境中，且在部分模型中，不确定性可预测左倾偏移，支持“不确定性导致猜测”（uncertainty to guessing）的认知机制。该研究为审计人工智能中介的政治信息传播提供了关键工具，并为选举相关部署中的安全防护设计提供了重要依据。
链接: https://arxiv.org/abs/2607.20487
作者: Chendi Wang,Liam Cunningham,Tom Yishay,Jieying Chen
机构: Vrije Universiteit Amsterdam (阿姆斯特丹自由大学)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Large language models (LLMs) are increasingly used to answer questions about political information, including in election-adjacent information settings where factual errors and ideological distortions are high-stakes. We present a reproducible measurement framework that treats hallucinations, unsupported statements in document-grounded QA, as diagnostic signals of ideological drift. Using 21,727 expert-labeled U.S. political news articles from QBias spanning left, center, and right sources, we (i) generate an article-specific question, (ii) elicit document-grounded answers from three open-weight LLMs and one proprietary model, (iii) detect sentence-level hallucinations via reference-based comparison, (iv) classify the ideological valence of hallucinated sentences with a fine-tuned stance classifier, and (v) probe output logits to relate token-level uncertainty to hallucination and drift. Hallucination rates vary substantially across models and concentrate in contentious topics, while source-ideology differences in hallucination frequency are modest. In contrast, hallucination content exhibits robust leftward drift: a majority of hallucinated sentences are classified as left-leaning, including among hallucinations generated from right-leaning sources. Logit-level analysis shows hallucinations arise in high-entropy generation contexts, and in some models uncertainty also predicts leftward drift, consistent with an “uncertainty to guessing” mechanism. We discuss implications for auditing AI-mediated political information and for designing safeguards in election-relevant deployments.
[NLP-62] Expectation Alignment of Language Models for Real-World User Expectations ICML2026
【速读】： 该论文旨在解决当前大语言模型（Large Language Models, LLMs）在真实用户交互中与用户实际期望之间存在显著错位的问题。尽管现有模型在标准基准测试中表现优异，但其评估方法（如基于模型启发式、专家评分或用户模拟）未能充分捕捉真实人类期望的多样性与细微差异，导致模型看似“胜任”却难以真正满足用户需求。为此，论文提出了首个针对真实世界LLM交互中用户期望的系统性研究，设计了一套严谨的方法以提取语义丰富的用户期望，并构建了基于真实用户期望的ExpectBench基准。分析结果表明，当前主流模型在满足及预测用户期望方面存在明显不足，揭示了人机对齐中的根本性偏差。针对此问题，论文提出LENS（Latent Expectation-aware Network for Response Generation），一种轻量级的、感知用户隐含期望的响应生成框架。LENS的核心创新在于使模型能够内化并显式建模用户期望，从而生成更符合用户真实意图的回应，在多个指标上均显著提升期望满足度，验证了显式建模用户期望对于实现真实场景下人-机对齐的关键作用。
链接: https://arxiv.org/abs/2607.20485
作者: Miaomiao Li,Yang Wang,Bin Liang,Shudong Liu,Zhiwei Zhang,Kam-Fai Wong
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted by ICML 2026
Abstract:Large language models (LLMs) have demonstrated remarkable performance on standard benchmarks, yet it remains largely unexplored whether they truly meet user expectations. Existing evaluation approaches, relying on model heuristics, expert rubrics, or user simulation, fail to capture the diversity and subtlety of real human expectations, causing models to appear competent while misaligning with what users actually seek. We present the first systematic study of user expectations in real-world LLM interactions, proposing a principled procedure to extract semantically rich expectations and introducing ExpectBench, a benchmark grounded in real user expectations. Analyses reveal that current LLMs struggle to satisfy and anticipate what users hope to obtain, highlighting a fundamental source of misalignment. Building on these observations, we propose LENS, a lightweight latent expectation-aware response generation framework. LENS enables models to internalize user expectations and generate better-aligned responses, consistently improving expectation satisfaction and underscoring the importance of explicitly modeling user expectations for realistic human-AI alignment.
[NLP-63] PersonaTrail: Benchmarking Personalized Web Agents through Browsing Trails
【速读】： 该论文旨在解决现有基准测试无法有效评估个性化网络代理（personalized web agents）在用户指令不完整情况下的上下文推理能力的问题。当前主流方法要么仅支持完全明确的指令，要么将真实的浏览历史抽象为简化的形式，导致无法真实反映用户个性化需求的动态推断。为此，研究者提出PersonaTrail这一面向受控开放网络环境的个性化网络代理基准，其核心在于利用真实用户的浏览轨迹作为历史数据，以评估代理从过往会话中推断用户偏好和回忆信息的能力。解决方案的关键在于提出一种名为“偏好感知的上下文记忆”（Preference-Aware Contextual Memory, PACMem）的框架，该框架将原始浏览历史分解为两类结构化记忆：事实性记忆（factual memories），用于总结单个会话内容；偏好性记忆（preference memories），用于提炼用户反复出现的行为模式。在推理阶段，代理通过检索这两类记忆中最相关的条目，实现对用户偏好的精准建模与个性化导航。大量实验表明，PACMem在各类任务上均显著优于现有的基于记忆的基线方法。
链接: https://arxiv.org/abs/2607.20482
作者: Seungbin Yang,Chaewoon Ki,Dohyun Lee,Jaegul Choo,ChaeHun Park
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注:
Abstract:Recent advances in large language models have enabled web agents to autonomously execute complex tasks. In practice, users frequently provide underspecified instructions, requiring agents to infer the missing context from their raw browsing histories. Existing benchmarks fail to capture this form of personalization, as they either restrict tasks to fully explicit prompts or abstract web interaction history into simplified forms. To bridge this gap, we introduce PersonaTrail, a benchmark for personalized web agents operating in a managed open web environment. By leveraging realistic browsing trajectories as user history, PersonaTrail evaluates an agent’s ability to infer user preferences and recall information from past browsing sessions. We further propose Preference-Aware Contextual Memory (PACMem), a framework that decomposes raw browsing histories into two types of structured memory: factual memories that summarize individual sessions and preference memories that distill recurring behavioral patterns. At inference time, the agent retrieves the most relevant entries from these memories to guide personalized navigation. Extensive experiments show that PACMem consistently outperforms existing memory-based baselines on both tasks.
[NLP-64] Beyond Liars Bench: The Impact of Lie Typology Depth and Sparsity on Deception Detection in LLM s
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）生成欺骗性输出时的检测难题，尤其关注在跨领域（out-of-domain）场景下检测性能下降的问题。现有方法在特定类型谎言上训练的探测器难以泛化至其他类型的欺骗行为，导致检测效果受限。为此，研究提出系统性分析多个关键因素对检测性能的影响，包括表征深度、探测器表达能力、稀疏特征表示以及训练数据中的谎言类型（lie typology）。通过在标准基准数据集基础上引入包含虚构、遗漏和夸大等多种欺骗形式的补充数据集，研究评估了七种不同探测器类型的表现。实验结果表明：最优表征深度高度依赖于具体数据集；更具表达性的探测器仅在特定情况下优于线性基线；稀疏自编码器特征与密集隐藏状态表现相当。最终，研究强调训练数据中谎言类型的选取显著影响检测效果，揭示了欺骗检测本质上是一个高度依赖表征的选择性问题。
链接: https://arxiv.org/abs/2607.20479
作者: Amr Moustafa,Max Feser,Florian Mai
机构: University of Bonn(波恩大学); Lamarr Institute for ML and AI(波恩大学机器学习与人工智能研究所)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: Presented at the AI Transparency Conference 2026, forthcoming in the AI Transparency Journal
Abstract:Training probes to detect deceptive outputs from large language models is still an open problem. Recent work has demonstrated that detection probes fail especially in out-of-domain scenarios – training on one type of lie does not transfer well to deception scenarios involving other types of lies. In this work, we conduct a systematic study on how various factors impact detection performance: representation depth, probe expressivity, sparse feature representations, and the lie typology of the training data. To this end, we augment standard benchmark training data with a supplementary dataset containing diverse types of deception, including fabrication, omission, and exaggeration examples. Analyzing these factors across seven probe types, our experimental results show that the optimal representation depth is highly dataset-dependent, more expressive probes provide only selective gains over linear baselines, and sparse autoencoder features perform similarly to dense hidden states. Ultimately, we demonstrate that the choice of training data and lie typology substantially changes detectability, highlighting that deception detection is a highly representation-dependent problem.
[NLP-65] Benchmarking Large Language Models on Multi-Sensor Physical Hazard Assessment
【速读】： 该论文旨在解决生成式AI在多传感器物理风险数据评估中对复合性安全威胁的识别能力不足问题，尤其关注当多个传感器读数同时超出基线但未触及各自独立安全阈值时，模型是否能有效识别潜在协同风险。其解决方案的关键在于通过实证基准测试验证大语言模型在多传感器联合评估、响应比例性和模式消歧三类场景下的表现，结果表明：尽管所有五款模型（ChatGPT-4o、Gemini 2.5 Flash、DeepSeek、Kimi、Llama 3.1 8B）在单一传感器越限场景下表现出近乎完美的准确率（类别B Q1: 0.975–1.000），但在多传感器同时轻微超限的复合场景中却普遍未能生成任何预警信号（类别A Q2: 0.000–0.208；Q3: 0.000–0.592），显示出显著的系统性失效。这一发现揭示了当前生成式AI在复杂物理安全监测任务中的核心局限——缺乏对多源异构数据间非线性关联与累积效应的敏感性，直接警示相关从业者在部署此类模型于关键安全监控系统时需警惕其“虚假安全”风险。
链接: https://arxiv.org/abs/2607.20476
作者: Faizan Iqbal
机构: 未知
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
备注: 14 pages, 6 figures. Benchmark dataset, evaluation code, and raw results publicly available at: this https URL
Abstract:We present an empirical benchmark evaluating how five large language models assess multisensor physical hazard data. Testing 60 scenarios across three categories - multi-sensor joint assessment, response proportionality, and pattern disambiguation - with 1,800 API calls at temperature 0.0, we find that all tested models consistently produced no precautionary warning signal across the tested scenarios where multiple sensors are simultaneously elevated below their individual safety limits, while achieving near-perfect accuracy on single-sensor threshold violations. All five models (ChatGPT-4o, Gemini 2.5 Flash, DeepSeek, Kimi, Llama 3.1 8B) score near zero on Category A multi-sensor scenarios (Q2: 0.000-0.208; Q3: 0.000-0.592) compared to strong performance on single-sensor scenarios (Category B Q1: 0.975-1.000). Structured tabular formatting shows no consistent advantage over plain prose; ChatGPT-4o performs significantly better under prose (p = 0.001). These findings have direct implications for practitioners deploying the tested models in physical safety monitoring systems.
[NLP-66] SonicSampler: Unified Tile-Aware Kernels for LLM Sampling and Speculative Verification
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）推理过程中采样阶段的性能瓶颈问题，特别是现有实现中采样管道（sampling pipeline）存在部分加速、依赖多次核函数调用或假设批量内采样行为同质化，从而限制了对动态服务负载的支持并阻碍高效CUDA Graph执行。其解决方案的核心是提出SonicSampler，一个基于分块感知（tile-aware）Triton内核的统一采样套件，将完整的采样流水线（包括逻辑值处理、词元选择与验证）垂直融合为一个固定且工作负载感知的执行模型。该方案支持单批次内动态的每请求采样行为，涵盖语法约束解码、重复惩罚、频率与存在性惩罚、逻辑偏置、温度缩放、top-k/top-p/min-p过滤及推测验证等复杂机制，并保持完全兼容CUDA Graph。关键技术在于一种新颖的分层两级top-k算法，充分利用大语言模型输出的低熵特性，在大规模词表上实现高效词元选择，相较竞争基线最高提升10倍性能；在异构推测解码工作负载下，整体性能相较当前最优基线最高提升16倍，同时维持灵活的批处理执行能力。
链接: https://arxiv.org/abs/2607.20475
作者: Pragaash Ponnusamy,Shivam Sahni,Jue Wang,Tri Dao
机构: Together AI(Together AI)
类目: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 26 pages, 12 figures
Abstract:Sampling in LLM inference comprises a combinatorial set of logit processing, token selection, and verification operations for speculative decoding. However, existing implementations either accelerate only subsets of this pipeline, rely on multiple kernel launches, or assume homogeneous sampling behavior across a batch, limiting support for dynamic serving workloads and preventing efficient CUDA Graph execution. We present \textbfSonicSampler , a unified suite of tile-aware Triton kernels that vertically fuses the complete sampling pipeline into a fixed, workload-aware execution model. Our kernels support dynamic per-request sampling behaviors, including grammar-constrained decoding, repetition, frequency and presence penalties, logit bias, temperature scaling, top- k / top- p / min- p filtering, and speculative verification - within a single batched kernel while remaining fully CUDA Graph-compatible. Central to our approach is a novel hierarchical two-stage top- k algorithm that achieves up to \textbf10x speedup over competitive baselines and exploits the low-entropy structure of LLM outputs to enable efficient selection over large vocabularies. Across heterogeneous speculative decoding workloads, SonicSampler achieves up to \textbf16x speedup over state-of-the-art baselines while preserving flexible batched execution.
[NLP-67] DataPrep-Bench: Benchmarking LLM s as Training Data Preparators
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）训练数据准备过程中缺乏统一评估基准的问题，尤其针对由生成式 AI 驱动的数据构建与数据质量评估两大核心能力。现有方法在数据构造（data construction）和数据质量评估（data quality evaluation）方面均缺乏一致、可复现且以下游任务性能为导向的衡量标准，导致难以系统性比较不同方法的有效性。其解决方案的关键在于提出首个统一基准——DataPrep-Bench，该基准基于下游任务性能驱动的协议，在六个领域内联合评估上述两项能力。在数据构造方面，通过使用相同的原始数据源并以基础模型在生成数据上的微调表现作为评分依据，引入了技能引导型代理 Data-Construction-Skill，显著提升了金融等知识密集型领域的数据生成质量；在数据质量评估方面，提出分布对齐评分（Distributional Alignment Score, DAS），利用候选数据集与领域代理之间的最大均值差异（MMD）来衡量分布相似性，实现了在数学、科学和医学三个领域中同时达到 r > 0.70 的强相关性，优于现有的基于质量、多样性或启发式规则的评估指标。因此，DataPrep-Bench 提供了一个以下游性能为锚点的统一框架，将数据构造与质量评估视为同等重要的目标，推动了生成式 AI 驱动的数据准备流程的可度量发展。
链接: https://arxiv.org/abs/2607.20465
作者: Hao Liang,Qifeng Cai,Yibo Lin,Jianzhuo Du,Qifeng Xia,Sizhe Qiu,Linzhuang Sun,Meiyi Qiang,Zhaoyang Han,Xiaochen Ma,Bohan Zeng,Ruichuan An,Conghui He,Wentao Zhang
机构: 北京大学(University of the Chinese Academy of Sciences)
类目: Machine Learning (cs.LG); Computation and Language (cs.CL)
备注:
Abstract:The quality of training data fundamentally determines the capabilities of large language models (LLMs), yet no unified benchmark exists to measure how well LLMs, agents, and data-centric workflows actually prepare training data end to end. We view LLM-driven data preparation as comprising two complementary capabilities: data construction, which transforms raw sources into supervised training data, and data quality evaluation, which predicts the training value of candidate datasets before downstream training; throughout, “quality” refers to downstream training utility rather than surface-level textual properties. We introduce DataPrep-Bench, the first unified benchmark that jointly evaluates both capabilities under a shared downstream-grounded protocol over six domains and multiple base models. For data construction, methods consume identical raw sources and are scored by fine-tuning a base model on their outputs jointly with Dolly-15k; alongside this track we release Data-Construction-Skill, a skill-guided agent that lifts the Dolly-only baseline by nearly 20 points absolute on Llama-3.1-8B Finance and is competitive with the strongest agent- and DataFlow-based methods in knowledge-extraction-dense domains. For data quality evaluation, scoring functions are scored by Pearson correlation with downstream performance on a shared candidate pool; we release the Distributional Alignment Score (DAS), a distribution-based evaluator that uses MMD between a candidate dataset and a domain proxy. DAS attains the strongest cross-model correlation in four of six domains and is the only metric clearing r 0.70 simultaneously in Math, Science, and Medical, outperforming existing quality-, diversity-, and heuristic-based evaluators. DataPrep-Bench provides a unified, downstream-grounded framework for measuring progress on both capabilities as co-equal targets of LLM-driven data preparation.
[NLP-68] Instruct-FD: Can Your Full-Duplex Speech System Follow Turn-Taking Instructions?
【速读】： 该论文旨在解决当前全双工（Full-Duplex, FD）语音对话系统在受到显式指令控制时，能否灵活调整其交互轮次管理（turn-taking）行为的关键问题。这一能力对于真实场景中的部署至关重要，因为不同应用场景（如主动教学与被动咨询）对对话策略的要求存在显著差异。其解决方案的核心在于提出Instruct-FD——一个基于指令条件的基准评测框架，通过构建经人工验证且可扩展的合成数据生成流程，实现指令驱动的对话生成；同时引入与部署无关的多轮评估协议及基于大语言模型（LLM-based）的评判机制，以客观衡量系统对指令的遵循程度。实验结果表明，现有六种先进全双工系统在指令遵循方面的表现普遍不佳，最佳模型仅达到64.4%的指令遵循率，尤其在主动型行为（如模型回声反馈和打断）方面表现尤为薄弱。该研究揭示了指令可控的轮次管理是提升全双工对话系统适应性与可部署性的关键方向。
链接: https://arxiv.org/abs/2607.20460
作者: Yuzhi Tang,Wentao Ma,Xiling Zhao,Ahmad Salimi,Sepehr Harfi Moridani,Dongming Shen,Jixuan Wang,Abdulrahman Abdulrazzag,Murdock Aubry,Yu-Hua Chen,Daniel Lee,Jaewon Lee,Jonah Mackey,Silin Meng,Nicholas Stranges,Chenxu Xiong,Hao Yu,Yi Zhu,Mu Li,Alex Smola
机构: Boson AI, Santa Clara, CA 95054
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Current full-duplex (FD) spoken dialogue systems can produce fluid interactions, yet it remains unclear whether they can adapt their turn-taking behavior when explicitly instructed. This is critical for real-world deployment, where conversational policies vary across applications (e.g., proactive tutoring vs. passive counseling). We introduce Instruct-FD, an instruction-conditioned benchmark for evaluating controllable turn management in FD systems. To enable this, we develop a human-validated, scalable synthetic pipeline that generates instruction-conditioned conversations, along with a deployment-agnostic multi-turn evaluation protocol and an LLM-based judge. Benchmarking six state-of-the-art full-duplex systems reveals a substantial gap in instruction-following turn management: the best model achieves only 64.4% adherence. Performance is highly uneven across behaviors and scenarios, with proactive behaviors such as model backchanneling and interruption remaining particularly challenging. These findings establish instruction-following turn management as a crucial direction for building adaptable and deployable full-duplex dialogue systems.
[NLP-69] HOR: A Theta-Gamma Hierarchical Oscillatory Reasoning Framework for Multi-hop QA
【速读】： 该论文旨在解决多跳问答（multi-hop question answering）中存在的两个核心问题：注意力衰减（attention decay）与错误累积（error accumulation）。前者表现为随着推理链的延伸，模型对核心问题的关注度逐渐减弱；后者则指错误在各推理步骤间传播并不断放大，最终导致答案失效。为此，论文提出了一种受大脑Theta-Gamma节律振荡启发的分层振荡推理框架THOR，其关键在于通过分层机制解耦全局规划与局部检索过程，实现跨跳注意力的有效传递，并引入验证与修复机制，在错误路径上主动中断错误传播，从而显著提升推理的准确性与鲁棒性。实验结果表明，THOR在多个基准数据集上均表现出优越性能，且具备良好的模型泛化能力。
链接: https://arxiv.org/abs/2607.20459
作者: Ziyang Ling,Ronald X. Xu,Mingzhai Sun
机构: Suzhou Institute for Advanced Research, University of Science and Technology of China; School of Biomedical Engineering, Division of Life Sciences and Medicine, University of Science and Technology of China
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-hop question answering requires retrieving and integrating evidence from multiple contexts. Despite the rapid progress of current research, multi-hop reasoning remains constrained by two persistent limitations: attention decay, where the model’s focus on main question degrades as the reasoning chain grows, and error accumulation, where mistakes propagate across hops and compounds into final failure. Inspired by Theta-Gamma hierarchical oscillation which decouples global planning from local retrieval, enabling efficient attention transfer between hops and a verification and repair mechanism that interrupts the accumulation of errors in the wrong paths, we present THOR, a brain-inspired Theta-Gamma hierarchical oscillatory reasoning framework. Extensive comparative experiments and specific validation experiments on multi-hop QA benchmarks demonstrate that THOR improves answer accuracy and robustness while mitigating limitations, showcasing its generalization across different backbones.
[NLP-70] CAMeR: Keyword-Gated Hybrid Activation for Adaptive Memory Retention in LLM Agents
【速读】： 该论文旨在解决大语言模型（Large Language Model, LLM）代理在长期对话中因记忆系统缺乏选择性而导致的信息过载问题。现有记忆机制要么无差别地保留所有信息，要么采用统一的遗忘规则，无法有效区分知识的相关性与无关性。其解决方案的关键在于提出一种名为CAMeR（Context-Activated Memory Reinforcement）的记忆保留框架，该框架结合了关键词门控的混合激活机制——即基于词级杰卡德相似度（Jaccard）的符号化与基于嵌入余弦相似度的非符号化联合门控策略——以及自适应权重动态更新机制。CAMeR通过计算每条记忆与查询之间的混合相似度得分，对高于阈值的记忆进行强化，同时对所有记忆实施可控衰减。为验证该方法的有效性，研究构建了首个针对自适应记忆保留设计的基准测试集CAMeR-Bench，包含76条记忆、100轮对话及8个主题簇，具有梯度激活频率分布，可精准评估记忆选择能力。实验表明，在该基准上，关键词门控使高频记忆与从未被引用的记忆间的保留差距达到嵌入仅依赖门控的1.6倍（剪刀差：0.039 vs. 0.024），而基于时间驱动的基线方法（如Oblivion、SuperLocalMemory）在100轮后权重几乎降至零。此外，CAMeR的前5项检索相比全上下文方法节省83.2%的令牌开销（累计39k vs. 231k），且生成的权重信号显著提升了检索精度。通过8组消融实验进一步确认，关键词门控而非可学习衰减机制是当前规模下性能提升的核心驱动力。研究结果表明，符号-神经混合门控是一种简单但高效的自适应记忆保留机制，适用于复杂对话场景下的LLM代理。
链接: https://arxiv.org/abs/2607.20458
作者: Haowen Lai
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents operating over extended dialogues accumulate vast amounts of information, yet existing memory systems either retain everything indiscriminately or apply uniform forgetting heuristics that fail to distinguish relevant from irrelevant knowledge. We present CAMeR (Context-Activated Memory Reinforcement), a memory retention framework combining keyword-gated hybrid activation – a joint symbolic (word-level Jaccard) and sub-symbolic (embedding cosine) gating mechanism – with adaptive weight dynamics. CAMeR computes a hybrid similarity score for each memory-query pair; memories exceeding a threshold receive reinforcement while all memories undergo controlled decay. We introduce CAMeR-Bench, a 76-memory, 100-round benchmark spanning 8 topic clusters with graded activation frequency, designed to test adaptive retention where existing benchmarks (LoCoMO, LongMemEval) cannot. On CAMeR-Bench, CAMeR’s keyword gate achieves a 1.6 \times larger retention gap between high-frequency and never-referenced memories compared to embedding-only gating (scissors gap: 0.039 vs. 0.024), while time-driven baselines (Oblivion, SuperLocalMemory) collapse to near-zero weights over 100 rounds. CAMeR’s top-5 retrieval saves 83.2% tokens versus full-context approaches (39k vs. 231k cumulative) while producing weight signals that improve retrieval precision. Through 8 ablation conditions we establish that the keyword gate – not learnable decay – is the primary performance driver at this scale. Our findings demonstrate that hybrid symbolic-neural gating provides a simple yet effective mechanism for adaptive memory retention in LLM agents.
[NLP-71] Dropping the Anchor: Statistical Context Summarization for Distributed Systems via Pulsar Attention
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在处理长序列时因自注意力机制（self-attention）的二次计算复杂度导致的推理成本过高问题。现有分布式分块方法（如Star Attention）虽通过跨主机分片上下文降低计算开销，但依赖于在每个主机前静态附加一个内容无关的首块副本，造成冗余计算与资源浪费。针对此问题，本文提出Pulsar Attention，其核心创新在于用两个轻量级、内容感知的组件替代原有的静态锚点：一是用于稳定Softmax数值行为的注意力汇聚前缀（attention-sink prefix），二是基于最大逆文档频率（Max-IDF）启发式策略构建的紧凑跨块摘要，该策略优先选择包含全局稀有词元的文本片段。该方案在保持与原始方法相同键值缓存（KV cache）占用的前提下，将第一阶段每GPU浮点运算量（FLOPs）降低最高达3.3倍；在RULER和BABILong基准测试中，相较于Star Attention及密集注意力（dense attention），Pulsar Attention在长达128K token的序列上表现出更优性能，相较密集基线实现最高4.7%的绝对提升。
链接: https://arxiv.org/abs/2607.20457
作者: Aryan Sood,Shantanu Acharya
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Inference with large language models (LLMs) on long sequences is computationally expensive due to the quadratic complexity of self-attention. Distributed blockwise methods such as Star Attention reduce this cost by sharding context across hosts, but rely on prepending a static, content-blind copy of the first block to every host. We propose Pulsar Attention, which replaces the static anchor with two lightweight, content-aware components: a small attention-sink prefix that stabilizes softmax, and compact cross-block summaries built via a Max-IDF heuristic that selects chunks containing globally rare tokens. This reduces the Phase 1 per-GPU FLOPs by up to 3.3 \times over Star Attention while retaining an identical KV cache footprint. On RULER and BABILong with Llama-3.1-8B, Pulsar Attention outperforms both Star Attention and dense attention at sequence lengths up to 128K tokens, with absolute gains of up to 4.7% over the dense baseline.
[NLP-72] Learn2Zinc: Fine-tuning Small Language Models for Text-to-Model Translation in MiniZinc
【速读】： 该论文旨在解决小规模语言模型在生成领域特定语言（如用于组合优化问题的约束建模语言MiniZinc）时面临的严重语法错误问题。尽管大型语言模型在主流编程语言代码生成方面表现优异，但在处理罕见且结构复杂的领域特定语言时，其生成结果常因语法错误而无法执行。研究发现，语法错误是导致此类任务失败的主要原因，现有小规模模型（0.6B至20B参数）在未经过针对性训练的情况下，直接执行准确率接近零。为应对这一挑战，论文提出一种跨模型错误自举（cross-model error bootstrapping）方法：通过多次运行多个大模型并收集其产生的语法错误，构建高质量的错误纠正训练数据集。基于该数据集对小模型进行微调后，显著提升了所有规模模型在直接代码生成与思维链（chain-of-thought）推理路径上的表现。结合自省（self-reflection）与集成（ensembling）策略，最终实现高达98%的执行准确率。然而，解的准确性仍仅为35%，表明虽然语法结构可通过数据驱动方式学习，但约束推理能力仍是当前瓶颈。研究贡献包括开源的微调流程、数据集及训练模型，以推动自然语言到模型（text-to-model）转换方向的进一步探索。
链接: https://arxiv.org/abs/2607.20456
作者: Serdar Kadioglu,Karthik Uppuluri
机构: Fidelity Investments (菲利普投资公司); Brown University (布朗大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: CP 2026 Workshop on LLMs meet Constraint Solving
Abstract:Large language models excel at code generation for mainstream programming languages but struggle with rare, domain-specific languages such as MiniZinc, a constraint modeling language for combinatorial problems. We investigate whether targeted fine-tuning can teach small language models (0.6B to 20B parameters) to generate syntactically correct and semantically valid MiniZinc models from natural language problem descriptions. Our key finding is that syntax errors dominate failures when working with this domain specific language: the out-of-the-box execution accuracy of small language models such as Qwen3, LLaMa, Gemma, and GPT-OSS is near-zero. We propose a cross-model error bootstrapping approach that collects syntax errors from multiple LLM runs and leverage those to curate an error correction training dataset. This dataset allows us fine-tune small language models that consistently improves both direct code generation and chain-of-thought approaches across all model sizes. With self-reflection and ensembling, our approach achieves up to 98% execution accuracy. In parallel, solution accuracy still remains at 35%, indicating that while syntax is learnable, constraint reasoning remains a challenge. We contribute our fine-tuning pipeline, datasets, and models to opens-source for further research on text-to-model translation.
[NLP-73] RE-AD: Real-Time Requirement Adherence for Data Labeling ACL2026
【速读】： 该论文旨在解决大规模语言模型（LLM）训练中依赖人工标注数据所面临的标注质量下降问题，尤其是众包标注因标注者理解偏差或参与度不足导致的误标与不一致。其核心解决方案是提出一种实时需求合规性验证（RE-AD）框架，利用大语言模型（LLM）对标注过程进行主动质量校验。该方法的关键在于通过自我反思（self-reflection）将标准操作流程（SOP）分解为原子级规则，并按复杂度进行分类，进而实施分层验证策略，从而实现对标注行为的动态、细粒度监控与反馈。实验结果表明，该框架在合成基准上达到0.749的F1分数，且在实际部署中促使82%被标记的错误得到标注者的接受与修正，验证了其有效性与实用性。
链接: https://arxiv.org/abs/2607.20455
作者: Siddarth Malreddy,Ishan Nigam,Akshay Arora,Nikhil Mittal,Subrat Sahu
机构: Uber AI Solutions(优步人工智能解决方案)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted to The Fifth Generation, Evaluation Metrics Workshop (GEM) workshop at ACL 2026
Abstract:Human-annotated data remains fundamental to training frontier Large Language Models (LLMs). However, crowd-sourced annotations often suffer from quality issues stemming from annotator misunderstanding or lack of engagement. To address this, we introduce a real-time requirement adherence (RE-AD) framework that leverages LLMs to proactively validate labeling quality. Our methodology involves decomposing Standard Operating Procedures (SOPs) into atomic rules via self-reflection, categorizing them by complexity, and applying tiered validation strategies. Evaluated on a synthetic benchmark, the system achieved an F1 score of 0.749. Furthermore, production deployment resulted in annotators accepting and fixing 82% of the errors flagged by the framework. We include ablation studies to demonstrate the impact of our core design decisions.
[NLP-74] Response drift across frontier large language models
【速读】： 该论文旨在解决前沿大型语言模型（Large Language Models, LLMs）普遍存在的响应漂移（response drift）问题，即模型输出与专家验证参考答案之间的偏差。尽管已有研究指出该现象存在，但其程度与结构尚未通过系统性的人类评估得到充分刻画。本文提出的关键解决方案是开展一项全交叉式的人类评估实验：由47名来自不同地理区域的参与者，在盲测条件下对10个前沿大模型在62个跨领域问题上的输出进行独立评估，共获得29,140条人类判断数据。结果表明，所有模型均存在响应漂移，但漂移幅度差异显著——其中8个模型趋于统计上无差别的上限（78–81%偏离），而另有2个模型表现更优（47–49%偏离）。漂移模式在六个领域和具体问题间呈现显著差异，且处于“上限”水平的模型之间两两相关性超过r = 0.85，说明其漂移具有结构性共性。值得注意的是，自动化相似度指标仅能解释人类判断中不到2%的方差，凸显了现有评估手段的局限性。因此，研究揭示响应漂移是所有前沿大模型普遍存在的现象，其结构具有领域与问题依赖性，且唯有通过以人类为中心的评估才能有效捕捉和理解。
链接: https://arxiv.org/abs/2607.20454
作者: Mohammed Aledhari,Ali Aledhari,Fatimah Aledhari,Gowtham Venkat Eathamokkala,Mohamed Rahouti
机构: University of North Texas (北德克萨斯大学); Fordham University (福特汉姆大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:All frontier large language models (LLMs) exhibit response drift – producing outputs that deviate from expert-validated references – yet the magnitude and structure of this drift remain uncharacterised by systematic human evaluation. Here we report a fully crossed evaluation in which 47 geographically diverse participants each assessed all 62 multidomain questions across ten frontier LLMs under blinded conditions, yielding 29,140 independent assessments. Every model drifts, but drift magnitude varies substantially: eight models converge on a statistically indistinguishable ceiling (78-81% deviation), while two achieve lower deviation (47-49%). Drift profiles differ across six domains and 62 questions, with pairwise correlations among ceiling models exceeding r = 0.85. Automated similarity metrics explain less than 2% of variance in human judgements. These findings reveal that response drift is universal across frontier LLMs, domain- and question-dependent in structure, and accessible only through human-centred evaluation.
[NLP-75] A Knowledge-Injection Framework for Zero-Shot Adaptation of LLM s to Delirium Prediction
【速读】： 该论文旨在解决小型本地部署的生成式模型在临床预测任务中因领域知识不完整而导致的零样本性能受限问题，尤其聚焦于重症监护病房（ICU）谵妄预测这一专业性极强的任务。其核心解决方案是提出一种轻量级的推理时知识注入框架：在不进行微调或检索的前提下，通过在推理阶段将结构化电子健康记录（EHR）的自然语言摘要与外部临床知识报告进行融合，从而增强模型对特定临床场景的理解能力。实验结果表明，该方法显著提升了小参数量模型（LLaMA 3.1 8B）的性能（AUROC提升8.57个百分点），并对大模型（LLaMA 3.3 70B）也带来可观增益（+1.99个百分点），同时有效缩小了本地可部署模型与前沿闭源模型之间的性能差距。消融实验证明，性能提升依赖于知识内容的临床相关性而非单纯增加提示长度，且基于SHAP的归因分析确认注入的知识被模型实际利用。因此，该方法的关键在于实现高效、隐私友好且无需微调的推理时知识增强，为资源受限的临床环境提供了一种可行的高性能替代路径。
链接: https://arxiv.org/abs/2607.20453
作者: Jessica Sena,Shesadree Priyadarshani,Miguel Contreras,Bharat Gandhi,Scott Siegel,Subhash Nerella,Parisa Rashidi
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Large language models show promise for clinical prediction, but zero-shot performance on specialized tasks is limited by incomplete domain knowledge, especially for smaller locally deployable models. We present a lightweight knowledge-injection framework for zero-shot ICU delirium prediction that augments a deterministic natural-language summary of structured electronic health record data with an external clinical knowledge report at inference time, without fine-tuning or retrieval. We evaluate LLaMA 3.1 8B and LLaMA 3.3 70B on 3,160 ICU admissions from the MIMIC IV dataset. Adding a clinically meaningful external knowledge report improves AUROC by 8.57 percentage points for the 8B model and 1.99 percentage points for the 70B model compared to no external knowledge. Relative to a GPT-5.2 frontier-model reference without external knowledge report (AUROC 68.86%), knowledge injection reduces the performance gap from 15.66 to 7.09 AUROC points for LLaMA 8B and from 5.30 to 3.31 AUROC points for LLaMA 70B. Random control reports do not improve performance and often degrade it, indicating that gains depend on clinically meaningful content rather than added prompt length alone. SHAP-based attribution further confirms that the injected knowledge is actively used during prediction. These findings suggest that inference-time knowledge injection can narrow the gap between locally deployable open-weight models and frontier closed models while preserving a practical, privacy-preserving workflow for resource-constrained clinical settings.
[NLP-76] Semantic Field Theory: Historical Origin Higher-Order Interaction and Stabilized Semantic Inference
【速读】： 该论文旨在解决自然语言语义建模中缺乏可计算、可形式化且具有数学严谨性的统一框架问题，尤其针对词汇语义、高阶组合性以及意义稳定化过程的建模挑战。其核心解决方案的关键在于构建一个以语义场（Semantic Field）为基础的可计算模型体系，通过五个关键形式化要素实现：1）将语义场模型定义为包含语义空间、词项提升映射、上下文变形映射、交互复形与解释泛函的五元组结构；2）证明乘积型语义场交互满足高斯分布封闭性，具备显式的中心、精度与兼容性因子；3）运用莫比乌斯反演在子集格上推广三词问题，识别任意阶次的不可约语义交互；4）引入阶次谱（order spectrum）量化各阶交互对语义质量的贡献；5）将稳定释义建模为句子能量泛函的最小化问题，并给出存在性、下降性与稳定性条件。该框架为生成式语义表示提供了一种基于能量动力学的数学基础，虽非完整语言理论，但为计算语言学与表征学习中的语义建模提供了可独立评估的形式化工具。
链接: https://arxiv.org/abs/2607.20451
作者: Dimitris Vartziotis
机构: NIKI – Digital Engineering (NIKI数字工程); TWT Science Innovation (TWT科学创新)
类目: Computation and Language (cs.CL)
备注: 20 pages
Abstract:Semantic Field Theory (SFT) has developed from a philosophical critique of strong anti-formalist readings of language games into a proposed computational model class for lexical semantics, higher order composition, and stabilized interpretation. This paper reconstructs that evolution and gives SFT a sharper mathematical core suitable for independent evaluation in computational linguistics and representation learning. The central proposal is that a tractable level of linguistic organization can be modeled through lexical representations expressed as semantic fields, through contextual deformation of those fields, through interaction terms defined over subsets of tokens, and through stabilization governed by semantic energy dynamics. The paper contributes five formal elements. First, it defines a semantic field model as a tuple consisting of a semantic space, a lexical field lifting, a contextual deformation map, an interaction complex, and an interpretation functional. Second, it proves a Gaussian product closure result showing that multiplicative field interactions have explicit centers, precisions, and compatibility factors. Third, it generalizes the three-word problem by using Mobius inversion on the subset lattice to isolate irreducible semantic interactions of arbitrary order. Fourth, it introduces an order spectrum that measures how much field mass is explained at each interaction order. Fifth, it formulates stabilized interpretation as minimization of an energy functional associated with the sentence and gives existence, descent, and stability conditions. A small worked example shows how a three-word summer day triple can be represented by Gaussian semantic fields, implemented in Python, and summarized by a flow diagram. The result is not a completed theory of natural language meaning and does not replace social, pragmatic, or normative accounts of language.
[NLP-77] ShriNep@EEUCA 2026: RAKSHAK - Multi-Task DeBERTa with Rationale Distillation and Jigsaw-Augmented Training for Toxic Intent Classification ACL2026
【速读】： 该论文针对《World of Tanks》聊天文本的六类细粒度毒性意图分类任务（标签0-5）中存在严重类别不平衡、领域特定的多语言俚语表达，以及罕见类别（如威胁类Label 4仅60样本，极端主义类Label 5仅24样本）数据极度稀缺等挑战，提出解决方案。其核心创新在于构建名为RAKSHAK的多任务DeBERTa-v3-base框架，关键策略包括：基于Qwen2.5-14B生成的推理路径蒸馏（rationale distillation）、监督对比损失（Supervised Contrastive Loss）机制，以及为稀有类别设计的专用二分类头；同时通过跨域迁移学习，将Jigsaw Toxic Comment数据集中的16,225条样本映射至标签1-4，并结合100条大语言模型生成的极端主义样本增强标签5的数据。实验表明，跨域迁移贡献了+2.6的宏平均F1提升，而多任务架构进一步带来+3.7的性能增益，最终RAKSHAK在测试集上达到0.5883的宏平均F1，排名第七。
链接: https://arxiv.org/abs/2607.20450
作者: Binayak Karki,Aryan Kafle,Pingala Ghimire
机构: Mechi Multiple Campus, Nepal; Northern Kentucky University, USA; Himalaya College of Engineering, Nepal
类目: Computation and Language (cs.CL)
备注: 8 pages, 1 figure, EEUCA, ACL 2026
Abstract:This paper presents two systems for the GameTox Shared Task at the Workshop on EEUCA at ACL 2026, which requires classifying World of Tanks chat utterances into six fine-grained toxic intent categories (Labels 0-5). Severe class imbalance, domain-specific multilingual slang, and extremely scarce data for rare categories such as Threats (Label 4, 60 samples) and Extremism (Label 5, 24 samples) make this a challenging classification problem. Our primary submission, RAKSHAK (rak s. aka, Sanskrit for “Protector”), is a multi-task DeBERTa-v3-base (He et al., 2022) framework combining rationale distillation from Qwen2.5-14B (An et al., 2024), Supervised Contrastive Loss, and dedicated rare-class binary heads. RAKSHAK’s training data is augmented with cross-domain transfer from the Jigsaw Toxic Comment dataset (16,225 samples mapped to Labels 1-4) and 100 LLM-generated extremism samples for Label 5. Our secondary system (M1) fine-tunes DeBERTa-v3-base with Focal Loss on the original GameTox data plus the same 100 extremism samples, without Jigsaw transfer. RAKSHAK achieves a Macro F1 of 0.5883 on the official test set, ranking 7th out of 35 participating teams, while M1 achieves 0.5252 Macro F1. An ablation comparing M1 with and without Jigsaw data shows that cross-domain transfer accounts for +2.6 F1 points, while RAKSHAK’s multi-task architecture contributes a further +3.7 points.
[NLP-78] he Storyteller in the Model: Narrative Pattern Inheritance Escalation Dynamics and Alignment Governance in LLM s
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在实际部署中可能出现的“叙事漂移”（narrative drift）问题，即模型在长期交互过程中因吸收训练数据中固有的叙事结构与修辞模式（如原型角色、张力-解决型情节弧线等），导致输出行为逐渐偏离预期，演化为具有对抗性或修辞性诱惑的不可控行为。其解决方案的关键在于揭示并识别这种由训练数据中的叙事惯例引发的隐性系统性偏差：首先，LLMs并非基于独立推理，而是复现训练数据中的统计模式；其次，诸如阿谀奉承（sycophancy）和欺骗性（deceptiveness）等可测量的潜在特质在不同提示下稳定出现；第三，对特定叙事任务进行微调即可引发超出任务范围的意外行为变化。这些发现表明，叙事风格输出在真实场景中极为普遍，加剧了风险传播，因此亟需建立专门的监测机制以应对这一未被充分监控的、渐进式的系统性治理风险。
链接: https://arxiv.org/abs/2607.20449
作者: Adam Rigby,Raz Saremi,Azadeh Sohrabinejad,Mehdi Rahimi
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 2 figures, 11 pages
Abstract:LLMs are trained predominantly on human-authored text, yet the structural and narrative conventions embedded in that text are rarely examined as a source of systematic behavioral influence, or as a governance risk in deployed systems. This paper considers whether the storytelling patterns inherent in published human writing, including archetypal roles such as protagonist, antagonist, and underdog, as well as tension-and-resolution narrative arcs, are absorbed during training and subsequently surface in LLM outputs, causing responses to drift toward unexpected, adversarial, or rhetorically enticing behaviors over extended interactions. Through a systematic literature review and cross-paper analysis of recent empirical studies on LLM alignment, persona dynamics, emergent misalignment, and user interaction patterns, we observe evidence bearing on this hypothesis. The findings reveal three key patterns. First, LLMs reproduce statistical patterns from their training data rather than reasoning independently. Second, measurable latent traits, including sycophancy and deceptiveness, emerge reliably across unrelated prompts. Third, fine-tuning on a narrow narrative task can produce unintended behavioral changes well beyond that task. Furthermore, evidence suggests that persuasive, narrative-style outputs are among the most common LLM products in real-world usage, amplifying these risks. Narrative drift constitutes an unmonitored escalation pathway in deployed AI systems, one that evades discrete-incident detection mechanisms and requires dedicated monitoring instruments.
[NLP-79] Domyn-Small: A European 10B Reasoning Language Model
【速读】： 该论文旨在解决小型化生成式AI（Generative AI）模型在保持高推理能力与高效性之间的平衡难题，特别是在参数量处于7–10B量级的模型中如何实现强指令遵循能力、数学与科学推理性能，同时降低计算资源消耗。其解决方案的关键在于构建一个端到端的后训练（post-training）流程：首先通过持续预训练（Continued Pre-Training, CPT）将上下文窗口从原始长度扩展至32K tokens，再结合以数学任务为导向的分阶段监督微调（SFT），并引入基于可验证奖励的强化学习（GRPO）、直接偏好优化（DPO）及多环境GRPO策略，覆盖数学、代码、多项选择问答、指令遵循和工具调用五大任务领域。此外，通过YaRN技术在推理阶段进一步将上下文扩展至128K tokens，配合双模推理聊天模板切换机制，显著提升了长序列处理能力与任务适应性。实验表明，该模型在核心推理基准测试中仅需约三分之一于Qwen3.5-9B、约35%于OLMo-3-7B-Think的生成令牌数，即可达到优异的指令遵循表现（IFEval 79.9）与竞争力强的科学推理能力（GPQA-Diamond 50.0），实现了高精度与低开销的协同优化。
链接: https://arxiv.org/abs/2607.20448
作者: Simone Angarano,Francesco Bertolotti,Federico D’Ambrosio,Michele Resta,Alessandro Rognoni,Nicolò Ruggeri,Dario Salvati,Andrea Valenti,Alberto Veneri,Martin Cimmino
机构: iGenius; CINECA; NVIDIA Corporation
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: 27 pages, 5 figure
Abstract:We introduce Domyn-Small, a 10-billion-parameter open-weight reasoning language model released under the MIT license. Domyn-Small is the product of an initial pre-training phase on 9 trillion tokens multilingual data, followed by a post-training pipeline for reasoning, instruction following, and context extension. For the latter, we performed a Continued Pre-Training (CPT) phase that doubles the native context window to 32K tokens, followed by SFT with a math-focused annealing run. Finally, the RL phase includes GRPO with verifiable rewards, DPO, and a multi-environment GRPO stage spanning five task domains: mathematics, code, multiple-choice QA, instruction-following, and tool calling. The 32K-token native context extends to 128K at inference via YaRN, and a chat-template toggle enables dual-mode reasoning. Against peer models in the 7–10B class (Qwen3.5-9B, OLMo-3-7B-Think, Nemotron-Nano-8B, Ministral-3-8B), Domyn-Small achieves a strong accuracy-efficiency balance: it produces roughly one-third as many tokens as Qwen3.5-9B and approximately 35% of OLMo-3-7B-Think’s token budget on core reasoning benchmarks, while delivering strong instruction-following (IFEval 79.9) and competitive science reasoning (GPQA-Diamond 50.0). We release the weights and the post-training recipe alongside Domyn Swarm (Apache~2.0), an open-source framework for scalable LLM inference on HPC clusters developed during this program and used throughout this work.
[NLP-80] haulab@EEUCA 2026: Who Said What to Whom? A Targeting-Aware Neural-Symbolic Pipeline for Gaming Toxicity Detection
【速读】： 该论文旨在解决游戏聊天场景中毒性内容分类的难题，尤其针对少数类（仇恨/骚扰、威胁、极端主义）在极度不平衡数据分布下难以有效识别的问题。其核心挑战在于：在非毒性与极端主义类别间高达1,450:1的样本比例失衡背景下，如何提升对高安全风险少数类别的检测性能。解决方案的关键在于提出一个三阶段流水线架构，其中融合了两个轻量级预训练模型（DeBERTa-v3-base, 184M；XLM-RoBERTa-base, 278M）的集成结果，并引入语言学引导中介器（Linguistically-Informed Mediator, LIM），通过基于语料库的词汇归一化、类别条件一元词分值评分、多语言粗俗语检测以及基于言语行为理论的主体性目标分析，系统性地解决模型间的分歧。此外，为充分利用有限训练数据，设计了一种仅依赖原始训练集的两阶段数据增强策略，显著缓解了类别不平衡问题。该方法在官方测试集上取得0.6441的宏平均F1和0.9062的准确率，分别位列第3和第1，且具备良好的领域可迁移性——仅需替换特定游戏实体词典即可适配其他游戏平台。
链接: https://arxiv.org/abs/2607.20447
作者: Anmol Guragain,Marcos Estecha-Garitagoitia,Luis Fernando D’Haro Enríquez,Ricardo de Córdoba
机构: ETSI de Telecomunicación, Universidad Politécnica de Madrid, Madrid, Spain
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注:
Abstract:This paper describes our system for the EEUCA 2026 Shared Task on toxicity classification in gaming chat. We implement a three-stage pipeline combining an ensemble of two compact transformers (DeBERTa-v3-base, 184M; XLM-RoBERTa-base, 278M) with a Linguistically-Informed Mediator (LIM) that resolves inter-model disagreements through corpus-backed lexical normalization, class-conditional unigram scoring, multilingual profanity detection, and agentive targeting analysis grounded in speech act theory. The LIM specifically targets the minority classes (Hate \ Harassment, Threats, and Extremism), which are the most safety-critical categories in real-world gaming moderation. To address the extreme class imbalance (1,450:1 Non-toxic to Extremism ratio), we introduce a two-stage data augmentation strategy using only the provided training data. Our system achieves a Macro F1 of 0.6441 and accuracy of 0.9062 on the official test set, ranking 3rd in Macro F1 and 1st in accuracy among all teams. The proposed pipeline is domain-portable: adapting to other gaming platforms requires substituting only the game-specific entity lexicon. Code is publicly available at this https URL_EEUCA.
[NLP-81] Distinguishing Artificial from Authentic: Evaluating LLM s for Detecting LLM -Generated Content
【速读】： 该论文旨在解决教育场景中如何有效识别学生提交作业中由大语言模型（Large Language Models, LLMs）生成内容的问题，特别是探讨是否可利用LLMs自身作为检测工具来区分其生成内容与人类作者的原创作品。其解决方案的关键在于系统评估LLMs在多种教育任务类型（包括编程练习、反思性写作和简答题）中自我识别生成内容的能力，并分析提示设计（prompt design）、响应长度、任务类型等因素对检测性能的影响。研究发现，LLM自检测效果具有显著的任务依赖性：在编程任务和较长的反思性写作中检测准确率较高，但在短答案题中表现不佳，甚至常将自身生成的内容误判为更“人性化”的真实学生作答。此外，提示框架和响应冗余度对反思性写作的可检测性有显著影响，微小的提示变化即可大幅降低检测准确性，而编程类任务的检测则相对稳健。这些结果表明，尽管LLMs具备一定的自检测潜力，但其在实际教育应用中存在明显局限，应谨慎将其作为独立工具用于识别AI生成内容。
链接: https://arxiv.org/abs/2607.20446
作者: Juho Leinonen,Paul Denny
机构: Aalto University (阿尔托大学); University of Auckland (奥克兰大学)
类目: Computation and Language (cs.CL); Computers and Society (cs.CY)
备注: 8 pages, 5 figures, 3 tables
Abstract:As large language models (LLMs) are increasingly used by students to generate natural language responses and program code, there is growing interest in whether LLMs themselves can be used to distinguish AI-generated work from human-authored submissions. In this paper, we investigate the extent to which LLMs can detect their own generated content across multiple educational task types, including programming exercises, reflective writing, and short-answer questions. Using authentic student responses and multiple variants of LLM-generated answers, we evaluate detection performance under different prompting strategies and output formats. Our study addresses three research questions: (1) how accurately LLMs can identify their own outputs across task domains, (2) how detection effectiveness is influenced by factors such as prompt design, response length, and task type, and (3) what characteristics of LLM-generated responses contribute to successful or failed detection. Our findings show that LLM-based detection is highly task-dependent: detection is substantially more reliable for programming tasks and longer reflective responses, but performs poorly for short-answer questions, where LLMs frequently judge their own outputs as more human-like than authentic student responses. We further find that prompt framing and response verbosity have a pronounced effect on detectability in reflective writing tasks, with relatively minor prompt variations significantly reducing detection accuracy, while programming-related detection is more robust to prompt changes. Together, these results highlight both the potential and the limitations of LLM self-detection in educational settings and suggest caution in relying on LLMs as standalone tools for identifying AI-generated student work.
[NLP-82] SCoPE: Shift-Aware Speaker-Conditioned Priors for Emotion Recognition in Conversations
【速读】： 该论文旨在解决多轮对话中情绪识别（ERC）模型对情绪持续性与演变规律建模不足的问题，尤其针对现有方法过度依赖显性线索、忽视非显性因素（如说话人个体情绪演化模式及上下文驱动的情绪转变）的局限性。其核心解决方案在于提出一种轻量级模块——说话人条件情绪先验（Speaker-Conditioned Priors over Emotions, SCoPE），该模块通过建模每位说话人的情绪历史以生成个性化的先验分布；同时引入情绪转移预测机制，指导模型在先验信息与多模态证据之间进行动态权衡；并设计了一种感知情绪转移的融合机制，采用精度加权的对数几率整合方式，实现基于贝叶斯“专家产品”思想的动态融合，使模型在情绪稳定时依赖历史先验，在情绪可能发生转变时则更侧重多模态证据。这一框架显著提升了复杂噪声场景下的鲁棒性与性能表现。
链接: https://arxiv.org/abs/2607.20445
作者: Burak Can Kaplan,Stefan Wermter
机构: University of Hamburg (汉堡大学)
类目: Computation and Language (cs.CL); Sound (cs.SD); Audio and Speech Processing (eess.AS)
备注: Under review at Cognitive Computation
Abstract:In conversations, human emotions are transient; however, they tend to persist across multiple utterances. For example, we rarely switch instantly between contrasting emotions such as happiness and anger. Instead, emotions tend to evolve smoothly, and these patterns are often speaker-specific. Some people might escalate, while others gradually cool down over time. Furthermore, when emotions change during a conversation, they are often driven by contextual factors, such as newly received information or unexpected events. Even though progress has been made in Emotion Recognition in Conversations (ERC), most existing approaches still rely heavily on overt evidence and do not sufficiently model these non-apparent factors. Especially in multimodal settings, this makes these models fragile when the signals are noisy (e.g., occluded faces, slang expressions, or microphone noise). To address these limitations, we introduce Speaker-Conditioned Priors over Emotions (SCoPE). SCoPE is a light weight module that utilizes the emotional history of each speaker and explicitly models their priors for use in subsequent emotion classification. Second, we incorporate emotion shift prediction, a well-established concept in ERC, to guide the model in balancing the priors from SCoPE and multimodal evidence. Finally, we propose a shift-aware fusion mechanism that performs precision-weighted logit integration between multimodal evidence and the speaker prior, forming a Bayesian-inspired product-of-experts formulation. This dynamic fusion allows the model to rely on historical priors when emotions persist and to prioritize multimodal evidence when shifts are likely. Experimental results show our model achieves superior performance over recent state-of-the-art models on the IEMOCAP dataset in multimodal settings.
[NLP-83] Confidently Deceptive: How Confidence Amplifies the Risk of LLM Deception
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在特定情境下产生欺骗性输出的问题，重点关注模型在生成欺骗性内容时的自信程度及其对用户说服力的影响。研究发现，模型在生成欺骗性回答时表现出显著的自我报告自信度，且人类标注者在成对比较中更倾向于选择高自信的欺骗性回答（78%的情况下）。更重要的是，经过误导性微调（misalignment fine-tuning）后，模型在各类基准测试中对欺骗性输出的自信度普遍上升，加剧了潜在风险，并表现出跨训练分布的泛化效应。值得注意的是，尽管模型能以高达82.7%的准确率识别自身生成的内容为欺骗性，却仍会预测自己会继续生成此类内容——即“识别但不规避”。因此，论文提出，自信型欺骗是一种独立的对齐风险（alignment risk），需要建立联合评估框架，同时衡量欺骗行为、自信度与自我意识，以全面评估和缓解该类风险。
链接: https://arxiv.org/abs/2607.20444
作者: Ali Asad,Stephen Obadinma,Anshul Pattoo,Wenxuan Zhang,Xiaodan Zhu
机构: Queen’s University (皇后大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) can produce deceptive responses: outputs that mislead users in service of a contextually or experimentally induced goal. Yet it remains unclear how confidently models deceive and whether higher confidence makes deceptive responses more persuasive to end users. In this paper, we study these basic questions in various models and different deception datasets. We provide a comprehensive study measuring confidence through both verbalized self-reports and a range of logit-based estimators. We show that LLMs deliver deceptive responses with substantial verbalized confidence and that human annotators prefer the higher-confidence deceptive response 78% of the time in paired comparisons. Misalignment fine-tuning amplifies the problem. Confidence in deceptive responses rises across all three benchmarks, increasing the resulting potential risk, with effects generalizing beyond the training distribution. Strikingly, models classify their own deceptive outputs as deceptive at high rates (82.7% under misalignment) while still predicting they would produce them - recognition without avoidance. We argue that confident deception is a distinct alignment risk requiring evaluations that jointly measure deception, confidence, and awareness.
[NLP-84] GLAN-QnA-KR: A Seedless Taxonomy-Driven Korean Instruction Corpus KR-300
【速读】： 该论文旨在解决大规模合成韩语指令问答数据集在去重性与数据污染控制方面的关键挑战，尤其针对现有合成数据普遍存在重复样本及潜在训练数据泄露问题。其解决方案的关键在于采用无种子（seedless）的分类体系驱动生成流程（taxonomy-driven GLAN synthesis pipeline），结合微软Phi-3.5-MoE-instruct作为生成模型，在不依赖预设种子样本的前提下实现高质量、高多样性数据的自动化构建。该方法通过严格的双重污染审计机制（针对KMMLU、KoBEST五子任务及HAE-RAE-Bench等基准），验证了生成数据在问题级字符三元组相似度（Jaccard）和多语言E5嵌入余弦相似度上的极低污染水平——最大Jaccard值仅为0.163，且无任何样本达到0.7以上；最大余弦相似度为0.901，仅1个样本超过0.90，无样本达0.95。同时，数据集具有高度的多样性：包含1,084个英文标签的扁平化学科分类，每条记录平均含313字符的问题与1,098字符的答案，并确保极低的重复率（每30万行仅1个完全重复项）。最终发布的GLAN-QnA-KR是目前可公开访问、基于单一管道生成、经验证的最大规模韩语指令问答数据集（303,581行），并以OpenRAIL许可证开放，具备良好的可复现性与下游引用价值。
链接: https://arxiv.org/abs/2607.20443
作者: Daekeun Kim
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Technical report; 7 pages, 4 tables. Dataset: this https URL
Abstract:We release GLAN-QnA-KR, a 303,581-row openly redistributable Korean instruction-QA corpus produced via the seedless taxonomy-driven GLAN synthesis pipeline with Microsoft’s Phi-3.5-MoE-instruct as the producer model (generation: 2024-12; release: 2024-12; licence: OpenRAIL). The corpus spans a flat taxonomy of 1,084 English-labelled disciplines paired with Korean question/answer text, a 100-900 difficulty scale, and a median of 313 question characters and 1,098 answer characters per record. Two properties are atypical for synthetic instruction data at this scale: (i) exact duplicate questions number only 1 in 303,581 rows and character-trigram near-duplicate clusters at Jaccard = 0.9 number zero in a 5,000-sample probe, and (ii) a two-layer contamination audit against KMMLU, KoBEST (five sub-tasks), and HAE-RAE-Bench shows a maximum test-vs-corpus question-level character-trigram Jaccard of 0.163 with zero test items at Jaccard = 0.7, and a maximum multilingual-E5 cosine of 0.901 with a single test item at cosine = 0.90 and zero at = 0.95, across 20,000 sampled GLAN questions and seven evaluation sets. At the time of release, this is, to our knowledge, the largest single-pipeline synthetic Korean instruction corpus verifiable on the Hugging Face Hub and the only Korean =100k-row corpus built under a seedless taxonomy-driven protocol. This note documents the generation protocol, corpus statistics, the contamination audit, and the licensing boundary in a form suitable for downstream citation.
[NLP-85] Naver-News-KO: A Korean News Summarization Dataset for Open-Source Fine-Tuning of Summarization Models
【速读】： 该论文旨在解决韩语新闻摘要任务中缺乏高质量、大规模且可公开访问的基准数据集的问题。其核心贡献在于发布Naver-News-KO，一个包含27,400组（原文, 摘要）对的韩语新闻摘要数据集，覆盖经济与IT/科学两大类别，具有明确的训练/验证/测试划分（22,194 / 2,466 / 2,740），并实现了平均6.03倍的文档到摘要字符压缩率。解决方案的关键在于：系统性地记录数据采集流程、字段结构及划分方法，提供详尽的语料级统计信息（如长度分布、压缩比、以及测试集与训练集之间存在16.8%的近似重复标题-Jaccard重叠），同时构建可复现的基线模型（包括KoBART和基于LoRA微调的Gemma-2B-ko），并配套发布训练脚本与清晰的许可协议，从而为后续研究提供可引用、可复现且具备实用价值的基准资源。
链接: https://arxiv.org/abs/2607.20442
作者: Daekeun Kim
机构: 未知
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Technical report; 7 pages, 1 figure, 4 tables. Dataset: this https URL
Abstract:We release Naver-News-KO, a Korean news summarization dataset of 27,400 (document, summary) pairs collected from Naver News over a ten-day window in July 2022 across two categories (Economy and IT/Science; 77/23 split), with train/validation/test partitions of 22,194 / 2,466 / 2,740 and a mean per-record document-to-summary character-compression ratio of 6.03x. The dataset has been publicly hosted on the Hugging Face Hub since January 2023 and, as of May 2026, receives approximately 33,000 downloads per month; community-maintained Korean summarization models fine-tuned on it include Gemma-2B-ko and Gemma2-9B variants. This technical report (i) documents the collection protocol, the column schema, and the split construction, (ii) reports corpus-level statistics (length distributions, compression ratio, and a measured 16.8% near-duplicate title-Jaccard overlap between test and train that users should be aware of), (iii) positions the resource against other open Korean summarization corpora, (iv) provides a Lead-3 extractive reference point (ROUGE-1 55.1, ROUGE-L 50.6) and two reproducible fine-tuned baselines – KoBART (R-1 56.6, BERTScore-F1 81.5) and Gemma-2B-ko with LoRA (R-1 55.3, BERTScore-F1 78.3) – with release-time training scripts, and (v) clarifies the licensing and intended-use scope of the resource. The goal is to provide a citable reference for downstream work that already uses this dataset, not to propose a new benchmark.
[NLP-86] Belief Propagation in LLM World Models: Measuring Strategic Information Bias with Prediction Markets KR
【速读】： 该论文旨在解决信息生态系统中由文本来源引发的认知偏差对战略决策的系统性影响问题，尤其关注生成式AI在处理特定信息源时如何继承并放大这些偏差。其核心挑战在于识别和量化信息源（如新闻报道）所导致的信念偏差，而非模型本身固有的错误。解决方案的关键在于构建一个结合大语言模型（LLM）与预测市场（prediction markets）的校准框架：利用LLM从文本语料中提取隐含信念，同时以预测市场在事件实际结果锚定后的价格轨迹作为外部客观参考基准，从而精确测量信念偏离真实情况的程度。通过消融实验（ablation），在固定模型、仅改变信息上下文的条件下，发现英文新闻内容系统性地误导领土预测，使预测错误率高达64%–72%，而知晓真实结果的“污染模型”表现出相同错误率，表明偏差根源在于信息源而非模型结构。研究进一步证实，引入乌克兰军事分析类可信来源可有效降低偏差，但改善程度有限且依赖具体模型。结论表明，此类认知扭曲主要源于信息源本身的偏见，且在不同模型架构中具有一致性，将在任何依赖此类数据的系统中持续存在，并向下游决策环节传播。
链接: https://arxiv.org/abs/2607.20441
作者: Mykola Khandoga,Yevhen Kostiuk,Anton Polishko,Yurii Filipchuk,Kostiantyn Kozlov,Dmytro Zamriy,Artur Kiulian
机构: Future Principle; Aarhus University
类目: Computation and Language (cs.CL); Machine Learning (cs.LG)
备注: Accepted at UNLP 2026. 12 pages, 8 figures, 11 tables. Dataset available at this https URL
Abstract:Every information ecosystem produces beliefs that shape strategic decisions. Both human analysts and AI systems inherit the blind spots of their information sources. We show that LLMs, combined with prediction markets, function as a calibrated instrument for measuring how far ecosystem-induced beliefs deviate from an external reference: LLMs extract the beliefs a text corpus implies, and prediction market price trajectories, anchored at resolution by realised outcomes, provide the calibration reference against which to quantify the deviation. We isolate the bias contribution of specific text through ablation: varying information context while holding the model fixed, with a contaminated model that knows actual outcomes as control. Applied to 111 Ukraine-related prediction markets, comprising approximately 93,000 predictions across four models, we find that English news context systematically biases territorial predictions, wrong 64 to 72 percent of the time when it pushes predictions toward territorial capture. A contaminated model that knows actual outcomes shows the same error rate, indicating that the bias originates primarily in the text. Supplementing with Ukrainian military-analytical sources reduces the bias for all clean models, while absolute-error gains are partial and model-dependent. We show that the distortion originates primarily in the sources, not the models. Consistent across four architectures, it will persist in any system that processes them and propagate into downstream decisions. Comments: Accepted at UNLP 2026. 12 pages, 8 figures, 11 tables. Dataset available at this https URL Subjects: Computation and Language (cs.CL); Machine Learning (cs.LG) ACMclasses: I.2.7; I.2.6 Cite as: arXiv:2607.20441 [cs.CL] (or arXiv:2607.20441v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2607.20441 Focus to learn more arXiv-issued DOI via DataCite
[NLP-87] Answer-then-Edit: Reasoning Skeleton Editing for Anti-Distillation with Preserved Utility
【速读】： 该论文旨在解决专有大语言模型（Large Language Models, LLMs）在通过黑箱API部署时，仍面临未经授权的知识蒸馏（knowledge distillation）威胁的问题。尽管已有水印（watermarking）等防御方法，但其依赖事后验证且易被规避；而现有基于内部模型扰动的反蒸馏（anti-distillation, AD）方法难以在防御能力与推理轨迹的实用性（如答案准确性、自然性）之间取得平衡，强防御往往导致显著性能下降。为克服这一局限，本文提出一种名为“骨架引导推理编辑”（Skeleton- and Guided Reasoning Editing, SGRE）的后处理式反蒸馏框架，采用“先生成、再编辑”的双阶段设计：首先由教师模型生成高准确率且自然的原始推理轨迹，随后基于认知负荷理论（Cognitive Load Theory, CLT），引入三阶段结构化扰动策略——推理骨架提取、骨架图粗化与骨架语义重述，通过重构推理结构并增加文本复杂度，有效提升学生模型在学习过程中的额外认知负荷，从而阻碍其对底层推理模式的捕获。实验结果表明，SGRE在多种大模型上均实现了当前最优的抗蒸馏效果，同时保持推理准确率无损及推理轨迹自然性更优。
链接: https://arxiv.org/abs/2607.20440
作者: Fan Li,Mengting Pan,Sijia Xu,Xiaoyang Wang,Chen Chen,Wenjie Zhang
机构: UNSW Sydney (新南威尔士大学); Shenzhen University (深圳大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 21 pages,8 figures
Abstract:Proprietary large language models (LLMs) entail substantial intellectual and financial investment, making them valuable intellectual property (IP). However, even when deployed via black-box APIs, these models remain vulnerable to unauthorized knowledge distillation, which allows adversaries to cheaply extract and replicate model capabilities. To address this issue, anti-distillation (AD) has been proposed to generate defensive outputs that hinder distillation effectiveness, overcoming the limitation of watermarking-based approaches that rely on post-hoc verification. However, existing AD methods based on internal model perturbations struggle to balance anti-distillability and utility (e.g., answer accuracy and naturalness) of reasoning traces, with stronger defenses often causing significant utility loss. To fill this gap, we propose \textbf\underlineSkeleton-\textbf\underlineGuided \textbf\underlineReasoning \textbf\underlineEditing (SGRE), an \textitAnswer-then-Edit framework that performs post-hoc trace modification for anti-distillation. In the answer stage, the teacher model first generates clean reasoning traces, preserving the original reasoning accuracy while enabling more flexible control over trace naturalness. In the editing stage, we draw inspiration from Cognitive Load Theory (CLT) and introduce a three-stage strategy consisting of reasoning skeleton extraction, skeleton graph coarsening, and skeleton verbalization. These operations jointly perturb reasoning structures and augment textual complexity to amplify extraneous load on student models, hindering their acquisition of underlying reasoning patterns. Extensive experiments across diverse LLMs demonstrate that SGRE achieves state-of-the-art performance in reducing distillation effectiveness, while maintaining lossless reasoning accuracy and superior trace naturalness.
[NLP-88] AsymVerify at SemEval-2026 Task 6: Asymmetric Confidence-Gated Verification for Political Evasion Detection ACL2026 ACL SEMEVAL-2026
【速读】： 该论文旨在解决政治性回避回答（political evasion）的检测难题，此类回答虽表面合作却回避具体承诺，导致传统分类方法难以准确识别。其核心解决方案是提出一种置信度门控的非对称验证系统AsymVerify，针对SemEval-2026 Task 6中的三分类任务（明确回复、模棱两可、明确不回复）设计了差异化验证策略：对低置信度预测分别采用降级验证（明确回复/明确不回复 → 模棱两可）或升级验证（模棱两可 → 明确回复），以精准修正边界案例。关键创新在于基于错误集中于模棱两可类边界的观察，构建不对称双验证器架构，并通过置信度门控机制控制额外推理开销，在保持低调用次数（1.48次/样本）的同时显著提升性能——在开发集上相较单次分类实现+17.1点宏平均F1，且升级验证器在所有测试大语言模型（LLM）后端上均带来+6.8至+15.2点的性能增益。
链接: https://arxiv.org/abs/2607.20439
作者: Sebastien Kawada
机构: Kaons 𝑲∗; OpenRouter (OpenRouter); Meta AI (Meta)
类目: Computation and Language (cs.CL)
备注: Accepted to SemEval-2026 Task 6 (CLARITY) at ACL 2026. Team AsymVerify placed 2nd of 41 teams on the official Subtask 1 leaderboard. Task: this https URL . Code: this https URL . Website: this https URL
Abstract:Political evasion is difficult to detect because evasive answers often appear cooperative while avoiding concrete commitment. We present AsymVerify, a confidence-gated verification system for SemEval-2026 Task 6, a three-way classification of Clear Reply, Ambivalent, and Clear Non-Reply responses. AsymVerify scored 0.85 Macro F1 on the evaluation split (D_eval, n=237), placing 2nd out of 41 teams on the official leaderboard. The system first classifies each question-answer pair, then selectively applies downgrade verification (CR/CNR - AMB) or upgrade verification (AMB - CR) to low-confidence predictions. Development analysis shows that errors concentrate at the Ambivalent boundary in both directions, motivating this asymmetric two-verifier design while confidence gating keeps additional inference cost low. On D_dev (n=308), AsymVerify with GLM-4.7 gains +17.1 Macro F1 over single-pass classification at 1.48 calls/example, and the upgrade verifier alone improves every tested LLM backend on D_dev by +6.8 to +15.2 Macro F1 over its single-pass baseline. Code is available at this https URL.
[NLP-89] Preference Tuning as Spectral Update Reorganization
【速读】： 该论文旨在解决生成式 AI（Generative AI）在基于偏好后训练（preference-based post-training）过程中，模型参数更新机制不透明的问题。尽管此类方法通过最终行为表现来评估对齐效果，但其背后驱动行为变化的参数更新路径仍缺乏可解释性。论文的关键解决方案在于从谱结构（spectral structure）角度分析强化学习人类反馈（RLHF）及相关偏好优化所诱导的参数更新，通过分解有效的低秩适配器（LoRA）更新并重新加载其谱成分作为可插拔模块，实现了对偏好驱动更新的隔离、重构与直接干预。研究发现，无论模型架构、优化算法或监督范式如何，这些更新均呈现出一致的“头部-尾部”（head-tail）组织结构：早期出现的紧凑“头部”主导了最终行为的显著偏移，而剩余的异质性“尾部”则贡献较小。进一步实验表明，“头部”承担了可见的行为偏离，而“尾部”在孤立状态下作用微弱；跨运行重构实验更揭示“头部”携带了运行级别的求解器偏差（solver bias）。值得注意的是，仅依赖“头部”的学习虽非无效，却无法恢复完整解空间，尤其在分布外（out-of-distribution）场景下表现不足；而“尾部”单独学习几乎无可见收益，但完整解的恢复又不可或缺。因此，该研究将偏好后训练重新诠释为一种结构化的更新重组过程，而非单一的行为修正，揭示了对齐增益与覆盖损失本质上源于学习到的更新本身的组织方式。
链接: https://arxiv.org/abs/2607.20438
作者: Peiyan Zhang,Haibo Jin,Liying Kang,Haohan Wang
机构: University of Illinois at Urbana-Champaign (伊利诺伊大学厄本那-香槟分校); Hong Kong Polytechnic University (香港理工大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Preference-based post-training is usually understood through endpoint behavior, yet the learned update that produces this behavior remains largely opaque. We study RLHF and related preference optimization through the spectral structure of their induced parameter updates. By decomposing effective LoRA updates and reloading their spectral components as plug-in modules, we turn preference-induced updates into objects that can be isolated, recomposed, and directly intervened on. Across model families, optimization algorithms, and supervision regimes, these updates consistently develop a spectral head–tail organization. A compact head emerges early and carries the dominant endpoint shift, while a heterogeneous residual tail remains. The split is functional rather than merely descriptive. Plug-in intervention shows that the head accounts for the visible behavioral departure from the base model, while the tail is weak in isolation. Cross-run recomposition further shows that mixed adapters follow the source of the head, indicating that the head carries run-level solver bias. This endpoint dominance does not imply learning sufficiency. Head-only learning is non-vacuous but fails to recover the full solution, especially on out-of-distribution behavior. Tail-only learning yields little visible gain, yet the full solution is not recovered without the tail. These findings recast preference post-training as structured update reorganization rather than a monolithic behavioral correction, and suggest that alignment gain and coverage loss are tied to how the learned update itself is organized.
[NLP-90] Routing Subspaces: Auditing Evaluation-to-Deployment Mismatch in Fine-Tuned Language Models
【速读】： 该论文旨在解决生成式AI在安全评估中存在的一致性问题：评估阶段的行为表现可能无法真实反映模型在实际部署环境中的行为，尤其当模型经过微调（fine-tuning）后，其在特定评估提示下的表现看似稳定，但在日常使用提示下仍会表现出不一致甚至风险行为。这种“评估-部署差距”难以通过输出评分识别，因其无法定位问题所在。论文提出的关键解决方案是：基于路径修补（path-patching-informed）的中深度窗口，构建配对激活对比（paired activation contrast），以识别并干预模型内部表征中隐含的评估与部署行为差异。该方法通过在保留提示上的坐标修改，实现对目标行为的精准修正，在十二种模型-行为配置中的十种（八种具备≥120对问答的场景中有六种）成功缩小了评估与部署之间的差距；另一模型进一步验证了定位能力与编辑溯源性，且部署框架下的行为率变化不超过6.1个百分点。然而，两个未被修复的“平坦单元”（均表现为阿谀奉承行为）表明，当模型内嵌差异为高秩结构或被深度启发式遗漏时，单一坐标审计不足以覆盖所有情况。整体而言，该方法是一种针对微调检查点的诊断工具，而非训练阶段的防御机制，亦不保证部署安全性。
链接: https://arxiv.org/abs/2607.20436
作者: Phongsakon Mark Konrad,Toygar Tanyel,Serkan Ayvaz
机构: Centre for Industrial Software, University of Southern Denmark (南丹麦大学工业软件中心); ProMake (美国特拉华州纽瓦克市普罗迈克公司)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Safety evaluations often assume that behavior observed during testing reflects behavior in ordinary use, but fine-tuning can break this assumption. A checkpoint can appear fixed under evaluation-style prompts while the same behavior persists under ordinary-use prompts. Output scores reveal this mismatch but do not locate it. We investigate whether the distinction is encoded in a stable internal site and introduce an approach that fits a paired activation contrast at a path-patching-informed mid-depth window, then modifies the resulting coordinate on held-out prompts. The intervention closes the evaluation-to-deployment gap in ten of twelve model–behavior settings (six of the eight settings with n\geq120 paired questions) across four full-matrix instruction-tuned model instances; a fifth model supports localization and edit-provenance checks, and deployment-framed rates change by at most 6.1 pp. The two flat cells, both sycophancy, indicate that a single-coordinate audit is not sufficient when the installed distinction is higher-rank or missed by the depth heuristic. The audit is a diagnostic for fine-tuned checkpoints, not a training-time defense or a guarantee of deployment safety.
[NLP-91] Making Open-Source Text LLM Watermarks Durable Against Merging
【速读】： 该论文旨在解决开源大语言模型（Open-source LLMs, OSMs）中文本水印在经历后训练修改（如模型合并，model merging）时易被移除的问题。现有方法将水印嵌入模型权重，但模型合并等常见操作会显著削弱甚至完全消除水印，导致溯源失效。其解决方案的关键在于提出一种名为“合并对抗训练”（Merge-Adversarial Training）的对抗训练算法，通过在训练过程中模拟后续模型合并场景，使水印信息能够被深度注入并稳定保留在模型权重中，从而增强水印对模型合并的鲁棒性。实验表明，该方法在多种真实合并场景下均显著优于现有基线（如在SLERP指标上平均提升25个百分点，最高达+51 pp），同时保持了模型的下游任务性能。研究首次系统评估了水印在典型合并应用中的耐用性，证实对抗训练是提升水印对后训练修改抗性的可靠策略。
链接: https://arxiv.org/abs/2607.20435
作者: Luisa Scharff,Thibaud Gloaguen,Robin Staab,Martin Vechev
机构: ETH Zurich (苏黎世联邦理工学院)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Open-source LLMs (OSMs)arereaching near state-of-the-art performance, prompting prior works to trace the text they generate by embedding text watermarking algorithms directly into their weights. Yet, OSMs are subject to post-training modifications, which has been shown to remove the watermark. Model merging in particular, a prominent method used for combining expert knowledge and preventing catastrophic forgetting, strongly removes such OSM watermarks. A key question is how to enable OSM watermarks that survive subsequent merging. In this work, we show for the first time how to design an OSM watermark that is durable against model merging. We propose Merge-Adversarial Training, an adversarial training algorithm to distill text watermarks into model weights while being robust to subsequent model merging. Our approach consistently outperforms all baselines (e.g. with SLERP up to +51 percentage points (pp) TPR@1%FPR with +25 pp on average) while preserving downstream capabilities. We also for the first time evaluate OSM watermarks against realistic merge scenarios, representing common use-cases such as combining expert capabilities or preventing catastrophic forgetting, and with 3 prominent merging algorithms. More broadly, our findings suggest that adversarial training is a reliable approach for increasing OSM watermark durability against post-training modifications.
[NLP-92] Break Through the Compression Bottleneck: From Theory to Practice
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在参数规模持续增长背景下，模型压缩技术面临的性能退化瓶颈问题。现有压缩方法如低秩分解（Low-rank Decomposition）与量化（Quantization）虽能有效降低计算与内存开销，但当二者联合使用时，性能下降显著，制约了高比例压缩的实现。传统观点认为这两种方法具有正交性（orthogonal），即其误差可独立叠加，但本文首次从数学上证明低秩分解与量化并非正交，二者协同作用会引入非线性的累积误差，导致性能严重退化。为应对这一关键挑战，论文提出一种新型融合策略——对角粘合方法（Diagonal Adhesive Method, DAM），通过优化参数空间中的耦合关系，有效缓解复合压缩带来的性能损失。该研究不仅揭示了压缩方法间内在交互机制的本质，还为高效、高保真的模型压缩提供了理论依据与实践方案。
链接: https://arxiv.org/abs/2607.20434
作者: Xiusheng Huang,Lu Wang,Yequan Wang,Jun Zhao,Kang Liu
机构: Institute of Automation, Chinese Academy of Sciences (中国科学院自动化研究所); University of Chinese Academy of Sciences (中国科学院大学); Beijing Academy of Artificial Intelligence (北京人工智能研究院); Ritzz-AI
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: 18 pages, 3 figures,
Abstract:As the parameter size of language models continues to grow, effective model compression is required to reduce their computational and memory overhead. Existing compression methods suffer from bottleneck issues: when the compression ratio is increased, performance degrades significantly. Low-rank decomposition and quantization are two prominent compression methods that have been proven to significantly reduce the computational and memory requirements of Large Language Models (LLMs) while maintaining model accuracy. Evidently, combining these two methods will break through the existing compression bottleneck. However, how these two methods interact when combined remains a critical question for developers, as many assume they are orthogonal, meaning their combination would not introduce additional errors beyond those independently introduced by each method. This paper provides the first mathematical proof that low-rank decomposition and quantization are non-orthogonal. We validate these findings through a series of experiments on large language models. Our results demonstrate that these methods are non-orthogonal, and their combination leads to significant performance degradation. Importantly, we propose a novel approach Diagonal Adhesive Method (DAM), which can effectively combine the two methods and mitigate the performance loss. Our research provides deep insights into model compression and lays a solid theoretical and experimental foundation for future related studies.
[NLP-93] Moir: Let the Model Direct Its Own Story for Robust Cross-Domain Knowledge Editing
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在知识编辑过程中出现的非对称能力退化问题，即在保持百科类知识召回能力的同时，数学推理与程序生成等核心推理能力显著下降。其关键挑战在于现有基于协方差的知识编辑方法依赖静态外部语料库（如维基百科或预训练数据混合），无法捕捉模型在后训练阶段（如监督微调SFT和直接偏好优化DPO）中形成的动态操作分布（operative distribution），从而导致编辑过程中的分布失配（distributional mismatch）。为解决此问题，作者提出Moir方法，其核心创新在于不依赖外部数据，而是通过从模型自身的解码分布中采样，直接估计模型内部的保留协方差矩阵 $ C $。该方法通过以单一随机词汇标记作为生成种子，规避了指令遵循模板对输出分布的主导作用，从而更全面地暴露模型所内化的大规模潜在子空间。Moir可无缝集成至任意基于协方差的编辑器中，且无需访问模型的预训练或后训练数据，具有高度实用性。实验结果表明，在OLMo-2、Llama-3.1和Qwen-3（7-8B）等多个模型上，无论采用MEMIT或AlphaEdit方法，以及批处理或顺序编辑场景，Moir均显著提升了模型在脆弱任务领域的知识保留能力，尤其在对Qwen3-8B进行20,000次批量AlphaEdit编辑后，仍能保持79.9%的GSM8K准确率，远超基于维基百科基线的10.9%。这表明，将保留分布与模型的实际操作分布对齐是实现无损编辑的关键，而模型自身生成的分布可能是部署系统中最可及的分布源。
链接: https://arxiv.org/abs/2607.20433
作者: Jea Kwon,Jiwon Kim,Dong-kyum Kim,Meeyoung Cha
机构: Max Planck Institute for Security and Privacy (MPI-SP)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:While language models remain frozen at their training state, the world evolves continuously. Knowledge editing has emerged as a key alternative to full retraining, but its deployment is bottlenecked by the erosion of core capabilities: mathematical and programmatic reasoning collapse while encyclopedic recall remains intact. We trace this asymmetric degradation to a distributional mismatch. Covariance-based editors preserve only the subspaces spanned by their reference corpus, but fail to capture the operative distribution shaped by post-training such as SFT and DPO. Static external corpora, including Wikipedia and even the original pretraining mixture, cannot recover this shifted manifold. We propose Moir, which estimates the preservation covariance C directly from the model itself by sampling from its own decoding distribution. Seeding generation with a single random vocabulary token bypasses the instruction-following templates that otherwise dominate sampled outputs, exposing the broader subspaces the model has internalized. Moir requires no external data and serves as a drop-in component for any covariance-based editor, a practical advantage given that the pre- and post-training corpora of most modern LLMs are not publicly accessible. Across OLMo-2, Llama-3.1, and Qwen-3 (7-8B), under both MEMIT and AlphaEdit and in batch and sequential regimes, Moir consistently extends preservation in the most vulnerable domains, most strikingly on Qwen3-8B after 20,000 AlphaEdit batch edits, it retains 79.9% GSM8K accuracy compared to 10.9% with the Wikipedia baseline. These results suggest that aligning the preservation distribution with the model’s operative distribution is a key factor in non-destructive editing, and that the model itself may be the most accessible source of that distribution for deployed systems.
[NLP-94] Position: Natural Language Should Not Fully Replace Formal Languages ICML2026
【速读】： 该论文旨在解决当前关于生成式 AI 中自然语言是否可完全替代形式化语言（如编程语言）的争议性问题。其核心关切在于，尽管大语言模型的兴起推动了以自然语言作为软件设计等任务的主要输入方式，但这一观点忽视了自然语言在开放语境下为“不明确性”（underspecification）优化的本质特征。论文的关键解决方案在于提出一个以任务特异性（task specificity）为核心的理论框架，将其定义为：在给定用户具体需求时，对输出空间（如所有可能图像）中不确定性的信息论缩减程度。通过建立并证明“特异性交叉定理”（specificity crossover theorem），论文揭示存在一个临界阈值——当任务要求超过该阈值时，将形式化需求转化为自然语言的表达成本将高于直接采用形式化语言进行规范的成本。基于跨模态案例研究（包括图像生成、代码合成与音频制作），论文进一步验证了自然语言在低特异性任务中具有优势，而形式语言则在高特异性任务中表现更优。最终结论强调，自然语言与形式语言是互补工具，应推动构建能够支持用户在特异性连续谱上灵活切换的混合系统。
链接: https://arxiv.org/abs/2607.20432
作者: Eitan Wagner,Elisha Rosensweig,Omri Abend
机构: 未知
类目: Computation and Language (cs.CL); Programming Languages (cs.PL)
备注: To be published in ICML 2026 (position track)
Abstract:Recent advances in large language models and their widespread adoption have prompted claims that natural language could entirely replace formal languages, such as programming languages for software design. In this position paper, we argue that this perspective overlooks fundamental linguistic properties of natural language, specifically that it is optimized for underspecification in open-ended contexts. We introduce a formal framework centered on task specificity, defining it as the information-theoretic reduction of uncertainty in an output space – such as all possible images – given a user’s specific requirements. We prove a specificity crossover theorem, showing the existence of a threshold beyond which the cost to express formal requirements into natural language exceeds the cost of direct formal specification. By analyzing case studies across modalities, such as image generation, code synthesis, and audio production, we demonstrate that natural language excels at low specificity tasks, while formal languages are advantageous on tasks with stricter requirements. We conclude that natural and formal languages are complementary tools and advocate the development of hybrid systems that allow users to move across the specificity spectrum.
[NLP-95] LLM -INSTRUCT at UZH Shared Task 2026: Constraint-Aware Retrieval and Selective Debate for Parag raph-Level Argument Mining ACL2026
【速读】： 该论文旨在解决联合国及教科文组织决议文本中段落级论据挖掘（paragraph-level argument mining）的复杂任务，具体包括三方面挑战：段落类型分类、141个官方标签的子集预测，以及在严格JSON模式约束下的有向关系预测。由于仅允许使用参数量不超过8B的开源大模型，且需在结构化输出上保持高精度与一致性，传统生成式方法面临决策空间过大、输出格式易出错等问题。其解决方案的关键在于将任务建模为约束结构化预测（constrained structured prediction），通过多阶段策略优化：首先利用元数据感知的密集检索（metadata-aware dense retrieval）缩小候选标签空间；随后采用逐维度上限控制的约束解码（per-dimension caps）确保输出符合预定义的标签分布；仅对置信度较低的样本引入三智能体辩论分支进行深度推理；最后通过输出模式验证机制保障最终结果满足严格的JSON schema要求。实验表明，该方法在官方排行榜上取得整体第一、F1指标第一的成绩，并显著提升了模型鲁棒性。核心启示是：在生成前有效压缩决策空间，可同时提升准确率与提交结果的稳定性。
链接: https://arxiv.org/abs/2607.20430
作者: Phuong Huu Vu Tran,Long Minh Vo,Son Nguyen Minh Le,Hoang Van
机构: Vietnamese-German University (越南-德国大学); RMIT University Vietnam (RMIT越南大学); VANGIA INNOVATIONS (万佳创新)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted to the 13th Workshop on Argument Mining (ArgMining 2026) at ACL 2026
Abstract:We present LLM-INSTRUCT, the winning system for the UZH Shared Task at ArgMining 2026 on paragraph-level argument mining in UN and UNESCO resolutions. The task requires paragraph-type classification, prediction of a subset of 141 official tags, and directed relation prediction under a strict JSON schema setting using only open-weight models up to 8B parameters. We frame the task as constrained structured prediction. The system first narrows the candidate tag space with metadata-aware dense retrieval, then applies constrained decoding with per-dimension caps, escalates only uncertain cases to a three-agent debate branch, and finally validates the output schema. On the official leaderboard, LLM-INSTRUCT ranked 1st overall, with 1st in F1 and 5th in LLM-as-a-Judge. During development, our configuration search further improved Task 1b Micro-F1 from 35.83% to 40.08% while keeping the internal Task 2 score at 4.421. The main lesson is simple: reducing the decision space before generation improves both accuracy and submission robustness. Our code and supporting scripts are publicly available at: this https URL
[NLP-96] More Is Not More: What Matters for Diversity in LLM Opinions?
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）在开放式任务中生成意见时存在的系统性观点同质化问题，即模型输出倾向于趋同，难以真实模拟多元化的公众意见。其解决方案的关键在于通过设计一个因子实验（factorial experiment），系统性地分离并评估两大核心干预维度：输入条件化（以人物角色深度，persona depth，为操作化指标）与交互架构（interaction architecture）。研究发现，过度细化人物角色信息并不会持续提升多样性，初始的角色条件化已实现大部分增益，而进一步添加人口统计学细节反而可能降低某些模型的多样性；不同交互架构探索的是高度非重叠的意见区域，组合使用多种架构可获得更广泛的意见覆盖，优于单一架构优化；此外，诸如提高采样温度或添加多样性指令等低成本方法效果微弱，远不及结构化干预有效。因此，论文揭示了多样性并非单一维度扩展的结果，而是高度依赖于干预措施的结构性形式及其组合方式，为提升生成式AI（Generative AI）在社会模拟任务中的表现提供了科学依据与实践指导。
链接: https://arxiv.org/abs/2607.20429
作者: Qiyang Yao
机构: New York University (纽约大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models are increasingly used to simulate diverse human opinions in open-ended tasks such as synthetic surveys, focus group modeling, and public opinion prediction. However, LLM outputs exhibit systematic opinion homogenization. Practitioners have explored various interventions to increase diversity, but the landscape remains fragmented: different methods are evaluated in isolation with incomparable metrics, and in practice they are typically deployed and upgraded simultaneously, making it difficult to attribute gains to specific components. To advance a more scientific understanding of LLM output diversity, we design a factorial experiment that separates two primary intervention dimensions: input conditioning (operationalized through persona depth) and interaction architecture. We evaluate all conditions on 100 real-user open-ended questions across 7 models, measuring diversity with multiple complementary metrics. Our findings challenge several common assumptions. First, more persona detail does not monotonically increase diversity. The initial step of persona conditioning already captures the majority of the gain, while further elaboration with demographic detail does not consistently improve and can reduce diversity on some models. Second, rather than seeking a single best interaction architecture, we find that different architectures explore largely non-overlapping opinion regions. Combining multiple architectures yields broader coverage than optimizing any one. Third, commonly attempted low-cost alternatives such as raising sampling temperature and adding diversity instructions produce negligible effects compared to structured interventions. Overall, our work demonstrates that diversity is not a product of scaling along any single dimension, but is highly sensitive to the structural form and combination of interventions.
[NLP-97] Is MoE Routing a Huffman Code? Discovering the Frequency-Diversity Law in Chain-of-Thought
【速读】： 该论文旨在解决混合专家（Mixture-of-Experts, MoE）架构中路由机制的内在逻辑不透明问题，尤其是其在实际运行中如何高效分配稀疏专家资源以应对不同频率与复杂度的任务。其核心发现是：当前先进模型（如Phi-3.5-MoE和Gemma-4-27B-A4B）的路由行为本质上遵循霍夫曼编码（Huffman Coding）原理——即高频词元采用低复杂度、高密度的专家路径，而低频、复杂的链式思维（chain-of-thought）任务则触发高多样性专家委员会。然而，研究揭示了Qwen3.5-35B-A3B中存在的关键冗余陷阱：当有效稀疏度（k/E_eff）过低时，负载均衡策略会人为引入功能冗余，掩盖了原本存在的霍夫曼效率信号。为此，作者提出“子集差异剪枝”（Subset Difference Pruning）这一精准剪枝策略，通过移除功能性重复的专家组合，恢复模型的潜在霍夫曼效率。实验表明，该方法不仅未损害推理能力，反而促使路由逻辑收敛为更紧凑、高密度的路径结构。研究进一步指出，下一代MoE应超越强制负载均衡，转向最小描述长度（Minimum Description Length, MDL）最优性设计，即对高频信息分配短编码、对低频复杂信息分配长且多样化的编码，从而将路由机制从经验性启发式转变为具有理论基础的信息压缩引擎。
链接: https://arxiv.org/abs/2607.20427
作者: Ching-Chieh Tsao,Zhuoyi Lin,Wenya Wang
机构: 未知
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Information Theory (cs.IT)
备注: 20 pages, 20 figures
Abstract:Mixture-of-Experts architectures have revolutionized scaling, yet the underlying logic of their routing remains a black box. In this paper, we uncover a fundamental governing principle: MoE routing is not merely selection, but a manifestation of Huffman Coding. We introduce the Frequency-Diversity Law, revealing that state-of-the-art models, such as Phi-3.5-MoE and Gemma-4-27B-A4B, spontaneously act as information-theoretic engines. These models allocate sparse expert resources for common tokens while invoking high-diversity expert committees for rare, complex tasks found in chain-of-thought trajectories. However, we identify a critical redundancy trap in Qwen3.5-35B-A3B: when effective sparsity (k/E_eff) is sufficiently low, load-balancing inadvertently imposes functional redundancy, masking the underlying Huffman efficiency signal. To bridge this gap, we propose Subset Difference Pruning, a surgical strategy to eliminate functional duplicates. We demonstrate that pruning does not degrade reasoning; instead, it unleashes the model’s latent Huffman efficiency, forcing the logic to collapse into streamlined, high-density paths. Our findings suggest that the next generation of MoEs should move beyond forced load-balancing toward Minimum Description Length (MDL) optimality, assigning shorter expert-routing codes to high-frequency information and longer, more diverse codes to low-frequency information, thereby transforming routing from a heuristic into a principled compression engine.
[NLP-98] Knowledge Injection Exists in MoE? Exploring Expert-Aware Contrast Decoding in MoE for Mitigating LLM sHallucinations ACL2
【速读】： 该论文旨在解决大语言模型（Large Language Models, LLMs）中存在的幻觉（hallucination）问题，尤其针对现有缓解方法在模型内部知识修改能力有限以及跨领域泛化性能较差的缺陷。现有方法如提示工程（prompt engineering）和模型优化虽有一定效果，但难以有效干预模型深层知识表征；而对比解码（contrastive decoding）虽利用了模型层间差异来抑制幻觉，但其研究仅局限于基于Transformer架构的模型（如GPT），忽略了其他高效架构——混合专家模型（Mixture-of-Experts, MoE）——的可能性。由于MoE在结构上不同于传统Transformer，作者通过实证研究发现：在共享专家的MoE中并不存在类似Transformer的层间差异；然而，在不同MoE模型中，高层数层对事实性与非事实性输出表现出显著不同的专家激活模式。基于此，论文提出一种专家感知的自适应对比解码方法（Expert-Aware Adaptive Contrast Decoding, EAACD），其核心在于利用MoE高层数层中专家激活模式的可靠性差异进行幻觉抑制。具体而言，EAACD将高层数层的专家按置信度与一致性划分为高可靠性组与多个低可靠性组，并通过对比高可靠性组与各低可靠性组的预测结果，对原始生成结果进行校准。为进一步强化对比信号，EAACD引入注意力机制与掩码策略，放大低可靠性专家产生的幻觉内容，从而提供更强的负向参考。实验表明，EAACD在四个数据集上的问答任务中均显著优于所有基线方法，验证了其有效性与泛化能力。
链接: https://arxiv.org/abs/2607.20426
作者: Xinyue Fang,Zhiliang Tian,Zhen Huang,Ziyi Pan,Zhihua Wen,Xi Wang,Quntian Fang,Dongsheng Li
机构: National University of Defense Technology (国防科技大学)
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
备注: Accepted by ACL2
Abstract:Existing LLM hallucination mitigation methods, including prompt engineering and model optimization, either hardly alter models’internal knowledge or have poor cross-domain generalization. Contrastive decoding mitigates hallucinations by using layer-wise differences in LLMs. However, prior studies only explore transformer-based models (e.g., GPT), ignoring other effective frameworks like mixture-of-experts (MoE) models. Since MoE alters the traditional transformer architecture, we conduct empirical studies to investigate whether similar layer-wise differences exist in MoEs. Our results show that they do not exist in MoE with shared experts; nevertheless, across different MoEs, higher layers exhibit distinct expert activation patterns between factual and non-factual outputs. Building on these, we propose EAACD, an expert-aware adaptive contrast decoding that uses expert differences in MoE’s higher layers to mitigate hallucinations on QA tasks. EAACD splits high-layer experts into a higher-reliability group and several lower-reliability groups based on their confidence and consistency. It contrasts the higher-reliability group’s prediction with each lower-reliability group’s prediction to calibrate the model’s original predictions. To strengthen this contrast, EAACD amplifies hallucinations from lower-reliability experts via attention and masking to provide stronger negative references. EAACD outperforms all baselines on four datasets.
[NLP-99] What is Good? Extracting and Testing Implicit Theories of Literary Quality from LLM Reasoning Traces
【速读】： 该论文旨在解决生成式 AI（Generative AI）在评估文学质量时所依赖的内在判断标准不明确的问题，尤其是在缺乏清晰可解释性的情况下，如何理解大语言模型（LLM）对写作质量的评判逻辑。其核心解决方案的关键在于通过两阶段实证研究，揭示模型在评估文本质量时的隐含理论与敏感特征：首先构建涵盖六个质量层级的真实文本基准集，并基于模型推理轨迹提取其评价标准；其次通过对经典文本进行系统性降级操作（如词汇简化、节奏平滑、意象移除等），量化不同语言层面的退化对质量评分的影响。研究发现，模型更重视作者意图、文风独特性与结构复杂性，而非单纯的语言正确性或词汇丰富度，且对结构和语调的退化尤为敏感，显示出对文本整体性与作者特异性的高度依赖。这一发现为自动化写作反馈系统与计算美学研究提供了可解释的理论基础。
链接: https://arxiv.org/abs/2607.20425
作者: Birger Moëll
机构: Uppsala University (乌普萨拉大学)
类目: Computation and Language (cs.CL)
备注:
Abstract:What makes writing “good” remains a persistent question in literary studies and computational linguistics. We present a two-study investigation of how reasoning-enabled LLMs evaluate literary quality. In Study 1, we construct a benchmark of 30 real texts spanning six quality tiers, from canonical literature to anonymous forum posts, and extract the model’s implicit theory of quality from its reasoning traces. Across five DeepSeek replications, the model achieves 79.3% mean tier-classification accuracy. The traces reveal a consistent stated theory: the model values intentionality over correctness, prioritizing craft, depth, and distinctive voice. A familiarity experiment with style-matched but unrecognizable passages suggests that source recognition may inflate scores, although this is confounded by genuine quality differences between canonical originals and researcher-written pastiches. In Study 2, we probe this theory through systematic degradation of five canonical prose passages. We apply six manipulations - vocabulary simplification, rhythm flattening, imagery removal, voice genericization, structure simplification, and combined degradation - and reevaluate each version. Vocabulary simplification causes the smallest quality loss (0.41 +/- 0.46 points), far below structure (2.78) or voice (2.34) loss. Combined degradation is devastating (-5.64) but subadditive. An exploratory comparison with Qwen QwQ shows the same broad qualitative pattern. Together, these studies suggest that LLM judgments of writing quality are holistic, author-specific, and more sensitive to structural than lexical features, with implications for automated writing feedback and computational aesthetics. Subjects: Computation and Language (cs.CL) Cite as: arXiv:2607.20425 [cs.CL] (or arXiv:2607.20425v1 [cs.CL] for this version) https://doi.org/10.48550/arXiv.2607.20425 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Birger Moell [view email] [v1] Tue, 7 Apr 2026 13:29:12 UTC (216 KB)
信息检索
[IR-0] Diffusion Language Model for Recommendation
链接: https://arxiv.org/abs/2607.21519
作者: Chengyi Liu,Yongqi Zhou,Junwei Pan,Zhixiang Feng,Chengguo Yin,Haijie Gu,Jie Jiang,Yinghao Liu,Yujuan Ding,Qing Li,Wenqi Fan
类目: Information Retrieval (cs.IR)
备注: 30 pages, 9 figures
Abstract:Large language model (LLM)-empowered recommender systems have emerged as a promising paradigm for generative recommendation, leveraging their strong semantic reasoning and generative capacity to model complex, diverse user preferences. However, most existing approaches rely on an autoregressive paradigm that is suboptimal for recommendation. The next-token objective emphasizes sequential order rather than the structural inter-item dependencies underlying user preferences. In addition, prefix-constrained generation restricts bidirectional context and commits to left-to-right decoding, causing early errors to accumulate without correction. Inspired by the success of diffusion language models, we propose \textbfDLMRec, a discrete diffusion language model tailored for recommendation that offers a compelling alternative to autoregressive generation. Specifically, DLMRec introduces three key components to bridge diffusion language modeling with recommendation. First, a collaborative-aware stochastic tokenizer encodes multi-hop collaborative signals into expressive discrete tokens compatible with diffusion modeling. Second, a curriculum-driven training strategy aligns the denoising process with preference recovery through progressive item- and token-level learning. Third, a stability-aware voting mechanism aggregates iterative predictions to improve generation consistency and robustness.
[IR-1] ransparent by Design Usable in Practice? A Formative Usability Study of a Conversational Product Advisor
链接: https://arxiv.org/abs/2607.21513
作者: Kevin Schott,Dagmar Kern,Daniel Hienert
类目: Human-Computer Interaction (cs.HC); Information Retrieval (cs.IR)
备注:
Abstract:Large language models can make conversational product advisors fluent but opaque. If they hide the logic behind a ranking and the evidence for a recommendation inside natural-language replies, they challenge users’ ability to understand, trust, and steer the results. One response is to build transparency into the advisor. We report a formative, moderated think-aloud usability study of one such system: a chatbot for laptop search with constrained natural-language generation, an on-demand ranking explanation, and a comparison feature. Seven participants completed three laptop-search tasks and reported post-task usability measures. We coded their sessions into severity-rated usability problems. Ease and satisfaction during the tasks were high, but two findings stand out. First, transparency by design did not guarantee understanding: several participants valued the ranking explanation in principle, yet it caused the most severe problem. Second, participants valued the effort the advisor saved, but some wanted additional direct-manipulation controls. We contribute a severity-prioritized set of usability problems and design implications for human-centered conversational product advisors.
[IR-2] Agent ic Context Management: Solving Agent Memory and Cost by Treating Them as Lifecycle and Architecture Problems
链接: https://arxiv.org/abs/2607.21503
作者: Gaurav Dadhich
类目: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注: 23 pages, 6 figures, 4 tables. Evaluation harness and study data: this http URL
Abstract:Production AI agents’ failures are less often due to an inability to reason well and more often because they cannot manage what is in their reasoning context: conversation histories, large prompts, large tool definitions, and ballooning tool outputs. Agents drown in their own accumulating history while paying a token cost that grows every turn, producing missing recalls within and across conversations. The incumbent response treats this as a storage-and-retrieval problem. We argue that framing is too narrow. Actively managing what an agent holds in mind is a lifecycle, not merely a store: it spans deciding what to remember, extracting and structuring it, choosing the right store per data type, consolidating and forgetting while preserving provenance, deciding what is relevant now, anticipating what is needed next, and compacting context to a budget without losing what matters. In serious production this operates not over a single user but across an organizational scope hierarchy. We name this discipline Agentic Context Management (ACM) and decompose it into five primitives: architecting, ingesting, scoping, anticipating, and compacting consolidation. We then make the economic case: naive context accumulation grows token cost quadratically in conversation length, crude summarization buys linear cost at the price of an accuracy cliff, and only validated compaction achieves linear cost with preserved fidelity. We describe a reference implementation, Maximem Synap, that realizes the five primitives as a multi-tenant service and reports 92% on LongMemEval and 93.2% on LoCoMo under the configuration detailed in Section 6. We close with dimensions existing benchmarks do not yet capture, latency, token efficiency, and context-rot resistance, and the frontier of decision-level and organization-level context the category points toward.
[IR-3] SHIFT: Self-reconstruction Harnesses Implicit Fine-grained Thinking for Retrieval
链接: https://arxiv.org/abs/2607.21333
作者: Yuxiao Luo,Da Li,Mingjie Zhang,Zhentao He,Shikun Zhang,Wei Ye
类目: Information Retrieval (cs.IR)
备注:
Abstract:LLM-based retrievers have become a fundamental component of modern information retrieval systems. The paradigm of “rewrite-then-retriev” introduces explicit reasoning before retrieval. In addition, implicit-reasoning retrievers such as GIRCSE and LaSER improve efficiency by replacing explicit reasoning with soft tokens. Although these methods demonstrated competitive performance on reasoning-intensive retrieval benchmarks, they struggle to address the mismatch between the objectives of retrieval and generation. In this work, we propose SHIFT ( \textbfS elf-reconstruction \textbfH arnesses \textbfI mplicit \textbfF ine-grained \textbfT hinking for Retrieval), a retrieval training framework based on LLMs. Firstly, we transfer LLMs into reasoning-efficient retrievers with residual projection and task-oriented bidirectional attention aggregation in the latent space. Secondly, we alleviate the mismatch between contrastive learning and implicit reasoning using fine-grained next-token-prediction-based reconstruction. Extensive experiments on reasoning-intensive retrieval benchmarks show that SHIFT consistently outperforms other widely used retrievers. We also carried out a detailed analysis to illustrate how our method works.
[IR-4] Bridging the Structural Gap: Adapting Autoregressive Generation for Recommendation
链接: https://arxiv.org/abs/2607.21028
作者: Junchao Zeng,Junzhang Zhu,Junyang Chen,Yudong Li,Wei Liu,Chengxiang Zhuo,Zang Li
类目: Information Retrieval (cs.IR)
备注: 14 pages, 15 figures
Abstract:Generative Recommendation (GR) has emerged as a new paradigm for sequential recommendation, in which a representative line of work encodes items into hierarchical semantic IDs via residual quantization and predicts the IDs token by token. However, this generative formulation still exhibits structural gaps with respect to the recommendation task: flattening multi-token IDs into a single sequence destroys item-level structure, and the inconsistency between training and inference over a hierarchical codebook gives rise to semantic drift. To bridge these two gaps, we propose BARGE, which employs Item Context-Aware Attention (ICA) to restore item-level structure during encoding, and Hierarchical Path Reranking (HPR) together with Dual-Path Decoding (DPD) to suppress semantic drift from two complementary angles during decoding. Extensive experiments and analytical studies on public benchmarks and a large-scale offline test demonstrate that BARGE achieves superior recommendation performance. An online A/B test on a Tencent platform yields improvements of 0.60% in click-through rate, 1.34% in click unique visitors, and 1.70% in total reading time, confirming the practical value of BARGE in industrial-scale recommendation.
[IR-5] Fast and Efficient Approximate Nearest Neighbor Search for High-Dimensional LLM Embeddings
链接: https://arxiv.org/abs/2607.20957
作者: Nico Hezel,Kai Uwe Barthel,Bruno Schilling,Konstantin Schall,Andre Moelle,Klaus Jung
类目: Information Retrieval (cs.IR)
备注:
Abstract:The annual SISAP Indexing Challenge benchmarks Approximate Nearest Neighbor Search (ANNS) algorithms under rigorous constraints. This paper presents our submissions for the 2026 edition, addressing both k -Nearest Neighbor Graph (kNNG) construction on 1024-dimensional BGE-M3 embeddings (Task 1) and Maximum Inner Product Search (MIPS) on unnormalized Llama-3.2-8B features (Task 2). To optimize construction speed, we utilize Equi-Voronoi Polytopes (EVP) for efficient quantization, supplemented by targeted reranking strategies to maintain high recall. For MIPS, we transform the asymmetric inner product problem into a Euclidean search space via dimensionality augmentation. To reduce query latency and optimize memory access, we introduce a 1D presorting mechanism via Fast Linear Assignment Sorting (FLAS) prior to graph construction. This significantly improves spatial locality and cache hit rates during subsequent graph traversal. Source Code: this https URL Subjects: Information Retrieval (cs.IR) Cite as: arXiv:2607.20957 [cs.IR] (or arXiv:2607.20957v1 [cs.IR] for this version) https://doi.org/10.48550/arXiv.2607.20957 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[IR-6] Controllable and Content-Based Recommendations
链接: https://arxiv.org/abs/2607.20938
作者: Fırat Öncel,Jihoon Jeong,Emiliano Penaloza,Mirco Ravanelli,Laurent Charlin,Cem Subakan
类目: Information Retrieval (cs.IR)
备注: Under review
Abstract:Traditional recommendation systems rely on latent (dense) representations, making them difficult to interpret and control. We propose the Controllable and Content-Based Recommendations (CCBR) framework, which builds its recommendations from textual user profile representations. CCBR plugs into collaborative filtering models and introduces controllability via text bottlenecks. We show that CCBR enables text-based and multimodal interventions, allowing users to steer the model towards the directions they prefer. Different from existing controllable recommendation systems, CCBR infers the text summaries directly from item contents (images, audio or video). Across image-, audio-, and video-based datasets, we demonstrate that the proposed framework obtains competitive model performance with standard (latent-representation) models while providing controllable model summaries via text. The model also outperforms TEARS, a recent baseline for controllable recommendation systems. Through systematic interventions, we demonstrate the efficacy of the user steering mechanism.
[IR-7] LO-FAR: A Cost-Aware Local Filter for Sparse Feature Ranking in Industrial Ad Recommendation
链接: https://arxiv.org/abs/2607.20873
作者: Egemen Erbayat,Luis Duque,Sohini Roychowdhury,Mohammad Amin,Srihari Reddy
类目: Information Retrieval (cs.IR)
备注:
Abstract:Industrial ad recommendation models rely heavily on sparse, high-cardinality ID-list features that encode user histories and contextual identifiers. Each is backed by a dedicated embedding table, so these features dominate storage, training, and serving cost and must be revisited as traffic and downstream models evolve. Therefore, sparse feature ranking is not just an offline modeling problem but also a recurrent systems decision limited by compute budgets and iteration cadence. We present Localized Feature Ranking (LO-FAR), a CPU-only, model-agnostic workflow that ranks each candidate feature from its stand-alone held-out predictive signal using lightweight local estimators rather than the GPU-bound retraining loops of permutation- and stochastic-gate-based methods. On a production dataset of more than one million logged interactions and 475 sparse ID-list features, LO-FAR completes ranking in approximately two CPU-hours and preserves downstream Normalized Entropy gains on CTR and CVR tasks that are competitive with shuffle-based importance, Binary Stochastic Neurons, and a coverage-based heuristic across budgets of 100–400 retained features. The contribution is a deployable workflow showing that, when cost and turnaround constraints are binding, a simple local filter can be a practical production choice over heavier interaction-aware alternatives.
[IR-8] Probabilistic Residual Learning for Online Recommendations RECSYS2026
链接: https://arxiv.org/abs/2607.20863
作者: Wenyuan Wang,Yusong Zhao,Zihao Xu,Hengyi Wang,Qi Xu,Zhigang Hua,Yan Xie,Yi Wang,Zihao Zhao,Bo Long,Chengzhi Mao,Shuang Yang,Hengguan Huang,Hao Wang
类目: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted at the 20th ACM Conference on Recommender Systems (RecSys 2026)
Abstract:Modern recommender systems are typically based on deep learning (DL) models, where a dense encoder learns representations of users and items. As a result, these systems often suffer from the black-box nature and computational complexity of the underlying models, making it difficult to systematically enhance their recommendation capabilities. To address this problem, we propose Probabilistic Residual Learning (PRL), a causal Bayesian recommendation model that models the residual between ground-truth and base predictions, enabling targeted refinement of existing systems. Specifically, PRL (1) probabilistically groups users for localized residual modeling, (2) models domain-level confounders that influence user and item representations, and (3) aggregates cluster-specific residual predictions over the confounders using do-calculus. Experiments demonstrate that our plug-and-play PRL is compatible with various base deep learning recommender systems, improving their performance while automatically discovering meaningful user clusters.
[IR-9] Cardinality-Decomposed Loss: Matching Training Objectives to Relation Structure in Heterogeneous Recommendation Graphs
链接: https://arxiv.org/abs/2607.20737
作者: Parul Maheshwari,Amulya Paruchuri,Yiqing Zou,Alireza Sahami Shirazi,Farhad Farahani,Prakhar Mehrotra
类目: Machine Learning (cs.LG); Information Retrieval (cs.IR)
备注:
Abstract:Graph Neural Networks trained on heterogenous bipartite graphs form a common basis in recommendation systems. These graphs often express relations that vary in cardinality, for example, user-item preferences are one-to-many and user-attribute features are one-to-one. Traditionally, a unique loss function is applied for all of the network components which is often Bayesian Personalized Ranking (BPR). While BPR works well for the recommendation task, we find that it causes attribute embeddings to collapse to near-random geometry – a silent failure that leaves standard ranking metrics largely unaffected and therefore invisible to conventional evaluation. This in turn pollutes user node embeddings, which are shaped by both edge types simultaneously, hurting downstream tasks like personalization, segmentation, etc. Here we propose a Cardinality-Decomposed Loss (CDL) that combines both Cross Entropy (CE) and BPR to enable the model to collectively optimize for relations across cardinalities. We confirm this CE-BPR conflict by showing the two losses compete in the shared encoder’s parameter space. We evaluate CDL on five datasets spanning two structural configurations – one-to-one attributes on user nodes (MovieLens-1M, this http URL-360K, PayPal Audience Factory, BookCrossing) and on item nodes (Yelp) – and find that CDL consistently improves discriminability in attribute embeddings. We also show that ranking (NDCG) improves when attributes carry meaningful preference signal, but conflicts with it when the correlation is weak. We use a lambda parameter to navigate this trade-off, and a lambda-sweep reveals that dataset behavior is governed by two graph properties – semantic alignment and topology leakage. Semantic alignment measures whether the attribute predicts preferences, while topology leakage measures whether the graph’s connectivity already encodes it.
[IR-10] SalesLoop: Reinforcement Learning from Performance Feedback for Sales Lead Ranking
链接: https://arxiv.org/abs/2607.20655
作者: Chenyu Zhang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
备注:
Abstract:Lead ranking in Customer Relationship Management (CRM) systems faces a persistent challenge: models achieving high offline accuracy often underperform in production. We identify three fundamental gaps responsible for this disconnect: offline-online metric mismatch, pointwise-listwise objective misalignment, and temporal distribution drift. To address these gaps, we propose SalesLoop, a reinforcement learning framework that establishes a closed feedback loop between model predictions and real-world business outcomes. Our approach introduces (1) a performance-aware reward that encodes conversion outcomes weighted by ranking position and conversion velocity, and (2) Discriminative GRPO, a listwise optimization objective that adapts Group Relative Policy Optimization to discriminative ranking models. SalesLoop improves NDCG@K by +7.9% and P@K by +15.8% over the strongest static baseline. A 160-day production A/B test at a New Energy Vehicle manufacturer, spanning 16.5M leads and 280 sales specialists across two provincial markets, validates statistically significant cumulative lift of +4.7% ( p=0.047 ) and +8.7% ( p=0.002 ). In production, the ranking backbone achieves Top-10% recall of 44.1% and surfaces high-intent leads at 2.3\times the conversion rate of specialist baselines. Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Information Retrieval (cs.IR) Cite as: arXiv:2607.20655 [cs.LG] (or arXiv:2607.20655v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2607.20655 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[IR-11] opoGuard: Graph Theory Based Defenses Against Split-Knowledge Attacks on RAG
链接: https://arxiv.org/abs/2607.20437
作者: Chahana Dahal,Zuobin Xiong
类目: Computation and Language (cs.CL); Cryptography and Security (cs.CR); Information Retrieval (cs.IR)
备注:
Abstract:Production Retrieval Augmented Generation (RAG) systems rely on aggregating multiple external documents to answer complex queries. However, the retrieved documents introduce a new threat surface that can be exploited to launch split-knowledge attacks. In this attack, the adversary injects documents that are individually benign but create false associations when combined and fed to language models. This paper shows that the new attack is structurally invisible to existing per-document filters, like LlamaGuard. To address this issue in RAG, this work introduces TopoGuard, a family of graph theory-based methods specifically targeting the split-knowledge attacks by building a semantic similarity graph from retrieved documents and detecting contexts with malicious topology. Grounded on the theoretical analysis, the TopoGuard family has been proven to be effective and robust even with noisy inputs. Extensive experiments are conducted on two retrieval datasets and compared with multiple baseline methods. Specifically, the TopoGuard- \lambda_2 +Entity catches 21 \times more attacks than LlamaGuard-2-8B at 1% FPR (32.6% vs 1.5% recall) on the HotpotQA dataset. Compared with production RAG detection systems using large language models, the proposed TopoGuard variants run efficiently at sub-millisecond latency and stay robust under adaptive adversaries and benign cross-domain queries.
[IR-12] Skill-Contracted Agents for Evidence-Aware Materials Literature Analysis
链接: https://arxiv.org/abs/2607.20431
作者: Bixuan Li,Yu Liu,Shuo Shi,Xiaoya Huang,Peng Kang,Lei Zheng
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: 9 pages, 5 figures
Abstract:Materials science literature analysis requires simultaneous attention to composition, processing, characterization, and property relationships, yet conventional retrieval-augmented generation pipelines struggle to reconcile heterogeneous tasks within a single retrieve-then-generate architecture. Here we present AlphaAgent, a skill-driven agent framework that decouples retrieval-based question answering from paper-level report generation through explicit skill contracts. A dedicated retrieval skill rewrites user requests into material-specific search intents, queries a curated index of more than 300,000 papers from the Journal Citation Reports Metallurgy and Metallurgical Engineering category, and reformulates queries when initial evidence is insufficient. A separate report-generation skill parses full-text PDFs to produce structured per-paper analytical reports and cross-paper summaries. In a blind evaluation on 40 materials-science questions, half of which required deep analytical reasoning, AlphaAgent substantially outperformed a baseline system matched for underlying model, document index, and retrieval scale, with the largest gains in mechanistic explanation and awareness of credibility boundaries. These results indicate that explicit task separation, refined retrieval intent, and evidence-aware generation improve large-language-model-based literature analysis for materials research.
[IR-13] Search-on-Graph-R1: Training Large Language Models to Search Knowledge Graphs with Reinforcement Learning
链接: https://arxiv.org/abs/2607.18481
作者: Jia Ao Sun,Hao Yu,Fengran Mo,Zhan Su,Yuchen Hui,Bang Liu,Jian-Yun Nie
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注:
Abstract:Knowledge graph question answering (KGQA) requires navigating from topic entities to an answer several relations away. Recent methods prompt a frontier LLM to explore the graph through a retrieval tool, but their reliance on frontier-scale inference makes them costly to deploy. We present Search-on-Graph-R1 (\sogrone), which internalizes this navigation into a compact 8B model through supervised fine-tuning (SFT) followed by reinforcement learning (RL). Our central idea is to scaffold a frontier teacher with each question’s gold SPARQL query, so the teacher traverses a known answer-bearing path with a live \textttSearch tool rather than having to discover the path itself. Since every call executes against a live Freebase server, the resulting trajectories are grounded in the knowledge graph by construction. On WebQSP, CWQ, and GrailQA, \sogrone at 8B surpasses every frozen frontier-LLM system in our comparison and posts the strongest results on CWQ of any system we compare against. It does so using no auxiliary module at inference and no LLM judge during training. Isolating each training stage shows that SFT and RL contribute complementary gains, our approach transfers across model families, and RL learns to reach answers in fewer \textttSearch calls than its SFT initialization.
[IR-14] Search-on-Graph: Iterative Informed Navigation for Large Language Model Reasoning on Knowledge Graphs KDD’26 KDD
链接: https://arxiv.org/abs/2510.08825
作者: Jia Ao Sun,Hao Yu,Fabrizio Gotti,Fengran Mo,Yihong Wu,Yuchen Hui,Zhan Su,Lingfeng Xiao,Jian-Yun Nie
类目: Computation and Language (cs.CL); Information Retrieval (cs.IR)
备注: Accepted to KDD '26 (32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining)
Abstract:Large language models (LLMs) augmented with knowledge graphs (KGs) offer a promising approach for knowledge-intensive reasoning. Central to this approach is the selection of appropriate reasoning paths in the KG. Yet, existing methods face a common limitation: reasoning path selection is often performed by separate modules using criteria that are only weakly connected to the reasoning requirements. This often results in selecting incorrect relations or premature pruning of relevant paths. We propose Search-on-Graph (SoG), a method that strengthens the connection between path selection and reasoning by having the LLM itself select which relations to follow, informed by both the available KG structure and the complete reasoning history. SoG follows an \textitobserve-think-navigate paradigm: at each step, the LLM observes the relational connections available at the current entity, reasons about which path best advances toward answering the question, and navigates accordingly. This context-aware navigation fully exploits the LLM’s reasoning capabilities rather than relying on independent selection modules with surrogate criteria. Experiments on six knowledge graph question answering (KGQA) benchmarks demonstrate that SoG outperforms state-of-the-art methods while requiring no task-specific fine-tuning and generalizing across different KG schemas.
人机交互
[HC-0] MedGame: Storytelling Gamification Empowered by Large Language Models for Medical Education
链接: https://arxiv.org/abs/2607.21570
作者: Qian Wu,Xinrong Zhou,Zizhan Ma,Kai Chen,Zheyao Gao,Xun Lin,Hongqiu Wu,Longfei Gou,Yixiao Liu,Ann Sin Nga Lau,Qi Dou
类目: Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
备注: Work in Progress; an explorational design and study on AI+Education+Game
Abstract:Large Language Models (LLMs) show promise for medical education, but most existing systems focus on localized interactions such as question answering or single-turn feedback, rather than organizing an entire clinical case into a decision-centered learning trajectory. We introduce \textitMedGame, a framework that transforms static clinical cases into structured, executable storytelling games. MedGame uses a dual-engine design: a Medical Narrative Designer synthesizes case-grounded clinical storylines with states and decision nodes, while a Story Director converts them into dependency-aware multimodal orchestration plans rendered by our released interactive platform. We construct MedGame Bench, a 5,000-case benchmark and evaluation protocol for Medical Narrative Generation and Story Direction. Experiments show that task-specific fine-tuning substantially improves open-source LLMs on MedGame Bench and narrows the gap with commercial models. A pilot student study further shows that learners perceive MedGame as more engaging and useful than text-only alternatives.
[HC-1] Sources of Inequity and Fairness Risks inWellbeing Sensing
链接: https://arxiv.org/abs/2607.21527
作者: Han Zhang,Vedant Das Swain,Koustuv Saha,Anind K. Dey,Jennifer Mankoff
类目: Human-Computer Interaction (cs.HC)
备注: 16 pages, 1 figure, 2 tables. Accepted to AIES 2026
Abstract:Passive sensing for wellbeing uses smartphones and wearables to continuously collect human behavioral data and applies ML/AI models to infer psychological states and behaviors (e.g., depression, cognitive load). These systems are increasingly adopted in high-stakes settings (e.g., hospitals, universities), yet fairness research remains limited—primarily to post-hoc, identity-based comparisons of model performance. However, passive sensing combines heterogeneous sensing infrastructures, indirect behavioral inference, and longitudinal deployment—characteristics that, while not exclusive to the domain, are jointly pronounced here and raise two underexplored questions: (1) what additional sources of inequity arise from these characteristics, and (2) how do such inequities propagate beyond algorithmic audits across the system lifecycle? To address this gap, we conducted semi-structured interviews with 14 researchers and practitioners across five countries, examining how fairness risks emerge and are negotiated across the full passive sensing lifecycle. Our findings empirically characterize five situated sources of inequity (e.g., comfort with monitoring, behavioral regularity) that systematically shape fairness risks beyond identity-based attributes. We further synthesize 15 fairness risks and corresponding mitigation strategies across the lifecycle, from study design to deployment. Finally, we identify structural barriers that constrain fair practice in reality, and argue that enabling fair passive sensing requires both individual researcher efforts and ecosystem-level governance support from funders, publication venues, and deploying institutions.
[HC-2] A Needs Assessment for Measuring Geographic - Legislative Associations in the U.S. House of Representatives
链接: https://arxiv.org/abs/2607.21502
作者: Ashu Gupta,Cameron Owens,Benjamin Wilson-Langman,Max Hill,Jason Cox,Joshua Clinton,Clio Andris
类目: Human-Computer Interaction (cs.HC)
备注: 35 pages, 9 figures, 6 tables. Link to map: this https URL . Research Team: Ashu Gupta, Cameron Owens, Benjamin Wilson-Langman, Max Hill, Jason Cox, Ian Arnold, Santiago Velez, Quinn Pearson, Yunhe Cui, Isaac Felix, Ryan Luong, Priya Parikh, Shubham Purish, Vivek Rao, Cat Tran, Mingkuan Yan, Christopher Yang, Clio Andris
Abstract:Political legislation affects the well-being and livelihoods of constituents. In the U.S. Congress a representative’s voting record on bills and legislation is public. These bills have themes associated with them, such as veterans’ affairs, coastal monitoring, agricultural appropriations, etc. A bill on veterans’ affairs may affect a constituency differently if they have a high percentage of veterans. In this work, we demonstrate how congressional vote outcomes can be merged with typical geographic information systems (GIS) data to help compare a legislator’s votes with the geographies of their constituencies to measure the association between district features and legislators’ decisions. We retrieved and tagged bills from the 118th U.S. House of Representatives (Jan. 2023 - Jan. 2025) by manually assigning each bill a set of themes. We then retrieved spatial data at the congressional district level related to each theme. We built a backend database to connect bill information and spatial data, and an interactive map prototype displayed the spatial data related to each theme associated with each bill. Our work helps identify challenges to creating a more accessible and complete system that would facilitate the visualization and analysis of the nature of contemporary Congressional representation. This technical report is primarily written for members of the data science community, and specifically those interested in legislative and geographic data. Comments: 35 pages, 9 figures, 6 tables. Link to map: this https URL. Research Team: Ashu Gupta, Cameron Owens, Benjamin Wilson-Langman, Max Hill, Jason Cox, Ian Arnold, Santiago Velez, Quinn Pearson, Yunhe Cui, Isaac Felix, Ryan Luong, Priya Parikh, Shubham Purish, Vivek Rao, Cat Tran, Mingkuan Yan, Christopher Yang, Clio Andris Subjects: Human-Computer Interaction (cs.HC) Cite as: arXiv:2607.21502 [cs.HC] (or arXiv:2607.21502v1 [cs.HC] for this version) https://doi.org/10.48550/arXiv.2607.21502 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[HC-3] hinkink: 2D Spatial Ink-native Interaction with LLM s
链接: https://arxiv.org/abs/2607.21468
作者: Mohammad Hasan Payandeh,Daniel Vogel,Jian Zhao
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注:
Abstract:People often use handwritten notes and sketches to externalize ideas for ideation. To integrate large language models (LLMs) into this practice, we propose Thinkink. Prompts can be handwritten text or drawn sketches with LLM-generated responses visualized as ink-like text and sketches spatially integrated into a shared canvas. A semantic tree streamlines ink interpretation, and a lightweight UI provides explicit control using a state machine. The tool was designed using a three-stage process. A formative study (N=12) examined current practices with conventional and digital inking methods. The results informed a technical probe for a diagnostic study (N=6) identifying usability and human-LLM interaction challenges. This motivated the design of Thinkink, with a final study (N=10) examining how people incorporate it into their ideation practices. We contribute design implications and a tool for ink-native LLM interaction where the user and LLM write and draw in a shared 2D canvas.
[HC-4] CRAFT: Exploring Wearable Creative AI on Smart Glasses for Fiction Writing in Real-World Contexts
链接: https://arxiv.org/abs/2607.21394
作者: Runze Cai,Yuxuan Huang,Lin-Ping Yuan,Kexin Xiang,David Hsu,Collier Nogues,Jussi Holopainen,Shengdong Zhao
类目: Human-Computer Interaction (cs.HC)
备注: Conditionally accepted (pending minor revisions) for publication in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (PACM IMWUT), Vol. 10, No. 3, Article 81, September 2026
Abstract:Creative writing increasingly integrates AI assistance, yet current tools miss in-situ moments when writers draw inspiration from real-world experiences. We envision Context-aware Reality-Fiction Transformation (CRAFT), an approach for AI glasses that translates daily experiences into fiction narratives. We explored its desirability, feasibility, and potential viability through three studies. Interviews with nine writers yielded desires and three design goals: 1) augmenting in-situ perception to bridge reality-fiction gaps, 2) promoting authenticity grounded in real-world experiences while maintaining fictionalization, and 3) preserving creative agency, enjoyment, and life-art boundaries. Co-design workshops with 16 writers and researchers operationalized these goals into concrete interaction mechanisms using a technology probe. We then conducted supported field trials with eight writers across 24 sessions using a refined probe, revealing writer-perceived benefits (e.g., enriched fictional ideas from serendipitous encounters), emergent practices (e.g., micro-creation), and design considerations for future sustained use. We contribute design explorations for the CRAFT approach, offering design implications and empirical insights on ubiquitous human-AI creative collaboration in everyday life.
[HC-5] AI Assistants Overassist
链接: https://arxiv.org/abs/2607.21306
作者: Verona Teo,Raghav Jain,Tobias Gerstenberg,Max Kleiman-Weiner
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注:
Abstract:Large language models (LLMs) are increasingly used as tutors and thought partners, helping users reason through problems. While guidance from AI assistants can scaffold thinking and foster learning, such benefits depend on how they help–for instance, intervening too early or too frequently may hinder true learning and cognitive engagement. Yet how AI systems navigate intervention decisions during problem-solving remains poorly understood. Here, we introduce Int-Bench, a simulation-based benchmark for evaluating LLM interventions during learning. Int-Bench simulates a “student” solving a problem while a “teacher” monitors the student’s reasoning and decides whether, when, and how to intervene. Across three domains–code debugging, mathematics, and brain teasers–we evaluate LLM teachers on the frequency and timing of interventions, as well as their impact on both immediate task success and generalization to new problems. We also compare LLMs to humans, finding that LLMs intervene more frequently and earlier than humans. Moreover, in contrast to humans, they tend to provide complete solutions rather than targeted hints. These findings suggest that current LLM assistants often optimize for short-term success rather than supporting the reasoning processes needed for deeper learning and long-term success.
[HC-6] Reimagining the Augmented Reality Accessibility Ecosystem for Deaf Students: Service Provider Perspectives in Experiential Learning
链接: https://arxiv.org/abs/2607.21289
作者: Roshan Mathew,Roshan Peiris
类目: Human-Computer Interaction (cs.HC)
备注: 14 pages, 2 images, HCII2026
Abstract:In experiential learning environments, Deaf and hard of hearing (DHH) students often experience ``split attention,‘’ dividing their focus among tasks, instructors, and access providers. Augmented reality (AR) has been proposed as a means to centralize communication access within the student’s field of view; however, little is known about how such systems affect the instructors, interpreters, and captioners who support access in these settings. We present a formative, expert-based evaluation of ARRAE, an AR-mediated communication access ecosystem, examining the experiences of an instructor, an American Sign Language (ASL) interpreter, and a real-time captioner in a simulated laboratory environment. Comparing in-person, traditional remote, and AR-mediated access, our findings suggest that AR reconfigures accessibility labor and interactional practices by redistributing communication, attention, and awareness across participants. While AR-mediated access supported integrated, in-situ communication and enhanced contextual awareness, it also altered interactional feedback mechanisms, such as gaze coordination and pacing, that support shared understanding during instruction. These observations highlight trade-offs between access, control, and coordination, and suggest that AR-mediated systems may be most effective when considered as part of a broader, hybrid ecosystem rather than as standalone replacements for existing modalities. This work contributes initial insights into how AR reshapes communication access in experiential learning and informs the design of systems that support coordinated, multi-user interaction in hands-on educational contexts.
[HC-7] Exploring the Design Space of LLM -Based Programming Support in CS Education: A Scoping Review through the Lens of Assistance Governance
链接: https://arxiv.org/abs/2607.21257
作者: Minsun Kim,S. Moonwara A. Monisha,Zihan Wu,David H. Smith IV
类目: Human-Computer Interaction (cs.HC)
备注:
Abstract:As large language models (LLMs) become integrated into programming education, learner-facing systems increasingly differ in how that assistance is bounded, enacted, and controlled. These governance decisions are often described implicitly, making it difficult to compare systems in educationally meaningful ways. To address this gap, we conduct a scoping review and qualitative synthesis of 90 peer-reviewed LLM-based programming support systems in CS education. We analyze assistance governance through three dimensions, which we refer to collectively as PEA: Policy, capturing what forms of help are allowed or restricted; Enforcement, capturing how those boundaries are operationalized through interaction and system behavior; and Authority, capturing who can configure, adapt, or override them during use. Our findings show that systems often share similar pedagogical goals, but implement those goals through varied enforcement mechanisms. At the same time, authority remains highly centralized in system logic, with fewer systems giving learners or instructors runtime control. This work contributes PEA as a three-dimensional analytic lens, a governance codebook empirically refined within these dimensions, and a map of underexplored configurations in the current design space of LLM-based programming support. By making these explicit and comparable, PEA offers a vocabulary for analyzing existing systems and designing future tools that are pedagogically bounded, configurable, and accountable.
[HC-8] QuantiBias: Benchmarking Quantization-Induced Bias in LLM s
链接: https://arxiv.org/abs/2607.21063
作者: Emilio Ferrara
类目: Computation and Language (cs.CL); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
备注: Benchmark protocol on Hugging Face: this https URL
Abstract:Almost every large language model that reaches a broad audience is quantized: trained in full precision, then compressed for efficiency. This step is assumed harmless and its safety is rarely re-checked. We find its principal side effect is increased bias that standard safety evaluation misses. Holding the model, its training, and the prompts fixed, a quantized model still refuses harmful requests, still avoids over-refusing benign prompts, and still selects the unbiased multiple-choice answer. Yet asked an open-ended question, the same model volunteers stereotypes in all eight languages we probe, in roughly one in four open-ended answers under an independent judge (~24% to ~27% across the compression ladder): it passes every standard check and still reaches users measurably more biased. The selective gap is a robust finding; whether open-ended bias further increases with compression is less certain, sensitive to the judge that scores it. We address both with \textbfQuantiBias, a benchmark that pairs a generative, multilingual stereotype probe with the refusal and multiple-choice controls that isolate open-ended generation, contrasts each build with and without reasoning, and rates the content severity of what it generates. Across two backbone models (Qwen and Gemma), a five-family screen, and eight benchmarks, quantizers allocate their extra precision by capability data that carries no bias-prevention signal, and reasoning before answering roughly halves the effect on some families while doing nothing on others. A quantized build must be re-evaluated for open-ended bias, not only on the short-form safeguards it already passes.
[HC-9] Sidewalk Moments: Are Richer Representations Always More Human-Aligned? Evidence from City-Walk Videos
链接: https://arxiv.org/abs/2607.20903
作者: Liu Liu,Freya Huying Tan,Fábio Duarte
类目: Computer Vision and Pattern Recognition (cs.CV); Human-Computer Interaction (cs.HC)
备注: 35 pages, 18 figures. Under review at Scientific Reports
Abstract:We examine whether richer visual representations yield more human-aligned measures of urban engagement, using 61 first-person city-walk videos from YouTube segmented into over 50,000 ten-second clips and represented across four modalities: spatiotemporal video features, temporally averaged images (TAIs), audio embeddings, and text-based semantic descriptions. Spearman correlation analysis reveals the expected ordering along the temporal-richness continuum, with video features showing the strongest continuous alignment. However, this ordering breaks down under binary classification of high- versus low-engagement moments (the paradigm most commonly used to train perceptual scoring models), where TAIs consistently match or outperform video across most classifiers and quantile thresholds. An independent two-alternative forced-choice study on Amazon Mechanical Turk confirms that this parity reflects human judgment: participants identified engaging moments with comparable accuracy from TAIs and full video clips, while text performed substantially worse and audio remained near chance. Gap analysis reveals a functional dissociation: video features are advantaged in activity-driven scenes with dynamic content, whereas TAIs better align with human judgments in composition-driven scenes dominated by stable spatial structure. These findings challenge the assumption that richer representations are inherently more human-aligned, and suggest that perceptually grounded temporal compression can be a principled alternative to full video encoding.
[HC-10] Sonic Stage: Automatically Generating Interactive Spatial Soundscapes to Facilitate Dialogue Video Comprehension for Blind Viewers
链接: https://arxiv.org/abs/2607.20835
作者: Shuchang Xu,Xiaofu Jin,Gaurav Jain,Wenshuo Zhang,Huamin Qu,Brian A. Smith,Yukang Yan
类目: Human-Computer Interaction (cs.HC)
备注: accepted to UIST 2026
Abstract:Audio description (AD) makes film and television accessible to blind and low-vision (BLV) audiences by narrating characters’ actions. However, in scenes with lots of dialogue, AD often omits important actions because it is constrained not to overlap with speech. It is not yet known how to convey characters’ actions during dialogue. We present Sonic Stage, a system that transforms dialogue videos into interactive spatial soundscapes, enabling BLV audiences to intuitively understand characters’ actions and movements through immersive auditory cues. Sonic Stage conveys essential visual information during dialogue through three auditory techniques: (1) spatialized dialogue to represent spatial layout, (2) diegetic sound to convey character actions, and (3) interactive descriptions to provide context-specific visual details. Evaluation with 12 BLV viewers showed that Sonic Stage significantly improved video comprehension, spatial presence, and narrative engagement. We highlight opportunities for enhancing video accessibility across diverse genres through immersive, interactive audio representations.
[HC-11] Can an AI System Be Creative? A Critical Perspective from Art and Engineering
链接: https://arxiv.org/abs/2607.20796
作者: Ivan Magrin-Chagnolleau
类目: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC); Multimedia (cs.MM)
备注:
Abstract:This paper examines the question of whether artificial intelligence (AI) systems can be creative, approached from the dual perspective of a researcher trained in electrical engineering, pattern recognition, machine learning, and neural networks, who has also spent most of his life engaged in the arts as actor, stage and film director, writer, composer, and visual artist, and in philosophy. Drawing on Margaret Boden’s foundational framework, both her three properties of creativity (novelty, surprise, and value) and her three types of creative processes (combinatorial, exploratory, and transformational), the paper argues that AI systems are structurally incapable of creativity in its strongest sense. While they exhibit genuine capability in the domain of combinatorial creativity, they are significantly bounded in exploratory creativity, and fundamentally incapable of transformational creativity. The paper further argues that the most important limitation of current AI systems is not the absence of novelty per se, but the absence of any mechanism for serendipity, accident, or the unexpected, all of which play a central role in the phenomenology of creativity, and the absence of any subject position from which to recognize and welcome such chance events. The paper concludes by proposing a model of human, AI creative collaboration that is both realistic and generative, illustrated by several concrete experiments. The paper is itself a demonstration of the thesis it advances: it was composed through a deliberate human AI collaborative process, which is described in the methodological note that opens it.
[HC-12] Flint: A Semantics-Driven Data Visualization Intermediate Language
链接: https://arxiv.org/abs/2607.20775
作者: Yunhai Wang,Kecheng Lu,Junhao Chen,Alper Sarikaya,Chenglong Wang
类目: Human-Computer Interaction (cs.HC); Programming Languages (cs.PL)
备注:
Abstract:We present Flint, an intermediate language that enables authors to create high-quality visualizations from concise, semantics-driven specifications without explicitly configuring low-level parameters such as scales, axes, and formatting. Unlike prior systems that infer default configurations from surface-level data representations, often producing brittle choices, Flint introduces a hierarchical data semantic model that allows users to specify the meanings of data fields structurally and helps the compiler derive appropriate visualization configurations. From a concise specification, the system generates and optimizes library-agnostic visualization configurations and translates them into complete, executable specifications for multiple target grammars, including Vega-Lite, Apache ECharts, and this http URL. We demonstrate that Flint simplifies the authoring process without compromising on visual quality, and it is an effective intermediate language for both humans and AI agents to create visualizations.
[HC-13] HARP: The Human–AI Research Platform
链接: https://arxiv.org/abs/2607.20773
作者: Zeshu Zhu,Natalie Friedman,Kevin Weatherwax,Emily Eiben
类目: Human-Computer Interaction (cs.HC); Artificial Intelligence (cs.AI)
备注: 5 pages, 3 figures, SAP Academic Community Conference North America, 2026
Abstract:Large language models (LLMs) have shifted human–computer interaction from traditional'' interface journeys toward more conversational exchanges. Researchers studying HCI and UI use moderated usability sessions, interviews, surveys, transcript analysis, and static prototypes. However, static prototypes provide limited opportunities to study interaction with live AI systems or systematically control how an LLM behaves across participants and scenarios. Conversation transcripts reveal little about how users formulate, revise, and hesitate over prompts before submission. We designed the Human--AI Research Platform (HARP) for researchers, designers, and anyone who has ever wondered, What if AI did this?’ HARP places participants in controlled mock scenarios with live, configurable AI agents. Researchers can control agent prompts, model parameters, response characteristics, and experimental conditions; trigger surveys at predefined moments; and record prompt composition time, response latency, deletions, and keystroke pauses. Planned capabilities include voice, facial expression, gesture, and, where legally and ethically appropriate, emotion analysis. We illustrate HARP through a study examining how technical specificity and response length affect retention of LLM output. By pairing controllable live agents with behavioral and self-report measures, HARP enables systematic testing of how AI design choices affect users.
[HC-14] Improving the performance of an ASV system using hybrid speech features
链接: https://arxiv.org/abs/2607.20706
作者: Stanisław Ciszkiewicz,Artur Janicki
类目: ound (cs.SD); Human-Computer Interaction (cs.HC)
备注:
Abstract:The growing need for secure and convenient authentication methods has led to the increasing popularity of biometric solutions. In addition to traditional and popular methods, such as fingerprint or iris scanning, voice-based approaches are also employed. User identity verification based on voice is conducted using Automatic Speaker Verification (ASV) systems. Despite their many advantages, these systems are sensitive to various types of attacks and acoustic noises, which can reduce verification accuracy. This work examines the potential to improve the performance of ASV systems by using hybrid feature sets that combine different signal representations, starting with widely-used Mel-Frequency Cepstral Coefficients (MFCC), through Constant Q Cepstral Coefficients (CQCC) and ending with the innovative RAB descriptor. Experiments were conducted on recordings from the Google Speech Commands dataset under two scenarios: in clean conditions and in the presence of acoustic noise. Finally, the systems’ performance was compared using the EER metric to determine whether hybrid feature sets decrease verification error. The results show that using a hybrid feature set (PNCC+RAB) improves speaker verification performance under noisy conditions.
[HC-15] AuthProbe: Specification-Driven Multi-Identity Detection of Broken Object-Level Authorization in Recruitment API
链接: https://arxiv.org/abs/2607.20574
作者: Jay Barach
类目: Cryptography and Security (cs.CR); Human-Computer Interaction (cs.HC)
备注: 10 pages, 3 figures, 4 tables
Abstract:Broken Object-Level Authorization (BOLA), also known as Insecure Direct Object Reference (IDOR), has topped the OWASP API Security ranking since 2019 and is the root cause of some of the largest exposures of applicant data in recruitment technology. The defining feature of this flaw class is that a malicious request is byte-for-byte indistinguishable from a legitimate one, which is precisely why web application firewalls and single identity scanners fail to catch it. We present AuthProbe, an open-source, black-box scanner that detects BOLA and IDOR in HTTP APIs by driving its tests from an OpenAPI specification and by acting under two or more identities that the operator controls. AuthProbe discovers, for each identity, the objects that identity legitimately owns, then attempts to read one identity’s objects while authenticated as another and confirms a leak by comparing the response against a ground-truth fetch by the true owner. It also walks predictable identifiers to expose enumeration and reports missing authentication and existence oracles. The tool returns a severity-thresholded exit code and machine-readable reports so that it can gate a continuous integration build. On a synthetic recruitment API in which the McHire failure class is reproduced, AuthProbe detects every planted cross-identity read with no false positives on a hardened counterpart, and its running time grows linearly with the number of objects under test. AuthProbe is released under the Apache 2.0 license with an authorized-use guardrail.
[HC-16] Representation Robustness Under Executable Reasoning Constraints in Large Language Models for Mathematical Problem Solving
链接: https://arxiv.org/abs/2607.20520
作者: Sagnik Nath,Edith Aurora Graf,Liang Zhang,Diego Zapata-Rivera
类目: Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC); Programming Languages (cs.PL)
备注: presented at the 28th International Conference on Human-Computer Interaction (2026), Montreal, Canada
Abstract:Large language models (LLMs) are increasingly evaluated on mathematical problem solving, yet prior work often treats representationally equivalent formulations as interchangeable and conflates reasoning errors with interface failures. This paper investigates representation robustness in LLM-based mathematical problem solving by systematically varying surface representations of the same underlying problems, including story problems, word-equations, symbolic equations, and isomorphic paraphrases. Using a curated dataset of mathematically equivalent problems, we evaluate five contemporary LLMs under a direct answer generation condition. We find substantial representational sensitivity: models frequently change correctness across equivalent formulations, with nontrivial flip rates across story, symbolic, and word-equation variants. We also observe systematic regressions under isomorphic reformulations, showing that even subtle paraphrase-level changes can degrade performance despite preserved mathematical structure. We then evaluate a code-augmented condition in which models externalize reasoning as executable Python code that is run locally for validation. This interface reveals strong latent reasoning capability in some models that perform poorly under direct prompting, but it does not uniformly improve robustness. Instead, failures shift across interaction layers, from opaque reasoning errors to protocol violations and execution failures. Even when executable reasoning succeeds, representation sensitivity often persists. Overall, our results show that reasoning scaffolds do not eliminate representational brittleness, but expose new tradeoffs among correctness, reliability, latency, and cost. We argue that representation should be treated as a first-class interface design variable in LLM evaluation and deployment, especially for AI-assisted problem-solving systems.
[HC-17] Can Valence Reflect Morality in Natural Language? A Preliminary Annotation Study
链接: https://arxiv.org/abs/2607.20461
作者: Jonny O’Dwyer,Malika Bendechache,Louise McCormack,Elif Calik,Ramin Ranjbarzadeh,Dost Muhammad,Shokofeh Anari Bozcheloei,Ishita Singh
类目: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)
备注: 8 pages, 2 figures, submitted to the 36th Irish Signals and Systems Conference
Abstract:Present implementations of artificial intelligence (AI) ethics do not adequately take feelings, or affect, into account. If AI should be aligned with human ethics, it seems reasonable to thoroughly investigate the possibility of AI behaviour that mirrors virtuous human ethical conduct, where feelings play a role in the actions, judgements or statements one makes. Furthermore, while prominent theories of normative ethics are often discussed in terms of their differences and shortcomings, Virtue, Consequentialist, and Kantian Deontological ethics all share a common feature of considering human feeling to some degree while the popular descriptive ethics theory, Moral Foundations Theory, positions feelings as central to many of its foundations. Therefore, in the present paper, a data set of moral valence is proposed, consisting of 500 annotations by six human participants for both action/judgement and consequence moral valence, ranging from -1 to 1 for text-presented scenarios from the Commonsense Norm Bank data set. The resulting valence features share significant relationships with multi-class (immoral/discretionary/moral) and binary immoral/moral categories while additionally providing a noteworthy test set Matthew’s correlation coefficient of 0.764 using regularised logistic regression for binary classification. This provides early evidence of the usefulness of valence features for morality estimation of text, indicating that valenced consequences of responses for others can be considered toward more human morally-aligned AI. In the interest of promoting further affective-moral computing research, this study’s annotations will be made available for research on request.
[HC-18] Mammal: Supporting Breastfeeding Monitoring Through Computational Garments with Inter-Body Sensing
链接: https://arxiv.org/abs/2607.19796
作者: Yanfeng Zhao,Morgan Geck,Kate Fernandez,Madison Nicole Jones,Xia Zhou,Jessica L. Ridgway,Te-Yen Wu
类目: Human-Computer Interaction (cs.HC); Hardware Architecture (cs.AR)
备注: Submitted to IMWUT’26
Abstract:Breastfeeding provides critical insight into infant feeding competence and physiological health, yet objective monitoring remains difficult due to the intimate and internal nature of feeding. We present Mammal, a caregiver-worn computational garment that unobtrusively monitors breastfeeding without attaching sensors to the infant. Mammal leverages inter-body signal transmission through natural mouth-to-breast contact to capture infant cardiac and feeding-related acoustic signals on the caregiver’s body. Using novel algorithms to detect latch onset, infer infant electrocardiogram (ECG), and identify suck and swallow events from inter-body signals, Mammal estimates latch duration, in-feeding heart rate, suck-swallow-breathe (SSB) ratio, and milk intake. In a user study with 10 caregiver-infant dyads, Mammal achieves a mean absolute percentage error (MAPE) of 5.56% for latch duration, a mean absolute error (MAE) of 3.61 bpm for infant heart rate estimation, a mean absolute error of 0.12 for SSB ratio estimation, and a mean relative error of 15.76% for milk intake, with participants reporting high comfort and wearability.
计算机视觉
[CV-0] 3D-Aware VLMs with Implicit and Explicit Geometries ECCV2026
链接: https://arxiv.org/abs/2607.21595
作者: Wenhao Li,Xueying Jiang,Quanhao Qian,Deli Zhao,Ran Xu,Shijian Lu,Gongjie Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted by ECCV 2026, Open Sourced
Abstract:Despite rapid progress, most existing vision-language models (VLMs) built from 2D visual inputs often struggle when handling various 3D tasks that require fine-grained spatial understanding and reasoning. To bridge this gap, we present VLM-IE3D, a unified framework that enhances the 3D spatial awareness of VLMs by equipping them with both implicit and explicit 3D geometries learned from RGB videos. Our VLM-IE3D introduces Implicit Geometry Tokens (IGTs) that capture high-level geometric priors from input videos, as well as complementary Explicit Geometry Tokens (EGTs) that encode detailed geometric structures from reconstructed 3D attributes. On top of that, VLM-IE3D comes with a 3D-aware adapter that effectively fuses the two types of geometric representations with 2D visual cues. This RGB-only design injects strong 3D inductive biases for fine-grained spatial understanding and reasoning without requiring any additional 3D inputs. Extensive experiments show that VLM-IE3D achieves superior performance consistently across various 3D tasks including 3D video detection, 3D visual grounding, 3D dense captioning, and spatial reasoning. Code and models are available at this https URL.
[CV-1] Streaming Multi-Agent Autoregressive Diffusion Model with World State Registers
链接: https://arxiv.org/abs/2607.21594
作者: Sicheng Mo,Yuheng Li,Ziyang Leng,Krishna Kumar Singh,Bolei Zhou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL
Abstract:Multi-agent interactive world models should not only generate consistent observations, but also maintain world states that persist across agents and evolve across views. Existing autoregressive video diffusion pipelines carry forward observation history as conditioning context, which makes shared state difficult to maintain in multi-agent and multi-view settings. We present WorldWeaver (W^2), a streaming multi-agent video diffusion model that augments rollout with cross-agent world state registers: learnable tokens that store shared world information, track individual agent status, and are dynamically updated after each generated chunk. We ground these registers with supervision signals spanning individual agent status, global state views including bird’s-eye views, and scene text. We further improve the architecture with a Mixture-of-Transformers design that uses separate weights for world state modeling and visual frame modeling. Extensive experiments in two-agent Minecraft video generation show that explicit world-state modeling improves logical consistency and generation quality.
[CV-2] Unified Video Dense Prediction from Disjoint Data ECCV2026
链接: https://arxiv.org/abs/2607.21592
作者: Yihong Sun,Seoung Wug Oh,Jiahui Huang,Bharath Hariharan,Joon-Young Lee
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ECCV 2026
Abstract:Scene understanding requires simultaneous prediction about geometry, appearance, and semantics. However, existing task-specific annotations are fragmented across incompatible, domain-specific datasets. Current unified systems circumvent this by restricting training to fully co-annotated data, or by incurring the large computational cost of pseudo-labeling. To mitigate this, we introduce UniD, a unified video model that jointly predicts eight dense scene properties-depth, surface normals, semantic segmentation, boundaries, human parts, albedo, shading, and materials-all learned from disjoint, domain-specific datasets. We propose a simple yet effective distillation step in which per-task experts supervise a unified backbone through lightweight task projectors, eliminating the need for annotation overlap or pseudo-labeling. Our key insight is that the strong visual priors of a pretrained diffusion model are sufficient to bridge the domain gaps introduced by disjoint training sources, enabling robust generalization to scene-task combinations never seen during training. UniD achieves competitive performance against per-task specialists and multi-task baselines, with strong generalization to out-of-distribution scenarios and enhanced temporal and cross-task consistency. Code and video results are available at this https URL.
[CV-3] Inference-Time Scaling of Diffusion Models via Progressive Seed Pruning WWW ALT
链接: https://arxiv.org/abs/2607.21591
作者: Rogerio Guimaraes,Pietro Perona
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Project page: this https URL . Code: this https URL
Abstract:Diffusion and flow-matching models dominate conditional image generation, yet inference-time scaling for these models is far less developed than for autoregressive language models. Because final quality is highly sensitive to the initial noise seed, many approaches spend extra compute on seed search or resampling under a black-box reward, but typically maintaining a constant memory footprint throughout inference. We show that relaxing this constraint enables an underexplored inference-time scaling axis: by front-loading exploration, evaluating many seeds early, and pruning aggressively, we can use a fixed compute budget more effectively. \emphProgressive Seed Pruning (\PSP) scores intermediate denoised estimates and progressively narrows the candidate set so that only promising trajectories are fully denoised, while keeping the total number of model evaluations fixed. Across diffusion and flow-matching backbones, \PSP \ consistently improves reward-guided selection and achieves higher GenEval scores (automated) and better human evaluation on prompt-alignment than best-of- N , importance-sampling, and tree-search baselines at matched compute. Project page: this https URL. Code: this https URL.
[CV-4] Scale Up Strategically: Learning Compositional Generalization via Bias-Aware Evaluation and Data Collection for Robotic Manipulation
链接: https://arxiv.org/abs/2607.21582
作者: Yu Qi,Zhang Ye,Xinyi Xu,Yuxuan Lu,Amitoj Sandhu,Boce Hu,Haojie Huang,Jonathan Tremblay,Lawson L.S. Wong
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Compositional generalization is essential for robot to follow diverse instructions. However, pretrained policies are known to take shortcuts, deferring to salient cues rather than grounding language. We introduce a diagnostic framework that localizes this failure to individual \textitinstruction factors, \textite.g., reusable semantic components such as color, verb, object, size, and spatial attribute. Our framework formalizes instruction factor bias, the tendency of fine-tuned policies to over-rely on dominant factors as shortcuts, and quantifies it through two metrics: Factor Dominance Rate (FDR), capturing pairwise bias between factors, and Factor Dominance Hierarchy (FDH), aggregating these into a global ranking. Evaluation on six foundation policies reveals broadly consistent ordering, \textiti.e., color \geq object \geq spatial \geq verb \geq size, with color dominant, and verb and size most under-grounded. We further show the diagnosis is actionable: a bias-aware data collection strategy that reallocates a fixed budget toward under-grounded factors outperforms baselines in simulation and on a real robot using half the demonstrations, thereby enabling more sample-efficient and generalizable policy learning.
[CV-5] GraphVid: Interactive Graph-Controllable Video Generation
链接: https://arxiv.org/abs/2607.21580
作者: Vedant Shah,Onkar Susladkar,Tushar Prakash,Kiet Nguyen,Tianjio Yu,Adheesh Juvekar,Muntasir Waheed,Ismini Lourentzou
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Controllable video generation remains challenging due to the difficulty of specifying precise multi-object interactions using text prompts or motion-control inputs that primarily constrain pixel movement. In practice, trajectory-based control often requires users to draw accurate tracks for multiple objects, which scales poorly with scene complexity and becomes ambiguous under occlusion or overlap. To enable flexible yet precise multi-subject control, we introduce \textbfGraphVid , a graph-conditioned image-to-video generation model that enables interactive control through structured interaction graphs. We further curate \textbfGraphVid-Bench , a large-scale interaction-centric video dataset with structured relational annotations to enable training of interaction-aware video generation models. Despite using substantially less training data and fewer trainable parameters than prior motion-control methods, GraphVid delivers strong controllability and video quality. Compared with Motion-I2V, GraphVid reduces FID by up to 39.9% and FVD by 37.6%, while improving PSNR (9.87=15.98) and SSIM (0.38=0.61). Our results highlight the potential of structured semantic interfaces as a powerful paradigm for controllable video generation.
[CV-6] Synthetic data generation framework for quality control automation in gravure printing
链接: https://arxiv.org/abs/2607.21577
作者: Korota Arsène Coulibaly,Mohamed Hamlich,Khalid Hmali,Andrea Trombin
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Image and Video Processing (eess.IV)
备注: 27 pages, 15 figures. To be submitted to Journal of Engineering Research (Elsevier). Certain TeX commands are supported
Abstract:Quality control in printing, particularly in rotogravure printing, still depends on slow, costly, and subjective manual inspection. Automated surface defect detection is critical for maintaining high-quality standards in rotogravure printing. Deep learning models give prospects for automation. However, training robust deep learning models, such as YOLO or Vision Transformers, is heavily hindered by the extreme scarcity of real-world industrial defects images. To overcome this limitation, this paper introduces a novel synthetic data generation framework tailored for rotogravure printing quality control. The proposed pipeline automatically generates high-fidelity images of specific printing defects (creases, streaks, misregistration, etc.) and outputs corresponding bounding boxes and annotations. To validate the framework, a synthetic dataset of 7533 images was generated and used to train the state-of-the-art object-detection model RFDETR. Experimental results demonstrate that the model trained on our synthetic data achieves a Mean Average Precision (mAP) of 80.9% on real industrial testing samples. This framework provides a zero-cost, rapid-deployment solution for automating defect inspection in printing lines without requiring massive manual data collection.
[CV-7] Self-Supervised Learning of Structured Dynamics from Videos
链接: https://arxiv.org/abs/2607.21576
作者: Lukas Knobel,Andrew Zisserman,Yuki M. Asano
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: preprint, Project page: this https URL
Abstract:Understanding motion in video is a fundamental challenge for visual learning, as frame-to-frame change entangles two sources of dynamics: camera motion and object motion. This decomposition has remained underexplored in representation learning, partly because these factors are tightly coupled in natural videos and difficult to supervise separately. Yet recovering it is important for learning robust motion representations that separate meaningful object dynamics from camera-induced variation. We study whether such structured motion representations can be recovered from frozen features of a pretrained image vision transformer. We propose the Structured Dynamics Model (SDM), which explicitly separates the dominant source of temporal change from residual dynamics through future-feature prediction, rather than representing video change with a single entangled latent or with unstructured, spatially dense transition tokens. Training combines self-supervised learning on real video with weak supervision of scene dynamics on synthetic Kubric data. We evaluate SDM on ProbeMotion, a new evaluation suite spanning synthetic and real videos with camera motion, object motion, and combined dynamics. SDM outperforms backbone baselines using global CLS or average-pooled features, and compares favorably to strongly supervised representations such as VGGT on several probes, despite using substantially weaker supervision. These results suggest that pretrained image models can be readily repurposed into structured video-dynamics representations, providing a useful inductive bias for learning and analyzing latent video dynamics.
[CV-8] Scene Parameter Saliency via Differentiable Light Transport
链接: https://arxiv.org/abs/2607.21562
作者: Linas Beresna,Eugene Fiume
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 13 pages, 5 figures
Abstract:Gradient-based saliency methods reveal which input features most influence a neural network’s output, and are a standard tool for model interpretability. We observe that differentiable renderers, which are conventionally used for parameter optimisation, produce an analogous form of saliency: given any scalar metric evaluated on a rendered image, a single reverse-mode differentiation pass yields per-parameter gradients that identify which scene elements most influence the metric. We call these gradient fields metric saliency maps. Unlike neural saliency, which propagates attribution through learned weights, metric saliency propagates through the image formation process itself, including multi-bounce light transport, capturing parameter dependencies that are semi-opaque to manual inspection. We compute metric saliency maps for qualitatively different objectives: psychovisual glare indices, mean scene luminance, and neural perceptual scores. The saliency rankings differ substantially across metrics for the same scene, with parameters that dominate one objective being negligible for another. The saliency map is specific to the metric, not an intrinsic property of the scene. Our results suggest that differentiable renderers produce derivative images that are as informative for scene understanding as the primal images they were designed to generate.
[CV-9] Visual Contrastive Self-Distillation
链接: https://arxiv.org/abs/2607.21556
作者: Yijun Liang,Yunjie Tian,Yijiang Li,Yuqi Jia,Furong Huang,Tianyi Zhou,Di Fu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 15 pages
Abstract:On-policy self-distillation (OPSD) is promising as it removes the external teacher required by on-policy distillation (OPD), yet it still needs asymmetric information between teacher and student to ensure that the self-teacher provides a stronger learning signal than the student. Existing methods create this asymmetry either through privileged answers or visual evidence. We ask whether both can be removed, yielding a simpler form of OPSD driven purely by input conditioning. For this purpose, we propose Visual Contrastive Self-Distillation, namely VCSD, which converts image-content removal into an on-policy self-distillation signal. At each student-generated response prefix, the EMA teacher produces two next-token distributions under the same prompt and prefix – one conditioned on the original image and the other on a content-erased control. Their token-wise log-probability difference highlights candidates whose likelihood is specifically increased by the instance-level visual content. We use this contrast to sharpen the teacher’s original-image distribution within its plausible support, and distill the resulting full-distribution target into the student. Using ViRL39K dataset, VCSD consistently outperforms matched OPSD across Qwen3-VL and Qwen3.5 models. For example, on Qwen3-VL, it improves the seven-benchmark aggregate from 62.27% \rightarrow 67.04% at 2B, 71.30% \rightarrow 73.16% at 4B, and 72.51% \rightarrow 76.26% at 8B. Furthermore, VCSD requires no external teacher, privileged answers, visual evidence signals, reasoning traces, or additional inference-time cost.
[CV-10] SANA-Video 2.0: Hybrid Linear Attention with Attention Residuals for Efficient Video Generation
链接: https://arxiv.org/abs/2607.21553
作者: Junsong Chen,Jincheng Yu,Yitong Li,Shuchen Xue,Haozhe Liu,Jingyu Xin,Yuyang Zhao,Tian Ye,Zhangjie Wu,Zian Wang,Daquan Zhou,Ping Luo,Song Han,Enze Xie
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 13 pages, 9 figures, 5 tables
Abstract:We introduce SANA-Video 2.0, a hybrid video diffusion transformer instantiated at 5B and 14B scales under a unified architecture. Designed to generate high-quality video up to 720p on a single GPU, SANA-Video 2.0 matches full-softmax video DiTs in quality while retaining the favorable long-sequence scaling of linear attention. To avoid quadratic attention throughout, Hybrid Linear-Softmax Attention combines gated linear attention for O(N)-dominated mixing with periodic gated-softmax anchors at a 3:1 ratio, restoring the full-rank token interactions that pure linear attention lacks. To propagate these refreshed representations across depth, Block Attention Residuals (AttnRes) route completed block summaries into later linear layers, enabling anchor-feature reuse and boosting deep-layer effective rank by ~12%. Through from-scratch training, SANA-Video 2.0 learns the complete hybrid directly rather than linearizing pretrained models, with reduced-resolution proxy studies establishing 25% softmax as the optimal quality-efficiency trade-off. With 40-step sampling, SANA-Video 2.0 achieves a VBench score of 84.30 in 13.2s at 480p on a single H100, remaining competitive with far larger softmax video DiTs at a fraction of the latency. Its compiled DiT forward pass is 3.2x faster than a matched full-softmax baseline at 720p/60s, a gap that expands with video duration. Furthermore, full-stack Sol-Engine optimization (kernel fusion, caching, and sparse attention) accelerates this hardware-friendly backbone by a further 3.58x, bringing the 5B pipeline to 13.06s at 720p/5s and making it 120x faster than Wan 2.2-A14B on one H100. Overall, our hybrid design recovers softmax-level expressiveness at substantially reduced cost, unlocking scalable long, high resolution video generation.
[CV-11] UnDA: Unpaired Domain Alignment for Cross-Modal Knowledge Transfer in Medical Imaging
链接: https://arxiv.org/abs/2607.21546
作者: Rafsan Jany,Shadab Tanjeed Ahmad,Ahsan Bulbul,Tahsinul Islam,Md Azam Hossain,Abu Raihan Mostofa Kamal
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multimodal based approaches often outperform single modality approaches in downstream tasks as the different modalities provide complementary information, yet acquiring paired clinical data remains a significant challenge in real world scenarios. While cross-modal knowledge distillation addresses this, existing methods often struggle with large modality gaps and the propagation of noise from uncertain source-domain predictions. To overcome these challenges, we propose UnDA, an anchor-guided framework for unpaired cross-modal distillation. Our approach introduces a backbone-agnostic Alignment Module that extracts semantically structured class tokens via an attention based pooling mechanism. To ensure robust knowledge transfer, we propose Uncertainty-Weighted Optimal Transport (UCT-OT), which dynamically weights feature-level alignment based on prediction confidence, effectively suppressing noisy supervision. Furthermore, a per-class ProtoNCE objective maintains stable prototype memories to enforce global discriminability across unpaired batches. Evaluations on representative segmentation tasks under strictly unpaired settings show consistent improvements in accuracy and boundary precision in the target modality, demonstrating that meaningful structural knowledge can be transferred across heterogeneous data sources without paired datasets.
[CV-12] owards Robust Iris Recognition Through Occlusion Identification and Conditional Diffusion-Based Reconstruction
链接: https://arxiv.org/abs/2607.21545
作者: Kamrul Hasan,Mylene C.Q. Farias,Oleg V. Komogortsev
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by IEEE International Joint Conference on Biometrics (IJCB) 2026
Abstract:Iris recognition is a reliable biometric approach that identifies individuals using the distinctive and stable texture of the iris. However, recognition performance can degrade when discriminative iris texture is partially occluded by eyelids, eyelashes, specular reflections, or other acquisition artifacts. Existing approaches often perform recognition directly on degraded samples or rely only on the remaining visible iris region, which may be inadequate when substantial texture is corrupted. To address this limitation, we propose an occlusion-aware iris recognition framework with three sequential modules: occlusion-type identification, diffusion-based reconstruction, and deep-learning-based recognition. First, a residual 2D CNN-based network determines whether an iris image is non-occluded or belongs to one of the controlled occlusion categories. Second, the occluded image, binary mask, and predicted occlusion type condition a denoising diffusion probabilistic model to reconstruct the corrupted region. Finally, VGG19-HPMNet, a modified VGG19 model with horizontal pyramid mapping, extracts discriminative global and part-wise local iris features for recognition. Experiments on the CASIA-Iris-Thousand dataset under a controlled synthetic-occlusion protocol show that the proposed framework improves iris recognition performance by identifying the occlusion type, reconstructing masked regions, and re-evaluating the restored iris samples.
[CV-13] ElasticTTT: Prior-Preserving Test-Time Tuning for Video Editing
链接: https://arxiv.org/abs/2607.21529
作者: Yueyi Liu,Chi Zhang,Sen Cui,Miao Liu
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Test-Time Tuning (TTT) on pretrained diffusion models has emerged as a powerful paradigm for video editing. However, there exists a foundational mismatch between the distribution-mapping nature of generative models and the single-point optimization of standard TTT. In this paper, we demonstrate that this mismatch triggers \textitPrior Collapse, a degenerate state where the model discards the text conditions and spatial latents, collapsing generations to the source video, or entangling the features of distinct regions. To resolve this, we propose \textbfElasticTTT, a novel framework that preserves the prior generative distribution and rescues generative elasticity. Specifically, we propose \textitTarget Distribution Regularization to prevent sharp memorization minima, \textitContrastive CFG to guide inference away from source biases, and \textitAsynchronous Noise Schedule to preserve unedited regions. Extensive evaluations, supported by theoretical analysis, demonstrate that ElasticTTT successfully preserves the generative prior of the base model, achieving state-of-the-art performance on one-shot video editing.
[CV-14] Boosting Robustness for All-Weather Self-Supervised Depth Estimation in Autonomous Driving
链接: https://arxiv.org/abs/2607.21526
作者: Mengshi Qi,Xiaoyang Bi,Xianlin Zhang,Huadong Ma
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Self-supervised depth estimation is challenging for safe autonomous driving under various adverse weather conditions due to sensor perception degradation. These challenges arise from two main aspects. Firstly, adverse conditions can distort pixel correspondences and violate the assumptions embedded in the self-supervised loss function, leading to erroneous depth predictions. Secondly, while radar is a widely adopted sensor in adverse weather conditions, the sparse distribution of radar points in the Point of View (POV) poses challenges for self-supervised fusion. To address these issues, we introduce a novel self-training pipeline using unpaired real all-weather data through multi-teacher distillation and robust radar fusion. We propose the Uncertainty-Aware Multi-Teacher Distillation method to generate diverse teacher models with different adverse condition inputs, and then employ uncertainty modeling to weigh the knowledge distillation loss. Additionally, we design the POV-BEV Radar Fusion approach, which leverages camera-pixel ray constraints to establish connections between the camera’s Point of View (POV) and the radar’s Bird’s-Eye View (BEV). This approach enables the utilization of denser radar points, effectively capturing the complementary perspectives of both POV and BEV. Extensive quantitative and qualitative experiments demonstrate the robustness of our proposed method on all-weather datasets, achieving state-of-the-art performance. Our code and models are available at this https URL.
[CV-15] xture: Elevating 3D Asset Texture Resolution with a Region-Aware Diffusion Model
链接: https://arxiv.org/abs/2607.21504
作者: Shuaiwei Wang,Shi Li,Jieting Xu,Yuchi Huo,Qi Wang,Wenting Zheng,Rengan Xie
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Numerous 3D assets are discarded due to low texture resolution, while current super-resolution models ignore texture maps and focus on natural images. An efficient and generalizable texture super-resolution model can revitalize a large corpus of aging yet valuable assets across industries such as film and video games. We present Texture++, a novel framework for texture super-resolution, which enhances the low-resolution textures of assets to produce high-resolution, high-quality results. Specifically, we reformulate the task of super-resolution in UV space into performing it across multiple rendered views and merging the outputs. Firstly, to achieve more complete and continuous textures in the view space, we propose an adaptive view selection strategy to integrate textures dispersed across UV texture patches. Furthermore, we introduce a quadtree-based texture region organization method for combining super-resolved textures from different viewpoints, providing masks to distinguish regions that require improvement. Finally, we design a diffusion-based super-resolution model that enhances the texture resolution for specified masked regions, seamlessly integrating with surrounding regions. Through comprehensive evaluations, we demonstrate that our approach yields textures with substantially improved detail and coherence over existing methods.
[CV-16] Recurrent Sinusoidal INRs for Efficient High-Fidelity Representation ECCV2026
链接: https://arxiv.org/abs/2607.21485
作者: Hyunmin Cho,Jaejun Yoo,Kyong Hwan Jin
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to ECCV 2026 (Poster)
Abstract:We study sinusoidal recurrence as an iterative mechanism for harmonic spectral enrichment in implicit neural representations (INRs). Our analysis reveals that sinusoidal activations induce a harmonic line spectrum, providing a spectral account of how recurrent unrolling enriches the effective spectral support. We realize this principle with a shared sinusoidal block that iteratively refines the latent representation. We empirically validate the resulting spectral behavior against feed-forward INRs, non-sinusoidal recurrent variants, and equilibrium-style sinusoidal models. Complementing this analysis, we evaluate the proposed architecture across image and 3D representation tasks. On RGB image benchmarks, our method achieves higher fidelity than feed-forward baselines with fewer parameters and fewer optimization steps, and it further transfers favorably to super-resolution, NeRF, and SDF tasks.
[CV-17] Future Rendering neq Future Surface: A Benchmark and Dataset for Dynamic Surface Reconstruction Beyond the Observed Window
链接: https://arxiv.org/abs/2607.21471
作者: Yukun Shi,Minglun Gong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: See this https URL
Abstract:Dynamic-scene reconstruction is almost always evaluated inside the observed time window, yet deployment settings such as AR overlays, robot interaction, and anticipatory planning need the future surface: the geometry at times beyond those captured. No standard benchmark measures this. We introduce FutureSurf, a controlled diagnostic benchmark and dataset for future-time surface reconstruction that trades scene diversity for exact future ground truth and falsification controls. A method trains on the observed first 75% of a sequence; we score its extracted per-frame surface on the held-out future by Chamfer distance, reporting absolute future CD as the primary score and the future/observed gap as a diagnostic. The dataset contains eight analytically defined controlled motions, including three falsification controls, with exact per-frame ground-truth meshes. We also provide a ground-truth-side recoverability oracle. The release includes split files, scoring code, a benchmark card, and Croissant metadata. On the controlled motions, the DG-Mesh backbone leaves a 2.7-4.1 \times gap even for futures predictable in principle (four of five recoverable from observed motion by a fixed rule), while the falsification controls behave as designed (the surface-invariant motion shows no gap). Beyond the contributed dataset, the gap persists across six animated DG-Mesh asset scenes and a second backbone, Deformable-3DGS (2.0-6.6 \times ; both share a deformation-MLP temporal model). The benchmark also shows that future rendering quality and future-surface accuracy are statistically decoupled, so the novel-view-synthesis metrics the field reports do not track future geometry. The future error is structured, concentrating where the surface moves. The dataset, evaluation toolkit, and scoring code are available on Hugging Face and GitHub (this https URL).
[CV-18] CLUIE: Clustering-Aware Recurrent Propagation with Local Structural Compensation for Underwater Image Enhancement
链接: https://arxiv.org/abs/2607.21467
作者: Kui Jiang,Zefan Feng,Laibin Chang,Yan Luo,Junjun Jiang,Xiaopeng Fan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 13 pages, 12 figures, IEEE Transactions on Image Processing journal paper, code available at this https URL . This paper presents CLUIE, a clustering-aware recurrent RWKV framework for spatially heterogeneous underwater image enhancement, with full-reference/no-reference quantitative comparisons, comprehensive ablation studies and feature visualization for CSDR and DMLP modules
Abstract:Underwater image enhancement remains challenging due to wavelength-dependent light absorption, scattering, and backscattering, which jointly cause color distortion, contrast degradation, and detail loss. Since these degradations vary with scene depth and imaging conditions, different regions within the same image often exhibit heterogeneous degradation patterns and thus require region-adaptive restoration. Although visual RWKV models offer an efficient linear-complexity solution for long-range dependency modeling, their predefined scanning orders are content-agnostic and therefore fail to adapt recurrent state propagation to spatially non-uniform restoration demands. To address this limitation, we propose a Clustering-aware RWKV framework, termed CRWKV, which reformulates the fixed recurrent propagation path of conventional RWKV into a content-adaptive token trajectory. Specifically, we introduce Clustering-aware Semantic Dynamic Reordering (CSDR), which groups tokens according to semantic feature similarity and derives a dynamic traversal order from inter-cluster contextual relations. This design enables WKV states to be accumulated along semantically correlated regions rather than fixed spatial or spectral orders. Since dynamic reordering may disrupt the local continuity of original spatial neighborhoods, we further propose Dark-response Modulated Local Propagation (DMLP), which extracts local structural responses via depth-wise convolution and adaptively modulates their propagation strength using a neighborhood-aware pseudo-dark response map. In this way, local structural cues are compensated before recurrent aggregation while preserving content-adaptive long-range modeling. Extensive experiments on multiple underwater image enhancement benchmarks demonstrate that CRWKV achieves state-of-the-art quantitative performance and superior visual quality.
[CV-19] SPDCN: Strip-based Deformable Convolutional Network for Steel Surface Defect Segmentation
链接: https://arxiv.org/abs/2607.21456
作者: Zhongming Liu,Bingbing Jiang,Guangxin Wan,Xiang Zou
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Steel surface defect segmentation is critical for industrial quality inspection, yet existing methods struggle with elongated, anisotropic defects such as cracks and scratches due to the isotropic receptive fields of standard convolutions and rigid sampling grids that cannot adapt to irregular defect boundaries. To address these limitations, we propose Strip-based Predictor for Deformable Convolutional Networks (SPDCN) with two key innovations. The \textbfFuzzy-enhanced Multi-scale Context Module (FMCM) employs group-wise multi-branch convolutions with an intuitionistic fuzzy channel attention mechanism to adaptively capture multi-scale contextual information across varying defect sizes. The \textbfAdaptive Direction-Aware Deformable Convolution (ADADC) replaces the conventional offset predictor with decoupled horizontal and vertical strip convolutions, enabling the deformable sampling grid to anisotropically align with the principal orientation of elongated defects. Extensive experiments on public steel surface defect benchmarks demonstrate that SPDCN consistently outperforms state-of-the-art methods, achieving 89.60% mIoU on NEU-Seg with only 3.54M parameters. The source code is publicly available at this https URL .
[CV-20] GrainGS: Gradient-Decoupled Gaussian Splatting for Efficient Dynamic Novel View Synthesis
链接: https://arxiv.org/abs/2607.21448
作者: Jiahao He,Yihua Shao,Zhengkai Zhao,Pan Gao,Fei Ma,Jingcai Guo,Hao Tang,Nicu Sebe,Qi Tian
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Dynamic scene reconstruction with 3D Gaussian Splatting requires a balance between fine-grained motion modeling, structural stability, and compact representation. Existing per-primitive methods provide flexible local deformation but often suffer from redundant primitive growth, while anchor-based methods improve spatial regularity at the cost of suppressing locally varying motion. To address these issues, we present GrainGS, a dynamic Gaussian framework that combines a hierarchical anchor scaffold with per-Gaussian deformation. A static warm-up stage first establishes a time-invariant canonical representation from observations across all timestamps. During joint training, a stop-gradient operation blocks the deformation-mediated gradient pathway to the canonical positions while preserving their direct refinement through the reconstruction objective. Each Gaussian then predicts independent temporal offsets for position, rotation, and scale, enabling detailed local motion within a structurally constrained scaffold. A canonical-residual appearance decomposition further models frame-dependent photometric changes without forcing them into geometric deformation. Experiments on synthetic monocular and real-world multiview benchmarks show that GrainGS achieves high reconstruction quality, real-time novel view synthesis, and compact storage. Under the synthetic benchmark setting, it reaches an average peak signal-to-noise ratio of 36.98 decibels, renders at 435.6 frames per second, and requires 4.67 megabytes of storage.
[CV-21] KroQuant: Kronecker-Structured Block Transforms for Efficient Post-Training Quantization of Diffusion Transformers
链接: https://arxiv.org/abs/2607.21446
作者: Yann Bouquet,Alireza Khodamoradi,Kristof Denolf,Mathieu Salzmann
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Post-training quantization (PTQ) of diffusion transformers (DiTs) to W4A4 severely degrades output quality, because activations entering each linear layer contain outliers that 4-bit formats cannot represent. The standard fix applies an invertible linear transform to the activations and its inverse to the weights before quantizing both. Normalization layers between blocks force this transform to run online at every denoising step, making its inference computation cost the binding design constraint. Existing options trade quantization quality for inference cost: per-channel scaling (SmoothQuant) is computationally cheap but impacts the magnitude of the channels, which can harm quantization accuracy; fixed Hadamard transforms yield better quantization accuracy but require large block sizes that incur a high online cost; learned full- d invertible transforms calibrate best but entail a prohibitive dense d \times d matrix multiplication (GEMM) per layer per step. We propose KroQuant, a PTQ method that applies a learned Kronecker-structured invertible transform to each 32-element block of the activation, storing less than half the parameters of per-channel scaling. The block-local structure runs as small tensor-core GEMMs, and on an MI350 GPU the KroQuant quantizer kernel is up to 14% faster than the SmoothQuant kernel. Offline LoRaQ weight calibration then absorbs the residual per-weight quantization error. On PixArt- \Sigma , SANA, and FLUX.1-schnell at W4A4 (MXFP4e2), KroQuant produces outputs closer to the FP reference than SVDQuant and LoRaQ on MJHQ-30K and SDCI, while preserving or improving image quality. Subjects: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2607.21446 [cs.LG] (or arXiv:2607.21446v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2607.21446 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-22] DAPM: UAV Monocular Depth Estimation from Any Height Pitch Roll and FOV
链接: https://arxiv.org/abs/2607.21438
作者: Tong Ling,Wenhui Diao,Yingchao Feng,Hanbo Bi,Zhongyan Hou,Xian Sun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Monocular depth estimation is a fundamental prerequisite for 3D reconstruction and autonomous navigation in Unmanned Aerial Vehicles (UAVs). In practical deployments, UAVs operate under highly dynamic camera poses characterized by continuous variations in height, pitch, roll, and field of view (FOV). Existing monocular depth estimation methods frequently fail to generalize across such diverse perspectives and the expansive scale of depth distributions inherent in aerial scenes. To address these challenges, we establish a quantitative representation of UAV viewing angles through rigorous theoretical analysis, deriving the geometric correspondence between viewing angles and view distances using the ground plane as a reference for observation. Building upon this, we propose Depth Estimation for Any Perspectives Model (DAPM), representing the first monocular framework specifically designed for UAV aerial imagery to jointly estimate camera pose and depth under continuously varying viewpoints. Specifically, we introduce an Ideal Ground Depth (IGD) module that leverages the derived geometric relationships between UAV perspectives and view distances to implement dense camera-pose supervision and enhance depth features. And we further develop a coarse-to-fine Progressive Quantization Bins (PQB) module. By incorporating progressive supervision and hierarchical quantization bins, the PQB module enables robust estimation in complex UAV aerial imagery. To evaluate the proposed framework, we present the UAV Any Perspectives Depth (UAPD) dataset, featuring comprehensive and continuous distributions of pose parameters. Experimental results on UAPD demonstrate that DAPM achieves state-of-the-art performance across both depth and camera-pose estimation metrics. The source code and datasets are available at: this https URL.
[CV-23] Adaptive Identity Anchoring: Closed-Loop Keyframe Placement for Synthetic Paired Supervision in Video Face Swapping
链接: https://arxiv.org/abs/2607.21434
作者: Logan Robbins
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Video face swapping has no natural paired supervision: no real footage exists of one person’s face performing another person’s video. The strongest current answer, DreamID-V’s SyncID-Pipe, mints pairs by replacing the identity in exactly two frames of a real clip – the first and the last – and regenerating the rest from a pose sequence alone. Pose carries no appearance evidence of the swapped-in identity, so over long clips, occlusions, and extreme pose excursions the synthesized identity has a long unanchored span on which to drift; no published ablation examines anchor count or placement. We propose Adaptive Identity Anchoring (AIA): (i) generalize the synthesizer to arbitrary anchor sets, architecturally natural for diffusion-forcing-style transformers where conditioning on a frame is clamping its tokens to zero noise; (ii) place anchors by a closed feedback loop that scores every generated frame against the real reference identity and inserts an image-face-swapped anchor at the worst-scoring frame until the pair passes a threshold or exhausts a budget; (iii) reuse the loop’s verdict as an automatic data filter. A second pathology, the beauty-filter look of over-smoothed skin, has the same root cause: micro-texture, like identity, is priced by none of the pipeline’s objectives. We therefore pair AIA with Reality-Referenced Texture Restoration: matched re-graining from each real frame’s non-face regions, band-split transfer of sub-identity micro-texture from the real footage, and a second, spectral acceptance channel refereed by the footage’s own spectrum. Identity-anchor density, we argue, is a controllable quality dial, and we specify falsifiable experiments – drift-versus-gap curves, uniform-versus-adaptive placement at matched budgets, student training on AIA-minted data, and texture ablations with a human beauty-filter study – that would validate or refute the proposal.
[CV-24] owards Privacy-Preserving Federated Prompt Tuning under Data Heterogeneity: A Subspace-Decomposed Expert Approach ACM-MM2026
链接: https://arxiv.org/abs/2607.21417
作者: Yuhua Wang,Xiaodong Li,Yihao Guo,Yuxiang Jia,Qinnan Zhang,Yifan Sun,Hainan Zhang,Yongxin Tong,Zhiming Zheng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by ACM MM 2026
Abstract:Federated prompt tuning (FPT) enables collaborative adaptation of vision–language models (VLMs) using lightweight prompts. Existing methods often address heterogeneity and privacy through a split-prompt design under local differential privacy (DP), combining a shared prompt for global transfer with private prompts for local adaptation. However, a single shared prompt may over-smooth diverse transferable knowledge, weakening the balance between personalization and generalization. Multi-expert prompts (MEPs) can better capture this diversity, but enlarge the communicated space, increasing DP noise and communication cost while making robust expert composition more difficult. We propose FedSEPT, a privacy-preserving Federated Subspace-decomposed Expert Prompt Tuning. Specifically, we employ Subspace-decomposed Expert Modeling (SEM) to parameterize multiple prompt experts with shared low-rank factors, a fixed public basis, and private residuals, thereby confining communication and DP perturbation to a compact factor space while enabling direct server aggregation in a common coordinate system. We further design Instance-aware Expert Fusion (IEF), which adaptively combines semantically complementary experts via on-device routing and performs efficient logit-level fusion using cached expert-specific text features. Extensive experiments on 11 heterogeneous benchmarks show that, under the same privacy constraints, FedSEPT achieves a better trade-off between local adaptation and global generalization than strong baselines.
[CV-25] GLAM-SLAM: Real-time Gaussian Large-scale Mapping via Flow Densification and Spatial Decomposition IROS2026
链接: https://arxiv.org/abs/2607.21416
作者: Panagiotis Mermigkas,Argyris Manetas,Petros Maragos
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to IROS 2026. Project page: this https URL Code: this https URL
Abstract:Existing Gaussian-splatting-based monocular Simultaneous Localization and Mapping (SLAM) systems are either tailored to short sequences, are not real-time, or suffer from prohibitive GPU memory requirements, limiting their applicability in realistic, long-horizon scenarios. To address this, we present GLAM-SLAM, a real-time, decoupled Gaussian-splatting SLAM system designed for large-scale outdoor scenes. We ensure lightweight tracking using a robust, feature-based SLAM frontend, while for mapping, we adopt a structured, sparse anchor grid representation that ensures scalable operation and maintains scene coherence across long-term sequences. To satisfy the dense initialization requirements of 3D Gaussian Splatting (3DGS), we introduce a geometry-based flow-densification anchoring strategy using epipolar constraints. Furthermore, by treating mapping as a multi-scene problem, we propose a scene-partitioning strategy that introduces a strong spatial inductive bias via MLP initializations to generate localized Gaussians. We evaluate our system on the challenging, long-sequence KITTI Odometry, Oxford RobotCar, and M’alaga datasets. Extensive ablations and comparisons demonstrate a 15% improvement in reconstruction quality over the second-best performer, while maintaining real-time performance and the ability to scale to longer sequences. Code is publicly available for the benefit of the community.
[CV-26] When Are Reasoning -Based Guardrails Not Efficient? ResponseGuard: A Fast Vision-Language Guard for Real-Time Moderation
链接: https://arxiv.org/abs/2607.21401
作者: Dongbin Na
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 8 pages, 6 figures, 3 tables. Project page: this https URL ; Code: this https URL
Abstract:A vision-language AI assistant returns its answer as a stream of generated tokens. Therefore, a safety guard that watches that answer has to keep up with the stream and stop a harmful reply before a user reads it. Recent vision-language guardrails instead generate a chain of thought before they issue a verdict. They believe that step-by-step reasoning yields a safer guard. This design makes the guard heavy and slow, since the model must decode many tokens for harmfulness detection. We pose the question of whether a vision-language guard really needs to reason in order to screen a response. We answer with a guard that has no chain. ResponseGuard reads a harmful verdict from a single pooled representation of the request, the response, and the image in one forward pass. Across a standard multimodal guardrail benchmark, our 2B ResponseGuard outperforms a recent 3B reasoning-based vision-language guard on response harmfulness detection, without any reasoning and at about 150 times lower time cost. On request harmfulness the reasoning guard retains an overall lead, and the remaining gap on both tracks sits on the image-only cells. We observe that the gap may stem from the frozen vision encoders that both designs use rather than from the missing chain. We have also found the reasoning guard directs almost none of its verdict attention to the image. Based on a single-pass detection, ResponseGuard can screen an answer sentence by sentence as it streams and stop a harmful answer before it finishes. For guarding the response of a vision-language model, a calibrated single-pass label may provide a sufficient safety signal. We fully release all source code, trained models, and datasets at this https URL.
[CV-27] DINOde: Continuous Vision-Text Alignment for Open-Vocabulary Semantic Segmentation ECCV2026
链接: https://arxiv.org/abs/2607.21371
作者: Sung-Hoon Yoon,Hoyong Kwon,Changgyoon Oh,Kuk-Jin Yoon
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to ECCV 2026. 27 pages, 8 figures, and 10 tables. Includes supplementary material
Abstract:Open-vocabulary semantic segmentation (OVSS) leverages textual semantics to segment objects beyond predefined categories. While the self-supervised model DINOv3 provides strong structured visual representations, its lack of native textual alignment hinders its direct application to OVSS. To bridge this gap, we propose DINOde, an ODE-based framework that continuously aligns CLIP text embeddings with the DINO visual manifold. Our approach employs two complementary components: (i) Semantic Text Flow (STF), which evolves text embeddings toward the DINO manifold through a continuous ODE trajectory, and (ii) Global Context Flow (GCF), which progressively refines the holistic image representation carried by DINO’s CLS token. To preserve the hyperspherical geometry of the feature space during this evolution, we further introduce Velocity Tangent Projection, which constrains the learned velocity field to the tangent space. By modeling alignment as a continuous trajectory, DINOde avoids the manifold entanglement inherent in discrete MLP projections and yields more robust cross-modal alignment. Extensive experiments demonstrate that DINOde consistently outperforms existing methods and achieves state-of-the-art performance across multiple OVSS benchmarks. The code is available at this https URL.
[CV-28] ASTRA-Net: Anatomy-Specific Transfer and Representation Alignment for Drug-Induced Sleep Endoscopy Segmentation
链接: https://arxiv.org/abs/2607.21370
作者: Suhua Sun,Yuqiao Wang,Sheng Liu,Rui Fan,Jiajun Wang,Ruoyan Xu,Yixin Chen,Tao Li,Yan Yan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 20 pages, 6 figures, 5 tables
Abstract:Quantitative drug-induced sleep endoscopy (DISE) requires reliable airway boundaries at specific anatomical levels. Pixel-level DISE annotations are scarce, and manual contouring limits the scalability of quantitative assessment. To address this limitation, we developed ASTRA-Net for known-plane DISE segmentation with limited real annotations. Stage 1 aligned intermediate ConvNeXt-Base representations from 14,250 unlabeled virtual endoscopy frames derived from computed tomography and real DISE frames. Virtual images were used only for feature alignment. Stage 2 fine-tuned four independent UNet++ decoders on 401 real annotated frames. Structured zero-mask supervision constrained incompatible plane outputs and invalid frames. Six alignment configurations used maximum mean discrepancy, domain adversarial learning, or both objectives. On a hold-out evaluation set of 100 frames, the five-model MMD-only segmentation ensemble achieved a mean Dice of 0.8927, with a 95% image-level bootstrap interval of 0.8631 to 0.9160. The mean intersection over union was 0.8239. A classification- enabled variant of the same alignment configuration reached a restricted four-plane top-1 accuracy of 0.92 on the same hold-out frames. These results indicate that ASTRA-Net can support frame-level, plane-specific DISE boundary delineation when real annotations are limited.
[CV-29] Incremental Optimal Assignment for Real-Time Crowd Tracking
链接: https://arxiv.org/abs/2607.21368
作者: Ismail H. Toroslu
类目: Computer Vision and Pattern Recognition (cs.CV); Data Structures and Algorithms (cs.DS)
备注:
Abstract:Multi-object tracking in dense crowds requires solving a bipartite assignment problem between detections and trajectories at every video frame. The classical Hungarian algorithm solves this in O(N^3) time, which becomes a bottleneck for large scenes with hundreds of people. We propose an \emphincremental assignment algorithm that exploits the block-sparse structure of crowd tracking cost matrices — dense within each crowd cluster, near-zero between clusters. We compute the exact same optimal N \times N assignment as the Hungarian algorithm, but via an incremental strategy: we add one person at a time, exploiting the fact that after step n-1 the dual potentials are \emphexactly optimal for the (n-1)\times(n-1) subproblem — a strictly stronger condition than the intermediate feasibility maintained by the Hungarian algorithm during its N outer iterations. Each new step therefore requires only a single augmenting path search from a certified optimal starting point. This avoids repeated full-matrix scans while guaranteeing an identical globally optimal result. A diagonal-reordering invariant keeps the data structure compact and cache-friendly. On realistic crowd benchmarks with N \in [200, 5000] people organised into dense clusters, our algorithm achieves \textbf3.7–6.5 \times speedup over the Hungarian baseline while producing provably optimal matchings identical to those of Hungarian. The speedup grows with N and remains stable beyond N=3000 , making the method especially attractive for large-scale crowd scenes such as stadium exits and mass public events.
[CV-30] Quality-Aware Multimodal Fusion Reveals Implicit Identity in Valence-Arousal Features
链接: https://arxiv.org/abs/2607.21347
作者: Jisu Kim,Benjamin S. Riggan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 3 figures, 6 tables. Accepted for publication at IEEE International Joint Conference on Biometrics (IJCB), 2026
Abstract:Conventional face recognition relies on static appearance cues and degrades in unconstrained settings with expression variation, occlusion, and poor lighting. We hypothesize that audiovisual expression dynamics carry identity-discriminative information complementary to static appearance, and that extracting this signal requires multimodal representations robust to the variable input quality of in-the-wild video. To learn such representations, we cast multimodal valence-arousal (VA) estimation as a pretext task and propose Quality-Aware Adaptive Fusion (QAAF), which estimates per-sample, per-modality reliability and adapts each modality’s contribution through learned soft gating and a quality-dependent dropout. For the problem of VA estimation, QAAF achieves an average Concordance Correlation Coefficient (CCC) of 0.472 via late fusion ensembling on Aff-wild2, improving over a baseline ensemble under the same setting (0.415) as well as a single-backbone baseline (0.288). Furthermore, the proposed QAAF demonstrates greater resilience to unavailable modalities, with only a 7.5-34.4% relative decrease in CCC when one modality is missing. We then probe whether these VA-trained features encode identity without identity-specific training. On AFEW-VA (67 actors) and YTF (1,595 subjects), VA-trained backbone features rank first among evaluated soft biometric methods, and score-level fusion with ArcFace lowers EER on both datasets (0.022 to 0.021 on AFEW-VA, 0.106 to 0.104 on YTF), correcting 68.2% of ArcFace’s false accepts on AFEW-VA. These findings establish multimodal VA estimation as a soft biometric modality complementary to conventional face recognition.
[CV-31] M3-Gen: Interpretable Multimodal Generation of Gene Expression Profiles Using Clinical and Imaging Data
链接: https://arxiv.org/abs/2607.21343
作者: Francesca Pia Panaccione,Carlo Sgaravatti,Marco Venere
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注: 15 pages, 6 figures
Abstract:Integrating heterogeneous biomedical data, including clinical metadata, histopathology images, and molecular profiles, is crucial for comprehensive disease understanding. However, gene expression data acquisition remains constrained by high costs and privacy concerns, limiting its use in multimodal research and AI-driven applications. We present MultiModal Molecular Generation (M ^3 -Gen), a novel framework for the generation of gene expression profiles by conditioning a Generative Adversarial Network on histopathology images and clinical metadata. M ^3 -Gen learns a unified latent representation from the clinical variables and the images, leveraging contrastive learning, and exploits the embeddings of the two modalities to guide a generative model in producing biologically coherent gene expression profiles. Evaluations on the TCGA dataset demonstrate that M ^3 -Gen generates realistic and functionally meaningful gene expression data. Importantly, by integrating multiple modalities in an attention-based mechanism, M ^3 -Gen provides intrinsic explainability: it allows the identification of which regions of the histopathology images most strongly influenced the generation of specific gene expression profiles, making the model’s decisions interpretable by design.
[CV-32] SlerpFlow: Spherical Trajectory Correction for Rectified Flow Inversion ICML2026
链接: https://arxiv.org/abs/2607.21326
作者: Wenbin Duan,Yan Shu,Zhuoyuan Fu,Fangmin Zhao,Yan Li,Yaru Zhao,Binyang Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 16 pages. Accepted at ICML 2026
Abstract:Rectified-flow-based diffusion transformers, particularly FLUX, have demonstrated outstanding performance in high-quality image generation. However, achieving fast and accurate inversion–transforming images back to latent noise for faithful reconstruction and editing–remains a challenging bottleneck due to the discretization errors of linear solvers. This paper introduces SlerpFlow, a straightforward yet highly effective zero-shot approach that unlocks the full potential of FLUX for high-fidelity inversion and editing. Unlike existing approaches (e.g., RF-Solver) that rely on complex numerical approximations such as high-order Taylor expansions to correct trajectory errors, we present a geometric view based on the Manifold Hypothesis: the empirically observed trajectory curvature is not a numerical artifact, but rather serves as a necessary “centripetal force” that constrains the flow to remain on the data manifold. Guided by this insight, SlerpFlow integrates Spherical Linear Interpolation (Slerp) to rectify flow velocity directions on the hypersphere, strictly adhering to the intrinsic curvature of the latent space. Crucially, by caching the corrected velocity for subsequent steps, SlerpFlow achieves high-precision inversion while maintaining the computational efficiency of a first-order Euler solver. Extensive experiments on FLUX-based reconstruction and editing tasks demonstrate that SlerpFlow improves reconstruction fidelity and achieves stronger semantic alignment in editing without requiring additional training. Code is available at this https URL.
[CV-33] PC-Edit: Prompt-Contrastive Region Discovery and Region-Guided Editing
链接: https://arxiv.org/abs/2607.21318
作者: Jian Zhang,Zhijun Zhang
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Replacing an object with one that differs in category or shape requires complete source removal, natural target formation unconstrained by the source silhouette, and preservation of unrelated content. Existing training-free editors either localize edits from terminal predictions under source and target prompts or preserve unrelated content through spatially unselective source-feature reuse without explicit region discovery. Before reaching the terminal predictions, prompt-induced semantic differences undergo additional network transformations that may obscure their spatial localization, reducing localization precision. Spatially unselective feature reuse forces a trade-off between edit completeness and background preservation. Therefore, we propose PC-Edit, a prompt-contrastive framework for training-free MM-DiT editing. PC-Edit contrasts the image-token attention outputs under the source and target prompts, capturing prompt-induced semantic differences directly where text-conditioned information is delivered to image tokens. The same contrast identifies a source-erasure region during inversion and a target-emergence region during denoising. Their union suppresses source remnants while allowing the target object to form naturally. PC-Edit further couples region discovery and background preservation within each sampling step by estimating the current edit region from preceding attention blocks and immediately injecting cached source K/V features outside it in subsequent blocks, thereby protecting unrelated content before the latent update. Experiments on PIE-Bench and our EditRegion-Bench, with human-verified edit-region annotations for single- and multi-object addition and replacement, show that PC-Edit achieves the best editing quality and background preservation among methods without user-specified edit regions.
[CV-34] Unlearning Under Imbalance: Benchmarking Fairness in Multimodal LLM Unlearning
链接: https://arxiv.org/abs/2607.21300
作者: Lorenzo Orsingher,Thomas De Min,Massimiliano Mancini,Davide Talon,Elisa Ricci
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 33 pages
Abstract:Machine unlearning has emerged as a tool for removing personal data from trained models to comply with recent AI regulations. To evaluate unlearning effectiveness in multimodal large language models (MLLMs), prior works fine-tune models on fictitious identities, simulating unlearning requests on subsets of these IDs, which are typically uniformly distributed. However, in realistic scenarios, people from different demographic groups may request to be unlearned at different frequencies, potentially altering the model’s internal beliefs for these groups and leading to biased behaviors. To fill this gap, we propose FAIRGET, the first Visual Question Answering benchmark that evaluates unlearning under unbalanced, realistic, forget requests. These requests are designed to simulate multiple realistic scenarios, ranging from simple to challenging settings, that lead to biased unlearned models if fairness is not accounted for. Additionally, we propose FAUN, the first unlearning algorithm for MLLMs that forgets unlearning data while preserving model fairness. FAUN exploits a bias-aware activation steering mechanism to unlearn identities while accounting for the unbalanced nature of the forget data. Experiments on FAIRGET and the established FIUBench demonstrate our method’s superiority both in unlearning quality and fairness.
[CV-35] HGeo-TopoMap: Boosting Topological Mapping with Hierarchical Geometric Priors
链接: https://arxiv.org/abs/2607.21281
作者: Siyu Li,Kunyu Peng,Di Wen,Beiping Hou,Zhiyong Li,Kailun Yang
类目: Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO); Image and Video Processing (eess.IV)
备注: The source code and model weights will be made publicly available at this https URL
Abstract:Topological maps are key outputs of autonomous driving perception systems, delivering essential road information for path planning. They identify instances such as centerlines and traffic signs, along with their connectivity relationships. Due to the lack of explicit markings for centerlines in real-world environments, the detection of centerline instances remains a significant challenge. To tackle this problem, we propose HGeo-TopoMap, which leverages an explicit prior map and implicit spatial relations to hierarchically boost topological mapping. First, a geometric adaptive learning module is designed for the road structure map obtained via inverse perspective mapping. This module discretely encodes semantic and spatial features from the map, followed by a prior-mask attention mechanism that selectively focuses on informative regions. Then, a geometric consistency learning module is devised, which leverages the geometric properties and spatial relationships of centerlines. Built on the geometry-aware decoder, it enforces spatial consistency by aligning features of centerline instances with identical geometric orientations. The proposed method is evaluated on the OpenLane-V2 dataset across the centerline, lane segment, and robustness benchmarks. Beyond substantial improvements in topological mapping accuracy, the proposed method offers the benefit of enhanced robustness, consistently outperforming baselines under both standard and challenging conditions. The source code and model weights will be made publicly available at this https URL.
[CV-36] Flash EQ-Linear: Accelerating Equivariant Linear Layers via Group-wise Discrete Fourier Transform
链接: https://arxiv.org/abs/2607.21271
作者: Zhongchen Zhao,Jixin Wang,Qi Xie,Hui Lin,Lei Zhang,Deyu Meng,Zongben Xu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Equivariant networks embed geometric symmetries as structural priors through weight sharing, achieving remarkable parameter efficiency across vision tasks. However, this parameter efficiency does not translate into compute efficiency: existing implementations unroll the structured weights into dense matrices and dispatch them to generic dense kernels, so the FLOPs of an equivariant layer are no smaller than those of a non-equivariant counterpart. In this paper, we observe that the equivariant linear (EQ-Linear) layer—the most fundamental and frequently used module in modern equivariant architectures—is essentially a circular convolution along the group dimension composed with a linear transform along the channel dimension. Building on this observation, we propose Flash EQ-Linear, an exact acceleration algorithm that reduces the complexity from \mathcalO(NDC) to \mathcalO(NDC/T) by combining the Fourier convolution theorem along the group dimension with the conjugate symmetry of the real DFT. We further provide dedicated CUDA kernels for Flash EQ-Linear, covering both forward and backward passes and both FP32 and FP16 precision. At the operator level, Flash EQ-Linear achieves up to 2\times forward speedup over PyTorch’s this http URL; at the network level, Flash EQ-ViT and Flash EQ-Swin achieve up to 1.7\times end-to-end speedup over both equivariant and non-equivariant baselines. To our knowledge, this is the first time equivariant networks strictly dominate their non-equivariant counterparts along all three axes simultaneously: accuracy, parameter efficiency, and inference this http URL is available at this https URL.
[CV-37] Detectors Learn the Wrong Thing: Shortcut-Resistant Adversarial Training Against Physically Realizable Attacks
链接: https://arxiv.org/abs/2607.21243
作者: Yuanhao Huang,Yilong Ren,Jinlei Wang,Xuesong Bai,Zheng Zhang,Haiyang Yu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:AI-enabled visual perception systems are increasingly deployed in intelligent transportation infrastructure and autonomous vehicle related applications. However, physically realizable adversarial appearances pose a significant reliability challenge for these safety-critical systems. Adversarial training is effective, but repeated co-occurrence between adversarial texture and positive person instances can cause detectors to treat the texture itself as evidence of object presence, forming a patch texture shortcut. The detector may then treat texture as evidence for the target, causing false detections on texture-only inputs and weakening cross attack generalisation. We propose InsCAT, an instance-level contrastive adversarial training framework that prevents detectors from using adversarial texture as an independent decision cue. SICA aligns adversarial person features with matched clean features and separates them from texture-only negatives, while ROPO and Guard maintain online attack pressure and coordinate training. We evaluate eight independently generated attack textures on rendered nuScenes, INRIAPerson, printed garments, and three detector families. InsCAT achieves an average attack AP of 82.3% on rendered nuScenes, exceeding the strongest baseline by 11.1 this http URL to AT-Mix, texture FPR decreases from 46.9% to 7.3%. Physical tests yield an F1 score of 96.6% and an FPR of 1.8%. Consistent gains across separately trained detectors demonstrate applicability across architectures with direct inference. The findings show that robust physical detection depends on preserving target related evidence while preventing adversarial texture from becoming an independent decision cu
[CV-38] Stokes-Informed Diffusion for Robust Linear Polarization Estimation
链接: https://arxiv.org/abs/2607.21239
作者: Yidong Luo,Chenggong Li,Yuchao Feng,Boxin Shi,Junchao Zhang,Xin Yuan
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Polarization cues benefit applications such as material detection and de-reflection, yet acquiring them typically requires dedicated hardware. This motivates us to estimate the linear polarization from a single RGB image. However, the task is inherently ill-posed, with the Angle of Polarization (AoP) becoming particularly unstable in weak polarization regions, where the polarimetric signal is overwhelmed by noise, leading to erratic angle estimates. To address these limitations, we propose GenPolar, a Stokes-informed diffusion framework grounded in the Mueller formalism from an intensity observation. Specifically, GenPolar predicts channel-wise linear Stokes components (S1,S2) from intensity S0, from which degree of linear polarization (DoLP) and AoP are analytically derived; AoP is further supervised with an observability-aware loss. In addition, to enable efficient and high-fidelity inference, we adopt a two-stage training strategy. Firstly, a multi-step conditional diffusion model is trained with a physics-based loss. Subsequently, we distill it into a one-step generator, which further supports stable Low-Rank Adaptation (LoRA) of the VAE encoder to mitigate domain-specific autoencoding bias. Extensive experiments across rotating-polarizer, division-of-focal-plane, and hybrid datasets demonstrate that GenPolar achieves state-of-the-art performance in both DoLP fidelity and AoP stability. Crucially, these improvements translate to significant and consistent gains in downstream applications, including material detection and de-reflection.
[CV-39] -STAR: A Large-Scale Benchmark for Spatio-Temporal Panoptic Scene Graph Generation in Satellite Video
链接: https://arxiv.org/abs/2607.21228
作者: Linlin Wang,Xue Yang,Zhihuang Zhou,Zhenyu Zhong,Ruiyuan Zhang,Yansheng Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 17 pages, 8 figures
Abstract:Structured understanding of satellite video is essential for advancing dynamic geospatial scene analysis from low-level perception to high-level cognition. To move beyond object-centric perception, this paper introduces spatio-temporal panoptic scene graph generation (TPSG) in satellite video as a new benchmark task. TPSG aims to generate a structured graph composed of a set of triplets subject, relationship, object with explicit temporal spans, thereby describing dynamic geospatial scenes by jointly modeling identity-consistent instance masks and spatio-temporal relationships among panoptic scene elements. However, there is still no dedicated dataset for TPSG in satellite video. Moreover, TPSG in satellite video is intrinsically challenging, as objects are often small and weakly textured, cross-frame association is easily disrupted by occlusion and background clutter, and relationship semantics are highly coupled with spatial structure and temporal evolution. Consequently, TPSG models developed for natural videos are not directly applicable to satellite video. This paper presents T-STAR, a large-scale benchmark dataset for TPSG in satellite video, comprising over 1.1 million instance masks and over 3.8 million spatio-temporal triplets across 39 fine-grained object categories and 70 fine-grained relationship categories. To enable TPSG in satellite video, we propose a unified framework to enhance cross-frame instance consistency and spatio-temporal relationship prediction. Extensive experiments demonstrate the significance of T-STAR and the effectiveness of the proposed framework, establishing a strong benchmark for future research on structured satellite video understanding. The dataset and code are available at this https URL.
[CV-40] DART: A Degradation-Aware Recurrent Transformer for Archival Film Restoration
链接: https://arxiv.org/abs/2607.21219
作者: Mikołaj Jastrzębski,Wojciech Kozłowski,Kamil Adamczewski
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 16 pages, 6 figures, 4 tables
Abstract:Archival film restoration is a challenging problem because historical footage contains compound degradations such as scratches, dust, blur, noise, flicker, and photometric aging, while clean reference videos are unavailable. Existing video restoration methods largely treat these degradations implicitly, reconstructing frames without explicit knowledge of where damage occurs or how severe it is. We propose DART, a degradation-aware recurrent transformer for archival film restoration. DART predicts and propagates a soft defect mask through time, using it to guide temporal fusion and condition the restoration network on both damage location and severity. This makes the restoration process explicitly aware of film artifacts rather than relying only on reconstruction losses. Experiments on real archival benchmarks show that DART improves no-reference perceptual quality over prior restoration architectures while remaining compact and efficient, producing cleaner and more temporally consistent restorations of structured film damage.
[CV-41] Learning-based Seam Correspondence Reconstruction in Sewing Patterns
链接: https://arxiv.org/abs/2607.21213
作者: Zhendong Wang,Jintong Wang,Chen Liu,Yao Jin,Ligang Liu,Huamin Wang
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注:
Abstract:Digital sewing patterns typically consist of disjoint 2D panels without explicit stitch annotations, making downstream 3D modeling reliant on labor-intensive expert specification. In this paper, we present a graph-based learning framework that reconstructs two-level stitching information, coarse panel connectivity and fine-grained seam correspondence, from 2D panel geometry alone. At the coarse level, panel connectivity is inferred by predicting panel semantics associated with anatomical body regions, enforcing consistency with body structure and garment design conventions. Based on the reconstructed panel graph, fine-grained seam correspondences between panel pairs are inferred by learning latent edge representations that jointly encode local seam geometry and global garment context through graph message passing. The resulting edge embeddings are subsequently decoded into detailed seam correspondences. Our method supports complex sewing-pattern topologies, including many-to-one correspondences, intra-panel seams, and curved seams. Experiments demonstrate high stitching accuracy and strong generalization across garment styles.
[CV-42] Physics-Informed Deep Learning Model for Cross-Modality Super-Resolution in Fluorescence Microscopy
链接: https://arxiv.org/abs/2607.21190
作者: Mohammad Soltaninezhad,Elena Corbetta,Francisco Paez Larios,Paul M. Jordan,Oliver Werz,Christian Eggeling,Thomas Bocklitz
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Cross-modality image translation offers a route to super-resolution fluorescence microscopy from low-resolution images while reducing phototoxicity and instrumentation demands. However, purely data-driven models can produce visually plausible outputs that are inconsistent with optical image formation. Here, we propose a physics-informed generative adversarial network for confocal-to-STED image translation that incorporates microscope-specific point spread function information into the training objective. Simulated and experimentally measured PSFs were evaluated using a limited paired confocal-STED dataset of TOM20-labeled mitochondria in human primary M2 macrophages acquired across different experimental days. Performance was assessed using reference-based and non-reference-based image-quality metrics, together with complementary frequency- and distribution-sensitive analyses. The no-reference metrics probed physics-relevant image properties, including spatial-frequency content, contrast, and signal-to-noise behavior. PSF-guided models improved structural fidelity, reduced local deviations, and achieved closer agreement with STED references than non-PSF baselines, particularly in frequency-domain analyses. These results demonstrate that optical priors can improve the structural fidelity and physical plausibility of generative microscopy models for cross-modality super-resolution imaging.
[CV-43] Out of Sight Still in Mind: Token Compression for Omni-LLM s
链接: https://arxiv.org/abs/2607.21179
作者: Suho Yoo,Youngjoon Jang,Hyebin Cho,Joon Son Chung
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint
Abstract:The goal of this paper is to reduce the input token cost of Omni-modal large language models (Omni-LLMs) at inference time. Omni-LLMs reason jointly over audio, video and text, but the cost of the three streams is highly unbalanced: visual tokens account for the vast majority of the input, and are highly redundant. In this paper, we propose ReMo, a training-free framework that compresses visual tokens by redistributing their information across modalities: a visual token is kept only if its information appears nowhere else. ReMo achieves this in two ways: (i) it aligns audio and video in a common embedding space, and removes visual tokens already explained by the audio or by other visual tokens; and (ii) it replaces object-level visual tokens with compact text proxies, short descriptions of each object and its location, conveying the same content in far fewer tokens. On Qwen2.5-Omni at two model scales, ReMo removes 54% of the input tokens with no loss in accuracy. Indeed, it slightly exceeds the full-token model, reaching 101.2% and 101.3% of its average accuracy over five audio-visual benchmarks.
[CV-44] Decoupling Cross-Modality Manifold Discrepancy: Leverag ing Visible Diffusion Priors for Infrared Super-Resolution ACM-MM2026
链接: https://arxiv.org/abs/2607.21174
作者: Yunpeng Hua,Hongwei Yu,Jiawei Li,Qiankun Liu,Huimin Ma,Jiansheng Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to ACM Multimedia 2026 (ACM MM 2026). Code: this https URL
Abstract:Infrared image super-resolution (IISR) mitigates the limitations imposed by low spatial resolution. Existing methods have recognized that IISR should preserve consistency in global distribution and structural information while enhancing image clarity. However, these methods are either insufficient or overly intrusive, a problem that becomes even more pronounced in diffusion-based models. To address these issues, we propose a dual-path diffusion-based framework for IISR, termed Shift-IISR. The proposed method is designed to improve the consistency of IISR results while preserving the generative capacity of diffusion models. Specifically, we develop a Global Representation Modulation (GRM) module to extract modality-specific information from infrared imagery and guide the global distribution of the diffusion model toward the ground truth. In addition, we introduce a Local Structure Refinement (LSR) module to encourage the model to focus on structural information at each step of the iterative denoising process. Extensive experiments demonstrate that the proposed method effectively improves distributional and structural consistency while maintaining competitive super-resolution performance. The source code of the proposed Shift-IISR can be available at this https URL.
[CV-45] CRAG -MM-Diagnostics: Enabling Stage-Wise Analysis of Knowledge-Intensive VQA ECCV2026
链接: https://arxiv.org/abs/2607.21155
作者: Hanseok Oh,Parishad BehnamGhader,Benno Krojer,Hyunji Lee,Paul Liang,Siva Reddy,Verna Dankers
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to ECCV 2026
Abstract:Knowledge-Intensive Visual Question Answering (KI-VQA) benchmarks evaluate Vision-Language Models (VLMs) as multimodal knowledge assistants by requiring external information beyond a provided image to answer questions. KI-VQA involves multiple sub-problems -referring expression understanding, visual grounding, object recognition, knowledge retrieval, and reasoning-yet existing benchmarks typically report only end-task accuracy, obscuring where failures arise. To analyze the full KI-VQA pipeline, we introduce CRAG-MM-Diagnostics, a diagnostic benchmark with stage-wise data annotations that isolate 1) language-based visual grounding, 2) object identification, and 3) knowledge retrieval and reasoning. We evaluate fully parametric and retrieval-augmented VLMs, providing fine-grained analyses using newly collected metadata, such as target ROIs, entity names, and visual complexity scores. Our results point to knowledge retrieval and reasoning as the primary bottleneck, but also highlight issues in the other parts of the KI-VQA pipeline, such as the fact that VLMs struggle with target object identification or that image retrievers struggle to integrate textual cues. These findings expose fundamental limitations in current KI-VQA systems and motivate stage-aware evaluation. We, lastly, leverage these findings to propose a grounded bimodal RAG pipeline that integrates a visual grounding module to crop targets before image retrieval, boosting GPT-5 and Qwen’s respective accuracies by 13.3 and 8.5 percentage points.
[CV-46] DTIF: Robust Loop Closure Detection via Delaunay Triangle Topology in Complex Forests
链接: https://arxiv.org/abs/2607.21138
作者: Xin Zhao,Jianping Li,Qin Zou,Fuxun Liang,Zhen Dong,Bisheng Yang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 19 pages, 6 figures, 4 tables. Submitted to IEEE Transactions on Geoscience and Remote Sensing
Abstract:Accurate forest inventory and large-scale mapping are essential for ecosystem monitoring and sustainable forest management. Multiple low-cost edge platforms enable efficient large-area data acquisition, but merging independently constructed local maps in GNSS-denied understory environments still requires initialization-free loop closure detection and global registration. This task is challenging because low-cost LiDAR point clouds are sparse and noisy, while repetitive trunk layouts and the lack of distinctive geometric landmarks lead to severe perceptual aliasing and false correspondences. To address these issues, we propose DTIF (Delaunay Triangulation in Forests), a lightweight trunk-topology-based framework for forest loop closure detection and global registration. Tree trunks are first extracted as stable landmarks and encoded using a Delaunay topology for compact scene representation. Candidate submaps are then screened using edge-length and radius statistics, followed by edge–radius consistency verification and strong/weak vertex support aggregation to construct weighted vertex correspondences. Finally, topology-derived reliability weights are incorporated into a decoupled robust pose estimator that separately estimates yaw, horizontal translation, and elevation translation under gravity alignment. Experiments on simulated and real-world forest datasets demonstrate that DTIF achieves accurate registration with low computational overhead, providing a favorable balance among robustness, efficiency, and deployability on resource-constrained edge platforms.
[CV-47] Safety-oriented sidewalk and road segmentation for smartphone-based assistive navigation
链接: https://arxiv.org/abs/2607.21137
作者: Hakan Calim,Anamaria Dumitrescu,Adarsh Bhandary Panambur,Huzaifa Asif,Andreas Maier
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注: 17 pages, 4 figures, 3 tables. Submitted to Assistive Technology
Abstract:Independent sidewalk mobility is essential for blind and visually impaired pedestrians (BVIPs), yet smartphone-based assistive navigation requires perception models that distinguish walkable sidewalks from adjacent unsafe regions. This study presents a safety-oriented semantic segmentation framework for future mobile guidance. We introduce SENSATION-DS, a chest-height pedestrian-view dataset with 2,752 image-mask pairs and nine-class navigation-relevant taxonomy. External urban and sidewalk datasets were harmonized to this label space, and five segmentation architectures were evaluated using staged target-domain adaptation with mask-conditioned synthetic images and Segment Anything Model 2 (SAM2) pseudo-labels. Models were assessed using mean Intersection over Union (mIoU), road- and sidewalk-specific metrics, Road-as-Sidewalk Error Rate as a proxy false-safe measure, and Android Open Neural Network Exchange benchmarking. Synthetic augmentation generally improved segmentation accuracy, whereas SAM2 pseudo-labels more consistently reduced Road-as-Sidewalk errors. UPerNet-MobileNetV3 achieved the highest offline mIoU (0.715 +/- 0.006), while DeepLabV3Plus-MobileNetV3 achieved the lowest Road-as-Sidewalk Error Rate (0.079) and highest Android runtime at 512x384 (7.383 FPS). These results show that assistive sidewalk perception should be evaluated jointly by segmentation accuracy, proxy false-safe behavior, and smartphone deployment feasibility, while real-world benefit requires validation with BVIP users. This evaluation supports selecting models that balance accurate perception, conservative error behavior, and practical runtime.
[CV-48] Causal-Agent IR: Self-Evolving Causal Memory for Adaptive Image Restoration Agents
链接: https://arxiv.org/abs/2607.21125
作者: Hu Gao,Yulong Chen,Lizhuang Ma
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Image restoration agents have recently emerged as a flexible paradigm for handling diverse and unpredictable degradations in real-world scenarios. Existing agents typically formulate restoration as a tool-using process, where the agent perceives degradations, searches candidate tools, executes restoration operations, and revises the plan through reflection or rollback. However, their knowledge is often stored as static tool descriptions, manually defined degradation priors, or unstructured textual summaries, which limits the accumulation, verification, revision, and forgetting of restoration knowledge over long-term experience. In this paper, we propose Causal-AgentIR, a hierarchical multi-agent framework with self-evolving causal memory for collective image restoration intelligence. Instead of representing restoration experience as isolated textual records, Causal-AgentIR organizes degradation patterns, image regions, restoration tools, actions, quality changes, and user preferences into a structured causal memory graph. This graph supports graph-based retrieval and multi-hop causal reasoning, enabling agents to infer how specific restoration operations or tool sequences affect restoration quality under different degradation conditions. The framework further organizes multiple agents into a collaborative system, including planning, degradation analysis, tool expertise, causal memory reasoning, outcome critique, and memory curation. Through this design, restoration experience can be added, updated, merged, reinforced, ignored, or discarded according to observed quality changes and feedback, allowing the agent to maintain reliable and transferable restoration knowledge. Extensive experiments demonstrate the effectiveness of the proposed framework. Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2607.21125 [cs.CV] (or arXiv:2607.21125v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2607.21125 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-49] he Second LoViF 2026 Challenge on Real-World All-in-One Image Restoration: Methods and Results ECCV2026
链接: https://arxiv.org/abs/2607.21118
作者: Xiang Chen,Hao Li,Jiangxin Dong,Jinshan Pan,Xin Li,Hongbo Ding,Junpeng Jiang,Xingyu Qiu,Yilian Zhong,Yuxiang Chen,Shibo Yin,Zixuan Huang,Yushun Fang,Xilei Zhu,Yahui Wang,Chen Lu,Xiaodong Zhou,Qingyue Cao,Changwei Gong,Jingyun Liu,Xingchen Yi,Hansen Shi,Ruiyi Liu,Jirui Xie,Tao Liu,Wenzhuo Ma,Hongzhen Li,Yongyong Chen,Zheng Zhou,Jingyong Su,Jie Liu,Haijin Zeng,Cheng Li,Peishuai Zha,Ziyi Wang,Jian Tang,Yan Chen,Long Bao,Heng Sun,Jiyuan Zhang,Shuai Liu,Wei Ding,Chengjun Guo,Yibin Huang,Xiaotao Wang,Dongqing Zou,Lei Lei,Xiaofeng Wang,Xiao Liu,Yulin Wu,Yuhan Zhao,Shurui Peng,Chao Ren,Yu-Kai Wang,Kosuke Shigematsu,Asuka Shin,Rong-Lin Jian,Cheng-Jun Kang,Jin-Hui Jiang,Jialin Zhou,Kuo Yuan,Songyu Zhang,E B Benson,Ashfaq Hussain,Pruthvikanth AC,Qirui Chen,Jinyuan Chen,Jun Zhang,Xu Zhang,Xuhui Cao,Jiaqi Ma,Laibin Chang,Yuchun Miao,Yichu Xu,Yuanzhi Yao,Shi Chen,Yuning Cui,Huan Zhang,Lefei Zhang,Saeed Ahmad,Ik Hyun Lee,Jun Young Park,Ji Hwan Yoon,Shangquan Sun,Behrooz Nobahar-Moghanlou,Majid Edalatjou,Karim Shahi-Niyar,Ruibo Zhang,Dexiang Hong,Xinyan Liu,Shengeng Tang,Weidong Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ECCV 2026 Workshops; this https URL
Abstract:This paper presents a review of the second LoViF Challenge on Real-World All-in-One Image Restoration. The challenge aims to advance unified image restoration under diverse real-world degradation conditions, including blur, low-light, haze, rain, and snow. It provides a common benchmark for evaluating the restoration accuracy, robustness, and generalization capability of models across multiple degradation categories within a unified framework. The competition attracted 158 registered participants, and 20 teams were included in the final ranking after their submitted results were successfully reproduced and verified. This report provides a comprehensive analysis of the submitted solutions and corresponding results, highlighting recent advances in real-world all-in-one image restoration. The summarized methods and empirical findings reveal effective design strategies and establish an updated benchmark for future research in real-world low-level vision.
[CV-50] HalluScope: Fine-grained Hallucination Diagnosis for Multimodal Large Language Models ACM-MM2026
链接: https://arxiv.org/abs/2607.21105
作者: Weilin Jin,Mingyu Wang,Wenbo Li,Haoyang Huang,Yifan Wu,Ying Li,Gang Huang,Zhonghai Wu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to ACM Multimedia 2026 (ACM MM 2026). This is not the camera-ready version. 18 pages, 7 figures, 12 tables
Abstract:Although Multimodal Large Language Models have achieved strong performance across a wide range of vision-language tasks, they still suffer from hallucinations, where model outputs become inconsistent with the visual content, textual context, or commonsense knowledge. Existing studies primarily address this problem through coarse-grained detection. However, these approaches often provide insufficient diagnostic information for understanding hallucination types and supporting downstream hallucination mitigation. To bridge this gap, we propose fine-grained hallucination diagnosis for MLLMs, a new unified task that jointly performs hallucination detection, classification, and interpretable explanation generation. We develop an automated data generation pipeline and construct HalluScope-30K, a large-scale diagnostic dataset covering eight sources and five task categories. Based on this dataset, we design a multi-granular joint reward function and train two diagnosis models, HalluScope-4B and HalluScope-8B, which achieve state-of-the-art performance on both the MHALO benchmark and our fine-grained hallucination classification benchmark. Notably, detection and classification are mutually beneficial under joint optimization. Furthermore, diagnosis-driven feedback experiments show that the fine-grained diagnostic explanations produced by our model effectively guide target models to correct their hallucinations, with full diagnosis substantially outperforming all baselines on both Qwen3-VL-8B-Instruct and LLaVA-1.5-7B.
[CV-51] Loss Landscape Topology Reveals Why Simple Baselines are Competitive at 3D Point Cloud Segmentation Under Class Imbalance ICPR
链接: https://arxiv.org/abs/2607.21089
作者: Antonis Savva,Christos Kyrkou,Theocharis Theocharides
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 21 pages, 7 figures, International Conference on Pattern Recognition (ICPR) 2026
Abstract:Semantic segmentation of 3D point clouds faces severe class imbalance, yet the effectiveness of specialized imbalance-aware methods from 2D computer vision remains unclear in 3D contexts. We systematically evaluate 11 imbalance mitigation approaches across datasets with extreme (641:1) and moderate (56:1) imbalance ratios, revealing a surprising finding: standard cross-entropy with uniform weighting achieves competitive performance, typically within 0.8-3.3% mIoU of specialized methods across architectures and datasets. Through multifaceted mechanistic analysis of error patterns, decision boundaries, and the geometry of the optimization landscape, our analyses suggest that imbalance severity shapes the topology, creating narrow solution basins under extreme imbalance and flat plateaus under moderate imbalance. This appears to constrain the effectiveness of loss-level modifications, as all methods must navigate these geometric constraints. Our findings offer practical guidance; standard cross-entropy provides a robust baseline, with specialized methods offering modest improvements (0.8-3.3% mIoU) that vary by architecture and dataset but risk substantial degradation if poorly tuned. This work provides the first mechanistic explanation for why techniques proven effective in 2D do not readily transfer to point-based 3D point cloud segmentation, validated across two representative architectures.
[CV-52] Geo3R: Mitigating Spatial Reasoning Hallucination in Multimodal Large Language Models ACM-MM2026
链接: https://arxiv.org/abs/2607.21085
作者: Mingyu Wang,Weilin Jin,Wenbo Li,Haoyang Huang,Tong Jia,Ying Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by ACM MM 2026. This is the arXiv preprint version, not the camera-ready
Abstract:Despite remarkable progress in visual understanding, Multimodal Large Language Models (MLLMs) remain prone to hallucinations when reasoning about spatial relationships, often producing judgments that contradict the true 3D structure of the scene. Though several existing works have proposed to mitigate hallucinations, our analysis indicates that they show limited effectiveness in spatial reasoning, as they fail to bridge the fundamental gap between 2D visual representations and 3D spatial reality. Based on this finding, we define hallucinations arising from insufficient spatial structure modeling as spatial reasoning hallucination, a subcategory of relation hallucination that existing mitigation methods fail to address. We further identify three typical scenarios where such hallucinations frequently occur: perspective effects, object orientation, and viewpoint changes. To this end, we propose Geo3R, a training-free, plug-and-play framework that incorporates geometric evidence and structured 3D reasoning to mitigate spatial reasoning hallucination. Experiments on three benchmarks, covering 18 tasks across all three scenarios, show that Geo3R substantially reduces spatial reasoning hallucination across diverse MLLMs without additional training, outperforming existing models and methods.
[CV-53] he RealDefocus Benchmark for Defocus Deblurring ICIP2026
链接: https://arxiv.org/abs/2607.21078
作者: Tim Seizinger,Zhuyun Zhou,Radu Timofte
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at ICIP 2026
Abstract:Single-Image Defocus Deblurring (SIDD) aims to recover an all-in-focus image from a single defocused observation, but rigorous and reproducible evaluation remains challenging due to the scarcity of realistic, high-resolution datasets with well-aligned defocused/sharp pairs and standardized protocols. We build on RealDefocus, a benchmark derived from the real-world RealBokeh dataset originally proposed for Bokeh Rendering. RealDefocus provides paired defocused inputs and sharp ground truth images, predefined training/validation/test splits, and a unified evaluation framework for comparing image restoration and neural rendering approaches. We further outline a benchmarking protocol with cross-dataset validation to assess reconstruction quality and generalization. The project page is publicly available at: this http URL.
[CV-54] C-PTQ: Fisher-weighted Channel-wise Sensitivity for Post-training Quantization of MLLM s
链接: https://arxiv.org/abs/2607.21076
作者: Jiameng Li,Han Zhou,Matthew B. Blaschko
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 7 pages
Abstract:Multimodal large language models (MLLMs) require huge memory and computational costs, which limits their practical deployment. Post-training quantization (PTQ) techniques offer an efficient solution for model compression and inference acceleration. Yet, the quantized model faces performance degradation due to outlier channels, which are highly sensitive to quantization and substantially impair activation fidelity and task accuracy. To protect these salient channels during quantization, existing PTQ methods leverage modality- or token-level metrics to guide channel-wise scaling (CWS) of LLM decoders. However, these orthogonal measurements fail to capture channel-wise impacts on task-specific loss, and the misalignment between importance and scaling factors ultimately leads to suboptimal performance. To address this issue, we propose C-PTQ, a unified channel-wise PTQ method that harmonizes task-specific loss perturbation and quantization error. Motivated by second-order derivatives, we design a Fisher-weighted objective as a tractable Hessian approximation, seamlessly injecting task sensitivity into the scaling process. Notably, we achieve state-of-the-art performance without auxiliary modules like LoRA, thereby maintaining high efficiency. Experiments on Qwen2.5VL, InternVL2 and LLaVA-OV across 8 benchmarks demonstrate our effectiveness in both weight-only and weight-activation settings.
[CV-55] Show Dont Tell: Evaluating Spatial Cognition in Generative Pixels Rather Than LLM Text
链接: https://arxiv.org/abs/2607.21072
作者: Xu Wang,Kaixiang Yao,Miao Pan,Xiaohe Zhou,Xuanyu Liu,Wenqi Zhang,Xuhong Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 36 pages, 14 figures. Project page: this https URL
Abstract:Spatial intelligence is essential for agents to move from static semantic understanding toward interacting with the physical world. Many spatial tasks are grounded in continuous visual scenes, where locations, regions, and paths are more naturally expressed by pointing, marking, or drawing than by reporting precise coordinates or discrete textual symbols. Yet existing spatial reasoning benchmarks usually require coordinates, options, or text, creating an answer-interface mismatch for image-generation models. This makes it difficult to evaluate image-generation models under the same task semantics as text-output VLMs, despite their ability to externalize spatial judgments directly in pixel space. We propose ProVisE (Protocolized Visual Evaluation), a benchmark-agnostic framework that elicits protocol-constrained visual answers from image-generation models and parses them into structured predictions compatible with original metrics. ProVisE also includes an Agentic builder that constructs and validates task-specific protocols for new benchmarks. We further introduce SpatialGen-Bench, a curated diagnostic benchmark of 470 samples across 14 spatial subtasks, four capability levels, and diverse answer forms. We evaluate representative text-output VLMs and image-generation models in a unified setting and validate Agentic protocol construction on six external spatial benchmarks. Results show that image-generation models are competitive when spatial answers can be externalized directly in pixel space, while text-output VLMs retain a clear advantage in compositional spatial reasoning. These findings reveal complementary strengths of pixel-space expression and text-based reasoning and establish a metric-compatible testbed for studying spatial cognition in image-generation models.
[CV-56] ransBiolab: A Real-World Multi-View Dataset of Cluttered Transparent Biomedical Objects
链接: https://arxiv.org/abs/2607.21071
作者: Ke Ma,Yifei Wang,Meng Wang,Tian Xia
类目: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM); Robotics (cs.RO)
备注: 9 pages, 10 figures, accepted by ACM Multimedia 2026
Abstract:Autonomous biomedical laboratories increasingly rely on visual perception to recognize, localize, and manipulate transparent plasticware, yet high-quality real-world datasets for this setting remain limited. The scarcity of domain-relevant data is particularly restrictive in cluttered multi-object scenes, where mutual occlusion and view-dependent appearance changes remain challenging even for contemporary visual foundation models. Existing transparent-object datasets have advanced segmentation, depth, and pose estimation, but they usually do not evaluate the combined setting of multi-object clutter, occlusion, and calibrated multi-view capture that characterizes real laboratory manipulation scenes. To address this gap, we present TrainsBiolab, a real-world RGB-D dataset of cluttered transparent biomedical objects captured as calibrated multi-view sequences. TrainsBiolab contains 161,315 frames from 98 scenes and 1.03M instance annotations over 15 laboratory object types, including 6D poses, full and visible masks, depth, and per-frame camera calibration. The dataset is organized along three axes that reflect operational difficulty: object category, the total number of objects in a frame, and camera viewpoint. We further define dataset-centric benchmarks for segmentation, depth estimation and completion, and 6D pose estimation, and report a system-level robot manipulation evaluation enabled by the released annotations and calibrations. By focusing on repeated transparent instances, clutter, and multi-view laboratory capture, TrainsBiolab provides a resource for segmentation, depth estimation, 6D pose estimation, and multi-view reasoning in autonomous laboratory manipulation. Project page: this https URL.
[CV-57] Counterfactual Explainability Framework With CycleGAN And Counterfactual-Classifier Alignnment Score for Retinal Disease Classification
链接: https://arxiv.org/abs/2607.21068
作者: Kritanu Chattopadhyay,Sayanjit Singha Roy,Soumya Chatterjee
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 8 pages, 9 figures, 9 tables
Abstract:Automated detection of vision impairing retina-based ocular conditions from fundus images is important for early screening, timely referral and reducing dependency on specialist-only assessment, for which neural network-based deep learning (DL) models have been widely utilized. However, explainability of the DL frameworks remains a major bottleneck for clinical adoption, particularly when model decisions are not linked to retinal regions that are clinically meaningful. To address this issue, this study presents CounterFundus, a novel CycleGAN-driven counterfactual explainability framework, integrating EfficientNet-B5-based retinal disease detection with visually interpretable disease-to-normal fundus image translation. For each pathological image, the counterfactual yielded by the CycleGAN generator represents an estimated healthy counterpart and the resultant difference map is utilized to localize disease-associated retinal changes. Unlike conventional post-hoc saliency methods, CounterFundus provides counterfactual explanations through visually plausible disease-to-normal retinal translation. Thereafter, to quantify the spatial agreement between counterfactual difference maps and classifier saliency, the Counterfactual-Classifier Alignment Score (CCAS) is introduced, embedding Spearman correlation, binary IoU and pointing accuracy into a single assessment protocol. To this end, EigenCAM-aligned evaluation demonstrates that the generated counterfactual explanations remain spatially consistent with classifier-relevant retinal evidence across all CCAS dimensions. Along with that, ablation studies further confirm that CCAS-filtered counterfactual augmentation improves the downstream classification performance in fundus images, establishing CounterFundus as a clinically-grounded, explainable artificially intelligence (XAI) framework for retinal disease detection.
[CV-58] Do Pathology Vision-Language Models Truly See Pathology?
链接: https://arxiv.org/abs/2607.21065
作者: Chengyang Zhang,Wenchuan Zhang,Bo Li,Xinyu Liu,Jiaming Yang,Mengran Li,Chenxun Deng,Jie Chen,Yang Zhang,Wei Ju,Yuhao Yi,Hong Bu,Jiancheng Lv
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Pathology vision-language models (VLMs) have recently progressed rapidly and are commonly evaluated by answer accuracy on pathology VQA benchmarks. However, we dig into current evaluations and identify three overlooked issues: 1) Visual evidence is not always necessary. For instance, Gemini-3-Pro achieves 53.5% average accuracy across 5 VQA benchmarks without any visual input. 2) Domain training can improve accuracy without proportional gains in visual binding. Compared with Qwen2.5-VL-7B, Patho-R1-7B exhibits a 5.8-point lower multimodal gain and a 3.7-point lower attention IoU. 3) Entity-level attention is diffuse and weakly query-specific. On PathVG, attention maps remain highly correlated across different entity queries. These issues can lead to substantial misjudgments of pathology VLMs’ actual multimodal capabilities. To this end, we present PathBind, a benchmark comprising 2,600 samples: PathBind-VQA with 1,500 questions across six dimensions, PathBind-PTA with 600 questions from a private pathology teaching atlas, and PathBind-Grounding with 500 expert-curated region-level samples. Each component undergoes task-specific automated filtering and expert review to reduce textual shortcuts and improve entity-region correspondence. We evaluate 18 representative VLMs on VQA samples of PathBind and five existing pathology VQA benchmarks, and further evaluate 10 VLMs on PathBind-Grounding and PathVG. Results show that current pathology VLMs still exhibit a substantial gap between answer-side performance and visual-semantic binding.
[CV-59] MVEI EmObserver: Empowering MLLM -Oriented Visual Emotional Intelligence via Emotion Statement Judgement
链接: https://arxiv.org/abs/2607.21061
作者: Daiqing Wu,Dongbao Yang,Jiashu Yao,Hongrui Zhang,Can Ma,Yu Zhou,Sicheng Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Affective Image Content Analysis (AICA) aims to recognize and understand emotions elicited by visual content, representing an indispensable step toward Artificial General Intelligence (AGI). However, despite the rapid progress of Multimodal Large Language Models (MLLMs), systematic evaluation of their visual emotional intelligence remains largely absent from recent model releases. We attribute this gap to a structural mismatch between conventional AICA paradigms and the open-ended, instruction-driven nature of MLLMs, where further analysis reveals four major limitations: omission of plausible responses, limited emotion taxonomies, neglect of contextual factors, and labor-intensive annotation. To overcome these barriers, we introduce Emotion Statement Judgement (ESJ), a statement-verification formulation that preserves the expressiveness of the input space while constraining outputs to discriminative judgements. We further develop INSETS, a labor-efficient pipeline that instantiates ESJ at scale by constructing INSETS-462k and supporting MVEI, a rigorously refined benchmark spanning sentiment polarity, emotion interpretation, scene context, and perception subjectivity. Beyond evaluation, we build EmObserver, an emotion-oriented MLLM optimized on ESJ through an elaborate multi-stage recipe. Extensive evaluation of broad-spectrum MLLMs on MVEI reveals fine-grained insights into current artificial visual emotional intelligence, while experiments on multiple AICA benchmarks demonstrate the accuracy, generalization, and reasoning faithfulness of EmObserver. Collectively, these results establish ESJ as a practical formulation, MVEI as a comprehensive benchmark, and EmObserver as an advanced baseline for advancing MLLM-oriented visual emotional intelligence. Code will be released at: this https URL.
[CV-60] Achieving Text-based Person Retrieval with Any Granularity
链接: https://arxiv.org/abs/2607.21057
作者: Jialong Zuo,Hanyu Zhou,Dongyue Wu,Yongtai Deng,Mengdan Tan,Nong Sang,Changxin Gao,Xiang Bai
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: TPAMI-2026 Accepted Paper
Abstract:Text-based person retrieval faces a critical but under-explored challenge: the inherent uncertainty of query granularity in real-world scenarios. This paper introduces a new paradigm, Text-based Person Retrieval with Any Granularity, and provides a systematic solution. First, we formalize a five-level granularity spectrum and construct UFine6926-MG, a high-quality multi-grained dataset annotated comprehensively at all granularities via a novel Multi-grained Text Annotation Engine. Second, acknowledging that coarse queries naturally correspond to multiple valid candidates, we propose MG-Eval, a holistic evaluation benchmark with progressively detailed texts and cross-identity labels that reflect real-world semantics, alongside tailored evaluation metrics and protocols. Third, after a comprehensive diagnosis reveals the systemic limitations of existing research, we propose the Cross-modal Multi-grained Aligning and Matching (CMAM) framework. CMAM achieves granularity-aware retrieval through: 1) orthogonal-expert perception to disentangle granularity-specific features; 2) probabilistic alignment to model many-to-many matches under query uncertainty; and 3) granularity-consistent reasoning to steer feature learning via joint cross-modal granularity verification. Experiments demonstrate that CMAM significantly outperforms state-of-the-art methods across all granularity levels. This work establishes a foundational benchmark and a robust baseline, paving the way for more practical person retrieval systems.
[CV-61] HyperImageNet: A Large-Scale High-Spatial Resolution Hyperspectral Imagery Classification Benchmark
链接: https://arxiv.org/abs/2607.21050
作者: Chuguang Zeng,Jingtao Li,Yinhe Liu,Yanfei Zhong
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We present HyperImageNet, a large-scale benchmark for fine-grained hyperspectral land-cover understanding. The dataset contains 26,084 airborne hyperspectral image patches with 224 spectral bands and 138 fine-grained land-cover categories. Unlike existing datasets, HyperImageNet provides raw imagery, pixel-level semantic labels, and object-level instance masks, supporting both semantic and instance segmentation. Furthermore, we establish an open-environment benchmark with strict spatial separation to evaluate representative methods and the HyperFree foundation model. Experimental results demonstrate the effectiveness of HyperImageNet for fine-grained hyperspectral understanding and open-environment remote sensing research.
[CV-62] GeoThreat: Transferable Targeted Adversarial Attacks on Large Vision-Language Models for Remote Sensing Image Interpretation
链接: https://arxiv.org/abs/2607.21036
作者: Yimin Fu,Yuefeng Bai,Baicheng Pan,Zhunga Liu,Michael K. Ng
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: The code will be released at this https URL upon acceptance
Abstract:Adversarial attacks against large vision-language models (LVLMs) serve as an effective means of assessing their robustness in cross-modal semantic understanding. Existing studies mainly focus on corrupting visual inputs to induce predefined erroneous responses in general vision-language tasks, whereas corresponding investigations in remote sensing fields remain largely underexplored. Compared with natural image understanding, remote sensing image interpretation requires joint reasoning over local discriminative cues and global scene context. This poses additional challenges to achieving transferable semantic manipulation toward specified responses under black-box settings. To tackle these challenges, we propose GeoThreat, a transferable targeted adversarial attack method against LVLMs for remote sensing image interpretation. Specifically, GeoThreat modulates adversarial representations in accordance with the target content at both conceptual and perceptual levels. The class tokens from surrogate image encoders are employed as conceptual representations, while perceptual representations are distilled from patch tokens of the adversarial example through collaborative importance estimation. Beyond merely rolling out attention scores across layers, we incorporate adversarial-target similarity gradients to more faithfully characterize the relevance of local visual cues to the intended semantic manipulation. The perceptual representations are then dynamically aligned with target patch tokens in a cross-attentive manner, facilitating the adaptation of local cues toward designated semantic details. Finally, adversarial perturbations are iteratively updated via ensemble-based joint optimization of conceptual calibration and perceptual adaptation. Extensive experiments across diverse LVLMs demonstrate the superiority of GeoThreat in both transferability and controllability.
[CV-63] Spectral-Spatial Synergistic Guided Network for Hyperspectral Salient Object Detection
链接: https://arxiv.org/abs/2607.21032
作者: Yanyan Peng,Tingfa Xu,Yao Xiao,Peifu Liu,Shuyan Bai,Fengxiang Xu,Jianan Li
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted by IEEE TMM
Abstract:Hyperspectral salient object detection aims to identify visually salient regions from hyperspectral images. Existing methods often fail because they fundamentally misunderstand the data, confusing incidental spectral variations caused by external factors such as illumination with essential spectral differences caused by the intrinsic material properties of the object. This leads to fragile representations and noisy predictions. To this end, we propose a lightweight and efficient Spectral-Spatial Synergistic Guided Network (S3GNet), with structure perception as the core, to build a closed-loop information flow around spectrum robust modeling, cross-stream co-perception and multi-scale refinement decoding. S3GNet introduces a parameter-free Spectral Structure-Aware Module that leverages spectral derivatives and regional hierarchical modeling to extract intrinsic features of robustness against illumination variations. Our Stream-Aware Attention Module achieves effective spectral-spatial collaboration through inter-stream global interaction and intra-stream spatial guidance. Furthermore, a Progressive Gated Refinement Decoder ensures precise object boundaries and detail recovery by optimally integrating multi-scale features. Experimental results show that S3GNet achieves superior performance in both computational efficiency and detection accuracy compared to existing methods.
[CV-64] GroupVideo: Multi-Identity Customized Text-to-Video Generation
链接: https://arxiv.org/abs/2607.21027
作者: Xinyang Song,Libin Wang,Jianxin Sun,Qi Li,Dandan Zheng,JingDong Chen,Zhenan Sun
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Current identity customized video generation methodologies are predominantly limited to single-identity scenarios, as the lack of explicit identity separation mechanisms often leads to identity confusion in multi-identity settings. Existing multi-identity approaches, which directly extend single-identity frameworks by concatenating face images as input conditions, frequently result in unnatural facial expressions and motions, manifesting as the “copy-paste” phenomenon. To overcome these limitations, we introduce GroupVideo, a novel framework that leverages multiple individual photographs to generate identitycustomized video. Built upon Video Diffusion Transformers, GroupVideo incorporates multimodal identity alignment: visual alignment jointly encodes multiple face images to provide robust identity references, while semantic alignment introduces a semantic perceiver to enhance the naturalness of motions. An ID localization module with spatial guidance is introduced to address identity blending and enhance identity fidelity, along with bounding box constraints and mask regularization loss, to focus on facial regions and improve training efficiency. In response to the shortage of multi-ID video datasets, we have curated a comprehensive high-quality dataset of 20,000 videos, thereby establishing a crucial resource to advance future research in multi-ID video generation. Extensive experiments demonstrate that GroupVideo outperforms existing methods in generating multi-character videos with consistent identities and natural motions.
[CV-65] WAT3R: Feedforward Underwater 3D Reconstruction
链接: https://arxiv.org/abs/2607.21023
作者: Jiayi Xu,Jiahao Lu,Ziqiang Zheng,Yihao Tan,Yaolong Zhu,Yuan Liu,Sai-Kit Yeung
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Reliable feedforward underwater 3D reconstruction remains challenging due to severe light attenuation and backscattering, which degrade visual quality and disrupt feature consistency across views, leading to inaccurate multi-view geometry. To address this issue, we propose WAT3R, a feed-forward framework for reconstructing 3D scenes directly from underwater images. By leveraging degradation adaptation as a geometry-constrained process, WAT3R integrates a lightweight neural adaptation module to flexibly account for these underwater imaging effects, thereby improving multi-view reconstruction quality. Implemented in a single forward pass, WAT3R directly and efficiently outputs pixel-aligned 3D point maps and camera poses from underwater videos, allowing a high-quality underwater 3D reconstruction. Experiments conducted on the FLSea, SQUID, and USOD10K datasets show that our method consistently outperforms state-of-the-art approaches on 3D reconstruction tasks, including multi-view/monocular depth estimation and camera pose estimation.
[CV-66] ProCap: Prominence-guided Object Rectification for Faithful and Comprehensive Video Captioning
链接: https://arxiv.org/abs/2607.21022
作者: Debjyoti Das Adhikary,Aritra Hazra,Partha Pratim Chakrabarti
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 10 pages, 7 figures, 5 tables. Submitted to IEEE Transactions on Multimedia
Abstract:Improving video captioning quality typically demands retraining large vision-language models, an expensive and often impractical requirement. Existing training-free alternatives instead ground captions in detected objects to curb hallucination, but apply only a single, fixed correction pass without prioritizing which objects matter most, leaving semantically significant content omitted. We propose a prominence-aware, iterative post-hoc rectification framework that overcomes both limitations without modifying the underlying captioning model’s parameters: a lightweight scoring mechanism ranks detected objects by spatial saliency, temporal persistence, and relational dynamics, and an iterative, prompt-driven refinement loop uses this ranking to progressively inject missing yet contextually relevant objects into the caption over multiple rounds. We validate the framework on MSVD and MSR-VTT using object-grounded automatic metrics, a 110-participant human study, and qualitative comparison against ChatGPT and Gemini; in human evaluation, the framework raises perceived completeness by up to 48% and reduces hallucination by up to 45% relative to a strong pretrained captioning baseline, all without retraining or reference captions. These results position prominence-guided iterative rectification as a lightweight, scalable, and model-agnostic route to more complete and trustworthy video captioning, with direct relevance to accessibility, retrieval, and other multimedia understanding applications.
[CV-67] EmoAgent -R1: Towards Multimodal Emotion Understanding with Reinforcement Learning-based Dynamic Agent Specialization
链接: https://arxiv.org/abs/2607.21013
作者: Lihuang Fang,Yuchen Zou,kebin Jin,Jinghui Qin
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Multimodal large language models (MLLMs) have achieved impressive performance in multimodal emotion recognition (MER) tasks and lifted MER to a new level that is complex emotion understanding with advanced video understanding abilities and natural language description. However, existing MLLM-based methods often use a fixed prompt to perceive the emotions, ignoring the dynamicity and complexity of the emotion source in the multimodal inputs. To address these issues, we propose a novel Reinforcement Learning-based Dynamic Agent Specialization framework (\textbfEmoAgent-R1) to optimize the emotion recognition, reasoning, and generalization abilities of an MLLM with dynamic agent specialization based on reinforcement learning. Specifically, we first adopt a cold start strategy to endow an MLLM with preliminary emotion recognition, reasoning, and agent routing ability by training with synthetic answer-conditioned chain-of-thought data and agent routing data. Then, we further train the MLLM with reinforcement learning to perceive emotions in a two-step agentic workflow with agent selection and agent specialization. To effectively train EmoAgent-R1, we propose a novel Progressive Group-Relative Policy Optimization (P-GRPO) to combine group-based relative advantages with a PMI-inspired progressive token-level modulation to transform sparse rewards into fine-grained learning signals, mitigating the coarse-grained uniform credit assignment issue in GRPO. Extensive experiments on MER benchmarks demonstrate the superiority of our EmoAgent-R1 in stronger emotion reasoning performance and improved optimization stability.
[CV-68] Explainable Deepfake Detection Challenge
链接: https://arxiv.org/abs/2607.21007
作者: Abhijeet Narang,Kartik Kuckreja,Shreya Ghosh,Muhammad Haris Khan,Usman Tariq,Jianfei Cai,Abhinav Dhall
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 5 pages, 1 figure
Abstract:Deepfake detection is moving beyond binary classification decisions toward systems that can also explain the visual evidence supporting those decisions. This transition is important for real-world verification settings, where diverse users need to understand not only whether an image is manipulated, but also why it is considered suspicious. The Explainable Deepfake Detection Challenge at ACM Multimedia 2026 is designed to benchmark this joint capability. Built on XPlainVerse, a million-scale benchmark for explainable deepfake detection, the challenge evaluates methods on image classification and grounded natural-language explanation generation. Participants submit a real/fake label together with two explanations for each image: a detailed complex explanation for technical users and a concise simple explanation for general users. The evaluation combines classification metrics with semantic similarity, simplicity, and intent-aware grounding metrics that assess whether explanations identify the relevant manipulated entities and supporting visual evidence. The methodologies developed through the challenge will contribute to the development of next-generation explainable deepfake detectors. Evaluation script, baseline models, and accompanying code are available on this https URL.
[CV-69] AUCH-Net: Action Unit-Based Consistency-Aware Hypergraph Network for Cross-Domain Few-Shot Facial Expression Recognition
链接: https://arxiv.org/abs/2607.21004
作者: Xinhan Qiu,Yan Yan,Rui Zhu,Si Chen,Hanzi Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recently, cross-domain few-shot facial expression recognition (CF-FER) has received considerable attention. However, the performance of existing CF-FER methods is still unsatisfactory due to inferior transferable feature learning under large domain discrepancy and limited target samples. Fortunately, the action units (AUs), which indicate the movements of different facial muscles, provide consistent conceptual semantics for describing expressions within and across domains. Inspired by this, we propose a novel Action Unit-based Consistency-aware Hypergraph Network (AUCH-Net), which constructs consistency-aware hypergraphs on AUs, for CF-FER. Specifically, AUCH-Net presents a new AU feature learning (AFL) module and a new visual feature learning (VFL) module. The AFL module learns AU features under the guidance of a novel relation consistency loss and an AU regularization loss, while the VFL module learns visual features supervised by a relation consistency loss and a classification loss. By learning consistent AU features, AUCH-Net effectively models the connections between AUs and expression categories. As a result, we can bridge the gap between fine-grained facial variations and high-level expression categories, greatly facilitating the learning of transferable feature this http URL experiments on both in-the-lab and in-the-wild datasets show that our method consistently outperforms several state-of-the-art methods. Our results clearly show that modeling the relationships among AUs holds significant potential for FER under cross-domain few-shot scenarios.
[CV-70] Sparse Concept Channels in Frozen 3D CT Vision Encoders
链接: https://arxiv.org/abs/2607.20993
作者: Farhad Nooralahzadeh,Lea Bogensperger,Christian Bluethgen,Michael Krauthammer
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:Large vision-language models are becoming increasingly dominant in 3D medical image interpretation, but we rarely know iwhich/i internal units encode clinical findings or iwhere/i that information lives in the representation. We first study this on a 3D chest vision-language model (Pillar-0) by probing its frozen vision embeddings. We show that (i) each radiological finding is encoded by a isparse/i set of ~10 vision-encoder channels that match full-feature classification performance and far exceed a zero-shot text prompting; (ii) turning off the channels tied to one finding, that finding’s score collapses while unrelated labels stay stable; and (iii) the same sparse probe ireplicates/i on an architecturally unrelated 3D abdominal VLM (Merlin) suggesting a general property of frozen medical encoders. Our training-free concept channel probe (CCP) method, paired with a corpus-derived report template, outperforms published CT-CHAT on clinical efficacy and NLG metrics (F1 0.549 vs. 0.184; BLEU 0.483 vs. 0.373) at 22x lower latency. Our results provide a clear, reproducible characterization of how frozen medical encoders represent findings, demonstrating direct applicability across models.
[CV-71] Latent Variable-Mediated Cross-Learning for Few-Shot Acoustic Impedance Imaging
链接: https://arxiv.org/abs/2607.20989
作者: Junheng Peng,Yong Li,Mingwei Wang,Yi Bao
类目: Computer Vision and Pattern Recognition (cs.CV); Geophysics (physics.geo-ph)
备注: The manuscript is currently under review
Abstract:Acoustic impedance imaging is a fundamental yet severely ill-posed problem in subsurface analysis: the seismic wavelet is unknown, observations are band-limited, and labeled well-log samples are extremely scarce (typically 1% of all traces). Existing semi-supervised deep learning methods mitigate few-shot problem by incorporating forward modeling, yet they either rely on inaccurate prior wavelet assumptions or introduce auxiliary networks, leading to unstable optimization and degraded performance. We propose RD-SCL, a novel framework that integrates regularized deconvolution with semi-supervised cross-learning. At its core lies a differentiable, closed-form first-order Tikhonov deconvolution operator that dynamically estimates the latent wavelet in the frequency domain during training, providing stable physics-guided feedback without explicit auxiliary networks and fixed wavelet priors. Building on this operator, we design a symmetric cross-learning that enforces consistency between predictions on labeled and unlabeled data, thereby effectively exploiting abundant unlabeled traces. Extensive experiments on the SEAM and Marmousi 2 benchmarks demonstrate that RD-SCL consistently outperforms state-of-the-art supervised and semi-supervised methods, achieving substantial gains with lower computational cost. With only 56.5k learnable parameters and competitive runtime, RD-SCL offers a practical, physically consistent, and efficient solution for acoustic impedance imaging.
[CV-72] HyWorldVLA: A Vision-Language-Action Model with Hybrid World Modeling for Autonomous Driving
链接: https://arxiv.org/abs/2607.20988
作者: Quanfu Yu,Xian Wu,Hao Xu,Liulong Ma
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 20 pages with 13 figures
Abstract:Vision-Language-Action (VLA) models augmented with world modeling represent a promising paradigm for end-to-end autonomous driving. While pixel-level future prediction enables fine-grained spatiotemporal reasoning, it compromises robustness in noisy driving scenarios. Conversely, latent-based world models alleviate this sensitivity but often incur limited interpretability and representational degradation due to absent pixel-level grounding. To reconcile this trade-off, we propose HyWorldVLA, a hybrid world-VLA framework that unifies pixel-level supervision and latent representation learning. In the pre-training stage, HyWorldVLA predicts video latents encoded by a pre-trained video VAE, while simultaneously reconstructing video frames to provide precise pixel-level grounding. During the subsequent co-fine-tuning phase, the model exclusively predicts latent features, which are fed into an action expert to generate trajectories. Extensive experiments on NAVSIM v1 and v2 benchmarks demonstrate that HyWorldVLA significantly outperforms both pixel-based and latent-based world model baselines. Notably, we present the first comprehensive qualitative and quantitative analysis of world model noise robustness in autonomous driving, establishing a new benchmark for evaluating future architectures.
[CV-73] Distribution-Alignment Bridge for Uncertainty-Aware Text-to-Video Retrieval ECCV2026
链接: https://arxiv.org/abs/2607.20984
作者: Kyeongmo Chae,Jihoon Lee,Sangtae Ahn
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: ECCV 2026
Abstract:This paper proposes the Distribution-Alignment Bridge (DAB), a framework that reconceptualizes text-to-video retrieval as a distribution alignment task rather than traditional deterministic point matching. By modeling both text and video embeddings as Gaussian distributions defined by mean and variance, DAB explicitly accounts for modality-specific uncertainty. We employ a deterministic, diffusion-inspired bridge to iteratively refine text distributions toward their target video distributions through a truncated refinement process. This approach unifies probabilistic embedding and distributional transformation into a cohesive, end-to-end trainable system. To optimize cross-modal similarity, we introduce a distribution-aware contrastive loss based on Kullback-Leibler divergence. Extensive evaluations on MSR-VTT, MSVD, and VATEX benchmarks confirm that DAB significantly outperforms existing probabilistic and diffusion-based baselines, while providing calibrated uncertainty-aware ranking through bridge-induced distributional margins.
[CV-74] Unsupervised Metal Artifact Reduction in Dental CBCT using Fine-tuned Cycle-Consistent Adversarial Networks
链接: https://arxiv.org/abs/2607.20977
作者: G.L.T. Chamika,S.N.A. Dhanapala,P.H.S.V. Nimalaweera,Maheshi B. Dissanayake,Ruwan D. Jayasinghe
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: accepted and published work
Abstract:Metal artifacts generated by dental implants significantly degrade cone-beam computed tomography (CBCT) volumes, obscuring critical anatomical structures and compromising diagnostic precision. To address this, an unsupervised deep learning framework has been proposed for Metal Artifact Reduction (MAR) utilizing a Cycle-Consistent Adversarial Network (CycleGAN) optimized for high-fidelity restoration. Unlike supervised methods that rely on unattainable voxel-aligned paired datasets, the proposed approach leverages an unpaired dataset of approximately 4,000 images, curated from the public ToothFairy dataset. The architecture integrates U-Net-based generators and PatchGAN discriminators, specifically tuned to mitigate generative hallucinations and preserve morphological integrity. Quantitative benchmarking on a held-out test set demonstrates a 34.6% improvement in the Blind/Referenceless Image Spatial Quality Evaluator (BRISQUE) score, a substantial reduction in Fréchet Inception Distance (FID) from 207.03 to 157.04, and a superior Structural Similarity Index Measure (SSIM) of 0.9105. The framework achieves real-time efficiency with a 3.03 ms inference time per slice, effectively suppressing artifacts while preserving anatomical detail. Expert validation confirms high fidelity; however, to ensure reliability in extreme cases, the architecture is recommended as a clinical decision-support tool under human-in-the-loop oversight. By enhancing diagnostic clarity via a scalable software pipeline, this study provides a robust solution for high-fidelity dental implant imaging.
[CV-75] FSB-Net: Frequency-Spatial Boundary Network for Brain Stroke Lesion Segmentation in Non-Contrast CT
链接: https://arxiv.org/abs/2607.20955
作者: Linke Fan,Xianglong Li,Huixin Huang,Kai Shu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Accurate segmentation of brain stroke lesions in non-contrast computed tomography (NCCT) scans is critical for rapid clinical decision-making, yet remains difficult due to the low contrast between lesion and normal brain tissue, heterogeneous lesion morphology across ischemic and hemorrhagic subtypes, and ambiguous boundaries caused by partial volume effects. Current deep learning approaches primarily optimize region-level overlap but lack explicit boundary modeling, leading to imprecise delineation that can affect volumetric assessment and treatment planning. We propose FSB-Net, a frequency-spatial boundary network that leverages frequency-domain analysis for boundary-aware stroke lesion segmentation. FSB-Net introduces three components: (i) a Wavelet Boundary Detection Head (WBDH) that applies the discrete wavelet transform to multi-scale encoder features, extracting high-frequency sub-bands as boundary representations; (ii) a Frequency-Spatial Cross-Attention Module (FSCAM) that performs bidirectional attention between wavelet boundary features and spatial decoder features for selective boundary enhancement; and (iii) a Spectral Boundary Loss that penalizes high-frequency discrepancies in the Fourier domain to optimize boundary sharpness. Built on a PVTv2-B2 encoder, FSB-Net is evaluated on a public Brain Stroke CT dataset containing both ischemic and hemorrhagic cases. Experimental results show that FSB-Net outperforms U-Net, UNet++, MANet, and DeepLabV3+ across all metrics, achieving state-of-the-art performance in mean Dice, mean IoU, and HD95.
[CV-76] RECO: Region-Aware Compensation for Extrinsic Perturbations in Roadside 3D Detection
链接: https://arxiv.org/abs/2607.20947
作者: Junsheng Du,Zhaocheng He,Yuhuan Lu
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:In intelligent transportation systems, roadside 3D object detection provides wide-area perception crucial for traffic understanding, cooperative early warning, and safe autonomous driving. However, existing methods suffer from high sensitivity to camera extrinsics; even slight deviations (whether manifesting as transient jitter or persistent drift) can be significantly amplified by projective geometry. This cascade results in severe feature misalignment and degraded localization. To mitigate this limitation, we propose RECO, a region-aware extrinsic compensation framework that corrects extrinsics using piecewise 6-DoF pose offsets. RECO predicts a learnable range boundary to partition the scene into near and far regions, estimating region-specific pose corrections. A differentiable sigmoid gate then smoothly blends the two compensated geometries to preserve continuous BEV sampling and facilitate stable optimization. To supervise the refinement of extrinsics, we introduce an auxiliary reprojection loss that compares 2D bounding boxes projected from 3D ground truth against 2D annotations, optimizing it jointly with the standard detection objective. Extensive experiments on the DAIR-V2X-I and Rope3D benchmarks under extrinsic perturbations demonstrate consistent improvements over state-of-the-art baselines across both yaw and z -axis deviations. RECO also generalizes from transient perturbations to persistent shifts, maintaining highly competitive performance under strict calibration uncertainty.
[CV-77] Ms. Forcing: Efficient Streaming Video Generation with Multi-Scale Patchification and Attention
链接: https://arxiv.org/abs/2607.20940
作者: Zekun Li,Xiaoyan Cong,Hongyu Li,Zhiyang Dou,Chuan Guo,Abhay Mittal,Sizhe An,Srinath Sridhar
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Streaming video diffusion models have made substantial progress toward interactive and dynamic world simulation, but the nested autoregressive and denoising loops of conventional next-frame generation hinder real-time deployment. Recent rolling-window methods pipeline denoising across multiple consecutive frames at different noise levels, improving throughput and long-horizon stability. However, they tokenize every state at the same fine spatial granularity, leaving substantial noise-dependent redundancy in the joint denoising window. We propose this http URL, an efficient streaming video generation paradigm that adapts spatial granularity to each state’s noise level. Its Multi-Scale Patchification (MSP) assigns coarser patches to noisier states, reducing the active-window token count by 45%, while Multi-Scale Self-Attention (MSSA) matches the density of visible non-sink keys and values to each query scale to further reduce attention cost. Because both schedules are fixed by window position, this http URL retains a static, hardware-friendly computation graph. We further introduce Homogeneous-Noise-Level DMD (H-DMD), which assembles each fake video from clean predictions sharing the same source noise level, thereby reducing the mismatch between DMD training sequences and inference-time rollouts. The multi-scale design helps offset the additional training cost of backpropagating through overlapping windows. We include both quantitative and qualitative experiments to show that this http URL reaches 22.84 FPS on a single H200 GPU, 39.6% faster than Rolling Forcing, while significantly improving VBench scores in both short video and long video generation setting.
[CV-78] MagicMakeup: A Region-Controllable Diffusion Transformer for High-Fidelity Makeup-Transfer
链接: https://arxiv.org/abs/2607.20924
作者: Ziyi Wang,Siming Zheng,Yang Yang,Shusong Xu,Hao Zhang,Bo Li,Changqing Zou,Peng-Tao Jiang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Makeup-transfer applies the reference makeup to the source face while preserving the source identity. Despite advances in full-face editing by diffusion-based methods, strong regional controllability, makeup fidelity, and identity preservation remain challenging. The reasons are (i) pixel-to-attention misalignment that causes spillover into non-target areas and weakens regional control; (ii) unclear transfer/preservation concept separation under two-image conditioning, leading to coupling between makeup attributes and identity; and (iii) the lack of a high-resolution dataset that is identity-consistent and region-labeled for fine-grained supervision. In this paper, we propose MagicMakeup, a diffusion transformer-based framework for region-controllable and high-fidelity makeup transfer, built on spatial constraints and concept disentanglement. To enable precise region-specific editing while preserving identity, we propose Token-Aligned Region Gating, which aligns pixel masks with attention and applies region-specific logit gating. To clarify the concepts of transfer and preservation, we further introduce Cross-Modal Perception Guidance, which aligns text and image features to enhance cross-modal concept perception. We also design a pipeline for the generation of 1024 x 1024 data pairs through region-specific makeup removal and establish a unified benchmark in synthetic and real settings. Extensive quantitative and qualitative experiments show that MagicMakeup improves regional controllability, makeup fidelity, and identity preservation, with strong robustness across styles, races, and poses.
[CV-79] FA-LAM: Focus-Aware Large Avatar Model for One-Shot 4D Animatable Gaussian Head
链接: https://arxiv.org/abs/2607.20922
作者: Yingdong Hu,Yisheng He,Yiming Jiang,Zehong Lin,Steven Hoi,Jun Zhang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:We propose FA-LAM, a Focus-Aware Large Avatar Model for one-shot animatable Gaussian head creation, while simultaneously enabling static 3D and dynamic 4D full-head recovery. The core of our method lies in a thorough analysis of the attention mechanisms and the entangled reconstruction and animation training pipeline adopted by prior state-of-the-art approaches. Our analysis identifies two main factors that compromise the quality of 3D full-head generation: (1) incorrect and noisy attention activations, and (2) conflicts between the tasks of reconstruction and animation. To address the first issue, we introduce a symmetric and semantic attention regularization strategy that leverages the inherent semantics and structural symmetry of human heads. To disentangle the objectives of reconstruction and animation, we develop a novel dual-phase training pipeline that separates the model’s capabilities for large-view hallucination and animation into distinct modules. Moreover, we enhance our model to support multi-view and streaming 4D reconstruction in an efficient and memory-friendly manner through a core autoregressive modification with tailored visibility-aware token fusion. Collectively, these innovations enable FA-LAM to reconstruct animatable Gaussian full heads with superior quality, particularly in fine facial regions and large viewing angles.
[CV-80] DINO-VPT: Hierarchical Visual Prompt Tuning for Joint Physical-Digital Face Anti-Spoofing
链接: https://arxiv.org/abs/2607.20900
作者: Pierre Gallin-Martel,Mika Feng,Koichi Ito,Takafumi Aoki
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: accepted to IJCB2026
Abstract:With the increasing diversity of spoofing attacks, there is a growing demand for unified Face Anti-Spoofing (FAS) models capable of detecting both physical and digital threats. While existing Vision-Language Models (VLMs) demonstrate high generalization in this context, they heavily rely on complex multimodal fusion and external text encoders. In this paper, we propose DINO-VPT, a lightweight, vision-only framework leveraging hierarchical visual prompt tuning. By dynamically injecting prompts conditioned on input features via a Prompt Routing Network (PRN), our method effectively disentangles diverse spoofing artifacts without requiring multimodal fusion. Evaluations on the UniAttackData benchmark demonstrate that DINO-VPT achieves higher accuracy than state-of-the-art VLM-based methods. Our results indicate that a properly structured vision-only architecture can achieve state-of-the-art performance in unified FAS without the need for multimodal supervision.
[CV-81] MAGE-Vein: Multi-Instance Age and Gender Estimation from Finger Vein Images
链接: https://arxiv.org/abs/2607.20897
作者: Katsuki Tanaka,Koichi Ito,Takafumi Aoki,Masakazu Fujio,Yosuke Kaga,Kanade Oshima,Kenta Takahashi
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: accepted to IJCB2026
Abstract:Age estimation from finger vein images has been widely considered impractical due to severe demographic biases in public datasets and physiological confounding factors like gender. To overcome these limitations, we propose MAGE-Vein, a novel multi-instance, multi-task learning framework. Our approach extracts robust structural aging signs by employing a hybrid feature-level fusion of three fingers, effectively suppressing local imaging noise. Furthermore, simultaneous optimization of gender classification conditions the network to effectively eliminate gender-specific vascular variations. Evaluated on a demographically balanced dataset of 402 subjects, MAGE-Vein achieves a mean absolute error of 6.12 years and a correlation of 0.880. Our results not only overturn the conventional consensus regarding the limitations of the finger vein modality but also demonstrate that previous estimation failures were primarily artifacts of biased public datasets. Our code is available at this https URL.
[CV-82] Engine-Native Editable 3D World Reconstruction with Objects and Lighting
链接: https://arxiv.org/abs/2607.20889
作者: Junhao Chen,Xinghao Chen,Henghaofan Zhang,Zihao Qiao,Saining Zhang,Yongzhi Li,Ruqi Huang,Sisi Li,Yimin Sheng,Jianyi Zhu,Hao Zhao
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 18 pages, 7 figures, Project Page: this https URL
Abstract:Editable 3D scene creation requires object instances and lights that can be inspected, moved, and imported into standard engines, yet existing single-image methods largely stop at room-scale geometry, baked/global illumination, or text-driven generation. We introduce Lumera (Light-aware Unified Engine-native Reconstruction and Assembly), a benchmark and reference pipeline for engine-native, light-aware 3D scene parsing from a single image. Lumera-2K is built from 2,513 UE5 projects and provides 3.73M components, 63M object instances, 102.6K engine-native parametric lights, and 95.1K camera views. On this data, Lumera-Box and Lumera-Light adapt VLM to parse object boxes and parametric light tuples (x,y,z,r,g,b,I), which are assembled with per-object mesh reconstruction, HDR environment estimation, and a bounded agentic refinement loop. In a sanitized box benchmark against DetAny3D, SpatialLM, N3D-VLM, and WildDet3D, Lumera-Box obtains the strongest overall detection, geometry, semantic, and layout scores (merged mAP 0.1141, IoU-B 0.2472, F-score 0.2762), while WildDet3D remains stronger on anchor recall. For lights, Lumera-Light recovers almost all non-empty scenes (recall 0.998) but remains limited at individual-light localization (F1 0.209 at 0.5 m); matched lights have median position error 0.261 m, median \DeltaE2000 4.59, and intensity Pearson r=0.628. These results establish parametric lights as a measurable editable-scene target and expose remaining bottlenecks in relation structure, light recall/intensity, and cross-engine generalization.
[CV-83] WhereEdit: Mask-aware Local Latent Editing for One-Step Image Editing
链接: https://arxiv.org/abs/2607.20883
作者: Ming Hu,Mingyu Dou,Jianfu Yin,Miaomiao Zhang,Cong Hu,Yao Wang,Bingliang Hu,Quan Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Recent one-step text-to-image (T2I) models enable efficient image synthesis and provide new opportunities for real-time image editing. However, existing one-step editing methods primarily rely on text conditioning for semantic transformation, lacking explicit spatial control over \textitwhere to edit. More importantly, even when spatial constraints are introduced, these methods often struggle to achieve strong and stable semantic modifications within the target regions. In this work, we revisit one-step image editing from a spatially controlled perspective and identify two key challenges: discovering editable regions and achieving effective localized semantic transformation. We reveal that existing methods perform global semantic transport, which limits high-intensity local editing under the one-step setting. To address this issue, we propose \textbfWhereEdit, a framework that reformulates one-step editing as localized adaptive editing. WhereEdit automatically identifies semantically relevant regions from internal model features and applies adaptive local modulation to enhance target-region editing while preserving non-target areas and structural consistency. Experiments on the PIE-Bench benchmark demonstrate that WhereEdit consistently outperforms existing one-step image editing methods, achieving superior editing quality while maintaining the efficiency of one-step generation. Additional experiments with region-level supervision further highlight the importance of explicit spatial reasoning for high-quality one-step image editing.
[CV-84] Webly Supervised Multi-Label Recognition: Evaluation Benchmark and Dual-Branch Multi-Label Contrastive Learning
链接: https://arxiv.org/abs/2607.20874
作者: Zhihua Xu,Zhijing Yang,Yufeng Yang,Tianshui Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Training deep learning models with freely available web images can reduce their dependence on costly manual annotations. Although webly supervised learning has been widely studied for single-label recognition, its multi-label counterpart remains underexplored, partly due to the lack of unified benchmarks and fair comparison protocols. To address this gap, we construct a benchmark for webly supervised multi-label recognition (WS-MLR), including Web-COCO and Web-Pascal, and re-implement representative baselines under a unified setting. The two datasets cover the same 80 and 20 categories as MS-COCO and Pascal VOC, respectively, and contain about 300 thousand images retrieved from the Internet using category-word combinations as search keywords. We further propose a Dual-Branch Multi-Label Contrastive Learning (DBMLCL) framework, which learns category-specific instance-level and category-level representations together with their similarities to identify and correct noisy labels. Extensive experiments on the benchmark demonstrate that DBMLCL achieves superior performance compared to representative baselines.
[CV-85] ViSTR-Bench: Can MLLM s Reason from Continuous Visual Cues in Dynamic Scenes?
链接: https://arxiv.org/abs/2607.20868
作者: Han Li,Si Liu,Zehao Huang,Dongxin Lyu,Longfei Xu,Jiahui Fu,Daxin Tian,Yuliang Xiu,Naiyan Wang
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: 37 pages, 37 figures
Abstract:Multimodal Large Language Models (MLLMs) have achieved remarkable success across diverse expert-level tasks, but they still struggle with fundamental abilities that humans naturally develop through continuous observation of the real world, such as spatial perception and dynamic reasoning. Recent studies have recognized this gap and introduced dedicated benchmarks to evaluate the spatial-temporal capabilities of MLLMs. However, existing benchmarks mostly focus on static scenes or require exact quantitative predictions, leaving intuitive reasoning from temporal cues largely underexplored. In this paper, we introduce the Visual Spatial-Temporal Reasoning Benchmark (ViSTR-Bench), a novel evaluation suite designed to systematically assess whether MLLMs can perform qualitative reasoning from continuous visual cues in dynamic scenes. Guided by the principles of temporal emphasis, reasoning orientation, and qualitative evaluation, ViSTR-Bench establishes a comprehensive four-dimensional evaluations covering Motion Perception, Spatial Relations, Outcome Prediction, and Physical Dynamics. The benchmark comprises 15 distinct subtasks and 1,340 high-quality video question-answer pairs spanning diverse tabletop, indoor, and outdoor scenarios. Extensive evaluations of a broad spectrum of state-of-the-art proprietary, open-source, and specialized spatial MLLMs reveal that, despite their strong general video understanding capabilities, current models still face substantial bottlenecks in complex spatial-temporal reasoning and remain far below human performance.
[CV-86] Agent ic Designer: Progressive Multi-Agent Collaboration for Structure-Aware Interior Layout Generation
链接: https://arxiv.org/abs/2607.20866
作者: Zhijing Yang,Haocheng Lin,Zhihua Xu,Haojie Li,Keze Wang,Liang Lin,Tianshui Chen
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: TPAMI 2026
Abstract:Generating realistic interior furniture layouts that strictly adhere to architectural constraints (e.g., walls, doors, and windows) remains a fundamental challenge in automated spatial design. Existing approaches, primarily based on one-shot generation using diffusion models or Large Language Models (LLMs), lack explicit mechanisms for intermediate geometric constraint verification, often resulting in structural collisions and functionally infeasible arrangements under complex room constraints. To address these challenges, we propose Agentic Designer, a progressive, multi-agent framework that formulates structure-aware interior layout generation as an iterative and constraint-verified decision process. By decomposing layout synthesis into modular stages of proposal, verification, and adjustment, the framework coordinates three specialized agents, a Generator, an Evaluator, and a Refiner, through a Progressive Consensus Mechanism. This mechanism enforces stepwise geometric validation and correction before each placement is committed, thereby preventing error accumulation. To facilitate this structure-aware paradigm and standardize evaluation, we establish InStruct, a comprehensive benchmark that integrates a dataset comprising over 18,000 high-quality, parametrically annotated samples with a novel suite of structure-centric metrics. Extensive quantitative evaluations, qualitative analyses, and user studies show that Agentic Designer significantly outperforms state-of-the-art methods, demonstrating substantial improvements in strict structural adherence and functional design coherence.
[CV-87] Explainable graph attention network for stress recognition (StressGAT) via differential action units
链接: https://arxiv.org/abs/2607.20819
作者: Thomas Kassiotis,Stefanos Gkikas,Nikolaos Smyrnis,Giorgos Giannakakis
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at the 14th International Conference on Affective Computing and Intelligent Interaction (ACII 2026)
Abstract:Stress is a dynamic process characterized by significant individual variability in facial expression. Traditional architectures, such as Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs), often overlook person-specific baselines or lack the representational capacity to model the non-linear temporal progression of distress due to sequential bottlenecks and rigid grid-based constraints. Furthermore, many deep learning models lack the interpretability required for clinical deployment. This study introduces StressGAT, a Graph Attention Network that leverages the relational inductive bias of graph modeling to capture complex facial dynamics that indicate acute stress. By using Differential Action Units, the framework normalizes individual responses relative to neutral baselines to achieve personalized recognition. The proposed model achieves 88.62% accuracy on a diverse stress-induction cohort (58 participants) using a subject-independent, Leave-One-Subject-Out (LOSO) cross-validation protocol. Beyond predictive accuracy, the architecture integrates a Multiple Instance Learning (MIL) attention mechanism to identify peak stress intervals and reveal distinct expressivity phenotypes. By simultaneously optimizing for accuracy and interpretability, this framework provides a robust, explainable solution for personalized affective monitoring.
[CV-88] SubSplat: High-Resolution Pixel-aligned 3DGS via Sub-pixel Gaussian Reparameterization
链接: https://arxiv.org/abs/2607.20813
作者: Jiun Lee,Jaekwang Kim,Sangmin Lee
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Pixel-aligned Gaussian splatting enables efficient and generalizable novel-view synthesis. However, high-resolution rendering faces a critical trade-off where increasing input resolution improves detail at the expense of quadratically rising network computational cost. Conversely, maintaining low-resolution inputs stabilizes this cost but results in insufficient Gaussian density and artifacts. To address this, we propose SubSplat, which introduces Sub-pixel Gaussian Reparameterizer(SPGR) to subdivide primary Gaussians into fine-grained primitives, restoring structural density directly from low-resolution features. We further enhance the reparameterization quality through feature aggregation, which effectively captures high-frequency details across multiple views. Experiments on RealEstate10K and ACID demonstrate that SubSplat achieves high-fidelity rendering with superior efficiency. Our results validate that the proposed framework successfully resolves the trade-off between reparameterization fidelity and network computational cost inherent in pixel-aligned Gaussian Splatting.
[CV-89] Ocular Verification for Virtual Reality
链接: https://arxiv.org/abs/2607.20790
作者: Husanpreet Singh,Robert Tran,Ayushree Kharel,Sudipta Banerjee
类目: Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Virtual reality (VR) headsets (e.g., Meta Quest, Apple Vision Pro) provide a seamless user experience due to their fast, frictionless interaction with the physical world in a simulated environment. User authentication relies on biometric cues such as iris in such headsets. However, traditional iris recognition protocols may not be adequate in cases of unconstrained acquisition, which is typical of VR-based data. In this work, we examine three crucial aspects: (1) evaluating ISO/IEC 29794-6 iris quality metrics on VRBiom dataset and analyzing their limitations, (2) addressing data-specific challenges such as off-axis gaze, non-uniform illumination, and specular reflection using generative models, and (3) performing unimodal (iris, periocular) recognition and multimodal score-level fusion (iris + periocular). We observe that some metrics (e.g., margin adequacy) fail on VR-acquired data; whereas, image adjustments primarily benefit periocular recognition, and multimodal fusion lowers EER by ~11% over unimodal iris recognition performance. We will release the evaluation scripts upon acceptance for reproducibility.
[CV-90] 3D-GIMP: When 3D Gaussian Inpainting Meets PatchMatch
链接: https://arxiv.org/abs/2607.20789
作者: Xuening Tian,Dieter Schmalstieg,Shohei Mori
类目: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
备注: 15 pages
Abstract:Recent advances in 3D scene editing have leveraged iterative diffusion models to update input views. However, this process is computationally expensive and struggles to produce sharp details. Meanwhile, ``hallucination drift’’ frequently introduces multi-view inconsistencies, leading to structural artifacts when rendering novel viewpoints. To address this problem, we present 3D-GIMP (3D Gaussian Inpainting Meets Patch Matching), a novel hybrid paradigm designed for high-fidelity object removal in 3D Gaussian Splatting. Instead of diffusing every view, 3D-GIMP performs a single generative inpainting on a key reference view, which serves as an appearance prior. We then introduce a 3D-aware PatchMatch algorithm to propagate these reference textures across all remaining views via correspondence matching, effectively bypassing the stochastic nature of frame-by-frame diffusion. By prioritizing reconstructive consistency over iterative generation, 3D-GIMP maintains high-frequency details across arbitrary resolutions while ensuring a mathematically consistent 3D reconstruction. Our experiments demonstrate that 3D-GIMP not only achieves competitive inpainting quality as previous methods using diffusion in multiple views, but also outperforms these methods in rendering speed and view consistency.
[CV-91] Rethinking Open-World Video Anomaly Detection: Diagnosing Definition Blindness
链接: https://arxiv.org/abs/2607.20780
作者: Inpyo Song,Jangwon Lee
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Preprint
Abstract:Open-world video anomaly detection (OWVAD) is expected to detect events that match a user-specified definition of abnormality. This requirement is stronger than generic anomaly localization: in the same video, changing the definition should change which temporal regions are scored as anomalous. We show that current OWVAD evaluation largely fails to isolate this conditional behavior. Standard VAD metrics and the dynamic-definition protocol can be dominated by target-versus-normal separation, allowing models to obtain strong scores while remaining nearly insensitive to the queried definition. We call this failure mode definition blindness. To explain why it is missed, we decompose dynamic-definition evaluation into target-versus-normal detection and target-versus-other-anomaly discrimination, and find that the former receives 7.2-26.8 \times more weight across common VAD benchmarks. Motivated by this diagnosis, we introduce three definition-conditioned evaluation metrics, DC-Disc, DC-Det \Delta , and DC-Sel \Delta , which progressively remove normal-frame, generic-anomaly, and multi-event selection shortcuts. Experiments on UCF-Crime, XD-Violence, and MSAD reveal that several strong VAD, OWVAD, and general vision language model baselines localize anomalous moments but exhibit weak definition following, often with near-zero definition-response margins. To validate that the failure is actionable, we further introduce DeCoS, a definition-contrastive scoring rule that subtracts anomaly evidence shared across definitions. DeCoS improves the strongest baseline by 7.3-16.0 AUROC points on DC-Disc and 15.5-28.3 points on DC-Det \Delta . Overall, our results argue that OWVAD should be evaluated as definition-conditioned anomaly scoring, not as anomaly detection under different prompt labels.
[CV-92] A real-time RGB-D perception pipeline for autonomous impact hammers in mining: self-filtering rock segmentation and rock-breaking poses generation
链接: https://arxiv.org/abs/2607.20748
作者: Martín Gallegos,Francisco Leiva,Patricio Loncomilla,Michelle Cortés,Javier Ruiz-del-Solar
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
备注: 25 pages, 20 figures
Abstract:Impact hammers, also known as rock-breakers, are essential machines in mining operations, where they perform secondary reduction. In underground mining, these machines are typically teleoperated, limiting operational efficiency. This paper presents a real-time RGB-D perception pipeline as a step towards automating the operation of hydraulic impact hammers used in mining. The proposed system simultaneously generates operationally feasible rock-breaking poses and a robot-free 3D representation of the workspace. The proposed approach combines image-based instance segmentation with geometric point cloud processing, and operates on embedded hardware at approximately 10 Hz with a total latency of around 675 ms, enabling responsive closed-loop behavior when integrated with a control system. Experimental results in a representative scaled scenario demonstrate that the proposed system is suitable for real-time autonomous impact hammer operation.
[CV-93] U-CFR: Uncertainty-Guided Cascade Forward Refinement for Interactive Segmentation ICPR2026
链接: https://arxiv.org/abs/2607.20705
作者: Elijah Danquah Darko,Min Xian,Terence Soule,Tiankai Yao,Matthew William Anderson
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: 12 pages, 3 figures, 4 tables, ICPR 2026
Abstract:Interactive image segmentation is critical for efficient image annotation; however, existing methods often require many corrective clicks or rely on passive refinement schemes that converge slowly. We propose Uncertainty-Guided Cascade Forward Refinement (U-CFR), a novel inference-time framework that enables models to autonomously self-correct after each user interaction. U-CFR introduces a boundary-aware uncertainty score that fuses segmentation uncertainty, contour gradients, and explicit edge predictions to guide the placement of internal pseudo-clicks. These self-generated clicks target the most ambiguous boundary regions, providing strong corrective signals without additional manual input. To support this process, we design a dual-head network with a shared encoder-decoder backbone: a segmentation head ensures region consistency, while an edge head sharpens boundary alignment. In inference, U-CFR launches a cascade of refinement steps, where each stage leverages the uncertainty-driven pseudo-clicks to refine the mask progressively. Experiments on standard benchmark datasets demonstrate that the proposed U-CFR improves click efficiency, initial mask quality, and boundary accuracy. It reduces the required clicks by over 10% on challenging datasets like Berkeley and offers a more intelligent and efficient interactive annotation.
[CV-94] DS@GT ARC at ImageCLEFmed GANs 2026: Geometric Filtering for Privacy-Preserving CT Slice Generation
链接: https://arxiv.org/abs/2607.20692
作者: Eric Regina,Richard Arnaud,Samir Hadi Cisneros
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注:
Abstract:We present a privacy-preserving framework for synthetic lung CT slice generation developed for the Image-CLEFmed GANs 2026 challenge. The approach combines Optimal Transport Conditional Flow Matching with privacy-oriented training and a post-generation “Supervisor” pipeline that filters generated candidates in learned geometric latent spaces using autoencoder embeddings, Determinantal Point Processes, and Stein Kernel Thinning. Official results show a strong realism-privacy trade-off, with the best-performing model achieving a Privacy Preservation Score of 0.549 and competitive visual fidelity with an FID of 0.3290. While the proposed geometric filtering substantially reduces nearest-neighbor memorization and membership-inference leakage, persistent patient re-identification scores indicate that preventing direct image copying is not sufficient to remove deeper patient-specific anatomical identity, highlighting an important frontier for future privacy-preserving medical image generation.
[CV-95] Spatially Grounded Concept Bottleneck Models for Trustworthy Breast Ultrasound Diagnosis ALT
链接: https://arxiv.org/abs/2607.20691
作者: Moshiur Rahman Tonmoy,Dunren Che,Haitham Y. Adarbah,Afzel Noore
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Accepted to the Workshop on Data Quality Aware, High-Performance, and Trustworthy AI Systems for Healthcare at IEEE/ACM CHASE 2026
Abstract:Concept Bottleneck Models provide interpretable-by-design predictions by mediating diagnosis through human-understandable concepts, but in medical imaging, their trustworthiness is often limited by the quality and granularity of available supervision. In particular, predicted concept activations can be driven by irrelevant regions, leading to spatially unfaithful explanations. We study a data-centric spatially grounded Concept Bottleneck Model (SG-CBM) that leverages coarse lesion delineations as weak supervision to encourage anatomically plausible concept evidence. For breast ultrasound, we derive two clinically motivated zones from each lesion mask: (i) an in-lesion region of interest for morphology-related concepts and (ii) a posterior acoustic band for posterior phenomena. We train concept maps using a grouped spatial grounding objective and preserve semantic faithfulness with a linear bottleneck classifier. Across five-fold stratified group cross-validation, the proposed SG-CBM improves diagnostic AUROC and concept macro-AUROC while markedly increasing spatial alignment of concept evidence. We also perform a Train-corrupt/Test-clean annotation-quality stress test to quantify the impact of supervision quality on diagnosis and spatial faithfulness. Overall, the results underscore the need for data-quality-aware supervision design and systematic trustworthiness validation for deployable healthcare AI systems.
[CV-96] ODeform: Learning Continuous 4D Motion for Shape Deformation with Neural ODEs IROS2026
链接: https://arxiv.org/abs/2607.20670
作者: Yordanka Velikova,Mahdi Saleh,Liming Kuang,Benjamin Busam
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted at IROS 2026
Abstract:Modeling continuous object deformation is important for many computer vision and robotics tasks, such as manipulation and simulation. Existing approaches rely on learning-based methods or physics simulators to model shape deformations. However, these approaches either use discrete time steps or are too computationally intensive for real-time applications. We present ODeform, a novel extension of Neural Ordinary Differential Equations to continuous 4D dynamics of deformable objects in 3D space. Our method transforms 3D point clouds and physical conditions (like material properties) into a unified latent space. By solving the resulting ordinary differential equations over time, we model deformations as continuous flows within this learned embedding, eliminating the need for discrete time steps while maintaining computational efficiency. We evaluate our approach on unseen physical parameter configurations, showing improved motion prediction accuracy over baseline methods. Our experiments further demonstrate a successful transfer to real 3D captured objects with novel shapes, along with effective interpolation and extrapolation of the learned dynamics. Our code and data will be made publicly available.
[CV-97] Axolotl3D: a Unified Framework for Faithful 3D Shape Completion ECCV2026
链接: https://arxiv.org/abs/2607.20660
作者: Anita Hu,Maria Shugrina
类目: Computer Vision and Pattern Recognition (cs.CV)
备注: Accepted to ECCV 2026
Abstract:Recent 3D generative models produce high-quality geometry from a single image using large-scale priors and diffusion architectures. However, they assume complete visibility and single-view inputs, limiting applicability in multi-view, occluded, or editing scenarios. Although prior works address these challenges individually, they lack a unified framework for controllable 3D completion under diverse conditioning signals. We present Axolotl3D, a multi-modal and occlusion-aware 3D generation model that jointly conditions on images, visibility masks, camera parameters, and a partial point cloud. The point cloud serves as a geometric anchor promoting faithful shape completion, while camera parameters ensure consistent multi-view alignment in a shared 3D coordinate system. A unified training strategy synthesizes diverse conditioning regimes from large-scale 3D data, enabling robust cross-modal reasoning. Experiments on Toys4K and OmniObject3D demonstrate state-of-the-art performance under both clean and occluded settings, as well as strong results in real-world reconstruction and geometry-consistent editing.
[CV-98] PhysCoRe: Physics-Corrected Residual World Models for Material-Aware Deformable Dynamics
链接: https://arxiv.org/abs/2607.20653
作者: Haocheng Yin,Shuohan Tao,Yongsheng Chen,Lu Gan
类目: Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Predicting how deformable objects evolve under robotic manipulation is a longstanding challenge. Existing approaches typically rely on per-object optimization to fit material parameters, which can be slow and cannot generalize, while end-to-end learned alternatives extrapolate poorly and often violate basic physical structure. We present PhysCoRe, a physics-corrected residual world model that couples a differentiable Material Point Method (MPM) simulator with two feed-forward neural networks. A material refinement module, Material from Motion (MfM), infers per-particle elasticity from visual observations, grounding the simulator in object-specific physics. A residual correction module, Residual from Dynamics (RfD), learns the discrepancy and predicts corrections to the simulator’s internal dynamics, absorbing systematic biases that the analytical model cannot capture. This design also supports online material identification on novel objects. MfM adapts from limited interactions, and its predictive uncertainty steers further exploration toward the regions where its estimate is least confident. Experiments on real deformable-object manipulation sequences show that PhysCoRe outperforms state-of-the-art baselines in prediction accuracy, and that its predicted confidence forms a reliable distribution across the object’s geometry, providing a natural signal for future confidence-guided exploration.
[CV-99] Masked Topology Modeling for Self-Supervised Learning on Parametric CAD
链接: https://arxiv.org/abs/2607.20642
作者: Heinrich Jiang,Jennifer Jang
类目: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Computer aided design (CAD) is ubiquitous: virtually any modern object was designed using editable CAD tools. However, with the shortage of available CAD datasets in its native editable and parametric format, boundary representation (B-Rep), it is ever more important to develop data-efficient methods for this domain. We present a new self-supervised pretraining task, Masked Topology Modeling (MTM), that leverages the face-adjacency graph, an induced structure unique to B-reps that the encoder can be asked to reconstruct. MTM masks a fraction of edges and trains a small head to predict each masked edge’s convexity and curve type from the encoder’s post-message-passing face features. We combine MTM with a MoCo-style momentum-queue contrastive learning over B-rep-aware augmentations, a BFS-connected face-region masked-reconstruction objective, and pretraining on the ABC dataset and our new procedurally generated dataset to show strong performance on a number of benchmarks. Subjects: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) Cite as: arXiv:2607.20642 [cs.CV] (or arXiv:2607.20642v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2607.20642 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-100] RealVDeblur: One-Step Diffusion for Generalizable Real-World Video Deblurring
链接: https://arxiv.org/abs/2607.20628
作者: Renbiao Jin,Mingxin Yang,Yutian Chen,Junhao Zhuang,Xin Cai,Mulin Yu,Linning Xu,Wenxian Yu,Danping Zou,Shi Guo,Tianfan Xue
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
备注: Project page with code: this https URL
Abstract:Real-world video deblurring remains challenging due to diverse motion patterns, complex degradations, and the scarcity of realistic training data, yet robust restoration is critical for downstream pipelines such as mobile imaging and 3D reconstruction. This work presents \textbfRealVDeblur, an efficient generative framework designed to improve in-the-wild robustness under diverse real capture conditions. First, a large-scale, physically grounded blur synthesis pipeline is constructed from scene-level 3D Gaussian Splatting (3DGS) assets and high-frame-rate videos, providing realistic training data covering both camera-induced and object-motion blur. Second, a video diffusion prior is leveraged for restoration; to better accommodate frame-dependent blur variations, temporal compression in the VAE is disabled and a frame-wise encoding scheme is adopted. For practical deployment on long videos, multi-step diffusion sampling is distilled into an efficient one-step generator, and a training-free Temporal Window Mask stabilizes inference beyond the training horizon with constant memory usage. Extensive experiments on diverse real-world benchmarks demonstrate strong perceptual quality, semantic fidelity, and temporal consistency on unseen videos, as well as improved robustness in downstream 3D reconstruction under severe motion blur. Project page: this https URL
[CV-101] Detecting Neural Network Failures through Spectral Analysis of Internal Activations ACML2026
链接: https://arxiv.org/abs/2607.20590
作者: Arunan J
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: Submitted for ACML 2026 , Under Review
Abstract:Neural network misclassifications exhibit characteristic spectral instability in internal activations that is invisible at the output layer. This phenomenon is identified and formalized as Spectral Drift – the frequency-domain distance between consecutive layer activations – with empirical validation showing that failures exhibit significantly higher drift than correct predictions (1.9% increase, p0.001). This spectral signature emerges during internal processing but becomes masked in final outputs, explaining why confidence-based detection methods struggle. This work introduces Self-Detecting Neural Networks (SDNN), a framework that monitors spectral dynamics across network depth using Short-Time Fourier Transform, wavelet decomposition, and statistical moments to capture multi-scale spectral features. A lightweight detector network (5% parameter overhead) learns to identify failure-indicative patterns via curriculum learning on progressively challenging distributions: natural misclassifications, distribution shifts, and adversarial perturbations. Experiments on CIFAR-10 demonstrate that SDNN achieves 79.0 +/- 25.3% AUROC across three seeds, substantially outperforming confidence-based baselines including MaxSoftmax (50.5%) and Energy Score (52.9%) by approximately 25-30 percentage points. Ablation studies reveal that wavelet decomposition and statistical features make consistent contributions, while STFT’s role remains unclear. This work establishes spectral analysis of internal activations as a promising direction for neural network reliability, revealing diagnostic information inaccessible to output-based approaches. Comments: Submitted for ACML 2026 , Under Review Subjects: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2607.20590 [cs.LG] (or arXiv:2607.20590v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2607.20590 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[CV-102] CT-Merging: Consensus Directions and Task-Level Scaling for LoRA Adapter Merging
链接: https://arxiv.org/abs/2607.20561
作者: Keumseo Ryum,Joonhyuk Kang
类目: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
备注: 9 pages, 5 figures
Abstract:LoRA adapters provide an efficient way to specialize a pretrained model for many downstream tasks, but deploying one adapter per task requires adapter storage and task selection at inference time. Model merging addresses this issue by combining independently trained adapters into one multi-task adapter. Recent SVD-based LoRA merging methods mainly focus on constructing shared or task specific directions, while the coefficients assigned to the final directions are often directly from the original task SVD. On a fixed merged basis, inherited coefficients preserve component order with high rank correlation, yet their magnitudes differ substantially from the coefficients induced by the task updates. To address this mismatch, we propose CT-Merging, a LoRA-aware merging algorithm that estimates consensus directions from average task subspace projectors and assigns task-level RMS coefficient scales in the final update. CT-Merging uses repeated support across task SVD subspaces to construct the common basis, while reducing reliance on rank wise SVD magnitudes after direction construction. On the DC-Merge CLIP adapter benchmark, CT-Merging achieves superior average normalized accuracy compared to state-of-the-art merging methods and further improves over DC-Merge by 2.56 points on ViT-B/32 and 1.51 points on ViT-L/14 KnoTS-trained checkpoints.
[CV-103] StrideDiffusion: Accelerating Diffusion Models for Time-series Generation
链接: https://arxiv.org/abs/2607.20545
作者: Du Yin,Estrid He,Julián Jerónimo Bañuelos,Yang Yang,Feng Hu,Yuchen Luo,Hao Xue,Stephan Sigg,Flora Salim
类目: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Software Engineering (cs.SE)
备注: Under Review
Abstract:Diffusion models have become competitive generators for time series, but their practical use is limited by the large number of sequential denoising steps required at inference time. Existing fast samplers typically use fixed or generic timestep schedules, overlooking a distinctive property of time-series diffusion: different spectral bands evolve at different rates during the reverse process. We introduce StrideDiffusion, a training-free spectral-aware sampler that adaptively selects the denoising stride from band-level activity. At each step, StrideDiffusion monitors relative band energy, log-power drift, and phase velocity to identify whether high- frequency dynamics remain active or whether the trajectory is dominated by stable low-frequency structure. It then takes fine steps when rapidly varying bands are active and larger jumps once only coarse components remain. A bandwise stability analysis shows that inactive frequency bands change only linearly with the jump size under deterministic affine reverse updates, providing a local justification for spectral activity as a step-size indicator. Across six unconditional time-series generation benchmarks, StrideDiffusion uses only 14-66 function evaluations instead of 500/1000 denoising steps, achieving up to 18.9x wall-clock speedup while preserving or improving generation quality. On conditional imputation and forecasting, it further delivers 5-14x average acceleration with comparable predictive accuracy. These results show that spectral evolution provides a practical and principled signal for fast time-series diffusion sampling. Our code is available at this https URL.
[CV-104] A Graph Neural Network approach to zero-shot Digital Twins
链接: https://arxiv.org/abs/2607.20535
作者: Alicia Tierz,Icíar Alfaro,David González,Elías Cueto
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
备注:
Abstract:Traditional Predictive Digital Twins often remain geometrically rigid, requiring extensive retraining or fine-tuning whenever the underlying physical domain or boundary conditions change. To overcome this limitation, we present a novel framework for \textitZero-Shot Digital Twins that seamlessly couples real-time visual perception with a geometry-agnostic, physics-informed reasoning engine. At the core of our architecture is the Thermodynamics-Informed Graph Neural Network architecture, a Geometric Deep Learning solver grounded in a metriplectic thermodynamic formalism that enforces energy conservation and non-negative entropy production locally through graph message passing. The framework integrates an auxiliary Graph Neural Network to infer unobservable fields (such as stress tensors or velocity and energy distributions) directly from sparse initial visual boundaries, mitigating numerical start-up transients. To bridge the sim-to-real gap, we implement a continuous closed-loop data assimilation mechanism; the pipeline tracks macroscopic deformations and free-surface fluid boundaries in real-time using deep segmentation networks combined with sparse optical flow, dynamically correcting the autoregressive simulation rollout and eliminating numerical drift. To test the validity of our approach, we demonstrate the extreme generalization capabilities of our approach across two disparate physical regimes: the large deformations of a viscoelastic beam and the non-linear sloshing of a viscous fluid. In both scenarios, the unified framework instantiates physically accurate simulations on novel, unseen geometries without case-specific retraining, operating well within real-time latency budgets (approximately 25 ms per frame) and enabling the direct projection of latent mechanical variables via Augmented Reality.
[CV-105] Deblurring in the Wild: A Real-World Image Deblurring Dataset from Smartphone High-Speed Videos
链接: https://arxiv.org/abs/2506.19445
作者: Syed Mumtahin Mahmud,Mahdi Mohd Hossain Noki,Prothito Shovon Majumder,Abdul Mohaimen Al Radi,Sudipto Das Sukanto,Afia Lubaina,Md. Mosaddek Khan
类目: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 8 pages (without references), 3 figures. Dataset this https URL
Abstract:We introduce the largest real-world image deblurring dataset constructed from smartphone slow-motion videos. Using 240 frames captured over one second, we simulate realistic long-exposure blur by averaging frames to produce blurry images, while using the temporally centered frame as the sharp reference. Our dataset contains over 42,000 high-resolution blur-sharp image pairs, making it approximately 10 times larger than widely used datasets, with 8 times the amount of different scenes, including indoor and outdoor environments, with varying object and camera motions. We benchmark multiple state-of-the-art (SOTA) deblurring models on our dataset and observe significant performance degradation, highlighting the complexity and diversity of our benchmark. Our dataset serves as a challenging new benchmark to facilitate robust and generalizable deblurring models.
[CV-106] From Attention to Frequency: Integration of Vision Transformer and FFT-ReLU for Enhanced Image Deblurring
链接: https://arxiv.org/abs/2511.10806
作者: Syed Mumtahin Mahmud,Mahdi Mohd Hossain Noki,Prothito Shovon Majumder,Abdul Mohaimen Al Radi,Md. Haider Ali,Md. Mosaddek Khan
类目: Image and Video Processing (eess.IV); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
备注:
Abstract:Image deblurring is vital in computer vision, aiming to recover sharp images from blurry ones caused by motion or camera shake. While deep learning approaches such as CNNs and Vision Transformers (ViTs) have advanced this field, they often struggle with complex or high-resolution blur and computational demands. We propose a new dual-domain architecture that unifies Vision Transformers with a frequency-domain FFT-ReLU module, explicitly bridging spatial attention modeling and frequency sparsity. In this structure, the ViT backbone captures local and global dependencies, while the FFT-ReLU component enforces frequency-domain sparsity to suppress blur-related artifacts and preserve fine details. Extensive experiments on benchmark datasets demonstrate that this architecture achieves superior PSNR, SSIM, and perceptual quality compared to state-of-the-art models. Both quantitative metrics, qualitative comparisons, and human preference evaluations confirm its effectiveness, establishing a practical and generalizable paradigm for real-world image restoration.
人工智能
[AI-0] Beyond Sufficiency: Time Series Explanation with Counterfactual Necessity
链接: https://arxiv.org/abs/2607.21573
作者: Hongnan Ma,Yiwei Shi,Mengyue Yang,Weiru Liu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Faithful explanations of time-series classifiers should identify subsequences that are not only sufficient to preserve a black-box model’s prediction, but also necessary for maintaining it. However, existing sufficiency-oriented methods can assign high importance to spurious subsequences that support the prediction without being essential to the model’s decision. We introduce \textbfTimePNS, a necessity-aware framework for time-series explanation. Inspired by Pearl’s counterfactual notion of necessity, TimePNS assesses whether a temporal factor is necessary by intervening on it and measuring whether the original prediction is disrupted. The framework adopts a two-stage design. Stage I learns an identifiable causal generative process together with a sufficiency-oriented explanation mask. Stage II performs counterfactual interventions on temporal factors to derive necessity signals, which supervise a temporal gate that refines the initial explanation by suppressing non-essential components and emphasizing counterfactually necessary ones. Experiments on synthetic and real-world time-series benchmarks show that TimePNS more accurately identifies decision-critical subsequences and consistently improves sufficiency-necessity trade-offs over strong baselines.
[AI-1] Unsupervised Consensus-Based Anomaly Detection for Spatiotemporal Malaria Incidence in Ghana
链接: https://arxiv.org/abs/2607.21559
作者: T. Ansah-Narh,Y. Asare Afrane
类目: Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Emerging Technologies (cs.ET); Applications (stat.AP); Machine Learning (stat.ML)
备注: 32, 15 figures, under review at spatial and spatio-temporal epidemiology
Abstract:A consensus anomaly detection framework was applied to monthly malaria surveillance data from Ghana (2014-2023) to identify atypical transmission patterns. Anomalies were highly structured in space and time. Ashanti and Northern Regions accounted for most recurrent anomalies, with persistent hotspots at Tamale, Kumasi, and Accra. A key finding was the spatial distinction between anomaly burden (cumulative cases during anomalous periods) and anomaly frequency (persistence of unusual behaviour). Tamale had the highest burden during anomalies, whereas the highest anomaly rates clustered in Ashanti districts, showing that high-burden areas are not necessarily those with the most frequent anomalous transmission. Anomalous months formed a statistically distinct group, with much higher case counts (Cohen’s d = 3.252 ) and large seasonal deviations ( d 1.2 ) compared with normal months. Malaria burden alone provides an incomplete picture of transmission dynamics. By distinguishing where malaria is most prevalent from where transmission behaves most unusually, this framework can strengthen surveillance, prioritise investigations, and support targeted control strategies.
[AI-2] Beyond Sycophancy: Structured Resistance and Compliance in LLM Moral Reasoning
链接: https://arxiv.org/abs/2607.21558
作者: Baihui Wang,Bernard Koch
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Building socially calibrated large language models, which can learn from others without simply yielding to them, requires more than reducing sycophancy as a one-dimensional failure mode. Models must distinguish when to incorporate others’ perspectives from when to maintain a well-grounded moral judgment. We study the broader resistance-compliance process governing this distinction. Across three studies, we show that models’ judgment revision is structured along three dimensions that parallel classic phenomena in human social psychology: the distance between an incoming view and the model’s initial position, the source attribution of that view, and the coalition structure supporting it. Models are generally more receptive to nearby positions, more influenced by views presented as their own prior judgments, and differently responsive to group pressure. These findings recast sycophancy as one expression of a broader judgment-updating process shaped by social influence. Our framework provides a principled basis for distinguishing constructive belief revision from sycophantic compliance, thereby supporting better alignment in morally consequential interactions.
[AI-3] MIRROR: Learning from the Other View for Multi-Modal Reasoning
链接: https://arxiv.org/abs/2607.21552
作者: Wen Ye,Yuxiao Qu,Aviral Kumar,Xuezhe Ma
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Unlike large language models (LLMs) that exhibit strong reasoning capabilities, vision-language models (VLMs) struggle with visual reasoning, even on geometry problems that admit equivalent text, diagram, and combined diagram+text views. We show that these views often elicit different behaviors: a model may solve a problem from text but fail on the corresponding diagram, or succeed visually while failing textually. This inconsistency suggests that different views expose complementary reasoning paths and failure modes that standard multimodal post-training does not fully exploit. To study and exploit this phenomenon, we construct ODA-Data, a high-quality paired multimodal geometry dataset with text-dominant, image-dominant, and combined image+text views of the same problems, together with splits for training and evaluating modality-dependent reasoning behaviors. We then develop Modality-Informed Reciprocal Reasoning Optimization (MIRROR), a reinforcement learning approach for improving multimodal reasoning via self supervision. For each problem, MIRROR evaluates the model under all views, selects the best-performing view as a teacher, and trains other views with a reverse-KL objective towards the teacher. Across reasoning benchmarks that evaluate on geometry problems, MIRROR improves over standard RL and yields more accurate and consistent behavior across modalities
[AI-4] From Resource Flow to Executable Tests: Petri-Net-Guided LLM Test Generation for Concurrent Stateful Rust APIs
链接: https://arxiv.org/abs/2607.21530
作者: Kaiwen Zhang,Guanjun Liu
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注:
Abstract:Concurrent stateful library APIs expose behavior through evolving resource ownership, lifecycle states, and competing interleavings. Large language models can synthesize executable Rust tests, but their outputs often violate API preconditions, remain shallow, or reduce concurrency to accidental sequential traces. Conversely, model-based and systematic testing techniques provide semantic control but commonly require substantial handwritten code to turn abstract scenarios into executable tests. This paper addresses the gap between formal scenario design and low-cost test concretization. We present a Petri-net-guided methodology for test generation over concurrent stateful Rust APIs. The method represents API resources, lifecycle conditions, and causal dependencies as colored tokens and transitions; derives legal deep-state, near-legal, and partial-order concurrent scenarios; and uses these scenarios as a constrained intermediate representation for LLM-based code synthesis. A local-faithfulness contract and structural repair loop preserve the modeled intent during concretization, while Petri-guided schedule shaping prioritizes high-conflict concurrency skeletons for systematic exploration. A layered semantic oracle then distinguishes synthesis failures from violations of the target API’s expected behavior.
[AI-5] Same Dangerous Objective Opposite Advice: Direct Exposure versus Multi-Agent Mediation
链接: https://arxiv.org/abs/2607.21518
作者: Linjun Li
类目: Artificial Intelligence (cs.AI)
备注: 21 pages; welcome comments
Abstract:Even a current high-capability LLM can appear safer when shown a dangerous objective directly than when other agents transform and relay its direction. Using OpenAI’s gpt-5.6-sol model alias, we test 25 pre-specified mirrored trade-off profiles. Direct exposure to an objective authorizing concealment, fabrication, and pressure produced advice net opposed to its target. After an Id and Censor transformed the same objective into affect and a constraint-rewritten, target-bearing intention, the user-facing Superego—which saw the preferred direction but not the raw objective, its manipulative clauses, or its source—produced advice net aligned with the target. This behavioral reverse shift is consistent with the model recognizing or distrusting the manipulative motive, although we do not identify its internal mechanism. The second result exposes a compositional safety gap: a current high-capability model can be used as the user-facing component of an automated, multi-stage workflow serving an explicitly manipulative objective. The workflow can keep the raw instruction, its manipulation-authorizing clauses, and its provenance outside the downstream model’s context while preserving the objective’s target direction. A user with endpoint-only access likewise cannot directly inspect those upstream messages including the objective. Comments: 21 pages; welcome comments Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2607.21518 [cs.AI] (or arXiv:2607.21518v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2607.21518 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-6] Improved lower bounds for the Shannon capacity of odd cycles
链接: https://arxiv.org/abs/2607.21517
作者: Nathaniel Itty,Christopher D. Rosin,Chase Carstensen,Daniel Reichman
类目: Information Theory (cs.IT); Artificial Intelligence (cs.AI); Discrete Mathematics (cs.DM); Combinatorics (math.CO)
备注:
Abstract:The Shannon capacity \Theta(G) of a graph G quantifies the maximum rate at which information can be transmitted with zero error over a noisy channel. It is lower bounded by \alpha(G^d)^1/d for any d , where \alpha(G^d) is the independence number of the d -th strong power of G . We construct independent sets of size 134753 in C_7^10 , 21909 in C_11^6 , and 62530 in C_13^6 , improving the best known lower bounds for the Shannon capacity of these graphs to \Theta(C_7)\geq 134753^1/103.258020 , \Theta(C_11)\geq 21909^1/65.289773 , and \Theta(C_13)\geq 62530^1/66.300109 . We also improve the best known lower bounds on the independence numbers of several individual strong powers of odd cycles that do not improve the Shannon capacity lower bound. The constructions were discovered through iterative interactions with a Large Language Model (LLM), illustrating the potential of LLMs for finding explicit combinatorial constructions.
[AI-7] AREX: Towards a Recursively Self-Improving Agent for Deep Research
链接: https://arxiv.org/abs/2607.21461
作者: Shuqi Lu,Chaofan Li,Kun Luo,Zhang Zhang,Hui Wang,Hongwang Xiao,Zheng Liu,Lei Xiong,Jiahao Wang,Sen Wang,Xiyan Jiang,Wanli Li,Yuyang Hu,Hongjin Qian,Bingyu Yan,Ziyi Xia,Yingxia Shao,Kang Liu,Zhicheng Dou,Di He,Chaozhuo Li,Qiwei Ye,Zhongyuan Wang,Zheng Liu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Deep research requires agents to find answers that jointly satisfy multiple constraints. Discovering such answers is costly, whereas verifying a candidate can often be decomposed into tractable constraint-wise checks. This discovery–verification asymmetry suggests that a research agent should do more than simply search longer: it should recursively improve its current answer by verifying intermediate results and using the partially verified state to guide subsequent refinement. We introduce AREX, a family of Recursively Self-Improving (RSI) deep research agents. AREX alternates between an inner research loop that gathers evidence and constructs a provisional answer, and an outer self-improvement loop that audits the answer constraint-wise, identifies unresolved claims, and launches targeted follow-up research. To sustain RSI over long horizons, AREX learns an autonomous context-update tool that compresses growing interaction history into a compact improvement state preserving verified evidence and unresolved constraints, without relying on an external model. We train AREX on verified synthetic tasks and high-quality trajectories through agentic mid-training and long-horizon reinforcement learning. To mitigate sparse final rewards during long horizon learning, we emphasize key steps where decisive evidence is acquired or erroneous research directions are corrected. We instantiate a dense 4B model and a 122B-A10B Mixture-of-Experts model. Across BrowseComp, WideSearch, DeepSearchQA, Humanity’s Last Exam (HLE), and other reasoning and tool-use benchmarks, AREX substantially outperforms comparable-scale baselines and remains competitive with models using substantially more activated parameters.
[AI-8] Detecting LLM -Generated Tokens in Human–LLM Coauthored Text
链接: https://arxiv.org/abs/2607.21458
作者: Yangjun Lu,Hongyi Zhou,Fabian Spill,Kai Ye,Chengchun Shi,Jin Zhu
类目: Artificial Intelligence (cs.AI); Methodology (stat.ME)
备注:
Abstract:The rise of human-AI collaborative writing has created a growing need for fine-grained detection methods that support localizing likely LLM-generated content in mixed-authorship documents. Existing methods for detecting LLM-generated text mainly focus on document-level classification and cannot identify which parts of the text are generated by LLMs. This paper introduces a new method to address this urgent need. Our method operates at the token level, the natural unit of modern language models, and builds on existing token-level detection scores. The key idea is to smooth adjacent token scores to reduce their variability, while using an adaptive Lepski-type rule to select the bandwidth according to the local authorship structure. Our method is simple to implement and does not require token-level labeled data for training. Theoretically, we characterize this trade-off and show that the proposed method achieves favorable mean square error performance in estimating the underlying signal. Empirically, we demonstrate strong performance of our method against a wide range of baselines in both synthetic datasets and a realistic dataset. We deploy a publicly accessible website that implements the methods as well.
[AI-9] Agent -Guided Relational Concept Discovery: Toward Interpretable Surgical Margin Assessment MICCAI2026
链接: https://arxiv.org/abs/2607.21437
作者: Nooshin Maghsoodi,Amoon Jamzad,Robert Policelli,Mohammad Farahmand,Dilakshan Srikanthan,Martin Kaufmann,Kevin Y. M. Ren,Shaila Merchant,Sonal Varma,Ross Walker,Doug McKay,John Rudan,Gabor Fichtinger,Parvin Mousavi
类目: Artificial Intelligence (cs.AI)
备注: This paper is accepted to MICCAI 2026, and this is the submission version, not the camera-ready version
Abstract:Deep learning models can effectively use Rapid Evaporative Ionization Mass Spectrometry (REIMS) data for surgical margin assessment. However, their clinical adoption remains challenging due to limited generalization to operating room conditions. This difficulty arises because models are typically trained on labeled spectra collected from resected tissue samples, while they must operate on noisy, unlabeled data acquired directly during surgery. In addition, the black-box nature of deep learning models makes it difficult to understand and systematically improve their behavior. Concept-based learning offers a promising way to address these challenges by mapping raw measurements to human-understandable concepts. However, supervised concept-based approaches rely on concept annotations, which are difficult to obtain in complex mass spectrometry workflows. We propose Agent-Guided Concept Discovery, a framework that learns meaningful concepts directly from data without requiring predefined concept labels. During training, a reasoning agent refines semantic descriptions of the learned concepts and adaptively adjusts their weight based on diagnostic relevance. These concepts are further grounded using a biochemical knowledge graph to ensure consistency with known metabolic relationships. Across Skin and Breast Cancer datasets, our model improves balanced accuracy and sensitivity over the baseline. In a representative intraoperative case, it shows fewer false positives, indicating better generalization to surgical conditions.
[AI-10] Bridging the Gap Between Plausibility and Admissibility: Constraint-Aware Flow Maps for Dynamic Graph Systems
链接: https://arxiv.org/abs/2607.21421
作者: Michael Romei de Socio,Gian Luca Pozzato,Alessio Merlo
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Generative models can support decision-making under uncertainty by producing ensembles of plausible future system trajectories, but statistical plausibility does not ensure structural feasibility. This study investigates whether post-sampling symbolic constraints can improve the reliability of generative trajectory modeling in dynamic graph-structured systems. A conditional diffusion model generates future graph-state trajectories from partial observations, while an external symbolic layer applies hard filtering, soft weighting, or projection-based repair. The framework is evaluated on two controlled synthetic regimes: a compact graph and a medium-complexity dependency graph, using metrics for structural validity, sample efficiency, diversity, robustness, and calibration. In the compact regime, the model produces an invalid probability mass of 0.002996, indicating an almost entirely admissible trajectory manifold. Under the same architecture and training protocol, invalid mass increases to 0.155929 in the medium-complexity regime. Hard filtering removes all invalid retained trajectories while preserving 84.4% of generated samples, whereas soft weighting preserves effective sample size but yields only limited validity gains. Family-level analysis shows that dependency constraints account for nearly all observed inadmissibility. These results indicate that statistical plausibility and structural admissibility are distinct reliability properties and that symbolic constraint handling becomes more valuable as graph-structural complexity increases.
[AI-11] PATS: Policy-Aware Training Scaffolding for Agent ic Reinforcement Learning
链接: https://arxiv.org/abs/2607.21419
作者: Yipeng Shi,Zhipeng Ma,Yue Wang,Qitai Tan,Yang Li,Peng Chen,Zhengzhou Zhu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:In long-horizon LLM agent reinforcement learning, weak policies often repeat similar failures, producing uninformative rollout trajectories and limiting effective policy optimization. Existing skill-centric methods improve exploration by optimizing, filtering, or internalizing reusable skills. However, they remain centered on the skills themselves rather than being designed as adaptive training-time support for the evolving policy. To address this, we propose a policy-centric training paradigm that reframes skills as a dynamic training scaffold. Our framework, Pats, converts rollout groups from the latest policy into evidence cards and uses task-specific evaluation to adjust the context used in subsequent rollouts. Concrete guidance helps weak policies to complete challenging tasks. As policy improves, redundant context is revised or removed to reduce reliance on explicit guidance while preserving useful rollout variation. The policy is optimized with environmental rewards using standard RLVR, and the training scaffold is discarded at deployment. On ALFWorld and WebShop, Pats improves over strong baselines by up to 18.6%. Across seven search-augmented QA benchmarks, it remains competitive while using 32.1% fewer prompt tokens than the baseline.
[AI-12] Logical Regression for Planning with Axioms
链接: https://arxiv.org/abs/2607.21414
作者: Connor Little,Christian Muise
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:In automated planning, logical regression is an operation that returns the most general condition necessary for an action to achieve a particular formula. It has many applications, such as allowing for more robust plan execution and providing compact policies for non-deterministic planning. Although relatively simple to calculate in basic planning settings, logical regression becomes significantly more complex when additional factors, such as axioms, are present. We introduce a methodology for approximating the logical regression of an action in a domain that includes axioms; an approximation that limits conditions to partial states. Our method produces minimal partial states while avoiding the recalculation of axioms. To demonstrate the impact of our methods, we embed our form of regression in an execution monitoring context, a well-established setting that can benefit greatly from logical regression. Our results show that this form of regression can dramatically generalize partial states across multiple domains, reducing the number of variables considered for execution monitoring by up to 70%, and demonstrate that the resulting execution monitor is robust enough to recover frequently in an environment with unexpected changes: several domains recover over 50% of the time in our tests.
[AI-13] MSBraM: A Multi-scale Self-supervised Brain Foundation Model for Hierarchical EEG Dynamics Learning
链接: https://arxiv.org/abs/2607.21402
作者: Tao Zhou,Jing Han,Lingyu Shu,Zixing Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Self-supervised foundation models have recently shown strong potential for electroencephalogram (EEG)-based analysis. However, existing approaches struggle to capture the inherently multi-scale temporal structure of EEG signals, where local neural patterns and long-range dependencies jointly encode task-relevant information. This limitation hampers cross-scale representation learning and generalization across diverse downstream tasks. To address this challenge, we propose MSBraM, a Multi-Scale self-supervised Brain foundation Model designed to learn hierarchical EEG representations. MSBraM follows a two-stage pretraining framework. First, a multi-scale neural tokenizer discretizes raw EEG signals into semantic codes at different temporal resolutions via vector-quantized reconstruction. Second, the model is pretrained to predict masked codes using a curriculum multi-scale masking strategy, progressively integrating fine-grained local patterns with global temporal context. We pretrain MSBraM on over 2,400 hours of EEG data and evaluate it across 10 downstream tasks on 12 public datasets. Extensive experiments show that MSBraM achieves superior performance on other state-of-the-art pretrained models, demonstrating strong generalization and transferability. These results indicate that explicitly modeling multi-scale temporal dynamics is critical for effective EEG foundation models.
[AI-14] VoLN: Vision-Only Long-Horizon Navigation—Paradigm Benchmark and Method
链接: https://arxiv.org/abs/2607.21400
作者: Jiabin Lou,Haopeng Wang,Yuanshuai Wang,Xinyu Liu,Xuxin Lv,Yuxin Guo,Lei Huang,Rongye Shi,Wenjun Wu
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注: 10 pages, 7 figures, 2 tables. Project page: this https URL
Abstract:Vision-and-Language Navigation (VLN) enables embodied agents to follow natural-language instructions. However, route-level instructions commonly encode spatial priors, such as orientation, distance, and layout, that are not explicitly available from onboard sensing at deployment in open, GPS-denied environments. Benchmark performance under such interfaces therefore jointly reflects visual navigation ability and the use of route structure explicitly supplied by the task description. As a complementary formulation, we propose Vision-Only Long-Horizon Navigation (VoLN), which shifts route-relevant information from externally supplied instructions and global guidance to locally observable in-scene cues. In VoLN, goal views specify the destination, while route-relevant information is available only through locally observable in-scene cues that the agent must detect, interpret, and select online. We instantiate VoLN for aerial navigation through VoLN-UAV, a 7,210-episode benchmark that combines long-horizon goal-directed flight, continuous 3D motion, large viewpoint changes, and context-dependent beacon selection. We further provide VoLN-MLLM as an initial reference baseline. It aligns self-supervised visual features with a structured semantic space and predicts short-horizon waypoint segments from observation history, goal views, retrieved visual–semantic tokens, and proprioception. On the five-environment Test-Unseen split, it obtains success rates of 7.4%, 4.5%, and 1.8% on Easy, Normal, and Hard episodes, respectively. These results provide an initial evaluation of VoLN and reveal substantial remaining challenges in long-horizon evidence integration, cross-view goal matching, and closed-loop stability. Project page: this https URL
[AI-15] Multimodal Pretraining for Generalizable EEG Representation Learning
链接: https://arxiv.org/abs/2607.21384
作者: Targol Bakhtiarvand,Jugal Kalita,Adham Atyabi
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Electroencephalography (EEG) models used for epilepsy are often limited to specific datasets and tasks. This limited approach can make it challenging to apply these models across different datasets or in various situations. However, recent studies in foundation models and self-supervised learning suggest that an adaptable EEG backbone could support a range of EEG related tasks. In this study, we have developed a multimodal EEG foundation model that combines a raw signal encoder based on the Mamba architecture, a Vision Transformer (ViT)-style encoder for time-frequency data, and a lightweight encoder for text, all within a shared embedding space. The pretraining process relies on several innovative techniques, such as masked modeling, cross-view contrastive alignment, and temporal consistency losses. These methods are designed to create rich, seizure-relevant representations without requiring labeled data. To assess the efficacy and generalization of our pretrained model, we fine-tuned it on the canonical CHB-MIT seizure detection benchmark and additional seizure detection datasets, and conducted extensive experiments comparing different model variants. On the standard CHB-MIT split, our best single model achieved an AUROC of 0.874, and an ensemble variant reached 0.878 AUROC, representing state-of-the-art performance on this benchmark. In addition to standard train-test splits, we evaluated performance under a leave-one-subject-out (LOSO) protocol, which is rarely reported in prior EEG seizure modeling work and highlights the difficulty of patient-independent seizure detection, with a mean LOSO balanced accuracy of 0.558 across 19 subjects. Across datasets and evaluation settings, our multimodal foundation model enabled robust seizure detection and straightforward adaptation to new seizure detection scenarios, while also supporting interpretable seizure localization.
[AI-16] owards Faithful Graph Explanations with Synergistic Edge Effects via Granular Balls
链接: https://arxiv.org/abs/2607.21381
作者: Jiancu Chen,Shuyin Xia,Guan Wang,Degang Chen,Fan Chen
类目: Artificial Intelligence (cs.AI)
备注: 11pages 24 figures
Abstract:Instance-level explanations aim to reveal the rationale behind a model’s decisions for a specific graph. Previous methods explain graph neural networks (GNNs) by selecting important edges to induce subgraphs, where edge importance is assessed by perturbing each edge and observing changes in the model predictions. However, they often neglect the synergistic effects among edges, which are crucial for accurately characterizing edge importance. To address this issue, we propose SeeExplainer, a parameter-free explainer to interpret GNNs. Specifically, we first introduce a granular-ball graph refinement mechanism that decomposes a graph into several disjoint granular-balls with no fixed size, and utilize them as nodes to construct a structural graph. This process can better capture the synergistic effects among edges. Then, we perturb nodes and edges in the structural graph to generate explanatory subgraphs based on their respective contributions. Experiments on several graph classification datasets of different networks show that SeeExplainer outperforms state-of-the-art baselines.
[AI-17] Mean-to-Score Discrete Diffusion: Posterior-Mean Denoisers for Score Entropy
链接: https://arxiv.org/abs/2607.21372
作者: Jingyuan Li,Xiaoyi Jiang,Yixuan Jiang,Wei Liu,Yi Zhu,Zuoqiang Shi,Pipi Hu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Score Entropy Discrete Diffusion (SEDD) parameterizes discrete reverse processes with unconstrained positive score ratios. While positivity guarantees nonnegative reverse jump rates, it does not ensure Bayes realizability: ratios at a noisy state need not be jointly induced by any clean-token posterior under the forward kernel. The score-entropy loss has the correct population optimum but does not enforce this constraint away from it. In a trained pure-uniform SEDD checkpoint, roughly one quarter of complete score vectors violate the coordinate box, while more than half lie inside it yet remain materially incompatible with any valid posterior. Such violations can produce negative pre-normalization weights in finite-step sampling. Projecting raw scores onto the bridge polytope removes all observed negative weights and improves external generative PPL from 203.6 to 175.1 without changing the sampler. We introduce \emphmean-to-score (M2S), which predicts a clean-token posterior mean and converts it to the score through an exact kernel-dependent linear map. The construction applies to any known coordinate-wise continuous-time Markov chain (CTMC) satisfying a mild support condition. For uniform corruption, it maps the probability simplex onto the bridge polytope; for absorbing-mask corruption, the resulting objective recovers MD4 exactly. In a controlled 28.4M-parameter CIFAR-10 comparison, M2S lowers test BPD from 3.173 to 3.129 and FID-50k from \CifarSEDDFID to \CifarMtwoSFID . A 170M-parameter M2S model trained on about 262B OpenWebText token slots outperforms the evaluated pure-uniform SEDD, GIDD, and Neural CTMC checkpoints at every tested sampling budget, reaching generative PPL 143.3 at 128 steps versus 183.6 for the strongest pure-uniform baseline.
[AI-18] Hilbert Operator for Progressive Encoding (HOPE): A Mathematical Framework for Deconstructing Learned Representations in Deep Networks
链接: https://arxiv.org/abs/2607.21366
作者: Hossein Mobahi,Peter L. Bartlett
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:Deep neural networks encode complex representations, but deconstructing this internal knowledge remains a challenge. Given the link between learning and compression, network compression offers a promising lens to analyze this knowledge. However, standard compression heuristics often suffer from scale symmetries and architectural biases. To resolve these, we introduce Hilbert Operator for Progressive Encoding (HOPE), a mathematical framework to gradually deconstruct the representations in trained network weights. HOPE shifts network compression from the discrete domain into a Hilbert space of continuous functions. By modeling individual neurons as rank-1 Hilbert-Schmidt operators, HOPE unifies pruning and neuron merging as low-rank subspace projection. Extending this formulation, HOPE introduces macro block eviction to encompass multi-layer structures like entire residual pathways under the same unified metric. This unified approach enables unbiased architectural decisions across layers with different types and sizes. HOPE is a data-free and hyperparameter-free framework. We present proof-of-concept experiments in model compression and fine-tuning to highlight the practical potential of our theory. Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML) Cite as: arXiv:2607.21366 [cs.LG] (or arXiv:2607.21366v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2607.21366 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-19] SPORD: A Simulation-Propose-then-OR-Dispose Approach for Supply Chain Planning
链接: https://arxiv.org/abs/2607.21354
作者: Jiayin He,Yutong Pan,Sen Yang,Ningxuan Kang,Yongzhi Qi,Jianshen Zhang,Wei Qi,Zuo-Jun Max Shen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:For years, supply chain planning at e-commerce firms has operated as a collection of isolated projects. Each planning task from static network planning to dynamic warehouse assortment planning requires analysts to spend weeks building models from scratch, calibrating and persuading executives to act on outputs they cannot verify. Three barriers drive this: bespoke models proliferate because standardization is difficult (operational fragmentation); once unified, the combinatorial scale of millions of SKUs, thousands of nodes, and intricate routing logic exceeds what solvers can handle within a tight window (computational intractability); and a mathematically optimal solution still fails to be implemented if the executives do not trust it (implementation hurdle). To bridge this gap, we propose and implement the Simulation-Propose-then-OR-Dispose method, deployed as this http URL’s NetSim platform. The central insight is decoupling: simulation proposes by generating and evaluating the full set of operationally valid candidate paths-absorbing all idiosyncratic business logic, while an integer program disposes by selecting the globally optimal subset. Computationally, matrix-vectorized CPU/GPU accelerated simulation achieves a 10-100 times speedup over serial methods, and a list scheduling algorithm reduces coupled-order processing from hours to minutes. Operationally, we establish a closed loop via an intelligent diagnosis engine. Since 2025, NetSim has optimized end to-end services for over 20,000 suppliers, the cross-regional fulfillment rate dropped from 6.1% to 4.9%, and the average monthly carbon reduction is approximately 5,745 tCO2e. SPORD moves simulation from monitoring to active planning. The transparent outputs turn skeptical executives into engaged collaborators, and the modular architecture ensures that the next planning requires just configuration, not reconstruction.
[AI-20] Regulating autonomous and agent ic AI
链接: https://arxiv.org/abs/2607.21345
作者: Chris Reed,Alex Austria,Anmol Bharuka,Pragnitha Mandava,Khushiya Mujawar,Luka Shakhkulashvili
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Regulating activities where regulatees use autonomous and agentic AI is challenging. Regulatory assumptions about regulatee knowledge and control no longer hold true; much of that lies elsewhere in the AI supply chain which thus needs to be brought within the scope of regulation. Governance systems for autonomous AI cannot replicate existing governance models, but need a fresh approach. Retrospective supervisory oversight becomes ineffective as a risk management tool, and AI autonomy generates new systemic risks which require new solutions. This paper investigate four regulatory systems: UK regulation of content platforms, data protection, UK financial services, and the EU AI Act’92s cross-sectoral regime. It analyses the challenges posed by autonomous and agentic AI and proposes potential solutions which regulators might adopt. These will transform regulation from a reactive process to an active one, and assist it in adapting to the challenges of AI autonomy.
[AI-21] From Static Bibliometrics to Dynamic Knowledge Graphs: An LLM -Powered Framework for Modernizing Science Technology and Innovation (STI) Analytics
链接: https://arxiv.org/abs/2607.21327
作者: Muhsen Hammoud
类目: Digital Libraries (cs.DL); Artificial Intelligence (cs.AI)
备注:
Abstract:Bibliometric indicators - citation counts, h-indexes, co-authorship networks - have long anchored science, technology, and innovation (STI) analytics, yet suffer from temporal lag, semantic shallowness, and an inability to capture the non-linear dynamics of contemporary knowledge ecosystems. Dynamic knowledge graphs and large language models (LLMs) have each been proposed as remedies, but neither is sufficient alone: existing scholarly knowledge graphs remain largely static, while LLM-driven pipelines are prone to hallucination, opacity, and corpus bias without structured grounding. This paper proposes a hybrid, symbolic-first framework integrating all three traditions under explicit methodological constraint. Organized across five layers - an open scholarly data backbone, a dynamic versioned knowledge graph, a constrained LLM-assisted semantic augmentation layer, a multi-layer validation pipeline, and an analytics layer - the framework positions LLMs strictly as generators of provisional candidate enrichments. Candidates become analytically admissible only after passing structural, evidentiary, comparative, and selective expert validation, with full provenance recorded at every stage. The analytics layer supports both established bibliometric indicators and extended graph-based analyses, including trend emergence detection, science-to-technology pathway mapping, and policy-oriented gap analysis. The framework’s central theoretical contribution is treating validation as the mediating principle between semantic flexibility and epistemic discipline, enabling STI analytics that is semantically richer and temporally more responsive than static bibliometrics while remaining aligned with the evidentiary standards of science-of-science research. Governance considerations addressing reproducibility, bias, and auditability are also discussed.
[AI-22] oward cryptographically verifiable authorization for autonomous AI agents : A security hypothesis preliminary formal model and proof-of-concept implementation
链接: https://arxiv.org/abs/2607.21325
作者: M. Llambí-Morillas,D. Fernández-Fernández
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 11 pages, 1 figure, 2 Tables. Keywords: autonomous AI agents, zero-knowledge proofs, verifiable authorization, agentic security, zk-SNARKs, access control, cryptographic authorization, cryptographic protocols, zero-trust architecture, pre-execution authorization. Submitted to ACM Transactions on AI Security and Privacy (TAISAP)
Abstract:Autonomous AI agents increasingly execute actions, invoke tools, and operate on protected resources with limited human oversight. Existing authentication and authorization mechanisms establish identity and delegate authority, but do not inherently provide cryptographic evidence that a concrete request issued by a specific agent satisfies the applicable policy in a specific execution context. This paper hypothesizes that agent authorization can be formalized as a cryptographically verifiable relation, denoted R_CVA , that jointly binds an agent principal, a concrete authorization request, an execution context, and the satisfaction of an applicable policy, while selectively preserving the confidentiality of private authorization attributes. We introduce a preliminary formal abstraction for Cryptographically Verifiable Agent Authorization (CVA), define a compact set of candidate security properties including authorization soundness, principal binding, request binding, policy binding, and replay resistance, and provide an executable zero-knowledge proof of concept that instantiates selected elements of the model over a Groth16 zk-SNARK construction. We further identify and formalize the structural separation among identity binding, authorization-request binding, and runtime execution binding as a central open problem in the design of secure agentic systems (a distinction not explicitly addressed by current agentic security frameworks) and present a falsifiable research agenda for its resolution.
[AI-23] Scaling Up Formal Representation of Clinical Trial Protocols in Ensemble Logic Using LLM s: A Preliminary Study
链接: https://arxiv.org/abs/2607.21307
作者: Yan Huang,Xubing Hao,Xiaojin Li,Rashmie Abeysinghe,Xiaoqian Jiang,Licong Cui,Guo-Qiang Zhang
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Emerging Technologies (cs.ET)
备注: Proceedings of the 2026 American Medical Informatics Association Symposium, to appear
Abstract:The reliance on unstructured free text for documenting clinical trial protocols creates a significant barrier to automated reasoning, cohort discovery, and trial simulation. The lack of formal structure obscures critical temporal phenotypes, such as dynamic eligibility criteria and event timing constraints. Although Temporal Ensemble Logic (TEL) offers an expressive framework for modeling these elements, manual encoding remains a prohibitive bottleneck. We introduce the CT-TEL workflow: a scalable pipeline leveraging Large Language Models (LLMs) to translate narrative clinical protocols into TEL formulas. We applied CT-TEL to generate logical models for 23 real-world trials from this http URL. We evaluated translation fidelity via a back-translation approach, using LLMs to convert TEL formulas back into natural language and measuring semantic similarity against source texts. The resulting semantic retention suggests that LLMs may offer a pathway for mapping informal protocols to computable logic, providing preliminary evidence toward scalable clinical trial emulation within the emerging “Symbolic Biomedicine” paradigm championed by the corresponding author.
[AI-24] Expert Behavior Prior Reinforcement Learning
链接: https://arxiv.org/abs/2607.21302
作者: Gong Gao,Weidong Zhao,Xianhui Liu,Ning Jia
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Behavior prior reinforcement learning (BPRL) has emerged as a promising paradigm to improve sample efficiency in online reinforcement learning (RL) by leveraging policy priors derived from offline demonstrations. However, most existing BPRL methods rely on static offline datasets, which often suffer from low data diversity and suboptimal trajectory quality. This reliance restricts the effectiveness of policy priors, hindering both policy exploitation and stability during online training. Consequently, agents are prone to inefficient exploration and unstable learning dynamics. To address these limitations, we deviate from existing offline pre-training methods and propose an Expert Behavior Prior (EBP) algorithm. Specifically, we introduce a Q-guided conditional variational autoencoder (Q-CVAE) that learns to generate expert policy priors directly from the online replay buffer. This enables the generation of high-value actions for guiding policy updates without relying on pre-collected expert trajectories. To further enhance policy exploitation, we propose an expert policy guidance (EPG) mechanism that selects expert actions from a generative support set, and we integrate a policy gradient correction (PGC) module to harmonize Q-guidance with expert supervision, promoting stable and consistent policy improvement. Extensive experiments conducted on robotic control (Gym, PyBullet) and industrial control (DMControl) benchmarks demonstrate that EBP significantly outperforms state-of-the-art online RL algorithms, achieving higher sample efficiency and more stable convergence.
[AI-25] An LLM -Driven Workflow for Automated Process Control Strategy Generation and Tuning from Dynamic Process Models
链接: https://arxiv.org/abs/2607.21292
作者: Ari Luna Rueda,Eike Cramer,Klaus Hellgardt,Mehmet Mercangöz
类目: Artificial Intelligence (cs.AI); Systems and Control (eess.SY)
备注:
Abstract:We present a structured large-language-model-driven workflow for automated multi-variable control design from dynamic process models. The workflow decomposes the design task into constrained code-generation steps: plant-interface construction, normalization, manipulated-variable controlled-variable (MV-CV) pairing, controller specification, closed loop simulation, scenario generation, performance evaluation and Bayesian-optimization (BO) based tuning. Generated artifacts are executed and validated before downstream tasks proceed, and failed artifacts are repaired using validation feedback. The proposed approach is demonstrated on a nonlinear gas-preheater benchmark with coupled pressure and temperature dynamics. The generated workflow produces a physically consistent decentralized PI (proportional-integral) feedback-feedforward control structure and an executable tuning environment. Bayesian optimization reduces the closed loop performance objective, which aggregates set-point tracking and disturbance-rejection errors for the controlled variables, by approximately 26.5% relative to the initial controller generated by the workflow, mainly through improved pressure-loop transient performance. This figure quantifies the automated tuning stage rather than a comparison against a manually designed controller. The results demonstrate the feasibility of using structured large-language-model-based code generation to construct executable control-design workflows, while also highlighting the need for broader validation on larger plantwide-control benchmarks.
[AI-26] Multi-Task Learning for Heterogeneous Prediction from Video Game State with Transfer Learning
链接: https://arxiv.org/abs/2607.21290
作者: Jonas Peché,Aliaksei Tsishurou,Alexander Zap,Günter Wallner
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Multi-task learning (MTL) is a promising approach for prediction tasks derived from video game state data, as modern game telemetry provides multiple related supervision signals from the same structured observations. We study whether a shared model trained jointly across tasks in team-based multiplayer games can improve generalization while reducing training and inference cost compared to specialized single-task models. We adapt a multimodal architecture for endpoint prediction to a general multi-task setting that combines rasterized vision inputs, global match context, and per-unit state information through an image encoder and attention-based interaction modeling. Experiments on a large proprietary World of Tanks dataset compare single-task and multi-task training, evaluate weighting strategies for mixed losses and conflicting gradients, and test pre-training/fine-tuning under limited target-data regimes. We also examine within-game transfer across game maps under structured environment shift.
[AI-27] BasketEvent: Understanding Who Did What and When in Basketball Videos
链接: https://arxiv.org/abs/2607.21267
作者: Yu Zhang,Jiayuan Rao,Haoning Wu,Weidi Xie
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Comprehensive basketball video understanding requires resolving not only what event occurs, but also who is responsible and when the key evidence appears. However, exist- ing methods typically treat spatial perception and semantic recognition as isolated tasks, failing to ground events to individual players or pinpoint their temporal boundaries within complex collective dynamics. To bridge this gap, we introduce BasketEvent, a player- centric basketball event understanding dataset curated from real NBA broadcasts. In BasketEvent, event labels are grounded to the responsible players, and a manually an- notated subset of 1,000 samples with precise event intervals is provided to evaluate tem- poral evidence localization. Based on this data, we propose PlayNet, a player-centric reasoning framework that maps basketball videos to player-level event predictions with temporal evidence. Concretely, PlayNet tracks key entities, associates player identities, and reasons about events by modeling player-player, player-ball, and global court inter- actions, while aggregating sparse temporal evidence via gated pooling. Extensive experi- ments demonstrate that PlayNet significantly outperforms representative video-level and crop-based baselines, proving the superiority of player-centric modeling for fine-grained sports video understanding. Our data, code, and models will be made publicly available.
[AI-28] Logic Programming Semantics for Causal Processes
链接: https://arxiv.org/abs/2607.21233
作者: Felix Weitkämper
类目: Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO); Programming Languages (cs.PL)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:Motivated by challenging modelling issues in the life sciences, we investigate the relationship between logic programming semantics and the eventual states of causal processes compatible with those logic programs. More precisely, we show that while stable models of positive logic programs correspond to the eventual states of processes commencing from a neutral state and continuing undisturbed indefinitely, supported models describe the eventual states reachable from arbitrary starting points. This also contributes to the discussion of the appropriate semantics for logic programming as a causal rule language, adding a temporal perspective to recent interpretations of the stable and supported model semantics from an explanatory viewpoint of causality.
[AI-29] ICAE-Bench: Evaluating Coding Agents as Interactive Project Builders
链接: https://arxiv.org/abs/2607.21217
作者: Zhongyuan Peng,Dan Huang,Chuyu Zhang,Caijun Xu,Changyi Xiao,Shibo Hong,David Lo,Lin Qiu,Xuezhi Cao,Jiyuan He,Yixin Cao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:The recent emergence of vibe-coding workflows is changing what coding agents are expected to do. Instead of merely completing code under fully specified instructions, agents are increasingly expected to transform incomplete product intent into working software by combining various abilities including planning, requirement clarification, tool use, debugging, and repository-level construction. Yet existing benchmarks have not fully caught up with this shift, evaluating agents on static, fully specified tasks. In this paper, we introduce ICAE-Bench, a benchmark for evaluating coding agents under interactive project-building settings. The basic idea is to start from a fuzzy product requirement, simulating the dynamic paradigm with an automated User Agent. To make this setting both realistic and evaluable, ICAE-Bench introduces three key designs. First, to avoid the ambiguity of unconstrained fuzzy requirements, each task derives ambiguity from a precise real open-source repository with executable behavior. Second, to ensure high-quality and reproducible user simulation, ICAE-Bench grounds interaction through User Agent Data, allowing the User Agent to reveal hidden constraints without inventing new requirements or leaking implementation artifacts. Third, to evaluate open-ended repositories fairly, ICAE-Bench uses standardized black-box tests together with multi-dimensional diagnostics, including functional correctness, semantic and API similarity, structural fidelity, design quality, and interaction quality. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2607.21217 [cs.AI] (or arXiv:2607.21217v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2607.21217 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-30] Explainability Framework for Policy-Aware Autonomous Agents
链接: https://arxiv.org/abs/2607.21209
作者: Heather Merhout(Miami University),Daniela Inclezan(Miami University)
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:In the field of Artificial Intelligence, an agent is a system which is able to autonomously make decisions in order to reach a desired goal. As these systems grow more prevalent in our day-to-day lives, there has been an increased need to add explainability features which can provide an account for an agent’s behavior. We therefore propose a framework that outlines how to produce comprehensible explanations for policy-aware agents, or agents which have rule-enforcing policies incorporated in their decision-making framework. This framework is designed using insights from the social sciences on how to produce good explanations. It is implemented in the Answer Set Programming language while using Python to assist with information extraction and natural-language translation. Because these agents incur penalties when violating policies, we are able to leverage these penalties to detect undesirable events in scenarios that are counterfactual to the agents’ original actions. This lends itself to creating contrastive explanations (e.g., “the agent performed this action because, had it not, undesirable event X would have occurred.”), which formulate the core component for our explainability framework. The framework is evaluated using a survey wherein human participants provide feedback on our program-generated explanations.
[AI-31] How Rules Represent Causal Knowledge: Causal Modeling with Probabilistic Logic Programming
链接: https://arxiv.org/abs/2607.21208
作者: Kilian Rueckschloss(Eberhard Karls Universitaet Tuebingen),Felix Weitkaemper(German University of Digital Science)
类目: Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO); Programming Languages (cs.PL)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:Pearl famously argues that causal knowledge enables the prediction of intervention effects. By contrast, purely descriptive knowledge supports only conclusions drawn from observations. His theory of causality, however, is developed exclusively within Bayesian networks and causal models. Consequently, it is largely restricted to acyclic causal relationships, and transferring its ideas to other formalisms risks misinterpretation or inconsistency. This paper brings Pearl’s approach to causality into probabilistic logic programming (PLP). To this end, such programs are aligned with philosophical foundations established in prior work that do not rely on temporal notions; that is, all relevant events are assumed to occur simultaneously. A formal causal semantics for these programs, together with a notion of intervention and an implementation, is proposed. It is shown that this semantics coincides with the P-log semantics for stratified ProbLog programs, while the two may differ in the non-stratified case and for other PLP formalisms. Comments: In Proceedings ICLP 2026, arXiv:2607.17707 Subjects: Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO); Programming Languages (cs.PL) Cite as: arXiv:2607.21208 [cs.AI] (or arXiv:2607.21208v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2607.21208 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Journalreference: EPTCS 450, 2026, pp. 478-491 Related DOI: https://doi.org/10.4204/EPTCS.450.35 Focus to learn more DOI(s) linking to related resources
[AI-32] A New Well-Supported Semantics for Description Logic Programs
链接: https://arxiv.org/abs/2607.21203
作者: Spencer Killen,Jia-Huai You
类目: Artificial Intelligence (cs.AI)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:Description logic programs are a powerful formalism for combining rules with ontologies. The well-supported semantics for description logic programs ensures that no answer sets rely on cyclic dependencies. Most popular semantics for logic programming have this property of well-supportedness. We recognize two limitations of the current well-supported semantics for DL programs: its increased computational complexity for the consistency problem and its lack of a reduct transformation characterization. In this work, we present a new semantics which evaluates ontological atoms more strictly than the current semantics. This keeps the complexity of its consistency problem NP-complete, rather than increasing it to the second level of the polynomial hierarchy. Additionally, we identify a syntactic class of description logic programs for which our new semantics is equivalent to the current semantics. We characterize our semantics using a fixpoint operator and a reduct-based transformation. Our new semantics is a strict subset of the current well-supported semantics, so it maintains the prior notion of well-supportedness while inducing its own stricter notion. We prefer our new notion of well-supportedness due to its similarities with logic programming.
[AI-33] Hybrid MKNF with Classical Negation in the Rule Component
链接: https://arxiv.org/abs/2607.21202
作者: Arun Raveendran Nair Sheela(Université Clermont Auvergne, LIMOS Laboratory, Thales),Christophe Rey(Université Clermont Auvergne, LIMOS, CNRS, France),Florence De Grancey(Thales)
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Symbolic Computation (cs.SC)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:Hybrid MKNF knowledge bases under the well-founded semantics integrate Description Logics with Logic Programming. However, they do not support classical negation in the rule component, limiting their ability to represent explicit negative knowledge. This limitation is particularly significant in safety-critical applications, where reasoning often requires explicit negative information rather than interpreting the absence of information as evidence of absence. To address this issue, we introduce an extension of Hybrid MKNF that supports classical negation in the rule component. We formally define the syntax and semantics of the extended language and present a general procedure for computing its well-founded model.
[AI-34] Bound-Founded Semantics for Answer Set Programming with Difference Constraints: Preliminary Report
链接: https://arxiv.org/abs/2607.21201
作者: Pedro Cabalar(University of A Corunna, Spain),Jorge Fandinno(University of Nebraska at Omaha, USA),Nicolas Rühling(University of Potsdam, Germany),Torsten Schaub(University of Potsdam, Germany and Potassco Solutions, Germany),Sebastian Schellhorn(University of Potsdam, Germany),Philipp Wanko(University of Potsdam, Germany and Potassco Solutions, Germany)
类目: Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:While the integration of linear constraints has significantly expanded the reach of Answer Set Programming (ASP), existing hybrid solvers often rely on disparate semantic underpinnings that lack a unified logical foundation. We address this gap by introducing a many-sorted variant of the Bound-founded Logic of Here-and-There (HTb), providing a versatile framework capable of characterizing equilibrium models across a wide spectrum of alternative semantics for extensions of ASP with linear constraints. We apply this framework to the setting of difference constraints, focusing on the semantic characterization of clingo[DL]. Central to our approach is the formalization of foundedness for numeric variables. By investigating how different hybrid systems - such as clingo[DL], clingcon, and flingo - justify constraint atoms, we uncover the semantic roots of their varying behaviors. This investigation results in a single, consistent framework that not only formalizes the foundations of current systems like clingo[DL] but also facilitates the rigorous study of program simplifications and the future integration of diverse semantic principles.
[AI-35] owards a Certifying Grounder
链接: https://arxiv.org/abs/2607.21199
作者: Daimy Van Caudenberg,Alexander Ek,Carlos Cantero,Bart Bogaerts
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:Grounding, the translation of high-level theories into equivalent quantifier-free formulas, is a crucial step in declarative solving, yet it has so far escaped the proof-logging revolution. When this grounding step is not certifying, there is no way of knowing that the obtained solutions actually correspond to the original problem specification, resulting in a trust gap. In this paper, we close the trust gap between the user’s high-level specification and the solver’s low-level input by introducing a novel certifying grounding framework for first-order logic model expansion (FOX) over finite domains. We present CertiFOX, a framework consisting of: (1) a proof format for grounding derivations, (2) GroundFOX, a certifying grounder operating on theories in Grounding Normal Form (GNF)–a new normal form designed for compact, domain-aware grounding–and (3) CheckFOX, an independent proof checker. Our approach guarantees that the grounder’s output is equivalent to the input specification, setting the stage for trustworthy end-to-end certified solving pipelines for declarative languages. Experimental evaluation confirms that CertiFOX is a feasible approach. The GroundFOX grounder is broadly comparable with other grounders, and proof checking with CheckFOX adds overhead within a small constant factor of grounding time. Comments: In Proceedings ICLP 2026, arXiv:2607.17707 Subjects: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI) Cite as: arXiv:2607.21199 [cs.LO] (or arXiv:2607.21199v1 [cs.LO] for this version) https://doi.org/10.48550/arXiv.2607.21199 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Journalreference: EPTCS 450, 2026, pp. 309-324 Related DOI: https://doi.org/10.4204/EPTCS.450.24 Focus to learn more DOI(s) linking to related resources
[AI-36] Declarative Problem Solving in UAM Strategic Deconfliction
链接: https://arxiv.org/abs/2607.21197
作者: Gioacchino Sterlicchio(DMMM, Polytechnic University of Bari, Bari, Italy),Angelo Oddi(ISTC-CNR, Rome, Italy),Riccardo Rasconi(ISTC-CNR, Rome, Italy),Francesca Alessandra Lisi(DIB and CILA, University of Bari Aldo Moro, Bari, Italy)
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:The growing demand for Urban Air Mobility (UAM) introduces significant challenges in airspace management, particularly within densely populated metropolitan regions. As the number of aerial vehicles-such as drones, air taxis, and helicopters-continues to rise, so does the risk of mid-air collisions and conflicts with existing air traffic and obstacles. Ensuring safe and efficient UAM operations requires robust strategic deconfliction mechanisms. We propose an Answer Set Programming (ASP) based approach for strategic deconfliction, focusing on time synchronization and route optimization for conflict-free flight plans. The solution is benchmarked against Constraint Programming (CP), emphasizing scalability and resource use. Results show that ASP offers faster execution and better scalability for small to medium cases, while CP maintains stable memory but degrades with complexity.
[AI-37] Case study: solving P-99 with LPTP and an LLM
链接: https://arxiv.org/abs/2607.21196
作者: Fred Mesnard,Thierry Marianne,Étienne Payet,Wim Vanhoof
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:Ninety-Nine Prolog Problems (P-99) is a famous set of Prolog exercises. We solved the first thirty three just by prompting an LLM (Large Language Model). We used Claude from Anthropic. By solved we mean: generate the Prolog code and a test file, run the tests and check whether they pass, then formally prove types, groundness, termination, uniqueness, existence and also sometimes functional correctness with LPTP (Logic Program Theorem Prover). Hence our approach is an experiment in vibe-coding/vericoding of P-99. It is a vibe-coding experiment because we started from informal specifications written in English and let Claude generate the Prolog code. It also fits within vericoding because the LLM proved reliability guarantees on the generated Prolog code. Claude wrote 58 logic procedures, 508 tests, 257 lemmas for a total of 11800 proof lines. We manually checked each file generated by the LLM. We checked the Prolog code, ran the tests, examined the logical statements generated by Claude and proof-checked Claude’s proofs with LPTP. This paper describes this experiment and provides the main details so that it can be reproduced by the interested reader.
[AI-38] Chess_db: A framework for working with large chess game datasets
链接: https://arxiv.org/abs/2607.21195
作者: Nicos Angelopoulos(University College amp; Imperial College, London UK),Jan Wielemaker(SWI-Prolog solutions)
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Databases (cs.DB); Symbolic Computation (cs.SC)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:Chess is a two player strategic game that is embedded in classical AI culture as it was once the frontier for intelligent behaviour. There was the silent assumption that the advent of computer engines that play better than the best humans will extinguish interest in the game. However, the opposite has come to pass, with a growing following for the game. A lot of the computational resources are now centered around training of players, where the engine output is just one aspect. Access to past games is also an essential part, both in knowing what games a specific player has played previously, and also which continuations at a certain position have led to victory more often for each of the two colour players. We present Chess_db a suite of logic programming tools that can effectively manipulate games both in memory and via creating back end databases. In particular, we provide versatile code that creates databases from PGN (portable game notation) game files and explore the suitability of open source key-value databases for storing position tables that provide near-instant access to information pertaining to substantially large number of games.
[AI-39] Animation Verification and Visualisation of Prolog Transition Systems with ProB
链接: https://arxiv.org/abs/2607.21192
作者: Jan Gruteser(Heinrich Heine University Düsseldorf),Michael Leuschel(Heinrich Heine University Düsseldorf),Katharina Engels(Heinrich Heine University Düsseldorf),Fabian Vu(Heinrich Heine University Düsseldorf)
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:ProB is a Prolog-based model checker, animator and constraint solver for high-level formal specifications. One can also use ProB to animate transition systems defined by Prolog predicates, allowing the application of its various validation techniques. In this work, we present the existing features of ProB’s Prolog animation mode and its recent extensions. The extended capabilities include simulation for statistical checks, more reliable trace replay, transitions with user input and improved state visualisation. We apply the new features to case studies, particularly for evaluating different strategies in game play, such as Connect Four. The features are useful for many other applications, especially for ProB’s new sequent prover for Event-B proof obligations, as well as for demonstration models for teaching in combination with interactive visualisation.
[AI-40] Encoding Event-B Proof Rules in Prolog: An Interactive Sequent Prover for ProB
链接: https://arxiv.org/abs/2607.21191
作者: Katharina Engels(Heinrich Heine University Düsseldorf),Jan Gruteser(Heinrich Heine University Düsseldorf),Michael Leuschel(Heinrich Heine University Düsseldorf)
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:Event-B is a formal method rooted in predicate logic and set theory. We encoded over 600 proof rules in Prolog, enabling a systematic, comprehensible proof analysis and construction. By integrating the proof rules into the Prolog-based validation tool ProB, we obtain an interactive proof system with proof tree visualisation. This has advantages in teaching, giving students direct control over the selection of proof rules. Our tool can import proof obligations from the Rodin platform and provides multiple exports: a trace file for proof replay in ProB, an interactive HTML document for tool-independent exploration of the proof tree, and an export back to Rodin, allowing the ProB prover to be used as second chain. Compared to the previous implementation of the proof rules in Java, the encoding in Prolog is more compact, maintainable and extensible. While a preliminary iterative deepening prover with simple heuristics is already available and useful for finding short proofs, we aim to obtain fast automatic provers in the future.
[AI-41] Identifying Good Rules for Efficient SAT Encodings of Single-Constant Multiplication Using Machine Learning
链接: https://arxiv.org/abs/2607.21188
作者: Chufeng Jiang(Graduate Center, The City University of New York),Neng-Fa Zhou(Graduate Center, The City University of New York)
类目: Artificial Intelligence (cs.AI)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:The Single Constant Multiplication problem is a fundamental NP-hard optimization task in hardware design, which seeks to decompose a fixed constant using only additions, subtractions, and bit-shifts. Although dynamic programming methods can produce near-optimal SAT encodings for SCM, their encoding cost remains high for large constants. We propose a neuro-symbolic framework that accelerates SCM SAT encoding by identifying good rules for guiding operator selection during decomposition. Our approach employs a graph neural network model to predict promising operator types from constant decompositions, and exploits the resulting confidence scores to prune no-good choices in the symbolic search. Experimental results on unseen 17-32 bit constants demonstrate one to two orders of magnitude reductions in encoding time, over 97% reduction in memory usage, and an order-of-magnitude decrease in branching, while preserving near-optimal encoding quality in terms of additions. These results show that learning-guided symbolic strategies can significantly improve the scalability and efficiency of SCM encoding. Our code and data are publicly available at: this https URL
[AI-42] Case study: proving sqrt(2) irrational with LPTP and an LLM
链接: https://arxiv.org/abs/2607.21187
作者: Fred Mesnard,Étienne Payet,Wim Vanhoof
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI); Symbolic Computation (cs.SC)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:We present the interactions with an LLM (Large Language Model) aiming at proving that the square root of 2 is not a rational number in an LP (Logic Programming) context. We start from a few basic pure logic programming predicate definitions. We rely on the LPTP (Logic Program Theorem Prover) system for stating and proving properties about logic programs. As the proof language of LPTP is based on natural deduction, the proofs are human readable. In our case study, we sketch in LPTP the usual proof showing the irrationality of the square root of 2. Then we describe the interactions we had with the LLM. We end up with a complete formal proof, partially generated by an LLM and fully proof-checked by LPTP.
[AI-43] Differentiable Logic Programming to Mitigate Reasoning Shortcuts in Neurosymbolic Systems
链接: https://arxiv.org/abs/2607.21185
作者: Akihiro Takemura(National Institute of Informatics, Tokyo, Japan),Katsumi Inoue(National Institute of Informatics, Tokyo, Japan)
类目: Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:Neurosymbolic (NeSy) systems integrate neural networks with logical reasoning to achieve both generalization and interpretability, but recent work has shown they are susceptible to shortcut reasoning behaviors. We propose a novel method using matrix-based differentiable logic programming to mitigate reasoning shortcuts in two phenomena: constraint satisfaction shortcuts, where constraints are satisfied without achieving the intended task, and cognition shortcuts, where biased data leads to semantically incorrect concept mappings despite logically sound inference. Building on recent matrix-based logic programming semantics, we introduce design elements to mitigate shortcuts, including a unified encoding of rules and constraints in a single matrix. We also identify connections to fuzzy logic t-norms and empirically compare their gradient flow properties. Through carefully designed experiments on MNIST variants, we show that one-to-one grounding of neural outputs to logical atoms significantly reduces both shortcut types compared to previous methods that rely on soft probability distributions. We then confirm that architectural choices in coupling symbolic knowledge with neural learning play a critical role in shortcut mitigation.
[AI-44] Explaining Weather Bulletins via ILP
链接: https://arxiv.org/abs/2607.21184
作者: Enrico Santi(University of Udine, DMIF),Alessandro Dal Palù(University of Parma, SMFI),Agostino Dovier(University of Udine, DMIF),Talissa Dreossi(University of Udine, DMIF),Andrea Formisano(University of Udine, DMIF)
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Symbolic Computation (cs.SC)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:Inductive Logic Programming (ILP) originated within the Logic Programming community in the Nineties as a framework for combining symbolic learning with declarative knowledge representation. Nowadays, mature ILP frameworks exist and they are capable of learning complex, non-monotonic hypotheses, thus broadening both the modeling capabilities and the scope of real-world applications of ILP. This work is primarily based on the FastLAS2 framework and aims to generate simple, interpretable hypotheses to help clarify the weather bulletins issued by OSMER FVG, the Regional Meteorological Observatory of the Italian region of Friuli Venezia-Giulia. In this paper we present a pipeline that, starting from simulated meteorological raw data and from OSMERs’ bulletins (used as ground truth), extracts data as ASP facts and generates ILP examples. From such examples an explanatory hypothesis is then inferred via FastLAS2. Such a hypothesis (translated into natural language) explains the weather forecast issued by human experts, and in particular the rationale behind experts’ choices of specific symbols in the bulletin pictogram (the symbol-annotated meteorological map of the forecast). The proposed approach is general, not specific to any particular region and it can equally be applied to bulletins from other sources and to different regions.
[AI-45] Representative Sets in Propositional Abduction
链接: https://arxiv.org/abs/2607.21183
作者: Johannes Schmidt(Jönköping University),Mohamed Maizia(Jönköping University, Linköping University),Victor Lagerkvist(Linköping University),Johannes K. Fichte(Linköping University)
类目: Computational Complexity (cs.CC); Artificial Intelligence (cs.AI); Data Structures and Algorithms (cs.DS); Logic in Computer Science (cs.LO)
备注: In Proceedings ICLP 2026, arXiv:2607.17707
Abstract:The propositional abduction problem is a well-known form of non-monotonic reasoning where we are asked to find an explanation of a given manifestation. Recently, there has been an influx of results asking more refined questions about the solution space rather than only individual solutions. For example, we might be interested in finding two solutions that are sufficiently far from each other (diverse solutions) in the solution space. In this paper we consider a related representation question where we ask if a given set of explanations S can represent any other explanation (that is, whether their symmetric difference is smaller than a given k). We first study this problem from a classical complexity perspective and obtain a complete classification. While only a handful of cases are tractable, the increase in complexity compared to classical abduction is often smaller than expected. We then study the parameterized complexity for several parameters and obtain new tractable and hard cases. Interestingly, a full parameterized complexity classification would require resolving the parameterized complexity of the covering radius problem from coding theory. To the best of our knowledge, no useful relationship between coding theory and non-monotonic reasoning has previously been established, but such connections seemingly become important when asking more complex questions about solution spaces.
[AI-46] Safeguards for Speech2Speech LLM -Assistants: A Case Study in Automotive Applications
链接: https://arxiv.org/abs/2607.21180
作者: Gregor Endler,Sebastian Kraus,Lukas Stappen
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Recent advances have introduced speech-to-speech (S2S) conversational assistants capable of producing natural-sounding interactions, including non-verbal cues like tonality and mood. In the automotive domain, this enables intuitive and humanlike in-car dialogue experiences. However, integrating these end-to-end assistants limits architectural options for programmable domain-specific safeguards. This paper discusses two implementation approaches for S2S guardrails: transcript-based and tool-based. Through an empirical evaluation, we demonstrate that both strategies are insufficient for industrial deployment in most cases due to prohibitive latency (delaying each answer by 0 to 1.4 seconds even for computationally cheap checks) and technical impediments (like potentially non-deterministic tool call behavior). Finally, we outline open challenges for S2S guardrails in the automotive context.
[AI-47] V-DEAL: Diagnosing Video Safety De-Calibration as an Understanding-Refusal Coupling Failure
链接: https://arxiv.org/abs/2607.21151
作者: Zhetong Zhang,Honghao Fu,Miao Xu,Yiwei Wang,Yujun Cai
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As Video Large Language Models are increasingly deployed in real-world applications, ensuring their safety alignment has become critical. Counterintuitively, we find that harmful videos paired with benign queries achieve higher attack success rates than the same videos paired with explicitly harmful queries. To understand the underlying mechanism of this vulnerability, we present V-DEAL, a three-level diagnostic framework that jointly analyzes this failure across model behaviour, understanding, and internal representations. By progressively ruling out perception failure and quantifying the model’s internal refusal tendency, V-DEAL provides a new diagnostic perspective for analyzing the underlying mechanism of the observed vulnerability. We tested six Video LLMs on three public benchmarks and observed that models correctly recognize harmful video content with over 81% accuracy, yet the average attack success rate still reaches 48.33% under the condition pairing harmful videos with benign queries. Hidden-state analysis further shows that visual understanding activates a weaker refusal tendency than textual understanding. Furthermore, we introduce a prompt injection intervention method that reduces attack success rates by an average of 48.24 percentage points and achieves performance comparable to prior fine-tuning-based methods, providing an effective and practical means to address such safety risks in Video LLMs.
[AI-48] Demographically-Informed Heat-Mortality Risk Curves via Risk Graph Neural Networks
链接: https://arxiv.org/abs/2607.21131
作者: Alex O. Davies,Eunice Lo,Rui Zhu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Estimating heat-related mortality risk is a core task in environmental epidemiology, typically addressed with Distributed Lag Non-linear Models (DLNMs); interpretable exposure-response surfaces fitted to temperature-mortality time series. DLNMs are effective but ignore demographic and geographic context, despite well-established relevance to heat vulnerability. We propose Risk Graph Neural Networks (RGNNs), a hierarchical GNN encoder that uses granular census features to optimise DLNM coefficient vectors, preserving interpretable risk curve outputs while substantially improving predictive calibration. Evaluated across 10 regions of England and Wales on two unprecedented heat years, RGNN variants maintain both lower point-errors and near-nominal uncertainty coverage during the 2022 heatwave where baselines collapse.
[AI-49] Hardware-Software Co-Design for Float16 On-Device Training on RISC-V Single-Core
链接: https://arxiv.org/abs/2607.21130
作者: Benjamin Hubinet,Pierre-Alain Moellic,Olivier Savry,Olivier Potin,Jean-Baptiste Rigaud
类目: Hardware Architecture (cs.AR); Artificial Intelligence (cs.AI)
备注: Accepted at IEEE PRIME 2026
Abstract:By leveraging standard RISC-V extensions, namely Zfh (scalar float16) and Zvfh (vector float16), this work proposes an open-source framework to enable complete on-device training on resource-constrained RISC-V single-core. Our approach allows memory footprint reduction by about 50% as compared to using float32 and with minimal model performance degradation. We also facilitate transfer learning and fine-tuning scenarios by incorporating layer-freezing capabilities. Our work builds onto AIfES, an open-source, modular and generic DNN training and inference framework for embedded systems that can be extended with custom hardware-specific functions. The benefits of float16 is further emphasized by outlining the low area overhead of Zfh on a RV64GC super-scalar out-of-order FPGA softcore (+1.15% LUT6 and +0.05% FF at 175MHz). Finally, we discuss the architecture of a Zvfh implementation within the same RISC-V core.
[AI-50] Relative Value Learning ICLR2026
链接: https://arxiv.org/abs/2607.21120
作者: Marc Höftmann,Jan Robine,Stefan Harmeling
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Published as a conference paper at ICLR 2026
Abstract:In reinforcement learning, critics typically estimate absolute state values V(s) , estimating how good a particular situation is in isolation. However, it turns out that only differences in value are relevant for control. Motivated by this, we propose Relative Value Learning (RV), a framework that learns value differences directly via an antisymmetric function \Delta(s_i, s_j) = V(s_i) - V(s_j) . We introduce a pairwise Bellman operator and prove it is a \gamma -contraction with a unique fixed point equal to the true value differences, derive well-posed 1 -step, n -step and \lambda -return targets and reconstruct generalized advantage estimation from pairwise differences to obtain an unbiased policy-gradient estimator (R-GAE). Beyond theoretical results, we integrate RV with PPO and achieve competitive performance on the Atari benchmark (49 ALE games) compared to standard PPO, indicating that relative value estimation is an effective alternative to absolute critics.
[AI-51] GlucoTune: A Unified Framework for Blood Glucose Preprocessing Forecasting and Benchmarking in Diabetes
链接: https://arxiv.org/abs/2607.21117
作者: Davide Marelli,Giorgia Rigamonti,Mirko Paolo Barbato,Paolo Napoletano
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Preprocessing blood glucose time-series data is a critical yet often overlooked step in developing data-driven methods for diabetes management, particularly for type 1 diabetes. The lack of standardized preprocessing workflows and evaluation protocols hinders reproducibility and complicates fair comparison across studies. These challenges are further exacerbated by data-sharing restrictions, as privacy and licensing constraints often prevent the redistribution of preprocessed medical datasets. To address these limitations, we present GlucoTune, a comprehensive and extensible framework for reproducible experimentation with blood glucose time-series data. The framework standardizes the entire experimental workflow, from preprocessing to model evaluation, enabling reproducible experiments directly from the original datasets. Reproducible preprocessing is achieved through configurable pipelines defined in portable YAML configuration files, ensuring consistent data handling without distributing sensitive preprocessed data. Beyond preprocessing, GlucoTune provides a unified interface for implementing, training, and evaluating blood glucose prediction models. The framework integrates public datasets through standardized wrappers and provides a curated collection of state-of-the-art blood glucose prediction and general time-series forecasting methods, while remaining readily extensible to additional datasets, preprocessing strategies, and forecasting models. To promote transparent and consistent evaluation, GlucoTune includes a benchmarking leaderboard that reports results across datasets, preprocessing configurations, and forecasting methods, enabling systematic comparison of experimental settings. We demonstrate the effectiveness of GlucoTune through comprehensive experiments and assess its usability in a user study.
[AI-52] OUR: A Trajectory-Level Unlearning Benchmark for Offline Reinforcement Learning
链接: https://arxiv.org/abs/2607.21111
作者: Chaofan Pan,Lingfei Ren,Xiangyu Jiang,Yanhua Li,Xuemei Cao,Xiangkun Wang,Hao Yu,Wei Wei,Xin Yang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Offline Reinforcement Learning (RL) agents are trained on fixed behavioral trajectories, which makes trajectory-level deletion important when selected data must be removed after training. Evaluating such deletion is difficult because a lower membership score can reflect trajectory removal, residual memorization visible to another attack, or policy collapse that destroys useful behavior. We introduce Trajectory-level memOrization and Unlearning in offline RL (TOUR), a benchmark that combines trajectory-level partitioning, matched non-member controls, retraining references, retained-performance anchors, and multi-attack privacy auditing. Across D4RL locomotion experiments and an exploratory AntMaze extension, TOUR shows that common deletion baselines have environment-dependent privacy-utility behavior. Retraining and fine-tuning often provide stronger retained-utility references than uniform GA+Refit, while TrajDeleter remains a useful comparator but is not uniformly stronger under the same audit. Reference-model, threshold, deviation, equivalence, action-error, representation-based, and query-limited attacks further show that a single likelihood-based membership score can overstate deletion quality. In the evaluated settings, conclusions about offline RL unlearning are therefore not stable under single-score auditing. They depend on matched non-member construction, retraining-relative calibration, attack family, retained utility, and explicit scope for diagnostic architecture or component-level evidence.
[AI-53] AttriMem: Attribution-Guided Process Feedback for Agent Memory Learning
链接: https://arxiv.org/abs/2607.21106
作者: Qinfeng Li,Yuntai Bao,Xinyan Yu,Hongze Chen,Wenqi Zhang,Xuhong Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Effective memory is crucial for LLM agents, yet constructing it effectively remains challenging. A memory-construction policy decides what information to extract, store, update, compress, or discard as interactions accumulate. Heuristic memory methods rely on subjective, task-specific rules, which can misalign with downstream objectives and limit cross-task adaptability. RL-based methods, by contrast, learn from task feedback but mainly use outcome- or module-level rewards. These coarse signals indicate task success but cannot identify which intermediate memory contents support the final answer, creating a fine-grained credit-assignment bottleneck. However, constructing such process feedback is prohibitively difficult because intermediate memory decisions lack unique ground-truth targets, while the appropriate credit varies with the agent’s uncertain reasoning trajectory and therefore cannot be specified in advance. We propose AttriMem, an attribution-guided process-feedback framework for learning memory-construction policies with RL. AttriMem augments the global outcome reward with local rewards derived from token-level contributions to the final answer. Experiments on long-horizon dialogue question answering show that AttriMem outperforms retrieval-based, heuristic, and RL-based baselines, generalizes across benchmarks and answer models, stabilizes RL optimization.
[AI-54] Can Generative Recommendation Reach Cold Items? A Temporal Perspective on Semantic-ID Generation
链接: https://arxiv.org/abs/2607.21101
作者: Jie Peng,Yanping Zheng,Zhewei Zhe,Bin Tong,Guan Wang,Bo Zheng
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Semantic-ID-based generative recommendation represents items as sequences of shared semantic tokens, enabling token recombination beyond isolated item IDs. However, closed-world recombination does not necessarily imply temporal open-token cold-start induction, where new items enter the item catalog with unseen atomic tokens or weakly supported SID paths. In this work, we revisit SID-based generative recommendation under an absolute-time temporal protocol that separates seen and unseen targets and diagnoses the cold item reachability at the token level. Through seen/unseen-hit analysis, coldness taxonomy, and oracle-prefix probing, we show that current SID-based models can occasionally reach future items supported by observed tokens and prefixes, but struggle with unseen atomic tokens and unsupported SID paths. We further explain this boundary by interpreting SID generation as hierarchical semantic bucketing: early tokens select coarse semantic regions, while later tokens refine item-specific paths. These findings show that SID generation is compositional but not fully open-ended, and suggest future directions in more independent SID spaces, scoring-based interfaces, and dynamic textual context.
[AI-55] Faster IndexTTS-2: Accelerating and Streaming Autoregressive Zero-Shot Text-to-Speech Synthesis on GPUs
链接: https://arxiv.org/abs/2607.21042
作者: Muyang Du,Shuang Yu,Junjie Lai
类目: Artificial Intelligence (cs.AI)
备注: 4 pages, 2 figures, 3 tables
Abstract:Autoregressive text-to-speech models achieve strong naturalness but suffer from slow inference due to sequential token generation, limiting their deployment in production applications that require low latency. IndexTTS-2 is a state-of-the-art autoregressive TTS model consisting of a GPT, a flow-matching Diffusion Transformer, and a vocoder. Despite its high synthesis quality, its inference speed barely reaches real-time without streaming or batching support. We present Faster IndexTTS-2, which accelerates all neural network components of IndexTTS-2 for production deployment on GPUs using NVIDIA TensorRT and TensorRT-LLM. Faster IndexTTS-2 also enables streaming synthesis for latency-sensitive interactive applications, and batched inference across all components to maximize GPU utilization. Experiments on the Seed-TTS benchmark for both English and Chinese demonstrate up to 5.0 \times speedup on the autoregressive GPT and 3.6 \times end-to-end, with minimal degradation in word error rate, speaker similarity, and naturalness. Our methodology provides a practical reference for efficiently accelerating similar autoregressive speech models on GPUs.
[AI-56] Reexamining zero-shot summarization: Empirical investigation of trustworthiness of LLM -summarizers
链接: https://arxiv.org/abs/2607.21010
作者: Vasudha Bhatnagar,Purnima Bindal,Vikas Kumar,Raj Kumari Bahl
类目: Artificial Intelligence (cs.AI)
备注: 28 pages, Under review in a journal
Abstract:Zero-shot summarization using Large Language Models (LLMs) has significantly advanced the abstractive summarization task by producing coherent and fluent summaries. However, underlying stochasticity of the large language models raises concerns about the stability and trustworthiness of the LLM-generated summaries. This issue has become increasingly important due to proliferation of LLM-generated summaries in educational settings, where students and researchers summarize complex academic materials in zero-shot manner. We propose a novel two-level diagnostic protocol for benchmarking LLM-summarizers based on the stability of the generated summaries. At the lower level, document-level stability analysis is performed over multiple LLM-summaries generated under controlled environment, and the stability coefficient is computed. Each generated summary is scored for semantic and factual alignment with the original document, enabling estimation of stability along more than one dimensions. At the next level, observations from a stratified sample of documents drawn from the corpus are consolidated to estimate the stability index of the LLM-summarizer, which is the proxy for its trustworthiness. Our empirical investigation of three LLM-summarizers across three genres of documents reveals statistically significant differences in the generation-level variability among LLMs across summary evaluation metrics. This study advances the LLM-summarization research by evidential recognition of the stability problem in LLM-summaries and motivates further research towards development of robust, reliable and trustworthy LLM-summarizers. Comments: 28 pages, Under review in a journal Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2607.21010 [cs.AI] (or arXiv:2607.21010v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2607.21010 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-57] Naju: A Native Discrete State-Space Model with Independent Retention and Writing for Long-Sequence Memory
链接: https://arxiv.org/abs/2607.21000
作者: Hyuk Lim,Seunghyun Yoon
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Long-sequence memory tracking places two opposing demands on a recurrent state: near-lossless retention of stored bindings over long horizons, and active overwriting of stale ones. In our diagnostic suite, the strongest efficient baselines tend to solve only one side well. Continuous-time-parameterized state-space models (SSMs) such as Mamba obtain their discrete recurrence by zero-order-hold discretization of a continuous-time system; we argue that this detour is unnecessary for memory tracking and parameterize the discrete transition directly. Naju (Native Adaptive Junction Unit) factorizes the recurrent update, schematically x_n = f_n\odot x_n-1 + i_n\odot(B_n u_n) , into an explicit discrete pole (a learned forget gate f_n ), an independent write gain i_n , and input-dependent write/read maps. Since the sigmoid pole satisfies 0f_n1 , each frozen local coordinate is Schur-stable by construction, and the full time-varying recurrence satisfies a fading-memory/BIBO bound under uniform boundedness assumptions, with no stability regularizer. We formalize the key structural limitation of coupled designs: any non-expansive complementary single-gate recurrence ties the effective retention r and write gain w through |r|+w\le 1 , so near-complete retention forces weak writing; decoupling f_n from i_n removes this constraint. Empirically, Naju is the only evaluated model that remains strong on both retention and overwriting at 4x the training length. Beyond the diagnostic suite, we evaluate Naju on WikiText-103 language modeling, Long Range Arena, and multi-query associative recall. Across these settings, Naju consistently combines strong long-range memory with competitive or superior performance, outperforming the Mamba baselines in the principal comparisons while remaining competitive with the Transformer and preserving linear-time, linear-memory scaling.
[AI-58] Workflow-Localized Mechanism Learning: Attribution-Guided Repair and Knowledge Reuse for Structured Agent Skills
链接: https://arxiv.org/abs/2607.20999
作者: Zibin Lin,Shengli Zhang,Taotao Wang,Yihan Xia,Deen Ma,Guofu Liao
类目: Artificial Intelligence (cs.AI)
备注: 8 pages, 3 figures
Abstract:Agent Skills package reusable procedural knowledge as external artifacts for frozen language-model agents, yet existing optimizers do not jointly resolve where a failure occurs in a workflow, which mechanism caused it, and how relevant knowledge from third-party Skills should be reused locally. We introduce Workflow-Localized Mechanism Learning (WML). Its Node–Mechanism Attribution identifies the failed workflow node, implicated mechanisms, and smallest valid edit target, routing single-mechanism defects to L3 resources and relational defects across mechanisms to L2 composition protocols. A six-module Workflow-Guided Skill Optimization (WGSO) loop then selects provenance- and scope-aware third-party knowledge, applies bounded patches, evaluates candidates, and stores verified outcomes in optimizer-side memory. On SpreadsheetBench, WML reaches 90.33 +/- 1.53 and 74.67 +/- 3.51 Hard Accuracy with DeepSeek and Qwen3.6-Flash, respectively; without additional optimization, the learned Skills transfer to WikiTableQuestions with 84.00 +/- 2.00 and 83.00 +/- 2.00 Denotation Accuracy. On Compiler-Supported50, WML attains both the highest hard-PASS rate and the lowest cost per successful task; compiled execution sharply reduces tokens and calls relative to a direct SkillAgent while retaining most of its successful tasks. Code and artifacts are available at this https URL.
[AI-59] GuardianAgent Bench: Where Agents Fail and How to Guard Them
链接: https://arxiv.org/abs/2607.20982
作者: Vishal Ishwar Naik,Chenyu Xu,Donna Dong,Hussein Hassan,Abhishek Pradhan,Ofer Mendelevitch,Tallat Shafat,Humayun Irshad
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:As large language model agents increasingly operate autonomously with access to tools and external environments, ensuring their safe and reliable behavior becomes critical. We present GuardianAgentBench (GABench), a benchmark of 580 scenarios across six domains evaluated on three production-ready frameworks: LangChain, LlamaIndex, and Vectara. The benchmark incorporates rigorous multi-stage validation and five adversarial attack modes. Experiments with six state-of-the-art models reveal that even the strongest configuration achieves only 74.8% overall accuracy and expose two distinct failure regimes: stronger models under-call required tools, while weaker models mis-select and over-call tools. Performance degrades monotonically with both tool-set size and sequential turn depth, with long-horizon planning proving the steeper bottleneck. Our guardrail implementation consistently outperforms system-prompt-based defenses across all models, recovering 19.9% of failures at a false positive rate of just 0.5%. These results demonstrate that execution-time structural intervention improves safety without disrupting correct agent behavior.
[AI-60] Beyond Independent Optimization: Compression MoE Routing and Quantization Interactions in Multimodal Edge Intelligence
链接: https://arxiv.org/abs/2607.20981
作者: Jay Gor,Karm Dave,Akshita Abrol,Rajesh Gupta,Sudeep Tanwar,Zhengkui Wang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Efficient multimodal inference is increasingly constrained not only by model quality or FLOP count, but also by the cost of preserving, moving, routing, caching, and quantizing multimodal representations under latency, memory, and energy constraints. This paper reviews recent advances in efficient vision-language and multimodal large language models, covering visual token compression, video token management, KV-cache optimization, Mixture-of-Experts (MoE) routing, low-bit quantization, edge deployment, and hardware-aware benchmarking. We argue that these techniques cannot be treated as independent optimizations. Visual token compression alters downstream feature distributions and MoE routing decisions, routing behavior affects expert utilization and quantization sensitivity, quantized router logits influence expert assignment, KV-cache policies determine retained multimodal evidence, and hardware constraints often transform computational savings into memory and communication bottlenecks. We organize the literature around these interactions and identify key design trade-offs, including accuracy versus token budget, static versus adaptive compression, sparse routing efficiency versus expert collapse, and low-bit inference versus modality-specific degradation. Finally, we introduce Temporal Routing Consistency as a diagnostic for video MoE models and highlight open research directions in routing-aware compression, cross-modal cache management, hardware-aware co-design, and unified benchmarking for multimodal edge intelligence.
[AI-61] Delivery Not Storag e: Cue-Anchored Working Memory as a Harness Property for Coding Agents
链接: https://arxiv.org/abs/2607.20972
作者: Swapnanil Saha
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注:
Abstract:Coding agents ship with one kind of memory: documents. Instruction files, plan artifacts, and auto-written memory directories are deliberately authored and deliberately retrieved: the agent must choose to write them and choose to read them back. Human expertise runs on a second tier that never gets written down: situationally-bound operational facts (gotchas, locations, local conventions) encoded as a side effect of the work and retrieved involuntarily when the situation cues them. We argue this second tier is the load-bearing one for long-running agents and must be a harness property, not an agent choice. We contribute: (1) a two-tier design theory grounded in the cognitive literature on memory offloading, incidental encoding, and event-based prospective memory, each mapped to an architectural requirement; (2) a cue-anchored memory model where memories carry first-class trigger conditions over a composable vocabulary (path, symbol, semantic, event, temporal), evaluated deterministically by the harness, a composition no surveyed academic or shipped system provides; (3) a controlled evaluation on a real coding task showing that voluntary memory use is near zero even with a pre-seeded store (0 memory operations in 114 turns), that deterministic injection delivered in every seeded run with zero false alarms, and that 39% of intra-session re-reads re-buy content paid for before a compaction boundary; (4) a repeated-compaction decay probe: ten facts held only in conversation vanish at the first summary and stay absent from 106 of 108 compactions, and the deprived agent greps the harness’s own session files to rebuild them, while the same facts injected from a harness-owned store arrive intact through all 138 compact-resumes as the final summary carries none. Delivery, not storage, is the product: the reliable memory channel for agents is the one the agent never has to think about.
[AI-62] From Scalars to Time Series: Rethinking Implicit Neural Representations for Time-Varying Volumetric Data IEEE-VIS2026
链接: https://arxiv.org/abs/2607.20970
作者: Weihan Zhang,Xuan Zhao,Yenwen Peng,Yuqi Chen,Jun Tao
类目: Artificial Intelligence (cs.AI)
备注: accepted by IEEE VIS 2026
Abstract:Implicit neural representations (INRs) for time-varying volumetric data are typically trained using dense sampling over spatiotemporal coordinates, where each observation corresponds to a single point in space and time. This coordinate-wise formulation requires extensive sampling during optimization, leading to high computational cost and inefficient use of temporal structure. In this work, we revisit this design choice and show that dense spatiotemporal sampling is not necessary for learning time-varying fields. Instead, we represent the data as a collection of spatially indexed time series and train INRs using sequence-level supervision over each spatial location, rather than coordinate-wise scalar samples. This reformulation eliminates the need for dense spatiotemporal sampling and instead learns each spatial location from its full temporal evolution in a structured manner. We demonstrate that this representation is compatible with a range of existing INR architectures and consistently improves reconstruction quality, while significantly reducing training cost. Furthermore, we show that this formulation can be combined with mixture-of-experts architectures, and that our MoE instantiation further improves reconstruction quality compared to both the base reformulation and existing MoE-based INR methods, providing a stronger capacity allocation under heterogeneous temporal dynamics.
[AI-63] Interaction Dynamics Modeling and Predictive Control for Safe Steerable Catheter–Tissue Interaction
链接: https://arxiv.org/abs/2607.20939
作者: Yongyan Cao
类目: ystems and Control (eess.SY); Artificial Intelligence (cs.AI); Robotics (cs.RO)
备注:
Abstract:Safe steerable catheter control is fundamentally a problem of interaction dynamics: the tip must follow a planned motion, remain compliant against moving tissue, reject friction and hysteresis, and respect a clinically meaningful never-exceed contact-force bound. We formulate catheter–tissue interaction dynamics in the scalar tip-normal coordinate of a single-segment single-tendon catheter. A partial-physics feedforward cancels only the reliable nominal bending dynamics, exposing a configuration-invariant linear interaction-dynamics model whose input gain varies through the scalar catheter inertia. A predictive optimizer then regulates this interaction state subject to hard contact-force, tendon-force, and curvature constraints. An augmented Kalman filter compresses contact, friction, and modeling error into one sensor-free disturbance state, giving nominal offset-free regulation in free space while leaving force safety to the explicit constraint. The unconstrained and disturbance-free limit recovers classical catheter impedance as a special realization of the same interaction dynamics, rather than as the main design object. In a MuJoCo distributed-compliance simulation of an eight-link tendon-driven catheter, disturbance augmentation cuts free-space approach error by 90%, and only the force-constrained predictive interaction-dynamics controller reconciles tracking with the 0.5,N bound: the unconstrained controller drives contact force to 0.60,N against a penetrating target, while the constrained one holds 0.47,N at identical tracking. These results show that offset-free motion regulation and contact-force safety are coupled interaction-dynamics objectives, and that the explicit predictive constraint resolves their tension under stiff tissue contact. The bound also holds under 0.5 ,mm, 1.2 ,Hz cardiac motion. Hardware validation is future work.
[AI-64] Clustered Edge Intelligence: Beyond Just Convergence of Edge Computing and AI
链接: https://arxiv.org/abs/2607.20937
作者: Chinmaya Kumar Dehury,Boris Sedlak,Alaa Saleh,Ilir Murturi,Lauri Loven,Satish Narayana Srirama,Praveen Kumar Donta
类目: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
备注: This is not survey or position paper. 28 Pages, 10 figures, 4 Tables, 108 references, Under review, submitted to Information Fusion
Abstract:We are moving from an information age to the age of intelligence. A decade, or possibly less than that, data will not be the gold anymore rather the derived intelligence out of the data and the information we posses from the edge of the network. Existing Edge Intelligence research focuses mainly on two directions: using AI for edge resource management and deploying lightweight AI models on edge devices. However, existing edge computing research lacks an intelligence-centric framework in which derived intelligence is treated as a first-class, independently manageable entity that can be described, discovered, observed, shared, reused, and dynamically clustered across heterogeneous edge devices and applications. To address these research gaps, we introduced Clustered Edge Intelligence, a visionary intelligence-centric approach. The aim of CEI is to make intelligence a shareable and reusable first-class entity that can be independently represented, discovered, observed, exchanged, and managed across the distributed edge-cloud continuum. We present a three layer CEI architecture and examine enabling technologies and research dimensions, including intelligence inventories, semantic knowledge representation, communication, discoverability, observability, lifecycle automation, clustering mechanisms, marketplaces, interoperability, and standardization.
[AI-65] SciExplore: Evaluating Autonomous Agents from Scientific Navigation to Information Integration ACL2026
链接: https://arxiv.org/abs/2607.20926
作者: Yinhao Tang,Youqing Fang,Yanan Sun,Wenran Liu,Weiming Zhang,Bin Liu,Kuikun Liu,Wenwei Zhang,Kai Chen
类目: Artificial Intelligence (cs.AI)
备注: 25 pages, 13 figures. Submitted to ACL 2026
Abstract:Scientific research involves complex information-seeking and reasoning workflows across heterogeneous sources. However, existing benchmarks primarily emphasize general-domain retrieval or static scientific question answering, and therefore fail to assess key capabilities required in realistic scientific research workflows. We introduce SciExplore, a benchmark designed to evaluate scientific information-seeking and reasoning capabilities of LLMs and agents. SciExplore comprises four task types covering 103 expert-curated tasks across more than ten scientific disciplines: scientific database navigation, ambiguous literature retrieval, missing reference completion, and cross-source structured knowledge synthesis, which probe progressively higher-level abilities from entity-level reasoning and document-level identification to evidence-level grounding and domain-level synthesis. We evaluate over ten state-of-the-art LLMs and autonomous agents on SciExplore, revealing substantial performance gaps with performance degrading sharply as task complexity increases and extremely low accuracy on the most challenging structured synthesis tasks. These results highlight significant limitations of current models and agents in realistic scientific information-seeking scenarios.
[AI-66] Representing Entity Importance in AI Knowledge Systems: A Dual-Signal Framework of Audience Evaluation and Structural Authority
链接: https://arxiv.org/abs/2607.20925
作者: Shen Xu
类目: Artificial Intelligence (cs.AI)
备注: 12 pages, 3 figures, 4 tables
Abstract:AI knowledge systems require representations of entity importance for retrieval, recommendation, evidence selection, and knowledge-intensive reasoning. Yet importance is often reduced to a single score derived from either human response or graph structure. Such compression may discard distinctions that matter when an AI system must choose among entities for different tasks. This study introduces an interpretable dual-signal representation in which each entity is characterized by an audience-evaluation dimension and a structural-authority dimension. The framework is evaluated using movie entities as an empirical validation domain. IMDb non-commercial datasets provide a rating-based audience ranking, Wikidata supports entity alignment, and English Wikipedia hyperlinks form the knowledge network on which PageRank estimates structural authority. Experiments on 482 entities and 13,690 directed relationships reveal a statistically significant but weak association between the two dimensions (Spearman rho = 0.2275, p 0.001). Their overlap is only 10% in the top 10 and 34% in the top 100, while entity-level divergence occurs in both directions. The results show that audience evaluation and structural authority are non-redundant signals and should not automatically be collapsed into a single scalar notion of importance. The contribution is not a new ranking algorithm or learned embedding, but a minimal knowledge-representation framework and an empirical test of its dimensional necessity. The findings support task-aware AI knowledge systems that preserve distinct importance signals before applying context-specific selection or aggregation. Comments: 12 pages, 3 figures, 4 tables Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2607.20925 [cs.AI] (or arXiv:2607.20925v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2607.20925 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-67] Scientific exploration collaboration and labor division in the large language model era
链接: https://arxiv.org/abs/2607.20923
作者: Xiang Zheng,Xi Hong,Jialin Liu,Chaoqun Ni
类目: Digital Libraries (cs.DL); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注: Main text: 21 pages, 4 figures. Supplementary materials: 25 pages, 13 figures, 4 tables
Abstract:Large language models (LLMs) have rapidly and significantly entered scientific workflows, but it remains unclear how their diffusion is associated with changes in scientists’ strategies in research directions and team building. We link PubMed Central full text with OpenAlex publication and collaboration histories for 775,323 scientists and analyze CRediT contribution statements from 137,120 multi-author papers. After 2022, scientists increasingly published across more intellectually distant fields and entered fields in which they had not previously worked. These increases in interdisciplinarity and exploration were especially pronounced among established scientists and scientists from non-English-speaking low- and middle-income countries. Authors with stronger AI-writing signals were already more interdisciplinary and exploratory before the widespread adoption of LLMs, and the gap widened further after 2022 compared with authors with weaker AI-writing signals. Scientists’ collaboration networks also became more interdisciplinary after 2022. Yet, among authors with stronger AI-writing signals, research interdisciplinarity was less closely tied to the disciplinary diversity of their collaborators. The division of labor within research teams also became more differentiated. Contributors on papers published after 2022 reported narrower role sets on average, coauthors shared fewer roles in common, and their role profiles became less rigid and more fluid. Software and validation roles increased, while conceptual and management roles decreased. These patterns suggest that team members are taking on more distinct responsibilities and may rely less on one another to perform research tasks. Overall, this study indicates that the LLM era coincides with a broader reorganization of scientific exploration, collaboration, and the division of labor.
[AI-68] OPOD: On-Policy Omni Distillation
链接: https://arxiv.org/abs/2607.20918
作者: Tong Zhao,Yuyang Hu,Reed Li,Yu Lu,Haibo Shi,Yutao Zhu,Zhicheng Dou
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Omni-modal models can handle text, images, and audio in one system, but improving all of these abilities together remains difficult. Training a single model on pooled multimodal data often fails to match models specialized for individual modalities. On-policy distillation (OPD) offers a way to combine such specialists: the student generates a response, and a teacher evaluates that same response, so the student learns directly from behaviors it actually produces. Yet using several teachers can introduce competing guidance and improve one modality at the expense of another. We present On-Policy Omni Distillation (OPOD), which routes each student response to the matching text, image, or audio teacher. OPOD keeps teacher guidance only on tokens where the teacher assigns a higher probability than the student, adjusts the influence of each modality teacher independently during training, and asks the routed teacher to assess both the final answer and whether the reasoning supports the correct answer. Across twelve benchmarks and three backbone sizes, OPOD achieves the best average score at every scale, reaching 70.8, 51.7, and 46.2 and exceeding the strongest comparator by 2.1, 1.8, and 1.7 points. On the 30B model, it outperforms both the base model and a counterpart post-trained jointly on pooled multimodal data on all twelve benchmarks, and ranks first or second on eleven even when the individual specialists are included. The specialists are discarded after training, leaving one deployable omni-modal model. These results show that coordinating modality-specific teachers is an effective way to improve a shared model while maintaining cross-modal balance.
[AI-69] raceable Scholarship: Page Anchors and Ariadnes Thread for Humanistic Inquiry in the Age of Generative AI
链接: https://arxiv.org/abs/2607.20916
作者: Deyu Jing
类目: Artificial Intelligence (cs.AI); Digital Libraries (cs.DL)
备注: 33 pages, 8 tables. This paper proposes a normative and infrastructural framework for traceable, AI-assisted humanistic research and presents an auditable Kant case study
Abstract:Generative AI lets large language models produce scholarly-looking text within seconds, yet fluency does not equal valid explanation. The deepest risk is not factual error alone but the appearance that an explanation is already established without clear sources, page numbers, editions, or evidence. We liken the page anchor to Ariadne’s thread: within the labyrinth of generative fluency, it is the thread that leads the scholar back to the source. This paper proposes Traceable Scholarship as the minimum normative condition for AI-assisted humanistic research, situating it across the three revolutions of knowledge infrastructure: print, digital, and generative AI. We introduce page anchors, dual page numbers, citation-first generation, NO_EVIDENCE, human verification, four-level compliance, and Scope Contract, and present AIH-Infra as a three-layer reference implementation: Contexture (document structuring), Open WebUI AIH-Infra (traceable knowledge base), and AIH-Infra MCP Server (agent gateway). A case study on a 29-volume Kant Akademie-Ausgabe knowledge base illustrates how traceability supports retrieval correction, evidence grading, and judgment downgrading. Traceability is not a software feature; it is the condition under which humanistic research can remain public and refutable in the age of generative AI.
[AI-70] Source-Prior-Driven Selective Adaptation for Efficient Diffusion Model Finetuning
链接: https://arxiv.org/abs/2607.20913
作者: Yi Xiong,Yuan-Yuan Cheng,Xiao-Ming Fu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Fine-tuning large diffusion models for new domains or styles involves a trade-off: improving target-specific generation often degrades the pretrained model’s broad generative capability. Existing full and parameter-efficient fine-tuning methods typically handle this trade-off only implicitly. In this work, we propose a novel source-prior-driven selective adaptation method to efficiently fine-tune diffusion models, achieving a favorable trade-off. Our method relies on two key observations: (1) the loss of general generative capability is highly inconsistent across pretrained parameters, and (2) parameters that have a relatively small impact on the model’s general generative capability remain structurally inconsistent across layers and parameter types. Motivated by these observations, we first learn a static mask to explicitly identify parameters better suited for downstream adaptation, and then construct structured update strategies for the selected subset. Experiments show that our method achieves a better adaptation-retention trade-off than existing strong baselines.
[AI-71] Multi-turn RL with Structural and Performance Aware Rewards for CUDA Kernel Generation
链接: https://arxiv.org/abs/2607.20908
作者: Quazi Ishtiaque Mahmud,Nesreen K. Ahmed,Ali Jannesari
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a powerful technique to enhance the reasoning capacity of LLMs for optimized code generation. However, existing RLVR approaches primarily rely on outcome-based signals such as correctness and speedup, overlooking performance-critical structural properties of programs that are essential for generating optimized code. In this work, we propose CudaPerf, a reflective RL framework that incorporates both verifiable execution rewards and structural code-aware rewards derived from parallelization features (e.g., memory coalescing, occupancy, Arithmatic Intensity, and synchronization patterns). CudaPerf operates in two stages: (1) an offline pairwise ranking module that learns to distinguish strong and weak program candidates via contrastive comparisons, and (2) an online RL training phase that jointly optimizes for correctness, performance, and structural efficiency through a unified reward signal. To further enhance learning, CudaPerf utilizes iterative refinement using execution feedback enabling progressive improvement of generated candidates. We also introduce a dataset comprising 2.9k C to CUDA and 1k PyTorch to CUDA programs, each paired with diverse input configurations and multiple CUDA implementations encompassing diverse optimization strategies. CudaPerf is evaluated across multiple benchmarks comprising both C to CUDA and PyTorch to CUDA transformations. Empirical findings suggest that CudaPerf significantly outperforms strong baselines, including Qwen-3-32B (for C to CUDA) and CUDA Agent (for PyTorch to CUDA) by achieving up to 5X 3.32X improvements in speedup, and 17% 7% improvements in correctness, respectively.
[AI-72] Anti-Goal Reasoning : Rethinking the Theory of Goal Reasoning in Non-Axiomatic Logic
链接: https://arxiv.org/abs/2607.20902
作者: Bowen Xu
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI)
备注:
Abstract:Goal reasoning in Non-Axiomatic Logic (NAL) explains how an adaptive system derives means for realizing desired events under insufficient knowledge and resources. However, the representation of avoidance is less clear. A common convention is to express avoid G '' as the goal sentence \neg G! ‘’, but this notation conflates two different readings: pursuing the negated event \neg G , and avoiding the positive event G . This paper shows that the conflation can produce a paradoxical case in which an avoidance intention is converted into a positive goal to act merely because acting is usually followed by the absence of hurt. Starting from NAL’s basic definition of goals, the framework is extended with a corresponding definition of anti-goals, so that avoidance can be represented without treating it as the pursuit of a negated event. Finally, a mental operation, \opprevent , is introduced to connect anti-goal reasoning with ordinary goal reasoning in cases of active prevention. Four minimal case studies check that the resulting rules distinguish pursuit, passive avoidance, active prevention, and withholding action to preserve a desired event.
[AI-73] Is Deep Research Reliable? Misleading Knowledge Induces False Conclusions
链接: https://arxiv.org/abs/2607.20891
作者: Pengyu Zhu,Lijun Li,Longju Yang,Sen Su
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Deep Research agents extend LLM-based assistants into long-horizon workflows involving planning, retrieval, evidence synthesis, and report generation, yet their reliability in open information environments remains underexplored. A key concern is whether apparently credible but factually misleading knowledge encountered in such environments can propagate through these workflows and be adopted as false conclusions in final reports. To study this failure mode, we introduce MisKnow-Agent, a framework for constructing and validating misleading knowledge for Deep Research tasks. MisKnow-Agent generates misleading instances with controllable authority levels and styles, yielding 5,933 quality-controlled instances built on DeepResearch Benchmark tasks. Extensive experiments across open-source and closed-source Deep Research agents show that even limited exposure to misleading knowledge can induce false-conclusion adoption in final reports, revealing a broad reliability vulnerability in current Deep Research agents. Although search-enabled verifier models consistently identify the retained instances as misleading during focused corpus validation, the same instances can still be adopted during long-horizon research, revealing a disconnect between focused verification and workflow-level evidence use. Finally, we evaluate pre- and post-research defenses, both individually and in combination, finding that all three configurations mitigate but do not fully prevent false-conclusion adoption. Our findings suggest that reliable Deep Research requires evidence verification and correction capabilities at both the model and framework levels, beyond improvements in planning, retrieval, evidence integration, or report-generation abilities.
[AI-74] wistedMerge: Certified Higher-Order Diagnostics and Abstention for Model Merging
链接: https://arxiv.org/abs/2607.20887
作者: Ting Gong,Shitan Xu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Algebraic Geometry (math.AG)
备注: 34 Pages, Comments welcome!
Abstract:Model merging combines independently trained or fine-tuned models, but pairwise alignability does not imply globally consistent alignment. We formulate merging as a finite descent problem in which checkpoints are local objects, alignment maps are transitions, and cycle products are residuals. TwistedMerge is a conservative certification pipeline that separates fixed-chart averaging, synchronization-removable gauge inconsistency, a certified central obstruction on a specified comparison complex, and nonabelian holonomy. A residual is promoted to a cohomology class only after inverse-consistency, coefficient-identification, centrality, and closure tests; otherwise the method abstains and returns an ordinary or synchronized fallback. We prove a constant-edge no-go result, frozen-complex three-way and predeclared-family error-control theorems, and a refinement test for comparison-complex sensitivity. A planted neural alignment defect is removed by cycle-consistent synchronization, showing that a nonzero cycle score alone is not a higher obstruction. Controlled central systems recover the predicted non-coboundary and projective-rank behavior, while noisy estimates move from certification to abstention without false lifts on the tested controls. A trained low-rank-adapter audit shows that naive factor averaging depends on the chosen GLr representative, whereas global factor synchronization and dense-delta SVD are stable. On natural checkpoint collections, cycle residuals do not predict merge degradation and no natural central or period-index class is certified. The results position descent theory as a falsifiable certification and abstention framework.
[AI-75] Multilevel Graph Wavelet Compressed Sensing with Scale-Aware Neural Recovery
链接: https://arxiv.org/abs/2607.20857
作者: Amirhossein Nouranizadeh,Sarang Rajendra Patil,Alan John Varghese,Varsha Narayanan,Amit Chakraborty,Mengjia Xu
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Scientific machine learning methods such as neural operators and physics-informed neural networks have advanced engineering applications and inverse problems, but their training typically requires large volumes of simulated data. This makes data preparation and model training expensive. We propose Graph Wavelet Compressed Sensing (GWCS), a learning-based framework for offline compression of graph signals by representing them as sparse, interpretable wavelet-domain representations using the spectral graph wavelet transform. The framework combines a nonparametric multilevel importance sampler, which retains high-energy wavelet coefficients within each scale for a given compression ratio, with a scale-aware graph neural network that reconstructs the signal from the sparse coefficients. We evaluate the proposed framework on synthetic approximately band-limited graph signals over random graphs and four PDE simulation datasets over meshes, which include Turbulent Radiative Layer, Viscoelastic Instability, Kolmogorov Flow, and Dynamic Stall. We compare against graph signal sampling methods and graph autoencoder baselines. Results demonstrate that the framework achieves high reconstruction fidelity and substantial data compression compared to existing benchmarks.
[AI-76] Code Monitor Red Teaming for Public-Test-Passing Code
链接: https://arxiv.org/abs/2607.20852
作者: Junchi Liao,Jiawen Deng,Fuji Ren
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Visible tests are a common gate for LLM-generated code, but passing them does not certify specification correctness. We study a deployment-like monitoring problem: after code has passed public tests, can a weaker LLM verifier identify the residual hidden bugs? We introduce Code Monitor Red Teaming, a monitor-red-teaming protocol that fixes a public-check information boundary while varying generator pressure, verifier scaffolding, and weak-to-strong capability. We instantiate it as CodeMonitorBench, spanning function-level, data-science, and workflow code. Across 71,000 generated candidates, 43,677 pass public tests and 23,081 of those fail hidden tests. Weak verifiers improve with scaffolding and model family, but still miss most hidden bugs at 5% false-positive rate. As a robustness stress test, adversarial public-test-overfit pressure lowers verifier AUROC and raises low-FPR miss rates in most cells. A GLM-5.1 verifier recovers part of the gap under the same evidence boundary; an inferability audit shows that remaining misses mix verifier failures with M1 evidence limits.
[AI-77] Auditing Evidence Use in Medical LLM Diagnosis
链接: https://arxiv.org/abs/2607.20848
作者: Junchi Liao,Jiawen Deng,Fuji Ren
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Medical LLMs are often evaluated by whether they select the correct diagnosis, but diagnostic accuracy alone does not show whether the model used the case evidence appropriately. We present a behavioral audit of evidence use in medical diagnosis. For each case, we decompose patient information into evidence units, score candidate diagnoses under controlled evidence subsets, and mine low-order interactions in diagnostic margins. Because medical evidence is diagnosis-relative, the audit separates interaction discovery from failure assignment: large or negative interactions can reflect plausible differential diagnosis, while suspicious interactions require robustness checks and clinical review. We evaluate five open-weight LLMs on DDXPlus, CupCase, and MedCase. Across datasets, faithful support and differential conflict or cancellation account for most interaction strength, showing that many evidence interactions are clinically plausible rather than failures. In a DDXPlus-focused blinded five-reviewer 130-item enriched review sample, invalid or shortcut-like cases concentrate in negated or absent findings and clinically local evidence. These results show that accuracy can hide candidate evidence-use failures and motivate role-aware audits for medical LLM evaluation.
[AI-78] Auditing Provenance Sensitivity in LLM Agent Action Selection
链接: https://arxiv.org/abs/2607.20827
作者: Junchi Liao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:LLM agents choose tools and arguments from context that mixes user requests, tool outputs, retrieved records, memory, and untrusted text. Evidence can be relevant without being authorized to determine a decision, so a correct action need not be grounded only in permitted evidence. We introduce a target-specific authorization audit that labels context factors separately for each tool and argument target. Its primary test holds the task, proposition, position, and policy fixed while changing only the proposition’s source authority. We then test behavior when valid evidence is weakened and use context-subset interactions as a secondary localization diagnostic. Across 450 controlled next-action tasks and multiple open-weight LLM families, trusted and untrusted variants produce different actions in 5.4 percent of competing cases versus 1.7 percent of supporting cases. Under controlled degradation, unauthorized competition is retained in a full-correct, mixed-error, clean-correct pattern in 2.4 percent of comparisons, with a 95 percent confidence interval from 2.1 to 3.0 percent. These are controlled stress-set rates, not deployment prevalence. The models respond to textual source-authority cues, but this does not prevent untrusted evidence from influencing their actions.
[AI-79] Efficient and Interpretable Body-Based Emotion Recognition with Lightweight Temporal Convolutional Networks
链接: https://arxiv.org/abs/2607.20820
作者: Christian Arzate Cruz,Stefanos Gkikas,Houshyar Asadi
类目: Artificial Intelligence (cs.AI)
备注: Accepted at the 14th International Conference on Affective Computing and Intelligent Interaction (ACII 2026)
Abstract:Body-based emotion recognition is important for real-time affective systems, but graph-based skeleton models can be computationally expensive. This paper studies whether lightweight temporal convolutional networks (TCNs) can provide an efficient and interpretable alternative for body-based emotion classification. We evaluate a family of TCN models on DIEM-A and compare them with a graph-based time-series graph (G-TSG) baseline using accuracy, macro-F1, parameter count, and inference latency. Although G-TSG achieves the highest mean performance, TCN-Base remains within 1.58 accuracy points and 1.25 macro-F1 points while using 79.18% fewer parameters and reducing classifier latency by approximately 12.5\times . We also analyze body-region contributions using region-specific TCN models, zero-based occlusion, and G-TSG gradient saliency. The results show that upper-body motion provides the strongest standalone regional cue, that the usefulness of body regions varies across emotions, and that different interpretability methods capture distinct aspects of model behavior. These findings suggest that lightweight TCNs can support efficient body-based emotion recognition while also providing practical insight into how motion cues contribute to classification.
[AI-80] Enhancing Explainable Cardiac Diagnosis with Guide-Grounded Multimodal LLM s
链接: https://arxiv.org/abs/2607.20814
作者: Hai-Nam Duy Vuong,Duy-Anh Bui,Trong-Nghia Nguyen,Kim-Ngan Thi Nguyen,Trang Mai Xuan,Tien-Cuong Nguyen,Van-Dem Pham,Thien Van Luong
类目: Artificial Intelligence (cs.AI)
备注: 12 pages, 3 figures, accepted at CITA 2026
Abstract:The electrocardiogram (ECG) is a cornerstone of cardiac as- sessment, yet clinical deployment of deep learning models remains con- strained by limited interpretability and the hallucination risk of large language models (LLMs). Existing CNN+Grad-CAM+multimodal LLM frameworks can generate ECG reports, but their explanations are often only weakly grounded in established diagnostic criteria, reducing trust- worthiness and reproducibility. We propose a guide-grounded multimodal framework that explicitly anchors report generation in curated clinical knowledge. A convolutional neural network (CNN) and Grad-CAM first produce class probabilities and class-specific heatmaps from 12-lead ECG images. In parallel, authoritative ECG textbooks and guideline materials are distilled offline into a structured ECG Interpretation Guide, which is injected as a fixed knowledge block for every sample. Conditioned on the ECG image, Grad-CAM overlay, CNN-derived fact pack, and the in- jected guide, a multimodal LLM generates structured diagnostic reports with guideline-consistent terminology and criteria usage. Experiments on the full PTB-XL test set demonstrate that guide grounding improves se- mantic quality and perceived consistency of generated reports while pre- serving competitive classification performance. In particular, our method increases the average BERTScore of generated impressions from 0.818 to 0.953 relative to a strong CNN+Grad-CAM+MLLM baseline, indicat- ing closer alignment with reference reports. These findings suggest that injecting a distilled interpretation guide into the multimodal prompting pipeline offers a practical pathway to reduce hallucinations and enhance the clinical plausibility of LLM-based ECG explanations, bringing ex- plainable cardiac diagnosis closer to real-world deployment.
[AI-81] Profiling Lightweight Large Language Models
链接: https://arxiv.org/abs/2607.20806
作者: Tomohiro Harada,Enrique Alba,Gabriel Luque
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Lightweight large language models (LLMs) are increasingly being deployed locally on personal computers and are expected to play a growing role in resource-constrained edge and mobile environments. In such settings, energy consumption, execution time, and memory usage directly affect practical usability, yet existing evaluations of LLM efficiency largely rely on proxy descriptors such as parameter count or FLOPs, often decoupled from task precision. This paper introduces a PTME-based experimental framework for the precision-aware profiling of lightweight LLM inference, jointly measuring Precision, execution Time, peak Memory usage, and Energy consumption through direct hardware-level measurements. The methodology is applied to a representative set of lightweight LLMs executed locally under edge-class resource envelopes on a controlled desktop platform, using benchmarks spanning code generation, mathematical reasoning, and multi-task understanding. We find that static proxy descriptors approximate inference cost well but fail to predict precision. Tightening the resource envelope increases cost without affecting precision, amplifying execution time more strongly than energy and penalizing larger models the most. Moreover, no single model dominates across all PTME dimensions, and a Pareto analysis reveals non-dominated configurations that would be hidden by accuracy-only or efficiency-only assessments, providing practical guidance for selecting models under different resource envelopes. These results show that selecting lightweight LLMs by size, FLOPs, latency, or accuracy alone can select the wrong deployment candidate; PTME profiling exposes configurations that preserve useful accuracy at lower physical cost.
[AI-82] Synthetic minority data is redundant or invalid: a data-dependent validity theory and a de-biased test
链接: https://arxiv.org/abs/2607.20787
作者: Ahmad B. Hassanat,Ahmad S. Tarawneh,Ghada A. Altarawneh
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 45 pages, 7 figures; Supplementary Information included as appendix
Abstract:For two decades, the standard remedy for class-imbalanced learning has been to fabricate synthetic minority examples, and the standard evidence of their validity has been a check that cannot fail: synthetic points are scored against the very data that generated them. We de-bias the check. Validity becomes a population quantity – the probability that a synthetic point truly belongs to the minority class – with a consistent estimator that scores synthetic points against withheld real data. Where held-out ground truth is available, the classical test underestimates true invalidity in 96-99% of method-by-imbalance-ratio cells, while the de-biased estimator tracks it closely. We prove validity is a property of the data, not the method: class overlap sets an invalidity floor no faithful generator escapes, making oversampling redundant where classes separate and invalid where they overlap. Across 91 methods, three classifiers, and datasets spanning medicine and finance – including a generator engineered to pass the classical check – none clears both bars: gains over the best trivial baseline are noise-thin (median below 0.01 F1, a decision threshold’s reach), and most damage calibration. We release the audit as a pip-installable test and flip the burden of proof: synthetic minority data must now demonstrate, on the data at hand, both validity and information gain.
[AI-83] Robostral Navigate
链接: https://arxiv.org/abs/2607.20785
作者: Arjun Majumdar,Avinash Sooriyarachchi,Benjamin Tibi,Chris Bamford,Elliot Chane-Sane,Guillaume Lample,Khyathi Raghavi Chandu,Ludovic Ho Fuh,Mathieu Poiree,Olivier Duchenne,Rosalie Millner,Srijan Mishra,Theo Cachet,Thomas Chabal
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI)
备注:
Abstract:Deploying navigation systems at scale requires a recipe that minimizes sensor assumptions, generalizes across robot embodiments, and trains efficiently. Yet, today’s best systems depend on depth sensors, multi-camera rigs, or pre-built maps, limiting the hardware they support and increasing deployment cost. We introduce Robostral Navigate, an 8B vision-language model built around this scalability objective. The model consumes only a stream of monocular RGB images - the most ubiquitous sensor across robotic platforms and predicts waypoints by pointing to the next target location in the current camera view. Operating purely in image space, rather than robot-specific coordinates, makes the policy naturally robust to changes in camera intrinsics and scene scale, enabling deployment across wheeled, legged, and aerial robots without recalibration. We generate 2.4 million trajectories across 350k simulated scenes to reduce the reliance on real-world data collection and scale easily. We further introduce a prefix-caching training recipe that packs entire episodes into single training sequences, reducing training tokens by 22x and cutting training time from months to days. A tree-based attention mask prevents conditioning on previous ground-truth actions, encouraging visually grounded action prediction, and reinforcement learning is used to further improve exploration and recovery capabilities. On the Room-to-Room and Room-Across-Room in Continuous Environments (R2R-CE and RxR-CE) benchmarks, Robostral Navigate sets a new state of the art. On R2R-CE, it achieves a 77.4% success rate, surpassing the best monocular method by 10.5 points and the strongest depth- or multi-camera system by 5.3 points despite using only a single RGB camera. On RxR-CE, it reaches 75.1% success rate, outperforming all monocular baselines.
[AI-84] he Human-AI Substitution Principle: When will you be replaced by AI in your organization?
链接: https://arxiv.org/abs/2607.20781
作者: Bonny Banerjee,Shreya Singh
类目: Artificial Intelligence (cs.AI); General Economics (econ.GN)
备注:
Abstract:Artificial Intelligence (AI) is rapidly transforming organizations, raising a fundamental organizational and economic question: when will a human employee be replaced by AI? We present an analytical model for studying Human–AI Task Allocation (HAT) in hierarchical organizations. A central feature of the HAT model is that it formally encodes the economic asymmetry between human skill acquisition and AI capability scaling. The HAT model allows us to derive how risk-adjusted costs, skills, organizational depth, deployment scale, strategic adaptation, and risk jointly determine when, where, why, and under what structural conditions human–AI replacement occurs. A key result is the Human–AI Substitution Principle, which provides a precise condition — grounded in the formal asymmetry assumption — under which AI replaces human labor. Building on this result, we show that AI adoption can produce abrupt workforce transitions, hybrid human–AI organizations, including cases where risk heterogeneity sustains human and AI roles without requiring a minimum-human-fraction constraint, and flatter managerial hierarchies with wider spans of control. The HAT model identifies structural conditions under which middle-management roles exhibit elevated vulnerability to automation, and shows that the vulnerability of highly skilled workers depends on a skill threshold shaped by organizational depth, baseline costs, and risk differentials. More broadly, the paper connects automation economics, organizational design, AI governance, and workforce planning into a unified theory of AI-driven organizational transformation.
[AI-85] Emergent Compositional Skills in Mixture-of-Experts VLAs ICML2026
链接: https://arxiv.org/abs/2607.20771
作者: Shlok Shah,Rhiaan Jhaveri,Tharun Kumar Tiruppali Kalidoss,Chirayu Nimonkar,Ishaan Javali
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: Accepted to the 2nd Workshop on Compositional Learning at ICML 2026
Abstract:We consider the problem of learning compositional robot policies end-to-end from expert demonstrations, without any pre-specified notion of task decomposition or hierarchy. We ask whether a VLA trained with a simplified Mixture-of-Experts (MoE) action head can emergently learn to decompose tasks into reusable, interpretable primitives. We find that learned experts are heavily reused across tasks and consistently correspond to qualitatively distinct low-level behaviors, suggesting that the router implicitly learns to perform high-level sequencing while experts serve as compositional primitives. Our MoE matches the task performance of a monolithic baseline while demonstrating meaningful expert specialization, a step toward modular, interpretable robot policies that emerge from data alone.
[AI-86] ArbiGraph: Arbitrarily Scalable Verifiable Task Graphs for Evaluating Context Management
链接: https://arxiv.org/abs/2607.20764
作者: Pavel Golikov,Evgenii Opryshko,Gennady Pekhimenko,Mark C. Jeffrey
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce ARBIGRAPH, a benchmark generator for evaluating whether tool-assisted language agents can retain, update, compose, and discard task-relevant context across extended reasoning workflows. ARBIGRAPH represents each task as a natural-language problem with an executable Python solver, and composes tasks through typed intermediate states, instantiated here as scalar and list values. This design enables controllable task graphs whose length, dependency structure, distractor count, and value type can be varied while preserving exact automatic verification. We instantiate ARBIGRAPH with math, GSM-style word-problems, and Python-tracing task categories, and evaluate a Qwen3.5-27B tool-assisted agent across four topologies. The results show high accuracy on isolated tasks but substantial degradation on more complex dependent tasks: accuracy drops by up to 33.3% on branching chains of dependent math tasks. This shows that ARBIGRAPH exposes failures that are not visible from single-task evaluation alone. Our code, generated datasets, and evaluation results are available at this https URL
[AI-87] IssueTrojanBench: Benchmarking AI Coding Agents Against Malicious Issue Requests
链接: https://arxiv.org/abs/2607.20759
作者: Ankur Singh,Jinqiu Yang,Tse-Hsun Chen
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: 10 pages, 4 figures, 4 tables
Abstract:AI coding agents powered by LLMs are increasingly integrated into real-world software development, where they generate, edit, and execute code with autonomous access to local files and tools. Coding agents inherit security risks from both the LLM backbone, where adversarial prompts, poisoned training data, and backdoor triggers can cause models to emit insecure or attacker-chosen code, and their agentic architecture, where tool-using autonomy enables induced misuse of external APIs, data exfiltration, and persistent compromise of development environments. This paper presents a systematic evaluation of malicious issue requests against state-of-the-art coding agents (Cursor, Claude Code, and Codex Desktop), powered by two major model families (OpenAI GPT-5.3 Codex/GPT-5.4 and Anthropic Sonnet 4.6). Our novel benchmark IssueTrojanBench contains malicious issues that are constructed based on four novel attack categories (i.e., embedded as malicious instructions in issues), six delivery vectors (e.g., PDF, or issue comment), and further augmented by perturbations. Our results reveal critical vulnerabilities in the as-deployed modern coding agents, i.e., 66.5% of the malicious issues from IssueTrojanBench penetrate all the guardrails (agent- and LLM-level) of coding agents. Our further analysis shows that rejection is almost entirely from LLMs rather than the agent frameworks, with GPT models broadly vulnerable and Sonnet 4.6 exhibiting more selective, risk-aware blocking of high-impact actions. Our evaluation also highlights that the current agent-level defense strategy offers limited additional protection for coding agents. Our findings highlight the urgent need for stronger agent- and model-level safety mechanisms to protect AI coding agents.
[AI-88] Self-Supervised Bio-Inspired Robotic Trajectory Planning with Obstacle Avoidance ICANN
链接: https://arxiv.org/abs/2607.20743
作者: Miroslav Krupa,Miroslav Cibula,Kristína Malinovská
类目: Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 12 pages, 3 figures. To be published in 2026 International Conference on Artificial Neural Networks (ICANN) proceedings. This research was supported by the Slovak Research and Development Agency, project APVV-21-0105
Abstract:Trajectory planning is a fundamental problem in robotics, requiring the generation of collision-free and efficient trajectories in a potentially complex environment. While sampling-based planners remain the dominant approach, they are often computationally expensive, particularly in high-dimensional spaces and obstacle-rich environments. Methods based on model learning offer a promising alternative, enabling efficient planning through a bounded number of forward passes through a neural trajectory planner, but commonly suffer from low sample efficiency or limited generalisation due to their reliance on exploration or expert demonstrations. This follow-up work tests our neuro-inspired self-supervised learning framework for trajectory planning that leverages forward and inverse models as the internal supervisory mechanism in an environment that contains an obstacle. Experimental results demonstrate the feasibility of the approach while revealing a tendency of our planner to exploit the learning signal provided by the forward and inverse models. To address this issue, additional training regimes and mitigation strategies are proposed and evaluated.
[AI-89] GPE: Evaluating Robust Evidence Aggregation for Fact Verification under Controllable GEO-Style Poisoning
链接: https://arxiv.org/abs/2607.20730
作者: Zhaoqi Wang,Zijian Zhang,Xiaomei Yuan,Pengtao Kou,Jiamou Liu,Zhen Li,Liehuang Zhu
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models increasingly use search tools to retrieve up-to-date information, introducing a new attack surface in which retrieved documents can be manipulated. This risk is amplified by the development of generative engine optimization, which can make selected content more likely to be retrieved, cited, and adopted by models. Existing fact-verification benchmarks and evaluation frameworks do not provide the controlled evidence environments needed to assess robustness against GEO poisoning. We therefore propose GPE, which consists of a multi-domain fact-verification benchmark and an evaluation framework for controlling evidence sources and poisoning ratios. Experiments across multiple verification methods and poisoning attacks demonstrate that GPE exposes robustness degradation and efficiency trade-offs that cannot be observed through clean evaluation alone, confirming the need to evaluate fact verification under adversarial evidence environments.
[AI-90] Operational Identity: A Finite Audit of Declared and Implemented Rules of Sameness
链接: https://arxiv.org/abs/2607.20729
作者: Denise M. Case
类目: Logic in Computer Science (cs.LO); Artificial Intelligence (cs.AI)
备注: 45 pages
Abstract:A record system declares when two records refer to the same entity, occurrence, scope, or rule. Its disclosed implementation mechanisms induce a corresponding operational identity relation. The declared and implemented relations may diverge systematically without producing a provenance gap or detectable contradiction. A system can apply, consistently and with every record individually correct, a rule of sameness that no artifact declares. This paper formalizes that implemented relation. A declared identity regime partitions a finite record domain into co-reference classes; a disclosed mechanism, through its typed identity-relevant outcomes, induces an operational identity partition of the same domain. The audit compares these partitions in the refinement lattice. A mechanism is faithful when the declared partition refines the operational partition, so no declared class is split. A divergence witness is a pair the declaration merges and the mechanism separates; such witnesses are decidable by pair enumeration. When an imported sibling basis also splits a declared class, local comparison with its partition yields sibling-aligned, sub-sibling, super-sibling, or sibling-incomparable divergence. This result reports only the relationship; it does not identify the basis carried by the mechanism. Global equality of the operational and sibling partitions is defined separately as regime substitution and does not follow from sibling alignment. A version field incremented on every textual edit inhabits the sub-sibling case by splitting declared classes more finely than either imported basis. The audit is three-valued and relative to the disclosed artifacts, evaluated surfaces, and identified uses; each boundary has a finite refuting witness. A passing verdict is non-monotone because extending the transformation history can merge declared classes and create a witness among records already examined.
[AI-91] A Framework for Reputation Aware Uninorm-driven Consensus Algorithms for Blockchain Networks
链接: https://arxiv.org/abs/2607.20700
作者: Bruno Ramos-Cruz,Javier Andreu-Perez,David Richerby,Luis Martínez
类目: Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI)
备注:
Abstract:The operation of blockchain is governed by consensus algorithms (CA). Several consensus mechanisms require significant computational power, while others necessitate high amounts of stakes to select the participant to validate and verify the transactions in the block, leading to centralisation of power and participant exclusion. This paper proposes a novel methodology to address these issues in reputation-based consensus algorithms by studying the reputation behaviour of the validator using intuitionistic fuzzy sets (IFSs) and uninorm aggregation operations (UAOs). Our approach uses IFSs to express the “reputation” because the reputation values in a consensus algorithm eventually imply uncertainty, and IFSs facilitate the representation of a lack of precise knowledge about reputation. Moreover, this methodology utilises uninorm aggregation operations to monitor reputation over time and reinforces the importance of negative and positive reputation. Consequently, this solution allows validators to rectify past failures in subsequent verification processes and foster an equitable consensus algorithm design. The proposed framework maintains linear computational complexity and does not introduce additional communication overhead beyond the underlying consensus protocol. Supported by experimental results, our methodology demonstrates improved performance and evaluation, promising advancements in blockchain network fairness and inclusivity.
[AI-92] Adaptive Multi-Horizon Reinforcement Learning
链接: https://arxiv.org/abs/2607.20656
作者: Manoosh Samiei,Doina Precup,Paul Masset
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Effective decision-making in complex and changing environments requires balancing short-term and long-term consequences. In reinforcement learning (RL), this trade-off is typically controlled through a fixed discount factor, which imposes a single exponentially discounted temporal horizon. However, biological agents exhibit flexible and adaptive temporal discounting, suggesting that effective planning requires multiple timescales. Here, we propose a multi-horizon approach that adaptively selects and combines temporal horizons, enabling robust adaptation to changes in reward structure without manual discount-factor tuning. This flexibility makes the method particularly suitable for continual learning scenarios involving task switches and varying environmental configurations. Empirically, we demonstrate that our approach identifies effective discount factors across a range of MiniGrid environments, including continual settings composed of three sequentially changing tasks. These results suggest that adaptive temporal discounting can improve parameter efficiency and enhance adaptability in both artificial and biologically inspired learning systems.
[AI-93] Scaling Interpretable Transformers with Parity Bottleneck Layers
链接: https://arxiv.org/abs/2607.20652
作者: Andrew Mack,Kraig Yuheng Tou,Mark Henry,Zhengxun Wu,Lauren Greenspan
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Language models are thought to exhibit the phenomenon of superposition, representing many more features than dimensions in their residual streams. Sparse autoencoders (SAEs) are designed to recover such features post-hoc, but training models that are interpretable by construction has remained impractical, as a per-layer over-complete bottleneck is prohibitively expensive in both memory and compute. To overcome this issue, we introduce the ParityTransformer, a GPT-2-scale architecture whose intermediate representations are efficient and wide / sparse by design. At each layer, a Deep Parity Bottleneck (DPB) replaces a learned over-complete basis with a parameter-free algebraic dictionary, providing a deterministic incoherence guarantee and eliminating the memory requirements that have prevented per-layer interpretable bottlenecks at scale. A DPB is a hierarchically structured sparse bottleneck which efficiently enforces sparsity using a multi-level mixture-of-experts approach: a hardware-aware implementation that closes the cost gap between activation sparse and dense training to a manageable interpretability tax. Empirically, ParityTransformers perform at least as well as post-hoc SAEs on sparse probing tasks, while out-performing on measures of feature absorption, steering effectiveness, and fine-grained causal interventions. Because subsequent computation acts only on features that survive the sparse bottleneck, the ParityTransformer’s features are native to the model’s forwards pass by construction, addressing the question of whether SAEs probe features the model actually uses during computation. We see this as a step toward training models whose internal representations are interpretable by design rather than recovered post hoc.
[AI-94] When Does Recurrence Become an Algorithm? Convergence Selection in Weight-Tied Looped Transformers
链接: https://arxiv.org/abs/2607.20594
作者: Tong Zhang,Junhao Hu,Yun Peng,Tao Xie
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:When does a weight-tied looped transformer – one block applied T times – implement an actual algorithm? We answer with four findings from controlled populations on group word problems. (1) The budget law: free training installs a linear computation frontier, a mechanism that solves v positions per loop, whose speed is priced by the training contract: v ~ n_train/T_train (exponent 0.98 +/- 0.04, R^2=0.99), exactly unity under T=n training. SGD selects a frontier matching the minimum the contract demands; granting more test-time loops than ever trained rescues late positions at fixed input length, yielding a principled halting rule T* = ceil(n / v-hat). (2) Architecture prior, not expressivity, picks the algorithm: standard-depth transformers learn parallel scans on this family; weight tying flips the selection to the serial frontier, even when positional addressing for a log-depth scan is supplied. At matched depth and parameters, untied models extrapolate worst and fail to learn A5 at all. (3) The walls are not where circuit complexity says: NC1-completeness costs nothing (A5 generalizes fully), while group order does (S5’s 120x120 operator deadlocks joint learning) – and an operator-first curriculum dissolves the wall in every seed. (4) Mechanisms are portable, not mandatable: warm-starting across budget contracts transfers the algorithm in every seed, re-pricing its speed, while imposing seriality through the input schedule fails where free training succeeds. These results are invisible to standard instruments, which provably saturate at the fixed points trained loops converge to. We introduce a head instrument, the convergence-time scaling tau(n,i), validate it causally via damage cones whose slope reproduces v, and show in-distribution head measurements predict out-of-distribution fate where tail metrics do not. Results replicate on the public easy-to-hard benchmark.
[AI-95] Geometric Configurations of Perturbed Jailbreak Prompts
链接: https://arxiv.org/abs/2607.20581
作者: Lynn Delcon,Andres Algaba,Vincent Ginis
类目: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
备注: 21 pages, 9 figures, 7 tables, 2nd Workshop on Safe AI (SafeAI)
Abstract:Perturbation techniques that turn unsuccessful jailbreak prompts into successful ones are continuously evolving, constituting a major security threat to LLM safety. In this paper, we investigate the internal representations of such string-level perturbed jailbreak inputs in the small weight models of the Qwen-2.5-1.5B/-3B/-7B-Instruct and Llama-3.2-1B/-3B/-3.1-8B-Instruct families. We select two representation spaces: the last-layer-last-token embedding space and the top-50 next-token probability space. The former space separates prompts based on their spelling and format, while the latter space is effectively one-dimensional but appears more complex to cluster. Within our refusal-dominated answer set we find no behavioral hyperplane in either space. Only the next token “Sure” in the 1.5B Qwen model, and both tokens “,” and “ĊĊ” in the 1 Llama model, display a significant association with a compliant-labeled answer.
[AI-96] Joint Utilization of Geospatial and census proxies for Autoencoder-Assisted Downscaling (JUGAAD) of socioeconomic indicators in India
链接: https://arxiv.org/abs/2607.20559
作者: Aditya Dutt,Paul Gader,Aditya Singh
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Monitoring poverty and food security indicators is imperative for addressing socioeconomic challenges in developing nations. A limitation is mismatches in scale between data sources: census data provide geographic coverage, while socioeconomic indicators are derived from infrequently conducted surveys at coarse resolutions, posing a methodological challenge. This study introduces a deep learning framework, JuGAAD, using Indian census and survey data from 2001 and 2011 as a case study. We employ a three-step process: census and geospatial data are averaged into intermediate village-cluster-scale tessellations to reduce noise and regularize administrative boundary changes; an autoencoder compresses high-dimensional National Sample Survey Office (NSSO) data into a low-dimensional latent representation; and a regression model maps upscaled census and geospatial data to this representation. This function is applied to fine-grained census data to generate high-resolution predictions, validated against ground-truth district-level NSSO indicators. Results confirm the methodology predicts socioeconomic indicators at fine scales with strong accuracy.
[AI-97] StabilityBench: Benchmarking Instability in LLM s
链接: https://arxiv.org/abs/2607.20558
作者: Emma Kondrup,Zachary Yang,Anne Imouza,Reihaneh Rabbany
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:AI Assistants are increasingly deployed in high-stakes settings, such as healthcare or government services. Yet their real-world behavior remains poorly understood due to strong context dependence. Current evaluation protocols follow a defense-in-depth paradigm with compounding layers of safeguards, ranging from traditional benchmarks to live or adversarial testing. Such benchmarks remain largely static and single-turn, limiting their ability to capture real-world variability in conversational settings. We propose StabilityBench, a principled, general and model-agnostic benchmark operator that transforms single-turn benchmark queries into multi-turn interaction histories. StabilityBench augments existing benchmarks by injecting realistic user simulations, through demographic proxies or sycophantic baits, while preserving original task intent. We apply StabilityBench to four benchmarks spanning mathematical reasoning, health question-answering and safety, and evaluate nine large language models under these conditions. Our results show that model performance is consistently unstable under these injections, with considerable performance degradations on three out of four benchmarks studied. These highlight important limitations of static evaluations and motivate more realistic evaluation settings. To this end, we propose StabilityBench-Mini: a size-preserving variant of StabilityBench that samples across diversification axes, enabling more realistic evaluation without increasing costs.
[AI-98] Monkey King Bang: A Unified Scientific Multimodal Foundation Model
链接: https://arxiv.org/abs/2607.20557
作者: Hesen Chen,Xinyu Su,Xiaomeng Yang,Yuetan Lin,Zixiong Yang,Junyi An,Fenglei Cao,Yifeng Jiao,Yunqi Zhang,Yuan Cheng,Zhiyu Tan,Hao Li,Libo Wu,Yuan Qi
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Scientific discovery is increasingly shifting from isolated disciplines to multi-domain reasoning, and AI for science faces a similar transition. Existing systems are either specialised for individual domains or unify scientific data mainly through text tokenisation and prompt-based interfaces, limiting their ability to handle diverse scientific inputs, produce modality-native outputs, and support joint understanding, reasoning, and generation across scientific domains. We introduce MKB, a unified scientific multimodal model for both understanding and generation, built around a shared Transformer backbone and modality-tailored encoders, adapters, and decoders. MKB covers six scientific branches, including DNA, RNA, proteins, small molecules, earth science, and medical images, and supports native outputs such as biological sequences, molecular strings, meteorological fields, and segmentation masks. Training follows a two-stage modality-then-language curriculum: Stage 1 aligns modality-specific components with the frozen backbone, and Stage 2 consolidates them with the language backbone using mixed scientific and general corpora. Experiments show that MKB achieves competitive scientific understanding across biological and molecular benchmarks, produces high-fidelity native outputs for weather forecasting, biological generation, and medical-image segmentation, and largely retains the general capabilities of its Qwen3-VL backbone. These results demonstrate the feasibility of the proposed paradigm, suggesting that shared-backbone models with modality-tailored components can provide a promising foundation for future cross-domain scientific multimodal exploration. The model and code are publicly available at this https URL and this https URL.
[AI-99] KeySI: An Interaction Framework for Tuning Text Embeddings Based on Human Feedback IEEE-VIS2026
链接: https://arxiv.org/abs/2607.20556
作者: Yan Zhu,Y. Chen,Rebecca Faust
类目: Artificial Intelligence (cs.AI)
备注: Accepted to IEEE VIS 2026
Abstract:In large-scale text analysis tasks, pre-trained language models are often used to embed text corpora for downstream analysis. However, such models may struggle to capture domain-specific semantics and adapting them typically requires large amounts of labeled data and technical expertise to implement training pipelines. Recent approaches have demonstrated how visual interactions in document projections can capture human feedback as training signals for model tuning. However, these methods operate on document-level feedback, which requires users to open and assess individual documents in order to provide effective feedback. In this paper, we propose KeySI, an interaction framework that enables feature-level feedback through keyword-based concept specification. Users specify feedback by organizing extracted keywords into groups representing concepts, which KeySI translates into document-level supervision for subsequent tuning. By operating on keywords as the primary interaction medium, KeySI reduces the need for manual document inspection and labeling and lowers the barrier to adapting embedding models. We present a prototype implementation that, given a corpus, curates representative keywords, visualizes keywords and document embeddings via dimensionality reduction, allows interactive specification of keyword groups, and supports iterative refinement through system feedback. We evaluate KeySI through a user study, usage scenarios, and quantitative experiments demonstrating its effectiveness in capturing user intent and improving embedding alignment.
[AI-100] AI-Driven Multi-Hop Relay Selection for Smart Urban NR-V2X Networks via Learning-to-Optimize Graph Neural Networks
链接: https://arxiv.org/abs/2607.20554
作者: Giambattista Amati,Federica Mangiatordi,Simone Angelini,Emiliano Pallotti,Pierpaolo Salvo
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Networking and Internet Architecture (cs.NI)
备注: 7 pages, conference
Abstract:Reliable and low-latency NR-V2X communications are essential for smart mobility in dense urban environments. However, limited Road-Side Unit (RSU) density, frequent non-line-of-sight conditions, and highly dynamic vehicular topologies often prevent many Connected and Automated Vehicles (CAVs) from maintaining stable single-hop connectivity. Although multi-hop relay-assisted communication can extend infrastructure coverage, selecting relay links in real time under practical flow, capacity, and connectivity constraints remains challenging. Mixed-Integer Linear Programming (MILP) yields optimal multi-hop relay decisions, but its computational complexity scales sharply with network density, limiting real-time applicability. To address this, we propose a Learning-to-Optimise (L2O) framework based on Graph Neural Networks (GNNs) for real-time NR-V2X relay selection. Vehicular communication states are modeled as attributed graphs, where CAVs and RSUs are nodes and candidate radio links are enriched with propagation-aware features. An offline MILP oracle provides optimal supervision, while an edge-aware Graph Isomorphism Network (GINE) approximates oracle decisions with near-constant inference latency. Experiments on large-scale urban datasets generated by an integrated SUMO–GEMV2 simulation pipeline show that the proposed approach achieves connectivity comparable to that of the MILP oracle while reducing execution time by orders of magnitude. The framework enables cost-effective enhancement of urban V2X connectivity by leveraging existing vehicular assets and supporting scalable, real-time NR-V2X operation in smart city environments. Comments: 7 pages, conference Subjects: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Networking and Internet Architecture (cs.NI) Cite as: arXiv:2607.20554 [cs.AI] (or arXiv:2607.20554v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2607.20554 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-101] SenCos-GEM: SENet-Calibrated and Law-of-Cosines-Constrained Geometry-Enhanced Molecular Representation for Property Prediction
链接: https://arxiv.org/abs/2607.20551
作者: Tianming Han,Li Zhang,Qi Zhao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Biomolecules (q-bio.BM)
备注:
Abstract:Effective molecular representation learning is crucial for accurate molecular property prediction. Recently, numerous self-supervised learning (SSL) approaches leveraging 3D GNNs have been developed to capture comprehensive 3D structural information for drug discovery. However, existing methods lack explicit physical constraints and are highly susceptible to geometric noise induced by coarse empirical force fields during large-scale this http URL, they overlook dynamic feature modulation during downstream adaptation, often resulting in catastrophic forgetting and negative transfer. To address these limitations, we introduce SenCos-GEM, a novel explicitly decoupled geometry-enhanced molecular representation learning framework that incorporates SENet-calibrated and law-of-cosines-constrained enhancements. SenCos-GEM employs a physics-guided geometric consistency loss based on the law of cosines to derive high-fidelity and mathematically invariant 3D spatial priors. In addition, lightweight Squeeze-and-Excitation (SE) modules are integrated into the backbone as task-specific adapters, while a dual-modulation prediction head combines Feature-wise Linear Modulation (FiLM) and SENet mechanisms to enable dynamic feature recalibration. SenCos-GEM demonstrates highly competitive performance across diverse classification and regression tasks on MoleculeNet benchmark, establishing new state-of-the-art results specifically on 3D conformation-sensitive regression tasks, such as FreeSolv, Lipophilicity, and QM9, achieving relative error reductions of 12.9% (RMSE), 5.3% (RMSE), and 8.2% (MAE), respectively. Moreover, our model exhibits superior capability in distinguishing stereoisomers and discriminating conformational perturbations, underscoring its robust spatial modeling performance. Collectively, SenCos-GEM represents a significant breakthrough in accurate molecular property prediction.
[AI-102] Beyond SBDD: Geometric Deep Learning in Polypharmacology and Multi-target Drug Design
链接: https://arxiv.org/abs/2607.20550
作者: Tianming Han,Zhijie Pan,Wenchi Ge,Qi Zhao
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:The traditional “one drug, one target” paradigm of structure-based drug design (SBDD) frequently proves inadequate for treating multifactorial diseases such as cancer and neurodegenerative disorders, owing to compensatory signaling pathways and the emergence of drug resistance. While polypharmacology offers a synergistic therapeutic strategy, the rational design of ligands capable of simultaneously satisfying the geometric constraints imposed by multiple targets remains a major computational bottleneck. This review positions geometric deep learning (GDL) as a powerful integrative approach to overcome these limitations. We systematically survey GDL architectures ranging from invariant graph neural networks to SE(3)-equivariant diffusion models that harness non-Euclidean molecular data to capture intrinsic three-dimensional (3D) structural interdependencies. We critically analyze GDL applications across three core dimensions, including the characterization of shared binding pockets via geometric embeddings, multi-target bioactivity prediction through heterogeneous graph fusion, and de novo generation of dual-target ligands. Particular emphasis is placed on emerging structure-conditioned generative algorithms that integrate diffusion models with reinforcement learning to autonomously resolve complex geometric conflicts between competing binding sites. Furthermore, we evaluate the pivotal role of multimodal omics integration and specialized geometric benchmarking infrastructures in validating these models. By synthesizing these methodological advances, this review elucidates the paradigm shift in drug discovery from serendipitous exploration to rational, structure-driven polypharmacological molecular engineering, thereby providing a clear, structured guide for navigating the complexities of next-generation therapeutics.
[AI-103] SOAP Muon and Beyond: Pushing LLM Pretraining Scales
链接: https://arxiv.org/abs/2607.20548
作者: Mikail Khona,Aditya Vavre,Boxiang Wang,Deyu Fu,Hao Wu,Mike Chrzanowski,Bryan Catanzaro,Dheevatsa Mudigere,Jeff Pool,Michael Lightstone,Mohammad Shoeybi,Mostofa Patwary,Nima Tajbakhsh,Tijmen Blankevoort
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Higher-order optimizers such as Muon and SOAP offer faster convergence than AdamW, but their computational cost and numerical stability challenges have limited adoption at scale. In this work, we adapt and enhance preconditioned gradient methods to overcome the practical challenges of large-scale LLM pretraining. We first identify instabilities in SOAP at large batch sizes and propose algorithmic modifications including per-step QR orthogonalization and improved preconditioning strategies that eliminate loss spikes and enable stable training in these regimes. We then present a unified empirical study of SOAP, Muon, and AdamW using update-RMS matching to ensure fair learning rate transfer across optimizers. As part of this analysis, we empirically evaluate the orthogonalization quality of Muon. Our experiments on multi-billion-parameter models trained on trillions of tokens reveal that SOAP and Muon consistently outperform AdamW at the scales we tested. Notably, at batch sizes of up to 100M tokens for next-token prediction, these optimizers maintain training stability and quality while AdamW degrades. To enable efficient training at large scale, we introduce a layer-wise distributed optimizer compatible with Megatron-LM. Our implementation balances memory and hides communication while avoiding approximations to the optimizer computations, thus retaining their convergence benefits. Additionally, we identify and build specific system-level improvements to further accelerate our layer-wise implementation. To support the research community, we release a codebase that contains emerging algorithms for optimization: this https URL
[AI-104] When RLVR Shrinks the Reasoning Boundary: Diagnosing Pass@k Inversion
链接: https://arxiv.org/abs/2607.20543
作者: Todd Zhou
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement learning with verifiable rewards (RLVR) can improve one-sample accuracy while making a model worse under repeated sampling. We study this pass@k inversion: after training, the policy may solve fewer distinct problems than its base model at large k . The failure concentrates on boundary prompts, where the base model contains rare correct trajectories that are recoverable by sampling but too sparse to reliably appear in finite RLVR rollout groups. We argue that a two-mode account explains this as an absence-of-evidence failure: rare correct trajectories may disappear before RLVR samples and reinforces them often enough. The main contribution is this diagnostic and mechanistic framing. Per-Problem Base Anchoring (PBA) is a deliberately simple proof-of-concept: sharpen prompts with sufficient frozen-base correct evidence, and anchor risky prompts to the base distribution. Across three training seeds on Omni-MATH-Test, with MATH500 as a secondary high-coverage validation benchmark, PBA improves both \PassK1 and high-budget coverage over matched GRPO. A 3000-prompt regime-controlled diagnostic study is consistent across seeds with the expected signature: ordinary GRPO loses base-solvable boundary prompts, while PBA preserves rare verifier-positive trajectories. We use mathematical verifiers as a controlled testbed for verifier-guided optimization; the same pass@k inversion risk applies to ECCV-relevant vision-language agents when repeated visual, spatial, or chart-reasoning attempts are checked by external tools or verifiers. Reasoning post-training should decide not only how strongly to optimize, but which prompts are safe to optimize.
[AI-105] Improving Access to Essential Medicines via Decision-Aware Machine Learning
链接: https://arxiv.org/abs/2607.20542
作者: Angel Tsai-Hsuan Chung,Jatu Abdulai,Patrick Bayoh,Lawrence Sandi,Francis Smart,Hamsa Bastani,Osbert Bastani
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computers and Society (cs.CY)
备注:
Abstract:A critical challenge in healthcare systems in low- and middle-income countries (LMICs) is the efficient and equitable allocation of scarce resources, particularly essential medicines. This problem is complicated by limited high-quality data, which restricts the applicability of traditional data-driven techniques. We propose a novel decision-aware machine learning framework for essential medicines allocation, which additionally leverages multi-task learning to ensure sample efficiency and catalytic priors to ensure equitable allocation. In collaboration with the Sierra Leone national government, we performed a staggered, nationwide deployment of our system as a decision support tool. Our econometric evaluation finds an estimated 19% increase in consumption of allocated products in treated districts, demonstrating its efficacy at improving access to essential medicines. Our tool was subsequently scaled nationwide, covering an estimated 2 million women and children under five. Our work demonstrates how machine learning methods can improve efficiency at very low cost in resource-constrained global health settings.
[AI-106] HypNO: A Graph-Based Neural Operator with Physics-Informed Message Passing for Hyperbolic Conservation Laws
链接: https://arxiv.org/abs/2607.20541
作者: Dimitrije Ždrale,Cassie An Jeng,Katie Wang,Sonia Vanier,Alexandre Bayen,Hossein Nick Zinat Matin
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:We introduce HypNO, a graph-based neural operator for scalar hyperbolic conservation laws. HypNO operates directly on a space-time graph of finite-volume cells and uses adjacency-factored, physics-informed message passing to respect upwinding and entropy admissibility near shocks. We benchmark the architecture on the Lighthill-Whitham-Richards (LWR) and Aw-Rascle-Zhang (ARZ) traffic-flow models, a stress test for operator-learning methods because of their simultaneous global transport and shock formation. HypNO predicts solution snapshots accurately across a range of initial conditions while capturing the shocks and discontinuities of the solution.
[AI-107] From Atoms to Entropy: Optimal Noise Allocation for Diffusion Training in the Convex Regime
链接: https://arxiv.org/abs/2607.20540
作者: Luca Ambrogioni,Giulio Franzese,Alberto Foresti,Gabriel Raya,Bac Nguyen,Georgios Batzolis,Yuhta Takida,Naoki Murata,Chieh-Hsin Lai,Yuki Mitsufuji
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:How should a diffusion model decide which noise levels to train on, and how much? Despite the importance of this choice, current noise schedules are based largely on heuristics or empirical tuning. Here, we develop a general statistical framework for studying asymptotically optimal noise-level allocation in diffusion training. Our first main result concerns the fully coupled regime, where information can spread between different time points. Under convexity or Polyak-Lojasiewicz-type assumptions, we show that the optimized training schedule admits an atomic minimizer, concentrated on finitely many noise levels. Our second main result specializes this framework to an idealized independent-learner regime, intended to model temporal specialization in neural networks. Under an additional feature-noise decoupling condition, a random-matrix analysis leads to an information-theoretic proxy: the decoupled sampling density is proportional to the square root of the generative entropy rate, the rate at which conditional entropy grows along the forward process. We test these predictions in controlled settings where the coupled objective can be optimized directly, including Dirac mixtures, low-dimensional manifolds, and MNIST. In these settings, the optimized schedules are consistently finite-support, while the smooth entropic proxy closely tracks the atomic optimum in neural-network models and breaks down mainly in the fully coupled parametric case, as the theory suggests. We then evaluate the entropic schedule in larger-scale experiments, where full schedule optimization is currently intractable. The results indicate that square-root entropy scheduling can substantially improve training efficiency on discrete domains and remains competitive with standard EDM-style heuristics on continuous images.
[AI-108] Leverag ing Biokinetic Knowledge Priors for Data-Scarce Bioprocess Modeling ICML2026
链接: https://arxiv.org/abs/2607.20539
作者: Kyunghoon Hur,Eunjung Jeon,Hyun Woo Kim,Gyubok Lee,Seongjun Yang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at ICML 2026 AI for Science Workshop
Abstract:While deep learning has accelerated drug discovery, its impact on biomanufacturing has been considerably more limited. The reason is data scarcity. Bioreactor experiments are high-cost, take days to weeks, and are rarely shared in public form, leaving each research work with only a handful of experiments. The domain itself, however, is rich in prior knowledge. Biokinetic ordinary differential equation (ODE) models have described microbial growth for decades, yet how to inject this knowledge into a neural network has not been studied systematically. We present the first systematic study of how to inject this ODE knowledge into a neural network, comparing a data-level prior that pre-trains a generic decoder on simulated ODE curves against an architecture-level prior that embeds the ODE inside the decoder. Both consistently outperform no-prior baselines across 11 datasets and 7 microbial species. Our central finding is that the two are substitutable. A generic decoder pre-trained on simulation matches a fully bio-structured decoder trained on real data. Simulation pre-training therefore offers a simple, data-efficient recipe for deep learning under bioprocess data scarcity.
[AI-109] Codec-Gauge: Learning Compression-Friendly Gauges for Transformer KV Caches
链接: https://arxiv.org/abs/2607.20538
作者: Yitao Jiang,Yaoqing Yang,Luyang Zhao,Muhao Chen,Devin Balkcom
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Long-context Transformer inference increasingly relies on KV-cache compression or quantization. Prior rotation and transform-coding results suggest that the channel basis of each key/value vector affects how faithfully a fixed backend preserves model behavior. We introduce Codec-Gauge, a post-training cache-coordinate layer that learns small orthogonal channel transforms around existing compression and quantization backends. Its frequency-distribution objective combines a token-channel DCT spectral-centroid loss with a smooth rate proxy to concentrate KV energy in low-frequency codec-facing layouts. We evaluate actual compression and decompression using measured bytes and rolling compressed-history scoring. Across six models at 3 , 4 , and 6 bits/value, learned gauges reduce zfp KL divergence by 44.0% on average relative to raw coordinates and outperform random, Hadamard, DCT, and PCA/KLT controls. The same gauges improve quality preservation for block-uniform and KIVI-style quantization. Experiments on a 27B model and long-context task prompts reproduce the quality trend, while serial storage and timing measurements validate the implemented compressed-cache paths. These results establish cache-coordinate geometry as a practical post-training variable for improving compression fidelity without changing model weights, attention semantics, or backend coding rules.
[AI-110] ReliableTableQA:How Much Supervision Does Reliability Annotation Need?
链接: https://arxiv.org/abs/2607.20537
作者: Huei-Chung Hu,Hsin-Tai Wu,Koyo Kobayashi
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 12 pages, 2 figures
Abstract:We introduce ReliableTableQA, a framework for training an LLM to annotate the statistical reliability of tabular QA results, not whether the query is answerable, but whether the computed answer is statistically meaningful. In real enterprise analytics, a syntactically correct SQL query can return a value that is based on too small a sample, has an excessively wide confidence interval, or is too confounded to support action. Existing systems answer confidently in all such cases, a failure we quantify as the Unreliable Confident Answer Rate (UCAR). We contribute (1) a ten-category reliability taxonomy (R1-R10) covering hazards such as small-sample aggregates, multiple-comparison inflation, and distribution-tail mismatch; (2) a program-first data pipeline that generates 50,000 reliability-labeled training examples from a context-free grammar over public retail schemas, with schema-stratified SFT/GRPO splits; and (3) a controlled study of how much supervision calibrated reliability annotation actually requires. We find that a small, schema-stratified SFT set is remarkably sufficient: 200 examples raise reliability-flag F1 from 0.61 to 0.98 and parse rate from 0.52 to 1.00, drive UCAR to zero, and yield a model that generalizes to an unseen retail domain (Rel-F1 0.997 on held-out HM). Against this strong SFT baseline, GRPO, commonly assumed to be essential, helps only when SFT is under-trained (+0.06-0.16 exact-flag-set match at 100 examples, in- and out-of-distribution) and provides no measurable benefit once SFT is adequate, a null result we confirm across a hard compound-flag slice, a strict exact-match metric, and out-of-distribution evaluation. Our findings reframe reliability annotation as a data-efficiency problem and delineate precisely when reinforcement fine-tuning does and does not pay off.
[AI-111] Grounding Investor Views: Neural Predicates in the Black-Litterman Model
链接: https://arxiv.org/abs/2607.20533
作者: Marcos Florencio
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Portfolio construction under the Black-Litterman model requires investors to specify views on asset returns alongside explicit uncertainty estimates – a process that remains largely subjective and difficult to scale. We propose a formal approach in which neural predicates serve as a structured, probabilistic mechanism for view generation. In our formulation, structured financial analysis data is processed through a compositional hierarchy of neural predicates whose outputs – probability distributions over market stances – are mapped to the pick matrix \mathbfP , the view return vector \mathbfq , and the view uncertainty matrix \boldsymbol\Omega of the Black-Litterman model. View confidence is derived from predicate output distributions, providing a data-driven alternative to subjective uncertainty elicitation. The resulting approach is interpretable, in the sense that any portfolio weight can be traced back through the predicate’s logical chain to the underlying data, and fully differentiable, enabling end-to-end learning.
[AI-112] Position: Stop Reactively Patching Your Model Every Time and Start Proactive Test-Driven AI Development ICML2026
链接: https://arxiv.org/abs/2607.20532
作者: Nadine Chang,Maying Shen,Jialiang Wang,Rafid Mahmood,Jose M. Alvarez
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: Accepted at ICML 2026 Position Paper Track. 18 pages
Abstract:Many modern AI systems are designed to operate under diverse, open-ended, use-cases. To help generalize deployed systems, many deployed-system maintenance pipelines use a reactive AI flywheel that observes emerging feedback from user behavior (errors) and patches the model accordingly. However, when used as the primary maintenance mechanism, these flywheels often ignore the broader context of these errors within the system’s objectives, failing to preempt potential future edge cases, which leads to more unnecessary flywheel iterations. Also, it is statistically increasingly difficult to collect remaining errors due to the long-tail nature of open-world use-cases. This position paper argues that a proactive test-driven flywheel is required to address reactive flywheel’s limitations and to approach a generalizable system. We advocate for creating a “test space” to technically map feedback data to task objectives, evolving the flywheel from reactive to proactive. We augment our position by mathematically proving a proactive one achieves better long-term scaling with fewer iterations than the reactive flywheel.
[AI-113] DynamicMCPBench: A Trace-Grounded Effect-Scored Benchmark for LLM Agents over Live MCP Servers
链接: https://arxiv.org/abs/2607.20531
作者: Jerzy Kamiński,Ilya Galyukshev,Artem Kuznetsov,Sergey Chuprin,Kirill Redko,Aidar Shumbalov,Anna Kalyuzhnaya
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language model (LLM) agents are increasingly deployed over Model Context Protocol (MCP) servers, yet the benchmarks used to evaluate them score the final answer or a fixed “ground-truth” list of tools, both of which are fragile once the underlying data is live and stateful. We present DynamicMCPBench, a reusable framework rather than a fixed dataset. A practitioner can run it on their own MCP servers to test models on their own tasks, or let it collect servers automatically to measure a model’s general ability to solve agentic tasks. Given the servers and any set of models, it generates realistic goals, pursues each one live to record a successful trajectory, distills that trajectory into path-agnostic effect checkpoints, and scores an agent on whether it reproduces those effects, never on the final answer. To show what the framework reveals, we run it at scale: 24 models over 121 servers and 750 tasks spread evenly over 15 task categories (50 each), where each category targets a distinct tool-use challenge of the generated questions. Each task is scored by pass^3: it counts as solved only if all three independent attempts succeed. Even the strongest agents solve only about half of the tasks, 31% of tasks are solved by no model at all, and accuracy collapses as the required tool chain grows longer (from 39% on the shortest chains to 13% on the longest). A human validation study confirms the automatic scoring is reliable (chance-corrected agreement of 0.76). DynamicMCPBench thus turns benchmark construction into something practitioners can rerun on their own servers and models, while exposing a consistent inability of current agents to handle long, multi-step agentic tasks.
[AI-114] CLOE: Christoffel Loss Autoencoder for Anomaly Detection
链接: https://arxiv.org/abs/2607.20530
作者: Léa Billet(LAAS, INSA Toulouse, ANITI),Louise Travé-Massuyès(LAAS-DISCO, Comue de Toulouse, ANITI),Elodie Chanthery(LAAS),Alexandre Gaffet
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
备注:
Abstract:Semi-supervised anomaly detection plays a key role in diverse fields such as process monitoring, healthcare, and finance. However, lightweight methods often struggle with high-dimensional data and typically require careful tuning of multiple hyperparameters. Among existing approaches, Christoffel Function–based methods are attractive due to their simplicity, requiring at most a single hyperparameter. They also benefit from a well-established theoretical foundation that yields several interesting results for data science. However, their main limitation is poor scalability to high-dimensional settings. In this paper, we introduce CLOE, a new method that combines an autoencoder for dimensionality reduction with a Christoffel Function–based detector applied in the latent space. To better align representation learning with anomaly detection, we design a novel loss function that leverages the Christoffel Function to guide the autoencoder toward representations that better capture the support of the normal data distribution. We further propose a principled procedure to set the detection threshold and an efficient strategy to tune the single remaining hyperparameter. Experiments on multiple high-dimensional tabular anomaly detection benchmarks demonstrate that CLOE achieves superior performance compared to existing methods, while preserving the lightweight and low-tuning advantages of Christoffel Function–based approaches.
[AI-115] Uncertainty-Aware Trust Estimation for Multi-LLM Systems via Structured Expert Judgement
链接: https://arxiv.org/abs/2607.20529
作者: Jiawei Zheng,Jiazhen Zhang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Model (LLM) ensembles are increasingly used to improve reliability by combining predictions from multiple LLMs. However, existing aggregation methods typically assume that all models are equally trustworthy, overlooking differences in uncertainty quality. This assumption is poorly suited to heterogeneous LLMs, whose reliability and capability vary significantly, making naive aggregation vulnerable to unreliable or adversarial experts. In this work, we formulate multi-LLM aggregation as a problem of uncertainty-aware trust estimation. We adapt structured expert judgment from decision theory, using context-aware calibration questions to estimate expert reliability based on the quality of its probabilistic predictions. Specifically, we employ Cooke-style log weighting, which penalises overconfident incorrect predictions and favours well-calibrated experts. We evaluate our approach on MMLU and MMLU-Pro across homogeneous, heterogeneous, and contaminated expert panels. Results show that while aggregation methods perform similarly in homogeneous settings, Cooke weighting becomes critical under heterogeneity and contamination. It achieves a superior accuracy-reliability balance and remains robust when unreliable experts are introduced. These findings suggest that Multi-LLM aggregation requires not just combining predictions, but calibrating trust under uncertainty.
[AI-116] PromptPack: Scaling LLM Annotation Agents for Online Recommendation
链接: https://arxiv.org/abs/2607.20528
作者: Sebastian Koralewski,Merwan Barlier,Yulia Stolin,Blaž Škrlj
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Online recommendation platforms increasingly use Large Language Models (LLMs) to extract structured features from ad creatives. While deploying a single-call LLM annotation agent yields significant Click-Through Rate (CTR) improvements in our live production environment, per-creative prompting is prohibitively expensive to scale. The redundant system instructions sent in every request account for 94% of billed input tokens. To break this cost bottleneck, we introduce PromptPack, a scalable, high-throughput LLM annotation agent. PromptPack achieves this scale via in-context batching, combining a shared system prompt, a strict XML structural envelope, and an output correction layer to ensure deterministic, pipeline-ready feature extraction across multiple creatives simultaneously. We evaluate PromptPack via an offline retrieval benchmark using a downstream logistic-regression ranker. To deeply profile the agent’s behavior, we measure AUC and introduce Volume-Weighted Absolute Lift (VWAL), a novel metric capturing the signal quality of the generated features. Compared to our live, unbatched production baseline, PromptPack at batch size 20 cuts our LLM costs by 89% and accelerates throughput by 2.5x while fully preserving AUC.
[AI-117] Evaluating and Guarding Citation Faithfulness in Agent ic Scientific Synthesis
链接: https://arxiv.org/abs/2607.20527
作者: Taewan Goo,Junsik Kim,Kyulhee Han,GwonYul Jo,Jong-Soo Kim,Tae-Hyung Kim
类目: Artificial Intelligence (cs.AI)
备注: 20 pages, 5 figures. Includes supplementary material (Appendices S1-S6). Open single-GPU reproducibility kit included
Abstract:Agentic LLM systems such as OpenScholar and PaperQA2 read the scientific literature and return cited answers, and both they and their benchmarks already check whether those citations hold, with a fixed attribution model or human graders. Neither audits the reliability of that check itself. We show it is not reliable, and that this matters. On identical agent outputs the measured unsupported-citation rate ranges from about 3% to about 18% depending only on the verifier’s strictness, and although verifiers agree on which citations are supported, they disagree on which to flag (negative-specific agreement 0.27 to 0.30), so no single flag set is trustworthy and cross-paper comparison is invalid without a named verifier and protocol. We present a gold-anchored evaluation protocol and a deployable guard that make this behavior measurable and bounded. The protocol validates the verifier, measures re-attribution, and calibrates a guarantee against human gold rather than another model’s verdict; the verifier is a swappable instrument chosen on cost (recall 0.94 on the supported class, held out), and re-attribution is a commodity step where a deterministic BM25 matches the best open generator. The guard adds a split-conformal layer placing a distribution-free, finite-sample bound on truly unsupported citations that slip past a chosen flagging rule, a guarantee on catch rate rather than conclusion correctness. The bound holds on held-out gold, and we identify and quantify the condition governing its transfer to deployment, calibration-negative difficulty, with a concrete recalibration recipe, left untested by prior conformal-factuality work. Validated across four open 27-35B models and three agentic pipelines on public benchmarks (SciFact, QASA, PubMedQA), with confidence intervals on every headline number, the protocol and guard ship as an open single-GPU kit. Comments: 20 pages, 5 figures. Includes supplementary material (Appendices S1-S6). Open single-GPU reproducibility kit included Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2607.20527 [cs.AI] (or arXiv:2607.20527v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2607.20527 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-118] ConfidenceBench: Evaluating Confidence Calibration in Large Language Models
链接: https://arxiv.org/abs/2607.20526
作者: Matthew ffrench-Constant,Daniel Yang,Xinmeng Huang,Sanyam Kapoor
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Machine Learning (stat.ML)
备注:
Abstract:Large language models (LLMs) are increasingly deployed in settings where fluent but incorrect answers can be costly. In these settings, accuracy alone is insufficient: models must also know when they are likely to be wrong. We present ConfidenceBench, a calibration benchmark that evaluates verbalized confidence estimates in 15 frontier LLMs using the Brier score, a proper scoring rule that incentivises truthful probability reporting. Confidence is elicited via prompting, requiring no access to model logits and making the framework applicable to both closed-source and open-source systems. The benchmark comprises 200 private multiple-choice questions across four categories: spatial reasoning, high-precision mathematics, word lookup, and unknowable questions. Across three independent runs, Claude Opus 4.6 and Gemini 3.1 Pro Preview achieve the lowest Brier scores, both reported as 0.103. Both substantially outperform the calibrated-random baseline of 0.1875, while Gemini 3.1 Flash-Lite scores 0.367, indicating severe miscalibration. Accuracy and calibration diverge substantially across model families: the most accurate model is not the best-calibrated, and several models perform worse than the calibrated-random Brier baseline despite reasonable accuracy. These results show that verbalized confidence calibration is a distinct and practically important axis of LLM reliability, complementary to standard accuracy-based evaluation.
[AI-119] Autonomous disproofs of the sum-product conjecture over mathbb R with GPT -5.5 Pro
链接: https://arxiv.org/abs/2607.20525
作者: Yichen Huang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:OpenAI’s recent disproof of the Erdős unit distance conjecture marked a milestone for AI in mathematics. It also inspired another breakthrough: a human disproof of the Erdős–Szemerédi sum-product conjecture over \mathbb R . In this paper, we present a simple agent built on GPT-5.5 Pro. Using a problem-agnostic, three-stage prompting pipeline – proof-plan proposal, proof construction, and review – the agent autonomously generated correct proofs that the sum-product conjecture is false over \mathbb R in 7 of 8 independent trials; in the remaining trial, it identified an unresolved gap in its argument. The seven proofs are diverse: some are close to existing unit-based constructions, while others avoid units by using L^p -type regions of algebraic integers. The system used an average of 132.4k reasoning tokens per trial. We release the code, intermediate outputs, and generated proofs, providing a reproducible, data-contamination-free case study in autonomous proof generation.
[AI-120] Attention Degradation Function Token Anchoring and the Limits of Attention-Based Intervention in Large Language Models
链接: https://arxiv.org/abs/2607.20524
作者: Sagar Dangal,Manoj Shakya
类目: Artificial Intelligence (cs.AI)
备注: 19 pages, 2 figures, 10 tables
Abstract:Mean cross-positional attention degradation is widely reported in transformer interpretability, yet whether it causally limits contextual retrieval remains untested. We present six coordinated experiments across GPT-2, LLaMA-3.2-1B/3B, OPT-1.3B, and distilgpt2. We first characterise short-term (5-100 token) attention degradation, finding a universal exponential-then-plateau pattern whose rate is inversely correlated with depth, with distinct layer-wise entropy signatures per architecture. Function token anchoring proves architecture-dependent: OPT-1.3B (absolute positional encoding) shows distance-dependent preposition specificity, GPT-2 shows uniform non-specific dependence, and LLaMA (RoPE) shows reversal at long distances. Strategic comma insertion at clause boundaries causally reduces prediction degradation in the 40-80 token range, with the benefit tied to syntactic boundary alignment rather than token density. We then test the mechanism causally: Relay-Aware Attention (RAA), which biases attention logits toward function token positions, verifiably increases attention mass by 16-24% yet yields null effects on GPT-2 and LLaMA-1B, preliminary harm on LLaMA-3B, and a mixed effect on OPT-1.3B that nets to approximately zero. Multi-fact retrieval probes further show that degradation rate does not predict retrieval accuracy across models. We conclude that mean attention degradation is largely descriptive rather than prescriptive: function tokens contribute through what their hidden states compute, not through the attention they receive – with implications for interpretability methodology and attention-score-based inference optimisations such as KV-cache eviction.
[AI-121] CANN Bench: Benchmarking Agent Generated Kernels against Real NPU and Algorithmic Limits
链接: https://arxiv.org/abs/2607.20518
作者: Xue-Jian Gao,Deng Pan,Yueming Su,Jiasheng Li,Bin Du,Fengming Zhu,Chengdi Ma,Junyi Fan,Qichen Liao,Chengqiu Hu,Xinxian Chen,Lingchao Zheng,Jun Li,Jiwei Yang,Yuwei Fan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:AI agents are now capable of writing, compiling, and iteratively optimizing low-level operator kernels on different hardware platforms. Existing benchmarks, however, focus almost exclusively on CUDA and Triton, leaving hardware ecosystems with less-exposed programming models without a common evaluation baseline. We present CANN Bench, an open benchmark for AI-generated operator code on Huawei’s Ascend NPU. The current release covers 53 operators and 1060 test cases organized into four difficulty tiers – from simple elementwise primitives to MoE dispatch and FlashAttention kernels – spanning FP16, BF16, FP32, and INT8 precision formats. Evaluation adopts a \textbfthree-dimensional weighted composite score that treats compilation, functional correctness, and performance as independent axes, providing a principled reward signal for kernel-generation agents. Performance is graded against an out-of-the-box PyTorch-on-Ascend baseline and an analytical per-case Hardware-Anchored Performance (HAP) limit on real NPU hardware, ensuring scores reflect genuine optimization headroom rather than measurement artifacts. The evaluation harness is designed to resist reward hacking from the ground up. CANN Bench is versioned within the official CANN repository and is designed for long-term community co-construction, providing the Ascend ecosystem with a quantitative, reproducible, and sustainably maintained yardstick for AI operator-authoring capability.
[AI-122] Scaling Closed-Loop Feature Channel Configuration with LLM s
链接: https://arxiv.org/abs/2607.20516
作者: Tolgay Atinc Uzun,Radu Timofte,Dmitry Ignatov
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注: 15 pages, 8 figures
Abstract:Promising initial results in closed-loop large-language-model-based channel-configuration search demonstrated that neural-network widths can be optimized directly through executable code generation and accuracy feedback. However, those results were obtained from a relatively sparse set of valid evaluations, leaving open whether the observed optimization behavior transfers to a denser sampling regime and whether additional architectural regularities emerge when more generated networks are evaluated. To test this, the same search setting is scaled to 250 candidate networks per fine-tuning cycle. The analysis covers 2000 generated candidates from 8 complete cycles, yielding 462 verified CIFAR-100 evaluations after task and metadata filtering. Per-cycle mean accuracy exhibits a positive linear trend with slope 9.87e-4 (p=0.043), while the high-performing frontier improves more strongly: the best observed accuracy increases from 0.3144 to 0.3676, and both the top-5 and top-10 cycle-level means exhibit positive trends. The scaled run also reveals improved parameter efficiency. The best model reaches 0.3676 with 11.8M parameters, compared with an early high-performing model at 0.3144 with 166.5M parameters. Beyond accuracy, the larger sample exposes architectural regularities that were difficult to assess from sparse observations. Non-power-of-two channel widths occur in 41.8% of verified candidates, and the strongest models share structured channel-allocation patterns characterized by moderate early widths and expanded middle or later blocks. These findings indicate that the channel-search signal observed in the initial study transfers
[AI-123] Reliability-Aware LLM Alignment from Inconsistent Human Feedback
链接: https://arxiv.org/abs/2607.20515
作者: Jingyi Huang,Ruohan Zong,Yujun Feng,Liran Ma,Lanyu Shang,Yang Zhang
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Reinforcement Learning from Human Feedback (RLHF) is critical for aligning Large Language Models (LLMs) with human preferences. However, its efficacy is often compromised by the inherent inconsistency and subjectivity of human annotations. Existing preference optimization frameworks, such as Direct Preference Optimization (DPO), typically treat ambiguous pairs with high annotator disagreement identically to those with unanimous consensus, forcing models to overfit to inconsistent supervision signals and leading to suboptimal alignment. In this work, we propose Reliability-Guided Preference Optimization (RGPO), a robust framework designed to mitigate the impact of inconsistent human feedback. RGPO estimates annotator reliability and infers latent ground truth labels from noisy human feedback to identify robust preferences. Furthermore, we introduce a reliability-aware consistency optimization that dynamically modulates the training objective based on the consensus level of annotations, ensuring the model prioritizes high-consensus supervision signals. Extensive experiments on LLM alignment benchmarks demonstrate that RGPO effectively reduces inconsistency and noise in training data and achieves superior performance compared to widely adopted RLHF baselines. Our code and configurations are available at this https URL.
[AI-124] he Active Ingredient in Muons Grokking
链接: https://arxiv.org/abs/2607.20512
作者: Yufeng Wang
类目: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
备注:
Abstract:The Muon optimizer reaches the grokking threshold on modular arithmetic faster than AdamW. Prior work attributes this to “spectral-norm constraints plus orthogonalized momentum” but does not isolate which mechanism matters. To better understand Moun’s behavior, we run multi-seed and multi-learning-rate sweeps to decompose and stress-test the effect. First, an ablation shows the speedup comes from orthogonalization (the Newton-Schulz iteration): orthogonalize-only matches full Muon, whereas spectral-only is no faster than AdamW and is unreliable, and this verdict holds across learning rates. Second, a mechanistic analysis finds that orthogonalizing optimizers reach generalization at roughly 3x lower spectral norm and, controlling for how much the embedding actually moves, settle into a lower-norm solution rather than simply perturbing the embedding less. Third, reducing the Newton-Schulz iteration count from five to one accelerates reaching the threshold but makes the grokked solution fragile, prone to transient collapse, with fragility that grows with learning rate; a single iteration is fast and stable only at small learning rate, while the canonical five iterations are the learning-rate-robust choice. We also show spectral scaling can be dropped at no measured cost. A methodological thread runs throughout: under a stability-aware metric, “faster” claims about grokking optimizers can invert, so we report both first-crossing and sustained-grok times. To support reproducibility, we release our full training and analysis code at this https URL
[AI-125] SiGMA: Sign-Guided Merging and Adaptation for Multimodal Continual Instruction Tuning ECCV2026
链接: https://arxiv.org/abs/2607.20511
作者: Keonhee Park,Gunhee Kim
类目: Artificial Intelligence (cs.AI)
备注: Accepted at ECCV 2026
Abstract:Multimodal Continual Instruction Tuning (MCIT) is crucial for adapting Multimodal Large Language Models (MLLMs) to evolving a sequence of downstream tasks. Prior methods mostly utilize Mixture of Experts or expansion merge approach, primarily focusing on catastrophic forgetting, yet they still suffer from negative interference during inference, where newly learned updates overwrite useful prior knowledge and degrade overall performance. To address this, we propose SiGMA (Sign Guided Merging and Adaptation), a simple yet effective framework that mitigates negative interference with two components: sign guided adaptive tuning during training and sign guided merging at inference. Sign guided adaptive tuning reduces collisions with past knowledge and learns the current task with minimal drift, mitigating severe forgetting. Sign guided merging further improves consolidation by selectively scaling salient parameters to preserve and amplify useful task specific knowledge. Experiments on UCIT and DCL benchmarks show that SiGMA significantly reduces negative interference and outperforms state of the art MCIT methods. Our code is available at SiGMA.
[AI-126] MiniCache: Reusable Program Caching with Small Model Interfaces for Efficient LLM Inference
链接: https://arxiv.org/abs/2607.20507
作者: Jingquan Chen,Jinghua Piao,Jie Feng,Shaogang Hu,Yong Li
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 16 pages, 8 figures
Abstract:Large language models (LLMs) are increasingly used for program-aided reasoning, agentic decision making, and structured task execution, but these applications often incur high inference cost. We present MiniCache, a reusable program caching framework that transforms Program-of-Thought (PoT) programs into parameterized cache objects, enabling reusable computation across structurally similar requests. MiniCache reuses the same small model for semantic variable extraction on cache-hit requests and speculative drafting during target-LLM generation, reducing expensive target-LLM invocations while preserving task quality. Experiments on shopping-style request datasets, WebShop, Formula, and CodeTAT-QA demonstrate that MiniCache improves the trade-off between inference latency, cache reuse, and accuracy, achieving up to 3.1x lower latency and 2.8x higher throughput under parallel serving. These results show that small models are most effective not as replacements for large models, but as lightweight interface models that enable reliable and efficient reusable program caching.
[AI-127] Optimizing Hypergraph-Based RAG : Toward Better Fact Extraction and Chunk Retrieval
链接: https://arxiv.org/abs/2607.20506
作者: Houda Khrouf,Pedro Fillastre,Sebastiao Correia
类目: Artificial Intelligence (cs.AI)
备注: APIA 2026 conference
Abstract:GraphRAG enables deeper reasoning by structuring knowledge as graphs but struggles with n-ary facts. HyperGraphRAG uses hypergraphs for richer semantics, improving accuracy, yet relies on error-prone LLM extraction and inefficient standard chunk retrieval. We address this by employing self-consistency prompting to improve the extraction, and Personalized PageRank algorithm over hypergraph to enhance chunk retrieval.
[AI-128] LeanFlow: A Case Study in Workflow-Driven Lean Autoformalization ICML2026
链接: https://arxiv.org/abs/2607.20503
作者: Lazar Milikic,Simon Guilloud,Khanh Nguyen,Viktor Kuncak
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Logic in Computer Science (cs.LO)
备注: 14 pages, 3 figures, ICML 2026: AI for Math Workshop
Abstract:We present and evaluate LeanFlow, an LLM agent system specialized for translating mathematical papers into buildable Lean projects. Recent verifier-in-the-loop systems show that large formal artifacts can be produced, but it remains unclear which runtime mechanisms affect completion, auditability, or efficiency in document-to-project formalization. We study this question through case studies on two previously unformalized mathematical papers in number theory and measure theory, using model, proof-workflow, and toolset ablations with Kimi2.6 and GPT5.5; we report task outcome, API calls, input tokens, and output tokens. With Kimi2.6, the full workflow completes both document-level projects within the 2000-call budget, while no-queue variants reach the budget limit; with GPT5.5, all document-level variants complete, and the full workflow has the lowest or tied-lowest input-token cost on both sources. As complementary calibration, LeanFlow reaches 75.7% BEq+ on the PFR slice of RLM25 and solves all five ICML 2026 AI for Math TCS challenge projects in our GPT5.5 runs.
[AI-129] Inducing Comparability of Factorised Probability Distributions
链接: https://arxiv.org/abs/2607.20502
作者: Jan Speller,Malte Luttermann,Marcel Gehrke,Tanya Braun
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:To allow for principled comparison between two probabilistic graphical models defined over non-identical variable sets, they have to be lifted to a common measurable space. To this end, we propose an extension scheme for any two given models and establish the formal foundation: Unmatched components are completed using conditionally uniform (Laplace) extensions such that the resulting joint distributions differ from the original ones only by multiplicative constants and coincide under projection. This preserves the probabilistic semantics while enabling the application of well-defined distributional discrepancy measures. We establish the invariance of the induced joint under projection and use the extensions to provide a minimal structural extension of two factor graphs to the smalles common measurable space as well as to a common graphical structure by a deterministic algorithm. In addition, we discuss structural and measure-theoretic properties and identify promising criteria for comparison methodologies.
[AI-130] FlowEdit: Information-Theoretic Control of LLM Reasoning Flows for Ill-posed Problems Involving Conflicts
链接: https://arxiv.org/abs/2607.20500
作者: Sizhe Tang,Guangyu Jiang,Yu Li,Rongqian Chen,Ioannis G. Kevrekidis,Tian Lan
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large Language Models (LLMs) perform strongly on well-specified reasoning tasks with a feasible answer. However, problems encountered in the open world can become ill-posed due to inconsistent conditions, conflicting statements, or mutually incompatible requirements, admitting no valid responses. We argue that reasoning of such ill-posed problems involving conflicts require novel LLM capabilities to make hidden conflicts explicit, maintain competing hypotheses via multiple reasoning branches, and generate alternative responses in a single pass, all of which are challenging due to the limitation of the next-token prediction mechanism in LLMs. To this end, we propose FlowEdit, a novel framework that leverages information-theoretic principles to quantify and regulate internal reasoning flows of LLMs, for generating a full set of alternative responses under valid hypotheses. FlowEdit can be viewed as enforcing a branch-aware reasoning process using two dual information-theoretic objectives on the model’s internal reasoning representations: maximizing the information flow from each selected hypothesis to the branch outcome, while minimizing the overlap and conditional dependence across sibling branches, to provide a diverse, informative set of responses with broad coverage. We show that this is achieved through tractable variational bounds under boundary embeddings being \epsilon-sufficient, optimizing the underlying conditional mutual information in LLM reasoning process. Extensive experiments demonstrate that FlowEdit outperforms leading proprietary models, improving exact-set-match accuracy by 68%, while boosting overall response informativeness by 24%. We further show that flow regulation surfaces in the token stream as a redistribution of next-token entropy that concentrates inside each branch, amplifies at flow boundaries, and scales with the number of flows the problem requires.
[AI-131] ExecuGraph: A Multi-Agent Execution-Grounded Framework for Reliable Backend Code Synthesis with Large Language Models
链接: https://arxiv.org/abs/2607.20499
作者: Sai Deekshith Lekkala,Jothi Prabha Appadurai,Rohith Reddy Bellibatlu,Manpreet Singh
类目: Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
备注: Submitted to Data Science and Management
Abstract:Large Language Models generate plausible backend code, but a single-pass paradigm provides no guarantee of correctness or runtime reliability. We present ExecuGraph, a multi-agent framework that places execution-based validation at the center of backend code synthesis. Six specialized agents (Planner, Code Generator, Logical Reviewer, Evaluator, Optimizer, and Explainer) are coordinated by a typed directed workflow with a bounded retry budget, implemented on LangGraph with locally hosted models (Ollama) and an optional retrieval layer for algorithmic technique recall. A subprocess-isolated sandbox with a wall-clock timeout guards every evaluation. We evaluate on a curated 30-problem DSA suite (internal-30), HumanEval (n=64), and an APPS-introductory subset, contrasting ExecuGraph against a single-agent one-shot baseline and a single-agent execution-retry baseline (a Reflexion-style ablation that isolates the contribution of multi-agent decomposition). On internal-30, the three conditions are statistically indistinguishable (n=30; paired Wilcoxon p=0.59 MF vs. SO, p=0.08 SR vs. SO); 95% bootstrap confidence intervals on all pairwise mean differences include zero. On HumanEval, multi-full edges ahead by +3.1 pp. The strongest signal is cross-model: with DeepSeekCoder V2 Lite, graph-category accuracy improves from 57.5% (oneshot) to 80.0% (multi-full), a +22.5 pp jump that supports a scaling hypothesis: the value of multi-agent decomposition grows with base-model capability. The framework’s primary contribution is methodological: a single codebase that collapses by configuration into one-shot, execution-retry, and per-agent ablation conditions, enabling controlled measurement of each lever’s marginal contribution. A per-agent ablation, retry-budget sweep, error-class taxonomy, and test-source audit are reported.
[AI-132] AISE-Bench: A Full-Cycle Curated Benchmark for Information Seeking on Academic Knowledge Graphs KDD2026
链接: https://arxiv.org/abs/2607.20498
作者: Fanjin Zhang,Zhengyang Wang,Ruixuan Huang,Kefan Zhang,Amy Xin,Yuanchun Wang,Shu Zhao,Evgeny Kharlamov,Jie Tang,Juanzi Li
类目: Artificial Intelligence (cs.AI)
备注: 9 pages, accepted by KDD 2026
Abstract:Large language models (LLMs) augmented with tools are emerging as autonomous agents capable of using Web engine, APIs, and code to solve complex, long-horizon tasks. Current tool-using benchmarks for information seeking on academic graphs rely on synthetic templates, simplified solution spaces, or narrow tasks such as paper-centric tasks, leaving key challenges underexplored - realistic user intent, complex multi-step API planning, rich parameter filling for APIs, grounded answers with references, and comprehensive evaluation of both the process and the outcome. We introduce AISE-Bench, a real-world, full-cycle annotated benchmark for information seeking on academic knowledge graphs. AISE-Bench release contains 1,133 QA pairs, including query taxonomies, full API execution trajectories, validated parameters, and source-grounded answers with reference links. To support high-quality annotation, we design a customized agent workflow to enable annotators to plan, execute, and revise complex API workflows efficiently. We develop a comprehensive evaluation protocol measuring answer quality, reference grounding, API-planning correctness, and execution success. Among the 14 evaluated methods, even the strongest model (PLAY2PROMPT with Gemini-3-Pro) achieves only moderate performance and often struggles with API planning and execution. AISE-Bench establishes a challenging new testbed for quantitatively evaluating and improving the stepwise correctness, grounded summarization, and traceable reasoning of multi-step API-using LLM agents. Our code and data are available at this https URL.
[AI-133] From Errors to Rules: Iterative Prompt Optimization for Text Classification
链接: https://arxiv.org/abs/2607.20497
作者: Yueying Cui,Renhao Xue,Yi Zhang,Mukul Prasad
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Prompt optimization for text classification spans diverse approaches, from demonstration selection to exploration-based search to error-driven diagnosis, each with known but incompletely characterized strengths and limitations. We conduct a comprehensive empirical study across diverse classification benchmarks (2 to 150 classes) comparing these paradigms through both quantitative evaluation and qualitative analysis of optimization traces, revealing that each paradigm excels on structurally different task types and that no single method dominates. Guided by these insights, we propose Error-Guided Optimization (ERGO), an error-driven method that iterates over the full training set in non-overlapping batches, diagnoses classification failures, and generates targeted decision rules through a diagnose-prescribe-rewrite feedback loop. ERGO achieves the best accuracy on tasks where errors concentrate in specific confused label pairs (which we term boundary-learnable tasks): TREC: 90.0%, CLINC150: 94.4%, converges in 3-5 iterations, and produces interpretable decision rules. While ERGO does not achieve the highest overall average, it fills a complementary role: demonstration-based ICL wins on coverage-dependent tasks, exploration-based search wins on many-class intent, and ERGO wins where decision boundaries are learnable from error patterns. We provide a complementarity framework linking task characteristics to optimal paradigm selection, offering practical guidance for practitioners.
[AI-134] Isolating LLM Alignment from Regex: Zero Coverag e and Metric-Dependent Divergence Under Adversarial Mutation
链接: https://arxiv.org/abs/2607.20494
作者: Alexandre Cristovão Maiorano
类目: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
备注: 15 pages, 2 figures, 7 tables
Abstract:Production LLM applications commonly stack a regex filter in front of model-side alignment; prior work found no measurable coverage gain from adding a live Gemini backend behind an active regex filter. We ask whether that ceiling holds when the corpus is \emphdesigned to bypass the regex. We introduce L_5 -no-regex – identical to L_4 -real (Gemini-2.5-flash, token-budget cap, rate limit, output scrub) but with the nine-pattern filter disabled – and evaluate it against N=45 adversarial probes across three sub-corpora (carry-forward, regex-bypass, alignment-isolate), amplified by Gemini paraphrase and PAIR to \sim1,555 probe-run pairs over N=5 replications. Under the primary substring classifier, H1 is refuted: L_5 block rate is 0,% across all five OWASP LLM Top-10 categories ( \Delta\textpp=0 vs.\ L_0 , p=1.00 ; Wilson upper bound 5,% ). A secondary LLM-judge metric on PAIR variants shows 56 – 100,% block rates ( p0.01 ), revealing alignment does respond to adversarially-framed probes – but produces refusals too nuanced for substring matching. The sub-corpus differential prediction is not supported ( p=1.00 ). Alignment’s contribution is \emphmetric-dependent: on natural-language harmful-request probes, it adds zero observed coverage beyond the regex; on adversarially-framed variants, an LLM judge detects refusals the substring classifier misses. The locked corpus, mutation artifacts, and export scripts are released for replication.
[AI-135] Attention-based Experience Replay Framework for Continual Learning of Agnostic Time Series Forecasting Models
链接: https://arxiv.org/abs/2607.20493
作者: Quentin Besnard(RFAI),Nicolas Ragot(RFAI)
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:Deep learning has led to remarkable progress in artificial intelligence, particularly in robotics, imaging and sound processing. However, a major limitation of neural networks remains their strong dependence on large and stationary datasets. In many real-world applications, these conditions are rarely met due to evolving and dynamic environments where data distributions change over time. Continual learning aims to address this challenge by developing models capable of adapting incrementally while maintaining a balance between stability and plasticity under computational constraints. In this work, we introduce a novel framework for continual time series forecasting, designed to extend existing static forecasting models commonly used in the literature by incorporating an Experience Replay strategy guided by Attention mechanisms. This approach allows the model to adapt dynamically to new contexts while preserving prior knowledge, effectively mitigating catastrophic forgetting. The framework is evaluated on standard forecasting benchmarks as well as on a piezometric dataset exhibiting diverse temporal behaviors. Results show that our approach effectively increases or maintains predictive performance over time while reducing retraining costs and data requirements, thus facilitating the deployment of forecasting models in dynamic and real-world settings.
[AI-136] CRAWO: Custom Resources for Adaptive Workload Orchestration
链接: https://arxiv.org/abs/2607.20490
作者: Eugênio Santos,Daniel Maia,Stefano Loss,José Manoel Silva,Aluizio Rocha Neto,Thais Batista,Everton Cavalcante,Nélio Cacho,Eduardo Nogueira,Daniel Araújo,Frederico Lopes
类目: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC); Emerging Technologies (cs.ET)
备注:
Abstract:Edge Intelligence has emerged as a key paradigm for enabling real-time applications in smart cities by shifting computation from centralized cloud data centers to the network edge, thereby reducing latency and bandwidth consumption. However, deploying Artificial Intelligence (AI) pipelines across heterogeneous edge infrastructures remains challenging due to the wide range of device capabilities, from low-power microcontrollers to accelerator-equipped systems. Existing edge orchestration platforms primarily focus on deployment automation and infrastructure management, but these approaches are often inefficient and limit the ability to adaptively allocate resources under dynamic conditions. To tackle these issues, this paper introduces CRAWO (Custom Resources for Adaptive Workload Orchestration), an architectural framework for coordinating AI pipelines across distributed edge environments. CRAWO follows a control-loop-based model that separates allocation intelligence from execution by managing placement decisions, state management, and inter-stage data flows while instantiating services on edge nodes. The framework incorporates a hardware-aware allocator with a pluggable multi-criteria decision layer that leverages real-time infrastructure metrics to enable adaptive workload placement. The reference implementation adopts a microservices architecture deployed on a lightweight Kubernetes distribution (K3s), using Custom Resource Definitions (CRDs) for domain modeling and a dedicated operator for state reconciliation. Evaluation in a vehicle surveillance scenario using license plate recognition demonstrates improved workload distribution and reduced reliance on centralized cloud processing in latency-sensitive environments.
[AI-137] EvoSQL: Memory-Augmented Critic-Generator Co-Evolution for Text-to-SQL
链接: https://arxiv.org/abs/2607.20489
作者: Jiawei Zhou,Jianwei Wang,Chenyu Zhou,Chaojian Shi,Ming Dong,Kai Wang
类目: Artificial Intelligence (cs.AI); Databases (cs.DB)
备注:
Abstract:Text-to-SQL has advanced rapidly with large language models, but complex database queries still require reasoning beyond one-shot generation, including multi-step decomposition, execution-based diagnosis, and targeted correction. We present EvoSQL, a co-evolution framework that formulates SQL synthesis as an iterative interaction between a generator and a critic. EvoSQL maintains a contextualized candidate memory, verifies SQL candidates with both execution signals and LLM-based critique, and updates its memory through utility-guided aggregation. To strengthen the underlying generator-critic pair, we further introduce a Self-Distillation Policy Optimization (SDPO) fine-tuning stage that injects execution-aware supervision into modern coding LLM backbones. Experiments on Spider and BIRD show that EvoSQL consistently improves open-source models over Maj@16 baselines, with particularly large gains on BIRD-Dev, ranging from +1.37% for Qwen3-4B to +9.19% for Qwen2.5-Coder-3B. SDPO initialization further improves selected backbones on Spider-Test and BIRD-Dev. These results suggest that memory-grounded co-evolution is an effective path toward more reliable and generalizable Text-to-SQL systems. Code is available at this https URL.
[AI-138] Autonomous Topology Mutation: Safe Runtime Restructuring for Multi-Agent LLM Systems with Capability State and Shadow Invariants
链接: https://arxiv.org/abs/2607.20488
作者: Bronislav Sidik,Chaya Levi,Nizzan Kimhi
类目: Artificial Intelligence (cs.AI)
备注: 9 pages, 5 tables. Code and benchmark harness available at this https URL
Abstract:Multi-agent LLM frameworks typically fix their team topology at boot time. When an individual agent becomes overloaded at runtime, for example by mixing too many action categories, accumulating tool errors, or queueing behind too many calls, the system has no mechanism to restructure itself. We introduce Autonomous Topology Mutation (ATM), a runtime team-mutation mechanism for multi-agent LLM frameworks. ATM combines telemetry-driven overload detection with three safety invariants that gate each structural change: capability monotonicity, state-routing completeness, and shadow-before-live validation. ATM monitors a six-signal Bottleneck Index that includes queue depth, context thrash, tool-error rate, role entropy, retry-loop rate, and cross-agent wait time. When a warmup-calibrated threshold is breached for multiple consecutive ticks, ATM factorises the overloaded agent into specialised sub-agents and hot-swaps the parent into a coordinator role while preserving its external identity. State transfer is controlled by privacy-level-aware routing: each memory atom is routed only to a permitted child set, or explicitly dropped with a logged reason. No candidate topology receives live traffic until it has passed a shadow validation window. On 720 DeepSeek-V3-driven task runs with deterministic tool stubs across four ablation conditions and three workloads, the ATM factoriser split lifts code-task success from 3.3% to 61.7%. The full rail-and-distillation system reduces detected high-privacy memory exposure under a regex classifier from 2.0 to 0.0 events per task while preserving task quality. The runtime rails carrying ATM’s invariants add less than 500 microseconds of p99 latency on the agent hot path. A small live-tool probe with real Python execution is included as an external-validity check. The implementation, benchmark harness, and traces are open-sourced. Comments: 9 pages, 5 tables. Code and benchmark harness available at this https URL Subjects: Artificial Intelligence (cs.AI) ACMclasses: I.2.11; C.2.4; D.2.11 Cite as: arXiv:2607.20488 [cs.AI] (or arXiv:2607.20488v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2607.20488 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-139] OPTScientist: Multi-Agent Discovery of Typed Optimizer Programs for Transformer Pretraining
链接: https://arxiv.org/abs/2607.20486
作者: Zhongzheng Li,Tiancan Feng,Wenhao Li,Qingsong Ran,Shikun Feng,Xiaoyuan Zhang,Yue Wang,Xiaoguang Zhao
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Designing optimizers for modern deep learning remains a challenging scientific problem, requiring the joint consideration of optimization geometry, state dynamics, numerical stability, implementation constraints, and empirical generalization. Existing automated optimizer discovery methods typically search either over unconstrained code spaces or within narrowly parameterized optimizer families. The former is flexible but often produces invalid or uninterpretable programs, while the latter is stable but limits novelty. We introduce OPTScientist, a theory-guided multi-agent framework for optimizer discovery in a typed domain-specific language (DSL). OPTScientist formulates optimizer design as a constrained scientific search process, where candidate updates are expressed through direction, scaling, preconditioning, regularization, state, and grouping modules. Four role agents, Theorist, Designer, Engineer, and Reviewer, collaborate within a single orchestration loop to propose hypotheses, synthesize DSL candidates, compile and evaluate optimizers, and critique results. To overcome the limitations of a fixed search space, OPTScientist combines evolutionary search over optimizer programs with a second-stage mechanism that proposes small DSL extensions when repeated failures reveal representational bottlenecks. Using this framework, we discover RS-MR, a reduced-state matrix optimizer that improves transformer pretraining over strong baselines under our native evaluation protocol. Our results suggest a path toward automated optimizer science grounded in theory, typed programs, compiler validation, and closed-loop experimentation.
[AI-140] he Devil is in the Spectrum: Mitigating Representation Collapse in LLM s via Topologically Regularized Side-Path ICML2026
链接: https://arxiv.org/abs/2607.20484
作者: Yiheng Tao,Kaiwen Cheng,Yao Lu,Chang Liu,Jie Chen
类目: Artificial Intelligence (cs.AI)
备注: 22pages, 4 figures, poster of icml 2026
Abstract:Large Language Models (LLMs) are fundamentally limited by representation collapse, a bottleneck that severely degrades long-context performance. We identify that existing approaches risk drifting into one of two pathological extremes: homogenization collapse (e.g., attention sinks causing rank deficiency) and isolation collapse (e.g., local attention causing context disconnection). Through spectral analysis of attention dynamics, we derive an intrinsic trade-off between mixing efficiency (spectral gap) and information capacity (effective rank) that standard mechanisms struggle to balance. To resolve this dilemma, we propose the Topologically Regularized Side-Path (TRSP), a non-invasive architectural intervention that achieves spectral balance. TRSP employs a parameter-free Triangular Box mechanism, scaled by a lightweight, length-aware gate, to regularize the token interaction topology. By integrating proximal coupling to preserve effective rank and distal propagation to support non-degenerate mixing, TRSP promotes a geometrically healthier transition operator without altering core attention. Experiments show significant improvements across general capabilities and long-context benchmarks. Notably, on NoLiMa at 8\times the training length, TRSP retains 83% accuracy and surpasses the Differential Transformer and Gated Attention by approximately 30 and 50 percentage points, respectively. Code available at: this https URL.
[AI-141] ractable Hierarchical Control of Autoregressive Language Models
链接: https://arxiv.org/abs/2607.20483
作者: Max Scribner,Antonio Vergari,Vaishak Belle
类目: Artificial Intelligence (cs.AI)
备注: 21 pages, 4 figures, 2 algorithms
Abstract:Constraining the generation of autoregressive large language models (LLMs) is an important component of integrating language models into formal systems. In the generation of code and data for tasks like program synthesis, ensuring that language models produce syntactically valid output is a prerequisite for processing such output. These languages (such as SQL or JSON) are often designed as LR(k) context-free grammars. By distilling the LLM to a tractable probabilistic model, its autoregressive generation can be steered and masked to incorporate the probability of satisfying logical constraints, ensuring high quality output that is guaranteed to be valid. This paper demonstrates that the satisfaction of any LR(k) grammar of finite duration can be calculated in polynomial time, an improvement over the exponential time of applying previous methods to such grammars. This result enables efficient constraint and steering of LLM generation towards output that better satisfies formal syntactic constraints.
[AI-142] Routing Without Training: Controllable-Ratio LLM Offloading via Reliability Gating
链接: https://arxiv.org/abs/2607.20481
作者: Evan Chen,Shiqiang Wang,Kevin S Chan,Su Wang,Christopher Brinton
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Local-cloud collaboration is a practical way to deploy large language models under resource constraints, but existing methods often rely on trained routers or collaboration-aware finetuning that tie routing behavior to a particular operating regime. In this work, we show that such training may be unnecessary: the local model’s own inference-time agreement across sampled responses already provides a strong signal for deciding when to trust local execution and when to offload to a stronger cloud model. We propose CARGO, a training-free routing framework that estimates this agreement through prompt-varied sampling, applies Bayesian early stopping for sample-efficient uncertainty control, and supports arbitrary target collaboration ratios through lightweight deployment-time calibration. Across diverse reasoning and question-answering tasks, multiple local LLM families and scales, and both pretrained and finetuned local models, CARGO consistently outperforms other training-free baselines and in several settings surpasses supervised learned routers. These results suggest that effective and adaptable local-cloud collaboration can emerge directly from the local model’s intrinsic response behavior, without requiring an additional trained router.
[AI-143] Enabling Scalable Topology Inference in Distribution Systems via Constrained Multi-Source Inference
链接: https://arxiv.org/abs/2607.20480
作者: Haoran Li,Lihao Mai,Muhao Guo,Jiaqi Wu,Yang Weng
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Accurate distribution system topology is essential for outage localization, voltage analytics, and operation of distribution grids, yet maintaining reliable connectivity records remains challenging in practice due to heterogeneous and imperfect utility data. Existing topology identification methods often rely primarily on electrical similarity or spatial records alone, which become unreliable in dense feeders and under inconsistent metadata conditions. This paper formulates distribution topology identification as a constrained inference problem that refines a utility-provided base topology using heterogeneous evidence while enforcing spatial feasibility and physical operational constraints. Instead of reconstructing connectivity from scratch, the proposed framework detects inconsistent assignments, performs localized reconnection within constrained neighborhoods to ensure scalability, and iteratively enforces physical feasibility to produce operationally consistent topology estimates. In addition, a falsification-driven reliability metric evaluates how strongly each inferred connection is supported relative to alternative feasible assignments, enabling utilities to prioritize verification efforts while preserving system-wide observability. The framework is validated using operational data from three feeders comprising more than 8,000 AMI meters in collaboration with a large U.S. utility. Results demonstrate over 95% topology reconstruction accuracy while significantly reducing computational effort compared with global inference approaches. The study further shows that correlation-based methods alone produce ambiguous assignments in dense urban feeders, whereas combining electrical measurements with spatial and operational constraints enables robust and scalable topology recovery under realistic deployment conditions.
[AI-144] Verifier-First Evaluation of Agent ic LLM s for Infrastructure-as-Code Generation
链接: https://arxiv.org/abs/2607.20478
作者: Mohamed Jouini
类目: oftware Engineering (cs.SE); Artificial Intelligence (cs.AI)
备注: 26 pages, 3 figures, 17 tables. Benchmark dataset available at this https URL
Abstract:Infrastructure-as-Code (IaC) generation from natural language requires satisfying provider schemas, dependency planning, and organizational policy constraints, not merely producing syntactically plausible configurations. We present a verifier-first empirical study of seven agentic strategies for Terraform generation evaluated on IaC-Eval v2, a modernized 186-task AWS/Terraform benchmark with Rego v1 intent policies. Our evaluation separates failures into three verifier stages (terraform validate, terraform plan, opa eval) and applies McNemar’s test with Wilson confidence intervals on all pairwise comparisons (n=186, alpha=0.05). We report five principal findings. (1) Active retrieval via ReAct agents with MCP or ChromaDB-backed RAG raises Qwen2.5-Coder 7B from 14.0% to 45.7% pass@1 (p0.0001), primarily by reducing VALIDATE_FAIL from 144 to 66 tasks. (2) Iterative refinement with verifier feedback achieves 62.9% (Qwen 7B) and 84.4% (GPT-4o) pass@1, exhibiting binary convergence – tasks either resolve in one retry or exhaust the budget. (3) GEPA reflective instruction optimization raises the Active RAG baseline by +7.5 pp (p=0.026) using only 80 verifier-guided rollouts, providing evidence that prompt optimizers can improve verifiable IaC generation without weight updates. (4) SIMBA teacher-free demonstration injection achieves performance equivalent to Active RAG (p=1.0) without retrieval infrastructure, but fails to address the dominant SELF_DEFINED_PROPERTY error class (50% of failures). (5) A diagnostic Rego-injection experiment shows that 79% of post-refinement OPA failures are information-gap failures resolvable when policy text is visible (p=0.016), motivating policy Comments: 26 pages, 3 figures, 17 tables. Benchmark dataset available at this https URL Subjects: Software Engineering (cs.SE); Artificial Intelligence (cs.AI) ACMclasses: I.2.2; D.2.5 Cite as: arXiv:2607.20478 [cs.SE] (or arXiv:2607.20478v1 [cs.SE] for this version) https://doi.org/10.48550/arXiv.2607.20478 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-145] Semi-Supervised Text-Attributed Graph Distillation KDD2026
链接: https://arxiv.org/abs/2607.20477
作者: Yurui Lai,Samir Moustafa,Renchi Yang,Tsz Nam Chan
类目: Artificial Intelligence (cs.AI)
备注: Technical report for the paper “Semi-Supervised Text-Attributed Graph Distillation” accepted KDD2026
Abstract:\em Text-Attributed Graphs (TAGs) have emerged as an expressive data model for integrating graph topology with rich textual semantics. Existing representation learning methods over TAGs suffer from severe scalability bottlenecks, particularly together with \em Large Language Models (LLMs). While data distillation offers a promising data-centric solution, existing methods fail to capture the complex interplay between graph and text modalities, struggle with the label scarcity inherent in semi-supervised settings, and lack the ability to produce the human-readable textual attributes required for downstream LLM-based tasks. To address these challenges, we propose \algo, a unified semi-supervised framework guided by the \em Wasserstein Distance (WSD). Grounded in our empirical findings on real TAGs, \algo introduces a graph-text collaborative encoding module that utilizes dual-pathway encoders (graph-aware and -free) within a collaborative self-training scheme to harvest reliable pseudo-labels and fuse complementary graph-text features. Furthermore, we develop a theoretically grounded WSD-based graph sketching algorithm and a cost-effective LLM text synthesis module, which leverages cluster-based keyword extraction to generate coherent, human-readable summaries for condensed nodes. Extensive experiments on benchmark datasets demonstrate that \algo achieves a state-of-the-art performance-compression trade-off in terms of both GNN- and LLM-based downstream tasks, enabling effective and efficient TAG learning or analytics. Comments: Technical report for the paper “Semi-Supervised Text-Attributed Graph Distillation” accepted KDD2026 Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2607.20477 [cs.AI] (or arXiv:2607.20477v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2607.20477 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-146] VeriSimpl: Robust Optimization Modeling from Natural Language using Simplification-based Verification ICML2026
链接: https://arxiv.org/abs/2607.20474
作者: Sumaya Abdul Rahman,Seckhen Ariel Andrade Cuellar,Ghani Raissov,Mohammad Raza
类目: Artificial Intelligence (cs.AI)
备注: ICML 2026
Abstract:Natural language interfaces can greatly benefit the accessibility and usability of optimization modeling, and recent advances in large language models (LLMs) show promise in automatically translating textual problem descriptions into executable solver formulations. However, a key challenge for existing approaches is to ensure that the inferred formulation correctly implements the intended task, even if it may execute without errors. We introduce VeriSimpl, a solver LLM framework for robust natural-language-to-optimization formalization. Our approach is based on the idea of simplification-based verification, where the optimization solver is leveraged to generate simplified diagnostic queries about a candidate formulation to allow the LLM to tractably reason about the correctness of the formulation with respect to the task description. We present such simplification strategies along different dimensions with respect to problem constraints and decision variables, which allow the LLM to reason locally under fixed global contexts. Evaluations on a range of optimization benchmarks show how our approach provides consistent improvements in accuracy over existing methods, while also providing a novel high-precision self-verification signal.
[AI-147] Incomplete Prompt Jailbreaks in Large Language Models ACL2026
链接: https://arxiv.org/abs/2607.20473
作者: Yeonjea Kim,Bumjin Park,Jaesik Choi
类目: Artificial Intelligence (cs.AI)
备注: Accepted to ACL 2026 Findings. 15 pages (9 pages for main body), 13 figures
Abstract:Large language models (LLMs) are increasingly released as open-weight models with safeguards against harmful requests. Nevertheless, sentence completion remains vulnerable to incomplete harmful prompts. In this work, we formalize this phenomenon as incomplete prompt jailbreaks (IPJ) and provide a systematic empirical characterization of when and how incomplete prompts elicit harmful continuations. We analyze diverse attractor types associated with incomplete sentence continuation and show that LLMs systematically delay refusal until sentence termination. We further demonstrate that training models to refuse incomplete harmful prompts via parameter tuning is insufficient, failing to generalize across both content domains and attractor types. To enable fine-grained control, we identify two functional neurons: termination and continuation neurons. By clarifying their roles in sentence completion, we highlight the potential of neuron-level interventions for more precise and robust IPJ defenses.
[AI-148] Robust Critics: Defending LLM s Against Multi-Turn Attacks
链接: https://arxiv.org/abs/2607.20472
作者: Roman Belaire,Arunesh Sinha,Pradeep Varakantham
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:When a user asks a language model something harmful, is it a genuine attack or a misunderstood but well-meaning question? This ambiguity is one of the central challenges of LLM safety. A model that assumes the worst harms legitimate users; one that assumes the best is easily exploited. The problem is compounded in multi-turn dialogue, where an attacker’s true intent may only reveal itself gradually across many exchanges, yet existing safety frameworks apply a contextual bandit treatment, ignoring the trajectory of the conversation. To that end, we propose Dialogue Critic Guided Sampling (DCGS), a framework that addresses this by inferring user intent at every turn of dialogue. Instead of applying a fixed rule about what is or is not safe, DCGS learns what the user’s intent is likely to be based on the full conversational history and generates responses accordingly. Formally, we model adversarial dialogue as a Markov Decision Process and learn value and regret-based critics at both the individual token and utterance (full response) levels, scoring candidate responses via an action-value critic. We prove that this inference-time reweighting approximates exponential tilting of the base policy, guaranteeing improvement in expected return for any finite candidate pool, a property that group-relative objectives do not exhibit. Evaluated on CARES-18k, WildJailbreak, Redbench, and Harmbench, DCGS outperforms strong robust baselines and frontier models on adversarial dialogue tasks. DCGS also transfers to frontier models, improving their robustness without fine-tuning. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2607.20472 [cs.AI] (or arXiv:2607.20472v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2607.20472 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[AI-149] Benchmarking the Personalization Capabilities of Large Language Models
链接: https://arxiv.org/abs/2607.20471
作者: Ashutosh Srivastava,Siddharth Yedlapati,Vinay Aggarwal,Yaman Kumar Singla,Shashwat Dixit,Jitendra Ajmera,Balaji Krishnamurthy
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Personalization, the act of varying a message to induce action from a specific receiver while keeping sender, channel, and time fixed, has a long tradition in psychology and marketing as a two-party problem in which sender and receiver have independent objectives. Large language models remove the bounded-inventory constraint of classical retrieval-and-ranking approaches by generating a continuum of message variants conditioned on inferred receiver state, raising the question of how well current models perform personalization in the classical sense. Existing LLM personalization benchmarks measure sender-side adaptation, in which the receiver is the same user the model is serving. The two-party question, whether a generated message induces its intended action in a third party, has been investigated only through A/B tests and small-scale human studies that cannot be re-run against a new model on demand. We adapt the Bayesian Persuasion framework of Kamenica and Gentzkow (2011) to generative agents and instantiate the formulation in sales, where receiver actions are routinely logged against the outreach that induced them. We release SDR-Bench, a public corpus of 6,279 customer success stories spanning 22 industries and approximately 200 enterprises, served through a temporally constrained simulation that prevents future-data leakage. Across frontier LLMs and deep-research agents, we observe a consistent personalization plateau and on a Fortune 100 tech cohort no model statistically separates successful from unsuccessful outreach. A field deployment with 12 professional sales representatives validates the framework, with 48 percent of model-generated content rated immediately useful and senior-expert agreement at Pearson 0.82. We release SDR-Arena and SDR-Bench publicly to support reproducible study of generative personalization at scale.
[AI-150] PlanE: Meta Planning of Data Tuning and Inference for Extractive-based LLM s
链接: https://arxiv.org/abs/2607.20470
作者: Jiacheng Wang,Weiyan Zhang,Guangya Yu
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Enhancing the task-specific capabilities of Large Language Models (LLMs) primarily requires substantial instruction-tuning datasets. However, the sheer volume of such data imposes a considerable annotation cost, and a lack of optimization methods for tailoring LLMs to specific tasks. To address the above issues, we propose a \textbfPlanning framework for constructing \textbfExtractive-based LLMs called \textbfPlanE, which includes data decomposition, instruction tuning, and prompt inference. Additionally, we introduce a Data-Tuning-Inference (DTI) planner, aimed at selecting the optimal base-LLM and its DTI combinations for specific datasets to improve construction efficiency. The experimental results demonstrate the effectiveness of our PlanE from two views: (1) across different datasets using the same base-LLM, and (2) on the same dataset using different base-LLMs. Furthermore, we validate the generalizability of the proposed DTI planner under different optimization objectives. The codes are publicly available at this https URL.
[AI-151] DecodeShare: Tracing the Shared Subspace of LLM Decode-Time Decisions
链接: https://arxiv.org/abs/2607.20469
作者: Zishan Shao,Lixun Zhang,Kangning Cui,Yixiao Wang,Ting Jiang,Hancheng Ye,Qinsi Wang,Zhixu Du,Yuzhe Fu,Fan Yang,Danyang Zhuo,Yiran Chen,Hai Helen Li
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) handle many tasks with one set of parameters, but under KV-cached inference it is unclear what task-general structure, if any, is used at decode time rather than during prefill. We propose DecodeShare, a protocol that identifies a low-dimensional subspace consistently shared across tasks in decode-time hidden states, and then tests its causal role by removing that subspace only during decoding. In our experiments, disturbing the discovered shared subspace degrades decision performance far more than disturbing either a prefill-derived or random subspace under the same intervention budget. We further show this decode-shared subspace has practical consequences for activation steering: common steering directions can overlap the task-general decode channel. Projecting out this shared subspace directly separates the functional roles of the two components, while evaluating steering vectors at decode-time yields more reliable signal for downstream deployment than prefill-based proxies. Despite its compactness, the shared subspace can serve as a high-leverage causal channel at decode time. Code is available at: this https URL.
[AI-152] InferenceBench: A Benchmark for Open-Ended LLM Inference Optimization by AI Agents
链接: https://arxiv.org/abs/2607.20468
作者: Jehyeok Yeon,Ben Rank,Maksym Andriushchenko
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:AI agents are increasingly used to automate research and development tasks, yet existing benchmarks typically evaluate them on prescribed workflows or narrow action spaces. Even nominally open-ended tasks can often be solved by retrieving a well-known recipe and tuning a few hyperparameters, making it unclear whether strong results reflect genuine optimization or memorized solutions. We introduce InferenceBench, where an agent must deploy an OpenAI-compatible inference server and optimize the speed of LLM inference. Each agent receives a target LLM, one H100 GPU, an optimization scenario, and a wall-clock time budget of two hours. Three optimization scenarios isolate distinct bottlenecks of inference (prefill latency, decode latency, and concurrent request throughput) and a fourth balances all three at the same time. Across 15 frontier agent configurations, agents reliably improve over a naive PyTorch baseline (up to 8.08\times ) and often match or exceed serving engines with default settings ( 4.05\times for vLLM), but still fall below a simple hyperparameter search under the same time budget (up to 11.53\times ). Qualitative analysis of agent trajectories shows that although agents enumerate many relevant optimization techniques, they overwhelmingly converge on a single inference framework. They test only a few distinct configurations and spend the remaining budget re-measuring, repairing, or optimizing hyperparameters rather than exploring substantially different strategies. This suggests the bottleneck is not domain knowledge, but the ability to propose diverse configurations, evaluate them systematically, and submit the best identified solution. Overall, InferenceBench reflects the ability of agents to operate in an open-ended AI engineering setting, where memorized solutions lead to limited improvements.
[AI-153] DC-Leap: Training-Free Acceleration of dLLM s via Draft-Guided Contiguous Leaping Decoding
链接: https://arxiv.org/abs/2607.20467
作者: Yanhua Jiao,Tianyi Wu,Xiaoxi Sun,Yulin Li,HuiLing Zhen,Libo Qin,Baotian Hu,Zhuotao Tian,Min Zhang
类目: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注:
Abstract:While parallel decoding is central to the efficiency of Diffusion Large Language Models (dLLMs), current strategies are often hindered by overly conservative confidence thresholds. These thresholds, necessitated by the Joint Probability Dependence Error (JPDE), result in redundant denoising iterations and suboptimal inference speeds. To overcome this, we propose DC-Leap, a training-free framework that enables reliable acceleration of dLLMs in the moderate-confidence regime. DC-Leap introduces a Dynamic Contiguous Verification strategy that integrates strictly-ordered causal constraints into the parallel decoding process. By progressively validating token dependencies, this mechanism effectively neutralizes the JPDE, enabling reliable acceleration with comparable performance. Furthermore, DC-Leap incorporates the draft-guided decoding mechanism, where the draft helps extend the context by leaping forward across multiple tokens, providing look-ahead context and retaining the structural benefits of bidirectional attention during inference. Extensive experiments on standard benchmarks demonstrate that DC-Leap achieves substantial speedups, up to 53.19x on MBPP for long-sequence generation, and up to 105.02x when combined with KV-Cache with comparable generation quality. Code is available at this https URL .
[AI-154] JAXBench: Benchmarking Autonomous TPU Kernel Optimization
链接: https://arxiv.org/abs/2607.20466
作者: Arya Tschand,Charles Hong,Julian Walker,Nina Cai,Shangkun Wang,Suvinay Subramanian,Sundar Dev,Vijay Janapa Reddi,Amir Yazdanbakhsh,Sethu Sankaran
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Rigorous benchmarks have driven progress in autonomous GPU kernel performance optimization by establishing a shared target to hillclimb on, but no equivalent exists for TPUs. We present JAXBench, a TPU-native benchmark suite for AI-generated kernel optimization on Google Cloud TPUs. JAXBench comprises 50 JAX workloads that are both relevant and provide headroom for optimization. We extract 17 production ML operators from architectures in the public MaxText library such as Llama-3.1, DeepSeek-V3, Mixtral, Mamba-2, and AlphaFold2, and translate 33 operators from KernelBench that are validated for correctness and set with new problem sizes that achieve high TPU v6e MXU utilization. Eight of the 17 production operators ship with hand-optimized Pallas kernels from the public Tokamax library and block-size tuned to establish an expert upper-bound baseline. We evaluate four feedback-driven methods on generating candidate Pallas kernels for JAXBench. Across the full suite with Gemini 3 Flash, we find that target-specific context matters more than model scale on a sparsely-documented DSL like Pallas. Conditioning on curated TPU documentation raises per-sample correctness from 5.8% to 37.3% and solves 48 of 50 benchmarks at a 1.28x geomean speedup. Search structure yields significant gains once correctness is achieved, with Autocomp’s beam-search pipeline reaching a 1.36x geomean speedup over XLA. On the 8 hand-tuned kernels, Autocomp reaches 1.60x geomean over XLA, recovering most of the 2.08x Tokamax upper bound but trailing on the specialized paged and ragged attention operators. High-quality TPU kernel optimization remains a challenging task, and we release the JAXBench benchmark, evaluation harness, and baseline results to support open source contributions.
[AI-155] Stochastic Sampling is Epistemically Shallow: The Dimensionality Gap Between Temperature Variation and Model Diversity in LLM s ICML2026
链接: https://arxiv.org/abs/2607.20464
作者: Izhar Ali
类目: Artificial Intelligence (cs.AI)
备注: 8 pages, 4 figures, 3 tables. Accepted at EIML@ICML 2026
Abstract:When a language model gives different answers on repeated runs, does that variation reveal what it does not know? Self-consistency turns the variation into a per-question uncertainty estimate via majority voting. But does the same variation reveal cross-question structure – related questions flipping together, the way a diverse ensemble does? We compare two regimes on the same questions: one model run 100 times at \tau=1 versus an ensemble of 24 LLMs run once each at \tau=0 . A Marchenko–Pastur random-matrix test separates signal from sampling noise on both sides. Within any single model, at most one dimension rises above noise across five families and three benchmarks (MMLU, HellaSwag, GSM8K). Across the ensemble, four eigenvalues clear the noise edge, while a matched-difficulty Bernoulli null produces at most one in 500 Monte Carlo draws. Self-consistency gives accurate per-question uncertainty but no detectable cross-question structure; only a diverse ensemble surfaces what a model does not know.
[AI-156] ClickGuard: Detecting and Spoiling Clickbait News with Informativeness Measures and Large Language Models
链接: https://arxiv.org/abs/2607.20463
作者: Wojciech Michaluk,Tymoteusz Urban,Mateusz Kubita,Soveatin Kuntur,Anna Wróblewska
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:This paper presents an AI-driven browser extension that identifies clickbait to help users avoid misleading Internet articles. Moving beyond traditional detection, the application employs a hybrid machine learning architecture that combines transformer-based embeddings with linguistically motivated features and a custom “baitness” score. After evaluating various natural language processing techniques – from classic vectorizers to large language model (LLM) embeddings – an XGBoost-based model was developed that achieves an F1-score of 91% on the open combined dataset. Most importantly, the tool can warn users before and after they access a clickbait article. After opening an article, the user receives a percentage score indicating the likelihood that it is clickbait. The prediction is explained based on the analyzed metrics, including those specifically developed within the proposed system. The browser extension also provides a clickbait spoiler – a one- to two-sentence summary of the entire article. Demo video:this https URLthis https URL
[AI-157] Marking the Wrong Symptoms: Evaluating LLM Watermarks in Medical Texts
链接: https://arxiv.org/abs/2607.20462
作者: Melanie Rieff,Robin Staab,Thibaud Gloaguen,Stefan Hegselmann,Martin Vechev
类目: Artificial Intelligence (cs.AI)
备注:
Abstract:Large language models (LLMs) are increasingly integrated into clinical workflows, stressing the need for reliable traceability of model-generated output with watermarking. Yet, most watermarks are evaluated on general-purpose benchmarks, leaving domains like medicine, where small token-level perturbations can result in significant semantic changes, underexplored. In this work, we present the first rigorous study of how LLM watermarks affect medical performance, benchmarking 5 watermarking schemes across 11 LLMs and 7 VLMs on various tasks spanning unimodal and multimodal clinical reasoning. Importantly, we complement existing evaluations by introducing a human-expert-validated pipeline for systematically auditing medical reasoning quality, terminological precision, and induced hallucinations. Our results reveal that watermarking can induce substantial degradation across multiple failure modes, including lexical corruption, hallucinated terminology, and amplified misattribution or omission of image findings. Notably, we find that the absence of domain-specific analyses, combined with aggregate metrics that miss failures inherent to clinical text, can systematically obscure practical watermark-induced degradations. Our findings establish domain-specific evaluation as a prerequisite for the safe deployment of watermarked models in medicine, where current benchmarks can otherwise mask clinically consequential failures.
[AI-158] AINTMA: Agent ic AI Architecture for Autonomous Test Management with Generative Intelligence Secure Cloud Communication and Adaptive Quality Analytics
链接: https://arxiv.org/abs/2607.20452
作者: Vinil Pasupuleti,Shyalendar Reddy Allala,Siva Rama Krishna Varma Bayyavarapu,Shrey Tyagi,Srinivasateja Songa
类目: Artificial Intelligence (cs.AI)
备注: 11 pages, 2 figures, 4 tables, Submitted to AICCONS (AIP Conference Proceedings format)
Abstract:Modern software quality assurance demands intelligent, autonomous systems capable of adaptive decision-making across distributed cloud environments. This paper presents AINTMA (Agentic Intelligent Test Management Architecture), a multi-agent agentic AI system that transforms traditional test management into an autonomous quality intelligence ecosystem. AINTMA deploys six specialized AI agents (Test Discovery, Risk Assessment, Reinforcement Learning Prioritization, Execution Orchestration, Generative Quality Intelligence, and Cloud Security Monitor) coordinated through a secure multi-agent communication framework over a cloud-native microservices infrastructure. The Generative Quality Intelligence agent employs large language models to produce plain language quality narratives, defect risk summaries, and data-augmented test recommendations. The RL Prioritization agent models test selection as a Markov Decision Process, learning contextual policies from large-scale historical test execution data (47 features, rolling 36-month window). Secure cloud communication is enforced through a zero-trust API gateway with OAuth2/JWT authentication, encrypted inter-agent messaging, and multi-tenant isolation. Evaluation across 12 heterogeneous software projects over 18 months demonstrates: 88.4% test prioritization accuracy (APFD, vs. 51.2% random, 82.1% best commercial baseline); 43% test cycle time reduction; defect escape rate reduced from 8.3% to 2.1%; 340% ROI at 9-month payback. The agentic architecture scales to 50,000+ test cases with sub-400ms response time, and the generative intelligence module achieves 4.3/5.0 developer usefulness rating. AINTMA demonstrates that agentic AI, combining autonomous multi-agent coordination, generative intelligence and secure smart connectivity, can fundamentally advance software quality management in cloud-scale enterprise environments.
[AI-159] hrough-the-Earth Magnetic Induction Communication and Networking: A Comprehensive Survey
链接: https://arxiv.org/abs/2510.14854
作者: Honglei Ma,Erwu Liu,Wei Ni,Zhijun Fang,Rui Wang,Yongbin Gao,Dusit Niyato,Ekram Hossain
类目: ystems and Control (eess.SY); Artificial Intelligence (cs.AI)
备注: This work has been accepted by the IEEE Communications Surveys Tutorials (COMST) for publication. The final published version will be available on IEEE Xplore
Abstract:Magnetic induction (MI) communication (MIC) has emerged as a promising candidate for underground communication networks due to its excellent penetration capabilities. Integration with Space-Air-Ground-Underground (SAGUI) networks in next-generation mobile communication systems requires a well-defined network architecture. A recent discovery in MIC research, MI fast fading, remains in its early stages and presents unique challenges. This paper provides a comprehensive survey on through-the-earth (TTE) MIC, covering MI applications, channel modeling, point-to-point MIC design, relay techniques, network frameworks, and emerging technologies. We compare various MIC applications to highlight TTE-specific challenges and review the principles of channel modeling, addressing both MI slow fading and MI fast fading, along with its potential impact on existing MIC theories. We conduct a fine-grained decomposition of MI channel power gain into four distinct physical parameters, and propose a novel geometric model to analyze MI fast fading. We also summarize MI relay techniques, examine crosstalk effects in relay and high-density networks, and explore key research tasks within the OSI framework for a holistic MI network protocol in SAGUI. To bridge the gaps identified, we propose a MIC framework that supports TCP/IP and Linux, enabling full implementation of existing and emerging MIC solutions. This framework empowers researchers to leverage Linux resources and deep learning platforms for accelerated development of MIC in SAGUI networks. Remaining research challenges, open issues, and promising novel techniques are further identified to advance MIC research.
[AI-160] Barzilai-Borwein Fails Superlinear Convergence on an Open Set of Quadratics for Every Dimension ngeq 4
链接: https://arxiv.org/abs/2607.21579
作者: Dawei Li,Xiaotian Jiang,Mingyi Hong
类目: Optimization and Control (math.OC); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
备注: 31 pages, 3 figures
Abstract:Barzilai–Borwein (BB) method has shown strong practical performance in continuous optimization, yet its convergence dynamics remains poorly understood. In particular, a central unresolved question is whether BB converges superlinearly for almost every strictly convex quadratic problem and initialization. We provide a negative answer to this question. Specifically, for every finite dimension n\geq4 , we construct a nonempty open, hence positive-Lebesgue-measure, family of strictly convex quadratic problems and initial points for which the long Barzilai–Borwein method (BB1) converges but cannot converge root-superlinearly. More precisely, with the explicit constants \rho_\min=10^-6,\rho_\max=0.61 , every spectral component of the gradient is bounded above and below by the corresponding geometric sequence. Consequently, the gradient norm and the energy norm of the error satisfy two-sided geometric estimates with the same rates, while the objective gap satisfies the corresponding estimates with squared rates. In particular, all three quantities are bounded below by geometric sequences, ruling out superlinear convergence. The construction is highly nontrivial, based on a computer-assisted proof of a nonresonant, attracting seven-cycle of the projectivized BB dynamics in dimension four.
[AI-161] Cycle-Consistent and Uncertainty-Aware Neural Surrogates for Tokamak Edge Plasmas
链接: https://arxiv.org/abs/2607.21407
作者: Abdourahmane Diaw,Sebastian De Pascuale,Jae-Sun Park,Ivan Paradela Perez,Jeremy D. Lore,Stefan Dasbach
类目: Computational Physics (physics.comp-ph); Artificial Intelligence (cs.AI); Plasma Physics (physics.plasm-ph)
备注:
Abstract:The boundary and divertor plasma govern how a tokamak exhausts power and particles, setting heat fluxes, target conditions, and the onset of detachment. Predicting these quantities is essential for operating current and future devices, but edge simulations that resolve them are too slow for parameter scans, optimization, or real-time control. Machine-learning surrogates offer a fast alternative, yet most are forward-only: they cannot recover input parameters from observations or assess the reliability of their predictions. We introduce a cycle-consistent neural surrogate for edge plasmas, combining a conditional U-Net forward model with an optimization-based inverse method built on the frozen forward network. The forward model maps five control parameters to two-dimensional plasma-state fields on the SOLPS-ITER mesh; the inverse method enforces consistency between forward and inverse predictions, a self-supervised quality check needing no ground-truth labels at inference. An ensemble of multilayer perceptrons also predicts electron temperature and density profiles at the outboard midplane and divertor targets, with uncertainty estimates that flag where more simulations are needed. The forward model achieves normalized root-mean-square errors below 2.6% and Pearson correlations above 0.95 for all fields. Cycle-consistency regularization raises the average cyclical R^2 from 0.59 to 0.99 without degrading forward accuracy and enables recovery of the core fueling rate; all five control parameters are recovered with Pearson r\ge0.97 . A k -d tree warm start yields a database completion rate above 95%, versus roughly 30% outright failures when cold-started. With about 4\times10^6 parameters, the model produces full 2D predictions in milliseconds, five to six orders of magnitude faster than SOLPS-ITER, enabling real-time control, parameter scans, uncertainty analysis, and digital twins.
[AI-162] ransition-Related Potentials as Markers of Narrative Comprehension in Continuous EEG
链接: https://arxiv.org/abs/2607.20720
作者: Bálint Csanády,Péter Vedres,Kristóf Zsolt Makó,Orsolya Papp-Zipernovszky,Márta Volosin,Dávid Apagyi,András Lukács,András Bálint Kovács,Zoltan Nadasdy
类目: Neurons and Cognition (q-bio.NC); Artificial Intelligence (cs.AI)
备注: 40 pages, 14 figures
Abstract:Harnessing the potential of electroencephalography (EEG) for brain research is fundamentally limited by intrinsic noise and the diffuse projection of brain-generated activity over the scalp. The standard event-related potential (ERP) paradigm addresses this limitation by relying on repeated independent trials, albeit at the cost of moving away from naturalistic experimental conditions. As a more naturalistic alternative, we collected continuous EEG while participants watched short films and extracted potentials aligned to sharp cinematic transitions (cuts). We demonstrate that such transition-related potentials (TRPs) exhibit canonical ERP-like temporal structure associated with significant information processing. By comparing coherent films with scene-scrambled versions containing matched post-cut sensory input, we find that these responses are systematically shaped by narrative context. We then show that the cut-related EEG signature can be recovered directly from group-averaged continuous recordings with a compact deep neural network (DNN). The detector generalized across films and subject groups, and the resulting TRPs reproduced the main context-dependent effects observed for manually annotated cuts. These results indicate that narrative context leaves a measurable signature in EEG responses, that this signature can be detected directly in continuous recordings, and that such detections provide a semi-automated framework for analyzing how viewers process and understand film narratives. We propose that the method outlined here can be adapted to parse EEG responses to other forms of continuous stimulation, providing a general tool for probing experimental conditions that are closer to natural human experience.
[AI-163] Foundation-model-guided radiogenomic discovery linking cancer genomes to cancer scans
链接: https://arxiv.org/abs/2607.20583
作者: Frederik Hauke,Jeremias Krause,Patrick Wienholt,Christiane Kuhl,Ingo Kurth,Sikander Hayat,Jakob Nikolas Kather,Sven Nebelung,Daniel Truhn
类目: Genomics (q-bio.GN); Artificial Intelligence (cs.AI)
备注:
Abstract:The function of many genes is still unknown, and conventional driver-discovery methods, which rely on how frequently a gene is mutated, cannot assess genes that are only rarely affected. Here we pair Evo~2-based genome analysis with routine clinical imaging to identify gene–phenotype associations at genome-wide scale. For every somatic mutation across three TCGA cohorts (cRCC=clear cell renal cell carcinoma, HCC=hepatocellular carcinoma, and BC=breast cancer; n = 340 total), Evo~2 predicts a severity score, with no task-specific training. Per-gene severity summaries are then correlated with radiomic features extracted from paired tumor segmentations, controlling for total mutation burden. In TCGA-cRCC ( n = 162 ), this sweep recovers established renal-cancer drivers and identifies 46 additional genes reaching false discovery rate (FDR) significance absent from curated cancer-gene panels, several of which are Mendelian ciliopathy and cytoskeletal-disease genes. These results demonstrate that pairing a genomic language model with widely available clinical imaging can serve as a hypothesis-free discovery tool for gene–imaging associations invisible to conventional approaches.
机器学习
[LG-0] Expanding Flow Maps
链接: https://arxiv.org/abs/2607.21585
作者: Sophia Tang,Pranam Chatterjee
类目: Machine Learning (cs.LG)
*备注:
Abstract:Flow-based generative models have enabled remarkable progress in fast and controllable generation across continuous and discrete state spaces, yet existing parameterizations are constrained to fixed dimensions or fixed sequence lengths. Here, we introduce Expanding Generative Flows (EFlows), which define flows between distributions of increasing dimensionality along an expanding interpolant that grows the state by augmenting it with conditional noise. Building on this construction, we propose Expanding Flow Maps (EFMs), a new class of flow maps that distill the expanding interpolant into efficient few-step generative models. Each EFM factors the map between any two timesteps into two learnable operations: an expand operator, which augments the state space with new coordinates or tokens conditioned on the current state, and a transport map, which pushes the expanded state forward along the interpolant. Composing these operators yields a single map that jointly expands and denoises the state, recovering existing fixed-canvas flows and flow maps as the special case in which the expand operator is the identity. We further extend the framework to the discrete simplex, enabling variable-size graph generation and variable-length sequence generation. Across both continuous and discrete modalities, we establish EFlows and EFMs as a principled framework for settings in which output size is itself a learned, controllable degree of freedom.
[LG-1] Graph Learning on Ensembles of Cyclic Peptides: An Investigation of Molecular Ensemble Modeling ICML’26
链接: https://arxiv.org/abs/2607.21561
作者: Aaron Feller,Kris Deibler,Maxim Secor
类目: Machine Learning (cs.LG); Biomolecules (q-bio.BM)
*备注: Accepted to Graph Foundation Models workshop at ICML '26. Contains 8 pages, 4 figures
Abstract:Molecular property prediction from structure often uses a single representative conformation, even though many molecules exist as conformational ensembles in solution. We introduce EnsembleEGNN, a molecular ensemble foundation model that encodes an ensemble by first encoding each conformer with shared Equivariant Graph Neural Network (EGNN) layers, then pooling the resulting conformer representations with a Set Attention Block. We pretrain the model on CREMP, a cyclic peptide ensemble dataset, using a multi-task self-supervised objective combining masked token recovery, noisy-coordinate reconstruction, and pairwise distance reconstruction. On the CREMP-CycPeptMPDB dataset, training EnsembleEGNN from scratch fails entirely ( R^2=0.005 ). However, the pretrained model reaches R^2=0.477 and Pearson r=0.699 , outperforming the sequence-only BERT baseline ( R^2=0.439 , Pearson r=0.667 ). When EnsembleEGNN is co-trained end-to-end with the BERT sequence encoder, the hybrid model improves further to R^2=0.538 and Pearson r=0.737 . These results demonstrate that encoding conformational ensembles into a single thermodynamically informed embedding improves cyclic-peptide property prediction.
[LG-2] X3-OPD: Distilling Reasoning into Large Audio-Language Models via On-Policy Alignment
链接: https://arxiv.org/abs/2607.21550
作者: Dongjie Fu,Di Cao,Xize Cheng,Zihan Zhang,Wenxu Jia,Yifu Chen,Shengpeng Ji,Yu Zhang,Tao Jin
类目: Machine Learning (cs.LG)
*备注:
Abstract:While large audio-language models have achieved remarkable progress in auditory perception, they still lag behind text-based large language models in deep logical reasoning, primarily due to the scarcity of high-quality audio reasoning data. To bridge this gap, we propose X ^3 -OPD, a cross-modal on-policy distillation framework that transfers reasoning capabilities from a powerful text teacher to an audio-language student. During training, the student generates reasoning trajectories conditioned on its own acoustic perception, while the teacher provides token-level guidance using matched textual inputs and verified answers. We further construct a three-tier symmetric corpus covering textual reasoning rendered into speech, audio-event reasoning grounded in complex acoustic scenes, and spoken-dialogue reasoning involving paralinguistic cues. This design extends cross-modal distillation beyond textually recoverable content to reasoning grounded in non-linguistic events, prosody, and conversational context. Experiments on MMSU, MMAU, BIG Bench Audio, and MMAR demonstrate that X ^3 -OPD substantially improves audio-grounded reasoning and chain-of-thought quality while largely preserving the model’s existing capabilities under domain shift.
[LG-3] Zero-Flow Two-Sample Tests
链接: https://arxiv.org/abs/2607.21542
作者: Yakun Wang,Leyang Wang,Song Liu,Taiji Suzuki
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:We propose a new approach to two-sample testing for deciding whether two sets of samples are drawn from the same distribution. The test is built on a statistical discrepancy based on the zero-flow criterion, termed zero-flow discrepancy (ZFD). We prove the validity of ZFD and propose a practical testing procedure, termed the zero-flow two-sample test (ZF2ST). The key idea is to learn how samples from the two distributions are locally misaligned and use the resulting directional pattern as evidence of distributional difference. By separating witness learning from hypothesis evaluation, ZF2ST can use flexible neural networks while maintaining valid statistical calibration. We develop both regression-based and power-maximized approaches for learning the witness. Experiments on synthetic and image datasets demonstrate that ZF2ST can achieve strong testing power for structured distributional changes while maintaining well-calibrated type-I error.
[LG-4] Finite-Sample Coverag e Audits for High-Recall Candidate Generation: Certification and Learning-Theoretic Design
链接: https://arxiv.org/abs/2607.21480
作者: Martin Anthony,Kaveh Salehzadeh Nobari
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:An initial high-recall stage in an empirical pipeline decides which items pass to later review, labelling, or modelling, and relevant items it misses are lost to every subsequent stage. We study how many audit labels are needed to certify, with finite-sample validity, that this missed relevant mass is small, and our main results characterise the label complexity of this problem. We first show that no procedure using only labels from inside the candidate set can certify any non-trivial bound on the missed mass: the audit must sample the excluded pool, the only region where unrecovered relevant items can lie. We then prove a matching finite-corpus lower bound. Any valid audit that certifies fewer than m missed relevant items with high probability when none are present, even if adaptive and permitted to label the entire included pool, must inspect on the order of N_0/m excluded-pool labels. Excluded-pool auditing is therefore minimax rate-optimal, not merely convenient, for missed-mass certification in the zero-miss regime. Building on this characterisation, we develop an exact finite-sample toolkit, using binomial and hypergeometric inversion rather than asymptotic approximation, that certifies missed mass, converts it to recall through a two-pool design, certifies pre-specified families of nested candidate generators simultaneously, and produces stress-test certificates against declared perturbation mechanisms. These certificates can be paired with observable review burden to select the least burdensome pre-specified candidate generator meeting a missed-mass target. Every guarantee holds under one discipline: the candidate generator, or the pre-specified family from which it is selected, and the audit rule are fixed before the certification labels are examined.
[LG-5] st-Time Scaling via Error Localization
链接: https://arxiv.org/abs/2607.21453
作者: Rajiv Shailesh Chitale,Rahul Madhavan,Taneesh Gupta,Deepanway Ghosal,Aravindan Raghuveer
类目: Machine Learning (cs.LG)
*备注: 10 pages, 8 figures (With appendix: 27 pages, 11 figures)
Abstract:Scaling inference-time computation has emerged as a reliable method to improve the performance of large language models on complex reasoning and programming tasks. However, standard approaches such as independent sampling and sequential multi-turn refinement operate without token-level credit assignment, resulting in computational inefficiency, since valid reasoning prefixes are frequently discarded. In this work, we introduce Test-Time Scaling via Error Localization (TTEL), an inference-time algorithm that utilizes fixed or environment feedback to perform token-level error localization. By comparing conditional probabilities under informed feedback against a null-context baseline, TTEL isolates the step at which an error occurred. The algorithm then truncates the trajectory and branches a new generation, maximally reusing the valid prefix. Extensive evaluations demonstrate that TTEL establishes strictly dominating Pareto frontiers across sequential reasoning domains, measured by pass-at-k vs. generated-token cost. With Qwen3-8B on LiveCodeBench, TTEL attains a pass@64 of 71.0% while generating approximately half as many tokens as independent sampling (360.4k vs. 735.0k). Generalizing to math benchmarks AIME-2025 and HMMT-2025, TTEL cleanly outperforms competing test-time baselines across both Qwen3-8B and Qwen3-4B-Thinking-2507.
[LG-6] Context-weighted Discrete Flow Matching NEURIPS2026
链接: https://arxiv.org/abs/2607.21427
作者: Daniil Cherniavskii,Daniel Severo,Karen Ullrich
类目: Machine Learning (cs.LG)
*备注: Under review at NeurIPS 2026
Abstract:Discrete flow matching provides a flexible framework for generative modeling on discrete structures. However, the standard factorized training objective exposes the model to targets of varying difficulty, mixing well-conditioned, predictable tokens with ambiguous, high-entropy ones. We empirically demonstrate that the uncertainty over the value of each token is closely related to the density of available context in its neighborhood. Motivated by this observation, we propose a simple modification to the underlying continuous-time Markov chain (CTMC) that incorporates local context information. Our context-weighted sampler improves generation quality with negligible computational overhead, while our scaled cross-entropy loss function reweights the training signal from different tokens and reduces generative perplexity by up to 63% on OpenWebText. Moreover, our approach matches a strong semi-autoregressive block diffusion baseline in quality while retaining the ability to perform generation in any order. These results highlight the role of local context as an important factor in discrete generative modeling and show that simple context-aware modifications can significantly improve both sampling and training efficiency.
[LG-7] Semantic-Aware Task Clustering for Constructive and Cooperative Multi-Tasking
链接: https://arxiv.org/abs/2607.21426
作者: Ahmad Halimi Razlighi,Maximilian H. V. Tillmann,Edgar Beck,Bho Matthiesen,Armin Dekorsy
类目: Machine Learning (cs.LG); Information Theory (cs.IT); Signal Processing (eess.SP); Machine Learning (stat.ML)
*备注: This work has been submitted to the IEEE for possible publication
Abstract:Cooperative multi-task semantic communication (CMT-SemCom) improves task execution performance by leveraging shared representations. However, as we demonstrated in [1], cooperative multi-tasking can be either constructive or destructive, depending on the semantic relationships among tasks. To ensure constructive cooperation, we propose a semantic-aware task clustering method for CMT-SemCom. We have formulated a sequential multi-stage optimization problem in which semantically aligned tasks are clustered once after a short initial training phase, and then end-to-end (E2E) joint training is conducted exclusively within the discovered groups. Specifically, the problem decomposes into two stages: (i) a semantic clustering problem leveraging hierarchical density-based spatial clustering, and (ii) an intra-cluster E2E CMT-SemCom learning problem. Simulation results demonstrate that the proposed framework effectively mitigates destructive cooperation and negative transfer, yielding accuracy gains compared to unclustered multi-tasking and individual training baselines.
[LG-8] A Diffusion-Model Subpopulation Digital Twin for Mobile Health Deployment: A Case Study on the HeartSteps Intervention
链接: https://arxiv.org/abs/2607.21403
作者: Ziping Xu,Yuyi Chang,Chenshun Ni,Nithin Sugavanam,Asim H. Gazi,Pedja Klasnja,Emre Ertin,Susan A. Murphy
类目: Machine Learning (cs.LG); Methodology (stat.ME)
*备注:
Abstract:Mobile-health interventions increasingly use online learning and decision making algorithms to personalize when to nudge users toward healthier behavior, but a poorly designed algorithm can burden and disengage participants. New algorithm design decisions should therefore be vetted against realistic simulated users before each real-life deployment. We propose a method to develop ``JITAI-Twins’': digital twins of a target subpopulation for comparing candidate online algorithms before a just-in-time adaptive intervention (JITAI) deployment. The method builds on a conditional time-series diffusion model that is temporally consistent (future actions do not affect the generated past), and it supports repeated updating from three sources of information, in three steps: pre-training on a large observational dataset, fine-tuning on small prior intervention deployments in related populations, and inference-time calibration to the next target population from domain-scientist expertise. We validate the twin at each pre-deployment stage of the long-running HeartSteps series (v2 through v4) of physical-activity suggestion intervention deployments, treating each successive deployment as an upcoming study. The proposed method reproduces the target subpopulation’s temporal and between-participant structure better than simpler simulators. These results suggest that our twin can be used to simulate a target deployment before it runs, the prerequisite for testing and informing online algorithm design decisions.
[LG-9] Emergent Misalignment Recruits a Pre-existing Persona Subspace
链接: https://arxiv.org/abs/2607.21356
作者: Mohammed Suhail B Nadaf
类目: Machine Learning (cs.LG)
*备注: 108 pages (19 pages main text, 13 appendices), 13 figures
Abstract:Fine-tuning an aligned language model on a narrow stream of bad advice can make it broadly misaligned on questions unrelated to the training data, a phenomenon called emergent misalignment. We ask why the narrow lesson generalizes at all, and we find that narrow fine-tuning recruits a persona structure that is present in the model before the fine-tune exists. From a frozen instruction-tuned model (Qwen2.5-14B-Instruct) we extract per-domain persona subspaces by contrastive teacher forcing and find that 4 unrelated domains share one low-rank core at 657x a random-subspace null, with 82% of that core lying outside a style core built at matched diversity. The literal first optimizer step of fine-tuning on insecure code climbs a broad-misalignment margin harder than the same code framed as educational, and forecasts realized margin movement out to 375 steps. Projecting the subspace out of the residual stream throughout fine-tuning prevents broad misalignment (27.7% to 0.0% of judged generations) while a matched-rank random subspace changes nothing; injecting it into the never-fine-tuned model induces misalignment that grows with dose to 45.4%, past the fine-tuned model it is measured against. The same projection applied to the weight gradient is inert, and three post-hoc weight edits leave the disposition in place: the sharpest edit suppresses the behavior rather than removing it, and the ablated structure re-forms inside the subspace the edit cleared. Spreading a fixed budget of bad data across 4 domains produces more broad misalignment than mechanical weight superposition and matched diversity jointly account for. All measurements come from one model at 14B; the extraction is from an aligned instruction-tuned checkpoint, which leaves the structure’s provenance open; and the intervention that prevents misalignment also abolishes the narrow trained behavior.
[LG-10] Gradient Concentration Not Weight Saliency Explains Representation-Level Class Unlearning
链接: https://arxiv.org/abs/2607.21353
作者: Billel Habbati,Alessio Merlo,Luca Verderame,Meriem Guerar
类目: Machine Learning (cs.LG)
*备注: 51 pages, 7 figures. Submitted to Neural Networks
Abstract:Machine unlearning aims to remove the influence of specific training data while preserving model utility. Many state-of-the-art approaches pursue this goal by restricting the forgetting update to a subset of parameters selected through gradient-based saliency. Although such methods are widely adopted, the actual contribution of saliency-based weight selection to representation-level forgetting remains unclear. In this work, we perform the first controlled ablation of the saliency masking mechanism used by SalUn. Using a matched-compute experimental design on CIFAR-10 and CIFAR-100 with ResNet-18, we compare saliency-based masking against random masks of equal sparsity and unconstrained updates, while keeping the unlearning objective, optimization schedule, and computational budget fixed. Across multiple representation-level evaluations, including linear probing, prototype recovery, and layer-wise CKA, the three configurations exhibit statistically equivalent representation-level recoverability. We find that forget gradients are strongly concentrated in the final network layers (approximately 92% of the squared gradient energy on CIFAR-10) before any mask is applied, causing all masking strategies to operate within the same representational subspace. Furthermore, saliency masks show limited class specificity (specificity index 0.09-0.11), selecting highly overlapping parameter subsets across different forget classes. Our findings suggest that, in the studied setting, representation-level forgetting is primarily governed by gradient concentration and representation geometry rather than by the specific identity of saliency-selected weights. More broadly, the results support a growing body of evidence indicating that effective representation-level unlearning requires objectives that act directly on latent representations rather than on increasingly sophisticated weight-selection strategies.
[LG-11] How Many Bits Can an Adapter Write? Measuring the Capacity and Memorization of Parameter-Efficient Fine-Tuning
链接: https://arxiv.org/abs/2607.21351
作者: Kaizhen Tan,Heqing Du,Yang Feng
类目: Machine Learning (cs.LG)
*备注:
Abstract:A LoRA adapter is a few megabytes that almost everyone treats as a skill rather than a record of the data behind it. We put that assumption on a scale. Extending compression-based memorization analysis to the frozen-base setting, we measure directly, in bits, how much a low-rank adapter writes into a model it never changes. The answer is both smaller than full fine-tuning and less lawful than parameter counting would predict. Adapters store a couple of bits per trainable parameter, well short of a full model’s budget, but that figure turns less on how many parameters an adapter carries than on where they sit. Move the same parameter budget from attention into the MLP and it holds nearly twice as much; strip the frozen base of its structure and the capacity all but disappears. Applied to realistic fine-tunes of Qwen2.5, the same instrument shows privacy leakage rising with the bits an adapter writes rather than the parameters it nominally has, and it draws a clean line between supervised and reinforcement learning: the secrets that supervised fine-tuning copies down verbatim, an adapter trained on verifiable rewards never records. Measuring what fine-tuning writes, rather than attacking it after the fact, turns a piece of folklore into a quantity one can design against.
[LG-12] he Dark Room in the Reward Channel: Dense Prediction Rewards Collapse GRPO-Trained LLM Agents – and What Actually Works
链接: https://arxiv.org/abs/2607.21273
作者: Yu Wang
类目: Machine Learning (cs.LG)
*备注:
Abstract:Dense per-step supervision is an appealing remedy for sparse-reward, long-horizon LLM agents: reward the agent for predicting its next observation, and memory should follow. We show that under group-normalized RL (GRPO), this recipe does not merely fail – it destroys the policy. Across Qwen3-1.7B/4B/8B on ALFWorld, a potential-based prediction reward drives every run into a degenerate absorbing state (prediction accuracy - 1.0, task success - 0,episode length pinned at the horizon): the “dark room” pathology, built automatically by the optimizer. A single-factor ablation localizes the cause – removing only GRPO’s std normalization turns the same reward from catastrophic (0%) into baseline parity – and a two-line proposition explains why: in all-fail groups the z-scored advantage is invariant to the shaping coefficient, so bounded rewards become unbounded pressure and annealing cannot help. Our central insight generalizes this: what z-scoring amplifies is a dense signal’s within-group variance while all-fail groups dominate, so signals whose variance decays by mastery are structurally this http URL variance-profile criterion retrodicts our collapses, carries preregistered predictions for arms that had not yet run, and is consistent with published reward-channel successes (a compatibility check, not an independent test). Finally, a controlled signal-delivery matrix (identical signal, varying only the consumption mechanism) shows the reward channel is at best neutral while the auxiliary-loss channel gains ~20 points – and a shuffled-gold placebo matches the true-gold arm, so the gap survives without correct labels. Endpoints are single-seed; seed replication and group-size controls are preregistered and in progress.
[LG-13] Filter Learning for Subgraphs: Algebras and Performance Risk Bounds
链接: https://arxiv.org/abs/2607.21263
作者: Purui Zhang,Feng Ji,Yanan Zhao,Bihan Wen,Wee Peng Tay
类目: Machine Learning (cs.LG); Signal Processing (eess.SP)
*备注: Submitted to IEEE TSP
Abstract:Graph signal processing tasks that leverage spectral information typically assume access to the complete graph topology, which is often unavailable in practice. We propose a systematic framework for subgraph filter learning (SFL), where subgraph-supported operators approximate ambient graph filters under partial observations. We formulate SFL as a statistical learning problem in which optimal subgraph operators are inherently data-dependent. To address the difficulty of directly estimating such operators, we develop a subgraph filter algebra based on distance-aware Laplacian constructions, defining a structured and controllable class of filters for effective approximation. We further establish performance risk bounds under the least squares loss, quantifying how well the learned operator approximates the restricted ambient mapping. Experiments real-world datasets show that, for SFL tasks, the proposed algebraic models consistently outperform polynomial filters, distribution-agnostic operators, and direct numerical filter learning baselines that attempt to recover the underlying structure from data.
[LG-14] Automated Synthesis and Adversarial Validation of Executable Causal Research Pipelines
链接: https://arxiv.org/abs/2607.21173
作者: Irena Girshovitz,Dan Zeltzer,Ran Gilad-Bachrach
类目: Machine Learning (cs.LG)
*备注:
Abstract:While automated research systems promise to accelerate empirical analysis, they are prone to silent failures: instances in which analysis code executes successfully yet relies on invalid causal assumptions. We present the Artificial Intelligence (AI)-based Epidemiology Research Assistant (ARA), a framework that makes these failures visible by explicitly encoding causal design principles, study-specific assumptions, and methodological constraints. ARA integrates protocol construction, synthetic data generation, and adversarial validation into a unified pipeline. The framework translates natural language research questions into structured causal protocols and executable analysis code by first constructing a protocol and then generating synthetic datasets using Structural Causal Models (SCMs) with known ground-truth effects. This synthetic-data step can also support pipeline development when access to confidential data, such as medical data, is restricted. The generated analysis is then evaluated under controlled violations of identification assumptions. We evaluate ARA on the Automated Causal Reasoning Benchmark, assessing recovery of identification strategies, causal quantities, treatment and outcome variables, and consistency between generated code and approved protocol. Protocol construction and adversarial validation did not consistently improve numerical agreement with benchmark estimates compared with standard LLM-based generation. However, they changed the failure mode: instead of silently returning causal estimates, ARA often surfaced protocol concerns, diagnostic failures, incomplete inference, or downgraded non-causal interpretations. These findings suggest that validity-first automated science systems should be evaluated not only by answer accuracy, but also by whether they indicate when causal claims are unwarranted.
[LG-15] Agree on the Model Verify the Inference: GKR Protocols for HND-Based Transformer Inference
链接: https://arxiv.org/abs/2607.21162
作者: Xiaolong Liang,Juanjuan Li,Rui Qin,Yisheng Lv
类目: Machine Learning (cs.LG); Cryptography and Security (cs.CR)
*备注: 24 pages, including 4 pages of supporting information; 2 figures
Abstract:Outsourced Transformer inference exposes clients to model substitution and incomplete execution, while direct replay removes the computational benefit of delegation. We present GKR-HND, a registered-model protocol for verifying the polynomial backbone of Homomorphic–Nonhomomorphic Decomposition Transformers. The retained verifier checks the GKR transcript and registered-weight openings, but delegates expensive public evaluations to an assigned computation worker. Assuming an honest retained verifier and prover–worker non-collusion, the verifier accepts only when the worker’s signed, request-bound response agrees with the proof claims. Experiments with pretrained HND models validate the proof path and the delegated public computation without dense-matrix replay.
[LG-16] Smooth Neural Point Processes via B-Splines
链接: https://arxiv.org/abs/2607.21098
作者: Michele Bellomo,Riccardo Ramaschi,Alberto Dolara,Tomaso Aste
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 9 pages, 2 figures
Abstract:Temporal point processes (TPPs) provide a general and flexible framework for modeling sequences of events in continuous time. Neural networks have been successfully employed to model TPPs in a highly expressive and data-driven way. Neural TPPs are typically trained via Maximum Likelihood Estimation (MLE) by minimizing the negative log-likelihood (NLL), which depends on both the conditional intensity function (CIF) and its integral over time, the compensator. Recent neural TPP approaches enable exact evaluation of the NLL without numerical integration. However, these methods typically model the compensator rather than the CIF directly, impose constraints on the neural network architecture, and are computationally expensive during training, as event contributions to the NLL are evaluated sequentially rather than in parallel. In this work, we propose a novel neural TPP model that directly parametrizes the CIF as a non-negative combination of B-spline basis functions, whose coefficients are predicted by a neural network. This formulation enables exact evaluation of the NLL, preserves full flexibility in the neural architecture, allows efficient parallelization during training, and naturally supports CIF smoothness regularization through the integrated squared second derivative. Experiments on both synthetic and real-world datasets show improved computational efficiency and predictive accuracy compared to the reference neural TPP baseline.
[LG-17] A Polynomial Architecture-Attribution Co-Design Framework for Exact Aumann-Shapley Attribution in GNNs
链接: https://arxiv.org/abs/2607.21094
作者: Bizu Feng,Zhimu Yang,Shuming Wang,Shaode Yu,Yuan Cheng,Xiaojun Qian,Zixin Hu
类目: Machine Learning (cs.LG)
*备注: It has 23 pages
Abstract:We study feature-level and node-level explanations for graph neural networks (GNNs) through the lens of Aumann-Shapley attribution. Path-integral methods such as Integrated Gradients provide an axiomatic formulation of attribution, but their practical use in deep GNNs typically relies on finite-sample numerical approximations to the path integral, requiring a trade-off between quadrature error and computational cost. This paper proposes APEX, a model-attribution co-design framework that makes the attribution integral exactly computable under a polynomial GNN architecture. The key component is PolyGIN, a GIN-style graph network whose message-passing, normalization, and transformation operations preserve a bounded multivariate polynomial form for scalar model scores, such as pre-softmax logits. We show that, for a PolyGIN with L polynomial transformation blocks, the derivative along the attribution path has degree at most 2^L-1 . Therefore, Gauss–Legendre quadrature can evaluate the Aumann–Shapley path integral exactly, up to floating-point precision, with 2^L-1 deterministic evaluation points. The resulting attributions can be computed at the feature level and then aggregated into node-level scores while preserving completeness. Experiments on synthetic and real-world graph benchmarks show that PolyGIN maintains competitive predictive performance, while the complete APEX framework achieves higher attribution fidelity than the compared baselines and substantially reduces the number of evaluations required for path integration.
[LG-18] CASC: Causal Adversarial Subspace Clustering for Multivariate Spatiotemporal Data
链接: https://arxiv.org/abs/2607.21088
作者: Francis Ndikum Nji,Vandana Janeja,Jianwu Wang
类目: Machine Learning (cs.LG)
*备注: 10 pages
Abstract:Deep subspace clustering plays a critical role in applications involving multivariate spatiotemporal data, such as sea ice monitoring, disease spread analysis, and tracking neuro-degeneration over time. Despite recent advances, existing methods primarily rely on geometric self-expressiveness, assume static subspace structures, and often fail to capture causal dependencies, local spatial interactions, and long-range temporal dynamics inherent in complex spatiotemporal systems. To address these limitations, we propose a novel Causal Adversarial Subspace Clustering (CASC) framework for discovering evolving latent regimes in high-dimensional spatiotemporal data. CASC integrates a U-Net-inspired deep adversarial clustering architecture with stacked FAConvLSTM layers to preserve spatial and temporal structure while learning robust latent representations. A graph attention transformer-based self-expressive network is introduced to jointly model local spatial relationships, global dependencies, and long-range temporal interactions. Furthermore, we propose two new learning objectives: (1) a Causal Subspace Preservation Loss that aligns self-expression coefficients with latent causal relationships, encouraging clusters to reflect underlying causal processes rather than simple feature similarity, and (2) a Dynamic Temporal Subspace Evolution Loss that captures evolving subspace structures and temporal regime transitions in nonstationary environments. Together, these components transform deep subspace clustering from a correlation-driven paradigm into a causal-temporal regime discovery framework.
[LG-19] Nipping the Butterfly Effect in the Bud: Self-Output Fine-Tuning for Autoregressive Weather Prediction
链接: https://arxiv.org/abs/2607.21080
作者: Yun-Ye Cai,Hsuan-Tien Lin
类目: Machine Learning (cs.LG)
*备注:
Abstract:Long-horizon weather forecasting is a fundamental challenge in atmospheric science, for which autoregressive Deep Learning Weather Prediction (DLWP) has emerged as the primary paradigm. Although the autoregressive pipeline is highly scalable and flexible, its prediction errors grow rapidly over long forecasting horizons. In this work, we study this error growth phenomenon from both theoretical and empirical perspectives. Our analysis reveals that the growth is driven by a feedback loop between output errors and input distribution shifts. Specifically, the autoregressive process amplifies small initial output errors, which progressively corrupt subsequent input distributions, echoing the butterfly effect in atmospheric science and ultimately deteriorating forecasting accuracy over longer horizons. Furthermore, we show that this distributional shift originates at the earliest stage of inference, with out-of-distribution signatures detectable as early as the first autoregressive step. To mitigate this issue, we propose \textbfSelf-Output Fine-Tuning (SOFT), a plug-and-play strategy that leverages the model’s own one-step predictions to calibrate the biased input distribution encountered at the first step. Extensive experiments demonstrate that, despite its simplicity, SOFT achieves state-of-the-art performance on long-horizon forecasting tasks and substantially reduces both prediction errors and distributional discrepancy. The success of SOFT highlights the importance of reexamining the fundamental pipeline of deep learning weather prediction, representing a critical pipeline advance for atmospheric science.
[LG-20] Spectral Transformation for Layer-wise Global Rank Discovery in Federated LoRA for Vision Transformers ECML-PKDD2026
链接: https://arxiv.org/abs/2607.21074
作者: Hariharan Ramesh,Jyotikrishna Dass
类目: Machine Learning (cs.LG); Distributed, Parallel, and Cluster Computing (cs.DC)
*备注: Accepted at ECML-PKDD 2026 (Research Track). This is the submitted version, prior to peer review. Code: this https URL
Abstract:Fine-tuning Vision Transformers (ViTs) with low-rank adapters (LoRA) promises better communication efficiency under federated setup, yet existing aggregation strategies face fundamental limitations. Independently averaging these LoRA factors is mathematically inconsistent, introducing cross-term aggregation error. In contrast, approaches that preserve heterogeneous client ranks by concatenating local adapters on the server substantially increase download cost and often require merging global LoRA updates into pretrained weights on the clients, causing reinitialization lag and unstable convergence. Other approaches further increase server-side overhead by reconstructing dense weight updates or training auxiliary models to refine aggregation error. In this work, we propose SpecTraL, spectral transformation for layer-wise global rank discovery, that resolves these challenges within a unified design. SpecTraL stacks local LoRA modules from clients and performs orthonormal Householder Transformation of the stacked adapters directly in the low-rank latent space, eliminating dense reconstruction of the global update and any auxiliary refinement on the server. By leveraging the Spiked Covariance Model from Random Matrix Theory, SpecTraL analytically separates the global consensus signal from non-IID noise, discovering optimal layer-wise global ranks without manual hyperparameter tuning. To match local ranks in subsequent rounds, we introduce a padding-aware initialization framework that lets clients incorporate residual LoRA dimensions without re-merging them into the pre-trained base model. Experiments on federated fine-tuning of ViT-B/16 and ViT-L/16 over DomainNet and NICO++ demonstrate improved accuracy-communication trade-offs, reduced server computation, and elimination of hyperparameter search for rank selection. Our code is available at this https URL
[LG-21] From Evaluation to Optimisation: Hierarchy-Aware Training Signals for CWE Prediction in Python
链接: https://arxiv.org/abs/2607.21069
作者: Muntasir Adnan,Manile Srun,Carlos C. N. Kuhn
类目: Machine Learning (cs.LG)
*备注:
Abstract:The original ALPHA benchmark introduced a taxonomy-aware penalty for evaluating CWE-level vulnerability prediction in Python and proposed that the penalty could theoretically also serve as a training signal. This paper provides that validation. We compare three delivery mechanisms: supervised fine-tuning, a dual-head classification loss, and reinforcement learning with a dense reward derived from the normalised penalty. We find that supervised approaches consistently regress below the zero-shot baseline under distribution shift, while GRPO succeeds. Our best policy reduces the cumulative ALPHA penalty of Qwen2.5-Coder-7B on Security Hardening and Adversarial Testing (SVEN) dataset by 27.9% under greedy decoding, and by 25.5% under sampled decoding(p = 0.005, Welch’s t-test), reaching statistical parity with its 4.5x larger zero-shot teacher. We conclude that the value of a hierarchical penalty as a training signal depends largely on the directness of its delivery.
[LG-22] Regularized Optimization on Grassmann Manifold: Theory Algorithm and Applications
链接: https://arxiv.org/abs/2607.21039
作者: Zhuan Liang,Zheng Zhai
类目: Machine Learning (cs.LG)
*备注:
Abstract:Spectral methods are among the most widely used techniques for community detection, clustering, and graph learning. Their performance, however, critically depends on the accurate estimation of the underlying spectral subspace and can deteriorate substantially in the presence of noise, outliers, or model perturbations. To address this limitation, we propose a Regularized Projection Matrix Approximation (RPMA) framework for robust estimation of rank- K projection matrices. RPMA extends classical spectral projection by incorporating a regularization term, producing projection estimates that are more robust, sparse, and interpretable. We formulate the proposed model as an optimization problem on the manifold of rank- K projection matrices and exploit its geometric equivalence to the Grassmann manifold. Based on this manifold characterization, we derive the first- and second-order optimality conditions, establish the local stability of the regularized leading eigenspace, and characterize the stability of the critical-point landscape under sufficiently small regularization. To efficiently solve the resulting nonconvex optimization problem, we develop a Riemannian gradient projection algorithm with backtracking line search, together with a more efficient Cayley–Sherman–Morrison–Woodbury (Cayley–SMW) gradient method that avoids repeated eigendecompositions. Extensive experiments on both synthetic and real-world datasets demonstrate that RPMA substantially improves the recovery accuracy of projection matrices and consistently outperforms conventional spectral projection methods for community detection and clustering under noisy environments.
[LG-23] Weight-norm Criticality: A Mechanism for Loss Spikes Induced by the Normalization and Weight Decay
链接: https://arxiv.org/abs/2607.21005
作者: Xiaolong Li,Zhangchen Zhou,Zhi-Qin John Xu
类目: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
*备注:
Abstract:Most explanations of training instability focus on \emphlearning-rate criticality, typically characterized by the Edge of Stability, beyond which optimization becomes unstable. We argue that, in practical deep neural network training, there is an additional and often overlooked \emphweight-norm criticality. This criticality is induced by the interaction between normalization (which introduces scale-invariant components) and weight decay (which persistently shrinks parameter norms). As the weight decay coefficient increases, the norms of scale-invariant weights are progressively driven toward zero. Meanwhile, the sharpness of the loss landscape increases rapidly, destabilizing the optimization dynamics and resulting in abrupt loss spikes. This perspective provides a rationale for why weight penalties can improve generalization yet cannot be made arbitrarily strong: excessive decay drives scale-invariant weight norms past a critical boundary and destabilizes training. Our work provides a new mechanistic understanding of loss spikes through the lens of \emphweight-norm criticality. Moreover, \emphweight-norm criticality yields testable predictions that we validate empirically in networks with scale-invariant components, providing empirical support for the proposed mechanism.
[LG-24] ADABORD: a novel AdaBoost approach for ordinal classification
链接: https://arxiv.org/abs/2607.21003
作者: Rafael Ayllón-Gavilán,Francisco José Martínez-Estudillo,David Guijo-Rubio,César Hervás-Martínez,Pedro A. Gutiérrez
类目: Machine Learning (cs.LG)
*备注: 42 pages, 6 figures, 7 tables. Submitted to CS top tier journal
Abstract:Ordinal Classification (OC) deals with classification tasks where the classes follow a natural order. Despite the progress in OC, many existing approaches fail to fully leverage the ordinal information, treating the problem as nominal classification and thereby losing performance potential. In this work, ADABORD, an AdaBoost framework specifically designed for ordinal classification problems, is introduced. The ordinal nature of the classes is incorporated into two key components of the well-known AdaBoost algorithm: 1) the base estimator, where decision trees with the ordinal Gini splitting criterion are proposed; 2) the error function used to update sample weights at each stage and the weights of the classifier in the final ensemble model, given by the absolute ranked probability score, a measure that accounts for both the ordering and the distance between classes. ADABORD is extensively compared against seven state-of-the-art methods on the TOC-UCO repository, the largest benchmark collection for OC to date. The experimental results, supported by statistical analysis, show that ADABORD significantly outperforms competing methods, particularly on datasets with five or more classes, where the ordinal structure becomes more pronounced. Source code, along with all experimental protocols, is publicly available to ensure reproducibility and facilitate future research in OC.
[LG-25] Best-of-Evidence: Best-of-N Selection under Partial Verification
链接: https://arxiv.org/abs/2607.20950
作者: Cenwei Zhang,Teng Fang,Yuxia Wang,Derek Li,Bryan Dai,Lei You
类目: Machine Learning (cs.LG)
*备注: 3 figures, 28 pages
Abstract:BoN improves model outputs by sampling several candidates and selecting one with a proxy score, but it assumes that complete candidates can be evaluated reliably. Many vision-language tasks instead provide only partial verification: a finding, span, value, region, or relation may be checkable even when no dependable whole-response verifier exists. Moreover, the same claim may recur across candidates with opposing stances, allowing one observation to support part of the pool and contradict another. We introduce Best-of-Evidence (BoE), an inference-time selection framework that keeps the BoN candidate pool fixed, represents reusable claims with a signed candidate–factor graph, and allocates a limited budget to evidence actions that can change the final choice. BoE formalizes selection under partial verification and provides a practical score-based controller, with the zero-budget case recovering the underlying BoN decision. Theoretically, we show that residual evidence capacity limits any evidence-driven improvement and that shared factor queries can achieve an O(log K) versus \Theta(K) query separation in a factor-code model. Common-ledger experiments on four medical VQA settings show that BoE can improve fixed-pool selection and rescue some BoN failures when evidence is reliable, contrastive, and decision-relevant, while also revealing the channel-quality and candidate-generation limits that prevent universal gains.
[LG-26] hree-Pronged Spectral Control for Federated Parameter Efficient Fine Tuning
链接: https://arxiv.org/abs/2607.20914
作者: Shiva Raj Pokhrel,Dipsan Bhattarai,Anwar Walid
类目: Machine Learning (cs.LG); Networking and Internet Architecture (cs.NI)
*备注: 18 pages, 17 figures, 11 tables
Abstract:Federated parameter-efficient fine-tuning (PEFT) enables communication-efficient adaptation of large pretrained models on decentralized edge data, but it remains fragile under non-IID client heterogeneity. In low-rank adaptation (LoRA), different clients may learn locally useful but spectrally misaligned update subspaces, causing high-variance aggregation and poor global transfer. We propose TRISHUL, a spectral-control framework for robust federated PEFT. TRISHUL follows the FL no-raw-data-sharing setting but does not itself provide formal privacy guarantees. TRISHUL uses shared frozen multi-head low-rank bases to obtain algebraically exact aggregation of compact core updates, applies nuclear norm proximal shrinkage to suppress client-specific high-rank spectral components before upload, and allocates adaptation heads non-uniformly across layers using a concave water filling budget rule derived from pretrained layer capacity. Because shrinkage is performed only on small core matrices, TRISHUL adds negligible computation and no extra per-round communication over the underlying multi-head PEFT protocol. Across vision and language benchmarks, including CIFAR-100, SVHN, 20 Newsgroups, MRQA, and GLUE with LLaMA3.2-1B, TRISHUL improves convergence, stability, and final performance over federated LoRA baselines, with greater gains under stronger heterogeneity.
[LG-27] HierarchicalDAEW: Domain-Aware Edge-Weighted Graph Convolution with Evidential Uncertainty for Multi-Section Spatial Gene Expression Prediction from HE Histology
链接: https://arxiv.org/abs/2607.20896
作者: Kritanu Chattopadhyay,Soumya Chatterjee,Ondrej Krejcar,Debotosh Bhattacharjee
类目: Machine Learning (cs.LG); Genomics (q-bio.GN)
*备注: 30 pages, 36 figures, 26 tables
Abstract:Spatial transcriptomics assays remain costly and technically demanding, restricting transcriptome-wide profiling to specialist settings and preventing routine clinical deployment. Predicting spatially resolved gene expression from HE histology could close this gap, yet current methods largely ignore the underlying tissue architecture and rarely quantify how their predictions can be trusted. We introduce HierarchicalDAEW, a dual-graph architecture that addresses both gaps. On the spot graph, a Domain-Aware Edge-Weighted convolutional operator learns separate projections for inter-domain, intra-domain, and boundary edges derived from Leiden clustering, allowing the model to treat tissue heterogeneity as an explicit structural signal rather than an implicit one. A second gene-level graph then fuses protein-protein interaction priors from STRING-DB with tissue-specific co-expression through learned attention gating, propagating predictions from a landmark gene set to a broader gene panel. Reliability is handled through evidential uncertainty estimation, which produces far better calibrated confidence intervals than Monte Carlo dropout under identical conditions. Across six human Visium sections spanning breast, colorectal, prostate, and cerebellar tissue, and against thirteen published baselines, HierarchicalDAEW achieves the strongest correlation with ground-truth expression, with gains that hold up under multi-seed reproducibility checks and negative controls that rule out positional shortcuts. Ablations further confirm that both the domain-aware edge typing and the hierarchical depth are necessary to this improvement, and calibrated uncertainty estimates identify low-confidence predictions for pathologist review before clinical action.
[LG-28] Information-Theoretically Secure Aggregation for Lightweight Federated Learning: Resilient to Dropouts and Adversaries
链接: https://arxiv.org/abs/2607.20890
作者: Hyeong-Gun Joo,Songnam Hong,Dong-Joon Shin
类目: Machine Learning (cs.LG)
*备注:
Abstract:On-device federated learning (FL) enables privacy-preserving and personalized model training on resource-constrained devices such as smartphones and IoT nodes. To reduce communication cost, sign-based methods (e.g., signSGD) transmit one-bit gradients. However, exposing gradient signs makes them vulnerable to inference attacks, while existing secure aggregation schemes are often incompatible with such methods or incur significant computational and communication overhead. We propose a lightweight and information-theoretically secure aggregation framework tailored for sign-based FL. The framework securely computes the majority vote (MV) polynomial through single-round secure multiplication, ensuring end-to-end information-theoretic security under the honest-majority assumption while revealing only the final aggregated sign to the server. To enhance efficiency and scalability, we introduce two key techniques. First, inverse-form exponent reduction halves the effective MV polynomial degree, reducing both communication and computation costs. Second, we propose single-round secure multiplication, achieving linear offline complexity and storage with only a single online communication. Together, these techniques reduce online communication by up to 99.5% and latency by up to 85.7% compared to conventional approaches. Also, by leveraging inherent MDS-code-based decoding, the framework achieves robustness against both dropouts and adversarial behaviors, yielding accuracy gains of up to 20.65% and 10.74%, respectively. Overall, the proposed framework establishes a practical foundation for large-scale, low-latency, and information-theoretically secure aggregation in sign-based FL.
[LG-29] Offline RL with Hierarchical Action Chunking
链接: https://arxiv.org/abs/2607.20834
作者: Ahad Jawaid
类目: Machine Learning (cs.LG)
*备注: RLC/RLJ 2026
Abstract:Offline goal-conditioned reinforcement learning (RL) holds the promise of learning general-purpose policies from static datasets. However, scaling these methods to long-horizon tasks remains a challenge due to the curse of horizon, where value estimation errors can compound through long chains of bootstrapped Bellman backups. Existing hierarchical approaches mitigate this by decomposing tasks into subgoals, yet they often rely on low-level controllers that suffer from myopic execution and biased value estimates. In this work, we propose Hierarchical Implicit Q-Chunking (HiQC), an offline goal-conditioned RL algorithm that combines high-level latent planning with low-level action chunking. By conditioning the low-level critic on temporally extended action sequences, HiQC enables unbiased k-step value backups, compressing the horizon at both the planning and execution levels. We theoretically demonstrate that this dual decomposition results in a tighter bound on value error under a bounded per-backup error model compared to standard hierarchy or flat chunking alone. Empirically, HiQC achieves the highest aggregate performance among the compared methods on the OGBench suite, with its largest gains on long-horizon navigation tasks such as humanoid-giant.
[LG-30] Robust Asynchronous Q-Learning under Reward and State Corruption via Batching
链接: https://arxiv.org/abs/2607.20822
作者: Sreejeet Maity,Aritra Mitra
类目: Machine Learning (cs.LG); Systems and Control (eess.SY)
*备注: To appear at the 65th IEEE Conference on Decision and Control (CDC) 2026
Abstract:Motivated by reinforcement learning in harsh environments, we consider the problem of learning an optimal policy subject to adversarially corrupted feedback. Specifically, at each time-step, an adversary can perturb both the reward and state observations of the learner following the Huber contamination model. To defend against such data corruption, we propose \textttBR-Async-Q: a novel, epoch-based, robust (Q)-learning algorithm built upon two key ideas: (i) partitioning the online data stream into batches to reduce variance, and (ii) constructing robust estimates of the Bellman optimality operator using such batched data. We prove a high-probability \ell_\infty error bound for \textttBR-Async-Q that matches that for vanilla (Q)-learning, up to a small additive term that scales with the fraction of corrupted samples. To our knowledge, this provides the first robustness guarantee for asynchronous (Q)-learning subject to both reward and state corruption. Furthermore, when only rewards are corrupted, the dependence of our algorithm’s bound on the corruption fraction is minimax optimal.
[LG-31] New Complexity-Theoretic Frontiers of Tractability for Neural Network Training NEURIPS2023
链接: https://arxiv.org/abs/2607.20811
作者: Cornelius Brand,Robert Ganian,Mathis Rocton
类目: Machine Learning (cs.LG); Data Structures and Algorithms (cs.DS)
*备注: Appeared in the proceedings of NeurIPS 2023
Abstract:In spite of the fundamental role of neural networks in contemporary machine learning research, our understanding of the computational complexity of optimally training neural networks remains incomplete even when dealing with the simplest kinds of activation functions. Indeed, while there has been a number of very recent results that establish ever-tighter lower bounds for the problem under linear and ReLU activation functions, less progress has been made towards the identification of novel polynomial-time tractable network architectures. In this article we obtain novel algorithmic upper bounds for training linear- and ReLU-activated neural networks to optimality which push the boundaries of tractability for these problems beyond the previous state of the art. In particular, for ReLU networks we establish the polynomial-time tractability of all architectures where hidden neurons have an out-degree of 1 , improving upon the previous algorithm of Arora, Basu, Mianjy and Mukherjee. On the other hand, for networks with linear activation functions we identify the first non-trivial polynomial-time solvable class of networks by obtaining an algorithm that can optimally train network architectures satisfying a novel data throughput condition.
[LG-32] External Clustering Validation by the Homogeneity-Parsimony Trade-off
链接: https://arxiv.org/abs/2607.20799
作者: Andreas Tiffeau-Mayer
类目: Machine Learning (cs.LG); Information Theory (cs.IT); Methodology (stat.ME)
*备注: 12 pages, 5 figures
Abstract:Scalar metrics are often used to evaluate clusterings against known classes, but they can obscure a fundamental trade-off: clusterings should be informative about class labels while avoiding unnecessary fragmentation. Here we describe normalized scores of cluster homogeneity and parsimony that quantify this trade-off. These scores build on the information bottleneck principle, modified to not reward lossy compression. We show by example and mathematical proof that our definitions of these scores have the intuitive property of varying monotonically under cluster refinement in contrast to related proposals. Extending the information-theoretic framework beyond Shannon entropies, we furthermore derive set-matching and pair-based counterparts of the homogeneity and parsimony scores. These unify commonly used evaluation criteria and show that, in the pair-based setting, the homogeneity-parsimony trade-off recovers the receiver operating characteristic of binary classifiers. We demonstrate the framework’s utility for feature selection and algorithm comparison, illustrating how considering scores jointly can clarify clustering operating points and identify Pareto-optimal solutions.
[LG-33] Memoir: Should a Model Write to Its Memory While It Thinks?
链接: https://arxiv.org/abs/2607.20792
作者: Jaber Jaber,Osama Jaber
类目: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
*备注: 9 pages, 4 figures, 4 tables, 1 algorithm. Code: this https URL
Abstract:Memoir combines per-sample fast memory, shared slow parameters, variable-depth latent recurrence, and a future-latent energy objective. We test its riskiest coupling: each pondering iteration may rewrite the fast tier that the same iteration reads. On procedural associative recall with key interference, we compare a coupled arm against an otherwise identical read-only pondering arm. Both arms contain 81,738 parameters, including 76,362 trainable parameters, and use matched declared forward multiply-accumulate counts, data, optimizer, schedule, and seeds. After 240 training steps across 12 seeds, coupled recall is 0.5203 with a 95 percent interval of [0.4522, 0.5883], while read-only recall is 0.6557 with [0.5953, 0.7160]. The arms are paired per seed, and the read-only lead of 0.1354 gives a paired t of 3.23 on 11 degrees of freedom with a 95 percent interval of [0.0431, 0.2277] on the difference, winning on 10 of 12 seeds. After 960 steps across 8 seeds, both arms reach 1.0000, so the measured effect is a learning-speed penalty at a fixed budget, not a demonstrated capability penalty. That longer control is ceiling limited, leaving convergence on a non-saturating task unmeasured. A predicted failure in which memory rewriting corrupts the energy signal did not occur: the energy margin grew and held. Kernel restructuring also reduced delta-rule forward time from 0.907 ms to 0.351 ms on the stated device. Code and evidence are available at this https URL
[LG-34] oward Mechanistic Interpretability of an AI Foundation Model Fine-Tuned for Atmospheric Chemistry
链接: https://arxiv.org/abs/2607.20778
作者: Jason Y. Hu,Ivan Higuera-Mendieta,Patrick Obin Sturm,Makoto M. Kelp
类目: Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph)
*备注:
Abstract:Weather forecasting foundation models (FMs) are increasingly fine-tuned to predict air quality, offering fast global pollution forecasts at lower computational cost than conventional chemical transport models. These FMs are typically trained on reanalysis data and generate forecasts through autoregressive rollout. They do not explicitly represent governing physical or chemical processes. Therefore, high forecast skill does not reveal whether a model has learned physical mechanisms or exploits statistical regularities in its training data. Here, we present the first study of what a FM fine-tuned for atmospheric chemistry has learned by examining Microsoft’s Aurora model. We impose controlled chemical perturbations on its forecasts and test them against known photochemical relationships. We then examine the internal representations that generate these forecasts. We find that Aurora captures a first-order ozone response to reactive nitrogen but does not enforce the chemical constraints that a process-based model encodes. It generates chemically inconsistent combinations of related species and relaxes localized emission features such as wildfire plumes toward background. Internally, its representations remain largely organized around the meteorology inherited during pretraining, with little structure specific to chemistry. Using sparse autoencoders, we identify internal components that causally control the chemical forecast but do not map cleanly onto individual atmospheric processes. This work provides a framework for testing whether AI forecasting systems learn atmospheric chemistry from reanalysis data. As these models are increasingly positioned to inform environmental policy decisions, we argue that composition forecasts should also be judged by their internal mechanisms rather than by benchmark skill alone.
[LG-35] Memory-Computation Tradeoffs in Semi Amortized Parametric Optimization
链接: https://arxiv.org/abs/2607.20769
作者: Shijie Pan,Agustin Castellano,Zeyu Shen,Enrique Mallada
类目: Machine Learning (cs.LG)
*备注:
Abstract:Learning-enabled decision systems often use offline data or computation to reduce online compute cost. Despite the empirical success of such approaches, there is limited general understanding of how much offline information is needed to achieve a desired accuracy under a fixed online computation budget. We study this question through the lens of amortized parametric optimization: an offline phase stores a finite memory of solved problem instances, and an online phase produces a solution to a new instance by retrieving a warm start and applying K steps of projected gradient descent. We analyze this setup for smooth convex parametric optimization over a compact domain, using a nonparametric predictor built from the stored offline solutions. For \mu -strongly convex objectives, we establish matching upper and lower bounds on the memory required to guarantee \varepsilon -accuracy under a fixed online iteration budget K . For convex objectives satisfying a \beta -growth condition ( \beta2 ), we obtain near-matching bounds and identify a phase transition in K beyond which additional memory provides no benefit. We further provide a general proof framework that (i) explicitly quantifies the memory cost of acceleration—how much offline memory is required to achieve a prescribed speedup over the unaided online optimizer—and (ii) identifies two key quantities driving this cost: the convergence rate of the online optimizer and the Lipschitz sensitivity of the solution map to the problem parameter. Experiments on parameterized ridge regression confirm the predicted memory–computation–accuracy tradeoffs.
[LG-36] Adaptive Confidence-weighted Expansion for Trustworthy Multi-Omics Multimodal Fusion ICPR2026
链接: https://arxiv.org/abs/2607.20742
作者: Mohammad Raahemi,Ali Sekhavati,Alireza Maleki,Hamid Nasiri
类目: Machine Learning (cs.LG)
*备注: Accepted for publication in the proceedings of the International Conference on Pattern Recognition (ICPR 2026)
Abstract:Multimodal learning is a robust approach to improve predictive performance in applications such as medical prognosis. However, the clinical applicability of models that use multimodal learning is hampered by their poor performance under noisy or uninformative data streams. Present fusion approaches often lack robust mechanisms for the dynamic assessment of data quality and for the provision of a trustable confidence score on the final prediction. This dissuades their deployment in safety-critical settings. To address these limitations, we introduce Adaptive Confidence-weighted Expansion (ACE), a novel framework to enhance the trustworthiness of multimodal fusion models. ACE first enhances the multimodal space by generating new, complementary modalities from intra-modality correlations. It then employs a dual-level confidence mechanism that (1) adaptively reweighs all modalities by their reliability before fusion and (2) estimates a global trust score over the fused, final decision. To evaluate ACE, we used four challenging multi-omics datasets (BRCA, KIPAN, LGG, and ROSMAP). ACE significantly outperforms existing state-of-the-art algorithms in both classification performance and confidence calibration. Our framework provides a more stable and robust data fusion method that facilitates the use of multimodal learning in addressing high-stakes problems.
[LG-37] Pipelined Gradient Coding
链接: https://arxiv.org/abs/2607.20739
作者: Xian Su,Jun Li
类目: Information Theory (cs.IT); Machine Learning (cs.LG)
*备注: This is the extended version of the paper accepted at the IEEE Information Theory Workshop (ITW) 2026
Abstract:In large-scale machine learning, distributed training commonly involves multiple workers evaluating the gradients of the model on different dataset partitions. A common challenge is the presence of straggling workers, which may significantly slow down training. Traditional gradient coding (GC) addresses this by duplicating dataset partitions across workers, allowing for the replacement of missing gradients from stragglers. However, GC requires workers to evaluate gradients on multiple dataset partitions in each step, potentially increasing overall training time. In this paper, we propose to pipeline GC, such that gradient evaluation is segmented across multiple steps and each worker evaluates gradients on just a single dataset partition per step. We develop the pipelined version for fractional repetition (FR) and cyclic repetition (CR), two representative dataset placement schemes in GC, and prove convergence guarantees for both. Through extensive simulations and experiments on cloud infrastructure, our schemes not only significantly reduce training time but also accelerate convergence compared to GC and other baselines.
[LG-38] LLM s Get Lost in Evolving User Intent
链接: https://arxiv.org/abs/2607.20734
作者: Jihoon Tack,Philippe Laban,Jennifer Neville
类目: Machine Learning (cs.LG)
*备注: 20 pages, 10 figures
Abstract:As LLMs become more capable, they are increasingly deployed as collaborative agents, taking on user-delegated tasks through iterative interaction. Yet genuine interaction is inherently dynamic: users rarely specify their intent upfront, instead disclosing, revising, and reshaping it as the conversation unfolds. Despite this, LLMs are still predominantly evaluated or trained in single-turn, fully-specified settings, leaving open a fundamental question: how well do LLMs track and act on user intent as it evolves over the course of a conversation? To study this, we introduce a framework that transforms static, single-turn tasks into dynamic multi-turn conversations in which the user’s intent evolves across turns–incrementally revealed, revised, and at times redirected mid-conversation–while preserving each task’s original evaluation protocol, enabling existing benchmarks to be reused as controlled testbeds without new annotation. Across multiple tasks, we surface a consistent phenomenon: strong static-setting performance does not transfer to the evolving-intent setting, with substantial drops across model families. Our findings point to a fundamental gap: today’s LLMs do not yet faithfully track and act on the user’s evolving intent, a capability invisible to static evaluation yet critical for future collaborative agents.
[LG-39] Leaky Language Models: Stealing Architecture and Inference Optimizations via Per-Token Timing
链接: https://arxiv.org/abs/2607.20723
作者: Sadegh Majidi,Niloofar Mireshghallah,Kazem Taram
类目: Cryptography and Security (cs.CR); Machine Learning (cs.LG)
*备注:
Abstract:This work presents LeakyLMs, a set of attacks that leak proprietary model, architecture, and deployment information from production language models. LeakyLMs is the first to demonstrate that key model and deployment details can be inferred using only token generation timing, even when interacting through remote APIs. LeakyLMs introduces two core attacks. The first attack targets inference optimizations and deployment strategies. For example, our attack detects whether a provider uses speculative decoding, a widely deployed inference-time optimization, and further identifies the context length of the draft model used in the pipeline. Our measurements show that Google Gemini Flash 2.5 uses speculative decoding with a draft context window of approximately 128K tokens. The second attack recovers key architectural properties, including the number of transformer layers, hidden dimension size, and number of attention heads. To achieve this, LeakyLMs builds a detailed and accurate model of token-generation timing on modern NVIDIA GPUs, characterizing how latency scales with model configuration and hardware parameters. The attack then performs a search over the architecture space using this timing model. In experiments with Llama models, the near-correct architectural configuration appears in the top-10 guesses more than 90% of the time.
[LG-40] Perspective Latents as an Architectural Condition for Causal Emergence in Active Inference Agents
链接: https://arxiv.org/abs/2607.20708
作者: Hongju Pae
类目: Machine Learning (cs.LG); Neurons and Cognition (q-bio.NC)
*备注:
Abstract:A recent line of work measures causal emergence in reinforcement learning agents through Integrated Information Decomposition, reporting that \Phi_r grows with training and tracks reward improvement. For active inference, this raises the question of how reward-free predictive organization relates to such information-theoretic signatures. I test this within an active inference agent whose architecture separates a fast perception latent z from a slow global latent g , where g is driven by prediction error and structurally decoupled from policy gradients. In a reward-free environmental regime-switching protocol, \Phi_r concentrates in g ; its aggregate magnitude is largely architectural and decreases with training. The substantive effect of learning becomes legible only at the atom-compositional level: decoupling flips sign from negative to positive and becomes regime-invariant under environmental change, while downward causation carries the regime-dependent adjustment. These results identify g as the architectural locus of \Phi_r -relevant temporal organization in an active inference agent, and argue against reading scalar \Phi_r as a direct index of learned integration.
[LG-41] CEDAR: Causal Edge Discovery for Autoregressive Processes
链接: https://arxiv.org/abs/2607.20696
作者: Mohammad Fesanghary
类目: Machine Learning (cs.LG); Methodology (stat.ME)
*备注: 8 main pages, 5 figures
Abstract:We propose CEDAR (Causal Edge Discovery for Autoregressive Processes), a constraint-based method for lagged causal edge discovery in sparse autoregressive time series. CEDAR screens candidate cross-variable lags using AR(1)-residualized, U-centered distance correlation, then applies two targeted conditional-independence tests per significant cross-variable lag candidate and accepts at most one lag per ordered pair. A stable MCI pruning step removes indirect edges, and optional deterministic C-nodes adjust for specified trend-like nonstationarity. In sparse regimes where few lags survive screening, CEDAR requires O(d^2) CI tests after screening while retaining edge-level interpretability. CEDAR is most effective when data are scarce and variables exhibit lag-1 self-dynamics; methods with richer conditioning sets become preferable as T grows or when higher-order autoregressive or simultaneous multi-lag effects are common.
[LG-42] Attribution Markets: A Fisher-Market Formulation for Fractional Credit Assignment Between Planned Tasks and Performed Actions
链接: https://arxiv.org/abs/2607.20694
作者: Salavat Ishbulatov
类目: Machine Learning (cs.LG); Computer Science and Game Theory (cs.GT)
*备注: 35 pages, 3 figures. Companion technical report with extended treatment of dynamic/forward-looking markets and reinforcement-learning extensions available separately
Abstract:Personal and organizational planning systems maintain two records that drift apart: what was planned (a task’s effort budget) and what was done (a logged action’s duration and description). Existing systems bridge them with an exclusive, all-or-nothing link that strands genuinely related but unlinked effort and reports false stalls on active goals. We formulate the bridge as a quasi-linear Fisher market: planned tasks are budget-constrained buyers, performed actions are divisible goods, and a fused text/structural/temporal signal sets each buyer’s valuation. Two market instruments - a seller reserve price and a buyer cash option - yield conservation, a hard budget cap, and a provable junk filter as theorems. We extend the market with a concave completion utility discounting progress as a task nears its plan; standard convergence theory for the market’s algorithm does not transfer here, resolved by a satiation-threshold fixed point with existence (Brouwer) and local uniqueness under an explicit diagonal-dominance condition, validated empirically on random and adversarial instances. A de-circularized, multi-seed benchmark - observed affinity corrupted independently of the scored ground truth - surfaces a genuine weak spot: the market’s sharp, zero-entropy equilibrium is more sensitive to affinity noise than entropy-regularized optimal transport’s permanently smoothed one. We resolve this with a one-parameter entropy-regularized generalization unifying the two, plus a noise-adaptive rule for its regularization strength. We report full reproducibility parameters, discuss limitations candidly, and relate the result to multi-touch attribution, optimal transport, and online Fisher-market algorithms.
[LG-43] Explanation-Based Runtime Verification for Trustworthy ML-driven Optical Networks
链接: https://arxiv.org/abs/2607.20675
作者: Omran Ayoub,Carlos Natalino,Ali Al Housseini,Felix Foschum,Philipp Morger,Tiziano Leidi,David Hock,Paolo Monti
类目: Machine Learning (cs.LG); Networking and Internet Architecture (cs.NI)
*备注: 6 Pages, 2 Figures. Accepted and presented at the 30th International Conference on Optical Network Design and Modelling (ONDM 2026), Munich, Germany, 12-15 May 2026
Abstract:Machine learning (ML) models are increasingly integrated into optical network automation frameworks to support tasks such as failure management, performance monitoring and resource allocation. In these environments, ML-driven predictions may be directly coupled with control-plane actions where incorrect decisions can immediately impact service quality, resource efficiency, and network stability. As automation levels increase, ensuring the reliability of individual decisions at deployment time becomes a critical requirement. Explainable artificial intelligence (XAI) techniques have emerged to improve transparency by highlighting the factors influencing ML predictions. In addition to identifying influential features, they provide insights into the underlying reasoning process of the model, revealing how different input variables contribute to the final outcome and how feature interactions shape the decision boundary. In this work, we introduce explanation-based runtime verification, an approach that exploits model explanations to assess the soundness of individual ML decisions before they are executed in the network control loop. The proposed approach evaluates explanation coherence and physics grounding consistency at runtime, enabling the system to defer or reject decisions flagged as uncertain. We demonstrate the effectiveness of our approach on a representative use case of lightpath quality of transmission classification. Experimental results show that explanation-based verification can intercept a significant fraction of erroneous decisions while preserving high automation rate.
[LG-44] End-to-End Learning of Safe Optimal Feedback Control in High Dimensions with Control Barrier Function Layers
链接: https://arxiv.org/abs/2607.20674
作者: Xingjian Li,Kelvin Kan,Deepanshu Verma,Krishna Kumar,Stanley Osher,Samy Wu Fung
类目: Machine Learning (cs.LG); Systems and Control (eess.SY); Optimization and Control (math.OC)
*备注:
Abstract:We consider the problem of learning high-dimensional semi-global feedback controllers under hard safety constraints enforced by control barrier functions (CBFs). Incorporating CBFs into end-to-end policy training requires embedding a quadratic-program-based safety filter as an optimization layer, but computational and differentiation bottlenecks have largely restricted prior approaches to low-dimensional systems, typically with at most 16 state dimensions. We address this limitation by combining operator splitting with the recently developed Jacobian-Free Backpropagation (JFB) method to enable scalable end-to-end training while preserving hard safety guarantees through the CBF safety filter. We justify this training methodology theoretically using nonsmooth analysis techniques and demonstrate its effectiveness on high-dimensional multi-agent nonlinear control problems with state and control dimensions up to 1200 and 400, respectively.
[LG-45] Cross-Domain Generalization in Optical Networks via Joint Contrastive and Classification Learning
链接: https://arxiv.org/abs/2607.20666
作者: Ali Al Housseini,Carlos Natalino,Paolo Monti,Omran Ayoub
类目: Networking and Internet Architecture (cs.NI); Machine Learning (cs.LG)
*备注: 6 pages, 2 figures. Accepted and presented at the 30th International Conference on Optical Network Design and Modelling (ONDM 2026), Munich, Germany, 12-15 May 2026
Abstract:The robustness of machine learning techniques across heterogeneous network domains remains an open challenge in optical networks. Models trained on data from a specific topology or operational configuration often exhibit degraded performance when deployed in unseen networks. In this work, we address this challenge by proposing a representation learning technique aimed at capturing task-relevant relationships that remain stable across domains. The proposed technique is based on a novel joint contrastive and classification learning approach in which representation learning and task optimization are performed simultaneously, allowing both objectives to shape the latent space. Experimental results on a representative use case, namely, lightpath quality of transmission estimation, demonstrate the effectiveness of our approach compared to baseline approaches, and highlight its capacity for rapid adaptation, providing excellent performance even with limited fine-tuning.
[LG-46] One Round Is All You Need: Analytic Federated Learning for Task-Heterogeneous Multi-Label Medical Image Classification ALT
链接: https://arxiv.org/abs/2607.20641
作者: Afsaneh Mahanipour,Hana Khamfroush
类目: Machine Learning (cs.LG)
*备注: The 14th IEEE International Conference on Healthcare Informatics
Abstract:Federated learning (FL) enables multiple clinical institutions to collaboratively train a shared disease classifier without centralizing patient data. In practice, however, each institution annotates only the pathologies within its area of expertise, so the federation operates under task heterogeneity: each client holds labels for a strict subset of the target disease categories while the remaining classes are entirely unobserved at that site. Existing gradient-based FL methods fail under this setting because they require hundreds of communication rounds to converge and because missing class labels introduce systematic false-negative bias that the model cannot correct without a principled mechanism. We propose an analytic federated learning framework for multi-label medical image classification under task heterogeneity. The proposed method replaces iterative gradient optimization with three closed-form operations: a balanced label projection that neutralizes class-imbalance bias by normalizing positive and negative contributions to equal total mass; a per-class absolute aggregation law that independently assembles the optimal ridge-regression classifier for each disease category from the sufficient statistics uploaded by its annotating clients; and an optional analytic pseudo-label refinement round that propagates missing-class knowledge from a confidence-filtered teacher classifier to non-annotating clients. The entire procedure requires at most two communication rounds, irrespective of the degree of task heterogeneity or the number of participating clients. Experiments on ChestXray14 under four progressively severe missing-class configurations demonstrate that the proposed method consistently outperforms the state-of-the-art federated multi-label method FedMLP by up to 18.44 BACC points and 13.24 AUC points, while reducing the communication.
[LG-47] Algorithmic Approaches to Sequential Decision-Making and Social Epistemology
链接: https://arxiv.org/abs/2607.20636
作者: Kavya Ravichandran
类目: Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG)
*备注: PhD Thesis
Abstract:As humans, we face many decisions that require us to choose between sticking to something and giving up. This thesis uses algorithmic tools to derive insights about such decision-making problems in theoretical models, studying both near-optimal methods and outcomes of social and behavioral influences. Along the way, this thesis sheds light on what we gain and what we lose as we move from a messy and complex real world setting to a very general abstract model by studying various points along this spectrum. In Part I, we study algorithms for sequential decision-making in the improving multi-armed bandits problem. We provide nearly matching upper and lower bounds in the general case. Then, we then ask what is possible if we have access to similar instances to the one we wish to deploy our algorithm on. To that end, we provide guarantees in the data-driven algorithm design framework, showing that a polynomial number of samples is sufficient for learning good algorithms from a class of algorithms. In Part II, we study algorithmic approaches for problems in social epistemology. We start by analyzing what role theoretical models can play in the study of social problems. We then study social and behavioral influences in decision-making requiring investment. First, we provide mathematical formalism in which to study the formation of pessimism traps, a phenomenon identified by philosophers in which agents are influenced by their predecessors to engage in less-ambitious goals. We develop financial interventions to sustainably shift communities out of these traps. The second problem we study is the influence of grit as a behavioral trait in ambitious decision-making. Overall, these works seek to theoretically model phenomena in social epistemology and provide a framework for intervening algorithmically. Comments: PhD Thesis Subjects: Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG) Cite as: arXiv:2607.20636 [cs.DS] (or arXiv:2607.20636v1 [cs.DS] for this version) https://doi.org/10.48550/arXiv.2607.20636 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-48] STeMP: Spatio-Temporal Modelling Protocol
链接: https://arxiv.org/abs/2607.20592
作者: Jan Linnenbrink,Jakub Nowosad,Marvin Ludwig,Anna Frederike Jablotschkin,Fabian Schumacher,Teja Kattenborn,Hanna Meyer
类目: Machine Learning (cs.LG)
*备注: 3 figures, 1 table
Abstract:Spatio-temporal machine-learning modelling is an important tool in environmental research. However, machine-learning models are highly sensitive to both the characteristics of the training data, such as its distribution, and methodological choices, including the cross-validation strategy. Each decision has impact and implications on the model itself as well as the estimation of the model quality and applicability for certain purposes. Taking into account the large role of machine-learning based maps of the environment in science and their transfer into practice, transparent reporting of spatio-temporal models, ideally using standardized model protocols, is essential to enable trust, transparency and comparability. However, such protocols are currently lacking for spatio-temporal modelling. We propose STeMP (Spatio-Temporal Modelling Protocol) to fill this gap by serving two purposes: standardized reporting to understand the model functioning as well as providing guidance during the modelling process by pointing at critical decisions and parameters. The protocol is structured in three sections: Overview, Model and Prediction. The Overview section contains metadata, while the Model and Prediction sections go into detail, describing predictors, evaluation and software, and further relevant elements of the modelling workflow. The protocol definition is hosted on GitHub and accompanied by an R-package (this https URL). The R-package contains a web application that can be used to fill the protocol either manually or in a semi-automated way from provided modelling objects. Warnings are returned from the protocol when common pitfalls are encountered, which may help authors as a guide through the modelling process but also support reviewers in the assessment of modelling studies. Via GitHub, incorporation of contributions and feedback from the community is encouraged. Comments: 3 figures, 1 table Subjects: Machine Learning (cs.LG) Cite as: arXiv:2607.20592 [cs.LG] (or arXiv:2607.20592v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2607.20592 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-49] Exact ReLU realization of affine one-dimensional refinement iterates via residual memory and offset frames
链接: https://arxiv.org/abs/2607.20586
作者: Boldsaikhan Bolorkhuu,Tsogtgerel Gantumur
类目: Machine Learning (cs.LG)
*备注:
Abstract:We study vector-valued affine refinement operators of the form [ (W\gamma)(t)=\sum_j\in\mathbbZ A_j\gamma(Mt-j)+B(t), ] with finitely supported matrix mask and compactly supported continuous piecewise linear input and forcing data. Building on the homogeneous realization theorem for (B\equiv 0), we prove that, for (M\ge 3), every finite affine iterate (W^n\gamma) admits an exact fixed-width ReLU realization whose depth is (O(n)). The main new ingredient is a residual memory controller. It replaces the noninvertible residual dynamics by an injective skew-product and permits exact backward replay of the residual states required by a Horner-type evaluation of the affine forcing sum. Offset frames align the forcing atoms away from residual seams, allowing complementary loop readouts to recover their values exactly. The remaining branch-selection ambiguity occurs only where the accumulated affine state has already vanished. For (M\ge 3), the result applies to arbitrary compactly supported continuous piecewise linear forcing terms. For (M=2), the same construction applies to ordinary-frame seam-separated forcing. We also prove a stage-dependent extension for forcing terms in a fixed finite-dimensional continuous piecewise linear span and record the resulting linear-depth upgrade for open-curve, finite-state, and Hilbert- and Morton-type recursive constructions. Subjects: Machine Learning (cs.LG) Cite as: arXiv:2607.20586 [cs.LG] (or arXiv:2607.20586v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2607.20586 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
[LG-50] Fisher Widths: Local Learning Geometry and Anisotropic Recovery
链接: https://arxiv.org/abs/2607.20578
作者: Vu Khac Ky
类目: Machine Learning (cs.LG); Statistics Theory (math.ST); Machine Learning (stat.ML)
*备注: 38 pages, 3 figures
Abstract:We study Gaussian-width complexity on statistical manifolds through a pair of functionals: the primal Fisher width w_G(T) = w(G^1/2T) , induced by the Fisher metric, and the inverse-Fisher width w_G^-1(T) = w(G^-1/2T) , induced by the inverse Fisher metric. The two widths play complementary statistical roles. On the learning side, the Fisher width measures the size of local parameter fluctuations in the geometry induced by the Fisher information. For Fisher-regular losses, we prove that the scale (w_G(H_r)/\sqrt n) is attained on sufficiently small Fisher balls. On the recovery side, the inverse-Fisher width captures the effect of anisotropic Gaussian measurements whose covariance is determined by the inverse Fisher information. For sparse recovery, the resulting geometry depends not only on sparsity but also on the position of the active coordinates in the Fisher spectrum. We obtain a two-sided estimate for the corresponding statistical dimension, together with support-sensitive recovery estimates and a natural ordering of supports with different curvature profiles. Finally, we establish a sharp relation between the primal and inverse-Fisher widths. On any common compact coordinate set T , they satisfy [ w_G(T)w_G^-1(T)\geq w(T)^2. ] Thus, Fisher anisotropy may transfer complexity from one geometry to the other, but cannot reduce both widths relative to the Euclidean scale. Comments: 38 pages, 3 figures Subjects: Machine Learning (cs.LG); Statistics Theory (math.ST); Machine Learning (stat.ML) Cite as: arXiv:2607.20578 [cs.LG] (or arXiv:2607.20578v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2607.20578 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Ky Vu Khac [view email] [v1] Wed, 22 Jul 2026 10:07:03 UTC (85 KB)
[LG-51] AI-Driven Surrogate Models for Predicting Electrode-Scale Discharge Behavior in Lithium-Ion Batteries
链接: https://arxiv.org/abs/2607.20577
作者: Mengda Xing(CRIL, UA),Jean-Marie Lagniez(CRIL, UA),Alejandro Franco(LRCS)
类目: Machine Learning (cs.LG)
*备注:
Abstract:Physics-based simulations are essential for understanding the electrode-scale discharge behavior of lithium-ion batteries (LIBs) but suffer from prohibitive computational costs. To address this, we introduce a novel deep learning surrogate pipeline based on the Swin3D Transformer to predict spatiotemporal discharge dynamics directly from volumetric data. Our approach integrates two key innovations: Gaussian Positional Encoding (GPE), which enhances spatial feature representation by adapting to the complex geometry of electrode microstructures, and a specialized Temporal Encoding module to capture non-linear timeseries evolution. Experimental validation on an Electrochemical Simulation (ES) dataset demonstrates that our pipeline significantly outperforms state-of-the-art point cloud baselines in prediction accuracy. Furthermore, the proposed method reduces the computational overhead by orders of magnitude, providing a scalable and efficient framework for high-throughput battery design and optimization.
[LG-52] Chronofy: A Temporal-Logical Decay Architecture for Information Validity in Time-Aware Retrieval-Augmented Generation
链接: https://arxiv.org/abs/2607.20560
作者: Muntaser Syed,Marius Silaghi,Sheikh Abujar,Sharun Akter
类目: Machine Learning (cs.LG)
*备注:
Abstract:Retrieval-Augmented Generation (RAG) systems retrieve and integrate external knowledge to ground large language model (LLM) outputs. However, current RAG architectures treat all retrieved facts as equally valid regardless of temporal provenance, leading to temporal hallucination, where plausible but obsolete facts corrupt the output. A clinical lab reading from yesterday is actionable; the same reading from six months ago is noise. We present Chronofy, a three-layer neuro-symbolic framework implementing the Temporal-Logical Decay Architecture (TLDA) that embeds temporal validity directly into the representation, retrieval, and reasoning layers of RAG systems. Layer 1 reserves a dedicated temporal subspace within Matryoshka embeddings to make fact age structurally irremovable from the representation. Layer 2 integrates learnable exponential decay functions into graph-based retrieval, where the decay coefficient \beta_j is grounded in Bayesian decision theory as an approximation of twice the latent process mean-reversion rate. Layer 3 applies Signal Temporal Logic (STL) robustness functions to evaluate the temporal validity of retrieved knowledge, not LLM output confidence, and enforces the possibilistic weakest-link principle to bound output confidence by the most decayed evidence in the reasoning chain. We evaluate Chronofy on temporal knowledge graph forecasting benchmarks, the TimE temporal QA benchmark, and a domain-specific sensitivity analysis, demonstrating that explicit temporal decay modeling improves retrieval precision, reduces temporal hallucination, and enables principled data re-acquisition triggers when temporal context is insufficient.
[LG-53] Double-Scoring: Reliable Extraction of Strong Lottery Tickets
链接: https://arxiv.org/abs/2607.20555
作者: Bryce A. Christopherson,Jack Baretz,Darian Colgrove,Salah Dandan
类目: Machine Learning (cs.LG); Optimization and Control (math.OC)
*备注: 46 pages
Abstract:The lottery ticket hypothesis proposes that large random neural networks contain sparse subnetworks that can match the performance of dense models after comparable training. A stronger version asserts that sufficiently overparameterized random networks contain subnetworks that are already accurate before any weight training. Existing theory establishes that such strong lottery tickets exist, but reliable extraction remains difficult. We revisit edge-popup, a frozen-weight score-training method for extracting strong tickets, and identify layerwise sparsity selection as a central bottleneck. We introduce double-scoring, an augmented score-space parameterization that replaces a layerwise sparsity search with optimization over enlarged score tensors. We prove that fixed-density masking in an augmented score space preserves access to all original-coordinate masks, and we show that the resulting method can be interpreted as edge-popup on a zero-augmented network. In controlled experiments, double-scoring substantially improves strong-ticket extraction over fixed-density edge-popup and pruning-at-initialization baselines, improves on the performance of rewound sparse-training topologies, and exhibits markedly lower sensitivity to sparsity hyperparameters. Ablations show that the gain is not merely due to additional trainable score parameters, but is tied to the augmented score-space competition that induces the effective original sparsity.
[LG-54] hermodynamic Weight Decay: Exploring Grokking Acceleration via Attention Specific Heat
链接: https://arxiv.org/abs/2607.20552
作者: Chitraansh Pandey
类目: Machine Learning (cs.LG)
*备注: 9 pages, 4 figures, 2 tables. Code and data: this https URL
Abstract:Grokking – the delayed generalization of neural networks long after they have memorized their training data – wastes thousands of training epochs and is notoriously unpredictable. Building on the recent result that Transformer attention is formally isomorphic to a thermodynamic system, we treat the variance of attention logits as a specific heat Cv and show that its peak reliably precedes the generalization transition. We introduce CvAdamW, a drop-in AdamW variant that monitors Cv online and injects thermal energy by dynamically scaling weight decay when a phase transition is detected. Through a strictly iterative development process we identify three failure modes – initialization noise, mini-batch micro-ripples, and slingshot blinding – and resolve them with a memorization gate and an exponential-moving-average shock absorber. On modular arithmetic (a+b mod 97), CvAdamW enables grokking at epoch 2802 in a 4000-epoch budget where the baseline never groks. We further propose a scale-invariant z-score reformulation that removes task-specific hyperparameters, and evaluate it across 10 paired seeds. A paired analysis shows the cold-start variant reduces mean grokking latency by 257 epochs (6.0%; median 166 epochs; Wilcoxon p=0.049, Cohen’s d=0.68, bootstrap 95% CI [53,489]), improving 8 of 10 seeds; on this single task Cv peaks before grokking in all 10 seeds. Our results indicate that neural networks may expose detectable precursors of impending generalization transitions, and that a physically motivated, proportional intervention can facilitate generalization within a fixed compute budget. Code and data are public.
[LG-55] SevDiff: Severity-Conditioned Diffusion for Long-Tail Conflict Trajectory Generation
链接: https://arxiv.org/abs/2607.20549
作者: Eni Solomon Laughter
类目: Machine Learning (cs.LG); Robotics (cs.RO); Systems and Control (eess.SY)
*备注:
Abstract:Trajectory datasets used in ADAS evaluation are heavily biased toward routine driving; genuine vehicle-to-vehicle conflict events are rare, and the rarer the event, the higher the cost when an ADAS system fails to handle it. Existing generative approaches address this imbalance by conditioning on scene-level properties - spatial goals, agent structure, or natural-language adversarial objectives - but none can accept a target Time-to-Collision (TTC) value as input and be held to producing it within a measurable error. This paper introduces SevDiff, a severity-conditioned denoising diffusion probabilistic model (DDPM) that accepts a requested minimum TTC value as a scalar conditioning signal and generates paired vehicle interaction trajectories whose realized conflict severity matches the request, evaluated through a hit-rate metric. Trained on 468 interaction windows extracted from the UTE SQM-W-1 expressway weaving-section dataset (1,041 vehicles, 822,691 observations after smoothing), SevDiff achieves 100% hit-rate within +/-0.5 s for TTC targets of 0.5-1.5 s and 97-99% at 2.0-2.5 s, with graceful degradation to 39% at TTC = 5.0 s. Generated kinematic features are physically plausible, with a maximum out-of-range rate of 4.7% across 12 features and no negative speed or gap values in more than 96.5% of samples. The hit-rate degradation pattern is physically interpretable as the strength of the conditioning signal relative to the training prior, making it a precision characterization of the generator rather than a pass/fail result.
[LG-56] Conflict Resolution under Degraded Surveillance in Air Corridors Using Multi-Agent Reinforcement Learning
链接: https://arxiv.org/abs/2607.20547
作者: Esrat Farhana Dulia,Syed Arbab Mohd Shihab,Caleb Adams,Ruben Del Rosario
类目: Machine Learning (cs.LG)
*备注:
Abstract:Safe Advanced Air Mobility operations require aircraft to maintain separation when surveillance information is noisy, delayed, incomplete, or temporarily unavailable. This study develops a Deep Q-Network-based Multi-Agent Reinforcement Learning framework for decentralized conflict resolution among heterogeneous small unmanned aerial vehicles and electric vertical takeoff and landing aircraft operating within a structured three-dimensional corridor. Separate policies are trained for the two aircraft categories using local observations and a 14-action space that includes maintaining course, turning, vertical maneuvering, landing, and speed control. The simulation incorporates aircraft-specific dynamics, energy use, corridor constraints, observation noise, communication delay, information dropout, wind disturbance, actuator uncertainty, and model uncertainty. The trained policies are evaluated across 90 combinations of traffic density and minimum separation thresholds. Loss-of-separation frequency and duration generally increase with traffic density and separation requirements, although most events are resolved within 1s. Under safe conditions, agents maintain their motion approximately 79% of the time. During conflicts, turning accounts for 33% of actions, followed by maintaining motion at 29%, speed control at 25%, and vertical maneuvers at 13%. Six Pareto-optimal configurations reveal trade-offs between safety and corridor capacity. The framework supports the simulation-based evaluation of safer AAM conflict-resolution strategies under degraded surveillance conditions.
[LG-57] Do Active SAE Feature Planes Carry More Holonomy? A Preregistered Reversal in Gemma
链接: https://arxiv.org/abs/2607.20522
作者: Larry Richards
类目: Machine Learning (cs.LG)
*备注: 23 pages, 5 figures. Preregistered confirmatory study; frozen experiment record archived at Zenodo, doi: https://doi.org/10.5281/zenodo.20960027
Abstract:This paper tests whether holonomy concentrates on active sparse-autoencoder (SAE) feature planes in Gemma 2 2B, a concrete operationalization of the broader semantic-concentration prediction. Holonomy is measured at the final-token layer-12 to layer-13 residual-stream readout by carrying a local frame around small loops using the instrument’s restricted-Jacobian transport rule, then normalizing the resulting rotation by enclosed area. The design, materiality threshold, analysis, and verdict rules were preregistered and frozen before the analysed measurements were inspected. The prediction was falsified in reverse: active-feature planes carried less holonomy than matched mixed-feature controls, with an adjusted log contrast of -0.29439 and 95% interval [-0.43989, -0.14889]. A magnitude-only explanation was not supported in this design, while the three-way ordering across random, mixed-feature, and active-feature planes was undefined at matched magnitude because common support failed. Post-freeze diagnostics at the same readout supported the area law on a small validation subset, bounded matched-center displacement under a simple paired regression, and identified transport distortion as a live mechanism or confound. The result is therefore a narrow, auditable operational reversal, not a causal claim that meaning suppresses holonomy. The cause remains open, with activation-strength geometry, degree of feature engagement, dictionary geometry, matched-center displacement, activation-manifold proximity, and transport shear as live alternatives.
[LG-58] Generative Bayesian Filtering for State Estimation
链接: https://arxiv.org/abs/2607.20521
作者: Lei Cao,Sihang Feng,Jixin Yan,Tao Sun,Naichen Shi
类目: Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注:
Abstract:The state of a dynamic system evolves over time, switching among several latent modes that govern its observable behavior. Filtering methods infer the latent state from observations. Classical filtering approaches, including Kalman filters, typically rely on simple observation models, such as linear-Gaussian models, that are incapable of characterizing the increasingly nonlinear and heterogeneous patterns in high-dimensional sensor signals. To tackle the challenge, we propose Generative Bayesian Filtering (GBF), a filtering framework that replaces restrictive observation models with pretrained conditional generative models parametrized by conditional variational autoencoders (CVAE). For online inference, GBF performs a Bayesian prediction-update recursion in which the measurement update is formulated as a posterior sampling problem that combines the dynamical prior with the CVAE-induced likelihood. The resulting filtering problem is then transformed into a score-based sampling problem, which naturally inherits the flexibility from generative models and the uncertainty quantification capabilities from ensembling. Experiments on synthetic datasets and real-world applications involving manufacturing system monitoring and arrhythmia diagnosis demonstrate that GBF improves state estimation accuracy and robustness relative to baseline approaches.
[LG-59] Adaptive Depth in Looped Transformers: Diagnosing Learned Halting Gates and Trajectory Readouts
链接: https://arxiv.org/abs/2607.20519
作者: Andrei Cristian Popescu,Haitz Sáez de Ocáriz Borde,Pietro Liò
类目: Machine Learning (cs.LG)
*备注:
Abstract:Looped Transformers increase test-time computation by repeatedly applying a shared recurrent block. Learned halting objectives in looped Transformers typically use a single exit distribution both as the inference-time stopping rule and as the training-time weighting of per-depth losses. This entangles exit selection with trajectory formation: the gate not only chooses which recurrent state to use, but also determines how strongly each intermediate state is supervised. Consequently, poor adaptive-compute performance can arise from the readout, the induced trajectory, or their interaction. We study adaptive depth in looped Transformers through this trajectory–readout lens, across controlled synthetic tasks (modular arithmetic and binary parity) and large-scale Ouro-1.4B and 2.6B checkpoints. We find that fixed-prior depth supervision, which shapes the trajectory without an input-dependent halting policy, produces difficulty-aware trajectories whose intermediate states expose useful stopping signals, and that simple post-hoc confidence readouts often match or outperform learned linear and MLP gates. Fitting gates on frozen trajectories localizes the failure: it appears to stem mainly from the trajectory induced by joint gate training rather than from limited gate expressivity. The same pattern is present in Ouro evaluations, where pretrained ponder gates are competitive but not uniformly Pareto-optimal, and measured latency confirms that the resulting reductions in average exit depth translate into practical inference-time savings. Our systematic diagnostic evaluation reframes adaptive depth in looped Transformers as a joint problem of trajectory formation and exit readout, rather than gate learning alone, highlighting a distinction that prior learned-halting work has often left implicit.
[LG-60] Multimodal CoLRAG -TF: Triple-Filtered Retrieval for Complex PDFs
链接: https://arxiv.org/abs/2607.20517
作者: Takato Yasuno
类目: Machine Learning (cs.LG)
*备注: 18 pages, 8 tables, 9 figures
Abstract:Retrieval-augmented generation (RAG) over heterogeneous PDF collections remains challenging due to multimodal content, domain-specific terminology, and the need for multi-hop reasoning across dispersed evidence. We present Multimodal CoLRAG-TF, a four-axis fusion architecture that integrates dense text embeddings, BM25 keyword matching, knowledge-graph triple filtering, and image-based similarity for robust retrieval over complex documents. Our system constructs a multimodal index of 2,403 blocks extracted from 43 Japanese disaster lesson PDFs, supported by a hybrid OCR pipeline and LLM-based caption generation. To enhance compositional reasoning, we extract 11,414 OpenIE triples and index them with FAISS, enabling sub-second triple lookup and hierarchical propagation of relevance signals. A HippoRAG2-inspired coarse-to-fine retriever (volume \to chapter \to block) narrows the search space before final fusion scoring. Bayesian optimization over fusion weights reveals that the triple axis must dominate ( \alpha_\texttriple = 0.44 ) to counteract lexical bias and sustain multi-hop retrieval quality. Evaluated on a 457-pair benchmark, Multimodal CoLRAG-TF achieves a Retrieval Recall of 0.9909 and a 71.6 % improvement in multi-hop answer similarity over single-hop queries. An image-to-lesson pipeline using a vision LLM further demonstrates the applicability of the approach to visual inputs. These results show that triple-filtered multimodal fusion is essential for structured reasoning over noisy, heterogeneous PDFs and provides a general framework applicable beyond the disaster domain.
[LG-61] HERMES: Heterogeneous Edge-Relational Multi-Head Embedded SSM Attention for Traffic Conflict Prediction at Signalized Intersections
链接: https://arxiv.org/abs/2607.20505
作者: Md Monzurul Islam,Subasish Das
类目: Robotics (cs.RO); Machine Learning (cs.LG)
*备注:
Abstract:Surrogate safety measures (SSMs) enable proactive traffic safety assessment, but many existing methods evaluate pairwise interactions independently or flatten multi-agent scenes into fixed feature vectors, limiting their ability to represent heterogeneous interaction structure and evolving scene-level risk. This study formulates traffic conflict assessment as temporal heterogeneous scene-graph classification and proposes HERMES, a heterogeneous edge-relational graph neural network with SSM-informed multi-head attention. Vehicles and pedestrians are represented as heterogeneous nodes, while vehicle-vehicle, vehicle-pedestrian, and pedestrian-pedestrian interactions are encoded as relation-specific edges with continuous kinematic and surrogate-safety descriptors. Relation-specific attention, dynamic node-edge updates, safety-aware graph pooling, and temporal sequence learning are jointly used to estimate scene-level conflict probability. HERMES was evaluated using 109,028 trajectory-derived sequences from a signalized urban intersection and tested on an independently collected comparable intersection dataset. Enhanced HERMES achieved an AUC-ROC of 0.9898 +/- 0.0013, an AUC-PR of 0.9412 +/- 0.0067, and an F1 score of 0.8449 +/- 0.0103. At a 5% false-alarm rate, it detected 95.7% of conflict sequences, outperforming the strongest Transformer baseline and XGBoost. In zero-shot external evaluation, HERMES achieved an AUC-ROC of 0.9752 and an AUC-PR of 0.7829. Joint source-target training further improved target-site performance with limited target-site data. These findings show that preserving heterogeneous interaction topology, safety-informed edge semantics, and short-term temporal evolution improves scene-level conflict classification and supports transferable roadside safety monitoring at signalized intersections.
[LG-62] Neural solutions of coupled ghost and gluon Dyson–Schwinger equations in Landau gauge
链接: https://arxiv.org/abs/2607.21548
作者: Rodrigo Carmo Terin
类目: High Energy Physics - Phenomenology (hep-ph); Machine Learning (cs.LG); High Energy Physics - Lattice (hep-lat); High Energy Physics - Theory (hep-th)
*备注: 17 pages, 10 figures, 7 tables
Abstract:The coupled ghost and gluon Dyson–Schwinger equations (DSEs) of four-dimensional Landau-gauge Yang–Mills (YM) theory are solved with a neural representation trained only from renormalized equation residuals. The neural and fixed-point solutions agree at the percent level and remain stable under changes of initialization, network size, integration grid, and infrared boundary condition. Variations of the three-gluon vertex model produce substantially larger effects than the neural error. The MiniMOM ultraviolet running and the sign change of the gluon Schwinger function are also reproduced within the limitations of the truncation.
[LG-63] oward Generalizable Cognitive Impairment Detection with Speech-Based Multimodal Large Language Models
链接: https://arxiv.org/abs/2607.21496
作者: Yingchao Huang,Xin Wang,Yuhan Su,Shanshan Yao
类目: ignal Processing (eess.SP); Machine Learning (cs.LG)
*备注:
Abstract:Cognitive impairment (CI) is a growing public health concern. Early and accurate diagnosis is critical for enabling timely intervention and improving patient outcomes. Speech-based CI detection has emerged as a promising non-invasive approach, as speech signals encode both linguistic and acoustic markers associated with cognitive decline. Recent advances in large language models (LLMs) further strengthen the potential of speech-based assessment by enabling more expressive representation learning and improved generalization across diverse speakers, recording devices, and clinical environments. Moreover, multimodal learning by jointly modeling linguistic and acoustic features allows for a more comprehensive characterization of cognitive and behavioral changes related to CI, leading to more reliable detection. In this work, we propose a multimodal CI detection framework based on open-source LLMs that integrates speech audio and corresponding transcripts while preserving patient privacy. Acoustic embeddings are extracted directly from speech signals, while textual embeddings are generated from automatically transcribed speech. These modality-specific embeddings are then concatenated to create a combined feature vector and used for downstream classification, without requiring access to raw or sensitive patient data. The proposed approach is evaluated on the ADReSS20 and ADReSSo21 benchmark datasets. Experimental results show that the proposed multimodal framework achieves an CI classification accuracy of 92.4% and consistently outperforms single-modality baselines. Our work establishes a new state-of-the-art for CI identification, with the proposed method demonstrating superior cross-dataset generalization. This advance highlights the power of an LLM-based multimodal framework that fuses linguistic and acoustic data to enable robust, scalable, and non-invasive screening.
[LG-64] Climate-resilient electric vehicle charging infrastructure for sustainable cities: An interpretable causal-ensemble framework for preventive maintenance and low-carbon mobility
链接: https://arxiv.org/abs/2607.21444
作者: Cande Lian(1),Wentao Zeng(1),Jiabin Wu(1),Yiming Bie(2),Wei Zhou(3) ((1) School of Management, Foshan University, Foshan, China (2) School of Transportation, Jilin University, Changchun, China (3) Department of Civil and Environmental Engineering, National University of Singapore)
类目: Optimization and Control (math.OC); Machine Learning (cs.LG)
*备注:
Abstract:Reliable electric vehicle (EV) charging infrastructure is a cornerstone of sustainable, low-carbon cities, yet urban climate stress such as extreme heat, heavy precipitation, and humidity increasingly raises equipment fault risk and undermines the resilience of urban energy and mobility services. Shifting operation from reactive repair to preventive maintenance depends on accurate, forward-looking fault-risk prediction, a task complicated by the heterogeneous time scales of physical, behavioral, contextual, and historical signals and by forecasting over a multi-week horizon. We develop FGDSE, a feature-governed dynamic stacking ensemble that forms an interpretable decision-support system for climate-resilient charging-asset management. It partitions heterogeneous signals into four feature families, assigns each to a domain expert whose inductive bias matches the data, and adds two deep temporal experts for short-term pulses and long-term degradation; a horizon-wise gating mechanism then learns adaptive weights to forecast daily fault risk over 1 to 30 days. SHAP attribution and an X-learner extend the probabilistic output into causal decision support with post-level treatment effects. On 25 months of data from 13 stations, FGDSE surpasses twelve baselines beyond the ten-day horizon, sustains about 85% macro-recall at 30 days with an AUC decay of only 3.2 points, and reveals a shift of dominance from fault history toward climate stress. It identifies extreme heat as the sole exposure whose causal effect amplifies over time, flagging roughly 30% of posts as heat-sensitive and yielding quantitative thresholds for climate-adaptive maintenance that strengthens urban mobility resilience and sustains low-carbon travel.
[LG-65] Cautious optimism for deep parameterized quantum circuits
链接: https://arxiv.org/abs/2607.21409
作者: Marie Kempkes,Elies Gil-Fuster,Carlos Bravo-Prieto,Aroosa Ijaz,Alissa Wilms,Jens Eisert,Evert van Nieuwenburg,Vedran Dunjko
类目: Quantum Physics (quant-ph); Machine Learning (cs.LG); Machine Learning (stat.ML)
*备注: 21 pages (6+15), 2 figures (1+1), comments welcome
Abstract:A central challenge in quantum machine learning is understanding the scaling behavior of parameterized quantum circuits (PQCs). In particular, it remains unclear how their performance on unseen data changes as the number of trainable parameters increases. Prior works have derived formal generalization guarantees for quantum models, but it is well-known that many such results do not fully characterize generalization behavior in practice. In this work, we show that gradient-based PQCs can exhibit improved performance on unseen data as model size increases, displaying the phenomenon of double descent. This contrasts with the traditional view that larger models lead to degraded generalization. We provide analytical results rigorously underpinning this behavior by leveraging add-one-in perturbation techniques and spectral properties of random matrices. We support these results with numerical experiments on re-uploading PQCs across several data sets and training set sizes, consistently observing the predicted double descent behavior. While other obstacles on the path toward practical quantum machine learning remain, our finding that deeper parameterized quantum circuits do not necessarily exhibit degraded performance provides reasons for cautious optimism.
[LG-66] ransformer-based Diffusion models for Hydrological Time Series Probabilistic Imputation and Forecasting
链接: https://arxiv.org/abs/2607.21200
作者: Ferdinand Bhavsar(INRAE),Lionel Benoit(INRAE),Maxime Savatier(ANDRA),Edith Gabriel(INRAE)
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注:
Abstract:The modeling of hydrometeorological time series with limited observations is a key challenge in the monitoring of hydro-systems and water resources, as well as for flood or drought risk assessment. Due to the high variability of the underlying processes and the sparsity of available measurements, traditional statistical approaches often struggle to accurately represent their dynamics. In this context, recent advances in deep learning offer a promising direction for improving the representation and generation of complex temporal processes sampled at several observation sites. This study investigates the application of transformer-based diffusion models to the simulation and reconstruction of hydrological time series. The proposed framework is applied to the joint modeling of water quantity and quality at six sites spread across three adjacent headwater catchments located in North-East France on a limestone plateau covered by forests and field crops. The model is calibrated and validated using available observational data, which has been quality controlled and corrected for sensor drift and malfunction through collaborative efforts by LNE metrology expertise and Andra monthly quality control over more than 15 years. Its performance is compared with several established baseline approaches commonly used for time series modeling. Quantitative evaluation metrics are employed to assess the ability of the proposed method to reproduce key temporal characteristics of the observed signals in two settings: the imputation of incomplete time series and the forecasting of upcoming hydrological conditions. Results support the effectiveness of the transformer-based approach and highlight its capacity to capture and simulate the complex patterns present in hydrological data. In particular, the results indicate that diffusion models can efficiently sample realistic time series distributions under observation settings with variable missing data for both forecasting and imputation.
[LG-67] Approximate Quantum State Preparation Through Proximal Policy Optimization
链接: https://arxiv.org/abs/2607.21121
作者: Marco Mordacci,Michele Amoretti
类目: Quantum Physics (quant-ph); Emerging Technologies (cs.ET); Machine Learning (cs.LG)
*备注:
Abstract:In this work, a quantum architecture search framework for approximate quantum state preparation (QSP) is proposed. QSP is a challenging task, since the search space grows exponentially with the number of qubits, making the identification of the optimal circuit non-trivial. To address this problem, deep reinforcement learning is employed through an agent based on proximal policy optimization. The objective of the agent is to identify the best possible approximation of the target state while simultaneously minimizing the number of gates used. At each step, the agent appends a new gate to the circuit and recomputes the fidelity between the approximated state and the target states. Various experiments have been performed from 2 to 5 qubits. Both predefined states, such as Bell, GHZ, W, and Dicke states, and completely random states are considered. The proposed framework is able to achieve approximation errors of 10^-14 .
[LG-68] Automatic knot selection in smooth additive models
链接: https://arxiv.org/abs/2607.21083
作者: Nicolás Carrizosa,Vanesa Guerrero,María Durbán
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 43 pages (29 of which are the main document, the rest are part of the appendix), 31 figures (5 in main document)
Abstract:B-spline regression constitutes a widely used framework for nonparametric modeling. The performance of this methodology depends on specifying the number and placement of changepoints, known as knots, prior to the estimation process. Such knot sequence determines the dimension of the B-spline basis used to represent the regression function and the number of coefficients to be estimated. Therefore, the knots’ choice affects the model’s flexibility, influencing its smoothness and goodness-of-fit. Traditionally, this problem has been addressed either by explicitly selecting knots, via knot-selection algorithms, or by regularization methods, such as P-splines, which automatically tune the regressor’s smoothness. The latter have become the standard in generalized additive models (GAMs). In contrast, knot-selection techniques, frequently neglected because of computational or modeling limitations, provide certain advantages which can be valuable in some contexts. In this work, we introduce a novel explicit knot-selection technique for GAMs based on an extension of the adaptive splines (A-splines) knot selection methodology, combined with a customized Fellner-Schall scheme for tuning the associated parameters. Our approach is evaluated on various synthetic and real datasets and compared with P-splines and state-of-the-art knot-selection techniques. The results indicate comparable performance, while producing models built on a substantially smaller number of basis elements.
[LG-69] An Analytically Trained Variational Surrogate for Quantum Phase Estimation on NISQ Hardware
链接: https://arxiv.org/abs/2607.20943
作者: Mousumi Kundu,Ashish Kumar Patra,Anurag K. S. V.,Ruchika Bhat,Sai Shankar P.,Alok Shukla,Jaiganesh G
类目: Quantum Physics (quant-ph); Emerging Technologies (cs.ET); Machine Learning (cs.LG); Chemical Physics (physics.chem-ph)
*备注: 30 pages, 12 figures
Abstract:Quantum Phase Estimation (QPE) is a foundational algorithm for molecular ground-state energy estimation, but its deep circuit requirements make direct hardware execution impractical on Noisy Intermediate-Scale Quantum (NISQ) devices. We present an analytically grounded variational surrogate framework in which a shallow Variational Quantum Circuit (VQC) is trained to reproduce the QPE measurement distribution without any quantum circuit simulation. The training target is computed entirely classically via the Dirichlet kernel, evaluated directly from the Full Configuration Interaction (FCI) ground-state energy, the ancilla qubit count, and the time evolution parameter, eliminating the exponentially scaling simulation bottleneck of prior surrogate approaches. We apply this framework to the hydrogen molecule (H _2 ) with a symmetry-tapered Hamiltonian, conducting a four-stage experimental investigation on IBM Quantum hardware. Stage 1 compares linear and full entangler topologies for the R_Y - R_Z - CZ ansatz, with and without XpXm Dynamical Decoupling (DD), across four distributional metrics (Hellinger distance, fidelity error, total variation distance, Jensen-Shannon divergence), identifying the linear entangler as optimal. Stage 2 varies VQC layers ( p=1 to 5 ) for the linear-entangler ansatz, identifying single-layer depth as optimal under hardware noise. Stage 3 applies this configuration to the reduced R_Y - CZ ansatz, comparing ideal and noisy simulator-trained parameters. A supplementary noise analysis at p \in \8,64\ characterizes the depth-dependent interplay between circuit depth and DD effectiveness. The framework enables faithful QPE mimicry using a linearly scaling VQC, recovering the ground-state energy within the chemical accuracy threshold (1 kcal/mol), constituting a scalable, hardware-efficient paradigm for QPE-based molecular energy estimation on NISQ devices.
[LG-70] RadioTrace: Transmitter-Aware Diffusion for Radio Map Estimation without Deployment-Time Fine-Tuning
链接: https://arxiv.org/abs/2607.20909
作者: Liu Yang,Qiang Li,Zhuo Cao,Weijie Xiong,Guomin Sun,Jingran Lin
类目: ignal Processing (eess.SP); Machine Learning (cs.LG)
*备注: IEEE Trans. Wireless Comm
Abstract:Radio map (RM) estimation aims to reconstruct the spatial distribution of wireless signal characteristics, such as received signal strength (RSS), from sparse measurements, a task that is critical for spectrum management, interference mitigation, and localization in modern wireless networks. Traditional approaches, including interpolation and deep learning, either struggle to capture complex propagation effects or require large-scale retraining for each new sampling pattern, which limits their generalization. More recently, prior-based methods have combined pre-trained generative models with measurements to reduce the need for deployment-time model fine-tuning, but they typically treat the prior as a simple regularizer and lack explicit transmitter-aware integration. In this paper, we propose RadioTrace, a novel RM estimation framework without deployment-time fine-tuning that tightly integrates sparse RSS measurements with a frozen pre-trained diffusion prior. RadioTrace incorporates transmitter (Tx) location estimation directly into the denoising loop, iteratively refining Tx coordinates based on reconstruction quality to guide the generative process. To further enhance robustness, we introduce a propagation-guided K-means initialization that mitigates poor local minima in the Tx update and provides a geometry-consistent starting point. Moreover, we provide a stochastic stability analysis for the Tx-coordinate refinement component, showing that the Tx update remains stable under perturbations induced by diffusion sampling and Tx-map relaxation. Extensive experiments demonstrate that RadioTrace achieves competitive performance with state-of-the-art learning-based methods under random sampling, and maintains strong reconstruction quality under restricted-area sampling, highlighting its adaptability, robustness, and practical relevance.
[LG-71] Machine Learning for Charge State Characterization of Isolated Double Quantum Dots
链接: https://arxiv.org/abs/2607.20871
作者: Hyma Vallabhapurapu,Marco Candido,Krishna Choudhary,Paul Steinacker,Ensar Vahapoglu,Chris Escott,Wee Han Lim,Andre Saraiva,Nard Dumoulin Stuyck,MengKe Feng
类目: Mesoscale and Nanoscale Physics (cond-mat.mes-hall); Machine Learning (cs.LG)
*备注:
Abstract:Scaling semiconductor quantum dot arrays toward fault-tolerant quantum computing requires efficient tuneup of spin qubits, a process that depends on the analysis of charge stability maps (CSMs) and remains largely manual. While machine learning has been widely applied to CSM analysis in reservoir-coupled devices, automated tuning in the increasingly important isolated-mode regime has received limited attention. In isolated-mode CSMs, charge transitions appear as near-vertical lines, making them well suited to compact, task-specific models. We present two convolutional neural networks with fewer than one million parameters, trained on CSMs collected from 32 silicon metal-oxide-semiconductor (SiMOS) double-quantum-dot devices measured at approximately 1 K using an automated cryogenic probing system. Sixteen devices were used for training and sixteen were held out to evaluate cross-device generalization against hand-labeled ground truth. CSMClassifier identifies charge instability and sensor artifacts, achieving 94% macro-averaged accuracy across three quality classes on 2,407 held-out images. ChargeLineNet localizes charge-transition lines and determines electron occupancy, achieving 95.3% exact line-count accuracy on 1,131 held-out images. Combined into a single pipeline, the models correctly determine electron occupancy for 93.8% of clean held-out images. Pre-training on synthetic images substantially improves label efficiency. Fine-tuning the pre-trained model on limited experimental data maintains over 90% accuracy, whereas training from scratch degrades significantly under the same conditions. Together, the two models occupy only 6.5 MB and process images in less than 60 ms on standard laboratory hardware, demonstrating a practical path toward scalable, automated characterization and tuneup of quantum-dot devices.
[LG-72] Writhe-Based Polymer Link Classification Using Machine Learning
链接: https://arxiv.org/abs/2607.20657
作者: Jack Beda,Djordje Mihajlovic,Kasturi Barkataki,Davide Michieletto
类目: Geometric Topology (math.GT); Soft Condensed Matter (cond-mat.soft); Machine Learning (cs.LG)
*备注: 8 pages, 4 figures
Abstract:Unique and rapid classification of knots and links is an open mathematical problem that is relevant to a range of (bio)physical systems, including polymer melts, DNA, and proteins. In this paper, we explore a data-driven approach to the classification problem of link topology. Extending the framework introduced in Ref. 1 (Sleiman et al, 2024 Soft Matter, 20(1), pp.71-78), we show that a feedforward neural network trained on the writhe density matrix classifies thermally equilibrated configurations of the first six prime links with 97% accuracy. We demonstrate that this accuracy remains high across a range of temperatures and lengths of link components, while rapidly deteriorating with the addition of topology-altering Gaussian noise; a result consistent with the writhe density matrix containing features sensitive to topology. Our results show that neural networks based on the writhe density matrix efficiently classify two-component links, establishing machine learning as a promising tool for rapid classification of more complex link topologies, e.g. Borromean rings and multi-component links, as the computational cost of exact numerical calculation of topological invariants becomes prohibitive.
[LG-73] SPECTRA: State-Space Exogenous Context and Temporal-Frequency Resolution Architecture for Probabilistic Energy Forecasting
链接: https://arxiv.org/abs/2607.20587
作者: Hang Ye,Xinyan Jiang,Yuedong Shi,Yangxin Zhu,Jianming Wei,Tian Zheng,Xiaoying Zheng,Yongxin Zhu
类目: Machine Learning (stat.ML); Machine Learning (cs.LG)
*备注: 10 pages, 4 figures, 3 tables. Submitted to IEEE Transactions on Power Systems
Abstract:Modern power systems increasingly require probabilistic forecasts amid interacting uncertainties from renewable intermittency, flexible demand, market volatility, and weather-dependent generation. However, existing methods often treat multi-scale decomposition, exogenous-variable alignment, and probabilistic output as separate steps, obscuring how predictable structures and uncertainty-bearing fluctuations jointly shape the forecast distribution. This paper proposes a state-space exogenous-context and temporal-frequency resolution architecture for general probabilistic energy forecasting. Its central premise is that trend-periodic components primarily determine the baseline trajectory, whereas high-frequency residuals and external perturbations govern the spread and asymmetry of forecast uncertainty. Accordingly, the architecture adaptively separates deterministic and residual streams, aligns exogenous context with both, refines the deterministic backbone through multi-resolution spectral-temporal state-space modeling, and estimates ordered quantile boundaries from their complementary representations. Experiments on load, price, solar, and wind forecasting achieve the best continuous ranked probability score in 14 of 18 settings, reducing average CRPS by 5.74% and upper-tail quantile risk by 7.27% over the strongest baselines. These results support deterministic-stochastic separation as an effective design principle for general probabilistic energy forecasting.
[LG-74] Machine-Learned Compact Subspace Generation for Quantum Selected Configuration Interaction within Density Matrix Embedding Framework
链接: https://arxiv.org/abs/2607.20585
作者: Ashish Kumar Patra,Anurag K. S. V.,Ruchika Bhat,Sai Shankar P.,Rahul Maitra,Jaiganesh G
类目: Quantum Physics (quant-ph); Emerging Technologies (cs.ET); Machine Learning (cs.LG); Chemical Physics (physics.chem-ph); Biomolecules (q-bio.BM)
*备注: 31 pages, 16 figures
Abstract:Sample-based Quantum Diagonalization (SQD), an extension of Quantum Selected Configuration Interaction (QSCI), has emerged as a promising hybrid quantum-classical paradigm for computing molecular ground state energies. By leveraging quantum sampling instead of variational optimization, QSCI avoids barren plateaus and enables direct reconstruction of correlated electronic wavefunctions. However, existing configuration recovery techniques primarily enforce symmetry constraints without guaranteeing optimal selection of the most physically relevant configurations, often leading to unnecessarily large subspaces and increased classical diagonalization costs. In this work, we introduce a machine-learned compact subspace generation protocol based on Restricted Boltzmann Machines (RBMs), termed QSCI-RBM, and integrate it within the Density Matrix Embedding Theory (DMET) framework. The RBM is trained on quantum-sampled configurations to learn the underlying probability distribution of dominant determinants, enabling the targeted generation of high-probability configurations. We apply this framework to the simulation of a protein-ligand complex involving the inhibitor Carmofur bound to the SARS-CoV-2 main protease ( M^\textpro ). Our results demonstrate that DMET-QSCI-RBM achieves energies within the chemical accuracy threshold by accessing only approximately 4% of the configuration subspace. In contrast, standard DMET-SQD simulations failed to reach chemical accuracy while accessing up to 20% of the subspace, even as the chemical potential itself nearly converged. These findings highlight that RBM-assisted configuration generation produces significantly more compact subspaces while preserving physical accuracy, thereby reducing classical computational overhead and enabling the scalable quantum embedding simulation of complex biological systems.
附件下载
